Skip to main content
Download PDF

From CFTR to a CF signalling network: a systems biology approach to study Cystic Fibrosis

BMC Genomics volume 25, Article number: 892 (2024) Cite this article

Abstract

Background

Cystic Fibrosis (CF) is a monogenic disease caused by mutations in the gene coding the Cystic Fibrosis Transmembrane Regulator (CFTR) protein, but its overall physio-pathology cannot be solely explained by the loss of the CFTR chloride channel function. Indeed, CFTR belongs to a yet not fully deciphered network of proteins participating in various signalling pathways.

Methods

We propose a systems biology approach to study how the absence of the CFTR protein at the membrane leads to perturbation of these pathways, resulting in a panel of deleterious CF cellular phenotypes.

Results

Based on publicly available transcriptomic datasets, we built and analyzed a CF network that recapitulates signalling dysregulations. The CF network topology and its resulting phenotypes were found to be consistent with CF pathology.

Conclusion

Analysis of the network topology highlighted a few proteins that may initiate the propagation of dysregulations, those that trigger CF cellular phenotypes, and suggested several candidate therapeutic targets. Although our research is focused on CF, the global approach proposed in the present paper could also be followed to study other rare monogenic diseases.

Peer Review reports

Background

Cystic fibrosis (CF) is the most common life-limiting autosomal disease in the Caucasian population, affecting about 162,000 patients worldwide, of which 105,000 are diagnosed [1]. It is caused by mutations in the CFTR gene encoding for the cystic fibrosis transmembrane conductance regulator (CFTR) protein, a chloride ion channel expressed at the apical membrane of polarized epithelial cells [2]. More than 2000 mutations in CFTR have been reported, but the deletion of the F508 amino-acid (F508del) is present in 70% of the mutated alleles in the Caucasian population, and most of the mutations lead to compromised transepithelial anion conductance [3]. Various organs are affected in CF, but the most severe symptoms are in the lungs, where the defective chloride transport leads to the dehydration of surface mucus, chronic bacterial infection, and inflammation, causing lung tissue damage and ultimately, respiratory insufficiency.

However, CF symptoms not only result from the loss of CFTR-mediated anion conductance, but also from perturbations of other CFTR-dependent biological functions [4]. Indeed, CFTR belongs to a protein-protein interaction (PPI) network [5, 6]. In the present paper, we assume that proteins interacting with wt-CFTR, but not with F508del-CFTR, may experience functional or abundance perturbations due to the loss of their interaction with wt-CFTR. This, in turn, could result in partial gain or loss of activity, and downstream propagation of functional dysregulations. In agreement with this idea, studies on CFTR -/- knockout mice [7], CFTR -/- knockout piglets [8], and cell lines in which CFTR is inactivated by the CRISPR/Cas9 technology [9] have reported that the absence of CFTR affects cell signalling and transcriptional regulation. The dysregulations may explain various and unrelated cellular phenotypes, including uncontrolled pro-inflammatory response [10], unbalanced oxidative stress with increased reactive oxygen species [11], impaired epithelial regeneration [12], or perturbation of cell junctions and cytoskeleton [13].

The most studied cellular process affected in CF is ionic transport across the plasma membrane. Ionic transport dysregulation has been an important issue in CF, as it can be mechanistically linked to the loss of CFTR chloride function and to typical symptoms such as mucus viscosity in a direct manner [4]. In this work, we chose to focus on signalling pathways because the mechanistic link between dysregulated pathways and the loss of a functional CFTR protein is poorly understood. Among others, pathways related to inflammation and the immune response seem to play a critical role in CF phenotypes and have been frequently reported [14, 15].

Our main contribution is to provide a network-based model that explores how the absence of CFTR can be mechanistically linked to signalling perturbations, ultimately leading to certain CF cellular phenotypes.

Systems biology approaches provide tools for building network models to reason on complex systems. Subsequent topological analysis or dynamic mathematical models performed on these networks allow us to study how different biological components interact to produce observed phenotypic properties. Another contribution of the article is that topological analysis of the proposed networks may also suggest and prioritize candidate therapeutic targets whose modulation could attenuate CF cellular phenotypes.

Systems biology approaches have seldom been implemented in monogenic diseases, but have been widely used in cancer, often referred to as a network disease [16], where intricate processes contribute to the emergence of unexpected and often non-intuitive phenotypes. Very few contributions have been devoted to systems biology approaches of CF. Previous studies have focused on the construction of the CFTR interactome that distinguishes PPI networks involving wt-CFTR and those involving the most frequent mutant F508del-CFTR [5, 17]. The latter led to the construction of a navigable knowledge map, the CyFi-MAP, that integrates all proteins known to be involved in the processing, maturation, retention and degradation of wt-CFTR and F508del-CFTR. Although the CyFi-MAP represents a key contribution for the problem of rescuing F508del-CFTR, this map does not address the question of how CFTR is related to CF signalling perturbations. Besides, other studies highlighted links between CFTR and signalling pathways involved in the disease (see [14] for a review), but they did not provide a global view of how CFTR relates to these molecular perturbations and their resulting phenotypes. Recently, transcriptomic data have been produced to identify differentially expressed genes in CF. These genes were connected within a PPI network, based on information available in PPI databases [18]. Although this network comprises genes that are consistent with current knowledge in CF, it does not contain CFTR, which prevents understanding the functional link between CFTR and the differentially expressed genes, or with CF cellular phenotypes.

To overcome these limitations, we propose an approach that focuses on signalling dysregulations. More precisely, combining publicly available transcriptomic data for CF and for control patients and signalling pathway information compiled in biological pathway databases, we built a comprehensive signalling network, referred to as the CF network hereafter. As detailed below, we connected CFTR to this network based on PPI information from the CyFi-MAP. Analysis of the CF network topology led to formulate hypotheses about key proteins and molecular mechanisms that functionally link CFTR to major CF cellular phenotypes. Finally, it also suggested candidate targets whose modulation may counteract these phenotypes.

Results

Global approach to building the CF network

In systems biology, various networks can be built to represent different types of biological information, such as gene regulatory networks, genetic interaction networks, signal transduction networks, metabolic networks, PPI networks, or disease networks. There is no universal technique that can be followed to construct networks, and the choice of their representation needs to be adapted to the question of interest. In the present study, we wish to establish a link between the absence of CFTR and the overall signalling dysregulations leading to the cellular phenotypes that characterize CF. Therefore, we chose to build a CF network focusing on the signalling pathways that are perturbed in the disease, and where dysregulations in one pathway may affect other pathways. In order to avoid potential bias in the CF literature, we adopted a data-driven approach based on publicly available transcriptomic studies. We considered that gene expression data could capture major dysregulated signalling pathways in CF cells, although some CF phenotypes might arise from biological events that are not detectable in the transcriptome of CF cells. In other words, we assume that the variation of gene expression can reflect changes in protein abundance and in protein activity. We are aware that this assumption has its limits [19], and that ideally, proteomic (and phospho-proteomic) data would be a natural choice for studying dysregulated pathways [20]. However, proteomic data detect only a few thousand proteins, whereas transcriptomic data cover the entire genome, making it possible to detect a wider range of dysregulations. Additionally, there is much more publicly available transcriptomic data for CF than proteomic data. Overall, we chose to use transcriptomic data as a proxy to detect perturbations that occur at the proteomic level in the cell, as it is frequently done in systems biology modelling [20].

Our study relies on a meta-analysis of public transcriptomic datasets for CF respiratory epithelial cells and their Non-Cystic Fibrosis (NCF) control counterparts, allowing the identification of the signalling pathways dysregulated in CF. Based on information available in pathway databases, these dysregulated pathways share many common proteins, which allowed to connect them into a network. As detailed below, CFTR was absent from this network, because it did not belong to any of the differentially expressed signalling pathways. However, we observed that several proteins of the network were known to interact with wt-CFTR either directly or via a single intermediate protein, but were not known to interact with F508del-CFTR. This important result was consistent with the proposed assumption that wt-CFTR interactors, which do not interact with F508del-CFTR, may be perturbed in CF and initiate the propagation of dysregulations within the CF network.

Figure 1 summarizes the global approach followed in the present study.

Fig. 1
figure 1

Global approach followed to build the CF network: A meta-analysis of CF transcriptomic data allowed the identification of dysregulated pathways and the construction of the corresponding CF network. This network comprises known CFTR interactors that can be viewed as source nodes initiating the propagation of dysregulations

Full size image

Selection of publicly available transcriptomic data

Many transcriptomic studies have been performed in CF over the last 15 years [14]. However, these data suffer from a few limitations that are obstacles to improve our understanding of CF. First, they consider a wide range of cell types, including native nasal or bronchial cells, primary cultures of these cells, whole blood, peripheral mononuclear cells, leukocytes, or immortalized cell lines. Therefore, comparison between studies to identify common key molecular determinants can lead to inconsistent results. Then, compared to studies on more common diseases such as cancer, most of CF transcriptomic studies have very few samples per condition (disease and control), decreasing the statistical power of these datasets when analysed alone. Finally, these studies rely on various experimental biological models and transcriptomic technologies which rarely lead to consistent results between studies [21], particularly when the analyses are performed at the gene level.

To try to overcome these limitations, we focused on studies considering only samples from human Airway Epithelial Cells (hAEC hereafter), i.e., bronchial, tracheal, or nasal cells. Indeed, functional modifications in these cells are expected to reflect some of the most severe symptoms in the lung. We included studies of cell lines or primary cultures, in order to gather a statistically significant number of samples, because as shown in Table 1, each dataset comprises a very limited number of samples. We also focused on studies on the F508del mutation, for which most data are available. We discarded two studies [22, 23] that provide transcriptomic data for other mutations, because the corresponding cells could display disparities with respect to F508del cells. We retrieved from the literature all the CF transcriptomic studies with publicly available data that matched these criteria (see Methods section), which led to 10 CF transcriptomic datasets shown in Table 1.

Table 1 Summary of the 10 studies included in the meta-analysis. The table indicates the first author of the publication, year of publication, tissue of origin, sampling method, cell origin, culture conditions, technology, number of disease (CF) and control (NCF) samples, dataset reference identifier, and publication reference
Full size table

The studies are still heterogeneous in terms of tissue sample (bronchial, tracheal or nasal), culture condition (cell-line or primary culture) and transcriptomic technology (micro-array or RNA-Seq). However, we kept the 10 studies in order to improve statistical significance, because the numbers of samples per condition are very small in all studies: the median number of samples was 5 for disease (CF) and control (NCF) conditions.

Meta-analysis of transcriptomic studies at the level of biological pathways

The most straightforward way to analyse transcriptomic data is to identify Differentially Expressed Genes (DEGs), and to search for biological pathways enriched in these DEGs. This approach failed in the present meta-analysis, because the number of DEGs common to at least 3 out of 7 studies was too small to be enriched in any pathway, even though many reference pathway databases were considered (the Hallmark gene sets from the MSigDB Database [31], the Pathway Interaction Database (PID) [32], the KEGG database [33]). In fact, it has become clear that, in complex diseases, identification of pathway dysregulations based on DEGs is not optimal and does not provide robust results [34].

