Recent genetic findings have highlighted the contributions of the immune system to Alzheimer’s disease pathogenesis. In today’s Nature Neuroscience, a large group of geneticists led by Alison Goate at the Icahn School of Medicine at Mount Sinai, New York, suggest that innate immune cells may play an even more central role than previously thought. The researchers identified a polymorphism that lowers expression of the microglial transcription factor PU.1 and delays the onset of AD. PU.1 controls microglial differentiation, identity, and function. Lo and behold, among PU.1’s downstream targets, the authors found numerous known AD risk genes, including TREM2, TYROBP, CD33, MS4A cluster genes, and ABCA7. This suggests that expression of these genes might be controlled by PU.1 in microglia and that they form part of an AD gene network in those cells. Moreover, most of these AD genes are expressed in microglia, macrophages, and monocytes, rather than in neurons. “That suggests that a large proportion of the genetic risk for late-onset AD is explained by genes that are expressed in myeloid cells, and not other cell types,” Goate told Alzforum.

Others found the implications fascinating. “This is an extremely important and insightful piece of work, which firmly puts the microglial response as key to Alzheimer pathogenesis,” John Hardy at University College London wrote to Alzforum (see full comment below). Oleg Butovsky at Brigham and Women’s Hospital, Boston, noted, “The big excitement here is that PU.1 may regulate all these molecules associated with Alzheimer’s.” Because PU.1 controls so many things, however, this protein itself is unlikely to make a good therapeutic target, the researchers stressed. Instead, they believe further studies of PU.1’s role in microglia and AD pathogenesis may reveal promising downstream targets.

A Search for Genes That Alter Alzheimer’s Pathogenesis
Numerous large GWAS have turned up about two dozen loci linked to AD (see Apr 2011 newsJul 2013 conference news). For most of these loci, however, it remains unclear what the functional variant is, and how the variant raises or lowers the risk of Alzheimer’s. To begin to answer these questions, joint first authors Kuan-lin Huang and Edoardo Marcora examined more than 8 million single nucleotide polymorphisms (SNPs) in a cohort of 14,406 cases and 25,849 controls drawn from the International Genomics of Alzheimer’s Disease (IGAP) study. Because the age at which AD strikes holds clues to the underlying pathogenesis, the authors looked for any association between these SNPs and age of onset. Goate noted that previous age-of-onset studies included only people with AD; by adding controls, the authors hoped to increase the power to pick up protective genes, which might be enriched in people without the disease.

The researchers found four loci that met genome-wide significance. Three—ApoE, PICALM, and BIN1— previously had been associated with age of onset, while the fourth, MS4A, had not. The MS4A locus encodes a family of microglial receptors. Like PU.1, the protective allele at the MS4A locus decreases expression of these genes. In addition, the authors identified 18 other SNPs that correlated with age of onset but missed genome-wide significance. Fourteen of those were new to AD research.

Intriguingly, besides ApoE, only one of these total 22 onset-related loci associated with changes in cerebrospinal fluid biomarkers. This was SNP rs1057233, which correlated with higher CSF Aβ42 and a later onset of AD. SNP rs1057233 lies near the genes SPI1 and CELF1, a spot in the genome previously fingered as harboring an AD risk factor. SPI1 encodes PU.1, whereas CELF1 regulates mRNA splicing.

Do Microglia Harbor Most AD Risk?
How might these 22 SNPs influence AD pathogenesis? Many SNPs occur in noncoding regions of the genome and affect gene expression rather than protein sequence. To look for effects on expression, the authors made use of the BRAINEAC data set, a free database maintained by the U.K. Brain Express Consortium that compiles information on how SNPs relate to gene expression in whole-brain homogenate from 10 brain regions. This analysis, however, turned up little.

Goate wondered if focusing on specific cell types, rather than whole-brain homogenate, might be more revealing. A previous study had reported an enrichment for AD risk SNPs among those known to alter gene expression in monocytes (May 2014 news). That study, however, only examined monocytes and T cells. For a broader view, Goate and colleagues surveyed enrichment of AD risk SNPs in regulatory regions from 220 different tissues and cell types using a statistical method called LD score regression (Bulik-Sullivan et al., 2015). It turned out that SNPs that modulate AD risk are significantly enriched in regions of the genome that regulate gene expression in cells of the myeloid and B-lymphoid lineage, where PU.1 is known to act as a master regulator, but not in brain cells such as neurons. The findings were specific for AD, as schizophrenia risk alleles showed no enrichment in myeloid cells but were enriched in the brain. In other words, AD-associated variants are more likely to affect gene expression in myeloid cells than in other tissues or cell types. If AD genes are expressed primarily in myeloid cells, then that might explain why previous studies have struggled to link GWAS variants to gene expression changes in whole brain, Goate noted. Because microglia make up only a small percentage of brain cells, the expression signal from them would be swamped in whole-brain homogenate.

