Phosphate Limitation Induces Drastic Physiological Changes, Virulence-Related Gene Expression, and Secondary Metabolite Production in Pseudovibrio sp. Strain FO-BEG1

Strain FO-BEG1 was cultivated under −P_i and +P_i conditions by using the carbohydrate-mineral (CM) medium, as described previously (20). The medium contained glucose (10 mmol liter⁻¹) as the only carbon source, ammonia (10 mmol liter⁻¹) as the only nitrogen source, and final P_i concentrations of 1.4 mmol liter⁻¹ and 0.1 mmol liter⁻¹ under +P_i and −P_i conditions, respectively. This resulted in ratios of C to N to P of 43:7:1 under +P_i conditions and 600:100:1 under −P_i conditions. The medium was buffered with Tris-HCl, and the pH was adjusted to 8. Erlenmeyer flasks (250 ml) were filled with 100 ml of medium, inoculated with 100 μl of a preculture grown under +P_i conditions, incubated at 28°C in the dark, and shaken at 120 rpm. Bacterial growth was monitored by measurement of optical density at 600 nm (OD₆₀₀) using an Eppendorf BioPhotometer (Eppendorf AG). The OD₆₀₀ was correlated with the cell number determined by using a Thoma counting chamber (Brand GmbH) (data not shown). Micrographs were taken throughout the growth period by using an Axioplan universal microscope (Carl Zeiss GmbH), and at least 100 cells for each time point and under each condition were measured by using the AxioVision Rel. 4.8 imagining system program (Carl Zeiss GmbH). During the growth period, the UV-visible (UV-Vis) spectra of the cell-free supernatants were recorded by using a Beckman DU 640 spectrophotometer (Beckman Coulter).

Glucose and acetate concentrations were determined by using a high-performance liquid chromatography (HPLC) system (Sykam GmbH), as described previously by Bondarev et al. (18). P_i concentrations were determined colorimetrically by the ascorbic acid method (21), using a SpectroDirect spectrophotometer (Aqualytic). The activity of alkaline phosphatase (AP) during bacterial growth was evaluated by monitoring the degradation of 4-para-nitrophenyl phosphate bis(tris) salt (pNPP) (purity, ≥97.00%; Sigma-Aldrich). Experiments were performed independently twice, always analyzing biological triplicates. The level of tropodithietic acid (TDA) was measured by means of reverse-phase HPLC (RP-HPLC), using pure TDA as a standard (purity of ≥98%; BioViotica Naturstoffe GmbH). The presence and quantification of polyphosphate were investigated by staining cells with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI), performed according to the method described previously by Kulakova et al. (22). Nile red was used for staining of polyhydroxyalkanoate (PHA) inclusions. Polar lipid analysis was carried out by the identification service of the DSMZ and B. J. Tindall (DSMZ, Braunschweig, Germany). Detailed information about materials and protocols can be found in the supplemental material.

To investigate the effect of P_i limitation on protein expression, we chose a mass spectrometry (MS) shotgun proteomic approach coupled with isotope-coded protein labeling (ICPL) (23), performed on cytoplasmatic, membrane, and extracellular proteins. Biological duplicates were analyzed for the cytoplasmatic proteome, whereas data for the membrane and the extracellular proteomes were obtained after combining biological triplicates. In all proteomic analyses, samples collected during the logarithmic phase (T₁) under −P_i conditions (−P_i T₁) were used as the reference for calculation of the relative protein abundances. For the cytoplasmatic proteome, we compared the protein expression levels of cells growing under −P_i and +P_i conditions collected during the logarithmic phase (−P_i T₁ versus +P_i T₁) (OD₆₀₀ of 0.51 ± 0.03 for −P_i T₁; OD₆₀₀ of 0.78 ± 0.04 for +P_i T₁). Additionally, in order to monitor the effect of prolonged P_i limitation, we compared the protein expression levels of logarithmic-phase and stationary-phase (T₂) cells growing under −P_i conditions (−P_i T₁ versus −P_i T₂) (OD₆₀₀ of 1.29 ± 0.04 for −P_i T₂). Cells growing logarithmically under P_i limitation were collected at the same growth stage where we previously detected the first increase in AP activity. The proteome analyses conducted on the membrane and extracellular proteins were performed according to the same rationale in an independent experiment. In addition, for these two proteomes, a comparison of the protein expression levels between logarithmic-phase cells growing under −P_i conditions and stationary-phase cells growing under +P_i conditions was performed (−P_i T₁ versus +P_i T₂) (OD₆₀₀ of 2.16 ± 0.01 for +P_i T₂). An overview of the ICPL scheme and of the sample comparisons can be found in Table S1 in the supplemental material.