Therefore, the meta-analysis was conducted at the pathway level. Many methods have been proposed to capture pathway dysregulations when they do not appear clearly based on enrichment from lists of DEGs [35,36,37,38]. In the present study, we used the Gene Set Enrichment Analysis (GSEA) approach [39]. GSEA was performed separately on each dataset. Signalling pathways were identified as over-activated or under-activated in hAEC CF cells based on the complete expression matrix of CF and NCF samples, and taking into account the expression level of all genes belonging to the same pathway. We used pathway definitions provided by the KEGG pathway database [33], because this database provides graphical pathway representations that also include phenotypes, which helped the analysis of the CF network, as detailed in Analysis of the CF network section. Because our study focuses on signalling, we tested 131 signalling pathways from the KEGG pathway database, and Differentially Expressed Pathways (DEPs, hereafter) were identified according to an adjusted p-value lower or equal to 0.25, as detailed in Identification of Differentially Expressed Pathways (DEPs) section. The number of up- and down-regulated pathways for each dataset is provided in Table 2

Table 2 Number of detected genes, CF and NCF samples, tested pathways and dysregulated pathways per study with a corrected p-value < 0.25 and a \(|log2FC| > 1\) threshold at the gene scale
Full size table

The analysis of DEPs showed that 15 of the 134 biological pathways tested were differentially expressed in at least 3 studies. However, a closer analysis highlighted discrepancies between studies. As shown in the heatmap presenting the GSEA Normalised Enrichment Score (NES) (Fig. 2), for these 15 common DEPs, the 10 datasets can be gathered into 2 subgroups: subgroup 1 comprising 7 datasets in which common DEPs tend to be up-regulated, while they tend to be down-regulated in subgroup 2 comprising the 3 other datasets. This appeared somewhat puzzling. Our hypothesis is that datasets belonging to subgroups 1 or 2 arise from studies in which the differentiation media used for the primary cultures did not favour the same cell type, and therefore, should not be analysed together.

Fig. 2
figure 2

Heatmap of the GSEA Normalized Enrichment Scores (NES) of the biological pathways differentially expressed in at least 3 studies. The datasets can be clustered in two subgroups based on their NES: Subgroup 1 and Subgroup 2, respectively in agreement and in contradiction with CF physio-pathology. Black boxes around the tiles represent the pathways significantly differentially expressed in the corresponding dataset

Full size image

This was confirmed by the Saint-Criq (UNC) and Saint-Criq (SC) datasets (see Table 1), belonging respectively to subgroups 1 and 2, where it was shown that the UNC and SC differentiation media (two common differentiation media used on CF and non-CF epithelia) significantly impact cell lineage in primary cultures of CF hAEC, and consequently, the resulting transcriptomic profiles [30]. In this study, it was shown that the UNC medium promoted differentiation into club and goblet cells, while the SC medium favoured the growth of ionocytes and ciliated cells. Consistent with this result, the Ling transcriptomic dataset, which belongs to subgroup 2, was also obtained from primary cultures of CF and NCF airway epithelia that were differentiated into ciliated pseudostratified airway cells [29]. Datasets from subgroup 2 appeared in contradiction with the main CF phenotypes. In particular, the TNF-\(\alpha\) signalling pathway or NF-\(\kappa\)B signalling pathway are down-regulated in this subgroup, although the over-activation of these pathways is a well-known feature of CF disease. Therefore, we only considered the 7 datasets belonging to subgroup 1 for further analysis.

In this subgroup, the transcriptomic analysis appears to be highly consistent, since among the 15 DEPs common to at least 3 studies, 5 are up-regulated in CF vs NCF samples in 4 studies (NOD-like receptor signalling pathway, Cytosolic DNA-sensing pathway, Cytokine-cytokine receptor signalling pathway, and Regulation of actin cytoskeleton), 2 are up-regulated in CF vs NCF samples in 5 studies (Osteoclast differentiation and Toll-like receptor signalling pathway), and the IL-17 signalling pathway is up-regulated in CF vs NCF in 6 studies.

Overall, the 15 DEPs common to at least 3 studies are in agreement with various known aspects of CF disease, which confirms that our analysis did capture relevant information about CF. In particular, besides the TNF-\(\alpha\) and NF-\(\kappa\)B signalling pathways well known to be up-regulated in CF, the IL-17 pathway contributes to CF lung disease [40], the differentiation of osteoclast is perturbed in CF [41], the Toll-like receptor signalling pathway modulates function, inflammation and infection of lung in CF [8, 42, 43], and CFTR plays a role in cell junction and actin cytoskeleton organization [13].

Building the CF network

The 15 individual DEPs of the KEGG database provide interesting information about what is dysregulated in CF, but a lot of these pathways are partially redundant and show a high overlap of genes and interactions, indicating that they are highly intertwined. A dysregulation in one of these pathways will have a consequence in another pathway. To study the connection between them, we propose to merge them into a single network called the CF network.

The DEPs were extracted with the OmniPathR package [44] and curated, as described in Up-dating Omnipath DEPs pathways section. The rules that were used to build and clean this network are detailed in Network building and pruning section. The network, comprising 330 nodes and 529 interactions, is not fully connected: it contains one main component including 317 nodes connected by 517 interactions, and two small additional components that are non connected to the main component, and called unconnected components hereafter (see Fig. 3). The shiny application https://njmmatthieu.shinyapps.io/shinyapp provides fast and convenient browsing of the CF network with different viewing options. The overall network can also be accessed as a Cytoscape session, in the sysbio-curie/CFnetwork_cystoscape GitHub repository for further analysis.

Fig. 3
figure 3

The CF network. (A) The main component comprises 317 nodes connected by 517 interactions and two small unconnected components shown in (B): the two unconnected components correspond to the TGF\(\beta\) and the JAK-STAT signalling pathways. The cellular phenotypes triggered by the sink nodes of the two components are surrounded by black contours

Full size image

Identification of CFTR interactors in the CF network

It is striking to note that CFTR does not belong to any of the 15 DEPs, and therefore, is not part of the network. In fact, CFTR is present in only 7 biological pathways of the KEGG database (ABC Transporters, cAMP signalling pathway, AMPK signalling pathway, tight junction, Gastric and acid secretion, pancreatic secretion and bile secretion), but these pathways did not belong to the DEPs. To answer the question of how CFTR may be related to the CF network, we relied on our hypothesis that CFTR interactors may be perturbed by the absence of a functional CFTR protein and propagate downstream molecular perturbations. More precisely, we searched in the CF network for proteins reported to be involved in protein-protein interactions (PPI) with wt-CFTR but not with F508del-CFTR, according to the CyFi-MAP [5]. We found that 4 direct interactors of wt-CFTR but not of F508del-CFTR belong to the CF network, namely CSNK2A1, PRKACA, SYK and TRADD. Furthermore, 4 additional proteins (EZR, SRC, PLCB1/3) present in the network also interact with wt-CFTR, but not with F508del-CFTR, through a single intermediate protein. Figure 4 shows these 8 proteins, the 4 intermediates, and their interactions with CFTR. The presence in the CF network of 8 first or second neighbours in the CFTR interactome is in itself an interesting result, because it constitutes a strong argument in favour of our assumption that CFTR interactors may propagate functional dysregulations into the network.

Fig. 4
figure 4

CFTR interactors in the CF network: Known protein-protein interactions involving CFTR interactors in CFTR PPI

Full size image

Analysis of the CF network

Extensive interpretation of this large network, which contains rich but complex information, is beyond the scope of the present paper. However, we will investigate how analysis of its topology can help tackle the two questions of interest: how the absence of the CFTR protein at the membrane leads to CF cellular phenotypes, and how therapeutic targets can be suggested from this network.

Topological description of the CF network

The final CF network comprises 330 proteins and 529 interactions. Interestingly, CFTR interactors are present only in the main component, because according to the CyFi-MAP, it would not have been possible to link CFTR to proteins of the two small unconnected components without adding a large number of intermediate nodes. One of the two unconnected components contains 10 proteins and 9 interactions, and corresponds to cascades of the JAK/STAT signalling pathway. The other contains 3 proteins and 3 interactions, and corresponds to a cascade of the Transforming Growth Factor Beta (TGF\(\beta\)) signalling pathway. In the present section, we will focus on the main component of the CF network, and the two unconnected components will be discussed in Analysis of the unconnected components in the CF network section.

The topological description of the main component will be organized around three types of remarkable nodes: (1) the source nodes, i.e., CFTR first or second neighbours that were used to connect CFTR to the network, as described in Building the CF network section; (2) the sink nodes, i.e., the nodes from which no edge leaves in the network, and whose activation finally triggers their associated phenotypes (for example, transcription factors are typical sink nodes); (3) the hubs, i.e. the nodes with high betweenness centrality, through which the flow of information that passes is high. Figure 5 illustrates where these remarkable nodes stand within the network’s topology.

Fig. 5
figure 5

Illustration of propagation of dysregulation and remarkable nodes in the CF network: the source nodes (orange disks) are wt-CFTR interactors (that do not interact with F508del-CFTR) or connected to wt-CFTR interactors via a single intermediate protein (magenta circles). Nodes with high betweenness centrality (purple disks) are proteins through which much information flows within the network. Sink nodes (blue triangles) modulate their corresponding phenotypes

Full size image

Source nodes and initiation of dysregulations

According to the CyFi-MAP, 8 first or second neighbours of wt-CFTR interactors whose interactions are lost with F508del-CFTR are present in the CF network: CSNK2A1, EZR, PLCB1, PLCB3,PRKACA, SRC, SYK and TRADD. In the absence of CFTR, these 8 proteins can be viewed as source nodes that may initiate dysregulations that subsequently propagate within the network and finally reach the sink nodes (see Fig. 5).

Perturbations of some of these source nodes in CF cells, or their role in CF cellular phenotypes, are sustained by various studies:

  • CSNK2A1, also known as CK2 (casein kinase 2), is strongly overactivated in CF vs wild-type cells [45].

  • Cellular levels of TRADD are controlled by its lysosomal degradation in a wt-CFTR-dependant manner, and this regulation is lost with F508del-CFTR and G551D-CFTR [46].

  • SRC was shown to be overexpressed and overactivated in CF cells [47].

  • PLCB3 is a known CF modifier gene, for which the loss of function S845L variant is associated with a mild progression of the pulmonary disease and a reduction of Pseudomonas aeroginosa-induced IL8 release. This indicates that PLCB3 plays a role in the inflammation phenotype in CF [48].

  • The active form of Ezrin (EZR) is mainly located in the apical region of wild type airway epithelial cells, while in their CF counterparts, it is diffusely expressed in its inactive state in the cytosol [49, 50].

  • The SYK and PRKACA kinases play key roles with respect to CFTR, since the former negatively regulates the amount of CFTR at the membrane through phosphorylation at Y512 [51], while the latter is a well-known regulator of the CFTR chloride channel conductance [52], but their implication as propagators of dysregulations has not been investigated yet.

