Characterization of the differentiation of human ... - Wiley Online Library

0 downloads 0 Views 1MB Size Report
(Becton Dickinson) containing Eagle's minimal essen- tial medium (MEM, Gibco) supplemented with 10% fe- tal calf serum (FCS, Gibco) referred to as 10% FCS-.
0 1993 Wiley-Liss, Inc.

Cytometry 14:783-792 (1993)

Characterization of the Differentiation of Human Colorectal Cancer Cell Lines by Means of Voronoi Diagrams Francis Darro,’ Anna Kruczynski, Chantal Etievant, Jean Martinez, Jean-Lambert Pasteels, and Robert Kiss2p3 Division de Cancerologie Experimentale I, Centre de Recherche Pierre Fabre, 81106 Castres, France. (F.D., A.K., C.E.);Laboratoire de Pharmaco-Endocrinologie, Faculte de Pharmacie, C.C.I.P.E., Montpellier, France (J.M.); and Laboratoire d’Histologie, Faculte de Medecine, Universite Libre de Bruxelles, Brussels, Belgium (J.-L.P.,R.K.) Received for publication September 21, 1992; accepted March 16, 1993

This paper describes differentiation in terms of population dynamics through the medium of Voronoi paving which enables (via digital cell image analysis) the structure of human LOVO and HCT-15 colorectal neoplastic cell colonies growing on histological slides to be characterized. Two other tests were also used, i.e., the colorimetric MTT assay that enables the cell growth level to be determined, and a test allowing the assessment of the proliferation index, i.e., the percentage of cells in the S phase of the cell cycle. The results show that these colorectal neoplastic cells exhibited a comparatively high level of organisation in terms of the topographical distribution of nuclei within the clones when the cells were cultivated in media containing even small amounts of fetal calf serum. On the other hand, certain chemically defined media completely overturned this “pseudo-tis-

As Mason (18) states, differentiation implies that a cell acquires certain structural and functional characteristics that endow it with the ability to undertake a specialized task. Differentiation may for example involve the secretion of certain specialized molecules that are not produced by an undifferentiated cell (18). Such differentiation features may be related, as Huschtscha et al. (8) report, to the proliferation of cells, which is regulated by a complex interplay of growth-stimulating and inhibiting factors, including polypeptide growth factors and extracellular matrix proteins. These authors (8) further report that many tumor cells in culture exhibit a marked reduction in their requirements for exogeneous growth factors and an increased ability to produce growth factors that act in an autocrine or paracrine fashion (4,21). Such a feature may

sular” architecture. Furthermore, the colorectal cells growing in media including fetal calf serum exhibited relatively large and dense clones, undergoing an increase in the density of these clones when hormones were added to the culture medium and, concomitantly, a decrease in their proliferation. In contrast, the cells growing in chemically defined media generally exhibited smaller clones whose cell proliferation was paradoxically greater than that of the cells referred to above. This seems to bring out the importance of the part played by the cell loss factor in this cell population dynamic. 0 1993 Wiley-Liss, Inc.

Key terms: Voronoi paving, feulgen staining, digital cell image analyses, differentiation, proliferation, cell population dynamic, colorectal cancer

correspond to differentiation processes undertaken by these neoplastic cells. These are the cell lines which very frequently serve as experimental models for investigators interested in the problems of the differentiation of human colorectal cancers, one of the most frequent and lethal cancers in our latitudes, and in their gastrin sensitivity. As evidence, gastrin plays an

’F.D. is the holder of a grant from C.I.F.R.E., France. ‘R.K. is a Research Associate with the “Fonds National de la Recherche Scientifique” (FNRS, Belgium). 3Address reprint requests to Robert Kiss, Ph.D., Laboratory of Histology, Faculty of Medicine, Free University of Brussels, 808 route de Lennik, 1070 Brussels, Belgium.

784

DARRO ET AL.

Table 1 Composition of the “S” and “ S N Media Used for Culturing HCT-15 and LOVO H u m a n Colorectal Neoplastic Cells” Barnes and Sato (3) medium Mixture of Ham’s F12 and DMEM medium 250 ng Iiml + 25 pg HT/ml + 100 ng PGF2aiml + 100 ng EGF/ml

+

Taub et al. (23) medium A Mixture of Ham’s F12 and DMEM medium + 5 pg I/ml + 5 pg HTiml + 50 nM HC/l + 5 pM T311 + 25 ng PGEliml

S medium (Fig. 1) MCDB-104 medium 5 pg Iiml + pg HTiml 100 ng PGF2aiml 100 ng EGFiml + 25 ng PGEliml SN medium (Fig. 1) MCDB-104 medium + 5 pg Iiml + 5 pg HTiml 50 nM HCil + 10 pM T3/1 100 ng PGF2aiml

+ + +

+ +

“The hormones and growth factors used above were insulin (I, bovin pancreas, Sigma), human transferin (HT, iron-saturated, Sigma), prostaglandin F2 alpha (PGF2a, synthetic prostaglandin, Sigma), prostaglandin E l (PGE1, synthetic prostaglandin, Sigma), epidermal growth factor (EGF, human recombinant, Sigma), 3,3’,5-triiodo-L-thyroxine (T3, Sigma), 17 hydroxy-corticosterone (HC, Sigma), 17-beta-oestradiol (E2, Sigma), and gastrin (G, human, Sigma). The SEG and SNEG media in Figures 2-5 were S and SN media supplemented with 10 nmol/l E2 + 10 nmolil G, respectively.

