Perinatal methimazole exposure impairs the distribution and function of layer 5 neurons in the mouse auditory cortex

Abstract

Methimazole (MMI) is an antithyroid drug often prescribed for hyperthyroid conditions. While perinatal MMI exposure is used to model hypothyroidism in rodents, resulting in auditory cortex malformation, mice are reportedly less effective in lowering serum thyroxine levels. This raises the question on MMI-induced hypothyroidism being the underlying cause of cortical malformation. Here we examined if and how perinatal MMI exposure in mice influenced the serum thyroxine level and the distribution and function of deep layer projection neurons in the auditory cortex. MMI exposure resulted in little changes in the thyroxine level and in auditory brainstem responses. However, MMI exposure misdistributed Ctip2-immunopositive, corticocollicular neurons (CCNs) toward the white matter in layer 5, independent of thyroid hormone-sensitive neurogranin expression. Morphologically, the MMI exposure increased immature dendritic spines in apical dendrites of pyramidal neurons without affecting the number of mature spines. Functionally, excitatory synaptic activities were elevated in retrogradely-identified CCNs and callosal projection neurons (CPNs), while the excitability of these neurons was differentially elevated in both cell types. MMI increased spike induction probability and decreased burst-like spike amplitude ratio in CCNs, while it decreased spike threshold and increased burst-like spike amplitude ratio in CPNs. MMI exposure consequently decreased neural circuit activity in the deep layers. These data indicate that perinatal MMI exposure could impair cortical development, dendritic spine maturation, and neuronal properties of layer 5 neurons without lowering the serum thyroxine level, implicating that maternal MMI exposure thyroid-independently endangers auditory cognition in the offspring.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-17482-4.

Introduction

Methimazole (MMI) is used clinically to treat hyperthyroid conditions such as Graves’ disease^1–3. Though beneficial to hyperthyroid patients, it could affect the fetus when exposed maternally through the placenta⁴. MMI could also affect newborns through the breast milk of feeding mothers⁵. Thus, caution is necessary for its use since hypothyroidism leads to irreversible intellectual disability in newborn children. In particular, the auditory system is susceptible to hypothyroidism during the fetal and neonatal periods^6,7, where clinical reports are associated with congenital and acquired hypothyroid conditions with deafness^8–11 and auditory orientation deficits in infants¹².

In animal models of hypothyroidism, administering MMI to pregnant and lactating rats results in their offspring exhibiting deficits in the cochlea and brainstem development^13,14 and auditory cortex formation^15,64. MMI exposure for even 3 days during the period of corticogenesis (from embryonic day (E)12 to E15) alters radial migration of cortical neurons¹⁷. In particular, the migration deficit of deep layer neurons is consistently observed⁶⁴. In general, the deep layer neurons are important for interhemispheric neuronal communication and the neuronal feedback to various subcortical nuclei, including the inferior colliculus and the medial geniculate nuclei of the thalamus^20,21. How these callosal and subcortical projection (corticocollicular and corticothalamic) neurons are affected by MMI structurally and functionally has not been well characterized.

Perinatal MMI exposure via pregnant rats dose-dependently reduces plasma thyroxine (T₄) levels, resulting in an alteration of ultrasound vocalization²² and the delay in acoustic startle response acquisition²³. MMI exposure likely inhibits maternal and offspring thyroid peroxidase, an essential enzyme to produce T₄ and triiodothyronine (T₃), affecting the early and late centers of the auditory system in their offspring. Importantly, since cortical malformation could arise due to peripheral alteration²⁴, the auditory cortex abnormality in hypothyroidism may be due to the peripheral damage in the early auditory center, rather than the influence of thyroid hormone deficits on cortical development. In this report, taking advantage of mice, which are less sensitive to MMI to induce hypothyroidism at least with the dose used in the rat studies²⁵, we examined the consequences of perinatal exposure of MMI on mice at the concentration that typically affected the rat auditory system.

Results

Serum T₄ levels and auditory brainstem responses were unaffected by MMI exposure

We exposed 0.02% MMI to dams through drinking water from embryonic day (E)10.5 and continued to give it to offspring after birth until the time of sacrifice for experiments. 0.02% MMI has been established to be sufficient to reduce the thyroid activity in the rats^25,26. MMI exposure starting from E10.5 was aligned with the genesis period of corticofugal neurons of layers 5 and 6, which begins around E11.5²⁷. Unexpectedly, MMI exposure resulted in little influence on serum T₄ levels in early newborns (Fig. 1a). Even the serum of pregnant mothers or the plasma extracted from fetal brains and trophoblasts at E13.5 showed little change in the serum T₄ levels (Fig. 1b). We note, however, that the body weight of offspring was reduced at P60-65 (Fig. 1c, *p = 0.00005), consistent with a previous report²⁸. In contrast to the body weight, there was no significant difference in the brain weight (Fig. 1d). Despite this, the cortical thickness of primary auditory cortex (A1) from the pia to the layer 6/white matter (L6/WM) border increased approximately 5% (Fig. 1e; *p = 0.038). This expansion in cortical thickness was likely contributed by about 10% enlargement of layers 2–4 (Fig. 1e; *p = 0.005).

Fig. 1.

Serum thyroxine (T₄) level and auditory brainstem responses (ABR). (a). Free T₄ levels of blood sera measured by ELISA at the indicated ages of newborns. Control (Ctrl, □) and MMI-exposed mice (MMI, ■) in (a-d). (b). Serum T₄ levels of maternal blood serum, fetal brain extracts, and trophoblast extracts at embryonic day (E) 13.5. For (a) and (b), 3 mice (N) for each condition. (c). Body weight measured at P10 (N = 10 mice), P20 (N = 7) and adult (P60-P65, N = 7). **p = 0.00005, ANOVA with post-hoc Bonferroni test. (d). Brain weight measured at P10 (N = 7), P20 (N = 4), and adults (P60-P65; N = 4). (e). The width of cortical layer (L) 1, L2-4, L5 and L6 at P10 (N = 5), P20 (N = 4), and adult (P60-P65; N = 4) for control (non-shaded) and MMI-exposed mice (shaded). *p < 0.05, **p < 0.001. (f–h). ABR recordings. (f). Traces for ABRs. I-V are positive peaks. (g). Amplitudes of ABR waves I (top)–V (bottom). The amplitude of wave I was measured from the baseline to the first positive peak. The amplitudes of waves II-V were from the positive peak of each wave to the immediately following negative peak. (h). Onset latency of ABR wave peaks. Onset latencies were measured from the stimulus onset to the peaks of waves I-V. In (g) and (h), no significant difference was detected based on two-way ANOVA for control (Ctrl, ○, N = 6) versus MMI-exposed mice (MMI, ●, N = 6).

Since these cortical changes may arise due to hypothyroidism-induced structural changes in inner ear deformation^29–31, we next examined the effect of MMI exposure on auditory brainstem responses (ABRs). The amplitudes and onset latencies of waves I ~ V were not significantly different between vehicle- and MMI-exposed mice (Fig. 1f-h). Although the precise origins of these waveforms are unclear, the wave I ~ IV may represent the VIIIth nerve, the cochlear nuclei, the superior olive, and the inferior colliculus, respectively, in rats or other species³². Therefore, our data suggest that the auditory structures in the cochlea and brainstem are essentially intact, and the cortical effects may not be due to functional deficit of the early auditory centers.

