Neurofibrillary tangles spread from the hippocampus into the neocortex via axonal highways, but only after Aβ pushes them out the door. So claims a study published in Nature Neuroscience on February 5. Researchers led by Keith Johnson at Massachusetts General Hospital in Boston combined multiple imaging modalities to connect Aβ accumulation to an eroding axonal circuitry and tau accumulation outside the medial temporal lobe. They found that tau fibrils spread via synaptically linked regions, as opposed to those that are merely nearby. Tau did this mainly in people with elevated Aβ. Tau’s appearance outside of the hippocampus then coincided with the first inklings of memory problems.

  • Researchers combined multiple imaging modalities to draw causal connections between AD pathology and cognition.
  • Aβ plaques predicted a crumbling white-matter tract connecting the hippocampus to the posterior cingulate cortex. This in turn predicted accumulation of tau there.
  • Tau accumulation in the PCC predicted a decline in memory. 

“This study is a terrific example of translational neuroscience,” commented Rik Ossenkoppele of VU University Medical Center in Amsterdam. “[The findings] emphasize the importance of Aβ as a potential trigger for downstream effects, while tau pathology might be the actual driver of neurodegeneration and subsequent cognitive decline.”

Ever since Braak staging traced the insidious trajectory of tau tangles from the medial temporal lobe (mTL) out into the cortex, researchers have sought to understand what triggers this dangerous shift, and how it proceeds (see Research Timeline). Animal studies have implicated transsynaptic spread as tau’s travel mode of choice, and fingered Aβ as an instigator of the process (de Calignon et al., 2012; Ahmed et al., 2014Pooler et al., 2015Dec 2017 news). The recent advent of tau-PET tracers has allowed researchers to address these relationships in living people, and cross-sectional findings have so far revealed that the movement of tau pathology from the mTL into the cortex is a pathological event that Aβ may initiate (Mar 2016 news; Aug 2016 news). 

Connecting the Boxes.

The study showed relationships among imaging biomarkers and memory (larger blue box). Because tau PET imaging was not yet available at baseline, relationships with some parameters (green boxes) could not be directly tested. [Courtesy of Jacobs et al., Nature Neuroscience, 2018.]

To clinch causal connections between Aβ, tau, neural circuitry, and ultimately, memory problems (see graphic at left), first author Heidi Jacobs and colleagues made use of multimodal longitudinal data from the Harvard Aging Brain Study (HABS). At baseline, the study’s 256 participants were cognitively healthy and averaged 73.5 years old. Over the following seven years, participants underwent annual cognitive tests as well as multiple forms of brain imaging. These included amyloid-PET scans, magnetic resonance imaging (MRI) to assess hippocampal volume, and diffusion tensor imaging (DTI) to interrogate the integrity of white-matter tracts. After the third year of the study, researchers added PET imaging with the tau pathology-specific tracer flortaucipir to the mix. 

First, they asked whether elevated Aβ correlated with hippocampal volume loss, which can serve as a proxy for neurodegeneration. They found that having neocortical Aβ accumulation based on PiB-PET uptake at baseline predicted steeper shrinkage of the hippocampus over the following six years. Though hippocampal volume loss can be attributed to factors other than AD pathology, the first tau-PET measurements taken in the entorhinal cortex between the third and fourth years of the study correlated tightly with the extent of hippocampal shrinkage at the nearest time point, suggesting hippocampal atrophy was indeed a rough proxy for tau pathology in this population.

The researchers next asked whether hippocampal atrophy would predict abnormalities in the hippocampal cingulum bundle. The HCB is a white-matter tract that projects from the hippocampus into the posterior cingulate cortex (PCC). They investigated this tract not only because it is easy to interrogate with DTI, but also because it represents a likely escape route for toxic forms of tau from the mTL. Interestingly, they found that low hippocampal volume at baseline predicted declining HCB diffusivity, an indicator of waning integrity, over time. In contrast, having a small hippocampus did not predict damage to the uncinate fasciculus. The UF is a white-matter tract close to the hippocampus, but does not innervate it. Notably, damage to the HCB did not predict hippocampal shrinkage, suggesting atrophy there preceded the white-matter damage.

Would crumbling HCB integrity coincide with the spread of tau pathology to the downstream PCC? Yes, according to tau PET data. The researchers reported that HCB damage at baseline predicted a steeper annual increase in PCC tau. Once again, white-matter abnormalities in the nearby, but unconnected, UF had no bearing on PCC tau. Furthermore, HCB abnormalities did not correlate with levels of tau in the inferior temporal cortex, a region adjacent to, but not tightly connected with, the hippocampus.

Strikingly, the researchers found that low HCB diffusivity only associated with a rise in PCC tau in people who already had elevated Aβ at baseline. The findings support a model in which Aβ triggers tau propagation along the HCB, which in turn leads to the accumulation of tau in the downstream PCC. 

