A global analysis on the differential regulation of RNA binding proteins (RBPs) by TNF–α as potential modulators of metabolic syndromes

Highlights

• •Genes involved in signaling and metabolic pathways leading to metabolic syndrome (MetS) are grouped as MetS genes.
• •Earlier reports from our lab shows TNF-α significantly modulates the expression of 56 MetS genes at the alternative splicing level which were predicted to interact with various RNA-binding proteins (RBPs) when exposed to TNF-α
• •RNA-Seq analysis identified 1218 unique genes regulated at the mRNA level by TNF-α, 204 of which are MetS genes, among them 10% are RBPs.
• •TNF-α changes the phosphorylation status of certain RBPs such as SR proteins, possibly by modulating the activity of the upstream kinases.
• •Taken together, our results show TNF-α influences the regulation of the RBPs at various levels, influencing changes of the alternative splicing pattern of the MetS genes.

Abstract

Metabolic syndrome (MetS) is associated with a group of conditions, which enhances the risk of diabetes, heart diseases and stroke in the affected individuals. Earlier reports from our lab have shown that Tumor necrosis factor-α (TNF-α) significantly modulates the expression of 56 genes at the alternative splicing level which are involved in various signaling and metabolic pathways (MetS genes) connected to MetS. These MetS genes were predicted to interact with various RNA-binding proteins (RBPs) when exposed to TNF-α, resulting changes in their alternative splicing patterns. Here we are presenting data of an RNA-Seq analysis, which identified 1218 unique, and significantly regulated genes by TNF-α, 15% of which are RBPs . Among the 1218 genes, 204 genes have been identified as MetS genes by the ingenuity pathway analysis, and 10% of the MetS genes are found as RBPs. Our results also show that TNF-α changes the phosphorylation status of certain RBPs such as SR proteins, crucial players in alternative splicing, possibly via changing the activation status of certain upstream signaling molecules which also act as upstream kinases for these proteins. Taken together, these findings suggest that TNF-α influences the regulation of the RBPs at the various levels for their expression, which may lead to the alteration of the splicing pattern of the MetS genes. MetS genes acting as RBPs and are modulated by TNF-α, predict the existence of highly interconnected mechanisms which require further analysis to understand their dual roles on the onset of these diseases.

Introduction

Metabolic syndromes (MetS) are associated with an assemblage of conditions including diabetes, obesity, insulin resistance, high triglyceride levels and cardiovascular risks. Inflammatory signals mostly mediated by the pro-inflammatory cytokines such as Tumor necrosis factor-α (TNF-α) and Interleukin-6 (IL-6), have stated to be the connecting link between these risk factors [1,2]. Infiltration of macrophages in hypoxic and enlarged adipocytes causes overproduction of inflammatory chemokines such as TNF-α [3], [4], [5], causing impairment of insulin signaling pathways and leading to insulin resistance and later stage of Type 2 diabetes (T2DM) [6,7]. Anti-TNF-α therapy has been proven efficient for various MetS [8].

Different signaling pathways modulated by TNF-α in MetS, are well documented [5,9,10] TNF-α-mediated activation of stress related MAPKs (p38 MAPK and JNK) and various other Ser/Thr kinases (IKKβ, PKCθ etc.) have reported to contribute towards insulin resistance. TNF-α may trigger the activation of other cytokines-mediated pathways creating a second wave of signaling [11]. Other than these pathways, modulation of synthesis and activation of signaling proteins such as PPAR-γ and GLUT4 may also contribute towards metabolic disorders [7,12,13]. Yet, the comprehensive role of TNF-α in MetS including its downstream effector targets have not been fully understood.

Earlier report from our lab shows that TNF-α induces modulation of alternative splicing of genes (including CREB5, NFKB1, NFKB2, TP53 etc.) involved in various signaling and metabolic pathways (henceforth referred to as MetS genes), connected to MetS [14]. Alternative splicing is a posttranscriptional mechanism that creates diversity of proteins from a similar genetic background and is modulated by various RNA binding proteins (RBPs). The RBPs, once loaded, may decide the fate of the pre-mRNAs for their splicing, polyadenylation and capping [15,16]. The RBPs may have defined (canonical) or undefined (non-canonical) RNA binding domains (RBDs) [17]; the latter group uses intrinsically disordered regions (IDRs) for RNA binding [16]. The other classes of RBPs without a recognizable structural RBD or an assigned function in RNA biology are known as “enigmRBP” [17]. These unconventional RBPs include protein with various biological functions like cytoskeleton remodeling, protein folding, ATP-binding and enzymatic functions in classic metabolic pathways. Almost 9% of conserved core RBPs between human and yeast are shown to be metabolic enzymes. The classic enzymes belonging to this category includes oxidoreductases, isomerases, hydrolases, isomerases, lyases and ligases [18], [19], [20]. Metabolic enzymes were identified as RNA binding proteins by several groups [18,[21], [22], [23], [24]]. A large-scale analysis revealed as many as 71 metabolic enzymes as RNA binding proteins in humans. Some of these enzymes are shown to bind to their own mRNAs, predicting a feedback loop (moonlighting activity), or contributors in modulation of enzyme activity [25,26]. These moonlighting enzymes include enzymes involved in glycolysis, tricarboxylic acid cycle, lipid metabolism and deoxynucleotide biosynthesis [25], [26], [27].Differential regulation of several RBP's such TTP, HuR, QKI, SRSF6 are reported in connection to various metabolic disorders [28,29]

A global analysis of RBPs regulated by TNF-α at the transcriptional as well as at the post-transcriptional level is yet to be performed for a better understanding of the TNF-α associated MetS. A recent study from our lab has reported a genome wide map of 13,395 unique RNA-RBP interactions (RPIs) upon TNF-α treatment on the 228 significantly regulated alternatively spliced genes. Among which, 22% of the total interactions were observed on the 56 MetS genes with 32 unique RBPs (predicted by RBPDP database) [30].This result demonstrated the potential involvement of the RBPs in the modulation of the MetS genes at the alternative splicing level.

