On the quaternary structure of human D‐3‐phosphoglycerate dehydrogenase

Abstract

D‐3‐phosphoglycerate dehydrogenase (PHGDH) catalyzes the NAD⁺‐dependent conversion of D‐3‐phospho‐glycerate to 3‐phosphohydroxypyruvate, the first step in the phosphorylated pathway for L‐serine (L‐Ser) biosynthesis. L‐Ser plays different relevant metabolic roles in eukaryotic cells: alterations in L‐Ser metabolism have been linked to serious neurological disorders. The human PHGDH (hPHGDH), showing a homotetrameric state in solution, is made of four domains, among which there are two regulatory domains at the C‐terminus: the aspartate kinase‐chorismate mutase‐tyrA prephenate dehydrogenase (ACT) and allosteric substrate‐binding (ASB) domains. The structure of hPHGDH was solved only for a truncated, dimeric form harboring the N‐terminal end containing the substrate and the cofactor binding domains. A model ensemble of the tetrameric hPHGDH was generated using AlphaFold coupled with molecular dynamics refinement. By analyzing the inter‐subunit interactions at the tetrameric interface, the residues F418, L478, P479, R454, and Y495 were selected and their role was studied by the alanine‐scanning mutagenesis approach. The F418A variant modifies the putative ASB, slightly alters the activity, the fraction of protein in the tetrameric state, and the protein stability; it seems relevant in dimers' recognition to yield the tetrameric oligomer. On the contrary, the R454A, L478A, P479A, and Y495A variants (ACT domain) determine a loss of the tetrameric assembly, resulting in low stability and misfolding, triggering the aggregation and hampering the activity. The predicted tetrameric interface seems mediated by residues at the ACT domain, and the tetramer formation seems crucial for proper folding of hPHGDH, which, in turn, is essential for both stability and functionality.

1. INTRODUCTION

L‐serine (L‐Ser) is a non‐essential proteinogenic amino acid that plays an important role in the metabolism of eukaryotic cells. It serves as a precursor for many molecules essential for proliferation, growth, cell differentiation, and functionality (Hirabayashi & Furuya, 2008; Kent, ¹⁹⁹⁵). Moreover, L‐Ser is the precursor of two neurotransmitters involved in glutamatergic transmission: glycine (Gly) and D‐serine (D‐Ser), which act as co‐agonists of N‐methyl‐D‐aspartate receptors (Ferreira et al., 2017; Papouin et al., ²⁰¹²; Wolosker et al., ²⁰¹⁶). L‐Ser is reversibly converted to Gly through the reaction catalyzed by the enzyme serine hydroxymethyltransferase (EC 2.1.2.1) (Wang et al., 2013) and to D‐Ser through the reversible reaction catalyzed by serine racemase (EC 5.1.1.18) (Wolosker et al., 1999). L‐Ser metabolism plays an important role in the development and functioning of the central nervous system (Furuya et al., 2008; Yoshida et al., ²⁰⁰⁴). It has been shown that patients with low levels of L‐Ser and Gly in plasma and cerebrospinal fluid are affected by serious neurological disorders (Tabatabaie et al., 2011; van der Crabben et al., ²⁰¹³).

In the human brain, L‐Ser is produced de novo through the phosphorylated pathway (PP) (Snell & Fell, 1990; Tabatabaie et al., ²⁰¹⁰). This irreversible cytosolic pathway starts with the glycolytic intermediate D‐3‐phosphoglycerate (3PG) and uses three enzymes (Murtas et al., 2020; Yang et al., ²⁰¹⁰): D‐3‐phosphoglycerate dehydrogenase (PHGDH, EC 1.1.1.95) catalyzes the conversion of 3PG to 3‐phosphohydroxypyruvate (PHP) using the oxidized form of nicotinamide adenine dinucleotide (NAD⁺) as cofactor (Murtas et al., 2020); phosphoserine aminotransferase (EC 2.6.1.52) converts PHP to 3‐phosphoserine (3PS) through a transamination reaction, using glutamate as an amino donor and pyridoxal 5‐phosphate as a cofactor (Murtas et al., 2020); phosphoserine phosphatase (EC 3.1.3.3) catalyzes the irreversible hydrolysis of 3PS to L‐Ser with the release of an inorganic phosphate (Murtas et al., 2020; Zhao et al., ²⁰²⁰).

The gene encoding the human PHGDH (hPHGDH) has been mapped on chromosome 12; it encodes a 533 amino acid long cytosolic protein with a predicted molecular mass of 56.8 kDa. At equilibrium, the reaction catalyzed by hPHGDH is almost totally shifted toward the formation of 3PG and is driven toward PHP by its own consumption in the subsequent reactions of the PP (Grant, 2012, 2018). PHGDH is a member of the isomer‐specific 2‐hydroxyacid dehydrogenase family, generally specific for substrates in D‐configuration (Grant, 1989). This enzyme is present in various organisms. Three types of PHGDH (types I, II, and III) have been classified according to size and domain composition: hPHGDH belongs to the type I group, along with its Mycobacterium tuberculosis (MtPHGDH), mouse, rat, and rabbit homologs, and shows a homotetrameric state in solution (Murtas et al., 2021). All types of PHGDH contain two common and conserved domains: the substrate‐binding domain and the cofactor‐binding domain. The type I members contain two additional regulatory domains at the C‐terminus: the aspartate kinase‐chorismate mutase‐tyrA prephenate dehydrogenase (ACT) and allosteric substrate‐binding (ASB) domains (Figure S1), showing low sequence conservation. In selected species, the ACT domain has been reported to function as a binding site for serine to provide feedback inhibition, although this regulatory mechanism was not evident for hPHGDH (Murtas et al., 2021). The 3D structure of hPHGDH has been solved only for a truncated form harboring the substrate and the cofactor binding domains (residues 6–308, named shPHGDH) (Unterlass et al., 2017), which is present in solution as a dimer. Recent efforts to solve the structure of full‐length hPHGDH by x‐ray crystallography or cryo‐EM were unsuccessful (Loredano Pollegioni, 2024, personal communication).

