Abstract Diabetes mellitus (DM) is a significant risk factor for peripheral arterial disease (PAD), and PAD is an independent predictor of cardiovascular disorders (CVDs). Growing evidence suggests that long non-coding RNAs (lncRNAs) significantly contribute to disease development and underlying complications, particularly affecting smooth muscle cells (SMCs). So far, no study has focused on transcriptome analysis of lncRNAs in PAD patients with and without DM. Tissue samples were obtained from our Vascular Biobank. Due to the sample’s heterogeneity, expression analysis of lncRNAs in whole tissue detected only ACTA2-AS1 with a 4.9-fold increase in PAD patients with DM. In contrast, transcriptomics of SMCs revealed 28 lncRNAs significantly differentially expressed between PAD with and without DM (FDR < 0.1). Sixteen lncRNAs were of unknown function, six were described in cancer, one connected with macrophages polarisation, and four were associated with CVDs, mainly with SMC function and phenotypic switch (NEAT1, MIR100HG, HIF1A-AS3, and MRI29B2CHG). The enrichment analysis detected additional lncRNAs H19, CARMN, FTX, and MEG3 linked with DM. Our study revealed several lncRNAs in diabetic PAD patients associated with the physiological function of SMCs. These lncRNAs might serve as potential therapeutic targets to improve the function of SMCs within the diseased tissue and, thus, the clinical outcome. Subject terms: Molecular medicine, Cardiovascular biology Introduction Peripheral arterial disease (PAD), defined as the narrowing of the peripheral arteries, is primarily caused by atherosclerotic changes within the vessel wall, particularly in the lower extremities. Diabetes mellitus (DM) is a significant risk factor for PAD, and PAD is an independent predictor of cardiovascular and cerebrovascular ischemic events, affecting both the quality and expectancy of life^[36]1–[37]3. In diabetic patients, atherosclerotic lesions occur earlier with rapid progression and are frequently asymptomatic, thus bearing a high risk of unexpected cardiac and cerebral complications^[38]2,[39]3. Multiple metabolic aberrations, such as overproduction of advanced glycation end-products (AGEs), increased oxidative stress, reactive oxygen species (ROS), enhanced inflammation, and dyslipidaemia, have been shown to aggravate PAD in patients with DM^[40]3–[41]5. The metabolic changes strongly affect the biological function of endothelial cells (ECs) and upregulate the expression of many inflammatory factors^[42]6–[43]8. These cytokines, in turn, promote atherosclerosis as well as apoptosis of ECs. Increased inflammation, as well as oxidative stress, facilitate the development of atherosclerotic lesions within the arterial wall. In PAD patients, atherosclerosis increases the risk of thrombosis and lower extremity ulceration^[44]1,[45]2. Furthermore, individuals with PAD suffering from DM are more susceptible to plaque rupture^[46]9,[47]10. In addition, hyperglycaemia in diabetic patients induces the production of ROS and AGEs, leading to vascular damage and diminished bioavailability of nitric oxide (NO)^[48]3,[49]6,[50]11,[51]12. NO and prostacyclin PGI2 are the essential vasoactive factors affecting the underlying smooth muscle cells (SMCs)^[52]3,[53]13–[54]16. The impaired ECs, together with other metabolic aberrations due to DM, impair the physiological function of SMCs by changing their phenotype in a proatherogenic manner, leading increasingly also to apoptosis^[55]17,[56]18. Growing evidence suggests that particularly long non-coding RNAs (lncRNAs), participating in many biological processes such as transcriptional regulation of oxidative stress, inflammation, atherosclerosis, as well as insulin sensitivity, may also affect PAD development, especially in patients suffering from DM^[57]19,[58]20. Furthermore, many lncRNAs have already been