Advertisement

Unexpected structures formed by the kinase RET C634R mutant extracellular domain suggest potential oncogenic mechanisms in MEN2A

Open AccessPublished:August 16, 2022DOI:https://doi.org/10.1016/j.jbc.2022.102380
      The RET receptor tyrosine kinase plays a pivotal role in cell survival, proliferation, and differentiation, and its abnormal activation leads to cancers through receptor fusions or point mutations. Mutations that disrupt the disulfide network in the extracellular domain (ECD) of RET drive multiple endocrine neoplasia type 2A (MEN2A), a hereditary syndrome associated with the development of thyroid cancers. However, structural details of how specific mutations affect RET are unclear. Here, we present the first structural insights into the ECD of the RET(C634R) mutant, the most common mutation in MEN2A. Using electron microscopy, we demonstrate that the C634R mutation causes ligand-independent dimerization of the RET ECD, revealing an unusual tail-to-tail conformation that is distinct from the ligand-induced signaling dimer of WT RET. Additionally, we show that the RETC634R ECD dimer can form complexes with at least two of the canonical RET ligands and that these complexes form very different structures than WT RET ECD upon ligand binding. In conclusion, this structural analysis of cysteine-mutant RET ECD suggests a potential key mechanism of cancer induction in MEN2A, both in the absence and presence of its native ligands, and may offer new targets for therapeutic intervention.

      Keywords

      Abbreviations:

      BN (Blue Native), CLD (cadherin-like domain), CRD (cysteine-rich domain), CTF (contrast transfer function), ECD (extracellular domain), EM (electron microscopy), GDF15 (growth and differentiation factor 15), GDNF (glial-cell line–derived neurotrophic factor), GFL (glial-cell line–derived neurotrophic factor ligand), GFRα (GDNF receptor α), GFRAL (GDNF receptor α-like), MEN2A (multiple endocrine neoplasia type 2A), MD (molecular dynamics), NS-EM (negative stain-electron microscopy), SEC (size-exclusion chromatography), SEC-MALS (size exclusion chromatography-coupled multiangle static laser light scattering), WT (wild type)
      Multiple endocrine neoplasia type 2A (MEN2A) is a hereditary polyglandular cancer syndrome that is typically characterized by childhood development of medullary thyroid carcinoma with a high probability of additionally suffering parathyroid and adrenal cancers. Genetically, MEN2A syndromes are caused by pathogenic germline RET variants, the most common of which contain activating mutations in one of six cysteine codons in the extracellular domain (codons 609, 611, 618, 620, 630, and 634 account for more than 98% of MEN2A cases) (
      • Mulligan L.M.
      • Eng C.
      • Healey C.S.
      • Clayton D.
      • Kwok J.B.J.
      • Gardner E.
      • et al.
      Specific mutations of the RET proto-oncogene are related to disease phenotype in MEN 2A and FMTC.
      ,
      • Eng C.
      • Clayton D.
      • Schuffenecker I.
      • Lenoir G.M.
      • Cote G.
      • Gagel R.F.
      • et al.
      The relationship between specific RET proto-oncogene mutations and disease phenotype in multiple endocrine neoplasia type 2.
      ,
      • Wagner S.
      • Zhu S.
      • Nicolescu A.
      • Mulligan L.
      Molecular mechanisms of RET receptor-mediated oncogenesis in multiple endocrine neoplasia 2.
      ). Among these, the C634F/G/R/S/W/Y mutations are both the most frequently identified and the most aggressive, being designated high risk by the American Thyroid Association (
      • Wells S.A.
      • Asa S.L.
      • Dralle H.
      • Elisei R.
      • Evans D.B.
      • Gagel R.F.
      • et al.
      Revised American thyroid association guidelines for the management of medullary thyroid carcinoma.
      ). Historically, these cancers have been associated with poor prognoses, developing and progressing throughout childhood, and have had limited treatment options. In 2020, the first highly RET-selective kinase inhibitor (Selpercatinib) was approved for the treatment of RET-mutant medullary thyroid carcinoma, showing marked and durable antitumor activity (
      • Shah M.H.
      • Sherman E.J.
      • Robinson B.
      • Solomon B.J.
      • Kang H.
      • Lorch J.H.
      • et al.
      Selpercatinib (LOXO-292) in patients with RET -mutant medullary thyroid cancer.
      ,
      • Markham A.
      Selpercatinib: first approval.
      ). Cysteine-mutant RET now represents a clinically validated drug target, and a comprehensive understanding of the molecular mechanisms of cysteine-mutant oncogenic activation of RET is required to aid in diagnosis and treatment.
      The extracellular domain (ECD) of RET comprises four cadherin-like domains (CLDs) and a cysteine-rich domain (CRD) that collectively coordinate ligand recognition and subsequent signal transduction. A variety of ligands have been identified to signal through RET, including the four glial-cell line–derived neurotrophic factor ligands (GFLs), glial-cell line–derived neurotrophic factor (GDNF), neurturin (NTRN), artemin (ARTN), and persephin (PSPN) (
      • Kotzbauer P.T.
      • Lampe P.A.
      • Heuckeroth R.O.
      • Golden J.P.
      • Creedon D.J.
      • Johnson Jr., E.M.
      • et al.
      Neurturin, a relative of glial-cell-line-derived neurotrophic factor.
      ,
      • Baloh R.H.
      • Tansey M.G.
      • Lampe P.A.
      • Fahrner T.J.
      • Enomoto H.
      • Simburger K.S.
      • et al.
      Artemin, a novel member of the GDNF ligand family, supports peripheral and central neurons and signals through the GFRα3-RET receptor complex.
      ,
      • Milbrandt J.
      • De Sauvage F.J.
      • Fahrner T.J.
      • Baloh R.H.
      • Leitner M.L.
      • Tansey M.G.
      • et al.
      Persephin, a novel neurotrophic factor related to GDNF and neurturin.
      ,
      • Trupp M.
      • Arenas E.
      • Fainzilber M.
      • Nilsson A.-S.
      • Sieber B.-A.
      • Grigoriou M.
      • et al.
      Functional receptor for GDNF encoded by the c-ret proto-oncogene.
      ,
      • Durbec P.
      • Marcos-Gutierrez C.V.
      • Kilkenny C.
      • Grigoriou M.
      • Wartiowaara K.
      • Suvanto P.
      • et al.
      GDNF signalling through the Ret receptor tyrosine kinase.
      ), and the structurally related growth and differentiation factor 15 (GDF15) (
      • Hsu J.-Y.
      • Crawley S.
      • Chen M.
      • Ayupova D.A.
      • Lindhout D.A.
      • Higbee J.
      • et al.
      Non-homeostatic body weight regulation through a brainstem-restricted receptor for GDF15.
      ,
      • Mullican S.E.
      • Lin-Schmidt X.
      • Chin C.-N.
      • Chavez J.A.
      • Furman J.L.
      • Armstrong A.A.
      • et al.
      GFRAL is the receptor for GDF15 and the ligand promotes weight loss in mice and nonhuman primates.
      ,
      • Yang L.
      • Chang C.-C.
      • Sun Z.
      • Madsen D.
      • Zhu H.
      • Padkjær S.B.
      • et al.
      GFRAL is the receptor for GDF15 and is required for the anti-obesity effects of the ligand.
      ,
      • Emmerson P.J.
      • Wang F.
      • Du Y.
      • Liu Q.
      • Pickard R.T.
      • Gonciarz M.D.
      • et al.
      The metabolic effects of GDF15 are mediated by the orphan receptor GFRAL.
      ). These RET ligands exist as soluble homodimers and their binding to RET is mediated through pairing with specific membrane-bound coreceptors, the expression of which is tissue dependent. Under normal physiology, activation occurs sequentially: the dimeric GFLs or GDF15 first recruit two copies of the GDNF receptors (GDNF α GFRα1-4) or GDNF receptor α-like (GFRAL) and this complex is further coordinated by two copies of RET, yielding a 2:2:2 hexameric tripartite complex that exhibits a “butterfly” conformation (
      • Li J.
      • Shang G.
      • Chen Y.-J.
      • Brautigam C.A.
      • Liou J.
      • Zhang X.
      • et al.
      Cryo-EM analyses reveal the common mechanism and diversification in the activation of RET by different ligands.
      ). Different ligand/coreceptor pairs result in distinct signaling outcomes: for example, the activation of the RET/GDNF/GFRα1 signaling pathway is critically involved in the maintenance of dopaminergic neurons (
      • Lin L.
      • Doherty D.
      • Lile J.
      • Bektesh S.
      • Collins F.
      GDNF: a glial cell line-derived neurotrophic factor for midbrain dopaminergic neurons.
      ), while elevated RET/GDF15/GFRAL signaling suppresses energy intake and is associated with the cancer cachexia syndrome (
      • Patel S.
      • Alvarez-Guaita A.
      • Melvin A.
      • Rimmington D.
      • Dattilo A.
      • Miedzybrodzka E.L.
      • et al.
      GDF15 provides an endocrine signal of nutritional stress in mice and humans.
      ). Although the ligands all activate intracellular signaling through dimerization of RET, there are substantial structural and conformational differences of the ligand, coreceptor, and RET in their complexes. Notably, the angles between the two “wings” of the RET complexes are approximately 60°, 105°, 108°, and 125° for the RET/GDF15/GFRAL, RET/NTRN/GFRα2, RET/ARTN/GFRα3, and RET/GDNF/GFRα1 complexes, respectively (
      • Li J.
      • Shang G.
      • Chen Y.-J.
      • Brautigam C.A.
      • Liou J.
      • Zhang X.
      • et al.
      Cryo-EM analyses reveal the common mechanism and diversification in the activation of RET by different ligands.
      ). These extracellular structural differences impact the relative conformations of the intracellular kinase domain dimerization and result in differential signaling outcomes (
      • Parkash V.
      • Leppänen V.M.
      • Virtanen H.
      • Jurvansuu J.M.
      • Bespalov M.M.
      • Sidorova Y.A.
      • et al.
      The structure of the glial cell line-derived neurotrophic factor-coreceptor complex: insights into RET signaling and heparin binding.
      ,
      • Coulpier M.
      • Anders J.
      • Ibáñez C.F.
      Coordinated activation of autophosphorylation sites in the RET receptor tyrosine kinase: importance of tyrosine 1062 for GDNF mediated neuronal differentiation and survival.
      ).
      The C634R mutation in RETCRD induces RET dimerization via a C630-C630 intermolecular disulfide bond and is one of the most common mutations in MEN2A cases (
      • Wagner S.
      • Zhu S.
      • Nicolescu A.
      • Mulligan L.
      Molecular mechanisms of RET receptor-mediated oncogenesis in multiple endocrine neoplasia 2.
      ,
      • Romei C.
      • Ciampi R.
      • Elisei R.
      A comprehensive overview of the role of the RET proto-oncogene in thyroid carcinoma.
      ). However, a structural understanding of how this dimerization leads to constitutive activation of the kinase domains has so far been lacking. Previous cellular studies have shown contradictory results regarding the response of RETC634R to ligand stimulation, with downstream phosphorylation being either increased upon the addition of GDNF (
      • Mograbi B.
      • Bocciardi R.
      • Bourget I.
      • Juhel T.
      • Farahi-Far D.
      • Romeo G.
      • et al.
      The sensitivity of activated cys Ret mutants to glial cell line-derived neurotrophic factor Is mandatory to rescue neuroectodermic cells from apoptosis.
      ,
      • Arighi E.
      • Popsueva A.
      • Degl’innocenti D.
      • Borrello M.G.
      • Carniti C.
      • Perälä N.M.
      • et al.
      Biological effects of the dual phenotypic Janus mutation of ret cosegregating with both multiple endocrine neoplasia type 2 and Hirschsprung’s disease.
      ) or unchanged (
      • Zhang M.
      • Liu Y.
      • Fu J.
      • Hu Y.
      • Sun Z.
      Functional analysis of RET with multiple endocrine neoplasia type 2.
      ). Therefore, it remains uncertain whether dimeric RETC634R is able to bind to its ligands and coreceptors extracellularly. Furthermore, if RETC634R does associate with its ligand and coreceptors, does the C630-C630 disulfide bond introduce conformational restrictions that prevent binding in the same 2:2:2 stoichiometry and “butterfly” conformation as the WT RET (RETWT)?
      In this study, we investigate the function of the RETC634R ECD oncogenic mutant and report electron microscopy (EM) models of both the dimeric RETC634R ECD mutant and its complex with GDF15/GFRAL as well as a cryo-EM model of the WT RET/GDF15/GFRAL extracellular domain complex. We report the first structural investigation of the unliganded RETC634R ECD dimer and show that the two RETC634R protomers associate with each other through the CRDs with the N-terminal domains pointing outward. We also observe and describe a unique “twin-butterfly” conformation of GDF15/GFRAL binding to RETC634R ECD in a mechanism that differs from that of the RETWT complexes and that of RETC634R/GDNF/GFRα1, forming a dodecameric tripartite complex. The results suggest the existence of divergent oncogenic activation mechanisms of the RETC634R mutant resulting from configurational changes of unliganded and liganded RETC634R extracellular domain complexes. These findings were further supported by molecular dynamics (MD) simulations, which found that the RETC634R ECD dimer can adopt a diverse range of activating conformations driven by aberrant disulfide stapling.

