ABSTRACT Changes in the levels of circulating proteins are associated with Alzheimer’s disease (AD), whereas their pathogenic roles in AD are unclear. Here, we identified soluble ST2 (sST2),

a decoy receptor of interleukin-33–ST2 signaling, as a new disease-causing factor in AD. Increased circulating sST2 level is associated with more severe pathological changes in female

individuals with AD. Genome-wide association analysis and CRISPR–Cas9 genome editing identified rs1921622, a genetic variant in an enhancer element of _IL1RL1_, which downregulates gene and

protein levels of sST2. Mendelian randomization analysis using genetic variants, including rs1921622, demonstrated that decreased sST2 levels lower AD risk and related endophenotypes in

modulated by a genetic variation and plays a disease-causing role in females with AD. SIMILAR CONTENT BEING VIEWED BY OTHERS MULTIFACETED ROLES OF APOE IN ALZHEIMER DISEASE Article 21 June

2024 EXOME SEQUENCING IDENTIFIES RARE DAMAGING VARIANTS IN _ATP8B4_ AND _ABCA1_ AS RISK FACTORS FOR ALZHEIMER’S DISEASE Article Open access 21 November 2022 RARE VARIANTS IN _IFFO1_, _DTNB,

NLRC3_ AND _SLC22A10_ ASSOCIATE WITH ALZHEIMER’S DISEASE CSF PROFILE OF NEURONAL INJURY AND INFLAMMATION Article Open access 16 February 2022 MAIN AD is the most common neurodegenerative

disease and a leading cause of mortality in older people1. Its pathological hallmarks include the extracellular accumulation of Aβ peptides, which form Aβ plaques, and intracellular

neurofibrillary tangles composed of hyperphosphorylated tau protein (P-tau)2. While the pathophysiological mechanisms underlying AD remain unclear, genome-wide association studies (GWASs)

reveal more than 40 AD-associated genes linked with microglial functions (for example_, APOE_, _TREM2_, _BIN1_ and _CD33_), suggesting that microglia play a key role in AD pathogenesis3,4,5.

In particular, _APOE_-ε4, the strongest known risk factor for sporadic AD after chronological age3, affects Aβ accumulation in AD6,7 through regulating the clustering of microglia around Aβ

and the subsequent degradation of Aβ plaques4,8,9. This suggests that microglial dysfunction plays an essential causative role in AD. Recent studies show that, besides genetic factors,

changes in secreted proteins in the brain milieu and/or circulatory system may also disrupt microglial activities and contribute to AD pathogenesis10,11. For example, soluble TREM2 protein

(sTREM2) level is increased in the cerebrospinal fluid (CSF) of individuals with AD12,13,14, and injection of sTREM2 in transgenic mouse models of amyloidosis alleviates Aβ accumulation by

enhancing the interaction between microglia and Aβ and subsequent Aβ phagocytosis15,16. Moreover, the soluble form of a full-length VCAM1 protein in endothelial cells, sVCAM1, is increased

in the plasma and CSF of individuals with AD17,18; this is suggested to mediate reduced hippocampal neurogenesis and a pro-inflammatory response by microglia during aging19. Other secreted

proteins that contribute to AD pathogenesis include soluble cytokine receptors comprising the ectodomains of membrane-bound cytokine receptors, which function as decoy receptors to attenuate

cytokine-mediated signaling20,21. In particular, sST2 is a secreted isoform of the interleukin-33 (IL-33) receptor ST2L (full-length ST2) that is produced by alternative promoter

activation22 (Extended Data Fig. 1). ST2L is expressed by microglia in the brain23, and activation of IL-33–ST2 signaling decreases Aβ accumulation via enhanced microglial Aβ-clearance

capacity in transgenic mouse models of amyloidosis24,25. Meanwhile, sST2 acts as a decoy receptor of IL-33 and inhibits IL-33–ST2 signaling26,27. Altered sST2 level in plasma is a biomarker

of several inflammatory and cardiac diseases28,29,30,31. Notably, recent evidence shows that sST2 levels are also elevated in the blood of individuals with mild cognitive impairment (MCI) or

AD24,32. Nonetheless, it is unclear what regulatory mechanisms underlie sST2 dysregulation and whether sST2 plays a pathological role in AD. In the present study, we investigated sST2

regulation in AD and its roles in disease pathogenesis. We showed that sST2 level increases in the blood and brains of females with AD and is positively associated with disease progression.

Moreover, we identified a single-nucleotide polymorphism (SNP), rs1921622, in _IL1RL1_ (the gene encoding sST2 and ST2L) that is associated with decreased sST2 expression in human

endothelial cells and decreased plasma and CSF sST2 levels. Mendelian randomization (MR) analysis showed that, in female _APOE_-ε4 carriers, decreased sST2 level results in reduced AD risk

and less-severe AD-associated endophenotypes, suggesting a causal effect of sST2 in AD. Subsequent single-nucleus transcriptomic profiling revealed that the rs1921622 A allele is associated

with enhanced microglial activation toward Aβ and lowered Aβ plaque load in female _APOE_-ε4 carriers with AD. Together, our findings suggest that a circulating protein, sST2, in the brain

milieu plays a key role in _APOE_-ε4-dependent AD pathogenesis in females—by modulating the activation and Aβ-clearance capacity of microglia—and therefore might be a novel target for AD

therapy. RESULTS SST2 IS ASSOCIATED WITH ALZHEIMER’S DISEASE AND ITS PATHOLOGICAL CHANGES To investigate how sST2 is involved in AD pathogenesis, we examined the associations between sST2

level and AD and its related endophenotypes. We measured the plasma sST2 level of Chinese individuals with AD (that is, those having Alzheimer’s dementia with a Montreal Cognitive Assessment

(MoCA) score < 21) and healthy controls (HCs) recruited in Hong Kong (the ‘Chinese_cohort_1’ hereafter; _n_ = 345 HCs and _n_ = 345 individuals with AD; Supplementary Table 1). We then

performed a linear regression analysis between plasma sST2 level and AD and its related endophenotypes, adjusting for age, sex, status of cardiovascular diseases (CVDs; that is, heart

disease, hypertension, diabetes mellitus and hyperlipidemia), body mass index (BMI) and education level, followed by multiple testing correction. The results show that plasma sST2 level was

higher in individuals with AD when compared to HCs (_β_ (effect size) = 2.072, false discovery rate (FDR) < 0.001; Fig. 1a). Moreover, plasma sST2 level was positively associated with the

AD-related endophenotypes we examined—namely, the decreased volumes of gray matter (_β_ = −0.695, FDR = 0.003; Fig. 1b) and increased levels of plasma biomarkers corresponding to AD (that

is, P-tau181 (ref. 33), _β_ = 0.413, FDR = 0.005; Fig. 1c) and neurodegeneration (that is, neurofilament light polypeptide (NfL)34, _β_ = 0.107, FDR = 0.004; Fig. 1d). Notably, we found a

differential regulation of plasma sST2 levels between sexes: while plasma sST2 level is lower in females than that in males, it exhibits a greater increase in AD in females (females, _β_ = 

2.235, FDR = 6.97 × 10−4; males, _β_ = 1.926, FDR = 0.148; Extended Data Fig. 2a). In particular, female _APOE_-ε4 carriers with AD had the greatest increase of plasma sST2 level compared to

that in other subgroups (_β_ = 3.833, FDR = 0.004; Extended Data Fig. 2b,c and Supplementary Table 2). Moreover, females showed stronger associations between plasma sST2 level and

AD-related endophenotypes (including gray matter volumes, plasma P-tau181 levels and plasma NfL levels) than those in males (Supplementary Table 2). These results suggest that plasma sST2

level is associated with AD and its related endophenotypes in a female-specific manner. We further showed that plasma and CSF sST2 levels are positively correlated within the same individual

(_R__2_ = 0.130, _P_ < 0.0001; _n_ = 107 individuals from the Stanford Alzheimer’s Disease Research Center (ADRC) cohort35; Fig. 1e). Accordingly, we performed a linear regression

analysis between CSF sST2 level and AD and Aβ plaque load, adjusting for age, sex and postmortem duration (PMD), followed by multiple testing correction. Concordant with the regulation of

plasma sST2 level in AD, in the UK Brain Banks Network (UKBBN) cohort (_n_ = 11 HCs, _n_ = 75 individuals with AD; Supplementary Data 1), the CSF sST2 level was higher in individuals with AD

than in HCs (_β_ = 6.605, FDR < 0.01; Fig. 1f), with a greater increase in females than that in males (females, _β_ = 7.766, FDR = 0.034; males, _β_ = 4.019, FDR = 0.241; Extended Data

Fig. 3a). Moreover, CSF sST2 level was positively associated with Aβ plaque load in the frontal cortex in individuals with AD (_β_ = 0.191, FDR = 0.031; Fig. 1g,h and Supplementary Fig. 1);

this association was also stronger in females than in males (females, _β_ = 0.219, FDR = 0.041; males, _β_ = 0.026, FDR = 0.787; Extended Data Fig. 3b). Taken together, these findings

suggest that sST2 levels in both the blood and CSF are increased in AD and associated with disease progression, specifically in female populations. SST2 LEVEL IS ASSOCIATED WITH A GENETIC

VARIANT OF _IL1RL1_ To understand how sST2 is regulated in AD, we examined the contribution of various factors to the changes of sST2 levels. Specifically, we examined the associations

between plasma/CSF sST2 levels and sex/age in three independent cohorts including Chinese_cohort_1, comprising HCs and individuals with AD, and a Japanese cohort36 and a European-descent

cohort (that is, INTERVAL and LonGenity cohorts37) comprising cognitively normal individuals. Our results show that plasma sST2 level was lower in females than in males (_β_ = −3.577, _P_ 

< 0.001) and, compared to males (Chinese, _R_2 = 0.001, _P_ = 0.866; European descent, _R_2 = 0.010, _P_ = 0.009), females show a greater association between plasma sST2 level and age

(Chinese, _R_2 = 0.037, _P_ = 0.002; European descent, _R_2 = 0.020, _P_ < 0.001; Extended Data Fig. 4a–c). Furthermore, CSF sST2 level was associated with both age and sex in the

Japanese cohort, and the association with age, specifically, was greater in females (_R__2_ = 0.152, _P_ = 0.001) than in males (_R_2 = 0.044, _P_ = 0.081; Extended Data Fig. 4d). These data

suggest that sST2 level is modulated by age and sex in both East Asian and European-descent populations. However, factors such as age and sex accounted for only 6.92% and 13.01% of the

variance of plasma sST2 levels and CSF sST2 levels, respectively (Extended Data Fig. 4e,f), suggesting that other factors modulate such differences. Pilot association studies have identified

several genetic variants in the _IL1RL1_ gene that are associated with plasma sST2 level38,39, suggesting that genetic factors contribute to the regulation of sST2 level. Nonetheless, given

that these identified SNPs form haplotype structures in this gene region40, these SNPs might simply be inherited together with the causal variants. Therefore, to identify the key genetic

modulator(s) of sST2, we used our whole-genome sequencing (WGS) dataset41 of Chinese_cohort_1 to perform a GWAS of plasma sST2 levels, adjusting for age, sex, AD diagnosis and population

structure. Accordingly, we identified 575 genetic variants that were significantly associated with plasma sST2 level (_P_ < 1 × 10−5) and found that these variants accounted for 54.86% of

the variation thereof (Fig. 2a and Supplementary Fig. 2a). Among these 575 variants, 79 in or near _IL1RL1_ that form a haplotype were most strongly associated with sST2 level (Fig. 2b,

Supplementary Fig. 2b and Supplementary Data 2), and our fine-mapping analysis identified the sentinel variant rs1921622 (G/A) as the putative causal variant (with 99.9% probability;

Supplementary Fig. 2c and Supplementary Table 3). In Chinese_cohort_1, the rs1921622 A allele was associated with a 20% decrease in plasma sST2 level in an allele dose-dependent manner (_β_ 

= −3.346, _P_ < 0.001; Fig. 2c). Moreover, CSF sST2 level was lower in rs1921622 A allele carriers than in noncarriers (UKBBN cohort; _β_ = −2.244, _P_ < 0.05; Fig. 2d). Notably,

rs1921622 alone accounted for 18.04% and 18.29% of the variance in plasma and CSF sST2 levels, respectively, which is greater than the contributions of age and sex. Hence, our fine-mapping

analysis using WGS data identified rs1921622 as a key genetic factor that modulates the plasma and CSF levels of sST2. THE RS1921622 LOCUS REGULATES SST2 IN BRAIN ENDOTHELIAL CELLS As

rs1921622 is a noncoding variant located in the intronic region of ST2L, which is downstream of the region encoding sST2, we examined whether it modulates the expression of sST2 and ST2L.

Association analysis using the Genotype-Tissue Expression (GTEx) dataset42,43 showed that, compared to noncarriers, rs1921622 A allele carriers exhibited a lower transcript level of sST2,

but not ST2L, in certain brain regions (for example, hippocampus and frontal cortex; _P_ < 0.05; Fig. 3a and Supplementary Table 4). Furthermore, an analysis of our previously released

human frontal cortex single-nucleus RNA-sequencing (snRNA-seq) dataset44 revealed that sST2 is exclusively expressed by endothelial cells (that is, _CLDN5_-expressing cells; Fig. 3b,c). In

addition, cell-type-specific association analysis showed that, compared to noncarriers, rs1921622 A allele carriers had a lower endothelial cell sST2 transcript level (_P_ < 0.01) and

fewer sST2-expressing endothelial cells (_P_ < 0.05; Fig. 3d). These results collectively indicate that the rs1921622 variant is associated with decreased sST2 expression in human brain

endothelial cells. To investigate the specific role of rs1921622 in the decreased sST2 expression in endothelial cells, we examined whether rs1921622 and the surrounding genomic region (Fig.

3e) regulate sST2 transcription. Given that noncoding variants commonly modulate gene expression by functioning as enhancer elements45, we examined enhancer activity at the rs1921622 locus

in the human cerebral microvascular endothelial cell line (hCMEC/D3). Administration of the cytokine IL-33 increased the expression and secretion of sST2 in hCMEC/D3 cells (both _P_ < 

0.001; Supplementary Fig. 3a,b). Moreover, chromatin immunoprecipitation (ChIP) assay showed that these IL-33-treated hCMEC/D3 cells exhibit increased occupancy of an active enhancer histone

mark (that is, acetylated histone H3 Lys27 (H3K27ac)) at the rs1921622 locus, with a concomitant higher level of trimethylated histone H3 Lys4 (H3K4me3) histone modification (indicating

active promoter regions) at the sST2 promoter region (both _P_ < 0.05; Fig. 3f and Supplementary Fig. 3c). These results suggest that rs1921622 is located at a potential enhancer element

of the sST2 gene. To demonstrate that the genomic region harboring the rs1921622 locus contributes to the regulation of sST2 expression, we deleted this region in hCMEC/D3 cells with a

CRISPR–Cas9-based approach. We generated two different hCMEC/D3 cell lines with biallelic 38- or 67-base-pair (bp) deletions (Δ38 bp and Δ67 bp, respectively) encompassing the rs1921622

locus (Fig. 3e and Supplementary Fig. 4). Notably, loss of the 38 or 67 bp flanking the rs1921622 locus decreased the sST2 transcript level in hCMEC/D3 cells (_P_ < 0.001; Fig. 3g) and

concomitantly abolished sST2 protein secretion (_P_ < 0.001; Fig. 3h). These results suggest that the rs1921622-containing region plays a putative regulatory role as an enhancer element

that regulates sST2 expression in endothelial cells. THE RS1921622 A ALLELE PROTECTS AGAINST ALZHEIMER’S DISEASE IN FEMALE _APOE_-Ε4 CARRIERS Dysregulation of a specific protein in a disease

may indicate that the protein is a causative factor of the disease, or its altered level may simply be due to tissue damage/reaction. Notably, for a protein that is regulated by genetic

variants, MR analysis examining the associations between the protein with its genetic modulators and a disease would help illustrate whether the protein has disease-causing effects46.

