TNF-α-Induced Expression of Adhesion Molecules in the Liver Is Under the Control of TNFR1—Relevance for Concanavalin A-Induced Hepatitis¹ Free

Tumor necrosis factor-α is a pleiotropic cytokine that mediates host defense as well as tumor necrosis. TNF is able to induce profound changes of the vascular endothelium. These mechanisms are likely to contribute to tissue injury in pathology such as disseminated intravascular coagulation seen in septic shock, adult respiratory distress syndrome, or cerebral malaria. The central role for TNF in up-regulating cell surface adhesion molecules such as ICAM-1, VCAM-1, or E-selectin, thereby promoting adhesion of leukocytes to and subsequent transmigration through the endothelial barrier, is well documented (1, 2).

TNF exerts its actions via two distinct surface receptors, the 55-kDa receptor (also termed TNFR1 or TNFR p55) and the 75-kDa receptor (TNFR2 or TNFR p75) (3). The biological actions of TNFR1 have been extensively studied (4, 5), whereas the role for TNFR2, which preferentially binds the 26-kDa membrane-bound (m)³ precursor form of soluble (s) TNF (6), is not completely understood. A previous report emphasized the in vivo relevance of TNFR2 by showing that mTNF-induced adhesion molecule expression via TNFR2 was inhibited in mice lacking this receptor (TNFR2^−/−), but not in mice lacking TNFR1 (TNFR1^−/−), thereby inducing resistance against experimental induced cerebral malaria (7).

In addition there is increasing evidence that acute or chronic liver disease is closely related to elevated TNF and TNFR (sTNFRs) plasma levels as well as to a strong induction of in situ expression of both TNFRs (8, 9, 10, 11). The direct hepatotoxic effect of TNF has been studied in various experimental T cell- and macrophage-dependent models (12). We recently described a CD4⁺ T cell, TNF (13) and IFN-γ-dependent (14) model of liver injury in mice, which is inducible by a single injection of the mitogenic plant lectin Con A (15). The 26-kDa membrane-bound precursor form of 17-kDa sTNF and both TNFRs have been shown to be central for the development of Con A hepatitis (16). Additional mediators such as the Fas ligand/Fas system or perforin/granzyme have been described either to contribute (17, 18) or not to contribute (19, 18) to liver injury induced by Con A. However, mTNF-, Fas ligand-, and perforin-mediated direct hepatotoxic effects require cell-to-cell contact between hepatocytes and leukocytes. A prerequisite for the direct action of effector cells is their adhesion to and transmigration through the endothelial barrier. The protective effect of anti-ICAM-1 mAbs (18) and Arg-Gly-Asp mimetics (which block binding of β₁ integrins to several extracellular matrix glycoproteins) (20) from Con A hepatitis confirmed this assumption.

Therefore, we designed a study, which had the objective to characterize the TNFR-dependent expression of adhesion molecules in Con A hepatitis and to evaluate the functional role of adhesion molecule up-regulation for Con A-induced CD4⁺ T cell recruitment and liver damage.

Male BALB/c and ICAM^−/− mice were obtained from the institute’s internal animal breeding house. TNFR^−/− mice were kindly provided by Dr. H. Bluethmann (Hoffmann LaRoche AG, Basel, Switzerland). The appropriate wild-type (wt) animals were used in all experiments in which knockout mice were used. Animals received human care according to the National Institutes of Health as well as to the legal requirements in Germany and were maintained under controlled conditions (22°C, 55% humidity, 12-h dark/light cycle) and had free access to standard laboratory chow (Altromin 1313) and water. When taken into the experiment, mice were 6–8 wk old and had a half weight of 18–25 g.

All reagents were injected in a total volume of 200 μl per 20-g mouse. Con A (Sigma, Deisenhofen, Germany) was injected i.v. in pyrogen-free saline. The dose of Con A was 20 and 18 mg/kg in BALB/c mice and TNFR^−/− mice, respectively (because of the lower sensitivity of BALB/c mice toward Con A-induced liver injury; Ref. 21). rmuTNF (kindly provided by Dr. G. R. Adolf, Bender & Co, Vienna, Austria) was injected i.v. in pyrogen-free saline in a concentration of 10 μg/kg.

