Article Text

Submicroscopic genomic alterations in Silver–Russell syndrome and Silver–Russell-like patients
  1. Sara Bruce1,2,
  2. Katariina Hannula-Jouppi2,
  3. Mari Puoskari1,2,
  4. Ingegerd Fransson1,
  5. Kalle O J Simola3,
  6. Marita Lipsanen-Nyman4,
  7. Juha Kere1,2
  1. 1Department of Biosciences and Nutrition, Karolinska Institutet, Huddinge, Sweden
  2. 2Department of Medical Genetics, University of Helsinki, and Folkhälsan Institutet of Genetics, Helsinki, Finland
  3. 3Department of Pediatrics, Tampere University Hospital, Tampere, Finland
  4. 4Hospital for Children and Adolescents, University of Helsinki, Finland
  1. Correspondence to Dr Sara Bruce, Department of Biosciences and Nutrition, Karolinska Institutet, Hälsovägen 7-9, Huddinge, Sweden; sara.bruce{at}


Background Silver–Russell syndrome (SRS, OMIM 180860) features fetal and postnatal growth restriction and variable dysmorphisms. Genetic and epigenetic aberrations on chromosomes 7 and 11 are commonly found in SRS. However, a large fraction of SRS cases remain with unknown genetic aetiology.

Methods 22 patients with a diagnosis of SRS (10 with H19 hypomethylation and 12 of unknown molecular aetiology) and their parents were studied with the Affymetrix 250K Sty microarray. Several analytical approaches were used to identify genomic aberrations such as copy number changes (CNCs), loss of heterozygosity (LOH) and uniparental disomy (UPD). Selected CNCs were verified with quantitative real-time PCR.

Results The largest unambiguous CNCs were found in patients with previously molecularly unexplained SRS with relatively mild phenotypes: a heterozygous deletion of chromosome 15q26.3 including the IGF1R gene (2.6 Mb), an atypical distal 22q11.2 deletion (1.1 Mb), and a pseudoautosomal region duplication (2.7 Mb) in a male patient. LOH regions of potential relevance to the SRS phenotype were also identified. Importantly, no duplications or UPD of chromosomes 7 or 11 were identified.

Conclusion Unexpected submicroscopic genomic events with pathogenic potential were found in three patients with molecularly unexplained SRS that was mild. The findings emphasise that SRS is heterogeneous in genetic aetiology beyond the major groups of H19 hypomethylation and maternal UPD7 and that unbiased genome-scale screens may reveal novel genotype–phenotype correlations.

  • Silver-Russell syndrome
  • copy number
  • genomic screen
  • IGF1R deletion
  • atypical distal 22q11.2 deletion
  • endocrinology
  • genetic screening
  • molecular genetics

Statistics from

Request Permissions

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.


Silver–Russell syndrome (SRS, OMIM 180860) presents with fetal-onset growth restriction and variable dysmorphisms, including relative macrocephaly, skeletal asymmetry and leanness. Molecular studies have shown that 5–15% of patients with SRS have maternal uniparental disomy of chromosome 7 (matUPD7) and 20–65% show hypomethylation of the H19 imprinting control region,1 and these molecular subgroups represent emerging clinical entities of the syndrome.2–5 The remaining 25–75% of cases of SRS are either molecularly unexplained (idiopathic) or present with more rare genetic aberrations such as maternal duplication of 7p or 11p.1 Further, a range of genomic events have been described in association with SRS, including deletions of 8q11-q12 and 15q26.3 (IGF1R), X monosomy, duplications of 1q32-q42, 7q34-qter and X (in males), trisomy 18, translocations involving 17q25, and mosaic matUPD11.1 6 We and others have shown that UPD can be readily diagnosed when patients and parents are screened with genotyping arrays.7 8

Microarrays with genome-scale coverage of either tiling path probes or single-nucleotide polymorphisms (SNPs) are routinely used to study human diseases.9 The Affymetrix genotyping arrays can be used for both signal intensity determination and genotyping,10 enabling the parallel assessment of both copy number and additional events such as UPD. Copy number variants (CNVs) are often found in control populations, and their phenotypic contribution is largely unknown.11 A community resource called the database of genomic variants (DGV) has been created to summarise all reported CNVs, facilitate comparisons between studies, and assess uniqueness in clinical studies (

To search for genomic structural variation that might contribute to SRS, we explored the genomes of 22 patients with SRS and their parents: 10 with H19 hypomethylation (SRS-hypo) and 12 with SRS of unknown molecular aetiology (SRS-ue).

