Neuropilin tolloid-like 1 (Neto1), is a CUB domain-containing transmembrane protein that was recently identified as a novel component of the NMDA receptor complex. Here, we have investigated the possible association of Neto1 with the amyloid precursor protein (APP)695/GluN1/GluN2A and APP695/GluN1/GluN2B NMDA receptor trafficking complexes that we have previously identified. Neto1HA was shown to co-immunoprecipitate with assembled NMDA receptors via GluN2A or GluN2B subunits; Neto1HA did not co-immunoprecipitate APP695FLAG. Co-immunoprecipitations from mammalian cells co-transfected with APP695FLAG, Neto1HA and GluN1/GluN2A or GluN1/GluN2B revealed that all four proteins co-exist within one macromolecular complex. Immunoprecipitations from native brain tissue similarly revealed the existence of a GluN1/GluN2A or GluN2B/APP/Neto1 complex. Neto1HA caused a reduction in the surface expression of both NMDA receptor subtypes, but had no effect on APP695FLAG- or PSD-95αc-Myc enhanced surface receptor expression. The Neto1 binding domain of GluN2A was mapped using GluN1/GluN2A chimeras and GluN2A truncation constructs. The extracellular GluN2A domain does not contribute to association with Neto1HA but deletion of the intracellular tail resulted in a loss of Neto-1HA co-immunoprecipitation which was paralleled by a loss of association between GluN2A and SAP102. Thus, Neto1 is concluded to be a component of APP/NMDA receptor trafficking complexes.
Neto1 is shown to be a constituent of APP/NMDA receptor trafficking complexes. Since Neto1 is shown to associate with GluN2 expressed alone and, APP with GluN1-1a alone, it is proposed that trafficking complexes are formed in the endoplasmic reticulum and contribute to regulation of NMDA receptor surface expression. Alternative scenarios are depicted since it is to be determined if respective associations are direct.
Neuropilin tolloid-like 1 (Neto1) and Neuropilin tolloid-like 2 (Neto2) are complement C1r/C1s, Uegf, Bmp1 (CUB) domain-containing transmembrane proteins that were recently identified as novel components of glutamatergic neurotransmitter receptor complexes. Neto1 and Neto2 have both been described as auxiliary subunits of kainate receptors since they were shown to alter the trafficking, channel kinetics and pharmacology of this glutamate receptor subtype in a subunit-dependent manner (Zhang et al. 2009; Copits et al. 2011; Straub et al. 2011a, b; Tang et al. 2011; Fisher and Mott 2012; Straub et al., 2011b). In C.elegans, SOL-1 and SOL-2, species orthologues of Neto, were shown to be auxiliary subunits of GLR-1, 2-amino-3-(3-hydroxy-5-methyl-isoxazol-4-yl)propanoic acid (AMPA) glutamatergic receptors (Wang et al. 2012). In addition, Neto1 was shown to be a novel component of the N-methyl-d-aspartate (NMDA) receptor complex that was critical for maintaining the abundance of GluN2A receptors within the post-synaptic density (Ng et al. 2009). Neto1 was enriched in the post-synaptic density where it was shown to co-distribute with the scaffold protein, post-synaptic density-95 (PSD-95), and GluN1 (Ng et al. 2009). Neto1 (−/−) mice were found to have deficits in long-term potentiation and NMDA receptor-dependent spatial learning and memory (Ng et al. 2009).
We have previously shown (Cousins et al. 2009), in agreement with Hoe et al. (2009), that amyloid precursor protein (APP) co-immunoprecipitates with assembled GluN1/GluN2 NMDA receptors. Over-expression of the neuronal splice variant of APP, APP695, results in an increase in NMDA receptor cell surface expression in cell lines (Cousins et al. 2009). Similarly, over-expression of APP770 resulted in an enhanced NMDA receptor expression in primary cultures of neuronal cells (Hoe et al. 2009). Both these findings are compatible with an ascribed role for APP as a mediator of neuronal trafficking mechanisms possibly functioning as an NMDA receptor auxiliary protein. Since Neto1 was reported to be associated with NMDA receptors and crucially, to play a role in NMDA receptor-mediated learning and memory, it was of interest to determine if Neto1 is a constituent of NMDA receptor/APP macromolecular complexes. We have carried out an investigation to determine if this is the case and we report the findings in this article.
