Long non‐coding RNAs influence the transcriptome in pulmonary arterial hypertension: the role of PAXIP1‐AS1

Abstract In idiopathic pulmonary arterial hypertension (IPAH), global transcriptional changes induce a smooth muscle cell phenotype characterised by excessive proliferation, migration, and apoptosis resistance. Long non‐coding RNAs (lncRNAs) are key regulators of cellular function. Using a compartment‐specific transcriptional profiling approach, we sought to investigate the link between transcriptional reprogramming by lncRNAs and the maladaptive smooth muscle cell phenotype in IPAH. Transcriptional profiling of small remodelled arteries from 18 IPAH patients and 17 controls revealed global perturbations in metabolic, neuronal, proliferative, and immunological processes. We demonstrated an IPAH‐specific lncRNA expression profile and identified the lncRNA PAXIP1‐AS1 as highly abundant. Comparative transcriptomic analysis and functional assays revealed an intrinsic role for PAXIP1‐AS1 in orchestrating the hyperproliferative and migratory actions of IPAH smooth muscle cells. Further, we showed that PAXIP1‐AS1 mechanistically interferes with the focal adhesion axis via regulation of expression and phosphorylation of its downstream target paxillin. Overall, we show that changes in the lncRNA transcriptome contribute to the disease‐specific transcriptional landscape in IPAH. Our results suggest that lncRNAs, such as PAXIP1‐AS1, can modulate smooth muscle cell function by affecting multiple IPAH‐specific transcriptional programmes. © 2018 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of Pathological Society of Great Britain and Ireland.


Introduction
Pulmonary arterial hypertension (PAH) is a severe and progressive disease which ultimately leads to right heart failure [1]. Within the vessel wall, multiple factors contribute to the increased pulmonary pressure, including cellular hyperplasia and extracellular matrix (ECM) deposition [2,3]. In the medial layer, smooth muscle cells (SMCs) show increased proliferation [4], invasive migratory capacity, and resistance to apoptosis [5]. These functional alterations are accompanied by a reprogramming of cellular respiration [6] and glucose expenditure [7]. PAH has a strong genetic predisposition, with several mutations contributing to the disease development [8]. Loss-of-function mutations in BMPR2 occur in over 70% of patients with familial PAH and in 25% of patients with the idiopathic form [9]. However, BMPR2 mutations show only 20% penetrance [10]. Thus, it is apparent that additional regulatory mechanisms are involved and factors other than solely mutations in coding genes are required for disease development.
The human genome contains approximately equal numbers of coding and non-coding genes (GENCODE Release, version 29). Despite being highly abundant, non-coding genes are still under-studied. The

K Jandl and H Thekkekara Puthenparampil et al
non-coding transcriptome includes many regulatory transcripts, which can be roughly classified into small non-coding RNAs (sncRNAs, < 200 nt), including the group of microRNAs (miRNAs), and long non-coding RNAs (lncRNAs, > 200 nt) [11]. Variations in disease complexity and penetrance of genetic variants are increasingly attributed to changes in gene expression of the non-coding genome, rather than to changes within protein-coding sequences [12].
In contrast to sncRNAs, which act via RNA binding, lncRNAs exert their regulatory effects, both at the transcriptional and at the post-transcriptional levels, via interactions with DNA, chromatin, and other RNA species [13]. In addition, lncRNAs can bind chromatin-modifying proteins such as histone modifiers, and thus act as epigenetic modulators [14,15]. In this respect, lncRNAs often regulate fundamental cellular processes, such as proliferation or apoptosis [16,17]. Due to the tight transcriptional regulation of lncRNAs and their specific subcellular localisation, a 'long non-coding' landscape can be generated, which is highly cell-type, tissue, and disease-specific [11,18,19].
