Individuals with FTD and aMCI were identified in a retrospective review of medical records at the Cleveland Clinic Lou Ruvo Center for Brain Health in Cleveland. These records included approximately 3,000 cases with cognitive complaints seen per year at our clinic, over each of the past 10 years.

The FTD samples (n = 7 svPPA and n = 8 rtvFTD) represented consecutive cases seen in our clinic who met study inclusion and diagnostic criteria. In order to compare the effects of FTD to AD pathology, an AD sample was also recruited. In an effort to roughly equate the clinical samples for disease severity, the inclusion criteria of 1–2 years of symptoms was imposed for AD as well, resulting in a sample in the amnestic mild cognitive impairment stage of AD in the aMCI stage of AD (n = 8). The aMCI sample was randomly selected to be matched to the FTD samples, representing the most recent cases who had similar demographic properties (age, sex, education).

Clinical diagnosis was established according to current criteria for the aMCI stage of AD (29), svPPA (7), and rtvFTD (19), as reviewed to consensus by a clinical team including a neuropsychologist (A.B.-J.), and two behavioral neurologists (J.A.P. and J.B.L.). Additional, inclusion criteria included having detailed neuropsychological testing, MRI scans and AD biomarkers to support diagnoses.

The svPPA and rtvFTD groups were differentiated based on visual inspection of the laterality of anterior temporal lobe atrophy. CSF biomarkers included abeta-42, total-tau, phosphorylated-tau181 (p-tau181), and abeta42/p-tau181 ratios using the Athena Diagnostics ADmark® ELISA test, as indicators of AD pathology. Imaging biomarkers included FDG-PET scans and Amyvid™ PET amyloid scans. Although rtvFTD is sometimes conceptualized as including the behavioral variant of FTD when associated with predominant anterior temporal atrophy (6), any individuals meeting criteria for the behavioral phenotype (30) were excluded from the current study.

A control group (n = 22) was recruited prospectively for this study, to aid interpretation of results by anchoring relative losses of memory and hippocampal volumes in the FTD and aMCI groups to control values. The control participants were recruited and completed behavioral testing at Cleveland State University. They were screened to be matched to the FTD and aMCI samples in terms of demographic variables (age, sex, handedness, education, race/ethnicity). A history of medical conditions that could affect cognition was exclusionary, including cognitive impairment from neurodegenerative disease, developmental impairments, strokes, epilepsy, or any other self-reported neurological, psychiatric, or pharmacological conditions that may affect language, cognition, or study performance. They were required to have corrected vision sufficient to read and hearing sufficient to engage in conversation. All controls performed within normative (asymptomatic) ranges on the brief version of the Mini Mental State Examination-2 (31), and had delayed recall scores within normative ranges on the Hopkins Verbal Learning Test-Revised (32) and Brief Visuospatial Memory Test-Revised (33). No AD biomarkers were available for the control group. The control group completed MRI scans on the same scanner using the same imaging protocol used for the clinical samples (see Imaging section below).

Verbal memory was tested via the Logical Memory subtest from the Wechsler Memory Scale (34), Hopkins Verbal Learning Test-Revised, or Rey Auditory Verbal Learning Test (35). Visual memory was tested by the Visual Reproduction subtest from Wechsler Memory Scale (third and fourth editions (34, 36)) or Brief Visuospatial Memory Test-Revised. Raw delayed recall scores were converted to standard scores (mean = 100, standard deviation = 15) based on sex and age-specific normative data from each measure, in accordance with the scoring instructions for each test. The Boston Naming Test (37) was administered as a measure of language and object recognition, with scores also standardized (mean = 100, standard deviation = 15) using sex, age, and education-specific norms. Total scores from the Montreal Cognitive Assessment (MoCA) (38) were examined as a global measure of cognitive functioning in the patient groups (not administered to controls).

All participants were required to be primary English speakers.

All participants completed MRI scanning at the Cleveland Clinic, on a 3T Siemens Skyra (Erlangen, Germany) using a 20-channel head coil. Volumetric data were derived using a T1 MPRAGE sequence from the Alzheimer’s Disease Neuroimaging Initiative protocol. The images were post-processed using NeuroQuant™ software (Cortech Labs Inc., La Jolla, California), including quantification of hippocampal volumes in each hemisphere, which are reported as a percentage of each participant’s total intracranial volume (ICV). Whole brain volumes (also expressed as a percentage of ICV) are reported as a measure of overall atrophy.

