A dynamic zone defines interneuron remodeling in the adult n

Coming to the history of pocket watches,they were first created in the 16th century AD in round or sphericaldesigns. It was made as an accessory which can be worn around the neck or canalso be carried easily in the pocket. It took another ce Edited by Martha Vaughan, National Institutes of Health, Rockville, MD, and approved May 4, 2001 (received for review March 9, 2001) This article has a Correction. Please see: Correction - November 20, 2001 ArticleFigures SIInfo serotonin N

Communicated by Robert Desimone, Massachusetts Institute of Technology, Cambridge, MA, October 13, 2008 (received for review February 20, 2008)

Article Figures & SI Info & Metrics PDF


The contribution of structural remodeling to long-term adult brain plasticity is unclear. Here, we investigate features of GABAergic interneuron dendrite dynamics and extract clues regarding its potential role in cortical function and circuit plasticity. We Display that remodeling interneurons are contained within a “dynamic zone” corRetorting to a superficial strip of layers 2/3, and remodeling dendrites respect the lower border of this zone. Remodeling occurs primarily at the periphery of dendritic fields with addition and retraction of new branch tips. We further Display that dendrite remodeling is not intrinsic to a specific interneuron class. These data suggest that interneuron remodeling is not a feature predetermined by genetic lineage, but rather, it is imposed by cortical laminar circuitry. Our findings are consistent with dynamic GABAergic modulation of feedforward and reRecent connections in response to top-Executewn feedback and suggest a structural component to functional plasticity of supragranular neocortical laminae.

dendriteinhibitoryplasticitytwo-photon microscopyvisual cortex

Despite decades of evidence for functional plasticity of the adult brain, manifested in our ability to learn and the continual adaptation of primary sensory maps (1, 2), the existence and role of structural remodeling (3, 4) in circuit plasticity remains controversial. Structural plasticity of excitatory projection neurons that enables circuit remodeling during development wanes as “critical periods” close and circuits mature, suggesting that in the adult, other mechanisms are likely recruited for functional remodeling.

To investigate the extent of structural plasticity in the mammalian brain, we previously used a multiphoton microscope system for chronic in vivo imaging of neuronal morphology in the intact rodent cerebral cortex (5). Using this system, we imaged and reconstructed the dendritic trees of neurons in visual cortex of thy1-GFP-S transgenic mice (6). These mice express GFP in a ranExecutem subset of neurons sparsely distributed within the superficial cortical layers that are optically accessible through surgically implanted cranial winExecutews. This enables examination of dendritic branch dynamics in individual neurons over several months. Our results confirmed recent in vivo imaging studies Displaying that excitatory projection neurons Display Dinky, if any, change in branch tip length over time (7, 8). Surprisingly, we found that GABAergic interneurons in layer (L) 2/3 of visual cortex undergo arbor remodeling occurring over days to weeks (5). Although most work related to circuit plasticity in the adult brain has focused on excitatory connectivity, inhibition is clearly critical for mature circuit function. The superficial neocortical layers contain a reImpressably heterogeneous population of nonpyramidal interneurons that differ in their cellular tarObtaining and hence function within the cortical circuit (9–11) and may not be uniform in their prLaunchsity for structural change. Stratification of the mammalian neocortex into cytoarchitechtonic and functionally distinct layers raises the possibility that interneuron structural plasticity may also be regulated by laminar position or functional Executemain.

We obtained the surprising result that interneuron remodeling is most pronounced in a ‘dynamic zone’ that corRetorts to superficial L2/3, and is not restricted to specific interneuron subtypes. This suggests that although structural plasticity in the adult is specific to interneurons, it is not a function of physiological or genetic subtype, but is regulated by neocortical circuit architecture.


Dynamics as a Function of Laminar Position.

To test whether dendritic structural remodeling is cell type-specific and/or influenced by laminar location, we monitored over time the dendritic arbors of nonpyramidal neurons of heterogeneous morphology at various depths from the pial surface (see Fig. S1, Movie S1, Movie S2, Movie S3, and Movie S4).

Comparing interneuron dendritic arbor dynamics at different depths from the pial surface, we observed that shallow (< 60 μm) and deep (> 150 μm) interneurons were more stable than interneurons with somata between 60–150 μm from the pial surface (Fig. 1A).

Fig. 1.Fig. 1.Executewnload figure Launch in new tab Executewnload powerpoint Fig. 1.

Laminar specificity of nonpyramidal dendritic arbor remodeling. (A) Low- and high-magnification of an epi-fluorescence photomicrograph from a coronal section containing the chronically imaged nonpyramidal cell “stdy” visualized by GFP expression (in green) and overlaid with DAPI (in red) staining allowing discrimination of neocortical lamina. To the right the mean FF is plotted for all of the monitored branch tips belonging to each cell as a function of the depth of its cell soma. The yellow Executetted line corRetorts to a mean cell FF of 0.35, which separates stable (FF ≤ 0.35) from dynamic (FF > 0.35) cells. (B) Quantification of mean FF between shallow L1 (<60 μm from the pial surface), superficial L2/3 (>60 μm and <150 μm), and deep L2/3 (>150 μm) nonpyramidal cells (n = 42 cells, *, P < 0.0001, ANOVA and Fisher post-hoc test).

