AD and PD are known as deleterious CNS disorders leading to death usually in elderly people. It has been shown that there are similarities between both diseases. Several studies have compared these diseases using different approaches including graph communities structure (Calderone et al. 2016), proteome analysis (Ping et al. 2018), and cognitive changes (Stern et al. 1993). Here, to gain molecular insights and systems-level similarities between these two diseases, we have used transcriptome datasets produced by cohorts of AD and PD brain samples. The brain cortex was selected as the main region for this comparison because of its affection by both diseases. Although individual WGCNA studies in PD (Chatterjee et al. 2017) and AD (Miller et al. 2008) have been conducted previously, they have not compared these two diseases directly. Our study contained datasets of cohort samples obtained from cortex tissues to reduce dissimilarities in the areas of the brain. Additionally, the availability of detailed functional characteristics and in-depth information regarding neurodegenerative disorders in the cortex makes this organ as a proper tissue for comparison of AD and PD (Lodato and Arlotta 2015; Braak et al. 2006). On the other hand, WGCNA approach is able to find a group of genes based on their pairwise correlations of the expression, relate them to the specific trait or traits, and can detect the most important genes as a hub. Therefore, it only uses expression values and reflects the real interactions among the genes based on the expression data.
Similar biological processes are involved in both AD and PD
Our analysis detected several rigorous modules related to the disease in each set of studies. By annotation of the modules, modules related to brain and synaptic transmission in all analyses were identified, however, different disease-specific processes were also observed. These processes could be related to the particular responses of the regions of the brain to the specific disease. Additionally, common processes in all datasets were also detected (including chemical synaptic transmission, nervous system development, type I interferon signaling pathway/immune responses, and transcription), which are in concurrence with previous studies (Mosley et al. 2012a). Gene expression patterns of these processes are also identical in AD and PD, indicating the existence of striking similarities between these diseases (Fig. 6d). It is well known that synaptic transmission is disrupted by neurodegenerative disorders (Marttinen et al. 2015). On the other hand, nervous system development is considered to be important in late neurodegenerative disorders (Palubinsky et al. 2012). Here, we found that these two processes are embedded in the same modules that represent a relationship between them. Our results indicate that almost all of the DEGs in the modules are down-regulated which is a representative of the suppressive effect of AD and PD on neurotransmission and neuron development in the cortex of the brain. It has also been detected that immune response plays an important role in neurodegenerative disorders (Doty et al. 2015). Accordingly, it was observed that in both diseases genes grouped for immune response are up-regulated, but the ratio of the DEGs to non-DEGs was higher in PD (~17%). Only three DEGs in the AD were identified, including STAT5A, SLC9A9, and PLCE1. Among these genes, STAT5A is known as a critical transcription factor involved in cytokine response and normal immune function (Lin et al. 2012), both are shown to be associated with AD (Doty et al. 2015). While, in PD, multiple immune response factors are involved in the disease including activated microglia, genetic mutations, and oxidative stresses (Doty et al. 2015). This phenomenon could explain the detection of a higher number of DEGs in PD in comparison to the AD. Transcription, as another common process, had an opposite expression pattern, however, the number of related DEGs was very low in the AD. Although alteration of transcription regulation has been previously reported (Tiwari and Pal 2017; Theuns and Van Broeckhoven 2000), the authors suggest that impacts on the transcription regulation should not be considered as a common response in AD and PD.
Hub genes perform roles in both AD and PD
Hubs are highly connected nodes in a network that usually perform a critical role in regulating other nodes. We used a literature review to indicate whether the hub genes are previously reported or they are new candidates for the diseases. To make a reliable and strong estimation, modules with at least five previously-reported hub genes were presented either in AD or PD. Interestingly, all the modules that share similar gene ontology annotation between diseases contained previously reported hub genes, which could be an indicator of the importance of the detected modules in the diseases. Generally, AD sets, either for frontal or temporal cortices, had more reported hub genes compared to PD. However, more DEGs were observed among the hub genes in PD, which shows that PD causes a major alteration in the transcriptome. To less extent, it was also the case for AD-TC but not AD-FC. Differentially expression analysis of the hub genes enriched the results obtained from WGCNA and showed the importance of the hub genes at systems-level to these diseases. Comparing hub genes in both AD and PD proposes the existence of a strong similarity at the molecular levels between these two neurodegenerative diseases. The aim of this study was to show the impact of individual hub genes on the disease, therefore, we have included information regarding gene expression level, protein level, epigenetic alterations, and SNPs polymorphism for these genes in AD compared to the control group. Looking at the common genes among the top 10% hubs, there were two genes, GABRB3, and PAK1, that were also present at the top 10 hubs in the AD-TC (Fig 6d). PAK1 is shown to be associated with AD (Ma et al. 2008a) and PD (Chandrasekaran and Bonchev 2013), however, GABRB3 is only linked to AD (Antonell et al. 2013). Interestingly, both of these genes are involved in synaptic functions (Parker et al. 2011; Nikolić 2008). Notably, these observations suggest that, at least in the main biological processes, some common molecular mechanisms are involved in AD and PD.
