Heparan sulfate (HS) proteoglycans represent a major component of the extracellular matrix and are critical for brain development. However, their function in the mature brain remains to be characterized. Here, acute enzymatic digestion of HS side chains was used to uncover how HSs support hippocampal function in vitro and in vivo. We found that long-term potentiation (LTP) of synaptic transmission at CA3–CA1 Schaffer collateral synapses was impaired after removal of highly sulfated HSs with heparinase 1. This reduction was associated with decreased Ca2+ influx during LTP induction, which was the consequence of a reduced excitability of CA1 pyramidal neurons. At the subcellular level, heparinase treatment resulted in reorganization of the distal axon initial segment, as detected by a reduction in ankyrin G expression. In vivo, digestion of HSs impaired context discrimination in a fear conditioning paradigm and oscillatory network activity in the low theta band after fear conditioning. Thus, HSs maintain neuronal excitability and, as a consequence, support synaptic plasticity and learning.
Heparan sulfate proteoglycans (HSPG) are present throughout the brain and are located on cell membranes but also in the extracellular matrix (ECM). HSPGs represent a heterogeneous group of proteins, among them the ECM molecules agrin and perlecan as well as cell surface molecules like syndecans and glypicans (Sarrazin et al. 2011). A common feature of HSPGs is covalently bound side chains of heparan sulfates (HSs), which are often the sites of interactions with ligands such as growth factors (Sarrazin et al. 2011; Matsuo and Kimura-Yoshida 2014).
These HSs are unbranched polysaccharides that are synthesized from N-glucuronic acid and N-acetylglucosamine before extensive modification by, for instance, sulfation. HSPGs play a well-established role during brain development. They are important for patterning of the brain, neurogenesis and crucial for axonal guidance (Yamaguchi et al. 2010). Consequently, most studies of HSPGs have focused on developmental aspects of HSPG signaling.
In addition, HSPGs can profoundly modulate memory and learning in adult animals and synaptic long-term plasticity in acute slices prepared from these (Senkov et al. 2014). Interestingly, a recent study demonstrated that conditional ablation of Ext1, a gene involved in HS synthesis, from a subpopulation of pyramidal neurons leads to an autistic phenotype (Irie et al. 2012). These findings indicate that HSPGs and HSs shape brain function on many levels from cellular properties to complex behavior.
However, both a developmental phenotype and the acute absence of the HSPGs or HSs may underlie such a deficit in the adult. Acute enzymatic digestion of ECM components has been used to identify specific roles of ECM components and to isolate their acute contribution to adult brain functions (Senkov et al. 2014). For example, enzymatic digestion of hyaluronic acid in intact brain tissue can be achieved within only 2 h (Kochlamazashvili et al. 2010), thus avoiding potential developmental side effects of genetic manipulations. Similar to hyaluronic acid, the function of HSs has been studied mainly by either genetic ablation of the HS bearing proteoglycans or enzymatic digestion of HS side chains. In dissociated hippocampal cultures, enzymatic removal of highly sulfated HSs with heparinase 1 affects synaptic scaling and decreases the mean firing rate of neurons (Korotchenko et al. 2014; Garau et al. 2015). In acute hippocampal slices, treatment with a combination of heparinases 1 and 3, which supposedly removes both low and highly sulfated HSs, impairs long-term potentiation (LTP) of synaptic transmission (Lauri et al. 1999; Dityatev et al. 2004). However, the type of involved HSs and the cellular and network mechanisms underlying this LTP impairment have remained unclear. In addition, it is mostly unknown how HSs acutely affect behavior or network activity in vivo.
Here, we addressed these open questions by a combination of electrophysiological, imaging, and behavioral experiments. Our results in vitro show that removal of highly sulfated HSs reduces CA1 pyramidal cell excitability, which in turn leads to decreased dendritic and spine Ca2+ entry and impaired LTP. In vivo, the reduced excitability is associated with altered oscillatory network activity in the low theta range and abolished contextual discrimination in a fear conditioning paradigm.
Materials and Methods
Slice Preparation and Heparinase Treatment
Acute hippocampal slices were prepared from 5 to 8-week-old male C57BL/6 mice and Thy1-YFP (Feng et al. 2000) mice as previously described (Kochlamazashvili et al. 2010; Anders et al. 2014). Briefly, acute hippocampal slices were obtained in full compliance with national and institutional guidelines on animal experimentation. Slices, 350 µm thick unless stated otherwise, were prepared in an ice-cold slicing solution containing (in mM): NaCl 60, sucrose 105, KCl 2.5, MgCl2 7, NaH2PO4 1.25, ascorbic acid 1.3, sodium pyruvate 3, NaHCO3 26, CaCl2 0.5, and glucose 10 (osmolarity 300–305 mOsm), and kept in the slicing solution at 34 °C for 15 min before being stored at room temperature (RT, 21–23 °C) in an extracellular solution containing (in mM) NaCl 126, KCl 2.5, MgSO4 1.3, NaH2PO4 1.25, NaHCO3 26, CaCl2 2, and glucose 10. This solution was also used for recordings at RT. Heparinase 1 from Flavobacterium heparinum (Sigma H2519, 1 U/mL) was used for enzymatic removal of HSs from hippocampal slices in custom-made aerated incubation chambers (2 mL volume, at 34 °C). For sham treatment, a subset of heparinase aliquots was heat-inactivated for 120 min at 60 °C. All solutions were continuously bubbled with 95% O2/5% CO2.
