Molecular and structural mechanisms of ZZ domain-mediated cargo recognition by autophagy receptor Nbr1

In selective autophagy, cargo selectivity is determined by autophagy receptors. However, it remains scarcely understood how autophagy receptors recognize specific protein cargos. In the fission yeast Schizosaccharomyces pombe, a selective autophagy pathway termed Nbr1-mediated vacuolar targeting (NVT) employs Nbr1, an autophagy receptor conserved across eukaryotes including humans, to target cytosolic hydrolases into the vacuole. Here, we identify two new NVT cargos, the mannosidase Ams1 and the aminopeptidase Ape4, that bind competitively to the first ZZ domain of Nbr1 (Nbr1-ZZ1). High-resolution cryo-EM analyses reveal how a single ZZ domain recognizes two distinct protein cargos. Nbr1-ZZ1 not only recognizes the N-termini of cargos via a conserved acidic pocket, similar to other characterized ZZ domains, but also engages additional parts of cargos in a cargo-specific manner. Our findings unveil a single-domain bispecific mechanism of autophagy cargo recognition, elucidate its underlying structural basis, and expand the understanding of ZZ domain-mediated protein-protein interactions.

1 level ( Figure S1C), possibly because its vacuolar targeting efficiency is lower than other 2 NVT cargos. Given that ubiquitination promotes NVT transport (Liu et al., 2015), we 3 adopted the method of artificial ubiquitin tethering to enhance the transport efficiency of 4 individual cargos (Zhu et al., 2017). When three tandemly linked ubiquitin (3xUb) fused 5 with the GFP-binding protein (GBP) was expressed in the cell, GFP-tagged Ape2, a known 6 NVT cargo, localized more prominently to the vacuole, and Ub tethering-enhanced vacuolar 7 targeting of Ape2 still required Nbr1 ( Figure S1D and S1E). Applying this method to Ape4 8 resulted in an obvious vacuole lumen localization of Ape4 in wild-type but not in nbr1Δ 9 cells ( Figure 1E).
If Ams1 and Ape4 are cargos of the NVT pathway, the ESCRT machinery but not the 11 macroautophagy machinery should be required for their vacuolar targeting. Indeed, Ams1 no 12 longer localized in the vacuole in cells lacking sst4, which encodes a subunit of the ESCRT-13 0 complex ( Figure 1D). In contrast, atg1Δ mutant showed a normal localization of Ams1 14 ( Figure 1D). In addition, Ams1-mECitrine processing was blocked by sst4Δ but not atg5Δ 15 ( Figure S1F and S1G). Similar to the situation of Ams1, Ub tethering-enhanced Ape4 16 localization in the vacuole required Sst4 but not Atg1 ( Figure 1E). These results demonstrate 17 that Ams1 and Ape4 are NVT cargos. 1 ZZ1 molecules bind symmetrically to one Ams1 tetramer ( Figure 4A and 4B). Remarkably, 2 each Nbr1-ZZ1 molecule simultaneously engages two Ams1 subunits through a continuous 3 binding interface. Specifically, the Nbr1-ZZ1 molecule shown at the top left of Figure 4A 4 and Figure 4B binds both the N-terminus of the Ams1_3 subunit (sub-interface I) and a 5 large surface of the Ams1_1 subunit (sub-interface II) ( Figure 4D-4F). This special binding 6 configuration corresponds to the fact that the N-terminal tail of each Ams1 subunit extends 7 away from the rest of the molecule and inserts into a cleft formed between two other Ams1 8 subunits (Zhang et al., 2020). Thus, the Ams1-ZZ1 complex represents an interesting 9 example of a protein-binding module-in this case a ZZ domain-recognizing the 10 quaternary structure of its oligomeric binding partner.

11
At sub-interface I, the N-terminal dipeptide (T2-L3) of Ams1 fits into an acidic pocket of 12 Nbr1-ZZ1 located between the hairpin loop and the β-sheet ( Figure 4D). The density map 13 unambiguously shows that T2 is the most N-terminal residue of Ams1 ( Figure S5E), 14 consistent with the expectation that the first translated methionine residue is removed in vivo 15 when the second residue has a small side chain (Bonissone et al., 2013). The N-terminal 16 dipeptide of Ams1 aligns in an antiparallel manner with residues 60-62 of Nbr1 ( Figure 4G).

