Structural basis of dynamic P5CS filaments

  1. Jiale Zhong
  2. Chen-Jun Guo
  3. Xian Zhou
  4. Chia-Chun Chang
  5. Boqi Yin
  6. Tianyi Zhang
  7. Huan-Huan Hu
  8. Guang-Ming Lu
  9. Ji-Long Liu  Is a corresponding author
  1. School of Life Science and Technology, ShanghaiTech University, China

Abstract

The bifunctional enzyme Δ1-pyrroline-5-carboxylate synthase (P5CS) is vital to the synthesis of proline and ornithine, playing an essential role in human health and agriculture. Pathogenic mutations in the P5CS gene (ALDH18A1) lead to neurocutaneous syndrome and skin relaxation connective tissue disease in humans, and P5CS deficiency seriously damages the ability to resist adversity in plants. We have recently found that P5CS forms cytoophidia in vivo and filaments in vitro. However, it is difficult to appreciate the function of P5CS filamentation without precise structures. Using cryo-electron microscopy, here we solve the structures of Drosophila full-length P5CS in three states at resolution from 3.1 to 4.3 Å. We observe distinct ligand-binding states and conformational changes for the GK and GPR domains, respectively. Divergent helical filaments are assembled by P5CS tetramers and stabilized by multiple interfaces. Point mutations disturbing those interfaces prevent P5CS filamentation and greatly reduce the enzymatic activity. Our findings reveal that filamentation is crucial for the coordination between the GK and GPR domains, providing a structural basis for the catalytic function of P5CS filaments.

Editor's evaluation

This paper reports the cryo-EM structures of Drosophila P5CS, an enzyme important in amino acid metabolism. This group had previously described P5CS filaments in Drosophila, and here show how the filaments are assembled. The paper describes structural changes that occur upon the binding of substrates and reaction intermediates, making a strong case for a conformational cycle that involves some loop movements. Importantly, the work shows that these movements occur in the context of the assembled filament. Point mutants that block filament assembly have reduced catalytic rates, suggesting that a role of the filament is to increase enzyme activity.

https://doi.org/10.7554/eLife.76107.sa0

Introduction

The bifunctional enzyme Δ1-pyrroline-5-carboxylate synthase (P5CS) is responsible for proline and ornithine metabolism (Baumgartner et al., 2000; Baumgartner et al., 2005; Hu et al., 2008; Pérez-Arellano et al., 2010). In humans, over 30 mutations in P5CS have been identified as the causes of rare diseases (Baumgartner et al., 2000; Baumgartner et al., 2005; Marco-Marín et al., 2020; Pérez-Arellano et al., 2010; Skidmore et al., 2011). In addition, the glutamine-proline regulatory axis has been considered a promising target for cancer therapy (Guo et al., 2020; Liu et al., 2012). In plants, proline synthesis is associated with plant stress resistance (Pérez-Arellano et al., 2010). Therefore, P5CS is of great significance in human health and agriculture.

Previous studies have revealed a characteristic compartmentation of enzymes via filamentation (Hunkeler et al., 2018; Johnson and Kollman, 2020; Liu, 2010; Park and Horton, 2019; Stoddard et al., 2020). This filamentous structure is membraneless and termed the cytoophidium for its appearance (Liu, 2010; Liu, 2016). The cytoophidium has emerged as a mechanism for the regulation of metabolic enzymes (Hansen et al., 2021; Liu, 2016; Zhou et al., 2021). Recently, we have shown that Drosophila P5CS forms cytoophidia in vivo and forms individual filaments in vitro (Zhang et al., 2020).

P5CS corresponds to two individual proteins in prokaryotes and some lower eukaryotes such as yeast. One is the glutamate kinase (GK, proB gene), and the other is γ-glutamyl phosphate reductase (GPR, proA gene) in Escherichia coli. Kinetic analysis suggest that bacterial GK and GPR form a complex (Gamper and Moses, 1974). The dual functions of P5CS in higher eukaryotes implicate that both GK and GPR have evolved into one single protein for coupling reactions. However, no structure of the full-length P5CS has been solved. The underlying mechanisms of the catalytic reaction and the function of filamentation remain unknown.

Using cryo-electron microscopy (cryo-EM), here we solve the structures of full-length P5CS in multiple filamentous states. We reconstruct Drosophila P5CS structures at 3.1–4.3 Å resolutions, providing detailed information of the P5CS filaments bound with different ligands. Our results describe the assembly mechanism of P5CS filaments, in which the GK domain forms tetramer and the GPR domain forms dimer structure, and both domains form specific interaction interfaces. Based on these structures, we propose a working model that filamentation is critical for the coordinated reactions between GK and GPR, the two domains of P5CS.

Results

Overall structures of P5CS filaments

The P5CS molecule contains two domains, GK and GPR, catalyzing the first and second steps in the biosynthesis of proline from glutamate. The GK domain catalyzes glutamate phosphorylation, and the GPR domain catalyzes the NADPH-dependent reduction of γ-glutamyl phosphate (G5P) to glutamate-γ-semialdehyde (GSA). The end product P5C, formed by a spontaneous cyclization reaction of GSA (Figure 1A), will be used by another enzyme P5C reductase (P5CR) to produce proline.

Figure 1 with 6 supplements see all
Bifunctional enzyme properties and cryo-electron microscopy (cryo-EM) analysis of P5CS filaments.

(A) Domain organization of Drosophila melanogaster P5CS, which consists of two domains, N-terminal glutamate kinase (GK) domain and C-terminal γ-glutamyl phosphate reductase (GPR) domain. Putative mitochondrial targeting sequence (MTS) is labeled in gray; the glutamate-binding domain (GBD) and the ATP-binding domain (ABD) of the GK domain are respectively shown in orange and yellow; the NADPH-binding domain (NBD), the catalytic domain (CD), and the oligomerization domain (OD) of the GPR domain are shown in cyan, purple, and pink, respectively. Bifunctional P5CS enzyme catalytic reaction and residue numbers for domain boundaries are shown. (B–D) Single-particle analysis for 3D reconstruction of P5CS filaments, three cryo-EM maps of P5CSGlu filament, P5CSGlu/ATPγS filament, and P5CSMix filament are colored by local resolution estimations. (E) The structures of the P5CS monomer and color codes for P5CS models are indicated. (F) The P5CS dimer. Two monomers (gray or color coded by domain) interact via GPR domain hairpins contact. (G) The P5CS tetramer (sphere representation) is formed via GK domain interaction (cartoon representation) between two P5CS dimers (gray or color coded by domain). (H) The sphere and cartoon representation of P5CS filaments. P5CS filaments are modeled by the cryo-EM map. The rotated view is shown in the right panel; its rise, twist, and width are indicated.

In order to solve the structure of P5CS filaments, we expressed and purified Drosophila melanogaster full-length P5CS proteins. First, we analyzed the APO and substrate-bound states of P5CS by negative staining (Figure 1—figure supplement 1A–D). In our previous study, we found that Drosophila P5CS in the APO state is hard to form filaments at low concentrations (<0.05 μM). The addition of glutamate to the P5CS samples induces micron-scale filaments (Zhang et al., 2020). Here, we observe that increasing P5CS concentration (>1 μM) also promotes the formation of filaments in the APO state. Our results show that the P5CS proteins can be self-assembled into filaments without ligands, and adding all substrates increases the length of filaments at the same concentration of the P5CS proteins. Consistent with our previous study, glutamate (a substrate of P5CS) promotes the formation and maintenance of Drosophila P5CS filaments (Zhang et al., 2020).

Subsequently, samples of the P5CS proteins incubated with different combinations of substrates were prepared for cryo-EM (Figure 1—figure supplement 1E–J). Filaments in three conditions with (1) glutamate (P5CSGlu), (2) glutamate and ATPγS (P5CSGlu/ATPγS), and (3) glutamate, ATP, and NADPH (P5CSMix) were imaged in cryo-EM for single-particle analysis (SPA). Long and flexible filaments of P5CS were observed under all the three conditions. After 3D classification and 3D reconstruction, the electron density maps of the P5CSGlu, P5CSGlu/ATPγS, and P5CSMix filaments reached resolutions of 4.0 Å, 4.2 Å, and 3.6 Å, respectively (Figure 1B–D, Figure 1—figure supplements 24). Using a separate focused refinement strategy, we obtained multiple conformational states of the GK domain tetramer (3.1–3.5 Å) and the GPR domain dimer (3.6–4. 3Å). The cryo-EM data and model refinement statistics are provided in Table 1. The N-terminus (residues 1–44) and three disordered segments in regions I, II, and III in the GK domain were invisible in our maps.

Table 1
Cryo-electron microscopy (cryo-EM) data statistics.
P5CSGlu filamentP5CSGlu/ATPγS filamentP5CSMix filament
Data collection and processing
EM equipmentTitan KriosTitan KriosTitan Krios
DetectorK3 cameraK3 cameraK3 camera
Magnification22,500×22,500×22,500×
Voltage (kV)300300300
Electron exposure (e–/Å2)727272
Defocus range (μm)–0.8 to –2.5–0.8 to –2.5–0.8 to –2.5
Pixel size (Å)0.530.530.53
Symmetry imposedD2D2D2
Number of collected movies4933640810,566
Initial particle images (no.)1,911,8431,563,5538,027,582
Final particle images (no.)432,746327,8411,412,498
Refinement
P5CS tetramerGK domainGPR domainP5CS tetramerGK domainGPR domainP5CS tetramerGK domainGPR domain closed formGPR domain
open form
EMDB IDEMD-31466EMD-31469EMD-32877EMD-31467EMD-32876EMD-32880EMD-31468EMD-32875EMD-32878EMD-32879
PDB code7F5T7F5X7WXF7F5U7WX47WXI7F5V7WX37WXG7WXH
Initial model used (PDB code)-4Q1T2H5G-4Q1T2H5G-4Q1T2H5G2H5G
Map resolution (Å)4.13.53.64.13.44.23.63.14.24.3
FSC threshold0.1430.1430.1430.1430.1430.1430.1430.1430.1430.143
Map resolution range (Å)3.8–8.03.4–5.23.5–5.03.4–8.03.2–4.74.1–5.33.3–7.83.0–4.14.1–5.94.0–5.5
Map sharpening B-factor (Å2)–120–120–120–100–70–200–80–80–150–150
Model composition
Non-hydrogen atoms20,43672446,49420,7447968652220,912817264946590
Protein residues270018968602740104086027601064860430
LigandsGGLGGL---RGP-RGP, ADPNAP-
Ions0000000Mg00
B factors2)
Protein1401501621437412113162123100
Ligand140150---145---121
R.m.s. deviations
Bond lengths (Å)0.0050.0050.0070.0070.0060.0070.0050.0060.0080.005
Bond angles (°)0.6780.540.7770.7860.5760.7880.6750.5540.8640.73
Validation
MolProbity score2.732.532.252.472.512.592.171.852.892.19
Clashscore47.487.9313.1723.568.224.5716.575.4534.612.87
Poor rotamers (%)06.120.2906.390.580.183.561.740.29
Ramachandran plot
Favored (%)89.0491.6787.1587.5992.881.7892.9697.2793.2989.25
Allowed (%)10.517.8912.6212.267.218.226.892.7316.7110.75
Disallowed (%)0.450.440.230.1500.580.15000
  1. GPR: γ-glutamyl phosphate reductase; GK: glutamate kinase; FSC: Fourier shell correlation.

One P5CS monomer can be roughly divided into five subdomains: (1) the glutamate-binding domain (GBD) and (2) the ATP-binding domain (ABD) at the GK domain; (3) the NADPH-binding domain (NBD), (4) the catalytic domain (CD), and (5) the oligomerization domain (OD) at the GPR domain (Figure 1A and E). In the model, two P5CS monomers dimerize through the interaction between their GPR domains, where the β21 at the CD interacts with the β24 at the OD of the other monomer (Figure 1F). This interaction connects two groups of hairpins and maintains the homodimer structure by a hydrogen bond network. Two P5CS dimers further assemble into a compact tetramer through the interaction at the GK domains. The P5CS tetramer serves as the building block of P5CS filaments (Figure 1G).

P5CS filament structures in all the three states showed characteristics of double helix (Figure 1—video 1, Figure 1—video 2). We chose the P5CSMix filament to display the details (Figure 1H). In the helical P5CS filament structure, the GK tetramers serve as the core of the filament, and the GPR dimers form left-handed double helix structures around the central axis. The overall diameter of P5CS filaments in all three states is 180 Å, while the helical twist is 68° and the helical rise is 60 Å (Figure 1H).

Structural comparison of ligand-bound GK domains

The GK domain of Drosophila P5CS is conserved with the GK protein in E. coli. Alignments of sequences and structures indicate that their secondary structures are similar as both exhibit a sandwich-like α3β8α4 topological folding (Figure 2—figure supplement 1A), which is a characteristic of the amino acid kinase (AAK) family (Marco-Marín et al., 2007; Pérez-Arellano et al., 2010; Ramón-Maiques et al., 2002).

