Selective pressure against horizontally acquired prokaryotic genes as a driving force of plastid evolution

The plastid organelle comprises a high proportion of nucleus-encoded proteins that were acquired from different prokaryotic donors via independent horizontal gene transfers following its primary endosymbiotic origin. What forces drove the targeting of these alien proteins to the plastid remains an unresolved evolutionary question. To better understand this process we screened for suitable candidate proteins to recapitulate their prokaryote-to-eukaryote transition. Here we identify the ancient horizontal transfer of a bacterial polyphenol oxidase (PPO) gene to the nuclear genome of an early land plant ancestor and infer the possible mechanism behind the plastidial localization of the encoded enzyme. Arabidopsis plants expressing PPO versions either lacking or harbouring a plastid-targeting signal allowed examining fitness consequences associated with its subcellular localization. Markedly, a deleterious effect on plant growth was highly correlated with PPO activity only when producing the non-targeted enzyme, suggesting that selection favoured the fixation of plastid-targeted protein versions. Our results reveal a possible evolutionary mechanism of how selection against heterologous genes encoding cytosolic proteins contributed in incrementing plastid proteome complexity from non-endosymbiotic gene sources, a process that may also impact mitochondrial evolution.

A large number of plant genes encoding plastid proteins originated via independent horizontal gene transfer (HGT) events from prokaryotes other than Cyanobacteria 1,2 , the accepted endosymbiotic precursor of modern-day plastids 3 . The vast majority of those genes are nuclear, and therefore their protein products are translated in the cytosol and subsequently imported into the plastid. This in itself is remarkable, because prokaryotic genes lack the information needed to target their encoded proteins to eukaryotic organelles and hence must have acquired it following insertion in the nuclear genome and transcriptional activation. Despite the profound evolutionary importance of this process, we know little about the forces underlying the targeting of these alien proteins to the plastid.
We addressed this conundrum using a novel approach that is based on the idea that studying the ancestral (cytosolic) and modern (plastidial) subcellular localization of HGT-derived proteins in vivo may provide clues about the evolutionary mechanisms that ultimately resulted in their organellar targeting. We accordingly screened for candidate proteins of non-photosynthetic function that were present exclusively in prokaryotes and plants, and whose plastidial localization in the later could not be explained by a strict functional requirement.
Scientific RepoRts | 6:19036 | DOI: 10.1038/srep19036 A uniquely suited candidate for a case study was found to be polyphenol oxidase (PPO; Enzyme Commission 1.14.18.1 or 1.10.3.1), a bacterial enzyme 4 whose nucleus-encoded plant homologues are almost ubiquitously present in plastids 5 . Although the physiological functions of PPO are not fully understood in plants, its capability to oxidize phenolic compounds (e.g. phenylpropanoids) into highly reactive quinones is thought to function as an "armed bomb" defence mechanism against phytophagous organisms functioning only when cell damage occurs and plastids break [6][7][8] . In bacteria, PPO enzymes are involved in the production of melanin-like polymers that protect bacterial cells and spores against UV radiation, heat, metal ions, oxidants, enzymatic hydrolysis, antimicrobial compounds and phagocytosis 9 .
We present here evidence indicating that an early common ancestor of land plants acquired a bacterial gene coding for PPO through an ancient HGT event during the plant colonization of land. Recapitulating the bacteria-to-plants transition of PPO in the model plant Arabidopsis thaliana (Brassicaceae; hereafter Arabidopsis) suggests that the plastidial targeting of PPO in plants is likely a consequence of selective pressures acting against the deleterious effects caused by a PPO enzyme functioning in the cytosol. Given that biological processes in eukaryotic cells are highly compartmentalized, selection against horizontally acquired genes encoding non-targeted heterologous proteins may represent a general route by which novel organellar protein versions can arise.

