rearrangement of synaptic connections with inhibitory neurons in developing mouse visual cortex

Rearrangement of Synaptic Connectionswith Inhibitory Neurons in Developing

Mouse Visual Cortex

AKIKO YAMASHITA, KATIA VALKOVA, YURI GONCHAR,

AND ANDREAS BURKHALTER*

Department of Anatomy and Neurobiology, Washington University School of Medicine,St. Louis, Missouri 63110

ABSTRACTCortical inhibition is determined in part by the organization of synaptic inputs to

�-aminobutyric acidergic (GABAergic) neurons. In adult rat visual cortex, feedforward (FF)and feedback (FB) connections that link lower with higher areas provide �10% of inputs toparvalbumin (PV)-expressing GABAergic neurons and �90% to non-GABAergic cells (Gon-char and Burkhalter [1999] J. Comp. Neurol. 406:346–360). Although the proportions ofthese targets are similar in both pathways, FF synapses prefer larger PV dendrites than FBsynapses, which may result in stronger inhibition in the FF than in the FB pathway (Goncharand Burkhalter [1999] J. Comp. Neurol. 406:346–360). To determine when during postnatal(P) development FF and FB inputs to PV and non-PV neurons acquire mature proportions,and whether the pathway-specific distributions of FF and FB inputs to PV dendrites developfrom a similar pattern, we studied FF and FB connections between area 17 and the higherorder lateromedial area (LM) in visual cortex of P15–42 mice. We found that the innervationratio of PV and non-PV neurons is mature at P15. Furthermore, the size distributions of PVdendrites contacted by FF and FB synapses were similar at P15 but changed during the thirdto sixth postnatal weeks so that, by P36–42, FF inputs preferred thick dendrites and FBsynapses favored thin PV dendrites. These results suggest that distinct FF and FB circuitsdevelop after eye opening by rearranging the distribution of excitatory synaptic inputs on thedendritic tree of PV neurons. The purpose of this transformation may be to adjust differen-tially the strengths of inhibition in FF and FB circuits. J. Comp. Neurol. 464:426–437, 2003.© 2003 Wiley-Liss, Inc.

Indexing terms: intracortical connections; extrastriate visual cortex; feedforward connections;

feedback connections; GABAergic neurons; parvalbumin

The relative magnitude and timing of excitation andinhibition are important for setting the operating gain, thedynamic range, and the stimulus selectivity of visual cor-tical neurons (Douglas et al., 1996; Varela et al., 1999;Turrigiano and Nelson, 2000; Schummers et al., 2002).Developmental studies in rat somatosensory cortex haveshown that the mature balance of excitation and inhibi-tion emerges during the first postnatal month (Luhmannand Prince, 1991). However, it is not known how differentcortical circuits acquire different balances of excitationand inhibition. Specifically, it is unclear whetherpathway-specific inhibition emerges through modificationof the excitatory input to interneurons or the inhibitoryoutput from these cells.

In adult rat visual cortex, feedforward (FF) and feed-back (FB) pathways that connect area 17 with the second-

ary lateromedial (LM) area provide �10% of their inputsto parvalbumin (PV)-expressing �-aminobutyric acidergic

Grant sponsor: National Institute of Health; Grant number: EY05936;Grant number: EY10214; Grant sponsor: McDonnell Foundation for Stud-ies of Higher Brain Function; Grant sponsor: Human Frontier ScienceProgram; Grant number: RC0123-2000B; Grant sponsor: Nihon UniversityFellowship (A.Y.).

*Correspondence to: Andreas Burkhalter, Department of Anatomy andNeurobiology, 8108, Washington University School of Medicine, 660 SouthEuclid Avenue, St. Louis, MO 63110. E-mail: [email protected]

Received 16 December 2002; Revised 6 February 2003; Accepted 30 April2003

DOI 10.1002/cne.10810Published online the week of August 11, 2003 in Wiley InterScience

(www.interscience.wiley.com).

THE JOURNAL OF COMPARATIVE NEUROLOGY 464:426–437 (2003)

© 2003 WILEY-LISS, INC.

(GABAergic) neurons and �90% to non-GABAergic cells(Gonchar and Burkhalter, 1999). Studies in the develop-ing rat and mouse somatosensory cortex have demon-strated that the ratio of asymmetric (putative excitatory)and symmetric GABAergic synapses is mature at P16(Micheva and Beaulieu, 1996; DeFelipe et al., 1997). How-ever, the proportions of FF and FB connections to PV andnon-PV neurons may differ from those in the cortical neu-ropil, and the subcellular distribution of inputs maychange across development.

Studies in rodent sensory cortex have shown that thegeneration of both asymmetric and symmetric synapsescontinues until the end of the fifth postnatal week (Blueand Parnavelas, 1983b; Micheva and Beaulieu, 1996; De-Felipe et al., 1997; White et al., 1997). This proliferativephase is accompanied by synapse elimination, which af-fects predominantly asymmetric contacts, with much lesseffect on GABAergic synapses (Micheva and Beaulieu,1996; DeFelipe et al., 1997; but see Blue and Parnavelas,1983b, for results obtained without identification ofGABAergic and non-GABAergic synapses). Thus, the re-modeling of FF and FB circuits may largely spareGABAergic outputs of PV neurons and affect mainly thenumber and subcellular distribution of asymmetric syn-apses on PV neurons. It is conceivable that these changesinfluence the balance of excitation and inhibition in FFand FB circuits.

Electron microscopic studies in adult rat visual cortexhave shown that FF axons preferentially terminate onthick PV dendrites, whereas FB connections favor thin PVdendrites (Gonchar and Burkhalter, 1999). This organiza-tion may provide differential pathway-specific electrotonicfiltering of synaptic inputs (Williams and Stuart, 2002)and make FF inputs to PV neurons more effective than FBinputs. As a result, primary excitatory FF responses maybe more strongly opposed by disynaptic inhibition thanresponses to FB inputs (Shao and Burkhalter, 1996,1999). Recordings in slices of developing rat visual cortexhave shown that pathway-specific inhibition in layer 2/3emerges after eye opening during the third to fourth post-natal weeks. Pathway-specific changes in the subcellulardistributions of inputs to inhibitory neurons may underliethese modifications. We therefore examined whether thedistributions of FF and FB synapses on PV dendrites inlayer 2/3 of mouse visual cortex are differentially alteredbetween eye opening and the end of the critical period(Gordon and Stryker, 1996).

The results show that the ratio of FF and FB inputs toPV and non-PV targets remains constant after P15. Sim-ilarly, the diameters of PV dendrites contacted by FF andFB synapses were similar at P15. However, during thethird to sixth postnatal weeks, the distributions ofpostsynaptic PV targets changed in pathway-specific fash-ion, so that mature FF inputs preferred thick PV den-drites and FB synapses favored thin ones.

MATERIALS AND METHODS

Experiments were performed on nine P14–16 and nineP36–42 C57Bl/6 mice. Animal housing and experimentalprocedures were approved by the Animal Studies Commit-tee and were in compliance with NIH guidelines.

