The C-terminal C1 cassette of the N-methyl-d-aspartate receptor 1 subunit contains a bi-partite nuclear localization sequence


and reprint requests to G. A. Dekaban, Gene Therapy and Molecular Virology Group, The John P. Robarts Research Institute, 100 Perth Drive, London, Ontario, Canada N6A 5K8. E-mail:


The N-methyl-d-aspartate receptor (NMDAR) is a multimeric transmembrane protein composed of at least two subunits. One subunit, NR1, is derived from a single gene and can be subdivided into three regions: the N-terminal extracellular domain, the transmembrane regions, and the C-terminal intracellular domain. The N-terminal domain is responsible for Mg2+ metal ion binding and channel activity, while the transmembrane domains are important for ion channel formation. The intracellular C-terminal domain is involved in regulating receptor activity and subcellular localization. Our recent experiments indicated that the intracellular C-terminal domain, when expressed independently, localizes almost exclusively in the nucleus. An examination of the amino acid sequence reveals the presence of a putative nuclear localization sequence (NLS) in the C1 cassette of the NR1 intracellular C-terminus. Using an expression vector designed to test whether a putative NLS sequence is a valid, functional NLS, we have demonstrated that a bi-partite NLS does in fact exist within the NR1-1 C-terminus. Computer algorithms identified a putative helix–loop–helix motif that spanned the C0C1 cassettes of the C-terminus. These data suggest that the NR1 subunit may represent another member of a family of transmembrane proteins that undergo intramembrane proteolysis, releasing a cytosolic peptide that is actively translocated to the nucleus leading to alterations in gene regulation.

Abbreviations used:

cyan fluorescent protein


Dulbecco's minimal essential media


enhanced green fluorescent protein


human embryonic kidney


hierarchical neural network


multivariate linear regression combination


nuclear localization sequence


N-methyl-d-aspartate receptor


pheochromocytoma 12


post-synaptic densities


yellow fluorescent protein.

The N-methyl-d-aspartate receptor (NMDAR) is a multimeric transmembrane glutamate receptor protein that plays a major role in learning and memory (Malenka and Nicoll 1999; Thompson 2000). The NMDAR is composed of at least two subunits: NR1 and NR2 (Monaghan et al. 1989; Kutsuwada et al. 1992; Monyer et al. 1992). Recently a third subunit, NR3, has been described (Sun et al. 1998; Das et al. 1998). The exact stoichiometry for each of the subunits has not yet been resolved. The NR1 subunit is essential for ion channel activity, while the NR2 subunit potentiates the channel activity of NR1 and is required for glutamate and glycine binding (Grimwood et al. 1995). The role of the NR3 subunit is still not entirely clear but appears to modify channel activity (Das et al. 1998; Perez-Otano et al. 2001).

The NR1 subunit, derived from a single gene, can be subdivided into three regions: the N-terminal extracellular domain, the transmembrane regions and the C-terminal intracellular domain (Moriyoshi et al. 1991). The N-terminal domain is responsible for ligand binding, Mg2+ ion binding and channel activity, whereas the transmembrane domains are important for ion channel formation. The intracellular C-terminal domain is involved in regulating receptor activity and subcellular localization. The four major C-terminal splice variants (NR1-1, NR1-2, NR1-3 and NR1-4) of the NR1 subunit differ by the presence or absence of three different cassettes, C1, C2 and C2′(Fig. 1; Sugihara et al. 1992). Each of these splice variants can exist in two forms depending on the presence (b form) or absence (a form) of an exon cassette in the N-terminal extracellular domain that modifies ion channel activity (Sugihara et al. 1992). The C0 cassette, common to all NR1 splice variants, appears to be involved in mediating interactions with the cytoskeleton as it binds to α-actinin-2 (Wyszynski et al. 1997). The C1 cassette found in the NR1-1 and NR1-3 cassettes has been reported to contain elements important to cytoskeletal interactions [tubulin (12) and neurofilament (13)], mediating receptor regulation and function via phosphorylation of sites within C1 (Tingley et al. 1993; 1997) and to signal transduction protein interactions (Ehlers et al. 1996). The C2′ of NR1-3 and NR1-4 contains a T/SXV motif that allows it to interact with proteins that contain PDZ domains (Garner et al. 2000) and thus may mediate direct recruitment and retention at post-synaptic densities (PSD; Hsueh and Sheng 1998; Kennedy 1998; O'Brien et al. 1998). No specific function has yet been ascribed to the C2 cassette located at the C-terminus of NR1-1 and NR1-2.

Figure 1.

Illustration of the different NR1-GFP fusion proteins used to map the NLS located in the intracytoplasmic C-terminal domain. The pNR1-C0C1C2 (containing the entire NR1-1 subunit fused to GFP or CFP) and pGFP-C0C1C2 (containing only the C-terminal domain of NR1-1) were expressed from the vector pEGFP-N. The remaining constructs were expressed from the vectors (p829 and p830) designed to confirm the presence or absence of a functional NLS (Sorg and Stamminger 1999). The p829 and p830 vectors differed only with respect to the opposite orientations in which GFP and β-galactosidase were fused to the sequence containing the putative NLS. Only the p830 orientation is presented.

