The impact of blood MCP-1 levels on Alzheimer’s disease with genetic variation of UNC5C and NAV3 loci

Abstract Background Previous study shows that monocyte chemoattractant protein-1 (MCP-1), which is implicated in the peripheral proinflammatory cascade and blood-brain barrier (BBB) disruption, modulates the genetic risks of AD in established AD loci. Methods In this study, we hypothesized that blood MCP-1 impacts the AD risk of genetic variants beyond known AD loci. We thus performed a genome-wide association study (GWAS) using the logistic regression via generalized estimating equations (GEE) and the Cox proportional-hazards models to examine the interactive effects between single nucleotide polymorphisms (SNPs) and blood MCP-1 level on AD in three cohorts: the Framingham Heart Study (FHS), Alzheimer’s Disease Neuroimaging Initiative (ADNI) and Religious Orders Study/Memory and Aging Project (ROSMAP). Results We identified SNPs in two genes, neuron navigator 3 ( NAV3 , also named Unc-53 Homolog 3, rs696468) (p < 7.55×10 − 9 ) and Unc-5 Netrin Receptor C ( UNC5C rs72659964) (p < 1.07×10 − 8 ) that showed an association between increasing levels of blood MCP-1 and AD. Elevating blood MCP-1 concentrations increased AD risk and AD pathology in genotypes of NAV3 (rs696468-CC) and UNC5C (rs72659964-AT + TT), but did not influence the other counterpart genotypes of these variants. Conclusions NAV3 and UNC5C are homologs and may increase AD risk through dysregulating the functions of neurite outgrowth and guidance. Overall, the association of risk alleles of NAV3 and UNC5C with AD is enhanced by peripheral MCP-1 level, suggesting that lowering the level of blood MCP-1 may reduce the risk of developing AD for people with these genotypes.


Introduction
The relationship between peripheral proin ammatory factors and late-onset Alzheimer's disease (AD) is largely unclear, and existing studies suggest that most of these factors have no or modest sensitivity and speci city for the prediction of AD risk.(1) Due to the blood-brain barrier (BBB), the effects of blood in ammatory factors on the brain are heavily regulated and vary between factors.Furthermore, in ammatory factors may interact with inherited genetic factors to modulate the risk for AD.For example, using the Framingham Heart Study (FHS), we found that within carriers of the AD risk genotype, apolipoprotein E4 (APOE 4), elevated blood C-reactive protein (CRP) levels are associated with increased AD risk.(2) This phenomenon is not observed among APOE 2 and APOE 3 carriers.Additionally, in ApoE4 carriers, elevated CRP is associated with the AD biomarker phosphorylated Tau (p-Tau) in the cerebral spinal uid (CSF).(3) Peripheral chronic in ammation causes a cascade of cytokines and chemokines to surge.Another CRPrelated proin ammatory factor is monocyte chemoattractant protein-1 (MCP-1), which is also known as C-C motif chemokine ligand 2 (CCL2), is expressed in the blood and brain, and is implicated in in ammatory cell recruitment and BBB disruption.(4)MCP-1 is produced by macrophages and activated astrocytes during in ammation within the central nervous system.The expression of MCP-1 is also enhanced by CRP during chronic in ammation.(5,6) The main function of MCP-1 as a chemoattractant is to drive leukocyte migration, especially monocytes/macrophages into damaged or infected tissues including AD brains (7)(8)(9), indicating that MCP-1 may play a role in activating microglia in the brain and thus leading to cognitive decline.(10)(11)(12)(13)(14)(15)(16) As reported, MCP-1 level in human CSF is associated with a faster rate of cognitive decline during the early stages of AD (13), and its overexpression in the brain promotes glial activation and accelerates tau pathology in a mouse model.(17) In addition, a higher plasma MCP-1 level is associated with greater severity of AD and mild cognitive impairment (MCI) and faster cognitive decline.(14) Recently, Cherry & Stein et al.(16) reported that MCP-1 protein levels in the dorsolateral prefrontal cortex (DLPFC) are correlated with the density of Iba1 + cells and CD68 + cells, increased chronic traumatic encephalopathy (CTE) severity, and are correlated with pTau independent of age at death and Aβ42 in AD and CTE.
However, the relationship between blood MCP-1 and the risk of developing AD is not consistent across different studies.(1) We hypothesized that the levels of blood MCP-1 differently impact vulnerable people with AD risk genes, to increase AD risk.Our previous study discovered that the blood MCP-1 level can modulate the genetic risks of AD for two established AD gene loci, APOE and HLA-DRB1.(18)However, there is no genome-wide search for the interaction effects between blood MCP-1 levels with genetic variants on AD risk.In this study, we conducted a Gene-by-Environment (G E) genome-wide association study (GWAS) by applying logistic regression utilizing generalized estimating equations (GEE).Top ndings were further tested with Cox proportional hazards models for the incidence of AD.By analyzing data from the FHS and the Alzheimer's Disease Neuroimaging Initiative (ADNI), for the rst time, at the genome-wide signi cance level (p < 5.0×10 − 8 ) we found that elevated blood MCP-1 increased AD risk, hippocampal atrophy, and AD neuropathology only in people carrying one genotype of SNPs in two gene loci: Neuron navigator 3 (NAV3 rs696468) and Unc-5 Netrin Receptor C (UNC5C rs72659964).Interestingly, both genes have been identi ed in C. elegans to be involved in neuron outgrowth and guidance and both are associated with AD in humans.(19,20) Further explorations in the ROSMAP cohort provided molecular mechanism evidence of the signi cant associations between AD and variant in NAV3 at multi-omics levels (i.e., DNA methylation, gene expression).

