Metabolic Effectors Secreted by Bacterial Pathogens: Essential Facilitators of Plastid Endosymbiosis?^,

Abstract

Under the endosymbiont hypothesis, over a billion years ago a heterotrophic eukaryote entered into a symbiotic relationship with a cyanobacterium (the cyanobiont). This partnership culminated in the plastid that has spread to forms as diverse as plants and diatoms. However, why primary plastid acquisition has not been repeated multiple times remains unclear. Here, we report a possible answer to this question by showing that primary plastid endosymbiosis was likely to have been primed by the secretion in the host cytosol of effector proteins from intracellular Chlamydiales pathogens. We provide evidence suggesting that the cyanobiont might have rescued its afflicted host by feeding photosynthetic carbon into a chlamydia-controlled assimilation pathway.

INTRODUCTION

Members of the photoautotrophic lineage, the Archaeplastida (also called Kingdom Plantae) consisting of the Chloroplastida (green algae and plants), Rhodophyceae (red algae), and Glaucophyta (glaucophytes) are the founding lineage of photosynthetic eukaryotes (Rodríguez-Ezpeleta et al., 2005, Reyes-Prieto et al., 2007; Chan et al., 2011). Despite the advantages offered by photosynthesis and the innumerable opportunities that presumably have existed for phagotrophic protists to recapitulate plastid acquisition, only one other case is known of primary plastid endosymbiosis, in the photosynthetic amoeba Paulinella (Nowack et al., 2008). To understand why primary endosymbiosis is exceedingly rare, clues have been sought in the nuclear genomes of Archaeplastida. The presence of a significant cyanobacterium-derived component in these genomes is known to have arisen from endosymbiotic gene transfer, which resulted in the movement of many cyanobiont genes to the host chromosomes (Martin et al., 2002; Reyes-Prieto et al., 2006; Moustafa and Bhattacharya, 2008). Recent analyses suggest that between 6 and 10% of nuclear genes in algae retain the signature of cyanobacterial origin with most encoding proteins that return to the plastid to express their function (Sato et al., 2005; Reyes-Prieto et al., 2006; Moustafa and Bhattacharya, 2008). The second major source of foreign genes in Archaeplastida is Chlamydia-like pathogens (Stephens et al., 1998; Huang and Gogarten, 2007; Moustafa et al., 2008; Becker et al., 2008; Fournier et al., 2009; reviewed in Horn, 2008). The existence of 30 to 50 cases of horizontal gene transfers (HGTs) from members of the order Chlamydiales suggests a significant role for these obligate intracellular bacteria in Archaeplastida evolution. This may have included metabolic integration of the newly acquired plastid (Huang and Gogarten, 2007; Becker et al., 2008; Moustafa et al., 2008; Fournier et al., 2009). This intriguing possibility is buttressed by the observation that a significant portion of the Chlamydiales-derived genes are shared by two or all three archaeplastidal lineages (i.e., green and red algae and Glaucophyta) and therefore must have been present in their common ancestor (Huang and Gogarten, 2007; Becker et al., 2008; Moustafa et al., 2008; Fournier et al., 2009; reviewed in Horn, 2008).

We show that the sole extant enzyme that could have been devoted to the assimilation of photosynthate in the host cytosol is an enzyme of chlamydial origin. Because such assimilation is necessary for the selection of plastid endosymbiosis, this enzyme had to be present at the onset of endosymbiosis in the host cytosol. This specific function defines a previously unsuspected class of effectors secreted by Chlamydiales bacteria. This finding suggests strongly that an intracellular bacterium ancestor to the extant order Chlamydiales, a cyanobacterium, and a eukaryotic host were indeed linked together in a stable tripartite symbiotic relationship that led to the establishment of the plant lineages.

RESULTS

Phylogenomic Approaches: Assessing the Chlamydialean Contribution to Archaeplastidial Genomes

We reanalyzed the contribution of Chlamydiales to the genomes of Archaeplastida using the additional data (e.g., >60k proteins from the red algae Calliarthron tuberculosum and Porphyridium cruentum; Chan et al., 2011) that have become available since previous analyses (Huang and Gogarten, 2007; Becker et al., 2008; Moustafa et al., 2008). The set of predicted proteins from all available Chlamydiales genomes (28,244 proteins) was used to query a local database that contained annotated protein sequences from RefSeq release 37 and genome data from other public databases (for details, see Methods). This phylogenomic analysis returned 23,470 RAxML maximum likelihood trees with clade support values for each phylogeny based on 100 bootstrap replicates. Sorting these trees at the bootstrap cutoff of 75% showed that the majority (67%) of all chlamydial proteins are of bacterial affiliation. However, 584 distinct chlamydial proteins formed a sister group to Archaeplastida (Figure 1A). The chlamydial proteins of eukaryote affiliation were grouped into clusters of protein families (conservatively, trees with greater than or equal to one protein shared among them were united into a cluster). This showed that the Archaeplastida were the major recipient of these lateral gene transfers (LTGs) with 48 protein families associated with this lineage (see Table 1 for a full list). A full display of the 48 phylogenetic trees can be found in Supplemental Data Set 1 online. Among these genes, we found 12 HGTs shared by two of the three archaeplastidial lineages and three of them shared by all three archaeplastidial clades (Figure 1B). Changing the bootstrap cutoff for Chlamydiales-Archaeplastida monophyly to 90% still yielded 30 protein families. At the 75% cutoff, within the Archaeplastida, the Chloroplastida (green algae and land plants) contain most of chlamydial genes followed by Rhodophyta (red algae). These numbers reflect the current availability of genome data (see Supplemental Table 1 online) with animals (976,563 proteins in the database) and Chloroplastida (506,307 proteins) being data rich and other groups, such as haptophytes (96,133 proteins), cryptophytes (38,939), and glaucophytes (57,737 proteins), currently being relatively data poor. The data-rich fungi (451,434 proteins) contain only 14 chlamydial genes and the Amoebozoa (131,139 proteins), which are frequently infected by chlamydial species, contain only 12 genes (Horn, 2008). Hence, the Archaeplastida are significantly enriched in chlamydial LGTs, with several dozen distinct genes from Chlamydiales present in this lineage.

Figure 1.

Chlamydial Genes in Archaeplastida.

(A) Distribution of BLASTx hits to proteins encoded on chlamydial genomes. The vast majority of proteins are, as expected, of bacterial affiliation, whereas the second largest class of hits is to Archaeplastida.

(B) The sister group relationships of Chlamydiales-derived proteins in Archaeplastida at bootstrap cutoff ≥90%, ≥75%, and when these numbers are added, including the four genes shared exclusively by Chlamydiales and Archaeplastida (see Table 1)

Table 1. Minimal List of Candidate HGTs between Chlamydiales and Archaeplastida.