We obtained a structure of the GK domain with the binding of glutamate in the P5CSGlu filament (Figure 2A) and a second structure of the GK domain with G5P-Mg-ADP in the P5CSMix filament (Figure 2B, Figure 2—figure supplement 1B). In the P5CSGlu/ATPγS filament, the ligands could not be determined due to incomplete densities (Figure 2—figure supplement 1C). The GK domain structure of the P5CSGlu/ATPγS filament is virtually identical to that of the P5CSMix filament (Figure 2—figure supplement 1D). We speculate that there are two ligand-binding modes (bound with Glu-Mg-ATPγS and G5P-Mg-ADP, respectively) in the P5CSGlu/ATPγS filament. These two modes may coexist in the active sites of the GK tetramer, thereby affecting the 3D reconstruction of the structures. The unexpected presence of G5P could be due to the contamination of ATP in the commercial ATPγS (80% pure) and all substrates were in excess during our sample preparation. Thus, no ligand was modeled in the GK domain structure of the P5CSGlu/ATPγS filament.

Figure 2 with 1 supplement see all
Conformational changes in the glutamate kinase (GK) domain-binding pocket.

(A) GK domain of the P5CSGlu filament, with glutamate shown as sticks with yellow carbons. The dashed lines represent disordered segments (residues 124–142, 211–232, and 275–297) in this model. (B) GK domain of the P5CSMix filament, with G5P, Mg+, and ADP shown as sticks with pink, green, and red carbons, respectively. The dashed lines represent disordered segments (residues 128–140, 214–228, and 282–295) in this model. (C, D) GK domain model surface representation showing the conformation of the binding pocket in the P5CSGlu filament or P5CSMix filament. The cryo-electron microscopy (cryo-EM) density of binding glutamate molecule in (C), and the binding complex of G5P, Mg+, and ADP in (D). The dashed lines represent ‘open loop’ and ‘closed loop’

In the GK domain, a valley-like pocket locates between GBD and ABD, providing the binding sites for glutamate, ATP, or their derivatives (Figure 2C and D). Glutamate binds to the active site of GBD vertically (Figure 2A and C, Figure 2—figure supplement 1E). In contrast, G5P and ADP extend towards each other in the P5CSMix filament, and glutamate at the binding site is converted into the intermediate G5P. At ABD of the P5CSMix filament, the phosphate donor ATP becomes an ADP, associating with an Mg2+ (Figure 2B and D).

Superimposing the GK tetramer in the P5CSGlu filament and that in the P5CSMix filament revealed that the major motion of the GK domain occurred at the region containing flexible loops or disordered segments, whereas the α3β8α4 fold showed minor movement (Figure 3A). Meanwhile, based on the disorder densities in region II (Figure 3—figure supplement 1A–D), we modeled the possible trend of the missing segment with a dashed line (Figure 3A). In the P5CSGlu filament, we speculate that the disordered segment in region II acts as a closed loop, which traps glutamate in GBD (Figures 2C and 3A). In the P5CSMix filament, the same segment shifts away from the top of the binding pocket and forms an open loop, in which residue M213 interacts with G5P (Figure 3A, Figure 3—figure supplement 1E). We notice that the closed loop has a steric clash with G5P, preventing the binding of G5P under such a conformation (Figure 3—figure supplement 1D). Our findings support the idea that region II at the GK domain engages in regulating the catalytic reaction.

Figure 3 with 1 supplement see all
Structural comparison of the two types of glutamate kinase (GK) domain.

(A) Comparison of one protomer of the GK domain tetramer in the P5CSGlu filament (green) and P5CSMix filament (blue-violet) on the right panel. On the left panel, the dashed lines in the model represent the open loop (blue-violet) and closed loop (green) in region II. (B) Superimposition of the GK domain tetramer in the P5CSglu filament (green) with the P5CSMix filament (blue-violet). Transitions from glutamate-bound-conformation to G5P-Mg-ADP-bound conformation are shown as curved arrows, indicating GK domain conformational changes in the P5CS filament.

The helix-helix structure (residues 105–113, 115–124) at region I of the P5CSGlu filament transforms into a helix-loop-helix structure (residues 105–119, 120–122, 123–128) in the P5CSMix filament (Figure 3A, Figure 3—figure supplement 1F). This helix-loop-helix structure is referred to as the ‘hook’ structure. The transformation of the hook structure results in new contact sites between neighbor tetramers in the vertical direction, which is evidenced by a rigid density in our map (Figure 1B–D, Figure 1—figure supplements 2 and 3).

On the other hand, we notice the conformational variation of the loop at region III and a loop (residues 64–77) of GBD shifting greatly by approximately 3 Å away (Figure 3A). The function of these conformational changes is unclear, which may relate to conformational changes of the active site. In order to investigate the conformational changes involved in the catalytic reaction, we further compared the tetramer structures of the GK domain in the P5CSGlu and P5CSMix filaments (Figure 3B). The GK domain of each protomer rotates approximately 10° around its central axis, causing the horizontal compression of the GK domain dimer. By comparing the structures of the GK domain with various ligands, we demonstrate the conformational changes, which may be associated with phosphorylation of the substrate glutamate.

Open and closed conformations of GPR domains

The GPR domain of P5CS belongs to the aldehyde dehydrogenase (ALDH) superfamily. ALDH family uses NAD(P)+ to catalyze the conversion of various aldehydes into their corresponding carboxylic acids. Many studies on ALDHs have shown that a conserved residue cysteine acts as the active site of nucleophile, forming thiohemiacetal intermediate with substrate (Koppaka et al., 2012; Liu et al., 1997; Perozich et al., 1999). Curiously, the NADPH-utilizing GPR domain of P5CS catalyzes the reverse reaction of ALDHs.

On the basis of P5CS structures, we display four different binding modes (Figure 4A–D) of the GPR domain. In the P5CSGlu filament, no ligand binds to the GPR domain (Figure 4A). In the P5CSGlu/ATPγS, however, we observed the density of a G5P at the CD active site (Figure 4B, Figure 4—figure supplement 1A). It might be a contamination of ATP, leading to the production of the substrate G5P. In this model, the binding mode of G5P (referred to as the G5P-binding state) is clearly solved. By focus refinement of the GPR dimer in the P5CSMix filament, we determined two additional states of the GPR domain (Figure 4C and D). One is the NADP(H)-binding state, when NADP(H) is present at NBD (Figure 4—figure supplement 1B). The other is the NADP(H)-released state, of which the cofactor binding site is empty.

Figure 4 with 4 supplements see all
γ-Glutamyl phosphate reductase (GPR) domain ligand-bound mode and its conformation.

(A) The cryo-electron microscopy (cryo-EM) density of the GPR dimer structure and cartoon model is represented as an unliganded state in the P5CSGlu filament (green). (B) GPR dimer structure of the G5P-binding state in the P5CSGlu/ATPγS filament (coral). The conformation of the G5P-binding pocket and G5P (orange) is shown as stick representation. (C) GPR dimer structure of the NADP(H)-binding state in the P5CSMix filament (blue-violet). The conformation of the NADP(H)-binding pocket with NADPH (cyan) is shown as stick representation. (D) GPR dimer structure of the NADP(H)-released state in the P5CSMix filament (yellow). (E) Structural differences in the G5P-binding state (coral) and NADP(H)-binding state (blue-violet) of the GPR domain. Ligands are colored as in (B, C). (F) Superimposition of either the NADPH-binding domain (NBD) or the Rossmann-fold of the GPR domain at the G5P-binding state and NADP(H)-binding state using a single protomer.

Conformational comparison of unliganded, G5P-binding and NADP(H)-released states shows that the overall structures of the GPR domain are similar (Figure 4—figure supplement 2A). The structure of the NADP(H)-release state, which has no bound ligand, is identical to the unliganded state. The binding of G5P to CD of the GPR domain does not lead to obvious conformational changes.

Next, we compared the structures of the G5P-binding state and NADPH-binding state (Figure 4E, Figure 4—figure supplement 2B). We have found that the structures of CD and OD are generally consistent in those two states, while NBD of the GPR domain in those two states differs greatly (Figure 4E). NBD contains consecutive alternating α-helices and β-strands (α252) architecture, which is known as the Rossmann-like fold for dinucleotide binding (Cheek et al., 2005). By superimposing the Rossmann-like fold and the entire NBD, we determined conformational changes between the GPR domain at the G5P-binding state and that at the NADP(H)-binding state (Figure 4F).

Upon NADP(H) binding, the residue R525 interacts with the adenine moiety. This interaction transforms the 525REE527 loop into an ordered structure that extends the α16 helix in the Rossmann-like fold (Figure 4F). Meanwhile, the entire NBD rotates approximately 15° along the cylinder axis (Figure 4—figure supplement 3A) and slides towards CD (Figure 4F; Figure 4—video 1). We hypothesize that the helix, when turns disordered, loses contact with the adenine moiety and then separates the cofactor from NBD via a conformational selection mechanism. A similar phenomenon was also observed in ALDH1L1 (Tsybovsky and Krupenko, 2011). This transformation contributes to bringing the nicotinamide ring of the NADP(H) close to the catalytic residue C598 of CD. Conformational changes triggered by the binding of NADPH subsequently initiate the transfer of the hydride ion from NADPH to the intermediate G5P (Figure 4—figure supplement 3B).

In our models, unliganded, G5P-binding and NADP(H)-released states represent as open conformation, and the NADP(H)-binding state represents as closed conformation. We propose that the P5CS filament accommodates the GPR domain in both open and closed conformations. Therefore, recurring transformations between these two conformations are essential for the catalytic cycle at the GPR domain.

Filamentation regulates the enzymatic reaction

In P5CSGlu, P5CSGlu/ATPγS, and P5CSMix filaments, neighboring GPR dimers interact with each other and form a helical structure. The interaction formed between F642-P644 of the contact loops in adjacent GPR domains, which appears as a CH/Pi interaction (Zondlo, 2013), is critical for the filamentation (Figure 5A and B, Figure 5—figure supplement 1). In P5CSGlu/ATPγS and P5CSMix filaments, the additional interface between hook structures pairs locks adjacent GK tetramers. The hook structure extrudes from the GK domain. In a GK tetramer, four hooks extrude toward two opposite directions to form a ‘spinning top’ arrangement. Therefore, two pairs of hooks in a GK tetramer interact with their counterparts in two adjacent GK tetramers (Figure 5A and C). The hook interaction forms strong contacts via hydrogen bonds (M119-R124, L121-M123) and salt bridges (E116-R124). Therefore, a combination of the GPR contact (for the double helix) and GK contact (for the axis) stabilizes the P5CS filament.

Figure 5 with 5 supplements see all
Assembly and interaction surfaces of the P5CS filament.

(A) P5CS filament assembly interface, the four P5CS protomers in one layer are colored in red, yellow, blue, and green. (B) Interaction between two adjacent γ-glutamyl phosphate reductase (GPR) domain dimers, residues F642 located at loop that interacts with P644 from another neighboring GPR domain dimer. (C) Model for hook structure interaction. (D) Enzyme activity analysis to examine P5CS wild-type or mutant proteins. All of the experiments were replicated three times (n = 3, mean ± SD).

Figure 5—source data 1

Enzymatic activity of wild-type and mutant Drosophila P5CS.

https://cdn.elifesciences.org/articles/76107/elife-76107-fig5-data1-v2.xlsx

To understand the function of P5CS filaments, we generated two mutants, R124A and F642A, which are predicted to abrogate the tetramer-tetramer contact sites of the GK domains and GPR domains, respectively. Negative stain of the mutant P5CS showed that the P5CSF642A mutant proteins did not assemble into a filament with or without ligands (Figure 5—figure supplement 2A). These results indicate that the interaction at the GPR domain interface is crucial for P5CS filamentation.

In contrast, the P5CSR124A mutant proteins formed long filaments in the APO state as well as in the presence glutamate (Figure 5—figure supplement 2B). We observed that glutamate-bound P5CSR124A filaments disassembled at the initial phase of adding ATP. Being incubated with all substrates, P5CSR124A formed shorter filaments than P5CSWT (Figure 5—figure supplement 2B).

We propose that the interactions among the hook pairs are required for stabilizing the filament during the transformation from the P5CSGlu filament to the P5CSGlu/ ATPγS filament or P5CSMix filament. We subsequently analyzed the activity of the wild-type P5CS and two mutants, R124A and F642A. The two mutants exhibited a dramatically compromised activity in comparison with the wild-type P5CS (Figure 5D), suggesting that the integrity of filament is essential to the catalytic reactions.

Discussion

The GK domain

We observed two ligand-binding modes in the GK domain. Due to the lack of ATP-bound structure, it is difficult to determine whether ATP plays a decisive role in these conformational changes. According to a previous report on the N-acetyl-l-glutamate kinase (NAGK), nucleoside is important for the conformational change of the AAK domain, and the structures are similar when bound by ADP or AMPPNP (Gil-Ortiz et al., 2011). Based on the similarity of sequences and structures between GK and NAGK (Marco-Marín et al., 2007), we propose that the conformation of the GK domain in the P5CSGlu filament would transform upon the binding of ATP, thereby triggering the formation of hook structure and completing the catalytic reaction. Although we solved the clear structure of the P5CSGlu filament, further research is needed to understand how the conformation of glutamate binding contributes to the extension of P5CS filaments.

