Integrin α₁β₁ is involved in the differentiation into myofibroblasts in adult reactive tissues in vivo

Abstract

Connective tissue cell activation is of importance during reactive conditions such as solid tumour growth, wound healing and pannus formation in rheumatoid arthritis. Here, we have compared connective tissue cells of mesenchymal origin in human tissues from these conditions and their normal counterparts using a panel of cell-type-specific markers. In particular, we investigated variations of integrin expression among connective tissue cell phenotypes. Connective tissue cell populations were defined based on their association with the microvasculature and their expression of activation markers. The phenotype of these cells varied according to the type of pathological connective tissue examined. Our morphological data from human tissues suggested that the α₁β₁ integrin, a collagen/laminin receptor, is involved in the differentiation of precursor cells into myofibroblasts. To mechanistically investigate this hypothesis, we employed experimental models for carcinoma growth and wound healing utilizing α₁ integrin-deficient mice. The data confirmed that the α₁β₁ integrin is of importance not only for the differentiation of mesenchymal cells into myofibroblasts but also for the neovascularization and connective tissue organization and emphasize the importance of myofibroblasts in the pathophysiology of tissue repair, inflammation and tumour growth.

Introduction

Stroma formation in solid tumours, chronic inflammatory lesions and tissue repair share several features including infiltration of inflammatory cells, activation of blood vessels and angiogenesis. Of further and central pathophysiological importance, persistent activation of connective tissue cells leads to excessive extracellular matrix (ECM) deposition, dominated by collagen type I, which, in turn, leads to fibrosis and ultimately organ dysfunction [1–5]. The amount of fibrosis is not necessarily linked to the severity of inflammation, indicating mechanisms, in part, distinct from those that regulate inflammation [6]. Thus, tissue damage due to severe inflammation can, in some instances, be reversible with the reinstatement of organ architecture and function [7]. The underlying processes that result, on the one hand, in reinstatement of organ function and, on the other hand, in a chronic state resulting in fibrosis and organ dysfunction despite similar initial pathophysiology are largely unknown.

Myofibroblasts, as defined by their expression of α-smooth muscle actin (α-SMA), play a central role in the deposition and organization of ECM and thus also in the formation of fibrotic tissue [5, 8, 9]. They are related to fibroblasts and exhibit a hybrid phenotype between fibroblasts and smooth muscle cells/pericytes [5]. It has been suggested that the latter two cell types are derived from a common cell lineage [5, 10–14]. The origin of the myofibroblast is yet unclear. Resident tissue fibroblasts [5, 14], vascular cells such as smooth muscle cells and/or pericytes [10–13] and bone marrow-derived precursor cells [14] have been suggested as potential sources.

The transition to the myofibroblast phenotype in culture depends on the concerted action of cytokines such as transforming growth factor (TGF)-β and specific ECM proteins such as the fibronectin splice variant ED-A and on mechanical tension [8, 15]. Several adhesion receptors belonging to the integrin family have been implicated in the regulation of the myofibroblast phenotype in vitro. The α4β1 and α5β1 integrins promote the myofibroblast phenotype by binding to fibronectin [16]. The αvβ1 integrin binds latent TGF-β, which in concert with the αvβ6 integrin leads to TGF-β activation [17] and the development of the myofibroblast phenotype. The αvβ3 and αvβ5 integrins down-regulate α-SMA expression in myofibroblasts via binding to vitronectin [18]. In vitro studies have also suggested that the α1β1 integrin is important for the up-regulation of α-SMA in cultured fibroblasts subjected to interstitial fluid flow [19]. Thus, adhesion and the nature of that adhesion seem to be important in the orchestration of events leading to acquisition of the myofibroblast phenotype.

In the present study, tissue analysis from different human pathologies suggested that the α1 integrin subunit, which associates exclusively with the β1 integrin subunit, forming the α1β1 integrin heterodimer [20, 21], a major collagen receptor, is important for the differentiation and maintenance of the myofibroblast phenotype, as defined by the expression of α-SMA. We tested this hypothesis using a genetic approach employing mice carrying a null mutation in the gene for the integrin α1 chain [22]. Our data demonstrate a previously unrecognized importance of adhesion via the α1β1 integrin in acquisition of the myofibroblast phenotype in vivo, which, in turn, is of central importance for the neoformation of vessels and supporting connective tissue structures.

Materials and methods

Surgical specimens

Full-thickness biopsies from human tissues were taken from: (i) colorectal adenocarcinoma (CC; n= 4), (ii) adjacent normal colon (n= 2), (iii) pannus formation from synevectomies due to rheumatoid arthritis (RA; n= 3), (iv) normal synovia (n= 3), (v) 7-day-old healing cutaneous wounds (n= 3) and (vi) adjacent normal skin (n= 2). The biopsies were snap-frozen. The human ethics committee at the Uppsala Academic Hospital, Sweden, approved the present study.

