Molecular Vision 2003; 9:74-79 <>
Received 20 December 2002 | Accepted 14 March 2003 | Published 17 March 2003

Keratocyte matrix interactions and thrombospondin 2

David J. Armstrong, Paul Hiscott, Mark Batterbury, Stephen Kaye

Unit of Ophthalmology, Department of Medicine, University of Liverpool, Liverpool, UK

Correspondence to: David Armstrong, Unit of Ophthalmology, Department of Medicine, UCD, Daulby Street, University of Liverpool, Liverpool L69 3GA, UK; Phone: +44 151 706 4017; FAX: +44 151 706 5934; email:


Purpose: To determine whether human keratocytes synthesize thrombospondins 2 and 3 (TSP-2, TSP-3) in a collagen matrix and the effect of addition of antibodies to TSP-2 and TSP-3 on keratocyte-populated collagen matrix contraction.

Methods: Keratocyte-populated collagen matrices were evaluated for TSP-2 and TSP-3 mRNA. Sections of matrices were stained by immunohistochemistry for TSP-2 and TSP-3. Keratocyte populated collagen matrices were treated with antibodies specific for TSP-2 and TSP-3 and these preparations were evaluated for contraction and keratocyte morphology.

Results: Keratocyte derived fibroblasts in collagen matrices contained TSP-2 and TSP-3 mRNA, and the cells were immunoreactive for both proteins. Compared to controls, an antibody specific to the N-terminal domain of TSP-2 significantly inhibited matrix contraction for up to 10 days at a concentration of 20 μg/ml antibody. At 2 μg/ml TSP-2 antibody concentration significant inhibition occurred up to 3 days. Removal of the antibody from the media reversed the inhibitory effect. Cultured keratocytes in TSP-2 treated collagen matrices appeared more rounded than keratocytes in control matrices. An antibody specific to TSP-3 had no effect on matrix contraction or keratocyte morphology.

Conclusions: Keratocyte derived fibroblasts synthesize TSP-2 and TSP-3 when seeded in collagen matrices. Antibody specific to TSP-2 reversibly inhibits matrix contraction. TSP-2 may play a key role in keratocyte/collagen matrix interactions, as may occur during corneal stromal repair.


The thrombospondins are a family of at least five secreted multidomain glycoproteins, known as thrombospondin 1, 2, 3, 4 (TSP-1 to -4) and cartilage oligomeric matrix protein (COMP, also termed thrombospondin 5 or TSP-5), respectively [1,2]. These macromolecules have a close structural relationship at their carboxyl-terminal regions where they all contain thrombospondin type 2 and type 3 repeats [3]. Conversely, the N-terminals of the proteins are distinct [2]. Indeed, TSP-1 and TSP-2 possess a procollagen homology domain and three thrombospondin type 1 repeats which the remaining three members of the family lack [4]. Furthermore, TSP-1 and TSP-2 form trimers whereas TSP-3, 4 and 5/COMP form pentomers. Because of their structure, the thrombospondins have been assigned to two subgroups; subgroup A (TSP-1 and TSP-2) and subgroup B (TSP-3, 4 and 5/COMP) [2].

The precise functions of the thrombospondins are unclear and somewhat controversial. They are particularly expressed during tissue formative processes, such as embryogenesis and repair but there are tissue distribution differences between the family members, which suggest that each member of the family has discrete and non-redundant roles. Thus TSP-1 seems to be the only thrombospondin present in platelets (where it represents about 40% of the α-granules) and it may function in platelet aggregation [5]. Conversely, the musculoskeletal system is a major site of TSP-2 and TSP-5/COMP expression. Both proteins are found in embryonic cartilage. TSP-2 is also present in embryonic tendon and skeletal muscle, while TSP-5/COMP persists in adult cartilage and appears in adult tendon [6,7]. Although TSP-3 and TSP-4 can be detected in muscle as well, they are both observed in adult neural tissue and TSP-3 is also highly expressed in developing and adult lung [8].

At the cellular level, the thrombospondins appear to regulate cell behaviour by modulating cell-matrix interactions [2]. The evidence for such a role comes partly from investigations concerning wound repair and thrombospondins. For example, treatment of cutaneous wounds with antisense tsp-1 oligonucleotides delays dermal healing and the cell-mediated phenomenon of wound shrinkage is also delayed in TSP-1-null mice [9]. Conversely, repair is accelerated and collagen fiber organization is abnormal in the wounds of TSP-2-null mice [10,11].

