Potential conflict of interest: Nothing to report.
This study was supported by Herzfeldersche Familienstiftung and Fonds zur Förderung der Wissenschaftlichen Forschung (projects 17630-B12 and 19920-B12).
Fibroblast growth factors (FGFs) and their high-affinity receptors [fibroblast growth factor receptors (FGFRs)] contribute to autocrine and paracrine growth stimulation in several nonliver cancer entities. Here we report that at least one member of the FGF8 subfamily (FGF8, FGF17, and FGF18) was up-regulated in 59% of 34 human hepatocellular carcinoma (HCC) samples that we investigated. The levels of the corresponding receptors (FGFR2, FGFR3, and FGFR4) were also elevated in the great majority of the HCC cases. Overall, 82% of the HCC cases showed overexpression of at least one FGF and/or FGFR. The functional implications of the deregulated FGF/FGFR system were investigated by the simulation of an insufficient blood supply. When HCC-1.2, HepG2, or Hep3B cells were subjected to serum withdrawal or the hypoxia-mimetic drug deferoxamine mesylate, the expression of FGF8 subfamily members increased dramatically. In the serum-starved cells, the incidence of apoptosis was elevated, whereas the addition of FGF8, FGF17, or FGF18 impaired apoptosis, which was associated with phosphorylation of extracellular signal-regulated kinase 1/2 and ribosomal protein S6. In contrast, down-modulation of FGF18 by small interfering RNA (siRNA) significantly reduced the viability of the hepatocarcinoma cells. siRNA targeting FGF18 also impaired the cells' potential to form clones at a low cell density or in soft agar. With respect to the tumor microenvironment, FGF17 and FGF18 stimulated the growth of HCC-derived myofibroblasts, and FGF8, FGF17, and FGF18 induced the proliferation and tube formation of hepatic endothelial cells. Conclusion: FGF8, FGF17, and FGF18 are involved in autocrine and paracrine signaling in HCC and enhance the survival of tumor cells under stress conditions, malignant behavior, and neoangiogenesis. Thus, the FGF8 subfamily supports the development and progression of hepatocellular malignancy. (HEPATOLOGY 2011)
Hepatocellular carcinoma (HCC) is the third-leading cause of cancer deaths worldwide.1 Important risk factors for this disease are persistent infections with hepatitis viruses and chronic steatohepatitis due to ethanol abuse and obesity, which contribute to the increasing incidence of HCC in industrialized countries. Most HCC cases have a very poor prognosis. This dramatic situation stems in part from the rare detection of tumors at early stages and from the high inherent resistance of HCC to chemotherapeutic agents. Much hope, therefore, is focused on obtaining a better understanding of the disturbed signaling pathways relevant to this disease in order to develop new preventive, diagnostic, and therapeutic options.2
HCC often emerges with a background of persistent liver injury, inflammation, and hepatocellular proliferation, which are characteristic of chronic hepatitis and cirrhosis.3 It is assumed that these liver diseases induce increasing aberrations in cellular signaling networks, which lead to the appearance of early precursor lesions of cancer.4, 5 These lesions overrespond to growth stimulatory cytokines and show enhanced proliferation and insufficient elimination of cells by apoptosis.4, 6, 7 This appears to select for premalignant and malignant cell populations with increasingly dysregulated downstream signaling pathways, such as the Ras, phosphoinositide 3-kinase, and wnt pathways.8-10 Hepatocarcinogenesis is dependent on the development of a tumor-specific microenvironment of inflammatory cells, small vessels, myofibroblast (MFs), and extracellular matrix components.11, 12 These epithelial-mesenchymal interactions in early and advanced stages of hepatocarcinogenesis are driven by various growth factor systems. In particular, the signaling pathways induced by erythroblastic leukemia viral oncogene homolog receptors, hepatocyte growth factor, and insulin-like growth factor have been determined to contribute to the development of liver tumors and their stroma.3, 7, 13 However, our current understanding of the complex tumor-stroma interactions is far from complete.
