Direct Comparison of the Tau PET Tracers ¹⁸F-Flortaucipir and ¹⁸F-MK-6240 in Human Subjects

DISCUSSION

Amyloid radiotracers in human subjects with typical AD dementia show a consistent pattern of pathology across tracers and subjects despite differences in dynamic signal range and nonspecific binding characteristics (33–36). The similarity in the brain distribution of specific signal between Aβ radiotracers has facilitated standardization techniques, such as centiloid scaling (37,38), and is likely attributable to a single human isoform of fibrillar Aβ constituting a distinct brain pathology—amyloid plaques—which is the dominant signal source for Aβ in vivo imaging agents (39,40). Tau is considerably more complex, with 6 human isoforms, more posttranslational modifications, and a greater diversity of pathologic lesions (e.g., NFTs, astroglial and oligodendroglial tau inclusions, or Pick bodies), stages of tangle maturity (41), and ultrastructural conformations (i.e., PHF, straight filaments, or twisted filaments), although in AD the most prevalent species are NFTs comprising a PHF combination of 3R and 4R tau isoforms (42,43). In vitro binding studies of flortaucipir and MK-6240 suggest low affinity to tau pathologies other than PHF-tau (17,29,44–46), although a few in vivo imaging studies have suggested a possible or weak sensitivity of ¹⁸F-flortaucipir to 4R-tau deposits in some primary tauopathies (47–49).

As suggested by autoradiography (44), the present study suggested that ¹⁸F-flortaucipir and ¹⁸F-MK-6240 are almost certainly detecting the same AD- and age-related tau pathology. This possibility is evidenced by the similarity in radiotracer retention patterns across a spectrum of disease (Fig. 1), highly correlated SUVR outcomes across Braak stage regions, and voxel-based analyses that show no differences in patterns of specific binding to pathologic tau deposits (Fig. 5). In some Braak regions, we observed a slightly elevated floor signal for ¹⁸F-flortaucipir compared with ¹⁸F-MK-6240, suggesting the former to have a slightly higher nonspecific retention. This was most apparent in Braak II, as is likely attributable to spill-in from off-target binding of ¹⁸F-flortaucipir in the choroid plexus.

Our approach to visual interpretation of ¹⁸F-flortaucipir and ¹⁸F-MK-6240 images was intended to mirror a clinical nuclear medicine environment, albeit with slightly more granularity than a global rating. Although we observed only 1 instance in our small cohort in which a subject (MCI1) was adjudicated overall to be positive in either the MTL or the NEO but not both, conceivably distinct ratings for MTL and NEO tau signal may provide some differentiation of the diverse tau phenotypes previously described (3) and, considering Aβ status, may help to differentiate AD pathology from normal aging processes (e.g., primary age-related tauopathy) characterized by MTL tau deposits that may occur independently of Aβ (50,51).

Although we observed differences in dynamic signal range and off-target binding between tracers, we expected either would perform well in visual assessments, in which off-target binding can be more easily accounted for. This expectation was supported by our results, which showed complete concordance between tracers in overall ratings for both the MTL and the NEO (Table 2) and substantial agreement between raters (κ > 0.73, Table 3), regardless of tracer or region.

Comparing visual assessments with diagnoses, both MTL and NEO tau pathology was present in all AD subjects. Among NC subjects, 7 of 9 showed no evidence of tau pathology in either the MTL or the NEO, whereas NC1 and NC4 were positive in both and had the highest ¹¹C-PiB SUVRs of all NC subjects, although only NC1 was quantitatively Aβ-positive (¹¹C-PiB CL = 50). The MCI subject, who was Aβ-negative (¹¹C-PiB CL = 0), was tau-positive only in MTL. This subject may be an example of primary age-related tauopathy.

Comparison of SUVR images shows that both ¹⁸F-flortaucipir and ¹⁸F-MK-6240 indicated the same tau pathology, although ¹⁸F-MK-6240 showed a nearly 2-fold higher dynamic range (Fig. 2) as indexed by regression slopes ranging from 1.45 to 1.98. A similarity between ¹⁸F-flortaucipir and ¹⁸F-MK-6240 in cerebellar gray matter SUVs, which are used to compute SUVR, indicated that this observation is not attributable to differences in nonspecific retention. In vitro saturation binding studies of ³H-MK-6240 and ³H-flortaucipir conducted using the same AD tissue homogenates showed ³H-MK-6240 to have a 3- to 10-fold higher affinity (K_D) for PHF-tau than did ³H-flortaucipir, as well as a 3- to 5-fold higher B_max/K_D ratio (17,46). Indeed, differences in the pharmacologic properties may reasonably explain the increased dynamic range of ¹⁸F-MK-6240 SUVR, although other factors such as nonspecific binding, radiotracer metabolism, and rates of plasma and reference region clearance may influence in vivo specific binding measures. Although both tracers appeared to be well suited to visual interpretation, the greater dynamic range of ¹⁸F-MK-6240 may represent an advantage for longitudinal studies of tau progression or treatment response, in which detecting small interval changes is key.

