ABSTRACT
Endothelial cells are intimately involved in a wide range of biological processes including reproduction, development and wound healing (Folkman, 1992), as well as pathological processes such as inflammatory disorders of the skin and joints (Abbot et al., 1992), diabetic retinopathy and tumour invasion (Folkman, 1992). This has led to major efforts over the past twenty years to isolate and culture endothelial cells from both animal and human tissues, in order to investigate their role further. The most common isolation has been from the human umbilical vein, largely because of the ease with which the umbilici can be obtained. Isolation involves cannulation of the vein and introduction of a proteolytic enzyme, followed by a five to fifteen minute incubation before flushing to yield an isolate of endothelial cells (Jaffe et al., 1973). Alternatively, it is possible to obtain endothelial cells from large vessels, including arteries, by gentle scraping of the intimal surface with, for example, a cotton-wool tip.
INTRODUCTION
Endothelial cells are intimately involved in a wide range of biological processes including reproduction, development and wound healing (Folkman, 1992), as well as pathological processes such as inflammatory disorders of the skin and joints (Abbot et al., 1992), diabetic retinopathy and tumour invasion (Folkman, 1992). This has led to major efforts over the past twenty years to isolate and culture endothelial cells from both animal and human tissues, in order to investigate their role further. The most common isolation has been from the human umbilical vein, largely because of the ease with which the umbilici can be obtained. Isolation involves cannulation of the vein and introduction of a proteolytic enzyme, followed by a five to fifteen minute incubation before flushing to yield an isolate of endothelial cells (Jaffe et al., 1973). Alternatively, it is possible to obtain endothelial cells from large vessels, including arteries, by gentle scraping of the intimal surface with, for example, a cotton-wool tip.
The use of endothelial cells from large vessels, particularly those of foetal origin, is of limited value in the study of adult diseases involving an interaction with the microvasculature. Increasingly it is becoming apparent that there is marked endothelial heterogeneity, related in some way to the function of the organ in which the cells are found. For example, the endothelium of the cerebral microvasculature has tight junctions, an important component of the blood-brain barrier, while the hepatic endothelium has comparatively large gaps to permit the passage of larger solutes. Other heterogeneity involves surface antigen expression and growth factor responsiveness (Kuzu et al., 1992; McCarthy et al., 1991). Despite much research, the isolation and culture of microvascular endothelium remains a major challenge. Ideally, one would work on endothelial cells isolated from the tissue in the disease process that one wishes to study. The alternatives that remain are: to work on cells isolated from human tissue from another organ, or from the same organ but a different species, or to return to the human umbilical vein endothelial cells (HUVEC). The aim of this paper is not to resolve this dilemma but to bring together and evaluate the methods described in the literature for isolating endothelial cells from both human and animal tissues and to provide a practical guide for researchers planning to undertake the isolation and culture of such cells.
TECHNIQUES FOR ISOLATING CAPILLARY ENDOTHELIUM
The isolation of cells from the capillary endothelium poses significant challenges not met in the HUVEC preparation. As it is not possible to cannulate the microvasculature, most methods rely upon physical dissociation of the endothelial cells by either homogenising or mincing the tissue before incubating it with proteolytic enzymes for varying lengths of time to release the endothelial cells.
The enzyme preparations employed to digest tissue vary considerably, with trypsin and collagenase being most commonly used, and pronase and dispase occasionally. The concentrations and length of exposure depend upon the tissue in question; in the case of adult human skin, the tissue is allowed to incubate with 0.5% trypsin overnight because of the residual epidermal layer, which must be digested in order to release the underlying dermal capillary endothelium (Bull et al., 1990). By contrast, more delicate tissues such as the corpus luteum (Bagavandoss and Wilks, 1991), the adrenal (Fawcett et al., 1991) or the synovial endothelium (Abbot et al., 1992) are substantially dissociated within as little as thirty minutes.
Unlike the HUVEC preparation where the enzyme is only in contact with the cells to be isolated, using these methodologies many different cells are exposed to the enzyme, inevitably yielding an impure isolate. The most common contaminants are fibroblasts, which are more vigorous and grow more rapidly in culture than most endothelial cells, rapidly taking over. To minimise this, the isolates are taken through a series of purification steps to permit the endothelial cells to establish themselves.
TECHNIQUES OF PURIFICATION
Early methods relied upon the identification of cobblestone endothelial colonies after they had been seeded and allowed to adhere. As well as the endothelial cells, however, contaminants may adhere, necessitating lengthy and time-consuming manual purification techniques, which must be repeated at regular intervals to ensure the endothelial cells are not overrun. In addition, it requires morphological identification of the endothelial cells early in culture, which can be difficult, especially where there are large numbers of contaminating cells. Another difficulty is the propensity of endothelial cells to change their morphology and behaviour depending upon the growth supplements in the media or the matrix onto which the cells are seeded. These reversible changes are discussed further in the section on growth requirements of the cell cultures.
