Temporal Cortical Thickness and Cognitive Associations among Typical and Atypical Phenotypes of Alzheimer’s Disease

INTRODUCTION

Structural magnetic resonance imaging (MRI) allows for the assessment of neurodegeneration (N) in Alzheimer’s disease (AD) via cortical thickness and volume, which along with pathologic amyloid-β (A) and tau (T) deposition are elements of the 2018 National Institute on Aging –Alzheimer’s Association Research Framework (AT/N Model) [1]. Typical amnestic AD (tAD) is associated with atrophy of the hippocampus [2] and cortical thinning across temporal, parietal, and frontal regions [3–5]. Temporal cortex is particularly affected in tAD and a temporal lobe meta-region of interest (ROI) has recently been used to capture the regions that have the most tau deposition in tAD [6–8]. The temporal meta-ROI includes the entorhinal cortex, amygdala, parahippocampal gyrus, fusiform, inferior temporal, and middle temporal regions [6]. Recent investigations using this temporal meta-ROI have demonstrated its value as a biomarker relating to disease progression and correlations with cognitive outcomes, particularly memory, in tAD [7, 8]. Imaging measures from the temporal lobe are also often utilized as outcome measures for clinical treatment trials in tAD [8].

The temporal lobe is also atrophic in patients with atypical clinical presentations of AD, including logopenic progressive aphasia (LPA), in which patients present with hesitant speech, anomia and poor sentence repetition [9], and posterior cortical atrophy (PCA), in which patients present with prominent visuospatial and perceptual problems [10, 11]. Cortical thinning and atrophy have been observed in the lateral temporal lobe in both LPA [12–16] and PCA [15, 18], although there appears to be regional variability in temporal lobe involvement across LPA, PCA, and tAD [15, 18–20]. Prior studies often focused on one atypical AD group compared to controls or tAD, and less attention has been paid specifically to evaluating temporal lobe regions in atypical AD compared to one another. A better understanding of how temporal thinning patterns differ across typical and atypical AD is important since the findings may help inform the appropriateness of temporal lobe focused imaging biomarkers for clinical treatment trials that include both typical and atypical AD.

There is evidence that different temporal lobe regions may relate to cognitive symptoms across the AD clinical spectrum, and that the defined primary cognitive domain according to phenotype may not always separate out according to syndromic expectation. Impaired episodic memory performance is the hallmark cognitive feature of tAD [21] and has been associated with atrophy of the hippocampus. By definition, memory cannot be the primary cognitive domain effected in atypical AD; however, memory deficits may co-occur [22–24]. Indeed, when typical and atypical AD are directly compared, the expected poorer episodic memory in tAD is not always found [23, 25–27]. Whether verbal or visual memory is assessed among AD variants may contribute to different conclusions [24], and therefore a thorough assessment of episodic memory in AD should include both modalities. Similarly, despite primary language deficit in LPA, object naming performance is at times comparable when directly compared to other AD variants [24, 26]. Spatial reasoning is one of the primary cognitive deficits in PCA and is more commonly impaired in PCA relative to other AD variants [19, 27], though comparable impairments have also been shown [26]. Studies have also found conflicting results regarding the neuroanatomic correlates of memory impairment in atypical AD. In LPA, one study found correlations in the medial temporal lobe [24], but another found correlations with the middle temporal gyrus [28]. In PCA, memory has been associated with hippocampal volumes, as is commonly found in tAD [19]. Object naming deficits are commonly found in the context of temporal lobe changes though the specific regions identified is mixed. For example, the left superior temporal lobe gyrus has been associated with naming deficits in LPA and tAD [12], as has the middle temporal lobe though not the superior temporal lobe in a separate study with similar disease groups [28]. In a recent study combining tAD and atypical AD, the left lateral and inferior temporal and medial temporal, but not entorhinal cortices were associated with naming deficits [29]. Notably, in the context of naming deficits, right temporal lobe atrophy particularly involving the inferior temporal gyrus as also been shown in tAD [12] and when tAD and atypical AD are combined [29]. The inferior temporal lobe has also been associated with spatial object perception in PCA [17]. These cognitive associations with temporal lobe regions occur within atypical AD phenotypes that tend to present with left greater than right temporoparietal atrophy in LPA [16] and bilateral posterior cortical atrophy, including posterior temporal more so in the right hemisphere in PCA [18]. Given the expected laterality differences in the atrophy patterns associated with cognition across the AD spectrum, evaluation of these cortical and cognitive relationships when comparing tAD to atypical AD should include both hemispheres.

