Loss of retinoid X receptor gamma subunit impairs group 1 mGluR mediated electrophysiological responses and group 1 mGluR dependent behaviors

Retinoid X receptors are members of the nuclear receptor family that regulate gene expression in response to retinoic acid and related ligands. Group 1 metabotropic glutamate receptors are G-protein coupled transmembrane receptors that activate intracellular signaling cascades in response to the neurotransmitter, glutamate. These two classes of molecules have been studied independently and found to play important roles in regulating neuronal physiology with potential clinical implications for disorders such as depression, schizophrenia, Parkinson’s and Alzheimer’s disease. Here we show that mice lacking the retinoid X receptor subunit, RXRγ, exhibit impairments in group 1 mGluR-mediated electrophysiological responses at hippocampal Schaffer collateral-CA1 pyramidal cell synapses, including impaired group 1 mGluR-dependent long-term synaptic depression (LTD), reduced group 1 mGluR-induced calcium release, and loss of group 1 mGluR-activated voltage-sensitive currents. These animals also exhibit impairments in a subset of group 1 mGluR-dependent behaviors, including motor performance, spatial object recognition, and prepulse inhibition. Together, these observations demonstrate convergence between the RXRγ and group 1 mGluR signaling pathways that may function to coordinate their regulation of neuronal activity. They also identify RXRγ as a potential target for the treatment of disorders in which group 1 mGluR signaling has been implicated.


RXRγ is required for group 1 mGluR-dependent long-term depression (LTD) of synaptic strength, Ca 2+ release from intracellular stores and voltage-dependent cation currents. RXRγ
has been reported previously to be required for induction of stimulus-evoked LTD at hippocampal Schaffer collateral-CA1 synapses 18 . We reproduced this result using a 2 Hz/10 min stimulation protocol that elicited persistent LTD in slices prepared from wild-type animals, but significantly less LTD in slices from RXRγ knockout littermates (Fig. 1A). Since LTD can be induced at these synapses via at least two distinct mechanisms-one dependent on NMDA receptor activity and another dependent on group 1 mGluR activity 33,34 , we sought to determine whether RXRγ was required for one or both of these forms of LTD. We elicited mGluR-dependent or NMDA receptor-dependent LTD in acute slices from wild-type and RXRγ knockout mice by direct pharmacological activation of these receptors using the agonists DHPG or NMDA. Surprisingly-given the dependence of low-frequency dependent LTD on NMDA receptor activation 35,36 (Fig. S1)-we found that RXRγ knockout significantly impaired LTD induced by application of the group 1 mGluR agonist, DHPG (Fig. 1B), but had no show time courses of averaged initial fEPSP slopes ± SEM at Schaffer collateral-CA1 synapses in hippocampal slices from wild-type (WT) vs. RXRγ knockout (KO) mice. In (A-D), mean responses relative to baseline were calculated for each slice over 50-55 min post-stimulation/application and compared between groups using Student's t-tests. (A) LTD at Schaffer collateral-CA1 synapses evoked by 2 Hz/10 min low frequency stimulation (black bar) was eliminated in hippocampal slices from RXRγ-KO mice (98.8 ± 4.6% of baseline, grey circles, N = 11 slices from 7 animals) compared to slices from wild-type animals (87.9 ± 3.7% of baseline, black circles, N = 14 slices from 7 animals) (t = 2.1, P = 0.045). (B) LTD evoked by 10 min bath application of 30 μM DHPG (solid bar) was significantly smaller in hippocampal slices from RXRγ-KO mice (82 ± 3.9% of baseline, grey circles, N = 15 slices from 10 animals) compared to wild-type mice (63 ± 2.6% of baseline, black circles, N = 12 slices from 5 animals) (t = 3.605, P = 0.0014). (C) LTD evoked by 5 min bath application of 20 μM NMDA (solid bar) was not significantly different in hippocampal slices from RXRγ-knockout mice (86 ± 5.3% of baseline, grey circles, N = 8 slices from 7 animals) compared to slices from wild-type mice (89 ± 3.9% of baseline, black circles, N = 7 slices from 7 animals) (t = 1.951, P = 0.0652). (D) LTD evoked by 2 Hz/10 min train of pairs of stimuli (interpulse interval 50 ms) in the presence of the NMDA receptor antagonist D-AP5 (10 µM) to isolate mGluR-dependent LTD, was converted to modest potentiation in hippocampal slices from RXRγ KO mice (113 ± 6.2% of baseline grey circles, N = 11 slices from 6 animals), compared to wild type mice (79 ± 4.4% of baseline black circles, N = 11 slices from 5 animals) (t = 87.11, P < 0.0001). (E) DHPGinduced Ca++ release measured using the calcium sensitive dye Calcium Green and two-photon laser scanning microscopy imaging was significantly smaller in RXRγ KO CA1 pyramidal neurons, compared to wild-type control neurons (WT 32.4 ± 7% increase, KO 11.5 ± 2.5% increase, P = 0.029, Student's t-test, t = 2.841, N = 4 cells per group). DHPG was bath applied to slices at 30 μM for 20 min beginning at time = 0. Mean ± SEM Calcium Green fluorescence intensity normalized to pretreatment baseline were calculated at 5, 10, 15 and 20 min of DHPG application. Data were plotted using Prism (https ://www.graph pad.com/scien tific -softw are/prism /) and the figure assembled using Affinity Designer (https ://affin ity.serif .com/en-gb/desig ner/) software.

