Metabolic connectivity as a driver of host and endosymbiont integration

Abstract

The origin of oxygenic photosynthesis in the Archaeplastida common ancestor was foundational for the evolution of multicellular life. It is very likely that the primary endosymbiosis that explains plastid origin relied initially on the establishment of a metabolic connection between the host cell and captured cyanobacterium. We posit that these connections were derived primarily from existing host-derived components. To test this idea, we used phylogenomic and network analysis to infer the phylogenetic origin and evolutionary history of 37 validated plastid innermost membrane (permeome) metabolite transporters from the model plant Arabidopsis thaliana. Our results show that 57% of these transporter genes are of eukaryotic origin and that the captured cyanobacterium made a relatively minor (albeit important) contribution to the process. We also tested the hypothesis that the bacterium-derived hexose-phosphate transporter UhpC might have been the primordial sugar transporter in the Archaeplastida ancestor. Bioinformatic and protein localization studies demonstrate that this protein in the extremophilic red algae Galdieria sulphuraria and Cyanidioschyzon merolae are plastid targeted. Given this protein is also localized in plastids in the glaucophyte alga Cyanophora paradoxa, we suggest it played a crucial role in early plastid endosymbiosis by connecting the endosymbiont and host carbon storage networks. In summary, our work significantly advances understanding of plastid integration and favors a host-centric view of endosymbiosis. Under this view, nuclear genes of either eukaryotic or bacterial (noncyanobacterial) origin provided key elements of the toolkit needed for establishing metabolic connections in the primordial Archaeplastida lineage.

The origin and establishment of the photosynthetic organelle, the plastid, is heralded as one of the most important biological innovations on our planet (1, 2). This primary endosymbiosis occurred more than a billion years ago and resulted from the engulfment and enslavement of a once free-living cyanobacterium by a phagotrophic protist (3). Primary plastid capture putatively occurred a single time in the common ancestor of the eukaryotic supergroup Archaeplastida (also known as Plantae) that comprises the green algae and land plants (Viridiplantae), red algae, and glaucophyte algae (4–6). Once established in these lineages, the plastid spread to other lineages such as diatoms, haptophytes, most dinoflagellates, and euglenids, through red or green algal secondary endosymbiosis, and in some dinoflagellates, through tertiary endosymbiosis of a secondary endosymbiont-containing alga (7, 8). The exceptional rarity of primary plastid endosymbiosis is supported by there being only one other known case of a cyanobacterium-derived photosynthetic organelle (9). This “chromatophore” is found in a single lineage of photosynthetic filose amoebae that includes Paulinella chromatophora and its sister taxa (10–14). This independent primary endosymbiosis likely occurred ∼60 Mya, and the plastid donor was a member of the α-cyanobacterium clade (15, 16).

Given the fundamental role of algae and plants as primary producers in aquatic and terrestrial habitats (17, 18), much attention has focused on elucidating the rules that underlie primary plastid origin in Archaeplastida and, more recently, in Paulinella. We have previously made the argument that a key, and likely fundamental, step in endosymbiont integration (i.e., enslavement) was linking the metabolism of the host and endosymbiont, thereby allowing regulatory pathways to evolve that would maximize connectivity of the partners, and as a result, host fitness (19–21). The major players in this process are transporters located in the innermost envelope membrane of plastids (the plastid envelope permeome) that are responsible for the controlled movement of metabolites to and from the endosymbiont (e.g., energy as photosynthetically fixed carbon; the presumed raison d’être for plastid origin). Our previous work showed that members of the nucleotide sugar transporter family [NST; within the drug/metabolite superfamily (DMS)] gave rise through gene duplication and divergence to a variety of plastidic sugar transporters in red algae and Viridiplantae (Fig. S1) (19, 22, 23). These genes encode the plastidic phosphate translocators (pPTs) that facilitate the strict counter exchange of a host-derived inorganic orthophosphate (P_i) for an endosymbiont-derived phosphorylated C3, C5, or C6 carbon compound (e.g., triose phosphate, xylulose-5-phosphate, glucose-6-phosphate). Along with the shared ancestry of the plastid protein import system (6, 24), this innovation provides one of the strongest pieces of evidence that two major members of the Archaeplastida (red algae and Viridiplantae) are monophyletic. The tree also shows that members of the “chromalveolates” (e.g., stramenopiles, apicomplexans, cryptophytes) gained their pPT homologs through red algal endosymbiosis. The retention of hexose phosphate transport as the primary carbon export mechanism in the third arm of the Archaeplastida, the Glaucophyta (6), provides another intriguing twist in the story of primary endosymbiosis and will be discussed in detail below. This transporter (UhpC) originated through horizontal gene transfer (HGT) from a bacterial source. The work on pPTs inspired us to look in more detail into the evolutionary history and functional diversification of other plastid-targeted transporters and here we present an analysis of these proteins in Archaeplastida. Our approach was to use phylogenomic and protein similarity network analysis of the validated plastidic transporters from Arabidopsis thaliana to deduce their evolutionary histories and origins (25). We also studied the phylogeny and cellular localization of UhpC proteins in red algae to gain insights into what may have been the ancestral pathway of sugar transport in Archaeplastida. These data, combined with recent evidence of apparent translocon-independent protein import to the photosynthetic organelle in Paulinella (26), provide a novel perspective on endosymbiont integration. Based on these data, we suggest that metabolic connectivity, whereby recruitment of existing host-derived transporters to the plastid innermost membrane, was likely an early and fundamental step in unlocking the metabolic potential of the captured cyanobacterium.