Given these data, the authors examined how their 22 onset-linked SNPs correlated with gene expression in myeloid cells. Because microglia cannot be isolated from brain in sufficient quantity for this type of analysis, the researchers used the Cardiogenics consortium data set of 738 monocyte and 593 macrophage samples. In these samples, 17 of the 22 SNPs correlated with expression changes in a nearby gene. Again the protective SNP rs1057233 stood out. This SNP, which lies in the untranslated tail region of SPI1, strongly associated with lower expression of SPI1 in both monocytes and macrophages. It also correlated with expression changes in three other nearby genes. A co-localization analysis suggested that this variant likely accounted for the AD risk signal in the SPI1/CELF1 locus by virtue of its ability to modulate PU.1 levels in myeloid cells.

PU.1 Takes Center Stage
Based on these results, the authors drilled down on SPI1. They first tried to pin down the functional variant responsible for expression changes. Rs1057233 is known to directly affect SPI1 expression by altering miRNA binding, making it a likely candidate (Hikami et al., 2011). However, three other SNPs in the SPI1 gene turned out to be inherited along with rs1057233. They change the DNA-binding motif of SPI1’s protein product, PU.1, suggesting they might affect gene expression as well. While the relative contribution of each SNP is unclear, the haplotype controls the variable expression of SPI1, the authors concluded.

To examine SPI1’s potential effect on AD, the authors searched for PU.1 binding sites at spots in the genome that have been linked to Alzheimer’s. Incredibly, among the 112 genes that lie within AD GWAS loci, 60 were expressed in human microglia and contained PU.1 binding sites. In addition to those listed above, they included MS4A4A, MS4A6A, PILRA and PILRB, TREML1 and others, but not ApoE. Overall, the authors estimated that AD heritability was enriched about 50-fold within PU.1 binding sites in monocytes and macrophages, significantly higher than what would be expected by chance.

If PU.1 controls so many AD genes, then what is the consequence of manipulating it? Knocking down its gene in the mouse BV2 microglial cell line, the authors found that the cells expressed less of several pro-inflammatory mediators, more of the lipid metabolism genes ApoE and CLU (ApoJ), and poorly phagocytosed a fungal pathogen mimic. Overexpressing the gene revved up phagocytosis.

Given that the protective haplotype lowers PU.1, the finding of impaired phagocytosis initially surprised her, Goate said. Other studies suggest that enhancing phagocytosis could benefit the AD brain (Jul 2016 news; Dec 2016 conference news). However, Goate noted that phagocytic uptake might not be the most important microglial characteristic in the context of AD. Effects on the clearance of the phagocytosed material or on microglial activation, migration, proliferation, or survival might supersede it. 

What Makes a “Good” Microglia?
In future work, Goate plans to knock down SPI1 in the microglia of adult AD mouse models and measure effects on inflammation, injury response, and amyloid deposition. She will also analyze how levels of PU.1 affect global gene expression. Butovsky speculated that lower levels of PU.1 might jolt microglia out of a homeostatic phenotype into a more active, protective mode, as has been suggested for TREM2 (Apr 2017 conference news). The authors agreed.

Before microglia can be targeted therapeutically, researchers will need to understand how to best modulate them to fend off disease, Butovsky added. “Pharma are interested in targeting these cells, but the biggest caveat is how to do it,” he said. Several current studies are exploring how microglial phenotypes change in different disease states (Jun 2017 newsJun 2017 news). “This research is revolutionizing the field,” Butovsky said.

Would modulating microglial genes such as SPI1 make a difference in disease? GWAS hits like the one at the SPI1/CELF1 locus typically nudge risk by only 10 percent or so, Goate noted. By contrast, rare TREM2 variants triple a person’s AD risk. However, the protective SPI1 haplotype is relatively common, suggesting it could influence disease risk in many people. A previous study in AD mice reported chromatin changes around SPI1, supporting the idea that its expression changes during the disease (Feb 2015 news on Gjoneska et al., 2015). Goate noted that the importance of identifying these genes rests not on their effect size or frequency, but the fact that they pinpoint critical biological pathways that may provide novel therapeutic targets and new biological insight into the disease.