Immunostaining for light microscopy

To study the expression of PV in neuronal cell bodiesand dendrites in the developing visual cortex, we stainedsections of three P16 and three P36–42 brains with anti-bodies against PV. Animals were deeply anesthetized withketamine (86 mg/kg, i.p.) and xylazine (13 mg/kg, i.p),injected intracardially with sodium nitrite (1%, 0.1 ml)and heparin (100 U/ml, 0.1 ml), and perfused through theleft ventricle with phosphate-buffered saline (PBS; pH7.4), followed by a mixture of 4% paraformaldehyde and0.1% glutaraldehyde in 0.1 M phosphate buffer (PB; pH7.4, 4°C). Brains were postfixed for 1–2 hours in the samefixative, equilibrated in 30% sucrose, and sectioned at 40�m on a freezing microtome in the coronal plane. Sectionswere treated sequentially for 30 minutes in sodium boro-hydride (1%), ethanol (50%), H2O2 (0.03%), and 10% fishgelatin (FG) and subsequently incubated in mouseanti-PV (1:5,000; Swant, Bellinzona, Switzerland; 18–36hours, 4°C). Sections were then washed in 2% FG andtreated in biotinylated horse anti-mouse IgG (2 hours,20°C), followed by avidin-biotinylated-horseradish perox-idase (HRP; Vectastain ABC Elite Kit; Vector, Burlin-game, CA), and processed for HRP reactivity with diami-nobenzidine (DAB; 0.05%) and H2O2 (0.005%). Stainedsections were mounted on glass slides, dehydrated in eth-anol, cleared in xylenes, rehydrated, and treated withAgNO3 (1.4%, 60°C) and HAuCl4 (0.2%; Jiang et al., 1993)to intensify the DAB reaction product. Selected sectionswere counterstained with cresyl violet and coverslippedwith DPX mounting medium.

Tracing of interareal connections

Neuronal connections between primary visual cortex(area 17) and the higher order visual area, LM, wererevealed by anterograde tracing with biocytin [Sigma, St.Louis, MO; 2.5% in Tris-buffered saline, pH 7.4 (TBS)] orwith biotinylated dextran amine (Molecular Probes, Eu-gene, OR; BDA-3,000 or BDA-10,000, 10% in 0.01 M phos-phate buffer, pH 7.4). For this purpose, six P14 mice wereanesthetized by inhalation of halothane (1% in air) andimmobilized by cementing the skull to a holder. The sixolder (P36–42) animals were anesthetized with ketamine(86 mg/kg, i.p.) and xylazine (13 mg/kg, i.p.) and secured ina head holder. Tracer injections into areas 17 and LMwere made by iontophoresis (5 �A positive current, 7seconds on/7 seconds off, 10–30 minutes) using pipetspulled from glass capillaries (1.2-mm outer diameter,0.6-mm inner diameter; F. Hear, New Brunswick, ME)with tip diameters of �15 �m. After a survival time of10–24 hours (pilot experiments have shown that rela-tively short survival times minimize retrograde labelingwithout affecting anterograde labeling), animals were per-fused through the heart with heparinized PBS, followedby a mixture of 4% paraformaldehyde and 0.5% glutaral-dehyde in PB at 4°C. Brains were postfixed for 2–4 hoursin the same fixative and stored overnight in PB at 4°C.Visual cortex was sectioned on a Vibratome at 40 �m inthe coronal plane.

For light microscopy, every fifth section was stained forbiocytin or BDA. Sections were pretreated in sodium boro-hydride (1%), H2O2 (0.03%), and 50% ethanol, followed byovernight incubation in avidin-biotinylated HRP complex.The HRP reaction was performed with 0.05% DAB and0.05% H2O2. Stained sections were mounted on glass

427INTERAREAL FEEDFORWARD AND FEEDBACK CONNECTIONS

slides, dehydrated in ethanol, cleared in xylenes, and cov-erslipped. To reveal layers and the area 17/LM border,selected sections were counterstained with cresyl violet.

Electron microscopy: double-labeling ofpathways and PV neurons

To study synaptic connections with PV neurons, FF andFB pathways between areas 17 and LM were traced byaxonal transport of biocytin or BDA, and sections thatcontained labeled axons were stained with antibodiesagainst PV. Biocytin or BDA was visualized with tetra-methylbenzidine (TMB), and PV immunoreactivity wasrevealed with DAB (Gonchar and Burkhalter, 1999). Forthis purpose, sections were incubated overnight (4°C) inavidin-biotinylated-HRP complex (Vectastain ABC EliteKit) and reacted in a phosphate-buffered solution (pH 6.2)containing TMB (0.005%), ammonium paratungstate(0.05%), NH4Cl (0.01%), and H2O2 (0.005%). The TMBreaction product was stabilized with DAB (0.05%), H2O2(0.005%), CoCl2 (0.002%), and NH4Cl (0.004%). To inacti-vate biocytin or BDA in anterogradely labeled fibers, sec-tions were incubated in avidin/biotin blocking reagent(Vector). This step was followed by sequential treatmentin ethanol (50%), FG (10%), normal goat serum (NGS; 1%PB, pH 7.4), rabbit anti-PV antibody (Sigma; 1:2,000, 48hours, 4°C), and biotinylated goat anti-rabbit secondaryantibody (1:200, 2 hours). After treatment with avidin-biotinylated HRP complex, sections were reacted withDAB (0.05%) and H2O2 (0.005%). Sections were addition-ally fixed in OsO4 (1%), dehydrated in increasing concen-trations of ethanol and propylene oxide, and embedded inDurcupan (Fluka, Ronkonkoma, NY). The sections weremounted between Teflon films, flattened between glassslides, and polymerized at 60°C for 48 hours.

Embedded sections were studied in the light micro-scope, and small regions of layer 2/3 of areas 17 and LMthat contained the core of FF and FB projections wereexcised and resectioned at 50 nm using an ultramicrotome(Ultracut E; Reichert). Serial thin sections were mountedonto 200-mesh copper grids and were counterstained withuranyl acetate (1%) and Reynold’s lead solution. Ultra-structural analyses were performed in a Jeol-100 electronmicroscope. To control for nonspecific staining by the sec-ondary antibody, incubation in primary antibody wasomitted.

Analysis

Proportion and density of PV neurons. The develop-ment of PV expression in cell bodies was analyzed quan-titatively by computer-assisted stereology using C.A.S.T.-Grid software (Olympus, Albertslund, Denmark). Todetermine the proportion of PV neurons, immunolabeledcells were counted in each layer in 1-mm-wide verticalstrips across the thickness of area 17. Sampling was per-formed in sections taken at 120-�m intervals from a ran-domly selected reference section for a maximum of eightsections per animal or a minimum of 500 PV neuronsacross three animals per age group. A similar strategywas used to determine cell density, except that cell countsin different layers were performed in fixed areas (0.1 � 1mm, long axis parallel to the pial surface). PV-stainedneurons were viewed with a 100� oil-immersion objective.Only PV-stained neurons whose nuclei were contained inthe section were counted. Counting was carried out byusing the optical disector. Sampling of different layers was

systematic (set sampling interval), uniform (mean of sam-ple equals mean of population), and random (each neuronin the population has the same probability of being sam-pled). The results were expressed as proportions of PVneurons per layer and PV neurons/mm3.