Proteins were labeled at the free amino groups with one of the four isotopic forms of nicotinoyloxysuccinamide. The complexity of the samples was then reduced via SDS-PAGE, and the proteins were digested into peptides and analyzed via liquid chromatography-electrospray ionization-tandem MS (LC-ESI-MS/MS). Since peptides with identical amino acid sequences derived from the four differentially labeled samples differ in mass, they appeared as sister peaks in the acquired MS spectra (multiplets; complete quadruplets if the peptides were detected in all four samples and incomplete quadruplets if not), with mass shifts corresponding to the difference between the masses of the labels (23). Peptides were then assigned to proteins by using the MASCOT suite, and the fold change of each protein was calculated as the median of the abundance ratios of the sister peptides. A fold change of ≤−1.75 or ≥1.75 was regarded as significant regulation. The regulated proteins were then divided into 21 categories according to their Clusters of Orthologous Groups (COG) classification. All upregulated proteins were used to perform a gene set enrichment analysis using their respective Gene Ontology (GO) annotations. Finally, for selected proteins of interest, we performed either a phylogenetic reconstruction or an analysis to identify their domain content. The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium (24) via the PRIDE partner repository with the data set identifiers PXD001621 and 10.6019/PXD00162 for the extracellular and membrane proteomes and PXD001624 and 10.6019/PXD001624 for the cytoplasmatic proteome. Detailed descriptions of the experimental procedures and the data analyses are given in the supplemental material.

The in silico screening of the Pho box in the genome of strain FO-BEG1 was based on methods described previously by Yuan et al. (25). The Pho box sequences proposed by those authors were used to build a position weight matrix (26) utilized to scan the intergenic regions of strain FO-BEG1. From this first scan, an initial gene list was obtained, and from this list, a total of 29 genes (see Table S2 in the supplemental material), with the best alignments against the position weight matrix and upregulated as determined by the proteomic approach, were selected and used to build a new position weight matrix specific for Pseudovibrio sp. strain FO-BEG1. The latter was then used to obtain the final list of genes that present a putative Pho box in their upstream regions. A detailed description of the method is provided in the supplemental material.

P_i limitation significantly depressed bacterial growth, leading to a final cell density 2.5 to 3.5 times lower than that observed under +P_i conditions (Fig. 1). The cultures growing under conditions of P_i surplus were characterized by a doubling time (t_d) of 4.33 h and a growth rate (μ) of 0.16 h⁻¹. P_i-limited cultures had similar growth during the first half of the logarithmic phase (t_d of 4.77 h; μ of 0.15 h⁻¹), but in the second half, when P_i was completely taken up by the cells (Fig. 1), the t_d increased to 7.68 h, and the μ dropped to 0.09 h⁻¹. Cells growing under +P_i conditions did not metabolize all P_i provided, whereas P_i was completely taken up under −P_i conditions during the first 30 h of growth (Fig. 1). After this growth stage, we detected a considerable increase in the activity of AP (see Fig. S1 in the supplemental material), which was negligible in cultures growing under +P_i conditions (data not shown).

Regardless of the difference in growth, glucose was metabolized completely under both conditions (Fig. 2). Under P_i limitation, uptake continued during the nonproliferating period, indicating a possible accumulation of carbon storage compounds. Consistently, we detected the presence of lipophilic inclusions, likely polyhydroxyalkanoate (PHA), starting after ∼37 h of growth (Fig. 2). Cells growing under P_i limitation underwent a progressive increase in cell length throughout the growth period. We detected cells up to 12 μm in length, and this elongation led to a pronounced increase in the cell surface (from 8.95 ± 1.59 μm² to 15.34 ± 3.49 μm²) (see Fig. S2 in the supplemental material).

During the proteomic analyses, we identified 1,862 and 1,866 proteins in the two replicates used for the cytoplasmatic proteome. In addition to these proteins, 1,809 and 258 proteins were identified for the membrane and the extracellular proteomes, respectively. For each proteome, the false discovery rate (27, 28) was always <0.5%. In the two replicates of the cytoplasmatic proteome, 655 and 651 proteins were quantified, and in total, according to the criteria we applied, 315 proteins were found to be regulated. In the extracellular and membrane proteomes, 80 and 364 proteins were quantified, respectively. Considering the low number of quantified proteins in the extracellular fraction and considering that the extracellular and membrane proteome analyses were performed by using the same bacterial cultures, we decided to combine these two data sets. In these two proteomes, 311 proteins were found to be regulated. The complete list of regulated proteins can be found in Tables S3 to S7 in the supplemental material.

