Abstract
Elevated intraocular pressure (IOP) triggers glaucoma by damaging the output neurons of the retina called retinal ganglion cells (RGCs). This leads to the loss of RGC signaling to visual centers of the brain such as the dorsolateral geniculate nucleus (dLGN), which is critical for processing and relaying information to the cortex for conscious vision. In response to altered levels of activity or synaptic input, neurons can homeostatically modulate postsynaptic neurotransmitter receptor numbers, allowing them to scale their synaptic responses to stabilize spike output. While prior work has indicated unaltered glutamate receptor properties in the glaucomatous dLGN, it is unknown whether glaucoma impacts dLGN inhibition. Here, using DBA/2J mice, which develop elevated IOP beginning at 6–7 months of age, we tested whether the strength of inhibitory synapses on dLGN thalamocortical relay neurons is altered in response to the disease state. We found an enhancement of feedforward disynaptic inhibition arising from local interneurons along with increased amplitude of quantal inhibitory synaptic currents. A combination of immunofluorescence staining for the γ-aminobutyric acid (GABA)A-α1 receptor subunit, peak-scaled nonstationary fluctuation analysis, and measures of homeostatic synaptic scaling pointed to an ∼1.4-fold increase in GABA receptors at postsynaptic inhibitory synapses, although several pieces of evidence indicate a nonuniform scaling across inhibitory synapses within individual relay neurons. Together, these results indicate an increase in inhibitory synaptic strength in the glaucomatous dLGN, potentially pointing toward homeostatic compensation for disruptions in network and neuronal function triggered by increased IOP.
- GABA
- glaucoma
- homeostatic plasticity
- inhibition
- synaptic scaling
- thalamus
Significance Statement
Elevated eye pressure in glaucoma leads to loss of retinal outputs to the dorsolateral geniculate nucleus (dLGN), which is critical for relaying information to the cortex for conscious vision. Alterations in neuronal activity, as could arise from excitatory synapse loss, can trigger homeostatic adaptations to synaptic function that attempt to maintain activity within a meaningful dynamic range, although whether this occurs uniformly at all synapses within a given neuron or is a nonuniform process is debated. Here, using a mouse model of glaucoma, we show that dLGN inhibitory synapses undergo nonuniform upregulation due to addition of postsynaptic γ-aminobutyric acid receptors. This is likely to be a neuronal adaptation to glaucomatous pathology in an important subcortical visual center.
Introduction
The goal of this study was to determine whether glaucoma, a disease that culminates in blindness from loss of retinal ganglion cells (RGCs) and their projections to the brain (Calkins, 2012; Weinreb et al., 2014), leads to homeostatic scaling of inhibitory synapses made onto thalamocortical (TC) relay neurons in the dorsolateral geniculate nucleus (dLGN) of the thalamus. The dLGN processes visual inputs directly from RGCs and transmits that information to the cortex, making it critical for conscious vision (Kerschensteiner and Guido, 2017; Guido, 2018; Liang and Chen, 2020). dLGN TC neurons receive inhibition from two principal sources—feedback projections from the thalamic reticular nucleus (TRN) and feedforward inhibition from local inhibitory interneurons (LINs)—that play important roles in shaping TC neuron receptive fields and setting resting membrane potential and firing mode (X. Wang et al., 2011; Hirsch et al., 2015; Cox and Beatty, 2017).
Synaptic scaling is a homeostatic neuronal adaptation, whereby altered synaptic inputs and/or disrupted activity levels lead neurons to modulate the synaptic strength to maintain action potential output (G. G. Turrigiano et al., 1998; G. Turrigiano, 2012). This phenomenon has been reported to occur at both inhibitory and excitatory synapses in response to activity modulation in dissociated cell cultures, in the visual cortex in vivo through intraocular tetrodotoxin (TTX) injection or visual deprivation, in the auditory system following hearing loss, and in response to seizure activity (Otis et al., 1994; G. G. Turrigiano et al., 1998; Desai et al., 2002; Wierenga et al., 2005; Ibata et al., 2008; Maffei and Turrigiano, 2008; Goold and Nicoll, 2010; Joseph and Turrigiano, 2017; Teichert et al., 2017; Gonzalez-Islas et al., 2023). Conventionally, synaptic scaling is thought to be the result of a global multiplicative process that triggers uniform changes at each of a neuron's synaptic inputs via somatic Ca2+ signals (Ibata et al., 2008; Goold and Nicoll, 2010; Wenner, 2011; G. Turrigiano, 2012; Joseph and Turrigiano, 2017). However, it is increasingly apparent that many neurons employ a “divergent” process, whereby scaling is nonuniform, with different synapses within individual neurons undergoing different degrees of scaling (Hanes et al., 2020; Koesters et al., 2024), although the mechanisms and conditions driving such scaling are still debated.
