G Model ARTICLE IN PRESS - 大阪大学 cite this article in press as: Ikezoe, K., et al., Decorrelation of sensory-evoked neuronal responses in rat barrel cortex during postnatal

N

Dp

Ka

5b

c

a

ARRAA

KSWMPNPCS

1

d1dtamsoRditec

SO

(

0h

ARTICLE IN PRESSG ModelSR-3456; No. of Pages 9

Neuroscience Research xxx (2012) xxx–xxx

Contents lists available at SciVerse ScienceDirect

Neuroscience Research

jo u r n al hom ep age: www.elsev ier .com/ locate /neures

ecorrelation of sensory-evoked neuronal responses in rat barrel cortex duringostnatal development

oji Ikezoea,b,∗, Hiroshi Tamuraa,b, Fumitaka Kimurac, Ichiro Fujitaa,b,∗

Laboratory for Cognitive Neuroscience, Graduate School of Engineering Science and Graduate School of Frontier Biosciences, Osaka University, 1-3 Machikaneyama, Toyonaka, Osaka60-8531, JapanCenter for Information and Neural Networks, Osaka University, 1-3 Machikaneyama, Toyonaka, Osaka 560-8531, JapanDepartment of Molecular Neuroscience, Graduate School of Medicine, Osaka University, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan

r t i c l e i n f o

rticle history:eceived 8 May 2012eceived in revised form 22 May 2012ccepted 24 May 2012vailable online xxx

eywords:omatosensory cortex

a b s t r a c t

The ability to detect and discriminate sensory stimuli greatly improves with age. To better understandthe neural basis of perceptual development, we studied the postnatal development of sensory responsesin cortical neurons. Specifically, we analyzed neuronal responses to single-whisker deflections in the pos-teromedial barrel subfield (PMBSF) of the rat primary somatosensory cortex. Responses of PMBSF neuronsshowed a long onset latency and duration in the first postnatal week, but became fast and transient overthe next few weeks. Trial-by-trial variations of single neuron responses did not change systematicallywith age, whereas the covariation of responses across trials between neurons (noise correlation) was high

hiskerultiple single-unit recording

oolingoise correlationopulation codingross-correlogram

on postnatal day 5–6 (P5–6), and gradually decreased with age to near zero by P30–31. Computationalanalyses showed that pooled responses of multiple neurons became more reliable across stimulus trialswith age. The period over which these changes occurred corresponds to the period when rats develop afull set of exploratory whisking behavior. We suggest that reduced noise correlation across a populationof neurons, in addition to sharpening the temporal characteristics of single neuron responses, may helpimprove behavioral performance.

12 El
ilicon probes © 20
. Introduction

In humans and many animal species, the ability to detect andiscriminate sensory inputs greatly improves with age (Aslin et al.,981; Atkinson, 2002). This improvement is supported by theevelopment of peripheral sensory organs, and neurons/circuits inhe central nervous system. Spike responses of individual neuronsre immature in infant animals. As animals grow older and accu-ulate sensory experiences, the sensitivity and selectivity of the

ingle neuron responses for the temporal and spatial propertiesf the sensory inputs acquire mature characteristics (Walsh andomand, 1992; Knudsen, 2002; Kiorpes and Movshon, 2004). Thisevelopment accounts for some, but not all, aspects of behavioral

mprovement (Kiorpes and Movshon, 2004). To better understand

Please cite this article in press as: Ikezoe, K., et al., Decorrelation ofpostnatal development. Neurosci. Res. (2012), http://dx.doi.org/10.10

he neural basis of perceptual development, it is necessary toxpand the analyses to include an assessment of the developmentalhanges in the sensory coding in a neuronal ensemble.

∗ Corresponding author at: Laboratory for Cognitive Neuroscience, Graduatechool of Frontier Biosciences, Osaka University, 1-3 Machikaneyama, Toyonaka,saka 560-8531, Japan. Tel.: +81 6 6850 6511; fax: +81 6 6850 6511.

E-mail addresses: [email protected] (K. Ikezoe), [email protected]. Fujita).

168-0102/$ – see front matter © 2012 Elsevier Ireland Ltd and the Japan Neuroscience Sttp://dx.doi.org/10.1016/j.neures.2012.05.009

sevier Ireland Ltd and the Japan Neuroscience Society. All rights reserved.

The activity of single cortical neurons varies across repeatedpresentations of an identical stimulus (Werner and Mountcastle,1963; Tolhurst et al., 1983). This variation makes single neuronsunreliable estimators of sensory stimuli. The neural representationof sensory stimuli is generally based on a population of neurons.The reliability of population coding is determined by the responsevariations of individual neurons and the correlation of the trial-by-trial fluctuations in their responses, or “noise correlation” (Abbottand Dayan, 1999; Averbeck et al., 2006). In particular, as noisecorrelation between neurons with similar stimulus preferencesapproaches zero, coding the sensory stimuli by the populationresponses becomes more reliable (Zohary et al., 1994; Shadlenet al., 1996). Although noise correlation may change during devel-opment, which then affects the perceptual and behavioral abilitiesof developing animals, to our knowledge this change has not beensystematically studied.

We investigated this problem by studying neurons in the pos-teromedial barrel subfield (PMBSF) of the rat somatosensory cortex.Rats explore their environment by moving their whiskers to locateand discriminate objects. The behavior and behavioral responses

sensory-evoked neuronal responses in rat barrel cortex during16/j.neures.2012.05.009

to whisker stimuli emerge in the second postnatal week andcontinue to develop up to the adult stage (Welker, 1964; Grantet al., 2012). Individual PMBSF neurons respond to whisker deflec-tions with a low probability (Simons, 1978). To encode whisker

ociety. All rights reserved.

dx.doi.org/10.1016/j.neures.2012.05.009


http://www.sciencedirect.com/science/journal/01680102

http://www.elsevier.com/locate/neures

mailto:[email protected]

mailto:[email protected]


IN PRESSG ModelN

2 ce Research xxx (2012) xxx–xxx

ds

o2pabadgtzsWgb

2

2

wtIe2

uIisemocbis

2

1s3pi22fi

abrtaiesstltta

2

riwcof

Fig. 1. Extracellular multiple single-unit recordings in the posterior medial barrelsubfield (PMBSF) of the primary somatosensory cortex. (A) Schematic diagram ofthe recording experiment. (B) Photograph of a P5 rat showing its whiskers. (C) Mul-tiple single-unit responses to whisker deflection were recorded using an electrodewith 16 vertically arrayed recording probes. All electrode penetrations were aimedperpendicularly to the cortical surface and were histologically verified from electriclesions (made at recording probes denoted as filled circles). Scars (enclosed withwhite dashes) were visible in the Nissl section. (D) Simultaneously recorded extra-cellular voltage signals. Eight neurons were isolated from this recording. The bottom

ARTICLESR-3456; No. of Pages 9

K. Ikezoe et al. / Neuroscien

eflection occurrences reliably, the activities of multiple neuronshould be used (Jadhav et al., 2009).

