Multiple weak linear motifs enhance recruitment and processivity in SPOP-mediated substrate ubiquitination

Abstract

Primary sequence motifs, with millimolar affinities for binding partners, are abundant in disordered protein regions. In multivalent interactions, such weak linear motifs can cooperate to recruit binding partners via avidity effects. If linear motifs recruit modifying enzymes, optimal placement of weak motifs may regulate access to modification sites. Weak motifs may thus exert stronger physiological relevance than suggested by their affinities, but molecular mechanisms of their function are still poorly understood. Herein, we use the N-terminal disordered region of the Hedgehog transcriptional regulator Gli3 (Gli3^1-90) to determine the role of weak motifs encoded in its primary sequence for the recruitment of its ubiquitin ligase CRL3^SPOP and the subsequent effect on ubiquitination efficiency. The substrate adaptor SPOP binds linear motifs through its MATH domain and forms higher-order oligomers through its oligomerization domains, rendering SPOP multivalent for its substrates. Gli3 has multiple weak SPOP binding motifs. We map three such motifs in Gli3^1-90, the weakest of which has a millimolar dissociation constant. Multivalency of ligase and substrate for each other facilitates enhanced ligase recruitment and stimulates Gli3^1-90 ubiquitination in in vitro ubiquitination assays. We speculate that the weak motifs enable processivity through avidity effects and by providing steric access to lysine residues that are otherwise not prioritized for polyubiquitination. Weak motifs may generally be employed in multivalent systems to act as gatekeepers in posttranslational modification.

Graphical Abstract

Introduction

Intrinsically disordered and flexible protein regions contain an abundance of primary sequence motifs that function to recruit binding partners, direct modifying enzymes and rewire signaling networks ^1–8. Many of these linear motifs are only weakly conserved and may have affinities for their binding partners that are too weak to probe by high throughput methods. This raises the question of how strong an individual interaction must be to have a physiological function. If multiple linear motifs in a protein interact with repeats of interaction domains in a binding partner, described as so-called multivalent interactions, even weak individual interactions can contribute through avidity effects ^9–14. Such multivalent interactions may thus serve to increase the affinity and specificity of modular interactions ¹⁵. In the context of substrate/enzyme interactions, multivalency can serve to regulate access to modification sites. Weak motifs may position an enzyme favorably relative to modification sites and thus contribute to the overall efficiency of the reaction to a higher degree than expected from their affinities. In multi-turnover reactions, avidity effects have the potential to reduce substrate off-rates and may thus serve to increase the processivity ¹⁶.

Many ubiquitin ligases recruit substrates by recognition of linear motifs (reviewed in ¹⁷). The existence of multiple weak motifs in substrates has been attributed to precise regulation of substrate levels, e.g. via the creation of switch-like dose response curves ^{18, 19}. Some ubiquitin ligases oligomerize and form dimers, tetramers or even higher-order oligomers ^{20–25, 16}. Abolishing their oligomerization often results in a decrease in their ubiquitination activity ^{14, 26, 27}. The combination of an oligomeric ubiquitin ligase with a substrate with multiple weak motifs may create interesting opportunities for regulation.

Herein, we characterize the function of weak linear motifs in a substrate for recruitment by the oligomeric substrate adaptor Speckle-type POZ protein (SPOP) of the cullin3-RING ligase (CRL3) ^28–31. SPOP recognizes primary sequence motifs with the consensus sequence Φ-π-S-S/T-S/T (Φ, nonpolar residue; π, polar residue) in its substrates and binds them in the substrate-binding groove of its MATH (for meprin and TRAF homology) domain (Fig. 1a,b) ³². SPOP recruits additional CRL3 subunits through its 3-box ³² and dimerizes through its BTB (for Bric à Brac, Tramtrack and Broad-Complex) domain; crystal structures of these dimers are known (Fig. 1b) ^{32, 16}. Recently, an additional oligomerization function has been assigned to the C-terminal BACK (for BTB and C-terminal Kelch) domain ^{16, 33} (Fig. 1a). The presence of 2 oligomerization domains facilitates self-association into higher-order SPOP homo-oligomers ¹⁶ (Fig. 1c), rendering SPOP multivalent for substrates. The SPOP substrates Gli2 and Gli3, which are transcriptional regulators of the Hedgehog signaling pathway, contain several regions that aid in their recognition by SPOP ³⁴. SPOP and SPOP substrates are thus multivalent for each other. Single strong binding motifs for monomeric ubiquitin ligases can efficiently target substrates for degradation ¹⁷. The multivalency of SPOP/substrate pairs suggests an additional layer of regulation of substrate levels.

Figure 1. Multivalency of Gli3 and SPOP.

(a) Cartoon diagram of SPOP. SPOP contains a MATH domain for substrate recognition, a BTB domain for dimerization ³², and a BACK domain with an additional oligomerization function ¹⁶. We used constructs encoding monomeric SPOP^MATH (28-166), dimeric SPOP^MATH-BTB (28-337) ³², and higher-order oligomer-forming SPOP^{MATH-BTB-BACK} (28-359) ¹⁶. (b) Ribbon diagram of the asymmetric SPOP^MATH-BTB dimer bound to 2 SPOP binding motif peptides, shown in orange in stick representation, PDB ID 3HQI ³². One monomer in the ribbon diagram is colored more lightly. (c) Schematic of the domain structure of SPOP dimer (left) and SPOP oligomer (right). The linear SPOP oligomer schematic is based on the dimer structures of the BTB and BACK domains ^{33, 32} and indicates the possibility for indefinite self-association. (d) Gli3 has a folded zinc finger (Zn finger) DNA-binding domain that encompasses residues 455-636; all other regions are predicted to be largely intrinsically disordered (see Figure S1, Supplementary Material). By using the consensus SPOP binding sequence (Φ-π-S-S/T-S/T; Φ, nonpolar residue, π, polar residue) ³² and allowing for 1 mismatch, we predicted 29 weak consensus SB motifs, which are shown as red bars.

SPOP mutations have been recently identified in endometrial and colorectal cancers ^35–37 among others, and SPOP is the most frequently mutated gene in prostate cancer ³⁵. Most SPOP mutations are clustered in the substrate-binding groove and abolish substrate binding ^{35, 38, 39}. Loss of SPOP expression occurs frequently in colorectal, gastric, and prostate tumors ⁴⁰. Therefore, SPOP acts as a tumor suppressor under physiological conditions ⁴¹, and loss-of-function mutations or a redirection of SPOP activity to a different set of substrates serves to drive tumorigenesis ^{42, 43}. Dissecting the mechanism of substrate recognition by SPOP is important to understand its normal function and role in pathogenesis.

