Structural insights into human propionyl-CoA carboxylase (PCC) and 3-methylcrotonyl-CoA carboxylase (MCC)

Reviewer #1 (Public review):

Summary:

The manuscript by Zhou et al offers new high resolution Cryo-EM structures of two human biotin-dependent enzymes: propionyl-CoA carboxylase (PCC) and methycrotonyl-CoA carboxylase (MCC). While X-ray crystal structures and Cryo-EM structures have previously been reported for bacterial and trypanosomal versions of MCC and for bacterial versions of PCC, this marks one of the first high resolution Cryo-EM structures of the human version of these enzymes. Using the biotin cofactor as an affinity tag, this team purified a group of four different human biotin-dependent carboxylases from cultured human Expi 293F (kidney) cells (PCC, MCC, acetyl-CoA carboxylase (ACC), and pyruvate carboxylase). Following further enrichment by size-exclusion chromatography, they were able to vitrify the sample and pick enough particles of MCC and PCC to separately refine the structures of both enzymes to relatively high average resolutions (the Cryo-EM structure of ACC also appears to have been determined from these same micrographs, though this is the subject of a separate publication). To determine the impact of substrate binding on the structure of these enzymes and to gain insights into substrate selectivity, they also separately incubated with propionyl-CoA and acetyl-CoA and vitrified the samples under active turnover conditions, yielding a set of cryo-EM structures for both MCC and PCC in the presence and absence of substrates and substrate analogues.

Strengths:

The manuscript has several strengths. It is clearly written, the figures are clear and the sample preparation methods appear to be well described. This study demonstrates that Cryo-EM is an ideal structural method to investigate the structure of these heterogeneous samples of large biotin-dependent enzymes. As a consequence, many new Cryo-EM structures of biotin-dependent enzymes are emerging, thanks to the natural inclusion of a built-in biotin affinity tag. While the authors report no major differences between the human and bacterial forms of these enzymes, it remains an important finding that they demonstrate how/if the structure of the human enzymes are or are not distinct from the bacterial enzymes. The MCC structures also provide evidence for a transition for BCCP-biotin from an exo-binding site to an endo-binding site in response to acetyl-CoA binding. This contributes to a growing number of biotin-dependent carboxylase structures that reveal BCCP-biotin binding at locations both inside (endo-) and outside (exo-) of the active site.

Weaknesses:

There are some minor weaknesses. Notably, there are not a lot of new insights coming from this paper. The structural comparisons between MCC and PCC have already been described in the literature and there were not a lot of significant changes (outside of the exo- to endo- transition) in the presence vs. absence of substrate analogues. There are sections of this manuscript that do not sufficiently clarify what represents a new insight from the current set of structures (there are few of them), vs. what is largely recapitulating what has been seen in previous structures.

There is not a great deal of depth of analysis in the discussion. For example, no new insights were gained with respect to the factors contributing to substrate selectivity (the factors contributing to selectivity for propionyl-CoA vs. acetyl-CoA in PCC). The authors acknowledge that they are limited in their interpretations as a consequence of the acyl groups being unresolved in all of the structures. They offer a simple, overarching and not particularly insightful explanation that the longer acyl group in propionyl-CoA may mediate stronger hydrophobic interactions that stabilize the alpha carbon of the acyl group at the proper position. The authors did not take the opportunity to describe the specific interactions that may be responsible for the stronger hydrophobic interaction nor do they offer any plausible explanation for how these might account for an astounding difference in the selectivity for propionyl-CoA vs. acetyl-CoA. Essentially, the authors concede that these cryo-EM structures offer no new insights into the structural basis for substrate selectivity in PCC, confirming that these structures do not yet fully capture the proper conformational states.

Some of these minor deficiencies aside, the overall aim of contributing new cryo-EM structures of the human MCC and PCC has been achieved. While I am not a cryo-EM expert, I see no flaws in the methodology or approach. While the contributions from these structures are somewhat incremental, it is nevertheless important to have these representative examples of the human enzymes and it is noteworthy to see a new example of the exo-binding site in a biotin-dependent enzyme.

Reviewer #2 (Public review):

Summary:

This paper reports the structures of two human biotin-dependent carboxylases. The authors used endogenously purified proteins and solved the structures in high resolutions. Based on the structures, they defined the binding site for acyl-CoA and biotin and reported the potential conformational changes in biotin position.

Strengths:

The authors effectively utilized the biotin of the two proteins and obtained homogeneous proteins from human cells. They determined the high-resolution structures of the two enzymes in apo and substrate-bound states.

Comments and questions to the manuscripts:

(1) I'm quite impressed with the protein purification and structure determination, but I think some functional characterization of the purified proteins should be included in the manuscript. The activity of enzymes should be the foundation of all structures and other speculations based on structures.

(2) In Figure 1B, the structure of MCC is shown as two layers of beta units and two layers of alpha units, while there is only one layer of alpha units resolved in the density maps. I suggest the authors show the structures resolved based on the density maps and show the complete structure with the docked layer in the supplementary figure.

