Introduction

Biological membranes are complex arrangements predominantly composed of lipids and both integral and surface-attached proteins (Nicolson, 2014). The primordial function of biological membranes was likely to act as a simple, semipermeable diffusion barrier separating the cell from its environment, and genomes from each other (Chen & Valde, 2010). Later, membranes and membrane proteins evolved to fulfil a multitude of cellular functions including transport, respiration and morphogenesis. Since the physicochemical state of biological membranes is highly sensitive to changes in the environment including temperature, osmolarity, salinity, pH or diet (Razin, 1967; Hazel, 1995; Ernst et al, 2016), careful homeostatic regulation of key membrane parameters such as thickness or fluidity is vital for cell function (Parsons & Rock, 2013; Ernst et al, 2016; Harayama & Riezman, 2018; Levental et al, 2020).

Arguably, the best studied membrane-adaptive process is homeoviscous adaptation that acts upon changes in temperature (Hazel, 1995; Parsons & Rock, 2013; Ernst et al, 2016). With increasing temperature, lipid bilayers exhibit reduced head group packing, increased fatty acid disorder and increased fluidity in terms of increased rotational and lateral diffusion of molecules (Chapman, 1975; Heimburg, 2007). While all of these membrane parameters are closely interconnected, membrane fluidity is thought to be the key property actively maintained at stable levels that optimally support vital membrane functions through homeoviscous adaptation processes (Hazel, 1995; Parsons & Rock, 2013; Ernst et al, 2016). All living organisms achieve this by actively adapting their lipid composition. Most commonly, this is obtained by altering the content of lipids carrying fluidity-promoting unsaturated fatty acids (UFA) or branched chain fatty acids (BCFA) and fluidity-reducing saturated fatty acids (SFA), respectively, thereby counteracting shifts in membrane fluidity (Hazel, 1995; Diomandé et al, 2015; Ernst et al, 2016).

While adaptive changes in lipid fatty acid composition as well as the regulatory processes involved are increasingly well characterised (Mansilla et al, 2004; Ernst et al, 2018; Ballweg et al, 2020), the cellular consequences of inadequate membrane fluidity are significantly less understood. Sufficiently high membrane fluidity has been implicated in promoting folding, catalytic activity and diffusion of membrane proteins (Lee, 2004; Andersen & Koeppe, 2007). Too high membrane fluidity, in turn, has been shown to increase proton permeability in vitro (Rossignol et al, 1982; van de Vossenberg et al, 1999), thus potentially hampering with efficient ion homeostasis and energy conservation (Valentine, 2007), while too low membrane fluidity impedes respiration due to reduced ubiquinone diffusivity (Budin et al, 2018). However, our understanding of the behaviour of biological membranes upon changing fluidity is predominantly based on in vitro and in silico studies with simplified model lipids, or in vitro studies with either natural lipid extracts or isolated membranes (Baumgart et al, 2007; Schäfer et al, 2011; Nickels et al, 2017).