Diameter of PV dendrites. Dendritic trees of PV-labeled layer 2/3 neurons (10 in each area 17 and LM ofthree P16 and three P36 animals) were imaged with aCCD camera (MagnaFire; Optronics, Goleta, CA) at1,250� and graphically reconstructed on a video monitor.The drawings were scanned into a computer and digitallyenlarged. The diameter of dendrites was measured at2-�m intervals from the soma with NIH Image software.For quantitative observations of dendritic branching, aSholl (1955) analysis was performed by counting the num-ber of intersections of dendritic processes with a series ofconcentric circles spaced at 10-�m intervals. The center ofthe circles was placed in the center of the soma.

Percentage of labeled synapses. Conventional crite-ria were used to classify symmetric and asymmetric syn-apses (Gray, 1959; Colonnier, 1968). Anterogradely la-beled FF and FB terminals were identified by the presenceof TMB crystals (Gonchar and Burkhalter, 1999). PV-immunoreactive profiles were distinguished by the con-tent of amorphous DAB reaction product. The physicaldisector method (Sterio, 1984) was used to obtain an un-biased estimate of the percentage of labeled FF and FBsynapses onto PV neurons. To obtain synaptic counts,identical regions of neuropil were photographed at10,000� in two consecutive sections and enlarged to20,000�. Synapses were counted if they were present inthe reference section but not in a serial look-up section.Counts were obtained from three FF and three FB con-nections labeled in three animals of each of the P16 andthe P36–42 age groups. The data from the same animalwere considered independent observations.

Size of dendritic and axon terminal profiles. Thesize of dendritic profiles was determined using photo-graphic enlargements of electron micrographs. Dendriteswere identified using conventional criteria (Peters et al.,1991). The diameter (D) of dendrites at synaptic contactswas determined by measuring the cross-sectional area (A)of the profile and substitution of A in the equation D �2�(A/�). When dendrites were cut parallel to the longitu-dinal axis, D was the minimal width perpendicular to theaxis. Because synaptic boutons have approximately spher-ical shapes, their size was determined by measuring thearea of axon terminal profiles.

Mitochondria. The content of mitochondria in presyn-aptic boutons and postsynaptic dendrites was determinedby counting mitochondrial profiles in cross sections of FFand FB axon terminals and PV dendrites.

Statistical analysis

The t-test was used for comparison of sample averages.Statistical comparisons of size distributions of postsynap-tic PV target profiles were made with the z-test. ANOVAwas used to compare the variance of the diameter of PVdendrites in different age groups. Throughout the study,the level of statistical significance was set at P � .05.

Preparation of figures

Electron photomicrographs were printed from negativesand scanned at 1,200 dpi resolution with a flatbed scanner(U-Max Power Look III). Adobe Photoshop 7.0 was used to

428 A. YAMASHITA ET AL.

make linear adjustments of brightness and contrast. Finalfigures were printed on a Fujix Pictography 3000 printer(Fuji Photo Film USA, Elmsford, NY).

RESULTS

Development of PV expression

In P16 and P36 animals, layers 2–6 of areas 17 con-tained numerous PV-immunoreactive somata, dendrites,

and axon terminals. Invariably, PV-positive cells had non-pyramidal morphologies and aspinous dendrites (Fig.1A,B). Axon terminals often formed baskets around un-stained cell bodies that resembled somata of pyramidalneurons. The distribution of PV neurons across layers wasnonuniform. In both age groups, the percentage of PVneurons contained in a 1-mm-wide vertical strip acrossthe thickness of cortex was highest in layer 5 (Table 1).The percentages of PV neurons in layers 2/3, 4, and 6 were

Fig. 1. Morphology of PV neurons and connections between visualcortical areas 17 and LM of P15 mouse. A: Dendritic tree of PV-immunoreactive neuron in layer 2/3 of area 17. PV staining extendsinto thin, higher order branches (arrows). Dendrites are aspinous, andmany of the initial segments are tapered. PV-labeled axon terminalsform pericellular baskets (arrowheads). B: PV neuron in layer 4 ofarea LM. Most of the dendritic arbor is contained in layer 4. Occa-sionally, branches invade layer 2/3 (top arrow). Most ascending den-drites turn back at the layer 3/4 border and branch in layer 4 (bottomarrows). C: Darkfield image of coronal section showing BDA-labeled

FF connections from area 17 to the higher visual area LM. Labeledfibers and boutons are dense in layer 2/3, layer 4, and the bottom halfof layer 5. Connections to layers 1 and 6 are sparse. Arrow markslateral border of area 17. Asterisk marks the center of the injectionsite. D: Darkfield image of coronal section showing BDA-labeled FBconnections from area LM to area 17. Labeled fibers and boutons aredense in layers 1, 2/3, 5, and 6 and sparse in layer 4. Arrow marks17/LM border. Lateral (L), dorsal (D). Scale bars � 10 �m in A,B, 1mm in C,D.


smaller, and no such neurons were found in layer 1 (Table1). At P16 and P36–42, the densities and percentages ofPV neurons across layers were similar (Table 1).

At P16, dendrites of PV neurons in layers 4 and 5 ofareas 17 and LM were intensely labeled, and immuno-staining spread throughout the entire dendritic tree. Al-though dendritic labeling was weaker in layer 2/3, in mostcases PV staining was observed in thin, higher orderbranches (Fig. 1A). Most dendrites in layer 2/3 of area 17originated from somata in the same layer. Invasion oflayer 2/3 by dendrites of PV neurons in layers 4–6 wasrare, but occasional branches of layer 4 and layer 5 neu-rons crossed the layer 3/4 border (Fig. 1B). Layer 4 cellscontributed to �5% of PV dendrites in layer 2/3. Most PVdendrites from layer 4 abruptly changed course at thelayer 3/4 border and branched in layer 4 (Fig. 1B). Theincidence of ascending dendrites from layers 4 and 5 wasconstant from P16 to P36–42.

The morphology of PV dendrites in areas 17 and LM didnot change noticeably during postnatal maturation. ASholl analysis showed that at P16 and P36–42 the num-ber of intersections (7–14) of dendritic branches with con-centric circles centered in the soma peaked at �30 �m.Thus, at P16 PV dendrites have acquired their maturebranching pattern, and no major new branches are addedduring the following 3–4 weeks. The total length of thedendritic arbor was slightly greater at P16 (area 17 472 148 �m, n � 10; LM 494 173 �m, n � 9) than at P36–42(area 17 426 57 �m, n � 10; LM 454 92 �m, n � 10);however, the difference was not statistically significant.Likewise, the branch order was similar in P16 (area 174.6 1; LM 4.3 1) and P36–42 (area 17 3.9 1; LM4.5 1) PV neurons.