Previous studies examined developmental changes of responsesf single PMBSF neurons (Armstrong-James, 1975; Stern et al.,001; Borgdorff et al., 2007; Shoykhet and Simons, 2008). In theresent study, we examined how responses of both single neuronsnd neuronal populations with a shared principal whisker developy performing multiple single-unit recordings of whisker-evokedctivity in an ensemble of neurons arranged along a column. Weemonstrate that during the second and third postnatal weeks, sin-le neuron responses become shorter in latency and duration, andhat noise correlation observed in 1-week old rats reduces to nearero by postnatal day 30–31. Simulation of population responseshows that the reduction improves the reliability of the responses.

e suggest that the improved coding at the levels of both sin-le neurons and a population of neurons contributes to the betterehavioral performance in older animals.

. Materials and Methods

.1. Animal preparation

Sprague-Dawley rats of either sex, reared in litters with their mother in a cage,ere used. All surgical, experimental, and animal care protocols were approved by

he Animal Experiment Committee of Osaka University and conform to the Nationalnstitutes of Health Guide for the Care and Use of Laboratory Animals. Generalxperimental procedures were similar to those described previously (Kimura et al.,010).

Neuronal recordings were performed in 32 rats at P5–31 (day of birth, P0)nder urethane anesthesia (1.2–1.8 g/kg intra-peritoneal injection; Fig. 1A and B).

f spontaneous body movements were observed, additional urethane was admin-stered. Local anesthesia (lidocaine) was administered subcutaneously to surgicalites. A metal plate with a center hole for electrode penetration was attached to thexposed skull with glue and dental cement (Fig. 1A), and was held by a custom-anufactured holder. Through the hole, craniotomy and duratomy were performed

ver the PMBSF on the right hemisphere. After inserting an electrode, the exposedortex was covered with 1–2% agar to prevent drying. Throughout the experiments,ody temperature was maintained at 37 ◦C with a thermostatically controlled heat-

ng pad. An electrocardiogram was used to continuously monitor the anesthetictate.

.2. Electrophysiological recording and histology

We performed extracellular unit recordings using a single shaft electrode with6 recording probes arranged linearly and spaced 50 �m apart (width, 123 �m at itshaft; thickness, 15 �m; surface area of a single probe, 413 �m2; catalog # a1×16-mm50-413, NeuroNexus Tech., Ann Arbor, MI; Fig. 1C). We inserted an electrodeerpendicularly to the cortical surface using a microdrive. All recording probes were

n the cortex and spanned 750 �m below the pial surface, corresponding to layers–5 (Fig. 1C). After the electrode stopped slipping in the cortex, recording session of00 or 800 trials was started. Voltage signals were amplified (× 10,000), band-passltered (0.5–3 kHz), digitized (16 bit, 20 kHz) and archived (Fig. 1D).

After the recording was completed, electrolytic lesions were made by passingnodal current through two probes. The rats were then overdosed with pentobar-ital sodium, and transcardially perfused with 4% paraformaldehyde. The brain wasemoved, infiltrated with sucrose for cryoprotection, flash-frozen in optimal cuttingemperature medium, and sectioned at 80 �m in the coronal plane. Lesions, whichppeared as cell-sparse zones in Nissl-stained sections (Fig. 1C), were successfullydentified in all samples. Histological reconstruction showed that the electrode pen-trations were made perpendicularly to the cortical surface as intended, and that theampled neurons were located between layers 2 and 5 (see Table 1 for the number oftimulus-responsive neurons in each layer and each age group). We did not attempto determine by cytochrome oxidase staining whether the sampled neurons wereocated in barrel columns or in septa between barrels, because the recording diame-er of each probe was estimated to be more than 100 �m. The distance correspondso the interval between three successive recording probes, and spikes derived from

single neuron were often recorded at three successive probes (Fig. 1D).

.3. Whisker deflection

We identified the “principal” whisker for which deflection evoked the largestesponses at a recording site by deflecting whiskers with a hand probe and listen-


ng to an audio monitor for voltage signals. All whiskers except for the principalhisker on the left (i.e., contralateral) side of the face were then trimmed. The prin-

ipal whisker was cut to 10 mm from the base and held with the “V”-shaped tipf the stimulating probe (Fig. 1A). The whisker was deflected in a ramp-and-holdashion with the probe by using a piezoelectric bimorph actuator (Fuji Ceramics,

trace indicates the position of the whisker stimulus probe (a piezoelectric bimorphactuator) monitored by a laser displacement sensor.

Fujinomiya, Japan). The resiliency of rat whiskers increases during the first postna-tal month. Therefore, in order to deflect the whisker reliably and evoke the maximalresponses in PMSBF neurons, we adjusted the location of the stimulating probealong the whisker according to the age of the rat: 2–3 mm (the closest distance pos-sible) from the base at P5–9 and 4–6 mm at all other ages. The tip moved 1.0 mmover 20 ms (mean velocity: 50 mm/s). This speed was high enough to maximizeresponses in adult rats (Armstrong-James and Fox, 1987). We monitored movementof the stimulus probe tip over time by using a non-contact laser displacement sensor(LK-G85, Keyence, Osaka) and recorded this information for further analysis (16 bit,spatial resolution = 0.5 �m, 20 kHz). In all experiments we observed no unintendedvibration (“ringing”) of the stimulus probe (Fig. 1D, bottom trace). We defined theonset and offset of deflection as the times when the stimulus probe moved 30 �maway from and back to the resting position, respectively. The duration between theonset and the offset of deflection was 330 ms. The inter-stimulus interval (ISI) was2 or 10 s. Neurons in rats younger than P13 hardly respond to the second or laterwhisker deflections with an ISI of 2 s (Armstrong-James, 1975). In such cases, weapplied whisker deflections with an ISI of 10 s. The number of trials was 200 or 800


for recordings with 10-s ISI or 2-s ISI, respectively. The total time for a trial thusranged from 1600 s to 2000 s. For rats older than P13 tested for the two ISIs, wecompared the results between the 2-s ISI experiments and the 10-s ISI experimentsand found no significant difference.


ARTICLE IN PRESSG ModelNSR-3456; No. of Pages 9

K. Ikezoe et al. / Neuroscience Research xxx (2012) xxx–xxx 3

Table 1Number of rats, stimulus-responsive neurons, and pairs for each age group.

Age (postnatal day) 5–6 7–9 10–12 13–15 18–21 30–31 Total

# Rats 4 11 6 5 4 2 32# Neurons in L2, L3 2 4 8 12 12 20 58# Neurons in L4 1 7 5 8 10 40 71# Neurons in L5 3 19 16 23 32 23 116

43 54 83 245211 444 2737 3483

2

awndtwWbbis

filabitfibcndGtrdfo

r1stx

al

C

wfitt(ottdsoC

rnias

s

P6 P8 P10

P14 P21 P31

200 ms

A

CB

E

Age (postnatal day)

Age (postnatal day)

200

100

0Re

sp

on

se

du

ratio

n (

ms)

035 25201510

035 25201510

Age (postnatal day)

Re

sp

on

se

on

se

t la

ten

cy

(ms)

Firin

g r

ate

(spik

es/s

)

Re

sp

on

se

ma

gn

itu

de

(sp

ike

s/t

ria

l)

1.6

0.8

0

Age (postnatal day)

5 30252015100

Kruskal-Wallis testp = 0.001

D

1.6

0.8

F

Age (postnatal day)

