Caulobacter crescentus Adapts to Phosphate Starvation by Synthesizing Anionic Glycoglycerolipids and a Novel Glycosphingolipid

Phosphate starvation induces elongation of both the cell body and stalk in C. crescentus (Fig. 1A). If we approximate the shapes of the cell body and stalk as cylinders, we can estimate that the total surface area of the cell, and thus the synthesis of membrane lipids, increases 6- to 7-fold upon phosphate limitation. This significant increase in membrane area suggests that C. crescentus must be able to produce alternatives to phospholipids during phosphate starvation. Indeed, when we compared the total lipid composition under phosphate-rich (1 mM phosphate) and phosphate-starved (1 µM phosphate) conditions by using normal phase liquid chromatography-tandem mass spectrometry (LC-MS/MS), we saw a dramatic decrease in phosphatidylglycerol and a corresponding increase in several species of glycolipids (Fig. 1B). In particular, phosphate starvation induced the synthesis of mono- and diglycosyl diacylglycerols as well as a GSL, hexose-hexuronic acid-ceramide (HexHexA-Cer; C₁₆ ceramide [d18:1/16:0]) (Fig. 1B). These lipid species were identified by exact mass measurement in conjunction with collision-induced dissociation (CID) tandem mass spectrometry. The presence of a GSL was unexpected since sphingolipids, while highly abundant in eukaryotes, are rarely found in bacteria. Some species in the Bacteroides, Porphyromonas, and Prevotella genera are capable of synthesizing the phosphosphingolipids ceramide phosphorylethanolamine and/or ceramide phosphoglycerol (13). Non-phosphate-containing GSLs have been described only for the family Sphingomonadaceae, in which they can function as a substitute for lipopolysaccharide (LPS) (14). Structural analyses of Sphingomonas GSLs have revealed carbohydrate moieties containing 1, 3, or 4 sugar units (Fig. 1C) (14, 15). In contrast, the GSL found in C. crescentus has two sugars and thus represents a novel bacterial GSL species, which we named GSL-2 (Fig. 1C).

Sphingolipid synthesis begins with the production of a ceramide molecule which is then modified with various polar groups. In eukaryotes, ceramide synthesis is a four-step process that begins with the condensation of serine and a fatty acyl coenzyme A (acyl-CoA) to produce 3-oxo-sphinganine in a reaction catalyzed by oxoamine synthase (16). While each of the enzymatic steps in ceramide synthesis has been well characterized in eukaryotes (17), only the oxoamine synthase enzyme required for the first step appears to be conserved in bacteria. Indeed, bacterial species have been identified encoding one or more oxoamine synthases, including serine palmitoyltransferase (Spt), 8-amino-7-oxononanoate synthase (BioF), 5-aminolevulinate synthase (HemA), and 2-amino-3-oxobutyrate coenzyme A ligase (Kbl) (18). C. crescentus encodes three putative oxoamine synthases: CCNA_01220 (BioF), CCNA_01417 (HemA), and CCNA_01647 (BioF).

To assess the role of the candidate oxoamine synthases in ceramide synthesis, total lipid composition was analyzed in wild-type, Δccna_01220, and Δccna_01647 cells. We were unable to obtain a deletion of ccna_01417 since this is an essential gene (19); however, recent biochemical analysis of CCNA_01417 reveals that it is most likely involved in the production and regulation of heme cofactors (20). Ceramides were completely absent in the Δccna_01220 strain (Fig. 2); therefore, we refer to this gene as ccbF, for Caulobacter crescentus BioF. Complementation of the ccbF deletion restored ceramide synthesis (see Fig. S1A in the supplemental material). Interestingly, deletion of BioF homolog CCNA_01647 had no effect on ceramide synthesis despite the fact that the protein has 51% similarity and 35% identity to CcbF (Fig. 2). We note that while the ccna_01647 deletion did not affect ceramide levels, we observed a reduction in MHDAG synthesis (Fig. 2), though the mechanism for this decrease is unknown.

