Comparative Plastid Genomics of Cryptomonas Species Reveals Fine-Scale Genomic Responses to Loss of Photosynthesis

Abstract

Loss of photosynthesis is a recurring theme in eukaryotic evolution. In organisms that have lost the ability to photosynthesize, nonphotosynthetic plastids are retained because they play essential roles in processes other than photosynthesis. The unicellular algal genus Cryptomonas contains both photosynthetic and nonphotosynthetic members, the latter having lost the ability to photosynthesize on at least three separate occasions. To elucidate the evolutionary processes underlying the loss of photosynthesis, we sequenced the plastid genomes of two nonphotosynthetic strains, Cryptomonas sp. CCAC1634B and SAG977-2f, as well as the genome of the phototroph Cryptomonas curvata CCAP979/52. These three genome sequences were compared with the previously sequenced plastid genome of the nonphotosynthetic species Cryptomonas paramecium CCAP977/2a as well as photosynthetic members of the Cryptomonadales, including C. curvata FBCC300012D. Intraspecies comparison between the two C. curvata strains showed that although their genome structures are stable, the substitution rates of their genes are relatively high. Although most photosynthesis-related genes, such as the psa and psb gene families, were found to have disappeared from the nonphotosynthetic strains, at least ten pseudogenes are retained in SAG977-2f. Although gene order is roughly shared among the plastid genomes of photosynthetic Cryptomonadales, genome rearrangements are seen more frequently in the smaller genomes of the nonphotosynthetic strains. Intriguingly, the light-independent protochlorophyllide reductase comprising chlB, L, and N is retained in nonphotosynthetic SAG977-2f and CCAC1634B. On the other hand, whereas CCAP977/2a retains ribulose-1,5-bisphosphate carboxylase/oxygenase-related genes, including rbcL, rbcS, and cbbX, the plastid genomes of the other two nonphotosynthetic strains have lost the ribulose-1,5-bisphosphate carboxylase/oxygenase protein-coding genes.

Introduction

Eukaryotic photosynthesis has evolved on multiple occasions through endosymbioses between nonphotosynthetic eukaryotes and photosynthetic partners. The progenitor of the primary plastid of Archaeplastida was very clearly a cyanobacterial endosymbiont, and subsequent eukaryote–eukaryote endosymbioses involving both red and green algal endosymbionts gave rise to various extant photosynthetic eukaryotes (Zimorski et al. 2014; Archibald 2015; Nowack and Weber 2018).

Plastids are the eukaryotic organelles that house the photosynthetic machinery that generates organic carbon in plant and algal cells. However, plastids are also the location of various other biochemical processes other than photosynthesis. Some plants and protists have secondarily lost the ability to photosynthesize but still retain a plastid, which generally retains a genome (a notable exception is the colorless green alga Polytomella) (Smith and Lee 2014; Hadariová et al. 2018). Such nonphotosynthetic plastids are known to maintain essential functions such as the biosynthesis of fatty acids, isoprenoids, and amino acids. The genomes of nonphotosynthetic plastids tend to still encode some essential proteins for plastid metabolism, despite reductions in size and the number of protein-coding genes. The protein-coding gene repertoire of the plastid genomes of nonphotosynthetic organisms varies, as does the suite of biological functions taking place in these derived organelles. For example, the plastid genomes of nonphotosynthetic Cryptomonas paramecium (Cryptophyceae), Euglena longa (Euglenophyceae), and Aneura mirabilis (parasitic plant) contain the gene for the large subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) (Gockel and Hachtel 2000; Wickett et al. 2008; Donaher et al. 2009), although other lineages, such as Nitzschia sp. (a diatom), Plasmodium spp. (Apicomplexa), Epifagus virginiana (a parasitic plant), and chrysophytes (Stramenopila) have lost RuBisCO and proteins related to its function (Wolfe et al. 1992; Wilson et al. 1996; Kamikawa, Tanifuji, et al. 2015; Dorrell et al. 2019). The retention of ATP synthase in some nonphotosynthetic plastids such as Nitzschia sp. (a diatom), C. paramecium (Cryptista), and several parasitic plants is also intriguing, but the exact roles of the ATP synthase complex in nonphotosynthetic plastids are debated (Wickett et al. 2008; Donaher et al. 2009; Kamikawa, Tanifuji, et al. 2015; Suzuki et al. 2018).

Cryptista is a unicellular eukaryotic group comprising plastid-lacking (i.e., Cyathomonadacea, Kathablepharidacea, and Palpitomonas) and plastid-bearing (i.e., Cryptomonadales) organisms (Yabuki et al. 2014; Hoef-Emden and Archibald 2016; Tanifuji and Onodera 2017; Cenci et al. 2018; Adl et al. 2019) (fig.1). Cryptomonadales acquired their photosynthetic capabilities by engulfing a red algal endosymbiont and retain the endosymbiont-derived plastid and nucleus (the latter called the nucleomorph) (Moore and Archibald 2009; Tanifuji and Archibald 2014; Tanifuji and Onodera 2017). Although nucleomorph genomes possess fewer than 500 protein-coding genes, only a small proportion of them are associated with plastid functions (31 and 18 in photosynthetic and nonphotosynthetic strains, respectively; Moore and Archibald 2009; Tanifuji and Archibald 2014; Tanifuji and Onodera 2017); the majority of plastid proteins in Cryptomonadales are encoded in the nuclear and plastid genomes (Curtis et al. 2012; Hempel et al. 2014; Gould et al. 2015). To date, eight cryptomonad plastid genomes have been sequenced (Douglas and Penny 1999; Khan et al. 2007; Donaher et al. 2009; Kim et al. 2015, 2017). Within the genus Cryptomonas, the plastid genomes of photosynthetic species C. curvata FBCC300012D and the secondarily nonphotosynthetic C. paramecium CCAP977/2a were sequenced (Donaher et al. 2009; Kim et al. 2017). However, phylogenetic surveys have revealed that three nonphotosynthetic Cryptomonas lineages are closely related to different photosynthetic species (Hoef-Emden 2005, 2007; Hoef-Emden et al. 2005), suggesting that members of the genus Cryptomonas have lost the ability to photosynthesize on at least three separate occasions. This provides an opportunity to study the dynamics of plastid genome structure and coding capacity as they relate to the loss of photosynthesis over short evolutionary timescales. Unfortunately, genomic sampling is presently sparse; only one plastid genome from a nonphotosynthetic Cryptomonas species has been sequenced, that of C. paramecium CCAP977/2a (Donaher et al. 2009).

