Competitive exclusion of clonal subpopulations in heterogeneous tumours after stromal injury.

Xenografted artificial heterogeneous tumours (AHTs) were created by admixing, in a ratio of 9:1 or 1:9, two clonal subpopulations (designated as clones A and D) obtained from a heterogeneous human colon adenocarcinoma. In unperturbed AHTs these percentages remain constant with increasing tumour size. At average volumes of 250 mm3, AHTs were X-irradiated (15 Gy) and changes in growth rate and composition assayed. A and D cells exhibited equivalent levels of survival after in vivo irradiation as determined by excision assay procedures. At about 2-3 weeks post-irradiation AHTs exhibited a significant enrichment of the majority population in both the 1:9 or 9:1 A:D AHTs. Additional studies were concomitantly performed to determine whether these changes were mostly a function of normal tissue damage or of parenchymal tumour cell killing. In these studies, the normal tissue only was irradiated, tumour cells were implanted one day after irradiation, and the composition of AHTs assayed as a function of time post-irradiation. In these studies, similar shifts in composition with similar kinetics to that seen in the in situ irradiations were found. We therefore propose that these compositional shifts are mainly a reflection of radiation damage to the stromal microenvironment, which is consequently unable to support tumour growth adequately leading to competitive exclusion of the minority subpopulation.

Many human solid cancers are clonally heterogeneous in composition . Mauro et al. (1986) and Hiddemann et al. (1986) have shown that approximately one-third of all human colorectal cancers contain two or more subpopulations based on flow cytometric analysis of DNA content. In this regard, we have been studying the biological characteristics of xenografted artificial heterogeneous tumours (AHTs). These are neoplasms comprised of admixtures of varying proportions of clonally related subpopulations (designated as A and D) originally derived from a human colon adenocarcinoma (Dexter et al., 1981). Zonality and compositional stability characteristics of unperturbed AHTs have been experimentally described (Leith et al., 1985(Leith et al., , 1987, and efforts to model AHT behaviour biomathematically have begun (Michelson et al., 1987a(Michelson et al., ,b,c, 1988Michelson, 1987). We have recently reported the responses of AHTs to treatment with single doses of mitomycin C (Leith et al., 1988a) as part of initial studies to determine the general nature of the response of multiclonal cancers to cytotoxic therapy. A similar focus on AHT behaviour for mammary carcinomas has been taken by Miller et al. (1987).
The tumour bed effect (TBE) is a well-documented phenomenon in which pre-irradiation of normal tissues modifies the subsequent growth behaviour of transplanted neoplasms in the damaged region (Hewitt & Blake, 1968;Urano & Suit, 1970;Jirtle et al., 1978;Trott & Kummemehr, 1983;Begg & Terry, 1983Ito et al., 1985;Milas et al., 1986). Due to the intimate relationship between tumour parenchyma and normal tissue stroma (Siemann et al., 1981), we thought that the relationship between TBE expression and tumour heterogeneity warranted investigation. While zonality and compositional stability aspects of unperturbed colon AHTs have been described (Leith et al., 1987), it is important to see if such characteristics would change in the face of a stressing agent (i.e. ionising radiation) that damages the local microenvironment, as this may have relevance to modelling of therapeutic concepts (e.g. Goldie & Coldman, 1979;Peters et al., 1986;Steel, 1988). In this manuscript, we present data examining the composition of AHTs irradiated in situ. These data are compared to changes in AHTs which themselves have not been irradiated, but have been transplanted to grow in previously irradiated sites (TBE studies) (Leith et al., 1988b

Tumour lines
The DLD-1 tumour system from which the clone A and D subpopulations were obtained (Dexter et al., 1981) has been described in detail. Briefly, the original biopsy specimen was histologically heterogeneous, and this cell line was designated as DLD-1. The A and D subpopulations were obtained by soft agar cloning of the DLD-1 parent line. These lines are distinct in morphology, in chromosome number and in their responses to a number of chemical and physical cytotoxic agents (Leith et al., 1982a(Leith et al., ,b, 1984. In vivo, clone A cells produce poorly differentiated tumours, while D cells produce moderately differentiated colon cancers (Dexter et al., 1981). The A and D lines are maintained in tissue culture according to previously published procedures (Leith et al., 1982a(Leith et al., , b, 1984 and are replenished from frozen stock every 3-4 months. Tumour disaggregation procedures We have previously published procedures for disaggregation of AHTs (Leith et al., 1985(Leith et al., , 1987(Leith et al., , 1988a. In the studies reported here, we disaggregated solid tumours from approximately days 7-70 after initial injection of cell suspensions. Neoplasms were excised and multiple samples per tumour were taken based on previously published considerations of intratumour zonality (Fidler & Hart, 1983;Leith et al., 1985Leith et al., , 1987Michelson et al., 1988). Samples were minced by scalpel, enzymatically dissociated (0.5% trypsin-EDTA, 40min, 37°C; Grand Island Biological Co., Grand Island, NY), counted by haemocytometer and single cells were seeded into 60 mm plastic dishes (Becton-Dickenson Laboratories, Rutherford, NH). Colonies developed at 37°C in a humidified incubator (NAPCO, Seattle, WA) with a 95% air 5% CO2 environment for about 14 days. Colonies were then fixed and stained using 0.5% crystal violet in absolute methanol. Colonies containing more than 50 cells were counted by eye for estimation of the overall colony forming efficiency (CFE) from each tumour.
