An inhibitory circuit-based enhancer of DYRK1A function reverses Dyrk1a-associated impairment in social recognition

Abstract

Heterozygous mutations in the dual-specificity tyrosine phosphorylation-regulated kinase 1a (Dyrk1a) gene define a syndromic form of autism spectrum disorder. The synaptic and circuit mechanisms mediating DYRK1A functions in social cognition are unclear. Here, we identify a social experience-sensitive mechanism in hippocampal mossy fiber-parvalbumin interneuron (PV IN) synapses by which DYRK1A recruits feedforward inhibition of CA3 and CA2 to promote social recognition. We employ genetic epistasis logic to identify a cytoskeletal protein, ABLIM3, as a synaptic substrate of DYRK1A. We demonstrate that Ablim3 downregulation in dentate granule cells of adult heterozygous Dyrk1a mice is sufficient to restore PV IN-mediated inhibition of CA3 and CA2 and social recognition. Acute chemogenetic activation of PV INs in CA3/CA2 of adult heterozygous Dyrk1a mice also rescued social recognition. Together, these findings illustrate how targeting DYRK1A synaptic and circuit substrates as "enhancers of DYRK1A function" harbors the potential to reverse Dyrk1a haploinsufficiency-associated circuit and cognition impairments.

Keywords: Ablim3; CA2; CA3; Dyrk1a; autism spectrum disorder; dentate gyrus; epistasis; feedforward inhibition; hippocampus; parvalbumin interneuron; social cognition.

Figure 1.. Dyrk1a is required in DGCs of adult mice for FFI connectivity, PV IN perisomatic contacts and social recognition

(A) Experimental design. Lentiviruses expressing GFP (CaMKIIα-GFP) were injected into the DG of 2-month old WT mice 2 weeks prior to habituation to a chamber (Context) for 7 days. On day 8 mice were exposed to the familiarized context or familiarized context with a social stimulus (novel juvenile mouse). (B) Representative images showing GFP⁺ MFTs (filopodial extensions, arrows) and quantification of the number of filopodia per vGLUT1⁺ MFTs (N = 8 mice per group). Scale bar, 5 μm. **p; < 0.01 using two-tailed unpaired t test with Welch’s correction. (C) Representative images and quantification of PV⁺ puncta density in CA2/CA3 stratum pyramidale (N =4 mice per group). Top, RGS14⁺ labeling defines CA2 subfield. Scale bar, 10 μm. *p < 0.05 using two-tailed unpaired t test with Welch’s correction (D) Images showing co-immunostaining of DYRK1A, ABLIM3 and ZO1 in hippocampus. Scale bar, 100 μm. Inset shows high magnification of dorsal CA3 subregion outlined in image. Scale bar, 5 μm. (E) Lentiviruses expressing CaMKIIα-Cre-T2A-GFP were injected into DG of 2-month old Dyrk1a^{f/f, f/+ and +/+} mice 2 weeks prior to processing for morphology analysis or use in social cognition behavioral paradigm. Representative images showing GFP⁺ MFTs (filopodia extensions, arrows) and quantification of the number of vGLUT1⁺ filopodia per MFT in CA3 (left) and PV⁺ puncta density in CA3 (right)(N = 5 for mice +/+; 4 mice for f/+; 6 mice for f/f). Scale bar, 5 μm (left); 10 μm (right). **p < 0.01; ***p < 0.001 using one-way ANOVA with Bonferroni post hoc test. (F) Schematic of social cognition task depicting social recognition (encoding: trial1, T1 and two trials of familiarization to social stimulus, T2-T3) and social memory discrimination trial (T4), each trial separated by 5 min intervals. (G) Quantification of total interaction time during T1-T3. Blue shaded box outlines social recognition phase (T1). (H) Quantification of empty cup vs. social stimulus interaction time during social recognition phase (T1) (left panel, blue shaded) and the discrimination trial (T4, right panel, yellow shaded) (N =13 mice per group). Emp, empty pencil cup; stim, stimulus mouse; fam, familiar mouse; novel, a new novel stimulus mouse. **p < 0.01 using two-way ANOVA with Bonferroni post hoc test. All experiments are performed in male mice and all data are displayed as mean ± SEM. See also Figure S1.