Antibodies and other reagents

The following antibodies were used: monoclonal antibody (mAb) F4/80, recognizing macrophages (Serotec, Raleigh, NC, USA); polyclonal antibody (pAb) anti-NG2, recognizing pericytes and smooth muscle cells [23] (Chemicon, Temecula, CA, USA); mAb anti-perlecan, recognizing heparan sulphate proteoglycan (Chemicon); FITC-labelled mAb anti-smooth muscle α-actin (α-SMA), recognizing smooth muscle cells, myofibroblasts and pericytes [24–26] (clone 1A4; Sigma, St. Louis, MO, USA); biotinylated mAb anti-CD31 (PECAM-1), recognizing endothelial cells (Pharmingen, San Diego, CA, USA); pAb anti-platelet-derived growth factor (PDGF) β-receptor (clone 958; Santa Cruz Biotechnology, Santa Cruz, CA, USA); mAb anti-reticular fibroblast marker (RFM), recognizing fibroblasts [27] (clone ER-TR7; Cederlane laboratories, Ontario, Canada); mAb ASO2 (Thy1/CD90), recognizing fibroblasts, pericytes and smooth muscle cells [28] (Dianova, Hamburg, Germany), mAb anti-high-molecular-weight-melanoma-associated antigen (HMW-MAA), expressed on activated pericytes and smooth muscle cells [26, 29, 30] (clone 225.28; Sanbios, Uden, The Netherlands); mAb PAL-E, recognizing endothelial cells (Sanbios); mAb anti-procollagen type I C propeptide [31] (Pierce Chemical Corp., Rockford, IL, USA); mAb PDGFR-B2, recognizing activated PDGF β-receptors when used at a concentration of 1 μg/ml [32]; mAb anti-α1 integrin subunit [20] (clone TS2/7; Dr. Timothy Springer, Boston Blood Center, Boston, MA, USA); mAb anti-α₂ integrin subunit (P1H5 and PIE6), mAb anti-α₃ integrin subunit (P1B3) and mAb anti-α₅ integrin subunit (PID6) [21, 33] (Dr. William Carter; Fred Hutchinson Cancer Research Center, Seattle, WA, USA); mAb anti-α₆ integrin subunit [34] (clone GOH3; Dr. Arnoud Sonnenberg, Netherlands Cancer Institute, Amsterdam, The Netherlands) and mAb anti-α₂ integrin subunit (CD49b), mAb anti-α₃ integrin subunit (Cdw49c) and mAb anti-α₅ integrin subunit (CDw49e) (Chemicon). The different mAbs used to detect the expression of α₂β₁ (PIE6, P1H5 and CD49b), α₃β₁ (P1B3 and Cdw49c) and α₅β₁ (P1D6 and CDw49e) integrins did not differ quantitatively or qualitatively (data not shown). The P1E6, P1B3 and P1D6 mAbs were used in the subsequent stainings.

The following secondary reagents were used: biotinylated (Fab)₂ fragments (DAKO, Glostrup, Denmark), biotinylated IgG (Vector Labs, Burlingame, CA, USA), fluorescein- and rhodamine-conjugated IgG (Vector Labs), alexa flour 350 streptavidin- and alexa flour 350-conjugated IgG (Molecular Probes, Eugene, OR, USA) Texas Red avidin D (Vector Labs), normal serum (Sigma) and non-immune IgG (Sigma). The antibodies were diluted in PBS, pH 7.4, containing 0.1% bovine serum albumin, 150 mmol/l tranexamic acid, 20 μg/ml aprotinin (3 TIU/mg), 1.8 mmol/l ethylenediaminetetraacetic acid and 2 mmol/l iodoacetic acid. The trichrome Masson kit (Sigma) was used according to the manufacturer’s instructions.

Immunolabelling, imaging and numerical analysis

The 6-μm tissue sections were fixed in 100% acetone and subjected to immunolabelling for light and immunofluorescence microscopy, as described previously [26, 35]. From each animal and each human specimen, 10 fields of view at a magnification of ×200 were chosen randomly and digitalized. Quantification of CD31, α-SMA and RFM was performed with the NIH imaging software version 1.62 (National Institute of Health, Bethesda, MD, USA) in the animal studies. Quantification of human samples was performed with the IC 300 image-processing system from Inovision Corp. (Research Triangle Park, NC, USA), as previously described [26, 35]. Statistical analysis was performed with the Student’s t-test and considered significant at P-values below 0.01.

In vivo Matrigel assay

The α₁-deficient mice [22], maintained on the 129sv/Tae background, and corresponding wild-type mice were bred at the animal facility at BMC, Uppsala, Sweden. The mice were anaesthetized by an intraperitoneal injection of 2.5% avertin (Sigma) in a volume of 200 μl saline and were then injected subcutaneously with a total of 200 μl of Matrigel™ (Costar; Fischer Scientific, Brightwaters, NY, USA), at a final concentration of 9.9 mg/ml, containing 1.4 μg/ml VEGF, 8 μg/ml FGF2 and 116 μg/ml bovine serum albumin (fatty acid-free), purchased from Sigma (St. Louis, MI, USA) and 500 nM sphingosine 1-phosphate from Biomol (Plymouth Meeting, PA, USA) [36]. The animals were killed by CO₂ narcosis after 14 days. All animal experiments were approved by the Ethical Committee for Animal Experiments in Uppsala, Sweden, and Scripps Institute, San Diego, CA, USA.

Experimental tumour mouse model

CT26 colon adenocarcinoma autografts were grown from subcutaneous inoculation of 1 × 10⁶ cells on the dorsum of syngeneic wild-type (n= 4) and α₁^−/− (n= 4) BALB/c mice (backcrossed to BALB/c for at least nine generations), as previously described [37].

Results

Immunofluorescence stainings, combined with computer-aided image analyses, were performed to characterize cell types with regard to the expression of particular antigens (markers) using a modification of a previously described methodology [26, 35] (Fig. 1). In a first step, immunohistochemistry was used to determine whether specific cell-type markers defined by the use of antibodies were expressed in vessels or in the interstitium (Fig. 1, step 1, and Fig. 2). In a second step, double immunofluorescence stainings (DIFS) using different combinations of antibodies directed to cell-type markers were used to determine their spatial interrelationship (Fig. 1, step 2, and Fig. 3). In a third step, the images from step 2 were analysed using computer-aided image analyses. A percentage value measuring the degree of co-localization, that is, spatial overlap between two different markers (X and Y), recognized by the respective antibodies on the same tissue section, was calculated (Fig. 1, step 3, Tables 1 and S1). A background co-localization ranging between 6% and 22%, with a mean of 19 ± 8%, was recorded for markers not expected to co-localize (Table 1) such as PAL-E (endothelial cells) and α-SMA (pericytes, smooth muscle cells and myofibroblasts). Furthermore, markers that were expected to completely co-localize such as PAL-E and PAL-E did not exceed a value of 80% due to restrictions in the technology (data not shown). These limits were in agreement with earlier studies using this technology [26, 35]. In a fourth step, data on the degree of co-localization from a number of different marker combinations, depicted as percentage values (Table 1 and Table S1), and the actual size of the individual marker populations, depicted as the number of pixels per field of vision (Table S2), was compiled and compared (Fig. 1, step 4). This procedure enabled us to determine the spatial distribution between the different markers in step 5 (Fig. 1, step 5). By including well-established cell-type-specific markers, we could correlate the spatial distribution of the markers under study in step 6 to in vivo structures/cell types of interest (Fig. 1, step 6) presented in Table 2.

