Objectives Porphyromonas gingivalis (P.g.) is discussed to be involved in triggering self-reactive immune responses. The aim of this study was to investigate the autocitrullinated prokaryotic peptidylarginine deiminase (PPAD) from P.g. CH2007 (RACH2007-PPAD) from a rheumatoid arthritis (RA) patient and a synthetic citrullinated PPAD peptide (CPP) containing the main autocitrullination site as potential targets for antibody reactivity in RA and to analyse the possibility of citrullinating native human proteins by PPAD in the context of RA.
Methods Recombinant RACH2007-PPAD was cloned and expressed in Escherichia coli. Purified RACH2007-PPAD and its enzymatic activity was analysed using two-dimensional electrophoresis, mass spectrometry, immunoblot and ELISA. Autoantibody response to different modified proteins and peptides was recorded and bioinformatically evaluated.
Results RACH2007-PPAD was capable to citrullinate major RA autoantigens, such as fibrinogen, vimentin, hnRNP-A2/B1, histone H1 and multiple peptides, which identify a common RG/RGG consensus motif. 33% of RA patients (n=30) revealed increased reactivity for α-cit-RACH2007-PPAD before RA onset. 77% of RA patients (n=99) presented α-cit-specific signals to CPP amino acids 57–71 which were positively correlated to α-CCP2 antibody levels. Interestingly, 48% of the α-CPP-positives were rheumatoidfactor IgM/anti-citrullinated peptide/protein antibodies (ACPA)-negative. Anti-CPP and α-RACH2007-PPAD antibody levels increase with age. Protein macroarrays that were citrullinated by RACH2007-PPAD and screened with RA patient sera (n=6) and controls (n=4) uncovered 16 RACH2007-PPAD citrullinated RA autoantigens and 9 autoantigens associated with lung diseases. We showed that the α-CPP response could be an important determinant in parenchymal changes in the lung at the time of RA diagnosis (n=106; p=0.018).
Conclusions RACH2007-PPAD induced internal citrullination of major RA autoantigens. Anti-RACH2007-PPAD correlates with ACPA levels and interstitial lung disease autoantigen reactivity, supporting an infection-based concept for induction of ACPAs via enzymatic mimicry.
- early rheumatoid arthritis
Statistics from Altmetric.com
What is already known about this subject?
Rheumatoid arthritis (RA) is one of the diseases for which the Porphyromonas gingivalis (P.g.) may have an important role in pathology. P.g. has been implied to lead immune abnormality in RA patients such as the activation of bacterial citrullination, anti-citrullinated peptide/protein antibodies (ACPAs) and immune responses via lung infection.
What does this study add?
RA-prokaryotic peptidylarginine deiminase (PPAD) were related to RA aetiology. Anti-RA-PPAD correlated with ACPA levels, and ILD autoantigens supporting an infection-based concept of induction of ACPAs via enzymatic mimicry.
PPAD was detected in synovial tissue and citrullinated internal arginines of major RA autoantigens such as fibrinogen, vimentin, hnRNP-A2/B1, histone H1 and multiple peptides, identifying a common RG/RGG consensus motif.
The failure of P.g. clearance induces bacterial citrullinated epitope (BCE)-ACPAs. This mechanism might be crucial to induce the general ACPA autoimmunity.
Anti-RA-PPAD correlated with ACPA levels, and interstitial lung disease (ILD) autoantigens supporting an infection-based concept of induction of ACPAs via enzymatic mimicry.
Anti-cit-specific α-RA-PPAD peptide response is a major determinant in parenchymal changes in the lung at the time of RA diagnosis.
How might this impact on clinical practice or future developments?
We revealed a novel link between the PPAD modification and pathology of RA. P.g. treatment with antibiotics could be considered in standard routine to prevent ILD at onset of RA. Our study will be a platform model of bacterial citrullinated epitopes to elucidate aetiology of rheumatic diseases.
