Impaired thalamic burst firing in fragile X syndrome

SUMMARY

The thalamus performs a critical role in sensory processing by gating the flow of sensory information to the neocortex and directing sensory-guided behaviors, functions that are disrupted in people with autism spectrum disorder (ASD). We have identified cellular changes in thalamic neurons in a mouse model of fragile X syndrome (FX) that alter how the thalamus transmits sensory information to neocortical circuits. In awake wild-type (WT) mice, thalamic neurons shift between two firing modes, burst and tonic, which may gate input to the neocortex. Thalamic neurons in FX mice, however, do not shift between these modes and instead fire primarily in the tonic mode. Voltage-clamp recordings reveal that macroscopic Ca²⁺ currents are significantly smaller at hyperpolarized membrane potentials in FX LGN neurons compared to those of the WT. In agreement with the voltage-clamp results, we demonstrate that the Ca²⁺-dependent low-threshold spike, which underlies normal bursting, is reduced or absent at membrane potentials at or around the resting membrane potential.

Graphical Abstract

In brief

Thalamic neurons relay specific information to the neocortex by switching between burst and tonic firing modes. Using in vivo and in vitro approaches, O’Shea et al. demonstrate that typical thalamic firing dynamics are disrupted in the lateral geniculate nucleus of fragile X mice due to a reduction in burst activity.

INTRODUCTION

In the face of complex sensory environments, animals must segment task-relevant features from the background to guide actions and avoid becoming overwhelmed by irrelevant noise. What constitutes target and background in the sensory landscape can vary rapidly, making flexible control over what sensory information gets processed by neocortical circuits crucial for guiding optimal behaviors.

People with ASD commonly struggle to detect, attend to, and interpret salient stimuli. This impairment is due to feeling overwhelmed by uninformative stimuli in “noisy” environments, hypersensitivity to normally innocuous features, and trouble redirecting attention when salient stimuli appear. Psychophysical studies demonstrated that these sensory abnormalities cannot be explained by changes in detection thresholds for low-level features.^1,2 Rather, there is a deficit in higher-order processing, apparent in tasks requiring the segmentation of task-relevant features from background.^2–7 This sort of hyposensitivity to behaviorally relevant stimuli in ASD stands in contrast with widely reported perceptual hypersensitivity to supra-threshold stimuli, which correlates with larger magnitudes of neocortical activation than in control groups, as measured by electroencephalogram.^8–10 These reports of both hypo- and hypersensitivities suggest that the relay of sensory information to the neocortex is dysregulated in ASD.

Given these sensory abnormalities, it is possible that typical relay of sensory information from the periphery to the neocortex, via the thalamus, is disrupted in ASD. While the thalamus is often considered a relay station, it also acts as a critical gate of information; the statistics of sensory inputs and modulatory feedback from the neocortex and brain stem alter how the thalamus conveys sensory information to the neocortex.^11–13 These inputs alter information transmission by shifting thalamic neurons between two modes of action potential (AP) firing: burst and tonic modes.¹⁴ Bursting is mediated by the generation of the low-threshold spike (LTS), a non-linear depolarization of membrane potential mediated by T-type Ca²⁺ channels, which are activated at membrane potentials near rest.^15–17 The combination of the statistics of the sensory input and top-down modulatory drive sets the membrane potential of a thalamic neuron, which determines whether it will fire in the burst or tonic mode in response to a stimulus.^18–20 One prominent hypothesis is that burst-mode firing acts to flag stimuli as novel and urgent, whereas tonic mode firing is associated with providing a faithful transmission of signals from the periphery.^13,21–25

The proposed role for thalamic firing dynamics in selectively gating sensory relay to the neocortex and the behavioral evidence of dysregulated sensory processing in ASD led us to compare thalamic population activity in wild-type (WT) and FMR1 knockout (fragile X syndrome [FX]) mice, a common experimental model for exploring the physiological basis of the autistic phenotype.²⁶ We found that burst firing in vivo is profoundly impaired in the FX thalamus, with these cells firing almost exclusively in the tonic mode, in stark contrast with the distribution of firing modes we recorded in the WT thalamus. To investigate the basis for impaired burst firing in FX at the cellular level, we performed in vitro whole-cell recordings in acute brain slices and found that the voltage sensitivity of the LTS was hyperpolarized in FX thalamic neurons, which contributes to reduced burst frequency in the FX thalamus.