Fig. 1.

—Schematic tree of Cryptista. Tree topology and classification were adopted from Adl et al. (2019), Hoef-Emden (2005, 2007), Hoef-Emden et al. (2005), and Yabuki et al. (2014). White and black circles indicate photosynthetic and nonphotosynthetic organisms, respectively. The plastid genomes sequenced in this study are shown by black arrowheads. The independent origins of nonphotosynthetic Cryptomonas were robustly suggested by the phylogenetic analyses based on nuclear internal transcribed spacer 2, nuclear LSU rRNA, and nucleomorph SSU rRNA sequences (Hoef-Emden 2005, 2007; Hoef-Emden et al. 2005).

To rectify this situation, we sequenced two plastid genomes of nonphotosynthetic Cryptomonas species, Cryptomonas sp. SAG977-2f and CCAC1634B, which independently lost the ability to photosynthesize and are distantly related to C. paramecium CCAP977/2a. We also sequenced the plastid genome of the photosynthetic species C. curvata CCAP979/52 and compared it to that of another strain, FBCC300012D, to provide the first insight into intraspecies diversity of plastid genomes in Cryptomonas. Comparative analyses of the plastid genomes of photosynthetic and nonphotosynthetic members of the genus Cryptomonas provide snapshots of the loss of photosynthesis, including the pseudogenization of photosynthesis-related genes and DNA elimination steps. Our results also support the idea that genome rearrangements are accelerated by the loss of inverted repeats (IRs) of ribosomal RNA operons (e.g., Palmer and Thompson 1982; Strauss et al. 1988; Wei et al. 2013; Vieira Ldo et al. 2014). Furthermore, unexpected variability in the repertoires of genes for photosynthetic functions was observed among the plastid genomes of nonphotosynthetic Cryptomonas.

Materials and Methods

Cell Culturing, DNA Extraction, and Genome Sequencing

Cryptomonas curvata CCAP979/52, Cryptomonas sp. SAG977-2f, and Cryptomonas sp. CCAC1634B were obtained from the Culture Collection of Algae and Protozoa (CCAP), the Sammlung von Algenkulturen der Universität Göttingen (SAG) and the Culture Collection of Algae at the University of Cologne (CCAC), respectively. Cryptomonas curvata CCAP979/52 was cultured in MWC medium under 12:12 light/dark conditions at 21 °C (Guillard and Lorenzen 1972). Cryptomonas sp. CCAC1634B was maintained in Warris-H medium (McFadden and Melkonian 1986) with 1% volume of bacteria standard medium under dark conditions at 18 °C. DNA extractions for C. curvata CCAP979/52 and Cryptomonas sp. CCAC1634B were done using the standard SDS–phenol/chloroform extraction method. For C. curvata CCAP979/52, total DNA was separated into nuclear and organelle DNA-enriched fractions (i.e., plastid, mitochondrion, and nucleomorph) by Hoechst dye (Sigma-Aldrich)–cesium chloride density gradient centrifugation at 35,000 rpm for 65 h at 4 °C, as described previously (Tanifuji et al. 2011, 2014; Moore et al. 2012). An organelle DNA-enriched fraction of C. curvata CCAP979/52 was subjected to sequencing library construction using the Nextera XT DNA Library Preparation Kit (Illumina), and DNA sequencing was carried out using a Mi-Seq instrument (Illumina). Because Cryptomonas sp. SAG977-2f did not grow well in the laboratory, cells received directly from SAG were immediately subjected to genome amplification using the REPLI-g Mini Kit (Qiagen) following the manufacturer’s instructions. The amplified genomic DNA of Cryptomonas sp. SAG977-2f and the total DNA of Cryptomonas sp. CCAC1634B were sent to Hokkaido System Science (Hokkaido, Japan) for TruSeq library construction and Illumina HiSeq sequencing.

