Molecular Basis of Lipopolysaccharide Heterogeneity in Escherichia coli

Background: LPS is essential for viability, although it is highly heterogeneous. Results: Synthesis of different glycoforms is regulated by differential expression of WaaZ (KdoIII transferase) and WaaR (glycosyltransferase). RpoE-transcribed rybB sRNA represses WaaR synthesis, and ppGpp is required for KdoIII incorporation in RpoE-inducing conditions. Conclusion: RpoE induction causes truncation of outer core and rhamnose addition to KdoIII. Significance: LPS alterations are crucial for outer membrane function. Mass spectrometric analyses of lipopolysaccharide (LPS) from isogenic Escherichia coli strains with nonpolar mutations in the waa locus or overexpression of their cognate genes revealed that waaZ and waaS are the structural genes required for the incorporation of the third 3-deoxy-α-d-manno-oct-2-ulosonic acid (Kdo) linked to Kdo disaccharide and rhamnose, respectively. The incorporation of rhamnose requires prior sequential incorporation of the Kdo trisaccharide. The minimal in vivo lipid A-anchored core structure Kdo2Hep2Hex2P1 in the LPS from ΔwaaO (lacking α-1,3-glucosyltransferase) could incorporate Kdo3Rha, without the overexpression of the waaZ and waaS genes. Examination of LPS heterogeneity revealed overlapping control by RpoE σ factor, two-component systems (BasS/R and PhoB/R), and ppGpp. Deletion of RpoE-specific anti-σ factor rseA led to near-exclusive incorporation of glycoforms with the third Kdo linked to Kdo disaccharide. This was accompanied by concomitant incorporation of rhamnose, linked to either the terminal third Kdo or to the second Kdo, depending upon the presence or absence of phosphoethanolamine on the second Kdo with truncation of the outer core. This truncation in ΔrseA was ascribed to decreased levels of WaaR glycosyltransferase, which was restored to wild-type levels, including overall LPS composition, upon the introduction of rybB sRNA deletion. Thus, ΔwaaR contained LPS primarily with Kdo3 without any requirement for lipid A modifications. Accumulation of a glycoform with Kdo3 and 4-amino-4-deoxy-l-arabinose in lipid A in ΔrseA required ppGpp, being abolished in a Δ(ppGpp0 rseA). Furthermore, Δ(waaZ lpxLMP) synthesizing tetraacylated lipid A exhibited synthetic lethality at 21–23°C pointing to the significance of the incorporation of the third Kdo.

Lipopolysaccharides (LPS) are the major amphiphilic constituents of the outer leaflet of the outer membrane (OM) 2 of Gram-negative bacteria, including Escherichia coli. Although LPS are highly heterogeneous in composition, they share a common architecture composed of a membrane-anchored phosphorylated and acylated ␤(136)-linked GlcN disaccharide, termed lipid A, to which a carbohydrate moiety of varying size is attached (1,2). The latter may be divided into a lipid A proximal core oligosaccharide and, in smooth-type bacteria, a distal O-antigen. The core oligosaccharide can be further divided into the inner and outer core. The lipid A part and the inner core are generally conserved in structure but often have nonstoichiometric substitutions. The 3-deoxy-␣-D-mannooct-2-ulosonic acid (Kdo)-lipid A portion of LPS defines the minimal structure required to support growth of E. coli up to 42°C and is the most conserved part in the LPS of Gram-negative bacteria (1)(2)(3). After the synthesis of lipid IV A precursor, the transfer of Kdo to it is a critical and essential step in the LPS biosynthesis. This Kdo incorporation ensures the incorporation of secondary laurate and myristate chains, resulting in the synthesis of hexaacylated lipid A and further extension by sequential addition of heptoses and hexoses (1)(2)(3). Thus, the gene encoding Kdo transferase can be deleted only under slow growth conditions of minimal medium at/or below 21°C without requirement of any suppressor mutations (4). Under such slow growth conditions at low temperatures, such mutants exhibit a limited incorporation of laurate and myristate chains (4). Thus, overexpression of either the lauroyl-or the myristosyltransferase in a waaA deletion could restore growth at 30 or 37°C but not above (5).
The biosynthesis of E. coli K-12 LPS is relatively well studied, and the main genetic and structural determinants are known (1,2). However, the overall composition of LPS is quite heterogeneous due to several nonstoichiometric substitutions (Fig. 1). Among the nonstoichiometric substitutions commonly observed in lipid A part are the addition of phosphoethanolamine (P-EtN) and 4-amino-4-deoxy-L-arabinose (L-Ara4N) (3). Such substitutions are not needed for growth under laboratory conditions but can have adaptive advantage in specific niches, because they confer resistance to cationic peptides like polymyxin B (3). The inner core structure can also be substituted nonstoichiometrically by residues, including phosphate, rhamnose (Rha), P-EtN, and additional Kdo (3). Two independent studies addressed structural aspects of substitution of additional Kdo linked to Kdo disaccharide (6,7). In one study, overexpression of the E. coli K-12 waaZ gene in an E. coli isolate with R1 core, which lacks the waaZ gene, was found to lead to the synthesis of increased amounts of Kdo(234)Kdo(234)Kdo trisaccharide in the inner core, which was accompanied by a truncation of the outer core (7). However, given the heterogeneity and overall complex composition of E. coli K-12 LPS, no obvious differences could be observed between a waaZ null as compared with the wild type, and only upon overexpression could differences be observed. In another study, using LPS from an E. coli K-12 strain, four different LPS glycoforms were purified, and one of the minor forms was found to contain an Kdo(234)[␣-L-Rha(135)]Kdo(234)Kdo-branched tetrasaccharide connected to lipid A, and it is designated as glycoform IV (6). This glycoform was found to have a truncation of the outer core, in which the terminal disaccharide L-␣-D-Hep(136)␣-D-Glc was missing. Both of these studies suggest that incorporation of additional (234)␣-Kdo on ␣-Kdo(234)␣-Kdo disaccharide causes truncation of the outer core, but its molecular basis remained unknown. Furthermore, the structural gene required for the addition of Rha to Kdo remained to be identified, and the functional and regulatory mechanisms that contribute to the presence of several glycoforms remained unaddressed.
We earlier showed that LPS of strains with tetraacylated lipid A due to lack of LpxL, LpxM, and LpxP late acyltransferases exhibited increased accumulation of glycoform IV with predicted presence of Rha, and P-EtN on the second Kdo in phosphate-limiting growth conditions (4). It is now established that the main function of the RpoE regulon is to ensure a correct assembly of outer membrane proteins (OMP) as well as to regulate functions involved in LPS translocation to the OM (8,9). Furthermore, RpoE not only responds to OMP misfolding, but also to the synthesis of defective LPS composed of either Kdo 2lipid IV A or of lipid IV A derivatives devoid of any glycosylation (4). However, whether LPS heterogeneity, including synthesis of LPS glycoform IV, is also regulated by this stress-responsive signal transduction is not known, because a ⌬(lpxL lpxM lpxP) strain also exhibits mild RpoE induction (4). Furthermore, it is not clear to what extent two-component systems other than known BasS/R-mediated lipid A modifications contribute to LPS heterogeneity. The importance of LPS core composition is further manifested in the discovery showing that out of the 51 genes, which are essential for bacterial growth at critical high temperature, products of 8 such genes are involved in LPS core assembly (10). Thus, in this study, we first identified structural genes that are required for the synthesis of glycoform with the third Kdo and Rha and their functional significance, using a panel of isogenic mutants with nonpolar mutation in various genes encoded in the waa locus. Next, we addressed the role of various regulators involved in sensing outer membrane alterations. This included analyses of LPS from strains with mutations in genes whose products are involved in lipid A modifications and the main regulatory control elements of the extracytoplasmic stress pathway under the control of RpoE/ RseA signal transduction. Furthermore, we addressed if any molecular switches control the relative abundance of different glycoforms and their impact on the OMP profile.
Generation of Null Mutations and Construction of Their Combinations-Nonpolar antibiotic-free deletion mutations of various genes were generated using the Red recombinase/ FLP-mediated recombination system (12). The coding sequence of each gene was replaced with either the kanamycin (aph) or chloramphenicol (cat) resistance cassette flanked by FRT recognition sequences, using plasmids pKD13 and pKD3 as templates (12), and recombined on the chromosome of BW25113 containing the Red recombinase-encoding plasmid pKD46. Gene replacements and their exact chromosomal locations were verified by PCR and further transduced in W3110. All the deletions were verified to be nonpolar. Construction of deletion derivatives of the rseA, rseB, rpoE, rybB, basS, eptB, and eptA genes and strains lacking all the late acyltransferases in W3110 were described previously (4). To construct ⌬(lpxL lpxM lpxP waaZ), ⌬waaZ deletion was introduced into ⌬(lpxL lpxM lpxP) strain on M9 medium at 21 or 30°C in the presence or absence of the lpxL-bearing plasmid. Because a null mutation in the waaU gene has not been described in E. coli K-12, a nonpolar deletion derivative was constructed using the Red recombinase/FLP-mediated recombination system in BW25113 and transduced in W3110 at 30 and 37°C, both on LB and M9 minimal medium. To construct relA spoT derivatives, relevant mutations were transduced into W3110 and into isogenic ⌬rseA on LA-rich medium. The presence of the suppres-sor-free relA spoT combination was verified by their inability to grow on amino acid-free M9 medium.