Colony identification procedures Each colony was visually inspected using phase contrast microscopy during the course of development and characterised as being of either clone A or clone D morphology, as each colony type has a unique appearance. Photomicrographs of these clone A and D colonies have appeared in several publications (Dexter et al., 1979(Dexter et al., , 1981. To validate this procedure, as previously described (Leith et al., 1985(Leith et al., , 1987, we selected colonies of each morphological type on a random basis from a number of different dishes and from different tumours. Individual colonies were trypsinised and these cells were allowed to proliferate until a sufficient number were available for karotyping. Because clone D or A cells contain 45-46 or 70-90 chromosomes respectively (Dexter et al., 1981), we could absolutely determine the ancestry of each colony and correlate this with the morphological identification. In no case was a colony of mixed chromosomal content noted, and in no case was there any discrepancy between the morphological and karyotypic assessments of colony identity. Therefore, we could measure the overall CFE from each tumour, and also determine the relative proportions of clone A and D cells for each admixture condition. We assayed, on average from both control and irradiated tumours, about 500 colonies per sample or about 3,000 colonies from each tumour. Therefore, even at long times post-irradiation, when tumour compositions were changing, this would still yield adequate numbers of the minority subpopulation for assessment. Also, as we were aware that a selection process would occur at long times post-irradiation, about twice the number of total colonies were scanned so as to end up with 50-60 positive identifications of colonies of the minority subpopulation.
Sampling procedures and mathematical techniques Sampling procedures have been previously described (Leith et al., 1985(Leith et al., , 1987 and were based on the design of Wallen et al. (1981). Generally six samples were taken from each tumour for cell yield, compositional and clonogenic studies.
Estimates of the range in the amount of growth delay produced were obtained by using the envelope of uncertainty generated in the individual volumetric growth curves by the standard errors of the mean on the tumour volumes as a function of time post-irradiation. Changes in percentage composition of AHTs with time were obtained by linear regression of probit transformed data (Goldstein, 1964;Finney, 1971).

Production of xenograft tumours
Mice bearing the nu/nu gene on an outbred Swiss background obtained from the Charles River Breeding Laboratories, Wilmington, MA were maintained in the Animal Resources Facilities of Brown University, Providence, RI. Mice were housed in a laminar flow hood (Thoren Industries, Pittsburgh, PA) under specific pathogenfree conditions, with sterilised food, bedding and water. Mice of both sexes of approximately 5-7 weeks of age were used in the studies.
For production of solid tumours, exponentially growing cells were enzymatically removed (0.03% trypsin-EDTA; GIBCO) from plastic flasks and resuspended in Hank's basic salt solution (GIBCO). A total of 1 x 107 cells was injected into the upper hip region in a total volume of 0.25 ml. Mice were ear tagged for individual monitoring, and were separated into various groups on a random basis (Leith et al., 1982a(Leith et al., ,b, 1984. Solid tumours were obtained after injection of either pure clone A or D cells alone, or after injection of 90% A: 10% D, 10%A:90%D or 50%A:50%D admixtures. Cells from these initial admixtures were plated into 60mm diameter plastic tissue culture dishes (Becton-Dickenson Labware, Rutherford, NJ) with 5ml of RPMI-1640 medium and colonies were allowed to develop. As clone A and D colonies have distinctly different morphologies as described previously (Dexter et al., 1979(Dexter et al., , 1981, it was possible to scan the developing colonies (10 x magnification, phase contrast microscopy), identify them as being either A or D colonies, and determine the relative percentage of each. As the colony-forming efficiencies of exponentially growing cells were essentially identical, these scans yielded the quoted values of the percentage of A:D cells injected.
