In a recent study, Bai et al. performed deep multilayer proteomics of postmortem brain samples of AD patients and 5xFAD mice models at 3, 6, and 12 months of age using mass spectrometry to identify the molecular network and biomarkers underlying Alzheimer’s disease (Bai et al., 2020). We used the human AD proteomic datasets to identify differentially expressed proteins in the disease state compared to the control samples. The human dataset contained 18 post-mortem samples each for the control and AD group. Each group had two independent pools (9 case per pool). The mean age of the control group was 83 years, whereas that of AD group was 76 years. The 5xFAD mice model (n = 4) and wild-type (n = 4) samples were then used to confirm the expression difference of the selected genes. Additional details on the samples and their proteomic profiling can be found in the original article (Bai et al., 2020). The differential expression of the proteomics data was performed through the amica tool² (Didusch et al., 2022) using the raw intensity values. The amica web-tool has been specifically designed for proteomic analysis and uses limma R package (Ritchie et al., 2015) to assess the differential expression of proteins.

The hippocampus mRNA expression dataset in the current study was generated in our group. The standardized dataset can be accessed through our GeneNetwork portal³ by selecting the group as “BXD Family,” type as “Hippocampus mRNA,” and the dataset name as “Hippocampus Consortium M430v2 (June 6) RMA” (GEO Series: GSE84767). A brief description of the experimental protocols used for generating this dataset is provided below, and additional details can be found on our GeneNetwork portal.

The hippocampus data used in the current study was from 71 genetically diverse strains of mice, including 67 BXD recombinant inbred strains, their parental strains (C57BL/6 J, DBA2/J), and two reciprocal F1 hybrids. The majority of the animals were between 45 and 90 days old (mean 66 days, ranging from 41 to 196 days). The mice were obtained from the University of Tennessee Health Science Center (UTHSC), the University of Alabama (UAB), or directly from the Jackson Laboratory. The animals were housed and maintained at UTHSC prior to execution. Animals were deeply anesthetized with 4–5% isoflurane in oxygen until loss of pedal and corneal reflexes, followed by cervical dislocation. Brains were then removed and placed in an RNAlater solution prior to dissection. Cerebella and olfactory bulbs were removed; brains were hemisected, and both hippocampi were dissected whole. All procedures involving mouse tissue were approved by the Institutional Animal Care and Use Committee at the University of Tennessee Health Science Center.

A pool of dissected tissue, typically from six hippocampi and three naive adults of the same strain, sex, and age, was collected and used to generate cRNA samples. A total of 215 RNA samples were extracted at UTHSC. Briefly, the RNA STAT-60 protocol was used according to the manufacturer’s instructions, which included tissue homogenization, RNA extraction, RNA pretreatment, RNA washing, and RNA purification. Finally, RNA purity was evaluated using the 260/280 nm absorbance ratio, and the samples with values greater than 1.8 were considered. The RNA integrity of the samples was assessed by the Agilent Bioanalyzer 2,100, and those with an RNA integrity number (RIN) greater than 8 were considered for microarray hybridization.

Samples were processed in the INIA Bioanalytical Core at the W. Harry Feinstone Center for Genomic Research, The University of Memphis. RNA samples from 2 to 3 animals of the same age, strain, and sex were hybridized to a single Affymetrix GeneChip Mouse Expression 430 2.0 short oligomer array. The Affymetrix 430v2 arrays consist of ~993,000 25-mer nucleotide probes that can profile the expression of approximately 39,000 transcripts and most known genes and expressed sequence tags. The raw microarray data were normalized using the robust multi-array average (RMA) method (Irizarry et al., 2003) and further converted to an improved z score (2z + 8) (Irizarry et al., 2003). Briefly, in Z-score normalization, instead of leaving the mean at 0 and the standard deviation at 1 unit, we shift up to a mean of 8 units and increase the spread by having a standard deviation of 2 units (2Z + 8 normalized data). This removes the negative values from the tables. Additional details on RNA extraction and hybridization protocols can be found in NCBI-GEO under the accession GSE84767.

Pearson correlation analysis was used to identify Cthrc1 co-expressed genes in the hippocampus or learning- and memory-related traits associated with the genes being explored. The R values with a p < 0.05 were considered statistically significant. The mRNA expression values from the “Hippocampus Consortium M430v2 (June 6) RMA” dataset were used for the correlation analysis. The hippocampal gene expression dataset and cognitive phenotype traits can be accessed through our GeneNetwork portal (Mulligan et al., 2017) (see Text footnote 3). The correlation analysis was performed using the corrplot v0.92 package⁴ in R.