Our recent studies have been directed toward understanding the role of the NMDAR in sympathetic preganglionic neurons after spinal cord injury within the context of the development of autonomic dysreflexia (Maiorov et al. 1997). We have focused our attention on the role of each of the individual NR1 C-terminal cassettes in the subcellular distribution of NMDARs containing the NR1-1, NR1-2 and NR1-4 splice variants as these are the most abundant NR1 splice variants in the spinal cord and in sensory neurons in the dorsal root ganglia (Tolle et al. 1993, 1995; Liu et al. 1994). As part of these studies, we recently reported that recombinant NR1 intracytoplasmic C-terminal domains containing the C1 cassette fused to the enhanced green fluorescent protein (EGFP) localize almost exclusively to the nuclei with little or no presence in the cytoplasm (Marsh et al. 2001). We also observed that the intracytoplasmic C-terminal domains of the NR1-2 and NR1-4 splice variants were located equally in both the cytoplasm and the nucleus, suggesting that, because of their small size, they could readily pass through nuclear pores, resulting in an even distribution throughout the cell. In this report we carried out studies to determine if specific nuclear localization signals were present in any of the NR1 C-terminal cassettes. A number of recent examples of transmembrane and membrane-associated proteins containing nuclear localization motifs have been reported (Brown et al. 2000; Fukuda et al. 2001; Miller et al. 2001; Zhang et al. 2001). These motifs impart novel subcellular localizations and novel functions that may have pathological significance in some cases. Our findings demonstrate that the C1 cassette but not the C0, C2 or C2′ cassettes contains a specific nuclear localization signal.

Experimental procedures

Cell lines

The E2 clone of pheochromocytoma 12 (PC12) cells was obtained from Dr R.A. Bradshaw (Wu and Bradshaw 1995) and grown in Dulbecco's minimal essential media (DMEM) in the presence of 5% fetal calf serum, 5% horse serum, 10 U/mL penicillin and 10 µg/mL streptomycin, 2 mm glutamine in 5% CO2. PC12 cells were differentiated with nerve growth factor as previously described (LeVatte et al. 1998). Only differentiated PC12 cells were used in this study. Human embryonic kidney (HEK) 293 cells were grown in a similar fashion except that only 10% fetal calf serum was used. All tissue culture reagents were obtained from Life Technologies Ltd. (Mississauga, ON, Canada).

Molecular clones

The rat cDNA clones for NR1-1, NR1-2, NR1-4 and NR2A were obtained as a kind gift from Shigetada Nakanishi (Sugihara et al. 1992; Ishii et al. 1993). The original NR1-1-GFP was obtained from Thomas Hughes (Marshall et al. 1995) and was further modified to contain the enhanced version of GFP (Clontech, Palo Alto, CA, USA) as previously described (Holmes et al. 2000). The expression vectors pHM829 and pHM830 were utilized to map the NLS by fusing a putative NLS sequence simultaneously between GFP and β-galactosidase. These vectors plus positive control vector pHM840, containing the SV40 T antigen NLS, were kindly provided by Dr T. Stamminger (Sorg and Stamminger 1999).

To map the NLS present in the C-terminal cassettes of NR1-1, recombinant PCR techniques were employed to generate the desired deletions or mutations. DNA sequencing on an ABI model 377 Automated DNA Sequencer was used to verify the correctness of the DNA sequence in all PCR products and that they were cloned in frame with respect to both GFP and β-galactosidase. The NR1-1 C-terminus containing the C0, C1 and C2 cassettes, individual cassettes or mutants were fused to either the N-terminus of enhanced GFP or enhanced cyan fluorescent protein (CFP, Clontech). In a similar fashion NR2A was fused to the enhanced yellow fluorescent protein (YFP, Clonetech) at the very end of the C-terminus after removal of the stop codon creating NR2A-YFP. Plasmids were purified using Qiagen (Mississauga, ON, Canada) plasmid purification columns (Tip-100 or Tip 500) according to the manufacturers' instructions.


All transfections were carried out with Lipofectamine Plus (Invitrogen Life Technologies, Mississauga, ON, Canada) according to the manufacturer's instructions. Transient transfections were carried out with 2 µg of plasmid DNA.HEK293 cells stably transfected with the pNR2A-YFP plasmid were selected for neomycin resistance. The stable pNR2A-YFP transfectants were subsequently cloned to obtain a population of cells uniformly expressing NR2A-YFP. Western blot analysis showed that the correct protein was being produced (data not shown, details to be published elsewhere).

Fluorescence microscopy

The recombinant proteins fused to the enhanced GFP, YFP and CFP were observed using fluorescent microscopy employing an Olympus IX-50 inverted microscope and filter cubes specific for EGFP (U-M41028), ECFP (U-M31044) and EYFP (U-M41028), all from Chroma Technology Corp. (Brattleboro, VT, USA). Nuclei were stained with the Hoescht 34295 (Sigma-Aldrich, Oakville, ON, Canada) stain as previously described (LeVatte et al. 1998). A Zeiss LSM-510 confocal microscope was also used to characterize GFP fusion protein fluorescence. Images were processed using Image Pro Plus software (Media Cybernetics Inc., Silver Spring, MD, USA), LSM 510 software (Zeiss, Thornwood, NY, USA), Adobe Photoshop 5.02 (Adobe Systems Inc., San Jose, CA, USA) and Corel Draw 8 (Corel Inc., Kanata, ON, Canada). β-galactosidase expression was detected by staining the cells with a rabbit antiβ-galactosidase polyclonal antibody (United States Biological, Swampscott, MA, USA) as previously described (Marsh et al. 2000). Staining of neuronal nuclei was achieved with a monoclonal antibody to the neuron-specific transcription factor NeuN (Chemicon, Mississauga, ON, Canada) as previously described (Marsh et al. 2000). Actin staining was performed using TRITC-conjugated phalloidin (Sigma-Aldrich) and is described elsewhere (Holmes et al. 2000).

