Massive Proteomics Studies Peg Glial Metabolism, Myelination, to AD

RNA sequencing studies of postmortem human brain samples have generated massive datasets implicating all manner of biological functions in Alzheimer’s disease. But how do these findings translate to proteins—where the biological rubber meets the road? Two massive studies by many of the same authors at Emory University School of Medicine in Atlanta set out to answer this question by surveying the proteomes of thousands of brains. The first, published April 13 in Nature Medicine and led by Allan Levey and Nicholas Seyfried, identified an uptick in proteins involved in sugar metabolism and in those specific to microglia and astrocytes in the brains of people with AD. The second, published May 18 in Nature Neuroscience and led by Levey, Seyfried, and Thomas Wingo, tied oligodendrocyte abnormalities to both AD and cerebral atherosclerosis. Curiously, people with atherosclerosis had more neurofilament light in their brain, regardless of AD status. NfL turns up time and again as a marker of neurodegeneration. Overall, the findings highlight the pivotal roles of microglia, metabolism, and myelination in the two brain diseases, which commonly occur together.

  • Thousands of brain proteomes implicate glial and metabolic disturbances in AD.
  • People with AD had more “protective” microglial profiles.
  • Oligodendrocyte proteins and neurofilaments higher in cerebral atherosclerosis, regardless of AD.

The findings demonstrate the devilish complexity of AD pathophysiology, commented Henrik Zetterberg of the University of Gothenburg, who called the studies “heroic.” Zetterberg was also impressed that the researchers detected many of the AD-related proteins in cerebrospinal fluid, opening up the possibility of tracking the proteins in living cohorts.

“A strong finding was that these clusters were reproduced across cohorts, tissues, and proteomic techniques,” wrote Betty Tijms of Amsterdam UMC and Pieter Jelle Visser of Maastricht University in the Netherlands. “The proteomic dataset is a major asset for the research community.”

Transcriptomics has implicated numerous biological functions—microglial responses chief among them—in AD pathogenesis (Jul 2018 conference news; May 2019 news). However, not every messenger RNA ultimately translates into a functional protein. Advances in proteomics now make it possible to take stock of the thousands of proteins mingling in the brain, and thanks to a massive collection of meticulously acquired postmortem brain samples, proteomes can be measured in thousands of brains.

As reported in Nature Medicine, first author Erik Johnson and colleagues used mass spectrometry to measure the proteomes of more than 2,000 brains as part of the Accelerated Medicines Partnership for AD (AMP-AD). For their initial analysis, they compared levels of more than 3,000 proteins in 453 dorsolateral prefrontal cortex samples from four well-characterized cohorts: the Baltimore Longitudinal Study of Aging (BLSA), the Banner Sun Health Research Institute, Mount Sinai School of Medicine Brain Bank, and Adult Changes in Thought Study (ACT). The brains came from controls, people with asymptomatic AD, and people who had been diagnosed with AD. The proteins fell into 13 different modules based on co-expression patterns. Of these, six modules significantly correlated with major hallmarks of AD, including Aβ plaques, tau tangles, and cognitive dysfunction. 

Differential Proteomics. Among 453 brain samples from four cohorts analyzed by mass spectrometry, proteins fell into 13 modules (M1–M13) based on co-expression patterns. Associations with aspects of AD pathophysiology are indicated in red (positive association) or blue (negative association). Models were assigned to cell types based on known functions. [Courtesy of Johnson et al., Nature Medicine, 2020.]

Modules designated M1, M3, and M4 had the strongest ties to AD pathophysiology. M1 and M3 were dialed down in AD brains compared with controls, while M4 proteins were upregulated in asymptomatic and symptomatic AD. M1 comprised neuronal proteins involved in synaptic function, M3 was flush with mitochondrial proteins, and M4, the module most strongly tied to disease, contained proteins involved in sugar metabolism as well as those expressed in microglia and astrocytes.

