Abstract
Juvenile zebra finches learn to sing by imitating conspecific songs of adults during a sensitive period early in life. Area X is a basal ganglia nucleus of the song control circuit specialized for song-related sensory–motor learning during song development. The structural plasticity and the molecular mechanisms regulating neuronal structure in Area X during song development and maturation are unclear. In this study, we examined the structure of spiny neurons, the main neuron type in Area X, at key stages of song development in male zebra finches. We report that dendritic arbor of spiny neurons expands during the sensitive period for song learning, and this initial growth is followed by pruning of dendrites and spines accompanied by changes in spine morphology as the song circuit matures. Previously, we showed that overexpression of miR-9 in Area X impairs song learning and performance and alters the expression of many genes that have important roles in neuronal structure and function (Shi et al., 2018). As an extension of that study, we report here that overexpression of miR-9 in spiny neurons in juvenile zebra finches reduces dendritic arbor complexity and spine density in a developmental stage-specific manner. We also show that miR-9 regulates the structural maintenance of spiny neurons in adulthood. Together, these findings reveal dynamic microstructural changes in the song circuit during the sensitive period of song development and provide evidence that miR-9 regulates neuronal structure during song development and maintenance.
Significance Statement
Song development in juvenile zebra finches provides a model to study sensitive period plasticity for language development and related neural developmental disorders in humans. Area X is a basal ganglia nucleus essential for song-related sensory–motor learning in the zebra finch. We show that dendritic arbor of spiny neurons in Area X undergoes an initial growth and expansion followed by pruning of dendrites and spines during song development and that this process is regulated by miR-9 in a developmental stage-specific manner. These findings reveal the temporal profiles of the structural development of key neurons in the basal ganglia song circuit and reveal a possible molecular mechanism for restricting sensitive period plasticity during vocal development.
Introduction
Juvenile zebra finches learn to sing by imitating conspecific songs of adults during a sensitive period early in life. The process of song learning and development involves two separate but overlapping phases, a sensory learning phase and a sensory–motor learning phase. Song learning begins at 25–30 d (d) of age. Juveniles first hear and memorize an adult song and begin to imitate the adult song by producing a highly variable subsong. After the sensitive period for sensory learning closes by 60 d, juveniles do not learn a new song, but they continue to consolidate the motor program of their song through auditory feedback-guided sensory–motor learning and practice. By 90 d, the sensory–motor phase is completed, and their song matures to a stable adult song, which they sing throughout life (Eales, 1985; Doupe and Kuhl, 1999; Tchernichovski et al., 2001).
Song behavior is controlled by a group of interconnected forebrain nuclei (Fig. 1A). Among them, Area X, an avian basal ganglia nucleus specialized for song, is critical for song learning in juveniles and for song maintenance in adult zebra finches (Sohrabji et al., 1990; Scharff and Nottebohm, 1991; Doupe et al., 2005; Fee and Goldberg, 2011). Area X contains multiple neuron types, including spiny neurons, pallidal projection neurons, cholinergic interneurons, low-threshold spike and fast-spiking interneurons, and a recently discovered group of excitatory glutamatergic interneurons (Farries and Perkel, 2002; Budzillo et al., 2017). The spiny neurons make up approximately two-thirds of all neurons in Area X and resemble the medium spiny neurons in the mammalian striatum in electrophysiological properties and in the expression of dopamine receptors (Farries and Perkel, 2002; Carrillo and Doupe, 2004; Reiner et al., 2004; Kubikova et al., 2010; Xiao et al., 2021). Area X spiny neurons receive glutamatergic excitatory inputs from the cortical nuclei HVC and the lateral magnocellular nucleus of the nidopallium (lMAN) to form part of the cortical–striatal circuit for song-related sensory and sensory–motor learning (Johnson et al., 1995; Ding and Perkel, 2004). The process of song development is presumably intertwined with the development and maturation of the song circuit. But the structural development of neurons in Area X and the molecular mechanisms that regulate the process remain unclear.
miR-9 is a brain-enriched miRNA that regulates the expression of transcription factors FOXP1 and FOXP2 (Otaegi et al., 2011; Shi et al., 2013). Mutations in FOXP1 or FOXP2 have been associated with language impairments and autism (Lai et al., 2001; Hamdan et al., 2010; den Hoed et al., 2021). In the zebra finch, proper expression of FoxP1 and FoxP2 in the song circuit is crucial for song learning and maintenance (Haesler et al., 2007; Murugan et al., 2013; Heston and White, 2015; Day et al., 2019; Norton et al., 2019; Garcia-Oscos et al., 2021; Xiao et al., 2021). We have shown that overexpression of miR-9 in Area X of juvenile zebra finches impairs song learning and performance and alters the expression of FoxP1 and FoxP2 and many genes encoding components of the synaptic machinery (Shi et al., 2018). The latter group includes synaptic adhesion proteins, synaptic scaffold proteins, and neurotransmitter receptors, suggesting that the behavioral phenotype of miR-9 overexpression can be mediated by alteration of neuronal structure in Area X. As an extension of the previous study and to further understand the role of miR-9 in regulating neuronal structure, in this study, we examined structural development of spiny neurons in Area X at key stages of song development and the effects of miR-9 overexpression on spiny neuron structure. We report that spiny neuron structure changes dynamically during song development. Specifically, following an initial growth and expansion, dendritic arbor and dendritic spines are pruned during song maturation. We further show that miR-9 modulates dendritic arbor and spine density of spiny neurons in an age-specific manner in juveniles and plays a role in the maintenance of spiny neuron structure in adulthood.
