Visualizing the chaperone-mediated folding trajectory of the G protein β5 β-propeller

SUMMARY

The Chaperonin Containing Tailless polypeptide 1 (CCT) complex is an essential protein folding machine with a diverse clientele of substrates, including many proteins with β-propeller domains. Here, we determined structures of human CCT in complex with its accessory co-chaperone, phosducin-like protein 1 (PhLP1), in the process of folding Gβ₅, a component of Regulator of G protein Signaling (RGS) complexes. Cryo-EM and image processing revealed an ensemble of distinct snapshots that represent the folding trajectory of Gβ₅ from an unfolded molten globule to a fully folded β-propeller. These structures reveal the mechanism by which CCT directs Gβ₅ folding through initiating specific intermolecular contacts that facilitate the sequential folding of individual β-sheets until the propeller closes into its native structure. This work directly visualizes chaperone-mediated protein folding and establishes that CCT orchestrates folding by stabilizing intermediates through interactions with surface residues that permit the hydrophobic core to coalesce into its folded state.

Graphical Abstarct

eTOC blurb

Wang*, Sass* et al. determined structures of the CCT-Gβ₅-PhLP1 complex that revealed Gβ₅ folding intermediates ranging from a molten globule to a fully folded β-propeller. These intermediates define a Gβ₅ folding trajectory guided by CCT interactions with surface residues that allow the hydrophobic core to coalesce into its folded state.

INTRODUCTION

G proteins mediate the transmission of a myriad of extracellular signals across the plasma membrane to the cell interior^1,2. The bridge for signals to traverse the plasma membrane is the G protein-coupled receptor (GPCR), a seven transmembrane helical structure that binds to extracellular ligands and undergoes conformational changes that propagate to the cytoplasmic surface³. Inside the cell, the activated GPCR interacts with the G protein heterotrimer (consisting of Gα, Gβ, and Gγ subunits) to release GDP from the Gα subunit and bind GTP^4,5. This nucleotide exchange disrupts the association of Gα-GTP with the GPCR and the Gβγ dimer, allowing both Gα-GTP and Gβγ to activate downstream effectors^6–9. Gα hydrolyzes GTP to GDP, thereby inactivating the Gα. However, for many Gα isoforms, intrinsic GTP hydrolysis is too slow for physiological responses and requires acceleration by Regulator of G protein Signaling (RGS) proteins^10,11. In neurons, members of the R7 family of RGS proteins associate with a unique isoform of Gβ (Gβ₅) to form dimers that interact with Gα and enhance its GTPase activity^12–14.

To perform their signaling function, G protein subunits must first be folded and then assembled into Gαβ_1–4γ or Gβ₅-RGS complexes by molecular chaperones^15–18. The primary chaperones for Gβ subunits are the cytosolic chaperonin CCT (also known as T-complex polypeptide 1 ring complex, or TRiC) and its co-chaperone phosducin-like protein 1 (PhLP1)^15,17,19,20. CCT is composed of two sets of eight paralogous subunits that form two back-to-back rings, thus creating a barrel-shaped structure with a central folding chamber into which nascent or otherwise unfolded proteins enter and are assisted in folding^21–23. Each CCT subunit is an ATPase that binds and hydrolyzes ATP to drive a conformational change that seals off the ends of the barrel and encapsulates the unfolded protein, allowing it to fold in the sequestered environment of the folding chamber^24–26. After phosphate release, the ring reopens and enables the folded protein to be released.

CCT has been estimated to fold up to 10% of the cytosolic proteome²⁷, particularly proteins with multiple domains and complex folding patterns²⁸. Gβ subunits belong to the β-propeller family of protein folds, a major class of CCT substrates with important physiological functions²⁹. This family of β-propellers contains seven WD40 repeat sequences that fold into structures consisting of seven β-sheets arranged into a radial, propeller-like structure (Figure 1A)^30–32. For β-propellers to fold completely, the N- and C-terminal β-strands of the β-propeller must be brought together to form the last β-sheet²⁰. Yet, the mechanism underlying CCT-mediated folding remains unresolved.

Figure 1. Reconstructions of Gβ₅-PhLP1-CCT complexes.

(A) Domain organization and structure of folded Gβ₅. B1–7, Blades 1–7. (B) Views of CCT-PhLP1-Gβ₅ reconstructions in open, intermediate, and closed states derived from a single dataset. CCT, gray; PhLP1, pink; Gβ₅, blue. (C) Conformational change in hydrophobic N-terminal arms of CCT5 and CCT7 between open and closed states. Density and model of additional ordered N-terminal residues observed in the closed state of CCT5 and CCT7 are colored red. (D) CCT closure creates a binding interface that stabilizes the CCT5 and CCT7 N-terminal arms (sequences indicated with hydrophobic residues in red). See also Figure S2.

The CCT co-chaperone, phosducin-like protein-1 (PhLP1), plays an important role in Gβ subunit folding and assembly into Gβγ and Gβ₅-RGS dimers. PhLP1 binds CCT and helps to fold Gβ_1–4 and Gβ₅ and then release them to interact with their Gγ or RGS binding partner^15,16,19,20. The vital role of PhLP1 in both Gβγ and Gβ₅-RGS dimer formation has been confirmed in vivo, where deletion of PhLP1 in mouse photoreceptor rods³³, cones³⁴, and striatal neurons³⁵ resulted in reduced levels of Gβγ and Gβ₅-RGS complexes and corresponding disruption of G protein signaling.

To understand the molecular mechanism underlying the roles of CCT and PhLP1 in Gβ₅ folding, we isolated Gβ₅ in complex with CCT and PhLP1 in the process of folding and used cryo-EM to determine an ensemble of structures that capture its complete folding trajectory. The structures reveal how CCT and PhLP1 orchestrate the Gβ₅ folding trajectory from an unfolded molten globule state to its fully folded β-propeller. Partially folded intermediate states are stabilized by hydrophilic interactions between surface residues of Gβ₅ and the apical domains of CCT that allow the hydrophobic core of Gβ₅ to coalesce into its β-propeller structure.

RESULTS

Reconstructions of CCT-PhLP1-Gβ₅ complex

Human CCT-PhLP1-Gβ₅ particles were isolated directly from HEK293T cells virally transduced with His₆-tagged human PhLP1 and Strep-tagged human Gβ₅ (Figure S1A). A two-step affinity purification scheme was used to first isolate PhLP1 complexes on a Ni²⁺-chelate column, followed by Gβ₅ complexes on a StrepTactin column. Coomassie blue stained SDS-PAGE showed strong bands corresponding to the expected molecular weights of CCT subunits, along with a band resulting from the co-migration of PhLP1 and Gβ₅. Immunoblot analysis confirmed the presence of PhLP1, Gβ₅ and CCT in the purified sample. ATPase assays showed the complex hydrolyzed ATP at a rate comparable to CCT alone, indicating that the complex was functional (Figure S1B).

Previous studies demonstrated that substrate folding is initiated by inducing a conformational change from an open to a closed state of CCT in an ATP-dependent manner^36,37. To capture CCT complexes in the act of folding Gβ₅, we incubated CCT-PhLP1-Gβ₅ particles with ATP and aluminum fluoride to generate the transition state analogue ADP-AlF_x and induce the closed form. The inclusion of ATP in the initial reaction permits at least one round of ATP hydrolysis, while the hydrolyzed nucleotide is presumably stabilized by AlF_x. Native gel analysis after nucleotide treatment confirmed a conformational shift in CCT, consistent with a transition from the open to the closed form as the complex becomes more compact^26,38 (Figure S1C).

Purified CCT-PhLP1-Gβ₅ particles were used for single particle cryo-EM imaging. Image classification revealed the enrichment of three distinct conformational states of CCT, including open, intermediate, and closed states with resolutions at 3.4 Å, 3.7 Å and 2.4 Å, respectively (Figures 1B and S2). The intermediate state shows CCT subunits with higher ATPase activity (CCT2, 4, 5) adopting a closed conformation, while those with lower ATPase activity remain in the open conformation, leaving the CCT barrel partially closed. Both the open and intermediate states contained additional densities within CCT at the apical and equatorial regions at poor local resolution. The size of the equatorial density roughly matched the expected volume of Gβ₅, and its position between the CCT rings was consistent with other unfolded substrates that have been reported previously^36,37,39. Our reconstruction also reveals N-terminal extensions contributed by both copies of CCT5 and CCT7 that tether the unfolded substrate within the center of the barrel (Figure 1C). These N-terminal regions are enriched in hydrophobic residues, suggesting that hydrophobic contacts partially contribute to the stabilization of the unfolded Gβ₅ substrate. These observations indicate that the open and intermediate states of CCT hold Gβ₅ in its cavity near the equatorial domain through interactions mediated by CCT5 and CCT7, but the substrate is held in a compact molten globule state.

The densities associated with the apical regions of CCT6 in each ring are consistent with the compactly folded thioredoxin domain (TXDN, residues 161–270) of PhLP1^37,40,41 (Figure S3). Although these densities are directly connected to CCT6, the poor local resolution in this region limited our ability to analyze the PhLP1-CCT6 interaction. Other regions of PhLP1 were not observed in our map. To validate the assignment of these densities to PhLP1, we performed cryo-EM imaging and computed a 3D reconstruction of human CCT particles that were extensively purified to remove bound substrates and co-chaperones (Figures S1D and S3B). This reconstruction was essentially identical to other published substrate-free structures of CCT. A difference map between this reconstruction and the open state of CCT-Gβ₅-PhLP1 indicate that the extra apical densities belong to PhLP1 (Figure S3C). We note that some density is still observed in the equatorial region despite the extensive washing, and likely represents unstructured N- and C-termini of CCT subunits or tightly bound endogenous substrate also observed by others^36,37,42 (compare Figs. S3A and S3B).

