*2.2. Patients*

A total of 34 HHT patient plasma samples were evaluated by quantitative reverse transcriptase–polymerase chain reaction (qRT-PCR). Seventeen of them were HHT1, and the remaining seventeen patients were HHT2. Plasma samples from 16 control healthy subjects were also assayed to establish the normal range of microRNAs (miRNAs). In all three groups, donors from di fferent sex and age were included. Peripheral venous blood samples were collected with ethylenediamine tetra-acetic acid (EDTA) as anticoagulant. Blood samples were centrifuged at 15,000 *g* for 15 min, and the resulting plasma was stored at −80 ◦C until analysis. Written informed consent was obtained from all the participants, or their legally authorized representative, in this study, and the protocol was supervised and received full approval from our Institutional Review Board (IRB) of the Spanish National Research Council (CSIC) with the ethical code number 075/2017. All HHT patients included in the present study were clinically diagnosed following the Curaçao criteria [3], and their genetic mutations were identified by sequencing. Their genetic characteristics and clinical manifestations are summarized in Table 1.

**Table 1.** Summary of genotypes and mutations of hereditary hemorrhagic telangiectasia (HHT) of patients 1.



**Table 1.** *Cont*.

1 A total of 33 HHT patients were included in this microRNA analyses. All HHT patients were clinically diagnosed following the Curaçao criteria [3]. Sixteen patients were genetically diagnosed as HHT1 as they harbor a mutation in *ENG*, whereas the remaining 17 patients were HHT2 with pathogenic mutations in *ALK1*; fs—frameshift mutation; Asterisks (\*) represent stop codons.

#### *2.3. Extraction of miRNAs from Plasma Samples*

miRNAs were extracted with the miRNeasy Micro kit (Qiagen, Hilden, Germany; #217084) according to the manufacturer's instructions and following a previously described protocol [25,31]. Briefly, 50 μL plasma samples from HHT patients were homogenized with 1 mL QIAzol ® Lysis Reagent and 6.25 × 10−<sup>3</sup> fmol/μL of the spike-in control cel-miR-39-3p (5 UCA CCG GGU GUA AAU CAG CUU G 3). Next, the homogenate was incubated at room temperature for 5 min, and then 200 μL chloroform was added to each sample, followed by centrifugation at 12,000× *g* for 5 min at 4 ◦C. Equal volumes from the upper aqueous phase were transferred to new tubes and mixed by pipetting with 1.5 volumes of 100% ethanol; finally total miRNAs were eluted in 40 μL RNase-free water.

#### *2.4. Quantitative Reverse Transcriptase–Polymerase Chain Reaction (qRT-PCR) Analysis of miRNAs*

qRT-PCR was performed to validate selected miRNAs. They were reverse-transcribed into cDNA using TaqMan ™ MicroRNA Reverse Transcription kit (Thermo Fisher Scientific, #4366596) following the manufacturer´s protocol. The reaction components were mixed with the corresponding reverse transcriptase (RT) primers of hsa-miR-370-3p, hsa-miR-214-3p, hsa-miR-10a-3p, hsa-miR-16-5p, and cel-miR-39-3p (Thermo Fisher Scientific, Waltham, MA, USA; #4427975; Assay ID 002275, 002306, 002288, 000391, and 000200, respectively). RT reaction was carried out as follows: 16 ◦C for 30 min, 42 ◦C for 30 min, 85 ◦C for 5 min, and 4 ◦C hold. The resulting cDNAs were used for quantitative real-time PCR experiments using Taqman Universal PCR Master Mix, no AmpErase ® UNG (Thermo Fisher Scientific, Waltham, MA, USA; #4324018), and the specific PCR primers (references detailed above). Thermal cycling was performed on LightCycler ® 96 detection system (Roche) as follows: 95 ◦C for 10 min, 40 cycles of 95 ◦C for 15s and 60 ◦C for 1 min. Relative quantification of individual miRNA expression was carried out with the 2-ΔΔCT method and normalized against two internal controls, miR-16-5p and the spike-in miRNA cel-miR-39-3p. In studies focused on cardiovascular disease, miR-16-5p is one of the miRNAs with the best performance as normalizer [32], whereas cel-miR-39-3p is one of the most used and reliable spike-in controls [25].

