TGF-β1, TGF-β Receptor II and ED-A Fibronectin Expression in Myofibroblast of Vitreoretinopathy

Abstract

purpose. Formation of scarlike epiretinal membranes (ERMs) constitutes potentially the end stage of evolution of proliferative vitreoretinopathy (PVR) and proliferative diabetic retinopathy (PDR). Among various cellular populations, ERMs contain cells with contractile features typical of myofibroblasts. The current study was conducted to investigate the presence of transforming growth factor (TGF)-β1, TGF-β receptor II (RII) and ED-A fibronectin (FN), the main inducers of myofibroblastic differentiation in ERMs in PDR and PVR.

methods. Samples of ERM were obtained from 23 patients during microsurgery for PVR or PDR. Electron microscopy, immunohistochemistry, and confocal microscopy with antibodies recognizing α-smooth muscle (SM) actin, desmin, TGF-β1, TGF-β receptors I and II, and ED-A FN were performed.

results. α-SM actin was detected in all ERMs, whereas desmin was present in 50% of the cases. ED-A FN was expressed in all ERMs in close relation with α-SM actin–positive myofibroblasts. In addition, TGF-β1 and TGF-β R II were always present, TGF-β RII being expressed in bothα -SM actin–positive and negative fibroblastic cells.

conclusions. Myofibroblast accumulation is a key event in ERM-associated traction retinal detachment occurring during PVR and PDR. The current results suggest that the presence of α-SM actin–positive myofibroblasts is probably dependent on the concomitant neoexpression of TGF-β1, TGF-β RII, and ED-A FN. The results furnish new data on the mechanism of α-SM actin stimulation in fibroblasts in a human pathologic setting.

Proliferative vitreoretinopathy (PVR) is a major cause of retinal detachment surgery failure. ¹ It is characterized by the formation at the vitreoretinal interface in the vitreous cavity of fibrous epiretinal membranes (ERMs) that results from inappropriate proliferation, migration, and differentiation of several cell types. ² The formation of such scarlike fibrovascular membranes also occurs at the end stage of proliferative diabetic retinopathy (PDR) ³ and a variety of ocular disorders such as occur after trauma to the posterior segment of the eye or chronic intraocular inflammation. Gradual contraction of the ERM causes a marked distortion of the retinal architecture, eventually resulting in complex retinal detachments. ^{1 4} Although modern vitreoretinal microsurgery has improved detachment treatment in PVR and PDR, the functional prognosis remains poor. It is therefore important to better understand the pathogenesis of this disorder, to devise strategies for its prevention. 

The development of a scarlike ERM appears to be caused by a complex sequence of cellular and biochemical events. After an initial phase, characterized by a gliotic reaction, ² fibrotic changes, similar to those characterizing the proliferative fibroblastic phase of wound healing and subsequent scar formation, ⁵ predominate. During the process of granulation tissue formation and contraction, fibroblasts modulate into myofibroblasts whose main feature is the expression of the smooth muscle (SM) differentiation marker α-SM actin and, more rarely, desmin and SM myosin heavy chains. ⁶ Myofibroblasts have been shown to be responsible for fibrocontractive situations such as hypertrophic scars, fibromatoses and stromal reaction to epithelial tumors. ^{7 8} The presence of cells showing myofibroblastic features in ERMs has been previously reported in different ophthalmologic diseases. ^{9 10 11} A single study has also described the presence of α-SM actin expressing myofibroblasts in ERM after PVR or PDR. ¹²  

In the present study, we systematically investigated the presence of myofibroblasts in ERMs of different origins, using electron microscopy and immunohistochemistry with antibodies against several cytoskeletal markers. We extended our studies to the examination of transforming growth factor (TGF)-β1, observed in vitreous samples of PVR, ¹³ TGF-β receptor II (RII), and the ED-A isoform of cellular fibronectin (FN), the main inductors of α-SM actin expression. ^{14 15 16}  

Our results show that ERMs in PVR and PDR are characterized by the diffuse presence of α-SM actin–positive myofibroblasts containing TGF-β1 and TGF-βRII and expressing ED-A FN. In some fibroblastic cells, TGF-βRII expression allegedly precedes α-SM actin’s appearance. Confocal analysis demonstrates for the first time the close connection of α-SM actin–containing stress fibers and extracellular ED-A FN. This reinforces the assumption that myofibroblast modulation is responsible for retinal tractional detachment and furnishes some new data on the mechanism of α-SM actin stimulation in a human pathologic setting. 

