Structure of the Mitochondrial Aminolevulinic Acid Synthase, a Key Heme Biosynthetic Enzyme

Summary

5-aminolevulinic acid synthase (ALAS) catalyzes the first step in heme biosynthesis. We present the crystal structure of a eukaryotic ALAS from S. cereivisae. In this homodimeric structure, one ALAS subunit contains covalently bound cofactor, pyridoxal 5’-phosphate (PLP), whereas the second is PLP-free. Comparison between the subunits reveals PLP-coupled reordering of the active site and of additional regions to achieve the active conformation of the enzyme. The eukaryotic C-terminal extension, a region altered in multiple human disease alleles, wraps around the dimer and contacts active-site-proximal residues. Mutational analysis demonstrates that this C-terminal region that engages the active site is important for ALAS activity. Our discovery of structural elements that change conformation upon PLP binding and of direct contact between the C-terminal extension and the active site thus provides a structural basis for investigation of disruptions in the first step of heme biosynthesis and resulting human disorders.

The eTOC blurb

Brown et al. determine structures of ALAS, a heme biosynthetic enzyme, that reveal how its PLP cofactor orders the active site. These structures also reveal the positioning of the eukaryote-specific C terminal extension, providing a framework for understanding the mechanism of erythroid disease-causing mutations.

Introduction

Nearly all living organisms require heme as an enzyme cofactor, enviromental sensor and for oxygen transport by hemoglobin (Ponka, 1997). Heme biosynthesis is greatly upregulated during erythropoisesis to supply ligand for hemoglobin. Heme biosynthesis must be tightly regulated in all cells as both a lack or an excess of heme or of biosynthetic intermediates leads to several human diseases (Besur et al., 2014; Chiabrando et al., 2014; Egger et al., 2006; To-Figueras et al., 2011). In non-plant eukaryotes, 5-aminolevulinic acid synthase (ALAS) catalyzes the first committed step of heme biosynthesis, the condensation of glycine and succinyl-CoA to yield 5-aminolevulinic acid (ALA), within the mitochondrial matrix (Gibson et al., 1958; Kikuchi et al., 1958). ALAS is a member of a large family of enzymes that employ pyridoxal 5’-phosphate (PLP, the active form of vitamin B6) as a cofactor. We previously discovered that the AAA+ unfoldase ClpX promotes ALA synthesis by accelerating PLP binding to ALAS (Kardon et al., 2015).

Previous work described the structures of ALAS from Rhodobacter capsulatus bound to PLP and substrates (Astner et al., 2005). However, we were limited in our understanding of features specific to eukaryotic ALAS homologs. Although α-proteobacterial and eukaryotic ALAS homologs share ~70% sequence similarity (Astner et al., 2005), eukaryotic enzymes contain a unique 35–60 residue C-terminal extension not present in bacteria (Munakata et al., 1993). This extension is implicated in two human diseases resulting from mutation of the erythroid-specific isoform, ALAS2. Several human disease mutations that truncate this C-terminal extension produce hyperactive ALAS enzymes. These gain-of-function mutations induce toxic accumulation of the heme precursor protoporphyrin IX, causing X-linked protoporphyria (XLPP) (Balwani et al., 1993; Balwani et al., 2013; Bishop et al., 2012; Ducamp et al., 2013; Whatley et al., 2008). Many partial loss-of-function mutations in ALAS2, including several that map to the C-terminal extension, cause the opposite phenotype, resulting in reduced heme production leading to X-linked sideroblastic anemia (XLSA). The C-terminal extension of ALAS2 also binds succinyl-CoA synthetase, possibly to provide succinyl-CoA directly to ALAS (Furuyama and Sassa, 2000). Certain XLSA alleles of ALAS2 that map to the C-terminal extension disrupt this protein-protein interaction causing X-linked sideroblastic anemia (XLSA) (Bergmann et al., 2010; Bishop et al., 2012; Cotter et al., 1992; Ducamp et al., 2013).

Here we present the X-ray crystal structure of eukaryotic ALAS from Saccharomyces cerevisiae. We crystallized an asymmetric structure with PLP bound to only one subunit of the homodimer. By comparing the two subunits in the dimer to each other, as well as to bacterial ALAS structures (Astner et al., 2005), we observe substantial structural changes coupled to PLP binding. Structural elements that become ordered upon PLP incorporation directly contribute to configuring the active site of the enzyme for substrate binding and production of ALA. This structure also reveals the architecture of the eukaryotic C-terminal extension. It packs in an extended conformation against a surface-exposed hydrophobic groove on the cis subunit and interacts with residues near the active site. Our structure of S. cerevisae ALAS thus indicates how PLP binding is coupled to ordering the active conformation of the enzyme and provides further information to help dissect how mutation of the eukaryotic C-terminal extension contributes to two erythroid diseases.

