Alzheimer’s disease (AD) is a chronic neurodegenerative disease characterized by cognitive impairment, progressive memory loss, dementia, and behavioral disturbances (Jack et al., 2018). On a molecular level, extracellular deposits of amyloid beta and the resulting neuritic plaques along with the intracellular accumulation of hyperphosphorylated tau protein into paired helical filaments (PHFs) and neurofibrillary tangles (NFTs) contribute to neuronal loss and the pathological hallmarks seen in AD brains (De Strooper and Karran, 2016). The severity of tau pathology, in particular, has been implicated as a major predictive factor for the extent of cognitive impairment, making it a promising therapeutic target for AD (Arriagada et al., 1992; Bejanin et al., 2017; Harrison et al., 2019; La Joie et al., 2020).

As one of the most abundant microtubule-associated proteins in the central nervous system, tau has been recognized to play major roles in promoting microtubule assembly and stabilization and in maintaining the normal morphology of neurons. Recent studies suggest that tau, upon alternative mRNA splicing, may also participate in the regulation of intracellular signaling pathways, development, and the viability of neurons (Avila et al., 2004; Wang and Mandelkow, 2016). In the human CNS, tau protein is translated from a 6-kb mRNA transcript generating a series of six tau protein isoforms of 37–46 kDa which result from alternative splicing of exons 2, 3, and 10 (Guo et al., 2017). The complexity of the tau isoforms is further increased by various posttranslational modifications (Wang and Liu, 2008). Thus, dysregulation of tau proteostasis can arise from a number of sources such as: splice site and/or missense mutations, changes in overall expression or isoform composition (Trabzuni et al., 2012; Strang et al., 2019), various post-translational modifications (Mair et al., 2016; Arakhamia et al., 2020), and epigenetic regulators (Lardenoije et al., 2015; Millan, 2017). These changes can lead to the abnormal aggregation of tau, resulting in the eventual seeding and spread of pathological tau species that contribute to a number of neurodegenerative diseases, collectively known as tauopathies, which includes AD (Gotz et al., 2019).

Therefore, elucidating the specific role of tau protein is imperative to unraveling the underlying mechanisms of these neurodegenerative diseases. To address this issue, the laboratory of Peter Davies created the htau mouse model (Andorfer et al., 2003). The htau mouse produces all six isoforms of human tau with no evidence of native, murine tau expression. Furthermore, htau mice develop pathologies that effectively model spatiotemporal and histopathological features seen in AD, such as age-dependent hyperphosphorylation of tau in the hippocampus and frontal cortex, somatic redistribution of tau to dendritic compartments, tau aggregation into NFT-like structures, increased inflammation, decreased cortical thickness, and neuronal cell death (Andorfer et al., 2003). However, although the htau mouse model elegantly recapitulates many of the histopathological features of AD, behavioral analyses of this model have yielded inconsistent results in the past (Table 1). Taking Morris Water Maze, one of the most commonly used tests in studies of AD models, as an example, half of the published studies reported cognitive deficits in htau mice, whereas the other half reported normal learning/memory (Supplementary Table 1). This lack of consistency led us to conduct the current study and verify the presence/robustness of cognitive phenotypes in this mouse model before using it to screen for novel treatment strategies. While mouse models have provided new insights into disease mechanisms, efforts to convert this knowledge into gains in the clinic have struggled. One of the main sources of concern is the animal models used in preclinical drug development. Thus, we would like to emphasize here that the value of our study goes beyond serving as a pivotal point of phenotypic characterization by providing the research community with key insights into critical factors that might substantially affect replicability of “disease phenotypes” and subsequent preclinical research on treatment development. Also, we attempted to understand our results in the context of existing literature, and to provide some insights on reproducibility of behavioral data.

No sex differences were found in any test employed in the current study. See Supplementary Material for statistical results of sex comparisons.