Overall, in the comparison between −P_i T₁ and +P_i T₁, the higher number of upregulated proteins was observed for COG categories E, G, C, and O (Fig. 3A and C), whereas in the comparison between −P_i T1 and −P_i T2, the majority of upregulated proteins belonged to COG category J (Fig. 3B and D). The gene set enrichment analyses performed by using the Gene Ontology classification of the upregulated proteins underlined that many processes related to phosphorus metabolism were significantly enriched in the logarithmic phase under −P_i conditions (T₁) (see Fig. S3 in the supplemental material). Additionally, in both proteomic data sets, many genes involved in transmembrane active transport were significantly enriched under −P_i T₁ conditions. Also, the categories referring to proteins involved in protein synthesis were significantly enriched in the comparison between −P_i T₁ and −P_i T₂, reflecting the decline of cell proliferation in the stationary phase (T₂) (see Fig. S3 in the supplemental material).

The in silico analysis of the PhoB binding site revealed that strain FO-BEG1 possesses an 18-nucleotide Pho box, characterized by two 5′-CTGTCAT-3′ repetitions separated by a region of 4 nucleotides (see Fig. S4 in the supplemental material). In the first repetition, the first position does not show a high degree of conservation. However, due to the absence of molecular data that could elucidate the Pho box structure of this bacterium, we assumed that it resembled the well-conserved 18-nucleotide sequence known for E. coli and Sinorhizobium meliloti (25). The list of all genes exhibiting a putative Pho box in their upstream regions is reported in Table S8 in the supplemental material.

In order to decrease the bias due to the growth phase and focus on the effect of P_i limitation on Pseudovibrio sp. FO-BEG1 physiology, in the rest of this report, we refer to the comparison between −P_i T₁ and +P_i T₁, unless otherwise stated. As reported previously for other bacteria (7, 13), in strain FO-BEG1, P_i limitation triggered the upregulation of several genes coding for proteins involved in P uptake and metabolism (COG category P; GO categories “phosphorelay response regulatory activity,” “organophosphate ester,” “organophosphonate,” and “inorganic phosphate transmembrane transport activity”) (Fig. 3A, C, and E; see also Fig. S3 and Tables S4 and S5 in the supplemental material). Among these proteins, we detected three out of four subunits (PSE_1688, PSE_1690, and PSE_1691) of the Pst high-affinity ABC transport system for P_i. In proximity to its gene cluster, there is a gene (PSE_1692) encoding a protein homologous to PhoU, expressed only under −P_i conditions. This gene belongs to the pst operon E. coli, and the encoded protein appears to play a role in signal transduction (7). Consistently, a PhoH-like protein (PSE_0499), described to be part of the Pho regulon in E. coli (7), was highly upregulated under P_i limitation. Finally, we detected a >11-fold upregulation of PhoB (PSE_1693), which is the transcription factor of the two-component regulatory system (TCRS) PhoR-PhoB, which governs the phosphate starvation response. A putative Pho box was identified in the upstream region of the first gene of the pst gene cluster, PSE_1688 (see Tables S4 and S5 in the supplemental material).

After the complete consumption of the provided P_i in −P_i cultures and the consequent increase in AP activity, we still observed cell proliferation (Fig. 1; see also Fig. S1 in the supplemental material). Since no other source of P was provided, we quantified the stored polyphosphate (poly-P_i), previously detected in the cells, in order to verify whether it was used as a P source. Indeed, under −P_i conditions, the amount of stored poly-P_i decreased from 87.3 ± 3.4 fg cell⁻¹ to 30.2 ± 2.4 fg cell⁻¹ immediately after complete P_i uptake (see Fig. S5 in the supplemental material). Consistently, under these conditions, we detected a >3-fold-upregulated protein homologous to a poly-P_i kinase 1 (Ppk1 [PSE_2769]), which catalyzes the reversible synthesis of poly-P_i from ATP (29).

Two proteins, PhnD (PSE_3629) and PhnC (PSE_3630), of the phosphonate ABC transport system were strongly upregulated, together with two proteins involved in the degradation of these molecules (PhnJ [PSE_4852] and PhnM [PSE_4857]) and three subunits of the ABC transport system for sn-glycerol-3-phosphate (UgpB [PSE_0472 and PSE_0680] and UgpC [PSE_0683]). Potential Pho boxes were identified in the upstream regions of the PSE_0680, PSE_3629, and PSE_3630 genes. Finally, we detected a strong upregulation of three nucleosidases (PSE_1783, PSE_1587, and PSE_2573), and for two of these nucleosidases (PSE_1587 and PSE_2573), a potential Pho box was identified within the promoter region of their genes.

