Local replication of tau seeds—not their spread from region to region in the brain—drives tau accumulation in Alzheimer’s disease. That is the provocative conclusion of a chemical kinetics study published October 29 in Science Advances. Researchers led by Tuomas Knowles and David Klenerman, both of the University of Cambridge, U.K., and Bradley Hyman of Massachusetts General Hospital, Boston, applied kinetic modeling of tau aggregation and spread to multiple sources of experimental data, ranging from the measurement of tau seeds extracted from human brain samples to tau PET imaging in living people.

  • Kinetic model finds that local replication of tau seeds sets the pace of tau accumulation in AD.
  • Spreading plays a negligible role.
  • Past Braak III, tau aggregates double every five years in neocortex.
  • In mice overexpressing tau, aggregates double every two weeks.

The scientists found that by Braak Stage III and beyond, spreading between regions accounts for a negligible amount of tau accumulation. Instead, local replication of tau aggregates within each region appears to be the rate-limiting factor. Furthermore, they report that, on average, it takes five years for an aggregate of tau in the neocortex to double. Notably, the findings suggest that by Braak Stage III, aggregation-ready tau must already be widely distributed throughout the neocortex, rather than just beginning its prion-like tear from the entorhinal cortex.

“The study fits with the idea that each neuron, or local set of neurons, has the ability to cause more replication of tau,” said Dennis Selkoe of Brigham and Women’s Hospital in Boston. “This analysis calls into question the emphasis in our field on the concept of neuron-to-neuron passage,” he added. Selkoe and others have expressed skepticism about the prominence of proteopathic propagation in neurodegenerative disease research (Apr 2016 Alzforum webinar). 

“This is an interesting and thoughtful modeling approach to understanding the temporal and spatial characteristics of tauopathy in the Alzheimeric brain,” commented Lary Walker of Emory University in Atlanta. “One translational implication of the model is that inhibiting the formation of tau seeds, rather than blocking spread, is the more promising means of impeding tauopathy,” he added.

Still, Selkoe and other commentators noted limitations to the new model. For one, they ask whether it is valid to use different measurements of tau accumulation—such as tau seeding capacity versus neuropathology versus tau PET tracer uptake—within the same kinetic model.

Knowles told Alzforum that the study was motivated by a simple question: What is the primary driver of tau’s aggregation throughout the brain in AD? To answer it, first author Georg Meisl and colleagues first derived a mathematical model of tau accumulation. It accounts for the rate of tau aggregation as well as spreading throughout the brain. Importantly, the model was agnostic to mechanisms of spread, and was limited to the analysis of spread between brain regions, rather than between cells within a given region.

How Does Tau Accumulate? In the concept of spreading (top left), seeds would move from one region to another. Replication involves local growth and multiplication of seeds. [Courtesy of Meisl et al., Science Advances, 2021.]

Equipped with a mathematical equation that took both replication and spreading into account, the researchers set out to answer their first question: Is the accumulation of tau aggregates in the neocortex limited by the rate of tau spreading, or replication? To answer it, they plugged different kinds of experimental data into their model. They used data from three studies that had quantified tau seeding activity in different brain regions in people who had died in different stages of AD (DeVos et al., 2018; Furman et al., 2017; Kaufman et al., 2018). Other postmortem studies quantified tau accumulation via immunohistochemistry (Gómez-Isla et al., 1997; Qian et al., 2017). Finally, the researchers included data from tau PET studies, which infer tangle accumulation via tracer uptake (Sanchez et al., 2021). 

Data from all these studies suggested that the distribution of tau aggregates fundamentally changes at Braak Stage III, when aggregates start being detected outside of the medial temporal lobe and in the neocortex. Because clinical symptoms also begin at this stage, the researchers designated Braak stage III as “T=0” in their equation, and modeled tau kinetics from this stage onward.

They found that the experimental data fit tightly with the kinetic model only when the system is replication-limited. In other words, tau's progression throughout the brain can be explained mostly by the replication rate of tau seeds within each region, not by the rate tau seeds spread between regions.

What Curbs Tau? Using experimental data, scientists modeled which process dictates the accumulation of tau pathology in human brain. In one scenario, spread limits the accumulation of tau (left). In another, replication dictates the pace of accumulation (right). [Courtesy of Meisl et al., Science Advances, 2021.]

