Article Text

PDF

Extended Report
Comparison of rheumatoid articular adipose and synovial tissue reactivity to proinflammatory stimuli: contribution to adipocytokine network
  1. Ewa Kontny1,
  2. Magdalena Plebanczyk1,
  3. Barbara Lisowska2,
  4. Malgorzata Olszewska2,
  5. Pawel Maldyk3,
  6. Włodzimierz Maslinski1
  1. 1Department of Pathophysiology and Immunology, Institute of Rheumatology, Warsaw, Poland
  2. 2Department of Anaesthesiology and Intensive Care, Institute of Rheumatology, Warsaw, Poland
  3. 3Department of Rheumoorthopaedic Surgery, Institute of Rheumatology, Warsaw, Poland
  1. Correspondence to Ewa Kontny, Department of Pathophysiology and Immunology, Institute of Rheumatology, Spartanska 1, 02-637, Warsaw, Poland; ewa.kontny{at}wp.pl

Abstract

Objectives (1) To compare spontaneous and stimuli-induced adipocytokine secretion by articular adipose tissue (AAT) and synovial membrane (SM) explants obtained from patients with rheumatoid arthritis (RA). (2) To investigate the biological activity of AAT and SM released factors.

Methods Tissues were obtained from patients undergoing joint replacement surgery. Tissue explants were treated with proinflammatory cytokines relevant to RA pathogenesis (interleukin 1β (IL-1β), tumour necrosis factor (TNF), interferon γ, IL-15, IL-17, IL-23). Selected adipocytokine (TNF, IL-6, IL-8, IL-1β, IL-1Ra, adiponectin, leptin) concentrations were measured in culture supernatants using ELISA. The biological activity of tissue-conditioned media was evaluated by measuring production of selected factors (IL-6, IL-8, Dickkopf-1, osteoprotegerin) by fibroblast-like synoviocytes (FLS).

Results Spontaneous cytokine release from AAT was ≤12% of that produced by SM, while leptin was secreted in similar amounts. AAT was highly reactive to proinflammatory cytokines (IL-1β>TNF). AAT treated with IL-1β released four times more leptin, similar amounts of IL-6 and IL-8 and about 20% of TNF, as compared with SM. Upon activation, the IL-1 receptor antagonist (IL-1Ra)/IL-1β ratio was higher in AAT than in SM cultures. Irrespective of activation status, SM produced twice as much adiponectin as AAT. Conditioned media from AAT and SM cultures similarly upregulated IL-6, IL-8, Dickkopf-1 and osteoprotegerin production by rheumatoid FLS.

Conclusion Rheumatoid AAT is highly reactive tissue which upon stimulation secretes considerable amounts of proinflammatory (IL-6, IL-8, TNF) and anti-inflammatory (IL-1Ra) cytokines and classical adipokines. This tissue releases biologically active factors that intensify pathogenic activities of rheumatoid FLS. Thus, AAT should be considered an important contributor to the pathological processes taking place in the RA joint.

Statistics from Altmetric.com

Introduction

Rheumatoid arthritis (RA) is a chronic systemic autoimmune disorder characterised by synovitis and progressive damage to articular cartilage and subchondral bone. These pathological processes lead eventually to joint deformity, dysfunction and disability in most affected individuals.1 In RA, numerous cytokines play a fundamental role in inflammation and joint destruction and are thought to originate primarily from synoviocytes and leucocytes that massively infiltrate synovial tissue.2 3

Recently, adipose tissue has been shown to synthesise and release highly bioactive substances: classical adipokines (eg, leptin, adiponectin) and various proinflammatory factors such as cytokines, chemokines, growth factors and complement components. These secretory molecules are collectively called adipo(cyto)kines.4 5 Adipose tissue is an active endocrine organ, heterogeneous with respect to embryonic origin, body distribution and function. In addition to playing a major role in thermogenesis, nutrient and energy homoeostasis, it is also involved in bone mass growth and modulation of the immune response.5,,9 This loose connective tissue is composed of extracellular matrix and various types of cells: fat cells—named adipocytes, and the non-fat cells of the stromal vascular fraction that contains preadipocytes, capillary endothelial cells, infiltrating leucocytes and multipotent stem cells.8 Fat cells of white adipose tissue release mostly classical adipokines, while non-fat cells secret principally cytokines and growth factors.10 In human obesity, accompanied by low-grade inflammation, enhanced infiltration of adipose tissue by leucocytes and systemic release of proinflammatory adipokines, white adipose tissue participates in inflammatory response development and obesity-associated metabolic abnormalities—namely, cardiovascular diseases and type 2 diabetes.8,,10 The action of adipokines in RA pathogenesis is currently one of the most intensely studied topics. It is likely that in RA the articular adipose tissue (AAT) contributes to local pathological processes, but this suggestion is supported mostly by circumstantial evidence.

