Introduction

The repeatability of evolution and its potential adaptive significance has fascinated scientists as far back as Charles Darwin, and the relative role of genetic changes and contingent events in shaping the possible evolutionary outcomes is an ongoing focus of study1,2,3. This repeatability manifests as convergent evolution, which describes the independent evolution of a similar phenotype in different species2,3. Parallel evolution is sometimes distinguished from convergent evolution when a phenotype has repeatedly evolved via the same underlying mutation or process, although ambiguity arises in assessing the biological basis of the convergent phenotypes4,5, potentially involving the same developmental alteration, pathway flux, shift in gene expression or even amino acid replacement. Regardless, both convergent and parallel evolution are usually framed in the context of adaptation, in which the recurrently evolved trait is assumed to repeatedly arise due to an adaptive benefit in at least some contexts, such as the multiple origins of powered flight in animals6, the repeated convergence of floral morphology among plants7,8, the multiple independent shifts from C3 to C4 or CAM photosynthesis in plants9,10, and the repeated convergence of various protein adaptations, such as hemoglobins that enable high-altitude respiration11,12 and enzymes that produce plant secondary metabolites13,14.

Yet, repeated evolution can also occur when the adaptive benefit is not apparent. One example relates to the maturation of c-type cytochromes, which are essential electron carriers in the electron transport chain (ETC) of mitochondria, chloroplasts, and prokaryotes. Cytochrome c maturation (CCM) requires the covalent attachment of a heme group to the apocytochrome, which occurs at conserved cysteines in a CXXCH motif to form the holocytochrome. CCM occurs by at least three distinct pathways: system I and II originated in prokaryotes and persist in mitochondria (system I) or plastids (system II) of diverse eukaryotes, whereas system III is specific to mitochondria of some eukaryotic lineages, including animals, fungi, amoebozoans, and various algae15,16. The sharing of systems between bacteria and organelles indicates that mitochondria and plastids likely inherited system I and II from their alpha-proteobacterial and cyanobacterial progenitors. In contrast, the sporadic distribution of system III among eukaryotes implies a complex evolutionary history involving repeated independent shifts from system I to system III15,16. Moreover, increased mitogenomic sampling from diverse eukaryotes17,18,19,20,21,22 has revealed that the loss of system I (and presumed shift to system III) is more prevalent than currently appreciated.

Although both CCM systems I and III catalyze heme attachment to mitochondrial c-type cytochromes, the two pathways differ in their components and complexity15,16. In E.coli, system I comprises one operon encoding eight Ccm proteins:23 CcmA and CcmB form an ABC transporter; CcmC and CcmD deliver heme to CcmE, a heme chaperone; CcmF is the putative heme synthase that transfers heme from CcmE to the apocytochrome; and CcmG and CcmH participate in a thioreduction pathway that reduces the disulfide bond to allow heme attachment24. In plants, six system I homologs have been identified: CCMA, CCME and CCMH are usually encoded by the nuclear genome, while CCMB, CCMC and CCMF (CCMF is often split into two or three genes) are usually encoded by the mitogenome25,26. By contrast, the eukaryote-specific system III is comparatively very simple, relying on a holocytochrome c synthase (HCCS) enzyme that binds heme and apocytochrome, catalyzes ligation, and releases the holocytochrome c16,27. In yeast, two HCCS homologs (CCHLp and CC1HLp) operate specifically on apocytochromes c and c1, respectively, whereas in animals, a single HCCS can act on both c-type mitochondrial apocytochromes28. Because a defect in heme attachment has implications in mitochondrial pathology and disease, human HCCS has been extensively studied in terms of its structure, mitochondrial membrane localization and mechanism for generating the mature cytochrome c29,30,31,32.

Nevertheless, our understanding of the evolutionary history of CCM systems among eukaryotes, especially the spread and functional divergence of the system III HCCS among distinct lineages, remains limited. The repeated and independent shifts of the mitochondrial CCM pathway from system I to III raises fundamental questions about repeated evolution via pathway cooption and the underlying genetic mechanisms and evolutionary forces that promote or deter pathway shifts in specific lineages. In particular, how did the same system III HCCS enzyme become so sporadically distributed among eukaryotes? The Archaeplastida (comprising land plants, green algae, red algae, and glaucophytes) provide an ideal test group for understanding the evolution of system I and III CCMs. Mitogenomic sequencing of lycophytes33,34,35, hornworts36,37,38, a liverwort39, a fern21, and various green and red algae22,40,41 demonstrated that all mitochondrial system I genes were repeatedly lost during Archaeplastida evolution, although the number of independent examples has not been examined comprehensively. Importantly, in the transcriptome of the fern Ophioglossum californicum, no transcripts were detected for the missing mitochondrial ccm genes, arguing against their transfer to the nuclear genome, or for any nuclear Ccm genes, indicating that the entire system I pathway was lost21. In Chlamydomonas reinhardtii, which also lacks system I genes in its mitochondrial and nuclear genomes, two putative system III HCCS genes were detected in the nuclear genome15,25,42, suggesting that the loss of system I was functionally coopted by a switch to system III (as in fungi and animals), although this inference awaits experimental confirmation.

In this work, to determine the number of shifts from system I to III in Archaeplastida, we conduct a comprehensive survey on the presence/absence of the mitochondrion-encoded members of the system I pathway and the nucleus-encoded components of systems I and III. To test the inference of cooption of system I by system III in diverse Archaeplastida lineages, we characterize the function of HCCS homologs from the fern Ceratopteris richardii, an emerging nonflowering plant model in the study of plant evolution43,44, and Chlamydomonas reinhardtii, a unicellular algal model amenable to a variety of genetic and biochemical analyses45,46 and whose mitochondrial respiration-deficient mutants can be viably maintained by photosynthesis under light conditions47,48. These properties make C. reinhardtii an ideal model system for investigating mitochondrial respiration and carbon metabolism.

Results

Recurrent loss of CCM System I from diverse Archaeplastida

Mitochondrial genome sequencing has revealed the loss of ccm genes (ccmB, ccmC, ccmFC, ccmFN) encoding members of the system I CCM pathway from various lineages of land plants, green algae, and red algae21,22,33,34,35,36,37,38,39,40,41. To comprehensively evaluate the frequency and phylogenetic distribution of mitochondrial ccm gene losses across Archaeplastida, a survey of mitochondrial genomes from diverse land plants, green algae, red algae, and glaucophytes was performed, which identified at least 11 independent losses of the full suite of mitochondrial ccm genes (Fig. 1). These losses occurred in disjunct archaeplastid lineages, including (1) leptosporangiate ferns, (2) eusporangiate ferns from Ophioglossum, (3) all lycophytes, (4) all hornworts, (5) the liverwort Treubia lacunosa, (6) Chaetosphaeridiales, (7) the klebsormidiophyte Entransia fimbriata, (8) the common ancestor of Chlorokybus and Mesostigma, (9) all chlorophytes, (10) all red algae in Rhodophytina, and (11) all glaucophytes.

