Abstract
For many decades, synaptic plasticity was believed to be restricted to excitatory transmission. However, in recent years, this view started to change, and now it is recognized that GABAergic synapses show distinct forms of activity-dependent long-term plasticity, but the underlying mechanisms remain obscure. Herein, we asked whether signaling mediated by β1 or β3 subunit-containing integrins might be involved in regulating the efficacy of GABAergic synapses, including the NMDA receptor-dependent inhibitory long-term potentiation (iLTP) in the hippocampus. We found that activation of β3 integrin with fibrinogen induced a stable depression, whereas inhibition of β1 integrin potentiated GABAergic synapses at CA1 pyramidal neurons in male mice. Additionally, compounds that interfere with the interaction of β1 or β3 integrins with extracellular matrix blocked the induction of NMDA-iLTP. In conclusion, we provide the first evidence that integrins are key players in regulating the endogenous modulatory mechanisms of GABAergic inhibition and plasticity in the hippocampus.
SIGNIFICANCE STATEMENT Epilepsy, schizophrenia, and anxiety are just a few medical conditions associated with dysfunctional inhibitory synaptic transmission. GABAergic synapses are known for their extraordinary susceptibility to modulation by endogenous factors and exogenous pharmacological agents. We describe here that integrins, adhesion proteins, play a key role in the modulation of inhibitory synaptic transmission. Specifically, we show that interference with integrin-dependent adhesion results in a variety of effects on the amplitude and frequency of GABAergic mIPSCs. Activation of β3 subunit-containing integrins induces inhibitory long-term depression, whereas the inhibition of β1 subunit-containing integrins induces iLTP. Our results unveil an important mechanism controlling synaptic inhibition, which opens new avenues into the usage of integrin-aimed pharmaceuticals as modulators of GABAergic synapses.
Introduction
The modulation of synaptic efficacy is a key mechanism in brain development as well as learning and memory, and its dysfunction is at the root of many neurologic and psychiatric disorders. For several decades, studies into the phenomena of synaptic plasticity were primarily focused on the excitatory transmission, whereas GABAergic drive was believed to show limited if any capacity to undergo plastic changes. However, over the past few years, this view started to change, and now it is recognized that also GABAergic synapses undergo activity-dependent long-term plasticity that relies on presynaptic or postsynaptic mechanisms and are regulated by cytoplasmic (Yap et al., 2021) and extracellular factors (Wiera et al., 2021). Among the multifarious types of GABAergic plasticity, a form of postsynaptically expressed NMDAR-dependent inhibitory long-term potentiation (iLTP) has been described in the hippocampus (Marsden et al., 2007) and cerebral cortex (Chiu et al., 2018). NMDA-iLTP is induced heterosynaptically through moderate NMDAR activation (Marsden et al., 2010) and involves an increase in synaptic GABA A receptor (GABAAR) abundance, which requires activation of CaMKII, phosphorylation of gephyrin, and GABAAR subunits (Petrini et al., 2014). However, the precise molecular mechanisms of NMDA-iLTP remain obscure. In excitatory transmission, the extracellular matrix (ECM) is recognized as a major player in regulating various aspects of synaptic plasticity and thereby in shaping learning and memory (Fawcett et al., 2019). Nevertheless, very little is known about the involvement of ECM in the plasticity of GABAergic synaptic transmission.
Integrins function as adhesion receptors bidirectionally linking changes in the extracellular matrix to intracellular events. There are 18 alpha and 8 beta integrin subunits in mammals, which assemble into 24 heterodimers with different affinity for extracellular proteins and peptides like collagens, laminin, and fibronectin (Park and Goda, 2016). Based on pharmacological and functional properties, brain integrins may be broadly divided into two classes containing β1 or β3 subunit, respectively. In hippocampal pyramidal neurons, activation of β1 subunit-containing integrins is a prerequisite for the consolidation of LTP (Kramár et al., 2006) and structural plasticity of dendritic spines (Michaluk et al., 2011), whereas β3 integrin is involved in stabilizing AMPARs in postsynaptic density and is required for homeostatic upscaling of excitatory synapses (Pozo et al., 2012). Additionally, the deficiency of β1 or β3 subunit has a different impact on behavior. The infusion of functional-blocking antibodies against β1-containing integrins impairs spatial memory (Babayan et al., 2012). In contrast, subunit β3-null mice displayed altered anxiety-related behavior (McGeachie et al., 2012) and dysfunctional social interactions similar to those observed in autism spectrum disorder mouse models (Carter et al., 2011). Interestingly, a comparable behavioral phenotype is routinely reported in mice with compromised GABAergic inhibitory transmission (Cellot and Cherubini, 2014; Rudolph and Möhler, 2014). This intriguing observation raises the question of whether integrins are involved in the regulation of GABAergic drive, particularly its plasticity. Demonstration that integrins, ubiquitous signaling proteins in the brain, play an important role in regulating modulatory mechanisms of GABAergic inhibition could have a fundamental importance for understanding brain physiology. Indeed, GABAergic synapses are known for their extraordinary susceptibility to modulation; and, for instance, discoveries of endogenous modulatory processes depending on endocannabinoids (Castillo et al., 2012), neurosteroids (Mukherjee et al., 2017), or endozepines (Christian and Huguenard, 2013) opened new avenues in studying GABAergic inhibition in the brain and developing new pharmaceuticals.
To address the involvement of integrins in GABAergic synaptic signaling and plasticity, we used different proteins and peptides that selectively activate or inhibit the β1 and β3 types of these signaling proteins. We found that activation of β3 integrin with fibrinogen induces a stable depression of the inhibitory synaptic transmission, whereas inhibition of the interaction between integrin β1 and ECM potentiates GABAergic synapses at hippocampal pyramidal neurons. Additionally, compounds that interfere with the interaction of β1 or β3 integrins with ECM block the induction of NMDA-iLTP. In conclusion, we provide evidence that integrins are key players in regulating the endogenous modulatory mechanisms of GABAergic inhibition and plasticity.
Materials and Methods
Animals and reagents
We performed all experiments on C57BL/6 male mice. Females were not employed to reduce the data variability resulting from the effects of estrous cycle on GABAergic transmission (Huang and Woolley, 2012; Tabatadze et al., 2015). All procedures were performed following the guidelines of the European Communities Council and approved by the Local Bioethics Committee for Experiments on Laboratory Animals. All crucial reagents are listed in Table 1; reagents not listed in Table 1 were purchased from Sigma-Aldrich.
Sources and catalog numbers of antibodies and crucial reagents used
Slice preparation
Acute hippocampal slices were prepared from C57BL/6 male mice 18–23 d after birth (Marsden et al., 2007; Wiera et al., 2021). The brain was quickly removed after decapitation and submerged in aCSF containing the following (in mm): 119 NaCl, 26.3 NaHCO3, 11 glucose, 2.5 KCl, 1 NaH2PO4, 1.3 MgSO4, 2.5 CaCl2, pH 7.4, bubbled with 95% O2/5% CO2. Transverse hippocampal slices were cut (350 µm) using a vibratome (VT1200S, Leica) and then incubated in the recovery chamber filled with the same aCSF at room temperature for at least 1.5 h before electrophysiological recordings.