The fibroblast growth factor (FGF) system is known to be widely involved in nonliver carcinogenesis.14-16 It stimulates the growth of premalignant and malignant cells, enhances neoangiogenesis, and modulates tumor cell adhesion and migration.6, 14-18 This system is composed of at least 23 ligands, which are grouped into 7 subfamilies and signal by activating tyrosine kinase receptors encoded by four genes [fibroblast growth factor receptor 1 (FGFR1), FGFR2, FGFR3, and FGFR4].14 FGF8, FGF17, and FGF18 constitute the FGF8 subfamily and share a high sequence homology and evolutionary relationship. Alternative splicing may generate four FGF8 isoforms. These FGF8 variants, FGF17, and FGF18 are presumed to activate IIIc isoforms of FGFR2 and FGFR3 as well as FGFR4.19 In the adult human organism, FGF8 expression is largely restricted to steroid hormone target tissues and occurs at higher levels in hormone-responsive tumors, such as prostate and breast cancer.14, 20 FGF17 is also up-regulated in prostate cancer and is an even more potent mitogen for the cancer cells than FGF8.21 Synovial sarcoma and ovarian and colon cancer are tumor entities showing frequent overexpression of FGF18.16, 22, 23 This growth factor also occurs at considerable levels in the vascular tissue. In the liver and other organs, endothelial cells are a source of FGF18 and contribute to paracrine growth stimulation of hepatocytes (S.S., unpublished data, 2010).24 Moreover, the hepatic overexpression of FGF18 in transgenic mice or the systemic administration of FGF18 induces hepatocyte proliferation and significant increases in liver weight.25 Despite the obvious importance of the FGF8 subfamily in several cancers, detailed and mechanistic studies of the role of this subfamily in the pathogenesis of HCC are not available.
In a parallel study, we found that several FGFs, including FGF18, stimulate DNA replication preferentially in initiated/premalignant hepatocytes isolated from rat livers. Furthermore, FGF18 was up-regulated in rat hepatocellular adenoma and carcinoma; this was the first evidence of the gain of autocrine function for this specific FGF (S.S., unpublished data, 2010). Here we investigated the effects of FGF18 and the other two FGF8 subfamily members on the growth and malignant behavior of human hepatic malignancies. Clinical material from HCC cases was used to study the expression of FGF8 subfamily members. For functional studies, we chose epithelial and mesenchymal cells established from the HCC cases.12 We show for the first time that FGF8 subfamily members are frequently up-regulated in HCC and have important autocrine and paracrine functions in advanced stages of human hepatocarcinogenesis.
Patients were subjected to surgical resection of their HCCs. Written, informed consent was obtained from each patient. Further details are provided in Supporting Information Table 1. No donor organs from executed prisoners or other institutionalized persons were used. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki; this was reflected by the approval of the ethics committee of the Medical University of Vienna. Sources of samples for healthy liver tissue are presented in Supporting Information Table 2.
Culture and Treatment of Cells.
The human cell lines HepG2 and Hep3B were obtained from the American Type Culture Collection (Rockville, MD). The HCC-derived epithelial hepatocarcinoma line (HCC-1.2) and myofibroblastoid cell lines (MF-12, MF-14, and MF-16) were recently established. A detailed characterization of all the cell lines has been provided elsewhere.12 Stock solutions of human recombinant FGF8 and FGF18 (BioVision, Old Middlefield, CA) and FGF17 (BioSource, Camarillo, CA) were prepared according to the manufacturers' instructions. Aliquots were added to the medium to provide the final concentrations, as indicated later.
Analyses of Cell Proliferation and Cell Death.
The number of viable cells was determined with the 3′-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay, which quantifies the degree of dye reduction by functional mitochondria (EZ4U, Biomedica, Vienna, Austria). DNA synthesis was assayed by [3H]-thymidine incorporation as described.6 For the determination of apoptosis, cells were incubated in 0.5 mL of a medium containing 0.6 μg/mL propidium iodide (Sigma, St. Louis, MO), and were analyzed with a FACSCalibur system (Becton-Dickinson, San Jose, CA).
Forty-eight hours after transfection (described later), cells were plated at a low density in a medium containing 10% fetal calf serum (FCS) or were suspended in 0.3% agar (Sigma) and 20% FCS/Roswell Park Memorial Institute (RPMI) medium and were seeded onto 0.6% agar and 20% FBS/RPMI medium. The numbers of clones were determined in at least two dishes per group and time point.
Tube Formation Assay.