17
The free α-amino group of T2 hydrogen bonds with the side chain amide group of N63 and accessible surface areas of 760 Å 2 and 227 Å 2 per molecule, respectively. The binding at 1 sub-interface II is mediated by shape complementarity and a large number of van der Waals, 2 hydrophobic, hydrogen bonding, and electrostatic interactions ( Figure 4E and 4F). In 3 particular, the loop between P97 and P99 of Nbr1 fits snugly into a cavity on the surface of 4 Ams1 ( Figure 4E).

5
The entire structure of Ams1, including the N-terminal dipeptide, does not change 6 between the free and ZZ1-bound states ( Figure S5F), indicating that Nbr1-ZZ1 recognizes a 7 preformed conformation of Ams1.

9
The cryo-EM structure of the Ape4-ZZ1 complex We applied the same fusion strategy to determine the structure of Ape4 in complex 11 with Nbr1-ZZ1. The fusion of Ape4 and Nbr1(53-129) was expressed and purified from S. 12 pombe, and analyzed by cryo-EM ( Figure 5A and S6). A density map was reconstructed 13 from 352,316 particles at 2.26 Å resolution ( Figure S7).
14 Like other M18 family metallopeptidases with solved structures, including the Cvt 15 pathway cargo S. cerevisiae Ape1 (ScApe1) (Chen et al., 2012;Chaikuad et al., 2012; partial occupancy ( Figure S6E). Two Nbr1-ZZ1 molecules in a pair are in contact with each 23 other, raising the possibility that they bind Ape4 in a cooperative manner. However, 3D 24 variability analysis revealed no correlation between the densities of the two Nbr1-ZZ1 25 molecules, suggesting that they rather bind independently ( Figure S6E).
shaped tail and the α1 helix of Ape4, burying a solvent accessible surface area of 359 Å 2 per 1 molecule ( Figure 5C). The interface is rather small, accounting for the low occupancy of 2 Nbr1 in the complex. The interface can be divided into two sub-interfaces. Sub-interface I 3 largely corresponds to sub-interface I in the Ams1-ZZ1 complex, with the N-terminal 4 tripeptide M1-Q2-L3 of Ape4 fitting into the acidic pocket of Nbr1-ZZ1 ( Figure 5D). The 5 N-terminal methionine of Ape4 is retained ( Figure S7E), because N-terminal methionine 6 excision does not happen when the second residue is glutamine (Bonissone et al., 2013). The 7 fact that Nbr1-ZZ1 uses the same acidic pocket to bind the N-termini of Ams1 and Ape4 8 explains why these two cargos exhibit mutually exclusive binding to Nbr1-ZZ1. The sub-9 interface II is formed by the contact between the hairpin loop of Nbr1-ZZ1 and the α1 helix 10 of Ape4, mainly via van der Waals interactions ( Figure 5C).

11
In Ape4, the N-terminal U-shaped tail packs against the α1 helix by forming a 12 hydrophobic cluster composed of M1, L3, M7 and the aliphatic part of K12 ( Figure 5C). It 13 is likely that these intra-molecular contacts help stabilize the conformation of the N-terminal binding surface specific for one cargo may compromise the binding of that cargo but not the 1 other cargo. Indeed, Nbr1-P97, which is situated at the Ams1-specific binding surface, when 2 mutated to arginine, disrupted the Nbr1-Ams1 interaction ( Figure 6A), and abolished the 3 ability of Nbr1 to mediate vacuolar targeting of Ams1 ( Figure 6B), but had no effect on the 4 Nbr1-Ape4 interaction and the ability of Nbr1 to mediate vacuolar targeting of Ape4 ( Figure   5 6C and 6D). Conversely, when we introduced into Nbr1 the L66R mutation, which affects a 6 key residue at the Ape4-specific binding surface, the Nbr1-Ape4 interaction but not the 7 Nbr1-Ams1 interaction was disrupted ( Figure 6A and 6C), and vacuolar targeting of Ape4 8 but not Ams1 was abrogated ( Figure 6B and 6D). The Nbr1-A61R mutation caused effects 9 similar to those of the Nbr1-L66R mutation ( Figure 6). Based on the structures, this 10 mutation is expected to cause a steric clash with Ape4 but not with Ams1 ( Figure 5C).

11
These results demonstrate that cargo recognition by Nbr1-ZZ1 requires both its interactions 12 with the N-termini of cargos and its interactions with non-N-terminal parts of cargos. fusion approach, we also tested whether preserving M1 in Ams1 affects Nbr1 binding and 3 found that N-terminal methionine removal is important for the Ams1-Nbr1 interaction 4 ( Figure 7D). Together, these results showed that the sequences of the N-terminal residues in 5 both Ams1 and Ape4 are important for Nbr1-ZZ1 binding.