To further verify the positions of Gβ₅ and PhLP1 in the CCT open state, we performed crosslinking mass spectrometry (XL-MS) of purified CCT-PhLP1-Gβ₅. Crosslinking distances were minimized when Gβ₅ was positioned in the mass between the CCT rings (Figure S4). For PhLP1, no crosslinks were observed between the PhLP1 thioredoxin domain and CCT, but several were identified between the N-terminal domain and CCT4 and 5 in the equatorial and intermediate regions, indicating that this domain extends deeper into the CCT folding chamber and is not the PhLP1 density associated with the apical domain of CCT6. Based on these results, it appears that unfolded Gβ₅ is tethered to the center of CCT in its open form by extended N-terminal arms from CCT5 and CCT7, and the complex is capped at the apical ends by the PhLP1 thioredoxin domain with the PhLP1 N-terminal domain extending into the CCT cavity toward Gβ₅.

Gβ₅ and PhLP1 become ordered in the closed state of CCT

In contrast to the open and intermediate states, the densities for Gβ₅ and PhLP1 became well-resolved in the closed state of CCT (Figures 1B and S2). Gβ₅ and PhLP1 occupy opposing ends of the closed CCT chamber in different positions compared to the open state (Figure 1B). Most notably, Gβ₅ shifts positions from being tethered between the CCT rings in the open state and moves to the apical region of CCT2, 5, and 7 in the closed state. Interestingly, the N-terminal arms of CCT5 and CCT7 fold back onto the CCT inner wall in the closed CCT structure (Figures 1C, D). The repositioning of the CCT5 and CCT7 N-terminal segments appear to be induced by conformational changes at the interfaces between CCT rings. Specifically, the inner surface of CCT in the open state contains a large gap between CCT7 and CCT1′ in the other ring (Figure 1D). Upon CCT closure, the gap is shortened and creates a binding interface that accommodates the N-terminal arm of CCT5. Similarly, a gap between CCT8 and CCT3′ is observed in the open state and becomes narrowed in the closed state to accommodate the N-terminal arm of CCT7. These conformational differences between the open and closed states indicate a mechanism in which the N-terminal arms hold unfolded substrate in the center of the open CCT barrel, followed by the withdrawal of the arms upon CCT closing to release substrate into the folding chamber.

The folding trajectory of Gβ₅

In the closed state, the consensus map of Gβ₅ contains density of a partially folded β-propeller with variable local resolution (Figure S2F). The presence of partially ordered Gβ₅ density prompted us to investigate if the reconstruction represented a mixture of different conformational states. In order to probe for conformational heterogeneity, we applied 3D variability analysis as implemented in cryoSPARC⁴³ using a focused mask over the folding chamber of closed CCT (Figure S5). The principal component exhibiting the greatest degree of variability revealed heterogeneity within Gβ₅ densities. Subsequent focused classification over Gβ₅ revealed an ensemble of 16 distinct classes that represent a range of Gβ₅ folding intermediates at variable local resolutions (Figures 2A and S6). The classes were assembled into a progression of Gβ₅ folding based on increasing completeness and improved local resolution of Gβ₅ density (classes 0–15), beginning with a molten globule state with poor local resolution (class 0) to a reconstruction with high local resolution that enabled the modeling of a fully folded β-propeller (class 15) (Supplemental Movies 1, 2). Each class represents a distinct snapshot of Gβ₅ folding, and models of folding intermediates were built into charge density that displayed sufficient local resolution for the separation of β-strands. The ordering of classes reveals a trajectory that begins with the migration of unfolded Gβ₅ from the equatorial to apical end of CCT, and the ordering of snapshots reveals the sequence of folding for each of the seven blades of the β-propeller fold (B1-B7). Throughout the trajectory, charge density at poor local resolution is observed in multiple positions, as if the densities were shifting as a molten globule to sample various conformations without adopting an ordered state. Models were not built into these unstructured portions, although classes 1–14 show that they are connected to well-resolved segments of Gβ₅.

Figure 2. The folding trajectory of Gβ₅.

(A) Focused views of eight folding intermediates reveal the stepwise folding process of Gβ₅ from a molten globule state (class 0) to a complete β-propeller (class 15). Well resolved Gβ₅ regions were determined by local resolution heat maps (left) and models are shown as colored ribbons (right), beginning in class 1. Unmodeled densities represent unfolded/disordered segments of Gβ₅ (red densities in local resolution maps). See Figure S6 and Supplemental Movie 1 for views of all sixteen classes. (B) Focused views of four classes demonstrate that B2 adopts a range of extended conformations during Gβ₅ folding. These extended conformations may be caused by slow folding of 3₁₀ helix residues (yellow). In the final step (class 15), B2 refolds into a canonical β-sheet and creates space for B1 to form and complete the β-propeller. (C) Clashes (blue lines) between B2 in class 14 and B1 in class 15 indicate that B2 sterically hinders B1 folding. See also Figure S5, S6, and Table 2.

In the first snapshot, Gβ₅ is poorly resolved as a molten globule in a position as if it were just released from the N-terminal arms of CCT5 and CCT7 upon CCT closure (class 0). In the ensuing class, the globule migrates towards the apical region and the outermost β-strand of B4 initiates contact with CCT5 (class 1). This blade can be attributed to B4 because it maintains this position in all the classes up to the completely folded β-propeller (class 15) in which the position of each blade can be unambiguously modeled based on the fit of the sequence into the charge density map, including the 3₁₀ helix that is unique to B2. The class 1 snapshot is followed by four classes that reveal the concurrent formation of B3 and B4 into complete four-stranded β-sheets (classes 2–5). Following B3 and B4, folding progresses radially around the propeller such that B2 and B5 begin to form (classes 5–11), followed by B6 (classes 12, 13), and B7 (class 14). Among classes 5 through 14, B2 does not assemble into a β-sheet and is instead stabilized in a range of extended conformations while B5, B6, and B7 fold sequentially. In the final class, B2 is rearranged into its native conformation and the formation of B1 seals the edges of the β-propeller and completes the folding process (class 15).

Folding is guided by stabilizing intra-molecular interactions within Gβ₅

As described above, B2 is observed in a range of extended conformations during the sequential folding of B5, B6, and B7 (Figure 2B). These classes account for approximately 53% of particles in the CCT closed state, which suggests that this incomplete B2 intermediate is stabilized during Gβ₅ folding. Interestingly, the fully folded B2 contains a 3₁₀ helix that is not found in other Gβ blades and isoforms (residues 183–187, sequence: NMAAK)³², and the folding of this motif may be slower than B5-B7. Consistent with this possibility, the extended conformations observed in B2 (classes 8–14) all display an incompletely folded 3₁₀ helix. In these conformations, B2 is partially stabilized through the hydrogen-bond contacts with the β-sheet of the neighboring B3; however, the B2 residues N-terminal to the 3₁₀ helix occupy positions that sterically obstructs the formation of B1 (residues 172–182, Figure 2C). Following the formation of B7, B2 undergoes a conformational change and reorganizes into its native β-sheet, which enables B1 to fold and complete the folding process of the β-propeller (class 15). The stabilization of B2 during the folding trajectory of Gβ₅ suggests a mechanism that delays the temporal folding of B1. By occluding the space needed for B1 formation, this ensures that all other blades have folded before B1 seals the edges to complete the β-propeller.

Folding is initiated and guided by specific interactions between CCT and Gβ₅

The first interaction between Gβ₅ and CCT is mediated by the formation of the outermost β-hairpin that initiates the folding of B4 (class 1, Gβ₅ residues 267–274). Interestingly, R269 within this hairpin is conserved in all Gβ proteins and is the only surface-exposed basic residue within a ~6,450 Ų patch of folded Gβ₅ (Figure 3A). This residue forms a complementary electrostatic interface with a strong acidic patch in CCT5 (Figure 3B). The interaction appears highly specific as the charge density map in class 1 reveals an electrostatic interaction between R269 of Gβ₅ (B4) and E256 of CCT5 (Figure 3C). The β-hairpin also appears to be stabilized by a hydrogen-bond between D267 of Gβ₅ with the indole nitrogen of W324 (CCT5) and a hydrophobic contact between V274 of Gβ₅ with the same tryptophan residue (Figure 3C). No other interactions between Gβ₅ and CCT were observed in this class, suggesting that the initiation of folding is driven by the three contacts between Gβ₅ and CCT5 (R269-E256, D267-W324, and V274-W324).

Figure 3. Gβ₅ folding is initiated by specific contacts between B4 and CCT5.

(A) The electrostatic potential map of Gβ₅ reveals a single surface-exposed basic residue (R269) within B4 (outlined) and surrounding regions. Red, surface-exposed acidic residues; blue, surface-exposed basic residues. (B) The electrostatic potential map of CCT5 reveals a surface-exposed negatively charged pocket that accommodates R269 of Gβ₅. The interaction is also further stabilized by additional contacts involving D267 and V274. (C) A series of electrostatic, hydrogen-bond and hydrophobic interactions stabilize the position of the outermost β-hairpin of B4 with CCT5. (D) The R269E mutation increases Gβ₅ association with CCT with or without RGS9. Note that the R269E charge reversal slows the migration of the variant in SDS gels compared to Gβ₅ WT (n = 6). (E) The R269E mutation reduces Gβ₅ association with RGS9 (n = 4). In D and E, bars represent the average ± standard deviation normalized to the WT Gβ₅ sample. Red *** p < 0.001, **** p < 0.0001 in t-tests compared to WT Gβ₅.

