Architecture of an asymmetric short chain/long chain hybrid acyl‑CoA carboxylase from Mycobacterium smegmatis

Abstract

Mycobacteria comprise an uneven number of genes coding for biotin carboxylase (BC) and acyl-CoA carboxyl transferase (CT) activities and an ε-subunit of unknown function. To unravel the composition of acyl-CoA carboxylase (ACCase) holoenzymes under native conditions, we pulled down endogenously expressed biotinylated proteins from Mycobacterium smegmatis. The extract included an unexpected 16-subunit ACCase holoenzyme, exhibiting distinct long-chain (LC) and short-chain (SC) CT activities. It consists of a central two-layered CT (AccD4)₂/(AccD5)₄ assembly that interacts with a C-terminal AccE5 dimer within its central hole. On each CT hexamer face, an asymmetric BC AccA3 tetramer is connected through a flexible AccE5 segment. This is preceded by a folded AccE5 segment that contributes to an asymmetric nine-stranded β−barrel with contributions from all four AccA3 subunits. This barrel allows the BC assembly to rotate ~90^o in the presence of acyl-CoA substrates. These data demonstrate that asymmetric, multi-substrate ACCases differ fundamentally from symmetric, single-substrate ACCases.

Structural analysis of endogenously expressed proteins in Mycobacterium smegmatis reveals an asymmetric short-chain/long-chain hybrid acyl‑CoA carboxylase that comprises 16 subunits with (AccA3)4–AccE5–(AccD4)₂/(AccD5)₄–AccE5–(AccA3)4 stoichiometry.

Introduction

ACCases are large multienzyme complexes that catalyze the ATP-dependent carboxylation of acyl-CoA derivatives of various lengths to generate precursors of various fatty acids and lipids with crucial implications in cancer, diabetes, obesity, as well as in infectious diseases^1–5. ACCases generally consist of BC and CT subunits within higher oligomeric assemblies, in which BC-linked biotin carboxyl carrier protein (BCCP) domains transfer BC-produced carboxybiotin to the CT active site for acyl-CoA carboxylation. All ACCases characterized to date have distinct acyl-CoA substrates and are arranged in highly symmetric complexes^1,6,7. Many mycobacteria contain an expanded repertoire of unevenly occurring ACCase genes, due to specific requirements for the biosynthesis of a complex cell envelope enriched with mycolic acids, which are critical for bacterial survival and pathogenicity^7,8. Although the existence of mycobacterial ACCase complexes with multiple substrate specificities was proposed previously⁸, the overall composition of these ACCase genes in stoichiometric holo complexes has remained unknown. Furthermore, although a functional role of the ε-subunit AccE5 in promoting catalytic AccD5 CT activity was established^9,10, any mechanistic insight into its contribution to the corresponding holo complex remained elusive.

Results and discussion

To uncover this conundrum, we purified endogenous ACCase complexes from Mycobacterium smegmatis (Mycolicibacterium smegmatis). Among the previously established AccD1/AccA1 and AccD2/AccA2 dodecameric assemblies⁷, we identified and characterized an endogenously expressed hybrid ACCase complex, which is composed of one distinct BC subunit (AccA3), two different CT subunits (AccD4, AccD5) and a unique ε-subunit (AccE5) subunit (Fig. 1a–d, Supplementary Table 1). Consistent with the presence of two different CT subunits, this hybrid complex exhibited catalytic activity for both SC (acetyl-CoA, propionyl-CoA) and LC (heptadecanoyl-CoA, stearoyl-CoA) substrates (Fig. 1e).

Fig. 1. Endogenous purification and catalytic activity of the AccA3/AccD4/AccD5/AccE5 holo complex.