important role in the growth control of normal and malignant gastrointestinal (GI) mucosae (20,251. Such gastrin-induced modification of the growth of normal and malignant GI tissues is mediated through gastrin receptors (20,22,26). However, Lahm et al. (15) argue that depending on the differentiation status, one molecule may exert different effects and that the same cell line may even respond in different ways to a single growth factor. Accordingly, although gastrin has been widely shown to stimulate the in vitro growth of colorectal cancers (7,14,24), Weinstock and Baldwin (26) and Kolori et al. (13)were unable to demonstrate such a gastrin-induced trophic effect in several GI cell lines. In our present work we looked a t the concept of differentiation in terms of population dynamics. To this end we created a software that enabled us to characterize (via digital cell image analysis) the structure of human LOVO (3) and HCT-15 (2) colorectal neoplastic cells growing on histological slides. This software is based on the concept of Voronoi paving which, in the same way as another similar method (Dirichlet domains), has been used successfully by numerous investigators interested in the problems of cell sociology (5,6,17,27,28). In describing the concepts of in vitro “pseudo-tissular” differentiation, our present study came up against two tests that enabled cell proliferation to be assessed. These were the MTT (3-[4,5-dimethylthiazol-2yL] -2,5-diphenytetrazolium bromide) test, which is a simple colorimetric assay enabling the growth level of cell lines to be determined (12,19), and a test based on the digital cell image analysis of Feulgen-stained nuclei (10,16) which allowed the assessment of the proliferation index, i.e., the percentage of cells in the S phase of the cell cycle.

MATERIALS AND METHODS Chemicals, Medium, and Cells The LOVO (CCL229) and HCT-15 (CCL225) cells were obtained from the American Type Culture Collection a t “passage 0.” The cells were maintained for 50 passages (1passage = 1 week of cell culture) a s monolayers cultured at 37°C in closed Falcon plastic dishes (Becton Dickinson) containing Eagle’s minimal essential medium (MEM, Gibco) supplemented with 10% fetal calf serum (FCS, Gibco) referred to a s 10% FCSMEM, in 1% FCS-MEM or in 1% medium supplemented with 10 nmolil estradiol-17-beta (E2, Sigma, France) + 10 nmol/l gastrin (G, Sigma, France), referred to as 1% EG FCS-MEM medium. From passage P51 to P80, the FCS concentration was reduced from 10% to 5% in control cell line (5% FCSMEM medium). At passage 80 (PSO) the MEM medium was replaced by a chemically defined medium in the shape of MCDB-104 (Gibco). From P80 to P89 the variant cells were cultured in this medium supplemented with either 5%FCS (referred to a s the “5% medium” in the figures in Results, variant no. l),or 1% FCS (“1% medium”, variant no. 2), or 1% FCS + 10 nmolil E2 + 10 nmol/l G (“1% EG medium”, variant no. 3). At passage 89 the cells either remained in the 5%, 1%,and 1% EG media or were cultured for 1passage (from P89 to P90) in serum-free chemically defined growth factorsupplemented “S” (variant no. 4) and “SN” (variant no. 5) media, whose compositions are given in Table 1. All the experiments relating to colony structure characterization and cell proliferation assessments were performed a t passage 91 (P91). According to the results given in the different figures, the S1 (variant no. 6) and SN1 (variant no. 7) cells were the cells cultured in S

VORONOI PAVING AND HUMAN NEOPLASTIC COLORECTAL CELL DIFFERENTIATION

and SN media from passage 89 to 91, while the S and SN variant cells were those cultured in S and SN media from passages 90 to 91. The same procedures were used with respect to the SEG and SNEG media (variants no. 8-11). Therefore, on the basis of the protocol described above, 11 variant cells were obtained with respect to each of the parent cell lines, i.e., the LOVO and the HCT-15 models. All the media were supplemented with a mixture of 0.6 mg/ml glutamine (Gibco), 200 IU/ml penicillin (Gibco), 200 pg/ml streptomycin (Gibco), and 0.1 mg/ml gentamycin (Gibco). In our experiments the FCS was heat-inactivated for 1 h at 56°C. The medium referred to a s S in the figures was the medium defined by Barnes and Sat0 (1)but modified by us in terms of the present experiment. In the same way, the SN medium was modified (Table 1)from the medium described by Taub et al. (23).

