Structure of the yeast Nup84-Nup133 complex details flexibility and reveals universal conservation of the membrane anchoring ALPS motif

The hallmark of the eukaryotic cell is the complex endomembrane system that compartmentalizes cellular functions. Transport into and out of the nucleus, occurs through the Nuclear Pore Complex (NPC). The heptameric Nup84 or Y complex is an essential scaffolding component of the NPC. Here we report two nanobody-bound structures: the full-length Nup84-Nup133 C-terminal domain complex and the Nup133 N-terminal domain, both from S. cerevisiae. Together with previously published structures, this work enables the structural description of the entire 575 kDa Y complex, from one species. The structure of Nup84-Nup1 33CTD details the high flexibility of this dimeric unit of the Y complex. Further, the Nup133NTD contains a structurally conserved amphipathic lipid packing sensor (ALPS) motif, confirmed by liposome interaction studies. The new structures reveal important details about the function of the Y complex that affect our understanding of NPC structure and assembly.


Structure of the Nup84-Nup133CTD-VHH-SAN8 complex
We solved the structure of full-length yeast Nup84-Nup133 C-terminal domain 96 (Nup133CTD, residues 521-1157) bound by a nanobody (VHH-SAN8) by single-wavelength 97 anomalous dispersion (SAD) to 6.4 Å resolution using an Anderson-Evans polyoxotungstate 98 derivative ( Table 1, Fig. 1b). We observed clear elongated, helical density in the initial, solvent-99 flattened experimental map. After multiple rounds of refinement and successively placing 100 helices into the density, we arrived at a map with clear enough density to employ both real-101 space docking and molecular replacement to place fragments of the structure that were 102 previously known 17,22,23 ( Supplementary Fig. 1). Following placement of roughly half of Nup84 103 and the C-terminal heel domain of Nup133, we were able to manually build the remaining 104 helices. Due to the limited resolution of the data, we did not build side chains for the model. 105 However, we are confident in assigning the sequence for both, Nup84 and Nup133 (see 106 Material and Methods for details). 107 Placement of the nanobody VHH-SAN8 was unambiguous as we observed large 108 difference density near the Nup84 crown element ( Supplementary Fig. 2). However, the electron 109 5 density around the three complementarity determining regions (CDRs) which presumably make 110 up the majority of the interface with Nup84 are not clear enough for chain tracing. Therefore, we 111 only rigid body positioned a nanobody model without CDRs in our structure. Additionally, we 112 solved the structure of Nup84-Nup133CTD bound to both VHH-SAN8 and a second nanobody, 113 VHH-SAN9, at 7.4 Å resolution (Table 1 and Supplementary Fig. 3). The doubly bound  Nup133CTD-VHH8/9 structure crystallized in the same crystal form as the singly bound Nup84-

119
Nup84 forms a continuous α-helical stack domain which, expectedly, belongs to the 120 ACE1 architecture. Its 28 α-helices form a rectangular block with dimensions of ~124 x 37 x 46 121 Å, with α1-7 zig-zagging in one direction and α8-28 folding over and zig-zagging the opposite 122 direction to the C-terminal end of the molecule (Fig. 1). The C-terminal helices interact with 123 Nup133CTD, which itself forms an elongated, highly arched stack of 26 α-helices (Fig. 1). The  Full length Nup84 has an ACE1 fold 136 With the full-length structure of Nup84, we can unambiguously detect the complete 137 topology of the ACE1 fold. We assign the crown to consists of α-helices 4-10, the trunk to α-138 helices 1-3 paired with α-helices 11-18, and the tail to α-helices 19-28 (Fig. 1 Interestingly, VHH-SAN8 binds at the interface between the crown and trunk elements, 157 thereby potentially rigidifying this portion of the protein (Fig. 1b). We hypothesize that this 158 interaction may stabilize a significant crystal packing interaction between Nup84 and a 159 symmetry related copy of Nup133 ( Supplementary Fig. 5). In support of this hypothesis, 160 previous attempts at crystallizing this complex without the nanobody yielded poorer diffracting 161 7 crystals, albeit in the same space group and with identical unit cell dimensions. With the solved 162 structure, we could now phase these nanobody-free Nup84-Nup133CTD crystals. We observe 163 exceptionally poor density for the Nup84 crown, supporting our hypothesis of the nanobody 164 stabilizing a conformationally dynamic area within Nup84 (data not shown).

