Deletion of the miR-379/miR-410 gene cluster at the imprinted Dlk1-Dio3 locus enhances anxiety-related behaviour

Abstract

The brain-specific miR-379/miR-410 gene cluster at the imprinted Dlk1-Dio3 domain is implicated in several aspects of brain development and function, particularly in fine-tuning the dendritic outgrowth and spine remodelling of hippocampal neurons. Whether it might influence behaviour and memory-related processes has not yet been explored at the whole organism level. We previously reported that constitutive deletion of the miR-379/miR-410 gene cluster affects metabolic adaptation in neonatal mice. Here, we examined the role of this cluster in adult brain functions by subjecting mice with the constitutive deletion to a battery of behavioural and cognitive tests. We found that the lack of miR-379/miR-410 expression is associated with abnormal emotional responses, as demonstrated by increased anxiety-related behaviour in unfamiliar environments. In contrast, spontaneous exploration, general locomotion, mood levels and sociability remained unaltered. Surprisingly, miR-379/miR-410-deficient mice also showed normal learning and spatial (or contextual) memory abilities in hippocampus-dependent tasks involving neuronal plasticity. Taken together, the imprinted miR-379/miR-410 gene cluster thus emerges as a novel regulator of the two main post-natal physiological processes previously associated with imprinted, protein-coding genes: behaviour and energy homeostasis.

Introduction

Genomic imprinting is an epigenetic phenomenon whereby certain genes are expressed only from one allele depending on which parent the allele was inherited from, i.e. for a given gene locus, only one of the two parental alleles is transcriptionally competent. Approximately 100 mammalian autosomal genes are known to display this parent-of-origin-specific mono-allelic expression, with many of them grouped into large chromosomal domains (1,2). Besides their well-established roles in pre-natal growth and placentation (3), it has become apparent that imprinted, protein-coding genes also modulate important post-natal physiological functions, notably energy homeostasis (4) and neurobehavioural processes (5–10).

Four evolutionarily distinct, imprinted chromosomal domains harbour large arrays of repeated DNA sequences producing dozens of small RNA species belonging to the microRNAs (miRNAs) and/or C/D small nucleolar RNAs (C/D snoRNAs) families, two classes of small non-coding RNAs that post-transcriptionally regulate gene expression (11,12). The physiological functions of these imprinted, small regulatory RNA genes, which are only found in placental mammals, remain poorly understood [(13–18); reviewed in 19,20]. One of the imprinted gene clusters, the so-called Dlk1-Dio3 chromosomal domain, comprises numerous miRNA genes expressed from the maternal allele, many of which are grouped in a single cluster called the miR-379/miR-410 gene cluster (also known as C14MC in human) (16,17). This is the largest placental mammal-specific miRNA gene cluster identified to date; it contains as many as 39 miRNA genes in an ∼40 kb-long region (Fig. 1A). Whereas during development and the immediate post-natal period the miR-379/miR-410 gene cluster is expressed in many non-neuronal tissues, in adulthood it is expressed mostly, if not exclusively, in the brain (22).

Figure 1.

The miR-379/miR-410 gene cluster at the imprinted Dlk1-Dio3 locus. (A) Schematic representation of the ∼1 Mbp imprinted Dlk1-Dio3 locus on mouse distal chromosome 12. Genes expressed from paternally inherited chromosomes (Begain, Dlk1, Rtl1 and Dio3) are depicted by black boxes and those expressed from maternally inherited chromosomes (miRNA and C/D snoRNA genes) by white ovals and triangles, respectively. Silent alleles are coloured in grey. Note that numerous miRNAs and C/D snoRNAs are embedded within and processed from introns of non-coding gene transcripts, namely Gtl2 (white box), Anti-Rtl1, Rian and Mirg. Differentially methylated regions are indicated by filled and open lollipops (methylated and unmethylated, respectively). The bigger lollipops depicted the germline-derived differentially methylated region, also referred to as intergenic differentially methylated region, imposing the mono-allelic expression of all imprinted genes over the entire domain (21). Drawing is not to scale. (B) The breeding strategy. A schematic representation of the two reciprocal parental crosses performed to generate ΔMat and ΔPat individuals and their WT littermates. For clarity, only a deletion inherited from a maternal grandparent is shown for the parents in the two crosses. Paternally and maternally inherited chromosomes are symbolized by black and white rectangles, respectively. Het: heterozygote. (C) The paternally inherited allele of the miR-379/miR-410 gene cluster is not expressed in ΔMat. Expression of selected miRNAs scattered within the miRNA cluster (miR-379, miR-134 and miR-410) was assayed by RT-qPCR (relative to U6 snRNA) using total RNA extracted from manually dissected brain regions of WT, ΔPat and ΔMat individuals (n = 4 or 5 per genotype). Expression levels of WT were arbitrarily set to 1.

