Recently, a large clinical study revealed an inverse correlation of individual risk of cancer versus Alzheimer’s disease (AD). However, no explanation exists for this anticorrelation at the molecular level; however, inflammation is crucial to the pathogenesis of both diseases, necessitating a need to understand differing signaling usage during inflammatory responses distinct to both diseases. Using a subpathway analysis approach, we identified numerous well-known and previously unknown pathways enriched in datasets from both diseases. Here, we present the quantitative importance of the inflammatory response in the two disease pathologies and summarize signal transduction pathways common to both diseases that are affected by inflammation.
Epidemiological evidence has revealed an inverse incidence between Alzheimer’s disease (AD) and cancer that increases exponentially among aged cohorts [
Of interest, immune response is intimately related to both diseases [
Despite continuous efforts to understand the individual molecular mechanisms of the two diseases, distinction of the global effects of immune response toward specific signal transduction usage in the two diseases has not been established. Here, we systematically inspected the two diseases representing phenotypically opposite cell fates, death and survival, by utilizing functional enrichment analysis and a systems biology approach [
Throughout the paper, we compared one colorectal cancer (CRC) dataset (GEO accession GSE4107) [
IPA functional enrichment of the CRC and the AD datasets. (a) Top 5 functional categories from “Diseases and Functions” ontology for the datasets are represented. The
Since cancer and AD are phenotypically opposite (cell survival versus cell death), we obtained oppositely expressed common genes between the two diseases. Based on all the genes’ fold-changes from the three datasets, we obtained the common genes as shown in Figure
For generating networks from the three datasets, we applied our previous subpathway-based systems biology approach [
While cancer and AD are two of the most common diseases worldwide (15.6 million versus 7.7 million new cases per year) relating to aging, their phenotypes are opposite: cell death (neurons) in AD versus survival (mostly epithelial cells) in cancer. Also, AD patients are less susceptible to cancer and vice versa [
We used Ingenuity Pathway Analysis (IPA) to perform functional pathway enrichment of early CRC and AD. IPA reported the top 5 functional categories from its “Diseases and Functions” ontology. In Figure
Figure
Out of the common genes in Figure
We further dissected the common genes (28 and 35 genes in white circles in the Venn diagram in Figure
Inflammation-associated genes common to both AD and CRC show opposite expression patterns. The 16 oppositely expressed common genes (in Figure
Functional category | Downregulated in AD and upregulated in CRC | Upregulated in AD and downregulated in CRC |
---|---|---|
Chemokine | PTPN6 |
BAD |
|
||
Inflammation relating to CRC | DDIT3 |
|
|
||
Inflammation relating to brain | CCR6 |
PPARD |
|
||
Cytokines relating to cancer | CD28 |
ABL1 |
|
||
Cytokines relating to brain | CD36 |
#Genes detected in the AD network from GSE12685 dataset.
Next, we constructed molecular networks for the two diseases. By applying our previous systems biology method to the three disease datasets, we obtained CRC and AD pathogenesis networks (Supplementary Figures S1–S3). We summarized the most significant 100 subpathways for each network (Supplementary Tables S3–S5) in order to see the signaling in detail. These subpathways were assigned to various pathways in CRC and AD (Supplementary Tables S3–S5), suggesting that, in addition to inflammatory response inferred by our functional enrichment comparison, those pathways (not assigned to inflammation) remain largely unexplored in CRC or AD. Of interest, we found pathways previously unassociated with the two diseases, including Hedgehog signaling, axon guidance, ECM-receptor interaction, and WNT signaling (Table
The AD datasets were prepared from frontal cortex synaptoneurosomes and hippocampi. Both brain regions include neurons, as well as astrocytes and microglia [
So, we inspected the 16 genes’ (in Table
KEGG pathways associated with the 16 oppositely expressed common genes (in Table
Gene symbols | Pathways | CRC (GSE4107) | AD (GSE12685) | AD (GSE1297) |
---|---|---|---|---|
PTPN6 | hsa04662_B_cell_receptor_signaling_pathway; hsa04630_Jak-STAT_signaling_pathway; hsa05140_Leishmaniasis | Up | Down | Down |
IRAK3 | hsa04722_Neurotrophin_signaling_pathway | |||
FLT3 | hsa05221_Acute_myeloid_leukemia | |||
DDIT3 | hsa04010_MAPK_signaling_pathway | |||
FAS | hsa04115_p53_signaling_pathway; hsa04650_Natural_killer_cell_mediated_cytotoxicity | |||
IRF3 | hsa04622_RIG-I-like_receptor_signaling_pathway; hsa04623_Cytosolic_DNA-sensing_pathway | |||
CCR6 | hsa04060_Cytokine-cytokine_receptor_interaction; hsa04062_Chemokine_signaling_pathway | |||
CD28 | hsa04660_T_cell_receptor_signaling_pathway; hsa05416_Viral_myocarditis | |||
FCER1G | hsa04650_Natural_killer_cell_mediated_cytotoxicity | |||
NGFR | hsa04722_Neurotrophin_signaling_pathway | |||
FN1 | hsa04512_ECM-receptor_interaction | |||
|
||||
BAD | hsa04510_Focal_adhesion; hsa05223_Non-small_cell_lung_cancer; hsa05210_Colorectal_cancer | Down | Up | Up |
CD36 | hsa03320_PPAR_signaling_pathway; hsa04512_ECM-receptor_interaction | |||
PPARD | hsa05221_Acute_myeloid_leukemia; hsa04310_Wnt_signaling_pathway | |||
ABL1 | hsa04012_ErbB_signaling_pathway; hsa04722_Neurotrophin_signaling_pathway | |||
EGFR | hsa05214_Glioma; hsa04012_ErbB_signaling_pathway |
Subpathways previously not associated with the two diseases. These subpathways were selected from the most significant 100 subpathways in each network. Subpathway (linear signaling flow) with fold-change (the numeral in parenthesis) of the disease group over the control group is represented in each dataset. The most significant 100 subpathways for each dataset are provided in Supplementary Tables S3–S5. The notation in the flow is “B <- A: A activates B” and “B ∣- A: A represses B.”
