Number of pericryptal fibroblasts correlates with density of distinct mast cell phenotypes in the crypt lamina propria of human duodenum: Implications for the homeostasis of villous architecture



Pericryptal fibroblasts (PFs), a class of myofibroblasts, have strongly been implicated in the regulation of villous structure because of their location close to crypts and their ability to secrete cytokines affecting intestinal epithelial cell proliferation and differentiation. Recently, mast cells (MCs) have also been involved in the homeostasis of villous architecture. As myofibroblasts arise in a wide variety of settings concurrently with a local increase in the number of tissue MCs, we calculated in this study the density of both PF and distinct pericryptal MC phenotypes in the mucosa of human duodenum showing normal, defective, or atrophic villous profiles. In addition, we evaluated the statistical association between PF-MC densities and each pattern of villous architecture. Finally, we correlated the density of PF with the density of pericryptal MC phenotypes. For this purpose, samples taken by endoscopy from 30 patients complaining of inflammatory bowel disorders were studied by immunohistochemistry. The densities of α-smooth muscle actin-positive PFs as well as tryptase-, chymase-, and c-kit-positive MCs were determined in the crypt lamina propria. Villous architecture was found to be significantly associated with the number of PFs and tryptase-, chymase-, c-kit-positive MCs in the lamina propria (ANOVA group effect P < 0.001). High density of both PFs and MCs was found in intestinal samples with normal villous morphology while lower densities were associated with defective or atrophic villous profiles (Tukey's test for multiple comparison P < 0.001). In addition, a significant correlation was found between PF density and the density of each pericryptal MC phenotype (vs. tryptase-positive MCs, r = 0.913; vs. chymase-positive MC, r = 0.905; vs. c-kit-positive MC, r = 0.927; P < 0.001 in all cases). This study provides morphological support for an important cooperation between PFs and MCs in maintaining normal villous architecture. Anat Rec Part A 288A:593–600, 2006. © 2006 Wiley-Liss, Inc.

The mucosa of the small intestine is made up of a single-layered columnar epithelium and an underlying mesenchymal lamina propria. The epithelial lining invaginates to form numerous crypts (the intestinal crypts of Lieberkühn) and protrudes to generate larger, finger-shaped projections called villi. This complex architecture is maintained by epithelial-stromal interactions, which are largely unknown (Del Buono et al.,1992). Epithelial cells arise from a fixed proliferative stem cell compartment located at or near the base of the intestinal crypts (Bjerkens and Cheng,1999; Brittan and Wright,2002). Each active stem cell appears to give rise to two population of long-lived daughters, one committed to producing absorptive enterocytes, the other committed to producing secretory cell lineages (goblet, enteroendocrine, and Paneth cells) (Bjerkens and Cheng,1999). With the exception of Paneth cells, most members of the other three lineages complete their terminal differentiation during a highly organized upward migration from the crypt to adjacent villi (Millis and Gordon,2001).

The gut epithelium has one of the most rapid proliferative rates in the body (Brittan and Wright,2002). In the mouse, the journey from the crypt to the villus tip is completed in 2–5 days and terminates with cells being removed by apoptosis and/or exfoliation. It has been recognized that descendant cell migration and differentiation follow a topologically well-organized pattern of directional movement along the villous axis (Millis and Gordon,2001). This process appears to be somehow regulated by local environmental factors, among which the pericryptal microvasculature and neuron plexus, the resident and inflammatory cells in the lamina propria, and the molecular composition of the extracellular matrix are believed to play important but still largely unspecified roles (Simon-Assmann et al.,1995; Bjerkens and Cheng,2001; Paris et al.,2001).

Understanding the organization of epithelial cell movement may provide critical clues about how villous architecture is established and maintained in health and disease. Intestinal crypts of Lieberkühn are surrounded by a layer of specialized mesenchymal cells known as pericryptal fibroblasts (PFs) or intestinal subepithelial myofibroblasts (Powell et al.,1999b; Serini and Gabbiani,1999). These smooth muscle cell-like fibroblasts can be identified by certain characteristic features of the cytoskeleton, particularly by the expression of α-smooth muscle actin (Sappino et al.,1990). PFs have the capacity to establish organized contacts, including gap junctions, thus forming a pericryptal fenestrated cell envelope (Joyce et al.,1987). These cells produce factors that regulate the crypt and villous epithelium (Brittan and Wright,2002). In addition, tritiated thymidine labeling studies have indicated that, like their overlying epithelial cells, PFs continuously proliferate, differentiate, and migrate upward from the base of the crypt to the tip of the villus (Marsh and Trier,1974). This puts them in strategic position to establish and maintain instructive communications with stem cells and their descendants. According to this model, the structural integrity and the functional activity of the PF sheath would be crucial in orchestrating a correct epithelial cell replication and migration rate (Goyal et al.,1998).

