Part 2 of a two-part report. See Part 1.

Even as tau imaging forever changes the way researchers study Alzheimer’s disease and other tauopathies, the first batch of PET tracers are plagued by problems such as noisy data and off-target binding. At the Alzheimer’s Association International Conference 2016, held July 22-28 in Toronto, researchers delved into technical aspects of tracer uptake and how to deal with these vexing problems. Mark Mintun of Eli Lilly’s subsidiary Avid Radiopharmaceuticals, Philadelphia, described a new method for subtracting non-specific background signal by using a different anatomical reference region. This, he said, produced more reliable results. Melissa Murray of the Mayo Clinic in Jacksonville, Florida, tackled non-specific binding by Lilly’s AV1451, still the most commonly used tau tracer; she argued that these signals are spatially distinct from tau tangles and do not cut into AV1451’s usefulness for AD. Other tau tracers are entering human trials; Genentech researchers described preliminary data from their tracer GTP1 that indicated it could detect small changes in tau load. Meanwhile, researchers from Merck presented two posters suggesting their tracer MK-6240 is more specific for neurofibrillary tangles than is AV1451. In addition, MK-6240 may recognize the tau aggregates found in Pick’s disease, which are structurally different from those found in AD. Researchers still do not know exactly what forms of tau aggregate current tracers bind in vivo, noted Christian Sorg of the Technical University of Munich.

“This is a rapidly moving field with a lot of momentum,” Gil Rabinovici at the University of California San Francisco Memory and Aging Center wrote to Alzforum. “There are a number of new tracers just entering human studies, and over the next year or two we will learn a lot about their specific strengths and weaknesses compared to each other.” (See full comment below.) Other companies including Roche and Piramal are developing tau tracers, as well. This spate of tau tracer development was sparked in part because Lilly/Avid is proving difficult in making flortaucipir available for other groups to use.

Noise Control
Data noise creates problems in any study, but particularly in longitudinal PET scans, it can obscure the subtle changes in tracer uptake over time that are the point of the whole exercise. Researchers try to reduce this by picking an anatomical reference region with a low and stable signal for comparison. For both amyloid and tau PET, early studies have typically used the cerebellum, which accumulates less AD pathology than other brain areas. Recently, however, several groups have reported that using white matter as a reference region for amyloid PET lowers scan-to-scan variability and allows researchers to detect small changes over time (see Jul 2014 conference news; Landau et al., 2015). Some researchers speculate that white matter works better because it lies in the same plane as the regions harboring pathology and any “wobble” in the scan data, such as from a slight change in head position, will affect both equally. The cerebellum lies lower in the brain, and may be subject to a different level of noise. 

Finding True Background.

AV1451 pixels in white matter separate into high-intensity (blue curve) and low-intensity (red curve) populations; the latter is thought to represent non-specific binding of tracer to brain. [Courtesy of Sudeepti Southekal and Mark Mintun.]

At AAIC, Mintun described how Avid has adapted white matter as a reference for use with tau PET. The challenge was that, unlike amyloid tracers, tau ligands light up white matter. This is probably not because tau aggregates there, but because the signal from tracer binding to nearby gray matter bleeds into the white matter, Mintun told Alzforum. The resolution of PET scans is too low to discriminate between close signals. To deal with this, the researchers analyzed the intensity of the signal in all the pixels in the white-matter regions. They separated pixels into two populations with high and low intensity, respectively (see image at left). The researchers decided the higher-intensity pixels reflected noise from the gray matter, and the lower-intensity pixels represented true non-specific background. The researchers used these latter pixels as their reference region. “The method identifies a subset of the white-matter pixels that appear to best estimate the reference uptake,” Mintun wrote to Alzforum. Use of this reference region lowered variability in test-retest data. It also heightened the difference in tau signal between cognitively normal and impaired people, resulting in more robust, reliable data, he said. “We believe improved methods for estimating reference region uptake should make PET tau imaging a more powerful biomarker in Alzheimer’s disease research,” Mintun wrote.

Besides their problem with random noise, existing tau tracers tend to bind to things other than tau tangles. Many speakers noted that AV1451 produces non-specific signals in the striatum and brainstem (see Feb 2016 conference news). Murray investigated this off-target binding by comparing autoradiography and immunohistochemistry in postmortem brain slices from 38 people who had a variety of neurodegenerative disorders, including Alzheimer’s, frontotemporal dementia, primary age-related tauopathy (PART), parkinsonism, Lewy body disease, and multiple-system atrophy. AV1451 bound neurofibrillary tangles, which comprise tau with both three- and four-repeat domains, in AD and PART brains, but it poorly labeled pure 3R and 4R tau aggregates present in other tauopathies. In addition, the tracer did not bind less-mature neurofibrillary tangles recognized by the CP13 antibody, she reported. AV1451 binding more closely matched the pattern of staining seen with the PHF-1 antibody, which recognizes mature neurofibrillary tangles. Several researchers told Alzforum that this is a potentially important finding that needs to be independently reproduced.