Figure 1.

Method used to allocate expression of individual markers to different in vivo cell types. Tissues were subjected to immunohistochemical staining using different markers and analysed with light microscopy (1). Double immunofluorescent staining using combinations of markers was performed (2). Using computer-aided image analysis, co-localization between different markers was quantified (3). Co-localization between different sets of markers was analysed (4). From the analysis in steps 2 and 4, the spatial distribution and their co-localization could be determined (5). Using cell-type-specific markers, the spatial distribution of markers could be correlated to in vivo structures/cell types (6).

Figure 2.

Expression of phenotypical markers in colorectal adenocarcinoma (CC) and pannus formation in rheumatoid arthritis (RA). Immunohistochemical staining was performed in 6-μm sections from CC (left column) and RA (right column) using mAbs to characterize expression profiles of phenotypical markers in relation to the vasculature for PAL-E (A, B), α₁β₁ (C, D), α-SMA (E, F) and ASO2 (G and H). Expression profiles were similar in microvascular structures (arrow) in both CC and RA. However, expression profiles differed in interstitial structures (arrowhead) between the two conditions. In CC, the expression profiles in interstitial structures were positive for α₁β₁ (C), α-SMA (E) and ASO2 (G). In contrast, the expression profiles in interstitial structures in RA were negative for α₁β₁ (D) and α-SMA (F), but positive for ASO2 (H). Tumour acinar structures (*). The bar represents 200 μm.

Figure 3.

Phenotypical characterization of connective tissue cells in colorectal adenocarcinoma (CC) and rheumatoid arthritis (RA). Triple immunofluorescence staining of 6-μm sections of CC (left column) and RA (right column) was performed with mAbs to characterize the degree of co-localization as well as the spatial distribution between PAL-E (A, B), α₁β₁ integrin (C, D) and α-SMA (E, F). Co-localization of α₁β₁ integrin and α-SMA is depicted in yellow (G, H). Coexpression of α₁β₁ integrin and α-SMA was seen in both CC (G) and RA (H). Vascular PAL-E-positive structures (arrow) in both CC (A) and RA (B) were positive for α₁β₁ integrin and α-SMA (G, H). In CC, interstitial structures (arrowhead) expressed α₁β₁ integrin and α-SMA (G). In contrast, interstitial structures (arrowhead) in RA were negative for α₁β₁ integrin and α-SMA (H). Tumour acinar structures (*). The bar represents 200 μm.

Table 1.

Biopsies from tumour stroma formation in colorectal adenocarcinoma, pannus formation in rheumatoid arthritis and cutaneous healing wounds

Percentage of smooth muscle α-actin-positive pixels that are also positive for	Colorectal adenocarcinoma	Rehumatoid arthritis	Wound healing
HMW-MAA	28 ± 11	49 ± 23	47 ± 19
PDGF β-receptor	33 ± 14	47 ± 16	46 ± 22
Integrin α1 subunit	80 ± 11	72 ± 9	86 ± 14
Integrin α5 subunit	80 ± 11	72 ± 14	86 ± 15
PAL-E	14 ± 8	22 ± 11	20 ± 10

Biopsies were stained by a double immunofluorescence technique with various combinations of mAbs. Co-localization between two different markers is depicted as “percentage of pixels that are positive for both marker X and Y”. Stained sections were analysed by computer imaging processing, as detailed in Materials and methods. The recorded percentage values represent the spatial distribution of two mAb markers in relation to each other, measured in percentage of pixels that co-localize (1 pixel equals 0.9 μm × 0.9 μm) [26, 35].

Table 2.

Spatial distribution of the markers used in this study in relation to the microvasculature based on the method used in Figure 1

	Normal skin		Normal colon		Colorectal adenocarcinoma		Rheumatiod arthritis		Wound healing
	EC (PAL-E)	P (α-SMA)	EC (PAL-E)	P (α-SMA)	EC (PAL-E)	P (HMW)	EC (PAL-E)	P (α-SMA)	EC (PAL-E)	P (α-SMA)
α-SMA	−	NA	−	NA	−	++	−	NA	−	NA
HMW-MAA	−	−	−	−	−	NA	−	++	−	++
PDGF β-receptor	−	−	−	−	−	++	−	++	−	++
a1β1 integrin	++	++	++	+	++	++	++	++	++	++
a2β1 integrin	++	+	++	+	++	−	−	−	−	−
a3β1 integrin	++	+	++	+	++	−	++	−	++	−
a5β1 integrin	−	++	+	−	++	++	++	++	++	++
a6β1 integrin	++	−	++	−	++	−	++	−	++	−

	Normal skin	Normal colon	Colorectal adenocarcinoma	Rheumatoid arthritis	Wound healing
	I	I	I (α-SMA)	I (HMW)	I (HMW)
α-SMA	−	−	NA	−	−
HMW-MAA	−	−	−	NA	NA
PDGF α-receptor	−	−	−	++	++
a1β1 integrin	−	−	++	−	−
a2β1 integrin	+	+	+	−	−
a3β1 integrin	−	−	−	−	−
a5β1 integrin	−	−	++	++	++
a6β1 integrin	−	−	−	−	−

Human tissues studied were normal skin, normal colon, tumour stroma formation in colorectal adenocarcinoma, pannus formation in rheumatoid arthritis and cutaneous healing wounds. Endothelial cells (EC) detected using PAL-E; pericytes (P) detected using α-SMA or HMW; interstitial fibroblast-like cells (I) detected using α-SMA or HMW.

