Delayed looming escape responses

To investigate whether subcortical visuomotor transformations could be affected by genetic mutations associated with ASD, we evaluated the behavioural responses to the innate LER paradigm. LER is largely independent of cortical input, requiring the SC and its downstream targets to initiate appropriate responses. For this purpose, we tested a genetic haploinsufficiency ASD mouse model affecting the SET-domain containing 5 (Setd5) [23], a histone-associated protein found in approximately 1% of patients with intellectual disability and ASD [23]. We subjected mutants (i.e., Setd5^+/−) and their wild-type (WT) siblings to our LER paradigm, where mice would automatically trigger a sequence of 5 consecutive looms upon entering a threat zone (Fig 1A and 1B; see Materials and methods). For all experiments, we only used animals without any craniofacial abnormalities, tooth displacement, or eye abnormalities, which are described to sometimes occur in our Setd5 mutant animals [23]. As previously reported, Setd5^+/− animals were, on average, slightly smaller than their WT siblings [23]. Please note that other models for the same gene show more drastic abnormalities not observed in our hands [28]. During the acclimatisation period, all animals showed similar exploratory behaviour and dynamics (S1A–S1G Fig), and seemingly similar behavioural responses, escaping to the shelter upon stimulus presentation (S1 Video). However, behavioural differences between the genotypes became evident when observing their reaction times (Fig 1C). While WT siblings responded robustly to the first stimulus presentation, their heterozygous siblings required, on average, more repetitions to elicit an escape response (Fig 1D), triggered more looms (Fig 1E), made more shelter exits (Fig 1F), had longer reaction times, and escaped with less vigour, reaching lower maximum speeds [29] (Fig 1G). The differences across genotypes were independent of the location, speed of the animal at stimulus onset or heading angle (S1H–S1J Fig), or sex (Fig 1C–1J, females; and S1L–S1N Fig, males). The difference in response reaction time is an intrinsic and not an adaptive property, since it is present from the first loom presentation the animal encounters (Fig 1H and 1G, left). However, the escape vigour of the Setd5^+/− animals during these first-ever encountered loom presentations was indistinguishable from their WT siblings (Fig 1H, right). This indicates that the escape delay differences are not due to intrinsic motor problems but emerge due to perceptual decision-making deficits—the process of extracting and integrating sensory information to initiate a subsequent action. These results highlight the LER as a robust behavioural assay to probe perceptual impairments and their underlying circuit dysfunctions in visuomotor processing in ASD.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. ASD mice exhibit delayed and less vigorous looming escape responses.

(A) LER paradigm showing the shelter’s location and the threat zone (left) and LER example (right). (B) Paradigm schematic. Day 0 (D0) was used for acclimatisation. D1-D5 consisted of an acclimatisation period (grey) followed by 3 LER tests (red). The looming stimulus consisted of 5 consecutive looms (right). (C) Raster plot of mouse speed during LER (white, dotted vertical lines denote the start of each loom; solid white line denotes the end of the stimulus) for Setd5^+/+ (upper, n = 14, 60 trials) and Setd5^+/− (lower, n = 14, 198 trials), sorted by reaction time. Bottom, distribution of reaction times for all Setd5^+/+ (black) and Setd5^+/− (red) trials (p < 0.001, two-sample Kolmogorov–Smirnov test). (D) Left, example trials based on whether the mouse responds within 1 of the 5-loom stimuli, after the fifth (>5), or not at all (NR, no response) from one Setd5^+/− mouse. Grey shaded areas represent the frames used to calculate the speed at stimulus onset (S_at) and the immediate response speed (S_im). Right, proportion of escapes to loom presentations. (E) Total looms triggered across the 5 test days (Setd5^+/+, 4.36 looms; Setd5^+/−, 16 looms, P = 0.013). (F) Total shelter exits across the 5 test days (Setd5^+/+, 6.0; Setd5^+/−, 24.5, P = 0.019). (G) Average reaction time (left) and maximum escape speed (right) per animal (reaction time, P = 0.001; max. escape speed, P < 0.001). (H) As (G), but only for the very first loom presentation (reaction time, P = 0.046; max. escape speed, P = 0.166). (I) Average immediate speed change following the stimulus presentation for all trials where the mice escape within the first loom presentation (left, Setd5^+/+, n = 14, black, p < 0.001; right, Setd5^+/−, n = 14, red, p < 0.001). S_at is the mean speed of the animal ±50 ms of stimulus onset, and S_im is the mean speed of the animal 300–800 ms after stimulus onset. (J) As (I) but for trials where the mice escape during or after the second loom (left, Setd5^+/+, n = 3, black, p = 0.007; right, Setd5^+/−, n = 11, red, p < 0.001). (K) Proportion of response types per genotype (X² = 103.9, p < 0.001, X² test of independence). P-values: Wilcoxon’s test, p-values: paired t test, unless specified. The data underlying this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3002668.g001

To test whether the delay in initiating an escape to the loom was due to the mutant mice simply detecting the stimulus later than their WT siblings, we examined the immediate change in the speed of the animals from the time of the first loom, selecting only the trials where the mice performed an escape to the shelter at any point during the looming events. For trials where the animals generated an escape within the first loom presentation, both WT and mutant animals increased their speed significantly upon stimulus onset (Fig 1I). For trials where the mice did not generate an escape within the first loom, which was the majority of trials for mutant animals but only a small fraction of trials for WT animals (Fig 1D), the mice significantly reduced their speed in the time immediately following the stimulus onset (Fig 1J). This demonstrates that Setd5^+/− animals detect and respond to the stimulus within a similar time frame as their WT siblings but preferentially perform a locomotor arrest instead of an escape response (see S1–S3 Videos). This arrest behaviour is more reminiscent of risk assessment behaviour as previously shown for loom stimuli of different contrast [29] rather than the defensive freezing response characterised by the sustained cessation of all movement, since the animals were still performing small movements of their head and upper body. Indicating that the animals are using this time to perform ongoing threat evaluation. This suggests that the delay in LER arises from difficulties in either evaluating the threat level of the stimulus or initiating an appropriate response.

Given that cortical malfunctions have been suggested as the cause of sensory differences in autism [15–17] and that the visual cortex has been shown to modulate the response magnitude of looming sensitive cells in the SC [30], we tested whether the behavioural responses are affected in cortex and hippocampus-specific conditional Setd5 animals (Setd5^+/fl; Emx1-Cre, S2 Fig). These animals showed no behavioural differences in their reaction times and vigour, or response kinetics (S2C–S2F Fig), in line with previous studies that show that subcortical pathways [29,31], namely, the SC and PAG, and not altered top-down modulation from cortical areas, are required for this behaviour.

