Transcriptional risk score for chronic kidney disease: relation to Alzheimer's disease Authors: Jung‐Min Pyun, Young Ho Park, Jiyun Hwang, SangYun Kim, Yongjin Yi, Kwangsik Nho, for the Alzheimer's Disease Neuroimaging Initiative Published: 2026 ABSTRACT AbstractINTRODUCTIONAlthough Alzheimer's disease (AD) is prevalent in individuals with chronic kidney disease (CKD), the mechanistic link remains unclear. We assessed the impact of CKD transcriptional risk on AD endophenotypes.METHODSWe calculated the CKD transcriptional risk score (CKD‐TRS) using blood transcriptomes from the Alzheimer's Disease Neuroimaging Initiative (n = 661) and AddNeuroMerge (n = 665) and brain transcriptomes from the Religious Orders Study and Rush Memory and Aging Project (n = 622) and analyzed its associations with longitudinal cognition, brain volume, amyloidopathy, tauopathy, and mediation effects.RESULTSHigher blood‐based CKD‐TRS was associated with accelerated cognitive decline, rapid temporal cortical atrophy, and faster amyloid accumulation. Brain‐based CKD‐TRS was similarly linked to faster cognitive decline and greater amyloid burden, with amyloidopathy partially mediating the CKD‐TRS–cognition association.DISCUSSIONOur findings suggest that CKD‐TRS is associated with progressive neurodegeneration in AD, mediated by amyloidopathy, highlighting a shared molecular pathway. BACKGROUND Cognitive impairment in chronic kidney disease (CKD) is common, with studies reporting a high prevalence of mild cognitive impairment (MCI) from 27% to 62% and dementia from 8% to 37%. 1 In particular, decreased renal function has been associated with increased risk of Alzheimer's disease (AD), 2 , 3 the most frequent neurodegenerative dementia characterized by amyloid beta (Aβ) and tau accumulation leading to neuronal injury and cognitive decline. Although higher risk of AD in patients with CKD has been observed, the underlying mechanism remains largely unknown. 4 The kidney and brain share physiological features (kidney–brain axis), including high perfusion, low vascular resistance, and autoregulation, which render both organs vulnerable to vascular risk factors such as hypertension and diabetes. 5 Moreover, uremic toxins and systemic inflammation in CKD have been proposed as contributing factors to cognitive impairment. 5 However, AD key pathology such as Aβ and tau have not been investigated as potential links between CKD and dementia. From a genetic perspective, genes such as KLOTHO and KIBRA, expressed in both kidney and brain, have been linked to cognitive impairment. 1 , 6 , 7 However, the contribution of CKD‐related genes to cognitive dysfunction and their underlying pathways remains unclear. The transcriptional risk score (TRS) is an individual risk score comprising a summation of polarized transcriptome profiles of disease‐relevant genes. 8 In AD studies, AD‐TRS has shown associations with AD diagnosis, AD pathologies, and cognitive function, suggesting TRS as a potential biomarker. 9 , 10 TRS could reveal relationships between the transcriptional patterns of disease‐associated risk genes and corresponding phenotypes and pathology. In this study, we developed transcriptional risk scores for CKD (CKD‐TRS) using blood (blood‐based) and brain (brain‐based) transcriptomic data to examine their associations with AD pathological changes – amyloidopathy (A), tauopathy (T), and neurodegeneration (N) – and cognitive decline. We assessed blood‐based CKD‐TRS in two independent cohorts (the Alzheimer's Disease Neuroimaging Initiative [ADNI] and the AddNeuroMerge [ANM]) for longitudinal changes in amyloid, tau, cortical thickness, and cognition and evaluated brain‐based CKD‐TRS in the Religious Orders Study and Rush Memory and Aging Project (ROSMAP) for associations with neuropathologically confirmed amyloidopathy, tauopathy, and cognitive decline (Figure 1). Study overview. We selected target genes for calculating CKD‐TRS using summary‐based Mendelian randomization with GWAS summary statistics for renal function and eQTL summary data. Using the target genes, we generated blood‐based and brain‐based CKD‐TRS. Then we performed association analyses of CKD‐TRS with biomarkers for amyloid (A), tau (T), neurodegeneration (N), and cognitive function. Additionally, mediation analysis and pathway enrichment analysis of target genes were performed. AD, Alzheimer's disease; ADNI, Alzheimer's Disease Neuroimaging Initiative; ANM, AddNeuroMerge; CERAD, Consortium to Establish a Registry for Alzheimer's Disease; CKD, chronic kidney disease; eQTL, expression quantitative trait loci; GWAS, genome‐wide association study; PET, positron emission tomography; ROSMAP, Religious Orders Study and Rush Memory and Aging Project; SNP, single nucleotide polymorphism. METHODS Participants For the discovery cohort of blood‐based CKD‐TRS, we utilized data from the ADNI database (http://adni.loni.usc.edu/), which was launched in 2003 as a collaboration between public and private sectors. The main objective of ADNI is to assess whether a combination of serial magnetic resonance imaging (MRI), positron