|Home | About | Journals | Submit | Contact Us | Français|
Ocular dominance plasticity (ODP) following monocular deprivation (MD) is a model of activity dependent neural plasticity that is restricted to an early critical period regulated by maturation of inhibition. Unique developmental plasticity mechanisms may improve outcomes following early brain injury. Our objective was to determine the effects of neonatal cerebral hypoxia ischemia (HI) on ODP. The rationale extends from observations that neonatal HI results in death of subplate neurons, a transient population known to influence development of inhibition. In rodents subjected to neonatal HI and controls, maps of visual response were derived from optical imaging during the critical period for ODP and changes in the balance of eye-specific response following MD were measured. In controls, MD results in a shift of the ocular dominance index (ODI) from a baseline of 0.15 to −0.10 (P<0.001). Neonatal HI with moderate cortical injury impairs this shift, ODI = 0.14 (P<0.01). Plasticity was intact in animals with mild injury and in those exposed to hypoxia alone. Neonatal HI resulted in decreased parvalbumin expression in hemispheres receiving HI compared with hypoxia alone: 23.4 vs. 35.0 cells/high power field (P=0.01), with no change in other markers of inhibitory or excitatory neurons. Despite abnormal inhibitory neuron phenotype, spontaneous activity of single units and development of orientation selective responses were intact following neonatal HI, while overall visual responses were reduced. Our data suggest that specific plasticity mechanisms are impaired following early brain injury and that the impairment is associated with altered inhibitory neuronal development and cortical activation.
The immature brain exhibits a unique capacity for large-scale anatomical rewiring of connections in response to changes in neuronal activity, a form of plasticity only observed during critical periods in development(Hensch, 2005). Such plasticity mechanisms have potential to contribute to recovery from early brain injury. Despite the existence of these forms of plasticity, neonatal brain injury often results in permanent, pervasive neurodevelopmental impairments(Ferriero, 2004). Recent evidence highlights a role for maturation of cortical inhibition in controlling the onset of critical period plasticity(Hensch and Fagiolini, 2005). During critical periods, sensory-driven neuronal activity refines cortical inputs into precise maps that reflect sensory experience. Prior to the critical period, immature circuits incorporate a transient neocortical cell population, subplate neurons, and are driven by spontaneous patterned activity(Feller and Scanziani, 2005). In the visual system, this endogenous activity arises in the retina in waves that are required for the development of precise topographic maps in visual cortex(Cang et al., 2005b; Cang et al., 2008). During this precritical period of cortical development(Feller and Scanziani, 2005), maturation of cortical inhibition is dependent upon the presence of subplate neurons(Kanold and Shatz, 2006) and is influenced by neuronal activity(Morales et al., 2002).
Subplate neurons are selectively vulnerable to neonatal hypoxia ischemia (HI)(McQuillen et al., 2003). In this study, we investigate the effect of neonatal HI and consequent subplate neuron death during the precritical period on the development of visual cortical maps, cortical plasticity and inhibitory neurons. This study focuses on one of the best-described models of cortical plasticity – visual system ocular dominance plasticity, in the setting of a standard translational rodent model of neonatal cerebral hypoxic-ischemic injury.
Timed pregnant Long Evans rats (Simonsen, Gilroy, CA) were allowed food and water ad libitum. Animals were placed together for less than 6 hours in the AM with day of breeding designated E0. All animal research was approved by the University of California San Francisco Committee on Animal Research and carried out in accordance with standards for humane animal treatment as outlined in the Policy on Humane Care and Use of Laboratory Animals.
The Rice-Vannucci procedure was performed at postnatal day two (P2) as described(McQuillen et al., 2003). Pups were anesthetized with isoflurane and the right common carotid artery was electro-cauterized. Pups were placed back with the mother for one hour to recover from anesthesia and then placed in chambers temperature-controlled by a circulating water bath to maintain a chamber temperature of 34C and a rat skin surface temperature of 36.5–37C. Chambers received a flow of 5.6% oxygen gas at a rate of 3 liters per minute. Duration of hypoxia ranged between 2 – 2.5 hours with the hypoxia period terminated if any animal appeared premorbid or the mortality for any individual litter reached 20%.
Animals were sacrificed at weekly intervals or raised to P32 for optical imaging followed by perfusion fixation, sectioning and histology. Injury severity was defined categorically (mild, moderate or severe) as described(McQuillen et al., 2003) based upon postmortem histology of coronal sections stained with cresyl violet. Specifically, injury is graded as: (1) mild - asymmetry of the lateral ventricles with minimal loss of cells in the lower neocortical layers (Figure 1A, D), (2) moderate – asymmetry with dilatation of the lateral ventricles, significant thinning of subplate (VIb) with more severe injury in entorhinal cortex (Figure 1B asterisk, D) or (3) severe – loss of cells in layer VIa and subplate with clear cyst formation laterally (Figure 1C asterisk, D). Histological injury grade was determined for all immunohistochemistry, optical imaging and physiology experiments. Because histological assessment of injury is impossible in tissue prepared for immunoblots, animals were preselected for moderate injury 24 hours after HI using magnetic resonance imaging (MRI) as described(Wendland et al., 2008). Diffusion-weighted MRI at twenty-four hours following the Rice-Vannucci procedure reliably predicts caspase-3 activation, cell death and histological grades of injury at seven days(Wendland et al., 2008). Animals with mild injury have no detectable changes on MRI (Figure 1A). Moderate injury produces enlargement of the lateral ventricle visible as a hypointense band below neocortex (Figure 1B). Severe injury is detected by enlargement of the lateral ventricle with diffusion hyperintensity of overlying cortex (Figure 1C). To provide an alternate, continuous measure of injury, the thickness of primary visual cortex surface (−5.88mm with respect to bregma) was measured from the bottom of the cortical plate to the pial surface in one-month old animals (Figure 1A–C, see boxed area for location of measurement). Mean±standard deviation area V1 cortical thickness measures 1332±17 μm in unmanipulated animals. Following HI, V1 cortical thickness measurements by injury severity category were: 1223±83 μm (mild), 1118±96 μm (moderate) and 857± 172 μm (severe).
