Structural and functional dissection of a higher-order oligomerization interface in yeast ceramide synthase

Abstract

Ceramide synthases (CerSs) are crucial enzymes in sphingolipid metabolism and have shown therapeutic potential for treating various metabolic disorders. However, their regulatory mechanisms remain poorly understood. In this study, we report the cryo-electron microscopy structure of a yeast CerS (yCerS), composed of a catalytic Lac1 subunit and a regulatory Lip1 subunit, organized into a higher-order 4:4 assembly. This assembly is formed by dimerization of two 2:2 Lac1-Lip1 subcomplexes via an interface primarily involving the Lac1 subunit. Notably, within this interface, the C-terminal transmembrane helix (TM8) of Lac1 adopts a dramatically twisted conformation and engages in extensive interactions with TMs 6/7/8 of the adjacent Lac1 subunit. This structural rearrangement sterically occludes the catalytic chamber and blocks acyl-CoA substrate entry. Functional assays further demonstrate that, although structurally reminiscent of an autoinhibitory conformation, this interface is required for the regulation of ceramide output and cellular adaption during perturbation of complex sphingolipid biosynthesis. Together, our findings uncover a complex oligomerization-mediated regulatory mechanism in yCerS, advancing the mechanistic understanding of ceramide synthesis control and highlighting the nuanced role of oligomerization in modulating CerS activity.

The yeast ceramide synthase assembles into a 4:4 complex through a Lac1 helix interface. This interface physically occludes the active site yet is essential for proper ceramide regulation, revealing sophisticated control beyond simple autoinhibition.

Introduction

Ceramides are central molecules in sphingolipid metabolic pathways and serve as precursors for the biosynthesis of all complex sphingolipids. The importance of ceramides is underscored by their capacity to regulate a diverse array of critical cellular processes, including apoptosis, differentiation, senescence, and stress response^1–8. Dysregulation of ceramide metabolism has been closely linked to the development and progression of various metabolic disorders, neurodegenerative diseases, and cancers^9–19. Given their essential roles in numerous biological processes, ceramide levels must be tightly controlled to maintain cellular homeostasis²⁰.

In all eukaryotes, ceramide synthases (CerSs) are enzymes embedded in the endoplasmic reticulum (ER) that catalyze the de novo ceramide synthesis by transferring an acyl chain from acyl-CoA to the amino group of a long-chain base (LCB)^21,22. Following synthesis in the ER, ceramides are transported to the Golgi apparatus, where they serve as precursors for the biosynthesis of complex sphingolipids^23–25. The ceramide synthase reaction represents a pivotal branch point in the sphingolipid metabolic network, as it not only consumes potentially toxic LCBs derived from both de novo synthesis and recycling, but also provides the necessary precursors for the production of complex sphingolipids^26,27. As key players in ceramide synthesis, CerSs have emerged as promising therapeutic targets for metabolic disorders and cancers^28–30. Moreover, inhibiting fungal CerSs holds therapeutic potential against clinically important human fungal pathogens that rely on sphingolipid synthesis for membrane integrity and virulence^31–33.

In the yeast Saccharomyces cerevisiae, two functionally redundant catalytic subunits, Lac1 and Lag1, are expressed^21,22. These subunits associate with the regulatory subunit Lip1 to form a functional complex³⁴. In contrast, mammals possess six ceramide synthase isoforms (CerS1–CerS6) that exhibit tissue-specific expression patterns and distinct preferences for fatty acyl chain lengths^35–43. Notably, no homologs of the regulatory Lip1 subunit have been identified in mammals³⁴. All CerS catalytic subunits share a conserved TRAM-Lag-CLN8 (TLC) domain, which comprises six transmembrane helices (TMs)⁴⁴. Within this domain, the Lag1p motif—a critical structural element spanning two TMs—features two conserved histidine and two conserved aspartic acid residues that are essential for catalysis⁴⁵. Overexpression of Lac1 and Lag1 in yeast results in only a modest increase in total ceramide levels²². Additionally, mammalian CerS mRNA expression varies across different tissues, yet the acyl chain composition of sphingolipids does not always correlate with CerS expression levels⁴⁰. These findings strongly imply that ceramide synthesis catalyzed by CerSs is subject to precise post-translational regulation.

Previous studies have made much progresses in elucidating the post-translational regulatory mechanisms of CerSs, particularly through processes such as dimerization and phosphorylation. Mammalian CerSs have been found to form both homo- and hetero-dimers^46,47, with their activity modulated by these dimerization events⁴⁷. This has led to the hypothesis that the acyl chain composition of ceramides in different tissues may depend on interactions between different CerS isoforms⁴⁷. A more recent investigation identified a unique C-terminal DxRSDxE motif in mammalian CerSs, which plays a pivotal role in regulating dimer formation and, consequently, impacts ceramide synthesis⁴⁸. This motif is absent in other members of the TLC family and is exclusively conserved in CerSs (except CerS1), including yeast Lac1 and Lag1 (Supplementary Fig. 1). However, the extent to which this conserved motif contributes to dimerization and regulation in yeast CerS remains to be clarified.

Beyond dimerization-mediated regulation, both yeast and mammalian CerSs are reported to undergo phosphorylation-dependent modulation. In yeast, the protein kinase Ypk1 directly phosphorylates Ser23/Ser24 in the N-terminal region of Lac1/Lag1, resulting in enhanced enzymatic activity⁴⁹. Furthermore, casein kinase 2 (CK2) regulates CerS function through phosphorylation of Ser393/Ser395/Ser397 in the C-terminal region of Lac1/Lag1, which is essential for optimal catalytic activity^50,51. Notably, this regulatory mechanism is evolutionarily conserved, as human CerS2-6 similarly undergo CK2-mediated phosphorylation at corresponding C-terminal sites (Supplementary Fig. 1), leading to increased enzymatic output⁵². Intriguingly, the C-terminal phosphorylation sites are situated within or in adjacent to the DxRSDxE motif (Supplementary Fig. 1). This positioning raises the possibility of coordinated regulation through a crosstalk between phosphorylation and dimerization.

The recently published high-resolution cryo-EM structures of both the yeast CerS complexes and human CerS6 in multiple functional states have laid a foundation for understanding the catalytic and inhibitory mechanisms of CerSs^53–56. Despite these advancements, the mechanisms underlying the regulation of CerS activity remain poorly understood.