Sink nodes and CF phenotypes

There are 35 sink nodes in the main component of the CF network that are reached from each of the 8 source nodes. The full list of sink nodes and their associated phenotypes are given in Table S1 of Additional file 1. Among them, we can cite:

  • NFKB1, NFKB2, RELA and RELB are part of the NF-\(\kappa\)B complex, a transcription factor that can be activated by various stimuli such as cytokines, oxidant radicals, bacterial or viral products. It controls the expression of pro-inflammatory genes, and is related to various phenotypes, including inflammation and cell survival/proliferation.

  • FOS and JUN are two subunits of the AP-1 transcription factor activated by the MAPK signalling pathways, and are associated with inflammation and proliferation phenotypes.

  • CASP3 and CASP7 caspases are the effectors of apoptosis.

  • CASP1 is a caspase known to be the effector of pyroptosis, a highly pro-inflammatory cell death mechanism.

  • 10 sink nodes belong to the regulation of actin cytoskeleton pathway, including ACTN4, ARPC5, PFN, MYL12B and VCL. These nodes are associated to various phenotypes related to cytoskeleton, including focal adhesion, adherens junction, and actin polymerisation.

  • IRF1, IRF3, IRF5, and IRF7, that are members of the IRF family of transcription factors involved in the innate immune response phenotype, and controlling expression of Type-1 interferons upon viral infection.

Importantly, the phenotypes associated to these sink nodes have already been described in the CF context. In particular:

  1. 1.

    The NF\(\kappa\)B and AP-1 transcription factors are complexes of sink nodes that mediate inflammation, the most studied phenotype of CF disease. In addition to the well-known activation of NF\(\kappa\)B in CF, AP-1 is one of the downstream transcription factors of the MAPK pathway that was shown to be activated in CF [53, 54], as shown in Fig. 6.

  2. 2.

    Controversial results were reported about apoptosis in CF epithelial cells. Some studies showed defective susceptibility of CF cells to pro-apoptotic stimuli [55, 56], while others observed increased apoptosis [26, 57,58,59]. All agree that apoptosis is dysregulated in CF.

  3. 3.

    The dysregulation of actin cytoskeleton in CF is well documented, with a disorganized actin cytoskeleton, absence of actin stress fibres [49, 60, 61], and disrupted tight junctions [62, 63].

  4. 4.

    Finally, most of these phenotypes are related to the innate immune response, and various works indicate a dysfunction in the innate immune response of CF patients [42, 64, 65].

Fig. 6
figure 6

Extract from the CF network showing the TRADD protein connected to the TNF-\(\alpha\) signalling pathway, and to 5 other sink nodes, including FOS and JUN which form the AP-1 transcription factor, downstream of the MAPK cascade. The cellular phenotypes triggered by the sink nodes are surrounded by black contours. Note that TRADD is connected to the 35 sink nodes, but only part of the nodes downstream of TRADD in the network are represented

Full size image

Betweenness centrality and flow of information

In a network, the betweenness centrality (BC) of a node is the number of shortest paths that pass through that node. This measure is a way of detecting the amount of influence a node has over the flow of information in a network. Nodes with high BC, referred to as hubs, may provide interesting therapeutic targets, because their inhibition may efficiently reduce the propagation of information within the network [66]. Therefore, we calculated the BC for all nodes of the CF network, as detailed in the Methods section. All nodes were then ranked according to this measure, and Fig. 7A displays the histogram of the BC score. Interestingly, most proteins have a BC score below 3000, and only a very limited number of proteins have a BC score above 6000 (ARHGEF12, IKBKE, LSP1, PIK3KC1, PYCARD, RAC1, TRAF2, TRAF3, TRAF6). The list of the top 30 proteins is provided in the Table S2 of Additional File 1. Among them, PI3KCA could be an interesting therapeutic target candidate and is discussed in the next section.

Fig. 7
figure 7

A Histogram of the betweenness centrality measures for all nodes in the CF signalling network; (B) Number of sink nodes to which each of the 8 source nodes are connected

Full size image

Biological insights from the topological analysis

A simple path analysis of the network shows that several source nodes may contribute collectively to the emergence of the CF phenotypes, which illustrates the complexity of the disease. Indeed, while the source nodes PRKACA, EZR and CSNK2A1 are upstream of a limited number of sink nodes, PLCB1/3, SRC, SYK and TRADD are upstream of the 35 sink nodes, i.e., there exists a path from each of these 6 source nodes to each of the 35 sink nodes (Fig. 7B).

For example, TRADD is known to be up-regulated in CF [67] and to participate in the uncontrolled inflammation (see Fig. 6). Interaction between wt-CFTR and TRADD enhances the degradation of TRADD, which controls the activity of this pathway, as demonstrated by Wang and colleagues [46]. This direct interaction is lost with F508del-CFTR, which may contribute to the dysregulation of TNF-\(\alpha\) and NF-\(\kappa\)B signalling pathways in CF. However, up-regulation of TRADD could also contribute to the inflammation phenotype through another route, by inducing over-activation of the MAPK pathway, and in particular of AP-1, one of its output transcription factors. In addition, as shown in Fig. 8, our network suggests that other source nodes than TRADD could also initiate dysregulation of the inflammation phenotype because they are also connected to the NF-\(\kappa\)B sink node. Among these sources, we can cite: (1) SYK, which would be consistent with its role in inflammation processes shown in other diseases [68, 69]; (2) PLCB1/3, which are consistent with previous studies reporting PLCB3 as a key modulator of IL8 expression in CF bronchial epithelial cells [70]; (3) CSNK2A1, whose hyperactivity could contribute to activation of NF-\(\kappa\)B by enhancing the phosphorylation and degradation of NFKBIA.

Fig. 8
figure 8

Subnetworks of the CF network illustrating the connections between the source nodes TRADD, SYK, PLCB1/3, and CSNK2A1 and the sink node NFKB1

Full size image

Overall, the number of source nodes and routes that may contribute to inflammation in CF illustrates the challenge posed by its modulation, in order to reduce the related clinical symptoms. Various anti-inflammatory drugs have been recently evaluated in clinical trials [71], but none of them target the source nodes of the present study. Our hypothesis is that these source nodes could be interesting candidate targets in CF.

Among the source nodes, three proteins have emerged as interesting candidate targets, because pharmacological modulators are available for them so that they could be experimentally tested on cellular CF models: SYK, PLCB1/3, and SRC. To our knowledge, none of these proteins have been proposed as a target for CF. Interestingly, SYK emerged as a potential target for the treatment of numerous diseases. Many inhibitors are known for this kinase, which would allow evaluation of their potential anti-inflammatory effect in CF cells. These inhibitors include one marketed drug (Fostamatinib), but other inhibitors are currently under investigation in clinical trials for a range of indications [72]. Importantly, since SYK is connected to the 35 sink nodes, its inhibition may also contribute to the modulation of other CF phenotypes than inflammation. In particular, it could modulate CF phenotypes associated to the 35 sink nodes and mentioned in Topological description of the CF network section, such as dysregulations in apoptosis, cytoskeleton or innate immune response. Similarly, our network suggests that PLCB3 could be an interesting target for inflammation in CF. This is consistent with the fact that PLCB3 silencing in CF bronchial epithelial cells exposed to Pseudomonas aeruginosa, reduces the expression of IL-8 chemokine [70]. The U73122 PLC inhibitor could be an interesting pharmacological tool to further evaluate this strategy. As in the case of SYK, PLCB3 is connected to the 35 sink nodes, which means that its inhibition may also improve other CF cellular phenotypes. Consistent with this idea, it was shown that treatment with an SRC inhibitor, another of the 6 source nodes upstream of the 35 sink nodes, decreased the inflammatory changes and improved cytoskeletal defects in F508del human cholangiocytes [73]. These results underline the interest of SRC as a candidate target for CF.

Besides source nodes, candidate therapeutic targets can be searched among hubs in the network, i.e. among the best ranked proteins according to the BC score. Additional arguments can be invoked to highlight the best candidates. In particular, the fact that a protein is known in the literature to play a role in the disease, and that pharmacological modulators (or even better, marketed drugs) are available to allow experimental validation, are important criteria. In line with these ideas, PI3KCA appears as an interesting candidate target. Indeed, several inhibitors are known for this kinase, including the marketed drug Alpelisib, which would allow experimental tests in CF models. It has been suggested as a candidate target in CF based on its role in many signalling pathways implicated in CF lung pathogenesis [74]. Our study provide a more comprehensive case in favour of PI3KCA as a candidate target in CF. First, this kinase belongs to the best ranked proteins with respect to the BC score (see Fig. 7A), which means that a significant part of the signalling dysregulations flow through its node in the network. In addition, PI3KCA is connected through the network to the 35 sink nodes, which means that its inhibition may not only modulate inflammation, but also all the CF cellular phenotypes related to the sink nodes.

Analysis of the unconnected components in the CF network

As mentioned in Topological description of the CF network section, Fig. 3 shows that the CF network comprises two small unconnected components that are part of the TGF\(\beta\) and JAK/STAT signalling pathways. Contrary to source nodes of the main component, dysregulation of the source nodes of these unconnected components (namely the 4 interleukins IL2, IL21, IL4 and IL6 for one component, and TGF\(\beta\) for the other) cannot be explained by the absence of CFTR in a direct manner, because they are not linked to CFTR within a single network. However, activation of a sink node of the main component may modulate the expression of a source node in an unconnected component, affecting the activity of this unconnected component. For example, activation of the AP-1 transcription factor (a sink node of the main component) due to activation of the MAPK pathway in the main component, regulates the expression of TGF\(\beta\). This example shows how dysregulations in one pathway may have consequences in other pathways of the CF network, even if they are not connected, illustrating once again the complexity of the disease. We propose that phenotypes arising from the two unconnected components could be defined as downstream phenotypes, as opposite to upstream phenotypes arising from dysregulations of the main component (discussed in Topological description of the CF network section).

The JAK-STAT component mediates various cellular processes, including cell growth and apoptosis, but the role of these cascades has not been widely studied in CF. The TGF\(\beta\) component leads to the activation of SMAD2, a transcriptional modulator that regulates multiple cellular phenotypes, including cell proliferation, apoptosis, and differentiation. High levels of TGF\(\beta\) have been associated with the severity of lung disease [75, 76], and this protein was proposed as a therapeutic target for CF [77]. Although TGF\(\beta\) appears as an important player in CF, according to our model, its deleterious effect results from upstream phenotypes that further activate this growth factor. Our study suggests that therapeutic targets should be chosen among proteins closer to CFTR in the network. In particular, among the source nodes of the main component (as discussed above), because they may more successfully limit the global propagation of molecular dysregulations within the overall network.

Methods

Datasets selection

Based on the search engines of the National Center for Biotechnology Information (NCBI) and the European Nucleotide Archive (ENA), we selected 10 datasets from 8 studies published between 2007 and 2021. The selection criteria to include CF transcriptomic datasets were the following: (1) they should correspond to human Airway Epithelial Cells (hAEC); (2) the cells should be homozygous for the most common mutation F508del; (3) the transcriptomic data should be publicly available. Therefore, studies including samples heterozygous for the F508del mutation ([22] and [23]) and studies with no data available ([78] and [79]) were not included. In addition, studies with less than two samples were excluded ([80] and [81]), as the subsequent statistical analyses require several samples per condition. The list of selected transcriptomic studies is provided in Table 1. The datasets can be downloaded from the European Nucleotide Archive (ENA), the BioStudies portal or the National Center for Biotechnology Information (NCBI) GEO database, with the corresponding accession numbers provided in the ‘Dataset’ column of Table 1.

