Structural dissection of ergosterol metabolism reveals a pathway optimized for membrane phase separation

Abstract

Sterols are among the most abundant lipids in eukaryotic cells yet are synthesized through notoriously long metabolic pathways. It has been proposed that the molecular evolution of such pathways must have required each step to increase the capacity of its product to condense and order phospholipids. Here, we carry out a systematic analysis of the ergosterol pathway that leverages the yeast vacuole's capacity to phase separate into ordered membrane domains. In the post-synthetic steps specific to ergosterol biosynthesis, we find that successive modifications act to oscillate ordering capacity, settling on a level that supports phase separation while retaining fluidity of the resulting domains. Simulations carried out with each intermediate showed how conformers in the sterol's alkyl tail are capable of modulating long-range ordering of phospholipids, which could underlie changes in phase behavior. Our results indicate that the complexity of sterol metabolism could have resulted from the need to balance lipid interactions required for membrane organization.

Fig. 1.. Biosynthesis of sterols and their role in membrane phase separation.

(A) Chemical structure of the fungal sterol ergosterol, with highlighted modifications grouped according to their occurrence during its metabolism. The canonical sterol pathway yields the ring system, conserved modifications among fungi and animals result in demethylation of α-face carbons (indicated by a Pacman symbol at removed methyl groups), while post-synthetic modifications exclusive to ergosterol primarily remodel the alkyl chain. (B) Schematic of the Bloch hypothesis, which proposes that the capacity for sterols to condense and order phospholipids rose continuously along the pathway. (C) Liquid phase separation between L_d and L_o domains is a sterol-dependent membrane property that can be observed in both synthetic systems [giant unilamellar vesicle (GUV)] and in cells (the yeast vacuole). In both GUVs and vacuoles, changes in ergosterol concentrations drive the loss or appearance of L_o/L_d phase separation. Shown are 3:1 dioleoyl PC (DOPC):dipalmitoyl PC (DPPC) GUVs prepared with 0 (left) and 20 mol % (right) ergosterol (top) and yeast vacuoles from an ergosterol depleted strain (ERG9 knockdown, left) compared to WT (right). Scale bars, 5 μm.

Fig. 2.. Early-stage intermediates show limited capacity for L_o/L_d phase separation.

(A) Schematic of canonical sterol-forming steps and conserved modifications, shared between ergosterol and cholesterol synthesis. FPP, farnesyl pyrophosphate. (B) Sterol composition of knockdowns in each early-stage step as measured by gas chromatography–mass spectrometry (GC-MS). n = 3 independent cultures; error bars = SD. (C) Representative vacuole morphologies for each strain, shown as three-dimensional (3D) projections generated from confocal z-stacks. Among mutants, only ERG24 knockdown still yields robust L_o domain formation, likely because of increased residual ergosterol content. Scale bars, 5 μm. Corresponding wide-field images with multiple cells are shown in fig. S1A. (D) Quantification of uniform (L_o) and nonuniform (solid) domain formation frequency in stationary phase cells. Gel-like domains are abundant in mutants that accumulated squalene or 2,3-oxidosqualene. n = 3 independent cultures with n > 100 cells each; error bars = SEM. (E) Early-stage uncyclized intermediates squalene (SQ) and 2,3-oxidosqualene (OX) at 15 mol % cause expansion of solid or gel-like domains formed by 5 mol % of the glycosphingolipid glucosylceramide (GlcCer) in 3:1 DOPC:DPPC GUVs. In contrast, ergosterol supports fluid L_o/L_d domains (right). The presence of multiple domains in GUVs allowed to equilibrate below the mixing temperatures indicates non-fluid domains that do not coalesce. Scale bars, 5 μm.

Fig. 3.. Post-synthetic sterol intermediates show a non-monotonic increase in ability to support membrane phase separation.

(A) Schematic of post-synthetic steps transforming zymosterol into ergosterol. (B) Sterol composition of mutants that isolate each of the post-synthetic steps as measured. Mutants in bold predominantly produce an ergosterol intermediate in the canonical pathway shown in (A), while those that are unbolded produce noncanonical intermediates as shown in fig. S4. n = 3 independent cultures; error bars = SD. (C) Representative Pho8-GFP distribution in each strain shown as 3D projections generated from Airyscan confocal z-stacks. Only erg3∆, erg5∆, and erg3∆erg5∆ and WT cells show robust domain formation in the vacuole, which is most abundant in the latter that synthesizes ergosterol. Scale bars, 5 μm. Wider-field images with multiple cells are shown in fig. S1A. (D) Quantification of domain formation frequency in fused vacuoles of stationary phase cells in 0.4% glucose. Mutants in bold predominantly produce an intermediate in the canonical pathway shown in (A). Ordering of strains and arrow indicating pathway progression correspond to accumulated intermediates, showing each canonical intermediate in order (bold), followed by their matching noncanonical products formed by Erg3, Erg5, and/or Erg4 reactions. Domains arise along the post-synthetic steps. n = 3 independent cultures; error bars = SEM.

Fig. 4.. Structural features of ergosterol intermediates that support membrane domains.