Individuals with svPPA and rtvFTD represent a minority of individuals treated at our clinic, and small sample sizes were obtained. As such, our study design placed emphasis on the directionality of means and deviation of each clinical sample from control values, as we did not have sufficient power to compare groups via interaction terms in full factorial models. Nonparametric tests were employed to be more robust to small sample sizes. Cohen’s d effect sizes were calculated where values >0.2, >0.5, and > 0.8 indicate small, medium, and large effects, respectively (39).

Demographic and clinical characteristics were compared across all groups. Continuous variables were compared using Kruskal-Wallis test with posthoc Wilcoxon rank sum tests, while categorical variables were compared using chi-square test.

To clearly delineate how the three clinical samples deviated from typical performance in each memory modality, “memory loss” values were constructed by expressing each standardized test score in terms of standard deviations from control values (i.e., a z-score conversion using the mean and standard deviation from the control sample). Similarly, “hippocampal volume loss” values were constructed for each clinical group, expressing each participant’s ICV-corrected volumes in standard deviations from the control sample. Memory loss and hippocampal volume loss were compared between clinical samples using effect sizes and Wilcoxon rank sum tests.

Our overall hypothesis was that asymmetry in hippocampal atrophy would be associated with differential performance across verbal and visual memory modalities. To evaluate this association, asymmetry in hippocampal volumes was quantified by calculating a “hippocampal difference score,” subtracting left from right volumes. Likewise, a “memory difference score” was calculated by subtracting verbal from visual test values. Spearman correlation coefficients were calculated between these two difference scores across the three clinical samples, and in the two FTD groups where we anticipated asymmetry. Given the a priori and directional nature of our hypothesis, a one-tailed test criteria was employed.

Statistical significance was established throughout at p < 0.05. Due to small sample sizes, emphasis was placed on magnitudes of effect and there was no formal adjustment for multiple comparisons. All statistical analyses were conducted using SAS version 9.4 (SAS Institute Inc., Cary NC).

The demographic and clinical composition of each group is shown in Table 1. The groups did not significantly differ in terms of age or education (p > 0.05 on Kruskal-Wallis tests); and also did not significantly differ in sex, handedness, or race/ethnicity (p > 0.05 on all chi-square tests). Given the small sample sizes included in this study, this does not necessarily mean the groups were fully-matched on these variables, but rather that no flagrant mismatches were detected via inferential comparisons.

CSF data for AD biomarkers were available for 7/8 members of the aMCI group, 3/7 in the svPPA group, and 7/8 in the rtvFTD group (additional biomarkers including brain FDG-PET and Amyvid™ PET supported diagnosis in others). The Kruskal-Wallis tests were significant for all CSF biomarkers (all p < 0.01), indicating group differences. Pairwise comparisons showed that abeta-42 concentrations and abeta-42/p-tau181 ratios were significantly lower in the aMCI group compared to the two FTD groups (p < 0.0.05; d range-2.25 to-4.95). The rtvFTD group showed significantly lower total-tau and p-tau181 concentrations compared to the other two clinical samples (all p < 0.05; d range-1.31 to-2.6).

Amyvid™ PET amyloid scans were available for one participant with aMCI (who was amyloid positive), one participant with svPPA (amyloid negative), and one participant with rtvFTD (amyloid negative). FDG-PET scans were available for three participants with svPPA and one participant with rtvFTD, all showing characteristic anterior temporal hypometabolism, with preserved regional metabolism in posterior cingulate and parietotemporal regions that are affected in AD. These biomarker results suggest that AD and non-AD neuropathologies were likely dominant in the aMCI and FTD groups, respectively.

MoCA total scores did not differ across the three clinical groups (p = 0.21), but Boston Naming Test scores did significantly differ across groups (p < 0.001). Pairwise comparisons showed that naming was most impaired in the svPPA and rtvFTD groups (who did not differ from one another), with lesser naming impairment in the aMCI group (all p < 0.05). This fits with previous findings that confrontation naming is greatly disrupted by the presence of single-word comprehension deficits and/or visual agnosia in svPPA and rtvFTD, as successful naming requires recognition of the object and linkage with its corresponding lexical representation (19, 40, 41).

All three clinical groups showed significantly lower memory scores compared to controls, on both the verbal and visual memory modalities (all p < 0.05; d range 1.98 to 3.3). All three clinical groups also showed significantly lower hippocampal volumes compared to controls (all p < 0.05, d range 1.09 to 3.18), except the svPPA group, which did not reach conventional statistical significance for right hippocampal values when compared to controls (p = 0.10, d = 0.78). Differences between clinical groups are further examined in the following sections.

