Cortical Network Dynamics Is Altered in Mouse Models of Huntington’s Disease

Abstract

Huntington’s disease (HD) is a neurodegenerative disorder characterized by involuntary movements, cognitive deficits, and psychiatric disturbances. Although evidence indicates that projections from motor cortical areas play a key role in the development of dysfunctional striatal activity and motor phenotype, little is known about the changes in cortical microcircuits and their role in the development of the HD phenotype. Here we used two-photon laser-scanning microscopy to evaluate network dynamics of motor cortical neurons in layers II/III in behaving transgenic R6/2 and knock-in Q175^+/− mice. Symptomatic R6/2 mice displayed increased motion manifested by a significantly greater number of motion epochs, whereas symptomatic Q175 mice displayed decreased motion. In both models, calcium transients in symptomatic mice displayed reduced amplitude, suggesting decreased bursting activity. Changes in frequency were genotype- and time-dependent; for R6/2 mice, the frequency was reduced during both motion and nonmotion, whereas in symptomatic Q175 mice, the reduction only occurred during nonmotion. In presymptomatic Q175 mice, frequency was increased during both behavioral states. Interneuronal correlation coefficients were generally decreased in both models, suggesting disrupted interneuronal communication in HD cerebral cortex. These results indicate similar and contrasting effects of the HD mutation on cortical ensemble activity depending on mouse model and disease stage.

Introduction

Huntington’s disease (HD) is an inherited, neurodegenerative disorder caused by an expansion of CAG (glutamine) repeats within the Huntington (HTT) gene, which codes for the protein huntingtin (Htt) (Gusella et al. 1983; The Huntington's Disease Collaborative Research Group 1993). Phenotypically, HD is manifested by psychiatric and cognitive disturbances, weight loss, and progressive development of dance-like involuntary movements known as chorea (Bates et al. 2002; Harper and Jones 2002). Neuropathologically, HD brains are characterized by a massive loss of medium-sized spiny neurons within the striatum and atrophy of the cerebral cortex caused by the loss of both cortical pyramidal neurons (CPN) and some types of interneurons (Macdonald and Halliday 2002). Although the relationship between neuronal damage and the development of the phenotype remains far from understood, a close relationship between altered neuronal function of the motor cortex and the development of the motor phenotype has been long recognized (Thu et al. 2010; Waldvogel et al. 2015). For example, neuronal imaging studies from presymptomatic patients show that dysfunctional activity in the cerebral cortex precedes changes in the striatum (Rosas et al. 2005; Rosas et al. 2008). In addition, functional studies reveal that somatosensory, attention, and cognitive deficits precede chorea and other motor abnormalities (Boecker et al. 1999; Paulsen et al. 2008; Beglinger et al. 2010). Thus, examining cortical alterations before the onset of striatal dysfunction is of vital importance. Furthermore, while reducing mutant Htt from the striatum or cerebral cortex improves the behavioral phenotype, optimal results can only be achieved when both cortical and striatal neurons are targeted (Wang et al. 2014).

Electrophysiological studies in mouse models have provided further evidence of cortical alterations throughout the progression of HD. Indeed, the firing patterns of CPNs change in HD mice. For example, burst and correlated firing were reduced in several mouse models of HD (Walker et al. 2008; Miller et al. 2011). Studies in brain slices demonstrated cortical hyperexcitability and increased synchronous firing after GABA_A receptor blockade (Cummings et al. 2009). However, a more complete picture of neuronal interactions in cortical microcircuits remains to be determined.

Calcium (Ca²⁺) imaging has become a versatile method to study the behavior and intercorrelations of neuronal ensembles in awake, behaving animals (Grienberger and Konnerth 2012; Peron et al. 2015). To gain insights into the role of the cerebral cortex in the HD phenotype, we examined network dynamics including changes in amplitude and frequency of Ca²⁺ transients and interneuronal correlations of primary motor area (M1) neurons in two models of HD, the R6/2 (a transgenic model of juvenile HD) and the Q175 (a knock-in model of adult-onset HD) using two-photon laser-scanning microscopy (2PLSM) imaging of large populations of neurons in behaving mice.

Methods

Animals

Mouse use and experimental procedures were performed in accordance with the United States Public Health Service Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care Committee at the University of California, Los Angeles (UCLA). All efforts were made to minimize the suffering and the number of animals used. Mice were group housed under controlled temperature (21–24°C) and humidity conditions (30–60%) and maintained on a 12 h light/dark cycle with free access to food and water. Behavioral and Ca²⁺ imaging data were obtained during the light cycle.

Male and female mice from the two different genetic mouse models of HD, the R6/2 (line 2810, 150 ± 5 CAG repeats represented as mean ± SE) and the Q175 (183 ± 1.8 CAG repeats), were used. Repeat length for all mice used was determined by Laragen Inc (Culver City, CA). We collected data from 9 R6/2 (4 males and 5 females) and 9 wild-type (WT) (6 males and 3 females) littermates 2–3 months old (WT: 77.9 ± 6.8 days old, R6/2: 65.5 ± 1.9 days old). We examined 2 different age groups of heterozygous Q175 mice: 2–3 months (presymptomatic; 4 male and 6 female Q175 102.9 ± 4.3 days old; 4 male and 4 female WT 103.8 ± 4.8 days old) and 12–18 months (symptomatic; 4 male and 2 female Q175 466.7 ± 33.0 days old; 1 male and 4 female WT 417.2 ± 5.8 days old). Three of the Q175 mice in the 12–18 months group were about 3 months older than the oldest WT mouse. There were no consistent differences in repeat length among age groups of Q175 mice. Data from males and females were pooled for analysis since we did not observe consistent differences between sexes. Given our low number of mice per each sex, our data do not have enough power to imply any differences between the sexes. However, it is important to note that in a previous well-powered electrophysiological study, the authors did not find sex differences and also pooled data from males and females (Walker et al. 2008).

Surgery

Each mouse received two surgeries, first one for virus injection and then one for craniotomy and headbar implantation. For the virus injection, mice were deeply anesthetized with isoflurane (4% vol/vol for induction and 1–1.5% vol/vol for maintenance) and placed in a stereotaxic frame. A hole in the skull was drilled with a small steel burr (0.33 mm diameter) at +1.0 mm A/P and + 1.5–1.7 mm M/L to bregma to target primary motor cortex (M1, Paxinos and Franklin, 2001). A Nanoject Microliter Injector (Drummond Scientific Company) with a glass pipette was used to inject 0.15 μL of AAV1.Syn.GCaMP6f.WPRE.SV40 (Penn Core Vector, ddTiter: 5.006 × 10¹²), an AAV containing a GCaMP6f construct for visualization of Ca²⁺ transients. Two sites (+1.0 mm A/P, +1.5 mm M/L to Bregma and + 1.0 mm A/P, +1.7 mm M/L) and two different depths (400 and 600 μm from the surface of the brain) were injected with the virus at a rate of 0.03 μL/min for a total of 0.6 μL of virus per mouse. After each injection, the pipette was left in place for 10 min to allow the diffusion of the virus. Following surgery, mice were injected subcutaneously with 0.5 mL saline, to prevent dehydration, and the antibiotic, Carprofen (5 mg/kg).