MMI affects the distribution of layer 5, but not layer 6, corticofugal projection neurons

To study the effects of MMI exposure on the number and distribution of corticofugal projection neurons in A1, corticocollicular projection neurons (CCNs) in layer 5 and corticothalamic projection neurons (CTNs) in layer 6 were identified using antibodies against the transcription factors COUP-TF interacting protein 2 (Ctip2) and Forkhead box protein P2 (Foxp2), respectively, which labeled about 97% and 93% of the respective projection neurons³³. Ctip2 was expressed in neurons of both layers 5 and 6 in adult mice (Fig. 2a, b) as observed previously for younger ages (P5, P10, P15, P20)³³. The Ctip2-expressing (Ctip2⁺) neurons distributed between 52.5 and 62.5% from the pia in the middle and superficial layer 5 significantly decreased in MMI (Fig. 2b left). Since the total Ctip2⁺ neurons in layer 5 were unaffected by MMI exposure (Fig. 2b right), and since the cumulative fraction plot of individual Ctip2⁺ neurons significantly shifted to the left (Fig. 2d), the decrease is likely due to altered cell distribution toward the white matter rather than the loss of Ctip2 expression. This downward shift contrasts to the phenomena observed at earlier postnatal ages, where MMI exposure distributed the Ctip2⁺ neurons slightly toward the pia at P10 and returned them to apparently normal (control) locations at P20 (Supplemental Fig. 1), suggesting that the distribution of Ctip2⁺ neurons is age-dependently sensitive to MMI exposure. The effects of MMI exposure on Ctip2⁺ neurons in the adult were limited to layer 5, since the total cell densities of Ctip2⁺ neurons and their distribution in layer 6 were unaffected (Fig. 2b, e). Meanwhile, MMI exposure did not alter the cell distribution or total cell density of Foxp2-expressing (Foxp2⁺) CTNs in layer 6 in the adult (Fig. 2c, e), though Foxp2⁺ cells were located slightly toward the white matter at P10 (Supplemental Fig. 1). These data suggest that the MMI exposure disrupts the migration of Ctip2⁺ CCNs in layer 5 without affecting them in layer 6, while neurogenesis of these neurons was unaffected by the exposure.

Fig. 2.

Distribution of Ctip2- and Foxp2-immunopositive neurons in adult A1. (a). Representative images in primary auditory cortex (A1) for Ctip2 (cyan), Foxp2 (magenta), and overlay (Merge) for control and MMI-exposed mice. Scale bar = 100 µm. (b). Cell density distribution of Ctip2⁺ neurons from the pia (0%) to the L6/WM border (100%), binned at 2.5% (left). Total cell densities of Ctip2⁺ neurons in layers 5 and 6 (right). Control (Ctrl, □) and MMI-exposed mice (MMI, ■). Asterisks are for 52.5–55.0% (p = 0.042), 55.0–57.5% (p = 0.005), 57.5–60.0% (p = 0.022), and 60.0–62.5% (p = 0.042), Student’s unpaired t-tests, N = 4 for Ctrl and MMI. (c). Cell density distribution of Foxp2⁺ neurons (left) and total cell densities (right). (d). Cumulative fraction plots of individual Ctip2⁺ neurons in layer 5. (vertical dashed lines) The ranges of superficial, middle, and deep layer 5 were 46–58%, 58–72%, and 72–84% from the pia, respectively. *p = 1.08 × 10^–16, K-S test (N = 4 for Ctrl and MMI). (e). Cumulative fraction plots of individual Ctip2⁺ (left; p = 0.50) and Foxp2⁺ neurons (right, p = 0.49) in layer 5 , K-S tests (N = 4 for Ctrl and MMI).

MMI exposure shifted the distributions of Ctip2⁺ neurons and Ctip2^-/neurogranin-expressing neurons in an opposite direction

One of the genes that are well recognized to be a direct regulatory target of thyroid hormone is neurogranin (Nrgn), a calmodulin-binding protein distributed to somatodendritic domains, including dendritic spines, in pyramidal neurons^34,35. Previous reports indicated that hypothyroidism reduced the transcript level of Nrgn in the cortex of embryonic³⁶, neonatal^37,38, and adult brains³⁹. We wondered if the distribution shift of CCNs due to MMI exposure involved an alteration of Nrgn expression. Nrgn-immunopositive (Nrgn⁺) somata were distributed throughout A1, except in layer 1 (Fig. 3a, b). MMI exposure tended to increase the cell densities of Nrgn⁺ neurons in layer 4 (42.5–45.0%, p = 0.077) and superficial layer 5 (45.0–47.5%, p = 0.096). Because Nrgn⁺Ctip2⁺ neurons were nearly absent there (Fig. 3c), these Nrgn⁺ cells were mostly Ctip2⁻ neurons (Fig. 3d). Meanwhile, Nrgn⁺ cell density decreased at 60.0–62.5% from pia (*p = 0.041). This would be due to the reduction of Ctip2⁺ neurons (Fig. 3c, see also Fig. 2b). Nrgn^-Ctip2⁺ cell densities were unaffected in any (sub)layers (Fig. 3e). The total cell densities of Nrgn⁺, Ctip2⁺, Nrgn⁺Ctip2⁺, Nrgn⁺Ctip2⁻, and Nrgn⁻Ctip2⁺ neurons in layer 5, however, were not significantly altered by MMI exposure (Fig. 3f, see also Fig. 2b). These data suggest that the main MMI effect is to alter the distribution of Nrgn⁺ cell somata, not cell loss.

Fig. 3.

Distribution of neurogranin- and Ctip2-immunopositive neurons in A1. (a). Representative images for Ctip2 (magenta), neurogranin (Nrgn, cyan), and overlay (Merge) for control and MMI-exposed mice. Scale bar = 100 µm. (b-e). Cell density distribution of Nrgn⁺ (b), Nrgn⁺Ctip2⁺ (c), Nrgn⁺Ctip2^- (d), and Nrgn^-Ctip2⁺ (e). *p = 0.04 (60.0–62.5%) in (b), *p = 0.013 (52.5–55.0%), p = 0.017 (55.0–57.5%), p = 0.03 (60.0–62.5%) in (c). (f). Total cell densities in layer 5 (top) and layer 6 (bottom). In (b-f), Student’s unpaired t-tests for control (Ctrl, □, N = 4) versus MMI-exposed mice (MMI, ■, N = 4). (g). Cumulative fraction plots for individual Nrgn⁺ neurons in layer 6 (p = 0.54), layer 5 (*p = 1.08 × 10^–5), and layers 2–4 (*p = 0.03). (h). Cumulative fraction plots for individual Nrgn⁺Ctip2⁺ neurons (left; *p = 1.34 × 10^–6), Nrgn⁺Ctip2⁻ neurons (middle; *p = 0.0001), and Nrgn⁻Ctip2⁺ neurons (right; *p = 1.34 × 10^–13) in layer 5. In (g) and (h), K-S tests for control (N = 4, dotted line) versus MMI-exposed mice (N = 4, solid line).

To examine how the distribution of Nrgn⁺ cells changes due to MMI exposure, the cumulative fraction plots of individual neurons in layers 2–4, layer 5, and layer 6 were analyzed (Fig. 3g). MMI shifted the cumulative fraction plot for layers 2–4 slightly to the left (*p = 0.032), suggesting a downward shift toward the border of layers 4 and 5. Meanwhile, MMI shifted the fraction plot to the right in layer 5 (*p = 0.00013), reflecting the reduction of the Nrgn⁺ cell densities in the middle layer 5 and their elevation in the superficial layer 5. The data suggest that MMI induces Nrgn⁺ cells of the middle layer 5 an upward shift toward the pia. MMI induced little change in layer 6. Overall, MMI exposure may redistribute Nrgn⁺ cells of layers 2–4 and layer 5 in an opposite direction toward the layer 4/5 border.

Among the Nrgn⁺ cells in layer 5, those with Ctip2⁺ were distributed downward toward the WM in the MMI-exposed mice, while those with Ctip2⁻ showed an upward shift toward the pia (Fig. 3h) like the total Nrgn⁺ cells in layer 5 (Fig. 3g). Since Nrgn⁺Ctip2⁺ neurons accounted for only ~ 10% of Nrgn⁺ population in layer 5, the overall influence of MMI exposure would be to shift the distribution of Nrgn⁺ neurons upward toward the pia in A1. In contrast to Nrgn⁺ cells, the cumulative fraction plot of Nrgn^-Ctip2⁺ cells shifted toward the WM in the MMI-exposed mice (Fig. 3h, see also Fig. 2d). Since both Nrgn⁺Ctip2⁺ and Nrgn⁻Ctip2⁺ cells shifted toward the WM without total cell density changes, the MMI-induced misdistribution of Ctip2⁺ cells did not rely on Nrgn expression. We conclude that MMI exposure misdistributed Ctip2⁺ CCNs and Nrgn⁺ neurons independent of one another in layer 5.

