Long noncoding RNAs (lncRNAs) have been highlighted to be involved in the pathological process of ischemic stroke (IS). The purpose of the present study was to investigate the expression profile of lncRNAs in peripheral blood mononuclear cells (PBMCs) of acute IS patients and to explore their utility as biomarkers of IS. Distinctive expression patterns of PBMC lncRNAs were identified by an lncRNA microarray and individual quantitative real-time PCR (qRT-PCR) in four independent sets for 206 IS, 179 healthy controls (HCs), and 55 patients with transient ischemic attack (TIA). A biomarker panel (lncRNA-based combination index) was established using logistic regression. LncRNA microarray analysis showed 70 up-regulated and 128 down-regulated lncRNAs in IS patients. Individual qRT-PCR validation demonstrated that three lncRNAs (linc-DHFRL1-4, SNHG15, and linc-FAM98A-3) were significantly up-regulated in IS patients compared with HCs and TIA patients. Longitudinal analysis of lncRNA expression up to 90 days after IS showed that linc-FAM98A-3 normalized to control levels by day 7, while SNHG15 remained increased, indicating the ability of lncRNAs to monitor IS dynamics. Receiver-operating characteristic (ROC) curve analysis showed that the lncRNA-based combination index outperformed serum brain-derived neurotrophic factor (BDNF) and neurone-specific enolase (NSE) in distinguishing IS patients from TIA patients and HCs with areas under ROC curve of more than 0.84. Furthermore, the combination index increased significantly after treatment and was correlated with neurological deficit severity of IS. The panel of these altered lncRNAs was associated with acute IS and could serve as a novel diagnostic method.

Introduction

Stroke remains a leading cause of serious long-term disability worldwide, with strokes in more than 80% of patients attributed to brain ischemia [1]. Intracranial thrombosis and extracranial embolism are the major causes of ischemic stroke (IS) [2]. The symptoms of transient ischemic attack (TIA), a harbinger of imminent stroke, are similar to those of IS and last for less than 24 h [3]. IS is currently diagnosed primarily based on neuroimaging, which has poor sensitivity for discriminating early IS and TIA, and conventional neuroimaging techniques are not applicable for monitoring dynamic changes in acute severe patients [4]. Existing blood markers of neurological deficits lack sufficient predictive ability due to limited specificity and delayed release into peripheral blood [5]. Therefore, more efforts are urgently needed to identify a novel approach for clinical evaluation and precise prediction of ischemic cerebrovascular diseases.

Previous studies have indicated that epigenetic biomarkers may facilitate clinical assessment and accurate prediction of various diseases [6]. Many long noncoding RNAs (lncRNAs) have been identified by genome-wide analyses [7], which are involved in a variety of physiological and pathological processes through transcriptional and post-transcriptional regulatory mechanisms [8]. Emerging studies have shown that lncRNAs play critical roles in the development of various pathological conditions, including cardiovascular diseases [9] and malignancies [10]. Furthermore, some lncRNAs are also involved in the development of IS [11,12]. However, system characteristics of lncRNAs from peripheral blood mononuclear cells (PBMCs) in IS patients and scientific assessments of their predictive ability have been rarely conducted. Furthermore, the expression patterns of PBMC lncRNAs in TIA patients remain unknown. Additionally, the correlations of PBMC lncRNAs levels with the neurological deficit severity of IS must still be explored.

Existing evidence has shown that PBMCs, an important component of the immune system, display the dysregulation of proinflammatory and proapoptotic genes after IS [13]. A systemic inflammatory response is triggered in response to IS [14]. Previous studies have reported PBMCs as appropriate sources of potential biomarkers for the diagnosis and prediction of stroke [15,16]. Furthermore, Bam and colleagues [17] reported dysregulation of PBMC miRNAs in IS patients by an miRNA microarray and the involvement of these miRNAs in alterations of the immune system in response to IS. However, the complete profile of PBMC lncRNAs in IS patients has not been investigated.

In the present study, we: (i) initially identified differentially expressed lncRNAs in PBMCs after acute IS using an lncRNA microarray, (ii) screened and validated the findings in three independent sets, (iii) elucidated their temporal profiles on the first, second, third, and seventh day of hospitalization and at 90 days after IS, and (iv) investigated their clinical significance as a biomarker for IS and TIA.

Materials and methods

Study population and study design

A total of 206 IS patients in the acute phase, 55 TIA patients and 179 healthy controls (HCs) were enrolled from May 2016 to November 2017. All samples taken from IS patients were obtained within 48 h of symptom onset. The samples in the discovery set, training set, and test set were obtained from the affiliated ZhongDa Hospital of Southeast University and the affiliated Nanjing Hospital of Nanjing Medical University; the samples of the acute IS patients and the HCs in the external set were collected from Dezhou People’s Hospital, and the samples of the TIA patients were collected from three hospitals. A total of 179 healthy, age- and sex-matched individuals who visited the three hospitals for a routine physical examination were obtained as HCs. Furthermore, HCs were matched with IS and TIA patients for medical history. Patients who had severe inflammatory, infective diseases, malignancies, and alcohol consumption of more than 40 g/day [18] were excluded. IS was diagnosed by an acute focal neurological deficit lasting for more than 24 h and a diffusion-weighted imaging-position lesion on MRI as well as a new lesion on a brain computed tomography (CT) scan [19]. The diagnosis of TIA was based on acute onset transient symptoms lasting for less than 24 h without evidence of infarction on CT or MRI [20]. We excluded IS patients and HCs with a history of hemorrhagic infarction, chronic kidney/liver diseases, peripheral arterial occlusive disease, active malignant disease, and inflammatory or infectious diseases. For the discovery set, we further excluded participants with a history of myocardial infarction, TIA, stroke, or silent infarction on neuroimaging. Written informed consent was obtained from each participant. This experimental protocol was approved by the Ethics Committee of Southeast University and was performed in accordance with the Declaration of Helsinki and institutional guidelines.

A multiset, case–control, hospital-based study was performed to screen and validate a panel of PBMC lncRNA biomarkers for IS. A flow chart of the study design is shown in Figure 1A. The study included four sets: a discovery set, a training set, a test set, and an external set. Briefly, a genome-wide lncRNA profiling analysis was performed amongst five acute IS patients and five HCs in the discovery set. Then, strictly candidate lncRNAs were screened in the training set and in 26 paired IS patients at pre-treatment and post-treatment (Figure 1B). For the training set, we aimed to screen the candidate lncRNAs in PBMCs from 32 IS patients and 32 HCs and established an lncRNA-based combination index by using a logistic regression model. Furthermore, we also investigated the ability of the combination index to monitor IS dynamics at pre- and post-treatment. For the test set, we further demonstrated that the expression of candidate lncRNAs was increased in IS patients, and we analyzed the diagnostic power of the lncRNA-based combination index in another independent set of 50 IS patients and 50 HCs. For the external set, we explored the discriminatory ability of the tested lncRNA-based combination index amongst 119 IS patients, 55 TIA patients, and 92 HCs. HCs were matched with IS patients for demographics and medical history. The clinical characteristics of four sets are shown in Table 1.

Flow chart of the study design and candidate lncRNAs

Figure 1
Flow chart of the study design and candidate lncRNAs

Flow chart of the study design (A) and the screening process for candidate lncRNAs (B).

Figure 1
Flow chart of the study design and candidate lncRNAs

Flow chart of the study design (A) and the screening process for candidate lncRNAs (B).

