Abstract
Fragile X syndrome (FXS) is the single most common monogenetic cause of autism spectrum disorders (ASDs) in humans. FXS is caused by loss of expression of the fragile X mental retardation protein (FMRP), an mRNA-binding protein encoded on the X chromosome involved in suppressing protein translation. Sensory processing deficits have been a major focus of studies of FXS in both humans and rodent models of FXS, but olfactory deficits remain poorly understood. Here, we conducted experiments in wild-type (WT) and Fmr1 knock-out (KO; Fmr1-/y) mice (males) that lack expression of the gene encoding FMRP to assess olfactory circuit and behavioral abnormalities. In patch-clamp recordings conducted in slices of the olfactory bulb, output mitral cells (MCs) in Fmr1 KO mice displayed greatly enhanced excitation under baseline conditions, as evidenced by a much higher rate of occurrence of spontaneous network-level events known as long-lasting depolarizations (LLDs). The higher probability of spontaneous LLDs (sLLDs), which appeared to be because of a decrease in GABAergic synaptic inhibition in glomeruli leading to more feedforward excitation, caused a reduction in the reliability of stimulation-evoked responses in MCs. In addition, in a go/no-go operant discrimination paradigm, we found that Fmr1 KO mice displayed impaired discrimination of odors in difficult tasks that involved odor mixtures but not altered discrimination of monomolecular odors. We suggest that the Fmr1 KO-induced reduction in MC response reliability is one plausible mechanism for the impaired fine odor discrimination.
SIGNIFICANCE STATEMENT Fragile X syndrome (FXS) in humans is associated with a range of debilitating deficits including aberrant sensory processing. One sensory system that has received comparatively little attention in studies in animal models of FXS is olfaction. Here, we report the first comprehensive physiological analysis of circuit defects in the olfactory bulb in the commonly-used Fmr1 knock-out (KO) mouse model of FXS. Our studies indicate that Fmr1 KO alters the local excitation/inhibition balance in the bulb, similar to what Fmr1 KO does in other brain circuits, but through a novel mechanism that involves enhanced feedforward excitation. Furthermore, Fmr1 KO mice display behavioral impairments in fine odor discrimination, an effect that may be explained by changes in neural response reliability.
Introduction
Fragile X syndrome (FXS) is the single most common monogenetic cause of autism spectrum disorders (ASDs) in humans. Individuals with FXS can exhibit a range of debilitating deficits in cognitive abilities, social interactions and communication, hyperactivity and anxiety, and sensory processing (Hagerman et al., 2017; Mila et al., 2018). FXS results from loss of expression of the Fragile X mental retardation protein (FMRP), an mRNA-binding protein encoded on the X chromosome involved in suppressing protein translation. Previous studies in FXS patients and animal models of FXS have demonstrated that impaired expression of FMRP causes defects in neuronal branching and dendritic spine morphology (Davis and Broadie, 2017; Bagni and Zukin, 2019). Alterations in specific neurotransmitter systems such as GABA are also implicated in FXS, serving as a basis for potential pharmacological treatments (Berry-Kravis et al., 2018; Milla et al., 2023). Much of the attention in mechanistic studies of FXS in recent years has focused on brain circuits that are involved in sensory processing (Sinclair et al., 2017; Rais et al., 2018). This in part reflects the fact that sensory deficits are considerable in patients with FXS across several sensory modalities, appearing for example as hypersensitivity to sensory stimuli or deficits in sensory gating (Miller et al., 1999; Frankland et al., 2004; Kogan et al., 2004; Baranek et al., 2008; Van der Molen et al., 2012). In addition, the relatively conserved nature of sensory circuits between humans and animals means that studies in animal models of FXS should be especially fruitful for understanding sensory processing deficits.
One sensory system however that has received less attention is olfaction (reviewed in (Bodaleo et al., 2019). In studies in the commonly-used Fmr1 knock-out (KO) mouse model of FXS, only a few publications have reported olfactory behavioral deficits (Larson et al., 2008; Schilit Nitenson et al., 2015), with somewhat conflicting results, and there are few reports examining effects of KO on olfactory neurons (Galvez et al., 2005; Scotto-Lomassese et al., 2011). The relative paucity of mechanistic studies of olfaction in FXS is surprising. For ASDs in general, there is increasing evidence that humans with ASDs suffer significant olfactory deficits (Dudova et al., 2011; Lane et al., 2014; Rozenkrantz et al., 2015; Boudjarane et al., 2017; Koehler et al., 2018); olfactory deficits also appear to extend to patients with FXS (Miller et al., 1999). In addition, olfaction is expected to be an excellent system to study FXS deficits in mouse models. Olfaction is a critical sensory system for mice, playing essential roles both in finding food and reproduction (Bhattacharyya and Bhalla, 2015; Gire et al., 2016; Ishii and Touhara, 2019), and, as such, behavioral experiments that assess olfactory function are well-developed and robust (Bodyak and Slotnick, 1999). Social communication in mice also relies significantly on olfaction (Bozdagi et al., 2010; Brang and Ramachandran, 2010).
Here, we sought to examine olfactory dysfunction in Fmr1 KO mice that lack expression of the gene encoding FMRP using both electrophysiological and behavioral approaches. In our electrophysiological experiments, conducted in brain slices, we focused our analyses on the primary olfactory processing structure, the olfactory bulb. The bulb is a rich site of both inhibitory neural interactions (Yokoi et al., 1995; Burton, 2017) as well as excitatory interactions that amplify neural responses to sensory input (Carlson et al., 2000; Schoppa and Westbrook, 2001; Christie and Westbrook, 2006; Vaaga and Westbrook, 2016). These processes may be expected to regulate olfactory sensitivity and/or discrimination. In our electrophysiological analyses, we found that Fmr1 KO caused a large increase in spontaneous, long-lasting excitatory events in output mitral cells (MCs). This increase, which appeared to be due at least in part to decreased inhibitory connections, reduced the reliability of the MC response to sensory afferent stimulation. In behavioral studies involving a go/no-go operant discrimination task, we found that Fmr1 KO impaired the ability of mice to discriminate monomolecular odors from odor mixtures, but did not impair easier discrimination tasks.
Materials and Methods
Ethical approval and experimental animals
All experiments were approved by the Institutional Animal Care and Use Committee at the University of Colorado Anschutz Medical Campus in accordance with guidelines set by the United States Department of Health and Human Services and outlined in the Public Health Service Policy on Humane Care and Use of Laboratory Animals.
Fmr1 KO mice (C57BL/6 background) were acquired from The Jackson Laboratory (JAX stock #003025) and bred in-house. Experimental Fmr1 KO mice (hemizygous Fmr1–/y males) were generated by crossing homozygous Fmr1 KO (Fmr1–/–) females with hemizygous Fmr1–/y males. Wild-type (WT) mice (C57BL/6 males) were obtained from Charles River Labs. Mice were maintained on a 12/12 h light/dark cycle with food and water ad libitum except under training and testing conditions (see below). For slice experiments, mice were at postnatal day (P)21–P42. For behavioral studies, mice were approximately six weeks old at start of experiments.
Olfactory bulb slice preparation and electrophysiological recordings
Acute horizontal olfactory bulb slices (300–350 μm in thickness) were prepared from mice following isoflurane anesthesia and decapitation as described previously (Zak and Schoppa, 2022). Experiments were conducted under an upright Zeiss Axioskop 2 FS Plus microscope (Carl Zeiss MicroImaging) fitted with differential interference contrast (DIC) video microscopy and a CCD camera (Hamamatsu). Identified cells were visualized with 10× or 40× Zeiss water-immersion objectives. Recordings were performed at 31–34°C.
