So far many independent studies have

So far, many independent studies have linked cholesterol metabolism–related genes with AD, either by showing associations with specific polymorphisms or with differential 87 7 australia levels (Beecham et al., 2014, Blain and Poirier, 2004, Lambert et al., 2009, Lambert et al., 2013, Leduc et al., 2015, Papassotiropoulos et al., 2005, Poirier et al., 1993, Sun et al., 2015, Wavrant-DeVrieze et al., 1997, Wollmer et al., 2007). Using a combinatory approach with genetic, transcriptomic, and LOAD biomarkers, association studies were undertaken to further elucidate how alterations in cholesterol metabolism–related genes can influence AD.
Methods
Results
Discussion Although altered cholesterol metabolism has been extensively implicated in the pathogenesis of AD through biological, epidemiological, and genetic studies, the molecular mechanisms linking cholesterol and LOAD pathology are still not well understood and contradictory results have been reported (Kivipelto and Solomon, 2006, Ricciarelli et al., 2012). As exemplified with BIN1 (Table 2), contradictory results can arise from the use of populations with different ethnic background (mixed population from USA vs. the homogeneous founder population of Quebec). Although complementary, the pathological biomarkers targeted by the polymorphic variants in Table 2 have different origin and represent slightly different processes (e.g., soluble Aβ42 measured in the CSF is normally inversely proportional to the number of amyloid plaques quantified in the neuropil, whereas soluble CSF pTau proportionally increases with tangle formation and reflects the NFT density measured by the pathologist). Yet, several cholesterol metabolism–related genes were found to contain genome-wide significant signals for AD risk in the largest GWAS published to date (Lambert et al., 2013). These include APOE, CLU (alias APOJ), and ABCA7 involved in extracellular transport and BIN, PICALM, and SORL1 regulating cholesterol-rich lipoprotein internalization. However, none except APOE were shown to be significantly associated with amyloid or tangle pathologies in AD (Farfel et al., 2016), raising the issue of the clinical utility of these so-called genetic risk variants. For example, CLU's rs11136000 (Table 3: IGAP) reached genome-wide significance in several AD association studies (Harold et al., 2009, Lambert et al., 2009) but failed to show any significant association with amyloid and tau pathologies in autopsied AD cases (Farfel et al., 2016), an observation that we replicated here with the ADNI and the QFP cohorts (Table 2). Yet, CLU gene expression was shown to be a marker for neurodegeneration (May et al., 1990, May, 1993), but its upregulation does not appear to be specific to AD. CLU is increased in dementia with Lewy bodies, Pick's disease, and Down's syndrome (Duguid et al., 1989, Sasaki et al., 2002, Stoltzner et al., 2000) and also in response to experimental hippocampal deafferentation in rodents (May et al., 1990). CLU's rs11136000 variant was associated with brain CLU expression levels in both the UKBEC (Table 2: microarray brain) and the QFP cohorts (CTL and LOAD brains, Fig. 3) but not with blood CLU expression levels (Table 2: microarray blood). Thus, rs11136000 is indeed a brain-related functional variant, but its association with AD is not specific or pathophysiology related. Interestingly, of all the cholesterol-related polymorphisms examined in Table 2, only SREBF2's rs2269657 showed significant dual associations with several LOAD pathological biomarkers and gene expression levels. In a previous GWAS of CSF Aβ, tau, and pTau in the ADNI cohort, rs2269657 nearly reached genome-wide significance with CSF Aβ42 concentrations (Supplementary Material from Kim et al. (2011)). Without focusing on cholesterol-related genes, SREBF2 was identified by the investigators as a potential risk candidate that required independent replication in a separate cohort.