Table 1
Characteristics of patients enrolled in the present study
CharacteristicsDiscovery setTraining setTest setExternal set
ISHCsPISHCsPISHCsPISTIAHCsP
n=5n=5n=32n=32n=50n=50n=119n=55n=92
Demographics 
  Male sex 3 (60.0) 1.000§ 24 (75.0) 20 (62.5) 0.281 29 (58.0) 26 (52.0) 0.547 64 (53.8) 35 (63.6) 56 (60.9) 0.388 
  Age* 68.6 ± 5.32 68.0 ± 6.67 0.879 66.9 ± 10.32 65.1 ± 8.62 0.544 69.8 ± 11.80 66.5 ± 9.68 0.064 67.5 ± 10.00 65.6 ± 10.77 66.1 ± 10.83 0.415 
Medical history 
  Hypertension 4 (80.0) 3 (60.0) 1.000§ 15 (46.9) 12 (37.5) 0.448 23 (46.0) 18 (36.0) 0.309 52 (43.7) 18 (32.7) 38 (41.3) 0.386 
  Hyperlipidemia 1 (20.0) 1 (20.0) 1.000§ 8 (25.0) 5 (15.6) 0.351 10 (20.0) 9 (18.0) 0.799 20 (16.8) 8 (14.5) 16 (17.4) 0.899 
Diabetes mellitus 2 (40.0) 2 (40.0) 1.000§ 4 (12.5) 2 (6.3) 0.668 7 (14.0) 4 (8.0) 0.338 23 (19.3) 9 (16.4) 28 (30.4) 0.075 
History of AF 1 (20.0) 0 (0) 1.000§ 2 (6.3) 0 (0) 0.492 2 (4.0) 1 (2.0) 0.558 8 (6.7) 3 (5.5) 3 (3.3) 0.535 
Physical examination 
  NIHSS 7 (4–9) NA NA 7 (3.5–9) NA NA 8 (4–10) NA NA 8 (4–11.5) NA NA NA 
TOAST classification 
  LA 2 (40.0) NA NA 8 (25.0) NA NA 13 (26.0) NA NA 36 (30.3) NA NA NA 
  CE 1 (20.0) NA NA 12 (37.5) NA NA 22 (44.0) NA NA 46 (38.7) NA NA NA 
  SA 2 (40.0) NA NA 7 (21.9) NA NA 8 (16.0) NA NA 22 (18.5) NA NA NA 
  Others 0 (0) NA NA 5 (15.6) NA NA 7 (14.0) NA NA 15 (12.6) NA NA NA 
Thrombolysis 1 (20.0) NA NA 3 (9.4) NA NA 6 (12.0) NA NA 14 (11.8) NA NA NA 
Laboratory data on admission 
Glucose (mmol/l) 4.96 (4.51–5.78) 4.96 (4.51–5.78) 0.546 4.96 (4.51–5.78) 5.32 (4.76–5.72) 0.222 5.26 (4.68–5.75) 5.18 (4.60–5.62) 0.577 5.39 (4.91–5.85) 5.44 (4.86–5.97) 5.42 (4.91–5.73) 0.754 
TG (mmol/l) 1.19 (1.02–1.58) 1.24 (0.88–1.49) 0.844 1.10 (0.86–1.63) 1.07 (0.82–1.47) 0.757 1.21 (0.88–1.78) 1.12 (0.93–1.58) 0.836 1.28 (0.96–1.61) 1.19 (1.02–1.61) 1.29 (1.07–1.68) 0.721 
TC (mmol/l) 4.88 (3.74–5.20) 4.62 (4.11–5.09) 0.528 4.61 (4.05–5.32) 4.40 (3.87–5.22) 0.405 4.66 (4.24–5.10) 4.80 (4.23–5.32) 0.432 4.91 (4.14–5.53) 4.86 (4.20–5.40) 4.98 (4.33–5.63) 0.264 
LDL-C (mmol/l) 2.54 (2.22–2.96) 2.74 (2.15–3.08) 0.095 2.79 (2.16–3.20) 2.70 (2.34–3.42) 0.851 2.88 (2.49–3.41) 2.83 (2.31–3.21) 0.662 2.97 (2.40–3.47) 2.83 (2.36–3.23) 3.00 (2.61–3.35) 0.398 
HDL-C (mmol/l) 1.09 (0.98–1.34) 1.15 (0.90–1.32) 0.108 1.12 (1.00–1.26) 1.11 (1.00–1.25) 0.835 1.21 (1.03–1.43) 1.22 (1.08–1.43) 0.482 1.24 (1.12–1.50) 1.18 (1.12–1.33) 1.29 (1.14–1.52) 0.101 
WBC count (109/l) 6.89 (4.68–7.23) 6.24 (5.05–7.11) 0.236 6.31 (5.29–7.98) 6.00 (4.99–7.72) 0.493 6.79 (5.87–8.40) 6.73 (5.58–7.68) 0.521 7.72 (6.04–9.73) 7.09 (5.62–8.00) 7.17 (5.72–9.35) 0.108 
CharacteristicsDiscovery setTraining setTest setExternal set
ISHCsPISHCsPISHCsPISTIAHCsP
n=5n=5n=32n=32n=50n=50n=119n=55n=92
Demographics 
  Male sex 3 (60.0) 1.000§ 24 (75.0) 20 (62.5) 0.281 29 (58.0) 26 (52.0) 0.547 64 (53.8) 35 (63.6) 56 (60.9) 0.388 
  Age* 68.6 ± 5.32 68.0 ± 6.67 0.879 66.9 ± 10.32 65.1 ± 8.62 0.544 69.8 ± 11.80 66.5 ± 9.68 0.064 67.5 ± 10.00 65.6 ± 10.77 66.1 ± 10.83 0.415 
Medical history 
  Hypertension 4 (80.0) 3 (60.0) 1.000§ 15 (46.9) 12 (37.5) 0.448 23 (46.0) 18 (36.0) 0.309 52 (43.7) 18 (32.7) 38 (41.3) 0.386 
  Hyperlipidemia 1 (20.0) 1 (20.0) 1.000§ 8 (25.0) 5 (15.6) 0.351 10 (20.0) 9 (18.0) 0.799 20 (16.8) 8 (14.5) 16 (17.4) 0.899 
Diabetes mellitus 2 (40.0) 2 (40.0) 1.000§ 4 (12.5) 2 (6.3) 0.668 7 (14.0) 4 (8.0) 0.338 23 (19.3) 9 (16.4) 28 (30.4) 0.075 
History of AF 1 (20.0) 0 (0) 1.000§ 2 (6.3) 0 (0) 0.492 2 (4.0) 1 (2.0) 0.558 8 (6.7) 3 (5.5) 3 (3.3) 0.535 
Physical examination 
  NIHSS 7 (4–9) NA NA 7 (3.5–9) NA NA 8 (4–10) NA NA 8 (4–11.5) NA NA NA 
TOAST classification 
  LA 2 (40.0) NA NA 8 (25.0) NA NA 13 (26.0) NA NA 36 (30.3) NA NA NA 
  CE 1 (20.0) NA NA 12 (37.5) NA NA 22 (44.0) NA NA 46 (38.7) NA NA NA 
  SA 2 (40.0) NA NA 7 (21.9) NA NA 8 (16.0) NA NA 22 (18.5) NA NA NA 
  Others 0 (0) NA NA 5 (15.6) NA NA 7 (14.0) NA NA 15 (12.6) NA NA NA 
Thrombolysis 1 (20.0) NA NA 3 (9.4) NA NA 6 (12.0) NA NA 14 (11.8) NA NA NA 
Laboratory data on admission 
Glucose (mmol/l) 4.96 (4.51–5.78) 4.96 (4.51–5.78) 0.546 4.96 (4.51–5.78) 5.32 (4.76–5.72) 0.222 5.26 (4.68–5.75) 5.18 (4.60–5.62) 0.577 5.39 (4.91–5.85) 5.44 (4.86–5.97) 5.42 (4.91–5.73) 0.754 
TG (mmol/l) 1.19 (1.02–1.58) 1.24 (0.88–1.49) 0.844 1.10 (0.86–1.63) 1.07 (0.82–1.47) 0.757 1.21 (0.88–1.78) 1.12 (0.93–1.58) 0.836 1.28 (0.96–1.61) 1.19 (1.02–1.61) 1.29 (1.07–1.68) 0.721 
TC (mmol/l) 4.88 (3.74–5.20) 4.62 (4.11–5.09) 0.528 4.61 (4.05–5.32) 4.40 (3.87–5.22) 0.405 4.66 (4.24–5.10) 4.80 (4.23–5.32) 0.432 4.91 (4.14–5.53) 4.86 (4.20–5.40) 4.98 (4.33–5.63) 0.264 
LDL-C (mmol/l) 2.54 (2.22–2.96) 2.74 (2.15–3.08) 0.095 2.79 (2.16–3.20) 2.70 (2.34–3.42) 0.851 2.88 (2.49–3.41) 2.83 (2.31–3.21) 0.662 2.97 (2.40–3.47) 2.83 (2.36–3.23) 3.00 (2.61–3.35) 0.398 
HDL-C (mmol/l) 1.09 (0.98–1.34) 1.15 (0.90–1.32) 0.108 1.12 (1.00–1.26) 1.11 (1.00–1.25) 0.835 1.21 (1.03–1.43) 1.22 (1.08–1.43) 0.482 1.24 (1.12–1.50) 1.18 (1.12–1.33) 1.29 (1.14–1.52) 0.101 
WBC count (109/l) 6.89 (4.68–7.23) 6.24 (5.05–7.11) 0.236 6.31 (5.29–7.98) 6.00 (4.99–7.72) 0.493 6.79 (5.87–8.40) 6.73 (5.58–7.68) 0.521 7.72 (6.04–9.73) 7.09 (5.62–8.00) 7.17 (5.72–9.35) 0.108 

Abbreviations: AF, atrial fibrillation; CE, cardioembolism; HDL-C, high-density lipoprotein cholesterol; LA, large artery atherosclerotic stroke; LDL-C, high-density lipoprotein cholesterol; NA, not available; NIHSS, National Institutes of Health Stroke Scale; SA, small artery stroke; TOAST, Trial of Org 10172 in Acute Stroke Treatment; WBC, white blood cell.

*

Data were expressed as mean ± S.D.

Data were expressed as median (quartile).

Undetermined/unclassified.

§

Fisher’s exact test.

Clinical assessment

Stroke etiology was determined in accordance with the Trial of Org 10172 in Acute Stroke Treatment (TOAST) [21]. Stroke severity was assessed by the National Institutes of Health Stroke Scale (NIHSS) scores. The enrolled IS patients were divided into three groups according to NIHSS scores: mild stroke (NIHSS ≤ 5, n=83), moderate stroke (6 ≤ NIHSS ≤ 13, n=73), and severe stroke (NIHSS ≥ 14, n=45) [22].