In the normal knee joint, AAT is organised into the largest Hoffa infrapatellar fat pad (IPFP), and three smaller fat pads. The IPFP is situated intracapsularly, and extrasynovially, but in close contact with synovial membrane (SM) and cartilage surface.11 12 Structural and histological alterations of AAT have been demonstrated in animal models and patients with various arthritides, suggesting that this tissue is also affected with disease.11,,15 Moreover, infiltration of immune cells (monocytes, granulocytes, T lymphocytes) has been observed in the IPFP in the joints of patients with osteoarthritis (OA), and in animal models for OA and RA.14,,16 There are also reports showing elevated expression or secretion of cytokines—that is, tumour necrosis factor (TNF), and interleukin 6 (IL-6), as well as growth factors and classic adipokines (adiponectin, adipsin, visfatin) by the IPFP of patients with OA.15,,18 However, corresponding evidence from patients with RA is missing. In addition, data concerning the effects of cytokines relevant to RA pathogenesis on AAT secretory activity are not available. Therefore, a comparative analysis of spontaneous and stimuli-induced classic adipokine (leptin, adiponectin) and cytokine (IL-1, IL-1 receptor antagonist (IL-1Ra), TNF, IL-6, IL-8) secretion by cultured explants of rheumatoid AAT and SM was carried out in this study. Moreover, biological activity of AAT and SM tissue-conditioned media was evaluated using rheumatoid fibroblast-like synoviocytes (FLS) as responder cells.

Patients and methods

Patients

A group of 34 patients (32F/2M) who fulfilled the American College of Rheumatology revised criteria for the diagnosis of RA stages III–IV was included in this study.19 The mean (range) patients' age and disease duration was 52 (29–66) and 19 (3–40) years, respectively. Tissue specimens were obtained from the knee joints at the time of total joint surgery, performed as a normal part of clinical care. All patients gave their informed consent and the study was approved by the Institute of Rheumatology Ethics Committee. Patients were treated with both methotrexate and steroids; none had received biological therapy. Methotrexate treatment was discontinued 1 week before surgery.

Tissue, cell culture and treatment

Tissues, transported in sterile buffer (phosphate-buffered saline, 5.5 mM glucose, 50 μg/ml gentamicin) were processed under sterile conditions within 2 h after removal from the patients. Tissue explants were prepared according to a previously described protocol.20 Briefly, AAT was separated from SM and every tissue was cut into 8–10 mg pieces. After extensive washing with phosphate-buffered saline, tissue explants (100 mg/ml/well) were placed into a 24-well culture plate and precultured for quiescence (26 h, 37°C, 5% CO2) in Dulbecco's modified eagle medium (DMEM) supplemented with antibiotics (50 μg/ml gentamicin, 100 mg/ml kanamycin). Before stimulation, culture medium (CM) was replaced three times (at 1, 18 and 26 h of preculture) with fresh CM. Tissue explants were stimulated for 18 h with recombinant human cytokines relevant to RA pathogenesis—namely, IL-1β (1 ng/ml), TNF (10 ng/ml) or 40 ng/ml of interferon γ (IFNγ), IL-15, IL-17 or IL-23 (all cytokines were from R&D systems, Minneapolis, MN, USA). Tissue explants cultured in medium alone or treated with 1 μg/ml of lipopolysaccharide (LPS) from Escherichia coli 055:B5 (LPS, Difco, Detroit, MI, USA) were used as the negative or positive controls, respectively. Finally, adipocytokine concentrations in culture supernatants were measured by specific ELISA. Experiments were performed using tissue explants obtained from a number of different patients with RA, and each culture was carried out in duplicate.

