Inflammatory cytokines induce the transformation of human dermal microvascular endothelial cells into myofibroblasts: a potential role in skin fibrogenesis


Marvin A. Karasek, PhD, Department of
Dermatology, Program in Epithelial Biology, Stanford
University School of Medicine, Stanford, CA 94305,
Tel: +1 650 723 7317
Fax: +1 650 723 8762


Background:  The myofibroblast plays a central role in wound contraction and in the pathology of fibrosis. The origin(s) of this important cell type in skin has not been firmly established.

Methods:  Human epithelioid dermal microvascular endothelial cells (HDMEC) were isolated from foreskin tissue and maintained in cell culture. The transformation of epithelioid HDMEC into myofibroblasts (EMT) was induced by the inflammatory cytokines interleukin-1β (IL-1β) or tumour necrosis factor-α (TNF-α), and the transformed cells were characterized by electron microscopy, immunohistochemistry and quantitative RT-PCR.

Results:  After short-term exposure to IL-1β or TNF-α (<3 days), EMT was reversible; after long-term exposure (>10 days), EMT was permanent. The transformed cells were identified as myofibroblasts by cytoplasmic microfilaments with dense bodies and attachment plaques, by the expression of α-smooth muscle actin, type I collagen and calponin, and by quantitative RT-PCR gene expression of type I collagen and α-smooth muscle actin.

Conclusions:  Long-term exposure to TNF-α or IL-1β induced the permanent transformation of HDMEC into myofibroblasts in cell culture. A similar transformation following chronic inflammatory stimulation in vivo may explain one source of myofibroblasts in skin fibrogenesis.

The normal microvasculature is lined by a continuous squamous epithelioid endothelium that regulates exchanges between the blood and the surrounding tissues. When injured or exposed to inflammatory or angiogenic mediators, epithelioid dermal microvascular endothelial cells (HDMEC) convert reversibly from an epithelioid to more spindle and inflammatory-like morphology.1,2 While several of the second messengers that control these changes in vitro have been described,1,3 the positive identification of the cell type(s) resulting from long-term chronic exposure of HDMEC to inflammatory cytokines up-regulated in fibrosis has not been reported.

In the embryo, the early stages of angiogenesis involve the transition of endothelial precursor cells to a spindle-shaped morphology.4 Similar precapillary cords have also been observed in the adult.5 Embryonic quail endothelial cells have been reported to transdifferentiate into mesenchymal cells expressing smooth muscle actin both in vivo and in vitro.6 Human umbilical vein endothelial cells, when deprived of fibroblast growth factor, differentiate into smooth muscle-like cells;7 in postembryonic aortic and kidney endothelial cells, Frid et al.8 and Sommer et al.9 have identified myofibroblasts as one of the cell types generated from the endothelium by transdifferentiation.

To maintain an epithelioid phenotype, a balance between the activity of protein kinase A (PKA) and protein kinase C (PKC) is required.1,3 Cyclic AMP (cAMP) or metabolic reactions that increase cAMP support an epithelioid morphology while a decrease in cAMP or agents that increase calcium and activate PKC causes a disruption of tight junctions and promotes the conversion of epithelioid cells to a spindle-shaped morphology. Activation of Protein Kinase Cα has been shown recently to be the specific PKC isozyme central to angiogenesis, wound closure and permeability,10 and each of these functions involves a change in HDMEC structure.

In this study, we describe the phenotypic changes that take place in HDMEC following exposure to the inflammatory cytokines interleukin-1β (IL-1β) and tumour necrosis factor-α (TNF-α), and we identify myofibroblasts as one of the cell types resulting from this transformation by ultrastrucutural, immunohistochemical, and genetic markers. A potential role for inflammation-induced EMT in the pathophysiology of skin fibrosis is presented.

Materials and methods


Dispase was purchased from Collaborative Research (Bedford, MA, USA); Trypsin-EDTA, Iscove's growth medium, and DMEM from Gibco BRL (Grand Island, NY, USA); IL-1β from BD Biosciences (Bedford, MA, USA); Dynabeads from Dynal Biotech Inc. (Lake Success, NY, USA); and EGM from Clonetech (Palo Alto, MD, USA). All primary antibodies were purchased from Santa Cruz (Santa Cruz Biotech, Santa Cruz, CA, USA). Alexaflour, fluorescent-labeled (488/594) secondary antibodies were purchased from Molecular probes (Eugene, OR, USA). The following reagents were purchased from Sigma (St. Louis, MO, USA): TNF-α; TNF-β; Lucifer Yellow CH dipotassium salt; and Ulex europaeus agglutinin 1.

