- Email: [email protected]
Phenotypic responses of mouse mammary tumours and normal mammary epithelium cultured in collagen gels: Correlation with tumour type and progression
TISSUE AND CELL, 1992 24 (6) 879-894 0 1992 Longman Group UK Ltd.
W. JONES*,
A. E. LEE*, P. RIDDLE*, R. R. DlLSt and R. C. HALLOWES”
PHENOTYPIC RESPONSES OF MOUSE MAMMARY TUMOURS AND NORMAL MAMMARY EPITHELIUM CULTURED IN COLLAGEN GELS: CORRELATION WITH TUMOUR TYPE AND PROGRESSION Keywords:
Mouse,
mammary
epithelium,
tumour,
phenotype,
collagen
gel
ABSTRACT. Mammary turnours in female BR6/Icrf mice and the corresponding contralatcral normal mammary glands were disaggregated with collagenase and the epithelial structures released (‘organoids’) separated from other cellular components by filtration. The organoidy were established in primary culture in a collagen matrix and the outgrowths obtained were studied by light microscopy and time-lapse cinemicroscopy. The pattern of three-dimensional outgrowths produced was found to be specific to the original tissue. Organoids from normal tissue formed a characteristic outgrowth designated Pattern A. Normal tissue from pregnant mice formed an additional characteristic outgrowth (Pattern A’) which has not been described previously. Pregnancy-dependent turnours produced a distinctive phenotypic outgrowth de\ignated Pattern D, whereas pregnancy-independent tumours gave a different distinctive Pattern B as well as a unique specific outgrowth designated Pattern C. Outgrowths of Pattern D from a pregnancy-dependent tumour were removed from culture and implanted into a syngeneic female mouse. Tumours arising in the host were found to be pregnancy-independent and showed phenotypic outgrowths in subsequent culture of pregnancy-independent Patterns B and C. The results show that the type of outgrowths in these cultures correlates with the biology of the tissue in uiuo and that changes in tumour progression in uivo are accompanied by alterations in phenotypic outgrowths in culture.
however in the phenotypic expression of a number of established mouse mammary lines and mouse mammary tissue when cultured in a collagen matrix (Lawler et al., 1983). although the interpretation of these results is complicated by the presence of stromal cells in the tissue fragments. The BR6/Icrf mouse has a relatively high incidence of hormone-dependent mammary tumours during pregnancy (Lee, 1970). These tumours regress at parturition and continue as hormone-dependent tumours for a number of pregnancies although most progress eventually to autonomously-growing tumours. Since each stage of this tumour progression is marked by distinctive pathological features we have used the tumourcontaining mammary gland of the BR6/Icrf mouse as a model to study neoplastic progression. A tissue disaggregation procedure
Introduction
Well-defined changes in cell structure and function accompany neoplastic progression. Whereas differences between populations of normal and tumour mammary epithelial cells persist in organ culture, studies with dispersed normal and tumour mammary cells cultured in a collagen matrix have failed to demonstrate differences in phenotypic behaviour (Yang et al., 1979; Richards et al., 1983). Differences have been described * Imperial Cancer Research Fund, Lincoln’s Inn Fields, London WCZA 3PX, UK and tDepartment of Biochemistry and Physiology, School of Animal and Microbial Sciences, University of Reading, P.O. Box 228, Whiteknights, Reading, RG6 2AJ. UK. Correspondence to: W. Jones, Department of Molecular Science, Glaxo Group Research Limited, Greenford Road, Greenford, Middlesex. UB6 OHE, UK. Received
21 April
1992 879
JONES ET AL.
880
(Jones et al., 1983) has been used which preserves the structural integrity of mammary parenchymal units whilst removing virtually all stromal components including fibroblasts. The defined populations of epithelial components produced (i.e. organoids) were then cultured in a collagen matrix to investigate differences in the patterns of phenotypic behaviour of atypic and neoplastic mammary epithelia compared with control mammary epithelia. We report that primary mammary tumours of the BR6/Icrf mouse cultured in a collagen matrix display characteristic phenotypic patterns of outgrowth which differ from the phenotypic outgrowths observed with normal mammary tissue and which can be used to determine the pathobiology of the mammary tumours.