In Vitro Electrophysiology
For recordings, hippocampal slices were transferred to a submersion-type recording chamber and superfused with extracellular solution at RT. For extracellular recordings, standard patch pipettes (3–4 MΩ) were filled with extracellular solution and inserted into the CA1 stratum radiatum. Whole-cell recordings from CA1 pyramidal cells were obtained using standard patch pipettes (3–4 MΩ) filled with an intracellular solution containing (in mM) KCH3O3S 135, HEPES 10, di-Tris-Phosphocreatine 10, MgCl2 4, Na2-ATP 4, Na-GTP 0.4 (pH adjusted to 7.2 using KOH, osmolarity 290–295 mOsm). Membrane-impermeable dyes Oregon Green 488 BAPTA-1 (OGB1, 200 µM, Invitrogen), OGB6F (250 µM), Fluo-4 (200 µM), and Alexa Fluor 594 hydrazide (20 µM, Invitrogen) were added to the intracellular solution as indicated. In a subset of recordings, KCH3O3S was replaced by equimolar CsCH3O3S. Data were recorded using MultiClamp 700B amplifiers, digitized (10–40 kHz) and stored for offline analysis. Whole-cell recordings were rejected if at any time the access resistance exceeded 20 MΩ or changed by more than 20%. Bridge balance and series resistance compensation was used as appropriate. For stimulation experiments, a bipolar concentric stimulation electrode was placed in the “stratum radiatum” at the border between CA2/3 and CA1. For theta-burst stimulations (TBS), 8 bursts consisting of 4 stimuli (0.1 ms each, 100 Hz) were delivered at 5 Hz (a single TBS cycle). The pattern was repeated up to 5 times (5x TBS) with an interval of 1 min between individual TBSs. For LTP experiments presented in Figure 1, the stimulation intensity was adjusted to obtain field excitatory postsynaptic potentials (fEPSPs) with a slope of 50% of the subthreshold maximum (subthreshold maximum: fastest-rising fEPSP without a clearly visible population spike in the decay phase) (Kochlamazashvili et al. 2010). For experiments presented in Figures 2 and 4, the stimulus intensity was set to 4 times the intensity needed to elicit a single action potential by a single test burst of 5 stimuli at 100 Hz to compensate for variable stimulation efficiencies between slices (Kochlamazashvili et al. 2010). In some experiments, N-methyl-d-aspartate receptor (NMDAR)-mediated fEPSPs were recorded in the presence of 0.1 mM Mg2+ and isolated pharmacologically by the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR)-inhibitor NBQX. In these experiments, a cut was made between CA3 and CA1.
Channel or receptor blockers D-AP5 (50 µM, Abcam), NBQX disodium salt (10 µM, Abcam), CGP52432 (5 µM, Abcam), nifedipine (20 µM, Sigma-Aldrich), picrotoxin (PTX, 100 µM, Abcam), and tetrodotoxin (TTX, 1 µM, Tocris) were added to the extracellular solution as indicated.
Immunocytochemistry and Immunohistochemistry
Mouse hippocampal neurons were cultured for 24 days in vitro and incubated with 0.5 U/mL of heparinase 1 for 2 h at 37 °C as described previously with some modifications (Dityatev et al, 2004) and fixed with 4% paraformaldehyde (PFA), permeabilized with 0.1% Triton-X-100 in phosphate-buffered saline (PBS) for 10 min, washed 3 times and blocked (0.1% Glycine + 0.1% Tween-20 + 10% Normal Goat Serum in PBS) for 45 min at RT. Anti-HS antibody A04B08V (dilution 1:2; the antibody was purified against mouse skeletal muscle HS and supposed to recognize 2-O and 6-O high sulfate group) (Dam et al. 2003) diluted in blocking buffer was applied for overnight followed by 3 washes with PBS. Incubation with secondary antibody (anti-VSV-G, clone P5D4, dilution 1:100) and tertiary antibody (Mouse, Alexa 488, dilution 1:1000) was carried out for 3 and 1 h, respectively. Coverslips were subsequently washed with PBS, incubated with DAPI (Invitrogen, reference number-D1306) for 10 min and mounted using fluoromount aqueous mounting media (Sigma-F4680). For hippocampal sections, the similar procedure was applied but permeabilization step was carried out with 0.5% Triton-X-100 in PBS for 10 min and anti-VSV-G, clone P5D4 antibody was used with the dilution 1:50.
For ankyrin G (AnkG) labeling, acute horizontal hippocampal slices (250 µm thick) of three 52–58-day-old Thy1-YFP animals (Feng et al. 2000) were prepared and treated with heparinase as described above. After heparinase/control incubation, the slices were immersion-fixed in 4% PFA in PBS (pH 7.4) at 4 °C for 2 h. Next, slices were washed for 3 × 30 min in PBS before incubation in a blocking solution (5% normal goat serum and 0.1% Triton-X-100 in PBS) at RT for 6 h. Primary antibodies against GFP (host: chicken, 1:5000, Abcam, ab13970) and AnkG (mouse monoclonal, 1:500, Neuromab, clone N106/36) were incubated in blocking solution at 4 °C for 40 h. After a 3 × 30 min PBS wash, secondary antibodies (goat antichicken Alexa Fluor 488, goat antimouse Alexa Fluor 568, both 1:200, Invitrogen) were incubated at RT for 2 h. Slices were again washed 3 × 30 min in PBS before being mounted with ProLong Gold Antifade (Invitrogen). Imaging was performed on a Leica SP8 confocal microscope with a glycerol immersion 63x, 1.3 NA objective and hybrid detectors. Initial image analysis was done by an experimenter blinded to the experimental conditions using ImageJ (NIH) and Microsoft Excel. CA1 pyramidal cell axon initial segments (AISs) were identified using the YFP-label and their typical morphology and confirmed by AnkG labeling. Axonal AnkG expression profiles were obtained from image stacks by average intensity projections of all imaging planes containing AnkG signal (max stack z-depth 5 µm). A line profile was then drawn along the AIS starting from the center of the soma and finishing just beyond the most distal AnkG signal (typically 40–45 µm from the soma). Further details are provided in Figure 5 and the corresponding Results section.
For AnkG labeling in dissociated cultures, mouse hippocampal cultures were treated with heparinase and heat-inactivated heparinase (control) as described above after 21 days in vitro (DIV21). Subsequently, cells were quickly rinsed with warm PBS and subjected to above-mentioned immunocytochemistry protocol to stain with antibodies against AnkG (mouse monoclonal, 1:1000) and MAP2 (guinea pig, 1:500). Imaging was performed with the experimenter blinded to the experimental groups on a Zeiss LSM 700 confocal microscope with a 63x/1.4 NA oil immersion objective. Image analysis was done using ImageJ (NIH) and Microsoft Excel. AISs were detected as AnkG positive immunofluorescence starting from the edge of the soma as defined by MAP2 positive immunostaining. Of note, 40 µm long fluorescence intensity line profiles were taken along the AIS starting from the edge of the soma.
