Dead-end intermediates in the enterobacterial common antigen pathway induce morphological defects in Escherichia coli by competing for undecaprenyl phosphate

Summary

Bacterial morphology is determined primarily by the architecture of the peptidoglycan (PG) cell wall, a mesh-like layer that encases the cell. To identify novel mechanisms that create or maintain cell shape in Escherichia coli, we used flow cytometry to screen a transposon insertion library and identified a wecE mutant that altered cell shape, causing cells to filament and swell. WecE is a sugar aminotransferase involved in the biosynthesis of enterobacterial common antigen (ECA), a non-essential outer membrane glycolipid of the Enterobacteriaceae. Loss of wecE interrupts biosynthesis of ECA and causes the accumulation of the undecaprenyl pyrophosphate linked intermediate ECA-lipid II. The wecE shape defects were reversed by: (i) preventing initiation of ECA biosynthesis, (ii) increasing the synthesis of the lipid carrier undecaprenyl phosphate (Und-P), (iii) diverting Und-P to PG synthesis, or (iv) promoting Und-P recycling. The results argue that the buildup of ECA-lipid II sequesters part of the pool of Und-P, which, in turn, adversely affects PG synthesis. The data strongly suggests there is competition for a common pool of Und-P, whose proper distribution to alternate metabolic pathways is required to maintain normal cell shape in E. coli.

Introduction

Bacteria display a wide diversity of shapes and sizes, and these morphologies remain remarkably constant from one generation to the next (reviewed in Young, 2006). With a few notable exceptions (e.g. Mycoplasma), bacterial morphology is determined primarily by the architecture of the peptidoglycan (PG) wall, a mesh-like layer that encases the cell and protects it against turgor force (reviewed in Vollmer et al., 2008a). PG consists of glycan strands composed of the repeating disaccharide, N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc), and these glycan strands are cross-linked to one another by short peptides extending from the MurNAc residue. In the Gram-negative bacterium Escherichia coli, PG biosynthesis begins in the cytoplasm and results in the formation of a GlcNAc-MurNAc-pentapeptide linked to the carrier lipid undecaprenyl phosphate (Und-P) to form lipid II, hereafter referred to as PG-lipid II (reviewed in Bouhss et al., 2008, Typas et al., 2012). PG-lipid II is translocated to the periplasmic space by a flippase (Mohammadi et al., 2011, Sham et al., 2014), and the newly flipped GlcNAc-MurNAc-pentapeptide moiety is inserted into the growing PG chain. PG synthesis is driven by a complex of proteins whose activities are directed by the actin homolog MreB during cell elongation and by the tubulin homolog FtsZ during cell division (reviewed in den Blaauwen et al., 2008, Young, 2010, Egan & Vollmer, 2013). In addition, PG hydrolases facilitate growth and separation of daughter cells by cleaving selected bonds within the PG layer (reviewed in Vollmer et al., 2008b, van Heijenoort, 2011). Mutations affecting PG synthesis or degradation can cause defects in cell wall assembly, leading to shape abnormalities that include filamentation, branching, chaining, rounding, spiraling, and swelling (de Pedro et al., 2003, Heidrich et al., 2001, Iwaya et al., 1978, Varma & Young, 2004).

Much of our understanding of how bacteria establish and maintain their shapes is derived from mutations that directly affect PG synthesis and degradation. However, new and unexpected connections between metabolism and morphogenesis are expanding our view of how cell shape is regulated (reviewed in Vadia & Levin, 2015). For example, in Bacillus subtilis, the accumulation of UDP-glucose facilitates an interaction between the glucosyltransferase UgtP and FtsZ (Weart et al., 2007). When grown in rich medium, the UgtP-FtsZ interaction disrupts assembly of the cell division apparatus, which delays cytokinesis and causes cells to grow longer before they divide. In E. coli, the glucosyltransferase OpgH mediates an analogous reaction (Hill et al., 2013). In Caulobacter crescentus, the oxidoreductase-like KidO and the NAD-dependent glutamate dehydrogenase GdhZ promote FtsZ-ring (Z-ring) disassembly. The levels of KidO and GdhZ oscillate with the cell cycle to stimulate cytokinesis and prevent premature assembly of the Z-ring (Radhakrishnan et al., 2010, Beaufay et al., 2015). Finally, E. coli mutants lacking the fatty acid synthase FabH cannot change size in response to nutrient availability, thus implicating fatty acid biosynthesis in cell size regulation (Yao et al., 2012).

These associations between cell shape and metabolism suggest that there are alternate ways by which bacteria can control their dimensions. Therefore, to identify new morphological mechanisms and regulators, we developed a flow cytometry-based screen to look for novel shape mutants in E. coli. Flow cytometry can distinguish minor shape alterations among bacterial populations and can be used to select for mutants that suppress highly aberrant morphologies, or that are enriched for shape defects (Meberg et al., 2004, Laubacher et al., 2013, Burke et al., 2013, Sycuro et al., 2013). This ability to analyze a heterogeneous population and capture cells with altered shapes makes flow cytometry a powerful tool for morphological studies.

Here, we screened a transposon insertion library by flow cytometry and found that a mutation in wecE created morphological abnormalities. WecE is involved in the biosynthesis of enterobacterial common antigen (ECA), a non-essential glycolipid found in the outer membrane of Enterobacteriaceae. A wecE mutant accumulates the ECA intermediate lipid II (ECA-lipid II), which triggers several cell envelope stress responses and confers sensitivity to bile salts (Danese et al., 1998). However, the mechanism by which ECA-lipid II exerts these effects is unknown. The present results suggest that the accumulation of ECA-lipid II (and more broadly other Und-P-utilizing glycan intermediates) sequesters part of the pool of free Und-P, which apparently restricts or alters PG synthesis. We conclude that alternate metabolic pathways compete for a common pool of Und-P, whose balanced distribution is required to maintain proper cell shape in E. coli.

