- Email: [email protected]
Comparative analysis of RNA-Seq data from brain and blood samples of Parkinson's disease
Biochemical and Biophysical Research Communications 484 (2017) 557e564
Contents lists available at ScienceDirect
Biochemical and Biophysical Research Communications journal homepage: www.elsevier.com/locate/ybbrc
Comparative analysis of RNA-Seq data from brain and blood samples of Parkinson's disease Paulami Chatterjee, Debjani Roy* Department of Biophysics, Bose Institute, Acharya J.C. Bose Centenary Building, P-1/12 C.I.T Scheme VII M, Kolkata, 700054 West Bengal, India
a r t i c l e i n f o
a b s t r a c t
Article history: Received 18 January 2017 Accepted 23 January 2017 Available online 26 January 2017
Parkinson's disease (PD) is the second most common neurodegenerative disorders throughout the world. In order to search for PD biomarkers, we performed a system-level study of RNA-Seq data from PD brain and blood samples. Differentially expressed miRs of RNA-Seq data were subjected to generate the Coexpression networks. Three highly co-expressed clusters were identified based on their correlation coefficient values and fold change ratio. SM2miR drugs of the miRs contained in the three highly coexpressed clusters were identified, and drugs common among these clusters were selected. Coexpressed miRs not previously known to be associated with PD were identified from both the samples. Functional enrichment analyses of these miR targets were done, and the pathways common and unique to both the samples were identified. Thus, our study presents a comparative analysis of miRs, their associated pathways, and drugs from brain and blood samples of PD that may help in system level understanding of this disease. miRs identified from our study may serve as biomarkers for PD. © 2017 Elsevier Inc. All rights reserved.
Keywords: Parkinson's disease RNA-Seq data miR Co-expression network Pathway analysis SM2miR drugs
1. Introduction Parkinson's disease (PD) is the second most common neurodegenerative disorders throughout the world. It is one of the chronic and progressive movement disorders in which symptoms get worse over time. Deregulation of microRNAs (miRs) has been implicated as one of the major causative factors in neurodegenerative diseases [1], including PD [2e4]. The previous study has pointed out that few miRs can be predominantly expressed in certain tissues, whereas the majority of miRs (82.9%) are neither specific for single tissues nor housekeeping miRs [5]. Therefore, knowledge of the expression pattern and distribution of different miRs in tissues is essential for the understanding of the physiological and pathological mechanisms associated with a disease [5]. Analyses of genomewide small RNA transcriptome provide an accurate and comprehensive knowledge of miR expression in the disease under different conditions [6]. RNA sequencing (RNA-Seq) which uses Next Generation Abbreviations: PD, Parkinson's Disease; miR, MicroRNA; RNA-Seq, RNA sequencing; DE, Differentially Expressed; Hclust, Hierarchical Clustering; ATC, Anatomic Therapeutic Classification. * Corresponding author. Department of Biophysics, Bose Institute, Acharya J.C. Bose Centenary Building, P-1/12 C.I.T Scheme VII M, Kolkata, 700054, India. E-mail addresses: [email protected] (P. Chatterjee), [email protected] (D. Roy). http://dx.doi.org/10.1016/j.bbrc.2017.01.121 0006-291X/© 2017 Elsevier Inc. All rights reserved.
Sequencing technology to quantify and discover RNAs in a cellular transcriptome [7,8], has been widely used for investigating the miRmediated pathogenesis of complex diseases. The high-resolution, genome-wide miR profile of PD obtained from RNA-seq offer insight into the molecular and pathological mechanisms that occur in the disease [9]. RNA-Seq mediated comparative blood transcriptome analysis of idiopathic and LRRK2 mutated PD has resulted in the identification of the significant difference in peripheral blood transcriptome [10]. Long non-coding RNAs and alternative splicing modulations have been identified from PD blood cells by RNA-Seq method [11]. However, the comparative analysis of miR profiles of RNA-Seq data from brain and blood samples has not been performed yet for PD. In this computational study, starting from PD brain and blood RNA-Seq data, we have identified disease-specific highly coexpressed miRs. Of these co-expressed miRs, we screened the miRs that are not previously known to be associated with PD. In this work, we studied highly co-expressed miRs, their associated pathways, and drugs from brain and blood samples that may help in system level understanding of this disease. The miRs identified from our study may serve as biomarkers for PD. 2. Methodology Fig. 1 depicts the workflow of our methodology.
558
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
2.1. Data collection RNA-Seq data of brain and blood samples of PD were retrieved from Gene Expression Omnibus (GEO) [Accession IDs- GSE72962 and GSE40915]. GSE72962 (Data 1) contains 29 PD and 33 neuropathologically-normal control from post-mortem prefrontal cortex Brodmann Area 9 (BA9) [9]. GSE40915 (Data 2) contains 7 PD patients before and after Deep Brain Simulation (DBS) treatment (which improves some of the PD motor symptoms) and 6 agematched healthy control samples from blood leukocytes [6]. 2.2. Differentially expressed miR selection In order to identify the differentially expressed (DE) miRs over
disease and normal conditions, paired two sample t-test was applied on each of the expression datasets. The t-test measures the statistical significance of the dataset in terms of test statistic t, which is given by:
xy t ¼ qffiffiffiffiffiffiffiffiffiffiffiffi2ffi sy s2x n þm Where x and y are the sample means, sx and sy are the sample standard deviations, n and m are the sample sizes for two samples, x andy. Under the null hypothesis, this test returns the probability (p-value) of observing a value as extreme or more extreme of the
Fig. 1. Methodology for the analysis of RNA-Seq data from brain and blood samples of PD.
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
test statistic. In our study, miRs with p-value <0.05 (95% confidence level) were selected as DE miRs in PD. 2.3. Agglomerative hierarchical clustering of the DE miRs In order to arrange the DE miRs into groups of similar expression patterns, we employed hierarchical clustering. Hierarchical clustering is a method that groups objects into clusters depending on their distance. It works on the idea that objects that are close to each other are more connected than to objects that are present at a distance. In agglomerative hierarchical clustering, each object is considered as a cluster; hence it starts with a single object and aggregates nearby objects into
559
clusters [12]. For the calculation of the distance between objects in the data matrix, we used the Euclidean distance and average linkage which uses the average distance between all pairs of objects in any two clusters. Given a mx-by-n data matrix X where the row vectors are x1, x2, …, xmx, and my-by-n data matrix Y where the row vectors are y1, y2, …., ymy, the Euclidean distance between the vectors xs and yt are defined as d2st ¼ ðxs yt Þðxs yt Þ0 . 2.4. Fold change calculation Fold change is a measure describing the extent of changes in expression of a gene in disease condition as compared to control.
Fig. 2. Co-expression network containing highly positively correlated miR pairs (r > 0.8). (2a) Co-expression network for PD Brain sample, (2b) Co-expression network for PD Blood sample. Here nodes represent highly correlated miRs and edges between miRs represent the significant co-expression relationship. Node size is mapped to the degree of each node in the network.
560
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
Fold change values were calculated for each of the DE miRs using MATLAB.
