Sensory experience alters neuronal circuits, which is believed to form the basis for learning and memory. On a microscopic level, structural changes of the neuronal network are prominently observable as experience-dependent addition and removal of cortical dendritic spines. By environmental enrichment, we here applied broad sensory stimulation to mice and followed the consequences to dendritic spines in the somatosensory cortex utilizing in vivo microscopy. Additionally to apical dendrites of layer V neurons, which are typically analyzed in in vivo imaging experiments, we investigated basal dendrites of layer II/III neurons and describe for the first time experience-dependent alterations on this population of dendrites. On both classes of cortical dendrites, enriched environment-induced substantial changes determined by increases in density and turnover of dendritic spines. Previously established spines were lost after enriched stimulation. A fraction of experience-induced gained spines survived for weeks, which might therefore be functionally integrated into the neuronal network. Furthermore, we observed an increased density of spines that appeared only transiently. Together, we speculate that the cognitive benefits seen in environmental-enriched animals might be a consequence of both, a higher connectivity of the neuronal network due to more established synapses and an enhanced flexibility due to more transient spines.
Plastic changes in the neuronal network are believed to form the basis for learning and memory (Bailey and Kandel 1993). These changes occur at synapses either on a molecular level altering the synaptic strength or on a structural level like through the addition and removal of synapses. In the mammalian neocortex, the ability of the network to modulate its connections is most prominent in the developing brain (Hubel and Wiesel 1970; Katz and Shatz 1996), but plasticity was found to remain throughout life (Stepanyants et al. 2002; Chklovskii et al. 2004).
Since Ramon y Cajal dendritic spines have been studied extensively, not least as they can be explored easily with microscopic techniques due to their characteristic morphology. Dendritic spines are small protrusions along the dendrites of pyramidal neurons with a size of 0.5–2 µm and constitute the postsynaptic site of the majority of excitatory synapses (Gray 1959; Nimchinsky et al. 2002). In organotypic slice cultures of hippocampal neurons, a local electrical stimulation, which was sufficient to induce long-term potentiation, lead to the de novo formation of dendritic spines (Engert and Bonhoeffer 1999; Maletic-Savatic et al. 1999; Nagerl et al. 2004, 2007). On the contrary, the retraction of spines was observed after low-frequency stimulation (Nagerl et al. 2004). Time-lapse imaging of dendritic spines in the brain of living mice was enabled by the combined use of multiphoton laser scanning microscopy with transgenic animals expressing fluorescent proteins in neurons (Grutzendler et al. 2002; Trachtenberg et al. 2002). In vivo imaging in the murine cortex revealed 2 classes of dendritic spines, due to their temporal characteristics. Spines that persist for 8 days are in large part stable for much longer time periods and are therefore defined as persistent spines, distinguished from transient spines (Holtmaat et al. 2005). Ultrastructural analysis of in vivo imaged dendritic spines showed that persistent spines have all features of a mature synapse, whereas transient spines lack a presynaptic partner (Knott et al. 2006). Similar results were obtained in cultured hippocampal neurons, confirming that spinogenesis precedes the formation of synapses (Nagerl et al. 2007).
Sensory manipulation was found to strongly affect spine dynamics in vivo in various areas of the cortex of adult mice (reviewed in Fu and Zuo 2011). Most of these investigations used adverse, denervating experimental paradigms like whisker trimming or visual deprivation. Only recently studies report that specific learning tasks like a grabing task, the rotarod, or fear conditioning are associated specifically with either the de novo appearance or, in the latter case, the loss of dendritic spines in corresponding cortical regions (Xu et al. 2009; Yang et al. 2009; Lai et al. 2012). Learning-induced new spines correlate positively with the behavioral outcome and were stable for months. Together, these observations suggest that the experience-dependent formation and the elimination of dendritic spines modulate the cortical neuronal network and might therefore represent a physical basis for life-long memory storage (Xu et al. 2009; Yang et al. 2009).
However, little is known about broad physiological stimulation on the kinetics of cortical dendritic spines. Here, we studied in detail structural plasticity induced by an extensive enriched environment in the somatosensory cortex. Environmental enrichment usually involves a combination of increased sensory, cognitive, and motor stimulation (van Praag et al. 2000). The enrichment stimulation has been shown to improve learning and memory (Meshi et al. 2006), ameliorate neuropathology in models of neurodegenerative diseases like Alzheimer's disease and Parkinsons disease (Bezard et al. 2003; Lazarov et al. 2005), and beneficially influence the consequences of brain injury (Kolb and Gibb 1991). On the structural level, an enriched environment alters brain weight (Bennett et al. 1969) and dendritic branching (Greenough et al. 1973; Mohammed et al. 2002) and increases hippocampal neurogenesis (Kempermann et al. 1997). The spine density of enriched animals was found to be elevated in static ex vivo analysis of rodent brain regions, including the somatosensory cortex (Kolb et al. 2003; Leggio et al. 2005; Gelfo et al. 2009).