In this study we used biophysical techniques and functional assays to characterize the role of weak SPOP binding motifs in a disordered region of Gli3 for recruitment of SPOP and for CRL3^SPOP-mediated ubiquitination. We used nuclear magnetic resonance (NMR) spectroscopy, fluorescence anisotropy (FA) binding assays, and analytical ultracentrifugation (AUC) to map SPOP-binding (SB) motifs in a model multivalent substrate. We determined affinities to monomeric forms of SPOP by Biolayer interferometry (BLI) and assessed how SPOP oligomerization affects recognition of the multivalent substrate. In vitro ubiquitination assays provided insight into the role of multivalency and the contribution of individual weak motifs for polyubiquitination. These studies provide understanding of the functional role of weak SPOP binding motifs in a multivalent interaction.

Results

To gain molecular insights into the biophysical mechanism of the multivalent interaction between Gli3 and SPOP, we first analyzed the primary sequence of Gli3 and predicted 29 SB motifs based on the known SB consensus motif sequence Φ-π-S-S/T-S/T ³², if we allowed for 1 mismatched position (Fig. 1d). These motifs are likely solvent accessible; apart from the folded zinc finger DNA-binding domain (residues 455-636 according to sequence alignment with Gli1 ⁴⁴), Gli3 is predicted to be largely intrinsically disordered (see Supplementary Fig. S1). The interaction of a highly multivalent substrate with oligomeric SPOP is likely complicated. To gain insight into this interaction, we characterized the interaction of the 90-residue N-terminal Gli3 fragment (Gli3^1-90), which we predicted to have 4 SB motifs, with monomeric, dimeric and oligomeric variants of SPOP. We determined the position and affinities of SB motifs and how they contribute to SPOP recruitment and Gli3 ubiquitination.

A multivalent Gli3 fragment binds to the MATH domain substrate-binding groove

The MATH domain of SPOP (SPOP^MATH) binds SB motifs in its groove across an antiparallel β-sheet (Figure 1b)³². To investigate whether multivalent substrates bind to SPOP^MATH via the groove alone or whether they have additional interactions with other regions of SPOP^MATH, we titrated Gli3^1-90 into ¹⁵N-labeled SPOP^MATH and followed perturbations of SPOP^MATH amide resonances by NMR (Fig. 2a). Resonance assignments of the free and bound states of SPOP^MATH (Supplementary Fig. S2) allowed us to calculate chemical shift perturbations (Fig. 2b). Strong perturbations were observed only for residues clustered around the canonical groove (Fig. 2c). Residues known to directly contact the SB motif in crystal structures of SPOP^MATH/motif complexes underwent the largest chemical shift perturbations. These perturbations were in slow exchange on the NMR time scale as observed by the progressive appearance of a bound-state signal while the signal of the free state disappeared (Supplementary Fig. S3). Specifically, residues Y87, F102, F125, K129, W131, F133 and K134, which are mutated in prostate cancer³⁵, exhibited the largest perturbations. This result confirms the crucial role of these residues in recognizing substrates and supports previous findings that SPOP^MATH mutants found in prostate cancer have substrate binding defects ^{38, 39}. Binding of a peptide derived from the Drosophila melanogaster protein Puckered (pPuc, Ac⁹¹–ENLACDEVTSTTSSST¹⁰⁶–NH₂), which contains one known strong SB motif, produced similar chemical shift perturbations (Figs. 2d–f). Additional perturbations mapping to the back of SPOP^MATH were small and observed for binding of both Gli3^1-90 and pPuc and therefore likely do not indicate additional interactions but rather small conformational changes of SPOP^MATH upon ligand binding. These findings confirm that multivalent Gli3^1-90 interacts with SPOP^MATH via its canonical substrate-binding groove.

Figure 2. Gli3^1-90 is recognized in the canonical MATH domain substrate-binding groove.

(a,d) ¹⁵N TROSY-HSQC spectra of free ¹⁵N SPOP^MATH (black) and its complexes with (a) multivalent Gli3^1-90 (red) and (d) monovalent Puc^91-106 (blue). (b, e) Combined ¹H and ¹⁵N chemical shift perturbations of ¹⁵N SPOP^MATH upon binding (b) Gli3^1-90 or (e) Puc^91-106 demonstrate the involvement of the same interface. (c, f) Chemical shift perturbations mapped onto the surface of SPOP^MATH in a color code from white to red (Gli3^1-90, c) or white to blue (Puc^91-106, f) show that residues forming the canonical groove undergo the largest perturbations. Residues frequently mutated in prostate cancer cases are labeled ³⁵.

Gli3^1-90 is intrinsically disordered

To map the location of SB motifs in the Gli3^1-90 sequence (Fig. 3a), we used NMR spectroscopy and first assigned the Gli3^1-90 resonances (Fig. 3b) from spectra recorded at 5° C where maximum chemical shift dispersion was observed. NMR spectroscopy is a powerful method for characterizing conformational and dynamic properties of IDPs ^45–47. The narrow line widths and small ¹H^N chemical shift dispersion of the amide proton signals ^{48, 49} showed that this region of Gli3 is disordered, in agreement with in silico predictions (see Supplementary Fig. S1) and previous reports ⁵⁰.

Figure 3. Gli3^1-90 is intrinsically disordered.

(a) Primary sequence of the Gli3^1-90 fragment. (b) ¹⁵N HSQC spectrum with resonance assignments, BMRB ID 26575. (c) ¹H-¹⁵N hetNOE values at 25 °C, and (d) Secondary structure propensity (SSP) scores ⁵³ calculated from C^α and C^β chemical shifts demonstrate that Gli3^1-90 is intrinsically disordered and samples transient structure.

¹H-¹⁵N heteronuclear NOE (hetNOE) values, which provide information on fast protein dynamics and flexibility ⁵¹, were mostly negative when measured at 600 MHz and 25 °C and slightly positive at the center of the protein at 800 MHz (Fig. 3c). The modulation of the hetNOE by the location in the polypeptide chain and the relatively strong modulation by the field strength was expected for an intrinsically disordered protein and indicated that the protein was flexible with some motional restriction ⁵². The region between resides 45 and 60, which exhibits the highest hetNOE values, also has higher transverse relaxation rates than the rest of the protein (Supplementary Fig. S4) suggesting the presence of transient interactions within this region.

Secondary structural propensities calculated from C^α and C^β chemical shifts ⁵³ demonstrated the presence of stretches with ≥10% helicity for Gli3^6-12, Gli3^18-29, Gli3^36-43, and Gli3^47-54 and up to 20% helicity for Gli3^77-86. The rest of the protein had a propensity to sample extended conformations (Fig. 3d). The disordered nature of Gli3^1-90 likely provides accessibility for SPOP to SB motifs for binding partners, including SPOP.

Gli3^1-90 has three weak SB motifs

In mapping the location of SB motifs in Gli3^1-90, we wanted to avoid bias towards the discovery of motifs with the known consensus motif, because weak binding motifs with distinct sequences may be able to contribute to the multivalent interaction with SPOP. We therefore implemented a three-pronged approach consisting of (1) a titration of SPOP^MATH into ¹⁵N-labeled Gli3^1-90; (2) division of the candidate sequences into 15 residue-long peptides and assessment of their ability to compete with pPuc for SPOP^MATH binding in a secondary fluorescence anisotropy (FA) binding assays; and (3) a tertiary NMR binding assay to monitor binding of peptides to ¹⁵N SPOP^MATH. This combined approach allowed the identification of SB motifs and the exclusion of false- positive candidate motifs.

