Paracrine Regulation of Germinal Center B Cell Adhesion through the c-Met–Hepatocyte Growth Factor/Scatter Factor Pathway

Mouse monoclonal antibodies used were anti– c-Met, DO24 (IgG2a) (Upstate Biotechnology, Lake Placid, NY); anti-CD38, OKT-10 (IgG1) (American Type Culture Collection [ATCC], Rockville, MD); FITC-conjugated anti-CD38, HIT2 (IgG1) (Caltag Laboratories, Burlingame, CA); biotin-conjugated anti-CD38, HIT2 (IgG1) (Caltag); anti-β1 integrin (CD29), 4B4 (IgG1) (Coulter Immunology, Hialeah, FL); anti-α4 integrin (CD49d), HP2/1 (IgG1) (Immunotech, Marseille, France); anti-α5 integrin (CD49e), SAM-1 (IgG2b) (40) (a gift from A. Sonnenberg, NKI, Amsterdam, The Netherlands); anti-α4β7, Act-1 (IgG1) (41) (a gift from A. Lazarovits, University of Western Ontario, London, Canada); anti-ICAM-1 (CD54), RR1/1 (IgG1) (42) (a gift from T. Springer, Harvard University, Boston, MA); anti-HGF/SF, 24612.111 (IgG1) (R&D Systems, Abingdon, UK); anti-CD3, OKT-3 (IgG2a) (ATCC); anti-DRC-1, R4/23 (IgM) (DAKO, Glostrup, Denmark); anti-CD19, HD37 (IgG1) (DAKO); and anti-phosphotyrosine, PY-20 (IgG2b) (Affiniti, Nottingham, UK). Polyclonal antibodies used were rabbit anti–c-Met, C-12 (IgG) (Santa Cruz Biotechnology, Santa Cruz, CA); goat anti-HGF/ SF (R&D Systems); FITC-conjugated rabbit anti-IgD (DAKO); RPE-conjugated goat anti–mouse (Southern Biotechnology, Birmingham, AL); AP-conjugated goat anti–mouse (total, IgG1, or IgG2a) (Southern Biotechnology); biotin-conjugated rabbit anti– mouse (DAKO); biotin-conjugated goat anti–mouse (DAKO); biotin-conjugated rabbit anti–goat (Vector Laboratories, Burlingame, CA); HRP-conjugated goat anti–rabbit (DAKO); and HRPconjugated goat anti–rabbit (DAKO). In addition we used a rat monoclonal anti-CD77, 38.13 (IgM) (43) (provided by J. Wiels, Institute Gustave-Roussy, Villejuif, France); and RPE-Cy5-conjugated streptavidin (DAKO).

The epidermoid carcinoma cell line A431, the lung fibroblast cell line MRC-5, and the B cell lines Raji, Namalwa, Daudi, Ramos, JY, and Nalm-6 were obtained from ATCC and cultured in RPMI 1640 (Gibco BRL/Life Technologies, Paisley, UK) supplemented with 10% FCS (Integro, Zaandam, The Netherlands). The Burkitt's lymphoma line EB4B (44) was provided by R. Jefferis (University of Birmingham, Edgbaston, UK) and cultured in 10% FCS/RPMI 1640.

B cells were isolated as described previously (29). Total B cell fractions were >97% pure as determined by FACS^® analysis.

B cells were cultured in Iscove's medium (Gibco BRL/Life Technologies) containing 10% FCS, 0.5% BSA, 50 μg/ml human transferrin (Sigma, Bornem, Belgium) and 5 μg/ml bovine pancreas insulin (Sigma). Some media were supplemented with 50 ng/ml phorbol-12-myristate-13-acetate (PMA; Sigma).

For CD40 ligation, B cells were cultured on irradiated (7,000 rad) CD40L-transfected or, as a control, wild-type L cells (45) (provided by J. Banchereau, Schering Plough, Dardilly, France). In specific experiments, culture media were supplemented with either pansorbin cells of Staphylococcus aureus strain Cowan I (0.002%; Calbiochem Novabiochem, La Jolla, CA), rabbit anti–human Igcoated beads (2 μg/ml) (BioRad Laboratories, Hercules, CA), recombinant human IL-2 (100 U/ml) (Eurocetus, Amsterdam, The Netherlands), recombinant human Il-4 (100 U/ml) (Genzyme Diagnostics, Cambridge, MA), or recombinant human IL-6 (1,000 U/ml) (CLB, Amsterdam, The Netherlands).

