Iron is known to bind to tau proteins (Sayre et al., 2000), induce tau protein phosphorylation, and play an important role in the accumulation of hyperphosphorylated tau to form neurofibrillary tangles (Yamamoto et al., 2002). In postmortem studies, iron deposition in the cortex, particularly the temporal cortex, has been associated with the severity of beta amyloid (Aβ) and tau pathology in Alzheimer’s disease (AD), supporting an important mechanistic role for iron in AD (van Duijn et al., 2017; Bulk et al., 2018; Ayton et al., 2021). Iron burden in the inferior temporal cortex, specifically, has been associated with the formation of neurofibrillary tangles (Ayton et al., 2021), with the temporal lobe showing severe iron accumulation and tau burden compared to other cortical regions (Bulk et al., 2018).

Quantitative susceptibility mapping (QSM) is a relatively new, non-invasive, magnetic resonance imaging (MRI) technique to capture cerebral iron. It can provide estimates of the local magnetic susceptibility in tissues at the voxel-level (Langkammer et al., 2012) and can detect paramagnetic (positive magnetic susceptibility) iron burden (Wang and Liu, 2015) and diamagnetic (negative magnetic susceptibility) myelin loss (Liu et al., 2015) in the brain. Several previous studies have assessed differences in susceptibility signatures in postmortem tissue and concluded that the bulk of the magnetic susceptibility in gray matter structures in AD is contributed by the iron content (Hallgren and Sourander, 1958; Langkammer et al., 2012; Tiepolt et al., 2020). There is evidence that QSM susceptibility in the temporal lobe is associated with tau deposition measured using positron emission tomography (PET) across the AD continuum (including cognitively unimpaired and impaired individuals) (Spotorno et al., 2020).

However, it is unclear whether iron deposition is related to the heterogeneity in tau deposition observed across atypical clinical presentations of AD. Atypical clinical presentations of AD are non-amnestic and typically characterized by the presence of visual, language, behavioral, executive, or motor difficulties (Galton et al., 2000; Graff-Radford et al., 2021). Spatial patterns of cortical tau accumulation on PET differ across the different variants of atypical AD, with prominent burden in the posterior regions in the visual variant, left dominant temporal regions in the language variant, temporo-parietal regions in the behavioral/executive variant and temporo-parieto-occipital regions with involvement of the sensorimotor regions in the motor variant (Ossenkoppele et al., 2016; Tetzloff et al., 2018). We have previously showed differing patterns of abnormal magnetic susceptibility in the cortex of the visual and language variants of atypical AD, specifically in the temporo-parieto-occipital regions and putamen in the visual variant and in the temporo-occipital regions, caudate, putamen and substantia nigra in the language variant (Singh et al., 2022). Bearing in mind, that iron is known to bind to tau, induce phosphorylation, support hyperphosphorylated tau accumulation and tangle formation in amnestic AD (Sayre et al., 2000; Yamamoto et al., 2002), we expect to observe a similar pathobiological role of iron in atypical presentations of AD. Therefore, the primary aim of this exploratory study was to investigate the regional relationship between magnetic susceptibility measured on QSM and tau uptake measured on PET in atypical AD to determine whether there is evidence that iron deposition may be related to the heterogeneity observed in tau deposition in AD. Considering the regional overlap between both modalities and the evidence of QSM-tau PET associations in the temporal lobe in amnestic AD (Spotorno et al., 2020), we hypothesized that we may observe local relationships, i.e., local regional colocalization between cortical tau uptake and increased susceptibility in key atypical AD regions such as temporo-parieto-occipital regions.

The clinical and demographic features of the cohort are shown in Table 1. The cohort had a median age of 64 years at the time of assessment, and 70% of them were female. Performance on the clinical tests was consistent with the patients’ AD diagnoses.

At the voxel-level, positive correlations were observed between magnetic susceptibility and tau-PET uptake in the left caudate after the RFT-based multiple comparison correction (Figure 1). At the ROI-level, no significant correlations were found after correction for multiple comparisons. However, uncorrected data showed a moderate association (absolute Spearman rho ≥0.50) in three ROI-pairs (shown with arrows) (Figure 2), namely between tau-PET uptake in the right superior occipital lobe and magnetic susceptibility in the right middle occipital lobe (r = 0.57, p < 0.001, FDR-corrected p = 0.59) and the left substantia nigra (r = −0.50, p < 0.01, FDR-corrected p = 0.96), and between tau-PET uptake in the right inferior occipital lobe and magnetic susceptibility in the right substantia nigra (r = −0.53, p < 0.001, FDR-corrected p = 0.95) (Figures 2, 3). Spearman rho for the relationship between susceptibility and tau-PET uptake in the left caudate was 0.37, p = 0.02, FDR-corrected p = 0.96. The sensitivity analyses within the visual variant showed no significant voxel-level clusters and ROI-level correlations after correction for multiple comparisons. The uncorrected ROI-level data showed an association between tau-PET uptake in the left superior occipital lobe and magnetic susceptibility in the left middle (r = 0.53, p = 0.01, FDR-corrected p > 0.99) and superior temporal lobes (r = 0.52, p = 0.01, FDR-corrected p > 0.99) (Supplementary Figure 1).

This study examined the relationship between iron burden measured on QSM and tau accumulation measured on AV-1451 PET in atypical clinical presentations of AD. We found some weak evidence for positive local relationships between susceptibility and tau-PET in the caudate nucleus and occipital lobe.