In this study, we determine that the yeast ceramide synthase Lac1-Lip1 complex can further assemble into a higher-order 4:4 complex through the association of two 2:2 Lac1-Lip1 subcomplexes via a dimerization interface involving the Lac1 subunit. Structural analysis reveals that this interface obstructs acyl-CoA substrate binding, suggesting a potential autoinhibitory conformation. However, functional studies show that this interface does not simply act as an off-switch; instead, it is required for the proper regulation of ceramide output and cellular adaptation under sphingolipid pathway stress. Thus, our findings uncover a complex oligomerization-mediated regulatory mechanism in yCerS.

Results

Biochemical characterization of the oligomeric states of the Lac1-Lip1 complex

In our previous studies, we determined the structures of the Lac1-Lip1 complex in both substrate-bound⁵³ and inhibitor-bound⁵⁴ states. All these structures were resolved as a 2:2 Lac1-Lip1 complex, where two 1:1 Lac1-Lip1 holoenzymes dimerize via Lip1. Intriguingly, during purification, a higher-order oligomeric form of the Lac1-Lip1 complex was consistently observed in size-exclusion chromatography (SEC) prior to the peak corresponding to the 2:2 Lac1-Lip1 complex, as confirmed by blue native (BN)-PAGE anaylysis (Fig. 1A). To estimate the relative abundance of higher-order oligomers, we pooled all fractions in Fig. 1A, analyzed the mixture by BN-PAGE and performed semi-quantitative densitometry; higher-order species constitute ~25% of the purified protein (Supplementary Fig. 2A). To examine whether Lac1-Lip1-Lag1 heterotetramers can also form higher‑order oligomers, we co‑expressed Strep‑tagged Lac1, Flag‑tagged Lag1, and untagged Lip1 in HEK 293F cells and purified the heterocomplex using tandem affinity purification. BN‑PAGE analysis of the purified material showed that the Lac1-Lip1-Lag1 complex likewise assembles into higher‑order oligomers (Supplementary Fig. 2B). Cross-linking experiments with glutaraldehyde (GA) on the later-eluting peak (peak 2) fractions verified the formation of the 2:2 Lac1-Lip1 complex (~132.4 kDa) in solution (Supplementary Fig. 2C). In contrast, the fractions from the earlier elution peak (peak 1) was cross-linked to form higher molecular weight oligomers, consistent with the estimated molecular weight of a 4:4 Lac1-Lip1 complex (~264.8 kDa) (Supplementary Fig. 2C).

Fig. 1. Structural determination of the 4:4 Lac1-Lip1 complex.

A Characterization of the oligomeric states of the purified Lac1-Lip1 complex. Left: representative SEC profile. Right: SDS-PAGE (top) and BN-PAGE (bottom) of the indicated SEC fractions, stained with Coomassie blue. Data are representative of three independent experiments. B Representative cryo-EM micrograph and 2D class averages of proteins from SEC peak 1. Blue and green circles mark representative 4:4 and 2:2 Lac1-Lip1 particles, respectively. C Perpendicular views of the cryo-EM map and structure of the 4:4 Lac1-Lip1 complex. The assembly comprises two 2:2 subcomplexes related by C2 symmetry. Source data are provided as a Source Data file.

To examine the oligomeric states of the Lac1-Lip1 complex in vivo, we enriched the endogenously expressed complex via anti‑Flag immunoprecipitation from ΔLag1 cells carrying chromosomally Flag-tagged Lac1 and then cross‑linked the captured material. SDS‑PAGE resolved two discrete bands whose mobilities matched those of the cross‑linked 2:2 and 4:4 assemblies seen with overexpressed protein (Supplementary Fig. 2C, D), confirming that both oligomers form natively in cells. We acknowledge that our immunoprecipitation-based analysis of endogenous complexes likely does not capture their true dynamic equilibrium or absolute stoichiometric ratio in vivo. Future studies will be required to quantitatively assess the native oligomeric landscape of yCerS.

Structural determination of the 4:4 Lac1-Lip1 complex

To unravel the structural basis underlying the higher-order assembly of the Lac1-Lip1 complex, we collected and concentrated the fractions from peak 1 of SEC (Fig. 1A), which were subsequently subjected to cryo-EM analysis (Fig. 1B). This approach differs from our previous work, where we used the later-eluting peak (peak 2) for structural determination^53,54. Detailed procedures for grid preparation, cryo-EM data collection, and structural determination are provided in Supplementary Fig. 3 and the “Methods” section. In line with the incomplete separation of the 2:2 and 4:4 oligomeric forms of the Lac1-Lip1 complexes by SEC, two distinct sizes of particles can be observed in the original cryo-EM micrographs, likely corresponding to the 2:2 complex and the 4:4 complex, respectively (Fig. 1B). According to the 3D classification results in Supplementary Fig. 3A, approximately 47% of particles belong to the 4:4 state and about 53% belong to the 2:2 state.

We then focused on the higher-order Lac1-Lip1 complex for structural determination, which yielded an EM map at a resolution of 3.36 Å with C2 symmetry imposed (Supplementary Fig. 3 and Supplementary Table 1). The well-resolved EM density of the higher-order Lac1-Lip1 complex enabled unambiguous modeling of most protein regions, immediately revealing that this higher-order complex is a 4:4 complex, formed by the further dimerization of two 2:2 Lac1-Lip1 subcomplexes (Fig. 1C). Although the map exhibits a slight anisotropy–as indicated by a cFAR value of 0.33–the overall reconstruction of the 4:4 Lac1-Lip1 complex, including the dimeric interface between Lac1 subunits, is well-resolved and structurally coherent. Consequently, the higher-order 4:4 assembly and the conformational details of the Lac1 subunits at the dimeric interface are accurately determined (Supplementary Fig. 4A). To facilitate illustration, we designated the four Lac1 molecules as Lac1-A and Lac1-B in one 2:2 Lac1-Lip1 complex, and Lac1-A’ and Lac1-B’ in the other 2:2 Lac1-Lip1 complex (Fig. 1C).

