This is Part 3 of a seven-part series. See also Part 1, Part 2, Part 4, Part 5, Part 6, and Part 7. View a PDF of the entire series.
1 February 2011. At the 5th annual Human Amyloid Imaging conference held in Miami, Florida, on 14-15 January 2011, one of the hottest topics was what’s beginning to emerge from multimodal imaging studies that are aiming for a more comprehensive view of what happens in the brain once amyloid deposits there. To be sure, it’s early days for that; hence, some results cut opposite ways, data can be provocative but numbers small, and methods not fully established, much less standardized. But it’s where the bleeding edge of amyloid research is at right now, so here for your thought and inspiration are brand-new slices of data and discussion.
Stefan Foerster of the Technical University in Munich, Germany, made a preliminary case that amyloid jams neuronal networks. He noted previous work showing that the anatomical pattern of amyloid deposition as measured by PIB-PET, and hypometabolism as measured by FDG-PET, partly overlap both with each other and with the default-mode network of functional connectivity. In a new collaborative study, 20 patients with very mild AD and 15 controls, all in their sixties, underwent a PIB and a FDG-PET scan at baseline and again 27 months later. The scientists analyzed these images using a method called statistical parametric mapping, by which they can compare digitized images voxel by voxel in a fashion unbiased by previous assumptions. What did the scientists find? The baseline amyloid pattern and the follow-up hypometabolism pattern were the most similar. At baseline the amyloid was broad and did not expand in time; the hypometabolism pattern was smaller at baseline but spread over the next 27 months to then overlap very well with baseline amyloid deposition, Foerster said. In essence, this would imply that hypometabolism of neurons follows amyloid deposition in their area with some delay in time.
To get a sense of which networks the most amyloid-prone brain areas normally engage, the researchers took the regions with peak levels of PIB retention in their AD patients at baseline as "seeds" and calculated their functional connectivity networks in a separate dataset of 27 volunteers in their twenties who had undergone resting-state fMRI scans. Intriguingly, Foerster said, the networks normally subserving those brain areas most vulnerable to amyloid deposition overlapped even better with the actual areas of FDG-PET hypometabolism in the AD patients at follow-up than did the default-mode network. Overall, this points in the direction that amyloid deposits in a given brain region disrupt neurons locally and also disrupt its functional connectivity network, and some years later this shows up as hypometabolism, i.e., neuronal dysfunction in those areas. “These findings support the amyloid hypothesis,” Foerster said.
On a poster, Liang Wang of Massachusetts General Hospital, Boston, showed roughly similar trends in a study of 42 cognitively normal older volunteers who had resting-state connectivity MR scans and PIB-PET scans. Wang used different methods in a different study design. But similar to Foerster, he also saw that when a given region’s amyloid burden was high, intrinsic activity tended to be incoherent and break down both locally and long range across nodes of the default network.
Foerster’s talk prompted both compliments and caveats. For example, Cliff Jack of the Mayo Clinic in Rochester, Minnesota, whose colleagues are working on this as well, cautioned that the temporal ordering of "amyloid-then-dysfunction" does not hold true across the brain. “Some brain areas have amyloid deposition early but do not become dysfunctional until late in the disease. Only some areas are connected in this way, where early amyloid deposition correlates soon after with hypometabolic networks,” Jack said. Foerster readily acknowledged that he was putting out a provocative concept to stimulate discussion and further study.
Consider another study on what might lie between amyloid deposition and cognition. Patrizia Vannini of Brigham and Women’s Hospital in Boston won the 2011 HAI Young Investigator Award for her talk about what amyloid does to one particular brain area indispensable for episodic memory that lies smack in the default-mode network. The posteromedial cortex drew Vannini’s attention because it is not only prone to early amyloid deposition, but also figures in both encoding and retrieving memories. In what’s called the "encoding/retrieval flip," this area suppresses its default-mode firing during memory encoding and activates during cued retrieval. Perhaps, the BWH researchers reasoned, this area is so prone to amyloid deposition exactly because it is constantly active in these ways, that is, produces a lot of Aβ? This would jibe with a discovery by Randy Buckner of MGH that something about the activity and the metabolism of connectivity hubs is conducive to amyloid deposition (Buckner et al., 2009; see also Hedden et al., 2009), and fit with the finding that neural activity drives up Aβ secretion (Cirrito et al., 2005).