Electrophysiology
The hippocampal slice was transferred to a submerged recording chamber and perfused at 2.5–3 ml/min with oxygenated aCSF (same composition as described before). Recording electrodes with a resistance of 3–4 MΩ were pulled from borosilicate glass capillaries and filled with intracellular solution containing the following (in mm): 10 potassium gluconate, 125 KCl, 1 EGTA, 10 HEPES, 4 MgATP, and 5 sucrose, pH 7.25, 295 mOsm. Whole-cell patch recordings were performed at room temperature using the MultiClamp 700B amplifier and Digidata 1550B digitizer (Molecular Devices). Pyramidal cells were identified electrophysiologically based on their firing pattern. Miniature IPSCs (mIPSCs) were recorded in the presence of 20 µm DNQX and 1 µm TTX at a holding voltage of −70 mV. During all recordings, input resistance was monitored, and the cells exhibiting >20% variation in this parameter were discarded from the analysis. Considering relatively low values of mIPSC amplitudes (<100 pA), series resistance was not compensated. iLTP was induced by transient exposure to NMDA (3 min, 20 µm) as described previously (Marsden et al., 2007). Miniature IPSCs were recorded for at least 10 min before and for 30 min after NMDA application. In the statistics we included only the cells for which we conducted stable whole-cell recordings for at least 40 min. In some cells stable recordings could be conducted for >1 h, and the observed plasticity/peptide effects were unchanged.
Miniature IPSCs were detected manually using the template procedure, and the amplitude and frequency were analyzed over 2 min periods using pClamp 10.7 software (Molecular Devices). The extent of mIPSC plasticity was assessed as the mean value of mIPSC amplitude at the time point 20–22 min after induction of NMDA-iLTP normalized to the mean baseline amplitude. The impact of integrins peptides and inhibitors on basal inhibitory transmission was estimated at the time point 16–18 min after compound application. Additionally, we analyzed the coefficient of variation (CV) for mIPSC amplitude calculated as CV−2 = (σ/µ)−2, where σ is the SD of mIPSC amplitude, and µ is the mean of mIPSC amplitude (calculated in a 2 min time window). The CV parameter is proportional to the presynaptic probability of neurotransmitter release (Brock et al., 2020).
Immunohistochemistry
Acute brain slices from electrophysiological recordings were fixed 25 min after the start of NMDA application during the stable phase of NMDA-iLTP. During the fixation procedure, hippocampal slices were kept for 24 h (4°C) in 2% paraformaldehyde and afterward rinsed three times with PBS, pH 7.4. Next, slices were embedded in gelatin (15%) and cut in cold PBS (4°C) into 40-µm-thick slices using a vibratome (VT1000S, Leica). Immunolabeling with primary antibodies was preceded by blocking nonspecific antibody binding using 10% normal horse serum (NHS) for 1 h at room temperature. Next, incubations with primary antibodies supplemented with 4% NHS in PBS were carried for 48 h at 4°C using constant stirring. Afterward, secondary antibodies (with 4% NHS in PBS) were applied for 24 h at 4°C. We used antibodies against total integrin β1, β3, and active form of β3 (1:100 concentration), anti-active integrin β1 (1:250), anti-vGAT (1:500), and secondary antibodies (1:1000). More detailed information about the antibodies used can be found in Table 1. Finally, sections were mounted on glass slides using a Fluoroshield with DAPI.
Image acquisition and analysis
Hippocampal slices were visualized using an Olympus Fluoview 1000S laser scanning confocal microscope at the University of Wroclaw in Poland. Images were acquired in the sequential mode as a z-stack with six 0.95-µm-thick sections, using a 60× oil immersion objective (PlanApo 1.35) and 473 and 635 nm excitation lasers. The analysis of confocal images was performed using ImageJ software (National Institutes of Health and Laboratory for Optical and Computational Instrumentation, University of Wisconsin) on maximum-intensity projection of z-stack images. We focused our analysis on the CA1 stratum radiatum. In our analysis, we omitted the immunofluorescence signal in blood vessels or glial cells.
We analyzed the total fluorescence signal from integrin puncta, defined as the integrated fluorescence density. Additionally, we exploited vesicular GABA transporter (vGAT) as a presynaptic marker of inhibitory synapses to determine the synaptic localization of integrins. Therefore, integrins were characterized as synaptic if they occupied a vGAT-positive area enlarged by two pixels in every direction. The density of analyzed signals corresponds to the number of integrin puncta divided by the analyzed area of stratum radiatum. The fraction of inhibitory synapses containing integrin was expressed as a total area of vGAT-positive integrin puncta divided by the area of vGAT staining and expressed as a percentage. The Manders' coefficient (M2) was analyzed in ImageJ using the JaCoP plug-in by simultaneous comparison of the individual slices of the confocal image (not flattened).
Experimental design and statistical analysis
For all tested hypotheses, at least three independent experiments were performed using slices prepared from different mice. Statistical analysis was performed with SigmaPlot 11 software, and all statistical tests used are described in the figure legends. The following a priori criteria were adopted during the data analysis: (1) No outliers were rejected, (2) slices were randomly assigned to treatments, (3) investigators were not blinded in relation to treatment used during recording/analysis, and (4) minimal sample numbers were established using power analysis based on previous experiments (Wiera et al., 2021) and assuming power level 0.80 as sufficient. The distribution normality of datasets was tested by the Shapiro–Wilk test followed by direct comparisons between two groups using two-tailed paired or unpaired Student's t test or Mann–Whitney U test (see text and figure legends for an indication of which test was used in a specific case). For multiple comparisons, one-way ANOVA with Tukey's post hoc test was used for data with normal distribution or Kruskal–Wallis one-way ANOVA on ranks with Dunn's post hoc test for data with non-normal distribution. The statistical significance was defined as *p < 0.05, **p < 0.01, ***p < 0.001. All data are presented as mean ± SEM.
Results
Arg-Gly-Asp-biding integrins are involved in the modulation of hippocampal GABAergic synaptic transmission and iLTP
The Arg-Gly-Asp (RGD) sequence has been identified in extracellular matrix proteins such as fibronectin or laminin. It has also been implicated in mechanisms underlying cell–cell and cell–ECM interaction. Most importantly, integrins, including those containing β1 or β3 subunits (in particular all αV containing, α5/β1 and α8/β1), act as receptors for these adhesion molecules by interacting with the RGD motif (Kapp et al., 2017). We therefore started our investigations on the impact of integrins on GABAergic transmission and plasticity by testing the effect of the linear synthetic RGD peptide [with a GRGDSP (Gly-Arg-Gly-Asp-Ser-Pro) sequence] on mIPSCs.
