Structural basis for the helical filament formation of Escherichia coli glutamine synthetase

Abstract

Escherichia coli glutamine synthetase (EcGS) spontaneously forms a dodecamer that catalytically converts glutamate to glutamine. EcGS stacks with other dodecamers to create a filament‐like polymer visible under transmission electron microscopy. Filamentous EcGS is induced by environmental metal ions. We used cryo‐electron microscopy (cryo‐EM) to decipher the structure of metal ion (nickel)‐induced EcGS helical filament at a sub‐3Å resolution. EcGS filament formation involves stacking of native dodecamers by chelating nickel ions to residues His5 and His13 in the first N‐terminal helix (H1). His5 and His13 from paired parallel H1 helices provide salt bridges and hydrogen bonds to tightly stack two dodecamers. One subunit of the EcGS filament hosts two nickel ions, whereas the dodecameric interface and the ATP/Mg‐binding site both host a nickel ion each. We reveal that upon adding glutamate or ATP for catalytic reactions, nickel‐induced EcGS filament reverts to individual dodecamers. Such tunable filament formation is often associated with stress responses. Our results provide detailed structural information on the mechanism underlying reversible and tunable EcGS filament formation.

Short abstract

PDB Code(s): 7W85;

Abbreviations

Cryo‐EM

cryogenic electron microscopy

DLS

dynamic light scattering

EcGS

Escherichia coli glutamine synthetase

GS

glutamine synthetase

1. INTRODUCTION

As one of the 20 building‐block amino acids, glutamine is synthesized in vivo by glutamine synthetases (GS). Glutamine synthesis is conserved from bacteria to eukaryotes. GS are classified into three groups based on their oligomerization. ¹ , ² Type I GS (GSI) are mainly found in prokaryotes and archaea, forming a stable dodecameric structure by stacking two hexagonal rings (see structure in Figure 1a). Type II GS (GSII) that assemble two pentameric rings as a decamer are found in eukaryotes and the obligate anaerobe Bacteroides fragilis. ³ Type III GS (GSIII) have to date only been identified in a few bacteria. ³ , ⁴ Although GSIII form a dual‐ring dodecamer like GSI, they comprise longer primary sequences (~730 amino acids) than either GSI (~460 amino acids) or GSII (~350 amino acids). Dodecameric GSI is 100 Å high and 140 Å wide. Notably, both the GSI hexagonal and GSII pentagonal rings form open pores lacking known catalytic function. The GSI hexagonal pore is 30–40 Å wide, with each subunit extruding a loop (residues 155–188) toward the center of the pore. Nevertheless, the pore is not associated with catalysis (Figure 1a).

FIGURE 1.

GS forms polymeric filaments upon addition of cobalt, zinc, or nickel ions. (a) The dodecameric structure of bacterial glutamine synthetase (GS) comprises two hexagonal rings, with the N‐terminus oriented outward (arrows). GS width, height, and pore size are 140, 98, and 40 Å, respectively. (b) 4 μg purified EcGS (Figure S1) in SDS‐gel reveals a single band around ~55 kDa suggesting a great purity for DLS measurements. (c) Particle sizes of EcGS without or with incubation of six different divalent metal ions, as measured by DLS, with Co²⁺‐, Ni²⁺‐, and Zn²⁺‐induced particles all being considerably larger than particles in the presence of Ca²⁺, Mg²⁺, or Mn²⁺. Hydrodynamic diameter (D_H) measurements are shown in the Table. (d) nsTEM of EcGS after mixing with one of six different metal ions for 1 hr. The TEM images are consistent with our DLS results, with EcGS retaining a dodecameric form in the presence of Mg²⁺, Ca²⁺, or Mn²⁺, but forming extended filaments when mixed with Co²⁺, Ni²⁺ or Zn²⁺. (e) The increasing D_H of EcGS is Ni²⁺ concentration dependent. Then, 2  μM EcGS titrated by a range of Ni²⁺ concentration from 1 μM to 20 mM showing a saturated particle size when [Ni²⁺] is >5 mM. Data points with D _H values greater than 30 nm are indicated

The catalytic mechanisms of GS have been revealed comprehensively based on a number of GS crystal structures from Salmonella typhimurium (StGS) (5), Bacillus subtilis (BsGS), ⁶ Mycobacterium tuberculosis (MtGS), ⁷ and Helicobacter pylori (HpGS). ⁸ Glutamate is first recruited to GS and coupled with an ATP to form an intermediate 𝛾‐glutamyl phosphate product. In the next step, ammonia is used to replace the phosphate group with nitrogen, thereby generating glutamine, representing a critical ammonia‐mediated metabolic reaction in cells. GS, glutamate synthase, and glutamate dehydrogenase are enzymes all associated with ammonia assimilation. ⁹ , ¹⁰