Further evidence that thrombospondins are involved in cell-matrix interactions during repair has emerged from our studies of keratocytes derived from adult cornea. Keratocytes in normal cornea contain TSP-1 mRNA but do not appear to express the protein [12,13]. Moreover, normal keratocytes in situ do not have detectable mRNA for any other member of the thrombospondin family [12]. However, keratocyte derived fibroblasts can synthesize TSP-1, -2, and -3 (though not apparently TSP-4 or TSP-5/COMP) in tissue culture (an environment that may be regarded as a simple wound model) and they express the same three thrombospondins in damaged adult corneal stroma [14]. Finally, the cells produce TSP-1 in the keratocyte-populated collagen matrix model of corneal stroma repair [13]. In this model, keratocyte derived fibroblasts exhibit many of the features observed in keratocytes in healing stromal wounds such as cell-mediated matrix contraction and fibroblastic or myofibroblastic morphology. Given the inhibitory effects of TSP-1 neutralisation in dermal repair, we added TSP-1 antibodies to the corneal stroma repair model, but this approach did not alter keratocyte behaviour in the matrices. Therefore, we hypothesised that one or both of the other thrombospondins produced by keratocyte-derived fibroblasts in damaged cornea (TSP-2 or TSP-3) may be key mediators of keratocyte activities in collagen matrices. To investigate this concept, we first determined whether keratocyte derived fibroblasts synthesise TSP-2 and/or TSP-3 in the stromal repair model and then, employing antibodies specific to these two thrombospondins, studied the effects of modulating the glycoproteins in the matrices.


Human keratocyte cell culture

Cell cultures of human corneal keratocytes were established from donor corneas, obtained from the local eye bank (Saint Paul's Eye Transplant Center, Royal Liverpool University Hospital, Liverpool, UK), as previously described [12]. Briefly, a corneal button with a diameter of 7 mm was punched out and Descemet's membrane (with attached endothelium) was stripped off and discarded. The epithelium was removed by alcohol treatment and scraping. The remaining stroma was rinsed in phosphate buffred saline (PBS) containing penicillin, streptomycin and fungizone. The stroma was then cut into small pieces that were transferred to 25 ml culture flasks and left for 5 min to enable adherence of the tissue to the flask surface. Fetal calf serum (FCS, 2 ml) was added to each flask, followed by 4 ml of a 1:1 mixture of Dulbecco's modified Eagle's medium with Ham's nutrient mixture F-12 (DMEM-F-12). All medium was supplemented with 10% FCS and 1% each of glutamine, penicillin, streptomycin and fungizone. The purity of the cultures was evaluated by negative immunostaining for cytokeratins (to exclude epithelial contamination) and by the kite- and spindle-shaped morphology of the cells in primary culture. Cultures were kept in a 5% CO2, humidified atmosphere at 37 °C and medium was changed every three days. During passage, keratocyte derived fibroblasts were detached using a 0.1% trypsin/EDTA solution and cultures were used in experiments between third and fifth passage.

Preparation of keratocyte-populated collagen matrices

Keratocyte-populated collagen matrices were prepared using the method descibed by Mazure and Grierson [15]. Rat-tail type I collagen (Sigma, Poole, UK) was dissolved at 5 mg/ml in 0.1% acetic acid (Sigma) to create a collagen stock solution. To prepare a collagen matrix at a final concentration of 1.5 mg/ml for each assay, 2.1 ml of concentrated culture medium (15 ml of 10X MEM, 35 ml distilled water, 1.5 ml penicillin/streptomycin, 1.5 ml glutamine, 1.5 ml fungizone and 3 ml of 7.5% sodium bicarbonate) was added to 3.6 ml collagen solution at 4 °C in the presence of sodium hydroxide. To this mixture, 0.9 ml of serum free medium (DMEM:F-12, 1:1) containing the appropriate number of cells was added. A range of cell concentrations (50,000-1,600,000 cells/ml) was initially investigated in order to determine a dose response for the relationship between cell density and the rate of matrix contraction. Gel contraction studies were performed in 24 well plates (Sterilin, Stone, UK), in which each well received 0.25 ml of the final mixture. The gels were transferred to a humidified 37 °C, 5% CO2 incubator where the matrix set within 1 min. After 10 min the matrices were overlayed with 1.5 ml of medium, detached from the base using a pipette tip and floated. Identical preparations 1 ml in size were made in six well plates to provide sufficient cells for the RT-PCR experiments.