Thin-layer chromatography (TLC) analysis of the cellular polar lipids revealed that under +P_i conditions, most lipids contained P, whereas under P_i limitation, amino lipids (ALs) and glycolipids (GLs) were dominant (Fig. 4). Consistently, under P_i limitation, several proteins involved in cell envelope biogenesis and lipid metabolism were upregulated (COG categories I and M) (Fig. 3A, C, and E; see also Tables S4 and S5 in the supplemental material). Among these proteins, we detected a >2-fold-upregulated protein containing a phosphatidic acid phosphatase type 2 domain (PSE_2249) that is conserved in enzymes involved in the catabolism of phospholipids. Additionally, in the upstream region of its gene, a potential Pho box was identified (see Tables S4 and S5 in the supplemental material). Also, a protein homologous to PlsX (PSE_3572), described to be involved in membrane lipid and phospholipid metabolism (30), was significantly upregulated in one replicate of the cytoplasmatic proteome. Finally, we detected an upregulated protein homologous to acyl coenzyme A (acyl-CoA) dehydrogenase (PSE_0035) and two upregulated proteins homologous to enoyl-CoA hydratase (PSE_0406 and PSE_3595) (COG category I) (Fig. 3; see also Tables S4 and S5 in the supplemental material), both of which are involved in the oxidation of fatty acids and could be required for the degradation of the exchanged lipids.

Among the cellular polar lipids detected under P_i limitation, four different types of GL were identified (GL1 to GL4) (Fig. 4). One GL (GL4) had a retention factor consistent with the S-containing lipid sulfoquinovosyl diacylglycerol (SQDG). In one of the proposed pathways for the synthesis of SQDG, the biosynthetic process starts from UDP-glucose, which is then converted into SQDG via the intermediate formation of UDP-sulfoquinovose (31). All genes required for the biosynthesis of these lipids were found in the genome of Pseudovibrio sp. FO-BEG1. Consistently, under −P_i conditions, we detected a >2-fold-upregulated protein homologous to a UTP-glucose-1-phosphate uridylyltransferase (PSE_3207) that is responsible for the synthesis of UDP-glucose and a 22-fold-upregulated protein homologous to SqdB that is responsible for the formation of UDP-sulfoquinovose (PSEp_0373) (see Tables S4 and S5 in the supplemental material). Intriguingly, a potential Pho box was identified upstream of a second copy of the sqdB gene, PSE_2321 (see Table S8 in the supplemental material).

Consistent with the rearrangement of the membrane composition, we detected the upregulation of two proteins homologous to the phage shock protein PspA (PSE_4192 and PSE_4593). These proteins are members of the phage shock regulon that is induced under membrane stress conditions, and they regulate the transcription of enzymes involved in the maintenance of membrane stability, proton motive force, and protein secretion (32). Under −P_i conditions, we detected an upregulated protein homologous to TolB (PSE_4678) belonging to the Tol-Pal complex and a bacterial outer membrane protein containing an OmpA domain (PSE_1259) described to interact with the Tol-Pal system. Different functions were proposed for this system, but it seems to be involved mainly in the maintenance of cell envelope stability (33). Additionally, several proteins likely to be involved in the biosynthesis of lipopolysaccharide and peptidoglycan were also upregulated (PSE_4236, PSE_3471, PSE_3402, and PSE_3559) (see Tables S4 and S5 in the supplemental material).

In accordance with the lipophilic inclusions detected via Nile red staining, under −P_i conditions, two phasin-like proteins (PhaP [PSE_1257 and PSE_3807]) were upregulated. These proteins are associated with PHA granules and have been shown to affect polymer synthesis (34). We identified a Pho box in the upstream region of the PSE_3807 gene, indicating that its transcription is likely under the direct control of PhoB. Consistently, the formation of inclusions started after 37 h of growth, when P_i in the medium was completely taken up by the cells (Fig. 1 and 2). Additionally, a protein homologous to PhaR (PSE_0173) was >3-fold upregulated under −P_i conditions. PhaR is a transcription regulator, and there is evidence suggesting that the expression of the phasin genes (phaP) and phaR genes is derepressed under PHA-accumulating conditions (35). Finally, a protein homologous to an acetyl-CoA acetyltransferase (PSE_0174) was also upregulated. This protein shared 68% similarity (99% query coverage; E value of 0.0) with the PhbA protein of S. meliloti strain Rm41 that is involved in the first steps of PHA synthesis (36).

P_i limitation also triggered significant changes in the central metabolism. Under −P_i conditions, citrate synthase (PSE_3391), the first enzyme in the tricarboxylic acid (TCA) cycle, was downregulated (see Tables S4 and S5 in the supplemental material). Instead, two enzymes involved in the acetate kinase-phosphotransacetylase (Pta-AckA) pathway (PSE_1086 and PSE_1087), in which acetyl-CoA is converted into acetate via acetyl-P_i, were upregulated. In agreement with this, we observed acetate production starting during the second half of the exponential phase (0.15 ± 0.02 mmol liter⁻¹) and reaching maximum levels (0.73 ± 0.09 mmol liter⁻¹) at the beginning of stationary phase. The acetate was subsequently taken up by the cells during the rest of the growth period (data not shown). A potential PhoB binding site was identified in the upstream region of the PSE_1086 gene, suggesting that the Pta-AckA pathway is directly regulated in response to P_i limitation.