This atypical miRNA cluster has recently gained much attention as a regulator of neuronal plasticity. Indeed, one miRNA encoded in the miR-379/miR-410 gene cluster, miR-134, is one of the best-studied activity-dependent miRNAs and has received the most attention (23). In primary cultures of hippocampal and cortical neurons, miR-134 localizes to the synapto-dendritic compartment and limits dendritic spine growth by inhibiting local translation of the mRNA encoding the cytoskeleton regulator LIM domain kinase 1 (24). It also promotes activity-dependent dendritogenesis by interfering with translation of the mRNA encoding the translational repressor pumilio RNA binding family member 2 (25), which is necessary for homeostatic synaptic depression in response to chronic activity (26). Furthermore, miR-134-dependent regulation of expression of the cyclic adenosine monophosphate response element-binding protein mediates the effect of the deacetylase Sirt1 on synaptic plasticity and memory formation (27). Strikingly, the activity of miR-134 in primary cortical neuron cultures appears limited to a subset of neurons—the somatostatin interneurons—wherein it regulates the production of the palmitoylation enzyme DHHC9 (28). Another activity-regulated miRNA encoded in the miR-379/miR-410 gene cluster, miR-485, modulates dendritic spine number and synapse formation by repressing the mRNA encoding the synaptic vesicle protein (SV2A) (29), whereas others (miR-369p, miR-496 and miR-543) are involved in neurogenesis and neuronal migration in the developing cortex (30). These examples and others in the literature point to important regulatory roles for the miR-379/miR-410 gene cluster in many aspects of neuronal physiology in the normal brain as well as in some pathological contexts such as status epilepticus (31–33).

The neuronal functions ascribed to miRNAs encoded by the miR-379/miR-410 cluster have been inferred mostly from over-expression or antisense-mediated silencing experiments. Despite their widespread use and utility in the field of miRNA research, these approaches can yield conclusions at odds with those drawn from conventional gene targeting strategies (discussed in 34,35). To have a complete picture of the biological relevance of the miR-379/miR-410 cluster in brain function, it is therefore important to investigate thoroughly the phenotypes resulting from its genetic ablation in vivo. Here, by using a constitutive knockout mouse model with a site-specific deletion encompassing the entire miRNA cluster (22), we have examined the involvement of these maternally expressed miRNA genes in adult behaviour with particular attention to memory-related processes.

We show that the lack of miR-379/miR-410 expression leads to increased anxiety-related phenotypes but, quite unexpectedly, does not impair hippocampus-dependent long-term memory and learning performances as one might have predicted from previous studies highlighting a role of these miRNAs in the plasticity of hippocampal neurons (24,25,27,32,36). This represents one of the very first studies that have explored, extensively and systematically, how genetic ablation of a given mammalian miRNA locus impacts on adult behaviours.

Results

A maternally inherited deletion of the miR-379/miR-410 cluster is associated with a partially penetrant neonatal lethality phenotype, very likely due to impaired metabolic adaptation at the transition from pre-natal to post-natal life (22). The surviving adult mice, however, do not display overt abnormalities when maintained in classical mouse husbandry conditions. To explore the biological relevance of the miR-379/miR-410 gene cluster in higher brain functions, we submitted 3- to 5-month-old male mice to a series of well-characterized behavioural tests. More precisely, we generated two classes of genetically comparable heterozygotes from two reciprocal parental crosses (Fig. 1B): one class of individuals carrying a constitutive deletion of paternal origin (the so-called ΔPat in which the miRNA genes should be expressed normally since only the paternal silent allele is deleted) and one class of individuals carrying a constitutive deletion of maternal origin (the so-called ΔMat in which the miRNA genes should no longer be expressed since the maternal active allele is deleted). ΔMat and ΔPat mice were mostly compared with their respective wild-type (WT) littermates, thus avoiding any potential confounding littermate effects. Given the maternal expression of the miRNA genes, any behavioural deficiencies caused by the loss of miR-379/miR-410 expression would be expected to be revealed in ΔMat, but not ΔPat mice. Our breeding strategy was validated by showing that three miRNAs encoded by sequences scattered along the cluster—namely miR-379, miR-134 and miR-410—were not detected by real-time quantitative polymerase chain reaction (RT-qPCR) analysis in several regions of the adult ΔMat mouse brain (Fig. 1C). As expected, their expression levels were similar in ΔPat and WT mice, although somewhat weaker, but not statistically significant, expression was seen in some brain areas in ΔPat mice. Thus, significant expression of the miRNA cluster from the normally silent paternal allele in the brains of adult ΔMat mice appears very unlikely.

Loss of expression of the miR-379/miR-410 gene cluster promotes anxiety but not depression-related behaviours

We first examined ΔMat mice for their spontaneous locomotion and level of interest in a novel environment by measuring the distance travelled and general behaviour in the open-field (OF) test. No difference in wall leaning [F(_3,40) = 0.4586, P = 0.7129], rearing [F_(3,40) = 0.5364, P = 0.6603] and locomotion [F_(3,40) = 2.221, P = 0.1020] was observed between the four genotypes, indicating that general locomotion and exploration are not affected by loss of miR-379/miR-410 expression (Fig. 2A). We then looked for anxiety-related behaviour by monitoring the approach-avoidance behaviour that rodents display when they explore unfamiliar environments. In this test, anxiety-related behaviour is reflected by avoidance of the central part of the arena. One-way analysis of variance (ANOVA) revealed a genotype effect on the time spent and the number of entries in the central zone in the OF test [F_(3,40) = 12.07, P < 0.0001 and F_(3,40) = 6.983, P < 0.001, respectively]. Both the time spent (Fig. 2B, left) and the number of entries (Fig. 2B, right) into the central area were reduced in ΔMat mice when compared with their WT littermates. As expected, the behaviour of ΔPat mice was indistinguishable from that of their WT littermates (Fig. 2B).