In both age groups, many primary and secondary den-drites were much thicker at the origin than at the tip (Fig.2A,B). Most primary and secondary dendrites, however,were thin at the origin and showed little tapering towardthe distal tip. Similarly, tapering was minimal in higherorder dendrites. Both P16 and P36–42 PV dendrites wereoften slightly beaded, which increased the variance of themean diameter. The variance of the diameter in thin andthick dendrites remained constant across development.This suggests that varicosities are not developmentallyregulated. The distribution of thin (�0.5 �m in diameter),medium-sized (0.5–1 �m), and thick branches (1 �m)across the dendritic tree was similar in both age groupsand both areas. At P16 and P36–42, at all distances fromthe soma, most PV dendrites were thin, whereas medium-sized and thick dendrites were rare (Fig. 3). Thick den-drites were present only within �30 �m of the soma;medium-sized branches were encountered up to �80 �m

from the soma, but, distal to this point, all dendrites werethin (Fig. 3). Point-by-point comparisons of the percent-ages of thin, medium-sized, and thick PV dendrites at P16and P36–42 within and between areas 17 and LM re-vealed no significant differences (t-test). These resultssuggest that the morphology of PV dendrites remainsconstant between P16 and P36–42 and that, during de-velopment, thick PV dendrites do not slim down, and thindendrites do not become bulkier.

FF and FB connections

Tracer injections into the part of the occipital cortexthat was characterized by a thick granular layer wereconsidered to be in area 17 (Paxinos and Franklin, 2001).Area 17 injections produced strong projections to the lat-eral geniculate nucleus (LGN) and very sparse connec-tions to the lateral posterior thalamic nucleus (LP). Strongintracortical projections were found to at least four loca-

TABLE 1. Development of PV Expression in Visual Cortex1

Layer 1 Layer 2/3 Layer 4 Layer 5 Layer 6

Proportion across thickness of cortexP16 0 (0) 14.9 (129) 29.2 (252) 33.6 (290) 22.2 (192)P36 0 (0) 15.1 (110) 21.7 (158) 36.5 (266) 26.7 (195)

Numerical density 1000 PV neurons/mm3

P16 0 1.52 3.69 4.28 3.45P36 0 1.50 3.24 4.67 3.74

1The proportion of PV somata in different layers was determined by counting (seeMaterials and Methods) immunolabeled cells in 1-mm-wide vertical strips across thethickness of area 17 in P16 and P36 mice. The numerical density of PV somata wasdetermined in 0.1 � 1-mm areas by unbiased sampling using computer-assisted stere-ology. The total number of cells studied is indicated in parentheses.

Fig. 2. Relationship between diameter of PV dendrites and dis-tance from soma in P16 (A) and P36–42 (B) mouse visual cortex. Eachplot represents data from 10 cells from three P16 and three P36–42mice. Thick dendrites (1 �m in diameter) represent primary orsecondary branches that are found within �30 �m of the soma.Medium-sized dendrites (0.5–1 �m in diameter) are represented inbranches of all orders and are found within �80 �m of the soma. Thindendrites (�0.5 �m in diameter) are found in primary, secondary,tertiary, and quaternary dendrites at any location between the somaand the distal tip. Proximal-to-distal tapering is most prominent inprimary and secondary dendrites. There are no significant (t-test)differences in the mean diameters at comparable locations betweenP16 and P36–42 dendrites. Diameters of P16 and P36–42 dendritesshow similar variances (ANOVA).


tions within secondary visual cortex lateral to area 17(V2L; Paxinos and Franklin, 2001), to temporal associa-tion and postrhinal cortex (Burwell and Amaral, 1998)and to medial secondary visual cortex and retrosplenialcortex. The intracortical connections resembled those pre-viously described by axonal tracing with 3H-proline (Ola-varria and Montero, 1989). The second projection fieldfrom the posterior pole, adjoining the lateral border ofarea 17, was considered, based on the study by Olavarriaand Montero (1989), to be the lateral medial area LM.Only cases in which retrograde labeling was minimal (�10labeled neurons per 10 consecutive 40-�m section) wereused for analysis. The pathway from area 17 terminatedin LM, in which bouton density was highest in layer 2/3,slightly lower in layer 4 and deep layer 5, and lowest inlayer 1, superficial layer 5, and layer 6 (Fig. 1C). Area LMinjections produced sparse terminal labeling in the LGNand revealed dense projections to the LP nucleus. LMprojections to area 17 involved mainly layers 1, 2/3, 5, and6 (Fig. 1D). Terminations in layer 4 were extremelysparse. Thus, each arm of the reciprocal 17-LM connectionshowed a distinct laminar organization.

Synaptic connections with PV neurons

Electron microscopic analyses of FF and FB axon ter-minals in P15 visual cortex revealed numerous synaptic

inputs to layer 2/3 PV neurons, which all contained roundvesicles. Some of these synapses showed thin postsynapticdensities and narrow synaptic clefts (Fig. 4A,B) and ap-peared less mature (Miller and Peters, 1981; Blue andParnavelas, 1983a) than those found at P36–42, in whichclefts were wider and postsynaptic densities were thicker(Figs. 4C,D, 5A–C). Occasionally, dense postsynaptic PVstaining tended to obscure the distinction between sym-metric and asymmetric synapses (Fig. 4D). Importantly,however, FF and FB boutons were always PV negative(Figs. 4A–C, 5A–D), which indicates that FF and FB syn-apses are not GABAergic and supports the notion thatthey originate from pyramidal neurons in areas 17 andLM, respectively.

Of the FF and FB connections observed at P15, 7.7%(45/642) of FF and 7.8% (58/743) of FB axon terminalsformed synapses onto PV neurons. At P36–42, the per-centage of inputs to PV neurons was slightly higher in theFF (9.7%, 54/555) than in the FB (7.1%, 48/676) pathway,but statistical testing revealed no significant age andpathway differences. Of all asymmetric synapses found inthe neuropil of areas LM and 17 of P15 visual cortex, FFconnections accounted for 6% (7/116) and FB inputs for4.3% (7/161), respectively. These proportions were slightlyhigher at P36–42 (FF 6.4%, 5/78; FB 6.4%, 9/136), but thedevelopmental changes were not statistically significant.

Qualitative observations revealed striking developmen-tal changes in the distribution of FF and FB synapses onPV dendrites. At P15, FF axon terminals in layer 2/3 ofarea LM were frequently found on thick, medium-sized,and thin PV dendrites (Fig. 4A,B). At P36–42, in contrast,FF inputs onto thin dendrites were rare, and most syn-apses were on medium-sized and thick dendrites (Fig.4C,D). In many cases, synapses onto thick dendrites werelocated proximally (Fig. 4D,D1). Similar to the distribu-tion in the FF pathway, FB axon terminals in P15 cortexformed synapses onto thick, medium-sized, and thin layer2/3 PV dendrites (Fig. 5A). However, unlike FF connec-tions, FB synapses in P36–42 mice strongly preferred thinPV dendrites (Fig. 5B,C).