      Results

      Characterization of the extracellular domain of dimeric RETC634R

      The expression of correctly folded and functional human RET requires the use of mammalian expression systems and we previously reported that the addition of a cleavable C-terminal Fc tag facilitates the expression of RETC634R ECD in its dimeric form, mediated by an intermolecular C630-C630 disulfide bond (
      • Liu Y.
      • de Castro Ribeiro O.
      • Robinson J.
      • Goldman A.
      Expression and purification of the extracellular domain of wild-type humanRET and the dimeric oncogenic mutant C634R.
      ). For this study, we further optimized the expression construct (see Experimental procedures for details) and were able to isolate dimeric RETC634R ECD in high homogeneity (Fig. 1, A and B). To investigate the structural configuration of dimeric RETC634R ECD, we used single particle negative stain EM (NS-EM) to generate a three-dimensional model. With a total of 3490 particles, both the single particles and the two-dimensional (2D) class averages (Fig. 1C) showed a clear “S” shape, suggesting a rotational C2 symmetry of the model. Initial model building and further model refinement performed with C2 symmetry agreed well with the 2D classes (Fig. 1, D and E). The three-dimensional (3D) refinement of the model converged at 22 Å resolution. The dimensions of the reconstructed model of the dimeric RETC634R are 60 Å × 190 Å × 80 Å and the shape resembles a nonplanar “S”, comprising two “C”-shaped RETC634R protomers that are connected tail-to-tail at the CRDs via a C630-C630 disulfide linkage. Segmentation of the 3D map (
      • Pintilie G.
      • Zhang Junjie
      • Chiu W.
      • Gossard D.
      Identifying components in 3D density maps of protein nanomachines by multi-scale segmentation.
      ) resulted in seven components, representing 2 × CLD1/2, 2 × CLD3, 2 × CLD4 and joint density in the center for 2 × CRDs (Fig. 1D). We docked one copy of RETECD (modified from PDB ID: 6Q2J) into each half of the density map, which gave a good overall match of both size and shape and defined the location of the individual domains in the segmented maps (Fig. 1F). There are no regions of joint density other than the CRDs, indicating that none of the CLDs interact across the dimer interface. Because the C-terminal residues E623-R635 of RETECD are not resolved in the previously published structures (i.e., PDB ID: 6Q2J) (
      • Li J.
      • Shang G.
      • Chen Y.-J.
      • Brautigam C.A.
      • Liou J.
      • Zhang X.
      • et al.
      Cryo-EM analyses reveal the common mechanism and diversification in the activation of RET by different ligands.
      ), we could not define the position of the C630-C630 disulfide bond precisely. However, the measured distance between the two P622 residues of each RETC634R protomer is ∼18 Å, leaving enough space to accommodate the missing region in the docked structures.
      Figure thumbnail gr1
      Figure 1Characterization of the extracellular domain of the RETC634R dimer. A, coomassie-stained SDS-PAGE image showing the dimeric RETC634R mutant under reducing and nonreducing conditions. B, sample homogeneity assessed using SEC-MALS. C, a representative micrograph and 2D class averages of RETC634R. Scale bar for 2D class images represents 10 nm. D, top and front views of the segmented maps of the NS-EM model of RETC634R dimer. The segmented maps are colored to highlight the CLD1/2s, CLD3s, CLD4s, and CRDs in brown, off-white, orange, and gray, respectively. E, comparison of representative 2D class averages and the projection images of the reconstructed NS-EM model. The scale bar represents 10 nm. F, top and front views of the NS-EM model of RETC634R dimer docked with two RETECD (yellow and brown, modified from PDB ID: 6Q2J). An enlarged view of the density between the C-termini of RETC634Rs is shown (left side, boxed), showing the distance between the two P622 residues in the docked structures. CLD, cadherin-like domain; CRD, cysteine-rich domain; EM, electron microscopy; ECD, extracellular domain; SEC, size-exclusion chromatography; SEC-MALS, size exclusion chromatography-coupled multiangle static laser light scattering.

      The extracellular domain of RETC634R binds to its ligands through a novel mechanism

      Dimeric RETC634R is generally considered to be constitutively activated in the absence of ligand binding. However, previous studies have shown that RETC634R signaling is further upregulated in respond to GDNF stimulation (
      • Mograbi B.
      • Bocciardi R.
      • Bourget I.
      • Juhel T.
      • Farahi-Far D.
      • Romeo G.
      • et al.
      The sensitivity of activated cys Ret mutants to glial cell line-derived neurotrophic factor Is mandatory to rescue neuroectodermic cells from apoptosis.
      ,
      • Arighi E.
      • Popsueva A.
      • Degl’innocenti D.
      • Borrello M.G.
      • Carniti C.
      • Perälä N.M.
      • et al.
      Biological effects of the dual phenotypic Janus mutation of ret cosegregating with both multiple endocrine neoplasia type 2 and Hirschsprung’s disease.
      ), which may play a role in oncogenesis. It is currently unclear whether GDNF binds RETC634R in the same manner as RETWT and whether other GFL/coreceptor pairs share this activation capacity. In order to probe the mechanism of ligand-induced hyperactivation of RETC634R, we compared how RETWT ECD and dimeric RETC634R ECD bind two distinct ligand/coreceptor pairs (GDNF/GFRα1 and GDF15/GFRAL) using Blue Native (BN) PAGE. As expected, we found that RETWT forms a stable 2:2:2 tripartite complex with GDNF/GFRα1, as indicated by the formation of a major band that migrates more slowly than any of its components (Fig. 2A). Interestingly, we found that dimeric RETC634R also complexes with GDNF/GFRα1 to produce a major band that migrates similarly, suggesting that it too assembles predominantly as a 2:2:2 tripartite complex. However, RETC634R/GDNF/GFRα1 also forms a band with an even slower migratory shift, which is probably a higher-order complex (red star, lane 4). The RETWT/GDNF/GFRα1 mixture contains a substantial amount of uncomplexed component proteins unlike the RETC634R/GDNF/GFRα1 mixture, suggesting that the RETC634R/GDNF/GFRα1 complex, especially 2:2:2, may be more favored. In support of this, we found that both the 2:2:2 and higher-order RETC634R/GDNF/GFRα1 complexes are more thermally stable than the RETWT/GDNF/GFRα1 complex (Fig. 2B, left panel and Fig. S1), indicating that the C630-C630 disulfide bond has a stabilizing effect.
      Figure thumbnail gr2
      Figure 2Extracellular domain complex formation of RET/GDNF/GFRα1 and RET/GDF15/GFRAL. A, RETC634R formed a higher order complex with GDNF/GFRα1 and GDF15/GFRAL (marked with red stars) compared with RETWT in BN PAGE. The complexes of RETWT/GDNF/GFRα1, RETC634R/GDNF/GFRα1, and RETWT/GDF15/GFRAL with a molar ratio of 2:2:2 are marked with black stars. The Fc-GDF15 showed two major bands on the gel with the strongest band being the dimeric Fc2-GDF15 and the lower fainter band being the monomeric Fc-GDF15 as a result of Fc dissociation. B, stability measurement of the tripartite complexes of GDNF/GFRα1 (left panel) and GDF15/GFRAL (right panel) with RETC634R and RETWT. Above: an example temperature challenge BN PAGE image of the 2:2:2 RETWT/GDNF/GFRα1 complex (please refer to for full image). Band intensity of the complexes treated at various temperatures was quantified using ImageJ (
      • Schneider C.A.
      • Rasband W.S.
      • Eliceiri K.W.
      NIH image to ImageJ: 25 years of image analysis.
      ). Normalized band intensity was used to plot the melting curves, fitted by nonlinear regression to a sigmoidal model. C, characterization of RETC634R/GDF15/GFRAL (yellow), RETC634R (blue) using SEC-MALS. The right Y-axis represents the calculated molecular weight (MW), while the left Y-axis is the differential refractive index (dRI). They are plotted against elution volume (X-axis). The calculated average molecular weights of the peaks are plotted (black), showing MWs of 185 kDa for RETC634R dimer and 805 kDa for RETC634R/GDF15/GFRAL. D, coomassie-stained SDS-PAGE image showing purified RETC634R, RETC634R/GDF15/GFRAL, and RETWT/GDF15/GFRAL under reducing and nonreducing conditions. BN, Blue Native; GFRAL, GDNF receptor α-like; SEC-MALS, size exclusion chromatography-coupled multiangle static laser light scattering.
      RETWT forms a major complex with GDF15/GFRAL that migrates similarly to RETWT/GDNF/GFRα1, which is consistent in that both complexes are known to share a 2:2:2 tripartite composition. Surprisingly, however, we found that RETC634R/GDF15/GFRAL forms one major higher-order complex that migrates substantially more slowly than the RETWT/GDF15/GFRAL complex but similarly to the weaker band in the RETC634R/GDNF/GFRα1 sample (Fig. 2A). We found that this higher-order complex is moderately thermally stabilized (Tm = 60.6 °C) relative to the WT complex (Tm = 57.8 °C) (Fig. 2B, right panel). These results suggest that dimeric RETC634R may be conformationally restricted from forming the 2:2:2 complex with GDF15/GFRAL but not with GDNF/GFRα1 or that the higher-order species is more favored in the case of GDF15/GFRAL.
      To characterize these higher-order RETC634R complexes further, we investigated the oligomeric state of RETC634R/GDF15/GFRAL using size-exclusion chromatography-coupled multiangle static laser light scattering (SEC-MALS). Analysis of the RETC634R/GDF15/GFRAL complex gave a calculated molecular weight of 805 kDa (Fig. 2C), approximately twice the size of the heterohexameric RETWT/GDF15/GFRAL (395 kDa, Fig. S2). To investigate whether the RETWT- and RETC634R-ligand complexes have the same molar ratio of RET:GDF15:GFRAL, we compared the composition of the individual subunits using nonreducing and reducing SDS-PAGE (Fig. 2D). Under reducing conditions, the RETC634R mutant runs at the same molecular weight as RETWT, demonstrating the reduction of the C630-C630 disulfide bond. Indeed, we found that the RETWT- and RETC634R-complexes have the same molar ratio of all three components under reducing conditions. Taken together with the BN PAGE and SEC-MALS data, these results suggest that RETC634R ECD forms a complex with GDF15 and GFRAL in a stoichiometry of 4:4:4 that must exist in a different configuration from the RETWT ECD complex. Indeed, we measured that the affinity of RETC634R ECD for GDF15/GFRAL (Kd = 8.5 μM) is actually weaker than that of RETWT ECD (Kd = 3.2 μM) (
      • Liu Y.
      • de Castro Ribeiro O.
      • Robinson J.
      • Goldman A.
      Expression and purification of the extracellular domain of wild-type humanRET and the dimeric oncogenic mutant C634R.
      ) (Fig. S3). We had initially anticipated that the dimeric nature of RETC634R might offer an entropic advantage compared to monomeric RETWT in forming a 2:2:2 hexameric ligand complex and that this would instead result in a stronger binding affinity. We inferred from this loss of affinity that dimeric RETC634R is likely to be geometrically restricted from binding the ligands in the same conformation as two independent copies of monomeric RETWT, supporting the hypothesis that RETC634R forms a complex with GDF15/GFRAL through a novel mechanism.