a, c Strep-Tactin affinity purification of biotinylated proteins from cell lysates of M. smegmatis mc²155 and M. smegmatis (ΔaccD1–ΔaccA1, ΔaccD2–Δ accA2) strains. SDS–PAGE of fractions eluted with 50 mM biotin. Lanes M: molecular-weight marker; lanes 1–3: sequential elutions with 50 mM biotin. Bands correspond to pyruvate carboxylase (PC, ~130 kDa), AccA1/AccA3 (~70 kDa), AccD1/AccD2/AccD4/AccD5 (~60 kDa), and AccE5 (~17 kDa). The estimated molecular weights are in agreement with the mass spectrometry data, except for AccE5, which appeared at a higher apparent molecular weight on SDS–PAGE, probably due to its partly intrinsically unfolded nature (Supplementary Table 1). As expected, there is no AccA1, AccD1, and AccD2 in the cell extract of the M. smegmatis (ΔaccD1–ΔaccA1, ΔaccD2–ΔaccA2) strain, according to the mass spectrometry data. AccA2 was not detectable under the experimental conditions. b, d size-exclusion chromatography profile of the purified AccA3/AccD4/AccD5/AccE5 holo complex and SDS–PAGE of fractions across the main elution peak. Fractions, used for biochemical and structural analysis, are marked by a black bar; fractions that also contain pyruvate carboxylase are marked by a grey bar and have been discarded from further characterization. e activity measurements on SC substrates (acetyl-CoA, propionyl-CoA) and LC substrates (heptadecanoyl-CoA, stearoyl-CoA). Heptadecanoyl-CoA and stearoyl-CoA did not ionize under the conditions used and therefore are not shown. The ion chromatograms of substrates (black) and products (red) are plotted at 0 and 2 hours: acetyl-CoA (m/z = 810.1330) & malonyl-CoA (m/z = 854.1229), propionyl-CoA (m/z = 824.1487) & methylmalonyl-CoA (m/z = 868.1385), 2-carboxy-heptadecanoyl-CoA (m/z = 1064.3576) and 2-carboxy-stearoyl-CoA (m/z = 1078.3733).Each plot is an overlay of three replicates. The maximum ion count in each measurement has been set to 100%. Uncropped gels are shown in Supplementary Fig. 5.

We then determined the high-resolution cryo-EM structures of this assembly, revealing a 16-subunit holoenzyme architecture with (AccA3)₄–AccE5–[(AccD4)₂/(AccD5)₄]–AccE5–(AccA3)₄ stoichiometry, both in the absence and presence of LC and SC acyl-CoA substrates (Table 1, Fig. 2a–c). Its core is formed by a central hybrid CT assembly with a hexameric (AccD4)₂/(AccD5)₄ stoichiometry in a double-ring arrangement with 32-symmetry despite the mixed AccD4/AccD5 composition.

Table 1.

Cryo-EM data collection, refinement, and validation statistics

	A3/D4/D5/E5	(A3)4/E5	A3/D4/D5/E5	A3/D4/D5/E5
Ligand			propionyl-CoA	arachidyl-CoA
EMDB ID	55,717	55,718	55,802	56,159
PDB ID	9T96	9T97	9TDM	9TR9
Data collection and processing
Magnification	130,000		130,000	130,000
Voltage (kV)	300		300	300
Detector/energy filter	K3/20 eV		K3/20 eV	K3/20 eV
Electron exposure (e–/Å2)	44.4		44.1	44.0
Initial particle images	1,279,569		639,8521	353,856
Final particle images	251,974		52,469	76,111
Map resolution (Å)	2.4	3.2	2.4	2.4
Refinement
Initial model used	AF3 model		#1	#3
Model resolution (Å)	2.3	3.3	2.4	2.4
CC mask	0.90	0.86	0.85	0.81
Model composition
Non-hydrogen atoms	42,209	15,386	57,621	40,162
Protein residues	5563	2034	7565	5253
Ligands	5	1	11	4
B factors (Å2)
Protein	81	80	56	58
Ligand	132	52	102	179
R.m.s. deviations
Bond lengths (Å)	0.0116	0.0106	0.0123	0.01
Bond angles (°)	1.82	1.67	1.79	1.65
Validation
MolProbity score	0.91	1.13	0.79	0.86
Clashscore	0.58	1.34	0.37	1.0
Poor rotamers (%)	0.23	0.32	0.63	0.46
Ramachandran plot
Favored (%)	96.65	96.04	97.14	97.70
Allowed (%)	3.19	3.86	2.63	2.22
Disallowed (%)	0.16	0.10	0.23	0.08

Fig. 2. Overall arrangement of the AccA3/AccD4/AccD5/AccE5 holo complex with dual substrate specificity for LC and SC acyl-CoAs.