GSL-2 has a novel glycosylation pattern consisting of a hexose and a hexuronic acid (Fig. 1C). Lipid glycosylation, in both eukaryotes and prokaryotes, is performed by GT-4 family glycosyltransferases; C. crescentus encodes 11 putative GT-4 glycosyltransferases (21). To identify the glycosyltransferases required for GSL-2 synthesis, we narrowed down the list of candidate genes using the following criteria: (i) genes upregulated upon phosphate starvation (22), (ii) nonessential genes (19), and (iii) genes without a direct homolog in E. coli, since E. coli does not produce GSLs. Of the 11 initial candidates, only 2 genes fit all three criteria: ccna_00792 and ccna_00793 (Fig. 3A and Table S4). Both candidate genes are part of the PhoB regulon, which contains genes upregulated under phosphate starvation conditions (Fig. 3B) (22). Deletion of ccna_00792 or ccna_00793 resulted in a complete lack of diglycosylated ceramide (Fig. 3C), whereas deleting ccna_00792 led to the accumulation of the monoglycosylated HexA-Cer species (Fig. 3D). Complementation of these deletions recovered ceramide glycosylation (Fig. S1B and C). These data support a model of sequential ceramide glycosylation by CCNA_00793 and CCNA_00792, which we are naming sphingolipid glycosyltransferases 1 and 2 (Sgt1 and -2), respectively (Fig. 3E). Sgt1 functions as a glucuronosyltransferase, adding a hexuronic acid, while Sgt2 is a glycosyltransferase responsible for adding a hexose sugar. Sgt1 and -2 appear to specifically glycosylate sphingolipids, as neither deletion affected the glycosylation of DAG-based lipids (Fig. 3C and D). The specificity of Sgt1 and -2 was further confirmed by heterologous expression in E. coli, in which we could detect only nonglycosylated lipids (Fig. S2A and B).

While GSL-2 synthesis occurs in response to phosphate starvation, neither ceramides nor GSL-2 appears to be necessary for stalk biogenesis or cell elongation (Fig. S3A). This is likely due to the sufficiency of upregulated DAG lipids under low-phosphate conditions (Fig. 1B). Unlike GSL-2, nonglycosylated ceramides are produced across a wide range of phosphate concentrations, albeit at lower levels in the presence of excess phosphate (Fig. S3B). Surprisingly, ccbF mRNA levels are reduced during phosphate starvation, despite higher levels of ceramide production (Fig. S3C). Restriction of ceramide glycosylation to growth environments in which phospholipid synthesis is limited appears to be critical for membrane homeostasis. Overexpression of Sgt1 and Sgt2 in high-phosphate media results in the production of both HexA-Cer and GSL-2 (Fig. 4A to C). Unlike physiological glycosphingolipid production during phosphate starvation, Sgt1 and -2 overexpression leads to an accumulation of HexA-Cer (compare Fig. 1B to Fig. 4A). Furthermore, we did not detect HexHexA-DAG under these conditions, providing additional evidence that Sgt1 and -2 specifically glycosylate ceramide lipids (Fig. 4C). The production of glycosphingolipids under high-phosphate conditions leads to cell lysis, as assessed by propidium iodide staining (Fig. 4D). This phenomenon may be due to an excess of anionic lipids and is consistent with our previous observation that overabundance of anionic phospholipids is detrimental in E. coli (23).

Under laboratory growth conditions, ceramides and GSLs are nonessential; cell growth is minimally perturbed in ccbF and sgt1 deletion strains whether grown in high-phosphate (1 mM) or low-phosphate (1 µM) minimal medium (Fig. 5A and B). GSL-producing Sphingomonas species are resistant to the antibiotic polymyxin B (24); therefore, we tested whether ceramides or GSL-2 conferred similar resistance on C. crescentus. Surprisingly, the deletion of ccbF increased resistance to polymyxin B under both high- and low-phosphate conditions (Fig. 5A and B). Under low-phosphate conditions, the Δsgt1 strain did not have increased polymyxin B resistance, demonstrating that the resistance phenotype was specifically due to the presence of ceramide lipids regardless of their glycosylation (Fig. 5B). Under both phosphate conditions, complementation of ccbF restored sensitivity to polymyxin B (Fig. 5A and B).

