Introduction

Plants, green algae, red algae and glaucophytes (supergroup Archaeplastida) are thought to have evolved through a singular primary endosymbiotic association between a heterotrophic eukaryotic host and a photosynthetic prokaryotic endosymbiont (cyanobacterium) (for reviews see Archibald 2015; Stiller 2014; Zimorski et al. 2014). Euglenophytes, cryptophytes, haptophytes, heterokontophytes, dinophytes, apicomplexans and chlorarachniophytes contain complex plastids enclosed by three or four membranes. Complex plastids are believed to have originated through a secondary endosymbiotic association between a eukaryotic host and a photosynthetic eukaryotic algal endosymbiont (for reviews see Archibald 2015; Stiller 2014; Vesteg et al. 2009a; Zimorski et al. 2014). The plastids of euglenids are thought to have evolved from a secondary endosymbiotic association between an ancestral phagocytic non-photosynthetic euglenozoan host and a photosynthetic eukaryotic green alga (Ahmadinejad et al. 2007; O’Neill et al. 2015a, b; Turmel et al. 2009; Vesteg et al. 2010; Yoshida et al. 2016) closely related to a prasinophyte Pyramimonas parkeae (Hrdá et al. 2012; Turmel et al. 2009). Secondary endosymbiosis in euglenids occurred only relatively recently, approximately 100 MY ago (Parfrey et al. 2011; Stiller 2014). Just as in primary endosymbiosis, the evolution of the euglenid secondary endosymbiont into a plastid involved the transfer of some endosymbiont, chlorophycean algal, nuclear and possibly plastid genes to the host nucleus with the concomitant evolution of a mechanism to return the gene products synthesized in the host cell cytoplasm to the plastid (Sláviková et al. 2005; Vacula et al. 1999; Vesteg et al. 2010). Photosynthetic ability has been lost many times in different eukaryotic lineages, while plastids and plastid genomes have been in many cases retained (Krause 2008; Janouškovec et al. 2015). Within euglenids, E. longa, a close relative of photosynthetic E. gracilis, has retained colorless non-photosynthetic plastids containing a residual plastome whose function is unknown.

The E. gracilis plastid genome is a closed circle encoding 97 currently predicted unique transcribed loci (Bennett and Triemer 2015). The number of polypeptide-encoding genes found in the E. gracilis plastid genome is among the lowest among plastid chromosomes of photosynthetic algae (Rogers et al. 2007). Each gene is present in one copy per genome except for the rRNA genes which are present in three copies. It is estimated that there are 200–1000 plastid genomes per E. gracilis cell (Rawson and Boerma 1976) distributed among ten plastids. The growth of E. gracilis in the presence of specific inhibitors of bacterial DNA (i.e., ofloxacin), RNA and protein synthesis (i.e., streptomycin) generally seems to have no effect on viability when E. gracilis is grown in the presence of various organic carbon sources, but such treatment leads to the permanent loss of the ability to form green colonies, a process termed bleaching (Polónyi et al. 1998; for review see Krajčovič et al. 2002). Bleached cells lack detectable intact plastid genomes, while low levels, 100–1000 fold less than in wild type cells, of possibly rearranged plastid genomic fragments have been detected (Heizmann et al. 1982; Hussein et al. 1982). Different fragments are found in different laboratory produced bleached mutants (Wang et al. 2004). Nuclear genes encoding plastid proteins are retained and transcribed in the bleached cells, but the transcripts are not translated (Schiff et al. 1991; Vesteg et al. 2009b). The loss during bleaching of most if not all of the plastid genome in the absence of nuclear gene loss makes E. gracilis an attractive model to study the reductive evolution of plastids.

Euglena longa is a naturally occurring colorless non-photosynthetic euglenoid. All recent euglenoid phylogenies based on molecular data (Bennett and Triemer 2015; Bicudo and Menezes 2016; Karnkowska et al. 2015; Milanowski et al. 2014; Triemer et al. 2006; Turmel et al. 2009) clearly show that green and secondarily colorless euglenophytes form a monophyletic group. Based on a revised classification of the Euglenophyceae using cytosolic SSU rDNA sequence comparisons and synapomorphic signatures in the SSU rRNA secondary structure (Marin et al. 2003), the former Astasia longa has been renamed E. longa. Since, except for the lack of a stigma and paraflagellar swelling, E. longa was indistinguishable by light microscopy from bleached mutants of E. gracilis, E. longa has until recently been viewed as a naturally occurring bleached E. gracilis (for review see Bodył 1996).

