Cellular and synaptic phenotypes lead to disrupted information processing in Fmr1-KO mouse layer 4 barrel cortex

Abstract

Sensory hypersensitivity is a common and debilitating feature of neurodevelopmental disorders such as Fragile X Syndrome (FXS). How developmental changes in neuronal function culminate in network dysfunction that underlies sensory hypersensitivities is unknown. By systematically studying cellular and synaptic properties of layer 4 neurons combined with cellular and network simulations, we explored how the array of phenotypes in Fmr1-knockout (KO) mice produce circuit pathology during development. We show that many of the cellular and synaptic pathologies in Fmr1-KO mice are antagonistic, mitigating circuit dysfunction, and hence may be compensatory to the primary pathology. Overall, the layer 4 network in the Fmr1-KO exhibits significant alterations in spike output in response to thalamocortical input and distorted sensory encoding. This developmental loss of layer 4 sensory encoding precision would contribute to subsequent developmental alterations in layer 4-to-layer 2/3 connectivity and plasticity observed in Fmr1-KO mice, and circuit dysfunction underlying sensory hypersensitivity.

Fig. 1

Altered intrinsic properties of Fmr1-KO Layer 4 SCs and FS interneurons. a Left: Example firing characteristics of layer 4 excitatory neurons in response to 500 ms hyper/depolarizing current injections (−40 pA, rheobase, 2× rheobase shown). Right: Passive membrane and intrinsic properties of L4 excitatory neurons: Input resistance (Fmr1^+/Y: 412 ± 33 MΩ, Fmr1^−/Y: 609 ± 43 MΩ, p = 0.0007), Membrane time constant (Fmr1^+/Y: 39 ± 3.7 ms, Fmr1^−/Y: 55 ± 3.2 ms, p = 0.0015), Rheobase current (p < 0.002, Fmr1^+/Y: 72 ± 6.6pA; N = 33, Fmr1^−/Y: 44 ± 4.0 pA; N = 37). Not shown: Membrane capacitance (Fmr1^+/Y: 94 ± 6.7 pF, Fmr1^−/Y: 89 ± 4.9 pF, p = 0.56), resting membrane potential (Fmr1^+/Y: −64 ± 1.7 mV, Fmr1^−/Y: −64 ± 1.3 mV, p = 0.88). All statistics herein: Student’s t-test, two-tailed. b Suprathreshold current-spike frequency (FI) responses of L4 Fmr1^−/Y excitatory neurons were significantly steeper for current injections > 30 pA (p = 0.02, Mann–Whitney, Fmr1^+/Y: 120 ± 10 Hz/nA, Fmr1^−/Y: 200 ± 12 Hz/nA (n: 28 Fmr1^+/Y, 28 Fmr1^−/Y). c SC firing rate during twice-rheobase current injections. Asterisks: p < 0.05, t-tests comparing values for each spike position in train, N: Fmr1^+/Y = 50 neurons, Fmr1^−/Y = 42 neurons. d SC action potential half-width and amplitude during entrained firing to twice-rheobase current injections. Statistics as (c). e Left: Example spike waveforms fired by FS interneurons in response to 500 ms depolarizing current injections (rheobase, 2× rheobase shown). Right: passive membrane and intrinsic properties of FS interneurons (Asterisks: p < 0.05, t-test, N (neurons): Fmr1^+/Y = 15, Fmr1^−/Y = 23). f–h FS overall action potential rate (f), firing rate accommodation (g) and amplitude accommodation (h) during entrained firing to twice-rheobase current injections. Statistics as in (e). i Reduced rate of AP firing in Fmr1^−/Y FS interneurons as analysed by instantaneous frequency (1/inter-spike interval) of first and last two APs in train Light colours indicate individual neurons, thick bars/lines are mean ± SEM for each genotype. Asterisks on left and right graphs indicate parameters significantly different between genotypes compared by t-test (p < 0.05, N (neurons): Fmr1^+/Y = 15, Fmr1^−/Y = 27). Asterisks on centre graph indicate significantly different mean frequencies between genotypes compared by one-way ANOVA with Bonferroni’s correction for multiple comparisons (p < 0.05, N: Fmr1^+/Y = 15, Fmr1^−/Y = 27)