Results and Discussion

Phylogenomic Analysis of Arabidopsis Plastidic Transporters.

Phylogenomic analysis of the 34 Arabidopsis thaliana plastid envelope transporters listed in ref. 25 and 3 others that were more recently described in this species [nitrite transporters At5g62720 (AtNITR2;1) and At3g47980 (AtNITR2;2), and the glycolate/glycerate transporter At1g32080 (PLGG1)] (27, 28) were used as queries against a comprehensive local genome database (for details, see Methods and Table S1). The resulting alignments and trees, using either the full alignment length or a gap-trimmed version (available for download at cyanophora.rutgers.edu/transporters/) were inspected to determine the phylogenetic origins of the plant transporter families. Most of these Viridiplantae proteins were found to be either nested with strong bootstrap support (generally >90%; Table S1) within a variety of eukaryotic lineages or to be associated with prokaryotes. A total of 57% of Arabidopsis plastidic transporters were of host (ancient eukaryotic; nine families) origin, only 8% (three families) were of cyanobacterial (putative endosymbiont) origin, 24% were derived from noncyanobacterial prokaryotes (presumably many via HGT) with four families putatively derived from Chlamydiae [i.e., PHT2.1, (NTT1, NTT2), (DiT1, DiT2.1, DiT2.2), HMA1], and a small number were either plant-specific or of uncertain affiliation (Fig. 1). The divergent origin of some transporter families is exemplified by the phylogeny of plastidic phosphate transporters (PHTs) shown in Fig. S2.

Fig. 1.

Results of phylogenomic analysis of A. thaliana plastid-targeted transporters. The specific contributions and their numbers made by the eukaryote host (brown text), cyanobacteria (blue text), prokaryotes other than cyanobacteria (gray text), and other sources are shown.

These results (and previous work) support the scenario that the current, and presumably the primordial, contribution to plastid metabolite transport was dominated by the retargeting of existing host-derived proteins to the plastid envelope permeome rather than by the wholesale repurposing of endosymbiont genes (22, 25, 29). The genes encoding these ancient transporters presumably underwent duplication(s) with one or more copies taking on plastid-specific functions (Figs. S1 and S2). This host-centric perspective has also been taken to suggest that the eukaryote rather than the endosymbiont was the major contributor to protein sorting components with the endosymbiont outer membrane being the initial target for integration (30, 31). A contrasting view (32) relies on genetic tinkering with endosymbiont genes to derive basic components of mitochondrial translocons (31). This lively discussion is far from settled, but it is clear that distinguishing between these hypotheses with regard to different endosymbiont traits depends not only on identifying the putative genetic toolkit for endosymbiont integration (with solute transport and protein import being obvious candidates) but equally importantly, on elucidating their phylogenetic history. Whereas explaining the origins of plastid protein translocon components still remain a challenge (33), here we are able to provide strong evidence for a host-dominated process with regard to plastid metabolite transport. This hypothesis is buttressed by the fact that permanent endosymbionts (e.g., the Paulinella chromatophore genome) are characterized by massive genome decay through outright loss or endosymbiotic gene transfer (EGT) (12–14). Therefore, innovations relating to the compartment are more likely to originate in the eukaryote gene/function-rich nuclear host genome or via HGT to the host from foreign sources.