Recently, we reported that in differentiated astrocytes, the enzymes of the PP are in close proximity, yielding a multimeric complex named the serinosome (Rabattoni et al., 2023). In this metabolon, the three enzymes are kinetically coupled under in vitro conditions: their assembly provides a channeling solution for the pathway intermediates, giving higher local concentration and efficient processing of intermediates, preventing their bulk equilibration with the cytosol, thus increasing the pathway flux and definitely pushing the process to L‐Ser production.

The structural and functional properties of the three human enzymes of the PP have been recently investigated (Marchesani et al., 2023, 2024; Murtas et al., ²⁰²¹). Due to the cytosolic location of the PP and its connection with glycolysis, knowledge of its organization is of main relevance in order to elucidate the reversible generation of the serinosome and the modulation of L‐Ser synthesis under both physiological and pathological conditions. In this work, we delve into the oligomerization mode of hPHGDH through a combination of in silico structural analyses and mutagenesis studies, allowing us to propose the involvement of specific regions/residues at the C‐terminal end of the enzyme in tetramer formation.

2. RESULTS

2.1. Prediction of quaternary structure of hPHGDH

The experimental structure of the N‐terminal part (corresponding to residues 6–308) of hPHDGH (Uniprot ID: O43175) is available as dimer (Protein Data Bank (PDB) ID: 6plg). This region is responsible for the dimerization and forms the substrate and NAD⁺ binding sites. The C‐terminal region (corresponding to positions 309–533) comprehends two domains, putatively responsible for allostery (ASB domain) and regulation (ACT domain) (Grant, 2018) (Figure S1). Moreover, it is known that the C‐terminal region of the homologous MtPHGDH (Uniprot ID: P9WNX3) mediates tetramerization. Therefore, considering its functions, it is of primary importance to know the structure of each hPHGDH subunit and of the full tetramer to explain the effects of molecular interactions with metabolic intermediates and the molecular mechanism by which pathogenic single‐point missense mutations, especially at the C‐terminal, are responsible for hPHGDH dysregulation/inactivity (see Table 1 in Murtas et al., ²⁰²⁴). This region, never solved experimentally, shows no similarity in primary sequence with the corresponding region of the MtPHGDH enzyme, thus not allowing for “template‐based” modeling (Grant, 2018).

TABLE 1.

List of pairwise non‐covalent interactions with a distance ≤6 Å at the tetrameric interface of the human D‐3‐phosphoglycerate dehydrogenase model, observed in at least 50% of the conformers sampled by ClustENMD analysis.

Residue 1	Residue 2	Type of interaction	Frequency (% of conformers)
Y495	Y495	π‐stacking	94.5
R454	E456	Salt bridge	86.0
P479	I482	Hydrophobic contact	68.0
L478	L492	Hydrophobic contact	67.5
L478	L482	Hydrophobic contact	66.0
P475	L492	Hydrophobic contact	60.0
L478	L478	Hydrophobic contact	55.0

At first, a model of the hPHGDH tetramer was attempted by AlphaFold‐Multimer: only the dimeric interface was predicted in accordance with the experimental structure, while large interpenetration occurred between the two dimers, with an ipTM score <0.7 (data not shown). In a second approach, we used the predicted structural model of hPHGDH C‐terminal region from the AlphaFold database, assuming the same mechanism observed for the tetramerization of MtPHGDH (see Section 4). The resulting model was refined by three conformational analyses guided by normal mode analysis and MD simulations of the sampled conformers. The structural diversity of the obtained ensemble was summarized by a cluster analysis, resulting in 18 clusters: the centroids of the first three clusters by size, representing 29.5%, 28.7%, and 27.9% of conformers (the fourth cluster represents only 2.9% of conformers), are reported in Figure 1a. The root mean square fluctuation of backbone atoms (Figure 1b) spikes in the first 100 residues of the N‐terminal and at the hinge region that connects the N‐terminal to the C‐terminal region (309–533). The C‐terminal portions that present the highest flexibility correspond to a loop rich in glycine and proline (412–421) with F418 as the only bulky residue (in the ASB domain), and to the ACT domain, that directly participates in forming the tetrameric interface (Figure 1c). F418 is not involved in any stable oligomeric interaction and is in the middle of a highly flexible loop forming a disordered arm facing the lower margin of the tetrameric interface, suggesting its possible role in improving the formation of the tetrameric contact between the C‐terminal domains of each dimer.

FIGURE 1.

Human D‐3‐phosphoglycerate dehydrogenase tetrameric model after ClustENMD refinement. (a) Backbone‐superposed centroid structures of the first three clusters resulting from cluster analysis of sampled conformers. Different colors are used to depict the four subunits and N‐and C‐terminal regions. (b) Root mean square fluctuation (RMSF) averaged over all sampled conformers and subunits and average Aggrescan3D values: scores >0 are reported for the ensemble of the full tetrameric model (black filled circles) and analyzing dimers only (red crosses). (c) Tetramerization interface with a focus on residues involved in inter‐chain interactions for >50% sampled conformers (reported in Table 1) depicted as ball‐and‐sticks with side‐chain oxygen, nitrogen, and polar hydrogen atoms in red, blue and white, respectively. L478 and P479 of different chains are overlapped in the visualization and named only once. The models' images were obtained with UCSF ChimeraX v.1.5 (Pettersen et al., 2021). ACT, aspartate kinase‐chorismate mutase‐tyrA prephenate dehydrogenase; ASB, allosteric substrate‐binding.