The present study aims to expand this current state of knowledge by a temporal description of enriched environment-induced alterations on cortical dendrites, utilizing in vivo microscopy. This allows to quantify changes in spine density at the same dendritic element and to follow the fate of individual spines over time. Additionally, we not only focused on typically in vivo imaged tuft apical dendrites, but also included with basal dendrites of layer II/III neurons a second, poorly characterized population of dendrites.
Materials and Methods
Animals and Housing
Female mice expressing yellow fluorescent protein (YFP) under control of Thy-1 promoter were used in all experiments (YFP-H line) (Feng et al. 2000). Control mice were singly housed in standard cages (30 × 15 × 20 cm). For environmental enrichment, a large cage (80 × 50 × 40 cm) was used with a maximum of 10 animals. The cage was equipped with various nesting materials, platforms, tunnels, huts, and toys of various materials and textures. Furthermore, at least 2 running wheels were provided at a time to allow voluntary physical activity. The content of the cage was relocated and substituted for every 2–3 days, and the location of food and drinking facilities was changed as well. For the infant-start experiment, pregnant mice were transferred to the enriched cage, and after birth the rearranging of the cage was halted for 10 days. At an age of 3 weeks, mothers and male littermates were removed from enrichment, matching the time when control mice were separated from their mother. After surgery or imaging, enriched mice were kept isolated until they fully recovered from anesthesia. Supplementary Figure S1 summarizes in detail the procedure of the infant-start and the adult-start experiment.
All procedures were in accordance with an animal protocol approved by the University of Munich and the government of upper Bavaria.
As a general side note, it shall be mentioned that the broad sensory stimulation provided by the enriched environment rather resembles the natural condition of rodents. Standard caging as it is used by us and others as the control situation might therefore represent an impoverished condition, which might impact the conclusions and the values of such experiments.
Cranial Window Surgery
For in vivo imaging, a chronic cranial window was prepared as described previously (Fuhrmann et al. 2007; Holtmaat et al. 2009). The mice were anesthetized with an intraperitoneal injection of ketamine/xylazine (0.14/0.01 mg/g body weight; WDT/Bayer Health Care). Additionally, dexamethasone (6 µg/g body weight; Sigma) was intraperitoneally administered immediately before surgery. Utilizing the open-skull preparation, a cranial window was placed above the somatosensory cortex. For repositioning during repetitive imaging a small metal bar, containing a hole for a screw, was glued next to the window. After surgery, mice received subcutaneously analgesic treatment with carprophen (7.5 µg/g body weight; Pfizer) and antibiotic treatment with cefotaxim (0.25 mg/g body weight; Pharmore).
Two-Photon In Vivo Imaging
Imaging began after a 3–4-week recovery period post surgery utilizing a LSM510 setup (Zeiss) being equipped with a MaiTai laser (Spectra Physics). For in vivo imaging, mice were anesthetized by an intraperitoneal injection of ketamine/xylazine. Imaging sessions lasted for no longer than 60 min, and laser power was kept below 50 mW to avoid phototoxic effects. Two-photon excitation of YFP was performed at 880 nm, and a ×40 0.8 N water-immersion objective (Zeiss) was used. Stereological coordinates were used to locate the somatosensory cortex. Overview images were taken at low resolution (≈0.4 µm/pixel per image frame in 3 µm steps) up to a depth of 350 µm. These overviews included all cell bodies of YFP-labeled layer II/III neurons, typically at depth of 200–300 µm (Supplementary Fig. S2). Consequently, apical tuft dendrites protruding below the depth of 350 µm arise from layer V pyramidal neurons. Dendritic elements were imaged at high resolution (≈0.1 µm/pixel per frame with 1 µm z-resolution). Care was taken to ensure similar fluorescence levels in space and time.
Image Processing and Data Analysis
Fluorescence images were deconvolved (AutoQuantX2, Media Cybernetics) and filtered with an edge-preserving algorithm, followed by a local contrast change (Imaris 5.0.1, Bitplane). Image processing was performed to enhance contrast, to reduce noise, and to sharpen details, without introducing artifacts (Supplementary Fig. S2). For illustration of in vivo imaged dendrites, distracting neighboring dendritic elements were removed using Adobe Photoshop. Dendritic spines were counted manually and blinded to the experimental condition. In time-series images, a dendritic spine was identified as the same if its location did not change within a range of 1 µm along the dendrite. Because of resolution limitations in z-dimension, only protrusions emanating laterally from the dendritic shaft were analyzed.