a Domain structure of subunits participating in complex formation, proportional to sequence length. Colors of structurally visible segments: AccA3, green shades; AccA3 BCCP domains, magenta; AccD4, violet; AccD5, blue/cyan shades; AccE5: orange/red. Abbreviations: β, β-barrel; +, positively charged BCCP linker; *, flexible linker in AccE5, connecting BC and CT assemblies; overall architecture of the AccA3/AccD4/AccD5/AccE5 holo complex without (b) and with (c) acyl-CoA substrates. Left, density presentation; central, sliced presentation across the center of the complex; right, sliced sections of (AccA3)₄AccE5 BC (S1) and (AccD4)₂/(AccD5)₄/(AccE5)₂ CT (S2) assemblies in top view. Structural elements are labeled, and geometric parameters are indicated. Acronym shortcuts used: AccA3, A3; AccD4, D4; AccD5, D5, AccE5, E5.

A unique feature of the complex is the critical anchoring and conformational flexibility of the AccE5 subunit dimer, which forms a central axis of this complex (Fig. 2b). Each AccE5 subunit is divided into an N-terminal α/β−hairpin motif and a small C-terminal α−helical hairpin domain, connected by a flexible linker (Figs. 2a, 3a–c). The AccE5 dimer places each of the two AccE5 subunits within one AccD4/(AccD5)₂ layer of the central CT assembly to connect to the vis-à-vis AccA3 tetramer through its flexible linker. The AccE5 C-terminal hairpin domain mediates distinct interactions with each AccD4 subunit mediated through its C-terminal helix, while the remaining hairpin structure interacts with both AccD5 subunits (Fig. 3a, c). These data illustrate that the AccE5-mediated interactions are critical for forming the observed AccD4/(AccD5)₂ hybrid assembly within each CT layer, as opposed to established homo-oligomeric CT assemblies without an AccE5-like subunit from reconstituted ACCase complexes^1,6,11.

Fig. 3. Structural elements that generate asymmetry, flexibility, and multi-substrate specificity of the AccA3/AccD4/AccD5/AccE5 holo complex.

a CT AccD4/ AccD5/AccE5 α-hairpin module; b BC (AccA3)₄/AccE5 β-barrel in the presence (central) and absence (right) of acyl-CoA ligands, indicating a 93° rotation between the two states. c cartoon presentation of the AccA3/AccD4/AccD5/AccE5 holo complex orientated as in Fig. 2 (left) and vertically rotated by 60°, allowing a view along the AccE5 dimeric twofold axis with the two AccD4 subunits of the hexameric CT complex in front. Bound arachidyl-CoA and propionyl-CoA substrates (only when visible) are shown in yellow spheres; d AccA3 BCCP loop/AccD4 interactions; e ACoA and propionyl-CoA substrates within corresponding densities. All other conventions are as in Fig. 2. Acronym shortcuts used: AccA3, A3; AccD4, D4; AccD5, D5, AccE5, E5, arachidyl-CoA, ACoA; propionyl-CoA, PCoA.

The N-terminal AccE5 α/β−hairpin motif contributes to an unusual 9-stranded β−barrel by inserting a single β−strand into a fourfold repeated array of two-stranded β−hairpins from four AccA3 subunits, thus generating a central AccE5-mediated AccA3 tetramerization motif (Fig. 3b, c). Due to the fourfold repeated β−sheet hairpin sequence of the AccA3 protomers, the same residue (R489) from three protomers forms equivalent salt bridges with three glutamate residues next to each other within the AccE5 α/β−hairpin sequence (E36, E40, E41), thus contributing to the structural integrity of the (AccA3)₄/AccE5 β−barrel tetramerization motif. Although similar β-barrel motifs are found in single BC subunits of canonical single-substrate ACCase complexes as well^1,6, in the structure of this hybrid SC/LC ACCase complex the barrel differs fundamentally due to the contribution of all four AccA3 subunits and the insertion of a single AccE5 β−strand, thus generating an asymmetric arrangement.