Materials and methods


We studied 22 patients (eight boys and 14 girls) with a diagnosis of SRS and their parents, 66 individuals in total. Ten patients had H19 hypomethylation (SRS-hypo) and 12 had unexplained SRS (SRS-ue), as previously reported.2 MatUPD7 had previously been excluded in all patients by microsatellite screening. Four patients with SRS-ue (7, 8, 9, 12) were diagnosed as having SRS by experienced clinical geneticists and a paediatric endocrinologist and referred to us for molecular evaluation without access to detailed clinical data. SRS was verified from patient records and photographs and/or by clinical evaluation when possible. The diagnosis of SRS was based on the following criteria: (1) small for gestational age (birth length and/or weight ≤2.0 SDS for gestational age); (2) postnatal growth retardation (height SDS below −2.5 at the age of 2 years); (3) relative macrocephaly (head circumference at least >1.5 SDS above the length SDS); (4) a typical SRS face with at least three of the following facial characteristics: a triangular face, micrognathia (leading to down-turned mouth corners and irregular teeth), prominent forehead, craniofacial disproportion in early life; and (5) at least one of the following relative criteria: asymmetry (limb length discrepancy and/or hemihypoplasia of skull, trunk, limbs), 5th finger clinodactyly and/or brachydactyly, low-set/dysmorphic ears, syndactyly of the 2nd and 3rd toes, cryptorchidism, feeding difficulties, speech delay/difficulties, and excessive sweating. All 10 SRS-hypo patients and four SRS-ue patients (1, 2, 6, 10) fulfilled these criteria. Four SRS-ue patients (3, 4, 5, 11) fulfilled all other criteria but had only one or two typical facial features, and were termed SRS-like. Families provided written informed consent. The study was approved by the ethics review board of the Hospital for Children and Adolescents, University of Helsinki, Finland and Karolinska Institutet, Sweden.

Affymetrix genotyping

The GeneChip Mapping 250K Sty gene array surveys 238 063 SNPs. Samples were prepared and hybridised according to the supplier's instructions (Affymetrix Inc, Santa Clara, California, USA). Arrays were scanned with a GeneChip Scanner 7G (Affymetrix Inc). All results have been submitted to ArrayExpress (E-MTAB-56).

Data analysis

Genotype-based analyses

GTYPE 4.1 (Affymetrix Inc) was used for basic quality metrics and analyses including BRLMM genotyping, average heterozygosity estimates, and pedigree checks. We used SNPTrio to identify inconsistencies in Mendelian inheritance of subsequent markers that suggested a genomic aberration.13

Intensity-based analyses

CNAG 2.0 was used for copy number change (CNC) and identification of loss-of-heterozygosity (LOH) region.14 Parents were used as reference for copy number estimations for patients. The Aroma.Affymetrix package was used in parallel for CNC identification.15 The program is implemented in the R software ( The analysis comprised several preprocessing steps of the raw intensity data, including allelic cross-talk calibration, quantile normalisation, probe-level summarisation (plm), and PCR fragment length normalisation. Copy number summarisation on log2-scale was called with the circular binary segmentation method from the DNAcopy package,16 using the robust average across all samples as reference.

Verification of CNCs with quantitative real-time PCR

Three selected findings were verified with PCR amplicons within the deleted/duplicated segment, using two copy number neutral amplicons as reference.17 Two amplicons were designed in Primer Express v2.0 (Applied Biosystems, Carlsbad, CA, USA), and one was derived from a previous report (online supplementary table 1).17 We performed quantitative real-time PCR analyses in 20 μl volumes with 1× Fast SYBR Green PCR Master Mix (Applied Biosystems), 10 ng genomic DNA, and optimised primer concentrations of 200–800 nmol/l (online supplementary table 1). Each amplicon was quantified in triplicate using the Fast SyBR program (95°C for 20 s followed by 40 cycles of 95°C for 3 s and 60°C for 30 s) on a 7500 Real-time PCR System machine (Applied Biosystems). Relative copy number estimates were derived by separate ΔΔCt calculations for the two copy number neutral amplicons. Three asymptomatic parents of growth-restricted children from our sample collection were included as controls when inferring relative copy number.