Materials and Methods
Animals
Animal usage was in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and was approved by the Louisiana State University School of Medicine Institutional Animal Care and Use Committee. All experiments were conducted in male zebra finches (Taeniopygia guttata). Juvenile zebra finches of specific ages were obtained from our in-house breeding colony, and all animals were housed under a 12 h (6 A.M.–6 P.M.) light/dark cycle.
Lentivirus production and stereotaxic injection
The lentiviral vector we used carries an mCherry fluorescent marker driven by the human ubiquitin C promoter (hUBC). The lenti-control virus is an empty vector, and the lenti-miR-9 virus expresses the zebra finch miR-9-3. Lentivirus packaging typically yields a titer of 1 × 109/ml. We diluted the viruses (>100-fold) in our injection to sparsely label individual neurons so they can be clearly traced without interference from neighboring neurons. For stereotaxic viral injections, typically we injected each bird with the lenti-control virus into Area X of one hemisphere and the lenti-miR-9 virus into Area X of the opposite hemisphere alternating which hemisphere received which virus. Details of the lentivirus constructs, virus packaging, and stereotaxic injection were described previously (Shi et al., 2018). Virus injection for juvenile birds was performed at approximately 25 d of age, and brains were collected at specified ages. Virus injection for adult birds was performed at 150–240 d of age, and brains were collected 1 month after virus injection.
Tissue processing and confocal imaging
Animals were perfused with PBS followed by 4% paraformaldehyde (PFA). Brains were postfixed for 8 h in 4% PFA solution at 4°C, followed by equilibration in 30% sucrose solution for 48 h at 4°C. Brains were embedded in Tissue-Tek, frozen, and cut into 80-µm-thick sagittal sections. DARPP-32 (dopamine- and cAMP-regulated phosphoprotein of 32 kDa) is a signaling molecule downstream of dopamine receptors and a marker for striatal medium spiny neurons (Greengard et al., 1999). Brain sections were stained with an anti-DARPP-32 antibody (see below for immunostaining procedures) and DAPI and were mounted with ProLong Glass mounting medium and covered with a #1 coverslip (Thermo Fisher Scientific). Images were acquired with an Olympus FV1200 confocal microscope equipped with lasers of 559 nm (mCherry), 473 nm (FITC/DARPP-32), and 405 nm (DAPI). The dendritic profile of each mCherry-labeled neuron was acquired using a 20× objective (NA = 0.75) with a 1.5 optical zoom. The frame size was 512 × 512 pixels, the z-stack interval was 0.364 µm, and the scanning speed was 8 µs/pixel. A laser power gradient was applied through the scanning depth. No deconvolution was performed. To acquire dendritic spine images, dendritic segments were imaged at higher resolution using a 100× oil-immersion objective (NA = 1.49) without optical zoom. The frame size was 1,024 × 1,024 pixels, the z-stack interval was 0.2 µm, and the scanning speed was 10 µs/pixel. The total scanning depth was 10–15 µm to cover the dendritic segments on the z-axis using a constant laser scanning power. No deconvolution was performed. In selecting dendritic segments for spine analysis, we required that the segments were from secondary or tertiary dendrites, at least 50 µm distal to the soma, 20–50 µm in length, and clear of crossing neighboring dendrites. Image data supporting the findings of this study are available from the corresponding author upon request.
Image analysis and quantification
From consecutive serial confocal images, a neuronal dendritic profile was reconstructed in three dimensions using the software package Imaris (Bitplane, 9.2.1). Dendrites of individual neurons were traced automatically using the Filament function of Imaris and confirmed by manual inspection. We required that all dendrites be connected to the soma, and disconnected dendrites were excluded from quantification. Dendrite length, branching, and Sholl analysis were measured and reported by Imaris. Spine quantification was performed using the Filament function of Imaris. Automatic thresholding was used to generate spine seed points. Spines were confirmed by visual inspection by a researcher requiring that each spine be connected to the dendritic shaft. Unconnected spines were excluded from quantification. Spine density was expressed as the number of spines per 10 µm dendrite. Spine head volume was reported by the Filament function of Imaris. For spine head volume analysis, data points at the lower and upper extremes representing tracing artifacts (top and bottom, 2.5%) were excluded from the analysis.
Immunohistochemistry and quantification
For DARPP-32 staining, brain sections were stained with a rabbit anti-DARPP-32 antibody (Abcam, ab40801; 1:750) overnight at 4°C, followed by incubation with an anti-rabbit secondary antibody (Invitrogen, 1:300) for 2 h at room temperature. After washing in PBS and staining DAPI, sections were mounted and coverslipped. For quantification of FoxP1, FoxP2, and FRMPD4, lentivirus was injected at 25–30 d, brains were collected at 60 d, and sagittal brain sections were stained with antibodies against FOXP1 (Sigma-Aldrich, HPA003876. 1:500), FOXP2 (Sigma-Aldrich, HPA000382. 1:500), and FRMPD4 (Sigma-Aldrich, SAB3500116. 1:500) for two overnights at 4°C followed by secondary antibody incubation. Imaging was performed similarly as described above for dendrites. To minimize variations across samples, conditions for immunostaining and confocal imaging were kept the same across all sections and all samples. Quantification of gene expression levels was performed by using the Surface function of Imaris. Imaris reports fluorescent signal intensity of DARPP-32 staining of individual neurons in a field of view of 140 µm × 140 µm within Area X (which typically contained one mCherry+/DARPP-32+ neuron and 60–70 DARPP-32+/mCherry− neurons, Fig. 1E). The DARPP-32 level of a virally transduced mCherry-labeled spiny neuron was normalized to and expressed as a percentage of the average level of all DARPP-32+ neurons in the field of view. Expression levels of FoxP1, FoxP2, and FRMPD4 were quantified similarly.
