YhdP, TamB, and YdbH Are Redundant but Essential for Growth and Lipid Homeostasis of the Gram-Negative Outer Membrane

ABSTRACT The bacterial cell envelope is the first line of defense and point of contact with the environment and other organisms. Envelope biogenesis is therefore crucial for the survival and physiology of bacteria and is often targeted by antimicrobials. Gram-negative bacteria have a multilayered envelope delimited by an inner and outer membrane (IM and OM, respectively). The OM is a barrier against many antimicrobials because of its asymmetric lipid structure, with phospholipids composing the inner leaflet and lipopolysaccharides (LPS) the outer leaflet. Since lipid synthesis occurs at the IM, phospholipids and LPS are transported across the cell envelope and asymmetrically assembled at the OM during growth. How phospholipids are transported to the OM remains unknown. Recently, the Escherichia coli protein YhdP has been proposed to participate in this process through an unknown mechanism. YhdP belongs to the AsmA-like clan and contains domains homologous to those found in lipid transporters. Here, we used genetics to investigate the six members of the AsmA-like clan of proteins in E. coli. Our data show that YhdP and its paralogs TamB and YdbH are redundant, but not equivalent, in performing an essential function in the cell envelope. Among the AsmA-like paralogs, only the combined loss of YhdP, TamB, and YdbH is lethal, and any of these three proteins is sufficient for growth. We also show that these proteins are required for OM lipid homeostasis and propose that they are the long-sought-after phospholipid transporters that are required for OM biogenesis.

unknown function (25). However, the effect of YhdP on anterograde phospholipid transport in mlaA* cells is independent of ECA CYC (18). How YhdP modulates the effect of ECA CYC on OM permeability and phospholipid transport in mlaA* strains remains unknown, as is whether YhdP affects phospholipid transport in wild-type cells.
It is logical to assume that anterograde phospholipid transport to the OM is essential for growth, since phospholipids are the lipid constituent of the inner leaflet of the OM and the OM is essential for survival. However, YhdP is not essential (24). In addition, the loss of YhdP slows down but does not abolish anterograde phospholipid transport in mlaA* cells (18). One interpretation of these facts is that YhdP only plays an accessory role in anterograde phospholipid transport. However, its predicted architecture suggests that YhdP directly transports phospholipids. YhdP is a 1,266-residue protein that belongs to the AsmA-like CL0401 protein clan (26). AsmA-like proteins contain a predicted IM-anchoring N-terminal a-helix and a large periplasmic domain that includes a chorein-N domain and the AsmA-like domain (Fig. 1A) (27,28). Chorein-N domains are present in eukaryotic intermembrane lipid transporters, while AsmAlike domains are composed of various numbers of b-taco domains whose structure resembles that of the domains forming the periplasmic Lpt bridge, which shields the fatty acyl chains of LPS molecules as they travel through the periplasm en route to the OM (Fig. 1A) (29,30). If YhdP indeed transported phospholipids, it would have to be functionally redundant with other transporters, since it is not essential for viability. Interestingly, E. coli encodes YhdP and five additional AsmA-like proteins (TamB [formerly YtfN], YdbH, AsmA, YicH, and YhjG [ Fig. 1; see also Fig. S1 in the supplemental material]), and functional redundancy has been suggested as an explanation for why factors involved in IM-to-OM phospholipid transport have remained elusive (1,31). These connections motivated the present study investigating the possible functional redundancy among AsmA-like proteins in E. coli. Using a genetic approach, we found that TamB, YhdP, and YdbH are redundant in performing a function that is essential in envelope biogenesis. We show that the combined loss of TamB, YhdP, and YdbH is lethal, and that mutants lacking five AsmA-like factors grow if they retain either TamB, YhdP, or YdbH. Based on our data and the fact that these proteins are predicted to have domains and overall structures similar to those present in lipid transporters, we propose that TamB, YhdP, and YdbH are the long-sought-after transporters of phospholipids between the IM and OM.

RESULTS
TamB and YhdP are functionally redundant in maintaining the barrier function of the OM. The six paralogs of AsmA-like proteins in E. coli contain a chorein-N domain that is present in some intermembrane eukaryotic transporters (see Fig. S1 in the supplemental material). In addition, the recently released AlphaFold Protein Structure Database predicts that they share an overall structure with Atg2, a protein that belongs to a new family of lipid transporters at membrane-contact sites between eukaryotic organelles ( Fig. S1) (32)(33)(34)(35). Based on these relationships and the fact that YhdP has been shown to affect phospholipid transport from the IM to the OM (18), we investigated a possible role for the AsmA-like protein family in OM biogenesis in E. coli. We generated mutants lacking one or more of the six paralogs and tested for defects in growth and OM permeability, as defects in OM structure affects these phenotypes (2,4). We first characterized the six single mutants. None of them exhibited growth defects in lysogeny broth (LB) at 37°C. To probe OM permeability, we monitored their resistance to bile salts and various antibiotics whose entry is limited by the OM (36). We did not detect OM permeability defects in the DydbH, DyhjG, and DyicH single mutants, but, as previously reported in various Gram-negative bacteria, the DasmA, DtamB, and DyhdP mutants showed a slight increase in OM permeability ( Fig. 1B and C, Table S1A) (24,25,(37)(38)(39).
