04 Nov 2017

In the October 31 Nature Communications, researchers led by Oskar Hansson of Lund University in Sweden reported that in cognitively normal people, parts of the default mode network (DMN) were among the first regions to accumulate Aβ deposits detected by PET imaging. The longitudinal study followed people who had abnormally low Aβ42 in their cerebrospinal fluid (CSF), but were deemed amyloid-negative on global PET scans. While these earliest deposits did not correlate with neurodegeneration, they did correlate with a loss of connectivity within and beyond the DMN.

In people with abnormal CSF Aβ42, PET first detects Aβ in and around the default mode network.
Early amyloid accumulation correlated with weak connectivity in and between the DMN and other regions.
There were no signs of neurodegeneration, cognitive decline, or changing brain metabolism.

“These results support the idea that synaptic dysfunction caused by amyloid aggregation is a very early event in Alzheimer’s disease,” commented Betty Tijms of VU University in Amsterdam (see full comment below). They also add more evidence that measures of brain connectivity are particularly sensitive at picking up subtle brain alterations associated with amyloid accumulation, she wrote.

Deposits of Aβ start forming in the brain decades before the cognitive symptoms of AD emerge, but where do they first take hold? Postmortem neuropathological studies indicated neocortical regions, but these are cross-sectional studies, limited to examining single points in time (Price and Morris, 1999; Thal et al., 2002). Similarly, a recent cross-sectional study led by Michel Grothe of the German Center for Neurodegenerative Diseases in Rostock used florbetapir-PET data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to conclude that Aβ deposits crop up earliest in the temporobasal and frontomedial regions, and that CSF concentrations of Aβ42 nudge downward even in the nascent stages of deposition (Oct 2017 news). The authors borrowed from neuropathological staging methods to develop a four-stage scheme of regional Aβ progression. However, without longitudinal data, the authors could not definitively conclude where Aβ deposition first started.

Aβ’s First Strike. Tracking the annual deposition rate of Aβ revealed a distinctive accumulation pattern in early accumulators (a) compared to non-accumulators (b), and late accumulators (c). [Courtesy of Palmqvist et al., Nature Communications, 2017.]

For the current study, co-first authors Sebastian Palmqvist and Michael Schöll and colleagues used longitudinal data from ADNI to address the question. Previously, the authors had leveraged this data to confirm what others had reported: Among cognitively normal people, CSF Aβ42 concentrations become abnormal before global amyloid-PET scans do (Palmqvist et al., 2016). Using this finding as a foundation, the researchers divided ADNI participants into three categories. They dubbed 218 people who tested negative on both CSF and PET as “non-accumulators,” 59 participants with abnormal CSF Aβ42 but normal amyloid-PET scans as “early accumulators,” and 191 people who tested positive for both as “late accumulators.”

At baseline, early accumulators had slightly higher neocortical standardized uptake value ratios (SUVRs) on florbetapir-PET scans than did non-accumulators, although uptake was still within the normal range. The two groups had similar baseline cognition, hippocampal volume, and CSF tau concentrations. By contrast, late accumulators had Aβ throughout the brain, which was accompanied by hippocampal atrophy, and they performed worse on cognitive tests than did early or non-accumulators.

To search for the earliest regions to develop Aβ deposits, the researchers tracked regional amyloid-PET signals in early versus non-accumulators over two years. They found that tracer uptake increased in the medial orbitofrontal cortex; anterior, posterior, and isthmus cingulate cortices; and precuneus in the early accumulators, but not in non-accumulators. In these brain regions, the average annual growth rate of Aβ in early accumulators was quadruple that of non-accumulators.

The regions roughly matched those reported by Grothe and colleagues. Grothe pointed out that compared to his study, Palmqvist found relatively less Aβ deposition in the inferior temporal lobe and more in the posterior medial cortex. “Although their methodological approaches are very different, both studies agree that that amyloid-PET scans considered ‘amyloid-negative’ by commonly used criteria can harbor notable regional amyloid signal,” Grothe wrote to Alzforum.

These early Aβ deposits did not appear to cause neurodegeneration or reduce glucose metabolism in the brain, as longitudinal measures of both showed no difference between early and non-accumulators. Late accumulators, however, did have distinct loss of gray-matter volume and waning brain metabolism over the two-year follow up.

