Chapter 3 - Structural Studies of the Doublecortin Family of MAPs
Introduction
The neuronal microtubule (MT)-associated protein (MAP) doublecortin (DCX) was first characterized in 1998, when it was discovered that mutations in the X-linked doublecortin (DCX) gene cause a syndrome called double cortex (or subcortical band heterotopia, SBH) in females and lissencephaly in males (des Portes et al., 1998, Gleeson et al., 1998). Lissencephaly and SBH patients exhibit a range of symptoms—including epilepsy, intellectual impairment, and infant death—that result from abnormal development of the cerebral cortex. Mutation of the DCX protein causes defective neuronal migration, such that the precisely structured layers of the cortex are poorly organized. Because the DCX gene is on the X chromosome, females with a DCX± genotype exhibit random inactivation of one of the two X chromosomes; this ensures that half of the cells have a functional copy of the gene and migrate correctly into a layered cortex. In contrast, the other half lack a functional copy and subsequently stop half-way through their journey through the developing cortex, creating a heterotopic band of gray matter lying between the cortex and the ventricle, the so-called double cortex. DCX y/- patients possess no functional copy of the DCX protein, resulting in lissencephaly, in which their cortex is abnormally thick and composed of four poorly ordered layers, and these patients exhibit more severe symptoms (Gleeson, 2000). Thus, DCX is essential for migration and differentiation in human neurons. It is enriched at the distal ends of neuronal processes and may regulate MTs in response to extracellular signals in these distal zones to facilitate path finding during development (Francis et al., 1999, Tint et al., 2009). DCX is also essential for neurogenesis in the adult brain (Jin, Wang, Xie, Mao, & Greenberg, 2010).
DCX is now known to be a member of a much larger family of MAPs involved in MT regulation during cell division, migration, and differentiation (Fig. 3.1). DCX–MAPs are widely distributed evolutionarily and are often essential (Reiner et al., 2006): in worms, the DCX–MAP ZYG-8 is important for MT stabilization during asymmetric cell division in embryos (Gönczy et al., 2001), and in flies, a DCX–MAP is essential for the development of mechanotransduction machinery (Bechstedt et al., 2010), demonstrating the conservation and importance of these MAPs. Doublecortin-like kinase (DCLK), initially known as doublecortin and CaM kinase-like 1 (DCAMKL1) (Burgess, Martinez, & Reiner, 1999), is the closest homologue to human DCX. It has several splicing isoforms, with a DCX-like isoform that is 72% identical to DCX. Full-length (FL) DCLK is a 729-amino-acid protein, with a C-terminal serine/threonine–protein kinase domain, similar to CaM kinase II. The similarity of DCX and DCLK is highlighted by their functional compensation in mice, where the activity of both proteins must be perturbed to recapitulate the severity of the human lissencephaly phenotype (Deuel et al., 2006, Kerjan et al., 2009, Koizumi et al., 2006). In humans, DCX paralogs also include RP1, a protein that is mutated in retinitis pigmentosa (a common form of inherited blindness) and whose MT-stabilizing activity is essential for photoreceptor cell development (Liu, Zuo, & Pierce, 2004). In addition, Dcdc2, another DCX–MAP encoding gene, is linked with developmental dyslexia (Meng et al., 2005), and its protein product is involved with the control of primary cilia size and activity in neurons (Massinen et al., 2011).
The DCX 366-amino-acid sequence (40 kDa) shows no homology to other classical neuronal MAPs. In fact, sequence analysis reveals that DCX is built from an N-terminal tandem of homologous 90-amino-acid (11 kDa) domains which were accordingly named DC domains; these domains are separated by a well-conserved but presumed unstructured linker and are followed by a presumed intrinsically unstructured C-terminal serine/proline-rich (S/P-rich) domain (Fig. 3.1). Point mutations causing lissencephaly cluster within the two DC domains and modify the DCX–MT interaction in transfected cells (Bahi-Buisson et al., 2013, Sapir et al., 2000, Taylor et al., 2000). These data established the DC domains of DCX as MT-binding domains and reinforced the importance of DCX–MT binding during neuronal cortical migration.
An NMR study revealed the solution, β-grasp-like structure of the recombinant N-terminal DC domain (N-DC), and initiated the exploration of structure–function relationships of disease-causing point mutations in this domain (Kim et al., 2003). This analysis allowed discrimination between mutations of buried residues—affecting folding and stability of the protein—and of surface residues affecting putative interactions with DCX’s binding partners, including tubulin and MTs. However, an equivalent study of C-DC and of longer constructs proved technically challenging, leaving many aspects of DCX’s unique MT-binding mechanism unresolved.
We were fascinated by the idea that although neurons are full of MT-stabilizing proteins, DCX has unique properties that cannot be functionally compensated for by other neuronal MAPs. To address this mystery, we used established structural approaches for studying MAPs (Amos and Hirose, 2007, Hoenger and Gross, 2008).