Next, we constructed mutants carrying multiple deletion alleles to determine if AsmA-like proteins are functionally redundant or function in different or the same pathways. We expected that loss of redundant factors would result in a phenotype greater than the one expected from adding the phenotype of the single mutants. In contrast, if the effect of combining deletions was additive, it would indicate that the respective factors function in independent pathways. A lack of phenotypic synergy or additivity would be expected if factors functioned in the same pathway. Lastly, a mutant allele might suppress phenotypes conferred by other alleles, revealing that the factors are somehow functionally related. We therefore constructed the 15 double mutants lacking AsmA-like proteins and analyzed the barrier function of their OM (Fig. 1B). We observed that adding either DydbH, DyhjG, or DyicH to any single mutant did not alter their respective OM permeability phenotypes (Table S1A). Since, on their own, these three alleles do not confer any phenotypes, these analyses did not further clarify the role of YdbH, YhjG, and YicH. In contrast, the three double mutants resulting from combining DasmA, DtamB, and DyhdP revealed key information. The DasmA allele behaved in an additive fashion with DtamB and DyhdP (Fig. 1C, Table S1A), indicating that AsmA performs a function that is different from those of TamB and YhdP. Previously, AsmA had been implicated in the folding of defective integral b-barrel OM proteins (OMPs) (37,40,41). Moreover, we uncovered negative synergistic effects between DtamB and DyhdP. Unlike the mildly defective single mutants, the DyhdP DtamB mutant showed severe OM permeability defects to bile salts and various antibiotics (Fig. 1C). Thus, our characterization suggests that while AsmA works independently, YhdP and TamB function in redundant fashion to maintain the barrier quality of the OM.
TamA is required for TamB's function. TamB forms a complex with the OM b-barrel protein TamA in E. coli (Fig. 1A) (42). TamA is a homolog of BamA, the b-barrel component of the BAM complex that assembles b-barrel proteins in the OM (43,44). In some organisms that lack TamA, TamB has been shown to interact with BamA instead (31,38). The TAM complex has been proposed to function in the assembly of a subset of OM b-barrel proteins, including autotransporters and fimbrial ushers (29,42,45,46). However, in some organisms, autotransporter biogenesis is not affected by the loss of TamB (47)(48)(49), and while biogenesis of autotransporters and ushers requires BAM, they are partially affected in tamAB mutants (46,50). In addition, some bacteria lack autotransporters and fimbrial ushers but possess TamB (and BamA but not TamA) (31,51). Therefore, the function of TamB remains unclear.
To determine if TamA is required for TamB's function, we analyzed the effects of the DtamA allele and found that it causes the same phenotypes as DtamB by itself and in combination with DyhdP (Table S1B). Furthermore, plasmid-encoded TamB complements DtamB DyhdP but not DtamA DyhdP with respect to OM permeability defects (Fig. S2A). We therefore conclude that the OM protein TamA is required for TamB's function in maintaining OM permeability.
The combined loss of TamB and YhdP causes pleiotropic defects in the cell envelope. In addition to having OM permeability defects, we observed that cells lacking YhdP and TamB display other phenotypes indicative of severe defects in cell envelope biogenesis. While its parental single mutants grow similarly to the wild type at 37°C in LB, the DyhdP DtamB mutant exhibits increased lysis that becomes more evident upon entry into stationary phase ( Fig. 2A). Indeed, analysis of concentrated cellfree filtrates from supernatants of overnight cultures showed higher protein content in samples from the DtamB DyhdP mutant than the wild-type strain (Fig. 2B). In contrast, samples from the DyhdP but not the DtamB single mutant only showed a slight increase in protein content. Furthermore, phase-contrast microscopy of exponentially growing cells showed that the increase in lysis in the DtamB DyhdP mutant is not limited to stationary-phase cultures. We could not detect lysis in wild-type and single-mutant cultures, but ghost (i.e., lysed) cells and membrane blebs were readily observed in DtamB DyhdP cultures ( Fig. 2C and D). Phase-contrast microscopy also revealed that while the overall morphology (length and width) of wild-type and single-mutant cells is similar, DyhdP DtamB cells are more irregularly shaped, tending to be shorter and wider than the wild type and single-mutant parents ( Fig. 2C to E).
