Animals
Generation of homozygous rTg(tauP301L)4510 transgenic mice, henceforth referred to as rTg4510 mice, has been described previously [43]. The rTg4510 and litter matched wildtype mice were bred on a mixed FVB/NCrl +129S6/SvEvTa background for Eli Lilly by Taconic (Germantown, Maryland, USA), licensed from the Mayo Clinic (Jacksonville, Florida, USA), and imported into the United Kingdom for study at the UCL Centre for Advanced Biomedical Imaging. Mice were kept in individually ventilated cages in groups of 3–5, with ad libitum access to food and water. Animal weight was monitored upon arrival and until experimentation, and all animals were deemed healthy prior to experiments.
For this study, import of 10 female 7.5 month of age rTg4510 and 10 female litter matched wildtype mice occurred two weeks prior to initial imaging studies and subsequent sacrifice, perfuse fixation, and sectioning for immunofluorescent analysis. All the experiments were performed in accordance with the Animals (Scientific Procedures) Act 1986 (ASPA) revised according to the European Directive 2010/63/EU and the UK Home Office (Scientific Procedures) Act (1986) with prior project approval from UCL’s internal Animal Welfare and Ethical Review Body.
Magnetic resonance imaging of mice
Image acquisition
All imaging was performed with a 9.4 T VNMRS horizontal bore scanner (Agilent Inc). A 72 mm inner diameter volume coil (Rapid Biomedical) was used for radiofrequency transmission, and signal was received using a 4-channel array head coil (Rapid Biomedical). Mice were placed in an induction box before anaesthesia was induced using 2% isoflurane at 1 L/min in 100% O2. Mice were subsequently positioned in an MRI-compatible head holder to minimize motion artefacts. Anaesthesia was maintained throughout imaging using 1.5% (± 0.2%) isoflurane at 1 L/min in 100% O2 delivered via a nose cone, which permitted spontaneous breathing of the mice. Core temperature and respiratory rate were monitored using a rectal probe and pressure pad, respectively (SA Instruments). Mice were maintained at ∼37 °C using heated water tubing and a warm air blower with a feedback system (SA Instruments). Respiration rate was maintained between 80 and 120 breaths per minute by manually adjusting the isoflurane vaporizer.
A 3-dimensional T2-weighted fast spin echo sequence was employed for structural imaging with the following parameters: field of view (FOV) = 19.2 × 16.8 × 12.0 mm, resolution = 150 × 150 × 150 μm, repetition time (TR) = 2500 ms, effective echo time = 43 ms, echo train length = 4, number of signal averages = 1, and imaging time = 1.5 h.
Mice were imaged in a random order, working through cages which housed a mixture of rTg4510 and litter matched wildtype mice.
Image analysis
For manual segmentation of the optic nerve and eye structures for extraction of volumes and signal intensities on T2-weighted images, ITK-SNAP segmentation software (v3.x) [49] was used. For extraction of brain volumes, a multi-atlas-based structural parcellation framework was used [25]. Using this framework, we extracted 4 structures of interest to the visual system and dementia: the cortex (containing the visual cortex), thalamus (containing the lateral geniculate nucleus), hippocampus, and superior colliculus. The lateral geniculate nucleus and superior colliculus were of interest as in both humans and mice they are key targets of optic tract axons, and contribute to visual processing, as well as acting as a relay to other destinations in the visual pathway. In addition, we extracted the volume of the whole brain for normalisation purposes. For this, brain images were oriented, non-uniformity corrected, and skull stripped. We adopted the publicly available in vivo mouse brain MRI atlas previously published for the framework [26]. First, the atlas images were registered affinely to the original MR image data using a block-matching algorithm [36]. Once complete, the STAPLE algorithm [25] was applied to fuse the resampled atlas masks together to create a consensus brain mask for each animal’s scans. A further non-rigid registration based on fast free-form deformation was then performed to correct any remaining local misalignment of the affinely registered atlas to the brain volumes [30]. The structural labels from the atlas were then transformed and resampled to the image space of the brain scans using the same affine and non-rigid transformation and fused using the STEPS label fusion algorithm [22] to create the final parcellated structures of interest: the cortex, thalamus, hippocampus, and superior colliculus. A previously published calibration protocol was used to correct gradient scaling errors in the data [32]. Both absolute volumes and volumes normalised to the extracted whole brain volume are presented.