For protein induction, the minimal coding sequence of waaZ and waaS genes was cloned in pET16b and pET24b expression vectors. To co-express waaZ and waaS, the waaZ gene under T7 transcription control was subcloned from pET16b (pSR7960) into pGK2055 (pET24b waaS) in the BglII site, resulting in plasmid pSR9298, wherein both the genes are expressed from individual T7 promoter (Table 1). For controlled complementation and mild induction (0.15 mM isopro-  DECEMBER 16, 2011 • VOLUME 286 • NUMBER 50 pyl 1-thio-␤-D-galactopyranoside), the waaZ gene was expressed in pCA24N. In this vector, the minimal coding sequence is under the tight ptac promoter (13). Protein Purification-Expression of hexa-His-tagged WaaZ variants was induced in E. coli BL21 strain by the addition of 1 mM isopropyl 1-thio-␤-D-galactopyranoside at an absorbance of 0.1 at 600 nm in a 1-liter culture. After an induction for 4 h at 37°C, cells were harvested by centrifugation at 7,000 rpm for 20 min. The pellet was resuspended in lysis buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 10 mM imidazole (buffer A)) and supplemented with lysozyme to a final concentration of 200 g ml Ϫ1 . The mixture was incubated on ice for 20 min, sonicated, and centrifuged at 45,000 ϫ g for 30 min at 4°C. Soluble proteins (15 ml) were applied over nickel-nitrilotriacetic acid beads (Qiagen), washed, and eluted with buffer A containing 100 mM imidazole.

RpoE-dependent and -independent Alterations in E. coli LPS
Construction of Chromosomal C-terminal FLAG Derivatives-Using pSUB11 as template (14) and oligonucleotides listed in supplemental Table S1, C-terminal 3ϫFLAG-tagged waaZ, waaR, and lpxM PCR products with aph marker were generated. PCR products were electroporated to generate chromosomal recombinants of BW25113 containing the Red recombinase-encoding plasmid pKD46. Correct chromosomal exchanges were verified by PCR and sequencing of PCR products from chromosomal FLAG-tagged derivatives. Whether the expression of the concerned gene was not altered was verified by the presence of the wild-type phenotype. The 3ϫFLAG derivatives were transduced into W3110, followed by pCP20mediated excision of aph cassette. Whenever required, the appropriate deletion mutations were further introduced into these FLAG epitope derivatives, and expression was revealed by Western blotting using FLAG-specific M2 monoclonal antibodies from Sigma (F3165).
LPS Extraction and Growth Analysis-Cultures of isogenic bacteria were grown in a rotary shaker at 190 rpm in phosphaterich M9 medium or in phosphate-limiting medium until an absorbance of 0.8 -1.0 at 600 nm with appropriate antibiotic at a permissive temperature depending upon the mutation(s). Four hundred-ml cultures were harvested by centrifugation at 7,000 rpm for 30 min and dried. LPS was extracted by the phenol/chloroform/petroleum ether procedure (15) and lyophilized. For the LPS analysis, lyophilized material was dispersed in water by sonication and resuspended at a concentration of 2 mg ml Ϫ1 .
Mass Spectrometry-Electrospray ionization Fourier transform ion cyclotron (ESI FT-ICR) mass spectrometry was performed in negative ion mode using an APEX QE (Bruker Daltonics) equipped with a 7-tesla actively shielded magnet and dual ESI-Maldi. LPS samples were dissolved at a concentration of ϳ10 ng l Ϫ1 and analyzed as described previously (4,16). Mass spectra were charge deconvoluted, and the mass numbers given refer to the monoisotopic peaks. Mass calibration was done externally using well characterized similar compounds of known structure (16).
Western Blot Analysis-To detect and estimate changes in the levels of WaaZ, WaaR, and LpxM, 25-ml cultures of C-terminally 3ϫFLAG-tagged derivatives were grown with shaking at 37°C in LB or 121 medium. Absorbances at 595 nm were measured at different intervals, and aliquots were drawn at different stages of bacterial growth as indicated. Samples were harvested by centrifugation at 3,000 ϫ g for 10 min. Protein amounts were measured by BCA kit, and 20 g of each sample was used. Samples were resuspended in SDS lysis buffer and were applied to 12% SDS-PAGE. After the electrophoresis, proteins in the gel were blotted to PVDF membrane. To determine any differences in OMP composition because of the switch from glycoform I to glycoform V LPS, whole cell lysates from cultures of strains with remarkable differences in the accumulation of either of the two glycoform were prepared. Cultures were grown up to identical absorbances, harvested by centrifugation as described above, and lysed in SDS lysis buffer. Equivalent amount of total protein (5 g) was subjected to 12% SDS-PAGE. The relative amounts of OmpA, OmpC, and OmpF were revealed by Western blotting, using corresponding antibodies as described previously (17)(18)(19).
␤-Galactosidase Assays-To measure the activity of the waaZ promoter, single copy chromosomal promoter fusions to the lacZ were constructed. The induction of the RpoE pathway was monitored in strains carrying the rpoHP3 promoter, whose construction has been previously described (20). Putative promoter region of the waaZ gene was amplified by PCR, using specific oligonucleotides (supplemental Table S1). After PCR amplification, gel-purified DNA was digested with EcoRI and BamHI, cloned in pRS551 vector, and transferred to chromosome in single copy by recombination with RS45, selecting for Kan-resistant lysogens as described previously for other promoter fusions (8,20,21). To measure ␤-galactosidase activity, isogenic bacterial strains carrying promoter fusions were grown with appropriate antibiotics at 37°C. Cultures were harvested by centrifugation and diluted to an A 595 of 0.02. Cultures were allowed to grow for another 90 min at 37°C, and ␤-galactosidase activity was measured at different growth intervals. At least four independent cultures were assayed for each mutant and isogenic parent.

RESULTS
Growth of E. coli in Phosphate-limiting Medium Induces Synthesis of Glycoform IV-We previously reported that growth of E. coli K-12 in phosphate-limiting 121 medium induces lipid A modifications because of the induction of BasS/R two-component system leading to the substitution of P-EtN and L-Ara4N (4). Growth in this medium also favors addition of P-EtN to the second Kdo (4). In this study, we analyzed in depth the molecular and structural basis of the incorporation of the third Kdo and any other alteration in the LPS inner and outer core. Comparison of LPS of the E. coli K-12 wild-type strain W3110 obtained from phosphate-rich (M9) versus phosphate-limiting (121) medium revealed several differences. They are manifested by the presence of mass peaks predicted to contain either the usual glycoform I or the relatively rare glycoform IV and nonstoichiometric substitutions by P-EtN and L-Ara4N (Fig. 1). Mass peaks predicted to be derivatives of glycoform IV were preponderant species in LPS samples obtained from growth in phosphate-limiting medium. They are represented by mass peaks at 3948.7, 4079.8, 4202.8, and 4298.9 Da (supplemental Fig. S1). These mass peaks can be explained by the presence of a third Kdo and Rha linked to the Kdo disaccharide with a concomitant truncation of the outer core with a predicted composition LA hexa (Kdo 3 RhaHep 3 Hex 3 P 2 ) (supplemental Fig. S1, A and B) (6,7). The mass differences among the above described mass peaks can be explained by additional substitution with P-EtN and L-Ara4N. Additional mass peaks at 3936.7, 4059. 8, 4190.8, and 4286.9 Da are predicted to be derivatives of glycoform I. These mass peaks can be explained as LA hexa (Kdo 2 Hep 4 Hex 4 P 2 ) accompanied by additional substitutions with P-EtN and L-Ara4N as indicated. As expected, we also observed a mass peak of 4489.9 Da, which can be explained by the addition of GlcNAc to glycoform I (supplemental Fig.   S1B). This addition of GlcNAc is derived from the O-antigen biosynthetic pathway (1). In contrast, LPS obtained from M9 medium revealed mass peaks at 3915.7 Da and several of its derivatives, predicted to correspond to typical glycoform I, with additional sodium and phosphate adducts (supplemental Fig.  S1A). Further mass peak at 3927.7 Da can be explained to be a derivative, corresponding to glycoform IV, with additional Na and P adducts (supplemental Fig. S1A). No mass peaks with predicted substitution of P-EtN and L-Ara4N were detected in the LPS from M9 growth conditions. The predicted modification of the inner core with the presence of the third Kdo and Rha, corresponding to glycoform IV FIGURE 1. Proposed LPS structures from E. coli K-12 in phosphate-limiting growth conditions of 121 minimal medium. Schematic drawing of LPS glycoform I, IV, and V compositions with various nonstoichiometric substitutions in the LPS core region is presented. Glycoform IV has Rha addition on the second Kdo. Glycoform V differs because Rha is added on the terminal third Kdo with P-EtN on the second Kdo. The cognate genes, whose products are involved at different steps, are indicated. DECEMBER 16, 2011 • VOLUME 286 • NUMBER 50 derivatives, was further supported by MS/MS analyses of isolated mass peaks corresponding to mass peaks at 3948.7, 4071.8, and 4202.8 Da (see below). Because substitution of P-EtN and L-Ara4N is indicative of correct translocation of LPS and contributes toward structural heterogeneity (3), we analyzed LPS of several isogenic nonpolar mutant derivatives to address the molecular basis of the third Kdo incorporation, using phosphate-limiting growth conditions. waaZ and waaS Are the Structural Genes for Incorporation of the Third Kdo and Rha, Respectively-Previous studies with overexpression of E. coli K-12 waaZ gene in E. coli strain with R1 LPS core suggested that the waaZ gene could be the structural gene encoding KdoIII transferase (7). Given the high complexity of LPS in standard LB medium, the structural details of LPS from a strain with chromosomal deletion of the waaZ gene and its individual contribution could not be analyzed in E. coli K-12 derivatives (7). However, mass spectrometric analyses of LPS extracted from the wild-type strain grown in 121 medium revealed clear distinction between glycoform I and glycoform IV (supplemental Fig. S1). Thus, LPS from ⌬waaZ and individual nonpolar deletion derivatives of all the genes in waa locus were analyzed. Significantly, LPS obtained from a ⌬waaZ derivative lacked typical mass peaks corresponding to complete glycoform IV with ions at 3948.7 Da. Such mass peaks at 3948.7 Da representing glycoform IV are present in the LPS obtained under the same growth conditions from the isogenic wild type. However, mass peaks at 3582.6 Da and its derivatives corresponding to LPS with a truncation of the outer core causing loss of terminal heptose and hexose are present. Such mass peaks are explained as LA hexa (Kdo 2 Hep 3 Hex 3 P 2 ) derivatives ( Fig.  2A). These mass peaks are characteristic of glycoform IV, but without the third Kdo and Rha. The presence of mass peaks at 3582.6 Da with the same truncation of the outer core as observed in the wild type with glycoform IV LPS and its deriv-FIGURE 2. WaaZ and WaaS are required for the incorporation of the third Kdo and Rha, respectively. Charge deconvoluted ESI FT-MS spectrum in negative ion mode of LPS from isogenic ⌬waaZ (A), ⌬waaZ ϩ pwaaZ ϩ (B), and ⌬waaS (C) strains. LPS was extracted from cultures grown at 37°C in 121 medium growth conditions. The mass numbers refer to monoisotopic peaks. The predicted composition with varying number of substitutions of P-EtN and with L-Ara4N substitution is indicated. Mass peaks corresponding to glycoforms IV and V, containing the third Kdo, are shown as rectangular boxes and glycoform I with complete core derivatives as circles. All mass peaks marked with a black star differ by 96 Da. The details of the composition of major mass peaks are described in supplemental Table S2. atives substituted with P-EtN and L-Ara4N suggests that the incorporation of the third Kdo may not be the reason for the truncation under these defined growth conditions. All such mass peaks also lacked Rha, as revealed by the loss of 146.1 Da, indicating that the incorporation of Rha required a prior addition of the third Kdo. The other mass peaks at 3936.7 Da and its derivatives represent typical ions corresponding to glycoform I with complete core and varying numbers of substitutions by P-EtN and L-Ara4N ( Fig. 2A).

