Navigating the future of clinical assessments

This scientific commentary refers to ‘Differentiation of mild cognitive impairment using an entorhinal cortex-based test of virtual reality navigation’ by Howett et al. (doi:10.1093/brain/awz116).

Located in the medial temporal lobe, the entorhinal cortex (EC) is particularly vulnerable to accumulation of tau neurofibrillary tangles early in Alzheimer’s disease. The EC contains a number of neuronal types vital for spatial navigation, in particular grid cells (Hafting et al., 2005), which are crucial for path integration (Fig. 1A). Behavioural measures of path integration may therefore provide an early cognitive marker of Alzheimer’s disease. In this issue of Brain, Howett et al. (2019) demonstrate the power of harnessing new virtual reality technology to create tests with greater sensitivity for detecting individuals at increased risk of Alzheimer’s disease.

Figure 1

Summary of study design, major results and implications. (A) Areas implicated in path integration include posterior-medial entorhinal cortex (pmEC), hippocampus (Hipp), medial prefrontal cortex (mPFC), posterior cingulate cortex (PCC), retrosplenial cortex (RSC) and precuneus (Prec). The temporal lobe is highly vulnerable to tau neurofibrillary tangles in ageing and Alzheimer’s disease. During the transition from ‘healthy ageing’ to Alzheimer’s disease, tau tangles spread to frontal and posterior-medial cortex, where amyloid-β also accumulates. (B) Overview and major results of the study by Howett et al. The virtual path integration (PI) task was performed within a virtual open arena with boundary cues projected to infinity. Participants walked an L-shaped path to three locations (cones 1–3), before being asked to walk back to their start location (return path to cone 1). The path integration task showed the highest accuracy in discriminating biomarker-positive and -negative patients with MCI. (C) There were three different return conditions: (i) the same virtual environment as during the outbound path; (ii) the same virtual environment without distal boundary cues; or (iii) the same virtual environment without surface details. Grid cells in A are adapted with permission from Hafting et al. (2005). Right brain in A is adapted with permission from Jagust (2018) by permission of Springer Nature.

Open in new tabDownload slide

In their study, a group of healthy elderly participants and a group of patients with mild cognitive impairment (MCI) performed an immersive triangle completion task (Fig. 1B). Participants wore a wireless head-mounted display and walked two legs of an outbound path marked by cones, rendered in one of three different virtual environments, before being asked to complete the triangle by returning to their start location. The return leg was performed under one of three different conditions (Fig. 1C). Performance on this task, assessed via Euclidean distance error (‘PI error’ in Fig. 1B), is assumed to rely on the ability to path integrate and thus on entorhinal grid cell integrity. In addition to path integration performance, the authors assessed standard neuropsychological measures, MRI-based atrophy measures as well as CSF-based measures of amyloid-β and tau pathology in a subset of the patients with MCI. Relative to healthy controls, patients with MCI performed worse on the path integration task, as well as on most standard neuropsychological tests. Notably, however, accuracy on the navigation task almost perfectly separated the patients with MCI according to their biomarker status, such that patients with abnormally high levels of amyloid-β/tau pathology (biomarker-positive) made significantly larger Euclidean errors than patients with normal biomarker levels (biomarker-negative). Furthermore, the navigation task outperformed standard neuropsychological tests in determining biomarker status.

Consistent with the findings of Howett et al., path integration abilities in older adults have been shown to correlate with functional measures of putative grid cell firing in the EC (Stangl et al., 2018). Although grid cells are found primarily in the rodent medial EC, which is thought to correspond to the human posterior-medial EC (pmEC) (Maass et al., 2015; Schröder et al., 2015), other brain regions have also been implicated in path integration (Fig. 1A). For example, visual path integration, in which the participant has access only to perceptual information to update their spatial location, recruits a diverse range of brain regions including the hippocampus and medial prefrontal cortex (Wolbers et al., 2007). Moreover, the retrosplenial cortex and posterior medial cortex are also thought to play an important role in path integration (Etienne and Jeffery, 2004). To examine the neural correlates of successful path integration, Howett et al. segmented structural MRI scans to assess the relationship between the integrity of different brain regions (including entorhinal subregions) and Euclidean distance error. In line with the hypothesis that grid cell dysfunction in EC explains path integration impairments, across the whole sample of controls and patients, pmEC (i.e. EC volumes on the three most posterior slices) and whole EC volumes showed the strongest relationship with distance errors. Subsequent analysis in Howett et al., however, showed that the volume of a number of different brain regions was associated with path integration performance. It is possible, therefore, that the deficits observed in the biomarker-positive group reflect pathology not limited to the EC.