The GPR domain

Aspartate-β-semialdehyde dehydrogenase (ASADH) catalyzes NADPH-dependent reductive dephosphorylation of β-aspartyl phosphate to aspartate-β-semialdehyde (Karsten and Viola, 1991). The GPR domain of P5CS and ASADH catalyzes the same type of reaction. Interestingly, the binding of NADP+ will change the cofactor binding domain of ASADH from open conformation to closed conformation (Hadfield et al., 2001). Thus, we speculate that their catalytic mechanisms have something in common. In the GPR domain of Drosophila P5CS, our data suggest that the catalytic residue C598 of CD attacks the G5P to form the first tetrahedral thioacetal intermediate in the reaction, and then expulsion of phosphate collapses to form a stable thioacyl enzyme intermediate. A hydride is then transferred to this intermediate from NADPH, with subsequent collapse to release the product GSA. Furthermore, the NADPH-binding site is located inside the filament, close to the GK domain. The G5P binding site is close to the external solution environment, which is proposed to facilitate the release of the product GSA/P5C (Figure 5—figure supplement 3, Figure 5—video 1). G5P can freely bind to the GPR domain in our model (G5P-binding state). However, in the closed conformation, when the nicotinamide ring of NADP(H) approaches the G5P-binding site, the substrate tunnel entrance may be blocked by NADP(H). This may affect the subsequent binding of G5P. Therefore, we speculate that the GPR domain should bind with G5P prior to NADPH binding. However, whether this mechanism is a preferred binding order needs to be further verified by kinetic experiments.

The P5CS filament

As mentioned in the ‘Results’ section, we observed that mutated residues R124A and F642A do not directly participate in the active sites, while they are crucial for filamentation. This suggests that the P5CS filamentation couples the reaction catalyzed between the GK domain and GPR domain through transferring unstable intermediate G5P (Pérez-Arellano et al., 2010; Seddon et al., 1989). Considering the distance between the GK and GPR domains is about 60 Å (Figure 5—figure supplement 3, Figure 5—video 1), we propose a model that P5CS filament may exhibit a scaffold architecture that stabilizes the relative position of the GK and GPR domains, the cooperation between which may produce electrostatic substrate channels that mediate the transfer of unstable intermediate G5P. In addition, P5CS filamentation may create a half-opened chamber with the active sites located at the inner part of the filament. Since the GK domain is catalytically faster than the GPR domain, the unstable intermediates G5P accumulate within the filament. This microenvironment may reduce the amount of G5P escaped into the solvent, thereby facilitating the rate-limiting reaction at the GPR domain.

The working model

Together, we propose a coupling catalytic reaction mechanism of Drosophila dynamic P5CS filament. In this proposed model, spontaneous filamentation occurs at the APO state, and elongation of P5CS filament is associated with the binding of glutamate. Upon the binding of glutamate, the binding pocket at the GK domain is bound by ATP; subsequently, conformational changes facilitate the formation of a hook structure and phosphorylation of glutamate, which produces G5P. When products of the GK domain dissociate from the active site, G5P would be trapped by the channel and chamber within the filament and further captured by the GPR domain. Next, NADPH binds to the GPR domain, triggering the conformational change into closed conformation, which brings the NADPH towards the catalytic residue C598 and facilitates the reaction. After this reaction, NADP+ and GSA will be released, and the GPR domain returns to its open conformation (Figure 6). This working model suggests that the GK and GPR domains undergo continuous conformational transition during catalysis, resulting in a dynamic filament.

Model of P5CS filament structural transitions during GSA/P5C synthesis.

The P5CS molecule polymerizes into filaments at the APO state or after binding with the glutamate. Upon ATP binding, the glutamate kinase (GK) domain initiates glutamate phosphorylation. The product leaves the pocket, and the GK domain subsequently repeats reaction cycle (left). Unstable G5P will be transported through channel and the half-open chamber inside the filament, and captured by the γ-glutamyl phosphate reductase (GPR) domain. NADPH binding to the GPR domain transforms the domain to closed conformation, which enables NADPH to approach the catalytic site and completes reductive dephosphorylation of G5P. The GSA/P5C will be released, and the GPR domain returns to the unliganded state with open conformation. The GPR domain then begins the next cycle (right).

In the P5CSGlu filament, the GK and GPR domains are likely in a stable conformational state, while vibration may occur in the GPR domain of P5CSGlu/ATPγS and P5CSMix filaments due to the binding of ligands. This notion could be supported by the differences in their local resolution (Figure 1B-D). We speculate that the swing of GPR in the catalytic reaction could destabilize the interaction between adjacent GPR domain dimers in the filament. Therefore, the extra interaction at the hook structure of the GK domain may be required for the stabilization of the filament. This proposed stabilization is consistent with negative stain data showing that the P5CSR124A mutant cannot stabilize the filament structure in the catalytic process and lose the ability to form the long filaments.

P5CS and human disease

Recently, accumulative evidences have shown that mutations on the human P5CS gene (ALDH18A1) is one of the causes of hyperammonemia, neurocutaneous syndromes, and motor neuron syndrome (Baumgartner et al., 2005; Magini et al., 2019; Marco-Marín et al., 2020; Pérez-Arellano et al., 2010). Such mutations may result in the loss of P5CS function in various degrees. Drosophila P5CS residue R124 in region I of the GK domain, which corresponds to R138 of human P5CS, is highly conserved among different eukaryotes (Figure 5—figure supplement 4). However, 17 residues in region I, including the hook structure, are absent in E. coli GK. Pathogenic mutations of R138 in human P5CS, which are proven to be the cause of autosomal-dominant cutis laxa, have been demonstrated with a decreased activity and a dispersed distribution in mitochondria (Fischer-Zirnsak et al., 2015; Yang et al., 2021). In the protein structure database, there is only the GPR domain structure available for human P5CS (PDB: 2H5G). Its overall structure is similar to the GPR domain of Drosophila P5CS (Figure 4—figure supplement 2B). Although it is still unknown whether human P5CS can form filament structure in vitro, it is reasonable to suspect that the filament-forming property is conserved between human and Drosophila P5CS based on their structural similarity. Our structure reveals that the R138 mutation on human P5CS could abrogate the interaction between hook structures of GK domains, and thereby destabilize the filament and coupling of reactions at the two domains.

In summary, our cryo-EM structures of Drosophila P5CS filament present the assembly mode of P5CS protein and provide a molecular basis for a further understanding of the reaction mechanism of the GK and GPR domains. In our proposed model, the coupling of the GK and GPR domains in the filamentous structure facilitates the catalytic reaction of the bifunctional enzyme P5CS. Additional structural studies of P5CS filaments are required to determine whether there is an underlying regulatory mechanism that transmits information between the GK and GPR domains in the tetramer and along the filament.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Gene (Drosophila melanogaster)P5CSGenBankNM_001259948
Strain, strain background (Escherichia coli)Transetta (DE3)TransGen Biotech
Recombinant DNA reagentpET28a-6His-SUMOIn house
Commercial assay or kitBCA Protein Concentration Determination Kit (Enhanced)BeyotimeP0010
Chemical compound, drugBenzamidine hydrochlorideSigma-Aldrich434760-5G
Chemical compound, drugPepstatin ASigma-AldrichP5318-25MG
Chemical compound, drugLeupeptin hydrochloride microbialSigma/AldrichL9783-100MG
Chemical compound, drugPMSFMDBioP006-5g
Chemical compound, drugNi-NTA AgaroseQIAGEN30250
Chemical compound, drugl-Glutamic acidSigma-AldrichG1251-100G
Chemical compound, drugATPTakara4041
Chemical compound, drugATP-gamma-SAbcamab138911
Chemical compound, drugNADPH tetrasodium saltRoche10107824001
OtherNitinol meshZhenjiang Lehua Electronic TechnologyM024-Au300-R12/13Cryo-EM grid preparation
OtherHoley Carbon FilmQuantifoilR1.2/1.3, 300 Mesh, CuCryo-EM grid preparation
Other400 mesh reinforced carbon support filmEMCNBZ31024aNegative staining
Software, algorithmUCSF Chimerahttps://doi.org/10.1002/jcc.20084https://www.cgl.ucsf.edu/chimera
Software, algorithmUCSF ChimeraXhttps://doi.org/10.1002/pro.3235https://www.cgl.ucsf.edu/chimerax/
Software, algorithmRELIONhttps://doi.org/10.7554/eLife.42166https://relion.readthedocs.io/en/latest/index.html#
Software, algorithmCoothttps://doi.org/10.1107/S0907444910007493https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Software, algorithmPhenixhttps://doi.org/10.1107/S2059798318006551https://phenix-online.org/

P5CS protein purification

Request a detailed protocol

The full-length D. melanogaster P5CS gene was cloned into a modified pET28a vector with a 6 × His SUMO tag fused at the N terminus; the fusion proteins were expressed in E. coli Transetta (DE3) cells overnight at 16°C after induction with 0.1 mM IPTG at OD600 range of 0.6–0.8. The remainder of purification was performed at 4°C. The harvested cells were sonicated under ice and purified by Ni-NTA agarose beads (QIAGEN) in lysis buffer (50 mM Tris-HCl pH 8.0, 500 mM NaCl, 10% glycerol, 20 mM imidazole, 1 mM PMSF, 5 mM β-mercaptoethanol, 5 mM benzamidine, 2 μg/ml leupeptin, and 2 μg/ml pepstatin). After in-column washing with lysis buffer, the proteins were eluted with elution buffer (50 mM Tris-HCl pH 8.0, 500 mM NaCl, 250 mM imidazole, 5 mM β-mercaptoethanol), peak fractions were treated with SUMO protease for 1 hr at 8°C. The P5CS proteins were further purified through HiLoad 16/600 Superdex 200pg gel-filtration chromatography (GE Healthcare) in column buffer (25 mM HEPES pH 7.5 and 100 mM KCl), peak fractions were collected, concentrated, and stored at –80°C before use.

Enzyme activity assays

Request a detailed protocol

The full-length wild-type or mutant P5CS (100 nM protein) activity was determined in the reaction buffer containing 25  mM HEPES pH 7.5, and 10 mM l-glutamate (Sigma), with added 20  mM MgCl2, 10  mM ATP (Takara), and 0.5  mM NADPH (Roche) used to initiate the reaction (Magini et al., 2019; Sabbioni et al., 2021), then the reaction was monitored at 37°C in an MD-SpectraMax i3 plate reader and absorbance at 340 nm was measured every 20 s for 10  min (one experiment, n  =  3). The NADPH concentration was converted from A340 with the standard curve determined at the same experiment.

Negative staining

Request a detailed protocol

Wild-type or mutation P5CS proteins were mixed with different substrate conditions. In brief, the final concentration was as follows: 25 mM HEPES pH 7.5, 100 mM KCl, 10 mM MgSO4, 100 mM l-glutamate, 10 mM ATP, and 0.5 mM NADPH. The prepared protein samples were applied to glow-discharged carbon-coated EM grids (400 mech, EMCN), and stained with 1% uranyl acetate. Negative-stain EM grids were photographed on a Tecnai Spirit G21 microscope (FEI).

Cryo-EM grid preparation and data collection

Request a detailed protocol

For cryo-EM, purified full-length P5CS was diluted to approximately 2 μM and dissolved in buffer containing 25 mM HEPES pH 7.5, 100 mM KCl, 10 mM MgSO4, and incubated with 20 mM l-glutamate for the P5CSGlu filament preparation. The P5CSGlu/ATPγS filament was added with an additional 0.5 mM ATPγS (Abcam) compared to the P5CSGlu filament. For the P5CSMix filament, P5CS proteins (2 μM) were incubated with 100 mM KCl, 10 mM MgSO4, 20 mM l-glutamate, 2 mM ATP, and 0.5 mM NADPH. All the samples were incubated for 1 hr on ice before vitrification. The P5CS filament samples were placed on H2/O2 glow-discharged holey carbon grids (Quantifoil Cu 300 mesh, R1.2/1.3) or amorphous alloy film (CryoMatrix M024-Au300-R12/13). Then, the grids were immediately blotted for 3.0 s and plunge-frozen in liquid ethane cooled by liquid nitrogen using Vitrobot (Thermo Fisher) at 4°C with 100% humidity. Images were collected on Titan Krios G3 (FEI) equipped with a K3 Summit direct electron detector (Gatan), operating in counting super-resolution mode at 300 kV with a total dose of 72 e2, subdivided into 50 frames in 4 s exposure using SerialEM (Mastronarde, 2005). The images were recorded at a nominal magnification of 22,500 × and a calibrated pixel size of 1.06 Å, with defocus ranging from 0.8 to 2.5 μm.

Image processing and 3D reconstruction

Request a detailed protocol

The whole-image analysis was performed with RELION3 (Zivanov et al., 2018). We used MotionCor2 (Zheng et al., 2017) and CTFFIND4 (Rohou and Grigorieff, 2015) via RELION GUI to pr-process the image, movie frames were aligned, and the contrast transfer function (CTF) parameters were estimated in this process. After manual selection, there are 4933 images for the P5CSGlu dataset, 6408 images for the P5CSGlu/ATPγS dataset, and 10,566 images for the P5CSMix dataset left for further processing. For the flexibility of P5CS filaments, SPA was carried out in our reconstructions and no helical symmetry was implied in the whole process. Reference-free particle picking built in RELION3 was performed. This process provides 1,994,786 particles for P5CSGlu, 2,024,372 particles for P5CSGlu/ATPγS, and 8,027,582 particles for P5CSMix. At first, the particles were extracted binning two or three times for the fast 2D classifications. Datasets were cleaned with several rounds of 2D classification and the bin factors were gradually reduced to one at the same time. After 3D classifications with C1 symmetry were applied, several classes were selected to do finer 3D classifications with D2 symmetry. Classes with the intact structure were retained for 3D refinement with D2 symmetry. For the 3D refinement, 432,746, 327,841, and 1,412,498 particles were used for each dataset. The maps including three P5CS tetramer layers were obtained. The relative motion between GK and GPR limited the refinement at a high resolution, so we used the partition reconstruction strategy to improve the resolution for both the GK and GPR domains. For the GK domain, we used continued local refinement to improve the resolution with a mask focus on the middle layer GK. Then, the Ctf-refinement and Bayesian polishing were performed for the remained particles and improved the resolution to 4. 1Å, 4.1 Å, and 3.6 Å for three-layer P5CS maps and 3.5 Å, 3.4 Å, and 3.1 Å for GK maps. For the GPR domain, particles were expended symmetry for the 3D classification without alignment. Several classes with the intact structure were selected and oriented; symmetry collapse was done at the same time. Then, 3D classifications and refinements with C2 symmetry were performed. For the P5CSMix, two different states of GPR were captured. Finally, we got 286,291, 348,804, 193,482, and 233,624 particles to construct maps for the GPR domain with 3.6 Å, 4.2 Å, 4.3 Å, and 4.2 Å resolutions. LocalRes was used to estimate the local resolution of our map.

