Smooth Muscle Enriched Long Noncoding RNA (SMILR) Regulates Cell Proliferation

Supplemental Digital Content is available in the text.

V essel wall remodeling is an integral pathological process central to cardiovascular diseases including atherogenesis, plaque rupture, and neointimal hyperplasia-associated vein graft failure and in-stent restenosis. 1,2 Resident vascular smooth muscle cells (VSMCs) are typically quiescent and contractile in the normal physiological state. However, following pathological or iatrogenic vascular injury, the release of cytokines and growth factors from VSMCs, aggregated platelets, and inflammatory cells on the damaged intimal surface leads to phenotypic switching of VSMCs and the adoption of a more synthetic, pro-proliferative, and promigratory state. 3 In the setting of the pathological injury of atherosclerosis, VSMCs not only contribute to the atherogenic process itself but can also engender plaque stabilization through the generation of a thick-capped fibroatheroma. For acute iatrogenic vascular injury, overexuberant proliferation of VSMC subpopulations promotes neointimal hyperplasia leading to luminal narrowing such as seen in vein graft failure or in-stent restenosis. 4 Phenotypic switching of VSMCs and release of cytokines and growth factors are therefore critical in vascular disease, and understanding the mechanisms involved is critical to gain insights into the pathology and identify new opportunities for therapies.

Clinical Perspective on p 2065
The highly conserved interleukin-1α (IL1α) and plateletderived growth factor (PDGF) pathways play prominent roles in VSMC-associated pathologies. 1,5 IL1α is a central mediator in the cytokine cascade and a potent activator of vascular cytokine production. Furthermore, previous studies have demonstrated that ligation injury results in reduced neointimal formation in IL1 receptor knockout mice. 6 Downstream mediators include the signaling molecules MEKK1 and p38 and the transcription factor NF-κB that activate mediators of inflammation and cellular migration. 7 PDGF is a potent mitogen and chemoattractant, and expression is increased following vascular injury. 8 Conversely, a reduction in PDGF expression reduces intimal thickening and cellular content of the neointima. 9 Activation of both IL1α and PDGF signaling pathways simultaneously can activate common downstream targets leading to additive or synergistic effects. This includes activation of NF-κB leading to the upregulation of metalloproteinase 3 and 9, 10 genes critical in the development of vasculoproliferative pathologies.
Over the past decade, there has been substantial interest in determining the complex interactions between hierarchical levels of gene regulation. Up to 90% of the human genome is transcribed at different developmental stages and only ≈2% of RNA molecules are translated into protein. 11 The functional complexity of organisms therefore appears to be reliant on noncoding RNA molecules. Noncoding RNAs are subdivided into several classes, including microRNA (miRNA) and long noncoding RNA (lncRNA). miRNAs are abundantly expressed in vascular tissues and play an important role in vascular pathology. Interestingly, recent studies have demonstrated that miRNAs are capable of being released into the blood from injured cells. These miRNAs are relatively stable and have been reported as biomarkers for several disease states, including myocardial infarction 12 and heart failure. 13,14 Although the role of miRNAs is reasonably established in the setting of cardiovascular pathology, relatively little is known about the role of lncRNAs. lncRNAs are capable of regulating target DNA, RNA, and protein at the pre-and posttranscriptional level. It is becoming clear that lncRNAs play a pivotal role in cellular physiology and pathology via localization in subpopulations of cells and through highly controlled temporal expression. 15 However, detailed insights into their regulation and biological roles are only beginning to emerge. In the vascular setting, SENCR and MALAT1 have been implicated in the control of vascular cell migration and endothelial cell sprouting, respectively. 16,17 Interestingly, SENCR is implicated in phenotypic switching of VSMCs to a more promigratory phenotype because knockdown of this lncRNA downregulates contractile genes. 17 A greater understanding of lncRNAs in quiescent and proliferative VSMCs may provide valuable insight into the specific roles of lncRNAs in response to pathological processes.

Human Tissue Samples
Surplus human saphenous vein tissue was obtained from patients undergoing coronary artery bypass grafting. Carotid plaques were obtained from patients undergoing endarterectomy following an acute and symptomatic neurovascular event. Human plasma samples were used from a previously published study: Carotid Ultrasound and Risk of Vascular disease in Europeans and South Asians (CURVES). 2 All patients gave their written, informed consent. All procedures had local ethical approval (06/S0703/110, 12/WS/0227, 09/S0703/118, and 12/NW/0036). All studies were approved by East and West Scotland Research Ethics Committees, and all experiments were conducted according to the principles expressed in the Declaration of Helsinki.