RESULTS

Firing patterns in the LGN of awake WT and FX mice

To compare the thalamic firing patterns in WT and FX mice, we densely sampled single-unit activity in the lateral geniculate nucleus (LGN) of awake WT and FX mice with the Neuropixels 1.0 probe.²⁷ We classify spikes as emerging from burst or tonic firing mechanisms by the preceding and subsequent interspike interval (ISI): the first AP in a burst is preceded by at least 100 ms of silence, followed by APs separated by less than 4 ms, and ending with an ISI longer than 4 ms; all other spikes are classified as tonic (Figure 1A).²⁸ The pattern of tonic and burst spiking is consistent across WT cells (n = 13 mice, mean = 46.3 ± 29.5 cells/mouse) (Figures 1B and S1A). In contrast, cells in the FX LGN fired primarily in the tonic mode (n = 6 mice, mean = 35.2 ± 30.1 cells/mouse) (Figures 1C and S1A). The frequency of burst events was significantly lower in FX than WT populations during both spontaneous and visually evoked activity (WT spontaneous mean = 0.16 ± 0.25, median = 0.07 bursts/s; FX spontaneous mean = 0.01 ± 0.03, median = 0.004 bursts/s; p < 0.01 for a Wilcoxon rank-sum test; WT evoked mean = 0.09 ± 0.11, median = 0.04 bursts/s; FX evoked mean = 0.01 ± 0.03, median = 0.004 bursts/s; p < 0.01 for a Wilcoxon rank-sum test) (Figure 1D, left). Even when FX cells did burst, the bursts contained significantly fewer APs compared to bursts from WT cells (WT mean = 2.94 ± 1.02 spikes/burst, FX mean = 2.46 ± 0.81 spikes/burst; p < 0.01 for a Wilcoxon rank-sum test) (Figure 1F).

Figure 1. Firing patterns in LGN of awake WT and FX mice.

(A) Left: example voltage trace of a WT LGN cell exhibiting tonic (black) and burst (orange) spikes. Right: example ISI distribution of a WT LGN cell, where each scatterpoint represents pre and post ISI for a single spike. Green boxes indicate the criteria for spikes being classified as part of a burst.

(B) Example ISI distributions of WT LGN cells.

(C) Example ISI distributions of FX LGN cells.

(D) Left: distribution of burst rates for WT (black) and FX (red) populations (spont p = 1.1 × 10⁻⁵⁸, evoked p = 9.5 × 10⁻⁵⁰). Right: distributions of baseline firing rates (p = 3.3 × 10⁻²⁸). Circles are the medians of the distributions. Lower and upper bars indicate the 25^th and 75^th percentiles of the distributions, respectively. *p < .01 for a one-sided Wilcoxon rank-sum test between FX and WT population data.

(E) As in (A) for the resampled WT dataset.

(F) Histogram of the number of spikes in each burst across the WT and FX LGN populations (p = 3.54 × 10⁻¹⁹⁵).

WT cells also had higher baseline firing rates than FX cells (WT mean = 10.3 ± 10.4 spikes/s, median = 6.9; FX mean = 3.2 ± 3.3, median = 2.4 spikes/s) (Figure 1D, right; Figure S1B). This difference in baseline firing rate could underlie the difference in burst frequency we observe between the genotypes. To explore this possibility, we resampled our WT dataset to create a sample with firing rate statistics matched to the FX dataset (Figure 1E, right). The distribution of burst frequencies for the resampled WT population remained significantly higher than that of the FX population, indicating that the disruption of thalamic firing dynamics in FX cannot be explained simply by a difference in the baseline firing rate of neurons across genotypes (WT resampled spontaneous mean = 0.13 ± 0.19, median = 0.05 bursts/s; WT resampled evoked mean = 0.07 ± 0.10, median = 0.04 bursts/s) (Figure 1E, left).

The differences in firing patterns between WT and FX LGN neurons could be due to differences in the behavioral states of the animals. Locomotion is known to increase baseline firing rates in the mouse LGN.²⁹ Thus, we hypothesized that burst firing would be suppressed during bouts of locomotion, as LGN neurons are in a more depolarized state and therefore less likely to shift from the tonic to the burst mode. In addition, if FX mice exhibit more locomotion than WT mice in our experimental setting, then this behavioral difference could contribute to the reduction in burst frequency in FX LGN, which we observed.

To test these predictions, we tracked the movement of mice on a wheel while recording from the LGN population. Indeed, we found that burst firing was significantly suppressed during bouts of locomotion, for 47% of WT LGN neurons and 53% of FX LGN neurons (Figures S4A–S4C; STAR Methods). However, these differences in thalamic firing dynamics between WT and FX populations could also not be explained by differences in locomotion, as WT and FX mice ran equal amounts during recordings (Figure S4D).

Deviations from Poisson statistics in WT and FX LGN populations

The ISI distributions for FX LGN neurons show clear qualitative differences from those of WT LGN neurons, though some FX neurons do exhibit firing patterns that meet the criteria to be classified as bursts (Figure 1C, bottom right). To quantify thalamic firing dynamics of WT and FX populations, we examined the interactions between successive spikes. Spike trains modeled by a Poisson process are characterized by an exponential distribution of ISIs, nut neurons can deviate from this process in characteristic ways. After an AP, there is a refractory period during which a successive AP is less likely, and during a burst, there is a narrow window during which a successive AP is more likely. One way to quantify the degree of burstiness and refractoriness in a neuron is to measure the deviation of the ISI distribution from an exponential fit.³⁰ While refractoriness results in relatively few spikes with short ISIs compared to the Poisson prediction, bursting generates spikes with short ISIs with an increased probability of spiking relative to the Poisson expectation for a given mean firing rate (Figure 2A). We established a bursting weight for each cell by taking the ratio of the empirical ISI distribution to the rate-matched Poisson expectation within the 1- to 4-ms range (Figure 2B). A weight of greater than 1 indicates that a cell bursts more frequently than expected for a Poisson process. A weight of less than 1 indicates that a cell fires primarily in the tonic mode, with the empirical ISI distribution showing relatively few spikes separated by small ISIs compared to the Poisson prediction because of the absolute and relative refractory periods.

