Endoplasmic reticulum chaperone genes encode effectors of long-term memory

Abstract

The mechanisms underlying memory loss associated with Alzheimer's disease and related dementias (ADRD) remain unclear, and no effective treatments exist. Fundamental studies have shown that a set of transcriptional regulatory proteins of the nuclear receptor 4a (Nr4a) family serve as molecular switches for long-term memory. Here, we show that Nr4a proteins regulate the transcription of genes encoding chaperones that localize to the endoplasmic reticulum (ER). These chaperones fold and traffic plasticity-related proteins to the cell surface during long-lasting forms of synaptic plasticity and memory. Dysregulation of Nr4a transcription factors and ER chaperones is linked to ADRD, and overexpressing Nr4a1 or the chaperone Hspa5 ameliorates long-term memory deficits in a tau-based mouse model of ADRD, pointing toward innovative therapeutic approaches for treating memory loss. Our findings establish a unique molecular concept underlying long-term memory and provide insights into the mechanistic basis of cognitive deficits in dementia.

Fig. 1.. A multiprotein ER chaperone complex is downstream of Nr4a transcription factors during memory consolidation.

(A) Schematic depicting the SOR procedure. Mice expressing the tetracycline transactivator (tTA) protein under the CaMKIIα promoter (CaMKIIα-tTA: control mice) and littermates expressing both CaMKIIα-tTA and the dominant-negative mutant Nr4A1 under the control of TetO promoter (CaMKIIα-tTA, TetO-Nr4A dominant negative: Nr4ADN mice) were trained in SOR and then tested after 24 hours. (B) Preference for the displaced object (DO; dotted line marks 33% chance) during training and testing. Two-way ANOVA: significant main effect of genotype (F_1,21 = 18.42, P = 0.0003) and significant main effect of sessions (F_1,21 = 8.417, P = 0.0085). Sidak’s multiple comparisons test: **P = 0.0054 (control mice, train versus 24-hour test), **P = 0.0011 (control 24-hour test versus Nr4ADN 24-hour test). Control: n = 11 (3F); Nr4ADN: 12 (4F). (C) Schematic depiction of RNA-seq experiment. Nr4ADN and control male mice were trained in SOR and euthanized 2 hours after training. mRNA was harvested from the dorsal hippocampus and processed for the preparation of an RNA-seq library. (D) Volcano plot illustrating significance (y axis) and magnitude (x axis) of the down-regulation (blue) and up-regulation (red) of genes in Nr4ADN mice. (E and F) Functional groupings of network of enriched categories for genes whose differential expression [(E), down-regulation; (F), up-regulation] was significant, using the ClueGO and CluePedia plugins of the Cytoscape software. Gene Ontology (GO) terms include Molecular Functions (MF) and Kyoto Encyclopedia of Genes and Genomes (KEGG) and are represented as nodes (κ score level ≥ 0.4), with node size representing the significance of the term enrichment. Only the most significant term in each group is shown.

Fig. 2.. A subset of the genes downstream of Nr4a is induced by learning.

(A) Quadrant plot based on total RNA-seq data. Genes induced by learning were identified on the basis of comparison of genes regulated in dorsal hippocampus of control male mice 2 hours after SOR (tTA⁺ Nr4ADN⁻, n = 2; tTA⁻ Nr4ADN⁻, n = 2) and homecage control mice (tTA⁺ Nr4ADN⁻, n = 2; tTA⁻ Nr4ADN⁻, n = 2). Quadrant plot comparing genes regulated by learning in control mice to genes regulated in Nr4ADN mice by learning. Genes up-regulated by SOR are down-regulated in Nr4ADN mice (labeled points). Size, opacity, and color intensity of each point reflect the minimum false discovery rate value for a gene between each experiment. (B and C) Expression of (B) Hspa5 and (C) Pdia6 mRNAs in C57BL/6J male mice trained in SOR and euthanized at the indicated times after training (1 hour: n = 9; 2 hours: n = 9; 3 hours: n = 9; 4 hours: n = 9), expressed as fold difference from that of mice handled only in the homecage (HC) (baseline controls, n = 10). One-way ANOVA: Hspa5: F_4,41 = 13.00, P < 0.0001. Sidak’s multiple comparisons tests: ***P < 0.0001 (HC versus 1 hour), ***P < 0.0001 (HC versus 2 hours), ***P < 0.0001 (HC versus 3 hours), and *P = 0.0158 (HC versus 4 hours); Pdia6: F_4,41 = 8.442, P < 0.0001. Sidak’s multiple comparisons tests: ***P = 0.0008 (HC versus 1 hour), ***P = 0.0001 (HC versus 2 hours), ***P < 0.0001 (HC versus 3 hours), and **P = 0.0054 (HC versus 4 hours). (D and E) Down-regulation of gene expression at 2 hours after SOR training in male Nr4ADN (n = 9) and control (n = 10) mice, as validated by qPCR. Unpaired t test: t₁₇ = 3.305, **P = 0.0042 (Hspa5); t₁₇ = 3.630, **P = 0.0021 (Pdia6). (F and G) Quantification of Western blot of lysates of synaptosomes isolated from the dorsal hippocampus of male Nr4ADN (n = 4) and control mice 2 hours after SOR training (n = 4). Unpaired t test: t₆ = 4.011, **P = 0.0070 (Hspa5); t₆ = 2.982, *P = 0.0246 (Pdia6).