emission tomography (PET) scans, various biological markers, and clinical neuropsychological assessments can effectively track the progression of MCI and early‐stage AD. For our study, we included 661 participants from the ADNI dataset who had available blood microarray data. The cohort comprised individuals categorized as cognitively unimpaired (CU) (n = 213), with MCI (n = 345), and those diagnosed with dementia due to AD (n = 103), as defined by the ADNI criteria (http://adni.loni.usc.edu/study‐design/). Specifically, the CU group included participants who had no memory complaints, a Mini‐Mental State Examination (MMSE) score between 24 and 30, a Clinical Dementia Rating (CDR) score of 0, and normal memory function based on the Logical Memory II subscale from the Wechsler Memory Scale‐Revised, with scores above education‐adjusted cutoffs (≥9 for 16+ years of education, ≥5 for 8 to 15 years, and ≥3 for 0 to 7 years of education). The MCI group consisted of participants who reported subjective memory concerns, had a MMSE score between 24 and 30, a CDR score of 0.5, and abnormal memory function, as indicated by scores below education‐adjusted cutoffs on the Logical Memory II subscale (≤11 for 16+ years of education, ≤9 for 8 to 15 years, and ≤6 for 0 to 7 years of education). The AD group was composed of individuals with subjective memory concerns, MMSE scores between 20 and 26, CDR scores of 0.5 or 1, and abnormal memory performance as assessed by the Logical Memory II subscale, along with meeting the National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer's Disease and Related Disorders Association (NINCDS/ADRDA) criteria for probable AD. We incorporated demographic data, including age at the time of RNA data collection, sex, education level, cognitive assessment results, and brain MRI scans data. These data were accessed from the ADNI website (https://ida.loni.usc.edu/login.jsp?project = ADNI). Informed consent was obtained from all participants, and ethical approval was granted by the investigators at each participating site. RESEARCH IN CONTEXT Systematic review: The authors reviewed the literature using PubMed and Google Scholar to identify studies on CKD‐TRS. However, the association between CKD‐TRS and AD endophenotypes has not been studied. Interpretation: We report detailed associations of CKD‐TRS with longitudinal cognitive decline, brain atrophy, and AD pathology. CKD‐TRS is associated with progressive neurodegeneration in AD, mediated by amyloidopathy, highlighting a shared molecular pathway. Future directions: Future studies should validate these findings in larger cohorts and explore whether targeting CKD‐related pathways can modify AD progression. Systematic review: The authors reviewed the literature using PubMed and Google Scholar to identify studies on CKD‐TRS. However, the association between CKD‐TRS and AD endophenotypes has not been studied. Systematic review: The authors reviewed the literature using PubMed and Google Scholar to identify studies on CKD‐TRS. However, the association between CKD‐TRS and AD endophenotypes has not been studied. Interpretation: We report detailed associations of CKD‐TRS with longitudinal cognitive decline, brain atrophy, and AD pathology. CKD‐TRS is associated with progressive neurodegeneration in AD, mediated by amyloidopathy, highlighting a shared molecular pathway. Interpretation: We report detailed associations of CKD‐TRS with longitudinal cognitive decline, brain atrophy, and AD pathology. CKD‐TRS is associated with progressive neurodegeneration in AD, mediated by amyloidopathy, highlighting a shared molecular pathway. Future directions: Future studies should validate these findings in larger cohorts and explore whether targeting CKD‐related pathways can modify AD progression. Future directions: Future studies should validate these findings in larger cohorts and explore whether targeting CKD‐related pathways can modify AD progression. For the replication cohort of blood‐based CKD‐TRS, we used data from the ANM study, a cross‐European consortium formed for the discovery of AD biomarkers. For this study, we included 665 participants from ANM who had available blood microarray data. The cohort included individuals categorized as CU (n = 238), with MCI (n = 207), and those diagnosed with dementia due to AD (n = 220). CU participants were defined as those with a MMSE score between 24 and 30 and a CDR score of 0. MCI participants were defined as those with a MMSE score between 24 and 30 and a CDR score of 0.5. The AD group included participants with a MMSE score ranging from 12 to 28, consistent with NINCDS‐ADRDA criteria for probable AD. We utilized demographic data, including age at the time of microarray data collection, sex, educational background, cognitive assessments, and brain MRI data. These data were accessed from https://www.synapse.org/#!Synapse:syn22156880. Written informed consent was obtained from all participants at the time of enrollment, which included permission for analysis and data sharing. The study protocol and consent forms were approved