Figure 2. Deviations from Poisson statistics in WT and FX LGN populations.

(A) Example ISI probability density function (PDF) for a bursting cell. The black solid line is the empirical ISI PDF of the cell, and the gray dashed line is the ISI distribution of a Poisson process with a matched mean firing rate.

(B) Example ISI distributions for a WT (black) and FX (red) cell. Inset: bursting weights computed for each cell.

(C) Example ISI distributions for the same WT cells as in Figure 1B, with a gray dashed line indicating the rate-matched Poisson expectation. Inset: the bursting weight for each example cell.

(D) As in (C) for the same example FX cells as in Figure 1C.

(E) Histogram of log₂ bursting weights for all cells in the WT and FX populations. Arrows indicate means of distributions.

This metric captures the stark difference in thalamic firing patterns between WT and FX populations. WT LGN neurons exhibit more spikes separated by 1–4 ms than the Poisson prediction, indicated by bursting weights greater than 1 (WT mean bursting weight = 1.87; one-sample right-tailed t test p = 9.4 × 10⁻¹⁴) (Figures 2C and 2E). By contrast, burst spikes were infrequent in FX LGN neurons, indicated by bursting weights less than 1 (FX mean bursting weight = 0.47, one-sample left-tailed t test, p = 5.7 × 10⁻¹¹) (Figures 2D and 2E).

One potential mechanistic explanation for impaired bursting in FX LGN is that the function of voltage-gated T-type Ca²⁺ channels, which switch thalamic neurons to burst mode at hyperpolarized membrane potentials, is altered in FX.^15,31–33 To examine how blockade of T-type Ca²⁺ channels affects typical thalamic firing dynamics, we recorded from WT LGN populations before and after intraperitoneal injection of a T-type channel blocker, ethosuximide (250 mg/kg).³⁴ Ethosuximide significantly reduced the rate of burst firing in the WT LGN, consistent with prior results in vitro (WT pre-ethosuximide mean = 0.14 ± 0.19 burst/s; WT post-ethosuximide mean = 0.02 ± 0.04 burst/s, n = 3 mice, p < 0.001 one-sided Wilcoxon rank-sum test) (Figure S1A).³⁵ In contrast to the effects on WT LGN neurons, ethosuximide did not alter the already low degree of burst firing in FX LGN neurons (FX pre-ethosuximide mean = 0.08 ± 0.17 burst/s; FX post-ethosuximide mean = 0.07 ± 0.12 burst/s; p = 0.81 one-sided Wilcoxon rank-sum test).

Thalamic neurons preferentially fire in tonic mode in FX

To examine whether a disruption in the function of T-type Ca²⁺ channels contributes to altered bursting in the FX LGN, we performed in vitro electrophysiological recordings from LGN neurons in WT and FX mice. Both WT and FX neurons showed an increase in AP firing rate in response to depolarizing current injections (Figure 3A). WT neurons, however, fired fewer APs compared to FX neurons, particularly at the smaller current injections (Figure 3B; mixed-factor ANOVA, main effect of between genotype and amplitude: F (11, 96) = 4.817, p < 0.0001). At larger current injections, the firing rate increased as a linear function of current amplitude, in line with the observation from in vivo recordings that the tonic mode represents a linear relay of sensory inputs.¹⁴ To guard against exaggeration of statistical differences, we also compared the slope of the linear portion of the input-output curve. We found that the input-output (F-I) slope was significantly larger in FX LGN neurons compared to the WT (FX: 0.15 ± 0.02 Hz/pA; WT: 0.06 ± 0.02 Hz/pA; unpaired t test t = 2.542, df = 11, p = 0.0274; Figure 3C). When neurons were held at −60 mV (near the resting V_m), however, WT neurons reliably fired a burst of APs in response to small, brief current injections (Figure 3D). In the burst mode, these neurons integrated inputs in a nonlinear manner, as increasing the duration of the current step up to 3,000 ms did not elicit more spikes after the initial burst, consistent with the proposed gating function of burst spiking. In contrast, FX neurons fired in the tonic mode and increased the number of spikes fired with increasing current duration (Figures 3D and 3E; mixed-factor ANOVA, interaction between genotype and duration: F(5, 30) = 3.913, p = 0.007). We found that the input-output (F-D) slope was significantly larger in FX LGN neurons compared to the WT (FX: 0.01 ± 0.0032 Hz/ms; WT: 0.002 ± 0.001 Hz/ms; unpaired t test t = 2.458, df = 12, p = 0.0302; Figure 3F). At rest, WT thalamic neurons respond to excitatory drive in the burst mode, amplifying sensory drive following a period of inactivity, whereas FX thalamic neurons respond to same excitatory drive in tonic mode.^21,24 There were no differences in the passive or AP properties between WT and FX neurons, suggesting that the differences in burst firing were not likely due to Na⁺ or K⁺ channel-dependent mechanisms (Table S1).