Vannini tested this hunch with an associative memory task for which 26 young and 41 old volunteers—half of them with, half without brain amyloid—lay in the scanner and tried to memorize which name went with which face flashed on a screen. Later, they were tested on how well they did. Initially, Vannini found that old people were less able than young people to "flip this switch" in brain activity. A bit like gymnasts who lose flexibility, their posteromedial flips between deactivation and activation happened within a smaller dynamic range. When she split the old volunteers by whether they had brain amyloid, she found that brain amyloid further limited this ability to modulate. Furthermore, people who modulated, i.e., "flipped," well tended to get the test right. This would mean that being able to mount a nimble, dynamic response in the posteromedial cortex is important for memory, and both age and amyloid get in the way of that.
Other scientists, too, are probing how brain activation during memory encoding changes as people age, and what, if anything, amyloid has to do with that. Elizabeth Mormino of the University of California, Berkeley, first pointed to discrepancies between amyloid and cognitive status from person to person. How can it be that some older people with high amyloid function better than others? Mormino investigated a hunch that the former manage to call on the prefrontal cortex to compensate for the presence of amyloid. She studied 37 people in their seventies and 15 in their twenties in a more complex recognition test in which volunteers see landscape pictures while lying in the scanner and later, outside the scanner, get tested on whether they had actually encoded the picture. In this way, Mormino and her colleagues visualized the cortical areas that were active while volunteers encoded what they saw. Roughly similar to Vannini, Mormino also saw greater activation in young than in old subjects, and a trend for further reduction with amyloid.
One result stood out. People with brain amyloid activated their prefrontal cortex significantly more than those without. “This was a striking pattern,” Mormino said, and added, “It is possible that the integrity of the prefrontal cortex enables these subjects to remain in the normal range despite having amyloid.” Perhaps people whose prefrontal cortex escapes other insults that can accompany aging—vascular disease, for example—withstand the degrading effects of amyloid longer than do their fellow elders with a less preserved prefrontal cortex. This, then, could be one potential explanation for how some people "live in peace with their amyloid" for some years.
It sounded appealing until the next two speakers presented what appeared to be the opposite result. Kristen Kennedy of the University of Texas at Dallas reported on an fMRI/amyloid PET study of 137 adults aged 30 to 89 who had an AV-45 (aka florbetapir/Amyvid) scan and performed essentially the same memory encoding task Mormino’s group uses. Here, older people with high amyloid activated their prefrontal cortex less during encoding than did amyloid-free age-matched fellow volunteers. In Kennedy’s hands, people with brain amyloid performed worse, and show a twin pattern of reduced prefrontal activation and failure to suppress the default-mode network as has been shown previously (Hedden et al., 2009).
Michael Devous, also of the UT group in Dallas, presented functional connectivity MR data available to date from the Dallas Lifespan Brain Study on the default-mode network, which is active at rest, and the salience network, one of many networks active during focused tasks. Overall, his work indicates that when amyloid deposits in the brains of cognitively normal people, connectivity in the default-mode network falls apart, whereas in the salience network, it shifts, such that some areas increase abnormally. In particular, Devous noted seeing a steady destruction of the prefrontal cortex components of the default-mode network in people with high brain amyloid. MGH’s Wang, too, had noted a disruption of connectivity in the prefrontal cortex in cognitively normal people with more amyloid.
Finally, a decrease in prefrontal activity in normal people with brain amyloid also showed up in the work of Prashanthi Vemuri of the Mayo Clinic in Rochester, Minnesota. Vemuri studies how functional connectivity changes along the spectrum from cognitively normal to overt AD. Vemuri’s group did PIB-PET imaging and resting-state fMRI in some 250 folks who participate in the longitudinal Mayo Clinic Study of Aging. She grouped them as cognitively normal with or without amyloid, MCI with amyloid, and AD. She excluded people with MCI but without amyloid, “because we wanted a pure group on the AD path.”
Viewed globally, Vemuri saw functional connectivity in cognitively normal people with PIB go up, possibly reflecting that the brain reorganizes its functional networks to compensate for the presence of amyloid pathology. In people who were progressively further along the path to AD, functional connectivity was lower. Vemuri analyzed the brain’s "connectome," i.e., multiple functional networks, to look for changes within and among 25 major networks that she had previously identified in volunteers without amyloid. In this way, she saw functional rearrangements at the early stage (i.e., disconnection of the frontal and temporal-parietal subsystems) and consistent decreases within and across networks at the later stages.