Methods

Tissue Samples

Twenty-three samples of ERM (Table 1) were obtained from patients during microsurgery performed at the Department of Ophthalmology of Geneva University Hospital (Table 1) . The cases were classified as PVR (n = 14) and PDR (n = 9; Fig. 1 ). The clinical manifestation of ERM had been present less than 4 months, and local or diffuse tractional retinal detachment had been present less than 2 months. In three cases, specimens were taken at two time points, the first one corresponding to the primary ERM and the second one to a recurrence of the disease. All procedures adhered to the tenets of the Declaration of Helsinki. 

All operations were performed by the same surgeon using the same technique. Briefly, conventional vitreous surgery was performed with a three-port system, including a separate infusion cannula, a fiber-optic light probe and a port for use of a vitrectomy probe, vitreoretinal pick, and intraocular forceps. The central vitreous gel was removed first with the vitreous cutter, and the ERM was separated from the retina by classic vitreoretinal procedures. 

Antibodies

We used the following primary antibodies: mouse monoclonal IgG2a recognizing α-SM actin, ¹⁷ affinity-purified rabbit polyclonal IgG against desmin, ¹⁸ affinity-purified rabbit polyclonal IgG recognizing SM myosin heavy chain types 1 and 2, ¹⁹ mouse monoclonal IgG1 against vimentin, rabbit polyclonal IgG specific for glial fibrillary-associated protein (GFAP), mouse monoclonal IgG1 recognizing all types of cytokeratin (anti-pankeratin, Lu-5, all from Dako, Glostrup, Denmark), mouse monoclonal IgG1 specific for ED-A type III domain of cellular FN (IST-9, gift from Luciano Zardi, National Institute for Cancer Research, Genova, Italy) ²⁰ , affinity-purified chicken polyclonal IgG against TGF-β1 (R&D, Minneapolis, MN), and two affinity-purified rabbit polyclonal IgGs against TGF-β receptor I and II (TGF-βRI and II; Santa Cruz Biotechnology, Santa Cruz, CA). 

Histology, Immunohistochemistry, and Immunofluorescence

Tissue samples (eight PVR and seven PDR membranes) were fixed in 4% buffered formaldehyde and embedded in paraffin. Sections (5-μm-thick) were stained with hematoxylin and eosin. Immunostaining was performed on sections adjacent to the sections stained histologically. The presence of α-SM actin, desmin, SM myosin, TGF-β1, and TGF-βRI and II was determined by means of the streptavidin-biotin complex peroxidase or alkalin-phosphatase method (Dako), as previously described. ²¹ Immunoreactivity of desmin and SM myosin was intensified by one to three microwave treatments for 5 minutes in 10 mM citrate buffer (pH 6.0) before the first antibody was used. Sections were treated with one of the primary antibodies 1 hour at room temperature, except for TGF-βRI and II antibodies which were applied overnight at 4°C. This was followed by incubation with goat anti-mouse (for α-SM actin and desmin), anti-rabbit (for SM myosin and TGF-βRI and II) or anti-chicken (for TGF-β1) biotinylated antibodies (Jackson ImmunoResearch, West Grove, PA) and treatment with streptavidin-biotin-peroxidase or alkaline-phosphatase complex. The determination of peroxidase and alkaline-phosphatase activities was performed with diaminobenzidine (Serva, Heidelberg, Germany) and fast red (Dako), respectively. Slides were counterstained with hemalun and mounted (Aquatex; Dako). 

Double immunostaining with α-SM actin and TGF-βRII was performed as previously described. ²¹ Briefly, paraffin sections (from three PVR and four PDR membranes), after treatment with H₂O₂ to inhibit endogenous peroxidase, were incubated with anti-TGF-βRII antibody overnight. This was followed by incubation with goat anti-rabbit biotinylated antibody and treatment with streptavidin-peroxidase, which was revealed with diaminobenzidine. After washing, anti-α-SM actin antibody was applied for 1 hour, followed by incubation with goat anti-mouse-alkaline-phosphatase–conjugated antibody, which was revealed with fast red. Slides were counterstained with hemalun and mounted. 