Phenome-wide association studies (PheWAS) have emerged as a viable reverse genetic strategy in humans (Li et al., 2018). Recently this approach has been applied to the BXDs, enabling the discovery of novel gene-phenotype associations. This has been possible owing to the availability of a large amount of genotypic data and thousands of phenotypic traits for BXDs, the largest mouse genetic reference population. Genes that contain high-impact variants, including missense, splice site, and frameshift mutations, as well as genes that have significant cis-e(p) QTLs in the BXD transcriptome and proteome datasets, were included in the PheWAS analysis. The association between the genetic variants of each gene, represented by the SNPs, as well as their cis-QTLs were associated with about 5,000 clinical phenotypes. Here, we used a multi-locus mixed-model approach (MLMM) to estimate the associations between each gene and cognitive traits.

In addition, associations between transcript/protein expression and phenotypic traits were estimated using the mixed model regression analysis through ePheWAS. Transcript-trait pairs with fewer than 15 overlapping lines were removed from the analysis. Bonferroni correction was used to perform phenotype-wide significance analysis. The PheWAS and ePheWAS analyses were performed on Systems Genetics at EPFL⁵ (Li et al., 2018), and those with a –log10(p) ≥ 2 were considered significant.

The functional enrichment analysis of Cthrc1-correlated genes was performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID)⁶ (Sherman et al., 2022) to identify significantly enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways and Gene Ontology biological processes (GO-BP). Mammalian Phenotype Ontology (MPO) analysis was performed using the WEB-based Gene SeT AnaLysis Toolkit (WebGestalt)⁷ (Liao et al., 2019). For this enrichment analysis, “protein coding genes” was selected as the reference set, while ‘minimum number of genes per category’ was kept as 5. The p-values were corrected using the Benjamini-Hochberg method for multiple testing, and annotations with an FDR-adjusted p < 0.05 were considered statistically significant. Furthermore, the genes were submitted to the Metascape tool⁸ (Zhou et al., 2019) to explore their relationships based on GO-BP terms in the form of a clustered network. The Metascape tool employs a heuristic algorithm to select the most informative terms from the GO clusters. It samples 20 top-score clusters, selects up to the 10 best-scoring terms (lowest p-values) within each cluster, then connects all term pairs with Kappa similarity above 0.3.

The Cthrc1-coexpressed genes that were involved in the neurodevelopmental pathways were analyzed to predict their functional association network using the GeneMANIA tool⁹ (Warde-Farley et al., 2010). It connects the genes using a very large set of functional association data, including protein and genetic interactions, pathways, co-expression, co-localization, and protein domain similarity. The GeneMANIA cytoscape app¹⁰ was used for further analysis and visualization of the network.

The genes important in cognition and AD were collected from multiple gene-function resources using the keywords “cognition” or “Alzheimer.” These keywords were searched in the following databases/repositories: Mouse Genome Informatics (MGI) phenotype/disease association (Human-Mouse: Disease Connection)¹¹ (Smith et al., 2018), the NHGRI-EBI GWAS Catalog¹² (Buniello et al., 2019), and the International Mouse Phenotyping Consortium (IMPC)¹³ (Groza et al., 2023). Translation of human gene lists to mouse orthologs was done using the bioDBnet tool¹⁴ (Mudunuri et al., 2009).

The brain cell type expression of the genes, including that of Cthrc1, was obtained from the Mouse Cell Atlas (MCA)¹⁵ database (Fei et al., 2022) using the tissue type as “Adult Brain.” The MCA database uses single-cell RNA sequencing to determine the cell type composition of major mouse organs.

The SH-SY5Y human neuroblastoma cell line was obtained from ATCC (CRL-2266). These cells were cultured in DMEM/F12 (1:1) containing 10% fetal bovine serum and maintained at 37 °C in a humidified 5% CO₂ incubator.

CTHRC1 CRISPR RNA (crRNA) (AUGGCAUUCCGGGU ACACCUGUUUUAGAGCUAUGCU) was synthesized by Azenta. crRNA annealed with tracrRNA (IDT), then incubated with Alt-R^® S.p. Cas9 Nuclease V3 (IDT). Generation of a CTHRC1 knockdown SH-SY5Y cell line was achieved with the Cas9/guide RNA ribonucleoprotein complex (Cas9/RNP) delivered directly to the cells by electroporation. All kits and equipment were provided by Thermo Fisher Scientific. Programmed condition (1,500 V, 20 ms pulse width, 2 pulses) was chosen for electroporation.