Tonsillar T cells were isolated as described for the B cell isolation, except that after the second Ficoll-Isopaque density gradient centrifugation the pellet was collected, washed, and resuspended in shock medium. The remaining B cells were removed by using a MACS magnetic cell separator (Miltenyi Biotec, Bergisch Gladbach, Germany) using anti-CD19. The T cell fraction was >98% pure as determined by FACS^® analysis.

Tonsillar stromal cells were isolated as described (46). The cells were cultured in 100-mm petri dishes (Costar, Cambridge, MA) containing 10% FCS/RPMI 1640. After 4 d nonadherend cells were removed.

FDC were isolated as described (29). The cells were cultured in Iscove's medium containing 10% Fetal Clone I serum (HyClone Laboratories, Logan, UT). These FDC-enriched cell cultures contained 10–15% DRC-1–positive cells.

c-Met transfected Namalwa cells (Nam^met) were obtained by electroporating Namalwa cells with the eukaryotic expression plasmid pA71d containing full-length c-met cDNA (a gift from G. Hartmann and E. Gherardi, University of Cambridge, Cambridge, UK). After 2 d in culture, transfectants were selected in culture medium containing 250 μg/ml hygromicin (Sigma). c-Met positive cells were sub-cloned by using a FACStar^plus® flow cytometer (Becton Dickinson, Mountain View, CA).

For analysis of tyrosine phosphorylation of the c-Met protein, cells were incubated overnight in serum-free RPMI 1640. Nam^met or EB4B cells were incubated in serum-free RPMI 1640 in the presence or absence of 200 ng/ml HGF/SF (R&D Systems). After 5 min at 37°C the cells were solubilized in ice-cold 2× lysis buffer containing 20 mM Tris-HCl (pH 8), 250 mM NaCl, 20% glycerol, 2% NP-40, 20 μg/ml aprotinin (Sigma), 20 μg/ml leupeptin (Sigma), 4 mM sodium orthovandate (Sigma), 10 mM EDTA, and 10 mM NaF. After 1 h at 4°C the insoluble nuclear material was removed by centrifugation at 1 × 10⁴ g at 4°C for 20 min after which the supernatant was precleared with protein A–Sepharose CL-4B (Pharmacia Biotech) for 45 min at 4°C. c-Met was precipitated with rabbit anti–c-Met coupled to protein A–Sepharose at 4°C for at least 2 h. The immune complexes were washed with lysis buffer and diluted in Laemmli sample buffer containing final concentrations of 62.5 mM Tris-HCl (pH 6.8), 2% SDS, 10% glycerol, 100 mM 2-mercaptoethanol (BioRad Laboratories), and 0.001% bromophenol blue. After boiling for 5 min, the samples were subjected to 8% SDS-PAGE. Western blotting was performed as described previously (48).

For analysis of c-Met in total cell lysates, cells were lysed in 50 mM Tris-HCl (pH 8), 150 mM NaCl, 1% NP-40, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 1 mM sodium orthovandate, 2 mM EDTA, and 5 mM NaF for 1 h at 4°C. After centrifugation at 1 × 10⁴ g and 4°C for 20 min, the supernatant was diluted in Laemmli sample buffer, boiled for 5 min and subjected to 8% SDS-PAGE. Western blotting was performed as described previously (48).

Expression of c-Met on tonsillar B cell subpopulations was studied using a triple staining technique (47). Staining was measured by using a FACSCalibur^® flow cytometer (Becton Dickinson).

Expression of c-Met in tonsillar tissue was analysed by single and double staining. For single staining cryostat tonsil sections were fixed in acetone for 10 min, washed in PBS and preincubated with 10% normal goat serum (Sera Lab, Sussex, UK) in PBS for 15 min. After incubating with the primary antibody for 1 h, endogenous peroxidases were blocked with 0.1% NaN₃, 0.3% H₂O₂, PBS for 10 min. Subsequently, the sections were stained with biotin-conjugated rabbit anti–mouse for 30 min, followed by an incubation with HRP-conjugated avidin–biotin complex for 30 min. Substrate was developed with 3,3-amino-9-ethylcarbazole (Sigma) for ∼10 min. Tissue sections were counterstained with Haematoxylin (Merck, Darmstadt, Germany).

