Investigating the link between miR-34a-5p and TLR6 signaling in sepsis-induced ARDS

Abstract

Acute lung injury (ALI)/acute respiratory distress syndrome (ARDS) are lung complications diagnosed by impaired gaseous exchanges leading to mortality. From the diverse etiologies, sepsis is a prominent contributor to ALI/ARDS. In the present study, we retrieved sepsis-induced ARDS mRNA expression profile and identified 883 differentially expressed genes (DEGs). Next, we established an ARDS-specific weighted gene co-expression network (WGCN) and picked the blue module as our hub module based on highly correlated network properties. Later we subjected all hub module DEGs to form an ARDS-specific 3-node feed-forward loop (FFL) whose highest-order subnetwork motif revealed one TF (STAT6), one miRNA (miR-34a-5p), and one mRNA (TLR6). Thereafter, we screened a natural product library and identified three lead molecules that showed promising binding affinity against TLR6. We then performed molecular dynamics simulations to evaluate the stability and binding free energy of the TLR6-lead molecule complexes. Our results suggest these lead molecules may be potential therapeutic candidates for treating sepsis-induced ALI/ARDS. In-silico studies on clinical datasets for sepsis-induced ARDS indicate a possible positive interaction between miR-34a and TLR6 and an antagonizing effect on STAT6 to promote inflammation. Also, the translational study on septic mice lungs by IHC staining reveals a hike in the expression of TLR6. We report here that miR-34a actively augments the effect of sepsis on lung epithelial cell apoptosis. This study suggests that miR-34a promotes TLR6 to heighten inflammation in sepsis-induced ALI/ARDS.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13205-023-03700-1.

Introduction

Acute lung injury (ALI)/acute respiratory distress syndrome (ARDS) are inflammatory lung complications inflicting a significant impact on longevity (Bersten et al. 2002; Kim and Hong 2016). The incidence and mortality of ALI/ARDS vary worldwide, ranging between 20 and 40% (Pham and Rubenfeld 2017; Schissel and Levy 2006; Sigurdsson et al. 2013). ALI/ARDS is also being reported as a leading cause of fatality in the severe acute respiratory syndrome (SARS) coronavirus two infections resulting in COVID-19 illness (Yang et al. 2020; Zhang et al. 2021) with $33\%$ incidence rate and $45\%$ mortality rate in COVID-19 associated ARDS (Denson et al. 2021, p. 19; Tzotzos et al. 2020). ALI/ARDS is manifested by acutely reduced respiratory capacity leading to severe hypoxemia (Schissel and Levy 2006). Diffused alveolar damage is one of the hallmarks of ARDS, marked by compromised barrier function, vascular permeability with an accumulation of protein-rich edema and increased lung weight which ensues in reduced gas exchanges and lung compliance (Cardinal-Fernández et al. 2017; Englert et al. 2019). The onset of ARDS is multifactorial; however, it may be categorized as pulmonary (e.g., pneumonia, aspiration) and extrapulmonary (e.g., sepsis) (Raghavendran and Napolitano 2011). Sepsis is a majorly contributing etiology to ARDS incidence (Wang et al. 2021), with a mortality rate of $60\%$ observed in a retrospective study (Bellani et al. 2016; Mikkelsen et al. 2013). Sepsis-induced ARDS shows lower PaO₂/FiO₂ ratios, thus deteriorating respiratory function (Hu et al. 2020). ARDS induced by sepsis exhibits a hike in inflammation wherein macrophages dynamically switch phenotypes towards a pro-inflammatory state, thereby promoting a cascade of signaling expressing cytokines, such as TNF-$\alpha$, IL1-$\beta$, and IL-6 (Matthay et al. 2012). Macrophages express TLRs actively on their surface membranes for distinguished ligands like PAMP’s and DAMP’s (Billack 2006; Grassin-Delyle et al. 2020). TLRs are extensively studied receptors, well known for recognizing infectious microbial PAMP ligands (Newton and Dixit 2012). However, a continuous provocation to inflammation in sepsis is aided by non-infectious endogenous ligands generated host tissue damage (Gentile and Moldawer 2013; Roh and Sohn 2018). Excessive neutrophil infiltration at the site and distant (lungs) from the site of infection is observed in sepsis (Grommes and Soehnlein 2011). This neutrophil condition becomes detrimental to the host tissue due to the release of enzymes (Ginzberg et al. 2001). The damaged host tissue causes the release of endogenous DAMP ligands (Gentile and Moldawer 2013; Ginzberg et al. 2001; Roh and Sohn 2018). The systemic inflammatory response in sepsis is actively triggered by DAMPs like high mobility group box-1 (HMGB-1), host proteins, histones, extracellular cold-inducible RNA binding protein (eCIRP), ATP, HSPs, fibrinogens, which are released post tissue injury or senescence (Denning et al. 2019; Gentile and Moldawer 2013; Schaefer 2014; Sohn et al. 2012, p. 4). DAMPs ardently initiate different TLR signaling pathways in immune cells (Zhang and Mosser 2008). Infection triggers macrophages to polarize to either phenotype to attain homeostasis. This process is efficiently regulated by microRNAs (miRNAs), which get differentially expressed in specific conditions (Curtale et al. 2019). miRNAs are $22$ nt long short chain oligonucleotides that bind to the 3′UTR of the messenger RNA (mRNA) via seed region (Bartel 2009), leading to mRNA decay by de-adenylation in RNA-induced silencing complex (RISC) (Ha and Kim 2014; Huntzinger and Izaurralde 2011). In this way miRNAs regulate diverse processes like cell division and differentiation, senescence, inflammation, and homeostasis by regulating the mRNA concentrations (O’Brien et al. 2018). TLR signaling is actively modulated by miRNAs either by regulating the expression of TLR receptors or the downstream adapter molecules (Arenas-Padilla and Mata-Haro 2018).miRNAs are reported to regulate different genes in sepsis (Szilágyi et al. 2019). miRNAs are swiftly dysregulated in ALI/ARDS and sepsis (Cao et al. 2016; Ferruelo et al. 2018; Jiang et al. 2020). miR-34a is one such miRNA associated and prominently upregulated in sepsis inflammation (Abdelaleem et al. 2022; Chen et al. 2020; Goodwin et al. 2015; Oteiza et al. 2021; Zhang et al. 2020, p. 1). miR-34a is widely allied with promoting inflammation and senescence (Bai et al. 2011; Chafik 2018; Chang et al. 2007; Khan et al. 2020, p. 4; Long et al. 2018; Mohan et al. 2015; Navarro and Lieberman 2015; Syed et al. 2017; Velasco-Torres et al. 2019; Wang et al. 2016). Experimental data of ALI/ARDS depicts a significant analogy between miR-34a and lung inflammation (Chen et al. 2020; Cheng et al. 2018, p. 3; Cui et al. 2017; Du et al. 2012; Khan et al. 2020). miR-34a also polarize macrophage to M1 phenotypes by downregulating anti-inflammatory genes such as KLF4, SIRT1, and Notch (Chen et al. 2020; Khan et al. 2020; Yamakuchi et al. 2008; Zhang et al. 2015). STAT6, which IL4 strongly induces, is significantly associated with the anti-inflammatory macrophage phenotype (Gong et al. 2017; Li et al. 2020; Nepal et al. 2019; Tugal et al. 2013; Yu et al. 2019, p. 24). Contrary to this, TLR6 in conjugation with TLR2, forms a heterodimer receptor for the DAMPs arising due to host tissue damage continuously promotes inflammation (Eckmann 2006; Oliveira-Nascimento et al. 2012; Wang et al. 2019). Unlike sepsis or ARDS alone, sepsis-induced ARDS is a life-threatening condition broadly characterized by lung edema leading to severe hypoxemia. In sepsis-induced ARDS, miR-34a promotes TLR6 expression, conjugating with TLR2 forming an active receptor for the non-infectious endogenous ligands which continuously triggers TLR2-TLR6 signaling resulting in uncontrolled inflammation and organ dysfunction.