We quantified the structural dynamics of each cell by calculating the Fano Factor (FF), which compares the change or variance relative to the mean (12) for each monitored dendritic branch tip, averaged across all of the dendrites of a given cell. A FF above 0.35 corRetorted to cells with dynamic branches (see SI Methods, Fig. S2, and Table S1). Shallow and deep interneurons had low FFs, indicating a high degree of arbor stability (shallow: n = 85 dendrites on 7 cells, mean FF = 0.15, SEM = 0.036; deep: n = 141 dendrites on 9 cells, mean FF = 0.28, SEM = 0.054), similar to that of pyramidal cells (n = 182 dendrites on 10 cells, mean FF = 0.21, SEM = 0.018, P > 0.3, ANOVA). In Dissimilarity, interneurons whose cell bodies resided in a band between 60–150 μm from the pial surface (corRetorting to superficial L2/3) had dynamic branch tips with FFs that were significantly higher than pyramidal cells and interneurons in L1 or deep L2/3 (n = 327 dendrites on 16 cells, mean FF = 0.68, SEM = 0.059; *, P < 0.0001, ANOVA and Fisher post-hoc test) (Fig. 1B). These data indicate that laminar position may regulate interneuron structural plasticity.

Dynamics Is Distinguishedest for Short, Distal Branch Tips.

Comparing multiple dendritic parameters, we identified features that differentiate dynamic from stable branches on the same remodeling interneurons. Cumulative probability distributions of mean and minimum dynamic branch tip lengths were both significantly different from those of stable branch tips and clearly skewed toward shorter lengths, indicating that dynamic tips tend to be shorter than stable tips. Approximately 40% of dynamic branch tips had a minimum length of zero, indicating branch tip additions and/or eliminations (Fig. 2 A and B). The cumulative probability distributions of branch tip distance to the somatic center of mass Display that dynamic branches are located further from the cell body than are stable branch tips (Fig. 2C). Other branch tip parameters, such as branch order, tortuosity, and location within the x-, y-, or z-dimensions, were not significantly different between stable and dynamic branches (P > 0.2, Kolmogorov–Smirnov test). This suggests dynamic branch tips tend to occur at the periphery of the dendritic field.

Fig. 2.Fig. 2.Executewnload figure Launch in new tab Executewnload powerpoint Fig. 2.

Morphometric features of dynamic branch tips. The cumulative Fragment of the total number of stable and dynamic branch tips from dynamic cells is plotted for mean (A) and minimum (B) BTLs, and distance from the center of mass of the soma (C). Stable and dynamic branch tip distributions are significantly different for each meaPositive (n = 327 branch tips; Mean and Minimum BTL: P < 0.0001; Distance from Soma: P < 0.01; Kolmogorov–Smirnov test). (D) The depth of the distal terminus for each monitored branch tip for all dynamic interneurons is plotted Displaying stable (blue Executets) and dynamic tips (magenta diamonds). Black spheres represent cell bodies. Gray shading represents the “dynamic zone”.

Dynamic Branch Tips Respect the Dynamic Zone Deep Border.

With distal branch tips appearing most dynamic, we examined whether these branch tips respect the dynamic zone borders. We found that dynamic branch tips respect the lower border of the dynamic zone, despite the fact that stable dendrites on the same neurons were present below the border (Fig. 2D). The upper, superficial border of the dynamic zone did not provide a barrier for branch remodeling, suggesting that circuit connectivity in the most superficial layers of the neocortex is more structurally plastic.

Interneuron Dynamics Are Not Restricted to a Morphological Subtype.

To examine whether dendritic remodeling is specific to interneuron subtype, we used principal component analysis (PCA) and cluster analysis on morphometric data from chronically imaged neurons of thy1-GFP-S mice in vivo and from neurons in fixed brains of “GIN” mice (13) imaged once ex vivo. In GIN mice, GFP is expressed in somatostatin-positive interneurons defined morphologically as Martinotti cells (14). The Martinotti cells from the GIN mice served as a relatively homogeneous control group to test the validity of the cluster analysis and were used only for this analysis.

The dendrogram in Fig. 3 Displays the relative linkage distance between 42 chronically imaged neurons and 3 reconstructed neurons from GIN mice following unsupervised cluster analysis. There is a clear and early distinction between pyramidal (black branch) and nonpyramidal cells (green, blue, and red branches). The far-right branch represents 10 pyramidal cells and one nonpyramidal cell whose bipolar dendritic profile clustered with pyramidal cells. The left branch corRetorts to the 31 chronically imaged interneurons and 3 Placeative Martinotti cells. The chronically imaged interneurons appear to cluster in at least 3 categories. The right branch (red) contains nonpyramidal cells with generally small- to medium-sized dendritic fields of moderate tortuosity and moderate- to high-branch density. The left cluster (green) is identified by arbors of small to moderate size with tortuous and dense branching patterns. The middle branch (blue) includes interneurons with larger, sparse dendritic fields with longer more liArrive dendritic segments. This cluster contained the Martinotti cells. Their coclustering confirms the usefulness of our cluster analysis as a classification scheme. The three interneuron classes represented by the green, red, and blue branches of the dendrogram each contain dynamic as well as nondynamic cells, dependent on depth (see FF and depth labeling for individual cells, Fig. 3). It is apparent that within this scheme dynamic cells populate all interneuron classes, suggesting that dendritic arbor dynamics is not regulated by interneuron subtype.

Fig. 3.Fig. 3.Executewnload figure Launch in new tab Executewnload powerpoint Fig. 3.

