Rogue Gene Networks Track with Neurodegeneration Across Diseases

A sporadic case of neurodegenerative disease likely represents the culmination of myriad gene-expression changes in the brain. How to make sense of the chaos, let alone restore order? In the January Nature Medicine, researchers led by Daniel Geschwind at the University of California, Los Angeles, identified two key networks of dysregulated genes—one in neurons, the other in glia—in multiple animal models of frontotemporal dementia (FTD) and Alzheimer’s disease. The same two gene networks were compromised in postmortem brain samples from people with these diseases. The researchers zeroed in on a microRNA and histone deacetylases as key drivers of these networks.

  • In P301L-tau mice, cadres of genes changed expression in diseased brain regions.
  • The same networks are perturbed in postmortem samples from multiple neurodegenerative diseases.
  • Suppressing microRNA-203 and histone deacetylases restored network expression.

“This Herculean systems-level approach identified important pathogenic pathways in neurodegeneration that may be targetable by antisense oligonucleotides or small molecules,” commented Thomas Kukar of Emory University School of Medicine in Atlanta.

In recent years, researchers have started measuring co-expression changes that correlate with neurodegeneration (May 2013 webinar; Jun 2018 news). However, distinguishing changes that cause neurodegeneration from those that merely respond to it has proven difficult. Identifying therapeutic targets among such gene networks has been harder still. Animal models, while useful, are highly inbred, yielding insights that may not translate to other mice, let alone to people.

Grappling with these issues, co-first authors Vivek Swarup and Flora Hinz and colleagues searched for common gene-expression patterns that underpin neurodegeneration, and for ways to modulate those changes. Reasoning that tau pathology is a common denominator across many neurodegenerative disorders, including FTD and AD, the researchers analyzed gene expression in mice expressing the P301L mutant of human tau. They crossed the mice to three background strains, then compared gene-expression profiles among the offspring. At six months, hyperphosphorylated tau and gliosis were evident in the hippocampus, cortex, and brain stem, though cells were not dying yet. Compared with wild-type mice, all three tau transgenics had marked—and similar—changes in gene expression in these affected brain regions. Some expression changes also occurred in the cerebellum but often in the opposite direction, whereby the most upregulated genes in the cerebellum tended to be downregulated in the cortex. Since the cerebellum was spared from tau pathology, the expression changes there may reflect a protective response.

Common Change. On different genetic backgrounds, tau pathology came with common changes in gene expression over time. [Courtesy of Swarup et al., Nature Medicine, 2019.]

The researchers hypothesized that common gene-expression changes in affected regions across the different tau strains represented a neurodegenerative signature. To test this, they ran a weighted gene co-expression network analysis, zeroing in on two modules. One contained neuronal genes involved in synaptic function; the other comprised microglial and astrocyte genes involved in neuroinflammation. Lo and behold, the synaptic module contained genes previously implicated in FTD and progressive supranuclear palsy, including SLC32A1, NSF, and ELAVL2. AD risk genes, such as TREM2, ApoE, CLU, and C1q, populated the inflammatory module. Regions of the brain afflicted with early stage tau pathology were marked by suppression of the neurodegeneration-associated synaptic module, aka NAS, while showing an uptick of the inflammatory module (NAI).

Modules of Mayhem. Neurodegeneration-associated synaptic genes (top) were suppressed, while neurodegeneration-associated inflammatory genes (bottom) were more active in affected brain regions of P301L tau mice. [Courtesy of Swarup et al., Nature Medicine, 2019.]

When do these gene-expression changes occur in the course of disease? Geschwind said they happen in response to tau pathology, but prior to overt neurodegeneration. This suggested to him that the co-expression changes may contribute to early disease, rather than simply being a consequence of neuronal damage.

Wondering if these gene cliques pop up in other models of neurodegeneration, the scientists found NAS and NAI to be up- and downregulated, respectively, in affected brain regions of PS2APP, APP/PS1, and CRND8 AD models. In a progranulin gene mutation model of FTD, similar changes appeared over time, suggesting that this neurodegenerative signature holds across multiple disease etiologies.

Raffaele Ferrari of University College London considered this important. “This indicates that some of the neurodegenerative processes across neurological conditions are the same or similar, suggesting that effective therapeutic measures could target multiple different neurological conditions almost equally,” Ferrari wrote to Alzforum. He is a member of the International Frontotemporal Dementia Genomics Consortium, which co-authored the study.