Experimental design and statistical analysis
Male zebra finches were used in this study because only males sing and have well-developed song circuitry. Lentiviruses expressing mCherry alone (control) or expressing mCherry and miR-9 were injected into Area X at specified ages via stereotaxic surgery and animals were killed at specified ages for analysis of spiny neuron structure. For neuron structure analysis, spiny neurons were identified by staining with an anti-DARPP-32 antibody. Image analysis was performed by an investigator blinded to sample identity except for the 45 d groups. The measurements of dendrites, spines, or gene expression levels of multiple neurons were averaged for each bird, and animal means ± SEM were reported and used in statistical analysis except for spine head size distribution analysis. The numbers of animals and neurons analyzed for each group are summarized in figure legends. Two-tailed t test or paired t test was used to compare two groups unless noted otherwise. Statistical differences among three or more groups were examined using one-way or two-way ANOVA. The distribution of spine head volume was analyzed by Kolmogorov–Smirnov test.
Results
Spiny neuron structure is dynamically regulated during song development
To investigate spiny neuron structure, we used a lentiviral vector that expresses the fluorescent marker mCherry to label and trace neurons for structural analysis. By diluting the virus concentration, we sparsely labeled individual neurons such that the dendritic arbor of one neuron was clearly separated from that of neighboring neurons, thus enabling quantitative analysis. We confirmed spiny neuron identity by immunohistochemical staining using an antibody against the spiny neuron marker DARPP-32 (Greengard et al., 1999). We imaged neurons labeled with virally expressed mCherry and DARPP-32 antibody staining using confocal microscopy and traced and quantified their dendritic arbor and dendritic spines using Imaris software. Example images of a double-labeled spiny neuron are shown in Figure 1C–G. Using these methods, we examined spiny neuron structure at key stages during song development of juvenile zebra finches. We injected the lentivirus expressing mCherry into Area X at 24–27 d, a time when song learning was just beginning, and we analyzed spiny neuron structure at 45, 60, 100, and 150 d. These time points were chosen as they represent typical milestones in song development.
Sparsely labeling of spiny neurons in Area X for quantitative structural analysis. A, A schematic diagram showing the song control circuit in a male zebra finch brain with Area X highlighted in green. B, The temporal trajectory of song development in male zebra finches and the experimental timeline for viral injection and neuronal structural analysis. C, A virally transduced and mCherry-labeled spiny neuron in Area X. D, Immunohistochemical staining of DARPP-32. E, Merge of images in C and D. F, An enlarged image showing part of the dendritic arbor in the lower-left quadrant in C in three dimensions with arrows pointing to example dendritic branching points. G, Reconstruction of the dendritic arbor of the spiny neuron shown in C in 3D using Imaris software. Colors are used to indicate the order of dendritic branching following each branch point. For example, purple labels primary dendrites, blue labels secondary dendrites, and green labels tertiary dendrites. Scale bars: C, 20 μm; F, 5 μm.
We first quantified dendritic branching (a dendritic branch point is where a new dendrite branch emerges along the dendrite shaft to form a bifurcation), which reflects the complexity of dendritic arbor, across the four developmental stages. Representative images of spiny neurons at 60 and 100 d are shown in Figure 2A. We found that the numbers of dendritic branch points of spiny neurons were significantly different across the four developmental stages [p = 0.0157, df = (3, 12), one-way ANOVA], exhibiting a biphasic profile. The number of dendritic branch points increased from 45 to 60 d and then decreased to 150 d. The mean numbers of dendritic branch points were 10 ± 0 at 45 d, 14 ± 1 at 60 d, 12 ± 0 at 100 d, and 10 ± 0 at 150 d (45 d vs 60 d, p = 0.0362, df = 7; 60 d vs 150 d, p = 0.0309, df = 6, t test; Fig. 2B). We also examined dendrite length of these spiny neurons of the four age groups. The average dendrite lengths (in µm) were 655.2 ± 13.1 at 45 d, 704.6 ± 38.9 at 60 d, 654.3 ± 30.1 at 100 d, and 638.8 ± 44 at 150 d (Fig. 2C). The total dendrite length did not change significantly across the four developmental stages [p = 0.534, df = (3, 12), one-way ANOVA]. One way to explain this lack of change in dendrite length might be that we only studied a portion of the entire dendritic arbor, since the long terminal dendrites could be lost when cutting 80-µm-thick tissue sections during sample processing. Overall, the patterns of dendritic branching suggest that the dendritic arbor of spiny neurons in Area X changes dynamically during song development with an initial phase of growth and expansion followed by pruning of dendritic branches as the song matures.
