Background: Rheumatoid arthritis (RA) has been associated with an increased risk of infections, but the underlying pathways have not yet been identified. Toll-like receptors (TLR) probably play a role in synovial inflammation and may also contribute to the understanding of the role of infections in RA.
Objectives: To investigate if the synovial expression of TLR3 and TLR7 in RA correlates with that of inflammatory cytokines, and to assess whether this has functional consequences for local cytokine production and to study potential links between the TLR3/7 axis and TLR4 in RA synovium.
Methods: Immunohistochemistry was used to study the expression of TLR3, TLR7, interferon α (IFNα), tumour necrosis factor α (TNFα) and interleukins IL1β, IL12, IL17 and IL18 in RA synovium obtained by arthroscopy from 34 patients with RA. Monocytes, monocyte-derived dendritic cells (MoDCs) and RA synovial fibroblasts were stimulated via TLR3 (poly-IC) and TLR7 (loxorubin), after which IL1β, IL6 and TNFα were measured by Luminex bead array technology. Following preincubation with IFNα, IL1β and IL18, TLR3 and TLR7 mRNA expression was assessed using real-time PCR. Cytokine production after preincubation with IFNα and subsequent TLR stimulation was measured.
Results: Synovial TLR3/7 expression was co-expressed with IFNα, IL1β and IL18, but not with TNFα, IL12 and IL17. Stimulation of TLR3/TLR7 on monocytes, MoDCs or synovial fibroblasts led to secretion of type I IFN but no biologically active IL1β or IL18 could be detected. Type I IFNα increased TLR3/7 mRNA expression whereas IL1β and IL18 did not. In spite of the fact that the mRNA level of TLR4 remained unchanged, IFNα enhanced the response to TLR4 agonists, a phenomenon that was clearly more marked in patients with RA.
Conclusion: Type I interferons are highly co-expressed with TLR3/TLR7 in RA synovium. They enhance TLR3/TLR7-mediated cytokine production and also TLR4-mediated responses.
Statistics from Altmetric.com
Rheumatoid arthritis (RA) is a systemic autoimmune disease which is characterised by chronic inflammation of the synovial joints. Although the cause of RA is still unidentified, a role for both genetic and environmental factors has repeatedly been advocated. As environmental factors, it has been suggested that bacteria and/or viruses may trigger autoimmunity in the host, and evidence for the presence of at least some viruses—including cytomegalovirus,1,2 Epstein-Barr virus3 and parvovirus B194—has been demonstrated. In line with these observations, it has been shown that dsRNA, which is a common feature of viruses, exerts clear arthrogenic properties,5 further substantiating the potential role of viruses in RA. Although recent studies have indicated a type I interferon (IFN) signature in the RA synovium, the underlying pathway for the role of type I IFN remains unknown.6
Accumulating evidence points to a role for the Toll-like receptor (TLR) family in the type I IFN-mediated response. TLR belong to the family of pattern recognition receptors, which were first identified to recognise microbial components known as pathogen-associated patterns. TLR are constitutively expressed by numerous immune cells and designed to detect and eliminate invading pathogens by activation of both innate as well as adaptive immune responses. TLR3/7 and 9 serve as receptors for viral nucleic acids, by which means they have a key role in antiviral immunity by inducing type I IFN production.7,8,9 In contrast, TLR2 and TLR4 elicit immune responses on binding of antigens from bacteria and host-derived molecules (so-called endogenous ligands or alarmins), leading to the production of inflammatory mediators including tumour necrosis factor α (TNFα) and interleukin 1β (IL1β).10,11,12,13 The increased expression of various TLR subtypes in the synovial tissue of patients with RA further substantiates the potential role of TLR in RA.
Recently, the role of TLR in arthritis was highlighted in experimental models.14,15 Although TLR and its ligands are abundant in the synovial compartment of patients with RA, it is hard to conceive that any trigger of a single TLR subtype would be sufficient to convert tolerance to immunity. If this is the case, then autoimmunity should regularly follow from infections. It is therefore more likely that more then one TLR ligand is needed for the initiation of a chronic and persisting inflammatory response as is seen in RA. Accordingly, recent research suggested that simultaneous or sequential triggering of different pathways is perhaps needed to set off autoimmunity.16 It is thus tantalising to speculate that simultaneous triggering of different TLR pathways might initiate a series of events that form the basis of the breakthrough of tolerance. Previous evidence from our group demonstrated a high expression of TLR2, 3, 4 and 7 in the synovial tissue from patients with RA and a synergistic effect with regard to cytokine production after stimulation of dendritic cells (DC) with two (or more) TLR subtypes. Here we sought evidence for a potential link between TLR2/4 and TLR3/7 pathways in RA inflammation.
