NMR resonance assignment and structure prediction of the C-terminal domain of the microtubule end-binding protein 3

End-binding proteins (EBs) associate with the growing microtubule plus ends to regulate microtubule dynamics as well as the interaction with intracellular structures. EB3 contributes to pathological vascular leakage through interacting with the inositol 1,4,5-trisphosphate receptor 3 (IP3R3), a calcium channel located at the endoplasmic reticulum membrane. The C-terminal domain of EB3 (residues 200–281) is functionally important for this interaction because it contains the effector binding sites, a prerequisite for EB3 activity and specificity. Structural data for this domain is limited. Here, we report the backbone chemical shift assignments for the human EB3 C-terminal domain and computationally explore its EB3 conformations. Backbone assignments, along with computational models, will allow future investigation of EB3 structural dynamics, interactions with effectors, and will facilitate the development of novel EB3 inhibitors.

In mammals, the EB family consists of three paralogues, EB1, EB2 and EB3, which share a high degree of sequence homology [21]. They are comprised of 260-300 residues organized into the N-and C-terminal domains connected with a variable linker. The N-terminal region presented by the calponin-homology domain binds the MT tip [22], whereas the C-terminal region is required for dimerization [23][24][25]. Dimerization of EBs is a prerequisite for binding to growing MTs as well as interaction with other +TIPs [26][27][28]. Additionally, the C-terminal region contains the SxIP and LxxPTPh motifs, which are necessary for specific binding of EB partners [24,[29][30][31], and the EE(Y/F) sequence that is recognized by other cytoskeleton-associated proteins [32-34], including cytoplasmic linker proteins [35], and kinesin [36]. Hence, the C-terminus likely plays a pivotal role in multiple diverse cellular processes.
Despite significant sequence conservation between EBs, they have distinct functions in cells [21, 37,38]. EBs differ in their expression patterns throughout mammalian tissues and have unique binding partners [7,21]. EB3, for example, associates with the F-actin-binding protein drebrin and with the E3 ubiquitin ligase SIAH-1, while EB1 and EB2 do not interact with these proteins [39,40]. Additionally, EB3 but not EB1 interacts with IP 3 R3 in endothelial cells [38]. Remarkably, genetic ablation of EB3 in endothelial cells protects from pathological vascular leakage and pulmonary edema, suggesting that targeting its function with pharmacological agents might provide a novel strategy for treating inflammatory lung diseases [38]. However, there is little information on EB3 structure to guide drug discovery efforts. Here, we present NMR assignments and in silico protein structure prediction of the human EB3 C-terminus (residues 200-281). Our results will provide a structural basis for design of novel EB3 inhibitors.

Protein expression and purification
Preparation of EB3-C-terminus (200-281) with an N-terminal 6X His-tag was performed as described previously [38]. Briefly, the DNA sequence encoding the last 81 amino acids of the EB3 C-terminus was cloned into a pET42a vector and transformed into the BL21 (DE3) strain of E. coli (Invitrogen). Bacteria were grown at 37˚C in M9 media containing 15 N and 13 C stable isotopes and 50 μg/ml kanamycin. Protein expression was induced at an OD 600 of 0.6-0.7, by 250 μM isopropyl 1-thio-β-D-galactopyranoside, after which the cells were cultured at 30˚C for 4 hr. Bacteria were harvested by low-speed centrifugation, and the pellets lysed by sonication in the buffer containing 150 mM NaCl, 5 mM 2-mercaptoethanol, 2 mM CaCl 2 , 10 mM imidazole, 2 mM phenylmethylsulfonyl fluoride (PMSF), 25 mM Tris HCl, pH 7.4. 6X. His-EB3-C-terminal domain was purified using Ni-NTA beads (Thermo Scientific) equilibrated with 50 column-volumes of binding buffer (25 mM Tris HCl, pH 7.4, 300 mM NaCl, 5 mM 2-mercaptoethanol, 2 mM PMSF). Bacterial lysate (50 ml) was added to the column and the beads were washed with 150 column-volumes of wash buffer (PBS supplemented with 2 mM CaCl 2 and the protease inhibitor cocktail (Sigma). After washing, 6X His-EB3-C-terminus was eluted with 150 mM imidazole. Imidazole was removed using a PD-10 desalting column (GE Life Sciences), and concentrated in an Amicon Ultra-15 with 10 kDa cut-off concentrator unit (Millipore, Inc.). The 6X His-tag was cleaved by 1.5% (w/w) recombinant TEV protease at 4˚C for 16 hr. Cleaved EB3-C-terminus was then subjected to gel filtration chromatography over tandem Superdex 200 HR 10/30 columns connected in series and controlled by an AKTA FPLC (GE Life Sciences).