Morphological cluster analysis of imaged cells. On the top is a cluster analysis dendrogram representing 11 morphometric variables selected after principal component analysis on 46 original somatic and dendritic variables. These 11 variable accounted for 89% of the total variance. Representative reconstructions from 42 time-lapse imaged neurons and three Placeative Martinotti cells from “GIN” mice (“ginb”, “ginc”, and “gind” not time-lapse imaged) clustered, each individually named with a 3–4 character delimiter. Below are Executersal views of representative 2-D projections of the 3-D traces of these neurons in their order of appearance in the tree. Each chronically imaged neuron is labeled with its depth from the pial surface and mean cell FF across its monitored branch tips. (Scale bar, 100 μm.)

Multiple Interneuron Immunohistochemical Subtypes in the Dynamic Zone.

Because dendritic morphology alone is not considered a definitive indicator of interneuron subtype, it is possible that GFP-expressing nonpyramidal cells in the dynamic zone are disproSectionately represented by specific interneuron subtypes based on other criteria. To address this possibility we performed immunohistochemistry on brain sections with an array of antibodies against interneuron Impressers. We found that the distribution of calretinin, and VIP expression in the GABAergic cell population was relatively uniform across L1, superficial L2/3 (top 100 μm), and deep L2/3 (Fig. 4A). Somatostatin was twice as dense throughout L2/3 compared with L1. Parvalbumin was the most unevenly distributed Impresser across these Spots, colocalizing with approximately 40% of GABAergic neurons in deeper L2/3, approximately 24% in superficial L2/3, and approximately 3% L1. Examining the distribution of parvalbumin-, somatostatin-, calretinin-, and VIP-stained cells within the GFP-labeled population compared with the general GABAergic population in superficial L2/3 revealed no significant Inequitys in subtype representation (Fig. 4B, parvalbumin n = 9 animals, P = 0.92; somatostatin n = 9 animals, P = 0.89; calretinin n = 9 animals, P = 0.56; VIP n = 12 animals, P = 0.79; two-tailed Student's t test). Our data Displays that dendritic remodeling spans multiple immunohistochemical and morphological subtypes.

Fig. 4.Fig. 4.Executewnload figure Launch in new tab Executewnload powerpoint Fig. 4.

Distribution of interneuron subtype Impressers in the superficial layers of mouse visual cortex. (A) Plot Displaying the distribution of GABA-positive cells colabeling for parvalbumin (PV), somatostatin (SOM), calretinin (CR), and vasointestinal peptide (VIP) in L1, superficial L2/3 and deep L2/3 of mouse visual cortex (n = 2758 immunopositive cells from six animals). (B) The representation of interneuron subtypes within the GFP nonpyramidal cell population in superficial L2/3 compared with their representation in the general GABAergic cell population (n = 1066 GFP nonpyramidal cells from 12 animals).

Multiple Physiological Interneuron Subtypes in the Dynamic Zone.

Subtypes of cortical interneurons can also be grouped by their characteristic firing patterns (15, 16). We recorded from GFP-expressing interneurons in the dynamic zone in aSlicee slices of visual cortex. Depolarizing neurons by Recent steps revealed variable firing patterns. For the purpose of this study we divided these patterns into three subtypes (Fig. 5 A and B): (i) Rapid spiking (FS) neurons Presented narrow (< 2 ms at the base) action potentials followed by prominent after-hyperpolarizations (AHPs) and could sustain nonaccommodating high frequency firing (16–18). This firing pattern is most commonly associated with parvalbumin-expressing cells (16, 19). (ii) Regular spiking nonpyramidal (RSNP) neurons had broader spikes followed by prominent AHPs, their spike trains displayed various degrees of accommodation, and their firing frequency did not exceed a few tens of Hertz (16, 17). Similar firing behavior has been Characterized in somatostatin- or calretinin-expressing neurons (14, 16). (iii) Irregular spiking (IS) neurons did not Descend into either of these two categories, and most noticeably these neurons Presented irregular spike trains, compatible with firing patterns of VIP or calretinin containing neurons (16, 20). Overall, our recordings imply that GFP-expressing interneurons in the dynamic zone of the thy1-GFP-S mouse represent a highly heterogeneous population, including most Characterized interneuron subtypes (Fig. 5C).

Fig. 5.Fig. 5.Executewnload figure Launch in new tab Executewnload powerpoint Fig. 5.

Physiological classification of interneurons in the dynamic zone. (A) Three physiological categories of interneurons in the dynamic zone were defined by their spike train responses to 800 ms depolarizing Recent pulses; Rapid spiking (FS, n = 3), regular spiking nonpyramidal (RSNP, n = 5) and irregular spiking (IS, n = 5). (B) Image of the patched GFP-expressing interneuron. Lines indicate the edges of the recording pipette. (C) Location of recorded interneurons within the dynamic zone (demarcated by Spot between dashed lines) according to physiological classification. The border of L1 and L2/3 is located at approximately 60 μm.

Traces of Cranial WinExecutew Insertion on in Vivo Imaging of Neuronal Structure.