In the field of pulmonary vascular remodelling, sncR-NAs such as miRNAs have been intensively studied and constitute well-recognised cellular control mechanisms [20][21][22][23][24]. In contrast, lncRNAs are under-investigated in this field. In the present study, we have characterised the gene expression profiles of coding and non-coding genes from small pulmonary arteries obtained via laser-capture microdissection (LCM) from IPAH and control lungs. We focus on the functional role of a highly deregulated lncRNA, PAXIP1-AS1, in perpetuating the migratory and proliferative phenotype of IPAH pulmonary artery smooth muscle cells (PASMCs).

Genome-wide expression profiling
Briefly, whole genome expression profiling was performed on material obtained from small pulmonary arteries (50-

Subcellular fractionation
RNA for the subcellular fractioning was obtained from PASMCs seeded at a density of 8 × 10 3 cells/cm 2 and allowed to grow to 80% confluency in 10-mm Petri dishes. The precipitation and subsequent purification of the RNA were performed as described elsewhere [33].

IPAH small remodelled arteries possess a distinct coding and long non-coding transcriptional profile
One hallmark of PAH is the remodelling of small distal pulmonary arteries [34]. We defined the gene expression profile in small pulmonary arteries of 18 IPAH patients and 17 controls. Based on a minimum significance (-log 10 p) of 3 and a minimum absolute log 2 fold change (LFC) of 1.25, 124 annotated protein-coding genes were classified as up-(76) or down-regulated (48) in IPAH compared with control donor samples ( Figure 1A and supplementary material, Table S1). A KEGG pathway analysis following gene set enrichment on the entire dataset indicated a perturbation in metabolic, immunological, neuronal, and proliferative processes in IPAH ( Figure 1B and supplementary material, Figure S1A). Connections between regulated genes of the top ten perturbed KEGG pathways are displayed in the supplementary material, Figure S1B. The individual profiles of the 50 genes with the highest significance are shown in the heatmap in Figure 1C. Gene ontology assigned these genes to several nodes, which were overrepresented in IPAH; these predominantly belonged to the groups of phospholipid and nitrogen metabolism, proliferation, immunological and neuronal responses, signal transduction (involving tyrosine receptor kinases and small GTPases), and transcriptional regulation ( Figure 1D). In addition to protein-coding genes, we identified a large number of differentially regulated non-coding RNAs. This heterogeneous group included long intergenic non-coding (linc) RNAs, antisense (as) RNAs, pseudogenes, and other long and small ncRNAs, as depicted in Figure 1E. By excluding non-annotated transcripts and transcripts shorter than 200 nt, we identified a total of 146 regulated non-coding genes (supplementary material, Table S2). Changes in the expression of non-coding RNA genes thus contribute substantially to the unique transcriptomic landscape of IPAH vessels ( Figure 1F). The lncRNA profiles of IPAH and donor groups were clearly distinct ( Figure 1G). The differential expression of seven of the ten most regulated lncRNAs ( Figure 1H) was analysed by qRT-PCR. The lncRNAs TUSC8 and PAXIP1-AS1 were verified to be upregulated in IPAH ( Figure 1I). As PAXIP1-AS1 was previously reported to be involved in proliferative events [35], we selected it for further analysis. Following a t-stochastic neighbouring embedding (t-SNE) dimension reduction of all detected genes, PAXIP1-AS1 expression levels at single patient level were visualised in a colour-by-expression code. Indeed, the heatmap overlay of PAXIP1-AS1 expression resembles the clear discrimination of IPAH and donor transcriptome in the t-SNE plot ( Figure 1J).

PAXIP1-AS1 expression is enhanced in human IPAH-PASMCs and knockdown reveals its involvement in focal adhesion and ECM-receptor interaction
PAXIP1-AS1 comprises a single exon located at chromosome 7q36.2 (position: 155003433-155005703; NC_000007.14, GRCh38.p12 according to the NCBI database). It is flanked on its 5 ′ side by a GC-rich region that separates it from its diverging coding gene PAXIP1 on the opposite strand. To identify whether PAXIP1-AS1 expression is limited to the lung, we explored its expression across multiple human tissues using the Genotype-Tissue Expression (GTEx) portal. Low ubiquitous expression was observed in multiple tissues, with a notable enrichment in the cerebellum (supplementary material, Figure S2A,B). Closer examination of PAXIP1-AS1 expression by qRT-PCR revealed that PAXIP1-AS1 was enriched in pulmonary arteries without adventitia, compared with pulmonary veins, bronchi, and pulmonary arterial adventitia (supplementary material, Figure S2C). Furthermore, in isolated cells, PAXIP1-AS1 expression was highest in parenchymal fibroblasts and PASMCs, followed by adventitial fibroblasts and pulmonary arterial endothelial cells (supplementary material, Figure S2D).