The base extracellular recording solution contained the following (in mm): 125 NaCl, 25 NaHCO3, 1.25 NaH2PO4, 25 glucose, three KCl, 1 MgCl2, 2 CaCl2 (pH 7.3 and adjusted to 295 mOsm), and was oxygenated (95% O2, 5% CO2). The pipette solution for whole-cell recordings of excitatory currents contained the following (in mm): 125 K-gluconate, 2 MgCl2, 0.025 CaCl2, 1 EGTA, 2 Na3ATP, 0.5 Na3GTP, and 10 HEPES (pH 7.3 with KOH). Recordings of spontaneous IPSCs (sIPSCs) were conducted with equimolar replacement of K-gluconate with cesium methanosulfonate. 100 μm Alexa 488 was added to the pipette solution to allow visualization of cell processes. Patch pipettes, fabricated from borosilicate glass, were pulled to a resistance of 2–3 MΩ for MCs and 4–6 MΩ for external tufted cells (eTCs). Current and voltage signals were recorded with a Multiclamp 700B amplifier (Molecular Devices), low-pass filtered at 1.8 kHz using an eight-pole Bessel filter, and digitized at 10 kHz. Data were acquired using Axograph X software. Recording sessions with access resistances higher than 15 MΩ were discarded. Reported membrane potential values were corrected for a 7-mV liquid junction potential for our recordings. Epifluorescence was captured by an Axiocam HSm (Zeiss) camera; images were acquired using AxioVision software (Zeiss).
Stimulation of OSN axons in some experiments was performed using a broken-tip patch pipette (5–10 μm in diameter) filled with extracellular solution. The stimulation pipette was placed in the olfactory nerve layer, 50–100 μm superficial to the glomerular layer. Current injections were delivered by a stimulus isolator (World Precision Instruments) under control of a TTL output from Axograph X software. The stimulus interval was 15 s.
Cell identity in the recordings was determined in part by visualizing Alexa 488 fluorescence signals. MCs were easily distinguished by their large size and location in the mitral cell layer. eTCs were distinguished by their position near the border between the glomerular and external plexiform layers, their relatively large, spindle-shaped medium-sized somas (10–15 μm in diameter), a single highly branched apical dendrite that filled most of a glomerulus, the absence of clear lateral dendrites, and a relatively low input resistance (∼0.2 GΩ; (Hayar et al., 2004). Our test tufted cells in fact reflected a mix of eTCs with cell bodies within the glomerular layer along with what are likely to be “superficial” tufted cells with cell bodies in the outermost portion of the external plexiform layer (within 60 μm of the glomerular layer). However, throughout the study, we refer to the cells as eTCs to distinguish them from tufted cells that reside in more inner regions of external plexiform layer that are more “mitral cell-like,” with large cell bodies and long lateral dendrites.
Analysis of electrophysiological data
To evaluate the variance of the baseline currents in MCs and eTCs, current traces were analyzed in windows of >3 s. To remove the contribution of potential drift in the current to the variance estimate, a sloping baseline was first estimated and subtracted from each trace.
Estimates of MC resting potentials were made immediately after whole-cell break-in. For Fmr1 KO mice, where MCs engaged in spontaneous long-lasting depolarizations (sLLDs), baseline resting potentials were estimated from periods between the depolarizations. In the analysis of MC spike activity in WT and Fmr1 KO mice, effort was made to keep the MC baseline potentials near the original estimated resting potentials. In some cases, this required injection of hyperpolarizing current into the cells.
An event detection routine in Axograph was used to detect fast spontaneous EPSCs and IPSCs (sEPSCs and sIPSCs) in MCs and eTCs. In the analysis of sEPSCs in MCs, there were in fact few if any events that could be clearly discerned by eye. The reported very low frequencies of sEPSCs reflected a few events that were selected by the event detection algorithm, but these events generally had low amplitude (<20 pA). In the analysis of sIPSCs in MCs and eTCs, we used a template for the event detection that was similar for WT and Fmr1 KO mice (decay time constant = 3.5 ms). There have been reports that Fmr1 KO can induce modest ∼30% changes in the decay kinetics of sIPSCs (Olmos-Serrano et al., 2010), but we verified by eye that the analysis was capturing the large majority of the events for both mouse types. We also found that making modest adjustments in the template kinetics (adjusting the decay time constant by 30%) did not substantially alter the number of detected events.
To evaluate the relationship between spontaneous and evoked excitatory currents in MCs, an estimate of the prestimulus current was obtained over a 20-ms window just before stimulation of OSNs. An estimate of the evoked current was obtained from the difference between the prestimulus current and the current measured 50–100 ms after stimulation (optimized for each recording to best capture the LLD near its peak). For all recordings conducted in the absence of gabazine (Gbz), a time point of 50 ms after stimulation was chosen for measurements of evoked currents.
Behavioral procedure and training
Odors were delivered using a computer-controlled olfactometer (Bodyak and Slotnick, 1999) that delivered a known concentration of an odor. Odors were made weekly with high purity [Sigma-Aldrich; vehicle = mineral oil (MO)] to a final volume of 10 ml. Volatilized odors (1/40 dilution with air) were presented to mice through an odor port. We tested the animals' ability to discriminate between different monomolecular odors as well as between a single component odor and 1:1 mixture that consisted of two odors (see Table 1). The total concentration of odors for each experiment was 1% in MO. When mixtures were used, the concentrations of the individual components were reduced to bring the total concentration to 1%.
Stimulus pairs presented during go/no-go olfactory discrimination task
For preparing mice for the go/no go behavioral task, mice were water deprived for 1–2 d until they reached 80% of the normal body weight. Behavior training was performed in the olfactometer chamber where mice move freely (Losacco et al., 2020). All mice were first trained to lick the waterspout to obtain water in the presence of odor (1% isoamyl acetate in mineral oil, v/v) in the “begin” task. Subsequently they learned to discriminate 1% isoamyl acetate (S+) versus mineral oil (S−) in the go/no-go discrimination task, followed by learning to discriminate other odor pairs. Data were analyzed for the go/no-go task, but not for the begin task.
In the go/no-go task, mice self-initiated trials by poking their head into the odor delivery port, breaking a light beam that was detected by a photodiode. Mice started the trial spontaneously by poking their nose into the odor spout and licking on the lick port. The delivery of the odors was started at a random time starting 1–1.5 s after nose poke and persisted for 2.5 s. The mice had to decide to either lick on the waterspout at least once during each of four 0.5 s bins in the 2-s response window to obtain a 10-μl water reward for the S+ odor or refrain from licking for the S− odor. Licks were detected as electrical connectivity between the waterspout and the ground plate on which they stand. The mice learned to refrain from licking for the unrewarded odor because of the unrewarded effort of sustained licking. For correct rejections mice left the spout shortly after the last lick that took place 0.5–1.8 s after odor onset. Performance was evaluated in blocks of 20 trials, with 10 S+ and 10 S− trials presented at random. A mouse was considered proficient at discrimination when they achieved ≥80% correct response rate during at least three blocks per day for 2 consecutive days.
Experimental design and statistical analyses
Data were analyzed using Axograph or Prism software (GraphPad). Values are generally expressed as mean ± SE. Significance was most commonly determined using two-tailed nonparametric tests, either the Wilcoxon signed-rank test or the Mann–Whitney U test. In instances in which more than two comparisons were made at once, the Kruskal–Wallis test was first performed to determine whether there were significant differences present within the large dataset. A value of p < 0.05 was considered significant (single asterisks in the figures), except if multiple comparisons of one parameter were made, in which case the Bonferroni correction was applied.
N values reported for electrophysiological experiments refer to the number of test cells, and all statistical comparisons were made based on measurements across different test cells. We also report the number of mice used for slice studies. N values for the behavioral studies reflect the number of test mice.
Results
To analyze circuit-level effects of Fmr1 KO in the olfactory bulb, we conducted whole-cell patch-clamp recordings in bulb slices prepared from both KO (hemizygous Fmr1–/y males) and wild-type (WT) mice. Neural circuit effects of Fmr1 KO in some other brain systems can vary in younger mice before age P21 (Vislay et al., 2013). In our slice studies, we used older mice (P21–P42) to enable comparisons with our behavioral studies that were conducted in older mice (≥P42).