RpoE-dependent and -independent Alterations in E. coli LPS
To confirm that the loss of the incorporation of the third Kdo was due to lack of WaaZ function, LPS from ⌬waaZ transformed with waaZ complementing plasmid was analyzed. In this plasmid the waaZ gene is expressed under the tight control of ptac promoter (13). As shown in Fig. 2B, typical mass peaks at 3948.7, 4071.8, and 4202.8 Da correspond to the predicted presence of glycoform IV derivatives with the third Kdo, as seen previously, in the LPS from plasmid-free wild type grown in 121 medium. Interestingly, mild induction of the waaZ gene product also revealed mass peak at 3514.7 Da and its derivatives. Such mass peaks correspond to predicted core containing LA hexa (Kdo 3 RhaHep 2 Hex 2 ), indicating that in vivo even these smaller structures can support the incorporation of the third Kdo. This was further supported by results from ⌬waaO mutants.
Analyses of LPS from ⌬waaS mutant revealed the presence of mass peaks at 3802.7 Da and its derivatives with a predicted incorporation of the third Kdo but lacking Rha (Fig. 2C). Furthermore, LPS of ⌬waaS also contained typical mass peaks corresponding to glycoform I derivatives with complete core (Hep 4 Hex 4 P 2 ) with or without additional predicted GlcNAc. Such mass peaks at 4286.9, 4489.9, and 4612.9 Da corresponding to glycoform I were also present in the isogenic wild type. The loss of Rha in the LPS of ⌬waaS, as compared with its presence, was also verified by GC/MS analyses (data not shown). Taken together, these results suggest that the incorporation of the third Kdo does not require prior incorporation of Rha, whereas Rha addition does require the addition of the third Kdo to the Kdo disaccharide. Furthermore, because the LPS extracted from ⌬waaS mutant contained otherwise normal glycoform I, the changes in composition are only specific to the loss of Rha incorporation in the inner core.
Addition of Rha in Inner Core Requires Prior Incorporation of the Third Kdo-Earlier work has shown that the LPS of E. coli B strains lack galactose and its core contains fewer heptoses and hexoses than other core types (22). However, the genetic basis of these LPS changes is not known. Furthermore, the confirmed presence of either Rha or the third Kdo is not clear in E. coli B. We first sequenced the waa region of E. coli B derivative BL21. DNA sequence analyses revealed that it has an IS element in the waaT gene with an overall chromosomal organization resembling the E. coli strains with R1 core. The waaT gene in the R1 core region encodes ␣-1,2-galactosyltransferase adding hexose at the same place where E. coli K-12 WaaR acts (23). Furthermore, both waaZ and waaS genes were found to be absent. Thus, this served as a good tool to further substantiate in vivo waaZ-and waaS-dependent modification of the inner core. Examination of LPS of three different E. coli B derivatives (BL21) revealed similar mass peaks, and the data for one repre-sentative are presented (Fig. 3A). The mass peak at 3420.6 Da can be explained as LA hexa (Kdo 2 Hep 3 Hex 2 P 2 P-EtN). Additional mass peaks at 3516.6, 3551.6, and 3647.7 Da, and their derivatives correspond to further modification by P-EtN, L-Ara4N, and also a derivative with additional 96 Da (Fig. 3A).
The lack of the third Kdo and Rha in BL21 LPS, even under growth conditions that favor their incorporation, is consistent with the absence of waaZ and waaS genes in its genome. Next, a plasmid containing either E. coli K-12 waaZ alone (pSR7960) or waaS (pGK2055) or waaZwaaS (pSR9298) with T7 polymerase-based expression system vectors was introduced in the same parental BL21 strain. The overexpression of the waaS gene alone revealed mass peaks like that of the parental BL21 strain without any new mass peaks (Fig. 3B). However, the induction of the waaZ gene expression resulted in new mass peaks, as represented by ions at 3859.7, 3982.7, and 4113.8 Da, which are predicted to arise from the addition of the third Kdo. This is in contrast to the presence of only Kdo 2 LPS forms in the parental strain or in waaS-overexpressing strains (Fig. 3C). However, co-overexpression of waaZ and waaS genes from the plasmid containing both genes behind individual T7 promoter revealed mass peaks with predicted incorporation of the third Kdo and also those with further addition of Rha. Thus, the addition of the third Kdo can explain mass peaks at 3859.7 and 3990.8 Da as indicated (Fig. 3C). The predicted further addition of Rha can explain the mass peak at 4136.8 Da. However, no mass peaks with the predicted addition of Rha without prior addition of KdoIII could be detected by mass spectrometric analysis. Thus, the incorporation of Kdo 3 ϩ Rha in E. coli B upon co-overexpression of both waaS and waaZ genes complements our results from ⌬waaS or ⌬waaZ E. coli K-12 derivatives. Thus, we conclude that waaS and waaZ genes are the structural genes required for the incorporation of Rha and the third Kdo linked to the Kdo disaccharide. Furthermore, these results also show that the incorporation of Rha requires prior and sequential addition of the third Kdo.
Structural Requirements for Incorporation of the Third Kdo in Vivo-We showed earlier the preference of glycoform IV LPS in strains synthesizing tetraacylated lipid A, however, with intact genes for LPS core biosynthesis (4). Here, we systematically analyzed isogenic in-frame nonpolar deletions in structural genes encoding different glycosyltransferases with respect to the ability to incorporate the third Kdo and Rha to define the minimal core structure required for the addition of the third Kdo. LPS extracted from ⌬waaC, ⌬waaF, ⌬waaG, and ⌬waaP mutants were found to lack the third Kdo and Rha. Even overexpression of the waaZ gene product from a plasmid could not reveal incorporation of the additional third Kdo using immunostaining with Kdo-specific antibodies and chemical and mass spectrometric analyses (data not shown). However, deletion derivatives of waaL and waaU genes, encoding O-antigen ligase and putative heptosyltransferase IV, respectively, were found to have both glycoform I and IV derivatives, indicating that their products do not effect overall the incorporation of the third Kdo. Concerning the minimal in vivo LPS structure that can allow the addition of the third Kdo and Rha, LPS of ⌬waaO mutant, even without any additional waaZ plasmid-borne overexpression, was found to incorporate the third Kdo and DECEMBER 16, 2011 • VOLUME 286 • NUMBER 50

RpoE-dependent and -independent Alterations in E. coli LPS
Rha as revealed by the presence of mass peaks at 3768.7 Da. This mass peak can be explained as hexaacylated LPS with Kdo 3 RhaHep 2 Hex 2 P with L-Ara4N and 2P-EtN residues (supplemental Fig. S2A). Furthermore, LPS from strain overexpressing the waaZ gene in ⌬waaO background revealed the presence of several mass peaks with the predicted incorporation of the third Kdo and Rha in both pentaacylated and hexaacylated derivatives (supplemental Fig. S2B). This is exemplified by the presence of mass peaks with ions at 3304.5, 3514.7, and 3637.7 Da. Such accumulation of pentaacylated mass peaks with the third Kdo and Rha in ⌬waaO derivative suggests that LPS forms lacking myristoyl secondary chain can serve as substrate for the incorporation of the third Kdo. This was further supported by the examination of LPS by mass spectroscopy of ⌬lpxM mutant with the intact core biosynthetic pathway. Such a mutant with only pentaacyl lipid A contains LPS derivatives of both glycoform I and IV (data not shown). Examination of LPS of either ⌬waaP or ⌬waaG strains with or without the waaZ gene overexpression did not reveal any mass peaks corresponding to the substitution with the third Kdo and Rha. These results suggest that both phosphorylation of HepI and a minimal outer core structure with two Hep and two Hex residues is required for the synthesis of derivatives with the third Kdo and Rha in vivo consistent with results of ⌬waaO mutants.
