A microbial eukaryote with a unique combination of purple bacteria and green algae as endosymbionts

A eukaryote with two contrasting intracellular photosymbionts has a unique combination of physiological properties.


INTRODUCTION
Symbioses between heterotrophs and photosynthetic algae (photosymbiosis or acquired phototrophy) are widespread and ecologically important in today's ecosystems (1)(2)(3)(4). Many eukaryotes have thus become mixotrophs, i.e., they combine predation and photosynthesis, by acquiring algal endosymbionts or chloroplasts from algae. Photoautotrophic endosymbionts often release photosynthetic products (e.g., sugars, organic acids, and oxygen) into their hosts, whereas hosts provide a nutrient-rich environment (e.g., nitrogen and minerals) and physical protection against predators and viruses (5-7). Today, virtually all known photosymbioses involve cyanobacteria or eukaryotic algae that perform oxygenic photosynthesis (1,2), except for the marine ciliate Strombidium purpureum, whose purple bacterial endosymbionts remain uncharacterized (see Supplementary Text for details) (8,9). Only one eukaryote has ever been described as harboring both green algae and purple bacteria in the same host cell (see below) (10,11). This is a puzzling combination as these two groups of photosynthesizers typically occupy different niches in nature. Green algae are aerobic and oxygenic, whereas purple bacteria are primarily anaerobic and anoxygenic (12). Moreover, green algae and purple bacteria absorb distinct wavebands because of the different photosynthetic pigments [e.g., the absorption peak of chlorophyll a is at ~600 to 700 nm and those of bacteriochlorophyll a at ~800 to 900 nm; (13,14)]. It is thus unknown how two endosymbionts with these contrasting photosynthetic physiologies came to coexist in the same host cytoplasm and how this complex symbiotic consortium makes a living in its environment.

Pseudoblepharisma tenue is a phagotroph with two photosynthetic endosymbionts
We found populations of a purple-green ciliate dwelling in the Sphagnum ponds of the Simmelried moorland (Constance, Germany), a "hotspot" for microbial diversity (Fig. 1, fig. S1, and movie S1) (15). This ciliate most closely resembled P. tenue KAHL, which was briefly described in 1926 as a very uncommon ciliate (10) and since then never studied in detail (11,16). The characteristic purple-green color of P. tenue is conferred by two different endosymbionts: purple bacteria and green algae (Fig. 1, A to C, and fig. S2, A to F). The presence of purple bacteria and green algae dividing in the cytoplasm of P. tenue suggests that both endosymbionts are inherited in a vertical fashion (Fig. 1, C and D). Moreover, food vacuoles with partially digested contents, which clearly differ from degraded endosymbionts, also suggest that P. tenue is capable of phagocytosis (Fig. 1B). The major morphological features of this ciliate-such as cell body shape and length (182.5 m on average, n = 6; table S1), posterior contractile vacuole (Fig. 1A), macro-and micronuclei (Fig. 1E), peristome and associated structures (Fig. 1F), and cortical granules (putative mucocysts; Fig. 1G)-all agree with the original description of P. tenue (10,11) and suggest an affiliation to the class Heterotrichea (Ciliophora). Unlike Blepharisma species, P. tenue lacks a conspicuous flap-like undulating membrane alongside the peristome and thus rather resembles Spirostomum species (Fig. 1F); however, Spirostomum species are often larger and more contractile than P. tenue. In addition, we also observed spherical cells of P. tenue that potentially are thin-walled resting cysts (Fig. 1H).