Experimental Schedule for Cell Growth Assessments Cell growth was assessed by means of the MTT assay as reported previously (12,19). Briefly, after the incubation of the cells for 96hrs i n the 11 different media (passage P91), the culture medium was removed and replaced with 100 pl MTT (Sigma) a t 1 mg/ml RPMI medium (Seromed, Germany). The plates were incubated for 3 h at 37°C and then centrifuged for 7 min at 400g.The medium was replaced with 100 p1 dimethylsulphoxide (DMSO). The multiwells were shaken on a plate-shaker for 10 min; they were then read on a BioTek Instrument Microplate Reader (EL 308) using a test wavelength of 570 nm and a reference one of 630 nm (12,19). The 11 variant cells from the parent HCT-15 and LOVO cell lines were incubated in a hormone-free medium for 24 h to ensure good plating conditions. In order to test the hormone sensitivity of the LOVO and the HCT-15 cells, a n experiment was carried out in which, after 24 h of plating, the cells were incubated for 24 h in either a hormone-free medium (control), or one supplemented either by 10 nmol/l G, or 0.1 ymol/l proglumide (a gastrin receptor antagonist, Sigma, France), or by a gastrin polyclonal antibody (4-8 nmolll, Dako, France). All the assays were performed in sextuplicate.

785

media described above. After 96 h of culture, five coverslips for each experimental condition were fixed for 15 min in EFA [Ethanol 96" (75~1,neutral Formol40% ( ~ O Vand ) , acetic acid, 5v]. After fixation, the coverslips (cell back-up) were mounted on histological slides with DPX (BDH Chemicals) and stained by the Feulgen reaction, a selective and quantitative (stoichiometric) staining of the DNA. The staining procedures used here were identical to those described previously (10,16,19). Cell image analyses were performed using a SAMBA 2005 (Alcatel-TITN, Grenoble, France) microscope image processor. We employed a 100 x magnification lens (Zeiss Axioplan microscope, numerical aperture: 1.30). Two hundred cell nuclei were selected a t random on each coverslip. The nuclear DNA content of each cell nucleus was assessed by means of its integrated optical density (IOD), which is measured on a 540 nm wavelength and computed a t each pixel (0.16 pm2/pixel) on 256 densitometric levels. We calculated the proliferation index (PI) a s previously described (10,161. The PI index represents the percentage of cells outside the major peak (regardless of its ploidy level) and the related peaks (peaks standing a t 50% of the modal value and/or double the modal value). In the current study, each DNA histogram was defined by 50 classes of IOD values. All the cell lines under study were tetraploid (data not shown), and we defined the PI value as corresponding to the seven classes following the GO-G1 peak, whose CVs varied from 3 to 6% (data now shown). Normal human bladder cell nuclei were used a s reference to establish the GO/G1 peak. This was carried out according to a methodology described elsewhere (11).

Description of Using Voronoi Paving to Characterize the Cell Colony Pattern The same biological samples (histological slides) as those described above were used in this experiment. In the same way, the same cell image processor was also used but with a 2 0 x magnification lens (numerical aperture: 0.50). Four colonies (clones) were analyzed per histological slide with respect to each of the 22 cell variants under study. Thus, a total of 20 clones were analyzed with respect to each of the 22 cell variants. The culture conditions were such that no confluence of cells was observed on the histological slides. We thereExperimental Schedule for Proliferation fore defined a cell clone as a cell population with Index Determination clearly circumscribed boundaries on a histological The proliferation index, i.e., the percentage of cells in slide. the S phase of the cell cycle, was determined by means Software was specifically set up for the present exof the digital cell image analyses of Feulgen-stained periment. This software is related to the use of the nuclei. Voronoi diagrams restricted to 2-D studies (17). The experimental schedule was identical to that preFigure 1 illustrates the numerical image of a n viously described (10,191. Briefly, 100,000 celldm1 me- HCT-15 cell clone. Each black oval represents a cell dium taken from passage 91 of the 22 cell sub-types nucleus whose centre is linked to each of its neighbourunder study were plated in 35 x 10 mm Petri dishes ing cell nuclei by a right-hand segment. The applica(Becton Dickinson), each containing a n 18 x 18 glass tion of Voronoi paving to such a cell clone makes it coverslip on its bottom and 3 ml of the 11 different possible to describe its pattern quantitatively. The de-

786

DARRO ET AL.

FIG.1. The use of Voronoi paving to describe quantitatively the pattern of a cell colony growing on a histological slide. This figure illustrates the numerical image of a HCT-15 cell clone. Each black oval represents a cell nucleus whose centre is linked to each of its neighbouring cell nuclei by means of a right-hand segment which is either external (ES) or internal (IS)in relation to the cell colony as a whole.

tails of this methodology are available in the paper by Marcelpoil and Usson (17). Briefly, these authors state that this method requires three major steps in its development. The first step in constructing a Voronoi diagram associates a polygonal form with each point of the population. The second step involves the elimination of points whose associated form has been altered due to the construction of the partition. Voronoi diagrams are very sensitive to touching objects. This is a drawback of the method. Thus, only “crude” clones were analyzed. In other words, if one wants to define the characteristic pattern of a given cell clone, no “object,” i.e., cell nucleus, can be eliminated; should this occur, the intrinsic characteristics of the structure of the clone cell under study would be modified. This elimination problem connected with the second step of Voronoi paving is encountered in relation to a large number of contexts. This did not apply to us because a limiting factor obliged us to confine our study to cell clones with fewer than 100 nuclei. The maximum size of a clone represented in its entirety on the video screen had to involve fewer than 100 cell nuclei in the case of the HCT-15 and LOVO lines studied here because of limits due to the video monitor and the 20 x enlargement. Thus, robustness testing against irrelevant local alterations in order which may cause a sudden change in the Voronoi diagram is essential. Two robustness checks were carried out. First, a robustness check was carried out by repeating the analysis a number of times on the same clone, which always kept similar (P > 0.05) characteristics (data not shown). A second robustness check was carried out according to the fact that

