APOE alleles’ association with cognitive function differs across Hispanic/Latino groups and genetic ancestry in the study of Latinos‐investigation of neurocognitive aging (HCHS/SOL)

Abstract Introduction Apolipoprotein E (APOE) alleles are associated with cognitive decline, mild cognitive impairment (MCI), and Alzheimer's disease in Whites, but have weaker and inconsistent effects reported in Latinos. We hypothesized that this heterogeneity is due to ancestry‐specific genetic effects. Methods We investigated the associations of the APOE alleles with significant cognitive decline and MCI in 4183 Latinos, stratified by six Latino backgrounds, and explored whether the proportion of continental genetic ancestry (European, African, and Amerindian) modifies these associations. Results APOE ε4 was associated with an increased risk of significant cognitive decline (odds ratio [OR] = 1.15, P‐value = 0.03), with the strongest association in Cubans (OR = 1.46, P‐value = 0.007). APOE‐ε2 was associated with decreased risk of MCI (OR = 0.37, P‐value = 0.04) in Puerto Ricans. Amerindian genetic ancestry was found to protect from the risk conferred by APOE ε4 on significant cognitive decline. Discussion Results suggest that APOE alleles' effects on cognitive outcomes differ across six Latino backgrounds and are modified by continental genetic ancestry.

(OR = 1.46, P-value = 0.007). APOE-ε2 was associated with decreased risk of MCI (OR = 0.37, P-value = 0.04) in Puerto Ricans. Amerindian genetic ancestry was found to protect from the risk conferred by APOE ε4 on significant cognitive decline.
Discussion: Results suggest that APOE alleles' effects on cognitive outcomes differ across six Latino backgrounds and are modified by continental genetic ancestry.

K E Y W O R D S
admixture, Alzheimer's disease, ancestry, apolipoprotein E, cognitive decline, genetic epidemiology, Hispanics/Latinos, mild cognitive impairment

BACKGROUND
Cognitive decline, mild cognitive impairment (MCI), and Alzheimer's disease and related dementias (ADRD) are a growing worldwide epidemic and one of the leading causes of death in the elderly population. 1 MCI is a prodromal cognitive impairment state preceding the more serious cognitive dysfunction characteristic of dementia. It can involve problems with memory, language, thinking, and judgment that are greater than normal age-related changes. 2,3 Cognitive decline is a normal process of aging; however, in ADRD patients it begins many years before dementia is diagnosed and accelerates during the course of the disease. 3 Self-reported cognitive decline has lately been introduced to the field, to extend ADRD risk diagnosis to an earlier stage before MCI. 4 Hispanics/Latinos (Latinos henceforth) are the fastest-growing ethnic group in the United States, 5 and suffer from higher rates of ADRD, compared to Whites. 6,7 The apoplipoprotein E (APOE) ε4 allele is the strongest known genetic risk factor for ADRD, 8 and it has also been linked to more rapid cognitive aging, such as increased cognitive decline and MCI. [9][10][11] The APOE gene exists as three polymorphic allelesε2, ε3, and ε4which are determined by two single nucleotide polymorphisms (SNPs: rs429358 and rs7412), that substitute amino acids in the protein, resulting in functional changes. In general, APOE ε4 confers increased risk for cognitive decline, MCI, and ADRD compared to the more common APOE ε3, whereas the APOE ε2 is considered neuroprotective. 12 24 We hypothesized that there are differential association effects of the APOE alleles in the six Latino background groups on significant cognitive decline and MCI, which could potentially explain the inconsistent reported literature results of the association of APOE alleles with cognitive function, MCI, and dementia in Latinos. To address this, we tested the association of APOE alleles with significant cognitive decline and MCI and stratified the analyses by the six Latino background groups. Next, we hypothesized that differences in proportions of continental ancestries in the six Latino background groups will explain the heterogeneous APOE alleles' effects on significant cognitive decline and MCI. The purpose of this study was to determine whether genetic ancestry modifies the effect of APOE alleles on the significant cognitive decline and MCI by testing interaction effects between the three Latino continental ancestries and the APOE alleles.

