MEASUREMENT OF SELF-RENEWAL IN CULTURE OF CLONOGENIC CELLS FROM HUMAN OVARIAN CARCINOMA

To test the identity of human tumour clonogenic cells and stem cells, a procedure was developed to allow quantitation of self-renewal capacity of human ovarian carcinoma clonogenic cells. Primary colonies grown from malignant effu-sions of 10 patients were disaggregated and replated; secondary colonies were observed to be similar to primary colonies in size, morphology and culture requirements. Density-gradient separation of tumour-cell populations demonstrated that not all primary clonogenic cells are capable of self-renewal during clonal expansion. Patient-to-patient variation in self-renewal capacity was shown to be significantly correlated with the concentration of the tumour-cell population in the effusion fluid, and preliminary evidence of a progressive increase in self-renewal was found in one patient. It was concluded that some, but not all, ovarian-tumour clonogenic cells have the stem-cell property of self-renewal, and that quantitation of such a property may identify an important prognostic variable. THE GROWTH AND RESPONSE to curative therapy of human tumours is a function of the properties of those neoplastic cells capable of tumour repopulation (stem cells) (Steel, 1977). Since no in situ assay is feasible for human tumour-repopulating cells, the frequency of tumour stem cells in a population is approximated by measurement of clonogenicity under semi-solid culture conditions. In this way, the clonogenic cell population has been estimated in a variety of human carcinomas and sarcomas (reviewed by Salmon, 1980). The success of such measurements in quantitating the tumour stem-cell population is not known. The defining property of stem cells is capacity for self-renewal. It is through this property that a cell is capable of initiating a self-maintaining clone and therefore, in the setting of a neoplastic cell population, capable of tumour re-population after sub-curative therapy. Theoretically, therefore, the self-renewal capacity of tumour stem cells will largely account for the biological growth and recovery characteristics of a tumour (Steel, 1977). This has been borne out by experimental measurement of the property of self-renewal of clonogenic cells in human acute myeloblastic leukaemia (Buick et al., 1979) in which a significant association is seen between low values of clonogenic cell self-renewal and good prognosis (Buick et al., 1981). The development of procedures to measure self-renewal of leukaemic clono-genic cells by replating procedures was due largely to ease of colony transfer through the use of methylcellulose, rather than agar, as the semi-solid component of the cultures. We have recently described the use of methylcellulose as a semi-solid support for the growth …

THE GROWTH AND RESPONSE to curative therapy of human tumours is a function of the properties of those neoplastic cells capable of tumour repopulation (stem cells) (Steel, 1977). Since no in situ assay is feasible for human tumour-repopulating cells, the frequency of tumour stem cells in a population is approximated by measurement of clonogenicity under semisolid culture conditions. In this way, the clonogenic cell population has been estimated in a variety of human carcinomas and sarcomas (reviewed by . The success of such measurements in quantitating the tumour stem-cell population is not known. The defining property of stem cells is capacity for self-renewal. It is through this property that a cell is capable of initiating a self-maintaining clone and therefore, in the setting of a neoplastic cell population, capable of tumour repopulation after sub-curative therapy. Theoretically, therefore, the self-renewal capacity of tumour stem cells will largely account for the biological growth and recovery characteristics of a tumour (Steel, 1977). This has been borne out by experimental measurement of the property of self-renewal of clonogenic cells in human acute myeloblastic leukaemia (Buick et al., 1979) in which a significant association is seen between low values of clonogenic cell self-renewal and good prognosis (Buick et al., 1981).
The development of procedures to measure self-renewal of leukaemic clonogenic cells by replating procedures was due largely to ease of colony transfer through the use of methylcellulose, rather than agar, as the semi-solid component of the cultures. We have recently described the use of methylcellulose as a semi-solid support for the growth of ovariancarcinoma clonogenic cells . This paper describes a replating procedure to assess the self-renewal capacity of clonogenic ovarian-carcinoma cells grown in methylcellulose-containing medium.