Rats received monocular deprivation (MD) to the contralateral eye (left) of the hemisphere subject to HI. MD lasted for 4 to 7 days during the peak of the critical period (P24-P36). Rats were anesthetized with isoflurane and eyelids were trimmed and sutured together with three mattress stitches. The eye was reopened and cleaned prior to imaging.
Rapid imaging of the intrinsic signal of cortical activity in response to a continuous period stimulation was performed using minor modifications of a protocol described for mice(Kalatsky and Stryker, 2003). Surgery: To prepare the animal for imaging, isoflurane anesthesia was used at induction then one of two anesthesia regimens was used: (1) Urethane (intraperitoneal 25% w/v urethane 0.5 ml/100gm body weight); (2) Pentobarbital/buprenorphine (intraperitoneal pentobarbital 2mg/100gm and intramuscular buprenorphine hydrochloride 0.02mg/kg). Both regimens included a pre-anesthetic dose of intramuscular chlorprothixene (0.2 mg/100g), subcutaneous atropine (5 mg/kg) and dexamethasone (0.2 mg/100g). Body temperature was maintained at 37.5°C and heart rate was monitored continuously. Supplemental intraperitoneal urethane (0.1 ml/100g) or pentobarbital doses (1mg/100g) were administered as required. A tracheotomy was performed and oxygen supplied to 1.5mm diameter glass tracheostomy tube. Animals were placed in a stereotaxic frame using ear bars and a craniotomy was performed to expose V1. Agarose 3% w/v and a glass cover slip were placed over the brain to prevent drying and to optimize optics during imaging. Agarose was sealed with liquid paper. Imaging: Temporally encoded intrinsic signal maps were recorded as described(Kalatsky and Stryker, 2003) and used to determine ocular dominance plasticity(Cang et al., 2005a). A high refresh rate monitor was placed 25cm away from eyes of animal for periodic stimulus presentation. A Dalsa 1M30 CCD camera (Dalsa, Waterloo, Canada) was controlled by custom software to record intrinsic signals from V1 that was illuminated with red light (610 ± 10 nm). Cortical surface images were taken using green light (546 ± 10 nm). The camera was focused down 600μm from pial surface before stimulus experiments. Periodic stimuli in the elevation directions (90° and 270°) were presented and averaged for both ipsilateral and contralateral eye individually to measure V1 response magnitude for each. The ipsilateral response map was used to template the region of interest when comparing response magnitudes for each eye. Response magnitudes were compared using Matlab to determine an ocular dominance index (ODI). An ODI is equal to the magnitude of response to contra (C) eye stimulation minus magnitude of response to ipsi (I) eye stimulation, over the summation of magnitude of responses [ODI = (C − I)/ (C + I)]. Nasotemporal (azimuth) maps were also recorded in most animals (0° and 180°) but were not used for ODI calculations. Retinotopic map scatter, an index of map quality(Cang et al., 2005b) was determined using Matlab.
Animals were euthanized with intraperitoneal pentobarbital (50 mg/kg) and transcardially perfusion-fixed with 4% w/v paraformaldehyde in 0.1M phosphate buffer. Brains were post-fixed overnight and then allowed to sink in 25% w/v sucrose 0.1 M PBS. Brains were cut on a freezing stage microtome at a width of 50 μm and stored in PBS and sodium azide at 4°C or in glycerol based cryoprotectant at −20°C. Peroxidase-DAB Immunohistochemistry: Sections were quenched in a 3% v/v hydrogen peroxide solution then blocked in 10% v/v Donkey serum with 0.1% v/v Triton X-100 in PBS. Sections were incubated overnight in mouse anti-parvalbumin primary antibodies (Sigma-Aldrich, Saint Louis, MO) at a 1:1000 dilution at 4°C. Sections were washed with 0.1M PBS with 0.1% v/v Tween-20 and then incubated in biotinylated donkey anti-mouse secondary antibodies (Jackson ImmunoResearch, West Grove PA) at 1:500 dilution for one hour at room temperature. Sections were incubated in ABC solution (Vector Labs, Burlingame, CA) and then stained in DAB solution with 1.2% w/v nickel ammonium sulfate per manufacturer’s instructions. Wisteria floribunda Lectin (WFA): Sections were quenched in 3% peroxide solution v/v and then blocked in 3% w/v BSA, 20mM Lysine, and 0.2% v/v Triton X-100 in PBS. Sections were incubated in 20μg/ml of lectin from Wisteria floribunda (Sigma-Aldrich, Saint Louis, MO, L-1766) overnight at 4°C. Sections were incubated in ABC solution and then stained in DAB solution per manufacturer’s instructions. Washes in PBS 0.1% Tween-20 were carried out between incubations. Immunofluorescence (IF): Sections were incubated in mouse primary antibodies (anti-parvalbumin, Sigma-Aldrich, P3088, 1:1000, anti-calbindin, Swant, CB38, 1:2000, anti-calretinin, Swant, 7699/4, 1:2000) and/or 20μg/ml of lectin from Wisteria floribunda (Sigma-Aldrich, L-1766) overnight at 4°C. Sections were washed and then incubated in anti-mouse and anti-rabbit florescent secondary antibody (1:500) (Invitrogen, Carlsbad, CA) and florescent avidin for lectin (1:500) (Invitrogen, Avidin Oregon Green). Sections received bisbenzamide for a nuclear counter-stain. Sections were mounted and cover-slipped with Vectashield mounting medium (Vector Labs)
Digital images of coronal sections stained for parvalbumin, calretinin, calbindin and WFA were quantified using the Multi Wavelength Cell Scoring tool in Metamorph 7.0 (Molecular Devices, Downington, PA) to determine single- and double-labeled cell density. Data was collected for neocortex at three points in the brain, −0.36mm (S1), −4.44mm (V1), −5.88mm (V1) with respect to bregma (The Rat Brain in Stereotaxic Coordinates 5th Edition, Paxinos et al., Elsevier Academic Press, 2005). Mean cell density for each location was averaged and compared between hypoxia alone and HI hemispheres.