FLS were isolated from rheumatoid SM and maintained as described previously.21 Quiescent cells (2×104/well) were treated for 24 h with 250 μl of AAT or SM CM collected from tissue cultures and supplemented with 10% of fetal calf serum. Cells cultured in medium (DMEM/antibiotics/10% fetal calf serum) alone were used as an untreated control. After the treatment, tissue CM was discarded, replaced with 1 ml of fresh CM and cell cultures were continued for an additional 24 h. Finally, the culture supernatants were collected and FLS lysates were prepared, as described previously.22 In both types of sample, IL-6, IL-8, Dickkopf-1 (DKK-1) and osteoprotegerin (OPG) concentrations were measured by ELISA to estimate levels of secreted and cell-associated factors, respectively.

Enzyme-linked immunosorbent assays

The ELISA for IL-6 and IL-8 were done as previously described,21 while ELISA for IL-1β, IL-1Ra, TNF, leptin, adiponectin, DKK-1, and OPG were performed using appropriate ELISA sets (DuoSets, R&D systems, Minneapolis, MN, USA), according to the manufacturer's protocols.

Statistical analysis

Values are shown throughout the paper as the median±SEM. Data were analysed using Statistica vol. 7.0 software. The Wilcoxon test was applied to evaluate the effect of stimuli, while comparison between AAT and SM cultures was done using the Mann–Whitney U test. Differences were considered significant for *p<0.05, **p<0.01 and ***p<0.001.

Results

Adipocytokine secretion by tissue explants

Spontaneous leptin secretion by AAT and SM explants did not differ significantly (table 1, figure 1). Applied stimuli failed to modify leptin release from SM, while all of them raised leptin secretion by AAT. Upon IL-1β and TNF treatments this tissue produced significantly more leptin than SM (figure 1). On the other hand, SM secreted spontaneously twice as much adiponectin as AAT, and in both tissues adiponectin release did not change upon stimulation.

As shown in table 1 and figure 2 spontaneous TNF secretion by AAT and SM explants was negligible but rose markedly upon LPS and IL-1β treatment, while other stimuli had little if any effects. Notably, upon stimulation with IL-1β the release of TNF from AAT was raised to a level reaching 21% of that detected in SM cultures. Similarly, AAT explants released spontaneously small amounts of IL-6 and IL-8, but this tissue was highly reactive to the majority of applied stimuli, with the exception of IL-17 and IL-23 which failed to elevate IL-8 secretion. By contrast, spontaneous IL-6 and IL-8 secretion by SM explants was prominent. However, in this tissue IL-6 release rose only in the presence of some (LPS, IL-1β, TNF, IFNγ) but not all stimuli, and no stimuli upregulated IL-8. These results show that AAT is highly reactive tissue and responds by IL-6 and IL-8 release to a broader range of stimuli than SM. Moreover, upon IL-1β treatment both tissues secreted comparable amounts of these cytokines. Also TNF raised IL-6 and IL-8 release from AAT to considerable levels (35% and 56% of that noted in analogous SM cultures, respectively). These observations suggest that in inflammatory conditions—for example, upon exposure to IL-1β or TNF, the contribution of AAT to local IL-6 and IL-8 production might be substantial.

Figure 1

Secretion of classical adipokines by rheumatoid articular adipose tissue (AAT) and synovial membrane (SM) explants. Tissue explants were cultured in culture medium alone (C–control) or treated with indicated stimuli, then adipokine concentrations in culture supernatants were measured by ELISA, as described in ‘Patients and methods’. Values are the median and SEM of 24–34 (AAT) or 10–20 (SM) experiments. Numbers above bars represent the percentage of adipokine release by AAT explants in relation to that produced by SM (100%). Asterisks above the bars indicate statistically significant differences between stimulated and control cultures. Asterisks inside the bars indicate statistically significant differences between AAT and SM cultures; *p<0.05, **p<0.01, ***p<0.001. IFN, interferon; IL, interleukin; LPS, lipopolysaccharide; TNF, tumour necrosis factor.