We have shown previously that during glaucoma, retinal output signaling to the dLGN is disrupted, which is apparent as an intraocular pressure (IOP)-dependent diminishment of RGC excitatory drive to postsynaptic TC neurons (Van Hook et al., 2020; J. C. Smith et al., 2022). We have also found that in both a mouse model of glaucoma and in mice where bilateral enucleation leads to a loss of RGC synaptic input to the dLGN, TC neurons display an increase in intrinsic excitability, firing more action potentials in response to depolarizing stimulation (Van Hook et al., 2020; Bhandari et al., 2022; J. C. Smith et al., 2022). Although disrupted synaptic function and altered TC neuron spiking seem to be likely triggers for homeostatic scaling of excitatory synapses in the dLGN, there were no detectable effects on quantal excitatory current amplitudes in those studies. However, as diminished drive from RGCs is expected to also disrupt other members of the dLGN synaptic network such as feedforward inhibition, it remains to be seen whether glaucoma alters dLGN inhibitory synapses. Here, we tested this possibility, finding an increase in the strength inhibitory synapses on dLGN TC neurons from 10- to 14-month-old mice with glaucoma. This was apparent as an ∼1.4-fold increase in the number of GABA (γ-aminobutyric acid) receptors present at inhibitory synapses without detectable change in single-channel conductance, although several parallel analysis approaches point toward nonuniform “divergent” scaling across the population of inhibitory synapses on individual TC neurons. Together, these results indicate an increase in synaptic strength at a population of inhibitory synapses in the dLGN in mice with elevated IOP, possibly pointing to a homeostatic adaptation to disrupted function of the visual pathway accompanying glaucomatous pathology.
Materials and Methods
Animals
For this study, we used male and female DBA/2J (D2; Jackson Laboratory 000671; RRID, IMSR_JAX:000671; John et al., 1998) and DBA/2Jgpnmb+ (D2-control; Jackson Laboratory 007048; RRID, IMSR_JAX:007048; Howell et al., 2007) mice, aged 10–14 months. Elevated IOP and the consequent glaucoma in DBA/2J mice are the result of iris stromal atrophy and iris pigment dispersion arising from homozygous mutations to the Gpnmb and Tyrp1 genes (GpnmbR150X and Tyrp1b mutations, respectively; Anderson et al., 2002; Howell et al., 2007). All animal procedures were approved by the Institutional Animal Care and Use Committee at the University of Nebraska Medical Center. Mice were housed in a 12/12 h light/dark cycle and provided food and water ad libitum. IOP measurements were taken in mice lightly anesthetized with isoflurane (2%) approximately monthly beginning at 2–5 months of age using a TonoLab rebound tonometer. D2 mice (N = 43) had significantly higher-peak IOPs than the D2-control mice (N = 41), with maximum IOP averages of both eyes being 15.6 ± 0.3 mmHg for D2-control mice and 23.9 ± 0.8 mmHg for D2 mice (p = 1.3 × 10−13, unpaired t test).
Slice preparation and electrophysiology
For patch-clamp recording, we prepared 250-μm-thick coronal slices or 300-μm-thick parasagittal slices (Turner and Salt, 1998; Chen and Regehr, 2000; J. C. Smith et al., 2022) containing the dLGN and optic tract using the protected recovery method (Ting et al., 2014, 2018). Mice were killed by CO2 asphyxiation and cervical dislocation and quickly decapitated. Brains were dissected into a slush of artificial cerebrospinal fluid (aCSF) comprised of as follows (in mM), 128 NaCl, 2.5 KCl, 1.25 NaH2PO4, 24 NaHCO3, 12.5 glucose, 2 CaCl2, and 2 MgSO4 and bubbled with 5% CO2/95% O2. After chilling in aCSF for ∼30 s, brains were hemisected using a razor blade with a cut oriented ∼5° from the medial longitudinal fissure and ∼20° from the horizontal plane. The cut surface was glued to a flat agar block positioned on the stage of a Leica VT1000S Vibratome which was subsequently submerged in ice-cold aCSF. The 300-µm-thick slices were cut, visually inspected to verify they contained dLGN and optic tract, and submerged in an N-methyl-D-glucamine–based solution (in mm: 92 NMDG, 2.5 KCl, 1.25 NaH2PO4, 25 glucose, 30 NaHCO3, 20 HEPES, 0.5 CaCl2, 10 MgSO4, 2 thiourea, 5 L-ascorbic acid, and 3 Na-pyruvate, bubbled with 5% CO2/95% O2 and warmed to 33°C) for 12 min. After the NMDG solution incubation, slices were transferred to room temperature aCSF and incubated for at least 1 h prior to initiating patch-clamp recording.
For patch-clamp recording, slices were transferred to a recording chamber, secured with a nylon net, positioned on the stage of an Olympus BX51WI microscope, and superfused with aCSF warmed to 30–33°C with an in-line solution heater (Van Hook, 2020). For optic tract stimulation in parasagittal slices, a broken patch pipette (opening, ∼50 microns) filled with aCSF was positioned in the optic tract ∼500 µm from the dLGN and used as a stimulating electrode. dLGN TC relay neurons in the dLGN core were targeted for whole-cell voltage–clamp recording based on the soma size and morphology using patch pipettes (4–7 MΩ) filled with a Cs-based solution (in mM: 120 Cs-methanesulfonate, 2 EGTA, 10 HEPES, 8 TEA-Cl, 5 ATP-Mg, 0.5 GTP-Na2, 5 phosphocreatine-Na2, 2 QX-314). Reported voltages are corrected for a 10 mV liquid junction potential. An Axon MultiClamp 700B amplifier and Digidata 1550 Digitizer (Molecular Devices) were used for electrophysiology data acquisition. Current stimuli (0.1–0.3 ms, up to 1 mA) were delivered to the optic tract using an A-M Systems Instruments Isolated Pulse Stimulator. Signals were sampled online at 10 kHz. For each cell, stimulus amplitude, sign, and duration were adjusted to the minimum stimulus strength needed to reliably obtain a maximal excitatory response amplitude measured at −70 mV. Feedforward disynaptic inhibition (Blitz and Regehr, 2005; Augustinaite and Heggelund, 2018) was measured in response to optic tract stimulation by voltage clamping at 0 mV. Series resistance was partially compensated (60–70%) using the amplifier circuitry. When used, as indicated in Results, pharmacological agents (CNQX, D-AP5, TTX, SR95531) were added to the superfusate from 1,000 times stock solutions and bath-applied. CNQX, D-AP5, and SR95531 (gabazine) were purchased from Tocris Bioscience, while TTX was purchased from Abcam. CNQX and SR95531 stocks were prepared in dimethylfulfoxide, while D-AP5 and TTX stocks were prepared in deionized water.