Stained sections were analysed by computer imaging processing, as detailed in Materials and methods. The recorded percentage values represent the spatial distribution of two mAb markers in relation to each other, measured in percentage of pixels that co-localize (1 pixel equals 0.9 μm × 0.9 μm) [26, 35].

−, less than 30% co-localization; +, between 30% and 60% co-localization; ++, more than 60% co-localization.

Expression of β₁ integrin heterodimers in colorectal adenocarcinoma, cutaneous healing wounds and pannus formation in rheumatoid arthritis

The distribution of β₁ integrin-associated α-subunits expressed in reactive tissues and their normal counterparts were studied focusing on the expression patterns in the microvascular compartment, including the endothelium and pericytes, and the non-vascular interstitial compartment, including fibroblasts. The α₁, α₂, α₃ and α₅ integrin subunits associate exclusively with the β₁ integrin subunit, whereas the α₆ integrin subunit is capable of associating with both β₁ and β₄ integrin subunits [20, 21]. The results are shown in Table 2. For detection of blood vessels, the expression of PAL-E was used as an endothelial marker (Fig. 2A and B and Fig. S1A and B). To identify the non-endothelial cell type in the microvasculature, pericytes and interstitial fibroblasts, immunohistochemical staining of tissue sections using antibodies recognizing PDGF β-receptors (Fig. S1C and D), HMW-MAA (Fig. S1E and F), α-SMA (Fig. 2E and F and Fig. S1G and H) and ASO2 (Fig. 2G and H) and fluorescent multi-labelling of combinations of these markers (Fig. 3A–H and Fig. S2A–J) were employed and analysed according to the method outlined in Fig. 1.

In the tumour stroma of colorectal adenocarcinoma, endothelial cells expressed α₁β₁, α₂β_1,α₃β_1,α₅β₁ and α₆ integrins (Table 2). Microvascular pericytes and interstitial fibroblasts had similar integrin expression profiles and mainly expressed α₁β₁ and α₅β₁ integrins (Table 2). The latter two cell types both expressed α-SMA but could be distinguished by their spatial relationship to the endothelium and their differential expression of HMW-MAA and PDGF β-receptors (Table 2 and Fig. S1C–F), as previously described [26]. Tumour acinar structures expressed α₂β_1,α₃β₁ and α₆ integrins (data not shown).

Endothelial cells in pannus formation in rheumatoid arthritis and in tissue situated immediately adjacent to maturing granulation tissue in 7-day-old cutaneous healing wounds expressed a similar integrin subunit profile compared with colorectal adenocarcinoma with the exception of α₂β₁ integrin, which was not detected in the endothelium in cutaneous healing wounds or pannus formation (Table 2). Similar to what was observed in colorectal adenocarcinoma, microvascular pericytes in rheumatoid arthritis and cutaneous healing wounds expressed the α₁β₁ and α₅β₁ integrins (Table 2). In contrast to colorectal adenocarcinoma, interstitial fibroblasts, while expressing α₅β₁ (Fig. S2E and F), did not express the α₁β₁ integrin (Fig. 2C and D and Fig. 3C–H). Thus, the results suggested that the α₁β₁ integrin could be used as a marker to differentiate pericytes from interstitial fibroblasts in rheumatoid arthritis and cutaneous healing wounds as well as in defining interstitial fibroblast populations in the reactive conditions under study.

Phenotypical characterization of connective tissue cells in tumour stroma, pannus formation in rheumatoid arthritis and cutaneous wound healing in human beings

The possibility that the α₁β₁ integrin was linked to specific populations of non-endothelial connective tissue cells was investigated using the method outlined in Fig. 1. Pericytes and interstitial fibroblasts could be defined in the tumour stroma in colorectal adenocarcinoma, pannus formation in rheumatoid arthritis and cutaneous healing wounds by their spatial relationship to the microvasculature (Fig. 3A–H and Fig. S2A–J) and differential expression pattern of integrins and cell-type-specific markers (Tables 1 and 2). Microvascular pericytes were defined as cells that were juxtapositioned to the endothelium. Interstitial fibroblasts include cells that are partly dissociated from the microvascular wall or connective tissue cells situated in the interstitium. The latter cells were often accumulated in the perivascular space. However, connective tissue cells were also observed that were not overtly associated with blood vessels. Interstitial fibroblasts could be subsetted into α-SMA-positive/ASO2-negative cells (Fig. 2E and G and Fig. 3E and G), termed myofibroblasts, and α-SMA-negative/ASO2-positive cells (Fig. 2F and H and Fig. 3F and H), termed fibroblasts.

As shown in Table 2, the expression of markers in microvascular pericytes was similar in the tumour stroma, pannus formation and cutaneous healing wounds. These cells expressed the α₁β₁ integrin, α-SMA, PDGF β-receptors, HMW-MAA and α₅β₁ integrin. In addition, α₂β₁ integrin was detected in microvascular pericytes in cutaneous healing wounds. In the normal tissue, pericytes were negative for PDGF β-receptors and HMW-MAA, consistent with previous studies identifying these markers as activation markers [26, 30].