Quantitative analyses of the diameter of PV dendriteswith synaptic inputs from FF and FB connections con-firmed our qualitative impression that inputs to PVneurons are rearranged during postnatal development.At P15, the size distributions of PV dendrites withsynaptic inputs from FF and FB axons were overlapping(Fig. 6A), and the mean diameters of FF- and FB-recipient dendrites were similar (Table 2). During mat-uration, the proportion of FF inputs to thin PV den-drites decreased significantly, whereas FB connectionsto thin PV dendrites increased and FB inputs to thickdendrites were lost (Fig. 6A). These opposing pathway-specific developmental trends were also reflected in asignificant decrease in the mean diameter of PV den-drites contacted by FB synapses (Table 2) and a ten-dency (missing statistical significance) of FF synapsesto contact thicker PV dendrites (Table 2).

The analysis of PV dendrites in areas 17 and LMshowed that thick, medium-sized, and thin dendrites havedistinct but overlapping proximal/distal distributions(Fig. 3). To study the developmental changes of inputs todifferent size categories of dendrites, we determined thepercentages of FF and FB synapses onto thin, medium-sized, and thick PV dendrites. At P15, the overall sizedistribution of PV-immunoreactive dendritic targets was

Fig. 3. Radial distribution of the percentages of thick, medium-sized, and thin dendrites in PV neurons at P16 and P36–42 in areas17 (A,B) and LM (C,D) of mouse visual cortex. Mean percentages of 10PV neurons from three animals of each age group. Most dendrites arethin (�0.5 �m in diameter). Even at proximal locations, thin den-drites account for at least 50% of all PV branches. Thick dendrites arerare at all locations and are absent distal to �30 �m of the soma.Medium-sized dendrites are absent distal to �80 �m from the soma.There are no significant (t-test) differences in the mean percentages ofP16 and P36–42 thick, medium-sized, and thin PV dendrites at anydistance from the soma.


Fig. 4. Electron micrographs of feedforward (FF) synapses ontoPV dendrites in layer 2/3 of area LM in developing mouse visualcortex. At P15, BDA-labeled (dark TMB crystals) FF axon terminalsfilled with round vesicles form asymmetric synapses (arrowheads)onto thin (A) and thick (B) PV-immunoreactive dendrites (dark, dif-fuse DAB reaction product). Although synapses are clearly asymmet-rical, the postsynaptic densities are immature and less conspicuous

than at P36 (C). At P36, BDA-labeled FF axons form asymmetricalsynapses (arrowheads) on thick (C,C1,D) PV dendrites, which areoften found on primary proximal PV trunks (D,D1). In the synapseshown in D, the postsynaptic density is partially obscured by densePV immunoperoxidase staining. Images in C and D are higher mag-nifications of the boxed areas shown in C1 and D1. Scale bars � 0.25�m in A,C,D, 0.35 �m in B, 1 �m in C1, 2 �m in D1.


similar in the FF and FB pathways (Fig. 6B). However,FB inputs showed a slight bias for thin dendrites thatwas not statistically significant. During maturation, theproportion of FF inputs to medium-sized and thick PVdendrites increased and the percentage of synapses ontothin PV branches decreased by �24% (z-test, P � .014;Fig. 6B). In the FB pathway, the developmental changespointed in the opposite direction; by P36– 42, the pro-portion of synaptic inputs to thick PV profiles was �26%larger (z-test, P � .01), and inputs to thick dendriteswere eliminated. These developmental changes werepathway specific and were not observed in the distribu-tion of unlabeled, unidentified asymmetric synapses(not FF or FB synapses) onto PV dendrites in the neu-ropil of areas 17 and LM (Table 2).

Developmental changes in FF and FB axonterminals

To study developmental changes in presynaptic inputs,we measured the size of FF and FB boutons. Within eachage group, the mean cross-sectional areas of FF and FBterminals that formed synaptic contacts with PV den-drites were similar (Table 3). However, the comparison ofP15 and P36–42 mice showed a significant developmentalreduction of bouton size in both pathways (Table 3). Asimilar reduction was found in the size of FF and FBboutons on PV-negative dendrites and of unidentified (un-labeled) terminals on PV dendrites (Table 3). Thus, thereduction in bouton size was neither pathway specific nortarget specific.

Fig. 5. Electron micrograph of feedback (FB) synapses onto PVdendrites in layer 2/3 of area 17 of developing mouse visual cortex. AtP15, BDA-labeled (dark TMB crystals) FB axon terminals filled withround vesicles form asymmetric synapses on thin (PV1) and thick(PV2) PV-immunoreactive dendrites (A). The postsynaptic density ofthe synapse on the thin dendrite is obscured by the dark PV immu-

nostaining of the dendrite. At P36, BDA-labeled FB axons form asym-metric synapses on thin PV dendrites (B,C). In addition to FB input(arrowheads), PV1 receives input from a PV-immunoreactive termi-nal (PV2) that contains flat vesicles and forms a symmetric synapse.Scale bars � 0.5 �m in A, 0.25 �m in B,C.


Qualitative observations in adult rat visual cortex sug-gested that FF boutons on PV dendrites contain moremitochondrial profiles than FB terminals (Gonchar and

Burkhalter, 1999). To determine whether this differenceemerges during postnatal development, we counted mito-chondrial profiles. At P15, the average numbers of profilesin FF and FB boutons on PV-positive dendrites were sim-ilar (Fig. 7A). During development, the number of mito-chondrial profiles in FF boutons increased. By P36–42, FFaxon terminals contained significantly more mitochondriathan at P15 and differed from P36–42 FB terminals,whose size remained constant between P15 and P36–42(Fig. 7A). Similar changes were observed in FF terminalson PV-negative dendrites (Fig. 7B), indicating that, dur-ing development, the number of mitochondrial profilesincreased in all FF terminals. Interestingly, however, mi-tochondrial profiles in unidentified asymmetric terminalson PV-positive and PV-negative dendrites in the sur-rounding neuropil remained constant after P15 (Fig.7C,D). Previous studies of cat geniculocortical boutonshave shown that the number of mitochondrial profiles isscaled linearly to the bouton size (Friedlander et al.,1991). Thus, because the size of FF and FB boutons re-mained constant across development (Table 3), our resultssuggest that in FF boutons the density or complexity ofmitochondria is selectively increased during maturation.In contrast, the content and complexity of mitochondria inFB terminals remained unchanged across development.

DISCUSSION

This study shows that in FF and FB pathways betweenareas 17 and LM of mouse visual cortex, inputs to PVneurons are remodeled during the critical period of post-natal development (Gordon and Stryker, 1996). The alter-ations involve the redistribution of FF and FB inputs toPV dendrites from a similar immature pattern to maturepathway-specific patterns in which FF connections favorthick and FB inputs prefer thin PV dendrites, similar tothose observed in adult rat visual cortex (Gonchar andBurkhalter, 1999). The results show that, unlike the de-velopmental remodeling of the subcellular distribution ofinputs, the ratio of FF and FB connections to PV andnon-PV neurons remains constant across development.

