Background

Non-coding RNA (ncRNA), which could not be translated to protein, can be divided into two categories, One is the long non-coding RNA (lncRNA), the other is small non-coding RNA such as miRNA and microRNA. In the past, lncRNA is considered as genome "dark matter"(Zhao et al., 2014). However, the current studies show that a large number of ncRNA exist in mammals and plays an important role in many biological processes, for example, imprinting control, immune response and cell differentiation (Taft et al., 2010; Mercer et al., 2009; Pang et al., 2006).In humans, the protein-coding genes in the genome sequence is only a small fraction (about 1.5%), more than 98% of the non-coding sequences of genes (Bertone et al., 2004). Mutations and disorders of ncRNAs often cause many human diseases, such as cancer (Mercer et al., 2009), cardiovascular disease (Congrains et al., 2012), neurodegenerative diseases (Johnson, 2012), even, ncRNAs can also be used as molecular markers of treatments of cancer (Bolton et al., 2014). In Alzheimer's disease, lncRNA BACE1AS is the antisense strand RNA of encoding β-secretase enzyme gene, β secretase enzyme can produce β- amyloid, and β- amyloid accumulation is one of the main reasons leading to the occurrence of Alzheimer's disease (Faghihi et al., 2008). Pseudogene PTENP1 is absent in many cancers (eg, colon cancer, prostate cancer), which regulate the expression of tumor suppressor PTEN by complete binding site with miRNA (Poliseno et al., 2010). lncRNA HOTTIP through the interaction with compound WDR5 / MLL to regulate the expression of HOXA locus, which plays an important role in the development, relapse and metastasis of leukemia and liver cancer (Wang et al., 2011; Quagliata et al., 2014).

Ion channels are transmembrane channels of controlling ion passive transport, and involved in various biological processes, for example, nerve cells generate electrical signals, glandular secretion, muscle activity, neurotransmitter release and so on. Voltage-gated anion selective channel (VDAC) regulates metabolic function of mitochondria by control metabolites, which is present in mitochondrial outer membrane and brain postsynaptic membrane of all eukaryotics. Studies show the expression of VDAC in the frontal cortex and thalamus of Alzheimer's disease patients (Yoo et al., 2001).

Ion channels and ncRNAs are involved in the occurrence of cancers and neurodegenerative diseases. It is essential to study relationships between them, but these researches are not sufficient. In this paper, we show a multiple-network which include ncRNAs, ion channels and others coding-gene. We predicted ncRNAs functions, and explore the relationship with function of ion channels and ncRNAs. We hope to deepen the understanding of Alzheimer's disease.

Results

Probe re-annotation of gene expression data sets

After probe re-annotation, there are 133,229 probes mapped to coding-genes and 60,858 probes only mapped to ncRNAs. To reduce noises and improve the accuracy of probe annotation information, we will map transcripts to gene, remove probes matching more than one gene and gene contains less than three probes. Finally, we got 22,018 probe-gene pairs, which contain 10,467 genes and 4,216 ncRNAs.

We use GSE48350 to evaluate the accuracy of the result of the probe re-annotation. GSE48350 contains 80 samples; we calculate the mean of correlation coefficient of expression values of probes belongs to the same gene. We find that after the probe re-annotation, the correlation coefficient of the expression values of the probe of the same gene is significantly increased (t test, p-value <2.2e-16, Figure 1).

Multiple co-expression network analysis

We preprocessed each expression profile dataset and calculated co-expression of each gene pair of each expression profile dataset, respectively. We received a total of 11,413,157 pairs of co-expressed genes. In order to improve the accuracy of multiple co-expression network, we further analyze the co-expression direction(positive, negative) of these gene-pairs in the different expression profile datasets, found that only 541 gene-pairs have same co-expression direction in more than 5 expression profile datasets, and 68,648 gene-pairs have same co-expression direction in more than 3 expression profile datasets(Table 1).