Fig. 2

Reduced FS IN–SC connectivity in Fmr1-KO, but no change in connection properties. a Monosynaptic connection probability between pairs of FS and SCs tested with paired whole-cell recording. Asterisks: p < 0.05, Chi-squared test. b, c Connection strengths (b), peak evoked monosynaptic current amplitude and current onset latencies (c) for connected pairs shown in (a). No differences in connection strength between genotype were observed for either connection direction (p > 0.05, Mann–Whitney, N’s as in (a)). d Monosynaptic connection probability between pairs of SCs located within the same barrel for recordings in slices taken from P10–11 mice of each genotype. No change in connection probability was observed (p > 0.05, Chi-squared test). Not shown: Significant reduction in SC–SC connectivity for Fmr1^−/Y at P12–15 (37/110, 8/54 tested pairs connected for Fmr1^+/Y and Fmr1^−/Y, respectively, χ2(1) = 7.78, p < 0.01). e, f No change in synaptic strengths (e) or peak monosynaptic EPSC amplitude (f) between the connected SC pairs from P10–11 Fmr1^+/Y and Fmr1^−/Y littermate mice shown in (d) (p > 0.05, Mann–Whitney, N’s as in (a))

Fig. 3

Altered thalamocortical FFI in P10–11 Fmr1-KO mouse. a Left: Trial-averaged voltage-clamp recordings showing direct thalamocortical EPSCs and FF-IPSCs from an example Fmr1^+/Y Layer 4 SC, Scale: 25 ms/500 pA. The strength of TC-FFI (“G/A ratio”) is quantified as the ratio of peak evoked current. Right: Strength of TC-FFI at P10–11. Points indicate neurons, bars are mean ± SEM. Unlike for Fmr1^+/Y recordings (28 neurons, max. 3 per animal), in Fmr1^−/Y neurons, some cells (8/38 neurons, max. 3 per animal) lacked FFI; light red bars indicate FFI strength of cells with G/A > 0, hollow markers. Including Fmr1^−/Y neurons with G/A = 0, average strength of FFI was not significantly different to that of Fmr1^+/Y, but excluding these neurons (dark red bars and solid markers), the average strength was elevated over that of FFI in wild-type recordings (Fmr1^+/Y vs. all Fmr1^−/Y neurons: p = 0.40, Fmr1^+/Y vs. Fmr1^−/Y neurons with G/A > 0: p = 0.005, Mann–Whitney). b Synaptic kinetics of currents underlying TC-FFI. Left: Example TC-evoked currents in from Fmr1^+/Y (blue) and Fmr1^−/Y (red), EPSCs and FF-IPSCs, individually scaled to peak amplitudes. Note the slower decay time constant and onset latency for Fmr1^−/Y FFI-PSCs (indicated by red and blue arrows). Right: Slower FFI synaptic kinetics for Fmr1^−/Y SCs. No significant genotype-dependent differences were observed in the same comparisons for kinetics of EPSCs. Asterisks indicate p < 0.05 (t-test), N’s (neurons): 23 Fmr1^+/Y/26 Fmr1^−/Y (EPSCs); 27 Fmr1^+/Y/28 Fmr1^−/Y (FF-IPSCs). c Example TC EPSPs from SCs receiving low, medium and high TC-FFI. Note the progressive curtailment of EPSP duration with increasing FFI strength, the lack of a Fmr1^+/Y example for G/A = 0, and the exaggerated IPSP for the high strength FFI Fmr1^−/Y example. d Slower TC EPSPs in Fmr1^−/Y. Full width at half-height (‘half-width’) of EPSPs from SCs (N (neurons) = 19 Fmr1^+/Y, 36 Fmr1^−/Y), bars are mean ± SEM, asterisks denote p < 0.05 (t-test). Not shown: the dependence of Fmr1-KO thalamocortical EPSP duration on the strength of FFI is distorted. Fmr1-KO Neurons with weaker/no FFI (G/A ratio < 2) showed specifically broadened EPSP duration (22.4 ± 7.30 ms vs. 68.4 ± 16 ms, Fmr1^+/Y vs. Fmr1^−/Y, p = 0.02), whereas this effect was dampened in neurons with stronger FFI (13.2 ± 8.70 ms vs. 37.1 ± 15.8 ms, Fmr1^+/Y vs. Fmr1^−/Y, p = 0.07)

