MicroRNA-mediated disruption of dendritogenesis during a critical period of development influences cognitive capacity later in life

Significance

This study expands our knowledge of small noncoding RNAs and their important roles in brain development and cognitive function. We describe a mechanism by which dendritic morphology and synaptic formation are altered during a critical period of development, eventually leading to altered capacity for learning and memory later in life.

Abstract

The prenatal period of cortical development is important for the establishment of neural circuitry and functional connectivity of the brain; however, the molecular mechanisms underlying this process remain unclear. Here we report that disruption of the actin–cytoskeletal network in the developing mouse prefrontal cortex alters dendritic morphogenesis and synapse formation, leading to enhanced formation of fear-related memory in adulthood. These effects are mediated by a brain-enriched microRNA, miR-9, through its negative regulation of diaphanous homologous protein 1 (Diap1), a key organizer of the actin cytoskeletal assembly. Our findings not only revealed important regulation of dendritogenesis and synaptogenesis during early brain development but also demonstrated a tight link between these early developmental events and cognitive functions later in life.

Perturbation during the critical period of perinatal cortical development influences the functional connectivity of the brain and can lead to an increased propensity toward neurological disorders such as anxiety, schizophrenia, and autism spectrum disorders (1–3). Neuronal maturation involves distinct, highly regulated events, including neuronal differentiation and migration, dendritogenesis, axon formation/guidance, and synaptogenesis, among others. However, precisely how these events are regulated and how they orchestrate brain development and cognitive function remain largely unknown.

Fear-related learning and memory play a significant role in the development of anxiety disorders. Cortical dysfunction is associated with emotional disturbances, which are underpinned by impaired fear extinction, and an inefficient termination of physiological stress responses (4, 5). The medial prefrontal cortex (mPFC) is a primary mediator of fear-related learning and memory (6, 7). It is evident that the actin cytoskeleton is involved in synaptic plasticity and neuronal morphogenesis underlying the formation of fear memory. For example, a disruption in the actin cytoskeleton assembly in the adult brain impairs both cued and contextual fear conditioning (8–10), and several actin-regulatory proteins have been shown to be involved in synaptic plasticity and neuronal morphogenesis associated with memory formation (11–15).

Small noncoding RNAs, and miRNAs in particular, have emerged as a major regulatory mechanism that precisely controls the level of gene expression. In invertebrates, miRNAs play essential roles in regulating developmental timing. For example, in Caenorhabditis elegans the succession of certain cell fates from first to second larval stage relies on the induction of miRNA lin-4 expression at the first larval stage and reduction of lin-14 activity via base-pairing interactions with its 3′ UTR (16, 17). In mice, miRNAs have been shown to be involved in rapidly fine-tuning the expression of their target mRNAs and in regulating cognitive function (18, 19). In our study, we found that the expression of the actin polymerization regulator diaphanous homologous protein 1 (Diap1) was significantly down-regulated in mouse cortical neurons from E16 to postnatal day 0 (P0), which is concomitant with a critical period of cortical dendritogenesis. We have also shown that the expression of a brain-enriched miRNA, miR-9, was inversely related to the expression of Diap1 (20). Furthermore, we identified a putative miR-9–binding site in the 3′ UTR of Diap1, indicating a potential regulatory relationship between Diap1 and miR-9 during dendritogenesis. These findings prompted an investigation into whether there is a functional relationship between key regulators of the actin cytoskeleton and miR-9 during prenatal development and whether dendritogenesis during the perinatal period will have an impact on cognition in adult life.

Results

Diap1 Regulates Dendritogenesis and Fear-Related Memory.

Diap1 belongs to an evolutionarily conserved family of formin-related proteins (21), and binding of profilin (Pfn) to Diap1 mediates fast barbed-end elongation that promotes long, unbranched actin filaments (22). To explore a possible role for Diap1 in neuronal maturation, we analyzed the pattern of cortical Diap1 expression from E12 to adulthood. Diap1 was highly expressed at E12 but progressively decreased to low levels around E18 in cortical neurons. The expression of Diap1 increased again from P1 to P7 and decreased afterward (Fig. 1A). In contrast to Diap1, the expression of a brain-specific isoform of Pfn, Pfn2, increased in the prenatal cortex, decreased postnatally, and then remained relatively constant after P7 (Fig. 1A). Diap1 expression anti-correlates with the critical period of cortical dendritogenesis after neuronal migration is complete (23).

Fig. 1.

Diap1 and Pfn2 regulation of dendritic morphology and fear memory is directly regulated by miR-9. (A, Left) Western blot analysis of the temporal expression of Diap1 and Pfn2 in developing and adult cortices. GAPDH is a loading control. (Right) Densitometry analysis of Diap1 and Pfn2 expression. (B, Upper) Schematic representation of in utero injection of plasmid followed by electroporation into mPFC progenitors at the ventricular zone at E14.5. (Scale bar, 50 µm.) (Lower) Knockdown of Diap1/Pfn2 starting from E14.5 increased dendrite complexity in the PL neurons [**P < 0.01, Kruskal–Wallis test; control (CTL) n = 42 neurons; Diap1/Pfn2 KD n = 37 neurons]. (C, Upper) Schematic representation of in utero injection of plasmid followed by auditory fear conditioning. (Lower Left) Knockdown of Diap1/Pfn2 in mPFC facilitated fear memory in adult male mice (**P < 0.01, Kruskal–Wallis test; CTL n = 12; Diap1/Pfn2 KD n = 12). Brains from littermates were fixed and used for anatomic analysis. CTL, control empty plasmid; CS, conditioned stimulus; Diap1/Pfn2 KD, Diap1 and Pfn2 shRNA knockdown constructs; Neg, electroporated without plasmid; US, unconditioned stimulus. (Lower Right) Knockdown of Diap1/Pfn2 has no effect on shock sensitivities (Kruskal–Wallis test, P = 0.6056; CTL n = 12; Diap1/Pfn2 KD n = 12). (D) ISH (Upper) combined with TaqMan-qPCR (Lower) analysis of miR-9 expression in mouse cortex from E14–P40. The mature miR-9 was expressed in the ventricular zone (VZ) and subventricular zone (SVZ) from E14 to adult. miR-9 expression reached highest level around E16–E18. miR-9 was expressed in neocortical neurons at a relatively low level postnatally (P7–P40). CP, cortical plate; SP, subplate. (E) Conservation of miR-9 target sequences in mammalian Diap1 3′ UTRs. (F, Upper) Western blot analysis of miR-9 regulation of Diap1 expression. E14.5 cortical neurons were isolated and transfected with miR-9 overexpression, knockdown, or control lentiviral particles [E14.5 + 6 d in vitro (DIV)]. (Lower) Densitometry analysis of the Western blotting. (G) Relative luciferase activity of reporter genes in miR-9 overexpressed and knocked down HEK293T cells (*P < 0.05, Wilcoxon–Mann–Whitney test). Luc-3′UTR, luciferase reporter constructs of miR-9 targets; Luc-3′UTRm, mutant control of miR-9 luciferase reporter construct; pRL-TK, luciferase internal control.