Galactose Addition Is Required for Incorporation of the Third Kdo in the Absence of WaaO-Because the LPS of E. coli K-12 contains Gal substitution on GlcI, we analyzed LPS of ⌬waaB mutants. The waaB gene in Salmonella enterica serovar Typhimurium has been shown to encode the glucosyl LPS 1,6-galactosyltransferase (24). Examination of LPS from ⌬waaB revealed mass peaks corresponding to two derivatives as follows: one with a predicted addition of WaaO-dependent glucose and another without such glucose incorporation, causing a truncation. The truncated derivatives correspond to Hep 2 Hex in the core region and lacked at the same time the third Kdo and Rha. However, mass peaks with a predicted core composition of Hep 3 Hex 2 were found to be predominantly substituted by the third Kdo and Rha (data not shown). Such derivatives can be explained to arise from the incorporation of the WaaO-dependent GlcII and the WaaQ-dependent side chain heptose in waaB mutants.
Because LPS of ⌬waaB exhibited considerable complexity because of variable amounts of Hep and Hex residues, a double nonpolar chromosomal ⌬(waaB-waaO) mutant was constructed to further investigate the minimal structure in vivo. This was necessary to answer the question whether prior incorporation of galactose and/or the second glucose is required for the incorporation of the third Kdo. Examination of its LPS revealed mass peaks predicted to correspond to LPS with two Kdo, two or three Hep, and a single Hex (supplemental Fig.  S2D). The mass peak at 3258.5 Da is predicted to contain the side chain heptose HepIII and the phosphorylated HepII (supplemental Table S2). Additional derivatives with mass peaks at 3389.6 and 3512.6 Da can be explained with additional L-Ara4N and further P-EtN substitutions. Most significantly, LPS of ⌬(waaB-waaO) did not reveal any mass peaks containing the third Kdo and Rha. Such a mutant did not confer the ability to incorporate the third Kdo even upon overexpression of the waaZ gene. These results thus suggest that the simultaneous lack of Gal and the GlcII does not allow the third Kdo incorporation. This could be due to structural constraints leading to reduced affinity of available minimal acceptor for WaaZ. Thus, the minimal in vivo structure able to accept the third Kdo is the form with two Hep residues in the inner core and two Hex residues in the outer core as shown with ⌬waaO mutant. Furthermore, the synthesis of glycoform IV and V does not require the presence of Gal attached to GlcI, when waaO is functional. These conclusions are supported by the results obtained from E. coli B derivatives, which lack Gal but can incorporate a third Kdo when the waaZ gene from E. coli K-12 was overexpressed (Fig. 3C).
As shown above, LPS of ⌬waaO, ⌬waaB, and ⌬(waaB-waaO) mutant strains contained several mass peaks with pentaacylated lipid A (supplemental Fig. S2). This could be either due to reduced amounts/activity of LpxM or due to structural preferences. Thus, we examined levels of LpxM-FLAG in waaO and waaB mutants as compared with the wild type. No significant differences were observed (data not shown), and this regulation (defect) needs further study.
Side Chain Heptose Addition Does Not Require WaaO and WaaB Function-Our data also revealed that neither the Gal addition (WaaB-dependent) nor the GlcII addition (WaaO-dependent) are prerequisite for the HepIII incorporation. This is evident from the presence of a mass peak at 3674.6 Da in the LPS from ⌬waaO mutant (supplemental Fig. S2A). Similarly, the mass peak at 3258.5 Da can be explained to include the phosphorylated HepII and additional substitutions with or without additional P-EtN and L-Ara4N as indicated (supplemental Fig. S2D).
Control of waaZ Expression-The molecular basis of the incorporation of the third Kdo was addressed by measuring waaZ-lacZ promoter activity and accumulation of chromosomal FLAG-tagged WaaZ by Western blot analyses. Cloning of regions upstream of the waaZ coding sequence in single copy pro-moter probe vectors identified an additional promoter located between the waaY and waaZ ORFs. The shift of the culture conditions from phosphate-rich M9 medium to phosphate-limiting medium revealed an increase by about 2-fold of the waaZ-lacZ promoter activity (Fig. 10A). These results were further supported by a nearly similar observed increase in the accumulation of WaaZ-FLAG upon Western blot analysis (Fig. 8C).
Induction of RpoE Leads to Preferential Accumulation of LPS Glycoform IV and V with the Third Kdo-The main function of the RpoE factor is to control OMP biogenesis and some of the steps of LPS translocation (8). RpoE in turn responds to outer membrane perturbations. This includes signals like imbalance in OMP composition and defects in the early steps of LPS biosynthesis (4,8,17,25). Thus, we analyzed LPS from mutants with either constitutively induced RpoE or strains with mutations in genes predicted to regulate LPS biosynthesis/modifications. Several isogenic nonpolar deletion derivatives were constructed, including strains deleted for the main negative regulators of RpoE like rseA and rseB (26,27). Deletion derivatives of hfq, mgrR, rpoS, and yrbC/D genes were also constructed and their LPS analyzed. Among these, hfq mutants show up-regulated RpoE activity, besides alterations in several functions controlled by noncoding RNAs and RpoS factor. Products of yrb/mla locus have been implicated in phospholipid migration to OM (28). mgrR encoding noncoding RNA has been implicated in the translational repression of the eptB gene product (phosphoethanolamine transferase for P-EtN addition to the second Kdo) (29). Among all of these, the most dramatic alteration in the LPS composition was that obtained from rseA mutants (Fig. 4). ⌬rseA mutants elicit constitutive induction of the rpoE regulon, given the known function of RseA acting as an antifactor for RpoE. In contrast, LPS of other mutants, including rseB, mgrR, rpoS and hfq, depicted mass peaks quite like that obtained from the wild type and were found to contain both glycoform I and IV derivatives (Fig. 5).
Mass spectrometric analyses of LPS from the ⌬rseA mutant revealed nearly the exclusive presence of mass peaks corresponding to the predicted incorporation of the third Kdo and Rha and a characteristic truncation with a loss of terminal Hep-Glc disaccharide. The expected glycoform IV with the third Kdo and Rha derivatives can be assigned to mass peaks at 3921.7, 3948.7, 4071.7, 4167.8, 4202.8, and 4298.9 Da (Fig. 4B). For example, mass peak with ion at 3948.7 Da can be assigned to LPS containing LA hexa (Kdo 3 RhaHep 3 Hex 3 P 2 P-EtN). The other mass peaks can be attributed to arise due to additional substitutions with one or two additional P-EtN and L-Ara4N residue(s). Incorporation of P-EtN and L-Ara4N was also observed from predicted and observed lipid A mass spectrometric analyses (data not shown). The mass peaks at 3921.7, 4167.8, and 4298.8 Da can be explained as new derivatives with an addition of 96 mass units. The origin of this modification cannot be explained on the basis of known E. coli LPS structures. This new modification seems to arise because of HexA substitution on HepIII accompanied by the loss of phosphate on HepII 3 (see under "Discussion").
Interestingly, only mass peaks with predicted hexaacylated LPS derivatives containing two Kdo residues in ⌬rseA background are represented by mass peaks at 3582.7 and 3705.6 Da. These can be explained as precursors of glycoform IV with characteristic absence of the terminal Hep-Hex disaccharide and the lack of the third Kdo and Rha with an overall composition of LA hexa (Kdo 2 Hep 3 Hex 3 P 2 P-EtN) (Fig. 4B). The lack of any major structural differences in ⌬rseB mutants as compared with dramatic changes in ⌬rseA has important implications. RseB is a minor negative regulator as compared with major negative control by RseA of RpoE factor (26,27). Thus the maximal induction of the RpoE regulon is required for near exclusive synthesis of glycoforms IV and V as observed in rseA mutants.