In Vivo Electrophysiology, Intrahippocampal Injection and Behavioral Testing
Adult (2–4-month old) male C57BL/6 J mice (Charles River) were used. At least 1 week before start of the experiments, mice were transferred to a small vivarium, where they were housed individually with food and water ad libitum on a reversed 12:12 light/dark cycle (light on at 9:00 PM). All behavioral experiments were performed in the afternoons during the dark phase of the cycle when mice are active, under constant temperature (22 ± 1 °C) and humidity (55 ± 5%). All treatments and behavioral procedures were conducted in accordance with ethical animal research standards defined by German law and approved by the Ethical Committee on Animal Health and Care of the State of Saxony-Anhalt, Germany, the licence number 42502-2-1159 DZNE.
Fabrication of Electrodes and Cannulas
We have used a previously described hybrid infusion-multielectrode (HIME) recording system for acute drug delivery at the place of acquisition of neuronal activity (Senkov et al. 2015). In short, a 5 electrodes bundle was made of tungsten wires insulated with Teflon and bound together in a stainless steel tubing. Electrode tips were arranged to aim to different major hippocampal layers, using a decrement of 250 μm between each electrode, starting from 2 mm length, to reach molecular layer of the dentate gyrus (DG), and ending with the shortest electrode, 1–1.1 mm, to record from stratum oriens and pyramidale. For fabrication of guide cannulas, we used the same stainless steel tubing as for electrodes.
Implantation of Electrodes and Cannulas
Chronic implantation of electrodes and cannulas in mice was performed as described previously (Kochlamazashvili et al. 2010; Senkov et al. 2015) (see Fig. 7a for a schematic of electrode and cannula placement). In brief, mice were anesthetized with 1–3% isoflurane (Baxter International, Inc.). Preparation for implantation included injection and application of analgesics, head hair shaving, and skin cleaning and application of an antiseptic as previously described (Senkov et al. 2015) before the mouse was placed in a stereotaxic frame (Narishige). All following procedures were performed under a surgical binocular microscope and on a heating pad to maintain mouse body temperature constant (34–36 °C) over surgery. A 10 mm diameter circular scalp incision was done and exposed skull bone was cleaned and dried. After marking coordinates for implantation, 3 small holes for anchoring and ground screws were drilled in 2 frontal bones and one in the left parietal at the border with the interparietal bone. Two frontal screws and a parietal screw were connected with a ground/reference wire. The left cannula was stereotactically implanted after drilling another hole and by using a self-made universal holder and fixed using acrylic. Next, hippocampal electrodes were implanted in the right parietal bone. After hippocampal electrodes, the right cannula was similarly implanted and secured with acrylic. Coordinates for bilateral cannulas: AP = 2.0 mm and L = ±2.2 mm from Bregma and midline, respectively. Coordinates for the 5 electrodes bundle reaching different layers of the hippocampus were: AP = −2.0 mm from Bregma and L = ±1.25 mm; DV = 2.0 mm from the brain surface. The whole system was secured to the bone with dental cement and the Neuralynx EIB-16 headstage was screwed into position. Finally, tungsten wire tails from electrodes were gently connected to the headstage. After the assembly was completed, all ground wires were soldered together and the upper cover “hat” as a Faraday cage was placed above the bottom pedestal and fixed to it with 2 small screws. Surgery typically lasted 90 min. Mice were allowed to recover for at least 5–7 days before behavioral test and recordings (postoperative analgesic: carprofen, 5 mg/kg b.w. s.c.).
For intrahippocampal injection, we used a digitally controlled infusion system (UltraMicroPump, UMP3, and Micro4 Controller, WPI) fed with a 10 μL Hamilton syringe and a NanoFil (35 GA) bevelled needle. The mouse was first anesthetized with 1–3% isoflurane and put into the stereotaxic frame. The headstage EIB-16 was then gently unscrewed and shifted to expose underneath electrodes and cannulas. The dummy cannula was gently removed from the left guide cannula before the NanoFil injecting needle was inserted into the guide cannula and active heparinase 1 (0.05 U/1 μL per site, Sigma-Aldrich) or, as control treatment, heat-inactivated heparinase 1 (30 min boiling in 100 °C) was injected at a rate of 3 nL/s. After waiting for another 5 min, the injection needle was replaced by the dummy and the same procedure was performed on the other hemisphere. The duration of the procedure including re-mounting of the headstage after injection was about 25–30 min.
A classical Pavlovian contextual fear conditioning (CFC) paradigm was used. In short, CFC was done at day 0 (d0) as follows: 1 day after intrahippocampal injections, mice were placed in a neutral context (CC−) for 5 min to record behavior and electroencephalography (EEG). 1–2 h later the same mice were placed into a conditioned context (CC+) for 5 min and 3x medium intensity foot shocks were applied (0.5 mA, 1 s) with an interval of 1 min. The protocol was 2 min exploration of CC+, then first shock, after 1 min second shock, and again after 1 min the third shock, mice were then given another minute in CC+ before being removed. EEG was not recorded during the training. The conditioned context (CC+) was a chamber (20×20×30 cm) with contrast black-and-white chess-like pattern on the walls and a metal grid on the floor (75% ethanol was used for cleaning). The neutral context (CC−) was the same chamber, but with gray walls and gray plastic floor, and another type of cleaning solution was used with a different smell. On d1 and d3, the mice were subjected to retrieval sessions. First, mice were placed in CC+ for 5 min and then 1–2 h later they were placed into CC− context for 5 min. All tests were recorded with video synchronized with EEG. For CFC, we used a computerized fear conditioning system (Ugo Basile). The behavior of mice during fear conditioning was recorded for offline analysis using an USB video camera and a behavioral video acquisition and analysis software (AnyMaze). All recorded movies were analyzed by a trained observer blinded to the treatment group. Total freezing time was calculated as the percentage of 5 min (in either of the 2 contexts, CC+ and CC−).
Intrahippocampal local field potentials (LFPs) were recorded by using a 64 channels recording system and the acquisition software Cheetah (both Neuralynx). LFPs were recorded at 3200 Hz sampling rate, a 0.1–300 Hz wide-band filter, ×1000 amplification, and stored for offline analysis. LFPs were analyzed using NeuroExplorer (5.0, Madison). To record from a freely moving mouse, we used a 5 m cable (Neuralynx) counterbalanced with an equal weight of the headstage and the cable itself. For further data analysis, we chose 2 electrodes targeting 2 most interesting structures (CA1 and DG) involved in contextual memory and contextual discrimination, and electrodes were selected based on following criteria: 1) length of electrodes; 2) histology depicting electrode traces (Fig. 6b); 3) lowest theta power is characteristic for CA1 stratum pyramidale, whereas highest theta power for molecular layer of DG (Fig. 6c); 4) high-frequency oscillations as ripples are characteristic for stratum pyramidale and/or radiatum.
