Address correspondence and reprint requests to Bernhard Lohkamp, Department of Medical Biochemistry and Biophysics, Karolinska Institutet, Tomtebodavägen 6, SE-17177 Stockholm, Sweden. E-mail: firstname.lastname@example.org
Collapsin response mediator protein-5 (CRMP-5) is the latest identified member of the CRMP cytosolic phosphoprotein family, which is crucial for neuronal development and repair. CRMPs exist as homo- and/or hetero-tetramers in vivo and participate in signaling transduction, cytoskeleton rearrangements, and endocytosis. CRMP-5 antagonizes many of the other CRMPs' functions either by directly interacting with them or by competing for their binding partners. We determined the crystal structures of a full length and a truncated version of human CRMP-5, both of which form a homo-tetramer similar to those observed in CRMP-1 and CRMP-2. However, solution studies indicate that CRMP-5 and CRMP-1 form weaker homo-tetramers compared with CRMP-2, and that divalent cations, Ca2+ and Mg2+, destabilize oligomers of CRMP-5 and CRMP-1, but promote CRMP-2 oligomerization. On the basis of comparative analysis of the CRMP-5 crystal structure, we identified residues that are crucial for determining the preference for hetero-oligomer or homo-oligomer formation. We also show that in spite of being the CRMP family member most closely related to dihydropyrimidinase, CRMP-5 does not have any detectable amidohydrolase activity. The presented findings provide new detailed insights into the structure, oligomerization, and regulation of CRMPs.
The collapsin response mediator protein (CRMP) phosphoprotein family was first described as an intracellular component of the transduction pathway of the extracellular semaphorin 3A signal (Goshima et al. 1995). Its most recent member was simultaneously identified by three independent studies and reported under the names CRMP-5 (Fukada et al. 2000), CRAM (CRMP3-associated molecule) (Inatome et al. 2000), and unc33-like phosphoprotein 6 (Horiuchi et al. 2000). It is also sometimes referred to as dihydropyrimidinase-related protein 5. While the amino acid sequences of CRMPs 1-4 are ~70-75% identical, CRMP-5 shares ~50% sequence identity with CRMPs 1-4 and dihydropyrimidinase (DHPase) (Figure S1). It is thus equally distant from both non-enzymatic and enzymatic homologs and was suggested to be categorized into a third subfamily (Fukada et al. 2000). Within the protein family, several CRMPs (1, 2, and 4) occur in two alternatively spliced isoforms, with the less ubiquitously expressed ones containing an extra N–terminal domain of unknown structure and function.
CRMP-5 is crucial for and has unique roles in neuronal development (Veyrac et al. 2011; Yamashita et al. 2011; Camdessanché et al. 2012). Like all the CRMP family members, it is highly expressed in the developing brain as well as regions of neurogenesis of the adult brain (Fukada et al. 2000; Charrier et al. 2003). CRMP-5 itself has been identified as a biomarker in neuroendocrine lung cancer (Meyronet et al. 2008), while CRMP-5 autoantibodies (also named CV2) are known markers for paraneoplastic neurological syndromes and associated cancers, for example, small-cell lung cancer and thymoma (Yu et al. 2001; Rogemond and Honnorat 2000; Werry et al. 2009). Furthermore, CRMP-5 was shown to be induced as a result of neurotoxicity (Berg et al. 2011) supporting its proposed role in neurogenesis. The expression pattern of the CRMP-5 gene overlaps with that of one or several CRMPs in different cells, compartments, and stages of the neuronal development (Ricard et al. 2001; Charrier et al. 2003; Veyrac et al. 2005). It is thus hypothesized that CRMP-5 forms hetero-oligomers with and modulates the function of other CRMP family members (Fukada et al. 2000).
As an intracellular component of the semaphorin 3A signal transduction pathway, CRMP-5 is involved in regulating filopodial dynamics and growth cone development by negatively regulating the sensitivity of the growth cone to semaphorin 3A, (Hotta et al. 2005). At a later stage of neuronal development, CRMP-5 is thought to regulate dendritic development by mediating brain-derived neurotrophic factor signaling in the CNS (Yamashita et al. 2011). CRMP-5 competes with CRMP-2 for interaction with tubulin dimers, thus impairing tubulin transport and subsequently arresting the tubulin polymerization and neurite outgrowth (Brot et al. 2010). CRMP-5 and CRMP-1 are also shown to have antagonistic effects, with CRMP-5 negatively regulating and CRMP-1 stimulating the proliferation of neural progenitors and transit amplifying neuroblasts (Veyrac et al. 2011). Moreover, CRMP-1 and CRMP-5 are thought to contribute synergistically to the regulation of the balance between functional integration and death of newborn neurons in the adult neurogenic areas, via pathways that exclude direct molecular interaction between these two proteins (Fukada et al. 2000; Su et al. 2007; Veyrac et al. 2011).
CRMPs preferably occur as hetero-tetramers in vivo, but form homo-tetramers when purified from bovine brain (Wang and Strittmatter 1997). The core region, which excludes the C-terminal approx 80 residues of CRMPs, is hereby sufficient for the oligomerization (Wang and Strittmatter 1997; Stenmark et al. 2007; Majava et al. 2008). Divalent cations such as Ca2+ and Mg2+ have been reported to promote homo-tetramer formation of CRMP-2 in solution (Majava et al. 2008), whereas in the crystalline state, murine CRMP-1 (15–490) as well as human CRMP-2 are homo-tetrameric even in absence of divalent cations (Deo et al. 2004; Stenmark et al. 2007). In solution, CRMP-5 was shown to form hetero-oligomers with CRMP-2, -3, and -4, but interestingly not with CRMP-1 (Fukada et al. 2000).