described to significantly affect the phenotype and the physiological behaviour of vascular cells in DM, as well as the crosstalk between ECs and SMCs^[59]21,[60]22. Thus, dysregulation of lncRNAs in DM patients suffering from PAD may significantly affect the biological function mainly observed in SMCs and thus aggravate the clinical outcome for these patients. Combining clinical data with omics analyses and bioinformatics is a useful strategy for improving diagnostic precision and pursuing personalised medicine in the treatment of CVDs. Particularly, bioinformatics analysis of big data and the use of artificial intelligence may help to discover novel useful targets or biomarkers with the potential for more accurate diagnosis and therapy^[61]23–[62]25. Interestingly, no study has so far focused on the analysis of lncRNAs in diabetic patients with PAD using RNA sequencing of the corresponding tissue. Performing an extended PubMed research, we found no corresponding relevant work dealing with PAD or DM patients and lncRNAs, particularly in humans. Therefore, having the advantage of possessing such tissue samples in our vascular biobank, we performed a detailed transcriptome analysis of PAD patients with and without DM, focusing particularly on lncRNAs and SMCs. Results Patient characteristics The clinical characteristics of the selected study patients included in the transcriptome analysis of lncRNA, associated diseases, and corresponding medication are summarised in Table [63]1. No significant differences were observed between the study groups of PAD patients with and without DM with regard to age, gender or any other collected clinical data. The mean age of PAD patients with and without DM was 74.9 ± 9.2 and 71.9 ± 11.2 years, 52.9% and 40.0% were of the male sex. Most of the patients suffered from chronic kidney disease (82.4% and 80.0%). Furthermore, 58.8% and 65.0 of the study participants received aspirin or clopidogrel, 64.8% and 40.0% beta-blockers and/or ACE inhibitors, and 64.7 and 75.0% were on statins. The only divergence was observed for diuretic intake, with 52.5% of individuals suffering from DM compared to 15.0% of patients without DM (P = 0.038). Table 1. Study patients' clinical data. DMplus DMminus P value (n = 17) (n = 20) Age (years) 74.9 ± 9.2 71.9 ± 11.2 0.240 Sex (male) 52.9% 40.0% 0.340 Hypertension 94.1% 80.0% 0.495 Hyperlipidaemia 70.6% 60.0% 0.495 Smoking 41.2% 35.0% 1.000 Chronic kidney disease 82.4% 80.0% 1.000 Cardiovascular disease 47.1% 25.0% 0.313 Stroke 5.9% 10.0% 1.000 Aspirin / Clopidogrel 58.8% 65.0% 0.508 Beta-blocker 64.8% 40.0% 0.194 ACE inhibitors 41.2% 45.0% 0.743 Statins 64.7% 75.0% 0.728 Diuretics 52.9% 15.0% 0.038 [64]Open in a new tab Pathomorphological analysis At first, we performed histological and immunohistochemical (IHC) staining of the FFPE samples from the diseased iliac artery (n = 79) to obtain an insight into the pathomorphology of the tissue of our study patients. The specimen characterisation revealed highly heterogeneous pathophysiological features (Fig. [65]1). Figure 1. [66]Figure 1 [67]Open in a new tab Selected histological examples of the tissue samples from PAD patients with and without DM. (a) and (b) Haemalaun-Eosin staining of PAD tissue with and without DM. Most samples were highly atherosclerotic and heavily calcified (arrows). (c) and (d) Smooth muscle actin staining. (e) and (f): Example of MYH10 (synthetic phenotype of SMCs) and MYH11 (contractile phenotype of SMCs) staining. High overlapping was observed. (g) Example of ECs staining using CD31. (h) Example of leukocyte staining using CD45. (i) Example of macrophage staining using CD68. Scale bars: 1 mm (a–d), 100 µm (e–i). Most samples had extended atherosclerotic lesions, inflammation, calcification, and were highly vascularised. Interestingly, staining with MYH10 and MYH11 to distinguish between