In total, 63 of 66 (95%) samples passed the quality criteria and were included in the final analyses. The excluded samples (BRLMM call rates <97%) included one patient (SRS-hypo-8) and the parents of SRS-ue-12. Data on individual call rates, gender, average heterozygosity, diagnosis and H19 methylation status are presented for all patients in table 1 and for all genotyped individuals in online supplementary table 2. Pedigree checks were performed in the 20 complete trios. The number of observed Mendelian inconsistencies per trio ranged between 238 and 623 (average 334 and median 302).

Table 1

Description of patients and genotyping characteristics

Structural variation

Genotype-based analyses

To identify potential regions of CNC or UPD, we searched for chromosomal enrichment of Mendelian inconsistencies using SNPTrio (table 2). No instances of whole-chromosome (or very large >5 Mb) uniparental inheritance suggesting UPD were found. Thus we concluded that there were no undiagnosed matUPD7 or matUPD11 cases among our patients. SNPTrio reported 15 stretches of Mendelian inconsistencies in 10 patients (table 2). Notably, a region in FBXL7 was reported to be maternally inherited in SRS-ue-1 and SRS-ue-2. Nine of the reported instances (60%) were verified by the intensity-based analysis, confirming them to be copy number aberrations (table 2).

Table 2

Regions of Mendelian inconsistency stretches reported by SNPTrio

Copy number analyses

Altogether 60 CNCs were identified with CNAG 2.0 (online supplementary table 3). A more sensitive detection method was provided by Aroma.Affymetrix with altogether 687 reported CNCs (online supplementary tables 4 and 5). Of these, 489 (71%) consisted of only two consecutive SNPs. In order to exclude false-positive results, the shorter CNCs were only considered when recurring CNV was being looked for (see below). As expected from the increased sensitivity, the average and median size of the CNCs reported by Aroma.Affymetrix were smaller than those from CNAG (online supplementary table 6). To compile a list of robust CNVs, we merged the lists from CNAG and Aroma.Affymetrix and identified 33 CNCs that were reported by both programs (table 3).

Table 3

Copy number changes detected by both CNAG and Aroma.Affymetrix

The largest unambiguous CNCs occurred in unexpected regions and were found in patients with idiopathic SRS (table 3 and figure 1A). In SRS-ue-6, a 2.6 Mb paternal deletion of 15q26.3 spanned 14 genes, from IGF1R to CSHY1, thus not including the telomere. In SRS-ue-4, a 1.1 Mb atypical distal 22q11.2 deletion spanned 15 genes from PI4KAP2 to PRAME. Further, a 2.7 Mb pseudoautosomal region (PAR) duplication spanned 16 genes from PLCXD1 to XG in a male patient, SRS-ue-5. All three CNCs were confirmed with quantitative real-time PCR (figure 1B). There were no CNVs in DGV that covered these whole regions.

Figure 1

Large copy number changes (CNCs) in patients with Silver–Russell syndrome (SRS). (A) Relative copy number as derived from the Aroma.Affymetrix program, using the circular binary segmentation (cbs) method.16 Each dot represents the relative copy number for one single nucleotide polymorphism (SNP), while the solid line represents the estimated smoothed average relative copy number, derived with the cbs-method. (B) Relative copy number estimates derived through the ΔΔCt method, using 6PDH (dark grey) and HMBS (light grey) as reference genes. Error bars represent standard error of the means of the copy number estimates. We confirm the heterozygous deletion of the IGF1R gene (15q26.3) in patient SRS-ue-6, the heterozygous atypical 22q11.2 deletion in SRS-ue-4, through a primer pair within the VPREB1, and the duplication of the pseudoautosomal region in SRS-ue-5, through a primer pair in the SLC25A6 (Xp22.33) gene. For primer sequences, see online supplementary table 1.

Large recurrent CNCs were found on 14q11.2, 15q11.2 and 17q21.3 in several patients (table 3 and online supplementary tables 3–5). However, these regions carry segmental duplications (, many CNVs have previously been reported, and they probably constitute normal interindividual variation in genomic architecture (table 3).