The following mAbs were dissolved in pyrogen-free saline and injected i.v. 15 min before Con A application: 1) 200 μg of rat anti-mouse E-selectin mAb (UZ4) (22), 200 μg of rat-anti-mouse P-selectin mAb (RB40.34) (23), or a combination of both; 2) 150 μg of rat anti-mouse ICAM-1 mAb (YN1/1.7.4) and 150 μg of rat anti-mouse CD11a mAb (M17/4, PharMingen, Hamburg, Germany); 3) 150 μg rat anti-mouse VCAM-1 mAb (429, PharMingen). As isotype control, we injected an anti-endothelial Ab, which does not interact with adhesion molecules (BR2). In vivo binding was controlled with immunofluorescent staining of i.v.-injected mAbs (slides were solely incubated with the Texas Red-labeled goat anti-rat secondary Ab).

For determination of circulating cytokines (TNF and IFN-γ) and sTNFRs by ELISA (see above), blood was taken from the tail vein at the indicated time points. After lethal anesthesia, blood was taken by cardiac puncture for ALT measurement as previously described (14, 24). Liver was then removed and frozen for determination of DNA fragmentation as described elsewhere (14) or for immunofluorescent analysis (see below).

Detection of TNF, IFN-γ, and IL-2 was performed as previously described (25). sTNFRs were measured using a commercial ELISA kit (Genzyme, Cambridge, MA), which was exactly performed according to the manufacturers instructions.

As primary mAbs: YN1/1.4.7 reacts with ICAM-1, MK2/1 with VCAM-1; mAb 10E9.6 reacts with murine E-selectin; RM4-5 recognizes CD4 (both purchased from PharMingen); primary rat mAbs were detected using goat anti-rat IgG tagged with Texas Red (Dianova, Hamburg, Germany). TNFR1 and TNFR2 were detected with a polyclonal rabbit anti-mouse Ab (Hy Cult Biotechnology, Uden, The Netherlands); primary rabbit Abs were detected with Cy 3-labeled goat anti-rabbit Ab (Dianova). Clone 104 detected CD45.2 and was directly labeled with FITC (PharMingen). Mouse endothelial cell Ag 32 (MECA-32) recognizes an endothelial Ag and was used as positive, rat IgG (PharMingen) as negative control, to exclude unspecific binding of the primary Ab.

Cryostat sections were performed at 10 μm, fixed in acetone/methanol (1:1) at 4°C for 10 min. After washing in PBS, slides were blocked with PBS containing 3% BSA and then incubated overnight with the primary Ab, at 4°C in a moist, light-protected chamber. After rinsing with PBS, binding sites were detected with the secondary Ab dissolved in PBS/BSA 3% for 1 h at room temperature. After washing in PBS, slides were mounted with TBS/glycerol (1:1), pH 8.6. Sections were then examined by confocal laser scanning microscopy (Bio-Rad 1000).

Total liver RNA was isolated by Clontech Kit (ClonTech, Palo Alto, CA) according to the manufacturer’s instructions. Two micrograms of total liver RNA was then reverse transcribed into cDNA using the Superscript II Polymerase (Life Technologies GmbH, Eggenstein, Germany). The sequences of the oligonucleotide primers are: 5′ TNFR1, CTG CTG TCA CTG GTG CTC DTG; 3′ TNFR1, CAC ACA CCG TGT CCT TGT CAG; 5′ TNFR2, GTC GCG CTG GTC TTC GAA C; 3′ TNFR2, CAC TTG CTC AGC CTC ATG. PCR was conducted in an automatic DNA thermal cycler Primus 96 Thermocycler (MWG-Biotech, Ebersberg, Germany). The cycle program was set to anneal at 58°C for a total of 32 cycles. The expected fragment lengths were 349 bp for TNFR1 and 413 bp for TNFR2. Amplicons were visualized by agarose gel electrophoresis. The gels were scanned by GelDoc 2000 (Bio-Rad, Richmond, CA). The relative quantities of TNFR1 and TNFR2 mRNA are presented as ratio between the intensity of TNFR1 and TNFR2 band relative to the intensity of the housekeeping gene β-actin.

