A rapid method for the isolation of metastasizing tumour cells from internal organs with the help of isopycnic density-gradient centrifugation in Percoll.

Metastasizing tumour cells from a DBA/2 mouse T-cell lymphoma could be separated from the invaded tissue by isopycnic centrifugation in continuous Percoll density gradients. The metastasizing tumour cells from spleen, liver and lung, derived from a cloned lymphoma-cell line, showed a buoyant density in Percoll of 1.060 +/- 0.010. They could be separated from the host tissue, which had a higher buoyant density in the case of the spleen cells or a lower density in the case of the dead liver or lung tissue. The separated tumour cells as removed from the gradients were viable, and could be analysed by in vitro and in vivo assays. The separation procedure did not affect the expression by the tumour cells of TATAs and H-2 antigens. Furthermore, the method seemed to be applicable to the separation of human tumour cells from mononuclear cells prepared from blood samples of tumour patients by Ficoll centrifugation.

The separated tumour cells as removed from the gradients were viable, and could be analysed by in vitro and in vivo assays. The separation procedure did not affect the expression by the tumour cells of TATAs and H-2 antigens. Furthermore, the method seemed to be applicable to the separation of human tumour cells from mononuclear cells prepared from blood samples of tumour patients by Ficoll centrifugation.
METASTASIZING TUMOUR CELLS from different organs have been compared to tumour cells from the respective primary tumour by several authors (Sugarbaker & Cohen, 1972;Deichman & Kluchareva, 1966;Fogel et al., 1977). In these studies, time-consuming in vivo or in vitro culture methods were used to amplify the organderived tumour cells. During these amplification procedures the uncontrolled outgrowth of minorities in the cell populations could have changed the behaviour of the population originally existing in the organ.
With these methods, some authors showed similarities, others found differences between the surface markers of tumour cells from the primary tumour and those from the organs. Therefore, we tried to develop a quick method for the direct separation and isolation of metastasizing tumour cells from different organs in the mouse.
Previously it had been found that cell types of different origin (Akerstrom et al., 1979), as well as cell organelles, possess unique densities (Price et al., 1979). Such density differences could be used to separate cell subpopulations (Gutierrez et al., 1979;Kurnick et al., 1979) and subcellular components in animals and man (Jenkins et al., 1979). The present paper deals with the establishment of a method which is able to separate small amounts of metastasizing tumour cells from large amounts of host tissue. The method has also been successfully applied for the separation of human tumour cells from blood samples of tumour patients.

MATERIAL AND METHODS
Tumour cell lines and propagation.-The aetiology and origin of the Eb and ESb lymphoma, and the other tumour lines used, is described in a previous publication (Bosslet et al., 1979).
The tumour-cell-lines or clones derived from them were propagated in vivo by i.p. transfer or in vitro by continuous passaging in culture medium as described previously by Schirrmacher et al. (1979a,b) and Schirrmacher & Bosslet (1980). Cloning was achieved by growing single cells in suspension culture in microtitre plates. Human multiplemyeloma cell lines (U 226, ARN 8, 8226) were a gift from Dr Gunther Hammerling, who obtained the lines from Jan Monowada, Roswell Park Memorial Institute, Buffalo, New York. They were propagated in culture as described for the mouse lymphoma lines.
Conditions of metastasis formation.-Uncloned ESb tumour cells from ascites or in vitro culture were washed twice and 105 cells injected s.c. into syngeneic DBA/2 & mice (Zentralinstitut fur Versuchstierkunde, Hannover, FGR). Two to three weeks later, internal organs of tumour-bearing animals were removed and processed as follows at 4°C.
Separation of tumour cells from host organ tissue.-Liver, lung and spleen tissue was squeezed through a stainless-steel mesh, suspended in PBS (Ca-and Mg-), clots were removed by Pasteur pipetting and sedimentation. Erythrocytes were lysed by gently suspending organ pellets for 10 sec with 1 ml of distilled water followed by washing with PBS. Alternatively, erythrocytes were pelleted by centrifugation for 10 min at 2000 rev/min (Model TJ-6 Centrifuge, Beckman) on a 4ml pad of 70%O Percoll Medium (described below). These cell suspensions were then separated on the Percoll gradient.