Results

Structure of eukaryotic ALAS

We crystallized a variant of S. cerevisiae Hem1 (hereafter called ALAS_Sc) and determined the structure to 2.7 Å resolution using molecular replacement (Figure 1A; Table 1). The variant used for crystallization lacked the mitochondrial targeting sequence (residues 1–57, (Vogtle et al., 2009)), and the first 13 residues of the mature N-terminus were disordered in the final refined structure. The overall fold was similar to R. capsulatus ALAS (ALAS_Rc) with an RMSD of ~1.1 Å for common Cα atoms (Figure S1A, (Astner et al., 2005)). ALAS_Sc, like ALAS_Rc and related PLP-dependent enzymes, is a homodimer with PLP-binding pockets located at the dimer interface. In our structure, however, one active site contained PLP covalently bound to the catalytic lysine (Lys337–PLP, Figure 1B), whereas the other active site was PLP free (Figure S1B). This asymmetric occupancy by PLP was exhibited by all three dimers in the crystallographic asymmetric unit. Comparison of these asymmetric active sites with each other and with ALAS_Rc structures thus provided a view of the conformational changes that accompany PLP binding to ALAS. Residues from both subunits contribute to each active site, so for clarity, the pyridoxyllysine-bonded subunit and its residues, will be labeled with a B (e.g. ALAS^B, B for bound) whereas the subunit and its residues from the protomer lacking this covalent moiety will be labeled A (e.g. ALAS^A).

Figure 1. Structure of ALAS from S. cerevisiae.

A. Cartoon representation of the ALAS dimer with subunits A (unliganded) and B (containing covalently bound lysine-PLP (Lys-PLP)) colored tan and green, respectively. Lys–PLP is shown as blue spheres. The unoccupied cofactor-binding pocket is outlined with a blue dotted line. The structured N- and C-termini from Subunit A are labeled with name and residue number.

B. PLP binding pocket containing the covalently bound cofactor with 2mFo-Fc electron density for Lys–PLP ligand contoured at 1.0 σ and shown as grey mesh. Residues that directly interact with PLP are shown as sticks and are labeled according to their subunit with a superscript notation. See also Figure S1.

Table 1.

Crystallographic data collection and refinement statistics

	PLP-ALAS58–548
Data Collection
Space Group	P 1
a, b, c (Å)	63.49, 113.81, 119.34
α, β, γ (°)	116.52, 98.18, 92.56
Wavelength (Å)	0.9792
Resolution (Å)	50-2.70 (2.75–2.70) a
Z (molecules per ASU)	6
Total/unique reflections	307193/80317
Completeness (%)	98.8 (98.2)
Mean I/σ(I)	13.06 (2.65)
Data Redundancy	3.8 (3.8)
Rmergeb	14.1 (64.1)
Wilson B	31.69
Refinement
Resolution Range	35.0–2.7
Rworkc	0.1862
Rfree	0.2318
RMSD bonds (Å)	0.004
RMSD angles (°)	0.636
Ramachandran Favored (%)	98.44
Ramachandran Allowed (%)	1.56
Ramachandran Outliers (%)	0
Average B (Å2)	40.5
PDB ID	5TXT

^aHighest resolution shell data are shown in parentheses.

^bR_merge = Σ_hklΣ_i |I_i(hkl) - 〈I(hkl)〉 | / Σ_hklΣ_iI_i(hkl) where I_i(hkl) is the i^th observation of a symmetry equivalent reflection hkl.

^cR_work = Σ_hkl║F_obs| - F_calc║ / Σ_hkl |F_obs|, calculated over the 97.5% of the data in the working set. R_free is equivalent to R_work except that it is calculated over the remaining 2.5% of the data.

Several lines of evidence indicate that the observed structural asymmetry in PLP binding was not a consequence of obligatory half-occupancy but instead resulted from hydrolysis of the active-site lysine bond with PLP followed by dissociation. These crystals were grown in PLP-free solvent, rather than the PLP-supplemented solvent of the ALAS_Rc crystals (Astner et al., 2005). In previous work, both active sites of ALAS were found to contribute to catalysis, although a single active site was sufficient for activity, albeit at a reduced rate (Turbeville et al., 2011). We also purified and crystallized ALAS_Sc in which the covalent Lys–PLP bond had been chemically cleaved, converting PLP to PLP-oxime. This structure is a symmetric dimer with presumably this PLP derivative remaining non-covalently bound in both pockets and at full occupancy (Table S1; Figure S1C). Finally, we monitored the stoichiometry of the ALAS-PLP complex in solution and found that purified ALAS_Sc was initially fully occupied by PLP, but this occupancy declined after several days (Figure S1D).

PLP binding is coupled to disorder-order transitions in three distinct regions of ALAS

Comparison of the two subunits of the asymmetric ALAS_Sc dimer revealed three structural elements that were sensitive to the presence of bound PLP, which we termed NT, GR, and CT. These PLP-responsive regions were ordered when proximal to the PLP-containing active site but disordered when near the PLP-free active site (Figures 2A, 2B). NT (near the mature N-terminus, residues 83–113) was only ordered in subunit B. NT^B includes part of helix α1, an extended loop, strand β1, and the loop leading to β2. GR, which overlaps with a conserved glycine-rich motif (Gong and Ferreira, 1995), by contrast, was ordered only in the A subunit and consists of residues 147–156. The ordered GR^A element packs against the PLP cofactor-binding pocket that was in the B subunit. Part of the corresponding GR sequence coordinates succinyl-CoA and is essential for enzyme activity (Astner et al., 2005; Gong and Ferreira, 1995; Gong et al., 1996). CT is comprised of residues 538–548, the last 11 residues of the eukaryote-specific C-terminal extension in ALAS_Sc. The ordered CT^A element from the PLP-free subunit also contributed to the assembled active site of the opposite B subunit.