The craniotomy and headbar implantation were performed 3 weeks after the virus injection to permit GCaMP6f to be expressed. Mice were deeply anesthetized with isoflurane (4% v/v for induction and 1–1.5% v/v for maintenance) and head-fixed in a stereotaxic frame. Lidocaine (0.1 mL) was injected under the scalp and the scalp was removed. The skull was cleaned and dried and the neck muscles were pushed back and glued with cyanoacrylate or super glue. A stainless-steel headbar was fixed with cyanoacrylate glue to the base of the skull to later head-fix the mouse under the 2PLSM (Scientifica). Once the headbar was firmly attached, a 5 mm craniotomy was made above the virus-injection sites. The skull was removed and a 5 mm circular glass plate (Warner Instruments) was permanently glued to the skull using cyanoacrylate. The exposed skull and the base of the head bar were covered with dental acrylic (Ortho-Jet). One week postsurgery, mice received a session to habituate to head fixation prior to Ca²⁺ imaging sessions. To verify injection placement, after completion of recording sessions, mice were deeply anesthetized with isoflurane (4% v/v) and were transcardially perfused with 0.9% saline followed by 10% paraformaldehyde. The brain was immediately removed and stored in 10% paraformaldehyde. Prior to sectioning, brains were placed in 10% sucrose for 24 h then 40 μm coronal sections were cut at −19°C using a cryostat (Leica) and mounted on slides.

Motion Recording and Analysis

The mouse was positioned in the headbar restrainer and placed under the 2PLSM while its body was positioned on a spherical treadmill composed of an 8 inch Styrofoam ball (Graham Sweet Studios) suspended by compressed air (Fig. 1A). The ball could move freely in both the anterior–posterior and lateral directions. Movement of the ball was monitored by optical sensors (Avago Technologies) and collected as voltage readings by a data acquisition system (Analog Devices) (Fig 1B). Voltage thresholds were set to detect periods of motion and nonmotion. Thresholds for detection of motion were determined by identifying the baseline voltage during periods of nonmotion for 3 videos each from 3 different mice and applied to all subsequent recordings. Data from the motion sensor channels were combined to determine periods of motion or nonmotion. A motion epoch was counted if the combined motion lasted for more than 200 ms and ended immediately prior to a sustained period of nonmotion greater than 200 ms. Recording sessions in which an animal displayed no motion periods or was in motion for the entire session were removed from analysis.

Figure 1.

Motion data for HD (gray) and WT (black) mice. (A) Picture of recording setup with mouse head restrained under a 20× objective and body positioned over a Styrofoam ball. (B) Example traces from motion sensors detecting forward and reverse motion. The solid black trace represents a WT mouse and the solid gray trace represents a R6/2 mouse. Black dashed lines indicate the motion threshold. (C) Average overall time spent in motion for each recording session in R6/2 mice (left; WT: n = 45, R6/2: n = 52) and Q175 mice (right) at 2–3 months (WT: n = 60, Q175: n = 79) and >12 months (WT: n = 47, Q175: n = 43), (D) average total number of motion epochs for each recording session in R6/2 mice (left; WT: n = 45, R6/2: n = 52) and Q175 mice (right) at 2–3 months (WT: n = 60, Q175: n = 79) and >12 months (WT: n = 47, Q175: n = 43), and (E) average duration of an epoch for each recording session in R6/2 mice (left; WT: n = 279, R6/2: n = 1073) and Q175 mice (right) at 2–3 months (WT: n = 658, Q175: n = 534) and >12 months (WT: n = 349, Q175: n = 279). Error bars indicate standard error of the means by mouse. ^*P < 0.05 and ^***P < 0.001 as determined by Student’s t-tests.

Ca²⁺ Transient Recording and Analysis

Ca²⁺ transient images were acquired using ScanImage software (Vidrio Technologies) and a 20×/0.95 NA objective (Olympus). Excitation was elicited with 920 nm light from a Titanium-Sapphire tunable laser (Coherent) with 32 kHz resonant scanning and detected with gallium arsenide phosphide (GaAsp) photomultiplier tubes (Hamamatsu). Movies (image size: 500 × 500 μm) were recorded from cortical layers II/III at 31 Hz for a total of 8000 frames (4.3 min). Ca²⁺ imaging videos were analyzed using routines custom-written in NIH ImageJ and MATLAB (Mathworks). First, the Ca²⁺ imaging videos were visually inspected and videos with excessive motion artifacts or apparent z-shifts were excluded from analysis. A maximum intensity image of each 2PLSM recording was generated in ImageJ and used to isolate neuronal somata. Regions of interest (ROIs) were identified as areas with intensities > 30% of the background intensity. Identified ROIs were eliminated if they were greater than 380 μm² and smaller than 152 μm². These values were based on somatic areas from CPNs obtained from our separate slice data. Each automatically identified neuron was manually inspected. Identified neurons were excluded from the analysis if they did not exhibit characteristics of a healthy neuron including morphological inconsistencies (i.e., oblong shape) or persistent fluorescence (i.e., lack of any detectable Ca²⁺ transient). Movies with less than 10 verified neurons were excluded from the analysis. Fluorescence intensities from neuronal ROIs were then converted to ΔF/F values over time where F was the background fluorescence intensity. Background fluorescence intensity was determined as the mode fluorescence in a 30 × 30 pixel window of the darkest region of the video that did not include a neuron or blood vessel.

Ca²⁺ transient peaks were detected by applying a MATLAB smoothing and peak detection function to the waveform. Amplitudes of individual Ca²⁺ transients were calculated as the difference between the peak of a transient and the baseline ΔF/F. The average amplitudes of the Ca²⁺ transient for a given neuron during motion were determined separately from Ca²⁺ transients occurring during nonmotion. Average Ca²⁺ transient amplitudes of all neurons for a given genotype during periods of motion or nonmotion were calculated and the standard error of the mean was based on values per neuron.

To determine frequency of Ca²⁺ transients, the number of Ca²⁺ transients occurring during either motion or nonmotion was divided by the total time in seconds spent in motion or nonmotion to provide a frequency in Hz. The average frequency for all neurons of a given genotype during periods of motion or nonmotion was calculated and the standard error of the mean was based on values per neuron.

To further analyze the network activity, neurons were classified into four different categories based on the mean amplitude and frequency of Ca²⁺ transients: 1) high amplitude and high frequency (HAHF), 2) high amplitude and low frequency (HALF), 3) low amplitude and high frequency (LAHF), and 4) low amplitude and low frequency (LALF). To classify these neurons, the average amplitude and frequency was determined for the WT mice during periods of motion and nonmotion separately. Values above the average were classified as high and values below the average were classified as low. The distributions and proportions of neurons in each category were determined.