Two-Photon Excitation (2PE) Ca2+ Imaging
CA1 pyramidal cells, their dendrites, and spines were visualized by 2-photon excitation (2PE) fluorescence microscopy. We used a Scientifica system (Scientifica) and a FV10MP imaging system (Olympus) optically linked to a femtosecond pulse laser (Vision S, Coherent, λ = 800 nm) both integrated with patch-clamp electrophysiology (Multiclamp 700B, Molecular Devices) (Kochlamazashvili et al. 2010; Henneberger and Rusakov 2012; Anders et al. 2014). Setups were equipped with 25× (NA 1.05) and 40× (NA 0.8) objectives (Olympus). The laser power was adjusted for depth in the tissue to obtain, on average, 2–3 mW in the focal plane. Fluorescent dyes were loaded via the whole-cell patch-clamp pipette. Once in whole-cell mode, Alexa Fluor 594 typically equilibrated across the neuron within 10–15 min and was used to visualize dendrites and spines. For fluorescence recordings from spines and dendrites, a region of interest was pseudo-randomly chosen. Rapid scans across a line covering spines and dendrites were performed (line scans, 0.4–0.8 kHz) synchronized to electrophysiology. After background correction, fluorescence intensity (F) or its changes (∆F) were normalized to the baseline fluorescence (F0) before any stimulus to obtain F/F0 or ∆F/F0. In analyses presented in Figures 2d and 3c peak F/F0 or ∆F/F0 values were corrected for decaying fluorescence from the previous stimulus (Kochlamazashvili et al. 2010). The decay time constant of Ca2+-dependent fluorescence did not differ between sham and heparinase-treated slices (monoexponential fit, sham 161.9 ± 13.0 ms, heparinase 137.1 ± 15.8 ms, n = 29 and 17, unpaired t-test, P = 0.354). For recording of Ca2+ transients, the membrane potential was adjusted to −70 mV (either in voltage clamp or via a constant holding current). For Schaffer collateral stimulation (5× TBS; see above, Fig. 2), the stimulus intensity was set to 4× the intensity to induce a single action potential in a burst. The NMDAR inhibitor AP5 (50 µM) was added in these experiments to prevent Ca2+ indicator saturation and to reduce variability (a modulation of NMDAR-mediated synaptic transmission by heparinase treatment has been excluded, as shown in Fig. 1b and3). In a subset of experiments, saturation of the Ca2+ indicator Fluo-4 was induced using a high-frequency train of 80 bAPs at 100 Hz. Resting [Ca2+] was estimated from the ratio F0/FMAX (see Ermolyuk et al. 2012). The low-affinity Ca2+ indicator OGB6F was used in a Cs+-based intracellular solution (see above) to measure the nifedipine-sensitive component of sustained Ca2+ influx into dendrites and spines during 1 s of depolarization from −50 to +20 mV in presence of TTX, NBQX, AP5, PTX, and CGP52432 (see
2PE Fluorescence Lifetime Imaging
The fluorescence lifetime of OGB1 and its Ca2+-dependent changes (Agronskaia et al. 2004; Zheng et al. 2015) were measured to obtain more quantitative information about the effect of HS removal on pyramidal cell Ca2+ signaling. Time-correlated single photon counting (TCSPC) was performed on a FV10MP Olympus microscope upgraded with a TCSPC module (Picoquant). OGB1 lifetimes were quantified using a photon count ratio (see3d). Resting [Ca2+] presented in
Numerical data are reported as mean ± SEM with n being the number of samples. In figures, asterisks indicate statistical significance (details in figure or figure legend). Student's t-tests and multiway ANOVA with suitable post-hoc tests were used as indicated and performed in Matlab (Mathworks), SPSS (IBM), and Origin (OriginLab) and P represents the level of significance. Significance levels are indicated in figures by asterisks (*P < 0.05, **P < 0.01, ***P < 0.001) unless stated otherwise.
Removal of HSs Impairs LTP
HSPGs can interact with a plethora of cellular components and thereby modify synaptic transmission and plasticity. We first tested if enzymatic digestion of HSs affects synaptic transmission and its LTP. HSs were removed from acute slices by incubating slices in extracellular solution containing either heparinase 1 (1 U/mL) or heat-inactivated heparinase 1 (control treatment). This protocol was sufficient to cleave HSs in dissociated cultures and acute hippocampal slices (see1a–c, 1b, 1b and 1c, Kochlamazashvili et al. 2010). LTP was induced by a single cycle or repeated TBS (Fig. 1d,e). We found no significant effect of heparinase treatment on LTP induced by a single cycle of TBS. However, when 5 cycles of TBS were delivered, an impairment of LTP in heparinase-treated slices was uncovered (Fig. 1d,e). LTP in slices receiving the control treatment (heat-inactivated heparinase) was not statistically different from native slices, which were not incubated (native 156.1 ± 5.9% vs. heat-inactivated heparinase 144.5 ± 4.8%, n = 8 both groups, P = 0.15, unpaired t-test). The reduction of LTP in heparinase-treated slices was not due to an effect on GABAergic synaptic inhibition, because the reduction persisted in the presence of the ionotropic GABA receptor inhibitor PTX (Fig. 1e). Together these results reveal a significant reduction of LTP induced by repeated TBS after removal of HSs without an effect on baseline excitatory transmission.