Mice were killed by exsanguination from the subclavian artery and vein and livers were then removed. Hepatic mononuclear cells (MNCs) were prepared as previously described (26). Briefly, the liver was pressed through a stainless steel mesh and suspended in RPMI 1640 medium with 10% FCS and 25 mM HEPES. The cell suspension was centrifuged at 1200 rpm for 5 min and resuspended in 20 ml medium and then centrifuged at 50 × g for 30 s. Supernatants were resuspended in 10 ml medium and cell suspension was overlaid on 5 ml histopaque solution (density: 1083, obtained from Sigma). After centrifugation at 600 × g for 25 min at 25°C, the visible interface was aspirated and resuspended in medium to perform cell counts. Cells were washed once again and resuspended in PBS/FCS 1% to perform flow cytometry.

FITC-labeled anti-CD45 (clone 104) and Texas Red-tagged anti-CD4 (clone RM4-5) were purchased from PharMingen. Cell suspensions of 10⁵ cells were stained with mAbs and analyzed by flow cytometer (Coulter Pharmaceutical, Palo Alto, CA). Dead cells were excluded by forward sideward scatter. The number of CD4⁺ T cells per liver was calculated by multiplication of the percentage of CD4⁺ T cells with the percentage of CD45⁺ leukocytes of total isolated MNCs.

Data of ALT, cytokines, and sTNFRs are expressed as mean ± SEM and analyzed by one-way ANOVA; in case of differences among groups (at least p < 0.05), data were subjected to Dunnett’s multiple comparison test of the control against all other groups with the computer program INSTAT2 (GraphPad Software, San Diego, CA).

Mice deficient of either one of the two TNFRs and the corresponding wt animals were i.v. injected with a dose of 18 mg/kg Con A. TNFR1^−/− and TNFR2^−/− mice failed to develop liver injury upon Con A injection, as assessed by determination of plasma ALT-levels (Table I). To exclude the possibility that this, also previously observed protective effect (16) was due to a reduced production of the central mediators TNF and IFN-γ, we detected TNF and IFN-γ plasma levels of TNFR^−/− mice. Circulating peak concentrations of TNF (14) were significantly elevated in both, TNFR1^−/− and TNFR2^−/− mice (Table I). Maximum plasma levels of IFN-γ (14) were significantly elevated in TNFR1^−/− mice but were comparably high to the wt in TNFR2^−/− mice (Table I).

To investigate whether TNFRs are up-regulated within the liver after systemic Con A injection, we performed RT-PCR and subsequent semiquantitative analysis of total liver RNA. Both TNFR mRNAs were constitutively expressed in the livers of healthy mice. Injection of Con A did not significantly change overall hepatic TNFR1 (Fig. 1,A) and TNFR2 (Fig. 1,B) mRNA production during the observed time-span of 8 h. As internal control, we analyzed TNFR1 and TNFR2 mRNA production in the spleen (Fig. 1, A and B), where both receptor transcripts were significantly induced upon Con A injection. TNFR1 and TNFR2 mRNA were undetectable in TNFR1^−/− and TNFR2^−/− mice, respectively (Fig. 1,C). Immunofluorescent staining of both TNFRs showed constitutive expression of both TNFRs in livers of healthy mice. However, TNFR1 and TNFR2 protein decreased as early as 1 h after injection of Con A, reaching minimum expression levels at 2 to 4 h after treatment (Fig. 2). Thereafter, TNFR expression increased, reaching almost basal levels in case of TNFR1 8 h after Con A challenge (Fig. 2). The expression level of TNFR2 8 h after Con A injection was more prominent compared with the constitutive expression of saline-treated animals (Fig. 2). No staining of either TNFR1 in TNFR1^−/− or TNFR2 in TNFR2^−/− mice was detectable, confirming the specific binding of the Abs (data not shown).

Determination of sTNFRs in the circulation of healthy mice by ELISA revealed that both receptors are continuously shed into the circulation (sTNFR1: 183 ± 78 pg/ml, sTNFR2: 2024 ± 63 pg/ml, cf Fig. 3). Con A injection induced a rapid increase of both sTNFRs in plasma, which parallels the decline that we observed by immunofluorescence (Fig. 3). We detected maximum levels of sTNFR1 (1354 ± 12 pg/ml; Fig. 3, solid line) and of sTNFR2 (5776 ± 769 pg/ml; Fig. 3, dotted line) 8 h after Con A injection. Thereafter, the levels of both soluble receptors declined slowly. Twenty-four hours after Con A injection the serum levels of both sTNFRs were still elevated. In comparison, TNF itself showed a sharp transient plasma peak with a maximum at 2 h after Con A injection (data not shown), as described previously (14).