Genome Assembly

For C. curvata CCAP979/52, a data set containing 7.75 million paired-end reads 300 base-pairs (bp)-long were obtained. After sequence quality check using FastQC software (https://www.bioinformatics.babraham.ac.uk/projects/fastqc, last accessed January 21, 2020), positions with average quality scores <30 were trimmed down and 271 (forward) and 221 (reverse)-base reads were retained. Paired-end reads (101 base-long; 88.75 million and 154.46 million) were obtained for Cryptomonas sp. SAG977-2f and Cryptomonas sp. CCAC1634B, respectively. The first and last four bases of raw reads were trimmed due to their low sequence quality. The remaining 94-bp reads were subjected to additional quality control steps. Trimmed sequence reads in which >80% of the bases had quality scores of >20 were extracted using FASTAX-Toolkit (Ver. 0.0.14) (http://hannonlab.cshl.edu/fastx_toolkit/, last accessed January 21, 2020). Preliminary assembly was carried out using SPAdes ver. 3.5.0 (Bankevich et al. 2012) with the remaining reads for each strain. The putative plastid genome sequences were retrieved from the resulting scaffolds by BlastN similarity searches using the plastid genome sequences of C. paramecium CCAP977/2a and Guillardia theta as queries (GenBank accession numbers: GQ358203 and AF041468 [Douglas and Penny 1999; Donaher et al. 2009]). Because the read depth coverages of the plastid genome scaffolds in the preliminary assemblies were “too high” for optimal assembly (>1,200×) in Cryptomonas sp. CCAC1634B, 10% of the raw reads were subjected to a second-round assembly (Tanifuji et al. 2014). The CCAC1634B plastid genome contig that resulted from this second-round assembly was nearly identical to that of the first round, differing only at the boundary regions. The assembly of data from Cryptomonas sp. SAG977-2f was performed using the whole raw reads with the single cell mode of SPAdes (Bankevich et al. 2012). Putative plastid genome scaffolds were then assembled into two and one scaffolds for Cryptomonas sp. SAG977-2f and Cryptomonas sp. CCAC1634B, respectively. The gaps between scaffolds were filled by PCR and Sanger sequencing.

Genome Annotation and Comparative Analyses

Protein and tRNA gene predictions were performed by MFANNOT (http://megasun.bch.umontreal.ca/cgi-bin/mfannot/mfannotInterface.pl, last accessed January 21, 2020) and tRNAscan-SE (Lowe and Eddy 1997). The ribosomal RNA operons were predicted by BlastN searches using known plastid RNA operon sequences as queries. Individual gene models were confirmed manually using Artemis software (Rutherford et al. 2000), and unannotated nucleotide sequences were subjected to BlastX searches against the NCBI nr database to find putative pseudogenes and protein-coding regions with weak similarity to known proteins/genes. Genome statistics such as GC content were also generated using Artemis.

In the C. curvata CCAP979/52 genome, a putative psbN-associated group II intron was predicted by MFANNOT and MFOLD (Zuker 2003), and the full intron structure and associated sequence elements were manually curated. This included identification of a reverse transcriptase coding region located 230-bp upstream of “domain V,” the splicing motif (5′-GUGYG…AY-3′) consensus sequence, and the psbN open reading frame itself. The predicted secondary structure of the intron was concordant with known group II introns in other Cryptomonadales (Khan and Archibald 2008 ; Kim et al. 2017).

Genome synteny analyses were carried out using the Mauve multiple genome aligner (Darling et al. 2004). Detection and analysis of sequence repeats was carried out using the mreps software (Kolpakov et al. 2003).

The plastid genome sequences and gene annotations determined in this study were deposited under DDBJ accession numbers LC484192 (C. curvata CCAP979/52), LC484193 (Cryptomonas sp. CCAC1634B), and LC484194 (Cryptomonas sp. SAG977-2f).

Synonymous and Nonsynonymous Substitution Rates Test

For estimation of synonymous (d_S) and nonsynonymous (d_N) substitution rates in the C. curvata CCAC979/52 and FBCC300012D plastid genomes, 146 genes from each genome were individually aligned by mafft (Katoh et al. 2005). Codon alignments were generated by pal2Nal.pl (Suyama et al. 2006) with the “no gap” option. The alignments were then concatenated resulting in a supermatrix containing 96,171 bp. To collect the rbcL gene sequences of Cryptomonadales, the rbcL sequence of C. curvata CCAP979/52 was used as a blast query against the NCBI nt database. Seventy-one rbcL sequences of Cryptomonadales were obtained and aligned by mafft (Katoh et al. 2005). Short sequences and gap sites were removed and a multiple alignment consisting of 64 Cryptomonadales sequences and 852 bp was generated. Sequences from redundant and unidentified species were removed. During this process, Cryptomonas, Rhodomonas, and Chroomonas were retained for the purposes of interspecific comparison. Substitution rates for all data sets were estimated using the codeml programs of PAML4.8 with the F3 x 4 codon model (Yang, 2007).

Results and Discussion

Intraspecific Plastid Genome Diversity in C. curvata

We determined the plastid genome sequence of the photosynthetic species C. curvata CCAP979/52 and compared it with that of the previously sequenced C. curvata FBCC300012D (fig. 2 and table 1). Together these two genomes serve as important reference points for investigating the loss of photosynthesis in Cryptomonas spp. In terms of structure and coding capacity, the two C. curvata plastid genomes are generally similar to those of other photosynthetic Cryptomonadales (table 1) (Douglas and Penny 1999; Khan et al. 2007; Kim et al. 2015, 2017). The plastid genome of the CCAP979/52 strain is 1,689-bp longer and has a lower GC content compared with FBCC300012D (i.e., 129,974 and 128,285 bp in size and 34.6% and 35.3% GC content in the CCAP979/52 and FBCC300012D strains, respectively) (table 1; Kim et al. 2017). The number of predicted protein-coding genes in C. curvata CCAP979/52 and FBCC300012D is 148 and 146, respectively (table 1 and supplementary table S1, Supplementary Material online). The total amount of noncoding DNA in the CCAP979/52 plastid genome is 19,325 bp, 1,117 bp less than that of strain FBCC300012D (table 1). Two genes, orf264 (annotated as ycf55 in Chroomonas) and ycf20, both showing significant sequence similarity to homologs in Chroomonas mesostigmatica, were detected in the plastid genome of the CCAP979/52 strain. However, the FBCC300012D plastid genome encodes only partial protein sequences of orf264 (ycf55) and ycf20, from which the N-terminal regions, including a viable initiation methionine, are missing. Although the ycf20 protein is encoded in the plastid genomes of Guillardia, Storeatula, and Chroomonas, it is missing in Rhodomonas and Teleaulax (Douglas and Penny 1999; Khan et al. 2007; Kim et al. 2015, 2017). Therefore, these two genes appear to be dispensable for plastid function and cell viability in Cryptomonadales. The gene order, including the position of a psbN intron, is identical between the two C. curvata plastid genomes, with the exception of the presence/absence of orf264 and ycf20 (figs. 2 and 3) (see below for more discussion).

