The cryo-EM structure of trypanosome 3-methylcrotonyl-CoA carboxylase provides mechanistic and dynamic insights into its enzymatic function

Summary

3-methylcrotonyl-CoA carboxylase (MCC) catalyzes the two-step, biotin-dependent production of 3-methylglutaconyl-CoA, an essential intermediate in leucine catabolism. Given its critical metabolic role, deficiencies in this enzyme associate with organic aciduria, while its overexpression is linked to tumor development. MCC is a dodecameric enzyme composed of six copies of each α- and β-subunit. We present the cryo-EM structure of the endogenous MCC holoenzyme from Trypanosoma brucei in a non-filamentous state at 2.4 Å resolution. Biotin is covalently bound to the BCCP domain of α-subunits and positioned in a non-canonical binding pocket near the active site of a neighboring β-subunit dimer. Moreover, flexibility of key residues at α-subunit interfaces and loops enables pivoting of α-subunit trimers to partly reduce the distance between α- and β-subunit active sites, required for MCC catalysis. Our results provide a structural framework to understand the enzymatic mechanism of eukaryotic MCCs and to assist drug discovery against trypanosome infections.

Graphical Abstract

eTOC blurb

3-methylcrotonyl-CoA carboxylase from Trypanosoma brucei is a key metabolic enzyme and a potential drug target against this human and animal pathogen. By applying advanced electron microscopy, Plaza-Pegueroles et al. uncover unexpected details regarding the mechanism and dynamic behavior of this enzyme, thus opening avenues for future drug discovery.

Introduction

Biotin-dependent carboxylases are broadly distributed in nature, playing a critical role in the metabolism of different substrates such as fatty acids, carbohydrates or amino acids.¹ Four of these enzymes are known in eukaryotes, i.e. acetyl-CoA carboxylase (ACC), pyruvate carboxylase (PC), propionyl-CoA carboxylase (PCC), and 3-methylcrotonyl-CoA carboxylase (MCC). As a mitochondrial enzyme, MCC catalyzes the carboxylation of 3-methylcrotonyl-CoA, a key intermediate in leucine degradation, to produce 3-methylglutaconyl-CoA. Subsequent steps in the breakdown of this branched amino acid ultimately generate acetyl-CoA and succinyl-CoA, used for energy production and other metabolic processes.^2,3 Consistently, MCC is highly expressed in the liver and kidneys.^4,5

In humans, deficiencies in MCC can lead to a rare genetic disorder known as 3-methylcrotonyl-CoA carboxylase deficiency or 3-methylcrotonylglycinuria,^3,6–9 which is the most frequently-detected organic aciduria in newborns worldwide.¹⁰ Clinical manifestations of this autosomal recessive defect include hypoglycemia, ketosis, hypotonia, seizures, and developmental and psychomotor delays.^11–14 On the other hand, increased MCC expression promotes proliferation of several cancer types, being crucial for the oncogenesis of hepatocellular carcinoma.¹⁵ Moreover, MCC overexpression in astrocytoma, glioblastoma, meningioma and oligodendroglioma tumors indicates the capacity of brain tumor cells to use leucine as metabolic substrate, pointing to leucine catabolism as a potential target in cancer treatment.^2,15 Augmented MCC expression was also observed in Leishmania species resistant to antimony, which suggests MCC as a putative drug target against leishmaniosis.¹⁶ Additionally, MCC overexpression enhances adaptation of Arabidopsis thaliana to alkalized soil, a major problem impacting agricultural production, by increasing tolerance to sodium carbonate and bicarbonate.¹⁷

The MCC enzymatic activity involves two sequential chemical steps,¹⁸ taking place at different and distant active sites^19,20 (Fig. 1A). The first step is a Mg²⁺/ATP-dependent carboxylation of biotin (vitamin B7 or vitamin H) using bicarbonate as donor.^18,21 The active center for this reaction resides in MCC α-subunits. The second step is a transfer of the carboxyl group from carboxylated biotin to 3-methylcrotonyl-CoA, which occurs in the active site formed by two MCC β-subunits.

Figure 1. Cryo-EM structure of TbMCC.

(A) Schematic representation of dodecameric MCC, showing the hexameric β-core and both trimeric α-caps. Active sites for both steps in the enzymatic reaction are shown with blue and red circles for the first and second steps, respectively. Arrows indicate the necessary movement of BCCP domain for biotin shuttling between active sites. Two arrows, from subunits α5 to β3/β4 and from subunits α4 to β5/β6, are not depicted as they occur at the back of the drawing. (B) Overall structure of T. brucei MCC showing symmetry operators (triangle for 3-fold axis and oval for 2-fold axis), with one of the six αβ-dimers colored in magenta and blue for α- and β-subunits, respectively. (C) Schematic and ribbon representations of α- and β-subunits of TbMCC, showing domains and motifs in both chains. (D) Degree of conservation of residues in the surface of an α/β dimer.

