Background Mutations in the PLCE1 gene encoding phospholipase C epsilon 1 (PLCɛ1) have been recently described in patients with early onset nephrotic syndrome (NS) and diffuse mesangial sclerosis (DMS). In addition, two cases of PLCE1 mutations associated with focal segmental glomerulosclerosis (FSGS) and later NS onset have been reported.
Method In order to better assess the spectrum of phenotypes associated with PLCE1 mutations, mutational analysis was performed in a worldwide cohort of 139 patients (95 familial cases belonging to 68 families and 44 sporadic cases) with steroid resistant NS presenting at a median age of 23.0 months (range 0–373).
Results Homozygous or compound heterozygous mutations were identified in 33% (8/24) of DMS cases. PLCE1 mutations were found in 8% (6/78) of FSGS cases without NPHS2 mutations. Nine were novel mutations. No clear genotype–phenotype correlation was observed, with either truncating or missense mutations detected in both DMS and FSGS, and leading to a similar renal evolution. Surprisingly, three unaffected and unrelated individuals were also found to carry the homozygous mutations identified in their respective families.
Conclusion PLCE1 is a major gene of DMS and is mutated in a non-negligible proportion of FSGS cases without NPHS2 mutations. Although additional variants in 19 candidate genes (16 other PLC genes, BRAF,IQGAP1 and NPHS1) were not identified, it is speculated that other modifier genes or environmental factors may play a role in the renal phenotype variability observed in individuals bearing PLCE1 mutations. This observation needs to be considered in the genetic counselling offered to patients.
- nephrotic syndrome
- diffuse mesangial sclerosis
- focal segmental glomerulosclerosis
- hereditary glomerular disease
- molecular genetics
- renal medicine
Statistics from Altmetric.com
- nephrotic syndrome
- diffuse mesangial sclerosis
- focal segmental glomerulosclerosis
- hereditary glomerular disease
- molecular genetics
- renal medicine
Nephrotic syndrome (NS) results from disruption of the renal filtration barrier composed of interdigitating podocyte foot processes linked by the slit diaphragm, and endothelial cells separated from podocytes by the glomerular basement membrane. It is characterised by massive proteinuria, hypoalbuminaemia, hyperlipidaemia, oedema, and podocyte foot process effacement by electron microscopy. NS is classified as steroid sensitive (SSNS) or steroid resistant (SRNS) and mutations in several genes, mostly coding for podocyte proteins, have been identified as causing SRNS in humans. Although kidney biopsy discloses focal and segmental glomerulosclerosis (FSGS) lesions for most patients with SRNS,1 a smaller proportion of children may present a particularly severe renal histological pattern named diffuse mesangial sclerosis (DMS) with an early presentation and a poor renal prognosis. DMS is characterised by mesangial expansion and sclerosis that evolves towards obliteration of the capillary lumen and contraction of the glomerular tuft. This type of renal histology has been described as part of syndromes such as Denys–Drash syndrome or Pierson syndrome, caused by mutations in the WT1 and LAMB2 genes, respectively. More recently, PLCE1 mutations have been found as a novel cause of DMS.2 Involvement of the PLCE1 gene (MIM *608414) in NS has been found using a combination of homozygosity mapping and cDNA microarrays from rat glomeruli. In the original series described by Hinkes et al,2 truncating PLCE1 mutations were identified in 12 siblings from six families diagnosed with early onset NS and DMS, while a homozygous missense mutation was identified in two siblings with a later onset of disease and FSGS (nephrotic syndrome, type 3; NPHS3—MIM #610725); until now, these individuals were the only reported cases of FSGS secondary to PLCE1 mutations. Subsequently, mutations in PLCE1 have been demonstrated as a major cause of isolated DMS, identified in 28.6% of 35 families in a worldwide cohort.3 Although PLCE1 mutations typically lead to a severe phenotype, full and sustained treatment responses have been reported in two individuals with truncating mutations2 and a pathogenic homozygous mutation has also been found in a phenotypically completely normal adult.4 This suggests that other factors may modify the effect of PLCE1 mutations.
