1

Neuron, volume 51

Supplemental Data

Subplate Neurons Regulate Maturation of Cortical Inhibition and

Outcome of Ocular Dominance Plasticity

Patrick O. Kanold and Carla J. Shatz

Materials and Methods

All experiments were performed according to the Harvard Medical School Institutional

Animal Care and Use Committee Protocol.

Surgery and Anesthesia

Ablations: Subplate neurons in visual cortex of 30 cats of both sexes were ablated

selectively between P6-P9 by focal injection of either p75-immunotoxin (ME20.4-SAP,

Advanced Targeting Systems, 0.5 µl, 0.25-1 mg/ml) or kainic acid (Sigma, 0.5 µl,

10mg/ml) (control injections: normal saline 0.5 µl) using sterile surgical techniques as

described previously (Ghosh and Shatz, 1992; Kanold et al., 2003; Lein et al., 1999).

Animals were anesthetized with 3-4% isofluorane and maintained during surgery at 3-

4%. Fluorescent latex microspheres (Lumafluor, Naples, Fl, 10% by volume) were added

to verify the injection sites. At these early ages subplate neuron ablations using either

immunotoxin or kainate are highly selective and leave neurons in the overlying cortical

plate almost completely intact (Ghosh and Shatz, 1992; Kanold et al., 2003; Lein et al.,

1999). At the early time ages used for investigation in this study (<P40) we did not see

2

evidence for atrophy of layer 4 neurons (see Fig. 1,3,4, S1) as has been occasionally seen

after very long survival times (Ghosh and Shatz, 1994). In fact as this study and our

previous study shows, the basic electrophysiological and morphological properties of

layer 4 neurons in ablated cortex are not grossly abnormal (Kanold et al., 2003). Both

ablation methods cause similar anatomical and functional deficits with immunotoxin

injections producing smaller sized ablations (Kanold et al., 2003). Indeed, methods of

subplate ablation (immunotoxin, N=2 animals vs. kainate, N=3 animals) did not give

different results for the DE projections after monocular deprivation (MD) (P>0.1). For

MD, a drop of ophthalmic local anesthetic, proparacaine HCL, was placed in one eye,

which was sutured closed under general anesthesia with isofluorane (see above). Tissue

from 5 unmanipulated animals was used for quantitative PCR and in situ hybridization

analysis of expression levels.

Minipump implantations

Animals (P6-P7) were anesthetized with 3-4% isofluorane and maintained during surgery

at 3-4%. A small craniotomy (~1mm) was performed overlying visual cortex and a

canula was inserted into the cortex (~3mm deep) (methods modified Bear et al. 1990).

The canula was connected to an osmotic minipump (Alzet 2002, delivering 0.5ul/h) filled

with 50mM DL-2-amino-5-phosphonovalerate (DL-AP5) and 10mM 6-nitro-7-

sulfamoylbenzo[f]quinoxaline-2,3-dione (NBQX, disodium salt), specific NMDA and

AMPA blockers (Sigma) in 0.9% saline or 0.9% saline alone (control). After 1-2 weeks

infusion time animals were euthanized with an IP overdose of sodium pentobarbital (200

mg/kg to effect) and the brains were rapidly removed. Coronal sections were cut (~1mm

thick) from visual cortex surrounding the infusion site (~2-3 mm from infusion site; in

3

this area a yellow/brown residue from the NBQX is clearly visible indicating

effectiveness of infusion) and from the contralateral visual cortex. RNA was extracted

and qPCR for KCC2, GABAA-α1 and HPRT was performed as described below. For

each animal 2-4 samples of the infused and contralateral hemisphere were compared. It is

likely that samples from the infused hemisphere also included some tissue not receiving

effective glutamatergic blockade due to distance from the infusion site. Thus in the

reported average data, the mRNA levels likely represent an underestimate of the mRNA

levels after infusion.

In-Situ hybridizations

Animals were sacrificed with an IP overdose of sodium pentobarbital (200 mg/kg to

effect) and the brains were rapidly removed and frozen in cryoprotective medium (M1,

Shandon). Horizontal sections were cut (12-15 µm) on a cryostat. In-situ hybridizations

were performed as described previously (Kanold et al., 2003; Lein et al., 1999). Template

sequences were generated by RT-PCR from P28 cat RNA using oligonucleotide primers

that span the regions of maximum nucleotide sequence dissimilarity between the different

GABAA receptor subunits but high similarity between mouse and human. KCC2 primers

were made to regions of high homology between mouse and human.