It has been reported previously that E. coli GS (EcGS) forms “strands” or “bundles” upon addition of cobalt ions, ¹¹ but the mechanism underlying the formation of filament‐like polymeric GS is not well understood. Another study used electron paramagnetic resonance ¹² to reveal that copper ions (Cu²⁺) can induce “stacking” of polymeric GS, with residues His5 and His13 in the first helix being proposed as potential sites coordinating two metal ions in the polymeric GS conformation. Nevertheless, details of the mechanism by which metals induce GS polymerization are lacking. Accordingly, a high‐resolution structure of polymeric GS is required to depict the atomic coordination of residue sidechains and metal ions. Many metabolic enzymes form active homo‐oligomers, and high‐order oligomerization can convert homo‐oligomers into long filamentous chains. Examples of such filamentous enzymes include glutaminase, ¹³ glutamate dehydrogenase, ¹⁴ and human CTP synthase. ¹⁵ The yeast glutamine synthetase Gln1 forms polymeric filaments in vivo under conditions of starvation or low cytosolic pH. ¹⁶ Filament formation can be reversed upon glucose provision, suggesting that yeast Gln1 filament may be responsive to stress.

In this study, we have resolved the structure of helical filaments of EcGS using cryo‐EM. Our structural analysis reveals interactions between metal ions and GS residues in the polymer. We also report dynamic light scattering DLS‐based measurements of diverse GS particle sizes under varied conditions of association with metal ions, chelators, or substrates, as well as for GS mutants. Our study demonstrates the mechanism of EcGS filament formation upon addition of divalent metal ions.

2. RESULTS

2.1. EcGS is a stable dodecamer

To explore if metal ions induce GS polymerization, we tested six divalent metal ions, that is, magnesium, calcium, manganese, cobalt, nickel, and zinc. GS polymerization results in substantial changes in particle size, so we adopted DLS to determine hydrodynamic diameters (D_H) of purified EcGS (Figures 1b and S1) in the presence of the six divalent metal ion types. Mg²⁺, Ca²⁺, and Mn²⁺ did not exert any effect on the D_H of EcGS (Figure 1c), with average particle diameters of ~18 nm, similar to the metal ion‐free conditions and to the structure shown in Figure 1a. In contrast, Co²⁺, Ni²⁺, and Zn²⁺ ions substantially increased EcGS particle diameters to ~1,500–5,500 nm upon incubation with these metal ions for 1 hr. This >60‐fold enhancement in particle size indicates that EcGS forms an aggregate in the presence of Co²⁺, Ni²⁺, and Zn²⁺ ions. However, the propensity for EcGS to form high‐order polymer varies according to the metal ion, with Zn²⁺ inducing much larger particle sizes than either Ni²⁺ or Co²⁺.

Addition of metal ions resulted in significantly greater turbidity of GS in solution, ¹¹ , ¹² with the insoluble GS potentially representing amorphous aggregates or ordered polymers. We used negative‐stain transmission electron microscopy (nsTEM) to examine the appearance of EcGS aggregates formed upon mixing with different metal ions. Consistent with our DLS data, GS mixed with Mg²⁺, Ca²⁺, or Mn²⁺ retained the native dodecamer conformation, whereas Ni²⁺, Co²⁺, and Zn²⁺ all resulted in higher‐order structures (Figure 1d). In these latter, the EcGS stack on top of each other, growing along with a single dimension to generate several long chains.

Next, we inspected the Ni²⁺‐induced structures shown in Figure 1d more closely, as they presented the best resolution. A nickel ion concentration‐dependent titration experiment revealed that EcGS formed polymeric filaments when Ni²⁺ is greater than 5 mM (Figure 1e). The saturation condition is then used for followed nsTEM and cryo‐EM studies. Under nsTEM, each subunit of the Ni²⁺‐induced filamentous polymer is similar to the native dodecamer (Figure 2a). The width and length of the EcGS polymer subunits shown in Figure 2a are ~15 and 10 nm, respectively, implying that cylindrical polymerization is simply the result of numerous consecutive stacking reactions, with Co²⁺ and Zn²⁺ ions likely exerting a similar effect to Ni²⁺. However, based on our nsTEM images, we could not resolve whether EcGS dodecamers stack as parallel or helical filaments. Given the hexagonal symmetry of EcGS and surface of dodecameric conformation, it is likely that the 12 N‐terminal helices (named helix 1 or H1) from the dodecameric subunits are crucial to polymeric extension (Figure 2a). The H1 helixes of bacterial GS are often comprised of a few negatively charged aspartate and glutamate residues, as well as histidines (Figure 2b). These Asp, Glu, and His residues may serve as hinge points that coordinate with metal ions. For example, residues His5 and His13 of EcGS have been linked to copper‐induced stacking reactions. ¹² Amino acid sequence alignment of bacterial GS sequences reveal that His5‐His13 of EcGS and Asp5‐Glu13 of MtGS are feasible sites for chelating nickel ions to induce filamentous formation. We generated H5A, H13A, and H5A‐H13A mutant variants of EcGS and purified them to validate the roles of the His5 and His13 residues in filamentation. Even though polymer formation was not completely abolished in the His‐to‐Ala single mutants (Figure 2c), polymer length still appeared greatly reduced, with only 3–6 stacked dodecameric units seen in the nsTEM images. EcGS polymerization was no longer observed for the H5A‐H13A dual mutant in the presence of Ni²⁺. When we mutated these histidine residues in EcGS to negatively charged Asp (H5D) and Glu (H13E) to mimic helix H1 of MtGSand BsGS, remarkably, GS polymerization was retained (Figure 2c). Thus, like EcGS, the GS of B. subtilis and M. tuberculosis likely form long filaments in the presence of certain metal ions.