MMI exposure causes formation of immature dendritic spines in layer 5 pyramidal neurons

The change in the distribution of Nrgn⁺ neurons implicates other changes related to Nrgn. Since this synaptic protein plays an important role in synaptic plasticity^40–42, and synaptic activity induces its local translation⁴³ and nuclear translocation⁴⁴, we speculated that the number of dendritic spines and their morphology might be affected. Using the Golgi’s method, we found that the proportion of pyramidal neurons with dendritic spines in proximal apical dendrites were 80% (36 of 45 neurons) in the control mice and 94.1% (48 of 51 neurons) in the MMI-exposed mice (Fig. 4a, b). We then categorized the morphology of dendritic spines into stubby, mushroom, thin, and branched based on the description of a previous report⁴⁵, while those that could not be categorized into any of the groups were considered as atypical (Fig. 4c). Within 0–50 µm from the soma, we found approximately two-fold increase in dendritic spines with the mushroom shape in MMI-exposed mice. We expanded the analysis to 50–100 µm from the soma among neurons with apical dendrites that were clearly isolated and distinguishable with other dendrites. Here too, MMI exposure increased the average number of mushroom-shaped dendritic spines about two-fold. In addition, there was a three-fold increase in thin-shaped, presumptive immature dendritic spines. Meanwhile, the average number of stubby, branched, or atypical-shaped spines was similar between the two groups. These data suggest an abnormal development of dendritic spines along the apical shaft of pyramidal neurons in layer 5 of MMI-exposed mice. Since the total number of other types of dendritic spines appears to increase without altering the number of mature stubby spines, we speculate that MMI exposure affects spine development due to enhanced spine formation or stabilization earlier in development and reduced pruning of immature spines during maturation.

Fig. 4.

Dendritic spine morphology of layer 5 pyramidal neurons in A1. (a). Golgi impregnated pyramidal neurons for control (Ctrl) and MMI-exposed mice (MMI). Scale bar = 25 µm. (b). Proportion of pyramidal neurons with dendritic spines (black) or without them (white). 36 out of 45 neurons in Ctrl (N = 3 mice) and 48 out of 51 neurons in MMI (N = 3) had dendritic spines. (c). The number of dendritic spines in 5 different morphological categories found in 0–50 μm (top) and 50–100 μm (bottom) from the soma of apical dendrites. (top) For 0–50 μm analysis, 36 neurons for Ctrl (□) and 48 neurons for MMI (■) were examined blind to the treatment. *p = 0.047 for Mushroom, *p = 0.047 for Branched. (bottom) For 50–100 μm analysis, 27 neurons for Ctrl (□) and 35 neurons for MMI (■) were examined. *p = 0.024 for Mushroom, *p = 0.011 for Thin. Welch’s t-tests.

MMI cell type-specifically affected the frequency and amplitude of sEPSCs

Given the increase in the number of dendritic spines in layer 5 neurons, we speculated that excitatory synaptic activities were altered. Since there are CCNs and callosal projection neurons (CPNs) in layer 5 that are born and migrate around the same time during early cortical formation, we decided to examine their functional differences. We first examined spontaneous EPSCs (sEPSCs) in the two major types of neurons while retrogradely labeling them with fluorescent Retrobeads injected into inferior colliculus for CCNs and contralateral auditory cortex for CPNs (Fig. 5). The analysis of sEPSCs recorded for 5 min in each cell showed that MMI exposure significantly increased the number of sEPSC events between 12 and 24 ms inter-event intervals (IEIs) (*p = 0.001 or < 0.001) and 9–16 pA amplitudes (*p < 0.0001) in CCNs (Fig. 5b, c; 12 cells, 8 mice for both control and MMI) and 14–24 ms IEI (*p < 0.05, see legend) and 9–16 pA (*p < 0.0001) in CPNs (Fig. 5i, j; 15 cells, 8 mice for control; 6 cells, 4 mice for MMI). However, the average amplitudes of total sEPSCs did not show any significant difference between MMI and control in both cell types (CCNs: control = 14.52 ± 0.63 pA, MMI = 14.60 ± 0.63 pA, p = 0.93; CPNs: control = 13.05 ± 0.55, MMI = 12.85 ± 0.27, pA, p = 0.75). Thus, the number of sEPSCs was elevated for those having a range of IEIs and amplitudes. Note that our criteria of sEPSC (i.e., 3 × SD of spontaneous baseline noise) detected clear events having 8 pA or more, thus biasing relatively higher amplitudes of sEPSCs.

Fig. 5.

Spontaneous EPSCs of CCNs and CPNs. (a, h). Sample traces for CCN (a) and CPN (h) in control (black) and MMI-exposed (red) mice. A portion of the traces (underbar) was expanded below. Scale bars: (top traces) horizontal, 1 s; vertical, 10 pA. (magnified traces) horizontal, 100 ms; vertical, 10 pA. (b, c, i, j). The number of sEPSC events of 5 min recordings for inter-event intervals (IEIs) and amplitudes in CCNs (b, c) and CPNs (i, j) for control (Ctrl, black) and MMI-exposed mice (red). (inset) Horizontal axis was magnified. Error bars represent SEM. Two-way ANOVA: F(1, 3322) = 82.27, ‡p < 0.0001, post-hoc Bonferroni tests: *p = 0.001 at 12 ms, p < 0.001 at 14 ~ 24 ms for IEIs and F(1, 11,022) = 21.12, ‡p < 0.0001, *p < 0.0001 at 9–16 pA for Amplitudes on Ctrl (n = 12 cells, N = 8 mice) versus MMI (n = 12, N = 8) in CCNs. F(1, 2869) = 104.4, ‡p < 0.0001, *p = 0.0011 at 14 ms, p < 0.001 at 16 ms, p = 0.0004 at 18 ms, p = 0.0018 at 20 ms, p = 0.0393 at 22 ms, and p = 0.0003 at 24 ms for IEIs and F(1, 9519) = 17.03, ‡p < 0.0001, * p< 0.0001 at 9–16 pA for Amplitudes on Ctrl (n = 15, N = 8) versus MMI (n = 6, N = 4) in CPN. (d-n). The fractional analysis for IEIs and amplitudes using 300 sEPSCs for each cell. Histograms and cumulative fraction plots for IEIs (CCN: (d, f); CPN: (k, m)) and amplitudes (CCN: (e, g); CPN: (l, n)). Individual and averaged data are in gray and black for control and pink and red for MMI, respectively. Dotted lines are for the traces in (a) and (h). †p = 0.048, K-S test. CCN: n = 11, N = 6 for control, n = 11, N = 7 for MMI; CPN: n = 11, N = 7 for control, n = 6, N = 4 for MMI.

Next, to examine the MMI effects on the distribution of IEIs or amplitudes, we examined the same number of sEPSCs (initial 300 sEPSCs) (Fig. 5d-g, k-n). MMI exposure significantly increased the IEIs of sEPSCs (Fig. 5d, f; †p = 0.048), but it did not affect their amplitudes (Fig. 5e, g) in CCNs (Ctrl: 11 cells, 6 mice; MMI: 11 cells, 7 mice). Meanwhile, MMI exposure did not affect either their frequencies (Fig. 5k, m) nor amplitudes (Fig. 5l, n) in CPNs (Ctrl: 11 cells, 7 mice; MMI: 6 cells, 4 mice for MMI). Overall, MMI exposure increased the number of events in CPNs and CCNs as expected for the increase in synapses (Fig. 4). However, while the presynaptic release frequency onto CCNs, but not CPNs, was increased by MMI exposure, the amplitudes of excitatory synapses onto CCNs or CPNs were unaffected.

