Diverse origins of chloroplast enzymes
The present study provided two different criteria for assessing the origins of chloroplast enzymes involved in membrane lipid synthesis, photosynthesis, gene expression, among others. The first criterion is phylogenetic branching, which has also been used to show chloroplast enzymes of cyanobacterial origin before. However, many proteins involved in photosynthesis are conserved in only cyanobacteria and chloroplasts, and it has been theoretically difficult to determine exactly whether the chloroplast enzymes originate from cyanobacteria, or chloroplast and cyanobacterial enzymes are sister groups (type 1c). The combined tree of PsaA and PsaB showed unambiguously that a Gloeobacter homolog is at the base of each lineage. Based on this result, we can safely, but tentatively, set the root with Gloeobacter as the first branching species for type 1c trees. Figure 2 shows four major types of phylogenetic relationship of chloroplast proteins with cyanobacterial homologs. As described in “Results”, some enzymes such as RpoB, RpoC, and Ndh proteins showed inconsistent results depending on taxon selection and phylogenetic methods, but most proteins analyzed in the present study showed consistent branching pattern with both ML and BI methods. In this respect, the classification of enzyme origins in Fig. 2 is reliable, and provides the basis of further consideration on evolutionary history. Type 1 phylogeny could be further classified according to the origin within the cyanobacterial clade. Many type 1 enzymes had a deep origin in the cyanobacterial clade after the Yellowstone strains, but, as described above, PsaA, PsaB, and MinD originated from α-cyanobacteria, whereas RbcL[G], PRK, RpoA, and ChlBLN originated from β-cyanobacteria. This suggests that even the enzymes of cyanobacterial origin could have different origins within cyanobacteria (see below).
Another criterion is relative stem length as defined in Fig. 4. I used a method previously applied to mitochondrial enzymes (Pittis and Gabaldón 2016), but stem length was estimated for each of the green and red lineages. This is justified by two reasons. First, to estimate the stem length of Archaeplastida as a whole, we need sequences in all three lineages, namely, the green, red, and glaucophyte lineages. In many phylogenetic trees in which a glaucophyte sequence is available, glaucophyte diverges first. In this respect, the estimate of the stem length depends on the presence of glaucophyte sequence. In addition, Cyanophora paradoxa was the only species of glaucophyte whose genomic data were available until recently. The estimation of stem length of all Archaeplastida is not always possible with the original Cyanophora genome data published in 2012 from Rutgars University. A new data published recently (Price et al. 2019) might be useful in future studies.
The second reason to estimate stem length for the green and red lineages was the fact that not all enzymes are perfectly conserved in both green and red lineages. As stated above, Ndh proteins, enzymes involved in peptidoglycan synthesis, and chlorophyll biosynthesis enzymes are present in the green lineage, but not in the red lineage. In contrast, many proteins in the dataset B37, B42, and F35 are conserved in the red lineage but not in the green lineage. I therefore prefer to estimate the stem length for each of the green and red lineages in all enzymes or datasets. Nevertheless, the high correlation of the two values in Fig. 5b indicates that the analysis was successful in general. Extreme values were found for some small proteins, but we can use the distribution of data points in further discussion.
These two criteria, namely, branching pattern and relative stem length, suggested that chloroplast proteins, whether they are encoded by the chloroplast or nuclear genomes, are phylogenetically highly diverse.
Diversity in gene acquisition times
I have to admit that there are possible pitfalls in the estimation of relative stem length. First, the estimate could be liable to tree shape. If tree shape is not unambiguously fixed, it is difficult to determine stem lengths. In most of my analyses, the branching pattern of the chloroplast clades was supported by both ML and BI methods as described above. I made many more trees that are not presented in Online Resources during the process of the present study, and confirmed the validity of the presented trees unless otherwise stated. Curiously, however, different trees of RpoB and RpoC gave similar values of VS/VL, RS/RL, and CL/VL. Second, selection of taxa can affect estimation of relative stem length, especially in determining leaf lengths, VL and RL. As described above, these values were determined as a median of branch lengths in the respective subtree. In the case of green lineage, I was able to use enough taxa to obtain a stable value of median, namely, removing or adding a taxon does not significantly affect the value of VL. In the case of red lineage, however, the number of available taxa was limited to five or six, depending on enzymes. Addition or removal of a taxon can affect RL value significantly in some enzymes. This could result in lower or higher estimate of RS/RL values as illustrated in Fig. 5b, such as AccB, F35, and B42. Third, relative stem length might not reflect actual acquisition time in the evolutionary history. If evolutionary rate was not identical for the time of diversification (VL or RL) and for the time from the acquisition to the common ancestor (VS or RS), the ratio VS/VL or RS/RL does not reflect the ratio of the two times.