We use GOSemSim package (Yu et al., 2010) of Bioconductor package to calculate the functional similarity of gene-pairs with same co-expression direction. We found the gene-pair have more similar function, if it has same co-expression direction in more expression profiles datasets (Figure 2). Since the co-expression genes may be involved in the same biological processes and have similar functions, in order to ensure that the nodes of co-expression in the network have more similar functions and the size of the network will not be too small, we retained gene-pairs which have same direction in more than 3 expression profile datasets to construct multiple-network.

The multiple-network (Figure 3) contains 4427 nodes, which include 3370 protein-coding genes (ion channel protein gene 397, other protein-coding genes 2973) and 1057 ncRNAs. There are 68,647 edges, and 2480 co-expression relationships between ion channel protein genes and ncRNAs.

We analyzed the degree distribution of nodes in the network, found that it obey the power-law distribution of biomolecular network nodes. Most nodes have small degree value, only small part of nodes has large degree value. There are 2,885 nodes (65%) that their degree is less than 20, more than half of nodes have less than 10 neighbor nodes. It indicates that the multiple-network which construct with our method have high reliability.

Function of ncRNA

Multiple-network includes 18,751 co-expression relationships between ncRNAs and protein-coding genes. There are 369 ncRNAs co-expressed with more than 10 coding-genes. We predicted these 369 ncRNAs function from the function annotation information of neighbor nodes by bioinformatics tool Database for Annotation, Visualization, and Integrated Discovery (DAVID) (Huang et al., 2009). There are 339 ncRNAs enriched to at least one GO term (Table 2, Supplement 1), each ncRNA average enriched to 30 GO terms. We find these ncRNA functions mainly include metabolism, development, expression and regulation, transport, and there are 94 ncRNAs enriched to GO terms which are associated with ion channel, such as ion transport, channel activity, ion transmembrane transport and so on.

Relations between ncRNAs and ion channels

We extract genes co-expressed with potassium channel protein gene, sodium channel protein gene, calcium channel protein gene, chloride channel protein gene from multiple-network (Figure 4). According to the previous results, based on neighbor node predict ncRNA functions, we found that most of functions of ncRNAs co-expressed with ion channel protein gene are associated with function of cell membrane, ion transport. In the multiple-network, there are 19 potassium channel protein genes, and 540 genes co-expressed with them (Figure 4A). In total of 60 ncRNAs that co-expression with potassium channel protein genes, of which 54 ncRNAs enrichment to at least 1 GO term. These GO terms are associated with functions of cell membrane, plasma membrane, ion transport, cation transporter. For example, lncRNA LINC00202-2 enrichment to 72 GO terms which mainly include ion transport, membrane composition, cation transporter, metal ion transport, gated channel activity. There are 410 coding-genes and 96 ncRNAs co-expression with 15 calcium channel genes in multiple-network (Figure 4B). 85 ncRNAs enrichment at least one GO term which are related on fuction of membrane component, ion transport. For example, ncRNA RP4-761J14.8 enrichment to 10 GO terms, these GO term mainly include ion transport, metal ion transport, 2-, 3- monovalent inorganic cation transporter, cation transporter, transition metal cation transporter(Figure 5B). We analyzed sodium channel gene in multiple-network, found 3 sodium channel genes are co-expressed with 160 coding-genes and 49 ncRNAs (Figure 4C). 47 ncRNAs enrichment to at least one GO term. These GO term mainly include transcriptional regulation, DNA binding, membrane component, ion transport and so on (Figure 5C). Moreover, there are 4 chloride channel genes co-expressed with 86 coding-genes and 30 ncRNAs in multiple-network (Figure 4D). These 30 ncRNAs were mainly enrichment to GO term, like ligand-dependent nuclear receptor activity, ion transport, an inorganic anion transporter, ion channel complex. For example, ncRNA AC011298.2 enrichment to GO term, like 2, 3 monovalent inorganic cation transport, embryonic organ development, anion transport (Figure 5D).

Relations between the Alzheimer's disease-related genes and ncRNAs

We obtained 1721 Alzheimer's disease-related genes from GENE Cards database (Safran et al., 2010), 387 genes are contained in our multiple co-expression network, involving in 16,747 co-expression relationships (14,614 relationships between Alzheimer's disease-related genes and coding-genes, 2,133 relationships between Alzheimer's disease-related genes and ncRNAs).