Fig. 1: Distribution of system I and system III genes for mitochondrial CCM in sequenced Archaeplastida species.
figure 1

The boxes indicate the presence (black or blue), absence (white), or uncertain status (gray with question mark) of system I and III CCM genes in each species. Species in blue lack system I but possess system III, and the inferred shifts from system I to III were counted and mapped onto the phylogenetic tree, whose relationships are based on recent studies93,94. The status of ccmB is uncertain in Galdieria40.

Full size image

To assess the functional status of the nuclear Ccm genes (CcmA, CcmE, CcmH) encoding members of the system I pathway, and to evaluate whether the lost mitochondrial ccm genes were functionally transferred to the nucleus, we surveyed the presence of these system I genes in published nuclear genomes and newly assembled transcriptomes from diverse Archaeplastida (Fig. 1). For all species that lost their mitochondrion-encoded system I genes, we failed to detect any copies of the lost mitochondrial genes in their nuclear genome/transcriptome, arguing against functional transfer of these genes from the mitochondrion to the nucleus, and we failed to detect the suite of nucleus-encoded Ccm genes, which strongly implies complete loss of system I from these species. In contrast, all species that retained their mitochondrial ccm genes also retained their nuclear Ccm genes, indicating a presumably intact system I. Our blast-based survey results are likely to be robust because we used system I homologs from multiple Archaeplastida sequences as queries, our findings for the mitochondrial encoded proteins were corroborated by available mitogenome annotations, and the presence/absence of system I genes was concordant between the mitochondrial and nuclear genomes of each species, even for those with fast-evolving mitogenomes such as C. reinhardtii41 and Galdieria sulphuraria40. Overall, our survey expands upon previously reported mitochondrial ccm losses by documenting at least 11 independent losses of the entire system I CCM pathway across Archaeplastida.

Repeated shifts to CCM system III among Archaeplastida that lost system I

As the maturation of mitochondrial cytochrome c and c1 holoforms is likely an essential process, an obvious question is raised: how does CCM proceed in Archaeplastida lineages that lack system I? C. reinhardtii possesses two genes (HCS1, HCS2) encoding homologs to human and fungal HCCS, suggesting that this species has switched to the system III CCM pathway. Using the C. reinhardtii genes as queries, we identified HCCS homologs in the nuclear genomes of Selaginella moellendorffii and Chondrus crispus, which also lack evidence of system I. Using these diverse HCCS sequences, we surveyed the remaining archaeplastid nuclear genomes and assembled transcriptomes for the presence of HCCS homologs. A homolog to HCCS was identified in all species that have lost system I, but no HCCS was detected in species that retained system I (Fig. 1). This dichotomous pattern strongly suggests that the recurrent loss of system I in diverse Archaeplastida was enabled by the presence of system III HCCS.

Plant and green algal HCCS localize to mitochondria and rescue yeast system III mutants

Although homologs to a system III HCCS were detected in all Archaeplastida lineages that have lost system I genes, no study has experimentally confirmed that any of these archaeplastid homologs are functional system III HCCS enzymes capable of maturation of mitochondrial cytochromes c. To assess whether Ceratopteris HCCS (CriHCCS) is targeted to the mitochondrion, a vector carrying the CriHCCS coding sequence fused with GFP was transiently agroinfiltrated into tobacco leaves from which protoplasts were isolated. Confocal laser-scanning microscopy revealed clear mitochondrial localization of the GFP fluorescence, as confirmed by counterstaining with MitoTracker (Fig. 2). The two Chlamydomonas homologs (CreHCS1 and CreHCS2) were also predicted to be localized to mitochondria (Supplementary Table 1).

Fig. 2: Subcellular localization of Ceratopteris HCCS (CriHCCS) in tobacco leaf protoplasts.
figure 2

The 35S:CriHCCS-GFP construct was transformed into tobacco leaves. GFP, MitoTracker signal and chlorophyll autofluorescence in the transformed tobacco cells were separately observed by confocal laser scanning microscopy and then resulting images were merged. The experiment was repeated three times with similar results. Scale bars, 5 μm.

Full size image

In human and yeast CCM via system III, HCCS is responsible for heme attachment to c-type apocytochromes. S. cerevisiae carries two versions of this enzyme: CCHLp (encoded by CYC3) and CC1HLp (encoded by CYT2) attach heme to mitochondrial cytochromes c and c1, respectively. Yeast cells with a deleted CYC3 are completely, or almost completely, deficient in cytochrome c and, thus, do not grow on complete medium containing a nonfermentable carbon source such as glycerol28. This ∆cyc3 strain, when transformed with CriHCCS constructed in a vector under the control of the constitutive promoter PGK1, was able to grow on glycerol medium after a 5-day incubation, suggesting that CriHCCS can functionally complement this respiratory deficient mutant and restore the capability of aerobic respiration (Fig. 3a). We also found that CriHCCS could complement the ∆cyt2 yeast mutant deficient in holocytochrome c1 synthase. In a parallel analysis of C. reinhardtii HCCS homologs, CreHCS1 was able to rescue the corresponding ∆cyc3 strain, but not ∆cyt2, in the yeast complementation assay, while CreHCS2 could rescue ∆cyt2 but not ∆cyc3 (Fig. 3b).

Fig. 3: Yeast complementation with Ceratopteris HCCS (CriHCCS) and Chlamydomonas HCS (CreHCS1 and CreHCS2) homologs.
figure 3

S. cerevisiae strain Y00367 (∆cyc3; CCHLp) and Y04936 (∆cyt2; CC1HLp) were transformed with a plasmid containing a Ceratopteris CriHCCS coding sequence or b Chlamydomonas CreHCS1 or CreHCS2 coding sequence, each under control of the PGK1 promoter. An empty vector was used as a negative transformation control. Equal amounts of serial dilutions of cells from exponentially grown cultures were spotted onto solid synthetic defined (SD) plates supplemented with 2% glucose or 3% glycerol and incubated at 25 °C. Each spot assay was repeated three times with similar results.