Eggnog	Description	Pfam	RefSeq	Bootstrap	Topology
COG0037	Predicted ATPase of the PP-loop superfamily implicated in cell cycle control	PF01171	gi16752030	90	G
COG0039	Malate/lactate dehydrogenases	PF00056	gi15605100	90	G
COG0042	tRNA-dihydrouridine synthase	PF01207	gi46400854	90	G
COG0144	tRNA and rRNA cytosine-C5-methylases	PF01189	gi297620390	90	R+Gl
COG0162	Tyrosyl-tRNA synthetase	PF00579	gi15834952	90	G+R
COG0217	Uncharacterized conserved protein	PF01709	gi46400603	90	G+R
COG0275	Predicted S-adenosylmethionine–dependent methyltransferase involved in cell envelope biogenesis	PF01795	gi46399586	90	G
COG0304	3-Oxoacyl-(acyl-carrier-protein) synthase	PF00109	gi298537950	90	G+R
COG0306	Phosphate/sulfate permeases	PF01384	gi15605425	75	G
COG0324	tRNA δ(2)-isopentenylpyrophosphate transferase	PF01715	gi46400518	90	G
COG0448*	GlgC; ADP-Glc pyrophosphorylasea		gi297375973	NA	G
COG0517	FOG: CBS domain	PF01380	gi46401057	90	G
COG0527	Aspartokinases	PF00696	gi297376397	90	Gl
COG0545	FKBP-type peptidyl-prolyl cis-trans isomerase1	PF00254	gi281497802	75	R
COG0547	Anthranilate phosphoribosyltransferase	PF00591	gi123763216	90	G+R
COG0564	Pseudouridylate synthases, 23S RNA-specific	PF00849	gi281498107	75	R
COG0566	rRNA methylases	PF00588	gi46399416	90	G
COG0574	Phosphoenolpyruvate synthase/pyruvate phosphate dikinase	PF02896	gi281499037	90	Gl
COG0588	Phosphoglycerate mutase 1	PF00300	gi15834720	90	G
COG0590	Cytosine/adenosine deaminases	PF00383	gi281498821	75	G+R
COG0720	6-Pyruvoyl-tetrahydropterin synthase	PF01242	gi297376398	90	R+Gl
COG0812	UDP-N-acetylmuramate dehydrogenase	PF02873	gi15834838	90	G
COG0821	Enzyme involved in the deoxyxylulose pathway of isoprenoid biosynthesis	PF04551	gi46400015	75	G
COG1054	Predicted sulfur transferase	PF00581	gi46399653	90	G+Gl
COG1092	Predicted SAM-dependent methyltransferases	PF10672	gi281500146	75	Gl
COG1164	Oligoendopeptidase F	PF01432	gi46400453	90	R
COG1165	2-Succinyl-6-hydroxy-2,4-cyclohexadiene-1-carboxylate synthase	PF02776	gi336482699	75	R
COG1169	Isochorismate synthase	PF00425	gi336482698	90	R
COG1181	d-Ala-d-Ala ligase and related ATP-grasp enzymes	PF01820	gi46399673	90	G
COG1187	16S rRNA uridine-516 pseudouridylate synthase and related pseudouridylate synthases	PF00849	gi270284888	90	G
COG1212	CMP-2-keto-3-deoxyoctulosonic acid synthetase	PF02348	gi281500494	75	G
COG1218	3-Phosphoadenosine 5-phosphosulfate (PAPS) 3-phosphatase	PF00459	gi334695582	75	G
COG1477	Membrane-associated lipoprotein involved in thiamine biosynthesis	PF02424	gi333410213	90	R
COG1496	Uncharacterized conserved protein	PF02578	gi46401143	75	Gl
COG1523	GlgX; type II secretory pathway, pullulanase PulA and related glycosidases	PF00128	gi46400381	90	G+R+Gl
COG1576	Uncharacterized conserved protein	PF02590	gi46400983	75	R
COG1611	Predicted Rossmann fold nucleotide binding protein	PF03641	gi46399856	90	Gl
COG1878	Predicted metal-dependent hydrolase	PF04199	gi336483621	75	G
COG1947	4-Diphosphocytidyl-2C-methyl-d-erythritol 2-phosphate synthase	PF00288	gi46400864	90	G+R+Gl
COG2217	Cation transport ATPase	PF00122	gi15834725	90	G+R
COG2265	SAM-dependent methyltransferases related to tRNA (uracil-5-)-methyltransferase	PF05958	gi46401273	90	G+R
COG3914	Predicted O-linked N-acetylglucosamine transferase, SPINDLY family	PF00534	gi336480908	75	R
euNOG05012	2-C-Methyl-d-erythritol 4-phosphate cytidylyltransferase	PF01128	gi46445961	75	G+R+Gl
KOG4197	FOG: PPR repeat	PF01535	gi336482601	NA	G
NOG04817	Putative uncharacterized protein		gi281499473	90	G
NOG04985	Protein involved in glycerolipid metabolism	PF12695	gi29834553	90	R
NOG05008	Glycerol-3 protein	PF01553	gi46400593	NA	G+R+Gl
NOG87709	Glycosyltransferase family A (GT-A)	PF01704	gi46399599	NA	G+Gl

The Archaeplastida are named as follows: G, Chloroplastida; R, Rhodophyta; and Gl, Glaucophyta. These candidate HGTs were identified using an automated phylogenomic procedure that took as the selection criterion monophyly of Chlamydiales and Archaeplastida to the exclusion of any other clade (except algae that contain secondary plastids, such as chromalveolates and euglenids) with bootstrap values ≥90 or ≥75%. Trees that only contained Chlamydiales and Archaeplastida are marked with “NA.” This list does not contain examples of chlamydial gene transfers that have a complex gene history, such as GlgA (this work) and the ATP translocator that has been previously described. The trees listed here are available in the same order in Supplemental Data Set 1 online. It should be noted that the putative gene functions result from an automated annotation procedure and should be interpreted with caution.

^aThe automated annotation GlgC-ADP-Glc pyrophosphorylase is justified by the presence of a GlgC-like domain within an otherwise unknown protein. This protein found in prasinophytes, stramenopiles, and chlamydiales does not define a classical and functional ADP-Glc pyrophosphorylase.