Primers:

GABA-A α1 (Genbank Accession number: NM_000806, 130-402 bp):

L: AGTCCATGATGGCTCAGACC R: CGGCTGTCCATAGCTTCTTC.

GABA-A α2 (Genbank Accession number: NM_000807,54-303 bp):

L:GCTGCAGTCTCGGTCTCTCT R: ATGTTAGCCAGCACCAACCT

4

GABA-A α3 (Genbank Accession number: NM_000808, 1441-1740 bp):

L: CCACCTATCCCATCAACCTG R: TTGCTGCACTGCCACTATCT

GABA-A γ2 (Genbank Accession number: X15376, 246-542 bp):

L: GGAGCACAGGAAGCTCAGTC R: CGTTCACTGGACCAATGCTA.

KCC2 (Genbank Accession number: AF208159, 4413-4918 bp):

L: TCCTCGCCAAAGACTGAAAT R: GTACCCAGTCCCAGATGGTG.

The template sequence was verified by sequencing or restriction mapping. The BDNF

template was obtained from Lein et al (Lein et al., 1999). S35-labeled riboprobes were

generated by in vitro transcription. After hybridization, sections were processed, dipped

with autoradiographic emulsion (Kodak NTB-2) and exposed for 3-6 weeks.

Expression levels were quantified by densitometry. Darkfield images were acquired with

a CCD camera and processed in MATLAB. Images were thresholded to remove

background signal and mean pixel luminance (total luminance/area) was measured

separately in all layers of the ablated and control regions. The borders between layers

were chosen according to adjacent cresyl violet stained sections (see Fig. 1D). The

threshold was computed as the 2x mean intensity outside the section. The threshold was

always lower than labeling intensity in all layers and the same threshold was applied in

ablated and control areas in each section (Fig. 1). The relative expression level in each

layer of ablated cortex was then computed relative to the same layer in unablated regions

in the same section.

5

Slice physiology

Slice preparation: Slices were obtained from animals (P25-P45) that received an IP

overdose of sodium pentobarbital (200 mg/kg to effect). A block of brain containing

visual cortex was removed rapidly and coronal slices (350 µm thick) were cut on a

vibratome in ice cold ACSF. ACSF contained (in mM): 130 NaCl, 3 KCl, 1.25 KH2PO4,

20 NaHCO3, 10 glucose, 1.3 MgSO4, 2.5 CaCl2 (pH 7.35-7.4, equilibrated with 95%O2-

5% CO2). Slices were incubated for at least 1 hour in ACSF at 30°C. Slices from the

subplate ablated hemisphere were collected within 5 mm of the injection site, which

roughly corresponds to the entire region in which subplate neurons have been ablated as

assessed anatomically (Ghosh and Shatz, 1992) or by in situ hybridization for BDNF

mRNA (Lein et al., 1999). The zone of ablation was confirmed in the slice by the

presence of fluorescent microspheres that had been coinjected at the time of ablation.

Slices from the unmanipulated hemisphere were also studied.

Perforated patch recordings:

Perforated patch recordings using gramicidin, which allows recording without disturbing

the internal Cl--concentration, were performed in slices from ablated or unmanipulated

hemisphere of 4 animals (n=30 cells in 30 subplate ablated slices; n=12 cells in 12

unmanipulated slices). Recordings were performed with a patch clamp amplifier in

voltage or current clamp using pipettes with input resistance of 4-8 MΩ. Electrodes were

filled with (in mM) 110 K-gluconate, 4 KCl, 4 NaCl, 0.2 CaCl2, 10 HEPES (free acid),

1.1 EGTA, 2 Mg-ATP, 1 MgCl2 and 5 glutathione (pH 7.2, 300 mOsm). Gramicidin from

frozen stock (20 mg/ml dissolved in DMSO) was added to the electrode solution to a

final concentration of 0.02 mg/ml. The solution was remade every 4 hours. Lucifer