PLCE1 gene encodes the phospholipase C epsilon 1 (PLCɛ1). Members of the phosphoinositide specific phospholipase C (PLC) family catalyse the hydrolysis of membrane phospholipids to generate the second messenger molecules inositol 1,4,5-trisphosphate (IP3) and diacylglycerol (DAG) that initiate intracellular pathways of cell growth and differentiation.5 Several distinct PLC enzymes have been identified in a variety of mammalian tissues and are divided into four classes: PLCβ, PLCγ, PLCδ and PLCɛ. Phospholipase catalytic domains (PLC_X and PLC_Y) and Ca2+ lipid binding domain (C2 domain) typically characterised all PLCs. PLCɛ1, which is highly expressed in podocytes, 2 is the most recently identified member of the PLC family; it contains, in addition to conserved PLC domains, a RasGEF_CDC25 (guanine nucleotide exchange factor for Ras-like small GTPases domain) and two C-terminal Ras-binding (RA) domains, RA1 and RA2 (RasGTP binding domain from guanine nucleotide exchange factors). While PLCβ, PLCγ, PLCδ are differentially regulated by heterotrimeric G-proteins, tyrosine kinases and calcium, it has been shown that PLCɛ1 is regulated by H-Ras (via RA1 and RA2 domains) and also interacts with IQGAP1 (IQ motif-containing GTPase-activating protein) and BRAF.2 5–7
In order to better define the phenotype spectrum of patients bearing PLCE1 mutations, we performed mutational analysis and clinical follow-up of a large worldwide cohort of patients with either familial or sporadic SRNS. In addition, we aimed to determine the role of candidate genes, including other PLC genes, as modifiers of disease in families for which both symptomatic and asymptomatic members had homozygous PLCE1 mutations.
Patients and methods
A total of 139 patients were included in the present study, all of which presented SRNS. In accordance with criteria established by the International Study of Kidney Disease in Children (ISKDC), NS was defined as the association of proteinuria >40 mg/m2/h, hypoalbuminaemia <2.5 g/dl, oedema, and hyperlipidaemia.1 Steroid resistance was defined as lack of response to 4 weeks of treatment with 60 mg/m2 per day of prednisone.8 None of the included patients presented relapse after transplantation, if performed. Ninety-five patients, belonging to 68 families (53 consanguineous families, 15 non-consanguineous multiplex families), were classified as familial cases, defined as the presence of one (or more) affected individual(s) in consanguineous families and two or more affected children in non-consanguineous families. Forty-four patients presented with sporadic SRNS. Mutations in NPHS2, encoding podocin, were excluded for all patients. Mutations of the exons 8 and 9 of the WT1 gene, encoding the Wilm's tumour 1 protein, were excluded in all phenotypically female patients. Mutations in NPHS1, encoding nephrin, were excluded in children which presented NS at less than 1 year of age without DMS on kidney histology. Clinical data, including ethnic origin, type of renal histological lesions and age at onset of NS and end stage kidney disease (ESKD), were recorded. Informed consent was obtained for all participating families and the study was approved by the Comité de Protection des Personnes «Ile de France II».
PLCE1 molecular analyses
Genomic DNA specimens were collected with informed consent and isolated from peripheral blood using standard procedures. Linkage of familial cases to the PLCE1 locus was assessed using four polymorphic microsatellite markers spanning 5.5 Mb and flanking the locus (D10S185, D10S1680, D10S574, D10S1726). When results were compatible with linkage to PLCE1 locus, and for all sporadic cases, the complete coding sequence and exon–intron boundaries of the PLCE1 gene were amplified by polymerase chain reaction (PCR) (33 exons, GenBank accession number NM_016341. 3); subsequently, both strands were sequenced using a Big Dye terminator cycle sequencing kit and analysed with an ABI Prism 3130xl DNA analyser (Applied Biosystems, Foster City, California, USA). PCR and sequencing primers are available upon request. Sequence chromatograms were analysed using the Sequencher software (Gene Codes Corporation, Ann Arbor, Michigan, USA). Segregation analysis of mutations in families was performed by direct DNA sequencing. Prediction of the functional impact of missense PLCE1 variants was obtained with the PolyPhen software (http://genetics.bwh.harvard.edu/pph/). The absence of these mutations among at least 50 normal individuals was confirmed. Mutational data were described using the nomenclature of the Human Genome Variation Society (www.hgvs.org/mutnomen). Positions of mutations were numbered with the A of the ATG-translation initiation codon in the reference cDNA sequence being 1.
