INTRODUCTION

Top Abstract Introduction References

THREE-DIMENSIONAL (3-D) cell cultures have been widely used in biomedical research since the early decades of this century. Although preceded by work of other scientists, Holtfreter (62) and later Moscona (104-107) pioneered the field by their systematic research on morphogenesis using spherical reaggregated cultures of embryonic or malignant cells. Numerous subsequent studies were based on these investigations with regard to techniques and strategies for understanding organogenesis or expression of malignancy. The spectrum of research on cell aggregates was enlarged considerably by the fundamental studies of Sutherland and associates (64, 145-147), who inaugurated multicellular tumor spheroids as an in vitro model for systematic studies on tumor cell response to therapy. As a consequence, therapeutically oriented studies became the major domain of research with cell spheroids, although such investigations also triggered a number of studies on basic biological mechanisms, such as the regulation of proliferation, differentiation, cell death, invasion, angiogenesis, or immune response (for reviews see Refs. 9, 74, 108, 144).

At present, research on 3-D cell systems is more vital and productive than ever. The potential of 3-D cell cultures is currently being exploited in so many areas of biomedical research that it is impossible to review all aspects of these studies completely. Nevertheless, one reason for the recent progress in research on multicell systems may be the increasing interaction between researchers working in different fields of biomedical science and using similar 3-D culture techniques. Over the past 5 years, there has been an increasing number of publications with molecular and cellular biologists, cellular and neurophysiologists, as well as clinical scientists as joint authors, and work on 3-D cell systems has been published in scientific journals of outstanding ranking. Such a research effort mirrors the common need for improved and more refined in vitro models as a link between cell-free systems or single cells and organs or whole organisms in vivo.

One major advantage of 3-D cell cultures is their well-defined geometry, which makes it possible to directly relate structure to function and which enables theoretical analyses, e.g., of diffusion fields. Consequently, the most promising data on these cultures may be obtained with techniques allowing for spatial resolution. Combining such approaches with molecular analysis has clearly demonstrated that, in comparison with conventional cultures, cells in 3-D cultures more closely resemble the in vivo situation with regard to cell shape and cellular environment, and shape and environment can determine gene expression and the biological behavior of the cells. One impressive example demonstrating the significance of the environment is the finding that ectopic implantation of embryonic cells can transform them to malignancy and gives rise to cancer, whereas the same cells lead to normal embryogenesis in the uterus; conversely, teratocarcinoma cells may undergo normal development when implanted into an embryo (97). From a critical point of view, it should be kept in mind that the complexity of 3-D cell systems is not only an advantage but also a limitation. There will always be a number of questions that can only be answered by investigations using single cells or cell-free systems. At the same time, 3-D cultures cannot completely replace the testing of biological mechanisms for their relevance in vivo, e.g., in knockout animals.

The present review is based on a selection of publications, mainly from the past 5 years, dealing with 3-D cell cultures. Besides "classical" topics of research on multicellular tumor spheroids, new developments and efforts in experimental tissue modeling and in the generation of artificial organs are considered. Additionally, recent progress in organogenesis using embryoid bodies is discussed.

	EXPERIMENTAL THERAPEUTICS

After intensive research on radiation effects in multicellular tumor spheroids in the 1970s and 1980s, the number of systematic studies in this field has decreased considerably during the first half of the present decade. A detailed and systematic report has been issued by Stuschke and colleagues (143), who identified the degree of differentiation as an important determinant of radioresponsiveness in various spheroid types of human origin. Thus spheroids mirror the radiosensitivity of differentiating tumors in vivo more closely than conventional cell cultures. In another recent example for the exploitation of specific properties of spheroids in radiobiology, Fritz and associates (43) have shown the benefit of certain fractionation schemes in a study on cell cycle effects of dose rate and superfractionation using V79 cells grown as monolayers or spheroids. Regrowth has been demonstrated by Durand (32) to be a dominant variable during fractionated radiotherapy using V79 spheroids.

In a cooperative research effort, the groups of Kerbel and Teicher (52, 75) demonstrated a multicellular-mediated resistance to alkylating drugs. Drug resistance induced in EMT6 tumors in mice was completely lost when the cancer cells were isolated and grown in monolayers but could be fully recapitulated when cells were cultured as multicellular spheroids. This has implications both for tumor treatment in the clinic and for the design of drug testing in vitro. A similar resistance to ionizing radiation was identified more than two decades ago by Durand and Sutherland (33) who created the term "contact effect" for this phenomenon. Considering the recent demonstration of a contact effect for drugs, it seems important and challenging to investigate the molecular mechanisms underlying such a resistance. Recent studies of drug and radiation resistance in spheroids have been reviewed by Olive and Durand (117). As a synopsis of the mechanisms possibly involved in the contact effect, the authors conclude that intercellular communication via gap junctions may not contribute significantly to this effect, even though, in the literature, this mechanism has been assumed to cause contact resistance for some time. Predominantly on the basis of work from their own groups, the authors favor the specific nuclear shape and DNA packing in spheroids with a more efficient DNA repair to be the major reason for multicellular resistance. Although expression and constitutive activity of repair enzymes seem to be unaffected by multicellular growth conditions, the change in the accessibility of the substrate to these enzymes appeared to be improved in spheroids compared with single cells, which may explain an increased activity of these enzymes in situ.