Isolation of skin microvascular endothelial cells

The isolation of HDMEC and the preparation of growth medium and the maintenance of cultures were carried out as described previously.11 All studies involving human tissue were approved by the Human Subjects Committee of Stanford University School of Medicine, Stanford, CA and were in accordance with the Declaration of Helsinki Principles. After the first passage, cells were purified with Ulex europaeus agglutinin 1 coated on magnetic Dynabeads.12 Purity of the cell population was confirmed by positive staining for von Willebrand factor (vWf) and platelet endothelial cell adhesion molecule-1 (PECAM-1). All experimental studies were performed with Ulex-purified passage 3 or passage 4 HDMEC with the exception of the outgrowth from explant cultures carried out with primary cultures. All growth studies of HDMEC were carried out at 37°C in a 20% oxygen-5% CO2-humidified incubator.

Explant cultures

To prepare explant cultures of partially intact vascular lumens, the timed release of HDMEC from the vasculature by dispase as described above is decreased to 6–10 h. Under these conditions, fragments of intact endothelium rather than individual HDMEC are released into the medium, and these fragments are identified by phase contrast microscopy. Following neutralization of dispase activity, the fragments settle quickly on plating and attach firmly to the plastic surface. Migration of cells from the endothelium begins within 3 h of plating and is monitored by phase contrast microscopy.


Microvascular endothelial cells were used at passage 3 or 4. Cells were plated into 12-well 16 mm plates with 1 ml of growth medium per well. TNF-α (50 ng/ml), TNF-β (50 ng/ml), or IL-1β (10 ng/ml) was added when cell monolayer reached 100% confluency. After specified time point, cells were washed once with cold PBS fixed with 100% ethanol for 5 min and stained with 1% crystal violet.

Electron microscopy

For transmission electron microscopy (TEM) analysis of IL-1β-activated cells and untreated, HDMEC were cultured on 8-mm glass cover slips. When the cell monolayer reached confluence, it was washed with PBS, immersed in 2% glutaraldehyde, and washed twice in PBS and postfixed with 1% osmium tetroxide (Polysciences, Inc., Warrington, PA, USA) in PBS for 1 h. After two 10-min washes in double-distilled water, specimens were stained in 0.25% uranyl acetate (Polysciences, Inc.) overnight. After 24 h, the specimens were washed with water and dehydrated through a graded series of alcohol and propylene oxide washes. Each sample was infiltrated sequentially with 2:1 and 1:1 propylene oxide-EPON (Resolution Performance Products, Houston, TX, USA) for 4 h, incubated overnight with 100% EPON, transferred to fresh EPON, and embedded and polymerized at 60°C for 24 h. Thin sections were collected on copper grids, stained with uranyl acetate and lead citrate, and viewed using a Phillips CM-12 transmission electron microscope.


For immunostaining studies, HDMEC (passage 3) were grown in 8-well glass chamber slides (Fischer Scientific, NH, USA), washed with PBS, fixed with 4% paraformaldehyde, and stained for 2 h at room temperature with antibodies against vWf, smooth muscle actin, type I collagen, type IV collagen, calponin, and α-smooth muscle actin at the dilutions recommended by the manufacturer. Fluorescent-labeled secondary antibody Alexaflour 488/594 was used to detect primary antibody binding. Fluorescent images were recorded with a Zeiss Axiovert 100m microscope.

Quantitative RT-PCR

HDMEC (passage 3) were grown to confluence in 60-mm culture dishes and RNA extracted using the Qiagen RNA extraction kit (Qiagen, Valencia, CA, USA). Comparative quantitative RT-PCR was carried out according to the manufacturer's instructions using the ‘Brilliant SYBR green qRT-PCR’ kit from Stratagene on a MX3000 real time PCR machine (Stratagene, La Jolla, CA, USA) on four genes using GAPDH as the normalizing gene to correct baseline values for all samples. Samples from each time point were compared with the 0-h control (referred to as calibrator sample for real time analysis). The software used computed the mean difference and the standard deviation for three independent triplicate wells. The primer sequences for each gene are S-GAPDH-QRT: GAGTCAACGGATTTGGTCGT; A-GAPDH-QRT: TTGATTTTGGAGGGATCTC; S-Smooth Muscle Actin-QRT: TTCAATGTCCCAGCCATGTA; A-Smooth Muscle Actin-QRT: GAAGGAATAGCCACGCTCAG; S-Type 1 collagen-QRT: GTGCTAAAGGTGCCAATGGT; A-Type 1 collagen-QRT: CTCCTCGCTTTCCTTCCTCT; S-Type 4 collagen-QRT:CCTCCAGGAGTACCAGGACA; A-Type 4 collagen-QRT: CTTTTTCCCCTTTGTCACCA; S-Factor8-QRT: ATGATTCCTGCCAGATTTGC; A-Factor8-QRT: AGACTCTTTGGTCCCCCTGT.