To examine the role of Diap1 and Pfn2 in regulating dendritic morphogenesis in cortical neurons, we delivered Diap1 and Pfn2 shRNAs into neural progenitors in the mouse mPFC at E14.5 by in utero electroporation (Fig. 1B, Upper and Fig. S1A). The in utero electroporation methodology allows specific targeting of neurons in the prelimbic frontal cortex (PL) of the mPFC, an area of the brain that regulates fear-related learning and memory (7). At P40, we imaged PL pyramidal neurons in cortical layers 2/3 and reconstructed them in 3D. The continuous knockdown of Diap1 and Pfn2 from E14.5 to adulthood led to a dramatic increase in dendritic complexity in the PL (Fig. 1B, Lower).

Fig. S1.

Diap1 and Pfn2 regulation of dendritic morphology and fear memory. (A, Upper) shRNA duplexes of Diap1 and Pfn2 were inserted immediately downstream of a CMV promoter of the plasmid pCAG-RFP. (Lower) Western blot analysis of the protein level of Diap1 and Pfn2 in 293T cells 24 h after transfection. β-Actin (ActB) was used as a loading control. (B) Knockdown of Diap1/Pfn2 shows no significant difference in contextual fear memory among all experimental conditions (Kruskal–Wallis test, P = 0.7988; CTL n = 12; Diap1/Pfn2 KD n = 12). Diap1/Pfn2 KD, Diap1 and Pfn2 shRNA knockdown constructs. (C) Knockdown of Diap1/Pfn2 had no effect on the acquisition of cued fear (Kruskal–Wallis test, P = 0.0655; CTL n = 12; Diap1/Pfn2 KD n = 12). Neg CTL, control electroporated without plasmid; RFP CTL, empty plasmid control.

We examined the effect of Diap1/Pfn2 knockdown on fear memory and found that continuous knockdown of Diap1/Pfn2 in PL neurons significantly enhanced the expression of auditory-cued fear memory (Fig. 1C) without changing pain sensitivity (Fig. 1C, Lower Right). However, we did not observe a dramatic effect on contextual fear memory (Fig. S1B). Given that CD-1 mice are low-freezing mice for both tone and context (24, 25), it is possible that we were not able to see an increase in context fear despite the effects of our experimental manipulations on both cue and context fear. Finally, knockdown of Diap1/Pfn2 in the PL did not appear to alter fear acquisition (Fig. S1C).

miR-9 Directly Regulates Diap1 Expression.

Using in situ hybridization (ISH) combined with immunohistochemistry and qPCR, we found that miR-9 was highly expressed in neural progenitor cells in the germinal zone and within newborn neurons in the cortical plate as early as E14.5 (Fig. 1D and Fig. S2A) (20). The expression of miR-9 gradually increased, reaching its maximum level at E16–E18. This pattern of expression persisted in cortical neurons and in subventricular zone progenitors in the postnatal brain (Fig. S2B). Intriguingly, this temporal expression pattern exhibited an inverse correlation with the expression pattern of Diap1 before P7 (Fig. 1A). Moreover, cortical neurons overexpressing miR-9 formed massive lamellipodia-like structures around the soma (Fig. S2C, Upper). Overexpression of miR-9 in mouse embryonic fibroblasts (MEFs) led to a disorganized pattern of F-actin filaments (Fig. S2C, Lower), which somewhat mimicked the phenotype of F-actin cytoskeleton in HeLa cells expressing deficient Diap1 (26). We further identified an evolutionarily conserved miR-9–binding site in the 3′ UTR of Diap1 mRNA in different species (www.targetscan.org/vert_71/ and www.microrna.org/microrna/home.do) (Fig. 1E).

Fig. S2.

miR-9 regulates the actin cytoskeleton in cortical neurons. (A) ISH shows that miR-9 is specifically labeled. A neuronal specific miRNA, miR-124, and a small noncoding RNA, U6, are positive controls. The scramble sequence is a negative control. (B) Identification of neural populations that express miR-9 by ISH combined with immunostaining analysis. miR-9 expression in cortical neurons can be detected at E14 and persists to adulthood. Tbr1 labels layer 6 neurons (green). Pax6 labels VZ progenitors (red). Tbr2 labels SVZ progenitors (green). (C) miR-9 regulates the actin cytoskeleton in cortical neurons and in MEFs. (Upper) Overexpression of miR-9 in cortical neurons forms lamellipodia-like structures around the soma. Knockdown of miR-9 in cortical neurons shows simplified neurites. (Lower) Overexpression of miR-9 perturbs the integrity of actin filaments in MEFs. F-actin filaments were labeled with cy3-conjugated phalloidin. CTL, the empty lentiviral vector FG12; miR-9, miR-9–overexpressing lentiviral construct; miR-9AS, miR-9 knockdown sponge.

To examine whether there is a functional interaction between miR-9 and Diap1, we analyzed protein levels of Diap1 in cultured neocortical neurons by overexpressing or knocking down miR-9. We found that overexpression of miR-9 decreased Diap1 protein levels by 60%. Conversely, knockdown of miR-9 (miR-9AS) in cultured cortical neurons increased Diap1 protein levels by 50–80% (Fig. 1F). Using a luciferase reporter assay, we found Diap1 to be a direct target of miR-9 (Fig. 1G and Fig. S3A). Moreover, overexpression or knockdown of miR-9 in cultured cortical neurons did not significantly alter the expression levels of related actin cytoskeleton regulators such as WAS protein family member 2 (Wave2), Rho/Rac guanine nucleotide exchange factor 2 (Arhgef2), actin-related protein 2/3 complex, subunit 1A (Arpc1a), protein phosphatase 1, regulatory subunit 12A (pPP1R12A), and P21 protein (Cdc42/Rac)-activated kinase 4 (PAK4), which have predicted miR-9–binding sites in their mRNA 3′ UTRs (Fig. S3B). Interestingly, although Pfn2 expression does not anti-correlate with miR-9 expression in either developing or adult cortices, miR-9 overexpression in cultured E14.5 cortical neurons still significantly down-regulates Pfn2 expression, suggesting that, although miR-9 can regulate Pfn2 expression, Pfn2 is also under the regulation of additional factors irrelevant to miR-9 (Fig. S3 C and D).