Miniature inhibitory postsynaptic currents (mIPSCs) were measured at 0 mV in the absence of stimulation using aCSF supplemented with 20 μM CNQX, 50 μM D-AP5, and 0.5–1 μM TTX. No series resistance compensation was used for mIPSC recordings due to the increased noise it contributes. mIPSCs were detected and analyzed using the MiniAnalysis software (Synaptosoft) with an amplitude threshold of 4.5 pA. Fits of mIPSC amplitude and baseline noise distributions revealed good separation of recording noise and mIPSCs with these detection parameters. mIPSC amplitude for individual cells is reported as the average amplitude of all events analyzed from that cell. Reported mIPSC frequency for each cell is the median instantaneous frequency of analyzed events from that cell.
We used two analysis approaches to test for and determine the degree of inhibitory synaptic scaling from mIPSC data. In the first, we used linear regression to fit the relationship of rank-ordered mIPSC amplitudes from D2-control and D2 TC neurons (G. G. Turrigiano et al., 1998; Hanes et al., 2020). For this, 1,900 mIPSCs from D2 recordings and 1,900 mIPSCs from D2-control recordings were randomly selected to ensure an equal number of mIPSCs from each cell from a genotype along with an equal overall number of events between D2 and D2-control. Based on the fit parameters of the best-fit line to the rank-order–matched scatterplot, we downscaled the D2 mIPSCs, excluded those falling below the mIPSC detection threshold (4.5 pA), and compared the cumulative distribution of downscaled mIPSCs with the control mIPSC distribution using a Komolgorov–Smirnov (K-S) test. In the second approach (Kim et al., 2012), we iteratively downscaled the population of D2 mIPSC amplitudes using scaling factors of 1.05–1.75. After pruning downscaled event amplitudes that fell below our detection threshold (4.5 pA), we compared the cumulative distributions of D2-control and downscaled D2 mIPSC amplitudes using a K-S test, using finer gradations of scaling factors to determine the one that produced the highest K-S p value.
Peak-scaled nonstationary fluctuation analysis
To estimate single-channel conductance and numbers of synaptic GABAA receptors in dLGN TC neurons, we performed a peak-scaled nonstationary fluctuation analysis of the decay phase of mIPSCs (Traynelis et al., 1993; Kilman et al., 2002). In cells with low-enough mIPSC frequency to ensure no overlap of the mIPSC decay with other events, we scaled the average peak-aligned mIPSC waveform to the peak amplitude of each event (18–68 events per cell) and calculated the variance over the decay phase. Cells were included in this analysis if there was no statistically significant linear correlation of mIPSC peak amplitude with decay time constant, indicating adequate voltage clamp. The relationship of variance to the mean amplitude during the IPSC decay was fit with a parabolic function as follows:
Histology
For immunofluorescence staining for the vesicular GABA transporter (vGAT) or GABAA receptor alpha subunit (GABAA-α1) in fixed brain sections, mice were killed by CO2 asphyxiation and cervical dislocation. Brains were rinsed briefly in phosphate-buffered saline (PBS) and then immersed in 4% paraformaldehyde in PBS for 4 h while rocking at room temperature. Brains were then rinsed three times for 15 min with PBS and immersed in 30% sucrose in PBS overnight for cryoprotection. Brains were then embedded in 3% agar, and 50-µm-thick sections were cut using a Leica VT1000S vibratome and mounted on Superfrost Plus slides, which were stored at −20°C. For immunofluorescence staining, slices were thawed and rinsed briefly in PBS. Mounted brain sections were blocked and permeabilized for 1 h at room temperature in a blocking buffer comprised of 0.5% TritonX-100 and 5.5% donkey serum in PBS. They were then incubated for two nights in primary antibody (1:500 rabbit-anti-GABAA-α1; Synaptic Systems 224 203; RRID, AB_2232180; or rabbit-anti-vGAT, Synaptic Systems 131 002; RRID, AB_887871) at 4°C, rinsed six times for 10 min in PBS, incubated for 2.5 h in an Alexa Fluor 568-conjugated secondary antibody (1:200 donkey-anti-rabbit IgG conjugated to Alexa Fluor 568; Thermo Fisher Scientific A10042; RRID, AB_2534017), rinsed three times for 5 min, and coverslipped with VectaShield Hardset.
Imaging was performed using a Scientifica SliceScope equipped with two-photon imaging capabilities including a MaiTai HP Ti:sapphire laser that was tuned to 800 nm with power set to 80 μW. A series of optical sections was acquired through the z-axis of the tissue in the dLGN core (0.3 μm spacing) with a 74 × 74 μm field of view (14 pixels/micron). Four images were acquired at each optical plane and averaged to reduce noise. For analysis, background was subtracted using a 50 pixel rolling ball for GABAA-α1. For GABAA-α1, the average pixel intensity of each image in the stack was measured and the average intensity values of the six peak optical sections through the image stack was used for the intensity. vGAT images were filtered with the “unsharp mask” tool (25 pixel radius; weight, 0.6). vGAT puncta were detected and analyzed, and regions of interest generated for each punctum using the Synapse Counter plug-in with a 10 μm rolling ball radius and a min–max size threshold of 0.26–39 μm2.