Figure 2.

ΔMat mice exhibit increased anxiety-related behaviour. (A) Exploratory behaviour and locomotor activity. Histograms show the number of times the mice leaned against the walls of the arena (proximal exploration, ‘Wall leaning’, top) and they reared (distal exploration, ‘Rearing’, middle) and their spontaneous locomotion (distance moved, bottom) in the circular open field during a 10 min session. (B) The OF test. Histograms show the time spent in (left) and the number of entries into (right) the centre area of the circular open field during a 6 min session. (C) The EPM test. Histograms show the time spent in (left) and the number of entries into (right) the open arms of the EPM apparatus during a 5 min session. (D) The NSF test. Histograms show the latency to eat the food pellet located in the centre of the box in a 10 min session. (E) The TS test. Histograms show the time spent immobile during the 6 min session. All data are presented as means ± SEM. Statistically significant differences, *P < 0.05, **P < 0.01 and ***P < 0.001. ΔMat versus ΔPat, ^#P < 0.05, ^##P < 0.01 and ^###P < 0.001. In all plots, the numbers of individuals analysed are indicated above each histogram. ns: not significant.

To confirm this increased anxiety-related behaviour, we performed a different anxiety test, the elevated plus maze (EPM) test, in which mice were placed in a plus-shaped apparatus with two open (‘stressful’) and two enclosed (‘protective’) arms elevated from the floor. No overall genotype effect was observed for the time spent and numbers of entries in the open arms [F_(3,40) = 2.130, P = 0.1130 and F_(3,40) = 2.184, P = 0.1063, respectively]. However, ΔMat mice avoided the open arms as shown by a reduced time spent (Fig. 2C, left) and fewer entries (Fig. 2C, right) into these parts of the maze when compared with their WT littermates. As expected, the behaviour of ΔPat was indistinguishable from that of their WT littermates (Fig. 2C). We, thus, conclude that miR-379/miR-410-deficient mice display abnormal reactivity to anxiety-causing environments.

To investigate whether other forms of emotion-related behaviours might also be affected by deleting this miRNA cluster, ΔMat mice were submitted to the novelty-suppressed feeding (NSF) test. This test consists of exposing food-deprived mice to a novel, brightly lit environment in which a food pellet is positioned in the centre. Any delay in the time taken to start eating in this environment indicates mixed anxio-depressive behaviour (37). Both ΔMat and ΔPat mice showed similar feeding latencies when compared with their WT littermates (Fig. 2D). One-way ANOVA indicated no significant genotype effect on the latency to feed [F_(3,41) = 0.7878, P = 0.5082]. Depression-like behaviour was further investigated by using the tail suspension (TS) test, in which the mice were suspended by their tails and unable to escape; the period of immobility gives a read-out of despair-related behaviour. The duration of immobility measured for ΔMat mice was similar to that of WT mice (Fig. 2E). As expected, ΔPat mutant animals behaved like their WT littermates. Again, ANOVA showed no difference between the four genotypes in this test [F_(3,42) = 0.0516, P = 0.9843]. Together these data show that the lack of miR-379/miR-410 expression is associated with enhanced anxiety-related behaviour that is not accompanied by any obvious depression-like symptoms.

Loss of expression of the miR-379/miR-410 gene cluster does not alter stress-induced corticosterone release

By releasing corticosterone (CORT) into the circulation, the hypothalamic–pituitary–adrenal (HPA) axis plays a central role in the mammalian stress response. We reasoned that the anxious behaviour of ΔMat mice might be due, at least partly, to defects in CORT levels. To test this hypothesis, we measured circulating CORT levels in ΔMat mice and WT littermates (Fig. 3). We found no differences between the ΔMat and WT mice [F_(1,12) = 0.1999, P = 0.6628]; as expected, their CORT levels were lower in the morning than in the evening [F_(1,12) = 16.70, P < 0.01]. We then restrained ΔMat and WT mice for 30 min in a ventilated 50 ml tube, which is known to elicit a robust increase in CORT levels. In these stress conditions, CORT levels rose dramatically soon after restraint [F_(1,12) = 250.6, P < 0.0001], and the magnitude of the response was similar in WT and ΔMat mice. We conclude that stress-mediated activation of the HPA axis occurs normally in the absence of miR-379/miR-410 gene cluster expression.

Figure 3.

Activation of the HPA axis is not impaired in ΔMat mice. Plasma CORT levels were assayed under basal conditions in the evening (7 pm) and morning (9 am) or shortly after a restraint stress (mice were placed in a ventilated 50 ml plastic tube for 30 min). The numbers of individuals analysed are indicated above the histograms. All data are presented as means ± SEM. ***P < 0.001 and **P < 0.01. ns: not significant.