Tissue and Cell Culture
All cells were maintained at 37°C in a humidified atmosphere containing 5% CO 2 . Primary human saphenous vein-derived endothelial cells (HSVECs) were isolated by a modified version of the protocol described by Jaffe and colleagues 18 and maintained in large-vessel endothelial cell culture medium supplemented with 20% fetal calf serum (Life Technologies, Paisley, UK). Primary human saphenous vein-derived smooth muscle cells (HSVSMCs) were isolated from medial explants 19 and maintained in Smooth Muscle Cell Growth Medium 2 (PromoCell, Heidelberg, Germany) with supplements. Human coronary artery VSMCs were purchased from Lonza (Basel, Switzerland) and maintained in VSMC media as above.

Sample Preparation for RNA-seq Library Construction and Analysis
HSVSMCs were plated, quiesced in medium containing 0.2% fetal calf serum for 48 hours before the stimulation with 10 ng/mL IL1α, 20 ng/mL PDGF (R&D Systems) or a combination of both for 72 hours. Total RNA was processed through miRNeasy kit (Qiagen, Hilden, Germany) following the manufacturer's instructions, treated with DNase 1 (amplification grade; Sigma, St. Louis, MO) to eliminate genomic DNA contamination and quantified by using a NanoDrop ND-1000 Spectrophotometer (Nano-Drop Technologies, Wilmington, DE). Following bioanalyzer quality control for RNA integrity number values >8, RNA sequencing (RNA-seq) was performed on ribosomal-depleted RNA using an Illumina Hiseq platform by Beckman Coulter Genomics. Paired-end sequencing was performed with a read depth of 70 million (n=4/group). RNAseq reads were processed and trimmed to ensure quality, adapter sequences were removed using Flexbar 20 and mapped to the Ensembl annotation of GRCh37.75 using the TopHat2 version 2.0.9. 21 The transcriptome was assembled from the aligned reads and quantified using Cufflinks version 2.2.1. 22 The differential expression levels between the groups were assessed using Cuffdiff version 2.2.1. 23 The data set is deposited in the Gene Expression Omnibus (GEO) repository, study number GSE69637. The biotype of each transcript was annotated according to the Ensembl database. Normalization and statistical analysis of differentially expressed transcripts were performed by using edgeR and data filtered to find transcripts that were differentially expressed (P<0.01) between 0.2% media and each treatment group. Differentially expressed lncRNAs, between control and both IL1α/PDGF treatments, were explored by using more stringent criteria (P<0.01, false discover rate [FDR] <0.01, log fold change>2) and filtered according to transcript abundance (fragments per kilobase of exon per million fragments mapped [FPKM]>1 in at least 1 group). Data outputs such as pie charts and heatmaps were generated using R. Ingenuity pathway analysis was performed by using protein-coding genes differentially expressed (FDR<0.01) from Edge R analysis.

Assessment of RNA Secretion From HSVSMC
RNA extraction on conditioned HSVSMC media was performed by using a standard volume (2 mL). The conditioned media was first centrifuged (10 minutes; 2000g; 4°C) to remove all cells and debris. After addition of 1.4 mL of QIAzol (Qiagen), 3 µL of Caenorhabditis elegans total RNA at 25 ng/µL was added to each sample. Following 5-minute incubation at room temperature, 140 µL of chloroform was added and samples centrifuged (15 minutes; 15 000g; 4°C). The clear upper aqueous phase was used to isolate RNA by using the miRNEasy mini kit (Qiagen) as previously described with alteration of the final wash step (75% ethanol in diethylpyrocarbonate water). Different quantities of total RNA were spiked and a dose-response effect was observed ( Figure

Gene Expression qRT-PCR
For gene expression analysis, cDNA for mRNA analysis was obtained from total RNA using the Multiscribe Reverse Transcriptase (Life Technologies, Paisley, UK). Quantitative real-time polymerase chain reaction (qRT-PCR) was performed using Power SYBR green (Life Technologies) with custom PCR primers (Eurofins MWG, Ebersberg, Germany), the specificity of these primers was confirmed by performing a melting curve and running their PCR produce on a gel ( Table I in the online-only Data Supplement -primer sequences). Ubiquitin C was selected as housekeeping gene because of its stability across all groups studied. Fold changes were calculated by using the 2 -ΔΔ Ct method. 23