Blood sampling and processing

Blood samples were obtained from IS and TIA patients upon admission. IS patients were further longitudinally sampled on the second, third, and seventh day of hospitalization and at 90 days after stroke. Samples from HCs were collected in the outpatient clinic and physical examination center. Whole blood (4–5 ml) was collected from each participant in a K2-EDTA plasma tube. PBMCs were isolated within 2 h of blood draw by using Ficoll (TBD science, Tianjin, China) gradient centrifugation. PBMCs were then transferred into 1-ml TRIzol Reagent in 1.5-ml centrifuge tubes (Invitrogen, Carlsbad, CA, U.S.A.) and stored at −80°C until RNA extraction.

lncRNA microarray profiling

For lncRNA microarray, total RNA was extracted from PBMCs by using TRIzol Reagent and purified with an mirVana miRNA Isolation Kit (Ambion, Austin, TX, U.S.A.) following the manufacturer’s instructions. The lncRNA microarray was analyzed using a Human LncRNA Array v4 (CapitalBio Technology, Beijing) with four identical arrays per slide (4 × 180K format) to evaluate the lncRNA profiles of 41000 human lncRNAs from multiple databases. Each RNA was measured by probes and repeated twice. A total of 4974 Agilent control probes were included in the array. The microarray image was transferred into spot intensity values by Feature Extraction software V10.7 (Agilent Technology, CA, U.S.A.). The lncRNA microarray data were exported by using the GeneSpring software V13.0 (Agilent Technology, CA, U.S.A.). To screen the differential lnRNAs, we used a fold change ≥ 2 and a Benjamini–Hochberg corrected P<0.05. Then, we used hierarchical clustering analysis on differential lncRNAs between IS patients and HCs.

RNA extraction and quantitative real-time PCR validation

Total RNA was extracted from PBMCs by using TRIzol Reagent (Invitrogen, CA, U.S.A.) following the manufacturer’s instructions, as previously described [23]. Briefly, after vortexing 1 ml TRIzol Reagent and PBMCs for 30 s and then resting at room temperature for 5 min, we added 200 μl chloroform. The mixture was vortexed for another 15 s and then centrifuged at 12000×g for 15 min at 4°C. Finally, total RNA was extracted according to the manufacturer’s protocol.

After reverse transcription using a PrimeScript RT reagent Kit with gDNA Eraser (Takara, Dalian, China), quantitative real-time PCR (qRT-PCR) was performed using an ABI 7500 System (Applied Biosystems, Carlsbad, CA, U.S.A.) and SYBR Premix Ex Taq™ II (Takara, Dalian, China). All experiments were repeated in triplicate. The Ct value of each measured lncRNA was normalized to the housekeeping gene (GAPDH). ΔCt was obtained by subtracting the Ct values of GAPDH from those of the candidate lncRNAs. ΔΔCt was then obtained by subtracting the ΔCt of the HCs from that of the IS patients. The relative expression of candidate lncRNAs was calculated by 2−ΔΔCt. The primers of candidate lncRNAs are listed in Supplementary Table S1.

RNA stability experiments

Several lncRNAs are stable in some extreme conditions. PBMC samples were obtained from IS patients and were incubated under extreme conditions, such as incubation at 4, −20, and −80°C for 2 h, digestion with RNase A for 2 and 4 h, repeated freeze-thaw cycles and alkaline or acidic condition. Subsequently, the raw Ct values of PBMC linc-DHFRL1-4, SNHG15 and linc-FAM98A-3 levels were determined by qRT-PCR.

Serum brain-derived neurotrophic factor and neurone-specific enolase measurements

Serum samples were obtained from 52 IS patients, 15 TIA patients on admission and 60 HCs at least 12 h after fasting in the test and external sets. Brain-derived neurotrophic factor (BDNF) and neurone-specific enolase (NSE) were measured by commercial ELISA kits according to the manufacturer’s protocols (Cusabio Biotech, Wuhan, China).

Statistical analysis

Continuous variables are presented as the mean (S.D.) or median (interquartile range, IQR). Comparisons of continuous variables between groups (n=2 groups) were performed by Student’s t test or the Mann–Whitney U-test. Comparisons of continuous variables amongst groups (n>2 groups) were performed by one-way ANOVA or the Mann–Whitney U-test as appropriate. Comparisons of categorical variables were performed by the Chi-squared test or Fisher’s exact test. Diagnostic performance of candidate biomarkers was analyzed by receiver-operating characteristic (ROC) curves and the area under the ROC curve (AUC). Nonparametric statistics were applied to compare the AUCs of candidate biomarkers [24]. Youden’s index was used to determine the optimal cut-off values of candidate lncRNAs. The regression coefficient of each candidate lncRNA was determined by using a logistic regression model, and a combination index was then constructed by employing their regression coefficients as weight values. The predictive ability of candidate lncRNAs was further assessed by logistic regression analysis. All statistical analyses were performed with GraphPad Prism 5.0 (CA, U.S.A.) and SPSS 20.0 (IBM Corporation, NY, U.S.A.). A P-value less than 0.05 was considered statistically significant.

Results

Characteristics of the study participants

Five acute IS patients and five HCs included in the discovery set were matched with respect to all clinical or demographic characteristics. In addition to the discovery set, an independent training set of 32 IS patients and 32 control subjects was investigated to verify the results of the discovery set. To further confirm the expression of screened lncRNA and the potential correlation between the expression of screened lncRNA and the risk of IS, we performed a case–control study consisting of 50 IS patients and 50 control subjects in the test set. Finally, the diagnostic value of the screened lncRNAs in the IS patients was explored in another independent external set of 119 IS patients, 55 TIA patients, and 92 control subjects. The characteristics of all sets are summarized in Table 1.

Discovery set: microarray profiling of lncRNAs associated with IS

To screen for differentially expressed lncRNAs associated with IS in the acute phase, we initially analyzed lncRNA expression in PBMCs from five IS patients and five controls using the ‘Human LncRNA Array v4’. The results showed 198 differentially expressed lncRNAs (fold-change > 2 and P<0.05) between the two groups according to the microarray data (Figure 2). Of these lncRNAs, 70 lncRNAs were up-regulated and 128 lncRNAs were down-regulated (Supplementary Table S2). To confirm the stability of the microarray data, we randomly selected five lncRNAs (linc-DHFRL1-4, ENST0000057109.1, SNHG15, ENST00000536112.1, and ENSG00000251002.3) from amongst the differentially expressed lncRNAs to detect their expression levels using qRT-PCR. qRT-PCR of these five lncRNAs showed that their expression signatures in the discovery set were consistent with the microarray profiling (Supplementary Figure S1). To identify target biomarkers, we selected the up-regulated lncRNAs in PBMCs from the IS patients as candidate lncRNAs. Of the 70 up-regulated lncRNAs, we chose the potential candidate biomarkers according to the following criteria: statistical significance of <0.005 and lncRNAs with no annotations excluded. After the selection procedure, eight candidate lncRNAs were further investigated in the following experiments (Figure 1B).

Distinctive expression profile of lncRNAs in PBMCs of IS patients

Figure 2
Distinctive expression profile of lncRNAs in PBMCs of IS patients

The heat map (A), scatter plot (B), and volcano plot (C) shows 198 differentially expressed lncRNAs between IS patients (n=5) and HCs (n=5) by the lncRNA microarray analysis.

Figure 2
Distinctive expression profile of lncRNAs in PBMCs of IS patients

The heat map (A), scatter plot (B), and volcano plot (C) shows 198 differentially expressed lncRNAs between IS patients (n=5) and HCs (n=5) by the lncRNA microarray analysis.

Biomarker screening and validation

To confirm the up-regulated expression of the eight candidate lncRNAs, their expression levels were measured by qRT-PCR in PBMC samples of 32 IS and 32 controls. As shown in Figure 3A, five of these lncRNAs were differentially expressed after IS (all P-values <0.05). We further confirmed the expression of the five lncRNAs by qRT-PCR in 26 pre- and post-treatment PBMC samples from IS patients. As shown in Figure 3B, the three lncRNAs (linc-DHFRL1-4, SNHG15, and linc-FAM98A-3) were significantly down-regulated after systematic treatment.

Biomarker screening and qRT-PCR validation

Figure 3
Biomarker screening and qRT-PCR validation

(A) ENST00000450016.1, ENST00000443162.1, linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 were differentially expressed in the training set of 32 IS patients and 32 HCs. (B) Expression levels of five lncRNAs in 26 IS patients at pre- and post-treatment. *P<0.05.

Figure 3
Biomarker screening and qRT-PCR validation

(A) ENST00000450016.1, ENST00000443162.1, linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 were differentially expressed in the training set of 32 IS patients and 32 HCs. (B) Expression levels of five lncRNAs in 26 IS patients at pre- and post-treatment. *P<0.05.

Stability of PBMC lncRNAs under different conditions

Previous studies have demonstrated that plasma lncRNAs are stable in some extreme conditions [25,26]. To examine the stability of three PBMC lncRNAs in IS, PBMC samples were treated under extreme conditions, such as incubation at 4, −20, and −80°C for 2 h, digestion with RNase A for 2 and 4 h, repeated freeze-thaw cycles and alkaline or acidic condition. Although the PBMCs were destroyed under these conditions, the raw Ct values of PBMC linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 levels were not significantly changed (Figure 4).

Stability of PBMC lncRNAs under different conditions

Figure 4
Stability of PBMC lncRNAs under different conditions

Stability of lncRNAs in PBMCs under extreme conditions: incubation at different temperature for 2 h (A); RNase A digestion for 2 and 4 h (B); repeated freeze-thaw cycles (C); and high or low pH (D).

Figure 4
Stability of PBMC lncRNAs under different conditions

Stability of lncRNAs in PBMCs under extreme conditions: incubation at different temperature for 2 h (A); RNase A digestion for 2 and 4 h (B); repeated freeze-thaw cycles (C); and high or low pH (D).

