Many proposed drugs for Alzheimer’s disease reverse memory impairments in mice but subsequently fail when tested in clinical trials. Could one problem be that mice and humans are tested on different forms of memory? Scientists in the lab of Steve Finkbeiner, University of California, San Francisco, are attempting to bridge the species gap with a new human equivalent of the Morris water maze. By collaborating with researchers who routinely use this test in mice, they developed a computer game that challenges people to find a buried treasure. Both tests assess spatial navigation, which is compromised early in Alzheimer’s disease. As reported January 19 in the Journal of Clinical Investigation, people with mild cognitive impairment (MCI) had similar problems learning and remembering the location of a target when compared to hAPP mice tested in the Morris water maze. The authors suggest that this could be a new way to assess human memory in clinical trials.

“For too long, cross-species translational measures have been absent in the Alzheimer's field,” wrote Jared Young, University of California San Diego, who co-authored an accompanying JCI commentary. “The simple methods suggested by Possin and colleagues might enable faster and more reliable screening for AD treatments,” he and colleagues wrote.

Buried Treasure Video Game: People navigate through a virtual field to hunt for buried treasure in this computer game that mimics the Morris water maze. [Possin et al., JCI.]

“There is a compelling rationale for developing a human version of the maze,” said John Growdon, Massachusetts General Hospital, Boston. “The one reported here looks particularly innovative.”

In the Morris water maze, mice swim through a round tank of opaque water and use visual cues to escape onto a hidden platform. Various labs have tried to imitate the Morris maze for people (for a review, see Boccia et al., 2014). A recent example is the real-life, dry-land version created by Jakub Hort of Charles University in Prague (see Jan 2013 conference news). That group has also developed a virtual version of the task (Nedelska et al., 2012). However, none of these tests have caught on for routine use in clinical trials. Scientists continue to use familiar pen-and-paper tests of cognition that measure cognitive skills such as verbal memory, naming objects, and copying figures. Some have developed tablet-based versions of some of those tests, but none tap spatial navigation. Prior adaptations of the Morris maze have differed from the mouse versions both in how they are administered and how the data are analyzed, said co-first author Kate Possin. Morris-like tests usually lack one or more of the components of the original maze, such as visible-target training, hidden-target learning, and/or probe trials. It is unclear which aspects of performance on the test are most sensitive to AD-related deficits in people.

Possin aimed to develop a people-friendly version of the Morris maze that was more directly comparable between mice and people. She created a computer game in which the player hunts for a treasure chest by navigating a car through a virtual field with a steering wheel and gas pedal. The player starts from a different location each time, and uses extra-maze cues to find the target, testing their allocentric navigation. Like visible-target training trials in the Morris water maze, the target is first clearly visible, appearing as a purple box on a plain, circular brown field. It becomes a treasure box when the car arrives. In hidden-target learning, the treasure chest becomes invisible, appearing only when the player lands on it. Similar to the mouse version of the test, the driver finds and remembers the location of the treasure based on its relation to surrounding houses, trees, and mountains. In “probe” trials, the target disappears completely, and the driver is given 90 seconds to drive over its spot as many times as possible.

Possin tested 89 healthy controls on the virtual maze, as well as 28 people who were diagnosed with MCI due to AD based on a neurological exam, neuropsychological assessment, and an informant interview. Collaborators at the Gladstone Institute for Neurological Disease in San Francisco, including co-first author Pascal Sanchez, also tested 106 mice, aged four to seven months, on the Morris water maze to directly compare human-mouse performance on the different aspects of the mazes. Half of those were wild-type, half were J20 mice that overexpress mutant human APP.

The researchers decided not to analyze their data with the repeated-measures ANOVA, since this statistical test does not account for non-linear learning curves, variability between trials, or trials that are stopped when the subject runs out of time. Instead, they used a rank-summary score based on how a subject performed relative to the other members. For instance, a mouse that came in third out of 20 on a given trial would get a score of 3/20. After averaging these scores, scientists can perform simpler statistics to assess differences between groups. Possin claims that this method is at least as sensitive as the repeated measures ANOVA at detecting differences between groups.

People mastered the visible-target learning whether or not they had MCI. Mice, on the other hand, had to overcome a tendency to swim around the edge of the pool, so they did worse on these initial trials. In contrast, mice performed more like people on hidden-target learning. Both J20 mice and MCI patients traveled farther than controls to find the target. In probe trials, J20 mice and people with MCI did not search as close to the target’s location as their respective controls.

The results suggest that distance traveled in hidden-target learning, and proximity to the target in probe trials, could be comparable measurements of learning and memory, respectively, between species, the authors wrote. Now that direct cognitive comparisons across species are available, the next step could be to assess spatial navigation in the computer task in longitudinal studies and as a secondary outcome in small clinical trials, said Possin. These will help work out how the variables change at different stages with disease progression. Some evidence implies that spatial navigation falters in neurologically healthy ApoE4 carriers, hinting that this variable could sensitively detect preclinical impairments (Berteau-Pavy et al., 2007). However, that is still speculative, Possin said.