To confirm injury, protein extracts were analyzed for glial fibrillary acidic protein (GFAP) expression, a marker of injury and astrogliosis that persists for weeks in the HI hemisphere following neonatal Rice-Vannucci procedure(Sizonenko et al., 2008). GFAP is only increased in HI hemispheres (actin-normalized optical density (O.D.) units±S.D. in HI = 1.63±0.04 vs. H = 1.37±0.05, P=0.003; vs. control = 1.34±0.09, P=0.009). Proteins were extracted from rat cortices and harvested with appropriate volume of cell lysis buffer (1X PBS, 1% NP-40, 0.5% Sodium deoxycholate, 0.2% SDS, 1X Complete Mini Protease Cocktail [Roche]). Extracts were spun at 13,000 RPM for 20 minutes at 4C, and the supernatant containing the proteins was collected. Protein content was estimated with a BCA protein assay kit (Pierce, Rockford, IL) using a BSA-based standard curve. Equal amounts of protein were analyzed by SDS-PAGE and transferred to a Polyvinylidene Difluoride (PVDF) membrane using rapid semidry immunoblotting techniques (Bio-Rad, Hercules, CA). PVDF blots (Bio-Rad) were blocked using Superblock T20 Blocking Buffer (Thermo Scientific, Rockford, IL) for one hour at RT, then washed 3 times for 10 minutes each in TBS with 0.05% Tween 20. Blots were then incubated in Superblock T20 with appropriate dilution of primary antibody overnight at 4°C, including anti-GFAP (Sigma-Aldrich) at 4 μg/ml, anti-CaM Kinase IIα (Affinity BioReagents, Golden, CO) at 0.75 μg/ml, anti-PSD-95 (Affinity Bioreagents) at 0.5μg/ml, anti-GAD65/67 (Chemicon, Temecula, CA) at 0.5 μg/ml, anti-NR2A (Upstate, Temecula, CA) at 0.5 μg/ml, anti-VGLUT1 at 1:50,000 dilution and anti-VGAT at 1:10,000 (Synaptic Systems, Goettingen, Germany). After washing 3 times 10 minutes in TBS with 0.05% Tween 20, blots were incubated for 2 hr at RT with HRP-conjugated secondary antibody, including 0.1 μg/ml goat anti-rabbit (CalBioChem, San Diego, CA) for anti-GAD65/67 anti-VGLUT1, anti-VGAT and anti-NR2A, and 0.1μg/ml goat anti-mouse (CalBioChem) for anti-CaM Kinase IIα and anti-PSD-95. Blots were visualized using the SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific), and captured on autoradiographic films (GE Healthcare, Buckinghamshire, UK). For loading controls, all blots were incubated with anti-β-actin (Sigma, St. Louis, MO) at 0.2 μg/ml overnight at 4C, followed by incubation with HRP-conjugated goat anti-mouse as described for other antibodies. Films were digitized with an Epson Perfection 4490 Photo scanner, and band optical densities (OD) were measured using ImageJ version 1.40g (NIH, USA). For each sample, we calculated the normalized ratio of protein to β-actin values, and these numbers were used to calculate the mean ± S.D..
Single unit recordings were made using experimental and analytical methods described(Niell and Stryker, 2008). Rats were anesthetized using the urethane regiment described above, secured in ear bars, and a small craniotomy was created above V1. A silicon 16-channel linear multielectrode (Neuronexus Technologies, model a1×16–3mm50–177) was inserted and allowed to stabilize before recording. Signals were acquired on a System 3 workstation (Tucker-Davis Technologies) and analyzed using Matlab. Single units were identified by clustering spike waveforms using the FastICA Matlab package and KlustaKwik (Harris et al., 2000). Units were classified as broad or narrow spiking based on trough-to-peak time. Of 106 successfully isolated units, 5 were narrow spiking and were excluded from subsequent analysis. Visual stimuli were generated using the Psychophysics Toolbox Matlab extensions (Brainard, 1997; Pelli, 1997) and included contrast-reversing checkerboards, drifting sinusoidal gratings and contrast modulated Gaussian noise movies. Responses to drifting gratings were used to construct an orientation tuning curve at the spatial frequency giving the maximal response. From this, we calculated the width of the tuned component and an orientation selectivity index (OSI), the ratio of the tuned versus unturned component of the response.
Data that was normally distributed was initially analyzed graphically with scatter or box plots and summarized by mean and standard deviation. Population difference testing was performed with unpaired, two-sided Student’s T-test or one-way ANOVA (normally distributed data) or Mann-Whitney U test. Multiple comparisons were corrected using Bonferroni method. Association was tested by correlation or linear regression. Difference in cumulative probability was analyzed using Kolmogorov-Smirnov test.
The response to visual stimulation was determined using optical imaging in rats that had received neonatal HI (HI, N=27, Rice-Vannucci model), unmanipulated littermates (N=21) or hypoxia alone (N=4). The Rice-Vannucci model of neonatal HI performed at early ages (P0-2) produces a range of injury in neocortex(Towfighi et al., 1997) that correlates with the extent of subplate neuron death and can be evaluated from routine histology(McQuillen et al., 2003). The model does not produce detectable injury to hippocampus or striatum. Thalamic injury is limited to the reticular nucleus of the thalamus(McQuillen et al., 2003).