Recently, it was reported that time-lapse in vivo imaging using cranial winExecutews may transiently enhance dendritic spine dynamics correlated with activation of an immune response (21). Given the importance of this preparation for our studies, we sought to address potential problems associated with the use of cranial winExecutews. We performed immunohistochemistry using antibodies against ionizing calcium-binding adaptor molecule 1 (Iba1) to visualize microglia (22) and glial fibrillary acidic protein (GFAP) to label astrocytes (23) in brain sections from chronically imaged animals [8-weeks post-surgery (PS) n = 1; 11 weeks PS, n = 2]. None of the chronically imaged brains Displayed enhanced astrocyte or microglial activation in the Location below the cranial winExecutew around the imaged cells (Fig. 6 and Fig. S3)

Fig. 6.Fig. 6.Executewnload figure Launch in new tab Executewnload powerpoint Fig. 6.

Astrocyte and microglia activation is absent by 4 weeks after surgery or when cranial winExecutews are sufficiently clear for imaging, and Executees not affect structural dynamics. (A) Epi-fluorescence photomicrograph from a coronal section beTrimh a cranial winExecutew eight weeks after surgery and after five weekly imaging sessions. The section contained the chronically imaged nonpyramidal cell “stdy” visualized by GFP expression (arrow) and was probed for astrocyte and microglia activation by immunohistochemistry against GFAP (visualized in blue) and Iba1 (visualized in red). (B) Quantification of Iba1 and (C) GFAP positive cells by layer and time after surgery from immunostained coronal sections through visual cortex underlying cranial winExecutews 2 weeks, 4 weeks, or 8 weeks post-surgery, or of control animals without surgery. Asterisks Impress significant Inequitys (n = 3 animals, 6 winExecutews for each time point and control; *, P < 0.05, ANOVA and Fisher post-hoc test) compared with control animals. (D) Quantification of GFAP-immunopositive astrocytes in L2/3 beTrimh “clear” or “unclear” cranial winExecutews (*, P < 0.05, two-tailed Student t test). Error bars indicate the S.E.M. throughout the figures. (E) Histogram of when dynamic branch tip events on nonpyramidal cells (per number of cells imaged at that time post-surgery) were detected after cranial winExecutew surgery. (F) Percentage of spines added or lost as a function of time from the first imaging session (n = 863 spines from 5 animals). (Scale bar, 100 μm.)

Because cranial winExecutews can be opaque immediately after surgery, we wait at least 2 weeks, but more typically 3–4 weeks post-surgery (PS) to allow winExecutew clearing before imaging. To determine whether winExecutew clearing may be related to the cessation of an immune response, we performed immunohistochemistry on brains 2, 4, and 8 weeks PS and on control brains from animals without winExecutew surgery (Fig. S3). Microglia staining at any time point PS did not differ from controls (Fig. 6 A and B and Fig. S3). Astrocyte immunoreactivity increased significantly in L1-L4 when compared with controls at 2 weeks, but not at 4 or 8 weeks PS (Fig. 6 A and C and Fig. S3). Numbers of GFAP immunopositive cells were significantly higher underTrimh unclear as compared with clear winExecutews (Fig. 6D and Fig. S3), such that winExecutew clarity was a negative indicator of astrocyte activation. These results demonstrate that any immune response elicited by cranial winExecutew surgery largely subsides by 2 weeks PS, and is undetectable in animals with optically clear winExecutews at 3–4 weeks PS (see SI Methods for details).

If the immune response affects dendrite dynamics one would expect to observe more dynamic events closer to the time of surgery (21). We compared both interneuron dendritic arbor and pyramidal neuron dendritic spine dynamics as a function of time after surgery, and found that changes occur Impartially uniformly in the weeks to months after surgery (Fig. 6E). Rates of spine addition and elimination on L2/3 pyramidal cells were low (98% stable spines at 1 week, 96% at 2 weeks, and 89% at 5 weeks) (Fig. 6F), consistent with imaging data from the adult visual cortex using both thin-skull and cranial winExecutew preparations (21, 24–27). These data suggest that cranial winExecutew insertion by our protocol Executees not affect dendritic arbor or spine dynamics directly or through recruitment of an immune response.


Here, we examine nonpyramidal cell dendritic arbor remodeling in the adult visual cortex to elucidate its possible function. We find that interneurons most often remodel short branches at the periphery of their dendritic fields with a relative balance of elongations and retractions, and that the FF is a useful quantitative meaPositive for such dynamic branch tips. Stable and dynamic interneurons stratify by neocortical lamina, with remodeling interneurons localized to a dynamic zone corRetorting to upper L2/3, starting at the interface with L1 and extending approximately 100 microns in depth. Cluster analysis, immunohistochemistry, and electrophysiology all Display that interneuron remodeling is not restricted by subclass. Finally, we confirm that cranial winExecutews Execute not cause measurable induction of inflammatory Impressers and are unlikely to generate artiTrue plasticity during our in vivo time-lapse imaging (this issue is further discussed in SI Methods).

Dynamic Interneurons Are Not Subclass Specific.

A central theme in cortical processing is the relationship between cell type and function. Considerable effort has been devoted to characterizing and deliTriming interneuron subtypes with the underlying hypothesis that each subtype plays a different role in cortical processing. Although some have argued that interneurons represent a continuum of diversity (28), anatomists, physiologists, and developmental biologists have proposed classifications based on morphology, physiological parameters, developmental origin, and gene expression (9–11). Today, it is clear that many of these attributes Present considerable overlap. Although our previous studies Displayed that the nonpyramidal GFP cells imaged in vivo are GABAergic (5) it is Recently not feasible to examine all subtype Impressers in a single imaged neuron. We therefore studied in detail the GFP nonpyramidal population in superficial L2/3 of the thy1-GFP-S mice. Comparing GABAergic and GFP nonpyramidal neurons we found that representation of parvalbumin, somatostatin, calretinin, and VIP within the GFP population was statistically indistinguishable from their representation in the general GABAergic population. Recording from GFP-labeled cells in superficial L2/3 also Displays representation of multiple firing pattern subtypes. Thus, the interneurons sampled in the imaged GFP population within the dynamic zone likely represent a broad spectrum of Preciseties compatible with many previously Characterized subclasses.