Coeffic

ient of variation

of re

sp

on

se

ma

gn

itu

de

2

6

4

05 3025201510



0

5

0

10

0

0

20

40

0

10

0

10

0

100

200

Kruskal-Wallis testp < 0.0001

035 25201510

Ma

gn

itu

de

of sp

on

tan

eo

us firin

g (

330 m

s, spik

es/t

rial) Layer 2/3

Layer 4

Layer 5

Fig. 2. Developmental changes in PMBSF neuronal response properties. (A) Peri-stimulus time histograms of representative neurons at the listed age. Dotted linesindicate the onset and offset of whisker deflection. (B–F) Developmental changes

Total # neurons 6 30 29

# Neuron pairs 2 30 59

.4. Data analysis

Each individual probe of an electrode could record spikes from multiple neurons,nd spikes from a single neuron were recorded by multiple probes simultaneouslyith different amplitudes and waveforms (Fig. 1D). From such multichannel sig-als, we carried out isolation and classification of spikes by an automated methodescribed previously (Tamura et al., 2004). Only neurons whose spikes were con-inuously monitored for an entire recording period and which were responsive tohisker deflections were used for analyses (p < 0.01, Wilcoxon’s signed rank test).e could isolate only a small number of neurons in 5–12 day-old rats (Table 1),

ecause spikes were smaller in amplitude and more difficult to be isolated fromackground multi-unit activity in these young rats. Although we attempted exper-

ments in a larger number of animals for P5–6 than for older ages, recordings wereuccessful only for six neurons at this age.

The magnitude of a response in each trial was defined as the number of spikesred during 330 ms between whisker deflection onset and offset. We chose this

ong time window, because the response latency and duration varied greatly acrossges (Fig. 2A). Responses after stimulus offset were excluded from our analysis,ecause onset responses were sometimes merged with offset responses, especially

n younger animals. The response magnitude of a single neuron was calculated ashe mean of the response magnitudes across trials. The magnitude of spontaneousring for each trial was the number of spikes during the 330 ms period immediatelyefore whisker deflection onset. The coefficient of variation (CV) of responses wasalculated as the ratio of the standard deviation to the mean of the response mag-itude. The response onset and the response duration were determined from spikeensity functions obtained by temporally convolving a sequence of spikes with aaussian function (sigma = 2 ms). The response onset was defined as the point in

ime when the firing rate first exceeded 3 standard deviations of spontaneous firingate for a period of 7 ms (temporal resolution, 0.1 ms). The response duration wasefined as the interval between the response onset and the time when the responseell below 25% of the response peak. We used neurons with a response magnitudef >0.4 spikes/trial to calculate the duration.

Noise correlation was calculated as the Pearson’s correlation coefficient of theesponse magnitudes of two neurons. The cross-correlogram (CCG; Perkel et al.,967; Kohn and Smith, 2005) was calculated by using binary time sequences ofpike occurrences during a period of 330 ms after whisker deflection onset. Theime resolution of the sequences was 1 ms. A binary time sequence was defined asij(t) which takes the value of 1 if neuron j fires a spike in trial i during tth time binfter stimulus onset or a value of 0 otherwise. The CCG (coincidence/spike) at timeag � was calculated as follows:

CG (�) =1/M

∑M

i=1

∑N

t=1xi

1(t)xi2(t + �)

�(�)√

�1�2

,

here M is the number of trials, N the number of time bins, and �1 and �2 the meanring rate (spikes/s) of neurons in a pair. �(�) = T − |�|, where T is the duration of arial in seconds. We smoothed the CCGs with a 5 ms kernel. To determine whetherwo neurons fired spikes synchronously, we compared CCGs with shuffled CCGsPerkel et al., 1967). We computed a shuffled CCG by randomizing the trial orderf one neuron in a pair and calculating the CCG. We repeated this calculation 1000imes and then calculated the mean and standard deviation of the shuffled CCGs ofhe neuron pair. If the peak of a measured CCG was larger than the mean + 3 standardeviations of the shuffled CCGs from −10 to +10 ms, and larger than the mean + 5tandard deviation of the measured CCG from −100 to +100 ms (excepting the periodf −10 to +10 ms), we determined that the neuron pair showed a significant peak inCG.

We assessed the effects of pooling activities in a population of neurons on theeliability of population responses. We predicted CVs of summed responses from aeuronal population as a function of the number of pooled neurons by using exper-

mentally obtained data (i.e., the medians of CVs and correlation coefficients in eachge group). This calculation was performed separately for different age groups. Theummed responses, s, are defined as

Please cite this article in press as: Ikezoe, K., et al., Decorrelation of sensory-evoked neuronal responses in rat barrel cortex duringpostnatal development. Neurosci. Res. (2012), http://dx.doi.org/10.1016/j.neures.2012.05.009

=n∑

i=1

xi,

with age: (B) response onset latency, (C) response duration, (D) response magnitude,(E) coefficient of variation, (F) magnitude of spontaneous firing. Graphs show medianvalues (thick black lines: all neurons, Colored lines: neurons in each layer). Error barsrepresent the 1st and 3rd quartiles. (For interpretation of the references to color inthis figure legend, the reader is referred to the web version of this article.)


ING ModelN

4 ce Re

wt(

C

wrbonvcT

C

Co

3

o(aaF2pl(o

3

sdCt

wtsarr1it

elbdhsrrwm

tt



here n is the number of neurons in a population, and xi is the response magnitude ofhe ith neuron. The ratio of the standard deviation to the mean of summed responsesCvsummedresponses) is defined as

vsummedresponses =

√Var(s)

E(s)=

√∑n

i=1

∑n

j=1�ij

∑n

i=1�i

,

here Var(s) and E(s) are the variance and the mean of the summed responses,espectively, �i the mean response magnitude of neuron i, and � ij the covarianceetween responses of neurons i and j. In the case where i = j, � ij is the variancef responses of a single neuron. For simplicity, we assumed that the responses ofeurons in the same age group have the same response magnitude (�i = �) andariance (�2

ii= �2). Responses of neuronal pairs were also assumed to share the same

ovariance (� ij: constant), resulting in sharing the same correlation coefficients (r).hen,

vsummedresponses =

√∑n

i=1

∑n

j=1�ij

∑n

i=1�i

= �

n�

√n + n(n − 1)r = Cv

n

√n + n(n − 1)r

v and r were the medians of the CVs and the correlation coefficients, respectively,btained experimentally for each age group.

. Results

We performed multiple single-unit recordings from layers 2–5f the PMBSF in anesthetized rats from postnatal day 5 (P5) to P31Table 1). Neurons above and below the recorded neuron share

whisker that evokes the maximal response (principal whisker)nd form a functional column (Woolsey and Van der Loos, 1970;ox, 2008; Simons, 1978; Petersen, 2007; but see Sato et al.,007). By inserting an electrode with a linear array of 16 recordingrobes perpendicular to the cortical surface, we recorded extracel-

ular activity from single neurons or simultaneously from multiple2–74) neurons vertically arrayed along the electrode. Typically allf the neurons shared the principal whisker (Fig. 1).

.1. Development of the response properties of individual neurons

We pooled the data from different layers for the following analy-es of single neuron responses, because the developmental changesescribed below were similar among different layers (Fig. 2B–F).olored thin lines for different layers follow the thick black linehat stands for the mean across layers.

