Hopanoids as functional analogues of cholesterol in bacterial membranes

Significance

The function of the cell membrane as a barrier and a matrix for biochemical activity relies on the properties imparted by lipids. In eukaryotes, sterols are crucial for modulating the molecular order of membranes. Sterol ordering provides the basis for membrane lateral segregation and promotes a fluid, mechanically robust plasma membrane. How do organisms that lack sterols determine membrane order? Hopanoids are bacterial membrane lipids that have been demonstrated to have sterol-like properties in vitro. We now explore the distribution of hopanoids and their effect on membranes in Methylobacterium extorquens. We find that hopanoids determine bacterial outer membrane order in a manner analogous to sterol ordering in the eukaryotic plasma membrane, and that their deletion impairs energy-dependent multidrug efflux.

Abstract

The functionality of cellular membranes relies on the molecular order imparted by lipids. In eukaryotes, sterols such as cholesterol modulate membrane order, yet they are not typically found in prokaryotes. The structurally similar bacterial hopanoids exhibit similar ordering properties as sterols in vitro, but their exact physiological role in living bacteria is relatively uncharted. We present evidence that hopanoids interact with glycolipids in bacterial outer membranes to form a highly ordered bilayer in a manner analogous to the interaction of sterols with sphingolipids in eukaryotic plasma membranes. Furthermore, multidrug transport is impaired in a hopanoid-deficient mutant of the gram-negative Methylobacterium extorquens, which introduces a link between membrane order and an energy-dependent, membrane-associated function in prokaryotes. Thus, we reveal a convergence in the architecture of bacterial and eukaryotic membranes and implicate the biosynthetic pathways of hopanoids and other order-modulating lipids as potential targets to fight pathogenic multidrug resistance.

Sterols (e.g., cholesterol; Fig. S1) are ubiquitous eukaryotic membrane lipids with a planar geometry that endows them with a propensity to constrain, and thereby order, lipid bilayers. The principal terms contributing to lipid order are the rotational freedom of motion and lateral packing of the lipids within the plane of the bilayer. Lipid order is directly linked to essential membrane properties, including fluidity, permeability, lateral segregation, and the propensity for membranes to bind and integrate other biomolecules (1). High plasma membrane order is a fundamental property shared across the domains of modern life, and it may have been a key factor in the selective fitness of primitive life (2). However, sterols as the primary membrane-ordering lipid present a conundrum because sterol biosynthesis requires molecular oxygen, but life was present on Earth at least a billion years before cyanobacteria first enriched the atmosphere with oxygen (3, 4). Are there lipids that could have promoted membrane ordering before sterol synthesis emerged?

Fig. S1.

Structure of lipids referred to in SI Materials and Methods.

All three domains of life share isoprenoid synthesis pathways that give rise to a broad suite of structurally homologous lipids, including sterols and hopanoids (e.g., diplopterol; Fig. S1). Hopanoids are some of the most ubiquitous cyclic isoprenoidal lipids in the sedimentary record, and they have been used as molecular proxies for ancient microbial life (5). Importantly, hopanoid synthesis does not require molecular oxygen, and hopanoids have been reported in sediments predating the enrichment of oxygen in Earth’s atmosphere (6, 7). Their discovery led to the proposal that they might serve as sterol surrogates in bacteria (8), especially because hopanoids and sterols share common structural features and are cyclized by closely related enzymes (9, 10).

One of the properties of sterols in eukaryotes is their ability to interact preferentially with lipids such as sphingomyelin (SM) to form liquid ordered phases in lipid membranes (11). This interaction provides the mechanistic basis for the underlying interconnectivity supporting membrane lateral segregation and promoting a fluid but mechanically robust plasma membrane (12). Early investigations showed that both sterols and hopanoids exhibit the ability to condense lipids (13, 14). We demonstrated that hopanoids are indeed bacterial sterol surrogates with respect to their ability to form a liquid ordered phase, and that they promote liquid-liquid phase separation in membranes (15). This observation decouples the evolution of ordered biochemically active liquid membranes from the requirement for molecular oxygen and suggests that the ability to subcompartmentalize membranes could have preceded the evolution of sterols. Subsequently, it was shown that hopanoid-based ordering can be tuned by structural modifications of their ring structure or polar side chain (16, 17).

It is clear that hopanoids in vitro exhibit an ability to order synthetic membranes in a sterol-like manner (15). However, it is not known whether hopanoids impart molecular order within bacterial membranes in vivo, and with which lipids they interact to achieve such order. Hopanoids have been identified in many gram-negative bacteria (18, 19), but their exact physiological role is unclear. Deletion of hopanoid synthesis is nonlethal in the bacteria that have been studied so far, but hopanoid-deficient mutants have been shown to exhibit increased sensitivity to antibiotics and detergents as well as susceptibility to stresses, including variation in pH, temperature, and osmotic pressure (20–24). It is not understood precisely how hopanoids are linked to antibiotic resistance and tolerance to stress, but understanding the role of hopanoids in shaping membrane properties would provide an important step toward bridging this gap.

To understand how the ordering capacity of hopanoids contributes to membrane functionality, it is necessary to examine membrane physiology in vivo. Here, we have explored the physiology of a hopanoid-producing organism from a physicochemical perspective by measuring the distribution of hopanoids and their effect on membranes in the gram-negative plant-associated bacterium Methylobacterium extorquens. We demonstrate that hopanoids preferentially interact with outer membrane (OM) glycolipids to produce highly ordered membranes, thus revealing a convergent strategy for the functional ordering of bacterial and eukaryotic surface membranes. We also identify impaired multidrug efflux as a phenotype of hopanoid deletion, which we propose accounts for sensitivity to chemical stresses. Our findings imply a link between membrane order and protein function in prokaryotes. They also suggest a possible lipid target to address bacterial multidrug resistance, further implicating isoprenoidal lipid biosynthesis as a bacterial Achilles heel (25, 26).

Results and Discussion

Hopanoids Are Localized in the OM and Determine Membrane Order.