The discovery of a circular 73 kb E. longa plastid genome (Gockel and Hachtel 2000) which is about half the size of the circular 143 kb E. gracilis plastid genome (Hallick et al. 1993) suggested that the colorless E. longa could not simply be a naturally bleached E. gracilis. All genes encoding photosynthesis-related proteins were lost from the E. longa plastid genome except for the rbcL gene encoding the large subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase. The retained E. longa genes are homologous to their E. gracilis counterparts. There are, however, significant differences in gene order between the two genomes. The reduced E. longa plastid genome is transcribed (Gockel et al. 1994; Gockel and Hachtel 2000) suggesting that the plastid genome contains at least one gene required for cell growth and viability. Divergence of E. gracilis and E. longa appears to have involved a selective loss and rearrangement of E. longa plastid genes rather than the expected random loss if E. longa were simply a naturally occurring bleached E. gracilis. Plastid genomes generally contain various repetitive sequences such as microsatellite repeats (George et al. 2015) which can likely mediate recombination events and rearrangements of plastid genomes. The presence of the reduced rearranged plastid genome and several profound differences (ultrastructural, biochemical, physiological) between bleached strains of E. gracilis and the colorless E. longa described decades ago (Blum et al. 1965; Rogers et al. 1972) led to the proposal (Bodył 1996) that it would be impossible to transform E. gracilis into E. longa directly via induced bleaching reopening the question of the mechanism for divergence of E. longa and E. gracilis.

The bleaching process is not well characterized at the molecular level. Streptomycin (SM) and ofloxacin (OF) are highly effective bleaching agents. SM binds specifically and irreversibly to Euglena chloroplast ribosomes (Schwartzbach and Schiff 1974). The fluoroquinolone antibacterial drug, OF, is an inhibitor of DNA gyrase (bacterial topoisomerase type II) and directly inhibits plastid DNA replication (Krajčovič et al. 1989). To understand the possible role of bleaching in the divergence of E. longa and E. gracilis, we have compared the effect of both drugs on cell growth and on the amount of specific plastid genes during several weeks of exposure to bleaching agents. Based on the results, we propose a hypothesis for the origin of the reduced plastid genome of E. longa.

Materials and methods

Strains and cultivation conditions

Axenic cultures of E. gracilis Pringsheim strain Z (“Samlung von Algenkulturen”, Göttingen, Germany) and E. longa strain 1204-17a (a gift from W. Hachtel, Bonn, Germany) were maintained dynamically (50 rpm, Heidolph Unimax 2010, Schwabach, Germany) at 27 °C with continuous illumination (30 µM photons m−2s−1) or in the dark in 50 ml Erlenmeyer flasks containing 30 ml of a modified Cramer and Myers (C–M) medium (Cramer and Myers 1952) supplemented with ethanol (0.8 %) and adjusted to pH 6.9 (Buetow and Padilla 1963). The temperature, 27 °C, ensures a high rate of cell division. Bleaching occurs in cells grown at temperatures above 33 °C. Growth under continuous illumination ensures that cells divide asynchronously. Ethanol is a carbon source supporting rapid growth of E. longa and E. gracilis. The strains of E. gracilis and E. longa used for the experiments are actually those with fully sequenced plastid genomes (Gockel and Hachtel 2000; Hallick et al. 1993). E. gracilis and E. longa cultures in the exponential phase of growth were inoculated axenically at a final density 5 × 103 and 5 × 104 cells per ml, respectively. After the start of drug treatment, cultures were axenically diluted weekly to a final density of 5 × 103/5 × 104 cells per ml with fresh medium containing the sterile antibiotics. OF (Sigma-Aldrich, St Luis, MO, USA) and SM (streptomycin sulfate salt) (Sigma-Aldrich, St Luis, MO, USA) were added to the medium to a final concentration 173 µM. Cultures grown in the absence of antibiotics were used as controls. The growth was monitored by axenically removing triplicate samples every 7 days (weekly) and cell number was determined by direct counting using a Bürker counting chamber. Each experiment was repeated at least three times and a representative experiment is presented. Bleaching was evaluated by determining the percentage of green colonies as described previously (Krajčovič et al. 1989).