We also observed that, on LB agar, the DyhdP DtamB mutant forms mucoid colonies, unlike the parent and wild-type strains (Fig. S3A). Mucoidy results when production of The number of cells undergoing constriction after septation (with shape of the number 8) was observed with phase-contrast microscopy, and images were processed with ObjectJ (85) to measure cell length and width of each daughter cell undergoing constriction during exponential growth in LB at 37°C. colanic acid surface capsule is upregulated by the Rcs envelope stress response via the RcsB response regulator in reaction to various signals, including cell surface stress (52). Capsule production involves the synthesis of intermediates that are linked to undecaprenyl-phosphate, an isoprenoid lipid carrier that is also required for the synthesis of the peptidoglycan cell wall (53). Since the cell wall determines cell shape and protects cells from osmotic lysis, we considered the possibility that lysis and altered cell shape in DyhdP DtamB cells could result from titration of undecaprenyl-phosphate from peptidoglycan synthesis by the upregulation of capsule production, as it has been reported when there is an imbalance in the production of polysaccharides whose synthesis rely on undecaprenyl-phosphate (54,55). To test this possibility, we deleted rcsB in the DyhdP DtamB mutant. As expected, DrcsB eliminated mucoidy in the DyhdP DtamB mutant; however, the resulting triple mutant still shows increased lysis, severe OM permeability defects, and altered shape (Fig. S3). In fact, DrcsB worsens the growth of DyhdP DtamB cells. This detrimental effect appears to be solely caused by the effect that RcsB has on capsule production, since only deleting a gene responsible for capsule synthesis (wcaJ) conferred the same phenotype as DrcsB (Fig. S3). These results suggest that the Rcs system senses surface stress in DyhdP DtamB cells; in agreement, we found that their mucoidy requires the OM lipoprotein sensor RcsF since introduction of DrcsF abolishes their mucoid phenotype (52). Together, our results indicate that the loss of TamB and YhdP severely compromises the integrity of the cell envelope and that activation of the Rcs envelope-stress response confers some protection by inducing synthesis of a surface polysaccharide capsule.
Lysis of the DtamB DyhdP mutant results from defects in OM lipid homeostasis. We have described that cells lacking TamB and YhdP exhibit phenotypes characteristic of OM biogenesis defects: increased permeability to bile salts and antibiotics, production of membrane blebs, and activation of the Rcs response. Not surprisingly, we found that the s E stress response is also activated in DyhdP DtamB cells (Fig. S4A). The s E system regulates synthesis of OM components and OM biogenesis factors in response to stresses, such as the misfolding of OMPs and off-pathway intermediates in LPS transport (56,57). The OM contains two major types of proteins, OMPs, which fold into b-barrel structures, and lipoproteins, which are anchored to the OM via an N-terminal anchor (58,59). OMP assembly is catalyzed by the b-barrel assembly machine (BAM), which is composed of the b-barrel protein BamA and the BamB-E lipoproteins (43,60). Given that both BamA and BamD are essential for BAM function, OMP assembly itself also requires proper biogenesis of OMPs and OM lipoproteins. We therefore analyzed levels and folding of the major OMPs OmpA and OmpC to monitor BAM function and thereby the biogenesis of OMPs and OM lipoproteins. We could not detect misfolding of OmpA and OmpC in cells lacking YhdP and/or TamB (Fig. S4B), suggesting that these proteins do not function in either OMP or lipoprotein biogenesis. We did observe that the DyhdP DtamB mutant, unlike its mutant parents, produces higher levels of LPS than the wild type, which might be the cause for the activation of the s E stress response if some of the molecules go off pathway (Fig. S4C).
Elevated LPS levels and increased lysis during stationary phase resemble phenotypes previously reported for the mlaA* mutant (19). As stated earlier, the dominantnegative MlaA* variant causes the mislocalization of phospholipids to the cell surface, which activates PldA, leading to an increase in LPS synthesis and a decrease in phospholipid synthesis (Fig. 3A) (23). The resulting PldA-dependent imbalance in OM lipid synthesis leads to loss of OM material through vesiculation and eventually results in lysis in stationary phase because cells cannot synthesize enough lipids to overcome the loss of OM material (19,23). Here, we sought to investigate if the stationary-phaseinduced lysis in DyhdP DtamB mutants was related to defects in the OM. Loss of OM material in mlaA* cells can be partially suppressed by the addition of Mg 21 (19). Similarly, adding 1 mM MgCl 2 to the growth medium partially reduces lysis in DyhdP DtamB cultures (Fig. S4D), suggesting that lysis in the DyhdP DtamB mutant results from the loss of OM material. If so, we expected that DyhdP DtamB cells would be hypersensitive to EDTA, since this chelator removes LPS from the cell surface and causes phospholipids to translocate to the outer leaflet of the OM (61). Indeed, the DyhdP DtamB mutant is hypersensitive to EDTA (MIC of 0.195 mM compared to 25 mM in the wild-type and single-mutant parents). We also found that adding subinhibitory amounts of EDTA (0.5 mM) to cultures of wild-type and DyhdP and DtamB single-mutant strains triggers lysis in stationary phase (Fig. S4D), which is more pronounced in the single mutants than in the wild type. We next explored if MlaA and PldA play a role in the lysis phenotype of DyhdP DtamB cells. While deleting mlaA increased lysis, deleting pldA decreased the rate and extent of lysis induced during stationary phase in DyhdP DtamB cells ( Fig. 3B and C).