Therefore, we investigated the causality between sST2 level and AD risk by examining the associations between the genetic modulators of sST2 (including rs1921622) and AD risk and its related

endophenotypes. Specifically, we performed the two-sample MR analysis of sST2 on AD risk using eight AD datasets as the discovery cohorts: the Chinese_cohort_1 dataset, the WGS and array

datasets of the Chinese_cohort_2 (ref. 47) and five public datasets from populations of European descent (that is, the Late Onset Alzheimer’s Disease (LOAD)48, Alzheimer’s Disease Center 1–3

(ADC1–3)49,50 and Alzheimer’s Disease Neuroimaging Initiative (ADNI) datasets; _n_ = 5,477 HCs, _n_ = 5,910 individuals with AD; Supplementary Tables 1 and 5). The results show that while

sST2 level was not associated with AD risk in all individuals (Chinese, _β_ = −0.023, FDR = 0.861; European descent, _β_ = 0.077, FDR = 0.131; Fig. 4a and Supplementary Table 6), increased

sST2 level was associated with increased AD risk in female _APOE-_ε4 carriers in both Chinese and European-descent populations (Chinese, _β_ = 0.772, FDR = 1.70 × 10−5; European-descent, _β_

 = 0.168, FDR = 0.023; Fig. 4a,b and Supplementary Table 6). In particular, among the genetic modulators of sST2, the top variant, the rs1921622 A allele, exerted an AD-protective effect in

female _APOE-_ε4 carriers (odds ratio (OR) = 0.757, meta-analysis _P_ value using Han and Eskin’s random-effects model (RE2 _P_)51 = 7.78 × 10−5, FDR = 5.44 × 10−4; Fig. 4c), whereas this

effect is absent in the overall population (OR = 0.942, RE2 _P_ = 0.044, FDR = 0.102) or other subgroups including male _APOE-_ε4 carriers (OR = 1.086, RE2 _P_ = 0.412, FDR = 0.577), male

_APOE-_ε4 noncarriers (OR = 0.960, RE2 _P_ = 0.586, FDR = 0.684) and female _APOE-_ε4 noncarriers (OR = 0.961, RE2 _P_ = 0.233, FDR = 0.408; Supplementary Tables 7 and 8). Interestingly,

both the MR analysis of sST2 and the meta-analysis of rs1921622 showed that the associations between AD risk and sST2/rs1921622 in female _APOE-_ε4 carriers are stronger in the Chinese

population than those in the European-descent populations (Chinese versus European descent, _P_ = 5.55 × 10−4 and 0.017 for the MR analysis of sST2 and meta-analysis of rs1921622,

respectively; Fig. 4a–c and Supplementary Tables 6–8), suggesting an ethnic-specific effect of sST2/rs1921622 in modulating AD risk. Next, we examined the effects of the rs1921622 A allele

on the AD-associated endophenotypes in the discovery cohorts. Consistent with its AD-protective effects (Supplementary Table 9), the rs1921622 A allele was associated with delayed onset age

of dementia (hazard ratio (HR) = 0.874, FDR = 0.011; Fig. 5a), better cognitive scores (_β_ = 1.622, FDR = 0.001; Fig. 5b) and larger entorhinal cortex volumes (_β_ = 0.214, FDR = 0.027;

Fig. 5c) in female _APOE-_ε4 carriers. To further confirm the AD-protective effects of the rs1921622 A allele, we examined AD-related endophenotypes in an independent replication cohort: the

Australian Imaging, Biomarkers and Lifestyle cohort (AIBL; _n_ = 190), in which the Aβ deposition (that is, Aβ+) in the brains of individuals was confirmed by positron emission

tomography52. Concordant with our findings from the discovery cohorts, in female Aβ+ _APOE-_ε4 carriers, the presence of the rs1921622 A allele was associated with improved cognitive

performance as indicated by AIBL Preclinical Alzheimer Cognitive Composite (AIBLPACC) score and scores related to cognitive subprocesses including attention processing, episodic recall and

recognition (all FDR < 0.05; Fig. 5d and Extended Data Fig. 5). Importantly, in a subgroup of female Aβ+ _APOE-_ε4 carriers whose gray matter volume was traced for 7 years, rs1921622 A

allele carriers showed a slower progression of gray matter atrophy (_β_ = −0.027) than that in noncarriers (_β_ = −0.159; FDR < 0.05; Fig. 5e and Extended Data Fig. 6). Thus, these

results validate the protective effects of the rs1921622 A allele against cognitive decline and gray matter atrophy among female _APOE-_ε4 carriers in an independent Aβ+ cohort. THE

RS1921622 A ALLELE IS ASSOCIATED WITH MICROGLIAL ACTIVATION IN FEMALE _APOE_-Ε4 CARRIERS WITH ALZHEIMER’S DISEASE Given our findings that the rs1921622 A allele exerts AD-protective effects

in female _APOE-_ε4 carriers, we subsequently examined whether this variant modulates Aβ deposition in postmortem human brains. Among individuals with AD, specifically in females, the

_APOE_-ε4 carriers exhibited greater Aβ deposition in the frontal cortex than that in the _APOE_-ε4 noncarriers (females, _β_ = 1.967, _P_ < 0.05; males, _β_ = −0.621, _P_ = 0.422;

Extended Data Fig. 7a,b). However, upon further stratification by rs1921622 genotype, female _APOE_-ε4 carriers harboring the rs1921622 A allele exhibited less Aβ deposition than those

without the allele (_β_ = −2.262, _P_ < 0.05; Fig. 6a,b and Extended Data Fig. 7c), suggesting that this allele attenuates the effects of _APOE_-ε4 on Aβ-related pathological changes in

females. Moreover, immunohistochemical analysis revealed that harboring _APOE_-ε4 in AD is associated with less microglial coverage of Aβ plaques (that is, decreased proportion of Aβ

colocalized with Iba-1+ microglia) in females (females, _β_ = −2.818, _P_ < 0.01; males, _β_ = −0.965, _P_ = 0.327; Extended Data Fig. 7d,e). In contrast, female _APOE_-ε4 carriers with

AD who co-harbored the rs1921622 A allele showed increased colocalization between Iba-1+ microglia and Aβ plaques (_β_ = 2.017, _P_ < 0.05; Fig. 6c,d and Extended Data Fig. 7f,g). Thus,

these results suggest that in female _APOE-_ε4 carriers with AD, the rs1921622 A allele is associated with enhanced microglia–Aβ interaction and decreased Aβ pathological lesions. Next, to

investigate the regulatory effects of rs1921622 on microglial activities at the molecular level, we conducted an association analysis using the microglial snRNA-seq dataset in the frontal

cortices44. We found a negative correlation between the effects of the rs1921622 A allele and the effects of CSF sST2 level on modulating microglial gene expression in female _APOE_-ε4

carriers with AD (_R__2_ = 0.860, _P_ < 0.0001; Extended Data Fig. 8), supporting the notion that the variant exerts its modulatory effects on microglia through the regulation of CSF sST2

level. Specifically, we identified 1,639 microglial genes that were associated with rs1921622 genotype: 428 and 1,211 genes were upregulated and downregulated, respectively, in individuals

carrying the rs1921622 A allele compared to noncarriers (FDR < 0.05; Fig. 6e). Gene Ontology (GO) analysis showed that those upregulated genes are associated with leukocyte migration (FDR

 = 1.2 × 10−4) and innate immune response (FDR = 5.4 × 10−3), whereas those with downregulated expression are mainly involved in protein refolding (FDR = 8.1 × 10−3) or mRNA splicing (FDR = 

7.7 × 10−2; Fig. 6f). Recent studies of microglia in mouse and human brains revealed a subset of ‘microglial activation genes’, including _CD74_, _APOE_ and _TREM2_, whose expressions are

upregulated in AD53,54 and which are involved in microglial Aβ phagocytosis4,55,56. Therefore, we investigated whether these genes are associated with rs1921622. Interestingly, among female

_APOE_-ε4 carriers with AD, the rs1921622 A allele was associated with increased expression of these microglial activation genes—specifically, increased transcript levels of _CD74_, _APOE_

and _TREM2_ in microglia as well as an increased proportion of _TMEM163_+ microglia (Fig. 6g). In contrast, the rs1921622 A allele was associated with decreased expression of homeostatic

genes including _SRGAP2_, _TMEM119_ and _P2RY12_, whose expressions commonly indicate a less-reactive microglial state4,55 (Fig. 6g). Therefore, these results suggest that the rs1921622 A

allele promotes the transition of microglia to a more activated state in female _APOE_-ε4 carriers with AD. SST2 TREATMENT IMPAIRS MICROGLIAL AΒ-CLEARANCE CAPACITY IN FEMALE MOUSE BRAINS Our

genetic analyses showed that sST2 and its genetic modulators modulate AD risk and related pathologies possibly by regulating microglial activation and Aβ clearance. To verify the pathogenic

roles of sST2 in AD, we examined the effect of sST2 injection on Aβ-associated pathological changes in an amyloidosis transgenic 5XFAD mouse model (Fig. 7a). Intracerebroventricular

administration of mouse sST2 to 3-month-old 5XFAD mice for 28 d increased the Aβ plaque burden in the cortical regions of female (_P_ < 0.05; Fig. 7b,c) but not male transgenic mice

(Extended Data Fig. 9). In particular, the quantities of both the filamentous form (that is, X-34+ diffuse fibrils without a dense core57) and the compact form (that is, X-34+ dense cores

with 4G8 halos) of Aβ plaques were increased in sST2-treated female 5XFAD mice (_P_ < 0.05; Fig. 7d,e and Supplementary Fig. 5). However, sST2 administration did not affect the burden of

less-toxic inert Aβ plaques (that is, X-34+ dense cores without 4G8 labeling) or the total number of Aβ plaques. Next, we examined how sST2 regulates the interactions between microglia and

Aβ plaques. In female 5XFAD mice, sST2 injection increased the total number of microglia, the clustering of microglia around Aβ plaques, and the number of proliferating (Ki67+) microglia in

the cortical regions (all _P_ < 0.05; Supplementary Fig. 6), suggesting that sST2 treatment promotes the proliferation of microglia in the brain. However, compared to the vehicle-injected

condition, the cortical regions in sST2-injected female 5XFAD mice showed reduced microglial coverage of Aβ plaques (Fig. 7f,g, Supplementary Fig. 7 and Supplementary Video 1), suggesting

that sST2 injection inhibits the microglial barrier formation around Aβ plaques. To verify the inhibitory effects of sST2 on microglial Aβ clearance, flow cytometry analysis showed that sST2

administration decreases the Aβ-phagocytic capacity of microglia as indicated by a decrease in the percentage of Aβ+ microglia (that is, 23.57% methoxy-X04+CD11b+ microglia in the

sST2-treated group versus 32.79% in the control group; _P_ < 0.01; Fig. 7h,i and Supplementary Fig. 8) and an 11.3–17.0% decrease in Aβ uptake by microglia (_P_ < 0.05; Extended Data

Fig. 10). Together, these findings demonstrate that increased brain sST2 level impairs microglial Aβ-clearance capacity and exacerbates Aβ accumulation in female transgenic mice, suggesting

its pathogenic roles in female AD. DISCUSSION Here, we report that an endothelial cell-secreted protein, sST2, is upregulated in female individuals with AD. Meanwhile, endothelial expression

of sST2 can be downregulated by the rs1921622 A allele, a genetic variant that exerts an AD-protective effect specifically in female _APOE_-ε4 carriers, a subgroup of people accounting for

24.3–33.6% and 32.5–50.4% of individuals with AD in Chinese and European-descent populations, respectively (Supplementary Tables 1 and 5). This variant is also associated with enhanced

microglial activation, increased microglia–Aβ plaque interaction and decreased Aβ deposition. Taken together, we demonstrate that sST2 regulates microglial activities and that changes in

sST2 levels in the brain milieu impair microglial Aβ-clearance capacity, thus modulating AD risk and its pathological changes. As sST2 only comprises the extracellular domain of the

functional receptor ST2L and is independently transcribed22, it is a decoy receptor for IL-33–ST2 signaling. ST2L expression is mainly found in microglia in the brains of both humans and

mice58,59,60, and IL-33–ST2 signaling regulates microglial activities in tissue repair, Aβ clearance and synapse engulfment24,25,61,62. The expression profile of sST2 differs between humans

and mice: in humans, sST2 is expressed by brain and peripheral endothelial cells63,64,65 and monocytes/macrophages60,66, mast cells67 and T cells68 in peripheral systems; while in mice, sST2

is expressed by microglia in the brain59 and fibroblasts69, mast cells67 and T cells68 in peripheral systems. However, given that sST2 is a secreted protein in the circulatory system24,70,

increased sST2 level likely impacts microglial functions and AD-related pathological changes in both humans and mice by blocking the binding of IL-33 to ST2L on microglia. In this study,

sST2 treatment perturbed the interaction between microglia and Aβ as well as the subsequent Aβ phagocytosis in a mouse model of amyloidosis (Fig. 7). This finding is consistent with our

previous observation that in these mice, IL-33 administration activates IL-33–ST2 signaling and initiates microglial chemotaxis toward Aβ plaques, subsequently enhancing Aβ

phagocytosis24,25. Further corroborating the pathological role of sST2 in microglia in AD, among female _APOE_-ε4 carriers with AD, those carrying the rs1921622 A allele (who have lower sST2

levels) exhibit enhanced microglia–Aβ plaque interactions and a smaller Aβ plaque area than noncarriers. Thus, perturbation of endogenous IL-33–ST2 signaling by sST2 may lead to impaired

microglial chemotaxis, Aβ uptake and barrier formation, which subsequently contribute to AD pathogenesis. What are the regulatory mechanisms of sST2? In mice, sST2 expression is regulated by

a proximal promoter located upstream of the 5′ untranslated region of sST2, which is increased by oncogenes, serum and other mitogenic stimuli69. While the regulatory mechanisms of sST2

expression might be different between humans and mice, sST2 regulation in humans remains largely unknown. Our analyses in Chinese and European-descent populations corroborate findings from

previous studies that sST2 levels in humans are associated with age, sex and genetic variants37,38, and reveal that genetic components play a dominant role in the regulation of sST2 levels,

accounting for 54.86% of the variance of these levels. In particular, our GWAS and fine-mapping analyses identified rs1921622 as a putative causal variant associated with sST2 (causal

probability of 99.99%), and our ChIP assay and CRISPR–Cas9 genome editing verified that the rs1921622-containing region is an enhancer element that regulates the sST2 gene in endothelial

cells. Thus, as rs1921622 is a key genetic modulator of sST2, future investigations of the epigenetic events at the rs1921622 locus may help elucidate the regulatory mechanisms of sST2.

Single-nucleus transcriptomic profiling of postmortem human brains shows that sST2 is mainly expressed in brain endothelial cells (Fig. 3). Dysregulation of sST2 in the brain exacerbates

AD-associated pathological changes, specifically by impairing the phagocytic capacity of microglia. Thus, we demonstrate a new pathogenic role of the brain vasculature in AD that regulates

microglial activities via sST2. Indeed, apart from Aβ and tau-related pathological changes, neurovascular dysfunction occurs early in AD and is implicated in its pathogenesis71. While the

exact pathological functions of this brain vasculature remain unclear, vasculature-secreted cytokines (for example, CXCL1) in the peripheral system regulate the activation and migration of

immune cells, thereby mediating the immune response in tissues72. Similarly, emerging studies suggest that the brain vasculature serves as a critical source of soluble inflammatory proteins

such as IL-1β, IL-6, IL-8, tumor necrosis factor, transforming growth factor and monocyte chemoattractant protein-1 (refs. 73,74). Therefore, besides sST2, other soluble factor-based

cross-talk between the vasculature and other cell types likely occurs in the brain. Accordingly, identifying the proteins involved in this cross-talk may provide insights into novel

pathological mechanisms of AD. We show that sST2 plays a disease-causing role in AD in a female-specific manner. Indeed, emerging evidence suggests that sex is an important factor that

underlies the heterogeneity of AD;75 compared to males, females are more likely to develop AD76 and experience faster cognitive decline after an MCI or AD diagnosis76,77. Given that sST2

inhibits IL-33–ST2 signaling and microglial activities, this sex-specific causal effect of sST2 in AD may be attributed to the differential regulation of IL-33–ST2 signaling and differential

microglial activities between sexes. Concordant with previous studies78, we show that microglia in female individuals with AD, particularly those who carry _APOE_-ε4, have significantly

lower Aβ phagocytic capacity when compared to male individuals (Extended Data Fig. 7). Meanwhile, compared to males, females show a greater increase of sST2 upon AD development. Hence, it

would be interesting to examine whether increased sST2 level in female brains exhibits a stronger inhibitory effect on microglial activities and impairs their Aβ-clearance capacity.

Moreover, sex hormones (for example, testosterone) regulate IL-33 expression in mast cells, maintaining a higher activity of IL-33–ST2 signaling in men79, which might protect men from being

affected by increased sST2 levels. Together, increased sST2 levels, resulting from aging or the absence of the rs1921622 A allele, might have more detrimental impacts in microglial

activities and functions in females, eventually leading to a greater risk of AD. Therefore, investigating the underlying mechanisms of this female-specific, disease-causing role of sST2 may

be key to understanding the differential progression of AD between sexes. In addition to its female-specific effects, our genetic analyses demonstrate that sST2/rs1921622 plays

disease-causing effects in females with AD, specifically those carrying _APOE_-ε4, suggesting a potential interaction between IL-33–ST2 signaling and ApoE. In the brain, ApoE is mainly

produced by astrocytes, and its expression is upregulated in microglia in AD53,80. Single-cell RNA-seq of amyloidosis mouse models revealed that a microglial subpopulation transitions from a

homeostatic state to an activated state termed ‘disease-associated microglia’ or ‘activated response microglia’4,55, with increased expression of microglial activation genes (including

_APOE_, _AXL_, _TREM2_ and _CD74_)53,54,55,56 that are crucial regulators of phagocytic processes81,82,83,84. In contrast, perturbation of ApoE functions in the brain abolishes the induction

of this microglial activation state, resulting in reduced Aβ phagocytic capacity4,55. Thus, ApoE-mediated microglial activation may protect against AD and be required for Aβ clearance and

brain homeostasis. Notably, the presence of the rs1921622 A allele modulates the transition of microglia from a homeostatic state to a more activated state in female _APOE_-ε4 carriers with

AD, characterized by the increased expression of the aforementioned microglial activation genes (Fig. 6). Therefore, IL-33–ST2 signaling and ApoE might converge to regulate the expression of

these specific genes in microglia, thereby modulating the activation state and Aβ-clearance capacity of microglia. Interestingly, in the periphery, IL-33 administration ameliorates the

formation of macrophage foam cells (lipid-laden macrophages) and the development of atherosclerotic plaques in the ApoE−/− model of atherosclerosis85. Therefore, it would be of interest to

determine whether the modulation of IL-33–ST2 signaling reduces the detrimental effects of _APOE_-ε4 on Aβ accumulation through the regulation of lipid metabolism in AD. Our findings suggest

that sST2 is a promising therapeutic target for AD. First, as sST2 is mainly expressed by endothelial cells, this enables cell-type-specific manipulation of sST2 expression, and that

manipulation may not need to cross the blood–brain barrier. Second, sST2 levels are elevated in individuals with MCI or early-stage AD24,32, suggesting the potential applicability of sST2 in

early intervention strategies. Third, the deletion of the rs1921622 locus in a human brain endothelial cell line with high efficacy demonstrates the feasibility of this approach to silence

sST2 expression and secretion without disrupting the activities of ST2L; this is because the epigenetic and transcriptional controls of sST2 are distinct from those of ST2L86, and the

rs1921622 A allele only modulates sST2 but not ST2L expression. Fourth, given our finding that the rs1921622 A allele is a common AD-associated variant, manipulations of sST2 by targeting

this genetic variant could be developed for specific subgroups of individuals who have high sST2 levels (for example, female individuals who carry _APOE_-ε4 but not the rs1921622 A allele,

accounting for 6.2–12.2% of individuals with AD), enabling patient stratification and precision medicine. Nonetheless, there are some knowledge gaps regarding the functions and regulation of

sST2 in AD. First, while MR analysis of cross-sectional studies shows that sST2 plays a causal role in AD, longitudinal studies investigating the associations between sST2 level and

cognitive decline and/or AD risk in cognitively normal populations would help support the notion that sST2 has disease-causing effects in AD. Meanwhile, given that such effects are only

observed in female _APOE_-ε4 carriers, they are likely masked in recent genetic studies of AD, in both the overall population (_P_ = 0.881)87 and _APOE_-ε4 carriers (_P_ = 0.798)88.

