In this autopsy immunohistology study, we hypothesized that the fornix, a major efferent outflow tract of the hippocampal formation, is a conduit for the propagation of p-MAPT neuropathology from the hippocampus to the basal forebrain in early AD. To test this hypothesis, we analyzed p-MAPT immunostaining in the fornix and its efferent targets in the basal forebrain in human brain autopsies with early to advanced AD neuropathologic changes. We expected that the fornix would recapitulate a core feature of the Braak and NIA/AA staging systems in the spread of p-MAPT neuropathology across connected brain regions with involvement downstream of the hippocampal formation. Indeed, in cases of B0-1 neurofibrillary degeneration with none or minimal involvement of the hippocampal formation by p-MAPT neuropathology, we observed no significant p-MAPT (AT8) immunostaining in the fornix or efferent target nuclei. In cases of B2 neurofibrillary degeneration, we observed p-MAPT staining in axonal processes of the fornix and in precommissural nuclei (septum and nucleus accumbens). The mammillary body, innervated by postcommissural fibers of the fornix, demonstrated tauopathy that was predominantly astroglial in stage B2 and that was mixed neuronal and astroglial in stage B3. P-MAPT neuropathology was sparse in the ATN, an efferent target of the fornix and of the mammillary bodies via the mammillothalamic tract, in NIA/AA stage B2 but rose significantly in NIA/AA stage B3. These results of our autopsy study demonstrate that the fornix is a conduit for p-MAPT neuropathology in AD and are compatible with p-MAPT neuropathology propagation from the hippocampal formation to the basal forebrain nuclei via the fornix as AD-related tauopathy progresses.
Braak and Braak [2–4] described the stereotypical propagation of p-MAPT neuropathology in AD which is the basis of the neurofibrillary degeneration staging system that bears their name. The serial propagation between synaptically connected regions, like the entorhinal cortex and the hippocampal formation, supports the hypothesis that axons and synapses play an important role. Mechanistically, recent work has supported the concept that MAPT neuropathology is propagated along connected brain regions and that the synapse itself is an important factor for, and perhaps a substrate of, transneuronal p-MAPT neuropathology propagation [5, 6, 8]. Our findings, while generally not surprising in the context of these lines of evidence, are important because they fill a gap in our knowledge of p-MAPT propagation in structures not routinely sampled in AD neuropathologic examinations.
Since cognitive impairment in AD is correlated with the spread of pathologic tau [14, 16], it is possible that the risk of progression from MCI to AD might be predicted by the spread of p-MAPT from the hippocampus via the fornix. Consistent with this premise, elevated CSF MAPT and p-MAPT have become important biomarkers in the clinical diagnosis of AD . Diffusion tensor imaging MRI studies have suggested that increased diffusion measures in the fornix may predict progression in AD and are suggested to relate to p-MAPT related axonal injury [1, 18], however, definitive pathologic correlation is lacking. Ongoing work in in vivo MAPT imaging  might also facilitate detection of p-MAPT propagation in the fornix.
We performed correlation analyses of p-MAPT staining in the fornix and basal forebrain nuclei to further probe the tauopathy patterns in these interconnected structures. The density of p-MAPT staining in the fornix and all the basal forebrain structures demonstrated significant correlation with the Braak stage of neurofibrillary degeneration, compatible with the progressive spread of tauopathy to these structures as AD neurofibrillary degeneration advances along the regional pattern described by Braak and Braak [2–4]. Interestingly, correlation analysis of the density of fornix tauopathy and the density of tauopathy in basal forebrain structures was not uniform. The strongest correlation between tauopathy in the fornix and nucleus basalis of Meynert suggests the possibility that nucleus basalis projections to the hippocampus through the fornix contribute to fornix tauopathy in addition to efferent hippocampal formation projections. Tauopathy in the fornix was also significantly correlated with tauopathy in the septum but was not significantly correlated with tauopathy in the NA or ATN. In our cohort, there were cases with relatively higher levels of tauopathy in the NA and ATN but lower levels of fornix p-MAPT that contributed to this lack of correlation. Since these cases represent NIA/AA staged B2 and B3, lack of concordance might be due to spread of p-MAPT neuropathology to the basal forebrain via other pathways, for example via the amygdala in the case of the NA, or could reflect decreases in fornix p-MAPT-positive axons due to axonal loss.