Fig. 2.

—Circular physical maps of the plastid genome of (A) Cryptomonas curvata CCAP979/52, (B) Cryptomonas sp. SAG977-2f, and (C) Cryptomonas sp. CCAP1634B. Genes shown on the outside the circle are transcribed clockwise. Annotated genes are colored according to the functional categories shown in the center. Recognizable pseudogenes are emphasized by arrowheads.

Table 1:

Overview of Cryptomonadales Plastid Genomes

	The Photosynthetic Plastid Genomes in Other Genera of Cryptomonadalesa	Cryptomonas curvata FBCC300012Da	Cryptomonas curvata CCAP979/52	Cryptomonas sp. SAG977-2f	Cryptomonas sp. CCAC1634B	Cryptomonas paramecium CCAP977/2aa
Genome size (bp)	121,524–140,953	128,285	129,974	106,661	80,503	77,717
No. of protein genes	143–149	147	148	90	84	82
No. of tRNAs	30–32	31	31	30	29	29
GC contents (%)	32–36	35.3	34.59	34.6	39.7	38.1
Amount of noncoding DNA (%)b	14,792–27,796 bp (12.2–20.3%)	20,442 bp (15.9%)	19,325 bp (14.9%)	32,281 bp (30.2%)	10,704 bp (13.3%)	9,917 bp (12.8%)

^aData adopted from Donaher et al. (2009), Douglas and Penny (1999), Khan et al. (2007), and Kim et al. (2015, 2017).

^bPseudogenes were not considered.

Fig. 3.

—Alignment of plastid genomes of photosynthetic and nonphotosynthetic Cryptomonadales. Images were generated using the Mauve genome alignment tool (Darling et al. 2004). Green, blue, red, and yellow boxes indicate corresponding locally collinear blocks (LCBs), conserved segments among genomes identified by Mauve (Darling et al. 2004). The upper three lines correspond to the plastid genomes of three select photosynthetic species (Guillardia theta [Douglas and Penny 1999], Chroomonas mesostigmatica [Kim et al. 2017], and Rhodomonas salina [Khan et al. 2007]). The middle section shows Cryptomonas curvata FBCC300012D and CCAP9779/52. The lower lines indicate the genomes of Cryptomonas sp. SAG977-2f, Cryptomonas sp. CCAP1634B, and Cryptomonas paramecium CCAP977/2a. The rRNA operons, including SSU and LSU rRNA regions shown in orange boxes, were added manually. Dotted lines show rearranged regions in CCAC1634B (dark yellow) and CCAP977/2a (dark green), respectively.

Pairwise comparison between the two C. curvata plastid genomes revealed an overall sequence identity level of 79% with 27,843 substitutions and 3,929 bp of gap sequences. The sequence identity of rRNA regions between the two C. curvata strains was determined to be 1,468/1,490 bp (99%) in the small subunit and 2,736/2,825 bp (97%) in the large subunit. The observed intraspecific genetic diversity seen in C. curvata is higher than that typically seen in other algae. For example, an analysis of 13 strains of the green alga Ostreococcus tauri revealed only 314 single nucleotide polymorphisms sites (15–156 substitutions against consensus sequence) in ∼72 kb of plastid genome sequence (Blanc-Mathieu et al. 2013). To obtain further insight into the genetic diversity between the two C. curvata strains, we measured synonymous (d_S) and nonsynonymous (d_N) substitution rates in a pairwise comparison of 146 plastid protein-coding genes as well as those in a concatenated data set (supplementary table S2, Supplementary Material online). The d_S and d_N values between the two C. curvata strains are 7.4467 (d_S) and 0.0716 (d_N) in an average of the genes and 2.6798 (d_S) and 0.0609 (d_N) in the concatenated data set. These substitution rates are higher than that observed in a multistrain examination of another Cryptomonadales species, Hemiselmis andersenii (d_S = 0.0017, d_N = 0.0007) (Grisdale et al. 2019).

Hoef-Emden (2005) and Hoef-Emden et al. (2005) hypothesized that an accelerated evolutionary rate in a photosynthetic ancestor of Cryptomonas triggered loss of photosynthesis in some present-day strains (Hoef-Emden 2005; Hoef-Emden et al. 2005). Indeed, evolutionary rates of the plastid rbcL and nucleomorph SSU rRNA genes in particular Cryptomonas species, including both nonphotosynthetic and photosynthetic species (designated as “long branching”), were found to be accelerated relative to other Cryptomonas species (Hoef-Emden 2005; Hoef-Emden et al. 2005). That said, genetic diversity between different genera within Cryptomonadales has not been investigated. Here, we measured d_S and d_N rates of available rbcL genes in Cryptomonadales (supplementary table S3, Supplementary Material online). The averages of d_S and d_N rates of rbcL genes among 12 Cryptomonas species were found to be 1.5286 (d_S) and 0.0134 (d_N) (0.53 and 0.0045 between the two C. curvata strains), which is higher than that seen for six Rhodomonas species (0.30 [d_S] and 0.0052 [d_N]) and six Hemiselmis species (0.17 [d_S] and 0.0095 [d_N]) (supplementary table S2, Supplementary Material online). Thus, in addition to the species within the long-branching lineage (Hoef-Emden 2005; Hoef-Emden et al. 2005), our data suggest that the evolutionary rate across the whole Cryptomonas genus is higher than that of other genera.

