Dark Side of Escherichia coli Biogenic Inner Membrane: Overabundance of Three Main Phospholipids on Cytoplasmic Leaflet

Abstract

Transmembrane (TM) asymmetry of glycerophospholipids in the plasma membrane is well established, yet still a complex and evolving area of study regarding the exact mechanisms, magnitude, and composition. The emerging controversy challenges “Fluid Mosaic Sandwich” model simplicity by combining compositional (different types of lipids on each side) and numerical (different total amounts of phospholipids) asymmetries. While Gram-negative Escherichia coli phosphatidylethanolamine (PE) distribution is asymmetric (mostly cytoplasmic), the TM distribution of phosphatidylglycerol (PG) and cardiolipin (CL) across the inner (cytoplasmic) membrane (IM) is unknown. By integrating novel in vivo mapping techniques for PE, PG, and CL, we made the unexpected observation that their TM distribution does not correspond to the distribution of their total mass between the two IM leaflets. Instead, the underpopulated periplasmic leaflet can be occupied by lipid A, diacylglycerol (DAG), lipoproteins, and glycolipid MPase, whereas CL is found predominantly in the cytoplasmic leaflet of the IM. While PG is significantly enriched in the cytoplasmic leaflet of the IM in logarithmically growing E. coli, its asymmetrical distribution is inverted in stationary-grown cells, cells lacking CL, or cells subjected to growth in a hypotonic medium. These data suggest that cells can minimize the numerical differences in TM distribution by triggering a biosynthetic TM “cleanup” response, preventing membrane collapse. The numerical asymmetric interleaflet distribution of phospholipids may serve specific functional roles, potentially governed by critical threshold levels of lipid A and DAG that determine their compatibility for colocalization within the same leaflet. The overabundance of glycerophospholipids in the cytoplasmic leaflet of the IM may be driven by packing and localization constraints imposed during cell division and impacted by the rate of phospholipid biosynthesis and intermembrane flows. This imbalance may generate or result from the lipid concentration gradients and downhill flows across IM and between IM and outer membrane (OM), and the necessity to maintain appropriate electrostatic balance across the IM.

Graphical Abstract

INTRODUCTION

Lipid-rich membranes are found in all living cells and provide many vital functions. The chemical scaffold of all membranes is an asymmetric lipid bilayer composed of various noncovalently bound, highly interactive lipid molecules with different structures and physical properties.¹

Lipid asymmetry is a leading architectural principle for most, if not all, biological membranes, including the inner (cytoplasmic) membrane (IM) of Gram-negative bacteria, as we demonstrated recently by applying a novel vectorial lipidomics approach using either radiometric, spectrophotometric, and mass-spectrometric methods.² Phosphatidylethanolamine (PE) is asymmetrically distributed across the IM, with about 75% in the inner leaflet.² However, the physiological significance and detailed molecular mechanisms by which phospholipid asymmetry is generated and maintained, as well as its specific function in bacterial membrane processes, remain unknown.³

The complex double-membrane organization of Gram-negative bacterial cell envelopes places unique biosynthetic and topological constraints on phospholipids that are initially formed in the IM. All enzymes required for phospholipid synthesis are associated with the biogenic and energy-transducing E. coli IM. Although phospholipids are found in both the IM and the OM of Gram-negative bacteria, the OM is highly asymmetric, with lipopolysaccharide (LPS) residing in the outer leaflet and phospholipids (predominantly PE) in the inner leaflet.⁴ Sequential utilization of two primary amine-specific vectorial nonpenetrating (i.e., IM outer leaflet specific) 2,4,6-trinitrobenzenesulfonic acid (TNBS) and penetrating (1,5-difluoro-2,4-dinitrobenzene (DFDNB)) probes led to the development of a novel approach for determining head-amino and acyl-group asymmetries in the IM of Gram-negative bacteria.² The assay determines the pool of PE that is either fully accessible to or protected from the reaction with DFDNB or TNBS, respectively. Using IMs entirely devoid of OM, uniformly oriented inside-out (ISO) membrane vesicles sequentially treated with TNBS and DFDNB allowed us to confidently assess PE asymmetry in the IM of E. coli and Yersinia pseudotuberculosis.² This was the first unambiguous demonstration of PE asymmetry in the IM of Gram-negative bacteria, which has long remained unknown and controversial. Considering that PE makes up 60–65% of total phospholipids in the IM, distributed asymmetrically, with 75% on the cytoplasmic leaflet and the remaining 25% on the periplasmic leaflet of the IM, and the conventionally assuming that the abundances of phospholipids on each leaflet are similar (e.g., 50 mol % on each leaflet), 45 to 48 mol % of the lipids on the cytoplasmic leaflet of the IM should be occupied by PE, leaving almost “no room” for the abundant anionic phospholipids phosphatidylglycerol (PG) and cardiolipin (CL). However, these anionic phospholipids also play a crucial role in the cytoplasmic leaflet of the IM. They are associated with several physiologically essential processes, such as SecA-mediated protein translocation through SecYEG translocon,⁵ optimal DNA replication via lipid-dependent rejuvenation of inactive DnaA protein,^6,7 and proper cell division.^8–10 Therefore, the question remains: how do other phospholipid species distribute across the IM?

Although PG was discovered in plants and bacteria more than 70 years ago^11,12 and plays a vital role in bacterial metabolism, surprisingly, the TM distribution of PG across the IM of E. coli has never been reported despite many investigations suggesting leaflet-specific involvement of this lipid in many biosynthetic pathways. Two pools of PG (fast- and slow-metabolizing) were identified in E. coli.¹³ One significant drain on the PG pool occurs during the biosynthesis of CL, a universal component of energy-transducing membranes. In most bacteria, it is synthesized via the condensation of 2 molecules of PG or PG and PE catalyzed by cardiolipin synthase A and B (ClsA/B) or ClsC, respectively.^14–16 The second significant drain on the PG pool is the transfer of an sn-1,2-diacylglycerol group from PG to prolipoproteins catalyzed by phosphatidylglycerol: prolipoprotein diacylglycerol transferase (Lgt), which is essential for growth.¹⁷ Diacylglycerol is added to the N-terminal cysteine thiol group of lipoproteins. The active site of Lgt faces the periplasm, suggesting that periplasmic-facing PG is used. Cells lacking PG are not viable, primarily due to the lethal accumulation of the major unmodified lipoprotein (Lpp) at the IM.¹⁷ The third significant drain of the PG pool is biosynthesis of anionic osmoregulated periplasmic glucans (OPGs) (formerly known as periplasmic membrane-derived oligosaccharides (MDOs)) discovered by Gene Kennedy¹⁸ as a byproduct of the fast turnover of E. coli PG, which donates the phosphoglycerol moiety to glucans in a process mediated by phospho-glyceroltransferase I (the product of the opgB/mdoB gene). MDOs primarily maintain periplasmic osmolarity during growth in low-osmolarity medium.¹⁹ PG serves as a major precursor of MDO/OPG biosynthesis in E. coli and of anionic OPGs in other Gram-negative bacteria grown in hypoosmotic media.^18–20 On the periplasmic leaflet of IM, PG also donates its diacylglycerol-phosphate (phosphatidic acid: PA) moiety to the phospholipid-linked enterobacterial common antigen ECA_PG, which ultimately reaches the cell surface, tethered to the outer leaflet of OM via a phosphoglyceride.²¹ If PG functions as a precursor for the maturation of lipoproteins, for MDO and ECA_PG, which are localized in the periplasm, it is expected to be needed topologically at the outer periplasmic leaflet of the IM.

E. coli has three distinct enzymes that synthesize CL: ClsA, ClsB, and ClsC.^14–16 Despite this very marked redundancy in CL biosynthesis with none of three gene products being essential under growth at laboratory conditions, this redundancy already suggests that CL is important for bacterial fitness, notably to support the efficiency of protein insertion and translocation and outer membrane folding and assembly;²² CL is required in vivo for the stability of bacterial translocon and optimal membrane protein translocation and insertion²³ and membrane solute transport.²⁴ Moreover, CL-deficient strains exhibit several deficiencies that would limit their survival, vitality, and adaptation under stressful conditions. The absence of CL activates the Rcs envelope stress response, which represses the assembly and production of flagella, disrupts surface adhesion and biofilm attachment, and reduces biofilm formation and growth.²⁵ Deficiencies in CL result in drastic and pleiotropic effects on E. coli’s cell envelope ultrastructure.²⁵ CL is involved in maintaining cell size and proper cell division²⁵ and in supporting Z-ring positioning and MinD oscillation (Rowlett, Dowhan, and Vitrac, unpublished). CL generated by ClsA promotes the transport of poorly tetraacylated tetraacyl KDO2-Lipid A (LPS) substrates by MsbA. Underacylated forms stuck in the IM,^26,27 suggesting that CL produced by ClsA acts as a limiting factor in the regulation of LPS biosynthesis and transport to the OM, supporting an optimal MsbA activity.

Current methods for quantifying and visualizing CL are hampered by a lack of specificity for this lipid. 10-N-nonyl acridine orange (NAO) is widely used for fluorescent detection of CL in bacteria and mitochondria. NAO exhibits a characteristic red shift of emission spectra, which was previously considered to be unique for the interaction between this probe and CL. Nevertheless, most of the currently available techniques do not differentiate between monoanionic PG and dianionic CL. Traditional approaches to defining anionic phospholipid content use promiscuous NAO²⁸ and TTAPE-Me.²⁹ Indeed, TTAPE-Me emits little or no fluorescence when interacting with net neutral (PE or PC) or positively charged lipids.²⁹ It was utilized to quantify CL in the exterior leaflet of liposomes^30,31 in binary lipid mixtures lacking other anionic lipids. We recently demonstrated that this probe was not specific for CL and detected PG with the same efficiency;³² therefore, it is not suitable for determining CL transbilayer distribution in the presence of PG and other anionic phospholipids.

Identification, visualization, and quantification of PG and especially CL in biological membranes is of great interest due to the critical structural and physiological roles of these lipids. However, the distribution of both PG and CL between the membrane monolayers in bacterial IM remains unknown. Thus, selective nonpermeant leaflet-specific probes are in great demand for mapping the distribution of these lipids within lipid bilayers. In this manuscript, we adapted the protocol to dissect the steady-state distribution of PE across the IM of living, intact E. coli cells and report a novel methodology for mapping the transmembrane distribution of PG. Our newly developed method first employs periodate (NaIO₄) oxidation of intact cells to convert the central hydroxyl of glycerol in the PG headgroup to an aldehyde (oxidized PG, designated as oxPG) group (Figure 2A). NaIO₄ permeates the outer membrane (OM) freely and converts the entire pool of PG to oxPG regardless of its distribution across the IM (Figure 2A). Anionic oxPG is converted further to a zwitterionic phospholipid by reaction of the aldehyde with membrane impermeable (hydrazinocarbonylmethyl)trimethylammonium chloride (betaine hydrazide, Girard reagent T), containing a quaternary amine (Figure 2A and Figure S2). The success of any chemical modification procedure for mapping membrane phospholipid asymmetry depends upon the inaccessibility of one side of the membrane to the modifying reagent. This requires that the membrane remain uncompromised and impermeable during the covalent modification. Aside from its simplicity, this approach has the advantage of using whole intact live cells to assess the sidedness of PE and PG, thereby avoiding problems associated with converting cells into membrane vesicles with uniform orientation. This protocol can be universally adapted for use with Gram-negative and Gram-positive intact cells, as well as with single-membrane cellular systems and liposomal systems (Figure 2B and Figure S1).

Figure 2.