Results
Plants horizontally acquired a bacterial PPO gene during colonization of land. Genes coding for PPO have been found in bacteria, plants and fungi 10,11 , but their phylogenetic relations have not been studied in detail. At the protein level, PPOs of all three groups harbour a catalytic tyrosinase (TYR) domain (PFAM00264) towards the N-terminus, found not only in PPOs but also in several proteins of all branches of life 12 (Fig. 1a). In plants, the C-terminal half of PPO has distinctive DWL (PFAM12142) and KFDV (PFAM12143) domains 5 . Both DWL and KFDV domains are identified as such by the Conserved Domains Database (CDD) 13 in PPOs of various unrelated bacteria like the free-living betaproteobacterium Chitinomonas koreensis, the plant pathogen betaproteobacterium Ralstonia solanacearum, the gammaproteobacterium Rheinheimera sp. and the cyanobacterium Gloeocapsa sp. Some bacterial PPOs possess only one of these domains, while PPOs of many other bacteria as well as all fungal PPOs lack recognizable DWL and FKDV domains (Fig. 1a,b). Sequence alignments of PPO peptides from plants, bacteria and fungi show that the TYR domain can be readily aligned. The DWL and the KFDV domains, although only distantly related, can also be aligned among these proteins, indicating that the PPOs of these organisms are phylogenetically related ( Supplementary Fig. S1). Notably, sequences encoding proteins with a PPO-like structure could not be retrieved from the complete genomes of the green algae Volvox carteri, Chlamydomonas reinhardtii, Micromonas putida and Chlorella sp., as well as from the genomes of red algae and other distant organisms like stramenopiles or haptophyte algae. Although some representatives of these groups have proteins containing TYR domains, DWL and KFDV domains could not be identified in any of them (Fig. 1a).
The survey described above indicates that PPO proteins are only found in land plants, fungi and some bacteria. To analyse the phylogenetic relations between these proteins, we focused on the TYR domain. A Maximum Likelihood analysis of the relationships between PPO tyrosinase (PPO-TYR) domains of land plants, bacteria and fungi reveals that land plant PPO-TYR domains form a well-supported monophyletic group with the PPO-TYR of several bacteria (Fig. 1b). Protein structure predictions of R. solanacearum (WP_003278615), Physcomitrella patens (AAX69084) and V. carteri (XP_002956492) further illustrate a greater structural similarity among land plant and bacterial TYR domains (Fig. 1c). The PPO-TYR domains of fungi are grouped in a separate, well-supported branch (Fig. 1b). Although green algae lack bona fide PPO proteins, we included in our phylogenetic analysis the only TYR domains that could be retrieved from the proteomes of Chlamydomonas and Volvox. As shown in Fig. 1b, the green algae TYR domains do not cluster with land plants, suggesting that land plants have not inherited PPO-TYR domains from a green algal ancestor by vertical inheritance. The patchy phylogenetic distribution of PPO genes as well as the phylogenetic incongruences observed for the PPO-TYR domains support the conclusion that PPO proteins are of ancient origin but have been subjected to many different HGT events during evolution. In particular, the fact that PPO-TYR domains of land plants, including the basal groups Bryophyta (P. patens) and Lycophyta (Selaginella moellendorffii), are all grouped together with bacteria and that PPO genes cannot be found in glaucophyte, red and green algal genomes, strongly suggest that a PPO-like gene was transferred from a bacterium to an early land plant ancestor circa 450-500 million years ago during the transition of plants to live on land 14 (Fig. 1d). Additional evidence supporting the PPO bacteria-to-plants HGT is the observation that, compared to the average of plant genes, which typically contain 5-6 introns 15,16 , plant PPOs have substantially fewer or no introns (0-2 introns) 5 , a characteristic feature of prokaryotic horizontally acquired genes 17 .