Figure 2. ALAS structure reveals regions that become ordered by PLP.

A. Overlay of ALAS_Sc subunits colored as in Fig. 1A, with emphasis on the PLP-ordered regions. NT (residues 83–113) is colored magenta. GR (residues 147–156) is colored orange. CT (residues 538–548) is colored cyan. All PLP-responsive regions are represented in three of the six molecules in the crystallographic asymmetric unit, however the lengths of NT and CT vary slightly depending on the molecule (NT in Chain D is residues 82–113 and CT in Chain C is residues 535–548). The segments reported here represent the shortest disordered stretch of residues among all comparable molecules.

B. Asymmetric ALAS_Sc homodimer (top panel is the same orientation as Figure 1A, Lys–PLP shown as blue spheres) with the NT, GR, and CT regions depicted at approximately 2X magnification, and colored as in panel A.

C. Overlay of PLP-binding pockets from both subunits. The major conformational changes between the unbound and bound pockets are shown as black arrows, with the arrowhead marking the position in the bound pocket. See also Figure S1B.

PLP-dependent differences in the active site

The two active sites in the asymmetric ALAS_Sc dimer displayed several distinct features (Figure 2C). In comparison to the active site in the PLP-free subunit A, Lys337^B, which forms the catalytically essential covalent bond with PLP, is rotated in toward the pocket. His209^B, likewise, rotated upward to provide hydrophobic stabilization of the PLP pyridinium ring. The side chains of Tyr183^B and Thr366^A shifted by 5–7 Å to coordinate the PLP phosphate group and formed an enclosed binding pocket (Figure 2C). Overall, PLP binding results in substantial movements of key residues leading to formation and stabilization of the functional conformation of the active site. The positioning of PLP-binding residues in the cofactor-containing pocket was very similar between eukaryotic ALAS_Sc and bacterial ALAS_Rc, and the identities of most of the residues mediating these interactions were also conserved (Table 2; Figure S2A).

Table 2.

Cofactor and Substrate binding residues as identified in (Astner et al., 2005)

ALASSc	ALASRc	ALAS2Hsa	Ligand binding role	Interaction
Arg91	Arg21	Arg163	sCoA	Succinate carboxylate coordination
Asn121	Asn54	Asn197	Gly	Gly α-carboxylate hydrogen bond
Thr150b	Thr83b	Thr226b	Gly / sCoA	Gating residue for Gly pocket specificityc/Succinate carboxylate coordination
Asn152b	Asn85b	Asn228b	sCoA	Succinate carboxylate coordination
Ile153b	Ile86b	Ile229b	sCoA	Succinate carboxylate coordination
Cys182	Ala115	Cys258	PLP / Gly	PLP phosphate coordinationc
Tyr183	Tyr116	Phe259	PLP	PLP phosphate coordination
Ser204	Ser137	Ser280	sCoA	Stabilize sCoA phosphoadenosine
Asp205	Asp138	Asp281	sCoA	Stabilize sCoA phosphoadenosine
Glu206	Ser139	Ala282	sCoA	Stabilize sCoA phosphoadenosine
His209	His142	His285	PLP / sCoA	Pyridine ring hydrophobic stabilization / delineates sCoA binding pocket
Ala210	Ala143	Ala286	sCoA	Delineates sCoA binding pocket
Ile213	Ile146	Ile289	sCoA	Stabilize sCoA phosphoadenosine
Ile216	Ile149	Ile292	sCoA	Stabilize sCoA phosphoadenosine
Lys223	Lys156	Lys299	sCoA	Stabilize sCoA phosphoadenosine
Ile225	Ile158	Val301	sCoA	Stabilize sCoA phosphoadenosine
Ser256	Ser189	Ser332	Gly	Gly α-carboxylate hydrogen bond
Asp281	Asp214	Asp357	PLP	PLP-salt bridge
Val283	Val216	Val359	PLP / Gly	Pyridine ring hydrophobic stabilizationc
His284	His217	His360	PLP	PLP-hydrogen bonding
Thr334	Thr245	Thr388	PLP	PLP phosphate coordination
Lys337	Lys248	Lys391	PLP	Covalent pyridoxyl-lysine bond
Phe365b	Phe276b	Phe419b	sCoA	delineates binding pocket
Thr366b	Ser277b	Thr420b	PLP	PLP phosphate coordination
Thr367b	Thr278b	Thr421b	PLP	PLP phosphate coordination
Thr452	Thr365	Thr508	sCoA	Succinate carboxylate coordination
Arg461	Arg374	Arg517	Gly	Gly α-carboxylate stabilization

^ahuman ALAS2

^bresidues from opposite subunit

^c(Shoolingin-Jordan et al., 2003)

