Since it started in 2007 as a one-day powwow of 150 early aficionados in Boston, the annual Human Amyloid Imaging meeting has become a fixture for anyone using imaging technology to study dementia-related pathology in the brain. At the 10th meeting, held January 13-15 in Miami Beach, Florida, 2½ days of talks and ample discussion drew nearly 350 attendees, some spilling into overflow space. This year, the meeting focused overwhelmingly on tau PET imaging and its many challenges. HAI scientists debated whether the current batch of tau ligands are suitable to answer questions such as where, when, and in what form tau aggregates, and how those aggregates affect the normal working of the brain. For Alzheimer’s disease, scientists are beginning to make inroads into this territory. In the case of non-AD tauopathies, the jury is still out, as the ligands bind poorly to the various forms of tau found in these dementias, and specificity remains a concern (see Part 3). HAI is co-organized each year by Keith Johnson, Massachusetts General Hospital, Boston; William Jagust, University of California, Berkeley; and William Klunk and Chet Mathis of the University of Pittsburgh.
Location Matters, But What’s With the Inferior Temporal Lobe?
In their effort to understand how Alzheimer’s develops, scientists are now using multimodal imaging to relate tau aggregation with structural and functional changes. Molly LaPoint, working with Johnson and Reisa Sperling at Brigham and Women’s Hospital, Boston, examined the relationship between accumulating tau in various regions of the brain and thinning of the cortex. LaPoint analyzed data from 99 normal older individuals in the Harvard Aging Brain Study (HABS). She identified 68 regions of interest (ROIs), 34 per hemisphere, for analysis. Then she correlated how much the cortex had shrunk in thickness, measured by two to three MRI scans taken over approximately three years, with how much of the tau tracer AV1451 those 68 areas had taken up in a PET scan done close to the volunteer’s last MRI. LaPoint looked for correlations between AV1451 binding and thinning within the same ROIs, on the one hand, and between tracer uptake specifically in the inferior temporal lobe with global cortical thinning elsewhere, on the other.
In a cross-sectional analysis of the whole cohort, LaPoint found that, for many brain regions, AV1451 uptake correlated with thinner cortices in the same region. The correlation was highest in the temporal lobe, where tangles are known to appear earliest as people age.
Looking at thinning over time, LaPoint found similar relationships among several ROIs mostly in the right hemisphere, including the right superior temporal gyrus, right middle temporal gyrus, and the right temporal pole; all were thinning faster with higher tau.
But it was not just local tau that correlated with a thinning cortex. Though less widespread, a similar pattern emerged between AV1451 uptake in the inferior temporal lobe and thinning in medial and lateral temporal lobe regions. Furthermore, higher AV1451 binding in the inferior temporal lobe also was linked with thinning of the right middle temporal, right fusiform, right anterior cingulate, and right and left parahippocampal gyri. Curiously, for many of these regions, thinning more strongly related to tau in the inferior temporal lobe than to local tau.
Researchers at HAI intensely discussed this asymmetrical pattern, particularly the apparent remote influence of inferior temporal lobe tau. Clifford Jack, Mayo Clinic Rochester, Minnesota, wondered why the correlations were tighter in the right hemisphere. After all, most people in this cohort would not be expected to be on their way to disease with such asymmetry. Sperling noted that everyone entering HABS must have a clinical dementia rating of zero, meaning that candidates with left cortical thinning who performed poorly in memory tests would have been excluded.
Some wondered why tau in the inferior temporal lobe seemed to hold more sway over cortical thickness than did local tau. Jagust asked if tau in the inferior temporal lobe exerts remote effects, or if its presence in this region merely signals that atrophy is afoot elsewhere in the brain. “In our view, it is mostly the latter,” said Johnson. Jagust then wondered if tau in other regions of the brain affects disease. “Many of us felt that, unlike Aβ, tau would help us understand why some regions are vulnerable and others aren’t,” he said, asking, “Can you predict atrophy by knowing where tau is? Or is it not important where tau is?” LaPoint cautioned that these are normal elderly whose pathology was perhaps insufficient to draw out such relationships.
Mark Mintun of AVID Radiopharmaceuticals in Philadelphia noted that in symptomatic stages of the disease, the distribution of tau deposits does match the regional dysfunction predicted by cognitive testing, suggesting that local tau is important. Jagust agreed that this is true in later stages, but said he was more interested in how the disease develops and spreads. “In late disease Aβ is everywhere, but there are focal effects. How does the spread of tau fit in?” he asked. Johnson noted that LaPoint was looking at large ROIs. “We may need to look at tighter distribution and spread of tau to make those correlations,” he said.
Functional Connectivity and Temporal Lobe Tau
Other researchers are studying functional networks to understand the consequences of where tau deposits. Aaron Schultz from Massachusetts General correlated AV1451 uptake with activity in four canonical cortical networks—default mode (DMN), ventral attention (VAN), dorsal attention, and frontoparietal control. Among 103 participants in the HABS, a correlation emerged between inferior temporal AV1451 and connectivity, but only among those 29 subjects who also had a positive Aβ scan as per PiB PET. Specifically, inferior temporal tau came with decreased connectivity in the VAN and DMN networks; the other two networks trended toward reduced connectivity as well. Curiously, tau in the entorhinal cortex—a brain region widely studied for its tau burden in many human and mouse studies—appeared unrelated to connectivity. This again pointed to something unique about the inferior temporal lobe and its ability to influence other areas of the brain.