Continuous density-gradient centrifugation in Percoll.-Percoll density medium was made isotonic by mixing 9 parts of Percoll (Pharmacia Fine Chemicals, Uppsala, Sweden) with 1 part of 10 x PBS. This solution was designated as 100% Percoll-medium. Further dilutions of Percoll-medium were performed with 1 x PBS as recommended for "simplicity" by Kurnick et al. (1979). The organor culture-derived cells were suspended in 1 ml of 100% Percoll-medium and transferred into 15ml Falcon tubes (Falcon, California, U.S.A., No. 2095) which had previously been coated with foetal calf serum. Fifty ,ul of densitymarker beads No. 2, 3, 4, 5, 6 (Pharmacia Fine Chemicals, Uppsala, Sweden) were added to each probe. Fourteen ml of a continuous gradient of Percoll-medium from 70 to 20%, prepared in a gradient mixer, was layered on to the 100% Percoll-medium pad containing between 106 and 5 x 107 suspended cells and other tissue material and the marker beads.
The tubes were centrifuged at 2000 rev/min for 10 min in the Model TJ-8 centrifuge without use of the brake. In the case of unlabelled material, the visible bands of host cells or tumour cells were removed from the gradients with Pasteur pipettes. The cells were washed twice in PBS and then analysed microscopically and immunologically.
Cytotoxicity assay.-The separated tumourcell and host-cell bands were removed from the gradients with Pasteur pipettes, washed twice and assayed as described in Schirrmacher et al. (1979b) for the expression of surface markers in a 4h 5lCr-release assay.

Stability of buoyant density of tumour celllinesfrom culture
In an attempt to determine the buoyant density of tumour cells from culture, 6 different murine tumour-cell lines in stationary growth phase were compared with each other and with normal DBA/2 SL. It can be seen from the data in Table I that the tumour cells have a significantly lower buoyant density than the normal cells. In linear Percoll density gradients, tumour cells were found in bands ranging from 1-051 to 1-068 g/ml, whereas normal SL were found in bands ranging from 1-075 to 1-082 g/ml. Table II contains the buoyant densities determined for various tumour cells taken from culture in different growth phases (i.e. in exponential (Day 1 and 2) or stationary (Day 3) phase).
Data are shown for cloned murine tumour-cell lines derived from low-meta- 1-060 1-078 1-075 5 x 106 cultured tumour cells 3 days after seeding 2 x 105 cells/ml and 1 x 107 normal DBA/2 lymphocytes were used for these experiments. Each gradient contained density-marker beads as internal density controls. * For details see Kerbel et al. (1978).  ARH and 8226). No significant changes were found in the densities of the different cell lines 1, 2 or 3 days after seeding in fresh culture medium.
The data in Table I and II indicate that the buoyant density of murine and human tumour-cell lines from culture is a stable parameter. This stability was a prerequisite for the physical separation of tumour cells from normal cells.
Separation of 75Se-methionine-labelled tumour cells from 5 lCr-labelled normal cells Double-labelling experiments with tumour cells and normal cells were performed to determine the optimal conditions and the efficiency of the Percoll separation technique. Tumour cells were labelled internally with 75Se-methionine and SL with Na2 51CrO4. They were then mixed and centrifuged for 10 min at 2000 rev/ min in a linear preformed Percoll gradient (20-70%). Fractions were harvested from the top, and the radioactivity of each fraction measured in 2 channels set for optimal y emission energy of 75Se and 51Cr. Fig. IA shows a double-label experiment in which 5 x 106 cloned ESb tumour cells (I-LIO) were separated from 5 x 107 normal DBA/2 spleen cells (A-A). A similar experiment in which the mixture of tumour cells and spleen cells was preincubated for 10 min in 0 8 x PBS in 100% Percoll in shown in Fig. lB. It can be observed from this type of experiment that a slight hypotonic pretreatment decreases the buoyant density of tumour cells more strongly than that of the SL, and improves the separation of tumour cells from normal cells, and thus the purity of each cell population.
