Multifaceted N-degron recognition and ubiquitylation by GID/CTLH E3 ligases

N-degron E3 ubiquitin ligases recognize specific residues at the N-termini of substrates. Although molecular details of N-degron recognition are known for several E3 ligases, the range of N-terminal motifs that can bind a given E3 substrate binding domain remains unclear. Here, studying the Gid4 and Gid10 substrate receptor subunits of yeast “GID”/human “CTLH” multiprotein E3 ligases, whose known substrates bear N-terminal prolines, we discovered capacity for high-affinity binding to diverse N-terminal sequences determined in part by context. Screening of phage displaying peptide libraries with exposed N-termini identified novel consensus motifs with non-Pro N-terminal residues distinctly binding Gid4 or Gid10 with high affinity. Structural data reveal that flexible loops in Gid4 and Gid10 conform to complementary folds of diverse interacting peptide sequences. Together with analysis of endogenous substrate degrons, the data show that degron identity, substrate domains harboring targeted lysines, and varying E3 ligase higher-order assemblies combinatorially determine efficiency of ubiquitylation and degradation.

However, the mechanistic roles of CTLH-mediated ubiquitylation in these pathways remain largely mysterious.
Recent genetic, biochemical and structural studies have revealed that the GID E3 is not a singular complex. Rather a core GID Ant complex (comprising Gid1, Gid5, Gid8, Gid2, Gid9 subunits) essentially anticipates shifts in environmental conditions that stimulate expression of interchangeable and mutually exclusive substrate-binding receptors -Gid4 (termed "yGid4" for yeast Gid4 hereafter) [17,33,34], Gid10 (yGid10 hereafter) [34][35][36] and Gid11 (yGid11 hereafter) [37]. Whereas yGid4 is expressed after glucose has been restored to carbon-starved yeast, yGid10 and yGid11 are upregulated upon other environmental perturbations including heat shock, osmotic stress as well as carbon, nitrogen and amino acid starvation. The resultant E3 complexes, GID SR4 , GID SR10 , and GID SR11 (where SR# refers to Gid substrate receptor), recognize distinct N-terminal sequences of their substrates [7,8,34,35,37]. In addition, another subunit, Gid7, can drive supramolecular assembly of two GID SR4 units into a complex named Chelator-GID SR4 to reflect its resemblance to an organometallic chelate capturing a smaller ligand through multiple contacts [38]. The cryo EM structure of a Chelator-GID SR4 complex with Fbp1 showed two opposing Gid4 molecules avidly binding N-degrons from different Fbp1 protomers. As such, Fbp1 is encapsulated within the center of the oval-shaped Chelator-GID SR4 . This assembly positions functionally-relevant target lysines from multiple Fbp1 protomers adjacent to two Chelator-GID SR4 catalytic centers.
The molecular details of GID/CTLH recognition of Pro/N-degrons were initially revealed from crystal structures of human Gid4 (referred to as hGid4 hereafter) bound to peptides with Nterminal prolines [10]. Although Pro/N-degron substrates of the CTLH E3 remain unknown, hGid4 is suitably well-behaved for biophysical and structural characterization, whereas yGid4 has limited solubility on its own [10]. Previously, the sequence PGLWKS was identified as binding hGid4 with highest affinity amongst all sequences tested, with a KD in the low micromolar range [10]. The crystallized peptide-binding region of hGid4, which superimposes with the substrate-binding domains of yGid4 and yGid10 in GID SR4 and GID SR10 , adopts an 8-stranded b-barrel with a central tunnel that binds the N-terminus of a peptide, or of the intrinsically-disordered N-terminal degron sequence of a substrate [10,34,36,38,39]. Loops between b-strands at the edge of the barrel bind residues downstream of the peptide's N-terminus. Interestingly, although GID SR4 was originally thought to exclusively bind peptides with an N-terminal Pro, hGid4 can also bind peptides with non-Pro hydrophobic N-termini such as Ile or Leu, albeit with at best ≈8-fold lower affinity [39].
Furthermore, yGid11 is thought to use a distinct structure to recognize substrate Thr/Ndegrons [37]. Collectively, these findings suggested that the landscape of GID/CTLH E3 substrates can extend beyond Pro/N-degron motifs.
Here, phage display screening identified peptides with non-Pro N-termini that not only bind hGid4, yGid4 and yGid10, but do so with comparable or higher affinity than the previously identified Pro-initiating sequences including Pro/N-degrons of ubiquitylation substrates.
Structural data reveal that loops in GID/CTLH substrate-binding domains adopt conformations complementary to partner peptide sequences downstream of the N-terminus.
Thus, sequence context is a determinant of N-terminal recognition by GID/CTLH substratebinding domains. In the context of natural substrates recognized by yGid4, not only the degron but also the associated domain harboring targeted lysine contribute to ubiquitylation by the core GID SR4 and its superassembly.