Study population
The HCHS/SOL is a multisite, prospective cohort study of diverse All participants in this analysis signed informed consent in their preferred language (Spanish/English) to use their genetic and non-genetic data. The study was reviewed and approved by the institutional review boards at all collaborating institutions.

Cognitive outcomes
We analyzed two binary cognitive variables that were previously constructed 30 based on cognitive tests and self-report: significant 2. Interpretation: Our findings suggest a differential association between the APOE ε2 and APOE ε4 alleles and risk for significant cognitive decline and MCI in the six Latino groups defined by country of origin. We also found that Amerindian genetic ancestry protects from the risk conferred by APOE ε4 on significant cognitive decline.
Inconsistent estimated associations in Latinos may be due to different admixture patterns of continental ancestry across Latino groups. Aging-Alzheimer's Association if they fit the following three criteria: 31 (1) a cognitive test score below -1SD based on SOL-INCA robust internal norms, (2) significant cognitive decline (described above), and (3) self-reported cognitive decline based on the E-Cog12. 28 Also, individuals that had both a cognitive deficit < -2SD in any neurocognitive test at the SOL-INCA exam and more than minimal cognitive impairment in the IADL scale, were classified as MCI. Additional information about the SOL-INCA cognitive assessment approach is provided in detail in Appendix 1 of a previous publication. 30

Genetic data
Genotyping, quality control, and continental ancestry inference were previously described. 23,32 In brief, genotyping was performed using Illumina custom array and genome-wide imputation was conducted with the 1000 Genome Project reference panel. 32 Principal components (PCs) were estimated using PC-Relate 33 and continental-ancestry proportions were calculated using ADMIXTURE software. 23,34 "Genetic analysis groups" were constructed based on a combination of self-identified Hispanic/Latino background and genetic similarity, and are classified as Cuban, Dominican, and Puerto

Rican (Caribbean groups); and Mexican, Central American, and South
American (Mainland groups). 23 APOE genotyping was performed using commercial TaqMan assays previously described. 20 APOE variants were in Hardy-Weinberg equilibrium (P > 0.05).

Statistical analysis
We provided descriptive statistics to characterize the demographic and cognitive outcomes and APOE allele distributions in the full analytic dataset and by background groups. We tested the associations between APOE ε2 and ε4 alleles with cognitive outcomes in the same model (APOE ε3 used as the reference allele) using a complex survey design from the R "survey" package, 35  Further association analyses of APOE ε4 and ε2 alleles with cognitive outcomes were done separately for each of the six genetic analysis groups and the significance of the results was evaluated through 10,000 permutations for each group, to protect from potential high type 1 error due to the low proportion of APOE variants. Effect modification by genetic analysis groups was tested by including multiplicative interaction terms of these groups with the APOE alleles followed by a Cochran's Q heterogeneity test accounting for correlations between effect estimates. Continental-ancestry proportion interaction with APOE alleles effects was tested separately for each of the three ancestries (European, African, and Amerindian), by including the continentalancestry proportion variable together with a multiplicative interaction term of ancestry with the APOE allele (analytic dataset n = 3618). 23 Power calculation analysis for the two cognitive outcomes (cognitive decline and MCI) was calculated using population risk allele frequencies. 36 For APOE ε4 association with cognitive decline, the effect size was estimated based on the Cuban group and for APOE ε4 association with MCI, the effect size was estimated based on the Puerto Rican group, because these groups showed significant effects.
This calculation did not account for age distribution in the background groups. 2615 women), with a mean age of 62.1 years. Figure S1 in supporting information illustrates the overlap between the two dichotomous cognitive outcomes. Thirty percent of the population was classified with significant cognitive decline (n = 1250); of these, one third were also classified as MCI (n = 430). Twelve more participants have solely MCI.

RESULTS
The distribution of the APOE alleles in our SOL-INCA analytic sample by genetic background groups demonstrates the differential distribution, similar to the results published by González et al. 20 Overall, APOE ε3 is the most frequent allele in all Latino background groups, while ε4 and ε2 are relatively rare. Proportion ancestry interaction with APOE alleles' effects on cognitive outcomes is presented in Table 4. A significant interaction effect was found between Amerindian ancestry and APOE ε4 on cognitive decline (OR = 0.47, 95% CI [0.24;0.93], P-value = 0.04), such that protection from the risk of cognitive decline in APOE ε4 carriers was associated with higher Amerindian proportion ancestry (Figure 1).