Here, we included individuals on the AD continuum such that all participants were amyloid-β positive. We aimed to determine whether tAD and atypical AD 1) differed on structural measures, specifically hippocampal atrophy and cortical thickness of temporal regions, and 2) whether cognitive performances differed between and within groups and correlated with the temporal lobe measures. For the first aim, we hypothesized that tAD would show hippocampal atrophy compared to the atypical variants, that LPA would show more atrophy in regions within the left temporal lobe and that PCA would show more atrophy in regions within the right temporal lobe. For the second aim, we hypothesized that language would be associated with predominantly left lateral temporal cortical thickness, predominantly in the LPA group, that the visuospatial function would most strongly correlate with posterior temporal fusiform regions predominantly in the PCA group, and that memory performances would most strongly correlate with medial temporal regions, particularly in the tAD group.

METHODS

RESULTS

The groups did not differ on age at scan, age at onset, education, or PiB SUVR (ps > 0.05). LPA had a shorter illness duration (M = 3.4, SD = 1.4) than tAD (M = 4.6, SD = 2.6) and PCA (M = 5.0, SD = 2.4) (F (2, 146) = 5.76, p = 0.004, η² = 0.07) (Table 1).

Neuropsychological testing

MMSE scores did not differ among groups [F (2, 150) = 1.24, p = 0.29]. Although all AD groups scored below average on memory measures, there was a significant difference in performance for all memory measures (LM: χ² (2, n = 140) = 12.71, p = 0.002; VR: χ² (2, n = 113) = 54.85, p < 0.001; AVLT: χ² (2, n = 130) = 13.10, p = 0.001). Post-hoc testing revealed that tAD had significantly worse performance across all verbal and visual memory measures relative to LPA and PCA (ps < 0.032). The Boston Naming Test differed among AD type (χ² (2, n = 144) = 35.05, p < 0.001), with LPA performing significantly worse relative to tAD and PCA (ps < 0.001). Finally, visuospatial function as measured by the Rey-Osterrieth Complex Figure Test differed among AD type (χ² (2, n = 132) = 35.95, p < 0.001), with PCA performing significantly worse than tAD and LPA (ps < 0.001). Please see Table 1 and Fig. 1 for neuropsychological group differences.

OPEN IN VIEWER

Fig. 1

Cognitive Performances Among Groups. Note: Typical AD = Amnestic Alzheimer’s disease. LPA = Logopenic Primary Progressive. Aphasia. PCA = Posterior Cortical Atrophy. LM =Logical Memory. VR = Visual Reproduction. AVLT = Auditory Verbal Learning Test. BNT = Boston Naming Test. Copy Score is Rey-Osterrieth Complex Figure Test, Copy Trial.

Neuroimaging

To examine imaging ROIs by diagnostic group, we first ran a model predicting imaging ROI value by group. We then calculated pairwise comparisons to assess whether volume/thickness of each region, examining left and right hemispheres separately, differed between tAD and atypical AD (see Fig. 2). We assessed left and right hemispheres separately given these are not equivocally bilateral diseases. Hippocampal volumes and entorhinal cortex thickness were smaller in tAD compared to both atypical AD variants (ps < 0.001). Right hippocampal volume was smaller in PCA compared to LPA (p < 0.001). Compared to tAD, LPA showed greater thinning of the left middle (p = 0.025) and inferior temporal (p = 0.016) and left fusiform gyri (p = 0.010) but were relatively spared in the right middle temporal thickness (p = 0.041). PCA compared to the tAD group showed greater thinning of the right inferior temporal lobe (p = 0.016) and bilateral fusiform (ps≤0.001). When comparing the atypical variants to one another, PCA showed greater thinning of the right hippocampus (p < 0.001), middle (p = 0.011), and inferior temporal (p = 0.009) regions and right fusiform (p < 0.001) compared to LPA. In contrast, LPA showed greater thinning of the left middle temporal (p = 0.016) and left entorhinal cortex (p = 0.011) compared to PCA (Fig. 2).

