Assembly status transition offers an avenue for allosteric activity modulation of a supramolecular enzyme

Nature has evolved many supramolecular proteins assembled in certain, sometimes even seemingly oversophisticated, morphological manners. The rationale behind such evolutionary efforts is often poorly understood. Here we provide atomic-resolution insights into how the dynamic building of a structurally complex enzyme with higher-order symmetry offers amenability to intricate allosteric regulation. We have established the functional coupling between enzymatic activity and protein morphological states of glutamine synthetase (GS), an old multi-subunit enzyme essential for cellular nitrogen metabolism. Cryo-EM structure determination of GS in both the catalytically active and inactive assembly states allows us to reveal an unanticipated self-assembly-induced dynamics-driven allosteric paradigm, in which the remote interactions between two subcomplex entities significantly rigidify the otherwise structurally fluctuating active sites, thereby regulating activity. We further show in vivo evidences that how the enzyme morphology transitions could be modulated by cellular factors on demand. Collectively, our data present an example of how assembly status transition offers an avenue for allosteric modulation, and sharpens our mechanistic understanding of allostery, dynamics, cooperativity, and other complex functional and regulatory properties of supramolecular enzymes.

distal to either the catalytic site or substrate binding regions with distances >20 Å and 188 hence are unlikely to be directly involved in catalytic reaction, we infer that the 189 stimulations of the enzymatic activity of CsGSIb upon residue perturbations are 190 attributed to allosteric effects induced by remote contacts between two pentamer rings. 191 As expected, the mutation of Y150F, which did not alter pentamer-decamer equilibrium 192 of CsGSIb (Fig. S3a), showed no noticeable change in enzymatic activity (Fig. S3b). 193