Biological pathways databases

We initially considered a total of 380 gene sets corresponding to 380 biological pathways: 50 Hallmark gene sets from the Molecular Signatures Database (MSigDB) [31], 196 from the Pathway Interaction Database (PID) [32] and 134 from the KEGG database, restricted to the Genetic Information Processing, Environmental Information Processing; Cellular Processes and Organismal systems subdivision. However, most of the analyses were performed using only KEGG database. Indeed, in the Hallmark and the PID databases, gene sets are defined as gene signatures rather than as biological pathways. Thus, the genes are not necessarily connected to each other through functional interactions. Conversely, gene sets retrieved from the KEGG database correspond to biological pathways defined as genes corresponding to proteins that participate in oriented molecular cascades. They are available in the form of maps on the KEGG website. In addition, the structure of the KEGG database allows the construction of a network that provides mechanistic interpretation. Therefore, gene set enrichment algorithms required to build the signalling network were performed based on the KEGG database. All interactions and nodes from each biological pathway of the KEGG database were retrieved thanks to the OmnipathR R package [44].

Preprocessing of RNA-Seq data

Limma was originally developed for differential expression analysis of microarray data, where the values are assumed to be normally distributed, and the variance independent of the mean. This is not the case for log2-counts per million (log-CPM) values in RNA-Seq data: expression distributions may vary across samples and methods modelling counts assume a quadratic mean-variance relationship. Therefore, for the RNA-Seq data, 3 steps of pre-processing are necessary before applying the statistical tests [82]: (1) low expressed genes are filtered (i.e. genes with less than 10 read counts in at least one sample in the condition with the minimum sample size); (2) normalisation using the method of trimmed mean of M-values (TMM) is performed [83]; (3) raw counts are converted to log-CPM and the mean-variance relationship is estimated with the voom method.

Identification of Differentially Expressed Pathways (DEPs)

For each of the 10 transcriptomic datasets, identification of DEPs was performed using the fgseaSimple function of the Bioconductor package fgsea [84], for fast preranked Gene Set Enrichment Analysis (GSEA) [39].

The fgseaSimple method takes two inputs: a gene-level signed statistics and a defined list of genes known as gene set. The method ranks the genes in descending order based on the chosen statistics, and then computes the Enrichment Score (ES) for the gene set. The ES reflects how often members of that gene set occur at the top (e.g., upregulated) or the bottom (e.g., downregulated) of the ranked gene list. To account for differences in gene sets size, a normalisation step is performed to obtain the Normalised Enrichment Score (NES). Besides, random gene sets are generated and their NES computed. These NES are then used to create a null distribution, from which the significance of the NES of the tested gene set is estimated. In our study, we used the t-statistics from the differential expression analysis comparing gene expression levels of CF sample to NCF samples as the control condition. In order to compare all the studies together, all the microarray and RNA-Seq datasets were processed using the same pipeline, involving the limma [85] and edgeR [86] packages. After removing technical outlier samples and retrieving gene symbols using the biomaRt package [87], differential expression analysis at the gene level was performed by fitting a linear model using weighted least squares for each gene.

Gene sets with size larger than 500 were excluded for statistical testing. The p-values of the gene sets were adjusted for multiple testing error with Benjamini-Hochberg (BH) procedure. Differentially Expressed Pathways (DEP)s were considered with a corrected p-value lower or equal to 0.25. If the NES is positive, the DEP is categorized as up-regulated, and if it is negative, the DEP is categorized as down-regulated.

Up-dating Omnipath DEPs pathways

The CF network was built from DEPs among pathways in the KEGG database, as extracted with the OmniPathR package. We observed a few inconsistencies between the corresponding list of genes and interactions downloaded with OmnipathR R package, and those in the ’up to date’ pathways maps, as they are displayed on KEGG website. Therefore, we updated the OmnipathR version of the KEGG pathways by adding (or removing) a few nodes or interactions, in order to map the OmnipathR pathways with their corresponding pathways in KEGG. For each modification, bibliographic references were manually checked into other databases stored in Omnipath, in particular in the high confident databases SignorDB [88], and the Human Reference Interactome [89]. In addition, in a few pathways, some interactions are labelled as “indirect” in KEGG database. They involve part of signalling cascades belonging to other biological pathways, and they are not detailed in the considered pathway. For example, part of the PI3K-AKT pathway belongs to the Toll-like receptor signalling pathway but is not detailed in this pathway (see KEGG map for Toll-like receptor signalling pathway). In such cases, in order to build the network based on complete cascades involving only direct interactions, we added the missing nodes and interactions.

All the pathways modifications and the corresponding codes used to perform these modifications are available in the following GitHub repository: sysbio-curie/CFnetwork.

Network building and pruning

In the KEGG database, most of the 15 common DEPs display the same overall topology: some cell-surface receptor proteins activate one or more intra-cellular signalling cascades that in turn activate downstream transcription factors, thus triggering corresponding phenotypes. For example, the NF-\(\kappa\)B pathway leads to the “inflammation” or “cell survival” phenotypes. However, 2 of the common DEPs, Cytokine-cytokine receptor interaction and Viral protein interaction with cytokine and cytokine receptor, are pathways that do not consist in such functional cascades. The Cytokine-cytokine receptor interaction pathway consists in a list of interactions between extracellular signal molecules and cell-surface receptors (see KEGG database to visualise this pathway’s topology). These interactions are also part of larger biological pathways that comprise their corresponding downstream cascades. This means that KEGG pathways are partially redundant (i.e. small pathways are part of larger pathways), which is also found in all commonly used pathway databases. In the case of the Cytokine-cytokine receptor interaction pathway, this DEP is dysregulated in the meta-analysis because some of the interactions between extracellular molecules and cell surface receptors are dysregulated, but not necessarily all of them. For example, interactions between TNF-\(\alpha\) and its receptors, or IL17 and its receptors are dysregulated, but this information is also present in the DEPs containing the complete corresponding cascades, i.e. the TNF-\(\alpha\) signalling pathway and the IL-17 signalling pathway. The same type of analysis also holds for the Viral protein interaction with cytokine and cytokine receptor DEP. Overall, from these 2 DEPs, we only retained the cell-surface receptors that are sources of downstream dysregulated cascades in our network. Overall, 25 cell surface receptors without downstream dysregulations in our CF transcriptomic data were removed from the network.

Finally, we also removed from the pathways all the interactions corresponding to genes targeted by transcription factors, downstream of the pathways’ cascades, because these target genes do not define the pathways themselves.

Network building and pruning were performed using the R packages tidyr v.1.2.1, and dplyr v.1.0.10. Transcription factors were identified using the R packages dorothea v.1.4.2 and hgnc v.0.1.2, which give access to the Dorothea [90] and HUGO collections [91], respectively.

Betweenness centrality score

The betweenness centrality (BC) score of a node n is defined by

$$\begin{aligned} \sum _{i \ne j, i \ne n, j \ne n} p_{i n j} / p_{i j} \end{aligned}$$

where \(p_{i j}\) is the total number of shortest paths between nodes i and j while \(p_{i n j}\) is the number of those shortest paths which pass though the vertex n.

BC scores were computed using the betweenness function of the R package igraph v.1.3.4 [92]. This package was also used for the other network topology analyses.

Network visualisation and figure generation

The networks, generated as data frames in R, were imported into Cytoscape v.3.9.0 [93] for visualisation. We designed different custom style for nodes and edges, which are available in the Cytoscape session and also saved as an independent file or JSON file, available in the sysbio-curie/CFnetwork_cystoscape GitHub repository. The hierarchical layout was used to emphasize the information flow from the source nodes to the sink nodes. The shiny application https://njmmatthieu.shinyapps.io/shinyapp was developed using the R packages cyjShiny v.1.0.42 [94] and shiny v.1.8.1.1.

Barplots were generated using the R package ggplot2 v.3.3.6, and didactic figures were created using the open-source platform diagrams.net.

Discussion

Using a pathway-based meta-analysis of transcriptomic data, we built the CF network that provides a global understanding of the signalling dysregulations in CF. An important contribution was to integrate data analyses to network reconstruction, while proposing a strategy to relate CFTR to proteins of the network, based on CFTR interactome. The CF network comprises a restricted number of source nodes that allow to relate the absence of CFTR to the downstream sink nodes triggering CF cellular phenotypes. The network reveals non-intuitive and complex routes relating absence of a functional CFTR to CF phenotypes. Overall, the CF network can be seen as a tool to formulate hypotheses and interpret experimental observations.

Although several transcriptomic datasets were gathered, the total number of samples globally included remains modest (57 CF and 46 control samples). Additional transcriptomic data may refine the list of dysregulated pathways, and help to improve the proposed CF network. To cope with the low number of samples per transcriptomic study, we opted for a meta-analysis combining various datasets, which highlighted that distinct differentiation media used for the primary cultures may favour different cell types, leading to inconsistent transcriptomic profiles and potential erroneous interpretations. This may explain why previous comparative transcriptomic studies reported incoherent signs of gene dysregulation (up- versus down-) between different datasets for many genes [21]. We observed the same phenomenon at the pathway level for datasets belonging to subgroup 1 or 2 (see Meta-analysis of transcriptomic studies at the level of biological pathways section). Clustering studies based on the heatmap of common DEPs appears to be a good tool to select consistent data in future meta-analysis.

All datasets included in the meta-analysis are from bulk RNA sequencing studies. In the past three years, CF airway epithelial single-cell RNAseq (sc-RNAseq) datasets have been reported [95, 96]. Such data allow the study of dysregulations at the cell type level, and could facilitate building of the CF networks for specific epithelial cell types. In particular, our meta-analysis results suggest that the inflammatory phenotype is more pronounced in CF secretory cells than in CF ciliated cells. According to our hypothesis, i.e. signalling dysregulations are due to the absence of wt-CFTR at the PM, this finding would align with a recent scRNA-seq study showing that CFTR expression is higher in secretory cells than in ciliated cells [97]. Therefore, building signalling networks for these two specific cell types deserved to be further explored, in order to study their relative contribution to the “bulk” network from the present study. Furthermore, CFTR is expressed in cell types beyond airway epithelial cells. Thus, refining this network within the context of these cell types could enhance our understanding of the role of CFTR in specific cells such as macrophages, where CFTR seems to present non-channel functions [98].

Prior knowledge gathered in the KEGG pathway database was used to identify and connect the signalling DEPs, but the proposed methodology can be followed using other pathways databases. Pathway names and definitions vary between databases, and therefore, the resulting network may slightly depend on the reference database that was used. Nevertheless, it should comprise globally the same interactions and the same proteins.