Our early work revealed several distinctive properties of DCX and its interaction with MTs (Moores et al., 2004, Moores et al., 2006). An emerging area of interest is the sensitivity of MAPs to tubulin-bound nucleotide (Maurer et al., 2011, Zanic et al., 2009); however, we found that DCX stabilizes MTs independent of the bound nucleotide, it does not affect the intrinsic GTPase of tubulin polymer, and the lattice parameters of the DCX-stabilized MTs reflect the nucleotide that is bound in the lattice (Fourniol et al., 2010). In our hands, DCX does not enhance MT growth rates but blocks depolymerization (Moores et al., 2006). All of these effects occur at substoichiometric or close to stoichiometric ratios of DCX:tubulin and appear to operate independent of MT bundling, which is the primary and apparently nonspecific outcome of DCX–MT interactions at superstoichiometric ratios of DCX both in vivo and in vitro (Sapir et al., 2000).
One of the most striking properties of DCX is its ability to specifically nucleate and bind MTs with a 13-protofilament (13-pf) architecture (Bechstedt and Brouhard, 2012, Moores et al., 2004). This is particularly important because although the number of pfs within MTs polymerized in vitro varies between 8 and 19, the almost exclusive architecture found in vivo is the 13-pf MT (McIntosh et al., 2009, Tilney et al., 1973, Wade and Chrétien, 1993). Cryo-electron microscopy (cryo-EM) continues to be an invaluable method for elucidating mechanism and function in the MT cytoskeleton and its binding partners (Amos and Hirose, 2007, Hoenger and Gross, 2008), and the stabilization mechanism and architecture specificity of DCX were in large part explained by cryo-EM reconstructions.
Our early, low-resolution (~ 30 Å) reconstruction was calculated using helical analysis of a few, rare 14-pf paclitaxel-stabilized MTs that could be decorated with a truncated construct of DCX (t-DCX, DCX 1–275, lacking the S/P-rich domain; Moores et al., 2004). Because of its architectural preference, FL DCX could not be analyzed using this method. This first reconstruction showed a globular density wedged between pfs and in contact with four tubulin monomers. The DCX density corresponds to only one DC domain while additional weak density that may be attributable to the rest of the protein was observed at higher radius. It was immediately obvious that DCX’s binding between pfs represents an excellent way to “staple” pfs together and thus increases MT stability. By binding at the junction between four tubulin monomers, DCX has the potential to strengthen both lateral contacts between pfs and also longitudinal contacts along the MT. In addition, this unique binding mode suggests the key to DCX’s specificity for 13-pf MTs: the width of the inter-pf valley varies with pf number and DC domain may have evolved to fit this binding site in the 13-pf MT wall (Fig. 3.2A).
This binding site does not overlap with those of the MT-based motors kinesin and dynein and, thus, would be predicted not to impede movement of these motors (Fig. 3.2B). Indeed, in an ensemble gliding assay, DCX–MTs were found to support kinesin motility and with only slightly decreased speed (Moores et al., 2006). This suggests that, in vivo, MTs stabilized by DCX (and its relatives) may act as tracks for intracellular transport; the ramifications of this for cargo delivery control are an active area of ongoing research (Deuel et al., 2006, Liu et al., 2012).
Yet, after this initial work, several key questions remained and have been the focus of our more recent research efforts. For example, the biochemical stoichiometry of binding to tubulin for both t-DCX and FL DCX is estimated to be 1:1, suggesting that each density in our reconstruction corresponds to a single DC domain, with the rest of the molecule disordered and therefore invisible in the reconstruction. However, it is possible that adjacent binding sites are occupied by N-DC and C-DC from the same molecule—the inter-DC linker is long enough that either a longitudinal or lateral configuration is possible. Our hypothesis was that in working with t-DCX—in the absence of the S/P-rich domain that is critical for conferring MT architecture specificity—on non-13-pf MTs at low resolution, we were compromising our structural experiment. In addition, the low resolution of the reconstruction prevented us from determining whether DCX binds between two dimers or at the corner of four dimers. We also wanted to understand how DCX would stabilize MTs in the absence of the stabilizing drug paclitaxel. Therefore, we have focused our efforts on subnanometer resolution structural studies of 13-pf MTs nucleated and stabilized by FL DCX.
Early biochemical studies used recombinant DCX expressed in E. coli (Sapir et al., 2000, Taylor et al., 2000). However, when we started our work on DCX, our bacterially expressed protein produced primarily large 3D bundles of MTs that were impossible to study structurally, even at low-protein concentrations. On the other hand, DCX protein expressed in Spodoptera frugiperda insect cell cultures was stable and could be readily studied in the variety of ways described below. Others have had success subsequently with molecular studies of bacterially expressed DCX including recapitulation of DCX’s selectivity for 13-pf MTs (Bechstedt & Brouhard, 2012), but the proteins’ concentrations used were typically lower than required for our structural studies.