hGid4 can bind peptides with a range of N-terminal sequences
We took advantage of the amenability of hGid4 to biophysical characterization to further characterize features of the PGLWKS sequence mediating interactions. To assess the importance of peptide length beyond the N-terminus, we examined chemical shift perturbations (CSPs) in 2D 1 H, 15 N-HSQC NMR spectra of [ 15 N]-labeled hGid4 mixed with the amino acid Pro, a Pro-Gly dipeptide, or the PGLWKS peptide (Fig. 1A). Although prior studies emphasized the importance of an N-terminal Pro [10,39], Pro alone only minimally influenced the spectrum. The Pro-Gly dipeptide elicited stronger CSPs, presumably due to the peptide bond directly interacting with hGid4, and suppressing repulsion by burying the negatively charged carboxylate of a single Pro in a hydrophobic environment (Fig. S1A).
The PGLWKS peptide showed the greatest CSPs and binding kinetics in the slow exchange regime at the NMR chemical shift time scale, indicating tight binding, and, therefore, importance of downstream residues.
Given the ability of a Pro-Gly dipeptide to bind hGid4, we examined importance of the Nterminal residue by testing commercially-available variants (Leu-Gly, Ala-Gly, and Gly-Gly along with Pro-Gly) for competing with a fluorescently-labeled PGLWKS peptide whose binding to hGid4 can be measured by fluorescence polarization (FP) (Fig. S1B). Although each of the dipeptides yielded sigmoidal curves, those with N-terminal Pro or Leu were superior (Fig. 1B). Pro-Gly showed a 15-fold lower IC50 than Leu-Gly, consistent with prior studies emphasizing the importance of an N-terminal Pro [39].
To examine roles of individual positions in the 6-residue PGLWKS sequence, we employed peptide spot arrays testing all natural amino acids in position 1, positions 2 and 3 together, position 4 or position 5 (Fig. S1C). Binding was detected after incubating the membranes with the substrate binding domain of hGid4, and immunoblotting with anti-hGid4 antibodies.
Overall, the data confirm the previous findings that out of the peptides tested PGLWKS is an optimal binder, and that N-terminal non-Pro hydrophobic residues are tolerated in the context of the downstream GLWKS sequence albeit with lower binding [10,39].
The peptide array data also highlighted the importance of context. Amongst the 400 possible combinations of residues 2 and 3, Gly is preferred at position 2 and Ile or Leu at position 3, mirroring the previously defined sequence preferences. The dynamic range of our assay suggested that downstream residues also contribute to specificity, by unveiling pronounced amino acid preference for bulky hydrophobics and some non-hydrophobic residues also at position 4. In agreement with the structural data [10], the 5 th position following the PGLW sequence tolerates many amino acids.
Despite this seemingly strong preference for an N-terminal Pro, we serendipitously visualized hGid4 recognizing a supposedly non-cognate sequence when we set out to visualize its structure in the absence of a peptide ligand by X-ray crystallography.
Unexpectedly, the electron density from data at 3 Å resolution showed the first visible Nterminal residue of one molecule of hGid4 inserted into the substrate binding tunnel of an adjacent hGid4 molecule in the crystal lattice ( Fig. 1C; Table S1). Perplexingly, this was not the first residue of the input hGid4 construct but Gly116 located 16 positions downstream. It appears that hGid4 underwent processing during crystallization, although it remains unknown how this neo-N-terminus was generated. Nonetheless, the potential for hGid4 to recognize a non-cognate N-terminal Gly was supported by re-examination of the published "apo" hGid4 crystal. In 6CCR.PDB, distinct crystal packing is also mediated by a peptide-like sequence (initiating with a Gly from the Tobacco Etch Virus (TEV) protease cleavage site, followed by hGid4 Gly116) inserting into the substrate binding tunnel of the neighboring molecule in the lattice (Fig. 1D). We speculate that these structurally-observed interactions were favored by the high concentration of protein during crystallization.
To test binding of our fortuitously identified hGid4-binding sequence in solution, we examined competition with the fluorescently-labeled PGLWKS peptide (Fig. 1C). Low solubility of the GVATSLLW peptide (hGid4 residues 116-122) precluded accurate measurement of IC50 using our competitive FP assay. Nonetheless, the data qualitatively indicated that the GVATSLLW peptide binds to hGid4 with lower affinity than PGLWKS, but more tightly than the Pro-Gly dipeptide.
Taken together with published work, the data confirmed hGid4's preference for binding to the previously-defined sequence PGLWKS, but they also highlighted capacity for hGid4 to recognize alternative N-termini. Moreover, given that specific combinations of residues downstream of the Pro-Gly substantially impact the interaction, we considered the possibility that hGid4 recognition of N-terminal sequences could be influenced by context.