Scientific Reports
| (2021) 11:5552 | https://doi.org/10.1038/s41598-021-84943-x www.nature.com/scientificreports/ significant effect on LTD induced by application of NMDA (Fig. 1C). An NMDA receptor-independent form of LTD that depends on group 1 mGluR activation, but may also involve contribution of additional receptor types, can be produced by administration of low frequency paired-pulse stimulation in the presence of an NMDA receptor antagonist [37][38][39][40] (Fig. S1). This protocol produced a robust LTD of evoked responses in slices from wildtype animals, but actually produced modestly potentiated responses in animals lacking RXRγ (Fig. 1D), further suggesting that RXRγ knockout impairs LTD elicited by group 1 mGluR activation.
Since these results suggested that RXRγ knockout interferes with LTD produced by group 1 mGluR activation, we sought to determine whether RXRγ might also be required for other group 1 mGluR-mediated electrophysiological responses. One such response is the release of Ca 2+ from internal stores that results from group 1 mGluR activation. To visualize dendritic [Ca 2+ ], we loaded CA1 pyramidal neurons in hippocampal slices with the calcium sensitive dye, Calcium Green-1 (100 µM), and measured calcium release in these cells using twophoton laser scanning microscopic imaging 41 . Bath application of the group 1 mGluR agonist, DHPG (30 μM), elicited a time-dependent change in mean fluorescence intensity that was significantly smaller in CA1 pyramidal neurons lacking RXRγ, compared to wild-type control neurons (Fig. 1E). These data suggest that, in addition to group 1 mGluR-dependent LTD, RXRγ is also required for normal group 1 mGluR-induced release of Ca 2+ from intracellular stores.
A third group 1 mGluR-mediated response that we assessed was group 1 mGluR-mediated activation of voltage-sensitive inward current 42 . To do this, we performed whole-cell patch-clamp recordings from CA1 pyramidal neurons in hippocampal slices from wild-type animals and their RXRγ knockout siblings. We found that, in wild-type cells, bath application of the group 1 mGluR agonist DHPG (30 µM) elicited a voltage-dependent inward current, but that this current was absent in cells from RXRγ knockout animals (Fig. 2). This blockade of DHPG-induced voltage-sensitive currents is consistent with the effect of RXRγ knockout on mGluR-dependent LTD and mGluR agonist-evoked calcium release, further supporting the conclusion that a lack of RXRγ results in marked impairment in group 1 mGluR signaling in these animals.
RXRγ does not affect group 1 mGluR expression. One of the ways in which loss of RXRγ could result in reduced group 1 mGluR signaling is through reduced group 1 mGluR expression. To test this possibility, we compared mGluR1 and mGlur5 expression in hippocampal homogenates and slices from wild-type and RXRγ knockout animals. Quantitative RT-PCR using mGluR1 or mGluR5 specific primers revealed no significant differences in RNA expression levels for these two genes between wild-types and knockouts (Fig. 3A), suggesting that loss of RXRγ does not affect group 1 mGluR signaling through reductions in mGluR1 or mGluR5 mRNA expression. To test for possible differences in mGluR1 or mGluR5 protein levels, we also performed quantitative western blot analysis on hippocampal homogenates from wild-type and RXRγ knockout animals and found comparable levels of these proteins in both groups (Fig. 3B). Finally, to test for possible qualitative differences in the distribution of these two receptors, we performed immunohistochemistry on hippocampal sections from these animals, and found no gross differences in mGluR1 or mGluR5 distribution when we compared wildtype and RXRγ knockout mice (Fig. 3C). Together, these data suggest that, while loss of RXRγ impairs group 1 mGluR signaling, this is not due to reduced mGluR1 or mGluR5 expression or gross alterations in group 1 mGluR distribution in these animals.
Since our data suggested that loss of RXRγ impairs group 1 mGluR-induced electrophysiological responses, we next sought to determine the extent to which loss of RXRγ might also affect disease-relevant behavioral processes known to involve group 1 mGluR activity.
RXRγ knockout does not alter anxiety in a novel open field environment or elevated plus maze. Group 1 mGluR antagonists have been found to produce anxiolytic effects in animal models (reviewed in 24 ). To assess effect of loss of RXRγ on anxiety, we compared the behavior of RXRγ knockout animals and wild type control siblings in a novel open field environment and in an elevated plus maze. We found that RXRγ knockout animals exhibited a slight reduction in the amount of time spent in the center of the open field that was not statistically significant compared to their wild type control siblings (Fig. 4A), and a significant reduction in the amount of time spent rearing (Fig. 4B). While these results are suggestive of a possible modest anxiogenic effect of the loss of RXRγ, we observed no difference between RXRγ knockout animals and wild type siblings in the amount of time spent in the open vs. closed arms of an elevated plus maze (Fig. 4C). Neither did these groups differ in the amount of time spent freezing during pre-shock exposure to a contextual fear conditioning apparatus (Fig. 5A). Both groups of animals travelled comparable distances during open field and elevated plus maze testing (Figs. S2A,C) and spent comparable amounts of time resting (immobile) in the open field (Fig. S2B), suggesting that their behavior in these tasks was not affected by differences in activity level. The behavior of RXRγ knockout animals we observe in the open field and elevated plus maze is consistent with previous studies on an independently generated line of RXRγ knockout mice 20 Fig. S1). (C) Comparable levels and distribution of mGluR1 and mGluR5 protein in the hippocampal CA1 region by qualitative immunohistochemistry. Representative images of immunostained hippocampal CA1 region tissue in sections prepared from 3 animals per genotype processed in parallel with anti-mGluR1 or anti-mGluR5 primary antibody or with primary antibody omitted. Data were plotted using Prism (https ://www.graph pad.com/scien tific -softw are/prism /) and images prepared using Image Studio (https ://www.licor .com/bio/image -studi o/) and Olympus Fluoview (http://www.olymp uscon focal .com/produ cts/fv100 0/fv100 0soft ware.html) software. The figure was assembled using Affinity Designer (https ://affin ity.serif .com/en-gb/desig ner/) software.  43,44 and basal ganglia circuits 45 , and mGluR1 (but not mGluR5 46 ) knockout mice have been found to exhibit deficits in motor coordination [47][48][49] . A role for group 1 mGluRs has also been suggested in Parkinson's disease and mGluR5 antagonists have shown promise for the treatment of levodopa-induced dyskinesia (reviewed in 24 ). To assess the effect of RXRγ knockout on motor performance and learning, we tested these animals and their wild-type siblings on an accelerating rotarod task. We found that loss of RXRγ resulted in a significant decrease in motor performance in this task (Fig. 4D) that was consistent with the trend observed previously for RXRγ knockout animals on a continuous speed, single day version of this task 23 . These data are consistent with the possibility that impaired group 1 mGluR signaling in the RXRγ knockout mice may contribute to reduced motor performance in these animals.