Fig. 4

Synapse-specific changes to short-term plasticity in P10–11 Fmr1-KO mouse. a Top: Example voltage-clamped synaptic currents during repetitive TC stimulation at 50 Hz. Note strong amplitude attenuation of TC-evoked and FFI currents in Fmr1^−/Y. Bottom: Instantaneous G/A ratios for the above traces calculated by dividing FF-IPSC amplitudes by EPSCs for each sampled point in time. Note: (1) the graded attenuation of G/A ratio in the Fmr1^+/Y and slow onset of G/A balance during stimulus train despite large amplitude FF-IPSC, (2) Temporally disorganised ratio in the Fmr1^−/Y. Arrows show stimulation times. Scale: 50 ms/100 pA, G/A = 1. b Short-term depression of EPSCs and FF-IPSCs during 5 × 50 Hz stimulation: Evoked current amplitudes normalized to steady-state amplitude. Error bars: mean ± SEM normalized peak amplitude after for each stimulus for Fmr1^+/Y (blue, N = 19 (EPSCs) and N = 11 (FF-IPSCs) neurons) and Fmr1^−/Y (red, N = 15 (EPSCs) and N = 12 (FF-IPSCs) neurons). Asterisks: significantly different stimulus responses between genotypes (t-test, p < 0.05). Shaded regions are best ± 95% CI fits to bi-exponential decay functions. For both EPSCs and FF-IPSCs, the rate of depression for Fmr1^−/Y responses was faster and a single (i.e. common) fit could not adequately explain the behaviour of both genotypes (Extra sum-of-squares F-test, EPSCs: p = 0.0007, F(2,163) = 7.65, IPSCs: p = 0.0002, F(2,111) = 9.45, N’s as above). c Data from (b) represented as stimulus-by-stimulus G/A ratios from FF-IPSCs by EPSCs. Inset shows data normalised to starting G/A ratios. Asterisks indicate stimuli with significant reductions in G/A ratio for Fmr1^−/Y data (t-test, N’s as in (b)). d Example short-term depression (50 Hz stimulus frequency) of unitary connections between FS-SC and SC-FS neurons (left, right) from paired recordings. Single trials shown. Scale: 50 mV, 10 pA/50 ms. e Short-term plasticity analysis in (b) but for unitary connections tested between connected pairs of FS and SC neurons (Asterisks: p < 0.05, t-test, fits: (p < 0.05, Extra sum-of-squares F-test N: FS to SC Fmr1^+/Y = 9, Fmr1^−/Y = 8, SC to FS: Fmr1^+/Y = 9, Fmr1^−/Y = 6). N’s indicate neurons). f As for (d) but for example connected SC–SC paired recording. Scale: 50 mV, 5 pA/25 ms. g Short-term depression of SC–SC connections was indistinguishable between genotypes (p < 0.05, t-test, N = 17 Fmr1^+/Y, 11 Fmr1^−/Y)

Fig. 5

Relaxed coincidence detection impairs frequency gating in Fmr1-KO L4 networks. a Hyper-summation of high-frequency thalamocortical input in Fmr1^−/Y SCs. Top: example current-clamp recordings showing voltage summation in response to five regular stimuli at frequencies between 5 and 50 Hz. Amplitude is normalized to that of steady-state EPSP. Below: Mean ± SEM normalised EPSP amplitude as a function of stimulus number. Short trains of TC stimuli at 20 and 50 Hz evoked stronger voltage summation in Fmr1^−/Y recordings: asterisks denote significantly elevated responses (p < 0.05, unpaired t-test, Fmr1^+/Y n = 7, Fmr1^−/Y n = 7). b Shifted sensitivity of L4 network to thalamocortical input frequency in Fmr1-KOs. Example current-clamp recordings (10 trials overlaid) showing transient network activity evoked by five repetitive thalamocortical stimuli at 5, 10, 20 and 50 Hz (scale bar: 100 ms, 10 mV). Note relaxed requirement of high-frequency stimulation for generating sustained intracortical activity in Fmr1^−/Y SCs. c Fraction of trials evoking network activity as a function of stimulus pattern. Stimulation frequencies below 20 Hz could not evoke firing (p(spiking) = 0 ± 0) in Fmr1^+/Y slices, but with low-moderate probability in Fmr1^−/Y slices. Asterisks denote stimulation frequencies demonstrating significantly elevated firing probabilities in Fmr1^−/Y recordings (p < 0.05, unpaired t-test, Fmr1^+/Y n = 12 slices from eight animals, Fmr1^−/Y n = 10 slices from 10 animals). The 5 × 50 Hz stimulation pattern could evoke firing with high reliability for both genotypes (p(spiking): Fmr1^+/Y = 0.87 ± 0.11, Fmr1^−/Y = 0.85 ± 0.09). Data represented in b is an extended dataset to that shown in Fig. 9f, g in Booker et al.