Results

A genetic shape screen identifies a wecE mutant

We began to screen for novel shape and division mutants in E. coli by using a cell sorting assay previously employed to enrich for spontaneous shape suppressor mutants (Laubacher et al., 2013). Wild type E. coli CS109 was mutagenized with EZTnKan-2 (Epicentre) to give a preliminary library of approximately 5,000 independent insertion mutants, which were pooled, grown in LB medium at 37°C and analyzed by flow cytometry. The shape distribution of the mutant population was nearly identical to CS109 (Fig. 1A and 1B). Given this, we defined a selection gate to sort aberrantly shaped cells from the mutant population. Burke et al. (Burke et al., 2013) demonstrated that longer cells exhibit an increase in side scatter width (SSC-W), a measure of the time it takes a particle to pass through the fixed alignment laser in a flow cytometer. To confirm this observation, we filamented CS109 cells with aztreonam, which inhibits PG synthesis at the septum (Fig. 1C) (Georgopapadakou et al., 1982). As expected, these elongated cells exhibited an increase in SSC-W and in side scatter height (SSC-H) (Fig. 1D).

Figure 1. A genetic screen for shape mutants.

(A) Phase-contrast micrograph of E. coli CS109 wild-type (WT) cells. The white bar represents 4 μm. (B) Flow cytometry of the WT population displayed as a dot plot with side scatter height (SSC-H) plotted against side scatter width (SSC-W). The box denotes the gate (approximately 0.1% of the population) used to sort cells from a transposon library constructed in CS109. The transposon library population looked similar to the WT when analyzed by flow cytometry. (C) Phase-contrast micrograph of WT cells treated with aztreonam (2 μg ml⁻¹) for 40 minutes. (D) Flow cytometry of aztreonam-treated cells. The dashed box represents the same population outlined in panel B. (E) Phase-contrast micrograph of a tatC insertion mutant identified by the shape screen. (F) Histograms of forward scatter area for cells from the ΔtatC::Tn mutant (grey-filled peak) and the WT (black line). Mean of the forward scatter area as reported in arbitrary units (AU): WT=1668 AU; ΔtatC::Tn=11,493 AU. (G) Phase-contrast micrograph of a wecE insertion mutant identified by the shape screen. WT cells (inset) are shown for comparison. (H) Histograms of forward scatter area for cells from the ΔwecE::Tn mutant (grey-filled peak) and the WT (black line). Mean of the forward scatter area: WT=1504 AU; ΔwecE::Tn=4423 AU. All flow cytometry data is from 100,000 events (cells). Strains: MAJ3 (WT), MAJ165 (ΔwecE::Tn), and MAJ166 (ΔtatC::Tn).

We created a sorting gate (Fig. 1D, dotted rectangle) that encompassed ~80% of the aztreonam-treated cells, and collected the mutant population from within this gate (Fig. 1B, solid rectangle). These cells represented those that were either longer or larger than normal unit-sized wild type cells. Approximately 600 cells were collected from this gate and grown in LB medium at 37°C. Of these, 50% grew after overnight incubation. Remarkably, 99% of cells looked normal when reanalyzed by flow cytometry (not shown). This high false positive rate was most likely caused by transient filamentation of cells in the population, a phenomenon that is not understood (Burke et al., 2013). A similar morphological instability occurs in the bacterium Helicobacter pylori, suggesting that transient shape alterations may be a general phenomenon (Sycuro et al., 2013). Nevertheless, from 300 candidates we isolated three shape mutants, and we mapped the insertion points of each transposon by arbitrary PCR (Bernhardt & de Boer, 2004). One mutant produced a dramatic change in the shape distribution of its population (Fig. 1E and 1F), and the Tn insertion point in this strain was mapped to tatC. TatC is an essential subunit of the twin-arginine translocation system (TatABC) (reviewed in Palmer & Berks, 2012) that exports folded proteins to the periplasmic space, including the PG cell wall amidases AmiA and AmiC (Bernhardt & de Boer, 2003, Ize et al., 2003). As expected, the insertion in tatC prevented AmiA and AmiC from processing PG during growth and division, causing the cells to grow as unseparated cells (Fig. 1E) (Bernhardt & de Boer, 2003, Ize et al., 2003). Thus, isolation of this mutant served as a positive proof-of-principle for the cell sorting approach described here. A second morphology-altering mutation mapped to a gene of unknown function, yfiH, whose characterization will be reported elsewhere (Jorgenson et al., unpublished).

The third shape mutant contained an insertion that mapped to codon 163 of the wecE gene, which encodes a TDP-4-keto-6-deoxy-D-glucose aminotransferase that is required for elongating lipid III during the synthesis of ECA (Fig. 2A) (Meier-Dieter et al., 1990, Hwang et al., 2004). The distribution of the forward scatter area (FSC-A) of the wecE::Tn mutant population was shifted to the right (Fig. 1H), suggesting that the cells were enlarged, and microscopy confirmed that mutant cells were longer and wider than wild type (Fig. 1G). To confirm that the shape defects were caused by inactivation of wecE alone, we deleted wecE (ΔwecE) by lambda Red recombination (Datsenko & Wanner, 2000). This ΔwecE mutant exhibited the same phenotypes as the wecE::Tn mutant (not shown). Close inspection of the ΔwecE mutant grown in LB broth revealed that cells grew longer at elevated temperature (compare 25°C vs 37°C) and became noticeably swollen (25% wider than wild type) at 37°C (Table 1). Moreover, while the growth rate of the ΔwecE mutant was similar to the parent strain when grown at 25°C, 30°C and 37°C in LB broth, the mutant underwent a limited lysis shortly after being shifted to 42°C before resuming growth at rate similar to that of the wild type (Fig. S1). All morphological and growth defects were rescued by expressing wecE from a plasmid (Fig. 3A and 3B, and not shown).

