We used 10- to 16-week-old TRH KO mice, which were generated previously (Yamada et al., 1997). All procedures for the care and treatment of animals were performed according to the Japanese Act on the Welfare and Management of Animals, and the Guidelines for Proper Conduct of Animal Experiments as issued by the Science Council of Japan. The experimental protocol was approved by the Institutional Committee of Gunma University (No. 17-026; 17-034). All efforts were made to minimize suffering and to reduce the number of animals that were used.

The motor control ability of mice was evaluated using a rotarod test (MK-610A/RKZ; Muromachi Kikai, Tokyo, Japan). Mice were subjected to four trials separated by 30-min intervals on the rod while accelerating from 5 to 50 rpm in 5 min. For rescue experiments of motor ability, saline or TRH (Sigma Aldrich, St. Louis, MO, United States) was administered by intraperitoneal injection (30 mg/kg body weight [BW]) to KO mice 10 min before the first trial of the rotarod test.

Mice were deeply anesthetized and perfused intracardially with 4% paraformaldehyde in 0.1 M phosphate buffer (PB). The entire brain was removed and immersed in 4% paraformaldehyde in 0.1 M PB for 8 h at 4°C. Parasagittal cerebellar slices (50 μm thickness) were prepared using a vibratome (VT1200S; Leica, Wetzlar, Germany). The tissue slices were permeabilized and blocked with 0.1 M PB containing fivefold-diluted G-Block (GenoStaff, Tokyo, Japan), 0.5% (w/v) Triton X-100, and 0.025% NaN₃ (blocking solution). They were then incubated in blocking solution containing the following antibodies overnight at room temperature (26°C): rabbit monoclonal anti-calbindin D-28K (1:500; C2724; Merck, Darmstadt, Germany), mouse monoclonal anti-NeuN (1:1,000; MAB377; Merck), and goat polyclonal anti-parvalbumin (1:200; PV-Go-Af460; Frontier Institute, Hokkaido, Japan). After washing six times with 0.1 M PB, the slices were incubated for 4 h at room temperature (26°C) in the blocking solution containing the following secondary antibodies (1:1,000; all purchased from Thermo Fisher Scientific, Waltham, MA, United States): Alexa Fluor 488-conjugated donkey anti-rabbit immunoglobulin G, Alexa Fluor 568-conjugated donkey anti-mouse immunoglobulin G, and Alexa Fluor 647-conjugated donkey anti-goat immunoglobulin G. After washing six times with 0.1 M PB at RT, the slices were mounted on glass slides using ProLong Diamond Antifade Reagent (P36961; Thermo Fisher Scientific). Each specimen was observed using a fluorescent microscope (BZ-X700; Keyence, Osaka, Japan) or a confocal laser-scanning microscope (LSM 800; Carl Zeiss, Oberkochen, Germany).

Molecular layer thickness was measured using ZEN 2 (blue edition) software (Carl Zeiss). Sagittal slices from the cerebellar vermis (±0.5 μm from the median) were immunostained with the anti-calbindin D-28K antibody and mounted. Twenty-four hours after the mounting, images were obtained from lobules 4/5 – 6 including the primary fissure and lobules 8 – 9 including the secondary fissure, using a confocal microscope (LSM800). The molecular layer thickness was measured 100 μm (primary) or 200 μm (secondary) from the inner end of the fissure (see Figure 1B). The molecular layer thickness of both sides of the fissure was measured using ZEN2, and the sum was defined as molecular layer thickness.

