Shaky Specificity of Tau PET Ligands Stokes Debate at HAI
The emergence of PET ligands for tau has the potential to revolutionize the study of the whole gamut of tauopathies. Researchers are tracking how neurofibrillary tangles appear and spread in real time, yielding invaluable insight into the pathological course of Alzheimer’s disease. For other tauopathies, however, the scans seem open to interpretation. As discussed at the Human Amyloid Imaging meeting, held January 13-15 in Miami Beach, Florida, the problem boils down to two properties biologists constantly bump into: sensitivity, i.e., how little can be detected, and specificity, i.e., lack of off-target binding. At this early point in this rapidly evolving field, it appears that the tau ligands available thus far poorly bind to the particular types of fibril that form in frontotemporal dementia, progressive supranuclear palsy, Pick’s disease, and other non-AD tauopathies. Moreover, the ligands appear to bind to other things as well. Complicating the issue even further, the ligands perform a “now you see me, now you don’t” act that has scientists scratching their heads.
Does It, or Doesn’t It? Maybe It’s the Assays …
The tau ligands’ penchant for prestidigitation was unveiled in Marta Marquié’s talk. Last year, Marquié, who works with Teresa Gomez-Isla at Massachusetts General Hospital, Boston, reported that the PET ligand AV1451 bound to tissue sections from AD brains, but not to tissue taken from people who had had other tauopathies, such as progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), or Pick’s disease (Marquié et al., 2015). AV1451 did, however, appear to visualize tau in vivo in those tauopathies. How could that be? This year at HAI, Marquié addressed this question by correlating in vitro binding directly with in vivo imaging data. She reported a detailed postmortem analysis of tissue samples taken at autopsy from two patients who had non-AD dementia and had had a tau PET scan within eight months of death. “These are the first autopsy-confirmed non-AD tauopathy cases imaged while alive that have been reported,” Gomez-Isla said.
One was a 56-year-old man who carried the P301L tau mutation and had been diagnosed with behavioral variant FTD. He had atypical parkinsonism and had been declining for 2.5 years. PET imaging with AV1451 showed the ligand bound predominantly in his basal ganglia and midbrain. Gomez-Isla noted that in vivo uptake of this tracer in midbrain and basal ganglia has been noticed in a high percentage of cognitively normal older adults who are not expected to harbor tau lesions in those regions. This man had much less AV1451 uptake in the inferior temporal and frontal cortex, where tau deposits in bvFTD. In contrast, his autopsy revealed tau-positive grains not only in the basal ganglia and white matter, but also extending into the cortex.
What was the ligand binding to? Marquié used phosphor screen and nuclear emulsion autoradiography to compare AV1451 binding to tissue slices with tau staining by immunohistochemistry. In short, the in vivo and in vitro signals only partially correlated. Marquié found no AV1451 binding in slices from any brain region except the entorhinal cortex and the substantia nigra. The latter matched with the live scan, but the former did not. Age- rather than disease-related tau deposits probably explain the entorhinal cortex binding, said Marquié. She noted this binding could serve as an internal positive control. Neuromelanin, which AV1451 is purported to bind, may account for some of the signal in the substantia nigra. Marquié interpreted the results to mean that while AV1451 binds to the paired helical fragments of tau found in Alzheimer’s disease, it does not bind the straight tau filaments known to occur in bvFTD and other non-AD tauopathies with an affinity that could explain in vivo signals.
Analysis of tissue from the second patient supported this idea. This 68-year-old man had a neuropathologically confirmed diagnosis of progressive supranuclear palsy (PSP). Similar to the bvFTD case, Marquié detected no lifetime AV1451 PET binding to any of the regions suspected of having high tau pathology other than the basal ganglia and midbrain. Immunohistochemical analysis of sections of his brain tissue, using antibodies that recognize tau pathology in PSP, detected abundant tau pathology in the brain stem, cerebellum, basal ganglia, midbrain, and numerous other regions. Once again, autoradiographs of brain slices showed AV1451 binding to only the entorhinal cortex and substantia nigra.
“Our autoradiography studies have demonstrated that the nearly universal midbrain in vivo signal seen in older adults, regardless of the presence or absence of tau pathology, is very likely due to off-target binding of this tracer to neuromelanin-containing neurons in the substantia nigra,” Gomez-Isla wrote to Alzforum. She does not know what gives rise to uptake of AV1451 in the basal ganglia in vivo, and did not detect any binding in this area by autoradiography or by in vitro binding assays. “That’s what makes us think that it could be, at least in part, a non-specific in vivo uptake related to technical or biological factors other than tau pathology itself,” she wrote.