Therefore, stratification of AD datasets according to sex and _APOE_-ε4 genotypes may help to better investigate the disease effects of sST2 or other AD-associated factors. Moreover, as we

observed that sST2/rs1921622 has stronger modulatory effects on AD risks in the Chinese population than in the European-descent populations, validation of this finding by future studies

using larger datasets would help clarify this potential ethnic-specific effect of sST2. Lastly, while an increased brain level of sST2 contributes to AD pathogenesis, given the positive

correlation between plasma and CSF sST2 levels, it would be interesting to examine whether peripheral sST2, produced by peripheral endothelial cells65, monocytes/macrophages60,66, T cells68

and mast cells67, is blood–brain barrier permeable and plays a pathogenic role in AD. Indeed, recent studies show that several drugs (for example, sacubitril/valsartan) can reduce peripheral

sST2 levels in individuals with heart failure89. It is of interest to examine whether these drugs can also regulate peripheral sST2 levels in individuals with AD and ameliorate the

disease-related pathological changes. In summary, we uncovered an alternative pathogenic mechanism of AD that involves microglial dysfunctions mediated by sST2. Dysregulation of endothelial

cell-secreted sST2 leads to increased plasma and CSF levels of sST2 and impairs Aβ clearance by microglia, resulting in increased Aβ accumulation in AD. Furthermore, we identified an

AD-protective genetic variant, rs1921622, which downregulates sST2 expression and attenuates the _APOE_-ε4-related risk and pathological changes of AD through the regulation of microglial

signaling. Thus, a better understanding of how sST2—a biomarker and potential drug target for AD—is genetically regulated can aid the design of AD intervention strategies and clinical

trials. METHODS RECRUITMENT FOR CHINESE_COHORT_1 This study was approved by the Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee at the

Prince of Wales Hospital, the Chinese University of Hong Kong and the Hong Kong University of Science and Technology (HKUST). All individuals provided written informed consent for both study

enrollment and sample collection. Specifically, we recruited a total of 690 Hong Kong Chinese individuals ≥ 60 years old, including 345 individuals with AD and 345 HCs, who visited the

Specialist Outpatient Department of the Prince of Wales Hospital at the Chinese University of Hong Kong from April 2013 to February 2018. We established a clinical diagnosis of AD based on

the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition90. All individuals underwent a medical history assessment, clinical assessment,

cognitive and functional assessments using the MoCA test, and neuroimaging assessment by MRI91,92, and only those with cognitive dysfunctions (that is, those having Alzheimer’s dementia with

MoCA score < 21) were included in the AD group in this study. Individuals with any notable neurological disease besides AD or a psychiatric disorder were excluded. We recorded age, sex,

years of education, medical history and history of CVDs. We prepared DNA and plasma samples from whole-blood samples and stored them at −80 °C until use. We used T1-weighted MRI to retrieve

brain imaging data from 192 individuals (_n_ = 77 individuals with AD, _n_ = 115 HCs) from the Prince of Wales Hospital. Raw imaging files were deidentified and processed by AccuBrain IV1.2

(BrainNow Medical Technology) for the analysis of gray matter volumes. UK BRAIN BANKS NETWORK DATASET We obtained the following samples of the Medical Research Council (MRC) UKBBN dataset

(Bristol Brain Bank): CSF samples, frontal cortex sections, frozen frontal cortex tissues and genomic DNA samples (Supplementary Data 1), from South West Dementia Brain Bank (SWDBB), which

receives approval from North Somerset and South Bristol Research Ethics Committee to operate as a research tissue bank. For our initial sample selection from the UKBBN dataset, we excluded

individuals with other neurodegenerative diseases, severe vascular diseases (for example, stroke), an intoxicated state, infection, prions, severe inflammatory diseases (for example,

autoimmune diseases), structural brain disorders, metabolic/nutritional diseases, trauma, delirium, genetic disorders (for example, Down syndrome) or other systemic diseases. For CSF

samples, we further selected samples with a PMD ≤ 30 h, yielding a total of 86 individuals (_n_ = 75 individuals with AD, _n_ = 11 HCs). In addition, we obtained snRNA-seq data from frozen

frontal cortical samples from the UKBBN (_n_ = 12 individuals with AD, _n_ = 9 HCs) from our previously published dataset44. OTHER COHORTS AND DATA FOR ASSOCIATION STUDIES We obtained the

following data for replication studies: (i) genomic, demographic and clinical data from Chinese_cohort_2 (ref. 47); (ii) genomic, demographic and clinical data from the LOAD Family Study48;

(iii) genomic, demographic and clinical data from the National Institute on Aging (NIA) ADC cohorts49,50; (iv) genomic, demographic, clinical and brain imaging data from the ADNI cohort; (v)

genomic, demographic and transcriptomic data from the GTEx dataset;42,43 (vi) plasma biomarker, CSF biomarker, demographic and clinical data from the Stanford ADRC cohort;35 (vii) plasma

biomarker and demographic data from the INTERVAL and LonGenity cohorts37 retrieved from the online database (https://twc-stanford.shinyapps.io/aging_plasma_proteome/); (viii) CSF biomarker

and demographic data from a Japanese cohort;36 and (ix) genomic, demographic, clinical and brain imaging data from the AIBL cohort52. Details related to sample and data collection are

presented in the Supplementary Notes. DNA AND PLASMA EXTRACTION FROM HUMAN BLOOD SAMPLES We collected whole-blood samples (3 ml) from individuals into K3EDTA tubes (VACUETTE). We centrifuged

the samples at 2,000_g_ for 15 min to separate the cell pellet and plasma. The plasma was collected, aliquoted and stored at −80 °C until use. We sent the cell pellets to the Centre for

PanorOmic Sciences (Genomics and Bioinformatics Cores, University of Hong Kong) for genomic DNA extraction using the QIAsymphony DSP DNA Midi Kit (Qiagen) on a QIAsymphony SP platform

(Qiagen). Genomic DNA was eluted with water or Elution Buffer ATE (Qiagen) and stored at 4 °C. We determined the DNA concentration by BioDrop µLITE+ (BioDrop). MEASUREMENT OF PROTEIN LEVELS

IN HUMAN SAMPLES AND CELL LINES We measured the plasma level of sST2 in 613 individuals from Chinese_cohort_1 (_n_ = 277 individuals with AD, _n_ = 336 HCs); CSF level of sST2 in 86

individuals from the UKBBN cohort (_n_ = 75 individuals with AD, _n_ = 11 HCs); and the level of sST2 secreted by hCMEC/D3 cells using the Human ST2/IL-33 R Quantikine ELISA Kit (DST200;

R&D Systems). The plasma levels of NfL (_n_ = 154 individuals with AD, _n_ = 135 HCs) and P-tau181 (_n_ = 156 individuals with AD, _n_ = 134 HCs) in individuals from Chinese_cohort_1

were measured by the Quanterix Accelerator Laboratory using the Quanterix NF-light SIMOA Assay Advantage Kit (103186) and P-Tau 181 Advantage V2 Kit (103714), respectively93. WHOLE-GENOME

SEQUENCING AND SINGLE-NUCLEOTIDE POLYMORPHISM ARRAY FOR GENOTYPING We submitted DNA samples from 427 individuals from Chinese_cohort_1 (_n_ = 233 individuals with AD, _n_ = 194 HCs) to

Novogene for library construction and WGS. The samples were sequenced on the Illumina HiSeq X platform (average depth: 5×), and individual genotypes were analyzed using the GotCloud

pipeline41. The genotype results, which were stored in VCF files, were used for principal-component analysis. The top five principal components were generated by PLINK software (v1.9)94 with

the following parameters: --pca header tabs, --maf 0.05, --hwe 0.00001 and --not-chr x y. We subjected the rest of the genomic DNA samples, including 263 from Chinese_cohort_1 (_n_ = 112

individuals with AD, _n_ = 151 HCs) and 113 from the UKBBN cohort (_n_ = 102 individuals with AD, _n_ = 11 HCs), to SNP array for the genotyping of chr2: 102966067 (GRCh37/hg19), _APOE_-ε2

and _APOE_-ε4 using TaqMan Assays rs1921622, C___1226146_10, 4351376; rs7412, C___904973_10, 4351376; and rs429358, C___3084793_20, 4351376, respectively (Thermo Fisher Scientific). We

performed real-time quantitative PCR (qPCR) using the 7500 Fast and QuantStudio 7 Flex Real-Time PCR System (Applied Biosystems). We stored the results in EDS files and input them into

TaqMan Genotyper Software (Applied Biosystems) for the joint genotyping of SNPs. IMMUNOHISTOCHEMICAL STAINING OF POSTMORTEM HUMAN BRAIN SECTIONS We obtained formalin-fixed,

paraffin-embedded, postmortem, frontal cortex sections from 78 individuals with AD from the UKBBN cohort. We first deparaffinized and rehydrated the sections with xylene and graded ethanol

solutions. To stain Aβ, we treated the sections with formic acid at room temperature (RT) for 5 min and quenched endogenous peroxidase activity with a 3% hydrogen peroxide solution. We then

incubated the sections with a mouse anti-human Aβ antibody (1:500 dilution; clone NAB228, SC-32277, Santa Cruz Biotechnology) overnight at 4 °C. After washing, we incubated the sections with

horseradish peroxidase (HRP)-labeled anti-mouse IgG (QD440-XAKE, RTU, BioGenex) and developed signals with 3,3′-diaminobenzidine (DAB) substrate (QD440-XAKE, BioGenex). To co-stain

microglia and Aβ protein, we performed double immunohistochemical staining; after deparaffinization and rehydration, the sections were treated with sodium citrate buffer (10 mM sodium

citrate, pH 6.0) for 25 min and blocked, then endogenous peroxidase activity was quenched by 3% hydrogen peroxide solution. We then incubated the sections with the mouse anti-human Aβ

antibody (SC-32277) and rabbit anti-human Iba-1 (1:100 dilution; 019-19741, polyclonal, FUJIFILM Wako Pure Chemical) overnight at 4 °C. After washing, we incubated the sections with

HRP-labeled anti-mouse and AP-labeled anti-rabbit (HK597-50K, Double Staining kit, BioGenex) followed by substrate development with DAB (QD440-XAKE, BioGenex) and Fast Red Substrate

(HK182-5KE, BioGenex). We then counterstained the sections with Mayer’s hematoxylin (HK100-9K, BioGenex) and mounted them with coverslips. We used Tris-buffered saline with 0.01% Triton

X-100 as the buffer for washing and to dilute primary antibodies. We took images with a ZEISS Axio Scan.Z1 scanner and processed them with ZEN microscope software v3.2 (ZEISS). To quantify

Aβ plaques, we took ten random images of each section. After background subtraction and threshold adjustment, we analyzed the Aβ plaques using the Analyze Particles function in Fiji (ImageJ

v1.53c). We determined the total Aβ area, number of Aβ plaques and median plaque size for each section. We calculated Aβ plaque load (percentage area) by dividing the total Aβ area by the

total image area (10 mm2). To quantify microglia–Aβ co-staining, we selected 20 random images of each section and processed them with the Colour Deconvolution function to separate the data

into three color channels (that is, DAB, Fast Red and hematoxylin). After adjusting the threshold, we selected Aβ plaques and microglia using the Create Selection function, then analyzed

them using the Analyze function. We determined the total Aβ area and Aβ area colocalized with Iba-1 staining. We calculated the Aβ plaque area colocalized with microglia (percentage total

Aβ) by dividing the Aβ area colocalized with Iba-1 staining by the total Aβ area. Two independent researchers performed section staining, image acquisition and image quantification; they

also randomly selected and quantified images in a blinded manner. ASSOCIATION ANALYSIS AND DATA VISUALIZATION FOR THE GWAS We performed the association analysis between SNPs and plasma sST2

level at the genome-wide level in Chinese_cohort_1 with PLINK software (v1.9)94, adjusting for age, sex, AD diagnosis and the top five principal components as covariates (given that they

have relatively larger eigenvalues with higher power in explaining the population variation in the initially calculated 20 principal components; Supplementary Data 3), using the following

parameters: --keep-allele-order, --linear, --ci 0.95, --hwe 0.00001 and --maf 0.05. To visualize the data, we generated a Manhattan plot and quantile–quantile plot using the manhattan()

function and qq() function of the R qqman package (v0.1.4), respectively. We generated regional plots for the _IL1RL1_ locus using LocusZoom. We performed fine-mapping analysis of the

effects of the _IL1RL1_ locus on plasma sST2 level using CAVIAR software (v2.2)95 and generated association test results and pairwise LD information using PLINK software (v1.9) with the

following parameters: --hwe 0.00001, --maf 0.05, --r, --matrix, --chr 2, --from-bp 102000000 and --to-bp 104000000. We generated the fine-mapped regional plot using the plot_ly() function of

the R plotly package (v4.9.1) and plotted LD and haplotype structures using Haploview (v4.2). To identify all independent sST2-associated variants (_r__2_ < 0.2), we subjected variants

with _P_ < 1 × 10−5 according to the sST2 GWAS to analysis by PLINK software (v1.9; parameters: --hwe 0.00001, --maf 0.05, --clump-p1 0.00001, --clump-r2 0.2, --chr 2 and --clump-kb

2000), yielding 29 independent sST2-associated variants (Supplementary Data 2). We used the calc.relimp() function of the R relaimpo package (v2.2-3)96,97 to quantify the contributions of

genetic factors (that is, the 29 independent sST2-associated variants) and nongenetic factors (that is, age and sex) to sST2-level variance. ASSOCIATION ANALYSIS OF RS1921622 IN

TRANSCRIPTOME DATASETS We used human tissue sST2 and ST2L transcript levels as well as rs1921622 genotype data from the GTEx dataset42,43 for the genotype–expression association test,

adjusting for age, sex, RNA integrity (that is, RNA integrity number) and population structure (that is, the top four principal components). We performed rank-based normalization of

transcript levels using the rntransform() function of the R GenABEL package (v1.8). We obtained the transcript levels of sST2 and ST2L in the human frontal cortex at the single-cell level by

realigning the FASTQ files of our previously published snRNA-seq dataset44 using a modified reference genome. Specifically, we separated the _IL1RL1_ region (chr2: 102,311,563–102,352,037)

in the GTF file of the original GRCh38/hg38 pre-mRNA reference genome into three parts: the sST2-specific region (chr2: 102,343,416–102,346,100), the ST2L-specific region (chr2:

102,311,563–102,337,147 and 102,346,101–102,352,037) and the overlapping region (chr2: 102,337,148–102,343,415). We generated a modified reference genome with Cell Ranger (v3.0.1) using the

new GTF file and original FASTA file. In the subsequent quality-control step, we performed the quantification of gene levels and cell-type identification using Seurat (v3.0)44. For the

association analysis between genotype and candidate gene expression in each cell cluster, we performed a linear regression analysis, adjusting for age, sex, AD diagnosis and PMD. The level

of significance was set at an FDR < 0.05. We performed a GO analysis of associated genes using DAVID Bioinformatics Resources98,99. ANALYSIS OF THE ASSOCIATION BETWEEN RS1921622 AND

ALZHEIMER’S DISEASE RISK We performed a meta-analysis to examine the effects of rs1921622 genotype on AD risk. Specifically, we determined the effect sizes (that is, log ORs) and standard

errors from eight AD datasets (that is, the Chinese_cohort_1 dataset, the WGS and array datasets of Chinese_cohort_2 and the LOAD, ADC1, ADC2, ADC3 and ADNI datasets) using logistic

regression with age, sex and the top five principal components as covariates. We summarized and processed the results by METASOFT (v2.0.0)51 to estimate the joint risk effects and

significance levels under Han and Eskin’s random-effects model (RE2)51. We further performed multiple testing correction using the RE2 _P_ values to generate the FDR values. We then input

the results into ForestPMPlot (v1.0.2) to generate forest plots for data visualization. TWO-SAMPLE MENDELIAN RANDOMIZATION ANALYSIS We performed the two-sample MR analyses of the

associations between sST2 level and AD risk in the overall population and subgroups of individuals in the Chinese and European-descent populations using the mr() function and

inverse‐variance weighted method of the R TwoSampleMR package (v0.5.6), followed by multiple testing correction. Accordingly, for the analysis in European-descent populations, we selected

the _cis_-regulating (chr2: 102,820,000–103,100,000; GRCh37/hg19) protein quantitative trait loci (pQTL) for plasma sST2 levels39 for downstream filtering. Specifically, we used the

statistical results from an independent plasma sST2 _cis_-regulating pQTL dataset from European-descent populations38 to filter out the loci that did not pass the genome-wide threshold (_P_ 

< 5 × 10−8) or had the opposite direction of regulating effects in the two datasets. Moreover, we further used the statistical results from the GTEx dataset42,43 to filter the loci, and

only kept the loci associated with the altered sST2 expression in the expressing tissues (that is, meta _P_ < 1 × 10−5 in the brain regions and lungs) but not the altered ST2L expression

in the expressing tissues (that is, _P_ > 0.1 in the lungs) for LD clumping by PLINK software (v1.9) with the following parameters: --maf 0.05, --hwe 0.00001, --clump-p1 0.00000005,

--clump-r2 0.2 and --clump-kb 2000. We obtained a total of seven independent instrumental variables (rs1921622, rs13001325, rs4851575, rs10515922, rs11123935, rs10200945 and rs951774), then

generated their effects on AD risk by the meta-analysis in European-descent populations using the LOAD, ADC1, ADC2, ADC3 and ADNI datasets. For the analysis in Chinese populations, we

subjected the _cis_-regulating (chr2: 102,820,000–103,100,000; GRCh37/hg19) pQTL for plasma sST2 levels in the Chinese_cohort_1 (WGS) dataset to LD clumping by PLINK software (v1.9) with the

following parameters: --maf 0.05, --hwe 0.00001, --clump-p1 0.00001, --clump-r2 0.5 and --clump-kb 2000. We obtained a total of seven independent instrumental variables (rs1921622,

rs55664618, rs62151861, rs2241116, rs1468790, rs1523199 and rs56238602), then generated their effects on AD risk by the association analysis in the Chinese_cohort_2 (WGS) dataset. MICE We

housed all mice in the HKUST Animal and Plant Care Facility. All animal experiments were approved by the HKUST Animal Ethics Committee and conducted in accordance with the Guidelines of the

Animal Care Facility of HKUST. We housed four mice of the same sex in each cage at 22 °C and at a relative humidity of 60%, with a 12-h light/dark cycle as well as food and water ad libitum.