Figure 2

Secretion of proinflammatory cytokines by rheumatoid articular adipose tissue (AAT) and synovial membrane (SM) explants. Tissue explants were cultured in culture medium alone (C–control) or treated with indicated stimuli, then cytokine concentrations in culture supernatants were measured by ELISA, as described in ‘Patients and methods’. Values are the median and SEM of 24–34 (AAT) or 10–20 (SM) experiments. Numbers above bars represent the percentage of cytokine release by AAT explants in relation to that produced by SM (100%). Asterisks above the bars indicate statistically significant differences between stimulated and control cultures. Asterisks inside the bars indicate statistically significant differences between AAT and SM cultures; *p<0.05, **p<0.01, ***p<0.001. IFN, interferon; IL, interleukin; LPS, lipopolysaccharide; TNF, tumour necrosis factor.

Table 1

Spontaneous adipocytokine secretion by AAT and SM explants

Tissue explants released spontaneously minute (SM) or negligible (AAT) amounts of IL-1β but considerable amounts of IL-1Ra (table 1, figure 3). In both SM and AAT cultures LPS, IL-1β and TNF further raised IL-1Ra secretion, while other stimuli failed to exert any significant effects. Interestingly, in SM cultures LPS and TNF potently elevated also IL-1β secretion, while this cytokine release from AAT was moderately raised only by LPS. Owing to these differences the calculated ratio of IL-1Ra to IL-1β was significantly higher in stimulated AAT than SM cultures, and upon LPS and TNF treatment AAT released 100–360 times more IL-1Ra than IL-1β, while SM secreted IL-1Ra only in 16–19-fold excess (figure 3).

Figure 3

Secretion of IL-1β and interleukin 1 receptor antagonist (IL-1Ra) by rheumatoid articular adipose tissue (AAT) and synovial membrane (SM) explants. Tissue explants were cultured in culture medium alone (C–control) or treated with indicated stimuli, then cytokine concentrations in culture supernatants were measured by ELISA, as described in ‘Patients and methods’. Values are the median and SEM of 24–34 (AAT) or 10–20 (SM) experiments. Numbers above bars represent the percentage of cytokine release by AAT explants in relation to that produced by SM (100%). Asterisks above the bars indicate statistically significant differences between stimulated and control cultures. Asterisks inside the bars indicate statistically significant differences between AAT and SM cultures; *p<0.05, **p<0.01, ***p<0.001. IFN, interferon; IL, interleukin; LPS, lipopolysaccharide; TNF, tumour necrosis factor.

Tissue CM activate rheumatoid FLS

To verify whether factors released by tissue explants are biologically active, the effect of CM from untreated, TNF- and IL-1β-stimulated cultures on FLS activity was evaluated. With this aim, secreted and cell-associated concentrations of select proinflammatory cytokines (IL-6, IL-8) and proteins regulating bone homoeostasis (DKK-1, OPG) were measured (figure 4). Untreated FLS produced huge amounts of DKK-1, significantly less OPG, and minute amounts of IL-6 and IL-8. These factors were mostly secreted, but substantial proportions were also cell-associated, except for IL-8. Preliminary experiments have shown that recombinant human IL-1β (10–1000 pg/ml) and TNF (25–2500 pg/ml), used at concentrations comparable with the amounts released from AAT and SM, significantly upregulate IL-8, IL-6 and OPG production in FLS (the fold increase within the range 30–100, 5–15 and 4–6, respectively). By contrast, DKK-1 production was slightly (twofold increase) raised only by TNF (data not shown). Importantly, in FLS also CM from AAT and SM cultures significantly elevated secreted and cell-associated levels of almost all tested factors (figure 4). Culture medium from activated AAT and SM cultures upregulated IL-6 and IL-8 more potently than CM from untreated tissues. However, CM from TNF-stimulated SM cultures triggered higher production of IL-6 than analogous AAT CM. Similar differences were noted between the levels of IL-8 production induced by CM from untreated tissues. There were no differences between the effects exerted by AAT CM and SM CM on OPG and DKK-1 production. Despite this, stimulatory effect of CM from activated AAT on OPG secretion was more pronounced than the effect of CM from untreated tissue, while OPG upregulation by SM CM was independent of tissue activation status. For DKK-1, CM from untreated and IL-1β-stimulated AAT and SM elevated this protein production equally. By contrast, CM from TNF-treated tissues exerted weaker (AAT/TNF), if any (SM/TNF), stimulatory effects. Thus, factors secreted by AAT and SM are biologically active and modify RA FLS activity.