In an attempt to glimpse even earlier Aβ deposition, the researchers next scrutinized PET scans of “CSF converters.” These are people who had normal CSF Aβ42 concentrations at baseline, which then dipped into the abnormal range during the two-year follow-up. Eleven people fit the bill. All developed Aβ plaques in the same locations—the left posterior cingulate and right medial orbitofrontal cortex. The statistically underpowered findings did not survive correction for multiple comparisons. Still, the researchers noted that the localization of deposits in CSF converters aligned well with a subset of affected regions in the early accumulators.

By aligning their amyloid-PET data with a brain network atlas, the researchers determined that several of the early Aβ-accumulating regions—the posterior cingulate cortex, precuneus, and medial orbitofrontal cortex in particular—overlapped with certain parts of the DMN. The frontoparietal network (FPN) also aligned with some affected regions, although to a lesser extent.

Would these findings replicate outside of the ADNI cohort? To find out, the researchers looked to the Swedish BioFINDER study. This longitudinal biomarker study collects CSF and conducts brain scans and cognitive tests on its participants, but so far only baseline data is available for analysis. The study also uses a different PET ligand (18F-flutemetamol ) and CSF Aβ immunoassay than does ADNI. Despite the differences in study methodology, the results were roughly the same: Compared to 219 people with normal CSF and normal PET readings, the 30 people with abnormal CSF but normal amyloid-PET accumulated Aβ in the same regions as did early accumulators in ADNI.

The researchers next asked whether these early deposits of Aβ affected connectivity in the brain. Drawing on BioFINDER resting-state functional MRI, the researchers found that in people with abnormal, i.e. low, CSF Aβ42, network connections flagged both within the DMN and between the DMN and the FPN. Surprisingly, in a subset of 80 participants whose Aβ42 CSF concentrations were in the normal range but close to the abnormal threshold, the reverse was true: Lower CSF Aβ42 correlated with higher connectivity.

Lost Connections.

Among early accumulators, correlation between CSF Aβ42 concentration and brain connectivity mapped to the DMN. *[Courtesy of Palmqvist et al., 2017.]*

The researchers offered several explanations for this unexpected correlation. For one, when cortical hubs such as the medial frontal lobe and posterior cingulate cortex are more active, they produce more Aβ, and that would start to drive accumulation in the brain, reducing it slightly in the CSF. Hypoconnectivity would ensue later on, as the accumulating Aβ starts to take a toll on synapses. Alternatively, it could be that Aβ accumulation in the parenchyma, though insufficient to lower CSF levels into the abnormal range, boosts neuronal activity, they proposed. Again, this uptick would ultimately wane as Aβ accumulates further.

Rachel Buckley and Aaron Schultz of Massachusetts General Hospital in Boston commented that while this period of hyperconnectivity has been documented in other studies, none have demonstrated it so early in the AD cascade. “Further investigation of this phenomena is clearly warranted and may have important implications for understanding both the consequences and the drivers of Aβ pathology,” they wrote to Alzforum.

What makes regions in the DMN particularly vulnerable to Aβ deposition? Hansson speculated that increased activity of the DMN—perhaps exacerbated by lack of cognitive or social interactions—might slightly enhance Aβ production in these regions over decades. This hypothesis is in line with what others in the field have proposed (May 2011 news). The DMN has long been known to harbor a particularly heavy burden of Aβ pathology, along with metabolic and functional deficits in people with AD (Mar 2004 news; Sep 2005 news; Aug 2009 news).

Tijms was fascinated by one part of the DMN in particular. Her lab has tied loss of connectivity in the precuneus to lower CSF Aβ42 concentrations. From its position at the back of the parietal lobe, the precuneus is functionally and anatomically connected to many other regions of the brain. “This puts the precuneus in an ideal position to link early pathological alterations in terms of amyloid deposition and later atrophy in other distant areas, like the medial temporal lobe,” she wrote to Alzforum.