Parasagittal cerebellar slices (250 μm in thickness) were prepared from mice. Briefly, mice were anesthetized deeply by inhalation of isoflurane (3%) and killed by decapitation. The whole brain was quickly dissected out and immersed for a 2–3 min in an ice-cold solution containing the following (in mM): 234 sucrose, 26 NaHCO₃, 2.5 KCl, 1.25 NaH₂PO₄, 11 glucose, 10 MgSO₄, and 0.5 CaCl₂ (pH 7.4 when bubbled with 95% O₂ and 5% CO₂). Parasagittal slices of the cerebellar vermis were obtained using a microslicer (ZERO1; Dosaka-EM, Kyoto, Japan). The slices were maintained in an extracellular solution containing (in mM): 125 NaCl, 2.5 KCl, 2 CaCl₂, 1 MgCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, 10 D-glucose, and 0.1 picrotoxin, bubbled continuously with a mixture of 95% O₂ and 5% CO₂ at room temperature (26°C) for at least 1 h before commencing recording. PCs were visualized using a 40× water-immersion objective attached to an upright microscope (Axioskop, Carl Zeiss). Whole-cell recordings were made from PCs at room temperature (26°C) and the slices were continuously perfused with the extracellular solution during the experiment. The extracellular solution contained 0.1 mM picrotoxin, except for the solution used to record spontaneous and miniature inhibitory postsynaptic currents (sIPSCs and mIPSCs, respectively). The resistance of the patch pipette was 3–6 MΩ when filled with intracellular solution containing (in mM): 122.5 Cs methane sulfonate, 17.5 CsCl, 8 NaCl, 2 Mg adenosine triphosphate, 0.3 Na guanosine triphosphate, 10 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 0.2 ethylene glycol-bis(β-aminoethyl ether)-N,N,N’,N’-tetraacetic acid, and 5 sucrose (pH 7.2, adjusted with CsOH). The intracellular solution used for IPSC recording contained (in mM): 127.5 CsCl, 2 CaCl₂, 1 MgCl₂, 2 MgATP, 0.3 NaGTP, 10 HEPES (pH 7.3, adjusted with CsOH).

Stimulation pipettes were filled with the extracellular solution and placed in the molecular and granule layer to activate parallel fibers (PFs) and climbing fibers (CFs), respectively. PCs were clamped at -70 mV or -10 mV to record PF- or CF-evoked excitatory postsynaptic currents (EPSCs), respectively. Liquid junction potentials were not corrected. To estimate the passive electrical properties of the recorded PCs, we applied 10-mV hyperpolarizing voltage pulses (from -70 mV to -80 mV, 200 ms duration). The averaged trace of the 10 current responses was used for the estimation. Membrane capacitance and input resistance were calculated from the integral of the capacitive charging current, and from the steady-state current amplitude measured late during the pulse, respectively.

Selective stimulation of PFs and CFs was confirmed by paired-pulse facilitation (PPF) and paired-pulse depression (PPD) of EPSC amplitudes with a 50 ms inter-stimulus interval, respectively. To isolate metabotropic glutamate receptor type 1 (mGluR1)-mediated slow EPSCs, repetitive PF stimuli (100 Hz) were applied in the presence of 20 μM 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide, which is a highly selective competitive antagonist of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-type glutamate receptors (Libbey et al., 2016).

For analysis of long-term depression (LTD), PF EPSCs were monitored every 10 s. We always confirmed the stability of PF EPSCs for at least 10 min before LTD induction. Series resistance was continuously monitored every 10 s by applying small hyperpolarizing pulses. Data were discarded when the resistance values changed by >20% of the basal value during the course of the experiment. To induce LTD, we applied conjunctive stimulation consisting of 30 single PF stimuli paired with single 200-ms depolarizing pulses (-70 mV to +10 mV) repeated at 1 Hz. Amplitudes of PF EPSCs were averaged every minute (six traces) and normalized by the averaged value of the responses over 10 min immediately before the conjunctive stimulation. Some LTD recordings were made in the presence of 100 nM TRH (Sigma-Aldrich), 30 μM 8-bromoguanosine 3’,5’-cyclic monophosphate (8-bromo-cGMP) (Sigma-Aldrich) or 100 μM N^G-Nitro-L-arginine methyl ester hydrochloride (L-NAME) (Dojindo Laboratory, Kumamoto, Japan) (Figures 5–7).

Inhibitory postsynaptic currents were recorded in the presence of 25 μM D-2-amino-5-phosphonopentanoic acid (D-APV) and 20 μM NBQX to block excitatory inputs. mIPSCs were recorded in the presence of 1 μM tetrodotoxin (TTX) to block action potentials, and events were detected using a semiautomatic event detector (Clampfit software, San Jose, CA, United States). mIPSC events whose were amplitudes less than 10 pA were discarded.

Significant differences were analyzed using Welch’s t-test and repeated-measures analysis of variance using GraphPad Prism 7 (GraphPad Software, San Diego, CA, United States). The data are expressed as the mean ± standard error of the mean.