slides have been rotated slightly and scanned again, and once more leading to similar (P > 0.05) clone pattern characteristics (data not shown). The third and final step included the parametrization and quantitation of the topographical information. Each of the node of intersection in the network illustrated in Figure 1 corresponds to the barycentre of a cell nucleus, which is either external (ES) or internal (IS) in relation to the cell colony a s a whole. The sum of all the ES enabled a perimeter to be drawn circumscribing a surface computable in numbers of pixels. We used this “mean clone area” (MCA1 parameter to determine the size of the cell clones. Figure 1shows that i t was also possible to calculate 1)the number of cells (cell nuclei) that a clone contained, which is defined a s the “number of nuclei per clone” (NNC), and 2) the average length of the right-hand segments which link all the nuclei of the same clone. This parameter is referred to as the “mean segment length” (MSL). Lastly, the level of organization (distribution) of a clone’s cell nuclei in relation to each other, as in the clone illustrated in Figure 1,can also be quantified by means of the “standard deviation of the mean segment length” (SDMSL). The weaker the SDMSL value, the higher a clone’s level of organisation. To sum up, Voronoi paving as employed by us here enabled four parameters to be determined with respect to cell clone patterns, i.e., the size of the clone (the mean clone area MCA), the number of nuclei that it contained (the number of cell nuclei per clone NNC) and the average length of the right-hand segments which link all the nuclei of the same clone (the “mean segment l e n g t h MSL), and its level of organisation (standard deviation of the mean segment length, SDMSL).

Statistical Analyses The assessments of the cell proliferation, the proliferation index, and the clone pattern characteristics are reported as means (k standard error of the mean, SEM), statistically compared by means of the Fisher F test (one-way variance analysis). Equality of variances was checked by the Bartlett test. Non-parameteric rank correlations were performed by means of the Kendall test. The levels of statistical significance represented in the figures are *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

RESULTS Characterisation of the Hormone Sensitivity of the HCT-15 and LOVO Cell Lines Figure 2 illustrates the influence of gastrin (GI, proglumide (P), and gastrin polyclonal antibody (GAB) on HCT-15 and LOVO cells growing in either 5% FCS(5%),1% FCS- (l%), or 1%FCS + 10 nmolil E2 + 10 nmol/l G- (1%EG) MCDB-154 supplemented media. In accordance with a method described previously (121, we arbitrarily defined a s 1 the quantity (number) of HCT-15 and LOVO cells living in each control condi-

VORONOI PAVING AND HUMAN NEOPLASTIC COLORECTAL CELL DIFFERENTIATION

significantly (P < 0.01 to P < 0.001) lowered the proliferation of 616 cell variants, while P significantly (P < 0.05 to P < 0.001) lowered such proliferation with respect to 4/6 cell variants only.

HCT 15 I ID m

2 Y J

w

0 J

0 U

t-

z

0 0

2

5 OO/

1%

1%EG

U

w

LOVO

m

I 3

z

-I -I

w

0 U

0

0 ta

U

5%

1%EG

1 o/a

MEDIUM TYPE FIG.2. The determination of cell growth by means of the colorimetric MTT assay under four experimental conditions, i.e., 1) control (white bars), 2) 10 nmolil gastrin (hatched bars), 3) 0.1 kmol/l proglumide (cross-hatched bars), and 4) gastrin polyclonal antibody (4-8 nmolil, black bars). The HCT-15 and LOVO cells grew in either 5% FCS- (5%), or 1%FCS- (1%)or 1%FCS 10 nmolil E2 10 nmolil G- (1%EG) MCDB-154 supplemented media. We arbitrarily define as 1 the quantity (number) of HCT-15 and LOVO cells living in each control condition a t 96 h. This was done in order to facilitate the comparison of the hormonal treatment effect in all the experimental conditions. After a 24 h plating period, the cells were submitted for 72 h to control or hormonal treatment before growth assessment. Results are reported as means (sextuplicate) 2 SEM which are statistically (Fisher F test) compared to the control value (*, P < 0.05; **, P < 0.01; ***, P i0.001).

+

787

+

tion (first series of white bars) a t 96 h. Thus, all optical density values assessed by the MTT test were expressed a s a ratio compared to a reference value equal to 1. Gastrin a t a 10 nmol/l concentration (second series of bars in Fig. 2) induced no statistically significant (P > 0.05) effect on the proliferation of the six cell variants under discussion in the present experiment. In sharp contrast, both 10 nmol/l proglumide (third series of bars in Fig. 2) and GAB (fourth series of black bars in Fig. 2) significantly modified the proliferation of the HCT-15 and LOVO cell variants. The GAB treatment