Constructs and antibodies
The constructs, pCISGluN1-1a, pCISGluN2A, pCISGluN2B, pCISGluN2BFLAG and pCI-neoAPP695FLAG were as previously described (Cik et al. 1993; Hawkins et al. 1999; Cousins et al. 2009). DNA encoding GluN2A 838-1464 was cloned into the BamHI/EcoRI restriction enzyme sites of pcDNA4 to generate pcDNA4GluN2A 838-1464 yielding the GluN2A C-terminal domain with an N-terminal His tag. Note that only the GluN1-1a splice variant was used thus pCISGluN1 and GluN1 are used throughout the manuscript to denote the GluNR1-1a splice variant. DNA encoding Neto1 and Neto 1-372 were cloned via EcoRV/Sal1 restriction enzyme sites of pCMV-4a to generate pCMVNeto1 and pCMVNeto1 1-372 yielding full length Neto1 and Neto1 1-372 with a C-terminal FLAG tag. pGW1PSD-95αc-Myc and pCMVneoSAP102c-Myc were kind gifts of Dr M. Sheng (Genentech Inc., San Francisco, CA, USA). pcDNA3.1Neto1HA was a kind gift from Drs R.R. McInnes (McGill University, Montreal, Canada) and M.W. Salter (University of Toronto, Toronto, Canada) and pCISGluN2A 1-877 a kind gift of Professor G Hardingham (University of Edinburgh).
Anti-GluN1 C2, anti-GluN2A (44–58), anti-GluN2A (1381–1394), anti-GluN2B (46–60) and anti-APP (679–695) antibodies were raised, generated and affinity-purified as previously described (Chazot et al. 1992; Hawkins et al. 1999; Groc et al. 2006; Cousins et al. 2009). Anti-Neto1 (447-533) antibodies were as in Chow et al. (2004) and were a kind gift from Drs R.R. McInnes (McGill University) and M. W. Salter (University of Toronto). Anti-HA and anti-synapse-associated protein 102 (SAP102) antibodies were from Abcam (Cambridge, UK); anti-FLAG M2 antibodies were from Sigma-Aldrich (Dorset, UK). Anti-GluN2B antibodies which were raised from a 30 kDa GluN2B C-terminal fusion protein were from Millipore (Billerica, MA, USA). Anti-His antibodies were from Life Technologies Ltd (Paisley, Scotland).
Generation of GluN1/GluN2A NMDA receptor chimeric subunits
Two sets of GluN1/GluN2A chimeras were generated by overlap extension PCR. For the first set, the respective N-terminal regions were swapped such that (GluN1)/GluN2A contained GluN1 1-561 followed by GluN2A 557-1464 and (GluN2A)GluN1 contained GluN2A 1-556 followed by GluN1 562-938. For the second set, the extracellular S2 domains of GluN1 and GluN2A were exchanged. Thus GluN1 (GluN2A S2)/GluN1 contained GluN2A 648-816 and GluN2A (GluN1 S2)/GluN2A contained GluN1 650-812.
Mammalian cell transfections
Human embryonic kidney (HEK) 293 cells were transfected using the calcium phosphate method as described (Cik et al. 1993). For double transfections where one clone was Neto1, 10 μg DNA were used at a 1 : 1 ratio, for example, 5 μg Neto1HA: 5 μg GluN1. For triple transfections a total of 20 μg DNA was used with a ratio of 1 : 3 GluN1 : GluN2A (10 μg total) with 10 μg Neto1HA, 10 μg APP695FLAG or 10 μg PSD-95αc-Myc. For quadruple transfections, a total of 30 μg DNA was used with a ratio of 1 : 3 GluN1 : GluN2A (10 μg total) with 10 μg Neto1HA and 10 μg APP695FLAG. HEK 293 cells expressing GluN1/GluN2 were cultured post-transfection in the presence of 1 mM ketamine to prevent cell cytotoxicity. Transfected cells were incubated for 24 h post-transfection and either assayed for cell surface NMDA receptor expression or they were harvested and used in immunoprecipitation assays.