      Structural comparison of the RETWT and RETC634R extracellular domain complexes

      In order to compare the character of the RETWT and RETC634R ligand complexes, we used EM to interrogate their structures. Accordingly, we first reconstituted the RETWT/GDF15/GFRAL complex and analyzed it by cryo-EM. We performed 3D model reconstruction with 142,083 particles of RETWT/GDF15/GFRAL using an initial model built from the 2D class averages, and the resulting refined model had an overall resolution of 8 Å (Fig. 3). Comparing our map to a previously published cryo-EM structure of the same complex (PDB ID: 6Q2J) (
      • Li J.
      • Shang G.
      • Chen Y.-J.
      • Brautigam C.A.
      • Liou J.
      • Zhang X.
      • et al.
      Cryo-EM analyses reveal the common mechanism and diversification in the activation of RET by different ligands.
      ), we observed only minor differences, most notably extra density, consistent with the additional glycosylation at certain sites. RETECD is heavily glycosylated, containing 11 glycosylation sites that influence the structure and function of RET (
      • Liu Y.
      • de Castro Ribeiro O.
      • Robinson J.
      • Goldman A.
      Expression and purification of the extracellular domain of wild-type humanRET and the dimeric oncogenic mutant C634R.
      ,
      • Bigalke J.M.
      • Aibara S.
      • Roth R.
      • Dahl G.
      • Gordon E.
      • Dorbéus S.
      • et al.
      Cryo-EM structure of the activated RET signaling complex reveals the importance of its cysteine-rich domain.
      ,
      • Kjær S.
      • Hanrahan S.
      • Totty N.
      • McDonald N.Q.
      Mammal-restricted elements predispose human RET to folding impairment by HSCR mutations.
      ). In our study, we used HEK293T cells to produce RETECD bearing near-native glycans and, at this resolution, densities of the base glycans at all the glycosylation sites could be observed (Fig. S4). Overall, two of the glycosylation site densities point toward the inside of the complex (N361 and N336) while the rest point outward.
      Figure thumbnail gr3
      Figure 3Cryo-EM 3D model of the extracellular domain complex of RETWT/GDF15/GFRAL. A, schematic of hRETECD, hGFRALECD, and Fc-tagged mature hGDF15 used in this study. B, a modified structure of RETWT/GDF15/GFRAL (PDB ID: 6Q2J) is fitted in the density. One N-acetylglucosamine (GlcNAc) is placed at each glycosylation sites to fit the density using Coot. The protein domains are colored according to the color scheme in (A). ECD, extracellular domain; GFRAL, GDNF receptor α-like; EM, electron microscopy.
      In parallel, using the same data set, a 30-Å map was generated from the previously published cryo-EM structure of RETWT/GDF15/GFRAL (PDB ID: 6Q2J) that was low-pass filtered to 60 Å and used as a reference map for subsequent 3D classification and auto-refinement. This refined map also has an overall resolution of 8 Å and is in close structural agreement with the de novo model that we reconstructed using the RELION-built initial model as reference map. Analysis of three 3D classes generated during the reconstruction revealed movement of the two “wings” of the RETWT/GDF15/GFRAL complex (Fig. 4). Class 1 showed the narrowest (89°) and widest (111°) angles along the x and y axis, respectively, while class 3 had the widest (94°) and narrowest (102°) angles. Class 2 shared similar angles with class 1 along the x (90°) and y axes (109°). To assess these changes of conformation, the WT complex structure was fitted individually into the three refined cryo-EM maps from 3D classification using rigid-body flexible fitting. The resulting structures were then superimposed along one half of the complex for structural comparison. Overall, RETECD showed the highest degree of conformational change relative to GDF15 and GFRAL (Fig. 4B). Morphing of the structures suggests that the two “wings” of the complex are flexible, showing an “out-in” bend and “front-back” twist movement (Movie S1).
      Figure thumbnail gr4
      Figure 4Conformational variations of the extracellular domain complex of RETWT/GDF15/GFRAL. A, cryo-EM models auto-refined from three classes of particles. The three models share the same configuration but vary in the angles between the two “wings” of the complex along the x (89° to 94°) and y (102° to 111°) axes. The angles were measured using the positions marked by the red circles. B, comparison of the Flex-EM modeled structures of RETWT/GDF15/GFRAL after aligning one half of the complexes. Modeled structures for class 1, 2, and 3 are colored gray, yellow, and cyan, respectively. C, analysis of the conformational change of class 3 relative to class 1 as defined by the RMSD value after alignment of half of the complex (region colored in gray). EM, electron microscopy; GFRAL, GDNF receptor α-like.
      The RETC634R/GDF15/GFRAL complex was reconstituted similarly to the RETWT complex and purified by gel filtration (Fig. 5A). Analysis of the RETC634R/GDF15/GFRAL complex by negative stain EM showed that the particles were monodisperse with a distinct change in morphology with both single particles and the resultant 2D class averages showing an overall “X” shape with apparent two-fold rotational symmetry (Fig. 5B). An initial model built without symmetry (Fig. S5) agreed well with the overall “X” shape observed in the 2D classes and was roughly twice the size of the RETWT complex. Further refinement was performed with no symmetry assigned (Fig. 5C). The dimensions of the reconstructed model of the RETC634R/GDF15/GFRAL complex are 190 Å × 130 Å × 140 Å, showing an “X” shape as the front view. Reprojections of the density match well with the 2D class averages, supporting the quality of the reconstructed model (Fig. 5D). In agreement with our hypothesis, two copies of the RETWT/GDF15/GFRAL structure could be docked back-to-back into the RETC634R/GDF15/GFRAL density map (Fig. 5E), confirming our observations in the biochemical and biophysical assays (Fig. 2). In particular, the top half of the density provides a much better fit for the complex than the lower half, with extra densities observed in the top half, which could be from Domain 1 (D1) in GFRAL that was not previously resolved. This conformational heterogeneity likely also explains the poorer resolution in the lower half. Given that the RETWT/GDF15/GFRAL complex has 2:2:2 stoichiometry and possesses C2 symmetry, we hypothesized that the RETC634R/GDF15/GFRAL complex, which has 4:4:4 stoichiometry and an “X” shaped 2D projection, would have D2 symmetry. However, assigning C2 or D2 symmetry in further refinements did not improve the resolution of the map (data not shown), likely due to conformational flexibility between the domains. At this resolution, we were able to assign the overall configuration of the complex with moderate confidence but were unable to model the disulfide linked loops fully. Since the two RETC634R protomers in a RETC634R dimer are linked by a disulfide bond and favor the formation of a 4:4:4 complex, it is likely that a pair of disulfide-linked RETC634R dimers form two semi-independent 2:2:2 complexes in trans (Fig. 5E). Although the disulfide linkage itself cannot be resolved at this resolution, the distance between the P622 residues in the two RETC634R protomers is ∼40 Å, which is longer than that of the unliganded RETC634R dimer but could be spanned by residues E623-R635 that are missing from the model and known to be flexible (Fig. 1F). This suggests that ligand binding drives a conformational change of the C-termini of the ECDs of dimeric RETC634R, separating them from each other to accommodate the second heterohexamer.
      Figure thumbnail gr5
      Figure 5Structural analysis of the extracellular domain complex of RETC634R/GDF15/GFRAL using negative stain EM. A, analysis of RETC634R/GDF15/GFRAL by SEC on a Superdex 200 10/300 increase column and BN PAGE. B, negative-stained particles from a representative field of the micrographs taken at 50,000 magnification and representative 2D class averages processed using EMAN2.3 with 11278 particles, presenting different views of the complex. Scale bar for 2D class images represents 10 nm. C, different views of the NS-EM model of RETC634R/GDF15/GFRAL with measured dimensions. D, comparison of representative 2D class averages and the projection images of the reconstructed NS-EM model. Scale bar represents 10 nm. E, side, front, and top views of the NS-EM model of RETC634R/GDF15/GFRAL docked with two RETWT/GDF15/GFRAL complexes (modified from PDB ID: 6Q2J). RET or RET′ protomers that belong to the same RETC634R cross-linked dimer have the same color (blue or yellow). GFRAL and GDF15 are colored brown and green, respectively. BN, Blue Native; EM, electron microscopy; GFRAL, GDNF receptor α-like; SEC, size- exclusion chromatography.
      To obtain higher resolution information regarding the structure of the RETC634R/GDF15/GFRAL ECD complex, cryo-EM grids were prepared using the same sample that was used for negative stain EM. The particles in the cryo-EM micrographs exhibit a similar “X” shape to that seen in negative stain EM (Fig. 6A). With a set of 288,927 RETC634R/GDF15/GFRAL particles, 2D class averages were processed and several classes looked similar to those of the RETWT complex (Fig. 6, B and C). Additionally, several 2D class averages were observed that are unique to the RETC634R complex (Fig. 6B, red stars). These unique 2D classes generally contain a region which is better resolved with the rest of the view blurred, suggesting flexible particles within the classes, so that good alignment was achieved only with part of the particles. 3D classification and refinement with C1 point symmetry result in a 3D model similar to that of a RETWT complex with a final resolution of 10 Å, and the structure of RETWT/GDF15/GFRAL (PDB ID: 6Q2J) could be fitted into the density (Fig. 6D). Notably, after docking one copy of the RETWT/GDF15/GFRAL complex into the map, extra density was observed below one of the two “wings” (Fig. 6D, dotted circle), suggesting the presence of additional structurally flexible components to the complex. However, further processing did not help to improve the resolution of the complex by either masked classification and refinement (
      • Serna M.
      Hands on methods for high resolution cryo-electron microscopy structures of heterogeneous macromolecular complexes.
      ), using the NS-EM model as an initial model or by applying D2 symmetry.
      Figure thumbnail gr6
      Figure 6Cryo-EM analysis of the extracellular domain complex of RETC634R/GDF15/GFRAL. A, representative cryo-EM field of RETC634R/GDF15/GFRAL. Scale bar represents 100 nm. B, 2D class average from 288927 cryo-EM particles of RETC634R/GDF15/GFRAL. Scale bar represents 10 nm. C, 2D class average from 244,672 cryo-EM particles of RETWT/GDF15/GFRAL. Scale bar represents 10 nm. Unique classes of RETC634R/GDF15/GFRAL are marked with red stars and representative similar classes between RETC634R/GDF15/GFRAL and RETWT/GDF15/GFRAL are marked with blue stars. D, different views of a 3D cryo-EM model of RETC634R/GDF15/GFRAL. A structure of RETWT/GDF15/GFRAL (cyan, PDB ID: 6Q2J) is fitted in the density. Extra density observed exclusively for the RETC634R/GDF15/GFRAL complex is highlighted with dotted black circles. EM, electron microscopy; GFRAL, GDNF receptor α-like.