DNA extraction

Total DNA was extracted essentially as described in Geimer et al. (2009). Briefly, cells were harvested weekly by centrifugation at 1000×g for 3 min, washed twice with ice-cold deionized water and resuspended in 3 ml buffer (100 mM Tris–HCl, 100 mM EDTA, 250 mM NaCl, pH 8.0) containing 8 µl of Proteinase K (20 mg/ml, Thermo Fisher Scientific, Waltham, MA, USA) per ml of buffer. N-lauroylsarcosine (20 %, 150 µl, Sigma-Aldrich, St Luis, MO, USA) was added and the suspension was lysed at 55 °C for 1 h. Cell debris was removed by centrifugation and DNA was precipitated using 2-propanol. The DNA pellet was resuspended in 1.5 ml of TE buffer, pH 8.0, and RNA was removed by incubation with RNase A (10 mg/ml, Thermo Fisher Scientific, Waltham, MA, USA) for 15 min at 37 °C. The sample was then extracted sequentially with phenol, phenol–chloroform (1:1) and chloroform-isoamylalcohol (24:1), and DNA was precipitated from the aqueous phase at −20 °C by the addition of 1/10 volume of 3 M sodium acetate (pH 5.2) and two volumes of absolute ethanol. After centrifugation at 8000×g for 15 min at 4 °C, the precipitated DNA was washed with 70 % ethanol and resuspended in deionized water (Sigma-Aldrich, St Luis, MO, USA). The concentration of DNA was determined spectrophotometrically (Nano Photometer™, Implen, München, Germany).

Quantitative PCR

Quantitative PCR (qPCR) was used to determine the presence and relative copy number of eight plastid genes: rrn16 and rrn23 encoding plastid 16S and 23S rRNAs, rpl2 and rpl16 encoding ribosomal proteins L2 and L16, rpoC2 encoding RNA polymerase subunit, tufA encoding elongation factor Tu, psbC (present only in E. gracilis) and rbcL, encoding 43 kDa protein of photosystem II and the ribulose-1,5-bisphosphate carboxylase/oxygenase large subunit, respectively. Mitochondrial cox1, and nuclear Act and RbcS genes encoding cytochrome c oxidase subunit I, actin and ribulose-1,5-bisphosphate carboxylase/oxygenase small subunit, respectively, were used as controls. Quantitative PCRs were performed using primers listed in Supplementary Table 1 (purchased from Sigma-Aldrich, St Luis, MO, USA and Microsynth, Balgach, Switzerland).

DNA samples isolated from the first, third and 6th week of E. gracilis drug treatment and all isolated DNA samples from E. longa drug treatments were used in qPCR reactions. DNA isolated from cultures grown in the absence of antibiotics was used as a template in control qPCRs. qPCR reactions were performed using StepOne Realtime PCR system (Applied Biosystems, Life Technologies Corporation, Carlsbad, CA, USA) and 1 × Brilliant II SYBR® Green QPCR Master Mix (Agilent Technologies, Santa Clara, CA, USA) according to manufacturer instructions. The 20 µl qPCR samples contained 0.3 µM primers, and 0.48 ng and 2.4 ng of E. gracilis and E. longa DNA, respectively. The qPCR program included 40 cycles of 95 °C for 15 s, 60/55 °C (E. gracilis/E. longa) for 20 s, and 72 °C for 30 s. The relative copy numbers of plastid, RbcS and cox1 genes were calculated using standard curve method for relative quantification (Larionov et al. 2005). Act gene was used for the data normalization. Six five fold dilutions of untreated control DNA sample were used for calibration. Each qPCR measurement was performed in duplicates or triplicates.

Bioinformatic analysis

ORFs present in E. longa plastid genome sequence (NCBI accesion ID NC_002652.1) were used as queries in NCBI BLAST + 2.3.0, tblastx, blastp and blastx searches (Altschul et al. 1997) to identify homologous sequences in reference genomic and protein sequence databases (refseq_genomic, refseq_protein), the NCBI nucleotide collection (nr/nt) database, the NCBI EST database (est), the protein databank database (pdb) and the swissprot database (swissprot) with e-value cutoff of 10−4. Clustal Omega alignment of homologous ORFs and Duf613 protein domains of unknown function was performed using Clustal Omega 1.2.1 and 1.2.2 (Sievers et al. 2011).

Results

SM, a specific inhibitor of protein synthesis on plastid ribosomes (Schwartzbach and Schiff 1974), and OF, a specific inhibitor of plastid DNA replication (Krajčovič et al. 1989) are among the best characterized bleaching agents in E. gracilis. Only white colonies were obtained when cells were plated 24 h after exposure to each antibacterial drug indicating that under our growth conditions, both inhibitors were effective bleaching agents. During six cycles of cell growth for 7 days followed by dilution into fresh inhibitor-containing media, the cell yield was unaffected by growth in the presence of SM or OF (Fig. 1). Thus, E. gracilis growth and viability over a 6-week period were not affected by inhibitors of plastid protein synthesis or DNA replication even though the cells were bleached and irreversibly lost the ability to form green colonies after only 24 h of inhibitor treatment.