First, we titrated deuterated SPOP^MATH into ¹⁵N-labeled Gli3^1-90 up to a 3.2-fold molar ratio and assigned the resonances of the complex. Resonances in three regions, at the N-terminus, in the middle and close to the C-terminus of the protein, were not assignable. We monitored chemical shift perturbations in ¹H-¹⁵N heteronuclear single quantum coherence (HSQC) spectra. As expected, many resonances, in particular serine and threonine resonances (see Supplementary Fig. S5), experienced chemical shift perturbations because SB motifs are serine/threonine-rich. Chemical shift perturbations were largest around the regions that could not be assigned in the complex, suggesting that these regions contained SB motifs (Fig. 4a).

Figure 4. Gli3^1-90 can bind up to three SPOP^MATH molecules.

(a) Chemical shift perturbations ((ΔH² + 0.2ΔN²)^1/2) for Gli3^1-90 upon SPOP^MATH binding (at a molar ratio of 1:3.2). Residues shown in grey could not be assigned in the complex spectrum. These regions correspond to the three SPOP binding motifs in Gli3^1-90. (b) Peptides containing candidate SB motifs compete with fluorescein-Puc^91-106 for binding to SPOP^MATH in a fluorescence anisotropy competition assay. [L] is the peptide concentration. Solid lines are non-linear least squares fits to a complete competitive binding model ⁵⁴. See Table 1 for K_D values. (c) Two-dimensional size-and-shape distribution analysis of SV AUC data of a Gli3^1-90-SPOP^MATH mixture. Shown are contour plots of c(M,f/f₀) of the transformed velocity data with 0 fringes/S (white) to maximum value fringes/S (red), with increasing color temperature indicating higher values. The dotted lines are lines of constant s-value.

We then generated 15-residue-long peptides covering these regions and used a FA competition assay ⁵⁴ to test their ability to bind to the SPOP^MATH groove by competing with fluorescein-labeled pPuc (F-pPuc). The assay yielded a dissociation constant (K_D) of 7.7 μM for pPuc (Table 1), which is in good agreement with previous data ³². Peptides encompassing regions Gli3^1-15, Gli3^30-45, and Gli3^70-84 competed with pPuc for binding to SPOP^MATH and are referred to as SBM1, SBM2, and SBM3, respectively, hereafter (Fig. 4b and Table 1). We further confirmed the interaction of some of these peptides by monitoring chemical shift perturbations in ¹H-¹⁵N HSQC spectra of ¹⁵N SPOP^MATH. The peptides elicited similar perturbation patterns, but perturbations were weaker than those for pPuc or Gli3^1-90, suggesting individual motifs had weak binding affinities (see Supplementary Fig. S6).

Table 1.

Peptide affinities to SPOP^MATH

Motif	Residue No.	Sequence	KD (μM)a
pPuc	91-106	Ac-ENLACDEVTSTTSSST-NH2	7.7 ± 0.2
SBM1	1-15	Ac-MEAQSHSSTTTEKK-NH2	~4100
Gli316-30	16-30	Ac-VENSIVKCSTRTDVS-NH2	nbb
SBM2	31-45	Ac-EKAVASSTTSNEDES-NH2	80 ± 3
mSBM2	31-45	Ac-EKAVAGGSGSNEDES-NH2	nbb
SBM3	71-84	Ac-KVSEEPSTSSDERA-NH2	560 ± 30
Gli341-59	41-59	Ac-NEDESPGQTYHRERRNAIT-NH2	nbb

^aErrors represent standard errors from triplicate experiments from FA competition experiments.

^bNo binding detected.

The binding affinities of SBM2 and SBM3 to SPOP^MATH were weak, 80 μM and 560 μM, respectively (Table 1). As expected, mutation of the Ser/Thr-rich region in SBM2 to a generic disordered sequence (VASSTTS to VAGGSGS, previously used by Zhuang et al. to abolish substrate binding ³²) decreased binding to below the detection limit of the assay. Although we were able to monitor the binding to SBM1 effectively by NMR, we estimated from FA that its K_D was as high as ~4 mM. SBM1 lacks the hydrophobic residue at the motif N-terminus, probably accounting for its weak affinity. The order in which the signals at the resonance positions of the free state in individual motifs lost signal intensity in NMR titrations was consistent with the relative motif affinities measured for individual peptides by FA competition assays, indicating that transient structure in Gli3^1-90 did not alter the overall order of SB motif affinities (Supplementary Fig. 7).

The regions N- and C-terminal to SBM2 showed chemical shift perturbations as well, and so we tested peptides Gli3^16-30 and Gli3^41-59; neither peptide bound to SPOP^MATH in the FA assay (Table 1, Figure 4b). Gli3^16-30 does not contain a conventional SB motif; it contains a cysteine in place of the serine that was previously suggested to be invariable (VKCST instead of Φ-π-S-S/T-S/T) ³². We speculated that this sequence might bind SPOP^MATH, but did not observe consistent binding by FA- or NMR-detected titration (Fig. 4b and Supplementary Fig. S6). Gli3^41-59 does not contain an SB motif-like sequence. In the unbound state, this region of Gli3^1-90 has the highest hetNOE and R₂ values (Fig. 3c and Supplementary Fig. S4), and therefore perturbation of its resonances may reflect a change in intramolecular interactions within Gli3^1-90 rather than intermolecular interactions with SPOP^{MATH 55, 56}.

Together, these NMR and FA binding experiments show that we have identified 3 specific SB motifs in the N-terminus of Gli3. Among mammals, the entire N-terminus of Gli3 is highly conserved, while among vertebrates, the sequence similarity in the N- terminus is highest within the SB motifs (Supplementary Fig. S8), supporting the hypothesis that these motifs play a role in recruiting CRL3^SPOP.

Gli3^1-90 can simultaneously bind up to three MATH domain molecules

To derive the stoichiometry of binding, we used sedimentation velocity (SV) AUC in combination with two-dimensional size-and-shape distribution analysis, which allows the determination of accurate molecular weights of several species with different frictional coefficients in a complex mixture ⁵⁷. In the presence of an excess of SPOP^MATH, up to 3 MATH domain molecules were able to bind to 1 Gli3^1-90 molecule (Fig. 4c). Furthermore, we observed both free SPOP^MATH and a signal at higher molecular weights in the f/f₀ value range of 1.2 to 1.6. We interpreted the latter to reflect a mixture of different 1:1 and 1:2 complexes, all likely having similar average molecular shapes and sizes. The 1:3 complex is easily distinguishable from the other signals because its larger M_w is combined with a large f/f₀ value of 2–2.5, which we attribute to an extended and possibly rigid shape. SV analysis of two dilutions of the complex showed reduction of the fractional population of the largest complex as expected and confirms that the data are self-consistent (Supplementary Fig. S9). These data support the notion that Gli3^1-90 can interact with 1 SPOP^MATH molecule through each of its 3 SB motifs demonstrating that they are accessible simultaneously.