Fold Change ¼ log2
ED EC
Where ED and EC represent the arithmetic mean of the expression values of a particular miR over disease conditions and control conditions respectively. Depending on the fold change values upregulated and down-regulated DE miRs were identified from each dataset. 2.5. Co-expression network construction Co-expression Network was constructed for each of the DE miR sets based on the Pearson Correlation coefficient (r) which computes all possible combination of pairs of DE miRs over control and disease conditions. The basic formula for computing r is
r¼
P XX Y Y nSx Sy
Where, X and Y are the scores of the variables whose Correlation Coefficient are being measured (here X and Y represent the expression values of two miRs), X and Y are their respective means and Sx and SY are the respective standard deviations, and n is the number of individuals or pairs of scores in the sample [12]. The Correlation Coefficient calculation was performed using MATLAB. Highly positively correlated DE miRs (r > 0.8) and highly negatively correlated DE miRs (r 0.8) were selected for this purpose. Visualization of the network was performed in Cytoscape [13] and the topological significance analysis was performed using Network Analyzer module of Cytoscape software [13]. 2.6. Identification of drugs and target mRNAs associated with DE miRs FDA approved drugs associated with the DE miRs were identified from SM2miR database [14]. Experimentally validated mRNA targets associated with the miRs were identified from DIANA TOOLS-TarBase v7.0 [15]. Functional enrichment analysis of these target mRNAs was performed in DAVID Bioinformatics Resources 6.8 [16,17]. Information on the experimentally validated diseaserelated genes of PD was obtained from PDbase database [18]. 3. Results 3.1. Identification of DE miRs Our study identified 321 miRs (p-value <0.05) differentially expressed in PD brain samples and 137 miRs (p-value<0.05) differentially expressed in PD blood samples. Interestingly, we found 56 miRs common between these two samples (Supplementary Table 1). We processed these two DE miR datasets using the same pipeline and subjected them to further analyses. 3.2. Hierarchical clustering analysis Hierarchical Clustering (Hclust) arranged the DE miRs into groups of miRs based on their similar expression patterns. The 321 DE miRs of Data 1 (Brain sample) were divided into five clusters consisting of 88, 110, 1, 25, and 97 miRs respectively. The 137 DE miRs of Data 2 (Blood samples) were also divided into five clusters consisting of 127, 7, 1, 1, and 1 miRs respectively. To ascertain the
function of these miRs, we studied the miR associated pathways of significant miR clusters in mirPath v.3 software of DIANA TOOLS. Most of the clusters from brain and blood samples contain several PD-related pathways (p-value <0.05) e.g. Axon guidance, Prion diseases, Ubiquitin mediated proteolysis, Endocytosis, etc. 3.3. Co-expression network construction Two Co-expression Networks were constructed from each of the DE miR sets of brain and blood samples. We selected highly positively correlated miR pairs (r > 0.8) (Fig. 2), and highly negatively correlated miR pairs (r 0.8) from both the samples (Fig. 3). We named highly positively correlated miR pairs as Co-expression Network 1 and highly negatively correlated miR pairs as Coexpression Network 2 for both the datasets. Thus from PD brain samples 70 DE miR pairs with r > 0.8 and 22 DE miR pairs with r 0.8 were obtained. From PD blood samples 394 DE miR pairs with r > 0.8 and 8 DE miR pairs with r 0.8 were obtained. Supplementary Table 2 and Supplementary Table 3 show the degree distribution pattern of all these highly co-expressed miRs in both the samples. 3.4. Clustering of the co-expressed miRs The highly co-expressed miRs obtained from both the samples were clustered based on their fold change value and co-expression correlation coefficient. Thus, three clusters (CoexpCluster1, CoexpCluster 2 and CoexpCluster 3) were obtained from both the samples. CoexpCluster 1 contains positively correlated upregulated miR pairs (where expression of both the miRs increase in disease sample compared to control), CoexpCluster 2 contains positively correlated down-regulated miR pairs (where expression of both the miRs decrease in disease sample compared to control), and CoexpCluster 3 contains Negatively correlated miR pairs (where expression of one miR increases and other one decreases in disease sample compared to control). It was found that CoexpCluster1 contained 47 miR pairs, CoexpCluster 2 contained 23 miR pairs, and CoexpCluster 3 contained 22 miR pairs in PD brain samples. It was found that CoexpCluster1 contained 12 miR pairs, CoexpCluster 2 contained 359 miR pairs, and CoexpCluster 3 contained 8 miR pairs in PD blood samples. These miR clusters were then subjected to further studies. 3.5. Identification of drugs associated with the miR clusters In order to study the drug-mediated regulation of miR expression, SM2miR drugs associated with the miRs of each cluster were identified. It was found that 41, 20 and 25 drugs were associated with the miRs of CoexpCluster 1, CoexpCluster 2, and CoexpCluster 3 respectively in Data 1, whereas, 30, 42, and 26 drugs were associated with the miRs of CoexpCluster 1, CoexpCluster 2, and CoexpCluster 3 respectively in Data 2. Interestingly, fourteen drugs in brain sample and six drugs in blood sample were common among these three miR clusters (Table 1). We focused on these common drugs of each sample and studied their associated miRs. Unknown miRs (which were not previously known to be involved in PD) associated with these common drugs were selected from each dataset. In this way, 41 unknown miRs from Data 1, and 30 unknown miRs from Data 2 were selected. The mRNA targets of these unknown miRs were studied from each dataset. 2229 experimentally validated mRNA targets (associated with 41 unknown miRs) were obtained from Data 1 and 1927 experimentally validated mRNA targets (associated with the 30 unknown miRs) were obtained from Data 2.
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
561
Fig. 3. Co-expression network containing highly negatively correlated miR pairs (r -0.8): (3a) Co-expression network for PD Brain sample, (3b) Co-expression network for PD Blood sample. Here nodes represent highly correlated miRs and edges between miRs represent the significant co-expression relationship. Node size is mapped to the degree of each node in the network.
Table 1 SM2miR drugs common among three miR clusters in brain and blood samples of PD.a SM2miR Drugs common among three clusters PD brain samples
PD blood samples
Drug ID
DrugName
ATC
DrugID
DrugName
ATC
DB01262 DB00544 DB01169 DB00169 DB00441 DB02546 DB00126 DB00624 DB00136 DB00014 DB00255 DB00396 DB00467 DB00295
Decitabine Fluorouracil Arsenic trioxide Cholecalciferol Gemcitabine Vorinostat Vitamin C Testosterone Calcitriol Goserelin Diethylstilbestrol Progesterone Enoxacin Morphine
Antineoplastic agents Antineoplastic agents Antineoplastic agents Musculo-skeletal system Antineoplastic agents Antineoplastic agents Not Available Genito-urinary system and sex hormone Alimentary tract and metabolism, Dermatologicals Antineoplastic agents Genito-urinary system and sex hormone Genito-urinary system and sex hormone Antiinfectives for systemic use Nervous System
DB00783 DB00467 DB00441 DB00544 DB01262 DB00255
Estradiol Enoxacin Gemcitabine Fluorouracil Decitabine Diethylstilbestrol
Genito-urinary system and sex hormone Antiinfectives for systemic use Antineoplastic agents Antineoplastic agents Antineoplastic agents Genito-urinary system and sex hormone
a
Drugs common between two samples are shown in bold.
