Heat Shock Protein 60 as a Mediator of Adipose Tissue Inflammation and Insulin Resistance

Primary human preadipocytes were obtained from subcutaneous adipose tissue from lean or overweight females undergoing elective plastic surgery (BMI 28.1 ± 1.1 kg/m², age 42.4 ± 2.8 years) and from PromoCell (Heidelberg, Germany) and were differentiated in vitro to adipocytes as described before (10). For isolation of mature adipocytes and the stromavascular fraction, the protocol was modified by decreasing the collagenase digestion period to 45 min. Mature adipocytes were collected by careful aspiration of the upper phase, while the lower phase was centrifuged at 1,100g to obtain the stromavascular fraction. All protocols were approved by the local ethics committee, and all participants gave written informed consent. Primary human SkMCs derived from healthy individuals (male: 16 and 21 years of age; female: 33 and 37 years of age) were obtained from PromoCell, cultivated, and differentiated as described before (10).

Antibodies against phospho–extracellular signal–related kinase (ERK)-1/2 (Thr202/Tyr204), phospho-p38 (Thr180/Tyr182), phospho-SAPK/JNK (p46,Thr183/Tyr185), phospho–nuclear factor (NF)-κB (p65, Ser536), phospho-Akt (Ser473), phospho-GSK3α/β (Ser21/9), and β-actin (clone 13E5) were obtained from Cell Signaling Technology (Danvers, MA). Antitubulin antibodies were obtained from Calbiochem (Merck Biosciences, Schwalbach, Germany). Hsp60 antibodies (clone 24/HSP60) were purchased from BD Biosciences (San Diego, CA), and horseradish peroxidase–conjugated goat anti-rabbit and goat anti-mouse secondary antibodies were from Pierce Thermo Scientific (Bonn, Germany). Recombinant human Hsp60 was obtained from Loke Aps Diagnostics (Risskov, Denmark) or from StressGen Biotechnologies (Victoria, BC, Canada). Lipopolysaccharide (LPS) (Escherichia coli, 026:B6), lipoteichoic acid, ovalbumin (OVA), and porcine insulin were purchased from Sigma-Aldrich (Steinheim, Germany). Recombinant TNF-α, IL-1β, and interferon (IFN)-γ were obtained from Miltenyi Biotech (Bergisch Gladbach, Germany). The enzyme-linked immunosorbent assay (ELISA) kit for phospho–insulin receptor substrate (IRS)-1 (Ser307) was purchased from Cell Signaling Technology.

Primary human subcutaneous adipocytes and SkMCs obtained at different differentiation time points (day 0–14 and day 0–8, respectively) were lysed in a buffer containing 50 mmol/L HEPES, pH 7.4, 1% Triton, and protease inhibitors (Roche Diagnostics, Mannheim, Germany). Cell lysates were analyzed for Hsp60 expression by Western blot using the Lumi-Imager system (Roche Diagnostics).

To analyze the release of Hsp60 from human preadipocytes, mature adipocytes and SkMC 3.5 × 10⁶ cells were seeded in 75-cm² cell culture flasks for differentiation, and adipocyte conditioned medium (CM) was generated as described earlier (11). Adipocyte CM was collected and concentrated 200-fold by Amicon Ultra centrifugal filter units (Millipore, Schwalbach, Germany) before Western blot analysis. To induce Hsp60 release upon inflammatory stress or Toll-like receptor (TLR) activation, cells were treated with LPS (1 μg/mL); lipoteichoic acid (5 μg/mL); a cytokine mixture composed of TNF-α, IL-1β, and IFN-γ (1,000 units/mL for each cytokine); or each cytokine individually for 48 h. Concentrated cell culture supernatants were analyzed for Hsp60 concentrations by ELISA (Cusabio Biotech, Newark, DE).

For Hsp60 binding studies, 0.5 × 10⁶ human adipocytes or SkMCs were either directly incubated with DyLight649-labeled (Pierce, Rockford, IL) Hsp60 (Hsp60*, 45 min, 4°C) or preincubated with unlabeled Hsp60 or OVA as described before (12).

