Article Text

Original article
A novel intellectual disability syndrome caused by GPI anchor deficiency due to homozygous mutations in PIGT
  1. Malin Kvarnung1,2,3,
  2. Daniel Nilsson1,2,3,4,
  3. Anna Lindstrand1,2,3,
  4. G Christoph Korenke5,
  5. Samuel C C Chiang6,
  6. Elisabeth Blennow1,2,3,
  7. Markus Bergmann7,
  8. Tommy Stödberg8,9,
  9. Outi Mäkitie1,2,10,
  10. Britt-Marie Anderlid1,2,3,8,9,
  11. Yenan T Bryceson6,
  12. Magnus Nordenskjöld1,2,3,
  13. Ann Nordgren1,2,3
  1. 1Department of Molecular Medicine and Surgery, Karolinska Institutet, Stockholm, Sweden
  2. 2Center of Molecular Medicine, Karolinska Institutet, Stockholm, Sweden
  3. 3Department of Clinical Genetics, Karolinska University Hospital, Stockholm, Sweden
  4. 4Science for Life Laboratory, Karolinska Institutet Science Park, Stockholm, Sweden
  5. 5Department of Neuropediatrics, Children's Hospital, Klinikum Oldenburg, Oldenburg, Germany
  6. 6Department of Medicine, Centre for Infectious Medicine, Karolinska Institutet, Karolinska University Hospital Huddinge, Stockholm, Sweden
  7. 7Institute for Neuropathology, Klinikum Bremen-Mitte, Bremen, Germany
  8. 8Department of Women's and Children's Health, Karolinska Institutet, Stockholm, Sweden
  9. 9Department of Child Neurology, Astrid Lindgren Children's Hospital, Karolinska University Hospital, Stockholm
  10. 10Children's Hospital, Helsinki University Central Hospital, University of Helsinki, and Folkhälsan Institute of Genetics, Helsinki, Finland
  1. Correspondence to Dr Malin Kvarnung and Ann Nordgren, Department of Molecular Medicine and Surgery, Karolinska Institutet, Stockholm 17176, Sweden; malin.kvarnung{at}, ann.nordgren{at}


Purpose To delineate the molecular basis for a novel autosomal recessive syndrome, characterised by distinct facial features, intellectual disability, hypotonia and seizures, in combination with abnormal skeletal, endocrine, and ophthalmologic findings.

Methods We examined four patients from a consanguineous kindred with a strikingly similar phenotype, by using whole exome sequencing (WES). Functional validation of the initial results were performed by flow cytometry determining surface expression of glycosylphosphatidylinositol (GPI) and GPI anchored proteins and, in addition, by in vivo assays on zebrafish embryos.

Results The results from WES identified a homozygous mutation, c.547A>C (p.Thr183Pro), in PIGT; Sanger sequencing of additional family members confirmed segregation with the disease. PIGT encodes phosphatidylinositol-glycan biosynthesis class T (PIG-T) protein, which is a subunit of the transamidase complex that catalyses the attachment of proteins to GPI. By flow cytometry, we found that granulocytes from the patients had reduced levels of the GPI anchored protein CD16b, supporting pathogenicity of the mutation. Further functional in vivo validation via morpholino mediated knockdown of the PIGT ortholog in zebrafish (pigt) showed that, unlike human wild-type PIGT mRNA, the p.Thr183Pro encoding mRNA failed to rescue gastrulation defects induced by the suppression of pigt.

Conclusions We identified mutations in PIGT as the cause of a novel autosomal recessive intellectual disability syndrome. Our results demonstrate a new pathogenic mechanism in the GPI anchor pathway and expand the clinical spectrum of disorders belonging to the group of GPI anchor deficiencies.

  • Clinical genetics
  • Epilepsy and seizures
  • Genetic screening/counselling
  • Developmental
  • 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.