RXRγ knockout impairs spatial novel object recognition. Group 1 mGluR signaling is involved in changes in synaptic efficacy that are thought to be required for normal hippocampus-dependent learning and memory, and genetic or pharmacological interference with group 1 mGluRs has been found to cause deficits in spatial learning and memory (reviewed in 50 ). To determine whether RXRγ might also be required for normal spatial learning and memory, we compared the performance of RXRγ knockout animals to that of their wild type siblings in a contextual fear conditioning task. Both mGluR1 and mGluR5 knockout animals have been found to exhibit deficits in this task 47,51,52 . However, the freezing response of RXRγ knockout animals was comparable to controls both during initial exposure to a novel context, and during re-exposure to that context 24 h after CS-US (E) Histogram of average percent time ± SEM spent exploring two familiar objects (object 1 and center object) and one novel object during a spatial version of a novel object recognition test revealed significant differences between RXRγ KO animals (N = 12) and wild type siblings (N = 12). 2-way ANOVA shows significant object × genotype interaction (F(2,44) = 4.102, P = 0.0233). Tukey's post-hoc test reveals significant differences in exploration time for object 1 vs. novel object for wild type (P = 0.0046), but not RXRγ KO animals (P = 0.8940). Data were plotted using Prism (https ://www.graph pad.com/scien tific -softw are/prism /) and the figure assembled using Affinity Designer (https ://affin ity.serif .com/en-gb/desig ner/) software.  5A). Loss of mGluR5 has also been suggested to interfere with the extinction of contextual fear conditioning 52 , however, RXRγ knockout animals exhibited a normal decline of the conditioned fear response during repeated exposure to the context in the absence of foot shock (Fig. 5A). Together these data suggest that loss of RXRγ does not alter contextual fear conditioning or extinction of the conditioned fear response. As a second test for the possible effects of loss of RXRγ on hippocampus-dependent learning and memory, we compared the performance of RXRγ knockout animals to that of their wild-type siblings during visible platform, hidden platform, and reversal phases of a Morris water maze task. Both RXRγ knockout and wild type siblings rapidly learned the visible version of the water maze, suggesting that no task-relevant differences in swimming ability, visual acuity, or motivation exist between these groups (Fig. 5B, days v1-2). Swimming speed was also comparable between these two groups during all stages of water maze testing (Fig. S3A). Neither were significant differences observed between mutants and controls in the acquisition of a hidden platform version of this task (Fig. 5B, days h1-5) or in performance during a probe trial conducted at the end of hidden platform training (day h5) (Fig. 5C, probe 1). This result is in agreement with previously reported data 18 , and despite the observation that mice lacking mGluR5 show significant impairments in this task 51,52 . mGluR5 knockout mice have also been reported to exhibit a deficit in the reversal phase of this task 52 -wherein they are required to navigate to the platform in a new location. However, we observed no significant differences between RXRγ knockout animals and their wild type siblings during reversal training (Fig. 5B, days r1-5) or in performance during probe trial conducted on the last day of reversal training (day r5) (Fig. 5C, probe 2).
Animals lacking RXRγ were previously reported to exhibit deficits in novel object recognition 22 . Two versions of this task exist-one that requires hippocampus-dependent spatial memory in which animals recognize familiar objects in novel locations 53 , and a second, non-spatial version, that requires group 1 mGluR-mediated LTD in the perirhinal cortex in which animals recognize novel objects in familiar locations 54 . Since a previous report of novel object recognition impairments in RXRγ knockout animals did not distinguish between these two modalities, we sought to determine which form(s) of novel object recognition might be affected by loss of RXRγ. We found that RXRγ knockout animals and wild-type siblings showed similar preferences for novel versus familiar objects in the non-spatial version of this task (Fig. 5D), suggesting that loss of RXRγ did not affect the recognition of novel objects in familiar locations. However, we did observe a significant difference in the performance of these groups in the spatial version of the novel object recognition task. In this version, RXRγ knockout animals failed to distinguish familiar objects in novel locations under conditions where their wild-type siblings showed a clear preference (Fig. 5E), suggesting that RXRγ is required for the spatial but not the non-spatial form novel object recognition. In neither version of the task did the RXRγ knockout affect total exploration time (Fig. S3B,C). The observation that loss of RXRγ impairs performance in a spatial novel object recognition, but not contextual fear www.nature.com/scientificreports/ conditioning or Morris water maze tasks, suggests that the novel object recognition impairment may be related to the reported association between mGluR-dependent LTD and the response to novel environments and stimuli 50 .
