The N-terminal domain of the myelin enzyme 2′,3′-cyclic nucleotide 3′-phosphodiesterase: direct molecular interaction with the calcium sensor calmodulin



2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase) is a quantitatively major enzyme in myelin, where it localizes to the non-compact regions and is bound to the membrane surface. Although its catalytic activity in vitro has been characterized, the physiological function and in vivo substrate of CNPase remain unknown. Especially the N-terminal domain has been poorly characterized; previously, we have shown it is involved in CNPase dimerization and RNA binding. Here, we show that purified CNPase binds to the calcium sensor protein calmodulin (CaM) in a calcium-dependent manner; the binding site is in the N-terminal domain of CNPase. CaM does not affect the phosphodiesterase activity of CNPase in vitro, nor does it influence polyadenylic acid binding. The colocalization of CNPase and CaM during Schwann cell myelination in culture was observed, and CaM antagonists induced the colocalization of CNPase with microtubules in differentiated CG-4 oligodendrocytes. An analysis of post-translational modifications of CNPase from rat brain revealed the presence of two novel phosphorylation sites on Tyr110 and Ser169 within the N-terminal domain. The results indicate a role for the N-terminal domain of CNPase in mediating multiple molecular interactions and provide a starting point for detailed structure-function studies on CNPase and its N-terminal domain.

Abbreviations used

bovine serum albumin




2′,3′-cyclic nucleotide 3′-phosphodiesterase


differentiated CG-4 cells


dorsal root ganglion




isothermal titration calorimetry


myelin basic protein


phosphate-buffered saline


sodium dodecyl sulfate–polyacrylamide gel electrophoresis


trifluoperazine dihydrochloride

Myelin is a highly specialized structure in the vertebrate nervous system, formed by the differentiated plasma membrane of a glial cell. The myelin membrane wraps itself tightly around the axon, forming a compactly packed, multilayered proteolipid complex with a very low content of aqueous solvent. The myelin membrane composition is unique, containing over 70% lipids, of which nearly 30% is cholesterol (Norton and Poduslo 1973). The myelin proteome is also unusual in that it contains a handful of proteins that are often specific to myelin and present in high local concentrations (de Monasterio-Schrader et al. 2012). The protein composition of myelin differs between the central and peripheral nervous systems (CNS and PNS, respectively), and different regions of the myelin sheath carry specific protein components. Despite the fact that many of the major myelin proteins were first described already in the 1960s and early 1970s, as a result of their high concentration in myelinated nervous tissues, little is still known about the structure–function relationships in myelin proteins and their complexes. Such information will be crucial to understanding the function of myelin proteins inside the myelin membrane multilayer.

2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase) is an enzyme with a well-known catalytic activity but an unknown function, despite its initial characterization already 50 years ago (Drummond et al. 1962). It is present at high local concentrations in the non-compact regions of myelin (Trapp et al. 1988). CNPase-deficient mice develop brain axonal swellings and motor deficits with age (Lappe-Siefke et al. 2003). Recent data have linked CNPase to catatonia-depression syndrome in both mice and humans (Hagemeyer et al. 2012).

CNPase contains two domains, of which the C-terminal phosphodiesterase domain has been well characterized. Crystal and solution structures for this catalytic domain have been solved, and mutagenesis has highlighted residues important for the CNPase catalytic activity (Lee et al. 2001; Kozlov et al. 2003; Sakamoto et al. 2005). We also recently solved the first crystal structures of the mouse CNPase phosphodiesterase domain complexed with nucleotide ligands (Myllykoski et al. 2012). CNPase has been suggested to play roles in RNA metabolism within myelin; the true substrates of CNPase in vivo could be RNA molecules, instead of the nucleotide substrates used in enzyme assays (Gravel et al. 2009), or RNA degradation products formed during RNA turnover (Verrier et al. 2012). However, details of CNPase interaction with other macromolecules, including RNA, are still lacking. Our recent results indicate that the CNPase N-terminal domain plays major roles in CNPase dimerization and RNA binding (Myllykoski et al. 2012). CNPase also binds cytoskeletal proteins and attaches to the plasma membrane (Braun et al. 1991; De Angelis and Braun 1996; Laezza et al. 1997; Lee et al. 2005). The CNPase isoform 2 is imported into mitochondria, where it has been suggested to regulate matrix calcium release (Lee et al. 2006; Azarashvili et al. 2009).

