Automated quantification of head-twitch response in mice via ear tag reporter coupled with biphasic detection

Abstract

Background:

Head-twitch response (HTR) is a manifestation of the serotonergic system behavioral pharmacology commonly used as a proxy of psychedelic drug action in rodents.

New method:

We developed a minimally invasive magnetic ear tag reporter and designed a detection system that performs a comprehensive characterization of each potential HTR event on an electromagnetic readout.

Results:

Magnetic ear tags were easy to install and generally well tolerated by the animals. On the low-threshold first phase of detection, the tags signal recorded in a magnetometer was filtered and screened for potential HTR events. On the second phase, the detector performed a comprehensive spectral analysis evaluation of each event and identified the HTR characteristic distribution of power density. Our system delivered satisfactory performance in the identification of pharmacologically-induced HTR and discrimination power against common non-HTR behaviors.

Comparison with existing methods:

Our system offers a high-throughput solution for studying HTR in mice employing minimally invasive procedures and superior standalone discriminative power compared to our previously reported fully-automated approach.

Conclusions:

High-throughput identification of HTR utilizing magnetic ear tagging and biphasic detection delivers satisfactory detection and discrimination power employing less invasive procedures.

1. Introduction

Psychedelic drugs such as lysergic acid diethylamide (LSD), psilocybin, dimethyltryptamine (DMT) or 2,5-dimethoxy-4-iodoamphetamine (DOI) produce in humans a characteristic altered state of consciousness that includes profound perceptual and cognitive disturbances (Nichols, 2016). Although these drugs bear a complex poly-pharmacology, their serotonergic action appears to be crucial to their actions. More specifically, their ability to activate the serotonin (or 5-hydroxytryptamine) 2A receptor (5-HT_2AR) appears to be central to these drugs ability to elicit psychedelic effects (Vollenweider et al., 1998). Interestingly, in rodent models, these drugs produce a characteristic rotational movement of the head of the animal commonly known as head-twitch response (HTR) (González-Maeso et al., 2007). Given the well-established qualitative match between drugs that elicit psychedelic action in humans and drugs that induce HTR in rodents, this behavior has been employed to model psychedelic activity, and as a surrogate to study the 5-HT_2AR behavioral pharmacology (Hanks and González-Maeso, 2013).

Assessment of HTR traditionally relied on the visual identification by direct observation of mouse behavior (González-Maeso et al., 2007). More recently, efforts have been redirected toward less tedious and time-consuming techniques based on principles of electromagnetic induction. These techniques involve the surgical implantation of a magnet on the surface of the mouse skull that act as a reporter of the mouse head movement. Assessment of the voltage signal generated by the rapid movement of the animal head in a coil that serves as a magnetometer can be performed visually by identifying the characteristics that define a HTR wavelet (Halberstadt and Geyer, 2013). Alternatively, as we previously reported, detection can be automated by identifying maxima in the filtered signal (de la Fuente Revenga et al., 2019). While our previously reported automated detection of HTR significantly reduces processing time compared to any visually-based analysis, and delivers satisfactory performance under standard testing conditions of pharmacologically-induced HTR by psychedelic drugs, this methodology may bear an unsatisfactory high rate of false positive detection in mice that exhibit jumping behavior.

To simplify the operational procedures and maintain the high-throughput capabilities, we developed a small magnetic ear tag reporter that substitutes the previously reported surgical procedure required for implanting magnets on the animal head. Additionally, considering the limitation of our previous automated detection routine, we have developed and validated a standalone multiphasic detection system capable of providing superior discrimination power against non-HTR behaviors.

2. Methods

2.1. Animals

Experiments were performed on adult (8-16 weeks old) C57BL/6 (Taconic) male mice. Except during testing, animals were housed in groups of up to 4 littermates with food and water ad libitum in a vivarium with a 12h light/dark cycle at 23°. Experiments were conducted in accordance with NIH guidelines, and were approved by the Virginia Commonwealth University Animal Care and Use Committee. All efforts were made to minimize animal suffering and the number of animals used.