Compared with their closest homolog, the enzyme DHPase, the sequence of CRMPs, is extended by a positively charged C-terminal region, which is highly susceptible to proteolysis (Deo et al. 2004 and R.P. unpublished results). In vivo, the calcium-activated protease, calpain, specifically cleaves CRMPs within the C-terminal region to produce 55- to 58 kDa products (Jiang et al. 2007). These truncated proteins have various effects on axonal growth and sometimes even induce cell death after translocation into the nucleus (Jiang et al. 2007; Zhang et al. 2007; Taghian et al. 2012). However, little is known about the direct effects of the C-terminal truncation on structure and function of CRMPs. Knowledge of the structure and oligomerization properties of CRMP-5 in comparison with other members of the CRMP family is thus required to understand its unique role.
Here, we report the crystal structures of the full-length (WT) and a truncated version (ΔC) of human CRMP-5 at 2.2 and 1.7 Å resolution, respectively. For the former, only residues 1–491 are visible in the electron density map, but a possible location of the C-terminal region can be identified. Comparison of the active site of DHPase with the corresponding site of CRMP-5 reveals small, but crucial changes that explain the lack of amidohydrolase activity. As expected, the overall structure of CRMP-5 is very similar to that of CRMP-1 and CRMP-2. Nevertheless, the observed differences allow conclusions about why CRMP-5 preferentially forms hetero-oligomers with CRMP-2 rather than CRMP-1. Analysis of the oligomerization states of CRMP-1, -2, and -5 at varying protein concentrations and in absence or presence of divalent cations revealed significant differences, which, in addition to protein phosphorylation/dephosphorylation events may be crucial for regulation of CRMP function and interaction with other signaling and cytoskeleton proteins. Such a regulation via oligomerization has already been described for CRMP-2, which dissociates into monomers when interacting with tubulin and calmodulin (Fukata et al. 2002; Zhang et al. 2009).
Materials and methods
Cloning and expression
The regions coding for CRMP-1 (13–490; NP_001304.1), CRMP-2 (13–490; NP_001377.1), and CRMP-5 (1–483: ΔC, and 1–564: WT; NP_064519.2) were amplified by PCR from the cDNA fragments. All the clones were prepared following the same procedure. The primers contained an approx 20 bp overhang corresponding to the recombination site in the expression vector. Ligation-independent cloning was used to subclone the PCR products into the pNIC28-Bsa4 vector (Gräslund et al. 2008). The coding region is hereby extended by 23 additional residues at the N-terminus, which includes a hexahistidine tag and a tobacco etch virus (TEV) protease cleavage site preceding the protein sequence. The plasmids were transformed into E. coli Rosetta (DE3)pLysS competent cells. For expression, cells were grown using ZYP-5052 auto-induction-rich medium (Studier 2005) supplemented with corresponding antibiotics kanamycin (100 μg/mL) and chloramphenicol (34 μg/mL). The cells expressing the target protein were kept under constant agitation at 37°C for 2 h. Then the temperature was reduced to 22°C for further 22 h of growth, after which cells were harvested by centrifugation at 4°C and the resulting pellets were frozen at −20°C.
For lysis, the cell pellets were resuspended in buffer A [50 mM HEPES, 300 mM NaCl, 10 mM imidazole, 10% (v/v) glycerol, pH 7.5 (25°C)] supplemented with 1 mg/mL lysozyme, 5 μg/mL DNase, and one complete EDTA-free protease inhibitor cocktail tablet (Roche, AB, Stockholm, Sweden) per 30 mL of resuspended cells, and incubated for 30 min at 4°C. Cells were further lysed using sonication followed by centrifugation at 38 000 g for 60 min at 4°C. To the supernatant, 1 mL of Ni-NTA agarose beads (Qiagen, Ab, Sollentuna, Sweden) per 10 g of cell pellet was added. After incubation at 4°C for 40 min under constant agitation, the Ni-NTA agarose beads were collected by filtration and washed with 20 mL buffer A. CRMP protein was eluted by stepwise increasing the imidazole concentration from 50 to 1000 mM in buffer A. Fractions containing target protein were pooled and, after addition of TEV protease in a 1 : 100 molar ratio, dialyzed overnight against cleavage buffer [20 mM Tris, 150 mM NaCl, 5 mM dithiothreitol, pH 8.0 (25°C)] at 4°C. For purification of WT CRMP-5 dialysis and cleavage of the N-terminal tag was omitted. The proteins were concentrated and for further purification applied to a gel-filtration column (Superdex 200 from GE Healthcare, Uppsala, Sweden), which was pre-equilibrated with buffer B [50 mM HEPES, 300 mM NaCl, 10% (v/v) glycerol, pH 7.5 (25°C)]. The major peak was pooled and concentrated to approx 15 mg/mL. The proteins were either immediately used for crystallization or shock frozen and stored at −80°C until further use.