Model building refinement and validation

Request a detailed protocol

Based on our maps with the near-atomic resolution, the model of the GK and GPR domains was generated with focused refinement maps in different states. The initial model of the GK and GPR domains was generated via swiss model regarding 4Q1T (GK from Burkholderia thailandensis) and 2H5G (human GPR domain) as a reference, respectively. Manual adjustment and building the missing regions were done in Coot (Emsley and Cowtan, 2004). Real space refinements were performed with Phenix (Adams et al., 2011). The full-length P5CS models were linked using the corresponding GK and GPR structures; the linker was generated in the Coot and refined via Phenix. Figures and movies were generated with UCSF Chimera (Pettersen et al., 2004) and ChimeraX (Goddard et al., 2018).

Data availability

Atomic models generated in this study have been deposited at the PDB under the accession codes 7F5T, 7F5U, 7F5V, 7F5X, 7WX3, 7WX4, 7WXF, 7WXG, 7WXH, 7WXI. Cryo-EM maps deposited to EMDB as: EMD-31466, EMD-31467, EMD-31468, EMD-31469, EMD-32875, EMD-32876, EMD-32877, EMD-32878, EMD-32879, EMD-32880. Source Data files have been provided for Figure 5D.

The following data sets were generated
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2021) RCSB Protein Data Bank
    ID 7F5T. Drosophila P5CS filament with glutamate.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2021) RCSB Protein Data Bank
    ID 7F5X. GK domain of Drosophila P5CS filament with glutamate.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) Electron Microscopy Data Bank
    ID EMD-32877. GPR domain of Drosophila P5CS filament with glutamate.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) RCSB Protein Data Bank
    ID 7WXF. GPR domain of Drosophila P5CS filament with glutamate.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2021) RCSB Protein Data Bank
    ID 7F5U. Drosophila P5CS filament with glutamate and ATPγS.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) RCSB Protein Data Bank
    ID 7WX4. GK domain of Drosophila P5CS filament with glutamate and ATPγS.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) RCSB Protein Data Bank
    ID 7WXI. GPR domain of Drosophila P5CS filament with glutamate and ATPγS.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2021) RCSB Protein Data Bank
    ID 7F5V. Drosophila P5CS filament with glutamate, ATP, and NADPH.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) RCSB Protein Data Bank
    ID 7WX3. GK domain of Drosophila P5CS filament with glutamate, ATP, and NADPH.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) RCSB Protein Data Bank
    ID 7WXG. GPR domain closed form of Drosophila P5CS filament with glutamate, ATP, and NADPH.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) RCSB Protein Data Bank
    ID 7WXH. GPR domain open form of Drosophila P5CS filament with glutamate, ATP, and NADPH.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2021) Electron Microscopy Data Bank
    ID EMD-31466. Drosophila P5CS filament with glutamate.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2021) Electron Microscopy Data Bank
    ID EMD-31469. GK domain of Drosophila P5CS filament with glutamate.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2021) Electron Microscopy Data Bank
    ID EMD-31467. Drosophila P5CS filament with glutamate and ATPγS.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) Electron Microscopy Data Bank
    ID EMD-32876. GK domain of Drosophila P5CS filament with glutamate and ATPγS.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) Electron Microscopy Data Bank
    ID EMD-32880. GPR domain of Drosophila P5CS filament with glutamate and ATPγS.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2021) Electron Microscopy Data Bank
    ID EMD-31468. Drosophila P5CS filament with glutamate, ATP, and NADPH.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) Electron Microscopy Data Bank
    ID EMD-32875. GK domain of Drosophila P5CS filament with glutamate, ATP, and NADPH.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) Electron Microscopy Data Bank
    ID EMD-32878. GPR domain closed form of Drosophila P5CS filament with glutamate, ATP, and NADPH.
    1. Zhong J
    2. Guo CJ
    3. Zhou X
    4. Liu JL
    (2022) Electron Microscopy Data Bank
    ID EMD-32879. GPR domain open form of Drosophila P5CS filament with glutamate, ATP, and NADPH.

References

    1. Emsley P
    2. Cowtan K
    (2004) Coot: model-building tools for molecular graphics
    Acta Crystallographica. Section D, Biological Crystallography 60:2126–2132.
    https://doi.org/10.1107/S0907444904019158
    1. Seddon AP
    2. Zhao KY
    3. Meister A
    (1989)
    Activation of glutamate by gamma-glutamate kinase: formation of gamma-cis-cycloglutamyl phosphate, an analog of gamma-glutamyl phosphate
    The Journal of Biological Chemistry 264:11326–11335.

Decision letter

  1. Edward H Egelman
    Reviewing Editor; University of Virginia, United States
  2. José D Faraldo-Gómez
    Senior Editor; National Heart, Lung and Blood Institute, National Institutes of Health, United States

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Structural basis of dynamic P5CS filaments" for consideration by eLife. Your article has been reviewed by 4 peer reviewers, one of wom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by José Faraldo-Gómez as the Senior Editor. All reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

1) The major criticism of all three reviewers was that the paper needs major rewriting both to improve the presentation and to improve its clarity.

2) Significant questions were raised about the need for more mechanistic insight. This may require some additional experiments such as those suggested, and would also require a better discussion in the paper about mechanisms. Further, some claims may need to be tempered in the absence of a structure with bound ATP.

Reviewer #1 (Recommendations for the authors):

The writing would greatly benefit from a professional editor.

The paper uses "spiral" in many places when they really mean "helical". A spiral is defined as: winding in a continuous and gradually widening (or tightening) curve, either around a central point on a flat plane or about an axis so as to form a cone. In contrast, a helix has a fixed radius.

Reviewer #2 (Recommendations for the authors):

1. Most importantly, how does filamentation facilitate the reaction? Is it merely by local concentration effects, a.k.a. "channeling" between active sites via the filament interior? Or is it via concerted or coupled conformational changes within the filament? Is it possible that particular interactions within the filament may be required to stabilize particular conformations of the enzyme required for its activity? If so, then the filament also performs this function in facilitating the enzyme reaction. One could address this by investigating whether or not each active site is fully functional in the absence of filamentation. This could be accomplished using the non-filament forming mutant enzymes, and testing for the presence of the product of the first step separately from that of the second, rather than for only the product of both steps, as done in Figure 5. If the reaction intermediate is too unstable, an active site mutation in the second step active site could be utilized to stall the second reaction step and enable G5P to be measured.

2. None of the structures contain ATP. One contains ADP. Speculation on what ATP does to the structure should be tempered by this fact. Structures of ATP dependent enzymes generally show distinct structures when bound to ATP vs. ADP, although this may or may not be relevant here. Still, the qualification should be made and the language more careful to reflect this fact.

3. Movement of NADPH towards a C598 is mentioned, and C598 is mentioned to be important for enzyme activity. What does C598 do? Where is it located? A figure showing the relationship between C598 and NADPH in the two conformations would be useful, as well as a description of why the close approach of these moieties is important. C598 is shown in Figure 4B, though not discussed in the text pertaining to that figure, and it is close to G5P. Could it merely be involved in G5P binding?

4. Figure 4B, inset – the close proximity of D715 to G5P would appear to not be a favorable interaction. Are these mediated by any cations, or is the pKa of D715 known to be raised perhaps to be involved in the reaction mechanism?

5. The document needs professional language editing throughout to make the text easier to read, although the issues did not impede this reviewer from understanding the main points of the manuscript.

6. Legends of Figures 2 and 3 should explain what the dashed lines represent. From the text, these appear to be disordered residues, but that is not clear in viewing the figure alone.

7. Figure 2: the structure with Glu/ATPgammaS is not shown, yet the text (page 9 line 156) implies it should be. Is it not shown because its conformation is the same as that with G5P-Mg-ADP? Also, the text refers to P5CS^Mix, but the figure shows "G5P-Mg-ADP bound". I gather that these are the same, but this should be made clear in one or both of the places they are discussed (i.e. the text and the figure). The figure legend mentions P5CS^Mix as having G5P-Mg-ADP bound, but the figure itself could also indicate which structure is which using the nomenclature of the text (i.e. P5CS^Glu/ATPgammaS and P5CS^Mix).

8. Page 9, line 165 "L-glutamate is bound in a vertical way". This is not obvious in the figure. Also, it is difficult to see the L-glutamate with the color scheme chosen (also true of the ADP).

9. Page 9, line 170, "the loop shifts away from the top of the binding pocket (Figure 3A)". Which loop? Can it be identified in the figure?

10. Page 9, line 171, M213 is mentioned, but not shown in the figure (sticks of the side chain can be seen if one looks very closely, but the residue should be labeled in the figure if it is mentioned in the text). What is the significance of M213? Why mention it?

11. Page 11, line 207, "G5P-binging", is this "G5P-binding"?

12. Page 12, line 221 and Figure 4F, why is it a "Rossmann-like fold" and not a "Rossman fold"?

13. Page 12, line 228, beta9-beta10 is mentioned in Figure 4F and Figure 4-Video 1, but not identified in the figure or video. Labeling the CD would also be helpful to connect to the text.

14. page 15, line 285. "First, the intermediate G5P is instability and the distance between the two reaction centers is about 60 Å". Does this mean G5P is unstable? How unstable? What is its lifetime under normal enzymatic conditions? A source for this information should also be cited.

15. Figure 5, S2 – it would be helpful to show roughly where the filament axis is in panel B – there is some indication in A, but the map is truncated making it not as obvious to the reader without additional guesswork. This would help to see how the filament may sequester to some degree G5P. Also, a view from the top may be helpful to identify spaces within the filament and where pores are located.

16. Page 16, line 307, "When the products of the GK domain dissociate from the pocket, the G5P is trapped within the filament and further captured by the GPR domain. Next, NADPH binds to the GPR domain, the conformational change brings the NADPH towards the catalytic residue C598, becoming the close conformation and facilitating the reaction". Is there evidence that NADPH does not bind until G5P is released? "close conformation" does this mean "closed conformation" or is it referring to the close proximity of C598? Again, the reader does not know the significance of C598 in how it participates in the reaction and this would be helpful to know here.

17. Page 16, line 319. Please add references to the statement in the first sentence of the paragraph regarding mutations and diseases.

18. Page 19, Enzyme assays. Please provide a reference for the method.

19. This reviewer is not a cryo-electron microscopist, so cannot comment on the quality of the data or its processing. However, is there a means to assess map/model agreement? For example, in crystallography one has the Rfree (in addition to quality indicators of the model geometry and x-ray data scaling, which has similar types of measures here) to determine the quality of the model. In cryoEM, perhaps a method such as cross-correlation and/or number of atoms within the envelope would be useful to assess how well the model and map agree?

20. Page 22, line 446, say what 4Q1T and 2h5g are (and why is one capitalized and one not?). It should be stated how these were used – was a homology model made with the P5CS sequence? How much of the model required changing?

21. Figure 3 and 4 – a color purple is indicated, but the structures in question appear blue.

22. Can the authors speculate on how ATP binding (really ADP) results in the formation of the hook? If not, can this be mentioned?

23. Page 12, Line 232, explain what ALDH family is and why this is relevant (presumably P5CS is a member, but this should be stated).

24. Why is are these new structures not analyzed and compared to the known crystal structure of human P5CS?

Reviewer #3 (Recommendations for the authors):

The weakness of the paper is the presentation in general. The authors should make an effort to improve the clarity of their descriptions and review the text carefully to correct grammatical issues.

This work reports high-quality cryo-EM reconstructions that reveal an impressive supramolecular filament of great beauty. My impression is that the description of the filament is correct, but that the structures offer a lot more of information that should allow to deepen in some crucial aspects such as the catalytic mechanisms, the communication of conformational changes or regulation between domains, which probably will be the subject of future publications.

My major criticism is that the manuscript requires a carefully re-writing to improve clarity and correct many language issues.

Some specific comments follow.

1. When describing the GK domain, the authors could reference the work describing the first G5K structure: Marco-Marín C, Gil-Ortiz F, Pérez-Arellano I, Cervera J, Fita I, Rubio V. A novel two-domain architecture within the amino acid kinase enzyme family revealed by the crystal structure of Escherichia coli glutamate 5-kinase. J Mol Biol. 2007 Apr 13;367(5):1431-46. doi: 10.1016/j.jmb.2007.01.073.

2. The G5K fold is described as "a sandwich-like α3β8α4 structure". Perhaps the authors could mention that this is fold is characteristic of the amino acid kinase family (http://pfam.xfam.org/family/PF00696), as predicted in: Ramón-Maiques S, Marina A, Gil-Ortiz F, Fita I, Rubio V. Structure of acetylglutamate kinase, a key enzyme for arginine biosynthesis and a prototype for the amino acid kinase enzyme family, during catalysis. Structure. 2002 Mar;10(3):329-42. doi: 10.1016/s0969-2126(02)00721-9.