Fig. S3.

miR-9 regulation of cytoskeleton-related genes. (A) To construct the Diap1 3′ UTR luciferase control plasmid (Luc-3′UTRm), the miR-9 seed-binding site in the Diap1 3′ UTR was replaced by a scrambled sequence using an overlapping PCR approach. (B) Western blot analysis of actin cytoskeleton-related genes Wave2, Arthgef2, Arpc1A, Ppp1r12a, and PAK4 expression in lysates from cortical neurons (E13 + 6 DIV) transfected with miR-9–overexpressing, knockdown, and control lentiviral particles. β-Actin (Actb) is a loading control. (C, Upper) Western blot analysis of endogenous Pfn2 expression in cortical neurons isolated from E14 embryos (miR-9–overexpressing, knockdown, and control lentiviral particles infected 4 h after plating, and cultured for 8 d). (Lower) Densitometry analysis of Western blotting. (D, Lower) Relative luciferase activity of reporter genes in miR-9–overexpressing and knocked-down HEK293T cells. (Upper) miR-9 seed-binding site mutations in the Pfn2 3′ UTR of the luciferase reporter construct (Luc-3′UTRm). *P < 0.05, Wilcoxon–Mann–Whitney test. CTL, control plasmid FG12; Luc-3′UTR, luciferase reporter constructs of miR-9 targets; Luc-3′UTRm, mutant control of the miR-9 luciferase reporter construct; miR-9, miR-9–overexpressing construct; miR-9AS, miR-9 bulged antisense construct; pRL-TK, luciferase internal control.

miR-9 Regulates Dendritic Structure and Synaptic Formation During Early Brain Development and Influences Fear Memory in Adult Mice.

To elucidate the role of miR-9 in the regulation of fear-related learning and memory, we delivered miR-9 overexpression, knockdown, or the control plasmid, respectively, into PL progenitors at E14.5 by in utero electroporation. At P40, an auditory-cued fear-conditioning test revealed a significant enhancement of the expression of fear memory in mice that had been electroporated with the miR-9 overexpression construct. Conversely, knockdown of miR-9 prevented the formation of fear memory in adult mice (Fig. 2A, Left). There was no effect on contextual fear memory (Fig. S4A). Fear acquisition and foot-shock reactivity did not differ among miR-9, miR-9AS, and control groups (Fig. 2A, Right and Fig. S4B). These data demonstrate that miR-9, a negative regulator for Diap1, also regulates fear memory.

Fig. 2.

miR-9 modulates dendritic morphogenesis, number of synaptic puncta, and spin density that may contribute to memory formation. (A, Left) Cued fear memory is significantly enhanced in mice that had been electroporated with miR-9 at E14.5. miR-9AS administered at E14.5 prevented the formation of fear memory in adult mice (**P < 0.01, *P < 0.05, Kruskal–Wallis test; CTL n = 16; miR-9 n = 8, miR-9AS n = 12). (Right) Shock sensitivity did not differ among different groups (Kruskal–Wallis test, P = 0.8208; CTL n = 16; miR-9 n = 8; miR-9AS n = 12). (B) Overexpression or knockdown of miR-9 or overexpression of Diap1 in PL cortical neurons starting from dendritogenesis altered dendritic morphology (*P < 0.05, **P < 0.01, Kruskal–Wallis test; CTL n = 29 neurons; miR-9 n = 33 neurons; miR-9AS n = 30 neurons; Diap1 n = 44 neurons).(Scale bar, 40 µm.) Diap1, Diap1 overexpression construct. (C, Upper) Overexpression or knockdown of miR-9 or overexpression of Diap1 starting from E14.5 in PL cortical neurons altered synaptic formation (**P < 0.01, Kruskal–Wallis test; CTL n = 14 neurons; miR-9 n = 17 neurons; miR-9AS n = 16 neurons; Diap1 n = 14 neurons). Shown are examples of a dendrite (GFP⁺) stained with synaptic markers. (Scale bar, 5 µm.) Presynaptic terminals were identified by staining for synapsin (Syn, red). Postsynaptic structures were identified by staining for postsynaptic density protein 95 (PSD95, blue). Synapses were identified by the close proximity of pre- and postsynaptic elements (≤0.5 µm, arrowheads in the bottom images). (Lower) 3D construction of PSD-95 (blue) and presynaptic synapsin (red) identified using the “create spots” algorithm in Imaris. Insets show boxed areas at high magnification. (Scale bar, 1 µm.) (D, Left) Quantification of the number of protrusions per 10 µm of dendrites. (Scale bar, 2 µm.) (Right) Overexpression or knockdown of miR-9 or overexpression of Diap1 in PL cortical neurons affected dendritic spine formation (**P < 0.01, Mann–Whitney test; CTL, n = 28 neurons; miR-9 n = 42 neurons; miR-9AS n = 38 neurons; Diap1 n = 19 neurons).

Fig. S4.

Perturbation of miR-9 expression in the mPFC does not affect contextual fear memory and fear acquisition. (A) The contextual fear memory did not show significant difference in all experimental conditions (Kruskal–Wallis test, P = 0.8268; CTL n = 16; miR-9 n = 8; miR-9AS n = 12). (B) Repeated pairings of the conditioned stimulus (CS) and unconditioned stimulus (US) increasingly elicited freezing in all groups, suggesting no significant differences in fear acquisition during training (Kruskal–Wallis test, P = 0.3960; CTL n = 16; miR-9 n = 8; miR-9AS n = 12).

We subsequently examined whether exogenous miR-9 overexpression from E14.5 would lead to any changes in dendritic morphology as well as synaptogenesis of PL neurons. At P40 we imaged PL pyramidal neurons in cortical layers 2/3 and reconstructed them in 3D. In agreement with the Diap1 and Pfn2 double-knockdown data, overexpression of miR-9 from E14.5 led to an increase in the number of dendrite branch points in PL neurons and in cultured cortical neurons, whereas knockdown of endogenous miR-9 or overexpression of Diap1 (lack of the miR-9–binding sequence) led to a significant reduction in the number of dendrite branch points (Fig. 2B and Fig. S5). Importantly, we found that overexpression of miR-9 increased synaptic density in PL neurons. In contrast, knockdown of miR-9 or overexpression of Diap1 impaired synaptic formation (Fig. 2C). Moreover, we showed that miR-9 increased the number of dendritic spines, whereas knockdown of miR-9 or overexpression Diap1 decreased their number (Fig. 2D). Taken together, these data demonstrate an essential role for miR-9 and Diap1 in the regulation of dendritic morphogenesis and synaptogenesis in the developing brain that may contribute to fear learning and memory later in life.

Fig. S5.

miR-9 modulates dendritic morphogenesis in cortical neurons. (A) miR-9 regulates dendritic morphology in cortical neurons in vitro. (Left) A schematic representation of the in vitro study of dendritic morphology. Dendritic complexity increased significantly in neurons overexpressing miR-9, whereas repression of miR-9 activity reduced dendritic complexity. *P < 0.05, Wilcoxon–Mann–Whitney test. (B) Overexpression of miR-9 in cortical neurons in the somatosensory region. **P < 0.01, Kruskal-Wallis test. (C) Semi-in vitro gain- and loss-of-function studies of miR-9 in neocortical neurons. Overexpression of miR-9 led to a significant increase in dendritic density, whereas knockdown of miR-9 dramatically reduced dendritic complexity. GFP CTL, control plasmid FG12; miR-9, miR-9–overexpressing plasmid; miR-9AS, miR-9 knockdown plasmid. **P < 0.01, Kruskal–Wallis test.

Coexpression of Diap1 Rescues the Dendritic Phenotype as Well as Alterations in Fear-Related Learning and Memory Deficit Elicited by miR-9 Overexpression.