I have two ideas to overcome this problem of evolutionary rate. First, the length of cyanobacterial sister group could provide a measure of changes in evolutionary rate. According to this criterion, the small values of VS/VL and RS/RL of RNA polymerase subunits were interpreted as a result of higher rate of evolution in the chloroplast branch. Second, we can assume that the selection pressure of functionally related proteins is similar. This might be true for the groups of enzymes involved in fatty acid synthesis, lipid biosynthesis, photosynthesis, protein synthesis, transcription, and carbon fixation. In this respect, highly diversified values of relative stem lengths obtained with functionally related proteins in each functional group could represent real differences in acquisition time of the enzymes.
Acquisition of chloroplast rRNA and house-keeping proteins (including ribosomal components and ATPase subunits, and possibly RNA polymerase subunits) (Table 1) may be taken as a canonical time of acquisition of chloroplast genes. The fact that these house-keeping proteins form discrete gene clusters within the chloroplast genome (Stoebe and Kowallik 1999) suggests that the clusters were introduced into the chloroplast in formation as a single unit. In my data, it is not clear whether RNA polymerase was acquired with ribosome. The argument on evolutionary rate as described above might not be sufficient for proving that RNA polymerase was acquired with ribosome, because the phylogenetic origin of RpoB and RpoC is not clearly resolved (either type 1 or type 2). Another concern is that, even in type 1 trees, the exact origins of RpoA, RpoB, and RpoC within the cyanobacterial clade were diverse. Further work is needed to solve the origin of RNA polymerase.
Potentially early acquisition of glycolipid synthesis enzymes
Acquisition of lipid biosynthetic ability by the chloroplast is divergent in origin and time. Origins of enzymes involved in glycolipid and phosphatidylglycerol synthesis are quite diverse, as shown by type 3 and type 4 phylogenies. The enzymes were potentially acquired earlier than the acquisition of chloroplast ribosome as shown by longer relative stem lengths. These include fatty acid synthase, KAS I/II, malonyltransferase, FabD, the second acyltransferase, ATS2, eukaryotic MGDG synthase, MGD1, and palmitoyl desaturase of phosphatidylglycerol, FAD4. Some type 2 enzymes also presented longer relative stem lengths. These include dehydratase FabZ, the two enzymes in SQDG synthesis, SQD1/SqdB and SQD2/SqdX. Type 3 enzymes also include the first acyltransferase, ATS1, and the second galactosylation enzyme, DGD1. Acquisition of all these components of lipid biosynthetic machinery by chloroplast must be different from the origin of chloroplast ribosomes, in both time and source. According to the larger values of relative stem length, it is quite possible that many of the lipid biosynthetic enzymes had been acquired before the acquisition of ribosomes. By contrast, DgdA, a chloroplast-encoded galactosyltransferase to produce DGDG in Cyanidiales red algae, originate from cyanobacteria (type 1) and presented a comparable value of RS/RL with rRNA, suggesting that it is acquired from cyanobacteria with ribosomal components. It is unlikely that DGD1 was acquired independently in the other red algae and the green lineage. Therefore, DGD1 in the red algae other than Cyanidiales must be acquired before the separation of the red and green lineages, suggesting that DGD1 and DgdA co-existed in early red algae, and possibly in early Archaeplastida.