Fig. 1
figure 1

E. gracilis growth in the light in the presence of the bleaching agent (SM, OF) and in the absence of bleaching agents (C). Cultures were inoculated at a final cell density of 5 × 103 per ml every week during the period of 6 weeks. The presence or absence of a bleaching agent appears to have little effect on the final cell density. Similar results were obtained for cultures grown in the dark

Full size image

The effect of SM and OF on E. longa growth was distinctly different from their effect on E. gracilis growth (Fig. 2). SM treated E. longa showed minimal growth as measured by a decrease in cell number during the 1st week and completely stopped growing in the 2nd week of SM treatment what is likely indicative of cell death and/or the inability to divide. OF was a less effective inhibitor of E. longa growth reducing cell yields by approximately 50 % (Fig. 2), but the cells died in the 5th week of growth with OF. The growth inhibitory effect of SM and OF indicates that in contrast to the E. gracilis plastid genome, the E. longa plastid genome is required for at least one metabolic function essential for cell growth and viability.

Fig. 2
figure 2

E. longa growth in the light in the presence of the bleaching agent (SM, OF) and in the absence of bleaching agents (C). Cultures were inoculated at a final cell density 5 × 104 cells per ml every week during the period of three (SM) and 5 weeks (OF). SM at a concentration having no effect on growth of E. gracilis massively inhibited E. longa growth and it completely killed E. longa in the 3rd week of treatment. OF at a concentration having no effect on growth of E. gracilis inhibits growth of E. longa in the first 4 weeks of treatment and it kills cells in the 5th week of treatment. Similar results were obtained for cultures grown in the dark

Full size image

qPCR was used to determine the fate of the plastid genome during the bleaching process. The copy number of all studied E. gracilis chloroplast genes except for rpl16 and rrn16 was reduced to approximately 15–20 % of the untreated control after the 1st week of growth with SM, while the copy number of all genes except rpl16 was reduced to less than 10 % of the control after the 1st week of OF treatment (Fig. 3a, b). After the 3rd week of growth with antibiotics, almost no plastid gene copies (except for rpl16) were detectable by qPCR. The levels of rpl16 were virtually unaffected during 6 weeks of E. gracilis growth with both OF and SM as seen for the levels of the mitochondrial cox1 gene and the nuclear RbcS gene (Fig. 3a, b). The absence of a change during bleaching in the relative amount of the mitochondrial cox1 gene is consistent with the recent study showing that the biochemical properties of respiratory chain complexes of stable bleached mutant W gm ZOflL are virtually identical to those of wild type cells (Krnáčová et al. 2015).

Fig. 3
figure 3

The relative number of plastid, mitochondrial and nuclear gene copies (in %) after SM and OF treatment of light-grown E. gracilis (a, b) and E. longa (c, d). Total DNA extracted from E. gracilis after the 1st, the 3rd and the 6th week of treatment and total DNA extracted from E. longa (c, d) after every week of treatment until the cells died were used for determining the relative copy number of analysed plastid (rrn16, rrn23, rpl2, rpl16, rpoC2, tufA, psbC and rbcL), nuclear (RbcS) and mitochondrial (cox1) genes. psbC is absent from the E. longa plastome. The relative copy numbers were calculated using the standard curve method for relative quantification. The nuclear Act gene was used for data normalization. The black columns in a, b, c and d represent untreated sample (100 %), x-axis—genes, y-axis—gene copy number in  % of control sample. a, b Green (dotted) columns (SM1, OF1)—% of copy number after the 1st week of E. gracilis treatment; yellow (striped) columns (SM3, OF3)—% of copy number after the 3rd week of E. gracilis treatment; orange (bricked) columns (SM6, OF6)—% of copy number after the 6th week of E. gracilis treatment. c Green (dotted) column (SM1)—% of copy number after the 1st week of E. longa SM treatment; yellow (striped) column (SM2)—% of copy number after the 2nd week of E. longa SM treatment. d Green (dotted) column (OF1)—% of copy number after the 1st week of OF treatment; yellow (striped) column (OF2)—% of copy number after the 2nd week of E. longa OF treatment; orange (bricked) column (OF3)—% of copy number after the 3rd week of E. longa OF treatment; red (checkered) column (OF4)—% of copy number after the 4th week of E. longa OF treatment. Similar results were obtained for cultures grown in the dark

Full size image

The 73 kb E. longa plastid genome is transcribed (Gockel and Hachtel 2000; Gockel et al. 1994). The sequence similarity of the retained E. longa plastid genes to their E. gracilis counterparts is high, but their order on the E. longa plastid chromosome is different from their order on the E. gracilis plastid chromosome. The levels of representative plastid genes encompassing all regions of the circular E. longa plastid genome during exposure to SM and OF were determined by qPCR (Fig. 3c, d). As seen for E. gracilis, the levels of the mitochondrial cox1 gene and the nuclear RbcS gene were unaffected after 2 and 4 weeks of growth in the presence of SM and OF, respectively (Fig. 3c, d). On the other hand, the gene copy number of all plastid genes (including rpl16 gene) systematically decreased in treated cells depending on the length of treatment. SM and OF treated E. longa cells were killed in the third and the 5th week of treatment, respectively, suggesting that loss of plastid genes is lethal for E. longa and that intact plastid genome is necessary for E. longa survival.