Repetitive looming escape responses

Although the innate LER does not require learning, repeated loom presentations cause adaptive behavioural changes, for example, the emergence of place avoidance of the threat zone [29]. Given that Setd5^+/− mice triggered 3 times more looming events and shelter exits than their WT siblings (Fig 1E and 1F), we explored if altered adaptation to repeated presentations could account for the observed difference in average reaction time and vigour (Fig 1G). For that purpose, we compared the intrinsic behavioural characteristics and the effect of the LER on the exploration strategies and behavioural adaptations across days (Fig 2A–2D). Before stimulus exposure, both cohorts had similar exploratory strategies (S1A–S1G Fig), indicating that their innate exploration strategies, and, thus, intrinsic levels of anxiety, cannot account for the differences in reaction time and vigour. After stimulus exposure, WT animals showed expected reductions in their exploratory behaviour following the initial exposure to the looming stimulus, making fewer exits than during the prestimulus exploration time (Fig 2B), eliciting fewer looms in total (Fig 2C) with relatively constant reaction time (Fig 2D). Strikingly, Setd5^+/− consistently triggered more looms and had no signs of sensitisation upon repeated exposures (Fig 2A–2D). After the initial decrease in exploration following the first stimulus presentation, they showed no further consolidation of place avoidance to the threat zone. The reaction time and vigour of the escape response remained longer and slower across days (Fig 2D), respectively, indicating that Setd5^+/− do not sensitise as their WT siblings do to the LER paradigm. We next tested if the delayed perceptual decisions (Fig 1) are related to the consolidation of place avoidance, 2 seemingly independent behaviours. We analysed the total number of shelter exits for each animal and compared them with their average reaction times and vigour (Fig 2E and 2F). Surprisingly, the strength of the place avoidance was a strong predictor of the reaction times, but not vigour, suggesting that the timing of the perceptual decision and the formation of place avoidance are intrinsically linked. To assess the limits of response adaptation and to ascertain if the presence of a reward would overcome the fear, we conducted the same behavioural experiment but with the presence of a food reward on the far side of the arena within the threat zone (Fig 2G–2J) and with no interstimulus interval restriction. Although WT siblings did make some attempts to retrieve the food reward, they quickly sensitised, rarely leaving the shelter and were unsuccessful in obtaining the food reward. Setd5^+/−, on the other hand, increased their shelter-leaving events, often persisting until reaching the reward (Fig 2G–2I and S2 Video). Despite the rapid and repeated exposure, Setd5^+/− mice continued to escape to the shelter upon stimulus presentation, for >10 consecutive presentations (Fig 2H–2J). Their reaction times showed a mild adaptation (Fig 2J, top), while vigour remained largely constant (Fig 2J, bottom). These results show that 2 behavioural traits, LER and place avoidance, strongly correlate, suggesting they share a common neuronal pathway.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Altered adaptation and repetitive behavioural phenotype to the LER.

(A) Left, graphic depicting exits where the mouse enters (dotted line, filled dot) or does not enter (dashed line, open dot) the threat zone. Right, ethogram of exploratory shelter exit behaviour during the prestimulus acclimatisation, the first and last test of the LER paradigm. Each row represents one animal. (B-D) Adaptation in the number of shelter exits, average number of looms triggered and reaction times across days (B, Setd5^+/+: p = 0.2189, Setd5^+/−: p = 0.4974; C, Setd5^+/+, p = 0.0087; Setd5^+/−, p = 0.2192; D, Setd5^+/+, p = 0.890; Setd5^+/−, p < 0.001). (E) Relationship between the number of shelter exits and the average reaction time per animal (Setd5^+/+, p = 0.626; Setd5^+/−, p = 0.002). (F) As (E) but for maximum escape speed per animal (Setd5^+/+, p = 0.547; Setd5^+/−, p = 0.053). (G) Left, reward trial example showing the location of the food reward within the threat zone. Right, example trajectories during the reward trials show the mouse’s position for the 3 s before triggering the loom (light grey) and the 6 s following the stimulus start (black or red). Filled dots represent the position of the mouse when the stimulus was triggered, grey square represents the shelter, and the yellow star shows the position of the food reward. (H) Number of looms triggered during the reward trial (Setd5^+/+, 1 bout; Setd5^+/−, 8 bouts, P = 0.005). (I) Ethograms of exits, as in (A), during the reward trials show an increased probability of Setd5^+/− mice leaving the shelter during a trial (number of exits, 2.12 for Setd5^+/+, 6.88 for Setd5^+/−, p = 0.057). (J) Reaction time (top panel, Setd5^+/−, 44 trials, r = 0.828, p = 0.0001) and escape vigour (bottom panel, Setd5^+/−, 44 trials, r = 0.184, p = 0.5116) during repeated presentations of the loom. Trials when the animal was interacting with the reward were excluded. P-values: Wilcoxon’s test, p-values: Pearson’s correlation test, unless specified. Plotted linear fits depict the statistically significant correlations. The data underlying this figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3002668.g002

Convergent behavioural deficits across ASD models

To test whether the deficiencies observed are specific to Setd5^+/− animals or might reveal a more general behavioural phenomenon across ASD models with distinct molecular and genetic aetiologies, we decided to test 2 additional highly penetrant mutations, Cullin3 (Cul3), a ubiquitin ligase-encoding gene, and Ptchd1 (S3A–S3C Fig), a member of the Patched family speculated to function as a Sonic hedgehog receptor [25]. We exposed sex-matched sibling pairs from these models to the same LER protocol as used for the Setd5 model and observed qualitatively similar behavioural deficits in reaction time (Ptchd1: Fig 3A–3D; Cul3: Fig 3I–3L) and place avoidance (Ptchd1: Fig 3E–3G; Cul3: Fig 3I–3L). Both cohorts required, on average, more repetitions to elicit an escape response (Fig 3A, 3B, 3I, and 3J), had longer reaction times between the stimulus start and the maximum speed (Fig 3C and 3K), displayed slower maximum speed compared to their WT siblings (Fig 3D and 3L), triggering more looms (Fig 3E and 3M) and leaving the shelter more often (Fig 3F and 3N), but had no differences in their behavioural strategies prior to the first loom presentation (Figs 3G–3O, S3D, and S3K). Similar to Setd5, these changes were not due to the animals not reacting to the first stimulus, as seen in their consistent arrest behaviour (S3E and S3L Fig), nor to differences in their exploratory behaviour (S3D and S3K Fig). This difference in response reaction time is an intrinsic and not an adaptive property since it is present during the first loom presentation (S3G and S3N Fig). However, they varied in strength, with Cul3^+/− mice showing a less severe phenotype than the Setd5^+/− or Ptchd1^Y/− models. Interestingly, slower reaction times and reduced vigour could also be observed in Ptchd^Y/− animals, despite their baseline hyperactivity (S3D and S3I Fig). Interestingly, in both cohorts, the strength of the place avoidance and the reaction times were significantly lower and longer, respectively (Fig 3C, 3F, 3K and 3N), being linearly correlated in Ptchd1^Y/−, but not Cul3^+/− (Fig 3H and 3P). These results suggest a general link between the timing of the perceptual decision and the formation of place avoidance, despite different molecular and developmental origins.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Conserved behavioural changes across etiologically distinct autism models.