The major hopanoids in M. extorquens are diplopterol and its methylated derivative 2-methyl-diplopterol (Fig. S1), which together make up most of the total hopanoid content (27). M. extorquens also produces extended side-chain polar hopanoids, known as bacteriohopanepolyols (BHPs), predominately composed of a bacteriohopanetetrol cyclitol ether (BHT-CE) and its guanidine-modified derivative (BHT-GCE) (28) (Fig. S1). Trace quantities (<1% of total hopanoid) of BHT and adenosylhopane can also be detected, but these trace quantities are most likely just biosynthetic intermediates. Using lipid thin-layer chromatography (TLC), we found that the major hopanoid diplopterol and its methylated derivative comprised roughly 19 mol% of the total lipids (phospholipids, LPS, and diplopterols), therefore representing a substantial component of the cellular lipidome (Fig. S2).

Fig. S2.

Cellular content of diplopterols. The percentage of cellular content diplopterols in M. extorquens WT is shown relative to total content of the major lipid classes (phospholipids and LPSs) based on analysis of biological triplicates. To illustrate the contribution of each lipid class to the membrane surface area, we show a rough estimate of the relative molecular area of each lipid class. The estimated area was calculated based on an assumed average molecular area for DOPC (60 Ų), kdo-lipid A (200 Ų), and diplopterol (38 Ų) at 25 mN/m measured on a Langmuir trough at room temperature and pH 7.4.

Previous work in a closely related organism, Methylobacterium organophilum, identified lipids with hopanoid-like properties (e.g., nonsaponifiable) in the OM; however, the structural identity of these lipids was not confirmed (29). To determine the intracellular distribution of hopanoids in M. extorquens, we purified membranes using gradient centrifugation and analyzed fractions by TLC, revealing that the major hopanoids are highly enriched in the OM fraction (Fig. 1). We confirmed hopanoid identity by MS fragmentation studies.

Fig. 1.

Hopanoids are enriched in the OM. A TLC plate shows phospholipid and hopanoid distributions in OM and inner membrane (IM) fractions. The major hopanoids detected are diplopterols (diplopterol and 2-methyl-diplopterol; Fig. S1) and polar hopanoids (BHT-GCE and BHT-CE; Fig. S1). Structures were confirmed by MS.

To characterize the effect of hopanoids on the membrane, we measured order in OM fractions purified from the WT and a hopanoid-deficient mutant of M. extorquens, ∆SHC (squalene–hopene cyclase deletion mutant). This mutant was constructed by knocking out the squalene-hopene cyclase gene essential for hopanoid synthesis. OM fractions from WT and ∆SHC were analyzed using a lipophilic fluorescent probe, 6-dodecanoyl-2-methylcarboxymethylaminonaphthalene (C-laurdan), that has been extensively used in studying membrane packing and fluidity (30, 31). We observed a significant difference in order between the WT and ∆SHC OMs, with significantly lower membrane order in the hopanoid-deficient mutant (Fig. 2). This effect was confirmed with another fluorescent probe, Di-4-ANEPPDHQ (Di-4), that responds similarly to membrane ordering (31) (Fig. S3). To confirm that the decrease in order was due to the absence of hopanoids and not to other compositional changes in the membrane of the mutant, we tested the effect of depleting hopanoids from the WT OM and, conversely, the effect of loading hopanoids into the ∆SHC OM using methyl-β-cyclodextrin. Hopanoid depletion decreased the order of the WT OM, and, analogously, loading the membranes with diplopterol increased the order of the ∆SHC OM, demonstrating that perturbed membrane order is reversible and based on ordering by diplopterol. Surprisingly, loading the hopanoid-mutant OM with cholesterol also increased its order. Therefore, we conclude that hopanoids, as well as cholesterol itself, have the potential to determine order in the outer bacterial membrane.

Fig. 2.

Hopanoids determine OM order. Membrane order was determined by C-laurdan GP of OM fractions from WT (gray), hopanoid-depleted WT (WT Depleted), hopanoid-deficient ∆SHC (red), and ∆SHC loaded with diplopterol (∆SHC + Dip) or cholesterol (∆SHC + Chol). Data represent average values from biological replicates (n = 3) measured at 30 °C and pH 7.4. One-tailed P values for WT vs. WT Depleted (P = 0.047) and WT vs. ∆SHC (P = 0.006) were made by an unpaired t test using Prism software (GraphPad).

Fig. S3.

Hopanoids determine OM order. Membrane order was determined by Di-4 GP of OM fractions from WT (gray) and hopanoid-deficient ∆SHC (red). Data represent average values from biological replicates (n = 3) measured at 30 °C and pH 7.4.

Hopanoids Are Predicted to Interact Preferentially with Lipid A in the OM.

It remained unclear which lipids hopanoids interact with in the OM to determine order. In eukaryotes, sterols interact with saturated sphingolipids to form a membrane that is highly ordered. We demonstrated that hopanoids also interact with sphingolipids to form liquid ordered membranes (15). However, with a few exceptions, bacteria do not produce sphingolipids (32). The hopanoid-containing bacterial OM is characterized by an asymmetric bilayer in which the inner leaflet contains mostly phospholipids and the outer leaflet contains lipid A, which is the conserved core of LPS (33). In M. extorquens, virtually all of the phospholipids are unsaturated, with roughly 90% containing double bonds in both acyl chains and the remaining 10% containing both saturated and unsaturated acyl chains (Fig. S4). In contrast, lipid A has a highly conserved structure that bears many features similar to sphingolipids, most notably the saturated acyl chains, amide-linked backbone, and hydroxylations (Fig. 3). We therefore hypothesized that hopanoids interact with lipid A in a manner analogous to the interaction of sterols with sphingolipids to promote order in the OM.

Fig. S4.

Total phospholipid unsaturation. The percentage of total phospholipids with zero (saturated), one, or two unsaturated acyl chains in M. extorquens WT was calculated from individual phospholipid species analyzed using shotgun MS. Notably, no saturated phospholipids were detected.

Fig. 3.

Diplopterol interacts preferentially with saturated lipids (e.g., lipid A). The ∆G^ex is shown for the interaction of diplopterol or cholesterol with lipid A, SM, and phospholipids with various degrees of unsaturation (indicated by a red arrow) from 1:2 mixtures (by molarity) of diplopterol/lipid or cholesterol/lipid. Positive values indicate a repulsive interaction, whereas negative values indicate an attractive/favorable interaction. The ∆G^ex values represent the average of replicates (n = 3) measured by Langmuir trough at room temperature and pH 7.4. Original data used for these calculations are included in Tables S1–S3.