Dendritic arbor of Area X spiny neurons is dynamically regulated during song development. A, Example images of virally transduced Area X spiny neurons of 60 d (left) and 100 d (right) male zebra finches. The maximal projection images are generated using Imaris with background noise signals removed. Scale bar, 20 μm. B, Quantitative analysis of dendritic branching of spiny neurons at 45, 60, 100, and 150 d. C, Quantitative analysis of total dendrite length of spiny neurons at 45, 60, 100, and 150 d. Bar charts are plotted as animal means ± SEM. *p < 0.05; unpaired t test. Sample size: 23, 30, 18, and 19 neurons from four, five, four, and three animals were analyzed for the 45, 60, 100, and 150 d groups, respectively.
Dendritic spines are the main site for excitatory synaptic transmission in the vertebrate brain, and dendritic spine density and spine morphology can critically influence synaptic properties (Berry and Nedivi, 2017; Stein and Zito, 2019). We next examined dendritic spine density and spine structure of spiny neurons during song development. We traced 20–50-µm-long dendritic segments typically from secondary or tertiary terminal dendrites from the virally labeled spiny neurons and quantified spine density using Imaris software. Representative images of dendritic segments with protruding spines from 60 and 150 d male zebra finches are shown in Figure 3A. Overall, the mean spine density of spiny neurons changed significantly during song development [p = 0.0187, df = (3, 12), one-way ANOVA]. The average dendritic spine densities were (expressed as the number of spines per 10 µm of dendrite): 12.0 ± 2.2 at 45 d, 14.9 ± 1.3 at 60 d, 13.6 ± 0.8 at 100 d, and 9.6 ± 0.6 at 150 d. Spine density at 60 d was significantly higher than spine density at 150 d (p = 0.025, df = 6, t test; Fig. 3B). These results indicate that spine abundance of spiny neurons is dynamically regulated as juvenile zebra finches learn to sing and their song matures.
Dendritic spine density and spine structure of spiny neurons are dynamically regulated during song development. A, Representative images of dendritic segments showing spines of mCherry-labeled Area X spiny neurons at 60 and 150 d (left and middle panels: scale bar, 5 μm). The right panel shows a reconstructed image of the 60 d dendritic segment (left) within the boxed area. B, Quantification of dendritic spine density of spiny neurons at 45, 60, 100, and 150 d. C, Cumulative frequency curves showing the relative distribution of spine head volumes of spiny neurons at 60 and 150 d. D, Spine density of small and large spines at 60 and 150 d. In B and D, bar charts are plotted as animal means ± SEM. *p < 0.05 and **p < 0.01; Kolmogorov–Smirnov test for C and unpaired t test for B and D. Sample size: 26, 32, 34, and 25 dendritic segments from four, five, four, and three animals were analyzed for the 45, 60, 100, and 150 d groups, respectively.
While studies of mammalian hippocampal or cortical neurons often classify dendritic spines into stubby, mushroom, and thin spine subtypes, more recent studies suggest that spines represent a continuum of morphologic shape, and spine head volume is directly associated with synaptic strength and cognitive functions (Harris and Stevens, 1989; Harris et al., 1992; Matsuzaki et al., 2004; Arellano et al., 2007; Tonnesen et al., 2014; Berry and Nedivi, 2017; Ofer et al., 2021). We chose to directly measure spine head volume and analyzed its distribution at the spine population level, focusing on the 60 and 150 d groups. The cumulative frequency curves revealed that the distribution of spine head volume at 150 d shifted rightward significantly in comparison with the 60 d group (p < 0.0001; Kolmogorov–Smirnov test; Fig. 3C). The observation that the 150 d group had fewer spines but relatively larger spine head volume prompted us to ask whether spines of a particular size were more prone to spine loss during song maturation. We divided spines into two populations according to the median value of spine head volumes of all spines of all age groups, which was 0.047 µm3. Spines with a head volume of <0.047 µm3 were classified as small spines and spines with a head volume of >0.047 µm3 as large spines. Then, we plotted the spine density of the small and large spines separately. We found that the density of small spines at 60 d was significantly higher than that at 150 d (8.2 ± 0.7 at 60 d vs 3.3 ± 0.3 at 150 d, p = 0.0028, df = 6, t test), whereas there was no significant difference between the densities of large spines of the two age groups (7.8 ± 1.7 at 60 d vs 6.5 ± 0.7 at 150 d, p = 0.618, df = 6, t test; Fig. 3D). Together, these results suggest that small spines are preferentially lost as songs and the underlying circuit mature.