#### *2.5. Statistical Analysis*

HHT patient and control groups were compared using Kruskal–Wallis test. Subsequent Mann–Whitney U tests were run to test for pairwise comparisons in a post hoc fashion; significance values were adjusted by the Bonferroni correction for multiple tests. Statistical analyses were carried out with the IBM SPSS Statistics version 25 (Windows10 64-bit) software (IBM Corp., Armonk, NY, USA). Box Whisker plots show median (central line), upper and lower quartiles (box), and range excluding outliers (whiskers). Asterisks indicate statistically significant values between selected conditions (\* *p* < 0.05; \*\* *p* < 0.01; ns, not significant).

## **3. Results**

#### *3.1. Identification of miRNA-10a, miRNA-214, and miRNA-370, Potentially Related to HHT, Using in Silico and Literature Data*

Recently, studies on circulating miRNAs as potential biomarkers in different types of diseases [33], including HHT [21,27], have received increasing attention. In order to identify specific biomarkers of HHT, we searched for miRNAs involved in angiogenesis and vascular homeostasis and predicted to target *ENG* and/or *ALK1* using three robust target prediction programs which use different algorithms: (i) microRNA [28]; (ii) Target Scan [29]; and (iii) MicroCosm [30]. Using this stringent approach, we selected miRNA-370, miRNA-10a, and miRNA-214 for further analyses.

miRNA-370 was considered to be very relevant to endoglin function since it was the only miRNA predicted to target *ENG* with a good mirSVR score (−0.1740), according to the microRNA database (www.microRNA.org). This website also describes miRNA-370 as an evolutionary well-conserved miRNA because its alignment with *ENG* is found in other species as well. Likewise, algorithms from Target Scan and MicroCosm programs indicate that *ENG* is predicted to be targeted by miRNA-370. In fact, the negative regulation of *ENG* by miRNA-370 has been validated in ovarian cancer cells where miRNA-370 suppresses their proliferation and promotes chemosensitivity to cisplatin by negatively regulating *ENG* [34]. Based on predictions of microRNA webs, miRNA-370 not only regulates *ENG*, but also other gene products with essential roles in endothelial cell biology and angiogenesis, some of which have been already validated as targets in the literature. One of these proteins is TGF-β receptor type 2 (TGFBR2) that is negatively regulated by miRNA-370. By acting via TβRII, miRNA-370 plays a potential role in hepatic ischaemia-reperfusion injury and indeed, its inhibition efficiently attenuates liver damage [35]. In addition, upregulation of miRNA-370 might promote the repair of amputated fingers by regulating angiogenesis through targeting Forkhead box protein O1 (FOXO1) [36]. miRNA-370 can also induce growth and tube formation inhibition, and apoptosis in endothelial cells [37,38]. The miRNA-370-induced endothelial effects may explain its anti-angiogenic activity, as well as its developmental regulation of cerebral aneurysms [38–40]. Noteworthy, the effects miRNA-370 are mediated by targeting, at least, a receptor for vascular endothelial growth factor (VEGF), which is the major driver of angiogenesis [38,39].

We next focused our interest on miRNA-10a and miRNA-214 because both are predicted to target *ENG* or *ALK1*, and their dysregulated expression in HHT patients with pulmonary AVMs (pAVMs) has been reported [26] Thus, *ENG* has been revealed, by MicroCosm and Target Scan, as a confident target of miRNA-214. A search in microRNA.org revealed that miRNA-214 is also expected to bind ALK1 mRNA at two different sites with high mirSVR scores (−0.1277 and −0.5112), and ranks in the first position among all miRNAs potentially targeting ALK1. Moreover, by targeting matrix metalloproteinase 8 (MMP8), hepatoma-derived growth factor (HDGF), brain-specific angiogenesis inhibitors (BAIs), and other vascular-related genes, miRNA-214 contributes to the pathogenesis of various cardiovascular conditions, including ischaemic heart diseases, angiogenesis, and cardiac hypertrophy [41–43]. Of note, miRNA-214 is a response element to hypoxia in patients with pulmonary arterial hypertension (PHA), a complication of HHT. Moreover, inhibition of miRNA-214 can ameliorate the symptoms of PHA in animal models, suggesting its use for the prevention and treatment of PHA in humans [41].

Bioinformatics analysis of microRNA database showed that miRNA-10a does not target *ENG*, but appears to regulate *ALK1* with a good mirSVR score (−0.3837). In addition, miRNA-10a targets endothelial gene products such as VEGF receptor 1 (FLT1), β-catenin, GATA-binding factor 6 (GATA6), or mib-1, which may account for its active regulatory role in endothelial cell biology and angiogenesis [44–46].