Hypoxia poses many medical problems in a variety of biological tissues, including the induction of radioresistance in solid tumors. With the use of techniques that are applicable in the clinic, the detection of hypoxic cells is associated with problems that have not yet been sufficiently solved. Recent progress has been made by the development of a new hypoxic marker, on the basis of the well-known oxygen-dependent metabolic activation of nitroheterocyclic drugs (88). The major improvement with the new drug EF5 is a considerable reduction of oxygen-independent adduct formation, which had generated a large variability in background labeling in different tissues when previous designs of nitroimidazoles were used. The authors demonstrated the hypoxia specificity of EF5 in multicellular spheroids. Furthermore, it was shown that EF5 accumulated in the inner hypoxic regions of HT-29 human colon adenocarcinoma spheroids and colocalized with elevated expression of vascular endothelial growth factor (VEGF) (159). This work illustrates the potential of using spheroids in studies on the regulation of genes that can be induced by hypoxia or other specific environmental factors. Spheroids were also used for in situ calibration of nitroimidazole binding by the combined use of autoradiographic labeling and oxygen microelectrodes (54). The authors suggest from their results that such calibration curves of grain density vs. oxygen tension can be used for measuring oxygen pressure in tissues, if the drug binding level of fully oxygenated regions can be determined. From these recent investigations with hypoxic markers, one may conclude that channeling research efforts into this field may still be worthwhile. Thus the generation of calibration curves in spheroids to be used for oxygen pressure measurements in tissues and the detection of novel markers such as EF5 [possibly with ¹⁹F labels using nuclear magnetic resonance (NMR) or with ¹⁸F labels using positron emission tomography] should all be tested for suitability and practicability.

Another recent approach to quantification of hypoxia in tissue samples, which has been applied intensively to multicellular spheroids, is the so-called comet assay that was reviewed by Fairbairn et al. (39). The method with several different versions allows for the determination of DNA strand breaks in individual cells and may be used to measure the radiobiologically hypoxic fraction in tumors and normal tissues (118).

Spheroids have been involved in recent efforts to increase the efficiency of tumor-targeted radionuclide therapy, as reviewed by Wheldon (166). The efficiency and specificity of radionuclide treatment for certain tumor entities have been documented in spheroids from human neuroblastoma NB1-G cells (162), human glioma U-343MgaC12 cells, and human prostatic adenocarcinoma DU 145 cells (37). In these studies, ¹²⁵I- or ¹³¹I-labeled antibodies were used that were directed against neurectodermal antigens of neuroblastomas or were conjugates based on epidermal growth factor (EGF). Radiation doses were achieved that were associated with a substantial sterilizing effect, the extent of which could be predicted on the basis of theoretical considerations. In similar spheroid studies, Mairs et al. (92) arrived at the conclusion that metaiodobenzylguanidine with no carrier added would achieve a significant therapeutic gain in neuroblastoma treatment compared with conventional radiopharmaceuticals. These conclusions could be confirmed by in vivo experiments with immunodeficient animals. The penetration of the EGF-based antibodies into spheroids was very sensitive to antibody binding and could be enhanced by saturation of peripheral binding sites or by increasing the external antibody concentration. This finding is in accordance with the "binding site barrier" phenomenon that was demonstrated, for example, in guinea pig micrometastases (129) and that is considered a major obstacle to the satisfactory monoclonal antibody treatment of bulky cancers. Using anti-carcinoembryonic antigen antibodies and respective Fab fragments in human colon adenocarcinoma spheroids, Langmuir and colleagues (81) demonstrated very clearly and systematically that antibody size and receptor density were the main determinants of antibody penetration, whereas tumor cell architecture appeared to have only a minor impact.