miR-9 overexpression impairs the structural development of spiny neurons in juvenile zebra finches
We have shown before that overexpression of miR-9 in Area X impairs song learning and performance and alters the expression of many genes that have important roles in nervous system development (Shi et al., 2018). To extend these findings, we examined the roles of miR-9 in regulating spiny neuron structure during song development using the methods described above by injecting a lenti-miR-9 virus into Area X. While dendritic branching at 45 d was not affected, the numbers of dendritic branch points of spiny neurons transduced with lenti-miR-9 were reduced by 35.7% at 60 d, 33.3% at 100 d, and 40.0% at 150 d compared with controls (p = 0.005058, df = 4 for 60 d; p = 0.000443, df = 3 for 100 d; p = 0.031504, df = 2 for 150 d; paired t test; Fig. 4B). miR-9’s effect on dendritic arbor complexity was also revealed by Sholl analysis which describes dendritic complexity by the number of intersections between dendrites and concentric circles centered on the neuronal soma in 3D. A higher number of intersections indicates a more complex dendritic arbor. As shown in Figure 4C, miR-9 reduced the numbers of crossings at various distances from the soma compared with the controls for both the 60 and 150 d groups [p = 0.0393, df = (11, 96) for 60 d; p = 0.0003, df = (11, 48) for 150 d; two-way ANOVA; Fig. 4C]. We also found that while dendrite length was not affected at 45 d, the total dendrite length of spiny neurons transduced with lenti-miR-9 was reduced by 27.4% at 60 d, 35.0% at 100 d, and 32.6% at 150 d compared with the controls (p = 0.0062, df = 4 for 60 d; p = 0.0053, df = 3 for 100 d; p = 0.027, df = 2 for 150 d; paired t test; Fig. 4D). Together, these results indicate that miR-9 plays a key role in regulating spiny neuron dendritic arborization during song development in an age-specific manner: while having no effect on dendritic arbor in young juveniles when both sensory and sensory–motor learning occur (e.g., 45 d), miR-9 regulates pruning of dendritic arbor of spiny neurons during song maturation and consolidation of the motor program of a song.
miR-9 overexpression reduces dendritic arbor of spiny neurons during song development. A, Example images of spiny neurons transduced with lenti-control or lenti-miR-9 virus analyzed at 60 d (note the control neuron is also shown in Fig. 2A). Scale bar, 20 μm. B, Dendritic branching is reduced in lenti-miR-9 transduced spiny neurons at 60, 100, and 150 d. C, Sholl profiles of lenti-control and lenti-miR-9 transduced spiny neurons analyzed at 60 and 150 d. In Sholl profiles, the x-axis indicates the distance from neuronal soma, and the y-axis indicates the number of intersections between neuronal dendrites and concentric circles centered on the neuron's soma in 3D. D, Dendritic length is reduced in lenti-miR-9 transduced spiny neurons at 60, 100, and 150 d. E, DARPP-32 expression is reduced in lenti-miR-9 transduced spiny neurons at 60, 100, and 150 d. In B, D, and E, bar charts are plotted as animal means ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001. Two-way ANOVA for C and paired t test for B, D, and E. Note the control data are replotted from the data shown in Figure 2 for comparison with birds injected with lenti-miR-9. For miR-9 groups, 18, 22, 20, and 17 neurons from four, five, four, and three animals were analyzed for the 45, 60, 100, and 150 d groups, respectively.
miR-9 is known to downregulate the expression of DARPP-32, a key component of the dopamine signaling pathway, in Area X (Shi et al., 2018). Since we cannot measure miR-9 expression level in the same neurons used in morphological analysis, we examined DARPP-32 protein level in these neurons by immunohistochemical staining as indirect evidence for miR-9 overexpression (an example of DARPP-32 immunostaining of virally transduced spiny neurons is shown in Fig. 1E). We found that DARPP-32 protein level was reduced in spiny neurons transduced with lenti-miR-9 compared with the controls by 21.3% at 60 d, 16.9% at 100 d, and 13.1% at 150 d (p = 0.0005, df = 4 for 60 d; p = 0.0081, df = 3 for 100 d; p = 0.0479, df = 2 for 150 d; paired t test; Fig. 4E). The reduction of DARPP-32 expression in individual spiny neurons transduced with lenti-miR-9 virus, in accordance with our previous finding by Western blot performed with protein homogenates from virally transduced Area X tissues (Shi et al., 2018), provides evidence that miR-9 was successfully overexpressed in virally transduced spiny neurons.
We next examined the effects of miR-9 overexpression on dendritic spines in the four age groups. Figure 5A shows example dendritic segments of spiny neurons transduced with lenti-control or lenti-miR-9 at 60 and 150 d. While spine density was not affected at 45 d, spine density was reduced in lenti-miR-9 transduced neurons compared with the controls by 45.7% at 60 d and 40.5% at 100 d (p = 0.005, df = 4 for 60 d; p = 0.032, df = 3 for 100 d; paired t test; Fig. 5B). There was a trend of spine reduction at 150 d (26.6%), but it did not reach statistical significance (p = 0.064, df = 2 for 150 d, paired t test; Fig. 5B). We also examined miR-9’s effects on spine head volume focusing on the 60 and 150 d groups. At 60 d, the distribution of spine head volume shifted rightward in lenti-miR-9 transduced spiny neurons compared with controls (p < 0.001; Kolmogorov–Smirnov test; Fig. 5C). At 150 d, the distribution of spine head volume shifted leftward compared with the controls (p < 0.001; Kolmogorov–Smirnov test; Fig. 5D). One way to explain the differential effects on spine head volume between 60 and 150 d can be that miR-9 preferentially reduces spines of a particular size group in an age-dependent manner. To test this possibility, we examined miR-9’s effects on small and large spines of the 60 and 150 d groups. We divided the total spine population into small and large spines according to the median spine head volume (0.047 µm3) as described before and calculated spine density for small and large spines separately. For the 60 d group, the density of small spines was significantly reduced by 46.2% in lenti-miR-9 transduced spiny neurons compared with the controls (p = 0.029, df = 4, one-tailed paired t test), whereas the density of large spines did not change significantly in lenti-miR-9 transduced spiny neurons (p = 0.147, df = 4, paired t test; Fig. 5E). For the 150 d group, while small spine density did not change, the density of large spines was significantly reduced by 36.1% in lenti-miR-9 transduced spiny neurons compared with the controls (p = 0.04, df = 2, paired t test; Fig. 5F). These results show that miR-9 overexpression reduces the number of spines in a spine size- and age-specific manner during song development and maturation.