      Conformational flexibility of the extracellular domains of RETC634R and the RETWT/GDF15/GFRAL complex from MD simulations

      To further investigate the conformational flexibility of the covalently linked RETC634R dimer and the RETWT/GDF15/GFRAL complex, we performed atomistic MD simulations. MD simulations of the RETWT/GDF15/GFRAL complex showed that the angle between the two wings oscillates in a normal distribution fashion with the peak in the range of 45 to 50° (Fig. 7). Alignment of the simulation trajectory by backbone revealed that the two RET protomers in the tripartite complex exhibit a higher degree of movement than GDF15 and GFRAL (Fig. S7, and Movie S2), in close agreement with our observations in the cryo-EM maps as described earlier (Fig. 4C). MD simulations of the RETC634R mutant dimer predicts a high degree of conformational flexibility, with the angle between the two RETC634R protomers ranging primarily from 110 to 140° (Fig. 7 and Movie S3) but without unraveling of the secondary structure (Fig. S6). Such observation is not surprising as the terminal region of the RETECD is unstructured and flexible. During the entire simulation runs, the RETC634R dimer did not once adopt a conformation similar to that of RETWT protomers in the RETWT/GDF15/GFRAL complex, suggesting that the RETWT conformation is not favored for RETC634R in the absence of its ligands. The further implication of this is that activation may well occur by a different mechanism. It is thus consistent with the formation of the 4:4:4 highly flexible RETC634R/GDF15/GFRAL complex described above (Figs. 5 and 6). However, binding of certain ligands, for example GDNF/GFRα1, to RETC634R can apparently force the receptor into a WT-like signaling complex (Fig. 2A).
      Figure thumbnail gr7
      Figure 7MD simulation of the extracellular domain complex of RETWT/GDF15/GRFRAL and RETC634R dimer. A, distribution of the angles between the two “wings” of the RETWT/GDF15/GRFRAL complex (P161GFRAL-C77GDF15-P161GFRAL’, PDB ID: 6Q2J) and the two RETC634R protomers of the RETC634R dimer (V262RET-C630RET-V262RET’). Three representative frames (F1, F2, and F3) of the RETWT/GDF15/GRFRAL complex (B) and the RETC634R dimer (C) aligned along one protomer. GDF15 dimer is colored in green, GFRAL in light blue, and different RET protomers in yellow and orange. GFRAL, GDNF receptor α-like.

      Discussion

      WT RET signaling is regulated by ligand-induced dimerization of the RET ECD, bringing the two intracellular kinase domains close to each other and thus mediating proximity-induced autophosphorylation and activating downstream signaling. The dynamic control of RET complex association and disassociation is essential for normal cellular functions. In MEN2A, RET mutations such as C634R reorganize the disulfide network of the RET ECD such that an intermolecular disulfide bond occurs, enforcing persistent and ligand-independent dimerization of RET. Despite its clinical importance, little is known about the molecular basis for the oncogenic activation of the RET C634R mutant due to a lack of structural studies. In this work, we structurally characterized the extracellular domain of the RETC634R mutant both with and without ligands and showed that the RETC634R dimer can further complex with the different canonical RET-activating ligands in divergent modes that are structurally distinct from RETWT and have the potential to assemble high-order signaling clusters. These discrete structures define the molecular mechanisms through which the RETC634R mutation exerts its oncogenicity.

      The RETC634R ECD dimer adopts an activating conformation in the absence of ligands

      In the absence of activating ligands, the extracellular domains of RETWT either exist as inactive monomers or dimerize in an inactive head-to-head conformation (
      • Li J.
      • Shang G.
      • Chen Y.-J.
      • Brautigam C.A.
      • Liou J.
      • Zhang X.
      • et al.
      Cryo-EM analyses reveal the common mechanism and diversification in the activation of RET by different ligands.
      ) (Fig. 8A). Our work and other EM studies have shown that activation of RETWT ECD is mediated by the formation of a 2:2:2 “butterfly” complex with its ligands, driving a tail-to-tail conformational rearrangement that promotes proximal kinase domain activation (Fig. 3B). We reveal that the aberrant disulfide bond of the oncogenic RETC634R ECD mutant enforces a dimeric “S” conformation that presumably results in constitutive activation of RET signaling in the full-length receptor (Figs. 8B, 1, C and E). The angle between the two docked RETC634R ECD protomers is approximately 140°, which is wider than that of the two RETWT protomers in complex with any of its ligands (PDB ID: 6Q2O, 6Q2N, and 6Q2J) but consistent with the modal value in our MD simulations (Fig. 7, A and C). Conversely, the distance measured between the P622 residues (ca. 18 Å) is three times shorter (42–50 Å). This implies that the unliganded RETC634R dimer may result in more efficient autoactivation than for a liganded RETWT complex, although the lack of structural information regarding full length RET precludes a comprehensive understanding of how these distances and angles affect the intracellular domains.
      Figure thumbnail gr8
      Figure 8Proposed models of the dimeric RETC634R and its interaction with its ligands. A, cartoon representation of the activation mechanism of full-length RETWT under normal physiology. B, two configurations of the reconstituted extracellular domain complexes of RETC634R/GDF15/GFRAL and RETC634R/GDF15/GFRAL representing ligand-dependent activation mechanisms of RETC634R. The two RET monomers are colored in yellow and brown. Instead of forming a “butterfly” conformation seen in the RETWT complexes, RETC634R complexes with the coreceptors (red) and GFLs (pink) adopt either a “butterfly” conformation or a novel “twin-butterfly” conformation. C, cartoon representation of our proposed model for full-length RETC634R activation by extrapolation from our extracellular domain structures. RET protomer 1 and 1’ (yellow and brown) forms one RETC634R dimer, while RET protomer 2 and 2’ (blue and light blue) forms another. GFRAL, GDNF receptor α-like.
      Whereas C630 and C634 are located at the end of the flexible C-terminal loop of RETECD, the other cysteine residues, such as C609, C611, C618, and C620 whose mutations are also associated with MEN2A and are cancer-inducing, albeit with better prognosis, sit in two strands of an antiparallel β-sheet. In WT RET, C609-C620 and C611-C618 form two intramolecular disulfide bonds. Interestingly, we found that expression of RETC620R ECD using the same approach and conditions as for RETC634R ECD also yielded a disulfide-crosslinked dimer after Fc-tag removal (Fig. S8). However, the extent of dimerization was less than for RETC634R, which correlates with the reduced oncogenic potency of RETC620R (
      • Wagner S.
      • Zhu S.
      • Nicolescu A.
      • Mulligan L.
      Molecular mechanisms of RET receptor-mediated oncogenesis in multiple endocrine neoplasia 2.
      ). We hypothesize that the formation of the C609-C609 intermolecular disulfide bond in RETC620R is less favored due to structural constraints, which may be the same for other cysteine mutation-induced oncogenic dimers of RET and explain why C634 mutations are the most aggressive. Equally, some of these dimerized RET mutants may not pass through the protein production machinery as efficiently, in agreement with previous reports that the C609, C611, C618, and C620 mutations impair RET folding and maturation (
      • Chappuis-Flament S.
      • Pasini A.
      • De Vita G.
      • Ségouffin-Cariou C.
      • Fusco A.
      • Attié T.
      • et al.
      Dual effect on the RET receptor of MEN 2 mutations affecting specific extracytoplasmic cysteines.
      ,
      • Ito S.
      • Iwashita T.
      • Asai N.
      • Murakami H.
      • Iwata Y.
      • Sobue G.
      • et al.
      Biological properties of Ret with cysteine mutations correlate with multiple endocrine neoplasia type 2A, familial medullary thyroid carcinoma, and Hirschsprung’s disease phenotype.
      ,
      • Carlomagno F.
      • Salvatore G.
      • Cirafici A.M.
      • De Vita G.
      • Melillo R.M.
      • De Francisas V.
      • et al.
      The different RET-activating capability of mutations of cysteine 620 or cysteine 634 correlates with the multiple endocrine neoplasia type 2 disease phenotype.
      ).

      Divergent activation of RETC634R is mediated by ligand-dependent complexation

      We show that RETWT ECD readily forms 2:2:2 tripartite complexes with both GDF15/GFRAL and GDNF/GFRα1 (Figs. 2 and 3), in accordance with previous reports (
      • Li J.
      • Shang G.
      • Chen Y.-J.
      • Brautigam C.A.
      • Liou J.
      • Zhang X.
      • et al.
      Cryo-EM analyses reveal the common mechanism and diversification in the activation of RET by different ligands.
      ,
      • Schlee S.
      • Carmillo P.
      • Whitty A.
      Quantitative analysis of the activation mechanism of the multicomponent growth-factor receptor Ret.
      ). We observed multiple conformations of the RETWT/GDF15/GFRAL complex using cryo-EM, revealing that the “wings” undergo a concerted twisting and bending motion (Fig. 4). Our MD simulations corroborate this finding and predict that the degree of these movements would be even larger at biologically relevant temperatures (Fig. 7). In particular, we found that, within the tripartite complex, the RETECD undergoes the greatest degree of movement (Fig. 4C), which is consistent with findings from previous studies of the RETWT/NTRN/GFRα2 (
      • Bigalke J.M.
      • Aibara S.
      • Roth R.
      • Dahl G.
      • Gordon E.
      • Dorbéus S.
      • et al.
      Cryo-EM structure of the activated RET signaling complex reveals the importance of its cysteine-rich domain.
      ) and the zebrafish RETWT/GDNF/GFRα1 (
      • Adams S.E.
      • Purkiss A.G.
      • Knowles P.P.
      • Nans A.
      • Briggs D.C.
      • Borg A.
      • et al.
      A two-site flexible clamp mechanism for RET-GDNF-GFRα1 assembly reveals both conformational adaptation and strict geometric spacing.
      ) complexes. Such conformational flexibility might be anticipated for a protein capable of binding as diverse range of ligands as RET but may also have limited the resolution of the cryo-EM map reconstructed with the data used in this study.
      We found that mutant RETC634R ECD is capable of forming both the archetypal 2:2:2 tripartite RET-ligand complex but also higher order complexes that have not been previously reported. We have assigned the major higher order complex as a 4:4:4 tripartite complex with a back-to-back twin butterfly configuration. Our studies suggest that, in contrast to RETWT, the preference for RETC634R to adopt these distinct complex configurations is ligand dependent. In the published structures of the RETWT/GDF15/GFRAL (PDB ID: 6Q2J) and RETWT/GDNF/GFRα1 (PDB ID: 6Q2N) complexes, the distance between the last resolved residues (P622) of each RETECD protomer are ∼50 Å and ∼44 Å, respectively. In the case of RETC634R/GDNF/GFRα1, we were able to model in the unresolved C-terminal residues of RETECD (622–632) as disordered chains containing the mutant C630-C630 disulfide bond. However, this is not at all possible for RETC634R/GDF15/GFRAL, on account of the increased RETECD-protomer separation, explaining the ligand-dependent architecture of the RETC634R complexes that we observed (Fig. 9). Furthermore, the RETC634R/GDNF/GFRα1 complex is more thermally stable relative to RETWT/GDNF/GFRα1 and this is likely explained by the entropic benefit provided by the C630-C630 RET stapling (Fig. 2B). If these distinct arrangements of the RETC634R ECD complexes translate to its full-length form in vivo, they could impact on the extent of its activation in different tissues. We anticipate that the complexes characterized here are specific to the C634R mutation and that the other oncogenic cysteine mutations of RET may impact on the ligand binding in different ways.
      Figure thumbnail gr9
      Figure 9Analysis of the potential C630-C630 disulfide bond formation for the RETC634R/GDF15/GFRAL and RETC634R/GDNF/GFRα1 extracellular domain complexes in a 2:2:2 configuration. A polypeptide chain of 8 and 10 unresolved amino acids is added to the C-terminal tail of each RETECD protomer in RETWT/GDF15/GFRAL (PDB code 6Q2J) and RETWT/GDNF/GFRα1 (PDB code 6Q2N) using Coot, respectively. Structural comparison and distance measurement were performed using PyMOL. ECD, extracellular domain; GFRAL, GDNF receptor α-like.