Mutational analyses of possible modifier genes, including other phospholipases C
In order to identify modifier genes in families with asymptomatic individuals who carried the same PLCE1 homozygous mutation as the affected patients, a total of 19 candidate genes were evaluated, including 16 genes encoding other PLC (PLCB1, PLCB2, PLCB3, PLCB4, PLCD1, PLCD3, PLCD4, PLCG1, PLCG2, PLCH1, PLCH2, PLCL1, PLCL2, PLCXD1, PLCXD3 and PLCZ1) and three genes encoding proteins that interact directly or indirectly with PLCɛ1 (BRAF, IQGAP1, NPHS1) (table 1). First, a whole genome wide search for linkage was performed using a 250k SNP array (GeneChip Human Mapping 250K Nsp Array from Affymetrix, Santa Clara, California, USA). The potential role of one of the genes cited above as a modifier of disease was excluded if the asymptomatic mutated patient shared the same haplotypes as at least one of the affected mutated patients. Then, for genes not excluded by this method, microsatellite markers were used to test each locus of the remaining genes: PLCB2 (D15S994, D15S968, D15S1006), PLCB4 (D20S889, D20S115, D20S186), PLCH1 (D3S1594, D3S1279, D3S1268), PLCL1 (D2S117, D2S311, D2S374, D2S318, D2S115), PLCL2 (D3S3701, D3S2338, D3S1293) and BRAF (D7S2560, D7S684, D7S2513, D7S661). Finally, the complete coding sequence and exon–intron boundaries of the genes not excluded by the whole genome wide scan or by microsatellite marker analysis were sequenced as described above (PLCH1, 23 exons, GenBank accession number NM_014996. 2; PLCL1, five exons, GenBank accession number NM_006226. 3; BRAF, 18 exons, GenBank accession number NM_004333. 4) in one asymptomatic mutated patient and one affected mutated patient of each of the families.
All values are expressed as mean±SE or median (range). Comparisons between two continuous variables were performed using the Mann–Whitney test. All tests were two sided, and values of p<0.05 were considered significant. Statistical analyses were performed using GraphPad Prism software version 5.01.
The characteristics of the patients are summarised in table 2. Most of the familial cases originated from the Middle East (33.8%), while most of the sporadic cases were from Europe (66.7%) (mainly from France). Among familial cases, renal histology disclosed FSGS, DMS and other renal lesions (mesangial proliferation, membranoproliferative glomerulonephritis, and terminal kidney) in 50, 10, and four families, respectively (renal biopsy not performed in four families). Among sporadic cases, FSGS was present in 28 cases, DMS in 14 cases and mesangial proliferation in two cases. Mean age at onset of NS was 46.2±5.7 months. Seventy-six patients reached ESKD at a mean age of 79.7±7.1 months.
Identification of PLCE1 mutations
Microsatellite marker analysis excluded linkage to the PLCE1 locus in 43 families. Therefore, results were compatible with linkage to the PLCE1 locus in 25 families as evidenced by homozygosity for the polymorphic markers flanking the PLCE1 locus. Disease-causing mutations were found in 12 families, comprising a total of 18 patients (table 3). Among sporadic cases, homozygous PLCE1 mutations were identified in three patients. Most were homozygous truncating mutations (nonsense or frameshift mutations), although compound heterozygote missense mutations (p.E1386V/p.P1890L) were detected in two siblings and a homozygous missense mutation (p.H1407D) was detected in another familial case. The p.E1386V was predicted to be a possibly pathogenic mutation (Polyphen score 1.94) and involved a very highly conserved amino acid from fishes to humans. The p.P1890L and p.H1407D were predicted to be probably pathogenic mutations with a Polyphen score of 2.55 and 2.50, respectively, involving amino acids conserved from C-elegans to humans, and included in either the C2 domain or the PLC_X catalytic domain of the protein. Nine of these mutations have not been previously reported (figure 1).
Phenotypic spectrum of PLCE1 mutations
Among the 12 families bearing PLCE1 mutations, renal biopsy disclosed FSGS in six and DMS in five (not performed in one family). This translates into a 12% (6/50) and 50% (5/10) PLCE1 detection rate among families with FSGS and DMS, respectively. All the three sporadic cases presented DMS on renal histology, meaning that PLCE1 mutations were detected in 21% (3/14) of sporadic DMS cases, but in none (0/28) of the sporadic FSGS cases. Globally, PLCE1 mutations were identified in 13% of our whole cohort (15/112), 33% (8/24) of DMS cases, and 8% (6/78) of FSGS cases. When kidney histology of several affected patients of the same family was available, all siblings had the same type of renal lesions.