By analyzing the inter‐subunit interactions at the tetrameric interface along the ensemble (reported in Table 1 and graphically exemplified in Figure 1c) (Pettersen et al., 2021), Y495 and the two charged residues R454 and E456 interact (for most of the conformers) by forming π‐stacking and a salt bridge interaction, respectively. The other interactions spanning more than 50% of the conformers are all contacts between the side chains of small hydrophobic residues, that include P475, L478, P479, I482, and L492.

Overall, the in silico predicted stability of the interactions at the tetramerization interface and the concomitant presence of substantial mobility of the C‐terminal region over generated conformers may indicate that large collective movements are present in the hPHGDH tetrameric enzyme assembly, allowing to regulate the substrate/cofactor accessibility and activity at the N‐terminal region. We performed an Aggrescan3D analysis over ensembles of tetramers and dimers (obtained by analyzing each dimer separately): results indicate that aggregation hotspots are overlapped at N‐terminal and ASB domains and that the removal of the tetrameric interaction will cause one aggregation hotspot to appear in the ACT domain of the dimers at the tetrameric interface (positions 478–482), and an additional one (positions 492–500, which involves many residues not directly participating in the interface) to be strengthened (Figure 1b). To validate the generated hPHGDH quaternary ensemble and assess the impact that the tetrameric structure has on the regulation of enzyme functionality, an experimental alanine‐scanning approach was performed on residues selected from those predicted to be relevant for tetramer formation. Five single‐point variants were generated (indicated in Figure S1). These were chosen to cover at least one of all the non‐covalent residue pairs across the interface (Table 1), with preference for residues with a predicted higher flexibility considered a criterion to favor a more severe interface destabilization. The modified positions do not correspond to known pathological single nucleotide polymorphisms (SNPs) in hPHGDH (Murtas et al., 2024).

2.2. Recombinant expression and purification of hPHGDH variants

For expression of the F418A, R454A, L478A, P479A, and Y495A hPHGDH variants, optimized protocols were set up for each variant, starting from the one previously set up for the wild‐type enzyme and some pathological variants (Murtas et al., 2021, 2024), see Supporting Information. The highest expression level was 90 mg/L for F418A, 58 mg/L for R454A, 5 mg/L for L478A, and 77 mg/L for P479A (Tables S1 and S2). For the wild‐type (WT) hPHGDH, 95 mg of purified recombinant protein/L was obtained, a figure comparable to the one for the F418A variant in terrific broth (TB) medium and for P479A in Luria Bertani (LB) medium. Because the expression of the Y495A variant was very low in all conditions tested (~2 mg/L), its biochemical characterization was not performed.

The hPHGDH variants were purified by chelating chromatography with a ≥90% purity degree (Figure S2). All the purified single‐point variants showed significantly lower specific activity than the wild‐type hPHGDH under standard conditions, below the detection level for both L478A and P479A variants (see Table S2).

The effect of temperature on the protein folding/unfolding equilibrium of hPHGDH variants was evaluated by monitoring the spectral signals related to the protein secondary structure, following the circular dichroism (CD) signal at 220 nm. Compared to the wild‐type enzyme, the signal for the L478A variant was different: a clear signal transition at increasing temperature corresponding to the denaturation step was less evident (data not shown). This result suggests that this hPHGDH variant is present in solution in a partially unfolded conformation. The other variants under study showed a melting temperature value similar to the wild‐type hPHGDH (Table S3).

2.3. Oligomerization state of variants and wild‐type hPHGDH

The oligomerization state of the variants and wild‐type hPHGDH was determined by size‐exclusion chromatography (SEC) at different protein concentrations. For the wild‐type enzyme, the elution volume is affected by the protein concentration: while at 10 mg protein/mL the main elution peak (V _e/V ₀ ratio ~ 1.35) corresponds to a ≥200 kDa tetrameric enzyme, with a minor left shoulder corresponding to an undefined aggregation state (V _e/V ₀ ratio ~1.15, eight monomers), at decreasing concentrations an apparent peak at a higher elution volume is observed (V _e/V ₀ ratio ~ 1.5, corresponding to a dimeric form of the protein), see Figure S3 reporting the deconvolution of the elution profiles. While at 3.33 mg protein/mL the peak is symmetric, pointing out the presence of the protein mainly in the tetrameric form, at 1 and 0.33 mg protein/mL the second flanking peak corresponding to ~140 kDa (dimeric state) is observed, suggesting the presence of two forms (dimer and tetramer) (Figure 2a).

FIGURE 2.

Size‐exclusion chromatography chromatograms and related Western blot analyses of human D‐3‐phosphoglycerate dehydrogenase (hPHGDH) wild‐type and variants. (a) Elution profile of wild‐type hPHGDH at 0.33, 1, 3.33, and 10 mg protein/mL. (b–d) Comparison of elution profile of wild‐type and variants of hPHGDH at a concentration of: (b) 0.1 mg/mL; (c) 1 mg/mL; (d) 10 mg/mL. Black: wild‐type; blue: F418A; green: R454A; orange: L478A; light‐blue: P479A. (e) Western blot analysis of the fractions collected from the separations reported in panel (d), corresponding to the following retention times: (A) 8–10 mL; (B) 10–12 mL; (C) 12–14 mL; (D) 14–16 mL; (E) 16‐18 mL; (F) 18–20 mL; (G) 20–22 mL; (H) 22–24 mL. MW, molecular weight standard proteins; St, hPHGDH standard.