Identification of superior hGid4-binding motifs not initiated by Pro
To discover alternative hGid4-binding sequences that do not initiate with Pro, we constructed a highly diverse N-terminal peptide phage-displayed library of 3.5×10 9 random octapeptides. The library was constructed after the signal peptide using 8 consecutive NNK degenerate codons encoding for all 20 natural amino acids and fused to the N-terminus phage coat protein. It is expected that Arg or Pro located next to the cleavage site (position +1) will be inexistent or strongly underrepresented because they are known to either inhibit the secretion of phages [40,41] or the signal peptidase cleavage [42,43], respectively.
The library was cycled through five rounds of selections following an established protocol [44] to enrich for phages displaying peptides that preferentially bound hGid4 (Fig 2A).
Phages from individual clones that bound to GST-hGid4 but not a control GST based on phage ELISA were subjected to DNA sequence analysis.
The screen yielded 41 unique sequences, none of which were overtly similar to the previously defined hGid4-binding consensus motif PGLWKS ( Fig. 2B; Table S2). A new consensus emerged with the following preferences: (1) hydrophobic residues at position 1, with Phe predominating; (2) Asp at position 2; (3) hydrophobic residues at positions 3 and 6, and to a lesser extent at position 5; and (4) small and polar residues at positions 4 and 7.
Unlike the PGLWKS sequence wherein the striking selectivity is predominantly for the first four residues, this new consensus extends through the seventh residue.
Although peptides with non-Pro hydrophobic N-termini were previously shown to bind hGid4, the tested sequences bound with one to two orders-of-magnitude lower affinity (KD for IGLWKS 16 µM, VGLWKS 36 µM) than to PGLWKS (KD=1.9 µM) (Fig. S2A) [39]. To determine how the newly identified sequences compare, we quantified interactions by isothermal titration calorimetry (ITC). Notably, the peptides of sequences FDVSWFMG and VDVNSLWA showed superior binding (KD=0.6 and 1.3 µM, respectively) to the best binder with an N-terminal Pro ( Fig. 2C and S2A). Moreover, the affinity for a sequence starting with a Trp (KD=7.1 µM for WDVSWV) was superior to the previously identified best binders initiating with a non-Pro hydrophobic residue. Thus, hGid4 is able to accommodate even the bulkiest hydrophobic sidechain at the N-terminus of an interacting peptide. Taken together, the data show hGid4 binds a wide range of peptide sequences, with affinity strongly influenced by residues downstream of the N-terminus.

hGid4 structural pliability enables recognition of various N-terminal sequences
To understand how hGid4 recognizes diverse sequences, we determined its crystal structure bound to the FDVSWFMG peptide (Fig. 3A, Table S1; all peptide residues except C-terminal Gly visible in density). Overlaying this structure with published coordinates for other hGid4 complexes revealed diverse N-termini protruding into a common central substrate-binding tunnel (Fig. S2B, Phe (our study), or Pro, Leu, Val, or newly recognized Gly [10,39]). The N-terminal amine groups are anchored through contacts with hGid4 Glu237 and Tyr258 at the tip of the substrate binding tunnel, and common hydrogen bonds to the peptide backbone.
The structures suggest that the varying peptide sequences are accommodated by complementary conformations of four hairpin loops (L1-L4) at the edge of the hGid4 substrate-binding tunnel (Fig. 3B). The L2, L3, and L4 loops are fully or partially invisible, and are presumably mobile, in the structure of apo-hGid4 assembled in a subcomplex with its interacting subunits from the CTLH E3 [38]. However, they are ordered and adopt different conformations when bound to the different peptides.
As compared to the structure with PGLWKS, the interactions with FDVSWFMG are relatively more dominated by hydrophobic rather than electrostatic contacts (Fig. 3C). The L2 and L3 loops are relatively further from the central axis of the hGid4 b-barrel to interact with more residues in the peptide sequence. The different position of the L2 loop is also required to accommodate the bulkier N-terminal Phe side-chain, compared to Pro or other residues (Fig.   3D). Meanwhile, repositioning of the L4 loop places hGid4 Gln282 to form a hydrogen bond with Asp at the peptide position 2. Moreover, upon binding to hGid4, FDVSWFMG itself adopts a structured conformation owing to multiple intrapeptide backbone hydrogen bonds as well as interaction of Asp2 sidechain with the sidechain and backbone amide of Ser4 ( Fig. 3E). Overall, the structures reveal pliability of the hGid4 substrate-binding tunnel enabling interactions with a range of N-terminal sequences, which themselves may also contribute interactions by conformational complementarity.