Patients with PLCE1 mutations presented NS at a mean age of 22.8±5.0 months leading to ESKD at a mean age of 39.1±6.3 months (two patients with FSGS did not reach ESKD after a short follow-up of 1 year and one patient without renal biopsy died at 11 months of NS complications). This contrasts with non-mutated patients which presented NS at a mean age of 50.9±6.7 months (p=0.12) leading to ESKD at a mean age of 92.6±8.4 months (p<0.01). In patients bearing PLCE1 mutations, no clear-cut genotype–phenotype correlations were observed. Truncating (mean age at onset 24.6±5.8 months, mean age at ESKD 37.1±6.7 months) and missense mutations (mean age at onset 12.3±1.5 months, mean age at ESKD 49.0±19.9 months) led to similar renal presentation and evolution (p=0.80 and p=0.41, respectively) and were detected in both FSGS and DMS histological forms. However, renal prognosis was worst in patients with DMS compared to those with FSGS (mean age at onset of NS 15.3±6.0 months vs 30.5±7.7 months, p=0.15, and mean age at ESKD 20.9±5.8 months vs 57.2±7.3 months, p<0.01).
Identification of asymptomatic individuals bearing homozygous PLCE1 mutations
Surprisingly, segregation analysis revealed that three unaffected individuals belonging to three unrelated families were also found to bear, in the homozygous state, the mutations identified in their respective affected relatives (figure 2). In family E, three out of six siblings of a consanguineous family from Pakistan presented with SRNS and FSGS on renal histology at ∼5 years of age and reached ESKD by the age of 6. Another brother had a past history of intermittent mild proteinuria but was free of renal manifestations at 12 years of age. The p.F839fs889X mutation was identified in these four patients. In family O, the index case was born from consanguineous parents of Turkish origin. She was diagnosed with SRNS and DMS at the age of 11 months and found to bear the p.R2150X mutation. Among her two siblings, one brother was homozygous for the mutation and had no proteinuria at 10 years of age. In family A, two out of three children of consanguineous parents from Pakistan developed SRNS at 5 months and 3 years of age with FSGS on renal biopsy, and reached ESKD when they were 1.4 and 3 years old, respectively. The p.R321X mutation was identified. Segregation analysis revealed that the father, who was also born from consanguineous parents, was homozygous for the mutation. Although he had presented hypertension and mild proteinuria as a teenager, his renal evaluation, at 46 years of age, was normal.
Search for modifier genes
To understand this phenomenon, we ought to identify mutations or variants in other genes that could modulate the clinical phenotype. We hypothesised a protective effect arising from other members of the phospholipase C family in the asymptomatic individuals. Alternatively, another hypothesis was that additional mutations or variants in BRAF, IQGAP1 or NPHS1 genes in symptomatic patients could explain the phenotypic difference. Analysis of whole genome DNA arrays in family E excluded the role of 13 of the 19 candidate genes as a modifier of the disease, since the asymptomatic mutated individual shared the same haplotypes as at least one of his affected siblings at these loci. Then, microsatellite marker analysis of the remaining genes (PLCB2, PLCB4, PLCH1, PLCL1, PLCL2 and BRAF) in family O showed that both the affected and the clinically unaffected case shared the same haplotypes for PLCB2, PLCB4 and PLCL2 genes, excluding them as being potentially involved in the phenotypic variability in the family. Therefore, the direct DNA sequencing of the three remaining genes was performed in one affected case and one asymptomatic case of family E, O and A, but did not allow identification of mutations or variants.
In 2006, positional cloning uncovered mutations in PLCE1 causing early onset NS. Since then, 17 different mutations have been reported (figure 1), among which only one was associated with FSGS. We therefore aimed to better assess the phenotype spectrum of patients bearing PLCE1 mutations in a large worldwide cohort of patients with SRNS.