165
For future in vivo studies, we tested by size-exclusion chromatography (SEC) whether 166 the nanobodies VHH-SAN8/9 interfere with Nup145C binding, Nup84's anchor to the Y-  (Fig. 3a). The hsNup133 'heel' lacks equivalent helices to α19, α24 and α25, making it 179 smaller than the scNup133 'heel', and having a different helical topology 23 . When we superpose 180 scNup133 and hsNup133 on the α-helices that interact with Nup84 (α15-17), we can observe a 181 large difference between the two 'heel' domains ( Fig. 3a). However, the helical topology at the 182 scNup84/hsNup107 binding interface overlays reasonably well, consistent with the higher  The 'blade' module of Nup133 is quite similar between yeast and human, only differing in 186 two additional α-helices at the N terminus of the human homolog. However, the heel position 187 8 relative to the arch module varies across the two structures, with the human blade module 188 swung 'out' ~90° relative to the yeast blade. By superposing the two blade modules, this motion 189 becomes more apparent (Fig. 3b). This creates a swing of ~41 Å at the C terminus of the   194 We solved the structure of the yeast Nup133 N-terminal domain (Nup133NTD, residues 195 55-481) at 2.1 Å resolution by molecular replacement, using the Vanderwaltozyma polyspora 196 Nup133NTD as a template 24 (Table 1, Fig. 4). Nup133NTD is a β-propeller consisting of seven β-197 sheets (blades) that are radially arranged around a solvent-accessible core (Fig. 4). Blades β1-6 198 each contain 4 β-strands that start near the core and trace outwards. Namely, strand 'A' is the 199 innermost strand, where strand 'D' is the outermost. Blade β7 is the exception, forming a 5-  (Fig. 4). Additionally, this structure has an elongated loop between 209 stands 'A' and 'B' in blade β1 that is ordered due to crystal packing contacts. The core structure 210 of Nup133NTD is conserved across the three species. The nanobody, VHH-SAN4, binds along strand 5D, via its large CDR3 loop (Fig. 4). The 212 opposite end of the nanobody facilitated packing with three symmetry related copies of 213 Nup133NTD, critical for lateral assembly within the crystal. Additionally, we solved the structure of 214 Nup133NTD bound by a second nanobody, VHH-SAN5, to 2.8Å resolution (Table 1 and 215 Supplementary Fig. 7). VHH-SAN5 binds an adjacent epitope to VHH-SAN4, also binding along 216 strand 5D using primarily its CDR3. There are additional contacts between CDR2 and the α-   231 After observing the same disordered DA34 loop in our structure that is present in the 232 human Nup133NTD, we wondered if the yeast Nup133 also has a functional ArfGAP1 lipid 233 packing sensing (ALPS) motif. A computationally determined ALPS motif has been identified in 234 scNup133NTD via homology modeling 24 . However, the hydrophobic moment of this loop is 235 weaker than the human DA34 loop due in part to the presence of an asparagine and lysine within 236 the hydrophobic half of the amphipathic helix (Fig. 5a). The increase in hydrophilicity in the 237 yeast Nup133NTD DA34 loop led some to conclude that this feature may not be conserved 238 between metazoa and yeast 36 . To directly test whether yeast Nup133NTD has a functional ALPS 239 motif, we performed liposome floatation assays. This assay determines whether a protein 240 interacts with liposomes through the floatation of a liposome-protein mixture to the top of an iso-241 osmotic gradient during ultra-centrifugation 46 . We observed that WT-Nup133NTD 'floats' with 242 liposomes comprising yeast polar lipids, but pellets without liposomes (Fig. 5b). 243 We also wondered if we could visualize any remodeling of liposomes by negative stain isosteric linker (GGGGSGGGGS) (Nup133NTDΔALPS), these protrusions did not occur (Fig. 5c).