To identify the E. longa plastid gene(s) required for cell growth and viability, we have attempted to search for potential homologs of all unassigned plastid ORFs in currently available public databases. Except for orf67 and orf519, all other E. longa ORFs are absent from the E. gracilis plastome. The BLAST searches for ORFs 57, 70, 76, 105, 122, 125a, 125b, 160, 162, 167, 211, 253, 263, 288 and 558 revealed no significant homology with any currently available sequences from other species. However, the ORFs 162, 167, 211, 253, 263, 288 and 558 share various degrees of amino acid sequence homology. ORFs 253, 263 and 288 possess two Duf613 domains, while ORFs 162, 167, 211 and 558 possess one Duf613 domain (Fig. 4, Supplementary Fig. 1). Duf613 is a protein domain of unknown function (PF04764) described in Pfam (Finn et al. 2016). The E. longa plastid genome is currently the only plastid genome where genes encoding putative proteins containing the Duf613 domain have been identified. The Clustal Omega alignment of the proteins encoded by orf104 and orf426 (Supplementary Fig. 2) indicates that orf104 is a shorter version of orf 426. The BLAST search for orf426 revealed many significant homologs in bacteria and animals which were mainly, however, only predicted hypothetical proteins of unknown function. The best BLAST hits included also APOBEC1 complementation factor and coagulation factor V. orf426 was only homologous to relatively short parts of these proteins and it would be difficult to envision a function for these proteins in E. longa plastids. Unfortunately, BLAST searches could not identify the function of the proteins encoded by the unassigned ORFs in E. longa plastids.

Fig. 4
figure 4

The ClustalO alignment of Duf613 domains present in ORFs encoded by E. longa plastid genome. ORFs 253, 263 and 288 contain two Duf613 domains (a, b), while ORFs 162, 167, 211, 558 contain one Duf613 domain. The positions of terminal aminoacids of Duf613 domains within ORFs are in parentheses

Full size image

Discussion

E. gracilis and E. longa responded very differently to antibacterial agents. SM and OF treatment caused bleaching of E. gracilis cells without affecting cell growth or viability. In contrast, E. longa growth and viability was inhibited by both SM and OF. The copy number of seven studied plastid genes, rrn16, rrn23, rpl2, tufA, psbC, rbcL and rpoC2 decreased when E. gracilis was grown in the presence of SM or OF, while the copy number of one gene, rpl16, remained unchanged even after 6 weeks of growth. The reduction of plastid gene copy number was slightly slower in cells grown with SM requiring 3 weeks of growth to eliminate rrn16, rrn23, rpl2, tufA, psbC, rbcL and rpoC2, while these genes were barely detectable after 1 week of growth with OF. Surprisingly, the level of rpl16 gene was unaffected even after 6 weeks of E. gracilis growth with SM or OF. In contrast to E. gracilis, the levels of all studied plastid genes including rpl16 decreased in a time dependent manner in E. longa grown in the presence of SM and OF. E. longa was killed after the third and the 5th week of growth with SM and OF, respectively, indicating that its growth is dependent upon a metabolic function of the plastid.

The retention of rpl16 gene in E. gracilis after the growth with SM and OF is intriguing. Since rpl16 was undetectable by reverse transcription PCR when oligo(dT) primed cDNA was used as a template, it is unlikely that a functional copy of this plastid gene is present in the nucleus (Záhonová et al. 2014). Moreover, the bleaching of E. gracilis strain FACHB47 by OF and SM resulted in rpl16 loss in all and all but one mutant, respectively (Wang et al. 2004). rpl16 gene was not detected in E. gracilis W 10 BSmL and W 3 BUL stable bleached mutant strains of E. gracilis strain bacillaris (data not shown) suggesting that it is exclusively present in plastids in this strain. In contrast, rpl16 gene is present in W gm ZOflL mutant derived from E. gracilis strain Z (data not shown) suggesting that the retention/loss of rpl16 gene after bleaching might be strain specific. In contrast to our experiments, the mutants produced by Wang et al. (2004) retained 16S rRNA gene instead of rpl16 gene. Wang et al. (2004) suggested that this gene was retained due to its position on the plastome close to the origin of replication. Since bleaching results in loss of most plastid genes, this would suggest that plastid genome replication depends exclusively on nucleus-encoded proteins. The retention of rpl16 by W gm ZOflL mutant derived from strain Z and in our bleaching experiments with strain Z suggests that plastid genome of strain Z might possess an alternative origin of replication located within or close to the rpl16 gene, which might be responsible for the retention of a rudimental nonfunctional plastid genome. In addition, W gm ZOflL mutant has also retained rpl14 and rpl5 which are present downstream from rpl16 gene, while it does not possess 16S rRNA gene (data not shown). Much more detailed molecular analyses after bleaching of various E. gracilis strains as well as of stable bleached mutants would be needed to explain the phenomenon of differential retention of plastid genes (or eventually of their nuclear copies) by different strains. It would be also of interest to find out, if the retained reduced genomes are circular or linear with various isoforms like in some plants (Oldenburg and Bendich 2016). Since it was not possible to circularize the plastome of E. gracilis bacillaris due to the presence of repetitive sequences in the region where the putative origin of replication is localized in strain Z (Bennet and Triemer 2015), it is possible that at least some euglenid circular-mapping plastomes are virtually linear and that they are replicated by a virus-like recombination-dependent mechanism similar to that one proposed for plants (Oldenburg and Bendich 2016). However, experimental evidence would be needed to verify this hypothesis.