“The study by Palmqvist et al. represents an important step toward a better characterization of the earliest amyloid-accumulating regions in the human brain,” commented Grothe. He added that future studies should investigate the mechanisms that underlie selective vulnerability of some regions to Aβ accumulation. “This may provide interesting new insights into the complex pathophysiologic mechanisms of Alzheimer's disease.”—Jessica Shugart

Comments

- Betty Tijms
  Amsterdam UMC, loc. VUmc
- Posted: 04 Nov 2017
This study by Palmqvist et al. shows in a rigorous way that when amyloid values in CSF are abnormal, it is possible to identify early amyloid increases in a supposedly normal PET image when focusing on specific anatomical areas. A particularly innovative aspect of this study is the in-depth investigation of the relationship between early amyloid accumulation and its effects on functional connectivity patterns within the same individual. Together, these results provide support for the idea that synaptic dysfunction caused by amyloid aggregation is a very early event in Alzheimer’s disease.

The hypothesised link between amyloid aggregation and brain activation is appealing, as it provides an elegant explanation as to why seemingly remote areas may share a vulnerability to amyloid pathology, while atrophy occurs much later in other distant brain areas. Rodent studies have provided the most convincing support that brain areas showing higher levels of neuronal activation and functional connectivity seem to show increased Aβ secretion, and at later points during development, are more prone to amyloid plaque deposition (Bero et al., 2012; 2011). The present and other studies (e.g., Villain et al., 2012) identified the precuneus amongst brain areas showing the earliest signs of amyloid accumulation. The precuneus is an associative brain area that connects to many other areas in the brain either anatomically and/or functionally. This puts the precuneus in an ideal position to link early pathological alterations in terms of amyloid deposition and later atrophy in other distant areas, like the medial temporal lobe. Potentially, atrophy might result from the loss of connections through a mechanism related to Hebbian learning: Areas that fire together wire together, and potentially in case of neurodegenerative diseases, might also die together.

Using structural MRI, we previously observed in non-demented individuals that loss of gray matter connectivity of the precuneus (and other brain areas) was associated with lower Aβ values in CSF that were mostly within the normal range (Tijms et al., 2016), and associated with clinical progression (Tijms et al., 2017). The results from Palmqvist et al. further encourage that measures of brain connectivity seem to be especially sensitive to pick up subtle brain alterations associated with amyloid accumulation.

The results imply an inverted U-curve relationship, with connectivity values first increasing for normal amyloid levels near the cut point and decreasing after abnormality is reached, possibly because at that point synapses become affected. However, maximum connectivity values for normal but near abnormal amyloid values were similar to those in the most abnormal range. The peak of functional connectivity at the cut point was about five times higher, and so this might be a peculiarity for that subset of individuals. Further longitudinal functional MRI in combination with amyloid imaging would be necessary to study the temporal dynamics of functional connectivity and amyloid alterations in more detail.

References:

Bero AW, Bauer AQ, Stewart FR, White BR, Cirrito JR, Raichle ME, Culver JP, Holtzman DM. Bidirectional relationship between functional connectivity and amyloid-β deposition in mouse brain. J Neurosci. 2012 Mar 28;32(13):4334-40. PubMed.

Bero AW, Yan P, Roh JH, Cirrito JR, Stewart FR, Raichle ME, Lee JM, Holtzman DM. Neuronal activity regulates the regional vulnerability to amyloid-β deposition. Nat Neurosci. 2011 Jun;14(6):750-6. Epub 2011 May 1 PubMed.

Tijms BM, ten Kate M, Wink AM, Visser PJ, Ecay M, Clerigue M, Estanga A, Garcia Sebastian M, Izagirre A, Villanua J, Martinez Lage P, van der Flier WM, Scheltens P, Sanz Arigita E, Barkhof F. Gray matter network disruptions and amyloid beta in cognitively normal adults. Neurobiol Aging. 2016 Jan;37:154-60. Epub 2015 Oct 22 PubMed.

Tijms BM, Ten Kate M, Gouw AA, Borta A, Verfaillie S, Teunissen CE, Scheltens P, Barkhof F, van der Flier WM. Gray matter networks and clinical progression in subjects with predementia Alzheimer's disease. Neurobiol Aging. 2018 Jan;61:75-81. Epub 2017 Sep 20 PubMed.

Villain N, Chételat G, Grassiot B, Bourgeat P, Jones G, Ellis KA, Ames D, Martins RN, Eustache F, Salvado O, Masters CL, Rowe CC, Villemagne VL, AIBL Research Group. Regional dynamics of amyloid-β deposition in healthy elderly, mild cognitive impairment and Alzheimer's disease: a voxelwise PiB-PET longitudinal study. Brain. 2012 Jul;135(Pt 7):2126-39. Epub 2012 May 23 PubMed.