Fmr1 KO induces spontaneous long-lasting excitatory events in mitral cells
In order to assess whether Fmr1 KO alters the behavior of neurons in the olfactory bulb, we began with whole-cell patch-clamp recordings from output mitral cells (MCs; see circuit in Fig. 1a). Upon achieving whole-cell mode at hyperpolarized membrane potentials (at Vhold = –77 mV), we immediately observed a dramatic difference in the current patterns of MCs in Fmr1 KO versus WT mice under baseline conditions, in the absence of local circuit stimulation (Fig. 1b). MCs from KO mice typically displayed robust, spontaneous inward current events that lasted for hundreds of milliseconds to seconds that were generally not apparent in WT mice. These spontaneous current events in MCs from Fmr1 KO mice varied from cell to cell in their magnitude and frequency of occurrence (see two KO examples in Fig. 1bi). In some cases (top example in Fig. 1bi), the current events occurred often enough that they appeared to be part of regular oscillations. In order to quantify the difference in spontaneous currents between WT and Fmr1 KO mice, we computed the variance of the currents recorded under baseline conditions, finding that it was on average ∼4-fold larger in Fmr1 KO versus WT mice across our recordings (Fig. 1c; Fmr1 KO mean ± SE = 248 ± 39 pA2, n = 26 cells from 13 mice; WT = 61 ± 12 pA2, n = 26 cells from 16 mice; p < 0.0001 in Mann–Whitney U test). There were a few MCs in WT mice across our population of recordings that displayed long-lasting spontaneous inward current events (examples with higher variance in Fig. 1c), but this was clearly much rarer in WT versus Fmr1 KO mice.
Fmr1 KO induces spontaneous excitation of MCs. a, Simplified olfactory bulb circuitry. Ouput mitral cells (MCs) receive both direct synaptic inputs (black) from olfactory sensory neurons (OSNs) as well as feedforward excitation from external tufted cells (eTCs). MCs also form reciprocal connections with GABAergic periglomerular (PG) cells and granule cells (GCs); eTCs form reciprocal connections with PG cells. Excitation from eTCs to MCs occurs in the absence of direct synaptic contacts (Gire et al., 2012) and is shown as a thicker gray arrow. Not shown are a number of other types of GABAergic cells in the bulb, e.g., short-axon cells that reside in the glomerular layer. b, Example voltage-clamp recordings (Vhold = –77 mV) from MCs in bulb slices taken from Fmr1 KO mice (two cells in bi) or a WT mouse (in bii). The recordings, which were conducted under baseline conditions in the absence of stimulation, showed spontaneous long-lasting inward current events in KO mice (examples at diagonal gray arrows) that were absent in WT. Note that, in the two example MCs from Fmr1 KO mice, there was significant variability in the rates of occurrence and magnitude of the spontaneous currents. c, The increase in spontaneous currents in Fmr1 KO mice was reflected as a large increase in the variance of the baseline current (*p < 0.0001 in Mann–Whitney U test). Error bars in this all other figures denote standard error (SE). d, Fmr1 KO mice did not display an increase in spontaneous fast EPSCs that could have reflected direct OSN inputs (ns = no significant difference). e, Example current-clamp recordings in MCs from Fmr1 KO (ei) or WT (eii) mice. In KO, but not WT, there were spontaneous phasic depolarizations that induced bursts of action potentials. f, Fmr1 KO did not alter MC resting potentials. The baseline resting potential for MCs from Fmr1 KO mice was estimated from periods between the spontaneous long-lasting depolarizations. g, Dye-fills (100 μm Alex 488) of example MCs from Fmr1 KO and WT mice.
Despite these changes in spontaneous prolonged excitatory currents in MCs, Fmr1 KO did not alter the frequency of kinetically fast spontaneous EPSCs (sEPSCs; Fmr1 KO: 0.029 ± 0.010 Hz, n = 14 cells randomly selected from 26 cell-set; WT: 0.031 ± 0.018 Hz, n = 10 cells; p = 0.49 in Mann–Whitney test; Fig. 1d). In MCs from both WT and Fmr1 KO mice, these events were very rarely observed, with estimated frequencies that were generally below 0.05 Hz. The similar very low frequencies for sEPSCs in WT and KO mice argued against the possibility that the enhanced long-lasting inward currents were because of enhanced spontaneous inputs from olfactory sensory neurons (OSNs).
To assess whether the enhanced spontaneous inward currents in MCs from Fmr1 KO mice enhanced spontaneous action potential firing, we next turned to current-clamp recordings. This showed clear differences between KO and WT (Fig. 1e). Whereas the membrane potential of MCs in WT cells remained flat near their resting potentials and never displayed spikes (n = 11 cells in 8 mice), MCs in Fmr1 KO mice displayed prolonged depolarizations (6.0 ± 0.5 mV, n = 6 cells in 5 mice) that had associated bursts of action potentials. The additional spontaneous action potentials in MCs from KO mice occurred despite the fact that the resting potentials of MCs in KO mice measured between the prolonged depolarizations (–55.9 ± 2.3 mV, n = 12 cells in 5 mice) were not significantly different from the resting potentials in MCs in WT mice (–53.5 ± 1.9 mV, n = 19 cells in 12 mice, p = 0.57 in Mann–Whitney U test; Fig. 1f). This argued that the additional MC action potentials in KO mice could be attributed to the spontaneous depolarizations induced by KO and not because of an overall depolarizing shift in resting potential.
In this initial analysis of MCs in Fmr1 KO mice, we did not conduct a thorough evaluation of cell anatomy. It was nevertheless notable that the basic MC morphology appeared to be preserved in KO mice (Fig. 1g). All of our test MCs in KO mice (n = 23) had multiple lateral dendrites, along with one apical dendrite. We did not observe an increase in the number of apical dendrites of MCs. A prior study (Galvez et al., 2005) reported a modest increase in the number of apical dendrites in MCs in Fmr1 KO mice that were back-crossed into the FVB strain of mice. The same study however reported no effect of KO on dendrite number in the C57BL/6 strain that we used.
Spontaneous slow excitatory events in MCs from Fmr1 KO mice reflect spontaneous LLDs that originate in glomeruli
We next sought to determine the mechanistic basis for the spontaneous prolonged excitatory events in MCs in Fmr1 KO mice. One clue was provided by a series of recordings from MCs from KO mice that had apical dendrites sectioned before reaching the glomerular layer (Fig. 2a); this sectioning was fortuitous, accomplished during the preparation of the bulb slices. In none of these MCs (n = 11 cells in 5 mice) did we observe slow excitatory current events, a fact that was also reflected in a large reduction in current variance in these MCs (41 ± 4 pA2, n = 11) versus those with intact apical dendrites (p < 0.0001 in Mann–Whitney U test; Fig. 2b). This argued that spontaneous slow excitation originated in the glomerular tuft of MCs. Interestingly, the current variance values in MCs with sectioned apical dendrites from KO mice were indistinguishable from variance values in intact MCs from WT mice (p = 0.65 in Mann–Whitney U test), arguing that the enhanced variance in intact MCs from KO mice (Fig. 1c) could be attributed mainly to mechanisms in the MC glomerular tuft. In terms of pharmacological profile, we found that the spontaneous slow excitatory currents in MCs from KO mice were enhanced by the GABAA receptor blocker gabazine (Gbz; 10 μm; 677 ± 286% increase in MC current variance; p = 0.0041 in Wilcoxon signed-rank test, n = 7 cells in 6 mice; variance in Gbz = 1926 ± 572 pA2; Fig. 2c) and eliminated by application of ionotropic glutamate receptor blockers 2,3-dioxo-6-nitro-7-sulfamoyl-benzo[f]quinoxaline (NBQX; 10 μm) and DL-2-amino-5-phosphonopentanoic acid (DL-AP5, 50 μm; data not shown; MC current variance in drugs = 22 ± 2 pA2, n = 5 cells in 3 mice). This argued that the excitatory current events in Fmr1 KO mice were normally downregulated by GABAergic inhibitory inputs and required glutamate receptor activation.