(A) Raster plot of mouse speed in response to the looming stimuli for Ptchd1^Y/+ (upper, n = 9, 49 trials) and Ptchd1^Y/− (lower, n = 9, 259 trials), sorted by reaction time. Bottom, distribution of reaction times for all Ptchd1^Y/+ (black) and Ptchd1^Y/− (orange) trials. (B) Left, example trials based on whether the mouse responds within 1 of the 5-loom stimuli, after the fifth (>5), or not at all (NR, no response) from 1 Ptchd1^Y/− mouse. Right, proportion of escape to loom presentations. Trials where mice escaped within the first loom: Ptchd1^Y/+, 0.826, Ptchd1^Y/−, 0.374, p = 0.004. (C) Average reaction time and (D) maximum escape speed per animal, for all trials where the mice escaped (reaction time; Ptchd1^Y/+, 0.411 s, Ptchd1^Y/−, 1.41 s, p = 0.002; maximum escape speed; Ptchd1^Y/+, 69.1 cm s⁻¹, Ptchd1^Y/−, 54.2 cm s⁻¹, p = 0.005). (E) Average total looms triggered per genotype across 5 days of testing (Ptchd1^Y/+, 5.44 looms; Ptchd1^Y/−, 28.8 looms, P = 0.004). (F) Average total shelter exits across the 5 test days (Ptchd1^Y/+, 15; Ptchd1^Y/−, 53, P = 0.025). (G) Ethogram of shelter exits during the prestimulus acclimatisation, first and last trial of the LER paradigm. Each row represents one animal, filled and open dots represent exits that crossed into the threat zone or not, respectively. (H) Relationship between the number of shelter exits and the average reaction time per animal (Ptchd1^Y/+, p = 0.126; Ptchd1^Y/−, r = 0.76, p = 0.012, Pearson’s correlation). (I-P) Same as (A-H) but for Cul3. (I) Cul3^+/+ (top, n = 10, 61 trials) and Cul3^+/− (bottom, n = 10, 112 trials). Bottom, distribution of reaction times (p < 0.001, two-way KS test). (J) Cul3^+/+, 0.892, Cul3^+/−, 0.347, p < 0.001, two-way KS test). (K) Reaction time; Cul3^+/+, 0.464 s, Cul3^+/−, 1.11 s, P < 0.001. (I) Maximum escape speed; Cul3^+/+, 62.7 cm s⁻¹, Cul3^+/−, 46.8 cm s⁻¹, P < 0.001). (M) (Cul3^+/+, 5.60 looms; Cul3^+/−, 12.9 looms, P = 0.023). (N) (Cul3^+/+, 47; Cul3^+/−, 89, P = 0.006). (P) (Cul3^+/+, p = 0.078; Cul3^+/−, p = 0.571, Pearson’s correlation). Box-and-whisker plots show median, IQR, and range. Shaded areas represent SEM. Lines are shaded areas, mean ± SEM, respectively. P-values are Wilcoxon’s test, p-values: two-sample Kolmogorov–Smirnov test, unless specified. The data underlying this figure can be found in S3 Data.

https://doi.org/10.1371/journal.pbio.3002668.g003

Deep medial SC, not retinal drive, is linked to delay in LER

Changes in average reaction times have previously been shown to depend on the saliency of the threat stimulus. Whereas repeated high-contrast looming stimuli evoke strong LER, low-contrast looms drive less vigorous escapes and longer reaction times [29], suggesting that deficits in visual processing could underlie the behavioural differences observed in Setd5^+/−. This possibility is aggravated by the fact that the expression of genes involved in eye development are known to be misregulated in Setd5^+/− animals [23]. To explore possible visual response changes, we recorded visually evoked responses across different layers of the SC, using 32-channel silicon probes in head-fixed, awake-behaving animals (S4A Fig), and determined the recording depth using current source density analysis (S4B and S4C Fig). In our setup, mutant and control animals had similar running probabilities and showed similar baseline firing properties (S4D Fig). We next assessed visual response properties to full-field flashes, visual stimuli that drive all neurons similarly, irrespective of their receptive field positions. First, we show that pupil dynamics are unaffected across genotypes (S4E Fig). Next, when clustering visually responsive cells into 10 clearly defined groups, we could observe similar proportions and identical response properties (S4F and S4G Fig). Identical responses were also observed in their spatiotemporal receptive fields and the proportion of ON and OFF-responsive cells (S4A–S4G Fig). Direction-selective responses (S4H and S4I Fig) were also indistinguishable across groups. Finally, we explored the response properties to looming stimuli across SC layers. Looms elicit robust and reproducible firing across genotypes (S4J Fig), with similar firing, peak rates, and time-to-peak responses across genotypes (S4K–S4M Fig). The overall response kinetics (S4N Fig) and the adaptation in average firing rates across presentations (S4O and S4P Fig) were indistinguishable across genotypes. Similarly, no differences were observed when comparing the visual response properties of Cul3 and Ptchd1 models (S5 Fig). The lack of any reportable visual response difference across the SC layers indicates that retino-colliculuar visual processing is largely unaltered and excludes the possibility that sensory processing can account for the observed visuomotor deficits and suggests that the main cause of the behavioural difference may reside in downstream areas. Given that optogenetic stimulation of deep medial SC (dmSC) neurons that project to the dPAG generates robust escape responses and that the dPAG encodes for the choice of escape and controls the escape vigour [29], we hypothesised that activating dmSC neurons would reveal behavioural differences between genotypes and delineate the underlying circuit impairments. To test this, we focused on the Setd5 mouse line and used unilateral in vivo channelrhodopsin-2 (ChR2) activation of VGluT2⁺ cells in the dmSC (Fig 4A–4C). In line with previous findings, activation of dmSC neurons in Setd5^+/+ animals instructed immediate shelter-directed escapes (Fig 4D) and were absent in controls (S6 Fig). Gradually increasing the laser power and, thus, optogenetic activation, led to corresponding decreases in the latency to respond and an increase in the escape vigour in Setd5^+/+ (Fig 4C). By contrast, activation of dmSC neurons in the Setd5^+/− background more often resulted in an initial arrest, followed by shelter-seeking escape responses (Fig 4F and 4G), mimicking the behavioural divergence observed for LER (Fig 1). Interestingly, this behavioural difference was evident only during strong optogenetic activations. At lower laser intensities, both genotypes exhibited a transition from arrest to weak escape behaviours (Fig 4H) with similarly low reaction time (Fig 4E and 4G) and vigour (Fig 4H). Correspondingly, the escape probabilities were identical at low optogenetic stimulation but diverged with increased laser power (Fig 4I). Given that changes in the strength of optogenetic activation mimic LER to looms of different contrast [29], we hypothesised that when probing animals to LER of different contrasts, both genotypes should show similar responses at lower contrast (Fig 4J). To test this, we performed the LER at 3 different contrast levels: 98%, 50%, and 20% (Fig 4K and 4L). As predicted, the LER to low-contrast looms did not differ across genotypes, showing similar reaction times (Fig 4K). Remarkably, in Setd5^+/− animals, the reaction time remained largely constant across contrast levels (Fig 4F and 4G), suggesting that the time required to initiate a response is independent of stimulus intensity. Together with the persistence of the arrest behaviour even at high levels of optogenetic activation (Fig 4H), this supports the idea that there is a functional bottleneck downstream of the sensory processing circuits within the SC that impairs the timely generation of escape responses.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Activation of deep SC neurons recapitulates delayed LER in Setd5^+/− animals.

