To collect T2D-AD shared genes and variants from the literature, we developed a study protocol using the PRISMA statement (Page et al., 2021). The public database PubMed¹ was searched from August 2001 to the 4th of September 2022. The search terms were limited to “Type 2 diabetes” AND “Alzheimer’s disease” AND “gene” OR “biomarker” OR “Proteomic” OR “Methylation.” Initially, we established a systematic screening for all articles published during that period according to their title and abstract relevance. Articles on animal models, in vivo studies, and mitochondrial DNA were excluded. The final selection criteria were as follows: (1) relevant articles, (2) available in English, (3) studies conducted on human samples, and (4) genetic, transcriptomic, proteomic, and methylation studies.

Merged data were pruned based on the physical distances between adjacent markers and linkage disequilibrium (LD). High-density markers that did not provide additional information were excluded. Next, pruning data were used to create a multidimensional scaling plot (MDS) to study the landscape of the selected common T2D-AD gene regions. To this end, a symmetric matrix of identity-by-state (IBS) distances for all pairs of individuals was based on the proportion of shared common alleles. This analysis was performed using the Plink and R software (R: The R Project for Statistical Computing, n.d).⁴

After calculating the allele frequencies of the T2D-AD shared variants, the populations were clustered according to their geographic origins. Four groups were generated: North African (NAF), East Asian (EAS), American (AMR), and European (EUR). The Chi-square test was used to compare the risk allele frequencies of candidate variants between NAF populations and other populations. Bonferroni’s adjustment was applied to the level of significance set at a value of p threshold of 5% divided by the number of studied variants. All analyses were conducted using the R software.

We used a Bayesian clustering algorithm, STRUCTURE Ver. 2.3.4 software (Pritchard et al., 2000; FALUSH et al., 2007) to explore the variability of the common T2D and AD variants in terms of population structure. The algorithm assigns samples within a hypothetical K number of ancestries. We set a range of possible numbers of clusters ranging from K = 2 to K = 10, and four trials were run for each K. The Markov Chain Monte Carlo iteration for each structure analysis was run for 10,000 after an initial burn-in period of 10,000 steps. To assess the most likely number of clusters, we calculated Delta K, as proposed by Evanno et al. (2005). The similarity of the runs at each K level was evaluated using CLUMPP software as implemented online (Jakobsson and Rosenberg, 2007). Distruct software was used to visualize the best alignment of subpopulations, inferring population substructure and individual assignment across the best runs at each K level.

The PubMed search based on our defined search terms yielded 226 results. Studies with irrelevant results and those conducted using animal models (n = 163) were excluded. Additionally, four relevant studies were excluded due to access issues. The PRISMA flow diagram for the selected studies is represented in Figure 2. Finally, 59 publications were included in the present study (Supplementary file 1; Table 1).

The majority of the studies used data from American (Native American, Latino American, African American, and Mexican), Asian (Han Chinese, Japanese, and Korean), and European ancestry populations.

The literature search revealed 231 variants and 363 genes shared between T2D and AD identified by these studies. The 363 genes include those mapped to the 231 variants. A total of 46 genes were replicated in different studies (Supplementary file 1; Tables 2, 3).

Our literature search revealed 231 variants and 363 genes in common between T2D and AD. Annotation of the 231 shared SNPs showed that MADD and EFCAB5 harbored the highest number of these variants (nine SNPs). MADD gene, also known as IG20, plays a critical role in the development of glucose intolerance (Dupuis et al., 2010; Hu et al., 2010; Strawbridge et al., 2011; Wagner et al., 2011) and AD (Del Villar and Miller, 2004; Hassan et al., 2021). In accordance with our findings, a large-scale genome-wide cross-trait study identified MADD as the only gene significantly associated with AD and fasting glucose exclusively in pituitary tissue; it is also the only shared gene found in both cross-trait Meta-analysis and Transcriptomic-wide association studies. Thus, the pituitary gland may link T2D and AD by regulating glucose metabolism and neuronal viability through MADD (Zhu et al., 2019).

