As more research groups are conducting brain imaging in Alzheimer’s disease, separately and on different continents, their data nonetheless converge into a confirmation overall of the disease’s basic progression pathway: plaques, then tangles, then cognitive decline (Perrin and Holtzman, 2009). Large longitudinal imaging studies—and lots and lots of correlation exercises with fluid biomarker and clinical measures—are now adding heft to the idea. Scientists are calculating thresholds of amyloid and tau accumulation associated with progression. They are pinpointing the brain regions involved and parsing how each pathology relates to cognitive decline. At the Human Amyloid Imaging conference, held January 15–17 in Miami, several groups reported that tangles spread beyond the medial temporal lobe only after plaques have pretty much carpeted the cortex, in agreement with prior work.
- Tangles spread widely only after plaques become abundant.
- More plaque and tangles portend clinical worsening.
- Even so, in some people, cognitive problems precede tangle spread.
“The data at HAI reinforce previous models of Alzheimer’s progression predicted by Cliff Jack and others,” noted Michael Devous of Avid Radiopharmaceuticals in Philadelphia (Dec 2011 news). “Now we have quantitative data to support those ideas,” Devous said.
Alas, in AD research, nothing is ever simple and unanimous. Studies disagreed on the exact threshold level of amyloid needed, and some said inflammation needs to interact with amyloid for it to spark tangles. Finally, researchers quantified how strongly plaque and tangles predict cognitive decline. There, too, overall convergence was tempered by the nuance that not all amyloid-positive people with cognitive impairment have a high tangle load, hinting at other factors at work. In part, some seeming discrepancies could be because different groups set thresholds differently, use their own quantification methods, or simply slice and dice large data sets differently for their own analyses.
Widespread Tangles Correlate With Impairment. Among cognitively healthy people with amyloid plaques (left), only a few accumulate tangles outside the medial temporal lobe (darker colors indicate higher percentage of people with pathology); in the cognitively impaired group (right), the percentage is much higher. [Courtesy of Susan Landau.]
Plaques Drive Tangles
Previous imaging studies have shown that the spread of amyloid through the cortex triggers tangles to break out of their previous confinement within the medial temporal region (Mar 2016 news; Aug 2016 news). How much amyloid does it take to unleash tangles? In Miami, Keith Johnson of Massachusetts General Hospital gave an answer. He analyzed a cohort of 131 participants in the Harvard Aging Brain Study (HABS) whose average age was 75. All but eight were cognitively healthy at baseline. Participants were followed longitudinally with PiB amyloid and flortaucipir tau scans for up to four timepoints over a median span of 2.5 years.
A regression analysis of the change in plaque and tangles found that amyloid load needed to reach a threshold of 1.1 to 1.2 SUVR for tau deposition to take off. The higher a person’s baseline amyloid load, the faster both plaques and tangles accumulated. Fast progression of both pathologies correlated with greater cognitive decline on the PACC cognitive composite over two years, in line with previous HABS findings (Jun 2019 news). The PACC is sensitive to change in preclinical populations (Jun 2014 news).
David Knopman and colleagues from the Mayo Clinic in Rochester, Minnesota, independently confirmed that tangles spread only when amyloid is high, but calculated a different threshold value. They stratified 175 cognitively healthy participants in the Mayo Clinic Study of Aging into people who were amyloid-negative, with PiB PET loads below 22 centiloids; those above the threshold, from 22 to 67; and a group whose amyloid load exceeded 67 centiloids. The researchers then examined change over about two years in flortaucipir PET signal in these people’s entorhinal, inferior temporal, and lateral parietal cortex, as well as in a composite of regions typically affected in AD.
At baseline, all participants with amyloid levels above background also had an elevated tau PET signal in most regions. However, only those whose amyloid was above 67 centiloids at baseline accumulated tangles over the next two years, mostly in the inferior temporal cortex and the AD composite regions. Notably, 67 centiloids corresponds to a PiB PET SUVR of 2.0—much higher than the amyloid threshold at which the Harvard group found tangle buildup. Knopman noted that the higher cutoff fits with previous data showing that cognitive impairment occurs later in disease, at high amyloid loads, and closely corresponds to tangle spread.