Rationale for oxidation of PG and subsequent covalent modification of oxPG provides guidance for developing robust in vitro and in vivo transbilayer PG mapping assays. (A) Schematic representation of PG oxidation and subsequent derivatization of oxPG by BtH. (B) TLC analysis of NaIO₄ oxidation of SUVs and subsequent derivatization of oxPG by BtH. SUVs prepared from POPG, E. coli PG, or a mixture of E. coli PE/PG were subjected to PG oxidation on ice at 37 °C with 50 mM periodate. Five mM BtH was used for subsequent oxPG derivatization. TLC separated extracted lipids with solvent 1: chloroform/methanol/acetic acid (60/25/8, v/v/v). Individual phospholipids were visualized by iodine vapor. PE and CL were placed as a standard reference. (C and D) TNBS or DFDNB pretreatment of wild-type E. coli intact cells prior to PG oxidation prevents the formation of undesired oxPG molecules. Wild-type E. coli cells were preincubated with either TNBS or DFDNB, and then subjected to PG oxidation. Extracted lipids were separated by TLC with solvent system 2: chloroform/methanol/ethanol/acetic acid (60/20/5/8, v/v/v/v). The PG/CL ratio in nontreated cells, as well as in cells treated with TNBS or DFDNB, was determined by three independent experiments. (E) TNBS and the following periodate-treated E. coli cells were incubated with the indicated concentration of BtH. Extracted lipids were separated by TLC using solvent system 2: chloroform/methanol/ethanol/acetic acid (60:20:5:8, v/v/v/v). (F) The ratio of Bt-PG and PG was plotted as a function of BtH concentration. (G) The estimated abundance of phospholipids across the IM and the established interleaflet phospholipid abundance imbalance. Filling the gaps in the periplasmic leaflet of IM with “missing” amphipathic DAG and lipid A molecules.

These newly developed approaches revealed that PG distributes asymmetrically in the IM. However, in contrast to our estimation, the transbilayer distribution of PE and PG across the IM appears to be incompatible with the expected distribution of their phospholipid mass between two leaflets, 50%/50% (cytoplasmic/periplasmic). The cytoplasmic leaflet of the IM exhibited the highest possible abundance of PE and an overabundance of PG compared to the periplasmic leaflet. This unexpected distribution of PG led us to propose that the TM distribution of phospholipids across the IM is not uniform, e.g., phospholipids are overabundant on the cytoplasmic leaflet of the IM. Such a distribution creates a significant deficiency of glycerophospholipids in the periplasmic leaflet, which is presumably filled with lipid A, DAG, lipoproteins and other molecules (glycolipid Membrane Protein Insertase: MPIase). The TM distribution of PG presumably reflects the relative rates of its synthesis and renewal, such as its translocation and utilization as metabolic precursors, or is supported by an active mechanism that remains obscure. With this methodology and the means to control lipid bilayer asymmetry genetically (ΔclsABC null) and biosynthetically (this work), we are in a position to investigate the physiological importance of membrane lipid asymmetry in diderm and monoderm bacteria. We can manipulate TM distribution using a protein engineering approach e.g orienting ClsA catalytic domain “manually” into periplasm by enforcing “negative outside bias” and effectively control the spatial delivery of CL specifically to the outer leaflet of the IM.³³ We have also developed a new approach to determine the TM distribution of CL in intact cells. Heterologously expressed mitochondrial CL-specific deacylase Cld1p “deacylates” nearly all of the CL, demonstrating for the first time that CL is near exclusively localized to the cytoplasmic leaflet of IM.

Most biological membrane models implicitly assume that both leaflets have similar lipid abundances. This traditional view assumes that because a bilayer is a “sandwich,” the top and bottom leaflets must have roughly the same number of lipid molecules to maintain structural integrity (a 1:1 ratio). Experimental methods used in the past to analyze living cells often struggled to differentiate the exact count between leaflets.³⁴ The emerging controversy centers on whether these two leaflets actually contain the same number of glycerophospholipid molecules or if one is “overpopulated”.^34–36 Recently published results suggest a significant “steady-state” TM abundance imbalance, with potentially more than 50% more glycerophospholipids in the cytoplasmic leaflet compared to the exoplasmic leaflet of eukaryotic plasma membrane, maintained by ATP-dependent transporters and potentially balanced by asymmetric cholesterol distribution.³⁶ How can a bilayer “survive” if one side is overcrowded? The bilayer survives phospholipid imbalance (overcrowding on one side) through a combination of mechanical adaptation to new topological constraints and metabolic regulation. In the eukaryotic plasma membrane, such a massive numerical imbalance is physically possible due to the unique behavior of cholesterol, which may spontaneously and rapidly flip between leaflets to buffer the mechanical and surface area stresses created by the glycerophospholipid imbalance.³⁶ It is distributed asymmetrically to “fill the gaps” or regulate the packing density, allowing the membrane to sustain higher phospholipid counts on one side without losing its structural integrity. While the existence of TM compositional asymmetry is firm, the exact quantification of this imbalance in different types of eukaryotic and prokaryotic cells and membranes (self-made, self-assembling biogenic bacterial cytoplasmic membrane and ER versus nonbiogenic eukaryotic plasma membrane)^1–3 the role of specific flippases,³³ and the precise, in vivo, leaflet-by-leaflet distribution of glycerophospholipids remains area of active research and debate.³⁴

RESULTS

Steady State Distribution of PE across the IM of Living E. coli

We previously developed a methodology for the in vivo determination of the primary amine headgroup asymmetry of PE and PS.² The methodology utilizes the single or sequential addition of two primary amine-specific probes with different chromogenic and permeability properties. Primary amine-specific membrane-impermeable TNBS is added, followed by membrane-permeable DFDNB. Radiometric, spectrophotometric, or mass spectrometric methods (for acyl chains and head groups) were used to quantify transbilayer lipid asymmetry. We determined the steady-state and newly synthesized PE (in radiolabeled pulse-chase experiment) pools after isolating uniformly oriented inside-out IM vesicles (ISOvs) fully devoid of OMs. We demonstrated for the first time that zwitterionic PE is asymmetrically distributed across the IM of E. coli and Yersinia pseudotuberculosis with about 75% in the inner (cytoplasmic) leaflet and 25% in the outer (periplasmic) leaflet. From these results, we can estimate that anionic PG and CL make up most of the remaining phospholipids in the IM, thus making the inner leaflet of the E. coli cytoplasmic membrane less negative than the outer leaflet.²

To ensure that lipid redistribution does not occur during French Press-aided ISOvs isolation, we simplified and further developed this protocol using intact cells. Although solid evidence was provided² that ISO vesicles formed during rupture of cells have the same uniform inside-out orientation, it cannot be a priori excluded that the reported asymmetry of PE² is not due to redistribution of PE and other phospholipids during the disruption and resealing process. Thus, radiolabeled intact E. coli cells were treated with TNBS, which penetrates the OM but not IM³⁷ and therefore expected to label predominantly the periplasmic leaflet of IM prepared by differential centrifugation (Figure 1A–a). In this assay PE sidedness can be estimated from the amount of TLC resolved radiolabeled PE and its derivatized product [trinitrophenyl-PE (TNP-PE)] (Figure 1A–b) obtained after TNBS treatment of intact ³²PO₄-labeled cells from which IM devoid of OM was isolated, as evidenced by Western blots of subcellular marker proteins for an OM protein (PldA) (Figure 1B). Figure 1C clearly demonstrates that only a portion of PE is protected from derivatization by TNBS verifying its established TM distribution across the IM of isolated uniformly oriented membrane vesicles.² The amount of PE that reacted with TNBS did not increase at concentrations above 3 mM (Figure 1C and D), indicating that TNBS labeled all surface-accessible amino groups on the IM under these conditions, as the reaction reached saturation of available primary amines. Upon TNBS treatment, a TNP-PE spot appeared on TLC, and the sum of the radioactivity in the PE and TNP-PE spots always remains nearly constant (Figure 1C left panel and D) at steady state labeling, demonstrating that IM fully maintains a previously established PE asymmetry with a 77%/23% (cytoplasmic/periplasmic leaflet) distribution in ISO vesicles (Figure 1D and ref 2).

Figure 1.

Side-specific covalent modification of PE as a tool to dissect the asymmetry of this lipid in living E. coli cells. (A-a) Two-membraned envelope of Gram-negative bacteria with special focus on the IM. (A-b, A-c, and B) The PE sidedness assay determines whether PE is fully accessible to or protected from TNBS reaction. The PE transmembrane sidedness can be estimated from the amount of TLC resolved radiolabeled PE and its derivatized product [trinitrophenyl-PE (TNP-PE)] obtained after TNBS treatment of intact ³²PO₄-labeled cells from which IM devoid of OM contamination was isolated, as evidenced by Western blots of subcellular marker proteins for an OM protein (PldA) (B). Two-step differential centrifugation efficiently separated E. coli OMs and IMs, as judged by marker OM enzyme assays. (C–E) PE modification by TNBS in E. coli intact cells. By employing ³²PO₄-steady state labeling, the mol % of TNP-PE (periplasmic leaflet of IM), unmodified PE (cytoplasmic leaflet of IM) was quantified. Wild-type or CL-deficient cells were diluted into fresh LB medium to an OD₆₀₀ of 0.05. Cells were grown to mid log phase (OD₆₀₀ 0.6–0.8) with 1 μCi/mL ³²PO₄. Collected [³²P]-labeled cells were treated with TNBS at the indicated concentration, followed by phospholipid extraction as described in Materials and Methods. Extracted individual phospholipids were separated by TLC using solvent system 1 (chloroform/methanol/acetic acid (60/25/8, v/v/v)) and visualized using a Phosphoimager (C). The mol % of PE and TNP-PE (D) and the TNP-PE/PE ratio (E) determined by two individual experiments were plotted as a function of TNBS concentration.

Theoretically, an asymmetry of one nonbilayer-prone phospholipid (PE) cannot be disconnected from the asymmetry of another (CL). Surprisingly, the elimination of nonbilayer CL did not affect the transmembrane distribution of nonbilayer PE (Figure 1C right panel and E). Most importantly, our new protocol eliminates the need to isolate IM fractions by differential centrifugation of cells disrupted by French Press on isopycnic sucrose density step gradients.

Experimental Rationale for Topological Mapping of PG in Intact E. coli Cells

Although phospholipids are found in both the IM and the OM, the OM is highly asymmetric, with LPS residing in the outer leaflet and phospholipids (predominantly PE) in the inner (periplasmic) leaflet.³⁸ Therefore, mapping of PG exclusively located in the IM can be performed in intact cells without subsequent isolation of a pure IM fraction if vectorial probes with different permeabilities across the IM are available.

PG’s headgroup comprises a second glyceryl moiety on the phosphate as a (1,2)-diol group, which can be converted to an aldehyde group by oxidation with NaIO₄ (Figure 2A) and therefore potentially can be used to selectively modify the PG in the cytoplasmic membrane of Gram-positive and Gram-negative bacteria and synthetic liposomes.³⁹ However, Lentz et al. demonstrated that a high concentration of periodate (50 mM) was sufficient to oxidize only half of the amount of PG in small unilamellar vesicles (SUVs); while the entire pool of PG was oxidized by periodate only after solubilization with detergent sodium cholate.³⁹ These observations suggest that periodate does not permeate the lipid bilayer of these SUVs; therefore, it can be used to modify PG in one leaflet of the SUVs selectively. Although these results were consistent with the distribution of PG in synthetic LUVs estimated from the concentration of formaldehyde released during the oxidation of the glycerol group of PG by periodate,⁴⁰ the entire pool of E. coli PG was instantly oxidized by periodate regardless of the orientation of vesicles isolated from E. coli (either right-side-out (RSO) or inside-out (ISO)), as well as in intact cells suggesting that the periodate is crossing the biological membrane so rapidly that oxidation of PG in both the inner and outer leaflets occurs near at the same rate.⁴¹ Indeed, the oxidation of PG was fast and complete, and all PG molecules were oxidized monophasically, demonstrating that PG behaves as a single pool with respect to its reactivity with the oxidizer. It was concluded that an application of periodate as a specific membrane side-specific 1,2-diol oxidizer of PG and experimental impairment of PG sidedness from potential biphasic kinetics of oxidation, attributed to stepwise covalent modification of different leaflets of IM, cannot be utilized irrespective of either concentration of the oxidizer or reaction temperature.