Religious Orders Study/Memory and Aging Project (ROSMAP)
ROSMAP data was further used to investigate whether cis-regulatory variants affect DNA methylation and gene expression in brain and their relationships with AD (Fig. 1).Plasma MCP-1 measurement MCP-1 levels (Exam 7) in FHS participants were measured using enzyme-linked immunosorbent assay (ELISA) with a Dade Behring BN100 nephelometer(26) from fasting blood samples that were collected at exam 7 from the antecubital vein (details had been previously described).(27) In ADNI-1, MCP-1 in plasma samples was collected using the Human Discovery MAP Panel and measurement platform.(28,29) MCP-1 level was log-transformed in the downstream analysis as a continuous variable.The median of the MCP-1 level was also used to de ne participants into low versus high MCP-1 groups as a categorical variable.

AD neuropathology
A subset of the FHS Offspring participants (n = 105) donated their brains, which were used for neuropathological characterization.For data analysis, three routine neuropathology variables were selected, including the Braak stage for neuro brillary degeneration, the CERAD score for density of neocortical neuritic plaques, and the CERAD semi-quantitative score for diffuse plaques.
Neuropathological brain evaluation was performed by neuropathologists blinded to all demographic and clinical information.(30) 78 participants had the above neuropathology variables and blood MCP-1 measurement were included in the analysis (Figure S1).

Genome-wide association studies in the FHS
The genome-wide association studies (GWAS) were performed using the Framingham analytical pipeline for common autosomal single nucleotide polymorphisms (SNPs) imputed using MACH with the 1000G European Ancestry reference panel from March 2012.In brief, a total number of 412,053 genotyped SNPs were used as input to the MACH program (http://genome.sph.umich.edu/wiki/minimac)for phasing.The nal genotype set was obtained by following the GIANT protocol for imputation (31) to the November 2010 release of 1000G EUR panel based only on individuals of European descent.The relationship between AD and the interaction term of MCP-1 and SNP dosage was tested using GEEPACK (logistic regression utilizing generalized estimating equations; accounts for relatedness in FHS) while adjusting for age, sex, years of education, and the rst 10 principal components (PCs).Additive models were assumed.Only autosomal SNPs (chr1-22) with minor allele frequency (MAF) ≥ 5% were chosen for the analysis to reduce false positives due to the sample size of FHS.Additionally, samples with genotyping rate less than 97% and with an excess number of heterozygote observations (P < 10 − 6 ) or Mendelian errors were removed.The Manhattan plot, QQ plot and genomic control (32) were used for visualization quality control and accounting for genomic in ation.LocusZoom (33) was used to present the regional information.P values < 5.0×10 − 8 were considered genome-wide statistically signi cant for the SNP-MCP-1 interactive effects for AD.