Experimental design and statistical analysis
Statistical analysis was performed using GraphPad 10. Normality of data was assessed using a D'Agostino and Pearson’s test. Significant differences between normally distributed datasets were assessed using a two-tailed nested t test for instances where we recorded multiple cells for individual mice. This approach avoids pitfalls arising from pseudoreplication (Eisner, 2021). Datasets that followed a logarithmic distribution were log-transformed prior to statistical testing. For instances where we made single measurements from individual mice (immunofluorescence staining), we used a conventional unpaired t test. A paired t test was used to test for effects of acute TTX application on individual cells. Data from individual cells are displayed as individual points for electrophysiology analyses and for individual mice for immunofluorescence analysis. A K-S test was used to compare cumulative distributions. An alpha value of 0.05 was selected as the threshold for statistical significance for all tests. For vGAT analysis, the intensity of each vGAT punctum was measured as the average pixel intensity within regions of interest corresponding to each detected punctum. vGAT density was calculated for each image, and D2 and D2-control measurements were compared using a t test. The vGAT punctum size and intensity were log-transformed before a comparison with a nested t test. Data are presented as mean ± SEM for normally distributed data and median(interquartile range, IQR) for non-normally distributed data, as indicated in the text. Sample sizes (number of mice and number of cells) are reported in the figure legends.
Results
To test the possibility that glaucoma leads to homeostatic alterations at feedforward inhibitory synapses in the dLGN, we prepared parasagittal slices containing the dLGN and stimulated RGC axons by positioning a stimulating electrode in the optic tract. The current stimulus amplitude was increased to reliably evoke a maximal response. By voltage-clamping TC neurons at the chloride and cationic equilibrium potentials (−70 and 0 mV, respectively), we were able to isolate both monosynaptic retinogeniculate excitatory input and disynaptic feedforward inhibition (Blitz and Regehr, 2005; Augustinaite and Heggelund, 2018), which arises from LINs (Fig. 1A–D). Confirming adequate isolation of inhibitory and excitatory responses, the synaptic current evoked while TC neurons were voltage clamped at 0 mV was entirely blocked by application of SR95531 (20 μM; 100 ± 2% block; n = 6 cells; p = 0.027 paired t test; Fig. 1A). Similar to our previous work, the amplitude of the retinogeniculate EPSC was lower in slices from D2 mice compared with D2-controls [median(IQR), D2-control, 1,734(1,188–2,120) pA; n = 28 cells, nine mice; D2, 668(334–1,417) pA, n = 39 cells, nine mice; p = 0.0014; nested t test of log-transformed data], indicating a loss of retinogeniculate input in the D2 mouse dLGN (Fig. 1E). The amplitude of the disynaptic IPSC was also somewhat lower in D2 relative to D2-control (D2-control, 856(624–1,216) pA; D2, 578(350–851) pA; p = 0.039; nested t test of log-transformed data), pointing to a reduction in the strength of feedforward inhibition from local interneurons (Fig. 1F).
Retinogeniculate excitation and feedforward disynaptic inhibition in dLGN brain slice recordings from D2-control and D2. A, Voltage-clamp recording of monosynaptic retinogeniculate EPSC (black, −70 mV) in response to optic tract stimulation in a brain slice from a D2-control mouse. The disynaptic IPSC was recorded at 0 mV (blue) and was entirely blocked by gabazine (SR95531, 20 µM, red). B, Group data of D2-control EPSC and IPSC amplitudes (28 cells from 9 mice). C, EPSC and IPSC recordings from a D2 mouse. D, Group data of EPSC and IPSC amplitudes from D2 brain slices (39 cells from 9 mice). E, F, Group data of EPSC and IPSC amplitudes. Nested t test performed on log-transformed data. **p < 0.01; *p < 0.05. G, Normalizing the IPSC to the EPSC amplitude shows a relative increase in the strength of feedforward inhibition in recordings from D2 brain slices. Nested t test performed on log-transformed data. *p < 0.05.
We next sought to determine whether this reduction in IPSC amplitude scaled with the reduced EPSC. This was undertaken to test whether the reduced feedforward inhibition simply reflects reduced strength of excitatory drive from RGC axons, which would be the case if inhibition was diminished in proportion to the diminishment of excitation. We assessed this by normalizing the IPSC to the EPSC amplitudes for individual cells (IPSC/EPSC; Fig. 1G). If that ratio is dissimilar in D2-control versus D2 recordings, then it would imply that mechanisms downstream of retinogeniculate excitation are acting on the feedforward inhibitory synapses to modulate their strength. In this analysis, the IPSC/EPSC was 0.53(0.42–0.73) [median(IQR)] for D2-control and 0.88(0.46–2.2) for D2 slices (p = 0.016, nested t test of log-transformed data). The higher IPSC/EPSC ratio in TC neurons from D2 mice indicates that feedforward inhibition, while reduced relative to controls, is higher than would be expected if it simply scaled with the reduced retinogeniculate drive. This implies a relative scaling up of feedforward inhibition to TC neurons in the D2 dLGN.