To ascertain that the observed synthesis of primarily glycoform IV was WaaZ-dependent, a double mutant ⌬(waaZ rseA) was constructed, and its LPS was analyzed. Data presented in Fig. 4E revealed mainly mass peaks corresponding to predicted glycoform I derivatives, lacking the third Kdo; these are represented by ion peaks from 3893.7 to 4489.9 Da (Fig. 4E). However, analyses of LPS of ⌬(waaZ rseA) also revealed the presence of mass peaks predicted to be precursors of glycoform IV without the third Kdo and Rha, with the truncation of the terminal Hep-Glc disaccharide. Such mass peaks resemble LPS derivatives of a ⌬waaR mutant, represented by the mass peak at 3582.6 Da and its derivatives. Furthermore, LPS of ⌬(waaZ rseA) unexpectedly was found to contain mass peaks, which correspond to the predicted addition of phosphate residues as revealed by peaks at 3893.7, 3973.7, and 4016.7 Da. The accumulation of such mass peaks of glycoform I derivatives in ⌬(waaZ rseA) LPS might be a compensatory mechanism for the loss of the third Kdo in such a background. Thus, the overall induction of RpoE because of ⌬rseA mutation seemed to result in reduced amounts or activity of WaaR, resulting in the accumulation of LPS with truncation after HexIII and activation of waaZ-dependent pathway for glycoform IV and V biosynthesis.
Position of Rha Dictated by the Presence of P-EtN on the Second Kdo-As shown above, the isolation of LPS from phosphate-limiting growth conditions led to a significant shift in the accumulation of glycoform with Kdo 3 Rha incorporated. This shift to Kdo 3 Rha is however most dramatic and pronounced in RpoE-inducing conditions as shown in the case of rseA mutants. We previously showed that 121 medium containing 2 mM Ca 2ϩ is sufficient to allow a nonstoichiometric substitution   DECEMBER 16, 2011 • VOLUME 286 • NUMBER 50 of P-EtN on the second Kdo (4). EptB is known to require Ca 2ϩ for its activity (30). Furthermore, its relative incorporation depends on the induction of the eptB gene encoding phosphoethanolamine transferase specific to the second Kdo. The transcription of the eptB gene is induced upon RpoE up-regulation. Until now, the structure assignment of glycoform IV is based on the analysis of an oligosaccharide, which lacked P-EtN on the second Kdo (6). Thus, we analyzed in detail the KdoRha linkage under the conditions of the presence or the absence of P-EtN on the second Kdo. It needs to be emphasized that the second Kdo is also a site for the nonstoichiometric substitution by P-EtN. To address it, we analyzed composition of several isolated mass peaks predicted to contain KdoRha from LPS obtained from strains either lacking the eptB gene or when its expression is favored like in rseA mutants.

RpoE-dependent and -independent Alterations in E. coli LPS
To demonstrate this switch, triply charged molecular ions from LPS obtained from either the wild-type (eptB ϩ ) or rseA mutants were isolated and fragmented by the collision-induced dissociation. Main fragment ions result from the cleavage of the labile LA-Kdo linkage, comprising single charged LA [LA-H ϩ ] 1Ϫ species and doubly charged core oligosaccharide fragment ions [Core-H 2 O-2H ϩ ] 2Ϫ and a group of triply charged fragments, which provide information on the Kdo substitution. MS/MS spectra with parental ion at 4071.7 and 4202.8 Da revealed mass peaks at 3705.6 and 3836.7 Da, which can be explained by simultaneous loss of 366.3 Da. They are explained to arise by Kdo ϩ Rha cleavage from the branched Kdo 3 -Rha tetrasaccharide (Fig. 6, B and C). This indicates that Rha is linked to the terminal KdoIII. Additional mass peaks from MS/MS spectra of parental ions at 4071.7 and 4202.8 Da revealed ion peaks at 3362.6 and 3493.6 Da, respectively. These are explained to arise from further simultaneous loss of (KdoP-EtN) residues from ions at 3705.6 and 3836.7 Da, respectively. These new LPS forms are hence designated as glycoform V ( Fig.  1 and supplemental Fig. S1). Data in Fig. 6, inset, are charge deconvoluted mass spectra showing the most relevant features of fragmentation. These results and examination of several other peaks, for example ion peaks at 4079.7 and 4298.8 corresponding to glycoform V derivatives with P-EtN on the second Kdo, always were found to have KdoIIIRha as shown here for parental ions at 4071.7 and 4202.8 Da. This interesting change caused by the substitution point of P-EtN on KdoII determining the presence of Rha on KdoIII was not observed when the ion at 3948.7 Da was subjected to MS/MS analysis. This parental ion was not found to have P-EtN on the second Kdo and can explain ions arising at 3728.7 Da by the loss of the third Kdo and further loss of Rha leading to the mass peak at 3582.6 Da (Fig. 6A). Thus, the presence of P-EtN on KdoII leads to the addition of Rha to the third Kdo. Because incorporation of Rha was found to require prior addition of the third Kdo, the most favored position for Rha is the third Kdo under the conditions of EptBdependent P-EtN addition on KdoII. Induction of RpoE, as is the case with ⌬rseA mutants, is known to exhibit constitutive transcriptional induction of the eptB gene (4,31).
To further substantiate the above results, the MS/MS analysis of isolated mass peaks at 3948.7, 4071.7, 4079.7, and 4202.8 Da were carried out from LPS obtained from a ⌬eptB strain. Fragmentation spectra of such ion peaks revealed that in all such cases the terminal third Kdo was cleaved without Rha, which can explain mass peaks at m/z 1241.9, 1286.2, and 1326.9 interpreted as (M Ϫ H) 3ϩ ions (Fig. 7). This is in contrast to fragmentation spectra of isolated ion peaks of 4071.7 and 4202.8 Da from eptB ϩ strains, which revealed simultaneous loss of Kdo ϩ Rha from the branched Kdo 3 Rha tetrasaccharide (Fig.  6, B and C). Fragmentation spectra of all four parental ion peaks from ⌬eptB showed that predicted Rha cleavage occurs only after the loss of the terminal third Kdo. Furthermore, no ion peaks corresponding to fragmentation of KdoP-EtN was observed in the fragmentation spectra of ion peaks from eptB mutants. Because EptB is known to function as P-EtN transferase specific to the second Kdo, we can conclude that the incorporation of P-EtN is required to synthesize glycoform V with a switch of Rha addition to the terminal third Kdo.
RpoE-dependent Enhanced Glycoform IV and V Accumulation-One possible explanation for the nearly exclusive presence of glycoform IV and V derivatives in ⌬rseA could be that E E RNA polymerase initiates the waaZ transcription. In such a scenario, ⌬rpoE mutants are expected not to contain glycoform IV and V derivatives and have LPS resembling that of ⌬waaZ mutants. However, analyses of LPS from a ⌬rpoE strain revealed the presence of characteristic mass peaks at 3948.7, 4079.8, and 4202.8 Da with predicted incorporation of the third Kdo and Rha (glycoform IV and V) and not like that of LPS from ⌬waaZ mutants (Fig. 4D). As can be seen from the mass spectrometric analysis of LPS obtained from ⌬rpoE, several mass peaks with predicted glycoform I with or without the addition of HexNAc and the nonstoichiometric addition of P-EtN and L-Ara4N are also present. Thus, the overall distribution of mass peaks in ⌬rpoE LPS resembled more or less that of the isogenic wild type. Mapping of the transcriptional initiation site further supported these results. In such an analysis, no obvious RpoEregulated promoter upstream of the waaZ gene region was detected (data not shown). Thus, the overall LPS composition with preferential synthesis of either glycoform I or IV derivatives is not due to direct RpoE-dependent transcriptional initiation of the waaZ gene.
RpoE-regulated Noncoding Small RNA rybB-mediated Regulation of Glycoform IV and V Accumulation-To further understand the molecular basis of LPS changes upon the RpoE induction, we analyzed LPS of several mutants with either individual deletions in RpoE regulon members or their deletion combinations with ⌬rseA. The RpoE factor is known to be induced upon imbalance or misfolding of porins (17,25). We studied the potential control by micA and rybB noncoding sRNAs. These sRNAs are known to control translation of major porins, and synthesis of these sRNAs is regulated by RpoE (32). A deletion of individual micA or rybB genes did not cause any major alterations in the composition of LPS (data not shown). However, LPS from the ⌬(rseA rybB) derivative prominently resulted in restoration of the glycoform I presence, resulting in LPS composition quite like the wild type (Fig. 4C). LPS of ⌬(rseA micA) was also found to contain some glycoform I derivatives but overall only a partial suppression of rseA phenotype in terms of preponderance of glycoform IV and V content (data not shown). The restoration of glycoform I accumulation as seen by characteristic mass peaks appearance and suppression of glyco-form IV and V accumulation with the third Kdo and Rha in ⌬(rseA rybB) was further addressed by examining levels of WaaZ and other effector proteins.