Our autopsy study also highlighted two additional tauopathies that can involve the mesiotemporal lobe and basal forebrain: PART  and ARTAG . PART is characterized by neurofibrillary degeneration in a pattern resembling AD but in the absence of significant Aβ neuropathology. Since we performed Aβ immunostains only on the middle and inferior temporal gyri and are unable to completely exclude the possibility of amyloid plaques in other neocortical regions, we categorized patients lacking amyloid plaques as possible PART. Seven brains in our study demonstrated possible PART confined to the entorhinal cortex with sparse involvement of the CA1 sector of the hippocampus (Braak stage I-II). Overall, p-MAPT staining was sparse in the fornix in cases with entorhinal neurofibrillary degeneration and there was no significant difference in fornix p-MAPT axonal staining between the 12 brains with B1 Alzheimer disease neuropathologic changes (8.7 ± 2.6 p-MAPT-positive axonal profiles) and the seven brains with possible Braak stage I-II PART (5.7 ± 2.2 p-MAPT-positive axonal profiles). Only one brain in our autopsy series demonstrated possible Braak stage III PART and it showed approximately ½ of the mean axonal p-MAPT staining for the B2 neurofibrillary degeneration group. These results suggest that fornix p-MAPT neuropathology can be seen in Braak stage ≥ III PART. Since dementia is uncommon in patients with PART , the finding of p-MAPT-related axonal damage in the fornix highlights the potential importance of in vivo Aβ imaging and biomarkers to differentiate patients with AD who are more likely to progress to dementia.
Additionally, several brains demonstrated abundant p-MAPT immunoreactive thorn-shaped astrocytes in the MB. We suspect that this finding is most likely related to ARTAG  as it was seen in all NIA/AA stages of neurofibrillary degeneration and its density correlated with patient age. However it is interesting to note that MB tau astrogliopathy was most prominent in NIA/AA stage B2 of neurofibrillary degeneration and preceded neuronal tau pathology in the MB in our autopsy cohort. Furthermore, there was a modest but statistically significant correlation between the combination of astroglial and neuronal tauopathy in the MB and tauopathy in the fornix. Although we favor these findings to represent ARTAG, they could indicate a role for tau astrogliopathy in tau propagation from the fornix to the mammillary body. Further studied are needed to address the significance of tau astrogliopathy in the MB.
The major limitations of our study are its cross-sectional design and its small sample size. We have inferred from the cross-sectional design of this autopsy study that the fornix propagates p-MAPT neuropathology from the hippocampus to efferent structures in the medial basal forebrain. However, in vivo tau imaging, DTI and post-mortem neuropathologic correlation in patients with neuropsychiatric evaluations will be needed to determine if this is the true progression sequence of AD-related tau pathology from the hippocampus and its significance with respect to the risk of progression from MCI to dementia. Out of concern for the possible impairment of antigen immunostaining with chronic formalin-fixation, we limited our study to archived autopsies no older than 2 years. Furthermore, as p-MAPT neuropathology can be associated with diverse brain diseases, we eliminated cases with brain pathologies that might affect the pattern of p-MAPT neuropathology or that might affect our immunohistochemical detection of p-MAPT neuropathology. Despite the resultant small sample size or our study, our data robustly demonstrate that axonal AT8 immunoreactivity in the fornix is present in the limbic stage of AD neuropathology and does not precede hippocampal involvement. It is possible that tau antibodies that detect earlier phosphorylation states or oligomeric tau could render additional insights into tau propagation among the medial temporal lobe and basal forebrain. Nonetheless, our findings largely support the expected model of AD-related p-MAPT neuropathology propagation from the hippocampus to the basal forebrain via the fornix in NIA/AA stages B2 and B3 of neurofibrillary degeneration. However, our findings are compatible with the possibilities that the fornix carries p-MAPT to the hippocampal formation from the nucleus basalis of Meynert and that basal forebrain nuclei may propagate p-MAPT neuropathology from structures other than the fornix.