Our study primarily extracted mRNA microarray dataset of sepsis-induced ARDS and identified differentially expressed genes (DEGs). These DEGs were subjected to a weighted gene co-expression (WGCN) construction followed by trait-linked hub module detection. ARDS-specific 3-node miRNA feed-forward loop (FFL) gave us an insight into key miRNAs, transcription factors (TFs), and the highest degree subnetwork motif comprising miR-34a-5p, STAT6, and TLR6. Afterwards, we employed the molecular docking and molecular dynamics (MD) method against TLR6 to discover potential hits for ARDS which was further validated utilizing MD simulation and MM_PBSA studies. Cecal ligation puncture (CLP)-induced sepsis in mice showed a spike in the expression of miR-34a, which was also observed in-vitro. Overexpression strategies were adopted by using miR-34a mimic in-vivo, which showed a considerable deterioration of the lung injury symptoms. Contrary to this miR-34a inhibitors alleviated the lung injury symptoms and enhanced the expression of STAT6. miR-34a was found to be positively regulating TLR6, simultaneously antagonizing STAT6. Our experimental data permits us to suggest miR-34a as a probable therapeutic target in dealing with the menace of sepsis-induced ARDS. The novelty of our work lies in the microarray data extraction, extrapolation of trait-linked hub genes and FFL analysis to understand the complex molecular mechanism and signaling pathways associated with sepsis and ARDS.

Materials and methods

ARDS-specific expression profile collection and differential expression analysis

We accessed the National Center for Biotechnology Information (NCBI)-Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/) (Clough and Barrett 2016) for obtaining the ARDS-associated mRNA expression profile with “acute respiratory distress syndrome” and “ARDS” being utilized as suitable keywords. The search results were further trimmed down as per the inclusion and exclusion criteria mentioned in the supplementary methods section. Processed expression file (i.e., series matrix) of the selected dataset was downloaded, and probe ID mapping to their equivalent HUGO Gene Nomenclature Committee (HGNC) symbol(s) was performed utilizing the official sequencing platform library-based Bioconductor package followed by duplicate genes removal. Only officially approved HGNC symbols were retained as per the HGNC multi-symbol checker (https://www.genenames.org/tools/multi-symbol-checker/). A two-sample unpaired t test assisted in computing the p value and ${\log }_2(\mathrm{fold}\mathrm{change})$ values of each gene across two patient groups via limma (Ritchie et al. 2015). $\mathrm{p}\mathrm{value}<0.05$ and $\left|{\log }_2,(\mathrm{fold}\mathrm{change})\right|>0.2$ was set as the threshold for identification of DEGs. Lastly, Pigengene package (Foroushani et al. 2017) was utilized to remove any outlying or noisy DEGs.

ARDS-specific WGCN formation and hub module detection

We utilized the WGCNA package (Langfelder and Horvath 2008; Zhang and Horvath 2005) to establish an ARDS-specific WGCN and identify significant modules having a high correlation with age. The process of WGCN formation is discussed in supplementary methods section. The dynamic tree-cut algorithm was applied to find modules from the dendrogram branches. For possible merging of any modules with highly co-expressed gene patterns, module eigengene (ME) and dissimilarity measures between MEs (MEdiss) were taken into account. Any modules with alike expression profiles were merged by observing the ME dendrogram based on Pearson correlation. Module membership (MM) (i.e., correlation between ME and gene expression profile), standard intramodular connectivity (k.in) and average gene significance (GS) of genes in a given module were computed for all modules. The supplementary methods section discusses the criteria for selecting the hub module and their genes.