When Shultz compared connectivity in people with different amounts of amyloid and tau pathologies, he found an inverted-U-shaped response. Low amyloid and tau predicted low connectivity across the VAN and DMN networks; medium levels of amyloid and tau came with high connectivity, and peak levels of Aβ and tau pathology once again came with reduced connectivity. This up-down curve may help explain conflicting reports variously linking amyloid to increased or decreased functional connectivity in preclinical groups, said Schultz.
Researchers at HAI wondered if amyloid contributes to the connectivity changes more than tau. Schultz believes tau pathology plays a key role because, in its absence, the functional connectivity correlations disappear. Gil Rabinovici, University of California, San Francisco, noted that Schultz had measured AV1451 uptake in a node of the brain that lies outside the affected networks, and wondered if tau in other regions might correlate better. Shultz next plans to tease out local disruptions, and how tau spread might affect connectivity.
What Does Tau Have to do with Atrophy, Functional Connectivity?
Possible damage from local tau emerged in functional analyses by Rik Ossenkoppele, University of California, San Francisco. Ossenkoppele reasoned that if tau deposition spread along functional networks, its distribution pattern in a given person should reflect the areas of the brain most affected by that person’s specific clinical form of dementia. For example, patients with posterior cortical atrophy (PCA) tend to have visual impairments, suggesting visual network dysfunction, while people with logopenic variant primary progressive aphasia (lvPPA) struggle to find words and parse sentences, indicating problems in their language centers.
Telltale Troika. In AD dementia, tau covariance (top) based on right middle frontal gyrus atrophy (middle) overlaps with functional connectivity maps of the right executive control network (bottom). [Image courtesy of Rik Ossenkoppele.]
Working with Rabinovici at UCSF and others, Ossenkoppele measured AV1451 uptake in eight people with lvPPA, seven with PCA, five with late-onset AD, and seven with early onset AD. All had brain Aβ as per PiB PET, except one lvPPA patient whose amyloid status was unknown. Ossenkoppele generated maps of tau covariance based on a method originally devised by William Seeley and colleagues (see Apr 2009 news). The idea is that when correlations of tau uptake among different brain regions hold up between different people, they are highly “covariant,” and therefore likely meaningful. The image above, for example, shows that uptake of AV1451 in three voxels varies in the same way, i.e. it covaries, across seven people (S1 to S7). With this voxel-wise approach, Ossenkoppele correlated tau uptake in four seed regions with uptake throughout the brain and across the 21 dementia patients. The starting seeds were in the right middle occipital gyrus, the left superior temporal gyrus, the right middle frontal gyrus, and the left posterior cingulate. He chose the first three because prior work indicated they correspond to areas where the brain shrinks most in PCA, lvPPA, and AD, respectively (see Migliaccio et al., 2009). He chose the left posterior cingulate as a common denominator because it shrinks in people who have any of these diseases.
Ossenkoppele compared how patterns of tau covariance matched up against eight known cognitive networks that had been previously established in 1,000 young adults. They are the higher visual, language, salience, sensorimotor, right and left executive control, and ventral and posterior default mode networks. The overlap was striking and supports the idea that tau deposits spread along functional networks in the brain. For each of the four atrophy seeds, the tau covariance matched the connectivity networks affected by the given disease. PCA, for example, affects the visual network, and this network in turn maps closely to tau covariance with the right middle occipital gyrus, which atrophies most in that disease. Likewise, tau patterns matched networks affected in lvPPA and early onset AD. Tau covariance with the right middle frontal gyrus overlapped with the right executive control network that atrophies in typical AD, while tau covariance with the left superior temporal gyrus seed, which atrophies in lvPPA, overlapped with the language network. Lastly, the left posterior cingulate tau covariance matched the posterior DMN, a region susceptible to Aβ deposition in all forms of AD.
Others praised the approach, and indeed it earned Ossenkoppele the 2016 HAI Young Investigator Award. Sperling wondered if testing covariance based on peak areas of AV1451 rather than atrophy might reveal a different picture. Ossenkoppele said he chose atrophy seeds because he wanted to look at variance by disease phenotype, but also because he wanted to avoid bias. “We preferred to select the seeds from an independent sample of patients and we feel that’s a strength of the study,” though he agreed with Sperling that analyzing covariance based on peak tau uptake would be another way to parse the data, he told Alzforum.
Jack wondered if the data could be explained by local vulnerability to tau pathology. Rabinovici noted that tau covariance with the language and executive control seeds was not contiguous, but jumped between separate areas that are functionally connected. This hints that tau spreads along neuronal projections that make up functional networks. “We need longitudinal analysis to address that question,” he said.
Confused? It gets worse: David Jones, Mayo Clinic, Rochester, Minnesota, cautioned that expression of different genes covaries in these networks, as well, and this may further complicate the interpretation of how tau spreads.—Tom Fagan
- Tau Takes Center Stage at 10th Human Amyloid Imaging Conference
- Network Connections: Missing Links in Neurodegeneration?
- Migliaccio R, Agosta F, Rascovsky K, Karydas A, Bonasera S, Rabinovici GD, Miller BL, Gorno-Tempini ML. Clinical syndromes associated with posterior atrophy: early age at onset AD spectrum. Neurology. 2009 Nov 10;73(19):1571-8. PubMed.
No Available Further Reading