Fig. 6

Low precision spiking of Fmr1-KO Layer 4 excitatory neurons to TC stimulation. a Example multi-trial raster of spikes recorded in cell-attached configuration from Fmr1^+/Y (top) and Fmr1^−/Y (bottom) SCs in response to TC-evoked L4 network activity at 50 Hz. Scale: 200 pA/100 ms. b Example calculation of spike density functions for Fmr1^+/Y and Fmr1^−/Y example neurons. Two hundred consecutive trials showing trial-trial variability in the timing of spikes recorded in cell-attached configuration (different cells from those shown in (a)). Bottom: Trial-averaged spike density estimate across trials for neurons shown above. Scale: 50 trials/100 ms. c Mean ± SEM spike probability density functions for responding SCs during the peri-stimulus period of TC-evoked network activity. Mean peak spike probability was reduced in Fmr1^−/Y recordings, calculated across the whole 1 s post-stimulus sampling period (Fmr1^+/Y: 0.023 ± 0.002 sipkes s⁻¹, vs. Fmr1^−/Y: 0.015 ± 0.0016 spikes s⁻¹, p = 0.008, t-test, N: Fmr1^+/Y: 21 neurons, Fmr1^−/Y: 16 neurons). Mean spike density averaged across the successive 200 ms window of was not significantly different between genotypes (t-test, p = 0.7). Scale: p(spike/5 ms) = 0.5%/100 ms. d Left: Spike-time statistics for the first spike fired per trial for SCs. Mean latency (left) and inter-trial precision (right) were significantly slower and reduced in Fmr1^−/Y neurons (p = 0.01 and p = 0.03, respectively. t-test, N’s as in (c)). Right: Spike rate and rate stability was significantly different in Fmr1^−/Y recordings compared with Fmr1^+/Y littermates: SCs fired at rates that were slower and more variable between trials. Plotted points are individual neurons, bars show mean ± SEM values. Asterisks indicate statistically significant differences between genotypes (p < 0.05, t-test, N’s as in (c))

Fig. 7

Modelling TC summation recapitulates spiking phenotypes of Fmr1-KO layer 4 SCs. a Schematic of modelling approach. Left: Five grouped covariates measured from Fmr1^+/Y and Fmr1^−/Y recordings used in simulation. Centre: Simulated synaptic inputs were tuned with kinetics of recorded currents. Right: Parameter spaces were explored in silico to find conditions that either enhanced or suppressed firing in the Fmr1-KO model compared with the WT model. b Left: Input frequency dependence of simulated spiking responses for model Layer 4 neurons receiving different strengths of FFI (G/A ratios between 0 and 10 tested). Coloured areas for each modelled genotype indicate combinations of FFI strength/stimulation frequency at which the models fired at least one spike per trial. Red indicates firing parameter ranges in addition to those of the WT model. Note: (1) moderate strength FFI in the WT model prevents spike firing even at high input frequencies (2) the increased number of simulation conditions leading to spiking in the Fmr1^−/Y model, (3) the insensitivity of spiking regulation in the Fmr1^−/Y model to inhibitory tone even with FFI strengths elevated to extreme levels (10 trials overlaid). Right: example traces for simulated spiking by the two models at different parameter combinations. Inset: note later and more variable spike times in the Fmr1^−/Y model. Scale: 20 mV/50 ms. c In addition to affecting the overall spike firing response shown in (b), genotype-dependent effects were observed in the latency, timing variability and count of spikes fired. Spikes fired later and with lower temporally precision in the Fmr1^−/Y model across a broad range of model conditions, even with the FFI strength increased to the extreme values as observed in the Fmr1^−/Y recordings. More conditions led to spiking in the Fmr1^−/Y, despite a slight decrease in numbers of spikes fired per trial across the distribution compared with Fmr1^+/Y simulations