Recapitulation of the PPO prokaryote-to-eukaryote transition. Having defined the bacteria-to-plants HGT of PPO, we set to experimentally recreate its horizontal acquisition by land plants. Our aim was to generate a model to study the effect of incorporating a nuclear gene coding for a protein of prokaryotic origin (PPO) that initially lacked information for a defined subcellular localization and eventually evolved a plastidial destination. Given that inferred ancestral proteins and proteins having evolved in distant lineages may result inactive or cause a stress derived from maladapted cell components when introduced into a distantly related host 18,19 , we decided to use a modified plant PPO to simulate the presumed ancestral cytosolic state of the enzyme rather than a resurrected or a bacterial protein. To ensure that the enzyme used would be active, we chose a PPO gene from Solanum tuberosum that we previously confirmed to encode an active PPO enzyme 20 . The plastid-targeting signal sequence of the PPO gene was identified in silico using ChloroP 21  with Ancestral localization) and one retaining it (version PPO M , for enzyme with Modern localization) (Fig. 2a). The recombinant proteins were expressed in Escherichia coli and the activity of PPO A and PPO M was confirmed in vivo based on the ability of both proteins to induce melanin synthesis (Fig. 2b). To determine the subcellular localization of PPO A and PPO M in plant cells, a GFP (green fluorescent protein) tag was fused to the C-terminal end of both PPO versions and the corresponding constructs were transiently expressed in Nicotiana benthamiana leaves. Laser-scanning confocal microscopy analyses showed that PPO A presents a cytosolic distribution pattern while PPO M is detected only in plastids (chloroplasts) (Fig. 2c), as expected. To experimentally study the PPO bacteria-to-plants transition (albeit in a modern context), we generated stably transformed Arabidopsis plants by Agrobacterium-mediated transformation with constructs for constitutive PPO A or PPO M expression. In agreement with the conclusion that PPO activity is not absolutely required for photosynthesis or any other primary plant function 5,6,20,22 , Arabidopsis does not contain any PPO genes 5 and thus it has lived without PPO for several million years 23 . We therefore chose Arabidopsis as an optimal plant system to have an approximation of the physiological impact of acquiring a PPO gene ab initio. After isolating independent homozygous lines containing a single T-DNA insertion in the nuclear genome, we selected three PPO A and three PPO M lines with different PPO gene-expression (Fig. 3a) and activity (Fig. 3b) levels. Expressing the cytosolic PPO A led to reduced growth rates, while expressing the plastid-localized PPO M had no evident phenotypic effect on transgenic lines compared to untransformed controls (Fig. 3c-e). The expression and activity of the cytosolic version inversely correlated (r = − 0.97 and r = − 0.99, respectively) with the surface area of rosette leaves (Fig. 3d), supporting the conclusion that the observed deleterious phenotype in PPO A lines was directly caused by the presence of such an extraplastidial PPO activity. Plants transformed with PPO A were also considerably delayed in terms of bolting time (Fig. 3e), which again correlated with its expression and activity (r = 0.93 and r = 0.99, respectively).
At the metabolic level, the qualitative profile (i.e. number of HPLC peaks) of phenylpropanoid compounds such as flavonols (Fig. 4a) and anthocyanins (Fig. 4b) was found to be very similar in PPO A , PPO M , and untransformed plants. However, the presence of an active cytosolic PPO enzyme in PPO A lines had a strong quantitative impact on anthocyanin levels, which were doubled relative to PPO M plants or wild-type controls (Fig. 4c), Anthocyanins are purple pigments widely used as visual markers of plant cell stress. For example, treatment with norflurazon generates albino seedlings in which the absence of photosynthetic pigments (carotenoids and chlorophylls) causes a strong photooxidative stress and a visually detected accumulation of anthocyanins that again is higher in PPO A lines (Fig. 4d). Together, these results suggest that PPO enzymes can act on unidentified  Data correspond to mean ± SEM derived from 10 biological replicates. Asterisks represent statistically significant differences (*P = 0.05, ***P = 0.001) according to the ANOVA followed by Newman-Keuls post-hoc test. cytosolic substrates to eventually unchain a stress response. If such response is sustained (e.g. in lines constitutively expressing the extraplastidial PPO A enzyme), a negative impact on plant growth occurs.

Discussion
Despite the relevant contribution of HGT-acquired prokaryotic proteins to the establishment of the plastid proteome 1,2 , the evolutionary forces that promoted their targeting to the plastid have remained an unresolved evolutionary question. Focusing on a carefully selected candidate gene (PPO) for a case study, we have aimed to shed some light on this subject.
Our results strongly suggest that an early common ancestor of land plants horizontally acquired a gene coding for PPO of bacterial origin, possibly as a consequence of life in close contact with soil bacteria (Fig. 1d). An early eukaryotic origin of PPO is less parsimonious and would require, at least, multiple independent PPO losses in the glaucophyte, red algae and green algae lineages, as well as in the animal lineage. Furthermore, since the last common ancestor of the supergroup Plantae 3 originated more than 1.2 billion years ago 24 , this scenario would also imply that the PPO gene was maintained for more than 700 million years only in the algal lineage leading to land plants to be then lost after the colonization of land. Interestingly, the land plant acquisition of PPO corresponds to the period when the phenylpropanoid biosynthetic pathway (a rich source of phenolic metabolites) emerged in land plants, also in part by the gain of horizontally acquired genes from soil bacteria or fungi 25 . These coincidental events most likely allowed unprecedented interactions between PPO and the phenylpropanoid pathway, setting the ground for evolutionary innovations that led to the development of defensive and secondary metabolic functions for PPO in land plants. For instance, keeping the PPO enzyme separated from phenylpropanoid substrates that could be converted into toxic products for invaders could have represented a novel resource-efficient defence mechanism that only becomes active when cells are attacked and components of different subcelular compartments are mixed.