Among familial forms of SRNS, PLCE1 mutations were found in 12% of families with FSGS and half of the families with DMS. Although a significant proportion (21%) of the DMS sporadic cases had PLCE1 mutations, none of the 28 sporadic cases with FSGS had mutations in this gene. If we had included all children with SRNS without any prior genetic testing, the detection rate of PLCE1 mutations in FSGS cases would have been lower, as NPHS2 mutations would have been identified for a significant proportion of them. In contrast, the high rate of PLCE1 mutations found in DMS cases in our study is probably not overestimated and is concordant with what has been previously published.3 Indeed, in our cohort, none of the patients presented signs of Denys–Drash or Pierson syndrome, which could have suggested a mutation in other DMS associated genes such as WT1 or LAMB2. In addition, WT1 mutations were excluded in all phenotypically female patients and NPHS2 mutations have never been found as a cause of DMS.17 Similar to previous studies,2–4 16 mutated individuals presented with relatively early NS (from 3 months to 6 years of age) leading to ESKD before the age of 7 years. In addition, Gbadegesin et al did not identify PLCE1 mutation in a cohort of 231 affected individuals with FSGS and later onset of NS (median 26 years, range 1–66) originated from 69 unrelated families (10 were sporadic cases).18 This confirms that PLCE1 is a major gene of DMS and that all subjects with this type of histological lesion should be screened for mutations in this gene. On the other hand, PLCE1 mutations seem to account for a small, although not negligible, proportion of FSGS cases for which NPHS2 mutations have been ruled out. Considering that the mean age of FSGS patients with PLCE1 mutations in our cohort was ∼2.5 years, PLCE1 genetic testing should probably be performed only in those with an early presentation, particularly if they are considered as familial cases.
Among the initial cohort reported by Hinkes et al,2 all patients with homozygous truncating PLCE1 mutations had DMS as opposed to two patients with missense mutations who exhibited FSGS with a later onset of NS. In a follow-up study, this group reported 10 truncating but no missense PLCE1 mutations in a cohort of 35 families with DMS.3 In contrast, we observed either truncating or missense mutations in both FSGS and DMS patients and no clear phenotype–genotype correlation. For example, the p.R1246X mutation was found in the homozygous state in patients with FSGS (families H and I) or DMS (patients J and K), with age at NS onset varying between 6 months and 5 years. In addition, in the present study, missense mutations did not lead to a milder course of disease; however, missense mutations were identified in only a few cases, thereby precluding a definite conclusion. Nevertheless, our data showed the clinical heterogeneity associated with mutations in the PLCE1 gene.
In this regard, we identified three unaffected individuals carrying the same homozygous mutation as their affected relatives, a finding particularly unusual for disease of recessive inheritance. One other previous case report of a family of three affected siblings with DMS also showed that the asymptomatic father, who was himself born from consanguineous parents, had the same homozygous PLCE1 mutation as his affected children.4 These data are also reminiscent of the unusual progression of two patients with DMS in the original description of PLCE1 mutations who seemed to respond to steroid or ciclosporin therapy.2 The above observations raise the possibility of oligogenic inheritance, with either a third deleterious mutation in a second gene explaining the appearance of a renal phenotype in patients bearing two PLCE1 mutations or the presence of a protective modifier allele, which could compensate for the PLCE1 dysfunction in asymptomatic individuals. Although oligogenic inheritance has been suggested in several cases of renal diseases,19–21 it is only in very rare cases that asymptomatic individuals presented two pathogenic mutations in one gene in contrast to affected individuals bearing an additional third mutation in a second gene.22 Indeed, oligenic inheritance has been clearly demonstrated in some cases of Bardet-Biedl syndrome (BBS), a genetically heterogeneous disorder characterised by pigmentary retinal dystrophy, polydactyly, obesity, developmental delay and renal defects. In a study of 163 BBS families, Katsanis et al found the presence of ‘three mutant alleles’ (either two BBS2 mutations and one BBS6 mutation or the opposite) in four pedigrees. More interestingly, they detected two BBS2 mutations but not BBS6 mutations in unaffected individuals in two pedigrees, suggesting that BBS may not be a single gene recessive disease but a complex trait requiring a total of three mutations in two different genes to manifest the phenotype. On the other hand, existence of a protective modifier gene has been demonstrated in a recent study on autosomal recessive spinal muscular atrophy (SMA).23 The detection of asymptomatic individuals (all females) carrying the same mutations in the SMN1 (survival motor neuron 1) gene as their affected siblings has suggested the influence of modifier genes. In this study, higher expression of plastin 3, encoded by PLS3, were found in unaffected SMN1-deleted females compared to their SMA-affected counterparts. Overexpression of PLS3 rescued the axon length and outgrowth defect associated with downregulation in motor neurons of SMA mouse embryos and in zebra fish, thereby explaining the absence of a neurological phenotype in some patients bearing SMN1 mutation.