Next, the scientists asked their model to calculate the doubling time of tau aggregates in different brain regions from Braak Stage III onward. Things were slow. On average, tau seeds doubled once every five years. This varied depending on which experimental dataset was used, ranging from three to nine years. However, regardless of how tau was measured—by seeding assay, immunocytochemistry, or PET—the model consistently found that tau aggregates replicate on the order of several years.

“While the proposed models include a number of assumptions, the consistency of the results across different datasets is compelling,” commented Gil Rabinovici of the University of California, San Francisco (full comment below).

If it takes so long for one tau seed to become two tau seeds, then how does the neocortex manage to fill with tangles as AD advances? The only explanation for this, the researchers claim, is that by Braak Stage III, tau seeds are already widely distributed throughout the neocortex, where they continue to form and replicate locally as disease worsens. How those aggregates get there in the first place—whether by spreading that took place prior to Braak Stage III, and/or by de novo aggregation in each region—this study could not address.

Meisl also put the model to use in P301S-tau transgenic mice, which overexpress aggregation-prone human tau throughout the brain. Lo and behold, the tau seed doubling time was only two weeks—lightning-fast compared to that observed in the human brain.

This implies that the human brain may be better equipped than the mouse brain to ward off tau aggregation. In support of this idea, the authors found that P301S tau replicates nearly as fast in the mouse brain as it does in a test tube, whereas in the human brain, wild-type tau replicates two orders of magnitude more slowly than it does in vitro.

The findings demonstrate how both in vitro and mouse models come up short as models of tauopathy, Selkoe said.

“The model indicates that the human brain may be relatively adept at slowing the growth and replication of tau seeds; understanding how it does so could disclose new therapeutic objectives for Alzheimer's and possibly other tauopathies as well,” commented Walker. Klenerman made a similar point, noting that improving upon the brain’s inherent anti-aggregation activity by just a little bit could have a strong impact on slowing tauopathy progression.

The trouble is, most of the tau-targeted therapies in development take aim at tau spreading, Knowles noted. “Our analysis shows that there are fundamental reasons why spreading is not a good target,” he said. “By Braak Stage III, spreading is effectively irrelevant from an intervention point of view.”

Rabinovici agreed that, if correct, the findings could have major implications for drug development.

Selkoe believes therapeutics that take aim at intracellular tau, such as anti-tau “intrabodies” or antisense oligonucleotides that douse tau expression, might be best suited to target this process. Knowles and Klenerman added that therapies aimed at beefing up the neuronal anti-aggregation machinery, such as supporting chaperones or enhancing protein degradation, could also do the trick.

To the mind of Tharick Pascoal, University of Pittsburg, Pennsylvania, the results call for further study of what exactly catalyzes local tau seeding in Braak III. This could be amyloid—and other mechanisms, too (full comment below).—Jessica Shugart


  1. This is an interesting and thoughtful modeling approach to understanding the temporal and spatial characteristics of tauopathy in the Alzheimeric brain. The findings indicate that both the replication and spreading of tau seeds contribute to the progression of tauopathy, but that the local replication of the seeds prevails after Braak Stage III. It seems that, as tauopathy passes through Stage III, long-distance spreading becomes increasingly secondary because tau seeds are already widely distributed in the brain. In addition, the burgeoning population of sick and dying neurons might release seeds locally to be taken up and replicated by other cells, while at the same time losing their ability to transport seeds to interconnected sites.

    In Alzheimer's disease, it is not possible to fully disentangle Aβ and tau; the authors speculate that the transition that occurs around Braak Stage III might be related to the rapid development of Aβ plaques at this time. In this light, it could be informative to determine how toxic oligomeric forms of both tau and Aβ—which are not evident histologically or via in vivo imaging—fit into the scheme. Might a shift in the molecular diversity of tau assemblies play into the Stage III transition? The prion protein, for instance, has been reported to constitute both infectious (seeding-competent) and toxic varieties that have different temporal and pathobiologic properties (Sandberg et al., 2011). 

    One translational implication of the model is that inhibiting the formation of tau seeds, rather than blocking spread, is the more promising means of impeding tauopathy. In addition, the model indicates that the human brain may be relatively adept at slowing the growth and replication of tau seeds; understanding how it does so could disclose new therapeutic objectives for Alzheimer's and possibly other tauopathies as well.