A CTHRC1 wild-type cDNA was cloned in the pLVX-TetOne-Puro vector. SH-SY5Y cells were then infected by the lentivirus encoding full-length CTHRC1 or empty virus.

The total cellular RNA was extracted by RNeasy^® Mini Kit (QIAGEN), and RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) was used to reverse transcribe into cDNA. Then real-time quantitative PCR was performed using SYBR^™ Green PCR Master Mix (Applied Biosystems). Data were analyzed using the 2^-ΔΔCT method after standardization with β-actin expression levels in each sample. The primers are listed in Supplementary Table 1.

Whole-cell lysates were prepared using RIPA buffer (Sigma), and protein concentrations were quantified with the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of protein were resolved on 4–12% Bis-Tris Midi Protein Gels (Thermo Fisher Scientific) and transferred onto membranes using the iBlot^™ 2 system (Thermo Fisher Scientific). Membranes were blocked with 5% milk for 1 h at room temperature and subsequently incubated with primary antibodies followed by HRP-conjugated secondary antibodies. The following primary antibodies were used: SNCAIP polyclonal antibody (Proteintech, #17818-1-AP; 1:300), Tau (Tau 46) monoclonal antibody (Santa Cruz Biotechnology, #sc-32274; 1:200), and Actin antibody (Cell Signaling Technology, #4967S; 1:1000). Secondary detection was performed using HRP-conjugated anti-rabbit IgG (Cell Signaling Technology, #7074; 1:2000) and HRP-conjugated anti-mouse IgG (Cell Signaling Technology, #7076; 1:2000).

The CTHRC1 gene encodes for collagen triple helix repeat containing 1. Mutations at this locus have been associated with Barrett esophagus and esophageal adenocarcinoma. The differential expression analysis at the protein level between AD and LPC identified CTHRC1 as the second most upregulated protein based on fold-change. It was found to be upregulated with approximately 5-fold (adj. p = 0.05) in AD patients compared to the controls (Figure 2A). This protein also showed a significant increase in its expression in 6-month-old 5xFAD mice compared to their wild-type counterparts (Figure 2B). Furthermore, the cell-type-specific expression of Cthrc1 based on single-cell RNA sequencing of mouse brain demonstrated its expression predominantly in astrocytes and oligodendrocyte progenitor cells. Cthrc1 was also found to be expressed to some extent in GABAergic neurons (Figure 2C).

Next, we investigated the mRNA expression of Cthrc1 in the hippocampus (the brain region primarily associated with learning and memory) of 69 BXD mice and two parental strains (C57BL/6 J or B6, and DBA/2 J or D2) using our GeneNetwork database. The mean expression of Cthrc1 (probe-set 1452968_at) across BXD strains was found to be 8.7 ± 0.19 SD. BXD31 had the least expression (value = 8.3), whereas BXD73a showed the highest expression (value = 9.2) for Cthrc1 mRNA with a fold difference of ~1.9 across the BXD strains. Furthermore, we observed a difference of ~1.3 between both the parental strains, with D2 exhibiting higher levels (value = 9.0) of Cthrc1 than that of B6 (value = 8.7) (Figure 2D). PheWAS analysis using thousands of genomic variants and >5,000 phenotypes from BXD mice suggested significant association of Cthrc1 SNPs with 10 nervous system-related phenotypes. The phenotypes that were included in the analysis belonged to locomotor activity, depression assay, fear conditioning, and other traits (Figure 2E). Similarly, ePheWAS analysis indicated that Cthrc1 expression in the BXD hippocampus modulates multiple nervous system-related traits, including phenotypes, such as hippocampus dentate gyrus cells and motor coordination (Figure 2F). When we correlated the hippocampal Cthrc1 mRNA levels with learning and memory phenotypes (n = 341) in BXD mice, the results indicated a significant (p < 0.05) correlation of 22 cognition-related phenotypes with Cthrc1 expression (Figure 2G), suggesting that Cthrc1 mRNA levels in the BXD hippocampus modulate the phenotypic traits related to cognition. The GWAS results indicated a significant association of variants located in the genomic locus of CTHRC1 with AD and other brain-related traits, highlighting a strong association of this gene with neurodevelopmental disorders (Figure 2H).