Early acquisition of SQDG and MGDG synthesis activities could provide advantages for a proto-algal cell, because phosphorus is a limiting element in aquatic environments. Phosphorus is an important element for all organisms, but its abundance is limited in the Earth crust, and much more limited in the seawater. Photosynthetic membranes of plants and algae are made of glycolipids and the content of phosphatidylglycerol is limited to about 10%. If the membranes should be made entirely of phospholipids, phosphorus requirement of plants and algae would be much higher. In many green algae, such as Chlamydomonas reinhardtii, phosphatidylcholine (PC) is replaced by diacylglyceryltrimethylhomoserine (DGTS) (Sato and Furuya 1985). Phosphate limitation is known to reduce phosphatidylcholine content and to increase DGTS content in some fungi (Senik et al. 2015). The use of glycolipids instead of phospholipids must be an adaptation of photosynthetic organisms to phosphorus-limited environments. The same argument must also be valid for early microorganisms living in phosphorus-limited water. The acquisition of SQDG and MGDG synthesis activities (possibly, DGDG synthesis, as well) could have provided selective advantage in early microorganisms, even before the acquisition of photosynthetic activity. As an interesting supporting evidence, phosphate limitation in Arabidopsis thaliana causes accumulation of DGDG (normally a chloroplast lipid synthesized in the envelope membranes) to the extraplastidic compartment to replace phospholipids (Härtel et al. 2000). This could be reminiscent of the ancient situation in the host cell before the chloroplast formation.
Acquisition of photosynthetic machinery
The origin(s) of photosynthetic machinery might be more complicated than through a single endosymbiotic event. Most of the photosynthesis-related components analyzed in the present study originated from cyanobacteria (type 1). Nevertheless, as described above, the exact origins of individual components are not identical. As already known, most chloroplast-encoded proteins have a deep origin within the cyanobacterial clade: namely, chloroplast proteins diverged early after the diversification of some basal groups (Yellowstone strains of Synechococcus, and Pseudanabaena) as shown in Online Resource 4, although the exact branch point is different in different studies (Ponce-Toledo et al. 2017; Shih et al. 2013). However, PsaA and PsaB are sister to α-cyanobacteria (Prochlorococcus, Synechococcus), whereas ChlBLN, RbcL [green lineage], RPE, and PRK originate from β-cyanobacteria (Nostoc, Cyanothece, Synechocystis, etc.). In addition, relative stem lengths of RbcL, PsaD, and the set A48 (23 photosynthetic proteins encoded by the chloroplast genome) are larger than that of rRNA. These results suggest that photosynthesis-related proteins are diverse in origins (Fig. 6). Analysis of individual proteins in the set A48 will shed more light on the origin of photosynthetic machinery, although this is not the main focus of the present study.
Possible explanations
The finding of the present study that the origin of chloroplast enzymes, especially those involved in the synthesis of chloroplast membranes, is quite diverse in both phylogeny and time of acquisition is not consistent with the currently accepted endosymbiotic theory on the origin of chloroplasts, involving a single event of endosymbiosis. Several ideas can resolve this conflict. First, an easy solution is to deny the validity of the phylogenetic analysis, especially, the estimation of acquisition times of enzymes. Second, the two hypotheses can be merged, namely, accepting both single event of endosymbiosis and many gene transfers to chloroplasts in addition. The first attitude is not justified, because the endosymbiotic theory has never been rigorously tested by phylogenetic analysis with massive genomic data, which became available after the acceptance of the endosymbiotic theory in the late 20th century (see next section). In the second possibility, people are willing to admit that many gene transfers occurred after the establishment of chloroplasts. The present study revealed, however, that many enzymes related to photosynthesis and membrane lipid synthesis could have been acquired before the acquisition of chloroplast ribosome. We can imagine another scenario, in which an initial chloroplast was acquired by endosymbiosis with the enzymes of photosynthesis and lipid biosynthesis. Then, the ribosome was replaced by a new one from another endosymbiont. This could be a good idea. But then, we have to ask what were endosymbioses and what were gene transfers. Bacterial contribution to the plastid proteome has been repeatedly reported (Qiu et al. 2013), and an entirely different scenario was presented by Martin’s group (Ku et al. 2015), in which various bacterial enzymes had been transferred to a special lineage of cyanobacteria, which then became an endosymbiont in the primary endosymbiosis that engendered chloroplasts. A problem in this hypothesis is that such ancestral endosymbiont must have possessed two redundant pathways of glycolipid synthesis (MDG1–DGD1 for chloroplasts, and MgdA–MgdE–DgdA for cyanobacteria). I consider more plausible to suppose that the chloroplast lipid biosynthesis system was established before the acquisition of chloroplast ribosomes, because glycolipids must have provided adaptive advantage even before the formation of chloroplasts.