by the Institutional Review Board at each participating site. For brain‐based CKD‐TRS, ROSMAP data were used. Clinical data were accessed from https://www.synapse.org/Synapse:syn3191087. For this study, we categorized diagnosis using the variable, clinical consensus diagnosis of cognitive status at time of death in the ROSMAP dataset. Participants with no cognitive impairment (coded value in ROSMAP dataset 1) as CU (n = 301), those with MCI and no other cause of cognitive impairment (coded value 2), and those with MCI and another cause of cognitive impairment (coded value 3) were defined as MCI (n = 168). Those with AD dementia and no other cause of cognitive impairment (coded value 4) and AD dementia and another cause of cognitive impairment (coded value 5) were defined as AD (n = 253). Blood‐based RNA expression microarray profiling In ADNI, peripheral blood samples were collected using PAXgene Blood RNA Tubes, and total RNA was purified using the PAXgene Blood RNA Kit (Qiagen Inc., Valencia, CA, USA). The Affymetrix Human Genome U219 Array (Affymetrix, Santa Clara, CA, USA) was used for expression profiling in the ADNI. Raw expression values were preprocessed using the robust multichip average normalization method and the robust spline normalization method in the ADNI. The details are described in https://ida.loni.usc.edu/pages/access/geneticData.jsp#206. In ANM, blood samples for the collection of gene expression data were taken at baseline. RNA samples from ANM were processed on Illumina Human HT‐12 v3 Expression BeadChips (Illumina Inc., San Diego, CA, USA). 11 Preprocessed raw data files, as well as following quality control, batch‐corrected expression values, are distributed via Synapse. We used the file from the Accelerating Medicines Partnership for Alzheimer's Disease (AMP‐AD) Knowledge Portal on Synapse (syn22286804 from https://www.synapse.org/#!Synapse:syn22286804). Brain RNA sequencing Details of RNA sequencing (RNA‐seq) and quality control procedures were provided in previously published work. 12 , 13 Normalized data with log counts/million (log CPM) of dorsolateral prefrontal cortex (DLPFC) from ROSMAP was downloaded from the AMP‐AD Knowledge Portal on Synapse (syn8456638) (https://www.synapse.org/#!Synapse:syn8456638). TRS calculation Target gene selection We integrated recent large‐scale genome‐wide association study (GWAS) summary statistics with publicly available blood expression quantitative trait loci (eQTL) data to identify functionally relevant genes whose expression levels are associated with the disease, using the summary data‐based Mendelian randomization analysis (SMR) analysis. 14 We used GWAS summary statistics of Stanzick et al. for CKD‐TRS. 15 Publicly available blood eQTL summary data were used for blood‐based TRS (https://eqtlgen.org/phase1.html), 16 and publicly available brain eQTL summary data were used for brain‐based TRS (data in the frontal cortex [Brodmann area 9] V8 release of the GTEx [https://gtexportal.org]). SMR analysis was used to prioritize target genes. False discovery rate (FDR)‐corrected p value < 0.05 was considered significant. Then, heterogeneity in dependent instruments (HEIDI) testing provided by SMR was applied to distinguish pleiotropy from linkage. Genes that passed a HEIDI test (p HEIDI ≥ 0.05) were used as target genes. TRS calculation TRS was calculated using the identified target genes as follows. 8 First, the expression levels of each gene were normalized to a standard normal distribution with a mean of0 and variance of 1. Second, we used the eQTL activity of disease‐associated genetic variants to determine the direction of risk for each gene's expression. Genes associated with increased expression levels due to disease risk alleles were categorized as high expression, while those associated with decreased expression levels were categorized as low expression. In rare cases, genes could be classified as having both high and low expression due to association with different single nucleotide polymorphisms (SNPs), leading to opposing eQTL effects. In such instances, the genes were excluded from the TRS calculation. Subsequently, we polarized gene expression values so that elevated risk could be reflected in the TRS. This was accomplished by reversing the sign of z‐scores for genes classified as low expression. Finally, we obtained the TRS for each individual by summing the polarized z‐scores over all target genes. Kidney function In ADNI, estimated glomerular filtration rate (eGFR) was calculated from serum creatinine values available in the ADNI dataset using the 2021 Chronic Kidney Disease Epidemiology Collaboration creatinine equation. 