Objective Pyrin-Associated Autoinflammation with Neutrophilic Dermatosis (PAAND) is a recently described monogenic autoinflammatory disease. The causal p.S242R MEFV mutation disrupts a binding motif of the regulatory 14-3-3 proteins within pyrin. Here, we investigate a family with clinical features consistent with PAAND in whom the novel p.E244K MEFV mutation, located in the +2 site of the 14-3-3 binding motif in pyrin, has been found.
Methods Multiplex cytokine analyses were performed on p.E244K patient and control serum. Peripheral blood mononuclear cells were stimulated ex vivo with lipopolysaccharide (LPS). In vitro, inflammasome complex formation was evaluated by flow cytometry of Apoptosis-associated Speck-like protein containing a Caspase recruitment domain (ASC) specks. Interleukin-1β (IL-1β) and IL-18 production was quantified by ELISA. The ability of the p.E244K pyrin mutation to interact with 14-3-3 was assessed by immunoprecipitation.
Results PAAND p.E244K patient serum displayed a different cytokine profile compared with patients with Familial Mediterranean Fever (FMF). In overexpression models, p.E244K pyrin was associated with decreased 14-3-3 binding and increased ASC speck formation. THP-1 monocytes expressing PAAND pyrin mutations demonstrated spontaneous caspase-1-dependent IL-1β and IL-18 secretion, as well as cell death, which were significantly greater than those of wild-type and the FMF-associated mutation p.M694V.
Conclusion In PAAND, disruption of the +2 position of a 14-3-3 binding motif in pyrin results in its constitutive activation, with spontaneous production of IL-1β and IL-18, associated with inflammatory cell death. The altered serum cytokine profile may explain the different clinical features exhibited by PAAND patients compared with those with FMF.
- autoinflammatory disease
Statistics from Altmetric.com
Pyrin-Associated Autoinflammation with Neutrophilic Dermatosis (PAAND) is a recently described monogenic autoinflammatory condition caused by a heterozygous mutation in the MEFV gene resulting in the p.S242R substitution in pyrin.1 The dominant clinical phenotype of prolonged fever and neutrophilic dermatosis (eg, acne, pyoderma gangrenosum), and potentially the mechanism of disease, differs from the classical pyrin-associated disease, Familial Mediterranean Fever (FMF).
The p.S242 site of pyrin forms a 14-3-3 binding motif.1 2 Although there are a number of variations of 14-3-3 recognition motifs reported, all contain a phosphorylated serine or threonine residue.3 4 In its inactive state, pyrin is phosphorylated by serine-threonine kinases PKN1 and PKN2 at residues p.S208 and p.S242, and is bound to 14-3-3 proteins.5 When triggered in response to RhoGTPase modifications, such as those induced by the pathogen Clostridium difficile, there is dephosphorylation of pyrin at p.S208 and p.S242 residues and loss of 14-3-3 binding.1 5 6 In vitro models show that the p.S242R pyrin mutation is constitutively dephosphorylated, with reduced 14-3-3 binding.1 The resulting increased pyrin inflammasome activation and enhanced IL-1β production appear to drive the pathology in PAAND.1
Here, we report a novel mutation in the MEFV gene in a family with clinical features of PAAND that results in an altered 14-3-3 binding motif and constitutive activation of pyrin. We also confirm phenotypic differences and identify cytokine differences between PAAND and FMF.
We investigated three symptomatic patients in one family. We used patients with homozygous p.M694V FMF as disease controls, and blood donors as healthy controls. This study was approved by the Hospital Clinic-IDIBAPS Ethics Committee.
Patient cell stimulation and analysis
Fresh serum samples were collected from patients and controls, and cytokine quantification was performed by Luminex multiplex assay. PAAND patients had active clinical features at the time of collection. For human IL-18 and IL-18BP, serum was assayed in multiplex on a Luminex Magpix system (Bio-Rad, Hercules, California, United States). Bio-Rad group II cytokine standard was used for IL-18, whereas recombinant human IL-18BPa-Fc chimeric protein (R&D Systems, Minneapolis, Minnesota, United States) was used as standard for IL-18BP.