Although PAXIP1-AS1 harbours a putative open-reading frame, online coding potential calculators suggested only a weak coding potential (supplementary material, Figure S3A,B and Figure 2A). PAXIP1-AS1 structure prediction resulted in a complex alignment and its high stability given by the calculated free energy suggests a functional importance ( Figure 2B). We therefore investigated the transcript for functional sequences  to further delineate possible biological functions. Several Alu elements, SRP RNA, A-and C-repeats, and an AU-rich element were identified (Figure 2A and  supplementary material, Table S3). Furthermore, the predictions suggest several binding sites for miRNA as well as transcription factors (supplementary material, Tables S4 and S5). Like most lncRNAs [36], PAXIP1-AS1 lacks conservation across species beyond primates.
To identify which cells in the pulmonary artery are responsible for increased PAXIP1-AS1 levels, we visualised its expression in human IPAH tissue using fluorescence in situ hybridisation (FISH). PAXIP1-AS1 was expressed in the lung with abundance in α-smooth muscle actin-expressing cells ( Figure 2C and supplementary material, Figure S4A). The expression of PAXIP1-AS1 was further validated in PASMCs isolated from donors and IPAH ( Figure 2D,E and supplementary material, Figure S4B). Interestingly, a similar increase was also observed in adventitial, but not parenchymal, IPAH fibroblasts (supplementary material, Figure S2E).
The subcellular localisation of lncRNAs is critical for determining their functional properties. Both FISH and subcellular fractionation experiments detected PAXIP1-AS1 in both the nuclear and the cytoplasmic compartments ( Figure 2D,F). Next, we checked whether PAXIP1-AS1 transcription can be modulated by mediators that are involved in PAH pathogenesis. Treatment of PASMCs with endothelin-1 (ET-1) suggested a susceptibility of PAXIP1-AS1 expression to the endothelial-derived mediator (supplementary material, Figure S5).
To delineate how PAXIP1-AS1 can influence PASMC function, we analysed the gene expression profile following GapmeR-mediated PAXIP1-AS1 knockdown. A clear change in the transcriptional programme was observed, as highlighted by the volcano plot ( Figure 3A and supplementary material, Table S6) and in the heatmap representation of the top 100 regulated genes ( Figure 3B). Notably, the profiles of untransfected samples were nearly identical to scrambled control-treated samples ( Figure 3B). To gain insights into the global changes induced after PAXIP1-AS1 knockdown in PASMCs, we performed a KEGG pathway analysis of all regulated transcripts. PAXIP1-AS1 silencing perturbed protein turnover, cytoskeletal arrangement at focal adhesions, extracellular matrix, and pathways implicated in metabolic and proliferative processes ( Figure 3C). The genes involved in these pathways are shown in Figure 3D and supplementary material, Table S7. Importantly, PXN, encoding the focal adhesion adaptor protein paxillin, was one of the top ten regulated genes in two different KEGG pathways (highlighted in Figure 3D). To delineate how PAXIP1-AS1 could exert its effects in IPAH, we investigated potential downstream effectors in IPAH arteries. To this end, we directly compared genes classified as differentially expressed from our two transcriptomic profiling approaches: (1) LCM arteries from IPAH versus donor, and (2) PASMCs after PAXIP1-AS1 knockdown. This comparison revealed 61 genes as regulated in common in both approaches ( Figure 3E and supplementary material, Table S8). Next, we narrowed the number of genes to 32 by selecting only genes that were regulated in opposite directions, thereby identifying a potential PAXIP1-AS1-regulated gene set in IPAH. Amongst those, several were implicated in proliferation and apoptosis processes, e.g. the cyclin G-associated kinase (GAK) and cullin 1 (CUL1) ( Figure 3F).