The researchers repeated their analysis in a separate batch of 340 DLPFC samples from the Religious Orders Study and Memory and Aging Project (ROS-MAP). They found much the same pattern. They also spotted several of the same AD modules among 111 temporal cortex samples from the Mayo Clinic, as well as precuneus samples from the BLSA cohort, suggesting the proteins’ association with disease was not limited to the DLPFC.

The researchers focused deeper on the most significant module—M4—finding that many of the proteins in this module happened to be products of genes near AD risk loci. This suggested that proteins in the module could be causing AD and not simply responding to it. Using previously published data from mouse models to infer function for the module protein, Johnson and colleagues estimated the majority of microglial proteins in the module are neuroprotective and anti-inflammatory, rather than harmful or pro-inflammatory (Rangaraju et al., 2018). Johnson hypothesized that increased expression of this module reflects an attempt by microglia to protect the brain. This may fail as disease progresses, the authors suggest.

Interestingly, many proteins in the M4 module play a role in glucose metabolism. Although the proteomic data cannot prove that the metabolic proteins came from glia, their tight co-expression with glia-specific proteins suggests that they may reflect a stepped-up metabolism in the cells as they exert their protective functions, Johnson told Alzforum. The microglial receptor TREM2, a major risk factor for AD, is believed to help drive microglial metabolism, although it was not among the differentially expressed proteins in this study (Aug 2017 news).

Finally, the researchers detected 27 of the proteins in the M4 module within the cerebrospinal fluid of two separate cohorts at Emory University. Ten proteins were significantly elevated in people with AD, including CD44, PRDX1, and DDAH2, and metabolic proteins lactate dehydrogenase (LDHB) and pyruvate kinase (PKM). Three—CD44, LDHB, and PKM—were also elevated in the CSF of asymptomatic people who had biomarker evidence of AD.

The CSF findings suggest that these proteins can be tracked longitudinally in living subjects, which could help explain their role in disease progression, Johnson and Zetterberg noted.

Intersection Between Cerebral Atherosclerosis and AD
In the second paper, first author Aliza Wingo and colleagues used proteomics to ask what are the effects of cerebral atherosclerosis, and are they shared with those seen in AD? Both disorders have been implicated in cognitive decline. The researchers sequenced the proteomes of 438 DLPFC samples from the ROS-MAP cohort, correlating with eight brain pathologies in addition to cerebral atherosclerosis: β-amyloid, tau tangles, gross infarcts, microinfarcts, cerebral amyloid angiopathy, TDP-43, Lewy bodies, and hippocampal sclerosis. They asked which proteins were altered in atherosclerosis, independent of the other pathologies.

Wingo used a protein-wide association study approach. PWAS are similar to the genetic GWAS, but using protein-expression differences instead of genetic differences. From the PWAS, 114 proteins emerged, including 32 that were more abundant and 82 that were scarcer in people with the cerebral blood vessel disease. Many of the more-abundant proteins hailed from oligodendrocytes and were involved in myelination, while less-abundant proteins played roles in mRNA processing and splicing. These associations held independently of cerebrovascular risk factors, including infarcts, diabetes, hypertension, and smoking.

In addition to the PWAS, Wingo also ran a weighted co-expression analysis on the entire proteomic dataset, identifying 31 modules of co-expressed proteins. Of these, five modules were independently linked to cerebral atherosclerosis. In a nutshell, the modules painted a picture of elevated oligodendrocyte differentiation, development, and remyelination; less RNA splicing and mRNA processing in neurons and astrocytes; and decreased synaptic signaling, regulation, and plasticity in the brains of people with CA.

How would the proteomic effects of AD match up? To address this, the researchers did another PWAS, this time fishing out 856 proteins that associated with a clinical AD diagnosis. Of these, 23 were among those that also associated with cerebral atherosclerosis. Notably, seven proteins, which were more abundant in both diseases, were part of an atherosclerosis-associated module steeped in oligodendrocyte proteins.