Some genes which are associated with ion channel protein gene also have relationships with the development of Alzheimer's disease, such as, ESR2 mutations will affect the expression levels of potassium ion channel gene, and also increases the risk of Alzheimer's disease (Pirskanen et al., 2005; Oterino et al., 2008). In the multiple-network, ESR2 co-expressed with 25 genes include 18 coding-genes and 7 ncRNAs (Figure 6A). According to predict ncRNA functions, we found lincRNA RP11-700H6.1 is associated with functions of ion transport and ion channel proteins. This further demonstrates the correlation between lncRNA and ion channels.

Voltage-gated anion channel VDAC2 is an Alzheimer's disease-related gene, and co-expressed with 6 ncRNAs (Figure 6B). 5 ncRNAs enrichment to at least one GO term, and these GO term is mainly about mitochondrial function (Figure 7B, 7C, 7D, 7E, 7F). Mitochondrial dysfunction and synaptic damage are the early phenomenon of Alzheimer's disease (Reddy, 2013). Therefore, we speculate that 6 ncRNAs co-expressed with VDAC2 may be involved in the pathogenesis of Alzheimer's disease.

Discussion

Probe annotation information were obtained by probe re-annotation method, and then combined with gene expression profile datasets of Alzheimer's disease, ion channel protein gene, the correlation of gene expression and direction consistency of co-expression in each expression profile dataset to constructed multiple-network. The multiple-network shows ncRNAs and ion channel protein genes have regulatory relations on expression and function, and VDAC2 and 5 ncRNAs may be involved in the pathogenesis of Alzheimer's disease.

ncRNA and ion channel protein genes, which are important biological macromolecules,  involved in many important biological processes and plays an important role in the development of many diseases. It has been found more than ten thousand ncRNA, and the research of relationships between ncRNAs and ion channels will help us to deepen understand the mechanisms of biological processes and diseases. However, most of the studies only predict ncRNA functions, while ignoring the relationship between ncRNAs and ion channels (Zhao et al., 2014). Both ncRNAs and ion channels are participated in neurodegenerative diseases and cancers. Explore relationships between ncRNAs and ion channel protein genes will help us to better understand the pathogenesis of diseases.

Materials and methods

Probe re-annotation

Human probe sequences were obtained from Affymetrix (http://www.affymetrix.com), Human Genome U133 Plus 2.0 Array contains 604,258 probe sequences. We obtained 99,436 human protein-coding gene transcript sequences and 37,197 human non-coding gene transcript sequences from the Ensembl database (Flicek et al., 2011). Taking into account the speed of update genomic data becomes faster and faster, we need to improve the accuracy of microarray probes annotation information (Lu et al., 2007; Zhang et al., 2005). In this work, we use the probe re-annotation method (Liao et al., 2011); remove the ambiguity of the probe, to obtain relatively accurate probe annotation information of coding genes and non-coding genes.

Firstly, we alignment probe sequences with protein-coding genes and non-coding genes use Bowtie (Langmead et al., 2009), separately. Secondly, we removed the probe both alignment protein-coding gene and non-coding gene. Finally, we mapped the protein-coding transcripts and non-coding transcripts to gene, and remained the probes only matched to a gene; removed gene matched less than 3 probes.

Construct multiple-network

7 Alzheimer's disease-related gene expression datasets were downloaded from the Gene Expression Omnibus (GEO) database (Edgar et al., 2002). We preprocessed each gene expression dataset by affy package (Gautier et al., 2004) of Bioconductor package (Gentleman et al., 2004), respectively. Secondly, we mapped probes to genes in each expression dataset, and averaged expression values of same genes. Thirdly, we calculated the variance of gene expression in each expression profile dataset, retained genes with variance at top 75%. Fourthly, we calculated the correlation coefficient of gene expression values of each gene-pair in each expression dataset using the R software WGCNA package (Langfelder and Horvath, 2008), and remained gene-pairs with the person correlation coefficient greater than 0.75 and p-value less than 0.01 Finally, we remained gene-pairs have same co-expression directions at more than 3 expression datasets to construct multiple-network.