Rat endothelial cells were isolated as described and were seeded onto growth factor–reduced Matrigel (Becton Dickinson, Franklin Lakes, NJ).7 Six hours after the addition of FGFs, the extent of tube formation was quantified by the measurement of the tube length with ImageJ software (National Institutes of Health, Bethesda, MD).
Total RNA, which was extracted from tissue specimens or cell lines, was subjected to quality control (BioAnalyzer 2100, Agilent, Santa Clara, CA). Samples with RNA integrity numbers greater than 8 (for cell lines) and greater than 7 (for tissue samples) were used for complementary DNA synthesis. Complementary DNA was analyzed with an ABI-Prism 7500 detection system (Applied Biosystems, Foster City, CA) and TaqMan-based assays (Applied Biosystems; see Supporting Information Table 3) as described.6-8 All data were analyzed in duplicate. Levels of messenger RNA (mRNA) were normalized to β2-microglobulin with ABI-Prism 7500 software and to a set of further reference genes with the applet geNorm (http://medgen.ugent.be/∼jvdesomp/genorm; see Supporting Information Table 1). Similar results were obtained with both approaches.
Small Interfering RNA (siRNA).
siRNA targeting human FGF18 (Silencer Select AM4392420) and nonsilencing scrambled small interfering RNA (siSCR; Silencer Select AM4390843) were obtained from Applied Biosystems. siRNAs were transfected at either 10 (Hep3B) or 20 nmol (HCC-1.2 and HepG2) with siLentFect (Bio-Rad, Hercules, CA) according to the manufacturer's instructions. The cells were incubated for at least 24 hours until the analyses were performed.
Antisera are listed in Supporting Information Table 3. For immunostaining, tissue samples were fixed in 10% buffered formalin, processed, and immunostained as described.6, 7 For immunoblotting, cells were suspended in a radio immunoprecipitation assay buffer [50 mM trishydroxymethylaminomethane (Tris)/hydrochloric acid, pH 7.4, 500 mM sodium chloride, 1% Igepal CA-630, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and 0.05% NaN3] containing protease inhibitors (Complete, Roche Diagnostics, Mannheim, Germany) and phosphatase inhibitors (1 mM orthovanadate and 10 mM sodium fluoride). After sonification, the solution was centrifuged at 10,000g for 10 minutes. To detect secreted FGF18, conditioned cell supernatants were collected after 48 hours of culture. FBS was admixed to the supernatant of serum-free cultures for a final concentration of 10% to adjust the difference to supernatants derived from conventional cultures. Purification was carried out through the addition of the cell supernatant to 50 μL of prewashed Heparin-Sepharose CL-6B (Pharmacia, Uppsala, Sweden). The slurry was incubated for 24 hours at 4°C with rotation, washed three times with 0.15 M sodium chloride and 0.01 M Tris/hydrochloric acid (pH 7.5), and washed twice with 0.45 M sodium chloride and 0.02 mM Tris/hydrochloric acid. Proteins were eluted by the addition of 20 μL of a loading buffer and were applied to 15% sodium dodecyl sulfate–polyacrylamide gels. After the immunoblotting, the detection of the signal followed published protocols.7,8
Up-Regulation of FGF8, FGF17, and FGF18 and Their Respective FGFRs in HCC.
We studied the expression level of FGF8, FGF17, and FGF18 by qRT-PCR. The primers applied to determine FGF8 transcripts covered the four splice variants (a, b, e, and f) described so far.14 In comparison with the surrounding tissue, there was an at least 2-fold up-regulation of one or more FGF8 subfamily members in 59% of the HCC cases investigated (Fig. 1 and Supporting Information Table 1). Interestingly, the expression levels of FGF8, FGF17, and FGF18 were considerably lower in healthy livers versus the tumor surrounding this stemmed mostly from heavily diseased organs containing numerous cirrhotic (premalignant) nodules (Supporting Information Table 2).
The FGF8, FGF17, and FGF18 protein contents of HCC were analyzed by immunohistochemistry. The extent and intensity of the stains were comparable to the transcript levels; this is exemplified by representative cases in Fig. 1 and Supporting Information Table 4. The positive immunostains of human HCC were most prominent in the malignant hepatocytes (Fig. 1C). Furthermore, recently established hepatocarcinoma cell lines expressed FGF8, FGF17, and FGF18 at mRNA levels close to those in HCC (Fig. 2 and Supporting Information Table 2). These findings suggest that the epithelial compartment in HCC is the major source of the elevated levels of the FGF8 subfamily.
