Clinical data of HHT patients and variant identification

Clinical diagnosis of HHT is determined based on the Curaçao criteria established by the HHT international committee29. These criteria include the presence of epistaxis (spontaneous and recurrent nose bleeds), telangiectasias (multiple, at characteristic sites such as lips, oral cavity, fingers and nose), visceral vascular lesions (gastrointestinal telangiectasias and/or arterio-venous malformations), and family history (a first-degree relative with HHT). The diagnosis of HHT is definite if three of the criteria are present, possible or suspected if two are present and unlikely if fewer than two are present.

As part of a molecular diagnosis routine conducted at the genetics department of the Pitié-Salpêtrière Hospital (Paris, France), 274 individuals with rare vascular diseases among which 53 are with suspected HHT have been screened for candidate pathogenic variants using a custom next-generation sequencing (NGS) targeted gene panel including HHT genes (ACVRL1, ENG, SMAD4, GDF2, RASA1 and EPHB4) and additional genes related to other hereditary vascular diseases (Supplementary Table 3). Sequencing was performed on genomic DNA extracted from whole blood. VCF files from sequenced individuals were scrutinized using the MORFEE bioinformatics tool30 in order to detect and annotate non-coding SNVs creating uAUGs (uAUG-SNVs) in the 5’UTR of the sequenced genes.

Ethics declaration

The patients provided written informed consent for their DNA material to be used for genetic analysis in the context of molecular diagnosis in accordance with the French bioethics’ laws (Commission Nationale de l’Informatique et des Libertés no 1774128). The aforementioned committee approved the study.

Nomenclature

DNA sequence variant nomenclature follows current recommendations of the HGVS31.

Plasmid constructs and expression in human cells

In order to evaluate the potential effect of ENG variants on the protein steady state levels, we performed the in vitro functional assay described by Labrouche et al.32. To do so, we started by the amplification of the long isoform of ENG (L-ENG; NM_001114753.3) cDNA from HeLa cells by using specific primers covering the entire 5’UTR and the CDS lacking its stop codon (ENG c.-303_c.1974), and cloned it in the pcDNA3.1/myc-His(-) plasmid (Invitrogen) in frame with a Myc-His tag to obtain the wild-type (WT) clone (Supplementary Table 4). The PCR reaction was performed using Phusion High Fidelity (HF) DNA polymerase (ThermoFisher) and the cloning was carried out after double digestion of the inserts and plasmid with BamHI and HindIII. Mutated clones carrying ENG variants identified in this study or described in the literature (Fig. 1a and Table 1)2,21,22,23 were prepared by directed mutagenesis on the WT generated clone, pcDNA3.1-L-ENG-WT (Fig. 1b), using specific back to back primers (Supplementary Table 4) and the Phusion HF DNA polymerase, followed by DpnI digestion of the template, phosphorylation of the generated PCR product, ligation and transformation of competent cells to obtain unique clones. In addition, we generated a supplemental construct in which we deleted the start codon of ENG (pcDNA3.1-L-ENG-c.1_2del, Supplementary Table 4) and introduced separately ENG variants (c.-142A>T, c.-127C>T, c.-79C>T, c.-68G>A and c.-10C>T) in the latter construct. In presence of ENG c.1_2del, the created uAUGs become in frame with the Myc tag, allowing the detection of potential proteins translated from the uAUGs. All the recombinant plasmids were verified by Sanger sequencing (Supplementary Table 4) (Genewiz).

In order to evaluate the effect of the identified variants on protein levels, HeLa cells (ATCC as original source) were transfected with the prepared pcDNA3.1-L-ENG constructs. HeLa were prepared in 6-well plates 24 h before the transfection with 4.5 × 105 cells/well in RPMI medium (Gibco-Invitrogen) supplemented with 10% fetal calf serum (Gibco-Invitrogen). Transfections were performed in duplicate with 1 µg of each plasmid using JetPRIME® reagent (Polyplus Transfection) according to the manufacturer’s recommendations. Empty pcDNA3.1/myc-His(-) plasmid was used as negative control. The day of the transfection, cell confluence was at 60–80%. Cells were harvested and lysed 48 h after transfection to extract total RNA and protein. For this purpose, cells have been scraped and collected in 500 µl of PBS. Then, they were split in 2 aliquots: 100 µl aliquot for RNA extraction and 400 µl aliquot for protein extraction, as indicated below.

As endoglin is mainly known for its endothelial function in HHT, we also assessed the ENG steady-state levels in HUVECs (Lonza). For this purpose, we first used the generated pcDNA3.1-L-ENG constructs to subclone the WT and mutant cDNA of ENG in pRRLsin-MND-MCS-WPRE plasmid, upstream of a MND promoter. Unlike for pcDNA3.1-L-ENG constructs, pRRLsin-MND-MCS-WPRE-L-ENG ones do not contain any tag and the CDS of ENG ends at its own stop codon (ENG c.-303_c.1977). Then, lentiviruses were produced at the platform of vectorology (Vect’UB) of the University of Bordeaux (https://www.tbmcore.u-bordeaux.fr/vectorologie/) for the WT and ENG c.-142A>T, c.-127C>T, c.-79C>T and c.-68G>A variants. Twenty-four hours before transduction, HUVECs were prepared in 6-well plates with 2.8 × 105 cells/well in EGM-2 medium (Endothelial Cell Growth Medium-2, Lonza). Cells were transduced in duplicate with 20 MOI (Multiplicity of Infection) of the generated lentiviruses by using 3.2 mg of protamine sulfate. Transduced cells were harvested with trypsin 72 h post-transduction and each well was transferred to a P100 plate. For each construct, one plate has been used for whole protein extraction, 48 h after the transfer to P100 plates and the duplicate plate was used to freeze transduced cells.