To test whether this enhancement of feedforward inhibition reflects changes in the quantal properties of inhibitory synaptic transmission in the dLGN, we next sought to record single-vesicle synaptic currents in the absence of stimulation. This was accomplished by voltage-clamping TC neurons at 0 mV in the presence of inhibitors of glutamatergic synapses (20 μM CNQX and 50 μM D-AP5). We found that acute application of TTX (0.5–1 μM) led to a reduction in both IPSC amplitude (16 ± 2 pA to 12 ± 1 pA; n = 13 cells; p = 0.009, paired t test) and frequency (48 ± 2 Hz to 33 ± 3 Hz; n = 13 cells; p = 8 × 10−6, paired t test), suggesting that even under glutamatergic blockade, spontaneous inhibition represents some level of action potential-driven multiquantal GABA release in the dLGN (Fig. 2). Therefore, we performed recordings in the presence of TTX along with glutamatergic blockers to isolate quantal mIPSCs. Under these conditions, we found that the mIPSC frequency was not significantly different between recordings from D2-control and D2 mice (D2-control, 34 ± 2 Hz; n = 19 cells; eight mice; D2, 31 ± 3 Hz; n = 22 cells; nine mice; p = 0.46, nested t test; Fig. 2E–G). However, the mIPSC amplitude was significantly higher in D2 mice compared with D2-controls [median(IQR); D2-control, 12.3(11.1–16.6) pA; D2, 18.9(14.4–23.8) pA; p = 0.0092, nested t test; Fig. 2H,I]. The mean mIPSC amplitude for all cells recorded from individual mice was weakly but significantly correlated with peak IOP measured (Fig. 2J; R2 = 0.26; p = 0.037), pointing toward a link between disease severity and effects on inhibitory synaptic function. There was no significant difference in mIPSC amplitude between male and female D2 mice (p = 0.24, nested t test). Thus, there appears to be an increase in quantal IPSC amplitude in the D2 dLGN, likely reflecting an IOP-related postsynaptic increase in inhibitory synaptic strength.
Measurements of quantal inhibitory currents in dLGN brain slices from D2 mice. A, Example traces of inhibitory synaptic currents measured in the absence of stimulation before (sIPSCs) and after (mIPSCs) application of 0.5 µM TTX. B, Example mIPSC waveforms from the example in A. C, Group data showing that TTX application led to a significant reduction in mIPSC amplitude (**p < 0.01, paired t test). D, Group data showing that TTX application led to a significant reduction in mIPSC frequency (p < 0.00001, paired t test). n = 13 cells. E, mIPSC recordings from D2-control and D2 mice obtained while voltage-clamping at 0 mV in the presence of 0.5 µM TTX, 20 µM CNQX, and 50 µM D-AP5. F, Noise-amplitude histograms show good separation of measured mIPSC amplitudes from recording noise for the examples shown. G, Group data of mIPSC frequency of recordings from D2-control (19 cells from 8 mice) and D2 (22 cells from 9 mice). Nested t test, ns p > 0.05. H, Example average waveforms of mIPSCs detected from the recordings in E. I, Group data of mIPSC amplitude, **p < 0.01, nested t test. J, The average mIPSC amplitude from all recorded cells for each mouse was linearly correlated with peak IOP.
We next applied two tests of synaptic scaling to our mIPSC amplitude data (Fig. 3). In the first (Fig. 3A–C), we used linear regression to fit the rank-order paired scatterplot of mIPSC amplitudes from D2-control and D2 recordings with a line (G. G. Turrigiano et al., 1998; Hanes et al., 2020). This produced a fit with a scaling factor (slope) of 1.45 and a y-intercept of −0.81 pA. We tested whether this linear regression was an accurate representation of the synaptic upscaling by downscaling the D2 mIPSC amplitudes by the 1.45 scaling factor and −0.81 pA y-intercept, excluding those that fell below our mIPSC detection threshold (4.5 pA), and using a K-S test to determine whether the distribution of downscaled D2 amplitudes was significantly different from the distribution of D2-control amplitudes. The cumulative distributions displayed some overlap but had a K-S test with p = 0.0071, indicating statistically significant nonoverlap of the two distributions (Fig. 3B). Likewise, when we used only the scaling factor (1.45) to downscale the D2 mIPSCs without incorporating the intercept, the cumulative distributions also showed statistically significant nonoverlap (p = 0.00026). When we examined the ratio of the rank-ordered D2 to D2-control mIPSC amplitudes as a function of D2-control amplitudes, we found considerable heterogeneity in the D2/D2-control amplitude ratios, rather than a consistent value (Fig. 3C), which points toward nonuniform synaptic scaling (Hanes et al., 2020).
Tests for inhibitory synaptic scaling. A, The plot of rank-ordered mIPSC amplitudes from D2 and D2-control recordings (1,900 mIPSCs from each group) fit using a linear regression model (solid line). Fit parameters had a slope of 1.45 and a y-intercept of −0.81. Dotted line, unity. B, Cumulative histograms of mIPSC amplitudes from recordings of D2-control (solid black line), D2 (solid red line), and D2 that were downscaled using the linear fit parameters from A. The downscaled mIPSC amplitude distribution significantly diverged from the D2-control distribution (p = 0.0071, K-S test). C, Rank-ordered mIPSC ratio plot shows that D2/D2-control ratios do not converge at a single scaling factor. D, Iterative scaling approach. K-S test p values from comparisons of the iteratively downscaled D2 distribution plotted against scaling factors. The scaling factor with the highest p value was 1.405, with p = 0.0031. E, Cumulative distributions of the D2-control mIPSC amplitudes and the D2 distribution downscaled by 1.405.