To determine whether miR-9 indeed functions as a negative regulator for Diap1, we carried out rescue experiments to examine the effects of coexpression of Diap1 and miR-9 on dendritic morphogenesis and synaptic formation in cortical neurons. The mRNA of the overexpressed Diap1 lacks the miR-9 3′ UTR binding site (Fig. S6). We showed that coexpression of Diap1 and miR-9 normalized dendritic branching morphology in both primary cortical neurons and in the pyramidal neurons of PL layers 2/3. Overexpression of miR-9 increased dendritic complexity, whereas overexpression of Diap1 alone in cortical neurons reduced dendritic branch points in vitro and in vivo (Fig. 3 A and B). Quantification of synapses revealed a reversal of the number of synaptic puncta to control levels in cultured cortical neurons that coexpressed exogenous Diap1 and miR-9. Knockdown of miR-9 and overexpression of Diap1 reduced the number of synapses (Fig. 3C). We further demonstrated that co-overexpression of Diap1 with miR-9 in the PL reverses the freezing phenotype that resulted from either miR-9 or Diap1 overexpression alone (Fig. 3D).

Fig. S6.

The expression efficiency of the Diap1-overexpressing construct. (Upper) Schematic of in-frame fusion of the Diap1-T2A-GFP construct to overexpress Diap1. GFP separated from Diap1 by T2A-mediated ribosomal skipping drives Diap1 expression. (Lower) Western blot analysis of the protein level of Diap1 in 293T cells 36 h after transfection. GAPDH is a loading control.

Fig. 3.

Coexpression of Diap1 in miR-9–overexpressing cortical neurons normalizes the number of dendritic branches and synaptic puncta and the increased fear memory elicited by the overexpression of miR-9. (A and B) Coexpression of Diap1 and miR-9 in cortical neurons rescued the dendritic morphology in cultured cortical neurons (CTL n = 13; Diap1 n = 23; miR-9 n = 16; miR-9/Diap1 n = 22) (A) and in the PL pyramidal neurons in mice (CTL n = 10; Diap1 n = 66; miR-9 n = 19; miR-9/Diap1 n = 34) (B). **P < 0.01, Kruskal–Wallis test. Anti-GFP immunostaining is shown in red. (Scale bars, 40 µm.) (C) Quantification of synapse in cultured cortical neurons visualized by immunocytochemistry. Overexpression or knockdown of miR-9 or overexpression of Diap1 starting from E15 altered synaptic formation in cortical neurons. Coexpression of miR-9 in Diap1-overexpressing cortical neurons elevated the number of synaptic puncta in cortical neurons. (Scale bar, 5 µm.) Insets in the bottom panels show the boxed area at high magnification. (Scale bar, 1 µm.) **P < 0.01, Kruskal–Wallis test. (D) Coexpression of Diap1 and miR-9 in mPFC cortical neurons reduced the increase in freezing level elicited by the overexpression of miR-9 alone in an auditory fear-conditioning paradigm (Kruskal–Wallis test, *P < 0.05, **P < 0.01; miR-9 n = 6; Diap1 n = 4; Diap1/miR-9 n = 12).

Interestingly, knockdown of Pfn2 alone does not significantly enhance animal freezing behavior, whereas knocking down Diap1 alone and only during the developmental period (E14.5–P0) via a doxycycline (Dox)-inducible driver is sufficient to promote dendritic complexity and facilitate fear-memory–related freezing behavior (Figs. S7 and S8). Moreover, restricted overexpression of miR-9 only during development (E14.5–P0) is also sufficient to lead to increased dendritic complexity in the PL neurons and enhanced fear memory (Figs. S7 and S8). Taken together, although both Pfn2 and Diap1 are bona fide targets for miR-9, only Diap1 mainly functions in regulating fear memory. In addition, our data clearly demonstrate a tight functional link between Diap1 and its modulator miR-9 in establishing neural circuits during the critical period of neuronal development. Perturbation of this process has life-long effects on fear-related learning and memory.

Fig. S7.

The sensitivity and specificity of Dox-inducible constructs. (A, Upper) miR-9 or miR-9AS was inserted downstream of the tet-repressor–regulated U6 promoter (pTight/U6). (Lower) Relative fold changes of luciferase activity. The miR-9 luciferase construct was cotransfected into HEK239T cells with the Dox-inducible miR-9–overexpressing plasmid (pSingle-miR-9) and/or miR-9 knockdown plasmid (pSingle-miR-9AS). A Renilla luciferase plasmid, pRL-TK, was used as an internal control. **P < 0.05, Wilcoxon–Mann–Whitney test. (B) Luciferase assay of miR-9 Dox-inducible constructs. There is no significant background expression of miR-9 in the absence of Dox. (C, Top) The bidirectional miR-9 reporter construct. Three perfect miR-9–binding sites were inserted immediately downstream of the d2eGFP sequence. (Middle) d2eGFP is undetectable after 2 d of Dox-induced miR-9 overexpression. The miR-9 reporter and the inducible plasmid pSingle-miR-9 were cotransfected into HEK293T cells. A final Dox concentration of 10 µg/mL was used to induce miR-9 overexpression. (Bottom) The recovery of d2eGFP expression 3 d after the withdrawal of Dox shows that the overexpression of miR-9 can be reversed by the removal of Dox. (D) The efficiency of the Dox-inducible Diap1 knockdown construct. (Upper) The schematic of the Diap1 shRNA knockdown construct. (Lower) Western blot analysis of the protein level of Diap1 in 293T cells 36 h after transfection. β-Actin (ActB) was used as the loading control.

Fig. S8.

Spatiotemporal-restricted manipulation of Diap1 and its regulator miR-9 in cortical neurons during the period of dendritic arborization significantly impaired dendritic morphogenesis, influencing fear-related learning and memory later in life. (A) Temporal-restricted perturbation of Diap1 and miR-9 expression in the PL neurons during dendritogenesis impaired dendritic morphology in adult mice (**P < 0.01, Kruskal–Wallis test; Diap1 n = 41; Diap1 CTL n = 33; miR-9 CTL n = 14; miR-9 n = 14; miR-9AS n = 13). The Dox-inducible plasmids were delivered to the mPFC progenitors at E14.5 by in utero electroporation. Dox was delivered in the mother’s drinking water at 2 mg/mL from E14–P0. (Scale bar, 40 µm.) (B) Temporal-restricted perturbation of Diap1 and miR-9 in the PL during the stage of dendritogenesis impaired fear memory in adult mice The Dox-inducible plasmids were delivered to the mPFC progenitors at E14.5 by in utero electroporation. The Dox was delivered in the mother’s drinking water at 2 mg/mL for 5 d (E14–P0). The mice were trained using a mild auditory fear-conditioning paradigm at P40 and were tested 2 d after training. (**P < 0.01, Kruskal–Wallis test; Diap1 n = 12; Diap1 CTL n = 13; miR-9 n = 18; miR-9AS n = 13; miR-9 CTL n = 18).