Since periodate can selectively modify the total pool of PG in either isolated ISO vesicles or intact cells, we decided to use this reaction as a prerequisite for the subsequent application of a side-specific, membrane-impermeable vectorial probe. We took advantage of the fact that an aldehyde group can be subsequently covalently modified by hydrazide via formation of a stable hydrazone bond (Figure 2A and Figure S2). The employment of such water-soluble and membrane non-permeable bulky hydrazide should allow the estimation or determination of the transmembrane distribution of PG across the IM. Since Huijbregts et al. evaluated the oxidation of PG by periodate based on the disappearance of PG from its original position on TLC,⁴¹ we first examined whether the entire pool of PG is successfully and fully oxidized, and whether a single spot of oxPG appears accordingly after thin layer chromatography (TLC) of extracted lipids. Small unilamellar vesicles (SUV) composed of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylglycerol POPG were treated with periodate on ice, followed by lipid extraction and development by TLC and visualization. Indeed, a reduction in the amount of POPG after periodate treatment and the appearance of a new spot corresponding to oxPG were observed (Figure S1). Consistent with previous findings,⁴⁰ not all PG was oxidized in liposomes; however, all PG was converted to oxPG by oxidation when SUVs were solubilized by deoxycholate (DOC) (Figure S1), indicating that periodate does not penetrate across the synthetic lipid bilayer as was demonstrated before.³⁹

Next, we test whether oxPG can be covalently modified by a set of hydrazide compounds. We tested three different hydrazide analogs, Lucifer Yellow, biocytin-hydrazide, and Girard’s reagent T (betaine hydrazide, BtH) (Figure 2B). A pitfall of this two-phase extraction was different yields in retrieving covalently modified lipids from the chloroform-rich layer as the basis of quantitative extractions and the avoidance of bias due to technical variability. Due to their different polarities, these hydrazine analogs were evaluated and thoroughly compared with an amount of covalently modified oxPG in a pooled chloroform extract that was qualitatively and quantitatively considered to represent an average standard extract with satisfactory quantitative recovery. Among them, due to the different polarity and solubility of oxPG covalent adducts in an upper methanol-rich layer, containing hydrophilic compounds, only BtH-modified oxPG (Bt-PG) was quantitatively extracted in lipid-rich organic phase by a conventional two-phase lipid extraction procedure and clearly separated on TLC (Figure 2C). Since the hydrazone bond is basically not stable at acidic conditions, the extraction was performed at neutral pH. The derivatization of oxPG by BtH did not proceed at low reaction temperature (Figure 2B); thus, the BtH treatment was performed at 37 °C (Figure 2C). SUVs made of E. coli PG exhibited the same disproportionality in PG oxidation (Figure 2C). Although oxPG exhibited nearly similar mobility to CL (Figure 2B), this problem was circumvented in the following experiments as described below. Due to the satisfactory quantitative recovery of covalently modified PG in the chloroform phase, full liposomal or IM impermeability, water-soluble BtH was chosen for the IM outer (periplasmic) leaflet specific covalent modification of oxPG in all the following experiments.

PE Contributes to the Formation of “Undesired” oxPG in E. coli Intact Cells

Since BtH is membrane impermeable, outer (periplasmic) leaflet PG (represented by Bt-PG after covalent modification of oxPG) can be distinguished from inner leaflet PG (oxPG) of IM to assess the transbilayer distribution of PG. Since PG near exclusively resides in the IM, we oxidized and chemically modified PG of intact E. coli cells without their fractionation into IM and OM. We found that the presence of high concentrations of diol-containing sucrose, sorbitol, or glycerol, which are typically added as cryoprotectants for the storage of membrane preparations at −80 °C, chemically scavenges periodate and drastically decreases its effective concentration and therefore should be avoided during the oxidation step.

Since we were unable to separate oxPG and CL on TLC using our conventional solvent system (Figure S1), we first used CL-deficient cells (BKT12: ΔclsABC) to test whether all bulk PG can be quantitatively converted to oxPG in intact E. coli cells. CL-deficient cells were grown with ³²PO₄ to radiolabel phospholipids in both leaflets of IM uniformly and were subsequently subjected to periodate oxidation. However, we observed that only about half the amount of PG was converted to oxPG when oxidation was performed at 37 °C (Figure S3A and B). Furthermore, we observed that several potential oxPG spots were migrated at unexpected (above CL) positions on TLC (indicated as oxPG* in Figure S3A). Performing PG oxidation on ice partially but not entirely suppressed the formation of undesired oxPG (Figure S3A). Since the formation of undesired oxPG* was not observed when SUVs were prepared either from synthetic POPG or E. coli PG (Figure 2B and C), the heterogeneity of fatty acid composition cannot explain the appearance of the undesired oxPG* spots in intact cells. The sum of radioactivity of oxPG and undesired oxPG* spots was equivalent to the original amount of PG in nontreated cells (Figure S3B), supporting that undesired oxPG* molecules were derived from PG.

The most abundant phospholipid in E. coli is PE, which accounts for 75–80 mol % of total phospholipids. The ethanolamine headgroup contains a primary amine that can react with the aldehyde group of oxPG to form a Schiff base (Figure S2). Although the stability of Schiff-base is lower than the hydrazone bond, the formation of Schiff-base between oxPG and PE, even if it occurs transiently, could be responsible for the formation of undesired form of oxPG*. To examine the role of PE’s primary amine in the formation of undesired oxPG*, the PE headgroup was derivatized with a membrane-permeable primary amine-reactive DFDNB before PG oxidation. In this case, the entire PG population was successfully converted to oxPG without the appearance of undesired oxPG* (Figure S4A). Furthermore, oxPG was successfully derivatized by subsequent BtH treatment (Figure S4A).

Preventing the formation of undesired oxPG* upon PG oxidation enabled us to seek out the solvent system that can distinguish oxPG from CL on TLC. We added ethanol to the current solvent system (chloroform/methanol/acetic acid (60/25/8, v/v/v)) and changed the proportion of ethanol and methanol. We found that increasing the ethanol percentage clearly prevents overlapping of oxPG and CL mobilities on TLC, and all phospholipids and PG derivatives were successfully separated on TLC developed with optimal solvent ratios, e.g., chloroform/methanol/ethanol/acetic acid (60/20/5/8, v/v/v/v) (Figure S4A).

To verify further the involvement of the primary amine of PE in the formation of undesired oxPG*, PE-deficient cells (pssA::kan) were subjected to PG oxidation. We observed that the formation of oxPG was significantly suppressed in the cells lacking PE, and most of oxPG migrated to the expected position as a single spot (Figure S5), indicating that the formation of undesired oxPG* molecules during PG oxidation is a consequence of PE presence. Consistent with this notion, the potential undesired oxPG appeared when SUVs containing both E. coli PE and PG were incubated with NaOI₄ (Figure 2B). The undesired oxPG* molecules could contain PE-oxPG adducts or N-formyl-PE since formaldehyde is produced from the oxidation of the terminal vicinal hydroxyl, and reactions of primary amine with aldehyde may produce the carbinolamines rather than the imine derivatives. Although the disruption of PE synthesis significantly but not fully suppressed the formation of undesired oxPG* during oxidation, DFDNB treatment completely prevented undesired oxPG* formation, even though this mutant does not contain PE (Figure S5). These results suggest that not only the primary amine on the PE headgroup, but also other primary amine-containing molecules, are involved in the formation of the undesired form of oxPG*. It is noteworthy that the amount of oxPG after oxidation was higher than expected in PE-deficient cells (Figure S5). However, the increase in the oxPG was not observed when cells were treated with DFDNB before oxidation (Figure S5). These results suggest that the observed reverse, energy-independent reaction, which may stimulate the conversion of CL to PG, is sensitive to DFDNB. Most likely, all biomolecules containing primary amines, especially PE, are involved in the formation of undesired oxPG* during oxidation; however, their generation can be prevented by DFDNB treatment before PG oxidation.

PG is Asymmetrically Distributed across the IM of Intact E. coli Cells

Since we circumvented the formation of undesired oxPG* by derivatizing the primary amine on the PE headgroup prior to PG oxidation, we intended to apply this approach to determine the distribution of PG across the IM in wild-type cells. However, unexpectedly, we observed that the amount of CL was significantly increased at the expense of PG after treatment of cells with DFDNB, and these changes in PG and CL levels were independent of oxidation (Figure 2C and D). Since the increase in CL level upon DFDNB treatment was concomitant with a decrease in PG level, a reciprocal relationship or interconversion between them was anticipated. Indeed, when CL-deficient cells were subjected to PE derivatization by DFDNB and following the PG oxidation, PG was quantitatively converted to oxPG, and the formation of undesired oxPG* was not observed in DFDNB-treated cells (Figure S4A). Thus, treatment of cells with DFDNB appears not to be appropriate to circumvent pitfalls in topological PG mapping. In contrast, we found that membrane-impermeable primary amine-reactive TNBS can also prevent the formation of undesired oxPG*. Moreover, the interconversions between PG and CL molecules did not occur upon TNBS treatment (Figure 2C and D). Considering that cells are trying to homeostatically adjust lipid composition in response to PE derivatization by net negatively charged DFDNB, only TNBS was found to be suitable for subsequent leaflet-specific modification of oxPG by BtH. Therefore, vectorial labeling of PE with TNBS should be done before oxidation and covalent modification of PG to prevent the formation of undesired oxPGs. Moreover, this rationale, using non-IM-permeant TNBS, allows us to simultaneously validate the asymmetry of PE and confirm that the IM’s integrity is not compromised by periodate treatment.

Next, we tested whether oxPG can be successfully modified by externally added BtH in intact E. coli cells. By employing ³²PO₄-steady state labeling, the mol % of betaine-PG (Bt-PG) (periplasmic leaflet of IM) and oxPG (cytoplasmic leaflet of IM) were quantified. TNBS-pretreated wild-type cells were subjected to PG oxidation, followed by incubation with BtH at various concentrations. Since oxPG derivatization by BtH was observed only when the reaction was performed at 37 °C and undesired oxPG could form at this temperature, even though cells are pretreated with TNBS, unreacted periodate was washed out of the reaction mixture before oxPG modification. The cells remain intact during PG oxidation since not all PE was derivatized by membrane-impermeable TNBS due to its established natural asymmetric distribution (Figure 1 and ref 2) and, as shown in Figure 2E, an oxPG spot was converted to Bt-PG in a concentration-dependent manner. The derivatization was saturated at 5 mM BtH (Figure 2E and F). At 5 mM BtH, about 40% of oxPG was derivatized (Figure 2E and F), suggesting that PG is distributed asymmetrically across IM with an approximate ratio of 60%/40% (cytoplasmic/periplasmic leaflet). We confirmed that Bt-PG formation was dependent on prior PG oxidation (Figure S6). To examine whether the determined PG distribution in intact cells is consistent with PG asymmetry, particularly across the IM, the IM fraction was isolated after PG oxidation and BtH derivatization. Consistent with the results obtained from intact cells, about 40% of oxPG was derivatized by BtH (Figure 2G). Furthermore, we observed that oxPG was absent in the isolated OM (Figure S7), which indicates that the ratio of Bt-PG to oxPG reflects the distribution of PG across the IM. The assay for TM distribution of PG relies on impermeability of water-soluble and bulky BtH to cross the IM and thereby its ability to react with PG only in the periplasmic leaflet leaving the cytoplasmic leaflet unreacted. These results strongly suggest that not only PE but also PG are distributed asymmetrically across the IM.

Although the absolute quantification of PG and CL in the diderm envelope is still challenging and resulting in “approximate” and very often contradicting measurements,^38,42 PG and CL may present in the periplasmic leaflet of the OM as followed from old and most recent independent studies. Irrespective to this view, our PG mapping assay determines the pool of PG in whole cell or in the IM that are either fully accessible to oxidation or protected from the reaction with BtH (Figure 2E (cells) and 2G (IMs)). The oxidation of the diderm envelope will lead to total oxidation of residual PG in the periplasmic leaflet of the OM and its subsequent covalent modification by BtH to Bt-PG. Thus, the amount of covalently modified PG (e.g Bt-PG) corresponding to periplasmically localized PG (IM+OM) can be potentially overestimated. However, we are reproducibly seeing more oxPG (Figure 2E) in cells or in the IM (Figure 2G) which corresponds only to the cytoplasmic leaflet of the IM because this form of PG is fully protected from the reaction with BtH. oxPG remains intact because these molecules are confined to the inaccessible inner (cytoplasmic) leaflet of the IM. Thus, these results hold true whether intact cells or the IMs fully devoid of the OM were utilized in the assay and support the model that the cytoplasmic leaflet of the IM is disproportionately populated by glycerophospholipids PE and PG to maintain membrane order and function. Sidedness of glycerophospholipids, however, can be altered during metabolic or morphological changes or in stressed E. coli cells.

Most importantly, all modifications with vectorial probes can be done in intact cells and our new protocol eliminates the needs for isolation of uniformly oriented inside-out inner membrane vesicles. Although the protocol has been developed using E. coli, it can be directly adapted to other Gram-negative bacteria, including pathogens. However, to ensure that a PG mapping assay is robust against the presence of variable amounts of anionic phospholipids in the OM of diderm bacteria, particularly when the Mla system is compromised⁴³ or overwhelmed or when pathogens accumulate high amounts of anionic phospholipids within their OM,^44,45 control experiments focusing on isolated IM fractions are required.