MMI exposure differentially affected CCNs and CPNs to increase the membrane excitability for spike induction

We then examined if and how MMI exposure altered the excitable properties of CCNs and CPNs. We recorded passive and active properties of these neurons labeled with the Retrobeads. Each neuron was stimulated by constant 100 ms, −50 pA square pulses and followed by random 1 sec square pulses ranging from −400 pA to +400 pA. This stimulus set was repeated three times.

The analysis of passive membrane properties showed distinct influences of MMI exposure on membrane excitability. In CCNs, resting membrane potential (RMP) and time constant (τ) were slightly increased without much change in input resistance (R_i) in MMI-exposed mice (Supplemental Fig. 2a-d), suggesting increased excitability with slowed responses. Meanwhile, in CPNs, MMI exposure slightly increased RMP like CCNs (Supplemental Fig. 2b) but lowered τ as well as R_i unlike CCNs (Supplemental Fig. 2c-d), suggesting increased excitability with fastened responses.

Next, we analyzed spike or action potential (AP) induction properties. Initial APs were singlet spikes (SS) in many neurons at low stimulus intensities but changed to “burst-like” multiplet spikes (BS) as the intensities increased in both cell types (Fig. 6a, b). There were also cells that showed BS at any stimulus intensities above 50 pA in both cell types (data not shown). Interestingly, MMI exposure increased the spike probability, including those with BS, in CCNs but not in CPNs (Fig. 6c). Given the similar increase in RMPs due to MMI exposure in the two cell types, the higher RMP of CCNs compared to CPNs in basal conditions may have contributed to the difference in spiking probability. Other contributing factors could include changes in initial spike threshold and spiking potentials (i.e., the difference between initial spike threshold and RMP). In CCNs, spike threshold was essentially unaffected (Fig. 6d), though spiking potential appeared slightly lowered by MMI exposure (Fig. 6e). This is probably due to the elevated RMPs, which potentially contributed to the increased spike probability. In CPNs, spike threshold was slightly lowered (Fig. 6d), while spiking potential was clearly decreased (Fig. 6e). Despite these changes, spike probability was unaffected by MMI exposure, perhaps due to the RMPs being relatively low to begin with.

Fig. 6.

Action potential properties of retrogradely-labeled CCNs and CPNs in layer 5. (a). Representative traces for square pulse current stimuli for control (Ctrl) and MMI-exposed mice in CCNs (top) and CPNs (bottom). (b). Magnified traces around the onset of current injections at 3 stimulus intensities. Early spikes (boxes) were further magnified and aligned at spike threshold (V_Th). Raster plot shows spike peak time variability (bottom right). (c). Probability of eliciting RS (black) or BS (red) in control (blank) and MMI-exposed mice (filled) for CCNs (left) and CPNs (right). The number of spike-bearing responses out of 3 stimuli were averaged as spike probability. *p < 0.05. (d, e). Spike threshold and spiking potential (Threshold-RMP) for CCNs (left) and CPNs (right) in control (white circle) and MMI-exposed (black circle) mice. Two-way ANOVA analyses: spike threshold, F(1, 135) = 0.00211, p = 0.963 for CCNs; F(1, 134) = 9.96, ^†p = 0.0020 for CPNs. spiking potential, F(1, 131) = 2.62, p = 0.108 for CCNs; F(1, 134) = 67.74, ^†p < 0.0001 for CPNs, post-hoc Bonferroni tests: *p < 0.05. CCNs: n = 12 cells for Ctrl (N = 8) and n = 12 cells for MMI (N = 8). CPNs: n = 16 cells for Ctrl (N = 8) and n = 8 cells for MMI (N = 5).

MMI differentially modified evoked spike properties in CCNs and CPNs

The spike patterns we detected were different from those previously reported, where CCNs exhibited initial BS followed by after-burst regular SS (AB-RS) (called intrinsic bursting, IB or type A), and CPNs showed adapting SS during stimulation (regular spiking, RS or type B), respectively^46,47. Since our spike patterns show both types in single neurons, depending on stimulus intensities, we developed a method to objectively distinguish between BS or SS, using inter-spike intervals (ISI) and peak amplitudes (Supplemental Fig. 3, 4, see Methods for details). The relationship between ISI and amplitudes clearly shows the different data distribution between IB and RS (Supplemental Fig. 3). However, the border between them was still difficult to draw, so we relied on the DBSCAN cluster analysis, which objectively identified BS as the “outliers” of RS clusters (Supplemental Fig. 4a). With this definition, the number of APs in BS, AB-RS and RS increased as stimulus intensities were increased (Supplemental Fig. 4b). The AP properties within BS or AB-RS were clearly different based on the half-width and peak amplitude of their 1st and 2nd APs (Supplemental Fig. 4c). The half-width of 1st APs was shorter than that of 2nd APs for BS, while no such difference was found for AB-RS. The peak amplitude of BS 1st APs was larger than that of AB-RS 1st APs. The peak amplitude of 2nd APs was smaller than that of 1st APs in BS, while no such difference was found in AB-RS. These analyses clearly distinguished BS and AB-RS.

Square pulse current injections at higher intensities (+100 ~ +400 pA in CCNs and +150 ~ +400 pA in CPNs) elicited IB in control mice (Fig. 6a, b). The threshold intensity (TI), the lowest stimulus intensity that elicited spikes reliably, varied from +100 pA to +300 pA in CCNs and +150 pA to +300 pA in CPNs. Stimulus intensities at 50 pA below TI (TI-50 pA) elicited the SS-only RS patterns in many CCNs and CPNs.

An analysis of the effect of MMI exposure on evoked APs of BS, AB-RS, and RS revealed differences mainly in BS between CCNs and CPNs. Because of the variable TI, we averaged the AP parameter values of BS, AB-RS, or RS obtained at TI, TI + 50 pA, and TI + 100 pA, respectively. In the BS of CCNs, MMI exposure slightly but significantly increased 1st peak amplitude, lowering the 2nd/1st amplitude ratio (Fig. 7a, b). Meanwhile, in CPNs, MMI exposure increased 2nd peak amplitudes more than 1st peak amplitude, resulting in significant increase in the ratio (Fig. 7e, f). For both 1st and 2nd APs of BS in CPNs, their half-widths were shortened (Fig. 7g). Further, the half-width of 1st APs of AB-RS was also reduced (Fig. 7h), as might be expected if the same ion channels contributed to BS and AB-RS spike width. No change in the half-width of BS APs (Fig. 7c) or AB-RS APs (Fig. 7d) was detected in CCNs. Other parameters such as 1st spike onset latency, ISI, and spike number in BS, AB-RS, and RS (Supplemental Table 1, 2, 3, respectively) were unaffected by MMI exposure in both CCNs and CPNs, except the significant reduction of RS spike adaptation in CPNs. Overall, the main effect of MMI exposure is to alter the amplitude and/or width of APs in BS in the layer 5 neurons without altering spike number.

Fig. 7.

Peak amplitude and half-width of BS in CCNs and CPNs. (a-b, e–f). Peak amplitudes of 1st (a, e, left) and 2nd (a, e, right) APs and their 2nd/1st ratio (b, f) in BS of CCNs (a-b) and CPNs (e–f). Control (Ctrl, white) vs. MMI (black). Two-way ANOVA: (a) F(1, 65) = 5.00, ‡p = 0.0289 (for 1st peak amplitude), (b) F(1, 65) = 12.0, ‡p = 0.0009. (e) F(1, 66) = 6.17,‡p = 0.0155 (for 2nd peak amplitude). (f) F(1, 66) = 6.35, ‡p = 0.0142. (c, g). Half-width of 1^st (left) and 2^nd (right) APs in BS of CCNs (c) and CPNs (g). Two-way ANOVA: (g, left) F(1, 66) = 9.48, ‡p = 0.0030, (g, right) F(1, 66) = 11.7, ‡p = 0.011. (d, h). Spike half-width of 1^st APs of AB-RS in CCNs (d) and CPNs (h). Two-way ANOVA: (h) F(1, 66) = 5.59, ‡p = 0.0210, post-hoc Bonferroni tests: *p < 0.05.