Determination of the Proliferation Index and the Clone Pattern Characteristics in the HCT-15 Cell Line The results are illustrated in Figure 3: all the cases of statistical significance (asterisks) relate to the “control” (5%experimental) condition. We selected this condition as control because it represented a relatively standard mode of culture vis-a-vis the cancer cells used here. The assessments of the “mean clone area” (MCA) parameter reveal that the cellular clone sizes in the 5%,1%,and 1%EG media were similar (P > 0.05). The HCT-15 cells growing for one passage in the S and SEG media exhibited clones whose mean sizes were also similar (P > 0.05) when compared to each other and were not statistically distinct (P > 0.05) from the HCT-15 clones cultured either in 5%, 1%,or 1%EG media. When the HCT-15 cell line was cultured for two passages in these S type media (S1 and SEGl), its clones exhibited a mean MCA value which was significantly (P < 0.001) lower than that of their corresponding variant, which was cultured for one passage only in this medium. The fact that E2 and G were present in the culture medium of the HCT-15 cells growing for one passage in the S medium type did not significantly (SEG versus S = P > 0.05) modify the mean MCA value, while such hormonal treatment significantly increased this mean MCA value when the HCT-15 cells grew for two passages in this S medium type (SEG1 a s opposed to S1 = P < 0.001). The HCT-15 cells growing in the four SN medium types (SN, SN1, SNEG, and SNEG1) exhibited clones whose mean sizes were dramatically (P < 0.001) lower than those of the clones from cultures in either the three 5%, 1%,and 1%EG media or the four S medium types. In the present case, the E2 + G hormonal treatment did not significantly modify the mean MCA value (SNEG versus SN = P > 0.05; SNEGl versus SN1 = P > 0.05) (Fig. 3). When the four S and the four SN medium types are taken into consideration, the assessments of the “mean number of nuclei per clone” (NNC) corroborate totally with what we report above with respect to the MCA parameter. By way of a n example, the mean NNC value in the HCT-15 variants growing for two passages in S medium type was significantly lower than that of HCT-15 cell variants growing for only one passage in such medium (S1 versus S = P < 0.001; SEGl versus SEG = P < 0.001). The influence of adding E2 + G to the S medium was also statistically significant with respect to the HCT-15 cells growing for two passages in this medium (SEG1 versus S1 = P < 0.01), while this hormonal treatment remained without apparent effect on the HCT-15 cells growing for only one passage in the S medium (SEG versus S = P > 0.05). In contrast,

788

DARRO ET AL. MEAN ?UMBER OF NUCLEI PER CLONE 1001

.L

MEAN SEGMENT LENGTH

:

601 -,-

RO

E w vl

. vl

60

40

20

5 z

a

5%

1%

IREG

S

SI

SEC SEGl

SN

SNI SNEG SNECI

5%

$.

I%

l%EC

S

SI

SEG SECl SN

SNI SNEC SNEGI

p:

d

MEAN CLONE AREA

p:

t: ffi p!

‘z-l

d

100000

S PHASE

;I;,

T

c *

I

8woo

m 40330

2 m 0

5%

I%IOU;

S

SI

S I X SECl

SN

S N I SNEC SNEGI

5%

1%

I%EG

S

SI

SU; SECI SN

SNI SNECSNEGI

MEDIUM TYPE FIG. 3. The determination of the clone pattern characteristics of the HCT-15 cells growing in 11 different media, whose definitions are given in Table 1.The proliferation index, i.e., the percentage of cells in the S phase of the cell cycle, was determined by means of digital cell image analyses of Feulgenstained nuclei. Results are reported as means (pentaplicate) ? SEM which are statistically (Fisher F test) compared to the control value (5%medium) (*, P < 0.05; **, P < 0.01;***, P < 0.001).

when the three 5%, 1%,and 1%EG media were taken into consideration, the assessments of the mean NNC values contributed information complementary to that obtained with the mean MCA value assessments. Indeed, whereas the mean MCA values were similar (P > 0.05) in the three 5%,1%,and 1%EG media, the mean NNC values of both the 1%EG and the 1%media were significantly (P < 0.01 and P < 0,001, respectively) higher than that of the 5% medium, with the 1%EG mean NNC value being more significantly (P < 0.05) higher than that of the 1%mean NNC value (Fig. 3). An examination of the mean values relating to the “mean segment length” revealed that the distance between the cell nuclei in a 1% or a 1%EG type clone was lower to a very significant degree (P < 0.001) than the distance between the cell nuclei in 5% type clone. This feature reflects the fact that 1)the sizes (MCA) of the cell clones from the 5%, 1%,and 1%EG media were similar and that 2) the NNC value of these clones was higher in the 1% and 1%EG media than in the 5% medium. A detailed examination of the results relating to the MSL parameter revealed that this parameter was a source of very valuable and eminently complementary information in relation to that supplied by the

MCA and NNC ones. In fact, we observed that the 1%EG cell clone type exhibited a cell density (NNC parameter) significantly (P < 0.05) higher than that of the 1%clone type of equal size (MCA parameter). Now, whereas it might have been expected that the mean MSL value of the 1%EG clone type would be lower than that of the 1%type, the mean MSL values were in fact similar (P > 0.05) for both types. To resolve this apparent contradiction we had to examine the values of the fourth parameter, i.e., the proliferation index (the percentage of the cells in the S phase), that we also illustrate in Figure 3. It seemed that the HCT-15 cells growing in the 1% EG medium proliferated significantly less (P < 0.01) than the HCT-15 cells growing in the 1%medium and that these, in their turn, proliferated significantly less (P < 0.001) than the HCT-15 cells growing in the 5% medium. The mean MSL values of the HCT-15 cells growing in the four type S media (S, S1, SEG, and SEG1) were similar (P > 0.05). An examination of the average S phase values indicated that these values were also significantly different in the four S type media. In fact, they diminished significantly (P < 0.05 to P < 0.01) in the sequence SEGl + SEG -+ S1- S (Fig. 3).