Because of the relatively slow decay in the GCaMP6f signal, ΔF/F traces were temporally deconvolved to estimate neuronal activity (Yaksi and Friedrich 2006) (see Fig. 2E). The deconvolved traces for each cell were then correlated to those of all other cells using a Pearson Product-Moment correlation coefficient. To determine a significant correlation value, we used a Monte Carlo simulation where each deconvolved trace was randomly shifted in time from 0.032 s, the time of a single frame, up to 259.32 s. The time shift only moved the deconvolved values forward in time while any remaining tail of the trace due to the forward time shift was moved to the beginning of the time-shifted trace. The random time shift did not shuffle the sequential order of the deconvolved Ca²⁺ transients. The shifted deconvolved trace was then correlated to all other nonshifted deconvolved traces. This random shifting and correlating process was repeated 100 times for each trace. The 99th percentile of the distribution from the correlation coefficients derived from the Monte Carlo simulation was used as the cutoff for statistically significant Pearson Product-Moment correlation coefficients. Average cutoff for a significant correlation was 0.114 ± 0.003. The proportion of significantly correlated pairs as well as the mean significant Pearson Product-Moment correlation coefficients were calculated. Mean significant Pearson Product-Moment correlation and proportion of significantly correlated pairs were averaged for each movie. Distances between pairs of cells were determined from center to center of each ROI. The mean significant Pearson Product-Moment correlation coefficients and proportion of significant correlations were binned by distances of 0–100 μm, 100–200 μm, 200–300 μm, and 300–400 μm between pairs of neurons and averaged by movie. In addition to measuring the Pearson Product-Moment correlation, the autocovariance and cross-covariance were measured by time-shifting each trace ± up to 4 s in 0.032 s increments and correlated the shifted trace to either the original trace or another neuron. Autocovariance and cross-covariance was averaged by movie and normalized to the maximum covariance for each given movie.

Figure 2.

Example Ca²⁺ transients of WT (A&B) and R6/2 (C&D) neurons during a recording session. (A) Maximum intensity image of an example WT recording session with 5 neurons labeled. Scale bar indicates 100 μm. (B) ΔF/F over a portion of the recording session for each neuron labeled in A. (C) Maximum intensity image of an example R6/2 recording session with 5 neurons labeled. Scale bar indicates 100 μm. (D) ΔF/F over a portion of the recording session for each neuron labeled in C. (E) Example Ca²⁺ transients of a WT neuron in black and the corresponding deconvolved trace in gray.

To further analyze the network activity of cortical neurons that were significantly correlated, neurons were classified into the 4 different categories described above: HAHF, HALF, LAHF and LALF. The mean correlation of neuron pairs within each category was determined.

Statistical Analyses

All statistical analyses were performed using SigmaStat. Student’s t-tests, appropriately designed analysis of variance (ANOVAs) with post hoc t-tests using the Holm-Sidak method, and Chi-square tests were used to measure statistical significance between and among experimental groups. In a few cases where data failed the normality test, a Mann–Whitney rank-sum test was performed. In most cases, data were represented as individual neuron averages; however, the proportion of significant correlations was averaged by movie in order to determine error. The average correlation values were also calculated by movie to estimate the proportion of significant correlations. Corresponding statistical tests are described in the results and in each figure legend. Significance was defined as P < 0.05.

Results

We examined locomotor activity, frequency and amplitude of Ca²⁺ transients, and interneuronal correlations in symptomatic R6/2 mice (>60 days) and in Q175 mice at presymptomatic (2–3 months) and symptomatic ages (>12 months).

Motion and Nonmotion of R6/2 Mice under Head-Restraint

To our knowledge, no studies have examined motion of HD mice during head-restraint, a necessary condition widely used in 2PLSM studies (Grienberger and Konnerth 2012). Thus, we first measured the amount of time the mice spent in motion or in nonmotion on the Styrofoam ball. Contrary to expectation, symptomatic R6/2 mice exhibited an increase in total time spent in motion compared to WT mice (t(95) = −6.83, P < 0.001; Fig. 1C). The increased time in motion was driven by a significant increase in the number of motion epochs (t(95) = −10.07, P < 0.001) rather than the duration of each motion epoch. We did not differentiate the types of motion, and epochs of motion included walking, running, shifting balance, or lateral swinging of the posterior body.

Motion and Nonmotion of Q175 Mice under Head-Restraint

We quantified the same three parameters of motion (total time in motion, total number of motion epochs and average duration of each epoch in a recording session) in Q175 mice (Fig. 1). At 2–3 months, the total time in motion was similar in both WT and Q175 mice. However, there was a significant reduction in number of epochs (genotype: F_1,237 = 9.5, P < 0.01, post hoc: t = 3.9, P < 0.001) coupled with an increase in epoch duration, although not statistically significant, which together produced the similarity of total time in motion. At >12 months, Q175 mice displayed significantly decreased time in motion (Age x genotype: F_1,237 = 10.3, P = 0.001, post hoc: t = 5.3, P < 0.001), an equal number of epochs, but a marked decrease in epoch duration (post hoc: t = 5.1, P < 0.001). Thus, motion patterns in Q175 mice reflect the hypokinesia expected in symptomatic HD model mice and as demonstrated previously by studies in this mouse model (Heikkinen et al. 2012; Menalled et al. 2012). Interestingly, the total time in motion (post hoc: t = 5.0, P < 0.001) and epoch duration (post hoc: t = 6.6, P < 0.001) were increased over age in the WT mice.

These results highlight important differences in locomotor activity under head-restraint between the two HD models and also as a function of age. R6/2 mice showed hyperkinesia compared to WTs whereas Q175 mice showed hypokinesia compared to WTs. Because of differences in locomotor activity, we separated subsequent analyses of Ca²⁺ transient amplitudes, frequencies and correlation patterns into epochs of motion and nonmotion.

Frequency and Amplitude of Ca²⁺ Transients in R6/2 Mice

Although a similar number of videos per mouse were recorded (Table 1), we found significantly fewer neurons per video in R6/2 compared to WT mice (t(95) = 2.70, P < 0.01; Table 1). Figure 2 shows typical images of labeled neurons in layers II/III of motor cortex and traces of Ca²⁺ transients in WT and R6/2 mice. In general, cortical neurons in WT mice generated more Ca²⁺ transients than R6/2 mice and the transients were of higher amplitudes. The frequency of Ca²⁺ transients was significantly decreased in cells from R6/2 compared to WT mice during both motion and nonmotion (genotype: F_1,9466 = 119.2, P < 0.001, post hoc motion: t = 3.11.6, P < 0.001, post hoc nonmotion: t = 3.9, P < 0.001) (Figs 2 and 3A). Interestingly, the frequency of Ca²⁺ transients was significantly higher during motion than during nonmotion in WTs (Behavior x Genotype: F_1,9466 = 29.8, P < 0.001, post hoc: t = 5.4, P < 0.001), whereas in R6/2 mice, the frequency was significantly higher during nonmotion compared to motion (post hoc: t = 2.5, P = 0.013). These findings suggest that Ca²⁺ activity and behavior in R6/2 mice could be dissociated. Similar to changes in frequency, the amplitude (ΔF/F) of Ca²⁺ transients was significantly reduced in neurons from R6/2 mice compared to WTs during both periods of motion and nonmotion (genotype: F_1,6972 = 948.1, P < 0.001, post hoc motion: t = 21.1, P < 0.001; post hoc nonmotion: t = 23.1, P < 0.001; Figs 2 and 3B). In addition, in cells from WT mice, the amplitude of Ca²⁺ transients was significantly higher during motion than during nonmotion (Behavior x Genotype: F_1,6972 = 7.0, P < 0.01, post hoc: t = 4.1, P < 0.001), whereas in cells from R6/2 mice this was not the case. Based on the literature, a larger amplitude Ca²⁺ transient is thought to indicate an increase in action potential number in a burst (Rothschild et al. 2010; Chen et al. 2013). Therefore, the decrease in Ca²⁺ transient amplitude in R6/2 mice could indicate an overall decrease in bursting, consistent with previous electrophysiological findings (Walker et al. 2008).

