Occlusion of activity dependent synaptic plasticity by late hypoxic long term potentiation after neonatal intermittent hypoxia

Abstract

To elucidate the mechanisms of memory impairment after chronic neonatal intermittent hypoxia (IH), we employed a mice model of severe IH administered at postnatal days 3 to 7. Since prior studies in this model did not demonstrate increased cell death, our primary hypothesis was that IH causes a functional disruption of synaptic plasticity in hippocampal neurons.

In vivo recordings of Schaffer collateral stimulation-induced synaptic responses during IH in the CA1 region of the hippocampus revealed pathological late phase hypoxic long term potentiation (hLTP) (154%) that lasted more than four hours and could be reversed by depotentiation with low frequency stimulation (LFS), or abolished by NMDA and PKA inhibitors (MK-801 and CMIQ). Furthermore, late phase hLTP partially occluded normal physiological LTP (pLTP) four hours after IH. Early and late hLTP phases were induced by neuronal depolarization and Ca²⁺ influx, determined with manganese enhanced MRI, and had increased both AMPA and NMDA – mediated currents. This was consistent with mechanisms of pLTP in neonates and also consistent with ischemic LTP described in vitro with OGD in adults. A decrease of pLTP was also recorded on hippocampal slices obtained 2 days after IH. This decrease was ameliorated by MK-801 injections prior to each IH session and restored by LFS depotentiation. Occlusion of pLTP and the observed decreased proportion of NMDA-only silent synapses after neonatal hLTP may explain long term memory, behavioral deficits and abnormal synaptogenesis and pruning following neonatal IH.

INTRODUCTION

Intermittent hypoxia (IH) is a common clinical condition in premature infants which occurs during apnea episodes due to immaturity of respiration control (Eichenwald, 2016). While small episodes of IH are considered to be of no clinical significance, more severe IH may have adverse effects on both structural and functional development in the cortex and hippocampus, and is also associated with a high risk of cognitive and language impairment (Poets, 2019). Animal models of chronic neonatal IH demonstrated impaired synaptogenesis (Cai et al., 2012) and hypomyelination of white and gray matter structures without significant cell death (Goussakov et al., 2019; Juliano et al., 2015) suggesting that mechanisms of the observed cognitive deficits after neonatal IH are related to functional synaptic impairment and disturbances to neuronal circuit maturation. In our previous study (Goussakov et al., 2019) early and persistent decrease of long term potentiation (LTP) after neonatal IH in P3-P7 mice was associated with increased synaptic excitability and memory deficits. We hypothesized that it could be a result of pathological synaptic potentiation occurring after neonatal IH by a mechanism similar to ischemic or anoxic LTP (iLTP) (Hammond et al., 1994). If this pathological potentiation persists to the late phase, it may interfere with normal activity dependent synaptic plasticity, and thus would explain memory deficits in Morris water maze and fear conditioning test in adult mice after neonatal IH (Goussakov et al., 2019).

The phenomenon of iLTP was first described in vitro (Crepel et al., 1993a; Crepel et al., 1993b) in the form of anoxia-aglycemia induced potentiation of field excitatory postsynaptic potentials (fEPSP) in the rat hippocampal CA1 region, mediated by NMDA receptors. It was soon recognized that this pathological form of synaptic potentiation shares similar mechanisms with physiological LTP (pLTP), induced by high frequency afferent electrical stimulation (Hammond et al., 1994; Hsu and Huang, 1997), which involves postsynaptic potentiation of both NMDA and AMPA receptor-mediated synaptic currents through the Ca²⁺ dependent processes (reviewed by (Di Filippo et al., 2008)). It has also been shown in vitro that iLTP and pLTP can influence each other; if one form of LTP is induced, the ability of the neurons to express the other form is occluded (Hsu and Huang, 1997; Lyubkin et al., 1997).

Apart from the indirect ex vivo evidence on brain slices obtained from animals after cardiac arrest (Orfila et al., 2018), there is currently a lack of evidence to support hypoxia/ischemia-induced LTP (hLTP) in live animal models. In the current study we explored whether severe neonatal IH induces early and late phases hypoxic LTP (E-hLTP and L-hLTP), and whether hLTP shares common mechanisms with pLTP. Second, we would like to determine whether pathological L-hLTP directly reduces synaptic plasticity by occluding pLTP.

METHODS

Experimental Neonatal Intermittent Hypoxia Model.

The study received approval by the Institutional Animal Care and Use Committees of NorthShore University HealthSystem.

C57BL/6J mice, originally obtained from the Jackson Laboratory (Bar Harbor, Maine), were bred at the NorthShore University HealthSystem animal facility. Neonatal mice of both genders from each litter were randomly assigned to the IH group or control group and subjected to IH paradigm as previously reported (Goussakov et al., 2019). Briefly, the IH paradigm consisted of intermittent exposure to humidified 5% O₂ balanced with 95% N₂ for 2.5 min followed by reoxygenation in room air for 10 min. During hypoxic exposure, pups were separated from their dams and placed into a 250 mL volume air-tight chamber in a temperature controlled neonatal incubator with an ambient temperature of 35 °C. The paradigm was initiated at postnatal day 3 (P3, day of birth considered as P0) and continued for 5 consecutive days. The number of daily hypoxic episodes was 20, 20, 16, 14, and 12, equally divided into morning and afternoon sessions. Between the sessions, mice were returned to their dams for 2 h for recovery and nursing. Control mice were also separated from their dams for the same period of time as the experimental animals, but kept in room air in the same incubator at 35 °C. Upon completion of IH, mice were taken for in vitro electrophysiology studies at P10-P15 and after 6 weeks. Experiments was conducted in a blind manner to IH group assignment.