Lac1-mediated dimerization interface

The dimerization interface between the two 2:2 Lac1-Lip1 subcomplexes is exclusively established through interactions between the adjacent Lac1-B and Lac1-B’ subunits near the cytosolic membrane surface, while the majority of the transmembrane regions remain spatially separated (Fig. 1C). Upon superimposing the structures of Lac1-A and Lac1-B subunits with our previously resolved Lac1 structure in the apo 2:2 Lac1-Lip1 complex (PDB 8Y2N), we observed that the Lac1-A subunit exhibits no significant structural changes (Fig. 2A). In contrast, the Lac1-B subunit undergoes a pronouced structural rearrangement, with the final transmembrane helix (TM8) twisting approximately 45° outward at Val367 (Fig. 2A), bringing the C-terminus of TM8 into close proximity with TMs 6/7/8 of the adjacent Lac1-B’, and vice versa (Fig. 2B). This conformational shift in TM8 of Lac1-B not only involves a distinct structural change but also enables resovling of additional residues at its C-terminus, extending to Asp391, whereas both the Lac1-A subunit and the previously resolved Lac1 structure could only be modeled up to Arg385 (Fig. 2A). In addition to the protein-mediated contacts between the two 2:2 Lac1-Lip1 subcomplexes, several lipid-like densities were observed in the luminal membrane leafleat, tightly sandwiched between Lac1-B and Lac1-B’ (Supplementary Fig. 4B). This observation suggests that lipids may also contribute to the assembly and/or stabilization of the higher-order 4:4 Lac1-Lip1 complex. These lipids may serve as regulators of the higher‑order assembly, fine‑tuning the activity of the complex—a hypothesis that merits further investigation.

Fig. 2. Lac1-mediated dimerization of yeast CerS.

A Superposition of Lac1 protomors (Lac1-A and Lac1-B) from the 4:4 Lac1–Lip1 complex onto the Lac1 protomor of the apo 2:2 complex (PDB 8Y2N). B Overview of the Lac1-mediated dimer interface. C Detailed views of the interface between the twisted TM8 of Lac1-B and TMs 6/7/8 of Lac1-B’. Interacting residues are shown as sticks; potential hydrogen bonds are denoted by red dashed lines. D Interface mutations (L379A/L383A/I387A) in Lac1 abolish 4:4 Lac1-Lip1 complex formation. Data are representative of three independent experiments. E BN-PAGE analysis of WT and Lac1^{L379A/L383A/I387A}–Lip1 complexes in cell lysates, showing their oligomeric states prior to purification. Data are representative of three independent experiments. Source data are provided as a Source Data file.

On the basis of the C2 symmetry of the structure, we analyze the dimerization interface between TM8 of Lac1-B and TMs 6/7/8 of Lac1-B’ as an illustrative example. TM8 of Lac1-B penetrates into a predominantly hydrophobic groove formed by TMs 6/7/8 of Lac1-B’ (Fig. 2B and Supplementary Fig. 5), and vice versa, thereby creating a substantial interface area of 1380.9 Ų. The primary interactions at this interface are mediated by extensive hydrophobic contacts, involving residues Trp371, Ile375, Phe376, Val378, Leu379, Ile382, Leu383, Trp384, Ile387, and Leu388 from TM8 of Lac1-B, and Leu289, Ala303, Phe304, Phe307, Ala308, Phe310, Val311, Trp314, Ile315, Trp371, Ile375, Val378, Leu379, Ile382, Leu383, Ile387 from TMs 6/7/8 of Lac1-B’ (Fig. 2C). Additionally, this dimerization interface is further stabilized by two potential hydrogen bonds: one between the side chain of Tyr380 in Lac1-B and the main chain of Phe304 in Lac1-B’, and another between the main chain of Leu388 in Lac1-B and the side chain of Asn296 in Lac1-B’ (Fig. 2C).

To validate the role of the Lac1-mediated dimerization interface in the assembly of the higher-order Lac1-Lip1 complex, we introduced mutations in three key hydrophobic residues that contribute to the Lac1-mediated dimerization interface. Specifically, Leu379, Leu383, and Ile387 on TM8 were simultaneously mutated to alanine residues, resulting in the Lac1^{L379A/L383A/I387A} mutant (Fig. 2C). The Lac1^{L379A/L383A/I387A}-Lip1 mutant complex was purified using the same protocol as the WT complex. SEC and BN-PAGE analyses demonstrated that the purified Lac1^{L379A/L383A/I387A}-Lip1 mutant complex failed to form the higher-order Lac1-Lip1 complex (Fig. 2D). Furthermore, we examined the oligomeric states of detergent-solubilized microsomes containing either the WT or the mutant complex by BN-PAGE (Fig. 2E). While a band (band 1) containing both Lac1 and Lip1 subunits, migrating at a position similar to that of the purified higher-order Lac1-Lip1 complex, was observed for the the WT complex, this higher-order Lac1-Lip1 complex band was absent in the Lac1^{L379A/L383A/I387A}-Lip1 mutant complex (Fig. 2E). These findings support the notion that the specific assembly of the higher-order Lac1-Lip1 complex is mediated through the Lac1 dimerization interface and is not an artifact induced by the protein purification process.

Dimerization-induced autoinhibition

To understand the potential functional implications of Lac1-mediated dimerization, we compared the structure of the 4:4 Lac1-Lip1 complex with the previously resolved C26-CoA-bound 2:2 Lac1-Lip1 complex structure (PDB 8IZD). Upon superimposing the Lac1-A structure onto the C26-CoA-bound Lac1 structure, we observed that Lac1 maintains a highly similar overall conformation, with only minor local changes in residues near the C26-CoA binding site, likely attributable to C26-CoA binding (Supplementary Fig. 6A, B). This high degree of conformational similarity suggests that Lac1-A/Lac1-A’ in the 4:4 Lac1-Lip1 complex retain an active conformation. However, when we superimposed the C26-CoA-bound Lac1 structure onto the Lac1-B’ subunit in the dimerized Lac1-B/Lac1-B’ structure, we found that the C-terminal region of the twisted TM8 of Lac1-B severely clashes with the CoA moiety of the C26-CoA substrate from the C26-CoA-bound structure (Fig. 3A). The C-terminal segment of Lac1-B adopts a configuration akin to a plug, positioned above the catalytic reaction chamber, thereby obstructing the entry of acyl-CoA substrates (Fig. 3B). This structural observation indicates that Lac1-B/Lac1-B’ in the 4:4 Lac1-Lip1 complex may represent an autoinhibited state.

Fig. 3. Dimerization-induced autoinhibition.

