Neonatal Hypoxia-Ischemia Causes Persistent Intracortical Circuit Changes in Layer 4 of Rat Auditory Cortex

Abstract

The connection between early brain injury and subsequent development of disorders is unknown. Neonatal hypoxia-ischemia (HI) alters circuits associated with subplate neurons (SPNs). SPNs are among the first maturing cortical neurons, project to thalamorecipient layer 4 (L4), and are required for the development of thalamocortical connections. Thus, early HI might influence L4 and such influence might persist. We investigated functional circuits to L4 neurons in neonatal rat HI models of different severities (mild and moderate) shortly after injury and at adolescence. We used laser-scanning photostimulation in slices of auditory cortex during P5–10 and P18–23. Mild injuries did not initially (P6/P7) alter the convergence of excitatory inputs from L2/3, but hyperconnectivity emerged by P8–10. Inputs from L4 showed initial hypoconnectivity which resolved by P8–10. Moderate injuries resulted in initial hypoconnectivity from both layers which resolved by P8–10 and led to persistent strengthening of connections. Inhibitory inputs to L4 cells showed similar changes. Functional changes were mirrored by reduced dendritic complexity. We also observed a persistent increase in similarity of L4 circuits, suggesting that HI interferes with developmental circuit refinement and diversification. Altogether, our results show that neonatal HI injuries lead to persistent changes in intracortical connections.

Introduction

Premature birth or disruptions in prenatal brain development result in increased risk of cognitive impairments. One relatively common disruption is hypoxic–ischemic brain injury, which results in deficits of attention, memory, executive function, and speed of processing, depending on the severity of the injury (Anderson and Arciniegas 2010). Changes in cortical circuits likely underlie these functional deficits and raise the question: What functional circuit changes occur following early injuries?

Hypoxia-ischemia (HI) is a disorder in which the preterm infant undergoes a lack of oxygen and blood flow to the brain. Overall, ~0.7–1.2 million infants each year are affected by HI globally (Lawn et al. 2005). For those who develop more moderate HI encephalopathy, ~0.2–0.3% of the 4 million infants born in the United States of America each year are affected and have mortality rates of about 6–30% (Raghuveer and Cox 2011). In addition, estimates of 30% of cases of neonatal encephalopathy (NE) in developed countries and 60% of cases of NE in developing countries are associated with intrapartum HI (Kurinczuk et al. 2010).

Specifically, injury to the premature white matter of the brain disrupts the normal maturation of the brain and consequently increases the risk of developing cerebral palsy and epilepsy in 20–30% of surviving infants (Kaindl et al. 2009; Raghuveer and Cox 2011). Rodent models of HI (McQuillen et al. 2003; Mikhailova et al. 2017; Millar et al. 2017) have shown that at least at young ages, cortical activity is reduced after HI (Failor et al. 2010; Ranasinghe et al. 2015). However, it is unclear which specific circuits are affected by HI and how this change translates into overall reduced network activity.

In early cortical development, one important neuronal circuit affected by HI and whose damage results in abnormal brain function is formed by subplate neurons (SPNs) (Kanold and Luhmann 2010; Kanold et al. 2019; Molnar et al. 2020). SPNs are a largely transient population of neurons (Kostovic and Rakic 1980, 1984, 1990; Luskin and Shatz 1985; Kanold and Luhmann 2010) and HI injuries can lesion a fraction of them (McQuillen et al. 2003; Mikhailova et al. 2017). While more severe injuries can damage deep cortical layers (Okusa et al. 2014; Millar et al. 2017), mild HI injuries that do not cause loss of SPNs (Okusa et al. 2014; Sheikh et al. 2019) can result in functional hyperconnectivity of circuits to SPNs, manifested as altered SPN morphology (Sheikh et al. 2019). In sheep, HI has also been associated with altered SPN morphology and excitability (McClendon et al. 2017; Millar et al. 2017).

SPNs play a key role in orchestrating the development of cortical layer 4 (L4) (Kanold and Luhmann 2010; Kanold et al. 2019; Molnar et al. 2020) and SPN lesions result in functional cortical changes reminiscent of the results of HI injuries (Kanold et al. 2003; Kanold and Shatz 2006; Failor et al. 2010; Tolner et al. 2012; Ranasinghe et al. 2015). In sensory cortices, SPNs are the first cortical neurons to receive thalamic inputs (Krmpotic-Nemanic et al. 1983; Shatz and Luskin 1986; Friauf and Shatz 1991; Herrmann et al. 1994; Hanganu et al. 2002; Higashi et al. 2005; Zhao et al. 2009), respond to sensory stimuli (Wess et al. 2017), and SPN circuits can be shaped by peripheral activity (Meng et al. 2021). For example, selective ablation of SPNs prevents thalamocortical patterning, functional maturation of thalamocortical circuits, normal cortical sensory responses, and alters plasticity during the critical period (Ghosh et al. 1990; Ghosh and Shatz 1992; Kanold et al. 2003; Kanold and Shatz 2006; Tolner et al. 2012). Moreover, SPN ablations prevent the development of early oscillatory cortical activity patterns (Tolner et al. 2012). While the effects of SPN damage on thalamocortical circuits have been established, effects of SPN damage by HI on intracortical circuits are unknown.

Since SPNs project to L4 (Finney et al. 1998; Zhao et al. 2009; Deng et al. 2017; Viswanathan et al. 2017) and since normal functional responses in the adult sensory cortices depend on mature thalamocortical processing, we hypothesized that mild HI, which alters SPN circuits (Sheikh et al. 2019), will also alter processing in L4. Moreover, while HI causes hyperconnectivity of SPNs at young ages (Sheikh et al. 2019), at later ages, a reduction in glutamate receptors as well as reduced activity in cortical neurons is observed (Ranasinghe et al. 2015). This raises the possibility that L4 circuits might not show hyperconnectivity after HI.

In order to determine if HI has effects beyond the subplate and if SPN damage from HI causes a change in intracortical circuits later in life, we used laser-scanning photostimulation (LSPS) to investigate the functional spatial connection patterns of L4 neurons after HI in a thalamocortical slice preparation of auditory cortex (ACX). Specifically, we compared connection patterns for excitatory and inhibitory connections in cells from animals subjected to two different severities of neonatal (P1/2) HI, and Control. Since cortical circuits change markedly during the early two postnatal weeks, we investigated L4 circuits at two distinct time periods: during postnatal day (P) 5–10, thus within 1 week post-HI, and 2 weeks later during P18–23.

