Imagine watching amyloid plaques grow peptide by peptide. As reported in the June 16 Science Advances, researchers led by Jörg Hanrieder, University of Gothenburg, Sweden, harnessed the power of mass spectrometry imaging and stable isotope labeling, aka iSILK, to track real-time plaque growth in exquisite detail in mice. They determined that Aβ1-42 forms an initial core, which then expands. Once plaques have begun growing, the mice begin to incorporate Aβ1-38 into the plaque periphery. In these mice, the whole process starts in the cortex, then spreads to the hippocampus. This new technique may be sensitive enough to answer questions about plaque pathology in Alzheimer’s disease.

  • iSILK tracks plaque formation in real time in mice.
  • Aβ1-42 builds the core, then Aβ38 attaches.
  • Plaques started in the cortex and spread to the hippocampus.

“This approach provides a new range and dimension to actively track pathology and physiology in vivo in the living brain with metabolic labeling and then spatially analyze by mass-spec imaging,” Randall Bateman, Washington University in St. Louis, wrote to Alzforum. Bateman and Hanrieder are collaborating to study the progression of pathological aggregates in people who had AD. “It’s important to understand the dynamics of growth and change of pathology to better help design prevention and treatment trials,” wrote Bateman.

In studying plaques, scientists can use histochemistry to see plaque composition or PET to track fibrils over time. However, both methods rely on antibodies or tracers that tag certain Aβ species. To get a less-biased view, researchers have begun using imaging mass spectrometry (IMS) to identify intact, unlabeled biomolecules, such as peptides and phospholipids, within tissue slices. Hanrieder has used this technique to identify lipids within plaques from mice (Kaya et al., 2018). Alas, despite IMS’ ability to capture the molecular makeup of substructures within tissue in remarkable detail, it only offers a snapshot.

Imaging Isotopes and Isoforms. Matrix-assisted laser desorption/ionization-imaging mass spectrometry, aka MALDI-IMS (top images), allowed researchers to track N15-labeled Aβ peptides in APP NL-G-F mice. Single-ion MALDI was able to discern where different-length Aβ isoforms were within plaques (bottom images). [Courtesy of Michno et al., Science Advances, 2021.]

Now, Hanrieder and colleagues have combined the spatial prowess of IMS with the ability of stable isotope labeling kinetics to track molecules in real time (see image above). SILK quantifies proteins that have incorporated specific isotopes. To investigate the dynamics of plaque formation, first author Wojciech Michno and colleagues ran a pulse/chase experiment. They fed 6- or 10-week-old APP NL-G-F knock-in mice Nitrogen15-labeled for four weeks, switched them to regular food, then analyzed their plaques at 17 weeks (see image below). In this way, they were able to distinguish Aβ made before or after plaques had begun to form, and use IMS to see where those peptides landed in the plaques.

Chasing Plaques. By feeding APP NL-G-F mice isotope-labeled protein for four weeks beginning at weeks 6 (top) and 10 (bottom), researchers tracked which Aβ isoforms were incorporated during plaque formation and growth. [Courtesy of Michno et al., Science Advances, 2021.]

In mice fed N15 from 6 to 10 weeks, labeled Aβ1-42 turned up mostly in the plaques' core, whereas in mice fed N15 from 10 to 14 weeks, labeled Aβ1-42 congregated mostly in the plaque periphery. This is consistent with dense cores forming first, followed by recruitment of Aβ to the more diffuse peripheries as the plaques grow outward. In support of this, mice continually fed N15 from 7 weeks on, which is shortly before plaques begin to form, had a mix of light and heavy Aβ in the dense cores, while only heavy Aβ was found in the plaque periphery.

Jochen Herms, Ludwig-Maximilians-University, Munich, is unconvinced. “I have doubts based on what we have seen in APP-KI of different ages, and in various studies, that ‘initial plaque formation in APPNL-G-F mice starts with formation of a dense core,'" he wrote to Alzforum. “I have the feeling that there is an overstatement because the resolution of some of the techniques may not be high enough,” he added. He believes Aβ fibrils, in low amounts or small sizes, might be involved but are not being detected.