(A) Timeline of the experimental protocol for optogenetic activation of the dmSC. (B) Video frame during an optogenetics trial. (C) Confocal micrograph of AAV-ChR2 expression in dmSC and optic fibre location reconstruction, scale bar: 200 μm. (D) Raster plot of mouse speed in response to optogenetic activation, sorted by laser intensity and (E) mean speed responses at increasing laser intensities for Setd5^+/+ (n = 4, 312 trials). (F, G) As (D, E) but for Setd5^+/− mice (n = 4, 291 trials). Blue-shaded areas represent the laser stimulation. (H) Subplots of trials in (D, F) showing the immediate change in speed upon light activation at different laser intensities for Setd5^+/+ (n = 4, paired t tests) and Setd5^+/− (n = 4, paired Wilcoxon’s tests) optogenetics trials. S_at is the mean speed of the animal ±50 ms of laser onset, and S_im is the mean speed of the animal 300–800 ms after laser onset. (I) Proportion of trials at different laser intensities that either show an increase (white), decrease (black), or no change (grey) in speed upon light activation for Setd5^+/+ (top) and Setd5^+/− (bottom). 0.01 mW mm⁻²: 35 trials, p = 0.649; 0.5 mW mm⁻²: 111 trials, p = 0.193; 5 mW mm⁻²: 50 trials, p < 0.001; 10 mW mm⁻²: 49 trials, p < 0.001; 15 mW mm⁻²: 61 trials, p = 0.004; 20 mW mm⁻²: 49 trials, p < 0.001, X² test of independence. (J) Schematic of the behavioural divergence between genotypes with increasing stimulus intensity (laser or loom). (K) Summary of mean ± SEM of reaction time to the LER paradigm at different stimulus contrasts (p = 0.021 for the interaction between genotype and contrast, Setd5^+/+, n = 13, 68 trials; Setd5^+/−, n = 13, 59 trials, repeated measures ANOVA. p = 0.018 for 98% contrast, with multiple comparisons and Bonferroni correction. Setd5^+/+, 24 trials; Setd5^+/−, 22 trials). (l) Proportion of trials at different contrast looms that show an increase (white), decrease (black), or no change (grey) in speed upon light activation (20%: p = 0.698; 50%, p = 0.026; 98%, p = 0.021, X² test of independence). The data underlying this figure can be found in S4 Data.

https://doi.org/10.1371/journal.pbio.3002668.g004

dPAG neurons are hypoexcitable in the Setd5 mouse model

Since our previous results point towards a disruption in the circuits downstream of the SC, we decided to investigate if the activity of the underlying dPAG is affected in the Setd5^+/− model (Fig 5). To test if any synaptic or intrinsic properties of dPAG neurons are impaired, we performed whole-cell patch-clamp recordings in slices (Fig 5A). First, using the same animals used for optogenetic stimulation in vivo, we tested if the synaptic properties of dmSC inputs to the dPAG are changed. Ex vivo optogenetic activation elicited monosynaptic excitatory postsynaptic current (EPSC) with an average current of −244 ± 9.65 pA in Setd5^+/+ and −228 ± 11.8 pA in Setd5^+/− (Fig 5A and 5B). At this frequency, repeated stimulations showed no facilitation in both genotypes (Fig 5B), indicating that dmSC inputs to the dPAG are unaltered. Next, we explored spontaneous neurotransmission by recording sEPSCs (S7A–S7C Fig), supporting the view that the synaptic properties remain similar at a circuit level. Finally, we probed the intrinsic properties of dPAG neurons. These experiments were done blind to the cell type, but we were able to classify them based on their firing statistics as excitatory and inhibitory [32] (S7F–S7J Fig). Although the membrane potential, input resistance, membrane constant, and capacitance were identical between genotypes (Fig 5C), we observed a stark decrease in the ability of these cells to generate action potentials in response to injected current in both excitatory and inhibitory neurons (Figs 5D and S7H), without affecting the rheobase (S7K Fig). These changes were accompanied by changes in the spike shape of Setd5^+/− animals (Fig 5E and 5F) that showed slightly higher depolarisation, lower after-hyperpolarisation and slower rise dynamics in comparison to their WT siblings (S7I and S7J Fig). The hypoexcitability phenotype is also pronounced when the same analysis is performed on dPAG cells, which have been shown to receive input from dmSCs (S7K Fig) using optogenetic activation of dmSC fibres (Fig 5A and 5B). Based on prior work, we know that stronger dPAG activation leads to more vigorous escapes and that the underlying biophysical mechanisms depend on dPAG neuronal synaptic integration and excitability [29]. Thus, intrinsic excitability changes in excitatory neurons are a proxy of the expected escape strength. Remarkably, the response characteristics of dPAG neurons in Setd5^+/+ and Setd5^+/− match their behavioural differences (Fig 1). At weak current injections (Fig 5D), optogenetic activation (Fig 4D–4H), or low contrast looms (Fig 4K), both genotypes are indistinguishable from each other. However, at stronger current injections Setd5^+/+ and Setd5^+/− diverge, with a hypoexcitability phenotype for Setd5^+/− (Fig 5D) that matches the delay in LER for optogenetic (Fig 4D and 4F) and visual (Figs 1 and 4K) activations. To determine if the hypoexcitability phenotype was present in the preceding visual pathway, we patched dmSC interneurons (S7L–S7R Fig). Here, no statistical significance was observed despite a hypoexcitability trend, indicating that the strength of the effect is area-specific but not ruling out other brain-wide deficiencies. In combination with previous work [29] and our optogenetic experiments (Fig 4), these results suggest intrinsic deficits in the dPAG account, at least partly, for the delayed LER phenotype in the Setd5^+/− mice (Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Setd5^+/− dPAG cells are hypoexcitable.