First, the baseline (control) mIPSCs were recorded from a CA1 pyramidal cell for at least 10 min. Then, while continuing to record from the same cell, the RGD peptide at a concentration of 0.5 mm was added by bath perfusion. Interestingly, RGD administration resulted in a progressive decrease in mIPSC amplitude (within ∼5 min after application), which stabilized at the relative level of 89.5 ± 2.9% (n = 12; Fig. 1A–D) reaching statistical significance (before, 56.2 ± 3.3 pA; after, 49.7 ± 2.9 pA, n = 12, p = 0.007; Fig. 1C). To test the specificity of this effect, a scrambled peptide [with a GRADSP (Gly-Arg-Ala-Asp-Ser-Pro) sequence] was used, and we did not observe any changes in mIPSC amplitude (100.2 ± 1.4%) with respect to the baseline (before, 46.2 ± 5.7 pA; after, 46.4 ± 5.7 pA, n = 7, p = 0.751; Fig. 1A–D). Neither RGD nor scrambled peptide had any significant effect on mIPSC frequency (RGD, 87.7 ± 8.0%; Scrambled, 98.4 ± 1.8%, p = 0.357; Fig. 1E,F) or the coefficient of variation of mIPSC amplitudes proportional to the probability of neurotransmitter (RGD, 1.06 ± 0.1; Scrambled, 1.07 ± 0.1, p = 0.967; Fig. 1G; Brock et al., 2020). Because previous reports implicated an important role of NMDA receptor activity in various forms of GABAergic plasticity (Muir et al., 2010; Petrini et al., 2014; Field et al., 2020), we next checked the effect of d-APV (50 µm) on RGD-induced depression of mIPSC amplitude, but no significant change was found (RGD, 90.6 ± 3.0%, n = 9; RGD plus d-APV, 83.3 ± 6.1%, n = 6, p = 0.251; Fig. 1H–J).
RGD peptide decreases the efficacy of GABAergic transmission and blocks NMDA-iLTP induction. A, Examples of raw mIPSC traces recorded before and after 16–18 min of RGD (GRGDSP, 0.5 mm) peptide treatment. Calibration, 50 pA, 1 s. B, Time course of mIPSC amplitude changes during the administration of RGD or scrambled peptide (GRADSP, 0.5 mm). C, D, Summary plots of mIPSC amplitude changes after 16–18 min of RGD administration and in control conditions (C) paired t test and (D) t test. Note a significant decrease in mIPSC amplitude after RGD application. E, Time courses of mIPSC frequency before and after application of RGD or scrambled peptide at t = 0 min. Frequency is expressed in comparison to baseline value before treatment. F, Effects of β1 integrin inhibitor on mIPSC frequency, assessed after 16–18 min of infusion; t test. G, Summary plot of changes in the coefficient of variation of mIPSC amplitudes (CV-2) after 16–18 min of RGD or scrambled peptide administration. Data are expressed as a ratio to baseline values; t test. H, mIPSC amplitude changes during treatment with RGD or RGD in the presence of NMDAR inhibitor D-APV (50 µm). I, J, Summary plots of mIPSC amplitude changes after 18–20 min of RGD administration in control conditions and in the presence of d-AP5. Paired t test (F), t test (G). K, Time course of mIPSC amplitude changes after NMDA administration in the presence of RGD or scrambled peptide. L, Averaged traces before (black) and after (red) NMDA application in the presence of RGD (top) or scrambled peptide (bottom). Calibration, 10 pA, 20 ms. M, Summary plot of mIPSC amplitude changes after 20–22 min of NMDA application in RGD-treated slices or control conditions, t test. Data points are shown as mean ± SEM. The numbers on the bars refer to the number of experiments; *p < 0.05, **p < 0.01, NS p >0.05.
A well-established form of GABAergic plasticity is heterosynaptic potentiation of mIPSC amplitude induced by activating NMDA receptors, referred to as NMDA-iLTP. This paradigm of plasticity has been extensively studied by us (Wiera et al., 2021) and previously by others (Pennacchietti et al., 2017; Chiu et al., 2018). We thus tested whether inhibition of RGD-binding integrins might interfere with NMDA-induced iLTP. To address this issue, we used a protocol that evokes stable iLTP as previously described (Marsden et al., 2007) by applying 20 µm NMDA for 3 min in the presence of a scrambled peptide and found a significant potentiation of mIPSC amplitude (119 ± 6%, n = 5, p = 0.016; Fig. 1K–M) similar to that observed on iLTP induction (Wiera et al., 2021). However, when repeating the same protocol in the presence of the RGD peptide (0.5 mm), iLTP was completely abolished (RGD, 95 ± 6%, n = 5, p = 0.473; Fig. 1K–M). Together, our experiments with RGD provide the first evidence that integrins are involved in GABAergic long-term depression, which is insensitive to NMDAR blockade, and play a key role in the induction of NMDA-iLTP.
The antagonist of β3 and β5 integrins blocks the induction of NMDA-iLTP
As already mentioned, the RGD sequence in the linear GRGDSP peptide used in our experiments interferes with various integrins (Kapp et al., 2017). Thus, we used cilengitide, which preferentially blocks αvβ3 and αvβ5 integrin subtypes, to further identify integrin types involved in iLTP (Pfaff et al., 1994; Dechantsreiter et al., 1999; Jaudon et al., 2021). We observed that bath application of cilengitide (2 µM) did not alter the baseline mIPSC amplitude (before, 36.2 ± 7.9 pA; after, 36.7 ± 8.4 pA, n = 4, p = 0.598; Fig. 2B,C), frequency (before, 2.2 ± 0.2 Hz; after, 2.3 ± 0.3 Hz, n = 4, p = 0.63), or coefficient of variation of mIPSC amplitudes (before, 5.4 ± 0.9; after, 4.9 ± 0.7, n = 6, p = 0.31). Interestingly, application of β3 and β5 inhibitor converted iLTP to iLTD (control iLTP, 117 ± 2%, n = 8; cilengitide, 87 ± 3%, n = 4, p < 0.001; Fig. 2D–G). These data indicate that αvβ3 or αvβ5 integrin activity is crucial for hippocampal NMDA-iLTP. However, the role of β1 subunit-containing integrins needs to be additionally considered.
Cilengitide (an inhibitor of β3/β5 integrins) blocks the induction of NMDA-iLTP. A, Examples of raw mIPSC traces recorded before and after 16–18 min of cilengitide treatment. Calibration, 40 pA, 1 s. B, Time course of mIPSC amplitude changes in the presence of cilengitide (2 µm). C, Summary plot of mIPSC amplitude measured from the pyramidal cell before and after cilengitide administration; paired t test. D, Averaged traces before (black) and after (red) iLTP induction in control slices (top) and cilengitide-treated group (bottom). Calibration, 10 pA, 20 ms. E, Time course of mIPSC amplitude after iLTP induction with NMDA, recorded from control slices and in the presence of cilengitide. F, G, Summary plots of mIPSC amplitude changes after 20–22 min of iLTP induction in control conditions and cilengitide-treated slices. Note the significant decrease in mIPSC amplitude after NMDA application in the presence of cilengitide; t test (F), paired t test (G). The average mIPSC amplitude determined for each pyramidal cell is shown as one circle. The numbers on the bars refer to the number of experiments *p < 0.05, ***p < 0.001, NS p >0.05.