Endothelial cells convert from an epithelioid to a mesenchymal-like morphology in explant cultures of intact microvessel

Figure 1 illustrates the normal phenotypic changes that occur during the initial stages in outgrowth of endothelial cells from an explant culture of a partially intact endothelium both by phase contrast microscopy (Fig. 1A) and after immunofluorescent staining with vWf (Fig. 1B). As shown in this figure, the endothelial cells which migrate from the explant convert from an epithelioid morphology (as seen in the open and partially intact lumen) to spindle-shaped, mesenchymal-like cells that continue to stain with vWf, though less intensively. After 24 h, all spindle-shaped cells revert back to an epithelioid morphology similar to that observed in the intact endothelium and endothelial cells isolated from microvessel by more extensive enzymatic release by dispase (Fig. 2A).

Figure 1.

Outgrowth of endothelial cells from blood vessel explants. A) Phase contrast micrograph of outgrowth of normal human dermal microvascular endothelial cells (HDMEC) from explants of the lining of a normal vessel demonstrating the spindle-shaped morphology of HDMEC in the first stages of angiogenesis. B) Immunofluorescent staining with von Willebrand factor of the same explant.

Figure 2.

Phase contrast micrographs on IL-1β and TNF-α TNF-β-activated human dermal microvascular endothelial cells at 24 h. A) dimethyl sulfoxide (DMSO) (control), B) IL-1β, C) TNF-α, D) TNF-β 24 h after activation. Photo-micrographs are representative areas of three repeat experiments. Bar: 250 µm. TNF-α, tumour necrosis factor-α.

IL-1β and TNF-α induce EMT

Figure 2 compares the changes in morphology of HDMEC induced by IL-1β (Fig. 2B, TNF-α (Fig. 2C), and TNF-β (Fig. 2D). As seen in this figure, both IL-1β and TNF-α induce EMT. In contrast, TNF-β has no effect on EMT. While both IL-1β and TNF-α induce EMT, the detailed morphology and growth characteristics of the transformed cells are similar but not identical.

Ultrastructural analysis of HDMEC following chronic exposure to IL-1β shows fiber characteristic of myofibroblasts

Figure 3 compares the ultra structure of both control and spindle-shaped cells following chronic exposure (20 days) to IL-1β. In vivo, myofibroblasts are connected to the extra cellular matrix by cell-to-stroma attachment sites trough trans membrane complexes of intracellular microfilaments (fibronexus).13

Figure 3.

Ultrastructural changes in human dermal microvascular endothelial cells (HDMEC) following IL-1β activation. Electron micrographs of A) control HDMEC monolayer (10K magnification) and B) exposure to IL-1β for 3 weeks (8K magnification). Note the appearance of large bundles of fibrils with the attachment plaques characteristic of myofibroblasts marked by arrowhead.

Within the cytoplasm, myofibroblasts contain bundles of microfilaments usually arranged parallel to the long axis of the cell. Dense bodies are interspersed among these fibers. As shown in Fig. 3B, filaments consistent with those observed in myofibroblasts in vivo are present in the spindle-shaped cells and absent in the epithelioid configuration (Fig. 3A).

Immunofluorescent expression of proteins associated with myofibroblasts

Figure 4 compares the immunohistochemical profile of epithelioid and spindle-shaped HDMEC following long-term exposure to IL-1β. As shown in this figure, epithelioid HDMEC show the characteristic staining of vWf located in a pattern consistent with that of Weible-Palade bodies (Fig. 4A). Type IV collagen is present both within the cells and deposited extracellularly (Fig. 4B). Significant expression of calponin, type 1 collagen, or α-SMA was not detected in epithelioid HDMEC. In contrast, permanently transformed spindle-shaped HDMEC show a marked up-regulation in the synthesis of calponin (Fig. 4C), type 1 collagen (Fig. 4D), and α-SMA (Fig. 4E) and do not synthesize vWf or type IV collagen. Negative or negligible stains for control and IL-1β-activated HDMEC are summarized in Table 1. The protein profile for IL-1β-activated HDMEC is consistent with a pattern observed in myofibroblasts both in vivo and in vitro.