OPEN IN VIEWER

Fig. 2

Linear Mixed-effects model results showing pairwise comparisons of temporal lobe ROIs between AD variants. Notes: Values represent estimated effect (circle) and 95% confidence interval (line) for each region and hemisphere, scaling volume to have effects in % change. tAD = Typical Alzheimer’s disease. LPA = Logopenic Primary Progressive Aphasia. PCA = Posterior Cortical Atrophy.

Associations between neuroimaging and neuropsychological performance

To assess the relationship between cognitive performance and neurodegeneration, we ran a mixed model with the cognitive scores predicting imaging values of the regions across the whole cohort regardless of diagnostic group (Fig. 3). This model showed a positive association with left and right hippocampal volume for all three memory tests (AVLT Delay, Logical Memory Delay, Visual Reproduction Delay ps≤0.002). Another way to view these data, for the AVLT Delay score for example, is to consider that as the memory score decreases by 1 SD (the standard deviation of this score across this data set), hippocampal volume is expected to decrease by a little more than 3% in the left and right hippocampus. Object naming performance showed a positive association with left middle and inferior temporal (ps < 0.001) and left fusiform (p < 0.001) and entorhinal thickness (p = 0.018), though somewhat surprisingly negatively correlated with hippocampal volume (ps≤0.003). Spatial reasoning performance was most strongly positively correlated with fusiform thickness, particularly on the right (p = 0.002, left fusiform: p = 0.032). Finally, we ran a mixed model with the cognitive scores predicting imaging values of the regions by diagnostic group (Fig. 4). This model resulted in most of the previous associations not being retained. In tAD, the left entorhinal cortex was associated with naming performance (p = 0.027) and the right entorhinal cortex was associated with Logical Memory Recall (p = 0.005). In LPA, the left fusiform was associated with naming performance (p = 0.016) and the right entorhinal cortex was associated with spatial reasoning (p = 0.038). No associations remained in PCA (see Supplementary Table 1).

OPEN IN VIEWER

Fig. 3

Linear Mixed-effects model with cognitive scores predicting temporal lobe ROI values across all AD patients. Notes: Values represent estimated effect (circle) and 95% confidence interval (line) for each test in each region and hemisphere, scaling volume to have effects in % change and centering and scaling the predictors to be on the SD scale. Typical AD = Typical Amnestic Alzheimer’s disease. LPA = Logopenic Primary Progressive Aphasia. PCA = Posterior Cortical Atrophy

OPEN IN VIEWER

Fig. 4

Linear Mixed-effects models with cognitive scores predicting temporal lobe ROI values within each AD variant. Notes: Values represent estimated effect (circle) and 95% confidence interval (line) for each test in each region and hemisphere, scaling volume to have effects in % change and centering and scaling the predictors to be on the SD scale. Typical AD = Typical Amnestic Alzheimer’s disease. LPA = Logopenic Primary Progressive Aphasia. PCA = Posterior Cortical Atrophy.

DISCUSSION

We aimed to explore the relationships between structural change in the temporal lobe and cognitive decline in typical and atypical AD. We hypothesized that tAD would show greater hippocampal atrophy compared to atypical AD, that there would be greater left temporal thinning in LPA, and greater right temporal thinning in PCA. We also hypothesized memory would be most impaired in tAD and most strongly correlated with medial temporal structures, that LPA would show greatest impairment on the language naming task and that the left lateral temporal regions would reveal the strongest correlation with naming performance, and that PCA would show greatest impairment on the visuospatial task and that poor performance would be associated with fusiform thinning. Overall, our hypotheses were partially supported.