the synthetic and contractile phenotype of SMCs showed high overlapping characteristics and no clear separation between these two morphologic features was observed (Fig. [68]1e and f). In order to reduce the broad heterogeneity of our study tissue, samples with a great extent of calcification, inflammation, and a low number of cells, which markedly differed from the average, were excluded. Consequently, from the 79 histologically characterised patients, 37 were finally included in the transcriptomics analysis (DMplus: n = 17, DMminus: n = 20). The semi-quantitative pathomorphological analysis of the included study samples revealed no significant differences between the study groups (Table [69]2). Table 2. Histological characterisation* of the tissue samples. DMplus DMminus P value (n = 17) (n = 20) Infiltrates 20.3% 25.5% 0.150 SMCs 43.4% 50.8% 0.822 Neovessel 27.3% 33.3% 0.233 Thrombus (extent) 29.3% 19.8% 0.111 Atherosclerosis (extent) 19.3% 18.3% 0.893 Calcification / extent 11.6% 15.8% 0.518 [70]Open in a new tab *Using HE and EvG staining, the individual pathological features of the tissue samples were evaluated semi-quantitatively. 100% corresponds to the highest observed occurrence of the individual features. Differential expression analysis of lncRNA in the whole PAD tissue samples First, we analysed the RNA from the whole tissue samples, focusing on lncRNAs. The volcano plot, comparing PAD patients with and without DM, is shown in Fig. [71]2. In total, 13,491 lncRNAs were detected, with 6672 above the threshold of 10 counts per million. Nevertheless, only one lncRNA, ACTA2-AS1, demonstrated significantly differential expression between the study groups (4.9-fold increase in DMplus samples, FDR = 0.031, P = 2.7E−06) (Fig. [72]2a, Table [73]3). Figure 2. [74]Figure 2 [75]Open in a new tab Volcano plot comparing differentially expressed lncRNAs in the whole tissue (a) and in SMCs (b) from PAD patients with and without DM (DMplus–over–DMminus). The plot shows − log10 transformed FDR as a function of the difference between study groups. The broken lines show 0.5 log2 fold change. Statistically significant differences: FDR < 0.1. Table 3. LncRNAs with significant expression differences between study groups (FDR P-value < 0.1). Gene symbol Fold change P-value FDR P-value Gene function (if described) Literature DMplus versus DMminus (whole tissue) ACTA2-AS1 4.90 2.7E−06 0.031 phenotypic switch SMCs, proliferation and migration of cancer cells ^[76]26–[77]28 DMplus versus DMminus (SMCs) ENSG00000289474 − 12.21 7.7E−27 5.8E−24 Novel transcript – NEAT1 − 2.35 3.2E−15 1.2E−12 SMC phenotype, proliferation, atherosclerosis, aneurysm ^[78]41,[79]44,[80]46,[81]52–[82]56 ENSG00000288156 − 0.23 2.1E−11 5.3E−09 Novel transcript – XIST − 3.23 3.4E−10 6.5E−08 SMC proliferation, migration, apoptosis ^[83]51,[84]57,[85]58 ENSG00000289901 − 3.73 2.5E−09 6.5E−08 Novel transcript – MIR222HG − 3.53 8.6E−09 1.1E−06 Macrophages polarisation ^[86]37,[87]38 LINC01220 − 0.40 8.3E−07 9.1E−05 Cancer ^[88]33 CYTOR − 2.95 2.3E−06 2.2E−04 Cell proliferation and migration in cancer ^[89]35 ENSG00000288794 − 2.39 3.4E−06 2.8E−04 Novel transcript – ENSG00000288928 − 2.86 3.8E−06 2.9E−04 Novel transcript – LINC00910 2.35 4.2E−06 2.9E−04 Colorectal and breast cancer ^[90]34 ENSG00000291174 2.41 6.8E−05 0.0043 Novel transcript – LINC00511 − 2.27 1.5E−04 0.0085 Cell proliferation in cancer ^[91]32 ENSG00000267520 − 2.04 2.1E−04 0.0111 Novel transcript – MIR23AHG − 1.71 3.9E−04 0.0196 Function unknown – RUFY1-AS1 1.88 8.6E−04 0.0409 Function unknown – ENSG00000271959 − 2.00 9.4E−04 0.0422 Novel transcript – ENSG00000248994 − 1.49 0.0012 0.0492 Novel transcript – MIR100HG − 1.99 0.0019 0.0692 Regulator of cell proliferation, cardiomyopathy ^[92]48 PURPL 1.60 0.0019 0.0692 Regulator of cell proliferation in cancer ^[93]30 ENSG00000281195 − 1.75 0.0020 