Mast cells (MCs) are bone marrow-derived secretory cells that largely populate the lamina propria of the intestinal mucosa, whereby they exert a number of functions related to both innate and acquired immunity (Bischoff et al.,1996,1999). MCs also exhibit a spectrum of nonimmunological responses leading to wound repair, extracellular matrix remodeling, angiogenesis, and restitution of tissue structure and function (Maurer et al.,2003). MCs produce and release important regulators of myofibroblast development and function (Gailit et al.,2001). Myofibroblasts in turn synthesize and secrete cytokines, which exert marked chemoattractant and differentiating functions upon tissue MC and MC precursors. Thus, a complex circuit of regulatory signals involving PFs and MCs may operate in the pericryptal lamina propria to maintain intestinal epithelial cells homeostasis and consequently a proper villous shape.

In a previous study, we demonstrated that low densities of both total and tryptase-reactive MCs in the human duodenal mucosa from different inflammatory bowel disorders were associated with defective villous architecture (Crivellato et al.,2003). These findings let us speculate on a possible link between MC activity and the pattern of villous profile, suggesting participation of MCs to the complex network of cellular and molecular signals affecting mucosal morphology (Crivellato et al.,2005). In the present study, we extended our previous investigations by searching for a link between villous architecture, PF density, and the occurrence of distinct MC phenotypes in the pericryptal duodenal mucosa, in three groups of patients with inflammatory intestinal disorders.


Patients and Specimen Selection

We first examined a certain number of duodenal bioptic specimens selected from the archives of the Institute of Pathology, University of Udine Medical School. Thirty specimens were eventually chosen. To be included in the study, patients had to meet in accord to the following general criteria: no history of previous malignancies; intestinal biopsies negative for tumor conditions; no treatment before bioptic removal. Thus, all specimens referred to nontumor pathologies and were characterized by difference in the morphology of surface villi. The lamina propria presented mild to moderate inflammatory infiltrate, made essentially up of lymphocytes and plasma cells. Samples with neutrophil- and eosinophil-rich infiltrates were excluded. No signs of necrosis, fibrosis, or vascular damage were observable. Samples were subdivided into three groups according to the structural pattern of villi (Fenoglio-Preiser et al.,1999). Group 1 (10 cases) corresponded to duodenal slides presenting normal villous profile. Villous height fell between 320 and 570 μm and corresponded to 60–75% of the total mucosa height. Villous breadth, taken at the middle third of the villus, ranged from 85 to 140 μm. Enterocytes did not exhibit changes in cell shape or brush border alterations. Group 2 (10 cases) corresponded to samples showing defective villous architecture. Villi appeared broadened and shortened, sometimes also branched or fused. The surface epithelium was cuboid with variable alterations in the brush border organization. Group 3 (10 cases) was made up of samples showing complete absence of villi (villous atrophy). These specimens referred to bioptic samples taken during endoscopic examination from 30 patients (7 males, 23 females; mean age, 40.2 ± 13.7 years; range, 17–75), who had attended the Gastroenterologic Unit of the Udine General Hospital from 1996 to 2002. Clinical charts were collected. Reported clinical symptoms included dyspepsia, recurrent abdominal pain, weight loss, diarrhea, or a combination of them. In all patients, clinical data, routine laboratory findings, serological investigations for autoantibodies and antigliadin antibodies, as well as microscopic and cultured stool examination had been recorded. All patients had undergone endoscopic study of the proximal and distal duodenum, and multiple biopsies from the distal duodenum had been removed for conventional diagnostic histopathology and electron microscopy. Biopsies had been taken when the bowel symptoms were in an active state. On the basis of clinical, endoscopic, serological, histopathological, and ultrastructural findings, the following diagnoses had been established: celiac disease (14 cases), erosive duodenitis (2 cases), enteritis of possible viral origin (4 cases), and functional diarrhea (10 cases; Table 1).