Neural network activity declined in the lower layer of the MMI-exposed mice

To examine how above-described abnormalities in neuronal distribution, morphology, and function consequently affect circuit activities in A1, we examined if the expression of c-Fos were affected by MMI exposure using brain sections prepared in the semi-sagittal plane of auditory thalamocortical slices⁴⁸. c-Fos is a reliable, neuronal activity-dependent nuclear marker⁴⁹, whose expression is altered in A1 by sensory experiences such as hearing loss⁵⁰, social sounds⁵¹, and noise rearing⁵². In control mice, c-Fos-immunopositive (c-Fos⁺) cells were found across A1 with relatively high densities in layers 3/4 and layer 6 (Fig. 8a). Although the total cell densities of c-Fos⁺ in layers 5 and 6 were similar between control and MMI-exposed animals (Fig. 8b), MMI exposure reduced c-Fos⁺ cell density within the sublaminas of layer 5 and layer 6: 37.0% at 80.0–82.5% from pia (*p = 0.02) and 40.0% at 92.5–95.0% from pia (*p = 0.02) (Fig. 8c).

Fig. 8.

Distribution of c-Fos- and Ctip2-immunopositive neurons in A1. (a). Representative images for Ctip2 (cyan), c-Fos (magenta), and overlay (Merge) in A1 of semi-sagittal sections. Scale bar = 100 µm. (b). Total cell densities of c-Fos⁺, c-Fos⁺/Ctip2⁺, c-Fos⁺/Ctip2^-, c-Fos^-/Ctip2⁺, c-Fos⁺/Nrgn⁺, and c-Fos⁺/Ctip2⁺/Nrgn⁺ neurons in layer 5 and layer 6 for control (Ctrl, □, N = 3) and MMI-exposed mice (MMI, ■, N = 3). *p = 0.031 for c-Fos⁺/Ctip2⁺ in layer 5, Student’s unpaired t-test. (c-h). Cell density distribution for (c) c-Fos⁺, (d) c-Fos⁺/Ctip2⁺, (e) c-Fos⁺/Ctip2^-, (f) c-Fos^-/Ctip2⁺, (g) c-Fos⁺/Nrgn⁺, and (h) c-Fos⁺/Ctip2⁺/Nrgn⁺ neurons. (c) *p = 0.020 (80.0–82.5%), *p = 0.019 (92.5—95.0%); (d) *p = 0.0013 (50.0–52.5%), *p = 0.011 (52.5—55.0%), *p = 0.030 (55.0—57.5%), *p = 0.031 (57.5–60.0%), *p = 0.0091 (65.0–67.5%); (e) *p = 0.041 (60.0–62.5%), *p = 0.013 (80.0–82.5%), *p = 0.049 (82.5–85.0%); (f) *p = 0.021 (55.0–57.5%); (g) *p = 0.032 (67.5–70.0%), *p = 0.028 (82.5–85.0%); (h) *p = 0.035 (57.5–60.0%), *p = 0.037 (80.0–82.5%), Student’s unpaired t-tests.

We then examined possible differences in c-Fos expression between Ctip2⁺ and Ctip2^- neurons. The total cell densities of c-Fos⁺ cells that co-expressed Ctip2 (c-Fos⁺Ctip2⁺) in layer 5 significantly declined (*p = 0.03) in MMI-exposed mice (Fig. 8b). The reduction was mainly detected in the superficial layer 5 (50–60% from pia; Fig. 8d, e). As for Ctip2⁻ cells, which include CPNs, there was no change in the total cell density (p = 0.27; Fig. 8b), but c-Fos was reduced at the bottom of layer 5 (80.0–82.5%: *p = 0.013, 82.5–85.0%: *p = 0.049, Fig. 8e). MMI exposure, therefore, differentially affects Ctip2⁺ CCNs and Ctip2^- neurons, depending on their location in layer 5.

We also found a reduced c-Fos expression among Nrgn⁺ neurons in layer 5 (Fig. 8g). This difference was largely contributed by the loss of c-Fos expression in Ctip2⁺ CCNs in layers 5/6 because the subtraction of the averaged data for control and MMI resulted in similar curves between c-Fos⁺Nrgn⁺ cells and c-Fos⁺Nrgn⁺Ctip2⁺ cells. Interestingly, c-Fos expression in Nrgn⁺Ctip2⁺ cells became nearly absent in layer 5 for MMI (Fig. 8h). Thus, Ctip2⁺ CCNs may be particularly sensitive to MMI-induced changes in neuronal activities.

Overall, these data indicate that MMI exposure reduces the neuronal activity in the lower layers, as reflected in the laminar specific subsets of CCNs and CPNs in A1, likely because of the alteration in their developmental and functional properties.

Discussion

The main finding of this study is that MMI exposure influences the distribution, morphology, and function of layer 5 neurons without significant reduction of serum T₄ serum levels and ABR deficits. MMI exposure of pregnant rats strongly deprives thyroid hormones (T₃ and T₄) in the serum, resulting in cochlear damage and ABR deficits¹³. In mice, however, similar MMI exposure results in little reduction of T₃ and T₄ in pregnant mice in the previous^25,53 or this study. It is possible that mice did not drink sufficient MMI-containing water to expose them to effective MMI concentration. The water consumption, however, was qualitatively similar between control and MMI-treated mice, and weight loss was detected as found previously^54,55. Other possible reasons for these differences will include the species difference in metabolism^56,57. Perhaps, relevant to this will be a report showing that MMI alone was not effective in suppressing the expression of thyroid hormone-regulated genes in mice⁵⁸. It should be noted that the lack of ABR deficits in mice is not likely due to insensitivity of mouse’s cochlea to hypothyroidism because inner ear damage were reported in mice exposed to another anti-thyroid drug, propylthiouracil (PTU)^22–24. Therefore, the absence of ABR abnormality in mice could be due to not enough MMI to reduce the thyroid hormone level to elicit cochlear damage, perhaps because of their metabolic differences^56,57. Alternatively, the sustained T₄ level might have also occurred due to compensation of thyroid hormones by the production of neonates’ own thyroid hormones without being affected by MMI because of a compensatory function of the mammary gland⁵⁹. Clinical studies in iodine deficient mothers have suggested that the mammary gland tends to concentrate iodine and secrete it into breast milk to provide sufficient supplementation for the newborn. Indeed, the thyroid function of their offspring was unchanged^5,60, although the amount of MMI excreted in milk was almost equal to the levels found in maternal serum^61–63. Thus, newborns may have been resistant to MMI because of sufficient thyroid hormone supply. Nonetheless, MMI exposure affected the distribution, morphology, and function of cortical neurons, particularly in layer 5.

How MMI exposure affected the somatic location of layer 5 neurons in A1 is unclear. In rats treated with MMI during gestation (from E14) and subsequent thyroidectomy several days after birth (at P6), CPNs were misdistributed in the adult age, where they were found mostly between layers 4 and 6, instead of between layers 2 and 6 with a peak in layer 3 and upper layer 4 that is typically detected in untreated cortices¹⁵. It was confirmed later that this misdistribution was due to migration deficit, not due to axonal projection deficit⁶⁴. These previous data in rats suggested migratory impedance of CPNs by chronic hypothyroidism. Our findings differ from the above studies, in that putative CPNs (Nrgn⁺Ctip2⁻ cells) increased in the superficial layer 5. We also found that it was Ctip2⁺ CCNs that exhibited a migratory impedance in the middle layer 5. Whether MMI itself has specific molecular targets in proliferating or migrating cells to impede cell migration remains to be investigated. To our knowledge, there is no report of MMI directly regulating neuronal migration. However, we speculate an indirect mechanism, where MMI regulates cortical formation by inhibiting flavin-containing monoxygenases (FMOs)⁶⁵. FMOs metabolize endogenous trace amines such as phenethylamine and tyramine⁶⁶. Trace amines activate trace amine-associated receptors to modulate the release of dopamine⁶⁷ and serotonin⁶⁸. These biogenic monoamines regulate corticogenesis and neuronal migration^69,70. FMOs also convert hypotaurine to taurine^71,72, which regulates radial migration of cortical neurons via GABA_ARs⁷³. Thus, MMI could affect cortical development by inhibiting FMOs and regulating the availability of metabolites of endogenous molecules. These mechanisms would not restrict MMI’s influence on A1 but affect other brain regions. Precise mechanisms underlying MMI effects on cortical malformation need further investigation.