Crystallization, structure determination, and refinement
Crystallization screening was performed for both the CRMP-5 WT and ΔC proteins using a Phoenix robot (Dunn Labortechnik, Thelenberg, Germany) with the sitting-drop vapor-diffusion method in a sparse matrix screen. The crystallization drops were imaged regularly using Rock Imager (Formulatrix, Inc., Waltham, MA, USA). Initial hits were optimized by varying the concentration of the precipitant and the pH of the buffer. Crystals of both CRMP-5 WT and ΔC grew to full size within 1 day at 20°C under similar conditions. Optimized, well-diffracting crystals of CRMP-5 WT grew within 12 h incubation in 25% (w/v) polyethylene glycol (PEG) 1500, SPG (2:7:7 - succinic acid:sodium dihydrogen phosphate:glycine) pH 6.0. Protein at 15 mg/mL concentration was mixed in an equal ratio with precipitant. Similarly, CRMP-5 ΔC crystallized in 23% (v/v) PEG 400, MMT (1:2:2 - DL-malic acid:MES:Tris base) pH 5.0 using a protein concentration of 18 mg/mL. Cryoprotection for both crystal forms was achieved by increasing the respective PEG concentration to 30%. X-ray diffraction data were collected at 100 K at the European Synchrotron Radiation Facility, Grenoble, France. The data sets were processed with MOSFLM and scaled using SCALA (Evans 2006; Battye et al. 2011). The structure of CRMP-5 ΔC was solved by molecular replacement using the program Phaser (McCoy et al. 2007) with a monomer of CRMP-2 (2GSE; (Stenmark et al. 2007)) as the search model. The initial solution had an R-factor of 49.0%. Model building was performed first by automated methods using Buccaneer (Cowtan 2006) and then manually using WinCoot (Emsley et al. 2010). CRMP-5 WT was solved by molecular replacement using the ΔC version of the structure as a model and the program Phaser. Refinement was performed using REFMAC5 (Murshudov et al. 2011). To characterize putative Zn-bound CRMP-5 structures, CRMP-5 ΔC crystals were soaked in reservoir solution supplemented with 5 mM of ZnCl2. Data sets were collected from crystals soaked for a short (5 min) as well as a longer (120 min) time period, both to 2.9 Å resolution. Atomic coordinates and structure factors of CRMP-5 WT, ΔC, and Zn-bound CRMP-5 ΔC have been deposited with the RCSB Protein Data Bank, with accession codes 4B90, 4B91, and 4B92, respectively.
The protein interfaces and interaction surfaces were calculated and analyzed using the PISA web server (Krissinel and Henrick 2007). The volume of the active-site pocket was calculated after removal of solvent molecules using the Pocket-Finder web server (Laurie and Jackson 2005). Hetero-dimers were modeled by superposition of one CRMP monomer onto another CRMP homo-tetramer or -dimer. These hetero-dimers were then submitted to the YASARA server for energy minimization (Krieger et al. 2009). Structural analysis was performed and figures prepared using PyMOL (http://www.pymol.org/).
Analytical gel filtration
Oligomerization analyses were performed with CRMP-1, -2, and -5 using a Superdex 200 5/150 GL column (GE Healthcare) connected to an ÄKTA purifier system with automatic sample application. All analyzed proteins were freshly purified and diluted in the respective buffer before sample injection. For the concentration-dependent oligomerization analysis, buffer HN (10 mM HEPES, 100 mM NaCl, pH 7.5) was chosen. Effects of divalent cations were analyzed with the buffers HNM (10 mM HEPES, 50 mM NaCl, 20 mM MgCl2, pH 7.5) and HNC (10 mM HEPES, 50 mM NaCl, 20 mM CaCl2, pH 7.5). Buffer pH was always adjusted at 7°C. All analyses were performed at least twice in independent experiments at 4°C. The molecular weight of the proteins was estimated using the marker proteins ferritin (400 kDa), catalase (232 kDa), albumin (67 kDa), ovalbumin (43 kDa), and chymotrypsinogen (25 kDa).
Differential scanning fluorimetry
For the differential scanning fluorimetry (DSF) experiments, 5 μg of protein were added to the tested buffers (total volume: 20 μL). This solution was incubated at 20°C for 15 min before adding 5 μL of 200-times diluted SYPRO Orange stock (5000 ×) (Sigma-Aldrich Sweden AB, Stockholm, Sweden). The experiments were performed in a 96-well thin-wall PCR plate (Bio-Rad Laboratories AB, Solna, Sweden) sealed using Optical-Quality Sealing Tape (Bio-Rad). An iQ5 Real-time PCR Detection System (Bio-Rad) calibrated with external well factor solution was used to monitor the changes in the fluorescence intensity of the fluorophore. The wavelengths for excitation and emission were 490 and 575 nm, respectively. The temperature of the samples was changed from 20 to 95°C at a heating rate of 1°C/min and the fluorescence was recorded every 0.2°C. The melting points of the proteins were determined using a Boltzmann model as described previously (Ericsson et al. 2006).
Enzyme activity assay
Hydrolysis of the dihydrouracil ring was measured spectrophotometrically by observing the change in the absorbance at 225 nm (Gojkovic et al. 2000). A UV-3000 spectrophotometer (Hitachi, High Technologies America, Inc., Dallas, TX, USA) was utilized (optical pathlength 10 mm). Activity assays were performed at 30°C in 0.1 M Tris buffer at pH 8.0. The reaction was started by addition of substrate dihydrouracil to a final concentration of 250 μM. Human DHPase (hDHPase) and human CRMP-1 were used as positive and negative control, respectively. The protein concentration was approx 1 μM for all proteins probed. For chemical rescue experiments, the assay was performed in presence of 50 mM carboxylic acids (acetic acid, propionic acid, or butyric acid) as well as in presence and absence of 50 μM ZnCl2. Protein concentrations were determined based on the absorbance at 280 nm measured using a Nanodrop spectrophotometer and the extinction coefficients estimated using the ExPASy ProtParam tool (Wilkins et al. 1999).