6

Yellow was added to the electrode solution to monitor seal integrity and for post-hoc cell

identification (see Fig. 4D). For recording, slices were held in a chamber on a fixed stage

microscope (Zeiss Axioskop FS2) and superfused (2-4 ml/min) with ACSF containing (in

mM): 130 NaCl, 5 KCl, 10 HEPES (acid), 25 glucose, 1.3 MgCl2, 2.5 CaCl2 (pH adjusted

to 7.35-7.4 with NaOH, oxygenated with 100%O2 and 100 µM DL-AP5, 10 µM CNQX

and 1 µM tetrodotoxin (TTX). The calculated Cl-reversal potential (ECl) of these

solutions in the whole cell configuration was -78 mV. Layer 4 was identified in visually

by the distance from the pial surface (~600-800 µm). The location of neurons was

confirmed after recording by immunohistochemistry. Upon seal formation access

resistance was monitored until stable. Typical access resistances were 18-

70MΩ. Membrane voltages were corrected for the estimated liquid junction potential. To

measure the effect of muscimol on the cell, muscimol was bath-applied (100 µM) and the

evoked current from a holding potential close to the resting potential (~ -60mV) was

measured. The integrity of the patch was monitored and confirmed by fluorescent

imaging of the Lucifer Yellow. Then the patch was ruptured and the resulting whole-cell

current was measured.

Ca2+ imaging:

Slices were loaded with Fura2-AM (Molecular Probes) using a 2 step loading protocol

(Schwartz et al., 1998). Slices were incubated in the dark in 1.5mM Fura2-AM in

99.9%DMSO and 0.1% Pluronic (Molecular Probes) for 2 min. Then slices were

incubated in the dark for 2-4h in 10µM Fura2-AM, 1%DMSO and 0.001%Pluronic and

washed in the dark for 15-30 min in fresh ACSF. For imaging, slices were held in a

chamber on a fixed stage microscope (Zeiss Axioskop FS2) and superfused (2-4 ml/min)

7

with ACSF containing (in mM): 130 NaCl, 5 KCl, 10 HEPES (acid), 25 glucose, 1.3

MgCl2, 2.5 CaCl2 (pH adjusted to 7.35-7.4 with NaOH, oxygenated with 100%O2).

Glutamatergic transmission was blocked with 100 µM DL-2-amino-5-phosphonovalerate

(DL-AP5) and 10 µM 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX). In some

experiments 0.5-1 µM TTX was added to the solution to block voltage-dependent Na-

currents. No difference was observed in experiments with and without TTX. Slices were

illuminated via a Lamda 10 filterwheel and shutter (Sutter) with a Halogen (Zeiss) or

Xenon (Optiquip) Arc lamp at 340nm and 380nm. Emitted fluorescence was imaged with

40x objective via a 510 nm filter and a Sensicam HQ (Cooke Instruments) using a

PowerMacintosh G4 running IPlab (Scanalytics). Muscimol was either bath-applied (50-

100 µM) or pressure-applied focally (250 µM) with a patch pipette positioned on the

surface or up to 50µm above of the slice using a picospritzer (WPI, 4 psi 500 ms-1s). A

small holding vacuum was applied in the inter-puff intervals to avoid muscimol leakage.

During bath application of muscimol images at 340 nm and 380 nm illumination were

acquired every 30s, whereas during pressure application experiments images were

acquired continuously at 380 nm illumination. Analysis was performed offline with IPlab

and MATLAB (The Mathworks) using custom routines. Changes in fluorescence were

judged as significant if the absolute change (∆F) exceeded 2 standard deviations of the

baseline.

Quantitative PCR

mRNA was obtained from animals (P0-P35) that received an IP overdose of sodium

pentobarbital (200 mg/kg to effect). A block of visual cortex from one hemisphere was

removed and homogenized in Trizol (Gibco). The other hemisphere was frozen in

8

cryopreservative medium and processed for in situ hybridizations. Total RNA was

extracted by chlorophorm and isopropanol precipitation. cDNA was generated from 1 µg

RNA by reverse transcription using Retroscript (Ambion) with OligoDT and Random

decamer primers, or iScript (Biorad). The cDNA was used for quantitative PCR using a