Figure 3. Thalamic neurons preferentially fire in tonic mode in FX.

(A) AP firing recorded from a single thalamic neuron in response to current injection. A dashed box shows initial APs expanded on the right.

(B) FX thalamic neurons fire more APs compared to the WT.

(C) The slope of the F-I curve is significantly larger in FX thalamic neurons compared to the WT.

(D) AP firing in response to current injections of variable duration.

(E) FX neurons show a linear increase in firing frequency consistent with tonic firing mode, while WT thalamic neurons fire a brief train of 2–3 APs consistent with burst firing mode.

(F) The slope of the F-D curve is significantly larger in FX thalamic neurons compared to the WT. n(cells/mice): WT = 7/5, FX = 6/5.

Data are represented as mean ± SEM.

Burst firing in thalamic neurons is mediated in part by Ca²⁺ currents. To test whether there are differences in Ca²⁺ currents between WT and FX LGN neurons, we made whole-cell voltage-clamp recordings of pharmacologically isolated Ca²⁺ currents. We found that Ca²⁺ currents were significantly smaller in FX LGN neurons at membrane potentials between −60 and −40 mV (Figures S5A and S5B; mixed-factor ANOVA, interaction between genotype and V_M: F(11,92) = 2.602, p = 0.0063). Although most studies describe changes in dendritic spines and dendritic branching in FX to account for any potential differences in cell size, we normalized the Ca²⁺ current by the whole-cell capacitance measured in voltage clamping.³⁶ In agreement with the Ca²⁺ current data above, the Ca²⁺ current density was significantly smaller in FX LGN neurons at membrane potentials between −60 and −30 mV (Figure S5C; mixed-factor ANOVA, interaction between genotype and V_M: F(11,110) = 2.917, p = 0.0021). In agreement with the in vivo results, these data suggest that low-threshold-activated (T-type) Ca²⁺ channels, which are activated in the voltage range between −60 and −40 mV and generate the LTS, may be altered in FX LGN neurons.

LTS generation is hyperpolarized in FX thalamic neurons

The lack of bursts in FX cells at rest could be due to an absence of the LTS altogether or a change in its voltage-dependence. We delivered a depolarizing current step in the presence of 1 μM tetrodotoxin (TTX) to block Na⁺-dependent APs from a series of membrane potentials and measured the LTS. We found that the LTS was generated in FX cells, but only at a more hyperpolarized V_M. While the LTS in WT cells was apparent at −60 mV, the LTS in FX cells only began to emerge at −70 mV (Figures 4A and 4B; mixed-factor ANOVA, interaction between genotype and V_M: F(2,20) = 8.958, p = 0.0017). Even when the LTS was evoked, the amplitude was significantly smaller in FX neurons compared to the WT at both −60 mV (FX mean = 1.2 ± 1.98 mV, WT mean = 13.7 ± 7.32 mV) and −70 mV (FX mean = 12.6 ± 3.8 mV, WT mean = 19.3 ± 3.4 mV) (Figure 4B). The contribution of T-type Ca²⁺ channels to the LTS was confirmed by application of 50 μM Ni²⁺, a blocker of the T-type Ca²⁺ channels, to the extracellular saline (Figure 4A).³⁷

Figure 4. LTS generation is hyperpolarized in FX thalamic neurons.

(A) Representative traces showing a burst of AP, isolation of the LTS by TTX, and block of the LTS by Ni²⁺.

(B) Summary plot showing that LTS initiation is hyperpolarized in FX thalamic neurons. n(cells/mice): WT = 6/4, FX = 6/5. *p < 0.05, ***p < 0.001.

(C) Example traces showing a rebound burst of APs following hyperpolarization, isolation of the LTS, and block of the LTS in WT and FX neurons.

(D and E) Individual WT and FX neurons showing raw (D) and normalized (E) LTS amplitude as a function of steady-state V_M.

(F) Midpoint of LTS generation is hyperpolarized in FX compared to WT thalamic neurons.

(G) LTS amplitude between FX and WT thalamic neurons. n(cells/mice): WT = 8/4, FX = 7/5.

Data are represented as mean ± SEM.