Co-expression network had been used to identify cell module and predict function of protein-coding genes (Luo et al., 2007; Sharan et al., 2007; Wren, 2009). Biological processes and cellular regulatory networks are very complex and involve many interactions between molecules (Luo et al., 2007), Co-expression network can analyze correlation of expression of biological molecules and extract the relevant biological processes, and its node represents a biomolecule, edges showing co-expression relationship. In our multiple co-express networks, node contains protein-coding genes, ncRNAs and ion channel protein genes. Genes which have similar expression patterns under different experimental conditions are more likely to have similar functions (Lee et al., 2004) or participate in relevant biological pathways (Eisen et al., 1998). To reduce the noise present in the microarray, to improve the accuracy of multiple co-expression networks, we used 7 gene expression datasets, remained gene-pairs have same co-expression direction in at least 3 expression datasets to construct the multiple co-expression network.

ncRNA function prediction

Hub-based approach is the most direct method of analysis node function, which is enrichment function of functional annotation information of direct neighbor node to predict the function of the node. We used the DAVID bioinformatics tool to analyze functions of neighbor genes of ncRNAs as ncRNA functions. In order to improve the accuracy of prediction ncRNA functions, we only select ncRNAs with more than 10 co-expression coding-gene. We further analysis these ncRNAs function, and explore the relationship between ion channel protein genes and ncRNAs.

References

Bertone P., Stolc V., Royce T.E., Rozowsky J.S., Urban A.E., Zhu X., Rinn J.L., Tongprasit W., Samanta M., and Weissman S., 2004, Global identification of human transcribed sequences with genome tiling arrays, Science, 306: 2242-2246

http://dx.doi.org/10.1126/science.1103388 PMid:15539566

Bolton E.M., Tuzova A.V., Walsh A.L., Lynch T., and Perry A.S., 2014, Noncoding RNAs in prostate cancer: the long and the short of it, Clinical Cancer Research, 20: 35-43

http://dx.doi.org/10.1158/1078-0432.CCR-13-1989 PMid:24146262

Congrains A., Kamide K., Oguro R., Yasuda O., Miyata K., Yamamoto E., Kawai T., Kusunoki H., Yamamoto H., and Takeya Y., 2012, Genetic variants at the 9p21 locus contribute to atherosclerosis through modulation of ANRIL and CDKN2A/B, Atherosclerosis, 220: 449-455

http://dx.doi.org/10.1016/j.atherosclerosis.2011.11.017 PMid:22178423

Edgar R., Domrachev M., and Lash A.E., 2002, Gene Expression Omnibus: NCBI gene expression and hybridization array data repository, Nucleic acids research, 30: 207-210

http://dx.doi.org/10.1093/nar/30.1.207 PMid:11752295 PMCid:PMC99122

Eisen M.B., Spellman P.T., Brown P.O., and Botstein D., 1998, Cluster analysis and display of genome-wide expression patterns, Proceedings of the National Academy of Sciences, 95: 14863-14868

http://dx.doi.org/10.1073/pnas.95.25.14863 PMid:9843981 PMCid:PMC24541

Faghihi M.A., Modarresi F., Khalil A.M., Wood D.E., Sahagan B.G., Morgan T.E., Finch C.E., Laurent Iii G.S., Kenny P.J., and Wahlestedt C., 2008, Expression of a noncoding RNA is elevated in Alzheimer's disease and drives rapid feed-forward regulation of β-secretase, Nature medicine, 14: 723-730

http://dx.doi.org/10.1038/nm1784 PMid:18587408 PMCid:PMC2826895

Flicek P., Amode M.R., Barrell D., Beal K., Brent S., Carvalho-Silva D., Clapham P., Coates G., Fairley S., and Fitzgerald S., 2011, Ensembl 2012, Nucleic acids research: gkr991 PMCid:PMC3245178