Several investigators have used glioma spheroids for studying the interaction of cancerous tissue with defense cells and in particular with lymphokine-activated killer (LAK) cells that have been generated by incubating peripheral blood lymphocytes with interleukin-2 (65). The clinical background of these studies is the general resistance of malignant gliomas to treatment with LAK cells or recombinant interleukin-2. Penetration of LAK cells into multicell spheroids from glioma cell lines was relatively poor, although cellular damage was documented at a distance from the invading LAK cells, which may be interpreted as an effect of membrane-damaging agents released by these cells (65, 66). Penetration of LAK cells was much better when "organotypic" glioma spheroids were used that were obtained from continuous cultures of tissue specimens (67). The authors were able to demonstrate a substantial difference in both penetration and toxicity between LAK and peripheral blood cells, which makes their model attractive for mechanistic studies with immunotherapy. This finding also shows possible limitations of spheroids in comparison with established cell lines: stromal elements, such as extracellular matrix, may be synthesized by the aggregated tumor cells or may be induced by cocultures with fibroblasts or by reversible implantation in animals. However, such an "extracellular matrix" or "microvasculature" may still not adequately mimic the situation of the respective stroma in vivo at least in some specific experiments.

	METABOLISM AND METABOLIC ENVIRONMENT

One "classical" incentive for using multicellular tumor spheroids was their similarity with the initial, avascular growth stage of malignancies, with micrometastases, and with intercapillary tumor microregions (reviewed in Refs. 108 and 144). The concentric arrangement of heterogeneous cell populations in spheroids with proliferative cells at the periphery, an intermediate zone with viable and clonogenic, yet quiescent cells, and a necrotic core in the center is clearly suggestive of a diffusion-limited tissue model. It is therefore not surprising that for many years the development of central necrosis in spheroids was attributed to an insufficient oxygen supply; similarly, it was anticipated that the emergence of cellular quiescence was a consequence of hypoxia. Early experimental evidence suggesting that these assumptions may not be true (21, 110, 111) has often been ignored. Currently, a large amount of data from various laboratories shows that proliferation arrest in spheroids is not elicited by depletion of substrates for energy metabolism and that the same is true, with a few exceptions, for the emergence of necrosis. Some of these data are briefly reviewed here. Investigations by Acker et al. (2) and Carlsson and Acker (20) indicate that there is no single factor, such as hypoxia, lack of nutrients, accumulation of waste products, or low pH, that alone is responsible for the development of necrosis. The authors speculate that a combination of some or all of these factors may lead to a "toxic environment," in which cells will no longer be able to maintain their intracellular homeostasis. The results of recent measurements of oxygen tensions and concentrations of glucose, lactate, and ATP in multicellular spheroids (17, 152, 161) argue against the hypothesis of Acker and colleagues. Obviously, most of the spheroids that are commonly used for such studies do not have a very hostile metabolic milieu at the time of the first emergence of necrosis, although hypoxia may develop during a further increase in spheroid diameter at a later stage of cultivation. It has been demonstrated previously that cells can adapt their metabolism to the specific environmental conditions in a 3-D arrangement by reducing their metabolic turnover rates (41, 109). This may explain the relatively shallow metabolic gradients seen in spheroids before the development of necrosis. Such shallow gradients are reflected by relatively small differences between external and central concentrations of metabolites or between external and central oxygen tensions in smaller spheroids. The ATP content in central regions of EMT6 spheroids decreased from levels of 1-3 µmol/g, commonly observed in single cells, to 0 µmol/g during spheroid cultivation (161). However, this drop was preceded by the incidence of massive cell death in the spheroid center, which suggests that ATP is hydrolyzed as a consequence of cell death and that a critical decrease in cellular ATP levels can be excluded as a mechanism involved in the induction of necrosis. These observations agree well with ³¹P-NMR measurements by Freyer et al. (42), who found a relatively constant level of energy-rich phosphates in EMT6 spheroids until cells were very close to death. Similarly, these studies confirmed that the emergence of cellular quiescence was not related to a decrease in nucleotide triphosphate content or pH.

There are certain types of spheroids that are different from those described above with regard to mechanisms involved in the induction of cell death. Monz et al. (103) have shown that the first emergence of necrosis and hypoxia coincide during growth of WiDr human colon adenocarcinoma spheroids. In this case, hypoxia may be the predominant necrotizing factor, but this may be only coincidental. Theoretical considerations suggest that one single limiting factor, such as lack of oxygen, can explain the kinetics of the development of central necrosis during spheroid growth (53), which would support the former interpretation. On the other hand, Kunz-Schughart et al. (78) have demonstrated a remarkable tolerance of hypoxia in spheroids from myc/ras-cotransfected embryonic rat MR1 fibroblasts where central oxygen pressure dropped to 0 mmHg, which is more than for 24 h before the first appearance of necrosis. This result argues in favor of the latter interpretation of the WiDr data or indicates that the development of necrosis in spheroids is a multifactorial event.