Structural basis for dynamics-driven allostery of GSII 194
To elucidate the allosteric mechanism of how the interactions between two GSII 195 pentameric rings remotely trigger enzymatic activity, we next employed single-particle 196 cryo-EM imaging technique and first determined the structures of GmGSβ2 decamer, as 197 well as that of the CsGSIb that adopts decameric configuration (thereafter named as 198 CsGSIb Dec ). 3D classifications of 104717 and 43876 particles for GmGSβ2 and 199 CsGSIb Dec , respectively, revealed that both molecules were arranged in D5 symmetry 200 with two pentameric rings stacked in a head-to-head manner ( Fig. 3a and 3b), a 201 strikingly conserved structural feature that have been widely observed for other type II 202 GS species [11][12][13] . Refinement of the GmGSβ2 and CsGSIb Dec structures yielded maps 203 with an average resolution of 2.9 Å and 3.3 Å, respectively, with literally identical 204 dimensions of 115 Å x 115 Å x 95 Å. As expected from the very high conservation in 205 amino acid sequence (Fig. 1a), decamer structures of GmGSβ2 and CsGSIb Dec are very 206 similar to each other, as well as to that of the GSII of maize (pdb accession number 207 2D3A), as highlighted by the root-mean-square deviation (r.m.s.d.) of 0.66-0.80 Å for 208 328-352 aligned C  atoms. Structural alignments reveal that the active sites in GmGSβ2 209 and CsGSIb Dec , as well as that in the maize GSII, are highly conserved (Fig. 3c), suggesting the catalysis mechanism of these three enzymes, once decamers are 211 formed, are essentially identical. The overall buried inter-ring surfaces for both GmGSβ2 212 and CsGSIb Dec amount to ~2000 Å 2 , i.e. approximately only 400 Å 2 per individual 213 monomer-monomer interaction. This highlights the weakness of the inter-ring contacts, 214 characteristic of type II GS. Indeed, in both structures the inter-ring contacts are 215 established by only a limited number of hydrophobic and polar interactions provided by 216 the residues 136-141 and 146-152 segments of each of the intervening subunits ( Fig.  217 S7), which behave as two gear teeth (thereafter named as tooth-1 and tooth-2, 218 respectively) interlocking the two pentameric rings (Fig. 3d-f). This observation is in line 219 with the above result that mutations to tooth-1 resulted in drastic change in oligomeric 220 states behavior (Fig. 2d-2g), and the mixed nature of the inter-ring interactions is 221 consistent with the MALS analysis result of CsGSIb under various buffer conditions (Fig.  222 S1b). Intriguingly, the local structure of the teeth regions that mediate inter-ring 223 interactions remains largely the same in GmGSβ2 and CsGSIb Dec (Fig. 3d-3e), 224 suggesting the dramatically different propensities of GmGSβ2 and CsGSIb for decamer 225 formation are due to the nature of the amino acids involved in inter-ring contacts, rather 226 than the structure. Although the residue of I143 is not directly involved in inter-ring 227 contact, we argue that its replacement with leucine may stabilize the conformation of 228 tooth-1 via its interaction with the residue of L134, thus playing an important role in 229 stabilizing decamer architecture ( Fig. 2d and 2e). In order to elucidate the mechanism of how the pentameric CsGSIb (thereafter named 231 as CsGSIb Pen ) demonstrates distinct enzymatic properties than CsGSIb Dec (Fig. 2h), we 232 next determined the structure of CsGSIb Pen . Lowering the sample concentration, which favored pentamer dissociation (Fig. S1a), allowed us to obtain sufficient number of 234 CsGSIb Pen particles, which, in turn, enabled us to solve the cryo-EM structure of the 235 inactive single-ringed GSII for the first time. Interestingly, we observed additional class 236 averages in which the two masses of density attributed to the pentameric rings are no 237 longer parallel (Fig. S8). These non-parallel ring particles may reflect intermediate 238  among which, the fragment around residues 260-334 is a major component making up 255 an integral catalytic site (Fig. 4c), while the segment of residues 140-166 comprising of 256 the two gear teeth is responsible for ring-ring interaction (Fig. 3d-f and Fig. 4d). As 257 electron density missing often reflects the conformational heterogeneity arising from 258 internal motions 17 , these observation strongly suggest that the conformation of 259 CsGSIb Pen active site is highly dynamic, contrasting sharply to the conformationally 260 largely homogeneous CsGSIb Dec . In support this, thermal shift assays show the melting 261 temperature (T m ) of wild-type CsGSIb is significantly lower than that of its mutants of 262 I143L or EVK138DIQ, indicating of structural instability for the pentameric GSII (Fig.  263 S10). We therefore conclude that the dramatically difference in the dynamic property of 264 catalytic sites accounts for the distinct activities of pentameric and decameric CsGSIb. 265 Taken together, our cryo-EM structures allow us to propose a dynamics-driven allosteric 266 mechanism of how the GSII activity is regulated by changes in oligomeric state: (1) The 267 active sites within isolated CsGSIb Pen rings are highly disordered and the unstable 268 catalytic environments render it catalytically inactive; (2) Upon stacking of two 269 pentameric rings and formation of a decamer, the signals of interactions mediated by 270 the gear teeth of each intervening subunit are allosterically propagated to the 271 active sites, which reduce their conformational dynamics and in turn, unlock the 272 catalytic potential of GSII (Fig. 4e). 273