RT-PCR and PCR amplification of keratocyte-populated collagen matrices

Keratocyte populated 3D collagen matrices (1 ml, 200,000 cells/ml) were removed from culture medium and briefly washed in PBS. The gels were then placed in a cryovial and snap frozen in liquid N2. The collagen matrices were then completely homogenised for 2 min in 600 μl of Buffer RLT (which contained 10 μl β-mercaptoethanol per 1 ml of buffer RLT, QIAGEN [QIAGEN, Ltd., Crawley, West Sussex, UK]). RNA was extracted from these samples (after homogenization) using the RNeasyTM Mini Kit (QIAGEN). This RNA was digested with Dnase 1 to ensure no cellular DNA was carried over from the extraction. First strand synthesis RT-PCR using oligo-dT primers was carried out to produce a cDNA copy of the RNA using Abgene's first strand synthesis RT-PCR kit. PCR was then performed using 2 μl of cDNA template, 0.5 μl each of forward and reverse (1 μg/μl) specific oligonucleotide primers (the sequences of which have been previously published [16]) to tsp-2 and tsp-3. PCR was carried out using one pair of oligonucleotides per reaction only. A total reaction volume of 25 μl was achieved by adding 22 μl of Abgene's PCR Master Mix (1.5 mM MgCl2). The cycling conditions for all primer pairs were as follows; stage 1, 94 °C for 2 min; stage 2, 30 cycles of 94 °C for 30 s; 57 °C for 30 s; 72 °C for 30 s; stage 3, 72 °C for 5 min. The validity of the primers had been previously established using restriction digest [16]. Control PCR was carried out by replacing the cDNA template with a control produced by omitting enzyme from the first strand synthesis step, this ensured no DNA contamination of the extracted RNA occurred. Positive control was observed using GAPDH primer pairs. PCR products were visualized by gel electrophoresis on a 2% agarose gel containing 0.5 μm/ml ethidium bromide.

Immunohistochemical staining of keratocyte-populated collagen matrices

After fixation in 10% neutral buffered formalin, matrices were dehydrated though an ascending series of graded ethanol and embedded in paraffin wax. Sections 5 μm thick of wax-embedded tissue were cut and placed on 3% aminopropylethoxysilane (APES, Sigma) coated glass slides. The sections were de-waxed and rehydrated through descending grades of alcohol to distilled water. Immunohistochemistry was performed on the sections as previously described [12]. Briefly, sections were incubated in 1% hydrogen peroxide, washed in PBS and non-specific protein binding was blocked with rabbit serum (Dako, High Wycombe, UK). Sections were then incubated overnight in a humid chamber at 4 °C, with affinity purified goat antibodies specific to TSP-2 or TSP-3 (sc-7655 and sc-7656 respectively, Santa Cruz Technology, Inc., Santa Cruz, CA) diluted at 1:100 (2 μg/ml) in PBS. Following three PBS washes, sections were incubated with biotinylated rabbit anti-goat secondary antibody (Dako). After further PBS washes, they were incubated with HRP conjugated streptavidin (Dako) and washed. Immunoreactive sites were visualized brown with diaminobenzidine and mounted for bright field or differential interference contrast microscopy. Negative controls were prepared by prior absorption of the primary antibody with the appropriate peptide antigen (Santa Cruz) and were otherwise subject to all immunohistochemical procedures.