Double staining was performed as described for the single staining, except that a cocktail of primary antibodies was used, which was detected by either a cocktail of AP-conjugated goat anti–mouse and HRP-conjugated goat anti–rabbit, or a cocktail of AP-conjugated goat anti–mouse IgG2a and HRP-conjugated goat anti–mouse IgG1. The second color was developed with Fast Blue BB (Sigma) for ∼10 min.

96-well flat-bottom plates (Costar) were coated overnight with 5 μg/ml human fibronectin (CLB) or 0.2 μg/ml recombinant human sVCAM-1 (R&D Systems) at 4°C. After blocking the plates with 4% BSA, RPMI 1640 (2 h at 37°C), B cells that had been pre-incubated with HGF/SF (R&D Systems), in the presence or absence of monoclonal antibodies, for 30 min at 37°C, were added. Then, the plates were centrifuged (3 min 800 rpm, no brake) and incubated at 37°C for 25 min. After washing the wells, the bound cells were fixed with 10% neutral buffered formalin solution (Sigma) and stained with Giemsa (Merck). Bound cells were quantified by using a color CCD camera (Sony) and NIH Image 1.60 software on an Apple Quadra 840AV.

96-well EIA/RIA plates (Costar) were coated overnight with mouse anti- HGF/SF immunoglobulins at 4°C. Then, the plates were washed and blocked with 4% BSA, PBS for 1 h at 37°C. Next, the wells were incubated with culture supernatants or with a HGF/SF concentration series for 2 h at 37°C, followed by an incubation with goat anti–c-Met immunoglobulins for 1 h at 37°C. Subsequently, the wells were incubated with biotin-conjugated rabbit anti–goat immunoglobulins for 60 min at 37°C followed by HRP-conjugated avidin–biotin complex (DAKO) for 1 h at 37°C. Substrate was developed with 1,2-phenylenediamine (Fluka Chemica, Buchs, Switzerland) in 50 mM KH₂PO₄, 50 mM Na₂HPO₄·2H₂O (pH 5.4) containing H₂O₂. The reaction was stopped with 1 N H₂SO₄ and the results were analysed at 492 nm using a microplate reader (BioRad Laboratories).

Total RNA was isolated with RNAsol (Cinna/Biotex Laboratories, Houston, TX) according to manufacturers description. First-strand cDNA synthesis was performed on total RNA by a standard reverse transcription reaction, using Moloney leukemia virus reverse transcriptase (Gibco BRL/Life Technologies) and p(dN)₆ random hexamers (Pharmacia Biotech). PCR was performed with Taq DNA Polymerase (Gibco BRL/Life Technologies), 200 μM dNTPs (Pharmacia Biotech) and 1.5 mM MgCl₂ in 1× PCR Buffer (both Gibco BRL/ Life Technologies). Primers used were HGF-1 (5′-CGACAGTGTTTCCCTTCTCG-3′) in combination with HGF-3 (5′-GGTGGGTGCAGACACAC-3′), or 5′β2M (5′-ATCCAG CGTACTCCAAAGATT-3′) in combination with 3′β2M (5′-CATGTCTCGATCCCACTTAAC-3′). PCR was started with a 5 min denaturation step at 95°C, after which amplification was performed in 35 cycles of denaturation at 95°C for 30 s, annealing at 60°C for 1 min and elongation at 72°C for 2 min. After a final elongation step for 10 min at 72°C, samples were cooled on ice and analysed by electrophoresis in a 1.5% agarose TBE gel containing ethidium bromide.

Expression of c-Met by human tonsillar B cells and by a panel of B cell lines was assessed by Western blotting and by FACS^® analysis. On Western blot, c-Met expression was hardly detectable in freshly isolated tonsillar B cells, but we observed a strong induction of c-Met (and the c-Met precursor [pre c-Met]) upon stimulation with the phorbolester PMA (Fig. 1,A). Furthermore, constitutive expression of c-Met was found in the Burkitt's lymphoma cell lines Raji and EB4B, but not in the Burkitt's lymphoma cell line Namalwa (Fig. 1,B) nor in the B cell lines Daudi, Ramos, JY, or Nalm-6 (data not shown). As positive controls, c-Met expression of the epidermoid carcinoma cell line A431 and of Namalwa cells stably transfected with c-Met (Nam^met) are shown (Fig. 1 B).