Current knowledge on the architecture of this enzymatic complex stems from the crystal structure of MCC from the bacterium Pseudomonas aeruginosa (PaMCC).²² The functional complex consists of six copies of each of α- and β-subunits, giving rise to an elongated, dodecameric barrel (Fig. 1A). β-subunits assemble into a hexameric core, which can be considered as a trimer of tail-to-tail β-dimers, positioned at the central part of the barrel. Two α-subunit trimers cap the β-subunit core on opposite sides. α-subunits comprise three domains connected by two short linkers.²³ The N-terminal biotin carboxylase (BC) domain occupies the edge of the barrel and contains the α-subunit active site (Fig. 1A, purple circles). The C-terminal biotin carboxyl carrier protein (BCCP) domain contains the Met-Lys-Met biotinylation motif,^24,25 whose exposed central lysine covalently bonds biotin to form biocytin.^26,27 The middle BT domain connects BC and BCCP domains via extended linkers and establishes interactions with the adjacent β-subunit. β-subunits are composed of an N-terminal N-dock domain, followed by two carboxyltransferase (N-CT and C-CT) domains that are connected through the CT linker. Location of the β-subunit active site at the β-dimer interface results in the formation of two active sites per dimer (Fig. 1A, red circles). In this active site, the 3-methylcrotonyl-CoA substrate is sandwiched between the C-CT from one β-subunit and two α-helices in the N-CT from the opposite subunit. The active sites of the two enzymatic reactions locate 80 Å apart (Fig. 1A, black arrows), which requires shuttling of the BCCP domain, facilitated by flexibility of the BT-BCCP linker.²² This swinging-domain model has also been proposed for other carboxylases.^28,29

An equivalent architecture has been found in the electron cryomicroscopy (cryo-EM) structure of MCC from the kinetoplastid protozoan Leishmania tarentolae (LtMCC), forming straight filaments composed by four to six dodecameric MCC barrels connected though α-subunit trimers.²³ In this structure, the BT-BCCP linker is partially fixed through an interaction with the N-dock domain of a neighboring β-subunit, which may hamper the BCCP domain to reach the α-subunit active site. Based on these hindrances, concomitant swinging of the BCCP and BC domains of the same α-subunit was suggested for LtMCC, in the so-called ‘dual-swinging-domains’ model. However, LtMCC oligomerization immobilizes BC domains, which are flexible only at the filament edges. Besides, in LtMCC filaments biotin is non-covalently liganded to the reactive lysine in the BCCP domain. These two features suggest that the LtMCC structure represents an inactive state of the enzyme.²³

Here, we applied cryo-EM to study the structure of the endogenous MCC holoenzyme from the parasite of humans and animals Trypanosoma brucei (TbMCC). The map exhibits an overall resolution of 2.4 Å while regions of the barrel core are resolved to 2.1 Å resolution, which reveals critical structural details. The resulting structure represents a resting state of the enzyme, where the BCCP domain is weakly-bound at an unexpected position next to a the β-subunit active site. Our cryo-EM analysis also provides insights into the dynamics of TbMCC and enables generation of a mechanistic model for MCC activity.

Results

Overall structure

We enriched biotinylated proteins from a T. brucei lysate using streptavidin agarose and found a barrel-like structure clearly distinguished in negative-staining electron microscopy grids (Fig. S1). The 3D reconstruction derived from this experiment combined with mass spectrometry fingerprinting of the bands observed in SDS-PAGE identified TbMCC as the main sample component (Fig. S1). Using single-particle cryo-EM, we obtained a map at 2.4 Å resolution that allowed full model building of this dodecameric structure (Fig. S2; Table 1). TbMCC presents a barrel-like structure with a height of 201 Å and a diameter of 144 Å (Fig. 1B, left), closely resembling those of PaMCC²² and LtMCC.²³ The barrel core is formed by a trimer of β-subunits dimers, while two trimers of α-subunits occupy the top and bottom of the barrel.

Table 1. Cryo-EM data collection and refinement statistics.

	TbMCC
Sample Support	C-flat 1.2/1.3, 300 mesh, continuous carbon film
Microscope	Titan Krios
Detector	K3
Voltage (KeV)	300
Number of Frames	40
Dose (e−/Å2)	38.3
Exposure time (s)	3
Pixel size (Å/pixel)	0.8238
Number of grids	1
Days of data collection	1
Collected micrographs	4530
Particles after 2D class	187681
Final number of particles	126391
Resolution (Å)	2.4
Sharpening B-factor (Å2)	−73.0
Ramachandran plot
Outliers (%)	0.08
Allowed (%)	3.38
Favoured (%)	96.53
Ramachandran plot Z-score
Whole	0.69
Helix	1.86
Sheet	0.65
Loop	−0.73
Map CC (around atoms)	0.89 (cc_mask)
RMSD bond lengths (Å)	0.003
RMSD bond angles (°)	0.477
All-atom clashscore	5.33
Rotamer outliers (%)	1.09
C-beta deviations	0.00
FSC(model-map) = 0.5	2.50 (masked), 2.67 (unmasked)

The structure exhibits 3-fold symmetry along the cylinder axis and 2-fold symmetry between barrel halves, such that one copy of the αβ complex constitutes the repeating unit (Fig. 1B, right). The β-subunit hexamer defines a large cavity (41481 ų) that is capped at the two α/β interfaces (Fig. S3). Two types of pores are formed by positively charged residues, one at the center of each β-subunit dimer and the second between two β-subunit dimers. While their internal diameter is small (about 4 Å), minor conformational changes may allow access of small ions into the large β-subunit cavity. In addition, each α-subunit trimer defines a cavity (31500 ų) that is open both at the edge of the barrel and through three additional apertures at α/β interfaces (Fig. S3).