Electrophysiology

In vivo fEPSP recording during intermittent hypoxia in neonatal mice

Neonatal mice at P5-P7 (n=5) were anesthetized via inhalation of 3% isoflurane delivered through a nose cone mask. The skull was exposed with a 10 mm skin incision and a small hole was drilled at coordinate 1.7 mm posterior bregma, 1.5 mm lateral central line. Three 12 MΩ tungsten electrodes, 127 μm diameter, 10 μm tip (#575400, A-M Systems) were glued together as an assembly and mounted onto a micromanipulator. The distance between bipolar stimulating electrodes was 50 μm and a recording electrode was placed 300 μm apart medially. The electrodes were lowered to approximately 1 mm below dura, directly above the dorsal hippocampus. The electrodes were further advanced by micrometer-steps during continued paired-pulse stimulation in order to get paired-pulse facilitation of recorded excitatory field potentials (fEPSP) as an indicator of correct electrodes placement in CA1 stratum radiatum. A 300 μM silver chloride reference electrode was inserted caudally under the skin and secured with acrylic glue onto the skull. Head was additionally immobilized with a cross-bar glued to the skull. The incision was irrigated with 2% lidocaine, and isoflurane was reduced and kept at 0.5% during recording, providing a sufficient level of anesthesia. Animal body temperature was kept at 35° during recording session by heating pad connected to temperature controller (TC-1000, CWE, Inc.). Respiration and heart rate was recorded with a pulse oximeter (MouseOx).The setup was stable for several hours of recording with a stable baseline. The setup did not cause distress to the animal based on absence of movements and a stable heart rate. Stimulation was delivered through an isolator (IsoFlex, I.M.P.I., Israel). Responses were recorded with Multiclamp 700B amplifier, digitized with Digidata 1440 and recorded in current clamp using pCLAMP-10 software (Molecular Devices, CA) with sampling rate 100 kHz, and low pass filter set at 5 kHz. Direct current (DC) potential recordings were obtained from same electrodes using common ground reference electrode. Recording of fEPSP was conducted during IH protocol and for several hours after with inter-stimulus intervals of 30 sec during IH and 1 min before and after IH. Gas mixture, either 21% O₂ or 5% O₂ (balanced with 95% N₂) was supplied from tanks, mixed with isoflurane and delivered to animal through the nose cone mask. After obtaining a stable baseline recording, the IH session started and consisted of 5 episodes of 5% O₂ hypoxia duration 2.5 min interleaved with 10 min of 21% O₂, similar to the IH protocol as above. After IH session, fEPSP recording continued for 4 hours, followed by a depotentiation protocol with low frequency stimulation or theta burst stimulations (2xTBS) and was recorded for 1 more hour.

Synaptic depotentiation with low frequency stimulation (LFS).

Depotentiation protocol consisted of 900, 0.1 sec square pulses, delivered at 2 Hz to Schaffer collateral/commissural – CA1 synapses. After one hour after LFS depotentiation, when fEPSP amplitude stabilized, pLTP was invoked with 2xTBS, as described above. The mean of the 10 min of fEPSP time course before TBS was taken as a baseline for calculation of depotentiation level and the last 10 min of recording after depotentiation was taken for calculation of LTP levels.

Physiological long term potentiation (pLTP) was elicited with two theta burst stimuli’s (2xTBS) separated by 1 min. Each TBS stimulation consisted of 10 trains with 4 pulses each, delivered at 100 Hz with 200 ms between the trains. Time course of fEPSP slope was recorded during 20 min before TBS and 60 min after to measure pLTP. LTP time course was calculated as percentage of baseline fEPSP slope for each recording and then averaged. For statistical comparison, the average of the last 10 min of each individual recording was taken.

After the recording, 1 mA, 15 second direct current was applied to produce a small electrolytic lesion. The mice were euthanized, and the brains were fixed and Nissl stained to confirm the location of the recording and stimulation site in CA1 stratum radiatum.

Ex vivo fEPSP recording after IH

Hippocampal slice preparation and maintenance.

Two days after completion of IH protocol, at P10, mice were deeply anesthetized with isoflurane (3%) and decapitated. Brains were rapidly removed and the right hippocampus, with adjoined cortex, was dissected out in ice cold artificial spinal fluid (aCSF) of the following composition (mM): 124 NaCl, 2.5 KCl, 25 NaHCO₃, 2 CaCl₂, 2 MgSO₄, 1.25 NaH₂PO₄, 10 HEPES, 10 D-glucose, bubbled with carbogen (95% O₂/5% CO₂) to maintain pH of 7.4. Osmolarity of this solution was 300±5 mOsm. Dorsal hippocampus slices were cut 300 μm thick using Leica vibrotome (VT1000S, Leica Biosystems). Slices were maintained in an incubation chamber filled with room temperature 21°C aCSF and bubbled with carbogen for at least 1 hour before recording. The same aCSF solution was used during recording and perfused 3 mL/min at 21°C. All drugs were purchased from (Sigma-Aldridge Inc., MO).

Field excitatory postsynaptic potential (fEPSP) recordings.

For extracellular recordings, slices were transferred to a recording chamber (model RC-24, Warner Instruments Hamden, CT, US) fEPSPs were recorded in stratum radiatum of CA1 region of hippocampus in response to electrical stimulation of Shaffer collateral/commissural axons with bipolar platinum-iridium electrode, 25 μm diameter of wires, placed approximately 300 μm apart from recording electrode. Recording electrodes were pulled from capillary tubes 1B150F-4 (World Precision Instruments Inc., FL) using P97 puller (Sutter Instrument, CA) and filled with aCSF (resistance 2–4 MΩ). Stimulus pulses with 0.1 ms width were delivered from stimulus isolator. Stimuli to evoke fEPSP were applied ones per minute in all recordings. Input-output curves were generated by increasing stimulus intensity in 20 μA increments from 0 to 300 μA and recording fEPSP responses. Stimulus intensity required to evoke fEPSPs of 40% of maximum amplitude (at the level of pop spike appearance) was determined and used for evoking all subsequent fEPSPs. Electrical signals were amplified using Axopatch-700B amplifier, digitized at the sampling rate 100 kHz, and recorded in current clamp using pCLAMP-10 software (Molecular Devices). fEPSP slope was used as a metric to compare individual fEPSPs changes in different experimental conditions. The difference in fEPSP or DC time course relative to baseline was tested with one sample t-test.

