Shared structural architecture of CTLH-CRA domains.

(a) Domain architecture of the core subunits containing LisH (L), CTLH, and CRA-motifs. (b) Crystal structure of the RanBP10-Twa1 complex (3.2 Å, PDB: 9SOC). LisH and CRAC helices are labeled. (c) Schematic representation of the modular CTLH core complex highlighting interchangeable subunits (RanBP9 ↔ RanBP10; muskelin ↔ Wdr26) and variable substrate receptors (Gid4-Armc8, Ypel5, and FAM72A), illustrating the combinatorial nature of subunit arrangements. Gray arrows mark CTLH-CRAN/CTLH-CRA domain heterodimerization interfaces in a composite model of known structures (PDB: 8PJN, 7NSC, 8QBN, 8TTQ, and newly solved 9SOC) fitted into available cryo-EM maps (EMDB: 12542, 12545, 12547, 17713, 17715, 18172, 45088, 45138, and 45186). (d) Schematic overview of CTLH-CRAN/CTLH-CRA domain fusion constructs showing their position and binding specificity within the CTLH core complex. Crystal structures of RanBP9, RanBP10, Twa1, Maea, and muskelin (from RanBP9-muskelin complex) reveal a conserved fold of the CTLH-CRA domain.

Structural determinants of RanBP9/10-muskelin CTLH-CRA specificity.

(a) 2.0 Å crystal structure of the RanBP9-muskelin CTLH-CRA domains (PDB: 9SNV) with close-up view of the interaction site. (b) Sequence alignment of CRA-interface residues highlighting the conserved leucine (marked *). (c) Molecular details of the RanBP9-muskelin interface overlaid with RanBP10 (PDB: 9SNF) and Twa1 (PDB: 9SOC). (d) ITC binding analysis of muskelin with various RanBP10 mutants and the muskelin mutant F686L with RanBP10. (e) Overlay of the RanBP9-muskelin interface with Maea (PDB: 9SNH). (f) SEC-MALS analysis of Maea and the double mutant (Y272A, Y275A). (g) 1.4 Å crystal structure of the Maea CTLH-CRA domain (PDB: 9SNH) showing details of the homodimerization interface.

Rewiring the binding specificity of RanBP10 from muskelin to Maea.

(a) SEC-MALS, (b) ITC, and (c) NAGE binding analysis of CTLH-CRA domains of muskelin and Maea with wild-type RanBP10, R10_G (Q519G), and R10_GLFF (Q519G, F543L, Y548F, V556F) mutants. (d) 1.7 Å crystal structure of the R10_GLFF CTLH-CRA domain (PDB: 9SNG) showing details of the homodimerization interface. Twa1-like mutations are highlighted in yellow. (e) 3.5 Å crystal structure of the R10_GLFF-Maea CTLH-CRA complex (PDB: 9SOH) and overlay (f) with the RanBP9-muskelin CTLH-CRA complex (PDB: 9SNV) illustrating the determinants of binding specificity.

Rewiring the binding specificity of Twa1 from Maea to muskelin.

(a) NAGE binding analysis of various Twa1 mutants to Maea and muskelin, accompanied by a depiction of Twa1 mutants sequence changes. (b) 3.3 Å crystal structure of the Twa1_14-Maea complex (PDB: 9SOI) showing molecular details of the interface, overlaid with muskelin (PDB: 9SNV). (c) SEC-MALS analysis of wild-type Twa1, Twa1_14, and Twa1_52 in complex with Maea or muskelin. (d) ITC analysis demonstrating that Twa1_14 and Twa1_52 bind muskelin, in contrast to wild-type Twa1.

Picomolar RanBP9/10 binding to muskelin and Wdr26.

(a) Analytical size exclusion chromatography of RanBP9/10-Twa1 complexes in the presence of muskelin or Wdr26. (b) SEC-MALS analysis of co-expressed CTLH-CRA domain complexes Twa1_52-Wdr26 and Twa1_52-muskelin. Theoretical MW is denoted (theo.). (c) ITC binding analysis of Twa1_52 to muskelin/Wdr26, followed by competitive titration with RanBP9/10-Twa1 complexes. KD values and stoichiometries (n) are indicated. (d) Cryo-EM maps of the mammalian CTLH complex fitted with available structural data highlights the CTLH-CRA dimerization interface in the context of the full-length assembly. An AlphaFold3 prediction of the RanBP9-Wdr26 CTLH-CRA domain overlaid with the RanBP9-muskelin structure (PDB: 9SNV) illustrates their conserved mode of specificity.

CTLH-CRA domain specificity code.