Discussion

Neurodevelopmental disorders can be triggered by environmental and genetic interference with normal brain development. It is well established that there are critical time windows for the establishment of certain types of neural plasticity, and at those times individuals are most vulnerable to external disturbances (27–30). In the current study we provided evidence that miR-9 and Diap1 function as part of an intrinsic program that guides the process of neuronal maturation (i.e., dendritogenesis and synaptogenesis) and has long-lasting effects on dendritic complexity and synaptic density in adulthood. Furthermore, to provide some temporal specificity, we used a Dox-inducible system to knock down only Diap1 or miR-9 or to overexpress miR-9 during the early developmental period, i.e., from E14.5 to P0, and found that temporally restricted perturbation of the miR-9–Diap1 axis was sufficient to provide long-lasting effects in dendritic morphology and fear-related learning and memory behavior (Figs. S7 and S8). These data suggest that a disruption in an intrinsic developmental program from the midembryonic stage has a significant impact on cognitive function in adult life.

Our previous and the current studies have shown that overexpressed miR-9 binds exclusively to endogenous target sequences, although it does not target to its binding sequence with seed-sequence mutations (Fig. 1G). To test the specificity of the miR-9 overexpression construct further, we built miR-9 scramble control (miR-9SC). The scramble sequence of the miR-9 mature sequence was generated using online software (www.genscript.com/tools/create-scrambled-sequence). Using overlapping PCR, the mature miR-9 and its 3p sequences of the original genomic sequence were replaced by its scrambled and scrambled-complementary sequences, respectively (Fig. S9A) (31). The sequence was introduced into the same backbone vector as the miR-9 overexpression vector. In the luciferase assay we found that, as with the empty vector, the scrambled sequence did not bind to miR-9–binding sequence (Fig. S9B). Moreover, overexpression of the miR-9SC does not alter dendritic morphology or synapse formation (Fig. S9 C and D). To determine miR-9 knockdown specificity, we constructed a mutant form of the miR-9 knockdown control by replacing the seed-binding sequence with a random sequence (Fig. S9A). A luciferase assay showed that overexpression of the miR-9 knockdown mutant fragment (miR-9ASMut) did not interfere with luciferase gene expression (Fig. S9B). We further showed that overexpression of miR-9ASMut did not affect dendritic branching or synaptic formation (Fig. S9 C and D), suggesting that the miR-9 knockdown sponge configuration is specific.

Fig. S9.

miR-9 scramble sequence (miR-9SC) and miR-9 knockdown mutant (miR-9ASMut) controls have the same specificity as the empty vector control. (A) The schematic of miR-9–overexpressing control. miR-9SC or miR-9ASMut was inserted immediately downstream of a human H1 promoter of a lentiviral vector, FG12. (B) The luciferase assays show that miR-9SC and miR-9ASMut have no effect on the activity of a firefly luciferase gene with the Diap1 3′ UTR (pIS0-Diap1-3′UTR). pIS0 is an empty luciferase vector without a 3′ UTR sequence. (C and D) Overexpression of miR-9SC and miR-9ASMut does not change dendritic complexity (C) and synaptic density (D) in cultured cortical neurons (E15 + 10 DIV) [Kruskal–Wallis test, *P < 0.05; CTL (FG12), n = 20; miR-9ASMut, n = 19; miR-9SC, n = 13; miR-9, n = 14].

We do have concern that the overexpression of small RNAs may, by itself, have certain cellular and organismic effects that may not depend solely on the specific sequences used. For example, it is possible that overexpression of any exogenous DNA fragment (i.e., GFP, ion channels, noncoding RNAs) in a hemostatic biological system may have some unwanted effects, such as attenuating the transcription and translation machineries of other endogenous genes, changing a cell’s physiological properties. However, with the extensive series of controls described above, we concluded that the nonspecific effects of miR-9 overexpression or knockdown configurations, if any, could contribute little, if at all, to the biological effect we observed. In addition, we are fully aware of the potential off-target issue associated with the shRNA knockdown approach (32). Our Diap1 overexpression and knockdown assays showed that Diap1 regulates dendritogenesis, synapse formation, and fear-related learning and memory. With both constant and temporally restricted gain- and loss-of-function approaches targeting two related endogenous components (Diap1 and miR-9) pointing to the same conclusion, we are confident that miR-9 plays an essential role in the regulation of Diap1 gene expression. Misregulation of miR-9 and Diap1 interaction during the critical period of dendritogenesis led to abnormalities in dendritic complexity and the establishment of neural circuitry, with substantial effects on fear learning and memory later in life. More importantly, Diap1–miR-9 rescue assays showed the function of Diap1 in regulating dendritogenesis, synaptic formation, and fear-related learning and memory (Fig. 3). It remains to be determined whether indirect effects of the overexpression or knockdown of miR-9 or Diap1 on their neighborhood neurons contribute to the neuronal morphology and behavioral phenotypes. In agreement with emerging findings on the important role of noncoding RNAs in rapidly shaping phenotypic outcomes in response to current environmental demands, our findings suggest that perturbation of miR-9 activity during early-life events can elicit sensitization toward subsequent stressors later in life and that this sensitization is manifested as enhanced fear-related learning in adulthood.

Materials and Methods

Plasmid Constructions.

The constructs of the miR-9 overexpression lentiviral plasmid and the miR-9 antisense sponge were built as described in our previous study (20) using Diap1 and pfn2 3′ UTR fragments with predicted miR-9–binding sites that were inserted immediately downstream of the luciferase reporter vector, pIS0. To regulate the timing of miR-9, Diap1, and Pfn2 expression, a miR-9–2 premiRNA fragment, miR-9 bulged antisense oligo-duplex, Diap1 shRNA, or Pfn2 shRNA was inserted downstream of a pTight/U6 promoter of tetON plasmids pSingle-tTs-shRNA and pLenti7.3. To build a miR-9 reporter construct, a short-lived eGFP, d2eGFP, was inserted into a pCAG-RFP-CMV vector to create the pCAG-RFP-CMV-d2eGFP construct. Three miR-9–binding sites then were inserted immediately downstream of d2eGFP to create the pCAG-RFP-CMV-d2eGFP-miR-9 AM reporter vector. To overexpress Diap1, Diap1 was PCR amplified from the MGC cDNA clone (Thermo Scientific) and subcloned to a modified plasmid, FUIGW. To test the specificity of the miR-9 overexpression and miR-9 knockdown constructs, we built miR-9SC and miR-9ASMut (Fig. S9). See SI Materials and Methods for additional details.

Subjects.

For in utero electroporation experiments, E14 ICR CD-1 mice were used (Charles River Laboratories). All behavior testing was conducted when the mice were 5–6 wk old, during the light phase in illuminated testing rooms following protocols approved by the Institutional Animal Care and Use Committee of the University of California, Los Angeles.

ISH.

ISH assays were carried out as described by Zhao et al. (20). See SI Materials and Methods for additional details.

Western Blot Analysis.