Results and discussion
Backbone assignments for the human EB3 C-terminal domain (200-281) were obtained using 350 μM uniformly 13 C and 15 N-labeled protein and triple resonance NMR experiments [44]. These data were subsequently deposited in the Biological Magnetic Resonance Databank (http://www.bmrb.wisc.edu/) [45] under the BMRB accession code 50003.
The 1 H, 15 N-HSQC spectrum of the EB3 C-terminus showed dispersed peaks indicative of a well-folded protein (Fig 1). The signal intensities were not uniform, suggestive of self-association or conformational dynamics in parts of the protein. We assigned 90% of 15 N and 1 H N resonances, as well as 89% of 13 Cα, and 54% of 13 Cβ signals. Assignment of all backbone resonances was precluded by inefficient transfers in three-dimensional experiments that were likely affected by undesirable relaxation processes. The glycine resonances in the C-terminal region were assigned based on 15 N-edited NOESY, as no signals for these residues were observed in the three-dimensional resonance assignment experiments.
Due to severe loss of signal in our NOESY experiments, we did not observe sufficient numbers of NOEs for NOE-based protein structure determination. Thus, the three-dimensional structure of the C-terminal domain of EB3 was modeled based on the highly homologous structure of the C-terminal domain of EB1 and the TALOS+ secondary structure results, using the iterative threading assembly refined algorithm on of I-TASSER web server (https:// zhanglab.ccmb.med.umich.edu/I-TASSER/) [47][48][49]. Consistent with the TALOS results and based on EB1 structure (PDB ID: 3GJO), five models generated here described the C-terminal domain of EB3 as an arrangement of three helices (Fig 3). Helices 1 (residues 202-237) and 2 (residues 246-256) had a fixed relative orientations, whereas helix 3 (residues 267-274 in models 1, residues 268-274 in model 2, residues 264-271 in model 3, residues 265-270 in model 4, and residues 265-280 in model 5) possessed a variable position and length (Fig 3). Further validation by comparing experimental and predicted 15 N chemical shifts of the five models was made using SHIFTX 2.0 (http://www.shiftx2.ca/) [50]. Using this comparison, we found that model 2 was the most consistent with experimental results presented here (Fig 4). Similar calculations were made for the EB1 crystal structure (PDB ID: 3GJO). The latter showed agreement between the experimentally-derived and predicted 15 N chemical shifts with R 2 correlation coefficients of 0.67 and 0.84 for BMRB depositions 34191 and 18371, respectively (Fig 4).

PLOS ONE
Structural characterization of end binding protein 3 (Fig 5), secondary structure restraints alone (S1 Fig), or without either EB1 homology or secondary structure information (S2 Fig). The best models based on the structure of EB1 with and without NMR-derived secondary structure restraints had comparable correlation coefficients of 0.68 and 0.69 for the predicted versus experimental 15 N chemical shifts, respectively (Figs 4 & 5), while removing EB1 homology restraints reduced these correlations (S1 and S2 Figs). This suggests that the structure of EB1 is essential for modelling plausible topology of the C-terminal domain of EB3.  Furthermore, analysis of signal intensities in the 1 H, 15 N HSQC spectrum of the C-terminal domain of EB3 indicated that enhanced relaxation processes might occur in the α-helix 3 region of the protein (Fig 6A), suggesting that this region likely samples multiple conformations. For instance, the signal intensities for H273, Q274, and Q275 were low, suggesting increased rigidity in this region of helix 3. Additionally, we have observed concentration dependent changes in the overlaid 1 H, 15 N HSQC spectrum of EB3 at 0.30 mM and 1mM (Fig  6B). These changes involve residues Q201, N206, V216, D224, Y226, K229, R231, E239, S242, E243, N244, V247, I248; G261, A263, I270, H273, and Q275. Residues 201-256 are the part of helix 1 and 2 as well as the flexible loop between these helixes in both selected Models (Figs 4  and 5). These three regions correspond to the dimeric interface in the C-terminal domain of EB1 (PDB ID: 3GJO). Hence, it is likely that the concentration dependent spectral changes can potentially reflect the chain exchange between EB3 dimers as observed with dimerization of the C-terminal domain of EB1 [25, 26, 31] and EB3 [23,37].
In summary, we provide assignments for the backbone resonances of the C-terminal domain of EB3. Chemical shift index analysis and molecular modeling suggest that the C-terminal domain of EB3 is highly helical and structurally similar to the C-terminal domain of EB1. The most distal C-terminal portion of EB3 significantly differs from the corresponding portion of EB1 in its amino acid sequence and forms a short helix that likely samples multiple positions relative to α-helices 1 and 2. These models of the C-terminal domain of EB3 can be useful for drug discovery effort.