Activation of the GSII by the 14-3-3 scaffold protein 274
Having mechanistically established the allosteric coupling between GSII activity and its 275 quaternary assembly status, we next asked whether there exist cellular factors that may 276 regulate the CsGSIb activity, potentially via favoring its decamer assembly. 14-3-3 277 proteins are an important family of scaffold proteins that bind and regulate many key 278 proteins involved in diverse intracellular processes in all eukaryotic organisms 18-20 . In 279 particular, self-dimerization of 14-3-3 proteins, which induces dimerization of their 280 clients, plays a key role in its functional scaffolding and subsequent activity regulation 281 18,19,21 . Moreover, it has been reported that 14-3-3 proteins act as an activator of GSs in 282 various plants 22-25 , although the detailed activation mechanism remains unclear. Based 283 on these findings, here we tentatively provide the missing link in mechanistically 284 assigning the role 14-3-3 proteins play in regulating GS activity: One protomer of the 14-285 3-3 protein recognizes one phosphorylated GSII pentamer, and its self-dimerization 286 brings two pentamer rings in close proximity and therefore promotes decamer assembly, 287 which, in turn, switches on the GS activity via allosteric rigidification of the catalytic sites 288 (Fig. 5a). One prerequisite for this proposal is that, for the GS species whose activities 289 being 14-3-3 protein-dependent, they must have an intrinsically weak decamer-forming 290 propensity that is to be overcome by 14-3-3. In support of this, the GS from Medicago 291 truncatula, whose activity is simulated upon binding to 14-3-3 protein 24 , has been 292 shown to exhibit a dynamic pentamer-decamer transition 12 , similar to the CsGSIb 293 presented here (Fig. 1d). Moreover, it has been shown that only the higher order 294 complex of tobacco GS-2 that is bound to 14-3-3 is catalytically active 25 . 295 To further support the above proposal, we then explored whether the activity of the 296 weak-decamer forming CsGSIb could also be regulated by 14-3-3 scaffold protein. Homology search against tea plant genome revealed several candidate tea plant Cs14-298 3-3 proteins (Fig. S11). Analysis of their coding genes' expression patterns in tea plant 299 tissues and in nitrogen assimilation or metabolism-related processes allowed us to 300 identify Cs14-3-3-1a and Cs14-3-3-1b genes that displayed expression patterns highly 301 similar to CsGSI genes ( Fig. 5a and S12). Moreover, the expression levels of Cs14-3-302 3-1a and Cs14-3-3-1b genes were regulated upon changes in the availability of 303 ammonia ( Fig. 5b and 5c), the substrate of GS, suggesting both Cs14-3-3-1a and Cs14-304 3-3-1b are physiologically related to GS. We then examined the in vivo interactions 305 between Cs14-3-3-1a and CsGSIb using the bimolecular fluorescence complementation 306 (BiFC) technique, which is based on complementation between two non-fluorescent 307 fragments of a fluorescent protein when they are brought together by interactions 308 between proteins fused to each fragment 26 . Cs14-3-3-1a or CsGSIb were fused in 309 frame with N-terminal half of a yellow florescence protein (NYFP) or C-terminal half of a 310 yellow florescence protein (CYFP), respectively, and expressed in tobacco leaf 311 epidermal cells alone or in various combinations, such as CsGSIb CYFP alone or together 312 with Cs14-3-3-1a NYFP . As expected, Cs14-3-3-1a and 1b could self-dimerize or form 313 heterodimers in plant cells (Fig. 5e), consistent with 14-3-3 scaffold proteins adopting a 314 dimeric structure 18,19,21 . Importantly, formation of the fluorescent complex clearly 315 demonstrated the interaction of Cs14-3-3-1a with CsGSIb ( Fig. 5f). In order to further 316 establish the functional relevance, we performed the RNA interference (RNAi) technique 317 to knock down the transcript level of Cs14-3-3-1a gene in hairy roots of chimerical 318 transgenic tea seedlings (Fig. 5h), and evaluated the impact on GS activity by 319 measuring the contents of GS catalysis product, glutamine. We show that, along with 320 the reduction in Cs14-3-3-1a transcript level, the glutamine contents (Fig. 5i), as well as 321 the crude enzyme activity (Fig. 5j), were drastically reduced, indicating the 14-3-3 322

