NR2A−/− Mice Lack Long-Term Potentiation But Retain NMDA Receptor and L-Type Ca2+ Channel-Dependent Long-Term Depression in the Juvenile Superior Colliculus

Whether the subunit composition of NMDA receptors (NMDARs) controls the direction of long-term plasticity is currently disputed. In the visual layers of NR2A−/− juvenile superior colliculus (SC), synapses lose miniature NMDAR currents, leaving NR2B-rich receptors in extrasynaptic regions. Compared with wild type (WT), evoked NMDAR currents in mutant neurons have slower rise and decay times and lower NMDAR/AMPAR current ratios. Moreover, NMDAR and L-type Ca2+ channel-dependent SC long-term potentiation (LTP) is absent in NR2A−/− cells, whereas both WT and mutant neurons show long-duration, low-frequency-induced, long-term depression (LLF-LTD) that is blocked by either AP-5, nimodipine, or Ro 25-6981 [R-(R,S)-α-(4-hydroxyphenyl)-β-methyl-4-(phenylmethyl)-1-piperidine propranol]. Thus, NMDAR currents or signaling localized at the postsynaptic density are essential to SC NMDAR-dependent LTP, whereas extrasynaptic or NR2B-rich NMDARs are necessary for LLF-LTD. However, synaptic NMDARs as well as the NR2A subunit are missing in NR2A−/− mice. Therefore, NR2 subunit-specific ligand binding/channel properties and/or separate signaling pathways interacting with NMDARs at synaptic versus extrasynaptic receptors could underlie these results.

The distribution of NMDARs is also important to their ability to induce LTP versus LTD. NR2A-rich receptors concentrate at subsynaptic sites, whereas NR2B-rich receptors are prominent in perisynaptic or extrasynaptic regions (Tovar and Westbrook, 1999;van Zundert et al., 2004;Thomas et al., 2006). Activation of subsynaptic NMDARs in hippocampal cultures induces LTP, whereas activation of extrasynaptic NMDARs produces LTD (Lu et al., 2001b). Pairing low-frequency stimulation with postsynaptic depolarization produces hippocampal LTP regardless of NR2A or NR2B activation (Berberich et al., 2005). However, pairing synaptic stimulation with pulses of depolarizing current is also reported to produce LTP even when NMDARs are blocked with AP-5 (Kullman et al., 1992).
In visual superior colliculus (SC) synapses of the NR2 Ϫ/Ϫ mouse, miniature NMDAR (mNMDAR) currents disappear before eye opening [postnatal day 13 (P13)], whereas extrasynaptic NR2B-NMDARs remain and mediate evoked NMDAR currents with prolonged rise times (Townsend et al., 2003). Mutant mice lacking the NR2A cytoplasmic domain show a similar decrease in synaptic NMDARs in hippocampal CA1 (Steigerwald et al., 2000). The NR2A cytoplasmic domain therefore appears necessary to bind NMDARs to the synapse, and their loss from that position in NR2A Ϫ/Ϫ mutants is consistent with the relatively poor binding of the remaining NR2B cytoplasmic tail to the mature scaffold postsynaptic density-95 (PSD-95) (Townsend et al., 2003). In rat SC, L-type Ca 2ϩ channels and NMDARs act synergistically to induce LTP (Zhao et al., 2006). Here we show that juvenile NR2A Ϫ/Ϫ mice (P15-P17) lack synaptic NMDARs and show no SC LTP but retain NR2B-NMDAR-and L-type Ca 2ϩ channel-dependent long-duration, low-frequency-induced, long-term depression (LLF-LTD). Blockade of NR2B-rich NMDARs eliminates LLF-LTP in both WT and NR2A Ϫ/Ϫ neurons. The data suggest that synaptic NMDARs are critical to SC LTP, whereas extrasynaptic NMDARs are critical to LLF-LTD but they cannot fully address the question of NMDAR subunit specificity in these responses.

Materials and Methods
Slice preparation. Wild-type (WT) C57BL/6 mice and NR2A Ϫ/Ϫ mice (Sakimura et al., 1995) (a gift from M. Mishina, University of Tokyo School of Medicine, Tokyo, Japan) with a C57BL/6 background were bred and maintained in Massachusetts Institute of Technology facilities. All animal procedures were in accord with approved Massachusetts Institute of Technology Animal Care and Use Committee protocols. Parasagittal SC slices were prepared from these mice at P15-P17 as described previously (Zhao et al., 2006).
Electrophysiology. All recordings used wholecell patch clamping from narrow field vertical neurons in the stratum griseum superficale of the SC. When plasticity was examined the artificial CSF (ACSF) and pipette solutions were as in the study of Zhao et al. (2006). For miniature and evoked current recording the pipette solution contained the following (in mM): 122.5 Csgluconate, 17.5 CsCl, 10 HEPES, 0.2 NaEGTA, 4 ATP-Mg, 0.4 GTP-Na, and 8 NaCl, pH adjusted to 7.3 with CsOH.
One cell per slice was used for evoked response recordings, one to two cells per slice for mEPSC recordings, and n is given as the number of experiments/the number of animals. Data are expressed as mean Ϯ SEM. Statistical significance (*p Ͻ 0.05 and **p Ͻ 0.01) was determined using Student's t and ANOVA tests as stated in the figure legends.

