Abstract Marfan syndrome (MFS) is associated with mutations in fibrillin-1 that predispose afflicted individuals to progressive thoracic aortic aneurysm (TAA) leading to dissection and rupture of the vessel wall. Here we combined computational and experimental approaches to identify and test FDA-approved drugs that may slow or even halt aneurysm progression. Computational analyses of transcriptomic data derived from the aortas of MFS patients and MFS mice (Fbn1^mgR/mgR mice) predicted that subcellular pathways associated with reduced muscle contractility are key TAA determinants that could be targeted with the GABA[B] receptor agonist baclofen. Systemic administration of baclofen to Fbn1^mgR/mgR mice validated our computational prediction by mitigating arterial disease progression at the cellular and physiological levels. Interestingly, baclofen improved muscle contraction–related subcellular pathways by upregulating a different set of genes than those downregulated in the aorta of vehicle-treated Fbn1^mgR/mgR mice. Distinct transcriptomic profiles were also associated with drug-treated MFS and wild-type mice. Thus, systems pharmacology approaches that compare patient- and mouse-derived transcriptomic data for subcellular pathway–based drug repurposing represent an effective strategy to identify potential new treatments of human diseases. Keywords: Therapeutics Keywords: Bioinformatics, Cardiovascular disease, Extracellular matrix __________________________________________________________________ Analysis of transcriptomic data from patients with Marfan syndrome and a mouse model of the disease identified baclofen as an effective drug treatment of thoracic aneurysm. Introduction The therapeutic promises of precision medicine largely rely on our ability to distinguish molecular determinants of pathogenesis from correlative, compensatory, or other concurrent physiological processes ([55]1). Diseases like cancer that arise from multiple genomic alterations can often be defined by individual determinants that enable targeted therapies ([56]2). However, the situation is more complicated for other progressive diseases. Many molecular determinants can control disease progression even in monogenic disorders, thus requiring systems-level approaches to identify underlying mechanisms. Systems pharmacology views disease as a malfunction of integrated molecular interactions — i.e., networks of multiple subcellular pathways (SCPs) — and drug treatment as a method to restore physiological response of disease-centered networks by targeting one or more of their components ([57]3). Analyses of transcriptomic data for underlying SCPs provides an unbiased strategy to identify biological targets in disease-related regulatory networks, and then use this information to predict known drugs and compounds that can be repurposed for new therapeutic applications. Thoracic aortic aneurysms (TAAs) are life-threatening pathologies characterized by progressive vessel dilation associated with smooth muscle cell (SMC) dysfunction, occasional localized inflammatory infiltrates, and severe maladaptive extracellular matrix (ECM) remodeling that, together, predispose the arterial wall to dissection and rupture leading to premature death ([58]4). Consistent with the progressively degenerative nature of the disease, inherited forms of TAA are accounted for by mutations in molecules involved in supporting tissue integrity and homeostasis, such as components of the ECM, SMC contractile apparatus and its mediators, and TGF-β signaling pathways ([59]4). Yet, the molecular mechanisms and SCPs that translate these different genetic defects into a seemingly identical pathology are not fully understood. As a result, prevention of untimely death from TAA complications currently relies on early detection by imaging and prophylactic repair by surgery. Unfortunately, disease progression is highly heterogeneous, clinical outcome is unpredictable, therapeutic options are limited, and surgical intervention carries significant morbidity/mortality risks. TAA with ensuing dissection and rupture of the vessel wall is the clinical hallmark of Marfan syndrome (MFS), a relatively common connective tissue disease associated with mutations in the gene that codes for the multifunctional ECM glycoprotein fibrillin-1 ([60]4, [61]5). Fibrillin-1 assemblies (microfibrils and elastic fibers) impart specific physical properties to tissues, distribute mechanical forces within and across them, communicate to multiple types of vessel wall cells through integrin receptors, and modulate