562
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
Table 2 Most significant over-representative KEGG pathways (p-value <0.05) associated with the miRs that are not previously known to be associated with PD identified from brain and blood samples. Common KEGG pathways associated with unknown miRs of both the PD brain and blood samples
Significant KEGG pathways uniquely associated with the PD brain sample
Significant KEGG pathways uniquely associated with the PD blood sample
hsa04110:Cell cycle
hsa04550:Signaling pathways regulating pluripotency of stem cells hsa04141:Protein processing in endoplasmic reticulum hsa04068:FoxO signaling pathway hsa04390:Hippo signaling pathway hsa05169:Epstein-Barr virus infection hsa05161:Hepatitis B hsa04151:PI3K-Akt signaling pathway hsa04710:Circadian rhythm hsa04152:AMPK signaling pathway hsa04540:Gap junction hsa04142:Lysosome hsa05120:Epithelial cell signaling in Helicobacter pylori infection hsa04150:mTOR signaling pathway hsa04730:Long-term depression hsa04728:Dopaminergic synapse hsa04320:Dorso-ventral axis formation hsa05221:Acute myeloid leukemia hsa05164:Influenza A hsa04668:TNF signaling pathway
hsa04512:ECM-receptor interaction
hsa04360:Axon guidance hsa04350:TGF-beta signaling pathway hsa04120:Ubiquitin mediated proteolysis hsa04722:Neurotrophin signaling pathway hsa04144:Endocytosis hsa05214:Glioma hsa04330:Notch signaling pathway hsa05200:Pathways in cancer hsa04510:Focal adhesion hsa04010:MAPK signaling pathway hsa05220:Chronic myeloid leukemia hsa04115:p53 signaling pathway hsa04520:Adherens junction hsa05223:Non-small cell lung cancer hsa04910:Insulin signaling pathway hsa04720:Long-term potentiation hsa04810:Regulation of actin cytoskeleton hsa04310:Wnt signaling pathway
3.6. Functional enrichment analysis Functional enrichment analyses were performed with the unknown miR targets (2229 and 1927 mRNA targets respectively) of both the samples. It was found that several disease-specific pathways were associated with the unknown miRs of the two datasets. We identified pathways common between brain and blood samples and pathways uniquely associated with each of the sample (Table 2).
4. Discussion The substantia nigra in the midbrain is the most affected brain region in PD but the analysis of this region from post-mortem PD brains may only highlight miRs associated with changes in cellular composition [9]. On the other hand, analysis of post-mortem prefrontal cortex, (a region which contains both dopaminergic neuron projections and pathological hallmarks [19,20], may lead to the identification of miRs associated with disease progression [9]. However, post-mortem brain tissue is not readily available, and entails RNA quality concerns [21]. There comes the importance of studying blood samples in PD. Blood leukocytes are readily obtained, and the RNA can be obtained at high quality from them [6]. Interestingly, the concept of blood-based diagnostic markers is a highly emerging field of study nowadays [22,23]. With this notion, in this computational study, we have selected RNA-Seq PD data obtained from prefrontal cortex and blood leukocytes, which provide the global miR profiling of PD. Co-expression network identified the highly correlated miR pairs that are differentially expressed in PD brain and blood samples. Differential fold change and co-expression patterns of miRs are evident in these three discernible clusters (CoexpCluster 1, CoexpCluster 2 and CoexpCluster 3). These highly correlated miR pairs may undergo a similar type of regulation pattern in PD. We searched SM2miR database to identify drugs that are involved in each of these miR cluster. Common drugs among these three clusters were identified (Table 1). Several drugs are found to be shared between two samples. Table 1 shows the detailed
hsa04912:GnRH signaling pathway hsa04916:Melanogenesis hsa00600:Sphingolipid metabolism
information on Anatomic Therapeutic Classification (ATC) of these drugs. These correlated miRs and their associated drugs may provide a detailed overview of the probable therapeutic targets of this disease. Common drug associated miRs that are not previously known to be involved in PD were selected from each sample (Fig. 1). In this way, 41 unknown miRs from Data 1, and 30 unknown miRs from Data 2 were selected. Unknown miRs uniquely associated with each
Table 3 miRs that are not previously known to be associated with PD identified from brain and blood samples. miRs from brain
miRs from blood
hsa-miR-124-5p hsa-miR-127-3p hsa-miR-1277-5p hsa-miR-129-2-3p hsa-miR-134-5p hsa-miR-148b-3p hsa-miR-148b-5p hsa-miR-16-2-3p hsa-miR-181c-5p hsa-miR-191-5p hsa-miR-20a-5p hsa-miR-212-3p hsa-miR-212-5p hsa-miR-339-5p hsa-miR-340-3p hsa-miR-342-3p hsa-miR-345-5p hsa-miR-34c-3p hsa-miR-372-3p hsa-miR-373-3p hsa-miR-381-3p hsa-miR-382-5p hsa-miR-411-5p hsa-miR-432-5p hsa-miR-491-5p hsa-miR-516a-5p hsa-miR-522-3p hsa-miR-628-5p hsa-miR-769-5p
hsa-miR-107 hsa-miR-1286 hsa-miR-130b-5p hsa-miR-181d hsa-miR-183-5p hsa-miR-18a-5p hsa-miR-1910 hsa-miR-2116 hsa-miR-223 hsa-miR-299 hsa-miR-29b-2-5p hsa-miR-30e hsa-miR-3133 hsa-miR-320d-1 hsa-miR-320d-2 hsa-miR-328 hsa-miR-4306 hsa-miR-486 hsa-miR-641 hsa-miR-98
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
563
Table 4 Detailed information of miRs that are not previously known to be associated with PD and common between brain and blood samples. Common unknown miRs
hsa-let-7d-5p hsa-let-7f-5p hsa-miR-106b-5p hsa-miR-15b-5p hsa-miR-181a-5p hsa-miR-324-5p hsa-miR-454-3p hsa-miR-671-5p hsa-miR-92a-3p hsa-miR-92b-3p hsa-miR-93-5p
Fold change percentage
mRNA Targets associated with the miRs
Hierarchical cluster number
Brain sample
Blood sample
Total targets from TarBase
Targets known in PD
Brain sample
Blood sample
2.94 1.59 3.53 8.36 2.15 3.4 4.47 9.9 2.15 4.33 2.35
120.15 53.23 108.79 110.74 70.66 66.59 110.12 153.37 177.76 102.7 63.67
39 41 97 359 27 58 276 200 33 337 23
7 4 20 62 9 11 61 42 6 55 4
4 4 5 5 3 5 5 2 4 4 5
1 1 1 1 1 1 1 1 1 1 1