Anti-citrullinated peptide/protein antibodies (ACPAs) to human autoantigens are highly specific for rheumatoid arthritis (RA) and able to predict the onset and severity of clinical disease.1–3 The human peptidylarginine deiminases (hPADs) can citrullinate different autoantigens such as vimentin, fibrinogen, α-enolase, histones in the synovial tissue, which act as targets for seroreactivity in RA.4–9 In early RA, 40%–50% of patients are ACPA-seronegative.10–12 ACPAs are highly cross-reactive and they bind a wide variety of citrullinated proteins, but the relevant target antigen in vivo and its tissue localisation is unknown.13 Porphyromonas gingivalis (P.g.), the main pathogen in periodontal disease (PD), is unique in prokaryotic PAD (PPAD) expression.14–16 It has been shown that P.g. may infect lung, brain, oesophagus, placenta, synovial fluid and it penetrates intracellularly into macrophages, epithelial, endothelial and smooth muscle cells. In this context, P.g. can suppress autophagic destruction within dendritic cells and migrates from oral cavity to distant organs.17–21 PPAD is genetically unrelated to hPADs but it performs the same reaction with a different catalytic mechanism. It can citrullinate preferentially C-terminal peptidylarginine and free L-arginine without any cofactor, in contrast to the hPADs that deiminate internal arginine residues and are calcium-dependent.15 22–26 Internally citrullinated arginines have only been described for bacterial proteins, human histone H327 and autocitrullinated PPAD in P.g.28 29
There is a potential connection between PD and RA as there are parallels in pathogenesis, clinical parameters, therapies and ACPA intensities.5–9 30 Significantly increased α-PPAD antibody response, especially in sera from RA patients, indicates that PPAD itself could be an antigen relevant to RA pathogenesis.28 29 Anti-P.g. antibodies in PD are associated with RA-related autoantibodies. RA-associated risks include cigarette smoke and high prevalence of lung disease including airway inflammation, as identified in established RA. The prevalence of inflammatory airway disease is associated with RA-related autoantibodies.20 Although P.g. DNA was found in whole blood, plasma, serum, synovial fluid, synovial tissue and dental plaques from RA patients, the exact mechanism of P.g. migration to the joints remains unclear.21 31–33
The specificity of immune responses to citrullinated PPAD was confirmed by the reactivity of RA sera to multiple synthetic citrullinated peptides.29 Different mechanisms were hypothesised in breaking immune tolerance to citrullinated proteins, one of them that PPAD might directly contribute by citrullinating bacterial and human proteins and leading to ACPA response.4 34 No direct link of P.g. infection to the citrullination of human autoantigens and ACPA response could be studied since C-terminal citrullination is not immunoreactive in RA patients.35
Several full-length and truncated PPAD variants (flPPAD/tPPAD) have been published with various enzyme characteristics. Rodríguez et al 26 showed active flPPAD (amino acid (aa) 556) and N-terminal tPPAD (aa513), where only the truncated protein version was stable and able to autocitrullinate. Quirke et al 29 verified a flPPAD in P.g. culture fractions and outer membrane vesicles (OMVs) and produced a recombinant flPPAD that was able to autocitrullinate. Konig et al 28 identified 47 kDa and 75–85 kDa cellular forms of tPPAD and purified recombinant tPPAD with increased enzymatic activity. Depending on the PPAD variants used, in combination with different methods, heterogeneous results for PPAD autocitrullination were reported. The skewing of potential human autoantigens modified by PPAD raises questions about the role of PPAD in the RA pathogenesis. Is autoantigen citrullination by P.g. random or substrate-specific to modulate the immune system, which can lead to inflammation in the joint and possibly also in other tissues?
In this study, we showed that PPAD from an RA patient (RACH2007-PPAD) may have pathogenic potential in RA. Citrullination of internal arginines by RACH2007-PPAD, produced at the same protein sequence site as hPADs, leads to cross-reactive epitopes and may become an ACPA target.
Our findings supported that ACPAs against a citrullinated PPAD peptide can be specifically associated with interstitial lung disease (ILD) in early RA. Specifically, we examined potentially new ACPA epitopes generated by a bacterial RACH2007-PPAD support lung involvement in RA in the context of P.g. infections.