P. tenue is a heterotrich ciliate inhabited by a known Chlorella strain and a new purple sulfur bacterium
To determine the phylogenetic affiliation of each symbiotic partner, we relied on a single-cell polymerase chain reaction (PCR) approach; attempts to establish cultures of P. tenue and its endosymbionts were unsuccessful. Phylogenetic analyses of the ribosomal RNA (rRNA) gene operon and the mitochondrial COI gene revealed that P. tenue is most closely related to ciliates of the genus Spirostomum ( Fig. 2A and fig. S3A). All Spirostomum species are purely heterotrophic predators, except for Spirostomum semivirescens, which harbors zoochlorellae (17)(18)(19); P. tenue, however, was phylogenetically distant from S. semivirescens in our phylogenetic trees ( Fig. 2A and  fig. S3A). Although P. tenue has a clear affinity to Spirostomum, it remains to be determined whether it is sister to the genus Spirostomum or represents a new lineage within it ( Fig. 2A and fig. S3A).
The algal endosymbionts are most closely related to Chlorella species (Trebouxiophyceae) according to phylogenetic analysis of the 18S rRNA-ITS2 and rbcL genes ( Fig. 2B and fig. S3B). This is consistent with the fact that many eukaryotes harbor Chlorella-like endosymbionts (i.e., "zoochlorellae") (2). Unlike most Chlorella species (20), the green algal endosymbionts did not have visible pyrenoids under the light microscope (Fig. 1C), but a pyrenoid-like structure was clear in our electron micrographs ( fig. S2, B and C). The rbcL gene of the green endosymbionts is identical to that of Chlorella sp. K10, an endosymbiont of the freshwater cnidarian Hydra viridissima strain K10 isolated in Switzerland ( Fig. 2B and fig.   S3B) (21). Because H. viridissima has co-speciated with its Chlorella endosymbionts (21), it is possible that the ancestor of P. tenue acquired Chlorella sp. K10 after it was released from a cnidarian host.
The bacterial endosymbionts are most closely related to the genus Thiodictyon (Chromatiaceae, Gammaproteobacteria) according to phylogenetic analyses of their 16S rRNA gene ( Fig. 2C and fig. S3C). We confirmed the localization of the Thiodictyon-related bacteria in the cytoplasm of P. tenue by fluorescence in situ hybridization (FISH) (Fig. 2, D and E, and movie S2; see Materials and Methods for details). Most members of the Chromatiaceae, known as "purple sulfur bacteria," are anaerobic and anoxygenic photosynthesizers that often use hydrogen sulfide as a photosynthetic electron donor and accumulate sulfur globules in their periplasm (22). The conspicuous purple-violet color of the bacterial endosymbionts (Fig. 1, B and C) and the abundant vesicular "chromatophores" (Fig. 2F and fig. S2, E and F) agree with them being a member of the Chromatiaceae (22). Absorbance spectra of the purple endosymbionts by hyperspectral microscopy reveal clear peaks in the infrared spectrum ( Fig. 2G and  fig. S4B) that are typical for bacteriochlorophyll-containing organisms (12). The Thiodictyon-related endosymbionts do not appear to accumulate highly refractile sulfur globules (Figs. 1, C and E, and 2F), and we also did not recover discrete sulfur signals through scanning electron microscopy-energy-dispersive x-ray spectroscopy. This suggests that the photosynthetic physiology of the Thiodictyonrelated endosymbiont might have diverged from that of its closest free-living relatives. Moreover, the Thiodictyon-related endosymbionts are of great interest as no other photosynthetic purple sulfur bacteria are known to be endosymbiotic (22); there are, however, several nonphotosynthetic, sulfur-oxidizing endosymbionts (e.g., "Candidatus Thiosymbion") (23). On the basis of its phylogenetic placement as sister to known Thiodictyon species (Fig. 2C) and the 0.91 average nucleotide identity to its closest genome-sequenced relative (see Materials and Methods), we propose a new candidate species for the purple bacterial endosymbiont of P. tenue [in accordance with published recommendations (24)]: "Candidatus Thiodictyon intracellulare" [(Chromatiales, Gammaproteobacteria) NC; NA; R; NAS (GenBank number MW203126), oligonucleotide sequence complementary to unique region of 16S rRNA is 5′-CTC TCA GAC TCT AGC GCG TCA-3′ (probe Thio643); S (P. tenue, Ciliophora; cytoplasm); probably microaer., reduced metabolism (probably no sulfur dissimilation); M]. Muñoz-Gómez, Kreutz, and Hess, Sci. Adv., 7: eabg4102, 2021. Etymology: intrā (Latin) = inside, cellula, -ae f (diminutive of cella) (Latin) = small room, cell. The epithet "intracellulare" refers to the intracellular lifestyle of the candidate species. Additional information: Rod-shaped, sometimes slightly curved, purplish cells colonizing the cytoplasm of P. tenue (Heterotrichea, Ciliophora) in large numbers (about 900 to 1300 bacterial cells per host cell); vegetative cells about 6 × 2 m on average, bounded by two membranes and an additional vacuolar membrane, and packed with vesicular chromatophores discernible in transmission electron microscopy (TEM) micrographs; no other life history stages observed; genetically defined by the 16S rRNA gene sequence MW203126 and the draft genome JADNEJ000000000 (both GenBank). Reference material: Resin block for TEM (NHMUK 2021.4.15.1; deposited in the "Protists Collection" at the Department of Life Sciences of the National History Museum in London (Cromwell Road, UK) containing a single host cell collected in pond II of the "Simmelried" moorland, Constance, Germany; 47.717767, 9.09375.