Nuclear extraction

Cultured cell nuclei were isolated and soluble proteins were extracted using a protocol modified from Ausubel et al. (1989). Briefly, 293 cells, PC12 cells, and both cell types stably expressing NR1-C0C1C2-GFP were harvested with and then lysed in 2.5 mL of a buffer consisting of 10 mm HEPES pH 7.9, 1.5 mm MgCl2, 10 mm KCl, 0.5 mm dithiothreitol (DTT), 0.1% Triton X-100, and one-quarter of a Complete Mini, EDTA Free protease inhibitor cocktail tablet (Roche Diagnostics GmBH, Mannheim, Germany). Lysis was allowed to occur for 2–3 min and monitored by microscopy. Nuclei were pelleted by centrifugation at 4°C for 5 min at 800 g. Nuclei were resuspended in two volumes (∼1 mL) of a solution consisting of 20 mm HEPES pH 7.9, 25% glycerol, 0.42 m NaCl, 1.5 mm MgCl2, 0.2 mm EDTA, 0.5 mm DTT, and one-sixth of a Complete Mini, EDTA Free protease inhibitor cocktail tablet (Roche). The nuclei were incubated on a rocker platform at 4°C for 30 min, and then centrifuged at 13 000 g for 30 min at 4°C. The supernatants were quantified via the Bio-Rad Protein Assay (Bio-Rad, Hercules, CA, USA) followed by electrophoresis and immunoblotting. Blots were analyzed using commercially available, affinity-purified, polyclonal rabbit anti-NMDAR1 antibody, raised against the C1 cassette (AB5046P; Chemicon International, Temecula, CA, USA), and affinity-purified, polyclonal rabbit anti-NMDAR1 antibody raised against the C2 cassette (AB1516; Chemicon International). An horseradish peroxidase (HRP)-coupled goat anti-rabbit antibody served as the secondary antibody (Jackson Immunoresearch, West Grove, PA, USA).

DNA and protein sequence analysis

The analysis of DNA and amino acid sequences was performed using the Lasergene99 sequence analysis software package (DNASTAR Inc., Madison, WI, USA) to search for DNA and protein homologies. NLS motifs were identified using the PSORT II program (Nakai and Kanehisa 1992) (accessed through the Expasy www server Similarities to the consensus motif of a helix-loop-helix domain were determined using the PROSITE database (Bucher and Bairoch 1994; Hofmann et al. 1999). Secondary structure algorithms were used to predict alpha-helical regions within peptides. The utilized programs were the multivariate linear regression combination (MLRC; Guermeur et al. 1999), the hierarchical neural network (HNN; Guermeur, Y., PhD Thesis), and Predator (Frishman and Argos 1996). The peptides were further analyzed using a Type-II Discrete State-space Model from the Protein Sequence Analysis server at the Biomolecular Engineering Research Center (Stultz et al. 1993, 1997; White et al. 1994). The displayed plots exhibit the calculated secondary structural features of each amino acid residue in terms of probability, with each isobar representing a probability value of 0.1.


Nuclear localization of the recombinant form of the intracytoplasmic C-terminal domain of NR1

As part of our ongoing studies we have made various protein fusions with GFP and the intracytoplasmic C-termini of the NR1-1 (C0C1C2), NR1-2 (C0C2) and NR1-4 (C0C2′) subunits that were produced by the expression vectors pGFPC0C1C2, pGFPC0C2 and pGFPC0C2′, respectively (Fig. 1). In transiently transfected HEK293 or NGF-differentiated PC12 cells (data not shown), fluorescence microscopy demonstrated that the GFP-C0C2 (Figs 2a–c) and GFP-C0C2′ (Figs 2d–f) C-terminal fusion proteins were located in the cytoplasm and to a greater extent in the nucleus 24 h post transfection. This was also observed in stably transfected HEK293 cells (not shown). However, the fusion protein from the pGFP-COC1C2 construct was found almost entirely in the nucleus of HEK293 cells (Figs 2g–i). Similar results were found in PC12 cells (Fig. 2j) 24 h after transfection and this distribution had not changed 96 h after the transfection (Fig. 2k). This highly concentrated nuclear distribution of the GFP-C0C1C2 fusion protein was in sharp contrast to that observed for the distribution of the complete recombinant NR1-1 subunit fused to GFP (pNR1-C0C1C2-GFP, see Fig. 1) which was found predominantly in the cytoplasm and was not visible in the nucleus (Fig. 2l).

Figure 2.

Transient expression of the intracytoplasmic C-terminal domain of NR1-1 (C0C1C2), NR1-2 (C0C2), and NR1-4 (C0C2′) fused to GFP in HEK293 and PC12 cells. Twenty-four hours after transfection, HEK293 cells were fixed and then stained with Hoescht 34295, a nuclear stain, and TRITC-phalloidin, an actin stain. Using confocal microscopy, cells were then imaged for the blue fluorescence of Hoescht 34295 (a,g) and the green fluorescence of GFP (b,e,h). The combined images showing colocalization between C-termini and nuclear staining are also shown (c,f,i). Localization of the NR1-1 C-terminus is also shown at 24 (j) and 96 h (k) in differentiated PC12 cells. Full-length NR1-1 (NR1-C0C1C2-GFP) is shown at 96 h after transfection in PC12 cells (l). TRITC-phalloidin was used as a counterstain to show cell outlines and is shown in red in each image. Scale bars represent 10 µm.

Identification of NLS motifs within the NR1-1 C-terminal cassettes

To determine the reason for the nuclear localization of our GFP fusion proteins, we examined the amino acid composition and sequence of the four NR1 C-terminal cassettes. Twenty-seven of the 105 amino acids comprising the C-terminus of NR1-1 are basic amino acids. This, and the relatively small size of the GFP fusion proteins, might partially explain the presence of the GFP-C0C2 and GFP-C0C2′ fusion proteins in the nucleus of transfected cells, but does not adequately explain the nearly exclusive nuclear localization of GFP-C0C1C2. However, as illustrated in Fig. 3, computer algorithms predicted several potential nuclear localization signal (NLS) motifs (Nakai and Kanehisa 1992; Stultz et al. 1993; Frishman and Argos 1996) including a sequence in the middle of the C1 cassette that resembled a bi-partite NLS. The putative C1 bi-partite NLS was consistent with the published consensus sequence for a bi-partite NLS and strongly resembled several well-characterized bi-partite NLS sequences (Michigami et al. 1999; Healy et al. 1999; Somasekaram et al. 1999). In addition, within the C0 cassette a four amino acid sequence (KRHK) was also identified as a potential NLS that might function as a mono-partite NLS or perhaps form another bi-partite NLS with additional groups of two or three basic residues located immediately downstream.

Figure 3.