CL is Near Exclusively Localized to the Cytoplasmic Leaflet of the IM

The empirically determined PG distribution contrasted with our estimate, which was based on the previously established² and reaffirmed (Figure 1E) asymmetry, showing that more than 70% of PE is on the cytoplasmic leaflet of the IM. Due to this highly asymmetric PE distribution, there is clearly not enough “room” to accommodate such a large amount of PG on the cytoplasmic leaflet of the IM, according to the conventional view that both membrane leaflets have similar glycerophospholipid abundances. These unexpected results are not artifacts arising from potential cleavage of hydrazone bonds during acidic TLC separation, as the same amount of Bt-PG was observed when TLC was developed with a neutral solvent (Figure S4B), because BtH-PG was unstable during TLC development using an alkaline solvent system (Figure S4C). From these results, we can estimate that the abundance of phospholipids between the two leaflets is not equal; more phospholipids are located on the cytoplasmic leaflet of the IM. However, a stable lipid bilayer, essential for structuring the cell envelope, requires proper distribution of amphipathic molecules between leaflets; thus, the underpopulated periplasmic leaflet of the IM must be filled by appropriate candidate molecules that can balance excess phospholipids in the cytoplasmic leaflet by accumulating in this exoplasmic leaflet.

CL accounting for 5–10 mol % in the IM is a potential lipid candidate to fill a less-populated periplasmic leaflet of the IM. Although CL was experimentally visualized on both sides of the IM, the exact distribution of CL was not determined.⁴⁶ To further establish the presence of CL on the cytoplasmic leaflet of the IM, we heterologously expressed the mitochondrial CL-specific deacylase Cld1p, which is associated with the matrixfacing leaflet of the mitochondrial IM to specifically deacylate CL.⁴⁷ Since Cld1p is a peripheral protein that lacks Sec and Tat signal sequences, it is expected to localize strictly to the cytoplasmic leaflet of the IM in E. coli. In E. coli, the generated lysophospholipids are normally reacylated by Aas⁴⁸ (Figure 3A). Thus, we disrupted the aas gene and examined the effect of heterologous expression of Cld1p on CL levels. We observed that the level of CL remained unchanged upon Cld1p expression in aas⁺ background; however, it decreased notably in Δaas cells, with accumulation of the fully deacylated form of CL (hg-CL) (Figure 3B), previously identified by us using mass spectrometry.⁴⁹ The hg-CL spot upon Cld1p expression did not appear in the clsA null strain, further supporting that it is derived from CL (Figure 3C).

Figure 3.

E. coli IM leaflet-specific diacylation of CL. (A) Schematic representation of the deacylation of CL by Cld1p and reacylation by Aas. (B and C) Wild-type cells (W3110) and its Δaas or ΔclsA derivatives harboring pBAD18-Cld1p were grown until early log phase with supplementation of 0.002% arabinose. Phospholipids were uniformly labeled by 1 μCi/ml ³²PO₄. (D and E) W3110 and its Δaas derivative were pulse-labeled for 1 min, followed by chasing for the indicated time. After lipid extraction, individual phospholipids were separated by TLC using chloroform/methanol/water/ammonia (60/37.5/3/1, v/v/v/v), and then detected by Phosphoimager.

During the experiments, we observed that not all CLs were deacylated by Cld1p (Figure 3B and C), implying that a subset of CL was delivered to the periplasmic side of the IM or sequestered around IM proteins, thereby avoiding Cld1p-mediated deacylation. To examine this possibility, pulse-chase experiments were performed. In the aas⁺ strain, CL was synthesized subsequently to PG production, and CL levels continued to increase throughout the chase period (Figure 3D and E–a). In contrast, in the Δaas strain, CL levels were plateaued after 10 min, concomitantly with the appearance of hg-CL (Figure 3D and E–b). This equilibrium likely results from CL escaping Cld1p-mediated diacylation by clustering of this lipid around proteins,⁵⁰ phase separation, or spatial restrictions, as an equal rate of CL synthesis and diacylation would prevent its accumulation in the Δaas strain. In addition, we observed that PG synthesis was upregulated/or its turnover was enhanced while PE synthesis was repressed in the Δaas strain, suggesting a cellular response to compensate for CL synthesis. This acute response explains the observation of an imbalance of interconverted CL and hg-CL levels detected in the cells with induced and noninduced Cld1p (Figure 3B).

All these results demonstrate for the first time that CL is predominantly localized to the cytoplasmic leaflet.

Effect of the Disruption of CL Synthesis on the Transmembrane Distribution of PE and PG across the IM

CL is synthesized via the condensation of 2 molecules of PG or PG and PE, catalyzed by ClsA/B or ClsC, respectively.^14,15 CL deficiency generated by the disruption of three cls genes is accompanied by a drastic increase of PG levels mainly at the cell poles, most likely to compensate for the anionic function of CL in the bacterial membrane.²⁸ Thus, we tested the effect of CL deficiency on PE and PG asymmetry. TNBS treatment of CL-deficient intact cells followed by isolation of IMs revealed that PE distribution across the IM in CL-deficient cells remained the same as in wild-type cells (Figure 1E). Therefore, the asymmetry of one nonbilayer-prone phospholipid (PE) can be disconnected from the asymmetry of another nonbilayer phospholipid (CL) because the elimination of nonbilayer CL did not affect the transmembrane distribution of nonbilayer PE. Next, we performed sequential PE and PG mapping of CL-deficient cells. We used 5 mM BtH as a saturated concentration for oxPG derivatization. We observed that about 70% of oxPG was converted to Bt-PG in CL-deficient cells (Figure S4A and Figure 2F), suggesting that PG asymmetry is inverted when CL synthesis is disrupted. These results indicate that either the consumption of PG by CL synthase or the presence of CL per se affects the transmembrane distribution of bilayer-prone PG rather than nonbilayer counterpart PE.

Candidate Amphipathic Molecules that can Balance Excess Phospholipids in the Cytoplasmic Leaflet by Accumulating in the Exoplasmic Leaflet

Although lipid A and diacylglycerol (DAG) are minor lipid species in the IM, these lipids could contribute to filling the underpopulated periplasmic leaflet of the IM. Thus, we intended to map the TM distribution of these lipids across the IM. To map the TM distribution of lipid A, we expressed Francisella LpxE, which efficiently dephosphorylates bulk E. coli lipid A at the periplasmic face of the IM (Figure 4A) and can therefore be used as a topological reporter.²⁶ LpxE acts on the periplasmic surface of the IM, removing the 1-phosphate group from di- and triphosphorylated lipid A species. While LpxE is generally known for converting 1-monophosphorylated lipid A to a dephosphorylated or 1-dephospho forms, it has been demonstrated to act on 1-diphosphate lipid A of diphosphorylated and triphosphorylated species, to remove the 1-position phosphate(s), yielding a 1-dephospho-4′-monophosphorylated species.^26,51 Upon heterologous expression of LpxE, lipid A is dephosphorylated, resulting in monophosphorylated lipid A that migrates much faster than the two other most abundant lipid A species. Upon expression of LpxE, 97% of lipid A was dephosphorylated (Figure 4A), suggesting that nearly all lipid A is localized in the periplasmic leaflet of IM.

Figure 4.

Estimated TM sidedness of lipid A and DAG in E. coli IM. (A) Lipid A profile of E. coli cells heterologous expressed Francisella novicida LpxE. Cells were uniformly labeled with ³²PO₄, lipid A was isolated and separated by TLC, and Phosphorimaging subsequently visualized the resulting bands. % LpxE was calculated by radioactivity of the monophosphate lipid A divided by total lipidA radioactivity. (B) DAG sidedness was estimated from the extent of its phosphorylation by endogenous DgkA (e.g., amount of synthesized PA) before and after detergent treatment of the identical amounts of uniformly oriented ISO made from E. coli cells outgrown on ¹⁴C-acetate.

DAG is also predominantly localized in the periplasmic leaflet based on the extent of conversion of cellular ¹⁴C-radiolabeled DAG into PA in uniformly oriented ISO vesicles assayed before and after solubilization of ISO vesicles with detergent (Figure 4B). These results suggest that lipid A and DAG could contribute to stabilizing the lipid bilayer by filling potential gaps in the underpopulated glycerophospholipid periplasmic leaflet of IM.

When LPS transport to the OM is disrupted newly synthesized LPS does not simply cease to be made, instead, it accumulates in the periplasmic leaflet of the IM. When LPS accumulates in the periplasmic leaflet of the IM, YejM allows the LapB/FtsH system to degrade LpxC, the enzyme responsible for the first committed step in Lipid A synthesis, thereby downregulating further production. However, the existence of YejM system does not prove that accumulation is unlikely; rather, it proves that accumulation occurs and serves as a crucial signal for the cell to prevent toxic overload of the IM. While precise numerical thresholds for “toxicity” are not explicitly defined, accumulation of lipid A (specifically its precursors like Lipid IV or LPS) in the IM is generally disruptive to membrane integrity, the threshold for nonmatured and matured lipid A/LPS accumulation that is considered toxic to E. coli is not a single, fixed number, but rather the point at which the transport capacity of the cell is overwhelmed, causing the IM to accumulate excessive LPS. Any significant, sustained increase above normal physiological levels (roughly >10% of total lipids) typically causes stress, but the exact threshold for lethal toxicity is context-dependent based on the efficiency of the MsbA transport. The viability of E. coli is contingent on a tight balance between phospholipids and LPS, which is generally maintained at a ratio of approximately 1:0.15. The toxic threshold is reached when the homeostasis of LPS biosynthesis and transport (mediated by Lpt and MsbA) is compromised, leading to significant accumulation in the IM, which generally occurs in the range of 30–40% of membrane composition in mutant studies. Particularly the toxicity appears to occur when LPS levels exceed the capacity of the IM MsbA, inter IM and OM Lpt transport system and the ability of the OM itself to accommodate it, leading to accumulation in the IM. In LpxL/LpxM mutants that accumulate up to 40% nonmatured, tetra-acylated LPS in the IM, growth is inhibited (however only above 33–42 °C).²⁶ While typically destined for the OM, the presence of LPS in the IM can become structural either in cells growing normally (10%) or responding to environmental stress (>10%), where imbalances in synthesis or transport require regulated modification and rebalancing.²⁶ The same scenario could apply to DAG threshold amounts and distribution since unusually high abundance of DAG is detected in E. coli during aerobic-to-anaerobic transition.⁵²

When radiolabeled intact E. coli cells were treated with TNBS, which freely penetrates the OM but not IM, and therefore expected to label the periplasmic leaflet of the IM predominantly, we got an essentially the same distribution ratio confirmed that IM is resealed with full maintenance of PE asymmetry (Figure 1C of submitted MS versus ref 2). These results demonstrate that randomization or preferred leaflet localization of lipids does not accompany IM rupture and resealing of vesicles during high-pressure lysis.⁵³ The sum of the radioactivity in the PE and TNP-PE spots always remains nearly constant in cells and cell-derived uniformly oriented ISO vesicles (a 75%/22% (cytoplasmic/periplasmic leaflet)) distribution. We expected the same trend in the nonbilayer DAG distribution. The in vivo experiments with heterologously expressed S. meliloti diacylglycerol lipase DglA⁵⁴ that degrades endogenous DAG can help to confirm this assumption due to the demonstrated capability to degrade DAG in vivo, as evidenced by reduced DAG levels. However, this kind of experiment does face limitations due to the fact that this patatin-like bacterial lipase provoke unintended cell lysis, making them problematic for studying TM distribution of DAG in specific physiological states like stationary growth or osmotic stress.

DAG Dynamic Redistribution can Compensate for IM Leaflet Volume Imbalance

All-Atom Molecular Dynamics Simulations (MDS) demonstrate the ability of DAG to redistribute rapidly between the IM leaflets in response to an imbalance of the monolayer volumes (Figure 5). In simulation, this imbalance was modeled by depleting lipid A in the periplasmic leaflet by up to 30% of its initial value, which may mimic its trafficking to the OM via the Lpt system.

Figure 5.