Calmodulin (CaM) is a ubiquitous calcium sensor, capable of interacting with and regulating hundreds of target proteins. CaM is also abundant in the nervous system, and especially its interactions with the myelin basic protein (MBP) have been characterized in detail also at the structural level (Majava et al. 2010; Bamm et al. 2011; Wang et al. 2011). Two high-throughput proteomics studies have been carried out to find CaM targets in the brain (Berggård et al. 2006; Zhang et al. 2006). In both studies, CNPase was listed as a putative interaction partner for CaM, but the interaction has not been characterized using purified components, as was done, for example, for CRMP-2 in a similar case (Zhang et al. 2009).

We have earlier prepared expression systems for different domains of mouse CNPase, and purified several forms of it in large scale, to carry out detailed studies on CNPase structure and function (Myllykoski and Kursula 2010). Here, we used full-length CNPase and both of its domains separately to map the interaction site of CaM, and to study the effects of CaM on CNPase activity. In addition, the localization of CNPase and CaM in cell culture, the effect of CaM antagonists on CNPase in cultured cells, and post-translational modifications in brain CNPase were studied. The results bring about novel aspects of CNPase structure and function.

Experimental procedures

Protein expression & purification

The expression and purification of different versions of murine CNPase have been previously described (Myllykoski and Kursula 2010). Briefly, CNPase cDNA corresponding to residues 20-398 (CNP_20-398), 25-398 (CNP_25-398), 20-420 (CNP_20-420), 20-185 (CNP_20-185), and 179-398 (CNP_179-398) of CNPase (numbering according to isoform 2; the first 20 residues contain the mitochondrial targeting sequence, which was left out of the constructs) were cloned into the pTH27 vector (Hammarström et al. 2006), and the recombinant proteins were expressed in E. coli and purified with Ni-affinity and size exclusion chromatography. Additional purification for the N-terminal and full-length constructs was achieved by calmodulin affinity chromatography as described below, or using Blue Sepharose 6 (GE Healthcare, Uppsala, Sweden). Calmodulin was expressed in E. coli using the pETCM vector and purified as previously described (Hayashi et al. 1998; Kursula and Majava 2007). In addition, two synthetic peptides representing possible CaM-binding sites in the N-terminal domain (pep1 - KTAWRLDCAQLKEKNQWQ; pep2 - KSTLARVIVDKYRDGTKMV) were purchased from SBS Genetech (Beijing, China); the N-termini were acetylated and the C-termini amidated.

Affinity chromatography

Concentrated CNP_20-398, CNP_20-420, CNP_20-185, and CNP_179-398 were diluted to 0.3 mg/mL with equilibration buffer (50 mM HEPES pH 7.0, 300 mM NaCl, 5% glycerol, 1 mM dithiothreitol (DTT), and 10 mM CaCl2). CNPase was applied into a pre-equilibrated column containing approximately 1.2 mL of calmodulin agarose (Sigma-Aldrich, St. Louis, MO, USA; P4385, expected to contain 0.5 mg CaM per mL). The column was first washed with equilibration buffer and then, with buffer containing 5 mM EGTA instead of CaCl2. The eluted fractions were analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) .

Size exclusion chromatography

A Superdex 200 10/300 column was equilibrated with a buffer containing 20 mM Bis-Tris pH 5.5, 0.2 M NaCl, 10 mM CaCl2, and 1 mM DTT. 20 nmol of CaM, CNP_20-398, and CNP_20-185 were applied into the column separately, and complexes were run using 20 nmol of CNPase mixed with 20 nmol of CaM. Elution was monitored by measuring absorbance at 280 nm and by SDS-PAGE analysis of the fractions.

Affinity pull down from human brain white matter lysate

An affinity pull down experiment using CaM-agarose from a human white matter lysate was carried out essentially as described before (Majava et al. 2010). Human brain samples were obtained during autopsy (Department of Pathology, Oulu Central Hospital, Finland). Permission to use human brain tissue for research was obtained from the Finnish Medico-Legal Council (permit 102/32/200/99). Blocks of ~2 cm3 of white matter were dissected from the cerebral hemispheres of an 89-year-old male patient, with no known neurological diseases, and stored at −20°C in 1 mM EDTA.