Results
Juvenile NR2A Ϫ/Ϫ SC neurons lose mNMDARcs Previous work documented loss of mNMDARcs in the NR2A Ϫ/Ϫ SC by P13 (Townsend et al., 2003). This difference from WT was retained in the P15-P17 pups used here. We recorded mEPSCs from WT and NR2A Ϫ/Ϫ neurons in the absence and presence of 50 M AP-5 (Fig. 1A). In WT, the ratios of decay time with AP-5/ without AP-5 were significantly lower than in NR2A Ϫ/Ϫ (Fig. 1B), indicating a significant contribution of mNMDARcs in WT com-pared with NR2A Ϫ/Ϫ . Neither mEPSC rise time (Fig. 1C) nor amplitude ratios (Fig. 1D) differed between the two strains. Townsend et al. (2003) also demonstrated at P11-P13 slower rise and decay time eNMDARcs in NR2A Ϫ/Ϫ SC neurons compared with WT, with no difference in AMPARc rise or decay times. We corroborated this after eye opening [eNMDARc rise times in NR2A knock-out NR2AKO Ϫ/Ϫ slower than eNMDARc rise times in WT, p Ͻ 0.001; eAMPARc rise times in NR2A Ϫ/Ϫ not different from eAMPARcs in WT, p ϭ 0.66; recorded at ϩ40 mV with NBQX and at Ϫ70 mV, respectively]. eNMDARc/eAMPARc ratios for both WT and mutant neurons were then compared at multiple stimulation intensities ( Fig. 2 A). Evoked NMDARc/ AMPARc rise time and decay time ratios were significantly higher in NR2A Ϫ/Ϫ than that in WT neurons, confirming respectively slower rise time eNMDARcs and more NR2B-rich NMDARs in NR2A Ϫ/Ϫ neurons. Amplitude ratios were significantly lower in NR2A Ϫ/Ϫ SC, indicating fewer total NMDARs in the mutant. The ratios remained constant for both strains over the full range of stimulating intensities. Differences between WT and NR2A Ϫ/Ϫ NMDAR currents were not caused by altered presynaptic release because PPRs did not differ between genotypes ( Fig. 2 B).