      Relevance to cancer

      Antibody-based studies have identified that GFRAL is expressed in multiple cancer tissues (Human Protein Atlas available from http://www.proteinatlas.org/) (
      • Berglund L.
      • Björling E.
      • Oksvold P.
      • Fagerberg L.
      • Asplund A.
      • Al-Khalili Szigyarto C.
      • et al.
      A genecentric human protein atlas for expression profiles based on antibodies.
      ) and its signaling via RETC634R may be biologically relevant under disease conditions. The ECD of RETC634R clearly forms a twin-butterfly complex with GDF15/GFRAL with 4:4:4 stoichiometry (Figs. 2 and 5), and we speculate that about 30% of the RETC634R/GDNF/GFRα1 also forms a 4:4:4 complex. The structures seen in this study obviously lack the spatial constraints imposed by the cell membrane. In our low-resolution NS-EM model, the CRDs of the docked dual-hexameric complex sit directly opposite from each other (Fig. 5E). However, given how flexible the RETC634R cross-linked dimer is (Fig. 7), the hexameric complexes in the membrane-bound full length 4:4:4 complex might be able to bend toward each other to allow more efficient anchoring of the TM domains into the membrane (Fig. 8C). Such a collapsed “twin-butterfly” 4:4:4 complex might even lead to the formation of locally clustered higher-order oligomers (
      • Li J.
      • Shang G.
      • Chen Y.-J.
      • Brautigam C.A.
      • Liou J.
      • Zhang X.
      • et al.
      Cryo-EM analyses reveal the common mechanism and diversification in the activation of RET by different ligands.
      ,
      • Tansey M.G.
      • Baloh R.H.
      • Milbrandt J.
      • Johnson E.M.
      GFRα-mediated localization of RET to lipid rafts is required for effective downstream signaling, differentiation, and neuronal survival.
      ) that have been characterized for related receptors such as of Eph receptors (
      • Himanen J.P.
      • Yermekbayeva L.
      • Janes P.W.
      • Walker J.R.
      • Xu K.
      • Atapattu L.
      • et al.
      Architecture of Eph receptor clusters.
      ) and endothelial growth factor receptor (
      • Needham S.R.
      • Roberts S.K.
      • Arkhipov A.
      • Mysore V.P.
      • Tynan C.J.
      • Zanetti-Domingues L.C.
      • et al.
      EGFR oligomerization organizes kinase-active dimers into competent signalling platforms.
      ).We speculate that the change of conformation from “butterfly” to “twin-butterfly” leads to different signaling outcomes and intensities (
      • Frêche B.
      • Guillaumot P.
      • Charmetant J.
      • Pelletier L.
      • Luquain C.
      • Christiansen D.
      • et al.
      Inducible dimerization of RET reveals a specific AKT deregulation in oncogenic signaling.
      ) and thus provides a structural explanation for oncogenesis. However, as we have only observed these configurations in the context of the isolated extracellular domains, it remains to be seen whether such complexes do form in vivo and how they might influence signaling in cancer. Nonetheless, these insights pave the way for further investigations into RET gain-of-function activation under disease conditions and potentially for the discovery of cancer therapies targeting the extracellular domain of RETC634R as well as the other dimerization-inducing oncogenic cysteine mutations.

      Experimental procedures

      DNA constructs, protein expression, and purification

      The gene encoding human RETECD (residues 1–635) bearing the C634R, C86R, and C216S mutations (RETC634R) with a C-terminal Thrombin protease cleavage site, Fc, and His8-Flag tags (RETC634R-Fc) was cloned into pcDNA3.1 vector in frame with the cytomegalovirus promotor (Tables S1 and S2). WT RETECD (residues 1–635) was subcloned into pcDNA3.1 vector with a Tobacco Etch Virus protease cleavage site and His8-Flag tags (RETWT). Human GDF15 (residues 195–308) was cloned into the same vector with a modified N-terminal Fc tag, followed by a thrombin cleavage site, and this construct was used for cotransfection with a separate construct for the expression of modified Fc as described previously (
      • Liu Y.
      • de Castro Ribeiro O.
      • Robinson J.
      • Goldman A.
      Expression and purification of the extracellular domain of wild-type humanRET and the dimeric oncogenic mutant C634R.
      ,
      • Xiong Y.
      • Walker K.
      • Min X.
      • Hale C.
      • Tran T.
      • Komorowski R.
      • et al.
      Long-acting MIC-1/GDF15 molecules to treat obesity: evidence from mice to monkeys.
      ). The construct generation and expression procedures of RETWT, Fc-GDF15, and human GFRAL (residues 19–351) with C-terminal His8-twin Strep tags were the same as previously described (
      • Liu Y.
      • de Castro Ribeiro O.
      • Robinson J.
      • Goldman A.
      Expression and purification of the extracellular domain of wild-type humanRET and the dimeric oncogenic mutant C634R.
      ). Human GDNF (residues 142–275) and human GFRα1 (residues 25–429) were subcloned into pk503.9 vector with N-terminal Flag-His8 tags followed by a thrombin cleavage site. The over-expression of recombinant RETC634R-Fc, RETWT, and Fc-GDF15 were achieved using transient expression in adherent HEK293T cells (American Type Culture Collection, CRL-3216), while GFRAL, GDNF, and GFRα1 were expressed using the baculovirus-infected insect cell system in Trichoplusia ni High Five (Hi5) and Spodoptera frugiperda (Sf 9) insect cells (Thermo Fisher Scientific) maintained as suspension cultures. All proteins were secreted into the cell culture medium and the culturing medium containing the secreted proteins was harvested 7- and 5-days post-transfection for RETC634R-Fc or RETWT and Fc-GDF15, respectively and after 3 days for GFRAL, GDNF, and GFRα1. The reconstruction of the GDNF/GFRα1 complex was achieved either by coexpression and purification or the assembly of the separately purified proteins.
      To purify RETC634R, Protein A resin (GenScript) was used to immobilize RETC634R-Fc for 2 hours, followed by a thorough washing step using washing buffer containing 20 mM Hepes pH 7.5, 150 mM NaCl, and 1 mM CaCl2. Biotinylated thrombin (Merck Millipore) was added to the resin and the sample was incubated at 4 °C for 12 h without agitation. After protease cleavage, the flow-through containing the cleaved proteins was collected and incubated with pre-equilibrated Strep-Tactin resin (IBA Lifesciences) to remove any unbound thrombin for 30 min at room temperature (RT). The supernatant was then collected and concentrated using Amicon concentrators (50 kDa). A final polishing step was performed by size-exclusion chromatography (SEC) on a Superdex 200 increase 10/300 Gl column (GE Healthcare) in 20 mM Hepes pH 7.5, 100 mM NaCl, and 1 mM CaCl2. Each elution fraction was subject to nonreducing SDS PAGE for purity evaluation and the peak fraction was directly used to prepare EM grids. For storage, fractions containing the dimeric RETC634R were combined and concentrated with Amicon concentrators (50 kDa). The final concentrated samples were stored at −70 °C in SEC buffer containing 7.5% glycerol.
      The purification of RETWT, Fc-GDF15, and GFRAL was performed as described before (
      • Liu Y.
      • de Castro Ribeiro O.
      • Robinson J.
      • Goldman A.
      Expression and purification of the extracellular domain of wild-type humanRET and the dimeric oncogenic mutant C634R.
      ). In brief, RETWT was purified using a two-step purification protocol, which included Ni-NTA (Qiagen) and anti-Flag (GenScript) affinity purifications, and the final sample was eluted from anti-Flag resin using poly-Flag peptide (Bimake). The peptide was removed by buffer exchange using micro bio-spin columns (Bio-Rad). To purify Fc-GDF15, medium containing the secreted protein was incubated with Protein A resin and eluted using low-pH buffer containing 100 mM sodium citrate pH 3.1 and 100 mM NaCl, followed by immediate neutralization to pH 7.5 using 1 M Tris pH 8.8. For storage, purified Fc-GDF15 was buffer-exchanged into Tris-buffered saline containing 7.5% glycerol. Strep-Tactin resin was used to affinity purify GFRAL and the bound protein was eluted using 5 mM d-desthiobiotin. The eluate was further purified by SEC on a Hiload Superdex 200 pg 16/600 (GE Healthcare) in 20 mM Hepes pH 7.2, 150 mM NaCl and buffer exchanged into storage buffer containing 20 mM Hepes pH 7.2, 150 mM NaCl, and 7.5% glycerol. To purify coexpressed GDNF/GFRα1 or GDNF, Ni-NTA resin was used to capture His-tagged proteins and the bound protein was eluted with TBS containing 300 mM imidazole. The eluate was incubated with thrombin protease for tag cleavage overnight at RT and was then concentrated using Amicon concentrators (10 kDa or 5 kDa) and purified on a Superdex 200 increase 10/300 Gl column (GE Healthcare) in 20 mM Hepes pH 7.5, 100 mM NaCl. The peak fractions containing the desired proteins were combined, concentrated, and were stored at −70 °C in SEC buffer containing 7.5% glycerol. GFRα1 was purified similarly to that of the GDNF/GFRα1 complex but without protease cleavage.

      Bio-layer interferometry technology system

      A bio-layer interferometry technology system instrument (FortéBio) was used to measure the binding affinity between RETC634R and Fc-GDF15/GFRAL, and the experimental set up was as previously described (
      • Liu Y.
      • de Castro Ribeiro O.
      • Robinson J.
      • Goldman A.
      Expression and purification of the extracellular domain of wild-type humanRET and the dimeric oncogenic mutant C634R.
      ). Fc-GDF15 (15 μM) was first immobilized on anti-hIgG Fc capture biosensors (FortéBio) and the biosensors bound with Fc-GDF15 were then saturated by binding to GFRAL (3 μM). Various concentrations of RETC634R (0.3, 1.2, 2.8, 7, 14, 28, and 42 μM) were used for the second binding step and the saturation signals at the end of the second binding steps were analyzed as a function of RETC634R concentration on a logarithmic scale. The dissociation constant (Kd) was obtained through fitting a nonlinear regression sigmoidal model using GraphPad Prism 8.

      BN PAGE

      The RET/Fc-GDF15/GFRAL tripartite complexes were prepared by incubating Fc-GDF15 dimer (2 μM) with GFRAL (4 μM) for 15 min, followed by the addition of either RETWT monomer (4 μM) or RETC634R dimer (2 μM) to a final volume of 10 μl. Similarly, the RET/GDF15/GFRAL complex was prepared by adding 0.01 U thrombin protease to the GDF15/GFRAL complex prior to the addition of RET. The formation of the RET/GDNF/GFRα1 complex was performed by incubating RET and copurified GDNF/GFRα1 at a molar ratio of 1:1 at the same concentrations. The mixtures were kept at RT for 30 min to allow complex formation. BN PAGE loading buffer was added to each sample before electrophoresis, which was performed at 100 V and 4 °C for 3.5 h (
      • Liu Y.
      • de Castro Ribeiro O.
      • Robinson J.
      • Goldman A.
      Expression and purification of the extracellular domain of wild-type humanRET and the dimeric oncogenic mutant C634R.
      ,
      • Wittig I.
      • Braun H.-P.
      • Schägger H.
      Blue native PAGE.
      ,
      • Liu Y.
      • Kaljunen H.
      • Pavić A.
      • Saarenpää T.
      • Himanen J.P.
      • Nikolov D.B.
      • et al.
      Binding of EphrinA5 to RET receptor tyrosine kinase: an in vitro study.
      ). Gels were destained with 20% methanol and 10% acetic acid.
      For the temperature-dependent stability measurement, the RET/GDF15/GFRAL and RET/GDNF/GFRα1 complexes were reconstituted as described above, aliquoted into PCR tubes, and heated for 5 min at 50 °C, 54 °C, 57 °C, 59 °C, 60 °C, 64 °C, 70 °C and 54 °C, 60 °C, 64 °C, 67 °C, 70 °C, 72 °C, 76 °C, 80 °C, 85 °C, respectively. After incubation, the samples were immediately placed on ice for 10 min, centrifuged, and mixed with BN PAGE loading buffer. The electrophoresis was performed as described above. The band intensities of the complexes were measured using ImageJ (
      • Schneider C.A.
      • Rasband W.S.
      • Eliceiri K.W.
      NIH image to ImageJ: 25 years of image analysis.
      ). Normalized intensities were analyzed against the different temperature points, and the melting temperature (Tm) was obtained through fitting a nonlinear regression sigmoidal model using GraphPad Prism 8.