Training set: diagnostic values of PBMC lncRNAs in acute IS

To confirm these results in the discovery set, we detected the PBMC expression levels of these potential biomarkers in a training set of 32 patients and 32 control subjects by qRT-PCR. The results showed a 3.17-fold increase in the expression of linc-DHFRL1-4, a 3.02-fold increase in SNHG15, and a 2.19-fold increase in linc-FAM98A-3 in patients with IS compared with HCs (Supplementary Figure S2). To further study the diagnostic power of the three identified lncRNAs (linc-DHFRL1-4, SNHG15, and linc-FAM98A-3) for IS patients, we performed an ROC curve analysis in the training set. We observed that the AUC was 0.711 for linc-DHFRL1-4, 0.756 for SNHG15 and 0.659 for linc-FAM98A-3 (Supplementary Figure S2). Their detailed sensitivity and specificity are summarized in Table 2. The largest Youden index was used to determine the optimal cut-off values of the three lncRNAs. We found that PBMC SNHG15 presented the best diagnostic ability, with a cut-off value of 9.39 (Table 2).

Table 2
Performance of three lncRNAs in the diagnosis of IS from HC and TIA
lncRNAAUC (95%CI)SensitivitySpecificityYouden indexCutoffTrue positiveTrue negativeFalse positiveFalse negative
Training set 
linc-DHFRL1-4 0.711 (0.584–0.838) 0.687 0.719 0.406 12.77 22 23 10 
SNHG15 0.756 (0.639–0.873) 0.594 0.844 0.438 9.39 19 27 13 
linc-FAM98A-3 0.659 (0.525–0.792) 0.594 0.688 0.282 7.37 19 22 10 13 
Combination index 0.842 (0.749–0.935) 0.813 0.719 0.532 14.67 26 23 
Test set 
linc-DHFRL1-4 0.718 (0.618–0.817) 0.740 0.640 0.380 12.77 37 32 18 13 
SNHG15 0.755 (0.656–0.854) 0.700 0.700 0.400 9.39 35 35 15 15 
linc-FAM98A-3 0.727 (0.628–0.827) 0.660 0.720 0.380 7.37 33 36 14 17 
Combination index 0.878 (0.810–0.945) 0.880 0.660 0.540 14.67 44 33 17 
Combination set 
linc-DHFRL1-4 0.715 (0.637–0.794) 0.720 0.671 0.391 12.77 59 55 27 23 
SNHG15 0.763 (0.689–0.836) 0.659 0.756 0.415 9.39 54 62 20 28 
linc-FAM98A-3 0.700 (0.621–0.780) 0.634 0.707 0.341 7.37 52 58 24 30 
Combination index 0.859 (0.803–0.915) 0.854 0.683 0.537 14.67 70 56 26 12 
External set (HC) 
linc-DHFRL1-4 0.784 (0.724–0.844) 0.739 0.630 0.369 12.77 88 58 34 31 
SNHG15 0.744 (0.672–0.815) 0.697 0.728 0.425 9.39 83 67 25 36 
linc-FAM98A-3 0.683 (0.611–0.754) 0.588 0.689 0.277 7.37 70 82 37 49 
Combination index 0.879 (0.834–0.924) 0.857 0.782 0.639 14.67 102 93 26 17 
External set (TIA) 
linc-DHFRL1-4 0.673 (0.598–0.742) 0.739 0.400 0.139 12.77 88 22 33 31 
SNHG15 0.724 (0.651–0.789) 0.697 0.673 0.370 9.39 83 37 18 36 
linc-FAM98A-3 0.675 (0.600–0.744) 0.588 0.655 0.243 7.37 70 36 19 49 
Combination index 0.847 (0.785–0.897) 0.857 0.636 0.493 14.67 102 35 20 17 
lncRNAAUC (95%CI)SensitivitySpecificityYouden indexCutoffTrue positiveTrue negativeFalse positiveFalse negative
Training set 
linc-DHFRL1-4 0.711 (0.584–0.838) 0.687 0.719 0.406 12.77 22 23 10 
SNHG15 0.756 (0.639–0.873) 0.594 0.844 0.438 9.39 19 27 13 
linc-FAM98A-3 0.659 (0.525–0.792) 0.594 0.688 0.282 7.37 19 22 10 13 
Combination index 0.842 (0.749–0.935) 0.813 0.719 0.532 14.67 26 23 
Test set 
linc-DHFRL1-4 0.718 (0.618–0.817) 0.740 0.640 0.380 12.77 37 32 18 13 
SNHG15 0.755 (0.656–0.854) 0.700 0.700 0.400 9.39 35 35 15 15 
linc-FAM98A-3 0.727 (0.628–0.827) 0.660 0.720 0.380 7.37 33 36 14 17 
Combination index 0.878 (0.810–0.945) 0.880 0.660 0.540 14.67 44 33 17 
Combination set 
linc-DHFRL1-4 0.715 (0.637–0.794) 0.720 0.671 0.391 12.77 59 55 27 23 
SNHG15 0.763 (0.689–0.836) 0.659 0.756 0.415 9.39 54 62 20 28 
linc-FAM98A-3 0.700 (0.621–0.780) 0.634 0.707 0.341 7.37 52 58 24 30 
Combination index 0.859 (0.803–0.915) 0.854 0.683 0.537 14.67 70 56 26 12 
External set (HC) 
linc-DHFRL1-4 0.784 (0.724–0.844) 0.739 0.630 0.369 12.77 88 58 34 31 
SNHG15 0.744 (0.672–0.815) 0.697 0.728 0.425 9.39 83 67 25 36 
linc-FAM98A-3 0.683 (0.611–0.754) 0.588 0.689 0.277 7.37 70 82 37 49 
Combination index 0.879 (0.834–0.924) 0.857 0.782 0.639 14.67 102 93 26 17 
External set (TIA) 
linc-DHFRL1-4 0.673 (0.598–0.742) 0.739 0.400 0.139 12.77 88 22 33 31 
SNHG15 0.724 (0.651–0.789) 0.697 0.673 0.370 9.39 83 37 18 36 
linc-FAM98A-3 0.675 (0.600–0.744) 0.588 0.655 0.243 7.37 70 36 19 49 
Combination index 0.847 (0.785–0.897) 0.857 0.636 0.493 14.67 102 35 20 17 

Abbreviation: CI, confidence interval.

To assess the diagnostic performance of the combined lncRNAs for IS patients, we first obtained the regression coefficient of each identified lncRNA (linc-DHFRL1-4, SNHG15, and linc-FAM98A-3) by using a logistic regression model, and then established a combination index based on three identified lncRNAs by employing their regression coefficients as weight values. The results showed that the combination index significantly decreased after IS (P<0.001, Figure 5A). Furthermore, the combination index outperformed the individual lncRNAs in IS identification (Table 2). Subsequently, we compared the AUC values of these lncRNAs (Figure 5B). The combination index showed a markedly higher AUC value than the other lncRNAs, indicating that it had the greatest diagnostic power with an AUC value of 0.842 (95% confidence interval (CI): 0.749–0.935). Taken together, a combination of these three lncRNAs could precisely distinguish IS patients from HCs.

Diagnostic utility of the lncRNA-based combination index for acute IS in the training set

Figure 5
Diagnostic utility of the lncRNA-based combination index for acute IS in the training set

(A) The combination index significantly decreased after IS. (B) ROC curves for the combination index and individual lncRNAs and their corresponding AUC values for discriminating IS patients from HCs. (C) The combination index significantly increased after treatment.

Figure 5
Diagnostic utility of the lncRNA-based combination index for acute IS in the training set

(A) The combination index significantly decreased after IS. (B) ROC curves for the combination index and individual lncRNAs and their corresponding AUC values for discriminating IS patients from HCs. (C) The combination index significantly increased after treatment.

We obtained 26 PBMC samples from the 32 IS patients after systematic treatment in the training set,t o assess whether the combination index changed after treatment. The results showed that the combination index significantly increased after treatment (P<0.001, Figure 5C), indicating its ability to monitor IS dynamics. Furthermore, the ΔCt values of these three lncRNAs were also increased in 26 post-treatment PBMC samples (Figure 3B).

Test set: independent validation of PBMC lnRNAs

To further independently validate the levels of PBMC lncRNAs in patients with IS, we investigated these three lncRNAs in a test set of 50 IS patients and 50 control subjects using qRT-PCR. We found that these three lncRNAs exhibited significantly increased expression (3.19-fold increase for linc-DHFRL1-4, 3.51-fold increase for SNHG15, and 2.46-fold increase for linc-FAM98A-3) in patients with IS compared with HCs, with AUC values ranging from 0.718 to 0.755 (all P<0.01, Supplementary Figure S3). Their detailed sensitivity and specificity were determined by the cut-off values in the training set (Table 2).

Similar to the results in the training set, the lncRNA-based combination index significantly decreased after IS (P<0.001, Figure 6A). The combination index showed the greatest diagnostic power with an AUC of 0.878 in the test set (Table 2). Furthermore, the diagnostic power of the combination index was superior to that of lncRNA alone (P<0.01, Figure 6B). In addition, a total of 36 PBMC samples were available after systematic treatment. The value of the combination index was significantly up-regulated after treatment (P<0.001, Figure 6C), indicating that it could monitor IS dynamics. The training and test sets were pooled to further analyze its diagnostic performance. ROC curves showed that the combination index outperformed single lncRNA in the detection of IS (P<0.01, Supplementary Figure S4), with an AUC of 0.859 (95% CI: 0.803–0.915) (Table 2).