PAXIP1-AS1 inhibition promoted apoptosis and inhibited PASMC proliferation and migration via its downstream target paxillin
To substantiate the involvement of PAXIP1-AS1 in proliferation and apoptosis, we used two complementary loss-of-function approaches, GapmeR and siRNA-mediated knockdown ( Figure 4A). Indeed, forced reduction of PAXIP1-AS1 potently reduced proliferation and elevated pro-apoptotic events in PASMCs ( Figure 4B,C). As focal adhesions and ECM-receptor interaction pathways were perturbed after PAXIP1-AS1 knockdown, we also investigated its role in migratory processes. As observed in a scratch wound healing assay, both knockdown approaches of PAXIP1-AS1 resulted in a marked reduction in the migratory potential of PASMCs ( Figure 4D,E).
The PAXIP1-AS1-paxillin axis contributes to the IPAH PASMC phenotype Changes in focal adhesions can contribute to disease progression. To delineate the intrinsic mechanism of PAXIP1-AS1 on changes in focal adhesions, we focused on paxillin as a downstream effector of PAXIP1-AS1 in healthy donor PASMCs. We confirmed that depletion of PAXIP1-AS1 resulted in decreased total and phospho (p)-paxillin protein levels ( Figure 5A,B). Since precise coordination between focal adhesions and the actin cytoskeleton is essential for cell migration, we visualised p-paxillin together with F-actin on donor PASMCs. After PAXIP1-AS1 knockdown, p-paxillin levels were decreased, whereas F-actin levels increased ( Figure 5C,D). Plotting of the cross-sectional F-actin signal suggested cytoskeletal rearrangement and increased stress fibre formation visualised in the number and width of peaks in response to the knockdown of PAXIP1-AS1 ( Figure 5E and supplementary material, Figure S6A,B). This suggests that knockdown of PAXIP1-AS1 and the reduced levels of paxillin trigger a stress response in the cells. Interestingly, the focal adhesion kinase (FAK), another focal adhesion protein, showed a similar tendency to downregulation after knockdown of PAXIP1-AS1 (supplementary material, Figure S7A-C). Next, we translated these findings into the patient context, and investigated paxillin in IPAH PASMCs. We detected increased total and p-paxillin protein levels, but no robust changes in mRNA expression in IPAH PASMCs compared with donors ( Figure 5F,G and supplementary material, Figure S8A). In IPAH PASMCs, the increased levels of p-paxillin were accompanied by an increase in F-actin levels and appeared locally associated with both ventral and distal stress fibres ( Figure 5H-J, and supplementary material, Figure S6C,D). Collectively, this suggested a link between the lncRNA PAXIP1-AS1 and IPAH PASMC function that depends on the cytoskeleton and the focal adhesion protein paxillin.
To further examine this link, we next performed PAXIP1-AS1 knockdown experiments in IPAH PASMCs. Here, PAXIP1-AS1 knockdown robustly reduced total paxillin protein levels, while F-actin levels were unaffected ( Figure 6A-C). Knockdown of PAXIP1-AS1 in IPAH PASMCs led to increased apoptotic susceptibility, which was rescued by simultaneous overexpression of paxillin ( Figure 6D,E and supplementary material, Figure S8B,C). While knockdown of PAXIP1-AS1 robustly decreased total protein paxillin levels, we could not observe consistent changes in the gene expression levels of paxillin (supplementary material, Figure S8C). Next, we overexpressed PAXIP1-AS1 in both donor and IPAH PASMCs ( Figure 6F) and investigated its effect on apoptosis resistance. Overexpression of the lncRNA PAXIP1-AS1 potently decreased the apoptosis susceptibility in both donor and IPAH PASMCs ( Figure 6G,H), suggesting a role in IPAH-related cellular dysfunction ( Figure 6I).