One implication of Knopman’s finding is that tau PET would not be a practical outcome measure in therapy trials in preclinical AD populations unless participants’ plaque burden was very high. Knopman calculated that if participants’ baseline amyloid burden was below 67 centiloids, a trial would need to enroll 2,000 to 4,000 people to see a 25 percent slowing on tau PET progression. With participants above a cutoff of 67 centiloids, 500 participants would suffice.
Another study took a regional, rather than quantitative, approach to address the question of how much amyloid is needed to kick off tangle spread. Hazal Ozlen and Sylvia Villeneuve of the Douglas Mental Health University Institute in Montréal compared NAV4694 amyloid PET, flortaucipir tau PET, and cognitive performance in 129 participants in the PREVENT-AD cohort. All were cognitively healthy, with a family history of AD. The authors had previously defined seven cortical regions that accumulate plaques early in disease: precuneus, posterior cingulate, rostral anterior cingulate, medial orbitofrontal, rostral middle frontal, superior frontal, and inferior parietal cortex (Villeneuve et al., 2015). In this cohort, 81 participants had no amyloid tracer uptake in these regions, while 28 did in some of the regions—most commonly the precuneus and posterior cingulate—and 20 had uptake in all seven regions.
Compared with the 81 amyloid-negative controls, the 28 people with regional amyloid had an elevated tau signal in entorhinal cortex and middle temporal gyrus. Participants with widespread plaques across all seven regions, on the other hand, also had widespread tangles in the inferior temporal cortex, amygdala, and fusiform, lingual and parahippocampal gyrus. Over up to seven years of follow-up, only the group with widespread amyloid declined on a test of delayed memory in the RBANS. In addition, only the widespread amyloid group reported subjective complaints about their cognition. “Widespread amyloid is needed for tau to get outside of the temporal lobe and cause cognitive impairment,” Villeneuve wrote to Alzforum.
In all of these cognitively healthy cohorts, tangle accumulation precipitated cognitive decline. Another study examined a population that already was cognitively impaired. In cross-sectional data, tangles were not essential for memory impairment. Susan Landau of the University of California, Berkeley, analyzed data from 709 ADNI participants with baseline flortaucipir and either florbetapir or florbetaben amyloid scans. Among them, 315 had positive amyloid scans; 142 of them were cognitively healthy, 173 impaired.
Landau’s work agreed with that of the other groups in that it was primarily the amyloid-positive participants who had accumulated tangles outside the medial temporal lobe. But then the researchers used 1.4 SUVR as the threshold for regional tau positivity, because a previous flortaucipir imaging study had identified that as the best cutoff for discriminating cognitively impaired people from amyloid-negative controls (Maass et al., 2017). And they found that, among the 315 amyloid-positive ADNI participants, the tau signal exceeded this threshold in regions outside the medial temporal lobe in 40 percent of the cognitively healthy, and 60 percent of the impaired (see image above). While low tau signal was frequent in participants who otherwise seemed to be on the AD pathway, the low-tau individuals were clinically indistinguishable from those with high tau.
“People were interested in the fact that there is such a high proportion, up to 40 percent, of amyloid-positive, impaired subjects with low tau,” Landau told Alzforum. “This is likely due to comorbid pathology, but is not fully understood.” In Miami, Julie Schneider of Rush University in Chicago noted that co-morbidities in AD brain include Lewy bodies, TDP-43 deposits, and vascular disease. In some studies, vascular pathology contributes to one-third of dementia cases, Schneider said.
As has been reported in other studies, in Landau’s ADNI analysis, a higher tau PET signal correlated with lower scores on the PACC. The findings reinforce that amyloid pathology drives tangles to spread beyond the medial temporal lobe, and that tangle buildup greater than 1.4 SUVR causes impairment, Landau said.