ARDS-specific 3-node miRNA-mRNA-TF network construction & topological analysis

ChEA v3.0 database (Keenan et al. 2019) was queried for obtaining significant ($\mathrm{p}\mathrm{value}<0.05$) human TFs interacting with ARDS-specific hub genes. The miRNAs relating to ARDS-specific hub genes and TFs (compiled from ChEA) were acquired from the miRWalk v3.0 database (Sticht et al. 2018). The process of picking miRNAs and FFL formation is discussed in supplementary methods section. The FFL’s topological properties were analyzed utilizing NetworkAnalyzer and CytoNCA Cytoscape plugins. The topological properties of the network are described by measurements of degree distribution $(\mathrm{P}(\mathrm{k})$), clustering coefficient ($\mathrm{C}(\mathrm{k})$), neighborhood connectivity (${\mathrm{C}}_{\mathrm{N}}(\mathrm{k})$), betweenness (${\mathrm{C}}_{\mathrm{B}}(\mathrm{k})$), closeness (${\mathrm{C}}_{\mathrm{C}}(\mathrm{k})$) and topological coefficient ($\mathrm{TC})$. We utilized these parameters to explain topological changes when the network is perturbed.

Preparation of the natural product database

The natural product library containing $6348$ compounds was selected from the MEGxp database, one of the most extensive NP chemical libraries available (AnalytiCon Discovery) for our high-throughput screening targeting the TIR domain of TLR6. These compounds were downloaded in “SDF format” and then converted to three-dimensional (3D) utilizing prepared ligand molecules in Biovia DS 2020.

Docking-based screening and MD simulation studies

A collection of $6348$ natural compounds from MEGxp database (https://ac-discovery.com) was screened to rapidly discover novel drug candidates against the TLR6 protein of TIR domain (PDB ID:4OM7) (Jang and Park 2014). The primary docking screening was carried out utilizing the LibDock program followed by CDOCKER docking module of Discovery Studio 2020. The predocking strategy; preparation of ligands and protein, filtration of compounds based on drug like properties, and binding site identification were performed as previously described (Jha et al. 2022b). Following the virtual screening of potential drug candidates, a $100\mathrm{ns}$ MD simulation using the Gromos96 43a1 force field parameters was run using the GROMACS 2019 software at constant temperature and pressure using an earlier protocol (Jha et al. 2022a, p. 2).

Molecular mechanics/Poisson‐Boltzmann surface area (MM/PBSA) binding free energy calculation

Binding free energies of all three drug-target complexes were computed via the MM/PBSA method utilizing the g_mmpbsa tool from last $20\mathrm{ns}$ stable trajectories (Kumari et al. 2014). The MM/PBSA calculation utilized the following equation:

$$\mathrm{\Delta }{G}_{\mathrm{binding}}=G_{\mathrm{complex}}-(G_{\mathrm{receptor}}+G_{\mathrm{ligand}})$$

where $\mathrm{\Delta }{G}_{\mathrm{binding}}$ represents the total binding energy of the drug‐target complex, G_receptor denotes the binding energy of free receptor, and G_ligand stands for the binding energy of unbound ligand.

Materials

Rabbit monoclonal anti-TLR6 from Abbkine, and mouse monoclonal anti-STAT6 from Biolegend, were employed for antigen detection. HRP-conjugate anti-mouse and anti-rabbit secondary antibodies from Santa Cruz Biotechnology were utilized. Merck's hematoxylin and Eosin were used to stain the mice’s lung section and check morphological changes. miR-34a inhibitors, mimics, and scramble were purchased from Qiagen.

Cecal ligation puncture model to induce ALI

C57BL/6 mice weighing 18–22 g and aged six to eight weeks were used. Pathogen-free environment with proper lighting conditions along with food and water ad-libitum. Committee for the Purpose of Control and Supervision of Experiments on Animals (CPCSEA), Government of India, followed guidelines to conduct the study with ethical approvals (File Number: 1358). Ketamine (100 mg/kg) and Xylazine (10 mg/kg) were administered to induce anesthesia in mice. miR-34a mimic and inhibitor and scramble are given at a concentration of 2 mg/kg body weight intraperitoneally. An 18-gauge needle was used to puncture the caecum already ligated at a 3/4 length from the terminal end. Through and through puncture was done induce infection in mice.

Histopathological analyses

Lungs harvested from CLP-induced septic mice were initially fixed in a $10\%$ neutral buffered formalin solution from Sigma and later embedded with paraffine. $5\mu \mathrm{m}$ thick sections are subjected to standard IHC protocol. TLR6 and STAT6 primary antibodies were utilized to probe the presence of respective antigens. Later, HRP conjugate universal secondary antibody from Vector Labs is used for respective primary antibody. DAB is utilized as a substrate for the HRP, and positively stained cells was visualized under light microscope (MEZI, Japan).

Statistical analyses

Two way-ANOVA was applied to check for the statistical significance of the quantified data. Mean values were considered with a $\mathrm{p}\mathrm{value}<0.05$.