To examine in detail the altered voltage dependence of the LTS in FX mice, we measured the LTS that followed a series of hyperpolarizing current steps from a holding potential of −55 mV (Figures 4C–4E). In agreement with the experiment above, the voltage sensitivity of the rebound LTS was hyperpolarized in FX thalamic neurons compared to the WT (Figure 4F; unpaired t test, t = 4.880, df = 13, p = 0.0003). Similar to the LTS measured by depolarizing current steps (Figures 4A and 4B), the contribution of T-type Ca²⁺ channels to the LTS following hyperpolarization was confirmed by application of 50 μM Ni²⁺ (Figure 4C).³⁷

DISCUSSION

We found that thalamic firing dynamics are profoundly disrupted in FX mice. In vivo recordings from awake mice revealed that, while thalamic neurons regularly transition between burst and tonic firing modes in the WT LGN, neurons in the FX LGN fire almost exclusively in the tonic mode. Furthermore, when FX neurons do fire bursts, they contain fewer APs than those from WT neurons. Using whole-cell recordings in acute slices, we found that FX neurons continue to respond to depolarizing inputs in the tonic mode at membrane potentials that evoke bursts in WT neurons. FX neurons are capable of generating burst APs, but only at more hyperpolarized membrane potentials. This change in firing pattern can be explained at least in part by hyperpolarization of the voltage sensitivity of the LTS, which contributes to reduced burst frequency and duration in FX neurons. These in vitro results provide a mechanistic basis for the dramatic reduction in burst frequency and duration observed in the LGN of FX mice in vivo. While the LTS remains in FX neurons, it requires greater membrane potential hyperpolarization for activation, which may only rarely occur in vivo. A smaller LTS when FX cells do transition to burst mode means less activation of voltage-dependent Na⁺ channels for driving APs, which will reduce the number of spikes in a burst.

There are additional mechanisms that may account for the functional disruption of bursts in the FX LGN. Modulatory inputs to the LGN from brain stem cholinergic projections or thalamic reticular nucleus GABAergic projections could be altered in FX.^12,38 Properties of the retinal drive to the LGN, which can evoke bursting, are also altered in FX mice.^39,40 These explanations are not mutually exclusive, as disruptions at both the network and cellular levels could contribute to the observed changes in thalamic firing dynamics in vivo as well as the changes in mean firing rates. In particular, our observation of a lower mean firing rate in FX neurons, despite these neurons being hyperexcitable in vitro in response to current steps, may be due to the reduced activity of afferent drive from the retina.^39,40

We observed considerable variability of in vivo burst rates across WT mice, in contrast to the FX population, for which bursting rates are nearly zero for all mice in our experimental sample (Figure S2A). We considered how both behavioral state and visual stimulus properties might contribute to the inter-subject variability in bursting rates.

While we have shown that locomotion suppresses burst firing in both WT and FX LGNs, differences in locomotion patterns across subjects did not contribute to the observed inter-subject variability in bursting rates. There was no correlation between average speed and average spontaneous bursting rates across experiments for both WT and FX cohorts (Pearson’s correlation coefficient = −0.002) (Figure S6A). Importantly, while locomotion strongly modulates LGN firing mode, bouts of running are quite rare relative to time spent stationary. Across experiments, mice spent most of the time at rest, with only sparse bouts of locomotion (WT mean = 1.9 ± 1.7% running, FX mean = 1.1 ± .9% running). Thus, while locomotion does shift LGN firing mode on a rapid timescale correlated to behavior, over the long duration of experimental time (30–60 min), small differences in the amount of running between subjects cannot account for variability in average spontaneous bursting rates. However, there are additional measures of behavioral state that we did not track but that are correlated with changes in thalamic activity. Measuring pupil diameter and bodily movements other than running could grant additional insights regarding variability in behavioral state of the animal both over the course of an experiment and across subjects.^41–43

We found that diverse features of visual stimuli did contribute to the inter-subject variability in burst rates. In particular, we find that natural movies evoke higher bursting rates than drifting gratings, in keeping with prior results from the cat LGN.¹⁹ For the same WT LGN neurons tracked across visual stimuli, evoked bursting rates were higher on average during presentation of natural movies as compared to drifting gratings (Figure S3A). In keeping with this observation, WT and FX mice that viewed both natural movies and gratings exhibited higher evoked bursting rates than those that viewed only gratings (Figure S3B). These findings indicate that some inter-subject variability in burst rate across the experimental population may be explained by differences in the properties of visual stimuli.

We used FX mice as a model for neurodevelopmental disruptions in ASD. FMR1 is conserved across species, and its loss in humans causes a severe form of ASD, FX. Recording in the FX thalamus allowed us to probe the neural mechanisms contributing to the perceptual impairments commonly reported in ASD. We propose that a change in the pattern of firing modes in the FX thalamus could disrupt sensory processing in neocortical regions by altering the normal patterns of thalamocortical relay. In the face of complex sensory environments, appropriate segmentation of task-relevant features from the background is essential to guide actions and avoid overwhelming the capacity of neocortical processing with irrelevant noise. What constitutes target and background in the sensory landscape can vary rapidly, making flexible control over what sensory information gets processed by neocortical circuits crucial for guiding optimal behaviors. A disruption of the neural mechanisms for dynamic gating of sensory relay could result in misallocation of neocortical processing resources to task-irrelevant stimuli as well as failure to adequately detect, attend to, and interpret salient inputs. These predicted impairments are in line with perceptual deficits commonly seen in people with ASD. Across sensory modalities, ASD subjects display a generally reduced capacity to identify targets in noise or to direct attention to stimuli that have high salience for their neurotypical counterparts.^2–7

Modulation of thalamic firing mode may control the selective gating of sensory information to the cortex as a function of both the “bottom-up” salience of peripheral inputs and “top-down” segmentation of task-relevant sensory information.^14,22,24 Experimental evidence supports the notion that the two firing modes play distinct roles in thalamocortical relay.^{13,18–20,24,25} In addition, inhibitory inputs from the thalamic reticular nucleus to first-order thalamic nuclei can alter the probability of bursting.¹² These findings implicate the dual firing modes as central to the dynamics of thalamocortical relay, with the capacity to be controlled by both the bottom-up salience of inputs and top-down, behaviorally relevant factors.