3. When describing the structure of the GPR domain, the authors could mention that there is a crystal structure of the human P5CS GPR domain available in the PDB (entry 2H5G) without an accompanying publication.

4. It is unclear what is the ligand content for each of the reported cryo-EM reconstructions. The filaments grown with glutamate and the non-hydrolyzable ATP analog (ATPgS) showed the product glutamate 5-phosphate in the GPR active site. But what is the content of the GK domain in this structure? The authors state that (Page 9. Line 160) "Three conformations of the GK domain with the different ligands were revealed clearly in our models". However, only two structures (bound to glutamate or in complex with glutamate 5-phosphate, ADP and Mg) are described and represented in the figures.

The authors could consider adding images with the densities for the bound glutamate 5-phosphate, ADP and Mg (as they have already done for the glutamate 5-phosphate and for NADPH)

Also, table 1 uses acronyms (GGL RGP NAP) for the bound ligands, but their meaning is not explained. The table also lists no ions bound in any structure, although the text mentions the presence of Mg2+.

5. (Page 10. Line 191-196) The authors refer to "ATP binding" and this is misleading because none of the structures has ATP bound.

6. They refer to four different states for the GPR domain: (Page 11, lines 199) "… four different binding modes of GPR domain". But the differences between the apo state and the NADP(H)-released state are not clear since both show an empty active site.

7. The work reports the structures of the GPR domain bound to glutamate 5-phosphate or to NADP. Perhaps the authors could combine these structures by superimposing the active sites and provide some additional details about the reaction mechanism. Perhaps a composed figure that shows the active site with the bound NADP and G5P could be illustrative. The description of the proposed catalytic mechanism (page 12, lines 233-235) is not clear.

8. The authors could discuss why in presence of glutamate and ATPgS, the GPR active site has glutamate 5-phosphate bound, whereas in the filament grown with all substrates (glutamate, ATP and NADPH), the GPR active site only shows NADP bound.

9. There is certain inconsistency regarding the conditions for the formation of P5CS filament formation. It is not clear whether filament formation requires glutamate or not. In a previous paper (Zhang 2020), the authors stated that "Purified P5CS in apo state could hardly form filaments" and that "Removing glutamate from solution almost abolished P5CS filament formation". However, the current work describes that "P5CS proteins self-assemble into filaments without the requirement of ligands" but also that "addition of substrates could enhance the length of filaments" and that "In consistent (sic) with our previous study, L-glutamate (…) is critical in promoting the formation and stability of P5CS filament". Perhaps the authors should explain more clearly whether glutamate is needed or not, the required concentration, and provide any additional variables that influence the formation of the filaments in vitro.

10. (Page 15, lines 301-303) "In the proposed model, spontaneous filamentation and elongation of P5CS is importantly associated with the binding of glutamate". To me, it is not clearly described in the manuscript how glutamate favors the formation or enhances the elongation of the P5CS filament. Perhaps the authors should reformulate this sentence or add a more detailed explanation.

11. It is not explained why ATP triggers the depolymerization of the filaments formed by mutant R124A, since according to the results, ATP favors the formation of the hook structures that glue the protein tetramers along the filament. One would expect that the nucleotide enhances rather than destabilizes the mutated filaments.

12. One important conclusion in the manuscript is that "The disruption of P5CS filament may result in uncoupled catalytic reactions of bifunctional P5CS and a reduced activity". The authors measured the complete two-step reaction in the WT and mutated proteins. The decreased activity could mean that in absence of filaments, the product of the GK reaction, is not properly channeled to the GPR domain. However, the decreased overall reaction activity could also be caused by a reduced G5K efficiency. Perhaps the authors could compare the G5K activity of the WT and mutated proteins. If the efficiency of the partial reaction is similar, this would strongly support that defects in filament formation are causing a defect in the channeling of the intermediate metabolite to the GPR domain.

13. Some important references are missing. In addition to those indicated above (points 1 and 2), the authors should include references for RELION, COOT, Phenix, Chimera and for PDB entry 4Q1T cited in the methods.

Reviewer #4 (Recommendations for the authors):

A detailed list of issues and questions follows here.

Figure 1 suppl 1 – scalebars should be included on all micrographs; panel 1B should include all 3 ligand states evaluated in the paper.

Figure 1 suppl 2 – it is hard to know which regions of the filaments are covered by the masks. Please show the mask superimposed on the filament structure.

Page 7 – the discussion of focused classification is a bit confusing. Does "multiple different conformational states" mean multiple conformations within each liganded state, or that each liganded state had a unique conformation. I think it's the later, but if the former then this should be explained and a supplemental figure showing the refinement classification strategy should be presented.

Page 8 – in comparison to existing crystal structures, it would be helpful to know the % identity to the E. coli enzyme, and the RMSD when the two structures are aligned (and, ideally, a supplemental image showing superposition of the structures). Similarly, there appears to be a structure of part of the human enzyme available, pdb id 2h5g, and it would be good to know how similar this structure is to the reported Drosophila structure.

Page 9 – disordered loop in region II – what is meant by "opening" and "closure" in terms of the disordered region. It is difficult to see in Figure 2 A/B, but it looks like the ordered region common to both structures is pretty much the same? If there are differences, this could perhaps be displayed differently.

Page 9 – "with the binding of different ligands" – does this mean that the ligands induce different conformational states, or that the same three conformations were observed regardless of ligand state? – Also, what are the "three conformational states"? This paragraph describes two?

Figure 2 A/B – very hard to assess the similarities and differences between the structures here. At a minimum the same regions should be shown for both structures. But from this it looks like the ordered part of the Glu-bound structure is the same as the "mix" structure, and the disordered loops have just been drawn in different positions. Maybe a superposition of the models would help clarify the differences? – Ah, there is a superposition in Figure 3A. Here it is clear, with the exception of one (or two?) residues if the Glu structure that point up (Figure 3A, left-hand side of the region II loop) the structures are the same. Is there additional evidence (weak density, perhaps?) to support where the disordered loops have been drawn? Showing the quality of the em density in this region is important to judge the conclusions being drawn about the potential movement of this loop.

Page 9, Figure 2 C/D – it is unclear where the "closure loop" is here, can this be highlighted?

Page 9 – How was the conversion of substrate to phosphorylated intermediate (G5P) assessed? Is this based just on the fit to the density, or is there orthogonal evidence (mass spec or something)? It's hard to judge the quality of the fit into the partial map shown in Figure 2D – a supplemental figure with the region around the G5P/ADP showing the quality of the model in this region and demonstrating a better fit of G5P/ADP than Glu/ATP would be helpful. Or a comparison to the Glu/ATPgS structure might be convincing.

Figure 3C – A supplementary figure showing the fit of the atomic model for the "hook" into density in Region 1 would be helpful in assessing the conformational change modeled here.

Page 10, last paragraph – the residues contacting the ligands are all in nearly the same positions in the two structures. Conformational changes described above are distal to the binding site.

Figure 5 supplement 1A – the quality of the negative stain images of F642A is insufficient to assess the protein quality. One cannot discern whether the protein is in a monomeric or tetrameric state, and the apo state micrograph would suggest that aggregation may be a problem for this mutant. Either better stain images or orthogonal data (circular dichroism, melting curve, etc.) are required to be certain that the protein is folded and stable upon introduction of the Phe to Ala mutation.

Figure 5 supplement 1B – While it appears that ATP does limit polymerization of the R124A mutant, a more quantitative measurement would be helpful, especially in interpretation of the enzyme activity data in Figure 5D. I suggest light scattering or ultracentrifugation could be used to quantify the fraction of enzyme in polymers.

Page 14, first paragraph – It is unclear why the R124A mutation would destabilize polymers. The P5CS(Glu) structure (Figure 1B) shows that the polymers are stable in the absence of longitudinal "hook" interactions that are presumably disrupted by this mutation.

Figure 5D – enzyme assays. The methods should indicate what concentration enzyme was used in this assay, and what concentration was imaged in negative stain. A major question is whether filaments are observed for the wildtype protein at whatever concentration was used for the enzyme assay. Presumably the answer is yes, but if this is not shown it would bring into question what causes the effect of the mutations on activity.

These data should be quantified by calculating specific activity of the enzymes under these conditions, which would allow comparison of these data with published values for P5CS.

This assay reads out the second step in a two-step reaction mechanism. If the function of filaments is to couple the two activities as asserted in the text (see comment below), then one would also expect that the point mutants would affect the rate of the first reaction. This should be tested using an assay to monitor ATP hydrolysis in the first step. If the rates of both reactions are reduced by the mutants this would be consistent with a coupling mechanism of the filament, but if only the second step is affected this would be consistent with the authors' hypothesis that the filament increases local intermediate concentration.

Page 15 – The rationale behind "coupling" of catalytic reactions by filament assembly is not clear. As the enzyme appears to undergo a complete ligand binding and catalytic cycle in the context of the filament, it is not clear how the filament contacts are coupling activities. While I realize it is asking a lot to add another cryo-EM structure, it seems that the structure of free P5CS tetramers would be an important piece of data to have in interpreting how filaments might be increasing activity.

Is there evidence to support an internal channel for G5P, as suggested here (at least I assume that is what is meant by "electrostatic channeling")? If there were a relevant channel stabilized by the filament, it should be seen in the wildtype enzyme filament structure.

The proposal that filaments function to create a locally high concentration of intermediates is interesting, but should be tested. One way to do this would be to monitor NADPH production in the presence of G5P as a substrate – at high G5P concentrations one would expect the mutant protein to have the same enzymatic rate as the wildtype.

At the bottom of page 15, in proposing the model for "filament catalysis" the statement that upon dissociation from the active site G5P is "trapped within the filament" is not well supported. The architecture of the filament is such that G5 would appear likely to be able to freely diffuse away from the filament.

Figure 5, supplement 3 – It would be helpful to include subdomain delineation and the locations of regions I, II, and III with the sequence alignment.

Page 16/17 – The potential link between filament assembly and human disease is intriguing. Is there evidence that the human enzyme forms filaments? It would be good to indicate how well-conserved the filament assembly interfaces are between human and Drosophila.

In a similar vein, the introduction mentions the role of P5CS in plants being of potential significance in agriculture. It would be good to indicate how well conserved filament assembly interfaces are in the plant sequences, and whether based on that one would anticipate that they assemble filaments similar to the Drosophila structure reported here.

https://doi.org/10.7554/eLife.76107.sa1

Author response

Reviewer #1 (Recommendations for the authors):

The writing would greatly benefit from a professional editor.

We have rewritten the paper to improve the clarity and the presentation.

The paper uses "spiral" in many places when they really mean "helical". A spiral is defined as: winding in a continuous and gradually widening (or tightening) curve, either around a central point on a flat plane or about an axis so as to form a cone. In contrast, a helix has a fixed radius.

Thanks for the explanation. We replaced “spiral” with “helical”.

Reviewer #2 (Recommendations for the authors):

1. Most importantly, how does filamentation facilitate the reaction? Is it merely by local concentration effects, a.k.a. "channeling" between active sites via the filament interior? Or is it via concerted or coupled conformational changes within the filament? Is it possible that particular interactions within the filament may be required to stabilize particular conformations of the enzyme required for its activity? If so, then the filament also performs this function in facilitating the enzyme reaction. One could address this by investigating whether or not each active site is fully functional in the absence of filamentation. This could be accomplished using the non-filament forming mutant enzymes, and testing for the presence of the product of the first step separately from that of the second, rather than for only the product of both steps, as done in Figure 5. If the reaction intermediate is too unstable, an active site mutation in the second step active site could be utilized to stall the second reaction step and enable G5P to be measured.

By analyzing the arrangement of P5CS tetramers inside the filament, we believe that one role of filamentation is to facilitate the enzyme reaction. Having screened a large number of mutations, we were unable to obtain stable non-filament forming mutants.

2. None of the structures contain ATP. One contains ADP. Speculation on what ATP does to the structure should be tempered by this fact. Structures of ATP dependent enzymes generally show distinct structures when bound to ATP vs. ADP, although this may or may not be relevant here. Still, the qualification should be made and the language more careful to reflect this fact.

We temper our claims in the absence of an ATP-bound structure.

3. Movement of NADPH towards a C598 is mentioned, and C598 is mentioned to be important for enzyme activity. What does C598 do? Where is it located? A figure showing the relationship between C598 and NADPH in the two conformations would be useful, as well as a description of why the close approach of these moieties is important. C598 is shown in Figure 4B, though not discussed in the text pertaining to that figure, and it is close to G5P. Could it merely be involved in G5P binding?

We add text at lines 228-237 and a supplementary figure (Figure 4—figure supplement 3), highlighting the role and location of C598:

“The GPR domain of P5CS belongs to aldehyde dehydrogenase (ALDH) superfamily. ALDHs family uses NAD(P)+ to catalyze the conversion of various aldehydes into their corresponding carboxylic acids. Many studies on ALDHs have shown that a conserved residue cysteine acts as the active site of nucleophile, forming thiohemiacetal intermediate with substrate (Koppaka et al., 2012; Liu et al., 1997; Perozich et al., 1999). Curiously, the NADPH-utilizing GPR domain of P5CS catalyzes the reverse reaction of ALDHs. ”

4. Figure 4B, inset – the close proximity of D715 to G5P would appear to not be a favorable interaction. Are these mediated by any cations, or is the pKa of D715 known to be raised perhaps to be involved in the reaction mechanism?