Material and methods
Cloning, expression and purification of RACH2007-PPAD
A P.g. CH2007 strain was isolated from the saliva of an RA patient. The N-terminal truncated and enzymatically active form of RACH2007-PPAD aa44–556 according to Rodríguez et al 26 was cloned and expressed as a recombinant poly-His-tagged fusion protein in Escherichia coli (figure 1A). The mass spectrometric (MS) and DNA sequence analysis revealed two undescribed mutations of RACH2007-PPAD in comparison with wild-type PPAD (WT PPAD) of P.g. W8316 (figure 1A, §; online supplementary S4.1/4.2). At the beginning of the catalytic domain at aa30, corresponding to aa73 of WT PPAD, RACH2007-PPAD exhibits phenylalanine (F) instead of leucine (L). Furthermore, in the immunoglobulin superfamily region at aa404, corresponding to aa447 of WT PPAD, glutamic acid (E) is mutated to valine (V).
Localisation of RACH2007-PPAD in P.g.
To determine the intracellular localisation of RACH2007-PPAD, RA-P.g. culture was fractionated into supernatant, whole cell pellet and vesicle pellet and analysed by western blot using an α-PPAD antibody (figure 1B–D). Merely a ~70 kDa variant was detectable in the culture supernatant (figure 1B), which corresponds to the predicted size of the full-length form of WT PPAD (P.g. W83).36 The extract prepared from whole cell pellet predominantly contained the 70 kDa variant and several smaller PPAD degradation products (figure 1C). In the vesicle fraction dominated a truncated variant of ~60 kDa (figure 1D). Enzymatic activity was detected in the cytoplasmic fraction using an antibody-based assay for PAD enzyme activity (ABAP; supplementary data figure 1 S1). In supernatant and whole cell fraction, citrullination activity was detected. The vesicle pellet fraction contained citrullinated and enzymatically inactive RACH2007-PPAD (figure 1D and supplementary results figure S1). PPAD localisation was not specific for RA using a specific α-PPAD peptide antibody to detect different arthritic synovial tissues (online supplementary results figure S2).
Autocitrullination of RACH2007-PPAD
PADs themselves can be subject to deimination of internal arginine by autocitrullination.28 29 37–39 This occurred over time at room temperature (RT) with loss of RACH2007-PPAD activity (online supplementary results figure S3D) and confirmed data by Rodríguez et al.26 Immediately after purification, citrullinated RACH2007-PPAD was absent as determined by immunoblotting (figure 2A, α-cit ab I, day 0). Additionally, in a two-dimensional gel separation we could detect a pH-dependent shift of RACH2007-PPAD to acidic pH after incubation at RT for 3 weeks, indicating that positively charged arginine residues had been converted to neutral citrulline (figure 2B). Furthermore, autocitrullination of RACH2007-PPAD was confirmed by MS analysis (online supplementary results figure S4.1). One autocitrullination site could be detected at R20, corresponding to R63 in WT PPAD, which has previously been shown to be citrullinated28 (figure 1A, *). Arginine citrullination by hPADs and by RACH2007-PPAD was analysed with ABAP under varying conditions (online supplementary results figure S3A-E). RACH2007-PPAD activity was calcium-independent over a wide pH range (pH 3–10) and its enzymatic activity could be preserved by storing RACH2007-PPAD at −80°C (online supplementary results figure S3C/E).
Major human autoantigens can be citrullinated by RACH2007-PPAD
So far, internal citrullination by PPAD was described only for one site in histone H3.27 To identify new citrullination sites, recombinant RACH2007-PPAD was incubated in vitro with recombinant vimentin (Vim) and fibrinogen α (FibA), and citrullinated peptides were analysed by MS (table 1). Additionally, RACH2007-PPAD citrullinated FibA and Vim were detected by monoclonal α-cit-antibodies and RA sera by western blot (online supplementary results figure S5 A/B). RACH2007-PPAD was also able to citrullinate different proteins from HeLa-extract (>170, 60, 55, 50 and 46 kDa) (online supplementary results figure S5C). Three of these specific citrullinated proteins (>170, 60 and 55 kDa) were also recognised by RA patient sera. RACH2007-PPAD citrullinated in vitro RA autoantigens hnRNP-A2/B1 and histone H1 (online supplementary results figure S5D). Significant citrullination was detected by ELISA using a monoclonal α-cit antibody.