A block of ~1 cm3 was cut from the frozen material and homogenized in 10 mL of CNPase solubilizing lysis buffer (Suda and Tsukada 1980), containing 10 mM HEPES (pH 8), 1 M ammonium acetate, 100 mM NaCl, 1% Triton X-100, 10 mM CaCl2 and EDTA-free Complete protease inhibitors (Roche, Penzberg, Germany). The homogenate was centrifuged for 20 min at 27 000 g, and the supernatant was collected. The supernatant was diluted 1 : 10, adjusted to contain 10 mM HEPES (pH 8), 100 mM ammonium acetate, 100 mM NaCl, 0.2% Triton X-100, 10 mM CaCl2, and centrifuged for 20 min at 27 000 g. 10 mL of the supernatant were mixed with 0.5 mL of CaM-agarose (Sigma, St. Louis, MO, USA) for 1 h. The unbound eluate was collected, the column was washed five times with 1 mL of diluted lysis buffer and five times with 1 mL of elution buffer, which contained 10 mM EGTA instead of CaCl2. The fractions were analyzed with SDS-PAGE and western blotting with a 1 : 1000 dilution of mouse monoclonal anti-CNPase (C5922, Sigma) as the primary antibody and a 1 : 100 dilution of horseradish peroxidase-conjugated goat anti-mouse IgG (32430, Thermo Scientific, Waltham, MA, USA) as the secondary antibody.

Interaction assays using intrinsic tryptophan fluorescence

Tryptophan fluorescence was used as a probe for a direct interaction between purified CNPase and CaM. CaM has no Trp residues, so the observed signal comes from the Trp residues on CNPase (5 Trp residues in full-length CNPase, of which 3 in the catalytic domain). Measurements were done at an excitation wavelength of 295 nm, and the emission spectrum was recorded between 310 and 450 nm. The measurement was done in 50 mM Bis-Tris pH 5.5, 200 mM NaCl, 10% glycerol, and 5 mM CaCl2. CaM was gradually added into a solution of 750 μL of 4 μM CNP_20-398 or CNP_179-398. Between additions, the solution was mixed for 3 min and fluorescence was measured.

Enzymatic activity assays

The activity of CNPase was measured as previously described (Sogin 1976; Myllykoski and Kursula 2010), except that Bis-Tris instead of MES was used as the reaction buffer, because MES slightly inhibits CNPase activity (unpublished data). All measurements were done in triplicate. The effect of CaM was assessed by adding different amounts of CaM from 3 fmol to 1 nmol into the reaction containing 1 pmol of CNP_20-398.

Polyadenylic acid-binding assays

Poly(A)-binding assays were carried out as previously described (Myllykoski et al. 2012). 10 mg of polyadenylic acid (Pharmacia, Uppsala, Sweden) were covalently coupled to 1 mL of CNBr-activated sepharose (GE Healthcare), according to the manufacturer's instructions. 50 μL of the coupled poly(A) sepharose were equilibrated with binding buffer (10 mM Tris-HCl at pH 7.5, 50 mM NaCl, 2.5 mM MgCl2, and 1 mM DTT). 500 μL of CNP_20-398 at 0.5 mg/mL, with and without 0.5 mg/mL CaM and 10 mM CaCl2, was mixed with poly(A) sepharose for 2 h at 4°C with rotation in microcentrifuge tubes. After mixing, the matrix was pelleted by centrifuging at 6600 g for 30 s, and the supernatant was removed. The matrix was washed five times with 500 μL of binding buffer and diluted to 100 μL. SDS-PAGE loading buffer was added, and 15 μL of the matrix were examined with SDS-PAGE.

Surface plasmon resonance

Surface plasmon resonance was used to analyze the CNPase–CaM interaction, with the Biacore T100 instrument (GE Healthcare). CaM was immobilized onto a CM5 chip as previously described (Majava et al. 2008), and the binding of CNP_20-398 and CNP_179-398 was analyzed. A channel with no immobilized CaM was used as a control, and the data were analyzed using the BiaEvaluation software (GE Healthcare).

Isothermal titration calorimetry

Isothermal titration calorimetry (ITC) was carried out essentially as described (Zhang et al. 2009). CNP_25-398, CaM, and the two CNPase peptides were dialyzed into the assay buffer (20 mM HEPES (pH 7.5), 100 mM NaCl, 10 mM CaCl2). ITC experiments were done on a Microcal VP-ITC (Microcal, Northampton, MA, USA) instrument at 30°C. For the CaM-CNPase titration, CNP_25-398 was in the cell at 19 μM and CaM in the syringe at 190 μM. For the CaM-peptide titrations, the CaM concentration in the cell was 16 μM and the peptide concentration in the syringe 200–300 μM. The data were analyzed with Microcal Origin.