Gautier L., Cope L., Bolstad B.M., and Irizarry R.A., 2004, affy—analysis of Affymetrix GeneChip data at the probe level, Bioinformatics, 20: 307-315

http://dx.doi.org/10.1093/bioinformatics/btg405 PMid:14960456

Gentleman R.C., Carey V.J., Bates D.M., Bolstad B., Dettling M., Dudoit S., Ellis B., Gautier L., Ge Y., and Gentry J., 2004, Bioconductor: open software development for computational biology and bioinformatics, Genome biology, 5: 1

http://dx.doi.org/10.1186/gb-2004-5-10-r80 PMid:15461798 PMCid:PMC545600

Huang D.W., Sherman B.T., and Lempicki R.A., 2009, Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources, Nature protocols, 4: 44-57

http://dx.doi.org/10.1038/nprot.2008.211 PMid:19131956

Johnson R., 2012, Long non-coding RNAs in Huntington's disease neurodegeneration, Neurobiology of disease, 46: 245-254

http://dx.doi.org/10.1016/j.nbd.2011.12.006 PMid:22202438

Langfelder P., and Horvath S., 2008, WGCNA: an R package for weighted correlation network analysis, BMC bioinformatics, 9: 1

http://dx.doi.org/10.1186/1471-2105-9-559 PMid:19114008 PMCid:PMC2631488

Langmead B., Trapnell C., Pop M., and Salzberg S.L., 2009, Ultrafast and memory-efficient alignment of short DNA sequences to the human genome, Genome biology, 10: 1

http://dx.doi.org/10.1186/gb-2009-10-3-r25 PMid:19261174 PMCid:PMC2690996

Lee H.K., Hsu A.K., Sajdak J., Qin J., and Pavlidis P., 2004, Coexpression analysis of human genes across many microarray data sets, Genome research, 14: 1085-1094

http://dx.doi.org/10.1101/gr.1910904 PMid:15173114 PMCid:PMC419787

Liao Q., Liu C., Yuan X., Kang S., Miao R., Xiao H., Zhao G., Luo H., Bu D., and Zhao H., 2011, Large-scale prediction of long non-coding RNA functions in a coding–non-coding gene co-expression network, Nucleic acids research, 39: 3864-3878

http://dx.doi.org/10.1093/nar/gkq1348 PMid:21247874 PMCid:PMC3089475

Lu J., Lee J.C., Salit M.L., and Cam M.C., 2007, Transcript-based redefinition of grouped oligonucleotide probe sets using AceView: high-resolution annotation for microarrays, BMC bioinformatics, 8: 1

http://dx.doi.org/10.1186/1471-2105-8-1

http://dx.doi.org/10.1186/1471-2105-8-108

PMid:17394657 PMCid:PMC1853115

Luo F., Yang Y., Zhong J., Gao H., Khan L., Thompson D.K., and Zhou J., 2007, Constructing gene co-expression networks and predicting functions of unknown genes by random matrix theory, BMC bioinformatics, 8: 1

http://dx.doi.org/10.1186/1471-2105-8-1

http://dx.doi.org/10.1186/1471-2105-8-299

PMid:17697349 PMCid:PMC2212665

Mercer T.R., Dinger M.E., and Mattick J.S., 2009, Long non-coding RNAs: insights into functions, Nature Reviews Genetics, 10: 155-159

http://dx.doi.org/10.1038/nrg2521 PMid:19188922

Oterino A., Toriello M., Cayón A., Castillo J., Colas R., Alonson‐Arranz A., Ruiz‐Alegria C., Quintela E., Monton F., and Ruiz‐Lavilla N., 2008, Multilocus analyses reveal involvement of the ESR1, ESR2, and FSHR genes in migraine, Headache: The Journal of Head and Face Pain, 48: 1438-1450

http://dx.doi.org/10.1111/j.1526-4610.2008.01294.x

Pang K.C., Frith M.C., and Mattick J.S., 2006, Rapid evolution of noncoding RNAs: lack of conservation does not mean lack of function, Trends in Genetics, 22: 1-5

http://dx.doi.org/10.1016/j.tig.2005.10.003 PMid:16290135

Pirskanen M., Hiltunen M., Mannermaa A., Helisalmi S., Lehtovirta M., Hänninen T., and Soininen H., 2005, Estrogen receptor beta gene variants are associated with increased risk of Alzheimer's disease in women, European journal of human genetics, 13: 1000-1006