Amino acid sequence alignment between the bi-partite NLS in the NR1-1 C1 cassette and known bi-partite NLS motifs. Amino acid sequence alignment of the putative NR1 bi-partite NLS (shaded sequence) in the C1 cassette with a consensus NLS (Mattaj and Englmeier 1998; Somasekaram et al. 1999) and known bi-partite NLS from other proteins. A putative monopartite NLS in C0 is shown in the box. GenBank assession numbers for the sequences compared are P15771 (nucleolin), P05221 (nucleoplasmin), P32320 (cytidine deaminase), S35560 (SRY), P17470 (Herpes simplex virus ICP-8) and J03258 (vitamin D receptor). The bi-partite NLS of 5-lipoxygenase was previously reported (Healy et al. 1999). The important amino acids in the putative NR1 NLS and in the known bi-partite NLS' are shown in bold type.

Mapping the NLS in the NR1-1 C-terminal cassettes

The following approach was undertaken to confirm whether the putative NLS motifs in the C0 or C1 cassettes functioned as a true NLS. First, to investigate the possibility that the 40–45 kDa GFP fusion proteins did not enter the nucleus by passive diffusion through nuclear pores and were not being retained due to their basic amino acid content, the NR1-1, NR1-2, and NR1-4 C-terminal cassettes were subcloned into expression vectors designed to specifically detect an NLS sequence (Sorg and Stamminger 1999). These expression vectors have multiple cloning sites located between the coding regions of GFP and β-galactosidase, creating a fusion protein in excess of 142 kDa. These fusion proteins were more than twice the size (60 kDa) that can passively diffuse through nuclear pores and thus they could only gain access to the nucleus by active transport mediated by an authentic NLS motif (Mattaj and Englmeier 1998; Sorg and Stamminger 1999). Two different vectors were employed that differed in their orientation: p829 expressed a GFP-β-galactosidase fusion protein, whereas p830 expressed a β-galactosidase-GFP protein. Some NLS function better when near the N-terminus and others are best located near the C-terminus of a protein (Sorg and Stamminger 1999). We addressed this issue by cloning the NR1 C-terminal cassettes in either position. If a NLS was detected with these constructs, then successive deletions of the C-terminal regions and mutations within the putative NLS were used to locate and confirm the specific region containing the putative NLS. The constructs used in these experiments are outlined in Fig. 1 and the results are presented in Table 1 and Figs 4, 5 and 6. Representative recombinant proteins were evaluated for the correctness of size and to verify that they were expressed to similar levels (data not shown).

Table 1.  Mapping of NR1-1 C-terminal nuclear localization sequence
Region testedNuclear localization of GFP fusion proteins
pEGFP-N1NLS vector p829NLS vector p830
GFP + NLSGFP + NLS + β-galβ-gal + NLS + GFP
  • a

    Equal distribution between cytoplasm and nucleus observed.

  • b

    The presence (+) or absence(–) of GFP/β-galactosidase in the nucleus was determined.

Bi-partite NLS of C1ND+/+b+/+b
Mutated bi-artite NLS of C1ND–/–b–/–b
Figure 4.

Mapping the NLS sequence in transiently transfected PC12 cells with constructs expressing the intracytoplasmic C-terminal domains of NR1-2 (C0C2, a,b) NR1-4 (C0C2′, c,d), and NR1-1 (C0C1C2, e,f) fused between GFP and β-gal. Forty-eight hours post-transfection confocal microscopy was employed to view the green fluorescence of GFP (a,c,e) and the blue fluorescence of the DNA stain Hoescht 34295 (b,d,f). The arrows indicate the position of the nuclei in the cells depicted in each panel. Those constructs containing the putative NLS (e,f) had GFP fluorescence colocalize with the blue fluorescence of the nuclear stain Hoescht 34295.

Figure 5.

Mapping the NLS sequence in transiently transfected PC12 cells with constructs expressing different regions of the NR1-1 C1 domain fused between GFP and β-gal. Forty-eight hours post-transfection confocal microscopy was employed to view the green fluorescence of GFP (a,c,e) and the blue fluorescence of the DNA stain Hoescht 34295 (b,d,f). The arrows indicate the position of the nuclei in the cells depicted in each panel. Those constructs containing the putative NLS had GFP fluorescence colocalize with the blue fluorescence of the nuclear stain Hoescht 34295.

Figure 6.

Mapping the NR1-1 C1 cassette NLS sequence in sensory neurons of organotypic cultures of rat DRG after biolistic gene transfer with constructs expressing different regions of the NR1-1 intracytoplasmic C-terminal domain fused between GFP and β-gal. 48 h after biolistic gene transfer the DRG were permeabilized and immunostained with a biotinylated antibody to the neuron-specific transcription factor NeuN and streptavidin-conjugated rhodamine. Confocal microscopy was employed to view the green fluorescence of GFP (a,c,e) and the red fluorescence of the rhodamine (b,d,f). Arrows indicate the position of the nuclei.

When the C-terminal regions of NR1-2 and NR1-4 were placed in the NLS vectors, the putative NLS sequences identified by sequence analysis within C0 were not functional. GFP fluorescence did not co-localize with the nuclear DNA-binding Hoescht stain but rather was excluded to the cytoplasm of transiently transfected HEK293 cells (Figs 4a–d). However, when the C0C1C2 C-terminal cassettes of NR1-1 were placed in the NLS vectors, the GFP fluorescence co-localized with the nuclear DNA binding Hoescht stain (Figs 4e and f).