We performed genome-wide eQTL mapping using the GEMMA mapping method to investigate the genomic regulation of Cthrc1 mRNA expression. A -log10(p-value) of 4 was used as the significance threshold to determine whether Cthrc1 expression is genomically regulated by epistatic loci. Our results identified mapping of a significant locus for Cthrc1 on Chr 15 at 38.99 Mb (rs3660608) (Figures 3A,B). The mapped locus was very close to the genomic position of the Cthrc1 gene (Chr 15 @ 39.08 Mb), indicating that Cthrc1 mRNA expression in the hippocampus is cis-regulated. Thus, Cthrc1 may genetically regulate other downstream genes in the hippocampus. Statistical analysis of Cthrc1 expression between the two parental genotypes (B6 and D2) according to the eQTL peak position (rs3660608) showed that BXD mice with the D2 allele exhibit significantly higher expression than those carrying the B6 allele (p = 3.48E-05, Figure 3C).

To gain better insights into the pathways and biological functions of Cthrc1 involved in the hippocampus, we performed a Pearson genetic correlation analysis between Cthrc1 mRNA expression and the entire BXD hippocampal transcriptome data [Hippocampus Consortium M430v2 (June 6) RMA]. At an uncorrected p-value of <0.05, 12,475 probes corresponding to 9,305 genes were found to be correlated with Cthrc1 expression. To further shortlist the genes that are correlated with high confidence, we selected those that had a mean expression of >7 in the BXD hippocampus. This resulted in a total of 9,728 probes corresponding to 7,705 genes. Large-scale transcriptomic datasets commonly yield thousands of correlated genes because hippocampal gene expression is highly modular and dominated by co-regulated networks and shared pathway activity (Oldham et al., 2008). In systems genetics resources such as the BXD panel, the large sample size further increases power to detect broad correlation structures, many of which reflect coordinated biological processes rather than direct regulation (Chesler et al., 2005). Thus, the correlation of Cthrc1 with such a large number of genes in the hippocampus is implicative of its key role in hippocampal biology and the associated functions. To validate this hypothesis, we performed functional analysis, including KEGG pathway, GO, and MPO enrichment. The results indicated significant representation of cognition-related annotations, further pointing towards the importance of Cthrc1 in neurodevelopmental disorders. With an FDR P cutoff of <0.05, we identified a total of 319 GO-BPs, 300 MPOs, and 170 KEGG pathways. As shown in Figure 4A, “protein transport” was found to be the most enriched GO-BP with 300 genes (FDR p = 2.67E-29), followed by “positive regulation of transcription from RNA polymerase II promoter” with 483 genes (FDR p = 9.14E-24). The top 50 GO-BPs included several nervous system-related annotations, such as “axon guidance,” “neuron projection development,” “axonogenesis,” and “hippocampus development” (Supplementary File 1). The network of GO-BP terms constructed using the Metascape tool revealed sharing of genes among the annotations related to axon development and hippocampus development. Furthermore, interactions among the terms related to synapse assembly, autophagy, and catabolic processes were also observed. Interestingly, the cluster associated with “cytoskeletal-dependent intracellular transport” formed an independent group and included terms related to synaptic vesicle transport and localization (Figure 4B). Furthermore, the results from the MPO enrichment analysis were found to be in similar lines with that of GO-BPs. Around 15 of the top 20 MPO terms enriched were related to the nervous system (Figure 4C). Some of the important enriched terms included “abnormal nervous system physiology” (n genes = 726; FDR p = 2.56E-16), “abnormal brain morphology” (n genes = 639; FDR p = 2.56E-14), “abnormal cognition” (n genes = 300; FDR p = 4.45E-11), “abnormal hippocampus morphology” (n genes = 145; FDR p = 1.44E-10), “abnormal learning/memory/conditioning” (n genes = 300; FDR p = 3.88E-11), and “abnormal neuron morphology” (n genes = 533; FDR p = 2.56E-11). KEGG pathway analysis demonstrated enrichment of multiple nervous system-related pathways. As shown in Figure 4D, “pathways of neurodegeneration” and “Alzheimer disease” were among the top 5 pathways, based on the percentage of genes involved. While “pathways of neurodegeneration” involved 197 Cthrc1-correlated genes, the AD pathway involved 155 genes. Together, both these pathways included 229 Cthrc1-correlated genes. The pathway analysis results thus implicate that Cthrc1 may be linked to the neurodegeneration pathways through these 229 genes. The complete list of significant GO-BPs, MPOs, and KEGG pathways is provided as Supplementary File 1.