Connective tissue cells in the tumour stroma of colorectal adenocarcinoma not associated with vessels expressed α-SMA and integrins α₁β₁ and α₅β₁, but not the PDGF β-receptor or HMW-MAA. On the basis of their expression of α-SMA and the positioning relative to the vasculature, these cells were identified as myofibroblasts. The corresponding population of cells in rheumatoid arthritis and cutaneous healing wounds expressed α₅β₁ integrins, ASO2, the PDGF β-receptor and HMW-MAA. They, however, did not express α-SMA or α₁β₁ integrins. Because of the lack of α-SMA, these cells did not meet the criteria for myofibroblasts and were therefore identified as fibroblasts. These findings provide further support that the expression of α₁β₁ integrins co-distributes with that of α-SMA.

Procollagen type I C-propeptide used as a marker for collagen type I synthesis was not expressed in pericytes in any of the conditions studied. However, connective tissue cells that express procollagen type I C-propeptide were often seen to be concentrated in the surrounding perivascular space (Fig. 4A–D). In colorectal adenocarcinoma, these procollagen type I C-propeptide-expressing cells also expressed α-SMA, thus identifying them as myofibroblasts (Fig. 4C). In contrast, in rheumatoid arthritis and cutaneous healing wounds, procollagen type I C-propeptide did not co-localize with α-SMA (Fig. 4D), but was expressed in HMW-MAA- and PDGF β-receptor-positive connective tissue cells corresponding to fibroblasts (data not shown). Thus, in tumour stroma, but not in rheumatoid arthritis and cutaneous healing wounds, α-SMA expression co-distributes with collagen type I synthesis. This would suggest that the myofibroblast phenotype is not an absolute prerequisite for collagen type I neoformation in connective tissue cells.

Figure 4.

Procollagen type I expression in colorectal adenocarcinoma (CC) and pannus formation in rheumatoid arthritis (RA). Immunohistochemistry (A, B) and double immunofluorescence (C, D) staining of sections of CC (A, C) and RA (B, D) were performed with mAbs recognizing procollagen type I c-propeptide (procollagen), used as a marker for neosynthesis of collagen type I and smooth muscle α-actin (α-SMA). Co-localization is depicted in yellow. (A–D) Neosynthesis of collagen type I was not expressed in the microvasculature (arrowhead), but was largely confined to interstitial cells in the perivascular space (arrow) in both conditions. In these cellular structures, co-localization with α-SMA was high in CC (C) in contrast to RA (D), where co-localization was low. The bar represents 40 μm.

The role of α₁β₁ integrins in modulating connective tissue cell phenotype and function in an in vivo Matrigel plug assay

The morphological data in human tissues suggest that the α₁β₁ integrin co-distributes with α-SMA and raise the possibility that the α₁β₁ integrin is important in the acquisition of the myofibroblast phenotype. To study this correlation in detail, two experimental in vivo assays were performed with wild-type and integrin α₁-deficient mice [22]. In the first set of experiments, Matrigels were inoculated subcutaneously in mice and harvested after 14 days [36].

As previously described [36], the wild-type mice displayed a substantial ingrowth of vessels into the Matrigel (Fig. 5A). Newly formed vessels were, to a large extent, surrounded by NG2- and PDGF β-receptor-expressing pericytes invested in a perlecan-positive basement membrane (data not shown and [36]). RFM-expressing fibroblasts were present in the Matrigel. These cells also expressed α-SMA. Because they were not associated with the microvasculature, they were defined as myofibroblasts (Fig. 5A and C). In the adjacent dermis, accumulations of connective tissue cells that expressed RFM and α-SMA were associated with the microvasculature, thus identifying them as pericytes. Some of these cells were also partly or completely dissociated from the vascular wall and accumulated in the perivascular interstitial space (Fig. 5E). Similar accumulations of perivascular cells were observed in the dermis adjacent to Matrigels implanted in α₁-deficient animals, although to a lesser extent. These cells were positive for RFM but negative for α-SMA (Fig. 5D). The Matrigels in α₁-deficient mice had a reduced density of CD31-positive structures compared with gels harvested from wild-type animals, suggesting a disturbance of angiogenesis in an α₁ integrin-deficient environment (Fig. 5A and B). Furthermore, the cellular content of Matrigels inoculated in α₁-deficient mice was positive for RFM but not for α-SMA (Fig. 5D). In addition, in both conditions, F4/80-expressing macrophages had infiltrated the Matrigel (data not shown).

Figure 5.

Deficient myofibroblast development and neoformation of blood vessels and supporting connective tissues in the in vivo Matrigel plug assay in α₁^−/− mice. Double immunofluorescence (A–F) staining using combinations of the mAbs, α-SMA, RFM and CD31, and trichrome (G–H) staining were performed on sections of Matrigel plugs together with adjacent skin that had been grown in the subcutaneous space in wild-type (WT; left column) and α₁^−/− (KO; right column) mice. Optical sections were obtained using confocal laser microscopy and viewed as composites (A–F). Co-localization is depicted in yellow. (A, B) Note ingrowth into the Matrigel (M) of CD31-positive microvessels (arrowhead) surrounded by α-SMA-positive myofibroblasts (arrow) in wild-type (A) but not in α₁^−/− mice (B). (C, D) Note ingrowth into the Matrigel (M) of myofibroblasts positive for both RFM and α-SMA in wild-type (C) mice in contrast to Matrigels in α₁^−/− (D) mice that contained RFM-positive fibroblasts that were negative for α-SMA (arrow). (E, F) Note RFM-expressing cells that are juxtapositioned to, partially detached (arrow) from, the CD31-positive endothelium (arrowhead) that have accumulated in the perivascular space in dermis (*) adjacent to the Matrigel (M) in wild-type (E) mice and, albeit to a lesser degree, also in α1^−/− (F) mice. (G, H) Note well-developed cellular connective tissue septa (arrow) in Matrigels in wild-type (G) but not in α1^−/− (H) mice, where only isolated accumulations of cells (arrow) were observed. Matrigel/dermal interface (dotted line). The bar represents 40 μm.