Our study showed that IL1β, IL18 and IFNα (the classic type I interferon) are co-expressed with TLR3/7 in RA synovium. In addition, we provide evidence that type I IFN increases the expression of TLR3 and TLR7 which, on stimulation, lead to an increased production of cytokines. Interestingly, type I IFN also induced a clear augmentation of TLR4-mediated production of pro-inflammatory mediators including TNFα, IL1β and IL18, which was clearly more potent in RA. These data show that TLR3/7-mediated stimulation indirectly lowers the threshold for TLR4-mediated immune activation, setting the stage for the vicious circle of inflammation observed in RA. Together, these observations underscore the potential role of viruses in RA and provide a rationale for interference with TLR signalling in this condition.
For immunohistochemistry, synovial biopsy specimens were obtained from the medial and lateral suprapatellar pouch of 24 patients with RA by small needle arthroscopy. A mean of 30 samples were obtained at each biopsy. For in vitro experiments, heparinised venous blood was collected from 10 patients with RA, 11 healthy volunteers and 3 patients with systemic sclerosis (SSc) of a diffuse cutaneous subtype. All the patients with RA were attending the Department of Rheumatology of the Radboud University Nijmegen Medical Centre; they fulfilled the American College of Rheumatology criteria for RA, gave informed consent and had a disease activity score (DAS) of >3.2, reflecting moderate to high disease activity.17 All the patients with SSc fulfilled the preliminary criteria of the American College of Rheumatology for SSc.18 SSc was classified as either a limited subtype or a diffuse subtype according to the extent of skin involved, as proposed by Leroy et al.19 Patients using high-dose prednisolone (>10 mg/day) and those receiving intra-articular steroids or anti-cytokine therapies (anti-TNFα and/or IL1Ra) were excluded from the study. For the arthroscopy, patients who had a history of bleeding, infectious disorders or were currently pregnant or lactating were excluded.
Immunohistochemical staining of synovial biopsies
Tissue samples were immediately fixed with 4% formaldehyde and embedded in paraffin. Staining of IL1β, IL12, IL17, IL18 and TNFα was performed as described previously.20,21,22 For TLR3, TLR7 and IFNα staining, sections were incubated for 60 min with monoclonal antibodies against human TLR3 (T-17), human TLR7 (V-20) or IFNα (FL-198), which were all obtained from Santa Cruz, California, USA. After this, endogenous peroxidase was blocked with 3% hydrogen peroxide in methanol for 15 min and subsequently the appropriate biotinylated secondary antibody (mouse anti-goat (Jackson ImmunoResearch, West Grove, Pennsylvania, USA)/swine anti-rabbit (DakoCytomation, Glostrup, Denmark)) was incubated for 30 min. For TLR3 and TLR7 staining, Vectastain ABC (Vector Laboratories, Burlingame, California, USA) reagent was incubated for 30 min, developed with diaminobenzidine (Sigma, St Louis, Missouri, USA) and counterstained with haematoxylin for 30 s. For IFNα staining, sections were incubated with streptavidin peroxidase (DakoCytomation), developed with diaminobenzidine and counterstained with haematoxylin for 30 s. Staining was semi-quantitatively scored on a 5-point scale (scores 0–4) at 200× magnification where a score of 0 represents no or minimal staining, 1 represents 10–20% positive cells, 2 represents 30–40%, 3 represents 50–60% and 4 represents staining of >60% of the cells.
Isolation and culturing of monocytes and monocyte-derived dendritic cells (MoDCs)
Peripheral blood mononuclear cells (PBMC) were isolated from heparinised venous blood using density gradient centrifugation over Ficoll-Paque (Amersham Biosciences, Roosendaal, The Netherlands). Low-density cells were collected and washed with citrated phosphate buffered saline/5% fetal calf serum (FCS). For monocytes, the CD14+ cell fraction was isolated using MACS cell separation according to the manufacturer’s instructions. Briefly, PBMC were incubated with anti-human CD14 MicroBeads (Miltenyi Biotec, Bergisch Gladbach, Germany) for 30 min, cells washed and the CD14+ fraction was separated from the CD14− fraction over a MACS MS separation column (Miltenyi Biotec). The CD14+ cell fraction was eluted from the column, washed again and resuspended in a concentration of 0.5×106 cells/ml in RPMI 1640 Dutch modification (Invitrogen Corporation, Carlsbad, California, USA) supplemented with 10% FCS, plated in 6-well plates and cultured overnight at 37°C and 5% carbon dioxide.