"Ca. Thiodictyon intracellulare" has a reduced and specialized metabolism
To gain further insights into the possible metabolic interactions between P. tenue and its photosynthetic endosymbionts, we sequenced the metagenome of this tripartite symbiotic consortium. The metagenome was assembled from three deeply sequenced short-read libraries made from amplified genomes of three separately isolated single cells (see Materials and Methods for details). The metagenomic scaffolds mostly group into three main clusters that correspond to nearly complete genomes for each symbiotic partner ( fig. S5). The genome of "Ca. T. intracellulare" is less than half the size of that of its closest known relative, "Candidatus Thiodictyon syntrophicum" [2.9 versus 7.74 million base pairs (Mbp); 3141 versus 6771 genes; table S2] (25). Such genome reduction is consistent with an endo symbiotic lifestyle (26) and has been accompanied by extensive gene loss and pseudogenization (figs. S6 and S7 and table S2).
"Ca. T. intracellulare" lost genes for major metabolic pathways such as the capacity to use reduced sulfur compounds as electron donors for photosynthesis (sulfur dissimilation), a hallmark of purple sulfur bacteria (27). On the basis of the absence or pseudogenization of several important genes, "Ca. T. intracellulare" lost the capacity to oxidize sulfide (e.g., sqrD and sqrF genes), elemental sulfur (e.g., dsrA and dsrB genes), and thiosulfate (e.g., soxY and soxZ) (data S1). These predicted losses are consistent with the absence of visible sulfur globules in the purple endosymbionts (Figs. 1, C and E, and 2F). Furthermore, the capacity to fix atmospheric dinitrogen (e.g., nifD and nifK) and hydrolyze urea (e.g., ureA, ureB, and ureC) as well as the ability to store excess nitrogen as cyanophycin (e.g., cphA) were also lost by "Ca. T. intracellulare" (data S1). However, "Ca. T. intracellulare" might be able to acquire nitrogen from its host because it retained transporters for glutamate and/or aspartate and ammonium ( Fig. 3 and data S1). Moreover, "Ca. T. intracellulare" retained many genes involved in pathways for carbon and energy metabolism, such as photo synthesis (e.g., photosystem and light-harvesting antennae), carbon fixation [RuBisCO (ribulose-1,5-bisphosphate carboxylase-oxygenase) and the Calvin cycle], respiration under hypoxic conditions (e.g., cytochrome cbb 3 ), fermentation (e.g., pyruvate:ferredoxin oxidoreductase and Hox [NiFe] hydrogenase), and the capacity to store excess carbon as glycogen and polyhydroxyalkanoate (PHA) granules (see Fig. 3 and data S1). Furthermore, "Ca. T. intracellulare" has transporters for small organic acids (i.e., acetate and succinate), glucose, maltose, and phosphate, which might be imported from the host cytoplasm (see Fig. 3, data S1, and Supplementary Text). Similar to Allochromatium vinosum, free-living Thiodictyon species are metabolic generalists that can use both reduced inorganic and organic compounds as electron donors (22,25,28). Therefore, the reduced and more specialized metabolism of "Ca. T. intracellulare" is likely to be the outcome of an endosymbiotic lifestyle in a nutrient-rich environment.
The ciliate host and the green algal endosymbiont genomes harbor genes for fermentation The endosymbiotic Chlorella sp. most probably has the capacity to photosynthesize oxygenically and respire aerobically as its free-living relatives do. However, this endosymbiotic green alga may also be able to ferment both in the light and in the dark (29,30). Chlorella sp. has retained multiple fermentative pathways in the cytosol, mitochondrion, and chloroplast that could result in the excretion of multiple fermentative end products, including d-lactate, acetate, ethanol, and molecular hydrogen P. tenue lives as a mixotroph by means of phagotrophy and anoxygenic photosynthesis P. tenue represents a physiologically complex and potentially flexible symbiotic consortium. This is because each symbiotic partner has more than one type of energy metabolism or mode of growth (Fig. 4, A to C). The physiological mode preferred by each symbiotic partner may be primarily determined by two major environmental factors: light and oxygen ( Fig. 4C and fig. S9). Despite the aforementioned metabolic complexity, which suggests ecological flexibility, the P. tenue symbiotic consortium appears to occupy a rather narrow niche in its environment. This is suggested by three observations. First, P. tenue preferentially dwells in hypoxic sediments rich in organic matter, where it phagocytoses and digests bacteria as food (see Fig. 1B for evidence of digestive vacuoles); this lifestyle is similar to that of Spirostomum species that are voracious bacterivores and often venture into hypoxic microenvironments to find their prey (18,31,32). Second, P. tenue co-occurs with diverse free-living purple sulfur bacteria ( fig. S1). The plant-covered and organic matterrich sediments where P. tenue is found presumably filter far-red and near-infrared light (13) and offer plenty of organic compounds and sulfide that select for metabolically versatile purple bacteria like the putative ancestor of "Ca. T. intracellulare" (22,25). Dwelling in hypoxic sediments might therefore favor anoxygenic photosynthesis by the purple symbionts over oxygenic photosynthesis by the green algal symbionts. Third, most of the volume of the P. tenue symbiotic consortium is occupied by "Ca. T. intracellulare" (Fig. 1A).
The purple endosymbionts represent about 26.4% of the volume of the whole symbiotic consortium, whereas the green endosymbionts only represent about 1.4% (table S1 and Supplementary Text). Together, these observations suggest that the physiology of the symbiotic consortium largely follows that of its purple endosymbionts, which may primarily photosynthesize anaerobically (e.g., "Ca. T. intracellulare" synthesizes bacteriochlorophyll anaerobically through an oxygenindependent BchE enzyme). We conclude that P. tenue is a mixotroph with a dual lifestyle that combines anoxygenic photosynthesis in the light and micro-aerobic phagotrophy (Fig. 4, A to C, and fig. S9).