It’s not just mice, either. Homologous modules emerged in postmortem brain samples from people with a range of neurodegenerative diseases. Both NAS and NAI networks were dysregulated in the cortices, but not cerebella, of people with tau-positive and tau-negative FTD, as well as in people with GRN-positive and GRN-negative FTD. Using mass spectrometry to test a subset of postmortem brain samples, the researchers concluded that these changes occurred at the protein level as well. The modules also cropped up in postmortem samples of people with AD, amyotrophic lateral sclerosis (ALS), and progressive supranuclear palsy (PSP), but not in samples from people with non-neurodegenerative brain disorders such as depression or schizophrenia. In people who had Aβ plaques, but no tau tangles or dementia, expression of these modules appeared normal.

Together, the findings suggested that while distinct genetic pathways instigate different neurodegenerative diseases, somewhere early in the pathogenic process gene-expression changes converge to compromise synaptic function and stoke neuroinflammation.

What controls this dysregulation? The researchers investigated microRNAs, which can suppress expression of groups of genes. Going back to the P301L-tau strain, they found altered expression of multiple microRNA networks in affected regions.

The single most upregulated microRNA was miR-203, which was elevated in autopsy tissue from people with AD and FTD. Its expression tracked up when NAS expression fell, suggesting miR-203 shuts down these neuronal genes. Indeed, overexpressing miR203 in cultured neurons or in the brains of wild-type mice suppressed genes in the synaptic module. In one-month-old P301L-tau mice, miR203 overexpression silenced multiple synaptic genes and increased apoptosis, whereas knocking down miR-203 prevented these changes.

The researchers also searched for small molecules that might counteract these neurodegenerative gene-expression changes. They screened Connectivity Map (CMap), a public resource of gene-expression responses to drugs, for compounds predicted to correct NAS and NAI changes. Four of the top 10 hits were histone deacetylase inhibitors. CMap analysis predicted these drugs would lift suppression of genes in the synaptic module, but suppress the inflammatory module. They researchers tested two HDAC inhibitors: scriptaid, which was the top hit, and suberanilo-hydroxamic acid. SAHA, also known as vorinostat, is currently undergoing a dose-finding trial in people with mild AD (see clinicaltrials.gov). Both compounds reduced cell death in neurons overexpressing miR-203, although SAHA became toxic above a concentration of 1 μM. In iPSC-derived neurons from people with FTD or AD, 0.5 μM SAHA restored expression of many genes in the NAS module. Neurons do not express many of the genes in the NAI module, and the researchers have yet to test whether the drugs restore normal expression to genes in that glial network.

Geschwind was surprised that histone deacetylase inhibitors emerged so clearly from the screen. He noted this does not mean they will make suitable drugs, given their potential for toxicity. Rather, he sees the finding as proof of principle that therapeutic approaches can emerge from complex gene network data.

Gerold Schmitt-Ulms of the University of Toronto said the study raises key questions, including how histone deacetylases and miR203 control the gene modules, and whether their inhibition can prevent cell death in vivo. “Dissecting the molecular underpinnings of their action may not be trivial, because both histone deacetylases and microRNAs typically have relatively low target specificity,” he added.

Peter Nelson of the University of Kentucky in Lexington was impressed that such a massive, multifaceted study would zoom in on specific therapeutic strategies. That said, Nelson is leery of a “one size fits all” approach to treating neurodegenerative disease. “There may be common nodes and bottlenecks, but these diseases are triggered by inherently different processes,” he said.

Noting that miR-203 is not normally expressed in the brain, Nelson wondered what its physiological relevance there might be. Walter Lukiw of Louisiana State University in New Orleans noted the same. Geschwind, however, considers miR203’s absence from the healthy brain a plus, since it would be a target only in the diseased brain.