17 Among the 661 individuals whose transcriptomic data were available in ADNI, 367 had matched time points for transcriptomic profiling and eGFR measurement and were included in the analysis. Biomarker data‐related kidney function was not available in the ANM and ROSMAP cohorts. Amyloid PET In ADNI, amyloid PET scans were performed using [18F] florbetaben (FBB) or [18F] florbetapir (FBP). Amyloid burden was assessed based on Centiloid (CL) value. Additional information on amyloid PET processing can be found online (https://adni.loni.usc.edu/methods/documents/). Longitudinal amyloid PET imaging data were included, which were obtained after mRNA acquisition. The follow‐up period was up to 48 months. Cerebrospinal fluid In ADNI, the concentrations of cerebrospinal fluid (CSF) phosphorylated tau at threonine 181 (p‐tau181) were measured using the micro‐bead‐based multiplex immunoassay, the INNO‐BIA AlzBio3 Research Use Only test (Fujirebio, Ghent, Belgium) on the Luminex platform. 18 Detailed information on data processing can be found online (https://adni.loni.usc.edu/methods/documents/). Longitudinal CSF data were included, which were obtained after mRNA acquisition. The follow‐up period was up to 48 months. Neuropathology data In the ROSMAP cohort, brain amyloidopathy was assessed using the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) score (1 = frequent, 2 = moderate, 3 = sparse, 4 = none). 19 Tauopathy was evaluated by Braak stage: 0 = none; I–II = entorhinal involvement; III–IV = limbic involvement; V–VI = moderate to severe neocortical involvement. 20 Cognitive assessment In ADNI and ANM, longitudinal cognitive performances were assessed by the 13‐item Alzheimer's Disease Assessment Scale–Cognitive Subscale (ADAS‐Cog 13). 21 ADAS‐Cog 13 assesses multiple cognitive domains including memory, language, praxis, and orientation, and higher scores indicate greater severity. In the ADNI and ANM cohorts, baseline for the longitudinal analysis was defined as the time point of transcriptomic data collection. In the ADNI cohort, the follow‐up period extended up to 48 months. In the ANM cohort, cognitive assessments were included with follow‐up periods of up to 12 months, due to a substantial decrease in the frequency of cognitive measurements. In the ROSMAP cohort, longitudinal cognitive changes were evaluated using global cognitive function scores averaging 19 tests with five domains including episodic memory, semantic memory, working memory, visuospatial ability/perceptual orientation, and perceptual speed. 12 In the ROSMAP cohort, baseline was defined as the time of death, when post mortem brain transcriptomic data were obtained. The follow‐up period extended up to 9 years prior to death. In the ADNI, ANM, and ROSMAP cohorts, all longitudinal cognitive assessment data were collected after mRNA acquisition. Brain MRI In ADNI, structural brain volume was estimated from T1‐weighted brain MRI scans using FreeSurfer (surfer.nmr.mgh.harvard.edu). 22 More details regarding neuroimaging processing are available online (http://adni.info.org). In ANM, MRI data acquisition was designed to be compatible with the ADNI. 23 The imaging protocol for both studies included a high‐resolution sagittal 3D T1‐weighted Magnetization‐Prepared Rapid Gradient Echo (MPRAGE) volume (voxel size 1.1 × 1.1× 1.2 mm3) and axial proton density/T2‐weighted fast spin echo images. The MPRAGE volume was acquired using a custom pulse sequence specifically designed for the ADNI study to ensure compatibility across scanners. 23 Full brain and skull coverage was required for both of the latter datasets and detailed quality control carried out on all MR images from both studies according to the ANM quality control procedure. 24 , 25 Magnetic field strengths were 1.5 Tesla or 3 Tesla in ADNI and 1.5 Tesla in ANM. We examined temporal cortical thicknesses as the region of interest, which is known as the signature area of AD. 26 The cortical thickness for the temporal lobe was calculated as the average value of right and left temporal pole, fusiform, superior temporal, inferior temporal, and middle temporal cortical thicknesses. 27 Longitudinal MRI data up to 3 years in ADNI and 2 years in ANM were used. Longitudinal MRI imaging data were included, which were obtained after mRNA acquisition. Statistical analysis Demographic characteristics were compared by diagnosis using t‐tests or chi‐squared (χ2) tests. Associations between blood‐based CKD‐TRS and kidney function was assessed using a linear regression model adjusted for age and sex. Associations between blood‐based CKD‐TRS and AD endophenotypes were examined with linear mixed models. Longitudinal cognitive changes of ADAS‐Cog13 in ADNI and ANM were modeled adjusting for age, sex, education, and diagnosis, with the CKD‐TRS × time interaction tested. Temporal cortical thickness on MRI in ADNI and ANM was analyzed adjusting for age, sex, intracranial volume, and diagnosis, with additional adjustment for magnetic field strength in ADNI. Amyloid accumulation assessed by CL on PET and CSF p‐tau181 in ADNI were each analyzed adjusting for age, sex, and diagnosis. In the ADNI and ROSMAP datasets, we performed the sensitivity analyses with the addition of diabetes status as a covariate. In the ANM cohort, diabetes data were not available. Finally, mediation analysis tested whether amyloidopathy mediated the association between CKD‐TRS and longitudinal cognitive decline. For mediation with longitudinal data, we applied a two‐stage slope‐based approach. Participant‐specific slopes of amyloid PET burden and ADAS‐Cog13 scores were estimated using linear mixed‐effects models with random slopes, merged with covariates (age, sex, diagnosis, education), and entered into causal mediation analysis to test whether amyloid changes mediated the association between blood‐based CKD‐TRS and cognitive decline. Pathway enrichment of 890 target genes from SMR results was performed using Metascape on Gene Ontology (GO) biological process (BP). Terms with p < 0.05, ≥ 3 genes, and enrichment factor > 1.5 were retained, with multiple testing corrected by the Benjamini‐Hochberg procedure. Clustering was based on kappa similarity (> 0.3). 