Inhibition of β1 integrin potentiates inhibitory synaptic transmission and interferes with the long-term GABAergic plasticity
To pursue the mechanisms underlying the impact of RGD on GABAergic plasticity, we used pharmacological tools affecting integrins containing β1subunit. We first analyzed the effect of ATN-161 peptide (with an Ac-PHSCN-NH2 sequence), a blocker of α5β1 integrins (Kapp et al., 2017), on mIPSCs in CA1 pyramidal neurons (Fig. 3A). Administration of 0.5 mm ATN-161, after stable 10 min baseline recordings, caused an increase in mIPSC amplitude to 111 ± 3% (before, 47.8 ± 5 pA; after, 53.3 ± 6 pA, n = 7, p = 0.021; Fig. 3B–D). Interestingly, ATN-161 induced mIPSC potentiation, which depended on NMDA receptors, as it was blocked in the presence of d-APV (ATN-161 with d-APV, 102.1 ± 2%, n = 5, p = 0.031; Fig. 3B–D). We also used a synthetic cyclic GA(C)RRETAWA(C)GA peptide (hereafter referred to as RRETAWA, 0.15 mm), another inhibitor of α5β1integrins (Koivunen et al., 1994; Mould et al., 1998; Humphries et al., 2000), which also potentiated the mIPSC amplitude (before, 39.9 ± 4 pA; after, 47.5 ± 6 pA, n = 8, p = 0.008; Fig. 3E–H). This effect was blocked by d-APV, suggesting that NMDA receptors are involved in RRETAWA-induced potentiation (RRETAWA, 119 ± 6% of baseline, n = 8; RRETAWA with d-APV, 103 ± 2%, n = 6, p = 0.013; Fig. 3F–H). Neither ATN-161 nor RRETAWA peptide changed mIPSC frequency (sham-treated control, 103.3 ± 1.5% of baseline, n = 5; ATN-161, 109.2 ± 4.0%, n = 5; RRETAWA, 90.3 ± 9.1%, n = 9; one-way ANOVA vs control, F(2,18) = 0.291, p = 0.751; Fig. 3I,J) or the coefficient of variation of mIPSC amplitudes normalized to baseline (sham-treated control, 1.00 ± 0.05, n = 5; ATN-161, 0.99 ± 0.07, n = 7; RRETAWA, 0.9 ± 0.12, n = 9; ANOVA on ranks vs control, H(2) = 4.655, p = 0.10; Fig. 3K). To further elucidate whether RRETAWA interferes with mechanisms underlying NMDA-induced iLTP, we applied RRETAWA at least 15 min before NMDA administration (which was present throughout the experiment) and found that this pretreatment prevented iLTP induction (iLTP, 117 ± 2%, n = 8; RRETAWA, 93 ± 2%, n = 4, p < 0.001; Fig. 3L–N). On the contrary, the application of RRETAWA peptide 15 min after NMDA treatment did not alter the time course of iLTP (iLTP, 117 ± 2%, n = 8; RRETAWA 15 min, 117.4 ± 4.3%, n = 4, p = 0.830; Fig. 3L–N). Altogether, the present results indicate that β1 integrin plays a crucial role in inhibitory long-term potentiation as its blockade elicits potentiation of mIPSC amplitude that depends on NMDA receptors. Moreover, mIPSC enhancement induced by RRETAWA treatment blocked NMDA-induced iLTP, further substantiating the notion that β1integrins are essential for GABAergic long-term plasticity.
Inhibition of β1 integrin-dependent adhesion augments GABAergic transmission and blocks the induction of NMDA-iLTP. A, Raw mIPSC traces before and 20 min after ATN-161 (0.5 mm) treatment. Calibration, 40 pA, 1 s. B, Time course of mIPSC amplitude changes in the presence of ATN-161 and APV applied before ATN-161 and maintained throughout the experiment. C, D, Summary plots of mIPSC amplitude changes after 18–20 min of ATN-161 administration in control conditions and in the presence of D-APV (50 µm); t test (C), paired t test (D). E, Raw mIPSC traces recorded before and 20–22 min after RRETAWA treatment. Calibration, 40 pA, 1 s. F, Time course of mIPSC amplitude changes in the presence of RRETAWA and APV applied before RRETAWA and maintained throughout the experiment. G, H, Summary plots of amplitude changes after 20–22 min of RRETAWA administration in control conditions and in the presence of NMDAR inhibitor (APV), Mann–Whitney rank sum test (G), paired t test (H). I, Time courses of mIPSC frequency before and after applying ATN-161 or RRETAWA peptide at t = 0 min. J, Effects of integrin β1 inhibitors on mIPSC frequency, assessed after 16–20 min of infusion. Frequencies are expressed in relation to baseline value before treatment (1-way ANOVA, F(2,18) = 0.291, p = 0.751). K, Summary plot of changes in the coefficient of variation of mIPSC amplitudes (CV−2) after 16–20 min of ATN-161 or RRETAWA administration. Data are expressed as a ratio to baseline values (ANOVA on ranks, H(2) = 4.655, p = 0.10). L, Time course of mIPSC amplitude changes after iLTP induction, recorded in control conditions and in the presence of RRETAWA peptide applied before or 15 min after NMDA treatment. M, Averaged traces before (black) and after (red) iLTP induction in control conditions (top) and RRETAWA-treated slices (bottom). Calibration, 10 pA, 20 ms. N, Summary plot of mIPSC amplitude change after 20–22 min of iLTP induction in control conditions and RRETAWA-treated slices (1-way ANOVA, with Holm–Sidak post hoc test, F(2,18) = 21.891, p < 0.001). Data points are shown as mean ± SEM. The numbers on the bars refer to the number of experiments; *p < 0.05, **p < 0.01, ***p < 0.001, NS p >0.05.