However, AV1451 lit up many areas that the tau antibodies PHF-1 and CP13 did not. She saw evidence of off-target binding to neuromelanin-containing cells in the substantia nigra, as well as lipofuscin in the lateral geniculate nucleus, hemoglobin in red blood cells, and melanin in the subpial membrane. Other researchers have previously speculated the tracer binds neuromelanin, a pigment found in dopaminergic and noradrenergic neurons. The tracer also bound to regions of calcification, such as in the blood vessels of the choroid plexus, which may explain why signal is often seen there, she suggested. These non-specific signals do not abrogate the tracer’s usefulness, Murray emphasized. Radiologists have to learn what to ignore on scans, she said. In good news for AD research, AV1451 did not label the four-repeat tau deposits found in aging-related tau astrogliopathy (see Lowe et al., 2016). Called ARTAG, this pathology can be widespread in the aging brain, and would have been a significant confounder, Murray said (see Kovacs et al., 2016). 

Some researchers are hoping that other tau tracers in development may be more specific for tangles. In two posters, Merck researchers made the case that their tracer MK-6240 selectively binds the neurofibrillary tangles found in AD. In postmortem studies of AD brains, the tracer bound tangles tightly, while producing no signal in control brains, reported Idriss Bennacef. In non-human primates, the ligand entered the brain rapidly and washed out quickly, and did not light up white matter (see Walji et al., 2016Hostetler et al., 2016). Zhizhen Zeng directly compared MK-6240 binding with AV1451 in slices from frozen AD and control brains. Zeng described similar binding to tangles by both tracers, but less off-target binding with MK-6240. In control brain tissue, clorgyline, a monoamine oxidase inhibitor, competed with AV1451 for binding to control brain tissue, but it did not compete with MK-6240, suggesting the latter has better specificity for tau aggregates.

“The Merck compound looks very interesting because of its unique chemical structure, which is quite different from existing compounds,” Rabinovici wrote to Alzforum. “The very early data presented in normal controls suggests it may have low ‘off-target’ signal. That said, we still need to see more human data, including positive signal in AD patients, to judge whether this will be a useful tracer.”

Bennacef noted the possibility that this tracer may detect forms of tau other than the paired helical filaments present in neurofibrillary tangles. In vitro, MK-6240 also reacts with pure 3R tau, which forms straight filaments, he reported. This form accumulates in Pick’s disease. The researchers are currently evaluating binding in healthy elderly and AD patients in a Phase 1 trial. Biogen recently announced that it will partner with Merck to test this tracer in drug studies.

Other tracers are also entering human studies. Sandra Sanabria Bohorquez of Genentech, South Francisco, presented data on the company’s tau tracer GTP1. In a study of six cognitively normal people, six with prodromal AD, five with mild AD, and five with moderate, tracer binding intensity increased at each disease stage and correlated with the amyloid PET signal, she reported. In preliminary longitudinal data, tau signal did not increase over six to nine months in a healthy control, but rose about 9 percent in a person with mild AD, and 5 percent in one with moderate AD. Accumulation was most notable in the temporal lobe and hippocampus. The results suggest the tracer is sensitive to small changes in tau load, Sanabria Bohorquez noted.

Brian Gordon of Washington University, St. Louis, appreciated seeing the data on the new tracers, but noted more histological and autoradiography work will be needed to understand exactly what each one is binding. “There were similarities between the tau patterns across the different tracers, but they were not exactly the same. While I am optimistic about their ultimate potential, we need to be cautious with these new tracers,” he wrote to Alzforum.—Madolyn Bowman Rogers


  1. At this year’s AAIC we saw early results from many groups using tau PET, especially with the T807/AV1451 and THK-5351 tracers. Many aspects of the data converged across groups and tracers, while some discrepancies highlighted challenges that the field needs to address moving forward.

    Across groups, we saw characteristic patterns of tau tracer uptake in the aging-AD continuum, which, to a large degree, conform to Braak staging of neurofibrillary pathology. Tau deposition in the medial temporal lobe increases with age regardless of the presence of Aβ, but spread of tau into the neocortex seems to be highly dependent on the co-presence of amyloid pathology. It is really this spread of tau that heralds imminent cognitive decline (e.g., as described by Elizabeth Mormino in her presentation). This suggests a synergy between Aβ and tau that leads to neurodegeneration and ultimately dementia. However, Aβ and tau are separated in space and time, and the mechanisms that drive this synergy are not yet understood. Across studies, it is also clear that neurodegeneration and clinical symptoms are much more strongly linked to tau than to Aβ, reproducing in vivo previous observations from autopsy studies.

    As expected with a new technology, there are many methodological aspects of tau PET that are still being worked out and these were debated at the conference—should we dichotomize tau PET in a binary way as positive/negative, as we have done with amyloid markers (as proposed by Cliff Jack), or should we take a more continuous or staged approach (championed by Victor Villemagne)? Should we use global or region-specific measures? What is the threshold at which we can truly be sure that tau PET binding represents true signal over noise? What is the best reference region for intensity normalization?—Mark Mintun’s presentation, for example, suggested that white matter is a better reference region for measuring longitudinal change, as opposed to cerebellar gray matter, which has been largely used thus far in cross-sectional work.

