Human aminolevulinate synthase structure reveals a eukaryotic-specific autoinhibitory loop regulating substrate binding and product release

5′-aminolevulinate synthase (ALAS) catalyzes the first step in heme biosynthesis, generating 5′-aminolevulinate from glycine and succinyl-CoA. Inherited frameshift indel mutations of human erythroid-specific isozyme ALAS2, within a C-terminal (Ct) extension of its catalytic core that is only present in higher eukaryotes, lead to gain-of-function X-linked protoporphyria (XLP). Here, we report the human ALAS2 crystal structure, revealing that its Ct-extension folds onto the catalytic core, sits atop the active site, and precludes binding of substrate succinyl-CoA. The Ct-extension is therefore an autoinhibitory element that must re-orient during catalysis, as supported by molecular dynamics simulations. Our data explain how Ct deletions in XLP alleviate autoinhibition and increase enzyme activity. Crystallography-based fragment screening reveals a binding hotspot around the Ct-extension, where fragments interfere with the Ct conformational dynamics and inhibit ALAS2 activity. These fragments represent a starting point to develop ALAS2 inhibitors as substrate reduction therapy for porphyria disorders that accumulate toxic heme intermediates.

5. Relevant details on the methodology of the molecular simulations are missing. How big was the simulation box, how were long-range electrostatics included, was the simulation performed at constant temperature and/or pressure, were bond-lengths constrained, what timestep is used? The Gromacs code is not listed under 'software' in the 'reporting summary'. 6. The identified fragments are tested for their inhibitory capacity by a single point experiment. For an inhibitor of this kind, it may be expected that full inhibition is not attainable. Determination of an IC50 would be appropriate to confirm a concentration dependent inhibition.
7. More importantly, the inhibition of the WT is not the relevant measure here. Would the authors not much rather show inhibitory behavior of these compounds against delAT or Q548X ? 8. Abbreviations ClpX and AAA+ are not explained. 9. Homology is not quantifiable. Either two proteins are homologous (have a common ancestor) or they are not. In line 83, the authors describe a low-homology region. I guess they mean a region with low sequence similarity. This is an interesting paper that gives a bit more detail on the inhibitory action of the C-terminal region of aminolevulinate synthase. Alas2 has been recently shown to be a good target to control certain porphyrins through gene silencing approaches so there is potential therefore for small molecule drugs.
The authors do repeat a lot of what is known already, but the paper does provide a structural interpretation of the C-terminal autoinhibition. The MD is pretty, but I am not sure it is that especially informative in this case.
1. The key crystallographic point addressed here is the conformation of the C-terminal region because the story revolves around the structure of this autoinhibitory loop. I would therefore expect to see an omit map for this region to confirm its architecture and its correct placement in the electron density.
2. I suspect the problem the authors have is that most of the C-terminal region is unstructured, flexible, or disordered. At least some of the C-terminal region should be well-structured to warrant the crystallographic build-up in the narrative.
Provided there is good evidence for the C-terminal structure, then Nature Communications seems about the right level for this paper.
Reviewer #3 (Remarks to the Author): A combination of crystallography, limited biochemical analysis of disease-causing mutations, and crystallography-based fragment screening were employed in this study on human ALAS2. MD simulations are included to support mechanistic inferences. The significance of the work is that the structural studies led to mapping of the C-terminal extension, providing a framework for formulating hypotheses about the molecular basis of disease-causing mutations for XLSA and XLP. Overall, the experiments seem to have been well conducted. However, the mechanistic insights to emerge from the study are limited. The fragment search led to the identification of ligands that were largely clustered in a hydrophobic pocket but had minimal inhibitory effects (~15% at 1 mM concentration).
A key conclusion that the R511E mutation destabilizes the structure of the Ct-extension was not experimentally supported and other inferences of gain-of-function XLP mutation, tended to be overstated. The absence of quantitative data (kinetic data are shown without error bars, the number of replicates are not mentioned, values for kinetic parameters are not provided for wild-type versus mutant ALAS2) diminishes the study's rigor.