Fast dynamic redistribution of DAG between cytoplasmic and periplasmic leaflets of IM revealed by MDS. (A) Snapshot of the balanced model membrane with 40 lipid A in the periplasmic leaflet. (B) Snapshot of the lipid A-depleted system with 28 lipid A molecules (30% depletion) in the periplasmic leaflet. Lipid A molecules are tan with cyan oligosaccharide moieties; PE lipids are blue; CL lipids are red; PG lipids are green; DAG molecules are pink and shown in the space-fill representation. Phosphorus atoms of the lipid phosphate groups are shown as large orange spheres. K⁺ and Cl⁻ ions are shown as small pink and blue spheres, respectively. DAG molecules redistribute into the periplasmic leaflet, filling the gaps left by depleted lipid A. (C) Visualization of the prevalent negative charge of the lipids in the periplasmic leaflet of IM due to the presence of PG and lipid A. Lipids are shown as lines. Phosphorus atoms of the lipid phosphate groups and phosphorylated lipid A sugars are shown as large orange spheres. K⁺ and Cl⁻ ions are shown as small red and blue spheres, respectively. (D) Distributions of the number of DAG molecules in the periplasmic leaflet depending on the depletion of lipid A. The total number of DAG molecules in the system is 30. (E) Number of DAG molecules in the periplasmic and cytoplasmic leaflets and in indeterminate orientation between the leaflets, depending on the depletion of lipid A. (F) Computed deuterium order parameters of the most abundant lipids in periplasmic and cytoplasmic leaflets depending on the depletion of lipid A. Order parameters for all chains of each lipid are averaged for clarity.

In the membrane with balanced leaflet volumes (Figure 5A), DAG is distributed roughly equally between them. When the lipid A depletion reaches 30% almost all DAG molecules migrate to the periplasmic leaflet to compensate for the volume imbalance. Figure 5D shows equilibrium distributions of the number of DAG molecules in the periplasmic leaflet depending on the depletion of lipid A molecules in this leaflet. It is clearly seen that in the initial system, on average, 10 out of 30 DAGs are located in the periplasmic leaflet. Subsequent depletion of lipid A by 10, 20, and 30% shifts the center of distribution to 15, 20, and 23, respectively. For the most depleted system all 30 DAGs may reside in the periplasmic leaflet in some trajectory frames (such as one shown in Figure 5B). Further depletion of lipid A beyond the 30%, when there are no more DAG molecules to redistribute and to compensate for volume imbalance, leads to formation of the gel phase in the lower monolayer (data not shown).

During simulations, DAG molecules sometimes reside in the middle of the membrane’s hydrophobic core, without a clear orientation of their glycerol moieties toward either leaflet. Figure 5E shows that the number of such molecules with indeterminate positioning is significant and may reach 20% (6 out of 30) in the balanced membrane and at small depletion values, decreasing gradually with increasing lipid A depletion in the periplasmic leaflet. It is clearly seen that DAG molecules from both the cytoplasmic leaflet and the middle of the bilayer are migrating to the periplasmic leaflet in response to lipid A depletion.

As shown in Figure 5F, there is a well-defined, systematic change in the order parameter upon lipid A depletion. The order of the periplasmic leaflet progressively decreases with lipid A depletion and DAG influx from the cytoplasmic leaflet. In contrast, the order of the cytoplasmic leaflet progressively increases with lipid A depletion and DAG efflux to the periplasmic leaflet.

Notably, flip-flops of DAG molecules happen continuously during the simulations following the rapid initial stage of filling the gaps left by removed lipid A (Figure S8). After ~200 ns of equilibration the number of DAG molecules fluctuates randomly around well-defined average values without any further significant systematic drift. It is clearly seen that individual replicas, started from the initial state with different lipid A molecules removed, differ from each other and may have significant random fluctuations (like replica #3 of 10% depletion) but eventually all converge into similar DAG distributions.

Compensatory effect of DAG is also evident from the quick equilibration of the system box size upon removal of lipid A molecules (Figure S9). These observations indicate that simulation time of 1 μs is enough to reach proper equilibrium and to sample the DAG flip-flops adequately. It is also evident that initial filling of the free volume left by removing lipid A molecules occurs very fast (within ~50 ns) even when transversal motion of DAG molecules is forbidden by applying restraints (“restr” curve in Figure S9). In this case the voids are filled by adjacent lipids and water molecules instead of immediate flow of DAGs. However, when the restraints are lifted after 100 ns the DAG molecules still redistribute in the same way and to the same amount as in the case without restraints. This proves that the inter-leaflet redistribution of DAG is not an artifact caused by an abrupt removal of lipid A from the periplasmic leaflet.

Our data suggest that DAG redistributes rapidly between the leaflets of IM and can compensate for a significant depletion of lipid A in the periplasmic leaflet (up to 30% in our simulations). Redistribution of DAG from cytoplasmic to periplasmic leaflet leads to an increase in the lipid ordering in the former and a decrease in the latter. Still, the membrane remains in a liquid-disordered phase and maintains structural integrity over a wide range of lipid A content.

This allows us to hypothesize that DAG molecules act like a highly mobile “sealing compound”, which compensates for a free volume in either of the leaflets when it appears due to physical or metabolic lipid imbalance.

It is important to note that modeling realistic kinetics of DAG redistribution is challenging and beyond the scope of this study. Although the simulated DAG flip-flop rate might be different from the experimentally observed one (and likely to be overestimated), this does not change the qualitative hypothesis about its role as a “sealing compound” that diffuses between the membrane leaflets much faster than the phospholipids and thus compensates the lipid deficiency in either of the leaflets.

Testing the Biogenic Origin of Glycerophospholipid Imbalance

The compositional and numerical asymmetric interleaflet distribution of phospholipids may serve specific functional roles, potentially governed by the competitive availability of lipid substrates or by the localization constraints on product deposition outside the periplasmic leaflet. Although observed TM glycerophospholipid abundance imbalance is normally required to maintain a long-term nonequilibrium state, the asymmetrical distribution of PE and PG in the IM can be assumed to be an unstable state which can be potentially dissipated or altered if “refill” and “removal” of these lipids on the cytoplasmic and periplasmic leaflets of the IM are changed, respectively.

The biosynthesis and growth requirements of the diderm cell envelope are determined by the competitive availability of lipid substrates and constraints on product deposition. Thus, IM asymmetry can be also driven by packing and localization constraints imposed during cell division and impacted by the rate of phospholipid biosynthesis and intermembrane flows. While PG is significantly enriched in the cytoplasmic leaflet of the IM in logarithmic-growth E. coli (60%/40% (cytoplasmic/periplasmic leaflet distribution)) (Figure 2), its asymmetrical distribution is inverted in stationery grown cells (Figure 6A) (32%/68%, cytoplasmic/periplasmic leaflet distribution), suggesting the existence of mechanism for maintaining IM TM homeostasis in cells transitioning from active growth (logarithmic phase) to a dormant state (stationary phase). Since TM distribution of PE was not affected, the increasing accumulation of highly charged (dianionic) CL in the membrane (hallmark of stationery grown cells) could necessitate a mechanism to balance the TM charge and physical stress. The cell most likely maintains such an asymmetric distribution of anionic lipids between the inner and outer leaflets of the IM to counteract the charge and packing stresses caused by this increased CL.

Figure 6.

Alteration of transmembrane phospholipid distribution in stressed cells. The steady state TM distribution of main phospholipids in stationery grown E. coli cells (A) or cells challenged to growth on hypotonic media (B). TLC analysis of TNBS modification and subsequent NaIO₄ oxidation and derivatization of oxPG by BtH of wild-type E. coli intact cells entering the deep stationary phase of growth or experiencing a shift from high to low osmolarity (as described in legend to Figure 2).

A low-osmolarity medium triggers a specific metabolic shift to maintain periplasmic turgor pressure and maintain the integrity of stretched IM. TM distribution of CL, PG, and PE can be rebalanced to maintain an electrostatic balance and integrity of the IM. In low-osmolarity media, E. coli synthesizes large amounts of MDOs/OPGs (more than 15 times more and up to 5% of dry cell weight) and accumulates them in the periplasm to balance osmotic pressure by creating an “osmotic buffer” and preventing lysis.^18,55 MDO biosynthesis involves taking sn-1-phosphoglycerol residues from PG and transferring them to the maturing highly branched glucose oligomers in the periplasm. Thus, PG is preferentially and continuously consumed from the outer (periplasmic) leaflet of the IM. Therefore, high MDO synthesis in low osmolarity should theoretically drive a high rate of PG turnover particularly on the outer (periplasmic) leaflet of the IM and favor its asymmetric TM redistribution across the IM (causing potential transient depletion at the periplasmic face compared to the inner leaflet?), as PG is converted to DAG to fuel MDO decoration. We decided to analyze the sidedness of PG and PE in cells challenged to grow on hypotonic media (Figure 6B). Surprisingly and counterintuitively, we detected much more PG on the periplasmic leaflet of the IM in steady-state labeling conditions (17%/83% (cytoplasmic/periplasmic leaflet) distribution).

Because MDO synthesis is so high, PG is rapidly degraded (high turnover) to supply the necessary phosphoglycerol residues for the periplasmic MDO. Surprisingly, however, the high-rate biosynthesis of MDO in low-osmolarity conditions, triggering a substantial rapid turnover of PG, does not reduce its steady-state levels, suggesting the existence of a mechanism which allows PG to accumulate or remain higher. Inverted steady-state asymmetry of PG in normal and osmotically challenged cells suggests that TM imbalance of PG could be served for purpose to keep TM distribution in nonequilibrium states and maintain a slow relaxation of bilayer phospholipid asymmetry at different growth conditions. While the cell utilizes a controlled, imbalanced glycerophospholipid TM distribution (numerical asymmetry) to function, it simultaneously uses homeostatic mechanisms to prevent this imbalance from becoming pathological under stress, bringing the lipid distribution toward a functional, regulated, yet still nonequilibrium, state. Beyond involving MDO, low osmolarity (hypotonic) down-shock causes a rapid influx of water and increases the cell volume. As cell volume increases, the cytoplasm pushes against the cell wall, placing massive mechanical tension on the IM, causing it to stretch. These events may also trigger specific transient changes in membrane lipid composition and remodeling to manage membrane tension. The tension in the lipid bilayer stretches open mechanosensitive channels (such as MscL and MscS).⁵⁶ Thus, low osmolarity can provoke the cells to redistribute PG toward the periplasmic leaflet, highlighting the existence of an active response to hypotonic conditions (low osmolarity) to manage membrane tension, regulate activity of mechanosensitive channels during cell swelling in order to prevent lysis.

Since the ratio of TNP-PE to PE remains constant during osmotic downshock (Figure 6B), the TM distribution of PE is maintained (75/25% (cytoplasmic/periplasmic leaflet) distribution) suggesting that the cell can actively compensate for osmotically related membrane stress and membrane tension by maintenance of TM distribution of PE.

The anionic nature of MDOs (due to the PG-derived phosphoglycerol) allows them to act as an “anion” in the periplasm, helping to maintain a Donnan potential and high osmotic pressure when external salt concentrations are low.⁵⁵ The maintenance of the net negatively charged periplasmic leaflet by asymmetric distribution of PG can act as an additional repulsive force to support the generation of the layer of “freely” swimming MDO molecules. Irrespective of the mechanism, the observed redistribution of PG may allow the cell to rebalance the charge distribution rapidly in response to changes in environmental osmotic stress or metabolic state in order to stabilize and maintain membrane flexibility and responsiveness, enabling the membrane to withstand sudden osmotic pressure and metabolic and lipid stresses. This points to a trade-off mechanism between structural integrity and electrostatic balance, both of which are required to maintain the appropriate compositional and numerical charge asymmetry that is likely essential for the functionality of membrane-bound proteins and overall envelope stress responses.