It seemed paradoxical that removing MlaA would have the opposite effect of removing PldA, given that the individual loss of each factor increases the amount of phospholipids at the cell surface. We reasoned that the increase in lysis of DyhdP  DtamB cells caused by the loss of MlaA might be dependent on PldA. Specifically, DmlaA should increase the amount of PldA's substrate (i.e., mislocalized phospholipids), which would upregulate LPS levels while decreasing phospholipid synthesis (23). Indeed, DmlaA further increases the already elevated LPS levels in a DyhdP DtamB mutant (Fig. 3D). We therefore built a DmlaA DpldA DyhdP DtamB mutant. Strikingly, the combined loss of DmlaA and DpldA fully suppresses lysis of DyhdP DtamB cells, yielding a wild-type growth pattern (Fig. 3B). As expected, the DmlaA DpldA DyhdP DtamB mutant also has lower LPS levels than the DyhdP DtamB DmlaA mutant. Nevertheless, the quadruple mutant remains sensitive to bile salts and antibiotics (Fig. 3E, Table S1C).
Together, our results suggest that DyhdP DtamB cells have OM structural defects that somehow lead to increased levels of phospholipids in their cell surface, which results in increased sensitivity to bile salts and antibiotics and activation of the Mla and PldA pathways. In DyhdP DtamB cells, both MlaA and PldA contribute to lysis by removing these mislocalized phospholipids. PldA's action is particularly detrimental to DyhdP DtamB cells because it induces LPS synthesis, causing a further imbalance in OM lipid composition (23). In agreement, adding fatty acids (oleic acid) to the medium exacerbates the lysis phenotype in DyhdP DtamB cultures, although a detergent-like effect could also contribute to or cause lysis in these mutant cells (Fig. S4D). If only MlaA is removed, DyhdP DtamB cells lyse even more because of PldA's upregulation of LPS levels; however, when both PldA and MlaA are removed, lysis is abolished and cells can grow like the wild type. Thus, these results suggest that TamB and YhdP are required for OM lipid homeostasis.
The AsmA-like protein family is essential for viability in E. coli. The main lipid constituents of the OM are LPS and phospholipids. Notably, the Gram-negative Borrelia does not produce LPS, yet it encodes tamB, which appears to be essential (38). In addition, one of us (N. Ruiz) previously took advantage of the reduced size of two Gram-negative endosymbionts carrying ,600 genes to search for envelope biogenesis factors and found tamB present in both despite one of these bacteria not having LPS biogenesis genes (51). Here, we expanded this search to five additional gammaproteobacterial endosymbionts and to YhdP. We still found no correlation between the presence of TamB and YhdP and that of LPS biogenesis proteins (Table S2), ruling out a direct role for TamB and YhdP in LPS biogenesis. The fact that these endosymbionts encode either tamB or yhdP even after having undergone massive gene loss strongly suggests that these proteins perform a crucial function in the physiology of these bacteria. Given that YhdP has been implicated in IM-to-OM transport of phospholipids (18), and its predicted protein architecture supports this proposed role (Fig. S1) (27,30), the simplest explanation of the evidence presented so far here and by others is that YhdP and TamB mediate phospholipid transport between the IM and OM. However, anterograde phospholipid transport is expected to be essential for building the OM, but even the combined loss of YhdP and TamB is not lethal. We therefore explored if the functional redundancy between TamB and YhdP extends to additional AsmA-like paralogs.
We constructed mutants lacking more than two AsmA-like proteins. Although our data (Fig. 1) suggest that AsmA functions independently of YhdP and TamB, we still included AsmA in our studies. We first combined DyhjG, DyicH, and DydbH because they do not confer defects individually or in pairs and then introduced DasmA, DyhdP, and DtamB, which on their own confer OM permeability defects. After constructing several mutants, we assayed their OM permeability and compared them to single and double mutants (Fig. 4A, Table S1A) and found that we could not generate a mutant lacking all six AsmA-like proteins (Fig. 4A), suggesting that the AsmA-like family of proteins are essential for viability in E. coli.