Full size image

CRISPR-generated Chlamydomonas System III mutants exhibit growth defects

Wild-type Chlamydomonas can grow heterotrophically in the dark when the growth medium is supplemented with carbon sources, such as acetate, or photoautotrophically in the light using CO2 as the carbon source45,49. Taking advantage of this characteristic, a number of respiratory-deficient strains for different ETC components, including complex I (NADH dehydrogenase), complex II (succinate dehydrogenase), complex III (ubiquinol: cytochrome c oxidoreductase), and complex IV (cytochrome c oxidase), have been isolated by mutagenesis in combination with phenotypic observation of their null or slow growth in heterotrophic conditions48,50,51. However, Chlamydomonas mutants with altered cytochrome c function are limited to mitochondrial genes, such as cob (encoding apocytochrome b of complex III)50,51 and cox1 (encoding the subunit 1 of complex IV)52.

Here, to further evaluate the function of Chlamydomonas HCS1 and HCS2, two nucleus-encoded genes for system III c-type cytochrome maturation, we obtained several independent CreHCS1 and CreHCS2 mutants by CRISPR-Cas9 mediated gene editing (Fig. 4). PCR amplification and sequencing of the HCS loci confirmed the occurrence of targeted mutations at the expected CreHCS1 and CreHCS2 CRISPR-Cas9 RNP cut sites. Among the sequenced mutants, HCS1-9 carries one nucleotide deletion in exon 1, resulting in a frameshift that leads to a premature stop codon and, consequently, a functional knock-out of CreHCS1; HCS1-25 has a 24 bp insertion (Fig. 4a). In HCS2 editing, two mutants (HCS2-22, HCS2-35) obtained from dark screening have large insertions (460 bp and 1160 bp, respectively), which likely disrupt the function of HCS2. Interestingly, HCS2-27, a colony identified by PCR screening before the implementation of dark screening, has an in-frame mutation (45 bp insertion) (Fig. 4b).

Fig. 4: Description and growth measurement of Chlamydomonas HCS mutants generated by CRISPR-Cas9.
figure 4

a Schematic of the CreHCS1 target site (red arrow) and the mutated region (shown in red) in mutant HCS1-9 (1 bp deletion) and HCS1-25 (24 bp insertion). b Schematic of the CreHCS2 target site (red arrow) and the insertions (the number of inserted nucleotides was labeled in red) of mutant HCS2-27, HCS2-22, and HCS2-35. c Growth curve of g1 and HCS mutant strains cultivated in TAP medium under light and dark conditions. Values shown are the average of three biological replicates ± SD. Pictures are representative of strains on day 3 from three experimental replicates. Source data are provided as a Source Data file.

Full size image

We next compared the growth of the WT strain (g1) and the HCS1 and HCS2 mutants cultivated in acetate-containing culture medium (TAP) under continuous light or dark conditions (Fig. 4c). During the first three days of light exposure, the knock-out mutants HCS1-9 and HCS2-22 exhibited a slower growth rate compared to that of WT. In dark conditions, HCS1-9 and HCS2-22 failed to grow, while WT showed continuous growth, although its rate began to decline after day three. HCS2-27 mutant behaved comparably with WT under light but grew slower in the dark, indicating a partial functional impairment in its heterotrophic growth caused by the in-frame mutation. Its dark-growth defect was confirmed in a spot assay (Supplementary Fig. 1a).

Chlamydomonas HCS mutants are defective in cytochrome c-dependent respiratory chain with altered TCA intermediate accumulation

TTC (2,3,5-triphenyltetrazolium chloride) is a colorless molecule that can be converted to formazan (a red compound) by the mitochondrial ETC with a functional cytochrome pathway53,54. When treated with TTC, Chlamydomonas WT g1 and the HCS2-27 cells turned dark brown (a mixture of green and red color), whereas the knock-outs HCS1-9 and HCS2-22 remained green (Supplementary Fig. 1b), indicating a disruption of mitochondrial ETC activity in the knock-out mutants due to the functional impairment of cytochrome pathway components (complex III and/or IV).

We also measured oxygen consumption rates of the cells grown in heterotrophic conditions (dark + acetate). Like land plants, Chlamydomonas has two respiratory pathways: the classical cytochrome oxidative phosphorylation pathway and an alternative pathway that branches from the ubiquinol pool and directly reduces O2 to H2O without passing through the cytochrome pathway. The classical cytochrome pathway is sensitive to cyanide (NaCN), whereas the alternative pathway is resistant to cyanide but sensitive to salicylhydroxamic acid (SHAM) (Fig. 5). The HCS1 and HCS2 knock-out mutants (HCS1-9, HCS2-22) exhibited nearly 70% reduction in total dark respiration rate (Table 1), in agreement with our observation that both mutants exhibited severe growth defects in the dark (Fig. 4c). The addition of cyanide alone reduced the respiration of WT cells to around 30%, but cyanide had much less effect on the respiration of the knock-out strains. In contrast, the addition of SHAM abolished the oxygen consumption in the knock-out mutants (Table 1). The sensitivity to SHAM but insensitivity to cyanide in the knock-out mutants indicate that their reduced respiration rate was due to the dysfunction of the cytochrome pathway (cyanide-sensitive), whereas the alternative oxidase pathway (SHAM-sensitive) is still functional. The HCS2-27 mutant had a dark respiration rate in between the WT and knock-out strains and showed intermediate effects in response to the inhibitors (Table 1), indicating a less severe effect on the cytochrome c oxidase pathway, in agreement with the results of growth measurement and TTC staining.

Fig. 5: A diagram showing the mitochondrial electron transport chain (mETC), the effect of HCS genes and inhibitors on ETC components, and the associated tricarboxylic acid (TCA) metabolic pathway in Chlamydomonas.
figure 5

Red arrows indicate the involvement of HCS1 and HCS2 in the maturation of cytochrome c (CytC) and cytochrome c1 (subunit of complex III) within the cytochrome oxidase pathway, which can be inhibited by cyanide (NaCN), independent from the SHAM-sensitive alternative oxidase pathway (AOX). Diagram of the TCA cycle with bar graphs showing levels of metabolites in the wild type (WT) and HCS2-27 mutant (HCS2) growing in TAP medium under normal light (NL; blue bars) or low light (LL; orange bars) conditions. Data are presented as mean values +/− SD (n  =  3, 6, 4, 3. n refers to biological replicates of each strain). Asterisks indicate significant differences between means evaluated by one-way ANOVA followed by Tukey’s multiple comparisons test (*p < 0.05; **p < 0.001). Source data are provided as a Source Data file.