Similarly, CFTR interactors present in the network were identified according to PPI information in the CyFi-MAP. If new CFTR interactors are identified, this information may help improve the network, highlighting new source nodes or routes for the propagation of dysregulations. In particular, missing interactions, because they are not present in pathway databases, or have not been discovered yet, may explain the presence of unconnected components. If they exist, their discovery would make it possible to link the two unconnected small components to the main component of the network. Nevertheless, targeting proteins as upstream as possible in the network, or among key hubs of the network, is still an interesting concept in order to prioritize candidate therapeutic targets.

An important issue of the paper was to explore the link between absence of the CFTR protein, and signalling dysregulations leading to CF cellular phenotypes. However, the precise definition of a diseased cellular phenotype is not clearly defined yet, and we used keywords provided in the KEGG database or in the Gene Cards database [99]. The present work proposes to answer this question in the context of systems biology studies. Associating phenotypes to the activity of outputs of the signalling cascades, referred to here as sink nodes, could be a first step towards the definition of the disease read-outs. This is of particular interest for in vitro evaluation of drug candidates, because drugs active in CF are expected to reduce the activity of these sink nodes.

We are aware that other types of dysregulations such as aberrant phosphorylations are not detectable in transcriptomic data used in the present study. Including results from other types of omic data such as proteomic, phospho-proteomic, or even from metabolomic, epigenetic or volatolomic data may help to refine the CF network. The methodology presented in this study to build a network from transcriptomic data is applicable to other omic modalities, as soon as they become available.

In the longer term, by providing a better understanding of CF molecular mechanisms, the proposed systems biology approach could help optimise treatment and define new therapeutic strategies. In this context, an important contribution was to propose candidate therapeutic targets (namely, PI3KCA, PLCB1/3, SYK and SRC), based in particular on the topological analysis of the network. The immediate next step would be to experimentally evaluate the interest of these proposed candidates. This could be performed on CF cellular models and their control counterparts, using relevant readouts that include inflammation biomarkers. Because these proteins play essential roles in the cell, their complete inhibition could trigger deleterious side effects. However, partial inhibition might help modulate CF cellular phenotypes without significant side effects. Interestingly, marketed drugs are available for some of these candidates, namely SYK and PLCB1/3, which further justifies their experimental evaluation.

The last approved treatment known as Highly Effective Modulator Treatment (HEMT), also known as Trikafta, a combination therapy of three modulators (VX-445/VX-661/VX-770), enabled significant decrease of sweat chloride concentration and demonstrated major improvement of lung function for patients carrying at least one F508del mutation [100]. To date, omic studies comparing Trikafta-treated and untreated samples homozygous for F508del are very limited. However, as soon as more studies are available, the methodology presented in this article could be applied to these data in order to better understand the overall mechanism of this new therapy by highlighting the molecular mechanisms modulated by the treatment.

Furthermore, the proposed approach was settled on data from hAEC cells homozygous for F508del, because publicly available data are more abundant for this most frequent mutation. Therefore, our CF network characterizes the disease caused by this mutation. It would be interesting to study to which extent the CF network would differ for other mutations. A recent paper indicates that DEGs in human bronchial epithelial cell lines bearing mutations from different classes share about 30% DEGs, while 70% of the DEGs are class specific [101]. It would be interesting to study if this still holds at the level of biological pathways, and whether the resulting network is strongly modified between different mutation classes, or not. The resulting networks could highlight similar targets (as in the present study), or class-specific targets, potentially extending the therapeutic arsenal available for CF patients who are not eligible for CFTR modulators.

In the same line, it is now clear that CF patients bearing the same mutation may present diseases of different severity. Although many factors can modulate disease severity, including environmental factors, it would be interesting to explore the impact of patients’ molecular profiles. Refining the CF network into a “personalized” network based on patient’s transcriptomic data would be an interesting tool to answer this question.

Beyond CF, reduced amounts of functional CFTR have also been observed in other diseases like chronic obstructive pulmonary disease (COPD) [102, 103], cigarette smoke [104] , or cancer [98, 105]. The network could provide a basis to explore the consequences of reduced CFTR activity in these diseases. To a greater extent, the global approach proposed in this work, could also be applied to study other monogenic diseases.

Conclusion

We presented building of the CF network, a signalling network gathering the molecular dysregulations caused by the absence of CFTR. We adopted a data-driven systems biology approach to retrieve CF dysregulated signalling pathways. These pathways were merged to build a signalling network, recapitulating the dysregulated cascades that flow from the source nodes (proteins connected to CFTR) to the sink nodes (proteins that trigger CF cellular phenotypes). Five of the source nodes are upstream of all the sink nodes in the CF network: PLCB1/3, TRADD, SRC, and SYK. These proteins may collectively initiate the emergence of CF phenotypes (together with the other three source nodes EZR, CSNK2A1, and PRKCA), illustrating the complexity of the disease. The topological analysis of the network also highlighted nodes with a high degree of betweenness centrality, which are other important players in the propagation of the dysregulations, including PI3KCA. Among these key source nodes and nodes with high degree of centrality, SYK, SRC, PLCB1/3 and PIK3CA appeared as interesting candidate therapeutic targets. Interestingly, specific inhibitors are known for these proteins, and even marketed drugs in the case of SYK and PI3KCA. They stand out as potential therapeutic candidates for drug repositioning, potentially allowing the modulation of various CF phenotypes. Finally, an important contribution of the present work is that the adopted global methodology of the CFTR context, although perfectible, did provide interesting results for CF, and can be used as a common framework for other monogenic diseases.

Availability of data and materials

The codes and datasets supporting the conclusions of this article are available in the following GitHub repository: https://github.com/sysbio-curie/CFnetwork.git. The transcriptomic datasets used in this study are publicly available via the European Nucleotide Archive (ENA), the BioStudies portal or the National Center for Biotechnology Information (NCBI) GEO database, with the corresponding accession numbers provided in the ‘Dataset’ column of Table 1. The Cytoscape session of the CF network, the TSV files of the nodes and the edges of the CF network and the XML files of the custom styles of the Cytoscape session required to reproduce the Cytoscape session are available in the following GitHub repository: https://github.com/sysbio-curie/CFnetwork_cytoscape.git.

Abbreviations

BC:

Betweenness Centrality

CF:

Cystic Fibrosis

CFTR:

Cystic Fibrosis Transmembrane Conductance Regulator

DEG:

Differentially Expressed Gene

DEP:

Differentially Expressed Pathway

GSEA:

Gene Set Enrichment Analysis

hAEC:

human Airway Epithelial Cells

NCF:

Non-cystic Fibrosis

NES:

Normalised Enrichment Score

PPI:

Protein-Protein Interaction

References

  1. Guo J, Garratt A, Hill A. Worldwide rates of diagnosis and effective treatment for cystic fibrosis. J Cyst Fibros. 2022;21(3):456–62. https://doi.org/10.1016/j.jcf.2022.01.009.

    Article  PubMed  Google Scholar 

  2. Seibert FS, Loo TW, Clarke DM, Riordan JR. Cystic Fibrosis: Channel, Catalytic, and Folding Properties of the CFTR Protein. J Bioenerg Biomembr. 1997;29(5):429–42. https://doi.org/10.1023/A:1022478822214.

    Article  CAS  PubMed  Google Scholar 

  3. Veit G, Avramescu RG, Chiang AN, Houck SA, Cai Z, Peters KW, et al. From CFTR biology toward combinatorial pharmacotherapy: expanded classification of cystic fibrosis mutations. Mol Biol Cell. 2016;27(3):424–33. Publisher: American Society for Cell Biology (mboc). https://doi.org/10.1091/mbc.e14-04-0935.

  4. Hanssens LS, Duchateau J, Casimir GJ. CFTR Protein: Not Just a Chloride Channel? Cells. 2021;10(11):2844. Number: 11 Publisher: Multidisciplinary Digital Publishing Institute. https://doi.org/10.3390/cells10112844.

  5. Pereira C, Mazein A, Farinha CM, Gray MA, Kunzelmann K, Ostaszewski M, et al. CyFi-MAP: an interactive pathway-based resource for cystic fibrosis. Sci Rep. 2021;11(1):22223. Number: 1 Publisher: Nature Publishing Group. https://doi.org/10.1038/s41598-021-01618-3.

  6. Farinha CM, Gentzsch M. Revisiting CFTR Interactions: Old Partners and New Players. Int J Mol Sci. 2021;22(24):13196. Number: 24 Publisher: Multidisciplinary Digital Publishing Institute. https://doi.org/10.3390/ijms222413196.

  7. Crites KSM, Morin G, Orlando V, Patey N, Cantin C, Martel J, et al. CFTR Knockdown induces proinflammatory changes in intestinal epithelial cells. J Inflamm. 2015;12(1):62. https://doi.org/10.1186/s12950-015-0107-y.

    Article  CAS  Google Scholar 

  8. Fleurot I, López-Gálvez R, Barbry P, Guillon A, Si-Tahar M, Bähr A, et al. TLR5 signalling is hyper-responsive in porcine cystic fibrosis airways epithelium. J Cyst Fibros. 2022;21(2):e117–21. https://doi.org/10.1016/j.jcf.2021.08.002.

    Article  CAS  PubMed  Google Scholar 

  9. Hao S, Roesch EA, Perez A, Weiner RL, Henderson LC, Cummings L, et al. Inactivation of CFTR by CRISPR/Cas9 alters transcriptional regulation of inflammatory pathways and other networks. J Cyst Fibros. 2020;19(1):34–9. https://doi.org/10.1016/j.jcf.2019.05.003.

    Article  CAS  PubMed  Google Scholar 

  10. Jacquot J, Tabary O, Le Rouzic P, Clement A. Airway epithelial cell inflammatory signalling in cystic fibrosis. Int J Biochem Cell Biol. 2008;40(9):1703–15. https://doi.org/10.1016/j.biocel.2008.02.002.

    Article  CAS  PubMed  Google Scholar 

  11. Jeanson L, Kelly M, Coste A, Guerrera IC, Fritsch J, Nguyen-Khoa T, et al. Oxidative stress induces unfolding protein response and inflammation in nasal polyposis. Allergy. 2012;67(3):403–12. https://doi.org/10.1111/j.1398-9995.2011.02769.x.

    Article  CAS  PubMed  Google Scholar 

  12. Conese M, Di Gioia S. Pathophysiology of Lung Disease and Wound Repair in Cystic Fibrosis. Pathophysiology. 2021;28(1):155–88. Number: 1 Publisher: Multidisciplinary Digital Publishing Institute. https://doi.org/10.3390/pathophysiology28010011.

  13. Pankonien I, Quaresma MC, Rodrigues CS, Amaral MD. CFTR, Cell Junctions and the Cytoskeleton. Int J Mol Sci. 2022;23(5):2688. Number: 5 Publisher: Multidisciplinary Digital Publishing Institute. https://doi.org/10.3390/ijms23052688.

  14. Ideozu JE, Zhang X, McColley S, Levy H. Transcriptome Profiling and Molecular Therapeutic Advances in Cystic Fibrosis: Recent Insights. Genes. 2019;10(3). https://doi.org/10.3390/genes10030180.