DISCUSSION

Although the OM of Gram-negative bacteria is a unique, asymmetric lipid bilayer containing phospholipids (primarily PE) in the inner (periplasmic) leaflet and lipopolysaccharides (LPS) in the outer leaflet, the IM is also believed to be asymmetric with respect to all three major phospholipids. However, both leaflets contain similar types of lipids at various concentrations.⁵⁷ We developed previously unavailable methods for determining the transbilayer distribution of PG using novel leaflet-specific vectorial molecular probes, which complement our recently published method for measuring PE bilayer asymmetry. We report adapting a simple chemical assay for oxidizing vicinal diol groups and covalently modifying them with a vectorial hydrazide probe to precisely determine PG pools specific to each leaflet in the IM of intact E. coli cells. Our new methodology first employs periodate (NaIO₄) oxidation to convert the central carbon of glycerol in the PG headgroup into an aldehyde (oxPG) and to selectively modify the total PG pool regardless of orientation (Figure 2A). Apparently, in our experiments, periodate easily permeates both the IM and OM, converting the entire PG pool to oxPG, while CL remains unmodified due to the absence of reactive vicinal diol groups. Cells remain intact as indicated by the established percentage of cytoplasmic PE protected from TNBS (Figure 2C and E). Extraction of total phospholipids from ³²PO₄-labeled cells after oxPG formation and treatment with IM-impermeable BtH was used to quantify the transbilayer distribution of oxPG (inner leaflet) and its betaine hydrazone derivative, Bt-PG (outer leaflet) (Figure 2A). This allows mapping and determination of PG and PE topography in each leaflet of the IM within the same radiolabeling experiment. The assay was further optimized by omitting acidic pH during extraction and using a neutral extraction system, since hydrazone bonds are labile and prone to breaking at low pH. Hydrazone bonds are acid-labile and undergo rapid hydrolysis in acidic environments (pH 6.5 and lower) where a water molecule attacks the C=N double bond to reform the original hydrazine/hydrazide and carbonyl (aldehyde) moieties. Potential formation of anionic N-formyl-PE, N-acyl-PE, or PE-ox-PG adducts during oxidation was also ruled out. Challenges related to the uncontrolled heterogeneity of undesired oxPG species and varying yields in recovering covalently modified oxPG from the chloroform-rich layer were successfully addressed. An acidic TCL system was found superior to an alkaline one due to the unexpected instability (hydrolysis) of the Bt-PG hydrazone bond, which is considered “labile” under alkaline conditions (Figure S4). While acidic conditions (protonation) facilitate hydrolysis, water is the necessary nucleophile for the actual bond cleavage. Since low pH triggers acid-catalyzed hydrolysis, water is excluded in our TLC solvent system for hydrazone compounds, to prevent hydrolysis during the separation process. By removing water from the mobile phase, the acid-labile hydrazone bond remains stable while moving on the TLC plate. Using a nonaqueous solvent system avoids the hydrolysis reaction during thin-layer chromatography, ensuring that the Bt-PG adduct can be analyzed intact. Our data demonstrates that BtH can serve as a new vectorial chemical probe, suitable for analyzing the transmembrane distribution of both head and acyl groups of PG (vectorial lipidomics). To characterize both headgroup and acyl asymmetry, the PG could be further analyzed by mass spectrometric approach to determine the headgroup and acyl asymmetry. This method is suitable for investigating transbilayer dynamics, steady-state sidedness, and the physiological significance of PG asymmetry in double-membraned Gram-negative bacteria, single-membraned Gram-positive bacteria, submitochondrial particles, single-membrane cellular or extracellular organelles (such as bacterial outer membrane-derived membrane vesicles), or liposomes.

While PE shows pronounced leaflet asymmetry dependent on cell shape,² it has been proposed that both anionic PG and CL tend to be enriched in the cytoplasmic leaflet of the IM and preferentially interact with membrane proteins.⁵⁰ CL, in particular, shows strong and specific binding to many IM proteins.⁵⁰ This asymmetric distribution is thought to be essential for functions like regulating membrane protein activity and maintaining membrane stability. Molecular dynamics simulations reveal a clear asymmetry between leaflets, with increased CL and PG binding on the inner side of these proteins.^50,58

By employing ³²PO₄-steady state labeling, the mol % of TNP-PE (periplasmic leaflet of IM), unmodified PE (cytoplasmic leaflet of IM), oxPG (cytoplasmic leaflet of IM), and betaine-PG (Bt-PG) (periplasmic leaflet of IM) was quantified. This study demonstrates that PG is asymmetrically distributed across the IM of intact E. coli, with an approximate ratio of 60%/40% (cytoplasmic/periplasmic leaflet) (15%/10% in absolute amounts) which is incompatible with the distribution of phospholipid mass PE and PG between the two leaflets of the IM. Independently and consistent with our findings, the lipid compositional asymmetry of E. coli was recently simulated using unequal amounts of PG in the bilayer leaflets, with 26.3% in the cytoplasmic leaflet and 12.8% in the periplasmic leaflet.⁵⁸ In the absence of “wet” data for PG distribution, compositional asymmetry of this negatively charged lipid was intentionally and voluntarily guided to match the “positive inside rule” postulating that net positively charged extramembrane domains reside at the cytoplasmic leaflet of the IM. The authors thought that this biased distribution of charges should require more phospholipids with negatively charged head groups present on this side of the membrane to “neutralize” opposite protein charges.^57,58 In addition, we found that the periplasmic leaflet of E. coli IM could be net negatively charged (Figure 5C), contrary to the mammalian cytoplasmic membrane, which indeed segregates its anionic lipids in the inner leaflet.^59,60 However, we currently have no explanation for the mechanisms underlying the generation of the asymmetry observed in the biogenic E. coli IM or for its physiological significance. To avoid a mass imbalance between the two membrane leaflets, rapid lipid flip-flop in the IM is required during cell growth and division. Obviously, a symmetrical and asymmetrical distribution would result whenever the net rate of outward movement of new lipid to the outer leaflet is either equal to or unequal to the rate of biosynthesis, respectively. Due to an interleaflet phospholipid abundance imbalance, the driving force for IM lipid asymmetry can arise from asymmetric packing order of two leaflets (Figure 7). The driving force for IM asymmetry can arise from the packing requirements imposed upon the bilayer system during cell division through disproportional distribution of nonbilayer prone lipids, PE, CL, DAG and hexa-acylated lipid A and due to existence of critical threshold amounts of lipid A and DAG, which may be compatible or not for localization in the same leaflet. Due to the overabundance of glycerophospholipids on the cytoplasmic leaflet and consistent with the thermodynamic sink notion (the underpopulated periplasmic leaflet), loose lipid packing and discontinuity of the monolayer structure of the periplasmic leaflet can increase the probability of phospholipids to flip spontaneously from one leaflet to the other leaflet by a retentive thermodynamic “slip-pop” mechanism.^1,3 Overabundance of glycerophospholipids on the cytoplasmic leaflet can create lipid concentration gradients and downhill flows to satisfy the envelope’s growth capacity. Thus, interleaflet phospholipid abundance imbalance may serve a functional purpose which could be determined by different rates of biosynthesis and translocation of different phospholipids on the cytoplasmic and periplasmic leaflets, respectively, and/or competitive availability of lipid substrate or deposition of its product outside of the periplasmic leaflet of IM (Figure 7).

Figure 7.

Biogenesis of Lpp linked to anabolism and catabolism of PE and PG in E. coli IM. An asymmetric phospholipid overabundance in the cytoplasmic membrane of E. coli is achieved by temporally unbalancing the net rates of synthesis and flip-flop, as well as the inter-IM and OM bidirectional flows, and by regulating biosynthetic precursors for the synthesis and modification of major envelope components: lipoproteins (Lpp), lipopolysaccharides, and an enterobacterial common antigen (ECA_PG). From a biosynthetic point of view, transient pools of PG and PE should be present and consumed on the periplasmic leaflet of IM to serve as metabolic precursors for Lpp maturation. PG serves as a diacylglycerol donor to the N-terminal cysteine thiol group of lipoproteins, a reaction catalyzed by the essential enzyme prolipoprotein diacylglyceryl transferase Lgt with its active site facing periplasm. PE continuously flows from the periplasmic leaflet of the IM to the periplasmic leaflet of the OM to build up the periplasmic leaflet of the OM during cell growth. Lnt protein transfers the palmitate moiety from the sn-1 position of PE to the N terminus of the diacylglycerol-linked Lpp precursor, resulting in a triacylated mature Lpp. At any given time, more and more PG and PE molecules are required at the outer leaflet of IM to generate mature triacylated lipoprotein precursor (Lpp) and provide material for the exclusive building of the PE-enriched periplasmic leaflet of the OM, respectively. PG provides its diacylglycerol-phosphate moiety to ECA_PG, which finally ends up at the cell surface tethered to the outer leaflet of OM. EptA (not shown), a phosphoethanolamine transferase, catalyzes continuous modification of newly synthesized lipid A core by adding a phosphoethanolamine moiety to lipid A in antibiotic- or cold-stressed cells. All these processes can create lipid concentration gradients and downhill flows, which should be coordinated with the rates of IM and OM growth to satisfy the envelope’s growth capacity under different environmental conditions.

While PG is significantly enriched in the cytoplasmic leaflet of the IM in logarithmically growth E. coli, its asymmetrical distribution is inverted in stationery grown cells, cells lacking CL, or when cells are subjected to growth in a hypotonic medium (Figure 6). Most likely the mechanisms governing the inverted PG asymmetric distribution by the “loss of function” (clsABC null) (Figure 2F) and “gain of function” (Figure 6A) of cls genes are different. PG asymmetry can be directly related to the increased amount of CL in the stationary phase through a still unknown mechanism where concentration of PG in the cytoplasmic leaflet of IM, coupled with stationary-phase stress, drives its preferential conversion to CL by any of available CL synthesizing enzymes. In stressed cells (like those in stationary phase where clsB and clsC are specifically upregulated), the log-phase asymmetry can be disrupted and induced synthases, particularly ClsB and ClsC, can act as a metabolic “drain,” converting PG into CL, at a faster rate to adapt membrane composition.

Under metabolic or osmotic stress, alterations in lipid synthesis in the IM and intermembrane flows can work together to minimize excessive numerical differences in the TM distribution of different glycerophospholipids to ensure membrane integrity and prevent cell death. While the cell utilizes a controlled, imbalanced glycerophospholipid TM distribution (numerical asymmetry) to function, it simultaneously uses homeostatic mechanisms to prevent this imbalance from becoming pathological under stress, bringing the lipid distribution back toward a functional, regulated, yet still nonequilibrium, state. This points to a trade-off mechanism between structural Integrity and electrostatic balance, which are required to maintain the appropriate compositional and numerical charge asymmetry, which is likely essential for the functionality of membrane-bound proteins and overall envelope stress responses.

The asymmetry in the IM cannot be considered without the maintenance of asymmetry in OM, the utilization of lipids as metabolic precursors, and the biogenesis of the diderm envelope. Theoretically, voids on the periplasmic leaflet of the IM can be continuously generated due to the utilization of PE and PG as precursors in the biosynthesis of lipoproteins, or the modification of lipid A (PE) if their “consumption” as precursors and transport to the OM happen much faster than the rate of biosynthesis and flip-flop across the IM (Figure 7). It is also possible that using a periplasmic leaflet as a less abundant/underpopulated membrane glycerophospholipid constituent is advantageous to the cell, as it can be more easily regulated enzymatically by consuming PG and PE as metabolic precursors. Therefore, dynamic lipid (re)distribution can be naturally balanced metabolically and thermodynamically and thus ‘passively’ instead of actively (by postulated but still unidentified flippases and floppases of E. coli glycerophospholipids) maintained.^1–3

Our finding of a highly asymmetric structure in the biogenic IM of E. coli, with glycerophospholipids overpopulating the cytoplasmic leaflet and phosphoglycolipids (lipid A) and neutral lipids overrepresented on the periplasmic leaflet, will revise the current view on the biogenic IM of Gram-negative bacteria. The IM is now viewed as a highly asymmetric structure with complex and multiple systems maintaining its lipid composition and organization (Figure 7). The dynamics across IM (Figure 7) and inter IM and outer membrane phospholipid transport systems^61,62 (well respected but not shown on Figure 7 for the sake of simplicity) for supplying or removing phospholipids to and from the outer membrane challenge previous assumptions about cytoplasmic membrane equilibrium.^2,3 The phospholipid movement between these membranes must also be coordinated with the transmembrane movement (flipping) of phospholipids across the IM’s bilayer itself (Figure 7). However, this asymmetry should be maintained by balancing the fluxes of specific phospholipids moving in the anterograde (from the IM to the OM) and retrograde (from the OM to the IM) directions. This coordination is essential to both generate the initial IM asymmetry and maintain the overall stability of the bilayer structures of both membranes.