VORONOI PAVING AND HUMAN NEOPLASTIC COLORECTAL CELL DIFFERENTIATION

789

MEAN SEGMENT LENGTH

MEAN NUMBER OF NUCLEI PER CLONE

*

50

IM

40 80

E

'O

.

20

v)

10

Li

(A

E

K (A

-!. t

t

k

t

z 3 >.

60

(A

5

0

5%

1%

I%EG

p: d

S

SI

SEG SEGl

SN

SNI SNEGSNEGI

40

5%

1%

I%EG

S

t

SEG SEGl

SN

SNI SNffiSNEGl

SN

SNI SNEGSNEGI

S PHASE

MEAN CLONE AREA

p!

SI

p: d

. 5%

1%

IZEG

S

SI

SEG SEGl

SN

SNI SNEGSNEGI

5%

1%

19EG

S

SI

SEG SEGl

.

MEDIUM TYPE FIG.4. The determination of the clone pattern characteristics of LOVO cells growing in 11 different media, whose definitions are given in Table 1. The legend is identical to that of Figure 3.

Determination of the Proliferation Index and the Clone Pattern Characteristics in the LOVO Cell Line The results are given in Figure 4. The mean MCA parameter assessments reveal that the LOVO cells exhibit similar characteristics to those recorded with respect to the HCT-15 cells (Fig. 3). Specific differences nevertheless occurred. Indeed, if it is true, like with respect to the HCT-15 cells, that the clones from the 5%,1%,and 1%EG media exhibited similar (P > 0.05) sizes and the clones from the S1 and SEGl media were significantly (P < 0.001) smaller than those of the corresponding clones from the S and SEG media, the LOVO cells seemed to resist the 4 SN medium types better (Fig. 4) than the HCT-15 cells (Fig. 3). Considering these four SN media, the mean MCA value was significantly (P < 0.001) lower in the SN1 medium than in the SN one. The mean MCA values in the SNEG and SNEGl media were similar (P > 0.05). The mean number of nuclei per clone was relatively similar in both the SN and SNEG media with only a statistical significance of P < 0.05, while this mean NNC value was lower to a highly significant degree (P < 0.001) in the SN1 medium compared to the SN one on the one hand, and in the SNEGl medium compared to the SNEG one on the other. Thus, taking into con-

sideration the LOVO cells growing in SN type media, it appears that 1)the SN1 clones were smaller than the SN ones, while the SNEGl clones were the same size as the SEG ones; 2) the SN1 clones had fewer cells than the SN ones, as was the case with the SNEGl clones in relation to the SNEG ones; 3) the cell nuclei were located at the same distance from each other in the SN and SN1 clones, while they were much further away in the SNEGl clones than in the SNEG ones; and 4) the SN1 clones proliferated significantly (P < 0.001) more than the SN ones, as was the case for the SNEGl clones in relation to the SNEG ones. An examination of Figure 4 shows that everything stated above for SN type media is also applicable to type S ones.

Description of the Level of Organization of the 11 Types of Cell Clones From the HCT-15 and LOVO Cell Lines An analysis of the MCA and NNC parameters shows that the HCT-15 cells (Fig. 3) seemed to have greater difficulty in growing in the S1, the SEG1, and the four types of SN media than in other culture media. The analysis in Figure 5 shows that this phenomenon was accompanied by an increase in the degree of disarray in the clones' topographical distribution of the cell nuclei

790

DARRO ET AL.

3o

>

LOVO CELL LINE

1

5%

I’%

l%EG

S

SI

SEG

SEGI

SN

SNI

SNEC SNEFI

HCT 15 CELL LINE

MEDIUM TYPE FIG.5. The determination of the level of organization (topographical distribution) of the cell nuclei in relation to each other within a clone by means of the standard deviation of the mean segment length (SDMSL). The LOVO and HCT-15 cells grew in 11 different media, whose definitions are given in Table 1.

in relation to each other. It should be remembered that we quantified the degree of order (or lack of it) by means of the standard deviation of the “mean segment length’ parameter (SDMSL). The maximum degree of disarray was observable in the S1 and SEGl clones. Regarding the LOVO cells, the maximum level of disarray was also obtained in the S1 clones (Fig. 5). We obtained a statistically significant correlation (P < 0.01 / Kendall test) between the development of the values of the MCA parameter and those of the SDMSL parameter.