WT C57BL6J mice were obtained from the Jackson Laboratory. The 5XFAD mice were generated as previously described by overexpressing the p.Lys670Asn/p.Met671Leu (Swedish), p.Ile716Val

(Florida) and p.Val717Ile (London) mutations in human _APP_ as well as the p.Met146Leu and p.Leu286Val mutations in human _PSEN1_ (ref. 100). We confirmed genotypes by PCR analysis of tail

or ear biopsy specimens. We performed all in vivo experiments on age-matched groups and randomly assigned the mice to the experimental conditions. We chose our sample sizes primarily based

on our experience with similar types of experiments. We conducted all animal experiments during the light phase. IN VIVO EXPERIMENTS IN MICE We delivered murine recombinant sST2-Fc

(1004-MR-050; R&D Systems) into 5XFAD mice (B6.Cg-Tg(APPSwFlLon,PSEN1*M146L*L286V)6799Vas/Mmjax) via mini-osmotic pumps (model 1004; Alzet) at 0.11 μl h−1. Specifically, we implanted the

pumps intracerebroventricularly above the right hemisphere and loaded them with murine recombinant sST2-Fc protein (240 ng per pump; 10 μg ml−1) or human IgG (as a control; 009-000-008;

Jackson ImmunoResearch) in artificial CSF (119 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl2·2H2O, 1 mM NaH2PO4·2H2O, 1.3 mM MgCl2·6H2O, 26.2 mM NaHCO3 and 11 mM d-glucose). After 28 d of

administration, the mice were anesthetized with isoflurane and transcardially perfused with PBS, and their brains were collected. IMMUNOHISTOCHEMICAL AND IMMUNOFLUORESCENCE STAINING OF MOUSE

BRAINS The left hemispheres of the mouse brains were fixed in 4% paraformaldehyde at 4 °C for 24 h, transferred to 30% sucrose and stored at 4 °C until sectioning. We cut the brains

coronally into 50-μm sections with a vibrating blade microtome (VT1000S, Leica) and stored them in cryoprotectant solution (30% glycerol, 30% ethylene glycol and PBS) at −20 °C until use.

For immunohistochemistry, we rinsed the sections with PBST (that is, 0.1% Triton X-100 in PBS), then treated them with formic acid at RT for 5 min for antigen retrieval, followed by 3%

hydrogen peroxide solution for 30 min to quench endogenous peroxidase activity. We blocked the sections in 5% horse serum in PBST for 2 h, then labeled them with 4G8 antibody (1:1,000

dilution; 800703, BioLegend) or D54D2 antibody (1:1,000 dilution; 8243S, CST) in blocking buffer overnight at 4 °C. The next day, we incubated the sections with biotin-conjugated anti-mouse

secondary antibodies (1:1,000 dilution; BA2000, Vector Laboratories) followed by an avidin–biotin–HRP complex (PK-6100, Vector Laboratories), and we developed signals with DAB kits (SK-4100,

Vector Laboratories or QD430-XAKE, BioGenex). Imaging was performed using a Zeiss Axio Scan.Z1 Slide Scanner and processed in ZEN v3.3 (ZEISS). We analyzed the Aβ plaque areas in cortical

sections using the Analyze Particles function in Fiji (ImageJ v1.53c). Specifically, we analyzed four brain sections from each mouse in cortical regions (~200–300 μm apart) and calculated

the average percentage of the cortical area occupied by Aβ plaques. For immunofluorescence analysis, we washed sections and incubated them in 1 μM X-34 for 10 min. We then washed them in

X-34 buffer (40% ethanol in PBS), and then in PBS. We then blocked sections for 2 h in blocking buffer (4% horse serum, 1% BSA and 0.3% Triton X-100 in PBS). Primary antibodies used in

experiments include mouse anti-Aβ (1:1,000 dilution; clone 4G8, 800703, BioLegend), rabbit anti-Iba-1 (1:1,000 dilution; 019-19741, Wako) and rat anti-Ki67 (1:200 dilution; clone SolA15,

14-5698-80, eBioscience); we diluted these in blocking buffer and incubated sections overnight at 4 °C. Sections were subsequently incubated with fluorophore-conjugated secondary antibodies

against mouse, rabbit and rat Ig (Alexa Fluor 488, 568 and 647; 1:1,000 dilution; Life Technologies) in blocking buffer for 2 h at RT, extensively washed in PBST, stained with SYTOX Green

(1:300,000 dilution; S7020, Life Technologies) or DAPI (1:5,000 dilution; D3571, Life Technologies) and mounted using FluorSave Reagent (345789, Millipore). We performed imaging using a

Leica TCS SP8 confocal microscope with a Leica ×40 oil immersion objective. We took five images from each mouse cortex with a step size of 1 μm for a total of 40 μm, then merged them into a

single image with maximum intensity _z_-projection. We identified three different plaque morphologies using anti-Aβ immunolabeling (that is, 4G8) and X-34 staining: (i) filamentous plaques

characterized by filamentous X-34 and 4G8 labeling with no plaque core, (ii) compact plaques characterized by 4G8 amyloid fibrils projecting radially outward with an X-34–labeled core and

(iii) inert plaques characterized by an X-34-labeled core with no 4G8 labeling. We used a custom macro to segment individual Aβ plaques and identified the type of each plaque manually. For

plaque-associated microglia, we counted the numbers of microglia surrounding small plaques (that is, radius ≤ 8 μm) and large plaques (that is, radius > 8 μm) manually, as defined by

DAPI+ nucleus staining within the barrier surrounding the plaques and processes in contact with the plaques. We defined the Ki67+ microglia as microglia with Ki67 signals within the nucleus.

To quantify microglial coverage of Aβ plaques, we selected only compact plaques. We used ten optical slices 0.5 μm apart through the center of the plaque for analysis. We processed all

images with a customized macro in Fiji (ImageJ v1.53c). On each slice, after adjusting the threshold, we determined the perimeters of the plaque using the Analyse Particles and Area to Line

functions. We also determined plaque perimeters and the arcs of plaque perimeters colocalized with microglial staining. We calculated the proportion of the plaque perimeter covered by

microglia by summing the arcs of the plaque perimeter across three-dimensional (3D) stacks in close contact (within 1 μm) with Iba-1-immunolabeled cells (~25 plaques per group). We conducted

the 3D reconstruction of microglia–plaque interaction using Imaris v9.7.2 (Oxford Instruments). ASSESSMENT OF MICROGLIAL AMYLOID-BETA PHAGOCYTIC CAPACITY We examined microglial Aβ

phagocytic capacity as previously described25. Briefly, we injected 4-month-old 5XFAD or WT mice intraperitoneally with methoxy-X04 (10 mg per kg body weight) to label the Aβ. We

anesthetized the mice with isoflurane 3 h after methoxy-X04 injection and perfused the left ventricle with ice-cold PBS. We isolated, minced and incubated their forebrains at 37 °C for 30 

min in 5 U ml−1 papain (LS003126) and 35 U ml−1 DNase I (LS002140; Worthington Biochemical) for enzymatic digestion. After incubation, we depleted myelin debris by 30% isotonic Percoll

(P1644; Sigma-Aldrich) gradient centrifugation and obtained mononuclear cell suspensions in DMEM/F12 medium with ice-cold 10% heat-inactivated fetal bovine serum. We prepared unstained

controls from mixtures of different sample cell suspensions for cell population identification. To label microglia, we used an Alexa Fluor 488-conjugated mouse CD11b antibody (1:200

dilution; 53-0112-82; eBioscience) to stain the cell suspensions for 45 min at 4 °C. We analyzed the resultant labeled cell suspensions using a BD Influx cell sorter flow cytometer. We

analyzed the recorded scatterplot data for the microglial cell population using FlowJo software v10.5.0 (TreeStar). CELL CULTURE The hCMEC/D3 cell line was purchased from Cedarlane and

cultured as previously described101. Briefly, we coated a tissue culture plate with 100 μg ml−1 type I collagen (Millipore) at 37 °C in 5% CO2 for 1 h. We subsequently washed the plate with

Dulbecco’s phosphate-buffered saline (DPBS) and replaced it with a complete culture medium (endothelial cell growth medium-2 (Lonza) supplemented with 5% fetal bovine serum (HyClone), 1%

chemically defined lipid concentrate (Gibco), 10 mM HEPES (Gibco), 5 μg ml−1 ascorbic acid (Sigma), 1.4 μM hydrocortisone (Sigma), 1 ng ml−1 basic fibroblast growth factor bFGF (PeproTech),

10 U ml−1 penicillin and 10 μg ml−1 streptomycin (Gibco)). Cultured cells were dissociated with 0.05% trypsin for 5 min, replated at 25,000 cells per cm2 and returned to culture at 37 °C in

a 5% CO2 incubator. Three to four days after seeding, the cells reached confluence and could be passaged. We used cells at passages 27–35 for our experiments. For the ChIP–qPCR experiment,

we fully changed the medium of the hCMEC/D3 cells 2 h before treatment. We then treated cultured cells with recombinant human IL-33 (BioLegend) or DPBS as a vehicle control for 24 h. To

evaluate the efficiency of the sgRNA editing in endothelial cells, we transfected 5 × 105 hCMEC/D3 cells with a single CRISPR construct by nucleofection using the Human Umbilical Vein

Endothelial Cell Nucleofector Kit (Lonza) with a Nucleofector 2b device (Lonza). One day after transfection, we changed the cultured medium to a complete culture medium with 1 μg ml−1

puromycin (Thermo Fisher Scientific). After 3 d of puromycin selection, we extracted genomic DNA using QuickExtract DNA Extraction Solution (Lucigen) followed by a T7EI (NEB) editing

efficiency test. All four sgRNAs exhibited high editing efficiency (data not shown). CRISPR/CAS9-MEDIATED GENOME DELETION OF THE RS1921622-HARBORING REGION To delete the region harboring

rs1921622 in hCMEC/D3 cells, we used a dual-guide, RNA-mediated knockout approach. The cells were genotyped from 300 bp upstream and downstream of rs1921622 by Sanger sequencing. We

performed screening of potential _Streptococcus pyogenes_ Cas9 (SpCas9)-guided RNAs using the CRISPR design tool (https://crispr.mit.edu/) 100 bp upstream and downstream of rs1921622. We

used the following sgRNAs: sgRNA-1, 5′-TTATGGACAGAATTAAGAAG-3′; sgRNA-2, 5′-CTGTCCATAAGATTTGAAAG-3′; sgRNA-3, 5′-AATTTTGTTCTGGTAGCCAT-3′; and sgRNA-4, 5′-GGTATTTCAGCTAGTGCCTA-3′. We

subcloned the sgRNAs into PX459v2, which contains an sgRNA cassette, human codon-optimized SpCas9 and a puromycin resistance gene. To generate a dual-gRNA-mediated deletion cell line, we

transfected the hCMEC/D3 cells with plasmids containing sgRNA-1/sgRNA-4 (targeted 67-bp deletion), sgRNA-2/sgRNA-3 (targeted 38-bp deletion) or PX459v2 as a no-sgRNA control. After 3 d of

puromycin selection (1 μg ml−1) starting from the day after transfection, we seeded the puromycin-resistant cells in two 24-well plates. We changed the culture medium twice a week. After 3

weeks, when a single colony was observed, we passaged the wells containing only one colony into a 12-well plate. We genotyped each clone and subjected those in which the targeted deletion

may have occurred to Sanger sequencing. This protocol generated six control lines, eight lines with a 38-bp deletion and ten lines with a 67-bp deletion. RNA EXTRACTION AND REAL-TIME PCR We

extracted RNA using a NucleoSpin RNA column (MACHEREY-NAGEL) according to the manufacturer’s instructions. We first synthesized single-stranded cDNA using the PrimeScript RT Reagent Kit

(Takara Bio). We performed quantitative real-time PCR with TaqMan probes (Thermo Fisher Scientific) and a Premix Ex Taq Kit (Takara Bio) using the 7500 Fast Real-Time PCR System (Applied

Biosystems). The mRNA expression was normalized to that of _HPRT1_. CHROMATIN IMMUNOPRECIPITATION COUPLED TO QUANTITATIVE PCR We modified the ChIP–qPCR from the chromatin immunoprecipitation

sequencing protocol as previously described25. In brief, we trypsinized cultured cells and washed them once with DPBS. Then, we immediately fixed 4 × 105 cells in suspension with 1.5%

paraformaldehyde at RT for 10 min and quenched them with 0.125 M glycine for 5 min. We subsequently centrifuged the cells at 500_g_ for 5 min. We resuspended the cell pellet in 0.25% SDS

sonication buffer (10 mM Tris (pH 8.0), 0.25% SDS, 2 mM EDTA and protease inhibitors) and subjected it to sonication into 200–500-bp fragments using the S220 Focused-ultrasonicator system

(Covaris). We diluted the DNA–protein mixture at a 1:1.5 ratio in equilibration buffer (10 mM Tris, 233 mM NaCl, 1.67% Triton X-100, 0.167% sodium-deoxycholate (Na-DOC), 1 mM EDTA and

protease inhibitors). We spared 5% of the DNA–protein mixture as input, while we incubated the remainder with antibodies against H3K4me3 (3 μl; 39159; Active Motif), H3K27ac (5 μl; 39133;

Active Motif) or normal IgG (6 μl; AB-105-C; R&D) overnight at 4 °C. Simultaneously, Protein A Dynabeads (Thermo Fisher Scientific) were washed twice and incubated with RIPA-LS buffer

(10 mM Tris-HCl (pH 8.0), 140 mM NaCl, 1 mM EDTA (pH 8.0), 0.1% SDS, 0.1% Na-DOC, 1% Triton X-100 and protease inhibitors), then supplemented with 0.1% BSA overnight at 4 °C. The next day,

we added 10 μl pre-blocked beads to each sample and incubated them for 2 h at 4 °C. We then washed protein-bound beads twice with ice-cold RIPA-LS buffer, twice with ice-cold RIPA-HS buffer

(10 mM Tris-HCl (pH 8.0), 500 mM NaCl, 1 mM EDTA (pH 8.0), 0.1% SDS, 0.1% Na-DOC, 1% Triton X-100 and protease inhibitors), twice with ice-cold RIPA-LiCl buffer (10 mM Tris-HCl (pH 8.0), 250

 mM LiCl, 1 mM EDTA (pH 8.0), 0.5% NP-40, 0.5% Na-DOC and protease inhibitors) and twice with ice-cold 10 mM Tris (pH 8.0) buffer on a precooled magnetic stand. We then reverse-crosslinked

the protein-bound and chromatin-bound beads with protease K (NEB) in ChIP elution buffer (10 mM Tris-HCl (pH 8.0), 300 mM NaCl, 5 mM EDTA (pH 8.0) and 10% SDS) at 55 °C for 1 h, then at 65 

°C for 6 h. We diluted the input to a final concentration of 0.4% SDS and 300 mM NaCl. Next, 2 μl protease K was added to the input and incubated at 55 °C for 1 h, then at 65 °C for 6 h. We

purified both the eluted ChIP and input DNA with a MinElute column (Qiagen). We performed real-time PCR for ChIP and input DNA using Power SYBR Green Master Mix (Applied Biosystems) with the

following primer pairs: sST2_promoter_F, 5′-GGGAAAAAAGCTTGACTTTG-3′; sST2_promoter_R, 5′-ATTTGACAAAGTTCACCCAG-3′; rs192_F, 5′-GCCACTTCTTAATTCTGTCC-3′; and rs192_R,

5′-GGTATTTCAGCTAGTGCCTA-3′. STATISTICS AND REPRODUCIBILITY No statistical methods were used to predetermine sample sizes, but our samples sizes are similar to those reported in previous

publications24,25,41,47. No data were excluded from the analyses, and the data distribution was assumed to be normal, but this was not formally tested. We assigned random codes to human

samples before experiments and analyses; and we randomly assigned nonhuman samples (for example, mice and cultured cells) into experimental groups before experiments and sample collection.

The investigators were blinded to disease diagnosis and other demographic information (for humans) and experimental groups and conditions (for mice and cultured cells) during experiments.