The above findings clearly show that B cells can express c-Met and, hence, might potentially be triggered via the HGF/SF–c-Met pathway. To demonstrate that the c-Met receptor on B cells can indeed be functionally activated by HGF/SF, we studied c-Met receptor phosphorylation on tyrosine residues in response to HGF/SF. As is shown in Fig. 2, HGF/ SF stimulation of EB4B cells as well as of Nam^met B cells resulted in an enhanced tyrosine phosphorylation of c-Met. This indicates that the HGF/SF–c-Met pathway on B cells is capable of signaling.

To investigate whether c-Met induction is a physiological phenomenon, that occurs also during antigen-specific B cell differentiation in vivo, we assessed the expression of c-Met on human tonsillar B cell subsets using FACS^® triple staining. The subsets studied, recently defined by Pascual et al. (13), were: the naive B cell subset, IgD⁺CD38⁻ (Bm1-2); two GC B cell subsets, IgD⁻CD38⁺CD77⁺ centroblasts (Bm3), and IgD⁻CD38⁺CD77⁻ centrocytes (Bm4); and an IgD⁻CD38⁻ memory B cell subset (Bm5). Fig. 3 shows that c-Met is expressed by CD38⁺CD77⁺ centroblasts (Bm3) and by a part of the CD38⁺CD77⁻ subset. This finding is supported by immunohistochemical studies on frozen sections of human tonsillar tissue: as is shown in Fig. 4, c-Met is predominantly expressed by lymphocytes within the dark zone of the GC, which contains rapidly dividing centroblasts and low numbers of FDC. These results mean that c-Met induction in vivo, occurs in GC-cells at a pre-selection stage, i.e., cells that have recently been recruited by antigen plus antigen-specific T lymphocytes in the T cell– rich extrafollicular microenvironment.

Ligation of the B cell antigen receptor (BCR) and CD40 plays a key role in the initiation of a T cell dependent B cell response and initiates the GC reaction (1, 49–51). In view of the expression of c-Met on centroblasts, i.e., on recent GC immigrants, we hypothesized that these receptors might also regulate c-Met expression. To address this hypothesis, the biological conditions for B cell activation were mimicked in vitro. Tonsillar B cells were cultured on CD40 ligand (CD40L) transfected L cells or, as a control, on wild-type L cells, in the presence or absence of BCR stimuli (anti-Ig antibodies or Staphylococcus aureus Cowans strain I [SAC]). As is shown in Fig. 5,A, concurrent ligation of CD40 and the BCR induced a strong transient induction of c-Met in human tonsillar B cells, peaking at 48 h. Single triggering of CD40 also strongly induced c-Met (Fig. 5,C) but single ligation of the BCR did not induce c-Met expression above control levels (untransfected L cells and medium alone) (Fig. 5, B and D). In approximately half of the experiments, concurrent CD40 and BCR stimulation resulted in a c-Met induction that was stronger than after CD40 ligation alone, suggesting synergy between the CD40 and BCR pathways. Stimulation by various cytokines including IL-2, IL-4, and IL-6 did not induce c-Met (data not shown).

These data clearly identify CD40-CD40L as a major pathway for induction of c-Met in B cells.

Cell motility and morphogenesis, major functions of the c-Met–HGF/SF pathway, are dependent on tightly controlled cell adhesion. This prompted us to study whether the c-Met–HGF/SF pathway might regulate B cell adhesion. Since c-Met–positive B cells represent a subset of tonsillar B cells that cannot be readily purified by negative selection procedures, we addressed this question by using Namalwa cells transfected with c-met cDNA (Nam^met). The expression of c-Met in this B cell lymphoma line and the wild-type control (Nam^wt) are shown in Fig. 1,B. We observed that HGF/SF induces a strongly augmented adhesion to both vascular cell adhesion molecule 1 (VCAM-1) and fibronectin of c-met transfected Namalwa B cells (Nam^met) (Fig. 6, A and B). This effect of HGF/SF on B cell adhesion was dose dependent and was not observed upon stimulation of wild-type Namalwa cells (Nam^wt). An increased adhesion in response to HGF/SF to both VCAM-1 and fibronectin was also observed with the Burkitt's lymphoma cell lines Raji and EB4B (data not shown).