To further assess the contribution of the salt bridge between Gβ₅ R269 and CCT5 E256 to Gβ₅ folding, we expressed a Gβ₅ R269E variant in cells and measured its ability to bind CCT with or without RGS9 (Fig. 3D). Gβ₅ R269E showed more than a 2-fold increase in binding to CCT compared to WT, and this effect was unchanged by co-expression with the RGS9 cognate binding partner. This increase in CCT binding is consistent with a model in which substrate folding is impaired, leading to the failure of substrate to release from CCT. In further support of this model, Gβ₅ R269E binding to RGS9 was decreased 5-fold compared to WT (Figure 3E). The reduced binding caused by R269E is unlikely to be caused by disruption of the Gβ₅ and RGS9 binding interface. The interface is extensive, covering 8,750 Ų and 92 residues of Gβ₅, including R269³². However, the R269 contribution to the interface appears negligible, consisting of two van der Waals contacts between its β carbon and side-chain carbons of V238 and S243 in the Gγ-like domain of RGS9³². Taken together, it appears that the R269E substitution disrupts CCT-mediated Gβ₅ folding, resulting in failure to mature into a state that can release from CCT to interact with RGS9. These findings therefore support the idea that the salt bridge between Gβ₅ R269 and CCT5 E256 initiates folding.

Additional electrostatic and hydrogen-bonding interactions that stabilize folding intermediates are observed in subsequent classes. Specifically, B4 is further anchored to CCT5 through hydrogen-bonds between S248 of Gβ₅ and K259 of CCT5 (Figure 4A). As other blades fold radially around B4, additional interactions are observed between B5 (R311) and CCT7 (E245/D297) (Figure 4B). Hydrogen-bonding interactions are also observed between main chain carbonyls in B3 (L229, L230) and the side chain of H314 in CCT2 (Figure 4C). Additional hydrogen bond interactions are observed between B5 with CCT5 and between B3 with CCT4 in class 15 (Figure 4D). In contrast, no Gβ₅-CCT interactions are observed between B1, B2, B6, and B7. Taken together, the stabilization of B3 with CCT2, B4 with CCT5, and B5 with CCT7 through polar interactions likely seeds the folding of Gβ₅ to enable the sequential formation of each blade, beginning with B4 (Figure 4D). This extensive hydrogen bond and electrostatic interaction network in the Gβ₅-CCT interface indicates that CCT stabilizes Gβ₅ folding through hydrophilic interactions with surface residues, leaving the hydrophobic core free to collapse into its folded conformation.

Figure 4. Gβ₅ folding is stabilized by a network of interactions between B3–5 and CCT2, 5, and 7.

Focused views of specific CCT interactions with Gβ₅ folding intermediates. (A) Hydrogen bond interaction observed between S248 of Gβ₅ and K259 of CCT5 in class 5. (B) Salt bridge formed between B5 (R311) and CCT7 (E245) in class 6. (C) Hydrogen bond formed between CCT2 (H314) and backbone carbonyls of B3 (L229, L230) in class 2. All interactions are maintained in subsequent classes. (D) Hydrogen-bond network formed between B3, B4, and B5 with specific CCT subunits in the fully folded state of Gβ₅ (class 15). All hydrogen bond interactions are shown as blue lines.

PhLP1 spatially confines Gβ₅ to a defined region within CCT

The two copies of the PhLP1 co-chaperone also relocate when CCT closes. One copy is no longer found in the structure, presumably because it has dissociated from CCT, while the other copy dissociates from its position on the apical domain of CCT6 to occupy the folding chamber opposite of Gβ₅ (referred to as the CCT′ chamber). The local resolution permitted building and refining an atomic model of PhLP1 encompassing residues 55–93 and 136–301, which include two extended helices (H1 and H2), followed by a conserved thioredoxin domain (TXDN), and ending with a structured C-terminal tail (Figures 5A and S7A). Contacts between PhLP1 and CCT are mediated by a network of polar interactions that span a large surface area across multiple CCT subunits (Figure 5B). H1 (residues 57–68) lies against CCT1′, 3′, and 4′, while the C-terminal tail of PhLP1 is highly acidic and extends through a deep, positively charged groove between CCT1′ and CCT4′ and into the exterior of CCT. H2 and the TXDN are well resolved, and the linker between these two regions associates with CCT1′, CCT3′, and CCT6′ through H-bonds.

Figure 5. PhLP1 forms an extensive network of interactions with CCT.

(A) Domain organization of PhLP1. N, negatively charged N-terminal motif; H1, helix 1; H2, helix 2; TXDN, thioredoxin domain; C, negatively charged C-terminal motif. (B) Cut-away view of PhLP1 (pink) interactions with CCT. Select zoomed-in views of PhLP1-CCT interactions shown: N-terminal negatively charged residues with CCT1 and CCT4 (top left), H2-TXDN linker with CCT3′, CCT1′, and CCT6′ (middle left), C-terminal negative charged residues with CCT1′ (bottom middle), and H1 with CCT1′ and CCT4′ (top right). Hydrogen bond interactions between labeled residues shown as blue lines (distances measured between 2.7–3.3 Å). Unfolded Gβ₅ density is colored light blue. (C) The closed state of CCT at the CCT1/4 interface is stabilized by complementary charge interactions from the negatively charged N- and C- motifs in PhLP1 (electrostatic surfaces shown, PhLP1 surfaces are outlined). See also Figure S7.

PhLP1 shares sequence and structure homology to the PhLP2A co-chaperone, the latter of which is best characterized for its role in CCT-mediated folding of actin. Interestingly, PhLP1 and PhLP2A interact with different components of CCT and are positioned in different regions relative to substrate (Figure S7B). While PhLP2A interacts with CCT primarily via its TXDN domain³⁷, the same domain in PhLP1 does not appear to directly contact CCT. In contrast, the position of PhLP1 is stabilized through an extensive network of interactions that span multiple CCT subunits, thereby increasing the avidity of PhLP1 binding and restraining its position to a defined region within CCT. The interactions between co-chaperones with substrate are also dissimilar. PhLP2A contributes an N-terminal α-helix that extends into the other folding chamber to bind actin³⁷, whereas an ordered interaction between PhLP1 and Gβ₅ is not observed.

Density for PhLP1 becomes disordered towards the N-terminus of H1 as it traverses into the Gβ₅-containing half of CCT, in proximity to unfolded Gβ₅ (Figure 5B). However, density emerges at the CCT1/4 interface approximately 11 Å from the unfolded Gβ₅ density. The interface is structurally equivalent to the opposing CCT1′/4′ interface that accommodates the acidic PhLP1 C-terminal tail. Interestingly, a series of acidic and phosphorylated residues near the N-terminus of PhLP1 (residues 18–29)¹⁶ are predicted to form an α-helix, and structural prediction algorithms indicate that this helix interacts with CCT1 in a position that overlaps with the observed density (Figure S7C). Based on this modeling, it appears that PhLP1 N- and C-termini span between the positively charged CCT1/4 interface in one folding chamber to the equivalent groove in the opposite chamber (Figure 5B). This conformation would indicate that PhLP1 passes around unfolded Gβ₅, although no direct interaction between PhLP1 and Gβ₅ is observed likely due to poor local resolution of the unstructured substrate. The PhLP1-CCT interactions also appear to stabilize the CCT1/4 interface, which otherwise would be expected to encounter charge repulsion from the strongly basic patches in CCT1 and 4 (Figure 5C). Interestingly, this same interface anchors the acidic E-hook of tubulin, which may explain why tubulin does not need a PhLP co-chaperone within the CCT chamber to facilitate folding³⁶.

To understand the functional significance of this extended PhLP1 structure, we prepared a series of PhLP1 truncations and measured their effect on Gβ₅ binding to CCT and the formation of Gβ₅-RGS9 dimers in co-immunoprecipitation experiments (Figure 6). Full-length PhLP1 increased Gβ₅ binding to CCT several fold compared to Gβ₅ alone, as expected¹⁹. In contrast, co-expression of RGS9 with PhLP1 led to a reduction in binding to CCT as Gβ₅ releases upon dimer formation with RGS9 (Figures 6B). The progressive truncation of PhLP1 N-terminal residues up to PhLP1 Δ129 caused a striking increase in Gβ₅ binding above that of full-length PhLP1 and a proportionate decrease in Gβ₅ binding to RGS9 (Figures 6B, C). The effects of the PhLP1 Δ136 truncation were reduced presumably because this truncation disrupted binding to CCT (Figure 6B, PhLP1 blot), which is consistent with previous work showing that PhLP1 residues 130–136 contributed to binding the open form of CCT⁴⁴. In contrast to the N-terminal deletions, the truncation of the C-terminal tail of PhLP1 that extends from the thioredoxin domain (residues 277–301) did not change Gβ₅ binding to CCT or RGS9 compared to full-length PhLP1, suggesting that the interaction of the PhLP1 C-terminal tail at the CCT1′/4′ interface does not influence Gβ₅ folding by CCT in the opposite chamber. Collectively, the increased Gβ₅ binding to CCT and the increased Gβ₅-RGS9 dimer formation with full-length PhLP1, coupled with the additional increases in Gβ₅ binding to CCT and the decreases in Gβ₅-RGS9 dimer formation with the N-terminal truncations, suggest a dual effect of PhLP1 on Gβ₅ folding in which the thioredoxin domain enhances Gβ₅ binding to CCT while the N-terminal extension mediates its release to interact with RGS9 (see Discussion).