28 For brain‐based CKD‐TRS, associations with longitudinal global cognition in ROSMAP were examined using linear mixed models adjusted for age, sex, education, diagnosis, and post mortem interval. Neuropathological amyloidopathy (CERAD) and tauopathy (Braak) were analyzed with linear regression adjusted for age at death, sex, diagnosis at death, and post mortem interval. Mediation analysis further tested whether CERAD scores mediated the relationship between brain‐based CKD‐TRS and cognitive decline, using participant‐specific cognitive slopes as outcomes. All analyses were performed in R (version 4.2.2), with statistical significance set at 0.05. RESULTS Participant characteristics For blood‐based CKD‐TRS, the discovery dataset (ADNI) included 661 participants, while the replication dataset (ANM) included 665 participants. For brain‐based CKD‐TRS, ROSMAP included 622 participants. Age, sex, and educational levels of participants are listed according to clinical diagnostic groups, CU, MCI, and AD (Table 1). TABLE 1Participant characteristics.Blood‐based CKD‐TRSBrain‐based CKD‐TRSVariablesADNI (n = 661)ANM (n = 665)ROSMAP (n = 622) CU (n = 213) MCI (n = 345) AD (n = 103) CU (n = 238) MCI (n = 207) AD (n = 220) CU (n = 201) MCI (n = 168) AD (n = 253) Male, no. (%) 106 (49.7) 201 (58.2) 65 (63.1) 94 (39.4) 89 (42.9) 74 (33.6) 78 (39.0) 62 (36.9) 82 (32.5) Age a , b 76.4 (6.3) 73.2 (7.9) 77.5 (7.7) 74.0 (6.6) 75.4 (6.4) 76.6 (6.6) 84.9 (5.2) 87.0 (4.1) 87.9 (3.5) Educational level b 16.4 (2.7) 15.9 (2.8) 15.8 (3.0) 10.9 (4.5) 9.6 (4.4) 8.5 (4.2) 16.3 (3.4) 16.2 (3.6) 16.4 (3.4) Diabetes, no. (%)8 (3.7)15 (4.3)5 (4.8)NANANA32 (15.9)12 (7.1)20 (7.9) a Age at RNA acquisition. b Values are presented as mean (SD).Abbreviations: AD, Alzheimer's disease; ADNI, Alzheimer's Disease Neuroimaging Initiative; ANM, AddNeuroMerge; CKD, chronic kidney disease; CU, cognitively unimpaired; MCI, mild cognitive impairment; ROSMAP, Religious Orders Study and Rush Memory and Aging Project; SD, standard deviation; TRS, transcriptional risk score.John Wiley & Sons, Ltd. Participant characteristics. CU (n = 213) MCI (n = 345) AD (n = 103) CU (n = 238) MCI (n = 207) AD (n = 220) CU (n = 201) MCI (n = 168) AD (n = 253) 106 (49.7) 201 (58.2) 65 (63.1) 94 (39.4) 89 (42.9) 74 (33.6) 78 (39.0) 62 (36.9) 82 (32.5) 76.4 (6.3) 73.2 (7.9) 77.5 (7.7) 74.0 (6.6) 75.4 (6.4) 76.6 (6.6) 84.9 (5.2) 87.0 (4.1) 87.9 (3.5) 16.4 (2.7) 15.9 (2.8) 15.8 (3.0) 10.9 (4.5) 9.6 (4.4) 8.5 (4.2) 16.3 (3.4) 16.2 (3.6) 16.4 (3.4) Age at RNA acquisition. Values are presented as mean (SD). Abbreviations: AD, Alzheimer's disease; ADNI, Alzheimer's Disease Neuroimaging Initiative; ANM, AddNeuroMerge; CKD, chronic kidney disease; CU, cognitively unimpaired; MCI, mild cognitive impairment; ROSMAP, Religious Orders Study and Rush Memory and Aging Project; SD, standard deviation; TRS, transcriptional risk score. Blood‐based CKD‐TRS calculation For the CKD‐TRS calculation, 904 target genes were prioritized from SMR 14 using GWAS summary statistics for renal function 15 and blood eQTL summary data. 16 First, we excluded 14 genes that were associated with multiple SNPs exhibiting opposing eQTL effects. Among 890 genes, 506 genes in ADNI and 221 genes in ANM were used for the CKD‐TRS calculation due to the availability of transcriptome data. The genes used for the TRS calculation are listed in Table S1. Association between blood‐based CKD‐TRS and kidney function Among 661 individuals in the ADNI cohort, 367 with concurrent transcriptomic and eGFR data were included in the analysis. Blood‐based CKD‐TRS was not significantly associated with eGFR (β = 0.059, p = 0.490). Association of blood‐based CKD‐TRS with longitudinal cognitive changes In ADNI, higher blood‐based CKD‐TRS was significantly associated with faster cognitive decline, as measured by ADAS‐Cog 13 (β < 0.001, p = 0.029; Figure 2, Table S2). This result was replicated in ANM (β = 0.004, p = 0.020; Table S3). After additionally adjusting for diabetes status, the significant association between CKD‐TRS and cognitive decline was still observed in ADNI (β < 0.001, p = 0.028; Table S4). Association between blood‐based CKD‐TRS with longitudinal cognitive changes. Longitudinal cognitive change measured by ADAS‐Cog 13 between first and fourth quartiles of blood‐based CKD‐TRS were compared in ADNI (A) and ANM (B) cohorts. AD, Alzheimer's disease; ADAS‐Cog 13, Alzheimer's