Spontaneous events in MCs from Fmr1 KO mice reflect currents that underlie long-lasting depolarizations (LLDs) that originate in apical dendrites. a, Example current recording (Vhold = –77 mV) from a MC with sectioned apical dendrite (at arrow in associated image) that lacked spontaneous excitatory currents. b, Summary: dendritic sectioning drastically reduced MC baseline current variance in Fmr1 KO mice (*p < 0.0001 in Mann–Whitney U test), bringing it to the level of WT MCs with intact apical dendrites (ns = no significant difference). Plot also shows variance values for MCs from WT mice with sectioned apical dendrites. c, Spontaneous currents in Fmr1 KO mice were enhanced by block of GABAA receptors with gabazine (Gbz; 10 μm), as seen previously for evoked LLDs (Carlson et al., 2000). Three superimposed raw traces from one MC are shown under each condition (ci), along with summary results from seven MCs (cii; *p = 0.0156 in Wilcoxon signed-rank test). d, Further evidence that the spontaneous long-lasting currents in Fmr1 KO mice are spontaneous LLDs was provided by the fact that the spontaneous currents occluded the generation of LLDs evoked by OSN stimulation. Occlusion is represented in the middle raw trace in di, when stimulation (100 µA) elicited no additional current, as well as by the negative correlation (R2 = 0.83, p < 0.0001; in dii) between the magnitude of prestimulus current versus the current evoked 100 ms poststimulation. These occlusion experiments were conducted in the presence of gabazine to enhance the frequency of sLLDs and facilitate analysis. e, Spontaneous events also occurred in WT mice when recordings were conducted in gabazine (10 μm). f, In gabazine, the baseline current variance values for MCs from Fmr1 KO and WT mice were not significantly different.
We wondered whether the spontaneous slow excitatory events were mechanistically similar to the well-characterized long-lasting depolarizations (LLDs) in MCs that are evoked by OSN stimulation (Carlson et al., 2000; Schoppa and Westbrook, 2001). Both the glomerular origin of the prolonged excitatory events along with their modulation by GABAA receptors fit with prior reports of LLDs. That the slow events are the same as stimulus-evoked LLDs was also supported by the fact that the spontaneous current events occluded the generation of LLDs that were evoked by electrical stimulation of OSNs (Fig. 2d). In experiments in which we monitored the timing of both spontaneous and stimulus-evoked events (in the presence of Gbz to increase the frequency of the slow events), we found that stimulation was ineffective at evoking substantial additional current in stimulus trials in which a spontaneous inward current event was already occurring at the point of stimulation (see middle example trace in Fig. 2di). Occlusion was also reflected by a strong negative correlation between the magnitude of the prestimulus current and the additional current evoked by OSN stimulation (Fig. 2dii; R2 = 0.86 ± 0.03, n = 4 cells in 2 mice; p ≤ 0.00037). Because of these various results, we refer to the spontaneous excitatory events in MCs as spontaneous LLDs (sLLDs) throughout the rest of this study.
As an additional point of the characterization of the sLLDs, we asked whether the sLLDs could be observed in WT mice under certain conditions. Prior analysis in the rat olfactory bulb (Carlson et al., 2000; Schoppa and Westbrook, 2001) has shown that current events that resemble LLDs can occur spontaneously in wild-type animals when GABAA receptors are blocked. Indeed, we also found this to be the case in our WT mice. When recordings were conducted in the presence of Gbz, MCs in WT mice displayed robust sLLDs (Fig. 2e; variance = 941 ± 180 pA2, n = 7 cells in 7 mice). The sLLDs in Gbz in WT mice were, moreover, not significantly different from the sLLDs in KO mice in Gbz, based on measurements of current variance (p = 0.62 in Mann–Whitney U test comparing MC current variance in KO and WT in Gbz; Fig. 2f).
The presence of sLLDs in WT mice in Gbz was important, as it indicated that the basic phenomenon of the sLLDs in MCs is not unique to Fmr1 KO mice; they also regularly occur in WT mice with relatively moderate changes in recording conditions. In addition, the fact that addition of the GABAA receptor blocker could make MC currents in WT mice resemble currents in KO mice provided one plausible hypothesis for specific network alterations in KO mice. It has been well demonstrated that LLDs are network-level events within bulbar glomeruli that are downregulated by local GABAergic inhibition (Carlson et al., 2000; Schoppa and Westbrook, 2001; Gire and Schoppa, 2009). The similar sLLDs in Fmr1 KO and WT mice in the presence of Gbz (Fig. 2f) thus argued that one of the main defects leading to much larger sLLDs in Fmr1 KO versus WT mice in the absence of Gbz is a reduction in GABAergic inhibition.
Enhanced sLLDs in MCs reflect alterations in the E/I balance at the level of eTCs rather than reduced inhibition on MCs
MCs in the olfactory bulb receive inhibitory synaptic inputs from a variety of GABAergic interneurons (Burton, 2017; Fig. 1a). These include (but are not limited to) granule cells, which synapse onto the lateral dendrites of MCs, and periglomerular (PG) cells that synapse onto MC apical dendrites. We next built on the pharmacological observation above (Fig. 2f) that suggested that the increase in sLLDs in Fmr1 KO mice could be because of reduced inhibition by recording spontaneous IPSCs (sIPSCs) in MCs. sIPSCs were recorded using a depolarizing holding potential (–7 mV) to help isolate chloride-mediated, outward-going synaptic events. In addition, studies were conducted in the presence of glutamate receptor blockers (10 μm NBQX and 50 μm DL-AP5) to remove any indirect effect on the sIPSCs that could result secondarily from enhanced spontaneous excitation in KO mice. Under these conditions, we were able to isolate synaptic events with the relatively fast kinetics that are typical for GABAA receptor-mediated events (Fig. 3a); these events were also blocked by the GABAA receptor blocker Gbz (10 μm; n = 5). In comparisons of the MC sIPSCs recorded under these conditions, we in fact found no substantial differences between Fmr1 KO and WT mice. MCs from WT and KO mice had similar sIPSC frequencies (Fig. 3b; Fmr1 KO: 1.70 ± 0.40 Hz, n = 15 cells in 7 mice; WT: 2.38 ± 0.36 Hz, n = 21 cells in 8 mice; p = 0.205 in Mann–Whitney U test) and amplitude (Fmr1 KO: 44 ± 3 pA, n = 15; WT: 43 ± 3 pA, n = 21; p = 0.534 in Mann–Whitney U test). Moreover, while Fmr1 KO has been reported to cause modest changes in sIPSC kinetics in other systems (Vislay et al., 2013), the decay kinetics of the sIPSCs were similar for MCs from Fmr1 KO and WT mice (data not shown; Fmr1 KO decay time constant: 4.6 ± 0.5 ms, n = 15; WT: 4.9 ± 0.3 ms, n = 21; p = 0.122 in Mann–Whitney U test). Thus, Fmr1 KO did not appear to significantly alter GABAergic synaptic connections onto MCs (but see Discussion).
Effects of Fmr1 KO on spontaneous IPSCs (sIPSCs) in MCs. a, Example recordings of sIPSCs in MCs (at Vhold = –7 mV) from Fmr1 KO (top) and WT (bottom) mice, conducted in the presence of glutamate receptor (GluR) blockers (10 μm NBQX and 50 μm DL-AP5). Raw traces (ai) and detected sIPSCs (aii) are shown for each MC. The thick traces in aii reflect averages of detected sIPSCs. b, Summary of sIPSC recordings in GluR blockers: Fmr1 KO did not alter the sIPSC frequency or amplitude (ns = no significant difference). c, Example recording of sIPSCs from a MC in Fmr1 KO mice made in the absence of GluR blockers (at Vhold = –57 mV). Note in the raw current trace (ci) the high rate of outward-going sIPSCs associated with each long-lasting inward current event. d, Example recording of sIPSCs from a MC in a WT mouse, made in the absence of GluR blockers (at Vhold = –57 mV). e, Summary: Fmr1 KO enhanced the sIPSC frequency when recordings were conducted in the absence of GluR blockers (*p = 0.020 in Mann–Whitney U test). For KO, frequency estimates were obtained from the entire recording, not just during the long-lasting inward current events.