Structure elucidation and quality of the structural models
In this study, we investigated the tertiary and quaternary structures of CRMP-5, both in crystalline state and in solution. Two different versions of CRMP-5, WT (1–564), and ΔC (1-483), were heterologously produced in good yield in E. coli. The N-terminal histidine tag of the CRMP-5 ΔC version was efficiently removed by TEV protease. For the WT version of CRMP-5, the N-terminal histidine tag plus additional linker residues (total 23 extra residues) were not cleaved off because of the flexible C-terminal region present within the protein (see below). Diffraction-quality crystals for both versions of CRMP-5 belonging to space group P41212 were obtained by sitting-drop vapor diffusion. X-ray diffraction data were collected to a resolution of 1.7 Å from ΔC crystals and 2.2 Å from the WT crystals. The CRMP-5 ΔC structure was determined by molecular replacement, using a monomer of the CRMP-2 (pdb code: 2GSE) crystal structure (Stenmark et al. 2007) as the search model. Later, the ΔC version of CRMP-5 was used as a model to solve the WT CRMP-5 structure. The final models for CRMP-5 WT and ΔC both contain two polypeptide chains in the asymmetric unit that form a dimer. For CRMP-5 ΔC residues 1–7, 482, and 483 of chain A, and 1–7 and 483 of chain B are missing in the final model, which shows an R-factor of 17.0% and an Rfree of 20.2%. Of the amino acid residues, 98.1% and 1.7% are in the favored and allowed regions of the Ramachandran plot (Lovell et al. 2003), respectively. Two residues (Gly 350 in both chains) are found just outside the allowed region, but are well resolved in the electron density. The structure is of very good quality as indicated by a MolProbity score of 1.25 [98th percentile (n = 9248, 1.70Å ± 0.25Å)] (Davis et al. 2007).
For WT CRMP-5, residues 1–6, 492–564, and 1–7, 493–564 of chains A and B, respectively, as well as all 23 residues added to the N-termini by the cloning procedure could not be traced. This includes a large portion of the C-terminal sequence from both molecules in the asymmetric unit, for which no electron density was observed. The preceding residues are pointing toward a solvent channel in the crystal, and it is thus likely that the C-terminal regions are unstructured (see below). The C-terminus of CRMPs is highly susceptible to proteolytic cleavage (Deo et al. 2004 and R.P. unpublished results). To determine whether potential proteolytic degradation products were crystallized instead of CRMP-5 WT, crystals were harvested and washed thoroughly before analysis by sodium dodecyl sulfate gel electrophoresis. No degradation products were detected. The solution of the drop from which the crystals were harvested was also analyzed and revealed no degradation either (Figure S2). As the crystals appeared within a few hours after the crystallization set up, it is less likely that proteolytic degradation could have occurred. For the remainder of the molecule the amino acid residues are well defined, with 98.6% in the favored and 1.2% in the allowed regions of the Ramachandran plot (Lovell et al. 2003). As for the ΔC structure, Gly 350 in both chains is the only outlier. The CRMP-5 WT structure was refined to a final R-factor and Rfree of 18.0% and 22.1%, respectively. The excellent quality of the structural model is reflected by a MolProbity score of 1.32 [100th percentile (n = 10167, 2.20Å ± 0.25Å)] (Davis et al. 2007). The data collection and refinement statistics are described in Table 1.
Table 1. Data collection and refinement statistics
CRMP-5 WT (1–564)
Values in parentheses are for the highest resolution shell.
, where I(hkl) is the intensity of reflection hkl and 〈I(hkl)〉 is the average intensity over all equivalent reflections.
. Rfree was calculated for a test set of reflections (5%) omitted from the refinement.
487 698 (72 025)
775 447 (62 235)
69 984 (10 044)
110 493 (14 077)
Number of reflections
49 886 (3 609)
104 852 (6 462)
Wilson B (Å2)
R.m.s.d. from ideal geometry
Overall structure of CRMP-5
CRMP-5 belongs to the amidohydrolase superfamily, a functionally rather diverse group of enzymes. Within this group, it most closely resembles the enzyme DHPase. Based on sequence similarity, the CRMP family is assigned to the Amidohydro_1 subfamily as a metal-dependent hydrolase (Accession No.: PF01979; Pfam; (Punta et al. 2012)), even though residues crucial for the coordination of the metal ions are lacking (see below). The monomer of CRMP-5 consists of two domains, a larger triosephosphate isomerase-like barrel domain, the hallmark of the amidohydrolase family, and a shorter β-sandwich domain consisting of N- and C-terminal residues, which has no known structural similarities (Fig. 1). Both CRMP-5 WT and ΔC versions crystallized in a tetragonal crystal system with two molecules forming a dimer in the asymmetric unit. The crystallographically independent monomers of CRMP-5 WT and ΔC are very similar to one another, showing an r.m.s. deviation of 0.2–0.4 Å (based on Cα positions). Superposition of the CRMP-5 monomer with CRMP-1, CRMP-2, and DHPase also resulted in low r.m.s.d. values (Table 2), indicating strong conservation of the core structure. The CRMP-5 WT structural model contains only 10 additional residues as compared with ΔC (residues 483-492). These residues form an extended loop protruding from the protomer toward the other monomer in the dimer (Fig. 2). A salt bridge is formed between the guanidinium group of R489 from the loops and the carboxyl group of E214 from the dimer mates, which anchors the extended loops at the opposite peripheries of the dimer interface. Similar anchoring of C-terminal stretches by an Arg-Glu pair has been observed in human (PDB ID: 2VR2) and a bacterial DHPase (Sinorhizobium meliloti) (Martínez-Rodríguez et al. 2010). However, the amino acid sequences and thus the C-terminal tails of DHPases are significantly shorter than those of CRMPs. Besides the salt bridge, there are almost no other interactions observed for residues 483-491. In fact, some of the side chains could not be modeled unambiguously because of the lack of electron density that can be attributed to its flexibility.
Table 2. Structural differences between CRMPs and DHP monomer. Observed root mean square deviation (RMSD) in Å for pairwise superposition of crystal structures
With exception of the C-terminus, the ΔC and WT CRMP-5 structures are very similar to one another (Fig. 1 and Table 2). As it was determined to higher resolution, the CRMP-5 ΔC structure was used for the following analyses unless otherwise stated.