Changes in thalamocortical signaling in FX were previously described for nonvisual areas of the cortex. Recordings from L2/3 neurons in the barrel cortex found that thalamus-driven cortical UP states were prolonged in FX mice.⁴⁴ Further, the critical period of developmental plasticity in the barrel cortex was disrupted in FX mice, with the fraction of silent synapses persisting longer into development.⁴⁵ Additionally, changes in intracortical connectivity between excitatory and inhibitory neurons in layer 4 of the FX barrel cortex results in hyperexcitability to thalamocortical inputs.⁴⁶ These prior studies suggest that a combination of synaptic and intrinsic changes contribute to impaired thalamocortical signaling in FX. Our results in this study now extend these observations to describe how changes in LTS generation in the thalamus may contribute to the altered sensitivity to sensory stimuli in FX syndrome.

If the dynamics of thalamocortical relay do indeed play the hypothesized role in sensory processing, then rescuing normal thalamic firing patterns in FX, and potentially other forms of ASD, could allow people with ASD to better function in complex environments. Our results show that a disruption in the voltage sensitivity of the LTS contributes to reduced burst frequency and burst duration in the FX thalamus, which guides a potential therapeutic goal of rescuing normal T-type Ca²⁺ channel function and expression. Previous in vitro studies demonstrated that rescuing fragile X messenger ribonucleoprotein (FMRP) expression in FX hippocampal and prefrontal neurons rescues HCN channel and potassium channel function.^47,48 Future work should explore whether rescuing FMRP expression in FX thalamic neurons restores the normal voltage sensitivity of LTS generation. Ultimately, showing that reduced thalamic bursting contributes to perceptual impairments and tying gain of function for voltage-sensitive ion channels to rescuing normal thalamic firing dynamics could have therapeutic potential for people with ASD experiencing altered sensitivity to complex sensory environments.

Limitations of the study

During in vivo experiments, the only behavioral variable we measured was locomotion. While we observed no difference in locomotion between WT and FX mice, it is unknown whether differences in other behavioral variables, such as eye and limb movements, whisking, or arousal state, contribute to the observed differences in thalamic firing dynamics. We identified a particular physiological disruption in thalamic neurons that contributes to reduced burst firing in FX mice, though additional disruptions of the retinothalamic and corticothalamic drive may also contribute to abnormal thalamic firing dynamics in FX. We found that the LTS that underlies burst firing is smaller or absent near the resting potential of FX LGN neurons. While we are confident that this difference alters the firing dynamics of thalamic neurons in FX mice, the data do not provide a biophysical mechanism for this difference. Although we found no difference in the calcium current measured at hyperpolarized potentials, data on potential differences in voltage-gated calcium channel function and expression should be investigated in the future.

RESOURCE AVAILABILITY

Lead contact

Requests for further information and resources or reagents should be directed to the lead contact, Darrin Brager (darrin.brager@unlv.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• All original data have been deposited at Figshare at https://doi.org/10.6084/m9.figshare.c.7975940 and are publicly available as of the date of publication.

• All original analysis code has been deposited at Figshare at https://doi.org/10.6084/m9.figshare.c.7975940 and is publicly available as of the date of publication.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Electrophysiology experiments were conducted using adult C57/BL6J (Jackson Labs #000664) and Fmr1 KO (Jackson Labs #003025) mice of both sexes housed on a 12-hour-light-dark cycle. All mice were 6 weeks of age or older. All animal procedures were approved by the University of Texas at Austin and University of Nevada, Las Vegas Institutional Animal Care and Use Committee which maintain Association for Assessment and Accreditation of Laboratory Animal Care accreditation.

METHOD DETAILS

Surgery

For all surgical procedures, mice were anesthetized with isoflurane (2.5% induction, 1%–1.5% surgery) and given two preoperative subcutaneous injections of analgesia (5 mg/kg carprofen, 3.5 mg/kg Ethiqa) and anti-inflammatory agent (dexamethasone, 1%). Mice were kept warm with a water-regulated pump pad. Each mouse underwent two surgical procedures. The first was the mounting of a metal frame over the visual cortex using dental acrylic, used for head fixation of the mouse during experiments. The second was a 1–2mm diameter craniotomy centered 2.5mm posterior of bregma and 2mm lateral of midline in the right hemisphere followed by a durotomy. Exposed brain tissue was preserved prior to experiments with KWIK-SIL silicone adhesive (World Precision Instruments). Surgical procedures were always performed on a separate day from electrophysiological recordings.