Molecular modeling

The sequence similarity of the N-terminal domain of CNPase to other proteins is low. To find homologs and build a model, the Phyre2 server (Kelley and Sternberg 2009) at was used. The best hit (sequence identity of 22% over 133 residues, model confidence 99.9%) was the crystal structure of T4 polynucleotide kinase [(Zhu et al. 2007), PDB entry 2IA5], and the homology model built by Phyre based on this template was directly used in further analyses. Secondary structure predictions of CNPase matched well those observed in the template, and the P-loop, the only highly conserved feature of the N-terminal domain, is located in the same position. Several other hits to structures from the same fold family were obtained from the Phyre2 server, highlighting the reliability of the result. The CaM-binding site of the N-terminal domain of CNPase was predicted using the CaM target database (Yap et al. 2000).

Analysis of CaM antagonists in GC-4 cell cultures

CG-4 cells, a rat oligodendrocyte progenitor cell line, were cultured as described previously (Louis et al. 1992). The CG-4 cells were routinely expanded in Dulbecco's Modified Eagle's Medium (D-MEM; Invitrogen, Carlsbad, CA, USA), supplemented with the N2 supplements (Invitrogen), 10 ng/mL biotin (Sigma) and 10 ng/mL human platelet-derived growth factor-AA (PDGF-AA, Sigma). The cells were passed every 5 days on polyornithine (10 μg/mL) and fibronectin (5 μg/mL)-coated coverslips at a seeding density of ~10–50 cells/mm2. To differentiate the CG-4 cells, the medium was exchanged to serum-free chemically defined D-MEM (CDM), supplemented with 10 μg/mL insulin, 0.5 μg/mL transferrin, 100 μg/mL bovine serum albumin (BSA) , 60 ng/mL progesterone, 16 μg/mL putrescine, 40 ng/mL sodium selenite, 60 ng/mL N-acetyl-l-cysteine, 5 μM forskolin, 40 ng/mL l-thyroxine (T3), 30 ng/mL 3,3′,5-triiodothyronine (T4), and 5 ng/mL neurotrophin-3 (NT-3). The culture was continued for 7 days.

Differentiated CG-4 (dCG-4) cells were treated with the calmodulin inhibitors W-7 HCl (W-7) (5 μM, Wako) and trifluoperazine dihydrochloride (TFP) (2.5 μM, Sigma) overnight. To determine the expression and distribution of CNPase and tubulin in the cultured cells, double immunofluorescence staining for CNPase and tubulin was performed as previously described (Watanabe et al. 2006). The dCG-4 cells were fixed in 4% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS, pH 7.4) for 15 min at 20°C. The fixed cells were permeabilized with 0.5% Triton X-100 for 15 min before intracellular staining and were treated with 10% normal goat serum (Vector Laboratories, Burlingame, CA, USA) in PBS for 30 min to block non-specific binding. Following washing, the cells were incubated with the primary CNPase mAb (1 : 25, Sigma) overnight at 4°C. After washing, the cells were incubated with goat anti-mouse IgG Alexa Fluor® 488 (1 : 1000) for 1 h at 20°C. For double immunolabeling, the same procedures were repeatedly performed. After overnight incubation with the 2nd primary antibody, α-tubulin pAb (1 : 1500, Abcam, Cambridge, UK, 15246), at 4°C, the cells were incubated with goat anti-rabbit IgG Alexa Fluor® 594 (1 : 1000) for 1 h at 20°C. The stained cells were mounted using Vectashield® mounting medium (Vector Laboratories, Burlingame, CA, USA). Fluorescence microscopy was performed with an Olympus confocal laser-scanning FV1000 system with an inverted microscope (IX81; Olympus, Tokyo, Japan).

Localization of CNPase and CaM in myelinating Schwann cells

Mice (C57BL/6J strain) employed for the establishment of the dorsal root ganglion (DRG) explant cultures were acquired from the Laboratory Animal Centre, University of Oulu, and the Animal Welfare Committee has approved their use and the protocols employed (permit number 026/10).

DRG explant cultures, containing endogenous Schwann cells, were prepared according to a modified protocol (Fex Svenningsen et al. 2003). Cultures were maintained under a 5% CO2 atmosphere, at 37°C. DRG isolation was performed as described (Päiväläinen et al. 2008). Briefly, the DRGs were removed from the spinal cords of 13.5-day-old mouse embryos and placed in L15 medium. The pooled DRGs were incubated with 2 μg/mL dispase in 2 mL of L15 medium at 37°C for 45 min, washed three times with L15 medium containing 10% HS, and seeded in a drop of DRG growth medium (Päiväläinen et al. 2008) onto 3D-Matrigel-coated 13-mm glass cover slips on four well cluster plates, with 1–2 DRGs on each cover slip. The next day, 250 μL of DRG growth medium were added into each well, and the explants were grown in DRG growth medium for 12 days, renewing the medium every 2–3 days, to allow the endogenous Schwann cells to proliferate and populate the axons. Myelination was then started by the addition of myelination medium to the cultures (Päiväläinen et al. 2008). Myelination was allowed to proceed for 10 days or 3 weeks, renewing the myelination medium every 3 days.