A Superimposition of the C26-CoA-bound Lac1 (PDB 8IZD) onto the Lac1-B’ subunit of the dimerized Lac1-B/Lac1-B′ structure. B The C-terminal end of Lac1-B sterically clashes with the CoA moiety of the C26-CoA observed in the substrate-bound structure. C In vitro CerS activity of Lac1-Lip1 complexes collected from the two SEC peaks indicated in Fig. 1A. Each data point represents the average ± SD of three independent experiments. Statistical significance was determined by two-tailed Student’s t-test: ***P = 0.0002. D The in vitro enzymatic activity of the WT complex from SEC peak 2 was nearly identical to that of the Lac1^{L379A/L383A/I387A}-Lip1 mutant complex. Data points represent the average ± SD from six independent experiments. Statistical significance was determined using a two-tailed Student’s t-test: “ns” indicates no significance (P = 0.4294). Source data are provided as a Source Data file.

Given that we were unable to fully separate the 4:4 and 2:2 Lac1-Lip1 complexes, as evidenced by BN-PAGE and cryo-EM analyses (Fig. 1A, B), we employed fractions from both SEC peaks for enzymatic analysis. The results revealed that the Lac1-Lip1 complex from peak 1 exhibited approximately 30% lower enzymatic activity compared to that from peak 2 (Fig. 3C). According to our autoinhibition model, each 4:4 oligomer contains two inhibitory and two active Lac1 subunits. Based on this stoichiometry and the particle ratios from 3D classification (~47% 4:4, ~53% 2:2; Supplementary Fig. 3A), we estimate that about 68% of Lac1 subunits in SEC peak 1 are in an active state [(2 × 47 + 2 × 53)/(4 × 47 + 2 × 53) ≈ 0.68]. This estimate closely matches the experimentally measured activity of peak 1 (Fig. 3C).

Furthermore, the enzymatic activity of the WT complex from SEC peak 2 was nearly identical to that of the Lac1^{L379A/L383A/I387A}-Lip1 mutant complex (Fig. 3D), which exists exclusively as a 2:2 complex (Fig. 2D, E). This is consistent with the observation that the Lac1-Lip1 complex in SEC peak 2 predominantly exists as a 2:2 species (Fig. 1A), and supports that the Lac1^{L379A/L383A/I387A} mutation does not impair the enzymatic activity of the complex in vitro.

In vivo functional analysis reveals complex regulatory role of the Lac1-mediated dimerization interface

To evaluate the physiological role of Lac1-mediated oligomerization in yeast cells, we conducted phenotypic analyses through spotting assays. To eliminate the influence of the Lag1 subunit, we generated Flag-tagged Lac1 in a targeted Lag1 deletion strain (designated Lac1^WTΔLag1) and its corresponding Lac1^{L379A/L383A/I387A} mutant strain (designated Lac1^{L379A/L383A/I387A}ΔLag1). While both strains exhibited comparable colony numbers under standard, non-stress conditions, distinct phenotypes emerged during growth in the presence of aureobasidin A (AbA)—an antibiotic that inhibits complex sphingolipid formation in yeast by targeting the IPC synthase (the Aur1‑Kei1 complex in S. cerevisiae) (Fig. 4A). Specifically, the Lac1^WTΔLag1 strain exhibited a more pronounced growth defect, whereas the Lac1^{L379A/L383A/I387A}ΔLag1 strain partially alleviated this phenotype (Fig. 4B). The observed growth differences were not due to variations in protein expression levels, as immunoblot analysis confirmed similar levels of the WT and mutant proteins under the tested conditions (Fig. 4C). These distinct phenotypes under conditions of complex sphingolipid depletion underscore a functional role for the Lac1-mediated dimerization interface in vivo.

Fig. 4. In vivo functional analysis of the Lac1-mediated dimerization interface.

A Simplified scheme of ceramide and complex sphingolipid biosynthesis in yeast. Aureobasidin A (AbA) inhibits the IPC synthase complex (Aur1-Kei1) and reduces complex sphingolipid production. B Spotting assay of the Lac1 interface mutant under AbA treatment. C WT and mutant Lac1 proteins are expressed at comparable levels. D LC-MS quantification of ceramide species in Lac1^WTΔLag1 and Lac1^{L379A/L383A/I387A}ΔLag1 cells after AbA treatment. Data are mean ± SD (n = 3). Statistical significance was determined by two-tailed Student’s t-test. P values are shown in the figure. Source data are provided as a Source Data file.

The structural observation that the twisted TM8 in the 4:4 assembly occludes the acyl-CoA binding site suggested a potential autoinhibitory mechanism. To test its functional relevance in vivo, we performed lipidomic analysis of both strains after AbA treatment. Surprisingly, the Lac1^{L379A/L383A/I387A}ΔLag1 mutant cells accumulated significantly less ceramide than WT cells, particularly ceramide species bearing a C26 fatty acyl chain, which represents the major acyl-CoA substrate for Lac1 (Fig. 4D and Supplementary Table 2). In contrast, inositol phosphorylceramide (IPC) levels were comparable between WT and mutant cells (Supplementary Fig. 7A, B and Supplementary Table 2). These findings indicate that the AbA resistance phenotype of the Lac1^{L379A/L383A/I387A} mutant arises primarily from reduced accumulation of very long-chain fatty acid (VLCFA)-containing ceramides, rather than from increased synthesis of complex sphingolipids. Notably, a prior study demonstrated that defects in VLCFA biosynthesis confer resistance to growth inhibition induced by AUR1 repression upon AbA treatment in yeast⁵⁷. Together with our data, these observations suggest that the Lac1^{L379A/L383A/I387A} mutant causes the decrease in VLCFA-containing ceramide accumulation, thereby conferring AbA resistance in yeast.

This unexpected finding—disruption of the putative autoinhibitory interface reduces ceramide accumulation—challenges a simple autoinhibition model. Since ceramide accumulation is generally believed to impair cell growth, the severe growth defect observed in WT relative to the Lac1^{L379A/L383A/I387A} mutant under AbA stress is consistent with this view (Fig. 4B). The accompanying improvement in growth under AbA stress suggests that the interface mutation alters the relationship between ceramide levels and cellular fitness. While the precise mechanism by which interface disruption leads to reduced ceramide accumulation under stress remains to be elucidated, these findings unequivocally demonstrate that this interface plays an essential regulatory role in vivo—one that extends beyond simple autoinhibition and is required for appropriate CerS output under conditions of sphingolipid pathway perturbation.