The Search for Parasite Candidate Genes That Could Have Participated in Plastid Endosymbiosis

The stable maintenance of energy parasites during the metabolic integration of plastids suggests that a chlamydia-like organism that infected the ancestor of the Archaeplastida adopted an essential function in establishment of the cyanobiont. This essential function likely existed at the time when the cyanobiont entered the host through phagocytosis and established the initial metabolic connection with the eukaryote. Recent phylogenetic and biochemical evidence suggest that a likely candidate was export of the bacterial-specific metabolite ADP-Glc from the cyanobiont to the host cytosol (Deschamps et al., 2008a) where the latter could be stored by conversion to glycogen. The immediate flow into the osmotically inert cytosolic carbon stores for delayed use by host metabolism was required because there was no possibility to adjust carbon supply and demand between the two partners (reviewed in Ball et al., 2011). This function must have required a nucleotide sugar translocator that was recruited to the cyanobacterial inner membrane from the host endomembrane system (Deschamps et al., 2008a; Colleoni et al., 2010). However, this scenario would also require a glucan synthase in the host cytosol able to use ADP-Glc to generate storage polysaccharide (Deschamps et al., 2008a; reviewed in Ball et al., 2011). The presence of this enzyme, deduced from pathway reconstruction, is surprising because eukaryotes lack the ability to synthesize or use the bacterium-specific ADP-Glc metabolite.

Extant green plants and algae synthesize starch from a minimum of five enzymes named GBSS, SSI, SSII, SSIII, and SSIV. Figure 2 shows that these enzymes fall into two major monophyletic groups consisting on the one hand of GBSSI, SSI, and SSII on their own (referred to as the GBSS-SSI-SSII clade) and on the other hand of SSIII, SSIV, and a number of related bacterial sequences from proteobacteria cyanobacteria and chlamydiales (the whole group being referred to as the SSIII-IV clade). The source of the GBSSI SSI SSII clade can be reasonably assumed to be the ancestor of the form of GBSSI found in extant cyanobacteria (Deschamps et al., 2008a; Ball et al., 2011). Indeed, if the entire clade is rooted with such a source in mind, then duplication of the chloroplastidial GBSSI followed by mutations that yielded soluble enzymes explain the selective appearance of SSI-SSII in the green algae. This must have coincided with the complex rewiring of starch metabolism from the cytosol to the chloroplast when the chloroplastidial lineage diverged (Deschamps et al., 2008b; reviewed in Ball et al., 2011). GBSSI has been recently discovered in a group of clade B unicellular diazotrophic cyanobacteria (Deschamps et al., 2008a) whose ancestors, according to Gupta (2009), could qualify as the plastid source. The biochemical properties of this diazotrophic cyanobacterium have been studied in detail (Deschamps et al., 2008a). The cyanobacterial enzyme has similar functions and biochemical properties to the plant enzyme, with one exception. This is its high selectivity for ADP-Glc; the cyanobacterial enzyme has no residual activity with UDP-Glc. This is in contrast with all archaeplastidial enzymes that use both purine and pyrimidine nucleotide sugars, unlike the soluble starch synthases of plants, red algae, and glaucophytes (Leloir et al., 1961; Deschamps et al., 2006; Shimonaga et al., 2007; Plancke et al., 2008). This observation is good agreement with both a proposed cyanobacterial origin of the GBSSI gene and an ancestral dual substrate pathway in the cytosol of the common ancestor. GBSSI is well known to display very low (nonphysiological) activity as a soluble enzyme (Dauvillée et al., 1999; Edwards et al., 1999). Since there are no reported examples of eukaryotes not derived from Archaeplastida that accumulate starch, we assume that at the very onset of plastid endosymbiosis the cytosolic pools consisted of glycogen that would not have supported GBSSI activity, thereby disqualifying it from early recruitment to polysaccharide biosynthesis. Because SSI-SSII originated from GBSSI, duplicated in the green lineage, and because analogous proteins have not been found elsewhere, despite searching hundreds of fully sequenced bacterial genomes, we also conclude that genes encoding SSI-SSII were not available at the onset of endosymbiosis.

Figure 2.

Phylogenetic Analysis of ADP-Glc–Specific Starch Synthases of Bacteria and Archaeplastida.

Unrooted GlgA (glycogen starch synthase) tree using the best-fit protein model (LG + G) and 1000 RAxML bootstrap replicates (alignments are given in Supplemental Data Sets 2 to 4 online). To identify the source of the ADPGlc-using glucan synthase involved in endosymbiosis, the phylogeny of archaeplastidal starch synthases compared with bacterial glycogen synthases was determined using maximum likelihood analysis with 1000 replica bootstrap analysis. Extant archaeplastidal candidates whose ancestors may have triggered the establishment of the symbiotic link are those enzymes that can be linked phylogenetically to ADP-Glc–requiring enzymes conserved in green algae and land plants (in green), which group into three classes known as granule-bound starch synthase (GBSS in the bottom part of figure), the soluble starch synthase (SS) SSI/SSII class (in the bottom part of the figure), and the SSIII/SSIV class (in the top and middle parts of the figure; reviewed in Ball and Morell (2003). The high bootstrap nodes unifying the GBSS-SSI-SSII and the SSIII-IV clades are highlighted in bold. Rhodophyceae starch synthase progenitors are not present in this tree because those organisms have lost the ability to polymerize starch from ADP-Glc and have retained only the very different host-derived soluble glycogen (starch) synthase that uses UDP-Glc (reviewed in Ball et al., 2011). However, GBSSI was transmitted to red algae (in red) and Glaucophyta (in dark blue) where it polymerizes amylose from UDP-Glc in the cytosol. These proteins still display the ability to use ADP-Glc. The source of the GBSSI, SSI, and SSII clade is proposed to be cyanobacterial and is displayed by an arrow (for reasons sustaining this source, see text and Deschamps et al., 2008a; Ball et al., 2011). As to the SSIII/SSIV clade, SSIII/SSIV was previously reported to result from an HGT from environmental chlamydia-like amoeba parasites P. amoebophila and P. acanthamoeba (Moustafa et al., 2008). However, the phylogeny displayed here suggests the presence of several LGT events not previously evidenced. Because of the importance of the issue, we devoted the Supplemental Methods 1 online to the detailed analysis of these results.

This leaves the ancestor of the SSIII-SSIV clade as the only extant enzyme whose ancestor could have been recruited at the onset of plastid endosymbiosis. This possibility is further strengthened by the finding of a homologous sequence in the genome of the glaucophyte Cyanophora paradoxa where it appears to be involved in starch synthesis using ADP-Glc in the cytosol (Price et al., 2012).