For experiments in which we measured the effect of ethosuximide on thalamic firing dynamics in vivo, ethosuximide (Sigma-Aldrich) via intraperitoneal injection (250 mg/kg) 30 min prior to recording from LGN populations.

In vivo electrophysiology

During all experiments, mice were awake, head restrained and moved freely on a custom-built wheel. Mice were habituated to handling and head restraint over 1 week following recovery from the craniotomy in sessions increasing from 15 min to 1 h. The axle of the wheel was connected to a quadrature encoder to track locomotion during the experiment.

LGN recordings were made using a Neuropixels 1.0 probe implanted acutely 2500 μm posterior of bregma and 2000 μm lateral of midline and lowered to 2500–3200 μm depth. The electrode tip was lowered to the ventral extent of the LGN as identified by clearly evoked visual responses in the neural activity. Raw traces were acquired at 30 kHz using the spikeGLX System.

Visual stimuli

A monochrome LED projector by Texas Instruments (Keynote Photonics) with spectral peak at 525 nm was used to generate stimuli with a 60Hz refresh rate onto a Teflon screen which provides a near-Lambertian surface.⁴⁹ The screen was 12.5 cm high x 32 cm wide, equating to approximately 64 ° × 116 ° of visual angle. Stimuli were coded using the Psychophysics Toolbox extension in MATLAB.

Drifting gratings were presented at a spatial frequency of .02 cycles/degree, and a temporal frequency of 2 Hz, drifting at 0, 45, 90, 135°, for 50 repeats of each stimulus. Each grating was preceded and followed by 1 s of a blank screen at uniform luminance. 50 repeats of 10 s natural movies were also presented with each repeat followed by a 6 s blank screen at uniform luminance. Movie 1 consisted of honeybees flying in a garden (Ian Nauhaus, UT Austin) and Movie 2 consisted of monkeys playing in snow (David Leopold, NIMH).

Acute slice preparation

Brain slices containing the visual thalamus were prepared and maintained as in prior studies.^50,51 Briefly, mice had free access to food and water and were housed in a reverse light-dark cycle of 12 h on/12 h off. Experiments used male 10–16 week-old wild type and fmr1 knockout (FX) mice on a C57/Bl6 background (JAX: strain #000664). Mice were anesthetized using a ketamine/xylazine cocktail (100/10 mg/kg) and then underwent cardiac perfusions with ice-cold saline consisting of (in mM): 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 0.5 CaCl₂, 7 MgCl₂, 7 dextrose, 205 sucrose, 1.3 ascorbic acid, and 3 sodium pyruvate (bubbled constantly with 95% O₂/5% CO₂ to maintain pH at ∼7.4). The brain was removed and sliced into 300 μM parasagittal sections using a vibrating tissue slicer (Vibratome 300, Vibratome Inc). The slices were placed in a chamber filled with artificial cerebral spinal fluid (aCSF) consisting of (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 2 CaCl₂, 2 MgCl₂, 10 dextrose, 1.3 ascorbic acid and 3 sodium pyruvate (bubbled constantly with 95% O₂/5% CO₂) for 30 min at 35°C and then held at room temperature until time of recording.

In vitro electrophysiology

Slices were placed individually, as needed, into a submerged recording chamber, continuously perfused with oxygenated extracellular saline containing (in mM): 125 NaCl, 3 KCl, 1.25 NaH₂PO4, 25 NaHCO₃, 2.0 CaCl₂, 1.0 MgCl₂, and 21 dextrose (pH 7.4) at 32°C–34°C. LGN cells were viewed with a 60x water immersion objective and Dodt contrast. Ionotropic glutamatergic and GABAergic synaptic transmission were blocked with 20 μM DNQX, 25 μM D-AP5, and 2 μM gabazine (Hello Bio).

Current clamp recordings were made using a Dagan BVC-700 amplifier and SutterPatch acquisition software (Sutter). Data were sampled at 20–50 kHz, filtered at 3 kHz, and then digitized with a Dendrite interface (Sutter). The internal recording solution consisted of (in mM): 120 K-gluconate, 16 KCl, 10 HEPES, 8 NaCl, 7 K₂-phosphocreatine, 0.3 Na-GTP, 4 Mg-ATP (pH corrected to 7.3 with KOH). Recording pipettes had a resistance of 4–6 MΩ when filled with the pipette solution. During current clamp recordings, series resistance was monitored throughout, and the experiment was discarded if it exceeded 30 MΩ or varied by >20%. The LTS was isolated by blocking Na⁺-dependent APs with TTX (1 μM) during whole-cell recordings. Following LTS recordings, NiCl₂ (50 μM) was applied to confirm that the measured LTS was mediated by T-type Ca²⁺channels.³⁷