For the remaining statistical analyses of human samples that are not mentioned in the above methods, we conducted a linear regression analysis to determine the significance of the

associations between AD-associated endophenotypes (that is, cognitive performance, brain region volume and AD biomarkers) and sST2 level and/or rs1921622 genotype, adjusted for age, sex and

other covariates as indicated in the main text and figure legends. We obtained the CSF sST2 level cutoff based on the CSF sST2 level with the maximum value of Youden’s index using the

optimal.cutpoints() function and the Youden method of the OptimalCutpoints package (v1.1-4) in R102. We performed Cox regression to examine the association between onset age of dementia and

the rs1921622 A allele using the coxph() function of the R survival package (v1.3-24), with sex and the top five principal components as covariates. We tested the proportional hazard

assumptions using the cox.zph() function of the R survival package (v1.3-24; Supplementary Data 4). The level of significance was set to _P_ < 0.05 or FDR < 0.05. For data

visualization, we used the plot() function of R to generate a volcano plot and the ggplot() function of the R ggplot2 package (v3.2.1) to generate dot plots. For the data obtained from mouse

and cell culture system experiments, we assessed the significance of differences by unpaired Student’s _t_-tests, or one-way or two-way analysis of variance followed by the Bonferroni post

hoc test as indicated. The level of significance was set at _P_ < 0.05. We generated all statistical plots using Prism v8.0 (GraphPad). REPORTING SUMMARY Further information on research

design is available in the Nature Research Reporting Summary linked to this article. DATA AVAILABILITY Source data are available with this paper. All statistical data associated with this

study are contained in the main text, Supplementary Information or Supplementary Data. All other data are available from the corresponding author upon reasonable request. The consent forms

signed by individuals from Chinese_cohort_1 state that the research content will be kept private under the supervision of the hospital and research team. Therefore, the phenotypic, genomic

and proteomic data of individuals will only be available and shared in formal collaborations. A review panel hosted at HKUST will process and review any applications for data sharing and

project collaboration and promptly notify applicants with the decision. Researchers may contact HKUST ([email protected]) for details about data sharing and project collaboration related to

the present study. The GRCh38/hg38 reference genome is available at https://hgdownload.soe.ucsc.edu/downloads.html. The human frontal cortex snRNA-seq dataset of the UKBBN cohort has been

deposited in the Gene Expression Omnibus (accession no. GSE157827). The genomic, demographic and clinical data of the LOAD cohort are available on the National Institutes of Health (NIH)

database of Genotypes and Phenotypes (dbGaP) project (accession no. phs000168.v2.p2). The genomic, demographic and clinical data of the ADC1–3 cohorts are available on the NIH dbGaP project

database (accession no. phs000372.v2.p1). The genomic, demographic, clinical and brain imaging data of the ADNI cohort are available at http://adni.loni.usc.edu/ upon request. The proteomic

and demographic data from the INTERVAL and LonGenity cohorts are available at https://twc-stanford.shinyapps.io/aging_plasma_proteome/. The genomic, demographic and transcriptomic data from

the GTEx cohort are available on the NIH dbGaP project database (accession no. phs000424.v6.p1). The genomic, demographic, clinical and brain imaging data of the AIBL cohort are available at

https://aibl.csiro.au/ upon request. CODE AVAILABILITY The codes for our statistical analyses and data visualization are available at https://github.com/yjiangah/sST2-in-AD/. REFERENCES *

Alzheimer’s Association. Alzheimer’s disease facts and figures. _Alzheimers Dement._ 15, 321–387 (2019). Google Scholar  * Jack, C. R. Jr et al. NIA‐AA research framework: toward a

biological definition of Alzheimer’s disease. _Alzheimers Dement._ 14, 535–562 (2018). Article  PubMed  PubMed Central  Google Scholar  * Lambert, J.-C. et al. Meta-analysis of 74,046

individuals identifies 11 new susceptibility loci for Alzheimer’s disease. _Nat. Genet._ 45, 1452–1458 (2013). Article  CAS  PubMed  PubMed Central  Google Scholar  * Frigerio, C. S. et al.

The major risk factors for Alzheimer’s disease: age, sex and genes modulate the microglia response to Aβ plaques. _Cell Rep._ 27, 1293–1306 (2019). Article  PubMed Central  CAS  Google

Scholar  * Yeh, F. L., Wang, Y., Tom, I., Gonzalez, L. C. & Sheng, M. TREM2 binds to apolipoproteins, including APOE and CLU/APOJ, and thereby facilitates uptake of amyloid-beta by

microglia. _Neuron_ 91, 328–340 (2016). Article  CAS  PubMed  Google Scholar  * Rebeck, G. W., Reiter, J. S., Strickland, D. K. & Hyman, B. T. Apolipoprotein E in sporadic Alzheimer’s

disease: allelic variation and receptor interactions. _Neuron_ 11, 575–580 (1993). Article  CAS  PubMed  Google Scholar  * Kok, E. et al. Apolipoprotein E–dependent accumulation of Alzheimer

disease–related lesions begins in middle age. _Ann. Neurol._ 65, 650–657 (2009). Article  CAS  PubMed  Google Scholar  * Castellano, J. M. et al. Human apoE isoforms differentially regulate

brain amyloid-β peptide clearance. _Sci. Transl. Med._ 3, 89ra57 (2011). Article  CAS  PubMed  PubMed Central  Google Scholar  * Stephen, T. L. et al. APOE genotype and sex affect

microglial interactions with plaques in Alzheimer’s disease mice. _Acta Neuropathol. Commun_. 7, 82 (2019). * Hickman, S. E., Allison, E. K. & El Khoury, J. Microglial dysfunction and

defective β-amyloid clearance pathways in aging Alzheimer’s disease mice. _J. Neurosci._ 28, 8354–8360 (2008). Article  CAS  PubMed  PubMed Central  Google Scholar  * Guillot-Sestier, M.-V.

et al. Il10 deficiency rebalances innate immunity to mitigate Alzheimer-like pathology. _Neuron_ 85, 534–548 (2015). Article  CAS  PubMed  PubMed Central  Google Scholar  * Suárez-Calvet, M.

et al. Early changes in CSF sTREM2 in dominantly inherited Alzheimer’s disease occur after amyloid deposition and neuronal injury. _Sci. Transl. Med._ 8, 369ra178 (2016). Article  PubMed 

PubMed Central  CAS  Google Scholar  * Ewers, M. et al. Increased soluble TREM2 in cerebrospinal fluid is associated with reduced cognitive and clinical decline in Alzheimer’s disease. _Sci.

Transl. Med._ 11, eaav6221 (2019). * Wood, H. Soluble TREM2 in CSF sheds light on microglial activation in AD. _Nat. Rev. Neurol._ 13, 65–65 (2017). Article  CAS  PubMed  Google Scholar  *

Zhong, L. et al. Soluble TREM2 induces inflammatory responses and enhances microglial survival. _J. Exp. Med._ 214, 597–607 (2017). Article  CAS  PubMed  PubMed Central  Google Scholar  *

Zhong, L. et al. Soluble TREM2 ameliorates pathological phenotypes by modulating microglial functions in an Alzheimer’s disease model. _Nat. Commun._ 10, 1–16 (2019). CAS  Google Scholar  *

Janelidze, S. et al. CSF biomarkers of neuroinflammation and cerebrovascular dysfunction in early Alzheimer disease. _Neurology_ 91, e867–e877 (2018). Article  CAS  PubMed  PubMed Central 

Google Scholar  * Huang, C.-W. et al. Clinical significance of circulating vascular cell adhesion molecule-1 to white matter disintegrity in Alzheimer’s dementia. _Thromb. Haemost._ 114,

1230–1240 (2015). Article  PubMed  Google Scholar  * Yousef, H. et al. Aged blood impairs hippocampal neural precursor activity and activates microglia via brain endothelial cell VCAM1.

_Nat. Med._ 25, 988–1000 (2019). Article  CAS  PubMed  PubMed Central  Google Scholar  * Chakrabarty, P. et al. TLR5 decoy receptor as a novel anti-amyloid therapeutic for Alzheimer’s

disease. _J. Exp. Med._ 215, 2247–2264 (2018). Article  CAS  PubMed  PubMed Central  Google Scholar  * Liu, Y. -L. et al. Amelioration of amyloid-β-induced deficits by DcR3 in an Alzheimer’s

disease model. _Mol. Neurodegener_. 12, 30 (2017). * Griesenauer, B. & Paczesny, S. The ST2/IL-33 axis in immune cells during inflammatory diseases. _Front. Immunol._ 8, 475 (2017).

Article  PubMed  PubMed Central  CAS  Google Scholar  * Yasuoka, S. et al. Production and functions of IL-33 in the central nervous system. _Brain Res._ 1385, 8–17 (2011). Article  CAS 

PubMed  Google Scholar  * Fu, A. K. et al. IL-33 ameliorates Alzheimer’s disease-like pathology and cognitive decline. _Proc. Natl Acad. Sci. USA_ 113, E2705–E2713 (2016). Article  CAS 

PubMed  PubMed Central  Google Scholar  * Lau, S.-F. et al. IL-33-PU.1 transcriptome reprogramming drives functional state transition and clearance activity of microglia in Alzheimer’s

disease. _Cell Rep._ 31, 107530 (2020). Article  CAS  PubMed  Google Scholar  * Trajkovic, V., Sweet, M. J. & Xu, D. T1/ST2—an IL-1 receptor-like modulator of immune responses. _Cytokine

Growth Factor Rev._ 15, 87–95 (2004). Article  CAS  PubMed  Google Scholar  * Kakkar, R. & Lee, R. T. The IL-33/ST2 pathway: therapeutic target and novel biomarker. _Nat. Rev. Drug

Discov._ 7, 827–840 (2008). Article  CAS  PubMed  PubMed Central  Google Scholar  * Homsak, E. & Gruson, D. Soluble ST2: a complex and diverse role in several diseases. _Clin. Chim.

Acta_ 507, 75–87 (2020). * Oshikawa, K. et al. Elevated soluble ST2 protein levels in sera of patients with asthma with an acute exacerbation. _Am. J. Respir. Crit. Care Med._ 164, 277–281

(2001). Article  CAS  PubMed  Google Scholar  * Bergis, D., Kassis, V. & Radeke, H. H. High plasma sST2 levels in gastric cancer and their association with metastatic disease. _Cancer

Biomark._ 16, 117–125 (2016). Article  CAS  PubMed  Google Scholar  * Villacorta, H. & Maisel, A. S. Soluble ST2 testing: a promising biomarker in the management of heart failure. _Arq.

Bras. Cardiol._ 106, 145–152 (2016). CAS  PubMed  PubMed Central  Google Scholar  * Saresella, M. et al. IL-33 and its decoy sST2 in patients with Alzheimer’s disease and mild cognitive

impairment. _J. Neuroinflammation_ 17, 174 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Karikari, T. K. et al. Blood phosphorylated tau 181 as a biomarker for Alzheimer’s

disease: a diagnostic performance and prediction modelling study using data from four prospective cohorts. _Lancet Neurol._ 19, 422–433 (2020). Article  CAS  PubMed  Google Scholar  *

Preische, O. et al. Serum neurofilament dynamics predicts neurodegeneration and clinical progression in presymptomatic Alzheimer’s disease. _Nat. Med._ 25, 277–283 (2019). Article  CAS 

PubMed  PubMed Central  Google Scholar  * Gate, D. et al. Clonally expanded CD8 T cells patrol the cerebrospinal fluid in Alzheimer’s disease. _Nature_ 577, 399–404 (2020). Article  CAS 

PubMed  PubMed Central  Google Scholar  * Sasayama, D. et al. Increased cerebrospinal fluid interleukin-6 levels in patients with schizophrenia and those with major depressive disorder. _J.

Psychiatric Res._ 47, 401–406 (2013). Article  Google Scholar  * Lehallier, B. et al. Undulating changes in human plasma proteome profiles across the lifespan. _Nat. Med._ 25, 1843–1850

(2019). Article  CAS  PubMed  PubMed Central  Google Scholar  * Ho, J. E. et al. Common genetic variation at the IL1RL1 locus regulates IL-33/ST2 signaling. _J. Clin. Invest._ 123, 4208–4218

(2013). Article  CAS  PubMed  PubMed Central  Google Scholar  * Sun, B. B. et al. Genomic atlas of the human plasma proteome. _Nature_ 558, 73–79 (2018). Article  CAS  PubMed  PubMed

Central  Google Scholar  * Grotenboer, N. S., Ketelaar, M. E., Koppelman, G. H. & Nawijn, M. C. Decoding asthma: translating genetic variation in IL33 and IL1RL1 into disease

pathophysiology. _J. Allergy Clin. Immunol._131, 856–865 (2013). Article  CAS  PubMed  Google Scholar  * Zhou, X. et al. Non-coding variability at the APOE locus contributes to the

Alzheimer’s risk. _Nat. Commun._ 10, 3310 (2019). Article  PubMed  PubMed Central  CAS  Google Scholar  * Lonsdale, J. et al. The Genotype-Tissue Expression project. _Nat. Genet._ 45,

580–585 (2013). Article  CAS  Google Scholar  * Consortium, G. The Genotype-Tissue Expression pilot analysis: multitissue gene regulation in humans. _Science_ 348, 648–660 (2015). Article 

CAS  Google Scholar  * Lau, S.-F., Cao, H., Fu, A. K. & Ip, N. Y. Single-nucleus transcriptome analysis reveals dysregulation of angiogenic endothelial cells and neuroprotective glia in

Alzheimer’s disease. _Proc. Natl Acad. Sci. USA_ 117, 25800–25809 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Li, W., Notani, D. & Rosenfeld, M. G. Enhancers as

non-coding RNA transcription units: recent insights and future perspectives. _Nat. Rev. Genet._ 17, 207–223 (2016). Article  Google Scholar  * Davies, N. M., Holmes, M. V. & Smith, G. D.

Reading Mendelian randomisation studies: a guide, glossary and checklist for clinicians. _BMJ_ 362, k601 (2018). * Zhou, X. et al. Identification of genetic risk factors in the Chinese

population implicates a role of immune system in Alzheimer’s disease pathogenesis. _Proc. Natl Acad. Sci. USA_ 115, 1697–1706 (2018). Article  CAS  PubMed  PubMed Central  Google Scholar  *

Lee, J. H., Cheng, R., Graff-Radford, N., Foroud, T. & Mayeux, R. Analyses of the National Institute on Aging late-onset alzheimer’s disease family study: implication of additional loci.

_Arch. Neurol._ 65, 1518–1526 (2008). Article  PubMed  PubMed Central  Google Scholar  * Naj, A. C. et al. Common variants at MS4A4/MS4A6E, CD2AP, CD33 and EPHA1 are associated with

late-onset Alzheimer’s disease. _Nat. Genet._ 43, 436–441 (2011). Article  CAS  PubMed  PubMed Central  Google Scholar  * Jun, G. et al. Meta-analysis confirms CR1, CLU and PICALM as

alzheimer disease risk loci and reveals interactions with APOE genotypes. _Arch. Neurol._ 67, 1473–1484 (2010). Article  PubMed  PubMed Central  Google Scholar  * Han, B. & Eskin, E.

Random-effects model aimed at discovering associations in meta-analysis of genome-wide association studies. _Am. J. Hum. Genet._ 88, 586–598 (2011). Article  CAS  PubMed  PubMed Central 

Google Scholar  * Kathryn, A. E. et al. The Australian Imaging, Biomarkers and Lifestyle (AIBL) study of aging: methodology and baseline characteristics of 1,112 individuals recruited for a

longitudinal study of Alzheimer’s disease. _Int. Psychogeriatr._ 21, 672–687 (2009). * Mathys, H. et al. Single-cell transcriptomic analysis of Alzheimer’s disease. _Nature_ 570, 332–337

(2019). Article  CAS  PubMed  PubMed Central  Google Scholar  * Zhou, Y. et al. Human and mouse single-nucleus transcriptomics reveal TREM2-dependent and TREM2-independent cellular responses

in Alzheimer’s disease. _Nat. Med._ 26, 131–142 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Keren-Shaul, H. et al. A unique microglia type associated with restricting

development of Alzheimer’s disease. _Cell_ 169, 1276–1290 (2017). Article  CAS  PubMed  Google Scholar  * Thrupp, N. et al. Single-nucleus RNA-seq is not suitable for detection of microglial

activation genes in humans. _Cell Rep._ 32, 108189 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Yuan, P. et al. TREM2 haplodeficiency in mice and humans impairs the

microglia barrier function leading to decreased amyloid compaction and severe axonal dystrophy. _Neuron_ 90, 724–739 (2016). Article  CAS  PubMed  PubMed Central  Google Scholar  * Wicher,

G. et al. Interleukin-33 promotes recruitment of microglia/macrophages in response to traumatic brain injury. _J. Neurotrauma_ 34, 3173–3182 (2017). Article  PubMed  Google Scholar  * Zhang,

Y. et al. An RNA-sequencing transcriptome and splicing database of glia, neurons and vascular cells of the cerebral cortex. _J. Neurosci._ 34, 11929–11947 (2014). Article  CAS  PubMed 

PubMed Central  Google Scholar  * Gosselin, D. et al. An environment-dependent transcriptional network specifies human microglia identity. _Science_ 356, eaal3222 (2017). * Wang, Y. et al.

Astrocyte-secreted IL-33 mediates homeostatic synaptic plasticity in the adult hippocampus. _Proc. Natl Acad. Sci. USA_ 118, e2020810118 (2021). * Nguyen, P. T. et al. Microglial remodeling

of the extracellular matrix promotes synapse plasticity. _Cell_ 182, 388–403 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Zhang, Y. et al. Purification and

characterization of progenitor and mature human astrocytes reveals transcriptional and functional differences with mouse. _Neuron_ 89, 37–53 (2016). Article  CAS  PubMed  Google Scholar  *

Kalari, K. R. et al. BBBomics—human blood–brain barrier transcriptomics hub. _Front. Neurosci._10, 71 (2016). Article  PubMed  PubMed Central  Google Scholar  * Demyanets, S. et al.

Components of the interleukin-33/ST2 system are differentially expressed and regulated in human cardiac cells and in cells of the cardiac vasculature. _J. Mol. Cellular Cardiol._ 60, 16–26

(2013). Article  CAS  Google Scholar  * Joshi, A. D. et al. Interleukin-33 contributes to both M1 and M2 chemokine marker expression in human macrophages. _BMC Immunol._ 11, 52 (2010).

Article  PubMed  PubMed Central  CAS  Google Scholar  * Bandara, G., Beaven, M. A., Olivera, A., Gilfillan, A. M. & Metcalfe, D. D. Activated mast cells synthesize and release soluble

ST2‐a decoy receptor for IL‐33. _Eur. J. Immunol._ 45, 3034–3044 (2015). Article  CAS  PubMed  PubMed Central  Google Scholar  * Zhang, J. et al. ST2 blockade reduces sST2-producing T cells

while maintaining protective mST2-expressing T cells during graft-versus-host disease. _Sci. Transl. Med._ 7, 308ra160 (2015). Article  PubMed  PubMed Central  Google Scholar  * Lipsky, B.