We find that after mild injuries (HI-Caut), the younger age group showed similar numbers of connections but those connections were from cells spanning a broader range of ACX. In contrast, moderate injuries (HI-Lig) caused hypoconnectivity of excitatory and inhibitory connections. We also find that for HI-Lig, there was weaker excitation from P6–7 for both L2/3 and L4, but a strengthening of excitation from P8–10 for both layers. We also observed a persistent increase in the similarity of circuit topology between cells. These functional changes were mirrored by reduced dendritic complexity. At P18–23, HI-Lig resulted in a strengthening of excitatory and inhibitory inputs within L4 and patchy increases in dendritic complexity. These results suggest that neonatal HI causes transient hypoconnectivity in L2/3 and L4 from P5–10, and that over time, most but not all connectivity differences resolve.

Methods

All procedures were approved by the University of Maryland Institutional Animal Care and Use Committee.

Animals

Timed-pregnant Sprague Dawley rats were obtained from Charles River. Time of birth was marked P0 (±0.5 day).

Hypoxia-Ischemia

Sprague–Dawley pups at postnatal day (P) 1–2 were anesthetized with 2–4% isoflurane for the Rice-Vannucci procedure (McQuillen et al. 2003; Failor et al. 2010; Ranasinghe et al. 2015; Sheikh et al. 2019). A small incision was made in the midline of the neck just above the sternal notch. The common carotid artery (CCA) was exposed and either coagulated or ligated with a single suture (#S-G618R13 PGA absorbable 6/0, AD Surgical), as done previously (Sheikh et al. 2019). Care was taken not to damage the sympathetic ganglion chain in order to avoid Horner’s Syndrome, which results in ptosis. The different surgical techniques were used to cause ischemia of different severities. Cauterization of the CCA caused a largely transient and thus milder form of ischemia (HI-Caut). Alternatively, ligating the CCA with a suture caused a permanent and moderate form of ischemia (HI-Lig). Because the CCA is proximal to the circle of Willis, both techniques only cause partial ischemia. As a control, we performed sham surgeries exposing the CCA.

After recovery, the pups were returned to the dam in normal room air for 2 h. Thereafter, the pups were placed in humidified chambers (Lexan, UMD Physics shop) and exposed to 5% oxygen (concentration monitored in chamber by UV Flux 25% Oxygen Sensor Module CM-0201, co2meter.com; GasLab® software), balanced nitrogen for 2–3 h and then returned to the dam as described previously (Sheikh et al. 2019). Chamber temperature was held constant by a water bath. During hypoxia, one pup from each litter was monitored for skin temperature. Skin temperature was kept constant at 35 °C. Surviving pups were removed if the total mortality percentage for the entire litter reached 30–50% on average. Chambers remained closed until the end of the hypoxia procedure.

Brain Slice Preparation and Solutions

Thalamocortical brain slices were prepared as previously described (Sheikh et al. 2019) using methods adopted from mice (Zhao et al. 2009; Viswanathan et al. 2012; Meng et al. 2014). Rats of either sex were used. Rats were deeply anesthetized with isoflurane prior to decapitation and removal of the brain. Thalamocortical slices (500 μm) were prepared using a vibrating microtome (Leica) in ice-cold artificial cerebrospinal fluid (ACSF) consisting of (in mM): 130 NaCl, 3 KCl, 1.25 NaH₂PO₄, 20 NaHCO₃, 10 glucose, 1.3 MgSO₄, 2.5 CaCl₂, pH 7.35–7.4, equilibrated with 95% O₂–5% CO₂. The slices were incubated in ACSF for 1 h at 30 °C and then kept at room temperature.

In Vitro Electrophysiology

Whole-cell recordings were performed with a patch-clamp amplifier at room temperature (Multiclamp 700B; Molecular Devices) as previously described (Sheikh et al. 2019). Electrodes were filled with (in mM): 115 cesium methanesulfonate (CsCH₃SO₃), 10 HEPES, 5 NaF, 10 EGTA, 15 CsCl, 3.5 MgATP, 3 QX-314, pH 7.25, and 300 mOsm. Biocytin or Neurobiotin (0.5%) was added to the electrode solution as needed. This electrode solution blocks most intrinsic conductances and increases the fidelity of recording synaptic inputs. To reduce the probability of multisynaptic events, all slices were perfused during recordings with a high-Mg²⁺ ACSF solution: 124 NaCl, 5 KCl, 1.23 NaH₂PO₄, 26 NaHCO₃, 10 glucose, 4 MgCl₂, and 4 CaCl₂. The electrode resistance in the bath was 4–10 MΩ. Data were acquired with a National Instruments AD board and EPHUS software (Suter et al. 2010). Membrane voltages were corrected for an estimated liquid junction potential of 10 mV. Activation profiles of neurons were created by recording in cell-attached or loose-patch mode while mapping the same area and recording action potentials. The location of the recording site for primary ACX was identified by landmarks (Cruikshank et al. 2002; Zhao et al. 2009; Viswanathan et al. 2012; Meng et al. 2014; Sheikh et al. 2019).