The dimerization interface between the two molecules in the asymmetric unit buries a solvent-accessible surface area of 1090 Å2 per monomer, and includes 10 hydrogen bonds and salt bridges (Fig. 2b). This interface is henceforward called ‘arm-lever’. A second dimer interface, referred to as ‘arm-arm’, is formed between molecules A and A' or B and B' related by crystallographic symmetry. It buries an average solvent-accessible surface area of 1115 Å2 per monomer (Fig. 2c), includes 16–18 hydrogen bonds and salt bridges and slightly more hydrophobic interactions than the arm-lever interface (Table 3). CRMP-5 uses both arm–lever and arm–arm dimer interfaces to form a homo-tetramer in the crystal (Fig. 2a). Similar tetramer assemblies have been observed in the crystal structures of CRMP-1 and -2, as well as in DHPase (Table 3) [Deo et al. 2004; Lohkamp et al. 2006; Stenmark et al. 2007, and hDHPase (PDB ID: 2VR2)].
Table 3. Properties of CRMP and DHP interfaces
ΔiG kcal/M (No. of HB, SB)
Buried surface area Å2
The values were obtained from the crystal structures by PISA analyses (Krissinel and Henrick 2007).
HB, hydrogen bond, SB, salt bridge.
2.4 (18, 3)
−12.7 (8, 4)
−9.6 (10, 5)/−9.5 (7, 1)
−0.2 (16, 5)/−3.2 (14, 2)
−9.3 (14, 2)/−9.1 (18, 2)
−13.3 (14, 10)
−10.3 (14, 4)/−9.5 (18, 2)
−4.4 (8, 3)
0.9 (24, 10)
−26.8 (10, 8)
CRMPs preferentially form hetero-tetramers rather than homo-tetramers in vivo (Wang and Strittmatter 1997; Fukada et al. 2000), but the molecular basis for this preference and its potential functional implications have not been identified yet. Based on sequence alignments and the crystal structures of CRMP-1 and CRMP-2, several residues were suggested to be crucial for the formation of hetero-tetramers (Deo et al. 2004; Stenmark et al. 2007). Analysis of the crystal structures of CRMP-5 agrees largely with these hypotheses and provides new insights into hetero-tetramer formation.
Although CRMPs are structurally very similar to one another, CRMP-5 prefers to form hetero-tetramers with all but CRMP-1, which was shown by yeast two hybrid assay as well as coimmunoprecipitation analysis (Wang and Strittmatter 1997; Fukada et al. 2000). We have confirmed these observations by heterologous coexpression in E. coli and copurification experiments (Supplementary material and Figure S3). On the basis of the CRMP-1 structure, Deo et al. proposed three pairs of hydrogen-bonding residues to be crucial for weakening the CRMP-1 affinity for CRMP-5 (Deo et al. 2004): Q379*(R372)-R481*(K474); E208*(G201)-R245*(R238) and K269*(M262)-Y316*(Y309) (asterisks mark CRMP-1 residues, the corresponding CRMP-5 residues are given in brackets). For example, it was proposed that the replacement of Q379* by an arginine in CRMP-5 would lead to unfavorable contacts with the corresponding hydrogen-bonding partner. However, the neighborhood of the two positively charged residues R372–K474 does not cause any steric or electrostatic clashes in the CRMP-5 homo-oligomer, because R372 is facing away from the interface and forms an intramolecular salt bridge. The same may occur upon hetero-tetramer formation, but even if the side chain of R372 would be involved in interface interactions, it would equally weaken the hetero-oligomer association of CRMP-5 with all other CRMPs. Therefore, discrimination against CRMP-1 as binding partner in hetero-oligomer assembly is probably not solely based on the above-mentioned residue pairs. Comparative analysis of the CRMP-5 crystal structure reveals additional structural differences that may be crucial in this respect. At the arm–lever interface, the side chains of N237 and K265 both form hydrogen bonds with E223, a residue belonging to the acidic patch of α-helix 7′ that is completely conserved among the CRMPs. K265 located in the loop connecting α-helix 8 and β-strand 16 is replaced by proline in CRMP-1, and by threonine in CRMP-2, and hence the hydrogen bond to E223 would be abolished in both types of hetero-oligomer. However, N237 belonging to α-helix 7 is present in CRMP-2, which thus can form at least one of the interface hydrogen bonds, while it is replaced by a glycine in CRMP-1 (Fig. 3a). Furthermore, in the CRMP-5 arm–arm interface, the amino group of K473 located in α-helix 16 donates three hydrogen bonds to residues E370, N371, and T23, respectively. As previously observed by Deo et al. (Deo et al. 2004), K473 is replaced by glutamine in CRMP-1, but conserved in all other CRMPs. Although the glutamine is involved in a similar hydrogen-bonding network in CRMP-1, these interactions are less favored (Fig 3b). Thus both the arm–arm and arm–lever interfaces are weakened between CRMP-5 and CRMP-1 as compared with the interaction with other CRMPs.
Both the arm–lever and arm–arm interfaces bury substantial solvent-exposed surface areas within the CRMP-5 homo-oligomer, and include several hydrogen bonds and salt bridges. Nevertheless, several residues in and around these interfaces have no direct interaction partner in the homo-oligomer, but may have one in the hetero-oligomer, which could explain the corresponding preference in oligomer formation. For example, in the arm–lever interface R238 from α-helix 7 lacks an electrostatic interaction partner. The nearby G201 is replaced with a glutamate in CRMPs 1–4, which could form a salt bridge with R238. Also, the replacement of S252 (α-helix 8) located close to Q263 by lysine or arginine in CRMPs 1-4 may enable an additional hydrogen-bonding interaction. As described above, K474 (α-helix 16) located in close vicinity to R372 does not contribute to the arm–arm interface in the CRMP-5 homo-oligomer, while the replacement of R372 by glutamine or glutamate in CRMPs 1-4 can lead to the formation of a salt bridge as seen in the CRMP-1 homo-tetramer. T318 from the loop connecting α-helix 11 and β-strand 17 forms a long hydrogen bond with M262; this bond may be shortened and hence strengthened on replacement of T318 by lysine or arginine in the other CRMPs. Finally, Q477 (α-helix 16) may engage in a favorable interaction with an arginine residue (H433 in CRMP-5) of CRMPs 2-4. There is also an overall increase in the number of hydrophobic interface residues contributing to hetero- compared with homo-oligomerization (Table 3).