Loss of expression of the miR-379/miR-410 gene cluster does not affect social preference behaviour

A potential role for imprinted genes in regulating adult social behaviour has been proposed previously (8,38). We, therefore, investigated the sociability of ΔMat mice by using the three-chamber test (Fig. 4). In this test, mice are presented with a free choice between spending time in a chamber containing an unfamiliar mouse (social) or in an empty chamber (non-social). During the habituation session, the four genotypes contacted equally the two empty cages, E1 and E2, excluding any bias for one of them (Fig. 4, left). Two-way ANOVA indicated no significant difference between the genotypes [F_(3,76) = 1.365, P = 0.260] or the different cages [F_(1,76) = 0.0002, P = 0.988]. After habituation to two empty cages, an unfamiliar mouse was introduced into one of the two cages (denoted hereafter as the stranger cage, S). The time spent contacting the stranger cage and the numbers of contacts made (data not shown) provide a measure of sociability (Fig. 4, right). During the sociability session, the presence of the stranger significantly increased the time spent contacting the cage containing the stranger, as determined by two-way ANOVA [F_(1,76) = 128, P < 0.0001], but there was no difference between the four analysed genotypes [F_(3,76) = 0.6764, P = 0.5692]. This demonstrated that ΔMat mice showed a similar preference to the ΔPat and WT mice for the stranger cage, thus indicating that social preference stimuli are unlikely to be impaired in miR-379/miR-410-deficient mice. Whether other types of social behaviours (e.g. social novelty, social dominance, social interactions etc.) are impacted by the maternally inherited deletion remains to be further examined.

Figure 4.

Social preference behaviour is not altered in ΔMat mice. Histograms show the time spent sniffing each empty cage (E1 and E2) during the habituation session (10 min) of the three-chamber test (left) and the time spent sniffing the empty cage (E) and the stranger cage (S) during the sociability session (10 min) (right). The numbers of individuals analysed are indicated in brackets under the histograms. All data are presented as means ± SEM. Statistically significant differences, ***P < 0.001. ns: not significant.

Loss of expression of the miR-379/miR-410 gene cluster does not affect spatial memory

Given the growing body of evidence indicating a role for the miR-379/miR-410 gene cluster in general, and miR-134 in particular, in dendritic outgrowth and spine remodelling (24–27,31,36), we performed two tests to assess learning and memory in our mutant mice (Fig. 5). The first, the object location test, is a hippocampus-dependent spatial task that assesses the ability of rodents to remember the location of a previously encountered object. It comprises a training session during which mice are allowed to explore two identical objects and a session one day later when the mice are exposed to the same two objects but one of them has been moved to a different location. Spatial memory is reflected by the mice spending more time exploring the displaced object. During the training session, the exploration time at both objects was similar for the four analysed genotypes [F_(3,70) = 0.7189, P = 0.5474] and, as expected, decreased over the three trials [F_(2,70) = 48.97, P < 0.0001] as the mice became familiar with the objects (Fig. 5A, left). No significant difference was observed between the genotypes in the three trials [F_(6,70) = 0.8655, P = 0.5247]. The mice contacted the two objects equally, thus excluding a bias for one of them (data not shown). When one object was displaced to a new location, ΔMat mice spent more time sniffing the moved object than the non-displaced one, suggesting that they remembered correctly the previous exposure (Fig. 5A, right). Again, the duration of exploration of the displaced object was statistically similar for the four genotypes [F_(3,38) = 0.3991, P = 0.7545].

Figure 5.

ΔMat mice display normal spatial memory. (A) The object location test. Histograms show the total exploration time contacting the two objects (O1 and O2) during the three habituation trials (left) and the preference for the displaced object (DO, for example object O2 in the drawing) when compared with random chance (50%, dotted line) (right). Note that the preference of the WT littermates of ΔMat for the DO was not statistically significant (P = 0.175). (B) The Morris water maze test. The curves show the distance moved to reach the hidden platform (filled circle in the target quadrant) during the four training sessions (left). Histograms show the number of times the annuli (open circles) were crossed in each quadrant (target, T; opposite, O; adjacent 1, A1; adjacent 2, A2) during the trial performed 24 h post-training. The numbers of individuals analysed are indicated in brackets under the histograms. All data are presented as means ± SEM. Statistically significant differences, *P < 0.05, **P < 0.01 and ***P < 0.001. ns: not significant.