Impairment of Ca2+ Influx During LTP Induction
The dependence of LTP on postsynaptic intracellular Ca2+ signaling is well-established (Lynch et al. 1983). Therefore, we tested if a reduced Ca2+ entry during LTP induction underlies the LTP impairment observed after heparinase treatment. To this end, 2PE Ca2+ imaging was performed in spines and dendrites of CA1 pyramidal cells held in the whole-cell patch-clamp configuration (Ca2+ indicator Fluo-4, Materials and Methods) during TBS of CA3–CA1 Schaffer collateral synapses (Fig. 2). In order to decrease interexperiment variability, the intensity of axonal stimulation was set to 4× the current required for a single action potential to occur during a single test burst of 5 stimuli (control 71.6 ± 5.9 µA, heparinase 55.2 ± 6.2 µA, n = 11 both groups, P = 0.19, unpaired t-test, see also Kochlamazashvili et al. 2010) and NMDARs were inhibited by AP5 (NMDAR-dependent synaptic transmission was not affected by heparinase, see above and Materials and Methods). Indeed, the Ca2+-dependent fluorescence transients evoked by TBS were reduced in heparinase-treated slices (Fig. 2d). This was not a consequence of indicator saturation, because a 2.5-fold higher stimulus intensity induced Fluo-4 fluorescence transients with 3–4-fold higher amplitudes (P = 0.019 and <0.001 for control and heparinase-treated slices, n = 4 and 6, unpaired t-tests). Thus, postsynaptic Ca2+ signals during TBS are reduced in heparinase-treated slices and this reduction is likely to underlie the observed reduction of LTP. Importantly, boosting Ca2+ entry during induction of LTP via L-type voltage-gated Ca2+ channels using BayK 8644 (15 µM) abolishes the effect of heparinase treatment on LTP (control + BayK 154.2 ± 6.6%, heparinase + BayK 151.6 ± 5.9, n = 8 both groups, unpaired t-test, P = 0.77). While this observation does not implicate L-type voltage-dependent Ca2+ channels (VGCCs) in the heparinase-induced reduction of Ca2+ entry during TBS, it demonstrates that the reduced Ca2+ entry is sufficient to explain the LTP deficit in heparinase-treated slices.
No Change of Depolarization-Induced Ca2+ Entry
To identify the mechanism underlying reduced Ca2+ entry during TBS, we further investigated how removal of HSs modifies Ca2+ signaling of CA1 pyramidal cells. VGCCs represent a main route of Ca2+ entry at spines and dendrites. We previously showed that hyaluronic acid, another component of the ECM, supports Ca2+ entry via L-type VGCCs and in turn LTP (Kochlamazashvili et al. 2010). Also, HSs can directly interact with L-type VGCC and thereby delay channel inactivation in HEK cells (Garau et al. 2015). Both observations suggest that removal of HSs from hippocampal slices could impair L-type channel function and thereby Ca2+ entry during TBS. This was tested by recording Ca2+ responses to trains of back-propagating action potentials (b-APs, 5 at 20 Hz, Fig. 3). However, no effect of heparinase treatment could be detected (Fig. 3c, n = 16 and 25 for control and heparinase treatment, 3-way repeated measures ANOVA, P = 0.97) irrespective of whether Ca2+ transients were recorded from spines or dendrites (P = 0.79).
Because dendritic APs could change their amplitude or width uncontrollably when propagating toward the recorded spines and dendrites, we also tested if Ca2+ responses evoked by somatic depolarization in voltage clamp are affected by heparinase treatment in the presence of the sodium channel blocker TTX and receptors antagonists (NBQX, AP5, PTX, CGP52432). To obtain a more quantitative estimate of Ca2+ concentrations and their changes, we used the Ca2+-dependent fluorescence lifetime of OGB1 as a readout of Ca2+ concentrations (fluorescence lifetime imaging, FLIM [Agronskaia et al. 2004; Zheng et al. 2015],3d, 3d). However, a HS-dependent change of L-type VGCC inactivation may not be detectable during brief depolarizations. Therefore, a separate set of experiments using the low-affinity indicator Oregon Green 488 BAPTA 6F (OGB6F) with long depolarization was performed (see
Heparinase Treatment Reduces Pyramidal Cell Excitability
The absence of an effect of heparinase treatment on Ca2+ entry induced by somatic voltage step commands and b-APs prompted us to explore if the decreased Ca2+ influx during TBS could be a consequence of reduced action potential firing of pyramidal cells. To address this, the number of action potentials fired by pyramidal cells during TBS was analyzed for each burst during TBS (Fig. 4a,b). This analysis revealed a significant reduction of spikes generated per burst when HSs had been removed by heparinase treatment (Fig. 4b). Such a decrease of spikes fired per burst can have its origin upstream of the soma and axon, for example a dampened dendritic integration of synaptic input, or at the site of action potential generation itself, for example reduced excitability of the AIS. To differentiate between the 2 scenarios, 40 ms long current pulses mimicking the TBS pattern were injected into CA1 pyramidal cells (Fig. 4c). Similar to synaptic TBS, the current pulse amplitude was set in each individual neuron to 4× the current required for a single spike upon a single current injection (no difference of currents between treatments,4c,d). This heparinase-induced reduction of pyramidal cell excitability is sufficient to explain the lower number of spikes generated during synaptic TBS in heparinase-treated slices, and also the reduced Ca2+ entry and LTP. Having already excluded an effect of heparinase treatment on synaptic transmission, this result also strongly suggests that removal of HSs is affecting spike generation in the AIS. One indicator of axonal excitability is the spike threshold (Scott et al. 2014), which was therefore analyzed. Indeed, we found that the second but not the first spike threshold was increased by heparinase treatment (Fig. 4e). This increase of the spike threshold became prominent during later bursts, matching the progressive reduction of spike numbers per burst (Fig. 4d). An axonal site of action of heparinase is also indirectly suggested by the absence of a heparinase effect on various other cell parameters. The peak spike voltage and maximum rate of depolarization represent indicators of somatic sodium channel availability. Neither was affected by heparinase treatment (see
To gain further insight into the mechanism of heparinase action, action potential trains were evoked by longer-lasting current injections of increasing amplitude (Fig. 4f, left panel). Again, a spike threshold increase was observed in heparinase-treated slices (Fig. 4f, right panel). However, the effect on the spike threshold became prominent only when stronger stimuli were used. This is in line with 2 previous observations in our present study. First, the current threshold for triggering a single spike within 40 ms (or 300 ms) of stimulus onset is not affected by removal of HSs (see4a–d). Thus, the experimental evidence suggests a role of HSs for maintaining spiking in response to strong input at the AIS.
We, therefore, tested if heparinase treatment modifies the AIS. The protein AnkG is the master organizer of the AIS (Yoshimura and Rasband 2014). Its axonal distribution indicates the position and length of the AIS and changes of AnkG labeling accompany structural changes of the AIS, which affect properties of action potential firing such as the current threshold and repetitive firing (Grubb and Burrone 2010; Yoshimura and Rasband 2014). Immunohistochemistry using an antibody against AnkG was performed on dissociated hippocampal cultures that received either a control or heparinase treatment (see5a). The distance from the AIS start to the soma-hillock transition was analyzed first (blinded to experimental condition). The start of the AIS was visually identified as the region where AnkG labeling displayed a sharp rise. Its distance to the soma-hillock transition was not affected by heparinase treatment (control 4.6 ± 0.26 µm, n = 135; heparinase 4.7 ± 0.23 µm, n = 154; P = 0.74, unpaired t-test), which is in agreement with our observations in dissociated cultures. However, we noted a considerable variability of the AIS start (see 5b,c). As predicted by experiments in dissociated cultures, the distal but not proximal axonal AnkG label was significantly reduced by heparinase treatment (Fig. 5d,e). This indicates that, indeed, HSs have a role in maintaining the AIS.