Samples were observed with a photomicroscope (Axiophot; Carl Zeiss, Oberkochen, Germany) using an oil immersion plan-neofluar× 40/1.3 objective. Images were acquired with a high-sensitivity camera (Photonic Science Coolview; Carl Zeiss) using an acquisition software (Image Access 2.04; Imagic, Zürich, Switzerland). Images were processed with analysis software (Adobe Photoshop 5.0; Adobe Systems, Mountain View, CA) and printed with a digital printer (Pictography 4000; Fujifilm, Tokyo, Japan). 

Two researchers estimated independently the surface and the intensity of the staining for each antibody without taking into account the labeling into vessel walls. The evaluation of the staining was graded as follows: −, no labeling; +, focal labeling; ++, labeling in approximately 50% of the tissue; +++, extensive labeling. No differences were found in data obtained by the two observers. 

For the study of FN distribution we used immunofluorescence, because the antigenic properties of ED-A FN are destroyed by formol fixation. Tissues (nine PVR and seven PDR membranes) were embedded in OCT 4583 (Miles, Naperville, IL) and snap-frozen in precooled liquid isopentane. Three-micrometer-thick cryostat sections were fixed in acetone at− 20°C for 5 minutes and air dried for 2 hours at room temperature. They were incubated with anti-α-SM actin and anti-vimentin, anti-GFAP, anti-pankeratin, anti-ED-A FN antibodies for 30 minutes at room temperature. Goat anti-mouse IgG2a conjugated with tetramethylrhodamine isothiocyanate (TRITC; Southern Biotechnology, Birmingham, AL), goat anti-mouse IgG1 conjugated with fluorescein isothiocyanate (FITC; Southern Biotechnology), or goat anti-rabbit IgG labeled with TRITC (Jackson ImmunoResearch) were used as secondary antibodies. Nuclei were labeled with 4′,6-diaminido-2-phenylindol-dihydroclorid (DAPI; Fluka, Buchs, Switzerland). Preparations were mounted with polyvinyl alcohol. The surface and intensity of staining for α-SM actin and ED-A FN were estimated as described for immunohistochemistry. 

Confocal Laser Scanning Microscopy

Double immunofluorescence staining was performed on whole ERMs (five PVR and one PDR) fixed in methanol at −20°C for 15 minutes. Specimens were stained with anti-α-SM actin and anti-ED-A FN antibodies as described, except that primary and secondary antibodies were incubated overnight at 4°C. 

Specimens were observed with a confocal laser scanning fluorescence inverted microscope (LSM 410; Carl Zeiss) equipped with two lasers used simultaneously: a helium-neon (He-Ne) laser (excitation wavelength at 543 nm) and an argon laser (excitation wavelength at 488 nm). The excitation spectra were separated by a dichroic beam splitter of 488 and 543 nm and the emission spectra of the two fluorochromes were separated by a 560-nm dichroic beam splitter. Two detectors were used in parallel and were preceded with a 590- to 610-nm (rhodamine channel) or a 510- to 525-nm (fluorescein channel) narrow-band barrier filter. The partial superposition of the emission spectra of the two fluorochromes was negligible. ¹⁶ The specimen was observed through an oil immersion plan-neofluar ×63/1.4 objective, and the visual field was enhanced by zooming in two times. Between 30 and 50 optical sections of 512 × 512 pixels separated by 0.2 μm were performed in the z axis. A three-dimensional image was reconstructed with Imaris software (Bitplane, Zürich, Switzerland) running on computer workstation (Octane; Silicon Graphics, Mountain View, CA). Images were printed with a digital printer (Fujifilm). Colocalization of α-SM actin (red) and ED-A FN (green) was evaluated by counting the number of pixels containing both stains (yellow) in all optical sections using the Colocalization software (Bitplane). 