Medial temporal regions, including bilateral hippocampal volumes and entorhinal cortices, were smaller in tAD compared to atypical AD variants. While this is consistent with our hypothesis given hippocampal atrophy as a biomarker of tAD [2], hippocampal atrophy has not always differentiated tAD from atypical AD in direct comparisons [24]. Interestingly, in our sample the right hippocampus was smaller in PCA compared to LPA. The medial temporal lobe findings here are consistent with prior work showing atrophy of hippocampal subfields tends to occur more bilaterally in typical AD and when present in PCA, it tends to occur worse on the right [45, 46]. PCA also showed greater right inferior temporal gyrus and bilateral posterior temporal (fusiform) thinning compared to tAD and additionally right middle temporal thinning compared to LPA. This bilateral posterior thinning with a greater right hemisphere bias of cortical degeneration is in keeping with the general pattern of posterior cortical volume loss in PCA [18] with additional support for a less common finding of specific involvement in posterior temporal fusiform to a degree that uniquely differentiates PCA from other AD variants. The LPA pattern of thinning was more left sided, with generally greater thinning of left lateral temporal regions, particularly with the left middle temporal gyrus differentiating LPA from other AD variants. This also extends some of our prior work showing the middle temporal gyri rate of atrophy separates nicely from healthy controls [47] by demonstrating lateralized thinning of the middle temporal gyri differentiating the atypical AD variants.

Consistent with clinical expectation, the tAD group had a prominent and greater deficit in episodic memory (verbal and visual memory), the LPA group had poorer language/naming and the PCA group had poorer visuospatial performance, respectively, compared to the other AD variants. The direct comparison of cognition in tAD and atypical AD is valuable as similar cognitive domains are often impaired relative to controls and the literature is somewhat mixed on whether the relative declines differ among AD variants [22, 48]. For example, Li et al. (2018) compared PCA to tAD on memory and spatial reasoning and found no verbal memory difference between groups though did find poorer spatial reasoning in PCA [27]. In another study, verbal memory was modestly stronger in LPA, but visual memory was poorer compared to tAD [24]. As we have discussed previously [22], deriving a neuropsychological assessment without the confound of language is very difficult if not impossible given the overlap of cognitive skills across domains. This is particularly salient while trying to assess a higher order cognitive skill such as memory in individuals with either a primary language or spatial deficit, as occurs with LPA and PCA respectively. As such, we chose to include common clinical measures to assess both verbal and visual memory to ensure that any separation of memory performance in tAD from atypical AD was not modality specific and arguably solely attributable to their primary cognitive deficit. With this approach, we demonstrated that regardless of modality episodic memory is most strikingly impaired in tAD relative to atypical AD.