In a joint comment to Alzforum, Evgenia Salta, Annerieke Sierksma, and Bart De Strooper of KU Leuven in Belgium praised the study for its thorough and systematic approach. They also cautioned that it relied on transcriptomics of bulk brain cells. “In light of recent reports on the many distinct, transcriptionally defined, microglial subtypes that seem to be differentially impacted by brain pathology or aging, analyzing modules that seemingly relate to a pool of different cell types, for instance microglia, astrocytes, and endothelia, precludes the ability to isolate the contribution of individual cell types to distinct forms of neurodegeneration,” they wrote. Geschwind agreed, and said that moving forward his lab will analyze these neurodegenerative networks using single-cell transcriptomics.—Jessica Shugart

Comments

  1. There is a pernicious trend in scientific publishing favoring unwieldy papers that appear to escape critical evaluation because their scientific merit can no longer be adequately assessed by reviewers. This paper is a standout, although it ironically also summarizes a Herculean body of work. It manages to stay focused on its objective to identify gene expression modules that may not only be associated with disease but have a higher chance to be causally linked to it by their enrichment in GWAS hits.

    Much of the study focuses on two neurodegeneration-associated gene-expression modules that emerged from deep RNA-seq analyses of four brain tissues obtained at two ages from mice expressing P301S tau on three distinct genomic backgrounds. Whereas the first module was associated with neurons and enriched in synaptic pathway genes (NAS), the other could be linked to microglial, astrocytes, and endothelial cells and fittingly was enriched in immune and inflammatory genes (NAI).

    Intriguingly, consistent gene expression changes within these modules were not only preserved in datasets from additional mouse models harboring pathological tau mutations but also in transgenic mice expressing distinct AD and FTD risk mutations, as well as the cortices of tau-negative and tau-positive FTD patients.

    Although the above suggests broad involvement of NAS and NAI modules in a variety of dementias, GWAS hits from AD mapped primarily to the NAI gene network and PSP/FTD genes were enriched in the NAS gene network, respectively. Thus, the data are consistent with an interpretation whereby there is cross-talk between these modules leading to convergent downstream etiology, but that the upstream causal pathways are distinct in these dementias.

    Noticing a strong inverse correlation between the expression of miR-203 and NAS gene products, the authors explored if this microRNA might be a driver that suppresses the expression of genes within this module. This turned out to be the case.

    Impressively, the study did not stop there but explored public data in the Connectivity Map (CMAP) database, which takes stock of how cell lines respond to drugs with gene-expression changes, for a compound that may revert the NAS and NAI module changes. A histone deacetylase inhibitor (Scriptaid), the top-scoring CMAP hit, then could be shown to reverse the death caused by miR-203 overexpression in a neuronal cell model.

    As with any good study, this work is no exception in that it raised many more questions than it could answer. Paramount among them loom how exactly histone deacetylases are linked to the NAS and NAI gene expression modules, and whether their inhibition can revert cell death in vivo. Dissecting the molecular underpinnings of their action may not be trivial, because both histone deacetylases and microRNAs typically have relatively low target specificity. However, the fact that the NAS module and its disease-relationship was also preserved in iPSC from human FTD patients carrying GRN mutations suggests that the latter might provide a suitable paradigm for follow-on investigations into the regulation of this module.

    The study should serve as a useful template for the application of integrated systems biology to other diseases. It makes for a refreshing read that I plan to recommend to students going forward.

    View all comments by Gerold Schmitt-Ulms

  2. This report on evolutionarily conserved gene networks mediating tau-related neurodegeneration by the group of Daniel Geschwind represents a very thorough and systematic study employing integrative co-expression network analyses in an impressive series of mouse and human datasets.

    By delving into the transcriptomic changes inferred by genetic background in conjunction to tau-induced neuropathology, the study adds to our realization of the significance of the genetic makeup for neurodegenerative disease progression. At the same time it emphasizes the usefulness of animal models for identifying possible disease-modifying molecular factors.

    The authors very elegantly dissect disease-specific transcriptional signatures that are region-, age-, or genetic background-directed and thereafter construct unique co-expression gene networks. Interestingly, the analyses converge into two modules that seem to be independent of genetic background and species; one associated with the neuronal/synaptic response and one with the immune response during tau-pathology progression.

    Even though this type of study cannot generally infer causality, the authors do a great job in further experimentally validating their findings and confirming the significance of microRNA regulators in neurodegeneration. Their observation that divergent genetic backgrounds in both mice and humans can lead to similar transcriptional responses that possibly associate with core pathological drivers or modifiers of tauopathies is one of the particularly exciting findings of this study. Moreover, that certain transcriptional modules consistently anti-correlated with disease progression may shed light onto novel approaches to harness endogenous neuroprotection mechanisms.

