Temporal tau asymmetry spectrum influences divergent behavior and language patterns in Alzheimer’s disease

Introduction

Beta-amyloid plaques and tau neurofibrillary tangles are the hallmark pathological findings in Alzheimer’s disease (AD).¹ Unlike amyloid, the spatial distribution of tau correlates with specific clinical symptoms, domain-specific cognitive impairments, and dementia severity.^2,3 While the burden and spatial distribution of tau have been extensively studied in relation to cognitive symptoms, the link between tau distribution and behavioral and emotional symptoms remains underexamined. Similarly, while the focus of disease modifying therapies is improving cognitive symptoms, behavioral and emotional endpoints are rarely considered.

Emotional and behavioral changes represent a major challenge to patients,⁴ caregivers,⁵ and the health care system.⁶ Insidious behavioral and emotional changes occur commonly in the mild cognitive impairment (MCI) and dementia stages of AD.^7,8 MCI is defined as impairment in one or more cognitive domains,⁹ overlooking the early emotional and behavioral changes commonly reported in patients with AD.^7,8 The stereotypical appearance of tau pathology in sporadic AD involves the transentorhinal region followed by further spread within the medial temporal lobe and subsequently into neocortical regions.^10,11 However, many neuropathological and neuroimaging studies highlight heterogeneity in the spatial pattern of neocortical tau tangle deposition and have identified asymmetric subtypes.^3,12–18 While asymmetric tau distribution is known in atypical AD presentations, temporal lobe tau asymmetry in typical amnestic AD has been understudied. Importantly, in addition to memory, the temporal lobes subserve aspects of language and behavior.^19–23

Although understudied in AD, socioemotional and behavioral symptoms have been deeply characterized in frontotemporal dementia (FTD), which presents with various combinations of behavioral, personality, mood, cognitive, and motor symptoms.^24,25 Notably, left- or right-predominant anterior temporal atrophy is associated with semantic variant primary progressive aphasia (svPPA) or semantic behavioral variant frontotemporal dementia (sbvFTD), respectively, and on the molecular levels both svPPA and sbvFTD show vulnerability to sporadic TAR DNA-binding protein-43 (TDP-43) type C pathology.²⁶ It is well established that left anterior temporal lobe involvement is associated with verbal semantic loss whereas right anterior temporal lobe involvement is associated with non-verbal and socioemotional deficits. ^10,11 We hypothesized that a similar pattern of graded right-socioemotional/behavioral to left-linguistic spectrum will be present in AD depending on asymmetric tau distribution in the temporal lobes. To address this hypothesis, we leveraged the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study to quantify asymmetric temporal tau burden using tau PET imaging and investigated the behavioral and cognitive patterns associated with right and left predominant tau burden.

Statistical analysis

Data analysis was performed with R version 4.0.2. Normality for continuous data was tested with the Shapiro–Wilk test. Means for continuous variables were compared with the Student’s t or the Mann-Whitney U tests. Baseline characteristics were compared using the χ2 or Fisher exact test for categorical variables. Homogeneity of variance was tested by Levene’s test. Given the majority of participants with clinical impairment were in the MCI stage and the number of dementia and MCI-other individuals was low (Table 1), we combined all impaired individuals into one cognitively impaired (CI) group. Statistical significance was determined at p-value<0.05.

View this table:

• View inline
• View popup
• Download powerpoint

Table 1.

Demographics, functional, diagnostic, APOE4 and Amyloid status information of the asymmetry-spectrum and asymmetry-extreme groups. HW: Hispanic White; NHW: Non-Hispanic White, NHA: Non-Hispanic Asian, NHB: Non-Hispanic Black, AIAN, UNK, +2 more than 2.

To understand the relationship between tau SUVR and LI, we conducted a univariate correlation between whole temporal SUVR and the absolute value of the LI of the whole temporal lobe (Figure 1). We also examined associations between LI with tau SUVR from each individual temporal ROI (Supplementary Figure 6), as well as the direct association between left SUVR and right SUVR within each individual temporal ROI (Supplementary Figure 7).

• Download figure
• Open in new tab

Figure 1: The relationship between whole temporal tau SUVR and absolute laterality index values.

Higher whole temporal SUVR is associated with higher temporal tau asymmetry. Asymmetry-extreme cases, colored in blue (left extreme) and red (right extreme), show elevated SUVR values in a range overlapping with the symmetric cases. Yellow line represents regression line for the entire sample.

To determine whether lateralized temporal tau burden was associated with cognitive and behavioral performance, we executed a series of linear regression models within the asymmetry-extreme groups. First, controlling for age, sex, and education, we examined the effect of extreme laterality group (left versus right) on cognitive composite scores and behavior (NPI total score). Next, we examined whether the association between cognition/behavior and elevated tau burden differed by laterality group by including an interaction term between whole temporal SUVR and laterality group into these models.