Before the investigation of the effect of TNFR deficiency on adhesion molecule expression in livers of Con A-treated mice, we analyzed their expression in wt mice. Immunofluorescent staining and subsequent confocal laser imaging revealed weak constitutive expression of ICAM-1 (Fig. 4,a) and VCAM-1 (Fig. 4,d) in livers of healthy mice, whereas no constitutive E-selectin (Fig. 4,g) and P-selectin (data not shown) expression was detectable. Time course analysis of the induction of adhesion molecule expression revealed a strong increase of ICAM-1 staining on sinusoids, central and portal veins as early as 6 h after challenge (data not shown), reaching high expression levels after 8–24 h (Fig. 4, b and c). VCAM-1 expression also started to increase after 6 h (data not shown), reaching highest expression 8 h after Con A injection (Fig. 4, e and f). Expression of E-selectin also increased as early as 6 h (data not shown) after Con A injection and was prominently expressed 8 h after challenge with staining of central and portal vein endothelial cells as well as of sinusoids (Fig. 4,h). Twenty-four hours after Con A, only minimal staining of E-selectin was detectable on postsinusoidal venules (Fig. 4 i). We failed to detect any staining of P-selectin in livers of Con A-treated mice (data not shown).

Immunofluorescent staining of the CD4 Ag revealed accumulation of CD4⁺ T cells as early as 6 h after Con A (data not shown). Eight hours (Fig. 4,k) after Con A injection, we found most of the CD4⁺ T cells in the hepatic parenchyma. Twenty-four hours after Con A, most of the CD4⁺ T cells were found as large cell clusters in the periportal region and almost no CD4⁺ T cells were detectable in the hepatic sinusoids (Fig. 4 L). These data are corroborated by the finding that 8 h after Con A injection, CD4⁺ T cell accumulation is induced 6.1 ± 1.2 (6.7 ± 1.3 × 10⁵ CD4⁺/CD45⁺ cells) fold vs 1.0 ± 0.2 (1.1 ± 0.2 × 10⁵ CD4⁺/CD45⁺cells) in saline-treated mice, as quantified by flow cytometric analysis of hepatic MNCs.

TNFR-deficient mice are resistant against Con A hepatitis. Therefore, we investigated whether deficiency in either one of both receptors alters the expression of adhesion molecules induced by Con A, thereby affecting the infiltration of the critical T cell population, i.e., CD4⁺ T cells, into hepatic tissue. To this end, we injected TNFR1^−/−, TNFR2^−/−, and corresponding wt mice (all three strains showed low constitutive ICAM-1 and VCAM-1 staining comparable to untreated BALB/c mice, cf Fig. 4, A and D) with Con A and analyzed the expression of ICAM-1, VCAM-1, and E-selectin 8 h after challenge. We failed to detect any difference in the staining intensity of ICAM-1 (Fig. 5, first column), VCAM-1 (Fig. 5, third column), or E-selectin (data not shown) between the TNFR-deficient and the corresponding wt mice. Accordingly, there was no difference in the accumulation of CD4⁺ T cells (data not shown).

To further elucidate the role of TNF in up-regulating adhesion molecules in the liver, we injected TNFR1^−/−, TNFR2^−/− as well as wt mice with 10 μg/kg rmuTNF. Eight hours later, we removed the livers for immunofluorescent staining analysis of adhesion molecule expression. ICAM-1 (Fig. 5, second column) and VCAM-1 (Fig. 5, fourth column) expression was markedly induced in the livers of TNFR2^−/− and wt mice, whereas no up-regulation was detectable in the livers of TNFR1^−/− mice. We were unable to detect E-selectin in the livers of TNF-treated TNFR1^−/−, TNFR2^−/−, and wt mice, respectively, most likely reflecting the fact that the concentration of rmuTNF used was not sufficient to induce E-selectin expression in the liver.

These results clearly demonstrate that TNF-mediated ICAM-1 as well as VCAM-1 up-regulation in the liver is under the control of the TNFR1.