Keratocyte-populated collagen matrix contraction experiments

Collagen matrices 0.25 ml in size were prepared at the optimal cell seeding density for contraction (200,000 cells/ml; see below). The affinity-purified goat IgG raised against TSP-2 and TSP-3 (see above) were used at final concentrations of 20 μg/ml and 2 μg/ml in serum free DMEM:F-12 (1:1) media. Pre-immune goat IgG (Santa Cruz) was diluted to an equivalent protein concentration in DMEM:F-12 (1:1) as a control. Each collagen matrix received 1.5 ml of the DMEM:F-12 (1:1) containing antibodies or pre-immune IgG immediately after the matrix had set. Each antibody dilution was conducted in quadruplicate. The gel area was measured on fine graph paper at days 1, 3, and 7 after preparation and represented as area of gel (with standard error of the mean). Each experiment was conducted a minimum of three times. Cell morphology within the collagen matrices was followed by phase contrast microscopy throughout the duration of the experiments. To determine whether the effects of the TSP-2 antibody were reversible, matrices were incubated with 20 μg/ml concentration of this antibody for seven days. On day seven medium was refreshed in all wells. Medium was replaced with serum free DMEM:F-12 (1:1) medium in four of the wells which previously contained 20 μg/ml of TSP-2 antibody. In these experiments, gel contraction was measured at days 1, 3, 7, and 10 after seeding. This experiment was conducted in quadruplicate.

Statistical analysis

Statistical analysis of data was performed using the statistics package SPSS (SPSS, Inc., Chicago, IL). Comparisons of data samples from contraction and adhesion assays were by analysis of variance (ANOVA), which allowed multiple comparisons between data sets of 2 or more. Data was presented as means (with standard deviation) unless otherwise stated and significance was expressed as p<0.01.


Optimal keratocyte seeding density

A range of cell concentrations (50,000-1,600,000 cells/ml) was initially investigated in order to determine a dose response for the relationship between cell density and the rate of matrix contraction. It was found that a concentration of 200,000 cells/ml of collagen matrix resulted in a matrix in which contraction could most readily be observed during the course of the experiments. Therefore, subsequent matrix contraction experiments were carried out using 200,000 cells/ml of collagen matrix.

Keratocyte derived fibroblasts within collagen matrices express TSP-2 and TSP-3

The integrity of the cDNA produced by first strand synthesis of the extracted RNA from the collagen matrices was confirmed by PCR for the house keeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH, Figure 1). RT-PCR revealed the presence of mRNA for TSP-2 and TSP-3 in the keratocyte populated collagen matrices (Figure 1). Cultured human corneal keratocyte derived fibroblasts exhibited immunoreactivity for TSP-2 (Figure 2A) and TSP-3 (data not shown) when seeded within a collagen matrix. The staining was abolished when the primary antibody was absorbed with purified antigen or replaced by pre-immune IgG (Figure 2B).

Effects of TSP antibodies on keratocyte-populated collagen matrix contraction

TSP-2 antibody at a concentration of 20 μg/ml in culture media, caused a significant inhibition of contraction of keratocyte derived fibroblast populated collagen matrices (Figure 3A) on days 1, 3, and 7 compared with the control group (p<0.01). TSP-2 antibody at a concentration of 2 μg/ml, caused a significant inhibition of contraction of keratocyte populated collagen matrices (Figure 3B) on days 1 and 3 only (p<0.01). By day 7 there was no significant difference in the contraction of the matrices treated with TSP-2 antibody at a concentration 2 μg/ml and control matrices. Conversely, there was no significant difference in contraction between the control group and TSP-3 antibody treated matrices irrespective of TSP-3 antibody concentration.

Keratocyte derived fibroblasts in collagen matrices treated with 20 μg/ml of TSP-2 antibody remained round compared to keratocytes in untreated or TSP-3 antibody treated matrices at day 3 (Figure 4A,C). By day 7, the keratocytes in 20 μg/ml of TSP-2 antibody treated matrices had extended small processes, but were less elongated/stellate than cells in the control and TSP-3 antibody treated matrices (Figure 4B,D). To assess whether the effects of 20 μg/ml TSP-2 antibody were reversible, matrices were incubated with medium containing 20 μg/ml TSP-2 antibody for 7 days. The medium was then removed and replaced with fresh medium containing either 20 μg/ml TSP-2 antibody or no antibody. Matrices treated after 7 days with the antibody exhibited persistent inhibition of contraction, whereas when the antibody was removed from the medium there was a significant increase in contraction (p<0.01, Figure 5).


