Transgenic mice
All procedures were approved by the Lilly UK Institutional Animal Care and Use Committee and were conducted in accordance with the Guide for Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, 1996) and the Animals (Scientific Procedures) Act 1986. All studies are reported in accordance with the ARRIVE guidelines for reporting experiments involving animals (27).
rTg4510 mice (129S6;FVB-Tg(Camk2a-tTA)1Mmay Tg(tet-o-MAPT*P301L)) were bred to be heterozygous for tetracycline and P301L alleles to produce regulatable double transgenics as first described (17, 20) by Taconic (Germantown, USA). Three of the four produced lines were used in the present study: heterozygous tetracycline transactivator carriers but wild-type for tau (tTA), wild-type for both tTA and tau (WT) and both tTA carriers and tau carriers (Tg). A total of 100 male mice (25 WT, 25 tTA, 50 Tg) were used in the study.
All mice underwent baseline T-maze and open-field locomotor activity (ofLMA) evaluation between 9 and 12 weeks of age. Tg mice were randomised to ensure that an even distribution of high- and low-performers from the T-maze received doxycycline treatment. Of the 50 Tg mice, 25 were given DOX treatment beginning at 13 weeks of age (two 10 mg/kg bolus oral doses of doxycycline followed by Harlan Teklab base diet 2016 containing 200 ppm doxycycline for the duration of the experiment). These transgene altered Tg mice were designated Tg + DOX.
Studies were run in a four-week cycle of tests - two weeks of EEG recordings with ad libitum food followed by two weeks of food restriction and behavioural testing (T-maze and ofLMA). The first week of EEG recordings were used for acclimatisation and were therefore not analysed. Animals completed seven test cycles, unless they were removed for welfare reasons, until they reached approximately 46 weeks old, when they were euthanised.
Surgical preparation
Mice underwent in-house surgery at 19.8–31.0 g and 11–17 weeks old. Subjects were anesthetized and sedated (2% isoflurane in 100% oxygen, 0.1 mg/kg medetomidine HCl subcutaneous (SC) injection) and fitted with a cranial implant affixed to the skull by dental acrylic and cyanoacrylate. The implant consisted of five stainless steel screws for local field potential (LFP) electroencephalogram (EEG) recording (two frontal screws placed 2 mm anterior to bregma and 2 mm on each side of the sagittal suture line, two occipital screws placed 3 mm caudal from bregma and 3 mm on each side of the sagittal suture line and one ground screw over the cerebellum). Two Teflon-coated stainless-steel wires were positioned under the nuchal trapezoid muscles for electromyogram (EMG) recording. Atipamezole (0.5 mg/kg, SC) was administered to reverse the medetomidine. Carprofen (5 mg/kg, SC) was administered pre-operatively, post-surgery and on the morning of the first post-operative day. Cefovecin (8 mg/kg, SC) was administered post-surgery.
Housing environment
Following a three-week recovery period, animals were housed individually in custom designed cages. Cranial implants were connected to ultra-low-torque slip-ring commutators (Hypnion, Inc., Lexington, MA, USA) by a flexible tether to allow unrestrained movement. A 12-hour light/dark cycle was maintained using white LEDs (35–40 lux) during the light period and red (> 680 nm) LEDs during the dark period, which enables infra-red recording. Food (2916 diet or TD.120782 diet containing 200 ppm doxycycline, Envigo, UK) and water were available ad libitum, ambient temperature was 23 ± 1 °C and relative humidity averaged 50%. A digital video camera allowed remote visual monitoring.
Behavioural Testing
Open-field locomotor activity A 40 × 40 × 30 cm Perspex arena was used to assess spontaneous locomotor activity during the light period but under complete darkness over 60-minute trials. The mice were untethered and movement within the arena was monitored using overhead infrared cameras (Sanyo VCV-3412P, Tracksys Ltd., UK). Cameras fed into a Quad compressor unit (Sanyo VDM-801P, Tracksys Ltd., UK) which relayed data to a computer running the image analysis software (Ethovision XT v8.5, Noldus, Netherlands) to calculate distance moved.
T-maze rewarded alternation Discrete-trial rewarded alternation was tested using a semi-automated T-maze (Apogee Engineering Analysis Solutions, Norwich, UK) as described previously (19). This apparatus was constructed of matte black, 8 cm wide Perspex with 20 cm high transparent Perspex walls. The external lengths of maze edges were 86 cm (choice end), 105 cm (return arm) and 22.5 cm (delay end). The centre arm was 83 cm in length and a door was located 63 cm from the choice point forming a holding area at the base of the start arm. The entry of an animal into specific areas of the maze was detected using infrared beam breaks and passed to a microcontroller (Arduino Mega 2560, RS Components, UK). Matlab programs automatically controlled the maze doors and test procedure, allowing it to run without human intervention. Rewards were delivered by three pellet dispensers, one located at the end of each reward arm and a third in the delay/holding area of the maze to encourage return of the animal to the starting point for the subsequent test phase or trial. Mice were trained following a two-stage protocol: forced alternation training and discrete-trial rewarded alternation testing. During the forced alternation training stage, mice were released from the holding area at the base of the T-maze, allowed to run along the centre-arm and forced to turn toward one of the reward areas to receive a sucrose pellet reward. Mice then returned to the holding area to collect a second reward pellet, after which another forced trial was initiated. At this stage, a 0 s inter-trial interval was used, and mice were trained for a maximum of 60 trials or 30 minutes daily. Once most mice were performing 40 training trials or more in a session, mice were moved on to the discrete-trial rewarded alternation protocol. Each trial consisted of two phases—a sample phase and a test phase. During the sample phase, mice were forced to turn toward the left or right arm and return to the starting/holding area. Two reward pellets were collected along the way, one at the end of the choice arm and one in the holding area. During the test phase of each trial, mice could choose between the two arms of the T-maze but were rewarded when visiting the novel arm only (i.e. arm not explored during the sample period). Forced left or right allocations during the sample phase were pseudo-randomized with no more than three consecutive sample runs to the same side. Mice could run for a maximum of 20 trials or 30 minutes daily over 3 days of testing. A 2 s inter-trial interval and a 5 s sample-to-test delay were used. The percentage of correct choices (primary outcome measure: number of correct choice/number of trials) as well as the choice latency (secondary outcome measure) were recorded and calculated for each animal.