Immunoprecipitations were carried out as previously described from either 1% (v/v) Triton X-100 100 000 g detergent extracts of HEK 293 transfected cells or 1% (w/v) sodium deoxycholate 100 000 g extracts of adult rat brain (Cousins et al. 2009). Immunoprecipitating antibodies used were anti-GluN1 C2, anti-GluN2A (1381–1394), anti-GluN2B (46–60), anti-APP (679–695), anti-His, anti-HA, anti-FLAG or non-immune Ig antibodies. The resulting immune pellets were analysed by sodium dodecyl sulphate–polyacrylamide gel electrophoresis followed by immunoblotting.
Immunoblotting was carried out using 7.5% (w/v) polyacrylamide slab mini-gels all as described previously (Papadakis et al. 2004). Primary antibodies used were as follows: anti-GluN1 C2; anti-GluN2A (1381–1394); anti-APP (679–695); anti-Neto1 (447-533), anti-GluN2B, anti-SAP102, anti-FLAG and anti-HA. Rabbit or mouse horseradish-linked secondary antibodies were used at a final dilution of 1 : 2000 and immunoreactivities were detected using the ECL Western blotting system (Amersham Biosciences Ltd., Little Chalfont, Bucks., UK). Immunoblots with brain tissue used the above secondary antibodies at a 1 : 10 000 dilution and subsequent development using the SuperSignal West Femto ECL substrate (Perbio Science UK Ltd., Northumberland, UK). For all the immunoblots of the immunoprecipitates, for double transfections, 50% of immune pellets were analysed by the precipitating primary antibody and 50% with the second different specificity antibody; for triple transfections and in brain samples where three different antibodies were used for immunoblotting of immune pellets, 14% of the immune pellets were analysed by the precipitating primary antibody and 42% of the pellets for each of the additional two specificity antibodies and for quadruple transfections and in brain where four different antibodies were used for immunoblotting of immune pellets, 12% of the immune pellets were analysed by the precipitating primary antibody, 25% of the pellets for two additional specificity antibodies and always 38% for anti-Neto1 activity.
Determination of NMDA receptor cell surface expression by ELISA
The surface expression of GluN1/GluN2A and GluN1/GluN2B receptors was measured using anti-GluN2A (44–58) and anti-GluN2B (46–60) antibodies, respectively, since NMDA receptors are only expressed at the cell surface when they are co-assembled as GluN1/GluN2A or GluN1/GluN2B receptors (McIlhinney et al. 1998). Cell surface ELISA was carried out as previously described (Papadakis et al. 2004; Cousins et al. 2008).
Analysis of results
All results are the means ± SEM and were analysed by the Student's two-tailed paired t-test.
Results and Discussion
Neto1HA co-immunoprecipitates with assembled heteromeric NMDA receptors via GluN2
The association of NMDA receptors with Neto1 is controversial. Ng et al. (2009) reported that anti-Neto1 antibodies immunoprecipitated GluN1, GluN2A and GluN2B from radio-immunoprecipitation assay buffer (a buffer containing the detergents, 1% Non-idet P-40, 0.1% sodium dodecyl sulfate and 0.5% deoxycholate)-solubilized extracts of synaptosomes. Co-immunoprecipitation was independent of PSD-95 since all three GluNs co-immunoprecipitated with a Neto1 construct, Neto1-Δ20HA, which lacked an intracellular domain and also did not co-immunoprecipitate with PSD-95 (Ng et al. 2009). In contrast, Straub et al. (2011a, b) using 1% Triton to solubilize rat brain lysates could not detect GluN1 or indeed PSD-95 in anti-Neto1 or anti-Neto2 immune pellets thus inferring that since all NMDA receptors contain GluN1, Neto1 and Neto2 do not associate with NMDA receptors. To resolve these different observations, we carried out immunoprecipitations from detergent extracts of HEK 293 cells transfected with the single clones GluN1, GluN2A, GluN2B; GluN1/GluN2A and GluN1/GluN2B binary combinations and 1% sodium deoxycholate extracts of rat brain. The results are shown in Figs 1-3.