For the second approach, we applied an iterative scaling strategy (Kim et al., 2012) by downscaling the D2 mIPSC amplitudes by a range of scaling factors (1.05–1.75), excluding resulting values that fell below the 4.5 pA detection threshold, and comparing the resulting cumulative distributions with the D2-control mIPSC amplitude distribution using a K-S test (Fig. 3D,E). In this case, a scaling factor of 1.405 produced a distribution with the highest K-S test p value (p = 0.0031), reflecting the closest alignment of the downscaled D2 mIPSC distribution with the D2-control distribution. Together, these analysis results indicate that inhibitory input to dLGN TC neurons undergoes synaptic scaling by a factor of ∼1.4–1.45 in D2 mice, although the range of D2/D2-control ratios and the observation that downscaled D2 distributions significantly diverge from the control distribution point away from a uniform scaling of all inhibitory synapses on individual TC neurons. In the Discussion section, we elaborate on interpretation of this and the potential underlying mechanisms.
These above results point to an enhancement of inhibition at the postsynaptic GABA receptors on dLGN TC neurons. To test whether this represents a modulation of single-channel conductance, as might occur by changes in receptor stoichiometry or modulation effects, versus a change in the number of synaptic GABA receptors, we performed a peak-scaled nonstationary fluctuation analysis of the mIPSC decay phase from recordings in D2-control and D2 TC neurons (Traynelis et al., 1993; Fig. 4). In the datasets used for this analysis, there was no significant correlation of mIPSC decay time constant with mIPSC amplitude, indicating adequate voltage clamp. We found that on average, D2-control inhibitory synapses had a single-channel conductance (γ) of 18.8 ± 2.4 pS (n = 15 cells, six mice), which was not significantly different from measurements obtained from D2 TC neurons (17.2 ± 2.5 pS; n = 15 cells, six mice; p = 0.64, nested t test). Synaptic GABAA receptors on dLGN TC neurons are likely comprised largely of α1β2γ2 subunits based on immunostaining and in situ hybridization studies (Persohn et al., 1992; Wisden et al., 1992; Fritschy and Mohler, 1995; Pirker et al., 2000). Our measured single-channel conductance values align closely with nonstationary fluctuation measurements and single-channel recordings from TC neurons in other thalamic nuclei which have a similar receptor complement (Browne et al., 2001; Schofield and Huguenard, 2007). Although we did not detect changes in the average single-channel conductance (e.g., Fig. 4E), we did find that D2 TC neurons had a higher GABA receptor number (N) at each synapse (27 ± 4) compared with D2-controls (15 ± 1; p = 0.02, nested t test; Fig. 4F). Thus, an increase in the number of postsynaptic GABA receptors on dLGN TC neurons appears to contribute to increased inhibitory synaptic strength in dLGN TC neurons in D2 mice. An important caveat that influences this interpretation is that recorded mIPSCs potentially arise from a multiple of sources including two distinct populations of LINs (Leist et al., 2016), TRN projections (Kerschensteiner and Guido, 2017; Guido, 2018), and pretectal inputs (S. Wang et al., 2002), meaning that electrophysiological measurements of mIPSC parameters and the nonstationary fluctuation analysis are average measurements of GABA receptor conductance and number across these sources.
Peak-scaled nonstationary fluctuation analysis of mIPSCs. A, B, Example individual mIPSC events (gray traces) and mean event (black trace) from a D2-control recording (A) and D2 recording (B). Bottom, the mIPSC decay time constants were not significantly correlated with mIPSC amplitudes indicating adequate voltage clamp. C, Variance–mean plot from examples in A and B, showing parabolic fit parameters including single-channel current (i) and peak number of open receptors (N). D, Variance–mean plot of averaged (±SEM) variance and mean of mIPSC decay phases from D2-control recordings (15 cells from 6 mice) and D2 recordings (15 cells from 6 mice). Parameters of the parabolic fit are single-channel current (i) and peak number of open receptors (N). E, Single-channel conductance, measured by correcting for a 70 mV Cl− driving force, was not significantly different between D2-control and D2 (p > 0.05, nested t test). F, N was higher in D2 recordings compared with D2-control (*p < 0.05, nested t test).
To test whether we could detect increased GABAA receptor expression in dLGN using an independent approach, we performed immunofluorescence staining for GABAA-α1 subunits in fixed dLGN sections from D2-control and D2 mice and analyzed the relative intensity of GABAA-α1 staining in optical sections representing the peak staining intensity from the dLGN core in each slice (Fig. 5). In this approach, we found that GABAA-α1 immunofluorescence intensity was higher in sections from D2 mice (33.7 ± 1.5; n = 17 mice) versus D2-control mice (26.9 ± 2.0; n = 12 mice; p = 0.0095, unpaired t test). This result supports physiological data pointing to an increase in GABAA receptor number at dLGN inhibitory synapses.
Increased GABAA-α1 immunofluorescence in dLGN of D2 mice. A, Two-photon confocal images of GABAA-α1 staining. Images are average intensity projections of six optical sections (0.3 µm spacing). B, Quantification of average pixel intensity shows an elevated GABAA-α1 staining in dLGN of D2 mice (p < 0.01, t test).