Common Features of Plastid Genomes in Nonphotosynthetic Cryptomonas Species

Complete circular mapping plastid genome sequences were successfully obtained for two nonphotosynthetic Cryptomonas species. The newly sequenced plastid genomes of Cryptomonas sp. SAG977-2f and Cryptomonas sp. CCAC1634B were estimated to be 106,661 and 80,503 bp in size, respectively. Other features, including GC content and amount of noncoding DNA, are shown in table 1. Cryptomonas sp. SAG977-2f and Cryptomonas sp. CCAC1634B contain 90 and 84 protein-coding genes in their plastid genomes, respectively (fig. 2, table 1, and supplementary table S1, Supplementary Material online). Relative to the photosynthetic Cryptomonadales, including those discussed above (Douglas and Penny 1999; Khan et al. 2007; Donaher et al. 2009; Kim et al. 2015, 2017), ∼60 protein-coding genes are missing from the Cryptomonas sp. SAG977-2f and Cryptomonas sp. CCAC1634B genomes, including photosystems I and II (i.e., the psa and psb gene families) and the cytochrome b₆/f complex (i.e., the pet gene family) (table 1 and supplementary table S1, Supplementary Material online), or have become pseudogenized (see section below), as seen in the previously sequenced C. paramecium CCAP977/2a genome (Douglas and Penny 1999; Khan et al. 2007; Donaher et al. 2009; Kim et al. 2015, 2017). The loss of photosynthesis-related genes is reported in other lineages of photosynthesis-lacking organisms (Hadariová et al. 2018). The number of tRNA gene copies was slightly reduced (by 1 or 2 genes) in the plastid genomes of nonphotosynthetic species relative to the photosynthetic C. curvata FBCC300012D (Kim et al. 2017). Although two copies of isoleucyl-tRNA with anticodon GAU (trnI [gau]) were found in the C. curvata CCAP977/52 and FBCC300012D plastid genomes, the nonphotosynthetic strains possessed a single copy of trnI (gau). In addition, alanyl-tRNA with anticodon UGC (trnA [ugc]) was present in a single copy in both the Cryptomonas sp. CCAC1634B and C. paramecium CCAP977/2a plastid genomes, whereas two strains of C. curvata and Cryptomonas sp. SAG977-2f contained two copies.

Out of ∼90 proteins encoded in the plastid genomes of three nonphotosynthetic Cryptomonas, more than half were associated with housekeeping functions such as translation and transcription (e.g., ribosomal proteins, initiation factors, and RNA polymerase) (fig. 2 and supplementary table S1, Supplementary Material online). The second largest proportion of retained genes are those involved in the ATP synthase complex. Genes for eight subunits of the ATP synthase complex were found, namely, atpA, atpB, atpD, atpE, atpF, atpG, atpH, and atpI, although atpF in C. paramecium CCAP977/2a was found to be a pseudogene (Donaher et al. 2009). Many (but not all) secondarily nonphotosynthetic eukaryotes that are distantly related to Cryptista, such as certain diatoms and parasitic plants, retain plastid ATP synthase complex genes in their plastid genomes (Wicke et al. 2013; Kamikawa, Tanifuji, et al. 2015; Kamikawa et al. 2018; Suzuki et al. 2018), indicating a constraint against the loss of those genes from the plastid genome even after the loss of photosynthesis. Kamikawa, Tanifuji, et al. (2015) hypothesized that in the plastid of nonphotosynthetic organisms the ATP synthase acts to maintain a proton gradient between the thylakoid lumen and stroma, which is required for the twin arginine translocator (Tat) system (Kamikawa, Tanifuji, et al. 2015). In addition to ATP synthase complex genes, a core protein of the Tat system (TatC) was also maintained in the plastid genomes of the three independently evolved nonphotosynthetic Cryptomonas lineages investigated here. Although nonphotosynthetic plastids generally lack conspicuous thylakoid structures, “thylakoid-like structures” were observed in nonphotosynthetic Nitzschia species (diatoms) (Kamikawa, Yubuki, et al. 2015), supporting the importance of, and evolutionary constraint for, retention of Tat system-mediated protein translocation into the thylakoid lumen in some nonphotosynthetic plastids. On the other hand, given that some nonphotosynthetic green algae, such as Prototheca and Helicosporidium, possess ATP synthase complex genes but lack the Tat system (Suzuki et al. 2018), the exact role(s) of the ATP synthase complex in nonphotosynthetic plastids is still unclear. As suggested previously (Kamikawa et al. 2017), the ATP synthase complex might also contribute to maintenance of a higher stromal pH so that certain plastid proteins can be activated under alkaline conditions. There may in fact be different reasons for the retention of the ATP synthase in plastid genomes of independently evolved nonphotosynthetic organisms.