In prior studies, we have shown with retrograde lipophilic dye labeling that visual thalamocortical connections persist following neonatal HI(McQuillen et al., 2003). These methods however are not sufficient to reveal organization or activation of visual cortex. Therefore we derived functional retinotopic maps of elevation (Figure 2A control phase map) and azimuth orientation (not shown) using an optimized method of intrinsic signal optical imaging(Kalatsky and Stryker, 2003). The method employs a continuous, temporally periodic stimulus consisting of drifting horizontal (Figure 2B) or vertical bars presented sequentially in both directions. The elevation map was used to determine the magnitude of eye-specific response(Cang et al., 2005a). Rodents typically show a stronger response to stimulation of the contralateral eye (Figure 2A control contralateral vs. ipsilateral response maps). Animals were raised to the peak of the critical period for ocular dominance plasticity (P28-32)(Fagiolini et al., 1994) before imaging (mean±S.D. age at imaging 32.9±3.5 days). Following imaging, animals that were subject to HI were sacrificed and the severity of brain injury was graded. A minority of animals (N=5/27, 19%) had developed severe neocortical injury with cyst formation of the lateral parietal-occipital lobe and these animals either lacked detectable response to visual stimulation (N=2) or had very weak response (N=3). These animals were not analyzed further. All other animals (N=22) had either mild injury (N=10/27, 37%) or moderate (N=12/27, 44%) neocortical injury. The appearance of retinotopic maps was unaltered by neonatal HI (Figure 2A HI phase map) or hypoxia alone (N=2, data not shown). Neonatal HI did not alter the expected contralateral bias of response strength at baseline (Figure 2A HI contralateral vs. ipsilateral response maps).
A subset of both HI-treated (N=17) and control (N=9) animals received monocular deprivation (MD) by eyelid suture. MD did not alter the appearance of retinotopic maps (Figure 2A monocular deprivation phase map). Brief MD during the critical period shifts the response strength in favor of the undeprived (ipsilateral) eye, representing ocular dominance plasticity (Figure 2A, monocular deprivation contralateral vs. ipsilateral response maps). Neither HI, nor MD significantly alters the area of cortex that responds to visual stimulation (P=0.7, one-way ANOVA, Figure 2C). Quality of the retinotopic maps can be assessed by smooth progression of the location of visual response to drifting bars(Cang et al., 2005b). Increasing scatter of the response location indicates degradation of the map. MD, but not HI, increases elevation map scatter (P=0.005, ANOVA). This result is attributable to the expected decrease in response strength of the deprived eye. Azimuth map scatter tended to be worse in animals subjected to HI but this difference was not significant (Figure 2E). Azimuth map quality appeared worse in general in this study employing rats in comparison with prior studies in mice(Cang et al., 2005b).
Having confirmed that neonatal cerebral HI with mild/moderate damage does not prevent the development of normal visual cortical response magnitude and organization, we next assessed the effect of neonatal hypoxia or HI on ocular dominance plasticity. To compare the change in eye-specific response following MD (Figure 3A), we derived an ocular dominance index (ODI, see methods) as described(Cang et al., 2005a) for each condition (Figure 3C). A response driven entirely by the contralateral eye would result in an ODI of 1, while a response restricted to the ipsilateral eye would result in an ODI of −1. In unmanipulated control animals at baseline, the mean±S.D. ODI is 0.15±0.07 reflecting the response bias toward contralateral eye input. Neonatal HI does not significantly alter the ODI prior to induction of plasticity by MD (0.24±0.1, P=1.0, one-way ANOVA with Bonferroni correction). In littermate control animals without neonatal HI, MD of 5.9±1.6 days was performed at a mean±S.D. age of 28.3±3.0 days, resulting in a shift in response in favor of the non-deprived (ipsilateral) eye (Figure 3A) and a significant decrease in the ODI (−0.10±0.12, P=0.002, control-baseline vs. control-MD, one-way ANOVA with Bonferroni correction). Animals receiving neonatal hypoxia alone followed by MD displayed a shift in the ODI in favor of the non-deprived eye of similar magnitude to control-MD littermates (−.09±0.07, P=1.0, one-way ANOVA with Bonferroni correction). In contrast, in animals exposed to neonatal HI resulting in moderate injury, MD failed to induce a shift in the response toward the open eye (Figure 3B), with a mean±S.D. ODI (0.14±0.13) that was not different from control baseline (no HI, no MD) and was significantly different from the shift seen following MD in control animals not exposed to HI (P<0.001, control-MD vs. HI-MD, one-way ANOVA with Bonferroni correction) and no different from control-baseline (no HI, no MD) animals (P=1.0, control vs. HI-MD, one-way ANOVA with Bonferroni correction). The duration of MD was similar in the two groups (5.3±1.7 vs. 6.1±1.5 days, control-MD vs. HI-MD, P=0.30, Student’s T-test). The blockade of cortical plasticity was present only in animals with moderate brain injury and was thus not a non-specific result of the neonatal HI procedure, as animals treated with HI that developed undetectable or mild injury have a shift in the ODI of similar magnitude to deprived animals not exposed to neonatal HI (−0.20±0.07, P=0.48, control-MD vs. HI-MD no injury, one-way ANOVA with Bonferroni correction). Imaging results did not differ between anesthesia regimens (urethane vs. pentobarbital, Figure 3C).
As an alternative to the categorical grading of injury, we analyzed the association between cortical thickness in area V1 as a continuous measure of injury and impaired cortical plasticity. Cortical thickness is negatively associated with ODI following MD (β=−0.001, P=0.03, 95% confidence interval −0.002 – −0.00001, linear regression). Analyzing the difference in thickness between contralateral (hypoxia alone) and ipsilateral (HI) hemispheres identifies an even stronger association between injury and ODI following MD (β=0.003, P<0.001, 95% confidence interval 0.002 – 0.005, linear regression). Impaired ocular dominance plasticity is not the result of a grossly delayed critical period, as brief MD (4–5 days) in mature animals (P60) exposed to neonatal HI did not result in a shift in the ODI toward the open eye (N=3, mean±S.D. ODI 0.20±0.03).