Results

ARDS-specific mRNA expression profile retrieval and differential expression analysis

The supplementary results section mentions the selected dataset's details and preprocessing. A total of $883$ DEGs were identified corresponding to $\mathrm{p}\mathrm{value}<0.05$ and $\left|{\log }_2,(\mathrm{fold}\mathrm{change})\right|>0.2$. We were left with $771$ DEGs after eliminating noise/outliers. A 2D volcano plot showing an overview of significant (up versus downregulated) and nonsignificant genes in the GSE66890 dataset was shown in Supplementary Figure S1A. Supplementary Figure S1B shows the expression-based heatmap of top $10$ up and downregulated ARDS-specific DEGs. Among these $20$ DEGs, MME [${\log }_2\left(\mathrm{fold}\mathrm{change}\right)=1.1]$ and OLFM4 [${\log }_2\left(\mathrm{fold}\mathrm{change}\right)=-1.4$] were the most up and downregulated ones. The age of samples varied from $19$ to $92$, with the maximum patients aged $67$, $71$, and $77$. Among all the samples, $42\%$ were constituted by females, $56\%$ by males, and $2\%$ by other gender. Supplementary Figure S1C shows the principal component analysis (PCA) plot demonstrating the $771$ DEGs expression distribution across all samples. The expression variability of all DEGs’ was reduced dimensionally with respect to sample type leading to distinct cluster formations. The scree plot in Supplementary Figure S1D shows the $\%$ of variances accounted for by the top $5$ principal components (PCs).

ARDS-specific WGCN formation and hub module detection

Expression data of $771$ ARDS-specific DEGs and respective age data of all samples were given as input to WGCNA. In compliance with the WGCNA guidelines, $\mathrm{\beta }=7$ (corresponding to scale-free $R^2=0.8$) was picked for the construction of WGCN. Supplementary Figure S2 shows plots for $\mathrm{\beta }$ in consideration with SFT criteria. Figure 1A shows a total of nine unique color-coded modules (i.e., grey, pink, black, red, green, yellow, brown, blue, turquoise) ranging in size from $11$ to $229$. Figure 1B shows a GS barplot correlated with age across all modules where the blue module ($\mathrm{module}\mathrm{significance}=0.197$) was highly correlated. This relationship was highly significant ($p\mathrm{value}={9\times 10}^{-173})$ and correlated ($\mathrm{cor}=0.8$). Figure 1C shows the gene co-expression network (GCN) as a TOM plot among all module genes. Due to the short merging height observed in ME dendrogram, there was no need for merging modules. The grey module was rejected for further analysis since it contained unassigned genes. Figure 1D displays an eigengene adjacency heatmap for finding groups of highly correlated eigengenes termed “meta-modules”. MM versus GS correlation trend and p-values for eight modules is shown in Supplementary Table S1. GS versus k.in correlation trend and p-values for eight modules is shown in Supplementary Table S2. The blue module was chosen as our hub module based on GS versus MM and GS versus k.in correlation pattern.

Fig. 1.

A Hierarchical clustering dendrogram of $771$ ARDS-specific DEGs clustered based on dissTOM and nine color-coded modules obtained using Dynamic Tree Cut. The sizes of modules were as follows: black ($47$), blue ($117$), brown ($116$), pink ($42$), red ($59$), green ($60$), yellow ($90$), turquoise ($229$), and grey ($11$). B Barplot showing the distribution of GS and error bars across nine modules. C Representation of WGCN as a TOM plot. Genes within columns and their corresponding rows were hierarchically clustered by cluster dendrograms (displayed along the top and left side of the plot). Progressively darker and lighter red colors within the matrix indicate higher and lower topological overlap among genes. Dark colored blocks along the diagonal signify the modules. D Eigengene adjacency heatmap shows the relatedness of $8$ co-expression gene modules. These values were based on linear regression coefficients of determination which were obtained utilizing the ME values from WGCNA. Red indicates a correlation between $0.5$ and $1$ while blue indicates a correlation between $0$ and $0.5$

ARDS-specific 3-node miRNA-mRNA-TF network construction & topological analysis

ARDS-specific 3-node miRNA-mRNA-TF network comprised $88$ nodes and $433$ edges as shown in Fig. 2A. Among these edges, $175$ belonged to miRNA-mRNA pairs, $73$ belonged to TF-mRNA pairs, and $185$ belonged to miRNA-TF pairs. Among the total nodes, $14$, $14$, and $60$ belonged to ARDS-specific TFs, mRNAs, and miRNAs. The degree range of miRNAs, mRNAs, and TFs in the FFL varied from $2$ to $15$, $3$ to $47$, and $6$ to $35$, respectively. The degrees (average) of TFs, mRNAs and miRNAs were $18.42$, $17.71$, and $6$. Based on essential centrality measures such as degree, betweenness, and closeness, the most strongly connected subnetwork motif within FFL included one TF (STAT6), one miRNA (miR-34a-5p), and one mRNA (TLR6) as shown in Fig. 2B. Thus, these regulatory elements act as hubs that might be crucial in ARDS progression. The topological properties of the FFL can be seen in Fig. 3. Because only a fewer higher-grade node numbers have significant $C_{\mathrm{B}}$, $C_{\mathrm{C}}$ and $T_{\mathrm{C}}$ values, the number of higher regulating hubs that can monitor the cancer network is smaller. Therefore, the network is dominated by moderately low-degree nodes (proteins/genes), so the network’s structure, functioning, and control are mainly performed by these low-degree proteins/genes. However, the sparsely distributed few major/leading hubs can play an important role in maintaining and regulating the stability of the cancer network. Upon study, our network demonstrated that power-law distributions were followed for the likelihood of distribution of node degree ($P(\mathrm{k})$), clustering coefficient ($C(\mathrm{k})$) and neighborhood connectivity ($C_{\mathrm{B}}$(k)) with negative exponents (Eq. 1). This power-law function shows that the network displayed hierarchical-scale free action with the system-level organization of communities. The distribution of power laws for $P(\mathrm{k})$, $C(\mathrm{k})$, $C_{\mathrm{B}}(\mathrm{k})$ and $T_{\mathrm{C}}(\mathrm{k})$ against the distribution of degrees with negative exponents indicating more system-level organization of modules (Eq. 2). This specifies the assortive nature of the modules, indicating the likelihood of rich club forming. The hubs play a crucial role in maintaining the stability and properties of the network.