http://dx.doi.org/10.1038/sj.ejhg.5201447 PMid:15944651

Poliseno L., Salmena L., Zhang J., Carver B., Haveman W.J., and Pandolfi P.P., 2010, A coding-independent function of gene and pseudogene mRNAs regulates tumour biology, Nature, 465: 1033-1038

http://dx.doi.org/10.1038/nature09144 PMid:20577206 PMCid:PMC3206313

Quagliata L., Matter M.S., Piscuoglio S., Arabi L., Ruiz C., Procino A., Kovac M., Moretti F., Makowska Z., and Boldanova T., 2014, Long noncoding RNA HOTTIP/HOXA13 expression is associated with disease progression and predicts outcome in hepatocellular carcinoma patients, Hepatology, 59: 911-923

http://dx.doi.org/10.1002/hep.26740 PMid:24114970 PMCid:PMC3943759

Reddy P.H., 2013, Is the mitochondrial outermembrane protein VDAC1 therapeutic target for Alzheimer's disease?, Biochimica et Biophysica Acta (BBA)-Molecular Basis of Disease, 1832: 67-75

http://dx.doi.org/10.1016/j.bbadis.2012.09.003 PMid:22995655 PMCid:PMC3518645

Safran M., Dalah I., Alexander J., Rosen N., Stein T.I., Shmoish M., Nativ N., Bahir I., Doniger T., and Krug H., 2010, GeneCards Version 3: the human gene integrator, Database, 2010: baq020

http://dx.doi.org/10.1093/database/baq020 PMid:20689021 PMCid:PMC2938269

Sharan R., Ulitsky I., and Shamir R., 2007, Network‐based prediction of protein function, Molecular systems biology, 3: 88

http://dx.doi.org/10.1038/msb4100129 PMid:17353930 PMCid:PMC1847944

Taft R.J., Pang K.C., Mercer T.R., Dinger M., and Mattick J.S., 2010, Non‐coding RNAs: regulators of disease, The Journal of pathology, 220: 126-139

http://dx.doi.org/10.1002/path.2638 PMid:19882673

Wang K.C., Yang Y.W., Liu B., Sanyal A., Corces-Zimmerman R., Chen Y., Lajoie B.R., Protacio A., Flynn R.A., and Gupta R.A., 2011, A long noncoding RNA maintains active chromatin to coordinate homeotic gene expression, Nature, 472: 120-124

http://dx.doi.org/10.1038/nature09819 PMid:21423168 PMCid:PMC3670758

Wren J.D., 2009, A global meta-analysis of microarray expression data to predict unknown gene functions and estimate the literature-data divide, Bioinformatics, 25: 1694-1701

http://dx.doi.org/10.1093/bioinformatics/btp290 PMid:19447786 PMCid:PMC2732319

Yoo B.C., Fountoulakis M., Cairns N., and Lubec G., 2001, Changes of voltage‐dependent anion‐selective channel proteins VDAC1 and VDAC2 brain levels in patients with Alzheimer's disease and Down Syndrome, Electrophoresis, 22: 172-179

http://dx.doi.org/10.1002/1522-2683(200101)22:1<172::AID-ELPS172>3.0.CO;2-P

Yu G., Li F., Qin Y., Bo X., Wu Y., and Wang S., 2010, GOSemSim: an R package for measuring semantic similarity among GO terms and gene products, Bioinformatics, 26: 976-978

http://dx.doi.org/10.1093/bioinformatics/btq064 PMid:20179076

Zhang J., Finney R.P., Clifford R.J., Derr L.K., and Buetow K.H., 2005, Detecting false expression signals in high-density oligonucleotide arrays by an in silico approach, Genomics, 85: 297-308

http://dx.doi.org/10.1016/j.ygeno.2004.11.004 PMid:15718097

Zhao Y., Luo H., Chen X., Xiao Y., and Chen R., 2014, Computational methods to predict long noncoding RNA functions based on co-expression network, RNA Mapping: Methods and Protocols: 209-218