P., Toy, D. Y., Swart, D. A., Smithgall, M. D. & Smith, D. Deletion of the ST2 proximal promoter disrupts fibroblast‐specific expression but does not reduce the amount of soluble ST2 in

circulation. _Eur. J. Immunol._ 42, 1863–1869 (2012). Article  CAS  PubMed  Google Scholar  * Kikuchi, M. et al. Altered behavior in mice overexpressing soluble ST2. _Mol. Brain_ 13, 74

(2020). * Zlokovic, B. V. Neurovascular pathways to neurodegeneration in Alzheimer’s disease and other disorders. _Nat. Rev. Neurosci._ 12, 723–738 (2011). Article  CAS  PubMed  PubMed

Central  Google Scholar  * Hilscher, M. B. et al. Mechanical stretch increases expression of CXCL1 in liver sinusoidal endothelial cells to recruit neutrophils, generate sinusoidal

microthombi and promote portal hypertension. _Gastroenterology_ 157, 193–209 (2019). Article  CAS  PubMed  Google Scholar  * Grammas, P. & Ovase, R. Inflammatory factors are elevated in

brain microvessels in Alzheimer’s disease. _Neurobiol. Aging_ 22, 837–842 (2001). Article  CAS  PubMed  Google Scholar  * Grammas, P. & Ovase, R. Cerebrovascular transforming growth

factor-β contributes to inflammation in the Alzheimer’s disease brain. _Am. J. Pathol._ 160, 1583–1587 (2002). Article  CAS  PubMed  PubMed Central  Google Scholar  * Ferretti, M. T. et al.

Sex differences in Alzheimer disease—the gateway to precision medicine. _Nat. Rev. Neurol._ 14, 457–469 (2018). Article  PubMed  Google Scholar  * Laws, K. R., Irvine, K. & Gale, T. M.

Sex differences in Alzheimer’s disease. _Curr. Opin. Psych._ 31, 133–139 (2018). Article  Google Scholar  * Laws, K. R., Irvine, K. & Gale, T. M. Sex differences in cognitive impairment

in Alzheimer’s disease. _World J. Psychiatry_ 6, 54–65 (2016). Article  PubMed  PubMed Central  Google Scholar  * Guillot-Sestier, M. -V. et al. Microglial metabolism is a pivotal factor in

sexual dimorphism in Alzheimer’s disease. _Commun. Biol_. 4, 711 (2021). * Russi, A. E., Ebel, M. E., Yang, Y. & Brown, M. A. Male-specific IL-33 expression regulates sex-dimorphic EAE

susceptibility. _Proc. Natl Acad. Sci. USA_ 115, E1520–E1529 (2018). Article  CAS  PubMed  PubMed Central  Google Scholar  * Fernandez, C. G., Hamby, M. E., McReynolds, M. L. & Ray, W.

J. The role of APOE4 in disrupting the homeostatic functions of astrocytes and microglia in aging and Alzheimer’s disease. _Front. Aging Neurosci._ 11, 14 (2019). Article  CAS  PubMed 

PubMed Central  Google Scholar  * Fourgeaud, L. et al. TAM receptors regulate multiple features of microglial physiology. _Nature_ 532, 240–244 (2016). Article  CAS  PubMed  PubMed Central 

Google Scholar  * Kim, J., Basak, J. M. & Holtzman, D. M. The role of apolipoprotein E in Alzheimer’s disease. _Neuron_ 63, 287–303 (2009). Article  CAS  PubMed  PubMed Central  Google

Scholar  * Wang, Y. et al. TREM2 lipid sensing sustains the microglial response in an Alzheimer’s disease model. _Cell_ 160, 1061–1071 (2015). Article  CAS  PubMed  PubMed Central  Google

Scholar  * Mittal, K. et al. CD4 T cells induce a subset of MHCII-expressing microglia that attenuates Alzheimer pathology. _iScience_ 16, 298–311 (2019). Article  CAS  PubMed  PubMed

Central  Google Scholar  * Miller, A. M. et al. IL-33 reduces the development of atherosclerosis. _J. Exp. Med._ 205, 339–346 (2008). Article  CAS  PubMed  PubMed Central  Google Scholar  *

Baba, Y. et al. Involvement of PU.1 in mast cell/basophil-specific function of the human IL1RL1/ST2 promoter. _Allergol. Int._ 61, 461–467 (2012). Article  CAS  PubMed  Google Scholar  *

Kunkle, B. W. et al. Genetic meta-analysis of diagnosed Alzheimer’s disease identifies new risk loci and implicates Aβ, tau, immunity and lipid processing. _Nat. Genet._ 51, 414–430 (2019).

Article  CAS  PubMed  PubMed Central  Google Scholar  * Jun, G. et al. A novel Alzheimer disease locus located near the gene encoding tau protein. _Mol. Psychiatry_ 21, 108–117 (2016).

Article  CAS  PubMed  Google Scholar  * O’Meara, E. et al. Independent prognostic value of serum soluble ST2 measurements in patients with heart failure and a reduced ejection fraction in

the PARADIGM-HF trial (prospective comparison of ARNI with ACEI to determine impact on global mortality and morbidity in heart failure). _Circ. Heart Fail._ 11, e004446 (2018). Article 

PubMed  CAS  Google Scholar  * American Psychiatric Association. _Diagnostic and Statistical Manual of Mental Disorders_ 5th edn. (American Psychiatric Association, 2013). * Pangman, V. C.,

Sloan, J. & Guse, L. An examination of psychometric properties of the Mini-Mental State Examination and the standardized mini-mental state examination: implications for clinical

practice. _Appl. Nurs. Res._ 13, 209–213 (2000). Article  CAS  PubMed  Google Scholar  * Nasreddine, Z. S. et al. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild

cognitive impairment. _J. Am. Geriatr. Soc._ 53, 695–699 (2005). Article  PubMed  Google Scholar  * Jiang, Y. et al. Large‐scale plasma proteomic profiling identifies a high‐performance

biomarker panel for Alzheimer’s disease screening and staging. _Alzheimers Dement._ 18, 88–102 (2021). * Purcell, S. et al. PLINK: a tool set for whole-genome association and

population-based linkage analyses. _Am. J. Hum. Genet._ 81, 559–575 (2007). Article  CAS  PubMed  PubMed Central  Google Scholar  * Hormozdiari, F., Kostem, E., Kang, E. Y., Pasaniuc, B.

& Eskin, E. Identifying causal variants at loci with multiple signals of association. _Genetics_ 198, 497–508 (2014). Article  CAS  PubMed  PubMed Central  Google Scholar  * Grömping, U.

Relative importance for linear regression in R: the package relaimpo. _J. Stat. Softw._ 17, 1–27 (2006). Article  Google Scholar  * Patin, E. et al. Natural variation in the parameters of

innate immune cells is preferentially driven by genetic factors. _Nat. Immunol._ 19, 302–314 (2018). Article  CAS  PubMed  Google Scholar  * Huang, D. W., Sherman, B. T. & Lempicki, R.

A. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. _Nucleic Acids Res._ 37, 1–13 (2009). Article  CAS  Google Scholar  * Sherman, B.

T. & Lempicki, R. A. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. _Nat. Protoc._ 4, 44–57 (2009). PubMed  Google Scholar  * Oakley, H. et

al. Intraneuronal β-amyloid aggregates, neurodegeneration, and neuron loss in transgenic mice with five familial Alzheimer’s disease mutations: potential factors in amyloid plaque

formation. _J. Neurosci._ 26, 10129–10140 (2006). Article  CAS  PubMed  PubMed Central  Google Scholar  * Weksler, B. et al. Blood–brain barrier‐specific properties of a human adult brain

endothelial cell line. _FASEB J._ 19, 1872–1874 (2005). Article  CAS  PubMed  Google Scholar  * Youden, W. J. Index for rating diagnostic tests. _Cancer_ 3, 32–35 (1950). Article  CAS 

PubMed  Google Scholar  Download references ACKNOWLEDGEMENTS We thank P. Kwan, H. Mok, C. Y. Ling and R. M. N. Lo for coordinating the collection of clinical data. We also thank P. -O.

Couraud (Institut national de la santé et de la recherche médicale) for providing advice on the cell line culturing. We thank J. K. Y. Lau, E. K. F. Tam, Y. Duan, A. Miranda, C. W. S. Kwong,

X. Yang, H. Cao, J. Xu, A. S. L. Yuen and G. P. O. Chiu for their excellent technical assistance as well as other members of the Ip Laboratory for their many helpful discussions. We thank

the SWDBB for providing brain tissue for this study. The SWDBB is part of the Brains for Dementia Research program, jointly funded by Alzheimer’s Research UK and Alzheimer’s Society and

supported by BRACE (Bristol Research into Alzheimer’s and Care of the Elderly) and the MRC. We thank the AIBL study (https://aibl.csiro.au/). The AIBL study is a consortium between the

Commonwealth Scientific and Industrial Research Organisation (CSIRO), Edith Cowan University, The Florey Institute and Austin Health. Partial financial support was provided by the US

Alzheimer’s Association, the Alzheimer’s Drug Discovery Foundation, an anonymous foundation, the Cooperative Research Centre for Mental Health, the CSIRO Science and Industry Endowment Fund,

the Dementia Collaborative Research Centres, the Victorian Government Operational Infrastructure Support program, the Australian Alzheimer’s Research Foundation, the National Health and MRC

and the Yulgilbar Foundation. We also thank M. Vacher for providing advice on the data analysis. Please refer to the Supplementary Notes for corresponding acknowledgments for the ADNI

dataset, Alzheimer’s Disease Genetics Consortium GWAS–NIA Alzheimer’s Disease Centers Cohort (ADC dataset), the NIA–LOAD dataset and the GTEx Project. Part of the data used in the

preparation of this article was obtained from the ADNI database. As such, the investigators within ADNI contributed to the design and implementation of ADNI and/or provided data but did not

participate in the analysis or writing of this report. A complete listing of ADNI investigators can be found in the Supplementary Notes and at

http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf. This study was supported in part by the National Key R&D Program of China (2018YFE0203600 and

2017YFE0190000); the Hong Kong Research Grants Council Theme-based Research Scheme (T13-605/18-W); the Area of Excellence Scheme of the University Grants Committee (AoE/M-604/16); the

Innovation and Technology Commission (ITCPD/17-9, MRP/042/18X and INNOHK18SC01); the National Natural Science Foundation of China (31671047); the Guangdong Provincial Key S&T Program

(2018B030336001); the Guangdong Provincial Fund for Basic and Applied Basic Research (2019B1515130004); the Shenzhen Knowledge Innovation Program (JCYJ20180507183642005 and

JCYJ20170413173717055); and the Chow Tai Fook Charity Foundation (CTFCF18SC01). Analyses in AIBL were supported through NHMRC (Australia) funding awarded to S.M.L. (APP1161706). The funders

had no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript. AUTHOR INFORMATION Author notes * Benoit Lehallier Present address:

Alkahest Inc, San Carlos, California, USA AUTHORS AND AFFILIATIONS * Division of Life Science, State Key Laboratory of Molecular Neuroscience, Molecular Neuroscience Center, The Hong Kong

University of Science and Technology, Clear Water Bay, Hong Kong, China Yuanbing Jiang, Xiaopu Zhou, Hiu Yi Wong, Li Ouyang, Fanny C. F. Ip, Vicky M. N. Chau, Shun-Fat Lau, Wei Wu, Daniel Y.

K. Wong, Heukjin Seo, Wing-Yu Fu, Nicole C. H. Lai, Yuewen Chen, Yu Chen, Estella P. S. Tong, Kin Y. Mok, Amy K. Y. Fu & Nancy Y. Ip * Hong Kong Center for Neurodegenerative Diseases,

Hong Kong Science Park, Hong Kong, China Yuanbing Jiang, Xiaopu Zhou, Hiu Yi Wong, Li Ouyang, Fanny C. F. Ip, Vicky M. N. Chau, Shun-Fat Lau, Wei Wu, Daniel Y. K. Wong, Wing-Yu Fu, Nicole C.

H. Lai, Estella P. S. Tong, Kin Y. Mok, John Hardy, Tony Wyss-Coray, Amy K. Y. Fu & Nancy Y. Ip * Guangdong Provincial Key Laboratory of Brain Science, Disease and Drug Development;

Shenzhen–Hong Kong Institute of Brain Science, HKUST Shenzhen Research Institute, Shenzhen, China Xiaopu Zhou, Fanny C. F. Ip, Yuewen Chen, Yu Chen, Amy K. Y. Fu & Nancy Y. Ip * The

Brain Cognition and Brain Disease Institute, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences; Shenzhen–Hong Kong Institute of Brain Science–Shenzhen Fundamental

Research Institutions, Shenzhen, China Yuewen Chen & Yu Chen * Gerald Choa Neuroscience Centre, Lui Che Woo Institute of Innovative Medicine, Therese Pei Fong Chow Research Centre for

Prevention of Dementia, Division of Neurology, Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong, China Vincent C. T. Mok * Therese Pei Fong Chow

Research Centre for Prevention of Dementia, Division of Geriatrics, Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong, China Timothy C. Y. Kwok *

Department of Neurodegenerative Disease, UCL Queen Square Institute of Neurology, London, UK Kin Y. Mok, Maryam Shoai & John Hardy * UK Dementia Research Institute, University College

London, London, UK Kin Y. Mok, Maryam Shoai & John Hardy * Department of Neurology and Neurological Sciences, Stanford University School of Medicine, Stanford, California, USA Benoit

Lehallier, Patricia Morán Losada & Tony Wyss-Coray * Wu Tsai Neurosciences Institute, Stanford University, Stanford, California, USA Patricia Morán Losada & Tony Wyss-Coray * Centre

for Precision Health, Edith Cowan University, Joondalup, Australia Eleanor O’Brien, Tenielle Porter & Simon M. Laws * Collaborative Genomics and Translation Group, School of Medical and

Health Sciences, Edith Cowan University, Joondalup, Australia Eleanor O’Brien, Tenielle Porter & Simon M. Laws * School of Pharmacy and Biomedical Sciences, Faculty of Health Sciences,

Curtin Health Innovation Research Institute, Curtin University, Bentley, Australia Tenielle Porter & Simon M. Laws * Institute for Advanced Study, The Hong Kong University of Science and

Technology, Clear Water Bay, Hong Kong, China John Hardy * The Phil and Penny Knight Initiative for Brain Resilience, Stanford University, Stanford, California, USA Tony Wyss-Coray * The

Florey Institute of Neuroscience and Mental Health, The University of Melbourne, Melbourne, Australia Colin L. Masters * UC San Francisco, San Francisco, California, USA Michael W. Weiner, 

Norbert Schuff, Howard J. Rosen, Bruce L. Miller, Thomas Neylan, Jacqueline Hayes & Shannon Finley * UC San Diego, San Diego, California, USA Paul Aisen, Leon Thal, James Brewer, Helen

Vanderswag, Adam Fleisher, Melissa Davis & Rosemary Morrison * Mayo Clinic, Rochester, New York, USA Ronald Petersen, Clifford R. Jack, Matthew Bernstein, Bret Borowski, Jeff Gunter, 

Matt Senjem, Prashanthi Vemuri, David Jones, Kejal Kantarci, Chad Ward, Sara S. Mason, Colleen S. Albers, David Knopman & Kris Johnson * UC Berkeley, Berkeley, California, USA William

Jagust & Susan Landau * University of Pennsylvania, Philadelphia, Pennsylvania, USA John Q. Trojanowski, Leslie M. Shaw, Virginia Lee, Magdalena Korecka, Michal Figurski, Steven E.

Arnold, Jason H. Karlawish & David Wolk * University of Southern California, Los Angeles, California, USA Arthur W. Toga, Karen Crawford, Scott Neu, Lon S. Schneider, Sonia Pawluczyk, 

Mauricio Beccera, Liberty Teodoro & Bryan M. Spann * UC Davis, Davis, California, USA Laurel Beckett, Zaven Khachaturian, Danielle Harvey, Norbert Schuff, Evan Fletcher, Owen Carmichael,

 John Olichney & Charles DeCarli * Brigham and Women’s Hospital/Harvard Medical School, Boston, Massachusetts, USA Robert C. Green, Reisa A. Sperling, Keith A. Johnson, Gad Marshall, 

Meghan Frey, Barton Lane, Allyson Rosen & Jared Tinklenberg * Indiana University, Bloomington, Indiana, USA Andrew J. Saykin, Tatiana M. Foroud, Li Shen, Kelley Faber, Sungeun Kim, 

Kwangsik Nho, Martin R. Farlow, Ann Marie Hake, Brandy R. Matthews, Scott Herring & Cynthia Hunt * Washington University in St. Louis, St. Louis, Missouri, USA John Morris, Marcus

Raichle, David Holtzman, Nigel J. Cairns, Erin Householder, Lisa Taylor-Reinwald, Beau Ances, Maria Carroll, Sue Leon, Mark A. Mintun, Stacy Schneider & Angela Oliver * Prevent

Alzheimer’s Disease 2020, Rockville, Maryland, USA Zaven Khachaturian & Lisa Raudin * Siemens, Munich, Germany Greg Sorensen * University of Pittsburgh, Pittsburgh, Pennsylvania, USA Lew

Kuller, Oscar L. Lopez & MaryAnn Oakley * Weill Cornell Medical College, Cornell University, New York City, New York, USA Steven Paul, Norman Relkin, Gloria Chaing & Lisa Raudin *

Albert Einstein College of Medicine, Yeshiva University, Bronx, New York, USA Peter Davies * Alzheimer’s Drug Discovery Foundation, New York City, New York, USA Howard Fillit & Chet

Mathis * Acumen Pharmaceuticals, Livermore, California, USA Franz Hefti * Northwestern University, Evanston and Chicago, Evanston, Illinois, USA Marek M. Mesulam, Diana Kerwin, Kristine

Lipowski, Chuang-Kuo Wu, Nancy Johnson & Jordan Grafman * National Institute of Mental Health, Rockville, Maryland, USA William Potter * Brown University, Providence, Rhode Island, USA

Peter Snyder * Eli Lilly, Indianapolis, Indiana, USA Adam Schwartz * University of Washington, Seattle, Washington, USA Tom Montine, Ronald G. Thomas, Michael Donohue, Sarah Walter, Devon

Gessert, Tamie Sather, Gus Jiminez & Elaine R. Peskind * UCLA, Los Angeles, California, USA Paul Thompson, Liana Apostolova, Kathleen Tingus, Ellen Woo, Daniel H. S. Silverman, Po H. Lu,

 George Bartzokis, Adrian Preda & Dana Nguyen * University of Michigan, Ann Arbor, Michigan, USA Robert A. Koeppe, Judith L. Heidebrink & Joanne L. Lord * University of Utah, Salt

