MCs comprise a large fraction of the cells in the dentate hilar region and are implicated in multiple pathological conditions (Scharfman 2016, Scharfman 2018). To investigate whether MCs are involved in the pathophysiology of FXS, we first examined the excitability of MCs in wildtype (WT) and Fmr1 knockout (KO) mice, the FXS mouse model. Action potentials (APs) were recorded from MCs at different membrane potentials (from -64 to -55 mV, set by constant current injection). Unexpectedly, we found that loss of FMRP significantly decreased excitability of MCs, as evident by the reduced number of APs fired (Figures 1B-1C; statistical data for every measurement in this study are listed in Supplementary file 1–Supplementary Table 1). This hypo-excitability is opposite to the changes previously observed in other excitatory neurons in FXS models, which typically exhibit hyperexcitability (Contractor, Klyachko & Portera-Cailliau 2015, Deng & Klyachko 2021). To verify this observation, we employed a ramp protocol to evoke APs and examined various AP parameters (Deng et al 2021, Deng et al 2019, Deng et al 2022). Similarly, we found a decreased number of APs fired (Figures 1D-1E), as well as increased voltage threshold, rheobase, and rheobase charge transfer in MCs of Fmr1 KO mice (Figures 1D and 1F-1H), confirming the hypo-excitable state of MCs in the absence of FMRP. There were no significant changes in AP maximum rise rate, rise time, fall time, duration, peak potential and amplitude in KO MCs (Figures 1I-1N), indicating that the transient Na⁺ current and fast activating K⁺ conductances are likely unaffected in MCs of Fmr1 KO mice.

The MC hypo-excitability in Fmr1 KO mice can be attributed to cell-autonomous and/or circuit defects. We first asked whether this defect was caused by changes in the intrinsic passive membrane properties of MCs by examining the resting membrane potential (RMP), cell capacitance and input resistance at RMP level, and found all of them to be unaffected in Fmr1 KO mice (Figures 2A-2C). However, since input resistance around the threshold level contributes directly to AP initiation, we examined it again at -45 mV (close to threshold level) using a depolarization step in the pharmacologically isolated MCs [using a combination of blockers against both glutamate and GABA receptors (in μM): 10 NMDA, 50 APV, 10 MPEP, 5 gabazine and 2 CGP55845] (Deng et al 2019, Deng et al 2022), as well as in the presence of 1 μM TTX and 10 μM CdCl₂ to block Na⁺ and Ca²⁺ channels, respectively. Under these conditions, the input resistance was significantly lower in Fmr1 KO than WT MCs (Figure 2D-2E), suggesting that the reduced input resistance around the threshold potential may be a primary contributor to MC hypo-excitability in Fmr1 KO mice.

Among conductances active at sub-threshold potentials, Kv7 channels are known to play a major role in determining the input resistance (Baculis, Zhang & Chung 2020, Greene & Hoshi 2017, Jones, Gamper & Gao 2021, van der Horst, Greenwood & Jepps 2020). We thus tested whether the reduced input resistance in Fmr1 KO MCs is associated with Kv7 dysfunction. We found that the Kv7 blocker XE991 (10 μM) increased input resistance in all tested MCs from both WT and Fmr1 KO (Figure 2E), with a significantly larger effect in KO mice (Figure 2F). Importantly, XE991 abolished the difference in input resistance at -45 mV between WT and Fmr1 KO cells (Figure 2E). Moreover, Kv7 inhibition with XE991 also caused a larger shift in the holding current in KO than WT MCs (Figure 3A), and a larger membrane depolarization in KO than WT MCs when the membrane potential was initially set at -45 mV (Figure 3B). We further observed that loss of FMRP increased Kv7 currents in MCs as measured by a ramp protocol (Figures 3C-3D). Together, these results strongly suggest that Kv7 function is increased in Fmr1 KO MCs.