It has been established that translocation of genes from the essentially prokaryotic plastid genome to the nucleus involves post-insertional genomic rearrangements rather than random integration into a genomic location conferring instant functionality [26][27][28][29] . On the other hand, the acquisition of plastid targeting peptides is a process that can evolve with relative ease from existing genome sequences 30,31 . According to these data, it is presumed that ancestral prokaryote-to-eukaryote horizontally transferred sequences typically involved an initial integration into the nuclear genome followed by the incorporation of eukaryotic transcriptional regulatory elements and, in many cases, the eventual acquisition of subcellular localization signals. Therefore, for most HGT events resulting in plastid-localized proteins, there must have been an evolutionary time when populations contained both the ancestral (i.e. non-targeted, cytosolic) version of the protein and the modern (i.e. targeted, plastid-localized) version that was ultimately retained. From a Darwinian fitness viewpoint, and independently of the defence mechanism proposed for PPO in land plants, our results allow proposing that the ancestral bacterial gene coding for a non-targeted PPO integrated in the nuclear genome of an early common ancestor of land plants, became transcriptionally active, and subsequently evolved a plastid-targeting sequence, resulting in a more beneficial variant that was hence selected. Similar to that observed in bacteria 9 , it is possible that the acquisition of a non-targeted (i.e. cytosolic) PPO by early land plants might have provided the ability to synthesize phenolic polymers or other compounds with protective effects. This putative adaptive advantage might have been turned into a disadvantage once the transition from aquatic to terrestrial environments was completed. Alternative evolutionary trajectories that also fit our results would be that, when acquired by the land-plant ancestor, the bacterial PPO gene integrated by chance in a genomic location simultaneously conferring both transcriptional activation and plastid targeting or that the acquisition of a plastid localization signal preceded the transcriptional activation of the gene. Under these hypotheses, eventual mutations or genomic rearrangements resulting in the loss of the plastid-localization signal would lead to a disadvantageous phenotype and the cytosolic PPO would not be retained. In either case, the finding that PPO A triggers a stress response (as deduced from the accumulation of anthocyanins) that could eventually lead to reduced growth rates when compared with PPO M implies that directional selection would increase the frequency of PPO M relative to PPO A after the horizontal acquisition of the bacterial gene, ultimately favouring the fixation of the plastidial enzyme within the population.
In stark contrast to the common assumption that newly acquired genes conferring a strongly negative fitness effect will irreversibly undergo gene degeneration or loss, we propose that the fortuitous acquisition of organelle targeting sequences can provide an alternative fate to such genes. The argument is made that a horizontally acquired prokaryotic gene coding for a delocalized protein could have a deleterious effect that is only context-specific in a eukaryotic cell where different biological processes are highly compartmentalized. The targeting of the foreign protein to an organelle such as the plastid could alleviate the deleterious effect, hence allowing the gene to endure in the recipient genome and thereafter serve as raw material for the evolution of novel functions. The plausible generality of the proposed evolutionary mechanism is supported by the well-known fact that cytosolic localization of heterologous proteins is often deleterious to plant growth when compared with plastid-localized versions [32][33][34][35] , probably due to interference with cytosolic metabolic pathways and the enormous capacity of the plastid to accumulate heterologous proteins.
A fundamental goal in biology is to elucidate the forces leading to evolutionary change. Previous work has revealed the unexpected broad contribution of prokaryotic proteins acquired via independent HGT events to the plastid proteome 1,2 , the processes leading to gene functionalization of prokaryote-like (plastid) sequences in the nucleus 26,27 and the evolutionary acquisition of plastid targeting sequences 30,31 . By studying the likely evolutionary history followed by a horizontally transferred bacterial gene after acquisition by land plants hundreds of millions of years ago, we have experimentally shown that selective pressure against heterologous genes encoding cytosolic proteins may be an evolutionary force driving plastid proteome complexity from non-endosymbiotic gene sources. Finally, it has not escaped our notice that, as with the plastid, the inference we have postulated may also apply to certain horizontally transferred prokaryotic genes whose encoded proteins have been ultimately targeted to the mitochondrion 36-38 .