Statistical testing was carried out using the GraphPad Prisms software. For column comparisons, student's t-test was conducted to control for statistical significance. To test time-series experiments, 2-way analysis of variance was performed (the significance level is indicated behind the corresponding graph), followed by Bonferroni post-testing for the single time-point comparison (the significance level is indicated above the corresponding data point). All results are reported as mean ± standard error of the mean (SEM), and differences were considered to be significant at P < 0.05.
Infant-Start Enrichment on Spines of Apical Tuft Dendrites of Layer V Neurons
In a first set of experiments, we studied mice that were kept in the enriched environment from birth to elucidate the maximal effect enrichment can have on the neuronal network. Therefore, pregnant mice were placed in the enriched cage that was designed as complex as possible, containing various objects and toys of different materials and textures (Fig. 1A). After weaning, only the female offspring remained in the enrichment. Cranial window implantation was performed at an age of 3 months (week 13) and at 4 months (week 17) 2-photon in vivo microscopy started (Fig. 1B, Supplementary Fig. S1). Parallel-treated single-caged mice at the same age served as controls.
The density of dendritic spines of apical tuft dendrites of layer V neurons (Fig. 1C) in the somatosensory cortex was elevated by a third in the enriched group (mean ± SEM; 0.46 ± 0.02 vs. 0.33 ± 0.02/µm dendrite, P > 0.001) and stayed stable over an imaging period of 30 days (Fig. 1D,E). Whereas the majority of dendritic spines were stable under both conditions, the turnover of spines in the enriched group was more than doubled. Specifically, the density of spines that were gained (0.034 ± 0.004 vs. 0.016 ± 0.002/µm dendrite, P > 0.001) and lost (0.035 ± 0.002 vs. 0.014 ± 0.002/µm dendrite, P > 0.001) between 2 imaging sessions was found to be dramatically increased in the enriched mice (Fig. 1D,F,G). The rates for gained and lost spines within the experimental groups are comparable, confirming the stable net spine density.
The fate of newly gained dendritic spines, which were formed within the time span of the first 3 imaging time points, was followed (Fig. 1H). A higher number of these gained spines persisted for at least 22 days in the enriched mice (0.032 ± 0.004 vs. 0.015 ± 0.003/µm dendrite, P = 0.0057) and are therefore believed to be functionally integrated into the neuronal network (Knott et al. 2006; Fig. 1H,J). Nevertheless, more of these new spines were only transiently present in the enriched mice (0.044 ± 0.005 vs. 0.013 ± 0.003/µm dendrite, P > 0.001; Fig. 1K).
Infant-Start Enrichment on Spines of Basal Dendrites of Layer II/III Neurons
In vivo microscopy of experience-dependent changes on basal dendrites of layer II/III neurons has not been reported so far, as these dendrites are located typically at depth of about 150–300 µm and are therefore relatively difficult to image (Supplementary Fig. S2). Under control conditions, the spine density of basal dendrites of layer II/III neurons (Fig. 1C) is higher when compared with apical tuft dendrites of layer V neurons (0.67 ± 0.02/µm dendrite). Infant-start enriched caging leads to an increased spine density of 0.84 ± 0.02/µm dendrite (P > 0.001; Fig. 2A,B). Spine turnover was found to be elevated as well, evidenced by enhanced gain of spines (0.047 ± 0.005 vs. 0.018 ± 0.002/µm dendrite, P > 0.001) and loss of spines (0.043 ± 0.004 vs. 0.020 ± 0.002/µm dendrite; Fig. 2C,D). Similar to the results for apical dendrites of layer V neurons, a higher number of newly gained spines persisted over 3 weeks (0.035 ± 0.006 vs. 0.020 ± 0.004/µm dendrite, P = 0.035), and the density of transient spines increased (0.043 ± 0.007 vs. 0.012 ± 0.003/µm dendrite, P > 0.001; Fig. 2E–H).
Adult-Start Enrichment Triggers Structural Plasticity of Apical Tuft Dendrites of Layer V Neurons
For the second part of our study, we were interested to elucidate how the mature neuronal network of single-caged mice reacts toward the broad stimulation of an enriched environment. For this purpose, we implanted a cranial window in adult female mice at an age of 4–5 months (17–21 weeks) that were caged under control conditions. Imaging started at 5–6 months of age (20–25 weeks) for 2 time points to record the baseline status of cortical dendrites. One day before the third imaging session, the mice were transferred to the enriched environment, and the stimulation-induced changes were followed at the previously imaged dendrites (Fig. 3A, Supplementary Fig. S2). Parallel-treated single-caged mice at a comparable age served as controls.