Figure 2.. ABLIM3 functions downstream of DYRK1A in mossy fibers to regulate FFI connectivity, PV IN perisomatic contacts and social recognition

(A) Schematic of experimental design and quantification of ABLIM3 levels in mossy fiber terminals in EIIaCre;Dyrk1a^{+/+ or f/+} mice (referred to as Dyrk1a^{+/+ or +/−}). (B) Representative images of ABLIM3 immunoreactivity in mossy fiber terminals in Dyrk1a^{+/+ or +/−} mice. Dashed lines indicate ABLIM3 in mossy fibers. Scale bar, 100 μm. (C) Quantification of ABLIM3 levels in mossy fiber terminals (MFTs) in 2-3 month old mice (In Dyrk1a^+/+, N = 5 mice for each group; in Dyrk1a^+/−, N = 3 mice for each group). **p < 0.01 using two-way ANOVA with Bonferroni post hoc test. (D) Conserved DYRK1A phosphorylation site in ABLIM3 based on consensus sequence. (E-F) Representative images of MFTs and PV puncta taken from mice following lentiviral overexpression of GFP (Ctrl), ABLIM3 wild-type (WT), phosphomimetic ABLIM3 (SD) or phospho-dead ABLIM3 (SA) mutants in DG of 2-month old WT mice. Representative images and quantification of number of filopodia per vGLUT1⁺ MFT in CA3 (top) and PV⁺ puncta density in CA3 (bottom). Scale bar, 5 μm (top); 10 μm (bottom). (N = 5 mice for vector control; 4 mice for WT; 5 mice for SD; 5 mice for SA). ***p < 0.001 using one-way ANOVA with Bonferroni post hoc test. (G) Working model for how DYRK1A-dependent ABLIM3 phosphorylation in MFTs recruits PV mediated perisomatic inhibition in CA3/CA2. Yellow color denotes DYRK1A levels and peach color denotes ABLIM3 levels in DGCs and mossy fibers. (H) Schematic of lentivirus injection of U6-Ablim3 shRNA-Ubi-Cre-T2A-GFP-cassette into DG of 2-month old Dyrk1a^{+/+, f/+ or f/f} mice. Representative images showing GFP⁺ MFTs (filopodia extensions, arrows, top) PV⁺ puncta density in CA3 (bottom). Scale bar, 5 μm (top); 10 μm (bottom). (I-J) Quantification of the number of vGLUT1⁺ filopodia per MFT (H) and PV⁺ puncta density (I) in CA3 stratum pyramidale (N = 5 mice for each group). p > 0.05 using one-way ANOVA with Bonferroni post hoc test. (K) Quantification of interaction time during social recognition phase (T1, left panel, blue shaded) and social memory discrimination phase (T4, right panel, yellow shaded). 2-month old mice injected with shNT (N = 9 mice for +/+; 5 mice for f/+; 9 mice for f/f mice) or shRNA (N = 8 mice for +/+; 6 mice for f/+; 9 mice for f/f mice). Emp, empty pencil cup; stim, stimulus mouse; fam, familiar mouse; novel, a new novel stimulus mouse. *p < 0.05; **p < 0.01 using two-way ANOVA with Bonferroni post hoc test. These experiments are performed in male mice (C, F, K) or male and female (I-J). All data are displayed as mean ± SEM. See also Figure S2-S4.