Historical considerations on the origin of endosymbiotic theory
It is helpful to review the origin of endosymbiotic theory itself. Historically, the endosymbiotic hypothesis was initially conceived by Mereschkowsky (1905), but had no impact on biology in the first half of the 20th century. The endosymbiotic theory was revived in the 1960s and 1970s by various scientists as a possibility to explain a phylogenetic inconsistency, namely, the similarity of cyanobacteria (called blue-green algae, then) and chloroplasts, and the similarity in cell mechanisms of eukaryotic organisms, including plants, algae, protists, and animals (for reviews of the time, see Stanier 1970, 1974). See my previous publication (Sato 2017) for historical overview on the endosymbiotic theory with a re-evaluation of the work of Lynn Margulis. At that time, gene transfer was known only in bacteria and viruses. No one considered a possibility of gene transfer from a prokaryote to a eukaryote. Various serious doubts were cast on the endosymbiotic theory until the mid 1970s, but an early phylogenetic analysis using rRNA and some chloroplast genes provided qualitative evidence for the cyanobacterial origin of chloroplasts (Gray and Doolittle 1982; Schwartz and Dayhoff 1978). Once the endosymbiotic origin was accepted, various discussions and doubts about it were forgotten, and a visual image of endosymbiosis became a standalone model that does not require further rigorous verification. People no longer paid attention to the concept that membranes are not inherited by themselves. Various visual explanations of the endosymbiotic origin of primary and secondary chloroplasts were presented in top journals in the 1980s and 1990s (see for example, Cavalier-Smith 2000), but no one doubted the reality of the images, forgetting to explain the origin of chloroplast membranes in reasonable ways.
From the early days, we have been showing that cyanobacteria and chloroplasts have different pathways of glycolipid synthesis (Sato and Murata 1982). Later identification of glycolipid synthesis genes in plants and cyanobacteria clearly established fundamental differences in the two systems (for a short review, see Sato and Awai 2016). In this respect, the chloroplast membranes are not heritages of cyanobacteria. Some people might still try to consider that enzymes in membrane lipid synthesis must be replaced after the primary endosymbiosis. The present study clearly showed that the acquisition times of the enzymes were diverse, and some enzymes could have been acquired before the acquisition of chloroplast ribosomes. Certainly, it is no longer possible to keep the simple endosymbiotic explanation. Rather, we have to assume many repeated endosymbiotic events with cyanobacteria, chlamydiae, green bacteria, and others, if we still use endosymbiotic explanation. There are discussions about the role of chlamydiae, working with cyanobacteria to establish chloroplasts (Cenci et al. 2017; Domman et al. 2015), but the problem might not be limited to chlamydiae. Do we need to keep the endosymbiotic explanation? We only need many gene transfers. Some transfers might be triggered by endosymbiosis, but as stated above, the endosymbiont membranes were never inherited by themselves. We cannot keep the visual image of endosymbiosis any more, because chloroplast membranes are quite different from cyanobacterial membranes.
Philosophical considerations on the endosymbiotic theory
Many readers might question whether I try to deny the endosymbiotic explanation as a whole or many gene transfers that I suppose above could still be explained within the framework of the endosymbiotic theory. It is pertinent to present a logical or philosophical analysis of the theory to answer the question or doubt.
The current general belief of the endosymbiotic notion comes from the following inference:
First premise: many chloroplast genes (or most parts of the chloroplast genome) originate from cyanobacteria.
Second premise: an organelle as a whole has a high evolutionary fitness, whereas its components do not.
Conclusion: chloroplast must be acquired as a whole from a cyanobacterial ancestor.
This is a very strong type of inference, which is hardly falsified by any experimental evidence. The notion “endosymbiotic origin” is a hard-core theory, which is supported by this type of very strong inference, and this is the point that I try to criticize. Can we relax this inference to present a more reasonable argument from the scientific viewpoint? As I argued, acquisition of glycolipids could have been beneficial for the ancient host cell (future algal cell) in phosphate-limited environments. Photosynthetic genes and gene expression machinery could be acquired independently, because they have different functions. We can relax the second premise by replacing it with functionality of smaller units. Some metabolic enzymes could be acquired alone. We might be able to hypothesize a neutral evolution for some components. In other words, we should not be too much constrained by the benefit that a component affords to the organelle. The “shopping bag model” could be an interesting hypothesis in this regard (Howe et al. 2008). Exaptation or pre-adaptation could explain acquisition of some enzymes of an entire pathway such as peptidoglycan synthesis. For example, the fact that the Arabidopsis MurE acts as a regulator in chloroplast development (Lin et al. 2017) could suggest that the initial acquisition of this enzyme could be uncoupled with other components of peptidoglycan synthesis. We might be able to relax the hard-core theory of endosymbiotic origin of chloroplasts by accumulation of various efforts like this. At this moment, I am not sure if the notion “endosymbiosis” is necessary to explain the origin of chloroplasts (or mitochondria). This should be demonstrated experimentally but not logically.