Previously, SSIII-IV was reported as an enzyme of chlamydial origin from examination of the sequence of the amoebal parasite Protochlamydia amoebophila (Moustafa et al., 2008). However, inclusion of the more diverged enzymes from animal parasites and the new sequence information from additional genomes suggests a more complex SSIII-SSIV phylogeny (Figure 2; see Supplemental Figure 1, Supplemental Table 2, and Supplemental Methods 1 online for the complex phylogeny of the glucan synthase). Our results suggest strongly that the source of the entire SSIII-IV enzyme clade (inclusive of a small subset of proteins from proteobacteria and cyanobacteria) was an enzyme form that evolved in chlamydial ancestors. The phylogeny is complicated further by the presence of several distinct LGT events. Indeed, an additional LGT from Chlamydiales to proteobacteria is evidenced and well supported by the phylogenetic evidence. This sequence was subsequently passed on from proteobacteria to a subset of cyanobacteria. In all these bacteria, SSIII-IV is one of several glycogen synthase types present, while Chlamydiales define the only group of organisms that consistently have the SSIII-IV enzyme as their sole glucan synthase. The phylogeny is in agreement with a chlamydial source and its transfer from the parasites to a subset of proteobacteria and from there to a subset of cyanobacteria. In addition, the Archaeplastida have also received a copy of a gene encoding the chlamydial enzyme. However, erosion of phylogenetic signal and low bootstrap support does not allow one to distinguish between direct transfer of SSIII-IV from Chlamydiales from indirect transfer via a proteobacterial intermediate. Nevertheless, cyanobacteria can be safely ruled out as potential donors. This complexity suggests that the chlamydial enzyme probably harbors an unusual biochemical property responsible for its wide dissemination and success in the bacterial world.

Bioinformatic Effector Predictions

Why would a chlamydial glucan-synthase be present in the host cytosol at the onset of plastid endosymbiosis? A clue is provided by the fact that all chlamydia-like pathogens are known to secrete a large number of protein effectors into their host cytosol. Hence, if a particular effector played a key role in establishing symbiotic carbon flux, it would have engaged the parasite, the cyanobiont, and the host in a tripartite symbiosis, each genome being required for the coding of enzymes establishing the common essential carbon flux. Virulence effectors are secreted by Chlamydiales through the type III secretion system that has been present for >1 billion years in chlamydia-like parasites that infect protists (Horn et al., 2004; reviewed in Horn, 2008). This syringe-like macromolecular machine, known as the type three secretion system (TTS), spans the two bacterial membranes and the inclusion membrane of the host that surrounds the parasites. Hence, if enzymes of chlamydial glycogen metabolism were being secreted into the host cytosol, they could have participated in the key biochemical events that established plastid endosymbiosis.

It is noteworthy that the majority of Chlamydiales contain a full suite of enzymes for glycogen metabolism (Iliffe-Lee and McClarty, 2000), even though very few appear to produce glycogen autonomously or to accumulate the polymer internally. This is in contrast with the majority of intracellular pathogens that have lost the enzymes of glycogen metabolism (Henrissat et al., 2002). The relevant enzymes include ADP-Glc pyrophosphorylase (AGPase; GlgC), glycogen synthase (GlgA), glycogen branching enzyme (GlgB), one or more glycogen phosphorylases (GlgP and MalP), a GlgX-type of direct debranching enzyme related to plant isoamylases, and an amylomaltase required to metabolize the oligosaccharides released through GlgX (MalQ). Because the genes encoding these enzymes of glycogen metabolism have been retained in the highly streamlined and reduced genomes of the Chlamydiales (1 Mb for the animal parasites; Horn, 2008), they likely serve an important function in these organisms. One possibility is that some enzymes in this pathway are cytosolic effectors that manipulate host glycogen metabolism.

Based on the proposed relationship between cytosolic glycogen metabolism and establishment of the endosymbiotic link, we used four TTS prediction algorithms (Arnold et al., 2009; Löwer and Schneider, 2009; Samudrala et al., 2009; Wang et al., 2010) to study all the enzymes of glycogen metabolism identified in the genome sequences of environmental Chlamydiales. We focused on the amoebal parasites P. amoebophila and Parachlamydia acanthamoeba because these taxa are the most frequently recovered sisters to archaeplastidial enzymes of chlamydial origin (Moustafa et al., 2008). The results (see Supplemental Table 3 online), summarized in Table 2, show that the isoamylase encoded by GlgX was found by all prediction programs to be a putative effector, AGPase, the enzyme that synthesizes ADP-Glc (encoded by GlgC) was scored as a potential effector in either P. amoebophila or P. acanthamoeba, respectively, by the PETS and effective T3 programs. Glycogen phosphorylase in P. acanthamoeba was identified as a candidate effector by the effective T3 program, whereas MalQ in P. amoebophila was identified as a potential effector by both the SIEVE and PETS programs.

Table 2. Bioinformatic Predictions and in Vivo Validation of Chlamydial Putative Effectors.

Protein	BPBAac	SIEVE	PETS	Effective T3	Secretion Assay in Shigella
P. amoebophila GlgA	−	−	−	−	Not tested (highly conserved)
P. acanthamoeba GlgA	−	−	−	−	+
P. amoebophila GlgB	−	−	−	+	−
P. acanthamoeba GlgB	−	−	+	-	−
P. amoebophila GlgC	N/A	N/A	N/A	N/A	N/A
P. acanthamoeba GlgC	−	−	+	+	+
P. amoebophila GlgP	−	−	−	−	Not tested
P. acanthamoeba GlgP-1	−	−	−	+	Chimera not expressed
P. acanthamoeba GlgP-2	−	−	−	−	+
P. amoebophila GlgX	+	+	+	+	+
P. acanthamoeba GlgX	N/A	N/A	N/A	N/A	N/A
P. amoebophila MalQ	−	+	+	−	+
P. acanthamoeba MalQ	−	−	−	−	−

The different putative effector prediction programs were used respectively with a cutoff of 0.4 (PETS), a SIEVE percentage score above 1, and an effector T3% score above 90% (Arnold et al., 2009; Löwer and Schneider, 2009; Samudrala et al., 2009; Wang et al., 2010). N/A refers to the presence of too many uncertainties on the translation start site to allow prediction or in vivo testing. We were unable to test GlgX from P. acanthamoeba and GlgC from P. amoebophila because of uncertainties in the reported N-terminal sequence deduced from the genome sequence. Because the N-terminal sequence of GlgA from P. acanthamoeba and Protochlamydia was highly conserved, we chose to test only the Parachlamydia sequence. GlgA, glycogen synthase; GlgB, glycogen branching enzyme; GlgC, ADP-Glc pyrophosphorylase; GlgX, debranching enzyme (isoamylase-like); MalQ, α-1,4 glucanotransferase (DPE2 maltase type).