Residue D715 is highly conserved and close to the active site, and it is not clear what role D715 plays in GPR domain reaction. It might be involved in the reaction mechanism.

5. The document needs professional language editing throughout to make the text easier to read, although the issues did not impede this reviewer from understanding the main points of the manuscript.

We have rewritten the paper to improve the clarity and the presentation.

6. Legends of Figures 2 and 3 should explain what the dashed lines represent. From the text, these appear to be disordered residues, but that is not clear in viewing the figure alone.

We have revised Figure legends of Figures 2 and 3:

“The dashed lines represent disordered segments in this model.”

“On the left panel, the dashed lines in the model represent the open-loop (blue-violet) and closed-loop (green) in region II.”

7. Figure 2: the structure with Glu/ATPgammaS is not shown, yet the text (page 9 line 156) implies it should be. Is it not shown because its conformation is the same as that with G5P-Mg-ADP? Also, the text refers to P5CS^Mix, but the figure shows "G5P-Mg-ADP bound". I gather that these are the same, but this should be made clear in one or both of the places they are discussed (i.e. the text and the figure). The figure legend mentions P5CS^Mix as having G5P-Mg-ADP bound, but the figure itself could also indicate which structure is which using the nomenclature of the text (i.e. P5CS^Glu/ATPgammaS and P5CS^Mix).

We add text at lines 162-169 and a supplementary figure (Figure 2—figure supplement 1C, D) to depict the GK domain structure of the P5CSGlu/ATPγS filament:

“We obtained a structure of the GK domain with the binding of glutamate in the P5CSGlu filament (Figure 2A) and a second structure of the GK domain with G5P-Mg-ADP in the P5CSMix filament (Figure 2B and Figure 2—figure supplement 1B). In the P5CSGlu/ATPγS filament, the ligands could not be determined due to incomplete densities (Figure 2—figure supplement 1C). The GK domain structure of the P5CSGlu/ATPγS filament is virtually identical to that of the P5CSMix filament (Figure 2—figure supplement 1D).”

The ligand-bound states of the GK domain correspond to ligand-bound states of P5CS filament were indicated in Figure 2.

8. Page 9, line 165 "L-glutamate is bound in a vertical way". This is not obvious in the figure. Also, it is difficult to see the L-glutamate with the color scheme chosen (also true of the ADP).

We have added a supplementary figure (Figure 2—figure supplement 1D) with a different angle to highlight the position of glutamate. We have selected an outline font for a clear display.

9. Page 9, line 170, "the loop shifts away from the top of the binding pocket (Figure 3A)". Which loop? Can it be identified in the figure?

We have changed “the loop” into “a disordered segment” and highlighted it in the figure (Figure 3A, Figure 3—figure supplement 1A-D).

10. Page 9, line 171, M213 is mentioned, but not shown in the figure (sticks of the side chain can be seen if one looks very closely, but the residue should be labeled in the figure if it is mentioned in the text). What is the significance of M213? Why mention it?

We added text and a supplementary figure (Figure 3A, Figure 3—figure supplement 1E), highlighting the role and location of M213.

11. Page 11, line 207, "G5P-binging", is this "G5P-binding"?

Yes, it should be “G5P-binding”, we have revised it.

12. Page 12, line 221 and Figure 4F, why is it a "Rossmann-like fold" and not a "Rossman fold"?

The classical Rossmann fold contains six β-strands, whereas Rossmann-like folds contain only five β-strands, just like the GPR domain in figure 4F.

13. Page 12, line 228, beta9-beta10 is mentioned in Figure 4F and Figure 4-Video 1, but not identified in the figure or video. Labeling the CD would also be helpful to connect to the text.

We revised the description about NBD rotation at lines 272-275 and added a supplementary figure (Figure 4—figure supplement 3) to show the cylinder axis:

“Meanwhile, the entire NBD rotates approximately 15° along the cylinder axis (Figure 4—figure supplement 3A) and slides towards CD (Figure 4F; Figure 4-Video 1).”

14. page 15, line 285. "First, the intermediate G5P is instability and the distance between the two reaction centers is about 60 Å". Does this mean G5P is unstable? How unstable? What is its lifetime under normal enzymatic conditions? A source for this information should also be cited.

This means G5P is unstable; relevant references have been added.

15. Figure 5, S2 – it would be helpful to show roughly where the filament axis is in panel B – there is some indication in A, but the map is truncated making it not as obvious to the reader without additional guesswork. This would help to see how the filament may sequester to some degree G5P. Also, a view from the top may be helpful to identify spaces within the filament and where pores are located.

We have added a video (Figure 5-Video 1) and a view from the top (Figure 5—figure supplement 3) for a better representation.

16. Page 16, line 307, "When the products of the GK domain dissociate from the pocket, the G5P is trapped within the filament and further captured by the GPR domain. Next, NADPH binds to the GPR domain, the conformational change brings the NADPH towards the catalytic residue C598, becoming the close conformation and facilitating the reaction". Is there evidence that NADPH does not bind until G5P is released? "close conformation" does this mean "closed conformation" or is it referring to the close proximity of C598? Again, the reader does not know the significance of C598 in how it participates in the reaction and this would be helpful to know here.

This means “closed conformation”. We have added supplementary figures (Figure 4—figure supplement 1, Figure 4—figure supplement 2), highlighting the role and location of C598. We have expanded the discussion of a catalytic model of the GPR domain at lines 359-365:

“In the GPR domain of Drosophila P5CS, our data suggest that the catalytic residue C598 of CD attacks the G5P to form the first tetrahedral thioacetal intermediate in the reaction, and then expulsion of phosphate collapses to form a stable thioacyl enzyme intermediate. A hydride is then transferred to this intermediate from NADPH, with subsequent collapse to release the product GSA.”

17. Page 16, line 319. Please add references to the statement in the first sentence of the paragraph regarding mutations and diseases.

Relevant references have been added.

18. Page 19, Enzyme assays. Please provide a reference for the method.

Relevant references have been added

19. This reviewer is not a cryo-electron microscopist, so cannot comment on the quality of the data or its processing. However, is there a means to assess map/model agreement? For example, in crystallography one has the Rfree (in addition to quality indicators of the model geometry and x-ray data scaling, which has similar types of measures here) to determine the quality of the model. In cryoEM, perhaps a method such as cross-correlation and/or number of atoms within the envelope would be useful to assess how well the model and map agree?

To assess the agreement of map and model, the cross-correlation can be found in the PDB validation reports. As for the validation of the map, it is achieved by the gold standard FSC curve which is in Figure 1—figure supplement 2-4.

20. Page 22, line 446, say what 4Q1T and 2h5g are (and why is one capitalized and one not?). It should be stated how these were used – was a homology model made with the P5CS sequence? How much of the model required changing?

We use Swiss-model to build the initial model; 4Q1T (GK from Burkholderia thailandensis) and 2H5G (human GPR domain) were used as reference models to predict GK and GPR respectively. Then, we manually tuned each residue in Coot, using Phenix to refine and validate the final model.

21. Figure 3 and 4 – a color purple is indicated, but the structures in question appear blue.

Thank you for your suggestion. We have changed “purple” into “blue-violet”.

22. Can the authors speculate on how ATP binding (really ADP) results in the formation of the hook? If not, can this be mentioned?

We have expanded the discussion of nucleotides binding resulting in the formation of the hook structure at lines 335-346:

“We observed two ligand binding modes in the GK domain. Due to the lack of ATP-bound structure, it is difficult to determine whether ATP plays a decisive role in these conformational changes. According to a previous report on the N-Acetyl-L-glutamate kinase (NAGK), nucleoside is important for the conformational change of the AAK domain, and the structures are similar when bound by ADP or AMPPNP (Gil-Ortiz et al., 2011). Based on the similarity of sequences and structures between GK and NAGK (Marco-Marin et al., 2007), we propose that the conformation of the GK domain in the P5CSGlu filament would transform upon the binding of ATP, thereby triggering the formation of hook structure and completing the catalytic reaction.”

23. Page 12, Line 232, explain what ALDH family is and why this is relevant (presumably P5CS is a member, but this should be stated).

Yes, P5CS is a member of the ALDH family. We have added text explaining the ALDH family.

24. Why is are these new structures not analyzed and compared to the known crystal structure of human P5CS?

We have added text and a supplementary figure (Figure 4—figure supplement 2B) to compare the structure of P5CS in Drosophila and that in human.

Reviewer #3 (Recommendations for the authors):

The weakness of the paper is the presentation in general. The authors should make an effort to improve the clarity of their descriptions and review the text carefully to correct grammatical issues.

This work reports high-quality cryo-EM reconstructions that reveal an impressive supramolecular filament of great beauty. My impression is that the description of the filament is correct, but that the structures offer a lot more of information that should allow to deepen in some crucial aspects such as the catalytic mechanisms, the communication of conformational changes or regulation between domains, which probably will be the subject of future publications.

My major criticism is that the manuscript requires a carefully re-writing to improve clarity and correct many language issues.

We have rewritten the paper to improve the clarity and the presentation.

Some specific comments follow.

1. When describing the GK domain, the authors could reference the work describing the first G5K structure: Marco-Marín C, Gil-Ortiz F, Pérez-Arellano I, Cervera J, Fita I, Rubio V. A novel two-domain architecture within the amino acid kinase enzyme family revealed by the crystal structure of Escherichia coli glutamate 5-kinase. J Mol Biol. 2007 Apr 13;367(5):1431-46. doi: 10.1016/j.jmb.2007.01.073.

Relevant references have been added.

2. The G5K fold is described as "a sandwich-like α3β8α4 structure". Perhaps the authors could mention that this is fold is characteristic of the amino acid kinase family (http://pfam.xfam.org/family/PF00696), as predicted in: Ramón-Maiques S, Marina A, Gil-Ortiz F, Fita I, Rubio V. Structure of acetylglutamate kinase, a key enzyme for arginine biosynthesis and a prototype for the amino acid kinase enzyme family, during catalysis. Structure. 2002 Mar;10(3):329-42. doi: 10.1016/s0969-2126(02)00721-9.

We have added text at lines 152-158:

“The GK domain of Drosophila P5CS is conserved with the GK protein in E. coli. Alignments of sequences and structures indicate that their secondary structures are similar as both exhibit a sandwich-like α3β8α4 topological folding (Figure 2—figure supplement 1A), which is a characteristic of the amino acid kinase (AAK) family (Marco-Marin et al., 2007; Perez-Arellano et al., 2010; Ramon-Maiques et al., 2002).”

3. When describing the structure of the GPR domain, the authors could mention that there is a crystal structure of the human P5CS GPR domain available in the PDB (entry 2H5G) without an accompanying publication.

We have added text and a supplementary figure (Figure 4—figure supplement 2B) to compare the structure of P5CS in Drosophila and that in human.

4. It is unclear what is the ligand content for each of the reported cryo-EM reconstructions. The filaments grown with glutamate and the non-hydrolyzable ATP analog (ATPgS) showed the product glutamate 5-phosphate in the GPR active site. But what is the content of the GK domain in this structure? The authors state that (Page 9. Line 160) "Three conformations of the GK domain with the different ligands were revealed clearly in our models". However, only two structures (bound to glutamate or in complex with glutamate 5-phosphate, ADP and Mg) are described and represented in the figures.

The authors could consider adding images with the densities for the bound glutamate 5-phosphate, ADP and Mg (as they have already done for the glutamate 5-phosphate and for NADPH)

Also, table 1 uses acronyms (GGL RGP NAP) for the bound ligands, but their meaning is not explained. The table also lists no ions bound in any structure, although the text mentions the presence of Mg2+.

We add text at “lines 162-169” and a supplementary figure (Figure 2—figure supplement 1C, D) to depict the GK domain structure of the P5CSGlu/ATPγS filament:

“We obtained a structure of the GK domain with the binding of glutamate in the P5CSGlu filament (Figure 2A) and a second structure of the GK domain with G5P-Mg-ADP in the P5CSMix filament (Figure 2B and Figure 2—figure supplement 1B). In the P5CSGlu/ATPγS filament, the ligands could not be determined due to incomplete densities (Figure 2—figure supplement 1C). The GK domain structure of the P5CSGlu/ATPγS filament is virtually identical to that of the P5CSMix filament (Figure 2—figure supplement 1D).”

We added a supplementary figure (Figure 2—figure supplement 1B) to show the cryo-EM densities for the bound G5P-Mg-ADP in the GK domain of P5CSGlu filament.

GGL, RGP, and NAP are component identifiers (3-letter code) which are consist of RCSB PDB, representing the bound ligands in our models. We also updated table 1 and added the Mg2+ ion to the list.

5. (Page 10. Line 191-196) The authors refer to "ATP binding" and this is misleading because none of the structures has ATP bound.

We have removed this paragraph and expanded the discussion of ATP bound at lines 335-346:

“We observed two ligand binding modes in the GK domain. Due to the lack of ATP-bound structure, it is difficult to determine whether ATP plays a decisive role in these conformational changes. According to a previous report on the N-Acetyl-L-glutamate kinase (NAGK), nucleoside is important for the conformational change of the AAK domain, and the structures are similar when bound by ADP or AMPPNP (Gil-Ortiz et al., 2011). Based on the similarity of sequences and structures between GK and NAGK (Marco-Marin et al., 2007), we propose that the conformation of the GK domain in the P5CSGlu filament would transform upon the binding of ATP, thereby triggering the formation of hook structure and completing the catalytic reaction.”