View all comments by Betty Tijms

- Rachel Buckley
  Massachusetts General Hospital/Harvard Medical School
- Aaron Schultz
  MGH - HMS
- Posted: 04 Nov 2017
The approach of using CSF and PET to identify earlier stages of β-amyloidosis is quite interesting. Understanding where individuals are in the time course of disease is very difficult, particularly in preclinical stages, and the dual CSF/PET Aβ approach to help identify individuals in very early stage of Aβ accumulation looks quite promising. Using this group of early accumulators to identify regions of measureable accumulation over a few years is another nice addition that can help support primary prevention initiatives. Whether or not these regions represent an insight into disease biology, as opposed to a reflection of the sensitivity limitations of Aβ-PET tracers, remains an open question, and will be an interesting area of future research.

The authors also report default mode hyper-connectivity in CSF-low/PET-negative individuals and hypoconnectivity in CSF-positive/PET-negative individuals. A period of hyperactivity and hyperconnectivity has been implicated in several prior publications, but none have demonstrated the effect this early in the AD cascade. Further investigation of this phenomena is clearly warranted and may have important implications for understanding both the consequences of Aβ pathology, as well as the drivers of Aβ pathology.

View all comments by Rachel Buckley View all comments by Aaron Schultz

- Michel Grothe
  German Center for Neurodegenerative Diseases (DZNE)
- Posted: 04 Nov 2017
Although our methodological approaches were very different, there is a striking agreement between our finding that amyloid-PET scans considered to be "amyloid-negative" according to commonly used criteria can harbor notable regional amyloid signals in specific neocortical association areas. A distinct strength of the study by Palmqvist et al. is the use of longitudinal PET imaging and CSF biomarker data, which allows the authors to determine regional (i.e. voxel-wise) rates of amyloid accumulation in individuals with CSF evidence of cerebral amyloidosis but before conventional global PET thresholds are reached. Moreover, they combine their data with several other imaging markers of brain dysfunction and degeneration. The demonstration that these early, regionally restricted amyloid deposits have a measurable effect on brain function as measured by rs-fMRI connectivity, which is not paralleled by effects on imaging markers of overt neurodegeneration, is particularly intriguing. A major difference to our in vivo staging analyses is that the early amyloid regions are identified by group-level statistics, and thus the degree to which all individuals adhere to this early accumulation pattern remains unknown—a limitation also acknowledged by the authors in the discussion section.

The areas identified by Palmqvist et al. overlap to a great extent with the areas that define the earliest amyloid stages according to our in vivo staging scheme. However, some differences are also evident, such as relatively less prominent involvement of the inferior temporal lobe and more prominent involvement of the posterior medial cortex (posterior cingulate/precuneus). These differences may be accounted for by several methodological factors related to image processing and analysis (such as use and choice of partial volume correction method, reference region, region-wise versus voxel-wise analyses, etc.). However, given the markedly different analytic approaches used to determine these regions, the convergence on a distinct set of heteromodal association areas is still striking. Palmqvist et al. also use quantitative methods to demonstrate that these early affected areas coincide with a well-described functional brain network known as the default mode network (DMN). This nicely aligns with the previous observation by our group that although advanced amyloid deposition is widespread throughout the entire cortex, the DMN stands out as the most severely affected of all functional brain networks (Grothe and Teipel, 2016).

The study by Palmqvist et al. represents an important step toward a better characterization of the earliest amyloid accumulating regions in the human brain. An exciting avenue for future research that emerges from these findings is the question of the underlying mechanisms for this regionally selective vulnerability to amyloid pathology. Why does amyloid preferentially accumulate in these areas? Which functional, structural, or molecular characteristics distinguish these vulnerable areas from those that are more resistant to the accumulation of pathology? Answering these questions may provide interesting new insights into the complex pathophysiologic mechanisms of Alzheimer's disease (Sep 2016 conference news).