The DRG explant cultures were fixed for immunocytochemistry after 12 days in DRG growth medium (= 0 days in myelination medium), and after 10 and 21 days in myelination medium. The cocultures were rinsed 3 times with PBS, fixed with 4% paraformaldehyde in PBS for 10 min at 20°C, and then, rinsed three times with PBS. Before the staining, the cells were permeabilized for 6 min with ice-cold methanol. The cultures were blocked with 5% goat serum in PBS at 4°C overnight. The primary antibodies (rabbit polyclonal anti-CNPase (Santa Cruz Biotechnology, Santa Cruz, CA, USA; dilution 1 : 75) and mouse monoclonal anti-CaM (Lab Vision, Kalamazoo, MI, USA; dilution 1 : 75) were diluted in 1% BSA in PBS, and incubation on the fixed cultures was carried out at 20°C for 80 min. The cultures were then washed three times for 5 min with PBS, incubated for 80 min with the secondary antibodies (goat anti-mouse Alexa Fluor 488 (Molecular Probes, Eugene, OR, USA), dilution 1 : 150 and goat anti-rabbit Alexa Fluor 546 (Molecular Probes), dilution 1 : 150 in 1% BSA in PBS) at 20°C, washed twice with PBS for 5 min and again overnight at 4°C in PBS. After rinsing twice with sterile H2O, the cultures were mounted with Immu-Mount on objective slides.

The immunolabeled cultures were examined with an OLYMPUS FluoView-1000 laser-scanning confocal microscope equipped with argon and HeNe1 lasers, and using a UPLSAPO 60x O (NA: 1.35) objective, or a UPLSAPO 100x O (NA:1.40) oil immersion objective. Pictures were taken using Z-stack (11–13 slices), and an optical slice thickness of 0.25–0.31 μm.

Identification of post-translational modifications

Most myelin proteomics experiments were carried out exactly as described before (Baer et al. 2009; Chen et al. 2010; Majava et al. 2010), including the preparation of myelin protein extracts, electrophoresis, in-gel digestion, and peptide analysis. Details are given in the supporting information.

Results and discussion

In two earlier studies (Berggård et al. 2006; Zhang et al. 2006), focused on the identification of CaM-binding proteins in the brain, CNPase was for the first time identified as a potential CaM target, but follow-up studies have not been performed to date. Our aim was to validate the direct molecular CNPase–CaM interaction and study its potential functional significance.

CNPase binds directly to CaM in a calcium-dependent manner

CNPase binding to calmodulin, the domain localization of the binding, and the dependence of the binding on calcium were studied by affinity chromatography on calmodulin-coupled agarose (Fig. 1a). CNP_20-398 and the N-terminal domain bound to CaM, while the C-terminal domain did not. Both bound fragments were eluted from the matrix, when calcium was removed by EGTA. The result shows that the CNPase N-terminal domain harbors a calcium-dependent binding site for CaM. Interestingly, the inclusion of the 22-residue C-terminal tail in the CNP_20-420 construct abolishes CaM binding in this assay. This tail is known to mediate the binding of CNPase to the membrane, and it is required for CNPase–tubulin interactions (De Angelis and Braun 1994; Lee et al. 2005). The structural basis for this effect is currently unknown, as the available crystal structures are missing the tail region. It is known, however, that the tail must be located close to both the N-terminal domain and the active site, and it may, hence, block the CaM-binding site (see below).

Figure 1.

Interaction between 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase) and CaM in vitro. (a) Affinity chromatography of CNPase on CaM-sepharose. CNPase was passed through CaM-agarose in the presence of calcium. The column was first washed with a calcium-containing buffer and then, eluted with an EGTA-containing buffer. The fractions were analyzed with sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using Coomassie staining. 1 - input sample; 2 - flow-through; 3 - combined calcium washes; 4–6 - EGTA eluates. (b) CaM affinity pull down from white matter lysate. The brain lysate was mixed with CaM-agarose and applied to a column. The column was washed first with a CaCl2-containing buffer and then, with an EGTA-containing buffer. Top: western blot with anti-CNPase; bottom: Coomassie staining. 1 - input sample; 2 - unbound sample; 3–8 - calcium washes; 9–13 - EGTA washes. CNPase is indicated with an asterisk and MBP with a double; in the EGTA elutions, a lot of myelin basic protein (MBP) is present, which gives a cross-reaction with the antibody. (c) Size exclusion chromatography indicates that the peaks of CNP_20-398 and CNP_20-185 are shifted when the proteins are mixed with CaM.