Yeast GID substrate receptors recognize natural degrons with suboptimal affinity
To extend our findings to the yeast GID system, we screened the phage peptide library for binders to the yGid4 and yGid10 substrate receptors. The selected consensus sequence binding yGid4 paralleled that for hGid4 ( Fig. 4A; Table S2), in agreement with their being true orthologs. Remarkably, despite high similarity to the Gid4s, and its only known endogenous substrate likewise initiating with a Pro [36], the selections with yGid10 identified 12 unique sequences, some with bulky hydrophobic residues and others with Gly prevalent at position 1, each followed by a distinct downstream pattern ( Fig. 4B; Table S3). By solving an X-ray structure of yGid10 bound to FWLPANLW peptide and superimposing it on its prior structure with N-terminus of its bona fide substrate Art2 [36], we confirmed that the novel sequence is accommodated by the previously characterized binding pocket of yGid10 (  Table S1). Moreover, conformations of the yGid10 loops varied in complexes with different peptides, suggesting like hGid4, yGid10 structural pliability allows recognition of various N-terminal sequences (Fig. S3B).
To test if the selected sequences can mediate binding of substrates for ubiquitylation, we connected a yGid4-and a yGid10-binding sequence to a lysine via a flexible linker designed based on prior structural modeling [38]. The peptides also had a C-terminal fluorescein for detection. Incubating the peptides with either GID SR4 or GID SR10 and ubiquitylation assay mixes revealed that each serves as a substrate only for its cognate E3 (Fig. 4D).
Finally, we sought to quantitatively compare binding of the new sequences to respective substrate receptors. Affinities of yGid10 for Phe and Gly-initiating sequences, measured by ITC, were, respectively, comparable to and 2-fold greater than for a peptide corresponding to the N-degron of a natural substrate Art2 [36] (Fig. 4E and S3C). Notably, the endogenous degron, and selected sequences, bind yGid10 10-to 20-fold more tightly than the Proinitiating sequence previously identified by a yeast two-hybrid screen [35]. Although yGid4 is not amenable to biophysical characterization, we could rank-order peptides by inhibition of ubiquitylation of a natural GID SR4 substrate Mdh2 (Fig. 4F). Comparing IC50 values for the different peptides led to two major conclusions: (1) the phage display-selected sequences are better competitors than N-terminal sequences of endogenous gluconeogenic substrates, and (2) natural substrate N-terminal sequences themselves exhibit varying suppressive effects, with degron of Mdh2 being the most potent, followed by those of Fbp1 and Icl1.