Modeling substrate binding

The high structural similarity of the yeast and bacterial catalytic domains (0.8 Å RMSD for ALAS_Sc residues 120–385 and ALAS_Rc residues 53–296) allowed us to use ALAS_Rc– substrate structures to generate a model of ALAS_Sc with bound glycine and succinyl-CoA (Figure 3A) (Astner et al., 2005). This model, in turn, enabled us to determine the impact of PLP-binding and eukaryote-specific structural elements on the active-site region (discussed below). During a reaction cycle, the glycine substrate replaces the active-site lysine as the covalent binding partner of PLP (Ferreira and Dailey, 1993; Jordan, 1990) and thus in the ALAS_Sc model, lysine-bound PLP from ALAS_Sc and glycine-bound PLP from ALAS_Rc overlap, as expected (Figure 3). Most of the ALAS_Sc residues predicted to interact with both substrates were conserved with ALAS_Rc (Table 2; Figures S2B–C), and occupied positions similar to those in substrate-free (but PLP-containing) ALAS_Rc. Some of these residues shifted outward in the ALAS_Rc-substrate co-complexes, compared to their positions in our substrate-free ALAS_Sc structure (Figure S2C). In summary, five of the six glycine-coordinating residues are conserved between ALAS_Sc and ALAS_Rc, whereas eleven residues from ALAS_Sc subunit B and four from subunit A are conserved and predicted to be important for succinyl-CoA binding (Figure 3B, Figures S2D–E).

Figure 3. Model of ALAS_Sc active sites with cofactor and substrates.

A. ALAS dimer colored as in Figure 1 containing bound Lys–PLP (blue spheres) with glycine-PLP (yellow spheres) and succinyl-CoA (purple spheres) modeled in. Substrates were modeled based upon superposition of ALAS_Sc onto ALAS_Rc structures (PDB 2BWP and 2BWO).

B. Close-up view of the PLP-containing ALAS_Sc active site with modeled substrates. Residues known to interact with substrates based on alignment with ALAS_Rc are shown as sticks. Residues that become ordered by PLP binding (Arg91^B, Thr150^A, Asn152^A, Ile153^A) or move after PLP binding (His209^B and Phe365^A) are labeled with red font.

C. Close-up view of the PLP-free ALAS_Sc active site with modeled substrates. Residues Arg91^A, Thr150^B, Asn152^B, and Ile153^B are disordered with respect to this active site. See also Figure S2.

Based on comparison between the two active sites in our asymmetric dimer, we conclude that PLP plays a key role in positioning residues for binding both glycine and succinyl-CoA substrates. In the PLP-free active site of ALAS_Sc, several substrate-binding residues were disordered or adopted different conformations from the active site with bound PLP (Figure 3C). These changes include destabilization of Thr150^B, Asn152^B, and Ile153^B from GR^B and Arg91^A from NT^A as well as changes in the positions of His209^A and Phe365^B, which are predicted to coordinate succinyl-CoA. Importantly, in our symmetric, PLP-cleaved structure, none of these residues were disordered, supporting the conclusion that PLP binding imposes conformational changes that prepare the enzyme for catalysis.

PLP-ordered elements interact with each other and the PLP-bound active site

The active site of ALAS^B is comprised of elements from both subunits including NT^B, GR^A, and CT^A (the latter two elements functioning in trans). There are multiple intra- and intermolecular interactions between all three PLP-responsive elements that stabilize the active site containing bound PLP (Figure 4, Figures S3A–B). NT^B and GR^A interact extensively via multiple main-chain and side-chain interactions (Figure 4A, Figure S3A). NT^B also makes several main-chain and side-chain contacts with CT^A (Figure S3B). In addition, the 11 residues of CT^A were packed next to GR^A, and of special interest, the backbone carbonyl of Ser543^A in CT^A formed part of a hydrogen-bonding network with Arg151^A in GR^A and Glu164^A in the ALAS core (Figure 4B). Thus, our structural analysis reveals a direct contact between the eukaryotic C-terminal extension of ALAS and its catalytic center. This interaction suggests possible mechanisms for controlling eukaryotic ALAS activity, which may be disrupted by disease alleles or modulated by interaction with other proteins (see below and Discussion).

Figure 4. PLP stabilizes regions required for substrate binding.

A. Interactions between NT (magenta), GR (orange), and the active-site loop (brown). Thr452^B, a conserved residue in the active-site loop, interacts with Met257^B, Ile448^B, and Arg91^B in NT. GR packs against part of NT (see Figure S3A).

B. Interactions between GR (orange) and CT (cyan) stabilized by PLP. Hydrogen-bonding network between Arg151^A from GR, Glu164^A from the core, and the backbone carbonyl of Ser543^A in CT (cyan) depicted as black dotted lines. ALAS_Sc pyridoxyl-lysine (blue), ALAS_Rc pyridoxyl-glycine (yellow), and ALAS_Rc succinyl-CoA (purple) are shown as spheres. All of the colored regions are disordered in the absence of PLP, including Thr150, Asn152 and Ile153, which directly contribute to succinyl-CoA binding. See also Figure S2C,E.