(A) Schematic of the experimental approach, in vitro patch clamp recordings (top) and micrograph of VGluT2⁺ dmSC projections to the dPAG infected with AAV9-ires-DIO-ChR2 (green) with dPAG biocytin filled and recorded cells (red and arrow heads, scale bar: 100 μm). (B) Top, whole-cell voltage clamp traces of example Setd5^+/+ and Setd5^+/− cells (black and red, respectively) responding to 10 Hz light stimulation (blue ticks). Amplitude (middle, p = 0.078) and relative EPSC amplitude (bottom, p = 0.565) of responses to sequential light pulses in a 10-Hz train. (C) Intrinsic properties of Setd5^+/+ (n = 6, 11 cells) and Setd5^+/− (n = 7, 14 cells) dPAG cells. Input resistance (P = 0.756, top left), membrane constant tau (P = 0.436, top right), membrane capacitance (P > 0.995, bottom left) and resting membrane potential (P = 0.213, bottom right). (D) Summary of the relationship between current injection and action potential firing for putative glutamatergic cells showing a strong reduction in firing (p < 0.001). Inset, representative example traces to a 50-pA current injection. Grey area indicates the current injection values that significantly differ between Setd5^+/+ cells (black) and Setd5^+/− cells (red) found by a multiple comparisons analysis with Tukey correction. (E) Average shape and (F) phase plane analysis of the action potentials generated in the rheobase sweep (Setd5^+/+, 13 cells, 42 spikes; Setd5^+/−, 14 cells, 72 spikes). (G) Summary of the relationship between current injection and action potential firing for all Setd5^+/+ cells (black, n = 18) and Setd5^+/− cells (red, n = 20) before and after (Setd5^+/+: blue, n = 12 cells; Setd5^+/−: purple, n = 13 cells) application of α-Dendrotoxin (α-DTX, 100 nM, p > 0.995 and p = 0.0147 for the effect of α-DTX on Setd5^+/+ and Setd5^+/− firing, respectively). Inset, representative example traces from Setd5^+/+ cells (blue) and Setd5^+/− (purple) cells after α-DTX application to a 50-pA current injection. (H) Effect of α-DTX on firing in response to 120 pA current injection. Before α-DTX (Setd5^+/+ versus Setd5^+/−; p = 0.0017), effect of α-DTX on Setd5^+/+ (before versus after α-DTX; P > 0.995) and Setd5^+/− (before versus after α-DTX; P < 0.001, Setd5^+/+ before versus Setd5^+/− after α-DTX; P > 0.995). Multiple comparison analysis after rm-ANOVA with Tukey correction. (I) α-DTX-sensitive current densities in Setd5^+/+ and Setd5^+/− dPAG neurons (p < 0.001). Inset, example of α-DTX-sensitive traces (Setd5^+/+: black, n = 7 cells; Setd5^+/−: red, n = 6 cells). Grey area indicates the current values that significantly differ between Setd5^+/+ cells (black) and Setd5^+/− cells (red) found by a multiple comparisons analysis with Tukey correction. (J) Action potential shape and (K) phase plane analysis of the action potentials generated in the rheobase current in Setd5^+/+ cells (without α-DTX, black, n = 19; with α-DTX, blue, n = 12), Setd5^+/− cells (without α-DTX, red, n = 21; with αDTX, purple, n = 13). The data underlying this figure can be found in S5 Data.

https://doi.org/10.1371/journal.pbio.3002668.g005

dPAG cells are hypoexcitable due to K_v channel misregulation

Next, we investigated the underlying mechanisms of the hypoexcitable phenotype of the Setd5^+/− dPAG cells. The changes in excitability and spike kinetics are indicative of differences in channel composition that are part of adaptive homeostatic mechanisms [33], suggesting the influence of voltage-gated potassium channels (Kv). Previously published work [23], which studied the transcriptomic changes brought about by a mutation in Setd5^+/−, highlighted significant up-regulation in the expression of Kv1.1 channels during development. To test whether the blockade of Kv1.1 would rescue the hypoexcitability phenotype, we applied 100 nM of α-Dendrotoxin (α-DTX), a specific Kv1.1, Kv1.2, and Kv1.6 blocker, and patched dPAG cells. While α-DTX had no effect on Setd5^+/+ excitability of dPAG cells, strikingly, Setd5^+/− cells reversed their hypoexcitability phenotype to WT levels (Fig 5G and 5H). Accordingly, we measured a drastically increased α-DTX sensitive outward current in Setd5^+/− dPAG cells (Fig 5I). In addition, after α-DTX application, the spike dynamics remained identical in Setd5^+/+, whereas Setd5^+/− became faster, mimicking the WT properties (Fig 5J and 5K). These differences are not due to an overexpression of Kv channels. Tissue-specific proteomic analysis showed no difference in the protein level of Kv channels in either the PAG (Fig 6A), SC, or cortex (S8A and S8B Fig). This is in accordance with western blot and immunohistochemical analysis (Figs 6B, 6D, S8C and S8D). Given that Kv1.1 has been recently shown to be involved in the homeostatic control of firing [33,34] through modulation of the axon initial segment (AIS) [35], these results indicate that the deficits might not arise due to direct changes in expression but due to homeostatic misregulation. Next, we tested if the dPAG neurons of Ptchd1^Y/− and Cul3^+/− animals showed similar physiological properties to Setd5^+/− neurons. As expected from the molecular differences among the tested ASD models, the physiological characteristics differed. Ptchd1 animals, despite having mostly identical intrinsic properties to their WT siblings (S9A–S9D Fig), also show a hypoexcitability phenotype at high current inputs (S9E–S9I and S9S Fig). However, this hypoexcitability phenotype cannot be rescued by α-DTX application (S9S Fig). Cul3 animals did not show any changes in either their current-firing relationship or intrinsic properties (S9J–S9R, S9U and S9V Fig). Overall, these results indicate that different models might have distinct neurodevelopmental origins or key molecular changes that give rise to particular behavioural traits. In particular, for Setd5, Kv1 channels appear as key players and targets to reverse the delayed LER phenotype and probe the causality of its correlation with place avoidance deficits.

To investigate whether the area-specific dPAG hypoexcitability phenotype mediated the delayed LER, we tested whether α-DTX application targeted directly to the dPAG could rescue the behavioural phenotype (Fig 6A). We implanted cannulas above the dPAG in Setd5^+/− and their WT littermates, a target-specific approach as visualised through neurobiotin injections (Fig 6B). Initially, we injected saline as a control in both cohorts and subsequently tested the LER (Fig 6C and 6D, top). As previously demonstrated (Fig 1), Setd5^+/− animals responded slower than their WT littermates and had no difference in their maximal escape speed during the first presentation day (Fig 6D, top). However, the behaviour of both cohorts were indistinguishable from each other when α-DTX was applied, escaping immediately after the loom presentation (Fig 6C and 6D, bottom, and S3 Video), with similar reaction times (Fig 6E) and vigour (Fig 6F), even despite Setd5^+/− animals normally adapting across days (Fig 2D). We then quantified the shelter exits of both cohorts (Fig 7G and 7H) to investigate whether the correlation between shelter exits and delayed LER (Fig 2C) could have a common origin. Remarkably, Setd5^+/− animals recovered their place avoidance as well after α-DTX application (Fig 6H), indicating that the timely initiation of actions and memory formation are closely linked. These results suggest that perceptual deficits can have far-reaching consequences for unrelated behavioural traits.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Protein levels of Kv channels are not changed in Setd5^+/− dPAG cells.