Expression and functional aspects of RBPs are mostly determined by various signaling pathways [31,32] modulated by extracellular cues including cytokines [33]. Apart from the transcriptional regulation, the functional aspects as well as their shuttling back and forth between the nucleo-cytoplasmic pools of some of the RBPs, especially true for the SR-proteins, depend on their phosphorylation status [34,35]. Reported kinases responsible for phosphorylation of these RBPs include SRPK1/2, CLK and PRP4 [36]. Other kinases involved in crucial cellular signaling cascades including PI3K, P38MAPK, PKB, mTORC etc., are also known to modulate the phosphorylation status of these RBPs, either directly or by recruiting mediator proteins, such as Sam68 and HuR [37].

In the present study, we have identified the regulation of RBPs at the mRNA and post-translational level by TNF-α. Our data obtained from an RNA-Seq analysis shows that 10% of the MetS genes, regulated by TNF-α, are RBPs. We also obtained results showing that the phosphorylation status of some of the SR-proteins, which are the crucial players in alternative splicing, are regulated by TNF-α treatment. The change in the phosphorylation status of these SR-proteins correlated with the phosphorylation and thus the activation status of two crucial Ser-Thr kinases; p38 MAPK and mTOR. Taken together, this study substantiates our prediction that regulation of RBPs by TNF-α, at various levels, potentially involving upstream signaling pathways, may contribute towards differential regulation of MetS genes and in turn, influences the onset of MetS.

Methodology

Mammalian cell culture: Human Embryonic kidney (HEK 293) cells (NCCS, Pune) and Human skeletal myoblasts (A12555, Thermo Scientific), were maintained in low glucose (HEK 293; 1g/liter) or in normal glucose (4.5 g/l) DMEM media (HiMedia) supplemented with 10% FBS (Gibco, South American origin) at 37 °C in 5% CO₂. A12555 cells were incubated for 48 h to induce differentiation as per the manufacture's recommendation (Thermoscientific). The cells were treated with TNF-α (Sigma) (10 ng /ml) for different periods in serum-deprived media. Most of the experiments were performed after 6 h of TNF-α treatment at the concentration of 10ng/ml in vitro.

RNA Isolation: Total RNA was extracted from the cells using RNA Xpress reagent (HiMedia) according to manufacturer's instruction. The isolated RNA was quantified using Nanodrop spectrophotometer (Thermofischer scientific). The quantified total RNA (1 µg) was used for cDNA synthesis following a protocol of High Capacity Reverse Transcription kit (Applied Biosystems, Thermoscientific).

Quantitative real time PCR: 2.5 ng of cDNA were used to perform qPCR. The reactions were performed using Brilliant Ш Ultra-Fast SYBR green qPCR Master Mix in AriaMx PCR System (Agilent Technologies) with the following cycling conditions: 95 °C for 3 min (initial denaturation), followed by 40 cycles of 95 °C for 30 s, 58 °C for 30 s and 72 °C for 60 s. 18srRNA served as the internal control. The primers used have been enlisted in in the Supplementary Table 1.

Transcriptome sequencing: A reference based transcriptome sequencing (for Homo sapiens) was performed from the total RNA isolated from one set of TNF-α treated and one set of untreated cells after the respective cDNA library preparation. No replicates of the RNA-Seq was performed for the data analysis. The protocol for RNA extraction, cDNA synthesis and NGS analysis is the same as was mentioned in our previous publication [14]. Cufflinks-2.1.1 was used for differential regulation analysis from RNA-seq samples.

RPKM (Reads per kilobase per million mapped reads) value was calculated using the formula, RPKM = 10^9 x C/N*L, Where C = number of mapping reads, N = Total number of reads, L = length of the transcript for each gene.

A minimal read count of 5 reads was used to filter out the genes not expressed. To determine the differentially expressed genes, Cuffdiff was used to compare the two categories (Treated vs Control). A two-fold change and p-value <0.05 cut-off was used to report the DEGs. Owing to the absence of replicates, no FDR correction was used.

Gene Ontology Analysis: WEB-based GEne SeT AnaLysis Toolkit (WebGestalt): Over-Representation Analysis (ORA) with the 1218 short-listed genes was performed using the WEB-based GEne SeT AnaLysis Toolkit (WebGestalt; doi: 10.1093/nar/gkz401), considering all the unique genes reported by the RNA-Seq analysis (12,715) as the background. Out of the 1218 genes, 1162 were mapped against the 11,464 unique genes (out of 12,715 genes) from the reference list. For significance level, FDR<0.05 was chosen.

Pathway analysis

Ingenuity pathway analysis (IPA): The enriched pathways from the data set and their ratios and significant scores (using the right-tailed Fisher exact test) were calculated using IPA. Calculation of the ratio and the p-values is the same as mentioned in our previous publication [14]. A threshold score of p-value 0.05 and 0.01 corresponding to a significance score of 1.3 and 2 respectively, was used to report the significantly enriched pathways.

KEGG pathway analysis: The enriched pathways and the number of genes mapped against each pathway are represented by KEGG. The pathways related to metabolic disorders such as type 2 diabetes were extracted. The distributions of mapped genes related to metabolic disorders were represented in the pie chart.

Network analysis

STRING database analysis: STRING database represents a network detailing the interaction between the genes or proteins used as input. Isolated nodes represent the genes without any interaction. The genes/proteins are represented as nodes and the edges represent the interactions among them. The output from the STRING DB showing the connections among the genes (nodes) (Supplementary Datasheet 3) was used to assess the pairwise connections among the genes. Genes having ten or more connections were considered as highly connected nodes.

IPA network analysis: IPA network analysis was performed as mentioned earlier [14]. Enrichment score was given to the focus molecules (genes from the user's list).