KEGG pathway | GSE4107 (CRC) subpathway; |
GSE1297 (AD) subpathway; |
GSE12685 (AD) subpathway; |
---|---|---|---|
Hedgehog signaling (hsa04340) | PTCH1 (1.863) <- GLI2 (2.878) ∣- CSNK1G1 (0.587); 0.000035 | PTCH2 (0.938) <- GLI3 (0.682) ∣- GSK3B (1.513); 0.0015 | |
WNT3 (3.147) <- GLI2 (2.878) ∣- CSNK1G1 (0.587); 0.000223 | |||
|
|||
Axon guidance (hsa04360) | PAK3 (0.732) <- RAC1 (0.943) ∣- PLXNB3 (1.627) <- SEMA4C (1.283); 0.0008 | CFL1 (1.157) ∣- LIMK1 (0.896) <- PAK4 (0.871) <- RAC3 (0.892) <- PLXNA3 (0.954) <- FES (0.841); 0.0011 | |
|
|||
WNT signaling (hsa04310) | JUN (4.179) <- TCF7L1 (2.735) <- CTNNB1 (2.562) ∣- GSK3B (0.735) ∣- DVL3 (1.608) <- FZD10 (6.256) <- WNT3 (3.147) <- PORCN (1.279); 0.000114 | ||
JUN (4.179) <- TCF7L1 (2.735) <- CTNNB1 (2.562) ∣- GSK3B (0.735) ∣- DVL3 (1.608) <- APC2 (2.201) <- AXIN2 (2.307) <- CSNK1A1 (1.963); 0.00016 | |||
|
|||
Pathways in cancer (hsa05200) | MMP2 (3.031) <- JUN (4.179) <- MAPK1 (2.425) <- MAP2K1 (1.162) <- ARAF (4.631) <- HRAS (1.027) <- SOS1 (1.624) <- GRB2 (1.613) <- IGF1R (2.299) <- IGF1 (2.529); 0.000022 | ||
|
|||
ECM-receptor interaction (hsa04512) | SDC2 (3.091) <- TNC (9.557); 0.000026 | SDC3 (0.849) <- COL5A2 (0.162); 0.003 | SDC1 (0.865) <- COL3A1 (0.865); 0.0017 |
SDC2 (3.091) <- FN1 (5.594); 0.000125 | |||
|
|||
Neurotrophin signaling (hsa04722) | BAD (1.279) ∣- AKT2 (0.856) <- PDK1 (0.943) <- PIK3CD (0.576) <- GAB1 (0.997) <- SHC2 (0.844) <- NTRK1 (0.945) <- NTF3 (0.784); 0.0008 |
Out of the 16 gene products, CD36 (a class B scavenger receptor) was found in microglia and vascular endothelial cells of AD patient brains [
Another intriguing observation was the opposite expression of a cell growth (antiapoptosis) gene,
In general, single gene expression analysis looks into highly differentially expressed genes under a certain cutoff (e.g.,
Curt Balch is the Chair of Bioscience Advising, IN, USA. This does not alter the result of the study and adherence to the journal publication policy. All the authors declare no potential competing interests.
This work was supported by grants from the National Cancer Center, Republic of Korea (Grant nos. NCC-1210460 and NCC-1510141-1 to Seungyoon Nam), and Korea Institute of Science and Technology Information (KISTI) (Grant no. P-14-SI-IA27 to Seok Jong Yu and Yongseong Cho), Republic of Korea. KISTI kindly provided Seungyoon Nam with a high-performance computing resource.