We did also conduct recordings of sIPSCs in MCs in the absence of glutamate receptor blockers. Under these conditions, we found that KO induced a large increase in the frequency of sIPSCs in MCs (Fig. 3c–e; Fmr1 KO = 2.82 ± 0.40 Hz, n = 24 cells; WT: 1.29 ± 0.32 Hz, n = 12 cells; p = 0.020 in Mann–Whitney U test). For these recordings, we isolated sIPSCs at –57 mV, which enabled us to record both the inward-going current underlying the sLLDs as well as outward sIPSCs at the same time. This showed that the sIPSCs in Fmr1 KO mice were mainly associated with sLLD events (Fig. 3ci), consistent with the enhanced sIPSCs in KO being a secondary result of enhanced excitation of MCs. One might ask why the enhanced sIPSCs in MCs did not by themselves reduce the co-occurring sLLDs in these recordings. A likely explanation is that most of the sIPSCs reflect inputs from cells such as GABAergic granule cells that target the lateral dendrites or soma of MCs (Schoppa, 2006), while LLDs are mainly impacted by inhibitory signals in glomeruli (Gire and Schoppa, 2009).
The sIPSC recordings conducted in the presence of GluR blockers (Fig. 3a,b) provided evidence that the enhanced sLLD probability in MCs of Fmr1 KO mice was not because of reduced inhibitory synaptic connections onto MCs. We next considered an alternate possibility that built on a well-characterized feature of the MC excitatory current, which is that a major component reflects feedforward signals from external tufted cells (Carlson et al., 2000; Schoppa and Westbrook, 2001; De Saint Jan et al., 2009; Najac et al., 2011; Gire et al., 2012; Vaaga and Westbrook, 2016). Moreover, direct excitation of an eTC can evoke MC LLDs, indicating that activation of the eTC-to-MC pathway can specifically drive LLDs. Hence, one mechanism that could underlie the enhanced sLLDs in MCs is a reduction in inhibitory synaptic inputs onto eTCs. This could shift the excitation/inhibition balance in eTCs in favor of excitation, in turn, resulting in more eTC-to-MC feedforward excitation (see hypothesized model in Fig. 4a).
Fmr1 KO reduces inhibition and enhances excitation of eTCs. a, Hypothesized model for Fmr1 KO-induced alterations in bulb circuit: KO reduces inhibitory synaptic input from PG cells onto eTCs. This in turn shifts the E/I balance in eTCs in favor of excitation, resulting in greater eTC-to-MC signaling (large gray arrow). b, Example recordings of sIPSCs in eTCs in Fmr1 KO and WT mice. bi, Dye fills (Alexa 488, 100 μm) of two test cells. For WT, two images were taken at different focal planes. bii, Currents recorded at Vhold = –7 mV from Fmr1 KO (top) and WT (bottom) mice in the presence of GluR blockers (10 μm NBQX and 50 μm DL-AP5). The upward deflections are sIPSCs. biii, Detected sIPSCs for the experiments in bii. c, Summary of sIPSC recordings in eTCs: Fmr1 KO reduced the frequency of the sIPSCs (*p = 0.035, Mann–Whitney U test) but did not alter sIPSC amplitude (ns = no significant difference). d, Example voltage-clamp recordings at a hyperpolarized membrane potential (Vhold = –77 mV) from eTCs in Fmr1 KO (top) and WT (bottom) mice. Recordings were conducted in the absence of GluR blockers. Note the repeated occurrence of spontaneous long-lasting inward current events in the eTC from KO mice, consistent with enhanced eTC excitation. e, The increase in spontaneous excitatory current events in eTCs from Fmr1 KO mice was reflected in a large increase in the variance of the baseline current (*p < 0.0001 in Mann–Whitney U test).
To investigate this hypothesis, we first recorded sIPSCs in eTCs (Fig. 4b), again using depolarized membrane potentials (Vhold = –7 mV) to isolate inhibitory events and conducting experiments in glutamate receptor blockers (10 μm NBQX and 50 μm DL-AP5) to reduce network-level effects of KO. eTCs were identified based on a cell-body location near the boundary between the glomerular and external plexiform layers, the presence of one large dendritic tuft that filled most of a glomerulus, and negligible lateral dendrites (see Methods). Clear sIPSCs could be observed in eTCs from both WT and Fmr1 KO mice, but we found that the sIPSCs occurred at a considerably lower frequency in KO mice (Fmr1 KO: 1.34 ± 0.36 Hz, n = 18 cells in 11 mice; WT: 2.51 ± 0.50 Hz, n = 18 cells in 10 mice; p = 0.035 in Mann–Whitney U test; Fig. 4c), a drop of nearly 50%. On the other hand, there was no change in the amplitude of the sIPSCs (Fmr1 KO: 44 ± 4 pA, n = 18; WT: 38 ± 4 pA, n = 18; p = 0.303 in Mann–Whitney U test).
Does the reduction in inhibition onto eTCs in Fmr1 KO mice result in enhanced excitation of eTCs, as also predicted by our hypothesized mechanism (Fig. 4a)? To assess this possibility, we recorded eTC currents under baseline conditions at a hyperpolarized voltage (Vhold = –77 mV; in this case in the absence of glutamate receptor blockers). This showed that Fmr1 KO induced a dramatic increase in spontaneous, long-lasting excitatory current events in eTCs (Fig. 4d), an effect that was associated with a ∼10-fold increase in current variance (Fmr1 KO = 210 ± 68 pA2, n = 12 cells in 5 mice; WT: 21 ± 6 pA2, n = 9 cells in 4 mice; p < 0.0001 in Mann–Whitney U test; Fig. 4e). The inward currents in eTCs from KO mice were eliminated by application of glutamate receptor blockers (NBQX, 10 μm, plus DL-AP5, 50 μm; data not shown; variance = 17 ± 5 pA2, n = 5 cells in 2 mice; p = 0.0013 in Mann–Whitney U test in comparison with Fmr1 KO control), indicating an underlying glutamatergic mechanism. It should be noted that Fmr1 KO did not alter fast sEPSCs in eTCs that could have reflected spontaneous inputs from OSNs (data not shown; Fmr1 KO frequency: 1.52 ± 0.64 Hz, n = 15 cells; WT frequency: 0.67 ± 0.32 Hz, n = 15 cells; p = 0.136 in Mann–Whitney U test; Fmr1 KO amplitude: 23.4 ± 2.0 pA, n = 14 cells; WT amplitude: 22.2 ± 3.3 pA, n = 12 cells; p = 0.394 in Mann–Whitney U test). We observed large variations in the prevalence of fast sEPSCs in eTCs of both Fmr1 KO and WT mice, with some eTCs displaying very few or no such events (examples in Fig. 4d) but others displaying a large number.
What are the prolonged spontaneous inward current events in eTCs? Prior pair-cell recordings and optogenetic studies have revealed that eTCs have the capacity not only to excite MCs (as shown in Fig. 4a), but also excite other eTCs at the same glomerulus (Gire et al., 2019; Zak and Schoppa, 2021). eTC-to-eTC excitation is also prolonged, reflecting the fact that it is not mediated by traditional synaptic mechanisms. Hence, the spontaneous long-lasting currents in eTCs in KO mice very likely reflected the excitatory signals from a local network of excited eTCs, thereby providing indirect evidence that Fmr1 KO induces spontaneous action potential firing in eTCs. It should also be emphasized that the eTC currents very likely did not reflect enhanced activity within the local MC network in KO mice, i.e., the signals did not reflect MC-to-eTC excitation. Based on the available data (De Saint Jan et al., 2009; Gire et al., 2012), MC activity does not appear to drive eTC currents.