Figure 2. ECA is not required to maintain cell shape in E. coli.

(A) ECA biosynthesis pathway. Abbreviations: P-Gc, Glucose 1-phosphate; UDP, uridine diphosphate; G, N-acetylglucosamine; dTDP, thymidine diphosphate; Ma, N-acetyl-D-mannosaminuronic acid; Gt, 4-acetamido-4,6-dideoxy-D-galactose. In E. coli, loss of RffH is compensated for by RmlA. RmlA is involved in the synthesis of dTDP-L-rhamnose, an O-antigen precursor (Stevenson et al., 1994). The figure was adapted from Paradis-Bleau et al. (Paradis-Bleau et al., 2014). (B) Phenotypes of ΔwecE mutant cells. Cells were grown at 37°C in LB to an OD₆₀₀=0.5–0.6, then fixed and photographed by phase-contrast microscopy. The white bar represents 4 μm. (C) Flow cytometry. Histograms of the FSC-A of live cells shown in panel B. The dashed line represents the mean forward scatter area of the WT. Mean of the forward scatter area is reported in arbitrary units (AU). Flow cytometry data is from 100,000 events (cells). Data is representative of two independent experiments. Strains: MAJ3 (WT), MAJ73 (ΔwecE), MAJ89 (ΔwecA), and MAJ90 (ΔwecEA).

Table 1.

Morphological phenotypes of an E. coli wecE mutant

Genotypea	Temp (°C)	No. of cells evaluated	Avg. length, μm, (SD)	Avg. width, μm (SD)	% cells with indicated no. of constrictions
					0	1	>1
WT	25	323	4.4 (1.2)	1.1 (0.1)	76	24	0
	30	335	4.5 (1.1)	1.1 (0.1)	72	28	0
	37	284	4.3 (1.0)	1.2 (0.1)	68	32	0
ΔwecE	25	308	5.8 (4.0)	1.2 (0.1)	73	27	0
	30	281	6.9 (4.5)	1.3 (0.1)	66	33	1
	37	315	7.5 (5.2)	1.5 (0.2)	71	28	1

^aStrains: MAJ3 (WT) and MAJ73 (ΔwecE).

Figure 3. Suppression of ΔwecE shape defects.

(A) Shape. ΔwecE cells harboring plasmids carrying the indicated genes were grown to an OD₆₀₀ = 0.5–0.6 at 37°C in LB containing 25 μM IPTG. Cells were fixed, then photographed by phase-contrast microscopy. The white bar represents 4 μm. (B) Flow cytometry. Histograms of the forward scatter area of live cells shown in panel A. The vertical dashed line represents the mean forward scatter area of the ΔwecE mutant containing a plasmid expressing wecE (blue graph). Mean of the forward scatter area is reported in arbitrary units (AU). Flow cytometry data is from 100,000 events (cells). Data is representative of two independent experiments. Strains: MAJ78 (vector), MAJ79 (wecE), MAJ92 (uppS), MAJ135 (bacA), MAJ181 (uppS), and MAJ238 (wbbL).

wecE mutants prevent the formation of ECA-lipid III and therefore accumulate the intermediate ECA-lipid II (Fig. 2A) (Danese et al., 1998). For unknown reasons, accumulating ECA-lipid II disrupts outer membrane integrity and produces sensitivity to sodium dodecyl sulfate and bile salts (Rick et al., 1988, Danese et al., 1998, Ramos-Morales et al., 2003). Predictably, the growth of our ΔwecE mutant was inhibited on LB containing 1% deoxycholate (Fig. 4A), confirming that the outer membrane was defective.

Figure 4. Suppression of ΔwecE bile salt sensitivity.

Overnight cultures were normalized to an OD₆₀₀ ~1.0, and 10-fold serial dilutions were plated onto LB or LB containing 1% deoxycholate. Strains shown are listed in the legend to Fig. 2 (panel A) and Fig. 3 (panel B). IPTG was not included in the plates shown in panel B. Data is representative of two independent experiments.

ECA pathway mutants also trigger the Rcs stress response and inhibit motility (Castelli et al., 2008, Castelli & Vescovi, 2011). As expected, motility was repressed in our ΔwecE mutant, and this defect was suppressed by deleting the Rcs regulator rcsB (Fig. S2). The Cpx stress response is also stimulated in wecE mutants and negatively regulates motility (Danese et al., 1998, Evans et al., 2013). To determine the contribution of the Cpx stress response on motility in the ΔwecE mutant, we deleted the Cpx response regulator, cpxR, from the ΔwecE mutant, but this had little effect (Fig. S2). Moreover, the migration of a ΔwecE ΔrcsB ΔcpxR mutant was indistinguishable from that of the ΔwecE ΔrcsB mutant (Fig. S2). These results indicate that the Cpx stress response does not significantly alter motility in ΔwecE mutants. Collectively, these data indicate that wecE is required to maintain normal cell shape and envelope integrity in E. coli.