Glycoform IV and V Accumulation Is Due to rybB-dependent Translational Repression of WaaR in ⌬rseA-Because the introduction of ⌬rybB in ⌬rseA caused suppression of glycoform IV and V accumulation, we examined the levels of WaaZ and WaaR from their chromosomal copy using C-terminal FLAG epitope derivatives. The rationale for this is the commonly observed truncation of the terminal Hep-Glc disaccharide in the outer core upon the incorporation of the third Kdo associated with the presence of glycoform IV (4,6,7). First, in-frame chromosomal waaR-FLAG and waaZ-FLAG were constructed in the wild-type strain and then such FLAG-FIGURE 6. Incorporation of P-EtN on the second Kdo results in switch of Rha addition to terminal third Kdo, leading to synthesis of glycoform V. A corresponds to fragmentation spectra of parent ion at 3948.7 Da representing glycoform IV. In accordance to its published fragmentation behavior (6), a sequential loss of the first Kdo, followed by the cleavage of a Rha, confirms the substitution of Rha at the middle Kdo. B and C depict fragmentation spectra of parent ions at 4071.7 and at 4202.8 Da, respectively, both representing glycoforms V. Here, instead of the cleavage of one Kdo alone only the cleavage of (Kdo ϩ Rha) is observed. Furthermore, the cleavage of labile linked L-Ara4N and P-EtN can be observed either from the complete molecular ions as from the singly charged LA fragment ion. DECEMBER 16, 2011 • VOLUME 286 • NUMBER 50 marked WaaZ and WaaR alleles were transduced in ⌬rseA and ⌬(rseA rybB) backgrounds. Whole cell extracts from such isogenic bacteria were prepared and analyzed. Consistent with the requirement of WaaZ for the incorporation of the third Kdo and the nearly exclusive glycoform IV and V presence in ⌬rseA, the level of WaaZ was higher in the ⌬rseA. This increase in WaaZ levels was particularly more prominent in the stationary phase, as compared with the WaaZ levels in the isogenic wild type (Fig. 8C, lanes 3 versus 4). Thus, in a ⌬rseA background the incorporation of the third Kdo is at least in part due to WaaZ accumulation in stationary phase, but it is not sufficient to explain LPS composed of mainly glycoform IV and V derivatives.

RpoE-dependent and -independent Alterations in E. coli LPS
As shown above, both ⌬rseA and its isogenic ⌬(rseA waaZ) mutant derivatives caused accumulation of LPS precursor with truncation of terminal Hex-Hep disaccharide. These results FIGURE 7. eptB mutants lack glycoform V. ESI FT-ICR mass spectra are shown of the negative ion mode after isolation of parental ions of mass peaks obtained from LPS of eptB mutants, which lack P-EtN on Kdo 2 . Isolated ions peaks with the third Kdo and Rha were subjected to fragmentation analysis, and mass numbers correspond to (M Ϫ H ϩ ) Ϫ ions for lipid A and (M Ϫ H) ϩ3 for derivatives from cleavage in the Kdo region. A-C corresponds to MS/MS spectra from isolated molecular species at 3948.7, 4079.8, and 4202.8 Da, respectively. As depicted in the mass spectra, all selected species include first the cleavage of a Kdo then followed by the cleavage of Rha, indicating glycoform IV.
argue that in ⌬rseA mutant accumulation of glycoform IV and V could be due to WaaR defect, which can explain the resulting truncation. Thus, WaaR-FLAG levels were examined by Western blot analyses. As shown in Fig. 8, A and B, levels of WaaR are reduced in a ⌬rseA and restored back to near wild-type levels in ⌬(rseA rybB) derivative. Taken together, these results demonstrate that WaaR synthesis is subjected to translation repression upon RpoE activation via rybB sRNA resulting in the incorporation of the third Kdo. Thus, glycoform IV and V accumulation is primarily due to reduction in WaaR amounts at translational level.
⌬waaR Mutants Exhibit Enhanced Incorporation of the Third Kdo Leading to Preferential Accumulation of Glycoform IV and V-Based on the results showing that reduction in WaaR levels in rseA mutants or overall growth conditions in 121 medium is responsible for inducing the synthesis of glycoform IV and V with truncation of the outer core, we investigated LPS of waaR mutants. Consistent with the above results, mutant strains lacking glycosyltransferase due to nonpolar waaR deletion were found to primarily contain LPS with glycoform IV and V derivatives (supplemental Fig. S2C). All the main mass peaks at 3948.7, 4079.8, 4202.8, and 4325.8 Da correspond to characteristic glycoform IV and V derivatives with additional substitutions by P-EtN and L-Ara4N as indicated (supplemental Fig.  S2C). Other mass peaks with lower molecular masses can be explained as a premature termination of LPS synthesis or as precursors. This is evident from their predicted composition as indicated (supplemental Fig. S2C). Results from ⌬waaR suggest that the lack of the waaR gene itself or the presence of presumably preferred precursor with Hep 3 Hex 3 composition (mass peaks at 3582.6 Da) highly favors the incorporation of the third Kdo, leading to the synthesis of glycoform IV and V. Analysis of LPS of a ⌬waaR mutant grown in phosphate-rich growth conditions also revealed several mass peaks with predicted glycoform IV incorporation (supplemental Fig. S2E). Thus, in vivo LPS from strain ⌬waaR is the preferred substrate for the incorporation of the third Kdo. Hence, we can conclude that levels of WaaR seem to determine the switch between glycoform I and IV. Also induction of lipid A modifications (BasS/R-and PhoB/ R-dependent) is not required in ⌬waaR for the incorporation of the third Kdo.
Presence of Glycoform IV and V with Three Kdo in ⌬rseA Is Abolished in ⌬(relA spoT rseA)-The alarmone ppGpp is a general signal of starvation stress. ppGpp regulates the activation of alternative factors upon entry into stationary phase and also competition among various factors (33). ppGpp is also known to be involved in regulation of the PhoB/R two-component system and RpoE (33,34). Because in this study we show that induction of RpoE and PhoB/R controls switches in LPS composition, we analyzed LPS from isogenic ppGpp 0 (relA spoT) and ⌬(relA spoT rseA) strains. Examination of the mass spectra of LPS from ⌬(relA spoT) and the isogenic wild type revealed similar mass peak composition regarding the presence of glycoform I, IV, and V (Fig. 9A). However, introduction of ⌬rseA mutation in ⌬(relA spoT) (ppGpp 0 background) abolished the accumulation of mass peaks corresponding to glycoform IV and V derivatives. This resulted in the lack of detectable amounts of LPS with the third Kdo and Rha, which are predominant in ⌬rseA parental strain (Fig. 9B). Thus, no mass peaks at 3948.7 Da and its derivatives were detected in ⌬(relA spoT rseA). On the contrary, only mass peaks at 3936.7 Da and its derivatives corresponding to glycoform I are present. These mass peaks are represented by observed ions at 4262.8 and 4385.9 Da. They are predicted to arise upon further substitutions by P-EtN and GlcNAc. Interestingly, not only accumulation of glycoform IV and V was suppressed but also no mass peaks with predicted L-Ara4N incorporation were observed, which are present in both ⌬rseA as well as ⌬(relA spoT) (Fig. 9,  C and D). Analyses of mass peaks corresponding to the lipid A part revealed only a mass peak at 1920.2 Da (LA hexa P-EtN) but not that corresponding to L-Ara4N addition. Thus, ppGpp seems to control both LPS core composition and lipid A modification in RpoE-inducing conditions. These data also show that in E. coli L-Ara4N incorporation not only requires BasS/R FIGURE 8. Repression of WaaR and induction of WaaZ induce the third Kdo incorporation but not that of OMP. Culture of the wild type (wt) and its derivatives with or without the C-terminal 3ϫFLAG tag fusion were grown to early log phase in LB medium at 37°C, washed, and adjusted to A 595 of 0.02 in 121 medium growth conditions. Aliquots of samples were drawn at different intervals and harvested by centrifugation. Equivalent amounts of proteins (20 g) were lysed in SDS sample buffer and resolved on 12% SDS-PAGE. Proteins were transferred to PVDF membrane and probed with specific antibodies. Samples from the wild type and its derivative with in-frame rseA deletion, both carrying chromosomal waaR replaced by waaR-FLAG tag, were probed with monoclonal M2 antibody against FLAG tag (A). Samples prepared from WaaR-FLAG-tagged derivatives, carrying mutations in ⌬rybB and ⌬(rybB rseA), were prepared and resolved on SDS-PAGE under identical conditions and levels of WaaR-FLAG were determined as described above (B). Western blots of samples prepared from the wild-type and isogenic ⌬rseA strains carrying waaZ-FLAG replacement of the chromosomal waaZ gene were probed with FLAG-specific antibody (C). D-F correspond to Western blot analysis to reveal the relative amounts of OmpC, OmpF, and OmpA, respectively, from isogenic strains grown in 121 medium. The relevant genotype of each strain, from which samples were prepared, is indicated on the top of D.
induction but also seems to be further fine-tuned by ppGpp. Interestingly, until now ppGpp has not been implicated in the control of lipid A or LPS core structural alterations in E. coli or in other organisms.