2.2. Drugs

(±)-2,5-Dimethoxy-4-iodoamphetamine hydrochloride (DOI), morphine sulphate pentahydrate and naloxone hydrochloride dihydrate were purchased from Millipore Sigma. (±)-SKF38393 hydrobromide and M100907 were purchased from Tocris. All drugs were dissolved in saline (0.9 % NaCl) to the appropriate volume (5-10 μl/g) and concentration for administration (i.p.). The hydrochloride salt of M100907 was formed in situ by addition of 1 eq. of HCl. Vehicle-treated condition denotes injection of saline (i.p.) to the equivalent volume of the drug administered.

2.2. Magnetic ear-tagging

The detection of the rapid rotational head movement that characterizes HTR (Fig. 1A) by electromagnetic induction requires the installation of a magnetic reporter. In this case, we employed small neodymium magnets (N50, 3 mm diameter x 1 mm height, 50 mg) glued with cyanoacrylate to the top surface of aluminum ear tags for rodents (Las Pias Ear Tag, Stoelting Co.), and covered them with a nitrocellulose coating to reduce exposure of the animal skin to the magnet (Fig. 1B inset). Mice were ear-tagged bilaterally through the pinna antihelix. The ear tags were placed so that the magnet would rest in the interior of the antihelix, lying on top of the antitragus, leaving the ear canal unobstructed (Fig. 1B). The back part of the ear tag rested on the opposite side of the pinna. This procedure could be performed by simple restraining and immobilization of the mouse’s head, however, greater precision can be attained employing sedation or a dose of anesthesia sufficient to reduce the movement of the animal. For this experimental procedure, a standard dose of ketamine and xylazine (120 mg/kg and 12 mg/kg, respectively) anesthesia were employed. Although mild reddening of the ear was observed, the magnetic tags were generally well tolerated by the animals, even if housed in groups.

Fig 1.

Head-twitch response (HTR) detection based on magnetic tagging.

(A) Schematic representation of the rotational movement of the head characteristic of a HTR.

(B) Detail of the location of the magnetic tag in the mouse ear. Inset (dashed blue rectangle) shows the detail of the tag with the neodymium magnet attached to the surface and the back part that holds the tag in place.

(C) Schematic representation of the two phases of HTR detection.

Voltage signal, spectral density and derivative of the spectral density corresponding to a HTR induced by DOI (D), ear-flicking during grooming induced by SKF38393 (E) and jumping during naloxone-precipitated opiate withdrawal (F).

(D through F, left panel). Unprocessed recorded signal (blue trace) and its corresponding trace after processing (orange trace). Shown are the width (double-head orange arrow) and magnitude (inverted orange triangle) for each peak, along with the magnitude threshold (dotted black line) of the detector first screening phases. The green double-inverted head arrow shows the segment employed in the spectral analysis (secondary detection phase).

(D through F, center panel). Periodogram showing the power spectral density distribution (solid color area) as a function of frequency for events that by-passed the first phase of detection. Graphical depiction of filters employed in the second phase of the detector: power threshold (red line), cumulative sum area (solid black line rectangle, 70-110 Hz band) and the exclusion region for global power maximum corresponding to a frequency <35 Hz (blue translucent rectangle).

(D through F, right panel). First derivative of the spectral density, zero-line (dotted black line). Sum of zero-crossings of the derivative (5-200 Hz) employed in the secondary phase of detection.

2.3. Data acquisition

The animals were allowed to recover for 1 week prior to testing. Repeated testing was performed after a washout period of 1 week in between testing sessions. Data acquisition was performed as previously described with minor modifications (de la Fuente Revenga et al., 2019). Briefly, the tagged animals were placed inside a plastic container surrounded by a coil (~500 turns 30 AWG enameled wire) which output was amplified (Pyle PP444 phono amplifier) and recorded at a 1000 Hz using a NI USB-6001 (National Instruments) data acquisition system. Recordings were performed using a MATLAB (Mathworks, R2018a version) driver with the corresponding National Instruments support package for further processing. Experiments involving visual identification of mouse behavior were videotaped at 60 FPS and 1080p resolution.