FIGURE 2.

The N‐terminal helix H1 responses to filament formation. (a) Magnification of polymeric EcGS filament, showing a few repeated dodecamers. The N‐terminus is exposed to the solution, so the N‐terminal helix H1 was postulated as being responsible for stacking. (b) Alignment of the ~20 N‐terminal residues of selected bacterial GS reveals two conserved sites highlighted in cyan that are associated with dodecameric stacking via nickel ions. (c) Single H5A and H13A mutations of EcGS indicate that both those histidine residues are essential to enable GS grow as a long filament. Some short filaments (3–7 dodecamers) are boxed and magnified. A dual H5A–H13A mutation completely abolished filament formation. An H5D–H13E mutation mimicking GS of Bacillus subtilis and Mycobacterium tuberculosis results in pronounced filament bundling

2.2. Cryo‐EM structure of EcGS filamentous polymer

We examined the long EcGS filaments induced by nickel ions in further detail using cryo‐EM. The representative motion‐corrected micrograph shown in Figure 3a reveals several long and straight GS filaments. We collected a total of 1,182 micrographs to perform 3D map reconstruction, selecting 400‐pixel segments (box size = 328.8 Å) to cover three times the height of dodecameric EcGS for 2D classification. In Figure 3b, we present six well‐aligned 2D classes reflecting different rotational views of GS filaments, with the hinges between two dodecamers and flexible loops highlighted by yellow and cyan arrows, respectively. The three dodecamers in a single 2D class are rotated, indicative of helical filaments. Therefore, we adopted cryoSPARC 3.2 ¹⁷ to obtain a 3.48 Å cryo‐EM map of EcGS filaments, with a helical rise of 97.35 Å and a helical twist of −18.15° (Figures 3c and S2). Single EcGS dodecamers in our map match well with known crystal structures of GS ⁵ , ⁶ , ⁷ , ⁸ implying that filamentous GS comprises rotated repeats of GS dodecamers without needing profound conformational changes. In our 3D structure (Figure 3d), all 12 individual subunits of dodecameric EcGS are sharply defined, apart from certain extrusion loops in the hollow pore. Each dodecamer adopts a symmetric hexagonal 2‐layer ring, with the rings being attached and rotated tail‐to‐tail. Many of these EcGS dodecamers are then stacked face‐to‐face to form cylindrical filaments.

FIGURE 3.

Cryo‐EM structure of EcGS filament. A cryo‐EM micrograph depicts long and unbranched EcGS filaments (a), and 2D class averages further illustrate rotation of two stacked dodecamers that imply EcGS filament helicity (b). Hinges between two dodecamers formed by the N‐terminal H1 helices are indicated by yellow arrows. Cyan arrows point to the flexible loops of GS monomer subunits. (c) Cryo‐EM map of EcGS filament. The map was refined according to helical reconstruction protocols in cryoSPARC 3.2. ¹⁷ One subunit monomer of GS has been colored cyan. (d) A single dodecamer from panel (C) was extracted to illustrate the 12 subunits and their structures. Yellow spheres on the surface of the hexagonal ring represent nickel ions. The dashed lines indicate unassigned residues 155–188, for which a cryo‐EM density could not be generated. (e) Stacking of two GS dodecamers. The green and yellow subunits form a two‐layer ring in one dodecamer. The interface between the two rings shown in the schematic highlights the interactions between the two H1 helixes. A nickel ion (gold) is indicated. (f) A simulated EcGS filament according to determined helical rise and twist values. The height of the dodecamer is ~98 Å and 20 repeats result in a 1,960 Å (196 nm) filament. The two‐layer ring in (e) is shown to illustrate the helical rotation from the first to the 22nd dodecamer. A dashed line on top of the two‐layer ring helps track locations. Images were processed and analyzed in ChimeraX 1.2.5 ³⁷

Careful inspection of the dodecamer–dodecamer interface of Figure 3e revealed that the N‐terminal H1 helixes of each dodecamer subunit face each other, resulting in six hinges between dodecamers. We simulated a 22‐dodecamer EcGS filament in RELION 3.1 ¹⁸ according to the helical rise and twist values reported above. As anticipated, the EcGS subunits were stacked, displaying a clockwise helical pattern along the filament‐axis (Figure 3f), with 20 stacked dodecamers accomplishing a full 360° rotation to generate a helical pitch of ~196 nm.