In Vivo Validation of Bioinformatic Predictions

To test these bioinformatic predictions, we relied on a previously established in vivo assay in Shigella flexneri (Subtil et al., 2001). This assay relies on the conservation of the structural determinants of the TTS within bacteria. S. flexneri is used to establish the presence of an N-terminal sequence that allows secretion through the TTS of a fused reporter gene, the calmodulin-dependent adenylate cyclase reporter from Bordetella pertussis. This system has been validated in screens for TTS in Chlamydia pneumoniae with a rate of false positives below 5% (Subtil et al., 2005). To test whether proteins of glycogen metabolism of P. acanthamoeba and P. amoebophila contained a functional TTS signal, we made fusions between the first 30 codons from each gene and the reporter gene and tested secretion of the reporter by S. flexneri (see Supplemental Table 4 online). The results shown in Supplemental Table 4 and Figure 3 not only confirm that GlgX (debranching enzyme) and GlgC/GlgP (ADP-Glc pyrophosphorylase and glycogen phosphorylase) are recognized as TTS substrates, but they also show that GlgA (glycogen synthase) carries a secretion signal recognized by S. flexneri. As a negative control for the secretion assay, we deleted the first 10 amino acids of GlgA and showed that the corresponding fusion protein was no longer secreted through the TTS (Figure 3).

Figure 3.

Chlamydial Enzymes Encode Functional TTS Signals.

The N-terminal 30 codons (see Supplemental Table 4 online) of the indicated genes were cloned upstream of cya and expressed in the ipaB and mxiD S. flexneri strains. The ipaB strain constitutively secretes TTS substrates (labeled TTS+), whereas the mxiD strain is defective for TTS (labeled TTS−) (Subtil et al., 2001). Exponential cultures expressing the fusion protein were fractionated into supernatants (S) and pellets (P). Samples were resolved by SDS-PAGE gels, transferred to a polyvinylidene difluoride (PVDF) membrane, and probed with anti-Cya to detect the chimera. Probing the membrane with anti-IpaD showed that TTS in the ipaB background was not impaired by transformation of the various constructs. Antibodies against the cAMP receptor protein (CRP) were used to control for bacterial lysis during fractionation. A GlgA construct in which the first 10 amino acids were deleted (Δ10GlgA/Cya) was included as a negative control. When expressed in the ipaB strain, the Δ10GlgA/Cya chimera was not secreted in the supernatant, in contrast with the GlgA/Cya chimera. In the mxiB (TTS defective) background, none of the chimeric proteins were recovered in the culture supernatant, demonstrating that their secretion in the ipaB background occurred by a TTS mechanism.

Consequences of the Effector Nature of the Enzymes of Glycogen Metabolism of Chlamydiales on the Priming of Plastid Endosymbiosis

The bioinformatic predictions and experimental evidence implied that combinations of parasite enzymes could work in combination to hijack host metabolism. Chlamydia-like pathogens potentially use ADP-Glc pyrophosphorylase, glycogen phosphorylase, and glycogen synthase to increase flux to glycogen in the cytosol of their hosts and recover maltotetraose through the action of the GlgX-type of glycogen debranching enzyme. ADP-Glc is an energy-rich, bacterial-specific metabolite, unrecognized by eukaryotes, whose costly synthesis would be induced in the cytosol by these putative chlamydial effectors (Figure 4), likely thereby weakening host defense against the pathogen. Maltotetraose is normally not produced by eukaryotic cytosolic glycogen metabolism, and no MOS malto-oligosaccharides degrading enzyme other than the eukaryotic dpe2 type of amylomaltase is likely to be present in this compartment (although glucosidases have been described in other eukaryotic compartments). Hence, the bacterial effectors would induce the synthesis of glycogen through ADP-Glc at the beginning of the infection cycle when the ratio of cytosolic ATP to Pi is high. At later stages when this ratio decreases, the presence of the unregulated glycogen phosphorylase of the chlamydial pathogen will bypass the highly regulated host enzyme and trigger conversion of the cytosolic glycogen pools to Glc-1-P and a carbon substrate (maltotetraose) that can be metabolized only by the pathogen. It is relatively straightforward to envision how introduction of a cyanobacterium into the host via phagocytosis would rapidly rescue a cell infected by such a pathogen and effectively convert the latter to a symbiont (Figure 5). The cyanobacterium would generate large amounts of ADP-Glc as a product of photosynthetic metabolism. This high-energy compound could flood the host metabolic system with substrate to fuel flux through glycogen and into the parasite, without draining host energy resources. Only one critical event would be required: recruitment of a host-derived nucleotide sugar translocator to the inner envelope of the cyanobacterium. Extant eukaryote transporters are able to facilitate movement of ADP-Glc across membranes even though they do not use that particular nucleotide sugar (Colleoni et al., 2010). The endosymbiotic link would thus be established within a tripartite system without the need for a novel protein or a HGT event followed by foreign gene expression in the host. Under this “ménage à trois” scenario, just two chlamydial functions are required to establish symbiosis: the glucan synthase that feeds the exported carbon into the glycogen pool (GlgA) and the GlgX debranching enzyme that generates the maltooligosaccharides required to feed the chlamydial symbionts, thereby maintaining its genes required for symbiosis. Hence, all other chlamydial glycogen metabolism effectors (GlgC, GlgP, and MalQ) may have been counter-selected after establishment of the tripartite symbiosis as they were no longer necessary following the switch from pathogenesis to symbiosis. Indeed, when the system had switched to symbiosis, it was no longer necessary for the chlamydia to tap ATP and Glc-1-P from host metabolism through the use of its GlgC (AGPase) effector, and there would likely be selection against unregulated breakdown of cytoplasmic glycogen pools by the chlamydial phosphorylase (GlgP). In agreement with this observation, the phylogenetic origin of other components of storage polysaccharide breakdown in extant Archaeplastida is of host (Eukaryotic) derivation (Deschamps et al., 2008a).

Figure 4.

Glycogen Metabolism Putative Effectors in Extant and Ancient Chlamydiales.