The whole cerebral cortex was lysed in 0.7% Nonidet P-40 lysis buffer with 1 mM phenylmethylsulfonyl fluoride (PFSF), 10 mM DTT, and a mixture of protease inhibitors. The antibodies used for Western blotting were rabbit anti-Diap1 (1:1,000; Abcam); mouse anti-Pfn2 antibody (1:500; Santa Cruz); mouse anti–β-actin (1:2,000; Sigma); rabbit anti-WAVE2 (1:1,000; Cell Signaling); rabbit anti–GEF-H1 (Arhgef2) (1:1,000), rabbit anti-PAK4 antibody (1:1,000), and mouse anti-MYPT1 (Ppp1r12a) (1:1,000; BD Bioscience); goat anti-Arpc1a (1:1,000; Abnova); mouse anti–β-actin (1:2,000; Sigma); and mouse anti-GAPDH (1:4,000; GeneTex).

Immunohistochemistry.

The primary antibodies used in this study were rabbit anti–T-box brain 1 antibody (Tbr1) (1:2,000; EMD Millipore); rabbit anti–T-box brain 2 antibody (Tbr2) (1:2,000; Millipore); mouse anti-Pax6 antibody (1:500; The Developmental Studies Hybridoma Bank); rabbit anti-GFP antibody (1:2,000; MBL); rabbit anti-PSD95 and mouse anti-synapsin antibodies (1:1,000; Synaptic Systems); guinea pig anti-PSD95 (1:1,000; Synaptic Systems); and guinea pig anti-synapsin (1:1,000; Synaptic Systems). Images were processed with software Imaris (Bitplane).

miRNA TaqMan-qPCR.

Total RNAs were extracted from the whole cerebral cortex of E14–P40 CD-1 mice or from in vitro-cultured cortical neurons. RT-PCR and qPCR were described in our previous study (20).

Luciferase Assay.

Details are given in SI Materials and Methods.

Primary Neuronal Transformation.

Details are given in SI Materials and Methods.

Administration of Dox.

Dox was administrated in the animal’s drinking water at a concentration of 2 mg/mL (33).

In Utero Electroporation.

The in utero electroporation procedure was carried out as described by Zhao et al. (20). We electroporated the plasmid to a single side of the prelimbic cortex.

Dendrite, Dendritic Spine, and Synaptic Puncta Imaging and Image Processes.

The entire profile of each GFP-labeled neuron to be quantified was acquired using a 25× oil immersion objective without optical zoom, NA = 0.8 (Plan-Apochromat; Zeiss). The dendritic branch point is defined as the number of branch bifurcations in the shortest path from the beginning point to a given point in the dendritic graph. Dendrites of individual neurons for each condition were drawn manually and calculated using the Filament function of Imaris software. The dendritic spines were acquired using a 63× oil immersion objective lens, NA = 1.4 (Planapo; Zeiss). The spines were analyzed as described by Swanger and Bassell (34). The synaptic puncta were acquired using a 63× oil immersion objective lens, NA = 1.4 (Planapo; Zeiss) at a resolution of 1,024 × 512 pixels. A z-step of 0.2-μm intervals was used. The pre- and postsynaptic puncta were analyzed as described by Fogarty et al. (35). See SI Materials and Methods for additional details.

Fear Conditioning.

Cued fear was induced with three pairings of a 2-min, 80-dB, white-noise conditional stimulus (CS) coterminating with a 2-s, 0.4-mA foot shock followed by a 2-min intertrial interval (ITI). Two days after fear acquisition all mice were tested in context B. After a 2-min acclimation, freezing was assessed during two 2-min CS presentations See SI Materials and Methods for additional details.

Pain Sensitivity Assays.

We assessed the pain processing by analyzing shock reactivity during the shock in fear conditioning. The unconditioned response to shock was examined by a motion index of each animal during footshock (Video Freeze Software; Med Associates).

Histology.

At the completion of behavioral testing (for all experiments), the brains were fixed and sectioned. The mice with a majority GFP- or RFP-labeled cells in the PL of the mPFC were used for behavior analysis.

Statistical Analysis.

All statistical significance was assessed using an alpha level of 0.05. Statistical analysis was performed using SAS 9.2 and/or GraphPad Prism software.

SI Materials and Methods

Plasmid Constructions.

To investigate Diap1 and Pfn2 mRNAs as miR-9 direct targets, we PCR-amplified Diap1 and Pfn2 3′ UTR fragments with predicted miR-9–binding sites and inserted the fragment immediately downstream of the luciferase gene. To generate Diap1 and Pfn2 luciferase control plasmids, miR-9 seed-binding regions were point-mutated. Diap1 3′ UTR forward oligo: GGGAGCTCGCAGTGTAGGAGTGGCCTGA; Diap1 3′ UTR reverse oligo: GGGCTAGCAAGGTTCACAAGTCCCATGC; Diap1 3′ UTR mutant forward oligo: CTGTTGTATGGTTGGGGCAGGGTCCGTGTTTG; Diap1 3′ UTR mutant reverse oligo: CTGCCCCAACCATACAACAGCCAGGGCTGGC; Pfn2 3′ UTR forward oligo: GGGAGCTCCAGTGACTGCACTTGGGACA; Pfn2 3′UTR reverse oligo: GGGCTAGCTGAGCTACTGCAATGACAGAATG; Pfn2 3′ UTR mutant forward oligo: GGAGAGGATACGCACTTCCTCCCACGACCTT; Pfn2 3′ UTR mutant reverse oligo: GAGGAAGTGCGTATCCTCTCCAAGCTGCACG.

To knock down endogenous Diap1 and Pfn2, the oligonucleotides contain the specific 19-nt target sequence followed by an 8-nt spacer sequence (TGAATTCT) and a 19-nt sequence complementary to the target. The sense oligonucleotides also contained the 4-bp 5′ end and the 2-bp 3′ end overhangs required for cloning into the Xho1 and Pac1 sites in the vector. Diap1 shRNA forward oligo: TCGAGCTGGTCAGAGCCATGGATAGAATTCAATCCATGGCTCTGACCAGCAT; Diap1 shRNA reverse oligo: CGACCAGTCTCGGTACCTATCTTAAGTTAGGTACCGAGACTGGTCG; Pfn2 shRNA forward oligo: TCGAGAGATAGCCTATACGTTGAAGAATTCATCAACGTATAGGCTATCTCAT; Pfn2 shRNA reverse oligo: CTCTATCGGATATGCAACTTCTTAAGTAGTTGCATATCCGATAGAG. The sense and antisense oligonucleotides were annealed and cloned immediately downstream of a CMV promoter followed by a SV40 polyA signal. The vector pCAG-RFP-CMV was derived from the plasmid pCAG-RFP (a kind gift from Joshua Trachtenberg, Department of Neurobiology, University of California, Los Angeles) by inserting a CMV promoter, multiple cloning sites, and a SV40 polyA signal sequence in an orientation opposite that of the CAG promoter. To regulate the timing of miR-9 expression, a miR-9-2 premiRNA fragment with 80-bp flanking sequences or a miR-9 bulged antisense oligo-duplex was inserted downstream of a pTight/U6 promoter of tetON plasmids pSingle-tTs-shRNA and pLenti7.3. To control the expression level of Diap1 and Pfn2 temporally, the Diap1 shRNA or Pfn2 shRNA oligo-duplex was inserted downstream of a pTight/U6 promoter of tetON plasmid pLenti7.3 (kindly provided by Xiaoqing Liu; Tongji University, Shanghai, China). To build a miR-9 reporter construct, a short-life eGFP (d2eGFP; Clontech) was inserted downstream of the CMV promoter of the pCAG-RFP-CMV vector to create the pCAG-RFP-CMV-d2eGFP construct. Three miR-9–binding sites were then inserted immediately downstream of d2eGFP to create the pCAG-RFP-CMV-d2eGFP-miR-9 AM miR-9 reporter vector. To overexpress Diap1, the Diap1-T2A was PCR amplified from the MGC cDNA clone (Thermo Scientific) using a primer pair containing the T2A sequence and was subcloned to the modified plasmid FUIGW. The T2A sequence is GAGGGCAGAGGAAGTCTTCTAACATGCGGTGACGTGGAGGAGAATCCCGGCCCT. To test the specificity of the miR-9 overexpression (miR-9) and miR-9 knockdown (miR-9AS) constructs, we built miR-9SC and miR-9ASMut (Fig. S9).

