Phase separation in the outer membrane of Escherichia coli

Abstract

Gram-negative bacteria are surrounded by a protective outer membrane (OM) with phospholipids in its inner leaflet and lipopolysaccharides (LPS) in its outer leaflet. The OM is also populated with many β-barrel outer-membrane proteins (OMPs), some of which have been shown to cluster into supramolecular assemblies. However, it remains unknown how abundant OMPs are organized across the entire bacterial surface and how this relates to the lipids in the membrane. Here, we reveal how the OM is organized from molecular to cellular length scales, using atomic force microscopy to visualize the OM of live bacteria, including engineered Escherichia coli strains and complemented by specific labeling of abundant OMPs. We find that a predominant OMP in the E. coli OM, the porin OmpF, forms a near-static network across the surface, which is interspersed with barren patches of LPS that grow and merge with other patches during cell elongation. Embedded within the porin network is OmpA, which forms noncovalent interactions to the underlying cell wall. When the OM is destabilized by mislocalization of phospholipids to the outer leaflet, a new phase appears, correlating with bacterial sensitivity to harsh environments. We conclude that the OM is a mosaic of phase-separated LPS-rich and OMP-rich regions, the maintenance of which is essential to the integrity of the membrane and hence to the lifestyle of a gram-negative bacterium.

Keywords: atomic force microscopy; gram-negative bacteria; outer membrane; phase separation.

Fig. 1.

The OM contains a dense, crowded network of trimeric porins. (A) Large AFM phase scans show a MG1655 cell at low resolution, and (B) images of the nanoscale architecture of the entire OM can be produced by superimposing small, high-resolution phase images. (C) Enlarged phase and height images of the region marked by the dashed box in A and B show the OM covered by a network of ∼8-nm-wide pores. (D) Western blot showing variation in the levels of expression of OmpF and OmpC by the removal of ompR and its reintroduction on an inducible plasmid. (E) Number of pores per square micrometer detected in AFM images, showing that removal of ompR leads to the disappearance of the pores. Subsequent reintroduction of ompR leads to an increase in pores with OmpF and OmpC expression. Each data point corresponds to one cell with at least three independent experiments for each condition. (F) Typical phase images used for the quantification in E. [Scale bars: (A) 500 nm and (C and F) 50 nm.] Color phase (measured in degrees [deg]) and height scales are (A) 7 deg, (B) 1.5 deg, (C) 1.5 deg and 5 nm, and (F) 2 deg, 2 deg, 1 deg, 2 deg, and 1 deg. ns = P > 0.5.

Fig. 2.

Within the trimeric porin network, distinct pore-free patches that behave as liquid phases can be seen. (A) AFM phase image with patches highlighted by dashed lines. (B) Height image of the same area, showing that the patches protrude by about 1 nm. These regions are also extremely smooth, with height variations of less than 0.5 nm. (C) At timescales consistent with cell division, under these experimental conditions, patches merge, grow, and split apart. (D) Schematic of OmpF labeling by colicin N^1-185mCherry. Phase and height images of the same area are used to independently localize patches and labels, respectively. Quantification of the labels per area shows that OmpF colocalizes with the pore network. (E) OmpA is labeled by expressing ompA with a streptavidin binding peptide in an outer loop and adding streptavidin. Quantification of the labels per area shows that OmpA also colocalizes with pore networks. Each data point corresponds to a single image, in which images were recorded from three independent experiments with at least one cell per experiment. [Scale bars: (B and E) 100 nm and (C) 50 nm.] Color (phase/height) scales are (A) 1.5 deg, (B) 5 nm, (C) 1.5 deg, (D) 1.5 deg and 5 nm, and (E) 0.3 deg and 5 nm. ** = P < 10⁻² and *** = P < 10⁻⁴ from a paired two-way Student’s t test.

Fig. 3.

Patches are LPS-enriched domains. (A) Western blot showing changes in LPS levels. (B) For low LPS levels (lpxC101), the cell area covered by patches is significantly smaller than for high LPS levels (lpxC_R230L). Reintroduction of O-antigen, and hence longer LPS (+wbbL), results in this area being almost twice that measured for WT (MG1655). Data were recorded in at least three independent experiments per condition; each data point represents one cell. (C) Longer LPS chains result in larger patches, and measurements for lower LPS expression suggest smaller patches. (D) Patch morphology (here quantified by the aspect ratio) does not noticeably vary with LPS expression. Each data point represents an individual patch from cells used in B. (E) Typical phase images used to quantify B–D. (Scale bar: 50 nm.) Color (phase/height) scale is 1.5 deg. * = P < 0.05 and ** = P < 10⁻².

Fig. 4.

Outer-leaflet phospholipids lead to the formation of new domains. (A) AFM phase and height images of cells with mutations that disrupt lipid asymmetry in the OM. (B) Whole-cell phase images of an MG1655 and a ΔpldA ΔmlaA cell showing the extent of membrane reorganization with abundant phospholipids. (C) Height profiles of dashed lines in the AFM images in A. (D) For ΔpldA ΔmlaA cells, a significantly larger fraction of the bacterial surface is covered by pore-free patches of either type, compared with WT and single mutants. Data were recorded in at least three independent experiments per condition; each data point represents one cell. (E) The mean area of each individual patch varies. ΔpldA ΔmlaA cells also have a greater spread of patch sizes. Each data point represents an individual patch from cells used in D. (F) The mean aspect ratios of ΔpldA ΔmlaA cells is higher than single mutants; an example of an elongated patch can be seen in A. [Scale bars: (A) 50 nm and (B) 200 nm.] Color (phase/height) scales are (A) 0.75 deg and 5, 4, 5 and 5 nm. * = P < 0.05 and *** = P < 10⁻⁴.