Milos Ikonomovic, working with HAI co-organizer William Klunk at the University of Pittsburgh and collaborators, drew similar conclusions about AV1451 from his ex vivo binding studies. Ikonomovic compared ligand binding in brain tissue from three people who had had PSP, two who had had CBD, and one with Pick’s disease; he used two AD cases with confirmed tau pathology as positive controls. He also had brain slices from three cases of FTLD with TDP-43 inclusions—a tau-negative form of dementia. Both tritiated and F18-labelled AV1451 bound strongly to frontal cortex in AD samples but weakly or below control levels in frontal cortex from the other tauopathies, even though immunohistochemistry and ELISAs detected much more tau pathology in Pick’s disease and CBD than in PSP and FTLD-TDP. “We believe that the tau aggregates that form in non-AD tauopathies are distinct from the paired helical fibrils that form in AD,” said Ikonomovic. He added that this could explain why tau ligands poorly bind in non-AD tauopathies
These presentations prompted a searching discussion of whether the in vivo PET imaging or the autoradiography and other in vitro binding assays were more likely to be correct. Interestingly, Marquié found that in vivo AV1451 uptake in the pallidum, a part of the basal ganglia, correlated with tau levels there as measured by semi-denaturing detergent acrylamide gel electrophoresis. SDD-PAGE is gentle enough to leave amyloid-like polymers intact. “The correlation of AV1451 signal with tau levels measured by SDD-PAGE in pallidum in this case matches our in vivo findings,” noted Gil Rabinovici, University of California, San Francisco. “This highlights the globus pallidus as the area of highest signal, and supports the notion that this signal may be related to tau,” he said.
Attendees spent considerable time debating why the autoradiographs are turning up negative. “It is important to recognize that these assays are sensitive to experimental conditions,” said Rabinovici. In another presentation, Giorgio Attardo from Avid Radiopharmaceuticals, the supplier of AV1451, noted that autoradiograph protocols usually include a 70 percent ethanol wash, which may be too harsh. Attardo used signal matching techniques to correlate AV1451 binding with AT-8 tau antibody immunoreactivity. This technique hinges on using the identical tissue slice for both autoradiography and immunohistochemistry. Attardo found that the two signals overlapped well in brain tissue from AD patients, with area correlations ranging from 88 to 92 percent, while this correlation was at best 26 percent in tissue from PSP patients and 63 percent in tissue from Pick’s disease patients. Leaving out the ethanol wash led to muddled autoradiogram tests of normal tissue sections, Attardo said, but reducing the ethanol to 30 percent revealed a second binding site that could be blocked by non-radiolabeled AV1451. “It may be that a relatively low affinity binding of AV1451 to tau aggregates in PSP is washed out during tissue preparation,” said Rabinovici.
Klunk noted that the harsh washing conditions were needed because AV1451 is such a sticky compound. “It could be, then, that all we can expect to detect in vitro are the tangles in AD,” he said. That is because AV1451’s affinity for those tangles is much higher. Klunk added that the field remains challenged to figure out if tracers are being artifactually removed during in vitro biding assays, or if the signal in vivo is spurious. “I think our current assays and current compounds don’t allow us to fully answer these questions,” he told Alzforum.
Others suggested measuring the specificity of these ligands by running a competition assay in vivo with unlabeled ligand. That led to nods of support, but some quick mental math by HAI co-organizer Chet Mathis, University of Pittsburgh, squashed the idea, for AV1451 at least. “The dose is the problem,” Mathis said. “We would need several hundred milligrams of cold AV1451, and I doubt the toxicity and solubility would allow it.” Mathis suggested testing this in animal models. Perhaps such a “cold” competition assay might help identify non-tau binding sites? In his presentation, David Jones, Mayo Clinic, Rochester, Minnesota, reported that people with a form of FTD caused by progranulin mutations bind AV1451 in the brain. As Reisa Sperling of MGH noted, those patients do not have tau deposits on autopsy (see Part 4 of this series). One idea is that AV1451 might bind monoamine oxidase and, in fact, Attardo found that the MAO inhibitor clorgiline blocked AV1451 binding in slices of striatum. Clorgiline or isocarboxazid, another MAO inhibitor, might be safely given in sufficient doses to block AV1451, Attardo said.
Finally, some researchers raised the tantalizing possibility that AV1451 and other tau ligands might be binding something altogether different. Even if that something is not tau, it could be informative. Almost no off-target binding of AV1451 occurs in 30-year-olds, for example, implying that the target is not something normally abundant in the brain. “That, plus Jones’ data on FTD, suggests to me that AV1451 might bind some other marker of neurodegeneration,” said Sperling.—Tom Fagan
- Marquié M, Normandin MD, Vanderburg CR, Costantino IM, Bien EA, Rycyna LG, Klunk WE, Mathis CA, Ikonomovic MD, Debnath ML, Vasdev N, Dickerson BC, Gomperts SN, Growdon JH, Johnson KA, Frosch MP, Hyman BT, Gómez-Isla T. Validating novel tau positron emission tomography tracer [F-18]-AV-1451 (T807) on postmortem brain tissue. Ann Neurol. 2015 Nov;78(5):787-800. Epub 2015 Sep 25 PubMed.
No Available Further Reading
No Available Comments
Make a Comment
To make a comment you must login or register.