Table 1.

Number of neurons and movies in R6/2 and WT littermates

	# Mice	# Movies	Mean movie/mouse	# Cells	Cells/movie	Cell range
WT	9	45	5.0 ± 0.9	2868	55.2 ± 4.9	10–131
R6/2	9	52	5.8 ± 1.2	2049	39.4 ± 3.4**	10–108

Number of mice, movies, and neurons recorded in R6/2 and WT littermates. ^**P < 0.01.

Figure 3.

(A) Mean frequency of Ca²⁺ transients (Hz) from HD (gray) and WT (black) mice during motion and nonmotion. Left graph shows frequency for R6/2 mice during motion (WT: n = 1348, R6/2: n = 1182) and nonmotion (WT: n = 2583, R6/2: n = 1832). Middle and right graph show frequency for Q175 mice at 2–3 months and >12 months during motion (middle; WT 2–3 months: n = 1931, Q175 2–3 months: n = 2322, WT > 12 months: n = 1595, Q175 > 12 months: n = 1054) and nonmotion (right; WT 2–3 months: n = 2895, Q175 2–3 months: n = 4103, WT > 12 months: n = 1898, Q175 > 12 months: n = 1843). Error bars indicate standard error of the mean within neurons. (B) Mean amplitude of Ca²⁺ transients (ΔF/F) from HD (gray) and WT (black) mice during motion and nonmotion. Left graph shows amplitude for R6/2 mice during motion (WT: n = 1348, R6/2: n = 1182) and nonmotion (WT: n = 2583, R6/2: n = 1832). Middle and right graph shows amplitude for Q175 mice at 2–3 months and >12 months during motion (middle; WT 2–3 months: n = 1931, Q175 2–3 months: n = 2322, WT > 12 months: n = 1595, Q175 > 12 months: n = 1054) and nonmotion (right; WT 2–3 months: n = 2895, Q175 2–3 months: n = 4103, WT > 12 months: n = 1898, Q175 > 12 months: n = 1843). Error bars indicate standard error of the mean within neurons. ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001 as determined by a two-way ANOVA with post hoc Holm-Sidak tests or Mann–Whitney U-ranked sum tests.

When neurons were subcategorized by their amplitude and frequency, as described in the Methods section, we first noted a marked increase in variability during motion versus nonmotion categories in both WT and R6/2 mice (Fig. 4A). In general, during both motion and nonmotion, proportionately more neurons from both WT and R6/2 mice were in the low amplitude (LA) subcategory. Interestingly, WT mice demonstrated a reversal in proportions in the LAHF where there was a greater proportion of LAHF neurons compared to R6/2 mice in motion (χ² = 10.1, P = 0.001; Fig. 4A,B), but a lower proportion in WT mice compared to R6/2 mice during nonmotion (χ² = 16.0, P < 0.001; Fig. 5A,B). Overall, the greatest proportion of neurons displayed LALF in R6/2 mice during both motion and nonmotion, and the differences were statistically significant during both motion (χ² = 315.4, P < 0.001; Fig. 4A,B) and nonmotion (χ² = 193.6, P < 0.001; Fig. 5A,B). There were proportionately fewer neurons in the high amplitude (HA) subcategory, and there were significant decreases in the proportions of cells from R6/2 mice in HAHF and HALF subcategories compared to WT mice during both motion (HAHF: χ² = 212.4, P < 0.001; HALF: χ² = 72.3, P < 0.001) and nonmotion (HAHF: χ² = 259.1, P < 0.001; HALF: χ² = 140.9, P < 0.001).

Figure 4.

(A) Scatter plot of the amplitude and frequency of each neuron during motion for HD (gray) and WT (black) mice. Amplitude is on the x-axis and frequency is on the y-axis. Dashed lines indicate average WT amplitude (vertical) and frequency (horizontal). Each point represents the mean amplitude and frequency of a given neuron. Top set of scatter plots shows amplitude and frequency for R6/2 mice (WT: n = 1348, R6/2: n = 1182) and bottom sets of scatter plots show amplitude and frequency for Q175 mice at 2–3 months (left; WT: n = 1931, Q175: n = 2322) and >12 months (right; WT: n = 1595, Q175: n = 1054). (B) Proportion of significantly correlated neurons in each category of high versus low amplitude or frequency neurons during motion. Left graph shows the distribution of R6/2 neurons, and the right graph shows the distribution of Q175 neurons at 2–3 months and >12 months. LAHF represents low amplitude Ca²⁺ transients with high frequency; LALF represents low amplitude Ca²⁺ transients with low frequency; HAHF represents high amplitude Ca²⁺ transients with high frequency neurons; HALF represents HA Ca²⁺ transients with low frequency neurons. ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001 as determined by Chi square tests.

Figure 5.

(A) Scatter plot of the amplitude and frequency of each neuron during nonmotion for HD (gray) and WT (black) mice. Amplitude is on the x-axis and frequency is on the y-axis. Dashed lines indicate average WT amplitude (vertical) and frequency (horizontal). Each point represents the mean amplitude and frequency of a given neuron. Top set of scatter plots shows amplitude and frequency for R6/2 mice (WT: n = 2583, R6/2: n = 1832) and bottom sets of scatter plots show amplitude and frequency for Q175 mice at 2–3 months (left; WT: n = 2895, Q175: n = 4103) and >12 months (right; WT: n = 1898, Q175: n = 1843). (B) Proportion of significantly correlated neurons in each category of high versus low amplitude or frequency neurons during nonmotion. Left graph shows the distribution of R6/2 neurons, and the right graph shows the distribution of Q175 neurons at 2–3 months and >12 months. LAHF represents low amplitude Ca²⁺ transients with high frequency; LALF represents low amplitude Ca²⁺ transients with low frequency; HAHF represents HA Ca²⁺ transients with high frequency neurons; HALF represents HA Ca²⁺ transients with low frequency neurons. ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001 as determined by Chi square tests.

Frequency and Amplitude of Ca²⁺ Transients in Q175 Mice

A similar number of neurons and videos per mouse were recorded from both Q175 and WT mice in both age groups (Table 2). Although symptomatic R6/2 mice showed a significant decrease in frequency of Ca²⁺ transients during motion and nonmotion (see Fig. 3A), in Q175 mice the frequency of transients was increased compared to WTs during motion (genotype main effect: F_1,11 426 = 3.9, P < 0.05) in both the 2–3 months age group (U = 6308842.0, P < 0.01) and the >12 months age group (U = 2310624.5, P < 0.001, Mann–Whitney test) (Fig. 3A). For both Q175 and WTs, the frequency of Ca²⁺ transients decreased significantly with age during periods of motion (age: F_1,11 426 = 11.6, P < 0.001) (Fig. 3A). During nonmotion, Q175 mice showed opposite changes in frequency with age. At 2–3 months, the frequency was significantly increased in Q175 compared to WT mice (age x genotype: F_1,11 426 = 187.3, P < 0.001; post hoc: t = 16.4, P < 0.001) whereas at >12 months, the frequency was significantly decreased in Q175 compared to WT mice during nonmotion (post hoc: t = 4.7, P < 0.001) (Fig. 3A). These differences were due to both an increase in frequency in WT mice with age (post hoc: t = 10.9, P < 0.001) and a decrease in frequency in Q175 mice with age during nonmotion (post hoc: t = 8.4, P < 0.001).