CSF extraction, perfuse fixation and tissue preparation for histology
Following in vivo MR imaging, animals were terminally anaesthetised with an overdose of Euthanal administered via intraperitoneal injection. A midline incision was made at a midpoint between the skull base and the occipital margin to the first vertebrae. The underlying muscles were parted to expose the atlanto-occipital membrane and dura mater overlaying the cisterna magna. The area was cleaned and a durotomy performed with a 23-gauge needle, allowing CSF to be collected using a narrow bore pipette tip. The thoracic cavities were then opened and the animals were intracardially perfused through the left ventricle: first with 15–20 mL of saline (0.9%) and heparin; second with 50 mL of Formalin, at a flow rate of 3 mL per minute. Following perfusion, the animal was decapitated, de-fleshed, and the head stored at 4 °C and soaked in Formalin for 9 weeks. Brains were removed from the skull and processed using the Tissue TEK VIP processor (GMI Inc.) and embedded in paraffin wax. 6 μm thick sections of the brain in the sagittal plane were collected using a rotary microtome and mounted on glass slides for immunohistochemistry. The eye globes were enucleated, processed (Leica ASP3005) and embedded in paraffin in an orientation that facilitated sectioning in transverse planes. Ice cooled paraffin tissue blocks were sectioned (Leica RM2235 Microtome using S35-PFM feather microtome blades) to generate 4 μm sections which were then placed in a 45 °C water bath. Once mounted onto Super-Frost Plus slides, sections were drip-dried for 2–5 min before heat fixing for 60 mins. Haematoxylin and Eosin (H&E) staining (Autostainer Leica CV5030) and light microscopy (Olympus multi-head light microscope with Micropixx camera software) enabled semi-quantitative morphometric analysis (Image J) and identification of each sections anatomical location in the retinal peripheral-central axis. Sections within 25 μm of the optic nerve head were subsequently used for immunohistochemical staining. The central and peripheral retina were defined as the most lateral and central non-fragmented region of each tissue section (Additional file 1: Figure S1). H&E derived semi-quantitative data included Retinal Ganglion Cell Layer (RGCL) nuclear densities and Inner Nuclear Layer (INL) nuclear densities (number of nuclei occupying a measured area of RGCL or INL respectively), and Inner Plexiform Layer (IPL) thickness relative to total retinal thickness. This data was generated for peripheral and central retinal regions.
Quantification of CSF tau
The collected CSF was centrifuged briefly to collect any red blood cell contaminates, the supernatant removed and frozen at − 20 °C until further analysis by ELISA. 2 μl water was added to the blood pellet and snap frozen on dry ice to ensure hypotonic freeze-thaw release of haemoglobin from any red blood cells present. Each sample was measured at 417 nm on a NanoDrop spectrophotometer to quantify percentage blood contamination of each CSF sample. This method allows for measurement of blood contamination down to 0.0001%, well below that detectable by eye (~ 0.01%). Lack of significant blood contamination was verified by an average percentage contamination of 0.003% (±0.0006%). Tau content of CSF samples were quantified using both the Human Tau (total), and Human pTau [pS199] ELISA Kits (Invitrogen, UK) as per the manufacturer’s instructions.
Immunohistochemistry
For immunohistochemistry in the retina, retinal sections from 5 rTg4510 and 4 wildtype mice were deparaffinised and hydrated by sequential immersion in xylene, 3 successive ethanol solutions (100, 90, and 70%) and distilled water. As tau proteins largely have intracellular distributions, permeabilisation of nuclear and plasma membranes was carried out using detergents: Tween-20 (0.05%: Sigma-Aldrich P1379) in citrate buffer (Citric acid, anhydrous, Sigma-Aldrich C1857-100G), and Triton-X-100 (0.1%: Sigma-Aldrich T8787-50ML) in phosphate buffered saline (PBS: Gibco 20,012–019) forming PBST. Antigen retrieval was achieved through heating sections in a microwave (setting to 800 W) in citrate buffer (pH 6.0) for 5 mins before slides were left to gradually cool to room temperature (RT) for 20 mins. Following two 5 min PBS washes, sections were blocked with goat serum (Sigma-Aldrich, G9023-10 M: 1:20 in PBST) for 1 h at RT. After performing two 3 min PBST washes sections were incubated with primary anti-pTau (Thermo Scientific MN1020, mouse monoclonal AT8: 1:100 dilution in PBST) and anti-Tau (Dako, rabbit polyclonal A0024: 1:2000 dilution in PBST) antibodies for 1 h at RT. A0024 binds to total tau independent of phosphorylation (amino acids 243–441) whereas AT8 binds to aberrantly phosphorylated tau species (pS202/pT205). Negative controls, where no primary antibody incubation was performed, were conducted alongside each experimental run (Additional file 1: Figure S2). Following three washes with PBST (5 mins each) sections were incubated with fluorophore conjugated complementary secondary antibodies at RT for 1 h (Alexa-Fluor 568 goat anti-mouse A21124 and Alexa-Fluor 647 goat anti-rabbit A21244, Life Technologies: 1:200 dilution in PBST). This, and all subsequent stages were conducted in the dark to prevent photobleaching. Sections underwent three further 5 min PBST washes, and were then incubated with Hoechst stain (16.2 mM stock solution diluted to 1:5000 in PBS) for 15 mins at RT to provide blue nuclear staining. Two 3 min PBS washes followed. Slides were coverslipped (Mansel-Gläser 24 × 60 mm) following the application of Vectashield anti-fade mounting medium (Vector Laboratories, H-1000) and stored at 4 °C until imaged. Fluorescent confocal microscopy (Zeiss LSM510 confocal microscope with Zen 2 software) generated qualitative and semi-quantitative data. Constant standardised settings were used for comparative non-adjusted imaging, and optimised settings used for adjusted imaging. This semi-quantitative data included the number of pTau positive cells in the RGCL and INL, and the mean intensity per pixel of areas of individual retinal strata (RGCL, IPL, INL and photoreceptor inner segment (PIS)). Data was generated by using Zen & Image J softwares to analyse images of central and peripheral retinal regions. This data, like the semi-quantitative data acquired from H&E images, allowed comparison between rTg4510 and wildtype retinas, and between the central and peripheral retina.