Regarding the EFCAB5 gene, it encodes the EF-hand calcium-binding domain 5. Our results are in accordance with those of Karki et al. (2020) highlighting the importance of this gene inT2D and AD development. Among the nine identified variants in this gene, two SNPs (rs9902453 and rs7221743) are associated with coffee consumption (Cheung et al., 2012; Cornelis et al., 2015). In this context functional studies showed the protective role of coffee consumption against AD (Kwok et al., 2016; Zhou et al., 2018) and T2D-associated memory impairment through adenosine A2 receptor (ADORA2A) blockage (Duarte et al., 2019). Thus, we suggest that variants in EFCAB5 could affect memory impairment in T2D subjects.

We found six SNPs located in NECTIN2, which encodes the nectin cell adhesion molecule 2 protein involved in T-cell signaling (Zhu et al., 2016). These variants have been previously reported to be shared between T2D and AD (Wang et al., 2017). Indeed, these variants are associated with lipid metabolite measurements, emphasizing their critical role in the development of AD in T2D patients (Xiao et al., 2022).

CST3 encodes cystatin C inhibitors of cysteine proteases (Maniwa et al., 2020). It is one of the genes harboring the highest number of common SNPs, between T2D and AD (Karki et al., 2020). A recent study showed that CST3 protein aggregation abolishes its function and slightly increases amyloid-beta 1–40 (Aβ1-40) fibril formation, enhancing neurodegeneration (Sheikh et al., 2021). Furthermore, the exogenous Cystatin C induces impairment of insulin signaling in hippocampal neurons, which could promote cognitive decline and AD development (Luo et al., 2018). In contrast, other studies have suggested that Cystatin C exerts neuroprotective effects by inhibiting cysteine proteases, rescuing neurodegeneration, inhibiting Aβ oligomerization and amyloid fibril formation, inducing autophagy, and neurogenesis (Mathews and Levy, 2016). This discrepancy may be explained by Cystatin C conformation or its levels in the brain. Indeed, the shared T2D-AD variant (rs1064039) was previously associated with reduced Cystatin C levels owing to impaired signal peptide cleavage (Benussi et al., 2003). We hypothesized that CST3 plays a crucial role in the development of T2D-induced AD pathology through the regulation of cerebral amyloid angiopathy and insulin signaling in a dose dependent manner. Further studies are needed to determine its exact function in T2D-inducing AD condition.

We found five shared SNPs between T2D and AD within the APOE gene. These findings are in line with the literature. APOE is a leading factor for AD development in T2D subjects (Zhen et al., 2018; Shinohara et al., 2020). Three variants among the five were associated with body mass index (Yengo et al., 2018). Our findings support the synergic effect of obesity and APOE genotype on the development of T2D and AD (Jones and Rebeck, 2018).

AGPAT1 encodes for 1-acylglycerol-3-phosphate O-acyltransferase 1. It harbors five SNPs among the 231 T2D-AD shared variants identified in this study. Deletion of this gene induces low glucose and lipid plasma levels, as well as neurological disturbances (Agarwal et al., 2017). These findings emphasize the importance of AGPAT1 in the regulation of glucose homeostasis and neuron viability. Further studies are needed to investigate its role in T2D and AD.

The functional annotation of the coding variants revealed six probably damaging SNPs located in APOE, AGER, AGT, TP53, ICAM1, and EPHX2 gene. All the variants have been identified by the Karki et al. (2020), study. Interestingly the two variants rs7412 and rs2070600 have been reported by other studies (Wang et al., 2016; Kim et al., 2022). Minor allele (T) carriers of the variant rs7412 are classified as APOEε2 carriers. It has been largely proven that APOEε2 has a protective effect against AD (Shinohara et al., 2016). Controversially, Shinohara et al., (2020) showed that APOEε2 accelerates cognitive decline in diabetic patients by 4 years. This could be explained by the synergic effect between diabetes (Hardigan et al., 2016; Peng et al., 2021) and the APOEε2 genotype in enhancing neurovascular impairment and tauopathies (Kim et al., 2022).