For MoDCs, PBMC were allowed to adhere for 1 h at 37°C in RPMI 1640 Dutch modification supplemented with 2% human serum (PAA Laboratories, Pasching, Austria) in 25 cm2 cell culture flasks (Corning, New York, USA). Adherent monocytes were cultured in RPMI-1640 Dutch modification supplemented with 10% FCS and antibiotic-antimycotic (Life Technologies) in the presence of IL4 (500 U/ml, Schering-Plough, Amstelveen, The Netherlands) and granulocyte-macrophage colony stimulating factor (800 U/ml, Schering-Plough) for 6 days. Fresh culture medium with the same supplements was added on day 3. Immature DCs were harvested at day 6, resuspended in fresh cytokine-containing culture medium, transferred to 6-well culture plates (Corning) in a concentration of 0.5×106 cells/ml and cultured for 16 h at 37°C and 5% carbon dioxide.
Isolation and culturing of RA synovial fibroblasts
Immediately after surgery the synovial tissue was minced and digested with dispase at 37°C for 60 min. After washing, the cells were grown in Dulbecco’s minimum essential medium (Gibco Invitrogen, Basel, Switzerland) supplemented with 10% FCS, 50 IU/ml penicillin/streptomycin, 2 mM L-glutamine, 10 mM HEPES and 0.2% Fungizone (all from Gibco Invitrogen). Cell cultures were maintained at 37°C in a humidified incubator in an atmosphere of 5% carbon dioxide. For the experiments, cultured synovial fibroblasts between passages 4 and 8 were grown in 12-well culture plates (6×104 synovial fibroblasts/well) and cultured for 16 h at 37°C and 5% carbon dioxide.
Stimulation of monocytes, MoDCs and RA synovial fibroblasts
To study TLR mRNA expression on cytokine stimulation, after a resting period of 16 h monocytes (which express TLR7), MoDCs and synovial fibroblasts (which both express TLR3) were stimulated for 8 h with TNFα, IL1β, IFNα, IL12, IL17, IL18 (all R&D Systems, Minneapolis, Minnesota, USA). Culture supernatants were removed and 1 ml TRIzol reagent (Sigma) was added to the cells and stored at −20°C until RNA isolation was performed.
To study functional upregulation of TLR, after a resting period of 16 h monocytes, MoDCs and RA synovial fibroblasts were stimulated with IFNα (R&D Systems) for 24 h and subsequently stimulated with the TLR3 and TLR7 agonists poly(I:C) and loxoribin, respectively (both Invivogen, San Diego, USA) or medium. All non-TLR4 ligands were tested for lipopolysaccharide (LPS) contamination using the Lumilus assay and all were negative. After a further 24 h the culture supernatants were collected and stored at −20°C until cytokine measurement was performed. In addition, to investigate the role of type I IFN, various cell types were preincubated with 100 U/ml IFNα for 16 h.
RNA isolation and real-time PCR
Total RNA was extracted in 1 ml TRIzol reagent, an improved single-step RNA isolation method based on the method described by Chomczynski et al.23 Quantitative real-time PCR was performed using the ABI/Prism 7000 sequence detection system (Applied Biosystems, Foster City, California, USA). PCR conditions were as follows: 2 min at 50°C and 10 min at 95°C followed by 40 cycles of 15 s at 95°C and 1 min at 60°C, with data collection in the last 30 s. All PCR were performed with SYBR Green Master mix (Applied Biosystems), 10 ng cDNA and a primer concentration of 300 nmol/l in a total volume of 25 μl. Quantification of the PCR signals was performed by comparing the cycle threshold value (Ct) of the gene of interest of each sample with the Ct values of the reference gene GAPDH (glyceraldehyde 3-phosphate dehydrogenase). Primer sequences for gene expression analysis for hGAPDH, hTLR3, hTLR7 and IL1β are shown in table 1.