Spontaneous LLDs in MCs from Fmr1 KO mice reduce the reliability of the MC response to OSN stimulation
As a final set of electrophysiological studies in olfactory bulb slices, we asked a more functional question about the relevance of the spontaneous LLDs in MCs in Fmr1 KO mice: What impact do the sLLDs have on the MC response to stimulation of sensory afferents? To address this issue, we recorded the excitatory current response in MCs (at Vhold = –77 mV) following electrical stimulation of OSNs (100 µs; 45–200 µA). These experiments were similar to the occlusion experiments that were conducted earlier on Fmr1 KO mice (Fig. 2d), except we expanded the analysis here to include both Fmr1 KO and WT mice. In addition, we performed the experiments in the absence of GABAA receptor blockade to better mimic the physiological situation. Under these conditions, MCs from both Fmr1 KO and WT mice displayed evoked current responses that were dominated by long-lasting current components including LLDs (see two examples in Fig. 5a; Carlson et al., 2000; Schoppa and Westbrook, 2001).
sLLDs in Fmr1 KO mice reduce the reliability of the MC response to OSN stimulation. (ai) Example MC current responses to OSN stimulation in Fmr1 KO (top; 175 µA) or WT (bottom; 45 µA) mice. Four stimulus trials are shown for each cell, offset in KO and superimposed in WT. Note that, for KO, the magnitude of the evoked current versus the current just before stimulation varies considerably from trial to trial, being large in some cases (trials 1 and 3) and small (trials 2 and 4) in others, depending on the incidence of sLLDs. In contrast, the WT current was similar trial to trial. (aii) The evoked current magnitude across stimulus trials for the experiments in ai. Current magnitude was measured from the difference in the prestimulus current versus the current measured 50 ms after stimulation. b, Summary: values for the variance of the evoked current normalized to the mean evoked current across all MC recordings. Note the much larger normalized variance values for Fmr1 KO. *p = 0.0030, Mann–Whitney U test. c, Subdividing the 10 MC recordings from Fmr1 KO mice as a function of the variance of the prestimulus current revealed that MCs with more sLLDs under baseline conditions (baseline variance >100 pA2) had greater variability in their evoked responses versus MCs with fewer sLLDs (baseline variance <100 pA2). *p = 0.038, Mann–Whitney U test. d, The mean magnitude of the evoked current per MC did not vary between Fmr1 KO and WT mice (ns = no significant difference).
To evaluate the effect of prior spontaneous current activity on the evoked current response, we measured the evoked current amplitude 50 ms following OSN stimulation across many stimulus trials. We chose a delayed time point in our analysis to capture the effect of Fmr1 KO on the LLD that underlies the large majority of the excitatory charge in MCs following OSN stimulation under most conditions (Najac et al., 2011; Gire et al., 2012; Vaaga and Westbrook, 2016). This showed one clear effect of KO, which was a decrease in the reliability of the current response to OSN stimulation. Whereas the current responses in MCs from WT mice were quite reproducible from trial to trial (see WT example in Fig. 5ai,aii), the responses in MCs from Fmr1 KO mice were highly variable between trials and appeared to depend on whether sLLDs just preceded OSN stimulation (see KO example in Fig. 5ai,aii). The reduction in the reliability of the MC current for Fmr1 KO versus WT mice was also reflected as a large, 10-fold increase in values for the variance of the evoked current response normalized to the mean (Fig. 5b; Fmr1 KO = 5.37 ± 2.26 pA, n = 10 cells in 5 mice; WT: 0.54 ± 0.21 pA, n = 9 cells in 4 mice; p = 0.0030 in Mann–Whitney U test). As further evidence that the decrease in reliability of the MC evoked current in KO was because of the sLLDs, we found that MCs with high current variance under baseline conditions (i.e., with more sLLDs; variance > 100 pA2) had larger variance/mean values for the evoked currents (8.12 ± 3.37 pA, n = 6 cells) versus MCs with lower baseline variance (variance < 100 pA2; 1.24 ± 0.65 pA, n = 4 cells; p = 0.038 in Mann–Whitney U test; Fig. 5c). Interestingly, we found that Fmr1 KO did not significantly alter the mean magnitude of the evoked current responses across MCs (Fmr1 KO = 143 ± 27 pA, n = 10 cells; WT: 173 ± 33 pA, n = 9 cells; p = 0.661 in Mann–Whitney U test; Fig. 5d). While the sLLDs in Fmr1 KO mice should naturally have reduced the magnitude of the mean evoked current for some of the individual MCs (see KO example in Fig. 5aii), other factors such as the large cell-to-cell variation in the mean evoked current magnitude likely obscured such an effect across the population of recordings.
In the present study, we did not further evaluate the effect of Fmr1 KO on all components of the evoked MC current response to OSN stimulation. As can be seen from the example traces in Figure 5a, MCs in both WT and Fmr1 KO mice displayed complex prolonged currents that included both faster components lasting a few hundred milliseconds as well as longer components that lasted more than a second. Furthermore, MCs from both WT and KO mice often displayed prominent fast, short-onset evoked currents that likely reflected direct OSN inputs (Najac et al., 2011; data not shown). Analysis of all of the evoked current components was beyond the scope of this study. The important point is that, when we focused on a time point (50 ms poststimulation) designed to capture the effect of Fmr1 KO on the stimulus-evoked LLD, the presence of sLLDs in Fmr1 KO mice greatly reduced the reliability of the MC response to OSN stimulation.
Fmr1 KO mice display deficits in fine odor discrimination
To ascertain the effects of Fmr1 KO on olfactory behavior, we compared the performance of WT and KO mice in a go/no-go operant discrimination paradigm in which mice were tasked with discriminating two odors, one that was associated with a water reward (S+) and a second (S–) that was not (Fig. 6a). In the regimen we used (see sequence of stimuli in Table 1), mice first were trained in the go/no-go paradigm wherein they were tasked with discriminating a training odor (1% isoamyl acetate, S+) from mineral oil (MO; S–), followed by discrimination of one of the test odors 2-heptanone (2-Hept, 1% in MO) from MO. Subsequently, mice discriminated various test odor pairs. Responses of a mouse were evaluated in four 20-trial blocks on a given day, with a response counted as correct either when the mouse licked in response to an S+ odor (“hit”) or when it did not lick in response to an S− odor (“correct rejection”). Mice were considered proficient in the discrimination of a test odor pair when they achieved ≥80% correct responses in three consecutive blocks on a given day and if they displayed proficiency on 2 consecutive days. The same seven WT and seven Fmr1 KO mice were used for all stimulus combinations.
Fmr1 KO does not alter discrimination of monomolecular odors. a, Go/no-go olfactory discrimination paradigm. ai, Mouse self-initiated trials by poking head into the odor delivery port, breaking a photodiode beam. aii, Sequence of steps. When the mouse entered the port, the odor valve opened with odor delivery being initiated 1–1.5 s later and lasting for 2.5 s. The last 2 s of odor delivery was the response window, when the licking of the mouse was assayed to determine whether it was discriminating a rewarded S+ odor from an unrewarded S– odor. The mouse received a water reward if it licked at least once during each of four 0.5-s blocks of the response window on delivery of the S+ odor. (aiii) A response was counted as correct (checkmark) if the mouse either licked in response to the S+ odor (Hit) or did not lick in response to S– (Correct rejection). b, Fmr1 KO and WT mice were similar at discriminating 2-heptanone (2-Hept) from 3-heptanone (3-Hept), as reflected in the fact that the mice took a similar number of days as WT mice to reach proficiency (≥80% correct responses; bi; ns = no significant difference) and displayed similar correct response rates on reaching proficiency (bii). Data in bii reflect correct response rates for each of four consecutive blocks on 2 d in which proficient performance was displayed. *p = 0.0115 in Mann–Whitney U test for block 3. The correct response rates for individual mice are superimposed on the bar graphs, color-coded for each mouse. c, Fmr1 KO and WT mice were similar at discriminating ethyl acetate (EA) from propyl acetate (PA). d, For EA versus PA discrimination, an additional day of experiments was added in which the S+ and S– odors were reversed after the third of five blocks. Reversal resulted in a large decrease in correct response rate. *p = 0.0020 in Wilcoxon signed-rank test for Fmr1 KO comparing block 3 with both blocks 4 and 5 (n = 6, including 3 WT and 3 KO mice).