The ‘classical’ Kv7 current is carried predominantly by heteromeric KCNQ2/KCNQ3 channels (Brown & Passmore 2009). KCNQ2 and KCNQ3 have been identified as targets of FMRP translational regulation (Darnell et al 2011) and are highly expressed in the hippocampus (Tzingounis et al 2010). We thus examined if KCNQ2 and KCNQ3 levels were altered by FMRP loss, but found no changes in expression of either isoform in the whole brain lysate as measured by Western blot (Figure 3—figure supplement 1A-1B). To exclude the possibility that expression variation among brain regions masked KCNQ changes in a particular area, we performed the same experiment on dentate gyrus lysate only. We found no detectable changes in expression of KCNQ2 or KCNQ3 in Fmr1 KO mice (Figure 3—figure supplement 1C-1D). We did not further investigate the precise mechanisms underlying enhancement of Kv7 function in the absence of FMRP, since the present study primarily focuses on the functional consequences of abnormal cellular and circuit excitability.

To verify the role of elevated Kv7 function in decreased excitability of Fmr1 KO MCs, we first confirmed that hypo-excitability is still observed in these cells when they were isolated from circuit activity using blockers of both glutamate and GABA receptors (as above). We found this to be the case, as evident by increased threshold, rheobase and rheobase charge transfer (Figures 3E-3G), as well as decreased number of APs (Figure 3H) in circuit-isolated MCs of Fmr1 KO comparing to WT mice. Importantly, Kv7 blocker XE991 abolished all these differences in excitability metrics between genotypes (Figures 3E-3H).

Together, these results indicate that the hypo-excitability of MCs in Fmr1 KO mice has primarily a cell-autonomous origin and can be attributed to abnormally elevated Kv7 function.

Next we examined if, in addition to cell-autonomous Kv7-mediated defects, there is also a circuit contribution to MC hypo-excitability. First, we measured spontaneous and miniature excitatory postsynaptic currents (sEPSCs and mEPSCs), and found that the excitatory drive onto MCs was increased, as evident by the elevated number and instantaneous frequency of both sEPSCs and mEPSCs in KO MCs (Figures 4A, 4B, Figure 4—figure supplement 1A-1B), without changes in their amplitudes (Figures 4E and Figure 4—figure supplement 1C). We note that this increase in excitatory drive cannot explain the MCs’ hypo-excitability; rather these results likely reflect the hyperexcitable state of granule cells (Deng et al 2022), which provide the majority of excitatory drive onto MCs. We next examined inhibitory inputs onto MCs and found that the number and instantaneous frequency of spontaneous inhibitory postsynaptic currents (sIPSCs) was also increased in KO MCs (Figures 4C-4D) without changes in amplitude (Figure 4F). However, we did not observe significant differences in miniature inhibitory postsynaptic currents (mIPSCs) in KO MCs (Figure 4—figure supplement 1D-1F). Because the overall circuit contribution to MC excitability is determined by the net effect of both excitatory and inhibitory drives, we examined the E/I ratio of these inputs, a parameter which integrates changes in both excitatory and inhibitory drives into a single variable and thus allows us to determine the net circuit effect on MCs excitability. Given that the amplitudes of sEPSCs and sIPSCs were indistinguishable between genotypes (Figures 4E-4F), we could estimate the E/I ratio simply by using the mean frequency of spontaneous synaptic events (Figures 4G-4H), and observed an increased E/I ratio in Fmr1 KO MCs (↑ ∼20%, Figure 4I) which acts in opposition to the intrinsic hypo-excitability observed in MCs. Thus these adaptive circuit changes likely function as a compensatory mechanism to the intrinsic MCs’ hypo-excitability in Fmr1 KO mice, similarly to adaptive circuit changes also seen in the cortex (Antoine et al 2019). However, this circuit adaptation is insufficient to compensate for MC excitability defects because we observed overall hypo-excitability in Fmr1 KO MCs in the intact circuits.