MATERIALS AND METHODS
Patients.-Ovarian-carcinoma patients were nndergoing routine clinical care at the Ontario Cancer Institute. No patient had received cytotoxic therapy less than 1 month before the study. Malignant effusions were obtained by paracentesis into heparinized (10 u/ml) vacuum bottles. Cells were harvested by centrifugation at 600 g for 10 min and resuspended in McCoy's 5A medium +10%O heat-inactivated foetal calf serum (HIFCS). Mononuclear cells were prepared by Ficoll-Hypaque (density, 1.077) centrifugation (2000 g, 20 min). The tumour-cell-rich fraction was removed and washed twice in McCoy's/10% HIFCS. The resulting suspension was passed through needles of decreasing size to 23-gauge. Viability (trypan blue exclusion) for all cell suspensions was > 90%.
Tumourcolony formation. Ovariantumour cell colonies were grown with the enrichments described by Hamburger et al. (1978) as modified by .

One-ml layers of agar (0.5% w/v) in enriched
McCoy's 5A medium containing 10% HIFCS were formed in 35mm plastic Petri dishes (Falcon). Tumour-cell populations were suspended in a plating layer of 1 ml of 0.8% (w/v) methylcellulose (Dow Chemical, methocel, 4000 cP, premium grade) in enriched CMRL with 15% horse serum. No conditioned medium was used. Cultures were incubated at 37°C in a 7.5% CO2 humidified atmosphere of air. Colonies (defined as aggregates of 20 or more cells) were scored with an inverted microscope at 100 x magnification.
Secondary plating by transfer of pooled primary plates.-After primary colony growth for 7 days, the methylcellulose layer was harvested by pipette after dilution with McCoy's 5A/10% HIFCS. Cells were washed twice in McCoy's 5A and colonies were mechanically disrupted by passage of the cell suspension through needles of decreasing size to 25-gauge. The cells were then replated under identical conditions to the primary plating procedure (1 ml in 35mm dishes) or in 0-1 ml in flat-bottom microtitre wells (Limbro) with identical constitution of under and upper layer . The single-cell nature of the suspension was confirmed by microscopy. Morphological examinations of Papanicolaou-stained colonies plucked from methylcellulose were made on primary and secondary colonies. Secondary plating by 8ingle-colony-tran8fer.
-After primary colony growth for 7 days, the methylcellulose layer was harvested and cells washed as above. The suspension, containing single cells and clusters of up to 30 cells, was layered on a solution of BSA (7 % w/v). Colonies were allowed to sediment at 1 g for 1 min, after which the overlayer (containing only single cells) was removed. The suspension of cell clusters was sedimented (1000 g, 5 min) and resuspended gently in PBS/citrate. The cluster concentration was adjusted to 10 ml and 100 ,ul aliquoted to each of 300 wells of flat-bottomed microtitre trays (Limbro). Wells were examined microscopically, and those containing a single colony >20 cells were selected. The selected colonies were disaggregated by passage (x 10) through 23-gauge needles. Microscopic examination of all colonies indicated complete disaggregation to single cells. 104 irradiated (4000 Gy) autologous cells were added to each disaggregated colony, and the suspension plated in a methylcellulose layer over an agar underlayer in microwells, as described above. After incubation for 7 days at 37°C in a 7.5% CO2 humidified atmosphere of air the colony growth in each well was assessed.
Computation of self-renewal ratio from pooled plate transfer.-Self-renewal ratio is defined as the number of secondary colonies per primary colony and is computed as: PE2 (cols/well) x (cell recovery/primary dish xl -4) PEI (cols/dish) Density separation of cells.-Density separation of cells was performed on discontinuous bovine serum albumin (BSA) gradients. 5-23% and 17-35% BSA gradients were constructed in 12ml tubes by layering 10 lml aliquots of solutions of decreasing BSA percentages. 20-40x 106 cells in 0 5 ml of McCoy's 5A were layered on top of each gradient. After centrifugation at 600 g for 30 min the gradient showing best fractionation was selected and consecutive layers removed. The cells in the fractions were collected by centrifugation, washed once with McCoy's 5A and counted before plating for primary colony growth as described.
Differential assessment of cells in malignant effusions.-Differential assessment was based on methods described previously . Briefly, positive identification of tumour cells was based on characteristics observed on Wright-Giemsa-and Papanicolaou-stained slides, and negative identification by visualization of rosette-forming cells with sheep red blood cells, and cells capable of latex phagocytosis.