Do SPOP dimers exhibit cooperative substrate binding?

Next, we explored whether the interaction of dimeric SPOP with substrates was dominated by avidity or cooperativity effects. We expected to find avidity effects in the SPOP/Gli3^1-90 interaction, i.e. the first SB motif/MATH interaction would result in an increased local concentration of additional interaction partners and thus enhance binding. However, the binding of ligands to oligomeric proteins is sometimes cooperative i.e., depends on the number of ligands already bound ^58–60. A monovalent ligand cannot exhibit avidity toward dimeric SPOP, and can therefore be used to probe for cooperativity.

We used a direct FA binding assay in which we titrated either the SPOP^MATH domain or the SPOP^MATH-BTB dimer into the fluorescently labeled Puc peptide. Both interactions yielded the same K_D of ~6 μM within error (Fig. 5). It is unlikely that the K_D for the second MATH domain becomes too large to measure once the first MATH domain is bound, because a crystal structure of the SPOP dimer with 2 bound pPuc peptides exists ³² (see Fig. 1b). Therefore, the affinity of one MATH domain in the SPOP dimer is independent of the free or bound status of the other MATH domain. We conclude that the SPOP dimer does not exhibit macroscopic cooperativity toward monovalent substrates. Avidity effects may instead play a role in its interactions with multivalent substrates.

Figure 5. SPOP dimer binds monovalent substrate without cooperativity.

Direct fluorescence anisotropy binding data for monomeric SPOP^MATH (black) and dimeric SPOP^MATH-BTB (red) to monovalent Puc^91-106. [P] indicates the protein concentration.

Multivalent interactions afford avidity

To test for the existence of avidity effects, we determined the apparent affinities of SPOP constructs with increasing multivalency, i.e. monomeric, dimeric and higher-order oligomeric SPOP, to Gli3^1-90. Higher-order oligomeric SPOP accesses a distribution of different oligomers depending on the protein concentration, though the dimer is always the predominant species (Marzahn, Mittag; unpublished data). We used a BLI assay in which we immobilized His-tagged Gli3^1-90 on the sensor and monitored binding of the three SPOP constructs to it. Their concentrations were normalized to the number of protomers in solution, i.e. solutions of monomeric, dimeric and oligomeric SPOP at the same concentration contained the same number of binding sites for SB motifs.

For SPOP^MATH we obtained a K_D of 21 μM (Table 2 and Supplementary Fig. S10). The increased apparent affinity, when compared with the highest microscopic affinity (80 μM for SBM2, Fig. 4b, Table 1), may be caused by (i) the high local concentration of motifs and therefore an increased probability for a MATH domain molecule to rebind after its release; (ii) an entropic advantage because of a greater number of different bound states in a multivalent substrate; or (iii) modifications of the affinities from flanking sequences outside of the core binding SB motifs (e.g. through long-range electrostatic interactions).

Table 2.

Gli3^1-90 affinities to SPOP constructs of different association state from kinetic fits of BLI data

SPOP construct	association state	Kd (μM)a
SPOPMATH	monomer	21±2
SPOPMATH-BTB	dimer	28±6
SPOPMATH-BTB-BACK	higher-order oligomer	1.5±0.1

^aErrors represent standard errors from triplicate experiments from BLI.

With dimeric SPOP^MATH-BTB, we do not see a significant change in the affinity compared to the monomeric SPOP^MATH. While we had expected an increased affinity due to engagement of two SB motifs via the two MATH domains, these data indicate that the SB motifs in Gli3^1-90 are not adequately spaced for avidity to the SPOP dimer. In contrast, the higher-order oligomer-forming SPOP^{MATH-BTB-BACK} exhibited increased binding to Gli^1-90 with a K_D of 1.5 μM for Gli3^1-90 (Table 2). The multivalency of both proteins for each other thus led to an affinity increase due to avidity effects.

Contribution of weak SB motifs to ubiquitination

As we observed increased substrate binding by oligomeric SPOP, we wondered whether this would also lead to enhanced CRL3^SPOP-mediated ubiquitination activity. We utilized His-Gli3^1-90 in standard in vitro ubiquitination assays with recombinant substrate, ubiquitin, E1, E2, and neddylated CRL3^{SPOP 32}. Transfer of ubiquitin moieties onto lysine residues of His-Gli3^1-90 were monitored by Western Blot against the His-tag. A mixture of multi-mono- or poly-ubiquitinated species resulted in a regular laddering appearance (Fig. 6a, b, Supplementary Fig. S11). We compared CRL3^SPOP activity in the context of dimeric SPOP^MATH-BTB vs. oligomeric SPOP^{MATH-BTB-BACK}. We observed efficient ubiquitination after a 20 min reaction for oligomeric SPOP but smaller ubiquitinated species with dimeric SPOP (Fig. 6a). These results are in agreement with previously reported enhancement of ubiquitination efficiency upon SPOP dimerization ³² and oligomerization ¹⁶. Together with our analysis of substrate binding, the data suggest that the multivalency of oligomeric SPOP and Gli3^1-90 for each other enhances recruitment of the substrate, thereby enhancing ubiquitination efficiency.

Figure 6. Weak SB motifs stimulate CRL3^SPOP-mediated in vitro ubiquitination of Gli3^1-90.

(a) Western blot showing in vitro ubiquitination of His-Gli3^1-90 by neddylated CRL^SPOP with dimeric SPOP^MATH-BTB or oligomeric SPOP^{MATH-BTB-BACK}. His-Gli3^1-90 is detected with an anti-6x-His antibody. (b, d) Western blot showing in vitro ubiquitination of wild-type (wt) and mutant His-Gli3^1-90 substrates by neddylated CRL^SPOP with oligomeric SPOP^{MATH-BTB-BACK} with (b) wt ubiquitin or (d) a lysine-less ubiquitin variant (Ub noK), which sustains only mono-ubiquitination on multiple lysines. mSBM1, mSBM2, mSBM3, and combinations refer to inactivating mutations of the 3 SB motifs. (c) Schematic representation of ubiquitin-modified lysine residues (yellow) in wt Gli3^1-90 substrate. (e) Model proposing the role of weak SB motifs and SPOP oligomerization in substrate recruitment and polyubiquitination. Dimeric SPOP recruits substrates with lower affinity and may not provide suitable steric access to lysine residues on growing polyubiquitin chains. In contrast, oligomeric CRL^SPOP (a tetramer is shown for clarity) mediates enhanced recruitment via avidity effects and effective polyubiquitination through sterically favorable positioning relative to multiple catalytic centers in the oligomeric CRL. SB motifs are depicted as pink bars, the color saturation decreases for weaker motifs. Additional CRL components are not depicted for clarity.