The four FGF8 isoforms as well as FGF17 and FGF18 bind with considerable affinity to FGFR2, FGFR3, and FGFR4.19 We applied primers for qRT-PCR to detect all important splice variants of these receptors and found many HCC cases with elevated mRNA levels in comparison with the surrounding tissue (Fig. 1A and Supporting Information Table 1). Accordingly, 56% of the investigated HCC cases showed at least 2-fold up-regulation of one or more FGFRs. The immunostaining of FGFR2, FGFR3, and FGFR4 occurred predominantly in the epithelial HCC cells and matched the transcript levels well (Fig. 1C and Supporting Information Table 4). In conclusion, 82% of the studied HCC cases showed up-regulation of at least one FGF8 subfamily member and/or one FGFR. In every third tumor, the enhanced expression of at least one FGF and one FGFR coincided.
Serum Withdrawal and Simulation of Hypoxia Induce FGF8, FGF17, and FGF18 Expression.
Rapidly growing tumors such as HCC often suffer from an insufficient blood supply. Therefore, we asked whether the up-regulation of FGFs in HCC may be a response to a lack of serum and insufficient oxygen. When HCC-1.2, HepG2, and Hep3B cells were either serum-starved or kept under the hypoxia-mimetic drug deferoxamine mesylate, the transcript levels of FGF8, FGF17, and FGF18 were increased up to 40-fold within 48 hours above the already notable levels of controls (Fig. 2A,B and Supporting Information Table 2). Similar increases were obtained when the cells were kept under 1% oxygen (Supporting Information Fig. 1). Immunoblotting revealed that the up-regulated FGF18 mRNA in the serum-starved cells was paralleled not by an increased intracellular protein level (not shown) but instead by an elevated secretion of this growth factor to the culture supernatant. We estimated that 5 × 105 cells released at least 2 ng of FGF18 per mL of the medium when they were kept serum-free for 48 hours (Fig. 2C and Supporting Information Fig. 2).
FGF8, FGF17, and FGF18 May Act as Autocrine Survival Factors for Malignant Hepatocytes.
We asked whether the elevated production of FGF8 subfamily members due to a shortage of serum and oxygen confers any advantage to HCC-1.2 or Hep3B cells. When they were subjected to serum withdrawal for 48 hours, the apoptotic activity of the cells increased 6.7 ± 1.7-fold and 4.3 ± 1.1-fold, respectively [determined by fluorescence-activated cell sorting (FACS) analyses; n = 3]. As a result, 96 hours of serum withdrawal reduced the numbers of viable HCC-1.2 cells (0.3 ± 0.1-fold) and Hep3B cells (0.6 ± 0.1-fold) with respect to unstarved controls (determined by the EZ4U assay; n = 3). This indicates that the amounts of secreted FGF18 and presumably other FGF8 subfamily members in the medium were not sufficient for the cells to cope completely with the proapoptotic stimulus of serum withdrawal. The addition of 10 ng of recombinant FGF8, FGF17, or FGF18 per mL of the medium increased the viability of the starved cells significantly, suppressed their apoptotic activity, and enhanced the fraction of HCC-1.2 cells in the S-phase or G2/M-phase of the cell cycle (Fig. 3). In Hep3B cells, however, the rescue effect of the FGFs may predominantly be due to the inhibition of apoptosis because significant effects on the cell cycle were not evident. In conclusion, the increased production of FGF8 subfamily members with a lack of serum and and/or oxygen may enhance the survival of malignant hepatocytes.
Survival Factor Activity of FGF8, FGF17, and FGF18 Involves the Mitogen-Activated Protein (MAP) Kinase and Protein Kinase B (AKT) Pathway.
In this study, serum deprivation clearly reduced the levels of phosphorylated extracellular signal-regulated kinase (pERK) and phosphorylated S6 (pS6) and elevated the level of phosphorylated glycogen synthase kinase 3β (pGSK3β) in both HCC-1.2 and Hep3B cells (representative data are shown in Fig. 4). This may reflect a lack of extracellular signal-regulated kinase 1 (ERK1) stimulation by the growth factor–depleted serum-free medium, reduced MAP kinase signaling and translational activity in the cells, and concomitant inactivation of glycogen synthase kinase 3β (GSK3β). Treatment of the starved cells with FGF8, FGF17, or FGF18 partly reversed the effect of serum withdrawal and elevated the level of pERK within minutes. FGF8 instead reduced pGSK3β and elevated pS6, whereas FGF17 and FGF18 left the phosphorylated form of S6 and pGSK3β more or less unchanged.