Crystal structures determined in this study of CTLH-CRA domains highlight the specificity of the interaction interfaces and the position of the conserved leucine residue (*) for comparison. Mutational analyses demonstrate that binding specificity can be rewired bidirectionally: the RanBP10 quadruple mutant R10_GLFF (Q519G, F543L, Y548F, and V556F) successfully switched its preference from muskelin to Maea, while the engineered Twa1_52 mutant (A125G, Q126R, Δ127T, L140K, E144K, L147F, F152Y, and F160V) acquired RanBP10-like specificity toward muskelin. Together, these structures and functional swaps define the residue-level code underlying partner recognition within CTLH-CRA interfaces.

Data collection and refinement statistics of crystal structures from CTLH complex core subunits.

Comparison of LisH-CRAC interfaces.

Sequence alignment and logo (top) highlight conserved hydrophobic residues, like leucine or phenylalanine (marked with * or circled *), within the LisH motif. Gray shading marks the folding of the who helices, which explains a conserved glycine residue in the loop between them. Another conserved residue near the C-terminus is a glutamate which mediates backbone interactions with the opposing helix. Structural comparison of available LisH-CRAC dimers reveals a conserved helical arrangement. Muskelin was omitted from this comparison because it lacks a CRAC helix, while Lis1 and TBL1 have a CRAC helix despite lacking a CRAN segment. Black lines between the RanBP10 and Twa1 sequences indicate potential hydrogen bonds that may explain why Twa1 preferentially binds RanBP10 rather than forming homodimers.

Biochemical characterization of different CTLH-CRA domain complexes.

(a) SEC-MALS analyses of co-expressed CTLH-CRA domain complexes of RanBP9-muskelin and RanBP10-muskelin (shortened C-terminus). Measured molar masses match theoretical values (theo.), thus confirming a defined 1:1 stoichiometry. (b) SEC-MALS measurements of full-length dimeric Twa1, co-expressed with CTLH-CRA domains of Maea or Rmnd5a, reveal comparable 1:1 binding ratios. Twa1 homodimerization (LisH-CRAC-mediated) remains unaffected. Rmnd5a CTLH-CRA domain is soluble when co-expressed with Twa1. (c) Native agarose gel electrophoreses (NAGE) of RanBP10 mutants carrying combinations of Twa1-like substitutions (F543L, Y548F, V556F; L--, -F-, -FF, LF-, L-F, -FF, LFF) illustrate the progressive loss of muskelin binding with increasing substitutions (compare ITC, Fig. 2). (d) SEC-MALS analysis of ITC-derived complexes (Fig. 5). Addition of excess RanBP9-Twa1 to pre-assembled Twa1_52-muskelin or Twa1_52-Wdr26 complexes results in displacement of monomeric Twa1_52 and formation of muskelin/Wdr26-RanBP9-Twa1 complexes. The oligomeric state of muskelin (tetramer) and of Wdr26 (dimer) remain preserved.

ITC binding analysis of muskelin and RanBP10.

Representative ITC titrations of CTLH-CRA domains of muskelin and RanBP10. Muskelin F686L mutant was titrated against wild-type (wt) RanBP10, and muskelin wt was titrated against RanBP10 binding mutants (F543L, Y548F, V556F; and combination thereof: L--, -F-, -FF, LF-, L-F, -FF, LFF), corresponding to Fig. 2d and 3b. KD values and molar ratios (n) are shown as mean ± standard error from 3-4 replicates (6 for wt). Based on enthalpy and KD values, entropy (ΔS) and Gibbs free energy (ΔG) were calculated, revealing both enthalpic and entropic contributions to the binding reaction. Binding of RanBP10-Twa1 complex to full-length muskelin was too tight and could not be fitted.

Concentration-depended oligomerization of Maea and RanBP10 Q519G.

(a) SEC-MALS molar mass profiles of Maea CTLH-CRA domain across the elution peak illustrate dynamic concentration-dependent oligomerization. The most stable 0.3 ml peak fraction was used for analysis (middle panel). Molar masses were plotted against concentration, derived from the protein amount based on the refractive index (RI) signal in the peak. KD best-fit parameter and 95 % confidence interval of the one-site binding model are indicated, representing the dimerization of stable dimers. The theoretical molar mass (theo.) of a dimer was used for normalization. (b) Equivalent analysis of the RanBP10 Q519G mutant. (c) Muskelin-binding-impaired RanBP10 mutants (F543L, Y548F, V556F; and combinations thereof: L--, -F-, -FF, LF-, L-F, -FF, LFF) and wt RanBP10 are stable monomeric CTLH-CRA domains. (d) Structural overlay of CTLH-CRA domain interfaces of wt RanBP9 (similar to wt RanBP10), R10_GLFF, and Maea shows that Q552 (Q519 in RanBP10) restricts Maea binding and dimerization compared to Q519G mutation.