We found genes encoding iron–sulfur cluster assembly proteins (sufB and sufC), acetolactate synthase subunits (ilvB and ilvH), and an acyl carrier protein (acpP) in the three plastid genomes of nonphotosynthetic Cryptomonas strains; this suggests that the functions of iron–sulfur cluster assembly, amino acid biosynthesis, and lipid metabolism are functional. In addition, genes for a few secretory proteins (secA and secY), the chaperonin groEL, a protease (clpC), ferredoxin (petF), and plastid membrane protein cemA were commonly retained. In summary, the plastid genomes that have independently become nonphotosynthetic have lost the protein-coding genes directly related to photosynthesis (i.e., photosystems I and II and cytochrome b₆/f complex), but retained part of the ATP synthase complex, protein translocators, iron–sulfur cluster assembly, amino acid synthesis, lipid metabolism, and genes involved in their expression and maintenance (e.g., housekeeping genes and chaperonins).

Genome Reduction Steps: Pseudogenization and DNA Elimination

Among the three nonphotosynthetic species examined, we found that the plastid genome of SAG977-2f is the largest, being ∼26 kb larger than that of the other Cryptomonas spp. The main reason for this size variation is the larger intergenic spacers in the SAG977-2f genome. Whereas 12–20% of the plastid genomes in other Cryptomonadales are intergenic, regardless of whether or not they were photosynthetic, 30.2% of the SAG977-2f plastid genome is noncoding (table 1 and fig. 2). The total amount of intergenic DNA in the Cryptomonas sp. SAG977-2f, CCAC1634B, and C. paramecium CCAP977/2a plastid genomes is 32,281, 10,704, and 9, 917 bp (table 1), respectively. The Cryptomonas sp. SAG977-2f plastid genome contains 17 short (10–18 bp) repeats, similar to, or slightly lower than, the amount of obviously repetitive DNA in the plastid genomes of photosynthetic relatives (e.g., 10–20 and 18–23 bp repeats in C. curvata FBCC300012D and G. theta, respectively). Given that long stretches of repetitive sequence are conspicuously absent in the SAG977-2f plastid genome, it seems unlikely that their longer intergenic regions are the product of genome expansion.

A remarkable feature of the SAG977-2f plastid genome is the presence of pseudogenes. Although SAG977-2f has lost almost all of its plastid genes related to photosynthesis, as have strains CCAC1634B and CCAP977/2a, partial fragments homologous to at least ten genes (i.e., lysR, ftrB, psaF, psbB, psbV, ompR, ycf3, ycf39, rbcL, and cbbX) were identified in the plastid genome (fig. 2 and supplementary fig. S1, Supplementary Material online). The most striking example is lysR, where the 751-bp intergenic region between orf264 and rpl32 showed significant similarity to lysR of photosynthetic relatives. The pairwise alignment of the most significant high scoring pairs against the homolog of C. curvata CCAP979/52 is shown in supplementary figure S1, Supplementary Material online, which includes six in-frame stop codons, clearly indicating that this is a pseudogene of lysR. The alignment of another pseudogene, that of ycf3, is also shown in supplementary figure S1, Supplementary Material online. The intergenic region between infB and atpB showed similarity to ycf3 of photosynthetic relatives, and two high scoring pairs showed positive alignments at 39/59 amino acids (63%) and 24/37 amino acids (65%) in blast searches, respectively (supplementary fig. S1, Supplementary Material online). These ten pseudogenes are clearly nonfunctional due to their fragmentation, the presence of in-frame stop codons, and/or the absence of an initiation methionine codon.

Pseudogenes in nonphotosynthetic plastid genomes have also been seen in other lineages. For example, the parasitic plants Aneura mirabilis and Epifagus virginiana possess pseudogenes of photosystem proteins in their plastid genomes (Wolfe et al. 1992; Wickett et al. 2008). Intriguingly, despite the pseudogenization of some photosystem I and II proteins, open reading frames for some other photosystem subunits appeared to be intact, suggesting that photosynthesis was lost in these organisms very recently (Wickett et al. 2008). The existence of multiple pseudogenes in SAG977-2f, combined with the lack of detectable pseudogene counterparts in the other nonphotosynthetic Cryptomonas plastid genomes, suggests that the loss of photosynthesis in SAG977-2f occurred more recently in the lineage to which SAG977-2f belongs compared with the other nonphotosynthetic Cryptomonas. Interestingly, the total length of the ten pseudogenes in the SAG977-2f plastid genome is 3,609 bp, which does not account for the ∼20 kb of “extra” sequence compared with the plastid genomes of other nonphotosynthetic Cryptomonas (table 1). This suggests that at least some of these longer intergenic regions are in fact highly degenerate—and at present bioinformatically undetectable—pseudogenes that have yet to be purged from the genome. All things considered, given that the Cryptomonas strains considered here lost their photosynthetic capacities independently, this three-way comparison provides snapshots in the process of the loss of photosynthesis and accompanying genome reduction in eukaryotes.

The Loss of IRs of Ribosomal RNA Operons May Accelerate Genome Rearrangement

A notable structural difference between the plastid genomes of photosynthetic and nonphotosynthetic Cryptomonadales is the presence or absence of the IRs of ribosomal RNA operons. All of the photosynthetic Cryptomonadales plastid genomes from eight strains among six genera contain IRs (Douglas and Penny 1999; Khan et al. 2007; Kim et al. 2015, 2017). In contrast, one part of the IR disappeared from the plastid genomes of the nonphotosynthetic members (figs. 2 and 3). It is possible that the loss of IRs is related to the genome rearrangements observed in C. paramecium CCAP977/2a and Cryptomonas sp. CCAC1634B (fig. 3), given that the genomes of all photosynthetic Cryptomonadales, including C. curvata FBCC300012D and CCAC979/52, maintain genome synteny (i.e., conserved gene order), despite lineage-specific gene and intron losses (fig. 3) (Douglas and Penny 1999; Khan et al. 2007; Kim et al. 2015, 2017). In contrast, in the plastid genome of SAG977-2f, the gene order is the same as in the photosynthetic plastid genomes, with the exception of simple gene losses (figs. 2 and 3). Thus, the apparent lack of genome rearrangements observed in the SAG977-2f plastid genome might also reflect the relatively recent loss of photosynthesis in the organism (fig. 3).