The second test, the Morris water maze test, provides another measure of hippocampal-dependent learning and memory. In this test, mice are placed in a pool of opaque water. During the training phase, a platform is placed below the waterline and the mice must use visual clues placed around the pool to find the hidden platform and escape from the water. One day after the fourth training session, the platform is removed, the mice are allowed to swim for 60 s and the number of times they cross the position of a virtual platform in a target quadrant (the annulus) is monitored. The distance travelled in each of the four training sessions decreased [F_(3,78) = 15.53, P < 0.0001], as did the time spent to reach the platform (data not shown), indicating learning (Fig. 5B, left). There was no difference between the genotypes [F_(3,78) = 0.5654, P = 0.6427] or the interaction between session × genotype [F_(9,78) = 1.105, P = 0.3694]. Also, no significant difference was observed between the swimming speeds of the different genotypes during the learning sessions (data not shown). The next day, when the platform was removed, the four genotypes navigated preferentially within the target quadrant where the platform was previously located during the training sessions as shown by the number of times the annulus was crossed [F_(3,104) = 48.27, P < 0.0001] and the time spent (data not shown), demonstrating that they remembered correctly its position in the training sessions (Fig. 5B, right). There was no significant effect of genotype [F_(3,104) = 0.8779, P = 0.4552], showing that all mice exhibited a similar platform-seeking behaviour. From these two tests, we conclude that spatial memory is not altered in miR-379-/miR-410-deficient mice.

Loss of expression of the miR-379/miR-410 gene cluster does not affect contextual learning and memory

Finally, we assessed another form of long-term memory by examining the behaviour of ΔMat mice in the Pavlovian contextual fear conditioning test (Fig. 6). In this test, the mice learn to associate a neutral stimulus (a context and/or a single tone sound) with an aversive one (an electric shock to the foot), which triggers a conditioned fear or ‘freezing’ response characterized by the lack of movement. During the training session, the four genotypes displayed the same levels of freezing response (not shown). Twenty-four hours post-training, the mice were replaced in the same chamber (i.e. the same context), but neither the tone nor the aversive electric shock stimulus was applied (Fig. 6, left). All four genotypes showed an increase in their freezing response in this context when compared with the ‘baseline’ freezing scored before administration of the first electric shock during the training session [F_(1,26) = 231.6, P < 0.0001]. Moreover, levels of freezing were in the same range for the four genotypes [F_(3,29) = 2.730, P = 0.064]. We, therefore, conclude that ΔMat mice performed normally, indicating that the loss of the miR-379/miR-410 gene cluster does not alter contextual long-term memory. The mice were then placed in a completely different chamber (new context) in which the tone stimulus was applied, but not the electric shock (Fig. 6, right). In this new context, the freezing responses of the four genotypes were similar before the tone stimulus (pre-tone) [F_(3,29) = 1.294, P = 0.2973]; the responses all increased after the tone stimulus but, again, the responses of the four genotypes were similar [F_(3,29) = 1.164, P = 0.3423]. Two-way ANOVA showed a significant condition effect [F_(1,26) = 150.3, P < 0.0001], confirming that all genotypes increased their freezing response during presentation of the tone compared with the pre-tone condition. Altogether, these data indicate that both contextual and tone fear memories (hippocampus and amygdala-dependent tasks, respectively) are not affected by the lack of expression of the miR-379/miR-410 gene cluster.

Figure 6.

ΔMat mice display normal contextual and tone-cued fear memories. Contextual memory (left). Histograms show expression of conditioned fear in the training chamber before (baseline) and 24 h after conditioning (context). Note: differences in the ‘baseline’ freezing by the four genotypes were observed before administration of the electric shocks [F_(3,29) = 4.627, P = 0.01]; notably, the freezing of ΔMat mice was greater than that of their WT littermates. Tone–shock association memory (right). Histograms show expression of conditioned fear in a new context before (pre-tone) and during the tone stimulus. All data are presented as means ± SEM. Statistically significant differences, *P < 0.05. ns: not significant.

Discussion

Despite our increasing knowledge of the mode of the action of mammalian miRNAs in brain physiology and neurological disorders (reviewed in 39–41), the contribution of individual miRNA genes to behaviour is only beginning to be studied. A case in point is the mouse miR128-2 gene, inactivation of which causes hyperactivity and increases exploratory behaviour (42); also, knockout of the miR-132/122 gene impairs spatial learning and memory retrieval in mice (43). In this study, we investigated extensively the behaviour of knockout mice that have a constitutive, site-specific deletion of the miR-379/miR-410 gene cluster located in the imprinted Dlk1-Dio3 locus, which is expressed specifically in the brain and only from the maternally inherited allele (16,22). We found that adult mice that inherited the deletion from the mother (ΔMat mice) displayed increased reactivity to anxiety-causing stimuli with no depression-like symptoms. As expected, ΔPat mice that inherited the deletion from the father behaved like their WT littermates, further demonstrating that the differences in the emotional behaviour of the ΔMat mice are due to the deleted version of the miRNA cluster and not to other confounding factors (44). The observed anxiety-related avoidance of stressful environments is unlikely to be due to non-emotional factors since spontaneous exploratory locomotion, sociability, learning and memory remained unaltered. This phenotype, which seems not due to impaired CORT release in response to stress, should be interpreted as a transient, maladaptive response rather than a permanent trait (45) because the behaviour of ΔMat mice in their home cages, monitored over more than 4 years of colony management, was indistinguishable from that of their WT littermates. Of note, these emotional abnormalities are very likely task-dependent since they were revealed in the OF and EPM tests but not in the NSF test. Finally, whether the levels of anxiety-related phenotypes are different in ΔMat females (either exacerbated or diminished) remains an open question since the present study only focused on males.