Figure 6. Role of PhLP1 domains in Gβ₅ binding to CCT and RGS9.

(A) Domain organization of PhLP1 with truncation sites indicated. (B) Effects of co-transfection of full-length PhLP1 and several truncation variants on Gβ₅ co-immunoprecipitation (co-IP) with CCT or (C) RGS9. (D) Effects of co-transfection of the CK2 phosphorylation site variant S18–20A on Gβ₅ co-IP with CCT or (E) RGS9. Bars represent the average ± standard deviation normalized to the Gβ₅ alone sample (n = 3). Blue asterisks * mark co-IP bands of WT PhLP1 or PhLP1 truncations in CCT IPs with arrowheads indicating non-specific bands. Green asterisks * p < 0.05, *** p < 0.001 in t-tests compared to Gβ5 alone. Red asterisks * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001 in t-tests compared to the Gβ₅, PhLP1, RGS9 sample.

The N-terminal extension of PhLP1 harbors a protein kinase CK2 phosphorylation site at three consecutive serine residues (S18–20) that is essential for Gβγ assembly^15,16. Interestingly, these residues are part of the acidic stretch that interacts with the basic interface between CCT1 and 4. To assess the contribution of phosphorylated S18–20 to Gβ₅ folding, we measured the effect of a PhLP1 S18–20A alanine substitution variant on Gβ₅ binding to CCT and RGS9 (Figures 6D–E). This variant increased Gβ₅ binding to CCT and decreased binding to RGS9 almost to the same extent as the N-terminal Δ75 truncation, indicating that the phosphorylated S18–20 contributes significantly to Gβ₅ folding and Gβ₅-RGS dimer formation.

Discussion

The CCT complex is an essential and abundant molecular machine that is responsible for folding a diverse clientele of substrates, yet the underlying mechanism by which it facilitates folding is only now starting to be resolved for a small subset of CCT clients^36,37,42. Here, we isolated CCT-PhLP1 particles in the act of folding Gβ₅, a core component of the R7 family of RGS signaling complexes and performed cryo-EM and image processing to reveal its folding mechanism. The results enabled detailed molecular modeling for Gβ₅ folding from its unfolded, molten globule state to a fully folded β-propeller (Figures 7A, B and Supplemental Movies 1, 2). Our models establish that unfolded Gβ₅ initially associates with the open form of CCT in a space between the rings and is tethered by interactions with the hydrophobic N-terminal tails of CCT5 and CCT7 from both rings. ATP binding and progression to the ATP hydrolysis transition state drives a conformational change that brings the apical domains together to close the folding chamber and withdraws the N-terminal tails against the equatorial domains, thus releasing the unfolded Gβ₅ from between the CCT rings into the folding chamber.

Figure 7. Model of Gβ₅ folding by CCT.

(A) Beta-sheet topology of Gβ₅ depicting directionality of folding and interactions between B3, B4, and B5 with CCT2, 5, and 7, respectively. Schematic is not drawn to scale. (B) Model figure depicting key steps in CCT-mediated folding, from encapsulation of the unfolded protein in the open state, to progressive folding in the closed state, and the release of folded protein to interact with RGS9 and form the functional Gβ₅-RGS9 dimer. (C) Structures of CCT with different folding substrates: tubulin (left, PDB 7TUB), actin (middle, PDB 7NVM) and Gβ₅ (right, this study). CCT subunits are labeled and colored based on electrostatic surface charge within the inner wall (red, negatively charged surfaces are labeled; blue, positively charged surfaces).

Cryo-EM image processing revealed an ensemble of snapshots that capture the stepwise Gβ₅ folding process, beginning with the migration of the unfolded protein to the apical end of CCT2, 5, and 7. The folding process is initiated by specific interactions between the outer β-hairpin of B4 and the inner surface of the CCT5 apical domain, and folding continues radially around the β-propeller as hydrophilic interactions between the outer strands of B3 and B5 with adjacent CCT subunits stabilize their β-sheets. B2 initially adopts an intermediate extended conformation that interacts with the neighboring β-sheet of B3. B6 and B7 then fold without contacting CCT, with their folding presumably stabilized by the other folded blades. The folding of Gβ₅ is completed with the re-organization of B2 into its native β-sheet and the closing of the β-propeller through the formation of B1. This fully folded Gβ₅ is presumably then able to bind its RGS binding partner when CCT reopens upon release of the hydrolyzed phosphate.

Our structures reshape the understanding of CCT-dependent protein folding. Based on similarities to the type 1 chaperonin GroEL, previous models have postulated that CCT functions like an “Anfinsen cage” in which an unfolded substrate would bind to the chamber walls of the open form and then release into the space within the CCT folding chamber where it becomes encapsulated in the closed conformation^45–47. According to this model, CCT plays a passive role as the substrate folds spontaneously within the protected environment. In contrast, our work on Gβ₅ and the recent work of others on tubulin^36,42 reveals that CCT actively directs folding by forming specific interactions to initiate folding and stabilize folding intermediates. Whether CCT actively folds all substrates or whether a subset can be folded passively remains to be seen.

It is notable that folding appears to be initiated between a solvent-exposed β-hairpin of Gβ₅ with the interior surface of CCT5. Such a minimal cluster of contacts raises the question of how a segment comprising just eight residues (D267 through V274) defines the folding initiation of Gβ₅. Three distinct contacts are observed at this interface: an electrostatic interaction between R269 (Gβ₅) and E256 (CCT5), a hydrogen-bond between D267 (Gβ₅) and the indole nitrogen of W324 (CCT5) and hydrophobic packing between V274 (Gβ₅) and W324 (CCT5). Interestingly, D267 and R269 are conserved in all five isoforms of Gβ subunits (Gβ1–5), suggesting that these residues play a central role in ensuring proper folding. R269 stands out as the only surface-exposed positively charged residue within a large swath of Gβ₅, and the residue fits precisely in a pronounced negatively charged groove within CCT5 (Figures 3A, B). It is therefore plausible that the β-hairpin containing R269 samples a range of conformations in the molten globule state of Gβ₅ before its stabilization by E256 of CCT5 and other proximal interactions lock this intermediate in place to enable folding to progress. Consistent with this idea is our observation that the R269E mutation impairs Gβ₅ folding (Figure 3D, E).

A recent study examining CCT-dependent tubulin folding reported that folding is mediated by electrostatic interactions and hydrogen bonds between CCT and surface residues of tubulin, guiding the folding of individual domains that then consolidate into the native state³⁶. Interestingly, the interactions between CCT and tubulin are distinct from those of Gβ₅ observed in our study (Figure 7C). Negatively charged surface residues of tubulin associate with the inner walls of the predominantly positively charged CCT hemisphere encompassing CCT1, 3, 6 and 8^36,42. With Gβ_5, the outer blades of the β-propeller also form salt bridges and hydrogen bonds but do so with the inner walls of the predominantly negative charged hemisphere of CCT2, 5 and 7. Together, these findings establish a new paradigm for CCT-dependent protein folding in which CCT in the closed form presents a variety of electrostatic and hydrophilic binding sites that make substrate-specific contacts with surface residues that stabilize folding intermediates and allow the hydrophobic cores to coalesce into the native structure.

The CCT-PhLP1-Gβ₅ structures and the PhLP1 truncation data help explain the dual role of PhLP1 in Gβ₅ folding, with the PhLP1 C-terminal thioredoxin domain increasing Gβ₅ binding to CCT and the N-terminal α-helical domain enhancing release of Gβ₅ from CCT. The structures suggest that the thioredoxin domain acts allosterically by binding to the apical domain of CCT6 in a position that would inhibit CCT closure (Figure 1B), thus trapping Gβ₅ between the rings in an unfolded state. Evidence supporting this mechanism comes from the intermediate state of CCT in which the high ATPase hemisphere (CCT2, 4, 5 and 7) adopts a closed conformation while the low ATPase hemisphere (CCT1, 3, 6 and 8) remains in an open state and bound to the PhLP1 thioredoxin domain. In contrast, in the closed conformation the N-terminal domain forms the long H1 α-helix that traverses between the CCT rings into the Gβ₅ folding chamber (Figure 5). This positioning of H1 clashes with the location of Gβ₅ between the CCT rings in the open state. Thus, H1 may help displace Gβ₅ from between the rings as the chamber closes. Moreover, the PhLP1 N-terminal domain extends across the folding chamber containing Gβ₅ to the positively charged groove at the CCT1/4 interface where a sequence of three phosphorylated serine residues on PhLP1 (S18–20) bind (Figure 5). Notably, previous work has shown that phosphorylation of these sites is required for efficient Gβγ assembly^15,16, and our results extend this observation to Gβ₅-RGS9 assembly (Figures 6C, D). The location of phosphorylated S18–20 in the positively charged groove requires that the PhLP1 N-terminal extension traverse the folding chamber, possibly wrapping around Gβ₅ and constraining it to the hemisphere encompassing CCT2, 4, 5 and 7 where folding of the β-propeller occurs.