The 366-amino-acid isoform of human DCX was cloned in the Invitrogen Bac-to-Bac system by Dr Fiona Francis, Institut du Fer à Moulin, Paris. This isoform is a splicing variant from the 360-amino-acid form, containing five residues (GNDQD) inserted after residue 310 and one Val inserted after residue 342 (nomenclature of the 360-amino-acid form; NCBI Reference Sequence database); the significance of these differences is not understood. The engineered construct includes in its N-terminus a 6-Histidine affinity tag (His-tag), a TEV (tobacco etch virus) protease cleavage site (ENLYFQG, cleavage occurs between Q and G), and a Flag-tag (DYKDDDDK) used originally for immunofluorescence in transfected cells (Sapir et al., 2000).
The baculovirus genome containing the tagged DCX gene was transfected into Sf9 cells and amplified using standard methods (Bac-to-Bac Baculovirus Expression System, Life Technologies). Test expressions of DCX constructs were performed in small flasks; fewer cells with larger average diameters are indicative of baculovirus infection. Sf9 cells are usually round but the expression of DCX constructs causes their shape to change: they become lemon shaped or even grow a long cytoplasmic extension, which is presumably due to extensive MT bundling by DCX (up to 200 μm) (Fig. 3.3A).
For large-scale protein expression, cells were passaged to 450-mL cultures in 2.5-L spinners containing fresh medium, to a final concentration of 1 × 106 cells/mL. They were grown for 24 h, so that they reached ~ 2 × 106 cells/mL at the infection time point 24 h later. In our hands, 50 mL P2 was added to a 450-mL culture to achieve a multiplicity of infection (MOI = number of virions/number of cells) of 3–10, which stops cell multiplication and induces protein expression. Cultures are incubated for 48 h at 27 °C. Cells are harvested by centrifuging in 1-L tubes at 700 × g for 20 min at 4 °C in a Beckman Avanti J-20 I centrifuge. Pelleted cells were washed with 40 mL cold PBS per 1 L of culture and centrifuged in 50-mL tubes at 700 × g another 20 min in an Eppendorf 5810R centrifuge at 4 °C. Pellets were frozen in liquid nitrogen and stored at − 80 °C until protein extraction.
The protocol originally described in Moores et al. (2004) was modified slightly to improve the solubility and minimize protein degradation, together leading to better yields. All protein purification steps and centrifugations were performed at 4 °C to limit protease activity. Buffers used are listed in Table 3.1.
Extraction of cytoplasmic proteins is performed by incubating frozen cell pellets (typically 20 mL for 2-L cultures) with an equal volume of hypotonic lysis buffer (lysis—low salt in Table 3.1), for 30 min. The combined action of defrosting, osmolysis, and detergent (Triton) lyses the plasma membrane and releases cytoplasmic content. Freeze–thawing cycles cannot be employed as DCX degrades. A number of inhibitors are used to limit the action of a broad spectrum of proteases, and β-mercaptoethanol is indispensible to increase the solubility of DCX, which contains nine cysteines.
To extract DCX from large MT bundles, the salt concentration is brought to ~ 350 mM NaCl by addition of high-salt lysis buffer and incubation for 15 min. To maximize extraction, the lysate is homogenized using a Dounce homogenizer (Wheaton) and the resulting lysate is centrifuged for 45 min at 45,000 × g to pellet cell debris and inclusion bodies (Fig. 3.3B). A lipid fraction is present at the top of the supernatant. The nonlipid supernatant contains soluble proteins and is removed for further purification.
His-tagged DCX is purified by affinity chromatography: 1 mL Ni-NTA resin (Qiagen) per 1 L of culture is equilibrated with wash buffer, then added to the soluble lysate, and incubated for 1 h on a rotating wheel at 4 °C. The relatively high imidazole concentration (50 mM) reduces nonspecific binding. The resin is then deposited on a small gravity column and the flow-through discarded (Fig. 3.3B). Two successive 5-column volume washes help minimize the contaminants bound to the resin. The His-DCX is eluted with 250 mM imidazole. This step both purifies and concentrates the protein of interest, yielding typically 20–30 mg in 3-mL buffer, starting from a 2-L culture. However, His-tagged degradation products cannot be separated, and other contaminants may remain.
The purified 42 kDa His-DCX is loaded on a HiLoad Superdex 75 16/60 column (GE Healthcare) and elutes with a peak at 53 mL, monitored by AKTA FPLC (GE Healthcare) (Fig. 3.3C). This step purifies the protein and allows buffer exchange but dilutes it ~ 10-fold, typically yielding 8–12 mg protein in 10-mL buffer.
