What FDG PET ‘Sees’ in AD: Angry Microglia, Not Just Neurons

FDG PET imaging measures glucose uptake in the brain. It has long served as a proxy for neuronal health—but is it? In the October 13 Science Translational Medicine, researchers led by Christian Haass and Matthias Brendel at the German Center for Neurodegenerative Diseases in Munich make a case that much of the FDG PET signal comes from activated microglia, not neurons.

  • In mouse brain, much of the FDG PET signal arises from microglial activation.
  • In AD brain, the early rise in FDG uptake may reflect inflammation.
  • Later in AD, the signal crashes due to extensive neuronal death.

In mouse brain, FDG uptake coincided with that of the microglia-specific tracer TSPO. When microglia were ablated, the FDG signal fell. When the scientists isolated labeled cells, microglia proved to contain almost 30 times as much FDG as did neurons, demonstrating that they gobble up glucose.

Likewise, in AD brain, FDG and TSPO uptake correlated. Both were higher in relatively preserved regions. In regions with extensive neuron loss, however, the FDG signal dwindled and was no longer aligned with TSPO. This finding hints that neuroinflammation could be responsible for the elevated FDG signal seen early in Alzheimer’s disease, while the later drop reflects the death of neurons.

Other researchers called this an important contribution. “This paper convincingly shows that microglia and their activation state drive a portion of the FDG-PET signal in the brain,” David Holtzman at Washington University in St. Louis wrote to Alzforum, adding, “Use of this measure as a biomarker assessment tool will need to take this issue into account.”

Mony de Leon and Gary Gibson at Weill Cornell Medical College in New York agreed. “These observations offer caution in the interpretation of FDG PET signals,” they wrote (full comments below).

Hello Microglia? In PS2APP amyloidosis mice (left), FDG uptake (orange) is high. Killing off microglia (middle) drops the signal below that seen in wild-type controls (right). [Courtesy of Xiang et al., Science Translational Medicine.]

In general, waning FDG uptake is a biomarker of AD and related disorders; the regional pattern of hypometabolism helps clinicians discern which neurodegenerative disease a patient might have.

The FDG signal has long been believed to arise from glucose use by neurons. Curiously, though, the authors had previously found that the FDG signal drops in mice carrying a TREM2 mutation that impairs their microglial activation (Kleinberger et al., 2017). They were also puzzled by the fact that mouse models of amyloidosis often have elevated, rather than decreased, FDG uptake (Brendel et al., 2016). 

These findings led Brendel and colleagues to wonder if microglia could be contributing to the FDG signal. First author Xianyuan Xiang, now at the Chinese Academy of Sciences in Shenzhen, used a colony-stimulating factor 1 receptor inhibitor to kill off microglia in wild-type mouse brain. This reduced FDG uptake by about 10 percent, suggesting microglia contributed to the signal. In the PS2APP mouse model of amyloidosis, the difference was more dramatic. Eliminating microglia in them suppressed the FDG PET signal by about a quarter, below the level seen in wild-type mice (see image above). This decrease correlated with a drop in TSPO uptake. Knocking out TREM2 in APPPS1 mice abolished the rise in FDG PET usually seen as these animals age. Taken together, this suggested the elevated signal arose from activated microglia.

To confirm that microglia were taking up glucose, the authors injected 18F-FDG into wild-type mice, waited 30 minutes for it to spread, and then isolated neurons, astrocytes, and microglia from their brains. The authors measured 18F gamma emissions from each cell type to determine how much FDG they had taken up. This process, which the authors dubbed radiosorting, revealed that microglia took up 12 times as much tracer per cell as did astrocytes, and 28 times as much as neurons. By contrast, microglia lacking TREM2 took up about a third as much FDG as did wild-type microglia, again showing that the cells need to be activated to soak up glucose. Previous research has found that TREM2 is essential for microglial metabolism (Aug 2017 news). In addition, RNA-Seq of isolated cells revealed that microglia from amyloidosis mice boosted expression of glucose transporters, suggesting an explanation for enhanced uptake.