miR-9 overexpression reduces spine density and alters spine structure during song development. A, Representative images of dendritic segments of spiny neurons transduced with lenti-control or lenti-miR-9 viruses analyzed at 60 and 150 d, respectively. Scale bar, 5 μm. B, Spine density is reduced in lenti-miR-9 transduced spiny neurons during song development. C, D, Cumulative frequency curves showing the distributions of spine head volumes are altered in lenti-miR-9 transduced spiny neurons at 60 and 150 d. E, The density of small spines is reduced in lenti-miR-9 transduced spiny neurons at 60 d. F, The density of large spines, but not small spines, is reduced in lenti-miR-9 transduced spiny neurons at 150 d. In B, E, and F, bar charts are plotted as animal means ± SEM. *p < 0.05 and **p < 0.01; Kolmogorov–Smirnov test for C and D and paired t test for B, E, and F. Note the control data are replotted from the data shown in Figure 3 for comparison with birds injected with lenti-miR-9. For miR-9 groups, 19, 30, 21, and 21 dendritic segments from four, five, four, and three animals were analyzed for the 45, 60, 100, and 150 d groups, respectively.
miR-9 regulates the maintenance of spiny neuron structure in adult zebra finches
Male zebra finches reach adulthood at ∼90 d; from this time onward, they sing a stable adult song throughout life, which presumably is supported by a structurally and functionally mature song circuit. A mature song, however, needs active maintenance including proper gene expression in Area X (Day et al., 2019; Xiao et al., 2021). It is also known that miR-9 expression in Area X is regulated by singing behavior in adults (Shi et al., 2013). Therefore, we investigated whether miR-9 has a role in the maintenance of spiny neuron structure in a mature circuit. We injected lenti-miR-9 or lenti-control viruses into Area X of adult male zebra finches (5–10 months of age) and analyzed spiny neuron structure 1 month later. Compared with the controls, dendritic arbor complexity was reduced in spiny neurons transduced with lenti-miR-9 resulting in fewer dendritic branch points (11 ± 1 for lenti-control vs 8 ± 0 for lenti-miR-9; p = 0.035, df = 2, paired t test; Fig. 6A) and shorter total dendritic length (584 ± 44.1 µm for lenti-control vs 453.8 ± 41.5 µm for lenti-miR-9; p = 0.0123, df = 2, paired t test; Fig. 6B). The expression of DARPP-32 was reduced slightly (11.7%) in lenti-miR-9 transduced spiny neurons, but it was not statistically significant (p = 0.062, df = 2, paired t test; Fig. 6C).
miR-9 regulates spiny neuron structure maintenance in adult Area X. A, Dendritic branching is reduced in lenti-miR-9 transduced spiny neurons in adult Area X. B, Dendrite length is reduced in lenti-miR-9 transduced spiny neurons in adult Area X. C, DARPP-32 expression in spiny neurons transduced with lenti-miR-9 or lenti-control in adult Area X. D, Dendritic spine density of spiny neurons transduced with lenti-miR-9 or lenti-control in adult Area X. E, The densities of small and large spines of spiny neurons transduced with lenti-miR-9 or lenti-control in adult Area X. Bar charts are plotted as animal means ± SEM. *p < 0.05; paired t test. Sample size: 16 and 18 neurons from three animals were analyzed for dendrite branching and length, 14 and 15 neurons from three animals were analyzed for DARPP-32, and 19 and 18 dendritic segments from three animals were analyzed for spine density for the control and miR-9 groups, respectively.
We next examined the effects of miR-9 overexpression on spine density of spiny neurons in Area X of adults. There was no significant change in spine density in spiny neurons transduced with lenti-miR-9 compared with the controls (number of spines per 10 µm dendrite: 11.2 ± 1.8 for lenti-control vs 8.5 ± 0.5 for lenti-miR-9; p = 0.178, df = 2, paired t test; Fig. 6D). We also divided the spine populations of each group into small and large spines as before and examined miR-9’s effect on small and large spines separately. There was no change in the density of small spines; the density of large spines was reduced in lenti-miR-9 transduced spiny neurons, but the reduction did not reach statistical significance (number of spines per 10 µm dendrite: small spines, 3.6 ± 0.3 for lenti-control vs 4.6 ± 0.5 for lenti-miR-9; p = 0.186, df = 2; large spines, 7.7 ± 1.2 for lenti-control vs 4.6 ± 0.4 for lenti-miR-9; p = 0.067, df = 2, paired t test; Fig. 6E). Together, these results indicate that miR-9 plays a role in the maintenance of spiny neuron structure in Area X of adult zebra finches.