Finally, the His-tagged DCX is cleaved using recombinant tobacco etch virus protease (rTEV; gift from EMBL Heidelberg). Complete cleavage is achieved by adding 1% w/w rTEV (protease/fusion protein) and incubating at 4 °C overnight (Fig. 3.3D). A second Ni-NTA column (300 μL resin) retains uncleaved His-DCX as well as the rTEV; addition of 20 mM imidazole in the column buffer reduces unspecific interaction of the cleaved protein with the resin. In the original protocol, gel filtration was performed last but the modified order improved yields by reducing degradation and aggregation (see Fig. 3.3Cii). The His-tag may make the protein more soluble, while performing the gel filtration at an early stage helps eliminate contaminants including proteases that might cause both instability and degradation.
Finally, the ~ 10 mL dilute flow-through is concentrated in a 10,000-molecular-weight cut-off concentrator (Vivaspin 15R; Sartorius). Centrifugation at 4000 × g, always at 4 °C, for 1 h concentrates DCX approximately 5 × (~ 2 mL at 5 mg/mL). The concentrated protein is aliquoted and snap-frozen in liquid nitrogen and keeps for several years at − 80 °C.
In the different MT architectures polymerized in vitro, lateral contacts between pfs are usually homotypic—α–α and β–β—forming a so-called B-lattice. However, heterotypic contacts—α–β and β–α, the so-called A-lattice—also occur (Wade & Chrétien, 1993). Although the precise in vivo significance for MT architecture is far from clear, crucially for our studies, 13-pf MTs show a line of discontinuity called the seam, where the dominant B-lattice contacts are disrupted by a line of A-lattice contacts (Kikkawa, Ishikawa, Nakata, Wakabayashi, & Hirokawa, 1994). However, 13-pf MTs built entirely of A-lattice contacts have also been reported (des Georges et al., 2008).
Subnanometer resolution structure determination by single-particle processing requires an initial 3D reference model, which implies some prior knowledge of the structure of interest. In our case, it was well established that copolymerized DCX–MTs are 13-pf, but it was unclear whether they were made of mainly A- or B-lattice contacts or a mixture. Although previous reports showed that truncated DCX bound to paclitaxel-stabilized MTs that are built of a helical B-lattice (Moores et al., 2004), we considered the hypothesis that DCX–MTs might contain a proportion of A-lattice contacts. Cryo-ET permits 3D reconstruction from a single macromolecular complex without averaging or model bias and is thus ideally suited to determine the lattice parameters of a population of MTs. McIntosh et al. (2009) successfully applied cryo-ET to MTs extracted from cells and decorated with a kinesin motor domain, and observed primarily B-lattice MTs. We took a similar approach, using K340T93N, a 340-residue Thr93Asn mutant motor domain of rat conventional kinesin to emphasize the underlying lattice of our DCX–MTs (Crevel et al., 2004). This mutant has a very strong affinity for MTs facilitating full decoration but did not alter the stoichiometry of DCX binding (validated by cosedimentation assay, data not shown).
DCX–MTs were polymerized by incubating bovine tubulin (Cytoskeleton, Inc.) and human recombinant DCX in equimolar amounts (10 μM) at 37 °C for 1 h. To maximize decoration of MTs with DCX and kinesin, DCX–MTs were diluted 1:1 in the low-ionic-strength buffer BRB20 (20 mM PIPES, pH 6.8, 1 mM EGTA, 1 mM MgCl2) with 2 mM TCEP and adsorbed to glow-discharged lacey carbon grids (Agar). A second solution containing 30 μM DCX and 5 μM kinesin (K340T93N) in BRB20, 2 mM TCEP was mixed 3:1 with colloidal gold (Sigma) and applied to the grids. Grids were then transferred into a Vitrobot (FEI) set to 37 °C and 100% humidity, to prevent evaporation and consequent changes in ionic strength. They were blotted for 2 s and instantaneously vitrified by rapid plunging into liquid ethane.
For each tomogram, 61 images of the sample tilted from − 60° to + 60° were recorded on a 2k × 2k CCD (Gatan) on a Polara microscope (FEI Company) operating at 300 kV, at 6–8 μm defocus. Seven tomograms were reconstructed using the FEI software Inspect3D. Data collection and processing were performed with the generous help of Dr. Dan Clare (Birkbeck College).
A total of 5 μm of DCX–kinesin–MT tomographic reconstructions were visually inspected and all were found to be B-lattice 13-pf MTs (Fig. 3.4). This validated the use of B-lattice parameters in subsequent single-particle structure determination of DCX–MTs.
Using cryo-ET, we demonstrated that DCX promoted the assembly of B-lattice 13-pf MTs. These MTs contain a discontinuity—the seam—that means that they are not amenable to helical reconstruction. Instead 13-pf MTs can be reconstructed using so-called single-particle image processing. The method developed by Charles Sindelar (Yale, CT; Sindelar and Downing, 2007, Sindelar and Downing, 2010), referred to as “Chuff,” yielded an ~ 8 Å resolution reconstruction of kinesin-decorated, DCX-stabilized 13-pf MTs, the same sample that we had validated by cryo-ET (Section 2.2) (referred to below as DCX–K–MTs; Fourniol et al., 2010).