RXRγ knockout mice show reduced pre-pulse inhibition and working memory performance. Group 1 mGluR signaling has also been implicated in the pathology of schizophrenia (reviewed in 24 ).
In support of this hypothesis, mutations in mGluR1 and mGluR5 [55][56][57][58][59] , and changes in mGluR1 expression [60][61][62] are associated with schizophrenia in human patients, and genetic and pharmacological inhibition of group 1 mGluR signaling results in schizophrenia-related impairments in rodents both alone and in combination with NMDAR antagonist administration (reviewed in 63,64 ). While potentiation of NMDAR function was thought to underlie the salutary effects of group 1 mGluR positive allosteric modulators (PAMs) on schizophrenia-related phenotypes, recent evidence suggests that group 1 mGluR PAMs may exert these effects via NMDAR-independent mechanisms as well 65 . A possible role for retinoid signaling in schizophrenia pathogenesis and treatment has also been suggested 17 , and mice lacking RXRγ have been reported to exhibit deficits in working memory that are also a characteristic schizophrenia-associated cognitive impairment 21,22 . Our observation that loss of RXRγ impairs group 1 mGluR responses suggests that reduced group 1 mGluR activity may underlie, or contribute to schizophrenia-related impairments in RXRγ knockout mice.
To pursue this association further, we compared pre-pulse inhibition in RXRγ knockout mice and their wild type siblings. Prepulse inhibition is a behavioral paradigm designed to assess sensory-motor gaiting that has been found to be reduced in schizophrenic patients (reviewed in 66,67 ) as well as mGluR1 knockout, mGluR5 knockout, and NMDAR antagonist-treated mice [68][69][70][71][72][73] . We found reduced prepulse inhibition in RXRγ knockout mice across a range prepulse amplitudes that reached statistical significance at the highest amplitude tested (Fig. 6A). The RXRγ knockout mutation was without effect on either the amplitude of the startle response (Fig. S4A), or the startle response time (Fig. S4B), suggesting that the reduced prepulse inhibition we observed in these animals is the result of an impairment in sensory-motor gaiting.
Impairments in working memory are also considered an endophenotype of schizophrenia, and have been described previously in mice lacking RXRγ 21,22 . To verify this impairment, we tested RXRγ knockout animals in a Y-maze spontaneous alternation task as described in 22 and found a similar trend for reduced spontaneous alternation (Fig. 6B) that was consistent with, but lower in magnitude than previously reported.
RXRγ knockout mice show increased locomotor sensitization to cocaine. Group 1 mGluRs are involved at multiple levels in the synaptic changes that are thought to underlie the behavioral response to drugs of abuse 31,74 , and a majority of data suggest that inhibition of mGluR1/5 signaling may alleviate symptoms associated with addiction (reviewed in 24 ). To examine the effect of reduced RXRγ activity on addictionrelated behavioral responses, we compared the locomotor activity of RXRγ knockout and wild-type animals that received daily doses of 15 mg/kg cocaine. Surprisingly, we found significantly higher cocaine-induced locomotor sensitization in RXRγ knockout animals compared to their similarly treated wild-type siblings (Fig. 6C). This is in contrast to the reduced amphetamine-induced locomotor responses described in mice treated with the RXR antagonist HX531 75 and the reduced locomotor activity observed in RARβ/RXRγ and RXRβ/RXRγ double knockout mice that received a single cocaine injection 23 . The increased cocaine-induced locomotor sensitization we observe in RXRγ knockout animals may reflect the of loss of RXRγ during development, the selective interference with a subset of addiction-related group1 mGluR signaling functions, or mGluR-independent effects on the mechanisms underlying locomotor sensitization. However, it is also possible that this response is a consequence of altered drug metabolism stemming from the role that rexinoid signaling plays in regulating cytochrome expression in the liver 76,77 . RXRγ knockout blocks DHPG-induced LTD, but does not rescue increased locomotor activity in Fmr1 mutant mice. Group 1 mGluR activation causes an increase in the local translation of a subset of mRNAs in dendritic spines, leading to an increase in proteins that mediate group 1 mGluR-dependent LTD 78 . The activating effect of group 1 mGluRs on translation is opposed by the repressive effect that the fragile X mental retardation protein (FMRP) exerts on these mRNAs, and accordingly, an increase in group 1 mGluRdependent LTD is observed in Fmr1 mutant mice that lack FMRP 79 . This antagonistic relationship between the actions of group 1 mGluRs and FMRP on mRNA translation led to the proposal that inhibiting group 1 mGluR signaling might counter the effect of reduced FMRP function in patients with fragile X syndrome, and ameliorate some of the associated symptoms 80 . While this strategy has thus far proved ineffective in treating fragile X syndrome patients in clinical trials, multiple studies have reported that reducing group 1 mGluR signaling does effectively reverse fragile X syndrome-related phenotypes in mouse models of the disease 81,82 .