Electron Microscopy

Membrane tissue samples (two PVR and one PDR) were fixed in 1.5% glutaraldehyde in 0.1 M sodium cacodylate (Merck, Darmstadt, Germany) containing 1% sucrose for 3 hours. This was followed by fixation in 1% osmium tetroxide for 1 hour and subsequent dehydration and embedding in Epon. Semithin sections were stained with toluidine blue. Thin sections were treated with uranyl acetate and lead citrate and examined in an electron microscope (model 400; Philips, Eindhoven, The Netherlands). 

Results

Characterization of Myofibroblasts

Histologic examination showed that all ERMs contained several cell types (i.e., macrophages, retinal pigment epithelial [RPE] cells, and fibroblastic cells) and some small blood vessels. The results of immunostaining for cytoskeletal proteins are summarized in Table 1 .α -SM actin was detected in all ERMs obtained from eyes affected by PVR and PDR and was present in most of the fibroblastic cells (Fig. 2A ). Desmin was found in 50% of the specimens, either from PVR- or PDR-affected eyes. It was expressed in a small proportion of fibroblasts (approximately 10%) in five samples and in 50% of cells in two samples (Fig. 2B) . SM myosin heavy chains were present in only one sample of PDR and was graded as focal staining (data not shown). Double immunofluorescence staining with anti-α-SM actin and anti-vimentin or anti-GFAP or anti-keratin antibodies showed thatα -SM actin–positive cells expressed vimentin as well, but not GFAP or keratin. ED-A FN was strongly expressed in all ERMs, irrespective on the diagnosis (Table 1) , and localized around α-SM actin–positive cells. In specimens from a recurrence of the disease, cells expressed features similar to those in primary ERMs (Table 1) . 

Electron microscopy showed that cells present in PVR and PDR membranes exhibited typical features of myofibroblasts, ²² in particular their cytoplasm was filled with microfilament bundles (Fig. 3) . The extracellular material consisted essentially of collagen fibrils; thus, myofibroblasts were the major cell type present in ERMs. 

Expression of TGF-β1 and TGF-β Receptors

TGF-β1 was detected in all 15 PVR and PDR specimens, and labeling was graded from focal to extensive (Fig. 2C and Table 1 ). TGF-β1 was in general present in a smaller number of cells thanα -SM actin. TGF-βRII was expressed constantly in myofibroblasts in all specimens studied (Fig. 2D and Table 1 ), whereas TGF-βRI was negative. Double staining with α-SM actin and TGF-βRII (performed on three PVR and four PDR membranes) showed that fibroblastic cells expressed both; moreover, several cells positive for TGF-βRII alone were observed (Fig. 2E) . 

Three-Dimensional Reconstruction of Myofibroblasts

All PVR and PDR membranes used for confocal laser scanning microscopic analysis were strongly positive for α-SM actin and ED-A FN. Three-dimensional reconstruction by shadow projection showed α-SM actin–positive myofibroblasts (red) in which stress fibers appeared well defined (Fig. 4) . ED-A FN (green) was clearly detected in the extracellular compartment of the tissue and appeared as a network surrounding the myofibroblasts. Continuity between α-SM actin–positive stress fibers and ED-A FN filaments was visible (Fig. 4 , insert). However, only 0.058% ± 0.006% of the total pixel number exhibited a colocalization of α-SM actin and ED-A FN demonstrating that these proteins were almost never colocalized (yellow). 

Discussion

It is currently known that contraction of the ERMs is a cell-mediated event. ^{2 11 12 23} Two mechanisms, not entirely exclusive, have been proposed for the production of fibrotic tissue retraction: One involves an active contraction of myofibroblasts, ^{24 25} and the other is based on the motile activity of myofibroblasts that causes a remodeling of the surrounding extracellular matrix. ²⁶ Our results show that myofibroblasts, mainly of the type present in hypertrophic scars (i.e., expressing α-SM actin and rarely or never desmin and SM myosin heavy chains), ⁷ are the main cellular components of ERMs in both PVR and PDR. This suggests that myofibroblasts play a role in the production of retractile phenomena causing retinal detachment in both conditions, similar to other fibrocontractive diseases. ^{6 7 27} If untreated, ERMs tend to cause a progressive retinal detachment that begins locally but eventually may involve many retinal quadrants. It seems unlikely that the strong tractional force necessary to induce such a detachment results merely from extracellular matrix remodeling induced by isolated migrating fibroblasts. 