In previous work (Scott et al. 2014), we have demonstrated that repetitive firing expands the initially proximal and narrow action potential initiation site further into the axon of DG granular cells. A reduced integrity of the distal AIS after heparinase treatment is, therefore, expected to selectively affect later spikes during repetitive firing, which is in good agreement with the experimentally observed spiking behavior after heparinase treatment (Fig. 4).
Impairment of CFC
We next asked what the behavioral relevance of the LTP deficit after heparinase treatment might be. Selective impairment of LTP induced by strong stimuli at CA3–CA1 synapses has previously been associated with a deficit in CFC (Kochlamazashvili et al. 2010) or context discrimination (Jin et al. 2013). Therefore, we tested if heparinase treatment impairs fear conditioning or context discrimination in a CFC paradigm (Fig. 6a, see Materials and Methods). Injection of heparinase did not affect the level of spontaneous freezing/immobility before CFC at d0 in the neutral context (CC−) compared with control treatment (both conditions ∼20% freezing, Fig. 6b). This level of freezing before CFC is expected because mouse movement could be slightly constrained by the cable connection to the recording system (typical freezing level during fully unconstrained exploration of novel environment is 5–10% in mice).
Of note, 48 and 96 h after intrahippocampal heparinase injection, that is at d1 and d3, respectively, control- and heparinase-injected mice showed similar freezing in the conditioned context (CC+, Fig. 6b). Freezing times were 62.4 ± 5.7% (d1) and 57.6 ± 5.0% (d3) in heparinase-injected mice (n = 9) and 67.1 ± 3.8% (d1) and 60.5 ± 3.9% (d3) after control treatment (n = 10, repeated measures 2-way ANOVA: treatments P = 0.507, days P = 0.103). However, the freezing time in neutral contexts (CC−) was significantly different (P = 0.0093) between treatment groups and the test day was a significant factor (P = 0.031). Heparinase-injected mice froze in CC− as long as in CC+ (61.4 ± 4.7% and 53.3 ± 4.6% at d1 and d3, respectively) whereas the control group froze in CC− less than in CC+ (43.9 ± 4.5% and 39.1 ± 3.3% at d1 and d3, respectively; CC+ vs. CC− in the control group P < 0.001 at both days; CC+ vs. CC− in heparinase-injected animals P > 0.3 at both days; paired t-tests). We, therefore, calculated the context discrimination ratio ([Freezing% in CC+] − [Freezing% in CC−])/ ([Freezing% in CC+] + [Freezing% in CC−]), which revealed the inability of heparinase-injected mice to discriminate between the conditioned context CC+ and the unconditioned neutral CC− at any testing days (Fig. 6c), unlike control mice (P < 0.005 at d1 and P < 0.01 at d3).
Reduction of Oscillatory Network Activity In Vivo by Heparinase Treatment
In vivo, hippocampal neuronal activity can be synchronized in different frequency bands depending on the behavioral state. These hippocampal oscillations, in particular those at the theta frequency range, are implicated in many learning and memory processes in humans and animals (Colgin 2016). Therefore, we explored how low theta (Lθ, 4–7 Hz) and high theta (Hθ, 7–10 Hz) oscillations during CFC are affected by context, behavior and heparinase treatment (Fig. 7). Frequency analysis in the θ range revealed that when mice were actively exploring their environment, theta activity peaked in Hθ. However, when mice were freezing, independent of context, theta peak activity was shifted to Lθ (Fig. 7a–d). This shift was highly significant in both treatment groups (P < 0.0005) and was stable over time. The θ peak frequency during freezing (fr) events stayed within Lθ range around 6 Hz (5.8 ± 0.4 − 7 ± 0.4 Hz) for both treatment groups and recording areas (CA1 stratum pyramidale and DG stratum moleculare, Fig. 7e). Similarly, the θ peak frequency during exploration (exp) was mostly in the Hθ band ranging from 7.9 ± 0.2 to 9.1 ± 0.2 Hz (Fig. 7e).
Statistical analysis (general linear model with repeated measures; 4 within-subject factors: context, behavior, day after CFC, brain region; one between-subject factor: heparinase/control treatment) showed that only one factor, behavior (freezing versus exploration), highly significantly (P < 0.0005) modulated the θ frequency at d1 and d3. Furthermore, there were significant interactions between treatment and context (P = 0.003), as well as between treatment and day (P = 0.009). At d0, only data recorded in the CC− context were available for analysis of the θ frequency shift by design (Fig. 7e, left panel). To compare θ frequency shifts over the entire duration of the CFC experiments, another analysis was performed using only data obtained in CC− but from d0, d1, and d3. Again, analysis showed that only behavior (P < 0.0005) modulated the θ frequency. In addition, we found interactions between treatment and day (P = 0.023) as well as between day and behavior (P = 0.006) probably reflecting the fact that at d0 the peak Hθ (~9 Hz) was slightly higher compared with d1 and d3 (~8 Hz). To conclude, neither heparinase treatment nor context alone influenced the main θ frequency during exploratory and freezing behavior in mice.
We next tested if the degree of network synchronization is affected by heparinase treatment. Spectral power is a useful measure of synchronization of neuronal activity near the recording electrode. To compare the power of Lθ oscillations (4–7 Hz) during freezing versus exploration events, we normalized Lθ power during freezing over Lθ power during exploration in both contexts (CC+/CC−). We found that Lθ power doubled during freezing in CC+ at d1/d3, without an effect of heparinase treatment (Fig. 7f). However, statistical analysis based on the general linear model with repeated measures indicated a significant interaction between context and treatment at d1 and d3 (P = 0.003). This reflects that in control mice the relative Lθ power in CC+ during freezing tended to be higher than in CC−, whereas in the heparinase-treated group the relative Lθ power in CC− during freezing tended to be higher compared with CC+. This correlates well with a higher freezing of heparinase-injected mice in CC− and the loss of contextual discrimination after heparinase injection. Also, the relative Lθ power was different between treatment groups at d1 in CC− (DG only, repeated measures 2-way ANOVA, P < 0.05).