Although genetic association studies and gene expression profiling have suggested an involvement of miRNAs in anxiety disorders (46), little is known regarding how genetic alteration of miRNA expression levels in mice may cause anxiety-like phenotypes (47,48). Regarding the miR-379/miR-410 gene cluster, it has been shown that miR-134 is up-regulated upon acute stress or fear conditioning in rats (49,50) and expression of miR-323, another miRNA encoded in the miR-379/miR-410 gene cluster, is associated with anxiety phenotypes in mice (51). Despite these correlations, no previous robust functional data causally linked this miRNA cluster to emotional responses. In the absence of molecular clues, we speculate that one (or several) miRNAs encoded in the miR-379/miR-410 gene cluster function(s) to remediate changes in neural circuits triggered by stressful events. Such a ‘homeostasis guard’ hypothesis was already discussed for the stress-induced miR-34c (47). Finally, we cannot rule out the possibility that the metabolic deficiencies experienced by ΔMat pups (22) may contribute to their altered emotional state in adulthood. Indeed, adverse events in utero and in the perinatal period can give rise to long-term metabolic and/or brain disorders, including emotional phenotypes (52).

The lack of alterations in memory and learning may appear at odds with previous findings showing that ectopic expression or knock-down of miR-134 interferes with dendritic spine morphology, dendritic outgrowth and long-term potentiation (24,25,27,31,36). Indeed, one might expect ΔMat mice to display deficiencies in hippocampus-dependent cognitive tasks involving neuronal plasticity, as assessed by the object location, Morris water maze and contextual fear conditioning tests. Several non-mutually exclusive explanations might reconcile these apparent discrepancies. First, and paradoxically, complete lack of expression of the entire miRNA cluster may be less detrimental than down-regulation of miR-134 alone. Secondly, changes in dendritic spine morphology do not necessarily result in cognitive disabilities. For instance, silencing of miR-134 by the use of antagomirs increases CA3 pyramidal neuron spine volumes in vivo, yet it does not affect the performance of the mice in the object location task (32). Thirdly, compensatory and/or redundant pathways may operate in ΔMat animals, thus attenuating or even masking some neuronal phenotypes. These putative compensatory effects could also be causally linked to the survival of some mutant pups (22). Future studies using conditional mutant mice for which the miR-379/miR-410 is deleted post-natally might help to clarify this important issue. The trivial explanation that aberrant reactivation of expression from the normally silent paternal allele occurs in ΔMat mice can be ruled out, however, since homozygous null mutants behaved normally in the Morris water maze and contextual fear conditioning tests (data not shown). Fourthly, ectopic expression of a miRNA, often at supra-physiological levels, can generate off-target effects by increasing the magnitude of the repression of bona-fide targets or generate off-target effects on other mRNAs besides the biological target (35). Likewise, antisense technologies are also prone to off-target effects that might account for differences in phenotypes observed in some knockdown and gene knockout models (34,35,53). Finally, the miR-379/miR-410 gene cluster may participate in plasticity-related phenomena that have not been tested directly in this study. In this regard, it may be worth exploring the impact of loss of the miR-379/miR-410 gene cluster on acute pathologies such as status epilepticus, which induces synapse reorganization. Indeed, two independent groups have recently reported that inhibiting miR-134 protects from epileptic seizures (31–33). Whatever the explanation, our findings highlight the importance of testing in vivo how defective mammalian miRNA gene expression results in behavioural effects.

Many imprinted, protein-coding genes are expressed in the brain where they influence a broad range of behaviours including maternal care (54–56), suckling activities (57,58), milk release (56), ultrasonic vocalization (59), social interaction (60), exploration and risk taking (61), learning and memory (59,62) and emotional responses (63,64). In contrast, the involvement of imprinted, small non-coding RNA genes in behaviour has been poorly investigated so far. Interestingly, adult mice lacking the paternally expressed Snord116 gene array at the Snurf-Snrpn domain also displayed enhanced anxiety-related behaviour without alterations in memory or sociability (63,65) reminiscent of the phenotype we see in adult mice lacking the maternally expressed miR-379/miR-410 cluster. This suggests that these two large, imprinted small non-coding RNA gene arrays may have evolved in placental mammals to fine-tune regulatory pathways underlying emotional behaviour. It should be noted, however, that SNORD116-KO mice also suffer from early failure to thrive (63,66). As already discussed above, anxiety-phenotypes observed in these two small RNA KO mouse models may therefore originate, at least in part, from adverse early life experiences. Along this line of thought, it has been elegantly demonstrated for another imprinted gene—the paternally expressed Igf2 gene—that mismatches between placenta supply and embryo demand also leads to anxiety in adulthood (67). These observations point to the inherent difficulty in unravelling the complex, and very likely interlaced, physiological functions revealed by the constitutive disruption of imprinted genes (68).