The elution profiles and Western blot analysis of eluted fractions of hPHGDH variants are reported in Figure 2b–e in comparison to the wild‐type counterpart. The molecular mass corresponding to each fraction is reported in Table 2. At 0.1 mg/mL, both the wild‐type and F418A variants show a mixture of dimeric and tetrameric forms, while the R454A, L478A, and P479A elute close to the void volume. At 1 mg/mL, the wild‐type shows a tetrameric structure like other type I PHGDHs, the F418A variant elutes as a mixture of tetrameric and dimeric forms, the P479A and L478A elute both as an aggregate and a dimer, while the R454A only elutes as an aggregated form. At 10 mg/mL the wild‐type hPHGDH shows some signs of aggregation, eluting partially as an octamer while remaining mostly a tetramer (85%), the F418A variant elutes as a mixture of tetrameric and dimeric forms (the low‐intensity peak eluting at around 30 kDa corresponds to a residual contaminant protein, see the Western blot analysis in Figure 2e), the L478A and P479A variants show similar behavior as they elute both as an aggregate with a minimal portion as a dimer, while the R454A variant elutes fully as an aggregate. Native polyacrylamide gel electrophoresis (PAGE) analysis of 10 μg of each hPHGDH variant shows that the wild‐type protein migrates as a single band, while additional bands at lower mobility are apparent for all the other variants (Figure S4). Notably, large part of the L478A and P479A proteins are present as high molecular mass aggregates.

TABLE 2.

Main elution peaks, estimated molecular mass, and oligomerization state of variants and wild‐type human D‐3‐phosphoglycerate dehydrogenase, as determined by size‐exclusion chromatography analysis, at a concentration of 0.1, 1, and 10 mg/mL.

Protein (concentration and variant)	V e/V 0	MW (kDa)	Oligomeric state (number of monomers) a [% of each form]
0.1 mg/mL
Wild‐type b	1.39	178	4
	~1.6	~100	2
F418A b	1.38	153	4
	~1.6	~100	2
R454A	1.01	769	12 (12.9)
L478A	1.05	655	12 (10.9)
P479A	1.00	822	12 (13.7)
1 mg/mL
Wild‐type	1.32	258	4 (4.3)
F418A	1.38	205	4 (3.4) [74%]
	1.57	100	2 (1.7) [26%]
R454A	1.03	773	12 (12.9)
L478A	1.06	694	12 (11.6) [88%]
	1.59	93	2 (1.6) [12%]
P479A	1.01	827	12 (13.8) [70%]
	1.52	123	2 (2.21) [30%]
10 mg/mL
Wild‐type	1.15	467	8 (7.8) [15%]
	1.35	223	4 (3.7) [85%]
F418A	1.38	199	4 (3.3) [74%]
	1.56	98	2 (1.6) [26%]
R454A	1.04	711	12 (11.9)
L478A	1.05	692	12 (11.6) [95%]
	1.59	86	2 (1.45) [5%]
P479A	1.00	828	12 (13.9) [84%]
	1.52	116	2 (1.9) [16%]

Abbreviation: MW, molecular weight standard proteins.

^a Calculated from the calibration curve.

^b Because of the low absorbance signal, the relative amount of the two oligomers cannot be estimated.

To evaluate whether the oligomeric forms of the F418A variant (the only one partially resembling the wild‐type hPHGDH) are in equilibrium with the alternative oligomers, the first and main elution peak at 10 mg/mL (V _e/V ₀ from 0.92 to 1.12, Figure S5A) was separated again by SEC at a concentration of 0.5 and 0.1 mg/mL. A peak at the same elution volume is observed in both runs, suggesting that this oligomeric state is stable (Figure S5B,C). Similarly, the separation of the peaks at V _e/V ₀ 1.12–1.33 and 1.33–1.53 generated again elution peaks at the same elution volumes (Figure S5D,E), indicating that no interconversion of the different oligomeric forms is apparent at different protein concentrations.

In conclusion, the alanine scanning of the selected positions in the C‐terminal end of hPHGDH significantly alters the oligomeric state of the enzyme, frequently generating high molecular mass oligomers; the F418A variant is the one showing the least significant changes, although it generated a stable dimer.

2.4. Protein conformation of the hPHGDH variants

The study of the conformation of hPHGDH F418A, R454A, L478A, and P479A in their apoprotein and holoenzyme form (i.e., in complex with NAD⁺ and the reduced cofactor, NADH) was conducted by CD. Concerning the secondary structure content, the far‐UV CD spectra of the F418A and R454A variants and wild‐type hPHGDH show a similar trend, both in the apoprotein and in the NAD⁺‐ and NADH bound forms. On the other hand, the other two variants show a similar spectrum that differs from the wild‐type in all conditions (Figure 3, left panels). The far‐ultraviolet (UV) spectra were analyzed by the DichroWeb software with the Selecon3 method (Whitmore & Wallace, 2008), which allows the estimation of the secondary structure content. The analysis discloses a similar amount of secondary structure elements in the wild‐type, F418A, and R454A hPHGDH, whereas the L478A and P479A variants differ, with lower α‐helices content and higher β‐sheets than wild‐type hPHGDH (Table S4).

FIGURE 3.