The second SNP (rs2070600) located in AGER (advanced glycosylation end-product specific receptor), causes a conversion at position 82 from glycine to serine (G82S) responsible for a decrease of AGER proteolyze and increase of AGEs plasmatic levels (Serveaux-Dancer et al., 2019). Several studies shed the light on the role of AGEs in T2D and AD through oxidative stress and amyloid regulation mechanisms (Michailidis et al., 2022). As a result, we hypothesize that T2D patient carriers of these risk allele variants have a greater risk to develop AD.Likely, the functional annotation of non-coding variants revealed three top-ranked SNPs rs1544210, rs12679834, and rs515071 located, respectively, in HHEX, LPL, and ANK1 gene. The minor allele A of the variant rs1544210 is linked to an under-expression of EIF2S2P3 pseudogene in the substantia nigra and hypothalamus regions. EIF2SS2P3 is a pseudogene located in chromosome 10. Transcriptomic-wide association studies revealed a significant enrichment of EIF2S2P3 for depressive symptoms, T2D (Génin, 2020) and T2D patients skipping breakfast (Chen et al., 2022), but its function remains unclear. Available evidence showed an increased risk for dementia development in T2D patients with depression (Katon et al., 2012). These findings highlights the role of EIF2S2P3 pseudogene in the development of T2D-induced dementia through enhancing depression. Further studies should be conducted to assess the role of EIF2S2P3 in T2D and AD development.

The same allele was associated with TFB motif “USF1,” that regulates APOE gene a major linking factor between T2D and AD (Isotalo et al., 2012).

The second top-ranked SNP is rs12679834 located in the LPL gene. We found a TBF motif (NR2C2) with a high score associated with the G allele of this variant. NR2C2, also known as TR4, is an orphan nuclear receptor targeting many genes involved in metabolism including APOE (Fogarty et al., 2013). This variant was also marked by H3K4me1 and H3K27ac histone modification in three brain regions. Furthermore, an increased expression of LPL in microglia appears to have a protective effect against AD (Keren-Shaul et al., 2017) and obesity (Gao et al., 2017). We suggest that rs12679834 possesses a protective effect against T2D and AD development via increasing lipid and lipoprotein uptake in the Central nervous system (CNS).

The third SNP was rs515071 located in the ANK1 gene. The minor allele G of this variant was associated with decreased expression of ANK1 gene in the cerebellum and cerebellar hemisphere regions. In agreement with our results, the GG genotype of the rs515071 variant is associated with a greater risk for T2D (Sun et al., 2017, p. 1) and AD (Chi et al., 2016). A reduced expression of ANK1 gene could enhance T2D and AD development by affecting mediated metabolism, signal transduction (Sun et al., 2017), and inflammatory process (Morris et al., 2019).

Six SNPs reported in our study were found to affect miRNA molecules implicated in LOAD, insulin signaling, and T2D pathways. miR-1965-5p was enriched in the AD, T2D, and insulin signaling pathways. It has a positive effect on insulin biogenesis by enhancing insulin activity (Panda et al., 2014). miR-196b-5p down regulation has been implicated in innate immune response, apoptosis, and depression (Zhang et al., 2018). Inconsistent results have been found to be associated with the regulation trend in patients with AD (Pichler et al., 2017). We suggest that miR-196b-5p may play a protective role against T2D and AD development through insulin, immune response, and apoptosis regulation. However, further studies are required to elucidate their role in T2D and AD. Controversially, miR-378, previously associated with LOAD (Lusardi et al., 2017), was also enriched in the insulin-signaling pathway. Interestingly, miR-378 induces insulin resistance by targeting P110a and SIRT7 (Deng and Guo, 2019). Furthermore, genetic depletion of miR-378a-3p ameliorates inflammatory stress and insulin resistance via protein kinase R inhibition (Wang et al., 2021). Likewise, upregulation of this miRNA has also been found in patients with AD (Dong et al., 2021; Li and Cai, 2021). This evidence emphasizes its role in T2D-induced AD through Central insulin signaling impairment (Gabbouj et al., 2019).