    . Prion propagation and toxicity in vivo occur in two distinct mechanistic phases. Nature. 2011 Feb 24;470(7335):540-2. PubMed.

  2. This study represents an innovative, multidisciplinary effort to address a key question in the field: What are the most important mechanisms that drive the progression of tau pathology, a harbinger to neurodegeneration and cognitive decline in Alzheimer’s disease? By applying chemical modeling to an impressively diverse collection of datasets (ranging from seeding assays to quantitative neuropathology to tau PET), the investigators conclude that, at least at the symptomatic stage of AD (Braak Stage greater than III), local replication rather than distal spread drives increases in neocortical tau burden.

    While the proposed models include a number of assumptions, the consistency of the results across different datasets is compelling.

    If correct, the findings may have significant implications for drug development, guiding us to target the elements of tau biology that are most likely to impact disease progression. Even though clinical trials of tau-targeting therapies are well underway, this study highlights how little we actually know about the specific molecular mechanisms that drive tau aggregation and spread in AD.

    One wonders also if the kinetics of tau spread will be similar in other tauopathies, such as PSP, CBD, CTE, Pick’s, etc., or if they will vary depending on the specific structure and biochemistry of tau aggregates and the connectivity and susceptibility of the brain regions afflicted in each disease.

    Overall I found this to be a thought-provoking paper that raises many important questions for the field to address in future work.

  3. Meisl and colleagues showed results suggesting that tau seed accumulation distributes from Braak III to neocortex, preferentially via local replication rather than spreading. This is an interesting paper, in which the authors were able to organize complex mathematical concepts, experimental designs, and models into an elegant framework to describe tau seeds accumulation. Previous studies have proposed models to describe tau propagation, but I think Meisl's paper did a particularly good job of justifying mathematical assumptions with plausible biological inferences and some supporting experimental data.

    The authors’ results invite further studies. Their results showing that tau seeds distribute faster after Braak III are in line with previous postmortem and in vivo studies suggesting Braak III as a milestone for the development of AD dementia. Interestingly, they also found tau seeds in low concentrations in neocortical regions, even before Braak III. These results together suggest that further studies could be designed to elucidate the mechanistic underpinnings associated with the catalysis of existing tau seeding in Braak III, including, but not limited to, the role of amyloid.

    I would like to see more experiments supporting a comparison between replication and spreading of tau seeds, which, in my opinion, was heavily influenced by mathematical assumptions such as the chosen models.

    In summary, this interesting study adds new elements and raises new questions to advance our understanding of tau propagation in AD.

Make a Comment

To make a comment you must login or register.


Webinar Citations

  1. Webinar: Pathogenic Protein Spread? Let’s Think Again

Paper Citations

  1. . Synaptic Tau Seeding Precedes Tau Pathology in Human Alzheimer's Disease Brain. Front Neurosci. 2018;12:267. Epub 2018 Apr 24 PubMed.
  2. . Widespread tau seeding activity at early Braak stages. Acta Neuropathol. 2017 Jan;133(1):91-100. Epub 2016 Nov 22 PubMed.
  3. . Tau seeding activity begins in the transentorhinal/entorhinal regions and anticipates phospho-tau pathology in Alzheimer's disease and PART. Acta Neuropathol. 2018 Jul;136(1):57-67. Epub 2018 May 11 PubMed.
  4. . Neuronal loss correlates with but exceeds neurofibrillary tangles in Alzheimer's disease. Ann Neurol. 1997 Jan;41(1):17-24. PubMed.
  5. . Neurofibrillary Tangle Stage and the Rate of Progression of Alzheimer Symptoms: Modeling Using an Autopsy Cohort and Application to Clinical Trial Design. JAMA Neurol. 2017 May 1;74(5):540-548. PubMed.
  6. . The cortical origin and initial spread of medial temporal tauopathy in Alzheimer's disease assessed with positron emission tomography. Sci Transl Med. 2021 Jan 20;13(577) PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . In vivo rate-determining steps of tau seed accumulation in Alzheimer's disease. Sci Adv. 2021 Oct 29;7(44):eabh1448. PubMed.