Comparison of the circular dichroism spectra of wild‐type human D‐3‐phosphoglycerate dehydrogenase and the F418A, R454A, L478A, and P479A variants. The far‐UV spectra (left) and near‐UV (right) are reported for the apoprotein (top), NAD⁺‐associated holoenzyme (center), and NADH‐associated holoenzyme (bottom). Black: wild‐type; blue: F418A; green: R454A; red: L478A; light‐blue: P479A.

Concerning the tertiary structure, the near‐UV CD spectrum of the F418A variant resembles the one recorded for the wild‐type hPHGDH in the apoprotein form, with significant alterations in the presence of NAD⁺ or NADH. The spectra for R454A, L478A, and P479A variants are different from the ones for the wild‐type hPHGDH under all conditions (Figure 3, right panel), pointing to a significant alteration in protein conformation induced by the point substitutions.

2.5. Activity of the hPHGDH variants

The kinetic parameters of hPHGDH on 3PG in standard conditions (using an excess of both NAD⁺ and hydrazine monohydrate) were determined by a spectrophotometric assay, that is, by following the increase in the absorbance at 340 nm given by the production of NADH and through the removal of the reaction product PHP by hydrazine monohydrate (see Table 3). Notably, in standard conditions, the k _cat for the wild‐type was around 1.5‐ and 15‐fold higher compared to the F418A and all the other variants under study, respectively. In addition, an inhibition effect for F418A and P479A was apparent at increasing 3PG concentration (Figure S6). Since no significant alterations in the K _M value were apparent, the hPHGDH catalytic efficiency for the wild‐type is on average 2‐ and 20‐fold higher than the F418A and all the other variants, respectively. Notably, the most active hPHGDHs are the only two variants present as a tetramer at the lowest protein concentration in SEC analyses, thus suggesting a correlation between the tetrameric oligomeric state and the enzymatic activity, even for the variants.

TABLE 3.

Kinetic parameters of wild‐type, F418A, R454A, L478A, and P479A variants of human D‐3‐phosphoglycerate dehydrogenase at 37°C in 25 mM HEPES, pH 7.0. The kinetic parameters in the forward direction were determined under standard conditions (with excess of hydrazine monohydrate and NAD⁺); K _i was determined based on Equation (1) for a substrate inhibition effect; values are mean ± SD of three measurements.

Variants	k cat (s−1)	K M (mM)	k cat/K M (mM/s)	K i (mM)
Wild‐type	1.00 ± 0.03	1.15 ± 0.2	0.87	‐
F418A	0.655 ± 0.017	1.41 ± 0.06	0.465	51.4 ± 1.6
R454A	0.050 ± 0.002	1.30 ± 0.19	0.038	‐
L478A	0.067 ± 0.002	1.68 ± 0.24	0.040	‐
P479A	0.089 ± 0.001	1.48 ± 0.15	0.060	60.3 ± 8.6

2.6. Cofactor binding to the hPHGDH variants

The fluorescence spectrum of wild‐type hPHGDH following excitation at 280 nm shows a peak at 330 nm: the protein fluorescence intensity is quenched by adding NAD⁺ and NADH, indicating that cofactor binding induces a conformational change (Murtas et al., 2021). NAD⁺ binding to the wild‐type hPHGDH is a monophasic process with a K _d of 82 μM (Table 4, top), higher than that of the R454A variant. The F418A, L478A, and P479A variants display, on the contrary, a biphasic binding process (Figure S7), with a very low K _d for the first phase (between 1 and 2.4 μM), and a higher K _d for the second one associated with the largest change in fluorescence intensity (between 70% and 89%, see Table 4). The hPHGDH variants show a lower total emission change compared to the wild‐type (especially for the F418A and the L478A variants), suggesting a different exposure of the aromatic residues.

TABLE 4.

Binding of NAD⁺ (top) and NADH (bottom) to wild‐type and variants of hPHGDH, at 15°C and in 10 mM potassium phosphate buffer; K _d values were determined based on Equation (2), and at pH 7.0; values are mean ± SD (n = 2).

Variant	Wild‐type	F418A	R454A	L478A	P479A
NAD+
K d (μM)
First phase	81.8 ± 3.9	1.0 ± 0.1	22.6 ± 1.7	2.37 ± 0.55	2.2 ± 0.4
Second phase		449 ± 26		433.6 ± 61.9	34.0 ± 9.7
Fluorescence intensity change
Total	345	210	310	255	310
First phase (%)	100	12 ± 4	100	30 ± 2	27 ± 6
Second phase (%)		89 ± 4		70 ± 2	73 ± 6
NADH
K d (μM)
First phase	0.33 ± 0.05	1.4 ± 0.3	107.4 ± 9.7	98.4 ± 25.2	10.4 ± 2.0
Second phase	28.8 ± 7.7	79.5 ± 21.7
Fluorescence intensity change
Total	355	260	350	300	310
First phase (%)	28 ± 7	25 ± 5	100	100	100
Second phase (%)	72 ± 7	75 ± 5

NADH binding to the wild‐type hPHGDH is a biphasic process, with a K _d of 0.33 and 28.8 μM for the first and second phase, respectively (Table 4, bottom). The same biphasic process is clearly apparent for the F418A only; for L478A, P479A, and R454A variants, the experimental data are better fit as a monophasic process (Figure S8), whose K _d value corresponds to the one for the second phase of the wild‐type hPHGDH. The change in signal intensity is lower for the F418A and L478A compared to the wild‐type hPHGDH. Altogether, the introduced substitutions alter the cofactor binding to a limited extent.