Voltage-clamp recordings were made using an Axopatch 200B amplifier (Molecular Devices) and Sutterpatch acquisition software (Sutter). Data were sampled at 20–50 kHz, filtered at 3 kHz, and then digitized by a Dendrite interface (Sutter). The extracellular saline was as described above with in addition of 10 mM TEA-Cl, 5 mM CsCl, and 1 μM TTX. The internal recording solution consisted of (in mM): 120 Cs-gluconate, 10 HEPES, 10 CsCl, 10 EGTA, 4 NaCl, 3 TEA-Cl, 0.1 3,4 diaminopyridine, 7 K₂-phosphocreatine, 0.3 Na-GTP, 4 Mg-ATP (pH corrected to 7.3 with KOH). Recording pipettes had a resistance of 4–6 MΩ when filled with the pipette solution. Whole-cell Ca²⁺ currents were measured using a series of 300 ms depolarizing voltage steps a holding potential of −90 mV. Linear leakage and capacitive currents were digitally subtracted by scaling traces at smaller command voltages (P/4) in which no voltage-dependent current was activated.

QUANTIFICATION AND STATISTICAL ANALYSIS

In vivo electrophysiology

Putative single units were isolated from the raw traces offline using Kilosort3.⁵² We conducted further manual sorting of single units using the Phy GUI (https://github.com/cortex-lab/phy). We conducted further manual sorting of single units using the Phy GUI (https://github.com/cortex-lab/phy). Putative single units were accepted for further analysis if they exhibited less than 10% of 1ms refractory period violations (‘ContamPct’ in Phy) and had a consistent waveform throughout the experiment. Finally, single units were only included if they showed a significant visually evoked response, to at least one visual stimulus, above the spontaneous firing rate as quantified by a one-sided, two-sample t test with p < .01.

All further analysis of single unit data was performed using custom MATLAB code. Each spike from a single unit was defined as arising from the burst or tonic firing mode based on the pre and post ISI. The first AP in a burst is preceded by at least 100ms of silence, followed by APs separated by less than 4ms. A burst ends with an ISI greater than 4ms. In our data, as reported previously, the end of a burst tended to be followed by a period of at least 100ms of silence.²⁸ All spikes not meeting these criteria for bursting were classified as tonic spikes.

Bouts of locomotion were defined as timepoints in which the first derivative of the wheel position differed from zero for at least 500ms. At all other timepoints the mouse was considered to be stationary. All burst events could be classified as occurring during either stationary or running epochs. To quantify the degree of clustering of burst events in epochs of running versus rest, we employed a bootstrap t test that compared the burst time points for each single unit to a null distribution in which an equal number of bursts are randomly distributed in time throughout the experiment. For each LGN single unit, we used the MATLAB function rand-sample to shuffle the timing of all burst events. We then classify each shuffled burst event as coinciding with stationary or running epochs. We repeated this process 100 times for each single unit to build a null distribution of the degree to which burst events are clustered in stationary versus running epochs. Significant suppression of bursting during locomotion as compared to chance was defined as a p value less than .05 for the bootstrap t test, meaning that less than 5% of the shuffled iterations yielded a higher degree of clustering of burst events to stationary epochs than the actual data. Similarly, significant clustering of bursts to periods of locomotion was defined as a p value greater than .95.

In vitro electrophysiology

Input resistance was calculated from the linear portion of the voltage-current relationship in response to a family of 1.5 s current injections of −50 pA–50 pA in steps of 10 pA. Voltage sag responses were calculated from the ratio of maximum voltage deflection and steady-state voltage from the 40 pA current step. Membrane time constant was calculated using the 10 pA current step and was determined as the slow component of a double exponential fit to the voltage decay. Action potential firing frequency was calculated from the number of action potentials during a family of 1 s depolarizing current steps from 0 pA to 250 pA in steps of 25 pA. Action potential threshold was determined as the voltage where the first derivative exceeded 20 mV/ms. The maximum and minimum rate of action potential rise were determined using phase plane analysis. All the statistical tests (unpaired t tests, ANOVA, and repeated measures of ANOVA) were performed using Prism (GraphPad). All data are shown as the means ± standard error of the mean (SEM).

Supplementary Material

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.116309.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Experimental models: Organisms/strains
Mouse: C57BL/6J	Jackson Laboratories	000664
Mouse: B6.129P2-Fmr1tm1Cgr/J (Fmr1 KO)	Jackson Laboratories	003025
Software and algorithms
Sutter Patch	Sutter	https://www.sutter.com/amplifiers/sutterpatch
MATLAB	Mathworks Inc	https://www.mathworks.com/products/matlab.html
Kilosort3	Pachitariu, Sridhar, Pennington, & Stringer	https://github.com/MouseLand/Kilosort
Phy	Cyrille Rossant	https://github.com/cortex-lab/phy
Prism	Graph Pad	https://www.graphpad.com/features
Deposited Data
Electrophysiology data and code	This paper	https://doi.org/10.6084/m9.figshare.c.7975940

Highlights.

• Burst firing in vivo is reduced, and bursts contain fewer spikes in FX LGN neurons

• FX LGN neurons in vitro fire predominately in tonic mode

• Calcium currents at hyperpolarized membrane potential are smaller in FX LGN neurons

• Ca²⁺-dependent, low-threshold spike generation is hyperpolarized in FX LGN neurons