DISCUSSION In our present work we studied the structure of cell clones from the HCT-15 (2) and LOVO (3) human colorectal neoplastic cell lines that we grew in 11different culture media. Their degree of differentiation were analyzed in terms of their cell population dynamics. This type of problem concerning cell population dynamics

can be studied in particular with the help of mathematical tools such a s Voronoi paving and the Dirichlet domains (5,6,17,27,28). Using such techniques, it is possible to reduce many objects (cells, proteins, and crystals) to points, i.e., a cell’s center of gravity for example. As a consequence, it is possible to treat any such data sets as spatial point patterns, although as Marcelpoil and Usson argue (17), such studies were not feasible before the widespread use of computers because of the vast amount of data involved. This type of mathematical approach has already been used successfully in embryology where studies of considerable scope have been devoted either to cellular inter-actions or to cellular pattern analysis in order to elucidate the way cell differentiation acts on cell scattering and clustering or vice versa to organize the living matter (for review see 6, 27). Marcelpoil and Usson (17) report that a cell is not born in a n information-free environment but in a world consisting of signals sent out by its fellows and further argue that a cell’s environment has to be considered a s a n environment organized by elementary exchanges between the population members; communication plays a structural part, and studying the way cells structure their own environment should therefore increase knowledge about homeostatic tissular controls. Our results show that the various culture media used here dramatically modified the dynamics of the HCT-15 and LOVO cell populations. Before commenting on these results, we would like to make a slight digression with respect to the results of the characterization of these cells’ dynamic hormone sensitivity (MTT test). If the only reference had been to the response obtained in terms of the (zero) gastrin stimulation of the proliferation of the HCT-15 and LOVO cells, they would have been said to be hormone-insensitive and so relatively undifferentiated. But what is striking is that they have attained a high degree of differentiation and so failed to react to exogeneous gastrin stimulation. The present results obtained with gastrin antibody reveal in fact that these neoplastic cells are not only capable of synthesizing gastrin but are also able to use it to control their growth, corresponding to autocrine andlor paracrine regulation exercised by gastrin on human neoplastic colorectal cells (7,251. Whereas it might have been thought that the HCT-15 and LOVO cells growing in the 5% medium would be more undifferentiated than those growing in the 1% and 1%EGmedia, it should be noticed that they have the same gastrin-sensitivity profile a s the latter. In view of these results, which were obtained with the help of the MTT colorimetric test, i t would seem that the HCT-15 and LOVO cells growing in the three above-mentioned culture media displayed a n equivalent degree of differentiation. However, a study of their population dynamics indicates that this was not the case. Indeed, given identical culture times and plating conditions, the HCT-15 and LOVO cells cultured in the 5% medium had clones of more or less the same size as

VORONOI PAVING AND HUMAN NEOPLASTIC COLORECTAL CELL DIFFERENTIATION

the cells cultured in the 1%and 1%EG media, but were less dense and proliferated more rapidly (see Figs. 3,4). This might therefore correspond to a decrease of cell loss in the lines growing in the 1%and 1%EG type media in relation to those growing in the 5% medium. We are in the process of studying this hypothesis by computing the cell loss factor of these different cell lines by means of tritiated thymidine autoradiography (9). The results also show that the degree of organization of the nuclei (topographical distribution) was very much the same for the two cell lines in the three different types of culture media. In sharp contrast, this relatively structured growth lost its organization when the cells were cultured in chemically defined media. Also, such cells exhibited considerably smaller clones than their homologues cultured in FCS-supplemented media. Paradoxically, these characteristics were accompanied by a n increase in cell proliferation. Thus, everything occurred as though the cells’ homeostasis had been interrupted by transferring them from a weakly FCS-supplemented type of medium to a chemically defined one. This would perhaps trigger off distress signals in the cells, which would react by activating their proliferation in a n attempt to compensate for the pernicious influences of the culture conditions to which they were subjected. Contrary to what we observed for the 1%and 1%EG cells, which would probably see their cell loss factor diminished in relation to the 5% cells, it would seem that the opposite process was in operation here. In fact, even though the cells growing in type S1 or SEGl media activated their cell proliferation in a n extremely significant manner, their clones nevertheless remained significantly smaller (see Figs. 3, 4). We have already observed a phenomenon fairly close to the one described here, though in radically different experimental conditions. Indeed, we showed that in vivo, estradiol produces a marked increase in the tumoral cell loss in the MXT mouse mammary adenocarcinoma (9). In conclusion, the present experiments demonstrate that Voronoi paving is useful to describe differentiation processes in terms of cell population dynamics with use of human colorectal neoplastic cell lines as a biological model. The results show that when these cells were cultured in media containing even low amounts of fetal calf serum, they exhibited a relatively high level of organization in terms of the topographical distribution of the clone nuclei. However, certain chemically defined media completely overturned this “pseudo-tissular” architecture. Colorectal cells growing in media including fetal calf serum exhibited relatively large and dense clones, undergoing a n increase in the density of the clones when hormones were added to the culture medium and, concomitantly, a decrease in their proliferation. In contrast, the cells growing in chemically defined media generally exhibited smaller clones whose cell proliferation was paradoxically greater than the ones reported above.