Materials and Methods Mammary tumours and control tissue The mammary gland containing a tumour in the left vulva1 region which had shown hormone dependence during the previous pregnancy was removed from two 8- to 9-day pregnant BR6/Icrf mice (designated tumour A and tumour B). For comparison, pregnancy-independent tumours in the mammary fat pad of the left vulva1 region were removed from two virgin female BR6/Icrf mice (designated tumours C and D). With each of the four mice, the complimentary non-involved mammary gland was used as the control. Tissue disaggregation
Each mammary gland was disaggregated by limited digestion with collagenase as described by Jones et al. (1983). The epithelial structures released (organoids) were washed for three 30-min periods on a vertical rotary mixer at 37°C in Medium RPMI-1640 supplemented with 5% (v/v) foetal calf serum (Gibco) and filtered sequentially through 300, 95 and 51 micron pore-size Nylon monofilament cloth (H. Simon, Stockport, Cheshire). The organoid fractions collected on each filter, as well as the cellular material in the filtrate from the 51 micron filter, were suspended in warm Medium RPMI-1640 supplemented with 20% (v/v) foetal calf serum and lOpg/ml insulin (Sigma) and cultured separately in a collagen gel matrix.
Primary culture in collagen matrix
Each organoid fraction as well as the cellular material in the filtrate from the 51 micron filter was separately pelleted and mixed at 4°C with rat tail collagen at 2 mg/ml (Hallowes et al., 1980; Jones and Hosick, 1986). Portions (1 ml) of the mixture were layered over preformed collagen gel (0.5ml) in 35 mm diameter culture dishes and the dishes warmed to 37°C for 10min for the collagen gel to set. The cultures were overlayed with 2.0 ml of Medium RPMI-1640 supplemented with 5% (v/v) foetal calf serum and insulin (10 pg/ml) and maintained in a humidified atmosphere of 10% CO2 in air at 37°C for up to 20 days. The medium was changed after 24 hr and subsequently at 48-hr intervals. Secondary culture in collagen matrix
Once the characteristic phenotypic outgrowths from organoids had been established during primary culture in the collagen matrix, the individual gels were digested for 30 min at 37°C with 20ml of collagenase (Type lA; 50 u/ml; Sigma) in Medium RPMI-1640 supplemented with 10% (v/v) foetal calf serum on a vertical rotary mixer. The digested material was then allowed to settle under gravity for 10 min and the supernatant discarded. The sediment was washed for three 30-min periods in Medium RPMI-1640 supplemented with 10% (v/v) foetal calf serum at room temperature on a vertical rotary mixer. After each washing, the sediment was allowed to settle under gravity and the supernatant discarded. The final sediment was mixed with rat tail collagen at 4°C and again cultured in collagen gel as described above for primary culture. Implantation in vivo of specific phenotypic outgrowths and subsequent culture in collagen gel of the mammary tumours that developed
Once the characteristic phenotypic outgrowths from organoids prepared from tumour A had been established in primary culture, specific phenotypic outgrowths were removed individually from the gel under a dissecting microscope by cutting the gel around the structures with a pair of l&gauge hypodermic needles attached to 1 ml syringes. Fragments of collagen devoid of mammary structures were removed as controls.
D
tumour
t Normal
C
tumour
phenotype
Pregnancy-independent
patterns,
representing
Pattern
Pattern
Pattern
t Pattern
::::::::,>
Pregnancy-dependent tumour B
b{
pcri-tumoral
C
B
A
D
D
Phenotypic outgrowths in primary culture
and summary of results
b Pattern
epithelium
1. Design of experiment
Pregnancy-dependent tumour A
Normal
Tissue
Table
t
normal
2
1
from culture
2: tumour
1: tumour
selected
Mouse
Mouse
::::::1::::::
tissue,
*{
After implanting into pairs of syngeneic mice
of organoids
C
B
A
C
D
from pregnancy-dependent
* Pattern
* Pattern
+ Pattern
*Pattern
rPattern
Phenotypic outgrowths in secondary culture
tumour
A.
developed
b tumours
Mammary
After implanting into pairs of syngenic mice
?J
Y
z
8 F
z
E
!? 0 ::
z
JONES ET AL.