Leveraging the entire cohort, we examined associations between continuous levels of temporal tau asymmetry with cognitive and behavioral performance. In this analysis, the contribution of each lateralized ROI was examined while controlling for regional tau SUVR in the contralateral ROI. In these models, cognitive composite scores and behavior were the outcome variable, with left and right regional SUVRs are predictors (controlling for age, sex, and education).

Results

Among the entire sample of 858 subjects, 465 (54%) were on the left lateralized spectrum (LI>0) whereas 393 (46%) were on the right lateralized spectrum (LI<0) (Figure 1 and Table 1). There were no differences in age, education, sex, handedness, race/ethnicity, APOE4, CDR, functional diagnosis, or amyloid status between the two spectrum groups. In the extreme asymmetry groups, there were twenty-seven extreme left and twenty extreme right participants. There were no differences in age, education, sex, CDR-SOB, amyloid status or APOE4 between the left and right extreme groups. The prevalence of left handedness in the right-predominant group was comparable to the 10% prevalence reported in the general population,³³ whereas the prevalence was relatively higher in the left-predominant group (18% of the left-predominant group were left-handed). Higher whole temporal SUVR was associated with higher temporal tau asymmetry (Figure 1, Supplementary Figures 6 and 7), confirming that asymmetric cases had elevated SUVR values. Further, the range of elevated whole brain and whole temporal values overlapped with the symmetric cases. There were no autosomal dominant mutations careers in the entire cohort.

Extreme tau asymmetry group confers distinct cognitive and behavioral patterns

The extreme asymmetry groups did not exhibit differences in terms of behavior, memory, executive function, or language after adjusting for age, sex, and education (Box plots in Figure 2A and detailed in Table 2 model 1). Upon the incorporation of whole temporal tau into the model, higher tau was associated with poorer memory and executive performance; however, no associations were observed with behavior or language (Table 2 model 2).

• Download figure
• Open in new tab

Figure 2: Asymmetry-extreme group.

A-Boxplots: The asymmetry-extreme groups did not exhibit statistically significant differences on terms of behavior, memory, executive function, or language, although a trend is present for higher NPI in the right-extreme and lower language in the left-extreme. A-Scatterplots: There is a significant interaction between whole temporal tau by the right vs left extreme groups; higher right, but not left, tau is associated with worse behavior whereas higher left tau, not right, was associated with worse language symptoms. These distinctions were not present when predicting memory or executive functioning as higher right and left temporal tau were associated with worse memory and executive function. B. The first raw represents conjunction maps of the right-extreme group and left-extreme group. Figures were created using tau SUVR threshold of 0.25 (i.e., masks for tau SUVR values over 0.25 were created, binarized, and summed to create conjunction maps). The remaining rows are individual examples from each of the four patterns found upon visual inspection of the asymmetry-extreme groups. The four patterns of asymmetrical tau distribution along the anterior-posterior dimension are shown in B-B. anterior temporal (total of 6 right and 8 left), B-C. posterior (occipital) group (total of 3 right and 1 left) B-D. typical pattern (total of 6 right and 14 left), and B-E. typical pattern with frontal involvement (total of 3 right and 4 left).

View this table:

• View inline
• View popup
• Download powerpoint

Table.2: Series of linear regression models in the asymmetry-extreme groups (models 1-3) and in the asymmetry-spectrum (model 4).

Models 1: The asymmetry-extreme groups did not exhibit differences in terms of behavior, memory, executive function, or language after adjusting for age, sex, and education. Model 2: Upon the incorporation of whole temporal tau into the model, higher tau was associated with poorer memory and executive performance; however, no associations were observed with behavior or language. Model 3: Upon including the interaction of whole temporal tau by the right vs left extreme groups, higher right, but not left, tau was associated with worse behavior and higher left tau, not right, was associated with worse language symptoms. Model 4 shows the contribution of one ROI after correcting for demographics and contralateral tau. Model 4: In the entire cohort, when considering the contralateral tau effect, encompassing both right and left ROIs in the same models and correcting for age, sex, education, chiefly right ROIs displayed significant associations with behavior, while left ROIs were associated with language. Interestingly, more lateral temporal regions were significant (inferior temporal, middle temporal, superior temporal, fusiform gyrus) whereas not all the mesial temporal regions were significant (i.e., medial temporal lobe was trending for behavior on the right and not significant for language on the left). Although the whole temporal, medial temporal lobe, amygdala, and anterior temporal ROIs showed association between higher tau on the left, but not the right, and worse memory and executive function, many temporal ROIs (fusiform gyrus, inferior temporal, middle temporal, and entorhinal cortex displayed correlations with worse both memory and executive functions. NPI: neuropsychiatric inventory. Bold indicates p value less than 0.05. Italics indicate p value less than 0.

