Crystal structure of the Rab-binding domain of Rab11 family-interacting protein 2

The structure of the uncomplexed Rab-binding domain of Rab11 family-interacting protein 2 has been determined at 2.3 Å resolution. The structure reveals antiparallel α-helical dimers that associate in the crystal lattice through polar interactions. The dimers assemble into a four-helix bundle in the lattice that is brought together by a remarkable stack of arginine residues.


Introduction
Rab GTPases regulate membrane-trafficking pathways in eukaryotic cells via the recruitment of effector proteins to subcellular compartments (Hutagalung & Novick, 2011). Rab11 family-interacting protein 2 (FIP2) is a 512-residue effector that contains a Rab-binding domain (RBD) at its C-terminus. The RBD is shared by a family of effector proteins, which include Rab-coupling protein (RCP or FIP1), FIP2 and FIP3 (Hales et al., 2001). The N-terminus of this modular effector family is variable and consists of domains that include EF-hands, ERM domains, C2 domains and myosin V-binding domains. These effectors regulate membrane trafficking following their recruitment to subcellular compartments by Rab11, Rab14 and Rab25 GTPases. Interaction with Rabs is facilitated by the RBD, which is highly conserved in sequence and structure. The crystal structures of Rab11-FIP2, Rab11-FIP3 and Rab25-FIP2 complexes revealed that the RBD is a parallel -helical coiled coil. The dimers of FIP2 and FIP3 are stabilized by hydrophobic interactions, and the symmetric coiled coil binds to two Rab molecules on each side of the dimer.
The adaptor functions of FIP2 linking Rab11 membranes to the cytoskeleton are likely to involve dynamic conformational changes. Here, the crystal structure of an uncomplexed form of FIP2 was determined at 2.3 Å resolution by the ab initio phasing method in ARCIMBOLDO. In contrast to previous crystal structures of complexes, the structure of isolated FIP2 reveals the formation of antiparallel -helical dimers that are stabilized by polar interactions.

Macromolecule production
A 20 ml overnight culture was added to a conical flask containing 1 l sterile 2ÂYT broth medium along with a 1:1000 dilution of 30 mg ml À1 kanamycin. The culture was incubated at 37 C and 180 rev min À1 until an OD (A 600 ) reading between 0.6 and 0.8 was reached. At this point, protein expression was induced with 0.5 mM isopropyl -d-1-thiogalactopyranoside and the temperature of the shaking incubator was adjusted to the optimum temperature for expression of the construct. At 18 C, the culture was left to incubate overnight or for approximately 18 h. At 37 C, the culture was left to incubate for 3 h. To isolate the protein, 20 ml extraction buffer (300 mM NaCl, 10 mM Tris, 10 mM imidazole, 5 mM -mercaptoethanol) was used to resuspend the bacterial pellet from 1 l of culture. The cell pellet was homogenized and the solution was sonicated by a series of 2 min pulses (duty cycle 30%, output 5, Branson sonifier). Each sample was subjected to sonication three times, resting on ice between rounds of sonication. The lysate was spun in a floor centrifuge at 18 000 rev min À1 and 4 C for 30 min. The resulting supernatant was applied onto a nickel agarose gravity-flow chromatography column. The column was then washed thoroughly with extraction buffer (300 mM NaCl, 10 mM Tris, 10 mM imidazole, 5 mM -mercaptoethanol). The protein was eluted from the column with elution buffer (300 mM NaCl, 10 mM Tris, 200 mM imidazole, 5 mM -mercaptoethanol). The protein was cleaved with Tobacco etch virus (TEV) protease overnight in a cold room under dialysis with extraction buffer. After cleavage, the protein solution was reapplied onto a nickel agarose gravity-flow chromatography column. The flowthrough from the column, containing the cleaved protein, was collected. The protein was then dialyzed into low-salt buffer (5 mM NaCl, 10 mM Tris, 5 mM -mercaptoethanol) for 3 h. The protein was applied onto a Mono Q anionexchange column (GE Life Sciences) and a salt gradient was applied from low salt (5 mM NaCl, 10 mM Tris, 1 mM DTT) to high salt (1 M NaCl, 10 mM Tris, 1 mM DTT). The main resulting peak contained pure, cleaved FIP2, as demonstrated by SDS-PAGE. The purified FIP2 was then run on a Superdex 75 16/60 gel-filtration column (buffer: 150 mM NaCl, 10 mM Tris, 1 mM DTT). The resulting peak was taken and concentrated to >6 mg ml À1 for crystallization. The protein concentration was determined from the absorbance at 280 nm using an extinction coefficient of 5960 M À1 cm À1 . Macromoleculeproduction information is summarized in Table 1.