Cognitive phenotypes in htau mice have not been consistently reported in the past decade (Table 1). Overall, the existing literature suggests that the cognitive phenotype in aged htau mice, with standard behavioral testing methods, is not robust. Consequently, in the present study we set out to evaluate learning and memory in our own colony of aged htau mice. We tested 16–20 months old htau mice and age-matched controls in a battery of learning and memory tests commonly used to assess AD-like phenotypes in mouse models. While it is interesting that we found an impaired spatial recognition memory in aged htau mice tested in the Y maze, our study aslo revealed normal Morris Water Maze and fear conditioning performances, adding to the body of existing evidence that failed to show robust cognitive impairments in htau mice (Table 1). It is notable that in Morris Water Maze, our control mice failed to display a clear preference for the trained quadrant in probe trial 1 (Figure 2). This could be partially attributable to stress and fatigue from the training trials earlier in the day. Although both groups exhibited preference for the trained quadrant in probe trial 2, it is possible that the probe trial performed 2 h after the last training trial could have diminished the strength of preference exhibited in the probe trial performed 24 h later. This is due to the extinction effect, in the absence of a “refresher trial” between the two probe tests.

Given that the water maze is usually considered one of the most sensitive tests for detecting functional defects of the hippocampus (Vorhees and Williams, 2006; Kraeuter et al., 2019), we compared important parameters employed in each experiment, in an effort to comprehend why only about half of previous studies reported water maze deficits in htau mice. We noticed that the ratio of pool-to-platform size is 160 in studies that reported deficits, but only 110 in studies that reported no deficits. Interestingly, age or sex does not seem to be factors that affect result outcome (Supplementary Table 1). We argue that the larger the ratio of pool-to-platform size, the more difficult the task is. As such, the discrepancies in previous findings may be partially explained by task difficulty (which is related to pool-to-platform ratio).

Y maze task difficulty is higher than the water maze, due to its lack of strong incentive stimuli and repetitions. In the Y-Maze novel arm preference test, there is only one non-rewarded “learning trial” before the memory test is performed. In contrast, Morris Water Maze utilizes a strong incentive (escaping the water) and allows the animal multiple days of learning (with four trials/day). These important differences between Y-maze and Morris Water Maze may be helpful to understand why we discovered a deficit in the Y-maze test but not in the Morris Water Maze test. The fear conditioning test also utilizes a salient aversive stimulus—three 2-s foot shocks—during the conditioning phase. Therefore, behavioral tests that incorporate strong aversive stimuli might not be particularly sensitive to subtle differences in learning and memory performance (Redish and Touretzky, 1998; Broadbent et al., 2006).

Given the literature and our own data, we reason that cognitive impairments in htau mice may be too subtle for standard testing procedures to detect. Instead of arguing which tests may be more or less sensitive, we suggest that incorporating a number of different memory tests is a good strategy to ensure that multiple facets of the cognitive phenotype are comprehensively evaluated. In addition, resource permitting, it could be useful to incorporate low-stress operant tests such as the Bussey–Saksida touchscreen system (Horner et al., 2013; Mar et al., 2013; Oomen et al., 2013) which has high technical and conceptual similarities with the CANTAB Alzheimer’s test battery used to assess cognitive function in human AD patients (Swainson et al., 2001; Egerhazi et al., 2007).

Lastly, we want to emphasize the translational value of our novel findings. In this regard, using this humanized mouse model of late onset Alzheimer’s tau pathology, we found a spatial recognition memory impairment that might be relevant to some of the key cognitive deficits observed in patients with AD and related dementias. Spatial navigation, in particular, is emerging as a potential cost-effective biomarker to detect cognitive decline in incipient AD and related dementias in the preclinical stages, which has important implications for future diagnostics and treatment approaches (Coughlan et al., 2018).

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Institutional Animal Care and Use Committee of Columbia University.

IS-M and MY supervised the project. JC, YK, and ER performed research and conducted behavioral experiments. JC, MY, and IS-M analyzed and interpreted data and wrote the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.