To explore the possibility that RXRγ-dependent reductions in group 1 mGluR signaling might affect fragile X syndrome-related phenotypes in a mouse model of fragile X, we crossed RXRγ knockout mice with Fmr1 I304N mutant mice that carry a disease-linked point mutation in the gene encoding FMRP 83 . In wild-type hippocampal slices, DHPG treatment elicits a form of LTD that depends on the rapid induction of local protein synthesis, however, in Fmr1 I304N mutant mice, DHPG-induced LTD is no longer protein synthesis-dependent 83 . As shown in Fig. 7A,B, we found that the RXRγ/Fmr1 I304N double mutant mice showed significantly less LTD than mice carrying the Fmr1 I304N mutation alone, consistent with a general RXRγ-dependent impairment in group 1 mGluR signaling. However, we failed to observe significantly enhanced DHPG-induced LTD under these conditions in mice carrying the Fmr1 I304N mutation alone. To determine whether loss of RXRγ might also affect the behavioral phenotypes of mice that carry the Fmr1 I304N mutation, we compared the open field locomotor activity among siblings with these same mutant combinations. As reported previously 83  www.nature.com/scientificreports/ of RXRγ knockout on DHPG-induced LTD, loss of RXRγ did not significantly impact the locomotor activity of RXRγ/Fmr1 double mutant animals when compared to siblings that carried the Fmr1 knock-in mutation alone (Fig. 7C). Together, these data are consistent with the notion that, while loss of RXRγ impairs group 1 mGluR signaling, the actions of group 1 mGluR signaling and FMRP do not overlap completely, and therefore, impairments in group 1 mGluR signaling caused by RXRγ knockout may affect only a subset of Fmr1 mutant phenotypes.

Discussion
The electrophysiological data we present here suggest that RXRγ is required for normal group 1 mGluR-dependent electrophysiological responses in CA1 pyramidal neurons in the murine hippocampus. This conclusion is supported by impairments in mGluR agonist-induced LTD, mGluR agonist-induced calcium release, and mGluR agonist-induced activation of voltage-sensitive currents in RXRγ knockout mice. RXRγ knock out also impaired group 1 mGluR dependent LTD induced by paired-pulse stimulation in the presence of an NMDA receptor antagonist. These LTD impairments that we observe in the RXRγ knockout mice are similar to the impairments described in mice carrying a knockout mutation in a downstream effector of group 1 mGluR signaling-Gα q 84 . Given that loss of RXRγ also impairs LTD induced by low-frequency stimulation (LFS), and that this form of LTD depends on NMDA receptor activation, it is perhaps surprising to find that RXRγ knockout did not affect LTD induced by direct pharmacological activation of NMDA receptors 35,36 . While the basis for the effect of RXRγ knockout on LFS-induced LTD remains unclear, it may reflect impaired mGluR modulation of NMDA receptor activity, impaired mGluR-mediated metaplasticity, or another mGluR-dependent or independent effect on this form of LTD [85][86][87] . Future experiments to distinguish among these possible mechanisms may yield additional insights into the interaction between RXRγ and mGluR-dependent regulation of synaptic plasticity.
Given the involvement of group 1 mGluRs in multiple disease-relevant processes, we sought to determine the extent to which RXRγ-dependent impairments of group 1 mGluR signaling might affect behaviors in which these receptors are implicated. Our behavioral analysis identified impairments in a subset of group 1 mGluR-linked behaviors in mice lacking RXRγ. These impairments included: (1) impaired motor performance-suggesting that RXRγ-dependent decreases in group 1 mGluR signaling may underlie or contribute to this phenotype, (2) impaired prepulse inhibition that is consistent with the dependence of this behavior on normal group 1 mGluR activity and the role of group 1 mGluRs in this and other schizophrenia-related behaviors 24 , and (3) impaired recognition of familiar objects in novel locations, but not novel objects themselves, that is consistent with the reported relationship between hippocampal LTD and the response of mice to novel environments [88][89][90] . Despite these examples of correspondences between the effect of the RXRγ knockout mutation on mGluRdependent electrophysiological responses, and its effects on mGluR-dependent behaviors, we also identified examples of apparent dissociations. For example, we found that loss of RXRγ did not alter anxiety or extinction, despite the well-established anxiolytic effects of mGluR antagonists 24 and the evidence linking group 1 mGluR activity to extinction 52 . Loss of RXRγ also did not affect non-spatial novel object recognition learning, despite considerable evidence linking mGluR-dependent LTD in the perirhinal cortex to these behaviors 54 . While loss of RXRγ still impaired DHPG-induced LTD in Fmr1 mutant mice, loss of RXRγ had no detectable effect on increased locomotor activity in these animals. The differential effects of loss of RXRγ on mGluR-dependent behaviors suggest that loss of RXRγ may affect mGluR activity differently in different brain regions. For example, loss of RXRγ may affect hippocampus-dependent spatial novel object recognition, but not perirhinal-dependent non-spatial novel object recognition-or loss of RXRγ may modulate only a subset of the downstream effects of mGluR activation. These dissociations are perhaps to be expected given the number of downstream effectors of group 1 mGluRs, their wide distribution throughout the brain, and the multiple forms of synaptic plasticity in which group 1 mGluRs play a role 24,31,32 . An understanding of the basis for the differential effects of RXRγ on group 1 mGluR-dependent behaviors will require an examination of its modulation of additional mGluRdependent electrophysiological effects in different brain regions.
It will also be informative to determine the molecular basis for the genetic interaction between RXRγ knockout and group 1 mGluR signaling. Our data suggest that RXRγ-dependent changes in group 1 mGluR expression are not the basis for this interaction, raising the possibility that RXRγ-dependent changes in the expression of one or more group 1 mGluR downstream effectors or interacting proteins may lead to the impairments in group 1 mGluR mediated responses we observe. It is also possible that RXRγ exerts its effects on group 1 mGluR signaling via mechanisms that do not involve RXRγ-dependent changes in gene expression, but involve instead as yet unidentified direct molecular interactions with downstream targets analogous to the non-genomic actions of other nuclear hormone receptors 9 . An understanding of the molecular basis for the interaction between RXRγ and group 1 mGluR signaling may also suggest a role for RXRγ in mediating or modulating the electrophysiological and behavioral actions of other G-protein coupled receptors.