Ocular dominance plasticity is known to be dependent upon the normal development of cortical inhibitory neurons (reviewed in (Hensch and Fagiolini, 2005)) and inhibitory neuron development in turn is affected by premature loss of subplate neurons(Kanold and Shatz, 2006). As neonatal HI is associated with subplate neuron loss(McQuillen et al., 2003), we next examined the developmental expression of parvalbumin as a marker of one class of inhibitory interneurons (fast-spiking, basket) to determine if the timing or extent of parvalbumin expression was altered by neonatal HI. Parvalbumin is not expressed at high levels in the first postnatal week (Figure 4A – P7 control). By the second postnatal week, parvalbumin is expressed in all cortical layers, with prominent expression in layer V neuronal dendrites (Figure 4A – P14, control). After the third postnatal week, parvalbumin immunoreactive cell bodies are found in all cortical layers, with increased staining in the neuropil at older ages (Figure 4A – P21, P28, P60). Neonatal hypoxia alone does not alter the development of parvalbumin immunoreactivity (Figure 4A – hypoxia). Although neonatal HI does not alter the onset of parvalbumin expression, the magnitude of parvalbumin expression is decreased and is notably diminished in layer V dendrites at P14 (Figure 4 – P14, HI). At later ages the number of parvalbumin expressing cell bodies appears reduced suggesting a possible loss of inhibitory neurons, however neuropil staining is similar or more intense (Figure 4A – P21, P28, HI). Parvalbumin expression does not recover by immunohistochemistry after the critical period in mature animals exposed to neonatal HI (Figure 4A – HI P60 HI).
Parvalbumin expression is regulated by neuronal activity(Jiao et al., 2006) raising the possibility that decreased parvalbumin-expressing cell bodies could represent either loss of cells or decreased expression of the marker. To investigate this possibility, we examined the development of perineuronal nets (PNNs) visualized by binding of the lectin wisteria floribunda agglutinin (WFA). WFA-labeled PNNs are known to surround parvalbumin-expressing inhibitory neurons(Hartig et al., 1992; Brauer et al., 1993), although a minority are also found surrounding excitatory neurons(Alpar et al., 2006). Like parvalbumin, WFA-labeled PNNs are also not observed in rat neocortex in the first postnatal week (Figure 4B – P7, control). PNNs begin to form by P14 but do not reach the density seen in adult neocortex until P21. Neonatal HI results in the appearance of scattered cells with the appearance of activated microglia that are detected by WFA-lectin staining only in the first postnatal week (Figure 4B – P7 HI). Neither neonatal hypoxia-alone nor HI alters the timing of PNN development in cortex (Figure 4B – P7-28, hypoxia, HI). In contrast to parvalbumin, PNN expression is not diminished following neonatal HI.
Neocortical pyramidal neurons and excitatory synapses exhibit characteristic changes in the expression of proteins involved in glutamatergic neurotransmission during development(Petralia et al., 2005) and in response to neural activity(Cotrufo et al., 2003). These changes include switching of NMDA receptor subunits and alterations in the expression of presynaptic transporters and postsynaptic density (PSD) proteins. To explore whether the development of excitatory neurons and/or glutamatergic neurotransmission was altered by neonatal HI, we analyzed expression by immunoblot of Type II calcium/calmodulin-dependent protein kinase α (CaMKIIα), NMDA receptor 2A (NR2A), PSD-95 and vesicular glutamate transporter 1 (VGluT1) at two and four weeks of age in unmanipulated animals and following hypoxia or HI (Figure 5A). Consistent with prior reports(Petralia et al., 2005), expression of CaMKIIα, NR2A and PSD-95 all increase from two to four weeks of age in unmanipulated animals. Following hypoxia alone or HI, CaMKIIα, NR2A and VGluT1 levels do not differ from unmanipulated littermate controls, showing normal developmental upregulation. The only significant difference observed is an increase in PSD-95 expression at four weeks in both hypoxia and HI exposed hemispheres compared with unmanipulated control animals (actin-normalized optical density (O.D.) units ± S.D. in control = 0.83±0.12 vs. H = 1.37±0.17, P=0.01; vs. HI = 1.18±0.18, P=0.028). Unlike parvalbumin expression and ocular dominance plasticity, which are selectively decreased following HI but not hypoxia (Figure 3C, ,4A),4A), the elevation of PSD-95 expression is observed in both hypoxia and HI hemispheres, making this change unlikely to account for impaired plasticity observed only following HI.
GABA biosynthetic enzymes and reuptake transporters are also developmentally regulated and sensitive to changes in activity(Cotrufo et al., 2003; De Gois et al., 2005). To examine the effect of neonatal HI on these proteins, we measured levels of the 65 kDa and 67 kDa isoforms of glutamic acid decarboxylase (GAD65, GAD67) and vesicular GABA transporter (VGAT) by immunoblot (Figure 5B). In unmanipulated littermates, expression of GAD65, GAD67 and VGAT increase from two to four postnatal weeks. Neither hypoxia alone, nor HI alters levels of these proteins at two weeks. By four weeks, there is a trend to decreased expression of GAD65 and GAD67 in HI hemispheres that does not reach significance.