TamB, YhdP, and YdbH are redundant proteins essential for viability in E. coli. Since our genetic analyses indicated that TamB/YhdP and AsmA function in different pathways, we hypothesized that the synthetic lethality observed when attempting to construct a strain lacking all AsmA-like proteins could result from redundancy between TamB/YhdP and YdbH, YicH, and/or YhjG. We found that while the DtamB DyhdP DyicH and DtamB DyhdP DyhjG triple mutants were viable and phenotypically indistinguishable  Table S1A). The loss of all AsmA-like factors or the combined loss of TamB, YhdP, and YdbH is lethal. Data represent the average and standard deviation from three biological replicates. If not shown, standard deviation equals zero. (B) Chromosomal yhdP locus in YhdP depletion strains. Sequence encoding bla-araC-P BAD was inserted upstream of yhdP to decouple it from its native promoter (P yhdP ). The resulting recombinant locus has been engineered to have yhdP transcription under the arabinose-dependent activator AraC. (C) Cultures grown in LB with arabinose overnight at 37°C were diluted 1:10 from left to right and then stamped with a pin replicator onto LB agar containing or lacking arabinose. Growth of the DydbH DtamB P BAD ::yhdP mutant is arabinose dependent. (D) Depletion of YhdP in a quintuple (D5) or a DydbH DtamB double mutant is lethal. Overnight cultures of NR5921 (MG1655 DtamB::frt DydbH::kan yhdPX-1::bla araC P BAD ) and NR5850 (MG1655 DyhjG::frt DyicH::frt DydbH::frt DasmA::frt DtamB::kan yhdPX-1::bla araC P BAD ) were grown in LB with arabinose at 37°C. After a 1:5,000 dilution in LB with arabinose (1YhdP) or fucose (YhdP depletion), growth at 37°C was measured by monitoring the OD 600 . Data represent the average and standard deviation from three biological replicates. (E) Phase-contrast microscopy (100Â objective) of strain NR5921 grown in the presence of arabinose or fucose. White scale bar represents 5 mm. Data are representative of at least three independent experiments. from the DtamB DyhdP mutant, we could not build the DtamB DyhdP DydbH triple mutant (Fig. 4A), suggesting these three alleles are synthetic lethal.
We next constructed a YhdP depletion strain to better test if DtamB, DyhdP, and DydbH are synthetically lethal. By altering its promoter region, we placed yhdP transcription under the control of the arabinose-inducible promoter P BAD (Fig. 4B) (62). We confirmed that expression of yhdP is controlled by arabinose in strains carrying P BAD ::yhdP by showing that a DtamB P BAD ::yhdP strain exhibits OM permeability defects similar to those of a DtamB DyhdP mutant when grown in the presence of the anti-inducer D-fucose (an Larabinose analog) but not in the presence of the inducer arabinose ( Fig. S5A and B). Next, we constructed a DtamB DydbH P BAD ::yhdP strain in the presence of arabinose and determined that its growth is arabinose dependent (Fig. 4C and D). Thus, DtamB, DyhdP, and DydbH are indeed synthetically lethal. We also built a YhdP depletion strain lacking the genes encoding the other five AsmA-like proteins and determined that it behaves similarly to the DtamB DydbH P BAD ::yhdP strain (Fig. 4D). Further characterization of the DtamB DydbH P BAD ::yhdP strain showed that growth in the presence of D-fucose does not cause defects in the assembly of the major b-barrel protein OmpA but induces production of DegP, which is controlled by the s E stress response (56), and LPS (Fig. S5). Phasecontrast microscopy revealed that, in the presence of arabinose, the DtamB DydbH P BAD :: yhdP strain cells appear like wild-type rods, but, in the presence of fucose, the depletion strain undergoes lysis and exhibits morphological defects (Fig. 4E).
Given that the combined loss of mlaA and pldA suppresses lysis in DtamB DyhdP cells, we tested if it could also suppress the dependence on arabinose for growth of the DtamB DydbH P BAD ::yhdP strain. We found that it could not (Fig. S5F). However, we showed that TamA is essential in cells lacking YhdP and YdbH, confirming our earlier conclusion that TamA is required for TamB function (Fig. S2B). Altogether, our data demonstrate that TamB, YhdP, and YdbH are redundant in performing a function that is essential for growth of E. coli. Nevertheless, the difference in phenotypes observed in the three DtamB, DyhdP, and DydbH single mutants, and their corresponding double and triple mutants, also suggest that despite being functionally redundant, these proteins are not equivalent (Table S1A).
Gain-of-function substitutions in YdbH suppress defects in the DtamB DyhdP mutant. We next searched for suppressor mutations that could provide information about the function of TamB, YhdP, and YdbH using both reverse and forward genetics. We focused on isolating suppressors of the DtamB DyhdP double mutant. For the reverse-genetic approach, we tested whether the loss of the enterobacterial common antigen (ECA) caused by deleting wecA could suppress the increase in OM permeability and/or lysis, since, as stated earlier, it suppresses OM permeability defects in a DyhdP mutant (25). We did not observe changes in OM permeability in either the wild-type or the DtamB mutant, but, as previously reported, observed that DwecA suppresses the sensitivity of a DyhdP mutant to vancomycin (Table S1D) (25). However, DwecA does not suppress the OM permeability defects in the DtamB DyhdP mutant and, in fact, severely compromises its growth (Table S1D and data not shown). Thus, the ability of the loss of ECA to suppress the OM permeability defects caused by DyhdP requires TamB, and losing ECA has a detrimental effect on the fitness of DtamB DyhdP cells.