EuRBPDB analysis: The differentially regulated genes by TNF-α treatment were used as inputs for the EuRBPDB (Eukaryotic RNA binding protein database; http://eurbpdb.syshospital.org/search_rbp.php) database to retrieve the RBPs [17]. EuRBPDB updated version, v1.2 was used as the search engine with the specific chosen parameter “RBP type”. Data from our RNA-seq analysis (in terms of gene names), was fed into the EuRBPDB database using the “requests” python library. The search pages were analyzed to identify RBP type for the genes that are fed into it.

TRRUST analysis: The differentially regulated MetS genes by TNF-α treatment were used as inputs for the analysis into TRRUST (Transcriptionally regulatory relationships unraveled by sentence-based text mining) version 2 database (https://www.grnpedia.org/trrust/) to identify the transcription factors.

Western Blot analysis: Cells were exposed TNF-α for different time period (30’ to 6 h) and cell lysate was prepared using a lysis buffer (4x Lammeli buffer supplemented with 1X mammalian protease phosphatase inhibitor cocktail) and stored at -20 °C. After SDS-PAGE, wet transfer was carried out using PVDF membrane. After incubation with the primary and the secondary antibodies, the blots were developed using ECL plus western blot developing system (GE healthcare) on X-ray films. β-Actin served as the internal control for all the experiments. Antibodies used are as follows; p-SR (1:1000, MABE50, Merck), p-P38MAPK (1:1000, cell signaling Technology), p-mTOR (1:1000, CST), β-Actin (1:1000, R&D), Anti-Mouse HRP (1:2000, Sigma) and Anti-Rabbit HRP (1:2000, Cell signaling Technology).

Statistical analysis: Statistical evaluation for the p-values was done by paired, two-tailed t-test in each data set. The symbols *, ** and *** represent p < 0.05, < 0.01 and < 0.001, respectively.

Results

Genome-wide transcriptome analysis revealed the identity of differentially regulated genes by TNF- α:

Identification of the differentially regulated genes upon stimulation by TNF- α was carried out by a genome wide transcriptome analysis, with the total RNA extracted from the TNF-α treated and untreated HEK293 cells.

Considering a twofold change cut-off, a total number of 6056 and 5495 transcripts were found to be up and down regulated, respectively (Supplementary Datasheet 1). On excluding the transcripts with no gene names, the number got reduced to 4566 and 4465 for the up and down regulated transcripts, respectively. These transcripts corresponded to 3645 up-regulated and 3502 down-regulated genes (Fig. 1a). From this gene pool, considering a p-value cut-off <0.05, 1218 unique genes were sorted out, combining the up and down regulated ones (Fig. 1b and Supplementary Datasheet 1). Further analysis has been carried out with this gene pool.

Fig. 1.

Quantitative representation of the data obtained from the NextSeq analysis. (a) The bar diagrams depict the no. of transcripts and genes got up or down regulated with a two-fold cut-off, upon TNF-α treatment compared to the control (b) Representation of the no. of genes up and down regulated with p-value <0.05 and the total no. of unique genes.

Validation of RNA-seq data

Among the differentially expressed genes identified by RNA-seq analysis, a few candidate genes, which were already reported in the context of metabolic disorders, were short-listed for validation. We have previously reported the modulation of MetS genes at the alternative splicing level by TNF-α. Here we chose a few MetS genes whose expression is modulated at the mRNA level (Table 1).

Table 1.

Genes validated for regulation by TNF-α at the mRNA level and their association with MetS.

Gene Name	Chromosome	Transcript Start (bp)	Transcript End (bp)	FPKM control	FPKM TNF-Alpha	log2 (fold_change)	P_value	Regulation	Function/association with T2DM/metabolic disorders	References
IL8	chr4	74606222	74609433	0.439751	210.927	8.90584	0.0025	UP	Inflammatory cytokine	[1]
PI3KC2A	chr3	1.79E+08	1.79E+08	0.000294261	1.03101	11.7747	0.01255	UP	Produces the catalytic subunit of PI3kinase	[2,3]
SKP2	chr5	36152144	36184142	6.30767	0	-inf	0.056	DOWN	It has been reported to be associated with beta cell mass reduction in T2D	[4,5]
ACACA	Chr17	35441925	35766902	0.00339245	1.2181	8.48809	0.0148	UP	Involved in lipid metabolism	[6]
SMAD3	Chr15	67358194	67487533	5.44313	13.0264	1.25893	0.0305	UP	Reported to have role in obesity-associated metabolism by differentially regulating MPK38 activity in diet-induced obese mice.	[7,8]
GLS	chr2	191745546	191830270	21.6716	52.9999	1.29019	0.00125	UP	Reported as marker for obesity and diabetes, impaired glutamine metabolism in adipose tissue is shown in individuals with obesity and diabetes.	[9], [10], [11]
HDAC4	chr2	239969863	240322706	0.585792	1.94399	1.73056	0.07455	UP	Inhibition of HDAC4 by knockdown improved glucose metabolism in DIO mice by suppression of hepatic gluconeogenesis.	[12], [13], [14], [15]
GADD45A	chr1	68150859	68154021	24.1894	69.5631	1.52395	0.00145	UP	It was shown to be overexpressed in beta cells of diabetic mice	[16,17]
SMARCA2	chr9	2015341	2193623	1.54997	0.000228	-12.7247	0.00615	DOWN	SNP's in SMARCA2 are reported in obesity	[18,19]
ZRANB2*	chr1	71318035	71546972	0.0014248	1.77736	10.2848	0.02515	UP	Reported in metabolic syndromes	[20,21]
ZFP36*	chr19	39897486	39900052	5.00106	13.9937	1.48447	0.011	UP	Altered in insulin resistance, obesity and artherosclerosis	[21], [22], [23]
ELAVL2/HuB*	chr9	23690017	23826282	0.56393	0.000919422	-9.26057	0.0222	DOWN	It is reported to be downregulated in diabetes mice with hyperglycemia.	[21,24]
MBNL1*	Chr3	151980404	152183569	5.17392
2.18553	-1.24328	0.25355	DOWN	A down regulated is reported in high glucose conditions in MIN6 cells.	[21,25,26]
MBNL3*	ChrX	131503342	131624106	8.23063	3.03267	-1.44041	0.0202	DOWN	Reported in diabetes	[21,27,28]
SNRPA*	Chr19	41256758	41271297	54.9611	26.2377	-1.06677	0.02275	DOWN	It is reported in obesity	[21,29]

^⁎RBPs which were validated and published before [21].