Table 1. General and clinical data of the three groups of patients
 Group 1 (normal villi)Group 2 (defective villi)Group 3 (atrophic villi)
Sex1 M 9 F3 M 7 F3 M 7 F
Age (years)41.1 ± 8.838.4 ± 12.341.2 ± 19.3
Celiac disease 68
Erosive duodenitis  2
Viral enteritis 4 
Functional diarrhoea10  
Mean duration of the disease2.2 ± 0.8 (months)8.3 ± 3.4 (months)12.9 ± 4.2 (months)

Tissue Preparation and Immunohistochemistry

The material used for immunolabeling procedures was the same as for diagnostic histopathology. Duodenal biopsies fixed in 10% aqueous formalin and embedded in paraffin wax were cut perpendicularly to crypts. Immunohistochemical staining of PFs was performed using a mouse monoclonal antibody (MAb) against human α-smooth muscle actin (Dako, Glostrup, Denmark; 1A4). Mature MCs were stained using mouse MAbs against human MC proteases tryptase (Dako; AA1) and chymase (NeoMarkers, Fremont, CA; CC1). A rabbit polyclonal antibody antihuman c-kit (CD117; Dako) was used to identify both tissue MC precursors and mature MCs. A Mab against transforming growth factor-β (TGF-β; Novocastra Laboratories, Newcastle upon Tyne, U.K.; TGFB17) and a rabbit polyclonal antibody antitumor necrosis factor-α (TNF-α; Novus Biologicals, Littleton, CO) were also used to identify reactive MC in a limited number of specimens. Sections of 4 μm thickness were cut and applied to SuperFrost Plus slides (Menzel-Gläser, Braunschweig, Germany). Slides were dried overnight at 37°C, deparaffinized, and hydrated prior to antigen retrieval at high temperature. The immunoreaction was performed using the peroxidase technique. Briefly, after blockade of endogenous peroxidase activity with 3% H2O2 for 10 min at room temperature, sections were incubated with the primary antibodies in a Dako Autostainer Universal Staining System. Incubations with antibodies against α-smooth muscle actin, tryptase, chymase, and c-kit were protracted for 1 hr at room temperature. Incubations with antibodies against TGF-β and TNF-α were performed overnight at 4°C. Slides were then processed with ChemMate Dako EnVision Detection Kit for 30 min, counterstained with Mayer's hematoxylin for 3 min, dehydrated in a series of ethanols, cleared in xylene, and mounted in Eukitt. Negative controls were done by omitting the primary antibodies or by substituting them with irrelevant antibodies of the same class.

PF and MC Counts

The counts of α-smooth muscle actin-positive PFs, tryptase-positive MCs, chymase-positive MCs, and c-kit-positive MCs, and MC precursors were performed in sections cut perpendicularly to crypts and passing through the middle lamina propria. All counts were performed on a Zeiss Axioskop light microscope, using a micrometer grid fitted in a 10× eyepiece at a 100× objective magnification. Twenty five to 35 contiguous, nonoverlapping rectangular areas (each area measured 0.0117 mm2), covering the lamina propria of three sections per sample, were examined. PF and mucosal MC counts in each group were expressed as number of PFs/mm2 and MCs/mm2 lamina propria tissue, respectively.

Statistical Analysis

All results were expressed as mean (SD). To check for normality of data distribution, the Shapiro-Wilk test was performed. Levene's test was used to test homogeneity of variance among groups. Analysis of variance (ANOVA) testing was used for comparisons among groups. Tukey's honest significance difference test was used for confirmative testing. To assess the degree of correlation between PF density and the density of each MC phenotype, we used Pearson's r correlation coefficient. A linear regression model independently from villous structure was also estimated to predict the values of PF density as function of the density of tryptase-positive MCs and the values of c-kit-positive MCs and MC precursor density as function of PF density. The homogeneity of regression slopes in the three histological groups was tested by analysis of covariance. The software package SPSS (Statistical Package Social Sciences, SPSS, Chicago, IL) release 11.5.1 was used for data analysis. A P value < 0.05 defined statistical significance.