What metabolic interactions might underlie the P. tenue symbiosis
On the basis of our microscopic observations and inferences from metagenomic data, the following metabolic interactions among the symbiotic partners in P. tenue are most plausible. In the light and the absence of oxygen, the purple endosymbionts most probably photosynthesize anoxygenically, while both the ciliate host and the green endosymbionts ferment reduced organic compounds (Fig. 4, A and C); green algae such as Chlamydomonas reinhardtii or In light-oxia, all three partners respire aerobically, but the green algae also photosynthesize oxygenically; the chloroplasts of the green algal endosymbionts of P. tenue accumulate starch that suggests that they actively photosynthesize (see Supplementary Text). In dark-anoxia, the whole symbiotic consortium turns to fermentation of ingested food or stored carbon (36). Nevertheless, intermediate metabolic states between these four major modes might occur. See also fig. S8. the salamander-symbiont Oophila amblystomatis switch to hydrogenproducing fermentation in hypoxia and light [(33) and references therein] (it is also possible, in principle, that the green algae photosynthesize under hypoxia, and the oxygen released is quickly consumed by host mitochondria). The purple symbionts can use either hydrogen or organics as electron donors. Hydrogen might be released by fermentation from the green alga (by means of its [FeFe] hydrogenase) or by photo-fermentation from the purple symbiont itself (through an [NiFe] hydrogenase) (Figs. 3 and 4A). Organic compounds are produced by both algal (e.g., acetate and lactate) and host mitochondrial (acetate and propionate) fermentation ( Fig. 4A and fig. S8). The purple symbionts have, indeed, retained transporters for these small organic acids (i.e., acetate or succinate) and for glucose (see above; Fig. 3). The energy derived from the anoxygenic photosynthetic chain of the purple symbionts is likely used to fix inorganic carbon (CO 2 ) derived from the host metabolism and the environment. The fixed carbon can then be stored as glycogen or PHA granules (rarely seen in electron micrographs) or, more likely, released and stored in the host as starch (Fig. 4A). The latter is suggested by the presence of numerous electron-translucent granules in the cytoplasm of P. tenue [presumably amylopectin (34); fig. S2, E and F]. We expect that "Ca. T. intracellulare" supplies photosynthate in the form of small organic compounds to its host, as known from many other photosynthetic endosymbionts (7,(35)(36)(37)(38). As a nitrogen source, the purple symbionts (as well as the green algal symbionts) might import ammonium and glutamate/aspartate from the ciliate host ( Fig. 3; see above) as is also seen in many other photosymbioses (36,(39)(40)(41)(42). This agrees with the observation that "Ca. T. intracellulare" lost the capacity to store fixed nitrogen but retained ammonium and glutamate/aspartate transporters (Fig. 3).
In the dark, all symbiotic partners probably respire aerobically to fully oxidize reduced organic compounds, which derive from stored carbon (e.g., starch for the host and green alga and glycogen or PHA for the purple symbiont; Fig. 4B) or phagocytosed food (Figs. 1B and 4B). This is expected because each symbiotic partner has microaerobic respiratory chains ( Fig. 3 and fig. S8), and respiration is much more efficient than fermentation. Then, it might well be that the P. tenue symbiotic consortium moves into slightly more oxygenated microenvironments to respire micro-aerobically in the dark. Vertical migration in response to oxygen concentration is indeed known from various ciliates, including Spirostomum species (43,44). The purple endosymbionts can presumably survive under low oxygen concentrations as they retained superoxide dismutases for oxygen radical detoxification (data S1) and have micro-aerobic respiratory chains with a cytochrome cbb 3 (Fig. 3) that can scavenge oxygen (45). This agrees with the fact that P. tenue is found in the hypoxicto-oxic transition zones of muddy ponds.
At least two other physiological modes are allowed by the metabolic complexity of the P. tenue symbiotic consortium (see Fig. 4C). Moreover, intermediate metabolic states between these four major modes might also occur in transitional environments. For example, the green algal symbiont might perform some oxygenic photosynthesis (46) or a form of anoxygenic photosynthesis through cyclic electron flow under low-light and anaerobic conditions (47), and the purple symbionts could express their photosynthetic chain alongside its micro-aerobic respiratory chain (48). The accumulation of starch granules around pyrenoids in chloroplasts supports the notion that the green algae can photosynthesize oxygenically under some conditions (see fig. S2C and Supplementary Text). The P. tenue consortium is thus endowed with great physiological, and potentially also ecological, flexibility. This flexibility might allow P. tenue to temporarily survive under suboptimal conditions ( fig. S9) such as oxygenated waters; P. tenue survives for a few days but does not divide and eventually disappears in laboratory samples exposed to atmospheric oxygen. Given the complex physiology of each symbiotic partner, sequencing single-cell metatranscriptomes from this unusual symbiotic consortium under different environmental conditions (e.g., light-anoxia versus dark-oxia) is necessary. These future efforts will test the physiological hypotheses presented here and provide more certainty about the metabolic interactions between P. tenue and its symbionts.