These observations help explain the dual role of PhLP1 in first stabilizing and then releasing Gβ₅ from CCT. However, they do not effectively address the observed specificity of PhLP1 in which it is only known to facilitate the folding of Gβ substrates⁴⁸. Some specificity could be achieved by the geometric constraints created by PhLP1 traversing the CCT hemisphere that initiates Gβ₅ folding, but such specificity may require direct interactions between PhLP1 and Gβ₅ that were not observed in our structures. A conserved Gβ binding site on PhLP1 is found in H1 residues 57–70⁴⁰, and we have previously observed interactions between PhLP1 and a Gβ₅γ₂ dimer¹⁹. However, the 57–70 region associates with the N- and C-terminal tails of the CCT subunits as it passes between the CCT rings and does not contact Gβ₅ as it folds in the apical region of the closed structure (Figure 5B). Thus, further studies will be needed to fully identify the source of the specificity of PhLP1 for Gβ substrates.

Visualizing protein folding has been a major challenge in structural biology⁴⁹. However, recent advances in cryo-EM image processing that enable structural determination in a continuous array of functional states have allowed us to elucidate the CCT-mediated folding trajectory of the G protein β₅ subunit. This work establishes a structural framework to investigate the effects of pathogenic mutations in Gβ₅ that are associated with protein misfolding^50–52 and to determine the folding trajectories of other β-propeller proteins, including the other Gβ subunits, the mLST8 and Raptor subunits of mTOR complexes, and the BBS7 subunit of the BBSome complex, among others^20,39,53. With various CCT substrates now known to bind to different CCT subunits in their closed state, it appears likely that substrates have co-evolved with CCT to preserve optimal binding sites to initiate folding.

Limitations of the Study

This study does not address the mechanism by which Gβ₅ is recruited to and released from CCT. Gβ₅ may interact with CCT co-translationally, as has been reported for several yeast β-propeller proteins⁵⁴, or it may be delivered to CCT by a yet to be identified co-chaperone as is the case for actin and tubulin^36,38. Moreover, the mechanism by which RGS9 associates with Gβ₅ after folding by CCT remains to be clearly defined, although our study establishes the need for the PhLP1 N-terminal domain in this process.

STAR Methods

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Peter S. Shen (peter.shen@biochem.utah.edu).

Material Availability

All materials are available upon reasonable request to the lead contact, Peter S. Shen (peter.shen@biochem.utah.edu).

Data and Code Availability

• Cryo-EM reconstructions and atomic models have been deposited in the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB), respectively. The chemical crosslinking mass spectrometry data have been deposited to the MassIVE repository. Accession numbers for deposited structures and crosslinking mass spectrometry data are listed in the Key Resources Table. Original immunoblot images have been deposited at Mendeley Data and are publicly available as of the date of publication.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-CCT2 (Rabbit)	Abcam	Cat#Ab92746
anti-CCT5 (Rat)	Bio-Rad	Cat#MCA2178
anti-c-myc (Mouse)	Invitrogen	Cat#13-2500
anti-GFP (Rabbit)	Abcam	Cat#Ab6556
anti-HA (Rat)	Sigma-Aldrich	Cat#1186723001
anti-Strep (Mouse)	Genscript	Cat#A01732
IRDye 680RD Goat anti-Mouse IgG	LI-COR	Cat#926-68070
IRDye 800CW Goat anti-Rat IgG	LI-COR	Cat#926-32219
IRDye 800RD Goat anti-Rabbit IgG	LI-COR	Cat#926-32211
Chemicals, Peptides, and Recombinant Proteins
Acetonitrile	Fisher Scientific	Cat#A998-1
Aluminum nitrate	Thermo Fisher Scientific	Cat#12360
Ammonium bicarbonate	Sigma-Aldrich	Cat#A6141-25G
AspN protease	Millipore Sigma	Cat#P3303
ATP disodium salt hydrate	Sigma-Aldrich	Cat#A2383
Blasticidin	Invivogen	Cat#ant-bl-1
CCT-PhLP1-Gβ5	This study	N/A
CHAPS	Thermo Fisher Scientific	Cat#28300
D-desthiobiotin	Sigma-Aldrich	Cat#D1411-1G
DMEM/F-12 50/50	Corning	Cat#10-092-CV
FBS	Sigma-Aldrich	Cat#F0926-500ML
Formic acid	Fisher Scientific	Cat#UN1779
Guanidine hydrochloride	Sigma-Aldrich	Cat#G4505-100G
Halt protease inhibitor cocktail	Thermo Fisher Scientific	Cat#78439
HEPES	Sigma-Aldrich	Cat#H3375-500G
IGEPAL	Sigma-Aldrich	Cat#I8896-50ML
Imidazole	Sigma-Aldrich	Cat#12399-100G
Iodoacetimide	Sigma-Aldrich	Cat#I6125-5G
LC-SDA crosslinker	Thermo Fisher Scientific	Cat#26168
Lipofectamine 3000	Thermo Fisher Scientific	Cat#L3000-015
Magnisum chloride	VWR	Cat#0288-500G
Opti-MEM	Thermo Fisher Scientific	Cat#31985-070
PEI	Fisher Scientific	Cat#NC1014320
Phosphate Buffered Saline	Fisher Scientific	Cat#BP399-1
PMSF	Millipore Sigma	Cat#11359061001
Potassium chloride	VWR	Cat#MK6858-04
Puromycin	Invivogen	Cat#ant-pr-5b
Sequencing grade trypsin	Promega	Cat#V5111
Sodium chloride	Fisher Scientific	Cat#S271-3
Sodium fluoride	VWR	Cat#BDH9290-500G
Sodium hydroxide	VWR	Cat#7708-06
TBS blocking buffer	LI-COR	Cat#927-60001
TCEP	Sigma-Aldrich	Cat#C4706-10G
Trifluoroacetic acid	Sigma-Aldrich	Cat#T6508-100ML
TrisHCl	Promega	Cat#H5123
Tween20	Fisher Scientific	Cat#BP337-500
Zinc nitrate	Fisher Scientific	Cat#4344-01
Deposited Data
AlphaFold Gβ5 model	AlphaFold	AF-14775-F1
AlphaFold PhLP1 model	AlphaFold	AF-Q1337-F1
CCT reference model	Kelly et al.	PDB 7NVM
Crosslinking mass spectrometery data	This study	MassIVE dataset #MSV000092662
Closed state 0	This study	EMDB: EMD-40440; PDB 8SFF
Closed state 1	This study	EMDB: EMD-40452; PDB 8S8G
Closed state 2	This study	EMDB: EMD-40454; PDB 8SGC
Closed state 3	This study	EMDB: EMD-40453; PDB 8SG9
Closed state 4	This study	EMDB: EMD-40482; PDB 8SHA
Closed state 5	This study	EMDB: EMD-40488; PDB 8SHL
Closed state 6	This study	EMDB: EMD-40489; PDB 8SHN
Closed state 7	This study	EMDB: EMD-40486; PDB 8SHF
Closed state 8	This study	EMDB: EMD-40485; PDB 8SHE
Closed state 9	This study	EMDB: EMD-40487; PDB 8SHG
Closed state 10	This study	EMDB: EMD-40484; PDB 8SHD
Closed state 11	This study	EMDB: EMD-40490; PDB 8SHO
Closed state 12	This study	EMDB: EMD-40491; PDB 8SHP
Closed state 13	This study	EMDB: EMD-40492; PDB 8SHQ
Closed state 14	This study	EMDB: EMD-40494; PDB 8SHT
Closed state 15	This study	EMDB: EMD-40461; PDB 8SGL
Best resolved PhLP1 class	This study	EMDB: EMD-40481; PDB 8SH9
Open state	This study	EMDB: EMD-40439; PDB 8SFE
Experimental Models: Cell Lines
HEK293T cells	ATCC	CRL-3216
Recombinant DNA
pcDNA3.1 2xStrep- Gβ5	This study	N/A
pcDNA3.1 GFP	This study	N/A
pcDNA3.1 PhLP1-myc-6xHis	This study	N/A
pcDNA3.1 PhLP1-Δ129	This study	N/A
pcDNA3.1 PhLP1-Δ136	This study	N/A
pcDNA3.1 PhLP1-Δ277-301	This study	N/A
pcDNA3.1 PhLP1-Δ75	This study	N/A
pcDNA3.1 PhLP1-Δ75,Δ277-301	This study	N/A
pcDNA3.1 PhLP1-Δ91	This study	N/A
pcDNA3.1 S18-20A PhLP1-myc-6xHis	This study	N/A
pLenti-Blast PhLP1-myc-6xHis	This study	N/A
pLenti-Puro 2xStrep-HPC4-Gβ5	This study	N/A
Software and Algorithms
Chimera v1.15	Pettersen et al.	https://www.cgl.ucsf.edu/chimera/
ColabFold/AlphaFold2	Jumper et al. (2021)	https://alphafold.ebi.ac.uk/
COOT v0.9.6	Emsley et al.	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
cryoSPARC v3.3	Punjani et al.	https://cryosparc.com/
LI-COR ImageQuant	LI-COR	https://www.licor.com/bio/image-studio/
Phenix v1.19.2	Afonine et al.	https://phenix-online.org/
xiFDR v2.1.5.5	Mendes et al.	https://www.rappsilberlab.org/software/xifdr/
xiSEARCH v1.7.6.7	Mendes et al.	https://www.rappsilberlab.org/software/xisearch/
GraphPad Prism v10.0.2	GraphPad Prism	https://www.graphpad.com/
Other
595 Filter paper	Ted Pella, Inc.	Cat#47000-100
Protein A/G agarose beads	Santa Cruz Biotechnology	Cat#sc-2003
C18 stage tips	Thermo Fisher Scientific	Cat#87784
HisTrap HP 5mL column	Cytiva	Cat#17524802
Strep-Tactin resin	IBA	Cat#2-1201-010
UltrAuFoil R2/2 Au200 mesh grids	Quantifoil	https://www.quantifoil.com/products/ultraufoil
Mini-PROTEAN TGX precast gels, 4–20%, 15 well	Bio-Rad	Cat#4561096

EXPERIMENTAL MODEL DETAILS

Cell lines

HEK293T cells (a female cell line CRL-3216) were purchased from ATCC and stored in liquid N₂. Cells were cultured in DMEM/F-12 supplement with 10% FBS in T-25 flasks at 37°C in 5% CO₂ (v/v). Cells were passaged to maintain confluency between 10–90% and the passage number was kept under 20.