Noting that both fibrillar plaques with dense cores and diffuse plaques lacking cores are found in AD, Sabina Tahirovic and Michael Willem, also at LMU, wondered what these findings mean for people. “Plaque morphologies are not directly translatable from mouse to human, and it remains questionable whether fibrillar and more diffuse plaques actually follow the same evolutionary path,” they wrote (full comment below).

Recently, researchers at Mannheim University of Applied Sciences and elsewhere used IMS to resolve the three-dimensional structure of plaques from APP NL-G-F mice, finding an Aβ1-42-rich core surrounded by a Aβ1-38 shell (Enzlein et al., 2020). In contrast, Tahirovic and Willem said they spotted Aβ38 in the fibrillar coreS of APP NL-G-F mice.

Michno and colleagues found less Aβ1-38 in plaques than Aβ1-42. Aβ38 was evenly distributed through the core and periphery, whereas Aβ1-42 was mostly in the core. This suggested to the authors that the shorter peptide was added after plaques began forming. Colin Masters, University of Melbourne, Australia, didn’t find this unusual. “Aβ1-42 forming the nidus of the mouse plaque followed by Aβ1-38 is consistent with what we know about the human condition,” he wrote to Alzforum (full comment below).

Indeed, the relative amount of N15 in Aβ1-38 and Aβ1-42 throughout plaques was the same in mice continually fed the isotope. However, when the pulse/chase experiment began at six weeks, plaque Aβ1-38 contained 43 percent less N15 than did Aβ1-42, whereas when the experiment started at 10 weeks, plaque Aβ1-38 had four times as much isotope.

Taken together, the researchers concluded that Aβ1-38 is made and deposited into plaques after Aβ1-42. Why this happens is unclear. The authors posit that at some point after plaques begin to form, γ-secretase begins to generate more Aβ1-38.

Others interpreted these results differently. “I think it has nothing to do with the secretase because Aβ1-38 forms abundantly in these mice,” said co-author Henrik Zetterberg, also at U Gothenburg. “Rather, the already-formed plaques provide a surface for the typically soluble Aβ1-38 to bind to and misfold,” he told Alzforum. “It should also be considered that the on-off diffusion and the half-life of Aβ could be variable for different Aβ species,” wrote Tahirovic and Willem.

Researchers lamented that the mice used here make little Aβ1-40, the predominant Aβ isoform found in people with AD. “While the technology is impressive for gathering spatial and temporal data, that data are only as relevant to human AD as this particular mouse model is,” wrote William Klunk, University of Pittsburgh (full comment below).

As for where the plaques form, Michno found that in mice fed N15 at six weeks, cortical plaques had 1.75 times more labeled Aβ1-42 than did hippocampal plaques, suggesting plaques started to form in the cortex first. In support of this, in mice fed N15 at 10 weeks, cortical plaques had four times less labeled Aβ1-42 than those in the hippocampus. Masters cautioned that this is not what occurs in people. “In humans, the first deposits appear in the precuneus/posterior cingulate and involvement of the human hippocampus is a late-stage phenomenon, if it occurs at all,” he wrote.

Hanrieder and colleagues are doing similar experiments in APP NL-F mice, which start developing plaques at 6 months old. While these mice also have little Aβ1-40, they do not carry the Arctic mutation that accelerates Aβ1-42 aggregation, and thus have a longer course of amyloidosis. “NL-F mice are seen as a more realistic model of human disease,” Hanrieder said.

Looking to the future, scientists envision possible applications of this type of imaging. Mathias Jucker, University of Tübingen, Germany, and Lary Walker, Emory University, Atlanta, said they would like to see iSILK deployed to resolve the tiny soluble seeds that he thinks initiate plaques and drive disease (Nov 2020 news). Masters wondered about other aggregates. “It would be nice to see these sophisticated MS-based imaging techniques turned toward the neurofibrillary tangle,” he suggested.