Novel candidate central genes in AD and PD
Following the literature review, several new hub genes are detected and their direct relationship with AD or PD has not been reported so far. These genes could be considered as potential candidates for further studies because they are highly connected with the known genes associated with the diseases. To find a potential role for these genes we have investigated the pathways and processes that the genes belong to. This approach clearly unveiled potential mechanisms for the novel hub genes with a possible impact on the diseases. Summary of the results is presented in Table 2.
Table 2. Possible functions of hub genes identified by WGCNA in AD and PD and their relationship with the diseases has not been previously observed. Supporting references explain the possible role of the genes postulated by authors.
Samples
|
Modules
|
Gene symbol
|
Annotated function from Genecard/Uniprot databases
|
Possible association to the disease
|
Supporting reference
|
{Hroudová, 2014 #69}AD-FC
|
darkorange
|
LARGE
|
glycosyltransferase
|
Involvement in glycolisation
|
Jeanneau et al. 2018
|
darkorange2
|
RAP1GDS1
|
GDP/GTP exchange
|
involved in Rho GTPases
|
Aguilar et al. 2017
|
EI24
|
a putative tumor suppressor
|
may involve in apoptosis and autophagy
|
Obulesu and Lakshmi, 2014
|
darkslateblue
|
HLA-J
|
major Histocompatibility Complex
|
immune system
|
Ma et al. 2008
|
HLA-E
|
major Histocompatibility Complex
|
immune system
|
Ma et al. 2008
|
AD-TC
|
darkmagenta
|
HLA-F
|
major Histocompatibility Complex
|
immune system
|
Ma et al. 2008
|
green
|
INPP4A
|
phosphatase
|
by regulating (PI3K) signaling pathway
|
Fedele et al. 2010
|
CREG2
|
oxidoreductase activity and FMN binding
|
transcription corepressor activity
|
Kunita et al. 2002
|
orangered4
|
MLLT3
|
super elongation complex subunit
|
impact of neuron/cortex development
|
Büttner et al. 2010
|
FAM173B
|
mitochondrial N-lysine methyltransferase
|
mitochondrial dysfunction involved in AD
|
Hanneke et al. 2018
|
PD
|
plum1
|
ARFGEF1
|
intracellular vesicular trafficking
|
by guanine-nucleotide exchange on ARF1 and ARF3
|
Stafa et al. 2012
|
CAND1
|
Cullin-RING ubiquitin ligases
|
by involving in ubiquitin signaling pathway
|
Walden and Muqit, 2017
|
DHX36
|
regulating mRNA expression
|
important paralog, DHX57, is involved in aging of cerebellum
|
Lu et al. 2016
|
CPSF6
|
processing of 3' RNA cleavage and polyadenylation
|
Affecting on polyadenylation transcript of α-synuclein
|
Rhinn et al. 2012
|
salmon
|
LAPTM5
|
lysosomal transmembrane receptor
|
by affecting on lysosomal processes
|
Dehay et al. 2010
|
SASH3
|
cell signaling
|
important paralog, SASH1, is involved in PD
|
Zhang et al. 2005
|
DOCK2
|
cytoskeletal rearrangements and immune response
|
impacting immune response and cytoskeleton
|
Mosley et al. 2012; Pellegrini et al. 2017
|
ABI3
|
regulates actin polymerization
|
impacting immune response and cytoskeleton
|
Sims et al. 2017; Pellegrini et al. 2017
|
MYO1F
|
molecular motors
|
impacting immune response/intracellular movement
|
Kim et al. 2006; Pellegrini et al. 2017
|
SYK
|
non-receptor tyrosine kinase
|
innate and adaptive immunity/observed in AD
|
Schweig et al. 2017
|
IKZF1
|
transcription factor
|
involved in neurodegenerative diseases
|
Li et al. 2014
|
darkslateblue
|
OXCT1
|
Oxoacid CoA-Transferase
|
by impacting mitochondrial dysfunctionality in PD
|
Imamura et al. 2006
|
NAPB
|
SNAP protein
|
Involved in other neurodegenerative diseases, AD
|
Yoo et al. 2001
|
yellow
|
SYDE1
|
Rho GTPase-activating protein
|
involved in Rho GTPases
|
Hong and Sklar, 2014
|
NXN
|
redox-dependent negative regulator
|
impacting PD by involving in Wnt signaling
|
Arenas 2014
|
DOCK6
|
guanine nucleotide exchange factor
|
by activation other genes, including RAC1
|
Kim et al. 2018
|
EMP3
|
epithelial membrane protein
|
common in many neurodegenerative diseases
|
Li et al. 2014
|
TAGLN2
|
similar to the transgelin
|
important paralog, transgelin-3, is involved in PD mouse model
|
Triplett et al. 2015
|
We found 10 new hub genes in the AD that are related to metabolism (LARGE, INPP4A, FAM173B), immune response (HLA-J, HLA-E, HLA-F), transcription, and growth regulators (EI24, CREG2, MLLT3), and signaling (RAP1GDS1) (Table 2). LARGE encodes a glycosyltransferase. This gene could be affected by the AD as it has been shown that AD-associated peptidases shed glycosyltransferases (Jeanneau et al. 2018). INPP4A, a phosphatase encoding gene, can be associated with the AD by regulating (PI3K) signaling pathway, an important pathway in AD (Fedele et al. 2010). Neuroprotective roles of this enzyme have been also documented (Sasaki et al. 2010). FAM173B encodes a mitochondrial N-lysine methyltransferase that is involved in chronic pain (Willemen et al. 2018). Chronic pain could be related to mitochondrial dysfunction that is also observed in neurodegenerative diseases including AD (Hroudová et al. 2014). HLAs encode a major histocompatibility complex, proteins involved in immune responses, and is shown to be involved in AD (Ma et al. 2008b). EI24 is responsible for coding a putative tumor suppressor and may perform a role in apoptosis and autophagy during the course of AD (Obulesu and Lakshmi 2014). CREG2 encodes a protein that is involved in oxidoreductase and FMN binding. This gene is expressed in brain and its product may potentially be associated with AD both as a transcription corepressor and signaling molecule (Kunita et al. 2002). Encoded protein by MLLT3 gene is a super elongation complex subunit and has been detected during neurons and cortex development (Büttner et al. 2010). Therefore, it is possible that this gene would have some unknown role in the AD.