One of the fascinating properties of lipids is their ability to undergo phase transitions between distinct configurations that differ in terms of ability to form bilayers, membrane thickness and degree of lipid packing (Chapman, 1975). The biologically relevant bilayer-forming lipid phases are: (i) the liquid-disordered phase characterised by low packing density and high diffusion rates that forms the regular state of biological membranes, (ii) the cholesterol/hopanoid-dependent liquid-ordered phase that forms nanodomains (lipid rafts) found in biological membranes, both representing different fluid phases; and (iii) the gel or solid phase characterised by dense lipid packing with little lateral or rotational diffusion, which is generally assumed to be absent in biologically active membranes (Veatch, 2007; Sáenz et al, 2015; Schmid, 2017). In fact, the temperature associated with gel phase formation has been postulated to define the lower end of the temperature range able to support vital cell functions (Drobnis et al, 1993; Ghetler et al, 2005; Burns et al, 2017), and maintaining biological membranes in the correct phase (homeophasic regulation) has been suggested as an alternative rationale behind temperature-dependent lipid adaptation (Hazel, 1995). Finally, lipid phases can co-exist, resulting in separated membrane areas exhibiting distinctly different composition and characteristics (Baumgart et al, 2007; Elson et al, 2010; Heberle & Feigenson, 2011; Nickels et al, 2017; Shen et al, 2017). This principal mechanism of lipid domain formation is best studied in the context of lipid rafts (Lingwood & Simons, 2010; Shaw et al, 2021). Here, the co-existence of fluid liquid-disordered and liquid-ordered phases, and the associated protein segregation, has been demonstrated in membranes of living eukaryotic cells (Toulmay & Prinz, 2013; preprint: Shelby et al, 2021). In contrast, comprehensive in vivo studies on gel-fluid lipid phase separation in live cells have been challenging due to the tendency of cholesterol/hopanoids to suppress gel-fluid phase transitions (Heberle & Feigenson, 2011), and due to the difficulty to modify the membrane fatty acid composition and, thus, fluidity without inducing lipotoxicity (Shen et al, 2017; Budin et al, 2018).

While in vitro and in silico approaches with simplified lipid models have provided detailed insights into the complex physicochemical behaviour of lipid bilayers, testing the formed hypotheses and models in the context of protein-rich biological membranes is now crucial. Bacteria tolerate surprisingly drastic changes in their lipid composition and only possess one or two membrane layers as part of their cell envelope. Consequently, bacteria are both a suitable and a more tractable model to study the fundamental biological process linked to membrane fluidity and phase separation in vivo.

We analysed the biological importance of membrane homeoviscous adaptation in Escherichia coli (phylum Proteobacteria) and Bacillus subtilis (phylum Firmicutes), respectively. These organisms were chosen due to their prominence as Gram-negative and Gram-positive model organisms, and the different archetypes of membrane fatty acid composition (straight versus branched chain fatty acids) they represent. We have established protocols that allow the fatty acid composition of both organisms to be progressively altered and the cellular consequences to be directly monitored in growing cells. This approach allowed us to address three central questions linked to homeostatic regulation of membrane composition and fluidity: (i) what are the cellular consequences of an inadequate level of membrane fluidity that necessitate the extensive and conserved homeostatic regulatory processes, (ii) how do changes in lipid fatty acid composition translate to changes in membrane fluidity of living cells and (iii) what is the lipid phase behaviour in living cells with protein-crowded membranes and intact lipid domain organisation?

Our results demonstrate that too low membrane fluidity results in growth arrest in both organisms, which is accompanied by severe disturbances of the cell morphogenesis and ion homeostasis. Furthermore, too low fluidity triggers a striking, large-scale lipid phase separation into liquid-disordered and gel phase membranes, accompanied by segregation of otherwise disperse membrane proteins such as ATP synthase and glucose permease. Our results revealed that phase separation between liquid-disordered and gel state membranes is associated with loss of essential membrane functions, thereby limiting the range of membrane fluidity able to support life. At last, our findings demonstrating that gel-liquid phase separation and associated membrane protein segregation indeed occurs in protein-crowded, native plasma membranes of living cells, are fully consistent with the comparable phenomena observed in in vitro and in silico model systems (Baumgart et al, 2007; Veatch, 2007; Lingwood & Simons, 2010; Schäfer et al, 2011; Domański et al, 2012). Thus, the results provide strong in vivo support for the general validity of the respective models.

Results Depletion of BCFAs in B. subtilis

To modify the fatty acid composition in B. subtilis, we constructed a Δbkd Δdes deletion strain (Appendix Table S1). The bkd operon encodes enzymes catalysing the conversion of branched chain amino acids into intermediates for BCFA synthesis (Debarbouille et al, 1999). The lack of this activity can be complemented by supplementation with precursors 2-methylbutyric acid (MB) or isobutyric acid (IB) (Kaneda, 1977; Boudreaux et al, 1981). This provides the experimental means to control the lipid iso- and anteiso-BCFA composition (Appendix Fig S1A) normally responsible for the homeostatic adaptation of membrane fluidity in response to environmental changes (Diomandé et al, 2015). In addition, the strain is deficient for the lipid desaturase Des to prevent rapid adaptation of membrane fluidity by converting SFAs or BCFAs into UFAs (Diomandé et al, 2015). In the remaining text, the B. subtilis strain is labelled “Δbkd” for simplicity.