Knowledge of processes that are involved in the development of necrosis has been expanded considerably by the generation of spheroids from clonal rhabdomyosarcoma cells (69). These spheroids consist of a mixture of undifferentiated, mononuclear cells and of myotube-like, multinucleated giant cells that are formed by fusion of undifferentiated cells, as illustrated in Fig. 1. Although giant cells occur preferentially in the spheroid center, central necrosis does not emerge even at spheroid sizes of 1 mm and larger. Moreover, proliferating mononuclear cells can be identified by autoradiographic [³H]thymidine labeling in the center of such big spheroids, and the difference in the thymidine labeling index between peripheral and inner regions is much smaller than in all other spheroid types investigated up to now. Obviously, the incidence of differentiation is inversely correlated with the emergence of necrosis in these aggregates. This is true despite the fact that central oxygen pressure is very variable and very low in some of these spheroids as illustrated in Fig. 2. The large variability in central oxygen tension that is not found in undifferentiated spheroids can be explained by the pronounced variability in the proportion of giant cells in central spheroid areas, since there is an inverse relationship between these two quantities (see Fig. 3). Because myotube-like differentiation can be induced in monolayer cultures of these cells by an acidic rather than a hypoxic environment (unpublished data), the correlation shown in Fig. 3 may be the result of an increased induction of myotube-like differentiation by the acidic environment in central spheroid regions. This may be associated with a higher volume density of oxygen-consuming sites, higher local oxygen consumption, and lower steady-state oxygen tensions. The relevance of the numerical volume density of mitochondria for the local oxygen consumption rate and the steady-state oxygen pressure distribution in spheroids has been clearly demonstrated in previous studies (17). The data obtained in rhabdomyosarcoma spheroids illustrate how environment can determine the expression of a certain phenotype and how this expression can in turn determine the microenvironment. Such spheroids may therefore represent excellent models for studying the interaction between gene expression and micromilieu.

View larger version (149K): [in this window] [in a new window]   Fig. 1.   Paraffin section stained with hematoxylin and eosin through the center of a rhabdomyosarcoma spheroid (diameter, 840 µm) with multinucleated giant cells (asterisks). Bar: 80 µm.

View larger version (12K): [in this window] [in a new window]   Fig. 2.   Central PO2 in rhabdomyosarcoma spheroids as a function of spheroid diameter.

View larger version (13K): [in this window] [in a new window]   Fig. 3.   Central PO2 in rhabdomyosarcoma spheroids as a function of the proportion of myotube-like giant cells in central spheroid regions (linear regression: y = 1.21 × +91.7; R = 0.702; P < 0.002).

When the rhabdomyosarcoma spheroids are grown for extensive lengths of time, necrosis may eventually arise in central portions of the aggregates. This event is preceded by the occurrence of apoptosis mainly of the giant cells. This is illustrated in Fig. 4, showing a histological section through a rhabdomyosarcoma spheroid containing multinucleated giant cells, some of which are apoptotic. Apoptoses are visualized by immunostaining of DNA fragments with the TdT-mediated dUTP nick end labeling (TUNEL) technique (47). Such a sequence of events can also occur in undifferentiated spheroids, as shown for aggregates of V79 hamster lung cells in Fig. 5, A and C, where apoptoses were also labeled with the TUNEL technique in comparison with a general DNA staining with 4',6-diamidino-2-phenylindole dihydrochloride (Fig. 5, B and D). In this case it is obvious that apoptotic death is a single cell event that is distributed more or less uniformly within small spheroids and that can hardly be detected in V79 monolayers. With increasing spheroid size, apoptoses accumulate in the spheroid center and eventually "blend" together to form secondary necroses. Fragmentation of DNA in V79 spheroids can also be demonstrated by "laddering" in gel electrophoresis, as illustrated in Fig. 6. Because DNA fragmentation is not entirely specific for apoptosis, ultrastructural investigations with electron microscopy are required to supplement these studies on spheroids in the future. The signal for the induction of apoptosis in spheroids is not known to date, but induction of massive apoptosis by 3-D growth conditions has been demonstrated recently by Rak et al. (125) in spheroids of intestinal epithelial cells. Chemosensitization and induction of apoptosis by the retroviral wild-type p53 expression vector have been reported for p53-defective human lung cancer spheroids (44, 45). Recent data from our own laboratory indicate that oxidative stress may be involved in the induction of apoptosis in V79 spheroids, since glutathione is transiently increased during accumulation of apoptosis in central portions of spheroids (128).

View larger version (124K): [in this window] [in a new window]   Fig. 4.   TdT-mediated dUTP nick end labeling (TUNEL) staining of apoptotic multinucleated giant cells (thick arrow) in a rhabdomyosarcoma spheroid of the type shown in Fig. 1. Thin arrow, intact giant cell.