We proposed that DAG and lipid A are “missing lipids” or structural factors along with the structural threshold of lipoproteins (Figure 7) and the underpopulated periplasmic leaflet of IM, which could contribute to counterbalancing an IM interleaflet imbalance when glycerophospholipids overpopulate the cytoplasmic leaflet of IM. All-atom molecular dynamics simulations showed that DAG can spontaneously translocate (act as a “space filler”) and be redistributed, shifting the equilibrium and regulating the distribution of lipid A. However, we also found that physiological amounts of lipid A and DAG are still not enough to fill the periplasmic leaflet to avoid a numerical imbalance. A most abundant lipoprotein, Lpp, could transiently contribute to counterbalance the underpopulated periplasmic leaflet of the IM. Recent studies have demonstrated that the glycolipid MPIase (Membrane Protein Integrase), which facilitates membrane protein insertion and translocation, is distributed asymmetrically across the inner membrane,^63–65 with a predominance on the periplasmic leaflet (Sawasato et al., ACS Omega 2026, 11 (18), 27230–27241. https://doi.org/10.1021/acsomega.6c01244). MPIase may also serve as a candidate for a space filler. Our findings underscore a significantly more intricate and dynamic Gram-negative diderm cellular envelope than previously understood, providing new avenues for developing novel antibiotics.⁶⁶ Interestingly, our data are reminiscent of recent findings demonstrating that the eukaryotic plasma membrane cytosolic leaflet contains >50% of the cell’s total amount of phospholipids, and cholesterol sustains interleaflet glycerophospholipid imbalances by accumulating in the outer phospholipid-poor leaflet.³⁶ Although the precise, in vivo, leaflet-by-leaflet distribution of glycerophospholipids remains an area of active research and debate,³⁴ the observed differences in interleaflet glycerophospholipid imbalances shared by prokaryotes and eukaryotes are consistent with architectural principles that guide most, if not all, living biological membranes. Historically and traditionally, the “lipid bilayer” was viewed as a “sandwich” structure regarding phospholipid quantity, even if the types of phospholipids differed (compositional asymmetry). A new paradigm centers on whether two leaflets actually contain the same number of phospholipid molecules or if one is “overpopulated” (numerical asymmetry, different total amounts of glycerophospholipids).

The lipid asymmetry in biogenic self-synthesizing and self-assembling bacterial IM could be metabolically controlled in order to maintain uniform bilayer growth and asymmetric TM arrangement by (re) balancing temporally the net rates of synthesis and flip-flop across the IM, inter IM and OM bidirectional flows and bilayer chemical and physical properties as spontaneous response rendering dedicated flippases which are normally required to maintain, nonequilibrium state and slow relaxation of bilayer asymmetry in the IM redundant in biogenic (ER and bacterial) membranes.

CONCLUSIONS

Membrane asymmetry is primarily associated with lipid distribution. Membrane asymmetry refers to the two sides of a membrane being structurally, physically, and functionally different. Barriers to understanding the physiological significance of universal membrane lipid bilayer asymmetries are a lack of methods to precisely map the TM distribution of individual phospholipids in vivo and measure their flip-flops. While the existence of TM compositional asymmetry is firm, the exact quantification of in vivo leaflet-by-leaflet distribution of glycerophospholipids remains an area of active research and debate.³⁴ The scarcity and uncertainty of information regarding individual lipid TM distribution in the Gram-negative IM are due to the complexity of the diderm envelope and to the lack of suitable methods and vectorial probes to analyze the asymmetry and dynamics of phospholipid components in intact cells. Only the transbilayer asymmetry of PE has been determined by us recently by utilizing radiometric, spectrophotometric, or mass-spectrometric methods.^2,3 We demonstrated that the distribution of major phospholipid PE in the IM of Gram-negative bacteria is highly asymmetric, dynamic, and cell shape-dependent.^2,3 PG and CL are a ubiquitous net negatively charged phospholipid that present in almost all bacterial and archaebacterial species. However, their distribution between the membrane monolayers in bacterial IM remained unknown for a long time. In this study, we addressed this knowledge gap and developed a new methodology for determining the transbilayer distribution of PG and CL in intact cells, which complements our method for measuring PE bilayer asymmetry. Remarkably, we found that CL is localized near exclusively in the periplasmic leaflet of the IM. When PE, PG, and CL in vivo mapping techniques were combined together, surprisingly and unexpectedly, we found that their transbilayer distribution is incompatible with the distribution of their total mass between the two leaflets of IM and these phospholipids are overpopulated in the cytoplasmic leaflet of the IM. This uneven distribution creates a large deficiency of glycerophospholipids in the periplasmic leaflet. This TM abundance imbalance can create a transient lipid concentration gradient and possibly a downhill flow until equilibrium is reached.

At present, we have no explanation as to why the biogenic E. coli IM displays such compositional and numerical asymmetries. We propose that biosynthesis and translocation across biogenic IM must be intrinsically coupled and compositional, numerical and physical asymmetries are inherently linked. The fundamental principle is that numerical asymmetry creates large lateral tension imbalances between the leaflets. Mismatches in lipid numbers between the two IM leaflets generate “differential mechanical stress”, causing the compressed cytoplasmic leaflet to become “stiffened”. An overpopulated cytoplasmic leaflet, which is enriched in PE, PG, and CL (Figures 1–3), leads to compressive stress (negative tension) in that leaflet and, consequently, positive tension (stretching) in the less-packed periplasmic leaflet of the IM. Bilayer coupling/asymmetry stress in the E. coli IM can provide a plausible explanation for managing membrane tension under different physiological conditions. The osmotically challenged cells may alleviate this stress via “lipid redistribution,” particularly moving PG from the compressed cytoplasmic leaflet to the underpopulated (and less packed) stretched periplasmic leaflet (Figure 2F and Figure 6). In filamentous cells, continued IM synthesis without equivalent cell division leads to an overpopulated, physically stressed inner leaflet (high glycerophospholipid density). As we know, PE is a cone-shaped lipid that tends to overpopulate in the cytoplasmic leaflet of the IM (75%/25% (cytoplasmic/periplasmic leaflet)) of rod-sized cells.^2,3 The high demand for PE in outer OM growth and LPS biosynthesis encourages the translocation of PE to the outer leaflet of the IM or its consumption, resulting in inverted PE asymmetry (25%/75% (cytoplasmic/periplasmic leaflet))^2,3 and effectively reducing the packing pressure stress on the cytoplasmic leaflet. While lipid TM redistribution can relax the stress, cells may normally maintain numerical asymmetry (Figure 2G) of the IM (with tighter cytoplasmic leaflet packing) to sustain specialized functional states and support a physiologically important nonequilibrium state. Maintaining numerical asymmetry ensures the membrane is prepared for rapid response to stress (including membrane expansion) by storing “latent” capacity for rearrangement, acting as a “just-in-case” buffer against sudden environmental or cellular changes.

Our findings have broader implications for understanding the mechanical constraints that govern the establishment of compositional, numerical, molecular (different combinations of acyls with different length and unsaturation within the same lipid class of molecules), physical (lipid order and packing) lipid asymmetries in biogenic and nonbiogenic cytoplasmic membranes.

Still much remains to be discovered and several controversies remain to be resolved. The results presented in this manuscript raise additional questions to be addressed in the future. Obviously, both thermodynamic equilibrium and kinetic principles should serve as the sole mechanisms for creating and maintaining the dynamic asymmetric organization observed in the biogenic cytoplasmic membrane of bacteria. An identification of the mechanism (flippase-free or flippase-guided) controlling the asymmetric distribution of lipids across the bacterial IM is a guiding factor for future research focused on diderm bacteria.

MATERIALS AND METHODS

Materials

Sodium periodate (NaOI₄) and betaine hydrazide (Girard’s Reagent T) were purchased from Sigma-Aldrich. TNBS and DFDNB were purchased from Thermo Fisher Scientific and Sigma-Aldrich, respectively. Silica gel TLC plates were purchased from Merck. [³²P]PO₄ was obtained from PerkinElmer, Inc. E. coli PE, E. coli PG, and 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1′-rac-glycerol) (POPG) were purchased from Avanti Polar Lipids. Mini-PROTEAN TGX Precast Gels were obtained from Bio-Rad. SuperSignal West Pico PLUS Chemiluminescent Substrate was obtained from Thermo Scientific.

Bacterial Strains and Growth Conditions

Strains W3110 (λ⁻, IN(rrnD-rrnE)1, rph-1), BKT12 (W3110 ΔclsA, ΔclsB, ΔclsC), and AL95 (W3899 pssA93::kan, lacY::Tn9) were utilized in this manuscript. The cells were outgrown in LB medium (10 g of Bactotryptone, 5 g of Bacto-yeast extract, and 5 g of NaCl/liter), which was supplemented with 50 mM MgCl₂ when PE-deficient (AL95) cells were grown. All cells were grown at 37 °C.

Side-Specific Covalent Modification of PE in Intact Cells

For the TNBS modification of PE, 15 mL of logarithmically grown cells outgrown in the presence of carrier-free ³²PO₄ (1 μCi/ml) were resuspended in 1 mL of 100 mM HEPES/KOH (pH 7.5), 80 mM NaCl, and 1 mM EDTA. The TNBS was added to give a final concentration of 1, 3, or 10 mM by dilution from a fresh 5% commercial stock solution (Sigma) in water, and the reaction was stopped after incubation for 60 min on ice by the addition of 50 mL of 1 mM l-lysine to inactivate and scavenge unreacted TNBS. For the preparation of pure TNBS-modified inner membranes (IM) by differential centrifugation, the cells were sonicated in the presence of l-lysine for 1 min on ice using an amplitude of 15%. After removal of undisrupted cells and the OM fragments by centrifugation at 30,000g for 10 min, the inner membrane fraction was collected at 150,000g using a Beckman Coulter TLA-100 ultracentrifuge equipped with TLA-55 rotor. The IM were resuspended in 0.1 mL of 100 mM HEPES (pH 7.5) and 50 mM l-lysine. After the addition of an equal volume (0.1 mL) of 0.5 M NaCl in 0.1 M HCl, the lipids were extracted by first adding 0.6 mL of chloroform/methanol (1:2) to create a single-phase solution. After vigorous vortexing for 30 min, 0.2 mL of 0.5 M NaCl in 0.1 M HCl was added to convert the single-phase solution to a two-phase solution. After centrifugation at 20,800g for 5 min at room temperature, the lower phase was recovered and used for TLC. Extracted phospholipids were resolved by activated TLC using HPTLC 60 silica gel plates (EMD, Gibbstown, NJ) and developed with a solvent consisting of chloroform/methanol/acetic acid (65:25:8, v/v). Radiolabeled lipids were visualized and quantified using a Typhoon FLA 9500 (GE). Stored images were processed and quantified using ImageQuant software (GE) for scanning and analysis of the captured phosphor images and expressed as mol % of total phospholipid (correcting for two phosphates per molecule of CL) based on the intensity of the captured signal on the Phosphor screen generated by the radiolabeled spots on the TLC plate.

Oxidation of Liposomal PG with Sodium Periodate

100 μL of SUVs composed of synthetic POPG, E. coli PG or a mixture of E. coli PE and PG (the ratio of 1:1 (w/w)) were resuspended in 100 mM HEPES/KOH (pH 7.5) at final concentration of 5 mg/mL and subsequently subjected to PG oxidation by adding of equal volume of 100 mM NaOI₄, followed by incubating for 30 min on ice. For oxPG derivatization, BtH was added to give a final concentration of 5 mM and then incubated for 1 h at 37 °C. Phospholipids were extracted by adding 0.6 mL of chloroform/methanol (1:2) to create a single-phase solution. After vortexing for 30 min, 0.2 mL of 50 mM HEPES/KOH (pH 7.5) was added to convert the single-phase solution to a two-phase solution. After centrifugation at 20,800g for 5 min at room temperature, the lower phase was recovered and used for TLC analysis. Extracted lipids were separated by TLC with solvent system 2: chloroform/methanol/ethanol/acetic acid (60/25/8, v/v/v).