Other recent ADNI analyses agree with this. In a preprint on medRxiv, researchers led by Rik Ossenkoppele of Vrije University in Amsterdam report that, among 730 non-demented ADNI participants, only those who were amyloid-positive on baseline PET scans accumulated tangles over the next five years. Those who were amyloid-positive by CSF Aβ42, but negative by PET, did not (Reimand et al., 2020). This is in keeping with the idea that CSF changes before PET (Aug 2016 conference news).
How do plaques drive tangles? Perhaps through inflammation. Min Su Kang and Pedro Rosa-Neto of McGill University presented cross-sectional data from 95 participants in McGill’s Translational Biomarkers of Aging and Dementia (TRIAD) cohort. The majority were cognitively healthy, 18 had MCI, 10 Alzheimer’s dementia. All underwent scans with the amyloid tracer NAV4694, tau tracer MK-6240, and microglial activation PET tracer PBR28. Among the 35 amyloid-positive participants, amyloid load and microglial activation synergistically associated with tangles in precuneus, entorhinal cortex, basolateral temporal cortex, and medial frontal cortex, these authors claim, suggesting that these two pathologies together might spur tau pathology (see image above). In turn, tangle load correlated with worse performance on the MOCA, MMSE, and CDR-SB.
Similarly, Paul Edison of Imperial College London reported elevated PBR28 uptake in the medial and lateral temporal lobes of 18 people with AD compared with 16 healthy controls. The PBR28 signal correlated with amyloid PET in the temporal and occipital lobes, and with tau PET in temporal, parietal, and occipital lobes. In 30 people with MCI, on the other hand, the PBR28 signal was no higher than in controls. These findings, too, support a link between amyloidosis, microglial activation, and tangles in AD.
Pathology and Progression
On the question of how plaques and tangles relate to neurodegeneration and progression to dementia, new data from the Australian Imaging Biomarkers and Lifestyle Study of Ageing (AIBL) underscore the finding that high amyloid is required for a person to get worse. In Miami, Chris Rowe of the University of Melbourne, Australia, described a longitudinal analysis of 534 cognitively healthy people. At baseline, 73 percent of them had little amyloid, below 26 centiloids; 10 percent had moderate accumulation, defined as 26 to 50 centiloids; 14 percent were high, at 51 to 100; and 3 percent had more than 100 centiloids.
Over three years, 57 of the 534 participants progressed to MCI. That risk correlated strongly with their baseline amyloid load: People below 26 centiloid did not progress; those with moderate accumulation had a hazard ratio of 3.2; those with high loads, a ratio of 7; and for those with very high loads, risk of progression stood at a whopping 11.4. The steepness of cognitive decline echoed these numbers, with moderate baseline load associated with a slide on the PACC of 0.07 standard deviations per year, high loads with 0.32, and very high loads with 1.38.
Unlike ADNI, in AIBL the picture was similar for people who started the study already impaired. Among 125 people with mild cognitive impairment, 31 percent started out with no amyloid accumulation, 14 percent with moderate, 34 percent with high, and 20 percent with very high accumulation. Sixty people, almost half the cohort, developed dementia over three years. Two-thirds of them had high or very high amyloid loads; one-third, moderate. Those with moderate baseline amyloid declined by 0.1 per year on the PACC, those with more than 50 centiloid by 0.2 per year. “Greater than 50 centiloids of Aβ substantially increases the risk of clinical progression and rate of cognitive decline,” Rowe said.
Tangles, even more than plaques, are closely linked to progression and degeneration. Leonardo Iaccarino of the University of California, San Francisco, described a study of 51 amyloid-positive people with MCI or dementia. In them, flortaucipir binding in medial frontal, anterior temporal, and insular cortex correlated with the neuronal injury marker CSF NfL. Flortaucipir uptake in the left temporoparietal lobe correlated with the inflammation marker CSF YKL40, though only in older participants. NfL correlated with CSF total tau but not with CSF Aβ42. Thus, these markers of injury and inflammation seem to reflect tangles more than plaques, the researchers concluded.