DNA methylation and Gene expression analysis
ROSMAP data was used to investigate whether cis-regulatory variants (SNPs within 5KB for mQTL and 1 MB for eQTL of neighboring genes) affect DNA methylation (mQTL) and gene expression (eQTL) in brain as well as their relationships with AD. mQTL summary results for selected variants were obtained from Brain xQTLServe for ROSMAP (http://mostafavilab.stat.ubc.ca/xQTLServe/).Genotype data were generated from 2,093 individuals of European descent.Of these individuals, DNA methylation array (the Illumina In nium HumanMethylation 450K BeadChip) data was derived from fresh frozen post-mortem brain samples collected from the dorsolateral prefrontal cortex (DLPFC) region of 468 subjects.Gene expression RNA-seq (Illumina HiSeq) data were generated from the DLPFC of 494 individuals with an average sequencing depth of 90 million reads.A detailed description was previously described.(34) Furthermore, we examined the gene expression patterns across different AD brain regions using the Agora database (https://agora.adknowledgeportal.org/genes),which includes Cerebellum (CBE), DLPFC, Frontal Pole (FP), Inferior Frontal Gyrus (IFG), Parahippocampal Gyrus (PHG), Superior Temporal Gyrus (STG) and Temporal Cortex (TCX).This database was initially developed by the NIA-funded AMP-AD consortium that shared evidence in support of AD target discovery.

Statistical analysis
Analyses were performed using the R statistical environment (R 3.6.2).Several variables including sample size, age at baseline, sex, years of education, APOE ε4 status, and incident AD status were summarized as the baseline characteristics strati ed by MCP-1 levels.Group differences were assessed by analysis of variance (ANOVA) for normally distributed continuous variables, by Kruskal-Wallis rank sum test for continuous variables with skewed distributions, and by Chi-Square test of independence for categorical variables.
In strati ed genotype analysis, logistic regression was performed for AD status (prevalent + incident AD), using MCP-1 level (log-transformed continuous or high vs. low group) as the predictor, and adjusting for covariates (age at MCP-1 measurement, sex, years of education, and the rst 5 PCs).Additionally, both Cox proportional hazards regression and Kaplan-Meier survival analysis were conducted for AD incidence.Heterozygous and homozygous minor allele genotypes were combined in this section due to the relatively low MAF of the two SNPs (0.14 for rs696468 and 0.06 for rs72659964).
For neuropathology, ordinal regression was performed to examine whether the interactions between SNPs and MCP-1 associate with AD neuropathology traits including Braak stage, diffuse plaque CERAD score, neuritic plaque CERAD score, adjusting for the same covariates as described above.
To support the ndings in the FHS, the ADNI-1 cohort was analyzed using logistic regression models adjusted for age at baseline, sex, and years of education, then the results were meta-analyzed with that of FHS by inverse-variance weighted meta-analysis using METAL.(35)

Characteristics of the FHS population
The 2,884 FHS Generation 2 participants included in this study (Table 1) had an average follow-up period of 18.5 years (from exam 7), and 171 (5.93%) of the participants developed AD during this follow-up.Subjects were divided into four quartiles based on the MCP-1 serum concentration measured at baseline (Exam 7).Participants with the lowest MCP-1 concentration ( rst quantile) were the youngest (p < 0.001) and had the highest years of education (p < 0.001), compared to participants with higher MCP-1 quartiles.
Sex and APOE 4 genotypes did not show signi cant differences across MCP-1 quartiles.Although there was an overall increase in the incidence of AD with increasing concentration of blood MCP-1 quartile, the relationship was not linear (3.61% vs. 6.38% vs. 4.72% vs. 9.02%, p < 0.001), (Table 1).In addition, when directly testing the associations between continuous MCP-1 concentration and AD incidence after adjusting for confounders including age, sex and years of education, no statistically signi cant relationship was found (data not shown).Since we found that MCP-1 impacts the AD risk of genetic variants in APOE and HLA-DRB1 differently(18), we hypothesized that blood MCP-1 might modulate the effects of other genetic variants on AD risk.

