Promiscuous interactions of human septins: The GTP binding domain of SEPT7 forms filaments within the crystal

Abstract

SEPT7 G and SEPT7 G bind by molecular sieving (View interaction). In animal cells, septins have been associated with a wide range of intracellular processes involving membrane re-modelling events and chromosome segregation [1-5]. Their importance for the completion of cytokinesis is particularly well documented and even more so in the case of budding yeast [6-8]. Furthermore, in both yeast and animal cells, specific septins have been associated with the formation of diffusion barriers suggesting a role in protein segregation and compartmentalization [9-12]. Finally, septins play roles in non-dividing cells, such as neurons, where mammalian SEPT7 has been associated with regulating the morphological development of dendritic spines [13]. Several recent reviews have highlighted the increasing importance being given to the physiological roles of septins as well as their involvement in several pathological states [14-17]. The generic septin structure consists of three domains; a highly variable N-terminal domain, a central GTP-binding domain and a C-terminal domain which normally includes sequences compatible with a coiled coil [18, 19]. Towards the end of the N-terminal domain a polybasic region is believed to be important for membrane association [20]. Both in vivo and in vitro septins associate to form hetero-polymers, where the number of individual septins involved varies from species to species [16]. In mammalian septins, most hetero-filaments are believed to be composed of three different septins [18] with several such combinations reported in the literature [19, 13, 21, 22]. The best characterized to date is the SEPT2/SEPT6/SEPT7 hetero-polymer [23] which is the only one for which a crystal structure is currently available [24]. This structure shows that it is the GTP-binding domain which provides the two inter-subunit interfaces (named G and NC) necessary for filament assembly. Fourteen different septins have been identified within the human genome suggesting that a wide range of different hetero-filaments can potentially form. However, our structural knowledge is limited to the single complex described above and to two crystal structures of the GTP-binding domains of SEPT2 [24, 25]. This dearth of structural information severely limits our current understanding of filament formation and the factors which control correct assembly. Clearly there is a need for more structural information. Here we report a molecular replacement solution for the GTP-binding domain of human SEPT7 which shows it to employ the same G and NC interfaces for filament formation within the crystal lattice. The 1258 bp GTPase domain of the human SEPT7 coding sequence was amplified from a human brain cDNA library (GIBCO-BRL) by the polymerase chain reaction (PCR) based on the available human SEPT7_v1 sequence (NCBI Accession code NM_001788). Oligodeoxynucleotide primers for PCR amplification introduce both Nde I and Xho I restriction sites at the 5′ and 3′ end, respectively for cloning into the pET28a expression vector (Novagen) [26] modified to include a TEV cleavage site. The recombinant plasmid was sequenced on an ABI377 DNA sequencer to confirm its identity. Cells from a single colony were grown overnight at 37° C in 5 1L LB medium containing 30 μg/ml kanamycin to an O.D.600 of 0.7. The culture was then induced for 18 h with 0.4 mM IPTG at 18° C. The cells were harvested by centrifugation at 4° C. The pelleted cells were suspended in 80 ml of 25 mM Tris–HCl, pH 7.8 containing 5% glycerol and 100 μM GDP. Cell lysis was performed by the addition of lysozyme (0.5 mg/ml, 30 min incubation) followed by sonication at 4 °C. The cell lysate was clarified by centrifugation for 30 min at 4° C. The expressed GTP-binding domain of SEPT7 (SEPT7G), corresponding to residues 29 to 299 of SEPT7_v1, was initially isolated on a 3 ml metal affinity resin column (Talon – Clontech) pre-equilibrated with the lysis buffer and eluted with 6 ml lysis buffer containing 100 mM imidazole. The sample containing SEPT7G was further purified by size exclusion chromatography (Superdex 200 GL 16/60 column, GE Healthcare) in the same buffer at a flow rate of 1 ml/min. Nucleotide content was assessed using perchloric acid denaturation as described previously [27] followed by separation by HPLC (Alliance 2695, Waters) on a Protein Pak DEAE 5 PW column. SEPT7G crystallization conditions were screened using the hanging-drop vapour diffusion method and the Crystal Screen 1 and 2 matrices from Hampton Research at 18 °C followed by subsequent optimization. The best crystals were obtained in droplets containing 2 μl of protein solution at 8 mg/ml and 2 μl of precipitant solution containing 0.1 M Tris pH 8.4; 16% Glycerol; 1.5 M Ammonium Sulphate; 500 μM GDP; 12.5% Isopropanol. X-ray diffraction data were collected at 100 K on beamline ID29 of the ESRF (Grenoble) using a wavelength of 0.9627 Å and a PILATUS 6 M detector [28] in shutterless mode with fine slicing (0.125 s per image) and the oscillation set to half the mosaicity predicted by Mosflm [29] (). 528 images were used for data processing and scaling using the XDS program [30] to a resolution of 3.4 Å. The structure was solved by molecular replacement with Phaser [31] employing the known structure of the SEPT2 GTP-binding domain (pdb code 2QNR) after modification with Chainsaw [32] as the search model. The molecular replacement solution was submitted to density modification procedures with DM [33] and refinement with Phenix [34], using secondary structure restraints, Ramachandran restrains, NCS, group ADP and TLS. Where additional interpretable electron density was apparent, residues were built into the map according to the amino acid sequence. An unpublished structure of SEPT3 also aided in this process (unpublished data). The coordinates and structure factors have been deposited in the PDB under the code 3TW4. SEPT7G eluted from the size exclusion column with an apparent molecular mass of 69 KDa, consistent with a dimer and sufficiently pure for successful crystallization (Fig. 1 ). Under the conditions used, SEPT7G was found to be bound exclusively to GDP (Fig. 1c). It is known that the GTP-binding domains of both septin 2 and septin 4 also elute as dimers when purified in the absence of other septins [24, 35], however it is not clear in all cases which of the two possible interfaces (G or NC) is used for dimerization. In the