    We are still in early stages of understanding in detail which aspects of tau pathology are truly being targeted by our current ligands. A number of groups presented postmortem correlation studies at the meeting that shed new light on these points. Across studies, it seems clear that AV1451 (where most of the postmortem work has been done) has a higher affinity for PHF-tau composed of a mix of 3R/4R isoforms (as found in AD neurofibrillary tangles, chronic traumatic encephalopathy, and in some genetic tauopathies, including the MAPT R406W mutation presented by Ruben Smith from Oskar Hansson’s group), than for the straight or coiled filaments composed of primarily 4R or 3R tau, as present in non-AD tauopathies (PSP, CBD, AGD, Pick’s disease). However, studies have varied as to whether and to what degree current tracers can detect tau aggregates in these other diseases. Some early data suggest that the tracer may bind preferentially to more mature versus early phosphorylated tau aggregates (Melissa Murray’s presentation), a potentially important finding that needs to be independently reproduced.

    The question of specificity for tau, and the nature of so called “off-target” binding in basal ganglia, midbrain, and choroid plexus remains incompletely understood. This binding has been found in some, but not all, postmortem binding studies, but is omnipresent in vivo with most of the tracers. There are many hypotheses out there: binding to neuromelanin, MAO-A, and calcifications have all been proposed as possible non-tau binding targets, but none of these hypotheses have been proven or fully explain the in vivo results. Ultimately, we will need studies that compare tau PET results obtained during life to tau detected postmortem in the same patients to fully understand these issues.

    Overall, this is a rapidly moving field with a lot of momentum. There are a number of new tracers just entering human studies, and over the next year or two we will learn a lot about their specific strengths and weaknesses compared to each other. The Merck compound looks very interesting because of its unique chemical structure (quite different from the existing compounds), and because the very early data presented in normal controls suggests it may have low “off-target” signal. That said, we still need to see more human data, including positive signal in AD patients, to judge whether this will be a useful tracer.

  2. In general, I think tau imaging data is very promising, but there is a lot we have to learn. Increased Aβ pathology can occur decades before the onset of dementia, while elevations in tau pathology are thought to occur nearer to the onset of cognitive decline. When you think about the selection for clinical trials or perhaps even future prescription of interventional medications, it may be clinically normal individuals with both amyloid as well as tau pathology whom you target. Cerebrospinal fluid assays can measure tau pathology, but PET may be better able to detect early tau pathology in the brain that is focally localized to a few brain regions.

    Before AAIC, the majority of work I had seen had been with the AV1451 tracer from Avid. At AAIC I was able to see results with some new tracers. In particular the THK tracers that GE is licensing, but there were a number of tracers from other companies as well. There were similarities between the tau patterns across the different tracers, but they were not exactly the same. This is also common in tracers that bind to amyloid. For example PiB and AV45 (florbetapir) or any of the other F18 tracers do not show the exact same binding properties. Still, more histological and autoradiography work needs to be done so we better understand whether tau tracers are selectively binding to neurofibrillary tangles, if they are binding to other forms of tau, or if they are even binding to other targets unrelated to tau. While I am optimistic about their ultimate potential, we need to be cautious with these new tracers.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Tau PET Studies Agree—Tangles Follow Amyloid, Precede Atrophy
  2. References and Thresholds—Amyloid Imaging Protocols Debated at AAIC
  3. Shaky Specificity of Tau PET Ligands Stokes Debate at HAI

Antibody Citations

  1. Tau phos Ser202 (CP13)
  2. Tau phos Ser396/Ser404 (PHF-1)

Paper Citations

  1. . Measurement of longitudinal β-amyloid change with 18F-florbetapir PET and standardized uptake value ratios. J Nucl Med. 2015 Apr;56(4):567-74. Epub 2015 Mar 5 PubMed.
  2. . An autoradiographic evaluation of AV-1451 Tau PET in dementia. Acta Neuropathol Commun. 2016 Jun 13;4(1):58. PubMed.
  3. . Aging-related tau astrogliopathy (ARTAG): harmonized evaluation strategy. Acta Neuropathol. 2016 Jan;131(1):87-102. Epub 2015 Dec 10 PubMed.
  4. . Discovery of 6-(Fluoro-(18)F)-3-(1H-pyrrolo[2,3-c]pyridin-1-yl)isoquinolin-5-amine ([(18)F]-MK-6240): A Positron Emission Tomography (PET) Imaging Agent for Quantification of Neurofibrillary Tangles (NFTs). J Med Chem. 2016 May 26;59(10):4778-89. Epub 2016 May 5 PubMed.

Other Citations

  1. Hostetler et al., 2016

External Citations

  1. Phase 1 trial
  2. announced

Further Reading