The infection cycle of a typical eukaryote by a chlamydia-like organism is shown. Chlamydiales are internalized through phagocytosis. The pathogen modifies the surrounding vesicle into an inclusion membrane (shown in blue). These bacteria undergo two successive developmental forms: the infectious elementary body (EB) and the replicative reticulate body (RB). At the end of the cycle, reticulate bodies differentiate back to elementary bodies that are eventually released from the cell. Both elementary bodies and reticulate bodies possess a type III secretion apparatus that spans the two bacterial membranes and the inclusion membrane, allowing for the secretion of bacterial proteins directly into the host cytosol. In the enlargement, the inclusion membrane is depicted in blue, type III secretion apparatuses in pink, the eukaryote cytosol in yellow, and the inside of the chlamydial cell in white. The Glc residues of the host glycogen used as primers for synthesis are symbolized by black dots. Red dots represent accessible Glc, whereas the Glc-1-P generated by glycogen phosphorylase is represented by red dots with an attached “P.” The sequential nature of the biochemical reactions is depicted as numbers 1 to 5. The enlargement displays the particular type of putative effector combination present in extant P. acanthamoeba and possibly in the ancient parasite that would have predisposed a host cell to endosymbiosis. Briefly, carbon is diverted to the glycogen pools through the chlamydial ADP-Glc pyrophosphorylase putative effector (GlgC) making use of the high host cytosolic ATP and Glc-1-P pools at the beginning of the infection cycle. Polymerization into glycogen from ADP-Glc occurs through the chlamydial glycogen synthase (GlgA). The host glycogen synthase only uses UDP-Glc and is not represented. Branching could occur through the action of either the host or parasite branching enzyme, although we did not find clear evidence for a type III secretion signal in the two parasite branching enzymes tested (GlgB), indicating that branching might solely involve host enzymes. When the orthophosphate concentration rises and the cytosolic ATP and Glc-1-P decreases as a consequence of the infection, parasite phosphorylase (GlgP) will recess the outer chains of glycogen, terminating four residues away from each branch. This will allow the action of GlgX, whose possible substrates are restricted to such chains. The maltotetraose generated by GlgX is normally not metabolized by eukaryotes in the cytosol and could therefore be a substrate for import and catabolism within the parasites. The transporter for the import of MOS in the parasite is presently unknown. We omitted MalQ from the drawing since we do not know if the chlamydial version of this enzyme would behave more like a maltase or disproportionate type of α-1,4 glucanotransferase.

Figure 5.

Ménage à Trois.

The interdependent symbiosis between the cyanobiont, its host, and a chlamydial parasite is shown, displaying an effector combination akin to that hypothesized in Figure 4. Enzymes are colored with respect to their phylogenetic origin: in orange for host enzymes of eukaryotic glycogen metabolism, in pink for chlamydial effectors, and in blue for cyanobacterial enzymes. Only those enzymes of chlamydial origin that are required to establish the symbiotic flux and feed carbon into the parasite or the host are displayed. The other enzymes depicted in Figure 4 do not interfere with this biochemical flux. The GlgX chlamydial debranching enzyme is represented here by the name “iso” to denote its present function in Archaeplastida (isoamylase). The cyanobiont in blue exports the bacterial specific metabolite ADP-Glc by recruiting a family III nucleotide sugar translocator (colored in orange) from the host endomembrane system (colored in yellow) as recently proposed (Weber et al., 2006; Deschamps et al., 2008a; Colleoni et al., 2010). A possible primitive TOC (translocon of the outer chloroplast membrane) targeting system is drawn in gray. The ADP-Glc would be funneled to glycogen by the chlamydial effector glycogen synthase, whereas the host would still be able to polymerize Glc into glycogen from UDP-Glc. The dramatic increase of the cytosolic glycogen pools would have benefitted the host through the increased production of Glc and Glc-1-P that can be further metabolized. However, it would have equally benefitted the parasite by increasing the supply of MOS that may only be metabolized by the latter. BE, branching enzyme; iDBE, indirect debranching enzyme; ISO, glgX type of direct debranching enzyme; PHO, phosphorylase; SS-ADP, ADP-specific glycogen synthase; SS-UDP, UDP-specific glycogen synthase.

Isoamylase (GlgX) Defines a Chlamydial Effector Common to all Three Archaeplastidial Lineages

The isoamylase gene of Archaeplastida may have been derived from the chlamydial parasite’s GlgX protein (Brinkman et al., 2002; Huang and Gogarten, 2007; Becker et al., 2008; Moustafa et al., 2008). The Chlamydial protein has been reported to display structural properties analogous to the Escherichia coli GlgX protein, which is a maltotetraose-generating enzyme in glycogen catabolism (Jeanningros et al., 1976; Dauvillée et al., 2005). Among the plant isoamylase genes, all three isoamylases display a common origin (Figure 6; Abe et al., 1999; Hussain et al., 2003). Therefore, the anabolic function of isa1 (and isa2) in amylopectin synthesis likely evolved at a later stage, postendosymbiosis. As a result, we propose that glycogen accumulated in the ancestral symbiosis (Figure 5). In the phylogeny of GlgX, glycogen debranching enzymes (Figure 6; see Supplemental Figure 2 online), the enzymes from Archaeplastida and Chlamydiales, form a well-supported (maximum likelihood bootstrap support of 83%) monophyletic group to the exclusion of all other bacteria. This suggests a HGT event between these two lineages. The direction of gene transfer can be inferred from existing knowledge of storage polysaccharide metabolism. One of the major differences between bacterial and eukaryotic glycogen metabolism lies in the mode of hydrolysis of the α-1,6 linkages (Ball et al., 2011). Debranching enzymes that attack the α-1,6 branch directly and release debranched glucan chains are found only in bacteria and Archaea. Eukaryotes use the indirect debranching enzymes that produce Glc. No eukaryotes, with the notable exception of Archaeplastida, contain a direct type of debranching enzyme, such as GlgX or isoamylase. In addition, a candidate sequence encoding an indirect debranching enzyme was recently found in the C. paradoxa (Glaucophyta) genome, strongly suggesting that the ancestral host of the primary endosymbiosis contained such an enzyme, enabling eukaryotic glycogen catabolism in the cytosol (Price et al., 2012). From these observations, we conclude that the direction of gene transfer was from Chlamydiales to the common ancestor of the Archaeplastida. However, the relative positions of the different Archaeplastida are unresolved in the tree. In particular, the node grouping the Chloroplastida has low bootstrap support, as is the case for other basal nodes in the Archaeplastida.A possible explanation for this lack of resolution is the complex evolutionary history of the Chloroplastida with respect to debranching enzymes. We propose that the chlamydial gene was initially used in the cytosol of the Archaeplastida common ancestor but later retargeted to the plastid in Chloroplastida (Deschamps et al., 2008b). Therefore, acceleration of sequence evolution likely happened twice in Chloroplastida and only once in the Rhodophyta and Glaucophyta. This could have led to a stronger erosion of the phylogenetic signal in the genes of Chloroplastida. Similar problems exist in the more complex GlgA phylogeny and in the GBSSI-SSI-SSII clade (see Supplemental Figure 1 online).

Figure 6.

Phylogenetic Analysis of Bacterial and Archaeplastidal Glycogen Debranching Enzymes.

Starch debranching enzymes of Chloroplastida are known to play an important role both in polysaccharide synthesis and starch degradation. The maximum likelihood tree of these enzymes shows the three types of Chloroplastida (land plants and green algae) subunits (filled green triangles). Isoamylase1 (Isa1) is responsible for trimming misplaced preamylopectin branches that if unprocessed prevent polysaccharide crystallization, leading to glycogen formation, rather than starch. Isa2 is a noncatalytic subunit assembled in heteromultimers together with Isa1. Isa3 is involved in the breakdown of branches during starch degradation. The homologous enzymes of red algae (red filled triangle) and Glaucophyta (blue branches) are also shown in this tree. Analogous specialized roles in starch synthesis and degradation remain to be determined in these organisms. The novel glaucophyte sequences were isolated using RT-PCR. Alignments are shown in Supplemental Data Sets 3 to 5 online, and the full tree is displayed in Supplemental Figure 2 online.