PhoB/R and BasS/R Induction Is Required for Glycoform IV and V Presence in ⌬rseA but Not in Wild Type-We previously showed that phosphate limitation and components like Zn 2ϩ and Fe 3ϩ contribute to BasS/R-dependent lipid A modifications in 121 growth medium. Because growth of E. coli in this medium also caused a pronounced shift to the presence of glycoform IV and V, we analyzed the requirement of either BasS/R and/or PhoB/R two-component systems for this phenotype. Analyses of LPS extracted from isogenic ⌬basR and ⌬phoB bacteria grown in 121 medium revealed the presence of glycoform I with complete core and glycoform IV derivatives (supplemental Fig. S3A). However, LPS of ⌬basR mutant bacteria revealed much less heterogeneity as compared with either LPS from the wild-type or ⌬phoB mutant. This can be explained by the obvious lack of nonstoichiometric addition of P-EtN and L-Ara4N to lipid A (supplemental Fig. S3A). The mass peak at 4071.8 Da in ⌬basR can be assigned to glycoform IV and V with additional two P-EtN residues and that at 4167.8 Da with an additional substitution of 96 mass units. These results show that unlike P-EtN addition on lipid A, which is BasS/R-dependent, nonstoichiometric P-EtN additions on either the second Kdo or HepI are BasS/R-independent. This explains the presence of the mass peak at 4071.7 Da. Examination of LPS obtained from ⌬(basR phoB) derivative also revealed the presence of mass peaks corresponding to glycoforms I, IV, and V (supplemental Fig. S3B). However, significantly LPS of a strain ⌬(basR phoB rseA) contained mass peaks corresponding to glycoform I, IV, and V LPS derivatives in contrast to presence of mainly glycoform IV and V in ⌬rseA (supplemental Fig. S3C). These results argue that induction of both BasS/R and PhoB/R two-component systems contributes to synthesis/incorporation of glycoform IV and V in the ⌬rseA derivative, although individually both PhoB/R and BasS/R are dispensable in this switch of glycoforms. In support of these observations, LPS obtained from ⌬(eptA rseA) and ⌬(eptA arnT rseA) also revealed restoration of glycoform I synthesis in ⌬rseA as revealed by the presence of corresponding mass peaks (data not shown). However, this restoration of glycoform I in ⌬(eptA rseA) and ⌬(eptA arnT rseA) combinations is not to the wild-type extent as seen in ⌬(rseA rybB) mutants, arguing overlapping control by lipid A modification system as well as the translational control of WaaR by rybB noncoding RNA for the switch to glycoform IV and V.
Preferential Glycoform I or IV and Impact on OMP Levels-It is established that a ⌬rseA mutant exhibits constitutively elevated RpoE activity leading to reduced amounts of OMPs. This is ascribed to induction of RpoE-regulated noncoding RNAs, which down-regulate porin synthesis (32). The results presented in this work further revealed that such ⌬rseA bacteria primarily synthesize LPS composed of mainly glycoform IV and V derivatives. Thus, we asked whether the overall switch of glycoform I to glycoform IV and V is also reflected in any in vivo alterations of OMP levels. It is compelling because ⌬(rseA rybB) combination leads to restoration of OMP composition resulting in suppression of several defects of ⌬rseA (35). We also showed that ⌬(rseA rybB) exhibit a switch back to glycoform I synthesis. With the construction of panels of several single or various mutational combinations synthesizing either glycoform I or glycoform IV and V, we determined the relevance of the incorporation of the third Kdo (glycoform IV and V) for OMP composition. Whole cell lysates were prepared from the wild type, ⌬rseA, ⌬(relA spoT), ⌬(relA spoT rseA), ⌬waaR, and ⌬waaZ mutants. Levels of OmpF, OmpC, and OmpA were compared under identical growth conditions after Western blotting and were revealed using specific antibodies against each protein. Among the strains used, ⌬waaZ and ⌬(relA spoT rseA) were shown to have LPS composed mainly of glycoform I derivatives (Figs. 2A and 9B) and in contrast ⌬rseA and ⌬waaR possessed primarily LPS of glycoform IV and V in 121 growth conditions ( Fig. 4B and supplemental Fig. S2C). As expected, rseA mutants accumulated reduced levels of major OMPs, but ⌬waaR mutants with nearly similar LPS composition revealed OMP content (levels) quite like the wild type (Fig. 8, D-F). These results argue that the induction or preferential presence of glycoform IV and V does not contribute to OMP reduction. Examination of OMP levels of ⌬waaZ and ⌬(relA spoT rseA), however, showed highly reduced levels of OMPs in ⌬(relA spoT rseA), although ⌬waaZ possessed OMP levels nearly similar to the wild type, despite the two mutants representing LPS with only glycoform I derivatives. Interestingly, ⌬(relA spoT) mutants also contained reduced amounts of OMPs, which were not suppressed in ⌬(relA spoT rseA), although the LPS composition with respect to glycoform I versus glycoform IV and V was dramatically different. Taken together, these results suggest that in vivo preferential synthesis of either glycoform I or glycoform IV and V does not interfere in the OMP biogenesis significantly. These results are consistent with lack of any RpoE induction in ⌬waaR mutants (see below).
RpoE Signal Transduction in waaR Mutants, Consequences of Truncation of Outer Core-Because ⌬rseA and ⌬waaR revealed nearly similar LPS mass peaks corresponding primarily to glycoform IV and V derivatives with truncation of the outer core, we examined whether ⌬waaR also exhibits constitutive hyperinduction of the RpoE regulon, which responds to OM dysfunction as seen in ⌬rseA (26,35). A defined ⌬rseA mutant is known to exhibit a 6 -10-fold basal level increased transcriptional activity from RpoE-regulated promoters such as rpoHP3, rpoEP2, and htrA (26,27). Thus, a ⌬waaR mutation was introduced in the wild-type strain carrying single copy E E -transcribed lacZ promoter fusions. Data consistently revealed that no increase in the activity of a representative RpoE-regulated promoter fusion (rpoHP3-lacZ) occurred. Rather, in all tested growth conditions, the activity of rpoHP3-lacZ was lower in ⌬waaR mutant than in the isogenic parent (Fig. 10B). Thus, despite similar LPS composition of ⌬waaR and ⌬rseA, the observed increased presence of glycoform IV and V with the third Kdo does not result in RpoE activation. Thus, the preferential presence of glycoform IV and V can occur independent of RpoE activation as in ⌬waaR, and this LPS modification (increased accumulation/addition of the third Kdo) is not directly responsible for the RpoE induction.
Requirement of WaaZ in Strains Synthesizing Tetraacylated Lipid A with Intact waaC Gene-In this work, we showed that the constitutive induction of RpoE in ⌬rseA mutants induces the shift to preferential WaaZ-dependent glycoform IV and V incorporation. Earlier, we showed that strains lacking late acyltransferases, but with intact waaC, also showed preponderance of these glycoforms with the third Kdo. Thus, we addressed if this WaaZ-dependent incorporation of a glycoform with the third Kdo has a physiological function. Introduction of ⌬waaZ in ⌬rseA background did not confer any noticeable growth defects or suppress dramatic phenotypes like cell lysis or overall RpoE activity. However, a requirement for WaaZ was observed in ⌬(lpxL lpxM lpxP) mutants. ⌬(lpxL lpxM lpxP) mutants do not grow at 30°C on rich medium but can grow on minimal medium (4,36). In this study, we further show that suppressorfree ⌬(lpxL lpxM lpxP) mutants can be constructed on rich medium at 21 or 23°C but not at 30°C or above (Table 2). Next, to test the requirement of WaaZ defined nonpolar ⌬waaZ mutation was introduced in a ⌬(lpxL lpxM lpxP) strain. Thus, ⌬(lpxL lpxM lpxP waaZ) and ⌬(lpxM lpxP waaZ) mutants could be constructed at 21 or 23°C in the presence or absence  ). B, cultures of E. coli wild-type strain GK1111 carrying single copy chromosomal rpoHP3-lacZ promoter fusion or its isogenic derivative with ⌬waaR mutation were grown to early log phase in LB medium at 37°C. Cultures were washed and adjusted to A 595 of 0.02 in M9 or 121 medium. Aliquots of samples were drawn at different intervals. The A 595 was measured and analyzed for ␤-galactosidase activity as indicated above. Error bars represent S.E. of four independent measurements. ⌬(lpxM lpxP) ϩϩ a ϩϩ ϩϩ ϩϩ ϩϩ ⌬(lpxM lpxP) ϩ ⌬waaZ ϩϩ ϩϩ ϩϩ ϩϩ ϩ b ⌬(lpxM lpxP) ϩ ⌬lpxL ϩϩ ϩϩ ϩϩ Ϫ c ⌬(lpxL lpxM lpxP) ϩ ⌬waaZ ϩϩ ϩ b Ϫ Ϫ ⌬(lpxL lpxM lpxP) ϩ ⌬waaZ ϩ plpxL ϩ ϩϩ ϩϩ ϩϩ ϩϩ a ϩϩ indicates Ն500 colonies. b ϩ indicates 100 -500 colonies but is small in size. c Ϫ indicates inability to support colony forming capacity. of plasmid covering one of these genes by T4-mediated transductions on minimal M9 medium. However, viable ⌬(lpxL lpxM lpxP waaZ) were not obtained without complementing plasmid on rich medium at either 21 or 23°C (Table 2). Taken together, these results showed that WaaZ indeed has a physiological role in E. coli K-12 strains with tetraacylated lipid A with intact LPS core region.

DISCUSSION
In this study, using phosphate-limiting growth conditions, we found a pronounced shift to the presence of LPS derivatives with the third Kdo and Rha linked to lipid A-anchored Kdo disaccharide with a characteristic truncation of the outer core designated as glycoform IV. This glycoform is usually a minor component of LPS, whereas the major glycoform with a complete core is designated as glycoform I (LA hexa Kdo 2 Hep 4 Hex 4 P 2 ) (6). We demonstrated that the waaZ gene is the structural gene required for addition of the third Kdo to the Kdo disaccharide linked to lipid A. Accordingly, LPS of a ⌬waaZ strain contained only typical glycoform I and no glycoform IV derivatives with the third Kdo. However, mass peaks predicted to correspond to precursor(s) of glycoform IV with truncation of the terminal Hep-Hex without the third Kdo were found. These results suggested that the truncation of the outer core is not directly due to incorporation of the third Kdo. Interestingly, no mass peaks predicted to contain Rha were found in ⌬waaZ. Furthermore, we show that the addition of Rha requires WaaS function. ⌬waaS strain could add the third Kdo to Kdo disaccharide without the addition of Rha. These results argue that addition of Rha occurs only after the third Kdo is incorporated. Consistent with these findings, we showed that E. coli B strains, naturally lacking waaZ and waaS genes, can incorporate Rha in the LPS inner core only after WaaZ-dependent addition of the third Kdo.