What about the human brain? In 12 AD patients who participated in an imaging study, their FDG and TSPO signals coincided in the frontal cortex, which is relatively preserved at the AD stage; the signals did not coincide in their temporal and parietal lobes, which were undergoing neurodegeneration. Among 21 people suffering from a four-repeat tauopathy, this regional relationship was analogous. In them, FDG and TSPO correlated in their preserved parietal lobes, but not in the frontal and temporal lobes, where neurodegeneration was raging.

These results suggest that both neurons and microglia contribute to the FDG signal in neurodegenerative disease, with their relative contribution depending on how advanced the pathology is in each brain region. In a healthy brain, the microglial signal predominates, but once neurons start dying, they account for most of the hypometabolism, i.e., the drop in FDG signal. As neurons begin to degenerate in preclinical AD, increased microglial activity in response to proteopathy in the brain may mask the initial drop in neuronal metabolism, Brendel suggested. “That may explain why FDG PET is not a sensitive marker of early AD,” he said.

Microglial vs Neuron Signals. In AD brain (top panel), the FDG and TSPO signals correlate (gold) in the relatively preserved frontal cortex, but do not correlate (purple) in degenerating temporal, parietal, and occipital regions. In 4R tauopathy brains (bottom panel), the signals correlate in the preserved parietal/occipital lobe, but not in degenerating frontal and temporal regions. [Courtesy of Xiang et al., Science Translational Medicine.]

This interpretation belies earlier work, in which high FDG uptake in the hippocampi of preclinical AD patients was believed to indicate neuronal hyperactivation, perhaps as a form of compensation (Oct 2009 conference news; Benzinger et al., 2013; Ashraf et al., 2015). Tharick Pascoal, University of Pittsburgh School of Medicine, believes the microglial contribution may clarify other FDG findings, such as how amyloid causes FDG PET abnormalities in distant brain regions (Jun 2019 news). “The FDG network dysfunction reported in AD could be merely a proxy of microglial activation dysfunction,” he wrote (full comment below).

Intriguingly, another recent study arrived at the same conclusion as Brendel and colleagues. Working independently, researchers led by Dong Soo Lee at Seoul National University Hospital, Republic of Korea, devised a similar method to measure cellular FDG uptake, isolating hippocampal microglia and detecting gamma emissions. They found that microglia, but not other cell types, from 5XFAD mice took up more FDG than their wild-type counterparts, explaining the elevated signal in these mice. In people, elevated hippocampal FDG uptake associated with higher levels of sTREM2 in cerebrospinal fluid, again linking the signal to microglial activation (Choi et al., 2021). 

Brendel cautioned that his cell-sorting experiments do not paint a full picture of FDG uptake, because the isolation protocol captures only cell bodies, not neurites. Synapses are believed to binge on glucose, therefore Brendel is developing protocols to measure synaptic FDG uptake, as well. He is also investigating whether microglia use the glucose they vacuum up for their own metabolic needs, or whether they shuttle it to neurons.

Brendel believes that in the future, FDG could be used to monitor how microglia respond to therapies targeting them, such as TREM2 antibodies. He also noted that the radiosorting technique can be used together with other fluorinated PET tracers to determine what cell types take up those tracers.—Madolyn Bowman Rogers

Comments

  1. This paper convincingly shows that microglia, and their state of activation, drive a portion of the FDG-PET signal in the brain.

    It also shows that the FDG-PET signal correlates with microglial activation state in regions of the brain that have not undergone significant neurodegeneration. This appears to be the case since:

    1) the FDG-PET signal correlates with TSPO PET signal in mouse models with amyloid;

    2) the FDG-PET signal is decreased when microglia are removed or in TREM2 KO mice; and

    3) the positive correlation between TSPO-PET signal and FDG-PET signal is occurring in the human AD and FTD brain in regions that do not have significant neurodegeneration. 

    Showing the neurons are still likely driving a fair amount of the FDG-PET signal in human brain is that there is still a decrease in FDG-PET signal in the temporal and parietal region in AD, which corresponds well to where there is significant synaptic and neuronal loss occurring. These findings may explain why, in the very early preclinical period—i.e., ~25 years prior to clinical symptoms when amyloid is first accumulating—there is a relative increase in FDG-PET signal in autosomal-dominant AD. 

    These findings are important for another reason. As disease-modifying therapies for AD and other forms of neurodegeneration emerge, if the potential therapy results in increasing or decreasing microglial activation state, this will very likely affect the FDG-PET signal from brain. Thus, use of this measure as a biomarker assessment tool will need to take this issue into account.