To examine which type of lipids hopanoids would preferentially interact with, we compared the favorability of interaction of diplopterol with lipid A, SM, and synthetic phospholipids with varying degree of unsaturation (Fig. 3) by measuring the Gibbs excess free energy of mixing (∆G^ex) on a Langmuir trough. We made the same comparison using cholesterol in place of diplopterol. The ∆G^ex is a quantitative measure of the interaction between lipids, with negative values indicating a favorable/attractive interaction and positive values indicating a repulsive interaction. We revealed a key difference between diplopterol and cholesterol. Whereas cholesterol exhibits a favorable interaction with phospholipids of varying degrees of unsaturation, diplopterol exhibits a repulsive interaction with unsaturated phospholipids. Furthermore, the ∆G^ex values for the diplopterol-lipid A and cholesterol-SM are nearly identical and negative, thus confirming that interactions between both pairs of these lipids are favorable. This finding is consistent with our previous results showing that diplopterol orders lipid A but does not order unsaturated phospholipids (15). Thus, we show that diplopterol interacts favorably only with saturated lipids and that the interactions of diplopterol with lipid A and cholesterol with SM are thermodynamically analogous.

Deletion of Hopanoids Impairs Energy-Dependent Cellular Efflux.

Having established that hopanoids determine the order of the OM, we examined the impact of hopanoid deletion on the physiology of M. extorquens. The OM serves as the first barrier in gram-negative bacteria (33, 34). Reduced membrane order could lead to lower resistance to bilayer-disrupting agents (e.g., detergents), membrane permeability, and impaired membrane protein function (e.g., passive and active transporters) (35). In line with this reasoning, previous studies in other species of hopanoid-producing bacteria have shown that hopanoid deficiency sensitizes bacteria to antimicrobial compounds and membrane-disrupting agents, such as bile salts, peptides, and detergents (20, 22–24, 36). These observations have generally been interpreted as evidence that hopanoids maintain low permeability and mechanical robustness of the OM. However, it remained to be tested whether the chemical sensitivity caused by hopanoid deletion is also linked to the impairment of active processes, such as the energy-dependent transport of compounds out of the cell.

To examine the extent of detergent sensitivity as an indicator of membrane barrier function, we compared the influence of a nonionic detergent, Triton X-100 (TX-100), on WT and ∆SHC viability using a spot assay (Fig. 4). The lethal concentration of TX-100 was more than 1,000-fold lower in ∆SHC than in WT. Importantly this phenotype is also one of the classic symptoms of an impaired multidrug transport system (37).

Fig. 4.

Energy-dependent multidrug transport is impaired by hopanoid deletion. (A) Detergent sensitivity of WT and ∆SHC was determined by spot assay. (B) Membrane permeability and transport were assessed by monitoring the accumulation of the lipophilic dye NPN in the cell (n = 3 biological replicates). (C) ATP-dependent multidrug transport activity was examined by H33342 fluorescence assay (n = 3 biological replicates).

We examined whether hopanoid deletion affected membrane permeability and transport by monitoring the accumulation of the lipophilic dye 1-N-phenylnaphthylamine (NPN) in the cell. NPN fluorescence is low while in solution and high when the dye is present in the membrane environment (within the cell). WT M. extorquens showed an initial increase of fluorescence signal, followed by a decrease of signal, pointing toward the active efflux of the dye, as typically observed in healthy bacteria (38) (Fig. 4). The ∆SHC mutant, however, accumulated dye steadily until an equilibrium level was reached. This result is typical for bacteria that have been treated with ATP synthesis-blocking cyanide (38). The slope of NPN accumulation of the ∆SHC mutant is 2.2-fold higher than for the WT, indicating higher susceptibility to dye penetration of the mutant OM. However, membrane permeability changes alone cannot account for the more than 1,000-fold reduction in tolerance to detergent that we observed (37). Therefore, we hypothesized that these observations pointed toward impaired multidrug transport in the hopanoid-deficient mutant.

To test directly whether energy-dependent multidrug transport was impaired in the hopanoid-deficient mutant, we used a standard multidrug efflux assay utilizing the dye Hoechst 33342 (H33342). H33342 is a substrate for a broad range of bacterial multidrug transporters (39) and fluoresces only when bound to nucleic acids. Thus, its fluorescence intensity is a proxy for cellular uptake and efflux (40–42). We simultaneously provided cells with H33342 and energy-depleted them using the ionophore carbonyl cyanide 3-chlorophenylhydrazone (CCCP) in the absence of an utilizable carbon source (succinate) to inhibit energy-dependent efflux. Upon addition of succinate and withdrawing CCCP, ATP production was restored and the WT cells were rapidly able to efflux H33342, reducing fluorescence to levels observed in untreated cells (Fig. 4). In contrast, upon restoring ATP synthesis in ∆SHC cells, H33342 fluorescence exhibited much slower reduction, indicating impaired efflux. This result shows that energy-dependent multidrug transport is deficient in the hopanoid ∆SHC mutant.

Summary and Outlook

We demonstrate that hopanoids can determine order in the bacterial OM through their interaction with lipid A, analogous to the interaction of cholesterol with sphingolipids in eukaryotic plasma membranes. Hopanoids have been identified in the OM of diverse bacteria (29, 43–46), and two recent studies have reported a covalently linked hopanoid-lipid A compound in rhizobial plant-associated bacteria (47, 48). These observations suggest that hopanoid-lipid A ordering may be widespread among hopanoid-producing bacteria. Interestingly, hopanoids have also been observed in association with other specialized membranes, such as the thylakoid membranes in cyanobacteria and in the vesicle envelope of the root nodule symbionts Frankia spp. (44, 45, 49, 50). The exact role of hopanoids in these diverse bacterial membranes is not fully understood. Furthermore, bacteria that lack hopanoids may use other, possibly unknown, ordering lipids in their membranes. For instance, it has been shown that Borrelia burgdorferi, the causative agent of Lyme disease, can maintain membrane order by incorporating cholesterol from its host (51). It has also been proposed that proteins could serve a role in maintaining membrane order in bacteria that lack ordering lipids (2). Thus, although our understanding of the diverse mechanisms that bacteria use to modulate molecular order in the membrane remains incomplete, we reveal hopanoid/sterol ordering of saturated lipids (e.g., lipid A, sphingolipids) as a strategy for achieving ordered cell-surface membranes that are found from bacteria to eukaryotes.