Modifying the bacterial envelope structure and composition can also affect cellular interactions with bacteriophages. Phage ΦCr30 infects C. crescentus by attaching to the extracellular surface layer (S-layer) (25). Growth curves of wild-type and ccbF deletion strains infected with ΦCr30 in peptone-yeast extract (PYE) demonstrated that ceramides are important for increasing phage resistance (Fig. 5C); this initial screen was performed in PYE because ΦCr30 infections are generally inhibited in minimal media (26). The increased susceptibility of the ΔccbF strain could be attributed to either enhanced phage adsorption or an increased viral burst size. We tested the ability of ΦCr30 to adsorb to wild-type, ΔccbF, and Δsgt1 cells in Hutner base-imidazole-glucose-glutamate medium (HIGG)–1 µM phosphate. Both of the mutant strains had an enhanced rate of phage adsorption which was restored to normal upon complementation (Fig. 5D), suggesting that mature GSL-2 is required to inhibit phage adsorption. Measurements of burst size did not reveal any differences between the strains in low-phosphate medium (Fig. 5E); however, the ccbF and sgt1 deletion strains appear to have shorter latent periods, consistent with faster phage adsorption to these cells (Fig. 5E).

Phage ΦCr30 attaches to C. crescentus by binding to the cell envelope S-layer, a crystalline lattice composed of the protein RsaA (27). The S-layer is, in turn, anchored to the cell through interactions with the O-antigen domain of lipopolysaccharide (LPS) (28). A comparison of LPS and S-layer production in wild type and GSL mutants did not reveal any remarkable distinctions between the strains (Fig. S4A and B), suggesting that the enhanced phage adsorption observed in the GSL mutants is not due to an increase in S-layer production. The accessibility of the S-layer to phage is restricted by the production of an exopolysaccharide (EPS) capsule (29). To test whether the absence of ceramides disrupts EPS production, C. crescentus strains were grown in HIGG–1 µM phosphate and incubated with fluorescein isothiocyanate (FITC)-labeled dextran (29). The wide zone of exclusion around the wild-type and ΔccbF cells show that they produce EPS, in contrast to the non-EPS-producing ΔMGE strain (29) (Fig. S4C).

The strains, plasmids, and primers used in this study are described in Tables S1, S2, and S3, respectively. Details regarding strain construction are available in the supplemental materials. C. crescentus wild-type strain NA1000 and its derivatives were grown at 30°C in peptone-yeast extract (PYE) medium (39) for routine culturing. To control phosphate levels, C. crescentus was grown in Hutner base-imidazole-glucose-glutamate media (HIGG) with variable amounts of phosphate (1 to 1,000 µM) (40). E. coli strains were grown at 37°C in LB medium. When necessary, antibiotics were added at the following concentrations: kanamycin, 30 µg/ml in broth and 50 µg/ml in agar (abbreviated 30:50) for E. coli and 5:25 for C. crescentus; ampicillin, 50:100 for E. coli; tetracycline, 12:12 for E. coli and 1:2 for C. crescentus; gentamicin, 15:20 for E. coli and 0.5:5 for C. crescentus; and spectinomycin, 50:50 for E. coli and 25:100 for C. crescentus. Gene expression was induced in C. crescentus with either 0.3% (wt/vol) xylose or 0.5 mM vanillate. E. coli gene expression was induced with isopropyl-β-d-1-thiogalactopyranoside (IPTG; 1 mM). Phage titering was performed by adding 1 to 10 µl of ΦCr30 to 100 µl of an overnight culture of NA1000 in PYE. This mixture was added to 4 ml of soft agar (0.3% [wt/vol] agar in PYE) and overlaid on a PYE-agar plate. After solidifying, the plate was incubated overnight at 30°C and plaques were counted.