E. longa but not E. gracilis growth and viability were inhibited by both SM and OF indicating that E. longa but not E. gracilis survival is dependent upon the functional plastid. The Euglena plastid arose through an endosymbiotic association between an ancestral heterotrophic euglenozoan host and a prasinophyte alga (Hrdá et al. 2012; Turmel et al. 2009). Duplicate essential metabolic pathways were encoded by the ancestral host and endosymbiont genomes. The loss of essential cytoplasmic metabolic pathway(s) also occurring in the plastid was likely the pivotal event that triggered the divergence of E. longa, an obligate heterotroph with a reduced plastid whose growth is dependent upon a plastid metabolic activity, from E. gracilis, an organism with a photosynthetically competent chloroplast whose growth and viability is independent of a functional plastid. Whether the reduced E. longa plastid genome encodes a protein directly required for cell growth or whether the chloroplast is simply an essential intracellular compartment where nucleus-encoded proteins required for heterotrophic growth are localized remains uncertain. In the former case, a strong selection pressure would exist for retention of at least a portion of the plastid genome devoid of photosynthetic genes but containing the genes required for plastid replication, gene expression and any metabolic activity required for cell growth and viability. In the latter case, a selection pressure would only exist if the plastid genome was required for the maintenance of a membrane bounded metabolic compartment containing nuclear gene products.

The identity of the E. longa plastid-encoded protein required for cell growth remains elusive. Retention of the plastid-encoded ribulose-1, 5-bisphosphate carboxylase/oxygenase large subunit is highly suggestive that it might have an essential function in E. longa but not in E. gracilis. Rubisco in the absence of the Calvin cycle is involved in oil synthesis in developing plant embryos (Schwender et al. 2004). The Rubisco small subunit is encoded in E. longa nucleus, the large precursor is synthesized, but it does not appear to be processed into mature subunits or even transported to the plastid (Záhonová et al. 2016). The mRNA and protein levels of RbcL are much lower in E. longa than in E. gracilis (Záhonová et al. 2016). Thus the function (if any) of Rubisco in E. longa remains unknown (Záhonová et al. 2016). The most studied non-photosynthetic plastid, the apicoplast of apicomplexan parasites, harbours at least four essential metabolic activities—heme biosynthesis, fatty acids biosynthesis, biosynthesis of isoprenoids and Fe–S cluster assembly (Charan et al. 2014; Ralph et al. 2004). The mentioned metabolic pathways are also among candidates for deciphering the possible function of E. longa plastid. For example, E. gracilis synthesizes heme in mitochondria utilizing nucleus-encoded enzymes and in the plastid by a completely different pathway utilizing nucleus-encoded enzymes and plastid-encoded tRNAglu (Kořený and Oborník 2011). The E. longa plastid genome encodes the tRNAglu required for heme synthesis. If E. longa lost one of the nucleus-encoded genes required for mitochondrial heme biosynthesis, the plastid heme biosynthesis would be an E. longa plastid function required for cell growth and viability. Nevertheless, experimental evidence would be needed to infer the exact function(s) of E. longa plastids.