Quantification of blood vessel density showed a 90% reduction in the area covered by CD31-positive pixels in the Matrigels implanted in deficient animals compared with wild-type animals (Fig. 7A). In these animals, no formation of connective tissue septa in the Matrigel was observed compared with wild-type animals, where both vessels and myofibroblasts were primarily concentrated to networks of connective tissue septa extending from the normal adjacent tissue (Fig. 5G and H). However, in α₁-deficient mice, the Matrigels were not acellular as both macrophages (data not shown) and RFM-expressing fibroblasts were present (Fig. 5D). A 53% reduction in area covered by RFM-positive pixels (Fig. 7A) and a 97% reduction in the area covered by α-SMA-positive pixels were observed in α₁-deficient mice (Fig. 7A). Thus, RFM-expressing fibroblasts were capable of infiltrating the Matrigel, but they did not differentiate into myofibroblasts. Also, in contrast to wild-type mice, NG2-positive pericytes were absent in the gel (data not shown).

Figure 7.

Quantification of the number of pixels positive for markers expressed in the endothelium, pericytes, fibroblasts and myofibroblasts in syngeneic colorectal adenocarcinomas (CC) grown in wild-type and α₁^−/− mice. Matrigels (A) or syngeneic CCs grown in the subcutaneous space (B) in wild-type (filled bars) and α₁^−/− (empty bars) mice were extirpated, sectioned and triple stained with regard to the markers for the endothelial cells (CD31), pericytes/smooth muscle cells and myofibroblasts (α-SMA) and pericytes and fibroblasts (RFM). Average number of positive pixels per field of view was quantified. Mean ± SD. P-value according to Student’s t-test, *P < 0.01.

The role of α₁β₁ integrins in modulating connective tissue cell phenotype in tumour stroma formation in experimental colorectal adenocarcinoma

In a second experimental model, syngeneic CT26 colon adenocarcinoma cells were inoculated into the subcutaneous space in wild-type and α₁-deficient BALB/c mice. In the solid tumours that subsequently formed, the tumour stroma could be divided into two main compartments: a vascular compartment and a perivascular compartment. The expression of cell-type-specific markers and their spatial interrelationship were examined (Fig. 6A–J). In the tumour stroma of wild-type and α₁-deficient mice, a microvascular network consisting of endothelial cells that were partially covered by pericytes (Fig. 6G and H) and enveloped in the basal lamina (data not shown) was present. In both wild-type and α₁^−/− mice, these microvascular pericytes expressed α-SMA (Fig. 6G and H), RFM (Fig. 6I and J) and NG2 (data not shown). Furthermore, in both conditions, the majority of cells constituting the perivascular compartment were positive for RFM (Fig. 6I and J). In contrast to α₁-deficient mice, these perivascular cells also expressed α-SMA (Fig. 6I and J) in wild-type mice, thus identifying them as myofibroblasts. In α₁-deficient mice, the expression of α-SMA was largely confined to the vascular space (Fig. 6J). The fine fibrillar/reticular cellular extensions creating a meshwork interconnecting the microvasculature in tumours in wild-type mice (Fig. 6G and I) were not observed in tumours in α₁-deficient mice (Fig. 6H and J). Furthermore, the overall cellular density was lower in tumours in wild-type mice compared with α₁-deficient mice (Fig. 6K and L).

Figure 6.

Deficient myofibroblast development in a syngeneic experimental colorectal adenocarcinoma in α₁^−/− mice. Triple immunofluorescence (A–J) using combinations of the mAbs, CD31 (blue (b)), α-SMA (green (g)) and RFM (red (r)), and trichrome (K–L) staining were performed on sections of biopsies from CC grown syngeneically in the subcutaneous space in wild-type (A, C, E, G, I, K) and α₁^−/− (B, D, F, H, J, L) mice. Optical sections obtained using confocal laser microscopy and viewed as composites (A–J). Co-localization is depicted in yellow. (G, H) Note α-SMA-positive cells associated with CD31-positive vessels (arrowhead) in tumours from both wild-type and α₁^−/− mice. Note α-SMA-positive interstitial myofibroblasts (arrow) in the tumour stroma in tumours from wild-type mice (G) that are not associated with vessels and the apparent lack of similar cells (arrow) in tumours from α₁^−/− mice (H). (I, J) Note co-localization of RFM and α-SMA in cells associated to vessels (arrowhead) in tumours from both wild-type (I) and α₁^−/− (J) mice. Interstitial cells (arrowhead) express both RFM and α-SMA in tumours from wild-type mice (I) while only expressing RFM in tumours from α₁^−/− mice (J). (K, L) Note lower cell density in tumours from wild-type (K) mice compared with α₁^−/− (L) mice. The bar represents 20 μm.

A quantification made for the markers above (Fig. 7B) showed a 46% reduction in the area covered by CD31-positive pixels in tumours in α₁-deficient mice compared with wild-type mice. Moreover, in α₁-deficient mice, a 53% reduction in the area covered by α-SMA-positive pixels compared with tumours in wild-type mice could be observed. This was not due to a decrease in fibroblasts because there was a 48% increase in the area covered by RFM-positive pixels in tumours in α₁-deficient mice compared with wild-type mice.

Discussion

The results of this study demonstrate a previously unrecognized role of adhesion via the α₁β₁ integrin in the acquisition of the myofibroblast phenotype in vivo, which, in turn, is shown to be of central importance for the neoformation of vessels and supporting connective tissue structures. In the animal models employed, we were able to show that the α₁β₁ integrins plays a role in the differentiation of precursor cells into myofibroblasts and the establishment of blood vessels. Previously, in vitro experiments have shown that antibody-mediated blocking of α₁β₁ integrins in cultured fibroblasts reduces the expression of α-SMA when subjected to interstitial fluid flow [19]. The evidence in the present study suggests that the α₁β₁ integrins are also important for the development of myofibroblasts in vivo.