GID E3 supramolecular assembly differentially impacts catalytic efficiency toward different substrates
We were surprised by the differences in IC50 values for the naturally occurring degrons from the best-characterized GID E3 substrates, Fbp1 and Mdh2. We thus sought to compare ubiquitylation of the two substrates, which not only display different degrons but also distinct catalytic domains with unique constellations of lysines. Previous studies showed that ubiquitylation of both substrates depends on coordination of degron binding by yGid4 with placement of specific lysines in the ubiquitylation active site [34,38]. However, while GID SR4 is competent for Mdh2 degradation in vivo, a distinct E3 assembly -wherein the Gid7 subunit drives two GID SR4 complexes into an oval arrangement (Chelator-GID SR4 ) is specifically required for optimal ubiquitylation and degradation of Fbp1 [38]. Two Gid4 subunits in Chelator-GID SR4 simultaneously bind degrons from the oligomeric Fbp1, for simultaneous ubiquitylation of specific lysines on two Fbp1 protomers.
Much like for Fbp1, addition of Gid7 to GID SR4 was shown to affect Mdh2 ubiquitylation in vitro, albeit in a more nuanced way [38]. As a qualitative test for avid binding to two degrons from Mdh2 (whose dimeric state was confirmed by SEC-MALS (Fig. S4A) and homology modeling (Fig. S4B)) we performed competition assays with monovalent (GID SR4 alone or with addition of a truncated version of Gid7 that does not support supramolecular assembly) and bivalent (GID SR4 with Gid7 to form Chelator-GID SR4 ) versions of the E3, and lysineless monodentate (Mdh2 degron peptide) and bidentate (Mdh2 dimer) inhibitors (Fig. S4C). While the two inhibitors attenuated ubiquitylation of Mdh2 to a similar extent in reactions with the monovalent E3s, only the full-length Mdh2 complex substantially inhibited the bivalent Chelator-GID SR4 . This suggested that Chelator-GID SR4 is capable of avidly binding to Mdh2.
Thus, we quantified roles of the Fbp1 and Mdh2 degrons by measuring kinetic parameters upon titrating the two different GID E3 assemblies. In reactions with monovalent GID SR4 , the Km for Mdh2 was roughly 3-fold lower than for Fbp1, in accordance with differences in degron binding ( Fig. 5A and 5B). Although the higher-order Chelator-GID SR4 assembly improved the Km values for Fbp1 and for Mdh2, the extents differ such that the values are similar for both substrates. Formation of the higher-order Chelator-GID SR4 assembly also dramatically increased the reaction turnover number (kcat) for Fbp1, with a marginal increase for Mdh2 (8-vs. 1.4-times higher kcat, respectively), which was already relatively high in the reaction with monomeric GID SR4 ( Fig. 5C and S4D). Combined with its effects on Km, formation of the Chelator-GID SR4 assembly increased catalytic efficiency (kcat/Km) more than 100-times for Fbp1 and only 6-fold for Mdh2, which may rationalize Gid7-dependency of Fbp1 degradation.
Beyond avid substrate binding, the multipronged targeting of Fbp1 by Chelator-GID SR4 involves proper orientation of the substrate so that specific lysines in metabolic regulatory regions are simultaneously ubiquitylated [38]. To explain the lesser effect of Chelator-GID SR4 on catalytic efficiency toward Mdh2, we examined structural models. Briefly, after docking two substrate degrons into opposing Gid4 protomers, we rotated the tethered substrate to place the targeted lysines in the ubiquitylation active sites (Fig. S5A, S5B and S5C). As shown previously, docking either Fbp1 targeted lysine cluster (K32/K35 and K280/K281) places the other in the opposing active site ( Fig. 5D and S5B). For Mdh2, however, although the K360/K361 lysine clusters from both Mdh2 protomers could be modeled as simultaneously undergoing ubiquitylation, the two major targeted lysine clusters (K254/K256/K259 and K330) cannot be simultaneously situated in both Chelator-GID SR4 active sites (Fig. 5D, S5A and S5C). Thus, the distinct constellations of targeted lysines may also contribute to differences in efficiency of ubiquitylation.

Degron identity determines Km for ubiquitylation but differentially impacts glucoseinduced degradation of Mdh2 and Fbp1
To assess the roles of differential degron binding in the distinct contexts provided by the

Furthermore, as expected, the Km values for all substrates improved in reactions with
Chelator-GID SR4 . However, the relative impact seemed to scale with the way in which they are presented from the folded domain of a substrate rather than the degrons themselves (roughly 14-fold for Fbp1 and 11-fold for Fbp1 Mdh2 degron versus 4-fold for Mdh2 and 6-fold for Mdh2 Fbp1 degron ).
Effects in vivo were examined by monitoring glucose-induced degradation of the wild-type and mutant substrates. Degradation was examined using the promoter reference technique, which normalizes for translation of an exogenously expressed substrate (here, C-terminally 3xFLAG-tagged versions of Fbp1, Mdh2, Fbp1 Mdh2 degron and Mdh2 Fbp1 degron ) relative to a simultaneously expressed control [7,45]. As shown previously, Mdh2 was rapidly degraded in the wild-type yeast and the DGid7 strain ( Fig. 6C) [38]. However, turnover of the mutant version bearing the weaker Fbp1 degron was significantly slower in both genetic backgrounds. Thus, the Mdh2 degron is tailored to the Mdh2 substrate. In striking contrast, although the Mdh2 degron did subtly impact degradation of Fbp1, it was not sufficient to overcome dependency on Gid7 (Fig. 6D). Thus, substrate ubiquitylation, and turnover, depend not only on degron identity, but also on their associated targeted domains.