6. They refer to four different states for the GPR domain: (Page 11, lines 199) "… four different binding modes of GPR domain". But the differences between the apo state and the NADP(H)-released state are not clear since both show an empty active site.

We add text at lines 253-255 and a supplementary figure (Figure 4—figure supplement 2) to compare the NADP(H)-released state and the APO state.

“The structure of NADP(H)-release state, which has no bound ligand, is identical to the unliganded state.”

7. The work reports the structures of the GPR domain bound to glutamate 5-phosphate or to NADP. Perhaps the authors could combine these structures by superimposing the active sites and provide some additional details about the reaction mechanism. Perhaps a composed figure that shows the active site with the bound NADP and G5P could be illustrative. The description of the proposed catalytic mechanism (page 12, lines 233-235) is not clear.

We have added a figure (Figure 4—figure supplement 3B), showing the active site with the bound NADP and G5P. The description of the proposed catalytic mechanism has been rewritten in Discussion at lines 359-365:

“In GPR domain of Drosophila P5CS, our data suggest that the catalytic residue C598 of CD attacks the G5P to form the first tetrahedral thioacetal intermediate in the reaction, and then expulsion of phosphate collapses to form a stable thioacyl enzyme intermediate. A hydride is then transferred to this intermediate from NADPH, with subsequent collapse to release the product GSA.”

8. The authors could discuss why in presence of glutamate and ATPgS, the GPR active site has glutamate 5-phosphate bound, whereas in the filament grown with all substrates (glutamate, ATP and NADPH), the GPR active site only shows NADP bound.

We have added text at lines 243-244 and supplementary figures (Figure 2—figure supplement 1B and Figure 4—figure supplement 1) to discuss this result:

“It might be a contamination of ATP, leading to the production of the substrate G5P.”

“Figure 4—figure supplement 1. Representative cryo-electron microscopy (cryo-EM) densities for the active site of the γ-glutamyl phosphate reductase (GPR) domain.

(A) Cryo-EM map quality of G5P ligand in the active site of GPR domain in P5CSGlu/ATPγS filament. (B) Unmodeled densities in the active site of GPR domain at NADP(H)-binding state, which may be the reaction product: π or GSA/P5C.”

9. There is certain inconsistency regarding the conditions for the formation of P5CS filament formation. It is not clear whether filament formation requires glutamate or not. In a previous paper (Zhang 2020), the authors stated that "Purified P5CS in apo state could hardly form filaments" and that "Removing glutamate from solution almost abolished P5CS filament formation". However, the current work describes that "P5CS proteins self-assemble into filaments without the requirement of ligands" but also that "addition of substrates could enhance the length of filaments" and that "In consistent (sic) with our previous study, L-glutamate (…) is critical in promoting the formation and stability of P5CS filament". Perhaps the authors should explain more clearly whether glutamate is needed or not, the required concentration, and provide any additional variables that influence the formation of the filaments in vitro.

We have revised the main text at lines 99-110 in response to the comment:

“In our previous study, we found that Drosophila P5CS in the APO state is hard to form filaments at low concentrations (<0.05 μM). The addition of glutamate to the P5CS samples induces micron-scale filaments (Zhang et al., 2020). Here, we observe that increasing P5CS concentration (>1 μM) also promotes the formation of filaments in the APO state. Our results show that the P5CS proteins can be self-assembled into filaments without ligands, and adding substrates increases the length of filaments at the same concentration of the P5CS proteins. Consistent with our previous study, glutamate (a substrate of P5CS) promotes the formation and maintenance of Drosophila P5CS filaments (Zhang et al., 2020).”

10. (Page 15, lines 301-303) "In the proposed model, spontaneous filamentation and elongation of P5CS is importantly associated with the binding of glutamate". To me, it is not clearly described in the manuscript how glutamate favors the formation or enhances the elongation of the P5CS filament. Perhaps the authors should reformulate this sentence or add a more detailed explanation.

Revised as the following:

“In this proposed model, spontaneous filamentation occurs at APO state, and elongation of P5CS filament is associated with the binding of glutamate.”

We have expanded the discussion at lines 346-349 in response to the comment:

“Although we solved the clear structure of the P5CSGlu filament, further research is needed to understand how the conformation of glutamate binding contributes to the extension of P5CS filaments.”

11. It is not explained why ATP triggers the depolymerization of the filaments formed by mutant R124A, since according to the results, ATP favors the formation of the hook structures that glue the protein tetramers along the filament. One would expect that the nucleotide enhances rather than destabilizes the mutated filaments.

We have added text and expanded the discussion of short P5CSR124A filaments at lines 423-430:

“We speculate that the swing of GPR in the catalytic reaction could destabilize the interaction between adjacent GPR domain dimers in the filament. Therefore, the extra interaction at the hook structure of the GK domain may be required for the stabilization of the filament. This proposed stabilization is consistent with negative stain data showing that the P5CSR124A mutant cannot stabilize the filament structure in the catalytic process and lose the ability to form the long filaments.”

12. One important conclusion in the manuscript is that "The disruption of P5CS filament may result in uncoupled catalytic reactions of bifunctional P5CS and a reduced activity". The authors measured the complete two-step reaction in the WT and mutated proteins. The decreased activity could mean that in absence of filaments, the product of the GK reaction, is not properly channeled to the GPR domain. However, the decreased overall reaction activity could also be caused by a reduced G5K efficiency. Perhaps the authors could compare the G5K activity of the WT and mutated proteins. If the efficiency of the partial reaction is similar, this would strongly support that defects in filament formation are causing a defect in the channeling of the intermediate metabolite to the GPR domain.

We have tempered our claims and added a Discussion section for the potential working model.

13. Some important references are missing. In addition to those indicated above (points 1 and 2), the authors should include references for RELION, COOT, Phenix, Chimera and for PDB entry 4Q1T cited in the methods.

Relevant references have been added.

Reviewer #4 (Recommendations for the authors):

In support of, and addition to, the broad comments included in the public review, a detailed list of issues and questions follows here.

Figure 1 suppl 1 – scalebars should be included on all micrographs; panel 1B should include all 3 ligand states evaluated in the paper.

We have added scale bars to all micrographs, and cryo-EM micrographs of three ligand states are provided (Figure 2—figure supplement 1E-G).

Figure 1 suppl 2 – it is hard to know which regions of the filaments are covered by the masks. Please show the mask superimposed on the filament structure.

We have superimposed masks on corresponding filament structures in our flowchart

Page 7 – the discussion of focused classification is a bit confusing. Does "multiple different conformational states" mean multiple conformations within each liganded state, or that each liganded state had a unique conformation. I think it's the later, but if the former then this should be explained and a supplemental figure showing the refinement classification strategy should be presented.

It is the later. The word “different” has been removed at lines 121-124:

“Using separate focused refinement strategy, we obtained multiple conformational states of the GK domain tetramer (3.1 to 3.5 Å) and the GPR domain dimer (3.6 to 4.3 Å).”

Page 8 – in comparison to existing crystal structures, it would be helpful to know the % identity to the E. coli enzyme, and the RMSD when the two structures are aligned (and, ideally, a supplemental image showing superposition of the structures). Similarly, there appears to be a structure of part of the human enzyme available, pdb id 2h5g, and it would be good to know how similar this structure is to the reported Drosophila structure.

We have added text at lines 825-827 and a supplementary figure (Figure 2—figure supplement 1A) to show the comparison of E. coli GK and the GK domain of Drosophila P5CS:

“(A) The comparison of E. coli GK structure without PUA domain (tan; PDB: 2J5V) and Drosophila GK domain structure of P5CS (blue-violet; this study), with 30.56% sequence identity and RMSD value of 1.363 Å (198 atom pairs).”

We also add text at lines 464-449, 892-894, and a supplementary figure (Figure 4—figure supplement 2B) to compare the structure of P5CS in Drosophila and that in human:

“In the protein structure database, there is only the GPR domain structure available for human P5CS (PDB: 2H5G). Its overall structure is similar to the GPR domain of Drosophila P5CS (Figure 4—figure supplement 2B).”

“Compared with the GPR domain of Drosophila P5CS at NADP(H)-binding state (blue-violet), the GPR domain of human P5CS (cyan, PDB:2H5G) has 56.74% sequence identity and RMSD value of 1.531 Å (398 atom pairs).”

Page 9 – disordered loop in region II – what is meant by "opening" and "closure" in terms of the disordered region. It is difficult to see in Figure 2 A/B, but it looks like the ordered region common to both structures is pretty much the same? If there are differences, this could perhaps be displayed differently.

We have rewritten the description of the disordered region and suggested that they referred to open-loop and closed-loop at lines 190-202, which were shown in Figure 3A and Figure 3—figure supplement 1:

“Meanwhile, based on the disorder densities in region II (Figure 3—figure supplement 1A-D), we modeled the possible trend of the missing segment with a dashed line (Figure 3A). In the P5CSGlu filament, we speculate that the disordered segment in region II acts as a closed-loop, which traps glutamate in GBD (Figure 2C, Figure 3A). In the P5CSMix filament, the same segment shifts away from the top of the binding pocket and forms an open-loop, in which residue M213 interacts with G5P (Figure 3A, Figure 3—figure supplement 1E). We notice that closed-loop has a steric clash with G5P, preventing the binding of G5P under such a conformation (Figure 3—figure supplement 1D). Our findings support the idea that region II at the GK domain engages in regulating the catalytic reaction.”

Page 9 – "with the binding of different ligands" – does this mean that the ligands induce different conformational states, or that the same three conformations were observed regardless of ligand state? – Also, what are the "three conformational states"? This paragraph describes two?

We have rewritten the main text at lines161-168, and added supplementary figures ( figure 2—figure supplement 1C, D) to show three conformational states of the GK domain:

“We obtained a structure of the GK domain with the binding of glutamate in the P5CSGlu filament (Figure 2A) and a second structure of the GK domain with G5P-Mg-ADP in the P5CSMix filament (Figure 2B and Figure 2—figure supplement 1B). In the P5CSGlu/ATPγS filament, the ligands could not be determined due to incomplete densities (Figure 2—figure supplement 1C). The GK domain structure of the P5CSGlu/ATPγS filament is virtually identical to that of the P5CSMix filament (Figure 2—figure supplement 1D).”

Figure 2 A/B – very hard to assess the similarities and differences between the structures here. At a minimum the same regions should be shown for both structures. But from this it looks like the ordered part of the Glu-bound structure is the same as the "mix" structure, and the disordered loops have just been drawn in different positions. Maybe a superposition of the models would help clarify the differences? – Ah, there is a superposition in Figure 3A. Here it is clear, with the exception of one (or two?) residues if the Glu structure that point up (Figure 3A, left-hand side of the region II loop) the structures are the same. Is there additional evidence (weak density, perhaps?) to support where the disordered loops have been drawn? Showing the quality of the em density in this region is important to judge the conclusions being drawn about the potential movement of this loop.

We add text and supplementary figures (Figure 3A, Figure 3—figure supplement 1A-C) to show the quality of the EM density in region II.

Page 9, Figure 2 C/D – it is unclear where the "closure loop" is here, can this be highlighted?

We highlighted the “open-loop” and “closed-loop” in the Figure 2C and D.

Page 9 – How was the conversion of substrate to phosphorylated intermediate (G5P) assessed? Is this based just on the fit to the density, or is there orthogonal evidence (mass spec or something)? It's hard to judge the quality of the fit into the partial map shown in Figure 2D – a supplemental figure with the region around the G5P/ADP showing the quality of the model in this region and demonstrating a better fit of G5P/ADP than Glu/ATP would be helpful. Or a comparison to the Glu/ATPgS structure might be convincing.

We have added a supplementary figure (Figure 2—figure supplement 1B) to show that the binding modes of Glu/ATP and G5P/ADP ligands, indicating that the conformation of bound G5P/ADP is better than Glu/ATP.

Figure 3C – A supplementary figure showing the fit of the atomic model for the "hook" into density in Region 1 would be helpful in assessing the conformational change modeled here.

We have added a supplementary figure (Figure 3—figure supplement 1F) to show the conformational change of hook structure in region I, and cartoon models have overlaid the cryo-EM density of P5CSGlu filament shown as mesh.

And hook structure fitting into density was shown in Figure 1—figure supplement 4C.

Page 10, last paragraph – the residues contacting the ligands are all in nearly the same positions in the two structures. Conformational changes described above are distal to the binding site.

We have rewritten the description of GK domain structural changes at lines 221-224:

“By comparing the structures of GK domain with various ligands, we demonstrate the conformational changes, which may be associated with phosphorylation of the substrate glutamate.”

Figure 5 supplement 1A – the quality of the negative stain images of F642A is insufficient to assess the protein quality. One cannot discern whether the protein is in a monomeric or tetrameric state, and the apo state micrograph would suggest that aggregation may be a problem for this mutant. Either better stain images or orthogonal data (circular dichroism, melting curve, etc.) are required to be certain that the protein is folded and stable upon introduction of the Phe to Ala mutation.

According to a large number of mutant screening and structural analysis, we speculate that Drosophila P5CS cannot form a stable free tetramer, or the tetramer form only exists in the filament. We selected the high concentration protein, which will indeed affect the quality of staining. We found that aggregation is indeed a problem of F642A mutant. In our purification process, this mutant showed very low activity and was easy to inactivate, and the irregular spherical protein and no regular monomer or tetramer were observed by TEM.

Figure 5 supplement 1B – While it appears that ATP does limit polymerization of the R124A mutant, a more quantitative measurement would be helpful, especially in interpretation of the enzyme activity data in Figure 5D. I suggest light scattering or ultracentrifugation could be used to quantify the fraction of enzyme in polymers.