We compared growth of B. subtilis 168 used as wild-type (WT) and Δbkd cells at 37°C upon supplementation with BCFA precursors MB or IB (Fig 1A). While BCFA precursors had little impact on growth of WT cells, the auxotrophic Δbkd strain only grew in the presence of either of the precursors. Corresponding fatty acid analyses revealed large shifts in the composition of the Δbkd strain depending on the supplied precursor (Fig 1B and Appendix Fig S1B). As expected, cells supplemented with MB exhibited a high content (77%) of anteiso-BCFAs, whereas cells grown with IB showed a high content of iso-BCFAs (77%). To obtain cells depleted for both BCFA types, cells were grown in the presence of IB, followed by wash and incubation in precursor-free (PF) medium. This precursor depletion leads to growth arrest after about 90 min (Appendix Fig S1B), corresponding to an accumulated SFA content of ~50% (Fig 1B).

Figure 1. Membrane fatty acid composition-dependent growth of B. subtilis and E. coli

Growth of B. subtilis WT and fatty acid precursor-auxotrophic Δbkd cells in medium supplemented with precursor 2-methyl butyric acid (MB), isobutyric acid (IB) or grown precursor-free (PF). Fatty acid composition of B. subtilis WT cells grown in PF medium, and Δbkd grown with MB, IB or depleted for precursor for 90 min (IB→PF). For detailed analyses, see Appendix Fig S1B. Temperature-dependent growth of B. subtilis Δbkd on solid medium in serial 10-fold dilutions. For comparison between WT, Δdes and Δbkd Δdes cells, see Appendix Fig S1C. Temperature-dependent growth behaviour of E. coli WT and fabA(Ts), including a shift from 30 to 37°C as non-permissive temperature of fabA(Ts). Fatty acid composition of E. coli WT and fabA(Ts) cells grown at 30°C and shifted to different temperatures for 120 min. For detailed analyses, see Appendix Fig S2B and C.

Data information: (A) The diagram depicts mean and standard deviation (SD) of technical triplicates for each strain. (B, E) The histograms depict mean and SD of biological triplicates. (C, D) The experiments are representative of three independent repeats. CFA, cyclopropane fatty acid. Strains used: (A–C) B. subtilis 168, HS527; (D, E) E. coli MG1, MG4 (strains Y-Mel and UC1098, respectively, additionally encoding fluorescent ATP synthase (FOF1a-mNG)).

Source data are available online for this figure.

Analyses of Δbkd cells grown at different growth temperatures with different BCFA precursors (Fig 1C and Appendix Fig S1C) indicated that only MB, the precursor for anteiso-BCFAs, is capable for supporting robust growth at low temperatures (22°C). At 30°C and 37°C, growth was comparable in the presence of either MB or IB, while no growth was observed in the absence of precursors, demonstrating that a high BCFA level is essential for growth under these conditions. At 45°C, Δbkd could grow at low dilutions even in the absence of BCFA precursors (Fig 1C and Appendix Fig S1C). For these reasons, we chose growth with IB at 37°C, as the reference condition for B. subtilis Δbkd.