Smart Cycler (Cephaid). Real-time PCR reaction was carried out on a Smart Cycler

system (Cepheid, Sunnyvale, CA). A reaction mix contained 1X iQ SYBR Green

Supermix (Bio-Rad Laboratories, Hercules, CA), 100nM each oligonucleotide primers

and 10ng of cDNA in a 25-ul total volume. As internal normalizing controls we used

glyceraldehyde-3-phosphate dehydrogenase (GAPDH), hypoxanthine

phosphoribosyltransferase 1 (HPRT1) and 18s RNA. Only 18s remained constant over

the observed developmental period; thus all the KCC2 and GABAα1 developmental data

was normalized to 18s. To compare tissue at the same age both HPRT1 and 18s were

used for normalization. Primers for cat RNA were generated to regions highly

homologous in human and mouse and were confirmed by gel electrophoresis.

Primers for cat mRNA:

KCC2 (Genbank Accession number: AF208159, 435-563bp):

L:CACGGCCATCTCCATGAGTG, R:GTGCCCAGGTAGAAGCAGAG.

18srRNA (Genbank Accession number: X00686, 878-1096 bp):

L:CGCGGTTCTATTTTGTTGGT, R:AGTCGGCATCGTTTATGGTC.

GAPDH (Genbank Accession number: AF097177, 17-201bp):

L:GAGTCAACGGATTTGGTCGT, R:GACAAGCTTCCCGTTCTCAG.

HPRT1 (Genbank Accession number: NM000194, 235-429bp):

L: TGCTCGAGATGTGATGAAGG, R:TCCCCTGTTGACTGGTCATT

9

GABAA α1 (Genbank Accession number: NM_010250: 1386-1607 bp):

L: CCCGTTCAGTGGTTGTAGCA, R:CTCTGTTGAGCCAGAAGGAGAC

50 cycles of the PCR reaction were performed with the Smart Cycler with an annealing

temperature of 62°C and annealing time of 30s. The number of cycles it took the

fluorescence signal to pass an arbitrary threshold of 30 was determined by the Smart

Cycler software (typically 18-26 cycles). Difference between the obtained threshold cycle

(CT) (the Smart Cycler Operator Manual for details) for KCC2 (∆KCC2), GABAA α1

(∆GABAA α1) etc. and the normalizing control were calculated (∆KCC2= CT KCC2- CT 18s

or ∆KCC2= CT KCC2- CT HPRT1). For developmental profiles, the differences were

normalized to the oldest age. Expression levels were calculated from ∆KCC2, ∆GABAA

α1 etc. as 2^(-∆KCC2) and 2^(-∆GABAA α1) respectively.

Transneuronal Labeling

To visualize OD columns under normal circumstances and following MD, 2 mCi of L-

[2,3,4,5-3H] proline (Amersham) in 50 µl 0.9% saline was injected into the vitreous

chamber of one eye as described previously (Ghosh and Shatz, 1992). For these

injections, animals (aged P50-P70) were placed under general anesthesia (described

above) and a drop of ophthalmic local anesthetic, proparacaine HCl, was placed on the

eye. After 10-14 days (axonal transport time), animals were given an IP overdose of

sodium pentobarbital (200 mg/kg to effect). Brains were removed, frozen rapidly and

cryostat sections (20 µm) of visual cortex were cut in horizontal plane. Sections were

fixed in 4% paraformaldehyde in PBS, dipped in autoradiographic emulsion (Kodak

NTB-2) and exposed for 4-12 weeks. Dark-field images of silver grains were acquired

10

and analyzed in MATLAB (The Mathworks). To quantify the extent of the territory

belonging to the deprived (DE) or the non-deprived (NDE) eyes, linescans were made

along layer 4 of visual cortex: A line along the center of layer 4 was generated by

selecting 20-100 points, and then performing a cubic spline interpolation between these

points. At every pixel along this line, a perpendicular line through layer 4 (~200 um long)

was computed and the average signal intensity of pixels along this line was measured.