Finally, we sought to determine whether glaucoma led to disruption in the size and/or number of presynaptic GABAergic terminals in the dLGN, as changes to the presynaptic vGAT have been implicated in activity-dependent presynaptic homeostatic plasticity (De Gois et al., 2005; Prestigio et al., 2021; Fig. 6). To test this, we performed immunofluorescence staining of fixed dLGN slices for vGAT and quantified vGAT terminal size, density, and staining intensity. The size of detected vGAT puncta was similar in D2-control and D2 dLGN sections (D2-control, 1.06 ± 0.03 μm2; n = 13; D2, 0.99 ± 0.04 μm2; p = 0.17, nested t test). Additionally, the density of vGAT puncta was not significantly different between D2-control and D2 (D2-control, 66.7 ± 3.2 puncta/1,000 μm2; D2, 71.4 ± 2.5 puncta/1,000 μm2; p = 0.25, unpaired t test). The staining intensity of identified vGAT puncta was comparable between D2-control and D2 dLGN (p = 0.51, nested t test). Thus, there does not appear to be an effect of glaucoma on presynaptic GABAergic structures in the dLGN.
vGAT labeling is similar in D2-control and D2 dLGN. A, Example two-photon confocal images (single optical sections) of vGAT labeling in dLGN. B, Quantification of detected vGAT punctum density (ns p > 0.05, t test; D2-control n = 13 mice; D2 n = 15 mice). C, Quantification of a detected vGAT punctum size (ns p > 0.05, t test).
Discussion
The above results indicate that TC neurons in the dLGN of glaucomatous mice show an increase in synaptic inhibition mediated, at least in part, through an upscaling of postsynaptic GABAA receptors. We found that disynaptic feedforward inhibition, evoked by optic tract stimulation, is diminished in D2 mice, although not as extensively as is expected if it simply reflected diminished drive from RG inputs. This points toward mechanisms that elevate inhibitory synaptic strength, perhaps compensating for diminished RG-to-LIN synaptic drive. Measuring quantal inhibitory currents (mIPSCs), we found that mIPSC amplitude was increased without detectable changes in frequency, indicating postsynaptic upscaling of inhibitory synapses without alterations in inhibitory synapse number or vesicle release. A peak-scaled nonstationary fluctuation analysis indicated an increase in postsynaptic GABAA receptor number without change in their single-channel conductance. An important caveat here is that dLGN TC neurons receive inhibitory input from multiple sources including LINs (of which there are two distinct populations; Leist et al., 2016), projections from the TRN (Kerschensteiner and Guido, 2017; Guido, 2018), and others, such as pretectum (S. Wang et al., 2002), meaning that recorded mIPSCs and the nonstationary fluctuation analysis reflect the average properties of inhibitory synapses from each of these sources. Our electrophysiology data were supported by staining showing enhanced labeling of the major GABAA-α1 in the D2 dLGN tissue. Finally, analysis approaches testing for synaptic scaling point to an ∼1.4-fold upscaling of inhibitory synapses, although those analyses revealed considerable nonuniformity in the extent of scaling across the inhibitory inputs to dLGN TC neurons, likely reflecting “divergent” scaling (Hanes et al., 2020; Koesters et al., 2024).
Homeostatic synaptic scaling is conventionally understood as a multiplicative process whereby all synapses of a type within a given neuron are scaled by uniform factor in response to altered activity (G. Turrigiano, 2012, G. G. Turrigiano, 2017; Koesters et al., 2024). However, under many circumstances, scaling can be nonuniform, with synapses being scaled by divergent factors (Hanes et al., 2020; Koesters et al., 2024). Several analysis approaches have been used in attempts to distinguish these possibilities. These include testing whether the cumulative distributions of event amplitudes match after up- or downscaling based on the linear fit parameters comparing rank-ordered control and experimental events (G. G. Turrigiano et al., 1998; Hanes et al., 2020). When we performed this analysis, we found that downscaling our D2 mIPSCs by a scaling factor of 1.45 produced a distribution that had considerable overlap with the D2-control distribution. However, statistical comparison indicated quantitatively poor overlap suggesting that the D2 mIPSC amplitudes do not reflect a simple global multiplicative scaling process. Moreover, the peak K-S test p value obtained using the iterative approach (Kim et al., 2012) only reached 0.0031 (at a scaling factor of 1.405), again indicating quantitatively poor overlap of the distributions. Finally, a plot of the ratio of rank-ordered D2/D2-control mIPSC amplitudes as a function of D2-control amplitudes showed a wide range of D2/D2-control ratios without clear convergence on a single value (Hanes et al., 2020; Koesters et al., 2024). Combined, these analyses point away from uniform multiplicative scaling of TC neuron GABA synapses. TC neurons receive the bulk of their inhibition from two sources: projections from the TRN and inputs from local interneurons, with LINs providing both dendrodendritic (F2 terminals) and axodendritic (F1 terminals) inputs to TC neurons (Bickford, 2019), with other inhibitory inputs from pretectum (S. Wang et al., 2002). mIPSCs recorded from TC neurons thus reflect the combined input from each of these sources, with analyses of mIPSC properties representing averages across each of those populations. The nonuniformity of scaling across the measured mIPSCs might reflect differential scaling at each of these synapse types or perhaps even differential scaling of synapses within a type. Whether this is the case, along with the underlying mechanism(s), remains to be determined.
At the ages used in this study (10–14 months), D2 mice have fairly minimal RGC somatic loss (Buckingham et al., 2008; J. C. Smith et al., 2022), although predegenerative processes such as RGC dendritic retraction and synaptic disassembly, diminished axoplasmic transport, and loss of RGC synaptic output in the brain are well underway (Crish et al., 2010; Calkins, 2012; Berry et al., 2015; Cooper et al., 2016; M. A. Smith et al., 2016, Risner et al., 2018; Van Hook et al., 2020; J. C. Smith et al., 2022). Within the dLGN, TC relay neurons also demonstrate clear structural and functional alterations including dendritic remodeling, somatic atrophy, altered intrinsic membrane properties, and enhanced action potential firing (Van Hook et al., 2020; J. C. Smith et al., 2022). Despite the diminished excitatory synaptic drive and altered intrinsic excitability, we have not seen evidence for excitatory synaptic scaling in D2 TC neurons (Van Hook et al., 2020; J. C. Smith et al., 2022). This might be tied to mouse age, as prior work has shown that silencing of RGC spiking or altered visual experience during a developmental critical period can alter excitatory glutamate receptors at retinogeniculate and corticothalamic excitatory synapses, respectively (Krahe and Guido, 2011; Louros et al., 2014).