Modeling of ALAS_Sc with substrates also allowed us to examine the relationship between PLP binding and the active-site loop (residues 444–461), previously established as an important region controlling ALA release, the rate-limiting step of the catalytic cycle (Lendrihas et al., 2010). Prior studies proposed that movement of the active site loop to allow ALA release might be regulated allosterically, and that the C-terminal region, perhaps in response to cellular binding partners, might contribute to this allosteric regulation. Loss of the C-terminus (as in the XLPP C-terminal mutations) would thus mitigate this control, resulting in a hyperactive enzyme. (Fratz et al., 2015; Lendrihas et al., 2010). The active-site loops from both subunits in our ALAS_Sc structure adopted an open conformation regardless of PLP occupancy and the conformation of the CT element (Figure S3C). Our structure thus neither supports nor rules out this model. However, the portion of the C-terminal extension that is upstream of the CT element might enact this control, or the C-terminus might modulate loop dynamics that are not captured in a static crystal structure.

Modulation of catalysis by the C-terminal extension

In ALAS_Sc, the eukaryote-specific C-terminal extension adopted an extended conformation, largely devoid of secondary structural features. The extension from each subunit packed around the outer surface of the same subunit and then contacted the partner subunit via the extreme C-terminal segment (Figure 5A). Contacts between the C-terminal extension and enzyme core were largely hydrophobic and buried ~1400 Ų of solvent-accessible surface area. A hydrogen bond between the backbone carbonyl of Ser543^A in CT^A and the side-chain of the universally conserved ALAS residue Arg151^A in GR^A, which is directly between two succinyl-CoA binding residues (Table 2), appeared to stabilize the functional catalytic center (Figure 4B). To assess the importance of this interaction, we determined enzyme activity (V_max) for purified wild-type ALAS_Sc, a variant with Arg151 mutated to alanine (R151A), and a variant lacking residues 535–548 (ΔCT). The V_max of the R151A variant was 20% of that of wild-type. (Figure 5B). Thus, the conserved Arg151 residue is important for ALAS_Sc enzyme activity. The ΔCT variant also had reduced activity (~65% of wild type), revealing a role for the CT element in optimal ALAS_Sc enzyme function (Figure 5B). To further test the importance of the C-terminal extension in activity, we attempted to purify bacterially expressed variants with larger deletions of the C-terminal extension, but all were poorly soluble. Thus, truncations of the C-terminal extension beyond the CT element appear to destabilize the structure and/or folding of ALAS_Sc.

Figure 5. Structure and function of the ALAS eukaryote-specific C-terminal extension.

A. The asymmetric ALAS_Sc dimer is shown as a surface representation with subunits colored tan or green. The eukaryote-specific C-terminal extension (residues 489–548, cyan cartoon) wraps around the outer surface of the dimer and packs into a hydrophobic channel.

B. In vitro activity of ALAS_Sc WT, R151A mutant, or ΔCT (residues 58–534) at 30 °C. Experiments were performed in triplicate and error bars represent SEM.

C. Growth of yeast strains harboring C-terminal deletion mutants of ALAS_Sc. The indicated variants in HEM1 (the gene encoding ALAS_Sc) were integrated at its single genomic locus in W303a strain background (see Table S2). Five-fold serial dilutions from cell suspensions with OD₆₀₀ = 1 were spotted on YP + 2% agar, + 2% glucose or 3% glycerol, ± 50 µg/mL ALA as indicated, and grown for 2 d (+ glucose) or 3 d (+ glycerol) at 30 °C.

D. Cellular protein levels of C-terminal ALAS_Sc variants. ALAS_Sc was detected by FLAG antibody (M2 clone, Sigma), and the mitochondrial outer-membrane porin Por1 was probed as a loading control.

E. ALA levels in yeast cell extracts were measured using modified Ehrlich’s reagent (described in STAR methods) and normalized to wild type. Data are represented as mean ± SD; p ≤ 0.005 for ALA reduction in all HEM1 mutants (Student’s t-test, n = 3 (biological replicates)). See also Figure S4.

To determine the contribution of the C-terminal region of ALAS_Sc to in vivo function, we deleted the entire C-terminal extension (hem1ΔC; missing residues 493–548) or the CT segment (hem1ΔCT) from the single genomic copy of HEM1 in haploid yeast (Table S2). To allow assessment of protein levels, we also appended a C-terminal 3xFLAG tag to all ALAS_Sc variants. This epitope tag did not perturb wild type protein function in vivo, likely because the extreme C-terminal residue (Gln548) is surface-exposed and does not directly contact the active site (Figure 4B). The ΔCT variant supported normal growth on both fermentable and non-fermentable carbon sources (Figure 5C) and western blots revealed increased levels of the ΔCT protein relative to wild type (Figure 5D). This increase in ΔCT abundance may be due to feedback upregulation of expression to compensate for reduced activity, or to reduction in turnover of ALAS. In contrast, the ΔC mutant strain was not viable unless grown on medium supplemented with ALA (Figure 5C). The cellular level of ΔC variant protein was reduced compared to wild type (Figure 5D), again suggesting that truncation of the C-terminal extension destabilizes ALAS_Sc. To further probe the importance of the C-terminal region to ALAS activity in vivo, we assessed cellular ALA levels in the wild type and hem1 variant strains. To permit assessment of hem1ΔC, which requires ALA supplementation for growth, all strains were assayed following outgrowth (~4–5 generations) in ALA-free medium from an ALA-supplemented preculture. Both C-terminal truncation strains produced a low level of ALA that was statistically indistinguishable from the background signal of a complete null (hem1Δ) within the sensitivity of this assay (Figure 5E). Therefore, although the ΔCT variant produced enough ALA to sustain viability, loss of this structural element caused a substantial defect in ALA production in vivo. Taken together, the effects of C-terminal truncation on ALAS_Sc function in vitro and in vivo lead us to conclude that the eukaryote-specific C-terminal region of ALAS_Sc can strongly influence ALAS activity, stability and cellular function.