First, for single subunit transfections, Neto1HA was co-immunoprecipitated by anti-GluN2A and anti-GluN2B but not anti-GluN1 antibodies indicating, in agreement with Ng et al. (2009) that Neto1HA associates with GluN2A and GluN2B subunits (Fig. 1). Since Neto1HA does not associate with GluN1, in binary transfections, immunoprecipitations were carried out with anti-GluN1 antibodies thus if Neto1 and GluN2A or GluN2B were detected in immune pellets, it must mean that Neto1HA associates with assembled GluN1/GluN2A and GluN1/GluN2B NMDA receptors. The results in Fig. 2 show that this is indeed the case with GluN1, GluN2A or GluN2B and Neto1HA being found in all immune but not in control, non-immune pellets. Finally, anti-GluN2A antibodies immunoprecipitated Neto1 from detergent extracts of rat brain (Fig. 3). Thus, these results concur with those of Ng et al. (2009) in that at least Neto1 associates with GluN2-containing NMDA receptors. The differences between the two original reports may be explained by the detergent extraction conditions. NMDA receptors are notoriously difficult to solubilize from native tissue because of their embedding within the post-synaptic density. It was previously shown that 1% Triton, the conditions employed by Straub et al. (2011a, b), solubilizes a pool of unassembled GluN1 subunits (Chazot and Stephenson 1997) hence the detection of GluN1 in brain lysates but not in immune pellets because of the lack of soluble GluN2 subunits. However, harsher detergent conditions, that is, 1% sodium deoxycholate as used here or radio-immunoprecipiation assay buffer as in Ng et al. (2009) are required to solubilize GluN1/GluN2 heteromers. Since association with Neto1 is via GluN2, co-association is only detected in immune pellets from brain extracts solubilized with these detergents.
It should also be noted that here, immunoprecipitations were carried out from 100 000 g detergent extracts. Ng et al. (2009) used a 14 000 g supernatant which is not a bona fide soluble preparation, nevertheless, our findings concur.
GluN1/GluN2 assembled NMDA receptors form ternary complexes with APP and Neto1
To study the possible association of Neto1 with NMDA receptor/APP complexes first, it was investigated if APP695FLAG could associate with Neto1HA. Both were co-expressed in HEK 293 cells and immunoprecipitations carried out using anti-APP antibodies. APP695FLAG was found in immune pellets but Neto1 was not detected thus Neto1HA does not co-immunoprecipitate with APP695FLAG (Fig. 4). To determine if Neto1HA associates with NMDA receptor/APP695FLAG complexes, GluN1, GluN2A or GluN2B, APP695FLAG and Neto1HA were co-expressed in HEK 293 cells and immunoprecipitations carried out with anti-APP antibodies. Immunoreactivities for all four expressed proteins were detected in immune but not in non-immune pellets (Fig. 5). Since APP interacts with only GluN1; APP does not interact with GluN2 or Neto1 alone; GluN1 interacts with GluN2 and all four proteins are found in the immune pellet, it must mean that they are all co-associated in one macromolecular complex, that is, APP695FLAG/GluN1/GluN2/Neto1HA (Fig. 5).
To determine if the quaternary complex is found in native tissue, anti-APP immunoprecipitations were carried out from detergent extracts of adult rat brain (Fig. 6). GluN2A, GluN2B and Neto1 immunoreactivities were all detected in immune but not in control, non-immune pellets thus showing that all four proteins can be found in one complex. Further, this is the case for both GluN2A- and GluN2B-containing NMDA receptors (Fig. 6). Note that the existence of the APP/GluN1/GluN2/Neto1 quaternary complex is demonstrated by co-immunoprecipitation, although these experiments show that the four proteins are associated they do not show that they each interact directly. Association between different pairs may be because of an intermediary accessory protein as illustrated in Fig. 5.