Depletion of UFAs in E. coli

In E. coli, membrane fluidity is modulated by UFAs (Parsons & Rock, 2013) (Appendix Fig S2A). While synthesis of UFAs is essential for E. coli growth, a temperature-sensitive fabF fabA(Ts) mutant (Appendix Table S1), in which a shift to non-permissive growth temperatures leads to UFA depletion, has been isolated (Cronan & Gelmann, 1973). DNA sequencing of fabF and fabA from this rather old isolate revealed that FabF (β-ketoacyl-ACP synthase II) is non-functional due to S291N and G262S substitutions. FabA (β-hydroxyacyl-ACP-dehydratase/isomerase) carries a single G101D substitution (Rock et al, 1996). Based on the structure of the FabA head-to-tail homodimer (Nguyen et al, 2014), G101 is positioned at the border of the dimerisation interface. Consequently, the G101D substitution could plausibly cause thermosensitivity by destabilising the essential dimer structure of FabA at elevated temperatures, thereby provoking the gradual depletion of UFA (Cronan & Gelmann, 1973). Thus, this strain provides the experimental tool to control the UFA/SFA balance in growing cells. Throughout the text, the strain is labelled “fabA(Ts)” for simplicity.

While the growth of fabA(Ts) at 30°C is comparable to E. coli Y-Mel used as WT, transfer to non-permissive temperatures such as 37°C only supported growth for about 120 min, followed by growth arrest and onset of cell lysis (Fig 1D). Corresponding fatty acid analyses confirmed a strong, temperature-dependent decrease in UFAs (Fig 1E and Appendix Fig S2). In agreement with Cronan and Gelmann (1973), a minimal amount of 10–15% UFAs appeared to be essential to support growth (Fig 1E and Appendix Fig S2B). In comparison, WT cells showed only minor, temperature-dependent changes in fatty acid composition caused by homeoviscous adaptation towards increased SFA content at higher temperatures (Fig 1E and Appendix Fig S2C).

Reduced membrane fluidity in cells depleted of UFAs or BCFAs

To confirm that changes in fatty acid composition translate to shifts in in vivo membrane fluidity, we monitored changes in steady-state fluorescence anisotropy of 1,6-diphenyl-1,3,5-hexatriene (DPH), the rotational freedom of which is sensitive to acyl chain disorder and, thus, indirectly to the fluidity of lipid bilayers (Lentz, 1993). DPH anisotropy measurements with B. subtilis Δbkd revealed the highest membrane fluidity for cells with the highest anteiso-BCFA content (Fig 2A). Cells with high iso-BCFA content exhibited membrane fluidity levels slightly lower than those found for WT. These results confirm that anteiso-BCFAs promote higher membrane fluidity than the corresponding iso-forms in vivo; a difference previously based on in vitro evidence only (Lewis et al, 1987). The changes observed upon depletion of BCFAs, which is accompanied by accumulation of SFAs, were more drastic and resulted in a gradual reduction of membrane fluidity, ultimately leading to growth arrest (Fig 2A and Appendix Fig S1B).

Figure 2. Reduced membrane fluidity in cells depleted for UFAs or BCFAs

DPH anisotropy of B. subtilis WT or Δbkd cells supplemented either with MB, IB or depleted for precursor (IB→PF) for the times indicated. High DPH anisotropy indicates low membrane fluidity. DPH anisotropy of E. coli WT cells grown steady state at 30, 37°C or shifted from 30 to 37°C followed by immediate measurement. In addition, DPH anisotropy of fabA(Ts) cells grown steady state at 30°C or shifted from 30 to 37°C followed by measurement at the times indicated.

Data information: (A, B) The experiments are representative of three independent repeats. The histograms depict means and SD of technical triplicates, together with P values of an unpaired, two-sided t-test. Significance was assumed with ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, n.s., not significant. Strains used: (A) B. subtilis 168, HS527; (B) E. coli Y-Mel, UC1098.

Source data are available online for this figure.

DPH anisotropy measurements conducted with E. coli followed a similar trend (Fig 2B). Both E. coli WT and fabA(Ts) cells grown at 30°C exhibited an expected, immediate increase in membrane fluidity upon a shift to 37°C; a phenomenon that in WT cells is overtime counteracted by homeoviscous adaptation restoring membrane fluidity close to pre-shift levels. However, in fabA(Ts) continued growth at 37°C resulted in a gradual increase in DPH anisotropy, thus confirming a substantial reduction of membrane fluidity.