Whole-cell patch-clamp

Whole-cell patch clamp recordings were made in hippocampus CA1 pyramidal neurons in brain slices obtained from mice at 2 to 5 days after IH course and from age matched controls. Glass micropipettes were filled with intracellular solution of the following composition in mM; 83 K-gluconate, 30 KCl, 5 NaCl, 1 MgCl₂, 10 EGTA, 10 HEPES, 2 MgATP, and 0.1 NaGTP (adjusted to pH 7.3 with KOH), osmolarity290±5 mOsm. High Cl²⁺concentration was used to reproduce its intracellular concentration in neonatal animals. Filled electrodes had resistance of 2–4 MΏ. For all patch clamp recordings pipette capacitance and whole cell compensations was applied. Series resistance (Rs) was not compensated but monitored and recordings with more than 15% change in Rs were discarded.

Intrinsic and firing properties of CA1 pyramidal cell were measured, including input resistance (Rin), membrane potential (Vm), cell capacitance (Cm), firing threshold (FT), spikes amplitude (SA), spike width at the level of half amplitude (SW), after hyperpolarization (AHP) amplitude and duration (AHP-A and AHP-D), and spike on-of delta current (dI), determined as difference between threshold to spike current and stop spiking current in response to increasing and decreasing current ramps.

Synaptic depotentiation with low frequency stimulation and physiological long term potentiation protocols were the same as for in vivo experiments above. Paired pulse facilitation (PPF) was elicited by applying two consequent stimuli 70 ms apart. PPF was expressed as a ratio of amplitudes of the second fEPSP to the first.

AMPA and NMDA miniature currents (mEPSC) were measured in voltage clamp at −70 mV holding potential. In order to reveal and measure the NMDA component, the recording of mEPSC was first done in normal Mg²⁺ containing aCSF during 15 min and then after 15 min washing with Mg²⁺ free solution (Watt et al., 2000). The aCSF contained 1 μM tetrodotoxin (TTX) and 100 μM of picrotoxin (PTX) to prevent action potential and GABAergic events, and 20 μM glycine to saturate NMDA binding sites. Therefore, we recorded only glutamatergic events, confirmed by application of CNQX (50 μM) and AP5 (100 μM), AMPA and NMDA receptor blockers, respectively, at the end of each recording. Application of CNQX and AP5 completely removed miniature events confirming that all mEPSC’s were glutamatergic. 10 mM EGTA in pipette solution prevented occurrence of synaptic plasticity due to NMDA activation in Mg²⁺ free solution. AMPA current was measured at the maximum amplitude of response and NMDA current at 15 ms after stimulus artifact, where AMPA component of mEPSC is no longer present. mEPSCs recordings were analyzed using in-house software written on Matlab (MathWorks, Natick, MA, USA).

Measurement of silent synapses fraction by minimal stimulation technique.

Silent synapses are defined by a large failure rate of evoked EPSCs at a hyperpolarized membrane, comprised of AMPA receptors EPSCs, concurrent with a small failure rate to the same stimuli at depolarized membrane, comprised of NMDA or mixed NMDA and AMPA EPSCs (Liao et al., 1995; Marie et al., 2005). EPSCs were evoked using minimal stimulation (Isaac et al., 1995; Liao et al., 1995) and using method by (Hinds et al., 2003) with a fine tip (several micrometers) concentric electrode made from a patch pipette covered with electro-conductive carbon paint DAG-T-502 (Ted Pella, Inc., US). GABA receptors were blocked with 100 μM picrotoxin. Stimulus intensity, eliciting 50–60% synaptic responses at −70 mV, was determined during 50–100 consecutive trials. The same stimulation intensity was applied to evoke responses at +50 mV. Response failures were visually identified and counted. Failure rates to evoke EPSC at −70 and +50 mV were used to calculate the fraction of silent synapses assuming Poisson model of neurotransmitter release by using formula (1-ln(F₋₇₀)/ln(F₊₅₀)) (Liao et al., 1995) where F₋₇₀ and F₊₅₀ are the failure rates at −70 mV and at +50 mV membrane potential. AMPA and NMDA origination of EPSC’s were confirmed with application of CNQX (50 μM) and AP5 (100 μM), the inhibitors AMPA and NMDA channels respectively. For minimally evoked EPSC’s recordings K-gluconate was replaced with 110 Cs-methanesulfonate in equimolar concentration and pH was adjusted to 7.3 with CsOH. Osmolarity of intracellular solutions was adjusted to 290±5mOsm, patch electrode resistance was 2–4 MΩ. All recordings were performed at room temperature 21°C.

Drug treatments

MK-801, a selective and non-competitive NMDA receptor antagonist (0.25 mg/kg), and CMIQ, 4-Cyano-3-methylisoquinoline, a specific inhibitor of PKA, 0.02 mg per animal, (both from Sigma-Aldridge, MO) were applied via intracerebraventricularly injection. CMIQ was dissolved in 10% DMSO, 20% emulphor, and 70% saline, with the final concentration of 10 mg/ml. MK-801 was dissolved in sterile saline at final concentration 0.1 mg/ml. Drags were sterilized by filtering through 0.2 μm sterile syringe filter, aliquoted in sterile tubes and frozen at −20° C until use. Mice pups were anaesthetized with 3% isoflurane and 1μl of the drugs (CMIQ and MK-801) were injected with 33G needle through the skin and skull in each lateral ventricle 1 hr before IH session. For chronic IH, MK-801 was administered 1 hour before each IH session twice a day for five days from P3 to P7.