Areal hierarchy in mouse visual cortex

Studies in primate visual cortex have used the distinc-tive laminar projection patterns of FF and FB connectionsbetween different areas to derive an areal hierarchy (Fel-leman and Van Essen, 1991). The same approach hasrevealed that areas of rat visual cortex are hierarchically

Fig. 6. Size distribution of PV-immunoreactive dendritic profiles re-ceiving synaptic inputs from FF connections that link area 17 with areaLM (open bars) and from FB connections (solid bars) that project fromLM to area 17. A: At P15, the distributions of percentages of different-sized dendrites contacted by FF and FB synapses are similar. At P36–42, few FF synapses contact thin dendrites, whereas FB synapses arerarely found on thick dendrites. The total number of synapses analyzedis indicated in parentheses. B: Distribution of FF and FB synapses onthin (�0.5 �m in diameter), medium-sized (0.5 �1 �m in diameter),and thick (1 �m in diameter) profiles of PV dendrites. At P15, the sizedistributions of PV-immunoreactive dendrites contacted by FF and FBaxons are similar. At P36–42, the distributions in both pathways aresignificantly different (z-test). During postnatal weeks 3–5, the propor-tions of FF synapses on medium-sized and thick dendritic profiles sig-nificantly increase, and the proportion of contacts with thin dendritesdecreases. In contrast, during the same period, FB connections develop asignificantly stronger preference for thin dendrites and retract all inputsto thick dendrites.

TABLE 2. Developmental Changes in Mean Diameter of Postsynaptic PV Dendrites1

P15 P36–42

FF FB FF FB

FF and FB to PV� 0.69 0.99 (45) 0.64 1.24 (58) 0.8 0.86 (55) 0.43 0.23 (58)� P � 0.24 NS � � P � .01 �

� P � .09 NS �� P � .015 �

LM 17 LM 17

Non-FF Non-FB to PV� 0.74 0.33 (116) 0.68 0.62 (161) 0.77 0.55 (78) 0.64 0.37 (136)

1Mean diameter (�m SD) of PV� dendritic profiles contacted by feedforward (FF) and feedback (FB) synapses in layer 2/3 of areas 17 and LM. At P15, the mean diameter ofPV dendrites contacted by FF and FB connections is similar (NS, not significantly different; t-test). At P36–42, the average diameter of PV dendrites with FB inputs is significantlysmaller. The developmental change is due to a reduction in the diameter of PV dendrites in the FB pathway. The diameters of PV dendrites contacted by asymmetrical synapses(non-FF, non-FB, i.e., excluding FF and FB synapses) in the neuropil of areas 17 and LM are similar and remain constant across development. The total number of cells studiedis indicated in parentheses.


organized (Coogan and Burkhalter, 1993). Connectivityand functional mapping studies suggest that mouse visualcortex contains more than a single area (Wagor at al.,1980; Olavarria et al., 1982; Olavarria and Montero, 1989;Schuett et al., 2002; Kalatsky and Stryker, 2003). We have

found that two of these areas, areas 17 and LM, are linkedby pathways whose laminar organizations resemble FFand FB connections of rat visual cortex (Coogan andBurkhalter, 1990). This suggests that areas 17 and LM ofmouse visual cortex are part of a hierarchy in which LMranks above area 17.

Maturation of PV neurons and remodelingof inputs to these GABAergic cells

We have shown previously that the synaptic effects ofexcitatory FF and FB inputs on pyramidal cells areshaped by disynaptic inhibition from GABAergic, PV-expressing neurons (Shao and Burkhalter, 1996; Goncharand Burkhalter, 1999). In adult rats, these neurons re-ceive 10–12% of FF and FB inputs (Gonchar and Burkhal-ter, 1999). Here, we show that PV neurons are a majorGABAergic target of FF and FB connections in mousevisual cortex. In mouse, however, the proportions of FFand FB inputs to PV neurons are smaller (7–10%), whichsuggests that the ratio of interareal inputs to GABAergicand non-GABAergic cells is different. Alternatively, FFand FB connections in mouse visual cortex might be lessstrongly connected to PV neurons than in rat and maycontact additional GABAergic neurons that do not expressPV (Gonchar and Burkhalter, 1997).

Previous studies in the developing rat and mouse visualcortex have shown that PV is first expressed at �P11 andthat the proportion of PV neurons in layer 2/3 peaks atP21 (Alcantara et al., 1993; Del Rio et al., 1994). Ourunbiased quantitative analysis, however, shows that PVexpression develops more rapidly and that the laminardistribution and numerical density of PV neurons at P16resembles that at P36–42. The results further show that,by P15, the proportions of FF and FB inputs to PV neuronsare similar to those at P36–42, which suggests that theratio of FF and FB inputs to pyramidal cells and PV-expressing inhibitory neurons is mature prior to the end ofcortical synaptogenesis (Blue and Parnavelas, 1983b;Micheva and Beaulieu, 1996; DeFelipe et al., 1997), whichis well before FF and FB inputs are fully developed (Q.Wang et al., 2002). Moreover, the developmental con-stancy of the innervation ratio of PV and non-PV neuronsby FF and FB axons implies that the selection of excita-

Fig. 7. Number of mitochondria in FF and FB connections be-tween areas 17 and LM in developing mouse visual cortex. The his-tograms represent the mean number of mitochondrial profiles in FF(open bars) and FB (solid bars) axon terminals that are synapticallyconnected to PV-positive (PV�) and PV-negative (PV–) dendrites inlayer 2/3 of areas 17 and LM. The number of mitochondrial profiles inFF terminals on PV� (A) and PV– (B) dendrites significantly (t-test)increases from P15 to P36–42. No such increase is found in FBterminals on PV� and PV– dendrites. In the neuropil of layer 2/3 ofarea 17 and LM, the number of mitochondrial profiles in asymmetricterminals (excluding FF and FB terminals) on PV� (C) and PV– (D)dendrites remains constant between P15 and P36–42.

TABLE 3. Developmental Changes in Size of Boutons1

P15 P36–42

FF FB FF FB

FF and FB to PV� 1.65 1.95 (25) 1.22 1.01 (58) 0.80 0.58 (25) 0.83 0.52 (38)� P � .17 NS � � P � .42 NS �

� P � .03 �� P � .02 �

FF and FB to PV 1.27 1.56 (51) 1.16 1.07 (13) 0.80 0.50 (70) 0.79 0.49 (91)� P � .45 NS � � P � .32 NS �

� P � .01 �� P � .02 �

LM 17 LM 17

Non-FF Non-FB to PV� 1.02 1.19 (116) 0.83 0.71 (161) 0.72 0.65 (78) 0.76 0.46 (136)� P � .07 NS � � P � .33 NS �

� P � .02 �� P � .02 �

1Mean area (�m2 SD) of feedforward (FF), feedback (FB), and non-FF and non-FB (i.e., asymmetric terminals, excluding FF and FB) axon terminals forming asymmetricsynapses on PV� and PV dendrites in layer 2/3 of areas 17 and LM. At P15, the sizes of FF, FB, and unidentified asymmetric terminals (non-FF, non-FB) on both PV� and PV

dendrites are similar. From P15 to P36–42, there is a significant (t-test) reduction in the size of FF, FB, and asymmetric terminals on both PV� and PV dendrites. The totalnumber of synapses studied is indicated in parentheses.


tory and PV-positive inhibitory postsynaptic targets isindependent of the total strengths of the connections.