DISCUSSION
Overall, our study of degron recognition by the Gid4/Gid10-family of substrate receptors of GID/CTLH E3s leads to several conclusions. First, GID/CTLH E3 substrate receptors recognize a diverse range of N-terminal sequences, dictated not only by the N-terminal residue, but also the pattern of downstream amino acids ( Fig. 1 and S1). Second, such diverse N-terminal sequence recognition is achieved by the combination of (1) a deep substrate-binding tunnel culminating in a conserved Glu and Tyr placed to recognize the Nterminal amine (2)  . Notably, our randomized phage-display peptide library screen identified far tighter binders to yGid4 than known natural degrons, which themselves bind yGid4 with varying KDs. This approach identified yGid10-binding sequences on par with the only known natural degron, and with higher affinity than a sequence identified by yeast two-hybrid screening. Phage-display peptide library screening may thus prove to be a useful method for degron identification. Fourth, degron binding is only part of substrate recognition by GID E3s (Fig. 5 and 6). Rather, ubiquitylation and degradation depend on both the pairing of a degron with a substrate domain that presents lysines in a particular constellation, and configuration of the GID E3 in either a simplistic monovalent format or in a multivalent chelator assembly for optimal targeting. Some features of the high-affinity peptide binding by Gid4s and yGid10 parallel other enddegron E3s. In particular, several cullin-RING ligases, the founding family of multiprotein E3s with interchangeable substrate receptors [46], have recently been discovered to recognize either specific N-or C-terminal sequences as degrons [9,[47][48][49][50]. Structures showed half a dozen or more residues in C-degron sequences engaging deep clefts or tunnels in their cullin-RING ligase substrate receptors, thus conceptually paralleling the highaffinity interactions with Gid4s and yGid10 [11,[51][52][53][54]. Furthermore, much like Gid4s and Gid10 recognize diverse sequences, the substrate-binding site of a single cullin-RING ligase was recently shown to bind interchangeably to a C-degron or to a different substrate's internal sequence [53][54][55]. Another interesting parallel between Gid4/Gid10-type recognition and some Ubr-family E3s is potential to bind diverse N-terminal sequences. However, while our data show that a common binding cleft in the Gid4/Gid10 fold structurally accommodates diverse sequences with high-affinity, some Ubr-family E3s bind different N-terminal sequences using distinct N-degron-binding domains [56][57][58]. Moreover, many structurallycharacterized N-terminal peptide-bound E3s have shown shallower modes of recognition.
To-date, few GID E3 substrates have unambiguously been identified. Thus, our findings have implications for identification of new substrates. Most of the currently characterized substrates depend on co-translational generation of an N-terminal Pro. However, sequences initiating with bulky hydrophobic residues may be refractory to N-terminal processing enzymes such as Met aminopeptidase [67][68][69]. Nonetheless, post-translational processing could generate such N-termini. Several paradigms for post-translational generation of N-degrons have been established by studies of Ubr1 substrates. First, endoproteolytic cleavage -by caspases, calpains, separases, cathepsins and mitochondrial proteases [37,[70][71][72][73][74][75] -is responsible for the generation of myriad Arg/N-degron pathway substrates recognized by some Ubr-family E3s [2]. Notably, over 1800 human proteins have an FDI/V sequence within them, raising the possibility that the newly identified Gid4-and Gid10-binding motifs likewise could be exposed upon post-translational protein cleavage.
It is tempting to speculate that hydrophobic N-degrons in eukaryotes could involve Nterminal amino acid addition. Finally, yeast Ubr1 is modulated in an intricate manner: after HtrA-type protease cleavage, a portion of the protein Roq1 binds Ubr1 and alters its substrate specificity [86]. Notably, proteomic studies showed that the human CTLH complex itself associates with the HtrA-type protease HTRA2 [22,24,[87][88][89], known to be involved in mitochondrial quality control [90,91]. This raises the tantalizing possibility that the CTLH E3 might form a multienzyme targeting complex that integrates a regulatory cascade to generate its own substrates or regulatory partners.
Finally, our examination of degron-swapped actual GID E3 substrates Fbp1 and Mdh2 showed that N-terminal sequence is only part of the equation determining ubiquitylation and subsequent degradation. Mdh2 required its own degron and its ubiquitylation and degradation were impaired when substituted with the weaker degron from Fbp1, irrespective of capacity for GID SR4 to undergo Gid7-mediated superassembly. However, while either degron could support Fbp1 targeting, this requires Gid7-dependent formation of the Chelator-GID SR4 supramolecular chelate-like E3 configuration. Taken together, our data reveal that structural malleability of both the substrate receptor and the E3 supramolecular assembly endows GID E3 complexes -and presumably CTLH E3s as well -capacity to conform to diverse substrates, with varying degrons and associated targeted domains. Such structural malleability raises potential for regulation through modifications or interactions impacting the potential conformations of both the substrate binding domains and higherorder assemblies, and portends future studies will reveal how these features underlie biological functions of GID/CTLH E3s across eukaryotes. Moreover, our results highlight that turnover depends on structural complementarity between E3 and both the substrate degron and ubiquitylated domains, a principle of emerging importance for therapeutic development of targeted protein degradation.