Within the CRMP family, CRMP-5 is phylogenetically closest to DHPases, the enzymatically active homologs. Nevertheless, three of six residues coordinating the two catalytic zinc ions in hDHPase, namely K159* (Q158), H67* (S66), H69* (H68), H192* (H191), H248* (N247), and D326* (D325) (asterisks mark hDHPase residues, the corresponding CRMP-5 residues are given in brackets), are not conserved in CRMP-5. This includes the carboxylated lysine bridging the two ions of the di-zinc center in DHPases, which is replaced by glutamine in CRMP-5. After soaking CRMP-5 crystals in a ZnCl2-containing solution, two zinc ions were bound to each monomer as indicated by the observed electron density and anomalous signal. However, both ions were bound outside the active site at the surface of the CRMP-5 molecule (Figure S4), while no (anomalous) difference electron density was observed in the vicinity of the active-site region. Like CRMP-1 and -2, CRMP-5 still contains a recognizable active-site-like cavity, which is with a pocket volume of 277 Å3 comparable in size to the DHPase active site (251 Å3), but considerably smaller than the pocket found in CRMP-1 (540 Å 3) and CRMP-2 (580 Å3) (Laurie and Jackson 2005). The decrease in pocket size is primarily caused by R327 which is protruding into the pocket only in CRMP-5. However, if the side chain would reorient for involvement in hetero-oligomerization (see above), the volumes become comparable again. In addition, the guanidinium group of R327 partially occupies the binding site of the dihydropyrimidine substrates of DHPases, forming a hydrogen bond with the backbone carbonyl oxygen of residue 346 involved in dihydrouracil binding (Lohkamp et al. 2006).
After expression and purification as described, recombinant CRMP-5 did not show any amidohydrolase activity under standard assay conditions (Gojkovic et al. 2000). To exclude that the inactivity was caused by the absence of the zinc ion cofactors, assays were reperformed in presence of Zn2+. Again, no amidohydrolase activity was detected. As CRMP-5 is missing the lysine residue that is carboxylated and bridges the catalytic zinc ions in DHPase, addition of short-chain carboxylic acids for chemical rescue was probed in the enzyme assay. While successful in the analysis of hydantoinase mutants, chemical rescue did not result in any detectable amidohydrolase activity of CRMP-5.
Oligomeric state in solution
The oligomeric states of CRMP-1, -2, and -5 in solution were probed by analytical gel filtration. At a protein concentration of 1.0 mg/mL in HN buffer, CRMP-5 and CRMP-1 show a single peak in the elution profile corresponding to a monomer, whereas CRMP-2 elutes in two peaks representing tetramers and monomers. However, all three proteins crystallized in homo-tetrameric form (Deo et al. 2004; Stenmark et al. 2007). As physiological protein concentrations are usually low compared with those used for crystallization, different concentrations ranging from 0.25 mg/mL to 5.0 mg/mL of CRMP-1, -2, and -5 were subjected to oligomerization state analyses in HN buffer. CRMP-1 shows a single peak corresponding to a monomer at 0.25 and 0.5 mg/mL protein concentration with a symmetric Gaussian peak shape, whereas at 1.0 mg/mL the peak skews toward the higher molecular weight species. At 2.5 and 5.0 mg/mL, a shoulder indicating a second peak becomes evident that corresponds to a molecular weight between that of dimer and trimer (Fig. 4). CRMP-2 shows two peaks corresponding to a tetramer and monomer throughout the concentration series investigated. At 0.5 mg/mL, it displays an equal proportion of each species (Fig. 4). This ratio is shifted toward the tetrameric species at higher concentrations and monomers at lower concentrations. The total amount of monomer appears to remain constant up to the highest tested protein concentration (5.0 mg/mL). CRMP-5 showed only one peak with Gaussian shape throughout the concentration series, however, the calculated molecular weight shifted from monomer at 0.25 mg/mL to dimer at 5.0 mg/mL concentration.
Effect of divalent cations on CRMPs
CRMP-2 homo-tetramer formation was shown to be promoted by divalent cations such as Ca2+ and Mg2+ (Majava et al. 2008). Therefore, we investigated the effect of divalent cations on CRMP-5 and CRMP-1 as compared to that on CRMP-2 by analytical gel filtration and DSF. First, we replicated the analytical gel filtration experiments for CRMP-2 and could also show promotion of homo-tetramer formation by Mg2+ and Ca2+ ions (Fig. 5a), although the observed effect was not as pronounced as previously described. Both divalent cations had little effect on CRMP-1 and CRMP-5. In fact, for both the proteins, the peak profile indicates a slight shift toward the monomeric form instead (Fig. 5a). The melting point of CRMP-2, as determined by DSF, was increased by 1 and 2°C in the presence of Ca2+ and Mg2+ ions, respectively, whereas for both CRMP-1 and CRMP-5 the melting points were decreased. CRMP-1 presented a 2°C shift for both divalent cations and CRMP-5 a shift of 1 and 2°C for Ca2+ and Mg2+ ions, respectively (Fig. 5b).