These data are complex to interpret, though in Vemuri’s mind it still boils down to some simple messages. For one, the results in cognitively normal PIB-positive people reflect what amyloid alone does to connectivity, whereas the results on the MCI/PIB-positives and AD groups reflect what both amyloid and neurofibrillary pathology do together. For another, early on, while only amyloid pathology is at play, the two major subsystems affected—the frontal and the temporal/parietal cortex—react differently, with decreased connections with each other and increased connections within each subsystem. Once both pathologies are there, connectivity decreases overall consistently.
What gives? In discussion, scientists challenged each other as to what this data amount to when one area of focus, the prefrontal cortex, goes up in one study and down in another using the same fMRI paradigm. The studies were all different in how they judged a volunteer to be amyloid positive, which amyloid tracer was used, what sorts of people were enrolled, how ApoE might affect the outcomes, how the fMRI data were analyzed, and where researchers placed the seed region to visualize various networks. Moreover, noise, non-specific binding, and atrophy effects can blur data on small effects, cautioned Bill Jagust of UC Berkeley. “Connectivity networks are complicated to study. Given how nascent this field is, our interpretation must be cautious at this point,” Jagust told ARF. For this reason, scientists agreed, the field would do well to continue for some time longer with exploratory studies to let the technical issues play out, and not try to standardize methods just yet. (In contrast, standardization is beginning in MRI and well underway in CSF methodology for AD research.)
There was some disagreement about why the cognitive effects in these studies are so subtle. Some researchers found it “shocking” that the amyloid effect on cognitive measures in these studies is small. Others replied that they are small by definition because the studies enroll cognitively normal people. “Don’t forget that most patients with large effects have AD and amyloid,” said Reisa Sperling of Brigham and Women’s Hospital in Boston. The mystery, all agreed, lies in why some people can stay cognitively normal with amyloid in their brains for years.
Scientists also discussed how they could move from correlation to cause. “Right now, we have associations among amyloid, age, education, cognition, activity. How do we design studies so we can turn these associations into mechanistic relationships?” asked Steve Arnold of the University of Pennsylvania in Philadelphia. The answer, others said, lies in multimodal longitudinal studies such as ADNI, AIBL, DIAN, and increasingly others as longitudinal aging studies are bulking up on neuroimaging.
While multimodal imaging in AD research is young, the larger field of aging research has been transformed by the advent of amyloid PET. “Human amyloid imaging helps us parse which parts of aging are due to amyloid and which are due to other age-related insults,” said Sperling. One example for this was a functional imaging poster by Trey Hedden and colleagues at MGH. It showed that a significant fraction of older cognitively normal people was unable to dynamically modulate brain activity such that they would devote more attention to increasingly difficult tasks, and it ascribed this executive deficit to white matter hyperintensities, not to amyloid.
The overall change to the field has been “momentous,” said Randy Buckner of MGH in a keynote address, particularly when amyloid imaging is viewed together with other biomarkers. Collectively, these markers now enable scientists to view the progression of the disease in living people, Buckner said, and they have made explicit the staging model for the disease that was forming in people’s minds (see Jack et al., 2010; Perrin et al., 2009). “From this emerges our current focus on using these markers in asymptomatic individuals,” Buckner said. As this research advances, the staging model may shift, Buckner noted. Amyloid will stay earlier than atrophy, but the cognition curve may move to the left as scientists get better at detecting small impairments.
For a panoramic view, Bucker said he is inspired by a conceptual framework whereby metabolic conditions in early life set the stage for amyloid deposition in midlife, which itself precedes atrophy and then dementia in late life. These metabolic conditions are not yet understood, but aspects of glucose metabolism and mitochondrial function likely play a role (e.g., Valla et al., 2010; Vlassenko et al., 2010). By this framework, genetic risk factors such as ApoE and perhaps Tomm40 affect not so much the neurodegenerative phase late in life, which is the context in which they are frequently studied in experimental bench research. Rather, genetic risk factors primarily affect how a person’s brain gets wired up, Buckner suggested. In short, genetics define vulnerability by the time of young adulthood; the rest—metabolism, hubs, activity patterns, amyloid, tau, atrophy, cognition—unfolds from there.—Gabrielle Strobel.
This is Part 3 of a seven-part series. See also Part 1, Part 2, Part 4, Part 5, Part 6, and Part 7. View a PDF of the entire series.