Full size image
Table 1 In vivo measurement of dark respiration and the impact of SHAM and NaCN on the respiratory pathways of Chlamydomonas HCS mutants
Full size table

To further capture the impact of HCS mutations on the mitochondrial ETC, we performed metabolite profiling of the WT strain and the mutant strain HCS2-27 grown at normal and low light conditions. As suggested in our early experiments, HCS2-27 partially retained its mitochondrial function to support its heterotrophic growth and thus can be cultivated in parallel with the background WT strain for comparative metabolomics analysis. Using GC-MS analysis, we quantified a total of twenty-three primary metabolites, including sugars, amino acids, and the tricarboxylic acid (TCA) cycle intermediates (Supplementary Data 1). Under normal light conditions, no statistically significant differences in metabolite levels were detected between WT and HCS2-27, whereas in low light, the HCS2-27 strain accumulates several TCA cycle intermediates, namely succinate, malate, and citrate to higher levels than WT (Fig. 5). These results correlate with the observed dark-specific growth defects (Fig. 4c; Supplementary Fig. 1a) and the fact that mitochondrial respiratory defects in Chlamydomonas can be compensated for by photosynthesis under optimal illumination48,55. The data also indicate that the reduced activity of HCS2 alters mitochondrial metabolism, as evidenced by the fact that only the metabolites closely related to the TCA cycle were affected (Supplementary Fig. 2). More broadly, these CRISPR-generated mutants with novel defects in the mitochondrial cytochrome pathway will be valuable resources for the Chlamydomonas community, particularly in advancing research within the field of algal organelle biology.

System III HCCS is sufficient to complement an Arabidopsis system I mutant

The yeast complementation results demonstrate that archaeplastid HCCS homologs can function in other eukaryotes using system III (Fig. 3). Characterization of Chlamydomonas HCS mutants further confirmed their primary roles in the system III pathway (Figs. 4, 5). In plants, we hypothesize that CCM by system I is being repeatedly replaced by system III HCCS. To investigate whether a lethal system I CCM mutant can be functionally complemented by system III HCCS in plants, we used an Arabidopsis T-DNA insertion line, ccmh (SALK_046872) for transgenic study (Fig. 6). CCMH plays a central role as a mitochondrial thiol-disulfide oxidoreductase in cytochrome c maturation. Due to the essential function of CCMH, only heterozygosis of the T-DNA insert was found in the progeny of ccmh line, consistent with a previous report that ccmh mutation is lethal in its homozygote state56. Thus, we transformed heterozygous ccmh/CCMH plants with a system III HCCS construct (pUBQ10::CriHCCS). Progeny from the transformed ccmh/CCMH plants were selected and genotyped for the presence of the HCCS construct and the homozygous ccmh mutation (Fig. 6b). Seeds harvested from independent T3 lines (ccmh+CriHCCS) were used for further evaluation.

Fig. 6: System III HCCS is sufficient to complement Arabidopsis system I mutant.
figure 6

a Transgenic expression of CriHCCS rescued the seed lethality of the Arabidopsis CCMH mutant, resembling the developmental growth of the wild type Col-0. DAG, Days after Germination. Scale bar = 2 cm. b Genotyping analysis of ccmh complementation line. Transformants were PCR genotyped for ccmh T-DNA homozygosity using three primers (LBb1.3, T-DNA left border sequence; LP1 and RP1, Left and right primer flanking the T-DNA insertion site in the CCMH gene) and for the presence of the inserted pUBQ10:CriHCCS complementation construct using primers located within the CriHCCS coding sequence (F272 and R975). The analysis was repeated three times with same results. c Abnormal phenotype of homozygous seeds (denoted by asterisks) in the silique of self-fertilized heterozygous ccmh/CCMH plants was rescued in the ccmh+CriHCCS plant. Scale bar = 500 μm. d Photo of representative 60-day-old plants in long-day growth conditions. Scale bar = 2 cm. e Germination of seeds from Col-0, heterozygous ccmh/CCMH plants, and ccmh+CriHCCS plants at different times after transfer to germination conditions. The data are presented as means ± sd and the circles represent the number of values analyzed (n = 3 independent experiment). f Primary root length of these lines mentioned above. The boxplots show the minimum, first quartile, median, third quartile, and maximum primary root length of each line (n = 12). One-way ANOVA followed by Tukey’s multiple comparisons test was conducted to examine the variation. A representative picture of plants (corresponding to the lines in the boxplot from left to right) grown vertically on the MS plate was taken on day 14 after seed sowing. Scale bar = 1 cm. Source data are provided as a Source Data file.

Full size image

The growth rate of complemented plants resembled that of the WT under greenhouse conditions (Fig. 6a). Moreover, the abnormal seeds in the immature siliques of a heterozygous ccmh/CCMH plant were not observed in the complemented lines (Fig. 6c), suggesting the system III HCCS successfully rescued the embryo-lethality of ccmh homozygosity. The homozygous system I mutant line, rescued with the system III HCCS, produced seeds and was phenotypically indistinguishable from WT (Fig. 6d). Plate-based phenotyping demonstrated that all lines exhibited similar seed germination rates, measured as the percentage of seeds with visible root protrusion, after 18–42 h of transfer to germination conditions (Fig. 6e). Similarly, the root growth of ccmh+CriHCCS seeds exhibited no differences relative to the WT (Fig. 6f). Taken together, our complementation study revealed that exogenous expression of a system III HCCS gene is sufficient to rescue the seed lethality of Arabidopsis system I CCMH mutant and complement its essential function for producing viable seeds and enabling normal plant growth.

Phylogenetic analysis of eukaryotic HCCS homologs

The CCM survey demonstrated that all HCCS homologs detected among Archaeplastida are related in sequence (Fig. 1), and the functional assays demonstrated that fern HCCS and Chlamydomonas HCS1 and HCS2 are functional components of a system III pathway, capable of maturation of c-type apocytochromes (Figs. 2–6, Table 1). However, it is unclear how so many Archaeplastida lineages independently switched from system I to III, raising questions about the evolutionary mechanisms promoting these multiple pathway shifts in Archaeplastida and about the evolution of HCCS proteins capable of either bifunctional maturation of both cytochrome c and c1 or monofunctional maturation of only one c-type cytochrome.