In parallel to myofibroblast accumulation, we have shown that significant amounts of TGF-β1, as well as one of its specific receptors, TGF-βRII, are present in ERMs. TGF-β1 is mainly localized within macrophages or in connection with extracellular matrix. Our findings suggest that the initial stage of myofibroblast differentiation coincides with the presence of TGF-βRII followed byα -SM actin appearance. TGF-β1 has been shown to be the most important stimulus for α-SM actin and collagen synthesis. ^{14 15 28 29 30} Recently, our laboratory has shown that TGF-β1 needs the presence of the cellular ED-A FN variant to stimulate synthesis of both collagen and α-SM actin. ¹⁶ As described by Immonen et al., ³¹ we have observed that ED-A FN was present in ERMs in all samples studied. Therefore, factors known to be capable of producing contractile events resulting in retinal detachment are present in ERMs. Further studies are needed to establish the mediators of TGF-β1 synthesis by macrophages and other inflammatory cells. Presently, granulocyte macrophage colony-stimulating factor is thought likely to be responsible for this action. ^{32 33 34}  

Cells with the ultrastructural characteristics of myofibroblasts have been previously identified in contractile ERMs removed during surgery, ^{9 10} but their number was reported to be scarce, probably because most of these reports concerned nonvascular ERMs obtained during macular pucker surgery. More recently, α-SM actin expressing myofibroblasts were found in up to 90% of PVR and PDR specimens, ¹² but they were not evaluated quantitatively, nor was their organization within the membranes studied. In the present study, α-SM actin expressing myofibroblasts were found to be present in 100% of the membranes. Immunostaining for α-SM actin was prominent in all membranes and dense bundles of actin microfilaments forming stress fibers within the myofibroblasts were observed by electron microscopy, indicating a high contractile potential. Additionally, confocal microscopy clearly demonstrated that the myofibroblasts formed a dense sheet of cells, similar to those developing when fibroblasts are cultured in culture dishes, ³⁵ connected to one another and associated with ED-A FN. The continuity observed between α-SM actin–positive stress fibers and ED-A FN-positive extracellular bundles can be attributed to the fibronexus, another structure typical of myofibroblasts. ³⁶  

It has been suggested that RPE cell migration into the vitreous cavity during various vitreoretinal diseases secondarily differentiates into fibroblasts and stimulates the production of collagen and FN at the retinal surface by releasing various cytokines, including TGF-β1. ³⁷ However, transdifferentiation of RPE cells to fibroblasts is controversial. ¹¹ Indeed, RPE cells may adopt a fibroblastic appearance but remain keratin positive, thus retaining their epithelial nature. ¹¹ Hyalocytes have also been postulated as a possible source of fibroblasts, but they have macrophage rather than fibroblast characteristics. ³⁸  

In the present study, myofibroblasts in ERM specimens were positive for vimentin and α-SM actin and negative for keratin and GFAP. The absence of keratin staining excludes an RPE cell origin of the myofibroblasts. That specimens were constantly negative for GFAP and positive for vimentin suggests that myofibroblasts originate from astrocytes. However, we cannot exclude an origin from GFAP-negative nonactivated cells. ^{39 40} Finally, because ERM contain blood vessels, a pericytic origin of myofibroblasts is also possible. ^{41 42}  

In conclusion, we have shown that, similar to other fibrocontractive settings, ERMs are characterized by the presence of typical α-SM actin–positive myofibroblasts accompanied by TGF-β1, its specific receptor TGF-βRII, and ED-A FN. Further work examining the modulation of TGF-β1 and ED-A FN expression is needed to explore the mechanisms of ERM development and possibly to intervene in their onset and evolution. 

 Preliminary results of this work were presented by CJP at the Swiss Society of Ophthalmology Meeting, 1997, and at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May, 1998, and were summarized in Klinisch Monatsblatter Augenheilkunde 1998;212:356–358.

 Supported by the Swiss National Science Foundation, Grants 32-49543.96 and 31.50568.97.