Induction of NMDA-dependent iLTP is accompanied by increased activation of β1 integrin
To test whether induction of NMDA-dependent iLTP in mice hippocampal slices is associated with activation of perisynaptic integrins, we took advantage of conformation-specific antibodies that recognize an active form of integrin β1 (catalog #MAB2079Z, Merck; Fig. 4B; Luque et al., 1996; Babayan et al., 2012) and ligand-induced binding site of integrin β3 (catalog #MABT27, Merck; Fig. 4D; Wierzbicka-Patynowski et al., 1999; Diaz et al., 2020). Both antibodies recognize extended conformation of respective integrins, which are considered ligand bound. Additionally, we used two antibodies that do not distinguish active or inactive integrin conformations and therefore recognize total integrin β1 (catalog #MAB2252, Merck; Fig. 4A) and β3 (catalog #MAB1957, Merck; Fig. 4C). We found that 25 min after the induction of iLTP with NMDA, the integrated fluorescence density of active integrin β1 significantly increased in area CA1 stratum radiatum [120 ± 6%, p = 0.020 vs control (ctrl); Fig. 4E]. Signals from total integrin β1 (totβ1) as well as total or active β3 integrin were unchanged (totβ1, 98 ± 8%, p = 0.354 vs ctrl; totβ3, 91 ± 7%, p = 0.382 vs ctrl; actβ3, 107 ± 3%, p = 0.083 vs ctrl; Fig. 4E). Similarly, the density of active integrin β1 puncta increased after iLTP (ctrl, 0.0518 ± 0.0165 µm−2; iLTP, 0.2542 ± 0.0536 µm−2, p = 0.035; Fig. 4F).
Induction of NMDA-iLTP leads to increased level of active β1 integrin. A–D, Exemplary sections of hippocampal CA1 region immunostained against vGAT and different integrins [total integrin β1 (A), active β1 integrin (B), total integrin β3 (C), active β3 integrin (D)] in control conditions (top) and 25 min after induction of NMDA-iLTP (bottom). Enlarged representations of integrins localizing in the vicinity of the vGAT-positive synapses in stratum radiatum are shown in the top right corners. White asterisk on some panels indicates signal originating form blood vesels or glial cells not considered in the analysis. E–H, Statistic analysis performed in stratum radiatum region. Percentage of total fluorescence of different integrins in control conditions and 25 min after iLTP induction (E). Application of NMDA significantly increased integrated density of the active β1 integrin signal, contrary to the active β3 integrins. After iLTP induction, the number of integrin puncta detected in the stratum radiatum changes only in the active β1 integrin group (F). Fraction of integrin puncta colocalizing with vGAT-positive synapses increases after induction of iLTP only for the active β1 integrins (G). The increase in density of integrin signal colocalizing with inhibitory synapses is associated exclusively with activated integrins containing β1 subunits (H). Immunodetection was performed 25 min or 8 min after iLTP induction as indicated below the graphs. Data points are shown as mean ± SEM. The numbers on the bars refer to the number of slices from at least three different animals; *p < 0.05, **p < 0.01, NS p >0.05 (t test or Mann–Whitney test between the iLTP and control group).
Next, we examined immunostaining for β1 and β3 integrins colocalizing with GABAergic synapses. To this end, we performed double immunolabeling for integrins and vGAT, a marker of inhibitory synapses. We found that only a small percentage of vGAT immunopositive GABAergic synapses is associated with active β1 integrin in control conditions (0.9 ± 0.3%; Fig. 4G). Interestingly, the induction of iLTP led to a significant increase in the fraction of inhibitory synapses containing active β1 integrin (6.4 ± 1.5%, p = 0.005 vs ctrl; Fig. 4G). Similarly, the density of synapses immunopositive for vGAT and active β1 integrin increased after iLTP induction (ctrl, 0.0078 ± 0.0036 µm−2; iLTP, 0.0404 ± 0.0104 µm−2, p = 0.018; Fig. 4H). We analyzed the average area of vGAT-positive puncta as a control and found no significant changes after NMDA-iLTP in any of the analyzed groups (Table 2). Additionally, the immunoreactivity of ligand-bound, presumably active β3 integrins at an earlier time point after iLTP induction (8 min after NMDA infusion) is also unchanged in comparison with sham-treated controls (102 ± 3%, p = 0.610 vs ctrl; Fig. 4E). We also performed a colocalization analysis calculating the Manders' coefficient M2, which corresponds to a fraction of a vGAT-positive area of inhibitory synapses overlapping with the integrin signal. It is apparent from this study that after induction of iLTP there is a significant increase in the colocalization of only the active form of β1 integrin in vGAT-positive inhibitory synapses (ctrl, 0.029 ± 0.009; iLTP, 0.184 ± 0.031, p = 0.002; Fig. 5B). Overall, induction of NMDA-iLTP upregulated active β1 integrins and left the density of total β1 integrin unchanged as well as total or active β3 integrin. It suggests that the activation of β1 integrin during NMDA-iLTP does not require the synthesis of new integrins in the analyzed time window.
Percentage of the area occupied by vGAT-positive puncta of stratum radiatum in all analyzed groups
Induction of iLTP increases the colocalization of integrin β1 with inhibitory synapses. A, Exemplary images of hippocampal CA1 region with orthogonal views, immunostained against vGAT and different integrins in control conditions (top) and during the stable phase of iLTP (bottom). Scale bars, 4 µm (xy plane) and 2 µm (orthogonal views). B, The Manders' coefficient of vGAT-positive signal colocalizing with integrin. After iLTP induction the colocalization changes only for integrin β1 in extended conformation presumably in active state. Data points are shown as mean ± SEM. The numbers on the bars refer to the number of slices from at least three different animals; **p < 0.01, NS p >0.05 (t test or Mann–Whitney test between the iLTP and control group).
Activation of β3 integrins with fibrinogen downregulates GABAergic synaptic transmission
Described above, effects of integrin inhibitors and peptides clearly indicate the involvement of these signaling proteins in regulating GABAergic synaptic transmission and plasticity. To further substantiate these findings, we additionally considered the effect of integrin activation by thrombospondin-1 (TSP-1) and fibrinogen, two well-established endogenous ligands of β1 and β3 integrin, respectively (Charrier et al., 2010). First, we compared the amplitude and frequency of mIPSC before and after TSP-1 administration (2 mg/ml; Fig. 6A,B). After 28–30 min of TSP-1 application, mIPSC amplitude was unchanged (sham-treated control, 102 ± 3% of baseline, n = 5; TSP-1, 97.5 ± 2%, n = 7, p = 0.251; Fig. 6B–D), but the frequency significantly increased (sham, 103 ± 2% of baseline, n = 5; TSP-1, 116 ± 3%, n = 7, p = 0.016; Fig. 6E,F). Thus, the results indicate that the blockade of β1 integrin (with ATN-161 or RRETAWA) increased mIPSC amplitude but not frequency (Fig. 3), whereas its activation (with TSP-1) left mIPSC amplitude unchanged but increased its frequency (Fig. 6C,F). Additionally, we analyzed the coefficient of variation CV−2 of mIPSC amplitudes and found that the increased frequency of mIPSC after TSP-1 is accompanied by an augmented CV−2 of the mIPSC amplitudes (sham, 0.99 ± 0.05 normalized to baseline, n = 5; TSP-1, 1.16 ± 0.04, n = 7, p = 0.048; Fig. 6G), indicating that the activation of β1 integrin might increase the probability of presynaptic vesicle release.