Measurement of cytokines in culture supernatant
TNFα, IL1β, IL6, IL10 and IL12p70 levels were measured in the supernatant of the cell cultures using commercially available kits (Bio-Rad Laboratories, Hercules, California, USA) according to the manufacturer’s instructions.24 Cytokine levels were measured and analysed using the Bio-Plex system (Bio-Rad Laboratories) and data analysis was performed with Bio-Plex Manager software (Bio-Rad Laboratories). The secretion of IL18 was measured using an ELISA (Invitrogen, Biosource, USA) carried out using the manufacturer’s instructions. For the measurement of IFNα and IFNβ, ELISA kits were used and the ELISAs were performed according to the manufacturers’ protocols (Cell Science, Canton, Massachusetts, USA).
Correlations of the expression of TLR and cytokines in human synovial biopsies were calculated using the Pearson correlation test. Differences in mRNA expression and cytokine production on cell stimulation with cytokines and TLR agonists were calculated using the Mann-Whitney U test. p Values were two-sided and the level of significance was set at p<0.05.
The clinical and demographic features of the patients included in the immunohistochemistry studies are presented in table 2. All patients had active disease as defined by the inclusion criteria (DAS >3.2). Not surprisingly, patients with severe knee joint arthritis tended also to have more active disease. The patients with RA included in the in vitro studies on DCs (n = 10) also had active disease with a mean (SD) DAS of 3.8 (0.4) and were comparable with the patients included in the studies on synovial tissues.
Correlation of TLR3/7 expression with IL1β, IL18 and IFNα in RA synovium
Synovial biopsy specimens from 24 patients with active RA (DAS >3.2) stained for TLR subtypes (TLR2, 3, 4 and 7) and inflammatory cytokines IL12, IL17, IL18 and IL1β were used to study the association between TLRs and inflammatory mediators.20,21,22 Sequential slides from those used for the aforementioned markers were then stained for IFNα (fig 1A). TLR3/7 expression was correlated with the levels of expression of IFNα, IL1β and IL18 but not with IL12, IL17 and TNFα. Not unexpectedly, IFNα was clearly correlated with TLR3 and TLR7 expression in the lining and sublining of synovial biopsies (fig 1B, table 3). Moreover, both IL1β and IL18 were correlated with the expression of TLR in the lining, although these correlations were weaker than that observed between TLR and type I IFN. In contrast, neither IL1β nor IL18 was associated with either TLR3 or TLR7 expression in the sublining. Neither TNFα, IL12 nor IL-17 were correlated with the TLR expression levels in RA synovium.
TLR3/7-mediated stimulation of monocytes, MoDCs or synovial fibroblasts could not explain its correlation with IL1β/IL18
Since we found a correlation between TLR3/7 expression and the presence of IFNα, IL1β and IL18, we further investigated the functional relation between these mediators and TLR3/7. It is generally accepted that stimulation of TLR3/7 leads to the production of type I IFN DCs. In line with this, we found that TLR3/7-mediated stimulation of MoDCs resulted in clearly enhanced protein levels of both IFNα and IFNβ that reached similar levels in healthy donors and patients with RA (fig 2A). To investigate the potential relation between TLR3/7 and IL1β, TLR3-expressing MoDCs and synovial fibroblasts were stimulated with poly(IC) and TLR7-expressing monocytes were stimulated with loxoribine. All cell types investigated showed markedly increased expression of IL1β mRNA upon stimulation of either TLR3 or TLR7; however, IL1β protein could not be detected. The stimulation of monocytes, MoDCs or synovial fibroblasts with TLR3/7 did not lead to the secretion of IL18 proteins.
Regulation of TLR3/7 expression by IL1β, IL18 and type I IFN
Since TLR3/7 stimulation leads to the production of IFNα but not to IL1β or IL18, we investigated whether the correlation between TLR3/7 and IL1β/IL18 could be explained by IL1β/IL18-induced upregulation of these receptors. Monocytes (TLR7-expressing cells), MoDCs (TLR3-expressing cells) and RA synovial fibroblasts (TLR3-expressing cells) were cultured in the presence of IFNα, IL1β and IL18. IFNα significantly enhanced TLR3 mRNA expression on RA synovial fibroblasts and MoDCs and TLR7 mRNA expression on monocytes (fig 2B). This increase in TLR expression was similar in patients with RA (n = 5) and in healthy volunteers (n = 5), and was specific for TLR3 and TLR7 since the expression of TLR2 and TLR4 was not altered. In contrast, IL1β and IL18 had no effect on the expression of TLR3/7, excluding a direct role in their co-expression with TLR3/7.