$$\left(\begin{array}{c}P(k)\\ C(k)\\ C_N(k)\end{array}\right)\sim \left(\begin{array}{c}{k}^{-\alpha }\\ k^{-\beta }\\ k^{-\gamma }\end{array}\right);\left(\begin{array}{c}\alpha \\ \beta \\ \gamma \end{array}\right)\to \left(\begin{array}{c}-0.53\\ -0.67\\ -0.41\end{array}\right)$$ 1

Fig. 2.

A ARDS-specific 3-node miRNA FFL network comprising $88$ nodes and $433$ edges. Pink-colored circular nodes represent ARDS-specific miRNAs, green-colored diamond nodes represent significant human TFs, and cyan-colored nodes represent ARDS-specific hub DEGs. B ARDS-specific highest-order subnetwork FFL motif comprising one TF (STAT6), one miRNA (miR-34a-5p), and one hub DEG (TLR6)

Fig. 3.

Topological properties of PPI network. The degree distribution (p(k)), clustering coefficient (C(k)), average neighborhood connectivity ($C_{\mathrm{N}}$(k)), betweenness centrality ($C_{\mathrm{B}}$(k)), closeness centrality ($C_{\mathrm{C}}$(k)) and topological coefficient ($T_{\mathrm{C}}$(k)) as a function of k. The power law with exponent $\alpha$, $\mathrm{\beta }$, $\mathrm{\gamma }$, Φ, $\mathrm{\mu }$ and $\mathrm{\epsilon }$ for p(k), c(k), $C_{\mathrm{N}}$(k), $C_{\mathrm{B}}$(k), $C_{\mathrm{C}}$(k) and $T_{\mathrm{C}}$(k) respectively

The significance of hubs, and their control mechanisms, is defined by centrality estimation, such as $C_{\mathrm{B}}$ betweenness, $C_{\mathrm{C}}$ closeness and $T_{\mathrm{C}}(\mathrm{k}$) topological coefficient.

$$\left(\begin{array}{c}{C}_B(k)\\ \begin{array}{c}{C}_C(k)\\ T_C(k)\end{array}\end{array}\right)\sim \left(\begin{array}{c}{k}^{-\lambda }\\ \begin{array}{c}{k}^{-\delta }\\ k^{-\epsilon }\end{array}\end{array}\right);\left(\begin{array}{c}\gamma \\ \alpha \\ \beta \end{array}\right)\to \left(\begin{array}{c}1.12\\ 0.11\\ -0.05\end{array}\right)$$ 2

The power-law behaviour of these $C_{\mathrm{B}}$, $C_{\mathrm{C}}$ and $T_{\mathrm{C}}$ (centrality measurements) is again confirmed and validated by applying the power-law fitting method of Clauset et al. (Clauset et al. 2009) where p values are greater than $0.1$ and goodness of fit greater than $3.5$. The box-and-whisker plots showing the relative expression distribution of these $14$ ARDS-specific hub genes in sepsis alone and sepsis with ARDS patient samples are shown in Supplementary Fig. S3.

Virtual drug screening and molecular docking

4OM7 is a homodimer x-ray structure of TLR6's TIR domain. The active site was created with a sphere radius of $16.2{\mathrm{A}}^{\circ },$ and coordinates were $X=9.36636$, $Y=3.00432$, and $Z=17.2434$. Two docking algorithms in Biovia DS 2020, LibDock, and CDOCKER were utilized to assess the binding site's applicability. Initially, $6348$ natural compounds were docked into the predicted active site of TLR6's TIR domain, out of which $2342$ compounds were filtered utilizing this technique. The top $10\%$ of compounds ($234$) were then subjected to ADMET and TOPKAT filters to eliminate non-drug-like molecules and focus solely on drug-like molecules. Following that, $73$ drug-like compounds were docked utilizing the CDOCKER module of Biovia DS 2020 into the binding site of TLR6 protein to select compounds based on their ability to form favorable interactions and high CDOCKER energy with the active site. Supplementary Table S3 lists the top ten hit compounds with the highest CDOCKER energies and the amino acid residues involved in the interaction. Figure 4 depicts the receptor-ligand interactions of the selected top-3 hit compounds.

Fig. 4.

2D and 3D representation of TLR receptor residues interacting with A Lead 1, B Lead 2, and C Lead 3. Panel A, B, and C show 2D and 3D representations of the Toll-like receptor (TLR) receptor residues interacting with three different lead compounds, named Lead 1, Lead 2, and Lead 3. The 2D representations show the chemical structures of each lead compound and the specific residues in the TLR receptor that interact with the compounds. The 3D representations show the TLR receptor bound to each lead compound, indicating the orientation and spatial arrangement of the ligand and the receptor residues

MD simulation

Utilizing MD simulation, it is possible to examine the binding stability of protein–ligand docking complexes. To verify the stability and intermolecular contacts between the protein and ligands over a time of $100\mathrm{ns}$, the complexes from docking study of three natural chemicals (Lead 1, 2 and 3) that interact with TLR protein were examined in this study. The Gromacs 2019 interface was utilized to extract MD simulation trajectories, and the root-mean-square deviation (RMSD), root mean square fluctuation (RMSF), number of hydrogen bonds, and radius of gyration ($R_{\mathrm{g}}$) were utilized to describe the simulation findings. In MD simulation, the RMSD calculates the average distance caused by an atom's movement over a given time frame relative to a reference time frame. The RMSD value makes it possible to determine whether the simulation has equilibrated. Fluctuations between 1 and 4 A° are perfectly acceptable with respect to a reference protein structure. Backbone RMSD showed acceptable fluctuations in our three protein–ligand docking complexes, whereas Lead 1 showed least fluctuation (Fig. 5A). RMSF is required to characterize and determine the protein chain's local conformational change. Figure 5B and C show RMS fluctuations of chain A and chain B of TLR protein. Chain B showed greater fluctuation than chain A throughout the MD simulation, which showed greater flexibility in chain B.