miR-9 regulates the expression of FoxP1, FoxP2, and FRMPD4 in spiny neurons
To further understand the molecular mechanisms downstream from miR-9 in regulating neuronal structure, we examined the expressions of miR-9 downstream genes in lenti-miR-9 transduced individual spiny neurons in Area X focusing on FoxP1, FoxP2, and FRMPD4. FoxP1 and FoxP2 are transcription factors that regulate many neuronal genes important for neuronal structure. FRMPD4 is a postsynaptic scaffolding protein that has important roles in synaptogenesis and spine formation (Lee et al., 2008; Piard et al., 2018). In this experiment, we injected the lenti-miR-9 or lenti-control virus into Area X at ∼25 d and examined protein expression levels of these genes in virally transduced spiny neurons at 60 d by immunostaining (Fig. 7A). Compared to the controls, in spiny neurons transduced with lenti-miR-9, FoxP1 expression was significantly reduced by 23.4% (p = 0.017, df = 6, t test; Fig. 7B), FoxP2 expression was significantly reduced by 19.2% (p = 0.0146, df = 8, t test; Fig. 7C), and FRMPD4 expression was significantly reduced by 21.5% (p = 0.0003, df = 8, t test; Fig. 7D). Together, these results indicate that on an individual neuron level, lenti-miR-9 transduction reduces the expression of genes important for neuronal structure and that reduced expressions of these genes likely have contributed to the effects of miR-9 on spiny neuron structure.
miR-9 regulates gene expression in virally transduced spiny neurons. A, Example images of Area X sections stained with antibodies against FOXP1, FOXP2, and FRMPD4, respectively. The left panels show fluorescent signals (green) from antibody staining; the middle panels show spiny neurons sparsely transduced with lenti-control or lenti-miR-9 expressing mCherry; images in the left and middle panels are merged in the respective right panels. Scale bar, 40 µm. B–D, Reduced expression of FoxP1, FoxP2, and FRMPD4 in spiny neurons transduced with lenti-miR-9. Relative expression levels of FoxP1, FoxP2, and FRMPD4 in virally transduced spiny neurons (mCherry-labeled) are expressed as a percentage of the average expression levels of these genes in all neurons in the field of view. Data are presented as animal means ± SEM. *p < 0.05 and ***p < 0.001, unpaired t test. Sample size: 15 and 22 neurons from three and five animals were analyzed for FoxP1, 21 and 20 neurons from four and six animals were analyzed for FoxP2, and 12 and 15 neurons from four and six animals were analyzed for FRMPD4 for the control and miR-9 groups, respectively.
Discussion
We demonstrate that dendrites and spines of Area X spiny neurons, a key component of the cortical–striatal song circuit, change dynamically during song development. The period between 45 and 60 d is characterized by expansion of dendritic arbor, and this initial growth is followed by reduction of dendritic arbor complexity and pruning of dendritic spines until well into adulthood. These patterns are consistent with patterns of postnatal brain development in mammals, where an initial growth of dendrites and synapses is followed by a net loss of dendrites and synapses (Rakic et al., 1986; Zuo et al., 2005; Koleske, 2013; Forrest et al., 2018). Spine density of neurons in lMAN is known to increase between 3 and 5 weeks of age and decrease thereafter (Nixdorf-Bergweiler et al., 1995). While biphasic morphological change might be a common theme in neuronal development, specific neuron types appear to adapt different temporal trajectories to fulfill their functions. New spiny neurons are continuously being added to Area X and integrated into the functional circuit in adulthood (Kosubek-Langer et al., 2017). Future work is needed to learn whether the adult-born spiny neurons follow a similar biphasic developmental profile. We show that 60 d marks a turning point in spiny neuron development. This time also marks the closure of the sensitive period for learning a new song. The growth and expansion of dendritic arbor and spines before 60 d presumably provide a large synaptic field that allows heightened plasticity favorable to sensory and sensory–motor exploration. After 60 d, the capacity of juveniles to learn a new song drastically declines (Eales, 1985). The pruning of dendritic arbor and spines after 60 d reflects the structural constraints that restrict plasticity for song learning and lead to consolidation of the song motor program. Although many details remain unknown, the temporal profile of spiny neuron structure likely serves as a key mechanism for restricting sensitive period plasticity for song development.
Studies in mammals have shown that spine turnover contributes to refinement of a neural circuit and its plasticity for learning and memory (Berry and Nedivi, 2017; Stein and Zito, 2019). In the zebra finch, imaging of the HVC neurons connecting to Area X shows that their dendritic spines turnover rapidly and a higher number of spines and larger spine size correlate with a greater capacity for song learning (Roberts et al., 2010). Our results, although do not distinguish whether changes in spine density resulted from changes in spine formation, elimination, or a combination of both, provide a snapshot of spine dynamics at spine population levels over a period spanning stages of song development. Consistent with findings in mammals that spine elimination is linked to improvements in learning (Yang et al., 2009; Fu et al., 2012), we found a drastic reduction (approximately 36%) in spine density between 60 and 150 d as the song matures. The actual spine loss could be even higher if shortening of dendrites is considered. It is intriguing that spiny neurons in 60 d juveniles have more small spines than in 150 d adults and that small spines are selectively lost during song maturation (Fig. 3D). These findings are consistent with the idea that small spines, while being more transient, provide a site for plasticity favorable for learning, whereas large spines are more stable for retaining memory and circuit stability (Matsuzaki et al., 2004; Yang et al., 2009; Kasai et al., 2010; Lai et al., 2012). The selective loss of small spines perhaps reflects the decreased capacity for song learning and increased circuit stability that enables the lifelong stability of an adult song.