disease assessment scale–Cognitive Subscale 13‐item; ADNI, Alzheimer's Disease Neuroimaging Initiative; ANM, AddNeuroMerge; CKD, chronic kidney disease; TRS, transcriptional risk score. Association of blood‐based CKD‐TRS with longitudinal changes in brain atrophy Higher CKD‐TRS was significantly associated with faster temporal lobe cortical thinning (β < −0.001, p = 0.019) in ADNI (Figure 3, Table S2). This result was replicated in ANM (β < −0.001, p = 0.006; Table S3). The significant association between CKD‐TRS and faster temporal lobe cortical thinning was maintained after further adjustment for diabetes status (β < −0.001, p = 0.018; Table S4). Association of blood‐based CKD‐TRS with longitudinal changes in temporal cortical thickness. Longitudinal changes in temporal cortical thickness measured from MRI scans between first and fourth quartiles of blood‐based CKD‐TRS were compared in ADNI (A) and ANM (B) cohorts. ADNI, Alzheimer's Disease Neuroimaging Initiative; ANM, AddNeuroMerge; CKD, chronic kidney disease; MRI, magnetic resonance imaging; TRS, transcriptional risk score. Association of blood‐based CKD‐TRS with longitudinal changes of amyloid deposition on PET In ADNI, higher blood‐based CKD‐TRS was significantly associated with a more rapid increase in CL values (β = 0.028, p = 0.028; Table S2). The highest quartile (fourth quartile) group of blood‐based CKD‐TRS showed faster amyloid accumulation on PET compared to the lowest quartile group (first quartile) (Figure S1). In ANM, amyloid PET data were not available. The association between CKD‐TRS and amyloid accumulation remained significant after further adjustment for diabetes status (β = 0.028, p = 0.027; Table S4). Association of blood‐based CKD‐TRS with longitudinal changes of CSF p‐tau181 biomarker In ADNI, higher blood‐based CKD‐TRS was not associated with changes in CSF p‐tau181 over time (β = −0.006, p = 0.737; Table S2). In ANM, CSF biomarker data were not available. The association between CKD‐TRS and CSF p‐tau181 remained statistically significant even after further adjustment for diabetes status (β = −0.007, p = 0.727; Table S4). Mediation effect of longitudinal changes in amyloid deposition on relationship between blood‐based CKD‐TRS and longitudinal cognitive changes We conducted a causal mediation analysis to assess whether longitudinal changes in amyloid PET mediated the effect of blood‐based CKD‐TRS on changes in ADAS‐Cog 13 scores over time in ADNI. The analysis revealed no significant mediation effect. Specifically, the average causal mediation effect (ACME) was not statistically significant (β < −0.001, p = 0.490). Neither the average direct effect (ADE) (β < 0.001, p = 0.860) nor the total effect (β < 0.001, p = 0.890) was statistically significant. Pathway enrichment analysis of blood‐based CKD‐TRS target genes A total of 890 target genes identified from the SMR results for the blood‐based CKD‐TRS were used for pathway enrichment analysis based on GO BP terms. Terms such as organelle organization, protein phosphorylation, DNA metabolic process, protein polyubiquitination, and epithelial cell proliferation were included. The top 20 clusters of enriched GO BP terms are listed in Table S5. Pathways were grouped into clusters according to similarity. Cluster networks were visualized, excluding isolated clusters without network connections (Figure 4). All significant BP pathways are listed in Table S6. Pathway enrichment analysis of target genes used for blood‐based CKD‐TRS calculation. Target genes (890 genes) used for blood‐based CKD‐TRS calculation were enriched using Gene Ontology Biological Process pathways. Pathways were grouped into clusters depending on similarity, and a network visualization was generated for the top 20 most significant clusters. Isolated clusters without network connections were excluded from the visualization. CKD, chronic kidney disease; TRS, transcriptional risk score. Brain‐based CKD‐TRS calculation For the brain‐based CKD‐TRS calculation, 214 target genes were prioritized from SMR using GWAS summary statistics for renal function and brain eQTL summary data. First, we excluded three genes associated with multiple SNPs that exhibited opposing eQTL effects. We then used 154 genes in ROSMAP for the CKD‐TRS calculation based on the availability of transcriptome data. The genes included in the TRS calculation are listed in Table S1. Association of brain‐based CKD‐TRS with longitudinal cognitive changes In ROSMAP, higher brain‐based CKD‐TRS was associated with faster cognitive decline (β < −0.001, p = 0.032) (Figure 5, Table S7). Following additional adjustment for diabetes status, the significant association between CKD‐TRS and cognitive decline was maintained (β < −0.001, p = 0.032; Table S8). Association of brain‐based CKD‐TRS with longitudinal cognitive changes in the ROSMAP cohort. Longitudinal changes in cognition performance measured by global cognition between first and fourth quartiles of brain‐based CKD‐TRS were compared. CKD, chronic kidney disease; ROSMAP, Religious Orders Study and Rush Memory and Aging