Dynamics-driven allostery induced by assembly status transition 339
Allostery describes the mechanism that binding effector molecules at one site triggers 340 a conformational or dynamic change at a distant site, thereby affecting protein activity. 341 Allosteric regulation is a common mechanism to regulate protein function, playing 342 critical roles in various cellular activities ranging from the control of metabolic 343 mechanisms to signal-transduction pathways 27 . Based on the data from Allosteric 344 Database 28 , to date more than 1,900 proteins have been defined as allosteric. Most 345 allosteric modulators identified are small ligands or peptides, whereas in some rare cases the allosteric effects are induced by protein oligomerization in a rather simple 347 system 29 . Here we show that the oligomeric GS ring functions as positive modulator, 348 the largest allosteric modulator identified so far to our knowledge; and the ring-ring 349 association, which is motivated by 14-3-3 protein or other factors, leads to a transition of 350 assembly symmetry from C5 to D5 and subsequently triggers allosteric activation. 351 We show here that the assembly geometry of GS plays a critical role in determining 352 the protein functional motion properties (Fig. 4) by multiple post-translational mechanisms including nitration, oxidative turnover and 367 phosphorylation 37 , and by the 14-3-3 protein 22-25 . Our results reconcile with many of 368 the above observations, and allow us to gain a more complete picture of how GS activity is regulated in cells by an exquisite machinery (Fig. 6). While the protein 370 turnovers machineries to adjust cellular enzyme level can always provide means to 371 modulate pentamer-decamer transitions and thus deactivation-activation conversion of 372 GSII, we argue that phosphorylation-dephosphorylation processes, coupled with 14-3-3 373 binding and subsequent allosteric activation, may enable a more efficient regulatory way. 374 When sufficient reaction products are available demanding low glutamine synthesis 375 activity, GSII is kept in the dephosphorylated state by certain phosphatase and exists as 376 isolated inactive single-ringed pentamers. In the physiological context of high demand of 377 glutamine, phosphorylation of GSII by certain kinase prompts 14-3-3 protein binding, 378 and the intrinsic dimerization property of 14-3-3 recruits two GSII pentamer rings in 379 close proximity and in doing so, result in a rapid transition of quaternary assembly from 380 the pentamer to decamer, and eventually enzymatic activation. In this manner, the 381 poised GSII pentamer ring itself acts as a positive effector and the allosteric ring-ring 382 association offers a great advantage of immediate response to precisely meet the ever-383 changing metabolic needs, whereas the reversible assembly-disassembly behavior 384 enables a tunable mode for activity modulation. Indeed, the dynamic association-385 disassociation of GSII subcomplexes, a prerequisite for this modulatory machinery, 386 have been widely observed in various species including humans 38 , plants other than 387 Camellia sinensis reported here 12,15 and fungi 39 . Therefore, the dynamics-driven 388 allostery shown here may represent a general regulatory machinery harnessed by many 389 eukaryotes to ensure optimal utilization of nitrogen sources, and the infrastructure of 390 fragile ring-ring contacts evolutionarily chosen by many eukaryotes offers a convenient 391 and robust avenue for activity regulation.

Practical implications 393
As a crucial enzyme to all living organisms, which is involved in all aspects of nitrogen 394 metabolism, GS has emerged as an attractive target for drug design 40 and herbicidal 395 compounds development, as well as a suitable intervention point for the improvement of 396 crop yields 41 . However, because the overall geometry of the active site is the most 397 conserved structural element amongst GS enzymes 4,7,11 , the traditional strategy of 398 selective inhibition, which relies heavily on the subtle difference in the active sites from 399 different species, has only achieved limited success. Thus, the regulatory mechanism 400 discovered here will help guide the search for specific inhibitors of potential therapeutic

501
Cryo-electron microscopy image processing, 3D reconstruction, and analysis 502 All processing steps were performed using cryoSPARC 47 . A total of 4,051 raw movie stacks 503 acquired for CsGSIb and 1,777 raw movie stacks for GmGSβ2 were subjected to patch motion 504 correction and patch CTF estimation. An initial set of about 500 particles were manually picked to 505 generate 2D templates for auto-picking. The auto-picked particles were extracted by a box size of 506 512 pixel and then subjected to reference-free 2D classification. After particle screening using 2D 507 and 3D classification, the final 355,289 particles for CsGSIb and 115,795 particles for GmGSβ2 were 508 subjected to Ab-Initio Reconstitution and followed by 3D Refinement with C5 symmetry imposed.