Laser-Scanning Photostimulation

LSPS was performed as previously described (Sheikh et al. 2019); 0.5–1 mM caged glutamate (Ncm-Glu, [N-(6-nitro-7-coumarylmethyl-L-glutamate)]) (Muralidharan et al. 2016) was added to the ACSF. We typically stimulated an array of up to 30 × 30 sites with 40 μm spacing to enable us to probe areas of ~1 mm × 1 mm (Fig. 1B). At the magnification used, the image acquired by the camera sensor did not cover the entire radial extent of the rat cortical column from pia to white matter; therefore, we did not map the entire extent during each recording. Thus, we only investigated inputs to L4 neurons originating from L2/3 and within L4. Stimuli were applied at 1 Hz. Laser power in each experiment was adjusted to achieve reliable neuron activation and was <25 mW for all experiments (Fig. 1C). We typically did not record more than 5 cells per slice. To detect monosynaptically evoked postsynaptic currents (PSCs), we only included PSCs with latencies below 50 ms after the stimulation. We chose this threshold as spiking responses under our condition occurred up to 50 ms after stimulation in both control and HI-Lig (Supplementary Fig. 1) (Viswanathan et al. 2012; Meng et al. 2014; Nagode et al. 2017). AMPAR-mediated responses were recorded while holding cells at 70 mV, and GABAR-mediated responses were recorded at 0 mV holding potential (Fig 1E). Traces containing a short-latency (<8 ms) response were considered to result from direct activation of receptors on the patched cell and were excluded from analysis. Traces with latencies longer than 50 ms were discarded due to potential involvement of polysynaptic components (Fig. 1F). Analysis was performed blinded and essentially as previously described with custom software written in MATLAB, most of which are automated routines (Meng et al. 2014; Sheikh et al. 2019; Meng et al. 2021). Stimulus locations that showed PSCs were deemed connected and were used to derive binary connection maps. We aligned connection maps for L4 neurons in the population and averaged connection maps to derive a spatial connection probability map. In these maps, the value at each stimulus location indicates the fraction of L4 neurons that received input from that stimulus location. For easier visualization, we smoothed the maps when plotting. Layer boundaries were determined from infrared images as previously described (Meng et al. 2014; Sheikh et al. 2019).

Figure 1.

Resolution of LSPS is not altered by HI. (A) Experimental timeline and comparison between human gestational development and rat postnatal development. (B) Infrared image of P9 thalamocortical brain slice with electrode on a L4 ACX neuron (left). Schematic of LSPS presynaptic activation of cortical neurons via photolysis (355 nm) of caged glutamate. (C) Distribution of number of evoked action potentials. (D) Distribution of distance for which 80% of action potentials were evoked. For all plots, “*” is P < 0.05. (E) Schematic of LSPS performed as a pseudorandom activation in order to record both EPSCs and IPSCs. (F) Example traces acquired with photostimulation at different locations. The vertical blue line indicates the time of photostimulation (blue). The analysis window is delimited by the dashed and solid black lines, which mark 8 and 50 ms after laser onset. Insets show early part of response for two example traces. (G) Pseudocolored maps of a single P9 L4 neuron show evoked PSC charge for each stimulus location for 1 example L4 neuron. Direct responses were excluded from analysis (black pixels in the map). The white filled circle indicates soma location, and the horizontal white bars indicate layer boundaries and also serve as scale bars of 200 μm.

We then derived laminar measures for individual cells which was done on unsmoothed maps. We calculated the input area for each layer as a measure reflecting the number of presynaptic neurons in each layer projecting to the cell under study. Input area is calculated as the area within each layer that gave rise to PSCs. Mean charge is the average charge of PSCs from each stimulus location in each layer. Intralaminar integration distance is the extent in the rostro-caudal direction that encompasses connected stimulus locations in each layer. Since the tonotopic map is largely in the rostro-caudal axis, the intralaminar integration distance reflects integration across the tonotopic axis. While the input area and intralaminar integration are related, the input area will also show changes along the columnar (pia-ventricle) axis if more or less cells within a tonotopic place are recruited, for example, only L2 cells versus L2 and L3 cells. We calculated excitatory/inhibitory (EI) balance index in each layer for measures of input area and strength as (E − I)/(E + I), thus (Area_E − Area_I)/(Area_E + Area_I), resulting in a number that varied between −1 and 1, with 1 indicating dominant excitation and −1 indicating dominant inhibition (Meng et al. 2021). Since a measurement of E is not possible close to the soma due to direct responses, we excluded the direct area from both the E and I maps. Thus, this EI measure does not account for the contribution for cells from nearby locations but does allow analysis of the EI balance of inputs arising from different layers.

Spatial connection probability maps show the average connection pattern in each group. To visualize the diversity of connection patterns over the population of neurons in each group, we calculated the spatial correlation of the binary connection maps in each group by calculating the pairwise cross-correlations (Meng et al. 2020, 2021).

Dendritic Morphological Analysis

Recorded cells filled with Biocytin were stained and reconstructed in Neurolucida (MBF Bioscience). Sholl analysis was performed in Neurolucida as previously described (Sheikh et al. 2019; Meng et al. 2020).

Statistics

Results were analyzed by using the Multiple Comparison test. To perform multiple comparisons of group means, we used one-way ANOVA or Kruskal–Wallis test based upon normality of the data. Histology analysis results are plotted as boxplots using Rank Sum (based on Lilliefors test for normality) and deemed significant if P < 0.05 (Sheikh et al. 2019; Meng et al. 2020).

Results

In development, L4 receives excitatory input from SPNs (Zhao et al. 2009; Deng et al. 2017; Viswanathan et al. 2017). Because HI results in functional changes in SP connectivity (Sheikh et al. 2019), which might impact L4 connectivity (Kanold et al. 2003; Kanold 2009; Tolner et al. 2012), we sought to investigate if functional circuits within L4 also changed subsequent to HI.

We previously used two different HI models, both of which induced relatively mild cortical damage with varying severity (Sheikh et al. 2019). Importantly, none of these models caused overt reduction in the numbers of SPNs (Sheikh et al. 2019) and thus are categorized as mild based on the scale developed by McQuillen and colleagues (McQuillen et al. 2003; Failor et al. 2010; Mikhailova et al. 2017). In order to assess the state of functional cortical circuitry in L4, we chose two time periods reflecting two distinct stages of auditory development: P5–10, prior to the opening of the ear canals at ~P11, and after the critical period (P18–23) (Zhang et al. 2001) (Fig. 1A).

We recorded L4 neurons in thalamocortical slices of ACX in order to investigate the spatial pattern of functional connections to L4 neurons from the cortical plate (Fig. 1B). In ACX, L4 is mostly comprised of pyramidal cells (Smith and Populin 2001; Barbour and Callaway 2008; Chang and Kanold 2021). We performed LSPS in N = 232 cells (P5–10: 117 cells; 40 cells in 7 Control animals, 39 cells in 6 HI-Caut animals, and 38 cells in 8 HI-Lig animals; P18–23: 115 cells; 40 cells in 5 Control animals, 40 cells in 10 HI-Caut animals, and 35 cells in 7 HI-Lig animals; 1/6 rats for HI-Caut P5–10 L4 was also used in (Sheikh et al. 2019). All other rats for HI-Caut, HI-Lig, and Control were only used in this study).