“This is a very clever idea,” wrote Michael Weiner, UCSF. He proposed that this computer task be compared with other tests used to evaluate human subjects in clinical trials, saying, “It’s not clear if this new test offers improved sensitivity, specificity, or robustness for the evaluation of human subjects.”

Pierre Tariot, Banner Alzheimer’s Institute, Phoenix, agreed that the cross-species test was a good idea that deserves more study. "Finding translational cognitive testing paradigms has proven elusive,” he wrote to Alzforum. That some similar patterns exist between humans and transgenic animals is encouraging but not conclusive, Tariot added. He suggested testing whether there are converging patterns of responses to interventions.

Young suggested testing both mice and patients in the Morris maze after treatment with acetylcholinesterase inhibitors, one of the only treatments currently approved for patients with AD. This would show whether treatment effects in mice predict treatment effects in humans. Hort, of Charles University, previously reported that donepezil improved performance on his computerized version of the Morris maze (Hort et al., 2014). 

Hort has continued to use his virtual and real-life versions of the human Morris maze, and will soon present a head-to-head comparison of how mice and humans perform on the test after challenge with the anticholinergic drug scopolamine, which induces memory loss. Four external research teams have adopted his paradigm, Hort told Alzforum. He said that computer versions of the test are less sensitive than real-life versions, and noted that Possin’s does not distinguish allocentric from egocentric navigation, where the subject starts from the same location each time. Some data suggest that allocentric location, where a person depends on extra-maze cues to find a target, is most affected in AD and MCI (Serino et al., 2015).

Some researchers pointed out that major differences between species’ motivation to complete the task could still confound the comparability between measurements. Richard Keefe, Duke Institute for Brain Sciences, Durham, North Carolina, said that mice and humans process this maze information differently. After all, he said, “In this case, one species is swimming for its life and the other is playing a game.”—Gwyneth Dickey Zakaib


  1. I am convinced that spatial-memory paradigms have a good deal to offer in translating cognitive effects between preclinical and clinical studies. I think the sensitivity of this paradigm in patients with early stage Alzheimer's disease, as well as its responsiveness to anticholinesterase treatment, support its use in patient studies.

    Part of the intuitive appeal of this paradigm is that it better matches commonly employed preclinical paradigms, such as the Morris water maze. Our tests of episodic memory in patients are usually word-list or picture-learning tests; word-list learning tests are by definition language-mediated, and often visual stimuli are verbally labelled. Spatial-memory paradigms are perhaps less reliant on linguistic mediation, so the cross-species mechanisms of spatial memory tasks may be a better match. This might in turn better predict efficacy in humans of new drug entities.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Zuers—Can Spatial Navigation Guide Clinical Trials?

Research Models Citations

  1. J20 (PDGF-APPSw,Ind)

Therapeutics Citations

  1. Donepezil

Paper Citations

  1. . Neuropsychology of environmental navigation in humans: review and meta-analysis of FMRI studies in healthy participants. Neuropsychol Rev. 2014 Jun;24(2):236-51. Epub 2014 Feb 1 PubMed.
  2. . Spatial navigation impairment is proportional to right hippocampal volume. Proc Natl Acad Sci U S A. 2012 Feb 14;109(7):2590-4. PubMed.
  3. . Effects of sex and APOE epsilon4 on object recognition and spatial navigation in the elderly. Neuroscience. 2007 Jun 15;147(1):6-17. PubMed.
  4. . Effect of donepezil in Alzheimer disease can be measured by a computerized human analog of the Morris water maze. Neurodegener Dis. 2014;13(2-3):192-6. Epub 2013 Oct 30 PubMed.
  5. . Detecting early egocentric and allocentric impairments deficits in Alzheimer's disease: an experimental study with virtual reality. Front Aging Neurosci. 2015;7:88. Epub 2015 May 20 PubMed.

Further Reading


  1. . Spatial navigation measured by the Floor Maze Test in patients with subjective cognitive impairment, mild cognitive impairment, and mild Alzheimer's disease. Int Psychogeriatr. 2015 Aug;27(8):1401-9. Epub 2015 Feb 3 PubMed.
  2. . Impaired allocentric spatial memory underlying topographical disorientation. Rev Neurosci. 2006;17(1-2):239-51. PubMed.
  3. . Spatial cognition and the human navigation network in AD and MCI. Neurology. 2007 Sep 4;69(10):986-97. PubMed.
  4. . Neuropsychology of environmental navigation in humans: review and meta-analysis of FMRI studies in healthy participants. Neuropsychol Rev. 2014 Jun;24(2):236-51. Epub 2014 Feb 1 PubMed.
  5. . Neural correlates of spatial navigation changes in mild cognitive impairment and Alzheimer's disease. Front Behav Neurosci. 2014;8:89. Epub 2014 Mar 17 PubMed.

Primary Papers

  1. . Cross-species translation of the Morris maze for Alzheimer's disease. J Clin Invest. 2016 Feb;126(2):779-83. Epub 2016 Jan 19 PubMed.
  2. . Wet or dry: translatable "water mazes" for mice and humans. J Clin Invest. 2016 Feb;126(2):477-9. Epub 2016 Jan 19 PubMed.