Syndromes with a phenotype that includes intellectual disability and/or seizures are clinically and aetiologically heterogeneous. Among patients with a history of parental consanguinity, autosomal recessive monogenic traits predominate the aetiology. Nevertheless, there is extensive genetic heterogeneity in this subgroup, with many individuals harbouring unique disease-causing mutations in genes detected only in their family or in a small proportion of patients.1 In the present study, we used whole exome sequencing to delineate the genetic basis for a previously uncharacterised autosomal recessive syndrome. We examined four patients from the same kindred that displayed a strikingly similar phenotype, characterised by distinct facial features, intellectual disability, hypotonia and seizures, in combination with abnormal skeletal, endocrine, and ophthalmologic findings. Clinical and molecular investigations were performed on two sisters (V-1 and V-2), their female third cousins (monozygotic twins; V-4 and V-5) as well as the unaffected brother (V-3), and the healthy parents (IV-1, IV-2, IV-3 and IV-4), all belonging to one consanguineous Aramaic family originating from Turkey (figure 1 and, for an extended pedigree, see online supplementary figure S1).

Figure 1

Facial features of the patients with PIGT mutations. Patients at the approximate ages of 1 and 3 years, respectively. Note the high forehead with bitemporal narrowing, depressed nasal bridge, long philtrum with a deep groove and open mouth consistent with general hypotonia.

Subjects and methods

Human subjects

The study was approved by the regional ethics committee, Stockholm, and written informed consent was obtained from each participating individual or their respective legal guardians.

Exome sequencing

Libraries for sequencing on Illumina HiSeq2000 (Illumina) were prepared from genomic DNA derived from blood samples of affected individuals (V-1, V-2, V-4) as well as the healthy parents (IV-1, IV-2, IV-3, IV-4) and exome sequences enriched with Agilent SureSelect Human All Exon 50M (Agilent), according to the manufacturer’s instructions. Post-capture libraries were sequenced as 2×100 bp paired end reads on the Illumina sequencer. Reads were base-called using offline CASAVA (v 1.7, Illumina). Sample library preparation, sequencing, and initial bioinformatics up to base calling and demultiplexing was performed at the Science for Life Laboratory, Stockholm. An inhouse pipeline, freely available under a GPL licence, was used to process reads and arrive at candidate genes. Briefly, reads were mapped to the human reference genome (hg19) using Mosaik (v1.0.1388). Duplicate read pairs were removed using Mosaik DupSnoop. Variants were called using the samtools package (v.0.1.18).2 These were quality filtered (Q>=20), and annotated using ANNOVAR (v.2011 June 18).3 Variants were further filtered using ANNOVAR to remove those found at a 1000 genomes4 minor allele frequency (MAF) of 2% and above, as well as variants found in dispensable genes, truncated at a MAF of more than 1% in any 1000 genomes subpopulation, and variants not predicted to be damaging by PolyPhen25 (P threshold of 0.85). Non-synonymous variants, indels and putative splice site variants were retained. Further filtering against dbSNP1326 was performed. The samples were compared according to pedigree, and variants in concordance with a recessive inheritance model were retained.

Sanger sequencing

DNA was extracted from blood samples of the patients (V-1, V-2, V-4), the unaffected brother (V-3), as well as the healthy parents (IV-1, IV-2, IV-3, IV-4), followed by PCR amplification of PIGT exon 4 and Sanger sequencing, performed on an ABI 3730 genetic analyser (Applied Biosystems) using standard protocols. Primers used for PCR amplification of DNA fragments and Sanger sequencing were PIGT 4F 5′-CCCATCTCTGGAATGTGGAT-3′ and PIGT 4R 5′-CCATGCCAAGTGCTCCATAC-3′.

Flow cytometry

Peripheral blood from patients (V-1, V-2, V-4), the unaffected brother, the parents and, for comparison, healthy age matched controls, were collected in sodium heparin tubes and processed within 24 h. Whole blood was stained with fluorochrome conjugated antibodies to glycosylphosphatidylinositol (GPI) anchored protein CD16 (3G8, BD Bioscience), as well as CD45 (HI30, BD Bioscience), CD3 (S4.1, Invitrogen), and CD56 (NCAM16.2, BD Bioscience) to differentiate granulocyte, monocyte, lymphocyte, and natural killer (NK) cell populations. In addition, peripheral blood mononuclear cells were stained with fluorochrome conjugated antibodies to CD59 (H19, BD Bioscience). The surface expression of the GPI anchor itself was quantified using fluorochrome conjugated aerolysin (FLAER, Pinewood Scientific), a bacterial toxin that specifically binds all GPI anchors. Isotype antibodies were used as negative controls (MOPC-21, BD Bioscience). Whole blood was fixed after staining with FACS Lysing Solution (BD Bioscience). Data on cellular characteristics were acquired on a flow cytometer (LSR Fortessa, BD Bioscience) and analysed (FlowJo v9.5, TreeStar). Median fluorescence intensities (MFI) of the various GPI anchored proteins on different cell populations were determined.