SILK/IMS could even reveal if and how risk factors or treatments alter plaque dynamics. Zetterberg, Walker, and Carlo Condello, University of California, San Francisco, are also interested in using iSILK to better understand the incorporation of Aβ isoforms into blood vessel walls in cerebral amyloid angiopathy.—Chelsea Weidman Burke


  1. When, in 1937, John Steinbeck wrote Of Mice and Men, he was drawing on the words of Robert Burns that, “The best laid schemes o' mice an' men/Gang aft a-gley.” In this context, the natural history of Aβ amyloid plaques over 10 weeks in transgenic APP mice has to be cautiously interpreted when being compared to a 20- to 30-year period of Aβ aggregation in human AD. The elegant findings of Michno and colleagues that Aβ1-42 forms the nidus of the mouse plaque and is followed by Aβ1-38 is consistent with what we know about the human condition. But the spread from cortex to hippocampus may not be an exact replica of what is currently being disclosed by longitudinal Aβ-PET studies as people transition from preclinical to prodromal to clinical AD.

    There are so many interesting aspects thrown up by this paper it is difficult to know which should be highlighted. One observation that immediately caught my eye was that “an initial deposition of a small, tetrameric formyl thiophene acid (q-FTAA)-positive” Aβ1-42 core forms the initial deposit. This fits very nicely with our earlier observations that a tetramer (an anti-parallel dimer-dimer) may be at the heart of the aggregation process (Streltsov et al., 2011). But it raises many longstanding questions on how such an oligomer forms and diffuses toward the nidus of a plaque.

    Another observation would be that one would expect a 10-week-old mouse plaque to show substantial post-translational differences from a 20-year-old human plaque. The authors acknowledge this with respect to the ragged N-termini found in the human species, but the apparent relative preservation of the C-terminus (38 vs. 40 vs. 42) remains unexplained. Maybe it is a phenomenon related to the buried C-terminus compared to the exposed N-terminus? Certainly, a very topical issue in the light of aducanumab and other anti-amyloid therapeutic targets.

    The topographic distributions of Aβ plaques in “mice and men” should be different, since the underlying processes are fundamentally different. In humans, the first deposits appear in the precuneus/posterior cingulate/default mode network, for reasons that may be related to the normal function of APP in synaptic plasticity. Involvement of the human hippocampus is a late-stage phenomenon, if it occurs at all. In contrast, the transgenic APP mice are at the mercy of the promoters used to drive expression of the transgene, resulting in a topography that may have little relation to the normal function of APP.

    Finally, it would be nice to see these sophisticated MS-based imaging techniques turned toward the neurofibrillary tangle. Alas, the mice do not respond to the presence of aggregated Aβ in the same way that people do.


    . Crystal structure of the amyloid-β p3 fragment provides a model for oligomer formation in Alzheimer's disease. J Neurosci. 2011 Jan 26;31(4):1419-26. PubMed.

  2. As evidence grows for the importance of early intervention in Alzheimer's disease, an understanding of the initial pathogenic processes is essential. Michno and colleagues present iSILK as an auspicious method for following the development and compositional evolution of Aβ plaques in experimental mice. The study is a technical tour de force, and iSILK is a welcome tool for the longitudinal analysis of plaque formation.

    Because the study was performed in a genetically modified mouse model with three different AD-linked mutations in APP, the findings cannot yet be generalized to humans, in whom such multiple mutations don't occur. For instance, the Arctic mutation in humans results in cerebral amyloid angiopathy and ring-like Aβ plaques containing Aβ40 and Aβ42 but lacking a dense core, a pathologic phenotype not fully recapitulated in transgenic mice (e.g., Basun et al., 2008). 

    However, application of the method to different animal models, along with a comparison to related studies of plaques in humans, is likely to yield useful new insights into how plaques emerge, grow, and spread (Wildburger et al., 2018). The approach also might be employed to shed light on the inception of amyloid angiopathy, as well as its variable presence in the AD brain. It could also be informative to extend this general approach to the formation of the cryptic Aβ seeds that precede the appearance of plaques, and thus may be the prime mover of the pathogenic cascade in AD (Uhlmann et al., 2020). 


    . Clinical and neuropathological features of the arctic APP gene mutation causing early-onset Alzheimer disease. Arch Neurol. 2008 Apr;65(4):499-505. PubMed.