Figure 4

Conditioned media from rheumatoid articular adipose tissue (AAT) and synovial membrane (SM) enhance fibroblast-like synoviocytes (FLS) activities. Rheumatoid FLS were treated with conditioned media from untreated (AAT/C, SM/C), TNF- (AAT/TNF, SM/TNF) or IL-1β-stimulated (AAT/IL-1, SM/IL-1) tissue cultures. Cells cultured in medium alone were used as an untreated control (C). After replacement of conditioned media with culture medium, IL-6, IL-8, Dickkopf-1 (DKK-1) and osteoprotegerin concentrations were measured in culture supernatants (secreted) and cell-lysates (cell-associated) by ELISA, as described in ‘Patients and methods’. Values are the median and SEM of six experiments performed using FLS, AAT, and SM obtained from different patients with rheumatoid arthritis. Asterisks above the bars indicate statistically significant differences between untreated (C) and conditioned media-treated cultures. Asterisks inside the bars indicate statistically significant differences between cultures treated with AAT and SM conditioned media. # - statistically significant differences between FLS cultured in conditioned media from untreated versus IL-1β- or TNF-stimulated tissues; * and # p<0.05. IFN, interferon; IL, interleukin; LPS, lipopolysaccharide; TNF, tumour necrosis factor.

Discussion

Human adipocytes secrete small amounts of proinflammatory cytokines and are a major source of classical adipokines.10 However, other cell types express leptin (eg, chondrocytes, FLS, inflammatory cells) and adiponectin (eg, FLS, osteoblasts) as well.23,,28 Therefore, it is likely that in the rheumatoid joint proinflammatory cytokines and classic adipokines may originate from both non-fat and fat cells. Based on these findings we evaluated select adipocytokine (TNF, IL-6, IL-8, IL-1β, IL-1Ra, leptin and adiponectin) secretion by AAT and SM explants obtained from patients with RA. The aim of our study was to provide direct evidence clarifying the range of AAT contribution to local inflammatory response. Both spontaneous and stimuli-triggered release of tested adipocytokines was evaluated. As stimuli we used cytokines (IL-1β, TNF, IFNγ, IL-15, IL-17 and IL-23) that are produced in rheumatoid joints and have an established role in RA pathogenesis.2 3 Additionally, as a positive control we applied bacterial LPS, known to activate fat and non-fat cells via toll-like receptor 4.5

It has been suggested that classic adipokines participate in RA pathogenesis. However, data concerning leptin contribution to the inflammatory process and joint destruction are inconsistent. Both pro- and anti-inflammatory properties of leptin, as well as anabolic and detrimental effects of this adipokine on chondrocyte metabolism, were reported.4 29 For adiponectin, more coherent data show mostly its proinflammatory and pro-destructive effects exerted on the majority of cells involved in cartilage destruction.4 29 30 These results confirm that leptin and adiponectin are produced by both rheumatoid AAT and SM. In addition, we found that AAT and SM released spontaneously equivalent amounts of leptin, but upon IL-1β or TNF treatment this adipokine was secreted mostly by AAT. Thus, contribution of these tissues to leptin production is not equal and depends on tissue activation status (figure 1). Contrary to leptin, more adiponectin was produced by SM, irrespective of tissue activation.

For the cytokines, we found that spontaneous release of TNF, IL-1β, IL-1Ra, IL-6, IL-8 by AAT explants was low (≤12% of that released by SM), but these proportions changed drastically upon stimulation (figures 2 and 3). Secretion of IL-6 and IL-8 from AAT was potently elevated by IL-1β and TNF, while other stimulatory cytokines had weaker effects (figure 2). Interestingly, AAT responded to TNF and produced this cytokine (figures 1 and 2). Thus, similarly to SM, TNF can act on rheumatoid AAT in both an autocrine and paracrine way. Contrary to SM, AAT secreted little, if any, IL-1β (figure 3). Hence, it is likely that in rheumatoid joint AAT responds to IL-1β produced by other cells—for example, those present in SM. The biological activity of IL-1β is counteracted by its natural antagonist, IL-1Ra.31 Adipose tissue was reported to represent a major source of circulating IL-1Ra.32 33 Moreover, expression of IL-1Ra, but not IL-1α, was demonstrated in synovial adipose tissue of healthy donors.32 Our results support the suggestion that rheumatoid AAT also produces mostly IL-1Ra (figure 3). For successful blockade of IL-1 activity a large molecular excess (100-fold to 1000-fold) of IL-1Ra is required.31 Interestingly, activated AAT, but not SM, explants released IL-1Ra in an excess sufficient to protect tissue against activation by IL-1β (figure 3). These observations show that rheumatoid AAT is highly reactive to IL-1β delivered mostly in a paracrine way, but upon exposure to proinflammatory stimuli the predominance of IL-1Ra over IL-1β can restrict AAT reactivity to this cytokine by a regulatory feedback mechanism.