Statistical Analysis
Statistical analysis was performed according to figure legends. Data in graphs are shown on relative expression scales as referenced by Livak and Schmittgen. 24 Data are given as both mean±standard deviation (shown as bars and whiskers) and also as the individual points to clearly represent the data. Note that as the relative expression scale is inherently skewed; the bars indicate the geometric mean of the relative expression fold change, and the standard deviation whiskers denote the relative expression fold change equivalent to an increase of 1 standard deviation above the mean on the logtransformed scale. All statistical analysis is performed on the dCt scale (a logarithmic transformation of the data shown on the relative quantification in the plots). 24 No evidence of unequal variances across groups was found for any of analyses of the dCt scale data using the Levene test on Minitab version 17 before statistical analysis. Comparisons between 2 groups were analyzed using 2-tailed unpaired or paired Student t test. One-way analysis of variance with Tukey post hoc or 1-way analysis of variance multiple-comparisons test for pooled samples, via Graph Pad Prism version 5.0, was used for comparisons among ≥3 groups. Statistical significance is denoted by a P value of <0.05.

Induction of Inflammatory and Cell Cycle Pathways by IL1α and PDGF
We sought to identify lncRNAs that are regulated during the induction of proliferative and inflammatory pathways in HSVSMCs. RNAs were identified by using RNA-seq of HSVSMC treated for 72 hours ( Figure 1A). Activation of the IL1α and PDGF signaling pathways was confirmed by the presence of the inflammatory miRNA miR-146a (Figure1B) and induction of VSMC proliferation ( Figure 1C). The RNA sequencing obtained an average of 70 million reads per sample, with 93.5% aligning to the GRCh37 genome reference files. The majority of reads, under all conditions, corresponded to mRNA (49.6±0.48%; Figure 1D and Figure  IIA in the online-only Data Supplement). To identify the biological function, networks, and canonical pathways that were affected by VSMC stimulation, we performed ingenuity pathway analysis after RNA-seq analysis. Ingenuity pathway analysis confirmed that the mRNAs with altered expression following IL1α treatment were significantly enriched in pathways related to cellular movement and inflammatory disease (Table II in the online-only Data Supplement), whereas PDGF stimulation led to the marked enrichment in cell cycle pathways (Table III in the onlineonly Data Supplement). Interestingly, costimulation led to enrichment in cell cycle and cardiovascular development pathways (Table IV in the online-only Data Supplement). Further analysis of differentially expressed mRNAs with a stringent cutoff of FDR<0.01 identified 518 protein-coding genes altered following IL1α treatment and 540 following PDGF treatment. Notably, dual stimulation altered 1133 known protein-coding genes with 480 uniquely associated with dual stimulation and not affected by IL1α or PDGF treatment alone ( Figure 1E and Figure

Identification of Differentially Expressed lncRNAs in HSVSMCs Treated With IL1α and PDGF
We next assessed whether lncRNAs were dynamically regulated by growth factor and cytokine stimulation. Approximately 33% of reads in each condition aligned to known or predicted lncRNAs ( Figure IIIA in the online-only Data Supplement). Differential expression analysis confirmed substantial differences in lncRNA expression between control and stimulated cells. Using the stringent criteria FDR ≤0.01 and log2 fold change ≥2, to declare significance and FPKM >1, to confirm quantifiable expression we identified 224, 215, and 369 differentially expressed lncRNAs following IL1α, PDGF, or dual stimulation, respectively ( Figure  IIIA in the online-only Data Supplement). Because lncRNAs can typically contain multiple splice variants, the numbers quoted refer to a single consensus gene model and therefore do not reflect the multiple transcripts of each lncRNA. To determine whether specific biotypes of lncRNA were enriched following HSVSMC stimulation, those differentially expressed were further subdivided according to biotype in the Ensembl database. LncRNA biotypes are based on their relative orientation to protein-coding genes; intervening lncRNA (lincRNA), antisense, overlapping and processed transcripts. With the use of control and dual stimulation as an example, the distribution of different lncRNA biotypes was as follows: intervening (45.5%), antisense (45.3%), overlapping (1.4%), and processed transcripts (7.9%; Figure Table I