Project; TRS, transcriptional risk score. Association of brain‐based CKD‐TRS with CERAD score and Braak staging Higher brain‐based CKD‐TRS was significantly associated with higher amyloid burden assessed by the CERAD score (β = −0.006, p = 0.031) (Figure S2A, Table S7). There was no association between brain‐based CKD‐TRS and tau burden assessed by Braak stage (β = 0.003, p = 0.203) (Figure S2B, Table S7). These associations remained unchanged after further adjustment for diabetes (CERAD score: β = −0.006, p = 0.032; Braak stage: β = 0.003, p = 0.188; Table S8). Mediation effect of CERAD score on relationship between brain‐based CKD‐TRS and longitudinal cognitive changes We conducted a causal mediation analysis to evaluate whether the CERAD score mediated the effect of brain‐based CKD‐TRS on longitudinal changes in global cognition scores. ACME was statistically significant (β < −0.001, p = 0.014). The ADE was not significant (β < −0.001, p = 0.209), as was the total effect (β = −0.266, p = 0.088) (Figure 6). Mediation analysis of amyloidopathy between brain‐based CKD‐TRS and longitudinal cognitive decline. CERAD, Consortium to Establish a Registry for Alzheimer's Disease; CKD, chronic kidney disease; TRS, transcriptional risk score. DISCUSSION This study aimed to develop blood‐ and brain‐based CKD‐TRS and explore the association of CKD‐TRS with AD pathology and cognitive function. We demonstrated that a higher blood‐based CKD‐TRS was associated with faster cognitive decline, faster amyloid accumulation on PET, and more rapid thinning of the temporal lobe cortex. Significant association of blood‐based CKD‐TRS with cognition and temporal lobe thickness was replicated in another independent cohort. Moreover, a higher brain‐based CKD‐TRS was associated with faster cognitive decline and greater amyloid burden based on neuropathological data. Amyloidopathy mediated the relationship between brain‐based CKD‐TRS and longitudinal cognitive changes. These findings suggest that the transcriptional pattern of CKD risk genes is associated with amyloidopathy and subsequent cognitive decline. Previous studies reported an association between CKD and increased amyloid burden. Impaired kidney function has been linked to higher plasma Aβ42 levels 29 and lower CSF p‐tau/ Aβ42 ratios. 30 In the current study, we demonstrated that a higher blood‐based CKD‐TRS predicted faster amyloid accumulation on PET, while a higher brain‐based CKD‐TRS was associated with more severe amyloid pathology, as measured by CERAD scores. These results suggest that CKD and amyloidopathy may share transcriptional mechanisms and that elevated transcriptional risk for CKD could serve as a predictor for amyloid accumulation. Potential mechanisms linking CKD and amyloidopathy in the central nervous system (CNS) have been investigated from multiple perspectives. One study identified shared gene expression profiles between AD and CKD, highlighting common pathways such as KRAS‐signaling, transforming growth factor‐beta signaling, and Wnt‐β‐catenin‐signaling, the latter of which is related to amyloid production. 31 , 32 Additionally, impairment of the glymphatic system in the CNS may contribute to amyloid accumulation. The glymphatic system facilitates the clearance of proteins such as Aβ from the brain, and its dysfunction can promote amyloid accumulation in the CNS. Impairment of the glymphatic system has been reported in patients with CKD, with decreased aquaporin‐4 expression suggested as a potential cause. 33 Moreover, CKD impairs renal clearance of Aβ, and the resulting increase in peripheral Aβ may induce central amyloidopathy. 34 In our study, pathway enrichment analysis of target genes used for blood‐based CKD‐TRS revealed pathway clusters potentially involved in Aβ metabolism. For example, protein polyubiquitination is essential for clearing misfolded proteins such as Aβ via the ubiquitin‐proteasome system, and dysfunction of this system may exacerbate Aβ accumulation. 35 , 36 Pathways related to epithelial cell proliferation may contribute to blood–brain barrier disruption in AD, potentially impairing Aβ clearance and promoting neuroinflammation. Additionally, pathways involved in cell population proliferation may reflect glial activation, including microgliosis and astrogliosis, which can further exacerbate neuroinflammation. 37 , 38 Our findings also showed that amyloid pathology exerted an indirect mediating effect on the relationship between brain‐based CKD‐TRS and cognitive decline. Although direct and total effects were not statistically significant, the identification of a potential mediating factor is noteworthy. These results highlight the importance of amyloid burden in the link between CKD and cognitive decline, suggesting that targeting amyloidopathy may be a viable approach to mitigate cognitive impairment in individuals with CKD. In contrast to amyloid pathology, tau pathology was not significantly associated with CKD‐TRS in this study. While increased plasma tau levels have been reported in patients with CKD, neuropathological studies have not found a strong link between kidney dysfunction and tau deposition in the brain. 