Zebrafish embryo microinjection and manipulation

Morpholinos were designed by and obtained from Gene Tools; 1 nl of the indicated concentration was injected into wild-type (WT) zebrafish embryos at the one to four cell stage (n=42–63 embryos/injection, repeated at least twice; with masked scoring). MO-1 targeted the junction of exon 2/intron 2–3 (AAGCACATGTAAGCACTCACTCGGT) and MO-2 targeted the junction of exon 4/intron 4–5 (CAAATGTATCGATCTGGACCTTTGA). Embryos were reared at 23°C and scored live at 24 h post-fertilisation. To test morpholino (MO) efficiency, we harvested whole embryos in Trizol (Invitrogen), and generated oligo-dT primed cDNA (Superscript III, Invitrogen), for subsequent PCR. To generate human WT and p.Thr183Pro encoding mRNA, the full PIGT open reading frame was obtained from the Invitrogen ORF (open reading frame) collection and cloned into the pCS2+ plasmid, linearised with NotI, and performed in vitro transcription with the SP6 mMessage mMachine kit (Ambion). For rescue experiments, 9 ng of MO-1, 4 ng of MO-2, and 150 pg of PIGT mRNA was injected respectively. Live images of zebrafish embryos were acquired on a Leica DFC 230 microscope using Leica application Suite v4.1.0 (Leica Microsystems, Switzerland).


Clinical reports

The patients were born at full term after uneventful pregnancies and all presented at birth with hypotonia, mild macrosomia, and mild macrocephaly. From an early age, the children showed signs of impaired motor and cognitive development. Neurological examination at the age of 3 months confirmed a psychomotor delay with hypotonia, and a diagnosis of severe motor and intellectual disability was later established. No consistent pattern of intracranial pathology was identified on brain imaging, but rather variable findings such as frontotemporal atrophy and cerebellar hypoplasia, primitive Sylvian fissures indicating a possible neuronal migration disorder, and in one patient apparently normal brain structures. All patients developed seizures before 2 years of age, most commonly with an initial presentation of generalised tonic clonic seizures related to febrile episodes, and with normal findings on electroencephalogram (EEG). Subsequently, the patients developed different types of unprovoked seizures, confirmed by pathological findings on EEG.

Ophthalmologic examinations revealed almost identical findings in all patients, with impaired vision, as well as abnormal motility of the eyes. Auditory examination showed normal hearing. All children had tooth abnormalities, in particular premature loss of upper and lower incisors, as well as renal anomalies, predominantly nephrocalcinosis and ureteral dilation. One patient had unilateral renal cysts and dysplasia. Skeletal features were present in all and included pectus excavatum, scoliosis, short upper extremities, and abnormal skull shape due to premature closure of sutures, in addition to radiologic abnormalities—more specifically brachycephaly, slender and osteopenic long bones with relatively large secondary ossification centres, wide and long femoral necks with enlarged secondary ossification centres, short ulnae, and delayed bone age (see online supplementary figure S2).

Other observed abnormalities included inverted mammillae and cardiac abnormalities, such as restrictive cardiomyopathy and patent ductus arteriosus. The patients also presented mild dysmorphic facial features (figure 2). One of the patients (V-5) died from bronchopneumonia at the age of 3 years. Autopsy revealed atypical lobulation of the lungs with four and three lobes on the right and left side, respectively. Nephrocalcinosis was confirmed by the identification of lamellar crystal deposition. Furthermore, there was apparent global cerebral atrophy, predominantly frontal, as well as severe cerebellar atrophy with vermis hypoplasia. Microscopic examination showed loss of Purkinje cells as well as granule cells with focal proliferation of the Bergmann glia. Clinical findings are summarised in table 1.