Our finding that loss of RXRγ inhibits group 1 mGluR electrophysiological responses and affects a subset of group 1 mGluR-dependent behaviors raises the possibility that pharmacologically targeting RXRγ could represent a viable therapeutic avenue for the treatment of diseases in which alterations in group 1 mGluR signaling play a role. Obstacles to this approach include the complexity of RXR's interactions with multiple nuclear hormone receptor family members, and the lack of compounds that specifically target individual RXR isoforms. Nevertheless, RXR's have been targeted successfully in the treatment of cancer despite these obstacles 3,9 . In addition to the potential therapeutic implications of these results, our data point to an interaction between two signaling pathways with established roles in regulating neuronal physiology and synaptic plasticity that could function under normal physiological conditions to integrate the cellular response to coincident stimulation of these pathways. It is likely that similar points of interaction exist among other signaling pathways, and that their continued identification will ultimately transform our current view of the molecular mechanisms that regulate neuronal physiology into a complex web of interdependent molecular signaling interactions that offer new insights and targets for disease intervention.

Methods
Animals. RXRγ knockout animals were described previously 18 and were generously provided by R. Evans (Salk Institute). Animals were received in a mixed background and independently backcrossed for more than 5 generations to either C57BL6/J or 129SVEV/TAC wild type animals to generate congenic lines. Fmr1 I304N mutant mice 83 were obtained from the Jackson Laboratory (Bar Harbor, ME) in a C57BL6/J background and crossed to C57BL6/J congenic RXRγ mice to generate double heterozygous females in this background. All animals used for experiments were F1 hybrids of the indicated ages, generated by crossing heterozygous or double heterozygous C57BL6/J animals to 129SVEV/TAC RXRγ heterozygous animals. Animals were maintained under standard conditions in ventilated cages on a 12 h light-dark cycle and tested during the day. Wild-type and knockout siblings were housed together in groups of 3-5 and run on behavioral experiments in age-matched cohorts by experimenters blinded to genotype. All experiments were carried out in a manner consistent with NIH guidelines and approved by the Columbia University and New York Medical College Institutional Animal Care and Use Committees (IACUC), and described in accordance with the ARRIVE 2.0 guidelines 91 . Slice preparation. Acute hippocampal slices were prepared as described previously 92 . Wild-type and homozygous knockout mice of mixed sex, 3-6 months of age, were decapitated under deep isoflurane anesthesia, and the hippocampus plus entorhinal cortex dissected free from surrounding tissue and placed immediately in ice-cold artificial cerebrospinal fluid (ACSF) consisting of: 126 mM NaCl, 26 mM NaHCO 3 , 1.25 mM NaH 2 PO 4 , 5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , and 10 mM D-glucose, continuously gassed with 95% O 2 / 5% CO 2 (pH 7.2-7.4). For extracellular field potential recordings, 400 μm-thick transverse slices were cut using a vibrating tissue slicer (VT1200S, Leica Biosystems, Wetzlar, Germany), transferred to an interface recording chamber, maintained at 33 °C and continuously perfused with ACSF (3 mL/min), where they were incubated for a minimum of 90 min prior to the start of recording. For whole-cell patch recordings from single CA1 pyramidal neurons, slices were incubated for 1-6 h in oxygenated ACSF at room temperature prior to transfer to a submerged recording chamber continuously perfused with room temperature ACSF (5 mL/min).
Electrophysiological recordings. Extracellular population fEPSP recordings were made as described previously 93 using glass microelectrodes (2-3 MΩ when filled with ACSF) placed in stratum radiatum of the www.nature.com/scientificreports/ CA1 region of hippocampal slices under visual guidance, to a depth of 100-150 μm. Bipolar stainless-steel stimulation electrodes (FHC, Bowdoin, ME) were placed in stratum radiatum to activate Schaffer collateral afferents. For baseline recordings, synaptic inputs were stimulated once per min (150 μs square DC pulse). Baseline stimulus strength (10-200 μA) was adjusted to elicit a response ≈ 50% of the maximum fEPSP amplitude prior to the generation of a population action potential, and monitored for at least 30 min prior to induction of LTD. Slices in which there was a drift in baseline of > 5% for the 6 points prior to stimulation or drug treatment were excluded from further analysis. NMDA receptor-dependent LTD was elicited with a 10 min train of 2 Hz low frequency stimuli 94 , while group I mGluR-dependent LTD was evoked in the presence of the NMDAR antagonist D-2-amino-5-phosphonopentanoic acid (D-AP5, 10 µM) by a 2 Hz/10 min train of pairs of stimuli (50 ms pair interstimulus interval). Synaptic strength was quantified by measuring the maximum slope of the initial falling phase of the fEPSP, using a six-point interpolation least-squares linear regression analysis, which marched along the response until the maximum value was retrieved. Signals were collected with an Axoclamp-2A amplifier (Molecular Devices, San Jose, CA) filtered at 1 kHz, sampled at 10 kHz, and digitized and analyzed on an PC computer using DataWave Technologies software (DataWave, Loveland, CO). Whole-cell patch-clamp recordings (Axoclamp 700B, Molecular Devices) were obtained from CA1 pyramidal neurons visualized using infrared differential interference contrast optics on a Zeiss Axioskop FS upright microscope (Carl Zeiss, Thornwood, NY) as described previously 95 . The intracellular patch pipette filling solution contained: 130 mM CsMeSO 4 , 4 mM NaCl, 10 mM HEPES, 0.5 mM EGTA, 4 mM Mg-ATP, 0.3 mM Na-GTP, and 2 mM QX-314 (pH 7.25; 280-290 mOsm; 4-6 MΩ). EPSCs were acquired at 5 kHz sample frequency and filtered at 1 kHz with an eight-pole low-pass Bessel filter. After whole-cell configuration was established, membrane potential was voltage-clamped at − 60 mV. Recordings with leak currents of < − 100 pA, series resistances > 25 MΩ, or series resistances that changed by > 10% from beginning to end of experiments, were discarded. Voltage-dependent inward currents were evoked by stepping to − 80 mV, and then injecting a 1 s current ramp from − 80 mV to + 40 mV.