Effect of modifiers genes could also explain the absence of obvious renal phenotype in the Plce1 knockout mouse model, developed on a mixed C57/B6 X 129/S6 genetic background.2 24 Indeed, contrary to the zebra fish knockout model in which embryos injected with Plce1 morpholino show oedema at D4 of development and foot process effacement by electron microscopy,2 Plce1−/− mice do not exhibit any renal phenotype. However, as C57/B6 mice are resistant to the development of glomerulosclerosis,25 it might be of interest to assess the presence of a renal phenotype in Plce1 knockout mice from a different genetic background or under nephrosis promoting conditions.
Based on the preceding observations, we ought to identify mutations or variants in other genes that could explain the absence of renal phenotype in some patients with PLCE1 mutation. We first speculated a compensatory protective effect arising from other members of the PLC family in the unaffected individuals since the common role of these proteins is to initiate intracellular signalling through the generation of the second messenger molecules IP3 and diacylglycerol. Interestingly, some mouse models of PLC genes develop a renal phenotype, including increased kidney weight, polycystic kidneys, renal dysplasia, glomerulosclerosis or signs of glomerulonephritis (table 1). Moreover, PLC-γ1 has been shown to play a key role in the podocyte slit diaphragm. PLC-γ1 binds to phosphorylated nephrin, the main component of the slit, leading to PLC-γ1 activation and subsequent increase in IP3, and triggering of Ca2+ intracellular response.26 In our study, we did not identify a PLC candidate that was mutated in a recessive or dominant fashion in the affected individuals and not in the asymptomatic relative, or the reverse. We then used the same approach for candidate genes that are known to interact directly or indirectly with PLCɛ1 such as IQGAP1, BRAF and NPHS1. IQGAP1 is a known scaffold protein of the mitogen activated protein (MAP) kinase pathway27 and interacts with PLCɛ1,2 nephrin28 and BRAF.29 BRAF, which has been involved in various cancers,30 belongs to the RAF family of genes and binds H-Ras.31 It has also been implicated in the MAP kinase pathway and is considered as a third interaction partner of PLCɛ1,7 in addition to H-ras and IQGAP-1. In the present study, IQGAP1, BRAF and NPHS1 genes did not appear to explain the phenotypic variability observed in our patients. However, our approach cannot exclude an effect of different PLC or PLC partner genes in the three different families. Finally, it has to be stressed that, to date, only very few homozygous PLCE1 mutations detected in asymptomatic individuals or patients that responded to therapy have been reported and, surprisingly, all of them are males. This is reminiscent of the unaffected SMN1-deleted individuals, which were all females,23 and suggests the existence of gender specific modifiers of disease.
In conclusion, PLCE1 mutations are frequently found in familial and sporadic DMS cases. Although PLCE1 mutations had been rarely associated with FSGS, we showed that mutations in this gene are detected in a non-negligible proportion of familial FSGS cases without NPHS2 mutations. In this study, no clear genotype–phenotype correlation was present. Although variants in other phospholipase C genes do not seem to explain the absence of renal manifestations in some cases with PLCE1 mutations, we speculate that other modifier genes or environmental factors may play a role in the renal phenotype variability observed in individuals bearing PLCE1 mutations. More importantly, the clinical heterogeneity associated with PLCE1 mutations substantially increases the complexity of genetic counselling, including prenatal diagnosis, and needs to be considered in the discussion about renal prognosis with patients.
Financial support for this work was provided by grants from the Programme Hospitalier de Recherche Clinique (PHRC) AOM02123 (to C.A), the Association pour l'utilisation du Rein Artificiel (AURA) (to C.A), PodoNet (Clinical, Genetic and Experimental Research into Hereditary Diseases of the Podocyte) - project of the 2007 E-RARE program (to CA) and Fonds de la Recherche en Santé Québec (FRSQ) (Fellowship training Award to G.B.).
OB and GB contributed equally to this paper.
Competing interests None.
Patient consent Obtained.
Ethics approval This study was approved by the Comité de Protection des Personnes «Ile de France II».
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.