In this study, we found cell type specific influences of MMI exposure on layer 5 corticofugal projection neurons of mouse A1. The spiking probability or the number of spiking neurons in response to the low square pulse-current stimuli increased in CCNs of MMI-exposed mice, consistent with increased intrinsic excitability of CCNs. We could not find any properties that clearly explain the enhanced excitability, but increased resting potentials and membrane time constants may contribute to it. Meanwhile, in CPNs, the spiking potential became significantly smaller by the MMI exposure. This decreases in spiking potential along with decreased membrane time constants and increased EPSCs may result in enhanced spike firing in response to synaptic inputs to these neurons. These (and other) changes in the neuronal properties might reflect on the change in neural network activities. We found that the MMI exposure decreased the expression of c-Fos⁺ cells in the lower layers of A1, suggesting reduced network activities^50,74. Since CPNs suppress their contralateral cohort via fast spiking interneurons⁷⁵, their enhanced excitability may weaken the interhemispheric activity. This weakened activity will also suppress CCNs, which are normally facilitated by CPNs⁷⁵. Therefore, both types of centrifugal neurons may be suppressed in layer 5 of MMI-exposed mice, which will reflect on the decreased c-Fos expression.

The immature dendritic spine morphology and the reduced circuit activities in layer 5 with little changes in serum T₄ and ABR suggest a thyroid-independent influence of MMI on cortical formation without cochlear injuries. The increased number of mushroom- and thin-shaped dendritic spines on the apical dendrites of layer 5 pyramidal neurons contrasts with previous reports showing a decrease in the number of dendritic spines due to postnatal thyroidectomy⁷⁶ or noise-induced damage to hair cells and spiral ganglion⁷⁷, or a report showing little difference in dendritic spine density by the substantial elevation of ABR threshold in prestin knockout mice⁷⁸. Since peripheral injuries, which are often reported in hypothyroid models (see Introduction), will not increase the number of dendritic spines, the MMI exposure-dependent formation of immature dendritic spines will likely involve mechanisms that are independent of peripheral damage. It is important to note that there was no significant difference in the number of mature stubby spines between control and MMI-exposed mice. Normal development of dendritic spines requires proper spine maturation and elimination involving microglia⁷⁹. Thyroid hormone plays a major role in microglia development⁸⁰ with hypothyroidism reducing the densities of microglia and the ramified processes⁸¹. Since dendritic spines are normally reduced in number as animals mature properly⁸², it may be the elimination process of dendritic spines that MMI exposure compromised. MMI could also regulate spine formation or stabilization via yet unknown mechanisms, which might include the control of the bioamine release via FMO inhibition. Future investigation on the role of MMI on synapse formation would be crucial to further our understanding of its role in the cortical defect.

Lastly, our MMI-exposed mice exhibited features that were consistent with central auditory processing disorders found in the autism spectrum disorder (ASD). For example, patients of fragile X syndrome (FXS), the most common inherited disorder of ASD, exhibit an increased number of dendritic spines in layer 5 pyramidal neurons of the temporal cortex in postmortem studies⁸³ and greatly enhanced sensory evoked responses^84,85. Similarly, the mouse model of FXS, the Fmr1 knock-out mouse⁸⁶, also shows an increase in dendritic spine densities in layer 5 pyramidal neurons of the auditory cortex⁸⁷ along with sensory hyperactivity^88,89 and audiogenic seizure^90–92. It is of note that inferior colliculus plays a major role in audiogenic seizure and cognitive dysfunction, including impairment in auditory fear learning^93,94. Therefore, the functional alteration in CCNs we observed may contribute to these auditory processing disorders. Because MMI is a frequently prescribed drug, this study raises a concern for potential structural and functional damage in the developing auditory cortex. Future studies will reveal whether perinatally MMI-exposed mice exhibit ASD-like behaviors.

Methods

This study was performed in accordance with relevant guidelines and regulations. All methods were reported in accordance with ARRIVE guidelines.

Animals

All animal procedures were conducted according to the Guide for the Care and Use of Laboratory Animals (US National Research Council Committee, 2011) and approved by the Institutional Animal Care and Use Committee of Soka University (Approval No. 22001). Animals were housed in a vivarium with a 12/12 h light/dark cycle. FVB/NJcl strain mice (originally purchased from CLEA Japan, Tokyo) were laboratory grown and mated at postnatal days (P) 50–65. The day when vaginal plug was detected was termed as E0.5. The day that pups were born were termed P0. All animals were given ad libitum access to water and laboratory diet (CRF-1, OrientalBio Co. Ltd., Tokyo). To induce an early maternal hypothyroid condition, 0.02% methimazole (MMI; Sigma) was given in the drinking water to the parents starting from E10 to inhibit the production of thyroid hormone⁹⁵. The MMI water was continuously given until mice were sacrificed. Mouse body weight was measured just before sacrifice. Anesthetized mice were euthanized by decapitation during relevant sample preparations or after in vivo ABR recordings. For histochemical experiments, anesthetized mice were euthanized by cardiac perfusion of fixative.

ELISA for free thyroxine (T₄)

Samples containing thyroxine (T₄) were collected as follows: For serum T₄ level of fetal brains and trophoblasts, pregnant mothers at E13.5 were anaesthetized with urethane (1.3 g/kg) and transcardially perfused with 100 mg/kg of 6-propyl-2-thiouracil (PTU; Sigma) dissolved in 0.1 M phosphate buffer with 0.9% NaCl (PBS), and the uterine horn was removed into a dissection buffer (DB: 125 mM NaCl, 2.5 mM KCl, 2 mM CaCl₂, 10 mM glucose, 1.2 mM MgSO₄, 1.25 mM KH₂PO₄, 25 mM NaHCO₃), the embryo was dissected out, and the fetal brain and trophoblast were removed with the aid of a stereomicroscope. The removed tissues were placed in an extraction buffer (100 mM Tris, 150 mM NaCl, 1 mM EGTA, 1 mM EDTA, 1% Triton-X 100, 1 mM PTU) and homogenized on ice for 25 min. PTU was used to block the synthesis of the thyroid hormone and the conversion of T₄ to tri-iodothyronine (T₃). After the extraction, the samples were centrifuged at 13,000 rpm at 4℃ for 20 min. The supernatant was collected and stored at − 80℃ for later use. For the serum of pregnant mothers at E13.5 and animals after birth, ~0.2—0.5 ml blood was collected by either decapitation (P0, P5) or from the left ventricle (P10 to P20). Blood samples were left to clot at room temperature for 30–40 min before centrifuge at 12,000 rpm at 4℃ for 15 min. The supernatant was collected and stored at -80℃ until use.

The serum T₄ level was measured by using two Free T4 ELISA kits (KA4013, Abnova; CSB-E05080M, CUSABIO) following manufacturers’ protocols. Briefly, the plasma/serum samples were diluted with ddH₂O. 25 µl of the diluted samples, control serum, and calibration samples were added into a 96-well plate provided in the kits. 100 µl of the T₄-horseradish peroxidase conjugate solution was also added into the well for competition with samples and incubated at 37℃ for 60 min. The solutions were then discarded, and the wells were washed. Then, 100 µl of tetramethylbenzidine (chromogen) substrate solution was added into each well and incubated at room temperature for 15–20 min before adding 100 µl of 5% sulfuric acid to stop the chemical reaction. The samples were then spectrophotometrically measured at 450 nm using a microplate reader (Varoskan LUX, Thermo Fisher). A standard curve was plotted with the values collected from the calibration samples, and the concentration (ng/ml) of T₄ was estimated from the standard plot.