LSPS Reliably Activates Cortical Neurons in Both Age Periods

In order to ensure that our photostimulation has a similar spatial resolution at both time points and conditions, we performed cell-attached patch-clamp recordings together with LSPS on L2/3 and L4 neurons (N = 138 cells) from Control and HI-Lig groups for P5–10 and P18–23 (Fig. 1C,D; Supplementary Table 1) (P5–10: 18 Control cells for L2/3 and L4 each; 21 L2/3 HI-Lig cells and 17 L4 HI-Lig cells; P18–23: 17 L2/3 Control cells and 15 L4 Control cells; and 17 L2/3 HI-Lig cells and 15 L4 HI-Lig cells). To characterize the spatial resolution of LSPS, we targeted the laser beam to 900 locations with ~40 μm spacing to evoke action potentials in the recorded neuron (Fig. 1B). The numbers of evoked action potentials were similar between neurons from Control and HI-Lig in the older age group. In the younger age group, however, we find that L2/3 HI-Lig neurons showed a slightly larger number of action potentials (Fig. 1C, Supplementary Table 1). Next, we quantified the distance of locations from the soma of the recorded cell where action potentials could be evoked. While we find no difference in L2/3 cells at P5–10, and L2/3 and L4 cells at P18–23, similar to our prior results (Sheikh et al. 2019) (Fig. 1D), L4 cells at young ages showed a smaller average distance (median of 90 vs. 160 μm) (Fig. 1D, Supplementary Table 1). These results suggest that LSPS reliably and similarly activates L4 and L2/3 neurons in HI and Control but that at young ages, stimulation of L4 cells had to occur closer to the soma to be effective.

Moderate HI Results in Transient Hypoconnectivity at P6–7 Followed by Strengthening at P8–10, while Mild HI Results in Hyperconnectivity by P8–10

We first investigated in the younger age group if circuits to L4 neurons changed consequent to HI. We recorded from L4 neurons (N = 117 cells) at P5–10 and used LSPS with whole-cell patch-clamp recordings (Meng et al. 2014; Sheikh et al. 2019) to spatially map the connectivity of excitatory (AMPA) and inhibitory (GABA) inputs to A1 L4 neurons (Fig. 1E); 900 stimulation sites with ~40 μm spacing were mapped in a pseudorandom pattern (Viswanathan et al. 2012; Meng et al. 2014; Sheikh et al. 2019) (Fig. 1B). Photostimulation evokes spiking in the presynaptic neuron, and if the targeted neuron is connected to the recorded L4 neuron, a short-latency evoked PSC can be recorded. During voltage-clamp recordings when cells were held at −70 mV, photostimulation was able to evoke two types of responses: 1) a large inward current with short onset latency (<8 ms) and 2) a smaller amplitude inward current with a longer latency (Fig. 1F). The shorter latency response is attributable to direct activation of the recorded cell, whereas the longer latency response is dependent on synaptic release from presynaptic cells (Meng et al. 2014). We quantified the strength of the inputs by calculating the charge of the PSC and plotted this as 2D maps of the evoked PSC as a function of stimulus position (Fig. 1G). Excitatory and inhibitory inputs to each neuron are separately identified by stimulating and alternatively holding cells at the holding potentials of −70 and 0 mV.

We first analyzed the excitatory circuits associated with L4 neurons by holding cells at −70 mV (Fig. 2A). For each neuron, we summed the EPSC amplitudes at each distance bin resulting in a columnar input profile for each cell (Fig. 2B). Comparing these columnar input profiles across groups showed that cells from control animals received input from L4 but also from L2/3. For moderate HI-Lig, these columnar input profiles qualitatively showed differences across the age range, with cells from younger animals (P6–7) showing decreased inputs from L4 compared with cells from older animals (P8–10) and from control. These results suggested that moderate HI-Lig resulted in hypoconnectivity especially at young ages. In contrast, L4 cells in the mild HI-Caut condition seemed to receive more connections especially from L2/3.

Figure 2.

HI-Lig results in transient hypoconnectivity and later strengthening of excitation. (A) Schematic for EPSC recording in L4 and laser photostimulation for P6–10. (B) Laminar summation of excitatory (V_hold = −70 mV) connections at P6–10. Horizontal white bars indicate layer boundaries; 40 cells in 7 Control animals, 39 cells in 6 HI-Caut animals, 38 cells in 8 HI-Lig animals. (C) Spatial distribution of P(Connection) for excitatory connections at P6–7 and P8–10. Black scale bar is 200 μm. Horizontal white bars indicate layer boundaries. Areas around cell body where direct responses occurred are marked in black. (D) Relative amount of excitatory input to L4 neurons originating in L2/3 and L4 under the two HI conditions compared with control. Total input area (left), integration width along rostral-caudal axis (middle), and mean EPSC charge (right). (E) Schematic for EPSC recording with LSPS at P18–23. N = 40 cells in 5 Control animals, 40 cells in 10 HI-Caut animals, and 35 cells in 7 HI-Lig animals. (F) Laminar summation of excitatory (V_hold = −70 mV) connections at P18–23. Horizontal white bars indicate layer boundaries. (G) Spatial distribution of P(Connection) for excitatory connections at P18–23. (H) Relative amount of excitatory input to L4 neurons originating in L2/3 and L4 under the two HI conditions compared with control.

To get an overview of the changes in the 2D space, we aligned maps from all recorded cells in each age and treatment group to the soma position and calculated for every spatial location the probability of being connected to a given L4 neuron; the result is a 2D spatial connection probability map for each group (Fig. 2C). L4 neurons in Control animals showed extensive excitatory inputs from L4 and L2/3 at both ages. After HI-Caut, there seemed to be a decrease within L4 inputs at P6–7, while at P8–10, an increased connection probability emerged from L2/3 immediately above the recorded L4 neuron. In contrast, after HI-Lig, there seemed to be an overall decrease in the excitatory connection probability at P6–7, consistent with our columnar profiles.