Path to Problems. Imaging uncovered links among amyloid, a damaged HCB fiber track, posterior cingulate tau, and memory (see boxes). High diffusivity in HCB (red line, bottom) correlated with memory decline in people who tested positive for amyloid and tau (right graph), but not negative (left graph). [Courtesy of Jacobs et al., Nature Neuroscience, 2018.]

How does cognition fit into this cascade? That, after all, is what it’s all about. The scientists found that low white-matter integrity of the HCB at baseline predicted slippage of memory over the following six years, but not of executive function. They next classified participants as having either high or low PCC tau, based on a standardized PET uptake ratio of 1.28, and found that high tau drove the association between HCB integrity and memory decline. Narrowing things further, they found that among people with elevated PCC tau, the connection between HCB integrity and memory decline only occurred in those with elevated Aβ at baseline. 

Jacobs told Alzforum that this imaging study cannot nail down specific molecular mechanisms of transsynaptic transfer of tau. That said, the findings do support the idea that neocortical Aβ aggregation somehow incites tau pathology to spread from the mTL into the cortex via synaptic connections, rather than through simple diffusion into nearby regions. Based on the changes in diffusivity the researchers observed in the HCB, they proposed that tau’s propagation through the tract somehow disrupted the structure of both axons and myelin along the way. However, they also acknowledged the possibility that neurodegeneration in the hippocampus, on top of tau’s trailblazing, could have caused the alterations in the connected white-matter tract.

Zeshan Ahmed of Eli Lilly in Surrey, England, who previously reported that tau spread transsynaptically in a mouse model of tauopathy, expressed excitement that the human findings aligned well with those in various animal models. “A greater understanding of the underlying mechanisms of disease progression in human patients opens doors to new therapeutic strategies, but also helps validate the in vivo models we use to develop much-needed therapies,” he told Alzforum.

Michel Goedert of the MRC Laboratory of Molecular Biology in Cambridge, England, agreed. Goedert also broached the question of how this Aβ-associated tau propagation relates to tau’s behavior in other tauopathies. “One must bear in mind that prion-like spreading of tau aggregates is also believed to be important in sporadic tauopathies that lack Aβ deposits, such as Pick’s disease and progressive supranuclear palsy,” he wrote. “Does it follow that the spreading of tau pathology is less effective in those diseases?”

Ahmed also wondered about propagation in other tauopathies, speculating that other triggers might substitute for Aβ in those cases. “Maybe some forms of tau are more likely to aggregate, or are more concentrated for some reason,” he said. Mutations in tau can cause tauopathies other than AD. Ahmed added that much is left to learn about exactly how tau transfers from one neuron to another, and thus how best to target this propagation. Interestingly, recent findings have implicated presynaptic tau, either in the context of dystrophic axons congregating around Aβ plaques, or within cells expressing mutant forms of tau that adhere to synaptic vesicles, as a pathological form that could facilitate propagation (Dec 2017 newsFeb 2018 news). 

Jacobs and colleagues are continuing to collect longitudinal imaging and cognitive data on the HABS cohort, and plan to investigate tau’s propagation beyond the PCC should participants develop early AD. Given the connection between Aβ accumulation, the propagation of tau pathology, and subsequent memory problems, they proposed tau-PET imaging could serve as an additional outcome measure in Aβ-targeted trials.—Jessica Shugart



  1. Experimental studies showed that assembled mutant human tau propagates through connectivity, not proximity, in the absence of Aβ deposits (Ahmed et al., 2014). Additional experiments demonstrated that Aβ deposits can enhance the spreading of tau pathology (Pooler et al., 2015He et al., 2018). It is important to relate these animal studies to what may be going on in human brain. The present work, which used PET scanning and DTI, is therefore particularly welcome. Tau assemblies propagated through connectivity, not proximity, in an Aβ-dependent manner. The molecular mechanisms underlying tau aggregate propagation and its enhancement by Aβ remain to be elucidated. One must bear in mind that prion-like spreading of tau aggregates is also believed to be important in sporadic tauopathies that lack Aβ deposits, such as Pick’s disease and progressive supranuclear palsy. Does it follow that the spreading of tau pathology is less effective in those diseases?


    . A novel in vivo model of tau propagation with rapid and progressive neurofibrillary tangle pathology: the pattern of spread is determined by connectivity, not proximity. Acta Neuropathol. 2014 May;127(5):667-83. Epub 2014 Feb 16 PubMed.

    . Amyloid accelerates tau propagation and toxicity in a model of early Alzheimer's disease. Acta Neuropathol Commun. 2015 Mar 24;3:14. PubMed.

    . Amyloid-β plaques enhance Alzheimer's brain tau-seeded pathologies by facilitating neuritic plaque tau aggregation. Nat Med. 2018 Jan;24(1):29-38. Epub 2017 Dec 4 PubMed.