0.0692 Novel transcript – MZF1-AS1 1.86 0.0022 0.0693 Cancer ^[94]31 HIF1A-AS3 − 1.88 0.0023 0.0728 SMC phenotype ^[95]47,[96]50,[97]59,[98]60 ENSG00000290021 − 1.68 0.0023 0.0729 Novel transcript – ENSG00000279175 − 1.69 0.0024 0.0740 Novel transcript – MIR29B2CHG − 1.87 0.0032 0.0924 Heart failure ^[99]42 ENSG00000289404 − 1.50 0.0033 0.0925 Novel transcript – ENSG00000269940 − 1.72 0.0035 0.0940 Novel transcript – [100]Open in a new tab The heatmap of the most differentially expressed lncRNAs in the tissue of PAD patients with DM (DMplus) and without DM (DMminus) revealed broad heterogeneity of the individual samples regarding the expression of lncRNAs (Fig. [101]3). Even if many samples were clustered, each study group was distributed throughout the whole heatmap. Figure 3. [102]Figure 3 [103]Open in a new tab Heatmap of the most differentially expressed lncRNAs in the whole tissue of PAD patients with (DMplus) and without (DMminus) diabetes mellitus (rows indicate the expression of 241 most expressed lncRNAs, columns indicate the individual samples). Clustering was performed using conditional formatting features. Even if many samples are clustered, high heterogeneity was observed regarding the study groups. Differential expression analysis of lncRNA in SMCs of the PAD tissue samples Regardless of the preselection of the study samples according to their histology, wide heterogeneity of the lncRNA expression in the whole tissue was still observed. Therefore, we changed our experimental approach and focused only on the SMCs within the study samples. Consequently, we performed a dissection of the tissue and analysed the lncRNAs only from the microdissected SMC areas. The corresponding volcano plot is shown in Fig. [104]2b, and the heatmap of the clustering of the differentially expressed lncRNAs is shown in Fig. [105]4. The expression of lncRNA only from SMCs showed markedly higher homogeneity and better distribution (clustering) of the study samples (Fig. [106]4) compared to the heatmap of the whole tissue (Fig. [107]3). Figure 4. [108]Figure 4 [109]Open in a new tab Heatmap of the most differentially expressed lncRNAs in the SMCs from PAD patients with (DMplus) and without (DMminus) diabetes mellitus (rows indicate the expression of 28 most expressed lncRNAs, columns indicate the individual samples). Clustering was performed using conditional formatting features. In total, 14,499 lncRNAs could be detected, with 761 lncRNAs with counts above the threshold of 10 counts per million. From these 761 lncRNAs, 28 demonstrated significantly differential expression between the study groups (FDR < 0.1) (Table [110]3). Of these 28 lncRNAs, 14 were novel transcripts not yet described. The function of two other lncRNAs, MIR23AHG and RUFY1-AS1, is also unknown (Table [111]3). Six of the remaining lncRNAs have been described in different types of cancer (LINC01220, CYTOR, LINC00910, LINC00511, PURPL, and MZF1-AS1), one with macrophage polarisation (MIR222HG), and five of these lncRNAs could be associated with cardiovascular disorders (CVDs), mainly with SMC proliferation, migration, apoptosis, and phenotypic switch (NEAT1, XIST, MIR100HG, HIF1A-AS3, and MIR29B2CHG) (Table [112]3). The latter lncRNAs were further investigated and RT-PCR was performed to confirm the results from the transcriptome analysis (Fig. [113]5). Figure 5. [114]Figure 5 [115]Open in a new tab Box plots using RT-PCR to compare the expression of relevant lncRNAs in SMCs between the study groups (with and without DM). The lower part shows subsequent analysis of XIST and MIG222HG separated for sex because these two lncRNAs are expressed only on chromosome X. The expression was normalised for the housekeeping gene GAPDH. Furthermore, for better comparison, the expression of the individual lncRNAs of the study group without DM (w/o DM) was set as references equal 1.