Auditory brainstem responses

We recorded auditory brainstem responses (ABRs) as described previously⁹⁶. Briefly, mice were anesthetized using urethane (13 g/kg, i.p.) and xylazine (13 mg/kg, i.p.) and placed on a heating pad (~ 35℃) in a sound-proof chamber. A speaker (STAX, SR307) was placed ~ 15 cm in front of the mouse. White noise (WN) waveform was created by Web Pad Sound Editor using a PC (OPTIPLEX 790, Dell) and generated by the speaker connected to an amplifier (STAX, SRM-323S). Sound intensities for pure tone frequencies were calibrated prior to experiments using an ASA mini sound analyzer (Etani Electronics, Tokyo, JAPAN). An active electrode (< 1 kΩ) was inserted under the head skin at the forehead, a reference electrode was placed below the pinna of a left ear, and the ground electrode was placed below the pinna of a contralateral ear. WN (1 ms duration) was presented at 20 Hz for 25 s at a sound intensity in increasing levels between 25 dB SPL and 70 dB SPL at 5 dB increments. ABRs were amplified and filtered using a Biological Amplifier (Nihon Koden) and recorded using AxoGraph X. For each sound intensity, 500 responses were recorded and averaged with a sampling rate of 20 kHz. The amplitude of ABR wave I was measured from the baseline to the first positive peak. The amplitude of ABR wave II was measured from the second positive peak to the second negative peak. The amplitudes of remaining ABR waves were measured as for wave II.

Brain section preparation

Mice (P60-P75) were anaesthetized with urethane (1.4 g/kg, i.p.) and xylazine (13 mg/kg, i.p.). Trans-cardiac perfusion was performed with 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer, pH 7.4. Brains were removed from the skull and immersed in the fixative overnight. Before sectioning, brains were weighed. For coronal sections, brains were sectioned at 50 μm steps by a vibratome (DTK-1000, Dosaka, Japan). For “thalamocortical” sections that include anterior to posterior areas of A1, the frontal cortex and the cerebellum of fixed brains were excised near the bregma and the lambda, respectively, by cutting at an angle perpendicular to the cortical surface, and remaining brains were then placed the anterior surface down on the bench. At the posterior surface, a cut was made near the dorsolateral end of the cortex at 15–20° to the horizontal line that is perpendicular to the brain’s midline. The cut surface of the brain was glued to a cutting chamber. The brain was submerged with PBS and cut at 50 μm steps. Brain sections were collected in 24-well plates containing PBS and stored at 4℃ or at − 20℃ in 30% ethylene glycol and 30% glycerol mixture in PBS until use.

Immunofluorochemistry

Immunofluorochemistry was performed on free-floating sections using standard protocols. Blocking solution was prepared with 5% sera (donkey or goat sera; Sigma) and 0.3% Triton-X 100. Selected slices were incubated in the blocking solution at room temperature for 90 min before primary antibody incubation. The following primary antibodies were used: rabbit anti-Ctip2 antibody (1:500, ab18465, Abcam), goat anti-Foxp2 antibody (1:500, ab1307, Abcam), rabbit anti-Cux1 antibody (1:500, sc-13024, Santa Cruz Biotechnology), goat anti-c-Fos antibody (1:500, sc-52-G, Santa Cruz Biotechnology), rabbit anti-neurogranin antibody (1:500, AB5620, Millipore). Primary antibodies were incubated either at room temperature for 90 min or at 4˚C overnight (18–24 h). For secondary antibodies, Alexa Fluor 488-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch), Alexa Fluor-633-conjugated goat anti-rat IgG (Life technologies), and Alexa Fluor-555 donkey anti-goat IgG (Abcam) were used. Secondary antibodies were incubated at room temperature for 80 min. Sections were stained with Hoescht-33342 (5 μg/ml; Sigma) for 30 min and mounted with Vectashield antifade mounting medium (H1000, Vector Lab).

Immunofluorescence image acquisition and analysis

For the measurement of cortical and laminar thicknesses, fluorescent images were acquired from 3 to 6 coronal sections (−2.8 to −3.3 mm posterior from the bregma based on mouse brain atlas⁹⁷) for auditory cortex using an epi-fluorescent microscope (5 × Plan-NEOFLUAR lens, Axioskop 2, Zeiss) and analyzed using Axiovision software. In each coronal section, cortical thickness was measured from the pia to the layer 6 and white matter (L6/WM) border drawing a line perpendicular to both. Laminar thickness for each cortical layer was defined as the laminar band formed by cells with immunofluorescence staining against Cux1 for layers 2–4 and Foxp2 for layer 6. Layer 5 was defined as the laminar band between layer 4 and layer 6. The measurements for 3–6 coronal sections per mouse were averaged.

For cell distribution analysis, confocal images were acquired using a Carl Zeiss LSM 700 confocal microscope (10 × Zeiss Plan APO lens) or a Leica TCS SP8 confocal microscope (10 × Leica HC PL APO lens) at 2048 × 2048 resolution with a z-step of 2.0 or 3.5 µm for a total of 21–26 or 12–15 steps, respectively. For the identification of positive fluorescence staining, using NIH ImageJ software, the background fluorescence intensity in each section was corrected by subtracting the threshold intensity determined as the value that was 3 times larger than standard deviation (SD) of the background intensity of randomly selected unstained areas. Stained structures with more than 40.0 pixels were counted as positively identified cells. For cell counts, the region of interest (ROI) was 35–50 μm thick optical sections with a region delimited by the pia, the L6/WM border, and two lines that were parallel to and ± 0.2 mm distant from the line crossing the middle of A1, which was drawn at 15° to a horizontal line that crossed the middle of the ventral division of medial genicular nucleus (MGv)⁹⁸. For cell density analysis, the proportional cortical thickness was binned into 2.5% steps with 0% indicating the pial surface and 100% as the L6/WM border, and cell densities for each bin in each animal were obtained from the summation of cell counts divided by its respective volume for each section and averaging the cell density for 3 to 6 sections per animal. For cell distribution analysis, the cortical thickness was normalized to 100%, the proportional position of each cell was calculated from its X and Y coordinate as a distance from pia, and its distance was plotted in a cumulative fraction plot. The images shown in figures were adjusted for intensity using either Fiji, Zen (Zeiss) or Las X (Leica) software for optimum visualization.

Golgi staining

Trans-cardial perfusion was performed on adult mice (i.e., P60) with ice-cold PBS for 10 min at ~ 10 ml/min. The brains were processed with a Super Golgi kit (Cat. No. 003010, Bioenno Tech, LLC). Briefly, whole brains were immersed in Solution A (potassium chromate, mercuric chloride, and potassium chromate) for 10 days. When impregnation is ready, the brains were rinsed with deionized water, transferred into Solution B, and stored at room temperature in the dark. The solution was renewed after a day, and continued immersion for another day. The brains were sectioned with the vibratome at 100 µm thickness and placed on gelatin-coated slides, which were prepared by immersing glass slides in gelatin solution (gelatin and chromium potassium sulfate). The slides were washed with 0.01 M PBS containing Triton-X (PBS-T) for 20–30 min, and then placed in Solution C (ammonium hydroxide) for 20 min. After that, the slides were incubated in Solution D for 20 min, and then washed in 0.01 M PBS-T for 10 min. Finally, the slides were dehydrated with 100% ethanol and 100% xylene before mounting with a permanent mounting medium (H5000, Vectashield, Vector Lab).

For spine shape analysis of Golgi-stained sections, the brain tissues were imaged with a Keyence microscope (100× lens) with z-step 0.5 µm and the stitching function. The image was analyzed with Image J. First, the soma was identified with the edge of the soma marked as 0 μm, then the length of apical dendrite was measured between 0–50 and 50–100 μm and the number of spines within the area were counted. The spine was categorized based on the length of their head and neck^45,99.