The samples were implanted individually into the right flank of pairs of syngeneic nonpregnant host female BR6/Icrf mice which were palpated regularly for evidence of tumour formation in the implanted areas. Mice which developed tumours in the implanted areas were killed, and the mammary tumours (tumours l-3) were disaggregated and cultured in collagen gel as described above for the original tumours. A small piece of each tumour was processed for histological examination. Htitology
A fragment of each of the primary mammary tumours (tumours A-D), as well as of tumours l-3 arising from implantation of specific phenotypic outgrowths, was removed and fixed in formol-saline, dehydrated through a graded series of alcohols and embedded in paraffin wax. Sections (5 microns) were stained with haematoxylin and eosin and examined by light microscopy. Time-lapse cinemicroscopy
Representative cultures were filmed by timelapse cinemicroscopy using 2-set exposures at intervals of 4 min (Riddle, 1990). Tissue-matrix adhesion
Adhesion of mammary outgrowths to the collagen matrix was estimated by disrupting the gel in the vicinity of the structures with a pair of hypodermic needles. A dissecting microscope was used to observe whether the structures were released intact and free from adhering collagen, or were disrupted by the procedure. Results For convenience, Table 1 outlines the design of the experiments and summarises the results obtained. A. Phenotypic outgrowths of normal mammary epithelium
Organoids derived from preparations of normal and peritumoural mammary epithelium tissue produced three-dimensional branching tubular structures after 24 hr in culture. The prevalent phenotype (Pattern A; Fig. la) (see Jones and Hosick, 1986) consisted of fine pointed multicellular tubules protruding into the gel which became disorganised after 8 days in culture. The filtrate from the 51
micron filter, which contained very small clusters of cells and monodispersed cells, demonstrated a high degree of reorganisation of these cells in culture and produced phenotypic pattern A after about 48 hr. The second phenotypic pattern of outgrowth (Pattern A’; Fig. lb) which was only observed with normal and peritumoural tissue from pregnant mice consisted of a few broad, blunt projections extending from the organoids into the collagen matrix which persisted for up to 20 days in culture. A small proportion of the organoids showed growth patterns intermediate between Patterns A and A’ but these were not distinct enough to enable a separate phenotype pattern to be assigned. B. Pregnancy-dependent
tumours
Histological examination showed that the pregnancy-dependent tumours were well differentiated and had maintained the ductallobular organisation of the normal differentiated mammary gland in pregnancy although there was some epithelial hyperplasia (Fig.2a). Connective tissue stroma surrounded the epithelial units and the tumours were not invasive. There was some secretory material in the acini but much less than in the distended lumina of the acini of the complementary control gland (Fig. 2b). Multicellular epithelial organoids of various sizes were released from the tumour tissue on disaggregation. Although filtration efficiently separated these structures by size, it was not possible to distinguish between organoids from tumour and peritumoural normal tissue by shape or size. The filtrate from the 51 micron filter consisted mainly of monodispersed and small clusters of epithelial cells and stromal fibroblasts. Hence this filtration procedure resulted in the organoids collected on the 95 and 51 micron filters consisting predominantly of epithelial structures essentially devoid of contamination with fibroblasts and other stromal components. Phenotypic outgrowths of pregnancy-dependent tumours in primary culture. Limited
three-dimensional outgrowths from tumourderived organoids were observed after 24 hr, but phenotypic differences between tumour and complementary normal tissue did not become apparent until after 4 days.