Manganese enhanced magnetic resonance imaging

Manganese enhanced MRI was performed on neonatal mice undergoing IH to reveal brain areas with increased intracellular Ca²⁺ influx. Mn²⁺ contrast agent entered neurons via calcium voltage gated channels (Malheiros, Paiva et al. 2015) and NMDA receptors (Itoh et al., 2008). MnCl₂ (60 mg/kg, i.p.) was administered to neonatal mice at P4 (n=5). The injected solution consisted of 100 mM MnCl₂ (Sigma-Aldrich) in 0.01M PBS. The repeated injection of MnCl₂ with the same dose was done at P6. Animals received IH course as described above and were repeatedly scanned before the injection, and 24, 48, 72 and 96 h after the first injection. Magnetic resonance imaging was performed on a 14.1 T Bruker Avance imaging spectrometer (Bruker, Billerica, MA) using a 20-mm bird cage resonator. Mice were sedated with 3% isoflurane inhalation in air for induction and 1.5% for maintenance. Animals’ respiration rate and rectal temperature were monitored with a small animal physiological monitor (SAII’s Small Animal Instruments, NY). Body temperature was kept at 35°C by maintaining ambient temperature at 32 °C using the spectrometer gradient temperature controller. A T1-weighted three-dimensional fast low angle shot (FLASH) magnetic resonance imaging sequence was used with the following parameters TR/TE 133/4/3 ms, flip angle = 90 degrees, 12 averages, in-plane resolution 61×61 microns, slice thickness 0.3 mm, 12 slices. Imaging time was 5 min and 14sec. To ensure consistent and reproducible ROI placement between groups and between sessions for the different age of the same animal, imaging slices were positioned using anatomical landmarks. The imaging plane was oriented perpendicular to the plane passing through lowest portions of anterior and posterior cortex. Six axial slices of variable thickness (depending on animal age) we placed between anterior commissure and front end of superior colliculi on a sagittal localizer. Regions of interest (ROI) were manually drawn on CA1, CA3 and dentate gyrus of hippocampus on a single slice using the thickest portion of corpus callosum splenium as a landmark (slice 3). Image intensities in ROIs were normalized to the global brain intensity on the same slice.

Statistical analysis.

Data are presented as means ± standard error of means. All variables were checked for normality and homogeneity (using Shapiro Wilk and Levene’s tests). The variables were compared across treatment groups using Student T-test or one-way ANOVA. Post-hoc group comparisons were performed using Tukey-Kramer method. Significance was determined by p values ≤ 0.05. The effect of the IH factor on normalized signal intensity in serial MEMRI measurements was tested using repeated measures ANOVA, separately for each hippocampus region. RM AVOVA was implemented as a general linear model with fixed Group and Time factors and random Subject factor nested within Group factor in Minitab 16 software.

RESULTS

In vivo neonatal IH induced synaptic potentiation in hippocampus.

We have previously reported an early and persistent decrease of long term potentiation deficits after neonatal IH in hippocampal slices, measured 3 days and 6 weeks after IH (Goussakov et al., 2019). In the current study we examine whether such phenomenon exists in live intact animal and what is the mechanism of the late hLTP. fEPSPs in response to Schaffer collaterals stimulation in hippocampal CA1 stratum radiatum were recorded in lightly anesthetized P5-P7 mouse pups before, during and after a single IH session consisting of 5 hypoxic episodes (2.5 min 5% O₂, 10 min 21%). fEPSPs after IH were significantly increased relative to the baseline (149.5.0±17.15 %, one sample t-test t₄=2.9 n=5, p=0.04) within 30 min after the end of last IH episode (Fig. 1A). This increase persisted for at least 4 hr, indicating that both early and late phase hypoxic LTP were induced by the IH in vivo. Application of LFS depotentiation protocol (arrow, Fig. 1A) returned fEPSP back to the baseline level. Application of NMDA and PKA inhibitors (CMIQ and MK-801) prior to IH session abolished hLTP induction (Fig. 1B), whereas the LFS protocol did not induce LTD in control animals, indicating that hLTP was normalized by depotentiation mechanisms. The effects of inhibitor applications and LFS depotentiation confirmed that L-hLTP, induced in response to hypoxia, is both NMDA and PKA-dependent and occurs by mechanisms consistent with those involved in neonatal pLTP.

Figure 1.

Synaptic potentiation after IH in vivo. A. Time course of fEPSP in CA1 of control neonatal mice (n=5) during and after IH, followed by LFS depotentiation protocol. Red bars indicate 2.5 min episodes of hypoxia. Timing of LFS depotentiation protocol initiation is indicated by blue arrow. B. Treatment with either NMDA inhibitor MK-801or PKA inhibitor CMIQ (n=5 each group) prior IH session abolished hLTP induction. C. Physiological LTP, induced by TBS stimulation four hours after IH session (n=6), was reduced relative to sham controls (n=5), indicating occlusion by L-hLTP. Inserts are showing representative fEPSPs a different time points. Scale bar on inserts 0.2 mV/ 20 ms.

Occlusion of pLTP by late hLTP in vivo.

Since neonatal IH results in, we tested whether hLTP interfere with pLTP, that could represent a mechanism explaining memory deficits, observed after neonatal IH (Goussakov et al., 2019). Four hours after IH, after stabilization of fEPSP response at a new level, application of 2xTBS elicited only 114.6±6.1 % of pLTP in IH group relative to the new baseline (Fig. 1C), that was significantly lower than 154.6±8.45 % pLTP in sham controls of the same age (t₉= 2.40, p=0.036), indicating occlusion of pLTP after L-hLTP.