ECA is not required to maintain proper cell shape

The morphological defects observed in ΔwecE cells suggested two possibilities: either ECA was required to maintain proper cell shape, or else shape defects were caused by the accumulation of ECA-lipid II. To distinguish between these alternatives, we first deleted wecA, whose gene product initiates ECA biosynthesis by transferring N-acetylglucosamine-1-phosphate (GlcNAc-1-P) onto the essential lipid carrier undecaprenyl phosphate (Und-P) to make Und-PP-GlcNAc (ECA-lipid I) (Fig. 2A). If ECA was required to maintain normal cell shape, then a ΔwecA mutant should exhibit the same phenotype as the ΔwecE mutant. However, ΔwecA cells looked normal (Fig. 2B), although the forward scattered light of the ΔwecA mutant population shifted slightly to the right compared to the wild type (Fig. 2C). Since neither ΔwecA nor ΔwecE cells synthesize ECA (Fig. 2A), but the ΔwecA cells retained their normal rod shapes, the results demonstrated that ECA itself was not required to maintain wild type morphology. This suggested that accumulation of ECA-lipid II was responsible for the shape defects in ΔwecE cells. If true, then deleting wecA should reverse the shape defects caused by the wecE mutation, because ECA-lipid II would no longer be synthesized. Consistent with this prediction, the cells of a ΔwecE ΔwecA double mutant were of normal shape (Fig. 2B and 2C), indicating that the accumulation of ECA-lipid II was in some way responsible for the shape defects.

ECA-lipid II accumulation probably sequesters Und-P

The shape defects in ΔwecE cells suggested that ECA-lipid II might exert an indirect effect on PG synthesis by sequestering undecaprenyl phosphate (Und-P), the lipid carrier that transfers both PG and ECA intermediates across the inner membrane (reviewed in Bouhss et al., 2008). This scenario parallels what has been observed in Gram positive bacteria. Specifically, depleting an essential late acting enzyme from the wall teichoic acid (WTA) biosynthesis pathway causes the accumulation of undecaprenyl pyrophosphate (Und-PP)-linked WTA intermediates and reduces the incorporation of new PG (D’Elia et al., 2009). Interestingly, the addition of exogenous Und-P suppresses these effects (Farha et al., 2015). Thus, if a limited amount of free Und-P inhibited PG synthesis in the ΔwecE mutant, then synthesizing more Und-P should reverse the shape defects by supplying enough substrate to satisfy the needs of both pathways. We increased the pool of Und-P by overexpressing the Und-PP synthase, encoded by uppS (also known as ispU), which is involved in the de novo synthesis of Und-P. UppS overproduction in the ΔwecE mutant fully reversed the shape defects, as visualized by microscopy and flow cytometry (Fig. 3A and 3B), suggesting that accumulation of ECA-lipid II had reduced the pool of freely available Und-P, which thereby restricted or altered PG synthesis via competitive inhibition.

Overexpressing bacA, which converts Und-PP to Und-P, is also thought to increase the available pool of Und-P (Cain et al., 1993). However, overexpressing bacA did not reverse the shape defects of ΔwecE cells (Fig. 3A and 3B), and deleting bacA did not exacerbate the shape defect (Fig. S4C and S4D). It may be that BacA does not alter the overall pool size of Und-P in this context.

An alternate method for reversing the effects of substrate competition is to increase the competitive advantage of one pathway over another. MurA is a PG transferase that competes with WecA for the common starting substrate in both pathways, UDP-GlcNAc (Marquardt et al., 1992). Thus, redirecting UDP-GlcNAc into the PG pathway by overexpressing murA might reverse the shape defects of ΔwecE cells by increasing the flux of Und-P toward PG synthesis and restricting its availability for the ECA pathway. In fact, overexpressing murA suppressed the shape defects of ΔwecE cells, consistent with the competition hypothesis (Fig. 3A and 3B).

The ECA pathway also competes for substrates with the O-antigen pathway. However, most E. coli K-12 strains (including CS109) do not produce O-antigen because of an IS5 insertion into wbbL (Liu & Reeves, 1994). ECA-lipid I is the substrate for both WbbL, a rhamnose transferase in the O-antigen pathway, and for WecG in the ECA pathway (Fig. 2A). Therefore, redirecting ECA-lipid I might also increase the pool of free Und-P by promoting Und-P recycling through the O-antigen pathway and thus preventing this substrate from reaching the dead-end ECA pathway in ΔwecE cells. Indeed, multicopy expression of wbbL reversed the shape defects of ΔwecE cells (Fig. 3A and 3B), consistent with the idea that ECA-lipid II accumulation adversely affects the amount of Und-P available for PG synthesis. Reconstitution of the O-antigen by wbbL overexpression was verified by detection with Concanavalin A (not shown) (Ghosh & Young, 2005). Overexpression of the above genes had no discernable effect on cell shape (Fig. S3A and S3B).

Overall, the results strongly suggest that accumulation of ECA-lipid II in a ΔwecE mutant indirectly affects PG synthesis by sequestering part of the pool of freely available Und-P.

Correcting the shape defects of ΔwecE cells stabilizes the outer membrane

Defects in septal PG synthesis and cell division disrupt the barrier function of the outer membrane, conferring sensitivity to bile salts (Heidrich et al., 2002, Arends et al., 2010). If the accumulation of ECA-lipid II affected PG synthesis by the sequestration model, then correcting the shape defects by promoting the flux of Und-PP-linked intermediates into the PG pathway might also reverse deoxycholate sensitivity of the ΔwecE mutant. Indeed, overexpressing uppS and murA, but not bacA, restored deoxycholate resistance to the ΔwecE mutant (Fig. 4B). Overexpressing wbbL (Fig. 4B) or deleting wecA (Fig. 4A) only partially reversed the ΔwecE membrane defect, suggesting that membrane integrity is sensitive to minor alterations in cell shape, whose presence may be below our limit of detection. Alternately, membrane defects may be independent of cell shape under certain conditions. In either case, accumulation of ECA-lipid II in the ΔwecE mutant seems to confer bile salt sensitivity by restricting or altering PG synthesis.