(A) Micrographs of GUVs composed of a 3:1 ratio of DOPC:DPPC with 20 mol % of sterols extracted from each post-synthetic mutant. GUVs contain the L_d marker Texas Red DHPE and domains were allowed to coalesce at room temperature for >1 hour. Micrographs are arranged by their metabolic basis, with enzyme activity that converts their dominant sterol species shown through reaction arrows. In mixtures that support phase separation, the type is indicated by the inset text. Liquid domains are characterized by their capacity to coalesce. Extracts corresponding to yeast strains that show vacuole domains are highlighted in orange. The names of strains accumulating canonical intermediates are bolded. Quantification of domain type frequency across preparations is provided in fig. S7. Scale bars, 5 μm. (B) Structures of the predominant sterols from each strain categorized by their ability to either promote gel-like domains (solid), L_o domains (liquid), or no domains (none) in GUVs. B-ring and tail structural features are color coded to highlight differences in each category. Solid domain-promoting sterols contain a ∆⁸ B-ring, which is isomerized by Erg2, with the exception of ergosta-7-en-3β-ol that showed solid domains. Sterols that cannot form domains under these conditions feature a ∆⁷ B-ring and have an unsaturated branched tail. Sterols that form liquid (L_o) domains have a ∆⁷ B-ring and a tail that is reduced at C24(28) by Erg4.

Fig. 5.. Ordering capacities of ergosterol intermediates that support domains.

(A) The change in Laurdan GP for unsaturated (POPC) phospholipid vesicles upon incorporation of 20 mol % sterol was measured for extracts from each of the post-synthetic mutants at 30°C. ΔGP was calculated by subtracting the GP of POPC without sterol (−0.245 ± 0.011) from that measured with each of the sterol-containing samples. Mutants that accumulate canonical intermediates are indicated with bolded names. Strains whose sterols support membrane domains in GUVs are highlighted according to the type of domain. Liquid domain supporting sterols show low condensation of unsaturated phospholipids. n = 3 independent liposome preparations; error bars = SD. (B) The change in GP for saturated (DMPC) phospholipid vesicles upon incorporation of 20 mol % sterol at 30°C. Domain supporting sterols show moderate to high ordering of saturated phospholipids, which corresponds to the miscibility temperatures of the mixtures (fig. S8C). The GP value for pure DMPC was −0.078 ± 0.018. (C) The difference in GP between DMPC and POPC systems, both containing 20 mol % sterol, is largest for the ergosterol intermediates that also support liquid phase separation. (D) A model highlighting non-monotonic changes in ordering capacity in the ergosterol pathway and relationship to phase separation of coexisting domains of different types.

Fig. 6.. Atomistic models for phospholipid condensation by post-synthetic intermediates.

(A) Examples of representative structures of post-synthetic sterol intermediates in all-atom simulations. The arrow in the fecosterol image highlights the smooth α-face of the ring system, while that in the ergosterol image highlights the extended alkyl tail. (B) Increase in acyl chain ordering parameter (ΔS_CD) in with 20 mol % post-synthetic intermediates compared to pure DMPC. Systems are referred to by gene deletion corresponding to the subsequent enzyme in the pathway that acts upon them, e.g., erg2∆ for fecosterol extracted from erg2∆erg3∆erg5∆erg4∆, erg3∆ for episterol from erg3∆erg5∆erg4∆, and erg5∆ ergosta-5,7,24(28)-trienol from erg5∆erg4∆. (C) ΔS_CD in simulations correlates with measured miscibility temperatures (T_misc) of the same mixtures (R² = 0.915). Sterol-free DMPC has a T_misc of 24°C. Error bars = SD. (D to F), Correlation of S_CD with tail configuration shape relative to the sterol ring. For each dynamic conformation, 3D axes (indicated by the black axis in each panel) are determined by the moments of inertia of the ring system. Displacements of tail carbons are measured for each 3D axis, squared, and summed. The deviation of that sum from the mean is plotted on the x axis. The size of each mark represents their relative abundance for that specific sterol, which is indicated by its color and shape. Plotted are deviations out of the axis out of the ring plane (D), within the axis in the ring plane but perpendicular to the sterol long axis (E), and along the sterol long axis (F). Small deviations out of the ring plane are positively correlated with order, as are extensions along the axis. There is no correlation for deviations within the ring plane. These features are consistent with an extended tail conformation correlating with ordering. Error bars = SD of three independent simulations.

Fig. 7.. Post-synthetic intermediates modulate long-range ordering of saturated phospholipids.

(A) The frequency of the high-order DMPC populations, which contains mostly trans acyl chain conformers, increases in abundance with higher ordering sterols. Error bars and total distributions for each system and sterol-free DMPC are shown in fig. S10A. (B) The increase in higher order configurations tightly correlates with total ordering for each sterol system. Error bars = SD. (C) Sterol structural changes manifest in long-range ordering of DMPC. Panels show top-down view of the all-atom simulation with sterols, which are rendered as yellow circles, and DMPC chains, which are rendered as circles colored according to their order parameter as indicated in the key. Transient pockets of ordered DMPC acyl chains (blue) are observed in all systems but are most extensive in the solid-domain forming fecosterol (left) and least extensive in sterols that do not allow for demixing, like 24(28)-dehydroergosterol (middle). Ergosterol (right), which supports fluid domains, shows an intermediate level. Movies S1 and S2 show the dynamics of these components during simulation and further illustrate the differences between these three systems. In movie S2, DMPC acyl chains are color coded by S_CD, while, in movie S1, they are a uniform coloring to better highlight dynamics.