Our data show that multidrug efflux in M. extorquens is dependent on hopanoids. This finding may explain the basis for previous observations linking hopanoids to antibiotic resistance in the hopanoid-producing pathogen Burkholderia (22–24). One explanation for impaired efflux could be that reduced membrane order affects the functionality of some component of the multidrug transport system. However, it is also possible that hopanoids interact directly with certain proteins, modulating their action. There are diverse OM efflux proteins that interact with different types of transport systems to facilitate cellular efflux (52–55). The activity of an OM efflux protein might potentially be impaired by altered diffusivity or defective gating associated with a change in membrane order (52). For instance, it has been shown that the substrate affinity of proteins and receptors in the eukaryotic plasma membrane is modulated by cholesterol ordering (35, 56). Given the architectural convergence between the bacterial OM and the eukaryotic plasma membrane, similar mechanisms could account for the functional role of hopanoids in bacteria. However, a complete understanding of the mechanistic link between hopanoids and multidrug transport represents a complex problem that remains to be explored. Nonetheless, our results raise the interesting possibility that targeting the synthesis of bacterial lipids that determine membrane order may provide a novel approach to addressing antibiotic resistance.

Our observations raise several key questions surrounding the functional significance of hopanoids in membrane physiology and the role of the OM in bacterial cellular organization. It will be important to understand how lipid order influences the dynamics of OM proteins, such as their structure, insertion into the membrane, lateral diffusion, and dimerization. Furthermore, it remains unknown if the OM is capable of lateral compartmentalization as has been proposed to occur in other bacterial membranes (57). Do hopanoids promote such membrane organization? Recently, it was demonstrated that the bacterial actin homolog MreB contributes to the organization of the membrane (58). Are there links between hopanoids, OM compartmentalization, and bacterial cytoskeletal analogs? Our work provides a foundation for exploring these fundamental questions by introducing a bacterial system for studying the effect of lipid ordering on membrane function in vivo.

Materials and Methods

Materials.

SM, kdo-lipid A, DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine), DPPC (1,2-dipalmitoyl-sn-glycero-3-phosphocholine), POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine), and cholesterol were purchased from Avanti Polar Lipids. Diplopterol was purchased from Chiron AS. C-laurdan was a gift from B. R. Cho Korea University, Seoul, South Korea. Stock concentrations of lipids were measured by phosphate assay. Cholesterol and diplopterol were weighed out on a precision balance and solubilized in chloroform/methanol (2:1).

Methods for data presented in Figs. S1–S8 are described in SI Materials and Methods.

Fig. S8.

MS² spectra of polar hopanoids.

Table S1.

Isothermal measurements of pressure vs. area of individual lipids measured on Langmuir trough

	Cholesterol		Diplopterol		DOPC		POPC		DPPC		SM		Lipid A
Pressure, nN/m	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD
5	40	1.5	49	0.9	90	1.6	75	2.0	76	1.3	69	1.3	261	1.0
10	39	1.7	47	0.7	79	1.4	66	1.6	63	1.0	54	1.3	241	1.1
15	39	1.5	45	0.8	71	1.2	59	1.4	49	0.4	48	0.7	225	0.7
20	38	1.5	44	0.7	65	0.8	54	1.2	45	0.7	46	0.7	212	0.8
25	38	1.4	43	0.6	60	0.7	49	0.9	42	0.8	45	0.7	201	0.5
30	37.1	1.3	42.2	0.3	56	0.8	45	0.9	39	0.7	43	0.7	191	0.4

DOPC, (1,2-dioleoyl-sn-glycero-3-phosphocholine); DPPC, (1,2-dipalmitoyl-sn-glycero-3-phosphocholine); Lipid A, (Di[3-deoxy-D-manno-octulosonyl]-lipid A); POPC, (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine); SM, (N-stearoyl-D-erythro-sphingosylphosphorylcholine).

Table S2.

Isothermal measurements of pressure vs. area of cholesterol/lipid (1:2 mol/mol) mixture measured on Langmuir trough

	Cholesterol-DOPC		Cholesterol-POPC		Cholesterol-DPPC		Cholesterol-SM		Cholesterol-lipid A
Pressure, nN/m	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD
5	67	0.9	57	1.2	47	0.9	44	1.0	177	0.2
10	59	0.9	50	0.9	42	0.6	42	0.8	165	0.4
15	53	1.0	47	0.6	39	0.5	41	0.7	156	0.5
20	49	1.2	44	0.4	37	0.4	40	0.6	148	0.4
25	46	1.3	42	0.4	34	0.6	39	0.6	141	0.3
30	43	1.0	39	0.3	33	0.6	38	0.6	134	0.3

Table S3.

Isothermal measurements of pressure vs. area of diplopterol/lipid (1:2 mol/mol) mixture measured on Langmuir trough

	Diplopterol-DOPC		Diplopterol-POPC		Diplopterol-DPPC		Diplopterol-SM		Diplopterol-lipid A
Pressure, nN/m	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD	Å2/mol	SD
5	78	0.8	68	2.6	52	0.3	48	1.4	179	1.5
10	70	0.7	61	2.5	47	0.2	46	1.4	167	1.1
15	64	0.8	55	2.2	43	0.1	45	1.3	158	0.9
20	59	0.8	51	2.2	41	0.2	44	1.2	151	0.9
25	55	0.8	47	2.2	38	0.2	43	1.2	143	1.1
30	51	1.2	44	2.2	37	0.3	42	1.0	137	0.9

Media, Growth Conditions.

All Methylobacterium strains were grown at 30 °C (unless otherwise stated) in a minimal medium (59), referred to hereafter as Hypho medium, with 15 mM succinate as the sole carbon source. Escherichia coli strains were grown at 37 °C on solid LB. Suitable antibiotics were used for selection: 50 μg/mL ampicillin, 20 μg/mL chloramphenicol, 50 μg/mL kanamycin, 50 μg/mL rifamycin, 35 μg/mL streptomycin, and 10 μg/mL tetracycline.

Construction of Plasmids and Generation of Hopanoid-Deficient Mutant ∆SHC.