(3) In the introduction, I suggest the author provide more information about the previous studies about the structure and reaction mechanisms of BDCs, what is the knowledge gap, and what problem you will resolve with a higher resolution structure. For example, you mentioned in line 52 that G437 and A438 are catalytic residues, are these residues reported as catalytic residues or this is based on your structures? Has the catalytic mechanism been reported before? Has the role of biotin in catalytic reactions revealed in previous studies?

(4) In the discussion, the authors indicate that the movement of biotin could be related to the recognition of acyl-CoA in BDCs, however, they didn't observe a change in the propionyl-CoA bound MCC structure, which is contradictory to their speculation. What could be the explanation for the exception in the MCC structure?

(5) In the discussion, the authors indicate that the selectivity of PCC to different acyl-CoA is determined by the recognition of the acyl chain. However, there are no figures or descriptions about the recognition of the acyl chain by PCC and MCC. It will be more informative if they can show more details about substrate recognition in Figures 3 and 4.

(6) How are the solved structures compared with the latest Alphafold3 prediction?

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1 (Public Review):

Weaknesses:

There are some minor weaknesses.

Comment 1:Notably, there are not a lot of new insights coming from this paper. The structural comparisons between MCC and PCC have already been described in the literature and there were not a lot of significant changes (outside of the exo- to endo- transition) in the presence vs. absence of substrate analogues.

We agree that the structures of the human MCC and PCC holoenzymes are similar to their bacterial homologs. That is due to the conserved sequences and functions of MCC and PCC across different species.

Comment 2: There is not a great deal of depth of analysis in the discussion. For example, no new insights were gained with respect to the factors contributing to substrate selectivity (the factors contributing to selectivity for propionyl-CoA vs. acetyl-CoA in PCC). The authors state that the longer acyl group in propionyl-CoA may mediate stronger hydrophobic interactions that stabilize the alpha carbon of the acyl group at the proper position. This is not a particularly deep analysis and doesn't really require a cryo-EM structure to invoke. The authors did not take the opportunity to describe the specific interactions that may be responsible for the stronger hydrophobic interaction nor do they offer any plausible explanation for how these might account for an astounding difference in the selectivity for propionyl-CoA vs. acetyl-CoA. This suggests, perhaps, that these structures do not yet fully capture the proper conformational states.

We appreciate this comment. Unfortunately, in the cryo-EM maps of the PCC holoenzymes, the acyl groups were not resolved (fig. S6), so we were unable to analyze the specific interactions between the acyl-CoAs and PCC. We have revised the manuscript and acknowledged this limitation in the second paragraph of the discussion section:

“In the cryo-EM maps of the PCC holoenzymes, the acyl groups of acetyl-CoA and propionylCoA were not resolved (fig. S6), limiting the analysis of the interactions between the acyl groups and PCC. Nevertheless, the PCC-PCO and PCC-ACO structures determined in our study demonstrate that the conformations of the acyl-CoA binding pockets in the two structures are almost identical (Fig. 3F, fig. S7, B and C). In addition, the well resolved CoA groups of propionyl-CoA and acetyl-CoA bind at the same position in human PCC holoenzyme (Fig. 3F). These findings indicate that propionyl-CoA and acetyl-CoA bind to PCC with a similar binding mode.”

Comment 3: The authors also need to be careful with their over-interpretation of structure to invoke mechanisms of conformational change. A snapshot of the starting state (apo) and final state (ligand-bound) is insufficient to conclude *how* the enzyme transitioned between conformational states. I am constantly frustrated by structural reports in the biotin-dependent enzymes that invoke "induced conformational changes" with absolutely no experimental evidence to support such statements. Conformational changes that accompany ligand binding may occur through an induced conformational change or through conformational selection and structural snapshots of the starting point and the end point cannot offer any valid insight into which of these mechanisms is at play.

Point accepted. We have revised our manuscript to use conformational differences instead of conformational changes to describe the differences between the apo and ligand-bound states (see the last paragraph of the introduction section and the third paragraph of the discussion section).

Reviewer #2 (Public Review):

Comments and questions to the manuscripts:

Comment 1: I'm quite impressed with the protein purification and structure determination, but I think some functional characterization of the purified proteins should be included in the manuscript. The activity of enzymes should be the foundation of all structures and other speculations based on structures.

We appreciate this comment. However, since we purified the endogenous BDCs and the sample we obtained was a mixture of four BDCs, the enzymatic activity of this mixture cannot accurately reflect the catalytic activity of PCC or MCC holoenzyme. We have revised the manuscript and acknowledged this limitation in the first paragraph of the results section:

“We did not characterize the enzyme activities of the mixed BDCs because the current methods used to evaluate the carboxylase activities of BDCs, such as measuring the ATP hydrolysis or incorporation of radio-labeled CO2, are unable to differentiate the specific carboxylase activity of each BDC.”

Comment 2: In Figure 1B, the structure of MCC is shown as two layers of beta units and two layers of alpha units, while there is only one layer of alpha units resolved in the density maps. I suggest the authors show the structures resolved based on the density maps and show the complete structure with the docked layer in the supplementary figure.