We had observed a range of affinities for SB motif binding to SPOP and therefore postulated that the highly conserved motifs, even though weak, would play a distinct functional role. To test this hypothesis, we generated mutant Gli3^1-90 constructs that carried inactivating mutations ³² in 1, 2 or all 3 SB motifs. Strikingly, the mutation of SBM1, which had the weakest binding affinity, resulted in a substantial attenuation of ubiquitination (Fig. 6b). Mutation of SBM2, which had the strongest binding affinity and was expected to essentially abolish the recruitment of SPOP, only had a moderate effect on ubiquitination levels (Fig. 6b). In general, combinations of mutations led to increasing loss of ubiquitination. Interestingly, a Gli3^1-90 mutant that carried only SBM1 sustained relatively efficient ubiquitination. These results demonstrate that even weak motifs in Gli3^1-90 contribute to ubiquitination by CRL3^SPOP. Reducing the multivalency, either by restricting the SPOP oligomerization state to dimers or by inactivating SB motifs in Gli3^1-90, reduced the overall ubiquitination efficiency.

Several studies have shown the importance of an optimal placement of acceptor lysines relative to the catalytic cysteine in ubiquitin ligases ^61–64. We speculated that the unexpected loss of ubiquitination in the SBM1 mutant, and the lesser sensitivity of the SBM2 mutation, reflected the accessibility of modifiable lysine residues in Gli3^1-90. We characterized the lysine residues that were ubiquitinated in wt Gli3^1-90 by using tryptic digestion and liquid chromatography - tandem mass spectrometry (LC-MS/MS) analysis of reaction mixtures and identified 5 ubiquitinated lysine residues, K15, K22, K32, K70 and K87 (Fig. 6c for summary, Supplementary Fig. S12). We observed ubiquitination of the same lysine residues also for all mutants we tested, which included mSBM1, mSBM2, and mSBM23. While these experiments did not inform on the extent of ubiquitination at individual lysine residues, the results showed that not all SB motifs needed to be present simultaneously to provide access for lysine modification.

To further characterize whether the ubiquitination defect in some of the Gli3 mutants was a result of the modification of fewer lysine residues or less efficient building of polyubiquitin chains on all modifiable lysines, we next carried out ubiquitination reactions with a lysine-less ubiquitin mutant (Ub noK, in which all lysines are replaced with arginines). Ub noK does not support building of polyubiquitin chains but supports mono-ubiquitination at individual lysine residues. The reactions were carried out for 1 hr with higher E1 concentrations to reach an endpoint that showed the maximum number of modifiable lysine residues. We observed up to 5 modified lysine residues on wt Gli3^1-90 (Fig. 6d) in agreement with the LC-MS/MS results. Mutating one SB motif had only minor effects, with substrates still containing ~4 ubiquitinated lysines, although the ubiquitination patterns varied slightly among mutants, as observed by the bands on the Western blot (Fig. 6d). Most Gli3^1-90 mutants with less than 2 SB motifs were ubiquitinated on fewer lysines. The mSBM13 mutant, containing the highest affinity SBM2 motif, however, sustained ubiquitination on 4 lysine residues. The mSBM23 mutant, which contains only the weakest affinity SBM1 motif, was only modified on ~1–2 lysines. This is in contrast to the results with wt ubiquitin, which show that this mutant is still actively ubiquitinated (Fig 6b). Therefore, the weakest motif, SBM1, is required to promote efficient polyubiquitination of Gli3^1-90.

These results demonstrate that monoubiquitination is sustainable by the highest affinity motif, but that the presence of additional motifs, even if they have millimolar affinities, drives effective polyubiquitination, likely by enabling processivity of the ligase.

Discussion

Herein, we have demonstrated that weak linear binding motifs can strengthen protein/protein interactions and mediate function in multivalent systems, even if their dissociation constants are in the millimolar range. We identified three SB motifs in the intrinsically disordered N-terminus of Gli3; the strongest motif had a microscopic K_D of 80 μM, while the others were in the hundreds of micromolar to millimolar range (Table 1). Together, they mediated binding to oligomeric SPOP at a low micromolar affinity (Table 2). These observations support a model in which dual multivalency, i.e. multivalency of each binding partner for the other, enhances binding through avidity effects, where the resulting binding is stronger than each individual interaction. The likely mechanism is the increase of the local concentration of a second pair of binding partners once the first interaction is made ^{11, 9}. Interestingly, SPOP dimerization did not enhance binding of Gli3^1-90 compared to the SPOP monomer (Table 2). Although Gli3^1-90 supports binding of three MATH domain molecules simultaneously (Figure 4c), the SB motifs may not be optimally spaced for avidity with SPOP dimers.

In addition to the role of multivalency in substrate recruitment, multivalency was required for the full activity in an in vitro ubiquitination assay. Inactivating either SPOP’s ability to oligomerize or inactivating the weakest Gli3^1-90 SB motif led to a defect in Gli3^1-90 ubiquitination. However, all lysine residues that were ubiquitinated in wt Gli3^1-90 were also available for modification in the mutants. Therefore, we propose that the defect stems from a lack of processivity of CRL3^SPOP, caused by the following two mechanisms: (1) Mutant substrates with lower affinities, due to a lack of avidity with SPOP, experience higher off-rates from the ligase and consequently lower ubiquitination rates; and (2) Binding to three SB motifs mediates sterically favorable placement of multiple catalytic ligase centers to accommodate the changing distances in a growing ubiquitin chain. We conclude that SBM1, the weakest binding motif, plays a particularly important role in positioning the substrate relative to catalytic cysteines because of its disproportionate effect on polyubiquitination. For mSBM23, the mutant that only harbors SBM1, monoubiquitination is reduced to 1 site (Fig. 6d), but polyubiquitination of this mutant is more effective than that for others (Fig. 6b). We speculate that SPOP oligomerization has a similar functional impact as adding additional weak SB motifs to a substrate, i.e. that multivalency of the substrate and CRL3^SPOP for each other are decisive for processivity and thus polyubiquitination.

Our data suggest that even motifs that vary substantially from the consensus sequence support substrate ubiquitination, warranting a broader search for substrates than possible by sequence comparison with known motifs. How SPOP achieves substrate specificity despite its ability to recognize very weak motifs remains to be investigated.

Previous work showed that CRL3^SPOP-mediated in vitro ubiquitination of Puc is driven by its strongest motif and that contribution of the 2 weaker motifs is negligible ³². The inactivity of the weak motifs in Puc might be explained by the use of dimeric SPOP in the previous report and the concurrent lack of dual multivalency and avidity. The absolute affinity of the motifs in Puc vs Gli3^1-90 also differ; the strong Puc motif is the strongest known SB motif with a K_D of ~4 μM and may therefore dominate SPOP recruitment, even in the absence of SPOP oligomerization. We speculate that weak motifs are only functional in the context of highly multivalent substrates. They may serve to enable ultrasensitive substrate concentration/degradation responses ⁶⁵. In contrast, we speculate that substrates with few strong motifs are polyubiquitinated constitutively because they are readily recruited to the ligase.