An interaction term between laterality group and whole temporal tau SUVR showed distinct patterns with behavior and cognition within each group. Specifically, the right temporal extreme asymmetry group displayed increased behavioral symptoms as tau burden increased, while language symptoms remained relatively stable with increasing tau levels. On the other hand, the left temporal extreme asymmetry group exhibited lower language scores as tau burden increased, while behavioral symptoms remained relatively steady with rising tau levels (Scatter plots in Figure 2A and detailed in Table 2 model 3). These distinctions were not present when predicting memory or executive functioning.

Asymmetry spectrum of tau confers distinct cognitive and behavioral patterns

In the entire cohort, when controlling for contralateral tau, right regional temporal tau levels displayed significant associations with behavior, while left regional temporal tau levels were associated with language (Table 2, Figure 3, supplementary Figures 1-4). Interestingly, these effects were more consistent in lateral temporal regions (inferior temporal, middle temporal, superior temporal, fusiform gyrus) as opposed to medial temporal regions. The relationship between tau laterality was less clear for memory and executive function. Although the whole temporal, medial temporal lobe, amygdala, and anterior temporal ROIs showed associations between higher left asymmetric tau with worse memory and executive function, many other additional ROIs on both the left and right (fusiform gyrus, inferior temporal, middle temporal, and entorhinal cortex) displayed correlations with worse memory and executive function.

• Download figure
• Open in new tab

Figure 3. Association of right and left temporal tau with each of behavior as measures by the neuropsychiatric inventory (NPI) and cognition as measured by language, memory, and executive function composites.

Figure shows residualized contributions of the left and right temporal ROI after correcting for the contralateral temporal ROI as well as age, sex, and education. Upper left: Higher right-temporal but not left-temporal associated with worse behavioral symptoms. Upper right: Higher left-temporal but not right-temporal associated with worse language performance. Lower left: Higher left-temporal but not right-temporal associated with worse memory performance. Upper right: Higher left-temporal but not right-temporal associated with worse executive function performance. Similar plots are shown for each temporal ROI in Supplementary Materials Figure 1-4.

Discussion

This study characterized the presence of asymmetric patterns of tau accumulation in a mildly impaired AD cohort with a primarily amnestic presentation. Importantly, the degree of asymmetry within the temporal lobe corresponded to distinct behavioral and language deficits. Predominant right temporal tau accumulation was associated preferentially with behavioral symptoms, whereas predominant left temporal tau accumulation was linked preferentially to language symptoms. Furthermore, patterns of asymmetry were not consistent across individuals, suggesting multiple subtypes exist spanning the anterior-posterior axis. Interestingly, there were less clearly discernible differences between right and left predominant temporal tau groups in memory and executive function. These observations hold implications for clinical diagnosis, disease progression tracking, and therapeutic approaches. Our findings align with the established typical amnestic AD, limbic predominant, and hippocampal sparing schema^12,13 and extend this classification by noting additional heterogeneity regarding laterality, as well as additional subtypes such as an anterior temporal variant that aligns anatomically with frontotemporal dementia (FTD) subtypes. Altogether, this study highlights the need to integrate and develop additional socioemotional and neuropsychological measures that may be most impacted by tau laterality in the temporal lobes.

The early involvement of the temporal lobes in AD is well-documented. Aside from their role in episodic memory and spatial navigation, the temporal lobes function as central hubs for both verbal and non-verbal semantics. Verbal semantics encompass multi-modal knowledge of words and objects and are preferentially left temporal lateralized. Whereas non-verbal semantics involve multi-modal knowledge for socio-emotionally relevant concepts.^21,34,35 Extensive FTD, linguistic, and socioemotional literature show that semantic concepts are represented on a graded spectrum where both right and left temporal lobes seem to be required for semantic knowledge with the right being more specialized for non-verbal socioemotional concepts and the left for verbal semantics.^23,36,37 Over the past couple of decades, research has elucidated the networks responsible for emotion processing. Processing the emotions of others involves several steps including understanding the meaning of an external expression, internally experiencing the expressed emotion, attributing the behavior to the other rather than the self, and inhibiting one’s own perspective.^38,39 These processes localize to different but interconnected neuroanatomical circuits communicating between the insula, temporo-parietal junction, and the semantic appraisal network connecting the anterior temporal lobe and the orbitofrontal cortex. We presume that all these regions can also be afflicted by AD pathology. Previous studies have indicated that AD patients perform poorly on emotion perception tasks.⁴⁰ In the current study, the left-predominant group exhibited more pronounced language deficits whereas the right-predominant group showed associations with behavioral scores. This pattern of verbal/left and non-verbal right lateralization aligns with the literature in FTD, which indicates a spectrum of semantic loss on a spectrum spanning non-verbal socioemotional concepts in the right and verbal semantic on the left temporal lobes. Collectively, we propose that the interaction between tau and the language network in the left, and with the non-verbal socioemotional network in the right, could potentially be caused by the loss of verbal and non-verbal semantics, respectively.²¹