Fig. 5.

A RMSD, B RMS fluctuation of chain A, C RMS fluctuation of Chain B, D intermolecular hydrogen bonds between proteins, E hydrogen bond between leads and TLR receptor and F radius of gyration of apoprotein, TLR (black), lead 1 (red), lead 2 (green) and lead 3 (blue) during 100 ns md simulation

Furthermore, the number of H-bonds that played an important role in the stability of the protein–ligand complex was determined utilizing simulation trajectories. Figure 5D shows that all the complexes were stable, whereas Fig. 5E shows several H-bonds persist throughout the simulation between protein and top $3$ lead compounds. Lead 1 and lead 2 had maximum number of hydrogen bonds (five), whereas lead 3 had a maximum $4$ hydrogen bonds during MD simulation. The $R_{\mathrm{g}}$ was utilized to calculate the compactness of a protein structure. Figure 5F shows that lead $2$ had the lowest value of $R_{\mathrm{g}}$.

Binding free energy

The average binding free energy and its corresponding components obtained from the MM/PBSA calculation of all drug-target complexes are listed in Table 1. The table's energetic profile of all the lead molecules shows that all complexes are highly stable, although the TLR-Lead1 offers the highest stability. In the TLR-Lead1 complex, specific interaction in the form of hydrogen-bond or salt-bridge is almost three times higher than the non-specific interaction—Van der Waals interaction indicates a highly stable interaction between drug-target complex, and it maintains throughout the simulation. In the TLR-Lead2 complex, the same trend has been observed in TLR-Lead1. In the TLR-Lead3 complex, the Van der Waals interaction has dominated electrostatic interaction.

Table 1.

MM/PBSA binding free energy profile

S. no.	Compound	∆EVDW (kcal/mol)	∆EELE (kcal/mol)	∆GSOL-PB (kcal/mol)	∆GSOL-NP (kcal/mol)	∆Gbind-PBSA (kcal/mol)
1	TLR-Lead 1	$-24.42$	$-78.04$	$29.22$	$-2.49$	$-75.73$
2	TLR-Lead 2	$-10.40$	$-31.02$	$8.47$	$-1.07$	$-34.03$
3	TLR-Lead 3	$-35.03$	$-19.57$	$18.29$	$-3.05$	$-39.37$

Expression of miR-34a is hiked in sepsis-induced ALI deteriorating lung function

miR-34a is found culpable in promoting inflammation by regulating the expression of different genes. The involvement of miR-34a in the progression and worsening of lung injury has been proved by different research accounts [49, 69–71]. In-silico analyses of clinical data sets of sepsis-induced ARDS suggest a hike in miR-34a expression. We also tested the CLP-induced septic mice lungs to check the expression of miR-34a compared to normal control. Figure 6B shows a considerable hike in the septic mice lung, indicating a probable correlation between miR-34a and lung injury symptoms. For this purpose, we established a sepsis-induced lung injury model. Inhibitors and mimics were injected intraperitoneally to check miR-34a involvement in sepsis-induced pathophysiology. Scramble is utilized as a control for comparison. H & E staining exhibits the changes in the lung tissue from different groups (Fig. 6A). The CLP group combined with scramble & mimic showed a thickening around the alveoli, representing hyaline membrane formation. Also, neutrophilia in interstitial and alveolar space are profusely contributing to lung injury. These observations were limited in CLP scramble group. However, it got more prominent with the addition of miR-34 mimic. Contrary to this, much less damage to the lung tissue is seen in the CLP miR-34a inhibitor group (Fig. 6A).

Fig. 6.

Sepsis-induced expression of miR-34a. A H&E staining of lung tissue from the CLP-induced septic mice. Inhibitor & mimic of miR-34a and scramble is used to create different groups to check their effect on the lung pathology by light microscope. B CLP-induced sepsis gives rise to miR-34a expression in the lung tissue. Student t-test is applied to test the significance. $\ast p<0.05$ ($n=4$ animals)

miR-34a elevates TLR6 expression in septic lung tissue

Major finding of in-silico analyses is the probability of positive interaction between TLR6 and miR-34. TLR6 work in conjugation with TLR2 forming a heterodimer for ligand recognition, which are supposed to be the derivatives of host tissue along with infectious PAMPs. Extrapulmonary sepsis puts the pro-inflammatory cytokines into circulation and the non-infectious DAMP ligands. Thus, non-infectious endogenous ligands are also part of sepsis-induced lung injury’s overall inflammatory response. Based on the in-silico findings, we put forward the idea that TLR6 expression increases in septic mice which miR-34a regulates, and it actively participates in reaching the overly inflamed lung condition. TLR6 expression was probed in different groups belonging to scramble, miR-34a mimics, and inhibitors. According to the in-silico findings, TLR6 is markedly expressed in the CLP-induced septic lungs (Fig. 7A). However, adding miR-34a mimics to the group enhanced TLR6 staining unequivocally (Fig. 7A). Contrary to this observation, less and dampened expression is seen in CLP + miR-34a inhibitors, strengthening our idea of a probable relationship between miR-34a and TLR6 (Fig. 7A). Positively stained areas were quantified by imageJ software and graphically represented by Prism software (Fig. 7B).