SkMCs were used for glucose uptake experiments at 4 days after start of differentiation. Uptake of 2-deoxyglucose was measured for 2 h after 30 min exposure to insulin (100 nmol/L) as described before (13). To analyze the effect of Hsp60 on Akt and GSK3α/β phosphorylation, human subcutaneous adipocytes and SkMCs were preincubated with medium or Hsp60 (0.5–20 μg/mL) for 24 h. Afterward, cells were stimulated with insulin (100 nmol/L) for 10 min. To investigate the Hsp60-induced activation of signal proteins, cells were stimulated for 0–60 min with medium, 2.5 nmol/L TNF-α (positive control), or Hsp60 (1–20 μg/mL). Subsequently, cells were washed with cold PBS, lysed (1–2 h, 4°C), sonified, and centrifuged (15 min, 10,000g, 4°C). For detection of activated signal proteins, appropriate antibodies were applied. Signals were visualized by the Lumi-Imager system.

Human subcutaneous adipocytes or SkMCs (1 × 10⁵ cells each) were seeded and differentiated in 48-well cell culture plates and exposed to control medium, 0.001–20 μg/mL recombinant human Hsp60 (StressGen Biotechnologies) or 1 μg/mL LPS. After 24 h, concentrations of TNF-α, monocyte chemoattractant protein (MCP)-1, regulated on activation normal T cell expressed and secreted (RANTES), macrophage inflammatory protein-1α (MIP-1α), IL-6, and IL-8 were measured in cell supernatants by multiplex beads assay (Luminex, Austin, TX).

Hsp60 concentrations were determined by ELISA (Cusabio Biotech) in plasma obtained from 18 lean (BMI 23.4 ± 5.6 kg/m², age 56 ± 14 years) and 23 obese (15 without and 8 with type 2 diabetes) men (BMI 44.5 ± 5.6 kg/m², age 51 ± 14 years) in the fasted state before undergoing abdominal surgery at Ghent University Hospital. Anthropometric measurements were performed during preoperative examination. Subjects gave written informed consent to participate in this study, which was approved by the Ethical Review Board of the Ghent University Hospital and conducted according to most recent version of the Declaration of Helsinki. Adipose tissue lysates from paired subcutaneous and visceral fat of lean and obese individuals with or without type 2 diabetes were prepared as described before (14).

Data were expressed as means ± SEM. Statistical analysis was performed using the Student t test or ANOVA. Correlations were performed by Pearson product-moment correlation. Statistical analyses were done with JMP (SAS Institute, Cary, NC) or Prism (GraphPad Software, San Diego, CA). Differences were considered statistically significant at P < 0.05.

Human subcutaneous adipocytes express Hsp60 in comparable amounts in all differentiation stages (day 0–14) (Fig. 1A). Mature adipocytes freshly isolated from adipose tissue express significantly higher amounts of Hsp60 compared with the stromavascular fraction (Fig. 1B). To examine whether human subcutaneous adipocytes release Hsp60, a considerable amount of cell supernatant, termed “conditioned medium,” was harvested, concentrated 200-fold, and analyzed by Western blot, with adiponectin as a positive control (Fig. 1C). Hsp60 was already detected in 1 μL concentrated CM with increasing signal in 3 and 5 μL concentrated CM (Fig. 1C). Of note, Hsp60 concentration in CM was lower than adiponectin, since Hsp60 (in contrast to adiponectin) was only detectable by ELISA in concentrated CM. Adipocytes at day 14 of differentiation released about three times more Hsp60 than preadipocytes at day 0 (Fig. 1D). Moreover, human adipocytes were treated with proinflammatory cytokines (TNF-α, IL-1β, and IFN-γ, individually or as a mixture) to induce the release of Hsp60 under inflammatory conditions. The application of single cytokines evoked a significant Hsp60 secretion by IL-1β (3.9 ± 0.8-fold) and TNF-α (3.7 ± 0.3-fold), whereas the effect of IFN-γ was negligible (1.1 ± 0.3-fold) compared with medium control (Fig. 1E). Exposure of adipocytes to the cytokine mixture led to a 3.1 ± 0.9-fold secretion of Hsp60. The application of single cytokines to human preadipocytes revealed similar tendencies as for adipocytes (data not shown). Treatment of adipocytes with LPS as a TLR4 agonist stimulated Hsp60 release, whereas lipoteichoic acid as a TLR2 agonist had no effect on Hsp60 secretion (Fig. 1F).