A previous study showed that miR-125a-5p and miR-125b-5p shared between T2D and the insulin pathway could ameliorate gluconeogenesis, glycogen synthesis (Xu et al., 2018), and insulin sensitivity (Yu et al., 2019). Interestingly, the same miRNA was found to be downregulated in the gray matter of patients with AD (Wang et al., 2011). We found that miR-98-5p was enriched in the insulin-signaling pathway. Decreased expression of this miRNA has been observed in T2D patients (Kokkinopoulou et al., 2019). The same study also reported a negative correlation between miR-98-5p and insulin levels in patients. Interestingly, Chen et al. (2019) found that in AD mice, miR-98 binds to HEY2 inducing a decrease of Aβ production, improve oxidative stress, and mitochondrial dysfunction through activating the Notch signaling pathway. We hypothesized that low levels of miR-98-5p could serve as a biomarker for insulin resistance and Aβ aggregation. A recent study found a significant downregulation of miR-214-3p blood levels in T2D patients (Avgeris et al., 2020). Similarly, miR-124-3p is downregulated in patients with AD and animal models (Kou et al., 2020), suggesting its potential role as a biomarker and therapeutic target for insulin resistance (Cheng et al., 2020) and cognitive defects (Zhang et al., 2018).

For a more in-depth understanding of the shared genes, we conducted a pathway enrichment analysis. The results revealed that the replicated genes were mainly enriched in lipid subunit organization, positive regulation of protein binding, positive regulation of amyloid fibril formation, microglial cell activation, cholesterol metabolism, plasma lipoprotein assembly, and Statin inhibition of cholesterol production pathways. Our findings are supported by those of previous studies. Plasma protein binding (PPB) has been implicated in several mechanisms, particularly drug binding and pharmacokinetics (Smith et al., 2010). A recent study identified an enrichment of positive regulation of PPB in 3 × Tg-AD mice fed rosmarinic acid, a preventive molecule against AD (Yamamoto et al., 2021). Rosmarinic acid is a potent suppressor of Aβ and an inhibitor of phosphorylated tau accumulation (Yamamoto et al., 2021). Interestingly, Rosmarinic acid possesses a therapeutic effect against T2D through the remodeling of amyloid aggregates (Wu et al., 2021). We identified, for the first time, the implication of positive regulation of PPB in T2D and AD. We hypothesized that the regulation of the PPB pathway could serve as a potential therapeutic target for these diseases.

The accumulation of amyloid fibrils is a hallmark of several degenerative diseases including T2D and AD. Insulin resistance promotes the oxidative stress generation and proinflammatory cytokines secretion in beta-cells inducing mitochondrial dysfunction and accumulation of protein aggregates, including human islet amyloid polypeptide (hIAPP) (Rocha et al., 2020). The latter can across the blood–brain barrier (BBB) inducing AD pathology (Lupaescu et al., 2022; Marrano et al., 2023).

Furthermore, amyloid deposition causes microglial and astrocyte activation leading to cytotoxic molecules release (Lupaescu et al., 2022). Recent study has demonstrated that hyperinsulinemia impaired GLUT4 translocation inducing mitochondrial fission, microglial M1 polarization, and neuroinflammation (Yang et al., 2022). Moreover, long-term high fat diet induces microglial M1 polarization which explains obesity/diabetes-associated cognitive impairment (Wu et al., 2020).