Vascular Enriched Expression of RP11-94a24.1
The expression of each lncRNA was quantified in a range of 10 normal human tissues including specimens derived from brain, gastrointestinal, reproductive, and endocrine systems. In general, lncRNAs were expressed at relatively low levels across the tissue panel. However, we observed that RP11-94a24.1 was expressed highest in the heart, whereas RP11-91K9.1 and AC018647.3 showed preferential expression within the liver and brain, respectively. RP11-709B3.2 and RP11-760H22.2 displayed highest expression in the spleen and thyroid, respectively ( Figure VIA in the onlineonly Data Supplement). We next examined the expression of each lncRNA in primary HSVECs, HSVSMCs, and human coronary artery SMCs. All lncRNAs had higher expression in VSMCs of either venous or arterial lineage than in endothelial cells, suggesting VSMC enrichment ( Figure  VIB in the online-only Data Supplement). We also assessed whether the expression of these lncRNAs could be modulated by IL1α and PDGF in HSVECs as had been found in the HSVSMCs. Notably, subsequent downregulation of  Figure  3B). However, stimulation had no effect on RP11-94a24.1 expression ( Figure 3B), indicating selective regulation in HSVSMCs. Because of the expression of RP11-94a24.1 in HSVSMCs and its cell-specific induction in response to pathological mediators of vascular injury, we focused further studies on RP11-94a24.1. We termed this lncRNA, smooth muscle-induced lncRNA enhances replication (SMILR). SMILR expression was assessed through the use of 3 independent primer sets targeting differential exons of the lncRNA. qRT-PCR revealed similar Ct and fold changes among the 3 sets, further confirming our previous data ( Figure VII in the online-only Data Supplement). The longest open reading frame within SMILR is 57 amino acids. Analysis of this open reading frame using the Coding Potential Calculator (http:// cpc.cbi.pku.edu.cn) did not reveal any similarity to known protein-coding sequences suggesting that this RNA has no protein-coding potential (data not shown).

IL1α/PDGF Treatment Induces the Expression of SMILR in a Time-Dependent Manner
To investigate the longitudinal regulation of SMILR, we stimulated HSVSMCs with PDGF, IL1α, or a combination of both (1.5, 4, 24, 48, and 72 hours). We found significant upregulation of SMILR in response to PDGF as early as 4 hours after stimulation. By 24 hours, SMILR expression was increased by treatment with PDGF or IL1α, and both together, as well ( Figure   We performed RNA FISH to provide visuospatial information as to the location of SMILR within HSVMSCs. Negative control samples showed no fluorescent signal, whereas SNORD3 fluorescent activity confirmed the nuclear

A B
permeabilization of cells ( Figure 4A). In the absence of growth factor and cytokine stimulation, HSVSMCs typically exhibited between 0 and 3 positive fluorescent signals corresponding to SMILR localization ( Figure 4A). IL1α/PDGF treatment induced a marked increase in fluorescent signal within the nucleus and cytoplasm of HSVSMCs. Further specificity of the FISH probes was confirmed through the use of cells treated with either lentivirus containing SMILR or small interfering RNA (siRNA)-targeting SMILR. In each case, an increase and decrease in SMILR transcripts was observed ( Figure 4A). Quantification of FISH samples is provided in Figure 4B. In the absence of stimulation, 2±3.6 SMILR molecules were observed. Following stimulation, 25±5 individual SMILR molecules were observed within the nucleus and cytoplasm ( Figure 4B).

Identifying Upstream Mediators of SMILR Expression in HSVSMCs
It is well established that IL1α and PDGF work through distinct pathways leading to vascular cell activation. To assess the functional consequences of inhibition of these pathways on SMILR expression, selective pharmacological inhibitors AZD6244 (MEKK1) and SB 203580 (p38) were used ( Figure 5A). Following 60 minutes of pretreatment with inhibitors, VSMCs were stimulated with IL1α/PDGF, and the expression of SMILR was determined at 24 hours. Pretreatment with AZD6244 (5, 10, or 15 µmol/L) prevented the induction of SMILR in response to PDGF and IL1α ( Figure 5B), whereas inhibition of p38 with SB203580 induced a dose-dependent reduction in SMILR expression in response to PDGF and IL1α ( Figure 5C).