For immunohistochemistry in the brain, brain sections were deparaffinised and hydrated by sequential immersion in xylene, 3 successive ethanol solutions (100, 90, and 70%) and distilled water. Antigen retrieval was performed using the Lab Vision PT module system (Thermo Scientific), where sections were heated to 100 °C for 20 min in citrate buffer (TA-250-PM1X; Thermo Scientific). Slides were transferred to a Lab Vision Autostainer (Thermo Scientific) where the following incubations were performed: 10 min in H2O2 (0.3%); 30 min in normal goat serum (1:20; Vector Laboratories); 60 min in primary antibody for aberrantly phosphorylated (pS409) tau (PG-5; 1:8000 from Peter Davies, Albert Einstein College of Medicine, NY, USA); 30 min in biotinylated goat anti-mouse IgG (1:200, BA9200; Vector Laboratories); 30 min avidin-biotin complex solution (PK-7100; Vector Laboratories); 5 min in 3,3′-diaminobenzidine (SK-4105; Vector Laboratories). Apart from the last two steps, PBST was used for diluting reagents and washes between steps. Sections were then counterstained with haematoxylin before dehydration and cover-slipping. To quantify PG-5-positive tau pathology, stained sections were digitised using the Scanscope AT slide scanner (Aperio) at 20× magnification. Image J software was used to view the digitised tissue sections, threshold for immunoreactivity and delineate boundaries of the visual cortex, hippocampus, thalamus, superior colliculus and optic tract. Data is expressed as the percentage coverage of immunoreactivity in each region.
Magnetic resonance imaging of humans
Participants
Five male subjects fulfilling the criteria for the diagnosis of behavioural variant FTD [40] and carriers of a mutation in the MAPT gene (three 10 + 16 and two R406W mutations) were consecutively recruited from a tertiary referral cognitive disorders clinic at the National Hospital for Neurology and Neurosurgery, London, UK. Five healthy controls were also consecutively recruited (1 female, age range: mean(standard deviation) 62(11) years). Patients’ age: 64(3) years; disease duration: 9(3) years; age at onset: 55(1) years. A full clinical history including ophthalmic history was taken from all patients as part of the standard assessment. Hence the patients (and controls) included in this study were free of any retinal or eye disease which may have confounded the data presented.
Image acquisition
Volumetric T1- and T2-weighted MRI was performed in all subjects. MRI scans were acquired on a 3 T scanner (Tim Trio, Siemens) with the following sequences: (i) high-resolution isotropic 3D T1-weighted MPRAGE (sagittal orientation; TR = 2200 ms, TI = 900 ms, TE = 2.9 ms, flip angle = 10°, acquisition matrix = 256 × 256 and spatial resolution = 1.1 mm); and (ii) high-resolution isotropic 3D T2-weighted fast spin echo/SPACE (sagittal orientation; TR = 3200 ms, apparent TE = 105 ms, variable refocusing pulse flip angle to achieve T2-weighting, acquisition matrix = 256 × 256 and spatial resolution = 1.1 mm).
Image analysis
Acquired T1-weighted images were initially transformed into standard space by a rigid registration to the Montreal Neurological Institute (MNI305) template. Acquired T2 images were registered to the MNI305 template as previously described [6]. Segmentations of the optic nerve were performed manually on about 12 consecutive coronal slices on co-registered T1- and T2- MRIs using NiftyMIDAS (Centre for Medical Image Computing, UCL: http://cmic.cs.ucl.ac.uk/home/software/), starting from the most rostral slice where the eyeball was no longer visible. The intrarater intraclass correlation (ICC) for the optic nerve segmentation, computed with a two-way random effects model, was 0.85 (95% confidence intervals: 0.50–0.96).
Statistical analysis
Statistical comparisons between animal groups were performed via either a two-way analysis of variance (ANOVA) followed by post-hoc Bonferoni post-tests for multiple comparisons, or (un)paired t-tests for single comparisons, using GraphPad Prism (v5 for Windows, San Diego, CA, USA. Statistical comparisons between patient cohorts were performed via an ANOVA followed by post-hoc Bonferoni post-tests for multiple comparisons, using GraphPad Prism (v5 for Windows, San Diego, CA, USA).). All individual data points are shown with mean ± SEM for animals/participants in each group.