The DxRSDxE motif is nonessential for Lac1-mediated dimerization of yeast CerS

Beyond the highly conserved catalytic reaction chamber, there is very limited sequence identity between yeast Lac1 and mammalian CerSs (Fig. 5A). Sepecificlly, most residues of the twisted TM8 involved in Lac1-mediated dimerization are not strictly conserved in mammalian CerSs, though they share similar hydrophobic characteristics (Fig. 5B). In contrast, the DxRSDxE motif, previously shown to be crucial for dimerization in mammalian CerSs⁴⁸, displays greater conservation across yeast and mammalian CerSs (Fig. 5B). Notably, the DxRSDxE motif of Lac1 (residues 390–396) is positioned immediately downstream of the resolved twisted TM8 of Lac1-B (Fig. 5B). However, only the first two residues of the DxRSDxE motif could be modeled in the Lac1-B/Lac1-B’ structure (Fig. 5A), suggesting that this motif possesses considerable flexibility in the current structure.

Fig. 5. The DxRSDxE motif is nonessential for Lac1-mediated dimerization of yeast CerS.

A Most residues of the twisted TM8 involved in Lac1-mediated dimerization are not strictly conserved in mammalian CerSs. Lac1 is colored by ConSurf-derived conservation scores from an alignment of yeast Lac1/Lag1 and human CerS orthologues. B Sequence logo (WebLogo) of the DxRSDxE region adjacent to TM8, showing its conservation across yeast Lac1/Lag1 and human CerS proteins. C Mutations within the DxRSDxE motif have little impact on the formation of higher-order Lac1–Lip1 complex. Data are representative of three independent experiments. Source data are provided as a Source Data file.

To investigate whether the DxRSDxE motif contributes to the Lac1-mediated dimerization of the complex, we introduced collective alanine substitutions for all seven amino acids in the DxRSDxE motif (designated as Lac1^Mut7A). We co-expressed the mutant with Lip1 and purified the complex using the same protocol as that for the WT complex. Both SEC and BN-PAGE analyses showed that the mutant complexes can still form a higher-order Lac1-Lip1 complex similar to that of the WT complex (Fig. 5C). These results indicate that the DxRSDxE motif of yeast Lac1 is nonessential for Lac1-mediated dimerization of yeast CerS.

Lac1-mediated dimerization of yeast CerS is independent of phosphorylation

Given that the C-terminal phosphorylation sites of Lac1, encompassing residues Ser393, Ser395, and Ser397, are positioned immediately downstream of the twisted TM8 involved in Lac1-mediated dimerization (Supplementary Fig. 1), we sought to explore whether phosphorylation of Lac1 at the C-terminus influences Lac1-mediated CerS dimerization. To emulate a dephosphorylated state, we substituted the three serine residues with alanine (designated as Lac1^3A), rendering the protein incapable of phosphorylation at these sites under any conditions. Conversely, to mimic a phosphorylated state, we replaced these serine residues with glutamic acid (designated as Lac1^3E). Both the Lac1^3A-Lip1 and Lac1^3E-Lip1 mutant complexes were purified using the same protocol as the WT complex (Supplementary Fig. 8A).

Firstly, we assessed the enzymatic activity of the purified mutant complexes. The results demonstrated that the purified Lac1^3A-Lip1 complex exhibited markedly diminished activity compared to the WT complex (Supplementary Fig. 8B), indicating that Lac1 in the WT complex exists in a highly phosphorylated state. This was corroborated by a Pro-Q Diamond staining assay, which directly visualized the phosphorylated proteins (Supplementary Fig. 8C). The enzymatic activity of the purified Lac1^3E-Lip1 complex was approximately twice that of the Lac1^3A-Lip1 complex, yet still substantially lower than that of the WT complex (Supplementary Fig. 8B), suggesting that Lac1^3E can only partially mimic the phosphorylated state of Lac1. Nevertheless, these results imply that the regulatory effects of C-terminal phosphorylation on the Lac1-Lip1 complex can be, to some extent, recapitulated by both the dephosphomimetic and phosphomimetic mutations.

Analyses by SEC and BN-PAGE demonstrated that a higher-order oligomeric Lac1-Lip1 complex corresponding to the 4:4 complex could be observed for both Lac1^3A-Lip1 and Lac1^3E-Lip1 mutant complexes, similar to the WT complex (Supplementary Fig. 8A). These results indicate that the regulation of yeast CerS by Lac1-mediated dimerization is likely independent of C-terminal phosphorylation. However, considering that the inherently flexible C-terminal phosphorylation region remains unresolved in all reported yeast and mammalian CerS structures, further studies are required to elucidate the molecular mechanisms by which phosphorylation modulates CerS activity.

Discussion

CerSs catalyze de novo ceramide synthesis, serving as pivotal enzymes in sphingolipid metabolism across eukaryotes and holding significant potential as therapeutic targets for metabolic disorders, cancers, and antifungal therapies. However, the molecular mechanisms governing CerS activity regulation remain poorly understood. Our structural data clearly demonstrate that the twisted TM8 in the 4:4 assembly of yCerS sterically occludes the acyl-CoA binding site, providing a plausible mechanistic basis for autoinhibition (Fig. 6, left). However, our in vivo lipidomics data present an apparent paradox: disruption of this interface (Lac1^{L379A/L383A/I387A}) leads to reduced cellular ceramide levels under AbA-induced stress (Fig. 4D). This apparent contradiction suggests that the physiological function of this interface extends beyond being a simple static inhibitor in vivo (Fig. 6, right). We propose that the interface is not merely a static “off-switch” that dampens CerS activity when engaged. Rather, it serves as a regulatory hub that couples ceramide synthesis to cellular demand, stress signaling, or metabolic flux. In this view, the interface enables context-dependent tuning of CerS output, and its disruption uncouples the enzyme from these inputs, potentially resulting in inappropriately reduced ceramide production under conditions that normally require sustained or elevated ceramide production. Thus, the paradoxical in vivo phenotype does not invalidate the autoinhibitory mechanism but instead hints that the interface operates within a broader regulatory logic.

Fig. 6. A regulatory oligomerization interface in yeast ceramide synthase.