IL1α/PDGF Treatment Induces the Release of SMILR Into Conditioned Media
miRNAs have been reported to be secreted from cells as a means of cell-to-cell communication. 26 To investigate whether HSVSMCs release SMILR as an indication of expression, we modified a method commonly used to evaluate miRNA expression. 27 Because no endogenous control was stably expressed across all conditions in this study, an exogenous control was added to monitor extraction efficiency and to normalize data. Consequently, total RNA from C elegans was used as a spike-in, and ama-1 encoding polymerase II was chosen as a control for its high constitutive expression (see Methods). Interestingly, SMILR was detected at low levels in media from quiesced VSMCs and those stimulated by either PDGF or IL1α, whereas conditioned media obtained from VSMC stimulated by combination contained significantly higher levels of SMILR (4.8±4.5-fold; Figure 5D), consistent with the increased intracellular expression of SMILR following costimulation of VSMC. Thus, treatment with PDGF and IL1α increased intracellular and released levels of SMILR.
In addition, we sought to identify if SMILR was encapsulated within exosomes or microvesicles. We used both ultracentrifugation, to remove cell debris, and an exosome isolation kit. Figure XIA and XIB in the online-only Data Supplement confirms the presence of microvesicles and exosomes by using Nanosight technology and the expression of the previously described miR-143 within the exosomes/microvesicles. 28 Our data highlighted the expression of SMILR restricted to exosome-free media ( Figure IXC in the online-only Data Supplement) and the inability to detect SMILR expression in the exosome/microvesicle compartment using both techniques of isolation. This observation has been confirmed by agarose gel electrophoresis ( Figure XID in the online-only Data Supplement). Primer melting curves are also shown in Figure XIE in the online-only Data Supplement. Our data confirm that SMILR is secreted into the media and located in a nonexosome/microvesicle fraction. This could possibly be through interaction with specific membrane channels but requires additional experimentation.
In addition, we examined the release of SMILR following lentiviral overexpression in IL1-and PDGF-treated cells. Lentiviral overexpression resulted in a 10-fold increase in SMILR RNA intracellularly. However, only a marginal (not significant) increase was observed within conditioned media analyzed from infected cells ( Figure XIF in the online-only Data Supplement). When this media was transferred onto additional quiesced cells, no change in proliferation was detected ( Figure XIG in the online-only Data Supplement). This may suggest that the release of SMILR is under a stringent control mechanism and simply increasing SMILR expression via a lentiviral approach is not sufficient to induce the additional release of this lncRNA from the cells. In addition, these cells were stimulated with IL1α and PDGF, which strongly enhances SMILR expression in VSMC. The secretory machinery may have been saturated with the high levels of lncRNA within the cytoplasm. This has previously been demonstrated with miRNA where high levels of miR, via overexpression with miRNA mimics, saturated the exportin-5 pathway of endogenous miRNAs with fatal consequences. 29,30

Effect of Dicer Substrate siRNA-Mediated Knockdown of SMILR on HSVSMC Proliferation
We investigated the function of SMILR using dicer substrate siRNA (dsiRNA)-mediated knockdown and 5-ethynyl-2′deoxyuridine incorporation. Forty-eight hours after stimulation, IL1α and PDGF treatment induced a 34±15% increase in VSMC proliferation in comparison with control ( Figure XII in the online-only Data Supplement). dsiRNA SMILR caused 75±24% decrease in SMILR expression in comparison with dsi-control ( Figure 5E). Following SMILR knockdown with dsiRNA, VSMC proliferation was reduced by 56±15% ( Figure  5F). Results were confirmed through the use of a second dsiRNA targeting an alternative region of SMILR ( Figure  XIIIA and XIIIB in the online-only Data Supplement). No effect on the interferon pathway was observed via assessment of the response genes OAS1 and IRF7, which have previously been linked to dsiRNA off-target effects 31 (Figure XIIIC and XIIID in the online-only Data Supplement).
In addition, the effect of SMILR overexpression on VSMC proliferation was investigated. VSMCs were infected with SMILR lentivirus or empty control for 24 hours before stimulation. Infection at a multiplicity of infection of 25 and 50 produced a 5.5±3.5-and 11.4±4.7-fold increase in SMILR expression in comparison with the empty control, with no apparent toxicity effects ( Figure 5G). Overexpression produced a dose-dependent increase of 1.3±0.3-fold and 1.66±0.5-fold in VSMC proliferation, respectively ( Figure  5H), confirming the knockdown data.