METHOD DETAILS

Transfections

HEK-293T cells were transfected at 40–50 % confluency with 540 μl of transfection reagent containing 2 μg of an N-terminal 2x strep peptide-tagged human Gβ₅ cDNA or Gβ₅ R269E construct in the pcDNA3.1 vector and 2 μg of a C-terminal myc-His tagged human wild type PhLP1, PhLP1 truncation, or PhLP1 S18–20A cDNAs in the pcDNA3.1 vector along with 2 μg of a C-terminal HA tagged human RGS9 cDNAs in the pcDNA3.1 vector as indicated. Control transfections contained variable amounts of GFP in the same vector to make a total of 8 μg cDNA added.

Immunoprecipitation and Immunoblotting

Cells were washed in phosphate-buffered saline (PBS) and lysed in 500 μl of lysis buffer (PBS supplemented with 1% IGEPAL, 0.5 mM PMSF and Halt Protease Inhibitor Cocktail). Lysates were triturated 10 times through a 25-gauge 7/8” needle, then centrifuged at 21,000 × g for 10 minutes at 4°C to clarify the lysates. CCT was immunoprecipitated from the lysate by adding 4 μg of a CCT5 antibody (Bio-Rad MCA2178, 1:100 IP) and incubating for 30 minutes at 4°C, followed by adding 25 μl of protein A/G agarose beads for another 30 minutes at 4°C. RGS9 was immunoprecipitated from the lysate by adding 1 μg of a HA antibody (Sigma Aldrich 1186723001, 1:400 IP) following the same protocol. Immunoprecipitates were washed three times in lysis buffer, then resuspended in 30 μl of SDS-PAGE sample loading buffer. Proteins were separated by SDS-PAGE on 10% gels and transferred to nitrocellulose. The nitrocellulose was probed for 16 hrs at 4°C with the indicated primary antibodies at the following dilutions in TBS blocking buffer (LI-COR 927–60001): Strep (Genscript A01732, 1: 5,000), c-myc (Invitrogen 13–2500, 1: 1,000), CCT2 (Abcam, Ab92746, 1: 10,000), HA (Sigma Aldrich 1186723001, 1:1000), and GFP (Abcam Ab6556, 1: 5,000). Blots were wash 3 times with TBS-Tween wash buffer (50 mM TrisHCl pH 7.6, 150 mM NaCl, 50 mM NaOH, 0.05% Tween 20) and probed with appropriate IRDye-labeled secondary antibodies at 1:10,000 dilutions in TBS-Tween wash buffer: IRDye 680RD Goat Anti-Mouse (LI-COR 926–68070), IRDye 800RD Goat Anti-Rabbit (LI-COR 926–32211), and IRDye 800CW Goat Anti-Rat (LI-COR 926–32219). Blots were imaged using a LI-COR Odyssey infrared scanner, and band intensities were quantified with the LI-COR ImageQuant software.

Purification of CCT-PhLP1-Gβ₅

HEK-293T cells were cultured in DMEM/F-12 supplement with 10% FBS in T-25 flasks. At 40% confluency, cells were transduced with pLenti-Puro virus containing human Gβ5 with N-terminal 2X Strep and HPC4 tags. The cells were then treated with puromycin for one week to select for transduced cells. The selected cells were then transduced with pLenti-Blast virus containing human PhLP1 with N-terminal His₆ and Myc tags. The cells were then treated blasticidin for another week to select for transduced cells. The selected cell line was frozen in aliquots and used as a stable cell line for CCT-PhLP1-Gβ₅ purification.

HEK-293T cells stably expressing Strep-Gβ₅ and His-PhLP1 were grown to 80% confluency in T-175 tissue culture flasks, harvested and lysed in lysis buffer at 5 mL of buffer per gram of harvested cell pellet. The lysate was centrifuged at 30,000 × g for 20 minutes and filtered through a 0.45 μm filter followed by a 0.22 μm filter. The CCT-PhLP1-Gβ₅ complexes were then purified at 4° C using a tandem affinity approach. The filtered lysate was passed over a HisTrap HP 5 mL column equilibrated with 20 mM HEPES, 20 mM NaCl, 25 mM imidazole, 0.05% CHAPS, 1 mM TCEP, pH 7.5 for an hour. The column was then washed with 5 column volumes of equilibration buffer, and protein was eluted with linear gradient of 8 column volumes of 25 mM to 500 mM imidazole in equilibration buffer. Eluted fractions were analyzed by SDS-PAGE and Coomassie staining. Fractions containing PhLP1-CCT were combined and loaded onto 3 mL of Strep-Tactin resin equilibrated with 20 mM HEPES pH 7.5, 20 mM NaCl for an hour. The column was washed twice with one column volume of 20 mM HEPES pH 7.5, 150 mM NaCl, and then twice with one column volume of 20 mM HEPES pH 7.5, 20 mM NaCl. The CCT-PhLP1-Gβ₅ complexes were eluted with 3 column volumes of 20 mM HEPES pH 7.5, 20 mM NaCl, 0.05% CHAPS, 10 mM D-desthiobiotin, 1 mM TCEP, and 5 mM MgCl₂ and concentrated to 1.5 μg/μL using a 30 kDa cutoff filter and analyzed by SDS-PAGE and Coomassie staining and immunoblotting.

Endogenously expressed human CCT complex without Gβ₅ or PhLP1 was extensively purified from HEK-293T cells as described previously³⁹ and was also analyzed by SDS-PAGE and Coomassie staining and immunoblotting. To assess the functional state of this PhLP1-free CCT and the CCT-PhLP1-Gβ₅ complex, their ATPase activity was measured using the Malachite Green Phosphate Assay Kit (Sigma-Aldrich MAK307). This assay measures inorganic phosphate released upon ATP hydrolysis. ATP at 0.25 mM final concentration was added to freshly prepared CCT complexes (0.5 μM final concentration) in 20 mM HEPES, 20 mM NaCl, 5 mM MgCl₂, pH 7.5 in a total volume of 80 μl. ATP hydrolysis was allowed to proceed for 15 min. at 23°C, after which 20 μl of the Malachite Green reagent was added and incubated for an additional 15 min. at 23°C and the absorbance was measured at 640 nm.

Chemical crosslinking coupled with mass spectrometry

Purified CCT-Gβ₅-PhLP1 complex (50 to 150 μg) was diluted to 1 μg/μL in crosslinking buffer (20 mM HEPES, 20 mM NaCl, 5 mM MgCl₂, pH 7.5). NHS-Long chain-diazirine (LC-SDA, 150 μg) was dissolved in 6 μL acetonitrile and 144 μL crosslinking buffer to make a 1 mg/mL solution. The LC-SDA was added to the purified complex at 0.22 μg LC-SDA for every 1 μg of protein and the mixture was incubated in the dark at room temperature for 1 hr. The reaction was quenched with 50 mM TrisHCl. The sample was then loaded onto the lid of an Eppendorf tube and incubated on ice under UV light (365 nm source) 9 cm from the source for 30 min at 200 mJ/cm³. The crosslinked sample was denatured with 100 μL of 6M guanidine and then concentrated on a 10 kDa cutoff filter to a volume of 100 μL to remove peptide fragments. The sample was alkylated by adding iodoacetamide to 10 mM and incubated 15 min in the dark at room temperature. The sample was reduced by incubating for 30 min at 37 °C with 5 mM TCEP. Ammonium bicarbonate (400 μL of 500 mM) was added to quench the reaction. The sample was then digested overnight at 37°C with 0.5 μg of trypsin. After 16 hr, the digestion was stopped with 10% trifluoroacetic acid, and the sample was desalted using C18 stage tips equilibrated with 5% trifluoracetic acid. The sample was run over the column 5 times and washed twice with 100 μL of 5% trifluoroacetic acid. To elute, 50 μL of 70% acetonitrile was passed over the column 3 times, allowing the second elution volume to sit on the column for 1 min before eluting. The eluted protein was collected in an Eppendorf tube and vacuum concentrated to dryness. Sample was resuspended in 50 μL of AspN digestion buffer (50 mM TrisHCl, 0.5 mM zinc nitrate, pH 8.0), and 0.5 μg of AspN was added to the sample and incubated overnight at 37 °C. After 16 hr, the digestion was quenched with 10% trifluoroacetic acid. To enrich for crosslinked peptides, the sample was loaded onto a size exclusion column equilibrated with 2 column volumes of 30% acetonitrile and 0.1% trifluoroacetic acid in water. Sample was eluted with 1 column volume of 30% acetonitrile and 0.1% trifluoroacetic acid in water at a flow rate of 0.05 mL/min, collecting fractions every 100 μL. The first two high molecular weight fractions, composing the first half of the elution peak, were collected and desalted using C18 stage tips as previously described. The sample was vacuum concentrated to dryness and resuspended in 3% acetonitrile and 0.1% formic acid in water.