We next employed an unbiased selection for spontaneous mutants that restore growth of the DtamB DyhdP mutant in the presence of bile salts (i.e., MacConkey agar). After mapping suppressor mutations by genetic linkage, we identified four different missense mutations in ydbH that change conserved residues in the periplasmic region of YdbH, T303I, G305W, G308C, and G456V (Fig. 5). We found that these missense ydbH alleles also decrease sensitivity to hydrophobic antibiotics, but not vancomycin (Fig. 5B). Thus, these suppressor mutations do not fully restore wild-type phenotypes. Given that the loss of YdbH is lethal in the DtamB DyhdP mutant, we expected these alleles to be gain-of-function alleles. In agreement with these alleles being gain of function and not loss of function, introducing a plasmid carrying the wild-type ydbH locus also restores growth of the DtamB DyhdP mutant on MacConkey agar. Thus, altering YdbH function through mutations or increasing levels of wild-type YdbH function suppresses sensitivity to bile salts (Fig. 5A). Further characterization will be needed to understand how these changes suppress defects in DtamB DyhdP cells, but they support our conclusion that TamB, YhdP, and YdbH are functionally redundant but not equivalent.

DISCUSSION
The AsmA-like family of proteins has been recognized as conserved and specific to didermic Gram-negative bacteria, but determining its function has been difficult. The fact that many organisms, including E. coli, the model bacterium for studying the Gram-negative envelope, encode up to six AsmA-like paralogs (TamB, YhdP, YdbH, AsmA, YicH, and YhjG) has limited progress for our understanding of this protein family  Table shows Consurf (86) conservation scores (Cons. score) of residues in the regions where suppressing changes are localized in YdbH. Conservation scale ranges from 1 (most variable) to 9 (most conserved). Although these residues are conserved among YdbH orthologs, they are not conserved at the same positions in YhdP and TamB. (D) Model structure (see Fig. S1) of YdbH showing residues altered by suppressor mutations as green spheres. The cluster containing residues 303, 305, and 308 is located in a loop, while the side chain of residue 456 faces the interior of the tubular AsmA-like domain. (31). Indeed, although previous studies have linked AsmA, TamB, and YhdP to envelope biogenesis in various bacteria, their role remains mostly uncharacterized (18, 24, 25, 28, 37-42, 45, 47-49). To investigate whether redundancy was occluding their function in E. coli, we used reverse genetics to generate and characterize mutants lacking one or several AsmA-like proteins. Our work demonstrates that TamB, YhdP, and YdbH are redundant in performing a function that is critical for OM lipid homeostasis and essential for growth.
Previously, YhdP had been implicated in diffusive anterograde phospholipid transport between the IM and OM in mlaA* E. coli mutants, which have altered OM lipid structure (18). However, its role in anterograde phospholipid transport in wild-type cells remained unclear, especially because the loss of YhdP results in mild phenotypes but anterograde phospholipid transport is presumed to be essential for building the OM and thereby growth (24,25). Our work addresses these issues by showing that the combined loss of YhdP, TamB, and YdbH is lethal in an otherwise wild-type strain because these proteins are crucial in maintaining lipid homeostasis at the OM (Fig. 3  and 4). This essential role for growth is also underscored by the fact that at least one of these proteins is conserved in Gram-negative endosymbionts that encode fewer than 600 proteins in their reduced-size genomes (Table S2) (51). Importantly, TamB, YhdP, and YdbH are homologous to the eukaryotic proteins Vps13 and Atg2, which were recently shown to constitute a new family of lipid transporters at membrane-contact sites between organelles (33)(34)(35)63), and the less-characterized TIC236 and Mdm31/32 homologs, which are also needed for proper biogenesis of chloroplasts and mitochondria, respectively (27,30,(33)(34)(35)(63)(64)(65). Given this body of evidence and that anterograde IM-to-OM phospholipid transport is the only essential process required for OM biogenesis that is yet to be linked to the essentiality of any protein(s) (1), we propose that TamB, YhdP, and YdbH transport phospholipids between the IM and OM (Fig. 6).
Transport of phospholipids between the IM and OM has been proposed to occur by diffusion mediated either by proteinaceous bridge-like structures or by lipid bilayer assembly. The Lpt system (on the right) transports newly synthesized LPS molecules from the IM to the outer leaflet of the OM. We propose that TamB, YhdP, and YdbH transport phospholipids between the IM and OM. In this model, TamB, YhdP, and YdbH physically bridge the IM and OM by anchoring to the IM through their Nterminal a-helix (black segment) and likely to the OM via interactions with partners (green oval). The predicted structure of the large periplasmic region of these proteins would form a structure similar to the Lpt bridge and provide protection to the hydrophobic fatty acid chains of phospholipids as they travel through the periplasm. LPS transport is unidirectional and powered by ATP, while bidirectional phospholipid transport would be driven by diffusion through TamB, YhdP, and YdbH.