Sequence overview of Twa1 CTLH-CRA domain mutants.

To swap binding specificity and render the Twa1 CRA interface more RanBP9/10-like, a total of 52 different Twa1 mutants were purified, and tested for binding (see Suppl. Fig. 6).

NAGE binding analysis of Twa1 CTLH-CRA domain mutants.

Binding of 52 different Twa1 mutants (Suppl. Fig. 5) was analyzed at 200 μM against Maea CTLH-CRA (native binding partner) and muskelin CTLH-CRA (acquired RanBP9/10-like binding preference). NAGE served as a screening assay to identify promising candidates for subsequent biophysical validation.

ITC binding analysis of muskelin and Twa1 mutants.

ITC binding analysis analogous to Suppl. Fig. 3, in which muskelin was titrated against Twa1 CTLH-CRA domain mutants (see Suppl. Fig. 5). Representative titrations corresponding to Fig. 4d are shown. KD values and molar ratios (n) are presented as mean ± standard error from 3-4 replicates. Based on calculated entropy (ΔS) and Gibbs free energy (ΔG), the optimization process reduced unfavorable enthalpic contributions to binding.

Overlay of AlphaFold3-predicted CTLH-CRA interfaces across species.

AlphaFold3 predictions of human CTLH-CRA interfaces (Twa1-Rmnd5a, Twa1-Maea, RanBP9-muskelin, RanBP9-Wdr26, RanBP10-muskelin, and RanBP10-Wdr26) were compared with corresponding homologous interfaces from model organisms: Mus musculus, Xenopus laevis, Danio rerio, Caenorhabditis elegans, Drosophila melanogaster, Saccharomyces cerevisiae, and Arabidopsis thaliana (top to bottom). UniProt accession numbers are indicated in gray. Only protein sequences lacking the LisH-CRAC segment were used to avoid false interface prediction. When Rmnd5a was present, the closely related paralog Rmnd5b was omitted due to their high similarity (human, mouse). In D. rerio, which lacks Rmnd5a, the paralog Rmnd5b was used for prediction. This species also encodes two Twa1 homologs, Gid8a and Gid8b; only Gid8a is shown for clarity. D. rerio further contains a RanBP10 paralog, which X. laevis lacks. For S. cerevisiae, the experimental cryo-EM structure of Gid 1 (RanBP9 homolog) in complex with Gid7 (Wdr26 homolog) was used instead of a prediction for comparison (see also Suppl. Fig. 9) For M. musculus, the prediction was replaced by the experimental RanBP9-muskelin crystal structure.

CTLH-CRA interface of the yeast Gid1-Gid7.

(a) Cryo-EM structure of the Saccharomyces cerevisiae Gid complex (PDB: 7NSB), the yeast homolog of the CTLH complex. (b) Structural overlay of murine RanBP9 (from PDB: 9SNV) with its yeast homolog Gid1. (c) Close-up view of the Gid1-Gid7 CTLH-CRA interface. The N-terminus of Gid1 forms an additional contact with the Gid7 interface. The conserved leucine residue (*) is indicated for reference.

Non-CTLH complex CTLH-CRA domains.

(a) Structures of CRA interfaces from non-CTLH complex proteins – SMU1, Wdr47, TOPLESS, and TPR2 – reveal a conserved fold extending beyond the CTLH complex. (b) Sequence alignment based on CTLH-CRA domain structures highlights three positional classes of the conserved aromatic residue (dark gray). Overlay of Maea and TRP2 homodimers illustrates that in non-CTLH complex members, the aromatic residue is shifted to the opposite side of the interface, resulting in altered loop geometry. (c) Structures of SMU1, Wdr47, and TRP2 demonstrate that the LisH-CTLH-CRA module can engage additional binding partners beyond the canonical LisH-CRAC and CTLH-CRA interfaces. RED and IAA1 peptides bind to distinct hydrophobic grooves adjacent to the dimerization interface – interaction sites not observed in CTLH complex subunits.

Electrostatic surface representation of CTLH-CRA domains used in this study.

Although CTLH-CRA domains share a highly conserved overall fold, their electrostatic surface potential reveal differences beyond the conserved dimerization interface. These variations occur in additional surface grooves that differ in both charge and shape and may represent underexplored, subunit-specific interaction sites. While Twa1 displays an overall negatively charged surface, RanBP9 exhibits a comparatively neutral groove located C-terminally (left panel) to the dimerization interface.