While homeostatic synaptic scaling is thought to act on inhibitory and excitatory synapses in a coordinated fashion in response to altered activity levels, the triggers for scaling at each synapse type are different (Wenner, 2011; Gonzalez-Islas et al., 2023). This suggests that inhibitory synapses might show signs of synaptic scaling without parallel changes in excitation, depending on the trigger. As shown in studies using primary cultures of hippocampal and cortical pyramidal neurons, excitatory scaling is a cell-autonomous process whereby altered somatic Ca2+ concentration serves as an error signal linking perturbations in neuronal spiking behavior to activity of the Ca2+/calmodulin-dependent protein kinase IV (CaMKIV) that drives homeostatic trafficking of postsynaptic glutamate receptors (Goold and Nicoll, 2010; G. Turrigiano, 2012; Joseph and Turrigiano, 2017). Inhibitory synaptic scaling, on the other hand, operates independently of CaMKIV (Ibata et al., 2008; Goold and Nicoll, 2010; Joseph and Turrigiano, 2017), indicating it depends on another mechanism, possibly even independently of cell-autonomous spiking (Wenner, 2011). Of note, another work has suggested that altered synaptic transmission itself triggers excitatory scaling independently of spiking activity (Fong et al., 2015), while cell-autonomous spiking activity scales inhibitory synapses (Gonzalez-Islas et al., 2023), both in contrast to the “conventional” understanding and suggesting a diversity of mechanisms that might depend on cell type and methods used to modulate activity.
Our data do not resolve what specific facet of glaucomatous pathology and associated changes in dLGN function trigger GABAergic scaling in the D2 dLGN. Nor do our data currently answer whether this results from network-wide effects or instead from cell-autonomous or synapse-specific processes. Activity can clearly modulate inhibitory synaptic strength; in work by Kilman and colleagues, activity suppression via TTX led to downscaling of synaptic GABA receptor number in primary cortical cell cultures, although it is unclear whether that was cell-autonomous, as network effects of TTX would also reduce spike-driven synaptic activity (Kilman et al., 2002). Brain-derived neurotrophic factor (BDNF) is a noncell-autonomous trigger for homeostatic plasticity and has been shown to homeostatically influence inhibitory synaptic strength in hippocampal neuron primary cultures (Rutherford et al., 1998; Swanwick et al., 2006). BDNF transport along RGC axons is perturbed by increased eye pressure (Pease et al., 2000), and the resulting disruption in BDNF signaling might underlie the effects on dLGN inhibition we show here. Another potential non-cell-autonomous mechanism involves tumor necrosis factor α (TNFα), a cytokine released by glial cells in response to injury and altered neuronal activity levels and can regulate postsynaptic AMPA and GABA receptors (Stellwagen and Malenka, 2006; Pribiag and Stellwagen, 2013; Heir and Stellwagen, 2020; Heir et al., 2024). Glaucoma has been shown to lead to reactive glial responses in dLGN (Tribble et al., 2021), and we have shown previously that optic nerve injury triggers changes in microglia morphology consistent with reactive microgliosis (Bhandari et al., 2022). Release of TNFα in the glaucomatous dLGN, either from activated microglia or astrocytes, might be an important trigger of inhibitory synaptic scaling, but this possibility, along with whether it could have localized effects sufficient to trigger nonuniform scaling, remains to be tested.
Glaucoma is notorious for its relatively slow progression and asymptomatic presentation in early stages (Quigley and Broman, 2006; Tham et al., 2014; Weinreb et al., 2014). This has the potential of leading to late diagnosis, delaying treatment to a point where vision loss is irreversible. The GABAergic synaptic scaling we describe here might contribute to a vast array of adaptive processes that homeostatically preserve vision even as pathology progresses (Calkins, 2021). These include altered RGC excitability, changes in mitochondrial efficiency, glial responses, and changes in synaptic inputs and response properties of retinorecipient neuron populations in the brain (Calkins, 2021). We have shown previously that TC neuron intrinsic excitability is enhanced with increased IOP (Van Hook et al., 2020). Future work will need to explore the overall impact(s) of this altered intrinsic excitability in combination with lost retinogeniculate input (J. C. Smith et al., 2022) and potential for altered corticothalamic feedback excitation (Van Hook et al., 2020) on the overall ability of TC neurons to transform RGC input into spike output to the cortex. Further study is also needed to determine the cellular mechanisms of adaptation to glaucoma throughout the visual system and whether, when, and how the balance tips from sight-preserving adaptation to irreversible vision loss.
Footnotes
The authors declare no competing financial interests.
This work was supported by National Institutes of Health–National Eye Institute Grant R01EY030507 and a Career Advancement Award from Research to Prevent Blindness (RPB) and The Glaucoma Foundation to M.J.H. The funders had no role in study design, data collection, analysis, decision to publish, or manuscript preparation. We are grateful to Jennie Smith for performing intraocular pressure measurements.
- Received June 13, 2024.
- Accepted June 20, 2024.
- Copyright © 2024 Van Hook and McCool
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.