Diagnostic performance of the lncRNA-based combination index for acute IS in the test set

Figure 6
Diagnostic performance of the lncRNA-based combination index for acute IS in the test set

(A) The combination index significantly decreased after IS. (B) ROC curves for the combination index and individual lncRNAs and their corresponding AUC values for discriminating IS patients from HCs. (C) The combination index significantly increased after treatment.

Figure 6
Diagnostic performance of the lncRNA-based combination index for acute IS in the test set

(A) The combination index significantly decreased after IS. (B) ROC curves for the combination index and individual lncRNAs and their corresponding AUC values for discriminating IS patients from HCs. (C) The combination index significantly increased after treatment.

For evaluating the association between the combination index and clinical characteristics, we analyzed IS patients from the training and test sets. We divided these patients into low and high combination index groups based on the median of the index. The data showed that a low-level combination index was significantly associated with stroke severity (P=0.017), not stroke etiology (P=0.560, Table 3).

Table 3
Association of combination index with clinical characteristics in the training and test sets
CharacteristicsCombination index
Low valuesHigh valuesP
n=41 (%)n=41 (%)
Demographics 
  Male sex 26 (63.4) 27 (65.9) 0.817 
  Age* 68.85 ± 9.34 68.27 ± 11.83 0.804 
Medical history 
  Hypertension 18 (43.9) 20 (48.8) 0.658 
  Hyperlipidemia 12 (29.3) 6 (14.6) 0.109 
  Diabetes mellitus 6 (14.6) 5 (12.2) 0.746 
  History of AF 2 (4.9) 2 (4.9) 1.000 
Physical examination 
  NIHSS 9 (5–13) 7 (3.5–10) 0.017 
TOAST classification 
  LA 8 (19.5) 13 (31.7) 0.560 
  CE 19 (46.3) 15 (36.6)  
  SA 7 (17.1) 8 (19.5)  
  Others 7 (17.1) 5 (12.2)  
Thrombolysis 4 (9.8) 5 (12.2) 0.724 
Laboratory data on admission 
  Glucose (mmol/l) 5.28 (4.71–5.75) 4.96 (4.56–5.73) 0.277 
  TG (mmol/l) 1.22 (1.01–1.92) 1.05 (0.81–1.37) 0.105 
  TC (mmol/l) 4.43 (4.00–5.06) 4.82 (4.38–5.37) 0.060 
  LDL-C (mmol/l) 2.87 (2.31–3.38) 2.74 (2.31–3.32) 0.772 
  HDL-C (mmol/l) 1.18 (1.02–1.34) 1.15 (1.02–1.33) 0.990 
  WBC count (109/l) 6.74 (5.90–8.00) 6.47 (5.42–8.23) 0.921 
CharacteristicsCombination index
Low valuesHigh valuesP
n=41 (%)n=41 (%)
Demographics 
  Male sex 26 (63.4) 27 (65.9) 0.817 
  Age* 68.85 ± 9.34 68.27 ± 11.83 0.804 
Medical history 
  Hypertension 18 (43.9) 20 (48.8) 0.658 
  Hyperlipidemia 12 (29.3) 6 (14.6) 0.109 
  Diabetes mellitus 6 (14.6) 5 (12.2) 0.746 
  History of AF 2 (4.9) 2 (4.9) 1.000 
Physical examination 
  NIHSS 9 (5–13) 7 (3.5–10) 0.017 
TOAST classification 
  LA 8 (19.5) 13 (31.7) 0.560 
  CE 19 (46.3) 15 (36.6)  
  SA 7 (17.1) 8 (19.5)  
  Others 7 (17.1) 5 (12.2)  
Thrombolysis 4 (9.8) 5 (12.2) 0.724 
Laboratory data on admission 
  Glucose (mmol/l) 5.28 (4.71–5.75) 4.96 (4.56–5.73) 0.277 
  TG (mmol/l) 1.22 (1.01–1.92) 1.05 (0.81–1.37) 0.105 
  TC (mmol/l) 4.43 (4.00–5.06) 4.82 (4.38–5.37) 0.060 
  LDL-C (mmol/l) 2.87 (2.31–3.38) 2.74 (2.31–3.32) 0.772 
  HDL-C (mmol/l) 1.18 (1.02–1.34) 1.15 (1.02–1.33) 0.990 
  WBC count (109/l) 6.74 (5.90–8.00) 6.47 (5.42–8.23) 0.921 

Abbreviations: AF, atrial fibrillation; CE, cardioembolism; HDL-C, high-density lipoprotein cholesterol; LA, large artery atherosclerotic stroke; LDL-C, high-density lipoprotein cholesterol; SA, small artery stroke; WBC, white blood cell.

*

Data were expressed as mean ± S.D.

Data were expressed as median (quartile).

Undetermined/unclassified.

External set: external validation of PBMC lncRNAs

To further verify these findings, we examined circulating levels of linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 in an external set of 119 IS patients and 92 HCs. All three lncRNAs showed obviously increased expression levels in IS patients compared with s HCs (all P<0.001, Figure 7). IS patients had a lower combination index than HCs (P<0.001, Figure 8A). Furthermore, lncRNA-based combination index also outperformed the individual lncRNAs in distinguishing IS patients from HCs (Figure 8B), and its sensitivity and specificity were 0.857 and 0.782, respectively (Table 2). Collectively, these data from the external set further confirmed the results from the training and test sets.

Expression levels of lncRNAs in the external set

Figure 7
Expression levels of lncRNAs in the external set

Expression levels of linc-DHFRL1-4 (A), SNHG15 (B), and linc-FAM98A-3 (C) in the external set. *P<0.05, ***P<0.001.

Figure 7
Expression levels of lncRNAs in the external set

Expression levels of linc-DHFRL1-4 (A), SNHG15 (B), and linc-FAM98A-3 (C) in the external set. *P<0.05, ***P<0.001.

Diagnostic power of the lncRNA-based combination index for IS and TIA in the external set

Figure 8
Diagnostic power of the lncRNA-based combination index for IS and TIA in the external set

(A) Values of the combination index in IS patients, TIA patients, and HCs. ROC curves for the combination index and individual lncRNAs and their corresponding AUC values for discriminating IS patients from HCs (B) and TIA patients (C). ***P<0.001.

Figure 8
Diagnostic power of the lncRNA-based combination index for IS and TIA in the external set

(A) Values of the combination index in IS patients, TIA patients, and HCs. ROC curves for the combination index and individual lncRNAs and their corresponding AUC values for discriminating IS patients from HCs (B) and TIA patients (C). ***P<0.001.

A diagnostic biomarker should not only distinguish patients from HCs, but also identify patients from other similar symptoms. To study the diagnostic ability of the three identified lncRNAs for IS patients compared with other diseases, we enrolled 55 patients with TIA and evaluated their discriminatory ability. We observed that the three lncRNAs were significantly up-regulated in patients with IS compared with patients with TIA (P<0.001), and linc-DHFRL1-4 was increased in patients with TIA compared with HCs (P<0.05, Figure 7A). The combination index was also lower in IS patients than in TIA patients (P<0.001, Figure 8A). The ROC curve analysis showed that the combination index could distinguish IS from TIA with an AUC value of 0.847 (Figure 8C), and the corresponding sensitivity and specificity were 0.857 and 0.636, respectively (Table 2).

Temporal profile of PBMC lncRNAs levels after stroke

To further investigate the clinical utility of PBMC lncRNAs after stroke, we measured the expression levels of linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 on the first, second, third, and seventh day of hospitalization and at 90 days after stroke. A total of 16 PBMC samples were obtained from IS patients at five time points. We observed significant changes over time in the expression levels of three lncRNAs in the IS patients compared with the 16 HCs (P<0.001, Figure 9). Their expression levels were decreased from days 1 to 90 after IS (P<0.001). The SNHG15 expression in IS patients was significantly higher at all time points than the expression in HCs. Furthermore, the expression level of linc-FAM98A-3 remained high after day 2 with no significant difference compared with day 1. Additionally, the expression level of linc-FAM98A-3 in IS patients showed no significant difference compared with that in HCs from day 7.

Temporal expression profile of lncRNAs in PBMCs after IS

Figure 9
Temporal expression profile of lncRNAs in PBMCs after IS

The results for the 16 IS patients with available data for five time points after IS: linc-DHFRL1-4 (A), SNHG15 (B), and linc-FAM98A-3 (C). *P<0.05 compared with HC; #P<0.05 compared with day 1.

Figure 9
Temporal expression profile of lncRNAs in PBMCs after IS

The results for the 16 IS patients with available data for five time points after IS: linc-DHFRL1-4 (A), SNHG15 (B), and linc-FAM98A-3 (C). *P<0.05 compared with HC; #P<0.05 compared with day 1.

Predictive ability of PBMC lncRNAs in the training, test, and external sets

To investigate the predictive performance of the three identified lncRNAs for IS and TIA patients, we combined the three sets of patients (Supplementary Table S3) and performed a logistic regression analysis. We first employed the HCs as a reference group. The results showed that increased linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 expression levels and a lower combination index were independently associated with IS presence, and linc-DHFRL1-4 was independently associated with TIA presence. After adjusting for age, sex, medical history, and laboratory data on admission, the results remained stable. Subsequently, TIA patients were considered as the reference group. Increased linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 and a lower combination index were independently correlated with IS presence in the univariate and multivariate analyses. The odds ratio (ORs) of lncRNAs for IS or TIA presence are detailed in Table 4.