Figure 4.

Immunofluorescence of control and IL-1β-activated human dermal microvascular endothelial cells (HDMEC) following long-term exposure to IL-1β. HDMEC were activated continuously with IL-1β for 20 days (medium changed at 3-day time intervals). A) von Willebrand factor (vWf) (control HDMEC); B) type IV collagen (control HDMEC); C) calponin (IL-1β-treated HDMEC; D) type I collagen (IL-1β-treated HDMEC); E) α-SMA (IL-1β-treated HDMEC). Negligible or negative expression for vWf and type IV collagen was observed in IL-1β-activated HDMEC and for calponin, type I collagen, and α-SMA in control HDMEC. The staining patterns for all proteins are compared in Table 1. Bar: 250 µm.

Table 1.  Comparison of expression profile of proteins associated with myofibroblasts
Marker for EMTHDMECHDMEC activated with IL-1β for 20 days
  1. +++, highly expressed; –/+, negligible expression; –, absent expression.

von Willebrand factor+++
Type IV collagen+++
α-smooth muscle actin+++
Type I collagen–/++++

Comparison of the relative expression of vWf, type IV collagen, type 1 collagen and α-SMA genes following chronic exposure to IL-1β

Figure 5 illustrates the changes in relative gene expression following long-term exposure to IL-1β. In the first 6 h of exposure to IL-1β, there is a marked increase in both vWf (Fig. 5A) and in type IV (Fig. 5B) collagen gene expression followed by a major decrease at all subsequent time intervals. In contrast, no stimulation of type I collagen or α-SMA genes can be detected during the acute phase of exposure to IL-1β. After long-term exposure (>10 days), a major increase in the relative gene expression of both α-SMA (Fig. 5C) and type I collagen (Fig. 5D) is observed. This pattern of gene expressing is consistent with the protein expression patterns described in Fig. 4.

Figure 5.

Relative quantitative RT-PCR analysis of von Willebrand factor (vWf), type IV collagen and α-smooth muscle actin regulated during short-term and long-term exposure to IL-1β. Relative quantitative graphs comparing mRNA synthesis of GAPDH (normalizing gene) vs. A) vWf; B) type IV collagen; C) α-smooth muscle actin; and D) type I collagen, following activation with IL-1β for increasing time periods and fresh IL-1β added every 48 h. All time points are compared with zero time periods corresponding with time of adding IL-1β as depicted by arrow.


The transition of an epithelioid microvascular endothelial cell to a more spindle-shaped morphology is a normal and a reversible physiologic response of the vasculature in vivo to permit both the egress of leucocytes and the migration of endothelial cells from the vasculature in inflammation and during the first stages of angiogenesis. These transitions are clearly demonstrated in vitro in explant cultures of the vascular lumen where the migration of spindle-shaped cells from the endothelium occurs (Fig. 1) and after the exposure to two inflammatory cytokines (IL-1β and TNF-α) in dispersed endothelial cell populations (Fig. 2). That the spindle-shaped cells originate from the endothelium and not from surrounding pericytes or smooth muscle cells is shown by the retention of staining for vWf in the spindle-shaped cells and that all of the spindle-shaped cells revert to a classical epithelioid morphology within 24 h after migration. In epithelioid HDMEC, the actin cytoskeleton is arranged in dense peripheral bundles surrounding the perimeter of the cell surface in contrast to the rearrangement of actin filaments that transverse the long axis of the cell in parallel linear arrays in spindle-shaped cells.1 In the parallel actin filament orientation, cell migration from the vasculature into the surrounding medium or connective tissue can take place. In a completely dispersed HDMEC culture, a similar transition from an epithelioid to a spindle-shaped configuration can also be observed at the edge of expanding colonies and at the wound edge in confluent cultures that are disrupted by a scratch (unpublished observations). In each of these conditions, normal physiologic events are simulated, EMT is reversible and homeostasis is re-established.14 In earlier studies of the sequential changes with time in the expression of vWf and PECAM in HDMEC following IL-1β activation, vWF and PECAM immunostaining became weaker gradually at 3- and 6-day time intervals, and each was absent in most of the cells by day 10.13 As long as vWf and PECAM continued to be expressed in HDMEC, the morphologic change observed at the early time intervals was reversible.