We have rewritten the description of P5CSR124A filaments at lines 316-321:

“In contrast, the P5CSR124A mutant proteins formed long filaments in the APO state as well as in the presence glutamate (Figure 5—figure supplement 2B). We observed that glutamate-bound P5CSR124A filaments disassembled at the initial phase of adding ATP. Being incubated with all substrates, P5CSR124A formed shorter filaments than P5CSWT (Figure 5—figure supplement 2B).”

For negative stain sample preparation, all substrates are excessive. The effect of ATP on the R124A mutant can only be captured in the initial phase of the GK domain reaction. With the extension of incubation time, a short filament will form eventually. Our data show that ATP can affect the R124A mutant, highlighting the importance of the hook structure.

Page 14, first paragraph – It is unclear why the R124A mutation would destabilize polymers. The P5CS(Glu) structure (Figure 1B) shows that the polymers are stable in the absence of longitudinal "hook" interactions that are presumably disrupted by this mutation.

We have added text and expanded the discussion of short P5CSR124A filaments at lines 423-430:

“We speculate that the swing of GPR in the catalytic reaction could destabilize the interaction between adjacent GPR domain dimers in the filament. Therefore, the extra interaction at the hook structure of the GK domain may be required for the stabilization of the filament. This proposed stabilization is consistent with negative stain data showing that the P5CSR124A mutant cannot stabilize the filament structure in the catalytic process and lose the ability to form the long filaments.”

Figure 5D – enzyme assays. The methods should indicate what concentration enzyme was used in this assay, and what concentration was imaged in negative stain. A major question is whether filaments are observed for the wildtype protein at whatever concentration was used for the enzyme assay. Presumably the answer is yes, but if this is not shown it would bring into question what causes the effect of the mutations on activity.

These data should be quantified by calculating specific activity of the enzymes under these conditions, which would allow comparison of these data with published values for P5CS.

This assay reads out the second step in a two-step reaction mechanism. If the function of filaments is to couple the two activities as asserted in the text (see comment below), then one would also expect that the point mutants would affect the rate of the first reaction. This should be tested using an assay to monitor ATP hydrolysis in the first step

We used 100 nM protein in enzyme assays and revised in our Methods, this protein concentration actually can induce the formation of filaments. And all the protein concentrations are marked on negative stain micrographs.

Our aim for enzyme activity assays is to test whether the filament structure is necessary for its reaction. As for the mechanism, more studies are needed. From our results, we observed that the activity of the R124A mutant was greatly reduced. We speculate that it will not have much effect on the rate of the first reaction, because the residue is not close to the active site. If the rate of the first reaction decreases, the destruction of the hook structure may lead to the conformational change of the GK domain tetramer. We have expanded this discussion at lines 423-430:

“We speculate that the swing of GPR in the catalytic reaction could destabilize the interaction between adjacent GPR domain dimers in the filament. Therefore, the extra interaction at the hook structure of the GK domain may be required for the stabilization of the filament. This proposed stabilization is consistent with negative stain data showing that the P5CSR124A mutant cannot stabilize the filament structure in the catalytic process and lose the ability to form the long filaments.”

If the rates of both reactions are reduced by the mutants this would be consistent with a coupling mechanism of the filament, but if only the second step is affected this would be consistent with the authors' hypothesis that the filament increases local intermediate concentration.

Page 15 – The rationale behind "coupling" of catalytic reactions by filament assembly is not clear. As the enzyme appears to undergo a complete ligand binding and catalytic cycle in the context of the filament, it is not clear how the filament contacts are coupling activities. While I realize it is asking a lot to add another cryo-EM structure, it seems that the structure of free P5CS tetramers would be an important piece of data to have in interpreting how filaments might be increasing activity.

Is there evidence to support an internal channel for G5P, as suggested here (at least I assume that is what is meant by "electrostatic channeling"? If there were a relevant channel stabilized by the filament, it should be seen in the wildtype enzyme filament structure.

The proposal that filaments function to create a locally high concentration of intermediates is interesting, but should be tested. One way to do this would be to monitor NADPH production in the presence of G5P as a substrate – at high G5P concentrations one would expect the mutant protein to have the same enzymatic rate as the wildtype.

At the bottom of page 15, in proposing the model for "filament catalysis" the statement that upon dissociation from the active site G5P is "trapped within the filament" is not well supported. The architecture of the filament is such that G5 would appear likely to be able to freely diffuse away from the filament.

Figure 5, supplement 3 – It would be helpful to include subdomain delineation and the locations of regions I, II, and III with the sequence alignment.

We thank the reviewer for this suggestion and agree that is also a problem we want to solve in the future. We have tempered our claims and proposed a model for the discussion of the possible channel or chamber or communication between GK and GPR domains at lines 380-389:

“As mentioned in the ‘Results’ section, we observed that mutated residues R124A and F642A do not directly participate in the active sites, while they are crucial for filamentation. This suggests that the P5CS filamentation couples the reaction catalyzed between the GK domain and GPR domain through transferring unstable intermediate G5P(Pérez-Arellano et al., 2010; Seddon et al., 1989). Considering the distance between the GK and GPR domains is about 60 Å (Figure 5—figure supplement 3, Figure 5-Video 1Figure 5—video 1), we propose a model that P5CS filament may exhibit a scaffold architecture that stabilizes the relative position of the GK and GPR domains, the cooperation between which may produce electrostatic substrate channels that mediate the transfer of unstable intermediate G5P. In addition, P5CS filamentation may create a half-opened chamber with the active sites located at the inner part of the filament. Since the GK domain is catalytically faster than the GPR domain, the unstable intermediates G5P accumulate within the filament. This microenvironment may reduce the amount of G5P escaped into the solvent, thereby facilitating the rate-limiting reaction at the GPR domain.”

In our results, P5CS protein showed high filament-forming ability. The filamentation of P5CS is like a highly ordered factory to facilitate the two-step reaction, which is also an important feature of P5CS filament. Whether dynamic P5CS filaments produce transient conformations to promote the transfer of intermediates, resulting in the formation of the channel or chamber to couple filament reactions, more structural studies of P5CS filaments are required to determine an underlying regulatory mechanism that transmits information between the GK and GPR domain in the tetramer and along the filament.

We have updated this supplementary figure. Subdomain and these regions are labeled now.

Page 16/17 – The potential link between filament assembly and human disease is intriguing. Is there evidence that the human enzyme forms filaments? It would be good to indicate how well-conserved the filament assembly interfaces are between human and Drosophila.

In a similar vein, the introduction mentions the role of P5CS in plants being of potential significance in agriculture. It would be good to indicate how well conserved filament assembly interfaces are in the plant sequences, and whether based on that one would anticipate that they assemble filaments similar to the Drosophila structure reported here.

We have added text at lines 445-452:

“In the protein structure database, there is only the GPR domain structure available for human P5CS (PDB: 2H5G). Its overall structure is similar to the GPR domain of Drosophila P5CS (Figure 4—figure supplement 2B).Although it is still unknown whether human P5CS can form filament structure in vitro, it is reasonable to suspect that the filament-forming property is conserved between human and Drosophila P5CS based on their structural similarity.”

In the species we compared, the sequence of the interfaces is relatively conservative in animals, and less conservative in Arabidopsis. According to the P5CS structure of Arabidopsis predicted by AlphaFold2, its secondary structure on the interfaces are similar to that of Drosophila.

https://doi.org/10.7554/eLife.76107.sa2

Article and author information

Author details

  1. Jiale Zhong

    School of Life Science and Technology, ShanghaiTech University, Shanghai, China
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review and editing
    Contributed equally with
    Chen-Jun Guo and Xian Zhou
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5873-0450
  2. Chen-Jun Guo

    School of Life Science and Technology, ShanghaiTech University, Shanghai, China
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – review and editing
    Contributed equally with
    Jiale Zhong and Xian Zhou
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5342-4761
  3. Xian Zhou

    School of Life Science and Technology, ShanghaiTech University, Shanghai, China
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – review and editing
    Contributed equally with
    Jiale Zhong and Chen-Jun Guo
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0000-2415
  4. Chia-Chun Chang

    School of Life Science and Technology, ShanghaiTech University, Shanghai, China
    Contribution
    Conceptualization, Formal analysis, Investigation, Validation, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7942-6300
  5. Boqi Yin

    School of Life Science and Technology, ShanghaiTech University, Shanghai, China
    Contribution
    Data curation, Formal analysis, Investigation, Methodology, Validation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3974-9820
  6. Tianyi Zhang

    School of Life Science and Technology, ShanghaiTech University, Shanghai, China
    Contribution
    Data curation, Formal analysis, Investigation, Methodology, Validation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4632-6298
  7. Huan-Huan Hu

    School of Life Science and Technology, ShanghaiTech University, Shanghai, China
    Contribution
    Data curation, Formal analysis, Investigation, Validation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5997-530X
  8. Guang-Ming Lu

    School of Life Science and Technology, ShanghaiTech University, Shanghai, China
    Contribution
    Data curation, Formal analysis, Investigation, Validation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6607-2264
  9. Ji-Long Liu

    School of Life Science and Technology, ShanghaiTech University, Shanghai, China
    Contribution
    Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Validation, Writing – original draft, Writing – review and editing
    For correspondence
    liujl3@shanghaitech.edu.cn
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4834-8554

Funding

Ministry of Science and Technology of the People's Republic of China (2021YFA0804701-4)

  • Ji-Long Liu

National Natural Science Foundation of China (31771490)

  • Ji-Long Liu

Shanghai Science and Technology Commission (20JC1410500)

  • Ji-Long Liu

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Zhi-Jie Liu, Suwen Zhao, and Zherong Zhang for their helpful discussions. The EM data were collected at the ShanghaiTech Cryo-EM Imaging Facility. We also thank the Molecular and Cell Biology Core Facility (MCBCF) at the School of Life Science and Technology, ShanghaiTech University, for providing technical support. This work was supported by grants from the Ministry of Science and Technology of China (no. 2021YFA0804701-4), National Natural Science Foundation of China (no. 31771490), and Shanghai Science and Technology Commission (no. 20JC1410500).

Senior Editor

  1. José D Faraldo-Gómez, National Heart, Lung and Blood Institute, National Institutes of Health, United States

Reviewing Editor

  1. Edward H Egelman, University of Virginia, United States

Publication history

  1. Preprint posted: December 2, 2021 (view preprint)
  2. Received: December 3, 2021
  3. Accepted: March 13, 2022
  4. Accepted Manuscript published: March 14, 2022 (version 1)
  5. Version of Record published: March 29, 2022 (version 2)

Copyright

© 2022, Zhong et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 606
    Page views
  • 115
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Jiale Zhong
  2. Chen-Jun Guo
  3. Xian Zhou
  4. Chia-Chun Chang
  5. Boqi Yin
  6. Tianyi Zhang
  7. Huan-Huan Hu
  8. Guang-Ming Lu
  9. Ji-Long Liu
(2022)
Structural basis of dynamic P5CS filaments
eLife 11:e76107.
https://doi.org/10.7554/eLife.76107

Further reading

    1. Structural Biology and Molecular Biophysics
    Maicon Landim-Vieira et al.
    Research Article Updated

    Phosphorylation and acetylation of sarcomeric proteins are important for fine-tuning myocardial contractility. Here, we used bottom-up proteomics and label-free quantification to identify novel post-translational modifications (PTMs) on β-myosin heavy chain (β-MHC) in normal and failing human heart tissues. We report six acetylated lysines and two phosphorylated residues: K34-Ac, K58-Ac, S210-P, K213-Ac, T215-P, K429-Ac, K951-Ac, and K1195-Ac. K951-Ac was significantly reduced in both ischemic and nonischemic failing hearts compared to nondiseased hearts. Molecular dynamics (MD) simulations show that K951-Ac may impact stability of thick filament tail interactions and ultimately myosin head positioning. K58-Ac altered the solvent-exposed SH3 domain surface – known for protein–protein interactions – but did not appreciably change motor domain conformation or dynamics under conditions studied. Together, K213-Ac/T215-P altered loop 1’s structure and dynamics – known to regulate ADP-release, ATPase activity, and sliding velocity. Our study suggests that β-MHC acetylation levels may be influenced more by the PTM location than the type of heart disease since less protected acetylation sites are reduced in both heart failure groups. Additionally, these PTMs have potential to modulate interactions between β-MHC and other regulatory sarcomeric proteins, ADP-release rate of myosin, flexibility of the S2 region, and cardiac myofilament contractility in normal and failing hearts.

    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics
    Atefeh Rafiei et al.
    Research Article Updated

    Doublecortin (DCX) is a microtubule (MT)-associated protein that regulates MT structure and function during neuronal development and mutations in DCX lead to a spectrum of neurological disorders. The structural properties of MT-bound DCX that explain these disorders are incompletely determined. Here, we describe the molecular architecture of the DCX–MT complex through an integrative modeling approach that combines data from X-ray crystallography, cryo-electron microscopy, and a high-fidelity chemical crosslinking method. We demonstrate that DCX interacts with MTs through its N-terminal domain and induces a lattice-dependent self-association involving the C-terminal structured domain and its disordered tail, in a conformation that favors an open, domain-swapped state. The networked state can accommodate multiple different attachment points on the MT lattice, all of which orient the C-terminal tails away from the lattice. As numerous disease mutations cluster in the C-terminus, and regulatory phosphorylations cluster in its tail, our study shows that lattice-driven self-assembly is an important property of DCX.