To elucidate the functional role of the different adhesion molecules and their counterparts on leukocytes, we performed in vivo blocking studies, using mAbs against either ICAM-1/LFA-1 or VCAM-1 or E-selectin or P-selectin. But surprisingly, despite the strong expression of the adhesion receptors (see above), none of these Abs given either alone, i.e., anti-VCAM-1, anti-ICAM-1/anti-LFA-1, anti-E-selectin, anti-P-selectin (data not shown), or in combination (cf Fig. 6), or gene-targeted deletion of ICAM-1 (data not shown), prevented the development of acute hepatic failure induced by Con A. Plasma ALT levels were unaltered in normal mice passively immunized against E-selectin and P-selectin (Fig. 6,A) or in ICAM^−/− mice treated with anti-VCAM-1 mAb before Con A injection (Fig. 6,B) compared with Con A controls pretreated with a control IgG Ab. Moreover, the production of the central mediators TNF and INF-γ as well as of IL-2 was not affected (data not shown). However, flow cytometric analysis of isolated hepatic MNCs revealed that either anti-E-selectin mAb injection (Fig. 7,A) or the use of ICAM-1^−/− mice treated with anti-VCAM-1 mAb (Fig. 7 B) before Con A treatment significantly reduced total accumulation of CD4⁺ T cells ∼45 and 55%, respectively, compared with control mAb pretreated mice. In contrast, pretreatment of mice with either anti-ICAM-1 or anti-VCAM-1 mAb alone insignificantly inhibited total Con A-induced accumulation of CD4⁺ T cells by 20% (data not shown).

Inflammatory diseases, such as viral hepatitis (8, 9, 11) or hepatic inflammation due to alcohol abuse (10), are associated with enhanced sTNFR serum levels, as well as with a strong up-regulation of both TNFRs in situ. But, up to now, only a few experimental data exist concerning the in vivo regulation and function of TNFR1 and TNFR2 under conditions of acute inflammation.

It has been shown previously that TNFR1 and TNFR2 mRNA was rapidly induced in livers of mice suffering from TNF-dependent, carbon tetrachloride-induced liver damage (27). In contrast, in the Con A model, where liver injury depends on both TNFRs (16), we did not observe a transcriptional regulation of hepatic TNFR1 and TNFR2 mRNA, although both transcripts were significantly induced in the spleen within the same time-span after Con A injection. However, we observed a profound reduction of TNFR1 and TNFR2 expression in liver sections after Con A injection. This down-regulation is paralleled by a marked increase in both sTNFR serum levels. Elevated serum levels of soluble TNFRs, especially TNFR2 have often been described to correlate with TNF serum levels and disease activity in acute and chronic liver disease (i.e., hepatitis B or C or autoimmunohepatitis) in humans (8, 9, 11, 28, 29), which might point to a desensitizing mechanism, that protects hepatocytes from the high local levels of TNF. The contradictory observations between carbon tetrachloride- and Con A-induced liver injury are not yet clear, but it seems likely that the T cell mitogen Con A induces a different local cytokine milieu, thereby inducing a distinct transcriptional regulation of TNFR mRNA expression in hepatic tissue.

Leukocyte recruitment from the blood stream is a central feature of inflammatory responses, which is regulated by adhesion molecules expressed on endothelial cells and their counter receptors on leukocytes (2). Because expression of these adhesion molecules is induced during inflammation, we analyzed their expression pattern after Con A application. We found that, in contrast to selectins, ICAM-1 and VCAM-1 were constitutively expressed, which is in line with earlier reports. Upon Con A injection, a massive induction of hepatic ICAM-1, VCAM-1, and E-selectin expression was seen. This induction of adhesion molecules corroborates earlier reports showing an up-regulation of ICAM-1, VCAM-1, and E-selectin in the liver under inflammatory conditions in humans (30) and mice (31, 32). However, others described a lack of inducibility of selectins in human livers and excluded a role for selectin-mediated leukocyte recruitment via hepatic sinusoids (33, 34). Hence, the data concerning selectin expression and their functional role for leukocyte recruitment in liver microvasculature still remain to be elucidated. But one can conceive that selectin-mediated rolling mechanisms in hepatic sinusoids are not necessary for leukocyte endothelial interaction, because of the slow blood flow that allows continuous contact between leukocytes and endothelial cells.