Fig. 3.. Hspa5 regulates activity-dependent surface trafficking of NMDA receptor GluN2A.

(A and B) Cultured neurons were transduced with eGFP or Nr4ADN on DIV 16 or 17 and stimulated with KCl before live staining. (A) Costaining for dendrites (MAP2 antibody) and surface GluN2A by immunofluorescence (IF) in transduced cells after KCl stimulation. Scale bar, 10 μm. (B) Quantification of surface staining for GluN2A in (A) as mean fluorescence intensity. Mixed-effect analysis: significant adeno-associated virus (AAV) construct × treatment interaction: F_1,157 = 38.25, P < 0.0001. Sidak’s multiple comparisons test: ***P < 0.0001 (control eGFP versus stimulated eGFP) and ***P < 0.0001 (stimulated eGFP versus stimulated Nr4ADN). (C) Costaining for surface GluN2A and PSD95 by IF in transduced cells following KCl stimulation. Scale bar, 10 μm. (D) Quantification of surface GluN2A and PSD95 colocalization in (C). Mixed-effect analysis: significant AAV construct × treatment interaction: F_1,69 = 14.57, P = 0.0003. Sidak’s multiple comparison tests: **P = 0.0026 (control eGFP versus stimulated eGFP) and **P = 0.0012 (stimulated eGFP versus stimulated Nr4ADN). (E) Cultured neurons were transduced with Nr4ADN or Nr4ADN + Hspa5 on DIV 16 or 17 and stimulated with KCl before live staining. Costaining for surface GluN2A and PSD95 by IF in transduced cells following KCl stimulation. Scale bar, 10 μm. (F) Quantification of surface GluN2A and PSD95 colocalization in (E). Unpaired t test: t₃₀ = 3.727, ***P = 0.0008.

Fig. 4.. Chemical chaperone PBA reverses long-term memory and synaptic plasticity deficits in Nr4ADN mice.

(A) Nr4ADN mice were injected intraperitoneally with PBA (200 mg/kg, n = 9), sodium butyrate [NaBu; 200 mg/kg, n = 8 (2F)], or vehicle [n = 17 (2F)] immediately after SOR training and tested for long-term memory 24 hours later. Two-way ANOVA: significant treatment × session interaction (F_2,31 = 4.207, P = 0.0242). Sidak’s multiple comparisons tests: **P = 0.0011 (Nr4ADN mice-PBA, train versus 24-hour test), **P = 0.0036 (Nr4ADN-PBA 24-hour test versus Nr4ADN-vehicle 24-hour test), and **P = 0.0084 (Nr4ADN-PBA 24-hour test versus Nr4ADN-NaBu 24-hour test). (B) Male control mice were injected intraperitoneally with PBA (200 mg/kg, n = 6) or vehicle (n = 6) immediately after completion of SOR training and tested for long-term memory 24 hours later. Two-way ANOVA: significant main effect of sessions: F_1,10 = 33.46, P = 0.0002. Sidak’s multiple comparisons tests: *P = 0.0110 (control mice-vehicle, train versus test) and **P = 0.0018 (control mice-PBA, train versus test). (C and D) Effects of PBA on persistence of LTP. Expression of Nr4ADN attenuates persistence of LTP in hippocampal slices (Nr4ADN-veh), while bath treatment with 2 mM PBA rescues these LTP deficits (Nr4ADN-PBA) (two-way repeated-measures ANOVA, effect of PBA treatment: F_1,12 = 8.125, P = 0.0146). The mean fEPSP slope over the last 20 min of the recordings was enhanced in PBA-treated slices compared to vehicle-treated slices (PBA-treated: 187.3 ± 16.2%, n = 7 slices, four mice; vehicle-treated: 124.6 ± 14.9%, n = 7 slices, five mice; unpaired t test, *P = 0.0146). Treatment with 2 mM PBA had no significant effect on the baseline responses (pre-drug baseline, 20 min: 100.1 ± 0.13%; post-drug pre-induction baseline, 20 min: 90.63 ± 6.6%; paired t test, P = 0.2094). The representative fEPSP traces shown are sampled at baseline (black) and at the end of the recording (red). Scale bar (2 mV), 10 ms. Error bars indicate SEM.

Fig. 5.. The Nr4a proteins contribute to memory through downstream chaperone proteins.