Neuronal depolarization during IH is detected with direct current potential recording in vivo

It is important to determine physiological conditions leading to hLTP, in particular whether hLTP in neonates is related to neuronal depolarization, such in anoxic LTP in adults. Direct current (DC), potential recorded from the same electrodes used for fEPSP, dropped with each episode of IH to −29.0±6.6 mV at nadir of the last IH episode relative to baseline, t₄=4.41, p=0.015, followed by e recovery (Fig. 2). This observation indicates a transient neuronal depolarization of hippocampal neurons during IH (Hansen, 1978). At the end of IH session, DC potential gradually recovered and was not different from the baseline after IH (t₄ =1.27, p=0.26).

Figure. 2.

Intermittent decline of DC potential in hippocampus indicates neuronal depolarization during IH episodes in vivo at P5-P7 (n=5).

Increased Ca²⁺ uptake in hippocampus during IH on manganese enhanced MRI

As a result of hLTP, activation of NMDAR may lead to increased influx of Ca²⁺ to neurons that may trigger multiple downstream signaling pathways. We utilized manganese enhanced MRI with Mn²⁺ imaging contrast as an in vivo marker of neuronal Ca²⁺ uptake. Serial imaging of mouse pups was performed during IH course. Enhancement of hippocampal subfields was observed 24 hours after the systemic injection of MnCl₂ (Fig. 3) but in IH group the enhancement was significantly larger in CA1 (F_4,24=2.90, p=0.043, RM ANOVA) and CA3 (F_4,24=2.83, p=0.046, RM ANOVA). The contrast difference decreased on the subsequent scan (48 hours after the first contrast injection), presumably due to redistribution of the contrast. There was a difference in Group x Time interaction factor (p=0.43). The enhanced accumulation of the Mn²⁺contrast in IH group provides an evidence of increased neuronal Ca²⁺ uptake in hippocampal CA1 and CA3 during IH.

Figure 3.

Increase of intracellular Ca²⁺ uptake detected with manganese enhanced MRI during IH. Representative T1-weighted images of hippocampus in control (A) and IH (B) mouse pups at P7 (n=5). Regions of interest placement in hippocampus CA1 and CA3 is shown on A. C. Time course of signal intensity of selected ROIs, normalized to the total brain signal intensity. Mice were imaged before MnCl₂ contrast injection and 24, 48, 72 and 96 h after. Additional MnCl₂ contrast injection was administered at 48 hours after the first one. Imaging was performed at the same time for control and IH groups between sessions of IH (twice daily), indicated as red arrows.

Increased NMDA and AMPA receptors mediated currents and other synaptic properties after neonatal IH.

To examine what causes increase excitability of neurons after neonatal IH, we examine mEPSCs, intrinsic and firing properties of CA1 pyramidal neurons using patch clamp. Application of 100 μM NMDA selective inhibitor AP5 or ACSF with 2mM Mg ²⁺confirmed that mEPSC after 15 ms were mediated predominantly by NMDA receptors. The shape of mEPSCs’ curves in pyramidal CA1 cells in hippocampal slices was altered 2 days after IH, shown on Fig. 4, indicating increase of AMPA and NMDA receptor currents. When quantified by early and late components, (Fig. 4B, AMPA- and NMDA- dependent, correspondingly, see Methods), mean amplitude of AMPA mEPSCs in Mg²⁺ free aCSF was 15.1±2.0 pA in controls (n=5 cells, 5 mice) and 25.2±1.02 pA in IH (n=5 cells, 5 mice), t₈=4.53, p=0.0018. NMDA currents, measured at 15 ms after beginning of mEPSC, were 3.74±0.77 pA for control and 7.24±1.06 pA in IH, t₈=2.65, p=0.028. Mean frequency of mEPSCs was not different between control and IH groups (0.29±0.12 Hz vs. 0.19±0.14, t₈= 0.55, p=0.59).

Figure 4.

Increase of AMPA and NMDA miniature currents after neonatal IH (n=5) and in controls (n=5), measured at P10 in hippocampal slices. A. High concentration of Mg²⁺ or application of 100 μM NMDA selective inhibitor AP5 effectively abolished mEPSC after 15 ms B. Averaged mEPSCs after IH had larger amplitude. Insert is showing raw mEPSCs in control and IH slice recordings.. C. Significant increase both of AMPA and NMDA receptors – mediated currents in mEPSC.

We also examined basic and firing cellular properties of pyramidal neurons in CA1, which may indicate alterations in neuronal excitability (n =10 of IH, n=10 of controls, Table 1).

Table 1.

Basic and firing CA1 cells properties after IH (n = 5 controls, n = 10 IH).

	Rin, mΩ	Vm, mV	Cm, pF	Firing threshold, mV	Spike amplitude, mV	Spike width, ms	AHP amplitude, mV	AHP duration, ms	dI, pA
Control	244.6 ± 37.7	−62.0±1.4	43.6 ± 3.2	34.7 ± 3.3	67.5 ± 6.6	1.38 ± 0.02	−10.5 ± 3.5	66.7 ± 2.9	18.2 ± 5.6
IH	200.5 ± 25.7	−64.2 ± 1.8	47.0 ± 4.8	36.5 ± 1.4	73.5 ± 2.8	1.40 ± 0.15	−9.8 ± 1.1	73.0 ± 3.8	18.5 ± 4.4
t-test p-val.	0.33	0.45	0.64	0.73	0.34	0.28	0.86	0.73	0.91

No differences between control and IH groups was found in input resistance, membrane potential, cell capacitance, firing threshold, spikes width and amplitude, after hyperpolarization amplitude and duration, and spike on-of delta current. Consistent with unchanged after hyperpolarization parameters, IF curves showed no difference in firing properties between controls and IH (Fig. 5, controls slope 0.026 Hz/pA, 95% CI 0.023–0.028, IH slope 0.028 Hz/pA, 95% CI 0.026–0.030). This data suggest that intrinsic and firing properties of CA1 pyramidal cells were not changed after IH and observed increased excitability is due to synaptic changes.