2 , 39 This discrepancy suggests that elevated plasma tau in CKD may result from impaired peripheral clearance or systemic effects rather than CNS tau pathology. However, further research is needed to clarify this distinction. In addition, a higher CKD‐TRS was significantly associated with faster reductions in temporal cortical thickness, suggesting a potential molecular link between CKD and neurodegeneration. Previous brain MRI studies in patients with CKD have shown reduced cortical thickness in the temporal and frontal lobes. 40 , 41 Furthermore, a Mendelian randomization study demonstrated that impaired kidney function causally contributed to brain cortical atrophy, including the temporal lobe. 42 Our findings related to cognitive function are also consistent with previous clinical observations. In patients with CKD and end‐stage renal disease, impaired multi‐domain cognitive function, particularly in executive and memory functions, has been reported. 43 , 44 , 45 In animal studies, CKD induced neuronal death in the hippocampus and synaptic loss, resulting in reduced cholinergic activity and cognitive impairment. 46 , 47 Although executive dysfunction, memory impairment, and CKD‐related markers such as albuminuria have been proposed as predictors of dementia in CKD, no consensus has been reached on robust biomarkers. 45 , 48 CKD‐TRS, as a transcription‐based risk score, may serve as a promising biomarker for neurodegenerative changes, supporting the prediction of cognitive decline and the development of targeted therapeutic strategies. In the ADNI cohort, we did not observe a significant association between blood‐based CKD‐TRS and kidney function as assessed by eGFR. This finding suggests that CKD‐TRS may capture biological mechanisms underlying CKD susceptibility that are partially independent of glomerular filtration, which may explain its association with neurodegenerative outcomes despite the absence of a direct relationship with eGFR. 49 Furthermore, from a methodological perspective, the diagnosis of CKD requires evidence of persistent kidney dysfunction on at least two occasions separated by a minimum of 3 months. However, in ADNI, creatinine was measured only once. In addition, there was a temporal mismatch between creatinine assessment and transcriptomic profiling, resulting in a reduced sample size for analysis. The ADNI cohort also includes participants with relatively preserved renal function, and the limited variability in eGFR may have reduced the ability to detect associations. 50 This study has several limitations. First, the sample size available for analysis was moderate, which may have limited statistical power. However, the findings were successfully replicated in an independent dataset, supporting their robustness and potential applicability to similar cohorts. Second, for TRS calculation, we included only target genes identified through SMR analysis that were available in the ADNI and ANM datasets; however, the absence of certain genes may have influenced the findings. Third, microarray‐based gene expression data were used to calculate the TRS. As RNA‐seq becomes more widely available, future studies may replicate our findings using cohorts with RNA‐seq data. Lastly, although paired analysis of blood‐ and brain‐based CKD‐TRS within the same individuals would strengthen the validity of our findings, the cohort design of the present study precluded such direct comparison; future studies leveraging longitudinal cohorts with paired ante mortem and post mortem samples would be an important next step. In summary, we demonstrated that a higher blood‐based CKD‐TRS was associated with faster cognitive decline and greater temporal cortical atrophy, suggesting its potential utility as a biomarker for predicting neurodegenerative changes. Furthermore, brain‐based CKD‐TRS – representing more direct molecular changes in the brain – was also significantly associated with accelerated cognitive decline. Amyloid pathology was found to mediate this relationship, supporting a mechanistic link between CKD‐related transcriptional signatures and AD pathology. Overall, these findings highlight the potential of CKD‐TRS as a molecular bridge between kidney dysfunction and neurodegeneration. Future studies employing cell‐type‐specific transcriptomics, proteomics, and metabolomics are warranted to further elucidate the complex biological interplay between CKD and AD. CONFLICT OF INTEREST STATEMENT The authors declare no conflicts of interest. Author disclosures are available in the Supporting Information. CONSENT STATEMENT Informed consent was obtained from all participants, and ethical approval was granted by the investigators at each participating site from the ADNI, ANM, and ROSMAP. Supporting information Supporting Information Supporting Information Supporting Information Supporting Information Supporting Information Supporting Information Supporting Information Supporting Information Supporting Information Supporting Information Supporting Information