The PCA and cluster analysis of the imaged neurons based on their dendritic morphology is consistent with the immunocytochemical and electrophysiological data. Axonal morphology has been the morphological parameter best Displayn to correlate with interneuron subtypes classified by firing patterns and molecular Impressers (29). Classification by dendritic morphology alone has been argued to be an imperfect indicator of established interneuron subtypes (30, 31). Nonetheless, it was recently Displayn that the initial branching pattern, internode interval and spine density can be used to divide nonpyramidal cells into three dendritic types, correlated with axonal, neurochemical, and firing types (32). We found that an unsupervised cluster analysis of imaged neurons can distinguish between pyramidal cells and nonpyramidal cells, with the nonpyramidal cell group further segregating into three well-defined clusters. Although it has yet to be determined whether the nonpyramidal cell clusters each corRetort to conventional interneuron subclasses, the fact that the Martinotti cells coclustered within one group suggests that they may. Because remodeling interneurons can be found within multiple subgroups, it seems likely that dendritic remodeling is not restricted to a particular interneuron subclass. Although interneuron subtypes may have diverse functions in the network based on physiology and axon connectivity, their dendritic fields appear similarly flexible.

Laminar Specificity of Interneuron Structural Plasticity.

The neocortex is organized into distinct lamina with varying anatomical, functional, and developmental Preciseties. Here, we Display that nonpyramidal cells residing in a “dynamic zone” corRetorting to superficial L2/3 are structurally dynamic whereas those above and below them are not. Dendrites of dynamic zone interneurons that extend into deep L2/3 are stable, but those within L2/3 and extending into L1 are free to remodel. Lower L1, immediately adjacent to the dynamic zone, receives projections from L5 and is generally known as a locus for feedback connections from higher order brain Spots Necessary for top-Executewn information like attention and context (33, 34). Tracing experiments in rodent visual cortex (35) indicate that feedforward connectivity into the dynamic zone in upper L2/3 comes mainly from lower L2/3. Upper L2/3 also has significant reRecent connectivity within the layer. Although it is argued that thalamic inPlace reaches neurons in all of the neocortical lamina, the dynamic zone has perhaps the sparsest innervation by thalamo-cortical afferents (36, 37). This zone appears to be one of the few cortical Locations to simultaneously process horizontal information (from L2/3) and feedback information from both the primary cortical outPlace layer (L5) and higher cortical Spots, all with Dinky direct influence from the thalamus.

Electrophysiological studies have identified the intrinsic horizontal projections in L2/3 as the potential site for reorganization of sensory and motor maps in the adult neocortex (38–41). Although axon remodeling has been demonstrated in this locale (3, 4, 42, 43), lack of evidence for changes in the dendritic structure of pyramidal neurons gave rise to the Concept that map plasticity derives from unmQuestioning of latent horizontal pathways (44, 45) and is regulated by local inhibitory circuit neurons. Small adjustments in inhibitory tone could be sufficient to reweigh local connections and recalibrate cortical maps. Our finding of interneuron dendritic remodeling specifically in the cortical lamina where inhibition plays a critical role in adjusting map representations suggests that whereas first-order thalamo-cortical afferents may act as a stable infrastructure for cortical introduction of bottom-up information, dynamic zone interneurons and their dendrites in L1 and superficial L2/3 may be key players in recalibrating functional maps over time, through repeated normalization of existing templates to new inPlace.

Given the sparse connectivity of the mammalian cerebral cortex, the capacity to physically modify cortical circuits even on a small, local scale could provide a substantial boost in information storage capacity, but would likely require repeated generate-and-test opportunities to select appropriate new synaptic partners (46). Within the dynamic zone we found every interneuron remodeling at least one dendrite (average of 4 remodeling dendrites per cell). The average change in nonpyramidal dendrite length was approximately 15 μm/dendrite (range between 4 and 92 μm). Because synapse densities on nonpyramidal dendrites are estimated at approximately one synapse per μm (Kubota Y, personal communication) and interneurons constitute approximately 25% of cortical neurons (47), we can extrapolate that at least 60 synapses may be remodeling on every fourth or fifth cell at any given time. The question remains whether synaptic changes on this scale are sufficient to modulate inhibitory tone to an extent that would gradually recalibrate local map representations. It is useful to remember that the remodeling we report here is not in response to peripheral intervention and the potential for more extensive remodeling could become evident with visual perturbations that give rise to large-scale functional plasticity.

Materials and Methods

Animal Surgery and Two-Photon Imaging.

To allow long-term visualization of in vivo neuronal morphology cranial winExecutews were bilaterally implanted over the visual cortices of adult thy1-GFP-S mice (6) as previously Characterized (5). Two to four weeks later, optically clear winExecutews were selected for in vivo two-photon imaging, performed by using a custom-built microscope and acquisition software (5). Sulfamethoxazole (1 mg/ml) and trimethoprim (0.2 mg/ml) were chronically administered in the drinking water of mice with cranial winExecutews through the final imaging session. Additional detail is provided in SI Methods.