Figure 5.

Injected current – spike frequency dependence, measured at P10 in hippocampal slices in controls (n=5) and after IH (n=5). No difference found.

NMDA inhibitor administered before IH ameliorated pLTP impairment.

To test whether previously reported persistent decrease of pLTP in hippocampus CA1 after IH (Goussakov et al., 2019) involved occlusion by NMDA dependent hLTP mechanisms, NMDA channel inhibitor MK-801 was administered to pups daily before each IH session, and pLTP was measured on hippocampal slice preparation 2 days after completion of IH protocol (Fig. 6A). MK-801 application significantly improved pLTP (ANOVA F_3,29 = 5.75, p=0.0032, post-hoc MK-801 – 188.8±14.1%, n=10 vs. saline vehicle +IH group 117.08±18.3%, p=0.021, n=6. Consistent with the previous observation, the average level of LTP was significantly decreased in IH – 128.07±15.8%, n=8, p=0.045 and in IH with saline vehicle groups, n=6, p=0.022 relative to saline controls in room air – 188.3±14.9%. The pLTP in MK-801 treated group was not significantly different from controls, post-hoc p=0.99. Ability of NMDA inhibition to recover pLTP indicates that a mechanism of pLTP decrease after neonatal IH is occlusion by NMDA receptor dependent L-hLTP.

Figure 6.

Amelioration of pLTP impairment after neonatal IH by NMDAR inhibition, measured on hippocampal slices obtained 2–5 days after P3-P7 IH protocol. pLTP after TBS was reduced after IH (n=10) relative to naïve controls (n=7). MK-801 (n=11) and saline vehicle controls (n=5) were injected between P3-P7, 1 hour before each IH session. A. pLTP was induced with TBS on hippocampal slices 2 days after completion of IH protocol. Inserts are showing fEPSPs at different stages of experiment.

B. Application of MK-801 reduced synaptic hyperexcitability measured by input-out curves. Data were obtained from the same animals as in A.

Increased synaptic excitability in IH animals (Fig. 6B), quantified as amplitude of fEPSP at the level of maximal stimulation intensity 300 μA on input-output curves (ANOVA F_3,29 = 6.39, p=0.0018, post-hoc control −1.39±0.27 mV vs. hypoxia −2.47±0.22 mV, p=0.023) was normalized after MK-801 treatment (−1.41±0.27, post-hoc vs. controls p=0.99). Pair-pulse facilitation in control groups was 1.61±0.06 and was not different from in IH and IH+MK-801 groups (ANOVA F_3,41 = 0.66, p=0.57). Together with unchanged mEPSCs frequency, this observation indicates that release probability of glutamate was not altered.

Depotentiation with low frequency stimulation reverses synaptic potentiation and ameliorates pLTP deficits after IH.

To further elucidate mechanism of hLTP and occlusion of pLTP, and possible for therapeutic intervention, we used LFS for synaptic depotentiation in vivo in control mice (n=8) and 2–5 days after IH protocol (n=8). We first tested whether or not LFS induces long term depression (LTD) in neonatal mice by using three protocols with 900 stimulation pulses delivered at 0.5, 1 and 2 Hz to Schaffer collarals-CA1 synapses. In control animals, each protocol evoked a transient short term depression, but not LTD, with fEPSP recovering to the baseline within an hour (Figure 7A.). This observation was consistent with previous reports that 1–10 Hz LFS did not trigger LTD in mice CA1 (Goh and Manahan-Vaughan, 2013). After LFS and fEPSP recovery, LTP could be successfully induced by TBS (Fig. 7A). Subsequent LFS returned this potentiation to the baseline level, indicating that the 2Hz LFS protocol reverses LTP induced potentiation in neonatal P10-P15 mice by depotentiation mechanisms.

Figure 7.

Depotentiation with low-frequency stimulation (LFS) reversed hLTP and restored pLTP in Shaffer collaterals - CA1 synapses A. A typical time course of LFS, followed by pLTP induction by TBS, and then by second LFS in an individual P12 control animal. LSF did not induce LTD, but depotentiated LTP, induced by TBS. B. Time course of sequence LFS-TBS pLTP in hippocampal slices of P10-P15 control (n=5) and mice after IH (n=8). Only transient short term depression with LFS was observed in controls. LFS in slices after IH induced depotentiation and restored capacity to induce LTP. C. Summary of depotentiation levels and subsequent pLTP at 2–5 days after IH.

To test hypothesis that some synapses had been potentiated by L-hLTP during IH, depotentiation protocol with 2 Hz LFS was applied to hippocampal slices at 2–5 days (Fig. 7B) after neonatal IH. While the LFS protocol did not elicit LTD in control animals (94.9±10.2% of baseline, t₄= 0.66, p=0.54, n=5), it evoked depotentiation in IH slices at 65.3±4.7% of baseline, t₇=7.89, p<0.001, n=8. Subsequent TBS induced LTP was normalized in IH slices after depotentiation (239.1±29.8 % in IH vs. 203.2±12.3 % LTP in control, t₁₁=−0.9, p=0.38, Fig. 7C). The experiment indicates presence of potentiated synapses after neonatal IH that can be depotentiated, thus restoring synaptic capacity for pLTP, shown to be impaired in our previous study (Goussakov et al., 2019).