MAMMARY
TUMOUR
PHENOTYPES
IN COLLAGEN
GEL
883
Fig. 1. Phenotypic patterns of three-dimensional outgrowths established in primary collagen culture by mammary organoids. (a) Characteristic fine needle-like structures (Pattern A) observed with normal and peritumoural normal tissue; (b) amoeboid-like structures (Pattern A’) only observed with normal and peritumoural tissue from pregnant mice; (c) lichen-like wwtures (Pattern D) only observed with pregnancy-dependent turnours. The cells are spread as a sheet at the interface of the upper and lower layers of the collagen gel (see arrowheads): (d) spherical multicellular structures (see arrowheads) only observed with cellular material from the 51 micron filtrate of disaggregated pregnancy-dependent turnours. Fig. la x 130; Fig. lb x260; Fig. lc x130; Fig. Id x260 (phase contrastl.
Although ‘normal’ Patterns A and A’ of phenotype outgrowth (see Fig. la, b) occurred with organoids from the pregnancydependent tumour tissue, this was probably due to the presence of peritumoural normal tissue. In addition, a unique phenotype (Pattern D) with a lichen-like appearance (Fig. lc) was observed with all cultures of organoids prepared from the pregnancydependent turnours. This structure increased in size throughout the 20 days culture due, in part, to distension of the lumen. The filtrate from the 51 micron filter produced small
rounded clusters of cells by day 4 in culture which had a rather dark appearance under phase-contrast (Fig. Id). Tumourigenicpotential of specificphenotypic outgrowths in primary collagen gei after implantation into syngeneic mice. The phenotypic patterns of outgrowth established by organoids derived from tumour A were analysed further as follows. Phenotypic patterns of outgrowths common to both tumour and non-involved glands (Pattern A, Fig. la and Pattern A’, Fig. lb) and of outgrowths uni-
JONES ET AL.
884
que to the primary mammary tumours (Pattern D, Fig. lc) were dissected from the gels as well as areas of tissue-free gel as controls. Samples of each type of outgrowth were then implanted into pairs of syngeneic non-pregnant female BR6/Icrf mice. The two mice implanted with tissue that displayed phenotype Pattern D in primary culture developed palpable tumours at day 21 (mouse 1) and at day 32 (mouse 2) after implantation, i.e. these tumours arose independently of pregnancy. Mouse 1 was then mated at day 27 post-implantation and littered 21 days later. The mammary tumour carried by this mouse (tumour 1) did not regress after parturition and was removed at day 53 post-implantation. Mouse 2 was not mated and the mammary tumour (tumour 2) was removed on day 56 post-implantation. Both of these tumours, which were classified histologically as Type B mammary adenocarcinomas had maintained a degree of tissue organisation (Fig. 3a, b) despite gross hyperplasia and dysplasia (see Lee et al., 1983). This loss of pregnancy-dependence, together with the increasing degree of tissue disorganisation observed by histological examination, indicated that both tumours had progressed to a more autonomous form. One of the pair of mice implanted with fragments of tissue from control animals that displayed the phenotypic outgrowths common to both tumour and normal tissue (Patterns A and A’) in primary culture developed a tumour (tumour 3; Fig. 3c) 28 days after implantation. This tumour grew more slowly in vivo than tumours 2 and 3 and, when removed on day 75 post-implantation, had a very necrotic core. Neither the other mouse of this pair nor the pair of mice implanted with organoid- free areas of collagen gel produced tumours during the 103 days after implantation that they were observed. Phenotypic outgrowths in collagen gel of tumours l-3. Tumours l-3 were disag-
gregated and the organoids were recultured separately in collagen gels as described previously. Tumour 1, derived originally from material showing phenotype Pattern D in primary culture, gave rise within 12 days to outgrowths mainly displaying phenotype Pattern C in collagen gel culture. This pattern of outgrowth (see Fig. 4 b, c) is characteristic of