Transmembrane Mapping of PG in Intact Cells

E. coli cells outgrown until OD₆₀₀ 0.5 with 1 μCi/ml [³²P]PO₄ in 1 mL of LB medium were collected by centrifugation and then resuspended in the same volume of 100 mM HEPES/KOH (pH 7.5), 80 mM NaCl, and 1 mM EDTA. For TNBS derivatization of PE, 1 mL of resuspended cells was incubated with 3 mM TNBS for 1 h on ice. NaOI₄ was added to give a final concentration of 50 mM to convert PG to oxPG, followed by 15 min incubation on ice. After PG oxidation, cells were collected by centrifugation and washed with 100 mM HEPES/KOH (pH 7.5), 80 mM NaCl, and 1 mM EDTA twice. For leaflet-specific oxPG derivatization, the washed cells were suspended in the same buffer, and BtH was added to give an indicated final concentration of 5 mM. After incubating for 1 h at 37 °C, the reaction was stopped by collecting the cells by centrifugation and following washing with 100 mM HEPES/KOH (pH 7.5), 80 mM NaCl, and 1 mM EDTA twice. Washed cells were resuspended in 50 mM HEPES/KOH (pH 7.5), and then lipids were extracted by adding 0.6 mL of chloroform/methanol (1:2) to create a single-phase solution. After vortexing for 30 min, 0.2 mL of 50 mM HEPES/KOH (pH 7.5) was added to convert the single-phase solution to a two-phase solution. After centrifugation at 20,800g for 5 min at room temperature, the lower phase was recovered and used for TLC. Extracted lipids were separated by TLC with solvent system 2: chloroform/methanol/ethanol/acetic acid (60/20/5/8, v/v/v/v).

Transmembrane Mapping of CL in Intact Cells

Wild-type cells (W3110) and its Δaas or ΔclsA derivatives harboring pBAD18-Cld1p were grown until early log phase with supplementation of 0.002% arabinose. Phospholipids were uniformly labeled by 1 μCi/ml [³²P]PO₄. After lipid extraction, individual phospholipids were separated by TLC using chloroform/methanol/water/ammonia (60/37.5/3/1, v/v/v/v), and then detected by Phosphoimager Typhoon FLA 9500.

Transmembrane Mapping of DAG Sidedness in the IM

For preparation of uniformly and oriented inside-out vesicles (ISOv), cells were uniformly labeled with [1-¹⁴C]acetate (0.1 mCi/ml) (American Radiolabeled Chemicals) after dilution to A₆₀₀ of 0.05 to initiate logarithmic growth, followed by growth until OD₆₀₀ of 0.5 to 0.6. ISOv were prepared from E. coli wild-type cells as described previously by rupture of cells using a French press at 560 kg/cm² (8000 psi) with subsequent differential or isopycnic centrifugations.^2,67 DAG sidedness was estimated from the extent of its phosphorylation by endogenous DgkA (e.g., amount of phosphatidic acid (PA)) before and after treatment of ISOv with detergent. To determine the maximal pool of PA samples were diluted in buffer containing octylglucoside (OG) at a final concentration of 40 mM. Thirty min after the addition of ATP, the reaction was terminated by pelleting and extraction procedure. The total neutral lipids (DAG and FA) and PA were extracted separately and combined extract was developed then on the same TLC with a solvent system that balances the migration of nonpolar lipids (DAG, FA) with more polar phospholipids (PA). Therefore, both substrate (DAG) and PA (product of kinase reaction) can be seen on the same TLC. The selective extraction of neutral lipids was achieved using a 0.25 mL of 25% ethyl acetate/2,2,4-trimethylpentane (iso-octane) solvent mixture (v/v). PA can be selectively extracted to high purity by manipulating the pH of the extraction medium, specifically involving a switch from alkalinized methanol (pH 8–9) (to saponify interfering lipids) to acidified methanol (pH ~ 6), with subsequent partitioning of PA into an organic phase (such as hexane). This protocol allows to isolate PA from complex phospholipid mixtures and reach the purity near 100%.⁶⁸ To selectively extract PA, inside-out vesicles were washed first with 95% ethanol at 45 °C for 2 h. The lipids were extracted then from pellets using methanol with vigorous shaking for 1 h. The methanol extract was pooled and the pH of the methanol was raised to 9.0 using 0.1 N NaOH. After centrifugation at 3000g for 10 min at room temperature, the supernatant was collected and the pellet was discarded. The pH of the methanol extract was then lowered to pH 6 using 0.1 N HCl. 1/4 volume of hexane (relative to methanol extract) was added, the solvents were mixed vigorously and left to stand still for 2 h or centrifuged at 1000g to separate the PA-containing hexane phase from methanol phase. Two extracts (DAG+PA) were combined together, concentrated under vacuum using a SpeedVac concentrator and loaded to TLC which was developed in stepwise manner.

The TLC plates were developed half way in chloroform/acetic acid/methanol/water (70/25/5/2, v/v/v/v), dried, and completely developed in petroleum ether/diethyl ether/acetic acid (70/30/1, v/v/v). The lipids were detected then by Phosphoimager Typhoon FLA 9500.

Transmembrane Mapping of Lipid A in the IM

E. coli cells heterologously expressing Francisella LpxE from the pBAD18-LpxE plasmid were grown in LB supplemented with 0.2% arabinose and 1 μCi/mL of [³²P]PO₄ (PerkinElmer) to an OD₆₀₀ of 0.5 to prepare ISOv. To isolate lipid A, E. coli cells were suspended in 0.8 mL of PBS. 1 ml of chloroform and 2 mL of methanol were added, and the mixtures were vigorously shaken at room temperature for 30 min to make the one-phase Bligh–Dyer mixture (chloroform/methanol/water at a 1:2:0.8 ratio) and extract and remove glycerophospholipids. Mixtures were then centrifuged at 5000 rpm for 30 min. Pellets were washed once with a fresh one-phase Bligh–Dyer solution (3.5 mL) and centrifuged as above. Resulting pellets were suspended in 1.0 mL of sodium acetate (12.5 mM, pH 4.5) and heated at 100 °C for 30 min. After cooling to room temperature, the suspension was acidified to pH 1 with HCl and converted to a two-phase Bligh–Dyer mixture (chloroform/methanol/PBS = 2:2:1.8 v/v) by adding 0.8 mL of PBS, chloroform, and methanol (2 mL each). The lower organic phase of each mixture was retrieved after phase partitioning via centrifugation at 5,000 rpm for 30 min and used for analysis. TLC analysis of lipid A phosphorylated derivatives was performed in a pyridine, chloroform, 88% formic acid, water (50:50:16:5 v/v). Plates were exposed to a phosphor screen, imaged, and the percentage of lipid species quantified by densitometry. Radiolabeled lipid A derivatives were visualized and quantified using a Typhoon FLA 9500 (GE) and ImageQuantTM software (GE).

Molecular Dynamics Simulations

The model of the IM was built using the following considerations. For simplicity, we assumed that all phospholipids have one palmitoyl (18:1) and one oleyl (16:0) tail (POPE, POPG, and POGL). Cardiolipin is assumed to have two palmitoyl (18:1) and two oleyl (16:0) tails (PVCL2 in the CHARMM-GUI terms). Lipid A is modeled as an E. coli type 1 with six lipid tails, R1 core, and no O-antigens (ECLIPA in the CHARMM-GUI terms, total charge −10).

Initially, the lipid content was expressed as a mass percentage in order to be comparable with experimental data.² We assumed that the phospholipids constitute 40% of the total membrane mass, while lipid A, DAG, and other components, such as proteins, account for the remaining 60%. According to experiments on the total phospholipids, PG accounts for ~6% in the inner leaflet and ~3% in the outer one. PE accounts for ~67% in the inner leaflet and ~7% in the outer one. CL accounts for up to a maximum of 17% of the total and is located in the cytoplasmic leaflet.² We further assumed that DAG may constitute up to a maximally possible 20% of the total IM mass (which is still biologically plausible, though this would represent a significantly elevated concentration beyond normal steady-state levels of the total membrane mass) and is initially distributed equally between the leaflets. Such experimentally determined distribution of the phospholipids creates a very large deficiency of lipids in the periplasmic leaflet, and it is not entirely clear what fills it (see Discussion). In the computational model, for the sake of simplicity, we filled the deficiency of lipids in the periplasmic leaflet with lipid A. We are aware that this is unlikely to be physiological, but this allows us to keep the model system simple and computationally tractable. In order to compute the required amount of lipid A, we first converted mass ratios to the molar ratios with a factor of 1 for PG, PE, and DAG, a factor of 2 for CL, and a factor of 3 for lipid A, following the number of their FA tails. Then we computed the areas of both leaflets using the areas per lipid from the CHARMM-GUI lipid database and adjusted the number of lipid A molecules to make them close to equal. It is necessary to note that the areas per lipid computed for the pure bilayers are not very accurate for such an unusual asymmetric membrane and thus only allow achieving rough correspondence of the leaflet areas. As a result, lipid A constitutes about ~30% of the total inner membrane mass in our model, exceeding significantly the physiologically relevant WT amount in the IM (up to 10%). While there may be a varied transient amount of lipid A in IM during this transport process, it does not constitute a significant portion of the total IM mass in WT cells. However, lipid A can accumulate in the IM of lpx mutants, which can have 4-fold more LPS compared with WT.²⁶ The assumption that DAG and lipid A could constitute up to 20–30% of the total IM is a useful parameter for a study examining the effects of their “active” concentrations on lipid homeostasis within IM. It also reflects a condition of high local accumulation, which moves beyond normal physiological concentration to a range known to cause significant biophysical changes in the membrane. These assumptions allow us to explore the upper limits of DAG’s and lipid A in E. coli IM. Our initial tests have shown that not more than 8% mole fraction CL is tolerated in the cytoplasmic leaflet. The gel phase is formed for larger CL amounts, which is unlikely to be physiologically relevant. That is why we limit ourselves to a lower CL content of ~5.6% to allow enough tolerance for lipid redistribution in the course of simulations. The resulting composition of the system is shown in Table S1.

The membrane was constructed with CHARMM-GUI Membrane Builder^69,70 and solvated with ~17700 TIP3P water molecules. The number of K+ and Cl− ions corresponded to a 0.15 M solution and was adjusted to neutralize the net negative charge of the bilayer. The system was energy minimized and equilibrated for 1 μs.

Then we modeled the effect of LipidA depletion in the periplasmic leaflet by randomly deleting four LipidA molecules (10% of the initial amount) at a time and re-equilibrating the system. This procedure was repeated resulting in the systems containing 40, 36, 32, and 28 lipid A molecules. For each depletion level three systems were generated by removing different lipid A molecules at random and generating random initial velocities from the Maxwell distribution. Each system was simulated for 1 μs. The last 500 ns of each trajectory were used for analysis.

In order to confirm that the transversal movement of DAG molecules is not an artifact caused by the voids from removed lipid A molecules, we performed an additional simulation with depletion from 40 to 36 lipid A molecules where Z coordinate of DAG glycerol moiety was restrained by the harmonic potential. The system was simulated with such restraints for 100 ns. After that the restraints were removed and simulation continued with free DAG molecules.

All MD simulations were performed using the program Gromacs⁷¹ version 2024.4 at a pressure of 1 atm maintained by the Parrinello–Rahman barostat.⁷² The velocity rescale thermostat⁷³ was used with a reference temperature of 300 K. The Verlet cutoff scheme was used.⁷⁴ Force-switch cutoff of the van der Waals interactions was used between 1.0 and 1.2 nm. Long-range electrostatics was computed with the Particle Mesh Ewald method⁷⁵ with the cutoff of explicit short-range electrostatic interactions at 1.2 nm. An integration step of 2 fs was used in all simulations, with the bonds to hydrogen atoms converted to rigid constraints using the LINCS algorithm.⁷⁵ The CHARMM36 force field is used.⁷⁶ Optimized charges for diacylglycerol were used to ensure its correct trans-leaflet mobility.⁷⁷ Analysis was performed with the custom scripts using the Python bindings to the MolAR molecular modeling library.⁷⁸ Positions of the DAG molecules were determined by the Z coordinate of the center of mass of their glycerol moiety, Z_gly, relative to the membrane midline Z₀ defined as Z coordinate of the center of mass of the terminal carbon atoms of all lipid tails. If Z₀ − Z_gly > 0.4 nm, then DAG is considered to be in the periplasmic leaflet; Z₀ – Z_gly < −0.4 nm then DAG is considered to be in the cytoplasmic leaflet; otherwise it is considered to reside between the leaflets without clear preferential orientation.

Statistical Analysis

The results of quantitative experiments are shown as means for independent experiments performed multiple times, as indicated. Values are means (±S.D.) from three experiments, where each determination was done in duplicate. Variation between duplicates was ±3%. The statistical significance of mean differences was assessed by means of the two-tailed Student’s t test. Images represent three independent experiments.

Data Availability Statement

The presented data sets are available from the corresponding author upon request.

All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supporting Information. Bacterial strains and plasmids can be provided by a completed material transfer agreement.