Redundant but Essential Envelope-Biogenesis Factors ® structures fusing the IM and OM (1,9,13,18,19,66,67). In the latter model, TamB, YhdP, and YdbH could be responsible for building or maintaining the proposed membrane bridges or for preventing the flow and mislocalization of proteins or lipid-linked molecules between the IM and OM. However, based on the predicted architecture of their homologs, Vps13 and Atg2, and the structural models recently released by AlphaFold (Fig. S1) (32), we favor a model in which TamB, YhdP, and YdbH are proteinaceous bridges between the IM and OM that directly facilitate the flow of phospholipids (Fig. 6). Vps13 and Atg2 are large intermembrane bridge proteins that are anchored to the membrane of one organelle via an N-terminal a-helix; this anchor is then followed by a chorein-N domain, a large region rich in b-strands, and variable domains that are thought to mediate interactions with different factors at the membrane of another organelle (30,33,63,64,68). The large region rich in b-strands bridging the two membranes is modeled to fold into a structure that resembles the periplasmic bridge formed by the Lpt proteins that transports the glycolipid LPS from the IM to the OM (11,69,70) (Fig. 6). Specifically, they are thought to form an elongated tube-like structure with a lateral opening along the long axis that leads to a hydrophobic groove that interacts with phospholipids by protecting the hydrophobic fatty acid tails of multiple lipids as they travel through aqueous compartments from one membrane to another (33,68). TamB, YhdP, and YdbH also have an N-terminal a-helix that is followed by a chorein-N domain and a large portion rich in b-strands (.700 amino acids [aa]) large enough to cross the periplasm (27). Their structure is predicted to resemble that of Atg2 (Fig. S1). The eukaryotic homologs Vps13 and Atg2 contact two membranes from different organelles and associate with various partners through their C-terminal domains (30,64). To date, no partners have been identified for YhdP and YdbH, but TamB interacts, depending on the organism, with the OMP TamA or BamA (31,38,42). In agreement, our data show that TamA is required for TamB's function (Table S1B and Fig. S2). Based on the TamB-TamA interaction and that of their eukaryotic homologs with their partners, we propose that YhdP and YdbH are also likely to require an OM anchor (Fig. 6). We therefore suggest that TamB, YhdP, and YdbH constitute the membrane-contact sites proposed decades ago to mediate phospholipid transport between the IM and OM (66,67,71). In this model, TamB, YhdP, and YdbH would provide a bridge-like structure resembling that of the Lpt bridge that protects phospholipids as they travel across the periplasm (Fig. 6). However, unlike the unidirectional LPS transport mediated by the Lpt system, TamB, YhdP, and YdbH would support the previously reported bidirectional diffusive flow of phospholipids ( Fig. 6) (9,18,19,66,67).
Functional redundancy between TamB, YhdP, and YdbH would also explain why the identification of the mechanism for anterograde phospholipid transport has remained elusive despite the great progress made in the identification and characterization of factors involved in the biogenesis of other envelope components (1,13). Nevertheless, even though our data show that TamB, YhdP, and YdbH are redundant and any of them is sufficient to support growth of E. coli, these proteins are not functionally equivalent. Instead, our phenotypic analysis of the respective single and double mutants has revealed a functional hierarchy in the cell envelope that correlates with protein size (Fig. 1): TamB and YhdP play a more similar and important role than YdbH. The nature of this specialization is unknown, but possible explanations include differences in expression, cargo preference, and/or cellular localization as it occurs among the four human Vps13 paralogs (64). It is also possible that the specialization of these paralogs results from additional different functions that these proteins have evolved to perform. For example, TamB has been shown to affect the transport of a nonconserved subset of b-barrel proteins from the IM to the OM (28,29,42,46). Whether this is an additional function of TamB or a secondary effect of the loss of its primary role in OM lipid biogenesis needs to be determined.
Lastly, we note that growth of the OM lipid bilayer requires the balanced assembly of phospholipids at the inner leaflet and of LPS and other lipid-linked oligosaccharides at the outer leaflet of the OM (1). How the synthesis, transport, and assembly of these lipid components are coordinated is poorly understood, but studies like ours and those previously done on mlaA* show that imbalance leads to severe defects in asymmetry and even death when the growth of the two leaflets of the OM is out of balance (18,19). In mlaA* mutants, growth of the inner leaflet of the OM is compromised because of the aberrant translocation of phospholipids to the cell surface, which, in turn, causes the activation of PldA and downstream upregulation of LPS synthesis. These two events increase the synthesis of the main lipid component of the outer leaflet (LPS) and decrease the presence and synthesis, respectively, of the main lipid component of the inner leaflet (phospholipids). In our study, we found that a DtamB DyhdP mutant undergoes lysis that is suppressed by the loss of MlaA and PldA (Fig. 3). We propose that the compromised flow of phospholipids to the OM in the DtamB DyhdP mutant disrupts OM lipid homeostasis, somehow also leading to the activation of PldA and the subsequent upregulation of LPS synthesis. Upregulating the production of the main component of the outer leaflet of the OM (LPS) likely harms the cell through a futile cycle in which the LPS-driven growth of the outer leaflet of the OM would increase the demand for an already compromised transport of phospholipids, further driving the disruption of OM lipid homeostasis. How phospholipids are translocated across the OM in wild-type and DtamB DyhdP cells also remains unknown. Recent studies are just beginning to reveal how OM lipid asymmetry and LPS transport may be sensed by PldA and YejM to coordinate the synthesis of phospholipids and LPS through the regulated degradation of the LPS synthesis enzyme LpxC by FtsH/LepB (72)(73)(74)(75)(76)(77). Our work calls for investigating how TamB, YhdP, and YdbH may be integrated in these systems, as well as the role that other surface components such as ECA and capsule may play.