Major
1. Line 126. The Km values for glycine and succinyl-CoA are not reported as stated in Fig S4a,b. These need to be noted with SDs in the text.
2. Line 129. The data shown in Fig 4c does not establish interaction between ALAS2 and SUCLG1-SUCLA2, but rather a messy gel with many bands. If the authors consider this conclusion regarding protein-protein interaction to be important for this study, then MS data establishing the identity of the proteins present in the putative crosslinked bands needs to be shown.
3. Line 261/2: The statement that breaking the R511:E569 interaction renders the Ct-extension highly destabilized is not backed by experimental evidence. Furthermore, it is only marginally prone to proteolysis. Hence the conclusion that the R511 mutation is essential for maintaining the structural integrity of the Ct-extension is overstated.
4. His-tags were retained during characterization of mutant proteins, which is problematic. The authors need to remove the Hi-tags to make meaningful inferences about the resulting disease phenotype.
5. Line 272. The conclusion that the elongated C-terminal extension in the delAGTG variant is subjected to proteolysis is not experimentally supported. While a small amount of proteolysis is seen, its location is not mapped to the Ct-extension. 8. Line 283. Although it is stated that "we show here that delG has the same 2-to 3-fold wild-type efficiency (kcat/Km) with respect to succinyl-CoA as delAT and Q548X", no such data are provided.
9. Line 335. The statement, "a catalytically grounded state that prevents transition to the open conformation" is unclear. What is meant by catalytically grounded? And how can the authors rule out the much more like ensemble of conformational substates, with the equilibria being shifted by individual mutations?
10. Line 342. What is the experimental basis of the conclusion that kinetically, "the Ct-extension exerts more influence towards product release rather than succinyl-CoA binding?" Km can not be equated with substrate affinity. Minor: 1. Line 29. Change "That has evolved among higher eukaryotes" to "that is only present in higher eukaryotes". 3. Lines 90/91 (Supplemental section). Change "that could accommodate for the disordered" to "that could account for the disordered". 4. Line 184. Change "π−π" to "π−π interactions". 5. Line 191. FigS1 does not show the structure of succinyl-CoA bound ALAS2. Please correct.

The authors perform simulations of three distinct states of the protein (and one fourth, with a putative inhibitor bound)
. In order to model the substrate bound state, they start from a representative structure of the most common cluster of the simulation with PLP bound, but without substrate. This means that in that simulation, the Ct-extension already moves away from the position that is observed in the X-ray structure. That is a discrepancy that needs to be discussed. Seemingly, the dynamics in the Ct-extension already take place when the substrates are not bound? Response: As illustrated in the figure below, the orientation of the Ct-extension in the PLP-bound form (without substrate) is very similar to that in the crystal structure. The Ct-extension only moves away from its position in the crystal structure, after binding of the succinyl-CoA substrate (Fig. 3), as shown by complete breakage of the R511:E569 and R293/K299:E571 salt bridges in our simulation .
Orientation of the Ct-extension. The Ct-extension from monomer A in the energy minimized crystal structure (green) and in the representative structure from the simulation for PLP-bound ALAS2 (magenta). Response: It is correct that a few residues were missing in the crystal structure, namely 182-187 and 549-555 in chain A, and 183-187 and 549-557 in chain B. To model a complete hsALAS2 structure (137-578) with missing residues included, we first constructed a hsALAS2 homology model using the scALAS2 structure (PDB ID: 5TXT) as a template. This hsALAS2 homology model was then superimposed with the hsALAS2 crystal structure, such that coordinates for the missing residues were extracted from the homology model, and covalently linked to the crystal structure. A energy minimization step was then performed. The above procedures have now been included in the Materials and Methods section, page 18.

The authors have modeled missing residues in
The figure below shows the region 549-555 that was extracted from the homology model. We would like to make it clear that this small stretch of 9 amino acids connects the last helix of the catalytic domain to the Ct-extension, and is not involved in shielding the active site. The orientation of the Ctextension in our model is based solely on the crystal structure, and hence very unlikely to be biased by the small stretch taken from the homology model (please also see the figure in response to comment 1). The enhanced dynamics of helix 15 is seen only when bound with succinyl-CoA substrate in the simulation, as described in the response to the previous comment.
Modeling missing residues. The missing residues (residues 549-555) and the Ct-extension from monomer A of hsALAS2. For modeling the missing residues, a homology model of hsALAS2 was first constructed using scALAS2 (PDB ID: 5TXT) as a template. This homology model was structurally aligned with the crystal structure of hsALAS2, and the coordinates of the missing residues were extracted from the model. The extracted residues were then covalently linked to the crystal structure, and the structure was energy minimized.

Response:
We have now performed one replicate simulation for each of the three systems, by assigning a different set of initial velocities to all the atoms in the system. As shown in the figure below, the asymmetric behaviour in the salt bridge dynamics of the two monomeric subunits in the presence of substrates is observed in the replicate simulation; this is in agreement with the original simulation reported in the manuscript. We have added a sentence in the legend of Supplementary  Fig. 10 and 11 to indicate this is n=2 experiment.