α-subunits feature a semi-open BC domain and a structured BT-BCCP linker

The N-terminal BC domains of α-subunits (residues 11–470) define the edges of the dodecameric barrel. This domain can be divided into three structural motifs (Fig. 1C-D), as described for a homologous BC enzyme from Escherichia coli.³⁰ The A-motif (residues 11–139) and C-motif (residues 216–470) pack against each other at the center of the α-subunit trimer, with the A-motif contacting the BT domain of a nearby α-subunit. The middle B-motif (residues 140–215) or ‘BC lid’ extends away from the barrel in a semi-open conformation, as deduced from comparison with ATP-bound and Apo structures of the E. coli BC enzyme (Fig. 2A-B). Interestingly, the B-motif in TbMCC exhibits flexibility, as suggested by limited map resolution in this region (Fig. S2). Flexibility of the BC lid plays a role in closing the α-subunit active site for catalysis, such that BCCP binding immobilizes the lid.³¹ The C-motif harbors the active site for carboxylation of BCCP-bound biotin. Consistently, residues in this region are highly conserved (Fig. 1D; Fig. S4). While substrates for the first catalytic step are absent in the TbMCC structure, an extra density is observed in a pocket defined by positively-charged residues K247, R303 and R354 (Fig. 2C). The position of this density corresponds to the bicarbonate binding site, as seen in the BC site of the PC enzyme,³² suggesting the presence of an equivalent ion in our structure (Fig. 2D). PC is a tetrameric enzyme formed by two layers of identical monomers, with the BCCP translocating between active sites in the same layer. Comparison with the cryo-EM structure of PC suggests that R303 and E307 in TbMCC likely promote bicarbonate deprotonation while R354 stabilizes deprotonated biotin, which is the final CO2 acceptor. Interestingly, R354 in TbMCC contacts E250 through a salt bridge that in PC is formed after ATP hydrolysis to prevent the reverse reaction.³²

Figure 2. Structural details of α-subunits.

(A, B) Structural comparison of the BC domain with biotin carboxylase from E. coli complexed with ATP (panel A, PDB - 1DV2) or in the apo form (panel B, PDB – 1DV1), both in grey. The A-, B- and C-motifs in the BC domain of TbMCC are shown in red, yellow and pink, respectively. (C) Cryo-EM map contoured at 6.0 RMSD showing the bicarbonate binding site where a piece of density, marked with a black star, indicates the presence of an ion (not modeled). (D) Structural comparison of the BC active site of TbMCC, in pink, with the BC active site in pyruvate carboxylate from Lactococcus lactis, in light grey (PDB – 7ZZ3), showing the high degree of conservation in sequence and geometry of this region. (E) Cryo-EM map contoured at 6.0 RMSD around the BT-BCCP linker and surrounding regions. (F) Structural comparison of BCCP domain of TbMCC, in red, with a BCCP domain from E. coli, in grey (PDB - 1BDO).

The central BT domains of α-subunits (residues 477–589), comprising an axial α-helix enclosed by an eight-stranded β-sheet, locate between BC domains and the β-subunits hexamer (Fig. 1C). Residues connecting the α-helix and the β-sheet in the BT domain form the hook (residues 491–505). While it presents poor conservation among α-subunits, the hook constitutes the major interacting region with the adjacent β-subunit (Fig. 1D; Fig. S4). The BT-BCCP linker (residues 590–601), which is clearly seen in the electron density map (Fig. 2E), interacts with a β-subunit from a neighboring dimer. Finally, the C-terminal BCCP domain of α-subunits (residues 602–677) docks next to the active site of this β-subunit dimer (Fig. 1B). This domain adopts a capped β-sandwich fold,^27,33 with β-strands covering two sides of the sandwich and two symmetric hammerhead motifs (residues 620–634 and 657–671) (Fig. 2F). Lower than average resolution of the cryo-EM map in this region (Fig. S2) suggests some degree of flexibility for the BCCP domain.

β-subunits form active sites that are accessible to the substrates

The N-dock domain of β-subunits (residues 60–135) contains two helices, α2 (residues 90–115) and α3 (residues 119–130), that extend at the outer rim of the barrel core and establish extensive contacts with the hook motif of an adjacent α-subunit (Fig. 1C, 3A). The N-terminal extension of this domain, which is absent in PaMCC (Fig. S5), protrudes into a neighboring β-subunit dimer and includes a short helix α1 (residues 66–70) that interacts with the BT-BCCP linker of an adjacent α-subunit (Fig. 3A-B). Residue Y69 in helix α1 stacks between residues L592 and F596 from the BT-BCCP linker, partially limiting its mobility.

Figure 3. Structural details of β-subunits.

(A) Ribbon representation of the interaction region between α- and β-subunits. Note the hook in the BT domain interacting with all domains in the β-subunit and the short helix α1 in the N-dock domain (α1’ corresponds to an adjacent β-subunit, in grey). (B) Detailed view of the interaction between the BT-BCCP linker and helix α1’ in the N-dock domain from a nearby β-subunit, shown in grey. (C) Electrostatic surface around the biotin binding site in the β-subunit. A positively-charged region is the binding site for the 3-methylcrotonyl-CoA substrate (CoA in the structure of PaMCC is shown as reference, PDB - 3U9S). (D) Ribbon representation of the β-subunit active site, formed at the interface between two β-subunits in grey and wheat, showing capping helices α17 and α18 that contribute to the substrate-binding pocket. A dotted circle indicates the binding site for 3-methylcrotonyl-CoA.

The N- and C-terminal CT domains of β-subunits (residues 135–348 and 388–601) are connected through the CT linker (residues 349–387), which locates in the inner region of the MCC barrel next to the closest α-subunit (Fig. 1C, 3A). As observed in other MCCs, the active site for 3-methylcrotonyl-CoA carboxylation occupies the interface between N-CT and C-CT domains belonging to different β-subunits forming a dimer (Fig. 1B-C). Accordingly, these regions exhibit the highest conservation degree in β-subunits (Fig. 1D; Fig. S5). The cryo-EM map lacks density indicative of 3-methylcrotonyl-CoA presence in its binding pocket, located about 80 Å away from the α-subunit active site. This pocket features a positively-charged narrow tunnel leading to the position where 3-methylcrotonyl-CoA carboxylation takes place (Fig. 3C). Two α-helices (α17 and α18) in the C-CT domain also contribute to pocket formation, while they project away from the MCC barrel (Fig. 3D).