To identify potential E. longa essential plastid-encoded genes, we attempted to search for homologs of the proteins encoded by the E. longa plastid ORFs. BLAST searches failed to identify homologs of most unassigned E. longa plastid ORFs, although seven ORFs share one or two Duf613 domains of unknown function (Fig. 4). The study of interaction of the Duf613 domain with other plastid proteins might be the key for understanding the function (if any) of the proteins encoded by these ORFs. Many homologs of orf426 were identified, but most of them were hypothetical proteins. It would be difficult to explain the function of the best annotated hits such as coagulation factor V in E. longa plastids. orf122, orf153, orf211 and orf167 or orf125a, orf288 and orf160 are next to each other on the plastid genome suggesting that these ORFs encode portions of a single protein and the non-coding regions between the ORFs are introns. This was actually the case for the rpoA gene which was initially mis-annotated as orf67 and ycf67 in E. longa (Sheveleva et al. 2002). The function (if any) of the proteins encoded by the ORFs remains unknown.

Mixotrophy has been suggested as a driving force for the loss of photosynthesis in various algal lineages (Figueroa-Martinez et al. 2015) and it was a likely precondition for the reduction of the ancestral E. gracilis plastid genome during evolution of E. longa into an obligate heterotroph. The transition to heterotrophy would have been favored by growth in an organic carbon rich environment, and especially in the dark over long time periods. A paucity of carbon sources and natural light regime would rather favor the retention of photosynthesis. As pointed out by Bodył (1996), the divergent evolution of E. longa and E. gracilis was not confined to changes in the plastid genome but rather produced a number of biochemical, physiological and ultrastructural changes. The magnitude of the differences between E. longa and E. gracilis and the selective loss of photosynthetic but not other plastid genes would be expected to require slow divergence over millennia. The large time required for E. longa to diverge from E. gracilis led Bodył (1996) to conclude that divergence of E. longa from E. gracilis could not be the result of bleaching but rather resulted from gradual excision of small plastid genome fragments. qPCR measurements during bleaching show a relatively constant copy number of nuclear genes and a decreasing copy number of plastid genes. All plastid genes do not appear to be lost simultaneously as seen for the E. gracilis rpl16 gene whose copy number is unaffected during bleaching. Although bleaching clearly results in the loss of plastid genes, the rapidity of the process, a week for major genome losses, and the extent of the process, almost complete genome loss, appears to exclude highly concentrated bleaching agents as the causative factor in the reduction of the plastid genome during divergence of E. longa and E. gracilis triggered by a nuclear gene mutation making cell growth and viability of E. longa dependent upon a plastid metabolic function.

Bleaching agents are antibacterial compounds produced by various microorganisms which are ubiquitous in the environment, or physical factors such as elevated temperatures and UV radiation. It is, however, unrealistic to assume that natural populations of the common ancestor of E. longa and E. gracilis were continually exposed to bleaching agents for extended periods of time as done in this study, that they were exposed to the high concentrations of bleaching agents that completely inhibit plastid replication, transcription or translation as done in this study and that the bleaching agents present in the environment bind to their targets irreversibly as found for SM. It is more realistic to assume that natural populations of the ancestral E. longa were intermittently exposed to bleaching agents that do not bind their target irreversibly at subsaturating concentrations for short time periods.

We propose that a process termed “intermittent bleaching”, a short time exposure to subsaturating concentrations of reversible bleaching agents followed by growth in the absence of the bleaching agent, over a millennial time period provides a molecular explanation for the reduction of the plastid genome occurring during divergence of E. longa and E. gracilis. Prior to exposure to the bleaching agent, the individual plastids within a cell and the cells within the population are homoplasmic for plastid DNA having multiple copies of a circular genome. Intermittent short term exposure to a bleaching agent such as SM or OF would result in the loss of some plastid genes while other genes would be retained. Individual plastids within a cell and the cells within the population would become heteroplasmic having some genomes with a full complement of genes and a heterogenous assortment of other genomes having a copy of some genes and lacking other genes. Multiple inversions and translocations could potentially arise through recombination between the dissimilar genomes within a heteroplasmic plastid or cell. As the cells continue to grow in the absence of the bleaching agent, the altered plastid genomes replicate, segregate and the plastids within a cell and the cells within a population return to a homoplasmic state having genomes that have retained some genes and that have lost other genes.

After divergence of E. longa and E. gracilis triggered by transfer of an essential metabolic activity to the plastid, the dependence of cell growth and viability on a metabolically functional plastid would provide a selection force for reduced plastid genomes produced by “intermittent bleaching” which have retained non-photosynthetic plastid genes such as those required for plastid transcription and translation even if they have lost photosynthetic genes which are not needed for heterotrophic growth in the environment rich in organic carbon sources. Cells becoming homoplasmic for a genome with a deletion for an essential metabolic function would not be viable and this genome would be lost from the population while cells containing genomes with deletions of photosynthetic genes would be viable and the deleted genome would be maintained in the population. Over a long time period, successive rounds of “intermittent bleaching”, production of cells heteroplasmic for deleted plastid genomes and segregation producing cells homoplasmic for plastid genomes containing deletions, would lead to the loss of all plastid genes not required for cell growth and viability, the photosynthetic genes, producing the reduced plastid genome of E. longa.