Two other standouts—neurofilament light (NfL) and neurofilament medium (NfM)—were higher in cerebral atherosclerosis and AD, and increased with worsening cognition, from controls to MCI to AD. Neurofilament proteins are thought to reflect axonal damage, and NfL is a CSF biomarker for AD. However, the researchers found that levels of the neurofilament proteins only associated with cerebral atherosclerosis, and not with β-amyloid, tau tangles, or any of the other measured pathologies. Furthermore, the relationship between neurofilaments and AD was not significant when the researchers adjusted for cerebral atherosclerosis. Overall, the findings suggested that the abundant neurofilament proteins reflected damage inflicted by atherosclerotic plaques, not by Aβ or tau pathology. They also supported the idea that the brain-blood-vessel disease is a strong contributor to cognitive impairment.

Zetterberg was surprised by these findings, especially in light of recent studies in people with familial AD mutations, which have detected an uptick in CSF NfL 16–22 years prior to symptom onset, at ages when atherosclerosis in the brain is unlikely (Jan 2019 news; Aug 2019 conference news). One explanation for these seemingly contradictory findings is that in people with familial AD, Aβ and tau accumulation at a young age are sufficient to damage neurons and ultimately, to cause cognitive impairment, without cerebrovascular disease, Zetterberg said. Perhaps in late-onset AD, both vessel damage and AD pathology are needed to harm neurons and manifest cognitive decline. Another idea worth investigating is whether familial AD mutations wreak havoc on blood vessels, he said, noting that mutations in presenilin might interfere with blood-vessel development, while mutations in APP cause cerebral amyloid angiopathy (CAA)—a buildup of Aβ in blood vessels.

Of course, NfL biomarker studies only measure proteins released into CSF, while the current studies measure the proteins that remain in the brain. Seyfried told Alzforum that protein-labeling studies in mice may help track the cellular origin of proteins, as well as their movement into biofluids like CSF.

Jonathan Schott and Dylan Williams of the University College London were particularly intrigued by the relationship between CA and neurofilament proteins, also pointing out that NfL is known to rise in response to all manner of brain injuries, including those where CA is unlikely to exist. “While this is not likely to explain elevation of NfL in all conditions, it reinforces the importance of considering cerebrovascular disease both as a co-pathology and also as a potentially core (and modifiable) feature of late onset sporadic AD,” they wrote.—Jessica Shugart

Comments

  1. Johnson et al. provide fascinating findings on mass spectroscopy proteomics in brain tissue from over 2,000 pathologically defined individuals with AD, other dementias, and controls as part of the multicenter Accelerating Medicine Partnership for AD (AMP-AD) project.

    They clustered 3,334 proteins in a discovery cohort of 453 individuals with AD or controls and identified 13 modules by weighted co-expression network analysis (WCNA). A strong finding was that these clusters were reproduced across cohorts, tissues, and proteomic techniques. The proteomic dataset is a major asset for the research community.

    Module 4 showed the strongest association with amyloid CERAD, tau Braak, and MMSE scores. It included 186 proteins, which were on average increased in AD relative to controls. The model was enriched for microglia and astrocyte proteins, and MAGMA analyses suggest overrepresentation for AD SNPs. Remarkably, the proteins associated with this module were also increased in FTLD-TDP and CBD. This result seems at first glance unexpected, since amyloid pathology is the pathological hallmark for AD and not commonly seen in FTLD-TDP and CBD. Possibly, this overlap could be explained by abnormal tau processing, which is shared between these disorders. 

    Further analyses on module 4 proteins showed that they could be involved in an anti-inflammatory response, which the authors suggested may reflect a protective response. However, higher concentrations of module 4 proteins were also associated with worse cognitive impairment and pathology. 