A hopanoid-deficient mutant strain of M. extorquens PA1 was constructed using a modified procedure described previously (60). Details are described in SI Materials and Methods. Regions of the chromosome, including the deleted sequence, were amplified by PCR and sequenced to confirm generation of the desired, unmarked deletion. The deletion of hopanoid synthesis was further confirmed by the absence of hopanoids in a total lipid extract visualized by TLC (Fig. S6).

Fig. S6.

Confirmation of the absence of hopanoids in ∆SHC mutant. Bligh–Dyer (61) lipid extracts from WT, hopanoid-deficient ∆SHC, and a diplopterol standard were visualized by silica-gel TLC (mobile phase: 47:17:15:14:8 (vol/vol/vol/vol/vol) chloroform/acetone/methanol/acetic acid/water. Hopanoid identity was confirmed by MS.

Membrane Separation from M. extorquens WT and ∆SHC.

A membrane separation protocol was adapted from a protocol previously established for Methylobacterium (29). The method is described in detail in SI Materials and Methods.

Hopanoid Distribution in OM Fractions.

OM fractions were extracted using the procedure of Bligh and Dyer (61). The resulting extracts were loaded on a silica gel plate (HPTLC Silica gel 60 with concentrating zone; Merck) and resolved with chloroform/acetic acid/methanol/water [80:15:2:4 (vol/vol/vol/vol)] for phospholipids and polar hopanoids or with chloroform for diplopterols. The identity of the polar hopanoids (BHT-CE and BHT-GCE) and the diplopterols (diplopterol and 2-methyl-diplopterol) was confirmed by MS fragmentation of bands that were scraped from the plates and reextracted from the silica gel using the procedure of Bligh and Dyer (61), as described in SI Materials and Methods.

Methyl-β-Cyclodextrin Depletion and Loading of Membrane Fractions.

The depletion and loading of hopanoids into membranes were achieved using a method utilizing methyl-β-cyclodextrin (MβCD) (62). For depletion, membrane amounts were adjusted using the intensity of a lipid bilayer scattering peak at 425 nm (excitation wavelength λ_ex = 385 nm) (2) to 1 mM lipid and then depleted with 10 mM MβCD for 2 h on ice. For loading, a loading solution of 10:1 (by molarity) MβCD/cholesterol or diplopterol was prepared by equilibrating 20 mM MβCD with 2 mM cholesterol or diplopterol overnight at 30 °C. Membrane fraction amounts were then adjusted to ca. 2 mM lipid and loaded with 20 mM 10:1 MβCD/diplopterol or 10:1 MβCD/cholesterol and gently shaken for 2 h at 4 °C. To remove MβCD, the membrane fractions were pelleted by centrifugation at 70,000 × g for 90 min and then washed twice with buffer M; they were eventually pelleted and resuspended in sterile Hypho medium for subsequent analysis.

C-Laurdan and Di-4 Spectroscopy of Membrane Fractions.

Membrane fractions of WT and ∆SHC OMs from three independent cultures were prepared as described above. The membrane fractions were subjected to sonication for 5 min to promote formation of unilamellar membranes. The presence of bilayers and estimation of membrane amount were assessed in unlabeled membrane fractions by the intensity of lipid bilayer scattering. Membrane fractions adjusted to ca. 200 μM lipid were stained with 400 nM C-laurdan or Di-4 (0.2 mol%) and incubated on ice for 20 min. Labeled fractions were then equilibrated at 30 °C for 10 min. Spectra were recorded at a resolution of 1 nm on a Fluoromax-3 fluorescence spectrometer (Horriba) at a constant temperature of 30 °C. Excitation of C-laurdan was 385 nm, and excitation of Di-4 was 497 nm. The general polarization (GP) values were calculated from two emission bands: 400–460 nm (Ch1) and 470–530 nm (Ch2) for C-laurdan and 525–580 nm (Ch1) and 655–750 nm (Ch2) for Di-4, according to the following equation (30, 31):

$$\text{GP}=\left(I_{Ch\mathit{1}}-I_{Ch\mathit{2}}\right)/I_{Ch\mathit{1}}+I_{Ch\mathit{2}}.$$

Monolayers and ∆G^ex Calculations.

Isotherms of monolayers of synthetic and purified lipids prepared as described previously (63) were recorded using a 70-cm² Teflon Langmuir trough fitted with a motorized compression barrier equipped with a pressure sensor and Wilhelmy plate (Nima Technnology). The mean molecular areas for each mixture were estimated from the averages of isotherms from three monolayers that were prepared independently. The ∆G^ex was calculated by integrating the areas of lipid mixtures over pressures Π = 5, 10, 15, 20, and 25 mN/m according to Grzybek et al. (63), and as described in detail in SI Materials and Methods.

Detergent Sensitivity Spot Assay.

A spot assay was used to determine the sensitivity of the WT and ΔSHC to detergent. Cells in exponential growth adjusted to an OD₆₀₀ of 0.2 were treated with varying concentrations of TX-100 for 1 h. Cells were washed twice with medium to remove TX-100. After thoroughly mixing, 5 μL of each dilution was applied on a Hypho agar plate in a single drop, starting from the lowest to the highest dilution. The plates were incubated for 2–3 d at 30 °C.

NPN Uptake.

M. extorquens cultures were grown overnight to an OD₆₀₀ of 0.2–0.3 and harvested by centrifugation (5,000 × g for 10 min). The pellets were washed twice with medium and adjusted to a final OD₆₀₀ of 0.5 (8,000 × g for 2 min). One hundred eighty microliters per well of cells was transferred to a 96-well plate (black, clear, flat bottom; Sarstedt). The emission of NPN was recorded using a plate reader [Perkins Elmer Envison; filters (wavelength/bandwidth): excitation = 340/25 nm, emission = 450/8 nm]. Readouts were taken every minute, with 55 s of shaking between readouts. The background signal was measured for 5 min before NPN was added to a final concentration of 5 μM (5 μL of 185 μM NPN solution) per well. The uptake of dye was recorded for 90 min by measuring its emission under the same conditions. The entire assay was performed at room temperature. Succinate was present at all times at 15 mM.

H33342 ATP-Dependent Efflux Assay.