Protein preparation and western blot analysis

Whole protein extractions were performed with RIPA supplemented with protease inhibitors for a pellet collected from one well of HeLa cells or one P100 plate of HUVECs. Concentrations were measured by using the BCA protein assay kit (Pierce™) following the manufacturer’s instructions. Proteins were loaded on 10% SDS-PAGE gels in parallel with a protein prestained ladder (Euromedex) and transferred onto PVDF membranes (Bio-Rad) by using the trans-blot turbo transfer system (Bio-Rad). Membranes were incubated with monoclonal anti-(c-Myc Tag) antibody (Merck Millipore, #05–419) (HeLa extracts), anti-ENG (Abcam, #ab169545) (HUVECs extracts) to probe endoglin, and anti-β-actin (Cell Signaling, #4970) (HeLa and HUVECs extracts) as a loading control. These antibodies were used at 1/3000, 1/2000 and 1/3000 dilutions, respectively. Fluorescent goat anti-mouse IgG Alexa Fluor 700 (ThermoFisher, #A-21036, 1/5000) was used against anti-(c-Myc Tag) and goat anti-Rabbit IgG (H + L) Alexa Fluor 750 (Invitrogen, #A-21039, 1/5000) was used against anti-ENG and anti-β-actin. Odyssey Infrared Imaging System (Li-Cor Biosciences) in 700 and 750 channels was used to scan, reveal, and quantify the blots. For quantification, the average of each duplicate was computed from the quantified values and ENG levels for each sample were normalized to the corresponding β-actin levels then to the WT or the negative control levels (%). The two bands obtained for the Endoglin, corresponding to the more glycosylated (upper band) and less/non glycosylated (lower band) ENG monomers16, were taken together for the quantification. All blots were processed in parallel and derive from the same experiments.

RNA isolation and RT-qPCR analysis

In order to evaluate ENG transcript levels in transfected HeLa cells, total RNA was isolated from the collected pellets by using the RNeasy mini kit (Qiagen) following the manufacturer’s instructions. Extracted RNA was quantified and equal quantities were used for reverse transcription reaction, performed with the M-MLV reverse transcriptase (Promega). Then 20 ng of cDNA of each sample was used for the qPCR reaction (duplicate/sample) with ENG- or α-tubulin-specific primers (Supplementary Table 4) and the GoTaq qPCR Master mix (Promega), in the presence of CXR reference dye, in a final volume of 10 µl and 40 cycles of amplification on QuantStudio3 Real-Time PCR System (ThermoFisher). QuantStudio design and analysis software was used to analyze the results and transcript levels were normalized to the reference α-tubulin gene. 2−ΔΔCT method was used to calculate the relative amounts of ENG to α-Tubulin in different samples. ΔΔCT were calculated by taking into account the mean of qPCR duplicates followed by the mean of transfection duplicates for each sample. Reaction efficiency (90–110%) and melting curves were evaluated for each couple of primers.

Luciferase assay

A complementary in vitro assay was deployed to evaluate the effect of the identified variants on the promoter activity. For this purpose, the ENG promoter, containing the basal promoter and the region carrying major transcriptional regulatory elements, as defined in Rıus et al.24, was amplified by using specific primers (Supplementary Table 4). This promoter sequence corresponds to the 805 nucleotides located upstream of the main ATG. The amplified promoter has been cloned in pGL3-basic vector containing the CDS of Firefly luciferase to obtain the WT clone (Fig. 1e). Then, ENG variants c.-142C>T, c.-79C>T and c.-68G>A were introduced in parallel by directed mutagenesis, as described above for pcDNA3.1-L-ENG vectors. All the recombinant plasmids were verified by Sanger sequencing by using specific primers (Supplementary Table 4) (Genewiz). WT or mutant clones were co-transfected with a plasmid containing Renilla luciferase in triplicate in 96-well plate of HeLa cells. Forty-eight hours after the transfection, luciferase activity was measured by using the dual-glo luciferase assay system (Promega) directly in transfected wells by detecting luminescence with both Firefly and Renilla luciferases. Mean of the triplicates of Firefly/Renilla ratios of each sample was normalized to the WT.

Functional effect of ENG variants on BRE activity in vitro

The BRE assay described in Mallet et al.16 has been modified to assess the functionality of the 5’UTR ENG variants. Briefly, NIH-3T3 were seeded in 96-wells white plates (15,000 cells/well) in DMEM containing 1% fetal calf serum and transfected the following day by a mixture of plasmid (i) BRE luciferase reporter plasmid (75 ng) (ii) pRL-TK luc encoding Renilla luciferase (20 ng), pcDNA3-ALK1 (0.15 ng) and pcDNA3.1-L-ENG WT or 5’UTR-mutated constructs (0–10 ng/well). Four hours after transfection, cells were stimulated overnight with 5 pg/ml of BMP9 in serum-free medium (R&D Systems) and luciferase activity was measured with the twinlite Firefly and Renilla Luciferase Reporter Gene Assay System (PerkinElmer). Means of triplicate were calculated for each sample and Firefly/Renilla ratio of stimulated wells was normalized to that obtained in non-stimulated wells.

Statistical data analysis

Differential protein and RNA levels, luciferase activity, and BRE activity were assessed using two-factor analysis of variance followed by Tukey’s multiple comparison test. A threshold of p < 0.05 was used to declare statistical significance.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.