Biocytin is protected in a pocket adjacent to the β-subunit active site

The complete biocytin moiety is clearly defined in the cryo-EM map, which shows continuous density corresponding to the covalent bond linking biotin and the central lysine (K641) in the biotinylation motif of the BCCP domain (Fig. 4A). The biotin moiety protrudes from the BCCP domain towards the β-subunit dimer that is located laterally from the β-subunit interacting with the BT domain of the same α-subunit (Fig. 1B). Biotin, which comprises a thiophene ring next to an ureido ring, interacts with both β-subunits forming this lateral dimer (Fig. 4B). The thiophene ring occupies a hydrophobic pocket formed by residue M640 from the BCCP holding the biotin, residues L299 and A303 from the N-CT domain in the distant β-subunit of the dimer, and residues I432 and V466 from the C-CT domain in the proximal β-subunit of the dimer. Additionally, the ureido ring interacts with the main chain of residues T462 and F464 and the side chain of residue Q534 in the latter domain. Residues in this non-canonical binding pocket, including those interacting with biotin through the main chain, are conserved (Fig. S5). Interestingly, the biotin oxygen atom points towards a narrow tunnel connecting the non-canonical biotin-binding pocket with the binding site for 3-methylcrotonyl-CoA (Fig. 4C). In this unanticipated location that is ~10 Å away from the β-subunit active site, no density is observed for the biotin carboxyl moiety. These results suggests that, upon 3-methylcrotonyl-CoA carboxylation and product release, the BCCP domain may be retained next to the β-subunit active site, with biotin protected in an alternative pocket next to that required for carboxylation.

Figure 4. Structural details of the alternative biotin-binding pocket.

(A) Cryo-EM map contoured at 6.0 RMSD around biocytin. (B) Biocytin interactions inside the binding pocket. Note that residues from both β-subunits, in grey and wheat, are involved in the interaction. H-bonds are shown as purple dotted lines. (C) Tunnel connecting the alternative biotin-binding pocket and the substrate binding site. For clarity, CoA from the PaMCC structure (PDB - 3U9S) is shown as sticks and the distance from CoA in PaMCC to biotin in TbMCC is indicated.

BC domain flexibility contributes to approach of active sites

The largest inter-chain interaction is established between β-subunits forming dimers (3570 Ų area), which are placed tail-to-tail in the central region of the MCC barrel (Fig. 1B). A total of 28 hydrogen-bonds (H-bonds) and 12 salt bridges are observed in this interface (Table S1). The interface between β-subunits forming each trimeric ring is three times smaller (1360 Ų area), where 17 H-bonds and 12 salt bridges are observed (Table S1). These observations indicate that the β-subunit dimer is the primary building block for the assembly of the MCC barrel core. In contrast, the buried surface connecting adjacent α-subunits is smaller (500 Ų) (Fig. 1B), suggesting that α-trimers are more flexible than the β-core. This may also indicate that α-subunits attach to the MCC barrel at a later assembly stage, as shown for the PCC enzyme³⁴ and further supported by observation of pre-formed β-hexamers with only one α-trimer in negatively-stained grids (Fig. S1).

The main interacting region between α- and β-subunits is established between the BT domain of the former and residues from all domains in the latter (1120 Ų). A total of 12 H-bonds and 1 salt bridge are established (Table S1), most of them between the BT hook of α-subunits and the N-terminal region (residues 137–155) of the N-CT domain in β-subunits (Fig. 3A). Besides this principal contact with the closest β-subunit, each α-subunit interacts with the β-subunit located aside from the closest one (Fig. 1B). First, the N-terminal region (residues 64–78) in the N-dock domain of the aside β-subunit interacts with the BT domain and the BT-BCCP linker. Second, the BCCP domain interacts with the N-CT domain of the aside β-subunit (920 Ų). These lateral contacts are important for BCCP activities during MCC catalysis, as they partially limit BCCP swinging while allowing alternative binding configurations next to the β-subunit active site (see below).

To study the dynamics the MCC enzymatic complex, we analyzed the conformational distribution of the particles in our cryo-EM dataset. We observed that the β-subunit core is quite rigid and only minor rearrangements are apparent (Movie 1). In contrast, α-subunits are highly dynamic, especially the BC domains at the edges of the MCC barrel. On each edge of the barrel, BC domain trimers pivot over their corresponding BT domains to approach the β-subunit core. The BC-BT linker acts as a major hinge in this movement by rotating ~12°, so that the α-subunit active site approaches the β-subunit active site by ~7 Å (Fig. 5A). Unexpectedly, the three C-motifs in the BC domains forming each α-subunit trimer move together, such that the trimeric interaction is preserved. Consequently, when one BC domain approaches the β-subunit core, the other two BC domains in the trimer travel together, thus moving away from the core. This suggests that only one of the three α-subunit active sites can approach the β-subunit core at a time, then travels back to allow the approach of α-subunit another active site towards the core. Nevertheless, a second hinge region at the interface between the A- and C-motifs of each BC domain enables movement of the A- and B-motifs. This secondary movement, encompassing a ~20° rotation and reaching up to ~15 Å at the edge of the B-motif (Fig. 5A), appears independent among the different α-subunits of the MCC barrel. Consistent with these results, symmetry expansion followed by 3D variability analysis identified similar BC domain movements, though with a lower amplitude (Movie 2). Moreover, 3D variability analysis also captured motions in the BCCP domain, which pivots by ~10° over its biocytin moiety that remains bound to β-subunits while terminal residues of the BT-BCCP liker exhibit flexibility (Movie 2).

Figure 5. Flexibility analysis of TbMCC.

Comparison of the extreme conformations, in red an orange, captured through 3DFlex analysis. Two major movements of the BC domain are observed, the first involving a 12° swinging over the BC-BT linker and the second comprising a 20° rotation over the interface between the A- and C-motifs, with maximum displacement at the edge of the B-motif (BC lid).