Remarkably, the helix of the N-terminal AccE5 α/β−hairpin motif inserts into the central hole of the (AccA3)₄/AccE5 barrel, pointing its C-terminus followed by the flexible linker towards the vis-à-vis AccD4/(AccD5)₂ layer of the central hybrid CT complex (Fig. 2b middle panel, Fig. 3b, c). Due to its uneven composition, the (AccA3)₄/AccE5 β−barrel tilts by 11 degrees from the threefold axis of the central (AccD4)₂/(AccD5)₄ hybrid CT assembly. This asymmetry further propagates to the catalytic BC domains of the four AccA3 protomers that form two separate homodimers (Figs. 2b, c, 3c). In each dimer, one proximal protomer catalytic domain (AccA3p) directly interacts with central β−barrel residues of both AccA3 protomers (Fig. 2b, c right panel, Fig. 3c). The other protomer (AccA3d) is in a distal position, without any interactions with the central β−barrel. The two AccA3p/AccA3d dimers are related by uneven rotations around an axis established by the central (AccA3)₄/AccE5 β−barrel. The larger rotation of 191 degrees is around the barrel segment with a single AccE5 β−strand inserted (Fig. 2b, right panel).

Finally, each of the AccA3 protomers comprises a C-terminal BCCP domain that transfers carboxybiotin from the AccA3 BC active site to one of the CT AccD4 or AccD5 active sites (Fig. 2b left panel, Fig. 3c). However, the resulting 8:6 stoichiometry of 2 ×4 AccA3 and 6 AccD4/AccD5 subunits in the holo complex generates a mismatch between eight AccA3-connected BCCP domains and six CT active sites. Of the six BCCP domains visible in this structure, four are bound to the four AccD5 CT active sites (Fig. 3c). Since the linkers to the remaining sequences of their parent AccA3 subunits are invisible, the precise connections are unknown. The other two visible BCCP domains, located at opposite faces of the central hexameric CT complex, are loosely connected to the surfaces of those donating AccA3p catalytic BC domains that are most tilted towards the two vis-à-vis AccD4 subunits of the central hexameric CT complex (Fig. 2b, c left panel, Fig. 3c). While there are no direct contacts between the BCCP domains and the AccD4 subunits, a highly positively charged segment of the connecting linker (residues 511-519, sequence RKKPKPRKR) directly interacts with several residues on each AccD4 surface (Fig. 3c, d).

Cryo-EM structures of the (AccA3)₈/(AccD4)₂/(AccD5)₄/(AccE5)₂ holo complexes in the presence of the SC substrate propionyl-CoA and LC substrate arachidyl-CoA reveal that these substrates are specifically bound to the CT active site pockets of the AccD5 and AccD4 subunits, respectively (Figs. 2c, 3c, e). When comparing the three structures, we noticed a major conformational change in the holo complexes with bound acyl-CoA ligands, resulting in an even more tilted arrangement between the central hexameric CT assembly and the tetrameric BC assembly (Figs. 2b, c, 3c). This change is caused by a rotation of approximately 90 degrees of the central (AccA3)₄/AccE5 tetramer β−barrel, which propagates into a corresponding rotation of the associated AccA3 catalytic domain dimers (Fig. 2b, c right panels, Fig. 3b). The rotation effectively switches the positions of neighboring AccA3 subunits and AccD4/AccD5 subunits facing each other (Figs. 2b, c, 3c). Therefore, the two visible BCCP domains vis-à-vis the two AccD4 subunits switch their origin from one of the AccA3p subunits to AccA3d subunits. In the presence of acyl-CoA ligands, the BCCP domain on its own also becomes associated with the surface of the AccD4 subunit (Figs. 2c, 3c). This observation suggests a role for all eight AccA3 subunits in BCCP-mediated carboxybiotin delivery, involving two AccA3 subunits for each of the two AccD4 CT subunit active sites.

In summary, our data present the structure of a hybrid SC/LC ACCase complex that incorporates two different CT subunits. In addition, this complex is highly asymmetric and allows large-scale rotations of the tetrameric BC assembly upon acyl-CoA substrate binding, suggesting that these are part of the multifunctional catalytic mechanism. These properties that are confined to ACCase complexes from mycobacteria create opportunities for their specific targeting to interfere with their involvement in mycobacterial pathogenesis^4,5. Finally, our data demonstrate the potential of extracting endogenous multi-protein complexes from their natural source as a superior alternative to the classical reconstitution of expressed protein components.