2.4. Data processing and HTR detection

The signal processing script can be found in the supporting materials section. Firstly, the recorded signal was band-passed through a finite impulse response (FIR) filter (70-110 Hz). The absolute values of the filtered signal were doubled, had the baseline subtracted and were smoothed by applying a moving average. Screening of potential HTR was performed by identifying the maxima in the resulting peaks. More specifically, a potential HTR was annotated when it corresponded to a peak that met the following conditions (Fig. 1C): i) had a magnitude exceeding a threshold value of 0.02 |V|, ii) bore a ratio of prominence (maximum peak value relative to baseline) vs. peak height (maximum peak value relative to 0) below 0.95, iii) had a width (measured at half peak height) >20 and <150 ms and iv) was separated by a minimum of 200 ms from any other peak (Fig. 1D through 1F, left panels).

Segments of unprocessed data around the time stamps of the peaks that met all four conditions abovementioned were subjected to spectral density analysis in the second stage of the detector. The segments analyzed were ±2 x width length relative to the time stamp of each event. Their corresponding periodograms were computed by Fast Fourier Transform (Fig. 1D through 1F, center panels). The spectral density-based detection of HTR events analyzed the power density distribution in different bands of frequencies and changes in the derivative. More specifically, an event annotated in the first phase was confirmed to correspond to a HTR if its spectral analysis met the following conditions (Fig. 1C): v) maximum in the 70-110 Hz band of the spectrum exceeding a power threshold value of 0.005 V²/Hz, vi) cumulative sum of all power values in the 70-110 Hz band in excess of 0.05 V²/Hz, vii) frequency corresponding to the absolute maximum within the operative frequency range of the spectrum > 35 Hz and viii) sum of zero-crossings of the derivative of the spectrum density < 40 (between 5-200 Hz; Fig. 1D through 1F, right panels). For each event, the script annotated the parameters employed that characterization each event on a spreadsheet and generated an interactive graph should further visual inspection be needed.

3. Results

3.1. Magnetic ear tag-based detection of head-twitch responses

Rapid head movement that occurs during HTR in mice surgically implanted with a small neodymium magnet produces a characteristic sinusoidal voltage signal in the output of a coil as a result of the acceleration the magnet is subjected to during this event (Halberstadt and Geyer, 2013; de la Fuente Revenga et al., 2019). The spectral analysis of this wavelet is characterized by strong power density and maxima in two frequency bands (40-50 Hz and 80-100 Hz). Similarly, HTR (Fig. 1A) in mice bearing magnetic tags on both ears produced a sinusoidal wavelet of comparable appearance (multiple bipolar peaks), duration (~100 ms) and power density distribution in the 40-50 and 80-100 Hz frequency bands (Fig 1D). The raw signal corresponding to non-HTR events such as rearing and most grooming events produced minimal changes in the voltage signal relative to the baseline (not shown). Conversely, in some occasions, jumping and flicking of the ear tag following ear scratching or grooming resulted in irregular wavelets, abrupt voltage deflections or a combination of both (Fig. 1E and 1F, see also Supp. Fig. 1). Automated detection of HTR events was divided in two phases. The first phase was similar to the signal processing we previously reported for detection of HTR in mice bearing surgically implanted magnets (de la Fuente Revenga et al., 2019). In this case, we utilized the doubled moving-average smoothed absolute signal filtered between 70-110 Hz. Similarly, HTR appeared as unipolar peaks of width comparable to the duration of the event and magnitude proportional to the power density in the 70-110 Hz frequency band (Fig 1D, left and center panels). Detection parameters in this first stage included absolute peak prominence, prominence over baseline, duration of the event and separation from other peaks (see Fig. 1C and methods for details). As the first stage of the detector was intended as a screening for potential HTR events, we chose a highly permissive peak magnitude value. While this stage of detection captured HTR events of different magnitudes it also carried over some non-HTR behaviors that resulted in irregular wavelets (Fig. 1E and 1F, left panels). On the other hand, rapid voltage deflections were successfully screened out at this stage (Supp. Fig. 1).

