Background
The vast societal and economic burden of Alzheimer’s disease (AD) represents a growing problem [
1]. Despite extensive effort, AD drug discovery programmes have so far lacked late-stage clinical success [
2,
3]. While some of these failures have arisen from incomplete understanding of drug properties or inappropriate study design [
4,
5], many may also reflect true negative effects. Given the several hundred interventions reported to mitigate pathological and behavioural alterations in AD mouse models, it is reasonable to question the predictive validity of early AD drug discovery [
6]. Pre-clinical validation forms an integral part of nearly every drug project, yet justification of disease model as well as the endpoint(s) chosen often appears ad hoc, or simply lacking. In contrast, clinical efficacy of AD therapies is currently based on pre-defined cognitive test batteries and activities of daily living questionnaires. Therefore, a major consideration for future AD pre-clinical efforts would be to better understand the benefits and limitations of existing behavioural and functional assays, their relationship with disease progression (e.g. [
7]), and ultimately with clinical endpoints.
Contemporary research suggests tau rather than amyloid pathology is most closely associated with neurodegenerative processes [
8,
9] as well as cognitive and functional decline, prompting a shift towards tau-based drug discovery [
10]. Whilst current mouse models are unable to precisely recapitulate AD-like anatomical, temporal, and spatial progression of tau pathology, they still provide powerful tools for exploring disease-modifying therapies and are widely used in AD drug discovery [
11‐
13]. One of the most widely used models of tauopathy is the repressible rTg(tet-o-TauP301L)4510 (or rTg4510) mouse [
14,
15]. A significant advantage of this model is that expression of mutant tau is controllable by doxycycline administration, offering an ideal experimental positive control to determine functional effects of suppression of tau pathology. While the progression of tau pathology has been characterised thoroughly in this model [
14,
15], functional decline has often been described using cross-sectional comparisons of one or two endpoints; these vary from study to study and with very little justification of why these assays were selected.
The present work aims to extend the initial phenotypic characterisation of the rTg4510 mouse model [
15] and offer a detailed approach to the selection and validation of behavioural endpoints relevant for AD drug discovery. Male rTg4510 mice were first tested in seven different behavioural assays at an advanced pathological stage, between 12 and 15 months of age; the hypothesis was that rTg4510 mice would display robust and selective alterations in locomotor activity and spatial memory. The behavioural assays most sensitive to tau pathology were then selected and used in a separate and larger cohort of mice to assess longitudinally, from 4 to 12 months, the capacity of these endpoints to track progressive functional decline. The sensitivity of these functional endpoints to experimental modulation of tau pathology was tested in parallel via the administration of doxycycline from 4 months of age in a subgroup of rTg4510 mice. The hypothesis was that doxycycline would arrest the progression of tau burden and atrophy and thereby contribute to maintaining behavioural capacity at baseline levels, strengthening the hypothesis that such functional endpoints could be used to support future tau-based drug studies.
Discussion
The present study aimed to extend on the initial phenotypic characterisation of the rTg4510 mouse model and, in doing so, to define a common experimental approach to selection and validation of the most relevant cognitive and behavioural endpoints. Out of the seven behavioural assays tested at a late pathological stage, four were sensitive to tau pathology. Performance in these assays was shown to worsen progressively with age, to correlate well with developing tau pathology, and to be responsive to experimental modulation of tau expression by doxycycline treatment. The confluence of these three characteristics further illustrated that functional measures such as behavioural performance can be a meaningful proxy for underlying pathology, without the need for invasive measures.
The first study assessed 12-month-old male rTg4510 mice across a range of behavioural tasks to determine which would be sensitive to tau pathology. By this time point, bi-transgenic rTg4510 (CC) mice displayed considerable tau burden and brain atrophy, the extent of which was broadly comparable to that reported previously [
18]. CC mice exhibited striking levels of locomotor hyperactivity, profound stereotypy/circling behaviours, and yet intact balance and motor co-ordination. The propensity to become increasingly hyperactive and stereotyped with age is a common characteristic of tau transgenic models [
19‐
21] and needs to be considered when accounting for other behavioural impairments. For instance, male CC mouse performance was no better than chance in both T-maze rewarded alternation and Y-maze novelty preference tasks, though Y-maze continuous alternation performance was just as good as wild-type (WW) mice. Such apparent discrepancy between three tasks of short-term habituation and working memory are likely driven by stereotypy. Circling behaviour may indeed promote or prevent arm alternation respectively in continuous versus discrete-trial alternation paradigms. Several other groups have described behavioural deficits in rTg4510 mice, predominantly in the water maze, but rarely take into consideration how these effects might primarily be driven by stereotypy rather than cognitive dysfunction [
21]. Also of interest in this study was that male CC mice could acquire a non-spatial visual cue discrimination maze task as effectively as WW controls. This highlights the ability of the animals to overcome stereotypy in certain scenarios and suggests that some aspects of the rTg4510 CC mouse behavioural repertoire are still intact even at this advanced pathological stage. This stands in contrast to previous findings of impairments in visible platform water maze performance in rTg4510 CC mice [
16,
22], and suggests that not all tests of visual capacity may offer concordant outcomes. In conclusion, cognitive and behavioural deficits were observed in four out of the seven assays used. Open-field locomotor activity and T-maze alternation measured CC-related deficits of large effect size and lend themselves well to high-throughput longitudinal studies. Aversive Y-maze spatial reference memory and Y-maze spatial novelty preference tests were also capable of detecting CC deficits. However, these assays are of lower throughput and may evoke confounding practice effects after repeated testing; these assays are therefore more usefully employed in cross-sectional study designs.