Cells were spotted onto 1% agarose pads made in the corresponding growth medium. Phase microscopy was performed on a Nikon TiE inverted microscope equipped with a Prior Lumen 220PRO illumination system, Zyla sCMOS 5.5-megapixel camera, CFI Plan Apochromat 100× oil immersion objective (numerical aperture [NA] of 1.45 and working distance [WD] of 0.13 mm), and NIS Elements software for image acquisition. Cell and stalk dimensions were measured using Morphometrics (41) and ImageJ v. 1.48q (NIH), respectively. To measure membrane permeability, cells were grown in the presence of 1 µg/ml of propidium iodide. EPS production was assessed as previously described (29). Briefly, 500 microliters of cells grown in HIGG–1 µM phosphate were collected by centrifugation (14,000 × g, 5 min), and the pellet was resuspended in 30 µl of 0.5× phosphate-buffered saline (PBS). Ten microliters of the cell suspension was mixed with 5 µl of FITC-dextran (10 mg/ml; molecular weight [MW], 2 MDa; Sigma) and 1 µl of SlowFade Diamond mountant (Thermo Scientific). Two microliters of this mixture was spotted onto a glass slide, coverslipped, and sealed with vaseline-lanolin-paraffin (VALAP; 1:1:1) for imaging.

RNA was extracted from bacterial cultures using the Qiagen RNeasy kit. Following DNase digestion, RNA (5 ng/µl) was reverse transcribed using a high-capacity cDNA reverse transcription (RT) kit (Applied Biosystems). One microliter of cDNA was used as a template in a 10-µl quantitative RT-PCR (qRT-PCR) performed with Power SYBR reagent (Applied Biosystems). qRT-PCR was performed on an ABI QuantStudio 6 using the threshold cycle (ΔΔC_T) method. rpoD expression was used as the loading control.

C. crescentus strains were grown in 500 ml of HIGG with either 1 mM or 1 µM phosphate until reaching stationary phase. Sgt1 and -2 E. coli expression strains were grown overnight in 500 ml of LB medium with 1 mM IPTG to induce protein expression. Lipids were extracted by the method of Bligh and Dyer (42). Cells were harvested in glass tubes at 10,000 × g for 30 min, and the majority of the supernatant was removed; stalked C. crescentus organisms are very buoyant and do not form tight pellets, preventing the complete removal of supernatant. The cells were resuspended in the residual supernatant, 3.75 volumes of 1:2 (vol/vol) chloroform-methanol was added, and the samples were mixed by vortexing. Chloroform (1.25 volumes) and water (1.25 volumes) were added sequentially with vortexing to create a two-phase system, and the samples were centrifuged at 200 × g for 5 min at room temperature. The bottom, organic phase was transferred to a clean tube with a Pasteur pipette and washed twice in “authentic” upper phase. Subsequently, the residual organic phase with the lipids was collected and dried under argon.

Methods for liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS) have been described previously (43, 44). Briefly, normal phase LC was performed on an Agilent 1200 quaternary LC system equipped with an Ascentis Silica high-performance liquid chromatography (HPLC) column, 5 µm, 25 cm by 2.1 mm (Sigma-Aldrich, St. Louis, MO). The LC eluent (with a total flow rate of 300 µl/min) was introduced into the ESI source of a high-resolution TripleTOF5600 mass spectrometer (Applied Biosystems, Foster City, CA). Instrumental settings for negative-ion ESI and MS/MS analysis of lipid species were as follows: IS, −4,500 V; CUR, 20 lb/in²; GSI, 20 lb/in²; DP, −55 V; and FP, −150 V. The MS/MS analysis used nitrogen as the collision gas. Data analysis was performed using Analyst TF1.5 software (Applied Biosystems).