  2. Heidi Jacobs and colleagues performed a longitudinal multimodal imaging study to assess the relationships between Aβ (PiB-PET) and tau (flortaucipir PET) pathology, hippocampal volumes (T1-weighted MRI), white-matter tracts (DTI), and cognition (memory and executive-function composite scores) in 256 cognitively normal elderly. Using a stepwise approach, they tested a multifactorial hypothetical model flowing from previous imaging work from this group and others, as well as the basic science literature. Among many analyses presented in this excellent paper, I found two of them particularly interesting.

    In the first analysis, the authors tested whether diffusivity of the hippocampal cingulum bundle (HCB) was associated with accumulation of tau pathology in the posterior cingulate cortex (PCC). The HCB is a white-matter tract that connects the hippocampus with the PCC. Studying this tract in conjunction with the uncinate fasciculus (UF, a WM tract that innervates the medial temporal lobe but not the hippocampus) and tau load in the inferior temporal lobe (a region proximate to the hippocampus but less tightly connected to the HCB) as control conditions is of great interest for inferring spreading mechanisms of tau pathology. The authors found that mean baseline diffusivity was associated with increased PCC tau load over time, and that this relationship was absent for UF diffusivity (as predictor) and changes in inferior temporal cortex flortaucipir uptake (as dependent variable). Stratification by Aβ status revealed that this effect was strongly driven by Aβ-positive individuals. These findings provide further support for the hypothesis that propagation of tau pathology occurs through connectivity, rather than just proximity. The employment of DTI in this study provides convergent information to some recent cross-sectional studies in clinical populations that used the functional architecture (task-free fMRI) as a surrogate for the structural connectome (Hoenig et al., 2018; Cope et al., 2018). 

    In the second analysis, the authors assessed the inter-relationships between HCB diffusivity, PCC tau load, and cognitive function. They showed that baseline HCB diffusivity predicted memory decline, but this only reached significance in subjects with high PCC tau load. Further stratification by Aβ status showed that the Aβ-positive subjects in the sample—again—drove this interaction effect. This emphasizes the importance of Aβ deposition as a potential trigger for downstream effects, while tau pathology might be the actual driver of neurodegeneration and subsequent cognitive decline. These findings suggest that Aβ is an important focus for disease-modifying drugs, but is has to be targeted at very early, presymptomatic stages of Alzheimer’s disease to prevent the pathological chain of events from progressing.

    This study is a terrific example of translational neuroscience, in which cellular/molecular perspectives on mechanistic properties of neurodegenerative diseases were used to investigate these mechanisms on a macroscopic scale in living humans. Some debates at the microscopic level unfortunately cannot be broached by human imaging approaches due to some inherent limitations. For example, the perforant pathway (connecting entorhinal cortex with hippocampus) would have been a logical target for studying mechanisms of tau spreading along WM tracts, but this pathway cannot be captured with the current spatial resolution of DTI. Also, tau pathology in the hippocampus itself cannot be reliably measured using the current generation of tau PET tracers (due to off-target binding in proximate choroid plexus). Having said that, this study provides important insights into the spatiotemporal relationships between presence of Aβ, tau accumulation, alterations in white matter tracts, neurodegeneration and cognitive decline in preclinical AD.


    . Networks of tau distribution in Alzheimer's disease. Brain. 2018 Feb 1;141(2):568-581. PubMed.

    . Tau burden and the functional connectome in Alzheimer's disease and progressive supranuclear palsy. Brain. 2018 Feb 1;141(2):550-567. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Aβ Plaques: Breeding Ground for Toxic Tau?
  2. Tau PET Aligns Spread of Pathology with Alzheimer’s Staging
  3. Brain Imaging Suggests Aβ Unleashes the Deadly Side of Tau
  4. Tau Uses Synaptogyrin-3 to Clump Synaptic Vesicles

Paper Citations

  1. . Propagation of tau pathology in a model of early Alzheimer's disease. Neuron. 2012 Feb 23;73(4):685-97. PubMed.
  2. . A novel in vivo model of tau propagation with rapid and progressive neurofibrillary tangle pathology: the pattern of spread is determined by connectivity, not proximity. Acta Neuropathol. 2014 May;127(5):667-83. Epub 2014 Feb 16 PubMed.
  3. . Amyloid accelerates tau propagation and toxicity in a model of early Alzheimer's disease. Acta Neuropathol Commun. 2015 Mar 24;3:14. PubMed.

Other Citations

  1. Research Timeline

Further Reading

No Available Further Reading

Primary Papers

  1. . Structural tract alterations predict downstream tau accumulation in amyloid-positive older individuals. Nat Neurosci. 2018 Mar;21(3):424-431. Epub 2018 Feb 5 PubMed.