In addition to avidity, cooperativity is a possible mechanism for enhancing binding to oligomeric proteins. As cooperativity requires concerted changes of conformation or dynamics within the oligomeric protein upon binding the first binding partner, avidity may be simpler to realize through the use of a disordered protein with multiple motifs. The ubiquitous nature of disordered regions and linear motifs within them may attest to this fact.

Extrapolating from the 3 SB motifs we have mapped in Gli3^1-90, assuming constant density of SB motifs in all disordered regions, we would expect up to 46 SB motifs in full-length Gli3. Obviously, this deduction may not reflect reality; the density of SB motifs may be higher in the N-terminus. Nevertheless, previous reports have mapped several SPOP-recruiting regions in different parts of Gli3 ³⁴. In fact, a high density of short linear motifs has been reported in disordered regions of some proteins as a functional benefit and is therefore not unexpected^{66, 67}. We expect that the presence of many weak motifs in combination with higher-order oligomerization of a binding partner enables tight regulation of the interaction. In fact, higher-order oligomerization of enzymes and the assembly of signaling machines are emerging as a new paradigm of signal transduction and are thought to mediate signal amplification, ultrasensitive responses, noise reduction, and exquisite temporal and spatial signaling control (reviewed in⁶⁵).

The dysregulation of the Gli3/SPOP interaction may be expected to affect Gli3 turnover. Mutation of S8 and S36 in Gli3 have been reported in cancers (see www.cBioPortal.org, ^{68, 69}) and may negatively affect SPOP recruitment because these serines are located in motifs characterized herein, SBM1 and SBM2 (Figure 3a, b). SB motifs in other parts of Gli3 not characterized here may be affected similarly. Though mutation of several SB motifs in a substrate is an unlikely event, single point mutations in SPOP can greatly affect its multivalency. Indeed, mutations of BTB and BACK domain interface residues that may affect SPOP’s ability to oligomerize have been found in sequencing efforts of melanoma and uterine cancers (see www.cBioPortal.org, ^68–71). Mutations in the substrate binding groove of the MATH domain are usually not accompanied by loss of heterozygosity ³⁵. In the presence of wt and mutant SPOP, SPOP hetero-oligomers with reduced multivalency must form, because wt protomers are able to bind SB motifs while mutant protomers are not. While levels of substrates with many weak linear motifs may be exquisitely regulated in healthy cells, in the presence of wt/mutant SPOP hetero-oligomers, their turnover is likely impacted more strongly than that of substrates with individual strong motifs. SPOP mutations may thus affect subsets of substrates, potentially changing substrate selectivity ⁴³.

Despite tremendous progress in our understanding of individual systems, the rules of lysine selection by E2/E3/ubiquitin-like modifier trios ⁶², and for achieving monoubiquitination and polyubiquitination ^72–74 are still poorly understood in general. The presence of multiple linear motifs have the potential to drive lysine selection and determine the extent of polyubiquitination as we suggest here; altering the number of ‘active’ linear motifs by posttranslational modification or conformational changes thus adds an additional possible regulatory layer. If the extent of ligase oligomerization is regulated as well, as has been suggested previously ¹⁶, astounding combinatorial possibilities for substrate selection ensue.

Weak linear motifs, when appearing in multiples in disordered protein regions, have the potential for extensive function. A deeper understanding of the regulation of CRL3^SPOP activity toward multivalent substrates will ultimately provide insight into protein homeostasis in health and disease.

Methods

Constructs and cell cultures

The coding sequence for a Gli3 fragment spanning residues 1-90 (Gli3^1-90) was cloned into a pET28a vector and overexpressed in E. coli BL21 (DE3) RIPL cells (Agilent Technologies) in LB medium at 37°C with shaking at 280 rpm, and expression was induced with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). The expression of GST-SPOP^MATH (residues 28-166) and dimeric His-MBP-SPOP^MATH-BTB (residues 28-337) has been previously described ³². Briefly, these constructs, GST-SPOP^MATH-BTB and His-SUMO-SPOP^{MATH-BTB-BACK} (residues 28-359) were expressed in BL21 GOLD cells (Agilent Technologies) in LB medium for unlabeled samples or M9 minimal medium supplemented with¹⁵ N NH₄Cl and/or ¹³C glucose, depending on the desired labeling scheme. Expression was induced with 0.6 mM IPTG at OD₆₀₀ ~ 0.6 at 18°C with shaking at 280 rpm overnight. Cells were harvested by centrifugation and lysed with a Microfluidizer (Microfluidics).

Protein purification

His-tagged Gli3^1-90 was purified using 20 mL Ni-NTA resin in buffer containing 50 mM Tris pH 8, 300 mM NaCl, 30 mM imidazole, and 2 mM β-mercaptoethanol (β-ME), and eluted in buffer containing 25 mM Tris pH 8, 300 mM NaCl, 300 mM imidazole, and 2 mM β-ME. The resulting protein was cleaved with tobacco etch virus (TEV) protease under dialysis at 4°C overnight against 25 mM Tris pH 8, 500 mM NaCl, 20 mM imidazole, and 5 mM β-ME, or the His-tag left for BLI and ubiquitination assays. The cleaved product was passed through Ni-NTA resin to remove uncleaved protein and TEV protease.

GST-SPOP^MATH, GST-SPOP^MATH-BTB and GST-SPOP^{MATH-BTB-BACK} were purified using glutathione-Sepharose in PBS pH 7.3 and 5 mM DTT and eluted with PBS pH 7.3, 5 mM DTT, and 10 mM reduced L-glutathione. His-MBP-SPOP^MATH-BTB was purified on 20 mL of Ni-NTA resin, followed by simultaneous TEV cleavage and dialysis at 4°C into PBS pH 7.6 and 5 mM DTT overnight. Ion exchange chromatography (IEX) to separate the cleaved tags from SPOP was performed on a 5 mL HiTrap SP column (GE Healthcare, for SPOP^MATH) or a 5ml HiTrap Q column (for SPOP^MATH-BTB), using a gradient from 0 to 700 mM NaCl.

All proteins were further purified by size exclusion chromatography on a HiLoad 16/60 Superdex 75 or 200 gel filtration column in 20 mM Tris pH 7.6, 150 mM NaCl, and 5 mM DTT or in NMR buffer (see below).

The protein purity was at least 95%, as shown by SDS-PAGE and SV AUC (see Supplementary Fig. S13). Protein identities were confirmed by both top-down and bottom up mass spectrometry.