Methods

Bacterial strain and growth conditions

M. smegmatis mc²155 strain and the M. smegmatis (ΔaccD1-ΔaccA1, ΔaccD2-ΔaccA2) strains⁷ were initially grown in a pre-culture using 7H9 medium supplemented with 10% ADS (BSA 50 g/L, glucose 20 g/L, and NaCl 8.1 g/L) for 64 hours at 37 °C with shaking at 140 rpm. This pre-culture served as an inoculum for a 2 L culture. The main culture was grown in 7H9 broth supplemented with 0.2% glycerol, 0.1% glucose, and 0.05% Tween 80. Cultures were incubated at 37 °C with continuous shaking at 140 rpm for 48 hours until reaching an optical density at 600 nm of 2.5. No antibiotics were added to the growth medium.

Cell harvesting and lysis

Bacterial cells were harvested by centrifugation at 7000 × g for 60 minutes at 20 °C. The cell pellet was resuspended in lysis buffer containing 50 mM Tris-HCl (pH 7.5) and 300 mM NaCl. To prevent proteolysis, a protease inhibitor cocktail (Serva) was added to the resuspended cells. DNase I (1 mg/ml) was also included to reduce sample viscosity. Cells were lysed using a Sonoplus HD3200 sonicator at 45% amplitude for 6 minutes, using a 10-seconds on/5-seconds off pulse cycle, while keeping the sample on ice. The cell lysate was clarified by centrifugation at 19,000 × g for 1 hour at 4 °C, followed by filtration through a 0.45 µm filter.

Protein purification

The filtered lysate was incubated with pre-equilibrated streptavidin sepharose resin (Merck) for 2 hours at 4°C with gentle agitation. The resin was then extensively washed with lysis buffer to remove non-specifically bound proteins. The biotinylated protein complexes were eluted from the resin using 50 mM biotin in a buffer containing 50 mM Tris-HCl (pH 7.2) and 150 mM NaCl (Fig. 1a, c).

Elution fractions from strep-tactin affinity purification were analyzed by SDS-PAGE to assess protein purity. The pooled eluate was further purified by size exclusion chromatography using a Superose 6 10/300 column (GE Healthcare) equilibrated with 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 2 mM 1,4-dithiothreitol (DTT). Fractions containing the purified AccA3/AccD4/AccD5/AccE5 holo complex were pooled and concentrated using a centrifugal filter unit (Fig. 1b, d). The fractions indicated by black bars from M. smegmatis mc²155 strain were used for apo cryo-EM structure determination. The fractions from the M. smegmatis (ΔaccD1-ΔaccA1, ΔaccD2-ΔaccA2) strain were used for activity assays and cryo-EM studies of substrate-bound complexes.