The principal output of the dentate circuit is determined by granule cell firing, which is largely controlled by the balance of the excitatory inputs to granule cells from the stellate cells of the entorhinal cortex via the perforant path (PP), local excitatory inputs from MCs and inhibitory inputs from local interneurons. In the Fmr1 KO mice, stellate cells have normal excitability (Deng & Klyachko 2016); the granule cells are hyper-excitable (Deng et al 2022); while MCs are hypo-excitable. We thus next probed excitability of the hilar interneurons, the remaining component of this circuit that has not been examined thus far. Because these inhibitory neurons exhibit both morphological and electrophysiological diversity and no clear correlation between morphology and electrophysiological properties has been observed (Mott et al 1997), here we classified a total of 66 recorded interneurons into 3 types according to AP firing pattern: 1) fast-spiking interneurons (high-frequency non-adapting firing); 2) regular-spiking interneurons (slower and adapting firing); and 3) stuttering-like interneurons (high-frequency irregular bursting firing) (Golomb et al 2007) (Figure 5—figure supplement 1A). Chi-square test showed no significant differences in the ratios of the 3 types of interneurons between WT and KO mice (Figure 5—figure supplement 1A). Furthermore, we found that the passive membrane properties (RMP, capacitance and input resistance) and threshold of these interneurons were not significantly different between genotypes and among cell types (thus data were pooled, Figure 5—figure supplement 1B-1E). These results indicate that the intrinsic excitability of hilar interneurons is not affected significantly in the absence of FMRP, suggesting that alterations in synaptic drive onto these cells may play a critical role in determining changes in local inhibition in Fmr1 KOs.

Therefore, we next examined excitatory and inhibitory drives onto hilar interneurons by recording spontaneous and miniature synaptic inputs to these cells. We found that the number and instantaneous frequency of sEPSCs and mEPSCs were markedly decreased in Fmr1 KO mice (Figures 5A, 5B, and Figure 5—figure supplement 1F-1G) with no changes in amplitudes (Figures 5E and Figure 5—figure supplement 1H). Further, the number and instantaneous frequency of sIPSCs were also decreased in KO interneurons (Figures 5C-5D) without change in amplitude (Figure 5F), while no significant changes in mIPSCs was observed (Figure 5— figure supplement 1I-1K). We then estimated the balance of E/I inputs onto interneurons using the same approach as above for MCs and observed a markedly decreased E/I ratio of inputs onto Fmr1 KO interneurons (↓ ∼60%, Figure 5I). We note that changes in the excitatory drive onto interneurons include both mEPSC and sEPSC frequencies, which reflect not only potential deficits in excitability of their input cells, such as MCs, but also changes in synaptic connectivity/function, that may arise from homeostatic circuit reorganization/compensation (see Discussion).

Considering that excitability of stellate cells is largely unaffected and granule cells show profound hyperexcitability in Fmr1 KO mice, our observations suggest that it is the MCs that provide the dominant excitatory drive to hilar interneurons, causing a marked decrease in E/I ratio of inputs onto interneurons in Fmr1 KO mice. To determine if this is indeed the case, we took advantage of different properties of PP, granule cell and MC axonal terminals (Chancey et al 2014, Chiu & Castillo 2008, Shigemoto et al 1997). Specifically, granule cell and PP axonal terminals contain group II mGluRs (Shigemoto et al 1997), while MC axonal terminals express type I cannabinoid (CB1) receptors (Chancey et al 2014, Chiu & Castillo 2008). Accordingly, mGluR Group II agonist DCG-IV (1 μM) selectively inhibits granule cell-and PP-derived EPSCs onto hilar interneurons; while the CB1 agonist WIN 55212-2 (WIN, 5 μM) selectively inhibits MC-derived EPSCs onto these interneurons. In this analysis, for simplicity and better comparison among cells, we normalized the rate of sEPSCs to its own baseline before agonist application (i.e., normalized frequency). We observed that DCG-IV had little effect on the normalized frequency of sEPSCs recorded in interneurons in both genotypes (Figures 6A, 6C, 6D, 6F), indicating that granule cell-and PP-derived EPSCs comprise a limited portion of excitatory drive onto hilar interneurons in both genotypes. As a control for DCG-IV effectiveness, we observed that DCG-IV strongly reduced the normalized frequency of sEPSCs recorded from MCs that primarily receive excitatory inputs from granule cells (Figures 6G-6I). In contrast, even though WIN markedly reduced the normalized frequency of sEPSCs in interneurons of both WT and KO mice (Figures 6B, 6C, 6E and 6F), it failed to affect the normalized frequency of sEPSC recorded from mossy cells (Figures 6H, 6I), suggesting that MCs provide a significant proportion of excitatory drive onto hilar interneurons. Together with the observations above, this analysis supports the notion that the hypo-excitability of MCs in Fmr1 KO mice is a major factor contributing to the reduction of excitatory drive onto hilar interneurons, which ultimately results in reduced local inhibition.