Fig. 8

Relative contributions of different mechanisms to dysfunction in the Fmr1-KO model. Model parameter space explored for 16 different possible combinations of simulated Fmr1^+/Y (‘WT’) and Fmr1^−/Y (KO) conditions (four parameter groups, two possible genotypes, i.e. 4² combinations). Spike firing conditions (in response to five repetitive stimuli) are shown in green for each intermediate rescue scenario as well as Fmr1^−/Y to Fmr1^+/Y models. For each partial rescue scenario (either with Fmr1^+/Y or Fmr1^−/Y values, shown in blue and red, respectively), the total count of spike firing conditions across the whole 5–50 Hz and 0 < G/A < 10 range is shown compared with that of the wild-type simulation (i.e. full Fmr1^−/Y model fired > 1 spike(s) in 45% more simulated FFI strength/input frequency conditions compared with the Fmr1^+/Y model)

Fig. 9

A Fmr1-KO layer 4 model reproduces features of the network response to TC input. a Schematic of model circuit comprising recurrently connected pools of Ex and In neurons receiving simulated thalamocortical input. b Simulated membrane potential for two example Ex neurons (indicated by grey arrowheads in (d)) for TC input at 10 and 20 Hz stimulation frequency. Black dots indicate stimulus times. Scale: 100 ms/50 mV. c Frequency-dependent spike output of model Ex neurons. Asterisks denote significantly different spike mean spike counts between models (p < 0.05, t-test, 800 neurons per model, 5 random models, average of 10 trials each). d Exemplar full network spike rasters for Ex and In neurons (grey and green, respectively) showing firing patterns in response to 5 × 10–20 Hz model thalamocortical stimuli. Population histograms of Ex neurons overlaid in grey. Scale: 100 ms, 50% Synchrony. e Grand mean Ex (grey) and In (green) population spike density functions (5 random seeds, 10 repeats each) from Fmr1^+/Y and Fmr1^−/Y simulations. Note impaired E–I population interaction in Fmr1-KO simulations. f Phase plot summarizing rhythmic Ex–In population interaction in Fmr1^+/Y and Fmr1^−/Y models. Note reduced global synchrony and impaired recruitment of Inhibitory neurons in Fmr1^−/Y model

Fig. 10

Impaired sensory coding by neural ensembles in a model of Fmr1-KO layer 4 network. a Network-wide representation of an extra oddball stimulus inserted into a regular stimulus train (orange tick, single trial example) by firing changes. Circle sizes: fraction of neurons at each coordinate. Insets: simulated V_membrane of representative neurons for regular (black) and oddball (orange) trials. Fmr1^+/Y model neurons typically increased firing rates on the oddball trial, but Fmr1^−/Y neurons only weakly represented the presence of the oddball stimulus by change in firing rate (bulk of points on diagonal). b Different coding schemes underlie representation of sensory input details for simulated Fmr1^+/Y and Fmr1^−/Y networks. Fmr1^+/Y neurons typically increased their firing rate and advanced their first spike in response to an extra oddball stimulus. Fmr1^−/Y neurons showed inflexibility in their spike rate but a bidirectional change in first spike time. Size of points denotes mean spike train dissimilarity for each neuron (van Rossum distance between spike trains on regular and oddball trials). c Population histogram of spike train encoding of oddball vs. regular input patterns. Population average oddball sensitivity of individual cells was reduced in the Fmr1^−/Y network model (Kolmogorov–Smirnov test). d Classification of input pattern from the spike trains of cell ensembles in the layer 4 model, analogous to readout of input to layer 4 by layer 2/3 neurons. Schematic illustration of random sampling of neurons from the layer 4 network model for input to a linear classifier. e Impaired coding of input detail by ensembles of SCs neurons in the Fmr1^−/Y model. Mean leave-one-out decoder cross-validation error for Fmr1^+/Y and Fmr1^−/Y networks for randomly drawn ensembles of varying sizes between 10 and 500 neurons. At 20 Hz input frequency, ensembles comprised of 10 or 20 Fmr1^−/Y neurons performed significantly worse than Fmr1^+/Y ensembles (t-test, p < 0.05), and no better than chance at 20 and 50 Hz (t-test vs. responses with permuted stimulus labels, p > 0.05). All Fmr1^+/Y ensemble sizes performed better than chance at both frequencies. At 50 Hz, all Fmr1^−/Y ensemble sizes performed worse than corresponding Fmr1^+/Y ensembles (t-test, p < 0.05)