Table 2.

Mice, movies, and neurons recorded in Q175 and WT littermates

	# Mice	# Movies	Mean movie/mouse	# Cells	Cells/movie	Cell range
2–3 Months
WT	8	60	8.6 ± 0.5	3390	49.1 ± 2.6	11–91
Q175	10	79	7.9 ± 1.0	3838	48.6 ± 2.4	11–89
2–18 Months
WT	6	47	7.8 ± 1.3	2151	45.8 ± 3.2	11–103
Q175	6	43	7.2 ± 1.1	1963	45.9 ± 3.9	10–106

Number of mice, movies, and neurons recorded in Q175 and WT littermates at 2–3 months and 12–18 months.

Similar to symptomatic R6/2 mice (see Fig. 3B), both 2–3 months and >12 months Q175 mice displayed a decrease in the amplitude of Ca²⁺ transients compared to WTs during both motion (age x genotype: F_1,6898 = 39.2, P < 0.001; 2–3 months post hoc: t = 3.7, P < 0.001; > 12 months post hoc: t = 10.8, P < 0.001) and nonmotion (age x genotype: F_1,10 735 = 85.2, P < 0.001; 2–3 months post hoc: t = 4.6, P < 0.001; >12 months post hoc: t = 17.9, P < 0.001) (Fig. 3B). The amplitude of Ca²⁺ transients during motion increased with age in WTs (t = 7.9, P < 0.001) but not Q175 mice (Fig. 3B). During nonmotion, the amplitude of Ca²⁺ transients also increased with age in WT mice (t = 7.5, P < 0.001) but decreased with age in Q175 mice (t = 9.0, P < 0.001) (Fig. 3B).

When neurons were subcategorized by their amplitude and frequency during motion, it was very apparent that the great majority of neurons at both ages were in the low amplitude subcategories with the highest proportion in the LAHF subcategory (Fig. 4A,B). The proportions of LAHF neurons were increased in Q175 compared to WT mice for both age groups during motion (2–3 months: χ² = 10.3, P < 0.001; > 12 months: χ² = 147.5, P < 0.001). This effect was enhanced with age due to a decrease in the proportion of WT LAHF neurons with age (χ² = 30.4, P < 0.001) coupled with an increase in this proportion subcategory in Q175 mice with age (χ² = 40.4, P < 0.001). In contrast, the proportions of neurons in the LALF subcategory were decreased in Q175 compared to WT mice at both age groups during motion (2–3 months: χ² = 17.8, P < 0.001; >12 months: χ² = 12.1, P < 0.001) (Fig. 4B). The proportion of neurons in the HAHF subcategory was not significantly different for Q175 and WT mice at 2–3 months but was significantly reduced for Q175 at >12 months during motion (χ² = 16.6, P < 0.001). With age both WTs and Q175 showed a decrease in the proportion of neurons in the HAHF category during motion and the effect was greater in the Q175 mice (WT: χ² = 6.6, P = 0.01; Q175: χ² = 30.3, P < 0.001) (Fig. 4B). Overall, the subcategory with the lowest proportions of neurons during motion was HALF. The proportion of HALF neurons showed a reversal with age. At 2–3 months, the proportion of neurons in the HALF subcategory during motion was increased for Q175 compared to WT mice (χ² = 7.1, P < 0.01). At >12 months there were proportionately fewer Q175 neurons in this subcategory during motion (χ² = 82.0, P < 0.001) (Fig. 4B). This reversal during motion appeared to be primarily due to an increase in proportion of neurons in the HALF subcategory in WTs (χ² = 104.3, P < 0.001) and a simultaneous decrease in the proportion of neurons in this subcategory in Q175 mice (χ² = 12.7, P < 0.001) (Fig. 4B).

During nonmotion, it was apparent that there was generally less variability of the distributions of neurons in each of the subcategories compared to motion at both ages (compare Figs 4A and 5A). In addition, subcategory comparisons were unique during periods of motion compared to nonmotion. However, similar to the subcategory classification for motion, the greatest proportions of neurons during nonmotion were in the low amplitude categories. The proportion of neurons in the LAHF subcategory during nonmotion was increased in Q175 compared to WT mice at 2–3 months (χ² = 56.2, P < 0.001) but did not differ in the >12 months group. This age-dependent difference in effect was due to a significant increase in the proportion of neurons in the LAHF subcategory in WT mice with age during nonmotion (χ² = 37.9, P < 0.001) (Fig. 5B). The proportion of neurons in the LALF subcategory during nonmotion was decreased at 2–3 months in Q175 compared to WT mice (χ² = 87.8, P < 0.001) but increased at >12 months in Q175 compared to WT mice (χ² = 142.9, P < 0.001) (Fig. 5B). This reversal was due to a decrease in the proportion of neurons in the LALF subcategory in WT mice with age (χ² = 96.2, P < 0.001) coupled with an increase in this proportion in Q175 mice with age during nonmotion (χ² = 137.2, P < 0.001) (Fig. 5B). The proportion of neurons in the HAHF subcategory during nonmotion increased at 2–3 months in Q175 compared to WT mice (χ² = 62.4, P < 0.001) but decreased at >12 months in Q175 compared to WT mice (χ² = 135.3, P < 0.001) (Fig. 5B). This reversal was due to both an increase in the proportion of neurons in the HAHF category in WTs with age (χ² = 69.5, P < 0.001) and a decrease in the proportion in Q175s with age during nonmotion (χ² = 125.6, P < 0.001) (Fig. 5B). The proportion of neurons in the HALF subcategory was decreased in Q175 compared to WT mice at both ages during nonmotion (2–3 months: χ² = 22.1, P < 0.001; >12 months: χ² = 67.3, P < 0.001), and this effect was increased with age because there was a greater decrease in the proportion of neurons in this subcategory in Q175 mice compared to WT mice during nonmotion (WT: χ² = 5.7, P < 0.05; Q175: χ² = 58.7, P < 0.001) (Fig. 5B).

Overall, significant alterations in frequency and amplitude of Ca²⁺ transients were observed in both models. Reduced amplitude of Ca²⁺ transients was very consistent between models and across ages, while changes in frequency were more sui generis and complex. In R6/2 mice, reduced frequency occurred during both motion and nonmotion epochs, whereas in presymptomatic and symptomatic Q175 mice, the frequency of Ca²⁺ transients increased significantly during motion. However, during nonmotion epochs, the effects were reversed with age; there was increased frequency in presymptomatic Q175 mice compared to WT littermates but decreased frequency in symptomatic Q175 mice compared to WT littermates, consistent with decreases observed in R6/2 mice. Thus, the most consistent change in symptomatic HD mice occurred during nonmotion epochs and was manifested by a reduction in frequency and amplitude of Ca²⁺ transients compared to WT littermates.