The whole brain voxel-level analysis revealed a positive correlation between magnetic susceptibility and tau-PET uptake in the left caudate. This relationship was also significant, albeit weak, at the ROI-level. Greater magnetic susceptibility in the caudate has been noted in the language variant of AD when compared to cognitively unimpaired individuals (Singh et al., 2022). However, the caudate does not typically show elevated tau-PET uptake in the visual, language, behavioral, and dysexecutive variants of AD (Ossenkoppele et al., 2016; Tetzloff et al., 2018), although it can be affected in the motor variant of AD (Smith et al., 2017). It is important to note that of the 40 atypical AD patients in this study, only three patients were diagnosed with the motor variant and eight were diagnosed with the language variant of AD. One explanation for this finding could be that basal ganglia structures, including the caudate, are known to be “off-target” sites for AV-1451 tracer uptake (Marquie et al., 2015; Johnson et al., 2016), with off-target binding increasing with older age. Moreover, iron levels are known to increase steadily with age, particularly after middle age in the caudate nucleus (Hallgren and Sourander, 1958; Ramos et al., 2014). Likewise, previous studies have also shown that AV-1451 PET and iron-sensitive MR (R2* or QSM) signals positively correlated in the caudate, putamen, and pallidum in healthy controls, which is an unexpected finding as these individuals should not have iron-related tau deposition (Choi et al., 2018; Satoh et al., 2024). In our previous study, we performed the same voxel-based analysis on 67 cognitively normal controls and found positive correlations between tau PET and QSM in the bilateral caudate and pallidum (Satoh et al., 2024). Therefore, the presence of positively correlated signal between the two modalities in the caudate may be reflective of off-target binding to iron rather than biological relationships, such as the coexistence or the interaction between iron and tau (Yamamoto et al., 2002; Spotorno et al., 2020). The reason for a leftward asymmetry is unclear but has been reported before in the caudate nucleus (Holz et al., 2022).

No significant associations were noted between magnetic susceptibility and tau-PET uptake at the ROI-level after correction for multiple comparisons. However, uncorrected data showed a positive correlation between modalities in the occipital lobe. Although this finding is consistent with the spatial distribution of tau accumulation (Day et al., 2017; Singh et al., 2024) and greater occipital susceptibility reported in the visual and language variants of AD (Singh et al., 2022), the correlation is weak and does not survive correction. One could theorize the presence of a mechanistic relationship between iron and tau deposition in the occipital lobe in atypical AD, but these findings will have to be confirmed in a larger cohort. Lastly, the sensitivity analyses showed no significant associations after correction for multiple comparisons, but the uncorrected data showed an association between the occipital tau uptake and temporal susceptibility in the visual variant of AD. One could theorize that this finding could be suggesting toward a relationship between iron and tau deposition, as previous literature has shown significant tau uptake in occipital lobe (Day et al., 2017) and greater temporal lobe susceptibility (Singh et al., 2022) in visual AD. However, this subgroup analysis was underpowered, and the uncorrected results may include false positives, which should be interpreted cautiously. Overall, no positive evidence was found even in the best-case exploratory scenario once corrected, supporting the overall null finding.

A somewhat counterintuitive negative relationship was observed between substantia nigra susceptibility and occipital tau-PET uptake, whereby lower substantia nigra iron burden was associated with greater tau uptake in the occipital lobe. Substantia nigra is a major site for iron storage and is known to show a steady increase in iron levels with aging (Zecca et al., 2001). We have observed increased susceptibility in the substantia nigra in the language variant of AD but not the visual variant of AD (Singh et al., 2022). The reason for a negative association with occipital tau-PET uptake is unclear. This finding, although uncorrected, could be suggesting that iron deposition in brainstem nuclei may have a different mechanistic relationship with tau compared to iron in the cortex in AD. It is possible that susceptibility findings in cortical regions are influenced by myelin loss and white matter damage, whereas this is not expected to be a confound in brainstem grey matter nuclei. However, since the correlation did not survive correction for FDR, we suggest interpreting this finding cautiously and replicating these results in a larger cohort.

Strengths of this study include the consistent neuroimaging protocols and clinical evaluations with models that control for differences in age and sex effects. This is also the first study that directly compares the relationship between iron content measured on QSM and tau accumulation measured on AV-1451 PET in atypical clinical presentations of AD. Potential limitations include the relatively small sample size and the lack of power to examine relationships within each AD variant separately. Another limitation is the presence of off-target AV-1451 PET uptake (Marquie et al., 2015; Johnson et al., 2016) and age-related iron accumulation (Hallgren and Sourander, 1958; Ramos et al., 2014) in the basal ganglia structures. More specifically, the AV-1451 PET ligand is known to bind with iron-rich area in the basal ganglia, particularly to neuromelanin, which is known to accumulate with age (Marquie et al., 2015; Lowe et al., 2016; Choi et al., 2018; Langley et al., 2024). Additionally, AV-1451 PET and QSM signals are known to be positively correlated in the basal ganglia of healthy controls, suggesting toward iron-related tau deposition in these individuals (Choi et al., 2018; Satoh et al., 2024). Together, these issues make interpretation of the caudate findings challenging and raise questions on the existence of meaningful QSM-tau relationships and their utility in the deep nuclei. Lastly, our cohort consist predominantly of the language and visual variants of AD, so further studies will be needed to determine whether these findings can be generalized to cohorts with different phenotypic distributions. The reproducibility of these findings will also have to be confirmed in a larger cohort. Multi-site data aggregation would be needed to draw firm and comprehensive conclusions about this relatively rare disorder with diverse phenotypes.

Overall, these results show little evidence to suggest that iron deposition is strongly related to or driving regional patterns of tau deposition in atypical clinical presentations of AD. More work is needed in larger cohorts to better understand the pathophysiological role of iron in AD.