The resulting intensity line scan was low pass filtered (7 pt triangular), generating a curve

of labeling signal intensity versus distance along layer 4 (see Fig. S2 middle panels). The

resulting signal showed periodic fluctuations corresponding to OD columns. To minimize

effects of variations in labeling intensity (due to non-uniform diffusion of 3H-proline

within the eye and uneven illumination of the section) the line scan was further low pass

filtered (100-200 pixels, ~2 mm) and subtracted from the original linescan. The resulting

signal fluctuated periodically around the zero-intensity axis, corresponding to OD

columns (see Fig. S2 lower panels). We used 2 methods to measure the territory occupied

by the labeled eye, which gave similar results. First, since labeled eye column borders

(Fig. S2) are identifiable as large increases in labeling intensities, column borders were

generally detected at the half maximum intensity. The width of each DE or NDE column

was determined as the distance between 2 adjacent borders (Fig. S2). The fraction of area

occupied by radioactive label belonging to the DE in ablated or unablated regions was

calculated as DEwidth/(DEwidth+NDEwidth), where DEwidth and NDEwidth are the sum of all

column widths for the respective eye. A second measure was used to quantify the fraction

the area occupied by the DE or NDE: the total area occupied by radioactive label

belonging to the injected eye in which the line-scan has an intensity value above zero was

11

measured. The percent area occupied by the injected eye was computed as the fraction of

the line scan above zero relative to the total length of the line scan. The 2 methods gave

qualitatively similar results. A Student's T-test was performed to evaluate significance of

the changes.

Computational Modeling:

The subplate neuron (SPN) and the layer 4 neuron (L4N) were represented by "integrate

and fire neurons". Thus the membrane potential vSPN and vL4N at each time step t was

computed as:

vSPN(t)=ws*geye1(t)+ ws*geye2(t) (1)

vL4N(t)=weye1* geye1(t)+weye2* geye2(t) +wSPN*gSPN(t)+wspont+gspont (t)

(2)

geye1, geye2and gSPN are the synaptic conductances from thalamus and SPN respectively

and are of the form:

g(t)=exp(-(t-tspike+tdelay)/τsyn) (3)

tspike is the time of occurrence of the presynaptic spike and tdelay is the synaptic delay of 1

ms. τsyn gives the synaptic time constant and was set to 10 ms. weye1, weye2 and wSPN

denote the synaptic weights ("epsc amplitude"). The thalamic inputs of each eye to the

SPN (ws) were fixed at 0.5, thus there was no OD bias present in the SPN.

If vSPN or vL4N reached a threshold (0.4), a spike was generated and v was decreased by

vdelta(t) leading to a refractory period.

vdelta(t)=beta*exp(-(t-tspike)/trefr) (4)

12

To ensure model stability, vSPN and vL4N decayed slowly even if no spikes occurred. This

is equivalent to the existence of a leak current. Beta and trefr and the leak were adjusted to

10, 2 and 10%/ms respectively so that the cells fired with an average firing rate of ~1 Hz

(Kara and Reid, 2003; Kara et al., 2000).

If spikes occurred in the L4N, the synaptic weights were readjusted according to the

plasticity rule (see Fig. 5A), giving a plasticity factor ∆w(dt) as a function of the delay dt

(dt=tspike-tepsc) between the occurrence of the postsynaptic spike and the synaptic input.

w(t+dt)=w(t)*(1+∆w) (5)

The parameters of the learning rule were τS=3ms τW=20 ms. Parametric simulation

of 4900 different parameter sets (amplitude from 3*10^-3 to 21*10^-3 and τ from 1 to 30

ms) were performed for 4 different conditions (+subplate, -subplate, +subplate & MD, -

subplate & MD. 384/4900 parameter set replicated the experimental data. A parameter set

was judged as replicating the data when the results in all 4 conditions fulfilled the

following criteria:

1) +subplate: strengthening of thalamic inputs and weakening of subplate input (eye1 or

eye2 >> SP)

2) -subplate: no strengthening of eye1 or eye2 over initial strength (eye1 and eye 2 < 0.2)

3) +subplate&MD: strengthening of the open eye and weakening of both subplate inputs

and deprived eye (NDE>>DE, NDE>>SP<0.2)

4) -subplate&MD: No strengthening of either eye and paradoxical shift (DE, ND < 0.2,

DE >> NDE)

13

Synaptic weights were free to vary from their initial value between a range of 0 to 1. The

model was driven from its initial state by thalamic activity, which was simulated as

uncorrelated Poisson processes. The average LGN spike rates were 5-10 Hz if the eye

was open (Dan et al., 1996), and 0.3-1 Hz if the eye was closed. The spontaneous EPSC

inputs to the L4N was simulated as a 3rd uncorrelated Poisson process with a rate of 0.5

Hz (Kanold et al., 2003) and a weight of 0.7. Varying the absolute values of the firing

rates did not result in different results as long as the firing rates of the open eye were

larger than that of the closed eye and that of the spontaneous activity. All data presented

was simulated with rates of 5 Hz for the open eye and 0.3 Hz for the closed eye.