Antibodies against autocitrullinated RACH2007-PPAD and citrullinated vimentin are biomarkers in early RA
Several studies reported increasing levels of ACPA and rheumatoid factor (RF) shortly before or around onset of RA.3 40 41 Therefore, we analysed sera of early RA patients obtained before and after onset of disease by ELISA using vimentin as an antigen, which we previously citrullinated with RACH2007-PPAD, and detected a significantly higher antibody level in RA compared with normal donors (ND) (figure 3A; n=20; median=0; before/after onset p<0.0001). No significant difference could be detected before and after RA onset (figure 3A). Similarly, vimentin that was citrullinated with rabbit PAD reacted with antibodies in 40% of the RA patients before and 37% after or during onset (online supplementary figure S6) using non-citrullinated RACH2007-PPAD or citrullinated RACH2007-PPAD, respectively (figure 3B/C). Reactivity against citrullinated RACH2007-PPAD showed a sensitivity of 33% before and 33% after RA onset with 100% specificity compared with self-reported ND (figure 3C), whereas the non-cit-RACH2007-PPAD ELISA showed no specificity for RA (figure 3B). For α-RACH2007-PPAD or α-cit-RACH2007-PPAD, there was a shift of the antibody reactivity in approximately 30% of the patients around RA onset (figure 3B/C), and in 20% of RA patients we identified an increased optical density (OD) for α-cit-RACH2007-PPAD after RA onset (figure 3C). Also, higher levels of α-cit RACH2007-PPAD antibodies were weakly correlated with age in early RA patients (n=30; p=0.048) after RA onset (figure 3E).
Autocitrullinated RACH2007-PPAD is also recognised by sera of established RA patients
Anti-cit-RACH2007-PPAD IgG levels are significantly elevated in patients with established RA (n=36, median OD=0.15) compared with systemic lupus erythematosus (SLE) patients (n=30, median OD=0; p=0.018), osteoarthritis (OA) patients (median OD=0.005; p=0.02) and ND (n=23, median OD=0.004; p=0.045) (figure 3D). Thus, 38% of the RA patients were α-cit-RACH2007-PPAD ELISA-positive whereas none of the control patients or ND showed levels above the cut-off value. The single ODs of α-cit-RACH2007-PPAD and α-RACH2007-PPAD are presented in online supplementary results figure S10.
RA sera recognise the major autocitrullinated site in synthetic citrullinated PPAD peptide
Since antibodies against the isolated form of citrullinated RA-CH2007-PPAD from an RA patient were highly specific for RA, we analysed the main citrullinated site of RACH2007-PPAD in the in vitro synthesised CPP aa57–71. We investigated the sensitivity in two early and one established RA cohort. In the established RA cohort (n=99), we detected α-CPP antibody levels in 76% of the patients, α-PP (arginine-containing peptide) in 20% and ∆OD between α-CPP and α-PP (α-net-CPP) in 77% but not in ND (n=32) (figure 4A). The sensitivity of α-CPP is nearly doubled in comparison with α-cit-RACH2007-PPAD. The single citrullinated sites exhibit a lower background signal in ELISA. Furthermore, higher levels of α-CPP antibodies weakly correlated with age in these RA patients (n=91, p=0.0327) (figure 4B). In the established RA cohort positive α-net-CPP were found in 48% of the RF IgM/α-CCP2-negative (seronegative), 33% of the RF IgM negative and 50% of α-CCP2-negative RA patients. In seropositive patients, 72% were reactive, 69% in α-CCP2-positive and 80% in RF IgM positive (figure 4C–E). In this established RA cohort, we also found significant correlations between α-CPP antibody levels and α-CCP2, RF IgM, RF IgA and α-cit-RACH2007-PPAD antibody levels. Similarly, significant correlations were observed between α-PP antibody levels and α-CCP2, RF IgM, RF IgA antibody levels (online supplementary results table S1) and count of tender joints. Antibody levels against α-net-CPP were also positively correlated to α-CCP2 (online supplementary results table S1 and figure S9A/B). We also tested 106 early RA and 24 treatment-naive RA patients (n=130) and identified all α-CCP-2 positive (n=11) and all α-CCP2-negative (n=13) treatment-naive RA patients. No reactivity was detected in SLE (n=20), PsA (n=21), OA (n=20) and reactive arthritis patients (n=7) with one SLE serum as an exception (figure 5A). The pairwise ODs of RA cohorts, disease and healthy controls for α-CPP and α-PP are presented in online supplementary results figure S7/8.