The ménage à trois tripartite symbiosis hypothesis predicts that only two of the glycogen metabolism effector enzymes from the parasites were required to prime endosymbiosis. These are the two chlamydial HGTs that are observed in the storage polysaccharide metabolism network of Archaeplastida. These define one of the four (five if the nucleotide transporter (NTT) is taken into account) established cases of chlamydial HGTs in all three archaeplastidal lineages and one of the 10 to 11 cases of HGTs common to two out of the three archaeplastidal lineages. The possibility that it is by coincidence that these two cases would simultaneously define the two critical components of symbiotic carbon flux and the two cases of documented common HGTs is remote, and we believe this to be as close as possible to experimental demonstration of the tripartite symbiosis hypothesis.

DISCUSSION

The Chlamydial Imprint on the Genomes of Archaeplastida

Forty-eight genes have been found in this study that are likely candidates for HGT from Chlamydiales to the common ancestor of the Archaeplastida. However, the nature of the filters used in our phylogenomic analysis have systematically excluded genes that do not satisfy the criterion of monophyly of Chlamydiales and Archaeplastida (and their secondary endosymbiosis derivatives) proteins to the exclusion of all other clades (at bootstrap 75 or higher). Genes displaying more complex phylogenies, such as the glucan synthase gene GlgA discussed here, have been excluded. Interestingly, the paradigm of a parasitic function transferred to Archaeplastida (the ATP import protein NTT of that gave the name “energy parasites” to the Chlamydiales) would also have been excluded because of additional complexities in the phylogeny of these transporters and restricted gene distribution. Hence, the impact of Chlamydiales on the plant genomes may still be underappreciated. This impact of on plant genomes seems modest (50 genes) and appears restricted when compared with that of Cyanobacteria (∼1000 genes). Yet, this figure is of the same order of magnitude as the number of genes from Cyanobacteria that encode proteins not targeted to plastids and that are neither involved in photosynthesis nor in organelle maintenance and genome expression (the so-called nonplastid functions; Reyes-Prieto et al., 2006; Moustafa and Bhattacharya, 2008). In our view, the chlamydial symbiont compartment was likely not the site of an essential biochemical pathway and was maintained only because it encoded useful genes for the ménage à trois. Once all possible chlamydial genes had been transferred to the host, the compartment disappeared together with the chlamydial genome. It is well documented that eukaryotic lineages that have lost a plastid, have, as a consequence, lost the majority of their genes of cyanobacterial origin. Indeed, the nuclear genome of Cryptosporidium parvum, an apicomplexa parasite that has lost the apicoplast, has only five genes of possible cyanobacterial (two genes) or plant ancestry (three genes) (Huang et al., 2004). In this context, the maintenance of a 50 gene signatures in the Archaeplastida genomes underlines the biochemical importance of chlamydial gene products in the maintenance of the symbiosis.

The Ménage à Trois Allowed Unprecedented Levels of Metabolic Integration

After metabolic stabilization of the tripartite system, a great deal of subsequent evolution, including extensive HGT, was necessary before reaching the true photoautotrophic eukaryote that was the direct ancestor of extant archaeplastidial lineages. During that time, enzyme functions derived from the host, the cyanobiont, and the chlamydial symbiont were likely integrated and optimized through a combination of HGTs and novel effector evolution in the cytosol or elsewhere, including the membranes and stroma of the cyanobiont. The continued presence of a functional, temperate chlamydial symbiont would have been required for so long as the chlamydiae encoded useful effector genes. However, HGTs to the host nucleus of these effector genes would have rendered the chlamydial partner dispensable, leading ultimately to loss of the symbiont.

We can therefore expect that, as the ancestral archaeplastidial lineages diversified, some lineages lost their chlamydial symbiont early on because the symbiotic effector genes had all been transferred to the host nucleus. However, the switch from pathogenesis to symbiosis would have prompted the positive selection of mutations of the N termini of chlamydial proteins turning them into new effector proteins. This would provide an opposing force to parasite loss as a result of HGT of effector genes to the nucleus. In some lineages, the generation of novel effectors would have kept ahead of the HGT process and ensured that the chlamydial symbiont was maintained by natural selection. It is plausible that such lineages would have been at a selective advantage because they displayed higher levels of metabolic integration. This metabolic integration may in part explain how the cyanobiont reached the status of a true cellular organelle. The race between evolution of new effectors and HGT to the nucleus also explains why different archeaplastidial lineages may not necessarily display the same number and type of chlamydial HGTs. Another process that may have altered the pathogen’s signature in extant genomes consists of selective gene losses experienced by the three major archaeplastidial lineages. Of the three lineages, Rhodophyceae may be the clade that has experienced the greatest gene losses as it diverged from the other two archaeplastidial groups, while Glaucophyta may be the lineage that has maintained the highest number of genes from the ancestral protist. Starch metabolism offers an interesting example of this. For instance, Rhodophyceae have lost the ability to synthesize starch from ADP-Glc and as a consequence have lost the ancestral effector gene that established the tripartite symbiosis, while this gene was retained both in Chloroplastida and Glaucophyta. The combination of gene losses upon separation of the archaeplastidial lineages and of different timing of loss of the chlamydial symbiont can easily account for the pattern of chlamydial HGTs observed in Archaeplastida.

Our finding of at least 50 genes from Chlamydia (including the 48 shown in Table 1 and the GlgX debranching enzyme, the GlgA SSIII-SSIV, the plastid ATP translocator [NTT], and UhpC; Huang and Gogarten, 2007; Moustafa et al., 2008; Price et al., 2012) in algae and plants is consistent with long-term residency of the chlamydial symbiont in the archaeplastidial ancestor. This would provide continued selection for cyanobiont photosynthetic function to ameliorate deleterious effects of the energy parasite on host cell metabolism. Conversely, the photosynthetic function from the cyanobiont to provide abundant high-energy compounds to the tripartite system and its gradual modification to generate starch rather than glycogen would have significantly improved this partnership.