Discussion

In our asymmetric structure of the ALAS_Sc homodimer, only one of the two active sites contains PLP, providing a view of the enzyme in both an active, bound-cofactor state and an inactive, cofactor-free state. Comparison of these two states reveals prominent changes in the position and orientation of several PLP-binding residues. Based upon comparison with the ALAS_Rc–substrate structures, binding of PLP also positions and stabilizes multiple residues that are important for substrate binding, thus structuring the larger active site. Additionally, several of these substrate-binding residues are located in one of the three structural elements that together become ordered in proximity to a PLP-containing active site. As a consequence, the PLP-coupled changes restructure, stabilize, and ready the enzyme for ALA synthesis.

This work also supplies a structural framework for probing the mechanism by which mitochondrial ClpX promotes PLP binding to ALAS. The disordered structural elements in the PLP-free form of ALAS we observe may also represent regions that ClpX recognizes and/or remodels to promote PLP binding. Additionally, there are multiple XLSA mutations for which the mechanistic perturbation in ALAS is unknown (see Table S3 and (Astner et al., 2005)). Loss of contacts with ClpX at the site of mutation or perturbation of its remodeling of ALAS could underlie the phenotypes of some of these disease alleles.

Our work also provides the structure of the ALAS eukaryote-specific C-terminal extension. This element wraps around the outer surface of the ALAS dimer, placing a peptide segment adjacent to the C-terminus in close proximity to the active site of the opposite subunit. Indeed, residue Ser543^A, six residues from the C-terminus, interacts with residue Arg151^A within the GR element that modulates substrate binding in the active site (Figure 4B). This residue is universally conserved, and is mutated in an XLSA disease allele in humans (R227C) (Harigae and Furuyama, 2010; Katsurada et al., 2016). We found that an R151A substitution reduced ALAS_Sc activity by approximately 80% (Figure 5B). Our observation that PLP binding results in stabilization of the CT element as well as the glycine-rich GR element may provide insight into mechanisms of enzyme regulation and perturbations leading to XLPP. A previously proposed model for hyperactivation of ALAS by C-terminal truncations in XLPP posited that the C-terminal extension could stabilize the dynamics of the active-site loop, potentially by an allosteric mechanism (Fratz et al., 2015). C-terminal truncations might thus promote the open conformation of the active site and accelerate the rate-limiting step of product release. In our structure, we observe no direct contacts between the C-terminal extension and the active-site loop. Moreover, the active-site loop is in the open conformation irrespective of whether the CT element is ordered or disordered, and disruption of the interaction between CT and GR decreased rather than increased enzyme activity. One possible explanation for XLPP ALAS2 hyperactivation is the close proximity of the N-terminal portion of the C-terminal extension to the active-site loop, which may restrict the dynamics of this segment in some fashion. It is also uncertain whether the modest increase in isolated enzyme activity (three-fold or less) caused by XLPP C-terminal truncations of human ALAS2 accounts for the order-of-magnitude greater increase in porphyrin levels in patients (Ducamp et al., 2013). In addition to forming a direct contact with the ALAS_Sc active site, the C-terminal extension could modulate of ALAS function by mediating interaction with other mitochondrial proteins

One demonstrated interaction partner of the C-terminal extension of ALAS2 is the β subunit of succinyl-CoA synthetase (SCS). Binding of SCS to ALAS2 could facilitate direct transfer of succinyl-CoA to the ALAS active site (Furuyama and Sassa, 2000). We modeled residues demonstrated to perturb SCS-ALAS2 binding on to our ALAS_Sc structure (Figure S4). The proposed SCS-binding residues localize to solvent-exposed regions on the surface of the ALAS dimer, both on the conserved core of the enzyme (previously observed by Astner et al.), and on the eukaryote-specific C-terminal extension. Interestingly, the C-terminal sites that we model partly cap the succinyl-CoA binding pocket in our ALAS_Sc structure (Figure S4, (Furuyama and Sassa, 2000)). However, this model must be interpreted with caution; the sequence and length of the C-terminal extension differ between yeast and humans and thus the position of the C-terminal SCS-binding interface in vertebrate ALAS proteins may vary from this modeled site (Figure S4). Among the mutations found to perturb the human SCS-ALAS2 interaction, several within the C-terminal extension cause decreased heme synthesis and XLSA (Bishop et al., 2012; Furuyama and Sassa, 2000). Because truncation of this region can also hyperactivate ALAS leading to accumulation of the heme precursor protoporphyrin IX and XLPP, SCS binding and truncation of the C-terminal extension may induce similar conformational changes in the enzyme, perhaps through the CT-GR element contacts we observe. A recent study proposed that vertebrate ALAS2 forms a complex with other mitochondrial heme biosynthetic enzymes (Medlock et al., 2015).