The TbMMC structure represents a resting state

Superposition of the TbMCC structure with those of PaMCC and LtMCC shows an overall similarity, with root-mean-square deviations (RMSD) of 0.7 and 1.2 Å, respectively. Likely due to the filamentous nature of the large assemblies enriched during purification on density gradients, LtMCC presents a more extended barrel with a height of 212 Å, while PaMCC and TbMCC barrels are 201 Å high. The LtMCC extended barrel results from looser packing of α-subunits against β-subunits (Fig. 6A), further supporting the plasticity of α-subunits. BC and BT domains in α-subunits of LtMCC shift away from β-subunits by 4.5 and 3 Å, respectively, compared to their position in TbMCC. As a result, the contact surface between α and β-subunits is smaller in LtMCC than in TbMCC, i.e. 777 versus 1120 Ų. In contrast, PaMCC presents a larger interaction surface between α- and β-subunits (1492 Ų), mainly due to an extended hook motif in the α-subunit (Fig. S4). The extended PaMCC hook partially overlaps with β-subunit residues at the interface between the N-CT and C-CT domains, including the CT linker, in the structures of MCCs from eukaryotic parasites. Notably, β-subunit residue Q592 in TbMCC forms an H-bond with R183 that reinforces the interaction between the N-CT and C-CT domains (Fig. 6B). A stronger interaction is observed in LtMCC, where E691 connects R237 and R277 (Fig. 6C). In the PaMCC structure, residues in this position contact the α-subunit hook, with W537 placed between E130 and L543 (Fig. 6D). This indicates a tighter packing of β-subunits at the expense of a weaker interaction between α- and β-subunits in parasite MCCs.

Figure 6. Structural comparison of TbMCC with reported MCCs.

(A) Structural comparison with the cryo-EM structure of LtMCC filaments (PDB - 8F3D), in yellow, showing the displacement of α-trimers due to filament formation. (B) Structural comparison with the crystal structure of PaMCC (PDB - 3U9S), in yellow, where an open BC domain is involved in crystal contacts. (C) Similarities between LtMCC and TbMCC at the interface between α- and β-subunits. Purple and green dotted lines indicate H-bonds and salt bridges, respectively. (D) The presence of an extended hook in PaMCC alters the interface between α- and β-subunits. (E, G) Biocytin in TbMCC binds to a pocket formed by residues that are conserved in LtMCC but adopt a different configuration (E). Note that, in the latter, biotin is not attached covalently to the lysine residue in the biotinylation motif (G). (F, H) The biocytin-binding region in TbMCC is conserved in PaMCC (F), where the BCCP domain and biotin are more deeply inserted into the active site (marked with a red circle) for catalysis (H). Arrows indicate changes between the structures, while black labels correspond to TbMCC and yellow labels are used for LtMCC or PaMCC.

The α-subunit active site in TbMCC is narrower than in the structures of previously-reported MCCs (Fig. 6A-B). In contrast to E. coli BC structures,³⁰ open configurations of the BC domain in reported MCCs are likely due to contacts with neighboring dodecameric barrels in LtMCC filaments or PaMCC crystals, both involving large interaction areas in the absence of substrates (Fig. 2C). Regarding biotin binding in the vicinity of the β-subunit active site, biocytin in TbMCC occupies a non-canonical binding pocket formed by residues that are conserved in MCCs (Fig. 6E-F, S4, S5). Non-covalently bound biotin in LtMCC occupies an equivalent but rotated location (Fig. 6G), while biocytin in PaMCC complexed to CoA is placed ~5 Å closer to the substrate (Fig. 6H). This correlates with closure of this active site in PaMCC through movement of capping helices α17 and α18. In addition, the BT-BCCP linker in parasite MCC structures shows extensive interactions with helix α1 in the N-dock domain of a neighboring β-subunit (Fig. 3B), while this linker is disordered in PaMCC. Consistently, helix α1 is absent in PaMCC (Fig. S5) and the BT-BCCP linker in parasite MCCs presents a 4-residue insertion respect to other organisms (Fig. S4). This indicates that the BCCP domain is more loosely attached to the β-subunit in these organisms and likely explains why only one out of six BCCP domains is visible in the crystal structure of PaMCC. Altogether, these observations suggest that our TbMCC structure represents a resting state of the enzyme.

Discussion

The structure of MCC from T. brucei provides mechanistic details that deepen our understanding of this mitochondrial enzyme, essential for leucine metabolism. While the core of its cylindrical structure composed by six copies of the β-subunit superposes well with existing structures from P. aeruginosa and L. tarentolae, the outer portions of the cylinder exhibit interesting details. These regions, formed by α-subunit trimers, are now observed at high resolution in their native state, unaffected by crystal packing (as for PaMCC) or filament formation (as for LtMCC). Therefore, the intrinsic flexibility of the two α-trimers can be analyzed.

While the A- and C-motifs of the BC domain closely superpose with E. coli carboxylase, the B-motif presents an intermediate state between open and closed conformations, with a putative ion occupying the bicarbonate binding site in this highly-conserved region (Fig. 2). As seen in other carboxylases, binding of Mg²⁺/ATP induces this flexible B-motif to close the active site, which in PaMCC and LtMCC is wide-open likely due to contacts that are absent in our cryo-EM structure. Unexpectedly, dynamic analysis of cryo-EM data shows that not only the B-motif is highly flexible, but each α-trimer pivots so that BC domains can approach the β-subunit core (Movie 1). While this movement is limited in our analysis, this result confirms that α-trimers can contribute to reducing the distance that BCCP domains need to travel in order to swing between the α- and β-subunit active sites. However, the BT-BCCP linker in TbMCC is partly fixed by contacts with the N-dock domain of a neighboring β-subunit (Fig. 3), as also observed for LtMCC. Residues in the BT-BCCP linker contacting the N-dock domain are absent in PaMCC or mammalian MCCs (Fig. S4), suggesting that only parasite MCCs may partly restrict BCCP swinging. Nevertheless, mobility of BCCP in vivo might be greater than implied by the structure. Interestingly, PaMCC and mammalian MCCs present hook regions that are longer than those in parasite MCCs. As seen in the PaMCC structure, this involves a larger interaction area with β-subunits, which might restrict flexibility of α-trimers in species with longer hooks.