Together, “pathways of neurodegeneration” and “Alzheimer disease” pathways include 229 Cthrc1-correlated genes. Hence, we explored the interaction of these genes with Cthrc1 using functional association data from GeneMANIA to identify the primary partners of Cthrc1 and eventually link Cthrc1 to these pathways. The interaction network (Supplementary Figure 1A) contained 227 query genes (including Cthrc1) that were connected with 8,648 edges. The large number of interactions among these genes was expected owing to their involvement in the same pathway. We extracted the Cthrc1 subnetwork from the large primary network. The subnetwork contained 18 genes (including Cthrc1) connected with 97 edges (Supplementary Figure 1B). The subnetwork further revealed that Cthrc1 physically interacts with Wnt3a and Fzd6 while functionally interacting with the other 15 partners. The nodes BACE1, CASP9, SLC39A6, and GRIN2C were found to be differentially regulated at the protein level in 6-month-old 5xFAD mice compared to their wild-type counterparts. Furthermore, Fzd1, Nefm, Fzd6, Bace1, and Irs1 have been implicated in cognition-related properties or disorders based on collected data from MGI/IMPC/GWAS Catalog. Interestingly, several of these genes in the Cthrc1 subnetwork were negatively correlated with Cthrc1 mRNA expression in the BXD hippocampus (Table 1). Next, we wondered if Cthrc1 is linked to the well-known AD/neurodevelopmental disorder genes (ND genes) through its primary interactors. To this end, we explored the interaction of Cthrc1-primary interactors with the well-known ND genes, including App, Mapt, Apoe, Psen1, and Psen2. Our analysis revealed that several of the Cthrc1-primary interactors were directly connected to multiple ND genes. In particular, Tubb2b, Vdac1, and Nefm physically interact with one of the five ND genes. Furthermore, all five ND genes were found to be upregulated in AD compared to the control, either in the 5xFAD mouse model or in human samples at the protein level (Supplementary Figure 1C). Thus, the network analysis results demonstrated that although Cthrc1 does not directly interact with the ND genes, it affects their function through its primary interacting partners in the same pathway. Furthermore, it is noteworthy that all the ND genes shown in the network were significantly negatively correlated with Cthrc1 in the BXD hippocampus (Table 1), implying a protective role of Cthrc1 on neurodevelopmental disorders, such as AD.

While Cthrc1 is significantly correlated with multiple learning and memory phenotypes from BXD mice (Figure 2G), we intended to explore whether its primary interactors also correlate with the cognitive phenotypes. For the correlation analysis, a total of 341 learning and memory phenotypes belonging to different groups (e.g., Y-maze test, Morris water maze test, and contextual fear conditioning) were retrieved from our GeneNetwork portal. These were then correlated with the hippocampal mRNA expression of the 17 genes that directly interact with Cthrc1 in the functional association network (Supplementary Figure 1B). Of the total 341 phenotypes (Figure 5A), 190 learning- and memory-related traits were correlated with the hippocampal expression of at least one of the 17 genes (Figure 5B). Among these 190, 36% (n = 68) belonged to contextual fear conditioning, whereas 29.5% (n = 56) belonged to the y-maze test (Figure 5B). Furthermore, among the 17 genes, Nefl correlated with the most number of phenotypes, followed by Casp9 with 30 and 29 significant correlations, respectively. Grin2c correlated with the least number of learning and memory phenotypes (Figure 5C). The Y-maze test is a behavioral task that is used to study spatial learning and memory, which is underlined by the hippocampus (Kraeuter et al., 2019). We (and our collaborators) have collected several traits related to the Y-maze in BXD mice that can be accessed through our GeneNetwork portal. The correlation of the 17 Cthrc1-primary interactors with Y-maze phenotype traits indicated significant correlation with 56 of the total 96 traits phenotypes (Figures 5B,D). Of the 17 genes, Wnt3a was found to be correlated with the most number of Y-maze-related traits (n = 14), followed by Vdac1 (n = 13). Interestingly, Wnt3a was found to be physically interacting with Cthrc1 in the functional association network and was connected to all five ND genes. On the contrary, Vdac1 was functionally connected to Cthrc1; however, it was physically interacting with microtubule-associated protein tau (Mapt) in the network (Supplementary Figure 1C). The two other genes that correlated with a higher number of Y-maze traits were Casp9 and Bace1 (n = 12 and 11, respectively). There were seven Y-maze traits, each of which was significantly correlated with 4 Cthrc1-primary interactors (Figure 5D; Supplementary File 2).