Peptide synthesis

Fifteen residue peptides encompassing the SBMs were synthesized through the Hartwell Center, St. Jude Children’s Research Hospital, or purchased from GenScript. Each peptide allows for ~4 residues preceding and following the consensus sequence. All peptides were modified by N-terminal acetylation and C-terminal amidation.

Fluorescence anisotropy

An N-terminally fluorescently labeled Puc peptide (F-pPuc) with sequence ⁹¹Ac-ENLACDEVTSTTSSST-NH ¹⁰⁷ ₂ was purchased from GenScript. For direct FA binding assays, increasing concentrations of SPOP^MATH and SPOP^MATH-BTB were titrated into 40 nM F-pPuc in a buffer containing 20 mM Tris pH 7.4, 150 mM NaCl, 1% bovine serum albumin (BSA, Sigma), and 0.01% Triton, and the FA monitored with an EnVision multilabel plate reader (PerkinElmer) at 25°C. For competition experiments, increasing concentrations of each peptide were individually titrated into a mixture of 6 μM SPOP^MATH and 40 nM F-pPuc and the FA was monitored. Analysis was performed as described previously ⁵⁴.

Nuclear magnetic resonance spectroscopy

NMR data were acquired on Bruker Avance 600 and 800 MHz spectrometers equipped with TCI triple-resonance cryogenic probes and pulsed-field gradient units. All samples were prepared in an NMR buffer consisting of 50 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.76 mM KH₂PO₄, and 5 mM DTT pH 6.0 at 5°C. For assignment, a 1.5 mM ¹⁵N Gli3^1-90 sample was used to acquire standard triple-resonance backbone assignment experiments. These included a HNCACB and CBCA(CO)NH with 16 scans and 8 scans, respectively, and 1024 (¹H) × 28 (¹⁵N) × 60 (¹³C) complex data points, with 10 ppm, 22 ppm, and 70 ppm for ¹H, ¹⁵N and ¹³C sweep width, respectively. HNCO and HN(CA)CO were both acquired with 8 scans, 1024 (¹H) × 20 (¹⁵N) × 64 (¹³C) complex data points, with 10 ppm, 22 ppm, and 22 ppm for ¹H, ¹⁵N and ¹³C sweep widths, respectively.

Initial SPOP^MATH assignments were obtained from the authors who solved the solution NMR structure of SPOP^MATH (pdb: 2CR2) and the assignments were confirmed using 3D HNCA (1024 (¹H) x 24 (¹⁵N) x 40 (¹³C) complex data points, with 14 ppm (¹H), 32 ppm (¹⁵N) and 32 ppm (¹³C) sweep width) and CBCA(CO)NH (1024 (¹H) x 24 (¹⁵N) x 44 (¹³C) complex data points, with 14 ppm (¹H), 32 ppm (¹⁵N) and 72 ppm (¹³C) sweep width) experiments measured with 8 and 16 scans, respectively. Backbone resonances of residues E46, K74, G75, S96, S119-Y123 were not observed in the spectrum and therefore were unassigned. Since the binding pocket residues exhibited slow exchange in the SPOP^MATH-Puc (1:5) complex, this spectrum was reassigned using 3D HNCA, HNCO, HN(CA)CO and HNCACB experiments. For all of the 3D experiments the data were collected at 298 K for samples containing 0.5 mM SPOP^MATH in NMR buffer. The positions of the resonances in the binding pocket that were in slow exchange in SPOP^MATH-Gli3^1-90 (1:2) and SPOP^MATH-SBM2 (1:4) complexes were very close to that of SPOP^MATH-Puc complex and thus were assigned as for the SPOP^MATH-Puc complex. Data were processed using BRUKER Topspin version 3.2, NMRPipe (v.7.9) ⁷⁵ and analyzed using CARA (v.1.8.4) ⁷⁶.

¹H,¹⁵N TROSY-HSQC spectra of 350 μM ¹⁵N SPOP^MATH in NMR buffer with increasing concentrations of Gli3^1-90 (to molar ratios of 1:1.75) were collected with 1024 (¹H) × 90 (¹⁵N) complex data points and 32 scans with 12 ppm and 29 ppm for ¹H and ¹⁵N sweep widths, respectively. Similar spectra were recorded for the complexes in the presence of different peptides, with protein/peptide molar ratios of 1:5 for Puc, 1:8 for Gli3^16-30, 1:4 for SBM2, and 1:4 for SBM3.

Gradient-selected sensitivity-enhanced ¹H,¹⁵N HSQC spectra of 400 μM ¹⁵N Gli3^1-90 in NMR buffer with increasing concentrations of SPOP^MATH (to molar ratios of 1:3) were recorded with 1024 × 150 complex data points and 16 scans. Intensity ratios for each resonance were calculated from the intensities in the free state and from intensities at increasing molar ratios at the same resonance frequencies to combine chemical shift perturbations and broadening into a single parameter of intensity loss. hetNOEs were collected with a 2-sec relaxation delay and with and without a 5-sec presaturation delay, using 1024 × 200 complex data points and 32 scans with 10 ppm, and 24 ppm for ¹H, and ¹⁵N sweep widths, respectively. The hetNOE values were calculated from the ratio of peak intensities of the saturated and unsaturated spectra. All the spectra were referenced directly using DSS for the ¹H dimension and ¹³C and ¹⁵N frequencies were referenced indirectly.

Secondary structural propensities were calculated by using ¹³C^α and ¹³C^β chemical shifts using the SSP program ⁵³.

Analytical ultracentrifugation

Sedimentation velocity experiments of all protein samples were conducted in a ProteomeLab XL-I analytical ultracentrifuge (Beckman Coulter, Indianapolis, IN) following standard protocols unless mentioned otherwise ^{57, 77}. A mixture of Gli3^1-90 (156 μM)/SPOP^MATH (525 μM) and a dilution series thereof, in the ultracentrifugation buffer (100 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.76 mM KH₂PO₄, and 5 mM DTT at pH 6.0) was loaded into a cell assembly comprised of a double sector charcoal-filled centerpiece with a 12 mm path length and sapphire windows. The cell assembly, containing identical sample and reference buffer volumes of 400 μL, was placed in a rotor and temperature equilibrated at rest at 20 °C for 2 hours before it was accelerated from 0 to 50,000 rpm. Rayleigh interference optical data were collected continuously for 10 hours. The velocity data were modeled with diffusion-deconvoluted sedimentation coefficient distributions c(s) in SEDFIT (https://sedfitsedphat.nibib.nih.gov/software/default.aspx), using algebraic noise decomposition and with the signal-average frictional ratio and meniscus position refined with non-linear regression. Maximum entropy regularization was applied at a confidence level of P-0.70. A two-dimensional size-and-shape distribution model, c(s,f/f₀) (with the one dimension the s-distribution and the other the f/f₀-distribution) was also calculated with the same interference fringe displaced velocity data. The equidistant f/f₀-grid was from 1.0 to 3.0 with 0.13 steps, the linear s-grid from 0.5 to 5 S with 100 s-values, and Tikhonov-Phillips regularization at one standard deviation was applied. The data were transformed to a c(M,f/f₀) contour distribution plot with M the molecular mass and f/f₀ the frictional ratio. The dotted lines indicate lines of constant s-value. The distributions were not normalized ^{57, 78}.