Manual techniques include removal of contaminant colonies using a Pasteur pipette or cell scraper (Chung-Welch et al., 1988; Abbot et al., 1992) or killing by diathermy (Marks and Penny, 1986; Jackson et al., 1989). Although it has been used, diathermy is not usually available in most laboratories and is tedious. Even though these methods are time-consuming, laborious even, they should not be dismissed as they provide a useful means of removing smaller numbers of contaminants inevitably found in the culture. This can be facilitated by seeding the cells in a small area in the centre of the dish rather than across the whole surface area (Marks et al., 1985). Thus, if >50% of the colonies on a plate are endothelial, marking areas on the bottom of the plate where contaminating cells are found, and subsequently scraping off the non-endothelial colonies, can give rise to pure endothelial cultures.
Other procedures seek to remove the endothelial cells from their competitive environment, rather than to eliminate the contaminants. This is more successful than trying to remove contaminants manually when they prevail over endothelial cells. Once identified, the endothelial cells are removed by various techniques such as release with trypsin within cloning rings (Abbot et al., 1992; Bagavandoss et al., 1991; Fawcett et al., 1991) or with glass beads within cylinders (Diglio et al., 1982; Folkman, 1982). In the latter case the endothelial cells migrate from the beads onto the new surface. The use of magnetic beads is a refinement of this technique, facilitating the sorting of the cell-bearing beads and is described by Abbot et al. (1992). Briefly, tosyl-activated polystyrene beads are coated with, for example, Ulex europaeus lectin type 1, which recognizes endothelial cells. The beads are then added to the culture at the ratio of approximately three beads per cell (this ratio is critical, see Abbot, 1992) for ten minutes to allow binding before the cell-bearing beads are removed using a magnet. These beads are then washed with fucose (which displaces Ulex lectin) to release the endothelial cells. These methods can also be used to clone pure cultures of endothelial cells and avoids repeated exposure to trypsinization, which Bull et al. (1990) and Ryan (1984) claim affects their morphology, damages their growth rate and their receptor expression, and shortens the lifespan of the cells in culture. This has not been the experience in the authors’ laboratory, however, where endothelial cells are routinely removed with trypsin, without apparent deleterious effects on their life-span or properties.
Other purification techniques aim to minimise the numbers of contaminating cells from the digest actually being seeded in the first place. These include the use of simple nylon or mesh filters during the isolation procedure (Abbot et al., 1992; Detmar et al., 1992; Diglio et al., 1982; Marks et al., 1985; Schor and Schor, 1986). The pore size either permits the passage of the endothelial cells while retaining the contaminants or, alternatively, retains the endothelial cells, which are released by backwashing the filter. More complex methodologies include the use of fluorescence activated cell sorting (FACS) after labelling of the cells (Voyta et al., 1984), or the use of Percoll or dextran gradient centrifugation.
FACS sorting has been applied successfully to isolate endothelial cells, most noticeably after labelling with diI-Ac-LDL (Voyta et al., 1984). Acetyl-LDL is taken up only by endothelial cells and macrophages by the LDL scavenger receptor, and the latter do not proliferate in culture. In the authors’ experience, this is generally only successful if the endothelial cells comprise >30% of the cell population that is sorted. If they compromise less than 15% of the total population, the sorting of viable cultures has not proven possible. A potential way around this is to carry out a density gradient enrichment first and then to perform a FACS sort after the cells have recovered, say 24 hours later.
Percoll or dextran gradient centrifugation is in essence very simple and relies upon the different weights of the cells within the isolates to carry them to different points on the gradient on centrifugation. The endothelial fraction appears in a distinct region within the gradient, as demarcated by marker beads, and can be removed leaving the majority of the contaminants behind. A good description of this method is provided by either Marks et al. (1985) or Detmar et al. (1992). This technique is very useful where there is likely to be a significant degree of contamination. While not eliminating the manual weeding stage altogether, it clearly reduces the number of times this has to be performed (Marks et al., 1985). Another possibility is to use serum-free medium supplemented with vascular endothelial growth factor (VEGF), which to date is the only known growth factor that is specific for endothelial cells (Ferrara et al., 1991). This may stimulate the outgrowth of endothelial cell populations from mixed cultures. Until now, such an approach has been prohibitively expensive because of the lack of availability of VEGF.
A final technique that has been used with marked success is elutriation (see, for example, Irving et al., 1984). (Elutriation is separation by density; thus, cells are spun down (separation by mass) against a counterflow (separation by size).) This has the advantage of providing substantial quantities of pure cells but is limited by the availability of an elutriator.
GROWTH CONDITIONS
After obtaining a pure culture, the main problem is then to stimulate cell growth, which requires particular conditions according to the endothelial cell type. As can be seen in Table 1, the medium that has been employed varies considerably. Most isolates require mitogens, including endothelial cell growth supplement (essentially a crude brain extract) or tumour-cell conditioned medium, and show maximal growth when heparin is also added (Folkman et al., 1979; Marks et al., 1985). The medium is often also supplemented with whole human serum, with optimal growth at a serum concentration as high as 50% (Davison et al., 1980; Marks et al., 1985), although less is required when maternal pre-partum serum is used (Karasek, 1989).