Thus, we performed a phylogenetic analysis of Archaeplastida HCCS homologs in the broader context of HCCS homolog diversity across eukaryotes (Fig. 7a). Notably, each of the major eukaryotic lineages (as labeled in the figure) is monophyletic, except for stramenopiles group 2 for which oomycetes (stramenopiles 2a) did not cluster with diatoms and ochrophytes (stramenopiles 2b). The HCCS proteins from streptophytes (including the bifunctional fern HCCS) strongly clustered (92%/1.0 ML bootstrap/Bayesian posterior probability) with chlorophytes group 1 (including the monofunctional HCS1 from Chlamydomonas). These proteins also associate with a larger clade that includes the presumably bifunctional proteins from red algae and glaucophytes and one HCCS copy from stramenopiles and haptophytes (85%/1.0). In contrast, the chlorophyte group 2 proteins, including the monofunctional Chlamydomonas HCS2, are distant paralogs of these other Archaeplastida proteins. Chlorophytes group 2 instead associates with other eukaryotic groups including dictyostelids and stramenopiles group 2a. Overall, the bifunctional Archaeplastida HCCS enzymes are closely related to chlorophyte HCS1 (specific for apocytochrome c), whereas chlorophyte HCS2 (specific for cytochrome c1) is a distant paralog.

Fig. 7: Phylogenetic analysis of HCCS homologs.
figure 7

a Phylogeny of eukaryote HCCS homologs with branches compressed into major groups. The tree was rooted using midpoint rooting. The fully labeled ML tree is shown in Supplementary Fig. 3, and the Bayesian tree is shown in Supplementary Fig. 4. b Phylogeny of HCS homologs in land plants, green and red algae. The tree was rooted on red algae, based on the results in Fig. 7a. For both trees, bootstrap values < 50% and posterior probabilities <0.8 are not shown. Source data are provided as a Source Data file.

Full size image

Within Archaeplastida, the switch from system I to III has occurred at least 11 times (Fig. 1), prompting inquiries into the evolutionary relationships of HCCS homologs among Archaeplastida lineages that have switched to system III. At a broad level, HCCS relationships among Archaeplastida (Fig. 7b) are consistent with organismal relationships: chlorophytes (69%/0.97) and streptophytes (100%/1.0) each form a monophyletic group, except that the streptophyte Mesostigma grouped anomalously with chlorophytes (75%/1.0). Within streptophytes, the green algae form the expected grade, with Chlorokybales diverging earlier than the Coleochaetales relative to land plants (93%/1.0). Within land plants, however, ferns and lycophytes do not form expected monophyletic groups. Although branch support for many of these nonmonophyletic groupings was limited, Ophioglossum (a fern) is nested within a clade for Selaginella (a lycophyte) with strong branch support (87%/1.0). AU topology tests, forcing either fern monophyly, lycophyte monophyly, or fern and lycophyte monophyly, were rejected (P < 0.05), providing significant support against vertical inheritance and suggesting a history of horizontal transfer for HCCS among ferns and lycophytes.

Discussion

Here, in a survey for system I and III CCM genes in genomes and transcriptomes of diverse Archaeplastida, we documented at least 11 independent shifts from system I to III in distinct lineages (Fig. 1). This inference was based on the losses of nucleus- and mitochondrion-encoded system I genes from five land plant lineages (leptosporangiate ferns, the eusporangiate fern Ophioglossum, all lycophytes, all hornworts, and the liverwort Treubia), three streptophytic green algal lineages (Coleochaetales, the klebsormidiophyte Entransia, and the basal lineage Chlorokybus + Mesostigma), all chlorophytes, all red algae in Rhodophytina, and all glaucophytes. The switch from system I to III was inferred by the perfectly parallel presence of HCCS homologs in all archaeplastid lineages that lack system I, but not in any lineages that retained system I (Fig. 1), and by functional assays, which showed that plant HCCS homologs are targeted to mitochondria (Fig. 2; Supplementary Table 1) and are sufficient to functionally complement mutants of yeast system III (Fig. 3) and Arabidopsis system I (Fig. 6). Moreover, Chlamydomonas HCCS mutants exhibit reduced growth (Fig. 4; Supplementary Fig. 1a), impaired mitochondrial activity (Table 1; Supplementary Fig. 1b), and altered mitochondrion-related metabolic flux (Fig. 5), demonstrating that these identified HCCS homologs are indeed necessary for growth and mitochondrial function, particularly under heterotrophic conditions. These recurrent CCM pathway shifts in Archaeplastida greatly expand our understanding of the evolution of CCM pathways across eukaryotes, for which variation of the usage of systems I and III has also been documented, albeit to a lesser extent, in Opisthokonta20, Alveolata57,58, Cryptista59,60, and Discoba61,62.

Interestingly, within Archaeplastida, most lineages have a single HCCS that presumably catalyzes the maturation of apocytochromes c and c1. This inference was verified for fern HCCS, which could rescue yeast ∆cyc3 and ∆cyt2 deletion mutants (Fig. 3a). In chlorophytes, however, there are two divergent HCCS paralogs, and functional analyses of Chlamydomonas HCS1 and HCS2 showed that each homolog can only rescue one of two yeast system III mutants (Fig. 3b), indicating that each Chlamydomonas enzyme is specific for the maturation of either apocytochrome c or c1, but not both. The generalized or specialized functions of HCCS in Archaeplastida are also echoed across eukaryotes, where some eukaryotes (including animals and dictyostelids) have a single HCCS that is presumed, and experimentally confirmed for human HCCS28, to act on both apocytochromes, whereas many other eukaryotic lineages (including fungi, stramenopiles, haptophytes, and alveolates) have two distinct copies that are likely specific for the maturation of either apocytochrome c or c1, as demonstrated for yeast CCHLp and CC1HLp63,64.

From a mechanistic standpoint, the repeated CCM pathway shifts from system I to III, and the variable usage of either generalized or specialized HCCS enzymes, raise baffling questions about the repeatability of evolution. To explain how nearly a dozen different plant and algal lineages, and even more eukaryotic lineages, have independently switched in parallel to a homologous HCCS enzyme, two competing hypotheses have been proposed:15,25 systems I and III were both present in an early eukaryotic ancestor followed by differential loss in descendent lineages, or rampant horizontal gene transfer (HGT) spread HCCS in the disparate eukaryotic lineages that use system III. The explanation that both systems were ancestrally present and maintained over long evolutionary time periods, only for one or the other pathway to be lost in essentially all present-day lineages, is difficult to accept. To date, only a single eukaryotic lineage (the newly described Provora) has been documented to have both systems I and III65,66, and this inference was based only on sequencing (rather than functional) data. If both pathways were compatible for much of eukaryotic evolution, then why do we not find more living lineages that have both systems today? And why would two redundant pathways be maintained for so long? The alternative possibility that the HCCS gene has been repeatedly transferred horizontally among eukaryotes requires an extent of HGT that is likely to be contentious e.g.67,.