3. DISCUSSION

This study has been carried out to shed light on the role of the tetramerization on the structural and functional properties of hPHGDH, an essential enzyme for the serinosome and the modulation of L‐Ser synthesis, whose alterations are related to pathological mutations. An experimental structure of the full‐length tetrameric assembly was not possible before. Anyway, AlphaFold allowed us to obtain a 3D model ensemble of the tetrameric hPHGDH, which was validated here by an alanine‐scanning mutagenesis approach. A previous work on pathogenic variants showed how AlphaFold can build a reliable interaction model when only the C‐terminal domain is considered (ipTM score >0.75) (Murtas et al., 2024). The C‐terminal interaction model reported in that work is consistent with the domains' orientation and interactions reported in this study. Here, we used variants at the tetrameric interface to investigate their effects on hPHGDH biochemical and biophysical properties. This is relevant, as an altered hPHGDH activity and/or expression in humans have been related to several pathological conditions, such as serine deficiency disorder, Neu‐Laxova syndrome—the list of related SNPs in hPHGDH is reported in Table 1 of (Murtas et al., 2024)—and various types of cancer. Furthermore, an altered level of hPHGDH has been reported in the hippocampus during Alzheimer's disease progression (Le Douce et al., 2020; Maffioli et al., ²⁰²²; Yan et al., ²⁰²⁰), thus altering the levels of both serine enantiomers.

The wild‐type hPHGDH elutes in the SEC analysis largely as a homotetramer, prone to generating higher oligomeric forms at concentrations ≥10 mg/mL. The observed oligomerization state changes suggest that the F418 residue is involved in the interaction between two hPHGDH dimers; actually, the F418A variant elutes as a tetramer and dimer mixture at all protein concentrations (Table 2). On the other hand, the substitutions of the R454, L478, and P479 residues favor aggregation, which we interpreted as the loss of the wild‐type tetrameric interface mode, in line with the predicted aggregogenic property of the ACT domain without tetramerization (Figure 2b).

CD analysis shows a similarity in secondary structure composition between the wild‐type hPHGDH and the F418A and R454A variants, even when the enzyme is associated with NAD⁺ and, with slight alterations, with NADH. On the other hand, the L478A and P479A variants display a very different pattern of secondary structure elements compared to the wild‐type enzyme in all conditions. Regarding the tertiary structure, there is a minor difference between the F418A variant and the wild‐type hPHGDH even in the presence of NAD⁺, while a most significant variation is apparent in the presence of NADH. The far‐UV CD spectra for the R454A, L478A, and P479A variants deviate greatly from the wild‐type hPHGDH spectrum under all conditions. The thermal stability studies highlight a slight decrease in the T _m value for all the variants, which is more apparent for L478A (showing a >5°C decrease in T _m, Table S3).

From a functional point of view, substitution of the F418 residue leads to a slight alteration in the activity under standard conditions. On the contrary, the substitutions of R454, L478, and P479 compromise the activity, yielding a strong decrease in k _cat (Table 3, top). Notably, the K _M for 3PG was unchanged, suggesting that disallowing tetramer formation and inducing structural alterations with these substitutions at the tetrameric interface do not alter the substrate‐binding site. Contrary to the substrate, the substitutions affect the cofactor binding (altering the phases and/or the K _d, Table 4). By analyzing the tetrameric model of hPHGDH, it is apparent that the ASB domain has a similar flexibility to the first 100 residues of the N‐terminal part of the same subunit (Figure 2b). A principal component analysis of the backbone atoms of the ensemble indicates that nearly 50% of all motions are due to rotations of the C‐terminal part of each subunit, and a consequent shift of the domain forming the first 100 residues at the N‐terminal, which delimits a part of the cofactor binding cleft (Figure S9). Therefore, we propose that any change in the tetramerization at the ACT domain would affect the dynamics of the N‐terminal domain through the ASB domain. Further studies are required to better capture and validate such allosteric dynamics.

Overall, the presented results are in line with the bioinformatic model, which allows to provide a mechanistic hypothesis on the role of different regions at the tetrameric interface (Figure 4). In detail: (i) F418 could allow the predicted dynamic loop in the ASB domain to more efficiently recognize its analog in another dimer, yielding a more stable tetrameric oligomerization state; (ii) the other residues we predicted at the ACT tetrameric interface (R454, L478, P479, and Y495) seem to be essential to maintain a stable tetrameric oligomerization and to avoid the accumulation of the aggregogenic dimer, which results in large structural modifications of regions not involved in substrate or cofactor binding, and in the loss of most of the enzymatic activity.

FIGURE 4.

Mechanistic overview of the role of the designed variants within the context of the predicted human D‐3‐phosphoglycerate dehydrogenase (hPHGDH) tetrameric structure. Wild‐type hPHGDH is a tetramer; each individual chain is encoded by a different color. By substituting single residues predicted at the aspartate kinase‐chorismate mutase‐tyrA prephenate dehydrogenase (ACT) interface (R454, L478, P479, and Y495), highly aggregated states or dimers with a severe activity drop are generated, while by substituting the only bulky residue in the allosteric substrate‐binding (ASB) dynamic loop at the base of the interface (F418), the reduction of the tetramer population and the appearance of dimers is observed, with an overall halving of the enzymatic activity.

Collectively, the presented results indicate that the predicted tetrameric interface is mediated by residues at the ACT domain in a way that is similar to what previously described for the tetrameric structure of MtPHGDH (Dey et al., 2005), and the formation of the tetramer seems to be crucial for proper folding of hPHGDH, which, in turn, is essential for both its stability and functionality.