Altogether, these results show that in addition to SM, activated rheumatoid AAT also is a rich source of pro- (TNF, IL-6, IL-8) and anti-inflammatory (IL-1Ra) cytokines that can modify both tissue activities. Proinflammatory cytokines amplify the inflammatory process and may affect metabolism of adipocytes—for example, by inhibiting preadipocyte maturation, modifying secretion of classic adipokines and exerting lipolytic effects.34,,38 Owing to this last event they mobilise and direct energy from the adipocyte to other inflammatory cells.

In vivo cytokines act in a complex system of synergistic and antagonistic interactions. Therefore, their net biological effect is hardly predicable, even if their locally produced profile is more or less characterised. Thus, to finally confirm AAT contribution to the pathological processes taking place in the rheumatoid joint, we evaluated the biological activity of AAT and SM CM, using rheumatoid FLS, the key participants of chronic inflammation and joint destruction, as the responder cells. Rheumatoid FLS produce numerous biologically active agents, including proinflammatory cytokines (IL-6, IL-8) and bone homoeostasis regulating factors (OPG, DKK-1). Bone remodelling is conducted by the synthesis of bone matrix by osteoblasts, and its resorption by osteoclasts. Generation and survival of osteoclasts is supported by receptor activator of nuclear factor-κB ligand (RANKL), which acts by binding to a specific cell surface receptor (RANK). OPG is a non-signalling decoy receptor for RANKL, hence it protects bone from resorption.39 In contrast, DKK-1, a secreted protein antagonist of canonical Wnt signalling, mostly inhibits osteogenesis, but it also stimulates bone resorption by inducing RANKL and inhibiting OPG.40 41 In RA overactivation of the RANK/RANKL pathway with concomitant excessive production of DKK-1 is thought to be responsible for progressive articular bone loss and impaired rebuilding of this tissue, respectively.42 In addition, proinflammatory cytokines (TNF, IL-1, IL-6) contribute to inflammation, and via the above pathways they also support bone destruction.42

We found that CM from AAT and SM significantly upregulated IL-6, IL-8, OPG and DKK-1 production by FLS (figure 4), and thus we provide evidence that factors released by these tissues are biologically active. Although stimulatory activities of CM from both tissues were rather similar, some differences were seen (see ‘Results’ and figure 4). Interestingly, production of IL-6 and IL-8, but not DKK-1, was more potently elevated by CM from activated than untreated AAT and SM. Similarly, the magnitude of OPG upregulation was dependent on AAT activation, while such a relation was not found for SM. These observations support once more the proposal that AAT is highly reactive tissue and, especially upon IL-1β or TNF exposure, releases FLS activating factors. Our results do not allow identification of these factors. However, among those tested, TNF, IL-6 and leptin are plausible candidates, because in AAT they all are upregulated by IL-1β (figures 1-3), and IL-1β-treated AAT CM shows the strongest stimulatory effect on FLS activity (figure 4). However, this suggestion needs further study and more detailed characterisation of the factors secreted by AAT.

We report that activated rheumatoid AAT produces considerable amounts of pro- (TNF, IL-6, IL-8) and anti-inflammatory (IL-1Ra) cytokines, as well as classical adipokines (leptin, adiponectin). This tissue is highly reactive to IL-1β and TNF, which in vivo probably act on AAT by either a paracrine or a paracrine and autocrine way, respectively. For the first time, we showed that AAT-secreted factors are biologically active and stimulate rheumatoid FLS with similar potency to those originating from SM. Thus, AAT should be considered as a rich source of adipocytokines and other factors that participate in local pathological processes characteristic of RA.

References

View Abstract

Footnotes

  • Funding This work was sponsored by grant No N N402 369938 from the Polish Ministry of Science and Higher Education.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval The Institute of Rheumatology Ethics Committee.

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

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.