Lake City, Utah, USA Norm Foster, Pierre Tariot & Stephanie Reeder * Banner Alzheimer’s Institute, Phoenix, Arizona, USA Eric M. Reiman, Kewei Chen & Adam Fleisher * UC Irvine,

Irvine, California, USA Steven G. Potkin, Ruth A. Mulnard, Gaby Thai & Catherine McAdams-Ortiz * National Institute on Aging, Bethesda, Maryland, USA Neil Buckholtz & John Hsiao *

Johns Hopkins University, Baltimore, Maryland, USA Marylyn Albert, Chiadi Onyike, Daniel D’Agostino, Stephanie Kielb & Donna M. Simpson * Richard Frank Consulting, Washington, USA

Richard Frank * Oregon Health and Science University, Portland, Oregon, USA Jeffrey Kaye, Joseph Quinn, Betty Lind, Raina Carter & Sara Dolen * Baylor College of Medicine, Houston,

Texas, USA Rachelle S. Doody, Javier Villanueva-Meyer, Munir Chowdhury, Susan Rountree, Mimi Dang, Yaakov Stern, Lawrence S. Honig & Karen L. Bell * University of Alabama, Birmingham,

Alabama, USA Daniel Marson, Randall Griffith, David Clark, David Geldmacher, John Brockington & Erik Roberson * Mount Sinai School of Medicine, New York City, New York, USA Hillel

Grossman & Effie Mitsis * Rush University Medical Center, Chicago, Illinois, USA Leyla de Toledo-Morrell, Raj C. Shah, Debra Fleischman & Konstantinos Arfanakis * Wien Center, Miami,

Florida, USA Ranjan Duara, Daniel Varon, Maria T. Greig & Peggy Roberts * New York University, New York City, New York, USA James E. Galvin, Brittany Cerbone, Christina A. Michel, Henry

Rusinek, Mony J. de Leon, Lidia Glodzik & Susan De Santi * Duke University Medical Center, Durham, North Carolina, USA P. Murali Doraiswamy, Jeffrey R. Petrella, Terence Z. Wong & 

Olga James * University of Kentucky, Lexington, Kentucky, USA Charles D. Smith, Greg Jicha, Peter Hardy, Partha Sinha, Elizabeth Oates & Gary Conrad * University of Rochester Medical

Center, Rochester, New York, USA Anton P. Porsteinsson * University of Texas Southwestern Medical School, Dallas, Texas, USA Bonnie S. Goldstein, Kim Martin, Kelly M. Makino, M. Saleem

Ismail, Connie Brand, Kyle Womack, Dana Mathews, Mary Quiceno, Ramon Diaz-Arrastia, Richard King, Myron Weiner, Kristen Martin-Cook & Michael DeVous * Emory University, Atlanta, Georgia,

USA Allan I. Levey, James J. Lah & Janet S. Cellar * Medical Center, University of Kansas, Kansas City, Kansas, USA Jeffrey M. Burns, Heather S. Anderson & Russell H. Swerdlow *

Mayo Clinic, Jacksonville, Florida, USA Neill R. Graff-Radford, Francine Parfitt, Tracy Kendall & Heather Johnson * Yale University School of Medicine, New Haven, Connecticut, USA

Christopher H. van Dyck, Richard E. Carson & Martha G. MacAvoy * McGill University/Montreal-Jewish General Hospital, Montreal, Quebec, Canada Howard Chertkow, Howard Bergman & Chris

Hosein * University of British Columbia Clinic for Alzheimer’s Disease and Related Disorders, Vancouver, British Columbia, Canada Ging-Yuek Robin Hsiung, Howard Feldman, Benita Mudge & 

Michele Assaly * Cleveland Clinic Lou Ruvo Center for Brain Health, Las Vegas, Nevada, USA Charles Bernick & Donna Munic * St Joseph’s Health Care, London, Ontario, Canada Andrew

Kertesz, John Rogers, Dick Trost, Stephen Pasternak, Irina Rachinsky & Dick Drost * Premiere Research Institute, Palm Beach Neurology, Miami, Florida, USA Carl Sadowsky, Walter Martinez 

& Teresa Villena * Georgetown University Medical Center, Washington, D.C., USA Raymond Scott Turner, Kathleen Johnson & Brigid Reynolds * Banner Sun Health Research Institute, Sun

City, Arizona, USA Marwan N. Sabbagh, Christine M. Belden, Sandra A. Jacobson, Sherye A. Sirrel & Neil Kowall * Boston University, Boston, Massachusetts, USA Ronald Killiany, Andrew E.

Budson, Alexander Norbash & Patricia Lynn Johnson * Howard University, Washington, D.C., USA Joanne Allard * Case Western Reserve University, Cleveland, Ohio, USA Alan Lerner, Paula

Ogrocki & Leon Hudson * Neurological Care of CNY, Liverpool, New York, USA Smita Kittur * Parkwood Hospital, London, Ontario, Canada Michael Borrie, T-Y. Lee & Rob Bartha *

University of Wisconsin, Madison, Wisconsin, USA Sterling Johnson, Sanjay Asthana, Cynthia M. Carlsson, J. Jay Fruehling & Sandra Harding * Dent Neurologic Institute, Amherst, New York,

USA Vernice Bates, Horacio Capote & Michelle Rainka * Ohio State University, Columbus, Ohio, USA Douglas W. Scharre, Maria Kataki, Anahita Adeli, Eric C. Petrie & Gail Li * Albany

Medical College, Albany, New York, USA Earl A. Zimmerman, Dzintra Celmins & Alice D. Brown * Olin Neuropsychiatry Research Center, Hartford Hospital, Hartford, Connecticut, USA Godfrey

D. Pearlson, Karen Blank & Karen Anderson * Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire, USA Robert B. Santulli, Tamar J. Kitzmiller & Eben S. Schwartz * Wake Forest

University Health Sciences, Winston-Salem, North Carolina, USA Kaycee M. Sink, Jeff D. Williamson, Pradeep Garg & Franklin Watkins * Rhode Island Hospital, Providence, Rhode Island, USA

Brian R. Ott, Henry Querfurth & Geoffrey Tremont * Butler Hospital, Providence, Rhode Island, USA Stephen Salloway, Paul Malloy & Stephen Correia * Medical University South Carolina,

Charleston, South Carolina, USA Jacobo Mintzer, Kenneth Spicer, David Bachman & Dino Massoglia * Nathan Kline Institute, Orangeburg, New York, USA Nunzio Pomara, Raymundo Hernando &

 Antero Sarrael * University of Iowa College of Medicine, Iowa City, Iowa, USA Susan K. Schultz, Laura L. Boles Ponto, Hyungsub Shim & Karen Elizabeth Smith * USF Health Byrd Alzheimer’s

Institute, University of South Florida, Tampa, Florida, USA Amanda Smith, Kristin Fargher & Balebail Ashok Raj * Department of Defense, Arlington, Virginia, USA Karl Friedl * Stanford

University, Stanford, California, USA Jerome A. Yesavage, Joy L. Taylor & Ansgar J. Furst Authors * Yuanbing Jiang View author publications You can also search for this author inPubMed 

Google Scholar * Xiaopu Zhou View author publications You can also search for this author inPubMed Google Scholar * Hiu Yi Wong View author publications You can also search for this author

inPubMed Google Scholar * Li Ouyang View author publications You can also search for this author inPubMed Google Scholar * Fanny C. F. Ip View author publications You can also search for

this author inPubMed Google Scholar * Vicky M. N. Chau View author publications You can also search for this author inPubMed Google Scholar * Shun-Fat Lau View author publications You can

also search for this author inPubMed Google Scholar * Wei Wu View author publications You can also search for this author inPubMed Google Scholar * Daniel Y. K. Wong View author publications

You can also search for this author inPubMed Google Scholar * Heukjin Seo View author publications You can also search for this author inPubMed Google Scholar * Wing-Yu Fu View author

publications You can also search for this author inPubMed Google Scholar * Nicole C. H. Lai View author publications You can also search for this author inPubMed Google Scholar * Yuewen Chen

View author publications You can also search for this author inPubMed Google Scholar * Yu Chen View author publications You can also search for this author inPubMed Google Scholar * Estella

P. S. Tong View author publications You can also search for this author inPubMed Google Scholar * Vincent C. T. Mok View author publications You can also search for this author inPubMed 

Google Scholar * Timothy C. Y. Kwok View author publications You can also search for this author inPubMed Google Scholar * Kin Y. Mok View author publications You can also search for this

author inPubMed Google Scholar * Maryam Shoai View author publications You can also search for this author inPubMed Google Scholar * Benoit Lehallier View author publications You can also

search for this author inPubMed Google Scholar * Patricia Morán Losada View author publications You can also search for this author inPubMed Google Scholar * Eleanor O’Brien View author

publications You can also search for this author inPubMed Google Scholar * Tenielle Porter View author publications You can also search for this author inPubMed Google Scholar * Simon M.

Laws View author publications You can also search for this author inPubMed Google Scholar * John Hardy View author publications You can also search for this author inPubMed Google Scholar *

Tony Wyss-Coray View author publications You can also search for this author inPubMed Google Scholar * Colin L. Masters View author publications You can also search for this author inPubMed 

Google Scholar * Amy K. Y. Fu View author publications You can also search for this author inPubMed Google Scholar * Nancy Y. Ip View author publications You can also search for this author

inPubMed Google Scholar CONSORTIA ALZHEIMER’S DISEASE NEUROIMAGING INITIATIVE * Michael W. Weiner * , Paul Aisen * , Ronald Petersen * , Clifford R. Jack * , William Jagust * , John Q.

Trojanowski * , Arthur W. Toga * , Laurel Beckett * , Robert C. Green * , Andrew J. Saykin * , John Morris * , Leslie M. Shaw * , Zaven Khachaturian * , Greg Sorensen * , Lew Kuller * , 

Marcus Raichle * , Steven Paul * , Peter Davies * , Howard Fillit * , Franz Hefti * , David Holtzman * , Marek M. Mesulam * , William Potter * , Peter Snyder * , Adam Schwartz * , Tom

Montine * , Ronald G. Thomas * , Michael Donohue * , Sarah Walter * , Devon Gessert * , Tamie Sather * , Gus Jiminez * , Danielle Harvey * , Matthew Bernstein * , Paul Thompson * , Norbert

Schuff * , Bret Borowski * , Jeff Gunter * , Matt Senjem * , Prashanthi Vemuri * , David Jones * , Kejal Kantarci * , Chad Ward * , Robert A. Koeppe * , Norm Foster * , Eric M. Reiman * , 

Kewei Chen * , Chet Mathis * , Susan Landau * , Nigel J. Cairns * , Erin Householder * , Lisa Taylor-Reinwald * , Virginia Lee * , Magdalena Korecka * , Michal Figurski * , Karen Crawford *

, Scott Neu * , Tatiana M. Foroud * , Steven G. Potkin * , Li Shen * , Kelley Faber * , Sungeun Kim * , Kwangsik Nho * , Leon Thal * , Neil Buckholtz * , Marylyn Albert * , Richard Frank * ,

 John Hsiao * , Jeffrey Kaye * , Joseph Quinn * , Betty Lind * , Raina Carter * , Sara Dolen * , Lon S. Schneider * , Sonia Pawluczyk * , Mauricio Beccera * , Liberty Teodoro * , Bryan M.

Spann * , James Brewer * , Helen Vanderswag * , Adam Fleisher * , Judith L. Heidebrink * , Joanne L. Lord * , Sara S. Mason * , Colleen S. Albers * , David Knopman * , Kris Johnson * , 

Rachelle S. Doody * , Javier Villanueva-Meyer * , Munir Chowdhury * , Susan Rountree * , Mimi Dang * , Yaakov Stern * , Lawrence S. Honig * , Karen L. Bell * , Beau Ances * , Maria Carroll *

, Sue Leon * , Mark A. Mintun * , Stacy Schneider * , Angela Oliver * , Daniel Marson * , Randall Griffith * , David Clark * , David Geldmacher * , John Brockington * , Erik Roberson * , 

Hillel Grossman * , Effie Mitsis * , Leyla de Toledo-Morrell * , Raj C. Shah * , Ranjan Duara * , Daniel Varon * , Maria T. Greig * , Peggy Roberts * , Chiadi Onyike * , Daniel D’Agostino *

, Stephanie Kielb * , James E. Galvin * , Brittany Cerbone * , Christina A. Michel * , Henry Rusinek * , Mony J. de Leon * , Lidia Glodzik * , Susan De Santi * , P. Murali Doraiswamy * , 

Jeffrey R. Petrella * , Terence Z. Wong * , Steven E. Arnold * , Jason H. Karlawish * , David Wolk * , Charles D. Smith * , Greg Jicha * , Peter Hardy * , Partha Sinha * , Elizabeth Oates *

, Gary Conrad * , Oscar L. Lopez * , MaryAnn Oakley * , Donna M. Simpson * , Anton P. Porsteinsson * , Bonnie S. Goldstein * , Kim Martin * , Kelly M. Makino * , M. Saleem Ismail * , Connie

Brand * , Ruth A. Mulnard * , Gaby Thai * , Catherine McAdams-Ortiz * , Kyle Womack * , Dana Mathews * , Mary Quiceno * , Ramon Diaz-Arrastia * , Richard King * , Myron Weiner * , Kristen

Martin-Cook * , Michael DeVous * , Allan I. Levey * , James J. Lah * , Janet S. Cellar * , Jeffrey M. Burns * , Heather S. Anderson * , Russell H. Swerdlow * , Liana Apostolova * , Kathleen

Tingus * , Ellen Woo * , Daniel H. S. Silverman * , Po H. Lu * , George Bartzokis * , Neill R. Graff-Radford * , Francine Parfitt * , Tracy Kendall * , Heather Johnson * , Martin R. Farlow *

, Ann Marie Hake * , Brandy R. Matthews * , Scott Herring * , Cynthia Hunt * , Christopher H. van Dyck * , Richard E. Carson * , Martha G. MacAvoy * , Howard Chertkow * , Howard Bergman * ,

 Chris Hosein * , Ging-Yuek Robin Hsiung * , Howard Feldman * , Benita Mudge * , Michele Assaly * , Charles Bernick * , Donna Munic * , Andrew Kertesz * , John Rogers * , Dick Trost * , 

Diana Kerwin * , Kristine Lipowski * , Chuang-Kuo Wu * , Nancy Johnson * , Carl Sadowsky * , Walter Martinez * , Teresa Villena * , Raymond Scott Turner * , Kathleen Johnson * , Brigid

Reynolds * , Reisa A. Sperling * , Keith A. Johnson * , Gad Marshall * , Meghan Frey * , Barton Lane * , Allyson Rosen * , Jared Tinklenberg * , Marwan N. Sabbagh * , Christine M. Belden * ,

 Sandra A. Jacobson * , Sherye A. Sirrel * , Neil Kowall * , Ronald Killiany * , Andrew E. Budson * , Alexander Norbash * , Patricia Lynn Johnson * , Joanne Allard * , Alan Lerner * , Paula

Ogrocki * , Leon Hudson * , Evan Fletcher * , Owen Carmichael * , John Olichney * , Charles DeCarli * , Smita Kittur * , Michael Borrie * , T-Y. Lee * , Rob Bartha * , Sterling Johnson * , 

Sanjay Asthana * , Cynthia M. Carlsson * , Adrian Preda * , Dana Nguyen * , Pierre Tariot * , Stephanie Reeder * , Vernice Bates * , Horacio Capote * , Michelle Rainka * , Douglas W. Scharre

* , Maria Kataki * , Anahita Adeli * , Earl A. Zimmerman * , Dzintra Celmins * , Alice D. Brown * , Godfrey D. Pearlson * , Karen Blank * , Karen Anderson * , Robert B. Santulli * , Tamar

J. Kitzmiller * , Eben S. Schwartz * , Kaycee M. Sink * , Jeff D. Williamson * , Pradeep Garg * , Franklin Watkins * , Brian R. Ott * , Henry Querfurth * , Geoffrey Tremont * , Stephen

Salloway * , Paul Malloy * , Stephen Correia * , Howard J. Rosen * , Bruce L. Miller * , Jacobo Mintzer * , Kenneth Spicer * , David Bachman * , Stephen Pasternak * , Irina Rachinsky * , 

Dick Drost * , Nunzio Pomara * , Raymundo Hernando * , Antero Sarrael * , Susan K. Schultz * , Laura L. Boles Ponto * , Hyungsub Shim * , Karen Elizabeth Smith * , Norman Relkin * , Gloria

Chaing * , Lisa Raudin * , Amanda Smith * , Kristin Fargher * , Balebail Ashok Raj * , Thomas Neylan * , Jordan Grafman * , Melissa Davis * , Rosemary Morrison * , Jacqueline Hayes * , 

Shannon Finley * , Karl Friedl * , Debra Fleischman * , Konstantinos Arfanakis * , Olga James * , Dino Massoglia * , J. Jay Fruehling * , Sandra Harding * , Elaine R. Peskind * , Eric C.