Immunohistochemical staining allowed visualization of a series of immunostained cells in the pericryptal duodenal lamina propria. These were α-smooth muscle actin-reactive PFs, mature MCs with tryptase or chymase content, c-kit-positive MCs and MC precursors, TGF-β- as well as TNF-α-reactive MCs. Examination of microscopical slides revealed interesting details. In group 1, PFs localized close to crypts (Fig. 1A), while in group 2 and, even more, group 3, they were frequently mislocated and did not formed a well-organized pericryptal cell envelope (Fig. 1B and C). Tryptase-positive and chymase-positive MCs appeared as large round to oval cells in the crypt lamina prorpia (Figs. 1D–F and 2A–C). In some instances, they presented an elongated, fibroblast-like arrangement (Fig. 1E and F). A few intraepithelial MCs were also visible. C-kit-reactive MCs and MC-committed precursors showed a greater degree of cellular pleomorphism (Fig. 2D–E). We looked for TGF-β- or TNF-α-reactive cells in five samples belonging to group 1. Some large granular cells, with the general appearance of MCs, were identified in the pericryptal lamina propria as TGF-β- or TNF-α-reactive cells (Fig. 3). When immunostaining for TNF-α and tryptase were performed on serial sections, only a subset of tryptase-reactive MCs showed reactivity also to TNF-α (Fig. 3C and D).

Figure 1.

Histochemical pictures of α-smooth muscle actin-reactive pericryptal fibroblasts (AC) and tryptase-positive mast cells (DF) in the crypt lamina propria of duodenal samples with normal (A and D), defective (B and E), and atrophic villi (C and F). In A, pericryptal fibroblasts present a regular location around crypts, which is progressively lost in B and C. Tryptase-reactive mast cells reside in the crypt lamina propria and appear as round to oval cells. Scale bar = 100 μm.

Figure 2.

Histochemical pictures of chymase-reactive mast cells (AC) and c-kit-positive cells (DF) in the lamina propria of duodenal samples with normal (A and D), defective (B and E), and atrophic villi (C and F). Mast cells with chymase content show round to oval profiles, while c-kit-positive cells exhibit higher polymorphism. Scale bar = 100 μm.

Figure 3.

Histochemical pictures of TGF-β- (A and B), tryptase- (C), and TNF-α- (D) reactive mast cells in the lamina propria of duodenal samples with normal villi. Some large, granular, TGF-β-positive cells, with the general appearance of MCs, are indicated by arrows (A and B).> Immunostaining of serial sections shows that tryptase-positive mast cells are (arrows in C and D) or are not (arrowheads in C and D) also reactive for TNF-α. Scale bar = 100 μm.

Statistical Analysis

Table 2 shows the results of PF and MC counts in the three groups of samples. The mean values of PF density and the mean density values of each MC phenotype were significantly different in the three groups (ANOVA group effect P < 0.001). High values of PF and MC density were found indeed in intestinal samples showing normal villous profile, while low values were associated with defective or atrophic villi. Results of Tukey's test for multiple comparisons are also reported in Table 2.

Table 2. Tissue density of pericryptal fibroblasts and mast cells in the duodenal lamina propria of 30 patients with different villous architecture
 Group 1Group 2Group 31 vs 2*1 vs 3*2 vs 3*
  • *

    Tukey's test was used for comparisons among groups.

PF775.16 ± 135.23354.75 ± 66.84342.49 ± 95.92<0.001<0.0010.962
Tryptase-positive MC641.70 ± 69.85318.57 ± 56.21189.55 ± 101.80<0.001<0.0010.003
Chymase-positive MC529.54 ± 117.39239.94 ± 53.82201.57 ± 72.50<0.001<0.0010.581
C-kit-positive cells789.74 ± 182.18365.30 ± 108.31356.87 ± 74.41<0.001<0.0010.988

Correlation analyses showed that the values of PF density correlated with the density of each MC phenotype when data were considered as a whole, i.e., independently from villous structure: PFs vs. tryptase-positive MCs, r = 0.913 (P < 0.001); PFs vs. chymase-positive MCs, r = 0.905 (P < 0.001); PFs vs. c-kit-positive cells, r = 0.927 (P < 0.001). As expected, when correlation was calculated in each group, data were not significant due to the low number of considered cases.