Applying fluorescent-labeled Hsp60 (Hsp60*) to human preadipocytes and mature adipocytes revealed specific binding of Hsp60* to both cell populations (Fig. 2). Specificity was proven through inhibition of Hsp60*-binding by preincubation with the unlabeled ligand (Hsp60) up to 85.0% for preadipocytes (Fig. 2A) and up to 81.4% for adipocytes (Fig. 2B), whereas incubation with OVA was without any effect.

Human adipocytes revealed a dose-dependent significant decrease of insulin-stimulated Akt phosphorylation up to 55.0 ± 14.1% by 10 μg/mL Hsp60 (Fig. 3A). To elucidate the effect of Hsp60 on Akt phosphorylation, different mitogen-activated protein (MAP) kinases and the NF-κB pathway were investigated (Fig. 3B–F). Medium and TNF-α (2.5 nmol/L) were used as controls. A 4.6 ± 0.7-fold activation of the MAP kinase ERK1/2 occurred after a 10-min exposure to Hsp60 (1 μg/mL) (Fig. 3C). JNK activation reached its maximum after 10 min of incubation with 10 μg/mL Hsp60, since the phosphorylation increased 2.1 ± 1.2-fold over medium control (Fig. 3D). Activation of the MAP kinase p38 increased after 30 min (10 μg/mL Hsp60) up to 3.8 ± 0.7-fold (Fig. 3E), whereas the NF-κB pathway reached its activation peak (4.8 ± 2.8-fold) already after 10 min at the highest Hsp60 concentration (Fig. 3F).

Human adipocytes were exposed to different Hsp60 concentrations (0.001–20 μg/mL) to identify stimulatory Hsp60 concentrations and to investigate dose-dependent Hsp60 effects. Hsp60 concentrations <0.5 μg/mL did not result in a measurable secretion of proinflammatory mediators (data not shown). Treatment with Hsp60 (>0.5 μg/mL) led to a dose-dependent significant secretion of TNF-α (up to 14.0 ± 6.3-fold), RANTES (up to 7.1 ± 2.0-fold), MIP-1α (up to 15.6 ± 3.5-fold), and IL-8 (up to 15.2 ± 2.4-fold) by preadipocytes compared with untreated cells (Fig. 4A–F). In mature adipocytes, Hsp60-stimulated secretion of TNF-α (up to 20.0 ± 9.1-fold), MCP-1 (up to 22.7 ± 7.3-fold), RANTES (up to 1,900.0 ± 302.0-fold), MIP-1α (up to 140.0 ± 59.1-fold), IL-6 (up to 32.0 ± 3.5-fold), and IL-8 (up to 2.4 ± 0.5-fold) and occurred in a concentration-dependent manner compared with unstimulated adipocytes (Fig. 4A–F).

Hsp60-treated human SkMCs revealed a significant dose-dependent decrease in insulin-stimulated Akt phosphorylation from 77.3 ± 7.9% (for 1 μg/mL Hsp60) to 50.3 ± 4.6% (for 20 μg/mL Hsp60) compared with insulin control (Fig. 5A). Moreover, Hsp60 impaired insulin-stimulated phosphorylation of GSK3α and GSK3β (Fig. 5B and C). Our results demonstrate a significantly increased IRS-1 phosphorylation after application of insulin and Hsp60 (154.1 ± 12.4-fold) compared with insulin alone (126.8 ± 7.7-fold; Fig. 5D). Furthermore, Hsp60 exposure to human SkMCs significantly decreased insulin-stimulated glucose uptake (2,477.0 ± 29.3 cpm/well) compared with insulin-stimulated control (2,731.0 ± 134.2 cpm/well; Fig. 5E). Hsp60 treatment increases basal glucose uptake compared with control (1,491 cpm compared with 1,292 cpm). In previous studies, we also observed that SkMCs treated with adipocyte-CM display increased basal glucose uptake. Because both Hsp60 and adipocyte CM activate inflammatory and stress signaling pathways, one might suggest that activation of these pathways is responsible for this effect. To elucidate the role of Hsp60 in the activation of MAP kinases, Hsp60 and TNF-α were applied to SkMCs (Fig. 6A). Maximal activation of the MAP kinases ERK1/2 and JNK was reached after 30 min of incubation with 20 μg/mL Hsp60 (2.4 ± 0.3-fold increase of ERK1/2 and 2.4 ± 0.2-fold increase of JNK activation above medium control; Fig. 6A–C). The NF-κB pathway was activated in a more prolonged manner, with a maximal activation (2.3 ± 0.5-fold over medium control) after 60 min of exposure to 20 μg/mL Hsp60 (Fig. 6D). Activation of the MAP kinase p38 by Hsp60 could not be observed (data not shown).