Cholesterol metabolism involves energy metabolism, cell membrane composition, and myelination. Dysregulation of these biological processes induces several pathologies, mainly T2D and AD. Reports suggested that a long-term high-fat diet could induce AD by enhancing Aβ and phosphorylated tau accumulation (Czuba et al., 2017). Downregulation of cholesterol biogenesis has been observed in diabetic (Suzuki et al., 2010) and AD (Varma et al., 2021) brains. Cholesterol is biosynthesized in astrocytes via the Bloch pathway and is transported to neurons by APOE via the ABC transporter. For utilization by neurons, APOE-containing cholesterol should be absorbed by LRP1/LDLR (Czuba et al., 2017). A previous study successfully demonstrated that insulin resistance suppresses LRP1 expression, which may further compromise insulin signaling and cholesterol metabolism in neurons (Liu et al., 2015). Thus, our pathway enrichment analysis confirmed previous findings supporting the role of cholesterol metabolism and lipoprotein processes as linking factors between T2D and AD.

It is likely that genetic background plays an important role in the development of preventive strategies targeting modifier risk factors, such as T2D. Despite the high prevalence of T2D (Magliano et al., 2021) and AD (Nichols et al., 2022) in North African populations, we noticed an under or non-representation of these groups in the consortiums investigating these diseases (Martin et al., 2019). Taking all these evidences into consideration, it is important to dissect the genetic landscape of T2D-AD shared genes in North Africa in comparison with other well-studied populations. We determined the genetic landscape of T2D-AD shared genes in 829 individuals from 16 different populations whose genotyping data are publicly available [African ancestry in the South Western USA(ASW), Mexican ancestry living in Los Angeles, California, USA (MEX), Western European ancestry populations of Utah from the CEPH collection (CEU), Toscani people of Italy (TSI),South Spain (Spain_S), North Spain (Spain_N), North West of Spain (Spain_NW), Spain Basic populations (Spain_BASC), Han Chinese in Beijing, China (CHB),Chinese population of metropolitan Denver, Colorado, USA (CHD), Japanese in Tokyo, Japan; (JPT), Individuals from North Africa: Algeria (Algeria), Egypt (Egypt), Libya (Libya), Tunisia Douiret (TN_Ber), South Morocco (Morocco_S), and North Morocco (Morocco_N),],. MDS analysis showed genetic similarity among North African populations (Algeria, Egypt, TN_Ber, Morocco_N, Morocco_Sand Libya), reflected by a consistent cluster. A better individualization of the North African populations was identified when the MDS analysis was conducted at the population group level. A slight similarity between North African and Southwestern European populations (CEU, Spain-S, Spain_NW, and TSI) was detected. However, a great divergence between North African and East Asian populations was observed in the two MDS representations. These results were further confirmed by STRUCTURE analysis conducted on four clusters of populations: North African, European, East Asian, and American. STRUCTURE representation shows a high admixture of the genetic structure of North African populations consisting mainly of European and African components, with minimum penetrance of East Asian components. Our findings are consistent with those of previous studies. Indeed, several genes/polymorphisms in T2D (Chande et al., 2020) and AD (Rubin et al., 2021) are highly variable among ethnic groups. Similar genetic positioning was observed among North African, European, and Asian populations regarding Metabolic Syndrome (MetS) pharmacogenes (Jmel et al., 2018). It is important to note that genes explored by Jmel et al. (2018) were also investigated in our study because MetS share several mechanisms with T2D and AD (Hayden, 2019). The genetic positioning of the North African cluster could be explained by the high ethnic heterogeneity of these populations. North Africans are multi-ethnic populations with several ancestral components: Middle Eastern, Sub-Saharan African, European, and autochthonous (Arauna et al., 2019). The high heterogeneity of the T2D-AD genetic background in North African populations reflects previous historical events such as invasion and migration (Botigué et al., 2013; Arauna et al., 2017; Fregel et al., 2018). Our results also support the conserved and ancient divergence between the North African and East Asian populations going back 550 centuries ago (Tateno et al., 2014). Thus, the non-replication of some genetic biomarkers of T2D (Baroudi et al., 2009; BaroudiOuederni et al., 2009; Ezzidi et al., 2009; Turki et al., 2012) and AD (Smach et al., 2011, p. 1; Rassas et al., 2013; Landoulsi et al., 2018) in the North African group could be due to its high heterogeneity and unicity. Indeed, among 231 risk variants studied, only 49 SNPs were present in the North African group. This could be the result of ethnic selection because some AD variants are also ethnicity-specific biomarkers (Huang et al., 2017) or due to the limited size of the North African populations investigated.