To tease these competing theories apart, we used phylogenetic analyses (Fig. 7). Importantly, the HCCS homologs form a monophyletic group for each major eukaryotic lineage (or two weakly split subclades in the case of stramenopiles group 2), indicating that HCCS was already present in the common ancestor of each of these ancient eukaryotic groups (Fig. 7a). Similarly, the focused analysis of HCCS and HCS1 homologs among green plants (Fig. 7b) revealed consistent patterns at the deepest levels, in which chlorophytes and rhodophytes are each monophyletic, and streptophytes are monophyletic except for Mesostigma, which groups instead as sister to chlorophytes. Within land plants, however, HCCS relationships were inconsistent with organismal relationships: lycophytes and ferns are not monophyletic, but instead, Ophioglossum is nested within Selaginella while Isoetes associates with some leptosporangiate ferns. Thus, our phylogenetic analyses suggest a possible compromise, in which systems I and III may have been present in the archaeplastid common ancestor, prompting early decisions to retain one or the other system, followed by more recent HGT that prompted more recent system I to III switches among land plants.

We also assessed the evolutionary relationships among the single HCCS and dual HCCS homologs found among eukaryotes (Fig. 7a). In Archaeplastida, chlorophytes have two functionally distinct paralogs (HCS1 and HCS2 in Chlamydomonas), whereas all other Archaeplastida species that use system III have a single bifunctional HCCS. Interestingly, chlorophyte HCS1 (which is specific for apocytochrome c) strongly associates with the bifunctional HCCS from other Archaeplastida and with one HCCS enzyme from stramenopiles and haptophytes (whose apocytochrome affinities are unknown). In contrast, chlorophyte HCS2 (which is specific for apocytochrome c1) is a distant paralog of the Archaeplastida HCCS and HCS1 enzymes, instead more closely affiliating with the presumably bifunctional HCCS from slime molds and the second HCCS copy from stramenopiles group 2 A. This pattern suggests two possibilities. In one scenario, chlorophyte HCS1 evolved by specialization of the ancestral bifunctional HCCS, while chlorophyte HCS2 was acquired via HGT (although the exact donors can’t be gleaned from the tree due to limited branch support). Alternatively, it is possible that the ancestral eukaryote already had two distinct HCCS isoforms, and the functional generalization of one gene allowed for the loss of the other in most archaeplastids, whereas chlorophytes retained both ancestral paralogs as monofunctional enzymes. Functional characterization of the HCCS paralogs from diverse eukaryotes is needed to distinguish these possibilities.

This recurrent shifting from system I to III in so many eukaryotic lineages raises fundamental questions about the evolutionary forces driving repeated pathway shifts. First, the adaptive significance of the CCM pathway shift to system III is not apparent, whereas nearly all other examples of repeated evolution (e.g., powered flight, C4 and CAM photosynthesis, secondary metabolites) have an obvious adaptive benefit. From a functional standpoint, both system I and III catalyze the attachment of heme in the same orientation and position to the apocytochromes, resulting in identical holocytochrome end products, and our Arabidopsis HCCS lines indicate that a switch to system III has no obvious phenotype under optimal greenhouse conditions (Fig. 6). In this sense, if both systems are present in a lineage (either from vertical inheritance or horizontal acquisition), then the ultimate retention of system I or III may indeed be an arbitrary evolutionary coin flip with no functional consequences. On the other hand, system III is less complex than system I. Perhaps the simpler system III pathway is beneficial for many eukaryotes, while the increased complexity of system I may allow enhanced regulatory control over the mitochondrion in other eukaryotes. For example, increased mitochondrial complexity has been hypothesized as a regulatory control mechanism for the mitochondrial function of seed plants, for which mitochondria play an essential role in particular developmental stages, including pollen development, germination and early seedling growth68,69. Others have speculated that the switch to system III in eukaryotes was facilitated by the reduced need for diverse substrate specificity, as mitochondria require heme attachment to only two c-type cytochromes, whereas prokaryotes require the diverse substrate specificity of system I to attach heme to many more c-type cytochromes15.

Second, whereas the many examples of adaptive convergent evolution likely involve genetic or spatiotemporal regulatory changes to existing genes and pathways, the CCM shift from system I to III involves the swapping of two unrelated pathways. The most similar example to the CCM pathway evolution described here may be found in the recurrent replacement of translation elongation factor EF-1α by a related EF-Like protein, a switch that may not have any adaptive benefit and likely occurred either by HGT or differential loss of ancestrally coexisting factors70,71. Yet, even in that case, the two elongation factors are evolutionarily related, and the shift involves the cooption of a single enzyme rather than the cooption of the entire CCM system I pathway by an evolutionarily distinct system III pathway. Thus, our CCM work offers a unique view into the processes that can lead to recurrent evolutionary change at the pathway level and provides a tantalizing system to assess the underappreciated role of chance in recurrent evolution.

Methods

Survey of system I and system III genes for cytochrome c maturation

Sequences for system I genes were obtained from the annotated nuclear and mitochondrial genomes of Arabidopsis thaliana (land plant), Cyanidioschyzon merolae (red alga), and Klebsormidium nitens (green alga), while system III gene sequences were obtained from the nuclear genomes of Chlamydomonas reinhardtii (green alga), Chondrus crispus (red alga), and Selaginella moellendorffii (land plant) (Supplementary Table 2). Translated protein sequences of these system I and III genes were used in a tblastn v2.2.31 search (length > 100 bp; e-value < 1e−20) to query their presence/absence in available nuclear genome annotations, available mitochondrial genomes, and newly assembled transcriptomes and draft mitogenomes from diverse Archaeplastida (Supplementary Data 2). To assemble draft mitochondrial genomes, Illumina data sets were assembled with a range of kmer values (61,71,81,91) and the careful option in SPAdes v3.13.072. To assemble transcriptomes, RNA-seq data sets were assembled with default settings in Trinity v2.6.673.