      Complex reconstitution and purification for structural studies

      To form the RETWT/Fc-GDF15/GFRAL complex, Fc-GDF15 and GFRAL were first incubated at a molar ratio of 1:2 for 15 min at RT. RETWT was then added to the mixture so that the final molar ratio of RETWT, Fc-GDF15 dimer, and GFRAL was 2:1:2. The sample was purified using SEC after 30 min postincubation on a Superose 6 increase 10/300 Gl column in 20 mM Hepes pH 7.5, 100 mM NaCl, and 1 mM CaCl2. For the preparation of the RETC634R/GDF15/GFRAL complex, Fc-GDF15 and GFRAL at a molar ratio of 1:2 were first incubated together with 0.055 U thrombin per 100 μg Fc-GDF15 for 15 min at RT before the addition of the RETC634R dimer. The final molar ratio of RETC634R dimer, GDF15 dimer, and GFRAL was 1:1:2. The protein mixture was kept at RT for an additional 30 min and the complex separated from thrombin using SEC on a Superdex 200 increase 10/300 Gl column in 20 mM Hepes pH 7.5, 100 mM NaCl, and 1 mM CaCl2. For both complexes, the purity of each fraction from SEC was assessed using BN PAGE and the peak fractions were directly used to prepare electron microscopy grids without concentrating. For protein storage or other applications, fractions containing the complexes were pooled and concentrated with Amicon concentrators (50 kDa).

      Size exclusion chromatography-coupled multiangle static laser light scattering

      SEC-MALS was used to characterize the particle distribution and the oligomeric state of RETC634R dimer, RETC634R/GDF15/GFRAL, and RETWT/GDF15/GFRAL complexes. All measurements were performed with a HPLC system (Shimadzu) on a Superdex 200 10/300 Gl column with a flow rate at 0.25 ml/min at RT in SEC buffer containing 20 mM Hepes pH 7.5, 100 mM NaCl, and 1 mM CaCl2. The MALS system was equipped with MiniDAWN TREOS light scattering and Optilab rEX refractive index detectors (Wyatt Technology Corp) The data was first analyzed using ASTRA 6 software (Wyatt Technology Corp) and was exported and replotted using GraphPad Prism 8.

      NS-EM sample preparation

      Negative stain grids were glow-discharged using either a Pelco Easiglow unit with 10 mA for 30 s or a Quorum Emitech K100X Glow discharge unit with 30 mA for 45 s. For negative stain EM experiments, 3 μl of protein samples at 0.01 mg/ml for RET complexes or 0.005 mg/ml for RETC634R dimer were first applied to the glow-discharged 300 mesh copper grids with carbon coating for 1 min. The grids were blotted, washed once quickly with a 10 μl drop of 2% (w/v) uranyl acetate (UA) solution, blotted again, and incubated with a second 10 μl drop of UA for 45 s, followed by a final blotting step to remove the stain. The grids were left to air dry for 5 min at RT before imaging or storage.

      NS-EM data processing

      For the RETC634R dimer, 40 micrographs were collected using a Tecnai F20 TEM (FEI, 200 kV) at a nominal magnification of 50,000 magnification resulting in 2.25 Å/pixel. Grids of RETC634R/GDF15/GFRAL were imaged using the same microscope and the same settings with a total of 68 micrographs collected. NS-EM data processing was performed using EMAN2.3 (
      • Tang G.
      • Peng L.
      • Baldwin P.R.
      • Mann D.S.
      • Jiang W.
      • Rees I.
      • et al.
      EMAN2: an extensible image processing suite for electron microscopy.
      ). After micrograph evaluation, contrast transfer function (CTF) estimation and structure factor calculation were carried out. A total of 3490 and 11,278 particles were manually picked for the RETC634R dimer and the RETC634R/GDF15/GFRAL complex, respectively. Reference-free 2D classification, initial model building, and single-map refinement were performed in EMAN2.3. The initial model building was performed with C2 symmetry for the RETC634R dimer and the same symmetry was applied in subsequent refinement steps for the RETC634R dimer. For the 3D model reconstruction of the RETC634R/GDF15/GFRAL complex, no symmetry was applied in the initial model building and the model refinement.

      Cryo-EM sample preparation

      Cryo-EM grids were glow discharged using the same method as for the NS-EM grids. To prepare the specimens for data collection, 3 μl of protein samples at 0.12 mg/ml for both RETWT/GDF15/GFRAL and RETC634R/GDF15/GFRAL were applied to glow-discharged Quantifoil R1.2/1.3300 mesh gold grids for 5 s before blotting. The grids were prepared using either a Vitrobot Mark IV with 6 s blot time (blot force 6) or a Leica EM GP with 1.5 s blot time at 4 °C with 80% humidity.

      RETWT/Fc-GDF15/GFRAL cryo-EM data processing

      For RETWT/Fc-GDF15/GFRAL, the micrographs were collected on a Titan Krios electron microscope (300 kV) with a Falcon 3EC direct electron detector in integrative mode at 1.065 Å/pixel at 75,000 nominal magnification. A total of 6563 movies were collected with a defocus range of −1.6 to −3 μm with a dose per fraction of 1.5 e/Å2 and a total dose of 60.3 e/Å2 (Table S3). Most steps of image processing were performed using RELION 3 (Fig. S9). MotionCorr2 was used to perform beam-induced motion correction of the micrographs (
      • Zheng S.Q.
      • Palovcak E.
      • Armache J.-P.
      • Verba K.A.
      • Cheng Y.
      • Agard D.A.
      MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy.
      ) and GCTF was used for the CTF estimation (
      • Zhang K.
      Gctf: Real-time CTF determination and correction.
      ). The corrected micrographs were subject to particle auto-picking using crYOLO with the general model (
      • Wagner T.
      • Merino F.
      • Stabrin M.
      • Moriya T.
      • Antoni C.
      • Apelbaum A.
      • et al.
      SPHIRE-crYOLO is a fast and accurate fully automated particle picker for cryo-EM.
      ). A total of 1,135,874 particles were extracted after rescaling to a pixel size of 2.13 Å/pixel. Two rounds of 2D and 3D classification were performed to separate particles with different qualities. Two processing routes were performed for the 3D model reconstruction. In one route, 21 images from 2D class averages with 302,004 particles were selected for initial model building and one of the five initial models was used as a reference map for 3D classification. In the other route, a 30 Å map was generated based on the published cryo-EM structure of RETWT/GDF15/GFRAL (PDB ID: 6Q2J) using Chimera (
      • Pettersen E.F.
      • Goddard T.D.
      • Huang C.C.
      • Couch G.S.
      • Greenblatt D.M.
      • Meng E.C.
      • et al.
      UCSF Chimera--A visualization system for exploratory research and analysis.
      ) which was then 60 Å low-pass filtered and used as a reference map for subsequent 3D classification and refinement with C2 symmetry. The structure of RETWT/GDF15/GFRAL (PDB ID: 6Q2J) was fitted into the final density map using Chimera. Additional glycans (N-acetylglucosamine) at different glycosylation sites, not present in the published structure (PDB ID: 6Q2J), were added using Coot (
      • Emsley P.
      • Lohkamp B.
      • Scott W.G.
      • Cowtan K.
      Features and development of Coot.
      ) through manual model building to fit in the densities of the glycans in the refined density map. Analysis of different 3D classes was performed in Chimera.

      Rigid-body flexible fitting into cryo-EM density maps

      To prepare the structure of the RETECD/GDF15/GFRALECD complex (PDB ID: 6Q2J), glycans and ions were first removed from the original structure. Rigid bodies were identified using Ribfind (
      • Pandurangan A.P.
      • Topf M.
      Finding rigid bodies in protein structures: application to flexible fitting into cryoEM maps.
      ). The individual domains of RETECD, one GFRALECD monomer with one GDF15 monomer were set as rigid bodies with a Ribfind cut off of 30% for flexible fitting into individual cryo-EM maps at a resolution of 11 Å using Flex-EM (
      • Topf M.
      • Lasker K.
      • Webb B.
      • Wolfson H.
      • Chiu W.
      • Sali A.
      Protein structure fitting and refinement guided by cryo-EM density.
      ). The final models were refined using PHENIX real-space refinement (
      • Afonine P.V.
      • Poon B.K.
      • Read R.J.
      • Sobolev O.V.
      • Terwilliger T.C.
      • Urzhumtsev A.
      • et al.
      Real-space refinement in PHENIX for cryo-EM and crystallography.
      ) and assessed by rigid fitting into the corresponding cryo-EM maps in Chimera. The structures were superimposed onto the GDF15 dimers and morphing was performed to evaluate the conformational change.

      RETC634R/GDF15/GFRAL cryo-EM data processing

      For RETC634R/GDF15/GFRAL, the micrographs were collected on a Titan Krios electron microscope (300 kV) with a BioQuantum967 (K2 summit camera, Gatan) in counting mode at 1.07 Å/pixel at 130,000 nominal magnification. A total of 2852 movies were collected with a defocus range of -1.6 to -3 μm with a dose per fraction of 1.26 e/Å2 and a total dose of 62.88 e/Å2 (Table S3). Following motion correction and CTF estimation as described earlier using RELION 3, 309,003 particles were autopicked using crYOLO, rescaled, and extracted with 2.675 Å pixel size. The bad particles were removed through four rounds of 2D class averaging, and 18 images from 2D classes with 281,224 particles were selected for 3D classification (C1 point symmetry) using the low-pass–filtered map of RETWT/GDF15/GFRAL (PDB ID: 6Q2J) as described earlier as a reference map. One of the three maps from 3D classification was selected as the reference map for auto-refinement (C1 point symmetry).

      MD simulation

      We performed classical atomistic MD simulations on the WT RET/GDF15/GFRAL complex (PDB ID: 6Q2J) and the RETC634R dimer. The cross-linked RETC634R dimer was constructed by docking two copies of RET monomers adapted from the RET/GDF15/GFRAL complex (PDB ID: 6Q2J) in the reconstructed NS-EM map of the RETC634R dimer. Thirteen additional amino acids (623EDIQDPLCDELRR635) were added to the C-terminal region of RETECD using Coot, which are absent in the PDB file used, with a disulfide bond between the C630 residues of the two RETC634R protomers. Web-based CHARMM-GUI tools were used to construct the model systems (
      • Jo S.
      • Kim T.
      • Iyer V.G.
      • Im W.
      CHARMM-GUI: a web-based graphical user interface for CHARMM.
      ,
      • Lee J.
      • Cheng X.
      • Swails J.M.
      • Yeom M.S.
      • Eastman P.K.
      • Lemkul J.A.
      • et al.
      CHARMM-GUI input generator for NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM simulations using the CHARMM36 additive force field.
      ,
      • Brooks B.R.
      • Brooks C.L.
      • Mackerell A.D.
      • Nilsson L.
      • Petrella R.J.
      • Roux B.
      • et al.
      CHARMM: the biomolecular simulation program.
      ). Each model system consisted of protein, structurally resolved calcium ions and paucimannose modeled at the known glycosylation sites (
      • Park S.J.
      • Lee J.
      • Qi Y.
      • Kern N.R.
      • Lee H.S.
      • Jo S.
      • et al.
      CHARMM-GUI glycan modeler for modeling and simulation of carbohydrates and glycoconjugates.
      ,
      • Park S.-J.
      • Lee J.
      • Patel D.S.
      • Ma H.
      • Lee H.S.
      • Jo S.
      • et al.
      Glycan Reader is improved to recognize most sugar types and chemical modifications in the Protein Data Bank.
      ,
      • Jo S.
      • Song K.C.
      • Desaire H.
      • MacKerell A.D.
      • Im W.
      Glycan reader: automated sugar identification and simulation preparation for carbohydrates and glycoproteins.
      ) and was solvated in TIP3P water molecules, containing sodium and chloride ions at a concentration of 100 mM. The model system sizes ranged from ∼550,000 to ∼1,000,000 atoms depending on the simulation box sizes. Due to the irregular shape of the protein, the simulation box was rectangular and protein rotation was prevented by keeping harmonic constraints on individual Cα-atoms of residues: 432, 496, and 506 from RET protomers. The strength of the constraints varied from 2000 kJ mol-1 nm-2 to 50,000 kJ mol−1 nm−2 to ensure the constraints do not affect the results. Multiple independent 100 to 400 ns MD simulations of WT (∼1.18 μs) and mutant (∼1.49 μs) model systems were performed using simulation software GROMACS 2020.2 (
      • Páll S.
      • Abraham M.J.
      • Kutzner C.
      • Hess B.
      • Lindahl E.
      Tackling exascale software challenges in molecular dynamics simulations with GROMACS.
      ,
      • Abraham M.J.
      • Murtola T.
      • Schulz R.
      • Páll S.
      • Smith J.C.
      • Hess B.
      • et al.
      GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputers.
      ,
      • Van Der Spoel D.
      • Lindahl E.
      • Hess B.
      • Groenhof G.
      • Mark A.E.
      • Berendsen H.J.C.
      GROMACS: fast, flexible, and free.
      ) and CHARMM36 (
      • Mackerell A.D.
      • Feig M.
      • Brooks C.L.
      Extending the treatment of backbone energetics in protein force fields: limitations of gas-phase quantum mechanics in reproducing protein conformational distributions in molecular dynamics simulations.
      ,
      • MacKerell A.D.
      • Bashford D.
      • Bellott M.
      • Dunbrack R.L.
      • Evanseck J.D.
      • Field M.J.
      • et al.
      All-atom empirical potential for molecular modeling and dynamics studies of proteins.
      ) forcefield. After minimization and equilibration of the systems, the pressure and temperature were kept stable using Nose-Hoover thermostat (
      • Nosé S.
      A unified formulation of the constant temperature molecular dynamics methods.
      ,
      • Hoover W.G.
      Canonical dynamics: equilibrium phase-space distributions.
      ) and Parrinello-Rahman barostat (
      • Parrinello M.
      • Rahman A.
      Crystal structure and pair potentials: a molecular-dynamics study.
      ,
      • Parrinello M.
      • Rahman A.
      Polymorphic transitions in single crystals: a new molecular dynamics method.
      ). The timestep of simulations was 2 fs, which was achieved with LINCS algorithm (
      • Hess B.
      • Bekker H.
      • Berendsen H.J.C.
      • Fraaije J.G.E.M.
      LINCS: a linear constraint solver for molecular simulations.
      ), and the long-range electrostatics was handled with Particle-mesh-Ewald (
      • Darden T.
      • York D.
      • Pedersen L.
      Particle mesh Ewald: an N ⋅log( N ) method for Ewald sums in large systems.
      ) as implemented in GROMACS. The simulation trajectories were analyzed using software VMD (
      • Humphrey W.
      • Dalke A.
      • Schulten K.
      VMD: visual molecular dynamics.
      ).