Six of the 33 CNCs from table 3 overlapped with the SNPTrio results (table 2, including the 15q26.3 and 22q11.2 deletions), lending strong support for their true nature. Three of these deletions were inherited from a parent, and two were not reported to DGV: a maternal deletion of CLLU1, PLEKHG7 and EEA1 (SRS-ue-5) and a paternal deletion of an intronic region of PRKCA (SRS-hypo-1) (table 3). Two additional CNCs from the compiled list were not reported to DGV: a de novo deletion of CADM1 intronic region in SRS-hypo-7 and a paternally inherited duplication of C6orf204 and PLN in SRS-ue-2. No duplications of chromosome 7p or 11p were identified.

Common CNVs

In our search for recurring genomic aberrations in SRS, we wanted to avoid CNVs and focus on CNCs that were not reported to DGV. We started from the 687 CNCs reported by Aroma.Affymetrix, as small CNCs identified in many individuals are also of interest. We could not identify inherited CNCs that were not reported to DGV, recurring in SRS, and showing parent-of-origin effects (ie, potential imprinting effects). When searching for CNCs that were found in at least three patients but not in DGV, we identified 10 regions containing known genes: CNTNAP2, RERG, AGPAT5, CSRP1, ITGB8, SPOCK1, TFCP2, FRMD6, POU6F1 and SH3MD4 (online supplementary tables 4 and 5). None of these genes were found exclusively in either H19 hypomethylated or molecularly unexplained SRS.

Potential recessive disease loci

LOH regions were found in four patients (table 4). All identified regions contained >400 SNPs, although they were not appreciably larger than reported in, for example, the HapMap population.18 None overlapped with each other or with regions that were implicated by the SNPTrio analysis. SRS-ue-1 carried two independent LOH regions, suggesting potential cryptic relatedness between parents. Two LOH regions overlapped with genomic regions previously implicated in SRS: a 6 Mb LOH region of 13q31.3-q32.1 (containing GPC5, GPC6 and SOX21) in SRS-ue-1 and an 11 Mb LOH region of 8p11.21-q11.23 (over 20 genes from SLC20A2 to RB1CC1) in SRS-ue-3.

Table 4

LOH regions detected with CNAG 2.0

Clinical history of the three patients with idiopathic SRS with large CNCs

SRS-ue-6, with the 15q26.3 IGF1R deletion, was born by caesarean section because of severe intrauterine growth retardation and reduced amniotic fluid at 36 weeks' gestation with a small placenta (245 g). Her birth length was 40 cm and weight 1780 g (−4.4 and −3.1 SDS, respectively), and her head circumference was 35 cm (+1 SD). She had feeding problems during infancy. At 0.9 years, height SDS was −4.5 and weight for height −14%, which both gradually increased to −2.4 and +5%, respectively, by 5.7 years, with acceleration in bone age. Her weight for height markedly increased to +64% (59.5 kg) by 13.8 years, and height SDS remained constant at −2.5 (144 cm). Her parents and four siblings are of normal height. She fulfils typical SRS stigmata (table 5), but her overall appearance is mild SRS. Overall development is delayed, with especially speech delay from age 1.5 years, with pronounced dysarthria, a small vocabulary and lexical syntactic dysphasia diagnosed at 8 years. Her early motor development was normal, but at 8.7 years impaired fine motor control was noted. Karyotype and growth hormone (GH) and thyroid secretion are normal. At 13 years she was found to have raised serum insulin concentrations and slight insulin resistance. She has had raised blood pressure (average 128/80 mm Hg) from 12 years of age, correlating with obesity.

Table 5

Clinical characteristics of patients

SRS-ue-4, with the atypical 22q11.2 deletion, was born at 36 weeks' gestation after an uneventful pregnancy. Her birth length was 41 cm and weight 1930 g (−4.2 and −3.0 SDS, respectively) and she required a feeding tube for 1 week. Congenital bilateral hip displacement and a ventricular septal defect, which later closed spontaneously, were noticed. During the first year, length SDS remained at −4, weight for height at −10%, and head circumference SDS at −2. Growth restriction persisted throughout childhood, with height SDS at −3.0 and weight for height at −10%, with ∼2.5 years delay in bone age. GH therapy lasting 1.5 years only marginally improved her height SDS (−3 to −2.5). At age 12.4 years, her height was 134.7 cm (−2.5 SDS) and weight 28.5 kg (−3%). Her parents and one sibling are of normal height. She has a SRS-like appearance with upper limb asymmetry, thoracic scoliosis, slight abnormality in the lower sacrum, a triangular face with micrognathia, but slight hypothelorism and a narrow face, a wide chest, hips rotated in valgus, and poorly formed long bones (table 5). Development, thyroid, GH, insulin-like growth factor I (IGFI) and IGF binding protein 3 (IGFBP3) concentrations, and karyotype were normal (table 5).