Next, due to lack of crystal structure of EcGS, we compared our EcGS cryo‐EM structure with previously generated crystal structures from other bacterial GS. Structural similarities (based on root‐mean‐square deviation, RMSD) of a single EcGS subunit with those of MtGS (PDB ID: 1HTO), HpGS 1HTO(PDB ID: 5ZLI), and StGS (PDB ID: 2LGS) were calculated as RMSD = 2.4, 3.1, and 2.1 Å, respectively (Figure 4). Individual GS monomers exhibit profound structural similarity across these bacteria, except for some pore‐oriented spikes (residues 155–188; Figures 3d and 4). Although these spikes were not resolved for the EcGS filament as the respective cryo‐EM density is missing (Figures 3d and 4a), loop 393–407 (394–406 in StGS) responsible for negative‐feedback control ⁵ , ¹⁹ is clearly resolved in our cryo‐EM structure. The interaction between two EcGS subunits determines the stability of the two‐ring dodecamer. That interface involves a long β‐sheet (residues 133–153) and two C‐terminal helices (residues 436–456, 460–463). The assembly of two‐ring dodecamers is highly similar for the EcGS cryo‐EM and StGS crystal structures (RMSD = 2.1 Å). Overall, the individual subunits and interactions within one GS dodecamer of our cryo‐EM structure of filamentous EcGS are indistinguishable from the previously published crystal structure of StGS, with this latter sharing 98% identity to EcGS. The dodecameric structures of EcGS and StGS are also indistinguishable. Inspired by a bacterial flagellar protein structure, ²⁰ we also applied single particle reconstruction to our filamentous EcGS structure. The resulting 3D cryo‐EM map reconstruction focused on a single EcGS dodecamer has a resolution of 2.94 Å (Figure S2). There are no significant differences between our two cryo‐EM maps, but the 2.94 Å map provides more defined sidechain densities for model building.

FIGURE 4.

Structural comparison of GS subunits. Two views of monomer subunits from several bacterial GS structures are shown in (a) and (b). EcGS (blue), StGS (2LGS, green), MtGS (1HTO, gold), and HpGS (5ZLI, magenta) were selected to demonstrate structural similarity. Superimposed GS monomers are shown on the left, with EcGS shown in cartoon format and other GS shown as ribbons. Dashed lines represent the pore spike (155–188) and loop 394–406 missing from the EcGS (filament) and StGS (dodecamer) structures, respectively. The two‐layer subunit of one dodecamer is shown in (c), revealing identical EcGS and StGS dodecameric assemblies. The β‐sheet (residues 133–153) and α‐helices (residues 436–456 and 460–463) from one GS monomer interact with the adjacent GS monomer to stabilize the two‐layer dodecamer. Structural analysis and imaging were conducted in PyMOL 2.5.0 (Schrödinger)

2.3. Positioning of chelated nickel ions

To identify where nickel ions are located at the helix hinges, we conducted a focused map refinement on the interface between dodecamer rings. The refined map comprises two EcGS monomers of different dodecamers, with a density between the two parallel N‐terminal H1 helixes likely indicating the position of a single nickel ion (Figure 5a). Residues His5, Met9, Glu12, and His13 of the parallel H1 face each other and sandwich the Ni²⁺. We calculated distances of 3.9 and 4.3 Å for the ε‐nitrogen of His5 and the ε‐oxygen of Glu12 to the nickel ion, respectively. An estimated 2 Å hydration shell of Ni^{2+ 21} is then proposed to connect the solvated Ni²⁺ and protein residues. Therefore, the water‐mediated bonds stabilized the interactions of His5/Glu12 to Ni²⁺ during the filament formation. The two His13 residues are located 5.3 Å from the Ni²⁺ suggesting weak interactions. We identified that His13 forms a hydrogen bond link the two H1 of different dodecamers. The ε‐nitrogen of His13 and ε‐oxygen of Glu12 interact molecularly to further stabilize the two aligned helices via two hydrogen bonds. As the two H1 are aligned on the same horizon, 18° rotation of one dodecamer allows it to perfectly stack onto a previous one assembled on the filament. Thus, residues His5, Glu12, and His13 of each H1 helix play critical roles in stacking EcGS for filament formation.

FIGURE 5.

Nickel ions are clearly characterized in the GS filament map and structure. (a) The refined local map focused on the interface of two dodecamers reveals an apparent extra density between two H1 helixes. A nickel ion is then assigned to that density, showing distances between it and His5/Glu12 of 3.9–4.3 Å. A ~2 Å hydration shell for Ni²⁺ ion ²¹ is shown to illustrate water‐mediated contacts between His5/Glu12 to solvated Ni²⁺. The paired Glu12‐His13 hydrogen bonds indicated as yellow dashed lines between the two H1 helixes further stabilize the stacked dodecamers. (b) The GS monomeric subunit and extracted cryo‐EM map near the EcGS catalysis site reveal nickel ion occupation. The nickel ion is stabilized via interactions with three residues: Glu130, Glu358, and His270 within 3 Å

Next, we focused on the catalytic pocket of EcGS where ATP and glutamate are coupled via Mg²⁺ coordination to generate the intermediate product 𝛾‐glutamylphosphate. ¹ , ²² As expected, no density correlating to ATP or glutamate could be detected in our cryo‐EM structure (Figure 5b), implying that GS filaments represent polymerization of inactive GS dodecamers (apo form). Notably, we did detect an unknown density surrounded by charged residues, including Glu130, His270, and Glu358. Previous analysis has shown that a single Mg²⁺ ion coordinates with these residues and the phosphate groups of ATP at the catalytic pocket during catalysis. ⁵ Since our EcGS polymer was generated with Ni²⁺ in solution (and no Mg²⁺), the unknown density in our cryo‐EM map represents a single nickel ion. The ε‐nitrogen of His270 and both ε‐oxygens of Glu130 and Glu358 lie close to Ni²⁺ (2.1–2.6 Å), that is, at distances highly similar to previous data on Mg²⁺‐bound GS (StGS; PDB ID 2LGS). ⁵