On the other hand, in PD, we found a greater number of novel hub genes without any previous link to the disease. The novel hub genes were related to cytoskeleton and vesicle trafficking (ARFGEF1, ABI3, DOCK2, MYO1F, NAPB, TAGLN2), metabolism, and protein degradation (CAND1, SYK, OXCT1, NXN, DOCK6, LAPTM5), transcription regulation (DHX36, CPSF6, IKZF1), and signaling (EMP3, SASH3, SYDE1) (Table 2). ARFGEF1 encodes a protein that activates ARF1/ARF3 by replacement of the GDP with GTP. We could not find any direct report regarding the role of ARF1/ARF3 in PD. However, it is known that ARF1 is an essential regulator of Golgi morphology and the Golgi fragmentation is a feature of many neurodegenerative diseases, including PD (Rendón et al. 2013). Therefore, this could be suggested that ARFGEF1 is involved in PD by affecting ARF1 and consequently the Golgi morphology. ABI3, DOCK2, and MYO1F are involved in cytoskeletal rearrangements and immune responses. These genes are potentially associated with PD by affecting immune response and cytoskeleton (Mosley et al. 2012b; Pellegrini et al. 2017; Sims et al. 2017; Kim et al. 2006). NAPB encodes SNAP protein and is associated with numerous neurodegenerative diseases (Yoo et al. 2001). In the literature, TAGLN2 had no direct relationship with PD, but, its close paralog, TAGLN3, has a role in the mouse model for PD (Triplett et al. 2015). CAND1 encodes a Cullin-RING ubiquitin ligase that could be associated with PD by affecting the ubiquitin signaling pathway (Walden and Muqit 2017). SYK product is a non-receptor tyrosine kinase that is related to AD (Schweig et al. 2017). This gene could probably involve in PD by impacting innate and adaptive immunity. OXCT1 encodes Oxoacid CoA-transferase that catabolizes succinyl CoA and produces succinate. Protective effects of succinate against rotenone, an agent in the generation of PD models (Imamura et al. 2006), may suggest that alteration in OXCT1 expression could be linked to the disease. NXN, encoding redox-dependent negative regulator, could be involved in PD via modulation of the Wnt signaling pathway as previously suggested (Arenas 2014). DOCK6 encodes guanine nucleotide exchange factor and activate molecules such as RAC1 and CDC42, which are involved in actin cytoskeletal organization. DOCK6 may be involved in the functionality of dopaminergic neurons by impacting RAC1 (Kim et al. 2018). LAPTM5 encodes a lysosomal transmembrane receptor. This gene could be related to lysosomal depletion in PD (Dehay et al. 2010). DHX36, CPSF6, and IKZF1 encode proteins involved in transcription regulation. The paralog of DHX36, DHX57, is involved in aging of the cerebellum (Lu et al. 2016). CPSF6 is involved in RNA polyadenylation that could be involved in polyadenylation transcript of α-synuclein in PD (Rhinn et al. 2012). IKZF1 transcription factor has been detected in the meta-analysis of neurodegenerative diseases (Li et al. 2014). EMP3 encodes an epithelial membrane protein, and its upregulation has been observed in neurodegenerative diseases (Li et al. 2014). SASH3 had no direct relationship with neurodegeneration, but its paralog, SASH1, is shown to be involved in PD (Zhang et al. 2005). SYDE1 encodes Rho GTPase-activating protein and may involve in PD by affecting Rho GTPases (Hong and Sklar 2014).
These indirect, but relevant mechanisms for the association of hub genes to AD and PD represent the potency of using these genes as new candidates for therapeutic targets. Additionally, some of these genes are differentially expressed that can be used as markers for the diseases. And finally, the involvement of some of the candidates in other neurodegenerative diseases suggested the presence of a potential common molecular background among many of these disorders.