Functionality of IFNα-mediated upregulation of TLR3/7
As we found that the expression of TLR3 and TLR7 was strongly upregulated by IFNα, we examined whether this enhanced TLR expression was functional in terms of increased TLR-mediated cytokine production. TLR7-expressing monocytes and TLR3-expressing MoDCs and RA synovial fibroblasts (n = 8) were incubated with IFNα or medium and TLR3 or TLR7 were stimulated with poly(IC) and loxoribine as appropriate. As anticipated, TLR7 stimulation on monocytes and TLR3 stimulation on MoDCs and RA synovial fibroblasts led to production of IL6 and TNFα, which was significantly enhanced when the cells were preincubated with IFNα compared with cells stimulated with TLR3/7 ligands alone (fig 3). These data show that the upregulation of TLR3/7 by type I IFN is functional.
Augmentation of TLR4-mediated cytokine production by type I IFN provides the missing link explaining the correlation between TLR3/7 and IL1/IL18 expression
Since TLR4 triggering is recognised as a potent inducer of inflammatory mediators including IL1β and IL18, we postulated that stimulation of TLR3/7 leading to the production of type I IFN perhaps augments the TLR4 response ending up in the production of IL1β and IL18. To test this we prestimulated MoDCs with IFNα and then stimulated them with the TLR4 agonist LPS. Surprisingly, prestimulation with IFNα resulted in a marked potentiation (threefold) of TLR4-mediated secretion of TNFα by DC from healthy controls (3044 pg/ml vs 1000 pg/ml, p = 0.01, n = 5; fig 4). In addition, preincubation with type I IFN augmented TNF production by the combination of TLR3/4 ligands by almost threefold (1880 pg/ml vs 5010 pg/ml, p<0.01). Compared with DC from healthy controls, stimulation of DC from patients with RA with TLR4 led to significantly greater production of TNF, as previously described.22 Notably, DC from patients with RA (n = 4) produced a fourfold higher level of TNFα on stimulation with IFNα/LPS (12344 pg/ml vs 3044 pg/ml, p = 0.001) and IFNα/LPS/Poly-IC (28477 pg/ml vs 8484 pg/ml), p = 0.01) than DC from healthy individuals. To investigate whether the augmented TLR4-mediated stimulation by type I IFN could underlie the association between TLR3/7 and IL18, we measured the production of this crucial mediator. In corroboration with TNFα, the secretion of IL18 was markedly increased following prestimulation with type I IFN and stimulation with LPS (2155 pg/ml vs 4855 pg/ml, p = 0.02). The addition of a TLR3 agonist, however, did not amplify IL18 production as is seen with TNF secretion, which is in line with our observation that TLR3/7 stimulation does not result in IL18/IL1β production. The secretion of IL1β showed the same trend, but the absolute levels by MoDCs was low (data not shown). Together these data imply that the TLR4 response is potentiated by prestimulation with type I IFN, and this is even more pronounced in patients with RA. To investigate whether the IFN-mediated augmentation of the TLR4 response is RA-specific, we included DC from patients with SSc (n = 3) in the analysis. Although the absolute production of pro-inflammatory mediators tends to be higher than that seen in healthy controls, the magnitude of potentiation of the TLR4 response is similar to that observed in healthy controls and thus smaller than that in RA.
In this study we have shown that TLR3 and TLR7 in synovial tissue from patients with RA is associated with the presence of IL1β, IL18 and type I IFN. Since TLR3/7-mediated cell activation did not result in IL1β/IL18 secretion and TLR3/7 expression was not regulated by these mediators, we sought evidence for the underlying mechanism to explain the relationship between these mediators in the synovium. Preincubation of DCs with IFNα led to clear augmentation of TLR4-mediated responses. It is therefore tempting to speculate that TLR3/TLR7 triggering leads to the production of type I IFN which, in turn, augments TLR4-mediated triggering leading to the production of pro-inflammatory mediators, explaining the observed correlation between TLR3/7 and IL1β/IL18 expression. Since ligands for the TLR3/7 axis and TLR4 axis are abundantly present in the synovial compartment of patients with RA, this chain of events is likely to contribute to the inflammatory cascade in this condition.