Silencing of FGF18 Reduces Viability and Clone Formation.
To assess the role of FGF18 in the survival and malignant behavior of HCC-1.2, HepG2, and Hep3B cells, the expression of this growth factor was knocked down by siRNA. As demonstrated in Fig. 5, small interfering RNA targeting fibroblast growth factor 18 (siFGF18) somewhat elevated apoptotic activity, significantly reduced viability, and impaired the cells' potential to form clones at a low density (clonogenicity) and in soft agar. siSCR exerted no significant effect on FGF18 expression, viability, or clone formation (not shown). This is strong evidence that FGF18 is of essential importance for the survival and tumorigenic phenotype of the cells.
FGF8 Subfamily Members Are Involved in Tumor-Stroma Communication.
We asked whether FGF8, FGF17, and FGF18 also affect cells of the tumor stroma. HCC-derived MFs synthesize considerable amounts of vascular endothelial growth factor (vEGF), one of the key factors for neoangiogenesis.12 We found that FGF17 and FGF18 stimulate replicative DNA synthesis in MF cells. A similar effect was evident on the DNA replication of endothelial cells that were isolated from human tumor-bearing livers (Fig. 6). Furthermore, all three FGFs induced tube formation of endothelial cells, which is a further necessary step in neoangiogenesis. This suggests that FGF8 subfamily members favor the formation of new blood vessels in HCC directly and indirectly via the multiplication of vEGF-producing MFs.
Here we show for the first time that FGF8, FGF17, and FGF18 have more or less equal potency in enhancing neoangiogenesis and the aggressive behavior of malignant hepatocytes. Accordingly, at least one of these FGFs was up-regulated in the majority of the investigated HCC cases. This implies that the FGF8 subfamily members are crucial components in autocrine and paracrine loops supporting the autonomous growth of advanced stages of hepatocarcinogenesis, as outlined in the following.
Up-Regulation of FGF8 Subfamily Members in HCC May Involve the wnt Pathway and Hypoxia-Inducible Transcription Factors.
In this study, we found pronounced overexpression of FGF18 in a subset of human HCC cases. The human FGF18 gene harbors T cell factor/lymphoid enhancer-binding factor binding sites in the promoter region. Accordingly, FGF18 transcription is under the control of the β-catenin T cell factor/lymphoid enhancer-binding factor complex, as shown recently by our group and others.16, 27 In human HCC, the wnt/wingless signaling cascade often is activated by mutations in AXIN1, AXIN2, or the gene coding for β-catenin [catenin (cadherin-associated protein) β1] (CTNNB1) and/or through epigenetic silencing of wnt antagonists, such as the secreted frizzled-related protein.9, 10, 13 These disturbances in the wnt signaling cascade may contribute to the observed up-regulation of FGF18 in human HCC.
Here we found that the withdrawal of serum or oxygen is a potent inducer of all FGF8 subfamily members in HCC-1.2, HepG2, and Hep3B cells. These regimens simulate the conditions in rapidly expanding HCC with an inadequate blood supply. Generally, such conditions alter signaling cascades and gene expression patterns of the affected cells and lead to increased neoangiogenesis and glycolysis and decreased mitochondrial respiration. In our experiments, serum deprivation clearly elevated the phosphorylation of GSK3β at serine 9 in the hepatocarcinoma cells, and this may lead to reduced phosphorylation and degradation of β-catenin and increase the probability of β-catenin entering the nucleus and activating the transcription of FGF18. The molecular mechanisms underlying the induction of FGF8 and FGF17 by serum withdrawal are still unclear.