Interactions between GWAS-selected SNPs and blood MCP-1 on AD risk in the FHS
To genome-wide search genetic loci interacting with peripheral blood MCP-1 level to affect AD risk beyond established AD loci, we conducted a G E GWAS for AD with an interaction term between MCP-1 × and SNP as predictor using a logistic regression model (Fig. 2and Figure S1).As shown in the Manhattan plot after genomic control (Fig. 2A), 20 SNPs showed suggestive signi cance (p < 1.0×10 − 7 ).One locus on chromosome 12 showed genome-wide signi cant interactions with MCP-1 on AD risk (p < 5.0×10 − 8 ) (Table S1), including 18 SNPs in intronic region of gene NAV3 (rs696468 (MAF = 0.14) as sentinel SNP) as shown in Fig. 2B.Another locus on chromosome 4 reached suggestive signi cance (p < 1.0×10 − 7 )   including two SNPs in intronic region of UNC5C (rs72659964 (MAF = 0.06) as sentinel SNP), as presented in Fig. 2C.
Among SNPs passing a genome-wide signi cance (p < 5.0×10 − 8 ), sentinel SNP rs696468 in NAV3 was selected for further analyses.Another signi cant SNP rs72659964 (p < 1.0×10 − 7 in FHS and p < 5.0×10 − 8 in meta-analysis) in UNC5C was also selected for further analyses since UNC5C is a homolog of NAV3 and is associated with familial and sporadic AD.(36-38)As shown in Table 2, the SNP main effect alone on AD risk was not signi cant (p > 0.05).However, their interaction effects with the continuous MCP-1 level (log-transformed) on AD is signi cant (p = 1.64×10 − 8 for NAV3 and p = 8.36×10 − 8 for UNC5C in FHS) using logistic regression models.We next performed SNP strati ed analysis to test the association between MCP-1 and AD incidence using Cox proportional hazards regression models for AD risk after adjusting for age at baseline, sex, years of education, and PCs.Due to the relatively low MAF of the two SNPs (0.14 for rs696468 and 0.06 for rs72659964), heterozygous and homozygous minor allele genotypes were combined.As shown in Table S2-3, elevated MCP-1 concentration (log-transformed) was associated with a higher incidence of AD among NAV3 rs696468-CC carriers (HR = 2.68, 95% CI = 1.55, 4.62, p = 3.9×10 − 4 ).On the other hand, the elevated MCP-1 was negatively associated with AD risk among NAV3 rs696468-CT + TT carriers (HR = 0.25, 95% CI = 0.10, 0.60, p = 0.002).These dose-dependent relationships were also observed using different MCP-1 percentile cutoffs (Fig. 3A).Elevated MCP-1 concentration (log-transformed and percentile cutoffs) was associated with a higher incidence of AD among UNC5C rs72659964-AT + TT carriers (HR = 80.74, 95% CI = 8.32, 783.28, p = 1.5×10 − 4 ), but this increasing trend was largely attenuated for the UNC5C rs72659964-AA carriers (Table S2 and Fig. 3B).In Kaplan-Meier analyses, we further strati ed subjects into low and high MCP-1 group, and signi cantly lower AD-free probability was observed among subjects having high blood MCP-1 (75% percentile as cutoffs) and among subjects with NAV3 rs696468-CC carriers (p = 3.6×10 − 5 , Fig. 3C) and subjects with UNC5C rs72659964-AT + TT carriers (p = 6.1×10 − 4 , Fig. 3D) (50% percentile as cutoffs in Figure S2).In contrast, elevated MCP-1 showed no or attenuated positive associations with AD risk among persons with other genotypes for these SNPs.c.Fixed effect model was applied.