We have shown that, in addition to TSP-1, keratocyte derived fibroblasts synthesize TSP-2 and -3 in the keratocyte-populated collagen matrix model of corneal stromal repair. Moreover, addition of an antibody specific to the N-terminal of TSP-2 retarded keratocyte-mediated matrix contraction, with greater inhibition of contraction occurring at the higher antibody concentration. In addition, the effect of the antibody to TSP-2 appeared to be reversible in that contraction of the TSP-2 antibody-treated preparations began to approach the level of control preparations over a period of a week, and removal of antibody from the medium at 7 days hastened this reversal.

The finding that keratocyte derived fibroblasts express TSP-2 and TSP-3 in the collagen matrix model is not surprising, given that the cells have the ability to express these glycoproteins both in monolayer culture and in damaged corneal stroma [12]. The appearance of these two thrombospondins in the keratocytes of damaged corneal stroma suggests that they play a role in the injury response of the cells. Indeed, it is established that thrombospondins are involved in dermal repair [17]. Therefore, we investigated the effects of antibodies specific to TSP-2 and TSP-3 in the model of corneal fibrosis. In this model, keratocytes appear fibroblastic and the activities of the cells are manifest by contraction of the matrix.

Whereas contraction of keratocyte-populated matrices was unaffected by the presence of an antibody specific to TSP-3, the antibody specific to TSP-2 inhibited matrix contraction. Additionally, the inhibitory effect of the TSP-2 antibody diminished during the course of the experiment. This "recovery" in matrix contraction has led us to speculate that ongoing synthesis of TSP-2 by the keratocytes, might in turn exhaust the antibody supply. Indeed, following washout of antibody at 7 days matrix contraction reversed at a significantly greater rate than matrices in which the antibody was replenished. The inhibition of matrix contraction also appeared to be related to an inhibition of the acquisition of a fibroblastic shape by the cells, which suggests that there is a link between keratocyte phenotype and their ability to contract a matrix. On the other hand, although we found that the antibody to TSP-3 had no effect on keratocyte morphology or keratocyte-dependent matrix contraction, it is possible that antibodies directed to other domains of the TSP-3 molecule might impact on keratocyte behaviour and we cannot discount this possibility.

The finding that an antibody to TSP-2 inhibits keratocyte-mediated matrix contraction may, at first, seem at variance with findings that dermal wound repair is accelerated in TSP2-null mice [10,18]. However, TSP2-null mice exhibit an abnormal collagen architecture [18] and fibroblasts from TSP2-null mice fail to contract collagen gels while exhibiting abnormal attachment to a variety of matrix proteins [11,18,19]. Current evidence suggests that the abnormalities in TSP-2 deficiency relate to the ability of TSP-2 to bind and inactivate some matrix metalloproteinases (MMPs), notably MMP2 [19]. Excessive amounts of MMP2 may prohibit the attachment of fibroblastic cells to the matrix (perhaps either by proteolysis of the receptors or the matrix itself [20]) and hence inhibit key adhesion-dependent cell-matrix interactions such as matrix contraction and organization. Such a mechanism could explain the more rounded appearance of keratocytes in TSP-2 antibody treated matrices.

If TSP-2 plays a key role in collagen fiber organization and cell-mediated matrix during repair, it may be involved in the abnormal collagen fiber assembly and tractional astigmatism that follows some corneal stroma insults. Indeed, we have observed upregulation of TSP-2 expression in damaged stroma compared to normal cornea. These observations suggest that TSP-2 might be open to therapeutic modulation. However, TSP-2 is anti-angiogenic and we have suggested that the protein (possibly with TSP-1) may help suppress vascularization in some corneal wounds. Therefore, any therapeutic scheme would have to balance the apparent anti-angiogenic and pro-contractile properties of TSP-2. One approach may be to target, or employ combinations of, specific TSP-2 peptide fragments, since it is well established that different fragments subserve diverse functions of the molecule [2,4]. Indeed, it may ultimately be possible to manipulate both corneal vascularization and fibrosis using a TSP-2 peptide fragment-based regime.


The support of Action Research and Sport Aiding medical Research for KidS (SPARKS) is gratefully acknowledged.