To verify that the putative bi-partite NLS in the C1 cassette was mediating the nuclear localization, three additional constructs were made (Fig. 1). One consisted of only the C1 cassette (p829/830C1), another consisted of only the putative NLS (p829/830NLS), and the last was a mutant in which the basic amino acids of the NLS were substituted with glycines or alanines (p829/830ΔNLS). As shown in Fig. 5(a–d) and Table 1, the C1 cassette and the putative bi-partite NLS delivered the GFP-β-galactosidase fusion protein to the nucleus in both orientations. In both cases the GFP fluorescence co-localized with the DNA-binding Hoescht stain. As expected the mutated bi-partite NLS did not confer nuclear localization in either orientation with respect to the placement of GFP and β-galactosidase (Figs 5e and f and Table 1). The GFP fluorescence was restricted to the cytoplasm and did not co-localize with the Hoescht stain. A similar experiment was also performed for each of the constructs described above except that the transfected cells were stained with a monoclonal antibody to β-galactosidase (not shown). As observed by GFP fluorescence, the constructs that had an intact bi-partite NLS also had nuclei stained positively for β-galactosidase (summarized in Table 1). The construct containing the mutated NLS only had β-galactosidase-positive staining in the cytoplasm. These studies in tissue culture cells indicate that a bi-partite NLS exists in the C1 cassette of NR1-1.

To determine whether the C1 NLS functioned in neurons, the constructs were transferred into organotypic explant cultures of whole dorsal root ganglia using a biolistic gene gun delivery system as previously reported (Arnold et al. 1994; Thomas et al. 1998; Marsh et al. 2001). The NLS vectors containing the C1 cassette (p830C1, Figs 6a and b) or only the putative NLS (p830NLS, Figs 6c and d) had GFP fluorescence in the nucleus that co-localized with the nuclear staining of an antibody for the neuron-specific transcription factor neuN (Wolf et al. 1996; Mullen et al. 1992). Conversely, the GFP fluorescence produced by the p830ΔNLS vector with the mutated NLS was located only in the cytoplasm and did not co-localize with the neuron-specific neuN staining in the nucleus (Figs 6e and f).

Dectection of NR1-1 C-terminal fragments in nuclear extracts of 293 and PC12 cells

To ascertain whether or not immunoreactivity for NR1 and/or NR1 C-termini could be detected in the nucleus of a given cell, extracts were prepared from 293 and PC12 cells expressing NR1-GFP fusion proteins and then immunoblotted with appropriate antibodies. Whole-cell lysates (WC) of 293 cells stably expressing NR1-C0C1C2-GFP (4 µg protein loaded) exhibited both the full-length protein (Fig. 7a, arrow), and several low molecular weight fragments (bracket) that cross-reacted with antibodies directed against both the C2 (not shown) and C1 cassettes. These fragments were highly enriched in protein extracts from isolated nuclei (N; 4 µg protein loaded, note absence of NR1-C0C1C2-GFP from nuclear extracts). The fragments were somewhat lower in molecular mass in comparison to recombinant GFP-C0C1C2, which was stably expressed in 293 cells (arrowhead; 2 µg protein loaded). Recombinant GFP-C0C1C2 protein was also present in protein extracted from isolated nuclei (2 µg protein loaded) from recombinant GFP-C0C1C2 stably transfected cells. Protein (10 µg) from a control 293 whole-cell extract did not exhibit equivalent immunoreactive fragments. The NR1-GFP fusion protein NR1-GFP-C0C1C2, with GFP placed immediately after the fourth transmembrane domain of NR1, was also cleaved in 293 cells (Fig. 7b). The observed fragments (bracket) were highly enriched in protein extracted from isolated nuclei but were much less apparent in whole-cell extracts (10 µg protein each). The observed fragments cross-reacted with both C1 (not shown) and C2 antibodies, and exhibited a similar mobility to recombinant GFP-C0C1C2 protein (arrowhead; 2 µg protein loaded).

Figure 7.

Detection of NR1 C-terminal fragments in whole-cell and nuclear extracts of PC12 and HEK293 cells. Whole-cell (WC) and nuclear (N) extracts were prepared from HEK293 cells (a,b) or PC12 cells (c) expressing NR1-GFP fusion proteins. Following electrophoresis and membrane transfer, proteins were detected using C1 (a) and C2 (b,c) specific antibodies. Immunoreactivity for NR1-C0C1C2-GFP, GFP-C0C1C2, and C-terminal cleavage products are indicated by arrow, arrowhead and brackets, respectively.

Whole-cell lysates of PC12 cells stably expressing NR1-C0C1C2-GFP were also shown to exhibit peptide fragments (bracket) immunoreactive for both C1 (not shown) and C2 antibodies (Fig. 7c). Whole-cell lysates faintly exhibited these fragments, but the bands were enriched in protein extracted from cell nuclei (10 µg protein each). Recombinant GFP-C0C1C2 protein is included for comparison (arrowhead; 2 µg). Similar results were observed in transfected PC12 cells (data not shown). Whole-cell lysates or nuclear extracts of PC12 cells did not have detectable low molecular weight fragments that might correspond to the immunoreactive bands (tagged with GFP) generated from NR1-C0C1C2-GFP. However, NR1 expression levels were a small fraction of the level of NR1-C0C1C2-GFP expressed from the stable PC12s, which were themselves expressed many times below the levels of NR1-C0C1C2-GFP expressed in 293 cells.

Analysis of potential interactions between the recombinant NR1-1 C-terminal domain and NR2A

Differentiated PC12 cells and sensory neurons of the dorsal root ganglia both express NR2 subunits but the levels are comparatively low in comparison to native NR1 subunits (Huh and Wenthold 1999) and especially in comparison to the over-expressed NR1 C-terminal constructs. Thus, to determine if more equivalent levels of NR2 could affect the nuclear localization of the NR1-1 C-terminal constructs, experiments were carried out employing transfected HEK293 cells stably expressing NR2A fused to the enhanced YFP (NR2A-YFP). Normally these cells show a relatively even distribution of YFP fluorescence in the cytoplasm with little or no fluorescent puncta or patchiness near the plasma or nuclear membranes (Fig. 8e, arrow). These cells were transfected with a NR1-1 C-terminus similar to pGFP-C0C1C2 described in Fig. 1, except that it differed by the exchange of GFP for the enhanced CFP. The cyan fluorescence (pseudo-colored red) of CFP-C0C1C2 was primarily localized to the nucleus (Fig. 8a), whereas the yellow fluorescence of NR2A-YFP was maintained in the cytoplasm (Fig. 8b). Thus no appreciable alteration in their distribution occurred, as they did not co-localize (Fig. 8c). This is in contrast to the pattern observed when the full-length NR1-1 fused to CFP (NR1-CFP-C0C1C2) was transfected into the stably transfected NR2A-YFP-HEK293 cells (Figs 8d–f). The resulting coexpression resulted in co-localization of the NR2A-YFP with NR1-CFP-C0C1C2 in numerous puncta (cell indicated by arrowhead, Figs 8d–f). NR2A-YFP-HEK293 cells not transiently transfected with NR1-CFP-C0C1C2 exhibited a diffuse fluorescence throughout the cytoplasm (cell indicated by arrow, Figs 8e and f).