Discussion
To date, transcriptomic profiling of IPAH patients either has often lacked compartment-specific analysis or has been limited by the number of patients analysed. Here, we not only describe an IPAH-specific transcriptome in small remodelled arteries in a comparable large cohort of patients, but also address the potential link between dysfunctional lncRNA regulatory mechanisms and the transcriptional changes implicated in vascular remodelling.
In our compartment-specific gene expression analysis, we identified key transcriptional pathways that were perturbed in IPAH; those included metabolic, neuronal, proliferative, and immunological processes. These processes are tightly interconnected. In IPAH, increased proliferation and migration of SMCs can be associated with metabolic changes [37,38]. Adaptions of metabolic processes might be necessary to keep up with the increased energy expenditure or, vice versa, increased energy utilisation might induce a hyperproliferative phenotype. Interestingly, neuronal [39,40] as well as inflammatory mediators [41] can influence a cell's metabolism and growth response, and immunological as well neuronal components have been described in IPAH [42,43]. Given this tight connection of deregulated pathways in remodelled arteries in IPAH, it is apparent that targeting one mechanism might not be sufficient to induce reverse vascular remodelling. In contrast, upstream control mechanisms which simultaneously affect multiple  transcriptional pathways might be able to overcome this limitation.
LncRNAs have emerged as critical determinants in human diseases. They can regulate gene expression at epigenetic, transcriptional, and translational levels. Thus, changes in the lncRNA expression profile can be associated with pathological alterations within the coding transcriptome. In this respect, lncRNAs have great capacity for gene regulation [44]. In pulmonary vascular remodelling, little is known about the involvement of lncRNAs in disease development and progression. Only a few recent studies have shown their contribution to vascular smooth muscle cell proliferation or apoptosis (e.g. MALAT1 [45] or LnRPT [46]) and only one study has investigated the lncRNA profile in PAH patients using peripheral blood lymphocytes [47]. Here, we have demonstrated that transcriptional reprogramming in small remodelled arteries (intima and media) is associated with a specific lncRNA expression profile in IPAH. Importantly, the lncRNA transcriptome can be highly cell type and disease-specific [48], and thus can be used to stratify disease subtypes or monitor disease progression.
Among many differentially regulated lncRNAs, we identified PAXIP1-AS1 as strongly regulated in IPAH small pulmonary arteries, PASMCs, and adventitial fibroblasts. In general, PAXIP1-AS1 expression was not limited to the lung or any specific cell type, which differs from other reported ncRNAs, e.g. miR206 or miR126 [49][50][51]. Under physiological conditions, PAXIP1-AS1 expression is ubiquitous but comparably very low. Its upregulation during disease therefore suggests a crucial role. In that regard, we identified PAXIP1-AS1 as a critical regulator of PASMC function. The function of a given lncRNA depends on its subcellular localisation and the presence of functional RNA sequences on the transcript. In PASMCs, PAXIP1-AS1 was ubiquitously localised in the nucleus and in the cytoplasm, and its secondary structure harboured several functional elements. PAXIP1-AS1 contained several Alu elements in two separate regions. Alu elements belong to the group of short interspersed repeated sequences (SINEs) and are mostly associated with function in the cytoplasm, where they mediate mRNA stability [52]. The PAXIP1-AS1 transcript also harboured miRNA binding sites. Sponging of miRNAs is another feature that is associated with a cytoplasmic function of lncRNAs. We also identified A-and C-rich tracts on the PAXIP1-AS1 transcript. These tracts can potentially infer transcriptional regulation, either via epigenetic modulation or via interaction with transcription factors (TFs) [53,54]. Indeed, several binding sites for TFs were revealed on the PAXIP1-AS1 transcript. The list included some implicated in PH, such as HIF2α, HIF1α, NFAT, NFκB, and FOXO1 [55]. In general, the interaction of an lncRNA with a TF can be multifunctional; it can either inhibit gene expression by acting as a decoy for TF binding or facilitate gene expression by promoting TF accessibility/recruitment to enhancer elements [56]. Taken together, it is apparent that certain structural RNA elements can exert different functional effects depending on the subcellular localisation of the lncRNA. This spatio-temporal action of lncRNAs also applies to PAXIP1-AS1. In comparison to PASMCs, where PAXIP1-AS1 is located both in the cytoplasm and in the nucleus, it was found to be retained in the nucleus in HEK-293 cells [35]. This resulted in functional characteristics differing from our observation, such as opposite proliferative effects.