If upregulation of Kv7 function is largely limited to MCs in the dentate circuit of Fmr1 KO mice, then our findings suggest that inhibition of Kv7 channels would be more efficient to enhance MC excitability in KO than WT mice and thus may be more effective to boost the inhibitory drive onto Fmr1 KO granule cells, thus reducing granule cell hyperexcitability in KO mice. To test this hypothesis, we first simultaneously recorded sEPSCs and sIPSCs in the granule cells to get a better quantitative analysis of the E/I input onto these cells. Granule cells were held at -40 mV resulting in sEPSC appearing as a down-ward current, while sIPSC appearing as up-ward current, which was verified by selective blockers of GABA_A or AMPA and NMDA receptors (Figure 7—figure supplement 1A). We found that loss of FMRP did not affect the mean frequencies, amplitudes and charge transfer of sEPSCs and sIPSCs onto granule cells (Figures 7B, 7C and Figure 7—figure supplement 1B-1G), and thus did not alter the baseline E/I ratio as determined by any of these parameters (Figures 7D, Figure 7—figure supplement 1D-1G). In line with our hypothesis, circuit-wide inhibition of Kv7 channels with bath application of XE991 markedly increased sIPSC frequency in the granule cells of Fmr1 KO but not in WT mice (Figure 7C), while having no effect on sEPSC frequency in either genotype (Figure 7B). As a result, XE991 application shifted the E/I balance toward stronger inhibition in the Fmr1 KO, but not in WT mice (Figure 7D). These results indicate that circuit-wide inhibition of Kv7 channels could be effective in reducing granule cell hyperexcitability in the KO mice. We noted that XE991 also slightly reduced both sEPSC and sIPSC amplitudes, but to a similar extent in both genotypes (Figure 7—figure supplement 1B-1C), and without detectable changes in charge transfer (Figure 7—figure supplement 1E-1F).

To examine more directly if circuit-wide inhibition of Kv7 channels could be effective in reducing granule cell hyperexcitability in the KO mice, we performed analyses of granule cell firing – the ultimate measure of cellular excitability. APs were evoked by a multistep current injection (55–75 pA, Figure 7E). To disentangle the direct-and circuit-mediated-actions of Kv7 inhibition, we first synaptically isolated granule cells from the circuit using the same pharmacological approach as described above, and observed that loss of FMRP strongly increased the AP firing in granule cells (Figures 7E-7F, upper panels), as we reported previously (Deng et al 2022). Inhibition of Kv7 channels with XE991 (10 μM) further increased AP firing in both genotypes, with significantly higher number of APs still observed in the isolated granule cells of Fmr1 KO mice (Figures 7E-7F, lower panels). These results indicate that intrinsic hyperexcitability of granule cells in Fmr1 KO mice is largely unrelated to Kv7 channels. Indeed, we have demonstrated that this intrinsic defect is caused by abnormal extrasynaptic GABA_A receptor activity in the absence of FMRP (Deng et al 2022). We then performed the same recordings in granule cells with intact dentate circuit, and again observed markedly increased excitability in the Fmr1 KO, as evident by much larger number of APs fired by KO than WT granule cells (Figures 7G-7H, upper panels). In line with our hypothesis, circuit-wide inhibition of Kv7 channels with XE991 abolished the difference in the number of APs between genotypes (Figures 7G-7H, lower panels). Together with the observations above (Figure 7F, lower panel), these results indicate that XE991 corrected granule cell hyperexcitability in a circuit-dependent manner.