Table 4
Logistic regression analysis for predictive ability of three identified lncRNAs
VariableUnivariate analysisMultivariate analysis
OR (95% CI)POR (95% CI)P
IS* 
linc-DHFRL1-4 1.668 (1.469–1.895) <10−4 1.801 (1.564–2.075) <10−4 
SNHG15 1.813 (1.572–2.092) <10−4 1.912 (1.634–2.237) <10−4 
linc-FAM98A-3 1.527 (1.337–1.744) <10−4 1.578 (1.371–.816) <10−4 
Combination index 2.983 (2.389–3.726) <10−4 3.548 (2.729–4.613) <10−4 
TIA* 
linc-DHFRL1-4 1.302 (1.094–1.550) 0.003 1.241 (1.040–1.538) 0.009 
SNHG15 1.059 (0.918–1.221) 0.433 NA NA 
linc-FAM98A-3 1.110 (0.929–1.325) 0.250 NA NA 
Combination index 1.138 (0.951–1.362) 0.158 NA NA 
IS 
linc-DHFRL1-4 1.311 (1.118–1.536) 0.001 1.404 (1.231–1.700) 0.0003 
SNHG15 1.740 (1.437–2.103) <10−4 1.629 (1.311–1.986) <10−4 
linc-FAM98A-3 1.417 (1.173–1.711) 0.0003 1.522 (1.207–1.842) 0.0003 
Combination index 2.470 (1.900–3.210) <10−4 3.156 (2.328–4.007) <10−4 
VariableUnivariate analysisMultivariate analysis
OR (95% CI)POR (95% CI)P
IS* 
linc-DHFRL1-4 1.668 (1.469–1.895) <10−4 1.801 (1.564–2.075) <10−4 
SNHG15 1.813 (1.572–2.092) <10−4 1.912 (1.634–2.237) <10−4 
linc-FAM98A-3 1.527 (1.337–1.744) <10−4 1.578 (1.371–.816) <10−4 
Combination index 2.983 (2.389–3.726) <10−4 3.548 (2.729–4.613) <10−4 
TIA* 
linc-DHFRL1-4 1.302 (1.094–1.550) 0.003 1.241 (1.040–1.538) 0.009 
SNHG15 1.059 (0.918–1.221) 0.433 NA NA 
linc-FAM98A-3 1.110 (0.929–1.325) 0.250 NA NA 
Combination index 1.138 (0.951–1.362) 0.158 NA NA 
IS 
linc-DHFRL1-4 1.311 (1.118–1.536) 0.001 1.404 (1.231–1.700) 0.0003 
SNHG15 1.740 (1.437–2.103) <10−4 1.629 (1.311–1.986) <10−4 
linc-FAM98A-3 1.417 (1.173–1.711) 0.0003 1.522 (1.207–1.842) 0.0003 
Combination index 2.470 (1.900–3.210) <10−4 3.156 (2.328–4.007) <10−4 

Abbreviations: CI, confidence interval; IS, ischemic stroke; NA, not available; OR, odds ratio; TIA, transient ischemic stroke.

*

The reference group was HCs.

The reference group was TIA.

Adjustment for age, sex, medical history, and laboratory data on admission.

Predictive power of serum BDNF, NSE, and PBMC lncRNAs for neurological deficits

Serum BDNF concentrations have been identified to correlate with risk of incident stroke and TIA [27]. Serum NSE level, a sensitive and specific marker of brain damage, was associated with severity of neuronal damage and relative to stroke outcome [28,29]. To compare the ability of serum BDNF, serum NSE, and PBMC lncRNAs to predict neurological deficits, we measured serum BDNF and NSE levels by ELISA in 52 IS, 15 TIA patients, and 60 HCs of the test and external sets (Supplementary Table S4). The results showed that serum BDNF concentration was markedly down-regulated in IS patients compared with TIA patients and HCs (P<0.001, Figure 10A), while NSE concentration was significantly up-regulated in IS patients compared with HCs (P<0.001, Figure 10B). There was no significant difference in serum BDNF and NSE concentrations between TIA patients and HCs. The AUCs of serum BDNF and NSE for distinguishing IS patients from HCs were 0.789 and 0.709, respectively (Figure 10C). The AUC of BDNF for distinguishing IS from TIA was 0.723 (95% CI: 0.600–0.825, P=0.008). Taken together, these data demonstrated that the PBMC lncRNA-based combination index also outperformed serum BDNF and NSE in predicting neurological deficits.

Expression levels and diagnostic power of serum BDNF and NSE

Figure 10
Expression levels and diagnostic power of serum BDNF and NSE

Expression levels of serum BDNF (A) and NSE (B) were determined by ELISA. ROC curves for serum BDNF and NSE and their corresponding AUC values for discriminating IS patients from HCs (C). ***P<0.001.

Figure 10
Expression levels and diagnostic power of serum BDNF and NSE

Expression levels of serum BDNF (A) and NSE (B) were determined by ELISA. ROC curves for serum BDNF and NSE and their corresponding AUC values for discriminating IS patients from HCs (C). ***P<0.001.

Correlations of PBMC lncRNA levels with stroke severity

Given that lncRNAs are involved in IS with ischemia/reperfusion injury [30]. We hypothesized that the three identified PBMC lncRNAs may be associated with neurological deficit severity. To test this hypothesis, we analyzed their expression levels in different sets of IS patients classified by the NIHSS scores (Supplementary Table S5). The results showed that PBMC linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 levels in severe stroke were significantly higher than those in mild stroke; linc-DHFRL1-4 and SNHG15 levels in moderate stroke were also higher than those in mild stroke (Figure 11). Furthermore, the lncRNA-based combination index in severe and moderate stroke was significantly lower than mild stroke (P<0.001, Figure 11).

Correlations of lncRNA expression with neurological deficit severity of IS

Figure 11
Correlations of lncRNA expression with neurological deficit severity of IS

Correlations of lncRNA expression with neurological deficit severity of IS in the three sets (A–D). *P<0.05, ***P<0.001.

Figure 11
Correlations of lncRNA expression with neurological deficit severity of IS

Correlations of lncRNA expression with neurological deficit severity of IS in the three sets (A–D). *P<0.05, ***P<0.001.

Discussion

The present study successfully constructed a set of three PBMC lncRNAs (linc-DHFRL1-4, SNHG15, and linc-FAM98A-3), that were differentially expressed after acute IS based on an lncRNA microarray followed by individual qRT-PCR validation in three independent sets. We also identified their distinct expression profiles in IS and TIA patients. Our results revealed that these differentiated lncRNAs may be used to predict and discriminate the presence of IS from TIA and HC. Their expression signatures were significantly correlated with neurological deficit severity of IS and reflected dynamic changes in IS. We thus considered that these PBMC lncRNAs may function as diagnostic biomarkers for IS.

PBMCs can selectively migrate to and infiltrate ischemic cerebral tissue. PBMC-induced inflammation is considered a pivotal factor in aggravating cerebral infarcts and is, therefore, a potential therapeutic target [31,32]. Inflammatory and endothelial cells are activated by proinflammatory cytokines and chemokines, which contribute to form microvascular thrombosis [33]. In view of these observations, many investigators have begun to explore gene expression in PBMCs upon IS through high-throughput screening [14,34] and have demonstrated differences in the expression of several mRNAs between IS and HCs, which are diagnostic for IS. Moreover, the expression of miRNA, regulator of mRNA, was significantly different in PBMCs upon IS [17]. In this study, we sought to identify differentially expressed lncRNAs by lncRNA microarray analysis. We established a panel of three PBMC lncRNAs, which was confirmed in four independent sets. Collectively, PBMCs may be a promising diagnostic biomarker.

LncRNAs, a class of transcripts more than 200 nts without protein-coding properties, are involved in various biological and pathological processes [35]. Some lncRNAs have been identified to influence the development of neurological diseases [36,37]. Furthermore, emerging studies have reported lncRNA-mediated regulatory networks in IS [11,12,38]. However, the expression signatures of circulating lncRNAs in IS patients are rarely investigated, thus triggering our interest in utilizing specific lncRNAs as potential markers for IS identification. Previous studies have shown the diagnostic power of circulating lncRNAs for detecting disease. For example, a plasma-based lncRNA signature can discriminate patients with esophageal squamous cell carcinoma from HCs with an AUC value of 0.842 [25]. A serum five-lncRNA panel was reported as a diagnostic set of biomarkers for patients with clear cell renal cell carcinomas [39]. For IS, one study reported that plasma H19 can serve as a diagnostic biomarker [11]. Zhu and colleagues [40] has also reported that lncRNA MIAT in peripheral blood leukocytes could discriminate IS patients with a sensitivity of 0.741 and a specificity of 0.804. Nevertheless, the abovementioned studies focussing on lncRNAs were based on previous reports, and many lncRNAs remain uninvestigated.

Different from previous studies, we systematically analyzed lncRNA expression patterns in PBMCs from IS patients. Furthermore, we confirmed their independent predictive ability for onset of IS. To our knowledge, the expression levels of linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 in IS patients remain unclear. Therefore, the present study investigated their expression levels in IS patients for the first time and further focussed on their clinical significance as potential biomarkers for IS identification.