PMBSF neurons of P5–6 rats, the youngest animals we tested,ere capable of increasing firing rate in response to whisker deflec-

ions of 1 mm over 20 ms (Section 2 and Fig. 2A). They were notpontaneously active (Fig. 2F). The onset latency was long and vari-ble among neurons (median, 113 ms; range, 91–259 ms; n = 6). Theesponses lasted beyond the stimulus offset, and merged with theesponses caused by the stimulus offset. The response duration was08 ms (median, n = 4). We excluded two neurons with a low fir-

ng rate from this analysis because we could not reliably determineheir response duration.

The onset latency and the response duration gradually short-ned with age (p < 0.0001, Kruskal–Wallis test; Fig. 2A–C). Theargest changes occurred during the second and third week afterirth. Neurons from P30–31 rats responded to single whiskereflections with a short onset latency (median, 10 ms; n = 76), asas been previously reported (Simons, 1978). The responses werehort-lasting with a median duration of 12 ms (n = 21). Thus, bothesponse latency and duration were 10 times shorter in P30–31ats than in P5–6 rats. In some P30–31 neurons, the initial responseas followed by oscillatory activity (at ∼10 Hz) for several hundred


illiseconds.In contrast to the onset and duration of the responses,

he response magnitude and its trial-by-trial variation nei-her increased nor decreased monotonically with age. PMBSF

PRESSsearch xxx (2012) xxx–xxx

neurons fired spikes in response to whisker deflections infre-quently, resulting in a median firing rate of <0.7 spikes/trialthroughout development. The response magnitude (total numberof spikes) and the response peak height during the 330-ms periodof whisker deflection differed with age (p < 0.001, Kruskal–Wallistest). They were high on P5 and P15, and the change did not exhibita systematic trend (Fig. 2D; data not shown for the response peakheight). The rate of spontaneous firing (measured as the numberof spikes in the 330-ms period before the stimulus onset) was<0.1 spikes/trial, substantially lower than the sensory responsemagnitude at all ages (Fig. 2F, compare Fig. 2F with Fig. 2D).The spontaneous firing rate was statistically different across ages(p = 0.001, Kruskal–Wallis test), being highest at P13–21. Similarly,the trial-by-trial variation of responses as assessed by the coeffi-cient of variation [CV, (standard deviation)/mean] of the responsemagnitudes differed across ages (p = 0.002, Kruskal–Wallis test;Fig. 2E). However, we again found no systematic increase nordecrease with age.

Thus, the response properties of individual neurons change sys-tematically in time, but not in intensity (magnitude).

3.2. Pairwise correlation of trial-by-trial response variabilitybetween neurons

In the first and second weeks after birth, noise correlation wasstrongly positive between neurons. Fig. 3A shows the responses ofthree neurons recorded simultaneously from a P9 rat to repeateddeflections of the principal whisker. The three responded to onlya small fraction (4–6%) of repeated whisker deflection trials (neu-ron 1, 8/200; neuron 2, 10/200; neuron 3, 12/200). Despite the lowprobability of response from individual neurons, two or all threewere simultaneously activated in eight of the trials (see the ver-tically aligned raster plots in Fig. 3A, trial #: 3, 4, 11, 78, 97, 127,128, 186). In seven of the eight trials where neuron 1 responded,so too did neuron 2 and/or neuron 3. The variation of the responsemagnitude across trials was positively correlated between neuronpairs; i.e., noise correlation was positive (see Fig. 3B for the neu-ron pair 1–3; Spearman’s correlation coefficient rs = 0.86, p < 10−5).Overall, noise correlation was positive and statistically significantfor all possible combinations of neurons (0.67, 0.86, and 0.64 for theneuron pairs 1–2, 1–3, and 2–3, respectively; p < 0.01 for all pairs,filled bars; Fig. 3G, top panel).

In sharp contrast to strongly positive noise correlation observedin the P9 rat, responses were nearly independent between neuronsin P30–31 rats. For the 74 simultaneously recorded neurons from aP31 rat, only few trials had a large fraction of neurons respondingsimultaneously (Fig. 3E). Noise correlation was not significant formost neuron pairs (2373/2701 pairs; p ≥ 0.01). Data for an exam-ple pair (neuron pair 41–53) is shown in Fig. 3F (rs = 0.021, p = 0.5).The median noise correlation of the 74 neurons was close to zero(0.025; n = 2701; Fig. 3G, bottom panel) and slightly shifted posi-tively (p < 10−5, Wilcoxon’s signed rank test), as reported in layer2/3 neurons of mice (Poulet and Petersen, 2008)

The median noise correlation changed systematically with age(p < 0.0001, Kruskal–Wallis test, Fig. 3H); it was strongly positiveat P5–6 (2 pairs; 0.48, 0.37), but decreased gradually with devel-opment, reaching adult levels at P18–21. The distribution of noisecorrelation in a rat at an intermediate age (P13) was unimodal(p = 0.43, Hartigan’s dip test; Fig. 3G, middle panel), suggestingthat noise correlation between individual pairs decreased gradu-ally with age, rather than occurring in a stepwise fashion wherebythe proportion of pairs with adult-like noise correlation increased


over time.In P5–6 rats, we further examined noise correlation between

responses of single neurons and responses of background multi-ple neurons (multi-unit activity). Multi-unit activity was derived



K. Ikezoe et al. / Neuroscience Research xxx (2012) xxx–xxx 5

Fig. 3. Developmental changes in pairwise correlation of trial-by-trial responsevariations (noise correlation). (A, C, E) The magnitude of responses to repeated (200times) whisker stimulations of simultaneously recorded neurons in rats of variousages: P9 (A, 3 neurons, ISI = 10 s), P13 (C, 16 neurons, ISI = 2 s), and P31 (E, 74 neurons,ISI = 2 s). Normalized responses relative to the maximum response of each neuronare shown using the color scale to the right of each panel. (B, D, F) Scatter plots ofresponse magnitudes of two neurons recorded simultaneously. The dot color indi-cates the proportion of trials (see right color bar for scaling). Dashed lines denote themean response magnitude for each neuron. Neuronal pairs in B, D, and F are fromthe experiments shown in A, C, and E, respectively. (G) The distribution of noisecorrelation for the rats at P9 (top), P13 (middle) and P31 (bottom), also from theexperiments shown in A, C, and E, respectively. Vertical lines denote median values(0.67 for P9, 0.31 for P13, and 0.025 for P31). Filled bars represent pairs of neuronsw(

fiprwpw

o2twco

10 20 30

0

1

10 20 30

0

1

10 20 30

0

1

10 20 30

0

1

10 20 30

0

1

10 20 30

0

1

Age (postnatal day)

No

ise

co

rre

latio

n

L4-L4 pair


L4-L5 pair


L2/3-L5 pair


L2/3-L4 pair


L2/3-L2/3 pair


L5-L5 pair


3

2

4

9

16 741

1

14

62

311

2

34

44741

11

19

31

171

8

1

15

23

45

321 141

7

4232

631

Fig. 4. Developmental changes of noise correlation of pairs from the same or differ-ent layers underlying Fig. 3H. Lines represent the median values; error bars denotethe 1st and 3rd quartiles. The number of pairs to calculate the median values is

ith noise correlation significantly different from 0 (p < 0.01). (H) Noise correlationmedian values with 1st–3rd quartiles) plotted against age.

rom simultaneously recorded spikes from neurons other than thesolated single neuron. Careful monitoring of spike waveforms waserformed so that there was no cross talk between single neuronecording and multi-unit recording. The median noise correlationas highly positive (6 pairs; 0.11, 0.61, 0.57, 0.62, 0.47, 0.61), sup-orting the notion that responses of neurons are strongly correlatedith each other at P5–6.