As outlined by the GO enrichment analysis where the category “peroxidase activity” was significantly enriched among the upregulated proteins under P_i limitation (see Fig. S3 in the supplemental material), several proteins potentially involved in the oxidative stress response were upregulated under −P_i conditions. For example, we detected an upregulated protein having an alkyl hydroperoxide reductase (AhpC)/thiol-specific antioxidant (TSA) family domain (PSE_0180) that can control peroxide levels, and in Mycobacterium tuberculosis, it was described to in turn be reduced by an alkylhydroperoxidase (AhpD) (37). Consistently, a protein homologous to the latter was also upregulated under −P_i conditions (PSE_0181). Interestingly, Ahp proteins were shown to be specifically required for E. coli viability under conditions of prolonged P_i starvation (38). Two proteins homologous to bacterioferritin (Bfr [PSE_1030 and PSE_3844]), which are responsible for iron storage and protection against oxidative stress via iron detoxification (39), were also upregulated, and both their genes presented a Pho box in their upstream regions.

Under −P_i conditions, two proteins homologous to PotD and PotA (PSE_1679 and PSE_1681), which belong to the Pot ABC transporter system for putrescine and spermidine, were highly upregulated. The latter molecules are polyamines and play an important role in the protection of macromolecules against radicals (40). In addition, the potD gene presented a Pho box in its upstream region (see Tables S4 and S5 in the supplemental material), which is consistent with data from previous studies conducted on S. meliloti (25). Finally, a protein responsible for the synthesis of glutathione (PSE_0381), a molecule which can help cells protect macromolecules via scavenging reactive oxygen species (ROS), was also upregulated under −P_i conditions. Protection against oxidative stress also requires a secondary line of defense consisting of enzymes able to repair damaged macromolecules (41). Indeed, in strain FO-BEG1, P_i limitation also triggered the upregulation of several proteases, peptidases, chaperone-like proteins, and proteins involved in DNA repair (COG categories O and L) (Fig. 3A, C, and E; see also Tables S4 and S5 in the supplemental material).

Surprisingly, P_i limitation triggered the secretion of compounds that gave a yellow coloration to the cultures and the cell-free supernatants. The latter showed a characteristic UV-visible absorption spectrum, with a maximum absorbance at ∼320 nm and a broad shoulder at 400 nm (see Fig. S6 in the supplemental material). The color appeared when cells entered stationary phase, and its intensity increased during the remaining incubation time. RP-HPLC analysis revealed the presence of a compound with a retention time and a spectrum consistent with those of tropodithietic acid (TDA), an antibiotic that is produced, together with a yellow-brown pigment, by many bacteria belonging to the Roseobacter clade (42). Its production was detected only under −P_i conditions (1.55 mg liter⁻¹) and only during the stationary phase.

P_i limitation also influenced the expression of protein homologues, described for other bacteria, to be involved in secretion, cell-cell interactions, and virulence. A protein (PSE_2475) annotated as a transcription regulator belonging to a TCRS was upregulated >2-fold. This protein shared 44% similarity (query coverage, 96%; E value, ≤1e⁻⁵⁴) with the transcriptional regulator PhoP from E. coli and Salmonella enterica, and the phylogenetic analysis that we performed showed that PSE_2475 clusters together with PhoP sequences of S. enterica (see Fig. S7 in the supplemental material). PhoP is a member of the TCRS PhoP-PhoQ, which has been well studied in Salmonella strains and was shown to be activated by different environmental cues, and it regulates the expression of virulence genes (43).