In TbMCC biotin is covalently bound to lysine in the BCCP domain, unlike being sequestered at a non-reactive distance from the equivalent lysine in the case of LtMCC (Fig. 6). However, as opposed to PaMCC complexed to the substrate analog CoA, biocytin in TbMCC locates more superficially to the β-subunit active site, occupying a secondary pocket that connects to the active site through a narrow channel (Fig. 4). We interpret this as a putative resting position for the BCCP domain that may protect it from degradation and/or hamper undesired biotin interactions while MCC is dormant. The structure of this non-canonical biotin binding pocket may assist discovery of trypanocides. Moreover, as the residues in the pocket are conserved (Fig. S5), it may also assist the search for inhibitors of human MCC that could refrain tumor progression.

Based on available structures and our work, a model for MCCs can be proposed (Fig. 7). As shown for LtMCC, the filament state corresponds to an inactive configuration due to intra-filament interactions that immobilize BC domains. It follows that quiescent MCC in the mitochondrial matrix can be easily reactivated. This filament structure was not observed in TbMCC likely due to differences in the LtMCC purification protocol, which included an initial selection of large particles using density gradients. However, based on structural similarity between parasite MCCs and conservation of residues involved in LtMCC filament formation, the existence of TbMCC filaments cannot be discarded. Free dodecameric complexes, as in the TbMCC structure, represent a resting state in the reaction where biotin, already bound to lysine in the biotinylation motif, is shielded next to the β-subunit active site but ready to be translocated to the α-subunit active site for carboxylation. The coordinated movement of α-trimers observed in TbMCC suggest that both BC and BCCP domains could move for biotin carboxylation, as stated for the ‘dual-swinging domains’ model. According to our results, BCCP swinging can cover most of the distance between active sites, while α-trimer pivoting is more limited. Upon biotin carboxylation at the α-subunit active site, the BCCP domain can travel back to the interface between β-subunits for the second step in the enzymatic reaction. While this second step takes place at one β-dimer, α-trimer pivoting may allow concomitant carboxylation of a second biotin moiety at a neighboring α-subunit active site. The PaMCC structure, with CoA mimicking the carboxyl acceptor and carboxyl-free biotin in close proximity to the active site, represents the post-reaction state for the second step. Upon this reaction, MCC may either continue its enzymatic task, fall back in the resting state, or be stored in the inactive filamentous state. Additional structural analysis in the presence of substrates or analogs will shed further light on the mechanisms of MCC activity.

Figure 7. Mechanistic model for MCC catalysis.

Schematic representation of the proposed model for MCC activity, showing the different states represented by the three known MCC structures and a putative state for the first catalytic reaction, currently uncharacterized structurally. One β-subunit (labeled β1) is shown in blue with its capping helices as a smaller box, while three other β-subunits (labeled β2, β5, β6 for consistency with Fig. 1) are in brown. One α-subunit (labeled α1) is colored salmon, yellow, purple and red for BC domain, BT domain, BT-BCCP linker and BCCP domain, respectively, while three other α-subunits (labeled α2, α3, α6 for consistency with Fig. 1) are in white. Flexible regions in the TbMCC structure are shown as dotted arrows, while thick arrows indicate domain movements for the catalytic mechanism. In filaments, α-subunits are fixed by intra-filament interactions and BCCP domains locate next to β-subunits. In the solution resting state, BCCP domains remain in the same location with uncarboxylated biotin occupying a cavity near the β-subunit active sites, while BC domains can swing especially at the BC lid. A larger movement is needed, according to the dual-swinging-domains model, where BC and BCCP domains approach for biotin carboxylation at α-subunit active sites. For the second reaction, BCCP domains swing back for 3-methylcrotonyl carboxylation at β-subunit active sites upon closure of the capping helices over the substrate.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, C. Fernández-Tornero (cftornero@cib.csic.es).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• The cryo-EM map of TbMCC has been deposited in the Electron Microscopy Data Bank under accession number EMD-19492. The coordinates of TbMCC have been deposited in the Protein Data Bank under accession number 8RTH. All aforementioned deposited data are publicly available as of the date of publication and accession numbers are also listed in the key resources table.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
HEPES	Sigma-Aldrich	Cat# H4034
Ammonium sulfate	Sigma-Aldrich	Cat# A4418
Magnesium chloride	Sigma-Aldrich	Cat# M8266
Glycerol	Sigma-Aldrich	Cat# G5516
Triton X-100	Sigma-Aldrich	Cat# 93443
2-Mercaptoetanol	Sigma-Aldrich	Cat# M3148
Biotin	IBA Lifesciences	Cat# 2-1016-002
Trypsin	Thermo Fisher	Cat# 90058
cOmplete EDTA-free Protease Inhibitor Cocktail	Roche	Cat# 11873580001 (Sigma)
Critical commercial assays
Strep-tactin XT 4Flow FPLC column, 1 ml	IBA Lifesciences	Cat# 2-5023-001
Amicon Ultra Centrifugal Filter, 10 kDa MWCO	Sigma-Aldrich	Cat# UFC5010
C-flat 1.2/1.3, 300 mesh	Protochips	Cat# CF-1.2/1.3-3CU-50
Deposited data
X-ray crystal structure of PaMCC	Huang et al.22	PDB: 3U9S
CryoEM structure of LtMCC	Hu et al.23	PDB: 8F3D
X-ray crystal structure of E coli BC	Thoden et al.30	PDB: 1DV1
X-ray crystal structure of E coli BC (ATP-bound)	Thoden et al.30	PDB: 1DV2
CryoEM structure of PC	Lopez-Alonso et al.32	PDB: 7ZZ3
X-ray crystal structure of E coli BCCP	Athappilly & Hendrickson27	PDB: 1BDO
CryoEM structure of TbMCC	This paper	PDB: 8RTH
CryoEM map of TbMCC	This paper	EMDB: EMD-19492
Experimental models: Cell lines
Trypanosoma brucei strain Lister 427	ATCC	Cat# PRA-380
Software and algorithms
Relion	Zivanov et al.35	https://relion.readthedocs.io/en/release-3.1/
MotionCor2	Zheng et al.36	https://emcore.ucsf.edu/ucsf-software
CryoSPARC	Punjani et al.37	https://cryosparc.com
UCSF Chimera	Pettersen et al.38	https://www.rbvi.ucsf.edu/chimera/
ChimeraX	Meng et al.39	https://www.cgl.ucsf.edu/chimerax/
Coot	Emsley et al.40	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Clustal Omega	Sievers & Higgins41	http://www.ebi.ac.uk/Tools/msa/clustalo
Phenix	Liebschner et al.42	https://phenix-online.org/documentation/index.html
AlphaFold	Jumper et al.43	https://github.com/sokrypton/ColabFold
ESPript3	Gouet et al.44	https://espript.ibcp.fr/ESPript/ESPript/
CASTp	Tian et al.45	http://sts.bioe.uic.edu/
PDBePISA	Krissinel and Henrick46	https://www.ebi.ac.uk/pdbe/pisa/
Pymol	Schrödinger	https://pymol.org/2/