Peripheral blood mononuclear cells (PBMCs) were isolated using Histopaque-1077 (Sigma-Aldrich, St Louis, Missouri, United States) and treated with Escherichia coli LPS serotype 055:B5 (Sigma-Aldrich; 1 µg/mL, 2 hours at 37°C) or left untreated. IL-1β was measured on cell supernatants by ELISA (eBioscience, San Diego, California, United States) while other cytokine quantification was performed by Luminex multiplex assay as described above. Cells were fixed with 2% paraformaldehyde and stained for the detection of Apoptosis-associated Speck-like protein containing a Caspase recruitment domain (ASC) specks by Time of Flight Inflammasome Evaluation using the rabbit polyclonal antibody anti-ASC (N-15)-R (Santa Cruz Biotechnology, Dallas, Texas, United States) as previously described.7 Alternatively, for the detection of active caspase-1, PBMCs were incubated for 20 min with Fluorochrome Inhibitor of Caspases (FLICA)660 reagent (ImmunoChemistry Technologies, Bloomington, Minnesota, United States) and fixed following manufacturer’s recommendations. Monocytes were detected with the APC-vio770 mouse anti-human CD33 antibody (Miltenyi Biotech, Bergisch Gladbach, Germany) and with the APC-Cy7-conjugated anti-human CD14 antibody (TONBO Biosciences, San Diego, California, United States). Stained cells were acquired on a BD FACSCanto cytometer.
Heat maps representing cytokine expression profiles were created using Morpheus software (Broad Institute, Cambridge, Massachusetts, United States).
HEK293T cells were cultured in Dulbecco’s Modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum (FCS) and transfected with mCherry-pyrin or GST-pyrin,8 GFP-ASC,9 or V5-Proline Serine Threonine Phosphatase-Interacting-Protein 1 (PSTPIP1) (HsCD00438559, DNASU Plasmid repository) constructs using Lipofectamine (Life Technologies) according to manufacturer’s instructions. Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)/Cas9 techniques were used for generation of MEFV KO and CASP1 KO THP-1 cells, as has been described.1 10 These cells were cultured in Roswell Park Memorial Institute medium (RPMI) supplemented with 10% FCS.
Lentiviral infection of THP-1 cells
MEFV KO THP-1 cells were reconstituted with pyrin by lentiviral transduction. A lentiviral construct was generated through ligation of MEFV cDNA into BamHI and AgeI restriction sites on the pFUGW backbone after performing site directed mutagenesis of the BamHI restriction site within MEFV. Lentivirus was produced as previously described.10 One million THP-1 cells were seeded per well in six-well plates with 3.5 mL of virus and 24 µg of polybrene. A total of 6 million THP-1 cells were seeded per condition. Plates were centrifuged at 840 g for 3 hours and then incubated at 37°C overnight. Cells were collected the following day, washed in phosphate buffered saline (PBS), reseeded in fresh media and incubated at 37°C overnight. After a further 24 hours, live and dead cells were separated using Ficoll density gradient centrifugation (GE Healthcare, Chicago, Illinois, United States). Live cells were seeded for experiments. Supernatant was harvested after 24 hours for cytokine analysis by ELISA for IL-1β and IL-18 (DY201 and DY008, R&D Systems). Cytokines from cell culture supernatant were also quantified using Bio-Plex Pro Assay (Bio-Rad). Cell death was analysed by flow cytometry using propidium iodide (Sigma-Aldrich) staining at 1 µg/mL. Where indicated, priming of cells was performed with Pam3CSK4, a synthetic TLR1/2 agonist (InvivoGen, San Diego, California, United States). Cells were also lysed using radioimmunoprecipation assay buffer to assess expression of pyrin by western blotting.
Site-directed mutagenesis was performed using the QuickChange Lightning Kit (210519–5, Agilent Technologies, Santa Clara, California, United States) according to manufacturer’s instructions. Mutations were introduced to pyrin-expressing constructs using the following oligonucleotide primers:
Fluorescence microscopy and flow cytometry
HEK293T cells were transfected with 25 ng wild type (WT) or mutant mCherry-MEFV and 5 ng GFP-ASC, and ASC specks were quantified 16 hours later using flow cytometry, as previously described.7 Colocalisation experiments were performed using mCherry-MEFV and GFP-ASC transfected into 1×105 HEK 293 T cells seeded in ibidi chamber slides (ibidi GmbH, Munich, Germany). Images were taken with a Zeiss LSM 780 Confocal microscope and were processed using FIJI software (National Institutes of Health, Bethesda, Maryland, United States).