509
Four different conformational states were obtained for CsGSIb, resulting in a 3.3 Å density map for 510 CsGSIb Dec , 3.5 Å density map for CsGSIb Pen State I, 3.6 Å density map for CsGSIb Pen State II, and

514
Model building and structural refinement 515 Homology models of CsGSIb and GmGSβ2 were generated with the I-TASSER server 48 and 516 docked into the cryoEM maps using UCSF Chimera 49 . The sequences were mutated with 517 corresponding residues in CsGSIb and GmGSβ2, followed by rebuilding in Coot 50 . The missing 518 residues of CsGSIb Pen were not built due to the lack of corresponding densities. Real-space refinement of models with geometry and secondary structure restraints applied was performed using 520 PHENIX 51 . The final model was subjected to refinement and validation in PHENIX. The statistics of 521 cryo-EM data collection, refinement and model validation are summarized in Table S1.

522
Different nitrogen treatments for hydroponically grown tea cuttage seedlings 523 Two-year-old hydroponic tea cuttage seedlings were grown in a greenhouse at 20-25°C until new 524 tender roots emerged. These healthy tea seedlings were then transferred into hydroponic solutions

RNA isolation and qRT-PCR analysis 530
Tea plant tissues or root materials were ground in liquid nitrogen into fine powders for total RNA

542
All experiments were independently repeated three times, and relative expression levels were 543 measured using the 2−ΔCt method.

548
Determination of the subcellular localization of these GFP fusions was performed using tobacco leaf 549 infiltration as previously described (Zhao et al., 2011). Briefly, the pK7WGF2-Cs14-3-3-1a-GFP, 550 pK7WGF2-Cs14-3-3-1b-GFP, and pK7WGF2-CsGSIb-GFP plasmids were transformed into A. 551 tumefaciens strain EHA105, and selected positive colonies harboring these constructs were used for 552 plant transformation by infiltration. Acetosyringone-activated Agrobacterium cells were infiltrated into 553 the Nicotiana benthamiana leaves leaf abaxial epidermal surface, and the tobacco plants were 554 grown at room temperature for 3 days before imaging. Imaging of these GFP fusion proteins was 555 performed using a confocal microscope with a 100× water immersion objective and appropriate 556 software. The excitation wavelength was 488 nm, and emissions were collected at 500 nm.

GS activity assay from plant samples 589
To determine the total GS activity, 100 mg of frozen plant samples were grounded into fine powder tended to facilitate re-association of pentameric subcomplexes to some extent (Fig.  754 S1b), suggesting a mixture of electrostatic and hydrophobic interactions is responsible 755 for attaching of two pentameric rings. 756 To probe the effects of various ligands on the stability of CsGSIb, a series of 757 fluorescent dye-monitored thermal shift assays were carried out as described previously 758 13 . As shown in Fig. S1c, addition of magnesium ions or its combination with the 759 nonhydrolyzable ATP analog AMPPNP resulted in increases in the melting temperature 760 (T m ) of CsGSIb with ~6 or ~10 °C, respectively, while the presence of the substrate of glutamate showed no apparent effect on T m , consistent with previous observations 13 . In 762 contrast, as evidenced by SEC-MALS measurements (Fig. S1d), the presence of the 763 above ligands showed no appreciable effect on the decamer-forming properties. These 764 observations collectively indicate that while binding to substrate or cofactor rigidifies the 765 structural organization within individual pentamer rings, the inter-ring assembly of GSII 766 is, at least in vitro, not substrate-induced.