    In the related study, Wingo et al. tested the association of tissue proteomics with cerebral atherosclerosis (CA). Of the 8,362 proteins investigated in 438 individuals, 114 proteins showed an association with CA, of which 82 proteins showed lower concentrations with the presence of CA. These proteins were enriched for RNA processing. Another 32 proteins showed increased levels, and were enriched for oligodendrocyte development. WCNA was also performed in the total group, producing 31 modules. Five of these modules were related to CA, including a module with synaptic proteins, which were decreased relative to controls. Two of those modules also showed an association with a clinical AD dementia diagnosis, but not with amyloid plaque or tau tangle burden. This could mean that CA could contribute to dementia that shows clinical similarities with AD, independent of AD pathology.

    View all comments by Pieter Jelle Visser View all comments by Betty Tijms

  2. This impressive study demonstrates the potential of combining proteomics with detailed pathological phenotyping to yield important mechanistic insights into the causes of dementia. Using this approach, the authors elegantly demonstrate a number of molecular changes specific to atherosclerosis, as well as others shared with AD.

    A particularly intriguing conclusion is that the well-established relationship between AD and neurofilament light (NfL)—increasingly measured as a biomarker of neurodegeneration where it is thought to reflect damage to large-caliber myelinated axons—is mediated via cerebral atherosclerosis. While this is not likely to explain elevation of NfL in all conditions—CSF and blood NfL concentration is elevated in numerous neurodegenerative and non-neurodegenerative diseases, some of which, e.g. Huntington’s, familial Alzheimer’s, head injury, are unlikely to be mediated via conventional vascular risk factors—it reinforces the importance of considering cerebrovascular disease both as a co-pathology and also as a potentially core (and modifiable) feature of late-onset sporadic AD.

    Importantly, however, any temporal associations between these pathologies can only be inferred from autopsy studies: longitudinal in vivo studies incorporating biomarkers are required to determine the exact sequences of changes and their interrelationships.

    View all comments by Jonathan Schott View all comments by Dylan Williams

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. A Delicate Frontier: Human Microglia Focus of Attention at Keystone
  2. When It Comes to Alzheimer’s Disease, Do Human Microglia Even Give a DAM?
  3. Without TREM2, Microglia Run Out of Gas
  4. Neurofilament in Blood Foretells Early Onset Alzheimer’s
  5. Colombian Cohort Delivers Data on Blood NfL

Paper Citations

  1. Rangaraju S, Dammer EB, Raza SA, Rathakrishnan P, Xiao H, Gao T, Duong DM, Pennington MW, Lah JJ, Seyfried NT, Levey AI. Identification and therapeutic modulation of a pro-inflammatory subset of disease-associated-microglia in Alzheimer's disease. Mol Neurodegener. 2018 May 21;13(1):24. PubMed.

Further Reading

Primary Papers

  1. Johnson EC, Dammer EB, Duong DM, Ping L, Zhou M, Yin L, Higginbotham LA, Guajardo A, White B, Troncoso JC, Thambisetty M, Montine TJ, Lee EB, Trojanowski JQ, Beach TG, Reiman EM, Haroutunian V, Wang M, Schadt E, Zhang B, Dickson DW, Ertekin-Taner N, Golde TE, Petyuk VA, De Jager PL, Bennett DA, Wingo TS, Rangaraju S, Hajjar I, Shulman JM, Lah JJ, Levey AI, Seyfried NT. Large-scale proteomic analysis of Alzheimer's disease brain and cerebrospinal fluid reveals early changes in energy metabolism associated with microglia and astrocyte activation. Nat Med. 2020 May;26(5):769-780. Epub 2020 Apr 13 PubMed.
  2. Wingo AP, Fan W, Duong DM, Gerasimov ES, Dammer EB, Liu Y, Harerimana NV, White B, Thambisetty M, Troncoso JC, Kim N, Schneider JA, Hajjar IM, Lah JJ, Bennett DA, Seyfried NT, Levey AI, Wingo TS. Shared proteomic effects of cerebral atherosclerosis and Alzheimer's disease on the human brain. Nat Neurosci. 2020 Jun;23(6):696-700. Epub 2020 May 18 PubMed. Correction.

Annotate

To make an annotation you must Login or Register.