A protein homologous to TolC (PSE_3720) was found to be >2-fold upregulated under −P_i conditions. These proteins are versatile outer membrane transport proteins described to be involved in the efflux of noxious molecules and the secretion of antibiotics or virulence factors in other bacteria (44). The tolC gene is located in a genomic region encoding proteins for a type 1 secretion system (T1SS). Adjacent to this region is the largest gene of the whole genome (PSE_3716), containing five T1SS-143 repeat domains. This domain of about 143 amino acids is found in proteins of the genera Vibrio and Legionella, which share properties with RTX (repeats in toxin) proteins. These proteins are secreted by Gram-negative bacteria via the T1SS and are described to be virulence factors, having cytotoxic functions, proteolytic and lipolytic activities, or roles in adhesion and biofilm formation (45). Interestingly, PSE_3716 was upregulated in one replicate under −P_i conditions, and in agreement with data reported previously by Yoshida et al. (46), the whole genomic region seems to be under the direct control of PhoB, since the tolC gene had a potential Pho box within its promoter region (see Table S8 in the supplemental material). Another member of the RTX protein family, a hemolysin-like protein (PSE_4352), was upregulated >3-fold, and a Pho box was identified in the upstream region of its gene. Similarly, this protein was upregulated >9-fold in the comparison between −P_i T₁ and +P_i T₂ (see Table S7 in the supplemental material). In this comparison, a second hemolysin-like toxin, PSEp_0144, was also upregulated >2-fold. Both hemolysin-like proteins and the RTX-like protein contained one or more serralysin-like metalloprotease domains (see Fig. S8 in the supplemental material). Serralysins are endopeptidases that act as virulence factors and can cause tissue damage. A similar domain was also present in another upregulated protein, PSE_4952, which, like the large toxin PSE_3716, also presented a cadherin domain (see Fig. S8 in the supplemental material). Cadherins are an important family of calcium-dependent cell-cell adhesion molecules that play key roles in developmental processes and maintenance of tissue architecture, and they are implicated in the regulation of signaling events throughout the animal kingdom (47). Finally, we detected a 16-fold upregulation of a protein homologous to invasion-associated locus B (IalB [PSE_3194]), which is known to be a major virulence factor in Bartonella bacilliformis and is upregulated in response to environmental cues signaling vector-to-host transmission (48).

By comparing the protein expression levels between the logarithmic and the exponential phases under −P_i conditions, we observed that only a minority of proteins (14%) were upregulated during the stationary phase (Fig. 3B and D). Consistent with the reduction of cell proliferation (Fig. 1), many of these proteins were involved in the repression of DNA replication and gene transcription (COG categories K and L) (Fig. 3B and D; see also Tables S3 and S6 in the supplemental material). The lipid rearrangement seemed to continue during stationary phase, as a protein containing a phosphatidic acid phosphatase type 2 domain (PSE_2249) and a protein homologous to phospholipase D (PSE_0624), both likely involved in the degradation of the phospholipids, were more abundant than in the logarithmic phase. Similar fold changes were also observed for the two phasin-like proteins (PSE_1257 and PSE_3807). Interestingly, a protein homologous to a major ferric iron binding protein (FbpA [PSE_1253]) was >10-fold upregulated during the stationary phase. These proteins are part of the ABC system for the transport of iron(III) that was also shown to be significantly regulated in response to P_i limitation in S. meliloti (13).

P_i has essential metabolic, structural, and regulatory roles in all bacteria (11, 49). The environmental P_i concentration influences not only P metabolism but also the bacterial life-style and interspecies and interkingdom relations, by affecting secondary metabolite production and virulence gene expression (9, 10). Previous studies have addressed the effect of P_i limitation on, e.g., the antibiotic production of Streptomyces spp., on bacteria pathogenic for humans, or on plant-associated bacteria that are of agricultural interest (9, 13, 25). Much less is known about the response of heterotrophic marine bacteria to P_i limitation, despite growing evidence suggesting that this condition is common in many areas of the ocean (2, 3).

Figure 5 shows a schematic overview of the physiological adaptations observed in Pseudovibrio during growth under P_i limitation. As described previously for other strains (8, 13, 14), P_i limitation significantly depressed bacterial growth, and this was reflected in the downregulation of proteins involved in DNA replication and protein synthesis (COG categories C and E) (Fig. 3A, C, and E; see also Tables S4 and S5 in the supplemental material). The growth rate dropped concomitantly with the complete consumption of P_i (Fig. 1), and a consistent increase in AP activity was observed immediately afterwards (see Fig. S1 in the supplemental material). Different APs were identified in the genome of FO-BEG1 (e.g., PSE_1012 and PSE_2813), but none were quantified during the proteomic analysis. APs are enzymes that cleave P_i groups in organic molecules containing phosphoester bonds, and their expression has been widely used as an indicator of P limitation (50). Considering these data together, it is reasonable to state that in FO-BEG1, the P_i limitation response is activated at this growth stage, with an environmental P_i concentration of ∼4 μmol liter⁻¹ (see Fig. S1 in the supplemental material). These data are consistent with what has been previously described for E. coli (7). The strategy used by FO-BEG1 to scavenge P_i from the environment is characterized by the upregulation of ABC transporters for P_i and proteins involved in the uptake and degradation of organic molecules containing P. Genes encoding these proteins are typically present and expressed in bacteria adapted to thriving in marine regions with limited amounts of P_i (51). Even though the P_i concentration can be far below 4 μmol liter⁻¹ in these regions, the respective C to N to P ratios are often comparable to the values used in our experiments (4, 52). Altogether, these data underline the versatility of Pseudovibrio, and as far as P metabolism is concerned, they indicate that it is well adapted to thrive as a free-living organism in the open ocean, where P_i availability is often limited and the DOP can represent the only bioavailable P pool (1).