    . Amyloid-β Plaques in Clinical Alzheimer's Disease Brain Incorporate Stable Isotope Tracer In Vivo and Exhibit Nanoscale Heterogeneity. Front Neurol. 2018;9:169. Epub 2018 Mar 22 PubMed.

    . Acute targeting of pre-amyloid seeds in transgenic mice reduces Alzheimer-like pathology later in life. Nat Neurosci. 2020 Dec;23(12):1580-1588. Epub 2020 Nov 16 PubMed.

  3. I have read this paper with great interest. The molecular (type of Aβ peptides), spatial, and temporal resolution data, as well as the longitudinal information provided by the study, are quite unique.

    The study confirms previous observations, such as the seeding role of Aβ1-42 or the delayed deposition of hippocampal relative to cortical plaques, and provides important insights into early (in vivo) seeding of amyloid and plaque maturation. These early events are considered critical in amyloid pathology, but the mechanisms are poorly understood.

    I find particularly relevant the data about the deposition dynamics of different Aβ peptides across brain regions. In this regard, I would like to note that Aβ1-40 (the major Aβ species generated from wild-type APP) is not produced in the APPNL-G-F mice (due to the presence of the Iberian mutation [APP I716F]). It would be interesting to see if the dynamics of amyloid deposition changes and how it is modified when Aβ1-40 is present in the brain.

    I congratulate the authors for the elegant strategy.

  4. While the technology for gathering spatial and temporal data is impressive, that data is only as relevant to human AD as this particular mouse model is relevant to human AD—and that remains uncertain, particularly considering the novelty of this model and the fact that it is partly based on the Arctic mutation, which favors oligomeric Aβ over plaques.

    Beyond that, the notion of Aβ1-42 depositing first is in line with descriptions from postmortem human studies decades earlier. It is particularly unclear how important the Aβ1-38 species are in humans compared to Aβ1-40, which seems to play little role in this mouse model. 

    While technically elegant, it is not obvious to me how this study informs new avenues of therapeutic development. That is, agents that prevent accumulation and deposition of Aβ peptides, and/or clearance of oligomeric and fibrillar aggregates once formed, seem as relevant as ever.

  5. Michno et al. follow spatial Aβ aggregation dynamics in the well-established amyloidosis mouse model (APP-KI) which results in early Aβ deposition (Saito et al., 2014). The authors performed metabolic in vivo labeling with stable isotopes (SIL), followed by mass spectrometry imaging to measure SIL kinetics (iSILK) and put forward the hypothesis that Aβ pathology starts with the formation of dense-core Aβ42 deposits, followed by deposition of shorter Aβ38 peptides. We recently compared this APP-KI mouse line with the overexpression amyloidosis model APPPS1 (Radde et al., 2006) for their Aβ pathology and microglial proteomic profiles and were surprised to see the reduced levels of Thiazin red positive fibrillar Aβ (fibrillar cores) in the APP-KI model while diffuse Aβ material could be readily detected a lot earlier (Sebastian Monasor et al., 2020). We also verified this finding using the LCO probes (Sebastian Monasor et al., 2020) and PET imaging (Biechele et al., 2021). Thus, it would have been an advantage to also perform iSILK experiments in a model with a higher fibrillar Aβ content such as APPPS1. Exploring additional mouse models would also allow for generalization of the presented findings.

    Noteworthy, it remains questionable whether fibrillar and more diffuse plaques actually follow the same evolutionary path. Fibrillar plaques are often more prominent in later stages of AD and diffuse plaques are often referred to as immature plaques and such morphologies—lacking the fibrillar core—exist in AD brains (Walker, 2020). Translation of plaque morphologies from mouse to human is definitely not 1:1, but eventually we would like to know that what is observed in mouse models can be of relevance for human pathology. Moreover, we need to be cautious in interpreting diffuse deposits as something beneficial and dense core deposits as something detrimental. This very important issue remains controversial.

    A surprising result of the report is the association of Aβ38 with amyloid plaques much later than the early seeding Aβ42. This aspect is intriguing because the occurrence of Aβ38 is most likely linked to the presence of the Arctic mutation in this model, which was previously associated with Aβ38 deposition, primarily found in cerebral amyloid angiopathy (Moro et al., 2012). Of note, we detected Aβ38 also in the fibrillar core of Aβ plaques in the APP-KI mice (unpublished data), and the dynamics of the Aβ38 deposition should be further addressed by immunohistochemical and biochemical studies at different pathological stages. In addition, it should also be considered that the on-off diffusion and the half-life of Aβ could be variable for different Aβ species.