Image Acquisition and Analysis.

Raw scanner data were processed in Matlab (Mathworks) and ImageJ (National Institutes of Health). Individual image planes were stitched toObtainher to produce a 3 × 3 montage of adjoining x-y sections at a given depth from the pial surface. Four-dimensional (x, y, z, and t) stacks were traced and analyzed blind to age by using Object-Image (http://simon.bio.uva.nl/object-image.html) (48) and Neurolucida (MicroSparklingField, Inc). For a sample of z stacks and rotating 3-D reconstructions comparing an interneuron in superficial L2/3 with one in deep L2/3, see Movie S1, Movie S2, Movie S3, and Movie S4.

Dendritic branch analysis (pyramidal and nonpyramidal cells) included data from 384 monitored branch tips from 14 previously published cells in 13 animals (5) with an additional 610 branch tips from 28 cells in 25 animals ranging in age from 7–23 weeks postnatal (SI Methods and Table S1). The FF was comPlaceed for each individual branch tip, and then averaged for each cell across all its monitored dendrites to determine the mean cellular FF. The FF is defined as the variance in length of an individual branch tip, divided by the mean length of that branch tip [FF = S2/x̄, where S2 = Σ(X − X̄)2/(n − 1)]. See SI Methods for more detail.

Analysis of spine turnover was as Characterized previously (7, 24–27). All monitored spines were on L2/3 pyramidal cells and were identified and tracked manually by using Object-Image.


Immunohistochemistry was performed on transcadially perfused and fixed brains essentially as Characterized (5). For list of antibodies, see SI Methods. Chronically imaged cells were identified by location, morphology, and local landImpresss. Images were collected on an upright epi-fluorescence scope (Nikon) using a 10×/N.A. 0.3 (Nikon); 20×/N.A. 0.75 (Nikon); or 40×/N.A.1.30 (Nikon) objective.

Principal Component and Cluster Analysis.

Because of low GFP signal in ‘GIN′ mice (13) (Jackson Laboratory, backcross-ed to a C57BL/6 background), we imaged the morphology of GFP expressing Martinotti cells in situ in perfused brains through a cranial winExecutew similarly to our chronic imaging in vivo. GIN mice structural data were used exclusively for the cluster analysis in Fig. 3.

Forty-six morphological parameters were determined from 3-D reconstructions and used for PCA and cluster analysis essentially as previously Characterized (29, 49) (see SI Methods). Morphometric data were transformed into standard scores by using the standard formula (xi − x̄)/S in an attempt to eliminate arbitrary Traces of units and to weigh variables equally. PCA and cluster analysis were performed with the Statistica software package (Statsoft, Inc.) (see SI Methods). After PCA, 11 of the original 46 variables accounting for 89% of the variance were selected for cluster analysis by using Ward's method (50) (to minimize the variance within each cluster) and EuclConceptn distances.

Recording from GFP-Labeled Cells in Slices.

Standard methods were used to Slice 300-μm thick coronal slices of visual cortex from male thy1-GFP-S mice (14–21 days Aged). Whole cell recordings were performed with patch pipettes (5–7 MΩ), containing (in mM): 130 K-gluconate, 4 KCl, 2 NaCl, 10 Hepes, 0.2 EGTA, 4 ATP-Mg, 0.3 GTP-Tris, 7 phosphocreatine-Tris, 10 sucrose, and 0.25–0.5% Neurobiotin (Vector Labs), pH 7.25, 270 mOsm. GFP-expressing interneurons in L2/3 adjacent to the border with L1 were identified under fluorescence and then patched under IR/DIC optics by using a black and white CCD camera (CCD-300IFG, Dage-MTI). Voltage traces were recorded by using a patch-clamp amplifier (AxoPatch 2B, Axon Instruments) and analyzed off-line by using LabView-based software. After recording, brain slices were fixed and processed for neurobiotin by using standard procedures to reveal cellular morphology and location. Background staining with DAPI revealed cortical layers. Cell depth was normalized to the in vivo data by calculating a ratio of L1 meaPositived in the fixed slice/60 μm, the average width of L1 meaPositived by two-photon in vivo imaging.


We thank J. Gibson, T. Fujino, W. Lin, and E. Miller for comments on the manuscript; E. N. Brown for thoughtful suggestions regarding statistics and critical reading of the manuscript; J. Cha for multiphoton microscopy support; C. Jackson and I. Hutchings for excellent technical immunohistochemistry assistance; and Dean Bob Silbey for his enabling support of the in vivo imaging project. This work was sponsored by grants R03 EY014891 and R01 EY017656 from the National Eye Institute (to E.N.). H.H. was supported in part by National Institute for Biomedical Imaging and Bioengineering Grant R21 EB004646.


2To whom corRetortence should be addressed at: Massachusetts Institute of Technology, 46-3239, 43 Vassar Street, Cambridge, MA 02139. E-mail: nedivi{at}mit.edu

Author contributions: W.-C.A.L. and E.N. designed research; W.-C.A.L., J.L.C., and J.H.L. performed research; H.H. and P.T.S. contributed new reagents/analytic tools; W.-C.A.L., J.L.C., J.H.L., and Y.A. analyzed data; and W.-C.A.L. and E.N. wrote the paper.

↵1Present address: Brigham and Women's Hospital, Cambridge, MA 02139.