For this reason, tangles as seen by tau PET may make a good progression marker. Lilly’s Devous reported on 717 amyloid-positive people with MCI or dementia. The data were pooled from three in-house flortaucipir studies and a flortaucipir add-on to the solanezumab Expedition 3 trial. First, the researchers dichotomized flortaucipir scans into those that had an advanced tauopathy pattern, with uptake in parietal or frontal regions, and those that did not. The former had nearly double the risk of cognitive decline on the MMSE, ADAS, CDR-SB, and FAQ over the next 18 months.
The upshot was the same when the researchers evaluated tau pathology quantitatively using MUBADA, Lilly’s in-house measure for how extensive tangles are in the brain. They deemed a person’s neocortical flortaucipir signal positive if it was 2.5 standard deviations above the average value seen in healthy young controls. Nearly all MUBADA-positive participants were at Braak stage V or VI, and above a brain-wide cutoff of 1.1 SUVR on flortaucipir scans. Once again, the group defined in this way was most likely to decline on the four cognitive tests.
“You need to have tau pathology beyond the temporal lobe in order to have cognitive progression over 18 months,” Devous concluded.
Braak V or VI means a lot of tangles. Can tau PET pinpoint people earlier on? To find out, the Lilly team derived a second statistical measure, called Etau. It corresponds to tangle accumulation in mesial and inferior temporal regions, roughly Braak stage IV. In this same cohort, only 15 people were positive on Etau but not on MUBADA. And their cognition? They had some decline on the same four tests, midway between that of tau-negative and MUBADA-positive participants, suggesting that Etau might pick out an earlier AD population (Feb 2020 conference news).
That tangles in these regions harm cognition fits with the neuropathology literature and other studies. “We are excited about the ability of tau imaging to inform us of the risk of progression, which is otherwise hard to predict,” Devous told Alzforum. In toto, researchers believe that these kinds of quantitative data build a nuanced understanding of disease progression. This will inform how better to select trial participants, and to detect the effects of therapeutic intervention.—Madolyn Bowman Rogers
- Research Brief: Evidence Supports Model of AD Biomarker Progression
- Tau PET Aligns Spread of Pathology with Alzheimer’s Staging
- Brain Imaging Suggests Aβ Unleashes the Deadly Side of Tau
- Serial PET Nails It: Preclinical AD Means Amyloid, Tau, then Cognitive Decline
- Test Battery Picks Up Cognitive Decline in Normal Populations
- Tau PET: The Field Expands Rapidly
- Perrin RJ, Fagan AM, Holtzman DM. Multimodal techniques for diagnosis and prognosis of Alzheimer's disease. Nature. 2009 Oct 15;461(7266):916-22. PubMed.
- Villeneuve S, Rabinovici GD, Cohn-Sheehy BI, Madison C, Ayakta N, Ghosh PM, La Joie R, Arthur-Bentil SK, Vogel JW, Marks SM, Lehmann M, Rosen HJ, Reed B, Olichney J, Boxer AL, Miller BL, Borys E, Jin LW, Huang EJ, Grinberg LT, DeCarli C, Seeley WW, Jagust W. Existing Pittsburgh Compound-B positron emission tomography thresholds are too high: statistical and pathological evaluation. Brain. 2015 Jul;138(Pt 7):2020-33. Epub 2015 May 6 PubMed.
- Maass A, Landau S, Baker SL, Horng A, Lockhart SN, La Joie R, Rabinovici GD, Jagust WJ, Alzheimer's Disease Neuroimaging Initiative. Comparison of multiple tau-PET measures as biomarkers in aging and Alzheimer's disease. Neuroimage. 2017 Aug 15;157:448-463. Epub 2017 Jun 3 PubMed.
- Reimand J, Collij L, Scheltens P, Bouwman F, Ossenkoppele R. Amyloid-beta CSF/PET discordance vs tau load 5 years later: It takes two to tangle. medRxiv February 3, 2020.