2.4. EcGS filamentous polymerization is reversible

Our cryo‐EM map of EcGS filament has unveiled the mechanism by which chelated nickel ions promote filamentous GS extension. Such filament formation in other systems is known to be reversible, for example, the yeast glutamine synthetase Gln1. ¹⁶ Accordingly, we assessed if EcGS filamentation is also reversible by means of ethylenediaminetetraacetic acid (EDTA) treatment. First, we used DLS to monitor EcGS filaments formed over time in the presence of Ni²⁺ (Figure 6a). Ni²⁺ rapidly induced EcGS filamentation with particles attaining ~3,500 nm within 60 min and only slightly extending thereafter. We introduced 10 mM EDTA after 160 min of Ni²⁺‐induced filamentation to remove metal ions, resulting in particle sizes reducing dramatically (within 2 min) from 3,000–4,000 to 17–20 nm. Subsequent DLS revealed practically no EcGS filaments upon addition of EDTA, supporting that EDTA treatment effectively disassembled the stacked GS dodecamers. Negative‐stain TEM confirmed that the EDTA treatment completely disrupted all long EcGS filaments into individual dodecamers (Figure 6b).

FIGURE 6.

GS filament formation is reversible. (a) EcGS filament formation induced by Ni²⁺ was monitored by DLS, highlighting the rapid filament formation driven by Ni²⁺. Then, 10 mM EDTA was added to the two types of EcGS filament before immediately undergoing further DLS experiments. Particle sizes dropped to ~18 nm upon EDTA addition and remained that size thereafter. (b) Negative stain TEM of EcGS filament treated by EDTA in (A), revealing the absence of filament. (c) Particle sizes of EcGS filaments in the presence of glutamate, ATP, or ammonia chloride. As observed for EDTA, addition of negatively charged glutamate and ATP‐disrupted EcGS filaments into ~18‐nm dodecamers, with respective nsTEM images presented in panel (d). In contrast, addition of NH₄Cl had no impact on EcGS filaments. (e) A yeast Rsp5 HECT (369–807)‐EcGS fusion construct demonstrated that nickel‐induced filamentation no longer occurs when helix H1 is not solvent‐exposed. The TEM image displays four‐layer HECT‐EcGS particles in the presence of Ni²⁺

Since surface exposure of the catalytic pockets is indistinguishable for the filamentous and single dodecameric forms (Figures 4 and 5b), we wondered if EcGS filament is still capable of glutamine biosynthesis like dodecameric EcGS. Mixing EcGS filaments with the reactant glutamate or ATP (cofactor) resulted in particle sizes shrinking from >1,500 nm to essentially a dodecameric state (~20 nm; Figure 6c). Again, nsTEM confirmed the disappearance of filaments (Figure 6d). Negatively charged glutamate and ATP readily bind metal ions, so they likely sequester the metals coordinating the EcGS filaments. Thus, we infer that the filamentous form may not be associated with glutamine synthesis activity. GS also recruit ammonia to modify the intermediate product 𝛾‐glutamylphosphate to generate glutamine. We also examined by DLS and nsTEM the effects of NH₄ ⁺ treatment on filaments but, unlike the glutamate and ATP treatments, it did not induce EcGS filament disassembly (Figure 6c,d).

GS have been proposed to act as a scaffold protein for maltose‐binding protein (MBP) for cryo‐EM structure determination. ²³ The previously generated cryo‐EM map and structure of a MBP‐GS fusion protein (EMD‐4039) ²³ showed that 12 MBP proteins form two extra rings that sandwich the 12 GS subunits. In that MBP‐GS form, N‐terminal H1 is not solvent‐exposed. To confirm that H1 is the sole region responsible for high‐order assembly, we genetically fused a ~40 kDa HECT domain from yeast Rsp5 E3 ligase to the N‐terminus of EcGS. Thus, H1 in the HECT‐EcGS fusion protein was completely masked from the solvent, as well as from H1 of other dodecamers, so it can no longer respond to environmental nickel ions for assembly. Subsequent nsTEM imaging revealed that our HECT‐EcGS fusion protein (Figure 6e) exhibits a similar architecture to MBP‐GS. ²³ The four‐ring conformation of HECT‐EcGS was not altered upon adding nickel ions, implying that H1 of GS is exclusively responsible for divalent metal‐induced polymerization.

3. DISCUSSION AND CONCLUSION

The 469‐residue EcGS spontaneously forms a dodecamer. “Bundled” or filament‐like GS have been reported for decades, ⁵ , ¹² but a detailed mechanism of filamentation has been lacking. Here, we have determined a sub‐3Å cryo‐EM map of filamentous EcGS. Our structure reveals that the first helix (H1) is essential to bridging two native GS dodecamers. GS dodecameric stacking follows rules of helical filaments, with each GS dodecamer being rotated ~18° to enable it to load on top of existing filament via coordinating divalent metal ions (Ni²⁺, Co²⁺, and Zn²⁺). Our structural and mutational analyses have characterized that His5 and His13 in EcGS are key residues responsible for binding these metal ions, as well as for inter‐helical hydrogen bond formation. The salt bridges particularly between these two residues and metal ions, as well as the hydrogen bonds, stabilize dodecameric interactions to promote helical filament formation.