Differential expressions of IL8, PIK3C2A, and SKP2, ACACA, SMAD3, GLS, HDAC4, SMARCA2, GADD45A were validated by qPCR analysis. Other than these genes, expression level of ZFP36, HuB (ELAVL2), ZRANB2, SNRPA, MBNL-1 and-3, hnRNPR and hnRNPK was already validated in our previous report [30]. As shown in Fig. 2, IL8, PIK3C2A, ACACA, SMAD3, GLS, HDAC4 showed significant up-regulation whereas SKP2 showed a significant down-regulation compared to the control after 6 h of treatment with TNF-α. We did not find significant changes in the expression of SMARCA2 and significance level of GADD45A was not calculated as the data was obtained only from two independent experiments. Apart from these results, from as reported previously, we found significant upregulation of ZFP36, and ZRANB2, whereas HuB (ELAVL2), SNRPA, MBNL-1 and-3 showed significant down regulation upon TNF-α treatment. Expression level hnRNPR and hnRNPK did not show significant changes [30]. To test if the results observed in HEK293 cells are reproducible in skeletal muscle cells, which are actively responsive to TNF-α treatment, we repeated our experiments on a few of these genes in primary human skeletal muscle cells. As shown in Supplementary Fig. 1a–c, IL8 and PIK3C2A showed up-regulation where as SKP2 showed down-regulation, when treated with TNF-α for 6 and 24 h. These results confirmed that the RNA-Seq results of gene regulation could be extrapolated in TNF-α responsive primary skeletal muscle cells as well.

Fig. 2.

Validation of the data obtained from the NextSeq analysis. Change in expression of various genes was validated by qPCR analysis upon 6 h exposure of TNF-α. The results were normalized to the amount of 18s rRNA gene product. All reactions were done in triplicates. The statistical significance (by paired T-TEST) with p < 0.05 is depicted with an asterisk *. Standard error bars are calculated for each bar diagram.

Genes regulated by TNF-α are connected to various signaling and metabolic pathways contributing to MetS

To determine the biological significance of the differentially expressed genes, Gene Ontology enrichment or Over-Representation Analysis (ORA) was carried out using WebGestalt. The shortlisted 1218 genes were analyzed against all the unique genes (12,715), identified by the RNA-Seq analysis. Among these, the database could map 1162 genes against the 11,464 genes as the background. For the significance level, FDR<0.05 was chosen. Among the top most molecular functions of these genes, DNA-binding transcription factor activity was identified. Intracellular signal transduction, MAP kinase cascade, gene regulation etc. were among the top most biological processes carried out by these genes and the cytokine receptor interaction and TNF-signaling pathways were identified as the top most pathways (determined by KEGG). Detail of these findings are enlisted in Supplementary Datasheet 2.

To find out the MetS genes among the total gene pool regulated by TNF-α, we performed pathway analysis using the Ingenuity Pathway Analysis (IPA) tool as well as KEGG database for metabolic pathways. In IPA analysis, 1216 unique genes (out of 1218) were mapped, which were distributed into 123 different pathways with a significant enrichment cutoff of p < 0.05 (not shown). Considering a more stringent condition for pathway enrichment of p < 0.01, 55 different pathways were enlisted (Fig. 3a).

Fig. 3.

Pathway analysis of the regulated genes. (a) Significantly enriched 55 canonical pathways reported by IPA, using a p-value cut-off of <0.01 (threshold corresponding to 2). (b) Significantly enriched 22 MetS related canonical pathways reported by IPA, using a p-value cut-off of <0.05 (threshold corresponding to 1.3). Both in (a) and (b), the x-axis represents the canonical pathways for the dataset and the y-axis depicts the -log of p-value, calculated by Fisher's exact test right-tailed. The blue bars denote the –log (p-value) and the red dotted line denotes the ratio of the genes in the list to the total number of genes known to be present in that pathway. (c) A Pie-chart representation of the KEGG pathways of the percentage of mapped genes belonging to MetS related pathways (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

Out of these canonical pathways, we shortlisted the signaling and metabolic pathways, reported in connection to various metabolic disorders. We identified 204 MetS genes regulated at the mRNA level, among which, with a pathway enrichment cutoff p < 0.05, 163 MetS genes (out of 1218 genes, 13.4%) were identified, covering 22 pathways (Fig. 3b). The pathways and the contributory genes for each are enlisted in Supplementary Datasheet 3 and 4. The top 20 up and down regulated MetS genes are shown in Table 2 (and in Supplementary Datasheet 4).

Table 2.

Top 20 Up and Down regulated MetS genes.

Genes	FPKM Control	FPKM TNF-α
TNF	0	1.65827
CCL5	0	0.705567
TNFAIP3	0.385401	16.2795
HSPB8	0.805868	12.8964
ETS1	1.88644	25.6396
BMP2	1.55818	21.169
TNFRSF12A	6.33994	82.8965
NFKBIA	9.66627	122.506
MAFF	0.666177	7.12936
BIRC3	0.850546	6.04088
TJP2*	10.5992	61.0644
JUNB	5.01822	23.3707
NFKB1	6.28057	25.726
HSP90B1	283.504	1093.94
CALR*	215.179	701.741
CREB5	4.03798	12.6794
GADD45A*	24.1894	69.5631
F2RL1	26.7809	74.6718
GLS	21.6716	52.9999
ANXA1*	135.385	315.11
SOX3	96.9309	39.7044
PIK3R3	15.0785	6.16174
HES1	82.4111	32.7002
EGR1	141.134	55.6184
FDFT1	68.7745	26.5938
LRAT	48.9901	17.9232
INSIG1	37.3645	13.3097
SCD	53.9226	19.0706
AMOT	18.5961	6.14681
IGFBP5	4.46865	1.43933
FOS	43.5768	10.144
MAP2K6	1.83619	0.382339
MYH14	0.977115	0.000284365
AKAP2	1.09045	0.000278744
SMARCA2*	1.54997	0.000228991
LTB4R	0.528234	7.37E-05
HDAC4	0.505129	6.28E-05
BCL9	0.890363	0.000106116
INPP5J	0.95979	9.35E-05
THRB	1.02577	3.54E-05