Mass spectrometry

Protein bands were excised from the gel and cut into 1 mm³ pieces for in-gel digestion as described¹². Gel pieces were dehydrated with 100% acetonitrile, reduced with 10 mM DTT for 30 min at 56 °C, dehydrated again, and alkylated with 55 mM 2-chloroacetamide for 20 min at RT in the dark. Following a final dehydration, proteins were digested overnight at 37 °C using 2 ng/µL trypsin in 50 mM ammonium bicarbonate. Peptides were extracted by sonication for 15 min, followed by centrifugation. A second extraction was performed by sonication for 15 min, using 50% acetonitrile and 1% formic acid at twice the gel volume. The pooled supernatants were dried by vacuum centrifugation and reconstituted in 4% acetonitrile with 1% formic acid for LC-MS/MS analysis. LC-MS/MS peptide analysis was performed on an UltiMate 3000 RSLCnano LC system (Thermo Fisher Scientific) equipped with a µ-Precolumn C18 PepMap™ 100, 300 µmi.d. × 5 mm, 5 µm, 100 Å trapping cartridge (Thermo Fisher Scientific) and an analytical column (nanoEase™ M/Z HSS T3, 75 µm i.d. × 250 mm, 1.8 µm, 100 Å) (Waters). Samples were trapped at 30 µL/min in 0.05% trifluoroacetic acid for 6 min. Peptides were eluted at 0.3 µL/min using a gradient of solvent B (3% DMSO, 0.1% formic acid in acetonitrile) against solvent A (3% DMSO, 0.1% formic acid in water). Peptides were introduced into an Orbitrap Fusion™ Lumos™ Tribrid™ mass spectrometer (Thermo Fisher Scientific) via a Pico-Tip emitter (360 µm OD × 20 µm ID, 10 µm tip; CoAnn Technologies) with a 2.2 kV spray voltage. The instrument operated in positive ion mode with a capillary temperature of 275 °C. Full MS scans were acquired in the Orbitrap profile mode from m/z 350–1500 at 120,000 resolution (at m/z 200), with a 100 ms maximum injection time and standard AGC target. For MS/MS, the instrument operated in DDA mode, acquiring scans in the ion trap in rapid scan mode with a 35 ms maximum injection time and standard AGC target. Fragmentation was induced by HCD (30% normalized collision energy), and MS2 spectra were acquired in centroid mode. Database Search and Analysis. Raw files were converted to mzML format with MSConvert (ProteoWizard), applying peak picking for the 1000 most intense peaks, 64-bit encoding, and zlib compression. Files were searched usingMSFragger in FragPipe (22.1-build02) against FASTA databases (MycolicibacteriumSmegmatis_ATCC700084_UP000000757_ID246196_entries6602_26102022_dl11012023 and P4225 with common contaminants and reversed sequences). Search parameters included: carbamidomethylation (C, 57.0215) as a fixed modification; oxidation (M, 15.9949) and N-terminal acetylation (42.0106) as variable modifications. Mass error tolerances were 20 ppm (MS1) and 0.5 Da (MS2). Trypsin was specified as the protease, allowing a maximum of 2 missed cleavages and a minimum peptide length of 7. The FDR was set to 0.01 (1%) at both peptide and protein levels. The ‘Default’ FragPipe workflow was used with the following modifications: ionquant.maxlfq: 0, ionquant.mbr: 0, ionquant.minions: 2, ionquant.normalization: 0, ionquant.uniqueness: 1, msfragger.fragment_mass_tolerance: 0.5, msfragger.fragment_mass_units: 0, msfragger.misc.fragger.enzyme-dropdown-1: trypsin, msfragger.search_enzyme_name_1:trypsin,msfragger.search_enzyme_nocut_1: P, phi-report.filter: --sequential --prot 0.01 --razor, quantitation.run-label-free-quant: true.

Enzymatic activity assays

The assay reaction mix contained 50 mM HEPES (pH 7.5), 5 mM Mg-ATP, 5 mM NaHCO₃, and 100 µM of acyl-CoA substrates (acetyl-CoA, propionyl-CoA, heptadecanoyl-CoA, or stearoyl-CoA, separate reaction for each substrate). The reaction was initiated by adding 10 µM of purified AccA3/AccD4/AccD5/AccE5 holo complex. After 2 hours of incubation, the reaction aliquots were quenched by multiple rounds of freezing and thawing. The control reactions were done either without ATP or without the AccA3/AccD4/AccD5/AccE5 holo complex. All experiments were carried out at 25 °C.

The formation of the carboxylated products was detected using liquid chromatography-mass spectrometry (LC-MS). For LC, an Infinity 1290 II LC-MS system (Agilent) was utilized. Substrates and products were separated using an Agilent Poroshell EC-C18 3 mm × 150 mm, 2.7 μm column. Solvent A was 40 mM ammonium formate, pH 6.8 and solvent B was acetonitrile. The gradient (% solvent B) used was: 0–3 min, 2%; 3–30 min, 2-98%; 30–35 min, 98-2%; followed by a 3 min re-equilibration period at 2% B at a flow rate of 0.5 ml/min. The column temperature was kept constant at 20 °C.

The HPLC system was coupled with an Accurate Mass 6230 TOF apparatus (Agilent) for metabolite identification. To achieve dynamic mass axis calibration, a reference mass solution was continuously infused using an isocratic pump with a 100:1 splitter. ESI capillary was set at 3000 V and fragmentor was set at 110 V. The nebulizer pressure was set to 40 psi, while the nitrogen drying gas was delivered at a flow rate of 10 L/min and maintained at a temperature of 200 °C. The sheath gas was set to a temperature of 350 °C with a flow rate of 11 L/min to support efficient desolvation. Mass spectrometric data were acquired at a rate of 1 spectrum per second across an m/z range of 50–1200. The instrument consistently provided accurate mass measurements with a mass error below 5 ppm, a resolution ranging from 10,000 to 25,000 over the m/z range of 121–955, and a dynamic range spanning five orders of magnitude. Data were acquired in centroid mode using the 4 GHz (extended dynamic range) setting. The data were analyzed with the MassHunter Qualitative Analysis software (Agilent). A mass tolerance of <0.005 Da was applied for metabolite identities.