Network Analysis of Arabidopsis Plastidic Transporters.

To gain a phylogeny-independent perspective on plastid transporter evolution, we generated protein similarity networks with an all vs. all BLASTP analysis of the nonredundant list of database hits to the 34 Arabidopsis transporters described in ref. 25 (see Methods for details). Using a 70% query coverage cutoff resulted in these data forming 16 major (i.e., containing many members) connected components (gene families) that together include all of the transporter families, as well as 5 minor components that include highly diverged family members (e.g., CLT2, PIC1; Fig. 2). Placement of the major components (boxed) into transporter superfamilies using the Transporter Classification Database (TCDB; www.tcdb.org/) shows that most plastidic transporters are anciently diverged and derive from distinct superfamilies, with six components (boxed in gray field) comprised solely of eukaryotic or plant-specific sequences (Fig. 2). Nonetheless, several interesting components were identified, including the expected pPT family (e.g., PPT, TPT, XPT) that is derived from a single gene duplication of an existing NST in the ancestor of red algae and Viridiplantae (Fig. S1). Another family of interest contained the copper transporting P-type ATPases HMA1 and PAA1 (34–36). This component included the eukaryote derived ACA1/PEA1 (a calcium ATPase; Table S1) family that shares some links with the HMA1/PAA1 cluster (Figs. 2 and 3). To test this latter connection, we imposed a minimum pairwise protein identity threshold of 40%. This restriction was applied together with the standard e-value cutoff ≤10⁻⁵ and a minimum of 70% query hit coverage to produce a new network. Under this stringent condition (Fig. S3), the ACA1/PEA1 family became an independent component, whereas links remained between many of the copper transporters. Inspection of the protein alignment suggested that the links to ACA1/PEA1 were likely explained by the presence of two shared domains that resulted in the network interaction (Fig. 3A).

Fig. 2.

Network analysis of 34 Arabidopsis plastid-targeted transporters. Each connected component is identified with respect to transport function, superfamily classification, and the taxonomic composition of network nodes. Highly diverged family members that form independent components under the cutoff used for pairwise comparison (i.e., 70% coverage) are also shown. Components comprised solely of eukaryotic sequences are shown in the gray boxes.

Fig. 3.

Network and phylogenetic analysis of the plastid targeted HMA1-PAA1 copper transporters. (A) The network that includes the ACA1/PEA1 calcium ATPase of apparent eukaryotic origin that shares a weak connection to HMA1/PAA1 transporters based on domain sharing (Fig. S3). The nodes in this network are labeled according to taxonomic origin within prokaryotes and eukaryotes. (B) The same network was relabeled with nodes indicating distribution in different eukaryotic phyla. This image highlights the independent prokaryotic origins of the HMA/PAA1 family in eukaryotes. (C) RaxML tree (WAG + Γ model of evolution) of the HMA1/PAA1 data with redundant sequences removed. The results of 100 bootstrap replicates are shown at the branches. Cyanobacteria are in blue text, Viridiplantae are in green text, red algae are in red text, Chlamydiae are in magenta text, and all other taxa are in black text. The query transporter sequences from Arabidopsis are shown in boldface black text within each Viridiplantae clade.

Labeling this network in two different ways (Fig. 3 A and B) shows that both of the copper transporters in Arabidopsis and in other eukaryotes are derived from prokaryotic ancestors. Phylogenetic analysis of the combined alignment resolves gene origins with PAA1 having a cyanobacterial (endosymbiont) provenance, whereas HMA1 has a chlamydial origin (Fig. 3C). Intriguingly, both of these prokaryotes are implicated in plastid origin with Chlamydiae, providing some key genes required for starch synthesis and metabolic integration of the endosymbiont (37). The biological meaning of this network result is, however, unclear. Copper is an essential micronutrient that is a component of photosystems (cofactor of plastocyanin) and Cu/Zn superoxide dismutase that is involved in the dismutation of superoxide to hydrogen peroxide (38). However, due to its high toxicity (e.g., it can catalyze the production of free radicals), in plant cells, copper is associated with Cu chaperones (39, 40). Recent work suggests that the HMA1 and PAA1 transporters operate as distinct pathways for copper import into plastids (36), thereby putatively explaining the maintenance of two diverged copies in Archaeplastida.