Samples were analyzed by LC-MS using an EASY-nlc 1200 LC system coupled to a Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Fisher) equipped with an EASY-Spray source. Mobile phase A consisted of 3% acetonitrile and 0.1% formic acid in water and mobile phase B consisted of 80% acetonitrile and 0.1% formic acid in water. Peptides were loaded into an EASY-Spray column operated at a temperature of 35°C and a flow rate of 0.3 μL/min. Peptides were separated at a flow rate of 0.3 μL/min using the following gradient: 5% mobile phase B (0–2 minutes), 2–32% mobile phase B (2–87 min), 32–42% mobile phase B (87–102 min), 42–95% mobile phase B (102–112 min), 95% mobile phase B (112–122 min), 95–2% mobile phase B (122–125 min), 2% mobile phase B (125–127 min), 2–100% mobile phase B (127–129 min), 100% mobile phase B (129–132), 100–2% mobile phase B (132–133 min), 2% mobile phase B (133–135 min), 2–100% mobile phase B (135–137 min), 100% mobile phase B (137–140 min), 100–0% mobile phase B (140–142 min), 0% mobile phase B (144–150 min).

MS data were acquired using the Orbitrap to detect both MS and MS/MS spectral scans. The data were collected in a data-dependent mode with a cycle time of 3 s. MS1 spectra were acquired at a resolution of 120,000 using a scan range of 375–1,700 m/z and AGC target of 4.0 × 10⁵ with a maximum injection time of 50 s. Monoisotopic peak determination (MIPS) was activated, and only precursors with charge states between 2 and 6 with an intensity higher than 5,000 were selected for fragmentation. Selected precursors were fragmented by CID using a collision energy setting of 30%. MS2 spectra were acquired at an AGC of 10,000 and a maximum injection time of 35 ms. Dynamic exclusion was set to 60 s after 1 count.

The crosslink search was performed using xiSEARCH v1.7.6.7⁵⁵ with the following parameters: MS accuracy 6 ppm; MS/MS accuracy 20 ppm; enzymes trypsin and AspN; exclusion of non-covalent interactions activated; missed cleavages allowed 4. Carbamidomethylation of cysteine residues was set as a fixed modification, while the following were set as variable modifications: oxidation of methionine, LC-SDA crosslinker alone (mass modification: 223.13207784 Da), hydrolyzed LC-SDA (mass modification: 213.136494 Da), LC-SDA alkylation (mass modification: 181.110279 Da), and LC-SDA oxidation (mass modification: 211.120839 Da). LC-SDA specificity for lysine, serine, threonine, and tyrosine residues, and N-terminus for the NHS reaction, and any residues for the diazirine reaction. Spectra were searched against FASTA files containing the human sequences for each of the eight CCT subunits, Strep peptide and HPC4 tagged Gβ₅, and His₆ and Myc tagged PhLP1. A FASTA file containing the reverse sequences of the target proteins was also included as a decoy database. FDR was estimated at a 5% residue pair level using xiFDR v2.1.5.5, including only unique peptide spectra match with a minimum peptide length of 5 residues, excluding consecutive peptide links, and boosting residue pairs between different proteins.

As an internal control of the reliability of the dataset, crosslink Cα distances between residues were mapped within the equatorial domains, the most structurally stable region of CCT. The maximal possible crosslinking distance for LC-SDA is 30–40 Å, which takes into account the crosslinker arm (12.5 Å), the side chain lengths (variable up to 12.6 Å) plus any protein backbone dynamics around the two crosslink sites. Greater than 80 % of the crosslink distances were less than 40Å, indicating a good quality dataset.

Inducing the closed conformation of CCT

To 10 μl of 1.5 μg/μl co-purified CCT-PhLP1-Gβ₅, 0.5 μL of 600 mM KCl, and 1.5 μl each of 10 mM Al(NO₃)₃, 60 mM NaF, and 10 mM ATP were sequentially added in rapid succession to avoid Al(OH)₃ precipitation. The complex was then incubated at 37°C for an hour and centrifuged at 16,300 × g for 1 min to remove any precipitate.

Electron cryo-microscopy specimen preparation

UltrAuFoil R2/2 Au200 mesh grids (Quantifoil) were glow discharged for 1 min on both sides at 25 mA using a Pelco easiGlow unit (Ted Pella, Inc.). 3 μL of freshly prepared CCT-PhLP1-Gβ₅ sample treated with ATP and AlF_x were applied on the grids and blotted with filter paper (595 Filter Paper, Ted Pella, Inc.) for 2.5 s using a Mk. II Vitrobot (Thermo Fisher Scientific) with a wait time of 25 s and −1 mm offset. The grid was then plunge frozen into liquid ethane and stored in liquid nitrogen until cryo-EM imaging.

Cryo-EM movies were recorded using SerialEM v3.8⁵⁶ in super-resolution mode on a 300 kV Titan Krios (Thermo Fisher Scientific) equipped with a post-GIF K3 direct detector (Gatan, Inc.). A total of 25,455 movies were recorded at a nominal magnification of 81,000X, corresponding to a super-resolution pixel size of 0.529 Å with a total dose of 40.42 e/Ų and 40 frames per movie. Cryo-EM data collection parameters are summarized in Table 1.

Table 1.

Cryo-EM Data Collection Parameters.

Data collection
Microscope	Titan Krios G3
Voltage (kV)	300
Detector	Gatan K3
Data collection software	SerialEM
Nominal magnification	81,000x
Dose rate (e−/Å2/second)	16.1
Total number of frames	40
Total electron exposure (e−/Å2)	40.42
Defocus range (μm)	−0.8 to −1.5
Pixel size (Å)	0.529 (super-resolution)
Data processing
Number of micrographs	25,455
Initial particle picked	6,077,238

Image processing

Super-resolution cryo-EM movie frames were motion corrected, dose weighted, Fourier-binned 2X, and summed using cryoSPARC Live as implemented in cryoSPARC v3.3⁵⁷. CTF parameters were determined using patch CTF estimation. The total exposures were then curated using CTF fit resolution cutoff of 5 Å.

A total of 6,077,238 particles were selected across 19,137 curated micrographs using a blob picker with a 180–220 Å particle diameter restraint. Particles were then extracted with a box size of 300 pixels and Fourier cropped to 150 pixels (2x binned). Four rounds of 2D classifications were performed, and a total of 2,373,004 selected particles were then used for six classes of heterogeneous reconstructions with input volume of opened, intermediate, and closed CCT derived from ab initio reconstructions. A total of 1,895,426 particles were sorted into the closed state, and 279,293 particles were sorted into the open state.

Further heterogeneous refinement on open particles revealed that 137,857 particles were in a canonical open state and 114,711 particles were in an intermediate, partially-closed state with resolution at 3.4Å and 3.7Å, respectively. Heterogeneous refinement followed by non-uniform refinement was then performed on open particles, which result in 104,907 particles in the open state with 3.4Å resolution.

Closed particles were subjected to another round of heterogeneous refinement to remove junk particles. A total of 1,813,701 closed state particles were sorted into well-resolved 3D reconstructions and re-extracted to the unbinned 300-pixel box size. Three modes of focused 3D variability analysis (3DVA) with 6 Å filtered resolution and a mask covering the substrate density were performed. Initial analysis for 3DVA found that the Gβ₅ density in component 0 had significant continuous changes. 5 clusters of individual intermediate states were then resolved from component 0 using 3DVA display with 6Å filtered resolution and followed by non-uniform refinement. Each cluster was further processed using focused 3D classification without realignment of particle orientations using a mask that only covers the Gβ₅ density. Focused 3D classification was performed with five classes, 6Å target resolution, and PCA initialization mode. Five classes from each cluster were resolved and subjected to non-uniform refinement.

A total of 25 classes with a resolution range between 2.8–3.0 Å were generated from focused 3D classification. Of these, 16 distinct classes were resolved with different Gβ₅ folding intermediates. These classes were ordered based on increasing completeness of Gβ₅ density (classes 0–15).

The next component, component 1, from 3DVA showed significant heterogeneity within PhLP1 density. Further 3DVA display using this component resolved five clusters with various intensities of PhLP1 densities. The class with best resolved PhLP1 density was selected for non-uniform refinement for further resolution improvement. A detailed image processing workflow is presented in Supplementary Figure S5.

Model building and refinement

The models of Gβ₅ and PhLP1 in closed CCT were refined from AlphaFold Gβ₅ model (AF-O14775-F1) and AlphaFold PhLP1 model (AF-Q13371-F1)^58,59. The initial model was docked into each intermediate state map as a rigid body using UCSF Chimera (version 1.15)⁵⁷. A detailed model was manually rebuilt and refined in Coot v0.9.6⁶⁰. The PhLP1 model was further refined using real-space refinement in Phenix v1.19.2⁶¹. Gβ₅ was further refined with its neighboring CCT subunits (PDB: 7NVM³⁷), CCT 2, 4, 5 and 7, using real-space refinement in Phenix v1.19.2. The rest of the CCT subunits were added into the refined PhLP1 and Gβ₅ model using Chimera and combined into a single CCT-PhLP1-Gβ₅ model. The complex model was then subjected to real-space refinement in Phenix with a supplement of ADP, AF3, HOH, Mg²⁺ restraint .cif files. Model refinement statistics are summarized in Table 2.