Neonatal intermittent hypoxia reduces number of silent synapses

One of the possible mechanisms, explaining decrease of pLTP and possible long term memory impairment, could be abnormal synaptic unsilencing as a result of hLTP after IH. A proportion of NMDA-only silent synapses in hippocampal CA1 was measured in P10-P13 mice using the difference in failure rate to invoke synaptic response with a given stimulation intensity at hyperpolarized and depolarized membrane potentials (Liao et al., 1995), Fig. 8A. Stimulus intensity −70 mV was set to produce failure rates at about 50% (no difference between controls and IH, p=0.12).The difference in failure rate between hyperpolarized and depolarized membrane states was lower in IH slices than in controls, Fig 8B (10.4 ± 2.0%, n = 10 cells from 5 mice vs. 27.9 ± 2.6%, n=10 cells from 6 mice, t₁₈=4.74, p<0.001), and resulted in the decrease of the estimated proportion of silent synapses after neonatal IH (Fig.8C: 54.1±2.9 after IH and 24.2±4.2 with t₈=4.63, p=0.0017), that may limit capacity for pLTP.

Figure. 8.

Decrease of functionally silent synapses in hippocampus at P10-P15 after neonatal IH. A. Failures to produce EPSC response to afferent stimulation, indicated as open markers, are more frequent in hyperpolarized membrane state −70 mV (responses shown in insert as red, black curves indicate response failures) than in depolarized state +50mV, presumably due to the presence of synapses containing only NMDA receptors. B. Difference in the failure rate was decreased after neonatal IH. C. Percent of silent NMDA-only synapses in P10-P15 hippocampus, estimated from the failure rates, was lower after neonatal IH. Measurements were obtained from 10 cells from 5 IH mice and 10 cells from 6 control mice.

DISCUSSION

In the current study, we provided evidence for the first time that early, and most importantly, late phase hLTP occurs during neonatal IH, including direct demonstration of synaptic potentiation and depotentiation by in vivo and ex vivo fEPSP recording in hippocampus CA1. We confirmed that L-hLTP shared common pathways and occluded normal pLTP. Hypoxic LTP led to premature functional activation of silent synapses that may result in abnormal maturation of hippocampal circuits with long term adverse effect on learning and memory.

It has been established that normal pLTP and pathological iLTP share common molecular mechanisms leading to an increase of synaptic strength. This increase involves NMDAR activation after hypoxia/ischemia induced glutamate release and Ca²⁺ entry due to synchronized presynaptic and postsynaptic membrane depolarization (Di Filippo et al., 2008; Hammond et al., 1994; Maggio et al., 2015). Pre-synaptic membrane depolarization occurs with the arrival of action potentials after stimulation of synaptic afferents in the case of pLTP, or due to the Na⁺/K⁺-ATPase failure and inability to support pre- and postsynaptic resting potential in case of iLTP. The question remained whether or not such membrane depolarization occurs during IH episodes and weather this depolarization may induce sequence E-hLTP – L-hLTP within a clinically relevant time in live neonates, which are known to tolerate brain hypoxia much better than adults (Singer, 1999). Our experiments with in vivo direct current potential recordings and Mn²⁺ enhanced fMRI revealed periods of hippocampal CA1 cells depolarization during IH and intracellular Ca²⁺ entry, which is consistent with the established triggers of iLTP. A transient neuronal membrane depolarization, about 30% from the resting level which fully recovered in 10 min with reoxygenation, has been shown in hippocampal slices during anoxic episodes that resulted in synaptic potentiation later (Hsu and Huang, 1997). A similar magnitude of DC potential drop, up to −30mV, occurred at the end of IH episodes in our in vivo recordings, indicating that physiological conditions created during IH episodes in the hippocampus were sufficient for neuronal depolarization in order to induce sequence E-hLTP - L-hLTP. Negative DC potential deflection has been shown to reflect an neuronal membrane depolarization due to dissipation of neuronal membrane potential by an increase of extracellular potassium during anoxic depolarization (Hansen, 1978).

Recent data demonstrate that during early ischemia, ion channels remain functional and neurons are prone to increased spontaneous spike generation (Du et al., 2018). Transient partial depolarization in IH facilitates spontaneous spiking and a release of glutamate that, with simultaneous depolarization of pre- and post-synapses, triggers a Ca²⁺ influx via NMDARs. Such a condition may result in potentiation of CA3-to-CA1 synapses, similar by mechanisms to the classical Hebbian learning paradigm occurring in pLTP. Notably, the magnitude of hypoxic depolarization increased with subsequent repeated episodes of IH in vivo, suggesting that conditions for hLTP may require certain severity of hypoxia in brain. Accordingly, adverse long-term neurological and cognitive outcome in premature infants is associated mostly in cases of severe IH with prolonged apnea that lasts for more than 15–20 s, especially associated with bradycardia (Janvier et al., 2004; Pillekamp et al., 2007; Poets, 2019).

Since hLTP after neonatal IH has not been shown previously, we examined if it shares common mechanisms with pLTP. Neonatal L-hLTP was manifested by increased amplitudes of both AMPA and NMDA-mediated currents and was abolished by NMDA blocker on in vivo recordings. Elevated AMPAR and NMDAR currents can be attributed to increased channels’ conductance or increased channels’ density. To elucidate this mechanism, levels of AMPAR or NMDAR proteins could have been measured and this constitute a limitation of the study. Increased synaptic excitability in hippocampal slices obtained 2–5 days after IH was reversed by LFS depotentiation, commonly recognized as evidence that LTP has occurred (Jouvenceau et al., 2003). By comparing paired-pulse facilitation and mEPSCs’ frequency, we also established that L-hLTP (same as pLTP in CA1) is postsynaptic. A major difference between neonatal and adult LTP induction is the key role of PKA in the first postnatal week, instead of CamKII in adults (Yasuda et al., 2003). Around the second week of postnatal age, LTP is shown to depend on concurrent activation of CamKII and PKA (Wikstrom et al., 2003). In our recordings of live animals, hLTP after IH was abolished with prior treatment of animals with PKA inhibitor CMIQ. In addition to establishing NMDAR and PKA dependence of hLTP, we demonstrated by Mn²⁺ enhanced fMRI an increased intracellular Ca²⁺ entry after IH. This is a key event which triggers a molecular cascade and modification of the postsynaptic membrane, shared in all NMDA dependent forms of LTP. Therefore, despite differences in triggering, it is evident that hLTP after IH shares similar mechanisms with iLTP, described with in vitro anoxia and with normal pLTP (Hammond et al., 1994; Hsu and Huang, 1997).