The cell proliferation observed after the consumption of P_i was likely sustained by the degradation of stored poly-P_i, known to be an optimal intracellular P source (53). In the genome of strain FO-BEG1, several genes encoding proteins involved in poly-P_i degradation are present (e.g., PSE_2770). However, none could be quantified during the proteomic analyses. Instead, we detected upregulated Ppk1, which is an enzyme involved in the reversible synthesis of poly-P_i (29), and as described previously for other species, it can also be responsible for degradation (54, 55).

In E. coli and S. meliloti, the P_i limitation response is controlled by the TCRS PhoR-PhoB. Both genes were detected in the genome of FO-BEG1 (PSE_1693 and PSE_1687), and PhoB was highly upregulated under −P_i conditions. Moreover, we detected many similarities to other organisms concerning the regulated proteins and the genes that presented a potential Pho box in their promoter regions. Therefore, it is reasonable to state that in Pseudovibrio, the response to P_i limitation is controlled by the PhoR-PhoB system.

We showed that P_i limitation also greatly affects the carbon metabolism of Pseudovibrio sp. FO-BEG1, confirming the importance of P in the overall regulation of cell physiology. P_i-limited cells took up all glucose provided, and due to the lower cell density, the rate of consumption of carbon per cell was consequently higher than that in +P_i cultures. Our data suggest that some carbon might have been used for the synthesis of the inclusion PHAs, which might be accumulated by bacteria to increase survival in changing environments or as a sink for redundant reducing power under conditions of nutrient limitation and carbon surplus (56, 57). The observed PHA accumulation and the upregulation of genes involved in PHA synthesis, accompanied by the presence of a Pho box in the promoter region of one of these genes (PSE_1257, PSE_3807, PSE_0173, and PSE_0174) (see Tables S4 and S5 in the supplemental material), suggest that, as observed for Acinetobacter species (58), PHA accumulation can also be part of the Pho regulon in marine organisms.

Pseudovibrio underwent pronounced cell elongation under P_i limitation. This phenomenon has been observed frequently for P_i-starved cells and can be due to the activation of the SOS stress response (59–61). Longer cells would have an increased surface area for nutrient uptake but also increased cellular P demand for the synthesis of membrane phospholipids. However, we showed a rearrangement of the cellular lipid composition in favor of P-free lipids (Fig. 4). Changes of the membrane phospholipids were suggested to activate the psp response, which controls enzymes involved in membrane stability and the maintenance of proton motive force (32). Therefore, the upregulation of PspA and other proteins involved in the maintenance of cell envelope stability supports the hypothesis that Pseudovibrio experiences membrane stress during growth under P_i limitation, likely due to lipid rearrangement.

In oceanic surface waters, P uptake for phospholipid synthesis can represent an important part of the total bacterial P_i incorporation (62), since P_i present in phospholipids can constitute up to 36% of the total cellular P (63). During P_i limitation, levels of phospholipids can decrease by up to 97% (64, 65); therefore, membrane lipid rearrangements provide Pseudovibrio with an additional P source and at the same time decrease its cellular demand, resulting in a competitive advantage in environments depleted of this nutrient. SQDGs can play an important role in membrane restructuring under P_i-limited conditions (17, 66), but to the best of our knowledge, this is the first study where SQDG production was observed as a direct response to P_i limitation in heterotrophic marine bacteria. Our data support recent molecular evidence suggesting that the ability of heterotrophic marine bacteria to produce SQDG might be more widespread than previously thought (67).

As observed for other bacterial strains (14, 38, 68), under −P_i conditions, we detected several proteins potentially involved in protection against oxidative stress (see Tables S4 and S5 in the supplemental material). So far, it has not been clarified why cells growing under P_i limitation experience increased oxidative stress. It can be speculated that, as also suggest by Yuan et al. (68), the physiological changes induced by P_i limitation, especially the membrane lipid rearrangements, could increase cellular sensitivity to ROS, requiring the activation of defense mechanisms.

P_i limitation greatly influenced the metabolites secreted by Pseudovibrio sp. FO-BEG1 (see Fig. S6 in the supplemental material). We previously showed that P_i limitation drastically affected both the amount and composition of compounds secreted by the strain during growth (20). Among these compounds, in the present work we identified the potent antibiotic TDA, which is a broad-spectrum antibiotic for which stable resistant mutants have been found to be difficult to select (69). Consequently, TDA production confers an ecological advantage to Pseudovibrio when competing for limited resources. Moreover, the production of the antibiotic can also be beneficial for the hosts, providing protection against potentially pathogenic bacteria. It is worth considering that the presence of sulfur in the molecules of TDA and SQDGs suggests that P_i limitation also affects the overall sulfur metabolism and sulfur demand of Pseudovibrio sp. FO-BEG1. This speculation is supported by the higher number of compounds containing sulfur detected in the exometabolome of this strain growing under P_i limitation (20).