In the present study of human reactive tissues, interstitial fibroblasts differed in their marker expression profiles in tumour stroma compared with pannus formation and wound healing. One explanation for this might be tissue specificity or that these cells are derived from different populations of precursor cells [14]. Heterogeneity of fibrogenic cells not only between tissues but also within the same tissue has previously been shown [38–40]. For instance, in the liver, phenotypically divergent fibrogenic cells are derived from different populations of precursor cells [41, 42]. Although there are important pathophysiological similarities between stroma formation in solid tumours and wound healing, the composition of the matrix is not identical [4]. Thus, one possibility for this divergence is the differences in cytokine expression and composition of the ECM components of the microenvironment. It has been shown that spatially distinct connective tissue cell populations respond differently to different cytokines [15]. For instance, a distinct population of periportal cells in liver undergoes a PDGF-BB-dependent conversion to a pro-fibrotic phenotype distinct from myofibroblasts in septal fibrosis [43]. The current observations made in human reactive tissues suggest, and are supported by the experimental models used in this study, that interactions between the ECM and connective tissue cells modify the phenotype of the cells, and that this interaction uses the α₁β₁ integrin as a conduit for this purpose.

The cellular origins of fibroblasts/myofibroblasts are unclear. We have previously shown that microvascular pericytes become activated in tumour stroma, pannus in rheumatoid synovitis and healing wounds [26, 44, 45]. Upon activation, these pericytes express several markers including PDGF β-receptors, HMW-MAA and α-SMA. Thus, the activation of microvascular pericytes in these conditions are reflected in a distinct molecular marker profile. We and others have reported that pericytes, upon activation, detach from the endothelial lining and enter the perivascular space where they alter their phenotype [10, 26, 46–48]. During tumour-induced angiogenesis and cutaneous wound healing, pericytes undergo a mitotic burst that precedes cell division in the endothelium and interstitial fibroblasts, resulting in the accumulation of connective tissue cells in the perivascular space [49]. In the present study, accumulations of connective tissue cells were radially distributed and concentrated around the vasculature. Furthermore, the findings in human reactive conditions show similarities in the marker expression profile between the microvascular pericyte and the interstitial fibroblast/myofibroblast. Thus, one possibility is that fibroblasts/myofibroblasts originate from microvascular pericytes, where pericytes detach from the endothelium, proliferate and subsequently differentiate into fibroblasts/myofibroblasts. This notion is supported by previous studies in excessive dermal scarring and placenta formation [11, 13]. Furthermore, gene deletion studies in mice show that PDGF and their receptors, although important for the expansion of the pericyte population during embryogenesis, are crucial for the expansion of the fibroblast population in adult reactive tissues, suggesting a common origin of these two cell types [50, 51]. Pericytes would thereby link the processes of connective tissue formation and the microvasculature in a previously unrecognized way. However, the results of the present study do not definitively prove a pericyte origin of fibroblasts/myofibroblasts in reactive conditions and this requires further investigation.

We have previously shown that the Matrigel plugs implanted into the subcutaneous space in vivo become infiltrated by macrophages and fibroblasts that support the ingrowth of blood vessels and the neoformation of connective tissue septa [36]. In similar models, using Matrigel plugs containing VEGF-A-transfected cells and syngeneic tumour models in mice, inhibition of α₁β₁ and/or α₂β₁ integrins perturbed ingrowth of blood vessels and tumour growth [52–54]. Our present results using Matrigel plugs implanted in α₁-deficient and wild-type mice show that fibroblasts are capable of populating the Matrigel but are incapable of differentiating into myofibroblasts. In concordance with the present Matrigel studies, fibroblasts are generated within tumours but are unable to undergo myofibroblast differentiation. However, in contrast to the Matrigel studies, tumour vasculature was present and was partially covered by α-SMA-expressing pericytes. The degree of reduction in vascularization in tumours grown in α₁-deficient mice compared with wild-type mice is in agreement with previous studies using this model [37, 54]. The differences between the Matrigel and syngeneic colorectal adenocarcinoma models may partly be due to the more complex array of ECM and cytokines within the tumour microenvironment. Furthermore, in the tumour model, co-option and expansion of already existing vasculature may occur, limiting the need for the neoformation of vessels and surrounding stroma [55].

The differences in connective tissue cell phenotype in the Matrigels from α₁-deficient mice compared with wild-type mice may be due to the differences in the inflammatory response. Monocytes/macrophages express the α₁β₁ integrin. Therefore one possibility is that the differences in fibroblast phenotype in the α₁-deficient mice is due to the lack of inflammatory cells, which, in turn, produce cytokines that influence the fibroblast phenotype [56, 57]. In the present study using immunocompetent mice, no gross differences in macrophage infiltration/population in the Matrigel or the colorectal adenocarcinoma model were observed. This argues against the possibility that the differences in connective tissue cell phenotype was due to the lack of infiltration of α₁β₁ integrin-expressing inflammatory cells. The simplest explanation is that precursor cells derived from the adjacent dermis were unable to migrate/populate the Matrigel or the tumour stroma due to a lack of α₁β₁ integrin [41, 42]. However, in both α₁-deficient and wild-type mice, connective tissue cells were able to migrate into the Matrigel and populate the non-vascular component of the tumour stroma. This would argue against the possibility that a lack of precursor cells could explain the observed differences. Taken together, our results show that, regardless of the event that triggers the transition into the myofibroblast phenotype in these conditions, it is dependent on the cellular expression of the α₁β₁ integrin.