Histology
Animals were terminally anaesthetised with pentobarbital and cardiac perfused with saline. Brains were then removed and weighed. Brain samples were cut into two coronal blocks using an adult mouse coronal brain matrix (ASI Instruments Inc., Warren, MI, USA) and processed using the Tissue TEK VIP processor (GMI Inc., Ramsey, MN, USA) before being embedded in paraffin wax for coronal brain sectioning. Serial sections (6 µm) were taken using HM 200 and HM 355 rotary microtomes (Thermo Scientific Microm, Germany). Immunohistochemistry was performed using a primary antibody for tau phosphorylated at serine 409 (PG-5, 0.11 µg/ml from Peter Davies; Albert Einstein College of Medicine, Bronx, NY, USA) as previously described (21). Stained sections were digitised using the Scanscope XT slide scanner (Aperio, CA, USA) at 20 × magnification. Imagescope software (version 11.1.2.760; Aperio) was used to view the digitized tissue sections and delineate the regions of interest (ROIs) which included the hippocampus and cortex. PG-5 positive tau pathology was quantified in these ROIs using a pathology scoring algorithm (1–5) by a pathologist blinded to treatment. Due to heavy staining of some sections, a different approach was used to quantify atrophy. The height of the hippocampus and cortex were digitally measured to create linear indication of atrophy, averaged across the left and right sides of the brain.
Data collection, sleep staging and statistical analysis
The SCORE2004™ bioassay facilitated the acquisition of multiple concomitant physiological measurements from 30 animals simultaneously. Validation of the SCORE2004™ technology has been previously described (28–30). EEG signals, recorded as the differential between front and back contralateral skull screws, were amplified 10,000x, bandpass filtered at 1-300Hz and digitized at 400 Hz [Grass Corp., Quincy, MA, USA]. EMG signals were amplified 20,000x, bandpass filtered at 10-100Hz, integrated based on the root mean square (RMS) and recorded as arbitrary units per 10 s epoch. Arousal states were then classified on-line as NREM sleep, REM sleep, wake, or theta-dominated wake in 10 second epochs using EEG period and amplitude feature extraction and ranked membership algorithms. Sleep bouts were determined as a minimum of 3 consecutive epochs of NREM, or 2 consecutive epochs of REM. Individually optimized EEG-arousal-state templates and EMG criteria differentiated states of arousal. Sleep data are presented as least squares means. Spectral frequency bands were delta (0.1 to 4 Hz), theta (5.1 to 9 Hz), alpha (9.1 to 12 Hz), beta (12 to 20 Hz) and gamma (30 to 80 Hz). Total power was selected as 0.1 to 30 Hz. Many of the analyses were separated into light or dark period to reflect significant periods of sleep and wake respectively. An average of one full week of data for each 12-hour light or dark period was used from each animal every 4 weeks.
Data quality control was assured by regular inspection of EEG blinded to treatment group using a proprietary suite of programs (SCOREVIEW™, Hypnion, Inc., Lexington, MA, USA). Data were excluded contemporaneously from further analysis based on signal quality. Sample size, after quality control, for each treatment group at each time point ranged between 12 and 23. EEG signal quality reduced over time in some rTG4510 mice, particularly after week 36, to the extent that scoring became unreliable. The EEG and EMG data for these affected animals were omitted from the analysis, thus reducing sample size in this treatment group at weeks 40 and 44 to n = 4–7.
For video locomotor activity (video LMA) the feed was converted to greyscale with 720 × 480 pixels per frame. The difference between each frame was calculated as the sum of the total number of pixels that changed relative to the previous frame. A minimum threshold of at least 10-pixel values changed per frame was necessary to limit the effects of video compression.
Statistical analysis of the longitudinal measures were conducted using a repeated measures analysis of variance and reported as least square (LS) means. For all responses the model used strain, animal age, and the interaction of strain by age to estimate the mean response difference between each of the 6 pair-wise strains at each of the 7 animal age time points. In addition, ofLMA included the location of the experiment in the model due to known effects. A post-hoc Tukey-Kramer multiple comparison adjustment was used to control for the Type I error rate based on the 42 pair-wise comparisons of interest. An unstructured covariance matrix was used to model the correlation of observations on the same animal. The bouts and spectral power outcomes were analysed on the LOG scale, as well as the T-maze measure average choice latency and the ofLMA total distance response. Estimates have been back transformed and differences will be interpreted as multiplicative.
Tau pathology and hippocampal thickness were analysed by ANOVA and Tukey post-hoc test. A correlation analysis compared hippocampus area with the last observation per animal during the dark period of the electrophysiological and behavioural measures. Actual values and Spearman residuals were analysed across all 4 treatments.