View larger version (105K): [in this window] [in a new window]   Fig. 5.   TUNEL staining of apoptotic cells in V79 spheroids of 2 different sizes. A and B: diameter = 360 µm. C and D: diameter = 500 µm.

View larger version (98K): [in this window] [in a new window]   Fig. 6.   Visualization of DNA fragmentation by laddering in gel electrophoresis. Left to right: lanes represent the size standard followed by DNA gels obtained from V79 spheroids with diameters of 200, 350, and 500 µm.

In summary, the development of central areas in spheroids with metabolically inactive and/or structurally disintegrated cells is a complex process that is not well understood. For many years, cell death in spheroids has been designated "necrosis," and hypoxia has been presumed to terminate the life of cells in the spheroid center. This may be true for some spheroid types, such as those derived from WiDr colon adenocarcinomas, but there is a variety of spheroid types that develop necrosis in the absence of hypoxia or nutrient deprivation. Currently, evidence is accumulating for the occurrence of apoptosis with increasing frequency in an early stage of spheroid growth and for this to be involved in the emergence of central necrosis in the spheroid center. This is undoubtedly a controversial area, particularly due to methodological problems with the distinction between apoptosis and necrosis. Multicellular spheroids appear to be appropriate models for mechanistic studies in this field, and scientists may be encouraged by this discussion to use spheroids for the clarification of this critical issue.

Molecular aspects of growth regulation in spheroids of A-431 cells have been investigated intensively by Mansbridge et al. (95, 96), who demonstrated a reduced expression of EGF cell surface receptors and an enhancement of tyrosine phosphatases in spheroids compared with monolayers. Using the same squamous carcinoma cell line, Laderoute et al. (80) showed that transforming growth factor- (TGF-) synthesis is more pronounced in these spheroids than in monolayer cells. Microenvironment and cell shape appear to be responsible for these changes.

	MATHEMATICAL MODELING

A traditional field of research on spheroids is mathematical modeling, and advances in this area are still being made. This field is not extensively covered by this review but rather is exemplified by a few recent approaches. Using theoretical considerations, Chaplain and associates (22) have analyzed the impact of nonlinearity in either diffusion or production of a growth-inhibiting factor secreted by the spheroid cells on spheroid growth. Duechting et al. (30, 31) have used spheroids for the simulation of different treatment protocols for irradiation of tumors. From their theoretical considerations, the authors were able to predict many clinical observations and to give recommendations on ways of optimizing the therapy with regard to differential effectiveness of radiation in malignant and normal tissue. Tumor treatment with immunoconjugates has been modeled by Kwok et al. (79). They were able to derive macroscopic binding, dissociation, and diffusion constants for monoclonal antibodies in spheroids from experimental data that may be exploited in experimental studies for the improvement of targeting tumor cells with immunoconjugates.

	INVASION AND METASTASIS

One of the classical assays for studying invasion in vitro is the coculture of tumor spheroids with embryonic chick heart fragments, a model that is still used intensively in many laboratories. Recent advances elucidating molecular mechanisms involved in the invasive properties of cancer cells come from Mareel and colleagues (14, 119), who studied the invasion suppressor function of E-cadherin in MCF7 human mammary carcinoma spheroids. Using the same model system, Tritthart and colleagues (58, 139, 140) have investigated the anti-invasive properties of anticancer agents employing quantitative image analysis. The authors have developed an intriguing way of analyzing various aspects of cellular motility, such as directional migration, area and number of ruffling sites, and so forth, in mouse melanoma spheroids, and they were able to show in a quantitative manner how different anti-invasive agents were directed toward these different aspects of cellular invasiveness (60). An interesting study that used Rous sarcoma virus-infected rat cerebellar cells invading chick heart fragments showed that expression of the neural cell adhesion molecule NCAM is essential for spheroid growth but not for invasion (15) and that invasiveness of hamster glial cells is not necessarily linked to morphological differentiation induced by dibutyryl adenosine 3',5'-cyclic monophosphate (10). Makiyama et al. (93) grew spheroids from high- and low-metastatic clones of mouse sarcoma cells and demonstrated that the propensity of these cells to metastasize was correlated with their adhesiveness and invasiveness in the chick heart assay. To elucidate mechanisms that contribute to the low incidence of cancer in the lens of the eye, Messiaen et al. (101) have demonstrated that isolated lens cells acquire certain malignant properties under long-term culturing for >25 wk; thus cells became invasive in the chick heart fragment assay and tumorigenic in syngeneic animals. Even so, the stage of metastasis was never reached. Brauner and Hulser (16) have demonstrated that invasiveness may be associated with tumor cell-host cell interaction via gap junctions. Recently, this group has initiated very intriguing studies on the role of different connexins in cell-cell interaction and invasion at the single cell level (34, 156).