With respect to markers of Alzheimer’s disease pathology, higher levels of both total tau and amyloid-β pathology (i.e. lower CSF amyloid-β_1‐42) independently predicted distance errors among the patients with MCI. Fluid biomarkers, however, cannot tell us where in the brain the amyloid-β and tau pathology are located. Neuropathological and in vivo PET studies have demonstrated that tau and amyloid-β show different spatial and temporal patterns of accumulation in Alzheimer’s disease. Whereas amyloid-β initially aggregates in prefrontal and posterior-medial cortex, tau aggregates first in the anterior temporal lobe, particularly in lateral EC regions bordering perirhinal cortex (Braak and Braak, 1997). Tau pathology in the anterior-temporal lobe is common in old age, whereas the progression of tau tangle pathology to posterior-midline regions is seen in the presence of amyloid-β plaques and is related to cognitive symptoms. Thus, it is likely that the biomarker-positive MCI group has tau pathology that is already more widespread (Fig. 1A). As noted above, given that the precuneus, retrosplenial cortex and prefrontal cortex are also involved in path integration, the performance deficits in the biomarker-positive subjects could reflect tau or amyloid-β burden throughout the cortex. Further PET studies will be necessary to determine more precisely the relationship between the location of Alzheimer’s disease pathology and path integration deficits.

Although the number of trials was relatively small, the data by Howett et al. suggest that the availability of different cues in the return condition modulates performance of the biomarker-positive patient group. Specifically, when path integration performance was broken down into its constituent rotational and linear displacement components, the biomarker-positive group demonstrated greater rotational errors when either the distal landmarks or local surface details were removed (Fig. 1C). One interpretation is that the biomarker-positive group cannot reliably use body-based information to estimate rotation in the environment and therefore they attend more to visual cues to determine the angle of turns. This deficit could be mediated by impairments in neural populations in retrosplenial cortex or precuneus that support representations of the direction in which one is facing (Shine et al., 2016). Alternatively, this impairment may reflect deficits in the biomarker-positive group in identifying and attending to other distal cues that are consistent across all return conditions. For example, the sun in the virtual environment (Fig. 1B) provides a stable cue that, even when the more prominent distal landmark cues are removed, could provide information as to rotations in the environment. It may be the case that, unlike the biomarker-positive group, the healthy control and biomarker-negative groups can use these constant perceptual features of the world to support rotation estimates. In future studies it will be important to separate out the relative contributions of different cues, including through the use of a purely body-based condition, which has been shown to be sensitive to grid cell function (Stangl et al., 2018).

Although Howett et al. provide one of the first examples of immersive virtual reality as a clinical assessment tool sensitive to Alzheimer’s disease biomarker status, there was also a considerable amount of variability in the performance of the healthy control and biomarker-negative groups. Given that tau accumulates in EC in the course of normal ageing, and that grid cell-like representations are predictive of age-related path integration deficits (Stangl et al., 2018), the variability in control group performance could reflect age-related entorhinal tau pathology. Because there was no biomarker information available for the healthy controls, it is unclear whether any of these participants showed elevated amyloid-β/tau burden in the absence of cognitive symptoms. Future studies combining path integration tasks and molecular imaging will be necessary to understand the molecular underpinnings of path integration deficits and impaired grid cell function in cognitively unimpaired elderly. Given the importance of administering drug or lifestyle interventions as early as possible in the course of neurodegenerative disease, it is crucial to determine whether path integration paradigms can identify individuals at risk of Alzheimer’s disease in advance of cognitive symptoms. In this respect, it will be necessary to compare the predictive value of spatial navigation tasks with object memory tests, since the earliest accumulation of tau neurofibrillary tangles is seen in the anterolateral EC, a region associated with object processing (Berron et al., 2018). It is likely that a combination of different behavioural paradigms will be required to establish sensitive and disease-stage specific cognitive biomarkers of Alzheimer’s disease.

Given the additional time and resources required to implement virtual reality methods in clinics, studies such as that by Howett et al. will prove invaluable in demonstrating to clinicians the added sensitivity of such tasks in comparison to the gold-standard neuropsychological battery. Moreover, it will be important to investigate whether desktop PC/tablet versions of this task offer the same predictive power. This may well be the case given that the putative grid cell signal can be observed during navigation in the absence of body-based cues (Stangl et al., 2018). This could provide an even easier way of bringing to the patient’s bedside the diagnostic properties of spatial navigation tasks.

Competing interests

The authors report no competing interests.

References