Left, Structural basis of the 4:4 assembly. The active 2:2 Lac1-Lip1 complex (top) can further dimerize via a specific interface involving the TM8 helices of the Lac1-B/B’ subunits (bottom). In this 4:4 structure, the twisted conformation of TM8 physically occludes the acyl-CoA binding pocket (red blockade), suggesting a potential autoinhibitory mechanism. Right, Complex in vivo functional outcomes. Under conditions of complex sphingolipid depletion (AbA treatment), Lac1^WT cells (top) maintain an equilibrium between 2:2 and 4:4 assemblies and produce ceramides at regulated levels, yet exhibit growth defects. In contrast, cells harboring the Lac1^{L379A/L383A/I387A} interface mutation (bottom) fail to form the 4:4 complex, accumulate lower levels of ceramides under the same stress, yet show a partial growth rescue. This indicates that the oligomerization interface is not a simple off-switch but is essential for the proper regulation of ceramide accumulation and cellular adaptation to sphingolipid pathway stress.

Several possibilities, though speculative at this stage, could potentially account for the reduced ceramide accumulation observed in the interface mutant under stress. One possibility relates to the complex metabolic fate of ceramides within cells. Given that cellular ceramide levels reflect not only synthesis but also degradation, recycling, trafficking, and feedback regulation, the observed decrease in steady-state ceramide accumulation could stem from increased ceramide turnover, altered flux into downstream complex sphingolipids, or changes in ceramide export or recycling—processes that are not captured by in vitro activity measurements alone. Alternatively, the interface could be required for proper subcellular localization, retention, or stability of the CerS complex under conditions of sphingolipid pathway perturbation. It is also conceivable that the interface facilitates productive interactions with upstream regulators such as Ypk1 or CK2; its disruption might impair this coordination, indirectly reducing ceramide accumulation through defective phosphorylation-dependent modulation. While purely speculative at this point, one might also consider the possibility that the 4:4 assembly represents a “primed” or “regulated” state that, despite being partially inhibited, is poised for modulation in response to specific signals. Based on this hypothesis, disrupting this interface could potentially lock the complex in an unregulated 2:2 state that fails to respond appropriately under stress, leading to dysregulated ceramide synthesis. Distinguishing between these speculative models will require systematic investigation of ceramide metabolic flux, enzyme turnover, subcellular distribution, and protein–protein interaction networks in future studies.

While our research highlights a functional role of the Lac1-mediated dimerization interface in vivo under conditions of complex sphingolipid depletion, a critical gap remains in understanding the regulation of the Lac1-Lip1 complex’s oligomeric states within physiological contexts. A compelling area for future investigation would be to determine whether physiologically relevant stimuli could trigger transitions between the active and autoinhibited forms of the Lac1-Lip1 complex. This hypothesis is bolstered by prior studies demonstrating that mammalian CerS dimers form rapidly in response to curcumin-induced stimulation of ceramide synthesis, indicating dynamic regulatory transitions in oligomeric states in vivo⁴⁷. The coexistence of 2:2 and 4:4 Lac1-Lip1 assemblies in vivo, along with the mutant’s altered ceramide levels and partial growth rescue under stress, supports the view that reversible oligomerization can tune CerS activity in response to metabolic demand. Although our current study does not capture this interconversion dynamically in living cells, future real-time studies that monitor oligomer dynamics in response to metabolic signals could directly test this hypothesis. Further investigations are essential to clarify the physiological importance of Lac1-mediated dimerization in modulating yeast CerS activity and to identify potential stimuli that may drive regulatory transitions in its oligomeric states.

The recently resolved cryo-EM structure of human CerS6 (PDB 8QZ6) reveals a homodimeric conformation, with TM1 and TM3 forming the dimerization interface⁵⁵ (Supplementary Fig. 9A). The transmembrane domain of CerS6, comprising seven TMs, exhibits a highly conserved architecture relative to TMs 2-8 of Lac1 (Supplementary Fig. 9B). Notably, the final TM of CerS6 adopts a configuration analogous to that of the Lac1-A/Lac1-A’ subunits, rather than the twisted conformation observed in the Lac1-B/Lac1-B’ subunits of the 4:4 Lac1-Lip1 complex (Supplementary Fig. 9B). In the resolved CerS6 dimer structure, the DxRSDxE motif at the C-terminus of CerS6, which remains unmodeled, is oriented in opposing directions, rendering direct physical interaction between this motif in the two protomers implausible (Supplementary Fig. 9A). Nevertheless, it remains plausible that, beyond the established TM1/TM3 dimerization interface, the CerS6 dimer could further assemble into a higher-order oligomer through an additional interface involving the DxRSDxE motif—a hypothesis that warrants further investigation.

Our results revealed that the DxRSDxE dimerization motif in mammalian CerSs contributes little to the Lac1-mediated dimerization in yeast CerS, highlighting distinct oligomerization mechansims for yeast and mammalian CerSs. Consistent with the differences in the structural elements contributing to oligomerization in yeast and mammalian CerSs, it has been shown that mammalian CerS activity is increased by dimer formation⁴⁷, which contrasts with our structural indication of oligomerization-induced autoinhibition in yeast CerS. This divergence underscores the need for further research to elucidate the specific regulatory mechanisms and evolutionary adaptations that govern CerS activity regulation by oligomerization across different species.

Sphingolipid homeostasis is regulated at multiple steps of the biosynthetic pathway. Interestingly, oligomerization appears to be a recurring theme. A recent structural study of the yeast SPOT complex, which catalyzes the first and rate-limiting step, revealed that it exists as both monomeric and dimeric forms of SPOT heterotetramers⁵⁸. Although the functional consequence of SPOT dimerization remains to be fully established, the presence of dinstinct oligomeric states in this rate-limiting enzyme underscores it as a recurrent structural feature in the pathway. The observation of distinct oligomeric forms in both the SPOT and CerS complexes suggests that quaternary structural changes could represent a common—though likely mechanistically distinct—strategy for fine-tuning the activity of key enzymes during sphingolipid biosynthesis.

Despite these open questions, this study makes significant progress in the field by defining the structural architecture of a critical oligomerization interface in yCerS and demonstrating its functional relavence in vivo. The unexpected functional complexity revealed by our lipidomics data—wherein disruption of this interface decreases rather than increases ceramide accumulation—moves the narrative beyond a simple autoinhibition model and underscores the need to understand CerS regulation as an integrated, systems-level process. Although the mechanistic basis for this paradox remains to be elucidated, our work provides a conceptual framework for future investigations. Future studies aimed at identifying the signals that govern the 2:2 ↔ 4:4 equilibrium, and at elucidating how this interface integrates with phosphorylation- and lipid-mediated regulatory mechanisms, will yield deeper insights into sphingolipid homeostasis and may open avenues for modulating ceramide synthesis in disease contexts.