Calcium imaging. Calcium imaging was performed on proximal apical dendrites of CA1 pyramidal cells in acute hippocampal slice preparations as described previously 96 . A customized two-photon laser-scanning Olympus BX61WI microscope with a 60×/1.1 nA objective was used to detect Ca 2+ signals. A Mai/Tai laser (Solid-State Laser, Mountain View, CA) tuned to 820 nm was used for excitation, and image acquisition and storage was controlled by Olympus Fluoview FV300 software (Olympus, Melville, NY). In the transfluorescence pathway, a 565 nm dichroic mirror was used to separate green and red fluorescence, passed through HQ525/50 and HQ605/50 emission filters, to eliminate transmitted or reflected excitation light (Chroma Technology, Rockingham, VT), and detected simultaneously by two photomultiplier tubes. Neurons were loaded with dyes through the patch pipette for 20 min before commencing image acquisition. Alexa Fluor-594 was used to outline neuronal dendritic structure, and Calcium Green Western blots. Hippocampi were homogenized by sonication in 50 mM Tris pH 7.4/2% SDS followed by incubation at 95 °C for 5 min. Extracts were run in duplicate and resolved by denaturing PAGE and blotted to nylon membrane (Immobilon-FL, Millipore). Blocking and antibody incubations were for 1 h at room temperature in Tris-buffered saline, 0.05% tween 20, 3% non-fat dry milk. Washes were carried out TBS-Tween without milk. Proteins were detected using commercially available antibodies against mGluR1 (Millipore AB1551), mGluR5 (Millipore AB5675), and α-tubulin (Sigma T9026) in conjunction with dye conjugated anti-mouse or rabbit secondary antibodies (Li-COR, Lincoln, NE). Blots were scanned and integrated band intensities determined using an Odyssey infrared imaging system. mGluR1 or mGluR5 band intensities were normalized to the band intensity for tubulin in that lane as a loading control and data were presented as percent of the average normalized control value. Q-PCR. Total hippocampal RNA was prepared using Trizol Reagent and first strand cDNAs were prepared using the Superscript III first strand cDNA synthesis supermix for Q-PCR (Invitrogen) as per manufacturers instructions. Quantitect primer pairs were purchased from Qiagen. All primer pairs showed a linear response of Log(template dilution) vs. cycle threshold for template dilutions over 5 orders of magnitude. Optimal annealing temperatures, template dilutions and corresponding efficiencies were determined experimentally for each primer pair, and melting curves determined for each reaction showed a single peak under these conditions. Relative expression levels were calculated for each sample using the experimentally determined efficiency and the number of cycles to reach a florescence threshold set manually within the linear portion of a plot of the Log(florescence intensity) vs. cycle number. Values were normalized to those obtained using a primer pair specific for GAPDH run in parallel for each sample.
Immunohistochemistry. Brains  www.nature.com/scientificreports/ saline + 10% fetal bovine serum. Primary antibody incubations were carried out overnight at 4 °C in blocking buffer, secondary antibody incubations were carried out for 1 h at room temperature also in blocking buffer, and washes were TBS + 0.2% triton. Primary antibodies were from Millipore as above, and an Alexa-568 antirabbit from Invitrogen was used as a secondary. Stained sections were mounted in Fluorsave aqueous mounting medium (Calbiochem) and images were obtained by confocal microscopy (Olympus Fluoview FV1000 and associated software).
Open field. Behavior in a novel open field was assessed using Plexiglas activity chambers (model ENV-520; Med Associates, St. Albans, Vermont) (43.2 cm long × 43.2 cm wide × 30.5 cm high). Mice were placed in the open field and activity was recorded for 60 min. Behavioral measures were calculated using the Activity Monitor program (Med Associates) and "center" was defined as an area beginning 10 cm from the walls.
Elevated plus maze. Animals were placed on an elevated plus maze apparatus for 6 min and their position was tracked using an overhead digital video camera and Anymaze software (Stoelting, Wood Dale, IL). The apparatus consisted of a plus shaped track with arms 18 cm long and 5 cm wide. Two arms were enclosed on 3 sides and the entire apparatus was elevated 50 cm above the bench top and surrounded by a white curtain under ambient room lighting.
Accelerating rotarod. Mice were tested on an accelerating rotarod apparatus (Letica LE8200) over a period of 5 days. The first day consisted of 3 pre-training trials in which animals were habituated to the stationary apparatus 2 times for 3 min each, and then for an additional 3 min at 4 rpm. Animals that fell from the apparatus during these trials were returned to the rod until 3 min had elapsed. On subsequent days, animals were given 3 × 5 min trials per day with the rotation speed ramped from 4 to 20 rpm over the course of the trial. Animals that fell from the rod during these trials were returned to the rod once to discourage them from learning to fall as an escape strategy, however only the latency to the first fall was used for analysis. For all trails on all days, the intertrial interval was approximately 30 min during which animals were returned to their home cages.