Figure 3.. Ablim3 downregulation in DGCs of adult heterozygous Dyrk1a mice restores FFI connectivity and PV IN perisomatic contacts in CA3 and CA2

(A) Lentiviruses expressing Ablim3 shRNA-GFP or non-targeting shRNA (shNT-GFP) were injected into DG of 2-3 month old Dyrk1a^{+/+ or +/−} mice. Representative images showing viral expression of GFP in DG and RGS14 immunostaining to delineate CA2. Scale bar, 500 μm. (B) Representative images showing GFP+ MFTs and quantification of the number of vGLUT1⁺ filopodia per MFT (in Dyrk1a^+/+, N = 3 mice for shNT; N=4 mice for shRNA; in Dyrk1a^+/−, N=4 mice for each group) in CA3. Scale bar, 5 μm. **p < 0.01 using two-way ANOVA with Bonferroni post hoc test. (C-D) Representative images and quantification of PV⁺ puncta density in CA3 (C) and CA2 (D) subfields. Scale bar, 10 μm. *p < 0.05; **p < 0.01 using two-way ANOVA with Bonferroni post hoc test. These experiments are performed in male mice. All data are displayed as mean ± SEM. See also Figure S5.

Figure 4.. Ablim3 downregulation in DGCs of adult heterozygous Dyrk1a mice restores inhibitory synaptic transmission in DG-CA2

(A) Schematic depicting lentiviral-Ablim3-shRNA/shNT and rAAV5-CaMKIIα-ChR2-eYFP injection into DG, and AAV-S5E2-tdTomato injected into CA3/CA2 (left). Representative image of CA2 showing eYFP expression in MFTs in stratum lucidum (SL), tdTomato expressed in PV INs, RGS14 immunostaining of CA2, and immunostaining of biocytin-filled PN and PV INs. Scale bar, 50 μm. (B) Schematic depicting whole-cell patch-clamp recording of miniature excitatory postsynaptic current (mEPSC) from CA2 PV IN (left). Representative recording traces from each group (middle). Cumulative probability plots of mEPSC frequency and amplitude from PV INs (Kolmogorov-Smirnov test, *p < 0.05, n=8-9 cells, 2-3 cells per mouse, 3-4 mice per group). (C) Schematic depicting whole-cell patch-clamp recording of miniature inhibitory postsynaptic current (mIPSC) from CA2 pyramidal neurons (PN) (left). Representative traces from each group (middle). Cumulative probability plots of mIPSC frequency and amplitude from PNs (Kolmogorov-Smirnov test, *p < 0.05, n=8-9 cells, 2-3 cells per mouse, 3-4 mice per group). (D) Schematic depicting whole-cell patch-clamp recording of EPSC and IPSC from CA2 PN to paired pulse optical (473 nm) stimulation. Representative traces from each group and traces after perfusion of DCG-IV. Bar graphs indicate excitation to inhibition ratio, the amplitude of the first EPSC and IPSC response to paired pulse optical stimulation, and the EPSC and IPSC paired pulse ratios. (Two-way ANOVA with Tukey posthoc, *p < 0.05, n=8-11 cells, 2-3 cells per mouse, 3-4 mice per group). (E) Schematic depicting whole-cell patch-clamp recording of CA2 PV IN in current-clamp configuration. Representative traces from each group depict action potential response to current steps (200 pA, 500 ms). Line graph indicates number of potential responses to incremental current steps (Two-way RM ANOVA with Tukey posthoc, *p < 0.05, n=11-12 cells, 2-3 cells per mouse, 5-6 mice per group). All data are displayed as mean ± SEM. These experiments are performed in male and female mice. See also Figure S6.

Figure 5.. Ablim3 downregulation in DGCs of adult heterozygous Dyrk1a mice restores inhibitory synaptic transmission in DG-CA3