TIA is often considered a warning sign of imminent stroke and increases the risk of stroke [3]. Therefore, differentiating and predicting TIA presence would facilitate early intervention to prevent subsequent stroke and related disability. One study uncovered significant changes in serum miRNA expression in TIA patients, which may predict the risk of subsequent stroke [41]. However, the expression levels of PBMC lncRNAs and their potential clinical significance in TIA patients remain unclear. Strikingly, linc-DHFRL1-4, SNHG15, and linc-FAM98A-3 had good ability for discriminating TIA patients from IS patients. Unfortunately, only linc-DHFRL1-4 could discriminate between TIA patients and HCs. Taken together, these findings provide a new avenue regarding the use of PBMC lncRNAs as supporting biomarkers for TIA differentiation and prediction.

Understanding the molecular function of identified lncRNAs in IS would promote rapid intervention and stroke prevention. The three lncRNAs (linc-DHFRL1-4, SNHG15, and linc-FAM98A-3) have not been explored in IS, but lncRNA–gene interaction analysis showed that various immune-associated genes were correlated with the dysregulation of linc-DHFRL1-4, SNHG15, and linc-FAM98A-3, including interferon regulatory factor 9 (IRF9) (P<0.001). IRF9 is generally expressed and can be activated by interferon-γ (IFN-γ) [42]. Furthermore, IRF9 is also correlated with pathological stress and cell fate [43]. A recent study demonstrated that IRF9 facilitated the acetylation and activation of p53-mediated signaling through inhibition of Sirt1-mediated cell survival, aggravating cerebral injury [44]. Thus, these studies imply a role of dysregulated lncRNAs in the control of IRF9 upon IS.

The contributions of the present study are as follows: (i) the first analysis of genome-wide lncRNA microarrays of PBMC samples validated by individual qRT-PCR in three independent sets; (ii) construction and validation of a new lncRNA-based combination index; (iii)clarification of the expression patterns of the identified lncRNAs in IS and TIA; (iv) elimination of interference of other RNAs from red blood cells and extracellular RNAs, including exosomes, microvesicles, microparticles, and extracellular vesicles [45]; and (v) stable identification of lncRNAs in extreme conditions. However, several limitations of our study should be mentioned: (i) as the number of participants was relatively limited and all participants belonged to Chinese Han populations from different regions in China, the combination index in the diagnosis of IS should be evaluated in a larger number of patients from different races and countries. (ii) The lack of follow-up data on the prognostic value of the PBMC lncRNAs in IS patients hinder a better understanding of IS prevention and treatment. (iii) The lack of large number of patients on the temporal profile of PBMC lncRNAs levels after stroke. (iv) As patients with a history of stroke, MI, TIA, and infection were excluded, these markers may not apply to all patients with stroke. (v) The molecular mechanisms of PBMC lncRNAs in IS need to be further explored.

Summary

The present study applied a systematic strategy to identify a differential lncRNA expression profile in PBMCs of IS patients. These differentially expressed lncRNAs manifested discriminative and predictive potentials as new biomarkers for IS, allowing neurological deficit severity of IS to be estimated. Further studies are needed to clarify the molecular mechanisms of PBMC lncRNAs underlying cerebral ischemia.

Clinical perspectives

  • lncRNAs have been highlighted to be involved in the pathological process of IS.

  • The present study applied a systematic strategy to identify a differential lncRNA expression profile in PBMCs of IS patients.

  • These differentially expressed lncRNAs manifested discriminative and predictive potentials as new biomarkers for IS allowing estimate neurological deficit severity of IS to be estimated.

Competing interests

The authors declare that there are no competing interests associated with the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China [grant number 81271336]; the Science and Technology Development Plan of Nanjing Foundation [grant number 2016sc511020 (to F.-L.Y.)]; and the Fundamental Research Funds for the Central Universities [grant number KYZZ16_0130 (to Q.-W.D.)].

Author contribution

F.-L.Y. and Q.-W.D. participated in the design of the study and performed the statistical analysis. S.L., H.W., H.-L.S., L.Z., Z.-T.G., G.L., C.-Z.S., and H.-Q.Z. collected patient information and prepared the figures. Q.-W.D., S.L., and H.W. participated in the study design and co-ordination and helped in drafting the manuscript. All authors have read and approved the final manuscript.