Measurement of tumour size Tumours were measured by calipers in two orthogonal diameters, and volumes calculated using the formula for a prolate ellipsoid: where L and W are the major and minor diameters respectively. We have used this technique in previous work (Leith et al., 1982a(Leith et al., ,b, 1984(Leith et al., , 1988a. Average volumes with standard errors for each tumour group were then plotted as a function of time to obtain growth curves. Volume measurements began at about day 7 post-injection, and extended over the next 60-70 days. Animals with impaired mobility leading to feeding problems due to tumour size or with ulcerated tumours were killed. All measurements for all tumour groups were made by a single individual.

X-irradiations
Mice were irradiated two at a time using a Philips 250kVp X-ray machine operated at 20mA and 250kVp. A 4x6cm collimator was used so that only the right hindlimb and flank areas were exposed. Mice were lightly anaesthetised with Metofane (methoxyflurane; Pitman-Moore, Washington Crossing, NJ), restrained on a lucite irradiation platform and allowed to recover before irradiations. Irradiation distances were 33 cm and dose rates were 1 Gy per min. Exposure doses were measured with a Victoreen R-meter (Victoreen Co., Cleveland, OH) and converted to absorbed doses using appropriate temperature, pressure and Roentgen-Gy correction factors. For TBE studies, 15 Gy was delivered one day before cell injections. While we have previously documented the effects of 15 Gy irradiations on the TBE (Leith et al., 1988b), we repeated these experiments with the in situ irradiations to ensure comparability of results. For in situ irradiations, tumours were irradiated at an average volume of 250mm3. Control animals were sham irradiated.

Results
In Figure la-d, we show the cell yield (CY) data obtained from disaggregation of the various tumour types as a function of time after irradiation. There is a diminished yield from all irradiated tumours, which is evident by about 2-3 weeks after exposure, and these values never recover to control levels although there is convergence at long times post-irradiation. This convergence is due to a decrease in the CY from control tumours at large sizes. A decreased CY as a function of radiation dose has also been demonstrated by Vogler and Beck-Bornholdt (1988).
In Figure 2, we show the clonogenic cell survival from irradiated tumours as a function of time post-irradiation. These values have been normalised to the average colony forming efficiencies (CFEs) of unirradiated, control neoplasms. The CFEs for these controls were: clone A, 17.4% (1.0); clone D, 38.6% (3.7); 90% D: 10% A, 35.0% (4.1); and 90% A: 10% D, 20.9% (2.1) (values in parentheses are standard errors of the means). There is no difference among tumour groups in their survival versus time postirradiation. The survival level assayed immediately after irradiation is about 6 x 10-4, which agrees well with previously published data (Leith et al., 1984). At one day post-irradiation survival in all groups had risen to about 2 x 10-3. Thereafter, the observed survivals rise smoothly and attain values equal to control levels by about 10-15 days post-irradiation (error values are not shown for purposes of clarity, but the 95% confidence limits were typically about 8-30% of the mean survival). Therefore, even though a (CY decreases, beginning at about 14 days post-irradiation, the CFE was equal to that from unirradiated neoplasms. In Figure 3, we show the compositional data obtained from the differential scoring of colonies from the cell survival studies after the in situ irradiations. In a, we show data from the 90% A: 10% D tumours and in b, we show similar results for the 90% D: 10% A neoplasms. Both sets of data show similar trends. At about 15-20 days postirradiation, there is a clear indication of a change in the percentage composition as compared to the stable compositions of control tumours. Linear regression analysis of the probit transformed data (Goldstein, 1964;Finney, 1971), indicates that the slopes of the compositional responses seen in irradiated tumours in Figure 3a and b are significantly different from zero, and are significantly different from the 95% confidence limits on the slope of the regression fit of the data from control tumours. Extrapolation of the data from irradiated tumours indicates that it would take approximately 4-6 months to reach a composition level of 99.99% of the majority subpopulation. Also, in Figure 3 we have included data from mice which had tumours implanted after receiving 15 Gy to the normal tissue one day before implantation of tumour cells. Note that as these tumours grew in the damaged normal tissue, the composition of these AHTs changed in exactly the same manner and with the same timing as that seen for established tumours irradiated in situ. These data provide strong evidence that the noted changes in composition are a function of normal tissue damage, and have little to do with the direct cytotoxic effects of ionising radiation on parenchymal tumour cells.