Table 2.

3D reconstruction and model refinement statistics (1/3).

Structure	closed state 0	closed state 1	closed state 2	closed state 3	closed state 4	closed state 5
EM Databank Accession ID	EMD-40440	EMD-40452	EMD-40454	EMD-40453	EMD-40482	EMD-40488
Protein Data Bank Accession ID	8SFF	8S8G	8SGC	8SG9	8SHA	8SHL
Symmetry imposed
3D classification	C1	C1	C1	C1	C1	C1
3D refinement	C1	C1	C1	C1	C1	C1
Final particles	32,666	34,714	93,128	96,135	29,630	44,618
Map resolution (Å)
FSC 0.143 (unmasked)	3.8	3.8	3.4	3.4	3.8	3.7
FSC 0.143 (masked, corrected)	3.0	3.0	2.9	2.9	3.0	3.0
Model Refinement
Initial model used for CCT (PDB code)	7NVM
Initial model used for Gβ5 (AlphaFold protein structure database)	N/A	AF-O14775-F1
Initial model used for PhLP1 (AlphaFold protein structure database)	N/A	AF-Q13371-F1
Map correlation coefficient	0.88	0.870	0.880	0.880	0.840	0.880
Model composition
Non-hydrogen atoms	66,750	66,794	66,712	65,583	67,158	67,244
Protein residues	8,667	8,672	8,665	8,522	8,717	8,731
Ligands (ADP, AlFx, Mg2+)	16	16	16	16	16	16
R.m.s. deviations
Bond lengths (Å)	0.003	0.002	0.004	0.003	0.006	0.002
Bond angles (°)	0.535	0.517	0.540	0.523	0.609	0.510
Validation
MolProbity score	1.590	1.590	1.620	1.560	1.69	1.550
Clashscore	7.12	7.140	7.060	6.670	7.29	6.870
Poor rotamers (%)	0.00	0.000	0.000	0.000	0.000	0.000
Ramachandran plot
Favored (%)	96.84	96.770	96.480	96.870	95.76	97.010
Allowed (%)	3.15	3.220	3.490	3.130	4.21	2.970
Disallowed (%)	0.01	0.010	0.020	0.000	0.02	0.020
C-beta deviations (0.25 Å)	0.000	0.000	0.000	0.000	0.000	0.000
CaBLAM outliers (%)	1.820	1.870	1.900	1.770	1.95	1.780
EMRinger Score	2.47	2.68	2.70	2.78	2.56	2.47
Structure	closed state 6	closed state 7	closed state 8	closed state 9	closed state 10	closed state 11
EM Databank Accession ID	EMD-40489	EMD-40486	EMD-40485	EMD-40487	EMD-40484	EMD-40490
Protein Data Bank Accession ID	8SHN	8SHF	8SHE	8SHG	8SHD	8SHO
Symmetry imposed
3D classification	C1	C1	C1	C1	C1	C1
3D refinement	C1	C1	C1	C1	C1	C1
Final particles	100,823	37,284	112,259	120,668	84,821	34,066
Map resolution (Å)
FSC 0.143 (unmasked)	3.4	3.7	3.4	3.4	3.5	3.7
FSC 0.143 (masked, corrected)	2.8	3.0	2.8	2.8	2.9	3.0
Model Refinement
Initial model used for CCT (PDB code)	7NVM
Initial model used for Gβ5 (AlphaFold protein structure database)	AF-O14775-F1
Initial model used for PhLP1 (AlphaFold protein structure database)	AF-Q13371-F1
Map correlation coefficient	0.890	0.870	0.890	0.880	0.880	0.880
Model composition
Non-hydrogen atoms	65,907	67,779	66,232	67,666	66,069	67,765
Protein residues	8,566	8,803	8,609	8,787	8,588	8,796
Ligands (ADP, AlFx, Mg2+)	16	16	16	16	16	16
R.m.s. deviations
Bond lengths (Å)	0.004	0.002	0.002	0.003	0.002	0.005
Bond angles (°)	0.571	0.500	0.506	0.526	0.504	0.623
Validation
MolProbity score	1.560	1.590	1.540	1.610	1.510	1.660
Clashscore	6.620	7.240	6.560	6.990	6.700	7.360
Poor rotamers (%)	0.000	0.000	0.000	0.000	0.000	0.000
Ramachandran plot
Favored (%)	96.840	96.820	96.980	96.550	97.250	96.200
Allowed (%)	3.160	3.160	3.010	3.420	2.730	3.750
Disallowed (%)	0.000	0.020	0.010	0.020	0.020	0.050
C-beta deviations (0.25 Å)	0.000	0.000	0.000	0.000	0.000	0.000
CaBLAM outliers (%)	1.770	1.910	1.780	1.790	1.710	1.900
EMRinger Score	2.96	2.41	2.83	2.84	3.07	2.75
Structure	closed state 12	closed state 13	closed state 14	closed state 15	best PhLP1 class	open state
EM Databank Accession ID	EMD-40491	EMD-40492	EMD-40494	EMD-40461	EMD-40481	EMD-40439
Protein Data Bank Accession ID	8SHP	8SHQ	8SHT	8SGL	8SH9	8SFE
Symmetry imposed
3D classification	C1	C1	C1	C1	C1	C1
3D refinement	C1	C1	C1	C1	C1	C1
Final particles	43,246	107,887	33,752	69,572	175,869	104,907
Map resolution (Å)
FSC 0.143 (unmasked)	3.7	3.4	3.8	3.5	3.2	4.1
FSC 0.143 (masked, corrected)	3.0	2.9	3.0	2.9	2.7	3.4
Model Refinement
Initial model used for CCT (PDB code)	7NVM					6QB8
Initial model used for Gβ5 (AlphaFold protein structure database)	AF-O14775-F1					N/A
Initial model used for PhLP1 (AlphaFold protein structure database)	AF-Q13371-F1					N/A
Map correlation coefficient	0.84	0.89	0.83	0.890	0.83	0.8
Model composition
Non-hydrogen atoms	68,052	68,037	68,382	67,432	67,324	60,512
Protein residues	8,834	8,831	8,873	8,762	8,742	7,881
Ligands (ADP, AlFx, Mg2+)	16	16	16	16	16	16
R.m.s. deviations
Bond lengths (Å)	0.003	0.005	0.002	0.004	0.003	0.003
Bond angles (°)	0.522	0.565	0.517	0.551	0.524	0.651
Validation
MolProbity score	1.60	1.63	1.61	1.590	1.57	2.120
Clashscore	6.76	6.59	7.31	6.560	6.23	15.360
Poor rotamers (%)	0.000	0.000	0.00	0.000	0.000	0.000
Ramachandran plot
Favored (%)	96.49	96.12	96.73	96.530	96.50	93.390
Allowed (%)	3.50	3.84	3.25	3.470	3.5	6.55
Disallowed (%)	0.01	0.03	0.02	0.000	0.000	0.060
C-beta deviations (0.25 Å)	0.000	0.000	0.000	0.000	0.000	0.000
CaBLAM outliers (%)	1.88	1.95	1.86	1.820	1.88	3.210
EMRinger Score	2.74	2.9	2.74	2.74	3.40	0.85

AlphaFold structure prediction

ColabFold as implemented in AlphaFold2 was used to predict the interaction between CCT1 and the N-terminal negatively charged sequence of PhLP1 (residues 18–29)⁶². The input sequence for CCT1 was CCT1 A178-N217, L312-S405, and the input sequence for PhLP1 N-terminal domain was S18-D29 where S18-S20 were changed to glutamates as phosphomimetics of the phosphorylated serine residues^15,16. The final input sequence was: AIKYTDIRGQPRYPVNSVNILKAHGRSQMESMLISGYALN: LKRDLKRIAKASGATILSTLANLEGEETFEAAMLGQAEEVVQERICDDELILIKNTKARTSASIILRGANDFMCDEMERSLHDALCVVKRVLES:EEEEDEDSDHED. The template mode was set to pdb70. The num_recycles was set to 24.

QUANTIFICATION AND STATISTICAL ANALYSIS

Immunoblot band intensities were quantified with the LI-COR ImageQuant software and normalized to the control band in each blot. Normalized values were averaged and p values relative to the control were calculated by unpaired t tests using the GraphPad Prism software.

Supplementary Material

Movie 1

Supplemental Movie 1. Snapshots of classes 0–15 assembled into a “flipbook” of Gβ₅ folding, Related to Figure 2. The left panel shows the local resolution of Gβ₅ and surrounding CCT density and the model built into well resolved densities. The right panel shows the color-coded blades according to the figure scheme used throughout the paper.

Movie 2

Supplemental Movie 2. Animation of Gβ₅ folding by CCT and PhLP1, Related to Figure 2. Percentages indicated represent proportion of particles assigned to each class (see Supplemental Figure S5).

Highlights.

• Structures of a native CCT-Gβ₅-PhLP1 complex in open and closed states

• Structures reveal the folding trajectory of Gβ₅ from unfolded to folded states

• Folding is guided by specific interactions between CCT and Gβ₅ surface residues

• PhLP1 displaces Gβ₅ from between the CCT rings and confines it to a defined region

Inclusion and Diversity

We support inclusive, diverse, and equitable conduct of research. Accordingly, one or more of the authors of this paper self-identifies as an underrepresented ethnic minority in their field of research or within their geographical location, and one or more of the authors of this paper self-identifies as a gender minority in their field of research.