Intriguingly, not all bacterial GS possess two histidine residues at the same two positions as EcGS H1. An alignment of the entire GS sequences from 72 bacteria revealed that only seven species (~10%) have two histidine residues in H1 (Figure 7), including E. coli, Salmonella, and Shigella, implying that GS filament formation may be conserved among these species. We show that replacing EcGS His5 and His13 with negatively charged Asp or Glu still facilitated Ni²⁺‐induced filament formation (Figure 2c). In fact, our analysis of bacterial GS sequences in WebLOGO ²⁴ revealed that Asp and Glu are the two most common amino acids located at the positions occupied by His5 and His13 in EcGS. A phylogenetic tree of GS sequences (Figure 7b) illustrates one‐third of characterized bacterial GS possess the Asp/Glu pairing in H1. The GS of Staphylococcus and Streptomyces, as well as GlnA1 of Mycobacterium, are all members of the Asp/Glu subgroup. Four species of Archaebacteria and cyanobacteria exhibit a Glu/Glu pairing at these two positions, implying that their GS also potentially undergo filamentation. Enrichment for negatively charged residues in H1 is an alternative evolutionary mechanism to histidines for dodecameric bundling. Further structural determinations on purified GS from diverse bacterial species are needed to validate if Asp/Glu residues at the key positions can chelate metal ions and enable GS filamentation similarly. Notably, M. tuberculosis GlnA1, but not GlnA2, is abundant and essential for homeostasis, ²⁵ and we identified Asp/Glu and Phe/Arg residue pairings in H1 for GlnA1 and GlnA2, respectively, potentially providing insights into bacterial GlnA gene evolution.

FIGURE 7.

Evolutionary analysis of GS indicates a specific H5/H13 subgroup. Seventy‐two annotated bacterial GS sequences from UniProt were aligned in Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/) and illustrated using WebLogo ²⁴ (a) and a phylogenetic tree (b). The two crucial EcGS residues His5 and His13 are not common among most bacterial GS, only being found in seven of the assessed species (arrows in (a) and pink highlight in (b)), with Asp/Glu or Glu/Glu being more common (~33% of assessed species) (yellow or green highlight in (b)). Species names of several common bacteria have been labeled

Eukaryotic GS are relatively smaller than bacterial GS, as they form a decamer assembled from two pentameric rings. Yeast GS also exhibits an N‐terminal helix, ²⁶ though the first helix is oriented towards the pore. A crystal structure of yeast GS (ScGS, PDB ID 3FKY) illustrates stacked decamers in one asymmetric unit, though metal ion coordinations were not characterized for that crystal form. ScGS undergoes filament formation in vivo under conditions of nutrient shortage, ¹⁶ indicating that the recurring stacking reactions necessary for filamentation are promoted by low metabolite availability or low cytosolic pH. A mutational analysis of yeast GS ¹⁶ identified residues involved in decameric assembly, with charged residues potentially forming salt bridges and hydrogen bonds. However, the atomic detail of those interactions was not fully deciphered. It remains unclear if GS filament formation by E. coli and other bacteria is a stress response, though our data supports EcGS filament being inactive (Figure 6c,d). Further studies of EcGS under stress in vivo will be necessary to determine the physiological role, if any, of filamentous EcGS.

Many metabolic regulatory mechanisms are known to be controlled by enzyme filamentation, ²⁷ , ²⁸ , ²⁹ such as those mediated by yeast glutamate dehydrogenase, ²⁷ glutamine synthetase, ¹⁶ asparagine synthetase, ³⁰ and CTP synthase. ³¹ Here, we have demonstrated how metal ions can drive polymerization of EcGS. Nevertheless, Ni²⁺‐induced EcGS filaments cannot convert glutamate to glutamine, as the reagents glutamate and ATP immediately dissociate polymeric EcGS, even though the catalytic funnel for loading reaction substrate glutamate and cofactor ATP remains open. Intracellular yeast GS filamentation is known to silence glutamine synthesis and nitrogen assimilation, ¹⁶ and filamentous yeast CTP synthase displays reduced catalytic activity. ³¹ Thus, GS filamentation may be a universal mechanism to prevent continuous glutamine production in vivo, which would otherwise consume energy excessively and promote cellular proliferation. ³² Alternatively, bacterial GS filaments represent a stress response to temporarily suspend GS activity during dormancy or in harsh environments, with GS reverting to the dodecameric and active state immediately upon encountering sufficient nutrition to supply glutamine for the metabolic pathway.