^⁎Differentially expressed RBPs among the top 20 up and down regulated MetS genes

We also analyzed the pathways by KEGG database and short-listed them in connection to MetS. A total of 458 unique genes (out of 1218) were mapped in KEGG. Among the 291 canonical pathways, which were covered by these genes, 25 contributed to MetS (covered by 160 genes). Percentages of total number of genes participated in each of these pathways were calculated and represented in a pie-chart as shown in Fig. 3c. Details about these pathways and the genes participated under each of them are enlisted in Supplementary Datasheet 5.

The pathway analysis by both IPA and KEGG confirmed the TNF-α mediated regulation of the MetS genes.

Highly interconnected nodes among the TNF-α regulated MetS genes are transcriptional regulators

By using the STRING database (https://string-db.org), we have performed an interactome analysis of the MetS genes. Fig. 4 depicts a connection network between all the 204 MetS genes. The maximally interconnected nodes in these networks (Genes with ten or more connections, listed separately in Supplementary Datasheet 4) are mostly transcription regulators (including the SMAD family, NCOA family, NFKB family etc.) including nuclear hormone receptors. This data suggests that by mostly regulating highly connected transcription factors, TNF-α potentially influences the expression of a large array of genes, causing MetS. Other than the transcription factors, cytokine ligands and receptors, crucial enzymes (kinases, phosphatases), cell surface receptors, chromatin and histone modifiers are also found to be highly interconnected.

Fig. 4.

STRING database network in the MetS genes. All 204 unique MetS genes were mapped in the STRING database. The nodes in the network represent proteins. The node color depicts the level of interactors: query proteins and first shell of interactors (colored nodes); second shell of interactors (white nodes). The node content depicts the presence or absence of a known 3D structure: proteins of known or predicted 3D structure (filled nodes); proteins of unknown 3D structure (empty nodes). The edges in the network represent protein-protein associations. The known interactions: from curated databases (light blue) and experimentally determined (magenta). The predicted interactions: gene neighborhood (green); gene fusions (red); gene co-occurrence (dark-blue). Other interactions: text mining (yellow); co-expression (black) (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

We also performed IPA network analysis consisting of 70–140 genes in a network (for maximal representation of our genes of interest in a given network). A total of 104 MetS genes from the top 2 merged networks could be traced. The top scoring networks and the genes involved in each (including our genes of interest), are enlisted in Supplementary Datasheet 6.

EuRBPDB database reveals the identity of many MetS genes as RBPs, regulated by TNF-α

Previous report from our laboratory has predicted the interaction between RNA sequences and RBPs to regulate the MetS genes via modulating their alternative splicing [30]. Among the 32 predicted RBPs, 14 of them (ZRANB2, ZFP36, YTHDC1, SNRPA, RBMX, PTBP1, NONO, MBNL1, PUM1, FUS, ELAVL2, hnRNPR, hnRNPK, MBNL3) showed differential expression by TNF-α, according to the RNA-Seq analysis. We have validated the significant regulation of SNRPA, ZFP36, ELAVL2 and ZRANB2, published in our previous report [30] (Supplementary Datasheet 7). All these RBPs belonged to the canonical group with well-defined RBDs.

We retrieved the identity of the total number of RBPs among the differentially regulated genes, both in the total gene pool and among the MetS genes, with the help of the EuRBPDB database (a database for eukaryotic RBPs). This study revealed 1116 RBPs among the total number of differentially regulated genes with a two-fold of cut-off (as discussed in Fig. 1a); out of which 670 RBPs were identified as the canonical and 446 as the non-canonical RBPs (Fig. 5a). At a higher p-value≤0.05, 199 RBPs were identified among which 105 (out of 724 upregulated genes) and 94 (out of 599 downregulated genes) were up and downregulated respectively at the mRNA level (Refer to Fig. 1b). Excluding the overlapping RBPs whose different transcripts are represented in both the up and down regulated genes, we identified 178 unique RBPs (Supplementary Datasheet 7), comprising of 14.6 % of the total of 1218 unique genes (Fig. 1b). On the other hand, among the 204 MetS genes (Supplementary Datasheet 4), 22 RBPs (10.7 %) have been identified (Fig. 5b, Table 3 and Supplementary Datasheet 7). Some of the differentially expressed RBPs among the top 20 up and down regulated MetS genes are highlighted in Table 2.

Fig. 5.

Representation of the RBPs in the regulated genes.(a) Number of canonical and non-canonical RBPs among the up or down regulated RBPs with at least a two-fold regulation, obtained from the RNA-Seq analysis is plotted in a bar diagram. (b) Number of RBPs present in the up or down regulated genes, in total number of unique genes and in MetS genes are plotted. Their percentage of representation in each category is shown in parenthesis. Each category represents a regulation with a p-value<0.05 and least a two-fold regulation.

Table 3.

MetS genes as RBPs and their roles in alternative splicing and metabolic disorders.