DISCUSSION
We performed an association study of APOE alleles and cognitive outcomes in a large cohort of diverse middle-aged and older Latinos. Our main result showed an association between the APOE ε4 allele and the risk of significant cognitive decline. In the stratified analysis, this result remained significant only for Cubans. We also found an association between APOE ε2 and decreased risk of MCI only in Puerto Ricans, suggesting differential effects of the APOE alleles on cognitive function in the Latino background groups. We further discovered that an increased proportion of genetic Amerindian ancestry was associated with a protective effect from the risk of  of APOE alleles on significant cognitive decline is not feasible. 37 In the analysis stratified by "genetic analysis groups," we anticipated some relationship between APOE ε4 and significant cognitive decline specifically among Cubans because they have higher degrees of European ancestry. 23 Cubans in our analytic dataset are also older (a risk factor for cognitive decline) and more educated (a protective factor for cognitive decline) compared to the other Latino backgrounds; however, our models controlled for age and education. A previous study conducted in Cubans from Cuba also reported an association between APOE ε4 and incident of dementia with a stronger effect in middle-aged adults (<70 years) compared to older adults (>70 years). 38 The non-significant results for the APOE ε4 allele association with  (Table S1). Alternatively, it may result from limited statistical power eg, power = 0.26 for the South American group for the cognitive decline trait (Table S4), or it could present a true differential effect size of APOE ε4 on significant cognitive decline in the different Latino backgrounds.
The latter hypothesis is supported by the fact that the statistical power   [16][17][18] Unique to this study, continental-ancestry proportion, which captures genetic variation across the genomes, further revealed that an increased proportion of genetic Amerindian ancestry was associated with a protective effect from the risk of APOE ε4 on significant cognitive decline. This result is also compatible with the APOE ε4 significant risk effect we found in Latino Cubans, which were shown to have the lowest proportion of Amerindian ancestry among all Latino background groups (Table 1). By using all three ancestries of the admixed Latino population we could infer the protective effect of the Amerindian ancestry, rather than the European or African ancestries being harmful (if there were only two studied ancestries, we could not distinguish between the protective effect of one and the risk effect of the other).
This result is inconsistent with a recent report in Peruvians (78 AD cases and 128 controls) suggesting that Amerindian local ancestry in the APOE region is contributing to a strong risk for AD in APOE ε4 carriers. 39 Two other studies in Caribbean Hispanics suggest a protective effect of the African local ancestry in the APOE region from the risk of APO ε4 on AD. 40,41 Therefore, we also studied whether local ancestry at the APOE region modifies the effects of APOE on significant cognitive decline (results not shown). While the Amerindian local ancestry association was protective, it was not statistically significant (P-value >0.2). It is a topic of future research to perform a more comprehensive analysis of local ancestry at an expanded region around the APOE gene and potentially genome-wide, and search for specific genetic variants explaining the observed interaction of global Amerindian genetic ancestry with APOE ε4 in its effect on significant cognitive decline.
Our study also highlights a protective association between APOE ε2 and MCI solely in the Puerto Rican background group (Table 3), compatible with the known neuroprotective effect of APOE ε2. 8 The direction of this effect was similarly protective in several other Latino background groups; however, they were not significant, despite high power estimations especially for the Cuban and Mexican background groups (Table S4). APOE ε2 is relatively rare and its effect on cognitive function is less studied compared to APOE ε4, 42  Winner's curse bias might also explain our significant result in the stratified analyses. Overall, our differential association results in the Latino background groups may suggest a true differential genetic association between APOE alleles and cognitive outcomes related to admixed genomes. However, they could alternatively represent the lifestyle and environmental factors differing between the Latino ancestries, such as smoking, nutrition, alcohol consumption, physical activity, sleep phenotypes, air pollution, and metal exposure, that may also interact with APOE alleles and cognitive function risk. 45

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.