Correcting the shape defect of ΔwecE cells represses the Rcs response

When ECA-lipid II accumulates, the Rcs stress system is induced by an unknown mechanism (Castelli & Vescovi, 2011). Interestingly, minor alterations in PG structure also induce the Rcs stress response and inhibit motility (Laubacher & Ades, 2008, Evans et al., 2013). If deleting wecE stimulated the Rcs pathway, then increasing the amount of Und-P available for PG synthesis should reverse the motility defect. Consistent with this prediction, overexpressing uppS, murA, or wbbL, but not bacA, reversed the motility defect of the ΔwecE mutant (Fig. 5B). Deleting wecA also partially rescued motility (Fig. S2). Overexpressing these genes had little effect on WT motility (Fig. 5A). Thus, accumulation of ECA-lipid II probably triggers the Rcs stress response by limiting the amount of substrate available for PG synthesis.

Figure 5. Suppression of ΔwecE motility defects.

(A–B) Migration assays. Overnight cultures were spotted onto migration agar, incubated for 18 h at 30°C and then photographed. The dashed circles outline the outer extent of cell migration. Diameter data is from three independent experiments. Strains in panel A: MAJ76 (vector), MAJ77 (wecE), MAJ91 (uppS), MAJ134 (bacA), MAJ180 (murA), and MAJ237 (wbbL). Strain names in panel B are listed in the legend to Fig. 3

Other ECA mutants exhibit shape defects

If Und-P was sequestered by the accumulation of ECA-lipid II, then removing other proteins in the ECA pathway should have similar effects, as long as these mutants would also accumulate Und-PP-linked intermediates. As expected, a ΔwzxE (ECA flippase) mutant (Fig. 2A) exhibited morphological abnormalities similar to those of the ΔwecE mutant (Fig. 6A and 6B). Mutants lacking wecB, wecF, or wecG (Fig. 2A) also exhibited shape defects, though these were less severe than those of the ΔwecE mutant (Fig. 6A and 6B). A ΔwecA mutation (Fig. 2A) had little effect on cell shape (Fig. 2B and 2C), as did a ΔrffH mutation (Fig. 6A and 6B). However, lesions in rffH do not inhibit ECA synthesis because, in E. coli, RmlA substitutes for the loss of RffH (Fig. 2A) (Marolda & Valvano, 1995, Danese et al., 1998). The results support the interpretation that cell shape defects are caused by indirect effects associated with the accumulation of ECA intermediates.

Figure 6. Shape defects of ECA mutants.

(A) Shape. Cells with the indicated genotypes were grown in LB at 37°C to an OD₆₀₀=0.5–0.6, fixed, and photographed by phase contrast microscopy. (B) Flow cytometry. Histograms of the forward scatter area of live cells shown in panel A. The vertical dashed line represents the mean forward scatter area of the WT. Mean of the forward scatter area is reported in arbitrary units (AU). Flow cytometry data is from 100,000 events (cells). Strains: MAJ3 (WT), MAJ73 (ΔwecE), MAJ101 (ΔwecA), MAJ102 (ΔwecB), MAJ103 (ΔwecF), MAJ104 (ΔwzxE), MAJ118 (ΔrffH), and MAJ119 (ΔwecG).

The Rcs response does not exacerbate morphological defects in the ΔwecE mutant

The Rcs phosphorelay response initiates the synthesis of the extracellular polysaccharide colanic acid, whose biosynthesis also requires Und-P (reviewed in Majdalani & Gottesman, 2005, Whitfield, 2006). Synthesizing this compound would also increase the demand on the available pool of Und-P and might therefore exacerbate the morphological defects of ΔwecE cells. However, deleting rcsB did not reverse the shape defects of ΔwecE cells (Fig. S4C and S4D). Moreover, deleting wcaJ, whose gene product transfers colanic acid precursors onto Und-P (Patel et al., 2012), did not alleviate shape defects (Fig. S4C and S4D). Deleting rcsB or wcaJ had no effect on shape (Fig. S4A and S4B). Thus, the shape defects of ΔwecE cells were independent of the Rcs response.

Overexpressing elyC or mrcB does not reverse the morphological defects of ΔwecE cells

Paradis-Bleau et al. reasoned that a newly-discovered protein encoded by the elyC gene might be involved in moderating a competition between the PG and ECA biosynthesis pathways, probably at the level of Und-P metabolism (Paradis-Bleau et al., 2014). An elyC mutant exhibits growth and membrane defects that are reminiscent of those in a ΔwecE mutant, defects that are reversed by overexpressing uppS, murA, or mrcB, and by mutations that block ECA production (Paradis-Bleau et al., 2014). The parallels between the elyC and wecE phenotypes suggested that overexpressing elyC or mrcB might also reverse the shape defects of a ΔwecE mutant, perhaps by shifting Und-P utilization towards PG synthesis. However, overexpressing elyC or mrcB did not restore normal shape to the ΔwecE mutant (Fig. S5). While this was somewhat surprising, the function of ElyC is more important at lower temperatures, whereas the ΔwecE phenotypes are more prominent at higher temperatures. Thus, elyC and wecE appear to affect the balance of Und-P by different mechanisms.