Sample preparation was as described in Section 2.2.2. Low-dose images were collected on a Tecnai F20 FEG electron microscope (FEI Company), operating at 200 kV, 50,000 ×, and 0.8–2.9 μm defocus. Micrographs were recorded on film (SO-163; Kodak) and were digitized (SCAI scanner; Carl Zeiss, Inc.) to a final sampling of 1.4 Å/pixel.
Chuff is a fully automated set of scripts that simply require the user to set up the data and parameter files. To speed up image processing, and because finer sampling did not significantly improve resolution, micrographs digitized to 1.4 Å/pixel were binned by a factor of 2 to a final sampling of 2.8 Å/pixel. For our study, DCX–K–MTs were boxed with the EMAN Boxer program (option helix/normal; Ludtke, Baldwin, & Chiu, 1999) into particles/segments of 432 × 432 pixels, containing 7–8 tubulin dimer axial repeats. The input for the DCX–K–MT reconstruction was 4772 segments coming from 172 MTs and 63 cryo-electron micrographs.
We used the default initial 3D model of kinesin-decorated 13-pf B-lattice MT generated by Chuff for alignment. In particular, the kinesin head density bound every dimer greatly facilitates identification of the position of the MT seam. The absence of DCX in the initial model gave us confidence that additional density observed in the output 3D reconstruction is indeed a faithful reflection of the MT-bound DCX in our sample. To further minimize the risk of model bias, high-resolution information was deleted using a low-pass filter with a 15 Å cut-off. Two-dimensional projections of the model with φ varying between 0° and 359°, θ between 75° and 105°, both with a 1° increment, resulted in a set of 11,160 references.
In Chuff, a first round of reference-based alignment is performed in SPIDER (Frank et al., 1996), where MT segments are cross-correlated to the references. Each segment cut along one MT image is assigned a seam orientation (φ angle). Because the seam runs parallel to the axis of the MT, one would expect that the φ values assigned to segments from the same 13-pf MT would be very similar. Typically, however, they vary by multiples of ~ 28° (360/13) because cross-correlation scores computed between references whose seam is rotated by an integer numbers of pfs and relatively noisy cryo-EM images are very similar; hence, the best score does not always correspond to the actual orientation of the seam in the image (Sindelar & Downing, 2007). Alternatively, variations of the φ angle can arise from MTs with different architectures, and Chuff contains AWK scripts that analyze the output φ angles from reference-based alignment and decide whether each MT has consistent enough seam orientations (φ values) to be kept for the reconstruction. This decision depends on a number of parameters that can be defined by the user, in particular, the size of the angular window where φ values can be considered to be consistent between one another (default: 20°), and the minimum fraction of the boxes that must be in this window (default: 20%). The default selection parameters were successfully used to process DCX–K–MTs and approximately 90% of the dataset passed this selection process. If the MT is accepted, the angles assigned are edited so that all segments in the MT have φ values within the previously determined angular window. Subsequent rounds of reference-based alignment are restricted to this φ window and performed to refine alignment parameters and Euler angles before the final 3D reconstruction.
Reconstruction of DCX–K–MTs with no symmetry imposed yielded a 13.5 Å resolution 3D map (FREALIGN option helical_subunits = 0 (Grigorieff, 2007); Fig. 3.5A; EMDB ID 1787). This asymmetric reconstruction confirmed that the MT (cyan) is occupied by both kinesin motor domain (red) and DCX (yellow), which binds at the corner of four tubulin dimers, stabilizing both lateral and longitudinal tubulin–tubulin contacts. Remarkably, DCX does not bind at the A-lattice seam of the 13-pf MT (right panel, arrow), but only at the 12 B-lattice inter-pf grooves (Fourniol et al., 2010).
To gain further insight, the 12 inter-pf valleys bound by DCX were averaged together. The averaging of approximately 168,000 decorated tubulin dimers generated an 8.2-Å resolution map (FREALIGN option helical_subunits = 12; Fig. 3.5B; EMDB ID 1788). At this resolution, secondary structures are resolved: alpha helices and beta sheets appear, respectively, as rods and sheets of density.
The detail in our 8.2 Å structure of DCX–K–MTs allowed us to dock atomic structures of each constituent subunit into our reconstruction in order to generate a pseudo-atomic model of DCX-stabilized MTs. Such modeling provides invaluable data about the binding interfaces between the components and the conformational changes they undergo in the context of the physiologically relevant macromolecular complex. We focus here on the interactions between DCX and tubulin; the implications of kinesin binding on DCX have recently been discussed elsewhere (Liu et al., 2012).