Figure 8.

Analysis of potential interactions between the NR1-1 C-terminus and NR2A. pCFP-C0C1C2 was transiently transfected into HEK293 cells stably expressing NR2A-EYFP. The localization of the CFPC0C1C2 and NR2A-YFP proteins were evaluated after 72 h using conventional fluorescence microscopy employing the specific filter sets for ECFP (a,d) and EYFP (b,e). A combined overlay of CFP and YFP fluorescence is also shown (c,f). The arrow in panels (d–f) indicates an NR2A-YFP HEK293 cell not transfected with pNR1-CFP-C0C1C2. The arrowhead in panels (d–f) indicates an NR2A-YFP HEK293 cell transfected with pNR1-CFP-C0C1C2. For ease of viewing, all ECFP images have been pseudo-colored red. All scale bars represent 20 microns and refer to all the panels in the same row.

Analysis of NR1-1 C-terminal domain primary and secondary structure

In an effort to understand the presence of the NLS within the C-terminus of NR1-1, a computer-aided search of the C0, C1 and C2 cassettes for amino acid sequence motifs associated with nuclear proteins was conducted. A number of motifs associated with nuclear proteins (homeobox domains, transcription factors and DNA binding proteins) were identified but these motifs generally exhibited a low to very low degree of homology. However, a region overlapping the C0 and C1 domains was found to resemble the consensus motif for a myc-type helix–loop–helix (HLH) domain. An alignment of several myc-type HLH protein domains with the C0 and C1 cassette amino acid sequence was performed (Fig. 9). The resultant alignment demonstrated significant similarity between the C0-C1 region and the basic region that precedes the first amphipathic helix of myc-type HLH domain and at several residue sites containing key hydrophobic amino acids. The degree of similarity was less apparent within the second helix region, especially for those myc-type HLH proteins that lacked a basic region at the end of the second helix. An identity with a key isoleucine (position 63, Fig. 9) was maintained but in the next position that contained a highly conserved leucine (position 64, Fig. 9) a threonine was found.

Figure 9.

Amino acid alignment between the NR1-1 C0C1 cassettes and the HLH domain of the myc-type family of transcription factors. Vertical open boxes indicate important conserved amino acids and the shaded boxes indicate the positions of basic amino acids in myc-type HLH proteins. The underlined regions represent the general location of where the first and second helices of the HLH motif are likely to be found. The accession numbers for the sequences employed in this analysis are as follows: SCK1, CAA70217; Myc 7E, AAD15818; vMyc, AAA62582; USF1, AAA79690; vMyc; P12523; c-Myc, P52160; HELHELA, AAA18517; c-Myc, P06171; USF1, A43899; mTFE3, AAD48781; AP4, AAF44993; hE47hlh, AAC41693; HES5, NP_034549; SREBP2, NP_004590; Mxi1, NP_005953; Hesr1, AAD46771; Hes1, NP_005515; ARNT2 hlh, AAD09750; BMAL1, BAA76414; ISF1, T04073; HER4 hlh, CAA65997.

HLH domains have a specific secondary structure that is required for their function. Four different computer algorithms were used to predict the structure of two model myc-type HLH proteins (Figs 10a and b) and the NR1-1 C-terminus (Fig. 10c). These models predict that the C0-C1 region contains the requisite secondary structure elements of HLH proteins; namely two amphipathic helices separated by a loop/turn region. In Figs 9a and b the same analysis was carried out for the well-described HLH protein e47 (Voronova and Baltimore 1990; Zhang and Bina 1991) and human Mxi 1 (Zervos et al. 1993). The latter, like the NR1-1 C0-C1 putative HLH domain, has a basic motif at its C-terminal end. Thus, these computer-aided sequence analyses revealed that at both the primary and secondary structure levels the NR1-1 C0C1 region contains a putative HLH motif similar to that observed in known transcription factors (Kingston 1989; Voronova and Baltimore 1990).

Figure 10.

Computer-aided secondary structure analysis of the amino acids found in the NR1 C0C1 cassettes. Computer algorithms were used to analyze the secondary structural motifs of two prototype myc-type HLH domains [the human e47 transcription factor (Zhang and Bina 1991), panel a; human Mxi 1 (Zervos et al. 1993), panel b] and the NR1-1 C0C1 region (c). (a) The e47 helix I and helix II (according to the scheme of Voronova and Baltimore 1990) of the e47 helix–loop–helix domain are boxed, and the N-terminal basic, DNA-binding region is overlined. (b) The basic regions of the human Mxi 1 HLH domain that would flank the helix I and helix II regions are overlined. (c) Within the C0C1 region of NR1-1 the C0 cassette is unshaded and the C1 cassette is shaded. The basic regions that flank putative helical regions are overlined. A region bearing similarity to the consensus motif for the second helix of a helix-loop-helix domain as determined using the PROSITE database (Bucher and Bairoch 1994; Hofmann et al. 1999) is underlined in gray. Alpha-helical regions predicted by secondary structure algorithms within the known and putative HLH domains are indicated as black bars underneath the appropriate amino acid residues. The utilized programs were the multivariate linear regression combination (MLRC; Guermeur et al. 1999), the hierarchical neural network (HNN; Y. Guermeur, PhD thesis), and Predator (Frishman and Argos 1996). The known and predicted HLH domains were further analysed using a Type-II Discrete State-space Model from the Protein Sequence Analysis (PSA) server at the Biomolecular Engineering Research Center (Stultz et al. 1993, 1997; White et al. 1994). The PSA displayed plots exhibit the calculated secondary structural features of each amino acid residue in terms of probability, with each isobar representing a probability value of 0.1.