ACKNOWLDEGMENTS
We thank Chen G. and Pavlenco A. for construction of N-terminal peptide phage-displayed     The 2FO-FC electron density map corresponding to the peptide is shown as grey mesh contoured at 2s.
D. Fluorescent scans of SDS-PAGE gels after in vitro ubiquitylation of fluorescent model peptides harboring either a yGid4 or yGid10-binding sequence by GID Ant (comprising 2 copies each of Gid1 and Gid8, and one copy each of Gid5, Gid2 and Gid9) mixed with either yGid4 (∆1-115) or yGid10 (∆1-56) (forming GID SR4 or GID SR10 , respectively). The model peptides contained a corresponding phage display-determined consensus at the N-terminus connected to C-terminal fluorescein (indicated by an asterisk) with a flexible linker. E. ITC binding assays as in Fig. 2C but quantifying binding of several peptides to yGid10 (∆1-56). F. Competitive in vitro ubiquitylation assays probing binding of two novel Phe-and Leuinitiating sequences to yGid4 (∆1-115) as compared to N-termini of natural GID substrates (Mdh2, Fbp1 and Icl1). Unlabeled peptides were titrated to compete off binding of fluorescent Mdh2 (labeled with C-terminal fluorescein) to GID SR4 , thus attenuating its ubiquitylation. Normalized inhibition (fraction of ubiquitylated Mdh2 at varying concentration of unlabeled peptides divided by that in the absence of an inhibitor) was plotted against peptide concentration. Fitting to [inhibitor] vs. response model yielded IC50 values and its standard error based on 2 independent measurements.   Table summarizing values of Km for ubiquitylation of WT and degron-swapped substrates by the two versions of GID. C. In vivo glucose-induced degradation of exogenously expressed and C-terminally 3xFlagtagged Mdh2 as well as its degron-swapped versions quantified with a promoterreference technique. Levels of the substrates (relative to the level of DHFR) at different timepoints after switch from gluconeogenic to glycolytic conditions were divided by their levels before the switch (timepoint 0). For each substrate, the experiment was performed in WT and DGid7 yeast strains. Error bars represent standard deviation (n=3), whereas points represent the mean. D. In vivo assay as in (C) but with WT and degron-swapped Fbp1.

Data availability
The accession codes for the PDB models will be available in RCSB. All the unprocessed image data will be deposited to Mendeley Data.

Plasmid preparation and mutagenesis
All the genes encoding yeast GID subunits including the substrate receptors yGid4 and yGid10, as well as Fbp1 and Mdh2 substrates were amplified from S. cerevisiae BY4741 genomic DNA. The gene encoding hGid4 was codon-optimized for bacterial expression system and synthesized by GeneArt (Thermo Fisher Scientific).
All the recombinant constructs used for protein expression were generated by Gibson assembly method [102] and verified by DNA sequencing. The GID subunits were combined using the biGBac method [103] into a single baculoviral expression vector. All the plasmids used in this study are listed in the Reagent table.

Bacterial protein expression and purification
All Untagged WT ubiquitin used for in vitro assays was purified via glacial acetic acid method [105], followed by gravity S column ion exchange chromatography and SEC.

Insect cell protein expression and purification
All yeast GID complexes used in this study were expressed in insect cells.

Phage-displayed N-terminal peptide library construction and selections
A diverse octapeptide N-terminal phage-displayed library was generated for the identification of peptides binding to hGid4 (D1-99), yGid4 (D1-115) and yGid10 (D1-56). An IPTG-inducible Ptac promoter was utilized to drive the expression of open-reading frames encoding the fusion proteins in the following form: the stII secretion signal sequence, followed by a random octapeptide peptide, a GGGSGGG linker and the M13 bacteriophage gene-8 major coat protein (P8). The libraries were constructed by using oligonucleotide-directed mutagenesis with the phagemid pRSTOP4 as the template, as described [106]. The mutagenic oligonucleotides used for library construction were synthesized using with NNK degenerate codons (where N = A/C/G/T & K = G/T) that encode all 20 genetically encoded amino acids. The diversity of the library was 3.5 × 10 9 unique peptides.
The N-terminal peptide library was cycled through five rounds of binding selections against immobilized GST-tagged hGid4, yGid4, and yGid10, as described [44]. Pre-incubation of the phage pools against immobilized GST was performed before each round of selections to deplete non-specific binding peptides. For rounds four and five, 48 individual clones were isolated and tested for binding to the corresponding targets by phage ELISA [107], and clones with a strong and specific positive ELISA signal were Sanger sequenced. A total of 41, 12, and 12 unique peptide sequences were identified binding to hGid4, yGid4, and yGid10, respectively, and their sequences were aligned to identify common specificity motifs.
Oligonucleotide used for the Kunkel reaction to construct the library: GCTACAAATGCCTATGCANNKNNKNNKNNKNNKNNKNNKNNKGGTGGAGGATCCGGAG GA