E. longa is not the only known organism with a reduced plastid genome. The apicomplexans such as Plasmodium sp. and Toxoplasma sp. which share a common photosynthetic ancestor with photosynthetic chromerids such as Chromera sp. and Vitrella sp. (Janouškovec et al. 2010), parasites such as Helicosporidium and Prototheca that evolved from green algae and a holoparasitic plant Epiphagus genus (reviewed in Krause 2008; Vesteg et al. 2009a) are among the organisms known to contain reduced plastid genome. A parasite is an obligate heterotroph whose lifestyle precludes photosynthetic growth. Just as “intermittent bleaching” provides a molecular mechanism for reduction of the plastid genome of the obligate heterotroph E. longa after divergence from E. gracilis, it provides an explanation for the evolution of the reduced plastid genomes found in parasites with diverse evolutionary origins. If divergence of the parasitic and photosynthetic ancestors occurred after a nucleus-encoded cytoplasmic metabolic pathway was lost so that the plastid genome was required for that function, “intermittent bleaching” would produce plastid gene deletions. The requirement of some plastid genes for cell viability would select over millennia for obligate heterotrophs with reduced plastid genomes lacking photosynthetic genes.

The eukaryotes are currently divided into five supergroups—Opisthokonta, Amoebozoa, Excavata, Archaeplastida and SAR (Stramenopiles, Alveolata and Rhizaria), although the phylogenetic position of various groups such as haptophytes and cryptophytes is currently uncertain (Adl et al. 2012; Burki et al. 2012, 2016). Most of the major eukaryotic lineages unite heterotrophic and photosynthetic (or at least photobiont-containing) organisms. The number of endosymbiotic events occurring in the evolution of eukaryotes, the number of potential complete losses of plastids and whether plastid loss is a rare event have been a matter of extensive debate (Bodyɫ et al. 2009; Burki et al. 2008, 2012, 2016; Kim and Graham 2008). Parasitic plant Rafflesia lagascae (Molina et al. 2014) and the green algae of the Polytomella genus (Smith and Lee 2014) have retained a plastid compartment, but they seem to have completely lost the plastid genome. Non-photosynthetic colpodellids such as Colpodella angusta, Alphamonas edax and Voromonas pontica, which are more closely related to photosynthetic chromerids than apicomplexan parasites, also possess plastids but they apparently lack a plastid genome (Janouškovec et al. 2015). The basal branching apicomplexans, Cryptosporidium parvum (Abrahamsen et al. 2004; Zhu et al. 2000), Gregarine niphandrodesmay (Toso and Omoto 2007) and Ascogregarina taiwanensis (Templeton et al. 2010) lack a detectable plastid genome and likely plastid itself. In addition, it seems that eukaryotes can even lose any identifiable trace of mitochondria as demonstrated by the example of Monocercomonoides sp. (Karnkowska et al. 2016). The spontaneous rate of E. gracilis chloroplast loss is 1–2 % (Diamond and Schiff 1974). E. gracilis exposed to a bleaching agent loses most plastid genes within 2–6 weeks. An identifiable structure and plastid remnant are not seen in the experimentally produced stable SM bleached mutant W 10 BSmL (Osafune and Schiff 1983) demonstrating that bleaching can produce cells where traces of a plastid genome and identifiable plastid remnant are undetectable. Taken together, the spontaneous rate of E. gracilis chloroplast loss, the rapidity of E. gracilis plastid gene loss during bleaching and the production by SM bleaching of the mutant W 10 BSmL lacking traces of a plastid genome and physical plastid remnant similar to the situation in basal apicomplexans provides an experimental proof that loss of the plastid genome and an identifiable plastid structure is not a rare event. Complete loss of an ancestral plastid is a mechanistically simple event that could have happened during the evolution of multiple lineages such as in the ciliates related to photosynthetic dinoflagellates and chromerids, and in the oomycetes related to ochrophytes (photosynthetic stramenopiles). E. gracilis is to our knowledge one of the few plastid-containing if not the only organism where the plastid is not required for cell growth and viability suggesting that the rare event is evolution of a plastid that does not perform a metabolic function required for cell growth and viability, a prerequisite for facile plastid loss. Thus, a true understanding of the contribution of photosynthesis to eukaryotic evolution must await resolution of valid concerns (see for example Archibald 2008, 2015) regarding phylogenomic analysis, since plastid genome reduction, plastid genome loss and loss of an identifiable plastid structure have been experimentally shown to be processes that could easily have occurred by bleaching multiple times.