This specificity issue was addressed in the second phase of the detector that identified HTR based on the spectral characteristics of each event. As previously described, HTR showed strong power densities with maxima in two frequency bands (40-50 Hz and 80100 Hz) (Halberstadt and Geyer, 2013; de la Fuente Revenga et al., 2019) and a smooth spectral line (Fig. 1D center panel). Conversely, non-HTR events that bypassed the first filter showed weak power in the 70-100 Hz range, had an absolute maximum that generally corresponded to a frequency < 40 Hz and displayed a spiky spectral line (Fig. 1E and 1F, center panels). These differential characteristics were employed in the design of the second stage of the detector. This stage included evaluation of the maximum power and cumulative sum of power values within the 70-110 Hz band, frequency corresponding to the absolute maximum in the periodogram (Fig 1D through 1F, center panels) and number or zero-crossings by the derivative of the spectral density (Fig 1D through 1F, right panels) as direct measure of saw-like appearance of the spectral line (see methods for details).

3.2. Evaluation of the HTR detector performance

As abovementioned, certain non-HTR behaviors produce abrupt changes in the signal that can bypass the first screening stage of the detector. To evaluate the discriminative performance of the detector secondary stage at excluding behaviors prone to result in ear flicking events we induced grooming behavior following a previously described procedure (de la Fuente Revenga et al., 2019). Wood chips were intermittently sprinkled over 5 min on the head of tagged mice (n = 3). Upon visual assessment, none of the HTR identified by the detector corresponded with grooming events. Indeed, all HTR identified by the detector corresponded to actual HTR events registered on video. To further establish the discriminative power of the detector against grooming behavior we induced aggressive grooming bouts via administration of the dopamine D₁ agonist SKF38393 as previously described (de la Fuente Revenga et al., 2019). SKF38393 (10 mg/kg i.p.) was administered to mice bearing magnetic tags (n = 3) and the coil output was recorded for 15 min after the first 15 min following drug administration. In this case the recorded signal was analyzed visually. We found that all the events identified as HTR by the detection system were found to correspond with the characteristic signal of HTR occurrences supporting the notion that aggressive-grooming induced by SKF38393 does not interfere with the discriminative power of the detector. Considering the greater freedom of movement of a tagged ear compared to a magnet implanted on the animal skull, it is noteworthy that the multistage detector could successfully distinguish signal artifacts related to ear movements from HTR events.

Another source of abrupt changes in the voltage signal previously reported as a potential source of false positive detection is jumping behavior (de la Fuente Revenga et al., 2019). Jumping is observed in mice displaying escaping behavior and can be pharmacologically induced by precipitating opiate withdrawal (Kest et al., 2002). Tagged mice (n = 4) were administered naloxone (10 mg/kg, i.p.) that precipitated withdrawal in animals previously treated with morphine (50 mg/kg, i.p., 4 h prior). The mice were videotaped and the signal of the coils recorded during the 10 min following the administration of naloxone. After visual inspection of the video recordings, none of the HTR identified by the detection system was found to correspond with visually-identified jumping events (Fig. 2A). This confirmed that the double stage detector can, standalone, exert sufficient discriminative power to distinguish HTR from jumping events.

Fig 2.