    View all comments by David Holtzman

  2. The paper takes an old observation—that inflammation, central and peripheral, is associated with increased glucose metabolism—and insightfully and experimentally develops the idea that activated microglia are in large part responsible for this effect in AD models and AD. While prior studies have associated FDG PET increases with chronic inflammatory diseases including auto-immune disease, bowel inflammation, etc., little is known about the target cells accounting for the increased uptake.

    The finding that microglia manipulation can influence the FDG profile creates a very important opening to a long-questioned finding of transient increases in FDG PET uptake over the course of AD. While the authors examine living human AD and four-repeat tau groups, it remains unknown in humans what glucose metabolic changes are occurring in other cell groups when the metabolism of the microglia are increased, or conversely the impact of neuron metabolism damage on the time course of microglia activation.

    Finally, these observations offer caution in the interpretation of FDG PET signals.

    View all comments by Mony de Leon View all comments by Gary Gibson

  3. This is a very interesting paper, demonstrating that microglial activation contributes to FDG PET signal. The authors showed convincing evidence of influence from microglial activation on FDG uptake using experimental models and in vivo PET imaging. Previous studies have shown that astrocytes (Zimmer et al., 2017) play an important role in FDG uptake; now the authors suggest that microglial activation is the major player compared to both neurons and astrocytes.

    These results may contribute to our understanding of previous FDG observations that do not yet have a clear explanation. For example, increased microglial activation may be the major culprit in the FDG hypermetabolism seen in the early stages of AD.

    Several other FDG results could be revised based on their findings. For example, we have shown that amyloid leads to distal FDG PET network dysfunction (Jun 2019 news) and, more recently, that amyloid potentiates microglial activation PET network dysfunction leading to tau pathology (Sep 2021 news). The results presented by Xiang et al. may suggest that the FDG network dysfunction reported in AD could be merely a proxy of microglial activation dysfunction.

    Unfortunately, we will not be able to test this hypothesis, as the studies were performed in different cohorts that do not have both FDG and TSPO tracer. However, emerging datasets with glial markers measured in historical samples associated with FDG PET suggest that several cohorts will be able to shed light on these questions soon.

    Although the authors’ cell-sorting experiment showed compelling results, their conclusion that microglial activation is responsible for FDG signal could be balanced with the fact that, when we zoom out from the cellular resolution to the PET resolution, the fact that microglia represent a much smaller subpopulation in the brain than neurons and astrocytes may play a role in the final quantifiable FDG uptake in patients. The authors’ results showing often moderate, and lack of, correlation between FDG and TSPO PET in AD and 4RT may also point to a key role of neurons and astrocytes in the pathological FDG signal.

    In summary, this is important work that adds to the body of evidence suggesting that glial cells are key players in the FDG PET signal.

    References:

    Zimmer ER, Parent MJ, Souza DG, Leuzy A, Lecrux C, Kim HI, Gauthier S, Pellerin L, Hamel E, Rosa-Neto P. [(18)F]FDG PET signal is driven by astroglial glutamate transport. Nat Neurosci. 2017 Jan 30; PubMed.

    View all comments by Tharick Pascoal

  4. This article shows that glucose uptake is reduced in the brains of amyloid-bearing mice after depletion of microglia or after impairment of the ability of microglia to activate. Using very elegant cell-sorting technology after FDG injection, it shows at cellular resolution that microglia display higher glucose uptake than neurons and astrocytes.

    Interestingly, using [14C]-2-deoxyglucose glucose microscopic imaging by autoradiography in amyloid-bearing mice, we showed that hot spots of glucose uptake surround amyloid plaques, where microglia and astrocytes accumulate (Poisnel et al., 2012). It thus seems that microglia responsible for glucose uptake are in part associated with amyloid plaques, at least in amyloid-bearing mice.

    The article of Xiang et al. is clearly a great discovery with strong implications for interpretation of FDG-PET in the context of Alzheimer's disease.

    References:

    Poisnel G, Hérard AS, El Tannir El Tayara N, Bourrin E, Volk A, Kober F, Delatour B, Delzescaux T, Debeir T, Rooney T, Benavides J, Hantraye P, Dhenain M. Increased regional cerebral glucose uptake in an APP/PS1 model of Alzheimer's disease. Neurobiol Aging. 2011 Nov 11; PubMed.