Subjects.

For in utero electroporation experiments, E14 ICR CD-1 mice were used (Charles River Laboratories). Following the in utero electroporation, the male mice were housed four per cage, maintained on a 12-h light/dark schedule, and allowed free access to food and water. All behavior testing was conducted when the mice were 5–6 wk old, during the light phase in illuminated testing rooms following protocols approved by the Institutional Animal Care and Use Committee of the University of California, Los Angeles.

ISH.

It has been reported that miR-9 is expressed in the neuroepithelium near the ventricles at E11, in the dorsal telencephalon at E14, and in the SVZ in the adult brain. However, information on miR-9 spatiotemporal expression patterns has not been established, particularly in later embryonic and perinatal development and in the adult stage. To understand the spatiotemporal expression pattern of miR-9 during brain development, we systematically examined miR-9 expression using ISH to measure miR-9 expression from E12 to P40. The Exiqon miRCURY LNA Detection probes were catalog nos. 616486-340 (miR-124), 615819-340 (miR-9), and 699004-340 (scramble-miR).

Western Blot Analysis.

The whole cerebral cortex was lysed in 0.7% Nonidet P-40 lysis buffer with 1 mM PFSF, 10 mM DTT, and a mixture of protease inhibitors. After determination of the concentration, protein samples were denatured and fractionated on an SDS/PAGE gel and then were transferred to a nitrocellulose membrane (Bio-Rad) for immunoblotting. The antibodies used for Western blotting were rabbit anti-Diap1 (1:1,000; Abcam); mouse anti-Pfn2 antibody (1:500; Santa Cruz Biotechnology); mouse anti–β-actin (1:2,000; Sigma); rabbit anti-WAVE2 (1:1,000; Cell Signaling); rabbit anti–GEF-H1 (Arhgef2) (1:1,000; Cell Signaling); rabbit anti-PAK4 antibody (1:1,000; Cell Signaling); mouse anti-MYPT1 (Ppp1r12a) (1:1,000; BD Bioscience); goat anti-Arpc1a (1:1,000; Abnova); mouse anti–β-actin (1:2,000; Sigma); and mouse anti-GAPDH (1:4,000; GeneTex, Inc.). Secondary antibodies were HRP-conjugated goat anti-mouse, donkey anti-goat, and goat anti-rabbit IgG (Calbiochem; EMD Millipore). The signal was detected using an ECL plus chemiluminescence kit (PerkinElmer) on X-Omat Blue film (Kodak).

Immunohistochemistry.

The primary antibodies that were used in this study were rabbit anti-Tbr1 (1:2,000; EMD Millipore); rabbit anti-Tbr2 (1:2,000; Millipore); mouse anti-Pax6 antibody (1:500; The Developmental Studies Hybridoma Bank); rabbit anti-GFP antibody (1:2,000; MBL International); and rabbit anti-PSD95, mouse anti-synapsin antibodies, and guinea pig anti-PSD95 (all 1:1,000; Synaptic Systems GmbH). Fluorophore-conjugated secondary antibodies were purchased from Invitrogen (Molecular Probe, Thermo Fisher Scientific). Images were taken with a Carl Zeiss LSM510 or a Carl Zeiss LSM800 confocal microscope, processed with Imaris software (Bitplane), and composed with Adobe Photoshop.

miRNA TaqMan-qPCR.

Total RNAs were extracted from the whole cerebral cortex of E14–P40 CD-1 mice or from in vitro-cultured cortical neurons. RT-PCR and qPCR were performed using TaqMan microRNA qPCR kits (Applied Biosystems, Thermo Fisher Scientific).

The Bidirectional miR-9 Reporter Assay.

Three perfect miR-9–binding sites were inserted immediately downstream of the d2eGFP sequence. The miR-9 reporter and the inducible plasmid pSingle-miR-9 were cotransfected into HEK293T cells. The final Dox concentration of 10 µg/mL was used to induce miR-9 overexpression. Cells treated with Dox were removed after 2 d in culture.

Administration of Dox.

To induce tTs-dependent miR-9 overexpression or knockdown, Dox was administrated in the animal’s drinking water at a concentration of 2 mg/mL. Dox was dissolved in 1–2% sucrose to mask the bitter taste of Dox. The Dox water was kept in light-sensitive bottles to prevent light-induced degradation and was exchanged every 2 d.

Luciferase Assay.

To test the knockdown efficiency of FG12-miR-9AS (miR-9AS), HEK 293T cells (ATCC) were plated 12–16 h before transfection and were transfected with 40 ng of firefly luciferase target reporter plasmid pIS0-miR-9, 80 ng of each miR-9 overexpression plasmid [FG12-miR-9 (miR-9) and miR-9AS or miR-9 knockdown control plasmid FG12-miR-9ASm (miR-9ASm)], and 2 ng of pRL-TK (Renilla luciferase plasmid; Promega) as an internal control. All assays were performed 24 h after transfection with the dual luciferase assay kit (Promega) using a Bio-Rad luminometer (Bio-Rad). To examine diap1 and pfn2 as miR-9 direct targets, we transfected a diap1 3′ UTR luciferase reporter or a pfn2 3′ UTR luciferase reporter plasmid (20 ng) and miR-9 (80 ng) or miR-9AS (80 ng) into HEK293T cells. Renilla luciferase plasmid pRL-TK (2 ng) was used as an internal control. To test the specificity of miR-9 overexpression and miR-9 knockdown constructs, we used firefly luciferase target reporter plasmid pIS0-Diap1 3′ UTR, 80 ng of miR-9SC miR-9ASMut with 2 ng of pRL-TK as an internal control. To test the efficacy of tetON constructs, the tet-inducible miR-9–overexpressing construct pSingle-miR-9 and/or the miR-9 knockdown construct pSingle-miR-9AS and the luciferase reporter vectors were transfected to HEK293T cells. A final Dox concentration of 10 µg/mL was used to induce miR-9 overexpression or knockdown. Luciferase assays were performed 24 h after transfection. pRL-TK was an internal control.