Validation of the interactive effects of blood MCP-1 and SNPs in NAV3 and UNC5C for AD in the ADNI cohort
To replicate the ndings in the FHS, we analyzed the ADNI-1 cohort in the same way as for the FHS using logistic regression models and adjusted for the same confounders.We then meta-analyzed the summary statistic results from the FHS and ADNI-1 with the inverse-variance weighted method for the 21 suggestive signi cant SNPs (Table 2, Table 3 and Table S4).Indeed, this analysis increased the signi cance as we obtained genome-wide signi cant interactions for all the selected SNPs (P < 5.0×10 − 8 ) (Table 2 and Table S4).In genotype strati cation analysis, we observed the same direction effects of elevated blood MCP-1 for increasing AD risk in rs696468-CC (Z = 3.94, p = 8.1×10 − 4 ) and UNC5C rs72659964-AT + TT (Z = 3.35, p = 8.0×10 − 4 ) carriers in both cohorts (Table 3), while in other genotypes the associations were inconsistent or insigni cant between the two cohorts.FHS and ADNI-1 participants were divided into subgroups based on the genotype of NAV3 rs696468 CC vs. CT + TT or the genotype of UNC5C rs72659964 AA vs. AT + TT, and an analysis of the effects of blood MCP-1 levels (log-transformed) on AD was performed after strati cation by genotype.Logistic regression model was used with adjustments for age, sex and years of education.Heterozygous and homozygous minor allele genotypes were combined due to the relatively low MAF of the two SNPs (0.14 for rs696468 and 0.06 for rs72659964).
b. AD ~ log (MCP-1) + age + sex + years of education, strati ed by different genotypes with xed-effect model.

Associations of blood MCP-1 and brain neuropathology among different genotypes
To further investigate the impact of peripheral MCP-1 on AD brain pathology, the relationships between blood MCP-1, selected variants, and three neuropathology features were evaluated using ordinal regression in the FHS, in which 78 participants donated their brain after death.As shown in Table 4, consistently, NAV3 rs696468 signi cantly interacted with blood MCP-1 concentration for Braak score (p = 0.02), CERAD score (neocortical neuritic plaque) (p = 0.03), and CERAD semi-quantitative score (diffuse plaques) (p = 0.03).UNC5C rs72659964 signi cantly interacted with blood MCP-1 concentration for CERAD semi-quantitative score (diffuse plaques) (p < 0.001).The main effects of the two SNPs without counting peripheral MCP-1 for AD pathology were not signi cant.

Exploration of regulatory functions of variants in NAV3 UNC5C underlying AD
To investigate possible regulatory mechanism of rs696468 in NAV3, and whether NAV3 is the function target gene of rs696468, we rst performed mQTL and eQTL mapping by leveraging the available genotype, and DNA methylation and RNA-seq gene expression measured from the same DLPFC brain region in ROSMAP.(Fig. 4).Interestingly, NAV3 rs696468 was negatively associated with methylation level of CpG site cg20521863 (located within NAV3) (p = 0.003) (Fig. 4A), and the cg20521863 methylation level was negatively associated with NAV3 gene expression (p = 1.2×10 − 5 ) (Fig. 4B).Of note, AD had a higher methylation level of cg20521863 (p = 0.01) compared with normal controls in DLPFC region of ROSMAP participants (Fig. 4C).Consistently, using the Agora database, we discovered that NAV3 had a signi cantly lower level of expression in DLPFC (p = 0.02) and temporal cortex region (TCX) (p = 8.7×10 − 6 ) among AD cases compared with normal controls (Fig. 4E).Both NAV3 and UNC5C showed relatively high expression levels in brain related tissues in the Genotype-Tissue Expression (GTEx) database (Figure S3).
Meanwhile, we did not observe signi cant associations for UNC5C rs72659961/rs72659964 in any molecular QTL mapping databases we studied (data not shown).The regulatory functions of these variants should be further investigated in the future.