Another intriguing connection uncovered by the network analysis was between three members of the mitochondrial carrier superfamily (MCF) that includes the SAMT/C [counter exchange of S-adenosylmethionine (SAMT; cytosol) with S-adenosylhomocysteine (SAHC; plastid)], NDT1 [counter exchange of NAD⁺ (cytosol) with AMP or ADP (plastid)], and BT1 (in Arabidopsis, unidirectional flow of plastid AMP, ADP, ATP to the cytosol; Figs. S4 and S5]. These three adenosine-based transporters are of eukaryotic origin and have evolved differing transport activities during Archaeplastida evolution. Phylogenetic analysis of the combined alignment (Fig. S4C) shows that all of these plastid (re)targeted transporter families likely trace their origin to a single gene that was present in the common ancestor of eukaryotes. This hypothesis is supported by the absence of prokaryotic homologs of this gene family in the tree (i.e., using our search parameters) and the observation that the transporter subtrees contain a wide array of eukaryotic lineages (e.g., fungi, metazoans, red algae), suggesting ancient provenance in the ancestor of these taxa.

In summary, the phylogenomic and network analyses point to a fundamental role for the host cell in the evolution of metabolic connectivity of the endosymbiont. This development is not only in terms of the number of proteins that have been recruited to the envelope permeome but also in the crucial roles they play, from fixed carbon transport to the delivery of methyl donors (SAMT/C) to the plastid to facilitate prenyllipid biogenesis or to regulate the synthesis of aspartate-derived amino acids (41).

Evolution of the UhpC family.

A surprising finding of the analysis of the genome of the glaucophyte Cyanophora paradoxa is that in contrast to red algae and Viridiplantae, it does not encode NST-derived plastid-targeted sugar-phosphate transporters (Fig. S1) (6). Six nonplastidial, likely endomembrane-targeted, NST proteins are present in the Cyanophora genome, suggesting that the glaucophytes had split from the Archaeplastida before the gene duplications and acquisition of plastid-targeting signals that propelled pPT evolution in red algae and Viridiplantae. A closer inspection of Cyanophora gene models, however, turned up two novel candidates for plastid sugar-phosphate transport in this species. These genes encode homologs of bacterial membrane bound UhpC-type hexose-phosphate sensors, which in bacteria are part of an operon that is responsible for the uptake of glucose-6-phosphate (G-6-P) (42). In contrast to their bacterial ancestors, both Cyanophora UhpC homologs feature an N-terminal extension that displays characteristic features of glaucophyte plastid-targeting sequences (6). Analysis of the plastid inner envelope proteome of Cyanophora confirmed the presence of the two UhpCs in the envelope fraction (43). This finding was further corroborated by transient expression of fluorescent-tagged Cyanophora UhpCs in tobacco cells, where they localized to the periphery of the plastids (43). Whereas NST-derived plastid sugar-phosphate transporters are restricted to red algae and Viridiplantae, it now becomes clear that UhpC is widespread in algal members of the Archaeplastida (Fig. 4A; but lost in plants). In Cyanophora, UhpC is putatively responsible for the counter exchange of orthophosphate for G-6-P, thereby likely constituting the ancestral path of carbon export from plastids in the Archaeplastida (43).

Fig. 4.

Analysis of UhpC proteins in Archaeplastida. (A) RaxML tree (WAG + Γ model of evolution) with the results of 100 bootstrap replicates shown at the branches. Glaucophyta are in blue text, Viridiplantae are in green text, red algae are in red text, Chlamydiae are in magenta text, and all other prokaryotes are in black text. The four proteins for which localization data exist to indicate plastid inner envelope membrane targeting are shown in large boldface black text. The results of bioinformatic targeting predictions using ChloroP (cTP score; marked with [+] when predicted to be plastid targeted or [−] when not) are shown for each Achaeplastida UhpC, when complete proteins are available. (B) Expression of YFP-fusion constructs in Nicotiana benthamiana protoplasts. Confocal microscope pictures on isolated Nicotiana protoplasts expressing the YFP-fusion constructs driven by the ubiquitin 10 promoter. (Plate 1) Expression pattern of the predicted transit peptide of the UhpC homolog Gasu_03960 from Galdieria fused to YFP. Shown are YFP fluorescence (YFP) in green, chlorophyll autofluorescence (chlorophyll) in red and an overlay of the two pictures (merge). (Plate 2) Expression pattern of the full-length Gasu_03960 fused to YFP. (Plate 3) Expression pattern of the predicted transit peptide of the UhpC homolog CYME_CMQ264C from Cyanidioschyzon fused to YFP. (Plate 4) Expression pattern of the full-length CYME_CMQ264C fused to YFP.