sample were identified (Table 3). Additionally, common miRs between two samples were identified. (Table 4). Fold change values of these common unknown miRs indicated that these miRs show differential expression pattern between PD brain and blood samples. Most of these miRs were up-regulated in PD brain samples whereas they were mostly down-regulated in PD blood samples (Table 4). Co-expression network indicated that these common miRs are co-expressed with already known PD miRs. Moreover, hierarchical clustering also illustrated that these common miRs have similar expression pattern with already known PD miRs since they were clustered with known PD miRs. It was found that most of these common unknown miRs were clustered into cluster 4 and 5 of Data 1 (Table 4) with several other miRs already known to be involved in PD. In data 2, all of these common unknown miRs were clustered into cluster 1 (Table 4) which also contains several known miRs in PD. Further study on the targets of these common unknown miRs revealed that all of these miRs are associated with genes already known in PD (Table 4). Functional enrichment analysis of the targets of unknown miRs identified disease-specific significant pathways associated with the unknown miRs of the two datasets. Interestingly, we found several over-representative pathways shared between the two datasets (Table 2). We also identified the over-representative pathways uniquely associated with each dataset (Table 2). Our study performed a comparative analysis of expression profiles of miRs obtained from two different tissues (brain and blood) and identified disease-specific miRs and shed light on their expression pattern, drug-mediated regulation and significant pathways associated with them. The highly correlated differentially expressed miRs and associated over-representative pathways offer insight into the significant molecular mechanisms that occur in the disease. 5. Conclusion In this computational study, starting from PD brain and blood RNA-Seq data, we have identified disease-specific miRs that are shared between two samples. These common miRs have also been found to be targeting known PD genes suggesting their significance in PD etiology. We have also identified the unique miRs from both the samples that are not previously known to be associated with PD and shed light on their expression pattern, drug-mediated regulation and significant pathways associated with them. The Coexpression networks consisting of the differentially expressed miRs helped us to identify the highly correlated miR pairs from each sample. The highly correlated differentially expressed miRs and associated over-representative pathways offer insight into the significant molecular mechanisms that occur in the disease. To our knowledge, this is the first study that provides a comparative
analysis of brain and blood samples of PD RNA-Seq data and illustrated miR disease markers, drugs and significant pathways associated with this disease. This systems level study may open up the opportunities for future PD therapeutic developments. Conflict of interest The authors declare that they do not have any potential conflict of interest. Acknowledgement The authors would like to thank the Department of Biophysics, Bose Institute. This work is not funded by any external funding agency. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.bbrc.2017.01.121. References [1] A.W. Chan, J. Kocerha, The path to microRNA therapeutics in psychiatric and neurodegenerative disorders, Front. Genet. 3 (2012) 82. [2] J. Kim, K. Inoue, J. Ishii, W.B. Vanti, S.V. Voronov, E. Murchison, et al., A MicroRNA feedback circuit in midbrain dopamine neurons, Science 317 (2007) 1220e1224. [3] S. Gehrke, Y. Imai, N. Sokol, B. Lu, Pathogenic LRRK2 negatively regulates microRNA-mediated translational repression, Nature 466 (2010) 637e641. ~ ones-Moyano, S. Porta, G. Escaramís, R. Rabionet, S. Iraola, [4] E. Min B. Kagerbauer, et al., MicroRNA profiling of Parkinson's disease brains identifies early downregulation of miR-34b/c which modulate mitochondrial function, Hum. Mol. Genet. 20 (2011) 3067e3078. [5] N. Ludwig, P. Leidinger, K. Becker, C. Backes, T. Fehlmann, C. Pallasch, et al., Distribution of miRNA expression across human tissues, Nucleic Acids Res. 44 (2016) 3865e3877. [6] L. Soreq, N. Salomonis, M. Bronstein, D.S. Greenberg, Z. Israel, H. Bergman, et al., Small RNA sequencing-microarraay analyses in Parkinson leukocytes reveal deep brain stimulation-induced splicing changes that classify brain region transcriptomes, Front. Mol. Neurosci. 6 (2013) 10. [7] Z. Wang, M. Gerstein, M. Snyder, RNA-Seq: a revolutionary tool for transcriptomics, Nat. Rev. Genet. 10 (2009) 57e63. [8] C.C. Pritchard, H.H. Cheng, M. Tewari, MicroRNA profiling: approaches and considerations, Nat. Rev. Genet. 13 (2012) 358e369. [9] A.G. Hoss, A. Labadorf, T.G. Beach, J.C. Latourelle, R.H. Myers, microRNA profiles in Parkinson's disease prefrontal cortex, Front. Aging Neurosci. 8 (2016) 36. [10] J. Infante, C. Prieto, M. Sierra, P. S anchez-Juan, I. Gonz alez-Aramburu, nchez-Quintana, et al., Comparative blood transcriptome analysis in C. Sa idiopathic and LRRK2 G2019S-associated Parkinson's disease, Neurobiol. Aging. 38 (214) (2016) e1e5. [11] L. Soreq, A. Guffanti, N. Salomonis, A. Simchovitz, Z. Israel, H. Bergman, et al., Long non-coding RNA and alternative splicing modulations in Parkinson's leukocytes identified by RNA sequencing, PLoS Comput. Biol. 10 (2014) e1003517. [12] P. Chatterjee, M. Bhattacharyya, S. Bandyo padhyay, D. Roy, Studying the
564
[13]
[14]
[15]
[16]
[17]
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564 system- level involvement of microRNAs in Parkinson's disease, PLoS One 9 (2014) e93751. P. Shannon, A. Markiel, O. Ozier, N.S. Baliga, J.T. Wang, et al., Cytoscape: a software environment for integrated models of biomolecular interaction networks, Genome Res. 13 (2003) 2498e2504. X. Liu, S. Wang, F. Meng, J. Wang, Y. Zhang, E. Dai, et al., SM2miR:a database of the experimentally validated small molecules' effects on microRNA expression, Bioinformatics 29 (2013) 409e411. I.S. Vlachos, M.D. Paraskevopoulou, D. Karagkouni, G. Georgakilas, T. Vergoulis, I. Kanellos, et al., DIANA-TarBase v7.0: indexing more than half a million experimentally supported miRNA:miRNA interactions, Nucleic Acid. Res. 43 (2015) D153eD159. D.W. Huang, B.T. Sherman, R.A. Lempicki, Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists, Nucleic Acids Res. 37 (2009) 1e13. D.W. Huang, B.T. Sherman, R.A. Lempicki, Systematic and integrative analysis of large gene lists using DAVID Bioinformatics Resources, Nat. Protoc. 4 (2009) 44e57.
[18] J.O. Yang, W.Y. Kim, S.Y. Jeong, J.H. Oh, S. Jho, J. Bhak, et al., PDbase: a database of Parkinson's disease-related genes and genetic variations using substantia nigra ESTs, BMC Genomics 10 (2009) S32. [19] T.G. Beach, C.H. Adler, L. Lue, L.I. Sue, J. Bachalakuri, J. Henry-Watson, et al., Unified staging system for Lewy body disorders: correlation with nigrostriatal degeneration, cognitive impairment and motor dysfunction, Acta Neuropathol. 117 (2009) 613e634. , M. Carmona, Neuropathology of [20] I. Ferrer, A. Martinez, R. Blanco, E. Dalfo sporadic Parkinson disease before the appearance of parkinsonism: preclinical Parkinson disease, J. Neural Transm. 118 (2011) 821e839. [21] M. Atz, D. Walsh, P. Cartagena, J. Li, S. Evans, P. Choudary, et al., Methodological considerations for gene expression profiling of human brain, J. Neurosci. Methods. 163 (2007) 295e309. [22] M. Martins, A. Rosa, L.C. Guedes, B.V. Fonseca, K. Gotovac, S. Violante, et al., Convergence of miRNA expression profiling, alpha-synuclein interaction and GWAS in Parkinson's disease, PLoS One. 6 (2011) e25443. [23] R. Margis, C.R. Rieder, Identification of blood microRNAs associated to Parkinsonis disease, J. Biotechnol. 152 (2011) 96e101.