In this context, it is suggested that SAG977-2f is maintaining genome synteny, indicating that the loss of photosynthesis in Cryptomonas is not strictly associated with genome rearrangements. Although the three nonphotosynthetic Cryptomonas, namely, SAG977-2f, CCAC1634B, and CCAP977/2a, lost both photosynthesis and IRs, it is unclear whether the loss of photosynthesis and loss of the rRNA operon structure are always associated with each other in Cryptomonas. It is noteworthy that the loss of IRs has been seen in other lineages, such as photosynthetic green algae (Turmel et al. 2017). In contrast, all of the plastid genomes sequenced from nonphotosynthetic diatoms and a nonphotosynthetic chrysophyte retain IRs (Sabir et al. 2014; Kamikawa, Tanifuji, et al. 2015; Kamikawa et al. 2018; Dorrell et al. 2019). Many “apicoplast” genomes in apicomplexans are also extremely reduced but contain IRs (Arisue et al. 2012). These findings suggest that the loss of photosynthesis might be less commonly associated with the loss of an rRNA operon (Janouškovec et al. 2013; Turmel et al. 2017). More insight into the association between photosynthetic ability and loss of IRs in Cryptomonas will hopefully come from investigation of plastid genomes from organisms even more closely related to nonphotosynthetic Cryptomonas spp. than the C. curvata strains investigated herein.

The correlation between IR structure and frequency of genome rearrangements has been discussed elsewhere (e.g., Palmer and Thompson 1982; Strauss et al. 1988; Wei et al. 2013; Vieira Ldo et al. 2014). Strauss et al. (1988) hypothesized that IRs accelerate the frequency of homologous recombination between individual plastid genome molecules, stabilizing them against structural changes (Strauss et al. 1988). This idea has been discussed in the context of chlorarachniophytes, the other lineage of nucleomorph-bearing organisms (Tanifuji et al. 2016). In this case, a comparison of different chlorarachniophyte genera showed that whereas the IR-lacking mitochondrial genomes have undergone frequent rearrangements, the IR-containing plastid genomes exhibit a very high level of synteny (Tanifuji et al. 2016). Similar trends have been reported in green algae as well. For example, frequent genome rearrangements are observed in the IR-lacking plastid genomes of Chloropicon and Chloroparbula species (Turmel et al. 2019). The same pattern is seen here in Cryptomonadales. In this study, we have shown that genome rearrangements appear to be more frequent in plastid genomes without IRs (i.e., CCAP977/2a and CCAC1634B) than in IR-containing plastid genomes (fig. 3). Furthermore, genome rearrangements appear frequent in the mitochondrial genomes of Cryptomonadales (Tanifuji and Onodera 2017; Kim et al. 2018). All together, these findings speak to a correlation between the frequency of genome rearrangements and the presence/absence of IRs. Notably, because there is no reason why these phenomena would be limited to particular lineages, we expect the same tendency to be widely observable.

Diversity of Photosynthesis-Related Genes in Nonphotosynthetic Cryptomonas

Significant differences were observed in the variety of carbon fixation- and chlorophyll synthesis-related genes in nonphotosynthetic Cryptomonas. The plastid genome of C. paramecium CCAP977/2a contains genes for three components of RuBisCO, namely, rbcL, rbcS, and cbbX, despite them being absent in the SAG977-2f and CCAP1634B plastid genomes. The existence of residual RuBisCO-related genes in nonphotosynthetic organisms is found not only in Cryptomonas but also in several other eukaryotes; for example, some plants, dinoflagellates, stramenopiles, and euglenids retain RuBisCO genes without exhibiting the ability to photosynthesize (Wolfe and dePamphilis 1997; Gockel and Hachtel 2000; Sekiguchi et al. 2002; Sanchez-Puerta et al. 2007; Wickett et al. 2008). Because RuBisCO is the key carbon fixation enzyme, it is unclear why it is retained in those nonphotosynthetic plastids. One proposed explanation is that in nonphotosynthetic strains, RuBisCO is maintained for functions other than the Calvin–Benson cycle, improving carbon efficiency via conversion of hexose to pyruvate in lipid biosynthesis (Schwender et al. 2004; Krause 2008). If RuBisCO plays an essential role in C. paramecium CCAP977/2a, then RuBisCO-lacking Cryptomonas spp. SAG977-2a and CCAC1634B may have replaced its function with something else, either a different enzyme or a different process. Another possible explanation is that those RuBisCO proteins encoded in plastid genomes of nonphotosynthetic strains are undergoing evolutionary degradation prior to their complete loss. Consistent with this hypothesis, the RuBisCO protein sequence in the nonphotosynthetic euglenophyte E. longa is divergent compared with those of its photosynthetic relatives (Gockel and Hachtel 2000; Záhonová et al. 2016). In addition, the stability and abundance of RuBisCO in the cells are quite low, suggesting that the E. longa homolog is undergoing evolutionary degradation (Záhonová et al. 2016). Although a relaxation of selective constraint on RuBisCO genes in C. paramecium CCMP977/2a was suggested (Hoef-Emden et al. 2005), the d_N/d_S ratio of rbcL in our analyses was 0.0093, suggesting substantial functional constraints still exist (and/or its function has only very recently become unnecessary). At present, it is difficult to determine whether these genes are truly essential or in the process of being lost. Comprehensive comparative analyses between RuBisCO-retaining and RuBisCO-lacking nonphotosynthetic Cryptomonas species could help to explain the presence of RuBisCO in nonphotosynthetic plastids.