On apical tuft dendrites of layer V neurons, spine density increases after the mice were placed in the enriched cage (Fig. 3B). Significant differences were observed after 16 days, when spine density was elevated by about 20%. There was a dramatic increase in the formation of new spines at the first imaging session after enrichment stimulation (Fig. 3C). The enhanced density of gained spines remained at a high order for 2 weeks before declining to a level that was still above the baseline (Fig. 3C). Interestingly, the increase in lost spines started later and did not reach the range of gained spines (Fig. 3D). Questioning if changes in spine turnover were restricted to a cluster of specific dendrites, we plotted the gained and lost spine densities for every imaged dendrite (Fig. 3E). This analysis revealed an equal distribution of turnover per dendrite favoring a rather broad effect regarding all dendrites. Furthermore, this illustration highlights that the equilibrium of gain and loss of spines in control mice is shifted toward gained spines after environmental enrichment.
Next, we ask how established synapses react on the enrichment-associated stimulation. The fate of dendritic spines, which were stable over the 2 time points of baseline imaging and therefore considered as persistent spines, was followed over the whole period of enrichment. Whereas, in control animals, 86.8 ± 1.2% of established spines survived, enriched mice had a survival of only 80.4 ± 1.4% persistent spines (P > 0.001; Fig. 3F). This indicates a significant reorganization of the neuronal network after enriched environment.
Furthermore, we analyzed spines that were newly formed within the first 2 imaging sessions after enriched environment. A much higher number of gained spines remained stable over the following 4 weeks compared with control levels (0.044 ± 0.004 vs. 0.013 ± 0.003/µm dendrite, P > 0.001; Fig. 3H). The density of transient spines was elevated as well in the enriched group (0.025 ± 0.003 vs. 0.008 ± 0.002/µm dendrite, P > 0.001; Fig. 3J). Interestingly, comparing the survival of gained spines between the infant-start and the adult-start–enriched animals at a similar time point reveals a higher fraction of newly stable integrated spines in the latter group (infant start: 37% survival; adult start: 60% survival; Figs 1H and 3G).
Adult-Start Enrichment Triggered Structural Plasticity of Basal Dendrites of Layer II/III Neurons
The enrichment-induced changes of dendritic spines of basal dendrites of layer II/III recapitulate in a similar way the findings described above for apical tuft dendrites of layer V neurons. Spine density increases by more than 20% due to an instant rise in gained spines which is accompanied by a delayed and less prominent increase in lost spines (Fig. 4A–D). The observed changes in lost and gained spines are distributed evenly over the imaged dendritic branches (Fig. 4E). More persistent spines, which have existed under control conditions, get lost after acute enrichment stimulation over the period observed when compared with single-caged animals (12.7 ± 1.1% vs. 8.2 ± 1.2%, P < 0.001; Fig. 4F). Likewise, it turned out that enrichment-induced newly gained spines are to a higher proportion stable until the end of the imaging period (0.060 ± 0.006 vs. 0.016 ± 0.004/µm dendrite, P < 0.001; Fig. 4H). The density of transient spines is enhanced in the enriched group as well (0.021 ± 0.004 vs. 0.010 ± 0.003/µm dendrite, P < 0.05; Fig. 4J).
In the present study, we show that enriched environment strongly influences dendritic spines in the somatosensory cortex in both of our experimental in vivo paradigms, the infant-start and the adult-start enrichment. At 2 populations of dendrites, apical tuft dendrites of layer V neurons and basal dendrites of layer II/III neurons, the spine density is elevated, and the formation and elimination of spines are enhanced. More new spines are stabilized, and the sudden enrichment also leads to the elimination of established spines. This indicates that substantial and lasting changes of the neuronal network are induced by the enriched environment.
In postmortem tissue of rodents, studies have revealed an enrichment-induced increase in dendritic spine density in the hippocampus (Rampon et al. 2000) and striatum (Comery et al. 1995) as well as in various areas of the neocortex (Kolb et al. 2003; Leggio et al. 2005; Gelfo et al. 2009). In the cortex, this was consistent for apical and basal dendritic segments, although one study found no effect specifically on apical dendrites of layer V neurons in the parietal cortex (Gelfo et al. 2009). With the help of in vivo imaging, we confirmed that the enhanced cortical spine density induced by the enriched environment in an unbiased manner, as we were able to compare the same dendritic elements before and during enriched stimulation.