Table 1

Clinical findings in patients with PIGT mutations

Figure 2

Glycosylphosphatidylinositol (GPI) linked protein expression on granulocytes. Whole blood from patients (V-1, V-2, V-4), the unaffected sibling (V-3), parents (IV-1, IV-2, IV-3, IV-4), and healthy controls were stained with fluorochrome conjugated aerolysin (FLAER) or fluorochrome conjugated antibodies to Fc receptor CD16 expressed as the GPI linked protein CD16b on granulocytes or transmembrane protein CD16a on natural killer (NK) cells, as indicated. (A) Lines represent staining of proteins on indicated cell populations. Filled histograms represent respective isotype control staining. (B) Bar graphs depict relative median fluorescence intensities (R-MFI), obtained by subtracting staining intensities of isotype control from true signals.

Biochemical analyses revealed low plasma concentrations of alkaline phosphatase, and plasma calcium in the upper range or above normal reference values with concomitant hypercalciuria and suppression of parathyroid hormone values. Plasma phosphate was normal. The results were consistent over time and within the group of patients (see online supplementary table S1). Analyses of routine biochemical tests, thyroid function, urine organic acids, plasma amino acids, morphology and immunohistochemistry of muscle fibres, tests for mitochondrial and peroxisomal disorders, array comparative genomic hybridisation, and karyotyping were all normal.

Exome sequencing

In order to identify a potential causative mutation in this family, we performed whole exome capture and sequencing on genomic DNA derived from blood samples of affected individuals (V-1, V-2, V-4) as well as the healthy parents (IV-1, IV-2, IV-3, IV-4). Two called variants that matched the inheritance pattern were not previously seen in public databases and also predicted damaging by PolyPhen2. One of the variants, TUBB1 NM_030773:c.925C>T:p.Arg309Cys, was deemed unlikely to cause the patients’ symptoms. Mutations in this gene may cause macrothrombocytopenia (MIM 613112), a disorder of which the patients showed no signs. The remaining variant, PIGT NM_015937:c.547A>C:p.Thr183Pro, was predicted to be damaging by PolyPhen-25 and SIFT7 and was not detected in 6500 ESP (Exome Sequencing Project) exomes, 1000 genomes,8 200 Danish exomes9 or in a database of 100 local sample exomes. Furthermore, because the affected amino acid position is highly conserved throughout evolution, from yeast to man (see online supplementary figure S3), PIGT is a very strong candidate gene.

Sanger sequencing

Presence of PIGT c.547A>C was verified by Sanger sequencing and further analyses of unaffected family members confirmed that the mutation segregated with the disease, presuming an autosomal recessive inheritance pattern (figure 1).

PIGT function and effects of c.547A>C on cell surface expression of GPI and associated proteins

PIGT contains 12 exons and encodes the protein phosphatidylinositol-glycan biosynthesis class T (PIG-T), which is a 578 amino acid subunit of a heteropentameric enzyme complex that catalyses the attachment of proteins to glycosylphosphatidylinositol (GPI).10 ,11 GPI is a glycolipid structure that is synthesised in a multi-step process in the endoplasmic reticulum, followed by linkage to the C-terminus of certain proteins via a transamidation reaction and transport to the outer layer of cell membranes.12 Thus, GPI functions as a plasma membrane anchor for extracellular proteins. The molecule is ubiquitous in eukaryotic cells and essential for life.13 ,14 Given the function of PIGT, we investigated possible effects of the mutation on surface expression of GPI and GPI anchored proteins (GPI-APs). Binding of fluorochrome conjugated aerolysin (FLAER) to granulocytes was lower in individuals homozygous for the c.547A>C variant in PIGT (figure 3). Anti-CD16 antibodies revealed decreased expression of the GPI linked Fc receptor CD16b on granulocytes (figure 3). In contrast, the closely related non-GPI linked Fc receptor CD16a, that is recognised by the same antibody but encoded by a separate gene and expressed on NK cells and some T cells, was normally expressed on NK cells. Moreover, the levels of CD59 on mononuclear cells were lower in the patients compared to the parents and healthy controls (see online supplementary figure S4). The data suggest that the c.547A>C PIGT variant results in an impairment in membrane anchoring of GPI linked proteins in the patients.

Figure 3

Pedigree of the family with individual results of Sanger sequencing. Sanger sequencing of PIGT exon 4 confirmed the presence of the c.547A>C variant (arrow) in a homozygous state among the affected individuals (filled) (V-1, V-2, V-4) and a heterozygous state in the healthy parents (IV-1, IV-2, IV-3, IV-4). The unaffected brother (V-3) did not carry the variant on either allele.