Fig. 7.

miR-34a regulates the expression of TLR6. A IHC stained septic mice lung tissue sections to see the expression of TLR6 antigen. Different groups were created to see variability in expression. B Two-way ANOVA quantified TLR6 expression. $\ast \ast \ast p<0.001$, $\ast \ast \ast \ast p<0.0001$ ($n=4$ animals)

miR-34a accelerates the sepsis-induced apoptosis, which exacerbates lung injury

Apoptosis is a dynamic phenomenon significantly contributing to the pathophysiology of sepsis. In the acute phase, it has successfully fostered disease severity. In response to the changed inflammatory settings cells undergo apoptosis by altering the integrity of DNA. To check whether miR-34a augments sepsis-induced apoptosis, we checked the septic mice lung tissue by TUNEL assay, which detects the fragmented DNA. The terminal deoxynucleotidyl transferase enzyme utilizes the fragmented DNA of cells as a template. TUNEL-positive cells are observed as green, fluorescent dots in lung sections. Based on staining, we compared the different treatment groups. More TUNEL-positive cells were visible in the CLP group as compared to the sham control figure (Fig. 8A). Further, the number of positive cells showed a remarkable hike by adding miR-34a mimic. TUNEL-positive cells were quantified by ImageJ software and graphically represented to test the significance (Fig. 8B). The results help us derive that miR-34a profoundly augments the apoptosis induced by CLP-induced sepsis in mice lungs. (Figs. 8A, B).

Fig. 8.

miR-34a promotes apoptosis in CLP-induced septic lungs. A TUNEL staining of the paraffin-embedded septic mice lungs from different treatment groups. B Quantification of the positively stained cells by ImageJ software by two-way ANOVA. $\ast \ast \ast p<0.001$, $\ast \ast \ast \ast p<0.0001$ ($n=4$ animals)

Discussion

The major finding of the present study is that miR-34a significantly augments the expression of TLR6 in sepsis-induced lung injury. Simultaneously miR-34a checks the STAT6 signaling to promote inflammation in sepsis. Also, we report that epithelial cell apoptosis is accelerated by miR-34a, the leading cause of worsening lung injury symptoms, ultimately leading to mortality. FFLs can engineer recurrent network motifs to enhance gene regulation's robustness in mammalian genomes (Bo et al. 2021). A three-node FFL can be extended to form a four-node or five-node FFL by adding further interaction links. 3-node miRNA FFL comprises a TF, a miRNA, and a common gene target with miRNAs dominating or driving the whole network (Sun et al. 2012). Finally, in-silico efforts reveal the potential of natural compounds (Lead1, Lead2, and Lead3) as anti-sepsis candidate molecules against TLR6's TIR domain.

Sepsis is a systemic inflammatory setting that arises due to dysregulated immune response to microbial infection (Gyawali et al. 2019; Rello et al. 2017). Lungs are supposed to be most vulnerable and responsive to the changed inflammatory milieu (Hotchkiss et al. 2016, 1999). This rapidly induces alterations in lung functioning, which further aggravates disease symptoms. Thus, sepsis-induced lung injury is a major concern contributing to the global disease tally. Infectious PAMP ligands are the leading cause of initiation of any microbial infection; however, the continuous trigger for dysregulated immune response includes non-infectious endogenous DAMP ligands (Kawasaki and Kawai 2014; Takeda and Akira 2004). Extrapulmonary infection like sepsis generates a systemic inflammatory response flagged by the expression of pro-inflammatory mediators and their secretion into circulation, along with the DAMPs (Humphries et al. 2018). This results to the expression and activation of TLR receptors in the lungs appreciably (Radstake et al. 2004; Ramírez Cruz et al. 2004). TLR4-MyD88 dependent signaling leads to the activation of TF NFκB, which finally culminates in the expression of pro-inflammatory cytokines IL-1β, IL6, and TNFα (Kawasaki and Kawai 2014; Radstake et al. 2004). This further expresses different TLRs on the membranes of APCs and endothelial cells. The TLR signaling releases cytokines, which act as an attractant for the neutrophils to the site of infection and the injured tissue (Kremserova et al. 2016). Thus, neutrophils are inevitable in reducing microbial infection, but it also helps improve the symptoms of lung injury (Ginzberg et al. 2001; Grommes and Soehnlein 2011). However, scenario changes in the condition of systemic inflammation as in sepsis. PAMPs and DAMPs actively contribute to the continuous provocation of the immune response (Denning et al. 2019; Gentile and Moldawer 2013). This leads to lung neutrophilia, which becomes detrimental to host tissue. In this study, we propose that non-infectious endogenous ligands actively trigger the TLR6 signaling in sepsis and may significantly contribute to inflammation.