    Inevitably, the data also raise questions. Despite the depth and breadth of analysis, this work remains a bulk approach to brain transcriptomics. In light of recent reports on the many distinct, transcriptionally defined, microglial subtypes that seem to be differentially affected by brain pathology or aging, analyzing modules that relate to a pool of different cell types, like for instance microglia, astrocytes, and endothelia, precludes the ability to isolate the contribution of individual cell types to distinct forms of neurodegeneration.

    Similarly, limitations regarding network metrics may hinder the ability to unveil biological dynamics at the single-gene level. On the other side of the same spectrum, differential module enrichment for GWAS genes in AD (risk genes were enriched in the immune response module) compared to FTD and PSP (risk genes were enriched in the synaptic module), is a key observation derived from network-based analysis, and again underscores the hypothesis that genetic risk for AD is linked to an Aβ-, and not tau-induced immune response (Salih  et al., 2018; Sierksma et al., 2019; Felsky et al., 2019). 

    Lastly, it would have been of additional interest had the authors identified modules differentially regulated between human patient brain and animal disease models, as this would provide pivotal knowledge regarding the much-debated issue of the translational gap. In all, this study adds significant insights into disease mechanisms, possible dynamic network biomarkers, and novel disease-modifying strategies, and provides a wealth of information for interested researchers to further explore.

    References:

    Salih DA, Bayram S, Guelfi S, Reynolds RH, Shoai M, Ryten M, Brenton JW, Zhang D, Matarin M, Botia JA, Shah R, Brookes KJ, Guetta-Baranes T, Morgan K, Bellou E, Cummings DM, Escott-Price V, Hardy J. Genetic variability in response to amyloid beta deposition influences Alzheimer's disease risk. Brain Commun. 2019;1(1):fcz022. Epub 2019 Oct 10 PubMed.

    Sierksma A, Lu A, Mancuso R, Fattorelli N, Thrupp N, Salta E, Zoco J, Blum D, Buée L, De Strooper B, Fiers M. Novel Alzheimer risk genes determine the microglia response to amyloid-β but not to TAU pathology. EMBO Mol Med. 2020 Mar 6;12(3):e10606. Epub 2020 Jan 17 PubMed.

    Felsky D, Roostaei T, Nho K, Risacher SL, Bradshaw EM, Petyuk V, Schneider JA, Saykin A, Bennett DA, De Jager PL. Neuropathological correlates and genetic architecture of microglial activation in elderly human brain. Nat Commun. 2019 Jan 24;10(1):409. PubMed.

    View all comments by Bart De Strooper View all comments by Evgenia Salta View all comments by Annerieke Sierksma

  3. This elegant and thoroughly planned work allows us to appreciate a potential molecular cascade that possibly normalizes downstream processes of a spectrum of neurodegenerative conditions including AD, FTD, and PSP. This comprehensive study uses animal and human models, computational methods, and experimental validation to support results and interpretation.

    The approach is quite state of the art. A major strength here is that systems biology approaches prove their affordability to the study of complex disorders by representing a step forward from the classic one-gene-at-a-time, reductionist approach. The current study indicates a conserved change in expression patterns of genes involved in synaptic and inflammatory processes as a communal feature in specific cell sub-populations in brain areas topologically relevant to neurodegenerative diseases such as AD, FTD and PSP.

    It also indicates that aberrant expression patterns appear to happen downstream of pathology. They are thus a consequence rather than a cause of the degenerative process, almost independent from the putative genetic cause or risk factors. This could be relevant in that it indicates that possibly some of the neurodegenerative processes across neurological conditions are the same, therefore effective therapeutic measures could target multiple different neurological conditions.

    All the more, this work highlights potential drug targets, already existing tool compounds and their effect on a signature of degeneration. This work is also important in that it validates previous, purely computational, work that particularly highlighted a potential functionally relevant role for acetylases (e.g., EP300—see also Dec 2018 conference news) or other targets (e.g., ELAVL proteins) in the pathogenesis FTD (Ferrari et al., 2016; Ferrari et al., 2017). 

    This study provides a great deal of new and useful insight into neurodegeneration processes in FTD, especially suggesting where and how disease progression might be halted or slowed. Yet, it is not fully able to explain causality; it rather highlights conserved and convergent subcellular events as a signature of an already initiated and irreversible degenerative process. It would be thus warranted to discriminate whether these processes are a cause or a consequence, maybe by modulating them at early stage in mouse models, before brain damage signatures occur.