To further clarify this notion, we compared the effects of XE991 between isolated-(Figure 7F) and circuit-intact-granule cells (Figure 7H), and probed the origin of XE991 contribution to granule cell excitability. Specifically, even though the direct action of XE991 on granule cells was to increase excitability in both genotypes (Figure 7—figure supplement 2A), the circuit action of XE991 resulted in suppressing granule cell excitability in Fmr1 KO mice (-.J ∼6 APs, Figure 7—figure supplement 2B), while having minimal effect in WT mice (changes fluctuating around 0, Figure 7—figure supplement 2B). This is in line with the observations above that XE991 increased sIPSC frequency but not sEPSC frequency, and only in the KO but not in WT mice (Figures 7A-7D). Together with our findings that the interneuron activity is driven in a large part by excitatory input from the MCs, these results indicate that the primary site of action of XE991 is the MCs in Fmr1 KO mice. Thus, our findings support the notion that circuit-wide inhibition of Kv7 channels boosted up inhibitory drive onto granule cells and restored granule cell excitability in Fmr1 KO mice by enhancing the local inhibitory circuit function.

Granule cell excitability and dentate output are not only dependent on basal synaptic inputs (i.e., basal E/I balance, as evident in Figure 7), but more importantly these cellular/circuit functions are controlled by a delicately tuned dynamic E/I balance during circuit activity. By providing the major excitatory drive to local interneurons, MCs play a key role in controlling circuit inhibition onto granule cells via a three-synapse pathway (Granule cells→MCs→Interneurons→Granule cells, Figure 8—figure supplement 3A). Accordingly, we hypothesized that the Kv7-dependent hypo-excitability of MCs in the Fmr1 KO mice compromises the effectiveness of this critical inhibitory pathway.

To test this possibility, we recorded compound postsynaptic current (cPSC) in granule cells by stimulating PP and holding cells at -45 mV. This holding potential was chosen as it was an intermediate potential between the excitatory and inhibitory reversal potentials ensuring comparable driving force for excitatory and inhibitory conductances. The PP-stimulation-evoked cPSC is the summation of largely overlapping excitatory and inhibitory postsynaptic currents, which shows as an initial down-ward excitatory component and followed by an up-ward inhibitory component (Figure 8—figure supplement 1A). At the end of each recording, the pure EPSC was isolated by adding GABA_A receptor blocker gabazine (5 μM) (Figure 8—figure supplement 1B), which then was averaged to create an EPSC template for each cell in order to isolate underlying EPSC and IPSC from the cPSC and calculate E/I ratio (Figure 8—figure supplement 1C-1D). For better comparison among cells, we normalized the cPSC and underlying IPSC to their respective underlying EPSC, which reflects the PP stimulation intensity (Figure 8—figure supplement 1E).

In line with our observations above, loss of FMRP significantly decreased the inhibitory component of cPSC and the underlying IPSC (Figures 8D-8E, upper panels), while the excitatory component of cPSC was comparable between genotypes (Figure 8C, upper panel). As a result, E/I ratio of the inputs onto granule cells was abnormally increased in Fmr1 KO mice (Figures 8F-8G, upper panels). Moreover, we also observed a wider excitation window (defined as the duration of cPSC excitatory component, Figure 8—figure supplement 1C) in the granule cells of the KO mice (Figure 8H, upper panel). In support of our hypothesis, circuit-wide inhibition of Kv7 with XE991 significantly increased the inhibitory component of cPSC and underlying IPSC in Fmr1 KO (Figures 8D-8E, lower panels), but not in WT mice. Consequently, inhibition of Kv7 normalized E/I ratio (Figures 8F-8G, lower panels), as well as the excitation window in the granule cells of Fmr1 KO mice (Figure 8H, lower panel). These results indicate that Kv7-dependent hypo-excitability of MCs in Fmr1 KO mice is a critical defect in the dentate circuit and that circuit-wide inhibition of Kv7 is sufficient to normalize circuit E/I balance.