Our study reports the expression of TLR6 in CLP-induced septic mice lungs. TLR6 signaling culminates in the expression of pro-inflammatory genes. Lungs harvested from septic mice were subjected to IHC analyses to look for the expression of TLR6. Profuse TLR6 staining was observed in the CLP-only group. As per our hypothesis, we found miR-34a augmenting the expression of TLR6 in the lung sections from the CLP + miR-34a mimic group. TLR6 staining was quantified by imageJ software. The data obtained were significant enough to correlate miR-34a and TLR6 expression. Further, we claim that miR-34a aggravates lung inflammation in septic mice via TLR6 signaling and the canonical pathways. miRNAs are differentially expressed in reported cases of sepsis, underscoring their role in the progression or resolution of sepsis (Ghafouri-Fard et al. 2021). miR-34a is well acknowledged for hiking inflammation by polarizing macrophages to the M1 phenotype (Cheng et al. 2018; Khan et al. 2020) simultaneously down-regulating the anti-inflammatory mediators (Shetty et al. 2017). The expression of miR-34a is notable in the LPS-induced lung injury model. However, inhibition of miR-34a successfully mitigates lung injury characteristics overall, positively affecting survival (Chen et al. 2020). We also checked the expression of miR-34a in CLP-induced sepsis mice. Results indicate a marked increase in expression compared to the control group. Based on the results we can say that miR-34a plays a vital role in the sepsis pathophysiology.

Epithelial cell apoptosis is one of the contributors to ALI (Martin et al. 2005). miR-34a mediates p53 acetylation to accelerate epithelial cell apoptosis, thus actively promoting lung injury (Shetty et al. 2017). However, the inhibition of miR-34a in LPS-treated HUVEC cells slowed down apoptosis (Zhang et al. 2020). Apoptosis is commonly observed in sepsis where epithelial cell apoptosis is reported to contribute to the severity of lung injury in its acute phase. Contrary to this in the latter phase, immune cell apoptosis paramount’s to immune suppression (Guinee et al. 1996; Hattori et al. 2010; Hotchkiss et al. 2003; Kitamura et al. 2001; Lang and Matute-Bello 2009; Luan et al. 2015). Intra-tracheally instilled LPS induces apoptosis in epithelial and endothelial cells (Kawasaki et al. 2000; Ogata-Suetsugu et al. 2017). One of the attributes of apoptosis is DNA damage/ fragmentation (Maeyama et al. 2001; Perl et al. 2007). This fragmented DNA is a template for terminal deoxynucleotidyl transferase enzyme in TUNEL assay. In the present study, septic mice lung sections from the CLP group were subjected to a TUNEL assay to affirm sepsis-induced apoptosis. The findings were in coherence with the previously reported work. CLP group showed positive TUNEL staining convincing enough to pronounce that sepsis-induced inflammation is the underlying cause of epithelial cell death. Inclusion of miR-34a mimics had a compounding effect on apoptosis-induced cell death. Our study not only reiterates the effect of sepsis on lung epithelial cells but also ascertains the role of miR-34a in exacerbating inflammation. The strategy adopted by miR-34a in elevating inflammation is checking the expression of anti-inflammatory genes. Different research groups have reported miR-34a targeting the genes that foster the anti-inflammatory M2 phenotype of macrophages (Khan et al. 2020; Tundup et al. 2014; Zhang et al. 2020). STAT6 is fairly acclaimed for its involvement in modulating the phenotype of macrophages to the M2 state (Cai et al. 2019; Gong et al. 2017; Hu et al. 2018; Li et al. 2021). The expression of STAT6 accelerates the uptake of apoptotic neutrophils which is the key to the resolution of inflammation (Nepal et al. 2019). In a separate study, STAT6 has been validated as a target of miR-34a (Tundup et al. 2014). To analyze the contribution of the miR-34a-STAT6 axis in the deterioration of symptoms in septic mice, we checked the expression of STAT6 with conditions of inhibition and overexpression of miR-34a. Inhibition of miR-34a increased staining of STAT6 in the septic lung sections. Contrary to this observation, adding miR-34a mimic to the CLP group swiftly checks the expression of STAT6. Based on the translational study of STAT6 with caveat of miR-34a, we can afford to say that miR-34a actively targets STAT6 as one of the strategies to accelerate inflammation in sepsis-induced ALI/ARDS.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ALI

Acute lung injury

ARDS

Acute respiratory distress syndrome

SARS

Severe acute respiratory syndrome

miRNA

microRNA

RISC

RNA-induced silencing complex

DEGs

Differentially expressed genes

WGCNA

Weighted gene co-expression network analysis

FFL

Feed-forward loop

TFs

Transcription factors

MD

Molecular dynamics

CLP

Cecal ligation puncture

NCBI

National Center for Biotechnology Information

GEO

Gene expression omnibus

HGNC

HUGO Gene Nomenclature Committee

ME

Module Eigengene

MM

Module membership

GS

Gene significance

MM/PBSA

Molecular mechanics/Poisson‐Boltzmann surface area

RMSD

Root mean square deviation

RMSF

Root mean square fluctuation

$R_g$  

Radius of gyration

TLR6

Toll-like receptor 6

STAT6

Signal transducer and activator of transcription 6

Author contributions

MJK: conceptualization, methodology, software, formal analysis, data curation, writing—original draft, writing—review & editing, investigation, validation, visualization. PS: conceptualization, methodology, software, formal analysis, data curation, writing—original draft, writing—review & editing. PJ: software, data curation, writing—original draft, writing—review & editing. AN: software, writing—review & editing. MZM: software, data curation, writing—original draft, writing—review & editing. GB: writing—review & editing. BK: writing—review & editing. KP: software, data curation, writing—review & editing. SA: resources, writing—review & editing. MC: writing—review & editing. RD: writing—review & editing, supervision, project administration. IKS: writing—review & editing, resources, supervision, project administration. MAS: writing—review & editing, resources, supervision, project administration.

Funding

This research work did not receive any external funding.

Data availability statement

The data underlying this article is available in NCBI-GEO at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE66890, and can be accessed with GSE66890.

Declarations

Conflict of interest

The authors declare that they have are no conflict of interest.

Human ethics and consent to participate

Not applicable.

Consent for publication

Not applicable.