As a further test of interaction, CaM affinity pulldown was performed from a human white matter lysate (Fig. 1b). Using western blotting, a band corresponding to native CNPase was detected in the EGTA eluate. Previously, a similar experiment indicated that MBP is the major CaM-binding protein in a human white matter lysate (Majava et al. 2010), and indeed, we could again detect MBP in the affinity-purified fractions. Size exclusion chromatography of CNPase with CaM indicated the presence of a new species, corresponding to a complex, with both the CNP_20-398 and the N-terminal domain (Fig. 1c).

Surface plasmon resonance (Fig. 2a–c) was used to obtain an estimate for the CNPase–CaM affinity. Full-length CNPase bound to covalently immobilized CaM with an apparent Kd of 100 nM, while no interaction was seen between CaM and the phosphodiesterase domain (Fig. 2a). Binding in solution was also detected using ITC (Figure S1) and covalent cross-linking (unpublished data). Using intrinsic tryptophan fluorescence spectroscopy, binding between CNPase and CaM was also detected (Fig. 2d). A change was seen in the wavelength of the spectral maximum as a function of CaM concentration. Taken together, our results indicate that CNPase binds to CaM in a calcium-dependent manner, and that the interaction requires the presence of the N-terminal domain.

Figure 2.

Quantitative analysis of 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase)-CaM binding with surface plasmon resonance and fluorescence spectroscopy. (a) Binding of CNPase to immobilized CaM in surface plasmon resonance. CNP_20-398 binds to CaM strongly (black), while no interaction is detected with the catalytic domain (CNP_179-398; red). (b) Injections of a concentration series of CNP_20-398 over immobilized CaM. The kinetic analysis indicates Kd≈0.1 μM. (c) A binding plot of the response at the end of the injection versus CNPase concentration. (d) Tryptophan fluorescence analysis of CaM binding to CNPase. The catalytic domain shows no change in the spectrum (black, no CaM; red, CaM). CNP_20-398 has a shift of the fluorescence peak with CaM (blue, no CaM; green, CaM).

The presence of CaM does not affect the activity of CNPase

The enzymatic activity of CNPase was measured in the presence and absence of CaM. To get rid of possible contaminating nucleotides, CNPase was additionally purified with Blue Sepharose. No changes in CNPase activity in the presence of CaM were detected (Fig. 3a). This indicates that CaM binding does not directly regulate CNPase activity in vitro. However, the activity assay uses 2′,3′-cyclic NADP+ as the substrate, while the physiological substrate is likely to be an RNA molecule. Earlier, upon characterization of CNPase activity in bovine adrenal medulla, CaM was also found to have no effect on the activity in vitro toward 2′,3′-cyclic AMP (Tirrell and Coffee 1986). Considering the assumption that the physiological substrate of CNPase could be RNA, it cannot be ruled out, based on this assay, that CaM might indirectly regulate CNPase activity by influencing substrate binding.

Figure 3.

Analysis of the effects of CaM on 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase) activity and RNA binding. (a) Activity assay. CNPase activity was determined with the coupled enzyme assay (Sogin 1976) in the presence of a range of CaM concentrations. Absorption at 340 nm from the produced NADPH is plotted against time for the first 100 s of the reaction. (b) RNA binding. CNP_20-398 was mixed with agarose-coupled polyadenylic acid with and without CaM. 1 - input sample; 2 - unbound fraction; 3–6 - washes; 7 - sample of poly-A matrix after washes. CNPase is indicated with a single asterisk and CaM with a double.

Polyadenylic acid binding by CNPase

CNPase has previously been shown to bind RNA (Gravel et al. 2009; Myllykoski et al. 2012). We measured binding of purified full-length CNPase to sepharose-immobilized polyadenylic acid in the presence and absence of CaM. CaM had no significant effects toward the binding of RNA, when using polyadenylic acid binding as the model system (Fig. 3b). In an earlier report, it was suggested the N-terminal domain, containing the P-loop, would be a major site for binding nucleoside triphosphates, especially GTP (Stingo et al. 2007). It is, thus, possible that CaM may affect nucleotide binding by CNPase.