Methods
Gene sampling. Extensive BLAST 39 searches spanning the diversity of eukaryotic and prokaryotic life were performed to identify putative eukaryotic and prokaryotic PPO homologues. Additional searches were also performed excluding land plants (embryophytes, taxid: 3193). The searches for amino acid sequences were conducted against the GenBank non-redundant protein sequences database using the BLASTp tool of the NCBI (National Center for Biotechnology Information) BLAST server. The PPO-TYR domain was also used to search the genomes of unicellular eukaryotes deposited in the Joint Genome Institute (http://genome.jgi-psf.org/) and Phytozome (www.phytozome.net). To check the robustness of the gene sampling additional searches using PSI-BLAST 39 (5 iterations), tBLASTn and HMMER 40 were performed. Phylogenetic analysis. Tyrosinase domains of PPO proteins and other TYR-containing proteins were aligned with ClustalW 41 and edited with Jalview 2.8 42 to minimize gaps. Information about sequences used is described in detail in Supplementary Table S1. The tree in Fig. 1b used the LG protein substitution model 43 , but JTT 44 and Blosum62 45 yielded essentially the same results. Bootstrapped maximum likelihood phylogenetic trees were constructed with Phyml 3.0 46,47 . Gamma correction was used to account for heterogeneity in evolutionary rates with four discrete classes of sites, an estimated alpha parameter and an estimated proportion of invariable sites as implemented in Phyml. A total of 100 bootstrap replicates were used to assess robustness, and tree topology was optimized by subtree pruning and regrafting (SPR). The tree was visualized and edited with TreeDyn 48 . The Fig. 1d  The in silico detection of the subcellular localization signal was identified using ChloroP 21 (http://www.cbs. dtu.dk/services/ChloroP/). Sequences were then subcloned into plasmids pGWB405 53 , pET28 54 (modified for Gateway-compatible cloning) and pB7FWG2 55 . Constructs were confirmed by restriction mapping and DNA sequence analysis. Information about primers used is described in detail in Supplementary Table S2. Constructs in pGWB405 were used for transient expression in Nicotiana benthamiana leaves 56 . Constructs in pET28 were used for protein expression in Escherichia coli. Constructs in pB7FWG2 were used for stably Agrobacterium tumefaciens-mediated transformation of Arabidopsis thaliana plants (ecotype Col-0) 57 . Homozygous Arabidopsis lines containing a single T-DNA insertion were selected based on the segregation of the phosphinothricin (PPT) resistance marker. Initial segregation analysis was performed by seeding plants on sterile Murashige and Skoog (MS) medium plates supplemented with 20 μ g/mL PPT (Duchefa, Haarlem, Netherlands). A total of 25 PPO M and 29 PPO A homozygous lines were isolated; from which three independent lines of each PPO version were selected for further characterization. Arabidopsis plants were grown on soil in a climate controlled growth chamber (22 °C, 65-70% RH, and 60 μ mol m −2 s −1 photosynthetically active radiation) under 8 h of light/16 h of dark photoperiod for 4 weeks. For phenylpropanoid profiling, Arabidopsis seeds were surface-sterilized and sown on Petri plates with sterile MS medium containing 1% agar and 30 g/L sucrose. When indicated, plates were supplemented with 5 μ M norflurazon. After stratification for 3 days at 4 °C in the dark, the plates were incubated in growth chambers for 20 days at 22 °C under continuous light (130 μ mol m −2 s −1 photosynthetically active radiation).

Expression of recombinant PPO A and PPO M in E. coli cells.
Following transformation of competent E. coli cells strain Rosetta 2 (DE3) (Novagen, Merck KGaA, Darmstadt, Germany) with purified plasmids pET28-PPO A and pET28-PPO M , the activity of PPO A and PPO M was confirmed by growing the transformed E. coli cells for three days at 28 °C on LB agar plates supplemented with 34 μ g/mL chloramphenicol, 25 μ g/mL kanamycin, 0.5 mM IPTG, 40 μ g/ml CuSO 4 , and 600 μ g/ml chlorogenic acid, similarly to methods previously described 58 to detect melanin formation. PPO A and PPO M colonies were dark because the cells were able to produce melanin-like polymers.