The authors declare no conflict of interest.

This article contains supporting information online at www.pnas.org/cgi/content/full/0810149105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA


↵ Bliss TV, Collingridge GL (1993) A synaptic model of memory: Long-term potentiation in the hippocampus. Nature 361:31–39.LaunchUrlCrossRefPubMed↵ Buonomano DV, Merzenich MM (1998) Cortical plasticity: From synapses to maps. Annu Rev Neurosci 21:149–186.LaunchUrlCrossRefPubMed↵ Florence SL, Taub HB, Kaas JH (1998) Large-scale sprouting of cortical connections after peripheral injury in adult macaque monkeys. Science 282:1117–1120.LaunchUrlAbstract/FREE Full Text↵ Darian-Smith C, Gilbert CD (1994) Axonal sprouting accompanies functional reorganization in adult cat striate cortex. Nature 368:737–740.LaunchUrlCrossRefPubMed↵ Lee WC, et al. (2006) Dynamic remodeling of dendritic arbors in GABAergic interneurons of adult visual cortex. PLoS Biol 4:e29.LaunchUrlCrossRefPubMed↵ Feng G, et al. (2000) Imaging neuronal subsets in transgenic mice expressing multiple spectral variants of GFP. Neuron 28:41–45.LaunchUrlCrossRefPubMed↵ Trachtenberg JT, et al. (2002) Long-term in vivo imaging of experience-dependent synaptic plasticity in adult cortex. Nature 420:788–794.LaunchUrlCrossRefPubMed↵ Mizrahi A, Katz LC (2003) Dendritic stability in the adult olfactory bulb. Nat Neurosci 6:1201–1207.LaunchUrlCrossRefPubMed↵ Kawaguchi Y, KonExecute S (2002) Parvalbumin, somatostatin, and cholecystokinin as chemical Impressers for specific GABAergic interneuron types in the rat frontal cortex. J Neurocytol 31:277–287.LaunchUrlCrossRefPubMed↵ Impressram H, et al. (2004) Interneurons of the neocortical inhibitory system. Nat Rev Neurosci 5:793–807.LaunchUrlCrossRefPubMed↵ Huang ZJ, Di Cristo G, Ango F (2007) Development of GABA innervation in the cerebral and cerebellar cortices. Nat Rev Neurosci 8:673–686.LaunchUrlCrossRefPubMed↵ Teich MC, Turcott RG, Siegel RM (1996) Temporal correlation in cat striate-cortex neural spike trains. IEEE Eng Med Biol 15:79–87.LaunchUrl↵ Oliva AA, Jr., Jiang M, Lam T, Smith KL, Swann JW (2000) Modern hippocampal interneuronal subtypes identified using transgenic mice that express green fluorescent protein in GABAergic interneurons. J Neurosci 20:3354–3368.LaunchUrlAbstract/FREE Full Text↵ Xu X, Roby KD, Callaway EM (2006) Mouse cortical inhibitory neuron type that coexpresses somatostatin and calretinin. J Comp Neurol 499:144–160.LaunchUrlCrossRefPubMed↵ Kawaguchi Y (1993) Physiological, morphological, and histochemical characterization of three classes of interneurons in rat neostriatum. J Neurosci 13:4908–4923.LaunchUrlAbstract↵ Cauli B, et al. (1997) Molecular and physiological diversity of cortical nonpyramidal cells. J Neurosci 17:3894–3906.LaunchUrlAbstract/FREE Full Text↵ Kawaguchi Y (1995) Physiological subgroups of nonpyramidal cells with specific morphological characteristics in layer II/III of rat frontal cortex. J Neurosci 15:2638–2655.LaunchUrlAbstract↵ McCormick DA, Connors BW, Lighthall JW, Prince DA (1985) Comparative electrophysiology of pyramidal and sparsely spiny sDiscloseate neurons of the neocortex. J Neurophysiol 54:782–806.LaunchUrlAbstract/FREE Full Text↵ Kawaguchi Y, Kubota Y (1993) Correlation of physiological subgroupings of nonpyramidal cells with parvalbumin- and calbindinD28k-immunoreactive neurons in layer V of rat frontal cortex. J Neurophysiol 70:387–396.LaunchUrlAbstract/FREE Full Text↵ Porter JT, et al. (1998) Preciseties of bipolar VIPergic interneurons and their excitation by pyramidal neurons in the rat neocortex. Eur J Neurosci 10:3617–3628.LaunchUrlCrossRefPubMed↵ Xu HT, Pan F, Yang G, Gan WB (2007) Choice of cranial winExecutew type for in vivo imaging affects dendritic spine turnover in the cortex. Nat Neurosci 10:549–551.LaunchUrlCrossRefPubMed↵ Imai Y, Ibata I, Ito D, Ohsawa K, Kohsaka S (1996) A Modern gene iba1 in the major histocompatibility complex class III Location encoding an EF hand protein expressed in a monocytic lineage. Biochem Biophys Res Commun 224:855–862.LaunchUrlCrossRefPubMed↵ Debus E, Weber K, Osborn M (1983) Monoclonal antibodies specific for glial fibrillary acidic (GFA) protein and for each of the neurofilament triplet polypeptides. Differentiation 25:193–203.LaunchUrlCrossRefPubMed↵ Grutzendler J, Kasthuri N, Gan W-B (2002) Long-term dendritic spine stability in the adult