M. extorquens cultures were grown overnight to an OD₆₀₀ of 0.2–0.3 in Hypho medium at 30 °C. Cells were harvested by centrifugation (5,000 × g for 10 min) and washed once with medium lacking an utilizable carbon source (succinate). All following wash steps and resuspensions were performed using succinate-free Hypho medium containing 5 μM H33342 (8,000 × g for 2 min). The OD₆₀₀ was adjusted to 1.0 before freshly prepared CCCP in DMSO was added to a final concentration of 100 μM [∼1% (vol/vol) DMSO] to abolish ATP synthesis. The 1% DMSO did not inhibit growth of the WT or ΔSHC mutant. The mixtures were incubated in darkness for 1 h, washed twice to remove CCCP, and then resuspended to an OD₆₀₀ of 1.0. One hundred eighty microliters of the suspension was transferred to wells of a 96-well plate. The initial uptake/equilibrium of H33342 was recorded using a plate reader (Envison) for 45 min [filters (wavelength/ bandwidth): excitation = 340/25 nm, emission = 405/8 nm], and readings were taken for 45 cycles (every minute), with 55 s of shaking between readouts. Afterward, either glucose or succinate was added to a final concentration of 20 mM (5 μL of 740 mM solution) per well and the change in H33342 emission was recorded for an additional 60 min (60 cycles, 55-s delay with shaking). Glucose was used as a negative control because M. extorquens is unable to utilize it. All steps of the assay were performed at room temperature.

SI Materials and Methods

Materials.

SM, kdo-lipid A, DOPC, POPC, DPPC, and cholesterol were purchased from Avanti Polar Lipids. Diplopterol was purchased from Chiron AS. C-laurdan was a gift from Prof. B. R. Cho. Stock concentrations of lipids were measured by phosphate assay. Cholesterol and diplopterol were weighed out on a precision balance and solubilized in chloroform/methanol (2:1).

Medium, Growth Conditions.

All Methylobacterium strains were grown at 30 °C (unless otherwise stated) in a minimal medium (60), referred to hereafter as Hypho medium, with 15 mM succinate as the sole carbon source. E. coli strains were grown at 37 °C on solid LB. Suitable antibiotics were used for selection: 50 μg/mL ampicillin, 20 μg/mL chloramphenicol, 50 μg/mL kanamycin, 50 μg/mL rifamycin, 35 μg/mL streptomycin, and 10 μg/mL tetracycline.

Determination of Total Cellular LPS, Phospholipid, and Diplopterol Content.

LPS was quantified in whole cells using a colorimetric assay (64). Assays for phospholipid and diplopterol content were performed on Bligh–Dyer total lipid extracts of cells (62) from six OD units of cells in early exponential growth (OD₆₀₀ of 0.2–0.3). Total phospholipid content was determined by phosphate assay. Diplopterols (diplopterol and 2-methyl-diplopterol) were quantified using TLC against a diplopterol standard by comparing the band intensity of lipids charred with sulfuric acid. Lipid extracts were resolved on a silica gel plate (HPTLC Silica gel 60 with concentrating zone, no. 1.13748.0001, lot no. HX939023; Merck) with chloroform as the mobile phase. TLC plates were digitally scanned, and bands were integrated and quantified using Fiji (ImageJ; NIH) (65) (Fig. S5).

Fig. S5.

Quantification of diplopterols by TLC. Bligh–Dyer (61) lipid extracts of M. extorquens WT biological triplicates (WT1–WT3) were resolved by TLC, and diplopterols were quantified against a diplopterol standard.

Determination of Degree of Phospholipid Acyl Chain Unsaturation.

Lipid composition was interrogated using shotgun MS of total lipid Bligh–Dyer extracts (66). The total lipidome was found to be represented by four lipid classes: phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), and cardiolipin (CL). Extracts were diluted 25-fold with an acquisition mixture containing the 0.5 μM PC (34:0), PE (34:0), and PG (34:0) mixture and 0.05 μM CL (57:4) dissolved in acquisition mixture containing 7.5 mM ammonium acetate in 4:2:1 isopropanol/methanol/chloroform. The extracts were then analyzed with a Q Exactive tandem mass spectrometer (Thermo Fisher Scientific) equipped with a robotic nanoflow ion source, TriVersa NanoMate (Advion BioSciences). Spectra were acquired in negative MS mode with a mass resolution of R_{m/z 400} = 280,000 in the mass range of 400–1,000 m/z. Raw data were processed using LipidXplorer software (67). Total phospholipid unsaturation was calculated from the individual phospholipid species.

Construction of Plasmids and Generation of Hopanoid-Deficient Mutant ∆SHC.

A hopanoid-deficient mutant strain of M. extorquens PA1 was constructed using a modified procedure described previously (60). DNA sequences flanking the upstream and downstream regions of SHC were amplified by PCR. Following methods described by Gibson (68), each PCR primer contained a 30-bp overhang that complemented the adjacent region of the final assembled plasmid. Plasmid pAB194 was generated by combining three linear pieces of DNA: (i) the broad-host-range allelic exchange vector pCM433 (61) linearized with NotI, (ii) a PCR product amplifying a region upstream of SHC using primers AB-orf52uf (ATGGATGCATATGCTGCAGCTCGAGCGGCCGCCCGCGCCGCAGGAATTC) and AB-orf52ur (CGCATCGTTCTCGCCTCGTTC), and (iii) a PCR product amplifying a region downstream of SHC using primers AB-orf52df (GAGACAGTCGAACGAGGCGAGAACGATGCGGCAACCTGAAGCGGGGCAAC) and AB-orf52dr (GGTTAACACGCGTACGTAGGGCCCGCGGCCGCGATTGAGACCCGCGGGTCATC). These three products were combined with Phusion (New England Biolabs), Taq ligase, and T5 exonuclease in an isothermal reaction buffer containing Tris⋅HCl, MgCl₂, dNTPs, NAD, PEG 8000, and molecular-grade water in a reaction volume of 20 μL. The reaction proceeded at 50 °C for 1 h. One microliter of this reaction was subsequently transformed into New England Biolabs (NEB) 5-alpha chemically competent E. coli and propagated.