Overall, we show that miR-9 negatively regulates neuronal structure by pruning dendrites and spines. This is consistent with the increase of miR-9 expression (Shi et al., 2013) and the reduction in dendritic arbor and spine density in Area X (Figs. 2 and 3) during normal song development. The effect of miR-9 overexpression on neuron structure appears to be developmental stage-specific. Unlike in 60 d and older animals, miR-9 overexpression does not affect dendrites and spines at 45 d. This is unlikely due to insufficient expression of miR-9 3 weeks after lentivirus injection, because virally expressed mCherry signal at 45 d was comparable to that at older ages. We interpret this result as suggesting that at 45 d, mechanisms that promote structural growth counterbalance the pruning effect of miR-9. We also observe that miR-9 overexpression reduces the number of smaller spines at 60 d but reduces the number of larger spines at 150 d (Fig. 5E,F). This observation suggests that the small and large spines not only differ in their head sizes but also differ in their molecular compositions, and this molecular heterogeneity changes as the song circuit matures. The molecular diversity of spines can be attributed to developmental stages, neuron subtypes, expression of neurotransmitter receptors, and/or local protein synthesis at the synapses (van Oostrum and Schuman, 2025). Gene expression changes vastly in the song circuit during song development (Shi et al., 2021). Presumably, genes targeted by miR-9 also change in spiny neurons during this time, resulting in age-dependent effects of miR-9 on spines. Future studies are needed to explore the molecular diversity of spiny neuron spines during song learning and maturation.
Adult zebra finch song requires active maintenance and perturbation of gene expression in adult Area X leads to song variability (Lombardino and Nottebohm, 2000; Brainard and Doupe, 2001; Day et al., 2019; Xiao et al., 2021). The adult song also retains some levels of plasticity (Kao et al., 2005; Tumer and Brainard, 2007; Andalman and Fee, 2009), and neural variability in Area X is a source of variation in syllable fundamental frequency which allows for motor exploration and adaptive song plasticity in adults (Heston et al., 2018; Kojima et al., 2018). Adult males sing an undirected song when singing alone or a courtship song directed toward a female (Jarvis et al., 1998). The acoustic features of undirected song are more variable than those of directed song, accompanied by variable firing patterns in Area X and calcium signals in spiny neurons (Ali et al., 2013; Woolley et al., 2014; Singh Alvarado et al., 2021). These findings indicate that Area X neurons are central to the plasticity of adult song. Interestingly, miR-9 expression in Area X is dynamically regulated when birds sing undirected songs (Shi et al., 2013). The finding that miR-9 regulates spiny neuron structure in adult Area X, although moderately, supports miR-9’s role in the maintenance and plasticity of a mature song. It would be of interest to investigate whether miR-9 modulates dendrites and spines of spiny neurons in response to the social context of singing.
Among the downstream genes of miR-9, FOXP1 and FOXP2 are known to regulate the structure and function of the mammalian striatum (Enard et al., 2009; Anderson et al., 2020). In the zebra finch, knocking down FoxP2 expression in Area X reduces spine density of spiny neurons, and a higher level of FoxP2 is correlated with more complex dendrites and higher spine density in adult-born spiny neurons (Schulz et al., 2010; Kosubek-Langer and Scharff, 2020). Knocking down FoxP1 in the HVC neurons projecting to Area X alters their spine density and reduces spine turnover (Garcia-Oscos et al., 2021). In addition to FoxP1 and FoxP2, overexpression of miR-9 in Area X regulates directly or indirectly many genes encoding components of the synaptic machinery (Shi et al., 2018). FRMPD4 is a postsynaptic scaffolding protein; it contains a FERM domain that binds glutamate receptors and a PDZ binding domain that interacts with PSD-95. The interaction of FRMPD4 with multiple postsynaptic proteins is crucial for dendritic spine morphogenesis (Lee et al., 2008; Hu et al., 2012). Mutations in FRMPD4 disrupt spine formation in mice and cause language delay and/or psychiatric disturbances in humans (Piard et al., 2018). The downregulation of FoxP1, FoxP2, and FRMPD4 (Fig. 7) is likely part of the molecular network mediating miR-9’s effect on spiny neuron structure.
Dendrites and synapses are complex protein apparatuses made of hundreds of proteins (Sudhof, 2018; Koopmans et al., 2019; van Oostrum and Schuman, 2025), and coordinated expression of these proteins is critical for proper neuronal structure and function. Because a single miRNA can regulate many genes (Bartel, 2018), miR-9 is well suited for a role in coordinating changes in dendrite and spine structure by regulating gene expression in response to environmental and experiential influences. miR-9 is also expressed in the striatum of rodents, and it plays a role in alcohol tolerance via regulating the stability of mRNA splicing variants of a potassium channel (Pietrzykowski et al., 2008). The basal ganglia are central to a variety of neurological disorders, and many neural developmental or psychiatric disorders are rooted in disruptions of dendrite and spine structures (Forrest et al., 2018; Runge et al., 2020). Knowledge of the roles of miR-9 in regulating neuronal structure in the basal ganglia circuit will help understand the molecular mechanisms underlying these disorders.
Footnotes
This work was supported by NIH Grant R01MH105519 to X.L. Z.F. was supported in part by NIGMS/NIH Grant U54 GM104940, which funds the Louisiana Clinical and Translational Science Center. We thank Dr. L. Marrero for advice on sample preparation and imaging and J. Romagnoli and L. Li for technical assistance. We also thank M. Schwing for the graphic work.
The authors declare no competing financial interests.
- Correspondence should be addressed to XiaoChing Li at xli4{at}lsuhsc.edu.