MATERIALS AND METHODS
Bacterial strains. Strains were derived from wild-type strain MG1655 (78) and are listed in Table S3A in the supplemental material. Deletion alleles were derived from the Keio collection (79) and introduced into the appropriate strains by generalized P1vir transduction and selection for kanamycin resistance (80). When necessary, kanamycin resistance cassettes were excised by the Flp recombinase as previously described (81). Unless indicated, strains were grown in LB medium at 37°C either in liquid cultures with aeration or on solid medium containing 1.5% agar (80). When necessary, growth media were supplemented with ampicillin (25 or 125 mg/ml), L-arabinose (0.2%, wt/vol), bacitracin (100 mg/ml), chloramphenicol (20 mg/ml), erythromycin (25 mg/ml), vancomycin (25 or 75 mg/ml), kanamycin (25 mg/ml), tetracycline (25 mg/ml), MgCl 2 (1 mM), EDTA (0.5 mM), oleic acid (10 mg/ml). MacConkey agar was commercially available from BD (catalog no. BD281810). Except for YhdP depletion strains, to monitor growth of strains, overnight cultures were diluted 1:1,000 in LB with the appropriate additives. Diluted cultures were delivered (200 ml) into a 96-well plate that was incubated at 37°C with continuous double orbital shaking in an Epoch2 BioTek reader. Growth was monitored every 30 min by measuring absorbance at 600 nm (OD 600 ).
YhdP depletion. A YhdP depletion strain was constructed by recombineering as previously described (51). Briefly, primers 5YhdP_Pbad and YhdP_Pbad (Table S3B) were used to amplify the bla-araC-P BAD region of pKD46 (82). The PCR product was used for recombineering to insert bla-araC-P BAD between the 21 and 11 positions of the yhdP gene in the chromosome of the recombineering strain DY378 (83). Recombinants were selected at 30°C on medium containing 25 mg/ml ampicillin and confirmed by PCR analysis. The yhdP X(21:: bla araC-P BAD ) allele was introduced into various strains by using P1 vir transduction (80) by selecting ampicillin-resistant transductants. To deplete YhdP, cells grown overnight in LB with arabinose were washed once in LB and diluted 1:5,000 (vol/vol) in 25 ml LB with arabinose (1YhdP) or fucose (YhdP depletion) and grown at 37°C with shaking. Growth was monitored by measuring the OD 600 . For depletion on solid medium, overnight cultures grown in the presence of arabinose were serially diluted 1:10 in LB and stamped with a 48-pin replicator on the appropriate plates containing or lacking arabinose.
Efficiency of plating assay. Cultures grown overnight at 37°C in LB were serially diluted 1:10 (vol/ vol). Dilutions were transferred with a 48-pin replicator to various plates and incubated overnight at 37°C. Efficiency of plating values was calculated by dividing the highest dilution with growth for each strain under each condition by the highest dilution with growth for that the wild-type strain MG1655 on LB agar.
EDTA sensitivity assay. Cultures grown overnight at 37°C in LB were diluted 1:1,000 (vol/vol) in LB and delivered to 96-well plates. EDTA was added and serially diluted to generate a 1:2 range of concentrations. After overnight incubation at 37°C, the MIC was determined as the lowest concentration of EDTA that inhibited growth as determined by OD 600 .
Microscopy. Cells grown in liquid media as indicated in each experiment were layered on a 1% agarose pad in LB and imaged using phase-contrast with a 100Â oil immersion lens objective and a Nikon Eclipse Ti-E microscope equipped with a Nikon DS-QI1 cooled digital camera. See the supplemental material for details about cell measurements.
Suppressor selection and mapping. MacConkey-resistant suppressors were selected by plating 1 to 2 ml of overnight cultures of NR5161 on MacConkey agar plates. Selection plates were incubated overnight at 37°C. The ydbH suppressor alleles were mapped by P1 vir cotransduction frequency to tetracycline-resistant mini-Tn markers as described previously (84). Suppressor mutations were identified by amplifying the chromosomal ydbH locus via PCR and sequencing the resulting PCR product. A linked tet2-3 mini-Tn insertion [IGR(ynaE-uspF)::tet] was used to move ydbH suppressor alleles into various strains via P1 vir transduction and demonstrate that they are solely responsible for suppression.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.1 MB.

ACKNOWLEDGMENTS
This study was supported by the National Institute of General Medical Sciences award GM100951 and the National Institute of Allergy and Infectious Diseases award AI139271 (to N.R.).
We declare that we have no competing interest.