Discussion
Our study conducted G×E GWAS analysis to investigate the interactive in uence of genetic and internal environmental factors on AD development.Through this analysis, we discovered two genetic loci, NAV3 rs696468 and UNC5C rs72659964, that are implicated in AD under the in uence by blood MCP-1 levels (Fig. 2, Table 2).Speci cally, we found that elevated blood MCP-1 levels were associated with a higher risk of AD, but only among individuals who carry speci c alleles of the NAV3-rs696468-CC and UNC5C-rs72659964-AT + TT (Fig. 3, Table S2, and Table 3).In contrast, individuals with different genotypes of these genetic variants did not exhibit the same association between MCP-1 levels and AD risk.
Additionally, our study revealed a correlation between increased blood MCP-1 concentrations and older age (Table 1).This nding aligns with previous research, indicating that advancing age is linked to peripheral chronic in ammation leading to neurodegeneration.(39) Consequently, our study proposes a novel mechanism whereby proin ammatory factors in the peripheral system in aging may heighten the risk of neurodegenerative changes in the brain.
NAV3, also named Unc-53 Homolog 3, is a homolog of UNC5C which has a similar cellular function as UNC5C in the nervous system in C. elegans.(40) It is reported that some microRNAs, e.g., miR-29c, regulates NAV3 protein expression in an AD mouse model.(41) In a recent study, the association of AD and NAV3 was further highlighted by an AI-based approach using data from ROSMAP and Mayo RNAseq Study (https://adknowledgeportal.org).(42) Our further explorations in the ROSMAP cohort at epigenomic and transcriptomic levels (Fig. 4) indicate that NAV3 rs696468 plays a role in DNA methylation, which, in turn, is associated with NAV3 expression.Interestingly, we observed negative associations between NAV3 expression and AD in DLPFC and temporal cortex.It is possible that blood MCP-1 levels may act as a moderator for the association between AD and NAV3 expression, which is shown to be impacted by the NAV3 genetic loci.
NAV3 and UNC5C are homologs and have been identi ed in C. elegans to have neuron outgrowth and guidance and are associated with AD in humans.(19,20,43,44) Mechanistic studies on cell and animal models discovered that aberrant UNC5C might contribute to AD by activating death-associated protein kinase 1 (DAPK1) which is involved in AD pathogenesis with extensive involvement in aberrant tau, Aβ and neuronal apoptosis/autophagy. (20,45) In addition, deleting UNC5C from netrin-1-depleted mice can mitigate AD pathologies and reduces cognitive disorders.The δ-secretase truncates UNC5C and increases its neurotoxicity, contributing to AD pathogenesis.(36)In brain imaging, UNC5C loci has been reported to be associated with temporal volume and alter the atrophy of strategic regions of AD such as the

Figure 3 Kaplan-
Figure 3 The ROS is a longitudinal, epidemiologic clinical-pathological study of memory, motor, and functional problems in older Catholic nuns, priests, and brothers aged 65 years and older from across the United States.

Table 1
Basic characteristics, APOE, CRP and incidence of AD among different MCP-1 level groups in FHS A total of 2,884 subjects in the FHS were divided into four quartiles based on blood MCP-1 levels in the analysis.Means (SD) and Medians (Q1 = Q3) were reported.ANOVA was used to analyze continuous variables, while n (%) with the χ 2 test was used for categorical variables for the MCP-1 quartile comparisons.P values indicating statistical signi cance are shown.Abbreviations: AD, Alzheimer's disease; MCP-1, Monocyte Chemoattractant Protein-1.

Table 2
The relationships between AD, variants and the interaction of both genotypes and MCP-1 using logistic regression a. AD ~ SNP dosage + age + sex + years of education + PCs b.AD ~ SNP dosage:MCP-1 (log transformed) + SNP dosage + MCP-1 + age + sex + years of education + PCs; GWAS results using GEEPACK.

Table 3
Strati ed genotype analysis of MCP-1 levels on AD, adjusted by age, sex, years of education using logistic regression

Table 4
Relationship between Neuropathology scores and NAV3 or UNC5C alone vs. the interaction of either genotype with MCP-1 concentration using ordinal regression