We also showed that the presence of P-EtN on the second Kdo (penultimate Kdo) in branched Kdo 3 Rha tetrasaccharide leads to a switch of Rha addition to the third Kdo, hence defining glycoform V. The change of the position of Rha to the third Kdo was more pronounced under the conditions of RpoE induction, such as in a ⌬rseA mutant. MS/MS analysis of several isolated mass peaks corresponding to glycoform V with P-EtN on the second Kdo revealed Rha only on the third Kdo. Kdo 2 P-EtN-dependent switch of Rha substitution was confirmed by MS/MS analysis of isolated mass peaks of LPS from a ⌬eptB mutant. EptB is known to be required for the transfer of P-EtN to the second Kdo (30). Indeed all the mass peaks with the third Kdo in ⌬eptB were found to contain Rha substitution on the second Kdo, thus corresponding to only glycoform IV. In conclusion, P-EtN on the second Kdo dictates Rha addition to the third Kdo.
Structural and compositional analyses of LPS obtained from various mutants allowed us to demonstrate that in vivo the minimal LPS structure that can support the incorporation of the third Kdo to generate Kdo 3 Rha is Kdo 2 Hep 2 Hex 2 with requirement for phosphorylation of HepI. The second Hex residue can be either Gal or Glc. These results were authenticated by the analysis of LPS from ⌬(waaB-waaO), which could not support the incorporation of the third Kdo even upon overexpression of the waaZ gene product. This is in contrast to waaO mutants with intact waaB, containing LPS with a Kdo 3 Rha branched tetrasaccharide and hence the core with Hep 2 GlcGal. Because ⌬waaB, as well as E. coli B which naturally lack Gal, can incorporate LPS substituted with Kdo 3 Rha, showed that Gal addition in LPS is not a prerequisite for this modification.
We addressed the question whether any molecular switch regulates preferential synthesis of glycoform I or glycoform IV and V. This switch in phosphate-limiting growth conditions was in part ascribed to increased transcription of the waaZ gene. Because growth in this medium induces both PhoB/R and BasS/R two-component systems, LPS from individual ⌬phoB, ⌬basR and double ⌬(basR phoB) was analyzed. Individual ⌬phoB or ⌬basR mutants were found to contain both glycoform I and glycoform IV and V. LPS composition was enriched in glycoform IV in a ⌬waaR mutant even in growth conditions that are repressing for PhoB/R and BasS/R. This is an argument that the induction of PhoB/R and BasS/R systems is not absolutely essential for the synthesis of glycoform IV. Thus, these results suggest that additional control mechanisms for the incorporation of the third Kdo exist, particularly at the control of WaaZ versus WaaR levels.
Because several cell envelope functions including the amounts of OMPs are regulated by RpoE factor (8) and LPS is the major component of OM, we analyzed LPS from strains, which exhibit constitutive induction of RpoE. Interestingly, ⌬rseA mutants contained primarily LPS with glycoform V derivatives. RseA acts as RpoE-specific antifactor. Thus, ⌬rseA, genes which are transcribed by E E RNA polymerase, are constitutively up-regulated, although the synthesis of several OMP is repressed (8,26,37). However, ⌬rpoE was found to contain glycoforms I, IV, and V derivatives quite like the wildtype strain. This led us to investigate whether any RpoE-induced gene product might be responsible for the preponderance of glycoform IV and V in ⌬rseA. Because RpoE does not positively regulate waaZ transcription, it argued for potential regulation via a negative control process under the control of RpoE. As RpoE-inducible noncoding RNAs of rybB and micA are known to repress the synthesis of OMPs, LPS of ⌬(rseA rybB) and ⌬(rseA micA) strains was analyzed. Interestingly, introduction of ⌬rybB mutation in ⌬rseA suppressed the ⌬rseA-dependent glycoform IV and V preponderance. Thus, LPS of ⌬(rseA rybB) strain contained both glycoform I and glycoform IV and V derivatives quite like the wild type. Thus, rybB seems to down-regulate a step that promotes WaaZ synthesis and hence the incorporation of the third Kdo. This was shown to occur at repression of WaaR synthesis in ⌬rseA mutants and its restoration to wild-type levels in ⌬(rseA rybB). Consistent with the induction of glycoform IV and V synthesis in ⌬rseA, related to reduction in WaaR amounts, a defined waaR mutant synthesized LPS composed of mainly glycoform IV derivatives. Thus, the increase of WaaZ synthesis and reduction in WaaR levels are the main factors that cause a switch to the synthesis of glycoform IV as seen in ⌬rseA mutants.
Another mechanism of regulation of RpoE-dependent increased synthesis of glycoform IV and V derivatives with the third Kdo was discovered. This was found to occur via the alarmone ppGpp. ppGpp is known to alter the specificity of several DECEMBER 16, 2011 • VOLUME 286 • NUMBER 50 alternative factors. Further ppGpp and phosphate starvations have many overlapping functions (37). Analyses of LPS of ppGpp 0 mutants revealed no significant differences as compared with the wild type. However, introduction of ⌬rseA in ppGpp 0 mutants abolished accumulation of glycoform IV and V presence and virtually resulted in LPS composed of only glycoform I. Such a switch to exclusive synthesis of glycoform I was observed only in ⌬waaZ. Furthermore, ⌬(relA spoT rseA) mutants did not incorporate L-Ara4N on lipid A, although P-EtN could be found. These results for the first time implicate previously unknown overlap of control of RpoE/ppGpp with lipid A modification system under phosphate starvation conditions as well as inner core alterations specific to the incorporation of the third Kdo.

RpoE-dependent and -independent Alterations in E. coli LPS
Although lipid A modifications by P-EtN and L-Ara4N are known to confer resistance to cationic peptides, the significance of the switch from glycoform I to glycoform IV and V is not known other than synthesis of LPS lacking the O-polysaccharide (smooth to rough variation). Because we constructed several derivatives, which synthesize either only glycoform I (⌬waaZ, ⌬(rseA relA spoT)) or primarily glycoform IV and V (⌬rseA, ⌬waaR), we examined total OMP content specifically of OmpA, OmpF, and OmpC. Consistent with the OMP defect of rseA mutants, they exhibited reduced levels of all such OMP. However, waaR mutants, despite their similar LPS composition compared with the rseA mutants, showed no defect in OMP content. These results were further supported by the lack of any induction of the RpoE pathway in waaR mutants. We also show that ⌬(relA spoT) mutants have reduced OMP content despite normal LPS, but they exhibit a drastic reduction in OMP content and a total dramatic switch to glycoform I in ⌬(relA spoT rseA) derivative. However, it cannot be due to this LPS switch, because ⌬waaZ have the wild-type level of OMP despite the synthesis of only glycoform I. Thus, no direct relationship is obvious between switches in glycoforms and OMP content, although both are controlled via RpoE.
Finally, we found waaZ is required for growth of ⌬(lpxL lpxM lpxP) mutants synthesizing tetraacylated lipid A on rich medium at 21-23°C. A deletion derivative lacking all late acyltransferases ⌬(lpxL lpxM lpxP) cannot grow on rich medium at 30°C (4,36). We show that such a derivative can grow at lower temperatures but requires waaZ for growth in rich medium at 21-23°C. However, ⌬(lpxL lpxM lpxP waaZ) can be constructed at such lower temperatures but only on minimal medium. These results assume significance, because ⌬(lpxL lpxM lpxP) mutants synthesize predominantly glycoform V in lipid A modification conditions. Thus, waaZ induction confers a subtle growth advantage to ⌬(lpxL lpxM lpxP) derivatives and may explain its conservation in E. coli K-12.
In this work, we also found that LPS of E. coli K-12 and E. coli B contains a new modification resulting in an addition of 96 mass units in the core region. Our initial results show that this modification arises because of the incorporation of a HexA residue (176.01 Da) on HepIII with an accompanying loss of phosphate residues on HepII. This explains the observed additional 96 mass units presence. HepII is known to require WaaY for phosphorylation, and WaaQ is thought to be a HepIII transferase (38,39). The rationale for HexA addition on HepIII was supported by the lack of this novel modification in ⌬waaQ mutants and its enhanced presence in ⌬waaY derivatives. 3 We have identified a new gene responsible for this modification and is currently under further intensive investigation.
In conclusion, we show induction of PhoB/R, BasS/R, RpoE, and ppGpp jointly control structural alterations of both lipid A and core oligosaccharide composition. Out of these roles of RpoE induction, requirement of ppGpp and modulation of WaaR levels by rybB noncoding RNA in controlling LPS composition are addressed and described for the first time. Because ⌬waaR mutants do not exhibit elevated RpoE induction, have normal OMP content, and yet can have LPS with three Kdo residues (glycoform IV) even without lipid A modifications argue for major control in glycoform IV synthesis at WaaR levels. We also showed that the synthesis of glycoform V results from P-EtN presence on the second Kdo, which leads to a switch of Rha addition to the third Kdo.