These constructs were conjugated from E. coli into CM2730, a strain of PA1 that is nearly WT but has been engineered to reduce clumping through the removal of the celABC locus (69) and plated onto media containing tetracycline for selection of strains with a single-crossover homologous recombination. Colonies were picked, grown in liquid medium in the absence of antibiotics for 6 h, and subsequently plated onto succinate agar containing sucrose (5% wt/vol) selected against the lethal sacB locus, resulting in a second recombination event that removed the integrated plasmid to leave behind the deleted (or original) allele. Regions of the chromosome, including the deleted sequence, were amplified by PCR and sequenced to confirm generation of the desired unmarked deletion. The deletion of hopanoid synthesis was further confirmed by the absence of hopanoids in a total lipid extract visualized by TLC (Fig. S6).

Membrane Separation from M. extorquens WT and ∆SHC.

A membrane separation protocol was adapted from a protocol previously established for Methylobacterium (29). Two liters of culture was harvested at an OD₆₀₀ of 0.2 by centrifugation (5,000 × g at 24 °C for 15 min). The pellet was washed in 250 mL of buffer W [20 mM Hepes (pH 7.25) and 5 mM EDTA (pH 8.0)] and spun down. After resuspending in lysis buffer [20 mM Hepes (pH 7.25), 5 mM EDTA (pH 8.0), and 20% (wt/wt) sucrose] (pellet wet weight of 50 mL/g), the suspension was incubated (30 °C at 150 rpm in a shaker) for 15 min, after which lysozyme (Amresco) was added to a final concentration of 1 mg/mL and left for another 90 min. Cells were spun down (5,000 × g at 4 °C for 15 min) and resuspended in precooled buffer M [20 mM Hepes (pH 7.25), 5 mM EDTA (pH 8.0), 20% (wt/wt) sucrose, 25 mM MgCl₂, 1 mM PMSF, and 0.5 μg/mL DNase (DNase I; Sigma–Aldrich)] (wet weight of 25 mL/g, minimum of 50 mL). All subsequent steps were performed at 4 °C. Cells were lysed by passing them four times through an EmulsiFlex-C5 (Avestin) at a peak pressure of 17,500 psi. Intact cells and debris were pelleted (8,500 × g for 20 min) and discarded. The supernatant was pelleted (70,000 × g for 90 min). The resulting membrane pellet was resuspended in 500 μL of buffer M using a widened 1-mL pipette tip and a 20-gauge needle. The suspension was adjusted with 3 mL of 65% (wt/wt) sucrose in buffer M to yield a final concentration of 58% (wt/wt) sucrose. This 58% sucrose suspension was overlaid with 5 mL of 50% (wt/wt), 2 mL of 30% (wt/wt), and 2 mL of 5% (wt/wt) sucrose solutions and centrifuged for 16 h at 250,000 × g. The translucent inner membrane fraction localized at the interface between 5% (wt/wt) and 30% (wt/wt) sucrose, whereas the pink OM fraction localized at the interface between 30% (wt/wt) and 50% (wt/wt) sucrose, consistent with previous work (29).

Hopanoid Distribution in OM Fractions.

OM fractions were extracted using the procedure of Bligh and Dyer (61). The resulting extracts were loaded on a silica gel plate and resolved with 80:15:2:4 (vol/vol/vol/vol) chloroform/acetic acid/methanol/water for phospholipids and polar hopanoids or with chloroform for diplopterols. The identity of the polar hopanoids (BHT-CE and BHT-GCE) and diplopterols (diplopterol and 2-methyl-diplopterol) was confirmed by MS fragmentation of bands that were scraped from the plates and reextracted from the silica gel using a procedure of Bligh and Dyer (61). The extracts were dried down and redissolved in acquisition mixture containing 7.5 mM ammonium acetate in 4:2:1 isopropanol/methanol/chloroform and then analyzed with a Q Exactive tandem mass spectrometer equipped with a robotic nanoflow ion source (TriVersa NanoMate). Polar hopanoids were identified in negative ion mode with a mass resolution of R_{m/z 400} = 280,000 in the mass range of 400–1,000 m/z. Detected were molecular ion [M + Acetate]⁻ for BHT-CE (m/z = 766.5490) and BHT-GCE (m/z = 808.5712), as well as characteristic fragment ions shown in Figs. S7 and S8. Spectra for diplopterols were acquired in positive ion mode with a mass resolution of R_{m/z 400} = 140,000 in the mass range of 360–455 m/z. Intact parent ions [M + H]⁺ of diplopterols were not detected in MS spectra due to lability of the hydroxyl group and rapid in-source dehydroxylation. Therefore, diplopterols were identified by masses corresponding to the water loss [M − H₂O]⁺ fragments in MS for diplopterol (m/z = 411.3984) and 2-methyl-diplopterol (m/z = 425.4139).

Fig. S7.

Fragmentation structures and masses of polar hopanoids.

MβCD Depletion and Loading of Membrane Fractions.

The depletion and loading of hopanoids into membranes were achieved using a method utilizing MβCD (62). For depletion, membrane amounts were adjusted using the intensity of a lipid bilayer scattering peak at 425 nm (excitation wavelength λ_ex = 385 nm) (10) to 1 mM lipid and then depleted with 10 mM MβCD for 2 h on ice. For loading, a loading solution of 10:1 (by molarity) MβCD/cholesterol or diplopterol was prepared by equilibrating 20 mM MβCD with 2 mM cholesterol or diplopterol overnight at 30 °C. Membrane fraction amounts were then adjusted to ca. 2 mM lipid and loaded with 20 mM 10:1 MβCD/diplopterol or 10:1 MβCD/cholesterol and gently shaken for 2 h at 4 °C. To remove MβCD, the membrane fractions were pelleted by centrifugation at 70,000 × g for 90 min and washed twice with buffer M; they were eventually pelleted and resuspended in sterile Hypho medium for subsequent analysis.

C-Laurdan and Di-4 Spectroscopy of Membrane Fractions.