Dynamics of salt bridges involving E569.
The variation of the distance between the residues participating in salt bridges is shown as a function of simulation time. To calculate the distance between two residues, the center of a given residue was defined as the center of mass of the side-chain guanidinium/amine/carboxylate group, depending on the amino acid type. The distance between the residue centers was then calculated. While R511-E569 is a native salt bridge present in the crystal structure, E569-R572 is a non-native salt bridge.
Dynamics of salt bridges involving E571. The variation of the distance between the residues participating in salt bridges is shown as a function of simulation time. To calculate the distance between two residues, the center of a given residue was defined as the center of mass of the side-chain guanidinium/amine/carboxylate group, depending on the amino acid type. The distance between residue centers was then calculated. While R293-E571 and K299-E571 are native salt bridges present in the crystal structure, E571-R572 is a non-native salt bridge.
4. The PLP in supplemental figure 7 seems different from the structure in figure 1e. The carbonyl oxygen is coloured blue, and points in a very different direction in figure S7.

Response:
We have now provided a corrected version of Supplementary Fig. 7.

The identified fragments are tested for their inhibitory capacity by a single point experiment. For an inhibitor of this kind, it may be expected that full inhibition is not attainable. Determination of an IC50 would be appropriate to confirm a concentration dependent inhibition.
Response: The fragment hits identified from our crystallography screening are low-molecular-weight compounds (average < 300 Da). In our experience from several fragment screening campaigns, these initial fragment hits are not expected to exhibit much inhibitory potency. We therefore thank the reviewer for understanding that full inhibition would not be attainable, and as such there would not be much meaning to calculate IC50 (half-maximal inhibition). However, as a follow-up to our original inhibition studies of wild-type hsALAS2 activity in the presence of 1 mM fragment concentrations, we have now determined wild-type enzyme activity in a separate set of experiments using 1 mM and 5 mM concentrations of fragments 1 and 8 (fragments that showed significant inhibition in the original activity assays at 1 mM) and have observed concentration-dependent inhibition. These data are incorporated into Supplementary Table 4, and cited in the text on page 11.
7. More importantly, the inhibition of the WT is not the relevant measure here. Would the authors not much rather show inhibitory behavior of these compounds against delAT or Q548X? Response: We evaluated the inhibitory effect of fragments using wild-type protein, but not delAT or Q548X, for two reasons: -The fragment hits of interest to us (1-8) are clustered around the Ct-extension; this region would not be present in the truncated delAT and Q548X variants.
-The primary motivation for our fragment-based discovery is to provide inhibitor starting point as substrate reduction therapy for those erythroid porphyria disorders caused by enzyme defects downstream of ALAS2 (namely UROS, UROD, FECH). It was outside the scope of our current fragment campaign to target the gain-of-function erythroid protoporphyria due to delAT or Q548X; if it were, we should be using a construct more relevant to these variant proteins, such as the catalytic domain-alone protein, which we currently have not crystallised.
We have now clarified the above points in the discussion section.  Supplementary Fig 8, to confirm its topology and correct placement in the electron density.

I suspect the problem the authors have is that most of the C-terminal region is unstructured, flexible, or disordered. At least some of the C-terminal region should be well-structured to warrant the crystallographic build-up in the narrative.
Response: As shown in the omit map described above ( Supplementary Fig. 8a), the majority of the Ct-extension is ordered and can be modelled from well-defined electron density. The only disordered segments in the vicinity are residues 549-557 (refer to reviewer 1) which connect the catalytic domain to the beginning of the Ct-extension, as well as the last 7 residues at the C-terminus. Both segments are not expected to play a role in shielding and gating the enzyme's active site.
Reviewer #3 (Remarks to the Author): The absence of quantitative data (kinetic data are shown without error bars, the number of replicates are not mentioned, values for kinetic parameters are not provided for wild-type versus mutant ALAS2) diminishes the study's rigor.

Response:
We have now corrected all these in the relevant figures and text reporting enzyme kinetics data.
Major 1. Line 126. The Km values for glycine and succinyl-CoA are not reported as stated in Fig S4a,b. These need to be noted with SDs in the text.

Response:
The Km values have been added as insets to the progression curves in Supplementary Fig.  4a,b, and also to the text on page 5.
2. Line 129. The data shown in Fig 4c does not establish interaction between ALAS2 and SUCLG1-SUCLA2, but rather a messy gel with many bands. If the authors consider this conclusion regarding