Here we extended this analysis by generating in silico targeting predictions for the other Archaeplastida UhpCs to determine whether any of these might also be plastid destined. Our hypothesis was that if red and/or green algal UhpCs are also plastid targeted, then the common ancestor of Archaeplastida was likely to have relied on this bacterium-derived protein as the primordial transporter of a C6 carbon compound that can be fed directly into cytosolic glycogen biosynthesis after being converted into UDP-glucose (i.e., the ancestral location of starch synthesis in Archaeplastida before relocation to the Viridiplantae chloroplast). This HGT event could have occurred before (or perhaps coincident) with the retargeting of members of the NST family and before origin of the pPT genes (43). Use of the program ChloroP 1.1 (44) suggested that many complete UhpC sequences in red and green algae are likely to be plastid targeted (Fig. 4A). We tested this hypothesis for UhpCs in the red algae Galdieria sulphuraria (45) and Cyanidioschyzon merolae (46). To this end, we expressed fluorescent-tagged fusion proteins from these algae in Nicotiana benthamiana and observed their localization by fluorescence microscopy. The full-length proteins of G. sulphuraria (gene Gasu_03960) and of C. merolae (gene CYME_CMQ264C), as well as the first 112 and 200 amino acids of each protein corresponding to their putative transit peptides, were cloned in front of the N terminus of the yellow fluorescent protein (YFP). Observation of the protoplast isolated from leaves infiltrated with the constructs carrying the predicted transit peptides fused to the YFP shows that the fluorescent signal colocalizes with the chlorophyll autofluorescence (Fig. 4B, Plates 1 and 3). This result indicates that the red algal predicted transit peptides are indeed sufficient to target the YFP to the chloroplast. The fluorescent signal of the full-length proteins fused to the YFP for Gasu_03960 and CYME_CMQ264C surrounds the plastids, a pattern typical for proteins located to the inner envelope membrane (Fig. 4B, Plates 2 and 4). This observation is consistent with localization to the plastid inner envelope membrane as previously described (43).

These results can also interpreted from the perspective of the ménage à trois hypothesis of symbiont integration (37, 47). This hypothesis posits that plastid endosymbiosis putatively relied on three partners: the host, the cyanobacterium, and a chlamydial symbiont (hence, ménage à trois). Under this view, the UhpC transporter (of bacterial origin; see below) could have been transferred first to the cyanobacterium genome to allow the flow of fixed carbon to the inclusion vesicle (of chlamydial origin) that housed both prokaryotes. The G-6-P that was secreted from the cyanobacterium, after conversion to G-1-P by phosphoglucomutase, acted as substrate for the synthesis of either ADP- or UDP-glucose and subsequently glycogen in the vesicular space. Glycogen was also synthesized in the host cytosol with the use of chlamydial-derived enzymes (GlgA, GlgC) secreted by their type 3 secretion system (TTS). Regardless of the manner in which the ancient ménage à trois interaction operated, over time the UhpC (and other endosymbiont) genes were transferred to the host genome and retargeted to the plastid inner membrane to allow uptake of G-6-P for glycogen synthesis. Therefore, genes provided to the host by chlamydial cells, other bacteria, and the cyanobacterium forged the successful endosymbiosis, and the chlamydial symbiont (and their inclusion vesicles) was lost (47). The phylogenetic evidence for a chlamydial origin of UhpC in Archaeplastida is ambiguous because of the presence of many Proteobacteria in the tree that also forms a sister to the eukaryote clade. Regardless of the rooting scheme that is used, the direction of HGT in this case is, however, clear. Among eukaryotes, only Archaeplastida contain a bacterium-derived UhpC gene; therefore, the transfer was most likely from a prokaryote to a eukaryote cell.