Both figures show that the crosscontamination of the tumour cells with lymphocytes and vice versa is small (< 10%). Similar data were obtained with tumour cells admixed with normal cells from liver and lung tissue (data not shown). The recovery after the fractionation procedure varies between 60 and 80% of the input radioactivity. There is no significant change in the recovery if between 106 and 108 cells are admixed before the separation. Fig. 2 illustrates an experiment in which 5 x 107 human PBC separated from erythrocytes and granulocytes by Ficoll-Hypaque gradient centrifugation (Boyum, 1968) were labelled with 51Cr (A-A) and mixed with 75Se-methionine-labelled myeloma cells (L-Li) and centrifuged at equilibrium in a preformed linear Percoll gradient (70-20%). It can be seen that the human tumour cells could be separated only partially from the human peripheralblood mononuclear cells, which in this experiment had a density ranging from 1-40 to 1-060. There was an additional peak of peripheral-blood mononuclear cells at a density range from 1 080 to 1-085. These cells were clearly separtaed from the tumour cells.
Stability of surface markers on murine tumour cells after Percoll density centrifugation ESb-lymphoma cells mixed with normal DBA/2 spleen cells were separated on linear Percoll gradients, and the tumourcell band (1.065) and the lymphocyte band (1-078) were removed and labelled 4 6 * % 51Cr-release after 4h incubation of the 51Crlabelled target cells with a 40-fold excess of specific cytotoxic T lymphocytes (CTL). These were directed either against the TATA (DBA/2 anti-ESB), against H-2k (BALB/c anti-CBA) or against H-2d (C57BL/6 anti-DBA/2) membrane antigens. with 51Cr. In order to test the expression of tumour-associated transplantation antigen (TATA) (Bosslet et al., 1979) and of normal H-2d antigens, the cells were used as target cells in a 4h cytotoxicity test with anti ESb, anti H-2d or anti H-2k cytotoxic T lymphocytes (CTLs). The data in Table III demonstrate that the surface markers (TATA and H-2d) found on control tumour cells from tissue culture (Row 1) could also be detected on the tumour cells after separation in the Percoll gradient (Row 3). The same is true for the normal DBA/2 SL from Percoll (compare Rows 2 and 4). These data thus show that the Percoll density-gradient centrifugation does not detectably influence the expression of surface markers such as TATA and H-2 antigens.
In Table IV the results of an experiment are summarized in which 105 ESb TATA+ cloned tumour cells were injected s.c. into syngeneic DBA/2 mice. Eleven days later the internal organs were removed and the metastatic tumour cells separated from host tissue cells by linear Percoll density centrifugation. The recovery of tumour cells after Percoll separation was above 50%, that of host lymphocytes or of dead liver cells above 80%. The tumour cells isolated from the liver could be lysed by anti-ESb and anti-H-2d CTL as efficiently as the tumour cells from culture DISCUSSION We here describe a relatively simple method for the physical separation of tumour cells from host tissue. This separation can be done quantitatively, as shown by double-labelling experiments (Figs 1 & 2). Additionally more than 50 % of the organ-derived tumour cells can be isolated, so that artefacts which might be induced by the enrichment of minorities may be excluded. Furthermore, the surface characteristics of the separated mouse tumourcell populations and of the host cells were not detectably influenced by the separation procedure (Tables III and IV).
The separation profiles of human peripheral-blood lymphocytes and human meyloma cells indicate that it is not possible under these conditions to isolate tumour cells completely free of normal mononuclear cells. The density range in which human peripheral-blood cells band in our gradients is slightly lower than that obtained by Pertoft et al. (1979), Gutierrez et al. (1979 and Ulmer & Flad (1979), who used different centrifugation and Percolldilution protocols. We do not believe that these minor differences in density are due to the labelling procedure, because very similar density distributions were obtained with unlabelled PBC.
The method may have various practical applications in cancer research. Biological experiments in animal and human systems can use tumour cells freshly isolated and separated from the host, which has not been possible so far. The surfaces of these tumour cells can also be characterized by serological and biochemical procedures. This may help better to define the characteristics of metastatic and non-metastatic tumour cells. These advantages might render linear Percoll density gradients a useful tool in tumour immunology, especially in the relatively unexplored field of research into metastasis.