To identify the adhesion receptors on Nam^met responsible for enhanced VCAM-1 and fibronectin binding, antibody blocking experiments were performed. Nam^met expresses α4β1 and α4β7, which both are receptors for VCAM-1 and for an alternatively spliced segment (CS-1) of fibronectin, but expresses no detectable level of the fibronectin receptor α5β1 (data not shown). We observed that adhesion to VCAM-1 and fibronectin was completely blocked by mAbs against both the α4 and β1 integrin chain (Fig. 6, C and D). Since c-Met stimulation by HGF/SF did not lead to increased α4β1 expression (and also did not upregulate or induce α4β7 or α5β1) (data not shown), these results indicate that the c-Met–HGF/SF pathway enhances B cell adhesiveness through activation of the α4β1 integrin.

All together, the above data strongly favor a functional role of the c-Met–HGF/SF pathway in B cell differentiation, namely in the regulation of B cell adhesiveness. However, obviously, an in vivo biological role in this process would require the availability of HGF/SF within the lymphoid tissue microenvironment. This prompted us to (a) assay the production of HGF/SF by primary cultures of various tonsillar cell populations, and (b) study the expression of HGF/SF mRNA within these cell populations by RTPCR. Determinations of HGF/SF production by ELISA demonstrate that high levels of HGF/SF are produced by primary cultures of tonsillar stromal cells, including cultures of FDC-enriched tonsillar cell subfractions (Fig. 7,A). By contrast, tonsillar T- or B-lymphocytes, cultured in the presence or absence of various mitogenic stimuli, did not produce detectable levels of HGF/SF. Consistent with these results, in RT-PCR studies HGF/SF mRNA was exclusively detectable in tonsillar stromal cells (Fig. 7 B).

The products of proto-oncogenes are important regulatory molecules that exert a wide range of effects on basic cellular functions such as the control of cell growth and differentiation. The c-met proto-oncogene product is a receptor tyrosine kinase (52, 53) that binds HGF/SF, a mesenchymally derived cytokine with pleiotropic biological effects on proliferation, cell motility and morphogenesis of epithelial, endothelial, and myogenic cells (31–36, 54). More recently, the c-Met–HGF/SF pathway has also been implicated in the proliferation and differentiation of early hematopoietic progenitor cells (55–58) and in monocyte-macrophage differentiation (59). Here, we demonstrate that the c-Met–HGF/SF pathway is also operative during T cell– dependent B cell differentiation, where it is involved in the control of lymphocyte integrin function on B cells, a process regulating adhesion and homing of B cells within the lymphoid microenvironment.

We observed that stimulation of tonsillar B cells with phorbol ester PMA leads to a rapid c-Met induction (Fig. 1,A). Induction of c-Met expression upon protein kinase C (PKC) stimulation by phorbol ester has previously also been reported in epithelial cell lines (60). In addition, we observed constitutive expression of c-Met on several B cell lymphoma lines (Fig. 1,B). The c-Met receptor on these B cell lines is signaling competent, as triggering by HGF/SF resulted in enhanced tyrosine phosphorylation of c-Met (Fig. 2). These findings present the first direct evidence for expression of a functional c-Met receptor on B lymphocytes. Indirect evidence for a role of c-Met in B cells has previously been provided by Delaney et al. (61), who demonstrated that HGF/SF enhances immunoglobulin production by murine B cells. However, as c-Met expression was not studied and whole splenocyte cultures were used, indirect effects of HGF/SF were not ruled out. For T cells, Shaw and colleagues reported that HGF/SF stimulated the adhesion and migration of the memory subset. However, these target cells appeared not to express c-Met (62).

One of the key findings of our study is that concurrent CD40 and BCR ligation induces a strong transient expression of c-Met on B cells in vitro (Fig. 5). Presumably, BCR and CD40 mediated signals are also instrumental in the physiological induction of c-Met. This is suggested by the fact that c-Met is expressed in vivo on a subset of tonsillar centroblasts (CD38⁺CD77⁺) (Figs. 3 and 4). Centroblasts are the offspring of B cells that have recently been activated at extrafollicular sites by antigen plus accessory signals provided by antigen-specific T cells (1). These signals critically involve CD40/CD40L interactions: patients with x-linked hyper-IgM syndrome (due to mutated and consequently defective CD40L) do not develop GC and blocking of the CD40/CD40L pathway in mice leads to complete inhibition of GC reactions (49–51, 63, 64). Our results strongly suggest that c-Met induction is directly linked to the initiation of the B cell immune response. Indeed, we observed that dual ligation of CD40 and the BCR also induces c-Met on naive (IgD⁺CD38⁻) B cells (our own unpublished observation).