The final aim of the current study was to explore the relationship between changes in cortical thickness and cognition among and within the AD variants. Bilateral medial temporal structural thinning (entorhinal cortex and hippocampal atrophy) was associated with poorer memory performance, as expected given the established association between hippocampal volume and memory [19, 49]. The association of episodic memory deficit and hippocampal atrophy is consistent with similar studies that compare one atypical variant to tAD [19, 24], but here we show that this core hippocampal finding persists when combining three AD variants. Also, consistent with Leyton et al. (2017), we found that performance on the naming task was associated with left inferior temporal and fusiform thinning. The superior temporal gyrus was also associated with anomia in the Leyton et al. study, and while that was not an area of interest in the current study, we did find additional correlations with the left middle temporal and entorhinal cortices. Win et al. (2017) also found this association with naming and left middle temporal gyrus in tAD and LPA, though they did not show a significant association with medial temporal structures [28]. Our data suggest that in addition to the commonly found lateral temporal associations with naming, the medial temporal structures also play a role, perhaps given the role of semantic categorization of objects in the hippocampal-entorhinal system [50]. Relatedly, Leyton et al. found single word comprehension was associated with bilateral fusiform atrophy, suggesting a role in ventral stream semantic processing [12]. We found that while the left fusiform was associated with naming performance, right fusiform thinning was not, fitting more with a predominantly asymmetrical language processing ventral stream. Instead, our bilateral fusiform regions were associated with visuospatial performance, which has also been suggested to play a role in spatial perception and integration of objects in addition to faces [51]. The finding of posterior temporal right fusiform thinning in PCA relative to other AD variants and its association with cognition is an interesting finding that warrants further investigation in tAD and atypical AD. The fusiform has also been shown to separate cognitively unimpaired amyloid positive individuals from cognitively unimpaired amyloid negative individuals in a longitudinal tau imaging paradigm [7]. This prior work suggests that the fusiform may play an important role early in the AD disease process, and our data suggest this important role may continue into the later symptomatic stages of AD across the spectrum. Longitudinal studies across the AD spectrum are needed to assess the potential value of the fusiform as an AD biomarker.

Isolating structural and cognitive associations to disease specific groups, yielded fewer relationships. Naming was associated with left fusiform in LPA, again supporting the possible critical role of the fusiform as a biomarker, and left entorhinal in typical AD. Additionally, the right entorhinal cortex was somewhat surprisingly associated with spatial reasoning performance in LPA but more understandably associated with memory recall on the Logical Memory Delay test. We anticipated that naming performance would most strongly associate with left lateral temporal regions particularly in LPA but this was not fully supported here. More associations between a temporal regions and cognition within AD phenotypes may become evident if examined longitudinally, as suggested by a recent study following PPA patients (non-semantic, and therefore may include LPA or agrammatic) which showed greater rates of atrophy particularly in the temporal lobe in those who were amyloid positive [13].

An alternative explanation for the limited associations that remained after separating by AD variant may be related to sample size and heterogeneity across groups. For example, by restricting the analysis to look at associations between cortical thickness and cognition within each AD variant individually, while the homogeneity of the groups increase, the ranges of both structural integrity as measured by neuroimaging and distribution of clinical test scores are reduced (i.e., the cognitive decline in disease states over time lead to reduced ranges in cognitive tests). For example, the PCA group did not show any significant associations between structural change and cognitive performance. This group also had the widest confidence intervals, partly because this is also the smallest sample in the current study. Alternatively, there may not be an effect here to find, particularly with PCA given that the majority of cortical atrophy in this group tends to be bilateral parietal and occipital, in addition to posterior temporal [18]. This second explanation alone seems less likely, particularly for LPA and tAD, given the prior work demonstrating associations between structural change and cognitive decline in AD variants (e.g., [12, 24]). Perhaps, too, the cross-sectional nature of this study limits our ability to find differences among a dynamic process. Multi-modal imaging studies exploring associations between cognition and additional biomarkers, including PiB PET and tau PET, may also help to elucidate patterns of neurodegeneration and cognitive decline [13, 52].

Taken together, these findings have important research, clinical and potential treatment implications. They provide direct support of critical temporal lobe differences between typical and atypical AD, as well as the presence of pathognomonic cognitive differences that drive the phenotype even when directly compared to other AD variants. These important distinctions may help determine which patient would benefit from further biomarker investigation (e.g., PET scanning, CSF analysis, etc.) as having a high confidence in disease etiology and ensuring the patient is on the AD continuum is important in guiding appropriate treatment recommendations and informing clinical trial enrollment [53]. Temporal ROIs, in particular, may have important utility and require a smaller sample size than other modalities for a clinical trial in cognitively unimpaired, yet amyloid positive, individuals, for example [7]. Our study lends further support for focusing on temporal regions and cognitive associations, with specific attention paid to the phenotype of AD being examined. It will also be important to consider cognitive outcomes as measured by neuropsychological tests associated with each phenotype.