Activation of β3 integrin leads to the induction of iLTD. A, Raw mIPSC recording traces before and 30 min after sham, TSP-1, or fibrinogen treatment. Calibration, 40 pA, 2 s. B, mIPSC amplitude changes during sham, TSP-1 (for 30 min), or fibrinogen (for 10 min) treatment. The short-term presence of fibrinogen leads to the induction of iLTD. C, D, Summary plots of mIPSC amplitude changes after 28–30 min of sham, TSP-1 or fibrinogen administration; one-way ANOVA with Holm-Sidak post hoc test, F(2,18) = 10.6, p < 0.001(C), paired t test (D). E, mIPSC frequency changes during sham, TSP-1 (for 30 min), or fibrinogen (for 10 min) treatment. During fibrinogen treatment, the frequency significantly decreases and then returns to baseline after washout. TSP-1 treatment increases mIPSC frequency. F, Summary plot of mIPSC frequency changes after 28–30 min of sham, TSP-1, or fibrinogen administration (Kruskal–Wallis 1-way ANOVA on ranks with Dunn's post hoc test; H(2) = 11.7, p = 0.003). G, Summary plot of changes in the coefficient of variation of mIPSC amplitudes (CV−2) after 28–30 min of sham, TSP-1, or fibrinogen administration. Data are expressed as a ratio to baseline values (1-way ANOVA with Holm–Sidak post hoc test, F(2,19) = 3.73, p = 0.043). Data points are shown as mean ± SEM. The number on the bars refers to the number of recordings; *p < 0.05, ***p < 0.001, NS p >0.05.
We also measured the mIPSC before and after activation of β3 integrins by fibrinogen application (0.2 mg/ml; Fig. 6A). Fibrinogen was applied for 10 min, which resulted in a significant decrease of mIPSC amplitude to 87.4 ± 1.9% after 20 min of washout (before, 54.7 ± 5 pA; after, 47.9 ± 5 pA, n = 9, p < 0.001; Fig. 6B–D), whereas the frequency temporarily decreased during fibrinogen administration (Fig. 6E) but then returned to the control level (sham-treated control, 103 ± 2% of baseline, n = 5; TSP-1, 100 ± 1%, n = 7, p = 0.079; Fig. 6F). Moreover, the coefficient of variation of the mIPSC amplitudes (CV−2) remained unaffected 30 min after the activation of β3 integrins with fibrinogen (Fig. 6G). Thus, whereas the blockade of β3 integrin with cilengitide did not affect inhibitory transmission (Fig. 2), the activation of β3 integrins with fibrinogen led to a stable depression of mIPSC amplitude, which is referred to as fibrinogen-induced iLTD.
β3 integrins control iLTD via the activity of calcineurin and postsynaptic endocytosis
Numerous forms of GABAergic long-term plasticity depend on a postsynaptic calcium influx through NMDA receptors (Bannai et al., 2009; Chiu et al., 2018); hence, to investigate the signaling mechanisms involved in fibrinogen-induced iLTD, we asked whether the activity of NMDARs is required for this form of plasticity. We found that fibrinogen-induced iLTD was unchanged in the presence of the NMDAR antagonist d-APV (at 28–30 min after fibrinogen administration; ctrl, 87.8 ± 1.6% of baseline, n = 9; 50 µm d-APV, 86.6 ± 2.1%, n = 7, p = 0.665; Fig. 7A). Next, we assessed the role of Src, CaMKII, and PKC, that is, the three kinases typically involved in various types of long-term synaptic plasticity. Interestingly, fibrinogen-induced iLTD does not require the activity of these kinases, as it was observed in the presence of PP1 (Src inhibitor, 20 µm, 89.3 ± 1.8%, n = 7, p = 0.539 vs ctrl; Fig. 7D), KN93 (CaMKII inhibitor, 1 µm, 88.7 ± 3.6%, n = 5, p = 0.784 vs ctrl; Fig. 7A), and Go6976 (PKC inhibitor, 2 µm, 85.1 ± 4.1%, n = 6, p = 0.511 vs ctrl; Fig. 7D). Among phosphatases, protein phosphatase 2B (calcineurin) is known to be involved in numerous forms of GABAergic plasticity (Heifets et al., 2008; Muir et al., 2010) and is expressed presynaptically and postsynaptically. (Heifets et al., 2008; Muir et al., 2010). We found that treatment of slices with calcineurin inhibitors (5 µm of FK506 or 10 µm of cyclosporin A) destabilized the fibrinogen-induced iLTD, whose extent started to decay 15–20 min after induction (Fig. 7B) and 28–30 min after fibrinogen application iLTD magnitude was significantly smaller than in the sham-treated iLTD control group (FK506, 96.0 ± 1.4%, n = 8, p = 0.002 vs ctrl; cyclosporin A, 95.5 ± 1.6%, n = 7, p = 0.006 vs ctrl; Fig. 7D). We did not observe any effect of used inhibitors on basic mIPSC amplitude or frequency (Fig. 7G,H).
Calcineurin and endocytosis are necessary for iLTD induced by activation of integrin β3. A–D, Results of pharmacological studies of the molecular mechanism of iLTD induced by fibrinogen. The competitive NMDAR antagonist d-AP5 (50 µm) or CaMKII inhibitor KN93 (2 µm) do not affect iLTD (A). FK506 (20 µm) treatment (to block calcineurin) prevented the stabilization of iLTD following fibrinogen application (B). Postsynaptic endocytosis blocker dynasore (40 µm) prevented the maintenance of fibrinogen-induced iLTD (C). Summary of mIPSC amplitude changes after fibrinogen treatment in the presence of different inhibitors (1-way ANOVA with Holm–Sidak post hoc test, F(7,47) = 3.84, p = 0.002; D). E, F, Endocytosis inhibitor Pitstop-2 (20 µm) decreases the amplitude of mIPSC and blocks induction of fibrinogen-iLTD (Mann–Whitney rank sum test, U = 19, p = 0.836). G, H, The impact of different inhibitors on mIPSC amplitude (E) or frequency (F) assessed after 12–20 min of administration in relation to the baseline. For statistical comparisons, Kruskal–Wallis one-way ANOVA on ranks with Dunn's post hoc test was used for amplitudes, H(6) = 1.96, p = 0.92; one-way ANOVA was used with Tukey's post hoc test for frequencies, F(6,38) = 1.02, p = 0. 42). Data are shown as mean ± SEM. The numbers on the bars refer to the number of recordings. Every significant difference is marked by *p < 0.05, **p < 0.01, or NS p < 0.05.