Most of the evidence for iLTP originates from in vitro studies on hippocampal slices (Crepel et al., 1993b; Hammond et al., 1994) where early LTP was observed within a short time after OGD. This leaves the question of whether or not iLTP or hLTP occurs in live animals and, importantly, persists for a long enough time to interfere with physiological synaptic plasticity. Indirect ex vivo evidence of late phase iLTP has been observed in juvenile mice in hippocampal slices, weeks after a single episode of cardiac arrest (Dietz et al., 2018) by increased GluR1 and mEPSP amplitudes and by ability to restore elevated fEPSP to control levels with LFS depotentiation (Orfila et al., 2018). We now directly demonstrated, with continuous recording after IH in vivo, that E-hLTP transitioned to L-hLTP. This lasted for at least four hours, suggestive of de novo protein synthesis required for its maintenance and direct contribution to activity dependent synaptic plasticity impairment and memory deficits.

Since hLTP and iLTP share the same molecular mechanisms with pLTP, it has been suggested and supported by in vitro studies that physiologically relevant pLTP induced with TBS stimulation, and iLTP after OGD equally occluded each other (Hsu and Huang, 1997; Lyubkin et al., 1997). The occlusion of pLTP by L-hLTP was also directly observed in our data with the continuous in vivo recording after IH. Impairment of pLTP and increased synaptic excitability 2–5 days after IH in vivo were ameliorated by short term application of NMDA inhibitor before IH sessions and was restored after LFS depotentiation without causing long term synaptic depression in control animals. While the occlusion of synaptic plasticity after iLTP is a well-established phenomenon (Maggio et al., 2015), the exact mechanisms are largely unknown. It is believed that iLTP occludes pLTP through saturation of plasticity mechanisms, in particular, by limiting capacity in the postsynaptic membrane to incorporate new AMPA receptors after being potentiated by iLTP. However, there is no experimental support for this notion. It is unclear why AMPA receptors “get stuck” for days and weeks in potentiated post-synapse (Holmes and Ben-Ari, 2007). Limited capacity to incorporate AMPA for induction of new LTP or altered PKA/CamKII activity (Wang et al., 2014) after iLTP may directly explain short term, but is unlikely to explain permanent pLTP occlusion by L-hLTP. Gradual recovery of impaired synaptic plasticity has been described in juvenile mice after iLTP induced by cardiac arrest (Dietz et al., 2018). In our neonatal IH model, a decreased pLTP and memory deficits were found 6 weeks after neonatal IH (Goussakov et al., 2019), suggesting that additional mechanisms may be involved in long term impairment of learning and memory. Therefore further study focused on synaptogenesis and pruning after neonatal IH are needed

While both pLTP and hLTP increase synaptic strength, pLTP induces input-specific activation of distinct synapses, thus forming a basis of associative learning a memory. In contrast, hLTP results from an indiscriminate and non-input-specific Ca²⁺ influx, therefore inducing synaptic potentiation nonspecific for learning and memory. Besides limiting synaptic capacity for novel pLTP induction, neonatal hLTP may interfere with synaptic pruning and synaptogenesis and, therefore, neuronal circuit formation, which occurs during a critical period of brain development. Taking place during this period is an activity dependent, coordinated process of strengthening active connections, and an elimination of inactive connections including remaining silent synapses (Hensch, 2004). Functionally silent synapses, abundant in this period, are being unsilenced by incorporation of AMPA receptors after induction of pLTP in a timely and activity dependent manner (Durand et al., 1996; Fox, 1995). According to our data, abnormal premature synaptic unsilencing can also be triggered by hLTP during IH, limiting the pool of synapses available for pLTP (Kerchner and Nicoll, 2008) and altering normal activity dependent synaptic maturation by stabilizing abnormally potentiated synapses and preventing its elimination. As a result, hyper-connected networks may form (Kerchner and Nicoll, 2008), contributing to cognitive, behavioral, and memory deficits. Due to the global nature of IH insult, it may affect several cortical and subcortical networks. Premature synaptic unsilencing in neonates has been also implicated in disruption of subplate neuron development (Kanold et al., 2019) and auditory cortical patterning during this critical period of plasticity (Sun et al., 2018).

In conclusion, our study provides novel insights on alteration of activity dependent synaptic plasticity induced by a chronic IH in neonates. Triggered by IH, L-hLTP induced occlusion of pLTP may occur by direct saturation of synaptic plasticity and disrupt normal activity of dependent synaptic and neuronal circuit maturation by global synaptic potentiation, including abnormal synaptic unsilencing. These mechanisms are likely to play an important role in the sequence of events that finally result in abnormal hippocampal and, possibly, cortical network formation, contributing to long term cognitive and memory deficits.

Abbreviations

IH

intermittent hypoxia

pLTP

physiological LTP

hLTP

hypoxic LTP

fEPSP

field excitatory postsynaptic potentials

DC potential

direct current potential

LFS

low-frequency stimulation

(ROI)

Regions of interest

LTP

long term potentiation

iLTP

ischemic LTP

E-hLTP and L-hLTP

early and late phases hypoxic LTP

TBS

theta burst stimulations

mEPSC

miniature excitatory synaptic currents

LTD

long term depression