We appreciate this comment. We have shown the cryo-EM maps of the PCC and MCC holoenzymes in fig. S8 to indicate the unresolved regions in these structures. The BC domains in one layer of MCCα in the MCC-apo structure were not resolved. However, we think it would be better to show a complete structure in Fig. 1 to provide an overall view of the MCC holoenzyme. We have revised Fig. 1B and the figure legend to clearly point out which domains were not resolved in the cryo-EM map and were built in the structure through docking. We have also revised the main text to clearly describe which parts of the holoenzymes were not resolved in the cryo-EM maps and how the complete structures were built.

Comment 3: In the introduction, I suggest the author provide more information about the previous studies about the structure and reaction mechanisms of BDCs, what is the knowledge gap, and what problem you will resolve with a higher resolution structure. For example, you mentioned in line 52 that G437 and A438 are catalytic residues, are these residues reported as catalytic residues or this is based on your structures? Has the catalytic mechanism been reported before? Has the role of biotin in catalytic reactions revealed in previous studies?

Point accepted. It was reported that G419 and A420 in Streptomyces coelicolor PCC, corresponding to G437 and A438 in human PCCβ, were the catalytic residues for the secondstep carboxylation reaction (PMID: 15518551). The same study also reported the catalytic mechanism of the carboxyl transfer reaction. The role of biotin in the BDC-catalyzed carboxylation reactions has been extensively studied (PMIDs: 22869039, 28683917). We have revised the manuscript to introduce the catalytic mechanisms of BDCs elucidated through the investigation of prokaryotic BDCs in the fourth paragraph of the introduction section.

Comment 4: In the discussion, the authors indicate that the movement of biotin could be related to the recognition of acyl-CoA in BDCs, however, they didn't observe a change in the propionyl-CoA bound MCC structure, which is contradictory to their speculation. What could be the explanation for the exception in the MCC structure?

We appreciate this comment. We do not have a good explanation for why we did not observe a change in the propionyl-CoA bound MCC structure. It is noteworthy that neither acetyl-CoA nor propionyl-CoA is the natural substrate of MCC. Recently, a cryo-EM structure of the human MCC holoenzyme in complex with its natural substrate, 3-methylcrotonyl-CoA, has been resolved (PDB code: 8J4Z). In this structure, the binding site of biotin and the conformation of the CT domain closely resemble that in our acetyl-CoA-bound MCC structure. Therefore, the movement of biotin induced by acetyl-CoA binding mimics that induced by the binding of MCC's natural substrate, 3-methylcrotonyl-CoA, indicating that in comparison with propionylCoA, acetyl-CoA is closer to 3-methylcrotonyl-CoA regarding its ability to bind to MCC. We have discussed this possibility in the last paragraph of the discussion section. We have also added a supplementary figure (fig. S11) to compare the structures of human MCC holoenzyme in complex with acetyl-CoA and 3-methylcrotonyl-CoA.

Comment 5: In the discussion, the authors indicate that the selectivity of PCC to different acyl-CoA is determined by the recognition of the acyl chain. However, there are no figures or descriptions about the recognition of the acyl chain by PCC and MCC. It will be more informative if they can show more details about substrate recognition in Figures 3 and 4.

We appreciate this comment. Unfortunately, in the cryo-EM maps of the PCC holoenzymes, the acyl groups were not resolved (fig. S6), so we were unable to analyze the specific interactions between the acyl-CoAs and PCC. We have revised the manuscript and acknowledged this limitation in the second paragraph of the discussion section:

“In the cryo-EM maps of the PCC holoenzymes, the acyl groups of acetyl-CoA and propionylCoA were not resolved (fig. S6), limiting the analysis of the interactions between the acyl groups and PCC. Nevertheless, the PCC-PCO and PCC-ACO structures determined in our study demonstrate that the conformations of the acyl-CoA binding pockets in the two structures are almost identical (Fig. 3F, fig. S7, B and C). In addition, the well resolved CoA groups of propionyl-CoA and acetyl-CoA bind at the same position in human PCC holoenzyme (Fig. 3F). These findings indicate that propionyl-CoA and acetyl-CoA bind to PCC with a similar binding mode.”

Comment 6: How are the solved structures compared with the latest Alphafold3 prediction?

Since AlphaFold3 was not released when our manuscript was submitted, we did not compare the solved structures with the AlphaFold3 predictions. We have now carried out the predictions using Alphafold3. Due to the token limitation of the AlphaFold3 server, we can only include two α and six β subunits of human PCC or MCC in the prediction. The overall assembly patterns of the Alphafold3-predicted structures are similar to that of the cryo-EM structures. The RMSDs between PCCα, PCCβ, MCCα, and MCCβ in the apo cryo-EM structures and those in the AlphaFold3-predicted structures are 7.490 Å, 0.857 Å, 7.869 Å, and 1.845 Å, respectively. The PCCα and MCCα subunits adopt an open conformation in the cryo-EM structures but adopt a closed conformation in the AlphaFold-3 predicted structures, resulting in large RMSDs.