In the testing phase, we first compared the ability of WT and Fmr1 KO mice to discriminate the structurally similar monomolecular odors, 2-heptanone (2-Hept; 1%, S+) and 3-heptanone (3-Hept, 1%, S−). For this pair, we found that WT and KO mice displayed a similar level of discrimination, with each mouse type taking a similar number of days to reach proficiency (Fig. 6bi; p > 0.9999 in Mann–Whitney U test) and generally displaying similar correct response rates across the four trial blocks on reaching proficiency (Fig. 6bii; see Table 2 for individual p values). The one exception were values for the correct response rates for block 3, when KO mice performed slightly better than WT (p = 0.0115 with Mann–Whitney U test). Similar results were obtained when the test odor pairs were ethyl acetate (EA; 1%, S+) and propyl acetate (PA; 1%, S–), both in terms of the number of days to proficiency (Fig. 6ci; p = 0.706 in Mann–Whitney U test) and correct response rates on reaching proficiency (Fig. 6cii; Table 2). For the task involving EA/PA discrimination, an additional day of experiments was added in which the rewarded S+ and unrewarded S− odors were reversed after the third of five blocks (Fig. 6d). Reversal resulted in a large decrease in the correct response rate (p = 0.0020 in Wilcoxon signed-rank test comparing block three and block 4; n = 6, combining three WT and three KO mice). This argued that WT and Fmr1 KO mice were not relying on other cues associated with the S+ and S− stimuli such as auditory cues from the opening of the water reward valve. Taken together, these results imply that Fmr1 KO does not significantly impact the ability of mice to discriminate some monomolecular odors.
Contrasting results were obtained when the odor discrimination task was made more difficult by having mice discriminate monomolecular odors from odor mixtures. The experiments with mixtures were conducted immediately after the testing phase involving discrimination of monomolecular odors. When mice were tasked with discriminating either 2-Hept (1%, S+) from a 1:1 mixture of 2-Hept and 3-Hept (0.5% for both, S−) or EA (1%, S+) from a 1:1 EA/PA mixture (0.5% for both, S−), WT but not Fmr1 KO mice were successful in discriminating the stimuli (Fig. 7a,b). Correct response rates for Fmr1 KO mice were much lower than for WT mice for nearly all blocks for both types of mixture experiments (see p values in Table 2), remaining near 50% chance level. The poorer performance of KO mice in the mixture experiments occurred despite the fact that Fmr1 KO mice were generally given one additional day of testing versus WT mice (mean days of testing for the two mixture experiments: 2.3 ± 0.2 d for WT, n = 14; 3.0 ± 0.0 d for Fmr1 KO, n = 14; p < 0.0001 in Mann–Whitney U test). Importantly, in the mixture experiments involving EA/PA, we found that Fmr1 KO mice, after just failing at discriminating EA from EA/PA mixture, were successful at discriminating EA (1%, S+) from EA (1%, S−) on an additional day of retesting (Fig. 7c; p = 0.0313 in Wilcoxon signed-rank test comparing block four of retest day with block four of last day of mixture experiment, n = 4 KO mice). This argued that the impaired performance of KO mice in discriminating the monomolecular odors from the mixtures was not because of fatigue or loss of interest in the task. We also evaluated whether the incorrect responses for Fmr1 KO animals in the mixture experiments occurred more often in S+ versus S− trials, finding many more errors for S− trials (Fig. 7d; p < 0.0001 in Mann–Whitney U test for blocks 2, 3, and 4). The higher error rate in the S− trials, when mice licked in response to the unrewarded odor, would be expected because of the water-deprived state of the mice.
Statistical comparisons of performance of WT and Fmr1 KO mice in go/no-go discrimination
Fmr1 KO mice display deficits in more difficult discrimination tasks involving odor mixtures. a, Correct response rates when mice were discriminating 2-Hept from a 1:1 mixture of 2-Hept and 3-Hept. Note that, except for the first block, WT mice performed much better than Fmr1 KO mice. *p < 0.0001 in Mann–Whitney U test for blocks 2, 3, and 4. Data for both WT and Fmr1 KO mice reflect the last 2 d in which experiments were performed with this odor combination; Fmr1 KO mice were generally given 1 additional day beyond what WT mice took to reach proficiency (see main text). b, Fmr1 KO mice performed worse than WT mice in discrimination of EA from a 1:1 mixture of EA and PA. *p < 0.0001 in Mann–Whitney U test for all blocks. c, Fmr1 KO mice that failed to discriminate EA from the EA/PA mixture were successful at discriminating EA from PA on retesting. Data reflect four KO mice for block 4 on the last day of testing EA versus EA/PA mixture discrimination (left bar) and four consecutive blocks on a subsequent day of retesting with EA versus PA. *p = 0.0313 in Wilcoxon signed-rank test comparing block 4 of the 2 d. d, Most errors made by Fmr1 KO mice were during S– trials (“false alarms” in Fig. 6aiii) rather than S+ trials. The plotted data reflect the number of errors made in the half (10) of the 20 trials in each block that were either S– or S+. *p < 0.0001 in Mann–Whitney U test for blocks 2, 3, and 4. e, Evidence that impaired odor discrimination in Fmr1 KO mice was not because of cognitive impairments. In the training phase involving discrimination of isoamyl acetate (IA) from mineral oil (MO), KO mice took a similar number of days to reach proficiency (ei; ns = no significant difference) and displayed similar correct response rates on reaching proficiency (eii). Data in eii reflect the last 2 d of experiments with the IA/MO pairs.
One interpretation of the impaired performance of Fmr1 KO mice in the discrimination experiments involving odor mixtures is that KO mice have olfactory sensory processing deficits, but we also considered whether the observed results could have been explained by general cognitive or motor deficits in KO mice. While these latter mechanisms were difficult to exclude completely, evidence against them being important in our experiments was the fact that Fmr1 KO mice performed just as well as WT mice in monomolecular odor discrimination (Fig. 6b,c). In addition, when we analyzed the performance of KO mice during the initial training phase of the experiments, involving discrimination of isoamyl acetate from MO, we found no differences: Fmr1 KO and WT mice took a similar number of days to reach proficiency (Fig. 7ei; p = 0.0813 in Mann–Whitney U test) and displayed similar correct response rates on reaching proficiency (Fig. 7eii; see p values in Table 2). These results suggest that the reduced performance of Fmr1 KO mice in the go/no-go discrimination experiments involving odor mixtures was because of sensory processing deficits.
Discussion
In this study, we examined olfactory circuit and behavioral properties of the Fmr1 KO mouse model of FXS. Our main findings were that KO induced a large increase in the excitation of bulbar MCs under baseline conditions, i.e., in the absence of a stimulus, and that this was likely at least in part because of reduced inhibitory connections onto a different neuron type, the eTCs. The enhanced baseline excitatory responses were also found to reduce the reliability of sensory-evoked responses in MCs. In addition, in behavioral studies, we found that Fmr1 KO impaired fine odor discrimination capabilities.
Deficits in olfactory bulb circuitry because of Fmr1 KO
Evidence for disrupted olfactory bulb circuit properties was obtained first in whole-cell patch-clamp recordings from MCs. MCs from KO mice displayed a massive increase in spontaneous, long-lasting excitatory current events that were largely not present in WT mice. The events, which originated in glomeruli, also were associated with prolonged depolarizations and an increase in spontaneous action potential firing. Several lines of evidence suggest that the excitatory events were spontaneous versions of the well-characterized LLD events that dominate the MC response to stimulation of OSNs (hence, sLLDs; (Carlson et al., 2000; Schoppa and Westbrook, 2001; Vaaga and Westbrook, 2016). These included the slow time course of the spontaneous events, their glomerular origin, their modulation by GABAA receptor antagonists, as well as the fact that the spontaneous currents occluded generation of stimulus-evoked LLDs. One prior study (De Saint Jan et al., 2009) reported spontaneous long-lasting excitatory events in MCs of WT mice, different from our results indicating that such activity was uncommon in control mice. It is unclear what these different results reflect. The important point is that we observed a large increase in sLLDs in Fmr1 KO versus WT mice when we performed measurements under identical experimental conditions.