A previous investigation using both the truncated and full‐length hPHGDH forms and focusing on residues at the N‐terminal region of the protein (such as E108 and I121), concluded that “mutations that can completely disrupt the hPHGDH dimer show different abilities to interrupt the full‐length tetramer” (Xu et al., 2021). For example, the E108A and I121A substitutions fully inactivated the truncated enzyme while not affecting the activity of the full‐length enzyme. Our results highlight the link of correct dimer generation for homotetramer formation, avoiding higher aggregation states.

In conclusion, the residues identified in the bioinformatic model of the full‐length tetrameric hPHGDH play a relevant role in the tetramerization process, which is essential for both its stability and activity. The possible allosteric effects of the tetramerization on the NAD⁺/NADH cofactor binding require further structural and functional studies. Among the modified positions, the ones belonging to the ACT domain most strongly affect the structure–function relationships of hPHGDH. This work is of relevance also to shed light on the pathological mutations identified in hPHGDH and leading to serious neurological disorders (Méneret et al., 2012; Murtas et al., ²⁰²⁴) that in some cases could arise from serinosome alteration due to protein misfolding and aggregation, in addition to enzyme inactivation, thus opening to novel therapeutic approaches.

4. MATERIALS AND METHODS

4.1. Modeling

A homotetrameric model for Homo sapiens hPHGDH (Uniprot ID: O43175) was predicted by two approaches: (a) using AlphaFold‐Multimer v. 2.2.0 (Evans et al., 2021) through the ColabFold interface (https://github.com/sokrypton/ColabFold) (Mirdita et al., 2022) and v2 model, searching for homologous sequences with mmseq2 (Steinegger & Söding, 2017) from UniRef30 and Metagenomic databases, with 6 recycles and using the default for the rest of parameters; (b) assuming that the orientation of the C‐terminal region (corresponding to the sequence 309–533) is the same as found in the tetrameric structure of MtPHGDH (PDB ID: 1ygy), which was used as structural (not sequence) template to align the experimentally solved N‐terminal (PDB ID: 6plg, region 6–308) and the C‐terminal region extracted from the AlphaFold database (https://alphafold.ebi.ac.uk, ID: AF‐O43175‐F1, region 309–533): the two pieces of each subunit were joined by the loop refinement protocol of Modeller 10.2 (Fiser et al., 2008) from Lys310 to A320. Ten alternate conformations were generated using a high optimization level; the loop conformation with the best DOPE loopmodel score (Webb & Sali, 2008) was selected. The same protocol was used to add the first six residues (not present in the experimental structure) and to eliminate clashes at the interface of different subunits (maximum loop length for each refinement was set to 10). The final model was refined, as reported in the following paragraph.

4.2. Conformational analysis and aggregation prone region prediction

A conformational analysis was performed in triplicate, starting from the obtained tetrameric model of hPHGDH. The clustENMD method (Kaynak et al., 2021) of the ProDy package (Bakan et al., 2011) was used. For each replicate, the model was prepared by adding hydrogen atoms at pH 7 with the pdbfixer script (https://github.com/openmm/pdbfixer). The clustENMD sampling was performed by normal modes of an anisotropic network model followed by clustering and a molecular dynamics (MD) refinement, with the following parameters: five generations, using the first three global modes and generating 50 conformers with an average C _α root‐mean square deviation (RMSD) distance of 1 Å from the parent conformer of the previous generation, and increasing at each generation the maximum number of clusters sampled from 20 to 120. A short, classic MD simulation was performed after an energy minimization of each sampled conformer solvated in tip3p water molecules and 150 mM Na⁺ and Cl⁻ ions up to 1 nm distance, by using OpenMM 7.5.1 (Eastman et al., 2017) under the amber ff14SB force field (Maier et al., 2015), with a time step of 2.0 fs, at 30°C and increasing the number of MD steps by 1000 over generations (from 3000 to 8000). The sampled conformations were summarized by a cluster analysis with the gromos method of Gromacs 2019.6 package (Abraham et al., 2015), using backbone atom coordinates and a RMSD threshold of 4 Å. The non‐covalent interactions at tetrameric interfaces were predicted by ProLIF (https://github.com/chemosim-lab/ProLIF) (Bouysset & Fiorucci, 2021), using standard ProLIF definitions of non‐covalent interactions at the 6 Å distance threshold. Only interactions present in at least 50% of the samples were considered.

The ensemble of hPHGDH was analyzed with Aggrescan3D v.1.0.2 (https://bitbucket.org/lcbio/aggrescan3d/src/master/) (Kuriata et al., 2019). The same analysis was repeated on dimers obtained by trimming the two subunits forming the tetramerization interface in the sampled tetrameric ensemble.

4.3. Expression in Escherichia coli and purification of hPHGDH variants

Mutagenesis reactions were performed on the pETM‐His‐hPHGDH expression plasmid using the primers reported in Table S1. The conditions resulting in the highest production yield for the wild‐type hPHGDH were used to optimize the expression of each variant (Murtas et al., 2021), that is, growing transformed Escherichia coli BL21(DE3) cells until an OD_600nm value of 0.6 at 37°C before adding isopropyl β‐D‐1‐thiogalactopyranoside, followed by incubation at 20°C for 20 h. Cells were then harvested by centrifugation (6000 g, 30 min, 4°C) and stored at −20°C.

The His‐tagged hPHGDH variants were purified by HiTrap chelating chromatography (Amersham Biosciences, Amersham, UK) as reported for the wild‐type enzyme (Murtas et al., 2021). The final enzyme preparations were stored at −80°C in 50 mM HEPES, pH 7.0, 5% (v/v) glycerol. The enzyme concentration was determined spectrophotometrically by using the theoretical extinction coefficient at 280 nm (40.45/mM/cm).