In the second study, the four nominated behavioural assays from step 1 were used with the aim of tracking progressive tau-related functional decline from 4 to 12 months of age. Doxycycline was administered from 4.1 months, with the hypothesis that this would prevent further progression of tau pathology, and this in turn would preserve functional capacity. The regional distribution of tau pathology was consistent with that previously observed [
15,
23], reflecting the use of the calcium/calmodulin-dependent protein kinase II as a promoter for human tau transgene expression in these animals. Tau burden was apparent at 4 months and increased in an age-dependent manner. The progressive increase in tau pathology was associated with brain atrophy at 8 and 12 months, believed to be a result of synaptic degradation and neuronal loss [
15,
18,
24]. Early research suggested that whilst doxycycline suppression of transgene expression was able to prevent further neuronal loss, tau burden continued to accumulate [
18]. The present findings were, however, in line with recent literature [
24‐
27], and suggested that despite only moderate levels of tau transgene suppression (i.e. 50%) no further tau accumulation and atrophy were observed once doxycycline treatment was initiated. This level of transgene suppression is lower than previously reported by other groups [
14]. This is likely due to the food restriction protocol used in this study, where animals would have received less doxycycline than they would have if they were kept on an ad libitum diet and likely were subject to a lower exposure of the drug as a result. Nonetheless, the fact that only a partial suppression of tau expression by doxycycline was able to modulate subsequent progression of tau pathology and normalise behaviour provides encouraging support that potential clinical therapies do not necessarily have to be maximally effective to exhibit functionally beneficial effects.
Locomotor activity and rewarded T-maze alternation performance at 4 months worsened progressively until 12 months, at which point deficits were of similar magnitude to that observed in the first study, providing an internal replication of the dataset. Progressive decline of function was also observed in the aversive Y-maze spatial reference memory task, and was rescued by doxycycline as previously reported elsewhere in the water-maze task [
14,
15,
21,
28]. Doxycycline treatment effects were bi-directional, preventing both the increase in locomotor activity and the decrease in performance in all three other maze tasks. Such bi-directionality of effects lends weight to the interpretation that doxycycline was truly normalising behaviour. This is an important point, as drug studies often report a restorative effect in only one assay and one direction. Only by combining multiple functional endpoints from tests with different ancillary demands can confidence be raised on the meaningfulness of the observations. Other groups are beginning to take this into consideration when testing therapeutic interventions (e.g. [
29]). It is also important to highlight that this work only involved male rTg4510 mice. Sex differences have been reported by other groups [
16]; female rTg4510 mice have been shown to develop tau pathology at an earlier age and more aggressively than males. An important next step will be to validate the present findings in female mice.
Correlation analyses revealed that pathological and behavioural endpoints were closely related in CC mice. Locomotor activity and T-maze alternation performance displayed moderate to strong relationships with both measures of pathology, i.e. tau burden and atrophy. Such behavioural measures can therefore be taken as indicative of the level of underlying regional pathology and may offer a longitudinal, non-invasive, and high-throughput in-vivo biomarker of progression (and suppression) of tau pathology. Correlations may ultimately be investigated further with newer pre-clinical imaging techniques such as tau positron emission tomography (PET) tracers [
30,
31] and multi-parametric magnetic resonance imaging (MRI) [
24] that would allow longitudinal assessment of underlying pathology. Such imaging techniques also offer the ability to investigate key features of AD such as atrophy [
24], white matter changes [
32], and cerebrovascular reactivity [
33]. Pairing longitudinal imaging techniques with the behavioural work described here could potentially provide powerful insight into the relationship between functional decline and morphological alterations within diseased brains as the pathology develops.
Given the applied nature of the current work, it is important to comment on how these findings may fit into the bigger picture of AD drug discovery. All pre-clinical models have limitations that should be acknowledged and accounted for wherever possible. Several hundred interventions have been reported to normalise pathological and/or behavioural endpoints in AD mouse models [
2,
11‐
13], with some of these findings reported in rTg4510 mice [
26,
34‐
36]. It is highly unlikely that any one of the many existing mouse models of AD possesses complete predictive validity. Just as AD patients exhibit various degrees of regional pathological burden and diverse functional alterations [
37], AD mouse models also differ in these parameters and may display limited correspondence with human AD canonical Braak staging [
38]. Mouse models can also present with pathological and functional artefacts as a consequence of their construction. For instance, the tetracycline transactivator (TTA) used to control tau transgene expression in rTg4510 mice may be linked to developmental and functional abnormalities [
39‐
41]. Since no model is perfect, hypothesis testing in multiple models is always preferable as it offers increased confidence of drug effect via convergent validation [
42‐
44].