In this study of genetic contributions to cognitive decline in the earliest stages of AD, we investigated whether a priori defined PRS/PGS for AD, intelligence and educational attainment were associated with cognitive decline in Aβ + CU and MCI, as well as in unselected CU and MCI individuals irrespective of Aβ status. Our main findings were that the PRS for AD was associated with rate of cognitive decline in early stage AD (participants with abnormal Aβ), while intelligence PGS was protective in both early stage AD and in CU and MCI participants irrespective of Aβ status. Individual SNPs did not reach genome-wide significance for effects on longitudinal cognition, but several suggestive associations were found. Taken together, these findings suggest that a priori defined genetic risk factors for AD may influence the disease by affecting the rate of cognitive decline for people with the early AD, while genetic factors linked to intelligence may affect the overall rate of decline in cognition. It is possible that treatments that mimic or interfere with specific biological pathways, identified through these genetic analyses, may have a potential to affect the rate of disease progression in early stage AD.
Notably, the effect of an AD risk score (PRS 8) was stronger than that for APOE ε4 burden in early stage AD, but APOE ε4 burden showed a nominal level of effect in the full cohort. This finding suggests that in early stages of AD the genetic control of the rate of cognitive decline is largely unrelated to APOE. This agrees with our recent finding that effects of Aβ-burden on cognition did not vary by APOE genotype in very early stages of AD [25]. The effect of the Alz-PRS on cognitive decline was specific to the Aβ + group, indicating a symbiotic effect between Aβ pathology and genetic factors related to AD. The Alz-PRS had no effect on cognitive decline in CU and MCI participants irrespective of Aβ status, supporting that the genetic variants that are included in this PRS only affected cognition in the context of Aβ pathology. Previous studies of genetic risk factors for cognitive decline in early stages of AD using the Alzheimer's Disease Neuroimaging Initiative (ADNI) subjects [26] and Australian Imaging, Biomarkers and Lifestyle (AIBL) [27] found significant effects of APOE ɛ4 and PGSs associated with AD [26, 27]. A study focused on preclinical AD subjects from AIBL study used episodic memory PRS and predicted rates of cognitive decline in domains typically affected in the preclinical stages of AD [28]. Compared to those studies our study had more detailed follow up data, and also included subjects with a wider age range compared to most previous studies. Unlike those studies [26, 27, 28] our findings suggest that a well-defined AD PRS can predict cognitive decline over and above APOE ɛ4 at an early stage of AD. This is consistent with one study that used polygenic hazard score (PHS) over and above APOE ɛ4 to predict longitudinal clinical decline in older individuals with moderate to high amyloid load [29]. Such PRSs may potentially be used (together with other modalities) to improve early prognostics of risk individuals. Looking for the PRS effect on the early disease process, another recent study showed the association of AD PRS with an increased probability of MCI compared to normal individuals at 50 years of age [30]. But perhaps more importantly, these genetic results may aid in understanding of the metabolic abnormalities of AD at an early stage, as discussed further below.
The intelligence PGSs (PGS 4 and PGS 5 for the Aβ + cohort and PGS mainly PGS 6 and PGS 10 for the full cohort) were protective in both early stage AD and in CU and MCI participants irrespective of Aβ status. Four intelligence PGS (PGS 1, PGS 2, PGS 3 and PGS 4) were only significantly associated with baseline cognition. The involvement of these PGSs in cognitive decline is uncertain. Previous studies have reported different results for associations between intelligence and AD. For example, one study found that intellectual enrichment was not a significant predictor of amyloid or AD-pattern neurodegeneration [31]. Two other studies found that intellectual enrichment may have marginal effects on AD biomarkers but a greater impact on delaying onset of cognitive impairment [32, 33]. Higher intelligence (and perhaps relevant genetic variants) was also associated with lower risk of AD in a twin study [34]. Taken together, these and our findings supports a model where genetic factors that contribute to higher intelligence are protective against general cognitive decline and may delay the onset of symptoms of AD (despite not affecting the underlying biological processes of the disease).
There was no effect of education PGS, arguing against associations between educational attainment and rate of cognitive decline. This is in line with another recent study using educational attainment PGS to predict rate of cognitive decline in non-demented individuals [35]. Another recent study agreeing with our findings showed that education and cognitive function in midlife did not affect long-term brain Aβ accumulation [36].
Our bioinformatics analysis based on the GO enrichment analysis showed that most of the genes of Alz PRS 8 were involved with regulation of Aβ in the brain. The lower enrichment p-value indicates that the proteins are at least partially biologically connected, and associated with rate of cognitive decline in harmony. The pathway analysis showed “Clathrin-mediated endocytosis” as the top pathway hit in which the major genes of the AD PRS are involved. Previous studies have shown that this pathway plays a central role in the production of Aβ in neurons [37, 38]. Our pathway analysis also showed Trges pathway as a second top hit. Dysfunction of Tregs has been reported to be associated with the neurodegenerative disease, such as AD [39].
For the Intelligence PGSs (Intel PGS 5 [Aβ + cohort] and Intel PGS 6 [Full cohort]) we did not find any pathway or GO enrichment term related to Aβ regulation. This further confirms the findings that the genes relevant for general intelligence are not associated with dysregulation of Aβ, but rather affects processes that modulate decline in cognition independent of AD [31, 32, 33].
Our exploratory GWAS analysis (Table 4) found 8 SNPs (5 in Aβ + cohort and 3 in full BioFINDER cohort) associated (p ≤ 10− 6) with rate of cognitive decline, although not reaching genome-wide significance. The minor allele of rs113625117, an intronic variant of WDR59, was strongly associated with slower cognitive decline. WDR59 had been linked before to behavioural disorders [40], but there are no previous reports of associations between this gene and cognitive decline. PSMD11 carrying the intronic variant rs9912262 was found to be strongly associated with faster cognitive decline. The enzyme encoded by this gene form a pivotal component for the ubiquitin-proteasome system (UPS) and cellular protein quality control [41]. Several clinical and experimental studies have shown that UPS deregulation may contribute to neurodegenerative disorders, including AD, which supports the relevance of this finding [42, 43, 44, 45, 46]. The minor allele of intronic variant of MACROD2 (rs370696) was found to be protective in both Aβ positive and full BioFINDER cohort. This gene has been reported to be strongly associated with temporal lobe volumes [47], supporting its relevance for cognitive decline, and also with autism spectrum disorder [48], supporting its relevance for brain diseases in general. Out of 8 SNPs, 2 SNPs (rs10119946 and rs9883584) were present near pseudogenes (RPS15AP27 and RNU6-815P), whereas rs4573079 was present near a long non-coding RNA (LOC101929705). All these 3 SNPs were associated with slower cognitive decline. Recent studies have shown that although pseudogenes are not transcribed themselves, they may contribute to regulation of gene expression [49], making it possible that the variants identified here modulate cognitive decline through regulation of other (unknown) genes.