The magnitude of noise correlation depends on the spike countf the responses (de la Rocha et al., 2007; Cohen and Maunsell,009; Mitchell et al., 2009; Cohen and Kohn, 2011). In our data,he magnitude of noise correlation did indeed positively correlate


ith the geometric mean response of two neurons (Spearman’sorrelation coefficient rs = 0.29, p < 10−5), the response magnitudef the less responsive neuron of the pair (rs = 0.29, p < 10−5), and

shown beside each data point.

the more responsive neuron of the pair (rs = 0.24, p < 10−5). Theserelationships may explain the developmental decrease in noise cor-relation. However, the response magnitude of PMBSF neurons didnot decrease systematically with age (Fig. 2D). Noise correlationchanges did not correlate with the developmental changes of themedian response magnitude (p = 0.48). Therefore, the decrease innoise correlation with age cannot be explained by the developmen-tal change in the response magnitude.

It has been shown in several cortical areas that noise correlationbetween neurons decreases with their tangential distance acrossthe cortical surface (Kerr et al., 2007; Smith and Kohn, 2008). Inour experiments, neurons were recorded along an electrode witha linear array of recording probes inserted perpendicularly to thecortical layers (Fig. 1C). Neurons were recorded from the same tan-gential point in the cortex, but from different radial depths belowthe cortical surface. We examined whether the decrease in noisecorrelation with age was accounted for by any biased samplingof neuron pairs with different radial distances, finding no nega-tive correlation between the radial distance of neurons in a pairand noise correlation between them at each age or across ages(p = 0.15–1.0, one-tailed, Spearman’s rank correlation). Further-more, when we calculated noise correlation separately for neuronpairs in different layers, noise correlation decreased with age in all


cases (p < 0.0001, Kruskal–Wallis test; Fig. 4). It is unlikely then thatthe systematic decrease in noise correlation with age arises from



6 K. Ikezoe et al. / Neuroscience Research xxx (2012) xxx–xxxco

incid

en

ce

/sp

ike

0

0.03

0.03

P7

0 001001-

Time (ms)

P31

co

incid

en

ce

/sp

ike

0 001001-

Time (ms)

Peak w

idth

(F

WH

H, m

s)

0

20

5 10 15 20 25 30

Age (postnatal day)

5 10 15 20 25 30

Age (postnatal day)P

ea

k h

eig

ht

(co

incid

en

ce

/sp

ike

)

0

0

0.02

0.04

0.06

CA

DB

Fig. 5. Spike timing correlation at different ages. (A, B) Cross-correlograms (CCGs)of neuron pairs in a P7 (A) and a P31 rat (B). Thick lines: raw CCGs. Thin lines:CCGs calculated from trial-shuffled spike sequences (shuffle-predicted CCGs; seeSection 2). (C) Peak height of CCGs in each age group (median values with 1st–3rdquartiles). (D) Developmental change in peak width of CCGs (median values with1Bp

sa

ast1iatfspposP4(fg0cwwhaTc

3r

o

0

1

2

3

CV

of sum

med r

esponses

of th

e n

eu

ron

al p

oo

l

0

1

2

3

CV

of su

mm

ed

re

sp

on

se

s

of th

e n

eu

ron

al p

oo

l

2515105 3020

P31

P13

Age (Postnatal day)

1

2

5

10

20501001000

706050403020100

Number of neurons

B

A

Fig. 6. Calculated CVs of pooled responses as a function of age and the numberof pooled neurons. (A) CVs of pooled responses in calculated from spike data ofrepresentative rats on P13 and P31 are plotted against the number of neurons in a

st–3rd quartiles). Data for C and D are only from statistically significant peaks.ecause there was only one data point at P5–6, it was not connected to the dataoints for other ages.

ampling neurons of different distances across animals of differentge.

Noise correlation is the covariation of responses of a neuron paircross trials, i.e., the correlation of response variations on a timecale of seconds. We next examined whether the response correla-ion on a time scale of milliseconds (spike synchrony, Perkel et al.,967) changes with age. Spike synchrony is proposed to play an

mportant role in the representation of sensory stimuli (Gray, 1999)nd can influence noise correlation (Bair et al., 2001). At the sameime, the two might develop differently, because they can ariserom different sources (Smith and Kohn, 2008). To analyze spikeynchrony, we calculated cross-correlograms (CCGs) between aair of neurons (Perkel et al., 1967), finding statistically significanteaks (see Section 2) at a near-zero time lag in CCGs of a fractionf neuron pairs at all ages (Fig. 5A and B). The number of pairs withignificant peaks/the number of all pairs were 1/2 at P5–6, 5/30 at7–9, 11/59 at P10–12, 18/211 at P13–15, 60/444 at P18–21, and58/2737 at P30–31; the proportion of pairs with significant peakssee Section 2) was comparable across all ages except for P13–15,or which the proportion was lower. The peak height decreasedradually with age: the median peak height was 0.026 at P5–6 and.0089 at P30–31 (Fig. 5A–C; p < 0.0001, Kruskal–Wallis test). Byontrast, the median width of statistically significant peaks (fullidth at half height) was the largest during the second and thirdeek (Fig. 5A, B and D; p < 0.0001, Kruskal–Wallis test). The peakeight in CCGs was correlated with noise correlation at each agend all ages pooled together (p < 0.01, Spearman’s rank correlation).hus, the degree of spike synchrony decreased with age as did noiseorrelation.

.3. Developmental changes in the variations of pooled neuron


esponses

The effects of noise correlation on the population code dependn the manner in which the population responses are decoded

pool. The number on the line is age of the rat. (B) CVs of simulated pooled responsesfor pools of 1–1000 neurons are plotted against age. The numbers on the right denotethe number of pooled neurons.

(Abbott and Dayan, 1999). One decoding method involves poolingor summing activities across a population of neurons with sharedstimulus selectivity (Tolhurst et al., 1983; Shadlen et al., 1996;Jadhav et al., 2009). If trial-by-trial fluctuations of single neuronsare independent of each other, the pooling of responses across apopulation averages out the fluctuations of single neurons. Thus, apopulation of neurons can represent sensory stimuli more reliablythan can single neurons. If responses covary among member neu-rons in a pool, the fluctuations cannot be averaged out by summingthe responses among them.

Here, we examined the effects of summing responses across apopulation of neurons sharing a principal whisker on the signalingstability across trials (Fig. 6). We first estimated the mean CVs ofpooled responses of variously sized neuron populations based ontwo representative data sets, one from 16 neurons of a P13 rat andthe other from 74 neurons of a P31 rat (Fig. 6A). For the P13 rat,we calculated the mean CV for each pooling size. When we calcu-lated the CVs for a large pool size (>14) for the P31 rat, the number ofcombinations of neurons were enormous. Therefore, for a given sizeof >14 neurons, we randomly chose neurons from the data set tomake a neuron pool, and calculated a CV for the pooled responses.