In order to determine whether parvalbumin-expressing cells were lost following neonatal HI or whether the reduction in parvalbumin immunoreactivity simply reflected decreased expression, we combined fluorescent detection of parvalbumin and WFA-labeled PNNs and quantified the density of cells with the two labels between the hemisphere receiving HI and the hemisphere exposed to hypoxia alone. Single channel images reproduce the findings of peroxidase histochemistry (Figure 4A,B): parvalbumin, but not WFA, expression is decreased in the hemisphere receiving HI (Figure 6A). Although the overall density of WFA-expressing cells did not change, examination of the merged image reveals qualitatively more single-positive WFA-expressing cells (Figure 6A, merge arrows). To quantify this observation, measurements were made in neocortex at three matched rostral caudal levels from parietal to occipital cortex and averaged (see methods). Individual cells were identified by the bisbenzamide nuclear counter stain and determined to be expressing neither parvalbumin nor WFA, a single marker (parvalbumin or WFA) or both markers using an automated algorithm and Metamorph software (see methods). Consistent with limited neocortical damage following P2 HI, bisbenzamide cell density was similar between hemispheres (1115±78 vs. 1144±90 cells/high power field (HPF), P=0.46 Student’s T-test, Figure 6C). Parvalbumin-expressing cell density was significantly reduced in the HI hemisphere compared with hypoxia alone (23.4±2.9 vs. 35.0±7.3 cells/HPF, P=0.01 Student’s T-test, Figure 6C), while total (single + double positive) WFA-expressing cell density was similar between hemispheres (23.1±2.2 vs. 24.8±5.4 cells/HPF, P=0.55 Student’s T-test, Figure 6C). These results were due to loss of both parvalbumin single-labeled (14.7±2.2 vs. 21.7±2.4 cells/HPF, P=0.05 Student’s T-test, Figure 6B) and double-labeled cells (8.7±1.0 vs. 13.3±0.5 cells/HPF, P<0.001 Student’s T-test, Figure 6B). In contrast, WFA-labeled single-positive cells trended to an increased density in the hemisphere exposed to HI compared with hypoxia alone (14.4±1.9 vs. 11.4±1.6 cells/HPF, P=0.26, Student’s T-test, Figure 6C). Taken together, the decrease in parvalbumin-expressing cell density with maintenance of total (single + double-positive) WFA-expressing cell density supports the hypothesis that this class of inhibitory neurons has not been lost, but is simply expressing less parvalbumin.
Subclasses of inhibitory neurons can be defined by their expression of neuropeptides and calcium binding proteins(Gonchar and Burkhalter, 1997; Markram et al., 2004). In addition to parvalbumin, inhibitory neurons may express calbindin or calretinin(Gonchar and Burkhalter, 1997). To examine whether the effect of neonatal HI was restricted to parvalbumin-expressing inhibitory neurons, or if other classes of inhibitory neurons were also affected, we quantified the density of calbindin and calretinin-expressing cells in adjacent sections from the same animals used for the analysis of parvalbumin and WFA (Figure 6B). No significant differences were observed between hypoxia alone and HI hemispheres in the density of calbindin (7.1±2.7 vs. 6.9±0.4 cells/HPF, P=0.9, Student’s T-test, Figure 6C) or calretinin-expressing cells (4.1±0.8 vs. 5.0±1.2 cells/HPF, P=0.1, Student’s T-test, Figure 6C).
To determine whether the magnitude of loss of parvalbumin expression was associated with the magnitude of impaired ocular dominance plasticity, we examined the correlation between the ODI and the difference in parvalbumin cell density between hypoxia-exposed and HI-exposed hemispheres (Figure 6C). Increasing loss of parvalbumin expression was significantly and linearly correlated with ODI shift in the moderately injured group (Figure 6C, closed circles, P=0.0003, r = 0.99). Among the mildly injured animals, as with animals not exposed to neonatal HI, there was no significant difference in hemispheric parvalbumin density and these animals had intact ODP (Figure 6C, open circles). Including these animals in the analysis reduces the correlation strength, but the relationship remains significant (P=0.004, r = 0.82). Cortical thickness is also significantly associated with hemispheric parvalbumin difference (β=−0.04, P=0.02, 95% confidence interval −0.07 – −0.008, linear regression), as is the difference in hemispheric thickness (β=0.06, P=0.01, 95% confidence interval 0.02 – 0.10, linear regression).
To explore the functional consequences of altered parvalbumin expression following HI, we recorded 101 isolated single units from V1 in HI and unmanipulated animals using linear 16-channel multielectrodes (Figure 7A). Animals exposed to neonatal HI were selected for a moderate pattern of injury on MRI at P3 (see methods). Post-recording cortical thickness (mean±S.D. = 1193±109 μm) and hemispheric difference in parvalbumin cell density (mean±S.D. = 10.5±3.6 cells per HPF) were representative of the moderate injury group, although by histology two animals showed injury patterns that were more severe than typical. Electrodes were placed at a fixed depth spanning all cortical layers. Laminar localization was determined using current source density (CSD) analysis of the local field potential (LFP) recorded at each electrode site during presentation of a contrast reversing checkerboard pattern as described(Niell and Stryker, 2008). This allowed localization of layer 4, where sensory input first arrives, spreading into layer 2/3. Visually responsive units were observed across all cortical layers.
Fewer active units per electrode penetration were observed in HI animals (n=39 units, 5 animals, 5 penetrations) than control animals (n=62 units, 4 animals, 5 penetrations). To measure orientation and spatial selectivity we presented drifting sinusoidal gratings. Blank stimuli were included to measure spontaneous activity. Mean and distribution of spontaneous firing rates were the same for control and HI animals (mean±SD control=0.32±0.46 spikes/sec, HI=0.27±0.34 spikes/sec) (Figure 7B), indicating an absence of excitatory/inhibitory imbalance or widespread epileptiform activity in visual cortex of HI animals.
We calculated an orientation tuning curve at the spatial frequency yielding the highest response. Units were classified as visually responsive to gratings if the peak of the tuning curve exceeded 2 spikes/sec. Response of individual units to drifting gratings was not qualitatively different between groups (Figure 7D1, 7E1). Specifically, we found no evidence of prolonged discharge to visual stimulation. We observed a smaller proportion of visually responsive units in HI (30%; n=12 units) than in control animals (56%; n=36 units). However, the mean orientation selectivity index (OSI; see Methods) of responsive units was not significantly different between control and HI animals (mean±SD control=0.56±0.29, HI=0.67±0.34). Further, the half-width at half-maximum of the responsive component of the tuning curve was not different between groups (mean±SD control=29.78±8.39°, HI=32.35±11.49°). This suggests that mechanisms leading to the development of well-tuned responses are intact in HI animals.