6
To verify the role of the non-N-terminal binding mode in the Ape4-ZZ1 interaction, we 7 introduced into Ape4 the A9R mutation, which was expected to cause a steric clash in the 8 sub-interface II between Ape4 and Nbr1-ZZ1 ( Figure 5C). Indeed, this mutation abolished 9 the Ape4-Nbr1 interaction ( Figure 7C). In addition, we found that the Ape4-M7R mutation and the Ape4-K12E mutation, which are expected to destabilize the conformation of the U-11 shaped tail of Ape4, also individually abolished the Ape4-Nbr1 interaction ( Figure 7C).

12
These data suggest that packing of the N-terminal tail to the structural body of Ape4 is 13 important for Nbr1 binding.   Table S2 and plasmids used in this 11 study are listed in Table S3. Genetic methods of strain construction and composition of 12 media are as described (Forsburg and Rhind, 2006). The deletion strains used in this study 13 were constructed by PCR amplifying the deletion cassettes in the Bioneer deletion strains 14 and transforming PCR products into our laboratory strains or by standard PCR-based gene 15 targeting. All point mutations were generated by PCR-based mutagenesis. Strains expressing 16 proteins with different tags under native promoters were generated by PCR-based tagging.
camera. Images were obtained with a 100×1.4NA objective and analyzed with the 1 SoftWoRx software.
Immunoblotting assay examining vacuolar processing of mECitrine-tagged proteins 3 About 10 OD600 units of cells were harvested, mixed with 100 μl of Lysis Buffer, and 4 lysed using the bead-beating lysis method. The lysate was mixed with 2×SDS loading buffer 5 and boiled for 10 min. Samples were separated on a 12% SDS-PAGE gel and 6 immunoblotted with anti-GFP antibody. 7 Pil1 co-tethering assay 8 To examine a pair-wise protein-protein interaction, a bait protein was fused to Pil1-9 mCherry, and a prey protein was fused to mECitrine or CFP. Cells co-expressing both 10 proteins were grown to log phase for fluorescence microscopy. To image the plasma-11 membrane associated filament-like structures formed by a Pil1-fused protein and its 12 interactor, we acquired 8-10 optical Z-sections so that either the top or bottom plasma 13 membrane is in focus in one of the Z-sections. Images were processed by deconvolution 14 using the SoftWoRx software.    1 processed with a mask of 300 Å diameter. Particles were downsized 4-fold for 2D and 3D 2 classification in RELION-3.0-beta (Zivanov et al., 2018). The free Ams1 structure (EMD-3 30021) low-pass filtered to 30 Å was used as the initial model for 3D classification and 4 refinement. After one round of 2D classification and two rounds of 3D classification, 5 227,295 particles from high resolution classes were subjected for 3D reconstruction with D2 6 symmetry imposed in RELION-3.0-beta ( Figure S4D). After the raw movie datasets were 7 further processed in RELION for motion correction, particle polishing and CTF refinement,  Table S1.

22
To build the Ams1-Nbr1 complex structure, the model of Ams1 was derived from the 23 cryo-EM structure of free Ams1 (PDB code 6LZ1) and the Nbr1-ZZ1 domain structure was 24 de novo built. The model was refined against the sharpened cryo-EM map of Ams1-Nbr1 25 complex.

26
To build the Ape4-Nbr1 complex structure, a starting model of Ape4 was generated by domain determined in complex of Ams1 was fitted as rigid body into the unsharpened and 1 unfiltered cryo-EM map of the Ape4-Nbr1 complex. The final model of Ape4-Nbr1 complex 2 was refined against the LocScale map.

3
The current model of Ams1-Nbr1 complex contains residues 2-1077 of Ams1, residues 4 1-3 of the linker and residues 59-108 of Nbr1. The current model of Ape4-Nbr1 complex 5 contains residues 1-473 of Ape4, residues 1-2 of the linker and residues 59-108 of Nbr1.  To calculate FSC between model and map, the model was transformed into a density map in 9 Chimera. The resultant model map was resampled to the grid of the map used for refinement.

10
For cross-validation, the model was re-refined against a map reconstruction from the other 11 half of the data. The resolution of density map was estimated using the criteria of gold-

12
Ape4 was endogenously tagged at its C-terminus with GFP and in the same cells GFP-

9
(E) In an AP-MS analysis using Ape2 as bait, Nbr1 and Lap2, but not Ams1 and Ape4.