Early RA with parenchymal HRCT changes are associated with antibodies against the cit-PPAD peptide
Structural lung changes and ACPA enrichment in the lung are characteristic hallmarks of early RA.42 To investigate the ACPA response to a net-CPP, the Lung Investigation in Newly Diagnosed RA (LURA) study was examined. The α-net-CPP signal could be detected in 31% of the LURA cohort (n=106, α-CCP2-negative (n=35) and α-CCP2-positive (n=71) part of early RA (n=130)) (figure 5A). In a statistical analysis, parenchymal changes tend to be associated with (p=0.018) the α-net-CPP and interestingly airway changes were associated with the presence of α-PP (aa121–136; p=0.0158), which is an additional arginine epitope of PPAD.
RA patient sera identify specific RACH2007-PPAD citrullinated autoantigens homologous to bacterial P.g. proteins
The consensus sequence of RACH2007-PPAD modification sites identified by MS analysis (autocitrullinated RACH2007-PPAD, cit-vim and cit-fib) and protein microarray analysis identified RG/RGG regions as major target sites for RACH2007-PPAD (figure 5B). We determined reactivity against 458 RA autoantigens, which included 43 previously described major autoantigens, by using protein macroarrays in native form and after citrullination with RACH2007-PPAD. We analysed sera of six RA patients (CCP-positive n=3, CCP-negative n=3) and four OA control patients. Sixteen antigens were detected only when citrullinated by RACH2007-PPAD. Interestingly, these detected antigens were previously published as autoantigens in different diseases (figure 5C, online supplementary results figure S11, references in online supplementary data excel table 1). Eight proteins are citrullinated autoantigens in RA, five of these are specifically targeted in ILD patients and heat shock protein 90 (HSP90) is specifically targeted in RA with ILD.43 Nine autoantigens are highly expressed in the lung and five (FGB, FGG, SRP14, HARS and HSP90) are previously identified as autoantigens in ILD and arthritis patients, where HARS and HSP90 are homologous to P.g. proteins. Most of these antigens are found in P.g.-infected tissue (lung, gingiva, brain, colon) and carry the RG/RGG recognition site for citrullination. Conserved homologous regions of HSP70, HSP90, H2AFY2, GAPDH and HARS are also expressed in P.g. (online supplementary results, table S2).
A link between microbiome and RA has been discussed for years and proposed for P.g., but there is still no clear evidence.14 44 P.g., the main pathogen of PD, appears to be involved in the RA initiation and progression. It is the only known bacterium expressing a PPAD capable of inducing C-terminal citrullination of peptides derived from fibrinogen and enolase.45–47 We showed that RACH2007-PPAD causes not only C-terminal but also internal citrullination of arginines in vimentin and fibrinogen, which is so far known only for histone H3.27 This is confirmed by our MS data and other human proteins citrullinated by RACH2007-PPAD on protein macroarrays, which indicate an overlapping consensus sequence RG/RGG for citrullination by hPAD4 (table 1/figure 5B/C). Thereby, the regulation of protein aggregation and the inhibition of RGG methylation sites could also be influenced by P.g. RG/RGG was found previously citrullinated by hPAD4 regulating the protein aggregation and inhibiting RGG methylation sites.48 ACPAs in RA are associated with HLA class II histocompatibility antigen, DRB1 beta chain shared epitope (HLA-DRB1SE) and they recognise many citrullinated autoantigens, for example cit71vim, which was previously found in lung tissue.49 50 We could identified 71vim as citrullinated by RACH2007-PPAD and it was also specifically detected in our investigations by sera before onset of early RA patients (table 1/figure 3A). This supports the enzymatic mimicry hypothesis that P.g. infection and PPAD citrullination of human antigens can induce an ACPA response. Fibrinogen was previously found citrullinated in synovial tissue.5 The type of PPAD-mediated citrullination of fibrinogen that we have observed (table 1/online supplementary results figure S5A) has the potential to destroy the inhibitory effect of native fibrinogen on bacterial co-aggregation and can thus stabilise the microbial biofilm so that infection can spread further.51 OMVs have been demonstrated to co-aggregate many species of bacteria to each other and enhance the attachment of other pathogens to epithelial cells.52–54