This calculation was repeated 1000 times to obtain the mean CVof the pool size. In both rats, the mean CV of summed responsesbecame smaller as the number of pooled neurons increased. Thiseffect was stronger for smaller pools. As the number of neurons


ING ModelN

ce Res

iptrbtwtfrw

wsw(bpnrstl(l

4

riftodantpra

dasr(ctt

dJdaos22lpani(



n the pool increased, the effect became weaker and finally disap-eared (i.e., the effect was saturated). For example, an increase inhe number of pooled neurons from 1 to 2 or from 2 to 5 greatlyeduced the population CV in both rats; however, adding neuronseyond 10 decreased the CVs by a smaller amount. In the P13 rat,he impact of expanding the pool was rapidly attenuated. The CVas almost saturated for 16 neurons, and the CV (1.1) was higher

han that (0.64) in the P31 rat. In the P31 rat, the CV decreased evenor adding >10 neurons to the pool. When responses of all 74 neu-ons recorded simultaneously were pooled, the CV of the responseas 0.43.

For estimating the effects of pooling across a larger population,e performed computational simulation. From the median CV of

ingle neurons and the median noise correlation of neuron pairs,e estimated the CVs of pooled responses of up to 1000 neurons

Fig. 6B; see Section 2). As in Fig. 6A, the effect of decreasing CVsecame weaker for a larger pool, and was nearly saturated for aool of more than 50 neurons. For summed responses of 95–1000eurons, the mean CVs decreased with age (p = 0.058, Spearman’sank correlation). Because the CVs of single neurons did not changeystematically with age (Figs. 2E and 6B, uppermost line), the sys-ematic decrease in CVs of summed responses with age was mostikely caused by the age-dependent decrease in noise correlationFig. 3). The low noise correlation observed in older animals thuseads to reliable responses from a neuronal population.

. Discussion

We examined how whisker-evoked responses of PMBSF neu-ons changed during the first postnatal month. Responses byndividual neurons to whisker deflection occurrences becameaster and transient during the first few weeks after birth. In addi-ion to these changes, positive noise correlation between pairsf neurons arrayed vertically across cortical layers drasticallyecreased during the first three postnatal weeks. Computationalnalyses showed that the summed responses of a population ofeurons became gradually stable with age over repeated stimula-ion. Because behavioral performance improves during the sameeriod, it is possible both changes in responses of individual neu-ons and the correlative activity between an ensemble of neuronsre partly responsible.

We showed that PMBSF neurons respond to single-whiskereflection at P5–6 (Fig. 2), which is before rats start to exhibit activend long bursts of whisking (Welker, 1964; Grant et al., 2012). Ithould be noted that photo-stimulation of layer 4 neurons in P8at slice preparation fails to synaptically activate layer 2/3 neuronsBureau et al., 2004). The response we observed in vivo may beaused, not via layer 4-to-layer 2/3 synapses, but by direct activa-ion from thalamocortical axons diffusely arborizing from layers 2o 5 at this age (Wise and Jones, 1978).

The responses were slower in onset latency and longer inuration than the responses reported for P7 rats (Armstrong-

ames, 1975; Borgdorff et al., 2007). The slow and temporallyispersed responses of PMBSF neurons in P5–6 animals began tocquire adult-like characteristics over the next two weeks of devel-pment, leveling off around the third postnatal week. Previoustudies examined layer 2/3 neurons during P12–P20 (Stern et al.,001) and layer 4 neurons during P14–P65 (Shoykhet and Simons,008), and reported similar developmental tendencies. This change

ikely involves many factors such as maturation of membraneroperties of individual neurons, myelination of thalamocortical


xons (Jacobson, 1963) and functional recruitment of inhibitoryeurons into the neural circuit (Daw et al., 2007). Myelination

n thalamocortical axons makes their conduction velocity fasterSalami et al., 2003; Kimura et al., 2010). Inhibitory inputs sharpen

PRESSearch xxx (2012) xxx–xxx 7

temporal aspects of responses of PMBSF neurons to whisker deflec-tions (Zhu and Connors, 1999; Wilent and Contreras, 2004).

The second and third postnatal weeks correspond to the periodwhen rats develop a full set of exploratory whisking behavior,including recurrent protraction and retraction of the snout and vib-rissae, and a series of rapid head movements and fixations (Welker,1964). It has been shown that precisely timed spikes of PMBSFneurons to a whisker movement encode stimulus position andsurface texture (Jadhav et al., 2009; Panzeri et al., 2001). It is there-fore possible that the sharpening of temporal characteristics of theresponses is important for the developmentally improved neuralcoding for the temporal structure of the sensory inputs that repre-sent stimulus position and surface texture.

Besides changes in the temporal properties of individual neu-ronal responses, noise correlation among neurons sharing the sameprincipal whisker gradually decreased during development (Fig. 3).Because of this decrease, summed responses became reliable withage (Fig. 6). Theoretical studies predicted that noise correlationseverely limits an animal’s psychophysical performance by reduc-ing the reliability of the summed responses (Zohary et al., 1994;Shadlen et al., 1996). Moreover, recent single-unit studies haveshown that neurons in primate visual area V4 decorrelate trial-by-trial response variations between them when the monkey directsits attention to a task-relevant stimulus (Cohen and Maunsell,2009; Mitchell et al., 2009). This attention-dependent decrease innoise correlation explains most of the attentional improvement ofthe population signal, and is suggested to be the major contrib-utor to the improved psychophysical performance. The postnataldecrease in the noise correlation of PMBSF neurons may contributeto the developmental improvement of tactile sensation throughwhisking. This hypothesis would be experimentally tested if anytransgenic/knockout or pharmacological techniques can allow usto influence the timing of the decorrelation without affecting themagnitude and trial-variation of responses.

The developmental improvement in tactile sensation is likelyto be linked to age-dependent increases in whisking quality. Inadult rats, the contact of whiskers with an object during whisk-ing induces an immediate retraction of the whiskers (Mitchinsonet al., 2007). This retraction is necessary to collect information effec-tively by keeping a distance between the whiskers and the object.The retraction first emerges at P10, with its probability increas-ing systematically up until P21 (Grant et al., 2012). The increasedreliability of pooled PMBSF neuron responses paralleled the devel-opment of whisker retraction. PMBSF neurons send outputs tothe spinal trigeminal nucleus and can control whisker retractiondirectly without the involvement of the primary motor cortex(Matyas et al., 2010). Thus, the increased reliability may contributeto the development of the immediate retraction as a result of anincrease in the accuracy of sensory information coding.

One may concern that anesthesia effects of urethane may dif-fer across different ages, which could account for some of thedevelopmental changes of PMBSF neurons. Urethane activatesBa2+-sensitive K+ leak conductance, but does not alter glutamater-gic or GABAergic synaptic transmission (Sceniak and Maciver,2006). Urethane-anesthesia has been commonly used to exam-ine correlation of neuronal activities and population coding in ratsomatosensory cortex (Kerr et al., 2007; Golshani et al., 2009;Petersen et al., 2001). Because the degree of correlation of spon-taneous activity in urethane-anesthetized rats during postnataldevelopment is indistinguishable from that in unanesthetized rats(Golshani et al., 2009), we believe that our conclusion is unaffectedby the use of anesthesia. It should be noted that recordings in


unanesthetized animals will not work for the purpose of the presentstudy, because the animals will move their whiskers spontaneouslyand we cannot obtain neural activity under constant conditions andhence cannot assess noise correlation.