Recent developments of novel in vitro systems for invasion studies include 3-D organ cultures of human endometrium, which can be invaded by choriocarcinoma spheroids (55), and fetal rat brain aggregates as targets for the invasion by glioma spheroids either from cell lines (91, 121) or from primary material (36). The latter group of researchers has studied the influence of various growth factors on invasion and differentiation and was able to demonstrate a growth- and invasion-promoting effect of EGF on gliomas. Also, a barrier function of leptomeningial tissue against brain tumor cell invasion could be shown using 3-D culture systems (121). The potential of coculturing fibroblasts with human bladder cancer spheroids for mimicking invasion in vivo has been shown by Schuster et al. (133).

	CELL ADHESION

There is a vast amount of literature dealing with cell adhesion molecules and their role in cell aggregation. The majority of this work concentrates on defense cells and their homotypic and heterotypic aggregational behavior. It is not the purpose of this article to review even parts of these investigations, although some aspects of this area may be contained in some of the referenced papers. A second domain of research on adhesion molecules is the field of developmental biology, an area that is partially addressed in EXPERIMENTAL TISSUE MODELING and EMBRYOID BODIES. At this point, only some recent data on adhesion molecules in tumor spheroids and their relation to invasion and metastasis are briefly discussed.

In an elegant, quantitative study, Byers et al. (19) demonstrated how tumor spheroids can be used to elucidate the role of adhesion molecules in the dissemination of cancer cells from the primary lesion. Under the influence of physiological shear forces, cell-cell adhesion in breast and colon carcinoma spheroids was strong, if the cells expressed intact E-cadherin, and was reduced in the absence of this molecule or in the absence of a proper linkage of E-cadherin to the cytoskeleton. Initial evidence that carcinoembryonic antigen molecules mediate the homotypic aggregation of colorectal carcinoma cells in malignant effusions of patients comes from studies of Kitsuki et al. (73), which suggest a possible involvement of the ₁ integrin subunit. Various components of extracellular matrix (ECM) and different integrins have been analyzed by Paulus et al. (120) and Hauptmann et al. (56). In summary, these investigations show that synthesis of ECM components and their receptors can be largely influenced by the culture conditions; in general, expression of these molecules in 3-D cell cultures reflects the in vivo situation much better than that in monolayers. The findings of Sutherland and colleagues (160) support these data by demonstrating a selective downregulation of integrin receptors in squamous cell carcinoma spheroids, which mimics the conditions in vivo. An interesting aspect of manipulating cell adhesion for therapy has been shown by Reith et al. (126). The authors were able to prevent an invasion of glioma spheroids into fetal rat brain cell aggregates by incorporation of exogenous laminin during the formation of these aggregates.

Systematic work on adhesion and ECM molecules in tumor spheroids is rare, which is in contrast to the general knowledge about the importance of cell-cell and cell-matrix interaction in cell biology. Among numerous other problems, the role of ECM synthesis by tumor cells compared with stromal cells is not very well understood and could be studied mechanistically in 3-D cultures. A second example may be the interaction between the expression of adhesion molecules and metabolic gradients, which may be investigated in such culture systems to elucidate the impact of a tumor-typical environment on cellular adhesiveness.

	ANGIOGENESIS

For many years, there has been a strong, continuous requirement by scientists working with tumor spheroids for an appropriate in vitro assay of 3-D angiogenesis. Ever since the pioneering work of Folkman, partially reviewed by Mueller-Klieser (108), on angiogenesis in implanted tumor spheroids, there has been little success in the development of a simple assay for vascularization of cell aggregates in vitro. Spheroids seem to be resistant to invasion by endothelial cells in cocultures (116). It is obvious that many of these approaches with negative results have not been published and have come to the knowledge of the author solely by oral communication.

Considerable progress in using spheroids for studies on tumor angiogenesis has been made by Zwi et al. (181, 182), who developed a technique for implantation of spheroids into the peritoneal cavity, where they eventually become vascularized and from where they can be harvested for quantitative assessment of vascularization. Using this technique, some intriguing studies have recently originated from a collaboration between the groups of Keshet and Neeman (1, 137, 141), which were able to show that vascularization in these spheroids can be visualized by magnetic resonance microimaging. The studies impressively demonstrate the role of VEGF as a major determinant of the angiogenesis process in these glioma spheroids, including an upregulation of VEGF in hypoxic or hypoglycemic microenvironments.

Nehls and Drenckhahn (113) have developed a novel 3-D culture system for systematic studies on angiogenesis in vitro. The system consists of endothelial cells seeded on gelatin-coated microcarriers that are entrapped in a 3-D fibrin matrix. Apparently, there is great potential in this new technique for quantitative evaluation of in vivo-like angiogenesis under well-defined in vitro conditions.