In conclusion, the established fatty acid depletion procedures allow membrane fluidity to be controllably lowered to a point incapable of supporting growth in both organisms. In the following chapters, we use this approach to analyse which cellular processes are impaired by inadequate levels of membrane fluidity.

Consequences of low membrane fluidity on membrane diffusion barrier function

The prevalence of adaptive mechanisms maintaining membrane fluidity (Hazel, 1995) might indicate its importance for preserving the fundamental membrane barrier function. To analyse the consequences of too low membrane fluidity on membrane leakiness, we used the combination of two fluorescent dyes. Sytox Green is a membrane-impermeable, DNA-intercalating dye used to assess the integrity of bacterial plasma membranes in terms of permeability (Roth et al, 1997). DiSC3(5), a voltage-sensitive dye accumulating in cells with high membrane potential (te Winkel et al, 2016), indicates changes in either membrane ion conductivity or respiration.

Growing B. subtilis Δbkd cells, irrespectively of the supplied BCFA precursor, exhibited DiSC3(5) fluorescence signals comparable to those observed for WT (Fig 3A and B, and Appendix Fig S3A–C). This indicates that the corresponding changes in the membrane fatty acid composition and fluidity had surprisingly little impact on membrane potential. In contrast, depletion of BCFAs triggered a gradual membrane depolarisation that was, in a mild form, already detectable after 30 min. A complete membrane depolarisation was observed after 90 min coinciding with growth arrest (Appendix Fig S1B). However, membranes remained impermeable for Sytox Green (Fig 3A), demonstrating that the gradual membrane depolarisation was not caused by a simple disruption of membrane continuity. In contrast, even the severely BCFA-depleted membranes were fully capable of forming a continuous, tight diffusion barrier.

Figure 3. Consequences of low membrane fluidity on membrane diffusion barrier function

Images of B. subtilis WT and Δbkd cells co-labelled with the membrane potential-sensitive dye DISC3(5) and the membrane permeability indicator Sytox Green. Membrane properties were assessed for Δbkd cells grown in the presence of MB, IB or washed precursor-free (IB→PF) for the times indicated. As controls, WT cells were measured in the presence of depolarising antimicrobial peptide gramicidin ABC (gABC) or pore-forming lantibiotic Nisin. For cross-correlation between membrane depolarisation and membrane permeabilisation, see Appendix Fig S3A–C. Quantification of DISC3(5) fluorescence for cells (n = 100–142) depicted in panel A. Median represented by red line. Images of E. coli WT and fabA(Ts) cells co-labelled with the same indicator dyes as in panel A. Membrane properties were assessed for fabA(Ts) at 30°C and upon transfer to non-permissive 37°C for the times indicated. As controls, WT cells were incubated with the pore-forming antibiotic Polymyxin B (PolyB). For cross-correlation between membrane depolarisation and membrane permeabilisation, see Appendix Fig S3D and E. The integrity of the diffusion barrier function was additionally studied via ONPG permeability in a ΔlacY background (see Fig EV1). Quantification of DISC3(5) fluorescence for cells (n = 76–141) depicted in panel C. Median represented by red line.

Data information: (A–D) The experiments are representative of three independent repeats. (B, D) Red lines indicate the median. P values represent the results of unpaired, two-sided t-tests. Significance was assumed with ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, n.s., not significant. (A, C) Scale bar, 3 µm. Strains used: (A, B) B. subtilis 168, HS527; (C, D) E. coli Y-Mel, UC1098.

Source data are available online for this figure.