SRS-ue-5, with the PAR duplication, was born by caesarean section at 38 weeks' gestation with a small placenta (300 g). His birth length was 43 cm and weight 2000 g (−3.9 and −3.3 SDS, respectively). He has grown constantly at a height SDS −4 to −5 and a weight for height of +10%. He received GH treatment from age 13 to 14 years, without any response. At 16.8 years, he was 140.9 cm (−5.2 SDS) tall and weighed 36 kg (+12%). His parents and two siblings are of normal height. He has a SRS-like appearance with hemihypoplasia of the entire right side of the body including a 20 mm leg length discrepancy, but no facial asymmetry (table 5). Radiological examination revealed a horseshoe-shaped kidney, a bicuspid aortic valve, short IV and V metacarpals, slightly dysplastic metaphyses, exceptionally short 4th metatarsals, and poorly formed long bones, while the spine was normal. Development and serum GH and thyroid levels are normal.


Two major aetiological subgroups of SRS have been recognised, with alterations in chromosomes 7 and 11. To search for novel molecular subgroup(s) of SRS, we set out to screen the genomes of patients with SRS and their parents by a method that can detect structural genomic alterations as well as UPD. Our genome-scale approach revealed distinct, submicroscopic aberrations in three of 12 idiopathic SRS or SRS-like patients on chromosomes 15 and 22 and the PAR, but none on chromosomes 7 and 11. Further, as expected from studies of control populations,11 numerous aberrations (probably representing CNVs) were detected in our patients. We could not identify any unambiguous genomic aberrations in patients with H19 hypomethylated SRS at the resolution of the Affy250K Sty microarray. A genetic basis for the methylation aberration in these patients thus still remains to be explored.

All three patients with molecularly unexplained SRS with the large and rare genomic aberrations had growth restriction and dysmorphic features compatible with SRS diagnostic criteria, but clinically they fall into a mild SRS class and do not represent the classical SRS phenotype seen with H19 hypomethylation. A haploid deletion including the IGF1R gene was found in SRS-ue-6. This deletion, involving only 15q26.3, does not include the telomere and is the smallest deletion described to date that spans IGF1R.19–21 She shared typical symptoms of other reported patients with 15q26 deletions (including IGF1R) such as intrauterine and postnatal growth retardation, a triangular face, micrognathia and developmental delay (table 5).19–21 IGF1R is the prominent candidate gene for the growth phenotype, as patients with mutations in this gene resemble the patients with deletions.21 Further, a 15q trisomy overgrowth syndrome has recently been described, with IGF1R suggested as the candidate gene.22

We further identified a 22q11.21-q11.22 deletion (1.1 Mb) outside the typical DiGeorge region, but overlapping the atypical distal deletion region recently described to constitute a genomic disorder with symptoms distinct from DiGeorge (OMIM 611867) in SRS-ue-4.23 Interestingly, she shared many symptoms with reported patients including prenatal and postnatal growth restriction, clinodactyly and brachydactyly, cox valga and congenital bilateral hip displacement (table 5).23 However, most of these characteristics were described in isolation for distal deletion patients, and their compositae phenotypes were not SRS-like.23 An interesting candidate deleted gene for the SRS and growth restriction phenotype is MAPK1, as the IGF pathway is known to activate the MAPK-signalling pathway.24

A 2.7 Mb duplication of the PAR, containing several immune-related genes and the SHOX gene, was found in a male patient (SRS-ue-5). An extra dosage of SHOX might be expected to cause overgrowth; however, there are reports of SHOX trisomy with normal growth.25 Two boys with 47,XXY karyotype and SRS have been described,26 although the 47,XXY karyotype normally contributes to Klinefelter syndrome. The most similar PAR duplication that has been described in a male patient coincided with prepubertal-onset systemic lupus erythematosus.27 The patient was reported to have short stature, but no data were given on the extent and possible aetiology of the growth restriction. Our patient also inherited a 12q22 deletion of CLLU1, PLEKHG7 and EEA1 from his phenotypically normal mother. We note that the Affymetrix 250K array does not cover the Y chromosome, and, as our patient has not been karyotyped, the underlying structural arrangement is not known (eg, structural Y or X aberrations or XYY karyotype).