TNF is well known as a potent inducer of several adhesion molecules. Hence, we wondered whether mice deficient in either one of both TNFRs would exhibit a different pattern of adhesion molecule expression after Con A treatment. However, we could not find any difference between Con A-treated TNFR1^−/−, TNFR2^−/− mice, and the corresponding wt animals. Moreover CD4⁺ T cell infiltration was unaltered in TNFR^−/− mice compared with the wt. These findings implicate that the resistance of TNFR^−/− mice toward Con A was not due to an impaired induction of adhesion molecules by TNF. In contrast, a resistance of TNFR2^−/− mice against experimentally induced acute cerebral malaria, which resulted from a lack of mTNF-inducible ICAM-1 expression, was recently described by Lucas et al. (7). However, treatment with rmuTNF instead of Con A induced ICAM-1 and VCAM-1 only in TNFR2^−/− and wt mice, but not in TNFR1^−/− mice. Hence, our data provide evidence that ICAM-1 expression in the liver is exclusively under the control of TNFR1. These findings corroborate previous findings showing that in other cell systems and tissues, TNF-inducible adhesion molecule expression including ICAM-1 and VCAM-1 is mediated by TNFR1 (35, 36). The observation that Con A-treated TNFR^−/− mice express high levels of adhesion molecules, although TNF-induced expression of adhesion molecules in the liver is under control of TNFR1 suggest the involvement of other endothelium activating cytokines such as IL-1, IL-12, or IFN-γ (37, 38).

However, it remained unclear whether these adhesion molecules take part in the hepatic accumulation of activated CD4⁺ T cells, i.e., the critical cell population in this experimental model (15). In this study we provide evidence that although CD4⁺ T cell accumulation was significantly inhibited by blockade of E-selectin or by lack of ICAM-1 together with VCAM-1 blockade, these adhesion molecules play no functional role for the onset of liver disease induced by Con A. This is in contrast to previous reports describing inhibitory effects of either anti-E-selectin and anti-VCAM-1 mAb (39) or anti-ICAM-1 mAb but not anti-VCAM-1 mAb (18). These controversial findings might have been due to different pretreatment regiments, e.g., Ab pretreatment 24 and 48 h before Con A injection, which might have depleted intrahepatic lymphocytes critical for the development of liver injury. In contrast, in our study, we have pretreated mice with the respective Abs shortly before Con A administration and the pretreatment regiment was proven to be functional with respect to a significant inhibition of CD4⁺ T cell accumulation. Interestingly, accumulation of CD4⁺ T cells in hepatic sinusoids remained unaffected by blockade of adhesion molecules (data not shown). Hence, E-selectin, ICAM-1, and VCAM-1 contributed to CD4⁺ T cell recruitment via pre- and postsinusoidal veins, whereas accumulation of CD4⁺ T cells in sinusoids seems to be regulated differently. The findings that adhesion molecules were strongly expressed and that the accumulation of CD4⁺ T cells was partially inhibited by adhesion molecule blockade, whereas the development of liver disease remained unaffected, was surprising, because CD4⁺ T cells have been shown to be critical for the development of Con A hepatitis by Ab-dependent depletion of CD4⁺ T cells (15). It is conceivable that this procedure also depleted the CD4⁺ proportion of liver resident Vα14 NK T cells, which have been shown to be central for the onset of Con A-induced liver disease (40, 41, 42). Hence, it seems likely that Con A-induced stimulation of local immune cells, i.e., NK T cells in concert with thymus derived T cells, which have been shown to be abundant in the parenchymal space (43), and liver resident macrophages are sufficient to induce hepatocellular damage. Recruitment of CD4⁺ T cells from the circulation may then contribute to the elimination of harmful activated intrahepatic lymphocytes (i.e., T cells, NK T cells, or NK cells) (44). Hence, blocking ICAM-1, VCAM-1, or E-selectin as therapeutic approach to inhibit human liver disease has to be carefully considered.

We thank Dr. H. Bluethmann (F. Hoffmann-LaRoche AG, Basle, Switzerland) for kindly providing us TNFR knockout mice. We are indebted to Dr. G. R. Adolf (Bender & Co Vienna, Austria) for providing recombinant murine TNF. We are also indebted to Dr. D. Vestweber for providing anti-P-selectin mAb (23). We thank Dr. W. Neuhuber (Institute of Anatomy, University of Erlangen-Nürnberg, Erlangen, Germany) for experimental support regarding confocal laser scanning microscopy. The perfect technical assistance of Andrea Agli is gratefully acknowledged.