IFN type I is the key cytokine that regulates the innate immune response against viruses.25 Type I IFN is released upon transcription of interferon regulating factor (IRF)-3 or IRF-7, which can be induced following TLR3- or TLR7-mediated cell activation. TLR3 and TLR7 are highly expressed in synovial tissue from patients with RA,22 and TLR3/7 ligands such as cytomegalovirus, Epstein-Barr virus and parvovirus B19 have also been demonstrated in the joints of patients with RA.1,3,26 The observation that a type I IFN signature is present in the synovium in a substantial proportion of patients with RA supports the notion that TLR3 and/or TLR7 triggering is likely to occur in RA synovium. In view of this, it is tempting to speculate that exposure to TLR3/TLR7 agonists might sensitise the synovial milieu for TLR4 ligands via IFNα production which, in turn, acts as a key regulator in the sustained inflammation in RA. For several reasons, IFNβ therapy has been expected to have a beneficial effect in RA comparable to that seen in multiple sclerosis.27 However, IFNβ therapy was shown to be effective in animal models of collagen-induced arthritis but ineffective in several clinical trials in RA.28,29,30 The fact that IFN type I strongly upregulates TLR3/TLR7 expression and tunes TLR4 responses which, in turn, are continuously stimulated by ligands present in the synovial joints might, in fact, maintain the inflammatory processes and therefore lead to failure of IFN therapy. Identification of endogenous TLR agonists is of great interest in terms of autoimmune disorders. In particular for TLR4, several endogenous agonists have been described to date—for example, hyaluronan fragments, heparan sulfate, fibronectin and (small) heat shock proteins—which can all be released in RA joints as a result of inflammation-induced tissue injury and cell stress.11,12,13,31 For instance, Brentano et al described an endogenous ligand for TLR3, as endogenous RNA released from necrotic synovial fluid cells was able to activate RA synovial fibroblasts in a TLR3-dependent fashion.10 Highly conserved RNA sequences within small nuclear ribonucleoprotein particles are able to activate immune cells via TLR7 and could act as an endogenous autoantigen in systemic lupus erythematosus.32,33 Thus, triggering of the TLR3/7 pathways does not necessarily result from viruses but might well originate from the host itself. TLR-mediated immune responses in the synovial joints might lead to the release of endogenous TLR ligands originating from cells under stress or tissue damage. It is therefore not unlikely that a self-sustaining loop of TLR activation and generation of new endogenous TLR ligands might lead to a chronic inflammatory process as occurs during RA.
It has recently been demonstrated that stimulation of several different TLRs at the same time leads to synergistically induced levels of pro-inflammatory mediators.22,34 Our data indicate that cytokine levels induced in a synergistic fashion by co-stimulation of TLR3 and TLR4 were even more enhanced when cells were preincubated with type I IFN. The exact mechanisms underpinning the augmented TLR4 response after preincubation with type I IFN is not yet understood. From our data, we can conclude that this phenomenon is not explained by an increased expression of TLR4 itself since preincubation with type I IFN does not affect its mRNA expression. Perhaps the explanation for this lies in the upregulation of adaptor molecules or downregulation of intracellular inhibitors that are part of the TLR4 pathways.35 Since the augmentation of TLR4 responses by type I IFN was significantly more potent in patients with RA, further investigation into the causative pathways is highly desirable because it might open novel insights into treating this condition selectively.
In conclusion, this study has shown that the expression of TLR3 and TLR7 is associated with IFNα, IL1β and IL18 in synovial tissue from patients with RA. Furthermore, we have demonstrated the involvement of IFNα in the regulation of TLR3/7 expression and function and also in an augmented TLR4-mediated response. These observations might, at least partly, explain the role of type I IFN in the inflammatory cascade of RA.
Funding TRDJR was supported by a VENI and VIDI Laureate of the Netherlands Organization for Scientific Research (NWO).
Competing interests None.
Ethics approval The local Medical Ethics Committee approved the study protocol and patients gave their informed consent.
Contributors: Design: MFR, FB, LABJ, DK, WBvdB, TRDJR; material: PB, PLCMvR, TRDJR; experiments: MFR, MHW, FB; interpretation of data: MFR, MHW, FB, LABJ, DK, TRDJR; writing: MFR, FB, DK, LABJ, WBvdB, TRDJR.
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.