#### *3.2. Circulating Levels of miRNA-370, miRNA-10a, and miRNA-214 in HHT1 and HHT2 Patient Plasma*

Taken together, the above studies sugges<sup>t</sup> that miRNA-370, miRNA-10a, and miRNA-214 are predicted to target not only *ENG* and/or *ALK1*, but also other relevant gene products involved in vascular functions related to the pathophysiology of HHT. Therefore, the chosen miRNAs were next validated by qRT-PCR in the plasma samples from a cohort of HHT1 and HHT2 patients and healthy controls in order to assess their possible diagnostic or biomarker value.

#### 3.2.1. Circulating Levels of miRNA-370 are Decreased in HHT1 Patient Plasma

The expression levels of miRNA-370, as measured by qRT-PCR, were found to be significantly lower in plasma samples from HHT1 patients (Figure 1), using miRNA-16 (Figure 1A; *p* = 0.001) or cel-miR-39-3p (Figure 1B; *p* = 0.004), as normalizers when compared with healthy controls. The levels of miRNA-370 in HHT1 patients were also significantly lower than those in HHT2 patients when using miR-16 as a normalizer (Figure 1A; *p* = 0.019), and showed a not significant but clear decreasing trend versus HHT2 samples when using cel-miR-39-3p as a normalizer (Figure 1B; *p* = 0.116). By contrast, the expression levels of miRNA-370 in HHT2 patients were not significantly a ffected compared with healthy control individuals. The specific downregulation in HHT1 patients suggests that the decreased levels of miRNA-370 may have potential diagnostic utility as an HHT1 biomarker. Interestingly, since miRNA-370 is predicted to target *ENG* to downregulate its expression [34] and a deficient *ENG* expression underlies the pathogenicity of HHT1, both results also sugges<sup>t</sup> the existence of a common link between the expression levels of miRNA-370 and endoglin in HHT1 patients.

**Figure 1.** Quantitative Reverse Transcriptase–Polymerase Chain Reaction (qRT-PCR) of miRNA-370. Total plasma RNA was isolated from HHT1 and HHT2 patients and control subjects. Relative expression levels of miR-370 were measured by qRT-PCR using miR-16 ( **A**) or cel-miR-39-3p (**B**) as normalizers. The number of samples analyzed is indicated in parentheses. Symbols outside the box plot represent extreme values (◦) and outliers (\*) with their corresponding sample numbers. Experiments were performed in triplicates. (\* *p* < 0.05; \*\* *p* < 0.01; ns, not significant).

#### 3.2.2. Circulating Levels of miRNA-10a, but not of miRNA-214, are Dysregulated in HHT2 Patient Plasma

As measured by qRT-PCR, the expression levels of miRNA-10a were significantly higher in plasma samples from HHT2 patients (Figure 2), using miRNA-16 (Figure 2A; *p* = 0.026) or cel-miR-39-3p (Figure 2B; *p* = 0.024), as normalizers, when compared with healthy controls. By contrast, the expression levels of miRNA-10a in HHT1 patients were not significantly a ffected compared with healthy control individuals. Even more, the levels of miRNA-10a in HHT2 patients were significantly higher than those in HHT1 patients when using miR-16 (Figure 2A; *p* = 0.03) or cel-miR-39-3p (Figure 2B; *p* = 0.018), as normalizers. The specific upregulation in HHT2 patients suggests that the increased levels of miRNA-10a may have potential diagnostic utility as an HHT2 biomarker. Interestingly, since miRNA-10 potentially targets ALK1 and a deficient ALK1 expression underlies the pathogenicity of HHT2, these results also sugges<sup>t</sup> the existence of a common link between the expression levels of both miRNA-10a and ALK1 in HHT2 patients. Plasma levels of miRNA-214 were also measured by qRT-PCR, using miR-16 or cel-miR-39-3p as normalizers (Figure 2C,D). However, no statistically significant di fferences

were found in the expression levels of miRNA-214 from either HHT1 or HHT2 patients, when compared with healthy controls, or between each other (Figure 2C,D).

**Figure 2.** qRT-PCR of miRNA-10a and miRNA-214. Total plasma RNA was isolated from HHT1 and HHT2 patients and control subjects. Relative expression levels of miRNA-10a (**A**,**B**) and miRNA-214 (**C**,**D**), measured by qRT-PCR using miR-16 (**A**,**C**) or cel-miR-39-3p (**B**,**D**) as normalizers. Symbols outside the box plot represent extreme values (◦) and outliers (\*) with their corresponding sample numbers. The number of samples analyzed is indicated in parentheses. Experiments were performed in triplicates. (\* *p* < 0.05; ns, not significant).