Multidimensional analyses of PV dendrites suggest thatthe cells are nest basket cells (Y. Wang et al., 2002). Themorphology of their dendrites remains constant betweenP16 and P36–42, and there are no significant changes inthe complexity or the overall thickness of branches, theamount of radial tapering, the beadedness, and the pro-portions of thin, medium-sized, and thick dendrites acrossthe dendritic tree. This suggests that, similarly to the casefor Golgi-stained nonpyramidal neurons (Parnavelas andUylings, 1980; Miller, 1986), the structure of PV dendrites(revealed with antibodies against PV) in layer 2/3 of areas17 and LM is mature at P16. Thus, the opposing develop-mental trends of reducing FF inputs to thin PV dendritesand increasing FB inputs to thin dendrites are not due topathway-specific alterations in the diameter of postsynap-tic PV dendrites but reflect the rearrangement of presyn-aptic FF and FB inputs to different-sized dendrites.

Our analysis shows that PV dendrites are thinner attheir distal tips than at their base. Numerous thinbranches, however, represent primary dendrites that exitfrom the cell body, which makes it difficult to translatedendritic diameter into radial location. It therefore re-mains unresolved whether the developmental changes inthe size selection of PV dendrites reflect a redistribution ofFF and FB inputs along the proximal/distal axis or repre-sent changes in the preference of thin or thick dendrites.

At P15, FF connections in layer 2/3 of area LM aredenser than FB connections to area 17 (Q. Wang et al.,2002). Recordings in pyramidal neurons have shown that,at this stage of development, paradoxically, disynapticinhibition in the FB pathway is similar to that in the FFpathway (Shao and Burkhalter, 1998; Dong et al., 2000).This immature pattern of inhibition may be due to strongFB inputs to thick proximal dendrites, which in pyramidalcells were shown to be more effective than inputs to distaldendrites (Williams and Stuart, 2002). The highly effec-tive inputs to thick PV dendrites may compensate for theoverall sparse FB connections at P15 (Q. Wang et al.,2002). After P15, the density of FF and FB connectionsincreases (Q. Wang et al., 2002). In parallel, inhibition inthe FF pathway gains strength and inhibition in the FBpathway remains constant (Shao and Burkhalter, 1998;Dong et al., 2000). This physiological transformation ofthe FF circuit is accompanied by a reduction of inputs tothin dendrites and an increase in the relative strength ofinputs to medium-sized and thick PV dendrites. Thus, thechanges in the subcellular targeting of PV dendrites sug-gest that, during maturation, the effectiveness of FF andFB inputs to inhibitory neurons is continuously scaled tothe overall strength of the connections.

In the FF pathway, the developmental remodeling of theproportion of inputs to thin, medium-sized, and thick PVdendrites is consistent with the selective addition of syn-apses, but it remains unclear whether synapses are elim-inated from thin dendrites. In the FB pathway, by con-trast, the complete absence of inputs to thick PV dendritesat P36–42 suggests that, during maturation, FB synapsesare eliminated from thick branches. The reduction in theproportion of the most effective inputs (Williams and Stu-art, 2002) to PV neurons may compensate for the in-creased overall anatomical strength of FB connections (Q.Wang et al., 2002) and may keep inhibition in the FBcircuit constant across development. Elimination of asym-

metric synapses on interneurons has previously been ob-served in rat cerebellum and cerebral cortex (Crepel et al.,1976; Mariani and Changeux, 1981; Bahr and Wolff,1984).

Evidence for a role of activity

The pathway-specific developmental regulation of thenumber or complexity of mitochondria (Anderson et al.,1998) suggests that activity plays a role in shaping dis-tinct innervation patterns of PV neurons by FF and FBaxons. The coincidence of these changes with eye openingsuggests that they are linked to the increased metabolicdemand on FF-projecting neurons responding to visualinputs (Wong-Riley et al., 1989). Activity in the FB path-way might be less tightly linked to retinal inputs andmight be weaker (Maunsell et al., 1991; Mignard andMalpeli, 1991). Thus, the probability of coincident activa-tion of presynaptic inputs and postsynaptic responses inPV neurons may be lower in the FB pathway, and, duringdevelopment, FB inputs may loose the competition for thelargest dendrites in which the action potential firingthreshold is lowest (Stuart et al., 1997).

Functional relevance

The purpose of the circuit transformations describedhere might be to endow the mature FF pathway withstrong and the FB pathway with weak disynaptic inhibi-tion. In the FF pathway, this may increase the activationthreshold of the initial excitation, lower the gain of recur-rent intracortical excitation, and allow extraction of theoptimal signal of visual inputs. In contrast, weak inhibi-tion in the FB pathway may increase the gain of recurrentexcitation, and coincident activation of top-down path-ways (e.g., FB) with FF inputs may amplify visual re-sponses to low-salience stimuli (Hupe et al., 1998).

ACKNOWLEDGMENTS

We thank Mazen Kheirbek for excellent technical assis-tance, Dr. Hongwei Dong for advice with statistical anal-yses, and Dr. Quanxin Wang for helpful discussions. Wealso thank Dr. J.L. Price for generous access to the stere-ology microscope.

LITERATURE CITED

Alcantara S, Ferrer I, Soriano E. 1993. Postnatal development of parval-bumin and calbindin D28 immunoreactivities in the cerebral cortex ofthe rat. Anat Embryol 188:63–73.

Anderson JC, Binzegger T, Martin KAC, Rockland KS. 1998. The connec-tion from cortical area V1 to V5: a light and electron microscopic study.J Neurosci 18:10525–10540.

Bahr S, Wolff JR. 1985. Postnatal development of axosomatic synapses inthe rat visual cortex: morphogenesis and quantitative evaluation.J Comp Neurol 233:405–420.

Blue ME, Parnavelas JG. 1983a. The formation and maturation of syn-apses in the visual cortex of the rat. I. Qualitative analysis. J Neuro-cytol 12:599–616.

Blue ME, Parnavelas JG. 1983b. The formation and maturation of syn-apses in the visual cortex of the rat. II. Quantitative analysis. J Neu-rocytol 12:697–712.

Burwell RD, Amaral DG. 1998. Perirhinal and postrhinal cortices of therat: interconnectivity and connections with the entorhinal cortex.J Comp Neurol 391:293–321.

Colonnier M. 1968. Synaptic patterns of different cell types in the differentlaminae of he cat visual cortex: an electron microscopic study. BrainRes 9:268–287.


Coogan TA, Burkhalter A. 1990. Conserved patterns of cortico-corticalconnections define areal hierarchy in rat visual cortex. Exp Brain Res80:49–53.

Coogan TA, Burkhalter A. 1993. The hierarchical organization of rat visualcortex. J Neurosci 13:3749–3772.

Crepel F, Mariani J, Delhaye-Bouchard N. 1976. Evidence for a multipleinnervation of Purkinje cells by climbing fibers in the immature ratcerebellum. J Neurobiol 7:567–578.

DeFelipe J, Fairen A, Jones EG. 1997. Inhibitory synaptogenesis in mousesomatosensory cortex. Cereb Cortex 7:619–634.

DelRio JA, de Lecea L, Ferrer I, Soriano E. 1994. The development ofparvalbumin-immunoreactivity in the neocortex of the mouse. BrainRes Dev Brain Res 81:247–259.