A surprising implication of our hypothesis is that the primary plastid endosymbiosis was not restricted to the host and cyanobacterial partners, nor based solely on the acquisition by the host of photosynthetic function (Huang and Gogarten, 2007; Becker et al., 2008; Moustafa et al., 2008). Rather, photoautotrophy in eukaryotes could have resulted from an increased tolerance to a particular chlamydial infection, thereby transforming an interaction from pathogenic to symbiotic (Dale et al., 2002; Wernegreen, 2004). A second aspect of our hypothesis is that the advantage conferred by endosymbiosis arose immediately after the capture of the cyanobacterium, rather than after extensive endosymbiotic gene transfer and gene activation had occurred in the host nucleus (Bodyl et al., 2007). It is noteworthy that the presence of an additional partner in endosymbiosis with the ability to secrete useful proteins in the cytosol not only expanded the repertoire of genes available for metabolic integration and production of a new organelle, but it greatly facilitated the process by making key elements available in the cytosol of the host. These factors might explain the singularity of the ancient plastid endosymbiosis (excluding Paulinella) (Nowack et al., 2008) that fundamentally changed the evolutionary trajectory of life on our planet.

METHODS

Phylogenomic Methods and TTS Prediction Analysis

The complete set of Chlamydiales proteins (28,244 proteins as of January 2011) available from RefSeq (Pruitt et al., 2007) was phylogenomically processed using iTree (Moustafa et al., 2010) against a comprehensive database, which was assembled from the complete RefSeq proteins and additional algal and microbial genomes from the Joint Genome Initiative. We also supplemented the database with recently published data from key lineages, such as red algae, dinoflagellates, and the glaucophyte Cyanophora paradoxa (Price et al., 2012). The generated phylogenetic trees were searched for trees that indicated sister relationship between Chlamydiales and Archaeplastida using PhyloSort (Moustafa and Bhattacharya, 2008). Only trees with bootstrap values ≥75% were considered in subsequent analyses. The four methods of bioinformatic prediction of TTS signals are detailed elsewhere (Arnold et al., 2009; Löwer and Schneider, 2009; Samudrala et al., 2009; Wang et al., 2010).

Type III Secretion Assays

The first 30 codons of the genes under study were synthesized (Life Technologies) for cloning into puc19cya as described (Subtil et al., 2001). In three cases, the codon that initiates the translation could not be unambiguously identified, and more than 30 codons were included (Parachlamydia acanthamoeba GlgX, Protochlamydia amoebophila GlgP1, and α-1,4 glucano-transferase; see Supplemental Table 3 online). Constructs were made in Escherichia coli strain TG1, verified by sequencing, and transformed in Shigella flexneri strains SF401 and SF620 in which the mxiD and ipaB genes, respectively, have been inactivated (Subtil et al., 2001). Analysis of secretion in these strains was performed as described previously (Subtil et al., 2001). Antibodies against CRP, a cytosolic marker, were used to estimate the contamination of supernatant fractions with bacterial proteins as a result of bacterial lysis. Antibodies against IpaD, a type III secreted protein of Shigella, were used to verify that type III secretion occurred normally in the transformed ipaB strains.

Sequence Mining and Phylogenetic Tree Reconstruction

Amino acid sequences from P. amoebophila were first used as queries to look for similar protein sequences using BLASTP (Altschul et al., 1997) against a local genome database containing ∼300 bacterial and ∼60 eukaryotic complete annotated genomes. All homologous sequences (with an e-value inferior to 1.0e-05) were collected from each BLAST hit list and aligned together using the MAFFT software (Katoh et al., 2002). After automatic gaps removal in the resulting multiple alignments, maximum likelihood tree reconstructions were done using the software FastTree (Price et al., 2009). Trees were manually inspected to remove possible sequences duplicates, to refine sequence sampling, and to add additional sequences of interest that were not part the original set used for BLASTP. Trees were then realigned and recomputed with the TREEFINDER software (Jobb et al., 2004) using the LG + Gamma model. Protein sequences where collected from GenBank. Experimental trees (see Supplemental Figure 1 online) and candidate HGT trees (see Supplemental Data Set 1 online) were run with 100 bootstrap replicates. For the GlgA and GlgX trees (Figure 1; see Supplemental Figure 2 online), we used the ProtTest 2.4 to identify the best-fit model of protein evolution for our data (the LG + I + G + F model) and applied 1000 bootstrap replicates (http://www.ncbi.nlm.nih.gov/protein), the Joint Genome Institute (http://www.jgi.doe.gov/) and on genome home project websites for Galdieria sulphuraria (http://genomics.msu.edu/galdieria/), Cyanidioschyzon merolae (http://merolae.biol.s.u-tokyo.ac.jp/), and Arabidopsis thaliana (http://www.Arabidopsis.org/). Isoamylase sequences from C. paradoxa and GBSS sequences from unidentified Cyanobacterium Clg1 (GenBank taxonomy ID 197335) were produced in this study. The analysis of isoamylases from P. acanthamoeba showed that the presence of two open reading frames in the genome was due to an annotation error. We therefore decided to fuse these two open reading frames together.

Cloning of Isoamylase Sequences from Glaucophyta

mRNA were purified from 300 mL of exponential growth culture of C. paradoxa using the FastTrack 2.0 kit (Invitrogen). Reverse transcription reactions (200 ng of mRNA) were performed with SuperScript III and in the presence of GeneRacer oligo(dT) following the recommendations of the GeneRacer kit (Invitrogen). 3′ Rapid amplification of cDNA ends reactions were then performed using Platinum Taq DNA polymerase High Fidelity (Invitrogen) in the presence of 2% DMSO. Forward primers were designed according to available nucleotide sequences of contig 40664 (5′-CATCACCACACGAACTACACGGGGTGCGGGAACACGGTCAA-3′), contig 12528 (5′-GTCCGCGAGTTCAAGGAGAT-3′), and contig 38691 (5′-GTGCACGAGTTCAAGACGATGGTGCGGGAGCTGCACA-3′). PCR products were cloned in TOPO cloning vector (Invitrogen) and sent for sequencing.

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure 1. Phylogenetic Analysis of the SSIII-IV ADP-Glc–Specific Glucan Synthase.

• Supplemental Figure 2. Phylogenetic Analysis of the GlgX-Isoamylase Glycogen (Starch) Debranching Enzymes.

• Supplemental Table 1. The Number of Protein Sequences in the Database That Was Used for the Phylogenomic Analysis to Uncover the Origins (Based on Phyla) of Chlamydiales Proteins.

• Supplemental Table 2. Distribution of GT5-Glucan Synthases among SSIII-IV Containing Bacterial Species.

• Supplemental Table 3. Bioinformatic Analysis of Candidate Chlamydial Effectors.

• Supplemental Table 4. Sequence of the Inserts Cloned in the puc19cya Vector.

• Supplemental Methods 1. The Complex Phylogeny of the SSIII-SSIV Clades.

• Supplemental Data Set 1. Candidate HGT Trees.

• Supplemental Data Set 2. GlgA Alignments.

• Supplemental Data Set 3. GlgX Alignments.

• Supplemental Data Set 4. GlgA Alignments.

• Supplemental Data Set 5. GlgX Alignments.