Fear conditioning. All fear conditioning trials were carried out in conditioning chambers placed in sound attenuating chambers (Med Associates, St Albans, VT) and freezing during each trial was monitored continuously using a video tracking and analysis system (Freezeframe, Actimetrix Software, Wilmette, IL). On day one animals were given a 3 min training trial consisting of 2 min of pre-exposure followed by a 2 s 0.6 mA continuous foot shock and 58 s of post-shock exposure. On subsequent days animals were reintroduced to the conditioning chamber in the absence of shock for 10 min. Percent of time spent freezing during the first 2 min of these trails was used to monitor extinction of the freezing response.
Morris water maze. The experiment was carried out in a circular pool 120 cm in diameter filled with water made opaque with white paint. A video tracking system (HVS Image VP-118) was used to record and analyze each animal's behavior. A 10 cm square platform submerged 1 cm below the water surface was used as an escape platform and in the visible version of the task, a colored syringe barrel was affixed to the center to mark its location. For all phases of the experiment, 4 trials of no more than 120 s were given each day at 20 min inter-trial intervals. Animals that exceeded the 120 s limit were guided to the platform before being returned to their home cage. A single "shaping" trial in which animals were placed on the platform for 15 s, was performed before training on the first day of the visible, hidden and reversal phases of the task. Probe trials of 60 s, during which the platform was removed from the water maze, were performed at the end of training on days H5 and R5. In the visible platform phase of the task, the pool was surrounded by a curtain and the marked platform was placed in a different location on all four trials of a day. In the hidden and reversal phases, the curtain was removed to reveal the room cues and the hidden platform was placed in a fixed location for each phase.
Spatial and non-spatial object recognition. For all object recognition experiments, the apparatus consisted of an opaque plexiglass arena 71 cm square and 36 cm high. The arena was surrounded by a white curtain, and two prominent visual cues were affixed to two of the arena walls. Objects were glass, metal or ceramic items fixed in place with double stick tape (eg. padlock, glass chess piece, metal bottle cap, ceramic drawer knob). All objects, the arena and floor were cleaned with non-alcoholic antiseptic wipes between each trial. To habituate the animals to the testing environment, animals were handled in the testing room on the three days prior to testing. Animal movements were tracked and analyzed using Noldus Ethovision XT software and an overhead-mounted video camera. For the novel object version of the task animals were placed in the empty arena for 5 min first. Then three 5 min training trails were conducted in which three different objects were placed equidistantly in a diagonal line across the arena. Finally a 5 min. testing trial was conducted in which one of two peripheral objects was replaced with a new object. The delay between the empty arena trial and all of the training trials were 10 min, and the delay between the final training trial and the testing trial was 3 h. The spatial version of this task was carried out similarly except that instead of replacing one of the peripheral objects in the testing trail, it was moved to one of the two empty corners so that the three objects were now arranged in a "V".

Prepulse inhibition.
Testing was performed essentially as described in 97  www.nature.com/scientificreports/ through the experiment. The testing protocol consisted of 6 presentations of a 40 ms broadband 120 dB startle tone, followed by 10 pseudo-randomly interleaved presentations of background alone, prepulse trials consisting of stimuli at 3, 6 or 12 dB above background delivered 100 ms before the 40 ms broadband 120 dB startle tone, and non-prepulse trials in which the 120 dB startle tone was presented alone. This protocol was followed by another 6 presentations of the 120 dB startle tone alone. The interval between trials varied pseudorandomly between 15 and 30 s. PPI for each animal at a particular prepulse interval was calculated as the difference between the average startle magnitude on interspersed startle tone-alone trials and prepulse trials divided by the average startle magnitude on the same startle tone-alone trials.
Y-maze spontaneous alternation. Mice were tested individually for spontaneous alternation during a single 6 min trial in a custom fabricated Y-maze apparatus. The maze consisted of three arms 40 cm in length, 17 cm high and 4 cm wide, separated by 120 degrees located on a table in the center of the testing room under dim lighting. Animals were placed at the end of one of the arms in a psedorandomly assigned manner. Movements were recorded by an overhead mounted video camera and monitored by an observer blinded to genotype. Arm entries were scored when all four of the animal's paws entered an arm. The percent alternation was calculated as the number of triads containing entries into all three arms divided by the total number of triads (the total number of arms entered minus 2) × 100. Average number entries for all groups was 24.2 with a standard deviation of 5.5. Three RXRγ knockout and 2 wild-type mice with 13 or fewer arm entries were excluded from the analysis.
Behavioral sensitization to cocaine. Testing was conducted in the same apparatus used for open field behavior testing. Immediately before each 30 min trial, animals were given a single intraperitoneal injection of either 15 mg/kg cocaine or saline only. The testing protocol consisted of 3 days of habituation (1)(2)(3) in which all animals were administered saline only. Habituation was followed by 5 sensitization trials, conducted on days 4-8 and day 20, in which either cocaine or saline was administered as indicated. On day 22 all animals were tested again after receiving saline injections.
Statistical analysis. Tests for statistical significance between groups were performed using Prism 7 (Graphpad Software, San Diego, CA, USA). Student's unpaired, two-tailed t-tests were used for experiments involving 2 groups, and paired t-test for experiments in a single group comparing before to after drug application. 2-way ANOVA comparisons-with or without repeated measures as dictated by the experimental design-were used for analysis of all experiments involving multiple groups. Post-hoc comparisons for multiple group experiments were performed using Dunnett's tests when comparisons were made to a single control group, and Tukey's tests when comparisons were made among multiple groups. The results of these tests together with their associated P values are listed in the legends for their corresponding figure.