This discussion should be limited to the primary endosymbiosis of Archaeplastida, but not the secondary endosymbiosis. The secondary endosymbiosis is supported by cytological and genetic evidence, and we can trace the pathways of endosymbiosis by identifying intermediates having a nucleomorph. Initial formation of chloroplasts might not be a simple process.
Alternative hypotheses
As an alternative to the current, simplistic notion of endosymbiotic origin of chloroplasts, I present two hypotheses: hidden cyanobacterial lineage and host-directed chloroplasts formation. The first hypothesis is introduced to explain type 2 phylogeny of chloroplast proteins. This type of phylogenetic trees implicate that chloroplast proteins originate from an ancestor of cyanobacteria. This could be achieved by assuming either that gene transfer occurred from a cyanobacterial ancestor to a eukaryotic host before chloroplast formation, or that another, hidden cyanobacterial lineage diverging before the diversification of extant cyanobacteria provided genes to chloroplasts. If the relative stem length is long for all proteins showing type 2 phylogeny, then we can choose the first alternative, namely, ancient gene transfer from a cyanobacterial ancestor to a eukaryote. But MraY, FtsZ, NdhC, and NdhE had short stems. It is difficult to imagine transfer of the genes encoding these proteins from a cyanobacterial ancestor to chloroplasts. That is why I suppose that another, hidden lineage of cyanobacteria might have been kept in a dormant state (with a low evolutionary rate) and provided genes to chloroplasts at later times.
Type 2 phylogeny was consistently obtained for SQD1, SQD2, FtsZ, and other proteins listed in Fig. 2. However, we cannot completely exclude a possibility that type 2 trees could be artefact of phylogenetic reconstruction. Inconsistent phylogenetic trees (type 1 or type 2) obtained with RpoB, RpoC, and some Ndh proteins (see “Results”) could cast doubt about the reality of type 2 phylogeny. We cannot answer this in the current state of knowledge and technique. It is best to keep this question to eventually answer it in the future.
Host-directed chloroplast formation is an attractive alternative to the simplistic view on the endosymbiotic origin of chloroplasts. The results of the present study showed that the genes encoding the enzymes involved in glycolipid biosynthesis and photosynthesis-related proteins could have been present in a eukaryote (future host) before the formation of chloroplasts. This is similar to the premitochondrion hypothesis (Gray 2014) and the “pre-mitochondrial symbioses” hypothesis (Gabaldón 2018) for the origin of mitochondrial proteins, which tried to explain early acquisition of many mitochondrial proteins before the mitochondrial formation. As already explained, the glycolipids have advantage in adaptation to low phosphate environments. We can imagine a modified autogenous scenario, in which a eukaryotic future host acquired ability to form membranes consisting of glycolipids, then incorporated photosynthetic machinery from cyanobacteria into the membranes, and finally acquired ribosomes again from cyanobacteria. It is tempting to assume parallel scenarios for the origin of mitochondria and chloroplasts. However, falsification of endosymbiotic origin of chloroplasts could have a stronger impact than that of mitochondria, because chloroplasts and cyanobacteria share a clearly identified function, photosynthesis. The origin of eukaryotic cells and organelles could be far more complicated than imagined before.
An additional hypothesis is the fatty acid hypothesis. Early acquisition of membrane forming ability was also important from the point of view of fatty acid synthesis. Archaea do not use acyl lipids and have no ability of fatty acid synthesis (for a review, see Caforio and Driessen 2016). Early eukaryotes were obliged to make their membranes by obtaining fatty acids from bacterial preys. Acquisition of ability of fatty acid synthesis is preferable for eukaryotes, but this required a large reducing power. Photosynthetic production of a large reducing power was beneficial to fatty acid synthesis as well as carbon fixation. Early acquisition of fatty acid synthesis ability guaranteed ability to produce numerous photosynthetic membranes in a eukaryotic ancestor. Photosynthesis, fatty acid synthesis, and glycolipid synthesis were a tightly related trio that promoted evolution of photosynthetic eukaryotes, and this does not require endosymbiosis.