    Another point to consider is that the current study has used a number of different and variable models to study disease. The authors have done a superlative job minimizing most confounding variables, yet possibly, in the future, multi-omics data integration approaches generated from the same sample sources might promise to be even more informative for human disease.

    Another point to consider might be that of verifying whether other micro-RNAs, in addition to miR-203, appear to modulate any of the edges of the candidate modules? Finally, if the two disease modules highlighted here are shared across different disorders AD, FTD, ALS, and PSP, and different genetic models of any of these disorders exist, how do we account for the phenotypic differences between these disorders? In other words, do other molecular signatures contribute to disease pathogenesis, as well, and are they exclusive to FTD (with or without tau pathology, or with or without GRN mutations), AD, and PSP, etc.? How do we experimentally discriminate those?

    References:

    Ferrari R, Forabosco P, Vandrovcova J, Botía JA, Guelfi S, Warren JD, UK Brain Expression Consortium (UKBEC), Momeni P, Weale ME, Ryten M, Hardy J. Frontotemporal dementia: insights into the biological underpinnings of disease through gene co-expression network analysis. Mol Neurodegener. 2016 Feb 24;11:21. PubMed.

    Ferrari R, Lovering RC, Hardy J, Lewis PA, Manzoni C. Weighted Protein Interaction Network Analysis of Frontotemporal Dementia. J Proteome Res. 2017 Feb 3;16(2):999-1013. Epub 2017 Jan 12 PubMed.

    View all comments by Raffaele Ferrari

  4. This is an exciting advance that uses systems biology and in vivo validation to generate mechanistic insight into the causes of neurodegenerative diseases including FTD and AD. Although autosomal-dominant mutations in MAPT, GRN, and C9ORF72 are frequent causes of FTD, it is still a mystery how these mutations precisely lead to neuronal dysfunction and ultimately degeneration. Moreover, current animal models of FTD and AD do not fully recapitulate the human condition, which may partially explain the difficulty of identifying efficacious drugs in clinical trials for multiple neurodegenerative diseases.

    This work led by Vivek Swarup in Dan Geschwind’s lab at UCLA attempts to overcome these challenges by incorporating network analysis of transcriptomic and proteomic data from mouse models of AD/FTD and human brain FTD tissue. This approach identified modules or groups of synaptic-associated genes that decreased or inflammatory-associated genes that increased across disease models, and importantly also in human FTD. A key microRNA, miR-203, was identified that appears critical for synaptic dysfunction and neuronal death in vitro and in vivo. Subsequently, small-molecule histone deacetylase inhibitors were identified that could rescue the disease-associated changes in synaptic and inflammatory genes.

    Taken together, this herculean effort suggests this type of systems-level approach can identify important pathogenic pathways in neurodegeneration that may be targetable by antisense oligonucleotides or small molecules. Much work remains, but it will be exciting to understand how miR-203 and associated pathways are involved in neurodegeneration and if they can be targeted effectively as therapeutic targets in FTD, AD, and related dementias.

    View all comments by Thomas Kukar

Make a Comment

To make a comment you must login or register.

References

Webinar Citations

  1. Can Network Analysis Identify Pathological Pathways in Alzheimer’s

News Citations

  1. Culling Connection From Chaos, Alzheimer’s Genetic Network Study Pins PLXNB1 and INPPL1

Research Models Citations

  1. rTg(tauP301L)4510
  2. PS2APP
  3. APPswe/PSEN1dE9 (line 85)
  4. TgCRND8

External Citations

  1. Connectivity Map
  2. clinicaltrials.gov

Further Reading

Primary Papers

  1. Swarup V, Hinz FI, Rexach JE, Noguchi KI, Toyoshiba H, Oda A, Hirai K, Sarkar A, Seyfried NT, Cheng C, Haggarty SJ, International Frontotemporal Dementia Genomics Consortium, Grossman M, Van Deerlin VM, Trojanowski JQ, Lah JJ, Levey AI, Kondou S, Geschwind DH. Identification of evolutionarily conserved gene networks mediating neurodegenerative dementia. Nat Med. 2019 Jan;25(1):152-164. Epub 2018 Dec 3 PubMed.

Annotate

To make an annotation you must Login or Register.