CNPase and CaM in differentiated CG-4 cells

To study the interaction of CNPase with CaM in dCG-4 cells, we performed immunocytochemical studies with CaM inhibitors (W7 and TFP) using anti-CNPase and anti-tubulin antibodies (Fig. 4). CNPase and tubulin were strongly expressed in the processes and cell bodies of dCG-4 cells. CNPase was expressed at the tips of processes, such as growth cones, while tubulin was absent. Treatment of dCG-4 with 10 μM W7 or 5 μM TFP for 24 h caused morphological changes, including the disappearance of processes, a round shape, and cell death. Hence, lower doses of these CaM antagonists were used (Fig. 4). The dCG-4 cells were supplemented with 5 μM W7 or 2.5 μM TFP for 24 h. CNPase was not colocalized with tubulin in non-treated dCG-4 cells. On the other hand, following exposure to W7 and TFP, CNPase colocalized with tubulin in the cytoplasm of dCG-4. These results suggest that the interaction of CNPase with tubulin could be directly or indirectly regulated by CaM.

Figure 4.

The effects of CaM antagonists on 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase) localization in dCG-4 cells. The panels are as follows: top left – merge, top right – CNPase, bottom left – tubulin, bottom right – DIC. 50-μm scale bars are shown in the bottom left panels in white. 157, 118, and 132 cells were examined for the untreated, trifluoperazine dihydrochloride (TFP)-treated, and W7-treated experiments, respectively, and representative results are shown.

Localization of CNPase and CaM in myelinating Schwann cells

Double immunofluorescence microscopy analysis of CNPase and CaM expression in mouse DRG explant cultures reveals their partial colocalization during myelination (Fig. 5). The simultaneous presence of both CNPase and CaM in the same cellular compartments supports the possibility of a direct physical interaction between these two proteins. The apparent colocalization was stronger during early stages of myelination.

Figure 5.

2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase) and CaM expression during myelination. Confocal immunofluorescence microscopy reveals the partial colocalization (orange–yellow) of CNPase (red) and CaM (green) in mouse dorsal root ganglion (DRG) explant cultures after 0, 10, and 21 days in conditions permissive for myelination. Colocalization is most evident in immature structures (inset at 0 days). CaM appears to be absent, or very weakly expressed, in compacted sheaths (yellow arrows in insets at 10 and 21 days). Scale bars: 100 μm. The insets in the merged images are digitally magnified from the areas outlined with dotted lines.

The putative CaM- binding site of the CNPase N-terminal domain

To get a more detailed view of the N-terminal domain, a homology model was built (Fig. 6a). The predicted CaM-binding segment (Yap et al. 2000) is in the immediate vicinity of the P-loop, on the protein surface. CaM binding could, thus, regulate the binding of CNPase to the ligands of its N-terminal domain. We also noted another predicted amphipathic helix in the N-terminal domain (Fig. 6a). Interestingly, in an ITC experiment, the latter gave a strong signal indicative of binding (Figure S1).

Figure 6.

Models for 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase) structure and function. (a) Homology model for the CNPase N-terminal domain. The predicted CaM-binding site by the CaM target database is indicated in yellow, a second potentially amphipathic helix in pink, and the ATP-binding P-loop in red. An ADP molecule has been added to denote the ATP-binding site. The two phosphorylated residues are also shown. The locations of the C- and N-termini are indicated; the catalytic domain will follow directly after the C-terminus. (b) Schematic view of the current data on CNPase function at the molecular level. CNPase is comprised of two folded domains (green/red), and a C-terminal membrane anchor (orange). As the N- and C-termini of the catalytic domain are next to each other in the structure (Myllykoski et al. 2012), the C-terminal tail formed by the last 22 residues must lie close to the domain interface, as depicted. Although the catalytic domain performs the phosphodiesterase reaction (blue), our current and earlier data indicate a role for the N-terminal domain in dimerization, RNA binding, and CaM binding. CNPase also interacts with the actin and tubulin cytoskeleton, and it is likely that at least the C-terminus of CNPase is involved in these interactions.

Identification of two novel phosphorylation sites in brain CNPase

A mass spectrometric approach (Figure S2, Table S1) was taken to identify novel post-translational modifications in rat brain CNPase. Two previously unreported phosphorylation sites were detected and confirmed by phosphatase treatment (Figure S3). Tyr100 is predicted to lie in the middle of a long helix, on the surface of the N-terminal domain. Ser169 is located some 10 residues before the beginning of the C-terminal phosphodiesterase domain, in a region predicted to be helical and connecting the two domains. In the homology model of the N-terminal domain, both residues are solvent-accessible, and their phosphorylation could affect the interactions of CNPase with its ligands (Fig. 6a).