Contents lists available at ScienceDirect
Biochemical and Biophysical Research Communications journal homepage: www.elsevier.com/locate/ybbrc
Comparative analysis of RNA-Seq data from brain and blood samples of Parkinson's disease Paulami Chatterjee, Debjani Roy* Department of Biophysics, Bose Institute, Acharya J.C. Bose Centenary Building, P-1/12 C.I.T Scheme VII M, Kolkata, 700054 West Bengal, India
a r t i c l e i n f o
a b s t r a c t
Article history: Received 18 January 2017 Accepted 23 January 2017 Available online 26 January 2017
Parkinson's disease (PD) is the second most common neurodegenerative disorders throughout the world. In order to search for PD biomarkers, we performed a system-level study of RNA-Seq data from PD brain and blood samples. Differentially expressed miRs of RNA-Seq data were subjected to generate the Coexpression networks. Three highly co-expressed clusters were identified based on their correlation coefficient values and fold change ratio. SM2miR drugs of the miRs contained in the three highly coexpressed clusters were identified, and drugs common among these clusters were selected. Coexpressed miRs not previously known to be associated with PD were identified from both the samples. Functional enrichment analyses of these miR targets were done, and the pathways common and unique to both the samples were identified. Thus, our study presents a comparative analysis of miRs, their associated pathways, and drugs from brain and blood samples of PD that may help in system level understanding of this disease. miRs identified from our study may serve as biomarkers for PD. © 2017 Elsevier Inc. All rights reserved.
Keywords: Parkinson's disease RNA-Seq data miR Co-expression network Pathway analysis SM2miR drugs
1. Introduction Parkinson's disease (PD) is the second most common neurodegenerative disorders throughout the world. It is one of the chronic and progressive movement disorders in which symptoms get worse over time. Deregulation of microRNAs (miRs) has been implicated as one of the major causative factors in neurodegenerative diseases [1], including PD [2e4]. The previous study has pointed out that few miRs can be predominantly expressed in certain tissues, whereas the majority of miRs (82.9%) are neither specific for single tissues nor housekeeping miRs [5]. Therefore, knowledge of the expression pattern and distribution of different miRs in tissues is essential for the understanding of the physiological and pathological mechanisms associated with a disease [5]. Analyses of genomewide small RNA transcriptome provide an accurate and comprehensive knowledge of miR expression in the disease under different conditions [6]. RNA sequencing (RNA-Seq) which uses Next Generation Abbreviations: PD, Parkinson's Disease; miR, MicroRNA; RNA-Seq, RNA sequencing; DE, Differentially Expressed; Hclust, Hierarchical Clustering; ATC, Anatomic Therapeutic Classification. * Corresponding author. Department of Biophysics, Bose Institute, Acharya J.C. Bose Centenary Building, P-1/12 C.I.T Scheme VII M, Kolkata, 700054, India. E-mail addresses: [email protected] (P. Chatterjee), [email protected] (D. Roy). http://dx.doi.org/10.1016/j.bbrc.2017.01.121 0006-291X/© 2017 Elsevier Inc. All rights reserved.
Sequencing technology to quantify and discover RNAs in a cellular transcriptome [7,8], has been widely used for investigating the miRmediated pathogenesis of complex diseases. The high-resolution, genome-wide miR profile of PD obtained from RNA-seq offer insight into the molecular and pathological mechanisms that occur in the disease [9]. RNA-Seq mediated comparative blood transcriptome analysis of idiopathic and LRRK2 mutated PD has resulted in the identification of the significant difference in peripheral blood transcriptome [10]. Long non-coding RNAs and alternative splicing modulations have been identified from PD blood cells by RNA-Seq method [11]. However, the comparative analysis of miR profiles of RNA-Seq data from brain and blood samples has not been performed yet for PD. In this computational study, starting from PD brain and blood RNA-Seq data, we have identified disease-specific highly coexpressed miRs. Of these co-expressed miRs, we screened the miRs that are not previously known to be associated with PD. In this work, we studied highly co-expressed miRs, their associated pathways, and drugs from brain and blood samples that may help in system level understanding of this disease. The miRs identified from our study may serve as biomarkers for PD. 2. Methodology Fig. 1 depicts the workflow of our methodology.
558
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
2.1. Data collection RNA-Seq data of brain and blood samples of PD were retrieved from Gene Expression Omnibus (GEO) [Accession IDs- GSE72962 and GSE40915]. GSE72962 (Data 1) contains 29 PD and 33 neuropathologically-normal control from post-mortem prefrontal cortex Brodmann Area 9 (BA9) [9]. GSE40915 (Data 2) contains 7 PD patients before and after Deep Brain Simulation (DBS) treatment (which improves some of the PD motor symptoms) and 6 agematched healthy control samples from blood leukocytes [6]. 2.2. Differentially expressed miR selection In order to identify the differentially expressed (DE) miRs over
disease and normal conditions, paired two sample t-test was applied on each of the expression datasets. The t-test measures the statistical significance of the dataset in terms of test statistic t, which is given by:
xy t ¼ qffiffiffiffiffiffiffiffiffiffiffiffi2ffi sy s2x n þm Where x and y are the sample means, sx and sy are the sample standard deviations, n and m are the sample sizes for two samples, x andy. Under the null hypothesis, this test returns the probability (p-value) of observing a value as extreme or more extreme of the
Fig. 1. Methodology for the analysis of RNA-Seq data from brain and blood samples of PD.
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
test statistic. In our study, miRs with p-value <0.05 (95% confidence level) were selected as DE miRs in PD. 2.3. Agglomerative hierarchical clustering of the DE miRs In order to arrange the DE miRs into groups of similar expression patterns, we employed hierarchical clustering. Hierarchical clustering is a method that groups objects into clusters depending on their distance. It works on the idea that objects that are close to each other are more connected than to objects that are present at a distance. In agglomerative hierarchical clustering, each object is considered as a cluster; hence it starts with a single object and aggregates nearby objects into
559
clusters [12]. For the calculation of the distance between objects in the data matrix, we used the Euclidean distance and average linkage which uses the average distance between all pairs of objects in any two clusters. Given a mx-by-n data matrix X where the row vectors are x1, x2, …, xmx, and my-by-n data matrix Y where the row vectors are y1, y2, …., ymy, the Euclidean distance between the vectors xs and yt are defined as d2st ¼ ðxs yt Þðxs yt Þ0 . 2.4. Fold change calculation Fold change is a measure describing the extent of changes in expression of a gene in disease condition as compared to control.
Fig. 2. Co-expression network containing highly positively correlated miR pairs (r > 0.8). (2a) Co-expression network for PD Brain sample, (2b) Co-expression network for PD Blood sample. Here nodes represent highly correlated miRs and edges between miRs represent the significant co-expression relationship. Node size is mapped to the degree of each node in the network.
560
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
Fold change values were calculated for each of the DE miRs using MATLAB.
