Pathogenic variants in HGF give rise to childhood-to-late onset primary lymphoedema by loss of function

Abstract

Developmental and functional defects in the lymphatic system are responsible for primary lymphoedema (PL). PL is a chronic debilitating disease caused by increased accumulation of interstitial fluid, predisposing to inflammation, infections and fibrosis. There is no cure, only symptomatic treatment is available. Thirty-two genes or loci have been linked to PL, and another 22 are suggested, including Hepatocyte Growth Factor (HGF). We searched for HGF variants in 770 index patients from the Brussels PL cohort. We identified ten variants predicted to cause HGF loss-of-function (six nonsense, two frameshifts, and two splice-site changes; 1.3% of our cohort), and 14 missense variants predicted to be pathogenic in 17 families (2.21%). We studied co-segregation within families, mRNA stability for non-sense variants, and in vitro functional effects of the missense variants. Analyses of the mRNA of patient cells revealed degradation of the nonsense mutant allele. Reduced protein secretion was detected for nine of the 14 missense variants expressed in COS-7 cells. Stimulation of lymphatic endothelial cells with these 14 HGF variant proteins resulted in decreased activation of the downstream targets AKT and ERK1/2 for three of them. Clinically, HGF-associated PL was diverse, but predominantly bilateral in the lower limbs with onset varying from early childhood to adulthood. Finally, aggregation study in a second independent cohort underscored that rare likely pathogenic variants in HGF explain about 2% of PL. Therefore, HGF signalling seems crucial for lymphatic development and/or maintenance in human beings and HGF should be included in diagnostic genetic screens for PL.

Keywords: hepatocyte growth factor; lymphatic system; primary lymphoedema; whole-exome sequencing.

Figure 1

Schematic representation of HGF with position of variants detected in lymphoedema patients. HGF includes an N-terminal PAN/apple domain, four kringle domains and a serine protease-like domain. Domain positions given in amino acid numbers. Two shorter isoforms (NK1 & NK2) shown. The variants, organized by type, have an allele frequency < 0.0014 in GnomAD and have not been detected in more than three distinct pathologies in our internal database. Variants from this study, in bold; published variants [5, 6], in italics. One variant in our Brussels PL cohort and also reported, bold italics. SP, signal peptide. Nucleotide positions according to NG_016274.2 (NM_000601.6) and amino acids to Uniprot P14210. Variants between [ ] are allelic.

Figure 2

RNAs with premature HGF stop codons are degraded by nonsense-mediated mRNA decay. Electropherograms with genomic (g)DNA and lymphocytic/lymphoblastic cDNA sequences around the variants in heterozygous patients and unaffected control individuals (wild-type). Mutated nucleotide shown between vertical red lines and with “Y”. Frameshift-induced double sequence also marked with “Y”. Heterozygous mutant alleles are well detected in gDNA, hardly visible on cDNA. gDNA not available for sanger sequencing for LE-1181-10.

Figure 3

Effect of variants on HGF expression and secretion in COS-7 cells. Cells were transiently transfected with an empty vector or plasmids for wild-type and variant human HGF. NT, non-transfected. Standard deviations indicated with bars. (A) Representative western blot of cell lysates revealed expression of pro-HGF (~100 kDa) and β-actin (~42 kDa) as loading control. (B) Quantification of pro-HGF in western blots of cell lysates (n = 3). Fold-change concentrations relative to wild-type. (C) ELISA quantification of HGF in the supernatants (n = 3). Fold-change concentrations relative to wild-type. Significant differences with wild-type (evaluated using a two-tailed Wilcoxon test) indicated with ^* for P-value ≤ 0.05; ^** for P-value ≤ 0.01. Two gels were used to load the samples, separated by a dashed line.

Figure 4

Effect of HGF mutants on phosphorylation of intracellular signalling molecules in LECs. Stimulation of LECs with same amount of recombinant wild-type HGF (rec-HGF) or HGF-containing conditioned medium from COS-7 cells expressing an empty vector, wild-type HGF, or one of four selected mutants. NT, non-treated. (A) Representative western blots of stimulated LEC cell lysates with levels of p-ERK1/2 (p44/p42), total ERK1/2, p-AKT (Ser473), total AKT, p-GSK3-α/β (Ser9), total GSK3-β, p-WNK1 (Thr60), total-WNK1 and β-actin. (B-E) quantification of the corresponding western blots (n = 3) relative to wild-type. Significant differences with wild-type (evaluated using a two-tailed Wilcoxon test) indicated with ^* for P-value ≤ 0.05; ^** for P-value ≤ 0.01. Wild-type indicated with black bar; decreased level of phosphorylation is indicated in light grey bars. Gels were cropped to show only 4 missense variants. See the entire series and whole blots of variants in Supplementary Fig. 4.

Figure 5

Altered lymphatic drainage in lower limbs of HGF patients compared to a VEGFR3-associated Milroy disease. Images 4 h after injection. Control: Drainage without delay in both legs, number of lymph nodes in pelvis normal. HGF stop codon Cys96^* (LE-291-10): No drainage on right lower limb, altered drainage on left, with insufficient number of pelvic lymph nodes stained. HGF stop codon Arg502^* (LE-535-10): No drainage on lower limbs and no lymph nodes, with distal dermal back flow. HGF missense variant Lys649Asn (LE-197-10): No drainage on lower limbs, subcutaneous diffusion of the radiotracer in left leg. FLT4 (VEGFR3) Asn1042Ile: Patient with PL showing no drainage and no visible pelvic lymph nodes.