Abbreviations

     
  • AUC

    area under the ROC curve

  •  
  • BDNF

    brain-derived neurotrophic factor

  •  
  • CI

    confidence interval

  •  
  • Ct

    cycle threshold

  •  
  • CT

    computed tomography

  •  
  • GAPDH

    glyceraldehyde phosphate dehydrogenase

  •  
  • HC

    healthy control

  •  
  • IS

    ischemic stroke

  •  
  • lncRNA

    long noncoding RNA

  •  
  • IRF9

    interferon regulatory factor 9

  •  
  • MIAT

    myocardial infarction associated transcript

  •  
  • NIHSS

    National Institutes of Health Stroke Scale

  •  
  • NSE

    neurone-specific enolase

  •  
  • PBMC

    peripheral blood mononuclear cell

  •  
  • qRT-PCR

    quantitative real-time PCR

  •  
  • ROC

    receiver-operating characteristic

  •  
  • Sirt1

    sirtuin 1

  •  
  • SNHG15

    small nucleolar RNA host gene 15

  •  
  • TIA

    transient ischemic attack

References

References
1
GBD Mortality and Causes of Death Collaborators
(
2016
)
Global, regional, and national life expectancy, all-cause mortality, and cause-specific mortality for 249 causes of death, 1980-2015: a systematic analysis for the Global Burden of Disease Study 2015
.
Lancet
388
,
1459
1544
[PubMed]
2
Saenger
A.K.
and
Christenson
R.H.
(
2010
)
Stroke biomarkers: progress and challenges for diagnosis, prognosis, differentiation, and treatment
.
Clin. Chem.
56
,
21
33
[PubMed]
3
Easton
J.D.
,
Saver
J.L.
,
Albers
G.W.
,
Alberts
M.J.
,
Chaturvedi
S.
,
Feldmann
E.
et al
(
2009
)
Definition and evaluation of transient ischemic attack: a scientific statement for healthcare professionals from the American Heart Association/American Stroke Association Stroke Council; Council on Cardiovascular Surgery and Anesthesia; Council on Cardiovascular Radiology and Intervention; Council on Cardiovascular Nursing; and the Interdisciplinary Council on Peripheral Vascular Disease. The American Academy of Neurology affirms the value of this statement as an educational tool for neurologists
.
Stroke
40
,
2276
2293
[PubMed]
4
Mullins
M.E.
,
Schaefer
P.W.
,
Sorensen
A.G.
,
Halpern
E.F.
,
Ay
H.
,
He
J.
et al
(
2002
)
CT and conventional and diffusion-weighted MR imaging in acute stroke: study in 691 patients at presentation to the emergency department
.
Radiology
224
,
353
360
[PubMed]
5
Whiteley
W.
,
Tseng
M.C.
and
Sandercock
P.
(
2008
)
Blood biomarkers in the diagnosis of ischemic stroke: a systematic review
.
Stroke
39
,
2902
2909
[PubMed]
6
Amouyel
P.
(
2012
)
From genes to stroke subtypes
.
Lancet Neurol.
11
,
931
933
[PubMed]
7
Ponting
C.P.
,
Oliver
P.L.
and
Reik
W.
(
2009
)
Evolution and functions of long noncoding RNAs
.
Cell
136
,
629
641
[PubMed]
8
Wang
K.C.
and
Chang
H.Y.
(
2011
)
Molecular mechanisms of long noncoding RNAs
.
Mol. Cell
43
,
904
914
[PubMed]
9
Schonrock
N.
,
Harvey
R.P.
and
Mattick
J.S.
(
2012
)
Long noncoding RNAs in cardiac development and pathophysiology
.
Circ. Res.
111
,
1349
1362
[PubMed]
10
Spizzo
R.
,
Almeida
M.I.
,
Colombatti
A.
and
Calin
G.A.
(
2012
)
Long non-coding RNAs and cancer: a new frontier of translational research?
Oncogene
31
,
4577
4587
[PubMed]
11
Wang
J.
,
Zhao
H.
,
Fan
Z.
,
Li
G.
,
Ma
Q.
,
Tao
Z.
et al
(
2017
)
Long noncoding RNA H19 promotes neuroinflammation in ischemic stroke by driving histone deacetylase 1-dependent M1 microglial polarization
.
Stroke
48
,
2211
2221
[PubMed]
12
Zhang
X.
,
Tang
X.
,
Liu
K.
,
Hamblin
M.H.
and
Yin
K.J.
(
2017
)
Long noncoding RNA Malat1 regulates cerebrovascular pathologies in ischemic stroke
.
J. Neurosci.
37
,
1797
1806
[PubMed]
13
Kassner
S.S.
,
Kollmar
R.
,
Bonaterra
G.A.
,
Hildebrandt
W.
,
Schwab
S.
and
Kinscherf
R.
(
2009
)
The early immunological response to acute ischemic stroke: differential gene expression in subpopulations of mononuclear cells
.
Neuroscience
160
,
394
401
[PubMed]
14
Grond-Ginsbach
C.
,
Hummel
M.
,
Wiest
T.
,
Horstmann
S.
,
Pfleger
K.
,
Hergenhahn
M.
et al
(
2008
)
Gene expression in human peripheral blood mononuclear cells upon acute ischemic stroke
.
J. Neurol.
255
,
723
731
[PubMed]
15
Bian
F.
,
Simon
R.P.
,
Li
Y.
,
David
L.
,
Wainwright
J.
,
Hall
C.L.
et al
(
2014
)
Nascent proteomes in peripheral blood mononuclear cells as a novel source for biomarker discovery in human stroke
.
Stroke
45
,
1177
1179
[PubMed]
16
Beer
L.
,
Mildner
M.
,
Gyongyosi
M.
and
Ankersmit
H.J.
(
2016
)
Peripheral blood mononuclear cell secretome for tissue repair
.
Apoptosis
21
,
1336
1353
17
Bam
M.
,
Yang
X.
,
Sen
S.
,
Zumbrun
E.E.
,
Dennis
L.
,
Zhang
J.
et al
(
2018
)
Characterization of dysregulated miRNA in peripheral blood mononuclear cells from ischemic stroke patients
.
Mol. Neurobiol.
55
,
1419
1429
[PubMed]
18
Deng
Q.W.
,
Wang
H.
,
Sun
C.Z.
,
Xing
F.L.
,
Zhang
H.Q.
,
Zuo
L.
et al
(
2017
)
Triglyceride to high-density lipoprotein cholesterol ratio predicts worse outcomes after acute ischaemic stroke
.
Eur. J. Neurol.
24
,
283
291
[PubMed]
19
(
1989
)
Stroke–1989. Recommendations on stroke prevention, diagnosis, and therapy. Report of the WHO Task Force on Stroke and other Cerebrovascular Disorders
.
Stroke
20
,
1407
1431
[PubMed]
20
Sacco
R.L.
,
Kasner
S.E.
,
Broderick
J.P.
,
Caplan
L.R.
,
Connors
J.J.
,
Culebras
A.
et al
(
2013
)
An updated definition of stroke for the 21st century: a statement for healthcare professionals from the American Heart Association/American Stroke Association
.
Stroke
44
,
2064
2089
[PubMed]
21
Adams
H.P.
Jr
,
Bendixen
B.H.
,
Kappelle
L.J.
,
Biller
J.
,
Love
B.B.
,
Gordon
D.L.
et al
(
1993
)
Classification of subtype of acute ischemic stroke. Definitions for use in a multicenter clinical trial. TOAST. Trial of Org 10172 in Acute Stroke Treatment
.
Stroke
24
,
35
41
[PubMed]
22
Sarker
S.J.
,
Rudd
A.G.
,
Douiri
A.
and
Wolfe
C.D.
(
2012
)
Comparison of 2 extended activities of daily living scales with the Barthel Index and predictors of their outcomes: cohort study within the South London Stroke Register (SLSR)
.
Stroke
43
,
1362
1369
[PubMed]
23
Deng
Q.
,
He
B.
,
Gao
T.
,
Pan
Y.
,
Sun
H.
,
Xu
Y.
et al
(
2014
)
Up-regulation of 91H promotes tumor metastasis and predicts poor prognosis for patients with colorectal cancer
.
PLoS ONE
9
,
e103022
[PubMed]
24
DeLong
E.R.
,
DeLong
D.M.
and
Clarke-Pearson
D.L.
(
1988
)
Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach
.
Biometrics
44
,
837
845
[PubMed]
25
Tong
Y.S.
,
Wang
X.W.
,
Zhou
X.L.
,
Liu
Z.H.
,
Yang
T.X.
,
Shi
W.H.
et al
(
2015
)
Identification of the long non-coding RNA POU3F3 in plasma as a novel biomarker for diagnosis of esophageal squamous cell carcinoma
.
Mol. Cancer
14
,
3
[PubMed]
26
Zhang
K.
,
Shi
H.
,
Xi
H.
,
Wu
X.
,
Cui
J.
,
Gao
Y.
et al
(
2017
)
Genome-wide lncRNA microarray profiling identifies novel circulating lncRNAs for detection of gastric cancer
.
Theranostics
7
,
213
227
[PubMed]
27
Pikula
A.
,
Beiser
A.S.
,
Chen
T.C.
,
Preis
S.R.
,
Vorgias
D.
,
DeCarli
C.
et al
(
2013
)
Serum brain-derived neurotrophic factor and vascular endothelial growth factor levels are associated with risk of stroke and vascular brain injury: Framingham Study
.
Stroke
44
,
2768
2775
[PubMed]
28
Isgro
M.A.
,
Bottoni
P.
and
Scatena
R.
(
2015
)
Neuron-specific enolase as a biomarker: biochemical and clinical aspects
.
Adv. Exp. Med. Biol.
867
,
125
143
[PubMed]
29
Jauch
E.C.
,
Lindsell
C.
,
Broderick
J.
,
Fagan
S.C.
,
Tilley
B.C.
,
Levine
S.R.
et al
(
2006
)
Association of serial biochemical markers with acute ischemic stroke: the National Institute of Neurological Disorders and Stroke recombinant tissue plasminogen activator Stroke Study
.
Stroke
37
,
2508
2513
[PubMed]
30
Saugstad
J.A.
(
2015
)
Non-coding RNAs in stroke and neuroprotection
.
Front. Neurol.
6
,
50
[PubMed]
31
del Zoppo
G.J.
and
Hallenbeck
J.M.
(
2000
)
Advances in the vascular pathophysiology of ischemic stroke
.
Thromb. Res.
98
,
73
81
[PubMed]
32
Kochanek
P.M.
and
Hallenbeck
J.M.
(
1992
)
Polymorphonuclear leukocytes and monocytes/macrophages in the pathogenesis of cerebral ischemia and stroke
.
Stroke
23
,
1367
1379
[PubMed]
33
Wang
X.
,
Louden
C.
,
Yue
T.L.
,
Ellison
J.A.
,
Barone
F.C.
,
Solleveld
H.A.
et al
(
1998
)
Delayed expression of osteopontin after focal stroke in the rat
.
J. Neurosci.
18
,
2075
2083
[PubMed]
34
Moore
D.F.
,
Li
H.
,
Jeffries
N.
,
Wright
V.
,
Cooper
R.A.
Jr
,
Elkahloun
A.
et al
(
2005
)
Using peripheral blood mononuclear cells to determine a gene expression profile of acute ischemic stroke: a pilot investigation
.
Circulation
111
,
212
221
[PubMed]
35
Gong
C.
and
Maquat
L.E.
(
2011
)
lncRNAs transactivate STAU1-mediated mRNA decay by duplexing with 3′ UTRs via Alu elements
.
Nature
470
,
284
288
[PubMed]
36
Wu
P.
,
Zuo
X.
,
Deng
H.
,
Liu
X.
,
Liu
L.
and
Ji
A.
(
2013
)
Roles of long noncoding RNAs in brain development, functional diversification and neurodegenerative diseases
.
Brain Res. Bull.
97
,
69
80
[PubMed]
37
Tollervey
J.R.
,
Curk
T.
,
Rogelj
B.
,
Briese
M.
,
Cereda
M.
,
Kayikci
M.
et al
(
2011
)
Characterizing the RNA targets and position-dependent splicing regulation by TDP-43
.
Nat. Neurosci.
14
,
452
458
[PubMed]
38
Dykstra-Aiello
C.
,
Jickling
G.C.
,
Ander
B.P.
,
Shroff
N.
,
Zhan
X.
,
Liu
D.
et al
(
2016
)
Altered expression of long noncoding RNAs in blood after ischemic stroke and proximity to putative stroke risk loci
.
Stroke
47
,
2896
2903
[PubMed]
39
Wu
Y.
,
Wang
Y.Q.
,
Weng
W.W.
,
Zhang
Q.Y.
,
Yang
X.Q.
,
Gan
H.L.
et al
(
2016
)
A serum-circulating long noncoding RNA signature can discriminate between patients with clear cell renal cell carcinoma and healthy controls
.
Oncogenesis
5
,
e192
[PubMed]
40
Zhu
M.
,
Li
N.
,
Luo
P.
,
Jing
W.
,
Wen
X.
,
Liang
C.
et al
(
2018
)
Peripheral blood leukocyte expression of lncRNA MIAT and its diagnostic and prognostic value in ischemic stroke
.
J. Stroke Cerebrovasc. Dis.
27
,
326
337
[PubMed]
41
Wu
J.
,
Fan
C.L.
,
Ma
L.J.
,
Liu
T.
,
Wang
C.
,
Song
J.X.
et al
(
2017
)
Distinctive expression signatures of serum microRNAs in ischaemic stroke and transient ischaemic attack patients
.
Thromb. Haemost.
117
,
992
1001
[PubMed]
42
Tamura
T.
,
Yanai
H.
,
Savitsky
D.
and
Taniguchi
T.
(
2008
)
The IRF family transcription factors in immunity and oncogenesis
.
Annu. Rev. Immunol.
26
,
535
584
[PubMed]
43
Tsuno
T.
,
Mejido
J.
,
Zhao
T.
,
Schmeisser
H.
,
Morrow
A.
and
Zoon
K.C.
(
2009
)
IRF9 is a key factor for eliciting the antiproliferative activity of IFN-alpha
.
J. Immunother.
32
,
803
816
[PubMed]
44
Chen
H.Z.
,
Guo
S.
,
Li
Z.Z.
,
Lu
Y.
,
Jiang
D.S.
,
Zhang
R.
et al
(
2014
)
A critical role for interferon regulatory factor 9 in cerebral ischemic stroke
.
J. Neurosci.
34
,
11897
11912
[PubMed]
45
Ma
J.
,
Lin
Y.
,
Zhan
M.
,
Mann
D.L.
,
Stass
S.A.
and
Jiang
F.
(
2015
)
Differential miRNA expressions in peripheral blood mononuclear cells for diagnosis of lung cancer
.
Lab. Invest.
95
,
1197
1206
[PubMed]

Author notes

*

These authors contributed equally to this work.