High DiSC3(5) fluorescence signals and, thus, high membrane potential levels were also observed both for E. coli WT and fabA(Ts) grown at the permissive temperature of 30°C (Fig 3C and D, and Appendix Fig S3D and E), whereas depletion of fluidity-promoting UFA in E. coli fabA(Ts) triggered a gradual loss of membrane potential. However, compared to B. subtilis the loss was delayed and incomplete (Fig 3D). Again, the lack of Sytox Green staining revealed that the membranes were not impaired in their general diffusion barrier function (Fig 3C). To confirm this important finding, membrane permeability was also monitored by following LacZ-dependent hydrolysis of the chromogenic substrate ONPG (ortho-nitrophenyl β-D-galactopyranoside) in cells deficient for the uptake system LacY. While we were able to detect low level, LacZ-dependent ONPG hydrolysis in a lacY deletion background, no difference was observed between cells with native fatty acid composition and cells strongly depleted for UFAs undergoing lipid phase separation. Hence, the results provide an independent control for the lack of membrane permeabilisation (Fig EV1).

Click here to expand this figure.

Figure EV1. Very low membrane fluidity in E. coli does not trigger membrane permeabilisation for ortho-nitrophenyl β-D-galactopyranoside (ONPG)

The membrane permeability of ONPG was assessed in strains deficient for the active uptake system LacY simultaneously expressing lacZ from a plasmid-encoded leaky Ptac promoter (without addition of the inducer IPTG). The graph depicts ONPG hydrolysis rates measured for intact cells upon incubation at 30 and 37°C. In case of the temperature-sensitive fabA(Ts) strain, growth at the non-permissive temperature of 37°C was limited to 120 min. As controls, the ONPG conversion rates were measured in strains lacking the lacZ-expressing plasmid. Note the lack of significant difference in ONPG conversion rates upon strong depletion of unsaturated fatty acids (fabA(Ts) strain at 37°C for 120 min), implying the lack of detectable membrane permeabilisation due to phase separation.

Data information: The graph depicts mean and SD of biological triplicates. The P values represent the results of unpaired, two-sided t-tests. Insignificant changes (P > 0.1) are indicated with n.s. Strains used: E. coli Y-Mel.ΔlacY, UC1098.ΔlacY, Y-Mel.ΔlacY/pTM30.lacZ-His2, UC1098.ΔlacY/pTM30.lacZ-His2.

Source data are available online for this figure.

In summary, while membrane depolarisation is observed due to too low membrane fluidity in both organisms, the core permeability function of the plasma membrane is not compromised even upon conditions unable to support growth. This is consistent with a more subtle effect of low fluidity on membrane-associated biological processes maintaining ion homeostasis such as respiration.

Consequences of low membrane fluidity on cell morphogenesis

In rod-shaped bacteria, cell growth and morphogenesis are predominantly driven by two membrane-associated multiprotein complexes, the elongasome responsible for envelope expansion and rod shape determination (Typas et al, 2012), and the divisome responsible for cytokinesis (Adams & Errington, 2009). The main scaffold proteins for these prominent cellular machineries are the tubulin homolog FtsZ (Adams & Errington, 2009) and the actin homolog MreB (Typas et al, 2012). To assess the functionality of these key cellular machineries, we determined the localisation of corresponding GFP fusions upon depletion of fluidity-promoting fatty acids. Furthermore, by use of GFP-fused DNA-binding protein Hu (B. subtilis) (Köhler & Marahiel, 1997) or DNA staining with intercalating dye DAPI (E. coli), we analysed the cells for potential defects in chromosome replication and segregation.

In B. subtilis, depletion of BCFAs had no effect on nucleoid prevalence or morphology indicating the presence of largely functional DNA replication, segregation and compaction mechanisms (Fig 4A and Appendix Fig S4A). While no DNA-free cells indicative for defects in DNA replication and segregation were observed in E. coli either, a clear de-condensation of the nucleoid was evident at later stages of UFA depletion (Fig 4B and Appendix Fig S4B). Intriguingly, this process coincides with partial dissociation of RNA degradosome from the membrane as indicated by an increasingly cytoplasmic localisation of the key scaffold and membrane anchor protein RNase E in cells exhibiting very low fluidity (Fig EV2-EV5). These two processes appear to be causally linked since expression of a cytoplasmic variant of RNase E, which lacks an amphipathic helix essential for membrane binding, leads to a comparable decondensation of the nucleoid (Fig EV3A and B).