In Figure 4, we have attempted to compare the effects on tumour growth delay produced by irradiation only of curves. As the in situ irradiations were carried out when 19.8 tumours were approximately 250 mm3 in volume, we have used the time needed for tumours to grow to twice this size 98~e _oas an index of effect. As may be seen in Figure  neoplasms. 0 represents data from tumours irradiated in Superimposed on this shift for irradiated tumours was the day zero at an average tumour volume of 250mm3. O additional selection from a 10%A:90%D situation to a nts data from tumours that were implanted into normal situation even more enriched in D cells. Therefore, even that had been given 15 Gy of X-rays (day zero) one day though the selection process occurs in the 50% A: 50% D rradiation.
AHTs and is consistent with other data, we have chosen to present the data only from the stable 10% A:90% D and 90% A: 10% D AHTs. 40 15 Gy to tumour Discussion The primary result of this research has been the 30 demonstration that radiation damage to normal tissue stroma produces a situation in which the relative cellular composition of AHTs changes with time. Further, the data presented ( Figure 4)  dose irradiation, and have therefore established the ure 4 Plot of the time needed (days) for tumours to grow significance of the TBE effect for clinical radiotherapy, it is m 250mm to 500mm3 (volume doubling time). 0 represents obviously important to perform fractionated studies on a from tumours of varying proportions of clone A:clone D AHTs to determine if the selection processes still occur and s irradiated in situ at an average volume of 250 mm3. The A resents data taken from tumours of varying composition are equally effective. Preliminary data indicate that this is lanted into normal tissue 1 day after a 15Gy exposure, and indeed the case (Leith, unpublished data, 1988). luated after reaching a volume of 250mm3. 0 represents data The recent review by Steel (1988)  with regard to the sequence of application. As subpopulations within heterogeneous neoplasms often show diversity with regard to their responses to any cytotoxic agent , the number of subpopulations and their individual differential sensitivites to cytotoxic agents, as well as integrity of the tumour microenvironment at the start of, or at any point during, therapy will impact on the ultimate results. Note that for large single doses of X-rays, clone A and D tumours respond equivalently in terms of acute cell survival and repopulation kinetics (Figure 2). If there were a differential sensitivity between subpopulations, then the ultimate outcome after irradiation would be the product of the relative sensitivities and percentages of the total clonogenic population occupied within the neoplasm by each subpopulation. For example, we have shown that clone A cells are about 2.3 times more sensitive to mitomycin C than are clone D cells (Leith et al., 1988a). Treating an AHT that was 90% A: 10% D with different sequences of X-irradiation and mitomycin C might produce different outcomes. From the physiological rather than the cellular viewpoint, another reflection of microenvironmental damage is the production of an increased fraction of hypoxic cells in recurrent tumours (Leith, 1988).
If tumour multiclonality is noted in flow cytometric analysis of a biopsy specimen, uncertainty in therapeutic strategy may result (Hiddemann et al., 1986;Mauro et al., 1986). Should such a finding alter the therapeutic approach (i.e. patient stratification for possible assignment to alternative treatment schemes)? This suggests that multiple biopsy sampling may be needed to appreciate the intratumour 'zonality' aspects of the architecture of the neoplasm (Hiddemann et al., 1986;Fidler & Hart, 1983;Leith et al., 1987).
Effective but not totally curative cytotoxic therapy will also produce a situation in which the ultimate survival of a minority subpopulation(s) will become stochastic in nature. In this regard, the selection process produced by radiation damage to stroma would suggest that induction of new subpopulations with different (e.g. drug resistant) properties as postulated by the mutation hypothesis of Goldie and Coldman (1979) may not be as important as might be first thought, because the probability of extinction of new subpopulations arising in a damaged microenvironment might concomitantly be much higher (Michelson et al., 1987a). This research was supported by ACS grant PDT 243C.