In HDMEC, both IL-1β and TNF-α are able to induce morphologic changes similar to those observed in explant cultures. While each cytokine induces EMT, the morphologic changes are not identical as shown in Fig. 2. These differences may result from the ability of TNF-α to amplify the release of IL-1β from endothelial cells15 or by increasing the activation of Protein Kinase Cα.10 Although TNF-α and TNF-β are 50% homologous, TNF-β as shown in Fig. 2 was not able to induce EMT. These differences are similar to those described in umbilical vein endothelial cells in vitro where a divergent effect on hematopoietic growth factor production and a neutrophil adhesion molecule expression by TNF-α and TNF-β were described16 or in their capacities to generate IL-1β release form human umbilical vein endothelial cells.17

The lack of contact inhibition, the increase in cell size, and the ragged appearance of transformed HDMEC are morphologic characteristics typical of myofibroblasts in vitro. The cytoskeletal features of myofibroblasts in vivo also include intracytoplasmic bundles of microfilaments with dense bodies and plasmalemmal attachment plaques, As shown in Fig. 3, HDMEC permanently transformed with IL-1β have the ultrastructural hallmarks that identify myofibroblasts (microfilaments with dense bodies and plasmalemmal attachment plaques). In addition to cells displaying the major features of myofibroblasts, spindle-shaped cells without these typical ultrastructural characteristic and more typical of classical fibroblasts were also observed but at a lower frequency (results not shown). The origin of these cells has not been identified. At the immuno-histochemical level, α-SMA is one of the most reliable markers of myofibroblasts.18,19 As shown in Fig. 4C transformed HDMEC are strongly positive for α-SMA that are associated with the filaments.

The regulation of genes associated with inflammation is controlled by molecular signals that determine when and how often given genes are transcribed.20 Changes in the level of induced mRNA and encoded protein in the activated tissue have been equated to chemical equations where there is a lag in the time course of mRNA accumulation represented as sigmoidal curve of induced mRNA kinetics.21 We see two different phenomena of EMT during acute and chronic exposure to inflammatory cytokines. In the absence of further stimuli, gene expression tends to return back to steady state levels and is reversible. As depicted in Fig. 5, only long-term exposure to cytokines results in permanent transformation of HDMEC to myofibroblasts. Recent findings in tubular epithelial cells have also shown EMT to be a reversible process.22

Lastly, the transformation of HDMEC into myofibroblasts in vitro following chronic long-term exposure to inflammatory cytokines raises an important question of the potential role of the vasculature in skin diseases where inflammation is associated with the development of fibrosis. In studies of fibrosis in experimental animals, both TNF-α and IL-1β have been shown to induce fibrosis. In the rat peritoneum, overexpression of IL-1β or TNF-α led to an acute inflammatory response, increased expression of VEGF and TGF-β, and an increase in the vasculature, submesothelial thickening, and fibrosis.23 These reactions in animals simulate those observed in patients on long-term peritoneal dialysis. Direct injection of TNF-α into the mouse footpad also produced a mononuclear infiltrate and focal fibrosis.24 On the basis of the in vitro results of this study, an early (<3 days) anti-inflammatory treatment could be considered as a possible approach to prevent or minimize the development of fibrosis.

In studies of human systemic sclerosis, one of the earliest pathologic changes that can be detected is in the vasculature of clinically normal appearing skin. The early pathology is characterized by the appearance of an inflammatory cell infiltrate into the papillary and mid-dermis and platelet aggregation within vessels.25 Dermal fibrosis, loss of adenxae and vascularity follows. An attractive hypothesis is that immunologic injury to the vasculature and an associated inflammation may be the missing link and the triggering factor between the endothelial cell effacement and the presence of myofibroblasts. As early as 1989, Beranek26 was one of the first investigators to suggest and present evidence that the microvasculature may the cell type responsible for the fibroblasts observed in the formation of granulation tissue.

Excessive scarring caused by an increase and persistence of myofibroblasts frequently leads to both disfigurement and impairment of the affected organ. The need to prevent this process from occurring is clearly of a high medical priority. The second messengers and the enzymes that play a major role in maintaining vascular homeostasis have been reported.1,3,10,27 Although the extent to which the endothelium may contribute to matrix protein synthesis has yet to be fully investigated in vivo, the studies of this investigation provide a new paradigm which may begin to explain the increased accumulation of matrix proteins and myofibroblasts pathognomonic of skin fibrogenesis.


This was supported by Virginia and Karl Ludwig Cancer Foundation.