Fold Change ¼ log2
ED EC
Where ED and EC represent the arithmetic mean of the expression values of a particular miR over disease conditions and control conditions respectively. Depending on the fold change values upregulated and down-regulated DE miRs were identified from each dataset. 2.5. Co-expression network construction Co-expression Network was constructed for each of the DE miR sets based on the Pearson Correlation coefficient (r) which computes all possible combination of pairs of DE miRs over control and disease conditions. The basic formula for computing r is
r¼
P XX Y Y nSx Sy
Where, X and Y are the scores of the variables whose Correlation Coefficient are being measured (here X and Y represent the expression values of two miRs), X and Y are their respective means and Sx and SY are the respective standard deviations, and n is the number of individuals or pairs of scores in the sample [12]. The Correlation Coefficient calculation was performed using MATLAB. Highly positively correlated DE miRs (r > 0.8) and highly negatively correlated DE miRs (r 0.8) were selected for this purpose. Visualization of the network was performed in Cytoscape [13] and the topological significance analysis was performed using Network Analyzer module of Cytoscape software [13]. 2.6. Identification of drugs and target mRNAs associated with DE miRs FDA approved drugs associated with the DE miRs were identified from SM2miR database [14]. Experimentally validated mRNA targets associated with the miRs were identified from DIANA TOOLS-TarBase v7.0 [15]. Functional enrichment analysis of these target mRNAs was performed in DAVID Bioinformatics Resources 6.8 [16,17]. Information on the experimentally validated diseaserelated genes of PD was obtained from PDbase database [18]. 3. Results 3.1. Identification of DE miRs Our study identified 321 miRs (p-value <0.05) differentially expressed in PD brain samples and 137 miRs (p-value<0.05) differentially expressed in PD blood samples. Interestingly, we found 56 miRs common between these two samples (Supplementary Table 1). We processed these two DE miR datasets using the same pipeline and subjected them to further analyses. 3.2. Hierarchical clustering analysis Hierarchical Clustering (Hclust) arranged the DE miRs into groups of miRs based on their similar expression patterns. The 321 DE miRs of Data 1 (Brain sample) were divided into five clusters consisting of 88, 110, 1, 25, and 97 miRs respectively. The 137 DE miRs of Data 2 (Blood samples) were also divided into five clusters consisting of 127, 7, 1, 1, and 1 miRs respectively. To ascertain the
function of these miRs, we studied the miR associated pathways of significant miR clusters in mirPath v.3 software of DIANA TOOLS. Most of the clusters from brain and blood samples contain several PD-related pathways (p-value <0.05) e.g. Axon guidance, Prion diseases, Ubiquitin mediated proteolysis, Endocytosis, etc. 3.3. Co-expression network construction Two Co-expression Networks were constructed from each of the DE miR sets of brain and blood samples. We selected highly positively correlated miR pairs (r > 0.8) (Fig. 2), and highly negatively correlated miR pairs (r 0.8) from both the samples (Fig. 3). We named highly positively correlated miR pairs as Co-expression Network 1 and highly negatively correlated miR pairs as Coexpression Network 2 for both the datasets. Thus from PD brain samples 70 DE miR pairs with r > 0.8 and 22 DE miR pairs with r 0.8 were obtained. From PD blood samples 394 DE miR pairs with r > 0.8 and 8 DE miR pairs with r 0.8 were obtained. Supplementary Table 2 and Supplementary Table 3 show the degree distribution pattern of all these highly co-expressed miRs in both the samples. 3.4. Clustering of the co-expressed miRs The highly co-expressed miRs obtained from both the samples were clustered based on their fold change value and co-expression correlation coefficient. Thus, three clusters (CoexpCluster1, CoexpCluster 2 and CoexpCluster 3) were obtained from both the samples. CoexpCluster 1 contains positively correlated upregulated miR pairs (where expression of both the miRs increase in disease sample compared to control), CoexpCluster 2 contains positively correlated down-regulated miR pairs (where expression of both the miRs decrease in disease sample compared to control), and CoexpCluster 3 contains Negatively correlated miR pairs (where expression of one miR increases and other one decreases in disease sample compared to control). It was found that CoexpCluster1 contained 47 miR pairs, CoexpCluster 2 contained 23 miR pairs, and CoexpCluster 3 contained 22 miR pairs in PD brain samples. It was found that CoexpCluster1 contained 12 miR pairs, CoexpCluster 2 contained 359 miR pairs, and CoexpCluster 3 contained 8 miR pairs in PD blood samples. These miR clusters were then subjected to further studies. 3.5. Identification of drugs associated with the miR clusters In order to study the drug-mediated regulation of miR expression, SM2miR drugs associated with the miRs of each cluster were identified. It was found that 41, 20 and 25 drugs were associated with the miRs of CoexpCluster 1, CoexpCluster 2, and CoexpCluster 3 respectively in Data 1, whereas, 30, 42, and 26 drugs were associated with the miRs of CoexpCluster 1, CoexpCluster 2, and CoexpCluster 3 respectively in Data 2. Interestingly, fourteen drugs in brain sample and six drugs in blood sample were common among these three miR clusters (Table 1). We focused on these common drugs of each sample and studied their associated miRs. Unknown miRs (which were not previously known to be involved in PD) associated with these common drugs were selected from each dataset. In this way, 41 unknown miRs from Data 1, and 30 unknown miRs from Data 2 were selected. The mRNA targets of these unknown miRs were studied from each dataset. 2229 experimentally validated mRNA targets (associated with 41 unknown miRs) were obtained from Data 1 and 1927 experimentally validated mRNA targets (associated with the 30 unknown miRs) were obtained from Data 2.
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
561
Fig. 3. Co-expression network containing highly negatively correlated miR pairs (r -0.8): (3a) Co-expression network for PD Brain sample, (3b) Co-expression network for PD Blood sample. Here nodes represent highly correlated miRs and edges between miRs represent the significant co-expression relationship. Node size is mapped to the degree of each node in the network.
Table 1 SM2miR drugs common among three miR clusters in brain and blood samples of PD.a SM2miR Drugs common among three clusters PD brain samples
PD blood samples
Drug ID
DrugName
ATC
DrugID
DrugName
ATC
DB01262 DB00544 DB01169 DB00169 DB00441 DB02546 DB00126 DB00624 DB00136 DB00014 DB00255 DB00396 DB00467 DB00295
Decitabine Fluorouracil Arsenic trioxide Cholecalciferol Gemcitabine Vorinostat Vitamin C Testosterone Calcitriol Goserelin Diethylstilbestrol Progesterone Enoxacin Morphine
Antineoplastic agents Antineoplastic agents Antineoplastic agents Musculo-skeletal system Antineoplastic agents Antineoplastic agents Not Available Genito-urinary system and sex hormone Alimentary tract and metabolism, Dermatologicals Antineoplastic agents Genito-urinary system and sex hormone Genito-urinary system and sex hormone Antiinfectives for systemic use Nervous System
DB00783 DB00467 DB00441 DB00544 DB01262 DB00255
Estradiol Enoxacin Gemcitabine Fluorouracil Decitabine Diethylstilbestrol
Genito-urinary system and sex hormone Antiinfectives for systemic use Antineoplastic agents Antineoplastic agents Antineoplastic agents Genito-urinary system and sex hormone
a
Drugs common between two samples are shown in bold.