The cerebella of 14-week-old TRH-KO mice were compared morphologically to those of wild-type (WT) littermates. After perfusion fixation, the cerebella were cut into 50-μm-thick sections and double-immunostained for calbindin, a PC marker, and NeuN, a granule cell marker. The morphology of the cerebellar cortex was almost indistinguishable between the KO mice and WT littermates (Figure 1A). We measured molecular layer thickness quantitatively, and confirmed no significant differences between the KO mice and WT littermates (Figures 1B,C, Lobule 4/5, 6; P = 0.272, Lobule 8, 9; P = 0.728). Thus, TRH deficiency had no overt influence on cerebellar morphology.

As the behavior of TRH-KO mice in the home cages was indistinguishable from that of their WT littermates, we analyzed their motor ability more carefully using a rotarod apparatus. Numbers of males and females for the rotarod were adjusted between KO and WT mouse groups. However, since BW affects the rotarod performance, we compared the BW of 10- to 12-week-old KO mice with that of their WT littermates. The results showed no significant differences between genotypes (Figure 2A, P = 0.757). In the rotarod test, 10- to 14-week-old TRH-KO mice had similar performance to their WT littermates on the first trial. During the following trials, mice of both genotypes learned the task, and the time on the rod gradually increased. However, learning was significantly slower in the KO mice [Figure 2B, F(1,22) = 5.22, P = 0.032]. These results suggest that motor coordination was normal in TRH-KO mice, but that motor learning ability was significantly impaired.

As revealed in the rotarod test, TRH-KO mice had impaired motor learning. This might be due to a developmental abnormality or merely the absence of TRH-triggered signaling. To address this issue, we administered TRH (6 ml/kg BW, 5 mg/ml) or similar volume of saline intraperitoneally to TRH-KO mice and assessed the effects of TRH on motor learning 10 min after the injection (Figure 2C). TRH-injected KO mice had significantly better performance than saline-injected KO mice in the rotarod test [Figure 2D, F(1,14) = 9.44, P = 0.008]. WT mice were treated similarly with TRH or saline, but, TRH, like saline, had no significant influence on the rotarod performance (Supplementary Figure S1). Notably, TRH-administered KO mice showed favorable rotarod performance, almost comparable to that of TRH-treated WT mice. These results suggest that the motor learning deficit was likely due to an absence of TRH signaling, which was restored by application of exogenous TRH.

We next examined synaptic transmission at PF- and CF-PC synapses using the whole cell patch-clamp technique. Fast PF and CF EPSCs were elicited in PCs by electrical stimulation of PFs or CFs, respectively. Membrane capacitance (Figure 3A, KO: 1053.0 ± 68.5 pF, WT: 903.8 ± 88.9 pF, P = 0.19), and rise time and decay time constants of the fast PF and CF EPSCs (Table 1) were comparable between the KO mice and their WT littermates. We then examined the relationship between stimulus intensity and evoked PF EPSC amplitude, and found no difference between the KO mice and their WT littermates (Figure 3B). Moreover, there were no significant differences in short-term synaptic plasticity, as assessed based on the ratio of PPF in PF EPSCs and the ratio of PPD in CF EPSCs, between the KO mice and their WT littermates (Figures 3C,D).

Long-term depression at PF-PC synapses is thought to be a cellular basis of motor learning. We thus examined whether LTD could be induced at PF-PC synapses in TRH-KO mice. After recording stable PF-EPSCs for at least 10 min, conjunctive stimulation (electrical stimulation to PFs combined with depolarization of the recording PC) induced LTD at PF-PC synapses in WT mice. However, similar stimulation failed to induce LTD in TRH-KO mice (Figures 4A,B, KO: 108.0 ± 9.3%, WT: 82.1 ± 4.5%, P = 0.039).