Methods

Protein expression and purification

The yeast CerS complex was expressed and purified as previously described⁵⁴. In brief, the cDNA sequence of Lac1 was subcloned into the pCAG expression vector with a C-terminal Flag tag, while Lip1 cDNA was subcloned into the same vector without a tag. Both subunits were co-expressed in HEK 293 F cells (Invitrogen, Cat# R79007), which were then harvested and resuspended in lysis buffer (25 mM HEPES pH 7.0, 150 mM NaCl), supplemented with a protease inhibitor cocktail (Amresco). Membranes were solubilized using 1% (w/v) LMNG (Anatrace, Cat# NG310) at 4 °C for 2 h. After centrifugation at 37,000 × g for 1 h to remove insoluble material, the supernatant was applied to anti-Flag G1 affinity resin (GenScript, Cat# L00432) for purification. The eluted protein was concentrated and further purified by SEC on a Superose 6 Increase 10/300 GL column (Cytiva, Cat# 29091596) in SEC buffer (25 mM HEPES pH 7.0, 150 mM NaCl, and 0.01% LMNG). Protein fractions from each peak were collected separately and concentrated for subsequent biochemical or cryo-EM analysis.

Blue-native PAGE (BN-PAGE)

3–12% NativePAGE™ Bis-Tris Mini Protein Gels (Invitrogen, Cat# BN1003BOX) were used to determine the oligomeric states of yCerS complex. For SEC-purified protein, samples were mixed with 4× native sample buffer and Coomassie G-250 additive before being loaded onto the gels. A high molecular weight native electrophoresis protein marker (Biosharp, Cat# BL1335A) was used as the molecular weight ladder. The anode buffer consisted of 50 mM Bis-Tris and 50 mM Tricine (pH 6.8), while the dark blue cathode buffer was prepared by supplementing the anode buffer with 0.02% (w/v) Coomassie G-250. Electrophoresis was carried out at a constant 150 V for 90 min at RT, followed by Coomassie Blue staining for protein visualization.

To analyze the oligomeric states of yCerS complex in detergent-solubilized microsomes, HEK 293 F cells expressing C-terminally Flag-tagged Lac1 and N-terminally Twin-Strep-tagged Lip1 were harvested and resuspended in lysis buffer supplemented with 1% LMNG and protease inhibitors. The lysates were then solubilized by gentle rotation for 2 h at 4 °C. After centrifugation at 18,000 × g for 1 h, the supernatant was collected. Samples were mixed with 4× native sample buffer and Coomassie G-250 additive before being loaded onto the gels. Electrophoresis was initiated at 150 V for 15 min at RT using dark blue cathode buffer, followed by a switch to the light blue cathode buffer (anode buffer supplemented with 0.002% (w/v) Coomassie G-250) for an additional 75 min. Western blotting was subsequently performed to detect the target proteins.

Cross-linking assay

For cross-linking of purified proteins, 1.8 μg of SEC-fractionated protein was treated without or with glutaraldehyde (GA, Sigma-Aldrich, Cat# G5882) at final concentrations of 0.025%, 0.1%, or 0.4% (v/v) in SEC buffer. Each 20 μL reaction was incubated at 25 °C for 50 min and terminated by adding 100 mM Tris-HCl (pH 7.5). The samples were then resolved by SDS-PAGE and visualized via Coomassie Blue staining.

To probe the endogenous Lac1-Lip1 complex by cross-linking, we enriched it via anti-Flag immunoprecipitation from ΔLag1 cells expressing Flag-tagged Lac1 at its native genomic locus. The immunoprecipitated material was incubated at 25 °C for 40 min in a 5 μL reaction containing either no cross-linker or 0.025% or 0.1% (v/v) GA, quenched with 100 mM Tris-HCl (pH 7.5), and analyzed by SDS-PAGE and Western blot.

Western blot

Proteins were transferred to a PVDF membrane as described in respective protocols. The membrane was rinsed with deionized water, blocked with 5% non-fat milk in TBST buffer for overnight at RT, and then incubated for 2 h with primary antibodies against Flag or Strep Tag II (1:5000 dilution; Sangon, Cat# D191041 and Cat# D191106). After three washes with TBST, the membrane was incubated with HRP-conjugated goat anti-mouse IgG secondary antibody (1:10,000 dilution; Sangon, Cat# D110087) for 1 h, followed by three additional washes with TBST. Signal detection was performed using ECL reagents (Epizyme, Cat# SQ203), and images were captured with the Tanon 6100C chemiluminescent imaging system.

In vitro CerS activity assay

The in vitro ceramide synthase activity assay was conducted as previously described^53,54. Briefly, the standard reaction mixture consisted of 0.4 mM DTNB (Sigma-Aldrich, Cat# D8130), 100 μM DHS (Sigma-Aldrich, Cat# D3314), 200 μM C26-CoA (Avanti Polar Lipids, Cat# 870726 P), and 10% (v/v) DMSO in SEC buffer. Reactions were initiated by the addition of 300 nM purified yCerS protein and were carried out in 96-well microplates with a final volume of 50 μL at 37 °C. Absorbance at 412 nm was continuously monitored over a 60-min period using a BioTek Synergy H1 microplate reader. Enzymatic activity was determined from data points within the linear range. Statistical analysis was performed using GraphPad Prism 10.

Cryo-EM sample preparation and data collection

Fractions from peak 1 of SEC were pooled for cryo-EM analysis. A 3 μL aliquot of purified protein (12.5 mg/mL) was applied to glow-discharged Quantifoil R1.2/1.3 300-mesh copper grids. Grids were blotted for 4.5 s at 8 °C and 100% humidity with a blot force of 20, and plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). Automated data collection was performed on a 300 kV Titan Krios electron microscope equipped with a GIF Quantum energy filter (20 eV slit width; Gatan) and a K3 Summit direct electron detector (Gatan), controlled by EPU software. Super-resolution movies were acquired at a nominal magnification of ×81,000, with a defocus range of −2.0 to −1.0 μm. Each movie consisted of 32 frames over 2.3 s, yielding a total electron dose of 50 e⁻/Ų.