Immunoblot analysis of unstimulated SkMC lysates revealed a consistent expression of Hsp60 in all analyzed maturation states (day 0–8; Fig. 7A). Hsp60 expression in SkMCs is donor dependent and slightly, but not significantly, lower than in adipocytes (data not shown). To investigate if human SkMCs themselves serve as a source for extracellular Hsp60, cell supernatants of unstimulated SkMCs were generated, concentrated to the same extent as the CM from adipocytes, and analyzed by ELISA. Hsp60 release by human SkMCs was not detectable.

Hsp60* was applied to human myoblasts and found to bind specifically to these cells (Fig. 7B). The Hsp60* binding signal was drastically reduced (81.6% inhibition) only in the presence of unlabeled Hsp60, but not with OVA as the control (Fig. 7B). Human SkMCs were exposed to different Hsp60 concentrations (0.001–20 μg/mL), and measurable cytokine levels resulted from Hsp60 concentrations ≥0.5 μg/mL. Hsp60 (20 μg/mL) induced a significant release of MCP-1 (1.2 ± 0.2 ng/mL) compared with medium control (0.1 ± 0.0 ng/mL), IL-8 (1.7 ± 0.4 ng/mL) compared with medium control (0.1 ± 0.0 ng/mL), and IL-6 (0.4 ± 0.1 ng/mL) compared with medium control (0.1 ± 0.0 ng/mL) from human SkMCs (Fig. 7C–E).

Plasma Hsp60 levels were higher in obese (20.3 ± 11.0 ng/mL Hsp60) than in lean (12.6 ± 11.0 ng/mL Hsp60) men (Fig. 8A; for clinical characterization of patients, see Supplementary Table 1).Within the obese group, Hsp60 levels did not differ between males with (n = 13) or without (n = 8) type 2 diabetes (data not shown). Nevertheless, plasma Hsp60 related weakly but positively with BMI (P = 0.04; r = 0.34), diastolic (P = 0.04; r = 0.35) and systolic (P = 0.03; r = 0.35) blood pressure, plasma leptin (P = 0.04; r = 0.34), and homeostasis model assessment–insulin resistance (HOMA-IR) (P = 0.05; r = 0.35) and inversely with quantitative insulin sensitivity check index (QUICKI) (P = 0.04; r = 0.38). Hsp60 expression in visceral relative to subcutaneous adipose tissue was greater in obese patients with type 2 diabetes (n = 8) than in obese (n = 7) or lean (n = 9) patients without type 2 diabetes (Fig. 8B).

The current study demonstrates that Hsp60 is a novel adipokine that could contribute to inflammatory processes in an autocrine manner within adipose tissue and to the development of peripheral insulin resistance in an endocrine fashion. The observation that Hsp60 is not only expressed but also released clearly indicates that Hsp60 is an intracellular chaperone and also a secretion product. Other heat shock proteins can also be secreted from viable cells such as cardiomyocytes, glial cells, and peripheral blood mononuclear cells (15,16). To date, the origin of circulating Hsp60 remains elusive, but adipose tissue might be considered as one of the sources. It has been reported that circulating levels of Hsp60 were elevated in subjects with inflammatory diseases such as arthritis, atherosclerosis, or type 2 diabetes (9,17,18). Here, we demonstrate that obese patients with or without diabetes also have higher plasma Hsp60 concentrations, suggesting that adipose tissue is a possible source of circulating Hsp60. Analysis of Hsp60 levels in concentrated cell culture supernatants confirmed adipocytes as a putative origin of circulating Hsp60, since Hsp60 was mainly released from untreated human adipocytes, in a lesser extent from preadipocytes, but not from human SkMCs. The latter could be explained by Hsp60 concentrations below the detection limit, even in highly concentrated CM from SkMCs. However, further investigations are needed to clarify Hsp60 sources in vivo. Furthermore, we simulated inflammation by applying a cytokine mixture to human adipocytes, to induce the release of Hsp60. To elucidate the individual contribution of these cytokines to these processes, cells were also exposed to each cytokine individually. Hsp60 secretion was observed after the application of TNF-α and IL-1β in comparable amounts to that obtained after incubation with the cytokine mixture, but not after IFN-γ treatment. These results indicate that macrophages/monocytes and adipocytes themselves might be an inducer of Hsp60 release by adipocytes, since they have been described as sources of TNF-α and IL-1β (7,19). Besides IL-1β and TNF-α, other stimuli for the regulation of Hsp60 expression might be considered. Increased Hsp60 expression was reported for primary human astrocytes in response to cytokines as diverse as IL-1β and TNF-α but also IL-4, IL-6, and IL-10 (20).