The effect of Neto1HA on the surface expression of NMDA receptors
We previously showed that APP695FLAG enhanced the surface expression of both GluN1/GluN2A and GluN1/GluN2B NMDA receptors. Since we showed that APP and Neto1 are both part of the NMDA receptor macromolecular complex, it was of interest to determine the effect of Neto1 on NMDA receptor surface expression and APP695FLAG-enhanced NMDA receptor surface expression. Thus in parallel, GluN1/GluN2A and GluN1/GluN2B were both co-expressed with either PSD-95αc-Myc; SAP102c-Myc; APP695FLAG; Neto1HA; PSD-95αc-Myc + Neto1HA or APP695FLAG + Neto1HA and surface expression was determined 24 h post-transfection. The results are summarized in Fig. 7.
As previously reported, PSD-95αc-Myc and APP695FLAG both enhanced GluN1/GluN2A and GluN1/GluN2B surface expression; SAP102c-Myc had no effect (Rutter et al. 2002; Cousins et al. 2008, 2009). Interestingly, Neto1HA resulted in a significant decrease in surface expression for both GluN1/GluN2A (28 ± 5%, n = 13) and GluN1/GluN2B (30 ± 7%, n = 6) but with no change in total GluN1, GluN2A or GluN2B expression (Fig. 7). Neto1HA had no effect on PSD-95αc-Myc- or APP695FLAG-enhanced NMDA receptor expression (Fig. 7). Ng et al. (2009) reported that a lack of Neto1 as in Neto1 (−/−) mice does not alter the surface expression of GluN2A or GluN2B, rather a significant (~ 30%) decrease in GluN2A in the post-synaptic density was observed suggesting that Neto1 is involved in the targeting and/or stability of GluN2A-NMDA receptors at synapses. The findings here show that Neto1 can influence surface NMDA receptor expression although in the presence of PSD-95 as would be the case in vivo, the Neto1-induced decrease in surface expression is not apparent. A decrease in receptor cell surface expression with no change in total overall expression levels of GluN1, GluN2A and GluN2B subunits suggests that either Neto1 enhances receptor internalization or it can decrease receptor insertion into the plasma membrane. The presence and thus association of assembled GluN1/GluN2A receptors with PSD-95 may prevent these possibilities.
Intracellular GluN2A C-terminal domains mediate association of Neto1HA with NMDA receptors
Neto1 is a single transmembrane domain protein comprising of two extracellular CUB domains of ~ 110 amino acids, a low-density lipoprotein receptor domain class A motif and a cytoplasmic tail of ~ 167 amino acids that includes a C-terminal TRV class I PDZ binding motif (Fig. 8). Ng et al. (2009) showed that removal of the Neto1 cytoplasmic C-terminal tail and the transmembrane domain did not abolish Neto1/NMDA receptor interaction. Further, a Neto1 construct that contained only the signal sequence and the CUB1 domain retained the ability to associate with GluN2A; thus, it was concluded that the Neto1/NMDA receptor interaction is mediated via the extracellular Neto1 CUB1 domain. To determine where within the GluN2A extracellular domain Neto1 binds, a series of GluN1/GluN2A chimeras were generated, each was co-expressed with Neto1HA in HEK 293 cells and immunoprecipitations carried out. For the first generation chimeras, the GluN2A N-terminal LIVBP + the S1 domain were swapped for the corresponding regions of GluN1 and vice versa. For the second generation chimeras, GluN2A S2 was swapped for GluN1S2 domain and vice versa (Fig. 8). Surprisingly, it was found that replacing the GluN2A extracellular regions with GluN1 had no effect on the association with Neto1HA, that is, Neto1HA was always co-immunoprecipitated with constructs containing the GluN2A transmembrane and intracellular domains but not with constructs containing GluN2A extracellular domains (Fig. 8).
To investigate the association further, GluN1 + Neto1HA were co-expressed in parallel with a GluN2A C-terminal truncated construct, that is, GluN2A 1-877; with wild type, GluN2A and additionally, a tagged construct which encoded only the GluN2A intracellular domain, that is, GluN2A 838-1464. The results in Fig. 9 show that deletion of the GluN2A C-terminal domain results in the loss of co-immunoprecipitating Neto1HA. Conversely, Neto1HA is co-immunopreciptated with GluN2A 838-1464. This implies that it is the C-terminal GluN2A domain that mediates directly or indirectly, the association with Neto1HA. The reciprocal experiment was carried out whereby the Neto1 C-terminal tail was deleted 7 amino acids after the transmembrane domain, that is, Neto1 1 -372. Surprisingly, anti-FLAG immunoreactivity, that is, Neto1, was found in immune pellets for Neto1 1-372 co-expressed with full length GluN2A (Fig. 9).