Finally, we investigated Lθ power changes during CFC in heparinase-injected animals and control animals (Fig. 7g,h). There was no difference between treatment groups in Lθ during baseline recordings. However, the control group showed a 2-fold increase of Lθ power at post-training days d1 and d3 in both CC+ and CC− contexts (Fig. 7g) compared with the baseline preconditioning level at d0. In contrast, the network activity in the Lθ band of heparinase-injected animals was not elevated after CFC. Instead, the power of Lθ oscillations after CFC was almost unaffected at d1 and slightly elevated at d3 compared with the baseline period before CFC (Fig. 7g,h). Analysis revealed no strongly significant effect of context (P = 0.07) or brain region (P = 0.071). However, several other parameters significantly contributed to the Lθ changes: day (P = 0.01), brain region and day (P = 0.023) as well as heparinase treatment (P = 0.016).
Thus, the heparinase-induced impairment of context discrimination in CFC is associated with 2 distinct effects on theta oscillations. Heparinase injection affected the Lθ power increase during freezing after CFC in a context-dependent manner and prevented the post-CFC increase of Lθ power.
The acute contribution of HSs to hippocampal function was identified in vitro and in vivo by enzymatic digestion of HSs. We found a distinct deficit of LTP of CA3–CA1 Schaffer collateral synapses after removal of HSs when a strong LTP-inducing paradigm was used in vitro. A reduced Ca2+ entry during LTP induction was identified as the underlying mechanism. This was not caused by alterations of excitatory or inhibitory synaptic transmission, but by a reduced axonal excitability of CA1 pyramidal cells. In vivo, removal of HSs was accompanied by a loss of context discrimination and abnormalities in Lθ oscillations.
HSs Regulate LTP and CA1 Pyramidal Cell Excitability
Previous studies have implicated HSPGs as contributors to synaptic plasticity. For instance, genetic deletion of the HSPG syndecan-3 improved LTP of CA3–CA1 synapses (Kaksonen et al. 2002). In contrast, enzymatic removal of HSs using heparinase 3 decreased LTP at the same pathway (Lauri et al. 1999), but the cellular mechanism has remained unclear.
We now demonstrate that specific removal of highly sulfated HSs with heparinase 1 results in impaired LTP that is only detected when strong LTP-inducing stimuli are used. This was not the consequence of altered baseline excitatory synaptic transmission at CA3–CA1 synapses, in line with a previous study (Lauri et al. 1999). Instead, we observed a reduction of Ca2+ influx during LTP induction in HSs deficient hippocampal slices. This reduced Ca2+ influx is sufficient to explain the LTP deficit, because pharmacological boosting of Ca2+-entry using the L-type VGCC modulator BayK 8644 was sufficient to restore LTP. At the surface, these functional consequences of HSs removal are similar to observations after digestion of hyaluronic acid (Kochlamazashvili et al. 2010). In both cases, a reduced Ca2+ influx during LTP induction and decreased LTP was observed. However, mechanistically both ECM components control LTP differently. Hyaluronic acid is required for maintenance of somatodendritic L-type VGCC function in hippocampal neurons. In contrast, HSs do not appear to affect L-type VGCCs at these cells but rather support axonal excitability. Several experimental findings are in line with this conclusion.
First, no modulation of Ca2+ entry in response to b-APs or direct depolarization was detected using electrophysiology in combination with intensity and lifetime-based Ca2+ imaging. Therefore, an altered function of somatodendritic VGCCs at CA1 pyramidal cells cannot explain the decreased Ca2+ entry after heparinase treatment. This appears to be at odds with experiments in HEK293 cell transfected with L-type VGCC, in which HSs modulated voltage-dependent inactivation of Ba2+ currents (Garau et al. 2015). However, these culture experiments were done in the absence of Ca2+, which was not the case in present experiments even though they were performed in the presence of high-affinity Ca2+ indicators. Therefore, Ca2+-dependent inactivation in the present experiments may mask the voltage-dependent effects of heparinase treatment on VGCC function found in cultures. Thus, in physiological settings, that is with fully intact intracellular Ca2+ signaling, a direct modulation of VGCCs by HSs is likely to play a minor role. Second, the number of action potentials fired by CA1 pyramidal cells during LTP induction was significantly reduced in heparinase-treated slices and this effect could be reproduced by strong somatic current injections. Because the action potential-induced Ca2+ entry is not different between control and heparinase-treated slices, the reduced number of action potentials during LTP induction in heparinase-treated slices explains the decrease of Ca2+ entry during LTP induction. A plausible cause of reduced spiking is the increased threshold of action potentials in heparinase-treated slices. This increase of the spike threshold did only become apparent during TBS-like and strong current injections (Fig. 4e,f, see also
The AIS is the site of action potential generation. Its geometry, location, and the distribution and availability of its ion channels dynamically determine the site and threshold of spike initiation (Kole and Stuart 2012; Yoshimura and Rasband 2014). AnkG is a key organizing molecule of the AIS and important for ion channel clustering at the AIS and thus sets key parameters of neuronal spiking (Grubb and Burrone 2010; Kole and Stuart 2012; Yoshimura and Rasband 2014). In our experiments, heparinase treatment selectively reduced the distal but not proximal axonal density of AnkG. Given the importance of AnkG in orchestrating the clustering of ion channels in the AIS, the reduced presence of AnkG is expected to profoundly impair ion channel clustering in the distal AIS after heparinase treatment. Interestingly, we have previously observed in DG granular cell axons that the action potential initiation site is initially proximal and narrow but widens and extends further into the axon during repetitive spiking (Scott et al. 2014). This indicates that the impairment of the distal AIS may only manifest during repetitive spiking, which is in good agreement with our experimental observations. Thus, HSs help maintain distal AIS integrity and as a consequence repetitive spiking. This could involve changes of, for instance, sodium or potassium channel density in the distal AIS.