    Overall, elucidating plaque composition and their spatial distribution in the brain is highly relevant. Despite the technical limitations that may still have to be overcome with this technology, it is important to further develop this area of research. In particular, this technology may be helpful for the studies of amyloid plaques upon genetic modifications of microglial regulators as some of them (such as Trem2 or ApoE) contribute to changes in the amyloid plaque conformation (Biechele et al., 2021; Parhizkar et al., 2019). 

    Furthermore, this information may be relevant for the judgment of the amyloid plaque load upon Aβ immunotherapy as different antibodies may preferentially target and induce phagocytic clearance of denser fibrillar material while others may target more soluble (protofibrillar and oligomeric) Aβ species. Additionally, as already anticipated in this manuscript, the various familial AD mutations produce different Aβ species and iSILK technology may help in revealing Aβ diversity and their tissue distribution and facilitate our understanding of the underlying pathologies.


    . Glitter in the Darkness? Non-fibrillar β-amyloid Plaque Components Significantly Impact the β-amyloid PET Signal in Mouse Models of Alzheimer's Disease. J Nucl Med. 2021 May 20; PubMed.

    . APP mutations in the Aβ coding region are associated with abundant cerebral deposition of Aβ38. Acta Neuropathol. 2012 Dec;124(6):809-21. PubMed.

    . Loss of TREM2 function increases amyloid seeding but reduces plaque-associated ApoE. Nat Neurosci. 2019 Feb;22(2):191-204. Epub 2019 Jan 7 PubMed.

    . Abeta42-driven cerebral amyloidosis in transgenic mice reveals early and robust pathology. EMBO Rep. 2006 Sep;7(9):940-6. PubMed.

    . Single App knock-in mouse models of Alzheimer's disease. Nat Neurosci. 2014 May;17(5):661-3. Epub 2014 Apr 13 PubMed.

    . Fibrillar Aβ triggers microglial proteome alterations and dysfunction in Alzheimer mouse models. Elife. 2020 Jun 8;9 PubMed.

    . Aβ Plaques. Free Neuropathol. 2020;1 Epub 2020 Oct 30 PubMed.

  6. Molecular-specific maps of Aβ plaque formation in mouse brains represent an impressive demonstration of recent advances in imaging mass spectrometry. While it is unclear that the approach currently has the necessary sensitivity to detect plaques in human samples, the method described provides a wealth of complementary information to classical neuroimaging methods.

    In the future, imaging mass spectrometry may provide scientists with a detailed view of the location and interaction between plaque molecules and an array of experimental drug treatments.

Make a Comment

To make a comment you must login or register.


Research Models Citations

  1. APP NL-G-F Knock-in
  2. APP NL-F Knock-in

News Citations

  1. In Mice, Aducanumab Neutralizes Aβ Seeds

Paper Citations

  1. . Shedding Light on the Molecular Pathology of Amyloid Plaques in Transgenic Alzheimer's Disease Mice Using Multimodal MALDI Imaging Mass Spectrometry. ACS Chem Neurosci. 2018 Jul 18;9(7):1802-1817. Epub 2018 May 4 PubMed.
  2. . Computational Analysis of Alzheimer Amyloid Plaque Composition in 2D- and Elastically Reconstructed 3D-MALDI MS Images. Anal Chem. 2020 Nov 3;92(21):14484-14493. Epub 2020 Oct 14 PubMed.

Further Reading


  1. . Age and amyloid effects on human central nervous system amyloid-beta kinetics. Ann Neurol. 2015 Sep;78(3):439-53. Epub 2015 Jul 20 PubMed.

Primary Papers

  1. . Following spatial Aβ aggregation dynamics in evolving Alzheimer's disease pathology by imaging stable isotope labeling kinetics. Sci Adv. 2021 Jun;7(25) Print 2021 Jun PubMed.