Immunoprecipitation and western blotting
HEK293T cells (3×106 cells) were transfected with 5 µg of GST-tagged WT or mutant pyrin, with or without WT PSTPIP1. Where indicated, cells were treated with Clostridium difficile Toxin B protein (TcdB, 5 µg/mL, ab124001, Abcam, Cambridge, United Kingdom) 16 hours before harvesting. Cell lysates were generated 48 hours after transfection using 1% NP-40 lysis buffer supplemented with protease inhibitors and sodium orthovanadate. Immunoprecipitation of pyrin was performed using glutathione sepharose 4B (GE Healthcare). After washing, bound proteins were eluted from beads using 2x sodium dodecyl sulphate (SDS) buffer and boiling at 90°C. Immunoblots were prepared using 4%–12% Novex SDS-Polyacrylamide gel electrophoresis (Invitrogen, Carlsbad, California, United States) gels in MES running buffer, followed by transfer on to nitrocellulose membranes. Membranes were blocked with tween/tris-buffered saline (TBST) +3% bovine serine albumin (BSA) at room temperature and subsequently probed overnight at 4°C with antibodies against pan-14-3-3 (1:500 Santa Cruz #sc-629-G), 14-3-3τ (1:500 Santa Cruz #sc-59414), 14-3-3ε (1:1000 Biorbyt #orb6357), pSer 14-3-3 binding motif (1:500 Cell signalling #9601), pyrin (1:500 AdipoGen #AL196), p10 Caspase-1 (1:200 Santa Cruz #sc-515), IL-1β (1:1000 R&D #AB-401-NA), GST (1:1000 in-house), PSTPIP1 (1:500 Abnova #H00009051) and actin (1:5000 Santa Cruz #sc-1616). All antibodies were prepared in TBST +1% BSA.
Mann-Whitney non-parametric test was used for the analysis of data in figure 2. Two-tailed t-tests were performed in other analysis using Prism software (GraphPad Software, La Jolla, California, United States). Data are represented as mean+/−SEM unless otherwise specified (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001).
PAAND family with a novel mutation in MEFV
The index patient is a 43-year-old woman of Spanish descent with a 30-year history of chronic and severe pustular acne, severe hidradenitis suppurativa, recurrent pyoderma gangrenosum, recurrent long-lasting febrile episodes, neutrophilic panniculitis as well as polyarthralgia and oligoarthritis of small and large joints (figure 1A). Raised C-reactive protein (CRP), erythrocyte sedimentation rate (ESR) and peripheral blood neutrophil count persisted despite treatment with corticosteroids and the IL-1 receptor antagonist (IL-1Ra), anakinra (figure 1B). Long-lasting (8 years) clinical benefit was seen with the chimeric anti-tumor necrosis factor (TNF)-α monoclonal antibody infliximab. However, loss of efficacy of infliximab was observed and necessitated switching to the human anti-TNF-α monoclonal antibody adalimumab when symptoms recurred. Although a clinical diagnosis of pyogenic arthritis, pyoderma gangrenosum and acne (PAPA) syndrome was suspected, genetic testing of PSTPIP1 failed to reveal a pathogenic mutation. Pathogenic mutations in NCSTN, reported in familial cases of hidradenitis suppurativa, were absent.11 The recent description of PAAND, a condition with significant clinical overlap with PAPA syndrome, prompted exon 2 MEFV sequencing in this patient, which revealed the heterozygous c.730G >A transition in the MEFV gene encoding for the p.E244K mutation (figure 1C). This mutation was absent from the 1000 Genomes Project, Exome Aggregatium Consortium, Exome Variant Server and 250 Spanish healthy controls. Furthermore, it had not been reported on the INFEVERS database.12–14 The locus is highly conserved across species (figure 1D) and the amino acid substitution predicted to be damaging using MutationTaster,15 Sorting Intolerant from Tolerant16 and Polymorphism Phenotyping v2.17 Evaluation of the patient’s mother and brother, both of whom have had dermatitis and long-lasting (>30 years) severe nodulocystic acne affecting the face and trunk respectively, revealed the mutation of interest, suggesting an autosomal dominant disease with variable penetrance (see online supplementary figure S2).
Supplementary file 1
PAAND family has a cytokine profile distinct from FMF patients
Serum cytokine analysis of the proband, mother and brother revealed a unique profile when compared with FMF patients (n=5) and healthy controls (n=7), highlighted on a heat map of relative values (figure 2A). The increased serum IL-18 was explored further with the measurement of IL-18 binding protein (IL-18BP). IL-18BP has a high affinity for IL-18, and renders it biologically inactive.18 Free IL-18, rather than total, correlates better with disease activity in IL-18-driven conditions, such as haemophagocytic lymphohistiocytosis.19 Interestingly, in our PAAND patients, IL-18BP was significantly elevated when compared with healthy controls, and the ratio of total IL-18 to IL-18BP was similar (figure 2B). Therefore, analysis of free IL-18 revealed no significant increase (data not shown).