Remaining problems and concluding remarks
All our discussions are related to the primary endosymbiosis, namely, the formation of chloroplasts in the initial Archaeplastida. The monophyly of Archaeplastida is supported by phylogenetic analyses cited in Introduction, and the conservation of translocon components (Tic20, Tic21, Tic22, and Toc75). Another line of evidence for monophyly might be the overall similarity in phylogenetic types of various chloroplast proteins (with some exceptions) in the three lineages of Archaeplastida.
We have to analyze the secondary endosymbiosis in future studies. Data on diatoms are included in the phylogenetic trees in the present study, but the phylogenetic relationship of chloroplast enzymes of diatoms with other algae and plants was variable, either within the plant/algal clade, or outside it. Both the red algal enzymes and the enzymes of the secondary host could contribute to the proteome of the diatom chloroplasts. Nevertheless, the shape of phylogenetic tree is sometimes strange for the diatom clade, and we will have to analyze carefully the phylogenetic relationship of diatom chloroplasts with other chloroplasts.
Paulinella chromatophora and its related protists are interesting organisms that originate from a recent (probably about 100 million years ago) endosymbiosis of an α-cyanobacterium (Lhee et al. 2017; Nowack et al. 2008). I have included one or two Paulinella species in all the phylogenetic trees in the present study. All enzymes in the biosynthesis of lipids and fatty acids in the chromatophore are entirely encoded by the chromatophore genome, and originate from cyanobacteria. The same is true for all Calvin-Benson Cycle enzymes, photosynthetic proteins, and division proteins that were analyzed in the present study. These provided good control data that my phylogenetic analyses were properly performed. The only chromatophore enzyme currently known to originate from proteobacteria is MurF involved in peptidoglycan synthesis (Nowack et al. 2016; Sato and Takano 2017). In this respect, the chromatophores of Paulinella chromatophora and related photosynthetic Paulinella are semiautonomous endosymbionts/organelles, and cannot survive outside the host cell. Supporting evidence for semi-autonomy was obtained by lipid analysis and lipid biosynthetic studies recently performed in our laboratory. The situation is fundamentally different from that of chloroplasts of plants and algae. Paulinella could be a good model of endosymbiosis, but we have to distinguish between chromatophores and chloroplasts.
On the other hand, we can recognize an interesting similarity of chromatophores and chloroplasts. Singer et al. (2017) demonstrated that about 450 nuclear-encoded proteins are targeted to chromatophores, and that most of them probably originated from the ancestral eukaryotic host. They suggested re-targeting of these proteins. In other words, the host cell had prepared these re-targetable protein genes before the chromatophore endosymbiosis. This is in line with my hypothesis of host-directed chloroplast formation. Therefore, despite apparent differences in integration of the endosymbiont/organelle within the host cell, we can assume a similar scenario of formation of the endosymbiont/organelle, in which the host had already genes for the proteins to be used in the endosymbiont/organelle.
Diverse origins of chloroplast DNA replication machineries (Moriyama and Sato 2014) and nucleoid components (Sato 2001) conform to the complex history of chloroplast formation described above. DNA polymerase of viral origin was first used in mitochondria and then also used in chloroplasts. Various components related to DNA replication were replaced during the evolution of chloroplasts mainly in the green lineage, but some components were also replaced in the red lineage. All these processes are supposed to occur after the establishment of chloroplasts.
The origins of chloroplast enzymes are diverse, in both donor organisms and acquisition times. Especially, most enzymes involved in chloroplast membrane lipids did not originate from cyanobacteria, but rather from various bacteria and eukaryotes. We cannot adhere to the visual image of the endosymbiotic origin of chloroplasts, explicitly showing membrane heredity. In addition, some of the photosynthesis-related proteins and enzymes that originated from cyanobacteria could be acquired at diverse times, either before or after the acquisition of ribosomal components. The origin of chloroplasts might not be as simple as we hypothesize with the visual image of cyanobacterial endosymbiosis, which occurred once at a fixed time during the evolution of eukaryotes.