The symbiotic merger of disparate physiologies created an unusual ecological niche
By studying a unique purple-green ciliate, we found the first example of a photosynthetic purple "sulfur" bacterium (Chromatiaceae, Gammaproteobacteria) with an endosymbiotic lifestyle. This anaerobic purple bacterium shares its host cell with an oxygenic green alga (Chlorella sp.). This puzzling coexistence appears to be possible because of the physiological versatility offered by each symbiotic partner and the reductive specialization of the purple bacterial endosymbiont. For example, the purple bacterium lost sulfur dissimilation genes, and both symbionts and host have the genes for respiring micro-aerobically and fermenting anaerobically. Moreover, the evolutionary merging of two distinct photosymbionts and a facultatively anaerobic predator has created a symbiotic consortium with a complex set of physiological properties not shared by any other species known. The niche of this symbiotic consortium, however, is not simply the union of the niches of each symbiont but presumably encompasses most of the ciliate host and purple symbiont niches and partially overlaps with that of the green symbiont ( Fig. 4C and fig. S9). The outcome is a eukaryote that actively swims in hypoxic sediments and combines phagotrophy with anoxygenic photosynthesis. P. tenue thus represents an extraordinary example of how symbiosis merges disparate physiologies and allows emergent consortia to create novel ecological niches.

Sampling and storage of live material
Water and sediment samples were taken from pond II [naming according to (15)] of the Simmelried peat bog area near Hegne, Germany (47.717767, 9.09375) at a depth of about 30 cm. These samples were transported in 200-ml jars and lastly combined into 1-liter bottles with a loose screw cap to allow the formation of a bottom sediment and different zones. These bottles were kept at 15°C under dim, indirect light with a 14:10-hour light/dark cycle.