Furthermore, the inter-ethnic risk allele frequency comparison of the 49 variants between North African populations and other population groups revealed significant differences in allele frequency of 11 SNPs between North African, European, and East Asian populations located in DNM3, CFH, PPARG, ROHA, AGER, CLU, BDNF1, CST9, and PLCG1 genes.

MAF of two variants, rs4504922 and rs7539972, located in the dynamin 3 (DNM3) gene, was significantly different between North African and European populations and between North African and East Asian populations. DNM3 is enriched in the Fc gamma R-mediated phagocytosis pathway associated with AD and T2D (Hao et al., 2015; Caputo et al., 2020). These variants were previously identified as SNPs shared between T2D and AD (Caputo et al., 2020). We suggest that North African carriers of these risk allele variants may be at increased risk of T2D and AD.

The rs800292 G > A SNP, located in the complement factor H (CFH) gene, has been previously reported to be associated with a higher risk of age-related macular degeneration (Guindo-Martínez et al., 2021) and diabetic retinopathy (Wang et al., 2013).

Furthermore, the inter-ethnic comparison of risk alleles revealed a significant difference in the MAF of three variants (rs6809832, rs6997, and rs11715915) located in peroxisome proliferator activated receptor gamma (PPARG) and Ras homolog family member A (RHOA) genes between North African and East Asian populations. These variants have been associated with increased BMI and HbA1c levels (Merino et al., 2017; Yengo et al., 2018; Barton et al., 2021; Jurgens et al., 2023). We hypothesized that North African carriers of risk alleles of these variants may have an increased risk of developing obesity and insulin resistance pathologies, such as T2D and AD (Magliano et al., 2021; Nichols et al., 2022).

Moreover, we identified significant differences in the MAF of three SNPs (rs2070600, rs11136000, and rs6265) located in AGER, clusterin (CLU), and brain derived neurotrophic factor (BDNF1) between North African and East Asian populations. These variants have been previously associated with MCI/AD development in T2D subjects (Cai et al., 2016; Wang et al., 2016; Daily and Park, 2017; Stepler and Robinson, 2019; Bradley, 2020). The high MAF of these variants in North Africans could partly explain their greater risk of developing T2D-AD pathology, (Magliano et al., 2021; Alzheimer’s Disease Facts and Figures, n.d.).

The variant rs3004145 C > G, located downstream of the Cystatin C9 (CST9) gene, presents a significant MAF difference between North African and East Asian populations. This variant has previously been associated with elevated cystatin C levels in the European population (Jurgens et al., 2023). High serum cystatin C levels have been previously associated with an increased risk of T2D (Yuan et al., 2022), T2D-related neuropathy (Hu et al., 2014), and AD (Straface et al., 2005). Thus, we suggest that North African carriers of the rs3004145 G allele may have an increased risk of T2D-induced AD.

Finally, our statistical analysis revealed a significant MAF difference in the exonic variant rs753381 T > C between North African and European populations. This variant is located in the phospholipase C gamma 1 (PLCG1) gene, recently identified as a potential therapeutic target for T2D (Ganekal et al., 2023). It has previously been associated with metabolic syndrome (Brown and Walker, 2016) and elevated serum cholesterol, LDL, and ApoB levels in individuals of African, East Asian, European, Hispanic, and South Asian ancestry (Graham et al., 2021). We suggest that the differences in rs753381 C allele frequency in North Africa explain the low plasma levels of TC, LDL-C, and ApoB compared to European populations (Najah et al., 2013). Further studies are required to elucidate the relationship between this variant and T2D-induced AD development in North Africa.