UCSF Chimera (Pettersen et al., 2004) was used for visualization of 3D models and rigid-body fitting of atomic structures in the cryo-EM volumes. Several crystal structures of α- and β-tubulin (1JFF.pdb (Löwe, Li, Downing, & Nogales, 2001), 3HKE.pdb (Dorléans et al., 2009)) were fitted independently in their respective densities in the DCX–K–MT reconstruction because we wanted to evaluate the conformation of each monomer separately in our DCX-stabilized MTs formed in the absence of paclitaxel. The quality of the fits was assessed by calculating the cross-correlation between our structure and a simulated 8 Å model for each atomic structure: β-tubulin from 1JFF gave the best score of 0.651 in our β-tubulin density (3HKE β-tubulin chain B gave 0.599), α-tubulin from 1JFF and 3HKE chain A gave similar scores, respectively, 0.605 and 0.607, whereas a chimeric structure increased the cross-correlation slightly to 0.614 (3HKE chain A residues 31–61, 69–92, and 275–298, corresponding to the N, H2S3, and M loop regions, respectively, involved in lateral contact formation) were substituted into 1JFF chain A. To build the pseudo-atomic model of the DCX–MT complex, four tubulin subunits and the N-DC domain of DCX (1MJD.pdb, model 11, residues 46–140) were placed independently in the EM density using Chimera, and Flex-EM was used for the refinement of the multiple subunit fitting (Topf et al., 2008). The best multicomponent fit had a cross-correlation value of 0.819 (PDB ID 2XRP). This pseudo-atomic model allowed us to visualize the precision with which the DC domain contacts four tubulin dimers at its binding site and revealed several disease-causing mutations in both DCX and tubulin at their binding interface (Bahi-Buisson et al., 2013, Fourniol et al., 2010).
High-resolution information about MTs and straight pf structures has mainly been derived from paclitaxel-stabilized tubulin assemblies (Li et al., 2002, Löwe et al., 2001). In fact, paclitaxel appears to be a promiscuous stabilizer, binding MTs with varying pf numbers, sheets of antiparallel and inverted pfs (Nogales, Wolf, & Downing, 1998), or short-straight pfs in solution (Elie-Caille et al., 2007). These data imply that paclitaxel stabilizes a conformation of tubulin that favors straight pfs but that is not necessarily specific to MTs. However, because there has been no high-resolution model of MTs in the absence of paclitaxel, it has been unclear whether the conformation stabilized by the drug is a native state of tubulin or is induced by drug binding. We found that the paclitaxel-binding pocket is present in our reconstructions, demonstrating that paclitaxel stabilizes a native conformation of polymerized tubulin (Fig. 3.5C). Our data thus reveal a native structure of MTs bound solely by cellular ligands, where the straight pf structure appears virtually identical to that found in zinc-induced sheets.
Our structure confirms the importance of the lateral contacts formed by the M, N, and H2S3 loops previously visualized by Li et al. (2002) and more recently by Sui and Downing (2010). While the procedure employed by Downing and colleagues in these studies averaged together α- and β-tubulin, our reconstruction discriminates between them. This enabled us to clearly verify that α–α and β–β lateral contacts—and at a lower resolution, the seam α–β and β–α lateral contacts—have similar densities. This is despite the divergence of the M and N loop sequences between α- and β-tubulin.
These native tubulin–tubulin contacts within GDP MTs can now be compared with other MT reconstructions. We have recently visualized another structural state of MTs in the absence of stabilizing drugs, but containing the nonhydrolyzable GTP analog GTPγS and bound by another MAP—the fission yeast end binding protein (EB) Mal3 (Maurer, Fourniol, Bohner, Moores, & Surrey, 2012). GTPγS acts as a static mimic of the otherwise dynamic binding site that is specifically recognized by Mal3 and other EBs at growing MT ends (Maurer et al., 2011). The comparison of this MT end-like structure, with the GDP MT lattice structure in the presence of DCX, shows that growing MT ends possess an additional layer of lateral contacts at higher radius, which likely explains their action as a stabilizing cap (see also Yajima et al., 2012). Thus, the comparison of cryo-EM maps of MTs at secondary-structure resolution is revealing the structural basis of MT dynamic instability.
Cryo-EM has been an essential tool in shedding light on MT stabilization by DCX. Nevertheless, critical aspects of our current data continue to limit our understanding of the molecular mechanism of this essential protein. One intriguing aspect of our reconstructions continues to be that we currently only visualize a single DC-shaped density in our structures, that is, ~ 1/4 of the FL DCX molecule. This is consistent with the original low-resolution cryo-EM reconstruction of paclitaxel-stabilized MTs bound with the t-DCX construct (Fig. 3.2; Moores et al., 2004), thereby disproving the hypothesis that our ability to visualize the entire DCX molecule was limited by the original experimental conditions. In addition, our recent analysis of the structure of DCX–MTs in the absence of bound kinesin has revealed a specific conformation for the linker regions on either side of the bound DC domain that strongly suggests the density visualized in all our reconstructions corresponds to N-DC (Cierpicki et al., 2006, Liu et al., 2012).