In summary, although deletion of the miR-379/miR-410 gene cluster in mice does not lead to obvious gross abnormalities under normal laboratory conditions, as documented for most other miRNA knockout mice (35,69–71), metabolic and behavioural phenotypes become apparent when these mice must adapt rapidly whether to the metabolic stresses of birth and weaning (22) or, as shown in this study, to stressful environments (the OF and elevated-plus maze tests). The miR-379/miR-410 gene cluster emerges, therefore, as a novel player in the two most prevalent post-natal physiological functions associated with imprinted, protein-coding genes: the control of energy homeostasis and brain functions (68). In this respect, our findings are consistent with the emerging view that the imprinted Dlk1-Dio3 domain exerts an influence during phases of post-natal adaptation, in addition to its well-characterized roles in embryonic growth and placentation (68,72,73). To the best of our knowledge, our study also provides the first evidence of a contribution of the imprinted Dlk1-Dio3 domain to modulating adult behaviour. Genome-wide studies on anxiety in humans identified a link to a region in the Dlk1-Dio3 locus on chromosome 14q32 (74). Whether deficiencies in the biology of miRNAs (also named C14MC) also lead to emotional impairments in humans awaits further investigation.

Materials and Methods

Mice housing and breeding

All animal procedures were approved by the University of Toulouse and CNRS Institutional Animal Care Committee (FRBT_N01503.01). The animal housing facility met CNRS standards. Mice were housed in standard plastic cages with access to food (rodent chow diet) and water ad libitum in a temperature-controlled room, with a 12 h light–dark cycles. For mating, male and female animals were placed together in a cage overnight. Mating was performed overnight. Behavioural tests were performed blind during the light phase on 3- to 5-month-old male mice backcrossed on the C57BL/6J genetic background for 6–12 generations. Before testing, the mice were handled for 3 days.

CORT measurements and restraint stress

Blood samples were obtained from unstressed animals both in the morning (9 am) and evening (7 pm). The time between taking each cage from the animal facility to the experimental room and collecting the blood from all animals in that cage did not exceed 3 min. Samples were also collected after a 30 min restraint inside a 50 ml Falcon tube with holes to allow breathing. Blood was collected from the facial vein into capillary tubes containing 10% ethylenediaminetetraacetic acid. Samples were centrifuged at 10 000 rpm for 10 min, and the supernatant (plasma) was stored at −80°C. Total CORT in plasma was measured with an in-house radioimmunoassay described previously (75). Briefly, plasma samples were extracted with absolute ethanol and total CORT was measured by competition between cold CORT and labelled CORT for a specific anti-CORT antibody provided by Dr H. Vaudry (University of Rouen, France).

OF test

Locomotion and exploratory behaviour were measured in a circular arena (height, 30 cm; diameter, 40 cm) located in a room containing no conspicuous features and illuminated by a white light (60 W). Proximal and distal explorations were defined by wall leaning and rearing, respectively. Time spent in the centre of the arena was recorded using the video tracker software (Ethovison^®XT, Noldus, Netherlands). Baseline behaviours were scored during a 10 min session whereas locomotion and centre/periphery activity were only analysed during the first 6 min.

EPM test

To measure anxiety-related behaviour, mice were placed in an automated EPM (Imetronic^®, Pessac, France). This maze consisted of a plus (+)-shaped track with two closed and two open arms (30 × 10 × 20 cm) that extended from a central platform (10 × 10 cm). The apparatus was elevated 50 cm above the floor and was surrounded by a white curtain with no conspicuous cues. Each trial began with the mouse placed in the central zone and lasted 5 min. The number of entries into the closed and open arms and the time spent in each arm were monitored.

The NSF test

To measure anxiety and depression components of behaviour, the NSF test was carried out as described by Santarelli et al. (76). Briefly, food-deprived mice for 24 h were tested in a 50 × 50 × 20 cm box covered with bedding and illuminated by a 70 W lamp. A single pellet of food was placed on a white paper positioned in the centre of the box. All mice were tested individually for 10 min and the latency time before the mouse ate the pellet was scored as a measure of motivation. Mice that did not eat during the testing period were not included in the analysis. Immediately afterwards, each single mouse returned to its home cage and the amount of food consumed was measured for 5 min, serving as a control for change in appetite as a possible confounding factor.

Tail suspension test

The tail suspension test (TST) assesses depression-related behaviours (77). Mice are suspended by their tails with tape, such that they cannot escape or hold on to nearby surfaces. During the test, typically 6 min in duration, the resulting escape-oriented behaviours are quantified using the Bioseb TST software (Bioseb, Vitrolles, France). A specific strain gauge linked to a computer quantifies the time of mobility and immobility.

Sociability test

Social behaviour was measured in a rectangular three-chambered enclosure, 60 × 40 cm, with clear walls 22 cm high. Removable doors blocked access from the centre chamber to the outer chambers. Two cages, diameter 8 cm, height 15 cm, were placed in the two outer chambers. The test consists of two stages, which were conducted immediately after one another. Mice were first habituated to the apparatus for 10 min, with no mice in the cages. During the second stage, a stranger mouse was placed in one of the cages, and time spent interacting with the stranger mouse, and the empty cage, was measured. Interaction was measured manually and defined as time spent sniffing the stranger mouse or cage. Preference for the stranger mouse over the empty cage was used as a measure of sociability.