      Data availability

      All representative data are contained within the article.

      Supporting information

      This article contains supporting information.

      Conflict of interest

      Authors declare that they have no competing interests.

      Acknowledgments

      This work benefited from access to the Astbury Centre for Structural Molecular Biology at the University of Leeds (UK), an Instruct-ERIC centre. Financial support was provided by Instruct-ERIC (PID 12751 and 10661). Molecular graphics and analyses were performed with UCSF Chimera, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (NIH P41-GM103311). We are thankful to the Center for Scientific Computing (CSC), Finland for providing high-performance computing resources. We thank Dr Brendan Farrell for his guidance in EM grid preparation, David P. Klebl, Dr Anna J. Higgins, and Dr Miriam Walden (University of Leeds, UK) for their help and suggestions in cryo-EM data processing. The FEI Titan Krios microscopes were funded by the University of Leeds ( UoL ABSL award) and Wellcome Trust ( 108466/Z/15/Z ). We also thank the EM facility at the University of Leeds for cryo-EM data collection and the Electron Microscopy Unit, Institute of Biotechnology, Helsinki, for providing laboratory facilities, Benita Löflund and Pasi Laurinmäki (University of Helsinki, Finland) for technical assistance in cryo-EM. The facilities and expertise of the HiLIFE CryoEM, NMR, and Crystallization units at the University of Helsinki, a member of Instruct-ERIC Centre Finland, FINStruct, and Biocenter Finland are gratefully acknowledged

      Author contributions

      Y. L. and A. G. conceptualization; Y. L., O. H., G. B. C., V. S., S. P. M., and A. G. methodology; Y. L., O. C. R., and O. H. investigation; Y. L. and O. H. visualization; Y. L. and O. C. R. validation; Y. L. formal analysis; Y. L., V. S., and A. G. funding acquisition; Y. L. writing–original draft; Y. L., O. C. R., O. H., G. B. C., V. S., S. P. M., and A. G. writing–review and editing; V. S., S. P. M., and A. G. resources; V. S., S. P. M., and A. G. supervision.

      Funding and additional information

      This work is supported by Instruct-ERIC ( PID 12751 and 10661 ), the European Molecular Biology Organization short term fellowship (No. 8476 ), the Academy of Finland (No. 322609 ), the Academy of Finland (No. 286429 ), the Biotechnology and Biological Sciences Research Council ( BBSRC, BB/M021610/1 ), the Academy of Finland (No. 294652 ), the Sigrid Jusélius Foundation , the Jane and Aatos Erkko Foundation , and the Magnus Ehrnrooth Foundation .