Figure 4. Consequences of low membrane fluidity on and cell morphogenesis

Images of B. subtilis Δbkd cells stained with membrane dye FM 5-95 and expressing GFP fusions of DNA-binding protein HBsu (top), cell division protein FtsZ (middle) or cell elongation protein MreB (bottom). Cells were grown with IB or depleted for precursors for 90 min (IB→PF). For further examples and additional time points, see Appendix Fig S4A. Images of E. coli fabA(Ts) cells stained with FM 5–95 for the outer membrane and with DNA-intercalating dye DAPI (top) or expressing GFP sandwich (SW) fusions to the cell division protein FtsZ (middle) and the cell elongation protein MreB (bottom), respectively. Depicted are cells grown at 30°C or with a temperature shift to 37°C for 120 min. For a more detailed view on the influence of low membrane fluidity on membrane dissociation of RNase E as well as on the cell division machinery, see Figs EV2-EV5 and Appendix Fig S5. For further examples and additional time points, see Appendix Fig S4B.

Data information: (A, B) The experiments are representative of biological triplicates. Scale bar, 3 µm. Strains used: (A) B. subtilis HS541, HS548, HS549; (B) E. coli UC1098, BHH500, BHH501.

Click here to expand this figure.

Figure EV2. Very low membrane fluidity triggers partial dissociation of RNase E from the membrane in E. coli fabA(Ts)

A, B. Phase-contrast and fluorescence images of E. coli fabA(Ts) strain expressing (A) RNase E-YFP or (B) FOF1a-mNG. Cells were grown in LB at 30°C to an OD600 of 0.3, transferred to the non-permissive temperature of 37°C for 120 min followed by labelling with DAPI and fluorescence microscopy. Note the increasing cytoplasmic localisation of RNase E-YFP upon depletion of the membrane for UFA, which coincides with decondensation of the nucleoid (compare Fig 4B). C. Quantification of membrane association of RNase E. The degree of membrane association was quantified by automated detection of cells using phase-contrast images, defining a 3-pixel wide band around the periphery of the cell, and measuring the relative membrane association as a ratio between the mean peripheral fluorescence signal and the mean fluorescence of the whole cell. D. Relative membrane association of FOF1a-mNG and RNase E-YFP at 30°C and 37°C for 120 min in individual cells (n = 60). Red lines indicate the median.

Data information: (A–D) Experiments are representative of independent biological duplicates. (D) Red lines indicate the median. P values represent the results of unpaired, two-sided t-tests. Significance was assumed with ****P < 0.0001, n.s., not significant (A–C). Scale bar: 3 µm. Strains used: (A, C, D) E. coli UC1098/pVK207; (B, D) E. coli MG4.

Source data are available online for this figure.

Click here to expand this figure.

Figure EV3. Expression of cytoplasmic RNase E is sufficient to trigger decondensation of the nucleoid

Phase-contrast and fluorescence images of E. coli WT cells expressing plasmid-encoded full-length membrane-associated RNase E (Rne) and a corresponding construct encoding RNase E that lacks the membrane-binding amphipathic helix (RneΔAH), respectively. The cells depicted were grown in LB medium at 37°C to an OD600 of 0.3 followed by induction of rne with 0.2% (w/v) arabinose (Ara) for 60 min, labelling with DAPI for 15 min and fluorescence microscopy. Note the decondensation of the nucleoid observed upon expression of cytoplasmically located RneΔAH, but not in the presence of the native membrane-associated RNase E. Variance of nucleoid staining. The degree of nucleoid condensation was assessed by analysing the variance of DAPI fluorescence within the cell (n = 199–372). In this type of analysis, a more homogeneous fluorescence signal such as that caused by nucleoid decondensation results in a lower variance of the per pixel fluorescence intensity. Red lines indicate the median.