Phylogenetic analyses
Sanger reads were assembled on Benchling (https://benchling.com). Overlapping consensus sequences of the ciliate rRNA operon were assembled in Geneious (www.geneious.com). Representative sequences for the Heterotrichea (ciliate rRNA gene operon) were retrieved from Shazib et al. (19). Representative sequences for the Chlorellaceae (rbcL) were retrieved from a BLASTn search against the National Center for Biotechnology Information (NCBI) GenBank nr database. Representative sequences for the Chromatiaceae and Sedimenticolaceae (16S rRNA gene) were retrieved from the SILVA rRNA database (53). The COI and rbcL genes were aligned by codons with the MAFFT algorithm in TranslatorX (54). The 16S, 18S, 5.8S, D1D2-28S rRNA genes, and the ITS1 and ITS2 regions were aligned separately with MAFFT v7.310 (55) using the G-INS-I method and then trimmed with Trimal v1.4.rev15 (56) and the -automated1 option. The final selection of taxa was performed visually with the aim of maximizing phylogenetic diversity. The 18S, 5.8S, D1D2-28S rRNA genes, and the ITS1 and ITS2 regions of the ciliate rRNA operon were analyzed as separate partitions. Phylogenetic trees were inferred using IQ-TREE v.2.0.3 (57) and the best-fit model according to Bayesian information criterion as implemented in ModelFinder (58).

FISH and fluorescence microscopy
Using a 16S rRNA gene sequence obtained by Sanger sequencing that showed >98% identity with Thiodictyon species (Chromatiaceae, Gammaproteobacteria), we retrieved potential FISH probes with the "Design Probes" web tool of the software toolset DECIPHER under default hybridization conditions (59). One of the probe sequences with maximum specificity (score of 0), reasonable hybridization efficiency, and no reported cross-hybridizations with nontarget sequences was analyzed with the online tool "mathFISH" (60) and manually modified to optimize binding to the target sequence. The resulting probe Thio643 (5′-CTC TCA GAC TCT AGC GCG TCA-3′) was lastly checked for specificity in silico with the "probeMatch" database (61) and the SILVA probe match and evaluation tool "TestProbe 3.0." The probe showed no hits with zero mismatches in both databases and only few hits with one mismatch (18 hits in probeMatch: "unclassified Chromatiaceae"; 4 hits in TestProbe: uncultured Gammaproteobacteria) indicating high specificity (as of March 2020). The Cy3-conjugated probe Thio643-Cy3 was ordered (biomers.net GmbH, Ulm, Germany), and its binding to bacterial cells from P. tenue was evaluated under varying formamide concentrations (0 to 60%) using a modification of previously published FISH protocols (62,63). In brief, cells of P. tenue were isolated from natural samples, washed twice in distilled water, placed on gelatincoated glass slides (coated with 0.1% gelatin in water), and air-dried. Some ciliate cells were deliberately burst with a pipette tip to spread the endosymbionts. Cells on dry slides were fixed with formaldehyde (3.7% in water) for 10 min, then rinsed in distilled water, and air-dried. To minimize potential autofluorescence, photosynthetic pigments of bacteria and algae were first extracted from the samples by incubating the slides in a methanol:water mixture (8:2) at 65°C for 10 min, then rinsed in ethanol (100%), and air-dried. The cells were then covered with 50 to 100 l of hybridization buffer containing probe Thio643-Cy3 at an appropriate concentration (5 ng/liter) and 0 to 60% formamide [according to (63)] and then incubated in a moist chamber at 46°C for 90 min. After incubation, the slides were quickly rinsed with warm washing buffer (48°C) and subsequently incubated in washing buffer at 48°C for 25 min. Last, slides were rinsed with distilled water, air-dried, and mounted with SlowFade Diamond Antifade Mountant (Thermo Fisher Scientific).
Routine epifluorescence microscopy of FISH samples was done with the Zeiss Axiophot microscope equipped with phase contrast optics, the Zeiss filter set 43 HE (excitation, 550/25 nm; emission, 605/70 nm), and the camera AxioCam HRc (Carl Zeiss). For confocal laser scanning microscopy, a Leica TCS SPE system and the Leica Confocal Software (Leica Microsystems) were used. The samples were excited with lasers (561 nm for Cy3 and 488 nm for chlorophyll) and scanned in three dimensions with step sizes of <1 m. Confocal laser scanning microscopy (CLSM) stacks were processed and analyzed with the software LAS AF lite (Leica Microsystems) and the image processing package Fiji (64). This included the selection of pseudocolors for fluorescence channels, the adjustment of brightness, and the combination of different channels (merged channels) and/or of several focal planes (Z-projections).