To explore which brain cell types express Cthrc1-interactors, we analyzed the single-cell sequencing data of adult mouse brains from MCA. Our results revealed that many of the genes were predominantly expressed in four brain cell types, viz., astrocyte, microglia, oligodendrocyte, and GABAergic neuron, among the 15 brain cell types investigated. Furthermore, Bace1, Chuk, Vdac1, and Tubb2b were found to be expressed in the greatest number of cell types. While the first three genes, Bace1, Chuk, and Vdac1, were expressed in 12 of the 15 brain cell types, Tubb2b was found to be expressed in 11 of the 15 brain cell types investigated (Supplementary Figure 2). None of the genes were expressed in neutrophils. The expression of Tubb2b was found to be highest in GABAergic neurons, which was also highest among all cell types checked. In addition, it showed high expression in a few other cell types as well, including astrocyte and oligodendrocyte progenitor cells. The expression of Bace1 was found to be the highest in oligodendrocytes, whereas Chuk was found to be highly expressed in neurons. The expression of Vdac1 was found to be high in ependymal cells and granulocytes. Surprisingly, none of the investigated brain cells expressed Wnt3a or Wnt2 (Supplementary Figure 2).

We used GWASCatalog¹⁶ to explore the genetic linkage of Cthrc1-interacting genes with the neurodevelopmental or cognition related traits in humans. The mutations in 10 of the 17 genes explored were significantly associated with one or more brain related traits in the human population. The following genes were associated with significant cognition or neurodevelopmental related traits: BACE1, FZD1, INS, IRS1, NEFL, NEFM, PIK3R3, SNCAIP, and WNT3A. SNCAIP (synuclein alpha interacting protein) showed association with the maximum number of traits (n = 12). The traits were mainly related to cognitive performance, Parkinson’s disease, white matter, and neurofibrillary tangles (Supplementary Table 2). The next best gene in terms of the number of associations was NEFL, which was correlated with five traits. Both FZD6 and CTHRC1 were associated with AD related traits, indicating a close association between both these molecules, which was also evident from the functional association network analysis.

We analyzed the changes in the expression of key candidate genes following CTHRC1 knockdown or overexpression in the SH-SY5Y human neuroblastoma cell line, which is often used as an in vitro model for studying neuronal function and differentiation.

Firstly, we used sgRNA technology to knockdown the expression of CTHRC1 in SH-SY5Y cells using different sgRNA constructs. The preliminary data showed that CTHRC1-sgRNA3 could achieve 67% knockdown efficiency; hence, it was used for further knockdown experiments. Following CTHRC1 knockdown, the expression of SLC39A6 was found to be reduced, while that of GRIN2C was elevated, which was consistent with our in silico analysis. However, no significant changes in the expression levels of BACE1 and CASP9 were observed (Figure 6A).

Further, we overexpressed CTHRC1 in SH-SY5Y cells by using Dox during passages. Interestingly, overexpression of CTHRC1 led to downregulation of several genes, including SNCAIP, NEF1, FZD1, TUBB2B, VDAC1, PIK3R3, NEFM, CHUK, FZD6, WNT3A, and PSEN1, and upregulation of WNT2 (Figure 6B). Our western blotting data indicated that overexpression of CTHRC1 in SH-SY5Y cells led to tau degradation. Furthermore, a slight decrease in SNCAIP protein was observed following CTHRC1 overexpression, pointing out a potential protective role of this gene in AD (Figure 6C).

We constructed an integrated molecular network to visualize how CTHRC1 may interface with established Alzheimer’s disease pathways (Supplementary Figure 3). The resulting map highlights multiple points of convergence between CTHRC1-associated genes and core AD risk factors, including APP, PSEN1, PSEN2, and APOE. Notably, CTHRC1-linked signaling modules—including Wnt/Fzd1 activation, MAPT regulation, VDAC1-associated mitochondrial pathways, and SNCAIP/NEFM synaptic components—feed into APP processing and presenilin-dependent cleavage steps. These interactions converge on downstream processes such as amyloidogenic APP processing, tau-related mechanisms, mitochondrial stress, and caspase-mediated injury, collectively pointing toward a multilayered contribution of the CTHRC1 subnetwork to neurodegenerative outcomes.