Beyond that, the in vivo approach certainly provides important information on spine turnover. The experience-dependent increase in the formation and elimination of dendritic spines has been shown mainly with adverse sensory manipulations, and only recently physiological challenges have also been used for stimulation (Fu and Zuo 2011). Consistent with our findings, Yang et al. (2009) also found a rapid formation of new spines and an increased elimination of existing spines in the somatosensory cortex, when whiskers were permanently stimulated. In an enrichment similar to ours, Fu et al. (2012) observed a doubled spine formation in the motor cortex within 4 days. All of these studies exclusively focused on apical tuft dendrites and, therefore, we additionally describe here for the first time experience-dependent changes on the basal dendrites of layer II/III neurons in vivo. Although showing under standard conditions a higher spine density, basal dendrites of layer II/III neurons are influenced by the enriched environment in the same way as apical dendrites in every parameter recorded here.
The mammalian cerebral cortex exhibits highest plasticity in developmental stages, when synapses are initially overproduced, but substantially decrease to form the mature, more stable neuronal network (Rakic et al. 1986; Hua and Smith 2004; Petanjek et al. 2011). Similarly, in the murine cortex the density of dendritic spines was found to decline in postnatal development reaching a balance of continuous gain and loss of spines at an age of about 4 months (Holtmaat et al. 2005; Zuo et al. 2005). We here observed a more pronounced elevation of spine density in infant-start–enriched mice, which was not reached by the adult-start group. This difference might therefore result from experiencing the enriched environment during the postnatal neuronal system development. Although the adult neuronal circuits retain the capability of structural changes, the young network features a higher flexibility. Interestingly, in aged rats, the enrichment-induced increase in spine density did not attain a significant difference in the cortex, whereas in the hippocampus of the same mice, the elevation in spine density was still prominent (Darmopil et al. 2009; Rasin et al. 2011). Hence, development but also aging influences the ability of structural spine plasticity in a brain region-dependant manner.
In the adult-start experiment, the switch from control to enriched housing was accompanied by a substantial rearrangement of the neuronal network. Importantly, the changes of the neuronal system started with the formation and stabilization of new spines, which was counteracted at later time points by a loss of spines. Xu et al. (2009) describe for the grabbing task a similar kinetic with a rapid gain of spines, which is later on followed by the elimination of pre-existing spines. For this specific task, the loss of spines leads to a balancing of synapse number resulting in an unaltered spine density. The broad stimulation of enriched housing might therefore cover a different quality of structural changes, leading to a net gain in spine density.
But how and to what degree do these structural changes in the neuronal network, induced by enriched environment, reflect specific experiences? The fact that environmental enrichment influences brain structure at various stages, ranging from increased brain weight, increased gliogenesis and hippocampal neurogenesis, and enhanced synaptogenesis (Bennett et al. 1969; Kolb and Gibb 1991; Kempermann et al. 1997; Soffie et al. 1999) argues for a more general influence. Furthermore in several brain regions, neurotrophic factors like brain-derived neurotrophic factor and nerve growth factor (NGF) are increased by enriched housing (Ickes et al. 2000). In line with this is our finding that modifications in dendritic spine turnover are evenly distributed over the imaged dendritic branches. Additionally, the observation that apical tuft dendrites of layer V neurons and basal dendrites of layer II/III neurons are in a similar way affected by enriched stimulation further speaks for general impact of the enriched environment on the neuronal network.
On the other hand, our results on structural plasticity parallel what has been shown for specific learning tasks, namely a grabbing task and motor learning on the rotarod (Xu et al. 2009; Yang et al. 2009). In all studies, established spines get lost after the new stimulus. Additionally, out of a large pool of newly gained spines only few get stabilized over long time, which might therefore represent a physical correlate of new memorizable experiences. Hence, one could imagine a combination of both, one more general aspect and one specific aspect of structural changes of dendritic spines in the enriched mice. Interestingly, the 2 classes of spines, persistent spines and transient spines, both are significant elevated under enriched conditions. Therefore, we speculate that the cognitive benefits seen in environmental-enriched animals might result from 1) a higher connectivity of the neuronal network, due to an increased number of established synapses and 2) an enhanced flexibility, due to more transient spines that, if needed, could be transformed into functional synapses.
This work was supported by the Deutsche Forschungsgemeinschaft (SFB 596) and the German Federal Ministry of Education and Research (Bundesministerium für Bildung und Forschung, 13N11130).
Specifically, we thank Sonja Steinbach for her dedication and excellent technical support. We further thank M. Biel, S. Michalakis, K. Keppler, M. Fuhrmann, S. Burgold, M. Dorostkar, and A. Gupta for helpful advice in planning the experiments and preparing the manuscript. Conflict of Interest: None declared.