Cryo-EM data acquisition, processing, and 3D reconstruction

For structure determination of the AccA3/AccD4/AccD5/AccE5 holo complex, 3 µL of sample at a concentration of 4 mg/ml was applied to glow-discharged Quantifoil R2/1 300-mesh copper grids. The grids were blotted for 2 seconds at 4°C and 100% humidity before being plunge-frozen in a liquid ethane-propane mix using a Vitrobot Mark IV (Thermo Fisher Scientific). For samples in the presence of acyl-CoA substrates, cryo-EM grids were prepared by mixing AccA3/AccD4/AccD5/AccE5 holo complex at a concentration of 5 mg/ml with substrates to a final concentration of 1 mM propionyl-CoA (Merck) or 1 mM arachidyl-CoA (Avanti Lipids), 4 mM ATP, 20 mM NaHCO₃, and 5 mM MgCl₂. Acyl-CoA substrates were added from 10 mM stocks, and other substrates (ATP, NaHCO₃, and MgCl₂) were added from 100 mM stocks. R2/1 Quantifoil grids were glow-discharged for 120 s. The substrates were mixed immediately before being applied to the grids; the total time from mixing to plunge-freezing in the liquid ethane-propane mix was 45 seconds. The grids were blotted for 2.5 s with a blot force of −7 at 4 °C and 100% humidity using a Vitrobot Mark IV (Thermo Fisher Scientific).

Cryo-EM data for all samples were collected on a Titan Krios G3 microscope (Thermo Fisher Scientific) operating at 300 kV, equipped with a K3 direct electron detector (Gatan) and a Quantum energy filter (20 eV slit width) operating in counting mode. Data were recorded at a nominal magnification of 130,000x, corresponding to a calibrated pixel size of 0.68 Å/pixel. In total, 18,636 movies were recorded for AccA3/AccD4/AccD5/AccE5 holo complex in the absence of acyl-CoA ligands, 10,897 in the presence of propionyl-CoA, and 6,272 in the presence of arachidyl-CoA. Each movie consisted of 40 frames, with a total exposure of 44 e⁻/Ų and a defocus range of -0.5 to -2.25 µm. The EPU program (Thermo Fisher Scientific) was used for automated data acquisition.

All datasets were processed in cryoSPARC v4.2¹³. Movie frames were motion-corrected and dose-weighted using MotionCor2, and the contrast transfer function was estimated using CTFFIND4, both implemented within cryoSPARC. Particles were picked using a combination of blob and template pickers, extracted with a 560-pixel box size and binned 2x, and subjected to multiple rounds of 2D classification.

Particles from the best 2D classes were selected to generate an initial 3D model. After several rounds of heterogeneous refinement, a subset of particles was selected for a final refinement and re-extracted with a 560-pixel box size. The refinement yielded 3D reconstructions at a global resolution of 2.4 Å in the absence of any acyl-CoA ligands, the presence of propionyl-CoA, and the presence of arachidyl-CoA, as determined by the gold-standard Fourier shell correlation 0.143 criterion (Supplementary Figs. 1–4). The resolution achieved mainly arises from the central (AccD4)₂/(AccD5)₄/(AccE5)₂ core.

In the global map of the structure of the AccA3/AccD4/AccD5/AccE5 holo complex without acyl-CoA substrate, the electron density for both peripheral (AccA3)₄AccE5 assemblies was significantly weaker than for the central (AccD4)₂/(AccD5)₄/(AccE5)₂ core (Supplementary Fig. 1). In the structure of the AccA3/AccD4/AccD5/AccE5 holo complex in the presence of propionyl-CoA, one 3D class showed well-defined density for the complete complex (Supplementary Fig. 2). For the structure of the AccA3/AccD4/AccD5/AccE5 holo complex in the presence of arachidyl-CoA structure, one (AccA3)₄AccE5 assembly was well-resolved, while the other one was only poorly resolved (Supplementary Fig. 3). To improve the density in this region, focused local refinement was performed for all structures using a focused mask on the AccA3 region. This resulted in locally refined maps of one of the two (AccA3)₄AccE5 assemblies at 3.2 Å (no acyl-CoA ligands), 3.0 Å (propionyl-CoA), and 3.5 Å (arachidyl-CoA). For the second (AccA3)₄AccE5 assembly, a resolution of 3.3 Å was achieved for the structure in the presence of propionyl-CoA (Supplementary Figs. 1–4), whereas local refinement did not yield well-defined densities for the corresponding assemblies in the absence of acyl-CoA ligands and in the presence of arachidyl-CoA structure. Further details are listed in Table 1.