For polymyxin B sensitivity assays, C. crescentus cells were diluted to an optical density at 660 nm (OD₆₆₀) of 0.05 in HIGG with 1 mM or 1 µM phosphate, treated with 30 µg/ml of polymyxin B, and incubated with shaking (250 rpm and 30°C). Complementation gene expression was induced with 0.3% xylose. Samples were taken at the desired times for absorbance measurements (OD₆₆₀). For phage sensitivity assays, cells were grown in PYE and diluted to an OD₆₆₀ of 0.05. A total of 148 µl of culture was dispensed per well in a 96-well plate. To each well, 2 µl of water (control) or ΦCr30 (final concentration, 5 × 10⁵ PFU/ml) was added. To prevent evaporation, each well was overlaid with 100 µl of mineral oil. OD₆₆₀ was recorded every 20 min on a BMG Labtech CLARIOstar plate reader with incubation at 30°C with continuous shaking.

Phage adsorption and burst size were measured essentially as previously described (25). To measure phage adsorption, cells were grown in HIGG–1 µM phosphate and diluted to an OD₆₆₀ of 0.2. Cells (1 ml) were aliquoted into glass culture tubes, and 10⁵ PFU of ΦCr30 was added. Cultures were incubated with shaking at 30°C; at various time points, 10 µl of culture was removed and diluted into 1 ml of water-chloroform (9:1 [vol/vol]). Ten microliters of this mixture was used to titer the unbound phage as described above. To measure viral burst size, cells were grown in HIGG–1 µM phosphate and diluted to an OD₆₆₀ of 0.1. Cells (0.5 ml) were infected with 0.5 × 10⁵ PFU of ΦCr30 and incubated at 30°C for 15 min. The culture was diluted 1,000-fold into HIGG–1 µM phosphate; a 200-µl aliquot was removed every 15 min for titering as described above.

E. coli strains were grown overnight in LB medium with 1 mM IPTG to induce protein expression. A total of 500 µl of each strain was collected by centrifugation (6,000 × g for 2 min) and the pellet was resuspended in 100 µl of sample buffer. Protein samples were resolved on a 12% SDS-PAGE gel and stained with Coomassie blue for visualization.

Lipopolysaccharide (LPS) was purified essentially as previously described (45). Briefly, 5 ml of C. crescentus cells grown in HIGG–1 µM phosphate (OD₆₆₀ = 0.5) was collected and washed once in 10 mM HEPES (pH 7.2). Cells were resuspended in 250 µl of TE buffer (10 mM Tris, 1 mM EDTA [pH 7.2]) and frozen overnight at −20°C. Cells were thawed, treated with 1 µl of DNase (0.5 mg/ml), 20 µl of lysozyme (10 mg/ml), and 3 µl of MgCl₂ (1 M), and incubated at room temperature for 15 min. For each sample, 36.25 µl was mixed with 12.5 µl of 4× SDS sample buffer and boiled at 100°C for 10 min. After cooling to room temperature, 1.25 µl of proteinase K (20 mg/ml) was added and samples were incubated at 60°C for 1 h. LPS samples were resolved on a 12% SDS-PAGE gel and stained with a Pro-Q Emerald 300 LPS stain kit according to the manufacturer’s protocol (Thermo Scientific). Images were acquired on a Bio-Rad ChemiDoc MP using UV excitation and a 530-nm emission filter.

RsaA was purified essentially as previously described (46). Briefly, cells were grown overnight in HIGG–1 µM phosphate and 5 ml (OD₆₆₀ = 0.6) was collected by centrifugation. The cell pellets were washed twice in 5 ml of 10 mM HEPES (pH 7.2), resuspended in 200 µl of 100 mM HEPES (pH 2), and incubated at room temperature for 10 min. Cells were pelleted (10 min at 5,000 × g), and the supernatant containing RsaA was collected and neutralized with 2.8 µl of 10 N NaOH. RsaA samples in 1× sample buffer were resolved on a 7.5% SDS-PAGE gel without heat denaturing and stained with Krypton protein stain (Thermo Scientific). Images were acquired on a Bio-Rad ChemiDoc MP using green light-emitting diode (LED) excitation and a 605-nm emission filter.