Correlations between the Occurrence of Ca²⁺ Transients in Neuron Pairs of R6/2 Mice

As a measure of motor cortex network synchrony, we determined the mean Pearson-Product moment correlation coefficient from deconvolved traces of significantly correlated neuron pairs at zero-lag based on a Monte Carlo simulation and the proportion of significantly correlated pairs. As pointed out above, the cutoff for significantly correlated pairs was P < 0.01, and only correlation coefficients above the cutoff were included in the data set to be analyzed. As expected, when comparing the mean correlation coefficient value during motion and nonmotion in WT mice alone, the correlation coefficient was higher during motion than during nonmotion periods (genotype x behavior: F_1,190 = 10.2, P < 0.001, post hoc: t = 2.17, P < 0.05) (Fig. 6A). In contrast, in R6/2 mice, the correlation coefficient was lower during motion than during nonmotion (post hoc: t = 2.36, P < 0.05), providing further evidence of decoupling between Ca²⁺ transient activity and behavior. During motion, R6/2 neuron pairs were significantly less correlated than WT neuron pairs (post hoc: t = 4.80, P < 0.001). However, during nonmotion the magnitude of correlations did not significantly differ between R6/2 and WT neuron pairs (Fig. 6A,B). There were no significant effects of distance during motion or nonmotion when the changes in mean correlation coefficients were examined as a function of distance between neuron pairs (Fig. 6B). In addition, there were no significant differences in autocovariance or cross-covariance between WT and R6/2 mice (data not shown).

Figure 6.

(A) Mean significant Pearson Product-moment correlation coefficient during motion and nonmotion collapsed across distance for HD (gray) and WT (black) mice. Left graph shows distance-collapsed correlation values for R6/2 mice (WT: n = 45, Q175: n = 52) during motion (left) and nonmotion (right). Right graph shows distance-collapsed correlation values for Q175 mice at 2–3 months (WT: n = 60, Q175: n = 79) and >12 months (WT: n = 47, Q175: n = 43) during motion (left) and nonmotion (right). Error bars indicate standard error of the mean of each movie. (B) Mean significant Pearson Product-moment correlation coefficients during motion (top) and nonmotion (bottom) for mice as a product of distance between neuron pairs. Left graph shows correlations by distance for R6/2 mice (WT: n = 45, Q175: n = 52) during motion (top) and nonmotion (bottom). Middle graph shows correlations by distance for Q175 mice at 2–3 months (WT: n = 60, Q175: n = 79) during motion (top) and nonmotion (bottom). Right graph shows correlations by distance for Q175 mice at >12 months (WT: n = 47, Q175: n = 43) during motion (top) and nonmotion (bottom). Error bars indicate standard error of the mean of each movie. ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001 as determined by two-way ANOVAs with post hoc Holm-Sidak tests.

Although the magnitude of correlations differed during motion and nonmotion between WT and R6/2 mice, the proportion of correlated neuron pairs was not significantly different (Fig. 7A). However, in both WT (behavior: F_1,190 = 12.1, P < 0.001, post hoc: t = 2.5, P < 0.05) and R6/2 (post hoc: t = 2.4, P < 0.05) mice, the proportions of significant correlations were decreased during periods of motion compared to nonmotion suggesting a reduced network of neurons might be recruited during motion (Fig. 7A). When the intercorrelations were separated as a function of distance (from 100 to 400 μm), there was a significant main effect of distance on the proportion of significantly correlated neurons during both motion (F_3,33 505 = 3.3, P = 0.02) and nonmotion (F_3,46 744 = 2.8, P = 0.04) (Fig. 7B). As expected, proportionately more pairs were significantly correlated when they were less than 100 μm apart compared to when they were 400 μm apart both during motion (post hoc: t = 2.90, P < 0.01) and nonmotion (post hoc: t = 2.79, P < 0.01) (Fig. 7B). There was no significant interaction between distance and genotype nor genotype differences on the proportion of significant correlations during motion or nonmotion (Fig. 7B).

Figure 7.

(A) Mean proportion of significant Pearson Product-moment correlation coefficients during motion and nonmotion collapsed across distance for HD (gray) and WT (black) mice. Left graph shows distance-collapsed proportions for R6/2 mice (WT: n = 45, Q175: n = 52) during motion (left) and nonmotion (right). Right graph shows distance-collapsed proportions for Q175 mice at 2–3 months (WT: n = 60, Q175: n = 79) and >12 months (WT: n = 47, Q175: n = 43) during motion (left) and nonmotion (right). Error bars indicate standard error of the mean of each movie. (B) Mean proportion of significant Pearson Product-moment correlation coefficients during motion (top) and nonmotion (bottom) for R6/2 mice (WT: n = 45, Q175: n = 52) as a product of distance averaged by movie. Left graphs shows proportions by distance for R6/2 mice during motion (top) and nonmotion (bottom). Middle graph shows proportions by distance for Q175 mice at 2–3 months (WT: n = 60, Q175: n = 79) during motion (top) and nonmotion (bottom). Right graph shows proportions by distance for Q175 mice at >12 months (WT: n = 47, Q175: n = 43) during motion (top) and nonmotion (bottom). Error bars indicate standard error of the mean of each movie. ^*P < 0.05 and ^**P < 0.01 as determined by two-way ANOVAs with post hoc Holm-Sidak tests.

Correlations between the Occurrence of Ca²⁺ Transients in Neuron Pairs of Q175 Mice

Overall, there was a decrease in the mean correlation coefficient values exceeding P < 0.01 in Q175 compared to WT mice. During both motion and nonmotion, there was significant main effect of genotype on mean correlation coefficients between neurons pairs (motion: F_1,234 = 4.6, P < 0.05; nonmotion: F_1,234 = 8.7, P < 0.01) (Fig. 6A). However, post hoc t-tests revealed that during motion, the mean correlation coefficients were only significantly decreased in Q175 compared to WT mice in the > 12 months age group (post hoc: t = 2.5, P < 0.05) whereas during nonmotion, the mean correlation coefficients were significantly decreased in Q175 compared to WT mice at both the 2–3 months age group (post hoc: t = 2.5, P < 0.05) and the >12 months age group (U = 1272.0, P < 0.05) (Fig. 6A). When the changes in mean correlation coefficients were examined as a function of distance between neuron pairs, there were no significant effects of distance either during motion or nonmotion (Fig. 6B). However, during motion there was a significant decrease in mean correlation coefficients in Q175 compared to WTs at >12 months (genotype: F_1,327 = 31.9, P < 0.001). During nonmotion, there were also significant decreases in mean correlation coefficient in Q175 compared to WT mice at both ages (2–3 months genotype: F_1,554 = 27.2, P < 0.001; >12 months genotype: F_1,324 = 9.1, P < 0.01) (Fig. 6B). There were no significant differences in autocovariance or cross-covariance between WT and Q175 mice (data not shown).