  15. Ghigo A, Prono G, Riccardi E, De Rose V. Dysfunctional Inflammation in Cystic Fibrosis Airways: From Mechanisms to Novel Therapeutic Approaches. Int J Mol Sci. 2021;22(4):1952. Number: 4 Publisher: Multidisciplinary Digital Publishing Institute. https://doi.org/10.3390/ijms22041952.

  16. Hornberg JJ, Bruggeman FJ, Westerhoff HV, Lankelma J. Cancer: A Systems Biology disease. Biosystems. 2006;83(2):81–90. https://doi.org/10.1016/j.biosystems.2005.05.014.

    Article  CAS  PubMed  Google Scholar 

  17. Pankow S, Bamberger C, Calzolari D, Martínez-Bartolomé S, Lavallée-Adam M, Balch WE, et al. \(\Delta\)F508 CFTR interactome remodelling promotes rescue of cystic fibrosis. Nature. 2015;528(7583):510–6. Number: 7583 Publisher: Nature Publishing Group. https://doi.org/10.1038/nature15729.

  18. Trivedi TS, Bhadresha KP, Patel MP, Mankad AU, Rawal RM, Patel SK. Identification of hub genes associated with human cystic fibrosis: A Meta-analysis approach. Hum Gene. 2023;35:201139. https://doi.org/10.1016/j.humgen.2022.201139.

    Article  CAS  Google Scholar 

  19. Buccitelli C, Selbach M. mRNAs, proteins and the emerging principles of gene expression control. Nat Rev Genet. 2020;21(10):630–44. Number: 10 Publisher: Nature Publishing Group. https://doi.org/10.1038/s41576-020-0258-4.

  20. Szalai B, Saez-Rodriguez J. Why do pathway methods work better than they should? FEBS Lett. 2020;594(24):4189–200. https://doi.org/10.1002/1873-3468.14011.

    Article  CAS  PubMed  Google Scholar 

  21. Clarke LA, Sousa L, Barreto C, Amaral MD. Changes in transcriptome of native nasal epithelium expressing F508del-CFTR and intersecting data from comparable studies. Respir Res. 2013;14:38. https://doi.org/10.1186/1465-9921-14-38.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Virella-Lowell I, Herlihy JD, Liu B, Lopez C, Cruz P, Muller C, et al. Effects of CFTR, interleukin-10, and Pseudomonas aeruginosa on gene expression profiles in a CF bronchial epithelial cell Line. Mol Ther. 2004;10(3):562–73. Publisher: Elsevier. https://doi.org/10.1016/j.ymthe.2004.06.215.

  23. Rehman T, Karp PH, Tan P, Goodell BJ, Pezzulo AA, Thurman AL, et al. Inflammatory cytokines TNF-\(\upalpha\) and IL-17 enhance the efficacy of cystic fibrosis transmembrane conductance regulator modulators. J Clin Investig. 2021;131(16):150398. https://doi.org/10.1172/JCI150398.

  24. Verhaeghe C, Remouchamps C, Hennuy B, Vanderplasschen A, Chariot A, Tabruyn SP, et al. Role of IKK and ERK pathways in intrinsic inflammation of cystic fibrosis airways. Biochem Pharmacol. 2007;73(12):1982–94. https://doi.org/10.1016/j.bcp.2007.03.019.

    Article  CAS  PubMed  Google Scholar 

  25. Ogilvie V, Passmore M, Hyndman L, Jones L, Stevenson B, Wilson A, et al. Differential global gene expression in cystic fibrosis nasal and bronchial epithelium. Genomics. 2011;98(5):327–36. https://doi.org/10.1016/j.ygeno.2011.06.008.

    Article  CAS  PubMed  Google Scholar 

  26. Voisin G, Bouvet GF, Legendre P, Dagenais A, Massé C, Berthiaume Y. Oxidative stress modulates the expression of genes involved in cell survival in \(\Delta\)F508 cystic fibrosis airway epithelial cells. Physiol Genomics. 2014;46(17):634–46. Publisher: American Physiological Society. https://doi.org/10.1152/physiolgenomics.00003.2014.

  27. Balloy V, Varet H, Dillies MA, Proux C, Jagla B, Coppée JY, et al. Normal and Cystic Fibrosis Human Bronchial Epithelial Cells Infected with Pseudomonas aeruginosa Exhibit Distinct Gene Activation Patterns. PLoS ONE. 2015;10(10):e0140979. Publisher: Public Library of Science. https://doi.org/10.1371/journal.pone.0140979.

  28. Zoso A, Sofoluwe A, Bacchetta M, Chanson M. Transcriptomic profile of cystic fibrosis airway epithelial cells undergoing repair. Sci Data. 2019;6(1):1–7. Number: 1 Publisher: Nature Publishing Group. https://doi.org/10.1038/s41597-019-0256-6.

  29. Ling KM, Garratt LW, Gill EE, Lee AHY, Agudelo-Romero P, Sutanto EN, et al. Rhinovirus Infection Drives Complex Host Airway Molecular Responses in Children With Cystic Fibrosis. Front Immunol. 2020;11:1327. https://doi.org/10.3389/fimmu.2020.01327.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Saint-Criq V, Delpiano L, Casement J, Onuora JC, Lin J, Gray MA. Choice of Differentiation Media Significantly Impacts Cell Lineage and Response to CFTR Modulators in Fully Differentiated Primary Cultures of Cystic Fibrosis Human Airway Epithelial Cells. Cells. 2020;9(9):2137. Number: 9 Publisher: Multidisciplinary Digital Publishing Institute. https://doi.org/10.3390/cells9092137.

  31. Liberzon A, Birger C, Thorvaldsdóttir H, Ghandi M, Mesirov JP, Tamayo P. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst. 2015;1(6):417–25. https://doi.org/10.1016/j.cels.2015.12.004.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Schaefer CF, Anthony K, Krupa S, Buchoff J, Day M, Hannay T, et al. PID: the Pathway Interaction Database. Nucleic Acids Res. 2009;37(Database issue):D674–9. https://doi.org/10.1093/nar/gkn653.

  33. Kanehisa M, Furumichi M, Sato Y, Ishiguro-Watanabe M, Tanabe M. KEGG: integrating viruses and cellular organisms. Nucleic Acids Res. 2021;49(D1):D545–51. https://doi.org/10.1093/nar/gkaa970.

    Article  CAS  PubMed  Google Scholar 

  34. Wang K, Li M, Hakonarson H. Analysing biological pathways in genome-wide association studies. Nat Rev Genet. 2010;11(12):843–54. Number: 12 Publisher: Nature Publishing Group. https://doi.org/10.1038/nrg2884.

  35. Martignetti L, Calzone L, Bonnet E, Barillot E, Zinovyev A. ROMA: Representation and Quantification of Module Activity from Target Expression Data. Front Genet. 2016;7. https://doi.org/10.3389/fgene.2016.00018.

  36. Landais Y, Vallot C. Multi-modal quantification of pathway activity with MAYA. Nat Commun. 2023;14(1):1668. Number: 1 Publisher: Nature Publishing Group. https://doi.org/10.1038/s41467-023-37410-2.

  37. Schubert M, Klinger B, Klünemann M, Sieber A, Uhlitz F, Sauer S, et al. Perturbation-response genes reveal signaling footprints in cancer gene expression. Nat Commun. 2018;9(1):20. Number: 1 Publisher: Nature Publishing Group. https://doi.org/10.1038/s41467-017-02391-6.

  38. Vaske CJ, Benz SC, Sanborn JZ, Earl D, Szeto C, Zhu J, et al. Inference of patient-specific pathway activities from multi-dimensional cancer genomics data using PARADIGM. Bioinformatics. 2010;26(12):i237–45. https://doi.org/10.1093/bioinformatics/btq182.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci. 2005;102(43):15545–50. Publisher: National Academy of Sciences Section: Biological Sciences. https://doi.org/10.1073/pnas.0506580102.

  40. Hsu D, Taylor P, Fletcher D, van Heeckeren R, Eastman J, van Heeckeren A, et al. Interleukin-17 Pathophysiology and Therapeutic Intervention in Cystic Fibrosis Lung Infection and Inflammation. Infect Immun. 2016;84(9):2410–21. Publisher: American Society for Microbiology. https://doi.org/10.1128/iai.00284-16.

  41. Dumortier C, Danopoulos S, Velard F, Al Alam D. Bone cells differentiation: how CFTR mutations may rule the game of stem cells commitment? Front Cell Dev Biol. 2021;9:611921.

  42. Kosamo S, Hisert KB, Dmyterko V, Nguyen C, Black RA, Holden TD, et al. Strong toll-like receptor responses in cystic fibrosis patients are associated with higher lung function. J Cyst Fibros. 2020;19(4):608–13. https://doi.org/10.1016/j.jcf.2019.11.009.

    Article  CAS  PubMed  Google Scholar 

  43. Curutiu C, Iordache F, Lazar V, Pisoschi AM, Pop A, Chifiriuc MC, et al. Impact of Pseudomonas aeruginosa quorum sensing signaling molecules on adhesion and inflammatory markers in endothelial cells. Beilstein J Org Chem. 2018;14:2580–8. https://doi.org/10.3762/bjoc.14.235.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Türei D, Korcsmáros T, Saez-Rodriguez J. OmniPath: guidelines and gateway for literature-curated signaling pathway resources. Nat Methods. 2016;13(12):966–7. Number: 12 Publisher: Nature Publishing Group. https://doi.org/10.1038/nmeth.4077.

  45. Venerando A, Pagano MA, Tosoni K, Meggio F, Cassidy D, Stobbart M, et al. Understanding protein kinase CK2 mis-regulation upon F508del CFTR expression. Naunyn-Schmiedebergs Arch Pharmacol. 2011;384(4):473–88. https://doi.org/10.1007/s00210-011-0650-x.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Wang H, Cebotaru L, Lee HW, Yang Q, Pollard BS, Pollard HB, et al. CFTR Controls the Activity of NF-\(\upkappa\)B by Enhancing the Degradation of TRADD. Cell Physiol Biochem. 2016;40(5):1063–78. Publisher: Karger Publishers. https://doi.org/10.1159/000453162.

  47. Massip Copiz MM, Santa Coloma TA. c- Src and its role in cystic fibrosis. Eur J Cell Biol. 2016;95(10):401–13. https://doi.org/10.1016/j.ejcb.2016.08.001.

    Article  CAS  PubMed  Google Scholar 

  48. Rimessi A, Bezzerri V, Salvatori F, Tamanini A, Nigro F, Dechecchi MC, et al. PLCB3 Loss of Function Reduces Pseudomonas aeruginosa-Dependent IL-8 Release in Cystic Fibrosis. Am J Respir Cell Mol Biol. 2018;59(4):428–36. Publisher: American Thoracic Society - AJRCMB. https://doi.org/10.1165/rcmb.2017-0267OC.

  49. Favia M, Guerra L, Fanelli T, Cardone RA, Monterisi S, Di Sole F, et al. Na+/H+ Exchanger Regulatory Factor 1 Overexpression-dependent Increase of Cytoskeleton Organization Is Fundamental in the Rescue of F508del Cystic Fibrosis Transmembrane Conductance Regulator in Human Airway CFBE41o- Cells. Mol Biol Cell. 2010;21(1):73–86. Publisher: American Society for Cell Biology (mboc). https://doi.org/10.1091/mbc.e09-03-0185.