Estimation of the linear regression model to predict the pericryptal density of c-kit-positive MCs and MC precursors as function of PF density showed an intercept a = 29.171 (P = 0.473) with a slope b = 0.973 (P < 0.001) and a determination coefficient R2 = 0.859 (P < 0.001; Fig. 4). Calculation of slope values in each histological group showed that differences were not statistically significant (P = 0.321). Estimation of the linear regression model to predict PF density as a function of tryptase-positive MC density showed an intercept a = 101.715 (P = 0.10) with a slope b = 1.008 (P < 0.001) and a determination coefficient R2 = 0.834 (P < 0.001; Fig. 5). Slopes were not statistically different in the three histological groups (P = 0.171).

Figure 4.

Relation between pericryptal fibroblast densities and the frequencies of c-kit-positive cells in the whole examined material of human duodenal samples.

Figure 5.

Relation between the densities of pericryptal fibroblasts and the densities of pericryptal mast cells expressing tryptase reactivity in the duodenal lamina propria of examined samples.


PF/Epithelium Unit

The present study provides morphometric/quantitative evidence for a significant association between PF density and distinct patterns of villous architecture in the human duodenal mucosa (ANOVA group effect P < 0.001). High values of PF density are found indeed in intestinal samples showing normal villous profile (775.16 ± 135.23 cells/mm2), while low values are associated with defective or atrophic villi (354.75 ± 66.84 and 342.49 ± 95.92 cells/mm2, respectively; P < 0.001 in both cases). In addition, microscopic analysis reveals that PFs in groups 2 and 3 do not form well-organized envelopes around crypts, with PFs often situated at a certain distance from epithelial cells and without topographic relation with them (Fig. 1B and C).

This study suggests a close link between the microanatomical organization of the PF sheath and the general villous architecture. These quantitative findings fit well with previous functional data that emphasize the role of PFs in driving maturation and differentiation of crypt epithelial cells. PFs indeed secrete cytokines and growth factor peptides, such as TGF-β, hepatocyte growth factor (HGF), platelet-derived growth factor (PDGF)-A, and granulocyte/monocyte colony-stimulating factor (GM-CSF), which affect epithelial stem cell and descendant cell activity (Del Buono et al.,1992; Fritsch et al.,1997; Karlsson et al.,2000; Brittan and Wright,2002; Sennikov et al.,2002). PFs may also indirectly affect epithelial cell growth and differentiation by secreting extracellular matrix molecules and various components of the basement membrane, such as type I and IV collagen, fibronectin, tenascin, and proteoglycans (Powell et al.,1999a).

Cross-Talk Between PFs and MCs

In addition, this study demonstrates that PF density highly correlates with the densities of tryptase- or chymase-containing MCs as well as c-kt-reactive pericryptal cells (r = 0.913, 0.905, and 0.927, respectively; P < 0.001 in all cases). C-kit-positive cells in the crypt lamina propria encompass a composite cell population essentially made up of MC precursors and mature MCs. In humans, intestinal MCs arise from bone marrow-derived CD34+, CD13+, FcϵRI, c-kit+ precursors that undergo complete maturation once they have entered the lamina propria. Beside acquiring either tryptase or tryptase/chymase phenotypes, mature MCs express the receptor for the IgE Fc region (FcϵRI) and still maintain the stem cell factor (SCF) receptor.

The cross-talk between PFs and MCs possibly involves bidirectional instructive signals because PFs produce chemokines and cytokines [monocyte chemoattractant protein (MCP)-1, RANTES, and TGF-β], which are capable to attract MCs (Gruber et al.,1994; Romagnani et al.,1999). Remarkably, myofibroblasts are a rich source of SCF, the key regulator of MC “homing,” development, and differentiation (Bischoff and Dahinden,1992; Casola et al.,1997; Lorentz et al.,2002). Interestingly, our study reveals that PF and c-kit-positive cell densities present the highest correlation coefficient (r = 0.927). This suggests that PFs may contribute to attract and “home” c-kit-positive MC precursors in the pericryptal zone. In addition, PFs may cooperate in generating fully mature MCs that continue to express c-kit reactivity in addition to tryptase and/or chymase positivity. This second point is suggested by the high correlation coefficients observed between PF density and the occurrence of tryptase- and chymase-reactive MCs in the pericryptal lamina propria (r = 0.913 and 0.905, respectively). MCs, on the other hand, synthesize and release mediators such as TGF-β, TNF-α, PDGF, and SCF, which are capable of regulating myofibroblast differentiation and function (Metcalfe et al.,1997; Micera et al.,2001). Remarkably, our study shows that TGF-β and TNF-α are expressed by pericryptal MC subsets. TGF-β, in particular, is a key factor in generating phenotypically and functionally activated myofibroblasts (Schitt-Gräff et al.,1994). Recently, MC-secreted tryptase has been implicated in the process of skin fibroblast differentiation into mature myofibroblasts (Gailit et al.,2001). These in vitro functional data fit well with the high correlation coefficient found in our study between tryptase-positive MC and PF frequencies (r = 0.913). It is tempting to speculate that MCs, recruited in the pericryptal lamina propria via PF-secreted chemokines, may stimulate transformation of fibroblasts into PFs and drive their complete differentiation into fully mature PFs.