	EXPERIMENTAL TISSUE MODELING

Various biological tissues can be kept as explants in tissue or organ culture, maintaining cellular function for a limited period of time. Although this technique has been used for many years, considerable progress has been made more recently in generating bioartificial tissues in vitro from single cells of primary material or of cell lines. Some of this recent work is briefly highlighted here. The majority of these investigations was performed to study basic mechanisms of organ development, to develop artificial organs for replacement or support of natural organs after failure, to optimize bioproduction, or to obtain realistic pharmacotoxicological test systems. In principle, 3-D cultures were obtained either by exploiting spontaneous cell aggregation and by generating more or less spherical cellular conglomerates or by culturing cells on artificial substrates that induce cellular differentiation and maintain cellular function.

Over the past few years, intensive research efforts have been focused on the development of an artificial liver, and only a few exemplary studies can be mentioned here. In many of these approaches, hepatocyte spheroids (82, 153, 172, 177) or heterospheroids consisting of hepatocytes and fibroblasts or unspecified parenchymal cells (35, 149, 173) were used. The application of artificial support systems, such as porous gelatin sponges (87), agarose (136), or collagen (115), as well as the induction of aggregation by exogenous adhesion molecules (25), appear to be advantageous for efficient long-term culture of liver cells. Reconstitution of the specific geometry, microenvironment, and metabolism of the hepatic sinusoid has been achieved recently in a complex 3-D culture system by Bader and colleagues (5). The efficiency of liver cell cultures in approximating the in vivo situation has been shown not only for histological architecture and cellular metabolism but also for gene expression (114, 150). Tremendous progress with regard to practicability and effectiveness in the clinic has been made by growing pig hepatocytes in nonparenchymal coculture in an artificial 3-D capillary system, as developed by Gerlach and colleagues (50), who reviewed this field in a recent article (49).

Compared with liver, progress toward the clinical application of artificial pancreas is not as advanced. So far, spheroids have been obtained from insulin-secreting recombinant mouse pituitary AtT-20 cells and mouse insulinoma  TC3 cells (157). A promising study has been published by Beattie et al. (8), who achieved long-term viability and function of human fetal isletlike cell clusters implanted under the kidney capsule of athymic mice. A number of researchers have used isolated piscine islets, so-called Brockmann bodies, as "natural" pancreatic spheroids that can be implanted as xenografts after microencapsulation with alginate (131). Oxygen supply seems to play a crucial role for the survival and maintenance of function of such islets, which is the background of quantitative measurements of the oxygen tension distributions in Brockmann bodies (131, 132).

Of the hormone-producing tissues, the pituitary gland has been modeled by spheroids most extensively, with research focused mainly on hormone release, such as luteinizing hormone, after stimulation with the appropriate releasing hormone (134, 158). 3-D cultures of melatonin-secreting cells from the pineal gland represent a further hormone-related tissue model that has been established recently (72). In contrast to pineal cell monolayers, 3-D cultures remained functional for >3 wk and secreted melatonin when challenged with isoproterenol. These aggregates may therefore be useful for kinetic studies on the release of indole amines. Apart from some studies on collagen-embedded spheroids from "classical" FRTL-5 thyroid cells (27, 122), there are only a few recent reports on isolated adult thyroid cells that tend to form follicles in aggregation cultures (174). Nevertheless, thyroid cell spheroids allowed for systematic investigations concerning the role of cell motility, cell adhesion, and E-cadherin in the biogenesis of thyroid follicles (175).

Besides organotypic tissue or slice cultures, aggregating brain cell cultures represent an important part of the methodological repertoire in neuroscience and developmental biology. The scientific groups that are actively working with this model are so numerous that only a fragmentary selection of reports demonstrating the potential of 3-D brain cell cultures can be given here. There are a number of publications from the groups of Bjerkvig and Laerum on brain tumor invasion (e.g., see Refs. 9, 36, 91, 121, 126). Brain cell differentiation has been studied intensively in 3-D cultures (7, 51), as has the process of myelination and demyelination (71, 90, 100, 154) and neuronal degeneration (23). Brain cell aggregates have been used for studying the neural toxicity of lead (180), for analysis of neuronal behavior during human immunodeficiency virus infectious disease (76, 176), during Alzheimer's disease (48), or during Parkinson's disease (167). A specific role of the Muller glia for retinal development could be demonstrated in aggregation cultures of retinal cells by Willbold et al. (169). This field has been reviewed recently by Schmid et al. (130).