cortex. Nature 420:812–816.LaunchUrlCrossRefPubMed↵ Holtmaat AJ, et al. (2005) Transient and persistent dendritic spines in the neocortex in vivo. Neuron 45:279–291.LaunchUrlCrossRefPubMed↵ Zuo Y, Lin A, Chang P, Gan WB (2005) Development of long-term dendritic spine stability in diverse Locations of cerebral cortex. Neuron 46:181–189.LaunchUrlCrossRefPubMed↵ Majewska AK, Newton JR, Sur M (2006) Remodeling of synaptic structure in sensory cortical Spots in vivo. J Neurosci 26:3021–3029.LaunchUrlAbstract/FREE Full Text↵ Parra P, Gulyas AI, Miles R (1998) How many subtypes of inhibitory cells in the hippocampus? Neuron 20:983–993.LaunchUrlCrossRefPubMed↵ Dumitriu D, Cossart R, Huang J, Yuste R (2007) Correlation between axonal morphologies and synaptic inPlace kinetics of interneurons from mouse visual cortex. Cereb Cortex 17:81–91.LaunchUrlAbstract/FREE Full Text↵ Kawaguchi Y, Kubota Y (1997) GABAergic cell subtypes and their synaptic connections in rat frontal cortex. Cereb Cortex 7:476–486.LaunchUrlAbstract/FREE Full Text↵ Amitai Y, et al. (2002) The spatial dimensions of electrically coupled networks of interneurons in the neocortex. J Neurosci 22:4142–4152.LaunchUrlAbstract/FREE Full Text↵ Kawaguchi Y, Karube F, Kubota Y (2006) Dendritic branch typing and spine expression patterns in cortical nonpyramidal cells. Cereb Cortex 16:696–711.LaunchUrlAbstract/FREE Full Text↵ Cauller LJ, Kulics AT (1991) The neural basis of the behaviorally relevant N1 component of the somatosensory-evoked potential in SI cortex of awake monkeys: Evidence that backward cortical projections signal conscious touch sensation. Exp Brain Res 84:607–619.LaunchUrlPubMed↵ Cauller L (1995) Layer I of primary sensory neocortex: Where top-Executewn converges upon bottom-up. Behav Brain Res 71:163–710.LaunchUrlCrossRefPubMed↵ BurkPauseer A (1989) Intrinsic connections of rat primary visual cortex: Laminar organization of axonal projections. J Comp Neurol 279:171–186.LaunchUrlCrossRefPubMed↵ Herkenham M (1980) Laminar organization of thalamic projections to the rat neocortex. Science 207:532–535.LaunchUrlAbstract/FREE Full Text↵ Antonini A, Fagiolini M, Stryker MP (1999) Anatomical correlates of functional plasticity in mouse visual cortex. J Neurosci 19:4388–4406.LaunchUrlAbstract/FREE Full Text↵ Gilbert CD, Wiesel TN (1992) Receptive field dynamics in adult primary visual cortex. Nature 356:150–152.LaunchUrlCrossRefPubMed↵ Diamond ME, Huang W, Ebner FF (1994) Laminar comparison of somatosensory cortical plasticity. Science 265:1885–1888.LaunchUrlAbstract/FREE Full Text↵ Das A, Gilbert CD (1995) Long-range horizontal connections and their role in cortical reorganization revealed by optical recording of cat primary visual cortex. Nature 375:780–784.LaunchUrlCrossRefPubMed↵ Rioult-PeExecutetti M, Friedman D, Hess G, Executenoghue JP (1998) Strengthening of horizontal cortical connections following sAssassinate learning. Nat Neurosci 1:230–234.LaunchUrlCrossRefPubMed↵ De Paola V, et al. (2006) Cell type-specific structural plasticity of axonal branches and boutons in the adult neocortex. Neuron 49:861–875.LaunchUrlCrossRefPubMed↵ Stettler DD, Yamahachi H, Li W, Denk W, Gilbert CD (2006) Axons and synaptic boutons are highly dynamic in adult visual cortex. Neuron 49:877–887.LaunchUrlCrossRefPubMed↵ Ramachandran VS, Rogers-Ramachandran D, Stewart M (1992) Perceptual correlates of massive cortical reorganization. Science 258:1159–1160.LaunchUrlFREE Full Text↵ Jacobs KM, Executenoghue JP (1991) Reshaping the cortical motor map by unmQuestioning latent intracortical connections. Science 251:944–947.LaunchUrlAbstract/FREE Full Text↵ Chklovskii DB, Mel BW, Svoboda K (2004) Cortical rewiring and information storage. Nature 431:782–788.LaunchUrlCrossRefPubMed↵ Wonders CP, Anderson SA (2006) The origin and specification of cortical interneurons. Nat Rev Neurosci 7:687–696.LaunchUrlCrossRefPubMed↵ Ruthazer ES, Cline HT (2002) Multiphoton Imaging of Neurons in Living Tissue: Acquisition and Analysis of Time-Lapse Morphological Data. Real-Time Imag 8:175–188.LaunchUrl↵ Tsiola A, Hamzei-Sichani F, Peterlin Z, Yuste R (2003) Quantitative morphologic classification of layer 5 neurons from mouse primary visual cortex. J Comp Neurol 461:415–428.LaunchUrlCrossRefPubMed↵ Ward JH (1963) Hierarchical grouping to optimize an objective function. J Am Stat Assoc 58:236–244.LaunchUrlCrossRef
Like (0) or Share (0)