As both the migratory and morphogenic responses to HGF/SF are critically dependent on cell adhesion, could regulation of the c-Met–HGF/SF pathway also have a regulatory role in B cell adhesion? This idea is supported by the finding that HGF/SF augments adhesion of c-Met transfected Namalwa B cells as well as the c-Met expressing B cell lines Raji and EB4B to VCAM-1 and fibronectin (Fig. 6 and data not shown). This HGF/SF induced adhesion was mediated through activation of the integrin α4β1. Previous studies from our own and from other laboratories have shown an important role for α4β1 in GC formation (27– 29). In the GC, α4β1 mediates B cell adhesion to VCAM-1 on FDC, an interaction that regulates the formation of the microenvironment required for the affinity selection of GC B cells. Apart from establishing physical contact between B cells and FDC, α4β1 presumably contributes directly to the B cell selection process itself, as signaling through the α4β1–VCAM-1 pathway costimulates rescue of GC B cells from apoptosis (6, 29, 47). Furthermore, α4β1 also regulates cell adhesion to fibronectin (65), an important substrate for cell migration.

The above data strongly favor a functional role of the c-Met–HGF/SF pathway in B cell differentiation, namely in the regulation of B cell adhesiveness. Obviously, however, an in vivo biological role in this process requires the availability of HGF/SF within the lymphoid microenvironment. Interestingly, we indeed observed production of high levels of HGF/SF as well as expression of HGF/SF mRNA by tonsillar stromal cells. In contrast, tonsillar T- or B-lymphocytes were HGF/SF negative (Fig. 7).

Adhesion regulation is believed to be fundamental to the control of cell migration and microenvironmental homing during lymphocyte differentiation. This migration and homing within tissues, like recruitment from the blood, presumably is determined by an organized display of adhesive ligands and regulatory factors, specific for a given microenvironment (23). Thusfar, most studies on the regulation of integrin-mediated adhesion have focussed on cytokines of the chemokine family (66). Chemokines, which bind to G protein linked 7-transmembrane serpentine receptors (67, 68), have been shown to mediate chemotaxis and rapid functional activation of leukocyte integrins on myeloid cells, macrophages, and lymphocytes (69–73). HGF/SF belongs to the family of plasminogen-related growth factors (74), that also includes macrophage stimulating protein. These molecules, that are structurally unrelated to the chemokines, have the basic domain organization and mechanism of activation of the blood proteinase plasminogen, i.e., they are characterized by the presence of a kringle domain(s), an activation domain, and a serine proteinase domain. Our present data strongly support a physiological role of HGF/SF in the regulation of B cell adhesiveness, microenvironmental homing and in the morphogenesis of the GC. Local production of HGF/SF has been demonstrated surrounding blood vessels in inflammation (75, 76), and in this paper we demonstrate that HGF/SF is produced by FDC-enriched cell populations. In view of the pleiotropic effects of HGF/SF on other cell types, HGF/SF may have additional, as yet unkown, roles in antigen-specific B cell differentiation. In particular, cross-talk between integrins and c-Met–signaling pathways triggered by FDC might contribute to B cell survival (6, 29, 47). Interestingly, like the chemokines, HGF/SF has a high affinity for heparin, which is present on cell surfaces and in the ECM in the form of heparan sulphate proteoglycans. This heparan sulphate binding limits diffusion, thus allowing the development of chemotactic gradients and the localization of proadhesive activity to the appropriate lymphoid microenvironment (77).

We thank Dr. G. Koopman for helpful advise, Dr. J. Banchereau for CD40L transfected L cells, Drs. G. Hartmann and E. Gherardi for c-Met c-DNA, Mr. J.B.G. Mulder for technical assistance, and Drs. C.M. van Noesel, F.M. van den Berg, and M. Snoek for critical reading of the manuscript.

This study was supported by a grant from the University of Amsterdam.