Metabolic gene	Regulation with TNF-α in our RNA seq data	Type of RBP(based on EuRBPDB)	Function (ref excel for T2DM related)	Role in Splicing	Role in metabolic disorders
ANXA1	Up regulated	Non-canonical, no RBDs	Enzyme	Role in splicing not yet defined but reported as RNA binding [30]	Reduction in plasma annexin 1 is reported in human obesity [31,32].
PDIA6	Up regulated	Non-canonical, no RBDs	Enzyme	It regulates the kinases in turn regulates the splicing activity of XBP1 [33]	It is reported to regulate insulin secretion [33,34]
PPIB	Upregulated	Non-canonical, no RBDs	Enzyme	Not yet defined	Reported in diabetes, apoptosis of beta cells [35]
TJP2	Upregulated	Non-canonical, no RBDs	Kinase	Not yet defined	Reported in obesity related metabolic disorders [36]
GADD45A	Upregulated	Canonical RBPs	Gene involved in DNA repair	No direct link	Reported to have role in metabolic diseases [17]
HSP90B1	Upregulated	Non-canonical RBP	Heat shock protein	No direct link	Involved in lipid disorders [37]
PDIA3	Upregulated	Non-canonical RBP	Peptidase	Part of spliceosome complex	Linked to metabolic diseases [38]
PTPN23	Upregulated	Non-canonical RBP	Phosphatase	Helps SMN for its function in snRNP complex [39]	Reported in obesity related inflammation [40]
CALR	Upregulated	Canonical RBP	Transcriptional regulator	Not yet defined	Involved in T2DM [41]
PTPRG	Upregulated	Non-canonical RBP	phosphatase	Role not yet defined	Involved in metabolic disorders [42,43]
ACACA	Upregulated	Non-canonical RBP	Enzyme	Not yet defined	Involved in lipid metabolism [6]
POLR2A/RPB1	Upregulated	Canonical	Enzyme	Largest subunit of RNA polymerase and act as link between transcriptional control on splicing [44]	Reported in type 2 diabetic risk pathway [45]
PRMT1	Upregulated	Canonical	Enzyme	Several splicing regulators are the know targets of PRMT1 such as RBM15 thereby controlling splicing [46,47]	Upregulation is reported in metabolic diseases [48]
RPS6KA4	Up regulated	Canonical	Kinase(ribosomal kinase)	Not well documented	Reported in obesity related reproductive disorders [49]
AKAP13	Upregulated	Non-canonical RBP	Regulatory subunit of protein kinase A	Not well documented	Reported to be the hub genes for T2D and CVD [7]
ARHGEF2	Upregulated	Non-canonical RBP	Nucleotide exchange factor	No direct role	It is reported to be required for TNF-α mediated nuclear translocation of YAP/TAZ in white adipocytes, which is activated in adiposity [50].
PXN	Upregulated	Non-canonical RBP	Focal adhesion adapter protein	No direct role	Reported to have role in obesity, via Noggin protein in adipogenesis [51]
TLN2	Down regulated	Non-canonical RBP	Focal adhesion protein	No direct role	Related to obesity [52].
INPP5J	Down regulated	Canonical RBP	Phosphatase	No direct link	Reported in KEGG metabolic pathways
HNRNPK	Up regulated	Canonical RBP	Transcription regulator	Reported in splicing [53]	Reported in glucolipoxicity [54]
SMARCA2	Down regulated	Canonical RBP	Transcriptional regulator	Reported in splicing of several genes [55]	Reported in inflammation and insulin resistance [18]
ZNF423	Down regulated	Canonical RBP	Transcriptional regulator	Not defined	Dysregulated in obesity [56]

Taken together, these results show the presence of at least 10% RBPs among the MetS genes regulated by TNF-α. Some of these genes have been predicted to influence alternate splicing, contributing towards metabolic disorders.

TNF-α causes post-translational modification of RBPs

A previous report from our lab showed that TNF-α influences alternative splicing of many genes including several MetS genes [14] As the activity of some of the RBPs (SR proteins for example), which crucially regulates alternative splicing, is modulated by phosphorylation, we tested whether the phosphorylation status of the SR proteins is influenced upon TNF-α treatment. As shown in Fig. 6a, a time dependent exposure of TNF-α shows an initial increase in the phosphorylation status of various SR proteins (MW 75, 63, 40 and 30 KDa), which comes down almost to the basal level over time. The SR protein of 55 KDa did not show much change in the phosphorylation status compared to the control. This finding indicates that the activation status of these RBPs is influenced by TNF-α which modulates their phosphorylation level.

Fig. 6.

Effect of TNF-α on phosphorylation of SR and signaling proteins. A time course immunoblot analysis on phosphorylated (a) SR and (b) p38 (upper panel) and mTOR (lower panel) proteins. β-Actin is served as an internal control. Representative blots of n = 3 experiments are shown. Number below each lane for the P-p38 and P-mTOR shows the average fold changes compared to the control lane, normalized by the β-Actin expression. Arrows in part (a) depicts the corresponding molecular weight of the P-SR proteins.

Phosphorylation pattern of crucial kinases involved in various signaling cascades also changes upon TNF-α treatment

Since our RNA-seq data revealed the influence of TNF-α on various signaling cascades and since some of the same signaling cascades are known to influence the phosphorylation pattern of the SR proteins, we tested the activation status of two crucial kinases; p38 MAPK and mTOR, by checking their phosphorylation pattern. Involvement of these kinases in various metabolic syndromes and as the upstream kinases of the SR proteins are well-documented [38], [39], [40], [41]. As shown in Fig. 6b, with respect to a housekeeping gene, a time dependent exposure of 10ng/ml TNF-α shows an initial increase in the phosphorylation status of these two kinases until 1 h, which comes back at the basal level by 6 h. This change in the phosphorylation pattern correlates with the changes observed in the phosphorylation pattern of the SR proteins as seen in Fig. 6a. This data indicates that there is a strong possibility that by modulating the activity of crucial signaling pathways, TNF-α may influence the phosphorylation, thus activation status of various RBPs, which in turn may influence alternative splicing of the MetS genes.

Discussion

In our previous reports we have shown that TNF-α alters pre-mRNA splicing of the genes connected to MetS [14] and hypothesized that the differential binding of RBPs at the spliced sites of the pre-mRNA under TNF-α treatment, may influence alternative splicing of the MetS genes [30]. In this study, we focused on the differentially regulated RBPs under TNF-α treatment and the consequences in the metabolic syndromes. A predisposition to the pro-inflammatory cytokines such as TNF-α, may act as a major threat for developing these diseases [7,42,43]. Although there are reports suggesting the role of TNF-α in post-translational modifications of the downstream targets of various signaling pathways causing MetS [6,44], lacunae exist in understanding the precise mechanism of TNF-α-mediated regulations of genes connected to these diseases.