Similar to mean correlation coefficient values, the overall mean proportion of correlation coefficients exceeding P < 0.01 was significantly decreased during motion in Q175 compared to WT mice at both ages (genotype: F_1,234 = 4.5, P < 0.05; 2–3 months Mann–Whitney test P = 0.375, > 12 months P = 0.104) (Fig. 7A). When neuron pairs were separated into bins by distance during motion, there was a significant decrease in the proportion of significant correlation coefficients for longer distances at both ages regardless of genotype (Fig. 7B). In both 2–3 months Q175 and WT mice, the proportion of significant correlations was decreased during motion at 300 and 400 μm intervals between pairs (F_3,581 = 4.0, P < 0.01, post hoc: t = 2.7, P < 0.01 for 100 vs 300 μm, t = 3.2, P < 0.01 for 100 vs 400 μm) (Fig. 7B). At >12 months during motion, the proportion of significant correlations was significantly decreased at the 400 μm interval in both Q175 and WT mice (F_3,351 = 3.8, P = 0.01, post hoc: t = 3.2, P < 0.01 for 100 vs 400 μm). There also was a significant decrease in the proportion of correlations exceeding P < 0.01 level during motion in >12 months Q175 mice compared to WTs (F_1,351 = 9.3, P < 0.01) (Fig. 7B). During nonmotion, there were no significant differences in the proportion of significant correlations between WT and Q175 mice (Fig. 7A). Again, when neuron pairs were separated by distance, there were significant decreases in the proportion of correlation coefficients exceeding P < 0.01 for the longer distances at both ages but there were no genotype differences at either age (Fig. 7B). In both 2–3 months Q175 and WT mice, the mean proportions of correlations exceeding P < 0.01 were significantly decreased during motion at 300 and 400 μm intervals between pairs (F_3,581 = 4.0, P < 0.01, post hoc: t = 2.8, P < 0.01 for 100 vs 300 μm, post hoc: t = 3.2, P < 0.01 for 100 vs 400 μm) (Fig. 7B). At >12 months during nonmotion, the proportion of significant correlations decreased at the 400 μm interval (F_3,351 = 3.0, P < 0.05, post hoc: t = 2.9, P < 0.01 for 100 vs 400 μm) (Fig. 7B).

Interneuronal correlations represent bona fide surrogate markers of neuronal synchrony. The present results demonstrated significant reductions of interneuronal correlation coefficients in symptomatic HD mice. Interestingly, this effect was also present in presymptomatic Q175 mice during nonmotion epochs suggesting early deficits in neuronal synchrony that nonetheless is not reflected in behavioral deficits. This probably indicates that in presymptomatic Q175 mice, the motor cortex network is able to compensate for disruptions in connectivity during motion, but this compensation is no longer present at symptomatic ages. An important difference between models was observed in terms of the proportion of significant interneuronal correlations. Genotype effects in the proportion of significant correlations were only seen in Q175 mice and only during periods of motion, whereas in R6/2 mice, both WT and R6/2 mice showed a significant increase in the proportion of significant correlations during nonmotion.

Discussion

The main finding of this study is that significant alterations in behavior, cortical neuronal network activity, and intercorrelations between the activity patterns of pairs of neurons occur in both R6/2 transgenic and Q175 heterozygotic knock-in mice compared to their WT littermates. A number of these changes were similar in the two HD models, but others were specific to each model. Symptomatic R6/2 mice spent more time in motion than WTs. In contrast, symptomatic (>12 months) Q175 mice were less mobile. The amplitude of Ca²⁺ transients was reduced similarly in both models, but changes in frequency were more complex; for R6/2 mice, the frequency was reduced during both motion and nonmotion, whereas in symptomatic Q175 mice the reduction was observed only during nonmotion. Interestingly, in presymptomatic Q175 mice, the frequency was increased during both behavioral states. The correlation coefficients between pairs of neurons were generally decreased in both models except during nonmotion epochs in R6/2 mice. These results point to clear disruptions of cortical ensemble activity in HD; however, the effects of the HD mutation have similar and contrasting effects depending on the mouse model used as well as the stage of disease progression and the behavioral state. To the best of our knowledge, this is the first 2PLSM study examining changes in network dynamics in the cerebral cortex of awake, behaving pre- and symptomatic HD mouse models.

Examination of temporal and spatial correlations in cortical neuronal ensembles using 2PLSM and Ca²⁺ imaging has become a powerful tool to understand how local circuits behave in normal and pathological conditions (Peron et al. 2015; Goel et al. 2018). This is particularly important in HD as there is an important correlation between structural changes in the cerebral cortex and disease pathology in striatum (Rosas et al. 2008; Thu et al. 2010; Waldvogel et al. 2015). In addition, electrophysiological studies in mouse models have demonstrated abnormal discharge patterns and altered synaptic activity (Cummings et al. 2007; Walker et al. 2008; Cummings et al. 2009; Miller et al. 2011). However, these studies were limited in terms of spatial and temporal resolution as well as the number of neurons recorded at any given time. In the present study, we were able to visualize the activity of up to 50 neurons on average in the same recording session and correlate this activity with two behavioral states.

Behavioral Alterations in HD Model Mice

Multiple studies in rodent models of HD have reported behavioral changes that increase in severity with disease progression (Mangiarini et al. 1996; Slow et al. 2003; von Horsten et al. 2003; Gray et al. 2008; Heng et al. 2008; Menalled et al. 2009; Heikkinen et al. 2012). Symptomatic R6/2 mice display clasping and deficits in a number of behavioral tests including open field, rotarod, grip test, and running wheel (Mangiarini et al. 1996; Carter et al. 1999; Hickey et al. 2008; Andre et al. 2010; Reidling et al. 2018). Here, under head-restraint conditions, symptomatic R6/2 mice were more active than WTs, displaying a greater number of epochs of activity. This finding was unexpected and could be due to a decreased capacity of symptomatic R6/2 mice to maintain balance (Carter et al. 1999; Stack et al. 2005). To compensate for such a deficit, animals adjust their center of balance resulting in more measured motion. In addition, weight loss and reduced strength of skeletal muscles, which also express mHtt, could play an important role. Last but not least, running and incoordination when suspended on a rotating ball could also be influenced by the stress induced by head-restraint. It is known that R6/2 mice are more susceptible to stressful conditions than WTs (Mo et al. 2014; Dufour and McBride 2016).

Increased amounts of motion did not occur in Q175 mice even at fully symptomatic ages. In fact, with disease progression, the amount of motion decreased. This observation is consistent with previous behavioral studies in unrestrained conditions (Heikkinen et al. 2012; Menalled et al. 2012). The reason for this difference in behavior between R6/2 and Q175 mice under head-restraint is unknown. However, some of the behavioral changes in Q175 mice, particularly in heterozygotes, tend to be milder compared to those in R6/2 mice. Given that we saw behavioral differences, we separated Ca²⁺ imaging data into periods of motion and nonmotion.

Network Dynamics in the Cerebral Cortex of HD Model Mice

A similar number of videos per mouse were recorded in R6/2 mice compared to WT mice, but we found significantly fewer neurons per video in R6/2 mice. We also recorded a similar number of videos per mouse in both Q175 cohorts and a similar number of neurons per video. The reason for the reduced number of neurons found in R6/2 mice is presently unknown. Increased cell loss is unlikely as cortical neurodegeneration occurs very late in this model and mostly affects the cingulate cortex (Davies et al. 1997; Turmaine et al. 2000; Rattray et al. 2013), an area not included in our recordings. One possibility is that viral infection is less efficient in R6/2 brains. However, in our previous studies using this mouse model, we did not observe differences in viral transfection (Cepeda et al. 2013; Parievsky et al. 2017). Another possibility is that basal Ca²⁺ levels are increased in cortical neurons and, as a result, those cells would appear brighter and fluctuations in Ca²⁺ levels would be harder to detect. Previous studies have shown Ca²⁺ dysregulation in striatal and cortical neurons from HD mouse models (Andre et al. 2006; Bezprozvanny 2009) as well as increased basal Ca²⁺ levels (Hodgson et al. 1999; Hansson et al. 2001), which could preclude detection of further changes in Ca²⁺. Finally, unwanted side-effects of the GCaMP protein could compound with the HD mutation to enhance excitotoxicity. For example, GCaMP-containing calmodulin has been shown to interfere with the gating of L-type channels, thereby disrupting Ca²⁺ dynamics and gene expression (Yang et al. 2018). In spite of this limitation, we were able to unravel significant alterations of network dynamics in the cortex of HD model mice.