4.4. Activity assay and kinetic measurements

The hPHGDH activity in the forward direction was determined by monitoring the time course of NADH production (recording the increase in absorbance at 340 nm) at 37°C in 1 cm quartz cuvette with Jasco V‐750 spectrophotometer (Jasco Co., Cremella, Italy). The assay was performed under standard conditions (0.1 μM hPHGDH [2.39 μg], 2.5 mM 3PG, 1.5 mM NAD⁺, and 200 mM hydrazine monohydrate) in 25 mM 4‐(2‐hydroxyethyl)‐1‐piperazineethanesulfonic acid (HEPES) buffer, pH 7.0. The latter assay was used to determine the apparent kinetic parameters according to a classical Michaelis–Menten equation, using the initial velocity values determined at increasing 3PG concentrations (Pind et al., 2002) or the equation modified to account for substrate inhibition:

$$v_0=k_{\mathrm{cat}}\times \left[E_{\mathrm{t}}\right]\times \left[S\right]/\left(\left[S\right]+K_{\mathrm{M}}+{\left[S\right]}^2/K_{\mathrm{i}}\right)$$ (1)

where K _i is the substrate inhibition constant.

4.5. Determination of the oligomerization state

SEC was carried out using a Superdex 200 Increase 10/300 GL column (Cytiva, Milano, Italy; 1–600 kDa separation range) on an AKTA system. The column was equilibrated in 50 mM HEPES, pH 7.0, 0.15M NaCl, and loaded of 200 μL of hPHGDH variants (at different protein concentration) after spinning the sample at 13000 rpm for 15 min at 4°C. The presence of hPHGDH in the eluted peaks was evaluated by sodium dodecylsulphate‐polyacrylamide gel electrophoresis (SDS–PAGE) and Western blot analysis using a specific anti‐HisTag antibody (diluted 1:200, Santa Cruz Biotechnology Inc., Dallas, TX). The area of each peak was estimated by nonlinear curve‐fitting of the elution profile using PeakFit v4.12 software (Grafiti LLC 405 Waverley St, Palo Alto, CA).

Native PAGE was performed on a 9% acrylamide gel (stacking gel at 3% acrylamide) using a solution made of 3.03 g/L Tris and 14.4 g/L glycine, pH 8.3, as running buffer. The protein samples were mixed with a non‐denaturing sample buffer and loaded on the gel. The electrophoretic run was performed in the cold room at a constant voltage of 200 V. The pI of hPHGDH (6.15) and his net theoretical charge at pH 8.9 (corresponding to −25.4) were calculated using ProtPi (https://www.protpi.ch/).

4.6. Spectral measurements

CD spectra were recorded in 10 mM potassium phosphate, pH 7.0, at 20°C, using a Jasco J‐815 spectropolarimeter (Jasco Co., Cremella, Italy) (Caldinelli et al., 2010). The cell pathlength was 0.1 cm for measurements in the 200‐ to 250‐nm region (0.1 mg protein/mL) and 1 cm for measurements in the 250–350 nm range (0.5 mg protein/mL). Temperature ramp experiments were performed in 10 mM potassium phosphate, pH 7.0, with a temperature gradient of 0.5°C/min (20–100°C range) using a software‐driven, Peltier‐equipped CD spectropolarimeter and recording the signal at 220 nm (Caldinelli et al., 2009). Estimation of the secondary structure content was performed by means of DichroWeb software (Selcon3 method) (Miles et al., 2022; Sreerama & Woody, ²⁰⁰⁰).

The NAD⁺/NADH dissociation constants (K _d values) were estimated by titrating 1 μM (0.6 mg/mL) of each hPHGDH variant with increasing amounts of cofactor and recording the protein fluorescence quenching at 330 nm in 20 mM potassium phosphate, pH 7.0, at 15°C. K _d values were determined by fitting the data to a hyperbolic function (Molla et al., 2000):

$$\mathrm{\Delta }F=\left(\mathrm{\Delta }{F}_{\max }\times \left[\text{ligand}\right]\right)/\left[\text{ligand}\right]\times K_{\mathrm{d}},$$ (2)

where ΔF is the recorded fluorescence intensity change and ΔF _max is the maximal fluorescence intensity change.

AUTHOR CONTRIBUTIONS

DR and VR performed experiments and contributed to the drafting of the manuscript. MO performed the computational structural studies. LP and MO contributed to the discussion of the results and critically revised the manuscript. LP conceived the work and designed the experiments. All authors contributed to the drafting of the manuscript, have read and agreed to the published version of the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

Supporting information

Data S1. Details about hPHGDH variants expression and purification. Primary sequence of hPHGDH (Figure S1). Data about the primers used for the introduction of the mutations (Table S1), the expression in E. coli cells (Table S2) and the purification of F418A, R454A, L478A, and P479A hPHGDH variants (Figure S2). Additional data concerning the biochemical properties of hPHGDH variants: melting temperature (Table S3), secondary structure content (Table S4), SEC analyses (Figures S3 and S5), native PAGE (Figure S4), kinetics (Figure S6), NAD⁺ and NADH binding (Figures S7 and S8), and structural analysis (Figure S9).

Riva D, Orlando M, Rabattoni V, Pollegioni L. On the quaternary structure of human D‐3‐phosphoglycerate dehydrogenase. Protein Science. 2024;33(8):e5089. 10.1002/pro.5089