Whole brain volumes in the rtvFTD group were significantly lower than controls (p < 0.001, d = 2.46), and aMCI trended towards being lower than controls (p = 0.057; d = 0.99). The svPPA whole brain volumes, however, did not differ from controls (p = 0.35, d = 1.32). There were no significant differences in whole brain volumes between clinical groups (p = 0.57, d range 0.15 to 0.61).

The memory loss scores for each clinical sample are shown in Figure 1, expressed as standard deviations from control values. The aMCI group showed severe and symmetrical memory loss in both verbal and visual domains. The svPPA group showed a similar level of verbal memory loss to the aMCI group (p = 0.36; d = 0.30), but less visual memory loss (p = 0.10; d = 0.89). The rtvFTD group demonstrated the opposite pattern, with a similar degree of visual memory loss to the aMCI group (p = 0.83; d = 0.23) but less verbal memory loss than aMCI (p = 0.037; d = 0.86). Thus, the svPPA and rtvFTD groups showed similar memory loss to aMCI in their relative areas of weakness (verbal and visual material, respectively), and in their relative areas of preservation they showed loss when compared to controls (Table 1) but less severe than seen in aMCI.

Representative MRI scans from members of each group are shown in Figure 2A. Hippocampal volume loss scores for each clinical sample are shown in Figure 2B, expressed as standard deviations from control values. The svPPA group showed a greater volume loss than the aMCI group in the left hippocampus (p = 0.019; d = 1.5), while volumes in the right hemisphere did not differ from aMCI (p = 0.41; d = 0.49). The rtvFTD group showed the opposite pattern, with greater volume loss in the right (p < 0.01; d = 2.12) but not left hemisphere (p = 0.22; d = 0.65), compared to aMCI.

A nonparametric Spearman correlation between the two difference scores (left–right hippocampal volumes and verbal-visual memory scores) was significant when examined across all three clinical samples (Figure 3; r = 0.36; p = 0.048). This correlation was also significant when examined solely among the two FTD groups with significant asymmetry: the svPPA and rtvFTD groups (r = 0.48; p = 0.037).

The svPPA, rtvFTD, and aMCI groups all showed lower memory scores for both verbal and visual material, compared to neurotypical controls. All clinical samples also showed lower hippocampal volumes, in both hemispheres, compared to controls. When memory scores and hippocampal volumes in the clinical samples were re-expressed as “loss score” differences from control values, distinct brain-memory relationships in FTD became apparent which differed from those seen in aMCI, based on areas of relative preservation and loss.

In aMCI, symmetrical hippocampal atrophy was associated with equally poor memory for verbal and visual material. In FTD, our primary hypothesis was supported, such that asymmetry in hippocampal volumes was correlated with a disparity in scores between the verbal and visual memory testing platforms. Hippocampal volume loss in svPPA exceeded that of aMCI in the more atrophic left hemisphere, and memory loss in svPPA was greater for verbal than for visual material. The opposite pattern was observed in rtvFTD, such that right hippocampal volume loss was maximal (also exceeding that seen in aMCI), and memory loss in rtvFTD was greater for visual material. Both FTD groups showed memory loss equivalent to aMCI in their more affected memory domain (verbal for svPPA, visual for rtvFTD), while performance in their relatively preserved domain (visual for svPPA, verbal for rtvFTD) was intermediate to aMCI and controls. These findings reinforce the notion that in some respects svPPA and rtvFTD are mirror images of one another (21, 26).

These findings suggest that the relationship between hippocampal pathology and episodic memory in FTD differs from that seen in AD; lateralized atrophy in FTD is associated with domain-specific memory loss, leaving relative areas of preservation in memory.

The retrospective design employed in this study included small sample sizes. Emphasis was placed on the directionality of means and deviation of the FTD and aMCI groups from control values. As such, the results from this study can be considered as proof of concept rather than definitive evidence, demonstrating how domain-specific memory specializations in each hemisphere may play out with regards to temporal FTD variants. The current findings would benefit from replication with larger samples, employing more robust statistical models with group interaction terms.

