Integrative Brain Transcriptome Analysis Links Complement Component 4 and HSPA2 to the APOE ε2 Protective Effect in Alzheimer Disease

Rebecca Panitch, Junming Hu, Jaeyoon Chung, Congcong Zhu, Gaoyuan Meng, Weiming Xia, David A. Bennett, Kathryn L. Lunetta, Tsuneya Ikezu, Rhoda Au, Thor D. Stein, Lindsay A. Farrer, Gyungah R. Jun

Preprint posted on November 24, 2020

'Complement'ing APOE ε2 in Alzheimer’s protection: A meta-analysis of differentially expressed genes in humans.

Selected by Theresa Pohlkamp


Alzheimer’s disease (AD) is a neurodegenerative disease presenting with progressive forgetfulness. AD pathology is defined by deposition of neurotoxic Aβ particles into plaques between brain cells, and neurofibrillary tau-tangles within neurons. The APOE gene encoding the cholesterol transporter Apolipoprotein E is the most significant genetic risk factor for developing the most common, late onset form of AD (LOAD). Of the three common allelic isoforms (ε2, ε3, and ε4), APOE ε3 is most frequent and considered neutral with respect to AD. If you carry one (e.g. ε4/ε3) or two (ε4/ε4) alleles of ε4, your risk to develop AD increases stepwise; contrarily, if you carry one or two alleles of ε2 your risk decreases. Most scientific studies try to uncover the mechanism how APOE4 contributes to AD pathology, less often the focus is on APOE2 and its protective role. In this preprint, Panitch et al. investigate differentially expressed genes (DEGs) between AD cases and controls, with a special emphasis on ε2/ε3-carriers. Their meta-analysis identifies components of the complement signaling pathway as top DEGs, and a glia-enriched APOE ɛ2 related co-expression network including complement pathway genes.

The complement system has been implicated in diverse neurodegenerative conditions, in most scenarios it drives inflammation to exacerbate pathology (Morgan, 2018). The complement cascade comprises several secreted effector components (proteins), their fragments, and membrane-bound complement receptors (CRs). It is part of the innate immune response and contributes to the extent and limit thereof. In general, antibodies recognize antigens, for example presented on pathogens. In the classical complement cascade, the initial antibody-recognition component is within the C1-complex. Activated C1 splits C2 and C4 into their respective a and b parts. C4b and C2a combine to activate C3, whose b-part is recruited to form C4b2a3b, which in turn converts C5. The now converted C5a-part has potent inflammatory and chemotactic activities, and the C5b-part anchors on the target cell surface to recruit and polymerize with C6-C9. Together, they constitute the lytic membrane-attack complex, a cylindrical hole in the antigen-presenting target cell. CRs are mostly expressed on blood and immune cells; upon binding to effector protein fragments, they induce diverse intracellular immune responses. Overall, the complement system can have both beneficial and detrimental effects in various neurodegenerative disorders (Schartz and Tenner, 2020). Notably, CR1, which primarily acts inhibitory upon C3B/C4B-binding, is expressed by brain microglia and CR1 is among the eight highest genetic risk factors for LOAD (Bellenguez et al., 2020; Wightman et al., 2020). In addition, and likely relevant in AD, APOE binds the C1 complex to keep inflammation in check (Yin et al., 2019). Another study found increased expression of several complement-pathway genes in an AD-mouse model encoding human APOE ε3 as compared to ε4 (Fitz et al., 2020). Thus, a link between the AD-protective isoform APOE ε2 and the complement system is an exciting piece to add to the puzzle.



Panitch and colleagues quantified DEGs in AD and control groups of three gene expression data sets of postmortem brains (ROSMAP with 627, MAYO with 162, and FHS/BUADC with 193 individuals, respectively). When they compared DEGs between AD- and control-groups dependent on APOE genotype (ε2/ε3, ε3/ε3, ε3/ε4), they found transcriptome-wide significant upregulation of C4A, C4B, and GFAP for AD, with the greatest and most significant difference within the ε2/ε3-group (Fig. 1A). When they plotted C4A, C4B, and GFAP expression levels with the progression of tau- (Braak stage) and Aβ- (CERAD score) pathology (ROSMAP dataset), they found that the expression of each of the three genes positively correlated with both pathologies. More specifically, the correlation was significant for all three genes and Braak in the genetic groups ε2/ε3 and ε3/ε3 but not ε3/ε4 (Fig. 1A). Plaque load correlated significantly for all three genes only for ε2/ε3-, but not for ε3/ε3- or ε3/ε4-groups.

In a gene co-expression network enrichment analysis (WGCNA) incorporating all three datasets, they defined 23 networks in total, and one of them (“M01” Fig. 1B) was specific to the APOE ε2/ε3 AD-cohort. Next, they analyzed single nucleus RNA sequencing (snRNAseq) data, which were available from the ROSMAP dataset (FASTQ, 48 subjects), for neuron/glia cell-type specific expression profiles and gene enrichment in the 23 networks. Overall, the individual networks were enriched in specific cell-types, APOE genotype and/or AD/control-groups. However, closer attention was drawn to the M01 network, which contained several complement pathway genes, including C4A, C4B, and CR1, and was enriched in most glia types (astrocytes, but stronger in oligodendrocytes and their progenitor cells) except microglia, but not in neurons. Within M01, 19 genes (including C4A and C4B) correlated with tau- and Aβ-pathologies. Lastly, by analyzing the expression of the most relevant M01 genes with AD-related proteins (within FHS/BUADC samples), they found that C4A, C4B, GFAP, PHLPP1, HSPA2, and DOCK1 positively correlated with phosphorylation of tau but not with Aβ.