(A) Schematic depicting lentiviral-Ablim3-shRNA/shNT and rAAV₅-CaMKIIα-ChR2-eYFP injection into DG, and AAV-S5E2-tdTomato injected into CA3/CA2 (left). Representative image of CA2 showing eYFP expression in MFTs in the stratum lucidum (SL), tdTomato expressed in PV⁺ INs, and immunostaining of biocytin-filled PN and PV⁺ INs. Scale bar, 50 μm. (B) Schematic depicting whole-cell patch-clamp recording of mEPSC from CA3 PV IN (left). Representative recording traces from each group (middle). Cumulative probability plots of mEPSC frequency and amplitude from PV INs (Kolmogorov-Smirnov test, *p < 0.05, n=8-9 cells, 2-3 cells per mouse, 3-4 mice per group). (C) Schematic depicting whole-cell patch-clamp recording of mIPSC from CA3 PN (left). Representative traces from each group (middle). Cumulative probability plots of mIPSC frequency and amplitude from PNs (Kolmogorov-Smirnov test, *p < 0.05, n=8-9 cells, 2-3 cells per mouse, 3-4 mice per group). (D) Schematic depicting whole-cell patch-clamp recording of EPSC and IPSC from CA3 PN to paired pulse optical (473 nm) stimulation. Representative traces from each group and traces after perfusion of DCG-IV. Bar graphs indicate excitation to inhibition ratio, the amplitude of the first EPSC and IPSC response to paired pulse optical stimulation, and the EPSC and IPSC paired pulse ratios. (Two-way ANOVA with Tukey posthoc, *p < 0.05, n=8-11 cells, 2-3 cells per mouse, 3-4 mice per group). (E) Schematic depicting whole-cell patch-clamp recording of CA3 PV⁺ IN in current-clamp configuration. Representative traces from each group depict action potential response to current steps (200 pA, 500 ms). Line graph indicates number of potential responses to incremental current steps (Two-way RM ANOVA with Tukey posthoc, *p < 0.05, n=13-13 cells, 2-3 cells per mouse, 5-7 mice per group). All data are displayed as mean ± SEM. These experiments are performed in male and female mice. See also Figure S7.

Figure 6.. Ablim3 downregulation in DGCs of adult heterozygous Dyrk1a mice restores social recognition

(A) Lentiviruses expressing Ablim3 shRNA-GFP or non-targeting shRNA (shNT-GFP) were injected into DG of 2-3 month old Dyrk1a^{+/+ or +/−} mice. (B-F) Behavioral analysis of Dyrk1a^+/+ mice injected with shNT (N=11 mice) or shRNA (N=8 mice) and Dyrk1a^+/− mice injected with shNT (N=8 mice) or shRNA (N=9 mice) (see Figure S1F for behavioral testing schedule). (B) OF, open field, quantification of total distance travelled, percentage of distance traveled across the center arena, percentage time in center. (C) LD, light-dark box assay, quantification of time spent in the light compartment (seconds). (D) OR, object recognition, quantification of the time spent sniffing the object (seconds). (E) NOR, novel object recognition, quantification of the time spent sniffing the familiar object (triangle) and novel object (cube). (F) Quantification of cup vs. social stimulus interaction time during social recognition phase (T1) (left panel, blue shaded) and the discrimination trial (T4, right panel, yellow shaded) These experiments are performed in male mice. *p < 0.05; **p < 0.01; ***p < 0.001 using two-way ANOVA with Bonferroni post hoc test. All data are displayed as mean ± SEM.

Figure 7.. Acute chemogenetic activation of PV INs in CA3/CA2 of adult Dyrk1a^+/− mice is sufficient to rescue social recognition

(A) Left, Representative image showing expression of hM3D(Gq)DREADD-tdTom in PV INs in CA3/CA2. Scale bar, 500 μm. Right, Schematic of chemogenetic activation of PV INs schedule. 2-month old Dyrk1a^{+/+ or +/−} mice were injected with AAV-S5E2-Gq virus into dCA2/CA3 2 weeks prior to behavioral testing using a vehicle/CNO cross-over design. Mice were injected with saline or 1mg/kg CNO 30 min prior to behavioral testing. 24 hours later mice were counterbalanced for vehicle/CNO and behaviorally tested. (B-D) Chemogenetic activation of PV INs in CA3/CA2 of does not affect locomotion (OF) or anxiety-like (LD) behavior (B and C) and object recognition (D). (E) Quantification of cup vs. social stimulus interaction time during social recognition phase (T1) (left panel, blue shaded) and the discrimination trial (T4, right panel, yellow shaded) (N= 10 mice for Dyrk1a^+/+ group; N= 7 mice for Dyrk1a^+/− group). These experiments are performed in male and female mice. *p < 0.05; **p < 0.01; ***p < 0.001 using two-way ANOVA with Bonferroni post hoc test. All data are displayed as mean ± SEM. See also Figure S8.