With larger samples, it may also be possible to probe for additional factors that may mediate the apparent relationship between hippocampal asymmetry and memory domain observed in this study. For example, medial temporal lobe atrophy in svPPA does not occur in isolation, but alongside temporopolar and lateral temporal atrophy in the left-hemispheric language network. It is therefore possible that poor recall of verbal material is primarily driven by dysfunction in the language network rather than the episodic memory network (13). Likewise, medial temporal atrophy in rtvFTD co-occurs with atrophy in the object recognition network, including the right temporal pole and fusiform (19, 21, 42). Larger studies with targeted memory experiments could help to disentangle which networks are most directly responsible for apparent memory deficits in FTD.

Inclusion in this study was restricted to individuals with 1–2 years of symptoms. This yields certain strength to our study design: emergent anatomic changes are constrained within this early time frame (facilitating targeted group comparisons), and symptoms may still be relatively isolated to key domains of cognition. Restricting inclusion to initial years of symptoms does not, however, necessarily equate the groups in terms of their current severity of neuropathology burden, or the severity of those resultant symptoms. Although our retrospective study records did not include an overall measure of dementia severity (e.g., the Clinical Dementia Rating Scale (43)), the patient groups did not significantly differ in global cognition (as measured by MoCA total scores), which is highly correlated with functional impairment (44).

Unlike aMCI and AD, asymmetry is a core feature of brain-memory relationships in temporal variants of FTD. Lateralized hippocampal atrophy in our study was correlated with domain-specific memory impairments in FTD, supporting the inclusion of both verbal and visual memory tests during clinical assessment. Right hippocampal volume loss in svPPA was on par with that seen in aMCI, but visual memory impairments were not as severe. Conversely, left hippocampal volume loss in rtvFTD was similar to aMCI, but verbal memory was less affected. Further work is needed to determine why hippocampal atrophy triggered by AD pathology is generally more detrimental to episodic memory than atrophy caused by FTD pathology. It may be, for example, that posterior aspects of the hippocampus and/or other components of the wider episodic memory network remain relatively intact in FTD, as compared to AD (45, 46). It is also possible that as both verbal and visual domains of memory are affected in AD (unlike FTD where at least one domain is relatively spared), memory loss in AD is strikingly apparent in clinical evaluations. Future studies employing network analyses (rather than focusing on a single structure, as we did in this study) may reveal the extent to which verbal and visual memory are the product of distributed networks, and uncover additional mechanisms by which those functions are disrupted in diseases such as FTD.

In this study we focused on delayed recall, which represents only one aspect of episodic memory. Changes in autobiographical memory have been reported in FTD (47), but few studies have examined memory in detail specifically in the rtvFTD variant. Future studies may help to reveal to what extent other aspects of episodic memory may be affected in rtvFTD, such as immediate recall, recognition, and remote memory.

Even though TDP-43 is considered a common etiology of svPPA and rtvFTD, underlying tau pathology has also been reported in these variants (48). However, lower levels of p-tau in our rtvFTD cohort suggest that it is likely that our specific rtvFTD cases are more likely to be enriched for TDP-43 pathology than the svPPA cases (49). In cases where FTLD-tau is the underlying etiology for svPPA or rtvFTD, one can hypothesize that the episodic memory relationships noted with respect to TDP-43 related hippocampal atrophy (50) could be less prominent. This hypothesis needs to be carefully evaluated in future clinico-pathological studies.

It would also be helpful to achieve some sort of consensus as to the nature of the rtvFTD syndrome, and ultimately to settle on standardized diagnostic criteria. While there is general agreement that rtvFTD manifests as a bouquet of nonverbal symptoms, it remains unclear as to whether there is a core or signature impairment. The current findings suggest that nonverbal impairments in visual object processing are an important part of the syndrome. This fits with the finding that prosopagnosia is one of the most common symptoms in rtvFTD (19, 21). Impairments in object recognition (agnosia) are also apparent in svPPA, in cases where anterior temporal atrophy has become appreciable in the right hemisphere (i.e., bilateral) (42).

The rtvFTD syndrome has also been reported, however, to profoundly affect socioemotional function, person-specific knowledge, interoception, and reward processing (6, 18, 26). In keeping with these symptoms, some groups have adopted a wider view of rtvFTD, such that it also encompasses the behavioral variant of FTD when caused by predominant right anterior temporal atrophy (6). Others have questioned whether rtvFTD would be better conceptualized as manifesting in multiple distinct phenotypes rather than a single clinicopathologic entity (18, 19). Future studies may help to deepen our understanding of the relationships between proteinopathy, polymorphisms/mutations, structural changes, functional changes, and cognitive symptoms in rtvFTD, as has been the case in svPPA.