Cell Cultures.

To gain insight into the mechanism by which miR-9 regulates dendritic morphology, we delivered miR-9, miR-9AS, or control lentiviral particles (University of California, Los Angeles Vector Core) to isolated E13 cortical neurons and cultured them in neurobasal medium plus B27 and l-Gln (Invitrogen, Thermo Fisher Scientific).

HEK293T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) plus 5–10% FBS, l-Gln, and penicillin/streptomycin (Thermo Fisher Scientific). MEFs were cultured in DMEM plus 10% FBS, l-Gln, MEM nonessential amino acids, and penicillin/streptomycin (Thermo Fisher Scientific).

Primary Neuronal Transformation.

To study the role of miR-9 and Diap1 in regulating dendritic morphology and synapse formation, mouse E13 cortical neurons were isolated, and the plasmid (miR-9, miR-9AS, GFP CTL, Diap1 shRNA, Diap1T2AGFP, miR-9SC, or miR-9ASMut) was electroporated into the cells using a Nucleofector device (Lonza). The electroporated neurons were plated and cultured in neurobasal medium supplemented with B27 and l-Gln (Thermo Fisher Scientific).

Dendrite, Dendritic Spine, and Synaptic Puncta Imaging and Image Processes.

Zeiss LSM 510 and LSM 800 confocal microscopes were used to image the sections. GFP was excited using an argon 488-nm laser. The entire profile of each GFP-labeled neuron to be quantified was acquired using a 25× oil immersion objective without optical zoom, NA = 0.8 (Plan-Apochromat; Zeiss) with a frame size of 1,024 × 1,024 pixels or 512 × 512 pixels. The neuron was scanned at 1-µm intervals along the z axis. The dendritic branch point is defined as the number of branch bifurcations in the shortest path from the beginning point to a given point in the dendritic graph. A final high-definition 3D image of the dendrite was achieved by reconstructing these consecutive scans using Imaris software. Dendrites of individual neurons for each condition were drawn manually and calculated using the Filament function of the Imaris software. The range of dendritic branching numbers we obtained in this study is in complete agreement with the previously carefully documented analyses of dendritic morphologies of various cortical neurons.

The dendritic spines were acquired using a 63× oil immersion objective lens, NA = 1.4 (Planapo; Zeiss), and the dendrite within the frame was cropped according to the extent of dendrite with a frame size of 1,024 × 512 pixels. The z-stack interval was 0.1-μm increments through the entire visible dendrite. The spines were analyzed as described by Swanger and Bassell (34). Briefly, in Imaris Surpass mode, a region of interest was selected, and a new filament was created using the manual mode. We selected a dendrite that was distal to a dendritic branching point and void of crossing neurites. A minimum dendrite end diameter of 0.8 μm was used, and a single point was assigned as the dendrite end point. Automatic thresholds were used for dendrite surface rendering. To trace spines, the maximum spine length and minimum spine end diameter were set at 3 μm and 0.3 μm, respectively. Automatic thresholds were used to generate spine seed point and surface rendering. All automatically generated spines were confirmed/corrected manually.

The synaptic puncta were acquired using a 63× oil immersion objective lens, NA = 1.4 (Planapo; Zeiss) at a resolution of 1,024 × 512 pixels. A z-step of 0.2-μm intervals was used. The pre- and postsynaptic puncta were analyzed as described by Fogarty et al. (35). Briefly, the dendrite was marked by the Imaris create surface tool. The pre- and postsynaptic puncta were filtered based on their distance from the dendritic surface. All presynaptic florescence signals (synapsin) inside the dendrite and all postsynaptic florescence signals (PSD95) outside the dendrite were removed. The puncta were analyzed by the Imaris spot analysis tool. Puncta with diameters between 0.25 and 0.8 μm and distances of ≤0.5 μm from the dendritic surface were selected for synapse quantification analysis. To estimate the synaptic number, we analyzed pre- and postsynaptic spots opposed to each other within a threshold distance of 0.5 μm. All figures were composed with Adobe Photoshop.

Fear Conditioning.

To be able to detect both increases and decreases in fear expression, we used an auditory fear-conditioning protocol that normally produces low levels of fear memory in control mice. Two contexts (A and B) in separate rooms were used for cued-fear conditioning. Shuttle box compartments (Med Associates) measuring 20.3 × 15.9 × 21.3 cm served as context A, and box compartment (Med Associates) measuring 30.5 × 24.1 × 21 cm served as context B. Cued fear was induced with three pairings of a 2-min, 80-dB, white-noise CS coterminating with a 2-s, 0.4-mA foot shock followed by a 2-min ITI in context A. Both contexts had two transparent walls and stainless steel grid floors (3.2 mm in diameter, 8-mm centers); however, the grid floors in context B were covered with flat white acrylic inserts to minimize context generalization. Before testing context A was wiped down with 10% ethanol, and context B was wiped down with 10% methanol. Individual video cameras were mounted in the ceiling of each chamber and connected via a quad processor to a standard videocassette recorder and monitor for videotaping and scoring of freezing. Grid floors were connected to a scrambled shock source (Med Associates). Auditory stimuli (Med Associates) were delivered via a speaker in the chamber wall. Delivery of stimuli was controlled with a personal computer and Med-PC software through a Smart CTL Interface System (DIG-716; Med Associates).

Two days after fear acquisition all mice were tested in context B. After a 2-min acclimation, freezing was assessed during two 2-min CS presentations. At least 20 mice from each experimental condition were used for behavior testing. For behavioral freezing, the absence of all nonrespiratory movement was rated during all phases by an experienced investigator blind to subject treatment, using a 5-s instantaneous time-sampling technique. One day after the cue fear memory test, miR-9 mice and Diap1 mice received an 8-min context test in the training contexts.

Pain Sensitivity Assays.

We assessed the pain processing by analyzing shock reactivity during the shock in fear conditioning. The unconditioned response to shock was examined by a motion index of each animal during footshock (Video Freeze software from Med Associates).

Histology.

At the completion of behavioral testing (for all experiments), the mice were overdosed with isoflurane and perfused intracardially with PBS followed by 4% paraformaldehyde (PFA). The brains were removed and stored in 4% PFA overnight at 4 °C. Coronal sections (100–150 µm) were taken through the mPFC and mounted on slides. The sections were examined with a florescence microscope. Behavior analysis was conducted using data from mice with a majority GFP- or RFP-labeled cells in the PL of the mPFC (up to 40% electroporated male mice).

Statistical Analysis.

All statistical significance was assessed using an alpha level of 0.05. Statistical analysis was performed using SAS 9.2 and/or GraphPad Prism software.