Petrie * , Gail Li * , Jerome A. Yesavage * , Joy L. Taylor *  & Ansgar J. Furst CONTRIBUTIONS Y.J., X.Z., A.K.F. and N.Y.I. conceived the study; F.C.I., Y.C., E.P.T., V.C.M. and T.C.K.

organized recruitment and sample collection; Y.J., H.Y.W., L.O., V.M.C., W.W., D.Y.W., H.S. and N.C.L. performed the experiments; Y.J. and X.Z. set up the data-processing pipelines; Y.J.,

X.Z., H.Y.W., L.O., V.M.C., S.L., W.W., D.Y.W., H.S., W.F., N.C.L., K.Y.M., M.S., B.L., P.M.L., E.O., T.P., S.M.L., J.H., T.W., C.L.M., A.K.F. and N.Y.I. analyzed the data; and Y.J., A.K.F.

and N.Y.I. wrote the manuscript with input from all authors; the ADNI contributed part of the study data. CORRESPONDING AUTHOR Correspondence to Nancy Y. Ip. ETHICS DECLARATIONS COMPETING

INTERESTS All authors declare no competing interests. PEER REVIEW PEER REVIEW INFORMATION _Nature Aging_ thanks Keenan Walker, Sven van der Lee, Alberto Pascual and the other, anonymous,

reviewer(s) for their contribution to the peer review of this work. ADDITIONAL INFORMATION PUBLISHER’S NOTE Springer Nature remains neutral with regard to jurisdictional claims in published

maps and institutional affiliations. EXTENDED DATA EXTENDED DATA FIG. 1 DIAGRAMS ILLUSTRATING IL-33–ST2 SIGNALING. (a) Schematic showing the soluble ST2 (sST2) and full-length ST2 (ST2L)

transcripts. (b) The IL-33/ST2 signaling pathways. IL-33, interleukin 33. EXTENDED DATA FIG. 2 ASSOCIATIONS BETWEEN PLASMA SOLUBLE ST2 LEVELS AND ALZHEIMER’S DISEASE IN THE CHINESE_COHORT_1,

STRATIFIED BY SEX AND _APOE_-Ε4 GENOTYPES. (a) Individual plasma soluble ST2 (sST2) levels stratified by sex and disease phenotype (male: _n_ = 132 healthy controls [HCs], _n_ = 84

individuals with Alzheimer’s disease [AD]; female: _n_ = 204 HCs, _n_ = 193 individuals with AD). Test for effects of sex: _β_ = −3.848 and −3.257 in HCs and individuals with AD,

respectively; test for effects of AD: _β_ = 1.926 and 2.235 in males and females, respectively. (b, c) Individual plasma sST2 levels stratified by _APOE_-ε4 genotypes and disease phenotype

in (b) males (male ε4: _n_ = 28 HCs, _n_ = 29 individuals with AD; male non-ε4: _n_ = 104 HCs, _n_ = 55 individuals with AD), and (c) females (female ε4: _n_ = 25 HCs, _n_ = 71 individuals

with AD; female non-ε4: _n_ = 179 HCs, _n_ = 122 individuals with AD). Test for effects of _APOE_-ε4: _β_ = −1.809, −1.499, −0.248, and 2.425 in male HCs, male individuals with AD, female

HCs, and female individuals with AD, respectively; test for effects of AD: _β_ = 2.034, 1.098, 3.833, and 1.811 in male ε4, male non-ε4, female ε4, and female non-ε4, respectively. ε4,

_APOE_-ε4 carriers; non-ε4, _APOE_-ε4 noncarriers. Data are presented as box-and-whisker plots including maximum, 75th percentile, median, 25th percentile, and minimum values; plus signs (+)

denote corresponding mean values. Linear regression test, adjusted for age, cardiovascular disease status, body mass index (BMI), and education level, with multiple testing correction;

#_FDR_ < 0.05, ##_FDR_ < 0.01, ###_FDR_ < 0.001; *_FDR_ < 0.05, **_FDR_ < 0.01, ***_FDR_ < 0.001. _FDR_, false discovery rate. EXTENDED DATA FIG. 3 ASSOCIATIONS BETWEEN

CEREBROSPINAL FLUID SOLUBLE ST2 LEVELS AND ALZHEIMER’S DISEASE AND AMYLOID-BETA DEPOSITIONS IN THE UKBBN COHORT, STRATIFIED BY SEX. (a) Individual cerebrospinal fluid (CSF) soluble ST2

(sST2) levels stratified by sex and disease phenotype (_n_ = 4 male healthy controls [HCs], _n_ = 36 male individuals with Alzheimer’s disease [AD], _n_ = 7 female HCs, _n_ = 39 female

individuals with AD). Test for effects of sex: _β_ = −1.307 and −0.675 in HCs and individuals with AD, respectively; test for effects of AD: _β_ = 4.019 and 7.766 in males and females,

respectively. Data are presented as box-and-whisker plots including maximum, 75th percentile, median, 25th percentile, and minimum values; plus signs (+) denote corresponding mean values.

Linear regression test, adjusted for age and postmortem duration (PMD), with multiple testing correction; *_FDR_ < 0.05. (b) Associations between amyloid-beta (Aβ) staining in the

postmortem frontal cortex and CSF sST2 levels in male and female individuals with AD (_n_ = 23 males and 28 females in the UKBBN cohort, respectively). The regression lines and 95%

confidence intervals are indicated in red/blue and gray, respectively. Linear regression test, adjusted for age and PMD, with multiple testing correction. Test in males: _β_ = 0.0256,

Pearson’s _r__2_ = 0.0024, _FDR_ = 0.7866; test in females: _β_ = 0.2188, Pearson’s _r__2_ = 0.1053, _FDR_ = 0.0407. _FDR_, false discovery rate. EXTENDED DATA FIG. 4 EFFECTS OF SEX AND AGE

ON SOLUBLE ST2 LEVELS. (a) Individual plasma sST2 levels stratified by sex (_n_ = 216 males, and 397 females from Chinese_cohort_1). _β_ = −3.577 for females vs. males, _P_ = 1.58E − 12. (b)

Associations between plasma sST2 levels and age in males and females (_n_ = 216 males, and 397 females from Chinese_cohort_1). Test in males: _β_ = 0.0134, Pearson’s _r__2_ = 0.0010, _P_ = 

0.8661; test in females: _β_ = 0.1256, Pearson’s _r__2_ = 0.0366, _P_ = 0.0016. (c) Correlations between plasma sST2 level and age in males and females from the INTERVAL and LonGenity cohort

(_n_ = 1,685 males, and 1,616 females from INTERVAL cohort; _n_ = 432 males, and 530 females from LonGenity cohort). Test in INTERVAL cohort (ages 18–76), males: Pearson’s _r__2_ = 0.0169,

_P_ < 0.0001; females: Pearson’s _r__2_ = 0.0001, _P_ = 0.7715. Test in LonGenity cohort (ages 65–94), males: Pearson’s _r__2_ = 0.0100, _P_ = 0.0093; females: Pearson’s _r__2_ = 0.0196,

_P_ = 0.0001. (d) Correlations between CSF sST2 level and age in males and females from the Japanese cohort (_n_ = 68 males, _n_ = 65 females). Test in males: Pearson’s _r__2_ = 0.0441, _P_ 

= 0.0812; test in females: Pearson’s _r__2_ = 0.1521, _P_ = 0.0014. (e, f) Contributions of age and sex to the variance of sST2 levels. Numbers denote the proportions of plasma sST2 (e) and

CSF sST2 (f) variance explained by age, sex, and other factors in the Chinese_cohort_1 and Japanese cohort, respectively. Data in box-and-whisker plots are presented with maximum, 75th

percentile, median, 25th percentile, and minimum values; plus signs (+) denote mean values; data in regression lines are presented as slope (red/blue) and 95% CIs (gray). Statistical tests

were performed by linear regression analysis, additionally adjusted for AD diagnosis, CVD status, BMI, education level in the Chinese_cohort_1. *_P_ < 0.05, **_P_ < 0.01, ***_P_ < 

0.001. EXTENDED DATA FIG. 5 THE RS1921622 A ALLELE IS ASSOCIATED WITH BETTER COGNITIVE PERFORMANCE IN FEMALE _APOE_-Ε4 CARRIERS IN THE AIBL AΒ+ COHORT. Individual cognitive scores in (a)

overall Aβ+ _APOE_-ε4 carriers (_n_ = 122 G carriers [G/-], _n_ = 53 A/A carriers), (b) male Aβ+ _APOE_-ε4 carriers (_n_ = 63 G carriers [G/-], _n_ = 24 A/A carriers), and (c) female Aβ+

_APOE_-ε4 carriers (_n_ = 59 G carriers [G/-], _n_ = 29 A/A carriers) stratified according to rs1921622 genotype. Test in the overall Aβ+ _APOE_-ε4 carriers, _W_ = 3782.0, 3671.0, 3842.0,

and 3319.0 for the effects of rs1921622 genotype on AIBL Preclinical Alzheimer Cognitive Composite (AIBLPACC) score, attention processing score, episodic recall score, and recognition score,

respectively; test in the male Aβ+ _APOE_-ε4 carriers, _W_ = 784.0, 727.0, 823.5, and 740.0 for the effects of rs1921622 genotype on AIBLPACC score, attention processing score, episodic

recall score, and recognition score, respectively; test in the female Aβ+ _APOE_-ε4 carriers, _W_ = 1087.0, 1115.5, 1083.0, and 916.5 for the effects of rs1921622 genotype on AIBLPACC score,

attention processing score, episodic recall score, and recognition score, respectively. Data in box-and-whisker plots are presented with maximum, 75th percentile, median, 25th percentile,

and minimum values; plus signs (+) denote mean values. Statistical tests are performed by Wilcoxon rank-sum test, with multiple testing correction. *_FDR_ < 0.05, **_FDR_ < 0.01,

***_FDR_ < 0.001. _FDR_, false discovery rate. EXTENDED DATA FIG. 6 THE RS1921622 A ALLELE IS ASSOCIATED WITH SLOWER GRAY MATTER ATROPHY IN FEMALE _APOE_-Ε4 CARRIERS IN THE AIBL AΒ+

COHORT. (a, b) Individual baseline (a: _n_ = 128 G carriers [G/-], and 55 A/A carriers) and longitudinal gray matter volume (b: _n_ = 148 datapoints from 37 G/- carriers, and 55 datapoints

from 14 A/A carriers) in overall Aβ+ _APOE_-ε4 carriers stratified by rs1921622 genotype. Test on baseline data: _W_ = 4109.0, _FDR_ = 0.109. Test on longitudinal data: _β_ = −0.089 and

−0.051 in G/- and A/A carriers, respectively; _F_ = 3.830, _FDR_ = 0.087. (c, d) Individual baseline (c: _n_ = 64 G/- carriers, and 24 A/A carriers) and longitudinal gray matter volume (d:

_n_ = 98 datapoints from 25 G/- carriers, and 36 datapoints from 9 A/A carriers) in male Aβ+ _APOE_-ε4 carriers stratified by rs1921622 genotype. Test on baseline data: _W_ = 852.0, _FDR_ = 

0.434. Test on longitudinal data: _β_ = −0.069 and −0.053 in G/- and A/A carriers, respectively; _F_ = 0.272, _FDR_ = 0.607. (e, f) Individual baseline (e: _n_ = 64 G/- carriers, and 31 A/A

carriers) and longitudinal gray matter volume (f: _n_ = 50 datapoints from 12 G/- carriers, and 19 datapoints from 5 A/A carriers) in female Aβ+ _APOE_-ε4 carriers stratified by rs1921622

genotype. Test on baseline data: _W_ = 1240.0, _FDR_ = 0.109. Test on longitudinal data: _β_ = −0.159 and −0.027 in G/- and A/A carriers, respectively; _F_ = 8.804. Data in box-and-whisker

plots are presented with maximum, 75th percentile, median, 25th percentile, and minimum values; plus signs (+) denote mean values; data in regression lines are presented as slope (red/blue)

and 95% confidence intervals (gray). Statistical tests on baseline data were performed by Wilcoxon rank-sum test, with multiple testing correction; and statistical tests on longitudinal data

were performed by linear mixed model test, adjusted for baseline age and MRI scanners, with multiple testing correction. *_FDR_ < 0.05, **_FDR_ < 0.01, ***_FDR_ < 0.001. EXTENDED

DATA FIG. 7 THE RS1921622 A ALLELE RESTORES IMPAIRED MICROGLIAL ACTIVITIES TOWARD AMYLOID-BETA IN FEMALE _APOE_-Ε4 CARRIERS WITH ALZHEIMER’S DISEASE. (a, b) Representative images (a) and

quantification (b) of Aβ plaques in the frontal cortices of individuals with AD stratified by sex and _APOE_-ε4 genotype (_n_ = 14 male _APOE-_ε4 noncarriers [non-ε4], 22 male _APOE_-ε4

carriers [ε4], 19 female non-ε4, and 23 female ε4 from UKBBN cohort). Test for sex: _β_ = 2.871 (non-_APOE-_ε4) and 4.927 (_APOE-_ε4), _P_ = 2.09E − 2 (non_-_ε4) and 4.61E − 5 (ε4); for

_APOE_-ε4: _β_ = −0.621 (males) and 1.967 (females), _P_ = 4.22E − 1 (males) and 4.84E − 2 (females). Scale bar, 200 μm. (c) Representative images of Aβ plaques in the frontal cortices of

females with AD stratified by _APOE_-ε4 and rs1921622 genotype. Scale bar, 5 mm. (d, e) Representative image (d) and quantification (e) of Aβ plaques (brown) and Iba-1+ microglia (purple) in

the frontal cortices of individuals with AD stratified by sex and _APOE_-ε4 genotype (_n_ = 14 male non-ε4, 22 male ε4, 19 female non-ε4, and 23 female ε4). Test for sex: _β_ = −1.214

(non-ε4) and −4.133 (ε4), _P_ = 4.61E − 1 (non-ε4) and 3.52E − 4 (ε4); for _APOE_-ε4: _β_ = −0.965 (males) and −2.818 (females), _P_ = 3.27E − 1 (males) and 6.00E − 3 (females). Scale bar,

100 μm. (f, g) Representative image (f) and quantification (g) of Aβ plaques and microglia in the frontal cortices of females with AD stratified by _APOE_-ε4 and rs1921622 genotype (_n_ = 4 

G/G and 15 A/- carriers among non-ε4, 5 G/G carriers and 18 A/- carriers among ε4, respectively). Test for _APOE_-ε4: _β_ = −3.745 (G/G) and −1.944 (A/-), _P_ = 3.09E − 2 (G/G) and 9.32E − 2

(A/-); for rs1921622: _β_ = 0.781 (non-ε4) and 2.017 (ε4), _P_ = 7.70E − 1 (non-ε4) and 2.53E − 2 (ε4). Scale bar, 100 μm. Data in box-and-whisker plots are presented with maximum, 75th

percentile, median, 25th percentile, and minimum values; plus signs (+) denote mean values. Statistical tests were performed by linear regression analysis, adjusted for age, PMD. *_P_ < 

0.05, **_P_ < 0.01, ***_P_ < 0.001; #_P_ < 0.05, ##_P_ < 0.01, ###_P_ < 0.001. EXTENDED DATA FIG. 8 THE RS1921622 A ALLELE AND CEREBROSPINAL FLUID SOLUBLE ST2 LEVEL EXHIBIT

OPPOSITE EFFECTS ON THE MICROGLIAL TRANSCRIPTOME IN FEMALE _APOE_-Ε4 CARRIERS WITH ALZHEIMER’S DISEASE. Scatterplot showing the correlation between the normalized effect size (_β_) of

cerebrospinal fluid (CSF) soluble ST2 (sST2) levels and the rs1921622 A allele on microglial gene expression in the frontal cortices of female _APOE_-ε4 carriers with Alzheimer’s disease

(AD) (_n_ = 2,636 microglia from 8 individuals from the UKBBN cohort). EXTENDED DATA FIG. 9 ELEVATED BRAIN SOLUBLE ST2 LEVEL EXACERBATES AMYLOID-BETA DEPOSITION IN FEMALE BUT NOT MALE 5XFAD

MICE. (a, b) Amyloid-beta (Aβ) deposition in the cortices of 4-month-old male and female 5XFAD mice after 28-day intracerebroventricular delivery of soluble sST2 (sST2)-Fc or Fc as a

control. (a) Representative images of Aβ staining. Scale bar, 1 mm. (b) Quantification of Aβ plaques (% of total cortical area) (control: _n_ = 6 male and 5 female mice, sST2: _n_ = 6 male

and 5 female mice). _T_ = 0.590 (males) and 2.721 (females); _P_ = 0.568 (males) and 0.026 (females). Data in bar charts are mean + SEM. Statistical tests were performed as two-sided

unpaired Student’s _t_-tests. *_P_ < 0.05, **_P_ < 0.01, ***_P_ < 0.001. EXTENDED DATA FIG. 10 ELEVATED BRAIN SOLUBLE ST2 LEVEL LEADS TO DECREASED AMYLOID-BETA PHAGOCYTOSIS BY

MICROGLIA IN FEMALE 5XFAD MICE. (a, b) Microglial amyloid-beta (Aβ) uptake activity in the cortices of 4-month-old female 5XFAD mice after 28-day intracerebroventricular delivery of soluble

ST2 (sST2)-Fc or Fc as a control. Representative distribution (a) and quantification (b) show the mean fluorescent intensity (MFI) of methoxy-X04-labeled Aβ in the methoxy-X04+ CD11b+

microglia (control: _n_ = 7 mice, sST2: _n_ = 7 mice). _T_ = − 2.490 and −2.014 for the tests in microglial with medium Aβ uptake (MeX04MedCD11b+; _left_) and high Aβ uptake

(MeX04HighCD11b+; _right_), respectively; _P_ = 0.028 and 0.067, respectively. Data in bar charts are mean + SEM. Statistical tests were performed as two-sided unpaired Student’s _t_-tests.

*_P_ < 0.05, **_P_ < 0.01, ***_P_ < 0.001. SUPPLEMENTARY INFORMATION SUPPLEMENTARY INFORMATION Supplementary Notes 1–9, Figs. 1–8 and Tables 1–9 REPORTING SUMMARY SUPPLEMENTARY DATA

1 Supplementary Data 1–4. SUPPLEMENTARY VIDEO 1 3D reconstruction for Fig. 7f. SOURCE DATA SOURCE DATA FIG. 2 Statistical source data for Fig. 2a. SOURCE DATA FIG. 6 Statistical source data

for Fig. 6e–g. RIGHTS AND PERMISSIONS OPEN ACCESS This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation,

distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and

indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to

the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will

need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. Reprints and permissions ABOUT THIS ARTICLE

CITE THIS ARTICLE Jiang, Y., Zhou, X., Wong, H.Y. _et al._ An _IL1RL1_ genetic variant lowers soluble ST2 levels and the risk effects of _APOE_-ε4 in female patients with Alzheimer’s

disease. _Nat Aging_ 2, 616–634 (2022). https://doi.org/10.1038/s43587-022-00241-9 Download citation * Received: 16 June 2021 * Accepted: 23 May 2022 * Published: 15 July 2022 * Issue Date:

July 2022 * DOI: https://doi.org/10.1038/s43587-022-00241-9 SHARE THIS ARTICLE Anyone you share the following link with will be able to read this content: Get shareable link Sorry, a

shareable link is not currently available for this article. Copy to clipboard Provided by the Springer Nature SharedIt content-sharing initiative