(A) Volcano plot for differential protein levels between adult Setd5^+/+ and Setd5^+/− mice (n = 6 samples per genotype, cyan dots represent proteins annotated as ion-channels and purple dots represent Kv1.1, Kv1.2, and Kv1.6. Lines and shaded areas, mean ± SEM, respectively. Box-and-whisker plots show the median, IQR, and range. P-values are Wilcoxon’s rank sum test. p-values are two-way repeated measures ANOVA. Volcano plot horizontal dashed line represents the significance threshold (P-value < 0.1) from a two-sided moderated t test, while the vertical dashed lines indicate fold change values greater or lower than 0.4 between Setd5^+/+ and Setd5^+/−. (B, C) Tissue-specific western blots of Kv1.1 protein content in the dorsal periaqueductal grey (dPAG) for Setd5^+/+ and Setd5^+/− mice and (C) their quantification (dPAG: Setd5^+/+, 0.259; Setd5^+/−, 0.157, P = 0.090). (D) Schematic of SC and PAG regions of interest. (E, F) Antibody staining for Kv1.1 in (E) Setd5^+/+ and (F) Setd5^+/− (E, 55 μm projection; F, 30 μm projection). Arrowheads indicate somas stained for Kv1.1. Scale bar: D: 200 μm; E: 50 μm; F: 100 μm. P-values are two-tailed Wilcoxon’s signed-rank test. The data underlying this figure can be found in S6 and S7 Data.

https://doi.org/10.1371/journal.pbio.3002668.g006

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Pharmacological Kv channel block rescues delayed LER and place avoidance.

(A) Timeline of in vivo α-DTX cannula experiments. (B) Confocal micrograph of a coronal section of the location of the cannula above the dPAG as well as the expression of neurobiotin that was infused into the cannula at the end of the experiments. Scale bar: 500 μm. (C) Raster plots of mouse speed in response to the looming stimuli for Setd5^+/+ before (top, n = 6, 10 trials) and after (bottom, n = 6, 11 trials) infusion of α-DTX (500 nL, 500 nM) sorted by reaction time. (D) As for (C) but for Setd5^+/− before (top, n = 6, 15 trials) and after (bottom, n = 6, 13 trials) animals after infusion of α-DTX (500 nL, 500 nM). (E) Effect of α-DTX on reaction time. Before α-DTX (Setd5^+/+ versus Setd5^+/−; P = 0.017), effect of α-DTX on Setd5^+/+ (saline versus α-DTX; p = 0.104) and Setd5^+/− (saline versus α-DTX; p = 0.014) and after α-DTX (Setd5^+/+ versus Setd5^+/−; P = 0.571). (F) Effect of α-DTX on escape vigour. Before α-DTX (Setd5^+/+ versus Setd5^+/−; P = 0.052), effect of α-DTX on Setd5^+/+ (saline versus α-DTX; p = 0.760) and Setd5^+/− (saline versus α-DTX; p = 0.649) and after α-DTX (Setd5^+/+ versus Setd5^+/−; P = 0.075). (G) Shelter exit behaviour during the first LER trial when the mice are injected with saline (Setd5^+/+, black, top left; Setd5^+/−, red, bottom left) or α-DTX (Setd5^+/+, blue, top right; Setd5^+/−, purple, bottom right). Each row represents 1 animal, filled and open dots represent exits crossed into the threat zone or not, respectively. (H) Effect of α-DTX on shelter exits. Before α-DTX (Setd5^+/+ versus Setd5^+/−; P = 0.012), effect of α-DTX on Setd5^+/+ (saline versus α-DTX; p = 0.038) and Setd5^+/− (saline versus α-DTX; p = 0.003) and after α-DTX (Setd5^+/+ versus Setd5^+/−; P = 0.495). Markers represent the average values across all trials for individual animals. P-values: Wilcoxon’s test, p-values: paired t test. The data underlying this figure can be found in S8 Data.

https://doi.org/10.1371/journal.pbio.3002668.g007

Mice

The study was discussed and approved by the institutional ethics committee of the University of Veterinary Medicine Vienna and the Institute of Science and Technology Austria in accordance with good scientific practice guidelines and national legislation under license numbers BMWF-68.205/0023-II/3b/2014 and 66.018/0017-WF/V/3b/2017. Male and female adult Setd5^+/−, Setd5^+/fl::Emx1-Cre, Setd5^+/−::VGluT2-ires-Cre, Cul3^+/−, Ptchd1^Y/− mice were housed under a 12-h light/dark cycle (lights on at 07:00), with food and water available ad libitum. The animals were housed in groups of 2 to 6 animals per cage and were tested during the light phase. Animals used for the in vivo electrophysiological recordings or freely moving optogenetics experiments were group housed before surgery and individually housed after the surgery. Mice were chosen based on genotypes. Sex-matched animal pairs of control–mutant siblings from the same litters were compared to decrease variance due to age, environment, and genetic background.

Generation of Ptchd1^Y/− mice

Ptchd1^Y/− mice for analysis were generated using the CRISPR/Cas9 system, as described in [53]. In brief, in vitro transcribed sgRNAs targeting the exon 2 of Ptchd1 were microinjected into zygotes of C57BL/6NRj mice together with Cas9 mRNA. Offspring were screened by PCR for deletions using primers spanning the target sites (5′-GTAGGGCTGGAATCATGAGG-3′, 5′-CACATCCTTTGGTGTGATGC-3′). Deletion of Ptchd1 exon 2 was confirmed by Sanger sequencing analyses. The founder line was established from a female Ptchd1^+/− mouse with a 1.26-kb deletion. Analysis of Ptchd1 transcript level by RT-PCR confirmed impaired Ptchd1 expression in brain, cerebellum, and kidney of the Ptchd1^Y/− mice, compared to control mice.

Behavioural procedures

Viruses

The viruses used in this study: for optogenetic activation, adeno-associated virus (AAV) AAV9-EF1a-DIO-hChR2(E123T)T159C)-EYFP-WPRE-hGH (1 × 10¹³ viral genomes per ml (vg ml⁻¹), Addgene 35509); for control experiments, AAV9-Syn-Flex-GCaMP6m-WPRE-SV40 (1 × 10¹³ vg/ml, Addgene 100838).