In vivo assay on zebrafish embryos

To further investigate the function of PIGT and to assay phenotypic consequences of the homozygous c.547A>C variant, we used MO based suppression of pigt in zebrafish embryos. Reciprocal BLAST identified a single zebrafish PIGT ortholog (pigt; 69% protein sequence identity with human PIGT). Two different splice blocking (sb) MOs were injected at increasing concentrations into embryos at the one to four cell stage, and embryo batches were subsequently scored for gastrulation defects (shortened body axes, longer somites, and broad and kinked notochords) at the eight to 10 somites stage. Morphant embryos were subdivided into class I or class II morphants based on phenotype severity (figure 4A). For each MO independently, we observed a dose dependent increase in the number of abnormal morphant embryos and severity of the defects (see online supplementary figure S5B,C), due to aberrant pigt mRNA splicing as determined by reverse transcriptase PCR (9 ng and 4 ng for MO-1 and MO-2, respectively; see online supplementary figure S5A). We then used the gastrulation phenotype to further evaluate the pathogenicity of the homozygous c.547A>C variant. We compared the ability of equivalent doses of PIGT WT or c.547A>C mRNA to rescue the MO induced gastrulation. Co-injection of MO and WT human PIGT mRNA resulted in a significant rescue of the abnormal phenotype (26% abnormal embryos for MO+WT vs 51% for MO alone; p=0.0003; n=97–163 embryos/injection; MO-1 is shown, figure 4B). In contrast, co-injection with MO and PIGT mRNA harbouring the c.547A>C variant resulted in significantly higher numbers of aberrant embryos as compared with MO and PIGT mRNA WT, suggesting that the variant impairs enzyme function (figure 4B).

Figure 4

In vivo complementation studies of PIGT c.547A>C. Morpholino (MO) induced suppression of pigt in zebrafish embryos generates gastrulation defects. Live zebrafish embryo images were acquired on a Leica DFC 230 microscope using Leica application suite v4.1.0 (Leica Microsystems, Switzerland). Dorsal and side view of normal embryos, class 1 and class 2 morphants are shown (A). Quantification of live scoring of pigt MO and human mRNA (co)injections. Embryos were injected with the indicated dose (9 ng MO-1 and/or 150 pg PIGT mRNA) and scored live according to criteria shown in panel A. n=97–163 embryos/injection, repeated at least twice; with masked scoring; NS, not significant; all comparisons were made using χ2 tests (B).


The PIGT gene belongs to a group of at least 25 genes that are involved in the stepwise process of producing mature GPI-APs at the cell surface.15 Mutations in PIGT or other genes encoding different subunits of the GPI transamidase have not been previously reported as a cause of human disease. However, disruption of components acting earlier in the pathway are known to cause a wide range of clinical symptoms with variable severity. The first inherited defect to be identified was GPI deficiency (MIM 610293) caused by a mutation in the promoter region of PIGM.16 In recent years, recessive germline mutations in PIGV,17 PIGN,18 PIGL,19 PIGA,20 and PIGO21 have been linked to additional GPI deficiency syndromes (see online supplementary table S2).22–24 In addition, somatic mutations in PIGA are linked to paroxysmal nocturnal haemoglobinuria (MIM 300818).25 Moreover, genes required for the synthesis of molecules that act as donors of moieties essential for GPI biosynthesis indirectly affect this pathway.26 A number of these genes have been linked to congenital syndromes, due to defective synthesis of dolichol-phosphate-mannose (see online supplementary table S3).27–34

Comparing the identified c.547A>C PIGT variant with disease-causing mutations in other genes of the GPI biosynthesis pathway, there are clear similarities in that all mutations reported to date are hypomorphic on at least one allele and lead to partial loss of protein function with subsequent lowering of GPI-AP levels (see online supplementary table S2). These findings suggest that homo- or hemizygous null mutations in genes of this pathway would not be compatible with life, a suggestion further supported by the observation that complete disruption of GPI synthesis, by targeted deletion of PIGA, is embryonically lethal in mice.13