Subcellular localization of Ceratopteris HCCS and Chlamydomonas HCS

The Agrobacterium tumefaciens strain AGL1 was transformed with the full coding sequence of CriHCCS fused to the N-terminus of GFP under the control of the CamV 35 S promoter. Cloning was performed by Golden Gate cloning74,75. Transformed AGL1 were used to infiltrate 4–6 week old Nicotiana benthamiana leaves for transient expression of the gene construct. Briefly, the incubated AGL1 cell culture was resuspended in infiltration medium (10 mM MgCl2, 10 mM MES pH 5.7, 150 µM acetosyringone) to an OD600  =  1.0 and delivered to the underside of leaves using a blunt tipped syringe. Protein expression was monitored after 48 hours76. Protoplasts were prepared from leaves by cutting them into 5–10 mm squares in 10 ml of F-PIN (10 mM KNO3, 3 mM CaCl2, 1.5 mM MgSO4, 1 mM KH2PO4, 20 mM ammonium succinate, 2 mM MES, Murashige-Skoog basal salts, adjusted to 550 mOsm with sucrose, pH 5.7) containing 1% cellulase Onozuka R-10, 0.3% macerozyme R-10, and 0.1% BSA. After shaking for 90 min at 40 rpm in the dark, protoplasts were released with shaking at 150 rpm for 1 min. The filtered suspension was overlayed with 2 ml F-PCN (same as F-PIN but adjusted to 550 mOsm with glucose) and centrifuged for 10 min at 70 × g with low acceleration and deceleration. Intact protoplasts were collected from the interphase and washed with W5 buffer (150 mM NaCl, 125 mM CaCl2, 5 mM KCl, 2 mM MES, 550 mOsm, pH 5.7) by centrifuging for 10 min. Intact protoplasts were resuspended in 0.5 ml W5 buffer prior to microscopy77. Mitotracker (Life Technologies), a mitochondrial-specific dye, was added to the protoplast suspension to a final concentration of 500 nM. Fluorescence was observed with a Leica TCS SP5 confocal laser scanning microscope. Subcellular localization of Chlamydomonas HCS1 and HCS2 was predicted using WoLF PSORT (https://wolfpsort.hgc.jp/) and PredAlgo (http://lobosphaera.ibpc.fr/predalgo)78, a specific program dedicated to the prediction of protein localization to the nucleus, chloroplast or mitochondrion in green algae with high confidence.

Yeast complementation assay

The yeast strains Y00367 (BY4741; MATa; ura3Δ0; leu2Δ0; his3Δ1; YAL039c::kanMX4) and Y04936 (BY4741; MATa; ura3Δ0; leu2Δ0; his3Δ1; YKL087c::kanMX4), which contain a knock-out of the gene encoding either holocytochrome c synthase (CCHLp, ∆cyc3) or holocytochrome c1 synthase (CC1HLp, ∆cyt2), were ordered from Euroscarf (http://www.euroscarf.de). Both strains were transformed with a plasmid containing either the C. richardii HCCS (CriHCCS) full coding sequence or an empty vector lacZ under the control of the 3-Phosphoglycerate kinase 1 (PGK1) promoter. Both vectors were assembled using Golden Gate cloning techniques74,75. Two similar vectors were assembled by replacing the CriHCCS sequence with the full-length coding sequence of C. reinhardtii HCS1 (CreHCS1) or HCS2 (CreHCS2), and these vectors were also transformed into the ∆cyc3 and ∆cyt2 strains. Transformants were selected on solid synthetic defined (SD) media consisting of 0.67% (w/v) yeast nitrogen base without amino acids, supplemented with yeast synthetic dropout medium without Leucine (SD-Leu). For complementation tests, constructed yeast strains were grown on SD-Leu supplemented with either 2% (w/v) glucose or 3% (w/v) glycerol. Primer sequences used for making these constructs are provided in Supplementary Data 3.

Chlamydomonas strains and culture conditions

C. reinhardtii strain g1 (CC-5415; nit1, mt+) was used as the background strain for gene editing experiments. Unless noted otherwise, cultures were incubated under continuous illumination (150 μmol m−2 s−1) on an orbital shaker (180 rpm) at 25 °C. For transformation, an initial batch of cells was pre-cultured, after inoculation from one-week-old Tris-Acetate-Phosphate (TAP) agar plates, to an optical density of ~0.4 at 750 nm in TAP medium. A one-tenth aliquot was then transferred into fresh TAP medium supplemented with 1 μg mL−1 cyanocobalamin (vitamin B12), to enhance Chlamydomonas thermal tolerance. Prior to electroporation, cells were collected at the exponential growth phase (~2 × 106 cells mL−1) by centrifugation at 3000 × g and resuspended in TAP containing 40 mM sucrose and 1 μg mL−1 cyanocobalamin to a final density of ~2.2 × 108 cells mL−1.

CRISPR/Cas9 RNPs, transformation and screening

We used the commercially available Alt-R CRISPR-Cas9 System developed by Integrated DNA Technologies (IDT) for gene editing. The 20-nt guide sequence, corresponding to the target-specific protospacer region was designed using the online tool CHOPCHOP v2 (https://chopchop.cbu.uib.no)79. C. reinhardtii transformation was carried out by electroporation of the CRISPR/CAS9 RNP complex and single-stranded DNA that targets precise editing of the HCS1 or HCS2 locus80. To aid selection, the paromomycin resistance cassette (aphVIII transgene) amplified by high-fidelity PCR from the pSI103 plasmid81 was co-transfected into cells by electroporation. Electroporated cells diluted in 500 μL TAP medium containing 40 mM sucrose and 1 μg mL−1 cyanocobalamin were placed on an orbital shaker (50 rpm) under light (50 μmol m−2 s−1) for 2 h at room temperature. Cells were then heat shocked at 39 °C for 30 min in a water bath and placed back to the orbital shaker for a 36 h recovery. Cells from each electroporation were spread on two TAP agar plates containing 10 μg mL−1 paromomycin for selection. The plates were incubated at room temperature under continuous illumination (100 μmol m−2 s−1) until visible colonies appeared. About 100–150 colonies were randomly selected and screened with the Chelex colony PCR method82. To facilitate the screening process, we also transferred the selected colonies to fresh TAP plates and then kept the plates in the dark for a week. Colonies with reduced or no growth in the dark were chosen as priority for PCR screening. The oligonucleotides and primers used for these experiments are listed in Supplementary Data 3.

Measurement of Chlamydomonas growth and respiration

For growth curve measurement, cells were pre-cultured as described above and then inoculated to TAP medium. Three replicates of each strain were cultivated under continuous light or in the dark. Culture growth was examined by measuring daily absorbance at 750 nm over a period of 6 days. Ten-fold dilution spot test was conducted by spotting 10 μL of cell suspension (at an initial concentration of 103–104 cells/μL) on TAP agar plates and incubating the plates under light for 3 days or in the dark for 8 days.