RESULTS
Colony formation was observed from single cells derived from pooled primary colonies. The secondary colonies were similar to primary colonies in terms of colony morphology and in the number of cells per colony. Papanicolaou staining of picked secondary colonies indicated cells indistinguishable from primary colony cells, which themselves are a close resemblance to the tumour-cell population of the original sample.
A linear relationship was found between the number of pooled cells plated and secondary colony formation. Data from a representative experiment are shown in Fig. 1. For cell numbers between 3 x 103 and 105 in 35mm plates and between 103 and 3 x 104 in microtitre wells, linearity is apparent. The efficiency of clonogenicity is, interestingly, considerably higher in microtitre culture. This presumably cannot be due simply to a concentration effect, since linearity is evident. For the routine measurement of self-renewal, cells are plated at 103-104 cells/0 1 ml in microwell culture. The optimal time of primary-plate development was determined. Data for Patient 9 are shown in Fig. 2. Transfer was possible only after 6 days of primary culture, since this was when colony growth (minimum 20 cells) could be assessed. Six days proved to be the time of transfer for maximal secondary-colony quantitation in agar, and 8 days in methylcellulose. Subsequently, replating was standardized at Day 7, a compromise between the time required for primary-colony development and the negative influence of extended culture on secondary plating efficiency. Linearity of secondary-colony growth with respect to number of cells plated was not compromised at any of the intervals in Fig. 2 (data not shown).
Primary plating is routinely performed over a range of cell concentrations from 105 to 5 x 105 cells/ml since plating efficiency (PE) cannot be accurately  predicted . We therefore tested the dependence of secondary PE on the number of primary colonies used to derive the cell suspension. Since most cells on the primary plates at 7 days are non-proliferative (primary PE -0 01-0-1%), the relative contribution of colony-derived cells to the cell suspension should be small (we estimate a maximum of 1%) and constant over a range of primary plating concentrations. Table I shows representative data relating secon-dary PE to the number of colonies on the primary plates used to generate the cell suspensions. The average primary colony size was constant over the range of cell concentration for plating. Between 2 x 104 and 5 x 105 cells plated in primary culture (generating 19-5-705 colonies/plate), the self-renewal ratio was relatively constant.
To support the use of a ratio generated from replating of pooled primary plates, we performed a comparison of singlecolony transfer and pooled-plate transfer. Cells from Patient 9 (4th sample, Table  Ilb) were plated for primary colony growth. After 7 days the PE1 was assessed as 650 + 74 colonies/ 105 cells. Harvesting and disaggregating one such plate yielded 9.5 x 104 cells, which when replated generated 83-5 + 16 colonies/104 cells (PE2). The self-renewal ratio is 1-21.
Additional primary plates were used for single-colony transfer as described; 48 colonies were selected, disaggregated and replated with 1 04 irradiated autologous cells. The frequency distribution of secondary colonies arising from this procedure is shown in Fig. 3. The total of secondary  colonies generated was 55, a renewal ratio of 1-14.
The relationship of the clonogenic cells demonstrating self-renewal to the total clonogenic cell population was investigated by density-gradient centrifugation. We have previously reported the use of BSA density-gradient centrifugation to demonstrate considerable heterogeneity within the clonogenic tumour-cell population . The malignanteffusion cell population from Pt 9 was fractionated as described. Individual fractions were then placed in primary culture and the distribution of clonogenic cells derived. Those plates representing different density fractions which contained primary colonies were then subjected to the secondary plating procedure, allowing an assessment of the self-renewal potential in relation to the density of the primary clonogenic cell. The results of a typical experiment are in Fig. 4. The distributions of primary clonogenic cells and those cells with renewal capacity did not coincide; the self-renewal population appeared to be a subpopulation of the primary clonogenic population. Table Ila contains the data for primary and secondary PE and self-renewal ratio of the tumour cells from malignant ascites of 10 patients with ovarian carcinoma. Considerable heterogeneity is apparent both in terms of primary and secondary PE and in the relationship between the two. There appears to be only a weak relationship between primary and secondary PE; i.e., a high level of PE1 is not always associated with a high PE2. In addition, Table Ilb shows that, for Pt 9, the property of self-renewal is not constant over 3 months. Four consecutive effusions, harvested at monthly intervals, indicated a progressive increase in selfrenewal ratio. Table lIc provides evidence that the property of self-renewal can be quantitated from cryopreserved cells; although the absolute values of primary clonogenicity of frozen samples are compromised to varying degrees, the computed value of the property of self-renewal is relatively well conserved.