To further examine the response properties of units in V1, we next presented stimuli consisting of stochastic noise at a range of spatial and temporal frequencies (0.05 – 0.12 cycles per degree, dc – 4 Hz). This “noise movie” was modulated sinusoidally from 0 to maximal contrast, leading to a periodic modulation of firing rate in visually responsive cells. The peak response of a unit occurred at the frequency of contrast modulation (F1, the first harmonic at 0.1 Hz). The distribution of peak responses was significantly shifted towards lower values in HI than control animals (Figure 7C, p=0.006, Kolmogorov-Smirnov test). Next, we explored whether this population shift was due to overall reduced amplitude of responses in HI animals. Responsive units fire more with increasing stimulus contrast, leading us to classify a unit as responsive if its phase fell within ±30° of the mean phase for all units. Using these criteria, we deemed 16 of the control and 12 of the HI units not to be responsive to the stimulus. The average F1 response of the remaining phase-matched units was markedly reduced in HI animals (mean±SD control=1.09±1.00, HI=0.42±0.37 spikes/sec, p=0.006, Mann-Whitney U test). The lower percentage of visually responsive units suggests diminished responsiveness of visual cortex following neonatal HI.
Neonatal hypoxic ischemic brain injury often results in pervasive neurodevelopmental impairment even with the existence of unique plasticity mechanisms in the developing brain that might contribute to recovery from injury. Following injury, many aspects of cortical development proceed normally despite extensive subplate neuron loss, including the development of precise retinotopic maps, observed in the present study in all but the most severely injured animals, the upregulation of many activity-regulated proteins involved in excitatory or inhibitory neurotransmission and the establishment of highly selective visual responses. In this rodent model of neonatal HI, animals with moderate cortical damage exhibit impaired ocular dominance plasticity and this impairment is associated with defective development of parvalbumin-expressing inhibitory neurons. Despite this alteration of inhibitory neuron phenotype, baseline neuronal activity is unchanged and overall activation in visual cortex is diminished. These findings provide a link between a translational model of very early hypoxic-ischemic brain injury and emerging understanding of the maturation of cortical circuits involved in structural plasticity. The results may also provide insight into deficits noted in children suffering neonatal brain injury including cortical visual impairment and learning disability. Finally, these results suggest ocular dominance plasticity is an outcome measure for studies of early brain injury that may be useful for selecting and evaluating interventions aimed at repair and recovery.
During the initial postnatal weeks of cortical development, both glutamatergic(Arumugam et al., 2005; Lujan et al., 2005) and GABAergic(Ben-Ari, 2002) signaling undergo major developmental changes manifested in alterations of the expression levels of many proteins in these pathways. In this dynamic period, a balance of excitation and inhibition is necessary for experience dependent cortical plasticity(Hensch, 2005). Appreciation of the unique role of cortical inhibition has emerged from experiments with genetically modified mice lacking the inducible isoform of the GABA biosynthetic enzyme, GAD65(Hensch et al., 1998), overexpressing the neurotrophin BDNF(Hanover et al., 1999) or following pharmacologic manipulation of inhibition (Hata and Stryker, 1994; Hensch et al., 1998; Fagiolini and Hensch, 2000). These experimental models all disrupt activity-dependent cortical plasticity through separate manipulations of the development of inhibition or the balance of excitation and inhibition. The present results extend from a model not previously known to preferentially affect excitatory or inhibitory networks. The Rice-Vannucci procedure performed at postnatal day 3 or earlier leads to damage focused in the subplate and intermediate zone with loss of subplate neurons(McQuillen et al., 2003) and preoligodendrocytes(Back et al., 2002). Deep layers of neocortex (V/VI) are injured only in severely affected animals and superficial layers (I–IV) are thought to be unaffected.(McQuillen et al., 2003; Sizonenko et al., 2003). We find not only that early cerebral HI impairs subsequent ocular dominance plasticity, but also that this impairment is tightly linked to decreased expression of the inhibitory neuron marker parvalbumin. In fact, in the moderately injured group, the loss of parvalbumin expression explains more than 95% of the variance in ocular dominance plasticity following HI and MD, providing evidence for a link between development of inhibition and experience-dependent plasticity in a translational model of human disease.
GAD65 knockout mice display enhanced cortical activation in the form of prolonged discharges in response to visual stimuli(Hensch et al., 1998). Despite this defect, many aspects of visual cortical organization and function develop normally including retinotopy, baseline activity and orientation selectivity(Hensch et al., 1998). This normal development in GAD65 knockout mice should not be surprising because there is little or no cortical expression of GAD65 before P6, during the period of map formation (Kiser et al., 1998). Although neonatal HI results in diminished parvalbumin expression, we do not find any evidence for excessive cortical activity resulting from diminished inhibition. Specifically, we did not observe prolonged discharges of single neurons, altered spontaneous activity or alterations in the expression of activity-regulated proteins. In fact, neonatal HI resulted in reduced visual cortical responses even to an optimal visual stimulus. Results from these two models indicate that impaired ocular dominance plasticity can be associated with alterations of cortical responsiveness in either direction.
Neonatal HI occurs during a developmental period when GABAergic signaling is depolarizing and involved in the maturation of cortical circuits, synapses and excitatory neurotransmission (reviewed in (Akerman and Cline, 2007)). GAD65 is not involved at this early stage of maturation. Following neonatal HI, we observed that both parvalbumin expression and cortical responsiveness are reduced. Neuronal activity is known to regulate parvalbumin expression (Jiao et al., 2006), suggesting that altered inhibition might result from a primary defect of cortical activation. However, this decrease in cortical activation is not severe enough to interfere with other aspects of activity-dependent visual cortical development including retinotopic organization, orientation selectivity and maturation of glutamatergic and GABAergic signaling pathways. Defects in the development of both cortical activation and inhibition have been observed following neonatal subplate neuron ablation.