Physiologically OD is frequently measured on a discrete scale from 1 to 7 where a value

of 1 and 7 indicate monocular responses to contralateral and ipsilateral stimulation

respectively. A value of 4 indicates equal responses from the 2 eyes. Initial biases in OD

in layer 4 is present at early development (Albus and Wolf, 1984; Hubel and Wiesel,

1963; LeVay et al., 1978). This bias is between OD group 3 and 4 (with an estimated

mean 3.2) (Hubel and Wiesel, 1963; LeVay et al., 1978), thus slightly favoring the

contralateral eye. Here we measure OD on a finer scale using an OD bias index (OD bias

= [Ipsi-Contra]/[Ipsi +Contra]) that varies continuously from -1 to 1. A lack of OD bias

will be indicated by a value of 0, whereas monocular responses will have values of -1 and

1 respectively. The physiologically measured OD values can roughly be translated into

OD bias by calculating OD bias=[OD-4]/3, thus an OD of 3 is roughly equivalent of an

OD bias -0.33. We simulated initial contralateral biases within (initial bias between 0 and

-0.27) and outside the physiological range (initial bias < -0.27). An initial bias towards

14

the DE or NDE was simulated by reducing the initial weight of one input and increasing

the other in order to keep total synaptic input constant. The model was implemented in

C++ on a PowerMac G5. To generate the correlation plots we selected a fixed number of

layer 4 spikes and computed the time differences to spikes of the respective input in a

time window of ±50ms.

Figure S1.

15

Figure S2.

16

References Albus, K., and Wolf, W. (1984). Early post-natal development of neuronal function in the kitten's visual cortex: a laminar analysis. J Physiol 348, 153-185. Bear, M. F., Kleinschmidt, A., Gu, Q. A., and Singer, W. (1990). Disruption of experience-dependent synaptic modifications in striate cortex by infusion of an NMDA receptor antagonist. J Neurosci 10, 909-925. Dan, Y., Atick, J. J., and Reid, R. C. (1996). Efficient coding of natural scenes in the lateral geniculate nucleus: experimental test of a computational theory. J Neurosci 16, 3351-3362. Ghosh, A., and Shatz, C. J. (1992). Involvement of subplate neurons in the formation of ocular dominance columns. Science 255, 1441-1443. Ghosh, A., and Shatz, C. J. (1994). Segregation of geniculocortical afferents during the critical period: a role for subplate neurons. J Neurosci 14, 3862-3880. Hubel, D. H., and Wiesel, T. N. (1963). Receptive Fields of Cells in Striate Cortex of Very Young, Visually Inexperienced Kittens. J Neurophysiol 26, 994-1002. Kanold, P. O., Kara, P., Reid, R. C., and Shatz, C. J. (2003). Role of subplate neurons in functional maturation of visual cortical columns. Science 301, 521-525. Kara, P., and Reid, R. C. (2003). Efficacy of retinal spikes in driving cortical responses. J Neurosci 23, 8547-8557. Kara, P., Reinagel, P., and Reid, R. C. (2000). Low response variability in simultaneously recorded retinal, thalamic, and cortical neurons. Neuron 27, 635-646. Lein, E. S., Finney, E. M., McQuillen, P. S., and Shatz, C. J. (1999). Subplate neuron ablation alters neurotrophin expression and ocular dominance column formation. Proc Natl Acad Sci U S A 96, 13491-13495. LeVay, S., Stryker, M. P., and Shatz, C. J. (1978). Ocular dominance columns and their development in layer IV of the cat's visual cortex: a quantitative study. J Comp Neurol 179, 223-244. Schwartz, T. H., Rabinowitz, D., Unni, V., Kumar, V. S., Smetters, D. K., Tsiola, A., and Yuste, R. (1998). Networks of coactive neurons in developing layer 1. Neuron 20, 541-552.