Recent single-molecule studies have revealed the cooperative nature of MT binding by DCX, implying that DCX molecules contact each other when present at close to stoichiometric concentrations on the MT lattice (Bechstedt & Brouhard, 2012). Unfortunately, our structures currently provide no information about these interactions: N-DC—the only domain of DCX we have visualized—is not by itself sufficient either for nucleation, stabilization, or 13-pf specification, for which C-DC and the C-terminal domain are required (Kim et al., 2003, Moores et al., 2004, Sapir et al., 2000, Taylor et al., 2000). One explanation is that the samples used for our subnanometer resolution reconstructions are end-points of the DCX-mediated nucleation and stabilization process. Therefore, some of our current efforts are directed toward a structural understanding of early species in the process of DCX-mediated MT nucleation. A parallel possibility is that the missing ~ 75% of the DCX molecules are not sufficiently ordered for us to visualize using our current averaging procedures. By computationally investigating the conformational variability of DCX within our samples, we aim to gain further insight into its location and structure, albeit with a likely compromise of resolution in the resulting reconstructions.
Based on the availability of the structure C-DC of DCDC2 (PDB ID 2DNF), it seems likely that the conserved DC fold is important for C-DC function and may involve tubulin interaction and control of DCX cooperativity (Bechstedt and Brouhard, 2012, Kim et al., 2003). However, structural insight into the C-terminal S/P-rich domain of DCX—previously shown to be essential for DCX’s highly specific binding of 13-pf MTs (Moores et al., 2004)—remains lacking. As it is predicted to be disordered and highly charged due to phosphorylation (Reiner et al., 2004), one hypothesis is that the C-terminal domain does not make specific contact with tubulin but rather modulates the DCX–MT interaction indirectly. This would predict that single point mutations in the S/P-rich domain would only indirectly affect the interaction and in fact hardly any disease-causing mutations in this domain have been reported (Bahi-Buisson et al., 2013). To further test this hypothesis, it would be interesting to generate a DCX construct with a scrambled C-terminal sequence, keeping the same phosphorylation sites and overall charge.
Our structural insight into the DCX–MT interface provides a framework to localize and predict the severity of the gradually accumulating examples of disease-causing point mutations in DCX. Our results suggest that a spectrum of cellular effects would be seen, with some mutations totally disrupting the DCX–MT interaction, while others might have only minor effects on the MT interface and exert their mutagenic effects elsewhere within the network of DCX binding partners, including via intracellular trafficking control (Liu et al., 2012). It will be informative to continue to test the effects of these mutations in vitro to investigate their consequences on the DCX structure and function (Bechstedt & Brouhard, 2012). Clinically, the rarity of these mutations presents a challenge for establishing a statistically robust diagnostic link between genotype and phenotype; in vitro studies could provide functional parameters to predict the exact nature and severity of the disease based on the location and nature of the mutation (Bahi-Buisson et al., 2013). Multiple molecular pathways to disease may emerge from such studies with DCX forming a MT-based regulatory hub.
Section snippets
Acknowledgments
We thank Charles Sindelar (Yale University) for sharing his MT reconstruction scripts and members of the Birkbeck EM group for helpful discussions and advice. We are supported by The Wellcome Trust, New Life and Fédération pour la Recherche sur le Cerveau.
References (52)
- et al.
Doublecortin recognizes the 13-protofilament microtubule cooperatively and tracks microtubule ends
Developmental Cell
(2012) - et al.
Genetic interactions between doublecortin and doublecortin-like kinase in neuronal migration and axon outgrowth
Neuron
(2006) - et al.
Straight GDP-tubulin protofilaments form in the presence of taxol
Current Biology
(2007) - et al.
Doublecortin is a developmentally regulated, microtubule-associated protein expressed in migrating and differentiating neurons
Neuron
(1999) - et al.
SPIDER and WEB: Processing and visualization of images in 3D electron microscopy and related fields
Journal of Structural Biology
(1996) - et al.
Doublecortin, a brain-specific gene mutated in human X-linked lissencephaly and double cortex syndrome, encodes a putative signalling protein
Cell
(1998) - et al.
zyg-8, a gene required for spindle positioning in C. elegans, encodes a doublecortin-related kinase that promotes microtubule assembly
Developmental Cell
(2001) FREALIGN: High-resolution refinement of single particle structures
Journal of Structural Biology
(2007)- et al.
Structural investigations into microtubule-MAP complexes
Methods in Cell Biology
(2008) - et al.