Retrograde labeling

Mice at P45-61 were anesthetized with a gas mixture containing N₂O:O₂ (3:2) and isofluorane (2.5%). Body temperature was maintained around 35℃ by using a heat pad. Deeply anaesthetized mice were prepared for aseptic surgery and placed in the stereotaxic apparatus (SM-11, Narishige, Japan). Following lidocaine (1%, Wako, Japan) application, the dorsal surface of the skull was exposed, and a hole was drilled above the injection site with a dental drill. A glass pipette was made from a borosilicate glass (GD-1.5; Narishige) pulled to a taper length of ~ 10.0 mm and a tip diameter of 30–50 µm using a puller (P-97, Sutter Intstrument) and connected to a 0.5 µl Hamilton syringe. This injector was stereotaxically adjusted to the top of the injection site and inserted into the brain. The coordinate of injection sites was as follows: for contralateral A1 to label callosal projection neurons, 2.5 mm caudal from the bregma, 4.0 mm lateral from the midline, and 0.6–0.8 mm from the surface of the brain; for the central nucleus of inferior colliculus to label corticocollicular projection neurons, 0.5 mm caudal from lambda, 0.8–1.0 mm lateral from the midline, and 0.5 to 0.8 mm from the surface of the brain. For each injection, ~100 nl of Red or Green Retrobeads (Lumafluor Inc., USA) was delivered over a time span of 10 min with ~ 10 nl/min steps by pressure ejection. The injector remained in place for an additional 10 min before being withdrawn. Bupivacaine (0.25%, LKT laboratories, USA) was applied on the surface of the surgery area after suturing to minimize the post-surgery suffering. All the animals were returned to cage for 2-week recovery after injection.

In vitro electrophysiology

Mice were anesthetized with halothane or isoflurane at P60-75. After decapitation, brains were transferred into a petri dish containing ice cold artificial cerebrospinal fluid (ACSF: 125 mM NaCl, 2.5 mM KCl, 2 mM CaCl₂, 10 mM Glucose, 1.2 mM MgSO₄, 1.25 mM KH₂PO₄, 25 mM NaHCO₃, bubbled with 95% O₂/ 5% CO₂). Auditory thalamocortical slices were prepared using a vibrating blade tissue slicer (7000SMZ-2, Campden Instruments) as previously described⁴⁸. Briefly, a brain was blocked at the rostral and caudal ends, placed with the frontal end on the dish surface, and cut diagonally at the dorsolateral end of the left hemisphere. The angle of the diagonal cut to a horizontal line perpendicular to the midline was between 15° and 20°. A thalamocortical slice (450–500 μm) was laid in a chamber filled with oxygenated ACSF at room temperature for more than one and a half hours before recording. A thalamocortical slice was transferred into a recording chamber and maintained under continuous bath perfusion of ACSF (2–3 ml/min) at 25℃ (automatic temperature controller, TC-324B, Warner Instrument). Retrogradely labeled cells in layer 5 were detected under a water-immersion 40× objective lens (LUMPlan FL N, 40x/080 W, Olympus) and monitored using a CCD camera (Rolera em-c², Q-imaging) and its control software mounted on an upright fluorescence microscope (BX51 WI, Olympus) with fluorescent filter sets (U-MWIG3 for red beads and U-MWIB3 for green beads).

Whole-cell patch recordings were carried out with glass electrodes (4–6 MΩ) filled with an internal solution (140 mM K-gluconate, 10 mM KCl, 10 mM HEPES, 0.5 mM EGTA, 2 mM ATP-Mg, 0.4 mM GTP-Na₂, 10 mM phosphocreatine-Na₂-4H₂O, and 0.4% biocytin, pH-adjusted to 7.28 ~ 7.32 with KOH) at 280–290 mOsm/kg (Vapro 5600, Wescore). Glass electrodes (GD-15, Narishige) were pulled using a puller (P-97, Sutter Instrument). Membrane voltages or currents were obtained in the voltage-clamp mode or the current–clamp mode, respectively, using MultiClamp 700B (Axon CNS, Molecular Devices). For voltage clamp recording, whole cell capacitance and series resistance were compensated, and leak current was subtracted. Liquid junction potential was corrected (Pipette offset was auto-corrected throughout recordings). Signals were amplified and low-pass filtered at 10 kHz, and digitally sampled at 20 kHz (Digidata 1440A, Axon CNS, Molecular Devices). Membrane stability was monitored by −50 pA current injections. Data were discarded if input resistance changed more than 15% over the baseline. Recording was made using AxoGraph X on a PC (Dell). Membrane voltages were recorded in response to randomly-injected positive or negative square-pulse currents between −400 and +400 pA (1 sec duration) at 5 sec inter-pulse intervals. Meanwhile, spontaneous excitatory postsynaptic currents (sEPSCs) were recorded for 5 min, holding at − 65 mV to avoid chloride currents (the calculated equilibrium potential of chloride ion was − 65 mV).

Analysis of electrophysiological properties

The resting membrane potential (RMP) was measured during 50 ms before −50 pA square pulse injection. Input resistance was calculated by dividing the voltage difference between steady-state potential (during the last 30 ms of responses) and RMP by the injected current. Membrane time constant was measured at 63% of the rising phase of the square pulse responses. Sag potential was calculated as the difference between early hyperpolarized peak and steady-state potential of the last 100 ms during the responses to randomly injected 1 sec currents (−400, −300, −200, and −100 pA). Spike threshold was defined as the membrane potential at which the first derivative of the voltage (dV/dt) exceeded 20 mV/ms. Spike onset latency was the time from the stimulus onset to the time that the membrane potential reached the spike threshold. Spike peak amplitude was measured from the spike threshold. Spike half-width was measured at the half of the peak amplitude. Inter spike interval (ISI) was calculated as the difference between the time of a spike threshold and that of the next spike threshold. Spike probability was defined as the probability of spike occurrence in response to three current injections. The ISI ratio was the ratio of nth ISI and 1st ISI. These analyses were performed with Microsoft Excel.

To distinguish burst-like multiplets or regular spikes, we examined the relationship between ISIs of two consecutive spikes and the peak amplitude ratio of the first spike to subsequent spikes that were generated in response to random positive current injections (+50 – +400 pA, 50 pA increments) and amplitude ratios of the nth and 1st peak amplitudes using the density-based spatial clustering of applications with noise (DBSCAN) analysis using MATLAB (Mathworks). In the DBSCAN, a cluster of spikes was defined by the threshold for a neighborhood search radius (epsilon) of 0.25 and a minimum number of neighbors of 3. Spikes within this cluster were defined as regular spikes (RS), while those outside the RS clusters, except those with ISIs larger than RS ISIs, were defined as BS.

Spontaneous EPSCs were recorded for 5 min using AxoGraph X and analyzed for the onset time and peak amplitude using MiniAnalysis (Synaptosoft). Low-pass Butterworth filter of 2 kHz was applied to all traces prior to event detection. The detection threshold of EPSC peak amplitude was set to 8 pA. 5 min sEPSCs were analyzed for the total number of sEPSC events. For the fractional analyses of inter-event intervals (IEIs) and peak amplitudes, initial 300 sEPSCs were selected from the 5 min sEPSCs in each cell. Those cells without 300 sEPSCs were excluded from this analysis.

Statistical analysis

Statistical significance was determined using the Student’s unpaired t-test, unless otherwise stated. Two-way ANOVA with Bonferroni’s post-hoc tests, Welch’s t-test, or Mann–Whitney U tests were performed with GraphPad Prism 10. Kolmogorov–Smirnov (K-S) tests (α = 0.05) were performed with MATLAB (Mathworks). All averaged data are reported as mean ± standard error of the mean (SEM). While the number of animals used in this study is a potential limitation of the study, appropriate statistical analyses were performed.

Author contributions

M.C. and H.D.K. designed experiments. M.C., M.N., T.N., K.I., H.Sa. and H.Sh. conducted experiments. M.C., M.N, and H.D.K. analyzed the data. M.C., M.N., and H.D.K. wrote and edited the manuscript.

Data availability

All relevant data are within the paper.

Declarations

Competing interests

The authors declare no competing interests.