Confocal laser-scanning microscopy. Subcellular localization of GFP fluorescence and chlorophyll fluorescence was determined with an Olympus FV 1000 confocal laser-scanning microscope (Olympus, Tokyo, Japan) using an argon laser for excitation (at 488 nm) and a 500-510 nm filter for detection of GFP fluorescence and a 610-700 nm filter for detection of chlorophyll fluorescence.

Measurement of rosette leaves surface area. The surface area of rosette leaves was measured using
Photoshop (Adobe, California, USA) and GraphPad Prism 5.0a (GraphPad Software, California, USA) by analysing digital images (acquired from above) of 4-week-old plants and comparing them with size standard series ranging from 4 to 16 cm 2 .
Transcripts analysis. Leaf samples from 4-week-old plants were harvested and total RNA was extracted using the Maxwell 16 LEV simplyRNA Tissue Kit (Promega, Wisconsin, USA) according to the manufacturer's instructions. Purified RNAs were quantified by spectroscopy using a NanoDrop apparatus (Thermo Scientific, Massachusetts, USA) and RNA integrity was evaluated by agarose gel electrophoresis. The First Strand cDNA Synthesis Kit (Roche, Basel, Switzerland) was used to generate cDNA according to the manufacturer's instructions, using 50 pmol of an anchored poly(dT) primer [d(T 18 V)] and 2 μ g of total RNA. The relative mRNA abundance was evaluated via quantitative reverse transcription PCR (RT-qPCR) in a total reaction volume of 20 μ l using LightCycler 480 SYBR Green I Master (Roche, Basel, Switzerland) on a LightCycler 480 Real-Time PCR System (Roche, Basel, Switzerland) with 0.3 μ M of each specific sense and anti-sense primers. Three independent biological replicates of each sample and three technical replicates of each biological replicate were performed and the mean values were considered for further calculations. The normalized expression of PPO was calculated as described 59 using UBC (Arabidopsis ubiquitin-conjugating enzyme gene At5g25760, GenBank accession number: DQ027035) as the endogenous reference gene. Information about primers used is described in detail in Supplementary Table S2. PPO activity. Protein extracts were obtained from rosette leaves homogenized at a ratio of 50 mg of fresh weight to 1000 μ l of 20 mM phosphate buffer (pH 6). The homogenate was centrifuged at 16.000 × g and 4 °C for 5 min, and the supernatant was used for protein assessment of PPO activity. PPO activity assays were performed in 20 mM phosphate buffer (pH 6) containing 10 mM L-DOPA. The reaction was measured with a SpectraMax M3 spectrophotometer (Molecular Devices, California, USA) by the change in absorbance at 475 nm and 25 °C.
Phenylpropanoid profiling. Phenylpropanoids (specifically, flavonoids) were purified and analysed as described previously 60 with minor modifications. In brief, lyophilized Arabidopsis seedling were homogenized at a ratio of 1 mg of dry weight to 33 μ L of methanol:acetate:H 2 O (9:1:10) extraction solvent containing 0.1 mg/mL of naringenin as an internal standard for quantification. Homogenates were incubated in the dark at 4 °C for 30 min with agitation (1000 rpm) and then centrifuged at 14,000 × g and 4 °C for 5 min. Supernatants were recovered, filtered and analysed by high-performance liquid chromatography (HPLC) using an Agilent 1200 series HPLC equipment (Agilent Technologies, California, USA) with a XSelect CSH C18 (3.5 μ m, 4.6 × 100 mm) column (Waters, Massachusetts, USA). Flavonols and anthocyanins were determined at 320 and 520 nm, respectively. Scientific RepoRts | 6:19036 | DOI: 10.1038/srep19036 Data analysis. ANOVA followed by Newman-Keuls multiple comparison post-hoc tests were used to determine statistical significance for multiple groups. Pearson correlation coefficients (r values) were calculated using the means of the surface area of rosette leaves, PPO mRNA abundance and PPO activity. Statistical analysis was performed using GraphPad Prism 5.0a (GraphPad Software, California, USA).