Discussion

Bacteria produce their specific shapes by regulating PG synthesis as the cells elongate via an MreB-dependent mechanism (Billings et al., 2014, Tropini et al., 2014, Si et al., 2015), as they divide via an FtsZ-dependent mechanism (Varma & Young, 2004, Varma et al., 2007, Varma & Young, 2009, Potluri et al., 2012), or by mechanisms that involve accessory PG hydrolases (Heidrich et al., 2002, Priyadarshini et al., 2007, Singh et al., 2012). A more recent discovery is that intermediates of glucose metabolism inhibit FtsZ-driven cell division, thereby explaining the decades-old observation that fast growing cells are longer and larger (Weart et al., 2007, Hill et al., 2013). Here, we show that the morphology of E. coli is also sensitive to alterations in the distribution of shared lipid-linked precursors among pathways that synthesize PG, ECA, and O-antigen. This conclusion was triggered by finding that cells lacking an ECA pathway enzyme, the aminotransferase WecE, exhibited abnormal cell shapes. wecE mutants accumulate the Und-PP-linked compound, ECA-lipid II (Danese et al., 1998), a dead-end intermediate that might therefore sequester part of the pool of free Und-P, which is required for PG synthesis. The results imply that the PG and ECA biosynthetic pathways, as well as those that synthesize other glycan polymers, compete for a common pool of Und-P (Fig. 7) (Paradis-Bleau et al., 2014). In short, bacterial morphology is determined not only by the synthases and hydrolases that create and modify the cell wall, but also by the physiological processes that distribute shared precursors among these related pathways.

Figure 7. Model of substrate competition among metabolic pathways.

The enterobacterial common antigen (ECA), O-antigen, and peptidoglycan (PG) biosynthesis pathways compete for a pool of uridine diphosphate N-acetylglucosamine (UDP-G) and undecaprenyl phosphate (Und-P). E. coli K-12 strains do not normally produce O-antigen because of an insertion in wbbL. Short peptides (circles) are attached to the N-acetylmuramic acid (M) residue. ECA-lipid I is noted in blue. ECA-lipid II and III are noted in purple. PG-lipid I and II are noted in green. Note that the colanic acid synthesis pathway also utilizes Und-P, but this pathway is omitted from the schematic. Other abbreviations: Und-PP-G, undecaprenyl pyrophosphate N-acetylglucosamine.

Explaining the toxicity of Und-PP-linked polysaccharide intermediates

The deleterious effects triggered by the accumulation of various Und-PP-linked metabolic intermediates have been known for some time (Yuasa et al., 1969, Rick et al., 1988, Danese et al., 1998, Rick et al., 2003, Marolda et al., 2006, Tatar et al., 2007, Castelli & Vescovi, 2011), but the mechanism of this toxicity has remained elusive. In theory, there are three possible explanations. First, the end product (in this case ECA) might be required for creating the wild type shape of E. coli. However, cells lacking wecA grew as normally shaped rods, even though this mutation prevents the production of ECA (Meier-Dieter et al., 1990). Also, deleting wecA reversed the morphological defects of the wecE mutant. These results argue that the lack of ECA does not produce the negative physiological phenotypes. The second alternative is that the compound that accumulates (in this case ECA-lipid II) is toxic in and of itself, perhaps because it cannot be translocated to the periplasm and thereby poisons some essential reaction. This is the simplest and most direct interpretation (Danese et al., 1998, Rick et al., 2003, Marolda et al., 2006). However, the toxic effects of a wecE mutation are suppressed by increasing the precursor pool of Und-P. Since the amount of any putative intermediate should not change in this circumstance (and might even increase), it is unlikely that the accumulated compound behaves like a classic toxin, in the sense of inhibiting some other reaction directly and specifically.

The third and most likely possibility is that trapping more and more material in a dead-end intermediate is deleterious because it sequesters a precursor that is required for another purpose. In this case, accumulation of ECA-lipid II would reduce the pool of free Und-P available for other pathways, namely those that synthesize PG, O-antigen, or colanic acid (Fig. 7). Strong support for this interpretation is provided by the fact that increasing the total pool of Und-P reverses the phenotypic effects of a wecE mutation, as does redirecting more of the limited pool of free Und-P into the PG synthetic pathway. In addition, cells containing a temperature sensitive allele of uppS have decreased levels of Und-P and exhibit shape defects that look remarkably similar to those of a wecE mutant (Kato et al., 1999), further supporting the idea that reducing the availability of Und-P is detrimental to cell wall synthesis. In a possible counter-example, Danese et al. found it unlikely that ECA-lipid II accumulation would affect PG synthesis by sequestering Und-P because its effects were not reversed by overexpressing the Und-PP phosphatase bacA (Danese et al., 1998). However, our results suggest that accumulation of ECA-lipid II would also reduce the amount of free Und-PP, meaning that increasing or decreasing BacA levels may have little or no effect on the availability of Und-P. Taking the evidence as a whole, we conclude that the accumulation of Und-PP-linked intermediates indirectly restricts PG synthesis by sequestering Und-P, thereby impeding cell growth and producing morphological and membrane defects in E. coli.