BLI

Binding affinities of Gli3^1-90 to SPOP^MATH, SPOP^MATH-BTB, and SPOP^{MATH-BTB-BACK} were measured by BLI using an OctetRED instrument (ForteBio). All proteins used for BLI were buffer exchanged with 20 mM HEPES, 150 mM NaCl, 5mM DTT, 0.005% Tween 20, pH 7.4. N-terminal His₆-Gli3^1-90 (0.1 μM) was immobilized on Ni-NTA biosensor (ForteBio) for 300 seconds. For baseline and surface stability, the captured His₆-Gli3^1-90 was cross-linked for 60 seconds with a mixture of 0.1 M ethyl(dimethylaminopropyl) carbodiimide (EDC) and 0.025 M N-Hydroxysuccinimide (NHS) and subsequently quenched with 1 M ethanolamine for 60 seconds. Biosensors with immobilized His₆-Gli3^1-90 were subsequently dipped into different concentrations of SPOP samples for association until the signal reached a plateau. Binding of SPOP^{MATH-BTB-BACK} was performed at concentrations of 0, 0.01, 0.05, 0.1, 0.5, 1, 2.5 and 5 μM, binding of SPOP^MATH-BTB and SPOP^MATH was performed at concentrations of 0, 0.1, 0.5, 1, 2.5, 5, 10 and 20 μM. The concentrations of SPOP constructs with different valency were normalized to the number of protomers in solution, i.e. solutions of 1 μM monomeric, dimeric and oligomeric SPOP contained the same number of binding sites for SB motifs. Dissociation of SPOP was monitored by subsequently incubating the biosensors in the buffer (20 mM HEPES, 150 mM NaCl, 5mM DTT, 0.005% Tween 20, pH 7.4). The sensorgram at each SPOP concentration was background subtracted using a reference sensorgram (obtained from biosensors without immobilized Gli3^1-90). The dissociation constants were obtained by fitting the data to a kinetic model for one site binding.

In vitro ubiquitination assay

Gli3^1-90 ubiquitination was sampled in 50 mM Tris pH 7.5, 150 mM NaCl, 10 mM MgCl₂, 5 mM ATP, 1 mM DTT, and 2 mg/mL BSA at room temperature at time points from 0 to 20 min. The reaction mixture contained ubiquitinating enzymes at final concentrations of 0.25 μM UBA1 (E1; in assays with lysine-less ubiquitin 1 μM UBA1 to reach maximum ubiquitination more rapidly), 5 μM UbcH5B (E2), 5 μM NEDD8-CUL3-Rbx1 (E3), 5 μM SPOP (either SPOP^{MATH-BTB-BACK} or SPOP^MATH-BTB as substrate adaptor), 75 μM ubiquitin or a lysine-less ubiquitin mutant (both BostonBiochem) and 5 μM His₆-tagged Gli3^1-90 or mutant versions. E1, E2, and E3 were purified as described previously ^{79, 80, 16}. The substrate and products were visualized by Western blot with anti-His antibody. The mutant versions of Gli3^1-90 were obtained by site-directed mutagenesis by replacing the Ser/Thr-rich regions with the generic disordered sequence GGSGS.

Mass spectrometry analysis of ubiquitination patterns

Ubiquitination reactions were run on an SDS PAGE gel, and proteins in each lane were excised, reduced with DTT to break disulfide bonds and Cys residues were alkylated by iodoacetamide to allow the recovery of Cys-containing peptides. The gel bands were washed, dried in a vacuum concentrator and rehydrated with a buffer containing trypsin. Samples were digested overnight and the peptides were extracted using acetonitrile. The extracts were dried and each sample was then resuspended in 5% formic acid. For MS analysis, the dried peptides were analyzed on an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific) after separation on a 40cm X 75 μm id column packed with 1.9 μm C18 resin (Dr. Maisch GmbH, Germany). Separation was achieved by applying 10–40% buffer B gradient in 2 h (buffer A: 0.2% formic acid; buffer B: buffer A plus 70% ACN). The column was heated at 65°C by a butterfly portfolio heater (Phoenix S&T) to reduce backpressure. The mass spectrometer was operated in data-dependent and targeted mode with a survey scan in Orbitrap (60 000 resolution, 2 X 10⁵ AGC target and 50 ms maximal ion time). During data dependent acquisition, Orbitrap survey spectra were scheduled for execution at least every 3 s, with the embedded control system determining the number of MS/MS acquisitions executed during this period. A list of masses corresponding to all putative lysine modification sites of the Gli3^1-90 variants was used for targeted MS/MS acquisition to increase the sensitivity of detection. The parameters for MS/MS scans were HCD, 1 X 10⁵ AGC target, 128 ms maximal ion time, 0.4 m/z isolation window, 38 normalized collision energy, and 20 s dynamic exclusion. MS/MS raw files were converted into mzXML format and searched against Uniprot mouse database populated with user added protein sequences representing the mutated proteins by JUMP algorithm ⁸¹, a tag-based database search program.

Research Highlights.

• Disordered proteins contain abundant weak motifs with unclear functional relevance.

• We identify several weak SPOP binding motifs in the disordered N-terminus of Gli3.

• These motifs cooperate in recruiting oligomeric SPOP through avidity effects.

• A motif with a millimolar dissociation constant stimulates polyubiquitination.

• Weak motifs may facilitate post-translational modification through positioning of enzymes.

Abbreviations

AUC

analytical ultracentrifugation

BACK

BTB and C-terminal Kelch

β-ME

β-mercaptoethanol

BLI

Biolayer interferometry

BTB, and Bric à Brac

Tramtrack and Broad-Complex

CRL3

cullin3-RING ligase

CRL3^SPOP

CRL3 with substrate adaptor SPOP

F-pPuc

fluorescently labeled Puc peptide

FA

fluorescence anisotropy

Gli3^1-90

90-residue N-terminal Gli3 fragment

hetNOE

¹H-¹⁵N heteronuclear NOE

HSQC

heteronuclear single quantum coherence

K_D

dissociation constant

LC-MS/MS

liquid chromatography - tandem mass spectrometry

MATH

meprin and TRAF homology

NMR

nuclear magnetic resonance

pPuc

peptide derived from Puckered SB motif

SB

SPOP-binding

SPOP

Speckle-type POZ protein

SPOP^MATH

MATH domain of SPOP

SPOP^MATH-BTB

SPOP construct encompassing residues 28-337

SPOP^{MATH-BTB-BACK}

SPOP construct encompassing residues 28-359

SV

sedimentation velocity

TEV

tobacco etch virus

Ub noK

lysine-less ubiquitin mutant

wt

wildtype