Next, we quantified these qualitative observations by measuring the evoked inputs of individual cells from each layer. We calculated the input area for each layer as a measure reflecting the number of presynaptic neurons in each layer projecting to the cell under study. Quantification of the laminar input areas showed that the amount of input from L2/3 was larger after HI-Caut than in Control at P8–10 (Fig. 2D left, Supplementary Fig. 2, Supplementary Table 3). In contrast, inputs from within L4 originated from a smaller area at P6–7. Thus, mild HI results in an initial rebalancing of excitatory connections to L4 neurons and hyperconnectivity at P8–10. After moderate HI-Lig, excitatory L4 as well as L2/3 inputs to L4 neurons originate from a smaller area than Control at P6–7, and this difference resolves by P8–10 (Fig. 2D left, Supplementary Fig. 2, Supplementary Table 3). Therefore, HI-Lig results in an initial hypoconnectivity of excitatory connections from P6–7, which later resolves by P8–10. Given that after HI, activation distance of L4 but not L2/3 neurons is decreased, the hypoconnectivity from both L4 and L2/3 is likely not due to reduced activation but a consequence of circuit changes.

To investigate if there has been a redistribution of inputs to L4 within each lamina, we computed the integration distance of inputs from each layer which reflects integration across the tonotopic axis (Fig. 2D middle, Supplementary Table 3). Since the tonotopic map is largely in the rostro-caudal axis, the intralaminar integration distance reflects integration across the tonotopic axis. While the input area and intralaminar integration are related, the input area will also show changes along the columnar (pia-ventricle) axis if more or fewer cells within a tonotopic place are recruited. This analysis revealed that L4 input in HI-Caut at P8–10 tended to originate from a wider distance suggesting an increased integration across the tonotopic axis. However, HI-Lig resulted in narrower distance from only L2/3 input from P8–10 suggesting that HI-Lig decreased integration across the tonotopic axis for superficial input to L4 neurons.

Next, we investigated the strength of the inputs from each layer by calculating the mean EPSC size of all inputs originating from each layer. At P6–7, inputs from both layers showed a weakening after moderate HI-Lig (Fig. 2D right, Supplementary Fig. 2, Supplementary Table 3). After mild HI-Caut, inputs from L4 were also weaker at P6–7 (Fig. 2D right, Supplementary Fig. 2, Supplementary Table 3). By P8–10, excitatory inputs from both layers were larger and thus stronger in HI-Lig (Fig. 2D right, Supplementary Fig. 2, Supplementary Table 3). Therefore, both severities of injuries resulted in initial weakening of inputs from L4 from P6–7; however, HI-Lig also initially weakened L2/3 excitatory inputs. Thus, HI-Lig initially weakened a larger area of excitatory connections compared with HI-Caut. However, by P8–10, HI-Lig resulted in strengthening excitatory connections in both layers.

Altogether, these results suggest that following HI-Lig, there is a transient period of hypoconnectivity and weakening of excitatory connections to L4 that transitions to strengthening, while HI-Caut results in initial weakening of excitatory connections, but by P8–10 hyperconnectivity without transient hypoconnectivity from both layers.

Moderate HI Causes a Persistent Increase in Excitatory Connection Strength within L4 by the End of the Third Postnatal Week

Since HI results in altered circuits within subplate and L4 at young ages, we wondered if such circuit changes would persist later in development. Thus, we used LSPS to spatially map the connectivity of excitatory and inhibitory inputs to A1 L4 neurons (N = 115 cells) at the end of the third postnatal week (P18–23) (Fig. 2E). This age is towards the end of the auditory critical period (Zhang et al. 2001) and many circuits in ACX have matured (Oswald and Reyes 2008; Meng et al. 2020).

Qualitative inspection showed that excitatory inputs to L4 cells in HI-Lig and HI-Caut were similar to control (Fig. 2F,G). Indeed, quantitative analysis revealed that the area of connections to L4 and integration width were unchanged but that EPSC amplitudes from within L4 had increased in HI-Lig (Fig. 2H, Supplementary Fig. 3A, Supplementary Table 3). Thus, the circuit hyperconnectivity present at P8–10 had resolved by P18–23, but the strengthening of connection persisted.

Moderate HI Causes a Transient Hypoconnectivity of Inhibition, while after Mild HI Functional Inhibitory Hyperconnectivity Is Present by the Second Postnatal Week

Next, we investigated if early HI also resulted in changes to inhibitory inputs to L4 neurons. To answer this question, we held neurons at 0 mV, performed LSPS (Fig. 3A) (N = 117 cells), and analyzed the resulting IPSCs as above. Qualitatively, the columnar profiles (Fig. 3B) showed that the mild HI-Caut resulted in hyperconnectivity from both layers at P8–10. In contrast, HI-Lig resulted in transient hypoconnectivity at P6–7 that resolved by P8–10. These apparent differences in the columnar profiles between groups were also evident in the spatial connection probability maps (Fig. 3C). Quantification of the input area from L2/3 and L4 validated our qualitative observations (Fig. 3D left, Supplementary Fig. 2, Supplementary Table 4). In moderate HI-Lig, L4 neurons received fewer inhibitory inputs from both L2/3 and L4 at P6–7, and the hypoconnectivity from L4 resolved by P8–10 (Fig. 3D left, Supplementary Fig. 2, Supplementary Table 4). In contrast, after mild HI-Caut, the input area was larger at P8–10, indicating that the increased number of inhibitory inputs from both layers emerged by P8–10. In addition, HI-Caut resulted in wider integration distance of inhibitory connections across the tonotopic axis. Initially from P6–7, HI-Caut resulted in weakened inhibitory connections for input from both layers. However, by P8–10, the strength of inhibitory inputs from L2/3 and L4 were larger from HI-Caut. After HI-Lig, inhibitory connection strength weakened from P6–7 in both layers but later resolved. Therefore, moderate HI-Lig results in a transient period of hypoconnectivity of both excitatory and inhibitory connections from L2/3 and from within L4, while in mild HI-Caut, excitatory and inhibitory hyperconnectivity emerges by P8–10.

Figure 3.