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Trypanosoma brucei cells (Lister 427 strain) were grown as described in method details and harvested at cell density of ~2x10⁷ cells/ml.

METHOD DETAILS

Purification of 3-methylcrotonyl-CoA carboxylase from T. brucei

Procyclic (insect) Lister 427 strain of Trypanosoma brucei (ATCC) was cultured in roller bottles at 27 °C in SDM-79 media⁴⁷ supplemented with 10% fetal bovine serum. Parasites were harvested by centrifugation at 3000 g for 15 min, resuspended in PBS buffer supplemented with 6 mM sucrose (1/10 of culture volume), and collected by centrifugation under the same conditions. Cell pellet containing ~5x10¹⁰ parasites was flash frozen in liquid nitrogen and stored at −80 °C until use. Harvested cells were resuspended in lysis buffer containing 25 mM Hepes pH 7.5, 150 mM (NH₄)₂SO₄, 5 mM MgCl₂, 5% glycerol, 0.5% Triton X100, 2 mM β-mercaptoethanol and one tablet of EDTA-free protease inhibitors (Roche), and lysed by sonication. The lysate was clarified by centrifugation at 35000 rpm for 30 min in a 45Ti rotor (Beckman). The supernatant was filtered (0.45 μm) and loaded in a Strep-Tactin column (1 ml, IBA) equilibrated with buffer containing 25 mM Hepes pH 7.5, 150 mM (NH₄)₂SO₄, 5 mM MgCl₂, 5% glycerol and 2 mM β-mercaptoethanol. After washing the column with 10 ml of the same buffer, the sample was eluted with 3 ml of 25 mM Hepes pH 7.5, 150 mM (NH₄)₂SO₄, 5 mM MgCl₂, 5% glycerol, 2 mM β-mercaptoethanol and 50 mM biotine. The elution buffer was exchanged to the Strep-Tactin equilibration buffer using a 10-kDa cutoff centrifugal filter device (Amicon) to remove biotin from the sample. The concentrated sample was aliquoted, plunged into liquid nitrogen and stored at −80 °C.

Negative stain electron microscopy

Electron micrographs were recorded on a JEOL 1230 electron microscope operated at 100 kV. 25 images were collected using a CMOS TVIPS TemCam-F416 detector. Data processing was done using Relion 3.1.³⁵ A total of 37313 particles were picked and, after several rounds of 2D classification, 28348 particles were finally selected for 3D refinement. The final 3D volume reached an estimated resolution of 17 Å.

Identification of proteins by MALDI-TOF-TOF peptide mass fingerprinting

Individual SDS-PAGE protein bands were cut manually and digested with trypsin (Thermo). Samples were analyzed with an Autoflex III TOF/TOF mass spectrometer (Bruker-Daltonics).Typically, 1000 scans for peptide mass fingerprinting (PMF) and 2000 scans for MS/MS were collected. Automated analysis of mass data was performed using FlexAnalysis software version 3.4 (Bruker-Daltonics). MALDI–MS and MS/MS data were combined through the BioTools 3.2 program (Bruker-Daltonics) to SwissProt 2021_02 database using MASCOT software 2.6 (Matrix Science). Relevant search parameters were set as follows: enzyme, trypsin; fixed modifications, carbamidomethyl (C); oxidation (M); 1 missed cleavage allowed; peptide tolerance, 50 ppm; MS/MS tolerance, 0.5 Da. Protein scores greater than 75 were considered significant (p<0.05).