The NR1-1 C1 cassette contains a functional bi-partite nuclear localization sequence

We have demonstrated that a putative bi-partite NLS, present in the C1 cassette, is capable of mediating nuclear localization of the NR1-1 C-terminus in cultured cells and in explants of cultured rat dorsal root ganglia. Although we did not specifically test the C-terminus of NR1-3, we would expect that the NR1-3 C-terminus (C0C1C2′) would also localize in the nucleus due to the presence of the C1 cassette. A similar NLS has recently been described for a naturally occurring mutant form of human SKCa3, a Ca2+-activated K+ channel protein expressed in neurons (Miller et al. 2001). The NLS within this SKCa3 mutant imparts a very similar rapid and exclusive nuclear localization with a diffuse distribution thoughout the nucleus in transient transfection assays, as was observed for C0C1C2.

When the bi-partite NLS located in the C1 cassette was mutated, it no longer mediated active transport to the nucleus when placed in the NLS vectors used in this study. The other putative NLS in the C0 cassette do not appear capable of mediating nuclear localization of the NR1-2 or NR1-4 C-terminal cassettes in the NLS vectors. However, these putative NLS may contribute to the overall efficiency of the nuclear localization observed, as the entire C-terminal cassette of NR1-1 when placed within the NLS vector was almost exclusively located in the nucleus of transfected cells. When NR1-1 C-terminal cassettes were removed, reducing the sequence to only the C1 cassette or the NLS motif itself, the efficiency of localization to the nucleus appeared to be reduced. At present, it is not clear whether this is due to a decrease in efficiency of active nuclear import or to a reduction in nuclear retention, once transported to the nucleus. The latter may be due to the inability of these fusion proteins to interact with other nuclear proteins via the putative HLH domain, an interaction that would require the presence of both the C0 and the C1 cassettes.

Possible physiological significance of the C1 NLS

The physiological role of the NLS in the C1 cassette of NR1 remains to be elucidated. For the NR1 cytoplasmic C-terminal domain to have a physiological role, the portion of the protein that contains the NLS would have to be proteolytically liberated from the nearby transmembrane region of the NR1 subunit. Indeed, in this study we were able to detect appropriately sized peptide fragments displaying immunoreactivity against NR1 C-terminal antibodies in cells expressing NR1-1-GFP fusion proteins. Furthermore, these immunoreactive fragments were significantly enriched in nuclear extracts of 293 and PC12 cells, suggesting the potential for such cleavage and nuclear targeting to occur.

These fragments were observed regardless of the placement of GFP in the full-length NR1. Placement of GFP immediately after the fourth transmembrane region of NR1 led to the generation of cleavage products which were similar in size or slightly smaller than GFP-C0C1C2. This suggests that the proteolytic machinery that leads to the liberation of these fragments is targeted upstream or within the fourth transmembrane region, as only this type of cleavage could generate a peptide fragment that still contained C1 and C2 but was similar in size to GFP-C0C1C2. The only alternative explanation would be that the observed cleavages occurred within the GFP moiety, but if this were true, NR1-C0C1C2-GFP would not be expected to yield peptide fragments. However, peptide fragments immunoreactive for both C1 and C2 were observed.

The concept that an integral membrane protein can be cleaved, releasing an intracytoplasmic fragment that is transported to the nucleus to mediate a change in gene regulation, is not generally unique, but it is novel for the glutamate ionotropic receptors. As recently reviewed, several bacterial and mammalian transmembrane proteins are cleaved within the plasma or organelle membranes, releasing into the cytoplasm proteolytic fragments that contain NLS motifs that direct them to the nucleus and regulate gene transcription, a process now termed regulated intramembrane proteolysis or Rip (Brown et al. 2000). Notch is the best-characterized member of this family of proteins (Struhl and Adachi 1998).

Interestingly, expression of full-length Notch protein does not lead to the generation of high levels of the Notch intracellular domain (NICD). In fact, the generation of the NICD was not initially detectable (Fehon et al. 1991; Lieber et al. 1993; Rebay et al. 1993). Instead, a Gal4-VP16 transcription factor was placed within the NICD and was shown to transactivate a reporter gene, which demonstrated that the NICD was gaining access to the nucleus in response to stimulation with Delta (Struhl and Adachi 1998). Yet even these minute levels of NICD are physiologically relevant. Likewise, the amyloid precursor protein (APP), which is the other well-characterized neuronal Rip substrate, is cleaved to generate levels of the APP intracellular domain (AICD) that are barely detectable by western blot in overexpressing heterologous cells (Yu et al. 2001). The study of the NICD and AICD has been facilitated by using recombinant proteins consisting of either the intracellular domain alone, or in concert with the transmembrane region and small portions of the extracellular domain; a substrate particularly well suited for cleavage by presenilins (reviewed in Brown et al. 2000).

Along these lines, it would be surprising for Rip of NR1-1 to lead to large levels of the liberated intracellular domain. Indeed, GFP fluorescence is not observed in nuclei upon expression of NR1-C0C1C2-GFP or NR1-GFP-C0C1C2, and the C1 and C2 immunoreactive peptides are only observable upon overexpression of recombinant NR1-GFP fusion proteins. Likewise, endogenous expression of NR1 from PC12 cells did not yield detectable levels of equivalent immunoreactive peptides, which was likely exacerbated by the fact that endogenous expression of full-length NR1 was a small fraction of the expression of the recombinant fusion proteins. As with Notch, a genetic approach will likely be required to confirm or eliminate the possibility that physiologically relevant levels of the NR1 intracellular domain are generated by Rip.