In PASMCs, PAXIP1-AS1 knockdown interfered with the PAH-related cellular function by inhibiting migration and proliferation, while increasing apoptotic susceptibility. Accordingly, overexpression of PAXIP1-AS1 in both donor and IPAH PASMCs led to apoptosis resistance. This suggests that PAXIP1-AS1 can potentially modulate PASMC-mediated vascular remodelling events. The key pathways that were dependent on PAXIP1-AS1 and strongly perturbed after knockdown were ECM-receptor interaction and focal adhesions. Both pathways are tightly interconnected, as cell motility depends on a functional interaction of focal adhesions with the surrounding extracellular matrix. Thus, changes in focal adhesions are central to the migratory and proliferative maladaptation of PASMCs in PAH. Indeed, paxillin and other adapter proteins of the focal adhesion complex are strongly associated with IPAH pathology [57][58][59]. Our transcriptomic analysis and complementary cellular assays identified paxillin as a downstream target of PAXIP1-AS1. Knockdown of PAXIP1-AS1 potently reduced paxillin levels and concomitantly induced a cytoskeletal disruption in donor and IPAH PASMCs, which was accompanied by increased apoptotic susceptibility. Overexpression of PXN reversed this effect, and simultaneously increased PAXIP1-AS1 expression levels (supplementary material, Figure S8A). This suggests a reciprocal dependency of the PAXIP1-AS1-paxillin axis in the mediation of the IPAH phenotype. Interestingly, in addition to paxillin, PAXIP1-AS1 knockdown also resulted in decreased FAK levels. In IPAH, it has previously been observed that FAK and paxillin levels are co-regulated [59] -a notion substantiated by this study. Together, this suggests that PAXIP1-AS1 knockdown induced an impairment of the focal adhesion axis, from integrins to actin filaments. While targeting cell motility in vascular remodelling seems promising at first, a loss of function of any member of the focal adhesion complex might be deleterious, due to their involvement in fundamental cellular processes.
In addition to focal adhesion, lncRNA PAXIP1-AS1 also controlled other critical pathways of PAH pathology. Among the most significantly downregulated pathways after PAXIP1-AS1 knockdown in PASMCs were also proliferative processes, axon guidance, and sugar metabolism. Interestingly, these pathways mirror the processes that we identified in the compartment-specific transcriptomic analysis of small remodelled arteries. These findings indicate that (1) these processes are pivotal to the IPAH pathology, (2) IPAH-specific pathological changes are strongly associated with a tightly regulated transcriptional programme, and (3) lncRNAs can act as important regulators in these cellular events in IPAH pathology.
At the moment, it is still unclear whether PAXIP1-AS1 is causative in disease development or merely a consequence of remodelling processes. Two major limitations of this study in answering this question are (1) its lack of sequence conservation in animals, and (2) the use of human transplanted lungs that always reflect end-stage disease.
Taken together, a multitude of pathways and single molecules finally converge in the manifestation of the maladaptive cellular phenotype in IPAH. The identification of lncRNA PAXIP1-AS1 as a critical component of the dysfunctional control machinery adds another piece to the complex patho-mechanism of IPAH.  Table S1. Differentially regulated coding genes in IPAH Table S2. Differentially regulated non-coding genes in IPAH Table S3. Functional regions in PAXIP1-AS1 Table S4. Transcription factor binding sites in the PAXIP1-AS1 transcript Table S5. miRNA target sites in the PAXIP1-AS1 transcript Table S6. PAXIP1-AS1 knockdown: differentially regulated genes Table S7. PAXIP1-AS1 knockdown: list of genes from the top regulated pathways