As described above, Ng et al. (2009) found that Neto1 immunoprecipitates with PSD-95. SAP102 is a closely related scaffold protein that is endogenously expressed in HEK 293 cells (Sans et al. 2003; Lin et al. 2004). To investigate the possibility that it is SAP102 that mediates association with Neto1HA, Neto1HA was expressed alone in HEK 293 cells, immunoprecipitated and the immune pellet screened for Neto1HA and SAP102 immunoreactivity (Fig. 10). For these experiments, the sensitivity of the immunoblots was increased by using the SuperSignal West Femto ECL substrate and also analysing 80% of the immune pellet for SAP102. Under these conditions, SAP102 was detected in immune but not in control non-immune pellets (Fig. 10). Simultaneously, the GluN2A 1-877 truncation construct was co-expressed with GluN1 and SAP102; it was found that GluN2A 1-877 did not co-immunoprecipitate SAP102. Thus, there is a correlation between the capability of GluN2A to co-immunoprecipitate SAP102 and Neto1HA. However, SAP102 does not co-immunoprecipitate with Neto1FLAG 1-372 suggesting that GluN2A/Neto1 association is not mediated via SAP102.
In this article, we have identified a new player in the NMDA receptor/APP trafficking complexes. We have shown that Neto1 co-immunopreciptates with NMDA receptor/APP trafficking complexes via association with GluN2, in heterologous expression and in native brain tissue. Furthermore, we have shown that it is the GluN2A C-terminal domain that mediates Neto1 association. The original work of Ng et al. (2009) showed that Neto1 associated with assembled NMDA receptors via GluN2A and the Neto1 extracellular N-terminal CUB1 domain. We conclude that Neto1 associates with NMDA receptors via GluN2; however, our findings are difficult to reconcile with a direct association between GluN2 and Neto1 since the removal of the GluN2A extracellular domain has no effect on GluN2A/Neto1 co-immunoprecipitation. Rather, we demonstrate that it is the intracellular region of GluN2A that mediates association. Further, since Ng et al. (2009) showed that deletion of the C-terminal 20 amino acids of Neto1 and the results presented here which deleted a further 141 amino acids from the Neto1 intracellular domain showed that these truncated constructs retained the ability to co-immunoprecipitate with GluN2A, it is hard to visualize how association between the two proteins could be direct. Since it is endogenously expressed in HEK cells, SAP102 is a potential candidate for an intermediary protein since it co-immunoprecipitates with both Neto1 and GluN2A. However, SAP102 does not co-immunoprecipitate with Neto1FLAG 1-372 and thus such a possibility is excluded. Overall, these findings suggest that Neto1 is not an NMDA receptor auxillary subunit since the protein–protein interactions that we have determined imply that their association is indirect.
Nevertheless, a new player in NMDA receptor/APP trafficking complexes has been identified. Future studies will ascertain how APP and Neto1 interactions regulate the appropriate surface expression of this key neurotransmitter receptor.
We thank Professors R.R. McInnes (McGill University, Montreal, Canada) and M.W. Salter (University of Toronto, Toronto, Canada) for the clone, pcDNA3.1Neto1HA and anti-Neto1 (447-533) antibodies; Professors S. Nakanishi and M. Mishina for the gifts of the original NMDA receptor clones, Professor M. Sheng for pGW1PSD95αc-Myc, Professor Chris C.J. Miller (Institute of Psychiatry, King's College London, UK) for the gift of pCI-neoAPP695FLAG, Professor G Hardingham, University of Edinburgh for pCISGluN2A 1- 877 and Dr Kieran Brickley, UCL School of Pharmacy for help with antibody production. This work was funded by Alzheimer's Research, United Kingdom.