The requirement for human serum in the isolation of human microvascular endothelium is a recurrent theme. Plasma-poor serum does not support growth, perhaps reflecting the loss of calcium and citrate during dialysis (Marks et al., 1985).
One of the requirements for the adherence of human microvascular endothelial cells is that the plate is coated with a substance to provide an artificial extracellular surface onto which the cells can adhere. This is usually gelatin or fibronectin, the latter being favoured because contaminants bind less readily. However, endothelial cells also have a decreased affinity for fibronectin, often yielding a poor culture. Karasek (1989) has claimed that fibronectin is merely more expensive, without being more effective.
CHARACTERISATION OF ENDOTHELIAL CELLS IN CULTURE
A final but crucial step is a full characterisation of the cells as endothelial. Traditional markers, such as the von Wille-brand factor (factor VIII-related antigen) have been shown to be absent, or to be expressed at a very low level in certain capillary endothelial cells (Kuzu et al., 1992; see also McCarthy et al., 1991). It is postulated that the presence of Weibel-Palade bodies (of which vWF is a component) decreases with increasing distance from the heart and can be influenced by constituents in the medium (Imcke et al., 1991). CD 31 is an excellent pan-endothelial marker but, while retained by HUVEC and lung endothelial cells in culture, it is lost by many others. The best means of characterising endothelial cultures is to examine a series of properties and then to make an assessment. Characteristic (although not definitive) endothelial markers include factor VIII-related antigen, CD 31, CD 34, angiotensin converting enzyme (ACE), strong uptake of diI-Ac-LDL, prosta-cyclin secretion, staining with Ulex europaeus lectin type 1 (for human cells), Bandiera simplicifolia lectin (for rodent cells) and morphology; for example, the ‘cobblestone’ appearance at confluence and the formation of ‘tubes’ in Matrigel. These features are not universal or necessarily consistent within cultures. For example, the morphology of endothelial cells has been shown to change according to the matrix (Marks et al., 1985) or growth supplements (Karasek, 1989; McCarthy et al., 1991) added to the medium.
It is imperative that the cells be proven to be endothelial before using them in studies. An example where this was not the case was in the human omental ‘endothelial’ cells (Kern et al., 1983). These were shown to be positive for vWF, ACE and diI-Ac-LDL, and had a cobblestone appearance at confluence. They were studied for some years before it became apparent that they were in fact mesothelial cells, on the basis of the presence of cytokeratins 8 and 18, and of vimentin (van Hinsbergh et al., 1990; Potzsch et al., 1990; Visser et al., 1991; Latron et al., 1991; Chung-Welch et al., 1989). Despite this certain groups appear oblivious to the fact that these cells are not microvascular endothelial cells and they are still being described as a unique endothelium in that, for example, they express receptors for PDGF (Beitz et al., 1991). As noted above, no single endothelial characteristic is sufficient to confirm a culture as endothelial, and it may be worthwhile testing the cells to con-firm that they are not of a different origin; for example, Diglio et al. (1982) demonstrated the absence of smooth muscle antigens in their endothelial cell cultures.
TISSUES FROM WHICH THE ENDOTHELIUM HAS BEEN ISOLATED
Some of the human tissues for which techniques have been described are foetal or neonatal in origin; for example, neonatal foreskin (Imcke et al., 1991; Davison et al., 1980; Ruzsczak et al., 1990) and decidua (Gallery et al., 1991; Lindenbaum et al., 1991). The neonatal stage represents the time when human cells are undergoing the transition from foetal to adult status. These cells retain significant foetal characteristics, such as their ability to differentiate, and it has been noted that they require less complex culture conditions than do adult dermal endothelial cells (Davison et al., 1983). The cytogenetic differences between endothelial cells of adult and foetal origin have been described by Nichols et al. (1987). Bull et al. (1990) described a method for isolating adult skin endothelium, that offers the advantages of obtaining a more representative example of endothelium involved in disease processes and of obtaining the tissue more easily. Two methods are described for isolating the endothelial cells from synovium (Abbot et al., 1992; Jackson et al., 1989), offering the opportunity to investigate inflammatory joint disorders including the angiogenic mechanisms underlying pannus formation in rheumatoid arthritis.
CONCLUSION
The purpose of this Commentary has been to outline the methods currently used to obtain pure cultures of microvascular endothelium from both animal and human tissue. A major advance that has become apparent is the use of the density gradient to purify the cultures at an early stage, making the task of weeding out the remaining contaminants much easier and less time-consuming. The improvement in method for the isolation of microvascular endothelium means that this is now potentially feasible for any tissue. The major outstanding problem is the need to determine culture conditions that permit in vitro proliferation of microvascular endothelium from many tissues. The VEGF growth factor family may be useful here. Endothelial cell growth supplement (ECGS), which is routinely added to microvascular endothelial cultures, is prepared from brain tissue, rich in members of the fibroblast growth factor family but lacks VEGF (Plate et al., 1992). For those endothelial isolates that do not proliferate in ECGS, for example, that from decidua, addition of VEGF may be of use.