One of the most common mechanisms of iLTD expression is a reduction of postsynaptic GABAARs, which could occur because of dynamin-dependent endocytosis of membrane receptors (including synaptic ones; Bannai et al., 2009; Muir et al., 2010). To test this scenario, we treated neurons with a dynamin inhibitor, dynasore (80 µm), for 30 min, which did not affect basal mIPSC amplitude (Fig. 7E) but significantly decreased the extent of fibrinogen-induced iLTD (95.0 ± 1.7%, n = 5, p = 0.016 vs ctrl; Fig. 7C,D). Additionally, Pitstop2, another inhibitor of endocytosis (Dutta et al., 2012), also impaired the induction of iLTD by fibrinogen (Pitstop2 plus sham treatment, 94.4 ± 2.1%, n = 6; Pitstop2 plus fibrinogen, 92.7 ± 1.1%, n = 7, p = 0.84; Fig. 7E,F). Interestingly, Pitstop2 infusion (20 µm) led to significant decrease in mIPSC amplitude (before, 66.5 ± 6.2 pA; after, 62.8 ± 5.5 pA, p = 0.045; paired t test). Thus, activation of β3 integrin with fibrinogen induces long-term depression of inhibitory synaptic transmission, which requires the activity of calcineurin and receptor endocytosis.
CaMKII and Src kinases are required for α5β1-dependent iLTP
To elucidate the signaling pathways underlying long-term inhibitory plasticity dependent on α5β1 integrins (and induced by RRETAWA), we first considered the involvement of calcineurin, which was crucial for fibrinogen-induced iLTD. We found that pretreatment with calcineurin inhibitor FK506 (5 µm) did not prevent mIPSC amplitude potentiation with RRETAWA (RRETAWA, 119.3 ± 6.3%, n = 8; FK506, 107.4 ± 2.7%, n = 7, p = 0.104; Fig. 8A,C). Next, we checked the impact of PKC inhibitor Go6976 (2 µm) on RRETAWA-induced iLTP, but no significant effect was found (Go6976, 118.9 ± 4.5%, n = 6, p = 0.957; Fig. 8A,C). Subsequently, we evaluated the role of CaMKII, which has long been known to be important for synaptic plasticity. As we expected, in the presence of CaMKII inhibitor KN93 (2 µm), RRETAWA-induced iLTP was totally abolished (93 ± 6%, n = 8, p < 0.001; Fig. 8B,C). Moreover, when slices were incubated with Src-selective kinase inhibitor PP1 (20 µm), application of RRETAWA, instead of inducing iLTP (as in control conditions), resulted in mIPSC amplitude reduction (85.8 ± 3.3%, n = 7, p < 0.001; Fig. 8B,C). Altogether, these results indicate that CaMKII and Src kinases are key players in the signaling pathway underlying long-term inhibitory plasticity dependent on α5β1 integrins.
CaMKII and Src are required for iLTP induced by inhibition of integrin β1. A, B, Results of pharmacological studies of the molecular mechanism of iLTP induced by RRETAWA. A, The PKC inhibitor Go6976 (2 µm) and calcineurin inhibitor FK506 did not affect iLTP. B, CaMKII inhibitor KN93 and Src inhibitor PP1(20 µm) abolished RRETAWA-induced iLTP. C, Summary of mIPSC amplitude changes after RRETAWA treatment in the presence of different inhibitors (1-way ANOVA with Holm–Sidak post hoc test, F(4,30) = 8.23, p < 0.001). Data are shown as mean ± SEM. The numbers on the bars refer to the number of recordings. Every significant difference is marked by ***p < 0.001.
Discussion
A growing body of evidence demonstrates that numerous synaptic and perisynaptic adhesion proteins such as neuroligin-2 (Ali et al., 2020) or dystroglycan (Früh et al., 2016) regulate GABAergic inhibition. Our work extends this line of investigations by revealing that integrins control inhibitory long-term plasticity and by indicating the underlying signaling pathways. Most interestingly, we found that β1 and β3 subunit-containing integrins tend to mediate opposing plastic effects on GABAergic synapses, suggesting specialized modulatory mechanisms mediated by distinct integrin types. A previous study conducted on cultured spinal cord neurons suggested that β1 and β3 integrins could control the strength of glycinergic synapses by regulating the lateral diffusion of glycine receptor (Charrier et al., 2010). However, GABAergic and glycinergic synaptic transmissions, although both are involved in inhibition, show profound differences not only with respect to the different locus (the former primarily in the brain vs the latter one in the spinal cord) but also in their subcellular synaptic organization at the neuronal network level and the molecular design of synaptic and perisynaptic structures (Renner et al., 2008).
An important aspect of the roles of integrins described here is that they mediate potentiating or inhibitory plastic effects and affect distinct mechanisms of synaptic transmission. Indeed, selective inhibition of β1 integrin led to stable potentiation of mIPSC amplitude without any changes in frequency, whereas activation of β1 integrins with recombinant thrombospondin-1 left the amplitude of mIPSCs intact but increased their frequency. In contrast, activation of β3 integrins by fibrinogen induced iLTD, whereas the inhibition of β3 integrin activity did not affect mIPSC amplitudes or frequency. Thus, blocking β1 integrin has the opposite effect to activating β3 integrin, whereas activation and inhibition of β1 and β3 integrins, respectively, have no impact on the amplitude of the mIPSC. These findings indicate that the efficacy of GABAergic transmission in CA1 pyramidal cells is regulated potently by mechanisms dependent on the interaction of distinct types of integrins with ECM constituents.
Activation of β3 integrin induces iLTD
Our experiments aiming to elucidate the role of integrin β3 in the plasticity of inhibitory synapses revealed that activation of this adhesion protein with fibrinogen induces iLTD, which is associated with endocytosis of membrane receptors and requires the activity of the phosphatase calcineurin, but not kinases CaMKII, Src, or PKC (presented schematically in Fig. 9). These findings appear consistent with the expression mechanisms of several already described types of postsynaptic iLTD (Lu et al., 2000; Wang et al., 2003). For example, prolonged activation of NMDA receptors induces iLTD through calcineurin-dependent dispersal of GABAA receptors from the synapse after dephosphorylation of S327 on the γ subunit (Muir et al., 2010). Thus, the involvement of calcineurin in both NMDA-iLTD and fibrinogen-induced iLTD suggests that they may represent the same type of inhibitory plasticity. However, because fibrinogen-induced iLTD does not require the activity of NMDA receptors, the activation of integrin β3 may occur downstream of calcium influx through NMDA receptors on induction of NMDA-iLTD. Interestingly, the intracellular Ca2+ influx can also lead to the activation of integrins (Rowin et al., 1998).
Interference with β1 or β3 integrin-dependent adhesion results in the opposite effects on the GABAergic synaptic transmission. A, Inhibition of β1 subunit-containing integrins induces NMDAR-dependent inhibitory long-term potentiation through the activation of kinase Src and CaMKII. B, Activation of β3 subunit-containing integrins induces inhibitory long-term depression. Integrin-dependent iLTD requires the activity of phosphatase calcineurin and augmented endocytosis of membrane receptors.