The Metabolic Connectivity Model.

Given the strong evidence for eukaryotic or HGT-derived metabolite transporters in Archaeplastida, both of which are nuclear encoded, how might these proteins be directly implicated in early events in endosymbiont integration? For clues to answering this question, we can look at the recent plastid endosymbiosis in Paulinella. Work done by Nowack and Grossman (26) has shown that this amoeba apparently relies on the secretory system, via passage through the Golgi, to deliver essential photosystem proteins (PsaE, PsaK1, PsaK2) to the chromatophore. These proteins have been lost from the endosymbiont genome and are now encoded in the nucleus, synthesized in the cytoplasm, and trafficked into the chromatophore. The absence of a detectable terminal or internal targeting signal in these proteins suggests that the secretory system is capable of delivering mature proteins to an endosymbiont. We take these results as support for the idea that the initial metabolite transporters that serviced the Archaeplastida (as described above) plastid could also have been delivered by the secretory system to the inner envelope membrane to allow exchange of metabolites with the host cell (20). Our work shows that the UhpC hexose-phosphate transporter present in the Archaeplastida ancestor could have been an early player in fixed carbon transport, as envisioned in the ménage à trois hypothesis, or under a scenario that does not rely on an inclusion vesicle as the initial platform for glycogen synthesis.

The hypothesis of a host-dominated process for metabolite transport in the nascent plastid endosymbiont as postulated here would be greatly strengthened by evidence that crucial functions were provided by the eukaryote. Another example of such a function is the essential detoxification of the oxygenation product of rubisco, 2-phosphoglycolic acid (2-PG). Cyanobacteria possess at least three distinct metabolic routes to this end, including the bacterial-type glycolate pathway and a plant-like C₂-pathway (48, 49). The capacity to detoxify 2-PG inside the plastid must have been lost early during endosymbiosis because all members of the Archaeplastida rely on the export of glycolic acid from plastids to the cytoplasm for conversion to glycerate in mitochondria and/or peroxisomes. This process requires the export of glycolic acid from, and the import of glycerate into, the plastid, which is achieved by the recently discovered glycolate/glycerate transporter PLGG1, a eukaryote-derived protein (Table S1). Some cyanobacteria are able to excrete glycolate at significant levels, with 9% of net fixed carbon lost in air for heterocystous species and up to 60% lost in some strains under high oxygen conditions (50). This mechanism, however, represents a significant loss of carbon and of Calvin–Benson cycle intermediate products. Hence, recovery of part of this carbon by the eukaryotic photorespiration pathway would offer a significant selective advantage. The advantage of PLGG1 over other (unknown) cyanobacterial glycolate exporters is that it works in a counter exchange mode, which leads to the salvage of three of four carbon units lost via oxygenation/photorespiration.

In summary, the metabolic connectivity model we present here relies on functional diversification of organelle functions. Through targeting of host-derived transporter proteins to the organelle membrane, two previously distinct metabolic networks become increasingly interwoven and interdependent. This idea is supported by previous work that shows the growth of metabolic networks in prokaryotes is frequently achieved by the addition of transporter activities at the network periphery. That is, given a core set of metabolic functions, the network can be expanded by gaining access to additional substrates via the acquisition of genes encoding transporter proteins. These elaborations are driven by adaptation to changing environments (e.g., extracellular vs. intracellular) (51). In addition, it is widely recognized that free-living bacteria have access to a large pan-domain gene pool from which they can acquire novel metabolic and transporter genes. Therefore, the cyanobacterial plastid ancestor had a chimeric genome at the time of endosymbiosis, components of which were transferred to the eukaryote host by EGT. The vast prokaryotic gene pool also contributed to Archaeplastdia evolution via independent HGT events. The role of HGT in endosymbiont integration is of particular relevance because once cells are internalized they rarely gain genes, but rather undergo significant genome reduction (12–14). Hence, it is not surprising that, as we show here, growth in functional diversity of plastids is achieved by the retargeting of host encoded proteins or the acquisition by the host of foreign genes whose products can be retargeted to the organelle.

Methods

Network and Phylogenetic Analysis.