Moreover, Goate believes that microglia may be power players in many other neurodegenerative diseases. She pointed out that two key Parkinson’s genes, LRRK2 and GBA, are expressed primarily in myeloid cells. The frontotemporal dementia gene progranulin may also act in microglia (Oct 2014 news). “For neurodegenerative diseases, people tend to think that all the action is in the neurons. Genetics is showing us repeatedly that that’s not true,” Goate said.—Madolyn Bowman Rogers


  1. I think this is an extremely important and insightful piece of work, which firmly puts the microglial response as key to Alzheimer pathogenesis. It builds on the work of Jones et al. (2010), as well as our own work and on both human genetics and transgenic animals (Matarin et al., 2015; Gagliano et al., 2016). 

    These together clearly show that genetic variability in how the brain responds to Aβ deposition is key to determining who gets the disease. This work perhaps serves as the genetic underpinning of the "cellular phase" of Alzheimer’s disease (De Strooper and Karran, 2016). 

    This paper has been available for several months on bioRχiv. I think this also is an important event.  Too much science is being held up by the slow reviewing process. It is great that the authors chose to make this seminal work available before acceptance. I hope this, too, turns out to be a marker for the future.


    . Genetic evidence implicates the immune system and cholesterol metabolism in the aetiology of Alzheimer's disease. PLoS One. 2010;5(11):e13950. PubMed.

    . A genome-wide gene-expression analysis and database in transgenic mice during development of amyloid or tau pathology. Cell Rep. 2015 Feb 3;10(4):633-44. Epub 2015 Jan 22 PubMed.

    . Genomics implicates adaptive and innate immunity in Alzheimer's and Parkinson's diseases. Ann Clin Transl Neurol. 2016 Dec;3(12):924-933. Epub 2016 Nov 4 PubMed.

    . The Cellular Phase of Alzheimer's Disease. Cell. 2016 Feb 11;164(4):603-15. PubMed.

  2. Huang, Goate, and colleagues made remarkable findings that underscore the importance of noncoding genetic variation in regulation of gene expression relevant to disease. In particular, it highlights a function for the myeloid transcription factor PU.1 and the microglial response as a vital component of AD pathogenesis.

    The results are consistent with our published work showing that AD-associated genetic risk variants are enriched in the noncoding regions that regulate immune response genes, suggesting that predisposition to AD is encoded in the immune system, and furthermore identifying PU.1 as a master regulator of the AD immune response (Gjoneska et al., 2015). 

    The implications of this work are very significant, as it provides novel avenues for therapeutic interventions suggesting that regulation of PU.1 expression or its relevant target genes can be used as an effective strategy for AD treatment. It is an exciting possibility that warrants further exploration.


    . Conserved epigenomic signals in mice and humans reveal immune basis of Alzheimer's disease. Nature. 2015 Feb 19;518(7539):365-9. PubMed.

  3. This multicenter study led by Alison Goate conducted a large-scale, genome-wide survival analysis on thousands of AD cases and control samples to uncover loci associated with age of onset of AD. The authors discovered a CSF Aβ42-associated SNP in the previously reported CELF1 AD risk locus. They found that this locus was significantly associated through a protective allele with reduced SPI1 (PU.1), a lineage-determining transcription factor for myeloid cells.

    Recent analysis of microglia enhancers has revealed the molecular mechanisms, which regulate microglia gene expression programs (Gosselin et al., 2014; Matcovitch-Natan et al., 2016). Active enhancers in microglia contain DNA sequences bound to the macrophage lineage-determining factor PU.1 (Heinz et al., 2010). 

    PU.1 cooperates with microglia-specific enhancer regions of the Mef2 family, which is exclusively expressed in microglia as compared to other peripheral immune cells (Butovsky et al., 2014; Matcovitch-Natan et al., 2016). This factor has been proposed as the responsible partner of PU.1 in establishing a microglia-specific molecular signature (Lavin et al., 2014). In addition, motif analysis of PU.1 binding has revealed enrichment for consensus sequences for Smad3, Mef2, and Mafb (Gosselin et al., 2014). These studies suggest that these transcription factors cooperate with PU.1 in the establishment of microglia-specific enhancer profiles. Secondary transcription factors included SMAD proteins, which are induced by TGFβ and contribute to microglia transcription of specific target genes (Butovsky et al., 2014; Gosselin et al., 2014).

    The discovery in this study is unsurprising because recently, genetic and molecular evidence obtained from brain homogenates has implicated myeloid cells in the etiology of AD. This includes Bin1, Trem2 and CD33 molecules related to phagocytic and immunomodulatory functions of myeloid cells.

    However, myeloid cells such as microglia represent a minor fraction of the tissue in these analyses. Thus, the investigators analyzed cis-eQTL effects of the AAOS-associated SNPs in human monocytes and macrophages. They discovered the key transcriptional regulator of the myeloid cells. It provides the core transcriptional regulation of microglia and blood monocytes, and may regulate the expression of multiple AD-associated genes in myeloid cells. Thus, this study provides important evidence of the implication of innate immunity in the etiology and disease progression of AD.