The PhoR-PhoB system and the Pho regulon have been shown repeatedly to influence bacterial virulence (10, 70, 71). Accordingly, we detected several upregulated protein homologues, which have been described to be involved in prokaryote-eukaryote interactions in other bacteria. For example, TolB, OmpA, and PspA can be involved in adhesion to and invasion of host cells and interaction with the host defense mechanism and were reported to be upregulated during macrophage infection and biofilm formation (72–74). Several virulence factors are regulated by the TCRS PhoQ-PhoP, which has been well characterized in Salmonella species (43). The upregulated PSE_2475 protein formed a unique cluster with the transcription regulator PhoP from Salmonella enterica in the phylogenetic reconstruction that we performed (see Fig. S7 in the supplemental material), suggesting a similarity in function. In FO-BEG1, PhoP could have been activated by acetyl-P_i, derived from the upregulated Pta-AckA pathway. Acetyl-P_i can activate the response regulator proteins of TCRSs, independently of the respective histidine kinases, and was shown to directly activate PhoP (75). Therefore, it was suggested that the overexpression of the Pta-AckA pathway is a way to connect P_i limitation with the general induction of a large number of genes (76, 77).

The upregulation of TolC and the RTX-like toxin (PSE_3716) and the presence of a Pho box in the promoter region of the tolC gene are strong indications that the whole genomic region, which also contains two subunits required for a functional T1SS (hlyD [PSE_3718] and hlyB [PSE_3719]), is under the direct control of PhoB and is upregulated under P_i limitation. The RTX-like toxin possesses a serralysin-like metalloprotease domain and two galactose binding domains (see Fig. S8 in the supplemental material), which could interact with glycosylated groups of proteins present on the eukaryotic cell surface, such as cadherins and integrins. Consistently, integrin- and cadherin-like proteins were also described previously for sponges and corals (78, 79), which are the two major hosts of Pseudovibrio strains. This suggests that, as observed for other bacteria (80), FO-BEG1 could use this type of toxin to adhere to the host cell surface and penetrate via disruption of the host tissue. Serralysin-like domains were identified also in other proteins upregulated under −P_i conditions, the hemolysins PSE_4352 and PSEp_0144 and the PSE_4952 protein (see Fig. S8 in the supplemental material). Additionally, cadherin- and integrin-like domains were identified in both RTX-like proteins and the PSE_4952 protein, respectively. Cadherins and integrins have been shown to mediate cell-cell and cell-matrix interactions in metazoans, and proteins containing these domains have also been identified in different bacterial strains (81, 82). As in strain FO-BEG1, cadherins and integrins are found in bacterial proteins that have additional enzymatic and adhesion domains (81). Interestingly, it was shown previously that proteins containing these domains are involved in bacterial cell interactions, biofilm formation, adherence, interactions with the extracellular matrix, and eukaryotic cell toxicity (82–85). These data support our hypothesis that Pseudovibrio sp. FO-BEG1 could use these proteins, together with IalB, to adhere to and interact with the host cell.

Altogether, these data indicate that P_i limitation and the general stress response induced by P_i limitation enhance the expression of proteins involved in the establishment and maintenance of a symbiotic relationship with marine invertebrates such as sponges. Sponges are filter-feeding organisms capable of pumping hundreds of liters of seawater per day (86). Consequently, the colonization of such organisms might represent a strategy adopted by Pseudovibrio to overcome nutrient limitation encountered as a free-living organism in the surrounding water, since the association with the sponge will expose the bacterium to a continuous flow of water and nutrients.

In this work, we describe for the first time in detail the response of a marine heterotrophic bacterium to P_i limitation. We show that this nutrient regime has a pleiotropic effect on the physiology of Pseudovibrio sp. FO-BEG1. This suggests that in marine systems, such conditions could also shape the biogeochemical cycles of other nutrients by drastically affecting bacterial physiology. Additionally, our data show that P_i limitation has a significant influence on the production of secondary metabolites. Considering the interest in new bioactive molecules of marine origin and in the Pseudovibrio genus for exploitation of novel compounds, our work indicates that P_i limitation can represent a promising starting condition for triggering the production of these molecules. Finally, the proteomic analyses that we performed pointed out how P_i limitation, or the general stress response induced by it, also affects the expression of virulence-related genes in Pseudovibrio. Further molecular work could unravel the regulatory connections between the environmental nutrient regimes and the ability of these bacteria to colonize their hosts.