Inhibitory inputs to L4 show transient hypoconnectivity and later increased amplitude after HI. (A) Schematic of IPSC recording in L4 with LSPS. (B) Laminar summation of inhibitory connections (V_hold = 0 mV) at P6–10. Horizontal white bars indicate layer boundaries; 40 cells in 7 Control animals, 39 cells in 6 HI-Caut animals, 38 cells in 8 HI-Lig animals. (C) Spatial distribution of P(Connection) for inhibitory connections at P6–7 and P8–10. Layer borders indicated by white bars. Black scale bar is 200 μm. (D) Relative amount of excitatory input to L4 neurons originating L2/3 and L4 in the two HI conditions compared with control. Total input area (left), integration width along rostral-caudal axis (middle), and mean IPSC charge (right). (E) Schematic for ISPC recording with LSPS at P18–23. N = 40 cells in 5 Control animals, 40 cells in 10 HI-Caut animals, and 35 cells in 7 HI-Lig animals. (F) Laminar summation of inhibitory connections (V_hold = 0 mV) at P18–23. Horizontal white bars indicate layer boundaries. (G) Spatial distribution of P(Connection) at P18–23. Layer borders indicated by white bars. (H) Relative amount of inhibitory input to L4 neurons originating in L2/3 and L4 under the two HI conditions compared with control.

Moderate HI Causes a Persistent Increase in Inhibitory Circuit Strength within L4 by the End of the Third Postnatal Week

Next, we asked if inhibitory circuit changes would persist later in development. Thus, we used LSPS to spatially map the connectivity of inhibitory inputs to A1 L4 neurons (115 cells) at the end of the third postnatal week (P18–23) (Fig. 3E).

Qualitative inspection showed that inhibitory inputs to L4 cells in HI-Lig and HI-Caut were similar to control (Fig. 3F,G). The area of connections to L4 and integration width were unchanged, but IPSC amplitudes from within L4 had increased in HI-Lig (Fig. 3H, Supplementary Fig. 3B, Supplementary Table 4). These results mirrored the persistent changes in excitatory inputs and thus indicate that after HI-Lig, there is a persistent increase in the strength of excitatory and inhibitory inputs from within L4.

Investigating sex differences (Supplementary Fig. 6, Supplementary Table 6) showed that there were no sex differences seen in the majority of our measurements of connectivity from P18–23 across the different severities of injury. Both males and females had increases in excitatory L4 connection strength and the increased inhibitory connection strength for females echoes the group data. Thus, the sex of the animal did not appear to play a major role in determining connectivity changes seen after HI injury.

HI Causes Transient Excitation/Inhibition Imbalance between P6–10

Our HI procedures caused differential changes in excitatory and inhibitory circuits to L4 neurons. Since a balance of excitation to inhibition is required for normal brain function, we tested if the combined changes led to an imbalance of excitatory and inhibitory input. To investigate the combined changes, we calculated for each cell an EI-index as the normalized areal differences of the excitatory to inhibitory input, (E − I)/(E + I), for each cell (Sheikh et al. 2019). HI-Lig did not significantly alter the EI-Index from P6–10. However, HI-Caut resulted in a decreased EI-index at P8–10. The differences in the EI-index resolved by P18–23 (Supplementary Figs 4C and 5B, Supplementary Table 5). Altogether, our results indicate that both mild and moderate HI cause transient EI imbalances.

Both Mild and Moderate HI Cause Changes in the Circuit Diversity

Individual cells can vary in their inputs leading to an emergence in functional circuit diversity with age (Meng et al. 2020). In particular, circuits are most similar during a transient period in development (Meng et al. 2020). Thus, we investigated the spatial diversity of the circuits impinging on SPNs by calculating the similarity (spatial correlation) between connection maps within each population (Meng et al. 2020). This analysis showed that both HI-Caut and HI-Lig caused an initial decrease in excitatory circuit similarity at P6–7 followed by an increase in similarity at P8–10 that persisted through P18–23 (Fig. 4A). In addition, inhibitory circuits showed largely similar patterns (Fig. 4B). Thus, the developmental emergence of circuit diversity is impaired after both mild and moderate HI.

Figure 4.

HI increases the similarity of circuits to L4 neurons. (A) Top: Correlation between excitatory input maps in the three groups. Symbols show pairs and boxplots on right show averages. Bottom: Plotted are the differences to control. After HI, there was a transient decrease in similarity, but at older ages similarity increased. (B) Correlation between inhibitory input maps in the three groups.

After Moderate HI-Lig, L4 Cells Show Morphological Changes at the End of the First Postnatal Week which Resolve by the End of the Third Postnatal Week

So far, we revealed changes in the functional connectivity of L4 neurons as a result of HI injuries which might be linked to morphological changes. Prior work identified morphological changes after HI in L5 neurons that included reduced dendritic complexity and reduced glutamate receptor expression, especially in deep layers (Ranasinghe et al. 2015). Since we observed functional hypoconnectivity in L4 after HI-Lig at young ages, we tested if L4 cells had corresponding morphological changes. We reconstructed 24 L4 neurons (P6–10: N = 17 cells in 7 Control animals, N = 6 cells in 3 HI-Lig animals) using Neurolucida and quantified the numbers of dendrites, ends, and nodes. We found that L4 cells at young ages showed reduced dendritic complexity after HI, which is evidenced by the lower number of nodes and ends (Fig. 5A,B). Sholl analysis showed patches of reduced numbers of intersections at locations close to the soma at P5–10 indicating local hypoconnectivity (Fig. 5E).

Figure 5.

HI-Lig results in transient reduction in dendritic complexity of L4 neurons. (A) Exemplar Neurolucida reconstructions of recorded L4 neurons from P6–10. (B) Quantification of morphological properties for L4 P6–10 neurons. Total number of dendrites is unchanged after HI-Lig in both age groups (P = 0.11). Total numbers of ends and nodes are decreased after HI-Lig from P6–10 (P = 0.02; P = 0.04, respectively. Rank Sum). (C) Exemplar Neurolucida reconstructions of recorded L4 neurons from P18–23. (D) Quantification of morphological properties for L4 P18–23 neurons. Total numbers of dendrites, ends, and nodes are not different from control by P18–23 (dendrites: P = 0.58, ends: P = 0.68, nodes: P = 0.61; Rank Sum). (E,F) Sholl analysis for L4 P6–10 neurons (Control shown in black, HI-Lig shown in red) and L4 P18–23 neurons, respectively. Shown is the number of intersections of sequential serial spheres of 5 μm radial steps centered on to the soma of each neuron plotted as a function of radial distance. The bold line and shaded bands indicate the mean and the standard error, respectively.