The eukaryotic C-terminal extension, through its contact with the active site and influence on activity, thus may provide a hub for macromolecular interactions that facilitate and regulate heme biosynthesis.

STAR Methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Tania A. Baker (tabaker@mit.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

E. coli growth conditions

E. coli BL21(DE3) harboring the rare codon plasmid pRIL (Cam^R) were transformed with modified pET28b plasmid (Kan^R) and plated on LB/kan/cam agar plates. Starter cultures were grown at 30°C in LB supplemented with antibiotics for approximately 18 h. Large-scale expression (1L) were inoculated with 10 mL of starter culture and 1:1000 dilution of antibiotic and grown at 30°C to OD_600nm 0.5–0.7. Each culture was induced with 0.5 mM IPTG and proteins were expressed at 25°C for 4 h.

Yeast growth conditions

All strains were made in the W303 MATa background. Yeast were grown at 30°C, with shaking at 220 rpm for liquid cultures. Unless otherwise indicated, yeast were grown in 2% glucose, 1x CSM amino acids, and 0.67% w/v yeast nitrogen base supplemented with 50 µg/L ALA.

METHOD DETAILS

Protein expression and purification

Saccharomyces cerevisiae ALAS (Hem1) was expressed and purified as described (Kardon et al., 2015). Briefly, each construct was expressed in a modified pET28b vector containing an N-terminal hexahistidine tag followed by the SUMO fusion protein in E. coli strain BL21(DE3) harboring the rare codon plasmid pRIL (Stratagene). Site-directed mutagenesis was performed using primers synthesized by Integrated DNA Technologies (IDT) to generate yeast ALAS_58–548 R151A. Proteins were grown at 30 °C in LB media supplemented with kanamycin and chloramphenicol to an OD₆₀₀ ~0.5–0.7. Expression was induced with 0.5 mM IPTG and 10 mg/mL sodium ascorbate, and cultures were then shaken at 22 °C for 4–5 h, followed by cell ha rvesting. For purification, cells were resuspended in 25 mM HEPES-KOH (pH 8.0), 100 mM KCl, 400 mM NaCl, 2 mM MgCl₂, 20 mM imidazole, 10% glycerol, 10 mM 2-Mercaptoethanol, and 20 µM PLP and lysed via French press. The lysate was clarified by centrifugation and the supernatant added to Ni²⁺-NTA agarose resin equilibrated in buffer and incubated at 4 °C for 90 min. After washing the resin, the tagged protein was eluted with 250 mM imidazole. The His₆-SUMO tag was cleaved overnight with SUMO protease (Wang et al., 2007) followed by a second Ni²⁺-NTA step to remove the affinity tag. At this stage, the protein was split into two aliquots. The first, fully PLP-occupied portion was purified with a final gel-filtration step (Superdex 200 16/600) with buffer exchange into 25 mM HEPES-KOH (pH 7.6), 100 mM KCl, 10% glycerol, 1 mM DTT, and 20 µM PLP. The second portion was diluted five-fold into 0.1 M potassium phosphate (pH 7.5), 10% glycerol, 1 mM DTT and treated with 5 mM hydroxylamine-HCl overnight at 4 °C to cl eave the pyridoxyl-lysine bond. A final gel-filtration step with exchange into storage buffer lacking PLP yielded >95% pure protein.

Protein crystallization

Immediately before setting up crystallization trials, proteins were buffer exchanged into 25 mM HEPES (pH 7.5), 100 mM KCl, 1 mM DTT and concentrated using a 30 kDa molecular weight cutoff centrifugal filter (Amicon) to either 14.3 mg/mL (PLP-bound ALAS) or 12.9 mg/mL (PLP-cleaved ALAS). Crystallization plates were set up with a Phoenix liquid handler (Art Robbins) at 18 °C . For covalently bound ALAS-PLP, crystals grown in 0.1 M tri-sodium citrate (pH 5.5), 20% PEG 3000 were fully formed after two weeks, harvested after one month, and flash-frozen in liquid nitrogen after a brief transfer into well solution containing 20% glycerol for cryoprotection. The process was repeated for PLP-cleaved ALAS, however these crystals grew in 0.2 M potassium formate, 20% PEG 3350.

Data Collection and Structure Determination

High-resolution native data were collected at a single wavelength at NE-CAT beamline 24-ID-E using an ADSC Quantum 315 detector. Data were indexed, integrated, and scaled with DENZO and SCALEPACK from the HKL-2000 program package (Otwinowski and Minor, 1997). Phenix Xtriage confirmed the space group was P1 for PLP-ALAS and P2₁ for PLP-cleaved ALAS. Molecular replacement was performed for the higher-resolution PLP-cleaved ALAS data set with Phenix AutoMR using a single monomer of ALAS from R. capsulatus (PDB 2BWN with the cofactor removed) as the search model. The initial MR model and electron-density map were optimized with Phenix Autobuild. The PLP-cleaved ALAS structure contains two molecules (one biological dimer) and two PLP derivatives, neither of which are covalently bound to the protein. To solve the covalent ALAS-PLP structure, molecular replacement was performed using a single monomer from the PLP-cleaved ALAS model. A single MR solution contained six molecules (three biological dimers) in the ASU and three covalently bound PLP molecules (one per ALAS dimer). A representative view of the PLP electron density in PLP-bound subunit is shown in Figure 1B. Rounds of iterative model building and refinement were performed using Coot and Phenix, respectively (Adams et al., 2010; Emsley et al., 2010). All figures were generated using PyMOL (Schrodinger, 2015).