562
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
Table 2 Most significant over-representative KEGG pathways (p-value <0.05) associated with the miRs that are not previously known to be associated with PD identified from brain and blood samples. Common KEGG pathways associated with unknown miRs of both the PD brain and blood samples
Significant KEGG pathways uniquely associated with the PD brain sample
Significant KEGG pathways uniquely associated with the PD blood sample
hsa04110:Cell cycle
hsa04550:Signaling pathways regulating pluripotency of stem cells hsa04141:Protein processing in endoplasmic reticulum hsa04068:FoxO signaling pathway hsa04390:Hippo signaling pathway hsa05169:Epstein-Barr virus infection hsa05161:Hepatitis B hsa04151:PI3K-Akt signaling pathway hsa04710:Circadian rhythm hsa04152:AMPK signaling pathway hsa04540:Gap junction hsa04142:Lysosome hsa05120:Epithelial cell signaling in Helicobacter pylori infection hsa04150:mTOR signaling pathway hsa04730:Long-term depression hsa04728:Dopaminergic synapse hsa04320:Dorso-ventral axis formation hsa05221:Acute myeloid leukemia hsa05164:Influenza A hsa04668:TNF signaling pathway
hsa04512:ECM-receptor interaction
hsa04360:Axon guidance hsa04350:TGF-beta signaling pathway hsa04120:Ubiquitin mediated proteolysis hsa04722:Neurotrophin signaling pathway hsa04144:Endocytosis hsa05214:Glioma hsa04330:Notch signaling pathway hsa05200:Pathways in cancer hsa04510:Focal adhesion hsa04010:MAPK signaling pathway hsa05220:Chronic myeloid leukemia hsa04115:p53 signaling pathway hsa04520:Adherens junction hsa05223:Non-small cell lung cancer hsa04910:Insulin signaling pathway hsa04720:Long-term potentiation hsa04810:Regulation of actin cytoskeleton hsa04310:Wnt signaling pathway
3.6. Functional enrichment analysis Functional enrichment analyses were performed with the unknown miR targets (2229 and 1927 mRNA targets respectively) of both the samples. It was found that several disease-specific pathways were associated with the unknown miRs of the two datasets. We identified pathways common between brain and blood samples and pathways uniquely associated with each of the sample (Table 2).
4. Discussion The substantia nigra in the midbrain is the most affected brain region in PD but the analysis of this region from post-mortem PD brains may only highlight miRs associated with changes in cellular composition [9]. On the other hand, analysis of post-mortem prefrontal cortex, (a region which contains both dopaminergic neuron projections and pathological hallmarks [19,20], may lead to the identification of miRs associated with disease progression [9]. However, post-mortem brain tissue is not readily available, and entails RNA quality concerns [21]. There comes the importance of studying blood samples in PD. Blood leukocytes are readily obtained, and the RNA can be obtained at high quality from them [6]. Interestingly, the concept of blood-based diagnostic markers is a highly emerging field of study nowadays [22,23]. With this notion, in this computational study, we have selected RNA-Seq PD data obtained from prefrontal cortex and blood leukocytes, which provide the global miR profiling of PD. Co-expression network identified the highly correlated miR pairs that are differentially expressed in PD brain and blood samples. Differential fold change and co-expression patterns of miRs are evident in these three discernible clusters (CoexpCluster 1, CoexpCluster 2 and CoexpCluster 3). These highly correlated miR pairs may undergo a similar type of regulation pattern in PD. We searched SM2miR database to identify drugs that are involved in each of these miR cluster. Common drugs among these three clusters were identified (Table 1). Several drugs are found to be shared between two samples. Table 1 shows the detailed
hsa04912:GnRH signaling pathway hsa04916:Melanogenesis hsa00600:Sphingolipid metabolism
information on Anatomic Therapeutic Classification (ATC) of these drugs. These correlated miRs and their associated drugs may provide a detailed overview of the probable therapeutic targets of this disease. Common drug associated miRs that are not previously known to be involved in PD were selected from each sample (Fig. 1). In this way, 41 unknown miRs from Data 1, and 30 unknown miRs from Data 2 were selected. Unknown miRs uniquely associated with each
Table 3 miRs that are not previously known to be associated with PD identified from brain and blood samples. miRs from brain
miRs from blood
hsa-miR-124-5p hsa-miR-127-3p hsa-miR-1277-5p hsa-miR-129-2-3p hsa-miR-134-5p hsa-miR-148b-3p hsa-miR-148b-5p hsa-miR-16-2-3p hsa-miR-181c-5p hsa-miR-191-5p hsa-miR-20a-5p hsa-miR-212-3p hsa-miR-212-5p hsa-miR-339-5p hsa-miR-340-3p hsa-miR-342-3p hsa-miR-345-5p hsa-miR-34c-3p hsa-miR-372-3p hsa-miR-373-3p hsa-miR-381-3p hsa-miR-382-5p hsa-miR-411-5p hsa-miR-432-5p hsa-miR-491-5p hsa-miR-516a-5p hsa-miR-522-3p hsa-miR-628-5p hsa-miR-769-5p
hsa-miR-107 hsa-miR-1286 hsa-miR-130b-5p hsa-miR-181d hsa-miR-183-5p hsa-miR-18a-5p hsa-miR-1910 hsa-miR-2116 hsa-miR-223 hsa-miR-299 hsa-miR-29b-2-5p hsa-miR-30e hsa-miR-3133 hsa-miR-320d-1 hsa-miR-320d-2 hsa-miR-328 hsa-miR-4306 hsa-miR-486 hsa-miR-641 hsa-miR-98
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564
563
Table 4 Detailed information of miRs that are not previously known to be associated with PD and common between brain and blood samples. Common unknown miRs
hsa-let-7d-5p hsa-let-7f-5p hsa-miR-106b-5p hsa-miR-15b-5p hsa-miR-181a-5p hsa-miR-324-5p hsa-miR-454-3p hsa-miR-671-5p hsa-miR-92a-3p hsa-miR-92b-3p hsa-miR-93-5p
Fold change percentage
mRNA Targets associated with the miRs
Hierarchical cluster number
Brain sample
Blood sample
Total targets from TarBase
Targets known in PD
Brain sample
Blood sample
2.94 1.59 3.53 8.36 2.15 3.4 4.47 9.9 2.15 4.33 2.35
120.15 53.23 108.79 110.74 70.66 66.59 110.12 153.37 177.76 102.7 63.67
39 41 97 359 27 58 276 200 33 337 23
7 4 20 62 9 11 61 42 6 55 4
4 4 5 5 3 5 5 2 4 4 5
1 1 1 1 1 1 1 1 1 1 1