Membrane fractions of WT and ∆SHC OMs from three independent cultures were prepared as described above. The membrane fractions were subjected to sonication for 5 min to promote formation of unilamellar membranes. The presence of bilayers and the estimation of membrane amount were assessed in unlabeled membrane fractions by the intensity of lipid bilayer scattering. Membrane fractions adjusted to ca. 200 μM lipid were stained with 400 nM C-laurdan or Di-4 (0.2 mol%) and incubated on ice for 20 min. Labeled fractions were then equilibrated at 30 °C for 10 min. Spectra were recorded with a resolution of 1 nm on a Fluoromax-3 fluorescence spectrometer (Horriba) at a constant temperature of 30 °C. Excitation of C-laurdan was 385 nm, and excitation of Di-4 was 497 nm. The GP values were calculated from two emission bands: 400–460 nm (Ch1) and 470–530 nm (Ch2) for C-laurdan and 525–580 nm (Ch1) and 655–750 nm (Ch2) for Di-4 according to the following equation (30, 31):

$$\text{GP}=\left(I_{Ch\mathit{1}}-I_{Ch\mathit{2}}\right)/I_{Ch\mathit{1}}+I_{Ch\mathit{2}}.$$

Monolayers and ∆G^ex Calculations.

Isotherms of monolayers of synthetic and purified lipids prepared as described previously (63) were recorded using a 70-cm² Teflon Langmuir trough fitted with a motorized compression barrier equipped with a pressure sensor and Wilhelmy plate (Nima Technnology). Briefly, chloroform/methanol [2:1 (vol/vol)] solutions of pure lipids and lipid mixtures were prepared at lipid concentrations of 0.5 mg/mL. Monolayers were procured by injecting 10–20 μL of lipid solution into an aqueous phase maintained at a constant 25 °C. The water phase was composed of 150 mM NaCl, 3.3 mM sodium citrate, 3.3 mM sodium phosphate, 3.3 mM glycine, and 0.1 mM EDTA, with the pH adjusted to 7.4, 5.1, or 3.1 by HCl or NaOH.

The mean molecular areas (MMA) for each mixture were estimated from the averages of isotherms from three monolayers that were prepared independently. The theoretical mean area per molecule (lipid) for each mixture was calculated as follows:

$${\text{A}}_i={\text{X}}_1{\text{A}}_1+{\text{X}}_2{\text{A}}_2,$$

where A_i is the MMA of the mixture; X₁ and X₂ are the mole fractions of lipid 1 and lipid 2, respectively; and A₁ and A₂ are the MMAs of lipid 1 and lipid 2, respectively, at surface pressures Π = 5, 10, 15, 20, and 25 mN/m. The percent change in molecular area (condensation effect) was calculated as follows:

$$c=100\left({\text{A}}_i-{\text{A}}_o\right)/{\text{A}}_i,$$

where c is the percentage of condensation; A_o is the observed MMA at 5, 10, 15, 20, and 25 mN/m; and A_i is the theoretical MMA of the two lipids. The excess free energy of mixing ∆G^ex was calculated by integrating the areas of lipid mixture A₁₂ (containing A₁ at mole fraction X₁ and A₂ at mole fraction X₂) with pure lipids A₁ and A₂ over pressures Π = 5, 10, 15, 20, and 25 mN/m according to Grzybek et al. (63) with the following equation:

$$\mathrm{\Delta }{\text{G}}^{\text{ex}}={\displaystyle \underset{\mathrm{\Pi }0}{\overset{\mathrm{\Pi }}{\int }}A12d\mathrm{\Pi }-\text{X}1{\displaystyle \underset{\mathrm{\Pi }0}{\overset{\mathrm{\Pi }}{\int }}A1d\mathrm{\Pi }-{\text{X}}_2{\displaystyle \underset{\mathrm{\Pi }0}{\overset{\mathrm{\Pi }}{\int }}A2d\mathrm{\Pi }}}}.$$

Detergent Sensitivity Spot Assay.

A spot assay was used to determine the sensitivity of the WT and ΔSHC to detergent. Cells in exponential growth adjusted to an OD₆₀₀ of 0.2 were treated with varying concentrations of TX-100 for 1 h. Cells were washed twice with medium to remove TX-100. After thorough mixing, 5 μL of each dilution was applied on a Hypho agar plate in a single drop, proceeding from the lowest to the highest dilution. The plates were incubated for 2–3 d at 30 °C.

NPN Uptake.

M. extorquens cultures were grown overnight to an OD₆₀₀ of 0.2–0.3 and harvested by centrifugation (5,000 × g for 10 min). The pellets were washed twice with medium and adjusted to a final OD₆₀₀ of 0.5 (8,000 × g for 2 min). One hundred eighty microliters per well of cells was transferred to a 96-well plate. The emission of NPN was recorded using a plate reader [Perkins Elmer Envison; filters (wavelength/bandwidth): excitation = 340/25 nm, emission = 450/8 nm]. Readouts were taken every minute, with 55 s of shaking between readouts. The background signal was measured for 5 min before NPN was added to a final concentration of 5 μM (5 μL of 185 μM NPN solution) per well. The uptake of dye was recorded for 90 min by measuring its emission under the same conditions. The entire assay was performed at room temperature. Succinate was present at all times at 15 mM.

H33342 ATP-Dependent Efflux Assay.

M. extorquens cultures were grown overnight to an OD₆₀₀ of 0.2–0.3 in Hypho medium at 30 °C. Cells were harvested by centrifugation (5,000 × g for 10 min) and washed once with medium lacking a utilizable carbon source (succinate). All following wash steps and resuspensions were performed using a succinate-free Hypho medium containing 5 μM H33342 (8,000 × g for 2 min). The OD₆₀₀ was adjusted to 1.0 before freshly prepared CCCP in DMSO was added to a final concentration of 100 μM [∼1% (vol/vol) DMSO] to abolish ATP synthesis. DMSO at 1% did not inhibit growth of the WT or ΔSHC mutant. The mixtures were incubated in darkness for 1 h, washed twice to remove CCCP, and then resuspended to an OD₆₀₀ of 1.0. One hundred eighty microliters of the suspension was transferred to wells of a 96-well plate. The initial uptake/equilibrium of H33342 was recorded using a plate reader (Envison) for 45 min [filters (wavelength/ bandwidth): excitation = 340/25 nm, emission = 405/8 nm], and readings were taken for 45 cycles (every minute) with 55 s of shaking between readouts. Afterward, either glucose or succinate was added to a final concentration of 20 mM (5 μL of 740 mM solution) per well, and the change in H33342 emission was recorded for an additional 60 min (60 cycles, 55-s delay with shaking). Glucose was used as a negative control because M. extorquens is unable to use it. All steps of the assay were performed at room temperature.