ALAS activity assays

ALAS activity was determined using a colorimetric NAD⁺-coupled assay as described (Hunter and Ferreira, 1995). Briefly, 1 µM of ALAS in 25 mM Hepes, pH 7.6, 100mM KCl, 5 mM MgCl₂ was mixed with 100 mM glycine and 50 µM succinyl-CoA to produce ALA and CoA. The reaction mixture also contained 1 mM NAD+ and 1 mM α-ketoglutarate (AKG), which in the presence of 1 mM AKG dehydrogenase and 0.25 mM thiamine pyrophosphate, produces succinyl-CoA and NADH. This enabled measurement of the reduction of NAD+ to NADH to assess ALAS activity. The reaction was carried out at 30 °C and the increase in fluorescence over time was monitored at 340 nm in a plate reader using a 384-well plate. All activity assays were repeated in triplicate and reported errors represent the standard error of the mean.

PLP dissociation

Purified ALAS_Sc (20 µM monomer equivalents) was incubated in 25 mM HEPES-KOH (pH 7.6), 100 mM KCl, 5 mM MgCl₂, and 10% glycerol at 22 °C. At indicated times, aliquots were withdrawn and exchanged into fresh buffer using Zeba spin columns (ThermoFisher Scientific). PLP retained in the buffer-exchanged sample was converted to a free fluorescent form by treatment with semicarbazide-HCl, protein was removed by precipitation with perchloric acid, and fluorescence (excitation 380 nm; emission 460 nm) was monitored in the neutralized supernatant as described (Srivastava and Beutler, 1973). Fluorescence values were converted to PLP concentration by comparison with semicarbazide-treated PLP standards.

Western Blotting

Yeast cell extracts (prepared according to (von der Haar, 2007)) were separated by SDS-PAGE and tranferred to PVDF in Tris-glycine running buffer with stirring at 4°C for 1 h at 100V. The PVDF membrane was then incubated with gentle rotation at room temperature in the following series of solutions: ≥ 1h in 2% milk/TBST, ≥ 1h in 1:1000 mouse anti-FLAG or Por1 antibody in 2% milk/TBST, once briefly and twice ≥ 10 min in 2% milk/TBST, ≥ 1h in 1:3000 goat anti-mouse alkaline phosphatase-conjugated antibody in 2% milk/TBST, and once briefly in 2% milk/TBST and three times ≥ 10 min in TBST. The membrane was then incubated briefly with ECF reagent (a fluorogenic substrate for alkaline phosphatase) and scanned with a Typhoon FLA 9500 (ex. 473 nm, em. ≥575 nm).

ALA measurement

Yeast cells were grown to OD₆₀₀ of 1.0–1.2, and an equivalent of 10 mL of culture at OD₆₀₀=1.0 was rapidly harvested by vacuum filtration through a 25 mm diameter, 45 µm pore nylon membrane. Cell-laden filters were immediately immersed in 0.7 mL 10% trichloroacetic acid on ice. After 15 min, the extract was withdrawn and centrifuged for 10 min at 21,000 × g. 150 µL of supernatant was mixed with 50 µL volume 8% acetylacetone in 2 M sodium acetate and incubated at 90°C for 15 min. After 5 min at room temperature, 150 µL of the resulting solution was mixed in a 96-well clear polystyrene plate with an equal volume of modified Ehrlich’s reagent (2% w/v 4-(dimethylamino)benzaldehyde, 84% (v/v) glacial acetic acid, 16% (v/v) 70% perchloric acid; 4-(dimethylamino)benzaldehyde dissolved in acetic acid before mixing with perchloric acid (Mauzerall and Granick, 1956) and incubated for 15 min at room temperature, upon which the absorbance at 552 nm and 650 nm was measured. ALA content was proportional to A₅₅₂-A₆₅₀.

QUANTIFICATION AND STATISTICAL ANALYSIS

For in vitro ALAS activity assays, experiments were performed in triplicate for each protein construct and average ± SEM are reported. Data for ALA levels in yeast cell extracts are represented as mean ± SD; p ≤ 0.005 for ALA reduction in all HEM1 mutants (Student’s t-test, n = 3 biological replicates). For PLP dissociation measurements, means ± SD of samples from three independent incubations are shown. Statistical details of each experiment can be found in the figure legends.

DATA AND SOFTWARE AVAILABILITY

Data Resources

The PLP-cleaved ALAS_58–548 and PLP-bound ALAS_58–548 structures were deposited in the Protein Data Bank (Berman et al., 2000) with accession codes 5TXR and 5TXT, respectively.

Highlights.

• ALAS undergoes PLP cofactor-coupled ordering of three distinct structural elements

• PLP binding helps properly position substrate-binding residues in the active site

• The eukaryotic C terminal extension interacts directly with the active site region

• This structure provides a framework for understanding human ALAS disease alleles