We have determined the crystal structures of WT and a C-terminally truncated form of human CRMP-5, the latest identified member of the CRMP protein family. CRMP-5 crystallizes as a homo-tetramer similar to that previously described for CRMP-1 and CRMP-2 (Deo et al. 2004; Stenmark et al. 2007). Analysis using gel filtration revealed that, in contrast to CRMP-2 (Wang and Strittmatter 1997; Stenmark et al. 2007; Majava et al. 2008), CRMP-5 and CRMP-1 form relatively weakly associated homo-tetramers in solution (Fig. 4). However, all CRMPs were shown to have a preference for formation of hetero-tetramers over homo-tetramers in vivo, with the N-terminal core region (8-134 and 281-435) being essential and sufficient for the assembly of this oligomerization state (Wang and Strittmatter 1997; Fukada et al. 2000). On the basis of the existing and presented structural data, we identified residues at both the arm–arm and arm–lever dimer interfaces, which in the homo-oligomer are less ideally positioned or interacting, and may thus play a role in the discrimination between homo-and hetero-oligomerization states. Furthermore, we also identified an interaction hotspot in both the arm–arm and arm–lever dimer interfaces that could explain why CRMP-5 can form hetero-tetramers with all CRMPs except CRMP-1 (Fukada et al. 2000). Both are characterized by changes of positively charged (K265 and K474) to uncharged residues (Pro and Gln, respectively). In addition, we showed that CRMP-5 and CRMP-1 have a lower propensity to form homo-tetramers than CRMP-2, which may indicate that formation of hetero-tetramers is not favored either. Based on the available structural data, it is not possible to exclude that more than two different CRMPs constitute one hetero-tetramer, which would allow CRMP-5 and CRMP-1 to be present in one tetramer as diametrically opposed monomers that do not interact. Similarly, it was proposed earlier that the presence of multiple combinations of CRMP hetero-oligomers in a neuron is necessary for a functional neuronal network (Fukada et al. 2000; Deo et al. 2004). Nonetheless, the precise manner in which CRMP hetero-tetramers are formed, that is, whether by using arm–arm or arm–lever homo-dimers, is still unclear. Stenmark et al. proposed a CRMP-1–CRMP-2 hetero-tetramer model via the arm–arm interface using arm–lever-interfaced homo-dimers owing to the higher degree of sequence conservation (Stenmark et al. 2007). However, the arm–lever interface in CRMP-1–CRMP-2 shows a larger contribution from hydrophobic residues compared with the arm–arm interface. Unlike the CRMP-1–CRMP-2 hetero-tetramer analyses, other hetero-oligomer analyses do not show any obvious discrimination between the two interfaces. Therefore, it cannot be ruled out that the arm–lever interface is also used in the hetero-tetramer assembly, especially as there are residue pairs in both interfaces that seem to favor the hetero-tetramer. The functional importance of CRMP homo- and/or hetero-oligomerization is still unclear. One possibility would be that hetero-tetramer formation would bring different C-terminal regions into close proximity. These regions are least conserved (see supporting Table 2) and may allow CRMPs to interact with the same or different partners in a precise and fine-tuned manner (see below). Thus far, two in vivo studies reported a modulation of axon growth in correlation to the oligomerization of CRMPs. In one study, the e204 mutation in the unc-33 gene from C. elegans (residue D58 is mutated to Asn) causes paralysis resulting from defective axon growth (Brenner 1974). The D58N mutation hampers the oligomerization of UNC-33 and its binding with UNC-14 (RUN domain protein) and Kinesin Light Chain-2 (Tsuboi et al. 2005). D58 is conserved between CRMPs and UNC-33 (D64 in CRMP-5; D71 in CRMPs 1-4). In CRMP-2, the D71N mutant does not interfere with the oligomerization of the protein, but specifically with the interaction to Sra-1 (Rac-1 associated protein 1). Nevertheless, this also results in axon growth defects (Kawano et al. 2005). The corresponding residue D64 is located in the core of CRMP-5, in fact exactly at the center of the mass of monomer molecule, and hence distant to both oligomerization interfaces (Figure S5). The effects of the mutation seen in UNC-33 and CRMP-2 may not be based on a direct interference with the oligomerization, but an indirect effect caused by the destabilization of the monomer core. In the second study, alanine substitution mutants of residues 367-368 and 487-489 of mouse CRMP-1 produced functionally inactive proteins, even though both retained their ability to interact with plexin A1. The residues within the mutated region appear to be crucial as they are conserved among CRMPs. They are located in the vicinity of the arm–arm interface and might destabilize the functional tetramer (Deo et al. 2004) (Figure S5).
The divalent cations Ca2+ and Mg2+ were shown to promote CRMP-2 oligomerization in solution (Majava et al. 2008) and to bind to CRMP-2 in crystals (Stenmark et al. 2007; Majava et al. 2008). The Ca2+-and Mg2+-binding sites identified in the corresponding CRMP-2 crystal structures overlap and are found on the tetramer surface distant from dimer interfaces. It is therefore not directly evident how these cations promote oligomerization. Also, both cations appear to be highly coordinated by water molecules that mediate the contact to the protein. In particular, the Ca2+ ion is bound directly to the main chain carbonyl oxygen of residue 349, and indirectly via water molecules to the side chains of E353 or Q81 of CRMP-2. While these latter residues are conserved in CRMP-1, in CRMP-5, they are replaced by H346 and M71, respectively, which do not possess the size or charge to interact similarly with these cations. Later, Majava et al. reported two additional Ca2+-binding sites in CRMP-2, but the data have not yet been deposited in the protein data bank (PDB) (Majava et al. 2008). Although these sites are not directly involved in the tetramer interface formation, the site near the side chain of Q245 belonging to α-helix 7 possibly stabilizes the CRMP-2 monomer and thereby assists in the tetramer formation. Q245 is replaced by arginine in CRMP-5 and CRMP-1, making it an unlikely binding site in these two proteins. In line with this structural evidence, CRMP-5 and CRMP-1 oligomerization was not promoted by Ca2+ or Mg2+in vitro (Fig. 5), and the thermal stability of both proteins was decreased rather than increased in the presence of these ions (Fig. 4). In vivo, both cations are important in cellular signaling and development, and especially Ca2+ has an important role in axon guidance. Several proteins and functions connect calcium ions directly to CRMPs. For example, CRMP-2 interacts with calcium channels and calmodulin, and calpain is activated by the ion (Zhang et al. 2007, 2009; Brittain et al. 2009). However, the in vivo relevance and the functional role of divalent cation binding to CRMPs are still to be investigated.