We next reconstructed 34 L4 neurons from the older age group (P18–23: N = 14 cells in 6 Control animals, N = 20 cells in 7 HI-Lig animals). There was no difference in the numbers of dendrites, nodes, or ends between groups in this age range (Fig. 5C,D). However, Sholl analysis revealed patches of hyperconnectivity at P18–23 (Fig. 5F). Altogether, these anatomical measures mirror our functional measures and indicate that consequent to HI, there are transient morphological changes in L4. These results suggest that HI resulted in morphological abnormalities of L4 cells at young ages but that most morphological changes resolved by P18–23. Nevertheless, some morphological differences persisted, consistent with our physiological observations.

Discussion

We investigated the consequence of neonatal HI on the functional circuits to L4 from two different severities of injury. We find that both mild HI-Caut and moderate HI-Lig caused transient changes in excitatory and inhibitory circuits impinging on L4 neurons. While these circuit changes mostly resolved by P8–10, a persistent increase in excitatory and inhibitory connection strength remained. Moreover, we find that neonatal HI results in a persistent increase in the similarity of L4 circuits. Thus, early HI led to persistent changes in cortical circuits.

Our prior studies showed a hyperconnectivity of SPNs after mild HI (HI-Caut) and moderate HI (HI-Lig) (Sheikh et al. 2019). Here, we find that mild HI-Caut also results in hyperconnectivity and strengthened circuits to L4 that persist into adolescence. In contrast to our prior studies, we find that the moderate HI-Lig causes hypoconnectivity of L4 circuits at young ages as well as reduced excitability, as evidenced by reduction in the distance over which stimulation could evoke action potentials. Thus, moderate HI might alter early-born SPNs and later-born L4 neurons differently, possibly due to the differing maturational state. However, we also observed a mild shrinkage of the cortex (~15%) after HI-Lig, consistent with our prior observations (Sheikh et al. 2019). While the shrinkage can account for some fraction of the observed hypoconnectivity of ~ 50–80%, additional factors such as reduced connectivity are likely present. Indeed, we observed reduced dendritic complexity, consistent with the hypoconnectivity. The excitatory hypoconnectivity and reduced dendritic complexity in L4 after HI-Lig is consistent with decreased GluR1 expression in L4 and L6 at P3 and reduced dendritic complexity in L5 cells at P14 (Ranasinghe et al. 2015).

Our results show an effect of HI-Lig on inhibitory circuits to L4 which is consistent with prior studies showing decreased parvalbumin expression and altered inhibitory function consequent to neonatal HI (Failor et al. 2010). Thus, neonatal HI can alter both the connectivity pattern as well as connection strength of circuits to L4.

In addition, our results show that neonatal HI results in SPN hyperconnectivity (Sheikh et al. 2019) and strengthening of intra-L4 connections, but the mechanisms linking these early insults to circuit changes are unclear. At early ages, SPNs receive extensive inputs via NMDAR-only “silent” synapses (Meng et al. 2014; Kanold et al. 2019). In addition, hypoxia causes a depolarization (Luhmann et al. 1993; Luhmann 1996), which might be sufficient to “unsilence” cortical inputs to SPNs (Kanold et al. 2019). Thus, in this situation, the activity between SPNs and its inputs would be more correlated leading to AMPAR-mediated hyperconnectivity as observed experimentally (Sheikh et al. 2019). Since SPNs project to L4, such correlated activity could lead to the observed strengthening of intra-L4 connections and increased similarity of cortical circuits.

In contrast to hyperconnectivity of SPNs consequent to HI (Sheikh et al. 2019), here, we observe transient hypoconnectivity and decreased EPSC amplitude within L4, and later increased EPSC and IPSC amplitudes within L4 and patchy increases in dendritic complexity. This raises the question of how these two observations can be reconciled. It is possible that SPNs and L4 neurons are both directly affected by HI and that HI has differential effects on the two cell types. Indeed, more severe HI injuries cause a reduction of SPNs (McQuillen et al. 2003) and might also exert direct influences on L4, since severe cases of HI can damage deep cortical layers (Okusa et al. 2014; Millar et al. 2017). An alternative hypothesis is that the reduced connectivity of L4 cells is due to homeostatic changes in response to SP hyperconnectivity. The hyperconnectivity of SPNs likely results in higher SPN activity levels. Since SPNs project to L4 (Zhao et al. 2009; Deng et al. 2017; Viswanathan et al. 2017), there would be increased excitatory input to L4 which then could result in compensatory reduction of other cortical inputs to L4. In this scenario, there would be decreased excitatory drive to L4 leading to reduced activity in L4. Therefore, activity levels in L4 might not reach sufficient levels for other synapses to stabilize, thus preventing the developmental increase in connectivity.

Compared with the effects of HI on SPNs (Sheikh et al. 2019), effects on L4 neurons at young ages are more subtle and diverse, consistent with an increased vulnerability of SPNs to HI (McQuillen et al. 2003). However, the effects of HI injuries can be diverse but also depend on the severity and likely the timing of the injury, and the selective vulnerability of SPNs to HI injuries has been debated (Millar et al. 2017). It is clear that extent of histological SP damage, for example, SPN loss, by HI depends on injury severity. Mild HI neonatal injuries in rat do not cause loss of SPNs (Okusa et al. 2014; Sheikh et al. 2019) but can result in functional hyperconnectivity of circuits to SPNs and cause altered SPN morphology in A1 (Sheikh et al. 2019). However, more severe HI injuries in rat can lesion a fraction of SPNs (McQuillen et al. 2003; Okusa et al. 2014; Mikhailova et al. 2017). In addition, HI has also been associated with altered SPN morphology and excitability in fetal sheep (McClendon et al. 2017). However, HI, especially more severe injuries, can also damage deep cortical layers (Okusa et al. 2014; Millar et al. 2017). The differences between studies might be due to slight differences in induction protocols (e.g., duration of hypoxia), selection of pups (e.g., based on MRI [Ranasinghe et al. 2015; Mikhailova et al. 2017]), characterization of SPNs by birthdating (McQuillen et al. 2003), or marker expression (Okusa et al. 2014).