When activated, most inflammasome forming proteins, including pyrin, associate with the adaptor protein ASC to form a platform for procaspase-1 activation and cleavage of pro-IL-1β and pro-IL-18 to the mature forms.20 21 Monocytes isolated from PAAND patients showed increased ASC speck formation with LPS exposure, and there was a trend toward an increase at baseline (figure 2C). Caspase-1 activity as measured by YVAD-FLICA staining was increased in PAAND monocytes when treated with LPS (figure 2D). Given these results, it was surprising that IL-1β production from PAAND patient PBMCs in response to LPS was unaltered (figure 2E). Nevertheless, the total IL-18 secreted by PBMCs was increased compared with healthy controls, both at baseline and following LPS stimulation, as were levels of IL-1Ra (figure 2F,G).
p.E244K pyrin is associated with increased ASC speck formation
To determine whether the above results were indeed caused by the novel p.E244K pyrin mutation, we assessed ASC speck formation in vitro as a surrogate marker for inflammasome formation. Colocalisation experiments were performed by expression of both mCherry-tagged pyrin and GFP-ASC in HEK293T cells (figure 3A). There was minimal spontaneous ASC speck formation. As expected, WT pyrin augmented this response, but p.E244K pyrin did so further (figure 3B). This was quantified using flow cytometry (see online supplementary Figure S3), with p.E244K pyrin resulting in a similar percentage of cells with ASC speck formation compared with the other known PAAND mutation p.S242R, both of which were greater than WT and p.M694V pyrin (figure 3C,D).
p.E244K pyrin is associated with increased IL-1β, IL-18 and pyroptosis
Further functional studies were performed using THP-1 monocytes. MEFV KO or CASP1 KO THP-1 cells were reconstituted with MEFV using lentiviral transduction of WT or mutant cDNA. Even without stimulation, MEFV KO THP-1 cells expressing p.E244K pyrin displayed increased cell death (figure 4A), as well as IL-1β and IL-18 release (figure 4B,C). Surprisingly, this phenotype was present without ‘priming’ the inflammasome, which is usually required to induce pro-IL-1β expression.22 Interestingly, IL-1β production in both p.E244K and p.S242R pyrin-expressing MEFV KO THP-1 cells was significantly higher than cells expressing FMF associated p.M694V pyrin (figure 4C). Genetic deletion of caspase-1 prevented p.E244K and p.S242R pyrin-induced cytokine production as well as cell death, suggesting the caspase-1 dependent inflammatory cell death (pyroptosis) (figure 4A–C). However, genetic deletion of caspase-1 did not affect Pam3CSK4-induced priming of pro-IL-1β (figure 4D). These in vitro data support the hypothesis that inflammasome activation in p.E244K pyrin patients is responsible for excessive cytokine release and pyroptosis.
p.E244K pyrin does not alter PSTPIP1 binding
In PAPA syndrome, mutant PSTPIP1 is hyperphosphorylated and binds more strongly to pyrin.23 Given the clinical similarities between PAAND and PAPA syndrome, binding of PSTPIP1 to pyrin with and without PAAND mutations was assessed. Both GST-pyrin and PSTPIP1 were transfected into HEK293T cells and GST-immunoprecipitation performed. When comparing the binding of WT PSTPIP1 to WT, PAAND and FMF associated pyrin, no significant difference was observed. This suggests that the mechanism of this disease is not related to increased PSTPIP1 binding (see online supplementary figure S4).
p.E244K pyrin has reduced phosphorylation of 14-3-3 binding motif and reduced 14-3-3 binding
The initial report of PAAND showed that the mechanism of increased inflammasome activation was loss of 14-3-3 binding to pyrin and subsequent loss of autoinhibition.1 Given that p.E244 is the +2 position of a 14-3-3 binding motif (figure 1D), preliminary experiments were conducted to examine 14-3-3 binding to p.E244K pyrin. Serine residues at positions p.S208 and p.S242 have previously been shown to interact with 14-3-31,2 and were used as comparators. Immunoprecipitation was performed using GST-tagged WT and mutant pyrin transfected into HEK293T cells. This revealed reduced binding of an antibody that recognises phosphorylated serine in the 14-3-3 binding motif in mutants p.E244K and p.S242R pyrin, but not in the FMF-associated p.M649V pyrin (figure 5A). Binding of 14-3-3 to pyrin was also affected, following the same pattern. Further evaluation of binding of the 14-3-3τ and 14-3-3ε isoforms to these mutants, as well as p.S208A and p.S208A/S242R pyrin, showed no differences, suggesting both isoforms behave similarly (figure 5B). These data suggest that PAAND pyrin mutations result in reduced phosphorylation of the 14-3-3 binding motif and reduced 14-3-3 binding to pyrin.