local bioavailability of ECM-bound latent TGF-β complexes ([62]5). In spite of significant research effort, the molecular pathogenesis of arterial disease in MFS remains unresolved, therefore hindering advances in drug therapy. Earlier studies of MFS mice with nondissecting TAA (Fbn1^C1039G/+ mice) have correlated aneurysm onset and progression with increased TGF-β signaling in the media stimulated by improper angiotensin II (AngII) type I receptor (AT1r) activity ([63]6, [64]7). More recent findings indicate a more complex disease mechanism involving the gradual stratification of stress-stimulated interactions among different cell types and multiple regulatory pathways, of which the AT1r and TGF-β signaling pathways are a critical subset ([65]8–[66]14). An overview of regulatory pathways and networks associated with a given pathology can often be obtained by examining changes in gene expression during disease progression. Systems pharmacology approaches that consider drug targets as nodes within cellular regulatory networks can use differentially expressed genes (DEGs) to predict dysregulated SCPs that underlie cell-level mechanisms ([67]1, [68]3). Further, computational analyses of the pharmacologically induced perturbations of gene expression listed in the Connectivity Map (CMap) database can predict drugs to be repurposed to normalize dysregulated SCPs ([69]15). By revealing that SCPs associated with muscle cell contractility are downregulated in both aortic SMCs isolated from MFS patients and aortic tissue harvested from MFS mice, this strategy enabled us to predict that the GABA[B] receptor agonist baclofen could restore normal activity of these TAA-related SCPs. To test our computational prediction, we performed a series of in vivo and ex vivo analyses that demonstrated a statistically significant mitigation of TAA pathology at the cellular and physiological levels in baclofen- versus vehicle-treated MFS mice. Hence, we conclude that alterations in muscle contractility processes contributing to TAA development in MFS can be therapeutically targeted through GABA[B] receptors. Results Our study was organized into 3 successive lines of investigation. First, we identified shared SCPs from transcriptomic analyses of SMCs isolated from the dilated aorta of MFS patients and aortic tissue harvested from Fbn1^mgR/mgR mice, a validated animal model of early-onset progressively severe MFS ([70]16). Second, we used the gene expression profiles listed in the CMap database ([71]15) to predict FDA-approved drugs that could normalize the dysregulated SCPs in common between human and mouse aortic samples. Third, we tested a top-ranked drug prediction for its ability to modify TAA pathology in Fbn1^mgR/mgR mice and determined if the drug acted through the subcellular mechanisms predicted to be shared by the diseased aortas of MFS patients and MFS mice. Impaired SMC contractility is associated with TAA in both MFS patients and mice. RNA sequencing (RNA-Seq) was used to identify genes differentially expressed in the aortas of Fbn1^mgR/mgR (MFS) mice versus wild-type (WT) littermates, and in aortic SMCs of MFS versus non-MFS patients ([72]Figure 1, A and B, and [73]Supplemental Tables 1 and 2; supplemental material available online with this article; [74]https://doi.org/10.1172/jci.insight.127652DS1). MFS patients and mice were comparable with respect to the phenotypic severity as opposed to the disease stage in which the tissues had been harvested. Gene Ontology (GO) SCP enrichment analysis of upregulated and downregulated genes revealed that SCPs related to actin cytoskeleton dynamics and muscle contractility were top-ranked in the downregulated genes of both human SMCs and mouse aortas ([75]Figure 1, C and D, and [76]Supplemental Tables 3 and 4). By contrast, a similar analysis found no top-ranked disease-relevant SCPs in common between the upregulated genes of human and mouse aortic specimens ([77]Supplemental Figure 1, A and B). Using the top-ranked downregulated SCPs associated with muscle contraction as seed nodes in the GO topology, we identified all related SCP ancestors and offspring that contained at least 20 genes and were also related to SMC contractility ([78]Figure 1E and [79]Supplemental Table 5). Importantly, network analysis revealed that the mouse (blue) and the human (green) downregulated subnetworks shared 5 (seed or intermediate) GO-SCPs related to muscle contractility ([80]Figure 1E). As our study was being completed, Lino Cardenas at al. ([81]17) reported that genes coding for contractile proteins are