Mice have been generated containing gene deletions for many of the known integrin subunits, and in some cases, double knockouts have been developed [58]. Most integrin subunits are present in only a single heterodimer, that is, the α₁ integrin subunit, which only associates with the β₁ subunit to form the heterodimer α₁β₁. A few integrins can form heterodimers with multiple partners, that is, α_v, making the interpretation of gene ablation experiments difficult [59]. The lack of a phenotype may suggest overlapping functions among different proteins. It can also be due to compensatory mechanisms in response to ablation of a certain gene, that is, up-regulation of VEGF receptor 2 in β₃ null mice [60, 61]. Discordance between pharmacological and genetic studies in mice adds to the complexity of interpreting the potential role of integrins in vivo[60]. The latter may be partially explained by transdominant inhibition, where occupation of one type of integrin leads to inhibition of the function of other specific integrins in the same cell [62]. Furthermore, biological processes during embryonic development differ from those in adult tissues and during pathological conditions in the adult. The expression of the α₁β₁ integrin is important in adult reactive tissues, but not during embryonic tissue morphogenesis. The α₁-deficient mice are viable and fertile and, besides a hypocellular dermis with a low number of fibroblasts and increased collagen and collagenase synthesis, have no overt phenotype [22, 54]. Evidence supporting discrepancies in the role of the fibroblast/myofibroblast lineage in embryonic versus adult reactive tissues does exist [50, 51]. It has been shown that the α₁β₁ integrin plays a role in reactive tissues such as the stromal response in solid tumours and the neoformation of blood vessels with a supporting connective tissue [52–54]. One possibility is that embryonic tissue morphogenesis, compared with repair processes in the adult where damaged tissue is replaced, differs in cell–matrix interactions [7]. Furthermore, it has been shown that myofibroblasts are not a prominent feature in foetal tissue repair, a process that results in minimal scarring [7, 15]. Our results support the notion that, although the α₁β₁ integrin is non-essential during embryogenesis [22, 54], it is essential in modulating connective tissue cell phenotypes, and thereby plays an important role in reactive tissues in the adult.

Taken together, our results support the in vivo results in human pathologies and show that, regardless of the events that triggers the transition of mesenchymal cells into the myofibroblast phenotype, it is dependent on the cellular expression of the α₁β₁ integrin. Thus, modulation of the connective tissue phenotypes and their environment may shift the balance from deleterious fibrosis towards optimal tissue regeneration and repair.

Supporting Information

Fig. S1 Expression of phenotypical markers in colorectaladenocarcinoma (CC) and in pannus formation in rheumatoid arthritis(RA). Immunohistochemical staining was performed in sectionsfrom CC (left column) and RA (right column) using mAb?s in order tocharacterize expression profiles of phenotypical markers.Expression profiles were similar in microvascular structures(arrowhead) in both CC and RA for PAL-E (A and B),PDGF β-receptors (C and D), HMW-MAA (Eand F) and α-SMA (G and H). However,expression profiles differed in interstitial structures (arrow)between the two conditions. In CC interstitial structures werepositive for α-SMA (G), but to a lesser degree PDGFβ-receptors (C) and HMW-MAA (E). In contrast,the expression profile in interstitial structures in RA werepositive for PDGF β-receptors (D) and HMW-MAA(F), but not for α-SMA (H). Bar represents 20μm.

Fig. S2 Phenotypical characterisation of connective tissuecells in the stroma of colorectal adenocarcinoma (CC) and in pannusformation in rheumatoid arthritis (RA). Doubleimmunofluorescence-stained sections of CC (left column) and RA(right column) was performed using mAb?s in order to characterizethe degree of colocalization as well as the spatial distribution oftwo markers in relation to each other and to the microvasculature.Images were captured and analyzed using computer imaging analysis[26, 35]. Colocalization is depicted in yellow. mAbs usedrecognized PDGF β-receptors (PDGFR-β), activatedpericytes (HMW-MAA), smooth muscle α-actin (α-SMA),α1β1 integrin (α1β1) and &lapha;5β1integrin (α5β1). Expression profiles were similar inmicrovascular structures (arrowhead) in both CC and RA, and werepositive for PDGF β-receptors (A and B), HMW-MAA(C, D, G and H), α5β1 (E andF), α1β1 (G, H, I and J) andα-SMA (A?F, I and J). However, expressionprofiles differed in interstitial structures (arrow) between thetwo conditions. In CC, interstitial structures were to a lesserdegree positive for PDGF β-receptors (A) and HMW-MAA(C and G) while maintaining their expression ofα-SMA (A, C, E and I), α5β1(E) and α1β1 (G and I). In contrastthe expression profile in interstitial structures in RA were to amuch lesser degree positive for α-SMA (B, D, F andI) and α1β1 (H and J) whilemaintaining their expression of PDGF β-receptors and HMW-MAA(D and H) and α5β1 (F). Barrepresents 40 μm.

Table S1. Biopsies from tumour stroma formation in colorectal adenocarcinoma, pannus formation in rheumatoid arthritis and cutaneous healing wounds.

Biopsies were stained by a double immunofluorescence technique with various combinations of monoclonal antibodies. Co-localization between two different markers are depicted as ?percentage of pixels that are positive for both marker X and Y?. Stained sections were analysed by computer imaging processing, as detailed in Materials and methods. The recorded percentage values represent the spatial distribution of two mAb markers in relation to each other, measured in percentage of pixels that co-localize (1 pixel equals 0.9 μm× 0.9 μm) [26, 35].

Table S2. Average number of pixels per field of vision for the different markers.

Biopsies were stained by double immunofluorescence techniquewith various combinations of monoclonal antibodies. Stainedsections were analysed by computer imaging processing, as detailedin Materials and Methods. The recorded values represent the averagenumber of pixels (×10³) (1 pixel &equals; 0.9μm × 0.9 μm) present in each field of vision (300 μm× 300 μm) from stained sections of colorectaladenocarcinoma, normal colon, rheumatoid arthritis, wound healingand normal skin for the different mAb?s used in this study. Mean± S.D. (number of fields of vision).

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.