Here we present a report on global analysis to elucidate the regulation of the genes at the mRNA level by TNF-α, connected to MetS. In this study, we specially emphasized on the RBPs . We have identified changes in the phosphorylation status of certain RBPs upon exposure to TNF-α, which is correlated with the phosphorylation/activation status of some of the crucial kinases involved in signaling pathways. Together, these findings suggest that TNF-α regulates the expression of RBPs at various levels, potentially via modulating the activity of the upstream signaling pathways, eventually causing alteration of the splicing pattern of the MetS genes.

In this study we have identified 1218 regulated genes at the mRNA level (at least by two fold) by TNF-α, which are involved in major cellular functions including signaling pathways, cell-cycle regulations, DNA repair mechanisms and in gene regulation. In concordance with the previous studies, we subcategorized 204 genes, (including NFKB1, CREB5, JUNB, CCL5, TP53, PFKFB4, IL8, HYOU1, ACE2, PIK3C2A, VEGFA etc.), which could be connected to various metabolic syndromes (MetS genes). From the pathway analysis networks generated from our study, we have observed that the most significant pathways contributed by the MetS genes are the RAR-activation pathway, Wnt signaling, mTOR signaling, TNF-receptor signaling, MAPK signaling, PI3K signaling, Ras signaling, PTEN, HIPPO signaling pathways etc. which are already reported to be contributory to different metabolic syndromes [39,40,[45], [46], [47], [48], [49], [50]].

Post-transcriptional modifications of genes are orchestrated by more than 1542 RBPs in humans [51], [52], [53], [54]. An analysis with our data set using EuRBPDB, a comprehensive eukaryotic RBP database, revealed the identity of the RBPs, which represents ∼14% of the regulated total genes, and ∼10% of the MetS genes. As per our knowledge, this is the first report analyzing MetS genes for RBPs on a global scale under TNF-α treatment. Among the 22 MetS genes identified as RBPs, 13 belong to the non-canonical and 9 belong to the canonical RBPs. This is in concordance with the report that the majority of the RBPs lack defined RBDs [55]

We compared our non-canonical RBPs from the MetS gene dataset with the enigmRBPs; RBPs without a recognizable RBD or a known function in RNA biology (name coined by Beckmann et al) [56]. Two MetS genes from our dataset, PPIB and MARK2, are reported as enigmRBPs by the same group [56]. PPIB is reported as a player in the development T2DM due to islets dysfunction in diabetic mice [57] and MARK2, a polarity kinase (a Ser/Thr kinase which regulates cell polarity) is reported to have a central role in glucose metabolism and adiposity, however; the molecular mechanism is unknown [58]. Our report might reveal for the first time this protein's role in metabolic disorders via RNA binding.

MetS genes with enzymatic activity, such as ANXA1, PDIA6, ACACA, POLR2A and PRMT have been identified as RPBs according to the EuRBPDB database. So far, only a few reports exist of the enzymes, which show RNA binding properties [59]. Among these genes except ACACA (Acetyl-coA carboxylase), all have been reported as regulators of alternative splicing (Table 3). ACACA is an enzyme involved in fatty acid synthesis [60]. EuRBPDB database identifies ACACA as a RBP; however, its role in RNA biology has not yet identified, leaving a possibility that ACACA could classify as an enigmRBP.

Our EuRBPDB and RNA-Seq analysis reveals that an RBP Argininosuccinate synthase (a gene product of ASS1) which is also a metabolic enzyme with moonlighting activity [25], and connected to obesity related pancreatic cancer [61] and in hepatic lipid metabolism [62], is regulated transcriptionally by TNF-α. This finding endorses the ‘REM (RNA-enzyme-metabolite) hypothesis’ which predicts the role of enzymes in gene expression and intermediary metabolism by binding to RNA [25].

The EuRBPDB has also identified several MetS genes as RBPs, which act as transcriptional regulators. Some of the examples are hnRNPK, ZNF423, CALR and SMARCA2 (Table 3). SMARCA2/BRM plays a part in chromatin remodeling whose role in inflammation related to obesity and insulin resistance has been reported [63], Thus, our study is in line with the previous reports where chromatin-remodeling factors are shown as players of alternative splicing [64,65]. Post-translational modification of hnRNPK is reported in glucolipotoxicity in humans as well as in rats [66,67]. CALR, calcium regulatory protein, has reported to play a role in fibrosis of diabetic nephropathy [68]. Although the role of hnRNPK in alternative splicing is established [69], no direct connection of ZNF423 and CALR with AS has been reported so far.

There are 15 MetS genes from our dataset, which were identified as RBPs without any defined function in connection to alternative splicing (Table 3). A detailed mechanistic study needs to be further carried out to understand whether these proteins also contribute to the alternative splicing of the MetS genes.

While the IPA analysis of the entire gene set (among 1218 genes, 1211 could be mapped, Supplementary Datasheet 4) showed the presence of ∼ 18% transcription factors, IPA analysis of the 204 MetS genes, shows an enrichment of these factors (∼ 26%) (Supplementary Datasheet 4). To further confirm this, we searched the MetS genes in the Transcriptionally regulatory relationships unraveled by sentence based text mining (TRRUST) version 2 database (https://www.grnpedia.org/trrust/), and identified 52 transcription factors among the 204 MetS genes (Supplementary Datasheet 8) . These findings suggest that by modulating the expression of the highly connected transcription factors, TNF-α potentially influences the expression of a large array of genes, connected to MetS.

Interactome analysis has revealed that many of these transcription factors are highly interconnected (Fig. 4 and Supplementary Datasheet 4). This finding is in concordance with the study reported by Zelezniak, A et al, where they have identified a transcription factor regulatory network for metabolic genes in connection with T2DM [13].