Homotypic and heterotypic 3-D cultures, including fibroblasts and/or keratinocytes, have been used for studies on the formation of extracellular matrix (3, 25, 40, 148) and skin (63). Pioneering work on many aspects of skin development and carcinogenesis has come from the laboratory of Fusenig, including studies on hair follicle formation (84, 85) or mesenchymal influences on the differentiation of keratinocytes (84 ,86).

During the past 2-4 years, there has been a tremendous increase in the number of publications regarding the use of 3-D cultures for studying chondrogenesis. This includes research on molecular aspects of matrix formation (89, 94, 123) or chondrocyte differentiation (26, 29). Some of these studies in which fetal cells are employed are transitional between aggregation cultures and cultures of embryoid bodies (26), a cell system that is briefly reviewed below.

Another field in which aggregated cell cultures have been used with increasing frequency for the past 2-3 years is the 3-D preparation of heart cells, mainly of fetal origin. Sinoatrial node cell preparations (13), atrial cell preparations (83, 164), embryonic chick heart cell aggregates (24, 124, 178), or ventricular cell aggregates (151) served as potent models for electrophysiological and pharmacological investigations of the role of various ionic channels and electromechanical coupling or of the autonomic nervous system in the regulation of heart activity. Furthermore, myocardial cell aggregates were employed to study specific questions, such as mechanisms underlying defibrillation (155), pacemaker currents (18), or cardiac side effects of anticancer drugs (102).

There are various other tissues that have been modeled or remodeled by aggregation cultures, which are mentioned only briefly here. These include 3-D cultures of mesangial cells (4, 57), urothelial cells (12), and nasopharyngeal cells (11).

Among other factors, one advantage of aggregating cell cultures is that they can be derived from stable established cell lines, from fetal cells, or from primary material. Aggregate composition can be manipulated, e.g., by growing tumor spheroids with or without fibroblasts (68). The disadvantages may be a limited degree of differentiation and a limited availability of appropriate cells for mixed cultures. For example, which type of fibroblast should be used for heterospheroids may be a critical question, since fibroblasts may be different in different tissues of origin. Some of these problems may be solved or circumvented by culturing embryonic stem cells as multicellular aggregates, so-called embryonic bodies.

	EMBRYOID BODIES

Aggregating embryonic cell cultures have to be clearly distinguished from embryoid bodies that represent 3-D cultures of pluri- or omnipotent embryonic stem cells. Although embryoid bodies have been used for some time after the pioneering research of Evans and Kaufman (38) and Martin and Evans (97-99), there has been a recent burst of reports on mechanistic and molecular work with these powerful models that is, among other indicators, reflected by reviews in highly ranked journals (for a recent review see Ref. 70). Some basic properties of embryoid bodies and some recent advances in the field are briefly summarized.

Embryoid bodies are derived from embryonic stem cell lines that have retained their capacity of lineage commitment, i.e., of generating cells of the hematopoietic, endothelial, muscle, and neuronal lineages. The majority of publications covers hematopoiesis, a field in which knowledge of embryogenesis is most advanced and which is nearing the generation of artificial blood to be used in clinics (61). Much progress has come from the data of Keller and colleagues (165, 168), highlighting the potential of targeted mutations in embryoid bodies, from Nakano et al. (112) on lymphocyte differentiation, or from Zhang et al. (179), showing similar developmental processes in vitro and in vivo. The mechanisms of vasculogenesis have been investigated in embryoid bodies by Wang et al. (163), by Doetschman et al. (28), by Krah et al. (77), and by Heyward et al. (59), with the latter authors analyzing the role of cell adhesion molecules for the development of vessels and their response to inflammatory stimuli. Basic work on neurogenesis has been published by Strubing et al. (142), dealing with electrophysiological and immunocytochemical cellular properties, and by Bain et al. (6), concerning the inducing activity of retinoic acid. A very exciting field of research is the development of muscle cells. Rohwedel et al. (127) have shown that the development of skeletal muscle myocytes is very similar in embryoid bodies and in vivo with regard to the activation of muscle-related genes. The role of MyoD status has been investigated by Shani et al. (135), and the role of TGF- in muscular development has been studied systematically by Slager et al. (138). Pioneering research has been reported recently from the laboratories of Wobus and Hescheler concerning cardiomyogenesis. The authors generated beating embryoid bodies with cardiospecific receptors, ionic channels, and action potentials (170, 171). Gassmann et al. (46) have studied the oxygenation and oxygen-regulated gene expression in embryoid bodies in different oxygen environments, which may be relevant for early embryogenesis where embryonic cells reside at low oxygen tensions before implantation and vascularization.

There are many more interesting aspects of this field that cannot be discussed here. In summary, studies on embryoid bodies represent perhaps one of the most exciting fields of research with 3-D culture systems, and much progress with regard to our understanding of embryonic development and carcinogenesis may come from this research area in the near future.