    . Environment drives selection and function of enhancers controlling tissue-specific macrophage identities. Cell. 2014 Dec 4;159(6):1327-40. PubMed.

    . Microglia development follows a stepwise program to regulate brain homeostasis. Science. 2016 Aug 19;353(6301):aad8670. Epub 2016 Jun 23 PubMed.

    . Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol Cell. 2010 May 28;38(4):576-89. PubMed.

    . Identification of a unique TGF-β-dependent molecular and functional signature in microglia. Nat Neurosci. 2014 Jan;17(1):131-43. Epub 2013 Dec 8 PubMed.

    . Nitric oxide prevents aortic neointimal hyperplasia by controlling macrophage polarization. Arterioscler Thromb Vasc Biol. 2014 Aug;34(8):1739-46. Epub 2014 Jun 12 PubMed.

  4. This is a very interesting finding, providing an association of PU.1 with age of onset of Alzheimer's disease. The authors provide solid evidence of the roles of the detected genetic variability on monocyte function, but I guess we are all asking ourselves: Is this informing about altered microglial function in AD? This is an exciting prospect that we can frame on existing literature.

    In this sense, we previously reported an increased expression of PU.1 in microglia in both AD brain (Gómez-Nicola et al., 2013Olmos-Alonso et al., 2016) and in a model of AD-like pathology (Olmos-Alonso et al., 2016). Although PU.1 is expressed by microglia in the healthy brain, its levels are unregulated during AD, with functions not fully understood. However, we do know that PU.1 regulates the expression of the components of the CSF1R pathway, helping to drive a prominent microglial proliferative response that can be observed in AD, both in humans and mice (Olmos-Alonso et al., 2016). It is therefore tempting to link the observed roles of PU.1 with those of CSF1R, and therefore suggest that the observed beneficial effects of targeting CSF1R (Olmos-Alonso et al., 2016; Dagher et al., 2015; Spangenberg et al., 2016) are correlative with the now-reported beneficial effects of mutation of PU.1.

    It will be exciting to see follow-up studies on the specific roles of these mutations on microglial PU.1, as this will get us closer to a full understanding of the role of these cells in AD.


    . Pharmacological targeting of CSF1R inhibits microglial proliferation and prevents the progression of Alzheimer's-like pathology. Brain. 2016 Mar;139(Pt 3):891-907. Epub 2016 Jan 8 PubMed.

    . Regulation of microglial proliferation during chronic neurodegeneration. J Neurosci. 2013 Feb 6;33(6):2481-93. PubMed.

    . Colony-stimulating factor 1 receptor inhibition prevents microglial plaque association and improves cognition in 3xTg-AD mice. J Neuroinflammation. 2015 Aug 1;12:139. PubMed.

    . Eliminating microglia in Alzheimer's mice prevents neuronal loss without modulating amyloid-β pathology. Brain. 2016 Apr;139(Pt 4):1265-81. Epub 2016 Feb 26 PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Large Genetic Analysis Pays Off With New AD Risk Genes
  2. Pooled GWAS Reveals New Alzheimer’s Genes and Pathways
  3. Alzheimer's GWAS Hits Reflected in Monocyte Gene Expression
  4. TREM2 Helps Phagocytes Gobble Up Aβ Coated in Antibodies
  5. Inflammation Helps Microglia Clear Amyloid from AD Brains
  6. New Evidence Confirms TREM2 Binds Aβ, Drives Protective Response
  7. What Makes a Microglia? Tales from the Transcriptome
  8. Hot DAM: Specific Microglia Engulf Plaques
  9. Consortium Debuts Biggest Set of Human Epigenomes To Date
  10. Does Progranulin in Microglia Protect Against Alzheimer’s?

Paper Citations

  1. . LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet. 2015 Mar;47(3):291-5. Epub 2015 Feb 2 PubMed.
  2. . Association of a functional polymorphism in the 3'-untranslated region of SPI1 with systemic lupus erythematosus. Arthritis Rheum. 2011 Mar;63(3):755-63. PubMed.
  3. . Conserved epigenomic signals in mice and humans reveal immune basis of Alzheimer's disease. Nature. 2015 Feb 19;518(7539):365-9. PubMed.

External Citations

  1. SPI1 
  2. CELF1 
  4. LD score regression
  5. Cardiogenics consortium

Further Reading

Primary Papers

  1. . A common haplotype lowers PU.1 expression in myeloid cells and delays onset of Alzheimer's disease. Nat Neurosci. 2017 Aug;20(8):1052-1061. Epub 2017 Jun 19 PubMed.