Different evoked NMDAR/AMPAR current ratios between NR2A Ϫ/Ϫ and WT neurons
NR2A Ϫ/Ϫ mice lack SC LTP As in rat SC neurons, LTP could be induced with 20 Hz stimulation for 20 s at intensity of 30 -50% spike threshold (ST) in WT but not in NR2A Ϫ/Ϫ neurons (Fig.  3 A, B,E). Increasing induction intensity produced a small but significant depression in WT but not in NR2A Ϫ/Ϫ neurons (Fig. 3 A, B,E). Stimulating frequencies of 10 Hz at 30 -50% ST produced no change in both WT and NR2A Ϫ/Ϫ neurons (Fig.  3E), whereas 50 Hz stimulation produced a significant LTD in both WT and NR2A Ϫ/Ϫ neurons (Fig. 3 A, B,E). SC LTP shows similar frequency sensitivity in rats (Zhao et al., 2006). Using the effective 20 Hz, 30 -50% ST protocol in WT neurons, bath application of AP-5 or AP-5 plus Nim prevented plasticity, Nim or Nim plus Ro 25 resulted in LTD (Fig. 3C,E), Nim applied after stimulation blocked change in EPSP slopes (Fig.  3 D, E), and LTP switched to LTD with Nim during preinduction (Fig. 3 D, E). Therefore, L-type Ca 2ϩ channel activity is necessary for both the induction and maintenance of SC LTP in WT. In Each average is of 10 evoked currents. Calibration: 50 pA, 100 ms. Bottom, Summary plots showing averaged eNMDARc/eAMPARc ratios for rise time, decay time, and amplitude (n ϭ 6 and 2 for each genotype). Genotype had a significant effect on rise time ratio (F ϭ 3.7, df ϭ 7, p Ͻ 0.0001), decay time ratio (F ϭ 5.7, df ϭ 7, p Ͻ 0.0001), and amplitude ratio (F ϭ 4.9, df ϭ 7, p Ͻ 0.0001). NR2A Ϫ/Ϫ neurons had higher rise time and decay time ratios and lower amplitude ratios than WT neurons. Stimulus intensity had no effect ( p Ͼ 0.05) on any ratio, and there was no significant interaction between stimulus intensity and genotype ( p Ͼ 0.05) (two-factor ANOVA). B, PPR sample traces (left; calibration: 20 pA, 50 ms) and summary graph (right) showing no difference between WT and NR2A Ϫ/Ϫ neurons (WT, 0.95 Ϯ 0.06, n ϭ 14 and 3; NR2A Ϫ/Ϫ , 0.95 Ϯ 0.07, n ϭ 14 and 3; p ϭ 0.97). Each sample trace is the average of 10 paired-pulse responses. T, Threshold. White diamonds represent individual experiments, and black diamonds are means of all the experiments in that group. NR2A Ϫ/Ϫ cells, 20 Hz, 30 -50% ST did not change EPSP slopes in the presence of Nim (Fig. 3E). Nim alone did not change the baseline activation in either genotype (data not shown). The mechanisms of the LTD induced by high-frequency stimuli or L-type Ca 2ϩ channel blockade remain unknown. As suggested by Zhao et al. (2006), they could involve fatigue of presynaptic terminals and postsynaptic conductance that change excitability or entrainment of metabotropic glutamate receptormediated LTP. Nim versus Nim plus Ro 25 in WT suggest that NR2A-containing receptors are involved in the LTD induced by 20 Hz. Consistently, no LTD is induced in NR2A Ϫ/Ϫ mice in the presence of Nim. In contrast, LTD induced by 50 Hz appears to be NR2A independent.

LLF-LTD is unaltered in WT and
NR2A Ϫ/Ϫ SC neurons Significant LTD was induced with 900 stimuli at 30 -50% ST presented at 1 Hz (LLF stimulation) in WT (Fig. 4 A, E) and NR2A Ϫ/Ϫ (Fig. 4C,E) neurons. In both mouse strains, application of AP-5 ( Fig.  4 A, C,E), or Nim (Fig. 4 B, D,E) or Ro 25 (Fig. 4 B, D,E) alone eliminated this depression. Therefore, blockade of either NR2B-NMDARs or L-type Ca 2ϩ channels is sufficient to eliminate SC LTD whether or not NR2A subunits are present.
The present results are consistent with previous work in several preparations (Liu et al., 2004;Massey et al., 2004;Lu et al., 2001b): Namely, synaptic, NR2A-rich NMDARs are obligatory for the induction of SC LTP, and extrasynaptic NR2B-rich NMDARs are critical to SC LTD induced by long, low-frequency stimulation. However, three questions critical to the NR2A/ NR2B debate remain unanswered. Are there unique attributes of NR2A subunits that require a concentration of this protein at synapses to initiate NMDAR-dependent LTP? At P15-P17, PSD-95 is the dominant scaffold for synaptic SC NMDARs (van Zundert et al., 2004). Are the cytoplasmic signaling molecules specifically associated with PSD-95 critical to LTP? Do NR2B-rich NMDARs selectively drive depression or is their requirement in SC LLF-LTD attributable to a different signaling complex that they specifically associate with as a result of being predominantly extrasynaptic? Answers to these questions are important. They should reveal how one receptor produces two diametrically opposite effects on synaptic plasticity and brain function. LLF-LTD (Cont; 77.9 Ϯ 6, n ϭ 12 and 3, p Ͻ 0.01) was blocked by AP-5 (94.4 Ϯ 4.7, n ϭ 8 and 2, p ϭ 0.34), Nim (95.2 Ϯ 4.6, n ϭ 10 and 3, p ϭ 0.35), or Ro 25 (97.3 Ϯ 2, n ϭ 7 and 2, p ϭ 0.28). WT versus NR2A Ϫ/Ϫ LLF-LTD was not significantly different ( p ϭ 0.58). Sample traces represent averages of 30 EPSPs obtained during the 10 min of base line (1) and 30 -40 min after 900 stimuli at 1 Hz induction (2) in WT and NR2A Ϫ/Ϫ neurons (calibration: 10 mV, 50 ms). E, Summary of mean EPSP slope changes in WT and NR2A Ϫ/Ϫ neurons under the different pharmacological conditions. Format as in Figure 3E (paired-sample, two-tailed Student's t test). White diamonds represent individual experiments, and black diamonds are means of all the experiments in that group.