Subplate neurons are incorporated into a transient circuit receiving innervation from and projecting to thalamus, as well as sending a recurrent collateral to layer IV(Allendoerfer and Shatz, 1994). A hypothesis for the function of this transient circuit is that subplate neurons may provide ‘gain control’ over nascent thalamocortical innervation and serve to strengthen developing thalamic inputs(Kanold and Shatz, 2006). During early postnatal weeks, cortical glutamatergic neurotransmission is immature and weak(Lujan et al., 2005). Neocortical neurons are closely coupled by gap-junctions and are capable of spontaneous propagating network oscillations that arise in the subplate(Dupont et al., 2006). Neocortical gap junction down-regulation is driven by increased signaling through the NMDA receptor(Arumugam et al., 2005) and a shift to NMDA-dependent network oscillations in neocortex coincides with the normal developmental disappearance of subplate neurons(Dupont et al., 2006). Premature loss of subplate neurons therefore would be predicted to diminish early cortical activity, a hypothesis confirmed by direct recordings of visually-evoked single unit activity in the present results. The alternative explanation that parvalbumin-expressing cells have died in response to HI appears unlikely in light of the facts that density of WFA-labeled perineuronal nets were unchanged following HI and that a majority of PNNs surround parvalbumin-expressing cells(Hartig et al., 1992).
Injections of the glutamatergic excitotoxin, kainic acid(Chun and Shatz, 1988) or a specific immunotoxin(Kanold et al., 2003) have been used to destroy subplate neurons in developing neocortex of kittens. When performed in the neonatal period, this manipulation dramatically reduces cortical responsiveness(Kanold et al., 2003) and prevents the segregation of ocular dominance columns(Ghosh and Shatz, 1992) and the refinement of orientation maps(Kanold et al., 2003), processes known to require neuronal activity(Stryker and Harris, 1986; Ruthazer and Stryker, 1996). Kanold and Shatz have demonstrated that neonatal subplate neuron ablation, which results in weak or absent visual responses, also shows defective development of the glutamatergic thalamic axon to layer IV synapse(Kanold et al., 2003) that includes diminished layer IV expression of a potassium-chloride co-transporter (KCC2) required for GABA-induced hyperpolarization and a specific GABA receptor subunit (GABAAα1)(Kanold and Shatz, 2006) essential for plasticity(Fagiolini et al., 2004). These changes lead to anatomical plasticity(Kanold and Shatz, 2006) of a form that could not be evaluated in the present study because rodents lack ocular dominance columns(Antonini et al., 1999). The removal of subplate neurons is more complete following kainic acid injection in kitten(Lein et al., 1999), compared with neonatal HI in rodent(McQuillen et al., 2003) and the reduction of cortical responsiveness and selectivity is much more dramatic(Kanold et al., 2003). Finally, neonatal HI produces a spectrum of injury and leads to death of other cell types than subplate neurons, including preoligodendrocytes(Back et al., 2002) and deep layer neurons in the more severe cases.
A limitation of the present results relates to the inherent variability in the Rice-Vannucci model of HI that results in animals with a range of injury(Towfighi et al., 1997; McQuillen et al., 2003). To account for this variability, both categorical and continuous measurements of injury severity have been analyzed with similar conclusions drawn from each approach. This experimental weakness however, strengthens the translational value of the model as it reproduces variability observed in human neonatal HI(Ferriero, 2004). A minority of animals with severe injury had absent visual responses, a condition reminiscent of cortical visual impairment observed occasionally in premature infants with periventricular leukomalacia resulting from neonatal brain injury (Cioni et al., 1997; Lanzi et al., 1998). Disrupted plasticity observed in moderately injured animals may be a mechanism that limits recovery from injury, contributing to widespread impairments noted across multiple cognitive domains in newborns with extreme prematurity(Hack et al., 2002; van Baar et al., 2005).
In summary, we observed impaired ocular dominance plasticity in animals with moderate cortical injury following neonatal HI and the magnitude of impaired plasticity was tightly linked to a loss of parvalbumin expression. Despite altered inhibitory neuron phenotype, baseline neuronal activity and many physiologic characteristics of mature visual cortex were normal, while visually evoked activity was reduced. Taken together with the enhanced cortical activation observed in GAD65 knockout mice,(Hensch et al., 1998), these results suggest that ocular dominance plasticity may be uniquely dependent upon precise regulation of cortical activation. Impaired inhibitory neuron development and subplate neuron death seen in this model and the related cat kainate model, suggest a specific disruption of a unique, transient developmental circuit. A goal of this translational investigation is to develop clinically relevant strategies that restore plasticity and improve outcome. Persistent disruption of ocular dominance plasticity following a different neonatal manipulation (alcohol exposure) has been reported in a ferret model(Medina et al., 2003; Medina et al., 2005; Medina and Ramoa, 2005). Importantly, in this model plasticity could be restored by a pharmacologic treatment (phosphodiesterase type 1 inhibitor)(Medina et al., 2006). Strategies to improve outcome by prevention of neonatal subplate neuron injury will be dependent upon identifying mechanisms of subplate neuron selective vulnerability, such as the expression of GluR2-lacking AMPA receptors(Talos et al., 2006). Alternatively, the present results suggest the paradoxical ideas that treatment with a use-dependent GABA agonist (e.g. diazepam) or interventions to augment cortical activation may be therapies worthy of further study to restore activity dependent plasticity following injury.
A1, Amplitude and phase of F1 response for all units. Green lines mark ± 30 degrees of average response of all units. Phase-matched control (black) and HI (red) are within this range; units outside the range are in green. A2, enlarged view of A1 around F1=1 with average vectors of control (black) and HI (red) responses.
This work was supported by NIH grants K02 NS047098 and 5R01NS060896 to PSM. JC was an Aventis Pharmaceuticals Fellow of the Life Sciences Research Foundation and supported by a Knights Templar Eye Foundation Pediatric Ophthalmology Research Grant.