View all comments by Michel Grothe

- Elizabeth Mormino
  Stanford University School of Medicine
- Posted: 07 Nov 2017
I just recently commented on the recent Grothe staging paper and now would like to comment on this Palmqvist-Schöll staging paper (which is also a very nice paper!). As I said in my previous remarks, a very interesting aspect of this work is the goal of identifying very early amyloid accumulation with which to apply to individuals without clinical symptoms who may represent an ideal group for prevention trials. We know from postmortem work that traditional cut offs and global reads of amyloid PET data reflect high levels of plaque pathology, emphasizing that global PET measures are likely missing lower and intermediate levels of plaque pathology. There is also strong evidence to suggest that CSF amyloid changes before global elevations in PET. Thus, approaches that interrogate regional PET values during the earliest stages of accumulation may be insightful for understanding where and when signal emerges with amyloid PET during the preclinical stage of Alzheimer’s disease.

Interestingly, in looking at these papers side by side, there is little overlap in the early amyloid pattern described by Grothe (Figure 1, second column) versus Palmqvist-Schöll (Figure 1a). There seems to be overlap in the cingulate, but Grothe Stage 1 highlights the inferior temporal cortex whereas Palmqvist-Schöll Stage 1 highlights the default mode network. Although this may simply be due to methodological differences (Grothe used a cut-off-based approach on cross-sectional data; Palmqvist-Schöll used CSF-defined groups and examined change in amyloid PET over time), theses inconsistencies may represent an opportunity to further explore very early amyloid accumulation.

One possibility is that there is heterogeneity in the regions that are involved early in the disease across individuals, and that these two distinct approaches have different sensitivities for detecting these different patterns. Heterogeneity in the clinical presentation of Alzheimer’s disease patients has long been established, even among typical presentations, and although spatial patterns of amyloid deposition do not seem to correspond with this heterogeneity during the later stages of AD, distinct patterns of regional involvement may begin decades before symptoms. Thus, some individuals may accumulate preferentially in the inferior temporal cortex, and others in the precuneus, rather than all individuals undergoing the same pattern of accumulation over time. Therefore it may be worth deriving maps capturing “spread” over time at the individual subject level across all amyloid PET negative cases to understand whether regional diversity exists across individuals. Another important analysis will be to understand when these spatial patterns emerge, which may require examination of earlier ages (middle age).

Finally, now that these spatial maps have been derived across these two separate groups, the authors may consider sharing these masks and working together to understand this signal a little better. For instance, what is the correlation between the amyloid PET values extracted from Grothe Stage 1 and Palmqvist-Schöll Stage 1 maps among amyloid normal? Do extracted values from each of these maps provide independent information in predicting future cognitive decline? Overall, the approaches implemented across these two independent studies represent an important step toward establishing whether we have enough signal in amyloid PET to confidentially and reliably capture very early amyloid pathology before the familiar global pattern of accumulation is evident by PET.

View all comments by Elizabeth Mormino

- Marcus Raichle
  Washington University
- Posted: 28 Nov 2017
This is a very nice study that points ever more explicitly to the fact that selective vulnerability among brain networks is an important feature of AD. The brain’s default mode network (DMN) stands out in this regard and forces us to consider what it is about this network that underlies this vulnerability. Several things come to mind, but at the moment there is no definitive answer. Work by others (e.g., Margulies et al., 2016) places the DMN at the top of a hierarchy of brain networks conferring on this network a major responsibility for the overall functional organization of the brain. Genetically, it remains “youthful,” harboring genes concerned with synaptic plasticity (Goyal et al., 2014). Do the underlying mechanisms of such large-scale functions harbor the secret to the true cause of AD? We don’t know, but they certainly point to directions of inquiry that are likely to contain the answer.

References:

Margulies DS, Ghosh SS, Goulas A, Falkiewicz M, Huntenburg JM, Langs G, Bezgin G, Eickhoff SB, Castellanos FX, Petrides M, Jefferies E, Smallwood J. Situating the default-mode network along a principal gradient of macroscale cortical organization. Proc Natl Acad Sci U S A. 2016 Nov 1;113(44):12574-12579. Epub 2016 Oct 18 PubMed.

Goyal MS, Hawrylycz M, Miller JA, Snyder AZ, Raichle ME. Aerobic glycolysis in the human brain is associated with development and neotenous gene expression. Cell Metab. 2014 Jan 7;19(1):49-57. PubMed.

View all comments by Marcus Raichle