Although fibrinogen is widely recognized as an endogenous activator of β3 integrin, it needs to be stressed that it is not present in the ECM of the physiological brain. Thus, endogenous ligands of brain integrin β3 still need to be discovered. Nevertheless, it needs to be emphasized that on brain hemorrhage, blood fibrinogen enters the brain tissue. This can be important in the context of mechanisms of hemorrhage-induced epilepsy (Derex et al., 2021) that can be potentially driven by fibrinogen-induced suppression of inhibitory transmission. Furthermore, the leakage of the blood–brain barrier, underlying, for example, the pathogenesis of Alzheimer's disease (Brzdak et al., 2017), may lead to the presence of fibrinogen in brain neuropil (Ryu and McLarnon, 2009). Similarly, glial-derived brain tumors may ectopically express fibrinogen and release it into the extracellular microenvironment (Dzikowski et al., 2021), which may explain the frequent occurrence of epilepsy in glioblastoma patients (Armstrong et al., 2016).
Interestingly, genetic-linkage studies have identified an association between the ITGB3 gene, encoding the integrin β3 subunit with autism spectrum disorder (Jaudon et al., 2021). Additionally, mice lacking integrin β3 exhibit behavioral deficits in social memory and increased grooming, the abnormalities with a significant correspondence to autism spectrum disorder in humans (Carter et al., 2011). This points to a potential role for disturbances in integrin-dependent GABAergic plasticity in autism spectrum disorder.
Inhibition of β1 integrin-dependent adhesion results in iLTP
In the present report, we provide extensive pharmacological evidence that blockade of β1 subunit-containing integrins favors potentiation of mIPSCs. Moreover, pretreatment with RRETAWA occluded NMDA-induced iLTP (Fig. 3I), further underscoring the involvement of this integrin in the long-term potentiation of GABAergic synaptic currents. This is in contrast to what was observed for the excitatory synapses, for which the role of β1 integrins has been extensively studied. Indeed, function-blocking antibodies or genetic knockout of β1 integrin impairs the maintenance phase of LTP (Huang et al., 2006; Babayan et al., 2012). It was found that shortly after LTP induction, β1 integrins become activated for a few minutes (Babayan et al., 2012) to drive actin polymerization that consolidates LTP (Kramár et al., 2006). The mechanisms whereby β1 integrins participate in the iLTP are not clear. Colocalization of α3 integrin subunit with inhibitory synapses has been reported in cerebellar Purkinje neurons, where activation of α3β1 integrins was shown to suppress rebound potentiation induced by postsynaptic depolarization (Kawaguchi and Hirano, 2006). In our hands, usage of broad-spectrum integrin ligand peptides that disrupt the function of RGD-binding integrins (GRGDSP), β3/β5 integrins (cilengitide), or β1 integrins (RRETAWA) block NMDA-iLTP. This indicates that during the induction phase of NMDA-iLTP, some integrins, β1 subunit-containing ones especially, are involved. It is particularly interesting that after the induction of NMDA-iLTP, the density and puncta intensity of active β1 integrin increased, without any changes in the level of total β1, total β3, or active β3 integrins (Fig. 4). Importantly, we observed that ∼10% of integrin staining colocalized with inhibitory synapses in the CA1 stratum radiatum. This observation suggests that the involvement of integrins in mechanisms underlying GABAergic plasticity might be because of these proteins operating at inhibitory synapses. Still, some signaling is most likely additionally mediated in the extrasynaptic areas. Thus, our results provide evidence that the activation of integrins containing β1 and probably β5 subunit but not β3 are involved in the induction of NMDA-iLTP.
We observed that thrombospondin-1 used to activate β1 integrin (Charrier et al., 2010) increased mIPSC frequency without changing mIPSC amplitude, clearly demonstrating that some modulatory mechanisms mediated by this integrin are presynaptic. Furthermore, integrin β1 is expressed both presynaptically and postsynaptically, and presynaptic β1 integrin was found to bind postsynaptic ICAM-5 adhesion protein affecting the frequency of mEPSC and inhibiting the spine maturation (Ning et al., 2013). It is thus possible that activation of presynaptic integrin β1 at inhibitory synapses may affect quantal release changing the frequency of mIPSC (Wang et al., 2018).
Integrins are placed ideally to detect external signals and to convert this information into intracellular signaling. One of the most widespread extracellular processes affecting integrins is proteolysis mediated by various enzymes that may reveal integrin ligands by processing ECM. Recently, it was shown that the extracellular activity of matrix metalloprotease 3 (MMP-3) drives the expression of NMDA-iLTP (Wiera et al., 2021). It is thus possible that at inhibitory synapses, the proteolytic activity of MMP-3 leads to the release of extracellular peptides that bind and regulate integrin β1 during NMDA-iLTP. This hypothesis raises an essential question about endogenous ligands of neuronal β1integrins. Hippocampal neurons accumulate a cleaved form of α5 laminin that can bind to α3β1 integrin and control the synaptic transmission (Omar et al., 2017). Moreover, non-fibril-forming collagens that bind β1 integrin are widely expressed by neurons. For instance, a proteolytically released fragment of collagen XIX promotes the formation of inhibitory synapses (Su et al., 2016). Finally, hydrolyzis of brain chondroitin sulfate proteoglycans with chondroitinase activates β1 integrin and promotes the motility of dendritic spines (Orlando et al., 2012). Although it requires further studies in the context of GABAergic plasticity, peptides derived from brain chondroitin sulfate proteoglycans that form perineuronal nets may play the role of putative integrin ligands.
Conclusions
Here, we discovered the novel role of β1 and β3 subunit-containing integrins as key players in the endogenous modulatory systems capable of potently regulating the efficacy of inhibitory transmission and long-term plasticity. Interestingly, integrin-dependent modulation may occur bidirectionally, and the effects of β1 and β3 integrins tend to be opposing, indicating that distinct integrins could orchestrate various modulatory scenarios at GABAergic synapses. Consequently, given the diversity of inhibitory interneurons and their GABAergic synaptic connectivity, it is tempting to hypothesize that different integrins might specialize in regulating modulatory processes at distinct inhibitory synapses (Favuzzi and Rico, 2018). Future studies should explore the input specificity of GABAergic plasticity in the context of different integrins. Finally, understanding the modulatory role of integrin-dependent adhesion within local inhibitory microcircuits may open a promising perspective for studies on clinically relevant integrin subtype-specific pharmaceutical agents targeting GABAergic inhibition.
Footnotes
This work was supported by the National Science Center (Poland) Grant SONATA 2017/26/D/NZ4/00450 (to G.W.). J.W.M. was partially supported by National Science Center Grant 2018/31/B/NZ4/01998 (comparative studies of pyramidal cells vs interneurons in Task No. 2). P.B. was supported by the Ministry of Science and Higher Education (Poland) Scholarship SMN/16/1284/2020.
The authors declare no competing financial interests.
- Correspondence should be addressed to Grzegorz Wiera at grzegorz.wiera{at}umed.wroc.pl or Jerzy W. Mozrzymas at jerzy.mozrzymas{at}umed.wroc.pl