SMILR Expression Correlates With Other Proximal Genes
The expression of lincRNAs can correlate with the expression of adjacent genes and other RNAs within the genomic locale. 32 We therefore assessed the expression of genes and noncoding RNAs within 5 million base pairs of SMILR, from COL14A1 on the forward strand to FERIL6-AS1 on the reverse strand ( Figure 6A) by using the RNA-seq data set ( Figure 6B). Upregulation of SMILR was not associated with a widespread increase in transcriptional activity within the region ( Figure 6B). However, similar changes in expression in response to VSMC stimulation were observed in 2 proximal transcripts (HAS2 and HAS2-AS1). SMILR is located ≈750 kbp downstream of HAS2 on the same (reverse) strand and ≈350 kbp from ZHX2 and ≈750 kbp from HAS2-AS1 on the opposite strand of chromosome 8 ( Figure 6C). The upregulation of HAS2 was accompanied by an increase in HAS1 but not HAS3 following dual stimulation ( Figure 6D through 6F). Interestingly, IL1α and PDGF in combination had no effect on HAS3 expression because IL1α and PDGF have opposing effects on HAS3 expression (full graph with single stimulation, Figure XIIIE and XIIIF in the online-only Data Supplement). In addition to SMILR, upregulation of HAS2-AS1 was evident following IL1α and PDGF treatment, but not ZHX1 in the RNA-seq data (data not shown). This observation was validated by qRT-PCR ( Figure 6G through 6I).
It has been previously shown that lncRNA can modulate the expression of nearby protein-coding genes. Thus, the expression of proximal genes HAS2, ZHX2, and HAS2-AS1 was determined following SMILR knockdown. RNA interference-mediated knockdown of SMILR significantly altered levels of HAS2 mRNA. However, no change in the HAS2-AS1 lncRNA or the ZHX2 gene was observed via qRT-PCR ( Figure 6J through 6L). Results were confirmed by using a second siRNA-targeting SMILR ( Figure XIIIG through XIIII in the online-only Data Supplement). In addition, no effect on HAS1 or HAS3 expression was observed while SMILR siRNA was used, indicating that the effect of SMILR knockdown is specific to HAS2 and not all isoforms of HAS ( Figure 6M and 6N).

SMILR Expression Is Dysregulated in Unstable Human Carotid Plaques
To investigate the importance of SMILR in human vascular pathologies, we assessed levels of SMILR in unstable atherosclerotic plaques. We  Figure 7H through 7K for graphs of tracer uptake). Because noncoding RNAs have not been assessed in a similar sample set before, we first confirmed whether expression of a panel of miRNAs associated with atherosclerosis processes were dysregulated. 36 As expected, inflammation-associated miRNAs 146a and 146b were significantly upregulated in unstable plaques in comparison with adjacent quiescent tissue, whereas miR-29 and miR-204, which are inversely associated with osteoblastogenesis and arterial calcification, were downregulated in mineralized regions of the atherosclerotic plaque. 37,38 In addition, we also found a downregulation of the miR-143/145 cluster, which is associated with SMC differentiation and aortic aneurysm formation, 39 an event that has previously been linked to osteogenic differentiation of SMC ( Figure 7L). Thus expression of small noncoding RNAs (miRs) was associated with positron emission tomography/ computed tomography-defined high-risk plaques. Therefore, we used the same cohort of samples to assess SMILR, HAS2, and HAS2-AS1 levels. A 3.9±2.3-fold increase in SMILR expression was observed in high-risk plaques in comparison with adjacent stable regions of the carotid artery ( Figure  7M). Intriguingly, we also observed an increased in HAS2 ( Figure 7N) but not HAS2-AS1 ( Figure 7O).

SMILR Is Detectable in Human Plasma and Correlates With Inflammatory C-Reactive Protein
Because of the release of SMILR into conditioned media from VSMC following stimulation with inflammatory mediators, we evaluated whether SMILR was detectable in stored samples from a cohort of men with varying metabolic dysfunction.
These samples were ranked in order of the serological parameter C-reactive protein (CRP) levels into 3 groups: CRP <2, CRP 2 to 5, and CRP >5 mg/L representing broad tertiles of CRP. SMILR showed no difference in patients with CRP levels below 2 mg/L versus 2 to 5 mg/L. However, a significant increase in SMILR was observed when CRP concentrations were >5 mg/ml (0.008±0.006 for CRP <2 mg/L and 0.046±0.05 for CRP >5 mg/L; Figure 8A). Furthermore, a significant positive correlation was seen between SMILR and CRP (R 2 =0.33, P<0.0001; Figure 8B). There was no correlation between SMILR and additional risk factors including age (P=0.66), blood pressure (P=0.12), BMI (P=0.14), or social deprivation status (P=0.11; Table VI in the online-only Data Supplement). Melting curves and gel products of SMILR primers in plasma are shown in Figure XV in the online-only Data Supplement. Further information regarding the statistical analysis of SMILR CRP correlation can be found in Figure  XVI in the online-only Data Supplement.