The phenotypes of previously described GPI anchor deficiency syndromes are diverse, with different symptoms depending on which step of the biological process is involved, which in turn corresponds to mutations in specific genes (see online supplementary table S2). Common to all syndromes, including the one that we describe here, is a phenotype that includes intellectual disability and/or seizures. Many of the additional features present in the patients in this study can be seen among patients with other GPI anchor defects. Craniosynostosis and prominent sutures with asymmetric shape of the skull are documented in patients with mutations in PIGA, PIGO, and PIGN. Similar to our patients, individuals with mutation in the latter gene also have macrocephaly, macrosomia, and impaired vision with nystagmus. Hypoplasia of the cerebellum and vermis, as seen in one patient of the present study, has been reported among patients with mutations in PIGA and PIGN.

The findings of skeletal hypomineralisation, premature tooth loss, short arms, and nephrocalcinosis have not yet been documented among other patients with deficient GPI synthesis. Interestingly, all of these symptoms are present among patients with infantile hypophosphatasia (MIM 241500), an autosomal recessive disorder caused by mutations in ALPL, the gene encoding tissue non-specific alkaline phosphatase (TNS-ALP or TNAP).35 ,36 The phenotype of this syndrome also includes craniosynostosis, seizures, and hypotonia. Biochemically, hypophosphatasia is characterised by low concentrations of serum alkaline phosphatase, and hypercalcaemia and hypercalciuria, findings that were also present in the patients of this study. TNAP is a GPI anchored enzyme essential for bone mineralisation and central nervous system development.37 ,38 Our findings suggest that decreased concentrations of functionally active TNAP with resulting phenotypic effects could be caused not only by mutations in ALPL, but also by deficient linkage of the GPI anchor to TNAP.

Hyperphosphatasia with mental retardation syndrome, caused by mutations in PIGV or PIGO, is characterised by elevated concentrations of serum alkaline phosphatase as opposed to the reduced values seen in our patients. Previous in vitro studies concerning the effect of deficiencies in various steps of the GPI anchor pathway on extracellular ALP values have been performed. These studies demonstrate that defects in the late steps of GPI synthesis (PIGV-PIGF) result in release of ALP into the extracellular space, whereas defects encompassing the GPI transamidase components result in a reduced concentration of extracellular ALP, most likely due to intracellular degradation of the protein.39 Thus, the biochemical and molecular genetic findings in the patients of this study are clearly in line with previous molecular studies. Even though ALP may be one of the most studied GPI-APs, there are in total more than 150 human proteins linked to GPI.40 These proteins play diverse biological roles, such as hydrolytic enzymes, receptors, adhesion molecules, complement regulatory proteins as well as other immunologically important proteins,15 and their functional insufficiency may contribute to the full phenotype of our patients.

Summarising the results of molecular investigations, in vivo functional studies in zebrafish embryos, human phenotypic characterisation, and comparison with results of previous studies, we find strong evidence that a syndrome of intellectual disability, hypotonia, seizures, and skeletal and ophthalmologic findings seen in the patients of this study is caused by mutations in PIGT. Mutations in PIGT or other genes encoding different subunits of the GPI transamidase have not been previously reported as a cause of human disease. Thus, the results of this study demonstrate a new pathogenic mechanism in the GPI anchor pathway and expand the spectrum of disorders belonging to the emerging group of GPI anchor deficiencies.

Web resources

URLs for data presented:


Inhouse pipeline Etiologica,

1000 Genomes Project,


Online Mendelian Inheritance in Man (OMIM),

PolyPhen 2,




We would like to thank all members of the family participating in this study.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • MK, DN and AL contributed equally

  • Contributors All authors included in this manuscript fulfil the criteria of authorship. There is no one else who fulfils the criteria but has not been included as an author.

  • Funding Financial support was provided through the regional agreement on medical training and clinical research (ALF) between Stockholm County Council and Karolinska Intitutet and, in addition, by grants from the Swedish Research Council, Kronprinsessan Lovisa, Karolinska Institutet, Linnea and Joseph Carlsson foundation and Sabbatical Leave Programme of the European Society for Paediatric Endocrinology through an educational grant from Eli Lilly International Corporation (OM). Computer resources were provided by SNIC through Uppsala Multidisciplinary Center for Advanced Computational Science (UPPMAX) under project b2011162.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Regional Ethics Committee, Stockholm.

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