MCs and Villous Architecture

This study provides also evidence for a significant association between distinct pericryptal MC phenotypes and the pattern of villous architecture in the human duodenal mucosa (ANOVA group effect P < 0.001). High values of MCs expressing tryptase or chymase, as well as c-kit-positive MCs and MC precursors, are found indeed in samples with normal villous profile, while intermediate and low MC values are associated with defective or atrophic villi. The strongest association is between villous morphology and the density of tryptase-reactive MCs (Tukey's multiple comparison P < 0.001).

MCs might affect villous architecture in different ways. First, by secreting cytokines, such as PDGF, TGF-β, and GM-CSF, which drive intestinal epithelial cell growth and differentiation (Karlsson et al.,2000; Brittan and Wright,2002; Sennikov et al.,2002). Second, by releasing tryptase and chymase that affect stromal and epithelial cell functions, pericryptal extracellular matrix remodeling, and angiogenesis (Gruber et al.,1989; Blair et al.,1997). In addition, by releasing numerous angiogenic factors, such as vascular endothelial growth factor (VEGF), fibroblast growth factor-2 (FGF-2), histamine, heparin, TNF-α, and interleukin (IL)-8 (Grutzkau et al.,1998; Qu et al.,1998; Bischoff et al.,1999), which affect microrepair processes and prevent epithelial cell damage (Paris et al.,2001). As documented in the present study, a proportion of tryptase-positive MCs in the pericryptal lamina propria also expresses TNF-α reactivity. MCs may affect villous architecture by exerting neurotrophic and neurogenic effects via nerve growth factor (NGF) secretion (Tsui-Pierchala et al.,2002). Indeed, pericryptal nerves have recently been implicated in the regulation of crypt epithelial cell migration and differentiation (Bjerkens and Cheng,2001),

Taken together, the results of this study suggest a relationship between PFs, pericryptal MCs, and the architecture of intestinal villus. Our data provide morphological and morphometric support for a critical role of PFs in driving crypt epithelial cell proliferation and differentiation. As suggested by our study, defective PF organization may contribute to villous atrophy in inflammatory bowel conditions, such as celiac disease. Likewise, overproduction of this cell type is a recognized contributory factor to fibrotic diseases in many organs and may affect such bowel conditions as Crohn's disease (Graham,1995; Powell et al.,1999). In addition, our study suggests that PFs may guide c-kit-positive MC precursors close to the pericryptal zone and promote their maturation into tryptase- or chymase-positive MCs. MCs in turn may be crucial for modulating PF full differentiation, crypt epithelial cell function, and the trophism of specific structural constituents in the lamina propria. The pericryptal zone thus appears as a critical area conveying myriad of modulatory signals to crypt epithelial cells and villous stroma. In this context, MCs are candidate to play strategic functions. On the one side, they participate to immune and inflammatory processes taking place in the intestinal mucosa; on the other side, they are well endowed with mediators acting on distinct structural targets in the lamina propria and the overlying epithelium. Hence, we speculate that inflammatory events in the intestinal mucosa may be converted by pericryptal MCs into instructive signals leading to restoration of crypt epithelial cell structure and reconstitution of the pericryptal physiological environment. Further studies will hopefully shed light on how MCs synergistically cooperate with PFs to maintain crypt homeostasis and how they concur to promote a proper villous architecture.


This study was supported in part by grants from Ministero dell'Istruzione, Università e Ricerca, Rome, Italy, to the Sections of Anatomy and Pathology, Department of Medical and Morphological Research, University of Udine.