Finally, we asked what are the functional consequences of MC hypo-excitability and dentate circuit E/I imbalance in Fmr1 KO mice? The power of theta-gamma oscillations is particularly high in the dentate gyrus (Csicsvari et al 2003), and plays a critical role in many dentate functions, such as pattern separation (Leutgeb et al 2007), information coding (Mizuseki et al 2009, Pernia-Andrade & Jonas 2014), and thus learning/memory (Bott et al 2016, Lisman & Jensen 2013, Neves et al 2022). In order to evaluate the contribution of MC defects in Fmr1 KO mice to this critical dentate function, we used a protocol of double-oscillation interplay at the single-cell level (Hasselmo, Giocomo & Zilli 2007, Mircheva, Peralta & Toth 2019). In this paradigm, gamma (∼50Hz) frequency stimulation of PP suppresses granule cell output driven by PP stimulation at the theta (∼5Hz) frequency range (gamma-suppression, for brevity). The experiments included two stimulation protocols: a control protocol at 5 Hz (theta stimulation, whose intensity was adjusted to achieve AP probability of ∼0.5 and the intensity was then kept in the following test protocol for the same cell); and a test protocol with a gamma burst (5 stimuli at 50 Hz) included 200 ms before the 5 Hz theta train. Theta stimulation of the PP evoked a steady baseline firing in granule cells (Figures 8I-8K, upper panels). In line with previous studies (Hasselmo, Giocomo & Zilli 2007, Mircheva, Peralta & Toth 2019), a preceding gamma stimulation suppressed granule cell output in response to theta train in both genotypes (Figures 8I-8K, upper panels). However, this gamma-suppression was much less efficient in the KO than in WT mice, as evident by the significant higher AP probability in KO mice (Figures 8I-8K, upper panels). Considering that granule cells are the only output neurons in the dentate gyrus and thus integrate all dentate circuit operations, these results indicate that loss of FMRP causes abnormal dentate information processing, leading to excessive dentate output.

The dentate GABAergic system plays a critical role in the above gamma-suppression of granule cell output (Mircheva, Peralta & Toth 2019). Based on our findings, we hypothesized that the compromised dentate inhibitory pathway in the Fmr1 KO mice weakens gamma-suppression through enhancing the EPSP summation in granule cells. The EPSP measured in granule cells in response to the PP stimulation integrates both excitatory and inhibitory synaptic inputs onto granule cells, including the direct synaptic input from the PP and all the PP stimulation-associated feedforward and feedback synaptic inputs. In other words, the EPSP in granule cells integrates all dentate circuit “operations”. Mechanistically, once the integrated EPSP reaches the threshold level, it triggers an AP, which makes the EPSP amplitude a direct predictor of AP firing. We found that theta stimulation evoked a steady EPSP (Figure 8—figure supplement 2A-2B), and a preceding gamma burst markedly suppressed the EPSP in WT mice (Figure 8—figure supplement 2A-2B). In line with our results above, this suppression effect was significantly weaker in KO mice (Figure 8—figure supplement 2A-2B), suggesting that the compromised dentate inhibitory pathway weakens gamma-suppression through enhancing the EPSP summation in the KO mice (Figure 8—figure supplement 3B). If this is the case, our results predict that circuit-wide inhibition of Kv7 channels should dampen EPSP summation and restore the gamma-suppression. Indeed, we found that Kv7 blocker XE991 reduced the EPSP amplitude in KO mice to the WT level (Figure 8—figure supplement 2C-2D) and, most importantly, it normalized gamma-suppression of granule cell output during theta activity in Fmr1 KO mice (Figures 8J-8L and Figure 8—figure supplement 3C).

Collectively, these findings demonstrates that MCs hypo-excitability in Fmr1 KO mice results in circuit E/I imbalance that compromises the dentate information processing functions, and that circuit-based interventions, such as inhibition of Kv7, could be a potential therapeutic strategy to re-normalize the circuit E/I balance and ameliorate dentate dysfunction in FXS.