The 34 different validated A. thaliana plastid inner membrane transporter sequences listed in Fischer (25) and the three new transporters (27, 28) were used in our analyses. Each protein sequence from Arabidopsis was used to query via BLASTP an in-house database composed of The National Center for Biotechnology Information Reference sequence database (RefSeq) v59, the 672 proteomes publicly available from the Moore Marine Eukaryote Transcriptome Sequencing Project (MMETSP; camera.calit2.net/mmetsp/list.php), with additional eukaryote proteomes obtained from NCBI dbEST, (www.ncbi.nlm.nih.gov/dbEST/), and TBestDB. The BLASTP output for each transporter protein sequence was analyzed so that a maximum of 100 individual hits was retained with a limit of ≤12 hits per phylum to avoid oversampling taxon-rich phyla in the database. Each dataset was aligned using Clustal Omega (52), and phylogenetic trees were constructed with the full alignment using IQ-TREE v.0.9.6 (53) with automatic best-fit model selection and 1500 μLtra-fast bootstrap replicates (Fig. S2). We also used Gblocks (54) to trim each transporter alignment to conserved sites using a block size of 2 (−b4 = 2) and allowing all gaps (−b5 = a). These reduced alignments were used as input to IQ-TREE as described above to generate phylogenies and bootstrap support values. For some transporters (e.g., pPTs, PTs), the individual protein alignments were combined, and a RAxML (55) tree was inferred using the WAG + Γ model of sequence evolution and 100 bootstrap replicates.

For the network analysis, the complete set of transporters that was returned from the phylogenomic analysis was analyzed to remove all redundant sequences. This data set of 2,330 unique sequences was used in an all vs. all BLASTP analysis with a minimum e-value cutoff ≤10⁻⁵; i.e., a threshold that selected hits based on both bit score (the similarity measure) and length. This restriction returned short hits to regions with high similarity while allowing hits to longer regions with lower similarity. The lowest similarity value recorded for the hits was 18.15% identity. From these results, we applied a cutoff of 70% query coverage (i.e., any hit must cover at least 70% of the 2 proteins), resulting in a set of 21 connected components that encompass 1,727 proteins. The classification of proteins in these connected components was done using a greedy clustering algorithm based on the modularity (ratio of intra- vs. intergroup links) (56) that detects densely connected regions of sequences. New classes of transporters were manually defined to regroup transporter families listed in Fischer (25) that have homologs that belong the same community in the network. The taxonomic content of the network was analyzed and displayed using in-house R scripts (57) with the igraph package (Figs. 2 and 3).

The motivation to use both phylogenetic and network methods with the same data reflects the fact that these approaches provide complementary insights into transporter evolution. Phylogenetic methods assume that sequences are inherited via vertical descent and the goal here was to detect these instances, as well as subsets of the data that contradict this assumption due to well-resolved cases of endosymbiotic or horizontal gene transfer. In contrast, network analysis is a straightforward clustering approach with input sequences, in which no assumption is made regarding the mechanism of evolution. Sequence clustering is based on several criteria (e.g., hit coverage, similarity score) using a robust algorithm that detects densely connected sets of proteins. Importantly, the clustering allows us to detect sequences that are related through intermediates that would not be present in the same tree due to high pairwise sequence divergence between the most distantly related protein families. This property potentially allows us to merge regions of different trees into single clusters to identify ancient connections between transporters. We have previously applied this principle to detect ancient signals of monophyly between redox enzymes, allowing inferences into their origin and evolution in divergent prokaryotic lineages (58–60). Here the networks were used in a more limited fashion to potentially identify transporter family relationships that may not be revealed using phylogenetic methods.

Localization of Transporter Proteins in Transiently Transformed Tobacco Cells.

The coding sequences of the UhpC homologs from Galdieria sulphuraria (Gasu_03960) and Cyanidioschyzon merolae (CYME_CMQ264C) were amplified by PCR and cloned by Gibson cloning into the plant expression vector pUBC-YFP for C-terminal YFP-fusion under the control of the ubiquitin 10 promoter (61, 62). Agrobacterium tumefaciens transformation (strain GV3101), tobacco leaf infiltration, and protoplast isolation were done as previously described (43). Isolated protoplasts were observed 2–4 d after infiltration with an inverted Zeiss LSM 780 confocal laser-scanning microscope. YFP and chlorophyll were excited using the 488-nm laser line of an Argon laser, and the emission was collected at 517–578 and 590–690 nm, respectively.