Object location test

This test was adapted from one described previously (78) and conducted in a circular arena (height, 30 cm; diameter, 40 cm). After a 10 min habituation session in the arena, mice were placed in the arena with two identical objects for three consecutive 6 min sessions, each separated by a 3 min interval, during which the animals were returned to the holding cage. Mice that did not contact objects during the habituation session were excluded from the analysis. Twenty-four hours later, mice were submitted to the testing session, in which one of the two objects was repositioned, thereby changing the spatial configuration of the objects in the arena. Mice were allowed to explore the objects for 6 min. The time exploring the displaced object during the test session was scored manually to determine whether mice preferentially re-explored the displaced object during the spatial change session.

Morris water maze test

Hippocampus-dependent memory was evaluated in a circular swimming pool (height, 30 cm; diameter, 110 cm) filled to a depth of 15 cm with water maintained at 23 ± 1°C. The water was made opaque by the addition of a white non-toxic opacifier. A white painted platform (diameter, 9 cm) was placed inside the pool, 15.5 cm away from the pool wall. Several visual clues were attached to the walls of the experimental room. Four start positions were located around the perimeter of the pool, dividing its surface into four equal quadrants. At Day 1, mice were individually submitted to a single pre-training session (familiarization) consisting of three trials with a visible platform 0.5 cm above the surface of the water and always at the same location [as described previously (79)]. At the beginning of each trial, mice were placed in the maze and allowed to swim freely until they reached the platform. Mice were allowed to remain on the platform for an additional 60 s. The next day, mice were submitted to four consecutive sessions of three trials with an inter-session resting period during which they were returned to the home cage. Distances to reach the hidden platform located in the same quadrant (the target) were recorded by a video tracking system (Ethovison^®XT, Noldus, Netherlands). During the test (Day 3), the platform was removed and mice were allowed to search for the platform for 60 s. The number of annulus crossings [the number of times a mice crossed a small region (diameter 14 cm) equivalent to each of the four possible platform positions] was obtained to characterize the search pattern used by the mice.

Contextual and tone-cued fear conditioning test

Training and testing of contextual fear conditioning occurred in a rectangular polyvinyl chloride box (length, 35 cm; width, 20 cm; height, 25 cm). The chamber floor was made from stainless steel rods (diameter, 4 mm) spaced 1 cm apart and connected to a shock generator (Campden Instruments, UK). The experimental device was lit by a 60 W white bulb. On the training day, individual mice were placed in the conditioning chamber for 120 s before the onset of the conditioned stimulus (CS), a tone of 1500 Hz and 85 db for 30 s. The last 2 s of the CS was paired with the unconditioned stimulus (US), 0.70 mA of continuous electric shock to the feet. The first US was followed by another paired CS/US 120 s later. After an additional 30 s in the chamber, the mouse returned to its home cage. Twenty-four hours later, contextual conditioning was assessed by a 4 min session in the chamber in which the mice were trained (context). Two hours later, the mice were placed in a new context (consisting of a triangular chamber with white Plexiglas walls and floor, which had been cleaned with a 1% acetic acid) for 2 min (pre-tone), after which they were exposed to the CS for 2 min (tone). The percentage of time spent freezing (which is defined as the absence of all movement except for respiration) served as an index of fear memory. All behavioural devices cited above were cleaned with 70% ethanol between each mouse to remove olfactory cues.

RNA extraction and RT-qPCR

Total RNA was extracted using TRI reagent^® (Euromedex) followed by RNAse free-RQ1 DNAse (Promega) and proteinase K (Sigma) treatments. miRNA expression was measured using the miScript Reverse Transcription kit and the miScript SYBR Green PCR kit (Qiagen), using specific primers for miRNA (miR-134 (#MS00001568), miR-379 (#MS00011942) or miR-410 (#MS00002331)) and U6 RNA (#MS00014000) as an endogenous control. Complementary DNAs were amplified on an Analytik Jena Flexcycler.

Statistical methods

The Graphpad Prism^® (version 5.01) statistical software was used for data analysis. Data from the OF, EPM, NSF and TS tests were analysed by using one-way ANOVA. During habituation trials, object explorations were analysed by using two-way ANOVA (genotype × trials) with repeated measures. Preference for the displaced object was analysed by using one-way ANOVA and one simple t-test for the 50% chance comparison. Social interactions were analysed by using two-way ANOVA (genotype × cage). The Morris water maze test data were analysed by using two-way ANOVA (genotype × session) with repeated measures. Context and CS test data were analysed by using one- or two-way ANOVA (genotype × condition). Plasma CORT data were analysed by using two-way ANOVA (genotype × time or condition). All comparisons of two groups were analysed by a two-tailed Student's unpaired t-test.

Funding

This work was supported by grants from ANR (ImpmiR) and the Fédération de Recherche en Biologie de Toulouse (FRBT).

Acknowledgements

We are grateful to B. Guiard, B. Frances and J.-M. Lassalle for helpful discussions throughout this work and to H. Halley, S. Pech, J. Auriol and S. Ethuin for their technical assistance. We thank G. Canal for help with the R software and the ABC Facility of ANEXPLO Toulouse for the mouse husbandry.

Conflict of Interest statement. None declared.