Figure: A) Most significant DEGs between APOE groups (upper graphs) and their expression correlation with tau pathology (lower graphs). B) M01 gene co-expression network in APOE ε2/ε3 AD-group. DEGs in total samples and APOE ε2/ε3 group (turquoise), GWAS AD-risk genes differentially expressed in total samples (purple) and APOE ε2/ε3 group (red). (Reproduced and modified from original preprint, made available by a CC BY-NC-ND 4.0 international license)


Short discussion about the genetic hits GFAP, PHLPP1, HSPA2, and DOCK1:
The glial fibrillary acidic protein (GFAP) is upregulated in reactive astrocytes within the AD brain, which correlates with enrichment and secretion of complement components (Goetzl et al., 2018). PHLPP activation has been described to be dependent on specific CRs during proliferation and RTK-signaling (Strainic et al., 2020). DOCK1 and HSPA2 have not been linked to complement before. While HSPA2 is associated with AD-risk (Lancour et al., 2020), the DOCK1 gene product affects neuronal morphogenesis (Shi, 2013).


What I like about this preprint

I study the function of human APOE isoforms in mice. Recent research highlights that mouse glia in AD-models behave differently from human glia, also with respect to APOE genotype (Tcw et al., 2019). Thus, it is important to keep on track with human APOE isoform function in the species they evolved in and in which they contribute to AD risk. In humans, this preprint identifies an AD-related link between APOE, particularly isoform ε2, and complement pathway genes in glia, most robustly in oligodendrocytes, but not neurons. The enrichment of M01 genes particularly in oligodendrocytes further stresses the question of the involvement of white matter in AD pathology. While AD pathology is most obvious in the gray matter, myelination-deficits in the white matter have been observed (Butt et al., 2019). I appreciate that the authors broadly discuss their study’s weaknesses, for example that the absence of C4A/B-mRNAs in the snRNAseq data is possibly due to their extranuclear localization. They also clarify that their findings do not provide any mechanistical information about the interaction between APOE (ε2) and complement.


Questions for the authors

  1. In your study you adjusted the data for sex. Did you also look at the data in a sex-dependent manner? If so, are there any interesting insights with respect to complement pathway related genes?
  2. Several complement pathway related genes were upregulated in ε2/ε3-AD as compared to ε2/ε3-control. Thus, it could be interpreted that APOE ε2 fails to be protective when complement pathway related genes are upregulated. However, in your title and conclusion, you suggest that the complement system may confer a neuroprotective effect against AD through interaction with APOE ε2. Can you elaborate how you got to this conclusion and a scenario how this might work out?
  3. Given the data would exist, do you think comparing paired data of siblings, in each pair one being ε3/ε3-AD and the other ε3/ε2-control, would help us to identify APOE ε2-related protective gene networks / pathways?
  4. Have you seen alterations in APOE expression between AD/control, APOE genotype, and/or cell-types?



Bellenguez, C., Küçükali, F., Jansen, I., et al., 2020. Large meta-analysis of genome-wide association studies expands knowledge of the genetic etiology of Alzheimer’s disease and highlights potential translational opportunities. medRxiv 2020.10.01.20200659.

Butt, A.M., De La Rocha, I.C., Rivera, A., 2019. Oligodendroglial Cells in Alzheimer’s Disease. Adv Exp Med Biol 1175, 325–333.

Fitz, N.F., Wolfe, C.M., Playso, B.E., Biedrzycki, R.J., Lu, Y., Nam, K.N., Lefterov, I., Koldamova, R., 2020. Trem2 deficiency differentially affects phenotype and transcriptome of human APOE3 and APOE4 mice. Mol Neurodegener 15, 41.

Goetzl, E.J., Schwartz, J.B., Abner, E.L., Jicha, G.A., Kapogiannis, D., 2018. High complement levels in astrocyte-derived exosomes of Alzheimer’s disease. Ann Neurol 83, 544–552.

Lancour, D., Dupuis, J., Mayeux, R., Haines, J.L., Pericak-Vance, M.A., Schellenberg, G.C., Crovella, M., Farrer, L.A., Kasif, S., 2020. Analysis of brain region-specific co-expression networks reveals clustering of established and novel genes associated with Alzheimer disease. Alzheimers Res Ther 12.

Morgan, B.P., 2018. Complement in the pathogenesis of Alzheimer’s disease. Semin Immunopathol 40, 113–124.

Schartz, N.D., Tenner, A.J., 2020. The good, the bad, and the opportunities of the complement system in neurodegenerative disease. J Neuroinflammation 17.

Shi, L., 2013. Dock protein family in brain development and neurological disease. Commun Integr Biol 6.

Strainic, M.G., Pohlmann, E., Valley, C.C., Sammeta, A., Hussain, W., Lidke, D.S., Medof, M.E., 2020. RTK signaling requires C3ar1/C5ar1 and IL-6R joint signaling to repress dominant PTEN, SOCS1/3 and PHLPP restraint. FASEB J 34, 2105–2125.

Tcw, J., Liang, S.A., Qian, L., et al., 2019. Cholesterol and matrisome pathways dysregulated in human APOE ε4 glia. bioRxiv 713362.

Wightman, D.P., Jansen, I.E., Savage, J.E., et al., 2020. Largest GWAS (N=1,126,563) of Alzheimer’s Disease Implicates Microglia and Immune Cells. medRxiv 2020.11.20.20235275.

Yin, C., Ackermann, S., Ma, Z., et al., 2019. ApoE attenuates unresolvable inflammation by complex formation with activated C1q. Nat. Med. 25, 496–506.


Tags: alzheimer's disease, apoe3, gwas, healthy aging, neurodegeneration

Posted on: 17th December 2020


Read preprint (No Ratings Yet)

Have your say

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Sign up to customise the site to your preferences and to receive alerts

Register here