In general, the amounts of Hsp60 released by adipocytes are relatively low and not measurable in unconcentrated CM by ELISA, a problem that is often encountered in unconcentrated cell culture supernatants. In fact, various known adipokines that are found in high concentrations in the circulation are relatively low in adipocyte CM (21). Nevertheless, it can be hypothesized that extracellular levels of Hsp60 mediating auto- and paracrine effects might be significantly higher. Furthermore, the fact that circulating Hsp60 levels are higher in obese subjects than in lean control subjects and that Hsp60 concentrations are correlated with circulating leptin further supports the assumption that adipocytes are a source for circulating Hsp60 in vivo.

Hsp60-binding studies revealed that Hsp60 binds specifically and in a dose-dependent manner to adipocytes, representing the typical characteristics of a ligand-receptor interaction. These findings are consistent with results obtained from the murine adipocyte cell line 3T3-L1 (12). Attempts to characterize the Hsp60 receptor structure(s) on innate immune cells identified TLR2, TLR4, and CD14 as components responsible for the proinflammatory effects of Hsp60 (22–25). Activation of TLR4 but not TLR2, on the other hand, induces the release of Hsp60 from adipocytes. The identity of the Hsp60 receptor complex on human adipocytes, however, remains elusive.

Because Hsp60 can be released from human subcutaneous adipocytes and moreover binds specifically to human adipocytes, autocrine effects of Hsp60 can be assumed. We obtained Hsp60-mediated activation of the MAP kinases ERK1/2, JNK, and p38 and the transcription factor NF-κB in human adipocytes. The activation of similar MAP kinase patterns has been described for other adipokines, such as chemerin and MCP-1 (26), suggesting that different adipokines are capable of inducing similar proinflammatory responses. Moreover, Hsp60 induced secretion of proinflammatory mediators. All these Hsp60-mediated effects may indicate that Hsp60 could be involved in adipose tissue inflammation by both inducing acute proinflammatory signaling and an enhanced release of proinflammatory mediators. The resulting elevated adipokine levels, e.g., for TNF-α, IL-6, or MCP-1, are well described to be obesity-related and may indirectly induce the development of insulin resistance in human adipocytes and other insulin-sensitive tissues (1,27). The first evidence that Hsp60 might contribute to these processes is given, since Hsp60 provoked a significant decrease of the insulin-stimulated Akt and GSK3α/β phosphorylation in human adipocytes (Akt) and SkMCs (Akt, GSK3α/β), thereby triggering the development of an insulin resistance. These findings were confirmed by our results demonstrating significant effects of Hsp60 on IRS-1 phosphorylation as well as Hsp60-mediated impairment of glucose uptake in human SkMCs, indicating Hsp60 as a putative mediator in the development of insulin resistance. Therefore, Hsp60 might have direct effects on insulin signaling and also may indirectly induce insulin resistance by increasing proinflammatory adipokines, known to interfere with insulin signaling.

There are several putative mechanisms by which Hsp60 and cytokine concentrations could reach pathophysiological levels in obese and/or type 2 diabetic patients. Heat shock proteins are involved in the activation of innate immune cells and in the resulting macrophage-infiltration of adipose tissue by the release of chemokines such as MCP-1 (5,6,28). Macrophage infiltration positively correlated with increased adipocyte size and body mass in human subcutaneous tissue, leading to elevated cytokine levels, which finally may contribute to the development of insulin resistance in adipocytes (29,30). Therefore, besides adipose tissue macrophages, adipocytes themselves must be considered as important players in the development of obesity-related insulin resistance (31). However, other mechanisms such as obesity-related hypoxia might contribute to inflammatory processes in adipose tissue. Previously, it was depicted that hypoxia is associated with an increased Hsp60 expression in human vessels (32) and that Hsp60 was translocated to the plasma membrane in the heart (33). Because hypoxia has been reported to occur in obese individuals (34), one might speculate that hypoxic conditions could contribute to elevated Hsp60 expression levels, possibly leading to an increased release of Hsp60 observed in obese individuals.