Discussion
We have shown that stimulation of HSVSMCs with PDGF and IL1α increases expression of SMILR. This novel lincRNA increases cell proliferation, which may be linked to its ability to regulate the proximal gene HAS2. In a clinical setting, we found increased expression of SMILR in unstable atherosclerotic plaques suggesting an association with fundamentally important vascular pathologies linked to inflammation and VSMC proliferation. We also discovered that SMILR can be released from cells and is detectable in plasma from patients with enhanced inflammation and thus susceptibility to atherosclerosis. These findings support the growing body of evidence that noncoding RNAs can act as mediators to modulate disease pathways. Recent advances in RNA sequencing have demonstrated that previously thought genome deserts are in fact pervasively transcribed and are populated by lncRNAs. The use of paired-end sequencing allowed the accurate alignment of reads to the human genome (GRCh37), the 93% alignment rate met quality standards for the RNA-seq technique 40 and ensured that our RNA-seq provided a high-quality profile of the HSVSMC transcriptome during quiescent and stimulated conditions. Notably, in comparison with control cells, protein-coding genes accounted for 3-to 4-fold greater abundance than lncRNAs. Our RNA-seq depth of 70 million reads was sufficient to identify lncRNAs within VSMC; however, it should be noted that greater read depths and the use of refined capture-sequencing technique would be beneficial to offer greater annotation of specific areas within the genome.
Analysis of the RNA-seq data pinpointed SMILR as an IL1α/PDGF-responsive lincRNA located on chromosome 8, 750 kbp from the closest protein-coding gene, on the same strand. This gene, HAS2, encodes an enzyme that synthesizes hyaluronic acid (HA), a critical component of the extracellular matrix that accumulates in human restenotic and atherosclerotic lesions. 41,42 Our results show that knockdown of SMILR reduces HAS2 expression and VSMC proliferation. This mechanism of action is supported by a number of studies demonstrating that HA can enhance VSMC proliferation and migration. 43 Recent studies using transgenic mice with VSMCspecific overexpression of HA have found increased susceptibility to atherosclerosis 44 and enhanced neointima formation in response to cuff injury. 45 The ability of SMILR to specifically target HAS2 with no effect on HAS1 or HAS3 allows a means of specifically altering HAS2 expression, the main HAS isoform expressed and functioning in SMC pathology. 46 lncRNAs can regulate other RNAs via a number of mechanisms, 47 including changes in chromatin signatures within their locus. For example, the HOTAIR lncRNA is capable of repressing transcription in trans across 40 kbp of the HOXD locus. 48 Thus SMILR may act as an enhancer or scaffold via interaction with the promoter region, and potentially other transcription factors of HAS2, to regulate expression following inflammatory cytokine stimulation. However, further detailed coimmunoprecipitation or site-directed mutagenesis studies would be required to demonstrate whether SMILR participates in transcription factor complexes with NF-κβ or other transcription factors. Previous work has found that HAS2 is regulated by an additional lncRNA, HAS2-AS1. 49 Interestingly, our RNA-seq data show HAS2-AS1 expression was also upregulated by PDGF treatment alone and in combination with IL1α. However, knockdown of SMILR did not alter HAS2-AS1 expression. lncRNA HAS2-AS1 modulates chromatin structures around the gene to allow more efficient binding of the RNA polymerase 2, and enhanced HAS2 gene expression. 49 This suggests both SMILR and HAS2-AS1 can regulate HAS2 by independent mechanisms. Interestingly, knockdown of HAS2 did not affect either SMILR or HAS2-AS1 expression, indicating that the expression of these lncRNAs is not directly linked to HAS2 expression.
The composition of extracellular matrix assists in the determination of the stability of the atherosclerotic plaques, the phenotype of cells within it and the volume of neointima. During the progression of atherosclerosis, VSMCs are exposed to a plethora of signaling molecules, including inflammatory cytokines. Using the clinical gold-standard methods of [ 18 F] FDG and [ 18 F]fluoride positron emission tomography/computed tomography imaging to identify inflamed, necrotic, and mineralizing atherosclerotic plaque, 33,34 our results indicate that miRs 29, 143, 145, 146, and 204 are differentially expressed in unstable regions of atherosclerotic plaques. These miRs have previously been linked to VSMC differentiation, inflammatory cell signaling, 50 and VSMC calcification. 51 The strong association and colocalization of SMILR with this classical miRNA profile and focal [ 18 F]FDG and [ 18 F]fluoride uptake within atherosclerotic plaque suggests that SMILR may play a role in atherosclerosis through inflammatory and proliferative pathways. In keeping with our results showing HAS2 regulation by SMILR, HA content has been shown to reflect the progression of atherosclerotic disease and promotes vessel wall thickening. 52 Indeed, HA has been implicated in the recruitment of inflammatory cells, known to play a prominent role in the initiation and progression of atherosclerotic lesion to an unstable plaque phenotype.
Our results demonstrate that SMILR is upregulated by a combination of PDGF and IL1α in VSMCs but not endothelial cells, suggesting that modulation of SMILR could specifically alter VSMC proliferation without detrimental effects on vessel reendothelialization. If this is the case, it would provide a suitable candidate to improve antiproliferative therapies because current pharmacological agents target cell proliferation in a non-cell-specific manner, events that can lead to late stent thrombosis. 53 The ability to identify confidently a plaque, or patient, at particular risk of a major adverse cardiovascular event (ie, plaque rupture) remains an important goal of cardiovascular research. Long RNAs, both mRNA and noncoding RNA, have been previously shown to be stable in vivo for up to 3 weeks. 54 As such, the search for prognostic biomarkers has greatly increased in recent years. SMILR was expressed in both the nucleus and cytoplasm of cells following stimulation and was released into the media. It will be important to determine whether the cytoplasmic copies induce functional effects, such as regulation of gene expression through posttranslational mechanisms or if they are simply being processed for release. Dual transcriptional functions of lncRNAs have been shown previously, 55 but to date no reports of a single lncRNA affecting both transcription and translation have been published. The release of SMILR could affect function in neighboring cells, particularly in a vascular injury setting where phenotypic switching of VSMCs occurs in distinct areas of the vessel wall. In support of this theory, it has been shown that miR-143/145 can be transferred from VSMC into endothelial cells. 56 This transfer produced physiological effects Figure 8. SMILR is detectable within plasma samples and correlates with patient CRP levels. A, SMILR expression is increased in patients with higher CRP levels (n=13 CRP<2; n=13 CRP2-5; and n=15 CRP>5; *P<0.05, **P<0.01 via 1-way ANOVA). B, Correlation between SMILR expression and CRP levels (linear regression P<0.0001). CRP indicates C-reactive protein; lncRNA, long noncoding RNA; and UBC, ubiquitin.