Model building, refinement, validation, and analysis

An initial model of an AccD5 hexamer was predicted by AlphaFold 3¹⁴ was fitted into the cryo-EM density map of the AccA3/AccD4/AccD5/AccE5 holo complex in the absence of acyl-CoA ligands, using ChimeraX¹⁵. The sequences for AccD4 and the C-terminus of AccE5 were subsequently assigned based on high-resolution features in the map, allowing model building of the complete (AccD4)₂/(AccD5)₄/(AccE5)₂ core assembly. During this procedure, two AccD5 subunits were replaced by AccD4 subunits, thus generating a hybrid CT core complex with 4:2 AccD5/AccD4 stoichiometry. Iterative refinement was performed using Servalcat¹⁶ within the CCPEM Doppio suite. Rotamers and outliers were corrected in Isolde¹⁷ within ChimeraX and manually adjusted in Coot¹⁸.

For the AccA3/AccE5 assembly, an AlphaFold 3 model was used as the initial template. However, since the EM map considerably deviated from the predicted symmetric AccA3 tetrameric arrangement, extensive manual remodeling and refinement were required, using Coot¹⁸ and Servalcat¹⁶, respectively.

For the propionyl-CoA and arachidyl-CoA bound structures, the refined structure of the AccA3/AccD4/AccD5/AccE5 holo complex in the absence of acyl-CoA ligands was used as a template. While the central (AccD4)₂/(AccD5)₄/(AccE5)₂ core assembly fitted well into the corresponding maps, the (AccA3)4/AccE5 assembly did not, indicating a different conformation triggered by the presence of these acyl-CoA substrates. These regions were manually rebuilt in Coot, corrected for rotamer and Ramachandran outliers in Coot and Isolde, and further refined in Servalcat.

While the AccA3/AccD4/AccD5/AccE5 holo complex structures in the absence of acyl-CoA ligands and in the presence of arachidyl-CoA showed high-resolution density for only one of the two (AccA3)₄/AccE₅ tetramers, the propionyl-CoA structure revealed both flanking (AccA3)₄/AccE₅ assemblies at sufficient resolution for molecular interpretation. Therefore, the complete (AccA3)₈/(AccD4)₂/(AccD5)₄/(AccE5)₂ holo complex in the presence of propionyl-CoA structure and truncated (AccA3)₄/(AccD4)₂/(AccD5)₄/(AccE5)₂ complexes in the absence of acyl-CoA ligands and in the presence of arachidyl-CoA were refined and used for further analysis.

Sequence confidence was validated using the CheckMySequence program¹⁹ in the CCPEM Doppio suite. The final models were validated using the MolProbity server, also implemented in CCPEM Doppio.

Geometry analysis was carried out in PyMOL (Version 3.0, Schrödinger, LLC), using PSICO and Draw_Rotation_Axis plugins.

Statistics and Reproducibility

The data shown in Fig. 1e are overlays of three technical replicates. Since the ion counts were on an arbitrary scale, the maximum ion count in each measurement has been set to 100%.

Author contributions

E.M. and M.W. designed the project. E.M. and H.M.T. performed the experimental work. All authors analyzed the data. E.M. and M.W. wrote the paper. L.P.S.d.C. and M.W. supported the project.

Peer review

Peer review information

Communications Biology thanks Jeremy R Lohman, Hugo Gramajo, and the other anonymous reviewer for their contribution to the peer review of this work. Primary Handling Editor: Tobias Goris. A peer review file is available.

Funding

Open Access funding enabled and organized by Projekt DEAL.

Data availability

The cryo-EM density maps have been deposited in the Electron Microscopy Data Bank (EMDB) under accession codes EMD-55717, EMD-55718, EMD-55802, and EMD-56159. The corresponding coordinates have been deposited in the Protein Data Bank (PDB) under accession codes 9T96, 9T97, 9TDM, and 9TR9, respectively. All other source data can be obtained from a Supplementary Data file named “Supplementary Data.xlsx”.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10439-x.