Surgical procedures

Mice were anesthetised with an intraperitoneal (IP) injection of ketamine (95 mg kg⁻¹) and xylazine (4.5 mg kg⁻¹), followed by metamizol (20 mg kg⁻¹, IP), buprenorphine (0.05 mg kg⁻¹, IP), and meloxicam (2 mg kg⁻¹, subcutaneous (SC)). If needed, isoflurane (0.5% to2% in oxygen, 0.8 l min⁻¹) was used to maintain anesthesia. Fur on the top of the head was removed with an electric hair remover (Braun Precision Trimmer, PT 5010) and then the mice were placed in a stereotaxic frame (Model 962, Kopf Instruments) and fixed using the ear bars. Eyes were protected using Oleo Vital eye cream, and a topical analgesic (Xylocain 2% gel) was applied with a q-tip before cutting the skin. Craniotomies of approximately 1 mm diameter were made using 0.5 mm burrs (Fine Science Tools) and dental drill (Foredom, HP4-917), and viral vectors were delivered using pulled pipettes (World Precision Instruments, 1.4 mm OD, 0.53 mm ID, #504949) made with a micropipette puller (DMZ Zeitz-puller, Germany) and delivered using an automatic nanoinjector (World Precision Instruments, Nanoliter 2010) at 20 nl min⁻¹. The skin was closed after surgery using surgical glue (Vetbond, 3M, #1469SB).

Visual stimuli

All stimuli were created and presented using custom-made scripts in MATLAB (MathWorks), using the Psychophysics Toolbox extensions [54], and frames were presented at a refresh rate of 60 Hz.

In vivo electrophysiology

In vitro electrophysiology

Mice were deeply anesthetised via IP injection of ketamine (95 mg kg⁻¹) and xylazine (4.5 mg kg⁻¹), followed by transcardial perfusion with ice-cold, oxygenated (95% O₂, 5% CO₂) artificial cerebrospinal fluid (ACSF) containing (in mM): 118 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 1.5 MgSO₄, 1 CaCl₂, 10 Glucose, 3 Myo-inositol, 30 Sucrose, 30 NaHCO₃ (pH = 7.4). The brain was rapidly excised and coronal sections of 300 μm thickness containing the SC and PAG were cut using a Linear-Pro7 vibratome (Dosaka, Japan). Slices were left to recover for 20 min at 35°C, followed by a slow cool down to room temperature (RT) over 40 to 60 min. After recovery, 1 slice was transferred to the recording chamber (RC-26GLP, Warner Instruments, Holliston, MA, USA) and superfused with ACSF containing 2 mM CaCl₂ and 20 μM bicuculline methiodide (Tocris) at a rate of 3 to 4 ml/min at RT. In experiments using 100 nM α-dendrotoxin (Alomone Labs, Jerusalem, Israel), 0.1% BSA was added to the ACSF. Glass pipettes (B150-86-10, Sutter Instrument, Novato, CA, USA) with resistances of 3 to 4 MΩ were crafted using a P1000 horizontal pipette puller (Sutter Instrument) and filled with internal solution containing (in mM): 130 K-Gluconate, 10 KCl, 5 MgCl₂, 5 MgATP, 0.2 NaGTP, 0.5 EGTA, 5 HEPES (pH 7.4) adjusted with KOH. Biocytin (0.2% to 0.3%) was added to the internal solution for post hoc morphological reconstruction. Neurons of the dPAG or the dmSC were visually identified using an infrared differential interference contrast video system in a BX51 microscope (Olympus, Tokyo, Japan). Electrical signals were acquired at 20 to 50 kHz and filtered at 4 kHz using a Multiclamp 700B amplifier (Molecular Devices, San Jose, CA, USA) connected to a Digidata 1440A digitizer (Molecular Devices) with pClamp10 software (Molecular Devices). Spontaneous excitatory postsynaptic currents (sEPSCs) were recorded at –60 mV for at least 1 min. Intrinsic membrane properties and neuronal excitability were measured in current clamp mode by applying current steps from –20 to 200 pA for 1 s in 10 pA increments. To measure DTX-sensitive membrane currents, voltage steps from −80 to 50 mV in 10 mV increments were applied from a holding potential of −60 mV before and after a 10- to 15-min incubation in 100 nM α-dendrotoxin, and only 1 cell per slice was recorded. Access resistance was constantly monitored between protocols and recordings with access resistances exceeding 20 MΩ or with changes in access resistance or holding current by more than 20% were discarded. After recordings, the pipette was carefully withdrawn and the slice was transferred to 4% paraformaldehyde (PFA) in PBS solution.

Proteomics

Histology

For fluorescent immunostainings and histological confirmation of fibre and probe locations, adult Setd5^+/+ and Setd5^+/− mice were transcardially perfused with 4% PFA and their brains were dissected, postfixed overnight in 4% PFA, washed with 1X PBS, dehydrated in a 30% sucrose solution, and sectioned with a freezing microtome (SM2010R, Leica) to 100 μm (for probe/fibre/cannula confirmation) and 40 μm (for immunostainings) thick slices. For histological confirmation of probe and fibre locations, the sections were rinsed in PBS, stained with 4′,6-diamidino-2-phenylindole (DAPI; 2 mM in PBS) and mounted on Superfrost glass slides in a mounting medium containing 1,4-diazabicyclooctane (DABCO; Sigma, D27802) and Mowiol 4–88 (Sigma, 81381). For Kv1.1 immunostainings, tissue was rinsed and then blocked for 1 h at RT in a buffer solution containing 10% normal donkey serum (Abcam, ab7475), 0.1% Triton X-100 in PBS. Primary antibodies (Kv1.1, Rabbit, 1:200; Rb-Af400, Frontier Institute) were diluted in a blocking buffer and incubated for 48 h at 4°C. The tissue was then rinsed 3 times in PBS and left overnight to wash at 4°C. Secondary antibodies (Alexa-594 Donkey anti-rabbit, 1:1,000; R37119, Thermo Fisher Scientific) were diluted in 1X PBS and incubated for 2 h at RT in the dark, before being washed 3 more times in 1X PBS before being counterstained with DAPI and mounted as for histological sections described above. Sections were imaged with either a confocal microscope (LSM800, Zeiss), at 63X or 20X magnification, or a wide-field microscope (VS120, Olympus), at 10X magnification.

For visualisation of the dPAG/dmSC cells recorded in vitro and the neurobiotin infused into the cannula implanted mice, tissue sections were washed in 1X PBS for 10 min, then a solution of Streptavidin (1:100, 1 mg/ml, Thermo Fisher Scientific, S11227) in PBS with 0.3% Triton-X was added and the samples were left at 4°C overnight. The tissue was then washed 3 times with PBS, and in vitro slices were mounted into a custom designed chamber for thick slices and imaged with a confocal microscope (LSM800, Zeiss) at 20X (90 μm projection, 2 μm sections) and 40X (50 μm projection, 1 μm sections).