The most prominent physiological change was a reduction in the amplitude of Ca²⁺ transients in both symptomatic R6/2 mice and Q175 mice in both age cohorts. This amplitude reduction was apparent during periods of motion and nonmotion. In addition, the amplitude reduction became more pronounced with age in the Q175 mouse model suggesting that reductions in amplitude of Ca²⁺ transients may be a marker of disease progression. Since the amplitude of the Ca²⁺ transients is generally considered a measure of neuronal bursting (Rothschild et al. 2010; Chen et al. 2013), reduced amplitude of Ca²⁺ transients supports the hypothesis that cortical neurons in behaving R6/2 and Q175 mice probably do not burst as frequently as neurons in WT mice (Miller et al. 2011). One caveat is that lower amplitudes could also be affected by GCaMP6f expression or, as mentioned above, altered basal Ca²⁺ levels.

In terms of neuronal ensemble behavior, the frequency of Ca²⁺ transients was disrupted in the two models of HD, but the direction of changes was different. In R6/2 mice, the frequency of Ca²⁺ transients was lower than WTs during both motion and nonmotion. This finding points to a dissociation between behavioral state and Ca²⁺ transients in the motor cortex of this model. In contrast, as expected, in the Q175 model the frequency of Ca²⁺ transients was decreased during nonmotion but increased during motion. Notably, at the presymptomatic stage, the frequency of Ca²⁺ transients was increased during both behavioral states. This observation points to biphasic changes in cortical activity, similar to our previous observations of time-dependent alterations along the corticostriatal pathway (Joshi et al. 2009). These results are in general agreement with a recent study showing increased frequency of Ca²⁺ transients in presymptomatic R6/2 mice (Burgold et al. 2019). Data from the oldest mice of both sexes showed the greatest difference in amplitude and frequency of Ca²⁺ transients suggesting that these differences become more pronounced with age. This is consistent with other findings that alterations in CAG knock-in mouse models are exaggerated beyond 12 months (Lerner et al. 2012; McCourt et al. 2016).

When we subcategorized neurons based on their amplitudes and frequencies, we noted a marked increase in variability during motion compared to nonmotion in both R6/2 and Q175 mouse models. This result was independent of genotype and could indicate that the activity of a subpopulation of neurons was more drastically altered in the motion state than other neurons. We found the greatest proportion of neurons tended to be low amplitude neurons in both R6/2 and Q175 mouse models, and during motion, the neurons tended to shift away from being LALF neurons. This suggests that the LALF neurons could be the subpopulation of neurons that are more susceptible to activity changes during motion.

As a general trend, we saw that the correlation coefficients between significantly correlated pairs of neurons were decreased in R6/2 and Q175 mice compared to their WT littermates. In R6/2 mice, this decrease in intercorrelations was apparent only during periods of motion, whereas in Q175 mice, the disruptions in intercorrelations began during periods of nonmotion and developed into periods of motion with age. The proportion of significant correlations was only affected in Q175 mice and only during periods of motion. This suggests that in R6/2 mice, the size of the cortical motor network did not differ but the connectivity within the network was decreased as indicated by a decrease in intercorrelations. In Q175 mice, the proportions of significant correlations were decreased compared to WTs suggesting fewer neurons were involved in the motor network as well as a decrease in the connectivity within the network. Finally, although not genotype specific, both WT and HD mice showed a distance-dependent decrease in proportion of significantly correlated neuron pairs. This is consistent with other findings that suggest the cortex uses microcircuits of coordinated activity (Dombeck et al. 2009; Komiyama et al. 2010).

In conclusion, our results show that there is a clear disruption in cortical network activity in HD mice. However, some of the disruptions are model specific. Differences between the models themselves such as differences in age, method of genetic alteration (transgenic vs knock-in), and different background strains could explain the contrasting findings. For example, it has been noted that genetic background affects neuronal firing activity and locomotor behavior between WT lines (Walker et al. 2008). In particular, the 129sv strain (CAG140 knock-in, the parent line from which Q175 mice were derived) had a faster spontaneous firing and burst rate than the CBA × C57BL/6 hybrid (R6/2 line). They also demonstrated that C57BL/6 mice tend to perform better on learning and memory tasks than other strains, including CBA and 129sv. Important in the present context is the fact that 129sv mice exhibit less exploratory and locomotor behavior than C57BL/6 J mice.

In addition, R6/2 model mice replicate the juvenile form of HD whereas the Q175 is a better model of adult-onset HD. Although the former progresses very fast and displays epileptic seizures, which may represent alterations in neuronal excitability and synchrony, the latter presents with a milder form of HD and no seizures. The fact that the mean correlation coefficients and proportion of significant correlations were significantly higher during nonmotion than during motion in symptomatic R6/2 mice, indicating again a dissociation between behavioral state and Ca²⁺ ensemble activity, could set the stage for increased synchrony and a proconvulsant state. In contrast, most of the cortical disruptions in Q175 mice occur early, evolve with disease progression, and are dependent, in most cases, on the behavioral state. The current limitations of our procedure which requires a minimum of a 2-week window for GCaMP6f to express prevents us from being able to analyze cortical activity in the presymptomatic R6/2 mouse. While our data come from cortical layer 2/3, it is possible that the changes we see in neuronal ensemble activity reflect the activity of deeper cortical layers given that cortical layer 2/3 neurons receive projections from deeper cortical layers (Bode-Greuel et al. 1987; Thomson and Bannister 2003).

Although, in general terms, our study agrees with a previous electrophysiological study (Walker et al. 2008), an important difference in the R6/2 mice is that in such study the overall firing rate of cortical neurons was increased whereas in our study the frequency of Ca²⁺ transients was reduced during both motion and nonmotion. It is important to emphasize that in Walker’s study, recordings were obtained from the prefrontal cortex. The only work from Rebec’s group looking at motor cortex was a more recent study in BACHD mice (Estrada-Sanchez et al. 2015). Interestingly, results from this model are at odds with their previous study on R6/2 and CAG140 mice. For example, in BACHD mice the firing rate of motor cortex neurons was reduced, not increased. Even more remarkable, there was no difference in burst rate or duration compared to controls and they also appeared to have higher, instead of lower, correlated activity. This just reinforces the idea that careful comparison between multiple models is imperative when trying to understand HD mechanisms.

Our finding that Ca²⁺ dynamics is altered in CPNs from HD models represents a step forward in our understanding of human HD. For example, reduced amplitude of Ca²⁺ transients and reduced intercorrelations between CPNs could help explain cognitive and motor deficits in HD patients. Further, if indeed Ca²⁺ handling and basal Ca²⁺ levels are altered in diseased mice and humans, we could hypothesize that such Ca²⁺ dyshomeostasis plays a role in CPN loss. Overall, our data suggest that the cerebral cortex is an important target for early treatment and prevention (Wang et al. 2014; Estrada-Sanchez et al. 2015).