Neuropsychological Markers of Medial Perirhinal and Entorhinal Cortex Functioning are Impaired Twelve Years Preceding Diagnosis of Alzheimer’s Dementia

INTRODUCTION

The search for cognitive markers of preclinical Alzheimer’s dementia (AD) is informed by the pattern of progressive cortical neurofibrillary tangle (NFT) pathology in Alzheimer’s disease, as the largely intracellular, abnormally phosphorylated tau proteins are deemed responsible for functional deficits [1–3]. Episodic memory is commonly impaired first in the preclinical course of typical AD, predating the diagnosis by up to 10 years (see [4] and Supplementary Table 1). These findings are consistent with NFT-induced disruption of neuronal functioning in the entorhinal cortex (ERC) and hippocampus (HP) in the earliest stages of the disease. However, cortical NFT pathology begins in the medial perirhinal cortex (mPRC) before spreading to the ERC and HP [5, 6], suggesting that cognitive impairments associated with mPRC dysfunction may precede or accompany episodic memory impairments in the preclinical course of AD [6, 7].

Perirhinal (PRC) functions have traditionally been studied in non-human primates [8], and only more recently in humans (e.g., [9, 10]). Human neurocognitive research suggests that the PRC is required for processing semantic memories of individual objects, especially those that are easily confused with other similar objects [10, 11]. For example, patients with brain damage including the PRC are significantly impaired at processing semantically confusable versus semantically less confusable objects during picture naming and audiovisual crossmodal feature integration [12, 13] and in early AD patients, the disambiguation of semantically confusable objects specifically depended on mPRC (but not ERC) integrity [14]. Moreover, a delayed matching-to-sample task [15], which depends on the PRC [14, 15], has a high predictive value for progression from MCI to AD [7].

Longitudinal data are required to determine the relative temporal appearance of PRC and ERC/HP (i.e., episodic memory impairments) dysfunction in the preclinical course of AD. However, many well-established clinical neuropsychological tests used in current longitudinal studies are more meta-cognitive in nature, lacking the anatomical specificity to target the PRC. For example, Hirni et al. [16] demonstrated that both clinical tests of episodic memory and semantic object memory significantly predicted ERC/HP and PRC atrophy. However, linear regression analyses differentiated the relationships between PRC integrity and semantic memory functioning, and ERC/HP integrity and episodic memory functioning: when the opposing memory performance was controlled, episodic memory performance was associated with ERC/HP but not PRC integrity, whereas semantic memory was associated with both PRC and ERC/HP integrity [16]. These results indicate that when existing clinical neuropsychological data are used to assess PRC and ERC/HP functions, one way to increase anatomical specificity for medial temporal lobe substructures would be to control for the opposing memory performance.

This study aimed to determine the optimal combination of common clinical neuropsychological test variables to specifically predict mPRC and ERC functioning (prediction study). In a critical analysis, these formulae were applied to the neuropsychological test performance of longitudinal groups of initially optimally healthy individuals, some of whom progressed to AD, to determine whether and when mPRC and ERC functioning decline in the prodromal phase of AD (longitudinal study). If also mPRC functions decline in the prodromal phases of AD, corresponding neuropsychological measures may prove powerful diagnostic markers of very early AD.

METHODS

RESULTS

DISCUSSION

Common clinical neuropsychological tests significantly predicted the mean cortical thicknesses of the mPRC and ERC: both semantic and episodic memory measures were significant predictors of mPRC thickness, while only episodic memory significantly predicted ERC thickness. These findings were specific to mPRC and ERC structural integrity, since semantic and episodic test performance failed to significantly predict left temporal, parietal, or frontal lobe thickness. While these findings require confirmation in a larger longitudinal cohort, ideally with CSF and/or PET measures of pathological amyloid-β and tau load, the present approach, and formulae, may be applied in clinical settings to more specifically assess the functional integrity of these circumscribed anterior medial temporal lobe (aMTL) regions.