ING ModelN

8 ce Re

retcpB(Ptwscd

mpa(CnGgStsnrrbFtdmrtimettbcmtewa

A

wWeCJi2

R

A

A



We examined noise correlation between vertically aligned neu-ons, which typically shared their principal whisker. We did notxamine noise correlation between distant neurons across the cor-ical surface, i.e., neurons located in different barrel columns. Noiseorrelation between neurons with different principal whiskers hasotentially an impact on coding of deflection of multiple whiskers.y using 2-photon calcium imaging techniques, Golshani et al.2009) have shown that correlation in spontaneous activity ofMBSF neurons decreases with the tangential distance betweenhem and the correlation decreases during the second postnataleek. Whether the similar developmental events happen in sen-

ory responses should be examined to address how the sensoryoding of multiple whisker deflections changes during postnatalevelopment.

High noise correlation during the early postnatal period (Fig. 3)ay be advantageous for network formation at the expense of

erceptual performance. In the PMBSF, a substantial fraction ofxonal projections and synaptic connections develop postnatallyBender et al., 2003; Bureau et al., 2004; Ashby and Isaac, 2011).orrelated spontaneous neuronal activities are pronounced in theeural circuit at an early postnatal stage (Adelsberger et al., 2005;olshani et al., 2009). This correlated spontaneous activity is sug-ested to play an important role in circuit formation (Katz andhatz, 1996). Stimulus-driven activities also play a crucial role inhe formation of neuronal connections. For example, in the ratomatosensory system, trimming of whiskers in the early post-atal period results in low tactile performance, changes in neuralesponsesiveness, a decrease in coincidence firing of multiple neu-ons, aberrant synaptic circuits, and abnormal barrel innervationsy afferents (Carvell and Simons, 1996; Shepherd et al., 2003;ox, 2008; Ghoshal et al., 2009; Popescu and Ebner, 2010). Givenhat PMBSF neurons respond with a low probability to whiskereflections, the high noise correlation increases opportunities forultiple neurons to fire coincidently. This may help developmental

efinement of axonal projections and synaptic connections throughhe Hebbian mechanism, which requires temporally coincident fir-ngs in pre-synaptic and post-synaptic neurons (Hebb, 1949). One

ight expect that the noise correlation increases with age by thestablishment of synaptic connections. It turned out, however,hat PMBSF neurons in older animals had lower noise correla-ion (Fig. 3H). The decrease in noise correlation may be causedy development of other aspects of the cortex such as inhibitoryircuits (Kimura et al., 2010; Renart et al., 2010) and neuronalembrane properties (Poulet and Petersen, 2008). We suggest

hat the correlation of stimulus-driven activities contributes toxperience-dependent circuit formation in early postnatal life,hereas decorrelation of activities among a population of neurons

t later stages enables efficient coding of the sensory information.

cknowledgments

We thank Hidekazu Kaneko for providing spike-sorting soft-are; Guy Elston, Hiroshi M Shiozaki, Yumiko Yoshimura, and Lisau for comments on the manuscript; and Peter Karagiannis for

diting the English. This work was supported by grants from theore Research for Evolutional Science and Technology Program of

apan Science and Technology Agency, and by grants from the Min-stry of Culture, Sports, Science and Technology of Japan (08108015,0500285, 23135521, 23135522).

eferences


bbott, L.F., Dayan, P., 1999. The effect of correlated variability on the accuracy of apopulation code. Neural Comput. 11, 91–101.

delsberger, H., Garaschuk, O., Konnerth, A., 2005. Cortical calcium waves in restingnewborn mice. Nat. Neurosci. 8, 988–990.

PRESSsearch xxx (2012) xxx–xxx

Armstrong-James, M., 1975. The functional status and columnar organization of sin-gle cells responding to cutaneous stimulation in neonatal rat somatosensorycortex S1. J. Physiol. (Lond.) 246, 501–538.

Armstrong-James, M., Fox, K., 1987. Spatiotemporal convergence and divergence inthe rat S1 “barrel” cortex. J. Comp. Neurol. 263, 265–281.

Ashby, M.C., Isaac, J.T., 2011. Maturation of a recurrent excitatory neocortical circuitby experience-dependent unsilencing of newly formed dendritic spines. Neuron70, 510–521.

Aslin, R.N., Alberts, J.R., Petersen, M.R. (Eds.), 1981. Development of Perception:Psychobiological Perspectives. Academic Press, New York.

Atkinson, J., 2002. The Developing Visual Brain. Oxford University Press, Oxford, UK.Averbeck, B.B., Latham, P.E., Pouget, A., 2006. Neural correlations, population coding

and computation. Nat. Rev. Neurosci. 7, 358–366.Bair, W., Zohary, E., Newsome, W.T., 2001. Correlated firing in macaque visual area

MT: time scales and relationship to behavior. J. Neurosci. 21, 1676–1697.Bender, K.J., Rangel, J., Feldman, D.E., 2003. Development of columnar topography

in the excitatory layer 4 to layer 2/3 projection in rat barrel cortex. J. Neurosci.23, 8759–8770.

Borgdorff, A.J., Poulet, J.F., Petersen, C.C., 2007. Facilitating sensory responsesin developing mouse somatosensory barrel cortex. J. Neurophysiol. 97,2992–3003.

Bureau, I., Shepherd, G.M., Svoboda, K., 2004. Precise development of functional andanatomical columns in the neocortex. Neuron 42, 789–801.

Carvell, G.E., Simons, D.J., 1996. Abnormal tactile experience early in life disruptsactive touch. J. Neurosci. 16, 2750–2757.

Cohen, M.R., Kohn, A., 2011. Measuring and interpreting neuronal correlations. Nat.Neurosci. 14, 811–819.

Cohen, M.R., Maunsell, J.H., 2009. Attention improves performance primarily byreducing interneuronal correlations. Nat. Neurosci. 12, 1594–1600.

Daw, M.I., Ashby, M.C., Isaac, J.T., 2007. Coordinated developmental recruitmentof latent fast spiking interneurons in layer IV barrel cortex. Nat. Neurosci. 10,453–461.

Fox, K., 2008. Barrel Cortex. Cambridge University Press, New York.Golshani, P., Gonc alves, J.T., Khoshkhoo, S., Mostany, R., Smirnakis, S., Portera-

Cailliau, C., 2009. Internally mediated developmental desynchronization ofneocortical network activity. J. Neurosci. 29, 10890–10899.

Ghoshal, A., Pouget, P., Popescu, M., Ebner, F., 2009. Early bilateral sensory depri-vation blocks the development of coincident discharge in rat barrel cortex. J.Neurosci. 29, 2384–2392.

Grant, R.A., Mitchinson, B., Prescott, T.J., 2012. The development of whisker controlin rats in relation to locomotion. Dev. Psychobiol. 54, 151–168.

Gray, C.M., 1999. The temporal correlation hypothesis of visual feature integration:still alive and well. Neuron 24, 31–47.

Hebb, D.O., 1949. The Organization of Behavior: A Neuropsychological Theory. Wiley,New York.

Jacobson, S., 1963. Sequence of myelinization in the brain of the albino rat. A. Cerebralcortex thalamus and related structures. J. Comp. Neurol. 121, 5–29.

Jadhav, S.P., Wolfe, J., Feldman, D.E., 2009. Sparse temporal coding of elementarytactile features during active whisker sensation. Nat. Neurosci. 12, 792–800.

Katz, L.C., Shatz, C.J., 1996. Synaptic activity and the construction of cortical circuits.Science 274, 1133–1138.