Western blot

Tissue samples from the SC, PAG, and cortex were dissected from Setd5^+/+ (n = 3) and Setd5^+/− (n = 3) mice. PAG tissue was dissected using a 2-mm diameter biopsy punch (ID: 2 mm, OD: 3 mm, Fine Science Tools, 18035–02). Each purification was performed independently 3 times, and samples were stored at −80°C until further processing. Samples were Dounce homogenised 15 times with 300 mL RIPA buffer (50 mM Tris-HCl (pH 8), 150 mM NaCl, 0.2% sodium dodecyl sulphate, 0.5% p/p sodium deoxycholate, 1% Triton-X, 1x cOmplete EDTA-free protease inhibitor cocktail (Roche) and 1 mM PMSF (Sigma) using the loose pestle and sonicated for 10 s. The lysed samples were centrifuged for 20 min at 16,000g, and the supernatant was stored at −80°C. Total protein concentration was determined using the BCA protein assay (Pierce). Proteins were boiled for 5 min in Laemmli sample buffer before being separated at 200 V for 40 min using TGX stain-free precast gels (Bio-Rad). Stain-free protein gels were activated with UV light for 5 min and then transferred to PVDF membranes (Bio-Rad) using a turbo transfer system (Bio-Rad). Membranes were blocked for 1 h in TBST (Tris-buffered saline with 0.1% Tween 20) supplemented with 5% BSA (Sigma) and then incubated overnight at 4°C with anti-Kv1.1 primary antibody (Rb-Af400, Frontier Institute, 1:100). After primary antibody incubation, membranes were washed 4 times with TBST for 5 min and then incubated with anti-rabbit IRDye 800CW (LI-COR Biosciences, dilution 1:4,000) for 1 h at RT.

Quantification and statistical analysis

Analysis of optogenetics experiments

Trials were grouped according to the laser power used and pooled across animals from each genotype. As in the analysis of the LER experiments, the speed of animal was calculated across an 83-ms time bin, and the effect of laser stimulation was assessed by comparing the mean speed of the animal ±50 ms around the onset of the laser (S_at) and 300 to 800 ms after the onset of the laser (S_im). Trials from the 0.01 mW mm⁻² condition were not subject to the escape criteria since laser stimulation at this intensity did not elicit an observed response; however, trials from all other laser power conditions had to meet the escape criteria used in the standard LER experiments in order to be included in further analyses. Trials were grouped by laser power intensity, and S_at and S_im were compared for all trials by paired Wilcoxon signed-rank tests. Similar to the LER analysis, responses were individually grouped based on both the sign and the magnitude of the difference between S_at and S_im, into 3 groups where S_im is either more than 1 SD greater than S_at, more than 1 SD less than S_at or within 1 SD of S_at. Chi-squared tests were used to compare the distribution of the Setd5^+/+ and Setd5^+/− trials at different laser powers.

For the control experiments, with GcaMP-injected mice (S6 Fig) the Log₁₀(ΔSpeed) was calculated as the logarithmic of the ratio between S_im and S_at. The change in Log₁₀(ΔSpeed) across laser powers was analysed using a two-way repeated measures ANOVA looking at the effect of genotype, laser power and their interaction, followed by a multiple comparisons analysis with Bonferroni correction.

Analysis of in vivo silicon probe recordings

Analysis of in vitro patch clamp electrophysiological recordings

Whole-cell recordings were analysed in Clampfit (Molecular Devices) and MiniAnalysis Program (Synaptosoft). Resting membrane potential (RMP) was measured as the average membrane potential prior to current injection. Input resistance (R_in) was calculated from the slope of the linear fit of the hyperpolarising current/voltage relationship. Membrane time constant τ was measured from the exponential fit of the membrane potential change in response to a –10 or –20 pA current injection. Membrane capacitance (C_m) was calculated from C_m = τ/R_in. Action potential firing was quantified in Clampfit using the event detection function. Phase-plane plots were performed to investigate the dynamic changes in the membrane potential over time (dV/dt) and were generated using the spikes from the first current step necessary to elicit action potentials (rheobase current). To visualize DTX-sensitive currents, 2 to 3 voltage-step protocols under baseline and DTX conditions were averaged and then the average current trace after DTX incubation was subtracted from the average baseline current trace. Currents were quantified as the maximal current deflection during the 1-s voltage step in either direction. To correct for different neuronal membrane surface areas between cells, current responses were divided by Cm and expressed as current density (pA/pF). Soma size was calculated by fitting an oval to the projected image (63X) of the neurons and calculating the area of that oval.

Analysis of proteomics

MS files were searched in DiaNN version 1.8.1 in library-free mode against a Mus musculus proteome obtained from UniProtKB. Match-Between-Runs was turned off. Fixed cysteine modification was set to Carbamidomethylation. Variable modifications were set to Oxidation (M), Acetyl (protein N-term) and Phospho (STY). Data were filtered at 1% FDR. DiaNN’s output was reprocessed using in-house R scripts, starting from the main report table. MS1 intensities were renormalized using the Levenberg–Marquardt procedure to minimize sample-to-sample differences. The long format main report table was consolidated into a wide format peptidoforms table, summing up quantitative values where necessary. Peptidoform intensity values were renormalized using the Levenberg–Marquardt procedure and corrected using the ComBat function from the sva package to remove the replicate-related batch effect. Protein groups were inferred from observed peptides and quantified using an in-house algorithm that (i) computes a mean protein-level profile across samples using individual, normalized peptidoform profiles (“relative quantitation” step); (ii) following the best-flyer hypothesis, normalizes this profile to the mean intensity level of the most intense peptidoform (“unscaled absolute quantitation” step); for protein groups with at least 3 unique peptidoforms, only unique ones were used; otherwise, razor peptidoforms were also included; phosphopeptidoforms and their unmodified counterparts were excluded from the calculations. Estimated expression values were log10-converted and renormalised using the Levenberg–Marquardt procedure. P-values were adjusted using the Benjamini–Hochberg (FDR) method. Average log10 expression values were tested for significance using a two-sided moderated t test per samples group and a global F-test (limma). Significance thresholds were calculated using the Benjamini–Hochberg procedure for FDR values of 1%, 5%, 10%, and 20%. For all tests, regulated protein groups were defined as those with a significant P-value and a log2 ratio greater than 5% of control-to-average-control ratios. GO terms enrichment analysis was performed comparing for each test regulated against observed protein groups (topGO). Statistical analysis was performed for phospho-modified peptides as for protein groups, normalizing ratios to parent protein group(s).

Analysis of histological data

For comparison between the Kv1.1 expression in Setd5^+/+ and Setd5^+/− tissue 30 μm stacks were imaged with 1 μm sections at 63X magnification, or 55 μm stacks were imaged with 2 μm sections at 20X magnification, and the same laser power settings were used for all images at a given magnification.

Analysis of western blot data

The Bio-Rad ChemiDoc MP system was used for imaging and acquisition and images were quantified using ImageJ software. Protein levels are displayed as fold change. Student t test was used to test for significance (MATLAB, MathWorks). Data represent mean ± SEM obtained from 3 mice per genotype.

Statistics

Analyses were performed in custom-written MATLAB (2019a) scripts (MathWorks). Shapiro–Wilk and Levene’s tests were used to test the normality and the equal variances of the datasets, respectively, and parametric or nonparametric tests were used accordingly. Two-tailed tests were used unless specified. All statistical tests are reported in the text and appropriate figure legends (*p < 0.05, **p < 0.01, ***p < 0.001). In bar plots, the mean ± SEM are shown, unless otherwise stated. Boxplots display the median and IQR.