Dong H, Huang ZJ, Tonegawa S, Burkhalter A. 2000. Regulation of inhi-bition in developing interareal forward and feedback circuits in visualcortex of BDNF overexpressing mice. Soc Neurosci Abstr 26:855.

Douglas RJ, Mahowald M, Martin KAC, Stratford KJ. 1996. The role ofsynapses in cortical computation. J Neurocytol 25:893–911.

Felleman DJ, Van Essen DC. 1991. Distributed hierarchical processing inthe primate cerebral cortex. Cereb Cortex 1:1–48.

Friedlander MJ, Martin KAC, Wassenhove-Carthy D. 1991. Effects ofmonocular visual deprivation on the innervation of area 18 in the cat byindividual thalamocortical axons. J Neurosci 11:3268–3288.

Gonchar Y, Burkhalter A. 1997. Three distinct families of GABAergicneurons in rat visual cortex. Cereb Cortex 7:347–358.

Gonchar Y, Burkhalter A. 1999. Differential subcellular localization offorward and feedback interareal inputs to parvalbumin expressingGABAergic neurons in rat visual cortex. J Comp Neurol 406:346–360.

Gordon JA, Stryker MP. 1996. Experience-dependent plasticity of binocu-lar responses in the primary visual cortex of the mouse. J Neurosci16:3247–3286.

Gray EG. 1959. Axosomatic and axodendritic synapses in the cerebralcortex: an electron microscopic study. J Anat 93:420–433.

Hupe JM, James AC, Payne BR, Lomber SG, Girard P, Bullier J. 1998.Cortical feedback improves discrimination between figure and back-ground by V1, V2 and V3 neurons. Nature 394:784–787

Jiang X-P, Johnson RR, Burkhalter A. 1993. Visualization of dendriticmorphology of cortical projection neurons by retrograde tracing. Neu-rosci Methods 50:45–60.

Kalatsky VA, Stryker MP. 2003. New paradigm for optical imaging: tem-porally encoded maps of intrinsic signal. Neuron 38:529–545.

Luhmann HJ, Prince DA. 1991. Postnatal maturation of the GABAergicsystem in rat neocortex. J Neurophysiol 65:247–263.

Mariani J, Changeux JP. 1981. Ontogenesis of olivocerebellar relation-ships. I. Studies by intracellular recordings of the multiple innervationof Purkinje cells by climbing fibers in the developing rat cerebellum.J Neurosci 1:696–702.

Maunsell JHR, Sclar G, Nealey TA, DePriest DD. 1991. Extraretinalrepresentations in area V4 in the macaque monkey. Vis Neurosci7:561–573.

Micheva KD, Beaulieu C. 1996. Quantitative aspects of synaptogenesis inthe rat barrel field cortex with special reference to GABA circuitry.J Comp Neurol 373:340–354.

Mignard M, Malpeli JG. 1991. Paths of information flow through visualcortex. Science 251:1249–1251.

Miller M, Peters A. 1981. Maturation of rat visual cortex: II. A combinedGolgi-electron microscope study of pyramidal neurons. J Comp Neurol203:555–573.

Miller MW. 1986. Maturation of rat visual cortex. III. Postnatal morpho-genesis and synaptogenesis of local circuit neurons. Brain Res DevBrain Res 25:271–285.

Olavarria J, Montero VM. 1989. Organization of visual cortex in the mouserevealed by correlating callosal and striate-extrastriate connections.Vis Neurosci 3:59–69.

Olavarria J, Mignano LR, Van Sluyters RC. 1982. Pattern of extrastriatevisual areas connecting reciprocally with striate cortex in the mouse.Exp Neurol 78:775–779.

Parnavelas JG, Uylings HBM. 1980. The growth of non-pyramidal neuronsin the cortex of the rat: a morphometric study. Brain Res 193:373–382.

Paxinos G, Franklin KBJ. 2001. The mouse brain in stereotaxic coordi-nates, 2nd ed. New York: Academic Press.

Peters A, Palay SL, Webster HF. 1991. The fine structures of the nervoussystem. New York: Oxford University Press.

Schuett S, Bonhoeffer T, Hubener M. 2002. Mapping retinotopic structurein mouse visual cortex with optical imaging. J Neurosci 22:6549–6559.

Schummers J, Marino J, Sur M. 2002. Synaptic integration by V1 neuronsdepends on location within the orientation map. Neuron 36:969–978.

Shao Z, Burkhalter A. 1996. Different balance of excitation and inhibitionin forward and feedback circuits of rat visual cortex. J Neurosci 15:7353–7365.

Shao Z, Burkhalter A. 1998. Development of synaptic inhibition in intra-cortical pathways of rat visual cortex. Soc Neurosci Abstr 24:1518.

Shao Z, Burkhalter A. 1999. Role of GABAB receptor-mediated inhibitionin reciprocal interareal pathways of rat visual cortex. J Neurophysiol81:1014–1024.

Sholl DA. 1955. The organization of the visual cortex in the cat. J Anat89:33–46.

Sterio DC. 1984. The unbiased estimation of number and sizes of arbitraryparticles using the dissector. J Microsc 134:127–136.

Stuart G, Sprutson N, Sakmann B, Hausser M. 1997. Action potentialinitiation and backpropagation in neurons of the mammalian CNS.Trends Neurosci 20:125–131.

Turrigiano GG, Nelson SB. 2000. Hebb and homeostasis in neuronal plas-ticity. Curr Opin Neurobiol 10:358–364.

Varela JA, Song S, Turrigiano GG, Nelson SB. 1999. Differential depres-sion at excitatory and inhibitory synapses in visual cortex. J Neurosci19:4293–4304.

Wagor E, Mangini NJ, Pearlman. 1980. Retinotopic organization of striateand extrastriate visual cortex in the mouse. J Comp Neurol 193:187–202.

Wang Q, Nerbonne JM, Burkhalter A. 2002. Postnatal development offeedforward and feedback connections between lower and higher areasof mouse visual cortex. Soc Neurosci Abstr 130.12.

Wang Y, Gupta A, Toledo-Rodriguez M, Wu CZ, Markram H. 2002. Ana-tomical, physiological, molecular and circuit properties of nest basketcells in the developing somatosensory cortex. Cereb Cortex 12:395–410.

White EL, Weinfeld L, Lev DL. 1997. A survey of morphogenesis during theearly postnatal period in PMBSF barrels of mouse SmI cortex withemphasis on barrel D4. Somatosens Motor Res 14:34–55.

Williams SR, Stuart G. 2002. Dependence of EPSP efficacy on synapselocation in neocortical pyramidal neurons. Science 295:1907–1910.

Wong-Riley MTT, Trusk TC, Tripathi SC, Hoppe DA. 1989. Effect of retinalimpulse blockage on cytochrome oxidase-rich zones in the macaquestriate cortex: II. Quantitative electron-microscopic (EM) analysis ofneuropil. Vis Neurosci 2:499–514.


rearrangement of synaptic connections with inhibitory neurons in developing mouse visual cortex

Documents