LLDs are network-level events that occur probabilistically depending on the balance between excitation and GABAergic inhibition within a glomerulus (Carlson et al., 2000; Schoppa and Westbrook, 2001; Gire and Schoppa, 2009). Here, we were able to identify one likely mechanism whereby Fmr1 KO increased the probability of sLLDs in MCs (see model in Fig. 4a). Based on recordings on eTCs, it appears that KO caused both a reduction in inhibitory inputs onto eTCs as well as a large increase in the excitation of eTCs. eTCs are glutamatergic neurons in the bulb that provide feedforward excitation onto MCs within glomeruli (De Saint Jan et al., 2009; Najac et al., 2011; Gire et al., 2012), and so an enhancement in spontaneous eTC excitation should naturally lead to more feedforward excitation being transmitted to MCs. Our results providing evidence that Fmr1 KO reduces GABAergic inhibition fit with prior circuit-level effects that have been observed in Fmr1 KO mice in other brain regions (Cea-Del Rio and Huntsman, 2014; Takarae and Sweeney, 2017; Telias, 2019), as well as the reported effects of dFMRP deletion in the insect analog of the olfactory bulb, the antennal lobe (Gatto et al., 2014; Franco et al., 2017). Alterations in the GABAergic system have also been reported for human patients with FXS (D'Hulst et al., 2006, 2015; Morin-Parent et al., 2019).
Our studies certainly do not exclude some other mechanisms by which Fmr1 KO could have enhanced sLLDs in MCs. For example, one other target of Fmr1 KO in other systems are channels that mediate persistent sodium currents, where increases in these currents have been shown (Deng and Klyachko, 2016). Within eTCs, persistent sodium currents have been shown to play a critical role in regulating eTC excitation (Liu and Shipley, 2008); thus, enhancement of these currents could increase feedforward signaling from eTCs to MCs. Another possibility for the enhanced sLLDs in MCs is a reduction in some inhibitory connections onto MCs themselves. We found, in recordings of sIPSCs in MCs in glutamate receptor blockers (Fig. 3a,b), that there were no differences in sIPSC frequency and amplitude between Fmr1 KO and WT. While this result argued that most inhibitory connections onto MCs are not altered in KO, an analysis of all sIPSCs may have missed changes in some cell-specific inputs onto MCs, for example those derived from PG cells onto apical dendrites that likely would have the strongest impact on LLDs (Gire and Schoppa, 2009).
In prior studies, there has been some limited attention paid to the cellular effects caused by Fmr1 KO in the mouse olfactory bulb. For example, Scotto-Lomassese and colleagues (Scotto-Lomassese et al., 2011) reported that Fmr1 KO caused moderate increases in the number and length of dendritic spines on adult-born GABAergic granule cells. However, that study did not examine the physiological implications of those changes nor did they examine changes in other cell-types, making it difficult to compare our results to theirs. Our studies may perhaps be better compared with the prior results of Geramita and co-workers (Geramita et al., 2020) who examined olfactory bulb properties in two other models of ASDs (Cntnap2-KO and Shank3-KO mice). Using calcium imaging methods, they reported that the mutant mice displayed reductions in the amplitude and reliability of calcium responses within glomeruli to odor stimulation. In our studies, we also observed that Fmr1 KO reduced the reliability of sensory-evoked excitatory currents in MCs (Fig. 5). The fact that reductions in response reliability in the olfactory bulb have been observed across ASD models suggests a potential common mechanism by which different ASDs alter olfactory-related behavior (see below). Interestingly, reductions in the reliability of sensory-evoked responses have also been recently observed across multiple sensory systems in studies of human patients with ASDs (Dinstein et al., 2012; Haigh et al., 2015).
Deficits in olfactory behavior because of Fmr1 KO mice
In our analysis of olfactory behavior, we found that Fmr1 KO mice displayed significant deficits in a go/no-go olfactory discrimination task in which the mice had to distinguish a monomolecular odor from an odor mixture. In contrast, KO mice performed as well as WT mice in an easier task involving discrimination of two similar monomolecular odors. In the latter experiments, the mice required a similar number of days to reach proficiency after initiating the task and displayed similar levels of proficiency. These results, taken together, suggested that Fmr1 KO mice are generally quite good at discriminating odors, only showing impairments for difficult tasks. In addition, the results with monomolecular odor discrimination serve as a reasonable control for other possible explanations for the impaired performance in fine odor discrimination task for KO mice other than sensory deficits. For example, if the impaired performance in the mixture experiments were because of cognitive or motor deficits, the impairments likely would have appeared in the easier discrimination task.
Only a few prior studies have examined olfactory behavioral deficits in Fmr1 KO mice, with somewhat conflicting results. Schilit Nitenson and co-workers (Schilit Nitenson et al., 2015) provided evidence, using an olfactory cross-habituation task, that KO reduces olfactory sensitivity, although they did not find impairments in olfactory discrimination. In contrast, Larson and co-workers (Larson et al., 2008) found impaired olfactory discrimination in KO mice with the go/no-go task, both slower learning and increases in failure to discriminate, but no differences in olfactory sensitivity. Certainly at least some of differences in these results with respect to odor discrimination may have been explained by the different behavioral tests, but the results of our behavioral studies here also point to the importance of task difficulty, i.e., the odor pairs used, in what results might be expected in experiments that assess odor discrimination deficits. Task difficulty was not systematically accounted for in the prior studies.
How might the apparent deficits in odor discrimination in Fmr1 KO mice be related to the observed circuit-level effects of KO, including the dramatic increase in spontaneous LLDs? One model builds on our observation that the enhanced sLLDs reduced the reliability of the MC response to OSN stimulation. If we assume that an animal discriminates odors based on the differences in which subpopulations of MCs are activated by those odors (Yokoi et al., 1995), a decrease in reliability of the responses of any of the subpopulations should make accurate detection of those differences more difficult. Indeed, in other sensory systems, it has been established that the reliability of neural responses is highly predictive of behavior when difficult perceptual decisions are being made (Logothetis and Schall, 1989; Britten et al., 1996; Cook and Maunsell, 2002). Also, an increase in neural variability has been proposed to be a mechanism for deficits in sensory perception in ASDs (Dakin and Frith, 2005; Simmons et al., 2009). There are of course other explanations as well for the impaired performance of Fmr1 KO mice in go-no go olfactory discrimination. For example, the fact that MCs are more likely to undergo spontaneous excitation could mean that they would be more responsive to weak stimuli, an effect that would result in the broadening of the tuning curves of MCs. Broader neural tuning has already been implicated as being at least correlated with reductions in odor discrimination capabilities in a Drosophila model of FXS (Franco et al., 2017). Odor discrimination capability has also been correlated with odor-evoked γ frequency (40–100 Hz) oscillations in the olfactory bulb (Stopfer et al., 1997; Beshel et al., 2007; Lepousez and Lledo, 2013; Kay, 2015; Losacco et al., 2020). In the present study, we did not test directly whether Fmr1 KO mice have altered γ oscillations, but we did find that KO increased transient inhibitory synaptic activity in MCs that underlie these oscillations (when sIPSC recordings were performed in the absence of GluR blockers, Fig. 3c–e; Bazhenov et al., 2001; Galán et al., 2006; Schoppa, 2006).
A final point that should be made about our behavioral studies is that some care must be taken in interpreting the differences in the performance in the experiments involving mixtures versus those involving just monomolecular odors. Given the structure of our behavioral experiments, we cannot exclude the possibility that the reduced performance of KO mice in the mixture experiments reflected the fact they were overly cuing their response to the presence of the S+ odor. In the mixture experiments, the S+ stimulus was the S+ monomolecular odor used in the prior monomolecular discrimination phase of the studies, while the S− mixture included the same S+ odor plus a second, similar odor. If KO mice were cuing their response to the presence of the S+ odor, their impaired performance might more reflect deficiencies in generalizing between stimuli (Li et al., 2023) rather than deficits in odor discrimination per se.
Footnotes
This work was supported by National Institutes of Health Grants R01DC006640 (to N.E.S.) and F31DC017350 (to S.T.J.). We thank Dr. Diego Restrepo (University of Colorado Anschutz Medical Campus) and personnel in his lab for advice and assistance in training Dr. Kuruppath in the go/no-go olfactory discrimination behavioral task.
The authors declare no competing financial interests.
- Correspondence should be addressed to Nathan E. Schoppa at nathan.schoppa{at}cuanschutz.edu