Transmitted light microscopy
For high-resolution imaging of live cells, two microscopes were used: an Olympus BX 50 microscope and a Zeiss IM35 inverted microscope, both equipped with differential interference contrast optics, high-resolution objective lenses, and an electronic flash. Digital

of 11
images were taken with an Olympus E-P5 digital camera and Canon EOS 6D digital single lens reflex camera, respectively, and processed with Adobe Photoshop CS4 (Adobe Systems).

Transmission electron microscopy
About 30 cells of P. tenue were isolated from natural samples and collected in about 200 l of distilled water. This cell suspension was quickly injected into 200 l of freshly prepared fixative [80 l of 25% glutaraldehyde, 200 l of 4% osmium tetroxide solution, and 120 l of half-strength medium Waris-H (pH 7) with 1.5 mM Hepes], resulting in a final concentration of 2.5% glutaraldehyde and 1% osmium tetroxide. Cells were incubated for 1 min at room temperature and for another 45 min at 6°C, then quickly washed with distilled water (twice, supernatant was removed after cells settled), and dehydrated by a series of acetone:water mixtures (30, 50, 80, and 90%) and pure acetone (twice) for 5 min each at −20°C. Acetone was then replaced by 50% EPON in acetone, incubated for 24 hours at about 6°C for infiltration, and then the tube was opened and stored for another 24 hours in a fume hood to let the acetone evaporate. The sample was then mixed with fresh EPON and cured at 65°C for 1.5 days. Single ciliate cells were sectioned (~60 nm thickness) with a Leica EM UC7 ultramicrotome (Leica Microsystems), stained with uranyl acetate (1%, 15 min) and Reynolds' lead citrate solution (3.5 min), and imaged with a CM10 TEM (FEI Europe Main Office, Eindhoven, The Netherlands) and a Gatan ORIUS SC200W TEM charge-coupled device (CCD) camera (Gatan Inc., Pleasanton, CA).

Hyperspectral imaging
Several cells of P. tenue were picked from a natural sample, washed in demineralized water, placed onto a coverslip, and let air-dry. The dry cells were then covered by immersion oil and imaged with a CytoViva Hyperspectral Microscope (CytoViva Inc., AL, USA) in the visible and near-infrared range (400 to 1000 nm) using the Olympus UPlanFL N 100×/1.30 oil objective (Olympus Europa SE & Co. KG, Hamburg, Germany). The microscope was based on an Olympus BX43 stand equipped with a SPECIM spectrograph (Specim, Spectral Imaging Ltd., Oulu, Finland) and the pco.pixelfly USB CCD Camera (PCO AG, Kelheim, Germany). The sample was illuminated with transmitted light of a Fiber-Lite DC-950 halogen illuminator (Dolan-Jenner Industries) coupled into the Olympus BX43 stand. Hyperspectral images were recorded and analyzed with the ENVI software (CytoViva Inc., AL, USA). Per sample type (purple bacteria and green algae), the recorded transmission spectra of four rectangular areas (defined as regions of interest; see fig. S4A) were averaged and used for analysis. First, the averaged transmittance spectra were normalized to the recorded lamp spectrum (also obtained by four averaged measurements) and then converted to relative absorbance spectra for each sample type. The exact sampling areas and absolute absorbance spectra are shown in fig. S4A.

Single-cell whole-genome amplification and sequencing
Individual ciliates isolated, washed, and deposited in 0.2-ml tubes as detailed above were inspected under low magnification to confirm cell integrity and then snap-frozen in liquid nitrogen. The REPLI-g Advanced DNA Single Cell Kit (QIAGEN) was used to perform a single-cell whole-genome amplification (WGA) on three individual ciliates. Three different variations of the WGA protocol were performed: Variation 1 was applied on single cells and followed the default protocol. Variation 2 introduced a 65°C incubation step (10 min) to facilitate lysis of endosymbionts. Variation 3 followed the default protocol on a single-cell DNA extract obtained with Chelex 100 (see above for details). Illumina sequencing libraries were made with the Illumina TruSeq kit from the products of all three WGA variations. Illumina short-read sequencing (2 × 150 bp) was done on a NovaSeq Illumina sequencing platform at the Cologne Center for Genomics at the University of Cologne, Germany.