Crystallization
Purified FIP2 was concentrated to approximately 6 mg ml À1 prior to crystallization. The lead conditions for crystallization were obtained from a sparse-matrix screen (Structure Screen, Hampton Research) as a combination of cobalt chloride and 1,6-hexanediol at low pH. The optimal condition is shown in Table 2. The crystals took approximately two weeks to grow to maximum size. Crystallization information is summarized in Table 2.

Data collection and processing
Crystals were briefly soaked in the reservoir solution supplemented with 25% glycerol prior to data collection. Data were collected on beamline 24-ID-C at the Advanced Photon Source (APS), Argonne, Illinois, USA. Data were integrated using XDS (Kabsch, 2010) and were merged and scaled using AIMLESS (Evans, 2006). Data-collection and processing statistics are summarized in Table 3.

Structure solution and refinement
An initial model of the crystal structure was determined using ARCIMBOLDO (Rodríguez et al., 2009). Four partial helices were built in an automated fashion using the software, and the model was of sufficient quality for manual building into the electron-density map. In an iterative fashion, ARCIMBOLDO_LITE uses Phaser (molecular replacement) to search for short -helices, followed by SHELXE to connect them into longer polypeptides. Following the last round of  Table 1 Macromolecule-production information.   Fig. S1). The automated model building was remarkably successful: only five extra residues were built during further stages of manual refinement to give a total of 202 residues distributed over four -helices.
The additional residues and solutes (waters and hexanediol) were built by multiple rounds of model building and refine-ment through inspection of 2F o À F c maps using Coot (Emsley et al., 2010) and Phenix (Liebschner et al., 2019). MolProbity (Chen et al., 2010) was used for Ramachandran analysis. Refinement statistics are summarized in Table 4.