sample were identified (Table 3). Additionally, common miRs between two samples were identified. (Table 4). Fold change values of these common unknown miRs indicated that these miRs show differential expression pattern between PD brain and blood samples. Most of these miRs were up-regulated in PD brain samples whereas they were mostly down-regulated in PD blood samples (Table 4). Co-expression network indicated that these common miRs are co-expressed with already known PD miRs. Moreover, hierarchical clustering also illustrated that these common miRs have similar expression pattern with already known PD miRs since they were clustered with known PD miRs. It was found that most of these common unknown miRs were clustered into cluster 4 and 5 of Data 1 (Table 4) with several other miRs already known to be involved in PD. In data 2, all of these common unknown miRs were clustered into cluster 1 (Table 4) which also contains several known miRs in PD. Further study on the targets of these common unknown miRs revealed that all of these miRs are associated with genes already known in PD (Table 4). Functional enrichment analysis of the targets of unknown miRs identified disease-specific significant pathways associated with the unknown miRs of the two datasets. Interestingly, we found several over-representative pathways shared between the two datasets (Table 2). We also identified the over-representative pathways uniquely associated with each dataset (Table 2). Our study performed a comparative analysis of expression profiles of miRs obtained from two different tissues (brain and blood) and identified disease-specific miRs and shed light on their expression pattern, drug-mediated regulation and significant pathways associated with them. The highly correlated differentially expressed miRs and associated over-representative pathways offer insight into the significant molecular mechanisms that occur in the disease. 5. Conclusion In this computational study, starting from PD brain and blood RNA-Seq data, we have identified disease-specific miRs that are shared between two samples. These common miRs have also been found to be targeting known PD genes suggesting their significance in PD etiology. We have also identified the unique miRs from both the samples that are not previously known to be associated with PD and shed light on their expression pattern, drug-mediated regulation and significant pathways associated with them. The Coexpression networks consisting of the differentially expressed miRs helped us to identify the highly correlated miR pairs from each sample. The highly correlated differentially expressed miRs and associated over-representative pathways offer insight into the significant molecular mechanisms that occur in the disease. To our knowledge, this is the first study that provides a comparative
analysis of brain and blood samples of PD RNA-Seq data and illustrated miR disease markers, drugs and significant pathways associated with this disease. This systems level study may open up the opportunities for future PD therapeutic developments. Conflict of interest The authors declare that they do not have any potential conflict of interest. Acknowledgement The authors would like to thank the Department of Biophysics, Bose Institute. This work is not funded by any external funding agency. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.bbrc.2017.01.121. References [1] A.W. Chan, J. Kocerha, The path to microRNA therapeutics in psychiatric and neurodegenerative disorders, Front. Genet. 3 (2012) 82. [2] J. Kim, K. Inoue, J. Ishii, W.B. Vanti, S.V. Voronov, E. Murchison, et al., A MicroRNA feedback circuit in midbrain dopamine neurons, Science 317 (2007) 1220e1224. [3] S. Gehrke, Y. Imai, N. Sokol, B. Lu, Pathogenic LRRK2 negatively regulates microRNA-mediated translational repression, Nature 466 (2010) 637e641. ~ ones-Moyano, S. Porta, G. Escaramís, R. Rabionet, S. Iraola, [4] E. Min B. Kagerbauer, et al., MicroRNA profiling of Parkinson's disease brains identifies early downregulation of miR-34b/c which modulate mitochondrial function, Hum. Mol. Genet. 20 (2011) 3067e3078. [5] N. Ludwig, P. Leidinger, K. Becker, C. Backes, T. Fehlmann, C. Pallasch, et al., Distribution of miRNA expression across human tissues, Nucleic Acids Res. 44 (2016) 3865e3877. [6] L. Soreq, N. Salomonis, M. Bronstein, D.S. Greenberg, Z. Israel, H. Bergman, et al., Small RNA sequencing-microarraay analyses in Parkinson leukocytes reveal deep brain stimulation-induced splicing changes that classify brain region transcriptomes, Front. Mol. Neurosci. 6 (2013) 10. [7] Z. Wang, M. Gerstein, M. Snyder, RNA-Seq: a revolutionary tool for transcriptomics, Nat. Rev. Genet. 10 (2009) 57e63. [8] C.C. Pritchard, H.H. Cheng, M. Tewari, MicroRNA profiling: approaches and considerations, Nat. Rev. Genet. 13 (2012) 358e369. [9] A.G. Hoss, A. Labadorf, T.G. Beach, J.C. Latourelle, R.H. Myers, microRNA profiles in Parkinson's disease prefrontal cortex, Front. Aging Neurosci. 8 (2016) 36. [10] J. Infante, C. Prieto, M. Sierra, P. S anchez-Juan, I. Gonz alez-Aramburu, nchez-Quintana, et al., Comparative blood transcriptome analysis in C. Sa idiopathic and LRRK2 G2019S-associated Parkinson's disease, Neurobiol. Aging. 38 (214) (2016) e1e5. [11] L. Soreq, A. Guffanti, N. Salomonis, A. Simchovitz, Z. Israel, H. Bergman, et al., Long non-coding RNA and alternative splicing modulations in Parkinson's leukocytes identified by RNA sequencing, PLoS Comput. Biol. 10 (2014) e1003517. [12] P. Chatterjee, M. Bhattacharyya, S. Bandyo padhyay, D. Roy, Studying the
564
[13]
[14]
[15]
[16]
[17]
P. Chatterjee, D. Roy / Biochemical and Biophysical Research Communications 484 (2017) 557e564 system- level involvement of microRNAs in Parkinson's disease, PLoS One 9 (2014) e93751. P. Shannon, A. Markiel, O. Ozier, N.S. Baliga, J.T. Wang, et al., Cytoscape: a software environment for integrated models of biomolecular interaction networks, Genome Res. 13 (2003) 2498e2504. X. Liu, S. Wang, F. Meng, J. Wang, Y. Zhang, E. Dai, et al., SM2miR:a database of the experimentally validated small molecules' effects on microRNA expression, Bioinformatics 29 (2013) 409e411. I.S. Vlachos, M.D. Paraskevopoulou, D. Karagkouni, G. Georgakilas, T. Vergoulis, I. Kanellos, et al., DIANA-TarBase v7.0: indexing more than half a million experimentally supported miRNA:miRNA interactions, Nucleic Acid. Res. 43 (2015) D153eD159. D.W. Huang, B.T. Sherman, R.A. Lempicki, Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists, Nucleic Acids Res. 37 (2009) 1e13. D.W. Huang, B.T. Sherman, R.A. Lempicki, Systematic and integrative analysis of large gene lists using DAVID Bioinformatics Resources, Nat. Protoc. 4 (2009) 44e57.
[18] J.O. Yang, W.Y. Kim, S.Y. Jeong, J.H. Oh, S. Jho, J. Bhak, et al., PDbase: a database of Parkinson's disease-related genes and genetic variations using substantia nigra ESTs, BMC Genomics 10 (2009) S32. [19] T.G. Beach, C.H. Adler, L. Lue, L.I. Sue, J. Bachalakuri, J. Henry-Watson, et al., Unified staging system for Lewy body disorders: correlation with nigrostriatal degeneration, cognitive impairment and motor dysfunction, Acta Neuropathol. 117 (2009) 613e634. , M. Carmona, Neuropathology of [20] I. Ferrer, A. Martinez, R. Blanco, E. Dalfo sporadic Parkinson disease before the appearance of parkinsonism: preclinical Parkinson disease, J. Neural Transm. 118 (2011) 821e839. [21] M. Atz, D. Walsh, P. Cartagena, J. Li, S. Evans, P. Choudary, et al., Methodological considerations for gene expression profiling of human brain, J. Neurosci. Methods. 163 (2007) 295e309. [22] M. Martins, A. Rosa, L.C. Guedes, B.V. Fonseca, K. Gotovac, S. Violante, et al., Convergence of miRNA expression profiling, alpha-synuclein interaction and GWAS in Parkinson's disease, PLoS One. 6 (2011) e25443. [23] R. Margis, C.R. Rieder, Identification of blood microRNAs associated to Parkinsonis disease, J. Biotechnol. 152 (2011) 96e101.