 Submitted for publication February 3, 1999; revised November 30, 1999; accepted December 28, 1999.

 Commercial relationships policy: N.

 

Table 1.

 

Clinical and Immunohistochemical Features of Epiretinal Membranes

Case	Age	Sex	Disease	α-SM Actin	SM Myosin	Desmin	TGF-β1	TGF-βRII	ED-A FN
1*	57	M	§ >PVR	+++	−	++	++	ND	ND
2*	57	M	PVR	+++	−	+	+	ND	ND
3	60	M	PVR	+++	−	+	+	+++	+++
4	52	F	PVR	+++	−	−	+++	ND	+++
5	56	F	PVR, §	+++	ND	ND	ND	ND	+++
6	14	M	PVR, §	+++	−	+	++	+	+++
7, †	76	M	PVR	+++	−	−	+++	+	ND
8, †	76	M	PVR	++	−	−	++	++	ND
9	67	F	PVR	++	−	−	+	++	ND
10	61	M	PVR	+++	ND	ND	ND	ND	+++
11	88	M	PVR, §	+++	ND	ND	ND	ND	+++
12	60	M	PVR, §	+++	ND	ND	ND	ND	+++
13	68	M	PVR, §	+++	ND	ND	ND	ND	+++
14	84	M	PVR	+++	ND	ND	ND	ND	++
15	62	M	PDR	+++	−	+	+++	ND	+++
16	79	F	PDR	+++	−	+	+++	ND	+++
17, ‡	37	F	PDR, §	+++	+	++	++	+++	+++
18, ‡	37	F	PDR	+++	−	−	+++	+++	+++
19	69	F	PDR	+++	−	−	+++	+++	+++
20	61	F	PDR	+++	−	−	+++	+	ND
21	71	F	PDR	++	−	−	+	+++	ND
22	65	M	PDR	+++	ND	ND	ND	ND	++
23	34	F	PDR	++	ND	ND	ND	ND	++

 Pairs of symbols (

*  ,

†  ,

‡  ) represent two specimens from the same patient obtained at two time points. ND, not done because specimens were not appropriately fixed for the technique used.

§  Specimens used for confocal laser scanning microscopy.

Figure 1.

 

Preoperative (A) and postoperative (B) red-free photographs of a PDR-affected eye. Fibrovascular tissue extending from the optic nerve head causes tractional macular detachment and tractional retinal detachment nasally to the optic disc (A). The removal of preretinal fibrovascular tissue results in retinal reapplication (B). Multiple laser scars after macular focal and panretinal photocoagulation are seen in (B).

Figure 2.

 

Immunolocalization of α-SM actin (A), desmin (B), TGF-β1 (C), and TGF-β RII (D) in four representative ERMs from PDR (A, B) and PVR (C, D). Note the strong expression of α-SM actin (A) and desmin (B) in nearly all cells. TGF-β1 (C) and TGF-β RII (D) also show extensive intracellular labeling. Double immunostaining for α-SM actin and TGF-β RII in a representative ERM from PDR (E) shows that they are colocalized in many fibroblastic cells and thus appear red-brown because of the mixture of peroxidase and alkaline phosphatase staining (arrows). Several cells express only TGF-β RII and appear brown (arrowhead). Scale bars, (A through D) 50 μm; (E) 20 μm.

Figure 3.

 

Electron micrograph showing a typical myofibroblast (My) from a PDR membrane. Microfilament bundles are visible within the cell (arrows) and are illustrated at higher magnification in the inset. The myofibroblast is surrounded by collagen fibers. A macrophage (M) and another myofibroblast (My) are also visible. Magnification, ×7,800; inset, ×21,000.

Figure 4.

 

Three-dimensional reconstruction by shadow projection of α-SM actin–positive myofibroblasts (red) and ED-A FN (green) obtained after confocal laser scanning microscopy. Well-defined α-SM actin–positive stress fibers appear. ED-A FN forms a network surrounding the myofibroblasts. Scale bars, 10μ m.

The authors thank Catherine Cametti and Philippe Henchoz for expert technical assistance and Jean-Claude Rumbeli and Etienne Denkinger for photographic work.