Fluorescence polarization (FP) assays
To determine conditions for a competitive FP assay, we first performed the experiment in a non-competitive format. To compare binding of several unlabeled ligands to hGid4, we performed the FP measurements in a competitive format. Based on the FP plot from hGid4 titration experiment, we identified hGid4 concentration, which resulted in ~60% saturation of the FP signal (6.8 µM hGid4). Next, 2-fold dilution series of unlabeled competitors was prepared in FP buffer mixed with hGid4. After 5 min incubation, the measurement was performed as described above. The data were plotted relative to the FP signal in the absence of an inhibitor as a function of log(ligand concentration) and analyzed with log(inhibitor) vs.
response model to determine IC50 values. To determine relative inhibitory strength of the ligands, the determined IC50 values were divided by that of PGLWKS.

Screening of PGLWKS sequence for hGid4 binding using peptide spot array
The

Isothermal titration calorimetry (ITC) binding assays
To quantify binding of peptides to hGid4 (D1-115) and yGid10 (D1-56), we employed ITC. All To verify whether FDITGFS and GWLPPNL can be recognized by, respectively, yGid4 and yGid10 during ubiquitylation reaction (Fig. 4D), we performed an in vitro activity assay with model peptides, consisting of the respective N-terminal sequences connected to a single acceptor lysine with a 23-residue linker and C-terminal fluorescein (the length of the linker optimized based on GID SR4 structure as in [38]). To start the reaction, 0.2 µM E1 Uba1, 1 µM was incubated with Gid7 for 5 minutes on ice before the start of the reaction.
In order to test whether the preferred ubiquitylation sites within Mdh2 determined previously for GID SR4 [34] are also major ubiquitylation targets of Chelator-GID SR4 , we performed an activity assay with WT and mutant Mdh2, in which putative target lysine clusters Kinetic parameters for ubiquitylation of WT and degron-swapped versions of Fbp1 and Mdh2 were determined as described previously [38]. . To obtain kcat values, Vmax was divided by the E3 concentration: A0< = 2 7BC [DE] .

Yeast strain construction and growth conditions
The yeast strains used in this study are specified in the Reagents sequencing and immunoblotting to confirm protein expression.

In vivo yeast substrate degradation assays
In order to test the effect of degron identity on glucose-induced degradation of GID substrates, we monitored turnover of WT and degron-exchanged versions of Mdh2 and Fbp1, using the promoter reference technique adapted from [45]. Initially, WT and DGid7  , and further refinements were carried out with phenix.refine. Details of X-ray diffraction data collection and refinement statistics are listed in Table S1.

Figure S5: Structural modeling of Fbp1 and Mdh2 ubiquitylation by Chelator-GID SR4
A. In vitro assay of Mdh2-6xHis testing effect of mutating several of its lysines (previously determined to be preferred targets of GID SR4 by mass-spectrometry) on Chelator-GID SR4dependent ubiquitylation. Mdh2-6xHis and its ubiquitylated versions were visualized by anti-6xHis immunoblotting. Reactions were performed with WT and lysine-less ubiquitin (Ub; K0 has all Lys mutated Arg) B. Ubiquitylation model of Fbp1 (PDB: 7NS3, 7NS4, 7NS5, 7NSB; EMD-12557) involving juxtaposition of its target lysines (red and violet sticks, indicated by black circles) with Ubc8~Ub active site (red stars) generated by: (1) docking of two Fbp1 degrons (black dashes) into substrate binding cavities of two opposing Gid4 molecules (red cartoon) and (2) rotation of the folded Fbp1 domain (brown cartoon) so that its target lysines could simultaneously reach both active sites (the recruited and activated Ubc8~Ub intermediate shown as cyan (Ubc8) and yellow (Ub) surface) modeled by aligning a previous RING-E2~Ub structure (PDB: 5H7S) with Gid2 RING (grey cartoon). C. Ubiquitylation model of Mdh2 (blue cartoon; obtained by homology modeling) generated as in (B) but requiring a more pronounced shift of substrate receptor-scaffolding modules (Gid1-Gid8-Gid5-Gid4; grey cartoon) towards the center of the oval chelator assembly to enable capture of two Mdh2 degrons (black dashes). Besides K360/361, Mdh2 cannot be oriented so that its target lysines simultaneously engage both Ubc8~Ub active sites (red stars).  Table S1. X-ray crystallography data collection and refinement statistics Values for the highest-resolution shell are given in parentheses.