Interestingly, the HSPG perlecan is recruited by dystroglycan to nodes of Ranvier, where it binds gliomedin and enhances clustering of node of Ranvier components (Colombelli et al. 2015). Although perlecan may play a more prominent role in organizing the AIS in the peripheral nervous system (Colombelli et al. 2015), all major HSPGs including perlecan are expressed in the hippocampus (Karthikeyan et al. 1994; Shee et al. 1998; Lauri et al. 1999). Moreover, NrCam and Neurofascin-186 are 2 molecules that establish a link between the ECM and AnkG and are thus thought to contribute to sodium and potassium channels clustering (Nav1.x, Kv1, and Kv7 families) (Dityatev et al. 2010; Rasband 2010). In addition, other components of the ECM like tenascins may also affect sodium channel clustering/activity in the AIS (Srinivasan et al. 1998). To summarize, the most likely scenario is that heparinase treatment disrupts the AIS ECM and modifies distal AnkG levels and ion channel clustering. A number of interesting questions for future work emerge from our findings. For instance, it remains to be shown that removal of HSs specifically at the AIS is responsible for the observed effects. It will also be interesting to dissect if the observed reduction of AIS excitability is primarily caused by a reduction of distal AIS potassium or sodium channels clustering or both. In addition, it remains to be established, if the impairment of LTP and neuronal excitability after HS removal are specific to the CA1 region or a ubiquitous effect.
Previous studies revealed the importance of interaction between HSPG syndecan-3—heparin-binding growth-associated protein (HB-GAM or pleiotrophin) in synaptic plasticity (Lauri et al. 1999; Kaksonen et al. 2002; Pavlov et al. 2002). However, mice deficient in syndecan-3 exhibited enhanced rather than reduced LTP (Kaksonen et al. 2002), and hence do not mimic the effect of heparinase 1. Moreover, syndecan-3 seems to act via postsynaptic mechanisms rather than by influencing axonal excitability (Lauri et al. 1999; Melendez-Vasquez et al. 2005). Thus, our experiments revealed a novel function of HSs in neuronal excitability and LTP that is mediated by proteoglycans carrying highly sulfated HSs. The specific effects observed after heparinase 1 treatment may be related to a loss of specific glycoepitopes or to the generation of novel molecular recognitions sites with a terminal glucosamine, which are distinct from N-acetyl-glucosamine-containing sites generated by heparinase 3.
HSs Support Learning and Learning-Related Network Oscillations
Our data provide the first evidence that a specific and acute modification of HSs in the adult brain, that is cleavage of highly sulfated HSs in the hippocampus, can lead to impairment of context discrimination, but not learning per se. A similar combination of an impairment of context discrimination and modified LTP in the CA1 region in vitro has also been observed in other studies. For instance, inhibition of the Ras-GRF1 pathway in the CA1 region modulated LTP and context discrimination (Jin et al. 2013; Darcy et al. 2014). Thus, the present study uncovers a contribution of HSs for hippocampal LTP and context discrimination and also supports the notion that the CA1 region is involved in context discrimination (Place et al. 2012; Czerniawski and Guzowski 2014).
We also uncovered several interesting relationships between hippocampal θ oscillations, behavior, and removal of highly sulfated HSs by heparinase. First, the transition from exploration to freezing was accompanied by a shift from high to low θ oscillations independently of treatment. A similar Hθ to Lθ shift was first described by Seidenbecher et al. (2003), who provided examples illustrating that theta activity between 7 and 10 Hz prevails in the CA1 during risk-assessment behavior, and θ activity between 4 and 5 Hz appears in the lateral amygdala and CA1 during freezing. Qualitatively similar observations have been documented in various other studies (Machado et al. 2012; Ognjanovski et al. 2014; Tendler and Wagner 2015). Thus, a high to low theta frequency shift is strongly associated with freezing behavior. However, this relationship was not profoundly affected by removal of highly sulfated HSs and does, therefore, not explain the impaired context discrimination after heparinase injection.
In contrast, we found that fear conditioning increased the power of Lθ oscillations in control but much less in heparinase-injected animals. This effect was particularly prominent in the CA1 region and correlated well with the impaired excitability and LTP in the same region in vitro. It is intuitive that a reduced excitability of pyramidal cells reduces the power of network oscillations. For instance, recordings from mice with reduced activity of Nav1.6, a sodium channel enriched at the AIS, and nodes of Ranvier display a reduction of the θ power in awake animals (Pal et al. 2015).
This absence of a Lθ oscillations power increase after CFC in heparinase-injected animals was associated with impaired context discrimination. The most parsimonious explanation for these findings is that an Lθ oscillations power increase is required for successful context discrimination in CFC. Indeed, available data suggest that CA1 Lθ power is a predictor for successful fear-driven memory retrieval in rodents. For instance, the muscarinic cholinergic antagonist scopolamine disrupts CFC (Anagnostaras et al. 1999) whereas M1/M4 agonists improve CFC in aged mice (Si et al. 2010). At the same time, cholinergic innervation powerfully modulates hippocampal theta oscillation power and coherence (Hasselmo 2006; Vandecasteele et al. 2014). Also, inactivation of the ventral tegmental area with procaine in rats leads to impaired fear conditioning and decreased θ power of CA1 specifically at the Lθ frequency range (3–6 Hz) during immobility states, whereas control rats exhibit an increases in θ power after fear conditioning (Matulewicz et al. 2015). Therefore, we speculate that interfering with increase of Lθ may impair learning and/or retrieval of context memories or, as our data suggest, introduce an error in discrimination of 2 similar but yet different contexts.
In this study, we have explored the contribution of highly sulfated HSs to hippocampal function and behavior. Their acute removal impairs synaptic LTP by lowering postsynaptic axonal excitability in vitro. In vivo, we demonstrate a role of HSs in modulating theta oscillation and context discrimination in a fear conditioning paradigm. It remains to be established to what degree pyramidal cell excitability alone explains our in vivo observation and if changes of, for instance, interneuron function are involved too. Nonetheless, our results demonstrate that HSs are continuously required in the adult brain to maintain hippocampal plasticity and hippocampus-dependent memory.
The NRW-Rückkehrerprogramm (C.H.), Human Frontiers Science Program (C.H., HFSP RGY-0084/2012), UCL Excellence Fellowship (C.H.), BMBF (A.D., 01EW1308A) and EraNet TargetECM (A.D.), and German Research Foundation (DFG, C.H., SFB1089 B03, SPP1757 HE6949/1 and HE6949/3).
We thank Dr Zheng and Dr Rusakov (University College London, UK) for help with setting up fluorescence lifetime imaging. Conflict of Interest: None declared.