1. Bornstein P. Thrombospondins: structure and regulation of expression. FASEB J 1992; 6:3290-9.

2. Lawler J. The functions of thrombospondin-1 and-2. Curr Opin Cell Biol 2000; 12:634-40.

3. Chen H, Herndon ME, Lawler J. The cell biology of thrombospondin-1. Matrix Biol 2000; 19:597-614.

4. Adams JC, Tucker RP. The thrombospondin type 1 repeat (TSR) superfamily: diverse proteins with related roles in neuronal development. Dev Dyn 2000; 218:280-99.

5. Lawler J. The structural and functional properties of thrombospondin. Blood 1986; 67:1197-209.

6. Carron JA, Bowler WB, Wagstaff SC, Gallagher JA. Expression of members of the thrombospondin family by human skeletal tissues and cultured cells. Biochem Biophys Res Commun 1999; 263:389-91.

7. Newton G, Weremowicz S, Morton CC, Copeland NG, Gilbert DJ, Jenkins NA, Lawler J. Characterization of human and mouse cartilage oligomeric matrix protein. Genomics 1994; 24:435-9.

8. Arber S, Caroni P. Thrombospondin-4, an extracellular matrix protein expressed in the developing and adult nervous system promotes neurite outgrowth. J Cell Biol 1995; 131:1083-94.

9. Lawler J, Sunday M, Thibert V, Duquette M, George EL, Rayburn H, Hynes RO. Thrombospondin-1 is required for normal murine pulmonary homeostasis and its absence causes pneumonia. J Clin Invest 1998; 101:982-92.

10. Kyriakides TR, Tam JW, Bornstein P. Accelerated wound healing in mice with a disruption of the thrombospondin 2 gene. J Invest Dermatol 1999; 113:782-7.

11. Kyriakides TR, Leach KJ, Hoffman AS, Ratner BD, Bornstein P. Mice that lack the angiogenesis inhibitor, thrombospondin 2, mount an altered foreign body reaction characterized by increased vascularity. Proc Natl Acad Sci U S A 1999; 96:4449-54.

12. Armstrong DJ, Hiscott P, Batterbury M, Kaye S. Corneal stromal cells (keratocytes) express thrombospondins 2 and 3 in wound repair phenotype. Int J Biochem Cell Biol 2002; 34:588-93.

13. Hiscott P, Armstrong D, Batterbury M, Kaye S. Repair in avascular tissues: fibrosis in the transparent structures of the eye and thrombospondin 1. Histol Histopathol 1999; 14:1309-20.

14. Armstrong DJ, Hiscott P, Batterbury M, Kaye S. Expression and Distribution of thrombospondins in the normal and wounded cornea. Invest Ophthalmol Vis Sci 2001; 42:S901.

15. Mazure A, Grierson I. In vitro studies of the contractility of cell types involved in proliferative vitreoretinopathy. Invest Ophthalmol Vis Sci 1992; 33:3407-16.

16. Carron JA, Hiscott P, Hagan S, Sheridan CM, Magee R, Gallagher JA. Cultured human retinal pigment epithelial cells differentially express thrombospondin-1, -2, -3, and -4. Int J Biochem Cell Biol 2000; 32:1137-42.

17. Kyriakides TR, Zhu YH, Yang Z, Bornstein P. The distribution of the matricellular protein thrombospondin 2 in tissues of embryonic and adult mice. J Histochem Cytochem 1998; 46:1007-15.

18. Kyriakides TR, Zhu YH, Smith LT, Bain SD, Yang Z, Lin MT, Danielson KG, Iozzo RV, LaMarca M, McKinney CE, Ginns EI, Bornstein P. Mice that lack thrombospondin 2 display connective tissue abnormalities that are associated with disordered collagen fibrillogenesis, an increased vascular density, and a bleeding diathesis. J Cell Biol 1998; 140:419-30.

19. Yang Z, Kyriakides TR, Bornstein P. Matricellular proteins as modulators of cell-matrix interactions: adhesive defect in thrombospondin 2-null fibroblasts is a consequence of increased levels of matrix metalloproteinase-2. Mol Biol Cell 2000; 11:3353-64.

20. Adams JC, Kureishy N, Taylor AL. A role for syndecan-1 in coupling fascin spike formation by thrombospondin-1. J Cell Biol 2001; 152:1169-82.

Armstrong, Mol Vis 2003; 9:74-79 <>
©2003 Molecular Vision <>
ISSN 1090-0535