Metagenome assembly and analysis
Raw Illumina short reads were quality assessed and filtered, and adapters were trimmed with afterqc v.0.9.6 (-f 8 -t 2) (65). Coassembly of multiple sequenced libraries was done with SPAdes v.3.14.0 (--sc) (66). Scaffolds were then processed using Anvi'o v.6.1 (67) and the metagenomic pipeline. In brief, all scaffolds shorter than 1000 bp were discarded, and reads from each library were separately mapped on the scaffolds to obtain their coverage values. Taxonomic affiliations for each contig were assigned on the basis of a consensus of best DIAMOND (68) hits for each scaffold gene against the NCBI nr database as of October 2019. Scaffolds were organized according to the composition and coverage and manually visualized and inspected with anvi-interactive (67). The scaffolds in the simple metagenome clearly fell into three main clusters and closely followed the taxonomy of each of the three symbiotic partners in P. tenue as determined with PCR/Sanger sequencing. The scaffold cluster (bin) annotated as Thiodictyon was further manually refined with anvirefine (67) by relying on composition, coverage, taxonomic, and completeness/redundancy criteria. The metagenome-assembled genome for "Ca. T. intracellulare" was estimated to be 99.16% complete and 0.74% redundant (a measure of potential contamination) by CheckM v.1.0.18 (69) and the taxonomic workflow for the Gammaproteobacteria. Scaffold clusters that taxonomically corresponded to the P. tenue ciliate host and the endosymbiotic green alga Chlorella sp. were identified manually through Anvi'o and estimated to be 93.0 and 77.2% complete by BUSCO v.4.0.6 (70) based on the alveolata_odb10 and chlorophyta_odb10 lineage datasets, respectively.

Mitochondrial genome (mitochondrial DNA) annotation
Mitochondrial genome scaffolds were identified by BLASTp searches of mitochondrial DNA (mtDNA) proteins (AN) from Gruberia lanceolata (Heterotrichea, Ciliophora) (71) against the P. tenue metagenome. The scaffolds identified were then inspected through anvi-interactive, and they were found to cluster together according to their composition. Attempts to circularize and/or extend the scaffolds were done with NOVOPlasty v.3.8.2 (72). Several NOVOPlasty iterations were performed with different seeds. The resulting contigs were assembled in Geneious to resolve the scaffold organization manually. A preliminary annotation was carried out with MFannot (https://github.com/BFL-lab/Mfannot). Because of the extreme divergence of ciliate mtDNA protein-coding genes, all predicted open reading frames (ORFs) larger than 50 amino acids were searched against the NCBI GenBank nr database and a ciliate mtDNA database (containing all conserved ORFs of unknown function). In addition, pHMM searches against the Pfam database were performed.
Inference of the anaerobic energy metabolism of the P. tenue ciliate host and its green endosymbiont Chlorella sp. A set of query sequences reported for anaerobic enzymes across eukaryotes in (73,74) were retrieved and used for BLASTp, BLASTn, and Exonerate (75) searches against the draft genome of the P. tenue ciliate host. Similarly, a set of protein query sequences for enzymes annotated as involved in anaerobic energy metabolism and fermentative pathways in C. reinhardtii (29,30) were manually retrieved from UniProtKB. These were used to search the draft genome of the endosymbiotic Chlorella sp. using Exonerate and the options "--refine full --model protein2genome." Best BLAST and Exonerate hits were then searched against the nonredundant (nr) dataset to confirm their presumed homology.
Inference of the metabolism of the purple endosymbiont Ca. T. intracellulare The proteome of "Ca. T. intracellulare" was inferred with Prodigal v.2.6.3 (-p single) (76). Annotation of the proteome was done with InterProScan v.5.41-78.0, BlastKOALA, and KofamKOALA at the Kyoto Encyclopedia of Genes and Genomes server (77). Reciprocal BLAST searches against the predicted proteome of "Ca. T. intracellulare" were performed using query sequences for genes involved in sulfur metabolism described and annotated in (27). Reciprocal BLAST searches against the predicted proteome of "Ca. T. intracellulare" using the query sequences for metabolic genes present in the genomes of "Ca. T. syntrophicum" and A. vinosum DSM-180, as discussed in (25,28), were also performed. Moreover, additional reciprocal BLAST searches were performed for additional genes of interest not discussed in (25,28). Rough predictions of pseudogenes in the genomes of "Ca. T. syntrophicum" and Ca. T. intracellulare were performed with PGAP (78) and DFAST (79). A more exhaustive prediction of pseudogenes in the genome of "Ca. T. syntrophicum" was performed with Pseudofinder (80).