LOH has been documented in outbred control populations,18 but we screened patients with a diagnosis of SRS for LOH regions, potentially corresponding to novel recessive disease alleles. Through searching for LOH that overlapped with genomic regions previously implicated in SRS, we identified two interesting LOH regions on 8p11.21-q11.23 and 13q31-q32 in patients with idiopathic SRS (SRS-ue-3 and SRS-ue-1). SRS-ue-1, in particular, carried two independent LOH regions, suggesting parental cryptic relatedness. The 13q31-q32 region spans only three genes: GPC5, GPC6 and SOX21. GPC5 and GPC6 belong to the glypican family of glycoproteins which act as co-receptors to heparin-binding growth factor receptors and are potential regulators of morphogenesis and growth regulation.28 GPC5 is an interesting candidate gene for SRS or an SRS-like phenotype. It is expressed in the limb bud and is the closest homologue to GPC3, which causes the overgrowth syndrome, Simpson–Golabi–Behmel,28 29 sharing many characteristics with Beckwith–Wiedemann, which is associated with H19 hypermethylation and is considered to be opposite to SRS in terms of phenotypes and genomic causes.30 GPC3 also regulates the IGF–GH axis, which is disrupted in SRS.29 One girl with SRS-like features and a chromosome 13q deletion has been described.31 Our patient with LOH had prenatal and postnatal growth retardation, relative macrocephaly, a typical SRS face, low-set ears, clinodactyly, excessive sweating and a ventricular septal defect in early childhood. We are currently performing mutational and homozygosity analyses of the gene in our patients. The 11 Mb LOH on chromosome 8 overlaps with a deletion of 8q11-q12 that was described in a girl with SRS-like features.32 No obvious candidate gene for SRS or growth restriction is located within the overlapping region.

Future systematic genome-scale screens in molecularly unexplained SRS alongside detailed clinical descriptions will help to resolve novel distinct clinical and molecular groups, as have been resolved for matUPD7 associated with mild SRS4 and hypomethylation of the H19 imprinting control region, associated with classical/severe SRS.2 The new SRS phenotype–genotype correlations reported here possibly represent different branches of the SRS phenotype spectrum. All aberrations reported here are submicroscopic in size and thus undetectable through karyotyping, underlining the need for array technology in clinical genetics, where genotyping arrays provide the advantage of detecting UPD, duplications and LOH regions.

We note with interest that genomic alterations were identified in patients with mild SRS, whereas they appear to be rare in classical SRS. It might be argued that, in the presence of atypical molecular findings, diagnosis should be made according to molecular genetic aberrations. Until the discovery of matUPD7 and H19 hypomethylation, the diagnosis of SRS was made solely on clinical manifestations, for which no consistent universal criteria have been established. Dysmorphic evaluation is not unambiguous, and many dysmorphic features become milder with age, especially in SRS, making clinical diagnostics difficult in real life, let alone if evaluation is based on patient records and photographs as in the majority of studies. This phenotypic variability probably contributes to the heterogeneous genetic aetiology beyond the major groups of H19 hypomethylation and matUPD7. Perhaps the diagnosis of SRS should be used only for H19 hypomethylated cases, and other genetic entities be named on the basis of their molecular nature. This would decrease confusion in SRS research and increase both scientifically and clinically relevant data correlating with the molecular genetic entities, aiding clinicians to manage patients and their families.


We thank all patients and families, and also treating physicians for referring patients, especially Dr Hanna-Liisa Lenko. We thank Ms Riitta Lehtinen for excellent technical support and the BEA Affymetrix core facility at Novum.


Supplementary materials


  • Funding This work was supported by Magn Bergvalls stiftelse, the Swedish Research Council, Päivikki and Sakari Sohlberg Foundation, Sigrid Juselius Foundation, Helsinki University Hospital research funds, Finnish Foundation for Pediatric Research, and the Academy of Finland.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval This study was conducted with the approval of the ethics review board of the Hospital for Children and Adolescents, University of Helsinki, Finland and Karolinska Institutet, Sweden

  • Provenance and peer review Not commissioned; externally peer reviewed.