Molecular mimicry and cross-reactivity of ACPAs, which can react with different citrulline epitopes after processes of epitope spreading, have also been shown in connection with human α-enolase.8 55 Our protein macroarray analysis identified 16 RACH2007-PPAD citrullinated autoantigens, including human enzymes, HSPs, histones, hnRNPs and vimentin (figure 5C), all with P.g. homologies (online supplementary data excel table 2; query coverage 21%–87%). When both the human and the homologous foreign antigen are presented by antigen-specific B cells, T-cell priming could be achieved.12 56 In contrast to hPADs, RACH2007-PPAD does not need a cofactor and is enzymatically active (online supplementary results figure S3) even in inflamed environment like synovial fluid and could citrullinate calreticulin (citCRT), which we confirmed by our protein macroarray data (figure 5C). CitCRT is known to bind to RA shared epitopes of HLA-DR molecules and triggers pro-inflammatory events in adjacent cells. CRT has been described as an autoantigen in bronchiectasis patients prior to onset of RA, suggesting citCRT is an early PPAD target in RA.57
Anti-P.g. levels were higher in RA patients and higher in ACPA-positive patients. This supports the assumption that α-P.g. and ACPA immune response is involved in breaking immune tolerance to citrullinated self-antigens, a phenomenon that occurs in the pathogenesis of RA.58 We detected in 33% of RA patients an α-cit-PPAD reactivity before onset of RA (figure 3B) and more than three-quarters of established RA patients produce ACPA response to CPP (figure 4A). Almost half of RF IgM/α-CCP2-negative RA patients also recognise the BCE CPP within RACH2007-PPAD (figure 4E). RACH2007-PPAD can citrullinate both human and bacterial epitopes (table 1/online supplementary results figure S4.1/table S2). The citrullinated epitope can be the same for PPAD and hPADs, but protein macroarray data suggest additional epitopes in human proteins citrullinated by PPAD (figure 5B/C). This indicates that molecular and enzymatic mimicry may exist for protein modification by citrullination in bacteria and in humans.
In addition to smoking, which explains 35% of the risk for RF IgM/α-CCP2-negative RA, P.g. infection might trigger the autoimmune response in the lung via citrullination of the host proteins, when P.g. migrates from the gum to the lung and induces pathological damage.50 We identified RACH2007-PPAD citrullinated autoantigens recognised by RA patients, which are highly expressed in the lung and identified as autoantigens in arthritis patients with ILD (figure 5C/ online supplementary results figure S11/online supplementary data excel table 1).59 Anti-cit-HSP90 antibodies that we detected are highly specific for ILD in RA.43 60 Citrullination of antibacterial human proteins like histones and HSPs (figure 5C) might be of great importance for the local survival of P.g. Autoantibodies to antibacterial proteins may induce a failure of bacterial clearance in the lung of RA patients. Over time, inflammatory and autoimmune-mediated lung damage may occur due to the spread of P.g. from the lung to other regions of the body as suggested.42 A P.g. infection from the airways to the surrounding parenchymal structures may induce the citrullination of fibrinogen and vimentin, as shown in the sputum of high-risk patients.61 Furthermore, a link to microbial triggers is also suggested by repeatedly reported microbial DNA or antigen fragments in joints of RA and even of OA patients33 62 63 and by activated macrophages with patterns of microbial stimulation in synovial tissues of long-standing RA patients.64
Our findings supported the hypothesis of possible chronic triggering by microbes like P.g., which are capable to citrullinate and produce exogenously citrullinated human and bacterial epitopes and thus induce ACPA development, which may contribute to RA onset and progression. P.g. prophylaxis and treatment with antibiotics, when the RA patient is infected, should be standard in RA routine therapy.
We thank Prof Lars Klareskog for providing mcl α-citrulline antibodies 1325:01BO9 and 1325:04CO3.
Handling editor Josef S Smolen
Correction notice This article has been corrected since it published Online First. The legend for figure 5 has been corrected and a further affiliation for Zoltán Konthur has been added and the supplementary citation has been corrected in the results section.
Contributors MJ wrote the first draft of the manuscript. All authors involved in drafting the manuscript and approved the final version.
Funding The work was supported by grants from the project ArthroMark (grant number: 01EC1401A).
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Patient consent for publication Obtained.
Ethics approval The study was approved by the ethics committee of the Charité University Hospital Berlin, Germany and by the institutional ethics committees of the participating clinics.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.
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.