Given the central role of Und-P in the biosynthesis of various glycan polymers, it should come as little surprise that the sequestration hypothesis has been suggested to account for several phenotypes observed in a diverse set of organisms including: E. coli (Paradis-Bleau et al., 2014), Salmonella enterica (Yuasa et al., 1969, Liu et al., 2015), Shigella dysenteriae (Klena & Schnaitman, 1993), Pseudomonas aeruginosa (Burrows & Lam, 1999), Bacillus subtilis (D’Elia et al., 2009), and Streptococcus pneumoniae (Xayarath & Yother, 2007). Thus, the sequestration of Und-P into non-productive intermediate compounds may explain many previous results obtained across a fairly wide range of bacteria. For example, overproducing WecA in an E. coli wzxB wzxE mutant causes the cells to elongate and then lyse, presumably because they produce an elevated amount of ECA-lipid III that cannot be exported to the periplasm (Rick et al., 2003, Marolda et al., 2006). ECA-lipid III accumulation would be expected to sequester Und-P (see Fig. 2A), and excess WecA would force even more of the Und-P pool into this non-productive pathway at the expense of PG synthesis, thus leading to the morphological alterations and lysis. Growth and morphological defects are also observed in mutants that accumulate Und-PP-linked O-antigen intermediates. Such mutants are genetically unstable (Yuasa et al., 1969) and cause significant growth defects (Klena & Schnaitman, 1993), which suggested to Tatar et al. that these unprocessed intermediates were toxic (Tatar et al., 2007). However, given our current results, it seems just as likely that these dead-end intermediates sequestered part of the pool of Und-P, thereby affecting PG synthesis and producing the phenotypes. Finally, the accumulation of ECA intermediates disrupts the bacterial membrane, confers sensitivity to detergents and bile salts, and induces the Cpx and Rcs stress responses (Rick et al., 1988, Danese et al., 1998, Castelli & Vescovi, 2011), all of which can also be caused by inhibiting PG synthesis (Laubacher & Ades, 2008, Evans et al., 2013). Thus, the simplest unitary explanation that explains the effects associated with the accumulation of different compounds in related pathways is that in each case part of the pool of Und-P becomes trapped in an unusable intermediate. The similar phenotypes in these disparate systems would therefore arise from the common, indirect mechanism of inhibiting or otherwise affecting PG synthesis.

The above precursor sequestration interpretation coincides exactly with the explanation for how similar growth and morphological effects are produced by mutations in interrelated cell wall pathways in Gram positive bacteria. For example, in B. subtilis, blocking the late stages of wall teichoic acid (WTA) biosynthesis leads to shape defects and lysis (Brandt & Karamata, 1987, Briehl et al., 1989, Pooley et al., 1991, Bhavsar et al., 2001, D’Elia et al., 2006a), as it also does in Staphylococcus aureus (D’Elia et al., 2006b, Campbell et al., 2012). Mutations that block capsule synthesis in S. pneumoniae produce similar results (Xayarath & Yother, 2007). However, removing the first enzyme in these pathways is not lethal and, in fact, reverses the lethality of mutations affecting later steps in the pathway (D’Elia et al., 2006a, D’Elia et al., 2006b, Xayarath & Yother, 2007), results that exactly parallel the phenomenon we report here. These previous authors speculated that the lethality was caused either by a buildup of toxic intermediates or because these intermediates trapped so much Und-P that PG synthesis was impaired. This latter possibility is identical to the Und-P sequestration mechanism we find to be at work here. Indeed, while the current manuscript was in preparation and under review, Farha et al. distinguished between these two possibilities in S. aureus by showing that the effect of late stage WTA mutations were reversed by adding exogenous Und-P, strongly supporting the idea that toxic intermediate compounds were not at fault but that Und-P was being sequestered (Farha et al., 2015). We propose that this same phenomenon occurs in E. coli by virtue of the fact that different cell envelope pathways compete for this universal lipid carrier (Fig. 7).

Implications and unanswered questions

The fact that multiple synthetic pathways compete for Und-P raises several questions. First, does E. coli maintain a common pool of Und-P that can be accessed by any needy pathway? Although this seems most likely, current data cannot rule out the possibility that each pathway is fed by a dedicated sub-pool, though these might be supplemented by drawing from a shared pool. Second, regardless of whether there is one pool or several, it is not clear how the cell prioritizes and distributes Und-P among the various pathways, whether this be during normal growth conditions or in response to antibiotics or other environmental cues. PG synthesis is an obvious priority, but scenarios may arise in which other polysaccharides are deemed more vital in the short run. Can the cell redirect Und-P in these situations? Third, is the total amount of Und-P modulated in response to changing demands, and if so, how? On the surface, it would seem that a simple feedback mechanism could adjust the pool of Und-P, but whether this occurs is not known. However, even if such a homeostatic mechanism exists, it must not be very robust since it is obvious that significant deviations in the availability of Und-P can be produced by somewhat minor alterations in the demands made by competing pathways.

The present results also address the nature of one possible signal that has been suggested to induce the Rcs envelope stress response in E. coli (Majdalani & Gottesman, 2005). Defects in ECA biosynthesis trigger this response (thereby inhibiting bacterial motility), which seems to imply that the absence of periplasmic ECA is one of the signals that can induce the Rcs cascade (Castelli & Vescovi, 2011). However, our current results throw such a conclusion into doubt, because the ΔwecE motility defect can be rescued even when mature ECA structures are absent. Thus, the lack of ECA does not, by itself, function as an Rcs system inducer.

As mentioned earlier, mutations that impact cell shape are explained most easily by how they affect MreB-directed cell wall elongation or FtsZ-mediated division. PG synthesis is central to each process, so that in a wecE mutant each pathway is probably constrained by having decreased access to PG-lipid II. That a ΔwecE mutation elicits a filamentation phenotype suggests that cell division is more sensitive than is elongation to alterations in the pool of Und-P. Interestingly, in B. subtilis, PG-lipid II appears to be required for MreB to associate with the membrane and to form protein filaments (Schirner et al., 2015). The fact that an E. coli wecE mutant elongates suggests that a similar relationship may exist between PG-lipid II and one or more cell division proteins.