These formulae were used here to estimate longitudinal mPRC and ERC functioning in two independent groups of originally optimally healthy older individuals: those that went on to develop AD (NC-AD) and pairwise matched individuals who remained cognitively healthy (NC-NC). We note that stringent inclusion and exclusion criteria ensured that all participants were optimally healthy at baseline. The slight decrease in performance of the NC-NC participants over time most likely reflects the known slight decline in cognitive test performance (such as, for example, in episodic, source, semantic, and procedural memory probably due to reduced processing speed and increased liability to interference) associated with advancing age [33]. Both the estimated mPRC and ERC functioning of the NC-AD group was significantly inferior to that of the pairwise-matched NC-NC group at the earliest available testing, 12 years preceding diagnosis, and further declined circa 4 years preceding diagnosis, corresponding roughly to the timepoint when MCI is typically diagnosed [34, 35]. In contrast, the widely-used MMSE, a measure of global cognitive functioning, did not distinguish the NC-NC and NC-AD groups until immediately prior to the AD diagnosis. These findings are consistent with the earliest cortical deposition of AD neurofibrillary pathology in mPRC and ERC [5, 6] and suggest that not only episodic, but also semantic memory, may be early and specific cognitive markers of future AD.

Additionally, the present results are in line with a longitudinal MRI study demonstrating significant atrophy in medial temporal, inferolateral temporal, and parietal lobes and posterior cingulate in families with early-onset AD circa five years before the diagnosis of AD [36], as well as studies with sporadic AD patients which demonstrated atrophy of medial temporal lobe structures four to ten years before the AD diagnosis [37–41].

The majority of large-scale longitudinal studies of cognitive functioning in preclinical AD report that episodic memory is the first function to decline, up to ten years preceding the diagnosis of AD, while semantic memory becomes impaired many years later (see [2] and Supplementary Table 1). However, Amieva et al. [32] reported a notable exception to this rule: in the PAQUID study, semantic memory as measured by semantic fluency was significantly impaired at the earliest testing 12 years preceding the diagnosis of AD in their group of 350 NC-AD patients. Episodic memory became impaired three years later, i.e., 9 years preceding the diagnosis of AD. Similar findings were reported by Wilson et al. [42], who showed that an average of three semantic memory tests (i.e., verbal fluency, Boston Naming Test, and a measure of reading recognition) was affected earlier than episodic memory performance (76 versus 63 months, respectively) in a group of 462 NC-AD patients. Since the mPRC presumably codes for semantic object memories [14, 16], the findings of Amieva et al. [32] and Wilson et al. [42] are in line with the pattern of cortical NFT pathology in AD, which affects the mPRC before the ERC in the preclinical course of AD [5].

A commonly cited biomarker model of preclinical AD [43] states that cognitive dysfunction occurs in the last preclinical stages of the disease, i.e., at diagnosis. Specifically, NFT pathology occurs prior to cortical atrophy as measured by MRI, followed by clinical symptoms. Thus, at a given point in time, MR atrophy will be more abnormal than cognitive functioning (cf. [33], Figure 2A). However, longitudinal studies demonstrate that significant aMTL atrophy is apparent only circa 3.5 years prior to the AD diagnosis [44–46], i.e., after semantic and episodic memory impairments as shown here. Also, NFT pathogenesis supports the initial affectation of function followed by structural atrophy, since the earliest stages of tau pathology are associated with neuronal dysfunction but not yet significant atrophy [47]. For example, experimentally induced abnormal tau hyperphosphorylation in rats resulted in significant learning and memory impairments just months later [48], i.e., before the tissue loss could occur [47]. Thus, NFT-induced functional mPRC impairments appear early in the preclinical course of AD, offering clinicians an inexpensive, early and specific additional AD biomarker. The challenge to future studies will be to develop more specific cognitive measures of mPRC functioning than the current clinical neuropsychological tools [14, 23] by taking advantage of the wealth of primarily non-human primate research on these structures [8].