Behavioral symptoms in AD in relation to frontal cortex are relevant to a previously described variant labeled executive, or behavioral variant AD.^41–47 Comparisons between typical amnestic AD and behavioral variant AD point to divergent patterns of cortical atrophy,^41,42 glucose metabolism,⁴³ and tau accumulation.⁴⁸ However, these studies have limitations such as not including measures of behavior or considering both hemispheres simultaneously. Furthermore, pathological studies often fail to determine whether the right or left hemispheres are simultaneously or asymmetrically involved, mainly because they typically focus on examining one hemisphere. Additionally, in-vivo tau PET imaging studies have widely utilized a meta-temporal ROI to estimate tau burden,^44–47 which combines regions in the temporal lobes and collapsed across hemisphere. In our study, we demonstrate that even in a typical amnestic cohort of AD, there is asymmetric tau accumulation associated with distinct phenotypic patterns along the behavioral/language continuum. Consistent with the right temporal role in emotions in AD, research investigating cortical volumes showed that emotional contagion is linked to atrophy in the right lateral temporal regions.²² Moreover, reduced emotion prosody recognition has been associated with cortical atrophy in the right temporal pole and superior temporal sulcus in AD.⁴⁹ Furthermore, right medial temporal lobe involvement has been linked to paranoid delusions in mild AD.⁵⁰ In our study, the differences observed in the NPI were primarily influenced by increased anxiety, irritability/liability, and hallucinations. Intriguingly, a functional MRI investigation revealed a connection between heightened salience network connectivity in the anterior cingulate cortex and right insula areas, and specific NPI subcomponents including agitation, irritability, aberrant motor behavior, euphoria, and disinhibition.⁵¹

When examined individually, the right- and left-extreme participants exhibited variable degrees of frontal, parietal, and occipital involvement. When defining these extreme asymmetry groups, we did not restrict selection to cases that only had asymmetry in the right or left temporal lobe and it is possible that a given subject’s maximum tau accumulation was outside the temporal lobes. Further, the observed patterns of variability along the anterior-posterior axis adds to the complexity of phenotypic and pathological heterogeneity. Interestingly, the left-predominant group involved more of the contralateral hemisphere compared to the right-predominant group consistent with research showing cortical volume loss occurring earlier and progressing faster in the left hemisphere.^52,53 However, this may be a confound related to study design. Specifically, participants are recruited based on language centered tests and must meet ADNI mild cognitive impairment or dementia criteria for enrollment whereas advanced right predominant cases may be excluded due to more pronounced behavioral symptoms that are presumed to be due to non-AD etiologies. Additionally, the neuropsychological tools that evaluate memory and executive function are language-dependent biasing sample selection even further to left lateralized disease and highlighting the need for non-verbal neuropsychological batteries to understand the full phenotypic heterogeneity associated with AD pathology.

Mechanisms underlying asymmetry and heterogeneity in the spatial distribution of tau in AD remain unknown. Asymmetric patterns in AD are not specific to PET and have been found in cortical volumes,⁴⁹ white matter topology,⁵⁴ and structural connectivity networks.⁵⁰ Furthermore, early pathological investigations reported asymmetric hemispheric, hippocampal plaques and tangles, and lateralized differences within Braak and Braak stages in at least 15% of cases.^56–58 Gaining a better understanding of asymmetry in AD could offer insights into the underlying causes of selective vulnerability.

This study has limitations. It was performed in the ADNI cohort which primarily consists of amnestic AD. Consequently, the study’s findings likely underestimate the prevalence or degree of right-predominant patterns in the broader population. It is important to note that large cohorts with a broad range of clinical symptoms that have undergone tau PET are currently lacking. A notable limitation is the absence of in depth objective measures for assessing behavior, emotion, and language in ADNI. Our behavioral analyses rely on the NPI questionnaire, which likely shows limited sensitivity to detecting right-sided deficits in behavior and emotion.

In conclusion, this study highlights the importance of asymmetry in patterns of tau burden in AD patients. It is imperative to develop better behavioral, socioemotional, and language tests to understand this phenotypic heterogeneity. Therapeutic and diagnostic approaches in precision medicine would benefit from characterizing the asymmetry burden in the spatial distribution of tau in patients with AD, as this asymmetry has implications for clinical impairment, caregiver burden, and disease trajectories.

Data availability statement

Data for this study is publicly available on the ADNI LONI website https://adni.loni.usc.edu/

Conflict of interest statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

This study was supported by the National Institutes of Health ADSP-PHC U24 AG074855.