(A) Schematic of viral constructs used to drive expression of Hspa5 in excitatory neurons of the dorsal hippocampus of male C57BL/6J mice. AAV₉-CaMKIIα-eGFP served as vector control, and AAV₉-CaMKIIα-Hspa5-Tavi was used to drive expression of Hspa5 in excitatory neurons. The Tavi-tag can be identified by an antibody against a consensus biotinylation sequence. (B) Western blot of synaptosomes, showing mild Hspa5-Tavi expression within 1 week of infusion, and expression approximately equal to that of endogenous Hspa5 within 2 weeks of infusion. (C) Quantitation of data in (B). (D) Long-term memory (24-hour) assessment of Nr4ADN mice infused with AAV-eGFP or AAV-Hspa5-Tavi into dorsal hippocampus. Two-way ANOVA: significant AAV type × session interaction: F_1,16 = 6.985, P = 0.0177. Sidak’s multiple comparisons tests: **P = 0.0022 (AAV-Hspa5, train versus test), ***P = 0.0002 (AAV-Hspa5, 24-hour test versus AAV-eGFP, 24-hour test), while eGFP-infused Nr4ADN mice showed no preference toward the DO. AAV-Hspa5: n = 10 (4F) and AAV-eGFP: n = 8 (3F). (E) Schematic illustration of model wherein learning-induced expression of Nr4a1 drives Hspa5 expression to initiate protein folding in the ER that enables the expression of functional proteins at the synapse.

Fig. 6.. Down-regulation of Nr4a factors and their effector genes is linked to ADRD pathology.

(A and B) Expression profiles of NR4A1, NR4A2, and NR4A3 in the hippocampus, from the Allen Brain Institute Study of Aging, Dementia, and Traumatic Brain Injury, correlated with (A) Cerad scores, which reflect the density of neuritic plaques, and (B) Braak stages, which reflect the severity of neurofibrillary tangles. (C) Schematic depiction of control (CaMKIIα-tTA) or rTg4510 (CaMKIIα-tTA and TetO-hMAPT P301L) mice and Western blots of tau phosphorylation (AT8, phosphorylation at both Ser²⁰² and Thr²⁰⁵) in the dorsal hippocampus in the presence or absence of doxycycline (Dox). (D) Long-term memory in rTg4510 and control mice at 4 months of age, following training in SOR. Two-way ANOVA: significant main effect of genotype/treatment (F_2,29 = 4.792, P = 0.0159) and significant main effect of sessions (F_1,29 = 9.221, P = 0.0050). Sidak’s multiple comparisons test: **P = 0.0096 (control mice, train versus test), *P = 0.0439 (rTg4510-Dox mice, train versus test), **P = 0.0016 (rTg4510 mice, 24-hour test versus control mice, 24-hour test), and *P = 0.0140 (rTg4510 mice, 24-hour test versus rTg4510 Dox, 24-hour test). Control, n = 11 (3F); rTg4510, n = 8 (4F); and rTg4510 + Dox, n = 13 (7F). (E and F) rTg4510 and control mice (n = 9 per group) were trained in SOR, and 2 hours later, the dorsal hippocampus was collected and processed for RNA extraction and the analysis of gene expression. (E) Nr4a subfamily gene expression: unpaired t test: Nr4a1: t₁₆ = 4.878, ***P = 0.0002; Nr4a2: t₁₆ = 2.621, *P = 0.0185. (F) Hspa5 and Pdia6 gene expression: unpaired t test: Hspa5: t₁₆ = 3.692, **P = 0.0020; Pdia6: t₁₆ = 4.177, ***P = 0.0007.

Fig. 7.. Restoration of Nr4a1 or ER chaperone function reverses memory deficits in a mouse model of ADRD.

(A) Schematic depiction of AAV constructs used to drive expression of Nr4a1 in excitatory neurons in the dorsal hippocampus of 3-month-old rTg4510 mice. (B) Western blot showing expression of virally transduced Nr4a1-HA in dorsal hippocampus 4 weeks following infusion. (C) rTg4510 mice 3 months of age were infused with AAV₉-CaMKIIα-Nr4a1-HA or control vector (AAV₉-CaMKIIα-eGFP), and 4 weeks later, they were trained in SOR. Long-term memory was tested 24 hours after the training session. Two-way ANOVA: significant main effect of AAV-type infusion (F_1,13 = 6.597, P = 0.0234). Sidak’s multiple comparison tests: *P = 0.0127 (AAV-Nr4a1, 24-hour test versus AAV-eGFP, 24-hour test) and *P = 0.0489 (AAV-Nr4a1 train versus AAV-Nr4a1 24-hour test). AAV-Nr4A1: n = 8; AAV-eGFP: n = 7. (D) rTg4510 mice 3.5 months of age were infused with AAV₉CaMKIIα-Hspa5-Tavi or control vector (AAV₉-CaMKIIα-eGFP), and 2 weeks later, they were trained in SOR. Long-term memory was tested 24 hours after the training session. Two-way ANOVA: significant session × AAV interaction, F_1,10 = 8.767, P = 0.0143. Sidak’s multiple comparison tests: *P = 0.0419 (AAV-eGFP test versus AAV-Hspa5 test) and *P = 0.0419 (AAV-Hspa5 train versus AAV-Hspa5 test). AAV-eGFP: n = 6 (3F); AAV-Hspa5: n = 6 (3F).