Repetitive PF stimulation leads to glutamate spillover from the synaptic clefts between the PF terminals and the dendritic spines of PCs. This in turn leads to activation of Gq/11 protein-coupled mGluR1, which is localized postsynaptically at perisynaptic sites of PC dendritic spines (Hirai and Kano, 2018). Activation of mGluR1 leads to production of diacylglycerol and inositol-triphosphate, the latter of which triggers calcium release from the endoplasmic reticulum. The elevated calcium together with diacylglycerol activates protein kinase C (PKC), which plays a key role in LTD induction (Matsuda et al., 2000; Hirai, 2001; Chung et al., 2003). Thus, activation of mGluR1 and downstream signaling is indispensable for LTD at these synapses. To confirm proper mGluR1 activation following repetitive PF stimulation in TRH-KO mouse PCs, we assessed the generation of slow EPSCs. PFs were stimulated with five pulses at 100 Hz in the presence of NBQX to block AMPA receptor-mediated fast EPSCs. The amplitude of the evoked slow EPSC was plotted against the increasing electrical stimulation (Figure 4C). There were no statistically significant differences in the slow EPSC amplitude between TRH-KO mice and their WT littermates. These results suggest that TRH deficiency had no influence on the activation of mGluR1 in PCs.

Next, we examined whether failure of LTD induction in TRH-KO mice could be restored by exogenous TRH. Bath application of TRH (100 μM) to cerebellar slices from TRH-KO mice did not influence the PF EPSC amplitude (Figures 5A,B). However, PF stimulation paired with depolarization of the PC recording effectively induced LTD (Figures 5C,D, KO + TRH: 72.7 ± 5.9%, P = 0.011). Similar bath application of TRH to cerebellar slices from WT mice had no significant effect on LTD expression and the degree of PF EPSC depression (Figures 5C,D, WT: 82.2 ± 4.5%, WT + TRH: 74.7 ± 10.3%, P = 0.528). These results suggest that absence of LTD in the TRH-KO mice were likely due to an absence of TRH signaling, which was restored by application of exogenous TRH.

In the cerebellar cortex, TRH receptors (TRH-R2) are expressed in granule cells and molecular layer interneurons, but not in PCs (Heuer et al., 2000), while TRH was shown to increase cGMP in PCs (Mao et al., 1975; Mailman et al., 1979; Schlichter et al., 1980). NO is a possible mediator of the trans-synaptic event. NO is synthesized by neuronal NO synthase, which is selectively expressed in granule cells and molecular layer interneurons (Vincent and Kimura, 1992; Rodrigo et al., 1994; Schilling et al., 1994; Ihara et al., 2006). Thus, it is assumed that NO, which is produced in these cells upon TRH receptor activation, diffuses into dendritic spines of PCs, and activates a guanylyl cyclase to synthesize cGMP. However, considering the highly bioactive nature of TRH, different mechanisms may contribute to the rescue of cerebellar LTD by exogenous TRH in TRH-KO mice. To clarify this, TRH was perfused to cerebellar slices, with NO production blocked by L-NAME, a cell-permeable NO synthase blocker. Bath application of L-NAME (100 μM) did not affect the amplitude of PF EPSC (Figures 6A,B). However, TRH-mediated rescue of LTD was completely blocked by L-NAME (Figures 6C,D). Next, we applied 30 μM 8-bromo-cGMP, a cell permeable analog of cGMP to cerebellar slices in the absence of exogenous TRH. This concentration of 8-bromo-cGMP did not affect PF EPSC amplitude (Figures 7A,B). Conjunctive stimulation was then applied to induce LTD, resulting in robust LTD without application of exogenous TRH (Figures 7C,D, KO + 8-bromo-cGMP: 78.9 ± 9.0%, P = 0.048). These results suggest the involvement of the NO-cGMP pathway in the TRH-mediated rescue of cerebellar LTD in TRH-KO mice.

Previous studies have shown that TRH increased the frequency of GABA_A receptor-mediated sIPSCs without affecting mIPSCs in the hippocampus (Deng et al., 2006) and lateral hypothalamus (Zhang and Van Den Pol, 2012). Since TRH receptors are expressed in molecular layer interneurons (Heuer et al., 2000), inhibitory synaptic transmission from molecular layer interneurons to PCs may be altered in TRH-KO mice. To test this possibility, we recorded sIPSCs and mIPSCs from PCs, and compared the amplitudes and frequencies between KO mice and their WT littermates. However, there were no significant differences in both the amplitudes (P = 0.437) and frequencies (P = 0.893) of sIPSCs between the genotypes (n = 3 mice, 9 PCs in each group) (Supplementary Figures S2A–C). Next, we recorded mIPSCs after perfusing TTX. Again, there were no significant differences in both the amplitudes (P = 0.687) and frequencies (P = 0.966) of mIPSCs between the genotypes (n = 3 mice, 7 PCs in each group) (Supplementary Figures S2D–F).