Colorless TTC (2,3,5-triphenyltetrazolium chloride) turns red when under reducing conditions and has been used as a redox indicator of mitochondrial ETC functionality in microorganisms83,84. Our TTC staining for C. reinhardtii was performed on TAP agar plates (9 cm in diameter) overlaid with 15 mL agar medium containing 0.5 mg/mL TTC50. The plates were incubated at 30 °C for 5 h in the dark and then photographed.

Measurement of whole-cell oxygen consumption was carried out in a dark room using a Clark Electrode (Oxygraph, Hansatech Instruments)85. To assess the activity of mitochondrial complexes, sodium cyanide (NaCN) and salicylhydroxamic acid (SHAM) were used as specific inhibitors and injected from the top hole of the electrode chamber to cell cultures at a final concentration of 0.5 mM. Respiration rates were recorded for 3 min on each sample before and after the addition of each inhibitor. All measurements were conducted in triplicate with the same setup.

Metabolite profiling of Chlamydomonas HCS knock-down mutant

Chlamydomonas cells were pre-cultured as described above and then transferred into fresh TAP medium and grown to middle logarithmic phase (~2 × 106 cells mL−1) under two continuous light conditions at 150 μmol⋅m−2⋅s−1 (normal light) and 50 μmol m−2 s−1 (low light). Cells were harvested by centrifugation at 4000 × g for 1 min, washed twice, and resuspended in ultra-pure water to an OD750 of 1.0. Cells in one milliliter of the cell suspension were collected by vacuum filtration through a membrane filter (0.45 μm HV Durapore 25 mm diameter; Millipore Sigma, Burlington, MA, USA). The filter was rolled up and placed into a 2 mL microcentrifuge tube, frozen in liquid nitrogen, and stored at −80 °C. Cellular metabolites were extracted according to Obata, Schoenefeld86 with slight modifications. The filter was soaked in 700 μL methanol supplemented with 5 μL of 0.2 mg/mL ribitol solution and vigorously vortexed for 15 s. Following 15 min incubation at 70 °C with shaking, the filter was removed, the solution was centrifuged at 17,500 × g for 10 min, and the supernatant was mixed with 750 μL water and 325 μL chloroform by vortexing. After centrifugation at 17,500 × g for 10 min, 400 μL of the upper aqueous phase was taken into a new microcentrifuge tube and dried down in a centrifugal vacuum concentrator (CentriVap, Labconco, Kansas City, MO, USA). Dried metabolites were dissolved into 20 mg/mL methoxyamine hydrochloride solution in pyridine and incubated for 2 h at 37 °C with shaking for methoxyamination. The metabolites were further trimethylsylilated by incubating for 30 min at 37 °C with shaking after the addition of 70 μL N-Methyl-N-(trimethylsilyl) trifluoroacetamide (Millipore Sigma). Gas chromatography-mass spectrometry (GC-MS) data acquisition, peak annotation, and quantification were conducted87, and the peak heights of the representative ions for each metabolite peak were normalized by that of internal standard (ribitol) to represent relative levels of metabolites. The parameters used for the peak annotation are provided in Supplementary Data 1.

Functional complementation of Arabidopsis ccmh line

Seeds for a ccmh (At1g15220) T-DNA line in the Arabidopsis (Arabidopsis thaliana) Columbia-0 background were obtained from the Arabidopsis Biological Resource Center (Ohio State University, Columbus, OH). The insertional ccmh line (SALK_046872) was genotyped by PCR using the primers listed in Supplementary Data 3 and confirmed by sequencing.

A codon optimized CriHCCS CDS sequence (system III) was synthesized (GenScript, USA) and assembled downstream of the Ubiquitin 10 promoter to create the vector pUBQ10::CriHCCS. The constructed plasmids were then transformed into viable heterozygous ccmh/CCMH plants by the floral dip method via Agrobacterium tumefaciens strain GV3101. Homozygous ccmh plants due to complementation were isolated by selecting seeds on plates containing 10 μM glufosinate ammonium (BASTA, for transformant selection) and PCR genotyping. Progenies from independent transformants were verified for ccmh T-DNA homozygosity and CriHCCS presence by PCR genotyping with the primers listed in Supplementary Data 3.

Seeds collected from the T3 complemented lines and the WT plants were surface-sterilized (70% v/v ethanol and 50% bleach) and sown on Murashige and Skoog (MS) agar plates (half strength MS medium, 0.05% w/v MES and 0.75 w/v agar, pH 5.8) followed by stratification at 4 °C in the dark for 3 days. One-week-old seedlings were transferred to soil and grown under controlled long-day conditions (16 h light/8 h dark, 22 °C, 50% relative humidity) for growth-stage-based phenotypic observation. Ten days after flowering, siliques were dissected using a fine tweezer and visualized under a stereomicroscope (SMZ25, Nikon).

Seed germination test was conducted by transferring the MS plates after cold stratification to light (100 μmol m−2 s−1) at 22–24 °C. Germination rate was scored as the percentage of seeds with visible radicle protrusion at each time point. Approximately 100 seeds were used for each line. Seedlings from two-days-old germination plates were transferred to the surface of new MS plates and placed vertically in a growth chamber (16 h light and 8 h dark period with approximately 100 µmol m−2 s−1 illumination at 22 °C. Root length was measured at day 14 after seed sowing. Both experiments were repeated three times.

Phylogenetic analysis of eukaryotic HCCS homologs

Eukaryotic HCCS homologs (accession numbers listed in Supplementary Fig. 3) were identified by using Archaeplastida HCCS sequences (Supplementary Table 2) in an online blastp search (https://blast.ncbi.nlm.nih.gov/Blast.cgi) against the NCBI non-redundant protein database and in a tblastn v2.2.31 search against newly assembled transcriptomes (Supplementary Data 2). HCCS proteins were aligned using MAFFT version 7.24588 with the most accurate ‘linsi’ setting. Alignments were trimmed to remove poorly aligned regions using trimAl v1.289 with the trimgappy setting. Maximum-likelihood trees were built from the trimmed alignment with RAxML v8.2.490 using the PROTGAMMAAUTO model option, and branch support was assessed with 1000 rapid bootstrap replicates. Bayesian inference trees were constructed with MrBayes v3.2.791 using default priors and two independent runs, each consisting of one cold and three heated chains. Runs were completed after 5 million generations, with trees sampled every 1000 generations and the first 25% of the sampled trees discarded as burn-in. AU topology tests were performed with IQ-TREE v2.2.2.692 by comparing the ML tree to alternatively constrained trees.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.