According to our RNA-Seq data, combined with EuRBPDB, among the entire gene pool, 178 unique genes are RBPs. We have performed an interactome analysis of these RBPs using the STRING database, which shows strong interconnection between them (Supplementary Fig. S2). The RBPs such as hnRNPs, ELAVLs, SNRPA, ZRANB2 etc. are among the most connected in the network. Among them, SNRPA and ZRANB2 were reported from our lab as novel RBPs involved in metabolic syndromes [30]. Among the other highly inter-connected RBPs in the network, we found several MetS genes, such as PPIB, CALR, PDIA6, PDIA4, ARHGEF2, TJP2, GADD45A and SMARCA2. Thus, our analysis shows that the RBPs belong to the MetS gene family, cross talk to each other to coordinate their mode of actions to regulate metabolic syndromes.

RBPs such as SR proteins play pivotal roles in alternative splicing. Phosphorylation status of these proteins determines their state of their activation and localization inside the cell. We identified an initial increase up to 1 h in the phosphorylation status of several SR proteins, which decreases and, in some cases, comes back to the normal level upon longer exposure of TNF-α. Change in the phosphorylation status of the SR proteins and other splicing factors, for example SRp40 and SFRS10, has been reported in connection to metabolic diseases [70]

Post translational modifications of proteins by TNF-α are mediated via modulation of the activation of several signaling pathways [6,7]. Phosphorylation status of two such crucial mediators of cellular signaling; p38MAPK and mTOR, whose activities are known to be modulated by TNF-α [11,71], was tested. We observed that the phosphorylation/activation status of both of these kinases was initially increased and upon a longer exposure of TNF-α, comes back to the normal level. Similar pattern of regulation of p38MAPK and mTOR by TNF-α has been reported in connection with inflammatory bowel disease [71] and inflammation [72], respectively. SR proteins phosphorylation status has shown to exhibit a similar pattern as phosphorylated p38MAPK and mTOR. This observation strengthens the hypothesis that various signaling pathways could mediate the post-translational modifications of these RBPs, where these RBPs play as downstream targets.

The decrease in the phosphorylation status of the signaling kinases as well as the SR proteins upon longer exposure of TNF-α, after initial increase, could explain the activation of phosphatases in the cell. Our data has revealed that various phosphatases including PTPN23, PTPRG and INPP5J (Table 3) which contribute towards metabolic syndromes and also belong to the category of RBPs, are regulated by TNF-α. TNF-α is reported to stimulate PTPN23 in PTPN2 knockout cells which mimics inflammatory bowel disease (IBD) [73]. PTPRG overexpression has reported to be a reason for hepatic insulin resistance and has shown to be upregulated by TNF-α in inflammation mediated insulin resistance [74]. Understanding these phosphatases as RBPs in the context of metabolic syndrome will pave a way for a better treatment strategy.

Taken together, our results demonstrate the potential of the RBPs (including enigmRBPs and moonlighting RBPs) as novel drug targets to manage metabolic syndromes. Their roles in other post-transcriptional and epigenetic regulation of the MetS genes will be explored in future studies.

We compared our findings with the other available global databases for the genes involved in metabolic syndromes, especially with type II diabetes and the associated diseases. Databases named as Type 2 Diabetes Associated Complications (T2DiACoD; www.t2diacod.igib.res.in/) [75], which compiled their database from the T2DGADB (T2D Genetic Association Databases) [76] and the combined database from GoT2D(Genetics of type 2 Diabetes) and T2D-Genes (Type 2 Diabetes Genetic Exploration by next generation sequencing in multi-ethnic samples) consortia [77], was considered for comparison. From the T2DiACoD database, genes associated with the top five T2DM related diseases (namely atherosclerosis, diabetic nephropathy, neuropathy, retinopathy and cardio-vascular diseases) were compared with our MetS gene list. Among the 204 MetS genes identified by our work, 23genes are found to be reported in the context of one or more of the T2DM associated diseases, HDAC4 being the common one to be involved in all of these diseases (Supplementary Datasheet 9 and Supplementary Table 3). Among the 43 top ranked genes identified for T2DM associations [78], our list matches with 4 genes (NFKB1, NFKB2, ACACA and PIK3C2A). From the gene list known for anti-T2DM drug targets [78] (Insulin, Metformin, TZD, Sulfonylureas, GLP-1 receptor agonists, Amylin mimetic, Meglitinides, α-Glucosidase inhibitors etc.), our finding matches with 13 of the genes (Supplementary Datasheet 9 and Supplementary Table 4).

Since in this study we do not have the replicates of the RNA-Seq data, we strictly followed stringency in selecting the genes for interpreting our results. In all our analysis, genes regulated with a P value < 0.05 were selected with a minimum of 2-fold change cut-off. We also followed the stringency for pathway enrichment analysis, where p value <0.05 and <0.01 were used to identify the MetS genes. We compared our data with the published research work from other laboratories and databases available online. Moreover, we validated the RNA-Seq data in the laboratory by RT-PCR and qPCR analysis in different cell lines as shown in this manuscript and also published before by our group [14,30] (Table 4)

Table 4.

Transcriptionally significant genes reported as enigmRBP.

Gene names

DYNC2H1,RRP12,FNDC3B,CCDC9,CTTN,SPTBN1,HUWE1,ZC3H4,CALD1,IDH1,CAST,PPHLN1,PPIB,APEH,ASS1,CXorf57,EPPK1,FAM46A LBR,MARK2,PDIA4,PHF6,SOGA2,ZNF207,ZYX

Data availability

The sequencing data generated in this study has been deposited at NCBI Gene Expression Omnibus (GEO) under the accession number GSE182259.

CRediT authorship contribution statement

Jiss Maria Louis: Conceptualization, Methodology, Investigation, Validation, Writing – original draft, Formal analysis. Arjun Agarwal: Formal analysis. Sukanta Mondal: Formal analysis. Indrani Talukdar: Conceptualization, Writing – review & editing, Supervision, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.