Cryo-EM data processing

A dataset of 2608 micrographs were collected. Micrographs were motion-corrected using MotionCor2⁵⁹ with dose weighting, and defocus values were estimated with Gctf⁶⁰. The collected micrographs were curated and selected using a CTF resolution cutoff of 4 Å and an astigmatism threshold of less than 500 Å. Initial particle picking and extraction in Relion 3.0⁶¹ yielded 5,455,667 particles, which were subsequently processed in CryoSPARC⁶². Following 2D classification, 1,242,928 selected particles were subjected to ab initio reconstruction and heterogeneous refinement. Two distinct 3D classes exhibiting reasonable secondary structural features were identified: one corresponding to the Lac1-Lip1 (2:2) complex, comprising 32% of particles, and the other representing the Lac1-Lip1 (4:4) complex, accounting for 47% of particles. For the Lac1-Lip1 (2:2) particles, an additional round of heterogeneous refinement was performed, followed by non-uniform refinement with C1 and C2 symmetry, respectively. This yielded final 3D reconstructions at overall resolutions of 3.47 Å (C1) and 3.15 Å (C2) from 124,884 particles. For the Lac1-Lip1 (4:4) particles, a further round of 2D classification, heterogeneous refinement, and non-uniform refinement resulted in 219,798 high-quality particles. Subsequent local refinements under C1 and C2 symmetry produced final maps with overall resolutions of 3.59 Å and 3.36 Å, respectively. In an alternative approach, C2 symmetry expansion of the non-uniform refined particles, followed by C1 local refinement, generated a 3.37 Å map. Further local refinement using a mask focused on one yCerS protomer together with the Lac1-Lac1’ interface improved the resolution to 3.25 Å. All resolutions were determined using the gold-standard Fourier shell correlation 0.143 criterion⁶³ with high-resolution noise substitution⁶⁴.

Model building and refinement

The previously determined cryo-EM structure of the apo 2:2 Lac1-Lip1 complex (PDB 8Y2N) served as the initial model for the 4:4 assembly. The 2:2 Lac1-Lip1 model was docked into the 4:4 Lac1-Lip1 map in UCSF Chimera⁶⁵. Every residue was manually inspected and adjusted in Coot⁶⁶. Real-space refinement, incorporating secondary-structure and geometry restraints, was performed in Phenix⁶⁷. Model quality was validated with the tools in Phenix and with MolProbity⁶⁸; refinement statistics are summarized in Supplementary Table 1. All structural figures were prepared with PyMol⁶⁹ or Chimera.

Yeast growth assay

Saccharomyces cerevisiae strains (BY4742 background) were constructed via in vivo homologous recombination. The LAC1 gene was initially disrupted to generate a ΔLac1 strain. Subsequently, either the WT or mutant Flag-tagged LAC1 gene was precisely reintegrated at its native genomic locus. Finally, the LAG1 gene was deleted in the reconstituted strains. All mutations and genomic integrations were verified by PCR and DNA sequencing.

Yeast cells were inoculated in YPD medium (1% yeast extract, 2% peptone, 2% glucose) from an overnight pre-culture and grown to the exponential phase, followed by serial tenfold dilution in fresh YPD. Aliquots of 4 μL from each dilution were spotted onto YPD agar plates supplemented with the specified concentrations of AbA (Macklin, Cat# A937715). The plates were incubated at 30 °C for 2 days, and images were captured to assess colony size and density. A C-terminal Flag tag was appended to both WT and mutant Lac1; expression levels were verified by Western blotting with an anti-Flag antibody. Equal loading was confirmed using an anti-Histone H2B antibody (1:2000; Zenbio, Cat# 381041).

Lipidomics analysis

Yeast strains were grown to log phase in YPD medium supplemented with 10 nM AbA. Approximately 20 OD₆₀₀ equivalents of cells were harvested, washed, and snap-frozen in liquid nitrogen. Total protein concentration in yeast cells was determined using the Coomassie Protein Assay Reagent (Thermo Fisher Scientific), with BSA as the standard.

Ceramides were extracted as described previously⁷⁰, with the addition of internal standard (C12-ceramide). After drying under nitrogen, samples were reconstituted in 200 µL methanol. Sphingolipid analysis was performed using high-performance liquid chromatography (AB SCIEX ExionLC AD) coupled to a triple quadrupole mass spectrometer (AB SCIEX 4500). A Luna 3 µm C8 column (2.0 × 150 mm; Phenomenex) was used for separation. The mass spectrometer was operated in positive ionization mode. The mobile phase consisted of solvent A (water with 0.2% formic acid and 2 mM ammonium formate) and solvent B (methanol with 0.2% formic acid and 1 mM ammonium formate). The gradient was as follows: 80% B (0–4 min), increased linearly to 90% B (4–14 min), then to 99% B (14–34 min), held at 99% B (34–40 min), and finally decreased to 80% B at 40.1 min.

IPC analysis was performed by LipidALL Technologies as reported previously⁷¹. Briefly, lipids were extracted from yeast cells using a modified Bligh-Dyer protocol⁷². Cells were homogenized in chloroform:methanol (1:2, v/v), incubated (4 °C, 1 h), and subjected to phase separation with water:chloroform (1:1, v/v). After re‑extraction, the combined organic extracts were dried (SpeedVac, OH mode) and stored at −80 °C until analysis. Polar lipids were analyzed using normal‑phase (NP)-HPLC (Shimadzu Nexera 20‑AD) coupled to a Sciex QTRAP 6500 PLUS mass spectrometer. Individual lipid classes were separated using a TUP‑HB silica column (150 × 2.1 mm, 3 µm) with mobile phase A (chloroform:methanol:ammonium hydroxide, 89.5:10:0.5, v/v/v) and B (chloroform:methanol:ammonium hydroxide:water, 55:39:0.5:5.5, v/v/v/v). MRM transitions were set up for comparative analysis of diverse polar lipid species. C8-GluCer (Avanti Polar Lipids) was used as internal standard for quantification.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

X.G. conceived and supervised the project. X.G., Q.F. and T.X. designed the experiments. Q.F. and T.X. conducted the experiments. C.Y. and N.Y. contributed to the lipidomics analysis. All authors contributed to the data analysis. X.G. wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Jan-Hannes Schaefer and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The EM density maps and atomic coordinates for the 4:4 Lac1-Lip1 complex have been deposited in the EMDB under accession code EMD-65613 and in the PDB under the accession code 9W3Z, respectively. PDB codes of previously published structures used in this study are 8Y2N, 8IZD, and 8QZ6. All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Information. Source Data are provided as a Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71272-8.