Results and discussion
Here, we describe an effector domain from a Rab11 familyinteracting protein ( Arginine-stacking interactions stabilize a tetrameric FIP2 assembly. (a) The FIP2 tetramer in the lattice that is related by a twofold crystallographic axis. The axis runs through the middle of a stack of arginine residues that enable oligomerization into a four-helix bundle. (b) A view of the eight arginine residues from the four -helices. The view is a close-up of the transparent ellipse in (a). (c) A section of the electron density (2F o À F c , 1.5) within the region comprising the arginine-stacking interactions.  Comparisons of FIP2 in Rab11-bound and free states. (a) Complex of Rab11 (residues 7-173) with residues 448-503 of FIP2 (PDB entry 4c4p). The effector domain is at the extreme C-terminus of the 512-residue protein.
Residues 129-290 of FIP2 (arrows), which lie upstream of the Rab11 effector domain, comprise the myosin Vb-binding region. Following 5 of Rab11, a hypervariable region of 43 residues is prenylated at two cysteine residues near the C-terminus of the 216-residue protein. This flexible region was dispensed with to enable crystallization of the complex (Lall et al., 2013). (b) Superposition of chain A of FIP2 with one of the -helices from the Rab11-FIP2 complex. The parallel coiled coil is generated by a twofold symmetry operation from a 1:1 Rab11-FIP2 complex in the asymmetric unit. Therefore, the two -helices in the complex are identical. The Rab11-FIP2 ribbons are displayed with transparency. (c) Superposition of the -helices of FIP2. Identical segments of FIP2 were aligned by a secondary-structure matching algorithm using Coot. The 39 core residues in the -helices aligned with a root-mean-square (r.m.s.) deviation of 1.8 A for their C atoms.
aligned with their long axes in similar orientations (Fig. 1a).
The nature of these crystals may pose a challenge for long -helices as MR search models. Since crystal formation required the presence of a small amount of Co 2+ , data were also collected at the Co 2+ absorption edge. Although there was a weak anomalous signal, it was insufficient for phasing of the structure. Wide-search molecular replacement from the full Protein Data Bank was performed using Phaser (McCoy et al., 2007), as implemented in the Structural Biology Grid portal (https://sbgrid.org/; Stokes-Rees & Sliz, 2010). Candidate -helical proteins were identified, but none were suitable as initial models for further refinement.
Phasing was successfully performed using ARCIMBOLDO_ LITE as implemented within the CCP4 suite (Rodríguez et al., 2009;Winn et al., 2011). The crystal lattice of FIP2 can be described as an -helical tetramer in which the long axes of the helices are coincident with the c axis of the crystal (Fig. 1a). The asymmetric unit consists of two pairs of antiparallel -helices that are assembled through polar interactions (Fig. 1b). The polypeptides are of varying lengths depending on the extent of disorder at the N/C-termini. Although the segment 439-512 was subjected to crystallization, a minimum of the first seven residues and the last 15 residues are disordered. The crystal structure of FIP2 is in contrast to the previously determined NMR solution structure of FIP2 ( Fig. 1c; Wei et al., 2009) and the crystal structure of FIP2 in complex with Rab11 ( Fig. 1d; Jagoe, Jackson et al., 2006). These structures are parallel -helical dimers that are tightly associated by hydrophobic interactions.
The packing interactions within the -helical tetramer are fascinating and are worth closer inspection (Fig. 2). Dimers from the asymmetric unit form a stack of arginine residues with their symmetry-related dimers in the lattice (Fig. 2a). Each helix contributes two arginines that form two layers of a four-arginine stack in the middle of the -helical bundle (Fig. 2b). The distances between the C atoms in the stacked guanidino side chains are 3.5-3.8 Å . Aspartate residues form salt bridges with these arginines and presumably contribute to orienting the guanidino groups (not shown). The electrondensity map in this region reveals well ordered side chains (Fig. 2c). Arg-Arg interactions have been recognized for their significant contributions to protein assemblies (Neves et al., 2012;Vernon et al., 2018). The distance between the stacks (<4 Å ) is similar to the observed van der Waals distances from a survey of structures (Vernon et al., 2018). However, the detailed energetics and stabilization of stacked arginines in protein assemblies are poorly characterized and require further study.
A physiological model for complex formation involves Rab11 recruitment of pre-formed parallel dimers of cytosolic FIP2 to endosomes Eathiraj et al., 2006). This model is premised on the finding that switch 1 and switch 2 of a single Rab11 molecule interact with both -helices of FIP2 in a symmetric fashion to form a heterotetrameric complex (Fig. 3a). Also, FIP2 spontaneously forms dimers in solution at physiological pH (Jagoe, Jackson et al., 2006;. It is probable that parallel dimers dissociate to form the crystals observed here at pH 4.8. Therefore, the physiological relevance of the structure of uncomplexed FIP2 observed under these conditions is unknown. Nevertheless, it is intriguing to observe the conformational flexibility of the -helices of FIP2 under various conditions. Although the central -helical regions are similar in all structures, the N-and C-termini diverge significantly (Figs. 3b and 3c). Whether these variations reflect possible dynamic changes of FIP2 during membrane trafficking requires further investigation. Interestingly, crystals of the Rab11-FIP2 complex have been grown at pH 4.5 , albeit under different precipitant conditions. It is conceivable that FIP2 can exist in multiple conformational states and that Rab11 binds selectively to the parallel dimer during the crystallization process.
In summary, the uncomplexed structure of FIP2 reveals head-to-tail oligomers of -helices that are stabilized by polar interactions. The diffraction data and crystal structure may contribute to a useful archive for further improvement of techniques in macromolecular phasing.