In comparison with serum withdrawal, hypoxia is a more specific stimulus transduced by members of several transcription factor families, including the hypoxia inducible factor (HIF), aryl hydrocarbon receptor (AHR), E twenty-six (ETS), and metal-responsive transcription factor (MTF) families. FGF8, FGF17, and FGF18 contain multiple ETS- and MTF-responsive elements in the promoter region, and FGF8 and FGF17 are additionally equipped with HIF- and AHR-responsive elements (http://www.genomatrix.de). Accordingly, deferoxamine mesylate, a known activator of HIF, induced FGF8 subfamily members in hepatocarcinoma cells.28 In conclusion, the enhanced expression of FGF8, FGF17, and/or FGF18 in human HCC may result from deregulation of the wnt pathway and/or an inadequate blood supply.
FGF8 Subfamily Members Support the Survival and Aggressive Behavior of Malignant Hepatocytes.
Serum withdrawal greatly enhanced apoptotic activity and reduced the viability of hepatocarcinoma cells, whereas the addition of FGF8, FGF17, or FGF18 rescued cells from apoptosis. These effects appear to involve the ERK and AKT/mammalian target of rapamycin pathway.29, 30 A great variety of extracellular signals, including growth factors, activate ERK1 through phosphorylation of threonine 202 and tyrosine 204. The AKT/mammalian target of rapamycin pathway transmits stimulatory cues for protein synthesis and cell growth by inducing the phosphorylation of downstream substrates, including the S6 ribosomal protein at serines 235, 236, 240, and 244. This leads to an elevated translation, particularly of mRNAs involved in cell cycle progression and translation machinery. AKT also phosphorylates BAD (B cell lymphoma 2–associated death promoter) on serine 136; this mechanism may render various cell types resistant to apoptosis.29 It appears that in response to FGFs, the serum-starved hepatocarcinoma cells immediately reactivated protein synthesis and growth; this was reflected by rapidly elevated levels of pERK and pS6.
The impact of the FGF8 subfamily members on the survival and growth of HCC-1.2, HepG2, and Hep3B cells was also proven by the blockade of signaling via siFGF18, which elevated apoptotic activity, reduced viability, and impaired clone formation. Similar effects were evoked by the introduction of kinase-defective FGFR3 or FGFR4 into the hepatocarcinoma cells (unpublished data, 2010). FGFR3 and FGFR4 are the main receptors for the FGF8 subfamily members, and these dominant-negative constructs are expected to replace the endogenous receptors and/or impair their function via heterodimerization. This indicates that FGF18 and probably also the other FGF8 subfamily members contribute to the malignant phenotype of the cells. Antitumor effects of blocking FGF-induced signaling have been shown for several nonliver cancer entities.31 Approaches include small molecule inhibitors of the FGFR kinase activity, antibodies neutralizing FGFs or FGFRs, and antisense approaches. Our data suggest that the blockade of FGF8 subfamily members and/or their receptors might offer promising therapeutic options for malignancies of hepatocellular origin.
FGF8 Subfamily Members Increase the Proliferation of Tumor Stroma Cells.
We found that FGF8, FGF17, and FGF18 increase the replication of MFs. These cells have been recently established from HCC and are an essential component of the tumor stroma.12 MFs themselves produce FGF18, and the addition of this growth factor to confluent HCC-1.2 cells significantly accelerated the scratch closure time (unpublished observations, 2010). This may explain the recent observation that the supernatant of MF cells increased the motility of HCC cells.12 Because MF cells are a rich source of vEGF, FGF8 subfamily members may induce the formation of new blood vessels in HCC indirectly by increasing the number of MF cells.
The hepatic endothelium harbors several unique features and functions that may also apply to this cell type in HCC.32 We therefore used endothelial cells isolated from this organ and found that these cells replicated and/or differentiated into tubes when they were exposed to FGF8, FGF17, and FGF18. These FGFs may act in a paracrine way on the endothelial cells within HCC when they are released from the malignant hepatocytes in response to an insufficient blood supply. On the other hand, FGF18 is expressed in the hepatic sinus endothelium and may also contribute to neoangiogenesis in an autocrine fashion (S.S., unpublished data, 2010).
In conclusion, FGF8, FGF17, and FGF18 seem to act as important driving forces for malignant behavior and neoangiogenesis in advanced stages of hepatocarcinogenesis. Thus, the role of the FGF8 subfamily in the formation and progression of HCC deserves further and intense research efforts.
The excellent technical assistance of Krystyna Bukowska, Helga Koudelka, and Birgit Mir-Karner is gratefully acknowledged.