Finally, competition for Und-P provokes the intriguing possibility that this was the selective pressure that removed the O-antigen from most strains of E. coli K12 via a single IS element insertional event (Liu & Reeves, 1994). Because this strain has been incubated for decades under extremely growth-friendly laboratory conditions, it is easy to imagine that a growth advantage would accrue to cells that had eliminated one or more of the competitor pathways, thereby increasing the pool of Und-P available for the more relevant synthesis of PG. Because O-antigen is not essential under these conditions, it is a prime candidate for removal. In fact, there are two mutations that inactivate O-antigen synthesis in different strains of E. coli K12 (Liu & Reeves, 1994), and both truncate the pathway at its first, WbbL-directed step (see Fig. 7). Thus, no Und-PP-linked O-antigen intermediates accumulate in either case. It may be that these specific mutations were selected precisely because mutating any other downstream member of the pathway would produce dead-end Und-PP-linked intermediates that would produce growth defects as bad (or worse) than those associated with similar ECA intermediates. Whether these mutations arose for this reason is purely speculative, of course, but it seems clear that there is no real pressure to restore the O-antigen to E. coli K12, and perhaps the competition for Und-P explains why.

Experimental procedures

Media

Unless otherwise noted, cells were grown in Luria-Bertani (LB) broth (Difco). Plates contained 1.5% agar. Ampicillin, kanamycin, and tetracycline were used at 100, 50, and10 μg ml⁻¹, respectively. Migration plates contained 1% tryptone, 0.25% NaCl, and 0.25% agar.

Strains and plasmids

All strains, plasmids, and primers are listed in Tables S1, S2, and S3, respectively. Strain and plasmid construction is detailed in the supporting information. All plasmids used for in vivo experiments are derivatives of pDSW204 (Weiss et al., 1999). Oligonucleotide primers and synthetic genes were purchased from Eurofins Genomics (Huntsville, Alabama).

Screening a transposon insertion library for cell shape mutants

Wild type CS109 cells were mutagenized with EZ-Tn-Kan2 (Epicentre), as described previously (Bernhardt & de Boer, 2004). The mutagenesis yielded approximately 5,000 colonies, which were pooled and diluted to an OD₆₀₀ = 0.1 in LB medium. The mutant library was allowed to grow for 2 hours at 37°C to an OD₆₀₀ = 0.6–0.7, after which cells were prepared for flow cytometry, as described previously (Laubacher et al., 2013). Individual cells from the insertion library were sorted as described in the legend to Figure 1. Cells were sorted using a FACSAria III sorter from BD Biosciences (UAMS Flow Cytometry Core Facility) into a 96-well plate whose wells contained 0.15 mL LB. Plates were incubated at 37°C overnight, after which glycerol was added to a final concentration of 15% and the plates were stored at −80°C. Cells were rescreened for shape defects by growing overnight cultures in LB medium containing kanamycin at 37°C. Overnight cultures were diluted 1:100 in LB medium at 37°C, grown to OD₆₀₀ = 0.5–0.6, and analyzed by flow cytometry. Mutants that displayed a shift in FSC-A to the right (indicating larger cells) were confirmed by microscopy. The insertion point of the transposon was mapped by arbitrary PCR, as described (Bernhardt & de Boer, 2004).

Morphological analysis of the ΔwecE mutant

Overnight cultures grown at 37°C in LB medium were diluted 1:2000 in fresh LB medium and grown to OD₆₀₀ = 0.5–0.6. Cells were stained with the membrane dye FM4-64 (Invitrogen), and 4 μl were spotted onto 1% agarose pads and visualized by phase-contrast and fluorescence microscopy.

Rescue of ΔwecE shape defects

Cells harboring plasmids were grown at 37°C in LB medium containing ampicillin. Overnight cultures were diluted 1:2000 in LB medium containing 50 μg ml⁻¹ ampicillin and 25 μM IPTG and grown to an OD₆₀₀ = 0.5–0.6 at 37°C. Cells were fixed, then photographed by phase-contrast microscopy. Live cells were used for flow cytometry analysis.

Flow cytometry analysis

Live cells were prepared for flow cytometry as described (Laubacher et al., 2013). Overnight cultures grown in LB medium at 37°C were diluted 1:2000 in the same medium and grown to an OD₆₀₀ = 0.5–0.6. Cells from 1 ml of an exponentially-growing culture were pelleted by centrifugation, and the resulting pellet was resuspended in 1 ml of filtered phosphate-buffered saline (PBS) (137 mM NaCl, 3 mM KCl, 9 mM NaH₂PO₄, and 2 mM KH₂PO₄, pH 7.4). Cells were washed twice in PBS, after which the cells were diluted 1:10 in PBS (OD₆₀₀ ~ 0.05). Typical experiments analyzed 100,000 events (cells) using the forward- and side-scatter detectors in a BD LSRFortessa at the UAMS Flow Cytometry Core Facility.

Bile salt sensitivity assays

Overnight cultures grown in LB medium at 37°C were normalized to an OD₆₀₀ = 1.0 in the same medium, and 10-fold serial dilutions were plated onto LB or LB containing 1% deoxycholate. Plates were incubated overnight at 37°C and then photographed.

Migration assays

Cells from 1.5 μl of an overnight culture grown at 30°C in LB were spotted onto the surface of a migration plate (Evans et al., 2013). Plates were incubated at 30°C for 18 hours and then photographed.

Microscopy, image analysis, and figure construction

Our microscope and camera have been described previously (Vega & Young, 2014). Cell lengths were measured with cellSens Dimensions software version 1.6 (Olympus). Images were processed to adjust brightness and contrast using ImageJ (microscopy assays) (Schneider et al., 2012) or Adobe Photoshop (migration assays). All images were cropped and assembled in Adobe Illustrator.