4. MATERIALS AND METHODS

4.1. Cloning, expression, and purification of EcGS

E. coli GS gene GlnA (Uniprot: P0A9C5) was directly amplified from E. coli DH5α genomic DNA by a polymerase chain reaction (PCR), before being inserted into pRSFDuet‐1 vector with a designed N‐terminal hexahistidine‐maltose binding protein tag (HisMBP‐tag) and a TEV (tobacco etch virus) protease recognition sequence (ENLYFQ/GS; Figure S1). Gene amplifications with N‐terminal mutations including H5A, H13A, H5A‐H13A, H5D, H13E, and H5D‐H13E were achieved using altered 5′‐end primers. The Rsp5 (369–807)‐EcGS fusion construct was generated by two‐step PCR to fuse Rsp5 and GlnA genes. Rsp5 was directly amplified from yeast genomic DNA gifted by Dr Chung‐Rung Chang (National Tsing Hua University, Hsinchu, Taiwan). All EcGS variants were separately transformed overnight at 20–25°C into the BL21 DE3 cell line for overexpression until an optical density at 600 nm of >0.8 was achieved. His‐tagged EcGS in the supernatant of cell debris was loaded into manually packed cOmplete™ His‐Tag Purification Resin (Roche) and eluted using 250–300 mM imidazole. The N‐terminal HisMBP‐tag was then cleaved off using TEV protease. EcGS was further purified using a Sepharose 6 16/300 Increase size exclusion column (buffer: 25 mM Tris pH 7.6, 100–200 mM NaCl, 4 mM β‐mercaptoethanol) on an Akta FPLC Pure M system (Cytiva) at 4°C. Pure fractions of EcGS validated by SDS‐PAGE were combined and concentrated to 10–20 mg/ml. Concentrated EcGS was aliquoted and frozen in liquid nitrogen before long‐term storage at −80°C.

4.2. Dynamic light scattering

A Malvern Zetasizer Nano ZS system at the Academia Sinica Biophysics Core Facility was used to measure GS particle sizes under various conditions at 25°C. We mixed 0.1 mg/ml GS (25 mM Tris pH 7.6, 150 mM NaCl) and 0–20 mM divalent metal ions for 1 hr, before loading into a DLS cuvette. Real‐time particle sizes of GS in the presence of Ni²⁺ or Co²⁺ were monitored at 70‐s intervals. After 160 min, solution containing GS filaments was transferred to a test tube and thoroughly mixed with 10 mM EDTA, before being subjected again (2–3 min later) to DLS. Other treatments (40 mM Glu, 10 mM ATP or 10 mM NH₄Cl) were performed as per EDTA, but the GS were first mixed with 1 mM NiSO₄ for 1 hr and particles were observed to ensure GS filaments formed (threshold >1,500 nm). Particle sizes were measured at 70‐s intervals for 10–30 time‐points.

4.3. Negative‐stain TEM

Negatively stained grids of GS in dodecameric or filamentous form were visualized using a 120 kV JEOL 1400 electron microscope (JEOL, Japan) at a magnification of × 60,000. Filamentous EcGS was prepared by mixing GS (0.1 mg/ml) and divalent metal ions (MgCl₂, CaCl₂, MnCl₂, ZnSO₄, NiSO₄, or CuSO₄; 1 mM). Mixtures were equilibrated at room temperature for 1 hr prior to preparing negatively stained grids. Copper grids with carbon support film were glow‐discharged for 30 s before use. The dodecameric or filamentous EcGS in the test tubes was gently suspended, before immediately loading 4 μl of the sample on grids for negative staining. Excess GS was absorbed, and the grid was then stained with 2% uranyl acetate. Finally, grids were air‐dried for 12 hr or longer.

4.4. Cryo‐EM data collection and processing

Aliquots (4 μl) of Ni²⁺−treated GS filaments (25 mM Tris pH 7.6, 150 mM NaCl) were applied to Quantifoil Holey carbon grids (Cu, 200 mesh, R2/1 μm) at 100% humidity and 4°C, blotted with filter paper for 4 s, and vitrified in liquid ethane using a Vitrobot Mark IV system (Thermo). Data were collected using a 300 kV Titan Krios microscope (Thermo) equipped with a Gatan Quantum K2 Summit direct electron detector at the Academia Sinica Cryo‐EM Core Facility. The slit width was 20 eV. Automatic data collection with three exposures per hole was driven by EPU 1.2. software (Thermo). We collected 1,182 60‐frame movies with a physical pixel size of 0.822 Å/pixel in counting mode and a nominal magnification of × 1,650,000. The total electron dose and intended defocus range for a 60‐frame movie were 59.6 e⁻/Ų and −1 to −2 μm, respectively.

We employed cryoSPARC 3.2 to obtain the EcGS filament map using helical reconstruction. The EcGS filament dataset was processed using helical reconstruction and single particle analysis in cryoSPARC 2.14 and 3.2. ¹⁷ Movies were aligned using the full‐frame motion function, and CTF parameters were estimated by Patch CTF provided in cryoSPARC. We manually selected ~30 filaments to provide a good template for the filament tracer to select all filaments automatically. We then extracted 315,731 filament segments for 2D classifications, and 134,626 promising segments were used to perform helical reconstruction using the “Helical Refinement” tool in cryoSPARC 3.2. The initial model without applied symmetry was reconstructed at 4.9 Å. The initial helical rise of 97 Å and twist of −18° were determined using the symmetry search utility. These initial helical rise and twist values, as well as C6 symmetry, were then used to perform helical refinement, resulting in a map resolution of 3.58 Å using the gold standard FSC_0.143 criterion. ³³ The C6‐fold 3.58 Å map and calculated helical rise (97.5 Å) and helical twist (−18°) were used for subsequent non‐uniform helical refinement to obtain a 3.48‐Å helical reconstruction GS map. The final determined helical rise and helical twist values are 97.35 Å and −18.15°, respectively.