Cryo-EM sample preparation and data acquisition

C-flat copper 300 mesh 1.2/1.3 holey carbon grids (Protochips) covered with a home-made continuous carbon film were glow discharged for 20 s and a current of 25 mA using a GloQube device (Quorum). Then 3 μl of sample at 0.07 mg/ml concentration were applied to the glow-discharged grids and blotted for 2 s with blotting force −5 using a FEI Vitrobot (Thermo Fisher Scientific) at a temperature 10 °C and 100% humidity. Grids were vitrified by plunging into liquid ethane cooled with liquid nitrogen, and stored in liquid nitrogen for later imaging. Movies were acquired in a FEI Titan Krios (ThermoFisher) electron microscope working at an acceleration voltage of 300 keV. A total of 4530 movies (40 frames each) were collected with a K3 summit (Gatan) direct electron detector operated in ‘super-resolution’ mode, using the EPU software (ThermoFisher), with a defocus value range from −1 to −3 μm, a pixel size of 0.8238 Å/pixel, a dose rate of 18.361 e−/pixel/s and a total dose of 38.3 e−/Ų.

CryoEM image processing

Movies were aligned and motion corrected with MotionCor2³⁶ and subsequent processing was performed in CryoSPARC v4.4.³⁷ Patch CTF estimation was followed by visual inspection, and 157 movies were rejected using the ‘curate exposures’ job. From the remanent 4377 micrographs, initial picking was done with ‘blob picker’ using an ellipse template of 200 by 290 Å. These particles were 2D classified and the resulting 2D averages were used as templates to pick a total of 490k particles. After extraction with a 540 pixel box, a reference-free 2D classification was performed and the best classes, including 188k particles, were selected. Ab-initio reconstruction was used to generate an initial 3D model in the absence of imposed symmetry, while subsequent 3D refinements were performed enforcing D3 symmetry. To further clean the dataset, a heterogeneous refinement was performed using both the initial 3D model and a reference generated from discarded 2D classes as a decoy volume. This resulted in a final set of 126k particles that, after homogeneous refinement, yielded a map at 2.6 Å resolution. These particles underwent iterative rounds of local defocus refinement and CTF aberration refinement (tilt, trefoil, spherical aberration and tetrafoil). The final reconstruction was obtained by non-homogeneous refinement with minimize over per-particle scales toggled on, yielding a map with a global resolution of 2.4 Å. Running the same job but imposing either C1 (asymmetric) or C3 symmetry produced FSC curves showing overall resolutions of 3.0 and 2.5 Å, respectively. Local resolution was calculated and the ‘Orientation diagnostic’ job was preformed to asses directional FSC. Anisotropy sharpening was applied to the map refined with D3 symmetry, using Phenix⁴² and supplying the two half maps.

3DFlex analysis⁴⁸ as implemented in CryoSPARC v4.4 was used to assess the conformational distribution of the sample. The final set of particles and the map derived from the final refinement imposing D3 symmetry were cropped to a box size of 320 pixels and then downsampled to 160 pixels for training only. A 3DFlex mesh was prepared with 20 tetra cells and no segmentation, as the use of a segmented mesh showed only the movement of one monomer, likely due to inability to properly set the sub-mesh connections. Then, the 3DFlex model was trained with 64 hidden units and a rigidity of 2 in the absence of imposed symmetry. Latent distribution plots were inspected and volume series were obtained with ‘Flex Generator’ to visualize movements. While several numbers of latent dimensions were tried, the series of volumes showed an almost indistinguishable movement, so only the first dimension was considered. Finally, ‘Flex Reconstruct’ with 60 BFGS iterations produced a volume exhibiting increased resolvability in the BC region. For further assessment of conformational heterogeneity, symmetry was relaxed using symmetry expansion over the D3 symmetry, followed by cropping to a box size of 360 pixels and down-sampling to 144 pixels. Subsequent 3D variability analysis as implemented in CryoSPARC 4.4 was displayed in simple mode for better seizing continuous heterogeneity.

Model building and refinement

An AlphaFold2 predicted model⁴³ was used as initial reference for both chains. The model was docked into the sharpened map using UCSF Chimera.³⁸ Subsequently, the model was refined using Phenix⁴² against the portion of the map surrounding an asymmetric α/β-dimer. Real space refinement was done using secondary structure and Ramachandran restraints. Restraints were used for Cα atom positions belonging to the B-motif of BC domains, using as target model the initial geometry. A link was added between the NZ atom of K641 and the C11 atom of biotin in order to mimic biocytin. Finally the model was multiplied according to D3 symmetry and fitted in the complete sharpened map for the last rounds of refinement. During refinement manual modification and model building was done using Coot.⁴⁰ Contact surface properties were calculated using the PDBePISA.⁴⁶ Protein cavities were calculated using the CASTp web server.⁴⁵ Sequence alignments were performed with Clustal omega⁴¹ and represented with ESPript3.⁴⁴ Figures were prepared with Pymol (Schrödinger) and ChimeraX.³⁹

QUANTIFICATION AND STATISTICAL ANALYSIS

Cryo-EM data were processed with CryoSPARC and structural models were refined with Phenix. The corresponding statistics can be found in Table 1.

Supplementary Material

2

Movie 1. TbMCC flexibility resulting from 3DFlex analysis. Related to Figure 5. Two copies of the repeating α/β unit, with α-subunits in magenta and β-subunits two shades of blue, are highlighted while the remaining of the barrel is in grey. Note the movements of BC domains.

3

Movie 2. TbMCC flexibility resulting from 3D variability analysis. Related to Figure 5. Two copies of the repeating α/β unit, with α-subunits in magenta and β-subunits two shades of blue, are highlighted while the remaining of the barrel is in grey. Note the movements of BC and BCCP domains.

Highlights.

• The structure of Trypanosoma brucei MCC suggests a resting state of the enzyme

• Movements of the BC domain support a dual-swinging domains model

• Biotin occupies a non-canonical binding site next to the β-subunit active center

• BBCP domains are loosely bound to the β-core and pivot over their biocytin moiety