      Supporting information

      References

        • Mulligan L.M.
        • Eng C.
        • Healey C.S.
        • Clayton D.
        • Kwok J.B.J.
        • Gardner E.
        • et al.
        Specific mutations of the RET proto-oncogene are related to disease phenotype in MEN 2A and FMTC.
        Nat. Genet. 1994; 6: 70-74
        • Eng C.
        • Clayton D.
        • Schuffenecker I.
        • Lenoir G.M.
        • Cote G.
        • Gagel R.F.
        • et al.
        The relationship between specific RET proto-oncogene mutations and disease phenotype in multiple endocrine neoplasia type 2.
        JAMA. 1996; 276: 1575
        • Wagner S.
        • Zhu S.
        • Nicolescu A.
        • Mulligan L.
        Molecular mechanisms of RET receptor-mediated oncogenesis in multiple endocrine neoplasia 2.
        Clinics. 2012; 67: 77-84
        • Wells S.A.
        • Asa S.L.
        • Dralle H.
        • Elisei R.
        • Evans D.B.
        • Gagel R.F.
        • et al.
        Revised American thyroid association guidelines for the management of medullary thyroid carcinoma.
        Thyroid. 2015; 25: 567-610
        • Shah M.H.
        • Sherman E.J.
        • Robinson B.
        • Solomon B.J.
        • Kang H.
        • Lorch J.H.
        • et al.
        Selpercatinib (LOXO-292) in patients with RET -mutant medullary thyroid cancer.
        J. Clin. Oncol. 2020; 38: 3594
        • Markham A.
        Selpercatinib: first approval.
        Drugs. 2020; 80: 1119-1124
        • Kotzbauer P.T.
        • Lampe P.A.
        • Heuckeroth R.O.
        • Golden J.P.
        • Creedon D.J.
        • Johnson Jr., E.M.
        • et al.
        Neurturin, a relative of glial-cell-line-derived neurotrophic factor.
        Nature. 1996; 384: 467-470
        • Baloh R.H.
        • Tansey M.G.
        • Lampe P.A.
        • Fahrner T.J.
        • Enomoto H.
        • Simburger K.S.
        • et al.
        Artemin, a novel member of the GDNF ligand family, supports peripheral and central neurons and signals through the GFRα3-RET receptor complex.
        Neuron. 1998; 21: 1291-1302
        • Milbrandt J.
        • De Sauvage F.J.
        • Fahrner T.J.
        • Baloh R.H.
        • Leitner M.L.
        • Tansey M.G.
        • et al.
        Persephin, a novel neurotrophic factor related to GDNF and neurturin.
        Neuron. 1998; 20: 245-253
        • Trupp M.
        • Arenas E.
        • Fainzilber M.
        • Nilsson A.-S.
        • Sieber B.-A.
        • Grigoriou M.
        • et al.
        Functional receptor for GDNF encoded by the c-ret proto-oncogene.
        Nature. 1996; 381: 785-789
        • Durbec P.
        • Marcos-Gutierrez C.V.
        • Kilkenny C.
        • Grigoriou M.
        • Wartiowaara K.
        • Suvanto P.
        • et al.
        GDNF signalling through the Ret receptor tyrosine kinase.
        Nature. 1996; 381: 789-793
        • Hsu J.-Y.
        • Crawley S.
        • Chen M.
        • Ayupova D.A.
        • Lindhout D.A.
        • Higbee J.
        • et al.
        Non-homeostatic body weight regulation through a brainstem-restricted receptor for GDF15.
        Nature. 2017; 550: 255-259
        • Mullican S.E.
        • Lin-Schmidt X.
        • Chin C.-N.
        • Chavez J.A.
        • Furman J.L.
        • Armstrong A.A.
        • et al.
        GFRAL is the receptor for GDF15 and the ligand promotes weight loss in mice and nonhuman primates.
        Nat. Med. 2017; 23: 1150-1157
        • Yang L.
        • Chang C.-C.
        • Sun Z.
        • Madsen D.
        • Zhu H.
        • Padkjær S.B.
        • et al.
        GFRAL is the receptor for GDF15 and is required for the anti-obesity effects of the ligand.
        Nat. Med. 2017; 23: 1158-1166
        • Emmerson P.J.
        • Wang F.
        • Du Y.
        • Liu Q.
        • Pickard R.T.
        • Gonciarz M.D.
        • et al.
        The metabolic effects of GDF15 are mediated by the orphan receptor GFRAL.
        Nat. Med. 2017; 23: 1215-1219
        • Li J.
        • Shang G.
        • Chen Y.-J.
        • Brautigam C.A.
        • Liou J.
        • Zhang X.
        • et al.
        Cryo-EM analyses reveal the common mechanism and diversification in the activation of RET by different ligands.
        Elife. 2019; 8e47650
        • Lin L.
        • Doherty D.
        • Lile J.
        • Bektesh S.
        • Collins F.
        GDNF: a glial cell line-derived neurotrophic factor for midbrain dopaminergic neurons.
        Science. 1993; 260: 1130-1132
        • Patel S.
        • Alvarez-Guaita A.
        • Melvin A.
        • Rimmington D.
        • Dattilo A.
        • Miedzybrodzka E.L.
        • et al.
        GDF15 provides an endocrine signal of nutritional stress in mice and humans.
        Cell Metab. 2019; 29: 707-718.e8
        • Parkash V.
        • Leppänen V.M.
        • Virtanen H.
        • Jurvansuu J.M.
        • Bespalov M.M.
        • Sidorova Y.A.
        • et al.
        The structure of the glial cell line-derived neurotrophic factor-coreceptor complex: insights into RET signaling and heparin binding.
        J. Biol. Chem. 2008; 283: 35164-35172
        • Coulpier M.
        • Anders J.
        • Ibáñez C.F.
        Coordinated activation of autophosphorylation sites in the RET receptor tyrosine kinase: importance of tyrosine 1062 for GDNF mediated neuronal differentiation and survival.
        J. Biol. Chem. 2002; 277: 1991-1999
        • Romei C.
        • Ciampi R.
        • Elisei R.
        A comprehensive overview of the role of the RET proto-oncogene in thyroid carcinoma.
        Nat. Rev. Endocrinol. 2016; 12: 192-202
        • Mograbi B.
        • Bocciardi R.
        • Bourget I.
        • Juhel T.
        • Farahi-Far D.
        • Romeo G.
        • et al.
        The sensitivity of activated cys Ret mutants to glial cell line-derived neurotrophic factor Is mandatory to rescue neuroectodermic cells from apoptosis.
        Mol. Cell Biol. 2001; 21: 6719-6730
        • Arighi E.
        • Popsueva A.
        • Degl’innocenti D.
        • Borrello M.G.
        • Carniti C.
        • Perälä N.M.
        • et al.
        Biological effects of the dual phenotypic Janus mutation of ret cosegregating with both multiple endocrine neoplasia type 2 and Hirschsprung’s disease.
        Mol. Endocrinol. 2004; 18: 1004-1017
        • Zhang M.
        • Liu Y.
        • Fu J.
        • Hu Y.
        • Sun Z.
        Functional analysis of RET with multiple endocrine neoplasia type 2.
        J. Cancer Res. Ther. 2017; 13: 823
        • Liu Y.
        • de Castro Ribeiro O.
        • Robinson J.
        • Goldman A.
        Expression and purification of the extracellular domain of wild-type humanRET and the dimeric oncogenic mutant C634R.
        Int. J. Biol. Macromol. 2020; 164: 1621-1630
        • Pintilie G.
        • Zhang Junjie
        • Chiu W.
        • Gossard D.
        Identifying components in 3D density maps of protein nanomachines by multi-scale segmentation.
        IEEE/NIH Life Sci. Syst. Appl. Workshop. 2009; 2009: 44-47
        • Bigalke J.M.
        • Aibara S.
        • Roth R.
        • Dahl G.
        • Gordon E.
        • Dorbéus S.
        • et al.
        Cryo-EM structure of the activated RET signaling complex reveals the importance of its cysteine-rich domain.
        Sci. Adv. 2019; 5eaau4202
        • Kjær S.
        • Hanrahan S.
        • Totty N.
        • McDonald N.Q.
        Mammal-restricted elements predispose human RET to folding impairment by HSCR mutations.
        Nat. Struct. Mol. Biol. 2010; 17: 726-731
        • Serna M.
        Hands on methods for high resolution cryo-electron microscopy structures of heterogeneous macromolecular complexes.
        Front. Mol. Biosci. 2019; 6: 33
        • Chappuis-Flament S.
        • Pasini A.
        • De Vita G.
        • Ségouffin-Cariou C.
        • Fusco A.
        • Attié T.
        • et al.
        Dual effect on the RET receptor of MEN 2 mutations affecting specific extracytoplasmic cysteines.
        Oncogene. 1998; 17: 2851-2861
        • Ito S.
        • Iwashita T.
        • Asai N.
        • Murakami H.
        • Iwata Y.
        • Sobue G.
        • et al.
        Biological properties of Ret with cysteine mutations correlate with multiple endocrine neoplasia type 2A, familial medullary thyroid carcinoma, and Hirschsprung’s disease phenotype.
        Cancer Res. 1997; 57: 2870-2872
        • Carlomagno F.
        • Salvatore G.
        • Cirafici A.M.
        • De Vita G.
        • Melillo R.M.
        • De Francisas V.
        • et al.
        The different RET-activating capability of mutations of cysteine 620 or cysteine 634 correlates with the multiple endocrine neoplasia type 2 disease phenotype.
        Cancer Res. 1997; 57: 391-395
        • Schlee S.
        • Carmillo P.
        • Whitty A.
        Quantitative analysis of the activation mechanism of the multicomponent growth-factor receptor Ret.
        Nat. Chem. Biol. 2006; 2: 636-644
        • Adams S.E.
        • Purkiss A.G.
        • Knowles P.P.
        • Nans A.
        • Briggs D.C.
        • Borg A.
        • et al.
        A two-site flexible clamp mechanism for RET-GDNF-GFRα1 assembly reveals both conformational adaptation and strict geometric spacing.
        Structure. 2021; https://doi.org/10.1016/j.str.2020.12.012
        • Berglund L.
        • Björling E.
        • Oksvold P.
        • Fagerberg L.
        • Asplund A.
        • Al-Khalili Szigyarto C.
        • et al.
        A genecentric human protein atlas for expression profiles based on antibodies.
        Mol. Cell. Proteomics. 2008; 7: 2019-2027
        • Tansey M.G.
        • Baloh R.H.
        • Milbrandt J.
        • Johnson E.M.
        GFRα-mediated localization of RET to lipid rafts is required for effective downstream signaling, differentiation, and neuronal survival.
        Neuron. 2000; 25: 611-623
        • Himanen J.P.
        • Yermekbayeva L.
        • Janes P.W.
        • Walker J.R.
        • Xu K.
        • Atapattu L.
        • et al.
        Architecture of Eph receptor clusters.
        Proc. Natl. Acad. Sci. U. S. A. 2010; 107: 10860-10865
        • Needham S.R.
        • Roberts S.K.
        • Arkhipov A.
        • Mysore V.P.
        • Tynan C.J.
        • Zanetti-Domingues L.C.
        • et al.
        EGFR oligomerization organizes kinase-active dimers into competent signalling platforms.
        Nat. Commun. 2016; 713307
        • Frêche B.
        • Guillaumot P.
        • Charmetant J.
        • Pelletier L.
        • Luquain C.
        • Christiansen D.
        • et al.
        Inducible dimerization of RET reveals a specific AKT deregulation in oncogenic signaling.
        J. Biol. Chem. 2005; 280: 36584-36591
        • Xiong Y.
        • Walker K.
        • Min X.
        • Hale C.
        • Tran T.
        • Komorowski R.
        • et al.
        Long-acting MIC-1/GDF15 molecules to treat obesity: evidence from mice to monkeys.
        Sci. Transl. Med. 2017; 9eaan8732
        • Wittig I.
        • Braun H.-P.
        • Schägger H.
        Blue native PAGE.
        Nat. Protoc. 2006; 1: 418-428
        • Liu Y.
        • Kaljunen H.
        • Pavić A.
        • Saarenpää T.
        • Himanen J.P.
        • Nikolov D.B.
        • et al.
        Binding of EphrinA5 to RET receptor tyrosine kinase: an in vitro study.
        PLoS One. 2018; 13e0198291
        • Schneider C.A.
        • Rasband W.S.
        • Eliceiri K.W.
        NIH image to ImageJ: 25 years of image analysis.
        Nat. Met. 2012; 9: 671-675
        • Tang G.
        • Peng L.
        • Baldwin P.R.
        • Mann D.S.
        • Jiang W.
        • Rees I.
        • et al.
        EMAN2: an extensible image processing suite for electron microscopy.
        J. Struct. Biol. 2007; 157: 38-46
        • Zheng S.Q.
        • Palovcak E.
        • Armache J.-P.
        • Verba K.A.
        • Cheng Y.
        • Agard D.A.
        MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy.
        Nat. Met. 2017; 14: 331-332
        • Zhang K.
        Gctf: Real-time CTF determination and correction.
        J. Struct. Biol. 2016; 193: 1-12
        • Wagner T.
        • Merino F.
        • Stabrin M.
        • Moriya T.
        • Antoni C.
        • Apelbaum A.
        • et al.
        SPHIRE-crYOLO is a fast and accurate fully automated particle picker for cryo-EM.
        Commun. Biol. 2019; 2: 218
        • Pettersen E.F.
        • Goddard T.D.
        • Huang C.C.
        • Couch G.S.
        • Greenblatt D.M.
        • Meng E.C.
        • et al.
        UCSF Chimera--A visualization system for exploratory research and analysis.
        J. Comput. Chem. 2004; 25: 1605-1612
        • Emsley P.
        • Lohkamp B.
        • Scott W.G.
        • Cowtan K.
        Features and development of Coot.
        Acta Crystallogr. Sect. D Biol. Crystallogr. 2010; 66: 486-501
        • Pandurangan A.P.
        • Topf M.
        Finding rigid bodies in protein structures: application to flexible fitting into cryoEM maps.
        J. Struct. Biol. 2012; 177: 520-531
        • Topf M.
        • Lasker K.
        • Webb B.
        • Wolfson H.
        • Chiu W.
        • Sali A.
        Protein structure fitting and refinement guided by cryo-EM density.
        Structure. 2008; 16: 295-307
        • Afonine P.V.
        • Poon B.K.
        • Read R.J.
        • Sobolev O.V.
        • Terwilliger T.C.
        • Urzhumtsev A.
        • et al.
        Real-space refinement in PHENIX for cryo-EM and crystallography.
        Acta Crystallogr. Sect. D Struct. Biol. 2018; 74: 531-544
        • Jo S.
        • Kim T.
        • Iyer V.G.
        • Im W.
        CHARMM-GUI: a web-based graphical user interface for CHARMM.
        J. Comput. Chem. 2008; 29: 1859-1865
        • Lee J.
        • Cheng X.
        • Swails J.M.
        • Yeom M.S.
        • Eastman P.K.
        • Lemkul J.A.
        • et al.
        CHARMM-GUI input generator for NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM simulations using the CHARMM36 additive force field.
        J. Chem. Theor. Comput. 2016; 12: 405-413
        • Brooks B.R.
        • Brooks C.L.
        • Mackerell A.D.
        • Nilsson L.
        • Petrella R.J.
        • Roux B.
        • et al.
        CHARMM: the biomolecular simulation program.
        J. Comput. Chem. 2009; 30: 1545-1614
        • Park S.J.
        • Lee J.
        • Qi Y.
        • Kern N.R.
        • Lee H.S.
        • Jo S.
        • et al.
        CHARMM-GUI glycan modeler for modeling and simulation of carbohydrates and glycoconjugates.
        Glycobiology. 2019; 29: 320-331
        • Park S.-J.
        • Lee J.
        • Patel D.S.
        • Ma H.
        • Lee H.S.
        • Jo S.
        • et al.
        Glycan Reader is improved to recognize most sugar types and chemical modifications in the Protein Data Bank.
        Bioinformatics. 2017; 33: 3051-3057
        • Jo S.
        • Song K.C.
        • Desaire H.
        • MacKerell A.D.
        • Im W.
        Glycan reader: automated sugar identification and simulation preparation for carbohydrates and glycoproteins.
        J. Comput. Chem. 2011; 32: 3135-3141
        • Páll S.
        • Abraham M.J.
        • Kutzner C.
        • Hess B.
        • Lindahl E.
        Tackling exascale software challenges in molecular dynamics simulations with GROMACS.
        in: Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 8759. Springer Verlag, Heidelberg, Germany2015: 3-27
        • Abraham M.J.
        • Murtola T.
        • Schulz R.
        • Páll S.
        • Smith J.C.
        • Hess B.
        • et al.
        GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputers.
        SoftwareX. 2015; 1–2: 19-25
        • Van Der Spoel D.
        • Lindahl E.
        • Hess B.
        • Groenhof G.
        • Mark A.E.
        • Berendsen H.J.C.
        GROMACS: fast, flexible, and free.
        J. Comput. Chem. 2005; 26: 1701-1718
        • Mackerell A.D.
        • Feig M.
        • Brooks C.L.
        Extending the treatment of backbone energetics in protein force fields: limitations of gas-phase quantum mechanics in reproducing protein conformational distributions in molecular dynamics simulations.
        J. Comput. Chem. 2004; 25: 1400-1415
        • MacKerell A.D.
        • Bashford D.
        • Bellott M.
        • Dunbrack R.L.
        • Evanseck J.D.
        • Field M.J.
        • et al.
        All-atom empirical potential for molecular modeling and dynamics studies of proteins.
        J. Phys. Chem. B. 1998; 102: 3586-3616
        • Nosé S.
        A unified formulation of the constant temperature molecular dynamics methods.
        J. Chem. Phys. 1984; 81: 511-519
        • Hoover W.G.
        Canonical dynamics: equilibrium phase-space distributions.
        Phys. Rev. A. 1985; 31: 1695-1697
        • Parrinello M.
        • Rahman A.
        Crystal structure and pair potentials: a molecular-dynamics study.
        Phys. Rev. Lett. 1980; 45: 1196-1199
        • Parrinello M.
        • Rahman A.
        Polymorphic transitions in single crystals: a new molecular dynamics method.
        J. Appl. Phys. 1981; 52: 7182-7190
        • Hess B.
        • Bekker H.
        • Berendsen H.J.C.
        • Fraaije J.G.E.M.
        LINCS: a linear constraint solver for molecular simulations.
        J. Comput. Chem. 1997; 18: 1463-1472
        • Darden T.
        • York D.
        • Pedersen L.
        Particle mesh Ewald: an N ⋅log( N ) method for Ewald sums in large systems.
        J. Chem. Phys. 1993; 98: 10089-10092
        • Humphrey W.
        • Dalke A.
        • Schulten K.
        VMD: visual molecular dynamics.
        J. Mol. Graph. 1996; 14: 33-38