An examination of the dispersion of the SUVR data (Fig. 2) shows that for all regions except Braak II there was a subject cluster, with SUVRs of approximately 1 for both radiotracers in all subjects visually adjudicated to be negative, whereas the positive cases covered a much broader range, with few subjects overlapping with the negative cluster. The fact that there were only a few visually positive subjects with low SUVRs likely indicates that they represent subjects for whom the raters identified focally intense radiotracer uptake but that the focus of increased signal was diluted in the averaging of all voxels in the respective Braak stage region, such as the relatively large Braak III–VI regions shown in Figure 6. An example of such a subject is AD1 (Fig. 1), who showed a clear unilateral focus of increased uptake in the left precuneus that, by visual ratings, was indicated as positive for the MTL and NEO with both radiotracers, but ¹⁸F-flortaucipir and ¹⁸F-MK-6240 SUVRs for the Braak V region in this subject were only 1.16 and 1.09. This example highlights a potential limitation of sampling PET images of tau radiotracers in accordance with Braak staging. Another complication is that in some elderly subjects with or without cognitive impairment, significant MTL (Braak stage I–III region) pathology may occur independently of Aβ. These subjects may represent cases of primary age-related tauopathy. Therefore, classifying tau status (T+/T−) on the basis of MTL pathology alone would presumably reduce the specificity of a pathologic diagnosis of AD. It is likely that the most sensitive and specific indicators of tau lesions consistent with AD neuropathologic change will require tau PET positivity beyond MTL structures. However, the Braak IV region is relatively large and potentially suffers from the limitation of diluting a focal signal that might be the earliest indicator of neocortical spread of tau pathology. For this reason, others have moved toward a data-driven approach with a more granular tissue sampling strategy (52).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

Grouping of FreeSurfer regions to correspond to Braak pathologic stages (I–VI).

¹⁸F-flortaucipir images often exhibit elevated off-target binding in the striatum and choroid plexus (Fig. 4), as well as other tissues, which may occur independently of neurodegenerative disease pathology as previously reported (6). For ¹⁸F-MK-6240, the frequent observation of elevated signal arising from the meninges and other extracerebral structures (Fig. 3) was in accordance with previous observations (21). In some cases, it was apparent that spill-in of off-target ¹⁸F-MK-6240 signal from the meninges could impact the quantification of signal in cortical brain regions as well as the cerebellum, although the fact that meningeal off-target signal is not apparent on ¹⁸F-flortaucipir images yet the SUVR outcomes of ¹⁸F-flortaucipir and ¹⁸F-MK-6240 correlated strongly (Fig. 2) suggests that this effect does not represent a major confounder.

The off-target signal in bone observed with both tracers, most notably in female subjects with hyperostosis frontalis interna, does not appear to be completely explained by in vivo defluorination of these tracers, as we did not observe widespread bone uptake consistent with ¹⁸F-fluoride scans and the inspection of batch records from our ¹⁸F-MK-6240 radiosyntheses did not show evidence of significant residual ¹⁸F-fluoride in the injectate. Intracranial calcifications are a common and normal age-related radiographic finding, often described in the pineal gland, habenula, choroid plexus, basal ganglia, falx cerebri, dura mater, petroclinoid ligaments, superior sagittal sinus, and dentate nuclei of the cerebellum and the hippocampus (53). Interestingly, these regions overlap areas where off-target binding of these tau radiotracers is often observed, but inspection of low-dose CT scans showed no macroscopic calcifications in the choroid plexus of our study subjects.

Limitations of the present study include a small sample size and a limited range of pathology and degree of clinical impairment. Only 1 of 9 NC subjects was globally amyloid positive and the only MCI subject was amyloid negative. The 5 AD subjects showed mild to moderate cognitive impairment, and we observed relatively limited tau pathology in Braak stages V and VI across the sample. Given the small sample size of our pathologically heterogeneous cohort, in which patients with advanced tau pathology comprise over one third of the study cohort, it might not be possible to generalize our measures of interrater reliability to other subject cohorts. We would expect there to be considerably less agreement and lower reliability in studies of cohorts that comprised predominantly cognitively normal elderly individuals, in whom tau burden is less. Indeed, we observed some discordance between raters among nondemented subjects (Supplemental Fig. 4). In our study, ¹⁸F-MK-6240 injected doses were limited to 185 MBq to meet the organ dosimetry limits of the University of Pittsburgh’s Radioactive Drug Research Committee. Another limitation of the present study was the lack of measures of intrarater reliability, as the small size and high disease burden in our cohort would be expected to yield a high level of intrarater reliability that also could not be generalized to other cohorts.