The fact that not all of the NR1-1 subunit splice variants contain an NLS is not unique as the γ form of the double C2 protein contains an NLS while the membrane-associated α and β double C2 spliced variants do not, suggesting that the NLS imparts a novel although unknown function (Fukuda et al. 2001). The proteolytic cleavage of nuclear membrane-associated pro-interleukin-16 (pro-IL-16) by caspase 3 yields a protein fragment that translocates to the nucleus via a NLS sequence. The pro-IL-16 NLS protein fragment appears to function only after cleavage has occurred and nuclear accumulation is associated with growth arrest (Zhang et al. 2001). The cytoplasmic domain of neuregulin has also been shown to undergo translocation to the nucleus via a monopartite NLS in both cell lines and primary neurons (Bao et al. 1999). Neuregulins comprise a family of epidermal growth factor-related growth factors found in the central and peripheral nervous systems (Gassmann and Lemke 1997).

A search of the literature revealed that the C-terminus of the NR1 subunit is cleaved and liberated from the membrane fraction of normal murine forebrain and cerebellum in vivo (Chazot and Stephenson 1997). A cleavage of the NR1 C-terminus also appears to be induced by the extracellular protease thrombin (Gingrich et al. 2000). Thrombin is a protease that is most associated with the coagulation cascade (Goldsack et al. 1998) and it also has a well-established role in the CNS (Turgeon and Houenou 1997) and is present in high amounts after CNS trauma (Gingrich et al. 2000). The observed cleavage appeared to liberate a 12-kDa fragment of the NR1-1 C-terminus, which corresponds to the predicted weight of the intact C0, C1, and C2 cassettes. This would suggest that thrombin was directing cleavage to the juxtamembrane region of the extracellular fourth transmembrane domain of NR1-1. However, the authors of this study were unable to conclude that cleavage occurred within the intracellular space (Gingrich et al. 2000). Since there is no consensus thrombin cleavage site (according to the scheme of Backes et al. 2000) in either the intracellular, transmembrane or extracellular juxtamembrane region of NR1, thrombin may not be acting directly on NR1-1. Also, if the observed cleavage were truly intracellular, then thrombin must transmit an activating signal through the plasma membrane to direct the cleavage of NR1-1. Although there are several G protein-coupled receptors for thrombin (Coughlin 1999) that could possibly mediate such a signal, Gingrich et al. (2000) demonstrated convincingly that the observed thrombin-mediated cleavage of NR1-1 was independent of the best-characterized thrombin receptor, the protease activated receptor-1 (PAR1).

Interestingly, mice that overexpress the thrombin inhibitor protease nexin-1 (PN-1) exhibit enhanced NMDAR synaptic transmission, and PN-1 knockout mice manifest a reduction in NMDAR transmission that might be associated with reductions in the number of functional receptors (Luthi et al. 1997). As well, presenilin-1 knockout mice are more susceptible to glutamate excitotoxicity (Guo et al. 1999) and glutamate excitotoxicity appears to play a prominent role in Alzheimer disease pathogenesis (Guo et al. 1999). All of these data are consistent with a role for cellular proteases in directing Rip of NR1-1, although other explanations for these results are possible. Tissue plasminogen activator has also been recently demonstrated to interact with NR1-1, resulting in its cleavage and potentiation of NMDA receptor-mediated excitotoxicity (Nicole et al. 2000).

The physiological role of the NR1-1 C-terminal fragment as well as many of the newly described NLS motifs in membrane-associated proteins and proteolytic fragments remains to be determined (Healy et al. 1999; Craggs and Kellie 2001; Fukuda et al. 2001; Miller et al. 2001). The NR1-1 C1 cassette NLS may regulate expression and/or transcription of NR1 and NR2 or affect the regulation of expression of other genes. The C-terminus contains regions that resemble the HLH secondary structure of the myc family of transcription factors. It is interesting to note that one of the SREBP proteins (human SREBP2), a known member of the Rip family of proteins, also has a myc-type HLH domain, homologous to that of the putative HLH domain present in the C0C1 region of NR1-1, and appears to be involved in regulating gene transcription (Hua et al. 1993). This also raises the possibility that the cleavage of the NR1-1 C-terminus may occur at intracytoplasmic membranes. An endoplasmic reticulum retention sequence has recently been demonstrated to exist within the C1 cassette (Standley et al. 2000). Future experiments will be required to determine the significance of the NLS and the putative myc-type HLH protein–protein structural motif in the cytoplasmic C-terminal domain of NR1-1. Two calmodulin binding domains span these same C0C1 regions (Ehlers et al. 1996) and NMDA receptor activation induces translocation of calmodulin to the nucleus in a facilitated manner where it interacts with CaMKIV to alter the phosphorylation of the CREB transcription factor (Deisseroth et al. 1998). The binding of calmodulin to the NR1-1 C-terminus, and subsequent cleavage, may represent a potential mechanism by which calmodulin is targeted to the nucleus through interaction with a released C0C1 domain. A proteolyticly released NR1-1 C-terminal fragment may have a role in the development of the nervous system and in known NMDAR-associated functions such as learning and memory. Alternatively, the proteolytic cleavage of NR1-1, released as a consequence of the presence of abnormal levels of proteases such as thrombin after a CNS trauma, may yield proteolytic fragments with dominant negative effects that could contribute to glutamate-mediated excitotoxicity. We have already demonstrated that the expression of the NR1-1 C-terminus in DRG neurons results in a dominant negative effect that blocks membrane placement of NR1-1 subunits and, in a significant proportion of DRG neurons, induced apoptotic-like nuclear fragmentation (Marsh et al. 2001). This dominant negative effect is similar to that recently described for the mutant form of human SKCa3 (Miller et al. 2001). Together these results suggest that, under the appropriate circumstances, a released NR1-1 C-terminal peptide could have important physiological consequences.