A B
within endothelial cells, including modulation of angiogenesis. We also found that SMILR could be detected in the plasma of patients with higher CRP levels indicative of chronic low-grade inflammation. In light of our studies, we propose that this release could be attributable to the increased levels of SMILR in the diseased vasculature, although delineating whether plasma SMILR is simply a by-product of increased intracellular levels or is functionally active in disease pathology is difficult to definitely demonstrate. However, circulating levels of miR 143 and 145 are associated with in-stent restenosis and, as such, have been proposed as biomarkers. 57 The correlation of SMILR and high CRP further supports its expression in low-grade chronic inflammatory settings, and proliferative settings, as well. Further large clinical cohorts will be required to ascertain if SMILR has prognostic potential in inflammatory vascular disease and, if so, whether it provides enhanced prognostic value over current biomarkers.
Vessel renarrowing after surgical intervention and atherosclerosis remain significant clinical problems, and HA/HAS2/SMILR have emerged as key components of these pathological processes. Administration of an siRNA targeting SMILR could be used to prevent renarrowing after surgical intervention for acute coronary syndrome. The use of siRNAs has been proven to be effective in phase I clinical trials. Davis et al 58 recently showed a dose-dependent increase of siRNA delivered via nanoparticles and observed a reduction in the specific mRNA target. However, we must remain cautious, because early clinical trials in the setting of vein graft failure suggested that antisense oligonucleotides directed against E2F (edifoligide) were promising for the treatment of vein graft failure and atherosclerosis, but the subsequent phase 3 Project of Ex Vivo Vein Graft Engineering via Transfection IV (PREVENT IV) study yielded largely disappointing results. 59 However, these studies do demonstrate that the surgical setting of coronary artery bypass grafting provides the ideal clinical setting to evaluate the clinical efficacy of these targets by gene therapy, given that the vein can be transduced ex vivo at the time of surgery. Administration of siRNA directly to the vessel would obviate the need to administer siRNA systemically and thus reduce the possibility of off-target effects. Unfortunately, there are no clear homologues of SMILR in the mouse. It would, however, be important to determine whether other preclinical models of human vascular disease contain SMILR homologues, once this information becomes available.
Taken together, these observations broaden our awareness of the complex interplay between lncRNA and protein-coding genes. The emergence of lncRNAs as regulators of gene expression will undoubtedly alter our understanding of the complex regulation network of pathological VSMC proliferation in vascular disease and may provide a means to specifically target VSMC or identify patients at risk of major adverse vascular outcomes.