Kerr, J.N., de Kock, C.P., Greenberg, D.S., Bruno, R.M., Sakmann, B., Helmchen, F., 2007.Spatial organization of neuronal population responses in layer 2/3 of rat barrelcortex. J. Neurosci. 27, 13316–13328.

Kimura, F., Itami, C., Ikezoe, K., Tamura, H., Fujita, I., Yanagawa, Y., Obata, K., Ohshima,M., 2010. Fast activation of feedforward inhibitory neurons from thalamic inputand its relevance to the regulation of spike sequences in the barrel cortex. J.Physiol. (Lond.) 588, 2769–2787.

Kiorpes, L., Movshon, J.A., 2004. Neural limitations on visual development in pri-mates. In: Chalupa, L., Werner, J.S. (Eds.), The Visual Neurosciences. MIT Press,Cambridge, MA, pp. 159–173.

Knudsen, E.I., 2002. Instructed learning in the auditory localization pathway of thebarn owl. Nature 417, 322–328.

Kohn, A., Smith, M.A., 2005. Stimulus dependence of neuronal correlation in primaryvisual cortex of the macaque. J. Neurosci. 25, 3661–3673.

Matyas, F., Sreenivasan, V., Marbach, F., Wacongne, C., Barsy, B., Mateo, C., Aronoff, R.,Petersen, C.C., 2010. Motor control by sensory cortex. Science 330, 1240–1243.

Mitchell, J.F., Sundberg, K.A., Reynolds, J.H., 2009. Spatial attention decorrelatesintrinsic activity fluctuations in macaque area V4. Neuron 63, 879–888.

Mitchinson, B., Martin, C.J., Grant, R.A., Prescott, T.J., 2007. Feedback control in activesensing: rat exploratory whisking is modulated by environmental contact. Proc.Biol. Sci. 274, 1035–1041.

Panzeri, S., Petersen, R.S., Schultz, S.R., Lebedev, M., Diamond, M.E., 2001. The roleof spike timing in the coding of stimulus location in rat somatosensory cortex.Neuron 29, 769–777.

Perkel, D.H., Gerstein, G.L., Moore, G.P., 1967. Neuronal spike trains and stochasticpoint processes. II. Simultaneous spike trains. Biophys. J. 7, 419–440.

Petersen, R.S., Panzeri, S., Diamond, M.E., 2001. Population coding of stimulus loca-tion in rat somatosensory cortex. Neuron 32, 503–514.

Petersen, C.C., 2007. The functional organization of the barrel cortex. Neuron 56,339–355.


Popescu, M.V., Ebner, F.F., 2010. Neonatal sensory deprivation and the develop-ment of cortical function: unilateral and bilateral sensory deprivation result indifferent functional outcomes. J. Neurophysiol. 104, 98–107.

Poulet, J.F., Petersen, C.C., 2008. Internal brain state regulates membrane potentialsynchrony in barrel cortex of behaving mice. Nature 454, 881–885.


ING ModelN

ce Res

R

d

S

S

S

S

S

S

S

S

S



enart, A., de la Rocha, J., Bartho, P., Hollender, L., Parga, N., Reyes, A., Har-ris, K.D., 2010. The asynchronous state in cortical circuits. Science 327,587–590.

e la Rocha, J., Doiron, B., Shea-Brown, E., Josic, K., Reyes, A., 2007. Corre-lation between neural spike trains increases with firing rate. Nature 448,802–806.

alami, M., Itami, C., Tsumoto, T., Kimura, F., 2003. Change of conduction velocity byregional myelination yields constant latency irrespective of distance betweenthalamus and cortex. Proc. Natl. Acad. Sci. U.S.A. 100, 6174–6179.

ato, T.R., Gray, N.W., Mainen, Z.F., Svoboda, K., 2007. The functional microarchitec-ture of the mouse barrel cortex. PLoS Biol. 5, e189.

ceniak, M.P., Maciver, M.B., 2006. Cellular actions of urethane on rat visual corticalneurons in vitro. J. Neurophysiol. 95, 3865–3874.

hadlen, M.N., Britten, K.H., Newsome, W.T., Movshon, J.A., 1996. A computationalanalysis of the relationship between neuronal and behavioral responses to visualmotion. J. Neurosci. 16, 1486–1510.

hepherd, G.M., Pologruto, T.A., Svoboda, K., 2003. Circuit analysis of experience-dependent plasticity in the developing rat barrel cortex. Neuron 38, 277–289.

hoykhet, M., Simons, D.J., 2008. Development of thalamocortical response trans-formations in the rat whisker-barrel system. J. Neurophysiol. 99, 356–366.

imons, D.J., 1978. Response properties of vibrissa units in rat SI somatosensory


neocortex. J. Neurophysiol. 41, 798–820.mith, M.A., Kohn, A., 2008. Spatial and temporal scales of neuronal correlation in

primary visual cortex. J. Neurosci. 28, 12591–12603.tern, E.A., Maravall, M., Svoboda, K., 2001. Rapid development and plasticity of layer

2/3 maps in rat barrel cortex in vivo. Neuron 31, 305–315.

PRESSearch xxx (2012) xxx–xxx 9

Tamura, H., Kaneko, H., Kawasaki, K., Fujita, I., 2004. Presumed inhibitory neurons inthe macaque inferior temporal cortex: visual response properties and functionalinteractions with adjacent neurons. J. Neurophysiol. 91, 2782–2796.

Tolhurst, D.J., Movshon, J.A., Dean, A.F., 1983. The statistical reliability of signals insingle neurons in cat and monkey visual cortex. Vision Res. 23, 775–785.

Walsh, E.J., Romand, R., 1992. Functional development of the cochlea and thecochlear nerve. In: Romand, R. (Ed.), Development of Auditory and VestibularSystems, vol. 2. Elsevier Science, Amsterdam, pp. 161–219.

Welker, W.I., 1964. Analysis of sniffing of the albino rat. Behaviour 22, 223–244.Werner, G., Mountcastle, V.B., 1963. The variability of central neural activity in a

sensory system, and its implications for the central reflection of sensory events.J. Neurophysiol. 26, 958–977.

Wilent, W.B., Contreras, D., 2004. Synaptic responses to whisker deflections in ratbarrel cortex as a function of cortical layer and stimulus intensity. J. Neurosci.24, 3985–3998.

Wise, S.P., Jones, E.G., 1978. Developmental studies of thalamocortical and com-missural connections in the rat somatic sensory cortex. J. Comp. Neurol. 178,187–208.

Woolsey, T.A., Van der Loos, H., 1970. The structural organization of layer IV in thesomatosensory region (SI) of mouse cerebral cortex: the description of a corticalfield composed of discrete cytoarchitectonic units. Brain Res. 17, 205–242.


Zhu, J.J., Connors, B.W., 1999. Intrinsic firing patterns and whisker-evoked synap-tic responses of neurons in the rat barrel cortex. J. Neurophysiol. 81,1171–1183.

Zohary, E., Shadlen, M.N., Newsome, W.T., 1994. Correlated neuronal discharge rateand its implications for psychophysical performance. Nature 370, 140–143.


Documents

G Model ARTICLE IN PRESS - 大阪大学 cite this article in press as: Ikezoe, K., et al., Decorrelation of sensory-evoked neuronal responses in rat barrel cortex during postnatal