Doublecortin-like kinase functions with doublecortin to mediate fiber tract decussation and neuronal migration
Neuron
(2006)
Microtubule structure at 8 A resolution
Structure
Molecular basis for specific regulation of neuronal kinesin-3 motors by doublecortin family proteins
Molecular Cell
Refined structure of alpha beta-tubulin at 3.5 A resolution
Journal of Molecular Biology
EMAN: Semiautomated software for high-resolution single-particle reconstructions
Journal of Structural Biology
EBs recognize a nucleotide-dependent structural cap at growing microtubule ends
Cell
Lattice structure of cytoplasmic microtubules in a cultured Mammalian cell
Journal of Molecular Biology
Mechanism of microtubule stabilisation by doublecortin
Molecular Cell
Structural basis of interprotofilament interaction and lateral deformation of microtubules
Structure
Patient mutations in doublecortin define a repeated tubulin-binding domain
Journal of Biological Chemistry
Protein structure fitting and refinement guided by cryo-EM density
Structure
Cryoelectron microscopy of microtubules
Journal of Structural Biology
Studying the structure of microtubules by electron microscopy
Methods in Molecular Medicine
New insights into genotype-phenotype correlations for the DCX-related lissencephaly spectrum
Brain
A doublecortin containing microtubule-associated protein is implicated in mechanotransduction in Drosophila sensory cilia
Nature Communications
KIAA0369, doublecortin-like kinase, is expressed during brain development
Journal of Neuroscience Research
The DC-module of doublecortin: Dynamics, domain boundaries, and functional implications
Proteins
Cited by (17)
A Combinatorial MAP Code Dictates Polarized Microtubule Transport
2020, Developmental CellCitation Excerpt :Interestingly, MAP7 inhibits kinesin-3 but does not substantially affect dynein motility (Monroy et al., 2018). Doublecortin (DCX) and its paralog, doublecortin-like kinase-1 (DCLK1) robustly stimulate microtubule polymerization (Bechstedt and Brouhard, 2012; Fourniol et al., 2013; Patel et al., 2016) but are restricted to the distal dendrites and axonal growth cones (Lipka et al., 2016; Liu et al., 2012; Tint et al., 2009), indicating they may have specific roles commensurate with their localization patterns. Both MAPs have been reported to interact with the kinesin-3 motor domain and promote kinesin-3 cargo transport within the dendrites (Lipka et al., 2016; Liu et al., 2012).
Pathogenic E2K mutation of doublecortin X (DCX) alters microtubule stabilisation and actin filament association
2019, Biochemical and Biophysical Research CommunicationsCitation Excerpt :Indeed, severe developmental defects can arise as a result of impaired neuronal migration, as observed in individuals with mutations in DCX, a neuron-specific MT-associated protein that is also now known to associate indirectly with F-ACT [3–5]. Importantly, many pathogenic mutations mapped to the two homologous doublecortin (DC) domains disrupt DCX protein association with the cytoskeleton [6–9]. However, these structured DC domains (DC1: DCX residues 45-150; DC2: DCX residues 170-275) are flanked by additional sequences: an unstructured N-terminal sequence (DCX-N: DCX residues 1-44) and a C-terminal sequence (DCX-C: DCX residues 276-366) [10,11].
Doublecortin X (DCX) serine 28 phosphorylation is a regulatory switch, modulating association of DCX with microtubules and actin filaments
2019, Biochimica et Biophysica Acta - Molecular Cell ResearchCitation Excerpt :Structurally, the DCX protein consists of two homologous doublecortin (DC) domains, DC1 (the N-terminal DC domain, residues 45–150) and DC2 (the C-terminal DC domain, residues 170–275) that are defining features of the larger DCX family of proteins [9]. These DC domains are linked in tandem via a flexible unstructured region (linker) and additionally flanked by a likely unstructured N-terminal region (DCX-N, i.e. residues 1–44) and serine/proline-rich C-terminal region (DCX-C, i.e. residues 275–366) [9,10]. The structured DC domains interact with MTs, specifically via binding to four neighboring tubulin dimers [11].
Doublecortin-Like Kinases Promote Neuronal Survival and Induce Growth Cone Reformation via Distinct Mechanisms
2015, NeuronCitation Excerpt :Because of the existence of multiple neuronal MT-associated proteins (MAPs), we next examined whether overexpression of other MAPs could mimic the effects of DCX-270. As shown in Figure S7, AAV-mediated expression of Tau and EB3, two well-characterized MT-binding proteins (Goedert et al., 1991; Straube and Merdes, 2007), had no significant effects on neuronal survival and axon regeneration after optic nerve injury in WT or PTEN−/− backgrounds, suggesting that the unique MT-binding property of DCX isoforms (Fourniol et al., 2013) might be important for their effects on axon regeneration. In this regard, it has been suggested that different from other MAPs that interact directly on the surface of the MT protofilaments, DCXs bind in the recess between the protofilaments, which might be uniquely suitable for MT stabilization.
- 1
Current address: Cancer Research UK, London Research Institute, Lincoln's Inn Fields Laboratories, 44 Lincoln's Inn Fields, London WC2A 3LY, United Kingdom.