Another unexpected finding is that all three nonphotosynthetic Cryptomonas strains retain a partial chlorophyll biosynthetic pathway, albeit with different gene repertoires. The plastid genomes of Cryptomonas spp. SAG977-2a and CCAC1634B retain the light-independent protochlorophyllide reductase (LIPOR) consisting of chlL, N, and B. In contrast, C. paramecium CCAP977/2a lost the LIPOR components. All three nonphotosynthetic plastid genomes maintain Mg-protoporphyrin IX chelatase (chlI), which plays a role in the first step of the chlorophyll biosynthetic pathway, branching from the plastid heme biosynthetic pathway. In addition, genes for other subunits of Mg chelatase, chlD and chlH, and Mg-protoporphyrin IX methyltransferase (bchM) catalyzing next step of Mg-protoporphyrin IX chelatase were found in transcriptome data of C. paramecium CCAP977/2a (Keeling et al. 2014; Johnson et al. 2018). Overall, our bioinformatic investigations suggest that C. paramecium CCAP977/2a, Cryptomonas spp. SAG977-2f, and CCAP1634B retain at least part of the chlorophyll biosynthetic pathway for synthesizing protochlorophylide (CCAP977/2a) or chlorophylide a (CCAP1634B and SAG977-2f), which are precursors of chlorophyll a synthesis.

Extant cyanobacteria possess both LIPOR and light-dependent protochlorophyllide reductase (POR). Although these two enzymes are isofunctional, their origins differ. LIPOR likely arose from nitrogenase in anoxygenic photosynthetic bacteria (Fujita and Bauer 2003; Muraki et al. 2010). In addition, POR with high similarity to the short-chain dehydrogenase reductase evolved in oxygenic photosynthetic bacteria (Suzuki and Bauer 1995; Hunsperger et al. 2015). In eukaryotes, the genes encoding POR have been transferred to the nuclear genome. In contrast, the genes encoding LIPOR remain in some plastid genomes but have been lost outright in many others (i.e., not transferred to the nuclear genome). Additionally, gene duplication and horizontal transfer of POR gave rise to various distribution patterns of coding POR in eukaryotes (Hunsperger et al. 2015; Matsuo and Inagaki 2018). In photosynthetic Cryptomonadales, LIPOR was found to be intact in C. curvata and Storeatula sp. CCMP 1868 (Kim et al. 2017). In contrast, G. theta and Teleaulax amphioxeia completely lost LIPOR from their plastid genomes (Douglas and Penny 1999; Kim et al. 2015), and Rhodomonas salina, Chroomonas placoidea, and Chroomonas mesostigmatica appear to have LIPOR only as pseudogenes (Khan et al. 2007; Kim et al. 2017). These distribution patterns of LIPOR suggest that nuclear POR is the main POR in Cryptomonadales, whereas plastid LIPOR is dispensable. On the other hand, it is still unclear whether nonphotosynthetic Cryptomonas species are currently undergoing gene loss because, despite the complete loss or degradation of all other photosynthesis-related proteins, they have retained intact chlorophyll biosynthesis enzymes. Furthermore, because chlorophyll precursors are photo-toxins generating reactive oxygen species (ROS), it is likely that those proteins potentially causing ROS generation disappear at an early stage of the loss of photosynthesis. This implies that they might have biological significance, such as in diverting chlorophyll precursors required for still unknown plastid functions in nonphotosynthetic Cryptomonas species. Future study of their functions is necessary.

Concluding Remarks

Despite having lost photosynthesis independently, the three nonphotosynthetic Cryptomonas species examined herein show similar patterns of gene retention in their plastid genomes. Structural changes in their genomes presumably occurred convergently via the same basic steps: 1) pseudogenization of photosynthesis-related genes tolerated by the loss of photosynthesis (as in SAG977-2f), 2) elimination of pseudogenic DNA, resulting in genome reduction (as seen in the plastid genomes of all three nonphotosynthetic strains), and 3) genome rearrangement accelerated by the loss of the rRNA operon-associated IRs (shown in CCMP977/2a and CCAP1634B). It is not clear whether steps 2 and 3 occurred sequentially or roughly concurrently. Although three nonphotosynthetic Cryptomonas have all lost the IRs, it is unclear whether loss of photosynthesis is associated with, and possibly a consequence of, the absence of IRs in Cryptomonadales. At the same time, significant differences in the retention of photosynthesis-related protein-coding genes were found in three nonphotosynthetic plastid genomes in Cryptomonas. Intraspecies analyses between two C. curvata plastid genomes support the existence of accelerated evolutionally rates of the plastid genomes in Cryptomonas. The extent to which this rate increase is related to the repeated loss of photosynthesis in this genus is unclear. Plastid genome sequence data in Cryptomonadales are currently limited, and the fine-scale processes underlying the loss of photosynthesis in Cryptomonas are still uncertain. More genomic data from close photosynthetic relatives of nonphotosynthetic Cryptomonas are necessary in order to test the hypotheses present herein (e.g., the relationship between loss of IRs and genome rearrangements). Combined with the generation of experimental data, more genome sequences will help to elucidate the underlying causes of the loss of photosynthesis evolution in Cryptomonas.

Data deposition: Newly sequenced data have been deposited at DDBJ under the accessions LC484192 (Cryptomonas curvata CCAP979/52), LC484193 (Cryptomonas sp. CCAC1634B), and LC484194 (Cryptomonas sp. SAG977-2f).