The p.E244 position is important in 14-3-3 binding to pyrin
Phosphorylated serine in specific motifs is important for 14-3-3 binding. Previous reports had suggested that proline was required at the +2 position of the motif for interaction between 14-3-3s and their target protein, documented as RXX(pS)XP (figure 1D). However, subsequent reports show that proline in +2 position is present in only 50% of 14-3-3 binding motifs.24 To explore the importance of the +2 position in 14-3-3 binding and pyrin regulation, p.E244 pyrin was mutated to various amino acids. Glutamate (E) was substituted by aspartate (D) or arginine (R) to explore charge and polarity, respectively, or proline (P) to explore the effect of the canonical 14-3-3 binding motif. Flow cytometric analysis of these mutants showed an increase in ASC speck formation in p.E244R pyrin mutant, while p.E244D and p.E244P pyrin mutations did not further activate pyrin in this assay (figure 6A). Immunoprecipitation showed that p.E244R pyrin had reduced binding to 14-3-3 when compared with WT, similar to p.E244K (figure 6B). Interestingly, p.E244P pyrin had increased 14-3-3 binding, suggesting that this mutation could potentially suppress pyrin activation. To test this hypothesis, cells were treated with the RhoGTPase inhibitor, TcdB, to activate pyrin. Although p.E244P increased binding of 14-3-3 to pyrin, this was insufficient to prevent activation by TcdB (figure 6C). Furthermore, the double mutant p.E244P/M694V had no effect on this, highlighting again a distinct pathophysiological mechanism of FMF and PAAND (figure 6C).
The initial clinical suspicion of PAPA syndrome in the index patient highlights the striking clinical overlap between PAAND and PAPA syndrome, as noted in the original description of PAAND.1 Compared with the initial report, our family is distinct in suffering from polyarthritis as well as severe hidradenitis suppurativa, suggesting that even within the PAAND diagnosis, there is variability in clinical presentation, consistent with a spectrum of pyrin-associated features. Our results agree with the original description of PAAND, namely that excessive IL-1β is pyrin dependent. Although PAPA syndrome is also pyrin dependent,23 the exact mechanisms underlying the similar clinical presentations of PAAND and PAPA syndrome have not been elucidated. We suggest that patients with clinically suspected PAPA syndrome who test negative for PSTPIP1 mutations should undergo genetic evaluation of MEFV, with particular attention to the bases in exon 2 encoding 14-3-3 binding motifs.
The role of 14-3-3 in controlling the activation of pyrin is highlighted by this novel mutation causing PAAND. Reduced binding of 14-3-3 to pyrin was seen with both p.E244K and p.S242R pyrin, but not in the FMF-associated p.M694V mutation (figure 5). The loss of 14-3-3 binding following stimulation with TcdB suggests that 14-3-3 is required to maintain pyrin in an auto-inhibited state and reduced 14-3-3 binding to PAAND-associated pyrin leads to its auto-activation. We propose that with the same expression of pyrin across the mutants examined in our model, the PAAND pyrin is likely to be more active, with increased pyroptosis and availability of pro-IL-1 for cleavage. It is possible that PAAND is at one spectrum of pyrin-associated disorders in terms of severity, with PAAND pyrin being spontaneously active and FMF pyrin having a lower threshold for activation than WT pyrin.
The 14-3-3 binding motif of pyrin differs from the canonical RXX(pS)XP motif with a highly conserved glutamate at the +2 position. Substituting glutamate for proline or aspartate, non-polar and negatively charged amino acids, respectively, retained 14-3-3 binding to pyrin, whereas substitutions to lysine or arginine, both positively charged amino acids, do not appear to be tolerated. The structure of this region of pyrin has not been elucidated, making it difficult to predict the effect of amino acid substitutions. However, we demonstrate that the +2 position of the 14-3-3 binding motif is important, and that substitution at this site can alter the ability for 14-3-3 to bind to pyrin.