(A) Graphical depiction of the HTR detector performance in pharmacologically-induced jumping. (B) Correlation between automated and visually identified HTR in mice treated with different doses of DOI (0 - 0.6 mg/kg, n = 4). (C) Time course showing the effect of DOI on HTR counts, and the blocking effect of pretreatment (5 min prior) with M100907. The arrow indicates the time of administration of DOI or vehicle. (D) Quantification of the total count of HTR events during the first 30 min following the administration of DOI (1 mg/kg, n = 3), DOI + M100907 (0.1 mg/kg, n = 3) or vehicle (n = 6). (E) Scatter plot of cumulative sum of power between 70-110 Hz as a function of peak magnitude. (F) 3-D scatter plot of peak magnitude (x-axis), frequency corresponding to the global maxima in the spectrum (y-axis) and sum of zero-crossing of the spectral line derivative (z-axis) for events classified by the detector as HTR (blue) and non-HTR (red) in mice treated with DOI. Values represent mean ± S.E.M (C,D). One-way ANOVA with Bonferroni’s post hoc analysis: ***P<0.001 (D)

DOI is a potent 5-HT_2A agonist known to induce robust induction of HTR in rodents and psychedelic effects in human (Canal and Morgan, 2012; Shulgin and Shulgin, 1997). We evaluated the performance of the detector by administering tagged mice different doses of DOI (0 mg/kg, n = 4; 0.3 mg/kg n = 4; 0.6 mg/kg n = 4), and assessed HTR employing the automated detector as well as visual analysis the mouse behavior on tape (Fig. 2B). Both assessment methods were highly correlated (Fig. 2B, F[1,22]=4298). The automated detection rate of false positive and negative detection of the automated system were 0.75% and 3.17% respectively (~397 visually-identified HTR). Next, we assessed the effect of DOI (1 mg/kg, i.p., n = 3) on induction of HTR over time (vehicle group, n = 6). As expected, DOI produced a robust increase in HTR counts that gradually decreased over time (Fig. 2C). These effects were successfully blocked by the 5-HT2AR-selective antagonist M100907 (0.1 mg/kg, i.p., n = 3, 5 min prior) (Fig. 2C). Over the first 30 min following the injection of DOI, the pre-treatment with M100907 completely blocked DOI-induced HTR (Fig 2D) which resulted in HTR counts comparable to those of vehicle-treated mice treated (One-way ANOVA: F[2,9]=1.982, P < 0.001; post hoc: Vehicle vs. DOI P < 0.001, Vehicle vs. M100+DOI P = 0.422, DOI vs. M100+DOI P < 0.001).

In addition to the classification of HTR or non-HTR events, the detector generates an interactive figure that allows the visual inspection of the classification relative to the actual signal and the values of the parameters analyzed (Supp. Fig. 2). These parameters are also annotated on a spreadsheet should further analysis be needed. We next aimed to visualize the distribution of HTR and non-HTR events as classified by the detector (478 and 148 respectively) in the mice treated with DOI (1 mg/kg) within an orthogonal space defined by some of the parameters used for detection. Firstly, as abovementioned, we confirmed that the peak magnitude (i) and the cumulative sum of power between 70-100 Hz (vi) were highly correlated, regardless of the event category (Fig. 2E). More importantly, non-HTR events appeared collapsed closer to the origin, while HTR clustered around a centroid located at higher peak magnitude and power values. Additionally, the 3-D scatter plot of peak magnitude (i), frequency corresponding to the absolute maximum on each spectral analysis (vii) and sum of zero-crossings of the derivative of the spectral line (viii) (Fig. 2F) revealed distinct populations within the HTR and non-HTR categories. Thus, HTR events were distributed in two clusters defined by the two frequency bands that agglutinate the greater spectral density characteristic of this behavior (~40-50 and 80-100 Hz). HTR were also characterized by greater peak magnitude power and count of zero-crossings in the derivative of the spectral line as anticipated. Compared to HTR events, non-HTR were more sparsely distributed and were generally characterized by lower peak power and greater zero-crossings in the derivative. Two clusters of non-HTR could be identified based on the frequency corresponding to the global maxima: one at frequencies < 40 Hz and another one around 90 Hz, but in appearance distinct from the HTR cloud. This succinct visual analysis confirms that the multistage detector is capable of successfully distinguishing different populations of events consistent with the characteristics of HTR signals.