Metal binding and enzymatic activity of CRMP-5
As CRMP-5 is the closest relative to the enzyme DHPase within the CRMP family, it could be expected to have some residual DHPase activity. However, several residues in the active site, including the carboxylated lysine residue that coordinates the catalytic metal center in DHPase, are exchanged in CRMP-5, which effectively abolishes metal binding as shown by mutational analyses of hydantoinases (Huang et al. 2009). This was confirmed experimentally and consequently no residual amidohydrolase activity of CRMP-5 is detectable. Chemical rescue by small carboxylic acids that could take over the role of the carboxylated lysine in zinc binding is probably hindered sterically and electrostatically by the presence of the glutamine side chain at the equivalent position near the binuclear metal center site. Substrate binding is most likely affected as well, even though no side-chain-specific substrate interactions are made in DHPase (Lohkamp et al. 2006). In CRMP-5, R327 is protruding into the ‘active site’ cavity, partially occupying the substrate-binding site. In addition, its positive charge may interfere with zinc ion binding. It thus appears that evolution to a non-enzymatic entity with alternative functions in neuronal development is already completed in CRMP-5, and that it does not represent a ‘missing link’ between the enzymatic and non-enzymatic forms. The likelihood that CRMPs show enzymatic activity toward yet unidentified other substrates is low as this would require the evolution of a different, metal-independent catalytic machinery.
Whether DHPases can substitute CRMP function, or has roles distinct from those of CRMPs, in neuronal development remains to be seen. Interestingly, deficiency in DHPase activity is associated with neurological abnormalities (Henderson et al. 1993), although the mechanisms involved may not be connected to CRMPs.
Structure and function of the C-terminal domain of CRMP-5
Similar to previous studies, we were not able to resolve the full-length structure of a CRMP. Although we crystallized the full-length CRMP-5 protein, only residues up to 492 were visible in the experimental electron density map. The missing electron density for the remainder of the protein may be attributed to its unstructured character, in other words flexibility and lack of permanent secondary structure. FoldIndex predicts the C-terminal region to be intrinsically unfolded (Prilusky et al. 2005). Furthermore, while the theoretical isoelectric point of CRMP-5 WT is 6.7, this value increases to 10.14 for the C-terminal region alone (last 84 amino acids), rendering it highly basic. For the crystal structures of CRMP-1 and -2, C-terminal residues were not even included in the structure determination because of their proteolytic susceptibility, and thus no structural information is available (Deo et al. 2004; Stenmark et al. 2007). Interestingly, the resolved CRMP-5 residues in the presented structures correspond almost exactly to two putative calpain-cleaved products, indicating their biological relevance (Taghian et al. 2012). Based on the CRMP-5 WT structure, it can be postulated that the C-terminus is likely to extend away from its own protomer. Residues 483-492 do not contribute much to the arm–lever dimer interface, with only one salt bridge being formed by R489 from the end of the extended loop. Mutation of this residue to alanine in a bacterial DHPase was shown to have no effect on the tetramer assembly (Martínez-Rodríguez et al. 2010). The remainder of the C-terminal region of CRMPs harbors many functionally important residues such as the phosphorylation sites crucial for regulating CRMPs' interactions with various binding partners, and a KKEX motif (526-529), which is commonly present in type I microtubule-associated proteins (Noble et al. 1989). Tubulin was shown to bind near the C-terminus of CRMP-5, specifically to the region 475-522 (Brot et al. 2010). As most, if not all, of the phosphorylation sites in CRMP-5 are found in the C-terminal region, it is likely that tubulin binding to CRMP-5 is regulated by phosphorylation. This is, for example, observed in the protein stathmin, which also interacts with tubulin dimers in a phosphorylation-dependent manner (Charbaut et al. 2001) and shows a low sequence similarity to CRMP-5. However, it appears that the C-terminal domain is not directly involved in binding to tubulin as surface plasmon resonance experiments indicated a low μM affinity of CRMP-5 ΔC for tubulin, which is comparable with that of the WT protein and other CRMPs (R.P. unpublished result, Fukata et al. 2002). Interaction with tubulin or other proteins at the C-terminus is unlikely to affect the structure of the core protein, thereby changing binding properties and function. However, it is not possible to rule out that the C-terminus interferes with binding epitopes on the core of the CRMP, thus modulating interaction of CRMP with other proteins.
CRMPs are regulatory phosphoproteins abundantly present in the early stages of neuronal development. Their spatiotemporal expression pattern varies between the members of the family. As a hub protein in a complex network of protein–protein interactions, CRMPs have to be highly regulated. This appears to be mainly achieved by phosphorylation; however, there is growing evidence that oligomerization may play a role as well. The formation of homo- and/or hetero-oligomers is imperative for CRMPs function in vivo, but their nature of composition during the various stages of neuronal development remains elusive. Availability of structural details of CRMP-1, -2, and -5 at atomic level will help to elucidate the in vivo oligomerization effect on CRMP complexes with other proteins by assisting a targeted mutagenesis approach with the aim to stabilize or destabilize specific homo- or hetero-tetramers.
We thank Dr. Doreen Dobritzsch and Magnus Claesson, MSB, Karolinska Institutet, for help with data collection. We also would like to thank Dr. Martin Hammarström, SGC, Karolinska Institutet, for providing cDNA as well as CRMP-2 ΔC and hDHP expression clones. We acknowledge access to synchrotron radiation at the ESRF (Grenoble, France) and the MAX-lab (Lund, Sweden) and thank the beamline staff for helpful support. This work was supported by Swedish Research Council, Åke Wiberg Stiftelse and KI fonder.