case of SEPT7G it was imagined that this ambiguity could be resolved by determining its crystal structure. An initial molecular replacement solution in the space group P6122 (LLG = 307 and TFZ = 19.2) was obtained using the coordinates of the SEPT2 GTP-binding domain (2QNR) as a search model. However, during the revision of the current manuscript a structure for residues 29 to 297 of SEPT7 was published by Zent et al. [36] in space group P61. By reducing the space group in our case also to P61 the values for LLG and TFZ improved to 1058 and 36.5, respectively showing this to be the correct solution. The crystals have a Matthews’ coefficient of 3.0 Å3/Da with a solvent content of 58.5% and two molecules in the asymmetric unit. Further statistics can be found in Table 1 . The structure reported here is similar to that described by Zent et al. [36] with the exception of the region of the β7-β8 hairpin mentioned below. The final R and R free values are 28.9% and 36.7%, respectively and the structure reveals intermolecular crystal contacts between SEPT7G monomers which raise interesting questions concerning septin filament formation. Within the hexagonal lattice SEPT7G forms filaments (Fig. 2 ) similar to those described previously for the SEPT2/6/7 hetero-polymer [24]. In the latter, two different interfaces alternate along the filament, maintaining its continuity. The G interface involves elements close to the nucleotide binding site while the NC interface is stabilized by contacts involving the N- and C-terminal regions of the GTP-binding domain. In the hetero-polymer, contacts made between adjacent SEPT2 monomers and between SEPT6 and SEPT7 are of the NC type, whilst those between SEPT2 and SEPT6 or between two monomers of SEPT7 are of the G type (Fig. 2). These are believed to be the native contacts necessary for the formation of a physiological filament. Surprisingly, in our structure determination, we observe that SEPT7G alone is able to employ the same interfaces, leading to the formation of similar filaments within the crystal lattice. It is therefore not possible to unambiguously identify which of the two interfaces stabilizes the dimer in solution, although the G interface is considerably larger and looks the most plausible. Although unexpected, a similar phenomenon has been reported in the case of two independent structure determinations of the SEPT2 GTP-binding domain which also forms filaments within the crystal [24, 25]. Furthermore, it is consistent with the presence of GDP bound to SEPT7G since the only crystal structure which does not form filaments is that of the SEPT2 GTP-binding domain bound to a GTP analogue [37]. This suggests that the nature of the nucleotide and the capacity to form filaments may be related. These observations have led to the concept of promiscuous (non-native) associations amongst septins in which self-self interactions can form in the absence of a more physiologically appropriate binding partner. In the case of the SEPT2 filament the NC interface is native whilst the G interface is promiscuous as it is absent from the hetero-filament (Fig 2). The SEPT7G structure described here shows the opposite phenomenon to occur, since the filament presents native G interfaces together with promiscuous NC interfaces. The picture is therefore completed by demonstrating for the first time that the NC interface is also capable of promiscuous association. Fig. 2 shows that the SEPT7G filament is slightly foreshortened with respect to SEPT2. This is due to closer association of the subunits at the G interface. In the case of SEPT7G the relative orientation of the two monomers around this (native) interface is very similar to that observed in the hetero-filament. In SEPT2, for which the G interface is promiscuous, the need to accommodate F156 in place of its homologue S149 in SEPT7 is probably responsible for inducing a slight separation of the two subunits. This residue lies sandwiched between two septin-specific sequence motifs (sep2 and sep3) [38] and may be important for determining the affinity of self-self interactions at the G interface. It is noteworthy that only SEPT2-like (group III) septins possess phenylanaline at this position, suggesting a possible role in influencing group-specific septin interactions. On comparing SEPT7G to the structure of SEPT7 observed within the SEPT2/6/7 hetero-complex notable differences are observed at the G interface. In the structure reported here, there is partial ordering of the switch II region (residues 101 to 107) which make contact with the same region of the neighbouring subunit across the G interface. This is similar to that observed in the structure of SEPT3. On the other hand residues 58 to 65 (part of switch I) which are ordered in the hetero-oligomer, are disordered in SEPT7G. At the current time it is hard to rationalize all of these differences but the fact that we are studying an isolated domain may be part of the explanation. However, a degree of caution should be exercised in these interpretations given the limited resolution of both structures. In SEPT2G (2QNR) neither of the switches are ordered (Fig. 3 c) but the β7-β8 hairpin is clearly visible at the G-interface. We were unable to interpret the density for this region in our structure of SEPT7G at the current resolution although the recent publication by Zent et al. [36] presents a similar conformation to that seen in SEPT2. Little is currently known about the molecular determinants which control the correct assembly of septin hetero-filaments. However, our data, together with that of others, suggests that due to promiscuity, the G and NC interfaces alone may be insufficient to control selective association. On the one hand, this suggests that other structural elements may be critical for this process but on the other that the promiscuity of the G and NC interfaces may provide a means to generate a wide range of different filaments. It seems likely, based on recent biophysical characterization, that the sequences coding for the coiled coil regions of the C-terminal domains are at least partly responsible for favouring SEPT6-SEPT7 NC interfaces over SEPT7-SEPT7 NC interfaces, leading to specific filament assembly (unpublished results). We gratefully acknowledge the financial support of via its CEPID program (Project number 98/14138-2) to the CBME as well as both CNPq (# 550514/2011-2) and FAPESP (08/57910-0) for funding INBEQMeDI. V.H.B.S. was a fellowship recipient from FAPESP (Project number 07/53412-3).