    View all comments by Marc Dhenain

Make a Comment

To make a comment you must login or register.

References

Research Models Citations

  1. PS2APP
  2. APPPS1
  3. 5xFAD (B6SJL)

News Citations

  1. Without TREM2, Microglia Run Out of Gas
  2. St. Louis: Imaging Preclinical AD—Can You See it Coming in the Brain?
  3. Parsing Local and Distal Aβ Shows Links to Metabolism, Cognition

Paper Citations

  1. Kleinberger G, Brendel M, Mracsko E, Wefers B, Groeneweg L, Xiang X, Focke C, Deußing M, Suárez-Calvet M, Mazaheri F, Parhizkar S, Pettkus N, Wurst W, Feederle R, Bartenstein P, Mueggler T, Arzberger T, Knuesel I, Rominger A, Haass C. The FTD-like syndrome causing TREM2 T66M mutation impairs microglia function, brain perfusion, and glucose metabolism. EMBO J. 2017 Jul 3;36(13):1837-1853. Epub 2017 May 30 PubMed.
  2. Brendel M, Probst F, Jaworska A, Overhoff F, Korzhova V, Albert NL, Beck R, Lindner S, Gildehaus FJ, Baumann K, Bartenstein P, Kleinberger G, Haass C, Herms J, Rominger A. Glial Activation and Glucose Metabolism in a Transgenic Amyloid Mouse Model: A Triple-Tracer PET Study. J Nucl Med. 2016 Jun;57(6):954-60. Epub 2016 Feb 18 PubMed.
  3. Benzinger TL, Blazey T, Jack CR Jr, Koeppe RA, Su Y, Xiong C, Raichle ME, Snyder AZ, Ances BM, Bateman RJ, Cairns NJ, Fagan AM, Goate A, Marcus DS, Aisen PS, Christensen JJ, Ercole L, Hornbeck RC, Farrar AM, Aldea P, Jasielec MS, Owen CJ, Xie X, Mayeux R, Brickman A, McDade E, Klunk W, Mathis CA, Ringman J, Thompson PM, Ghetti B, Saykin AJ, Sperling RA, Johnson KA, Salloway S, Correia S, Schofield PR, Masters CL, Rowe C, Villemagne VL, Martins R, Ourselin S, Rossor MN, Fox NC, Cash DM, Weiner MW, Holtzman DM, Buckles VD, Moulder K, Morris JC. Regional variability of imaging biomarkers in autosomal dominant Alzheimer's disease. Proc Natl Acad Sci U S A. 2013 Nov 19;110(47):E4502-9. Epub 2013 Nov 5 PubMed.
  4. Ashraf A, Fan Z, Brooks DJ, Edison P. Cortical hypermetabolism in MCI subjects: a compensatory mechanism?. Eur J Nucl Med Mol Imaging. 2015 Mar;42(3):447-58. Epub 2014 Sep 30 PubMed.
  5. Choi H, Choi Y, Lee EJ, Kim H, Lee Y, Kwon S, Hwang DW, Lee DS, Alzheimer’s Disease Neuroimaging Initiative. Hippocampal glucose uptake as a surrogate of metabolic change of microglia in Alzheimer's disease. J Neuroinflammation. 2021 Aug 31;18(1):190. PubMed.

Further Reading

Primary Papers

  1. Xiang X, Wind K, Wiedemann T, Blume T, Shi Y, Briel N, Beyer L, Biechele G, Eckenweber F, Zatcepin A, Lammich S, Ribicic S, Tahirovic S, Willem M, Deussing M, Palleis C, Rauchmann BS, Gildehaus FJ, Lindner S, Spitz C, Franzmeier N, Baumann K, Rominger A, Bartenstein P, Ziegler S, Drzezga A, Respondek G, Buerger K, Perneczky R, Levin J, Höglinger GU, Herms J, Haass C, Brendel M. Microglial activation states drive glucose uptake and FDG-PET alterations in neurodegenerative diseases. Sci Transl Med. 2021 Oct 13;13(615):eabe5640. PubMed.

Annotate

To make an annotation you must Login or Register.