Also, in vitro studies showed that both SPNs and cortical L5 neurons can be affected by glucose deprivation (Albrecht et al. 2005). However, given differences in cortical maturation across the brain (Chang and Kanold 2021), neonatal HI might thus have a differing effect depending on the relative state of maturation. Since L4 receives inputs not just from subplate, the diversity of the circuit changes seen in our study likely partially reflects the variability of injury severity across animals and it is likely that similar diversity is present in human patients.

Our results also show that most circuit changes are resolved by P18–23. This is consistent with prior studies that showed that decreased EEG power consequent to HI at P6–11 apparently resolved by P14 (Ranasinghe et al. 2015). Since connections within L4 can powerfully control the flow of thalamic information through L4, the recruitment of L4 by sensory stimuli is likely being altered, consistent with abnormal sensory activity (Ranasinghe et al. 2015).

We find that after both severities of HI, there is an initial decrease in the circuit similarity between neurons, while later circuits in HI animals are more similar. Many cortical circuits undergo a developmental trajectory of exuberance and refinement (Murphy and Magness 1984; McMullen et al. 1988; Kossel et al. 1995; Antonini et al. 1998; Lendvai et al. 2000; Schachtele et al. 2011; Meng et al. 2020; Richards et al. 2020), under normal conditions, at least in A1 L2/3 and L1 (Meng et al. 2020), there is an early period when circuits are similar and cells become more heterogenous over development. Our results here suggest that neonatal HI injuries could interfere with this trajectory. This exuberant period of connectivity is thought to underlie the ability of the brain to adjust to the sensory environment during early critical periods (Meng et al. 2020); such a view would suggest that critical period plasticity might be disrupted in this model. Indeed, prior studies have shown that critical period plasticity is defective consequent to HI (Failor et al. 2010; Ranasinghe et al. 2015) and SP lesions (Kanold and Shatz 2006). Moreover, perinatal anoxia leads to altered auditory cortical processing (Strata et al. 2005). Thus, early HI injuries could lead to altered functional responses in the adult even though the underlying circuits at this age seem unchanged. However, we do find persistent changes in L4 circuitry consequent to HI. We find that there is decreased circuit diversity after HI. This increase in circuit similarity, or decreased circuit diversity, could increase the susceptibility for seizure generation (Rich et al. 2021) consistent with increased seizures after HI (Tekgul et al. 2006).

Our cell-attached recordings show that at P5–10, there is a slight increase in the evoked number of spikes in L2/3 neurons. Such increase might be expected to be reflected in increased EPSC amplitudes. However, we observe decreased amplitudes at P6–7 and increased amplitudes at P8–10. Thus, the effect of increased spike number seems minor at best. We also observed a decrease in the effective stimulation distance for L4 at P5–10 which is reflected in a smaller direct area at P5–7 HI-Lig (Fig. 2C). Thus, a fraction of the observed hypoconnectivity in L4 at P6–7 could be due to reduced excitability. However, since we do not observe hypoconnectivity at P8–10 and since HI-Caut does not show changes in direct area while having hypoconnectivity, this potential confounding effect is likely small. Moreover, we observe hypoconnectivity in L2/3, while cell-attached recordings show no changes in spiking distance. Importantly, at P18–23, we do not observe any changes in photoexcitability. Thus, altogether, we conclude that changes in the photoexcitability of cells likely have only a minor confounding effect on our interpretations.

HI studies in humans reveal sex-related differences (Rosenkrantz et al. 2019). In contrast, our study in rodents shows little influence of sex. Of the rodent studies that do reveal sex-related differences after HI, the experimental HI models were with different timelines (Hill and Fitch 2012). Thus, the influence of sex might be more relevant to injuries at slightly later developmental stages. Therefore, care has to be taken in comparing HI studies with respect to experimental model, injury timeline, and injury severity.

While our morphological analysis revealed a transient reduction in dendritic complexity, there was also a patchy increase in dendritic complexity by P18–23 after HI-Lig. Our morphological results show an increase in dendritic branches close to the soma and at ~200 μm distance. Given that we observe an increase in EPSC amplitude after HI-Lig, it could be that synapses are located closer to the soma after HI and thus be subject to reduced dendritic attenuation. However, given that we block most intrinsic conductances in our recordings, dendritic attenuation is minimal and we thus believe that the increase in EPSC amplitude reflects a genuine increase in the strength of specific synaptic inputs. Thus, we speculate that these inputs come from nearby L4 neurons.

In addition, the transient reduction in dendritic complexity at P6–10 HI is reflective of our physiological observations of hypoconnectivity and might indicate altered DNA synthesis (Kuan et al. 2004). Therefore, this transient dendritic reduction consequent to HI injury might provide a possible therapeutic window for intervention.

Overall, we find that HI causes transient hypoconnectivity in L4; persistent functional, anatomical changes in circuits to L4 such as stronger excitation and inhibition; decreased circuit diversity; and patchy dendritic complexity. However, the basic interlaminar circuitry remains intact indicating that potentially, the observed circuit changes might be reversible (Chang et al. 2005; Jantzie et al. 2015; Jantzie et al. 2018; Sun et al. 2020). In particular, approaches that can overcome the initial hypoconnectivity and prevent later circuit similarity might be effective.

Funding

POK supported by NIH R01DC009607. JPYK supported by NIH R01GM056481. AS supported by NIH T32DC000046.

Notes

P.O.K., A.S. designed research. P.O.K. supervised research. J.P.Y.K. contributed reagents. A.S. performed surgeries. A.S. performed in vitro LSPS experiments. X.M. wrote analysis code for L.S.P.S. experiments. A.S. and X.M. performed analysis. A.S., P.O.K. wrote the paper. We thank Dr Patrick McQuillen for his advice on the Rice-Vannucci procedure. We thank Dr Amal Isaiah and Ji Liu for surgical assistance, and Dr Catherine Carr for sharing her stereology Neurolucida system. We also thank Andrew Chang, Abanoub Gad, John Krstacic, Kevin Ross, Julia Alexieva, Jacob A. Bart, Shahzeib Syed, and Mathew Thomas for histological assistance. Conflict of Interest: The authors declare no conflict of interest.