  50. Wu Q, Eickelberg O. Ezrin in Asthma: A First Step to Early Biomarkers of Airway Epithelial Dysfunction. Am J Respir Crit Care Med. 2019;199(4):408–10. Publisher: American Thoracic Society - AJRCCM. https://doi.org/10.1164/rccm.201810-1964ED.

  51. Mendes AI, Matos P, Moniz S, Luz S, Amaral MD, Farinha CM, et al. Antagonistic Regulation of Cystic Fibrosis Transmembrane Conductance Regulator Cell Surface Expression by Protein Kinases WNK4 and Spleen Tyrosine Kinase. Mol Cell Biol. 2011;31(19):4076–86. Publisher: Taylor & Francis _eprint: https://doi.org/10.1128/MCB.05152-11.

  52. Egan M, Flotte T, Afione S, Solow R, Zeitlin PL, Carter BJ, et al. Defective regulation of outwardly rectifying Cl channels by protein kinase A corrected by insertion of CFTR. Nature. 1992;358(6387):581–4. Number: 6387 Publisher: Nature Publishing Group. https://doi.org/10.1038/358581a0.

  53. Bérubé J, Roussel L, Nattagh L, Rousseau S. Loss of Cystic Fibrosis Transmembrane Conductance Regulator Function Enhances Activation of p38 and ERK MAPKs, Increasing Interleukin-6 Synthesis in Airway Epithelial Cells Exposed to Pseudomonas aeruginosa. J Biol Chem. 2010;285(29):22299–307. Publisher: American Society for Biochemistry and Molecular Biology. https://doi.org/10.1074/jbc.M109.098566.

  54. Wellmerling J, Rayner RE, Chang SW, Kairis EL, Kim SH, Sharma A, et al. Targeting the EGFR-ERK axis using the compatible solute ectoine to stabilize CFTR mutant F508del. FASEB J Off Publ Fed Am Soc Exp Biol. 2022;36(5):e22270. https://doi.org/10.1096/fj.202100458RRR.

    Article  CAS  Google Scholar 

  55. Cannon CL, Kowalski MP, Stopak KS, Pier GB. Pseudomonas aeruginosa-Induced Apoptosis Is Defective in Respiratory Epithelial Cells Expressing Mutant Cystic Fibrosis Transmembrane Conductance Regulator. Am J Respir Cell Mol Biol. 2003;29(2):188–97. Publisher: American Thoracic Society - AJRCMB. https://doi.org/10.1165/rcmb.4898.

  56. Gottlieb RA, Dosanjh A. Mutant cystic fibrosis transmembrane conductance regulator inhibits acidification and apoptosis in C127 cells: possible relevance to cystic fibrosis. Proc Natl Acad Sci. 1996;93(8):3587–91. Publisher: Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.93.8.3587.

  57. Chen Q, Pandi SPS, Kerrigan L, McElvaney NG, Greene CM, Elborn JS, et al. Cystic fibrosis epithelial cells are primed for apoptosis as a result of increased Fas (CD95). J Cyst Fibros. 2018;17(5):616–23. https://doi.org/10.1016/j.jcf.2018.01.010.

    Article  CAS  PubMed  Google Scholar 

  58. Yalçin E, Talim B, Özçelik U, Doğru D, Çobanoğlu N, Pekcan S, et al. Does Defective Apoptosis Play A Role in Cystic Fibrosis Lung Disease? Arch Med Res. 2009;40(7):561–4. https://doi.org/10.1016/j.arcmed.2009.07.005.

    Article  CAS  PubMed  Google Scholar 

  59. Rottner M, Kunzelmann C, Mergey M, Freyssinet JM, Martínez MC. Exaggerated apoptosis and NF-kappaB activation in pancreatic and tracheal cystic fibrosis cells. FASEB J Off Publ Fed Am Soc Exp Biol. 2007;21(11):2939–48. https://doi.org/10.1096/fj.06-7614com.

    Article  Google Scholar 

  60. Lasalvia M, Castellani S, D’Antonio P, Perna G, Carbone A, Colia AL, et al. Human airway epithelial cells investigated by atomic force microscopy: A hint to cystic fibrosis epithelial pathology. Exp Cell Res. 2016;348(1):46–55. https://doi.org/10.1016/j.yexcr.2016.08.025.

    Article  CAS  PubMed  Google Scholar 

  61. Burat B, Reynaerts A, Baiwir D, Fléron M, Gohy S, Eppe G, et al. Sweat Proteomics in Cystic Fibrosis: Discovering Companion Biomarkers for Precision Medicine and Therapeutic Development. Cells. 2022;11(15):2358. Number: 15 Publisher: Multidisciplinary Digital Publishing Institute. https://doi.org/10.3390/cells11152358.

  62. De Lisle RC. Disrupted tight junctions in the small intestine of cystic fibrosis mice. Cell Tissue Res. 2014;355(1):131–42. Available from: https://doi.org/10.1007/s00441-013-1734-3. https://doi.org/10.1007/s00441-013-1734-3.

  63. Castellani S, Guerra L, Favia M, Di Gioia S, Casavola V, Conese M. NHERF1 and CFTR restore tight junction organisation and function in cystic fibrosis airway epithelial cells: role of ezrin and the RhoA/ROCK pathway. Lab Investig. 2012;92(11):1527–40. Number: 11 Publisher: Nature Publishing Group. https://doi.org/10.1038/labinvest.2012.123.

  64. Gillan JL, Chokshi M, Hardisty GR, Clohisey Hendry S, Prasca-Chamorro D, Robinson NJ, et al. CAGE sequencing reveals CFTR-dependent dysregulation of type I IFN signaling in activated cystic fibrosis macrophages. Sci Adv. 2023;9(21):eadg5128. Publisher: American Association for the Advancement of Science. https://doi.org/10.1126/sciadv.adg5128.

  65. Dugger DT, Fung M, Zlock L, Caldera S, Sharp L, Hays SR, et al. Cystic Fibrosis Lung Transplant Recipients Have Suppressed Airway Interferon Responses during Pseudomonas Infection. Cell Rep Med. 2020;1(4):100055. https://doi.org/10.1016/j.xcrm.2020.100055.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  66. Durón C, Pan Y, Gutmann DH, Hardin J, Radunskaya A. Variability of Betweenness Centrality and Its Effect on Identifying Essential Genes. Bull Math Biol. 2019;81(9):3655–73. Available from: https://doi.org/10.1007/s11538-018-0526-z. https://doi.org/10.1007/s11538-018-0526-z.

  67. Ferenc Karpati LH Bengt Wretlind. TNF-A and IL-8 in Consecutive Sputum Samples from Cystic Fibrosis Patients During Antibiotic Treatment. Scand J Infect Dis. 2000;32(1):75–9. Publisher: Taylor & Francis _eprint: https://doi.org/10.1080/00365540050164263.

  68. Riccaboni M, Bianchi I, Petrillo P. Spleen tyrosine kinases: biology, therapeutic targets and drugs. Drug Discov Today. 2010;15(13):517–30. https://doi.org/10.1016/j.drudis.2010.05.001.

    Article  CAS  PubMed  Google Scholar 

  69. Wong BR, Grossbard EB, Payan DG, Masuda ES. Targeting Syk as a treatment for allergic and autoimmune disorders. Expert Opin Investig Drugs. 2004;13(7):743–62. Publisher: Taylor & Francis _eprint: https://doi.org/10.1517/13543784.13.7.743.

  70. Bezzerri V, d’Adamo P, Rimessi A, Lanzara C, Crovella S, Nicolis E, et al. Phospholipase C-\(\upbeta\)3 Is a Key Modulator of IL-8 Expression in Cystic Fibrosis Bronchial Epithelial Cells. J Immunol. 2011;186(8):4946–58. https://doi.org/10.4049/jimmunol.1003535.

  71. Bell SC, Mall MA, Gutierrez H, Macek M, Madge S, Davies JC, et al. The Lancet Respiratory Medicine Commission on the Future of Care of Cystic Fibrosis. Lancet Respir Med. 2020;8(1):65–124. https://doi.org/10.1016/S2213-2600(19)30337-6.

    Article  CAS  PubMed  Google Scholar 

  72. Cooper N, Ghanima W, Hill QA, Nicolson PL, Markovtsov V, Kessler C. Recent advances in understanding spleen tyrosine kinase (SYK) in human biology and disease, with a focus on fostamatinib. Platelets. 2023;34(1):2131751. Publisher: Taylor & Francis _eprint: https://doi.org/10.1080/09537104.2022.2131751.

  73. Fiorotto R, Amenduni M, Mariotti V, Fabris L, Spirli C, Strazzabosco M. Src kinase inhibition reduces inflammatory and cytoskeletal changes in \(\Delta\)F508 human cholangiocytes and improves cystic fibrosis transmembrane conductance regulator correctors efficacy: Fiorotto, Amenduni, et al. Hepatology. 2018;67(3):972–88. Number: 3. https://doi.org/10.1002/hep.29400.

  74. Natarajan V. Is PI3K a Villain in Cystic Fibrosis? Am J Respir Cell Mol Biol. 2020;62(5):552–3. Publisher: American Thoracic Society - AJRCMB. https://doi.org/10.1165/rcmb.2020-0029ED.

  75. Dorfman R, Sandford A, Taylor C, Huang B, Frangolias D, Wang Y, et al. Complex two-gene modulation of lung disease severity in children with cystic fibrosis. J Clin Investig. 2008;118(3):1040–9. Publisher: American Society for Clinical Investigation. https://doi.org/10.1172/JCI33754.

  76. Sagwal S, Chauhan A, Kaur J, Prasad R, Singh M, Singh M. Association of Serum TGF-\(\upbeta\)1 Levels with Different Clinical Phenotypes of Cystic Fibrosis Exacerbation. Lung. 2020;198(2):377–83. https://doi.org/10.1007/s00408-020-00320-x.

  77. Kramer EL, Clancy JP. TGFB as a therapeutic target in cystic fibrosis. Expert Opin Ther Targets. 2018;22(2):177–89. Publisher: Taylor & Francis _eprint: https://doi.org/10.1080/14728222.2018.1406922.

  78. Zabner J, Scheetz TE, Almabrazi HG, Casavant TL, Huang J, Keshavjee S, et al. CFTR \(\Delta\)F508 mutation has minimal effect on the gene expression profile of differentiated human airway epithelia. American Journal of Physiology-Lung Cellular and Molecular Physiology. 2005;289(4):L545–53. Publisher: American Physiological Society. https://doi.org/10.1152/ajplung.00065.2005.