79 1

LITERATURE CITED 1. Barnes D, Sat0 G: Growth of a human mammary tumour cell line in a serum-free medium. Nature 281:388-389, 1979. 2. Dexter DL, Barbosa JA, Calabresi PN: Dimethyl formamide-induced alteration of cell culture characteristics and loss of tumorigenicity in cultured human colon carcinoma cells. Cancer Res 39:1020-1025, 1979. 3. Drewinko B, Romsdahl MM, Yang LY, Ahearn MJ, Trujillo JM: Establishment of a human carcinoembryonic antigen-producing colon adenocarcinoma cell line. Cancer Res 36467-475, 1976. 4. Goustin AS, Leof EB, Shipley GD, Moses HL: Growth factors and cancer. Cancer Res 46:1015-1029, 1986. 5. Honda H: Description of cellular patterns by Dirichlet domains: The two-dimensional case. J Theor Biol 72523-543, 1978. 6. Honda H: Geometrical models for cells in tissues. Int Rev Cytol 81:191-248, 1983. 7. Hoosein NM, Kiener PA, Curry RC, Brattain MG: Evidence for autocrine growth stimulation of cultured colon tumor cells by a gastridcholecystokinin-like peptide. Exp Cell Res 186:15-21, 1990. 8. Huschtscha L, Rozengurt E, Bodmer WF: Growth factor requirements of human colorectal tumour cells: Relations to cellular differentiation. Eur J Cancer 27:1680-1684, 1991. 9. Kiss R, Danguy A, Heuson JC, Paridaens R: Effect of estradiol on proliferation and on cell loss in the MXT mouse mammary tumor. Anticancer Res 6:1321-1328, 1986. 10. Kiss R, de Launoit Y, Wouters W, Deslypere J P , Lescrainier JP, Paridaens R, Vokaer A, DeCoster R, Pasteels JL: Inhibitory action of androstenedione on the proliferation and cell cycle kinetics of aromatase-free MXT and MCF-7 mammary tumor cell lines. Eur J Cancer Clin Oncol 25:837-843, 1989. 11. Kiss R, Gasperin P, Verhest A, Pasteels JL: Modification of the tumor ploidy level by the choice of the tissue taken as diploid reference in digital cell image analysis of Feulgen-stained nuclei. Mod Pathol 5:655-660, 1992. 12. Kiss R, Salmon I, Pauwels 0, Gras S, Danguy A, Etievant C, Pasteels JL, Martinez J: Influence of gastrin, estradiol and gonadoliberin on HCT-15 and LoVo in vitro culture human colorectal neoplastic cell proliferation and differentiation. Eur J Cancer 27: 1268-1274, 1991. 13. Kolori 0, Vaillot MT, Martin F: Growth responses of rat stomach cancer cells to gastro-entero-pancreatic hormones. Int J Cancer 30:65-67, 1982. 14. Kusyk CL, McNiel NO, Johnson LR. Stimulation of growth of a colon cancer cell line by gastrin. Am J Physiol251:597-601,1986. 15. Lahm H, Suardet L, Laurent PL, Fischer JR, Ceyhan A, Give1 JC, Odartchenko N: Growth regulation and co-stimulation of human colorectal cancer cell lines by insulin-like growth factor I, I1 and transforming growth factor alpha. Br J Cancer 65:341346, 1992. 16. Larsimont D, Kiss R, d’Olne D, de Launoit Y, Mattheiem W, Paridaens R, Pasteels JL, Gompel C : Correlation between nuclear cytomorphometric parameters and estrogen receptor levels in breast cancer. Cancer 63:2162-2168, 1989. 17. Marcelpoil R, Usson Y: Methods for the study of cellular sociology: Voronoi diagrams and parametrization of the spatial relationships. J Theor Biol 154:359-369, 1992. 18. Mason MD: Differentiation and cancer. Eur J Cancer 28:255-258, 1992. 19. Pauwels 0, Kiss R Digital morphonuclear analyses of sensitive versus resistant neoplastic neoplastic cells to vinca-alkaloid, alkylating and intercalating drugs. Cytometry 12:388-397,1991. 20. Singh P, Walker JP, Towsend CM Jr, Thompson JC: Role of gastrin and gastrin receptors on the growth of a transplantable mouse colon carcinoma (MC-26) in BALBc mice. Cancer Res 46: 1612-1616, 1986. 21. Sporn MB, Roberts AB: Autocrine growth factors and cancer. Nature 313:745-747, 1985. 22. Takeuchi K, Speir GR, Johnson LR: Mucosal gastrin receptor. I. Assay standardization and fulfillment of receptor criteria. Am J Physiol 237:E284-E294, 1979.

792

DARRO ET AL.

23. Taub M, Cluman L, Saier MH J r , Sat0 G: Growth of Madin-Darby canine kidney epithelial cell (MDCK) live in hormone-supplemented, serum-free medium. Proc Natl Acad Sci USA 76:33383342, 1979. 24. Watson SA, Durrant LG, Morris DL: Gastrin: growth enhancing effects on human gastric and colonic tumour cells. Br J Cancer 59:554-558, 1989. 25. Watson SA, Durrant LG, Wencyk PM, Watson AL, Morris DL:

Intracellular gastrin in human gastrointestinal tumor cells. J Natl Cancer Inst 83:866-871, 1991. 26. Weinstock J , Baldwin GS: Binding of gastrin 17 to human gastric carcinoma cell lines. Cancer Res 48:932-937, 1988. 27. Weliky M, Oster G: The mechanical basis of cell rearrangement. Development 109:373-388, 1990. 28. Zahn C T Graph-theoretical methods for detecting and describing gestalt clusters. IEEE Trans Comput 20:68-86, 1971.