Several studies indicate that circulating Hsp60 levels are increased not only in patients with type 2 diabetes but also in patients with coronary heart disease (35). In vitro studies further underline that Hsp60 has endocrine effects on cardiomyocytes. Hsp60 was found to be increased early in heart failure accompanied by increased release from cardiomyocytes, where it induces apoptosis via TLR4 (36,37). Furthermore, Hsp60 induces proliferation of vascular smooth muscle cells (38), which might also contribute to cardiovascular disease (39). In the current study, we revealed a positive association between elevated Hsp60 concentrations and blood pressure, which might contribute to the development of cardiovascular diseases. Our study is the first demonstrating that Hsp60 might also be a relevant mediator for skeletal muscle and adipose tissue insulin resistance, thereby contributing to the development of type 2 diabetes. In an acute way, Hsp60 activates proinflammatory signaling cascades in primary human SkMCs similarly to MCP-1 and chemerin (10,26). Prolonged incubation with Hsp60 induces insulin resistance at the level of Akt and GSK3α/β in these cells at both physiological and pathophysiological concentrations. These findings could be verified through Hsp60-mediated impaired glucose uptake in SkMCs and a positive association between Hsp60 concentration and HOMA-IR, as well as a negative correlation with QUICKI. Most interestingly, Hsp60 treatment of SkMCs results in a marked secretion of myokines such as MCP-1, which is associated with skeletal muscle inflammation (40). Increased release of MCP-1 in Hsp60-treated SkMCs is in line with a reported increase in MCP-1 secretion in these cells after stimulation with adipocyte CM (41). Because Hsp60 is released by adipocytes, Hsp60 content in adipocyte CM might at least partly explain its effect on MCP-1 release.

In summary, inflammatory stress induces the release of Hsp60 by human adipocytes, and Hsp60 exerts autocrine/paracrine effects on adipocytes characterized by an increased release of proinflammatory adipokines, increased inflammatory signaling, and insulin resistance. Furthermore, the current study reveals that Hsp60 has endocrine effects on SkMCs, inducing insulin resistance. Our clinical data reveal positive associations of circulating Hsp60 concentrations with BMI, leptin, HOMA-IR, and blood pressure. Therefore, there is rising evidence that circulating Hsp60 levels are increased in obesity, leading to the conception that Hsp60 might represent a novel adipokine involved in adipose tissue inflammation, thereby contributing to the development of insulin resistance in human adipocytes and SkMCs.

This work was supported by the Bundesministerium für Gesundheit and by the Ministerium für Innovation, Wissenschaft, Forschung und Technologie des Landes Nordrhein-Westfalen, in part by a grant from the Bundesministerium für Bildung und Forschung (BMBF) to the German Center for Diabetes Research (DZD e.V.), European Union COST (European Cooperation in Science and Technology), the Commission of the European Communities (Collaborative Project ADAPT [Adipokines as Drug Targets to Combat Adverse Effects of Excess Adipose Tissue]), and the Deutsche Forschungsgemeinschaft (DFG).

No potential conflicts of interest relevant to this article were reported.

T.M. and H.S. researched data and wrote the manuscript. P.Z., A.G., J.K., and S.F. researched data. D.M.O., P.P., and J.R. performed clinical research. T.M., H.S., and C.H. designed and initiated the experimental procedures. M.R., J.E., and C.H. contributed to discussion and reviewed the manuscript. C.H. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Professor Jutta Liebau, Department of Plastic Surgery, Florence-Nightingale-Hospital Düsseldorf, and Dr. Christoph Andree, Department of Plastic Surgery, Sana-Hospital Düsseldorf-Gerresheim, for support in obtaining adipose tissue samples. The technical assistance of Jutta Brüggemann, Angelika Horrighs, Andrea Cramer, and Manuela Elsen is gratefully acknowledged.