Implications of CaM binding for CNPase function

The function of CNPase can be divided into three distinct, but probably connected, aspects. First, the enzymatic phosphodiesterase activity of CNPase converts 2′,3′-cyclic nucleotides into 2′-nucleotides (Drummond et al. 1962). The biological role of this activity, and – more specifically – the high amount of such activity within the developing myelin sheath, has puzzled researchers through decades. CNPase activity and its 2′,3′-cyclic substrates were found to modulate the mitochondrial permeability transition pore, which releases matrix contents with apoptotic consequences after elevated matrix calcium levels (Azarashvili et al. 2009). Calmodulin, on the other hand, has been shown to regulate the uptake of calcium into the matrix (Moreau et al. 2006). Recently, 2′, 3′-cyclic nucleotides were detected in the brain of CNPase knock-out mice (Verrier et al. 2012). The catalytic activity of CNPase was also linked as a possible player in the 2′,3′-cyclic AMP-adenosine pathway (Verrier et al. 2012); free 2′,3′-cyclic AMP is a side product of mRNA degradation (Thompson et al. 1994) coupled to cell injury (Jackson et al. 2009; Jackson 2011). Our assays, however, revealed no direct effects on CNPase catalytic activity by CaM.

Another recently discovered function for CNPase is the binding of RNA molecules (Gravel et al. 2009). In our earlier study, we showed that the N-terminal domain is required for efficient RNA binding in vitro (Myllykoski et al. 2012). In this study, we did not detect any difference in polyadenylic acid binding with and without CaM, indicating that the role of the CaM interaction is unlikely to be related to RNA ligand binding.

A third property of CNPase is its affinity toward the cellular cytoskeleton, including both actin and tubulin (De Angelis and Braun 1996; Laezza et al. 1997). CNPase is copurified with cytoskeletal elements in the detergent-resistant fraction (Kim and Pfeiffer 1999), and its C-terminus, mediating linkage to the lipid membrane, is essential for interactions with tubulin (Lee et al. 2005). The cytoskeleton plays a leading role during myelination, and dynamic polymerization and depolymerization of microfilaments and microtubules are necessary for a functional outcome (Bauer et al. 2009). The influx of Ca2+ and depolymerization of microtubules have been reported as responses for antibodies against cell surface galactocerebroside in oligodendrocyte cultures (Dyer and Benjamins 1990). Also, as a response to axonal stimulation in frog sciatic nerve, Ca2+ influx into Schwann cell cytoplasm has been detected (Lev-Ram and Ellisman 1995). CNPase has been shown to promote microtubule polymerization and reorganization (Lee et al. 2005), and it can be speculated that CaM binding might modulate this effect, depending on the variation of intracellular [Ca2+].

Both CNPase and the myelin basic protein (MBP) interact with membrane surfaces, cytoskeletal proteins, and calmodulin. The binding to cytoskeletal components and calmodulin creates an interesting analogy between CNPase and MBP. MBP also is likely to be involved in the regulation of CNPase localization outside the compacted regions of myelin (Aggarwal et al. 2011). MBP and CNPase could be involved in a regulatory network involving myelin membrane surfaces, the myelinating cell cytoskeleton, CaM, and other factors.

A schematic view of the current data on the functional domains of CNPase is shown in Fig. 6b. Although binding partners for the N-terminal domain and structural properties and catalytic activity for the C-terminal domain have been studied, it still remains unclear, how all these properties are interconnected. Our results warrant further studies on the complex interactions of CNPase with other molecules.

Concluding remarks

We have shown that CNPase directly interacts with CaM in a calcium-dependent manner. Using in vitro functional assays, we were unable to detect an effect of CaM on CNPase enzymatic activity or RNA binding. It is possible that the interaction plays a role in an as-of-yet unidentified function of CNPase; such functions could be related to CNPase localization or its interactions with the cytoskeleton. When the structure and function of the CNPase N-terminal domain, as well as those of the full-length protein, have been better elucidated, CNPase function in vivo will also become more clear.


We thank Dr. Young-Hwa Song for supervising the fluorescence titrations. This study has been financially supported by the Academy of Finland (Finland), the Sigrid Jusélius Foundation (Finland), the Magnus Ehrnrooth Foundation (Finland), the Department of Biochemistry, University of Oulu (Finland), and the Research and Science Foundation of the City of Hamburg (Germany). The authors declare no conflicts of interest.