We manually selected ~200 particles to generate good templates for automatic particle selection for our single particle analysis. Then, 558,091 particles were picked and extracted using a downsized factor of 2‐ to 200‐pixel boxes. Of those, 117,242 particles were selected after 2D classification and re‐extracted to a box size of 400 pixels (pixel resolution 0.822 Å/pixel). Ab initio modeling was conducted in cryoSPARC 2.14, before applying 3D homogeneous refinement with C6 symmetry. The refined EcGS particles were further subjected to non‐uniform refinement by means of per‐particle CTF corrections using cryoSPARC 3.2. The final refined single particle reconstruction map of GS filament is 2.94 Å. A local refinement to mask two half‐dodecamers was subsequently performed to provide better detail of H1–H1 interactions.

4.5. GS atomic structure building

The dodecameric EcGS structure was homology modeled using the crystal structure of S. typhimurium GS (PDB ID 2LGS) ⁵ for subsequent refinements. Sequence identity between these two GS is 98%, only differing by nine residues. The model was initially aligned to the GS map using the “Dock in map” function of Phenix 1.18.2. ³⁴ The density of one GS subunit (monomeric form) was extracted for model building with manual inspection of all 469 residues using Coot 0.9.5. ³⁵ Loop 393–407 missing from the StGS template (PDB ID 2LGS) was manually built. The monomeric model was then refined by real‐space refinement in Phenix to improve the quality of the EcGS model. The EcGS monomer displaying a low Molprobity clashscore ³⁶ and a highly favorable percentile for the Ramachandran dihedral angle was then selected for alignment to the 2LGS structure to create a D6‐fold dodecamer. Sidechain clashes, disfavored dihedral angles, and outliers of rotamers in the subunit interfaces of the dodecameric EcGS, especially the tangled β‐sheets of two‐ring layers and the contacts between neighboring chains, were manually inspected to improve model quality. The structure was refined and polished iteratively to satisfy map and model agreement. The dimeric “EcGS dodecamers” and associated map were finally validated using the Phenix validation tool to confirm acceptable Molprobity scores, torsion angles, and model‐to‐map CC values for publication. Details of the cryo‐EM structure and maps are summarized in Table 1. We used the RELION 3.1 helical_toolbox ¹⁸ to construct 22 dodecameric GS repeats as representative GS helical filament based on calculated helical rise and twist values.

TABLE 1.

Cryo‐EM data collection, refinement, and validation statistics

	GS polymer (2 stacked dodecamers)
Data collection and processing
Accession codes (PDB/EMDB)	7W85/32,352
Microscope/detector	Krios/K2
Magnification	×165,000
Voltage	300 kV
Electron exposure (e−/Å2)	59.9
Defocus range (μm)	−1 to −2
Pixel size (Å)	0.822
Symmetry imposed	C6
Initial particle images (n)	315,731	558,091
Final particle images (n)	134,626	117,242
Reconstruction method	Helical reconstruction	Single particle reconstruction
Map resolution (Å) (FSC0.143)	3.48	2.94
Map resolution range (Å)	2.5–4.5	2.2–6.5
Map sharpening B factor (Å2)	−115.6	−93.2
Helical filament parameter	Twist (−18.15°), rise (97.35 Å)
Refinement
Initial model	2LGS
Model composition
Non‐hydrogen atoms	81,522
Protein residues	10,440
Ligands (Nickel)	42
B factor (Å2) (min/max/mean)
Protein	3.99/75.60/29.93
Ligand	30.00/108.90/57.58
R.m.s. deviations
Bond lengths (Å)	0.008
Bond angles (°)	1.086
Validation
MolProbity score	1.87
Clash score	9.32
Poor rotamers (%)	0
Ramachandran plot
Favored (%)	94.44
Allowed (%)	5.56
Disallowed (%)	0

AUTHOR CONTRIBUTIONS

Pei‐Chi Huang: Data curation (lead); writing – original draft (supporting). Shao‐Kang Chen: Data curation (supporting); writing – original draft (supporting). Wei‐Hung Chiang: Data curation (supporting). Meng‐Ru Ho: Data curation (supporting). Kuen‐Phon Wu: Conceptualization (lead); data curation (lead); formal analysis (lead); funding acquisition (lead); methodology (lead); writing – original draft (lead).

Supporting information

Appendix S1: Supporting information

Huang P‐C, Chen S‐K, Chiang W‐H, Ho M‐R, Wu K‐P. Structural basis for the helical filament formation of Escherichia coli glutamine synthetase. Protein Science. 2022;31(5):e4304. 10.1002/pro.4304