Although M694V pyrin results in increased inflammasome formation25, the mechanism of auto-activation still remains to be elucidated. We saw no discernible difference between WT and p.M694V pyrin with regards to 14-3-3 binding, and Van Gorp et al documented unaltered phosphorylation at position p.S242 in p.M694V pyrin transfected HEK293T cells, which is required for 14-3-3 binding.26 Interestingly, Park et al did see reduced 14-3-3ε binding in FMF-associated mutations.5 It is possible that subtle differences in the experimental approach may influence this result, and given that PAAND is a more severe disease, we would expect FMF mutations to have a smaller mechanistic effect on 14-3-3 binding.
In addition to 14-3-3 binding, the clinical presentation, mode of inheritance and biochemical status of PAAND differ from FMF. Although this study focuses on only two generations of one family, the heterozygous mutation and variable phenotype suggest a dominant disorder with variable penetrance, compared with the typically autosomal recessive inheritance of FMF. All members of the PAAND family have marked dermatological manifestations, further differentiating this condition from FMF.
Another distinction between these pyrin-associated conditions is evident from the serum cytokine profile, and cytokine production by PBMCs, both at baseline and after LPS-priming. Although the FMF patients were asymptomatic, there was evidence of systemic inflammation with raised CRP in four of the five controls (see online supplementary table S1). Furthermore, their serum cytokine profile was distinct from healthy controls as well as PAAND patients, suggesting that there are indeed differences that are not accounted for by symptom control.
Despite elevated IL-1β in these analyses, one patient with the p.E244K mutation did not improve with a trial of anakinra. Interestingly, the elevated IL-1Ra levels in this individual may explain why a recombinant IL-1Ra did not provide further benefit. Our FMF patients did not have elevated IL-1Ra levels, and a number of recent publications suggest that colchicine-resistant FMF can be adequately treated with IL-1 antagonism.27–29 Despite elevated IL-18 levels in PAAND PBMC secreted at baseline and in response to LPS (figure 2F), an increase in IL-18BP levels (figure 2B) suggests that targeting this pathway may not be as effective as shown for patients with activation of the Nod-Like Receptor CARD containing protein 4 (NLRC4) inflammasome.30 The clinical response to TNF inhibition in our patient suggests that this is an important cytokine in PAAND, even though TNF was not elevated in the serum of these patients. This may be because at the time of the study, the patient was receiving treatment with immunomodulatory drugs including adalimumab. Alternatively, increased cell death in PAAND (figure 4A) could release damage-associated molecular patterns that trigger local cytokine production in tissues such as the skin. Furthermore, it would be interesting to assess tissue specific cytokines and cell responses as these may reveal pathogenic factors not present in peripheral blood. Regardless, given the difficulty controlling disease activity and the need for multiple therapeutic agents, PAAND is likely to be driven by more than a single cytokine.
The p.E244K pyrin mutation in PAAND patients highlights the importance of the 14-3-3 binding motif in pyrin activation, in addition to the p.S242R mutation described originally. Our study suggests that although PAAND and FMF mutations are located in the same gene, they are distinct diseases clinically, with unique cytokine profiles, cellular responses and 14-3-3 binding.
We thank the patient and family for participation in this study.
JIA and SLM contributed equally.
Handling editor Tore K Kvien
Contributors FM, JIA, SLM: Conception and design of the work. FM, RL, DDN, HMB, JJMG, PMdeC, PJB, VG, AMV, SC, IPW, PP, JIA, SLM: Performed experiments, data collection, analysis and interpretation. All authors were involved in drafting and approval of the manuscript.
Funding IPW: Australian National Health andMedical Research Council Program (NHMRC) project grant (1113577). PP: Grants from Instituto Salud CarlosIII-FEDER (PS13/00174) and European Research Council (ERC-2013-CoG 614578). PP would like to acknowledge networking support by the COST Action BM-1406. HMB: Rio Hortega fellowship from Instituto Salud Carlos III (CM14/00008), JIA: CERCA Programme/Generalitat de Catalunya and SAF2015-68472-C2-1-R grant from the Spanish Ministry of Economy and Competitiveness and Fondo Europeo de Desarrollo Regional (FEDER). SLM: NHMRC projects grants (1099262, 1081299) Viertel Fellowship and funding from Glaxosmithkline.
Competing interests None declared.
Ethics approval Hospital Clinic-IDIBAPS Ethics Committee.
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.