An analysis of the relationship between this culture parameter and clinical features must await a larger sample. The extent of patient-to-patient variation in self-renewal ratio (0.003 to 1.5) is considerable, and prompted the analysis in Fig. 5. Selfrenewal is shown to be significantly correlated with the tumour-cell concentration in the effusion at the time of removal of the sample. (r = 0-88, P < 0.01). In contrast, a similar analysis of PEI against tumourcell concentration showed no significant relationship (r = 0 55).

DISCUSSION
Implied in the growing use of tumour clonogenic assays as prognostic indicators for tumouir therapy  is the assumption that such assays quantitate the tumour stem cells. Since the only defining property of a stem cell in any setting is the capacity for self-renewal, measurement of this property would allow a test of this important assumption. Additionally, the capacity of tumour clonogenic cells for self-renewal would theoretically be expected to be directly related to the clinical aggressiveness of the individual tumour. A laboratory procedure for such measurements would allow self-renewal to be studied as a target for experimental therapy.
In this paper, we describe a culturereplating procedure designed to quantitate the self-renewal capacity of ovariancarcinoma clonogenic cells. The characteristics of the procedure are consistent with this aim; cells can be identified within primary colonies which have identical capacity for clonal expansion to the progenitors of such primary colonies. The replating procedure yielded reproducible results when applied repeatedly to frozen aliquots of the same sample. It was considered likely that secondary colonies were derived from primary colonies, since cells which were non-clonogenic in primary culture were exposed to identical conditions in secondary culture; it seems unlikely that such a cell would demonstrate clonogenicity after 7 days in culture. This view is supported by the analysis of single colony transfer in one patient. The selfrenewal ratios generated directly (by single colony transfer) and indirectly (by pooled-plate transfer) are very simliar (1.14 and 1 21 respectively). This indicates that it is unlikely that proliferative units of less than colony size (< 20 cells) contribute materially to the computation of self-renewal ratio by the pooled-plate method.
The linearity experiment shown in Fig.  1 indicates that PE in secondary culture is considerably higher in microwell culture than in 35mm Petri-dish cultures. "Feeder" effects mediated by the concentration of supportive cells are not involved, since linearity is clearly apparent in both cases. The culture conditions differ in 2 important aspects: the depth of plating layer and the surface area of the top of the 0.5% agar underlayer. It is of interest that such differences in basic efficiency are not seen in the primary-plating procedure (data not shown).
In addition to the procedure to estimate self-renewal within individual colonies, we have used density-gradient fractionation to obtain information on the heterogeneity of renewal potential within the tumour clonogenic population (Fig. 4). The distribution of cells responsible for PEI and PE2 were found to be nonidentical, consistent with the concept that some but not all primary clonogenic cells could undergo self-renewal during clonal expansion in culture (i.e., not all clonogenic cells are stem cells). Such informaion is of fundamental importance in terms of the use of such assays to define prognostic variables for tumour therapy, since one would predict that only the therapeutic sensitivity of cells with renewal capacity will influence long-term tumour control. The marked patient-to-patient variation seen in self-renewal capacity (Table Ila) was significantly correlated with the tumour-cell concentration in the effusion fluid, and in addition preliminary evidence of progression is available in one patient (Table Ilb). These facts support the contention that the self-renewal property of tumour clonogenic cells may be an important indicator of tumour behaviour. Using the information from this selected group of advanced ovarian carcinoma patients, we are now collecting data from clonogenic cells derived from primary tumours at the time of initial clinical study, in the hope that the renewal properties of such cells can be identified as a factor involved in subsequent outcome of therapy.