Data information: (A, B) The experiments are representative of biological duplicates. (B) Red lines indicate the median. P values represent results of unpaired, two-sided t-tests. Significance was assumed with ****P < 0.0001, *P < 0.05, n.s., not significant. (A) Scale bar: 3 µm. Strains used: (A, B) E. coli MG1655/pJG130, MG1655/pJG131.

Source data are available online for this figure.

Click here to expand this figure.

Figure EV4. Destabilisation of E. coli divisome by deletion of the division regulator minC triggers hypersensitivity towards low membrane fluidity

Images of E. coli WT, fabA(Ts), fabA(Ts) ΔminC and ΔminC cells grown either at the permissive temperature (30°C) or for 180 min at 35°C, which is non-permissive for the fabA(Ts) ΔminC strain. Cells were stained with the outer membrane dye FM 5–95 prior to microscopy. Note the strong cell elongation of the fabA(Ts) ΔminC strain upon incubation at 35°C, which is indicative of a severe cell division defect. Quantification of cell length for cells (n = 100) depicted in panel A. Red lines indicate the median. Viability of the strains depicted in panel A upon incubation on agar plates in M9-glucose minimal medium overnight at different temperatures. The serial dilutions and spot assays were carried out with pre-cultures grown at 30°C to mid-log growth phase. Note the temperature hypersensitivity and loss of viability of strain fabA(Ts) ΔminC at 35°C, which indicates that the cell division process has become the limiting factor in tolerance towards low membrane fluidity in this strain. The temperature sensitivity of the strains fabA(Ts) ΔzapA and fabA(Ts) ΔzapB supported this view (see Appendix Fig S5).

Data information: (A–C) The experiments are representative of biological triplicates. (B) Red lines indicate the median, while the P values represent results of unpaired, two-sided t-tests. Significance was assumed with ****P < 0.0001, ***P < 0.001, **P < 0.01, n.s., not significant. (A) Scale bar: 3 µm. Strains used: (A–C) E. coli Y-Mel, UC1098, UC1098.ΔminC, JW1165-1.

Source data are available online for this figure.

Click here to expand this figure.

Figure EV5. Protein segregation induced by very low membrane fluidity is limited to the inner cytoplasmic membrane in E. coli

Membrane protein segregation was monitored in fabA(Ts) cells expressing as inner membrane marker FOF1a-mNG and as outer membrane marker OmpA-mCherry. Widefield microscopy images depict phase-contrast, fluorescence and overlay images for cells grown at permissive 30°C or at non-permissive 37°C for 120 min in M9-glucose minimal medium. Note the transition from disperse localisation into a segregated pattern in case of the inner membrane-localised FOF1a-mNG at 37°C, while the pattern of the outer membrane-localised OmpA-mCherry remains homogeneous at both growth temperatures. Super-resolution 2D-SIM (structured illumination microscopy) images of cells expressing FOF1a-mNG and OmpA-mCherry at both growth temperatures. Localisation and orientation of 5-pixel wide lines used to analyse fluorescence intensity profiles depicted in panel D. Fluorescence intensity line scans across the cells imaged with 2D-SIM microscopy. Note the small, but detectable outward shift between the inner membrane marker FOF1a-mNG and outer membrane marker OmpA-mCherry.

Data information: (A–D) Experiments are representative of biological triplicates (A–C). Scale bar: 3 µm. Strain used: (A–D) E. coli MG4/pGI10.

Source data are available online for this figure.

The cell division machinery, indicated by mid-cell localisation of FtsZ, turned out to be robust towards changes in membrane fluidity in B. subtilis, with only a weakening of the fluorescent mid-cell signal observed upon BCFA depletion (Fig 4A and Appendix Fig S4A). In contrast, a clear defect in divisome assembly was observed upon depletion of UFA in E. coli (Fig 4B and Appendix Fig S4B). To confirm that the E. coli cell division machinery indeed is stressed by low membrane fluidity