250
The results of these disparate assays demonstrate that the yeast Nup133NTD has an ALPS motif 251 in its DA34 loop, a conserved motif with human Nup133, that can bind to and modify curved 252 biological membranes.   (Fig. 7). The structure of Nup84-Nup133CTD reported here is flatter and straighter than 283 previous models suggested 25,32 (Fig. 6), resulting in a better fit for our new composite model.

284
The Nup133NTD is only approximately positioned relative to the Nup133CTD, as the connection 285 between the two is highly flexible. However, Nup133NTD is positioned with its ALPS motif 286 adjacent to the NE. The fit for the nucleoplasmic ring was slightly better, suggesting some 287 conformational differences between the Y complexes on the cytoplasmic and nucleoplasmic 288 faces of the NPC (Fig. 7b, see Materials and Methods). The difference between the cytoplasmic 289 and nucleoplasmic Y complex conformation is more striking in docking attempts using our 290 previous composite model (Fig. 7). While the fitting correlation score and positioning for the 291 cytoplasmic ring was comparable between both models, the previous model fit much more we can delineate the flexible elements within the structure. We can also now see the 314 conformational differences in the Y complex on different faces of the NPC (Fig. 7). These 315 13 differences were also noted in a cryo-ET study on Xenopus laevis NPCs, suggesting this 316 difference as a conserved feature of the assembly 3 . Future studies can employ techniques such 317 as molecular dynamics, moving the molecule at known hinge points to fit into the map, rather 318 than individual nups that look highly similar at >10 Å resolution, to ultimately get a high-319 resolution assembly model. Overall, the structures presented here provide additional elements 320 for better constructing composite NPC structures combining X-ray crystallographic and electron 321 microscopy methods.

322
The observation of the great flexibility of the Y complex has led many to speculate its 323 importance in NPC biology. One key contributor to this flexibility is the ACE1 fold in multiple Y can anchor into the nuclear envelope with its ALPS motif, it is not unlikely that the Y complex 345 ring of a newly assembling NPC would need to change shape as the inner and outer nuclear 346 membranes transition through the extreme curvatures required for NE fusion during the building 347 of a new NPC 37 . This would be especially critical in assembly into a pre-formed NE during 348 interphase or a closed-mitotic system, such as in S. cerevisiae. 349 The conformation of the yeast Nup133 ALPS motif also raises many questions as to its    re-suspended in 100ml 2YT, 0.1% glucose, 50µg/ml kanamycin and 50 µg/ml ampicillin.

415
Cultures were incubated overnight at 30°C, then centrifuged for 20min at 7,700 × g, followed by 416 phage precipitation from the resulting supernatant with 1% PEG-6000, 500mM NaCl at 4°C, and 417 re-suspended in PBS.  Following two rounds of phage panning, 96 colonies were isolated in 96-well round-440 bottom plates and grown to mid-log phase at 37°C in 200 µl 2YT, 10 µg/ml ampicillin, 5 µg/ml 441 tetracycline, induced with 3 mM IPTG and grown overnight at 30°C. Plates were centrifuged at 442 12,000g for 10 min, and 100 µl of supernatant was mixed with an equal volume of 5% (w/v) 443 nonfat dry milk in PBS. This mixture was added to an ELISA plate coated with 1 µg/ml Nup84-444 133 or Nup133NTD. Following four washes with 1% Tween-20 in PBS, anti-llama-HRP antibody 445 (Bethyl) was added at a 1:10,000 dilution in 5% (w/v) nonfat dry milk in PBS for 1 hour at 25°C.

446
The plate was developed with fast kinetic TMB (Sigma) and quenched with 1M HCl. Absorbance