4. Discussion

The magnetic ear tagging approach herein described offers a less invasive alternative for the study of HTR compared to surgical implantation of magnets (Halberstadt and Geyer, 2013; de la Fuente Revenga et al., 2019). Although sedation or anesthesia is recommended for greater accuracy, the tagging procedure can be satisfactorily be performed by manual restraining of the awake animal simplifying even further postoperative care and supervision. Another advantage of the tag system relates to its durability. In our experience, magnets surgically implanted can detach from the skull. This issue is virtually eliminated with the use of magnetic ear tags.

The magnetic ear tags were well tolerated in mice housed in groups. However, the use of tags does bear other limitations. Identification tags have been shown to lead to the appearance of ulcerations and deformation of the mouse ear (Kitagaki and Hirota, 2007). Similarly, we observed that some animals developed inflammatory processes ~2 months after the tagging. Possible non-mutually excluding explanations for these occurrences involve inflammatory reactions to the metallic makeup of the ear-tag and/or magnet in contact with the ear skin and the accumulation of dirt around the stem of the tag. Bearing this limitation in mind, HTR testing should be successfully performed within an operative range of ~6-8 weeks after the tagging. Another limitation relates to the use of adolescent and very young adults that, should the ear tags become attached by magnetic attraction, might not be strong enough to separate from their cage mates. Ensuring magnetic repulsion between the poles of the exposed magnet surfaces in all cage mates greatly reduces this issue. Alternatively, this inconvenience could be overcome by using smaller magnets or resorting to individual housing.

Compared to magnets physically attached to the skull surface, magnetic ear tags act as reporters of both the head and ear movement. Thus, certain behaviors such as ear scratching and grooming resulted in flicking of the ear/s that manifested in changes in the voltage signal. Ear flicking and jumping resulted in abrupt voltage deflections, irregular wavelets or the combination of both. Simple voltage deflections are appropriately buffered by the moving average; however, some irregular wavelets are able to by-pass the permissive threshold of the first stage of detection.

We previously reported an add-on secondary set up to discriminate jumping events based on the signal of a piezoelectric sensor located at the base of the magnetometer (de la Fuente Revenga et al., 2019). This approach, although successful, increased significantly the technical complexity of the hardware employed. As discussed in the results section, the smoothness of the spectral line and the spectral density distribution differ between HTR and non-HTR behaviors. These qualitative differences were successfully parameterized in the design of the secondary stage of the detector to identify HTR.

With regard to HTR detection, the multiphasic detector delivered a satisfactory performance as demonstrated by the high correlation against visually-detected HTR, and the low false positive and negative detection rates in mice treated with different doses of DOI. Moreover, we confirmed the involvement of the 5-HT_2AR in the action of DOI by blocking HTR with the selective antagonists M100907 (Canal and Morgan, 2012). Additionally, we evaluated the discriminative power of the detector against behaviors prone to result in abrupt changes in the voltage signal that could lead to false positive detection. Under our current testing conditions, the detector displayed remarkable discriminative power against grooming and jumping behaviors. Importantly, the combination of magnetic ear tags and detection based on spectral analysis for the identification of HTR retains the high-throughput capabilities of our previous approach while it delivers reliable measures of HTR occurrences even in the presence of common potential sources of false positive detection. These features, however, do not preclude performing the necessary preliminary observations when employing drugs or animal strains prone to exhibit other behaviors not contemplated here that could interfere with the detector performance.

5. Conclusion

Magnetic ear tagging is a suitable less-invasive alternative to surgical implantation of magnets for the study of HTR based on electromagnetic induction. The addition of a secondary detection stage based on spectral analysis of each potential HTR event delivers satisfactory sensitivity and specificity even in the presence of common sources of potential false positive detection.

Highlights.

• Ear tag reporter for the study of mouse behavior

• Multiphase detection of head-twitch

• Modelling serotonergic behavioral pharmacology