  79. Wright JM, Merlo CA, Reynolds JB, Zeitlin PL, Garcia JGN, Guggino WB, et al. Respiratory epithelial gene expression in patients with mild and severe cystic fibrosis lung disease. Am J Respir Cell Mol Biol. 2006;35(3):327–36. https://doi.org/10.1165/rcmb.2005-0359OC.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  80. Bampi GB, Rauscher R, Kirchner S, Oliver KE, Bijvelds MJC, Santos LA, et al. Global assessment of the integrated stress response in CF patient-derived airway and intestinal tissues. J Cyst Fibros. 2020;19(6):1021–6. https://doi.org/10.1016/j.jcf.2020.04.005.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  81. Veltman M, De Sanctis JB, Stolarczyk M, Klymiuk N, Bähr A, Brouwer RW, et al. CFTR correctors and antioxidants partially normalize lipid imbalance but not abnormal basal inflammatory cytokine profile in CF bronchial epithelial cells. Front Physiol. 2021;12:619442.

  82. Law CW, Alhamdoosh M, Su S, Dong X, Tian L, Smyth GK, et al. RNA-seq analysis is easy as 1-2-3 with limma, Glimma and edgeR. F1000Research. 2018;5:ISCB Comm J–1408. https://doi.org/10.12688/f1000research.9005.3.

  83. Robinson MD, Oshlack A. A scaling normalization method for differential expression analysis of RNA-seq data. Genome Biol. 2010;11(3):R25. https://doi.org/10.1186/gb-2010-11-3-r25.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  84. Korotkevich G, Sukhov V, Budin N, Shpak B, Artyomov MN, Sergushichev A. Fast gene set enrichment analysis. bioRxiv. 2021. Pages: 060012 Section: New Results. https://doi.org/10.1101/060012.

  85. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. https://doi.org/10.1093/nar/gkv007.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  86. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics. 2010;26(1):139–40. https://doi.org/10.1093/bioinformatics/btp616.

    Article  CAS  PubMed  Google Scholar 

  87. Durinck S, Spellman PT, Birney E, Huber W. Mapping Identifiers for the Integration of Genomic Datasets with the R/Bioconductor package biomaRt. Nat Protoc. 2009;4(8):1184–91. https://doi.org/10.1038/nprot.2009.97.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  88. Lo Surdo P, Iannuccelli M, Contino S, Castagnoli L, Licata L, Cesareni G, et al. SIGNOR 3.0, the SIGnaling network open resource 3.0: 2022 update. Nucleic Acids Res. 2022;gkac883. https://doi.org/10.1093/nar/gkac883.

  89. Drew K, Wallingford JB, Marcotte EM. hu.MAP 2.0: integration of over 15,000 proteomic experiments builds a global compendium of human multiprotein assemblies. Mol Syst Biol. 2021;17(5):e10016. Publisher: John Wiley & Sons, Ltd. https://doi.org/10.15252/msb.202010016.

  90. Garcia-Alonso L, Holland CH, Ibrahim MM, Turei D, Saez-Rodriguez J. Benchmark and integration of resources for the estimation of human transcription factor activities. Genome Res. 2019;29(8):1363–75. Company: Cold Spring Harbor Laboratory Press Distributor: Cold Spring Harbor Laboratory Press Institution: Cold Spring Harbor Laboratory Press Label: Cold Spring Harbor Laboratory Press Publisher: Cold Spring Harbor Lab. https://doi.org/10.1101/gr.240663.118.

  91. Seal RL, Braschi B, Gray K, Jones TEM, Tweedie S, Haim-Vilmovsky L, et al. Genenames.org: the HGNC resources in 2023. Nucleic Acids Res. 2023;51(D1):D1003–9. https://doi.org/10.1093/nar/gkac888.

  92. Csardi G, Nepusz T. The Igraph Software Package for Complex Network Research. InterJournal. 2005;Complex Systems:1695.

  93. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al. Cytoscape: A Software Environment for Integrated Models of Biomolecular Interaction Networks. Genome Res. 2003;13(11):2498–504. Company: Cold Spring Harbor Laboratory Press Distributor: Cold Spring Harbor Laboratory Press Institution: Cold Spring Harbor Laboratory Press Label: Cold Spring Harbor Laboratory Press Publisher: Cold Spring Harbor Lab. https://doi.org/10.1101/gr.1239303.

  94. Luna A, Shah O, Sander C, Shannon P. cyjShiny: A cytoscape.js R Shiny Widget for network visualization and analysis. PLoS ONE. 2023;18(8):e0285339. Publisher: Public Library of Science. https://doi.org/10.1371/journal.pone.0285339.

  95. Carraro G, Langerman J, Sabri S, Lorenzana Z, Purkayastha A, Zhang G, et al. Transcriptional analysis of cystic fibrosis airways at single-cell resolution reveals altered epithelial cell states and composition. Nat Med. 2021;27(5):806–14. Bandiera_abtest: a Cg_type: Nature Research Journals Number: 5 Primary_atype: Research Publisher: Nature Publishing Group Subject_term: Diseases;Respiratory tract diseases Subject_term_id: diseases;respiratory-tract-diseases. https://doi.org/10.1038/s41591-021-01332-7.

  96. Thurman AL, Li X, Villacreses R, Yu W, Gong H, Mather SE, et al. A Single-Cell Atlas of Large and Small Airways at Birth in a Porcine Model of Cystic Fibrosis. Am J Respir Cell Mol Biol. 2022;66(6):612–22. Publisher: American Thoracic Society - AJRCMB. https://doi.org/10.1165/rcmb.2021-0499OC.

  97. Okuda K, Dang H, Kobayashi Y, Carraro G, Nakano S, Chen G, et al. Secretory Cells Dominate Airway CFTR Expression and Function in Human Airway Superficial Epithelia. Am J Respir Crit Care Med. 2021;Publisher: American Thoracic Society. https://doi.org/10.1164/rccm.202008-3198OC.

  98. Duan Y, Li G, Xu M, Qi X, Deng M, Lin X, et al. CFTR is a negative regulator of \(\gamma \delta\) T cell IFN-\(\gamma\) production and antitumor immunity. Cell Mol Immunol. 2021;18(8):1934–44. Number: 8 Publisher: Nature Publishing Group. https://doi.org/10.1038/s41423-020-0499-3.

  99. Stelzer G, Rosen N, Plaschkes I, Zimmerman S, Twik M, Fishilevich S, et al. The GeneCards Suite: From Gene Data Mining to Disease Genome Sequence Analyses. Curr Protoc Bioinforma. 2016;54(1):1.30.1–1.30.33. _eprint: https://doi.org/10.1002/cpbi.5.

  100. Keating D, Marigowda G, Burr L, Daines C, Mall MA, McKone EF, et al. VX-445-Tezacaftor-Ivacaftor in Patients with Cystic Fibrosis and One or Two Phe508del Alleles. N Engl J Med. 2018;379(17):1612–20. Publisher: Massachusetts Medical Society _eprint: https://doi.org/10.1056/NEJMoa1807120.

  101. Santos L, Nascimento R, Duarte A, Railean V, Amaral MD, Harrison PT, et al. Mutation-class dependent signatures outweigh disease-associated processes in cystic fibrosis cells. Cell Biosci. 2023;13(1):26. https://doi.org/10.1186/s13578-023-00975-y.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  102. Saint-Criq V, Gray MA. Role of CFTR in epithelial physiology. Cell Mol Life Sci. 2017;74(1):93–115. https://doi.org/10.1007/s00018-016-2391-y.

    Article  CAS  PubMed  Google Scholar 

  103. Simões FB, Kmit A, Amaral MD. Cross-talk of inflammatory mediators and airway epithelium reveals the cystic fibrosis transmembrane conductance regulator as a major target. ERJ Open Res. 2021;7(4):00247–2021. https://doi.org/10.1183/23120541.00247-2021.

    Article  PubMed  PubMed Central  Google Scholar 

  104. Valdivieso AG, Dugour AV, Sotomayor V, Clauzure M, Figueroa JM, Santa-Coloma TA. N-acetyl cysteine reverts the proinflammatory state induced by cigarette smoke extract in lung Calu-3 cells. Redox Biol. 2018;16:294–302. https://doi.org/10.1016/j.redox.2018.03.006.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  105. Wang Y, Tang L, Yang L, Lv P, Mai S, Xu L, et al. DNA Methylation-Mediated Low Expression of CFTR Stimulates the Progression of Lung Adenocarcinoma. Biochem Genet. 2022;60(2):807–21. https://doi.org/10.1007/s10528-021-10128-w.

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

The authors would like to than Benoît Chevalier and Alexandre Hinzpeter for fruitful discussions on CFTR interactome, and Marco Ruscone for fruitful discussions on biological databases.

Funding

Fondation pour la Recherche Médicale (FRM), Vaincre La Mucoviscidose (VLM), La Fondation Dassault Systèmes, Fondation Maladies Rares, and MSD Avenir

Author information

Author notes

  1. Laurence Calzone and Véronique Stoven contributed equally to this work.

Authors and Affiliations

  1. Center for Computational Biology (CBIO), Mines Paris-PSL, 75006, Paris, France

    Matthieu Najm, Loredana Martignetti, Matthieu Cornet, Laurence Calzone & Véronique Stoven

  2. Institut Curie, Université PSL, 75005, Paris, France

    Matthieu Najm, Loredana Martignetti, Matthieu Cornet, Laurence Calzone & Véronique Stoven

  3. INSERM U900, 75005, Paris, France

    Matthieu Najm, Loredana Martignetti, Matthieu Cornet, Laurence Calzone & Véronique Stoven

  4. Institut Necker Enfants Malades, INSERM U1151, 75015, Paris, France

    Matthieu Cornet, Mairead Kelly-Aubert & Isabelle Sermet

  5. Université Paris Cité, 75015, Paris, France

    Mairead Kelly-Aubert & Isabelle Sermet

  6. Centre de Référence Maladies Rares, Mucoviscidose et Maladies Apparentées, Hôpital Necker Enfants Malades AP-HP Centre Paris Cité, 75015, Paris, France

    Isabelle Sermet

Authors
  1. Matthieu Najm
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Loredana Martignetti
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Matthieu Cornet
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Mairead Kelly-Aubert
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Isabelle Sermet
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Laurence Calzone
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Véronique Stoven
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

I.S. and V.S. initiated the project and obtained funding for the project. M.N., L.M., L.C. and V.S. contributed to the conception and design of the research. M.N. and L.M. designed the transcriptomic analysis. M.N. conducted the transcriptomic analyses, built the signalling network and performed network analyses under the supervision of L.C. and V.S.. M.C., M.K.A. and I.S. contributed to the interpretation of the data. M.N., L.C. and V.S. wrote the manuscript. L.M. I.S. and M.K.A edited the manuscript. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to Matthieu Najm, Laurence Calzone or Véronique Stoven.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

12864_2024_10752_MOESM1_ESM.pdf

Additional file 1: A PDF document with extended results concerning the sink nodes and their corresponding cellular phenotypes. Table S1: The 35 sink nodes of the CF network and their associated cellular phenotypes. Table S2: The top 30 proteins in the CF network according to their betweenness centrality scores.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Najm, M., Martignetti, L., Cornet, M. et al. From CFTR to a CF signalling network: a systems biology approach to study Cystic Fibrosis. BMC Genomics 25, 892 (2024). https://doi.org/10.1186/s12864-024-10752-x

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12864-024-10752-x

Keywords

BMC Genomics

ISSN: 1471-2164

Contact us