similarly repressed in the aorta of MFS mice with nondissecting TAA. Figure 1. DEG analyses correlate defective muscle contractility with TAA progression in MFS mice and patients. [82]Figure 1 [83]Open in a new tab Volcano plots of DEGs in (A) aortas of MFS mice sacrificed at P16 (n = 3 mutant mice and n = 3 WT mice) and (B) aortic SMCs of MFS patients aged between 22 and 32 years (n = 3 MFS patients and n = 3 non-MFS organ transplant donors). The log[2] of fold change is plotted against the negative log[10] of the P value; blue/green dots indicate significantly downregulated genes, and orange dots indicate significantly upregulated genes. Dots above the horizontal gray line indicate genes whose minus log[10](P value) was calculated as infinity (since its P value was below the lowest number that is supported by Cuffdiff and therefore given as zero). Blue/orange or green/orange numbers give total counts of significantly down- or upregulated genes in mouse and human samples, respectively. Top-ranked (according to P values) SCPs related to muscle contractility (blue or green boxes) from GO enrichment analysis of mouse aorta (C) or human SMC (D) downregulated genes. Integration of the enrichment results into GO-SCP topology identifies 2 muscle contractility-related subnetworks (E) that have 5 SCPs in common. SPCs predicted to be downregulated in MFS aortas and human aortic SMCs are in blue and green, respectively, and intermediate SCPs are in white. Solid lines connect parents with their children processes (“is_a” or “part_of” GO relationship), dashed lines indicate regulatory relationships (“regulates” GO relationship). Regulatory SCPs are visualized as rectangles, all other SCPs as circles. Identity of network components can be found in [84]Supplemental Table 5. Computational identification of baclofen as a potential anti-TAA drug. Next, we screened the CMap gene expression database ([85]15) to predict FDA-approved drugs that could upregulate the 5 muscle contractility-related SCPs found in common between the downregulated subnetworks of the mouse and human aortic specimens ([86]Figure 1E). Each of the CMap drugs had originally been tested under several different conditions (i.e., cancer cell lines, drug concentration, and treatment length) listed in the database as separate experiments. We subjected the upregulated human genes in the CMap database to GO-SCP enrichment analysis under the assumption that SCPs most significantly enriched in upregulated genes best reflect the main action of a given drug. It follows that a drug was not considered further if the associated top-ranked SCP was not disease relevant, even if a high-significance disease-relevant SCP was low-ranked. This approach identified 12 candidate drugs for which upregulation of a disease-relevant SCP was predicted at the top rank ([87]Figure 2A and [88]Supplemental Table 6); of them, only 4 were considered further because they had an annotated protein target in the database ([89]Figure 2B and [90]Supplemental Table 7). Based on the relatively high expression of their respective targets in the aorta of MFS mice, this shorter list of top-ranked drugs was narrowed down to the steroid fludrocortisone and the muscle relaxant baclofen ([91]Figure 2B). However, fludrocortisone was excluded from further analyses because of its potential to induce iatrogenic arterial hypertension, the major risk factor for thoracic aortic disease ([92]14). Figure 2. Baclofen is a predicted potential treatment for TAA. [93]Figure 2 [94]Open in a new tab (A) Top-ranked potential drug candidates identified from GO-SCP enrichment analysis of upregulated genes in the CMap database. Bars indicate the minus log[10] (P values) of the identified treatments, numbers correspond to the SCPs in the subnetworks of [95]Figure 1E. (B) Target proteins of the indicated drugs annotated in the Drug Bank database (see also [96]Supplemental Table 7); green bar graphs on the right indicate the expression levels of the corresponding mouse genes in the P16 MFS aorta. (C) GO-SCPs identified in the top-ranked list of SCPs derived from baclofen-stimulated genes in the annotated drug treatment baclofen number 2036 of the CMap database (see [97]Supplemental Figure 2A and B). (D) Top-ranked muscle contractility–related GO-SCPs in the combined list of upregulated genes in all baclofen treatments of the CMap database. In panels C and D, muscle contractility–related SCPs are in orange. Baclofen is a highly selective drug for GABA[B] receptors, as evidenced by 17 supporting references in the Drug Bank database