pregnancy-independent mouse mammary tumours in culture (see below). The main phenotype observed with organoids prepared from tumour 2 were amoeboid shaped and consisted of a large rounded body with up to 5 large multicellular protrusions with rounded tips (Pattern B; Fig. 4a). Tumour 2 also gave rise to a small proportion of the lichen-like phenotype Pattern D (Fig. 5) seen with the original tumour organoids in primary culture (cf. Fig. lc). A further proportion of organoids displayed sheet-like spreading of cells (data not shown) which was also seen with some of the organoids displaying phenotype Pattern D. This represents the growth of a sheet of cells along the interface between the upper and lower layers of the collagen gel where the organoids had settled. These phenotypic outgrowths were fully established within 9 days in culture. Tumour 3, which arose from implanted material displaying the ‘normal’ phenotype Patterns A and A’ in culture, gave rise predominantly to the lichen-like Pattern D associated with outgrowths in culture from pregnancy-dependent tumours (Fig. lc). Again this phenotypic outgrowth was fully established within 9 days in culture. With each of these tumours a proportion of the organoids displayed phenotypes in culture which were not classified further. Tumourigenicity of outgrowths in culture from tumour 2. Phenotypic outgrowths (Pat-
terns B and D) from cultures of organoids derived from tumour 2, as well as the remaining fragments of indeterminate phenotype, were dissected from the collagen gels and each implanted into pairs of syngenic mice as described previously. Mammary tumours developed in all of these mice within 59 days of implantation. Gross metastatic nodules were present in the lungs of the pair of mice implanted with material of phenotype Patterns B or D. One of the mice implanted with fragments of material of indeterminate phenotype showed micro-metastases in the lungs. C. Pregnancy-independent Histological examination that both tumour C and typical of adenocarcinomas this strain of mouse, i.e.,
tumours
(Fig. 6) showed tumour D were associated with they showed no
MAMMARY TUMOUR PHENOTYPES
IN COLLAGEN
GEL
Fig. 2. Histological appearance of (a) pregnancy-dependent tumour in the mammary gland showing maintained lobule-alveolar development but varying degrees of epithelial hyperplasia and less secretory material than in (b); (b) contralateral normal gland at day 9 of pregnancy showing extensive ductal lobule-alveolar development and dilated lumen containing secretory material x72.
885
886
JONES ET Al
MAMMARY
TUMOUR
PHENOTYPES
IN COLLAGEN
GEL
X87
888
JONES ET AL.
MAMMARY
TLMOUR
PHENOTYPES
IN COLLAGEN
evidence of the ductal-lobular organisation of the normal gland, the stroma was sparce and there was no evidence of necrosis. The complimentary control mammary gland of these mice showed no obvious pathological changes. Phenotype
outgrowths
in primary
culture.
After 2 days in culture the tumour-derived organoids from each of the three filters showed detectable outgrowths of both Pattern A (Fig. la) and Pattern B (Fig. 4a), as well as the unique Pattern C (see Fig. 4 b, C). The tumour-derived filtrate from the 51 micron filter produced a distinctive phenotype pattern consisting of a few small refractive clusters of cells which persisted throughout the 16day culture period. The development of the unique phenotype Pattern C was analysed by time-lapse cinemicroscopy (Fig. 7) for a period of 1.50hr. Representative frames from the film showed that the main body of the cluster of tumour cells maintained its spherical shape with minimum interaction with the collagen matrix. An occasional columnar protusion arose from the surface of the structure and extended some distance into the matrix as a stalk with a rounded tip (Fig. 7f) which usually regressed with time. The tips swelled and pinched off by budding, forming satellite spheres in the vicinity of the parent organoid
GEL
WY
(Fig. 7i) which appeared to consist of single cells or clusters of cells. Phenotypic outgrowth patterns in secondar? and tertiary culture. Multicellular structures
displaying normal phenotype outgrowths of Pattern A (Fig. la) or tumour-derived Patterns B or C (Fig. 4) in primary culture were released intact from the gels by digestion with collagenase. These structures were settled under gravity and the supernatant, which contained contaminating fibroblasts and monodispersed and small clusters of stromal cells, was discarded. The multicellular structures were re-established in collagen gel and secondary culture continued for 18 days. The gels were again treated with collagenase, contaminating fibroblasts and stromal cells removed as described previously and the multicellular structures again established in collagen gel (tertiary culture). After 3 days in secondary or in tertiary culture, each preparation had established its characteristic phenotype of outgrowth, i.e.. Pattern A, B or C. Tissue-matrix adhesion. Microdissection at the end of culture released structures of Pattern C intact and essentially free from collagen gel. By contrast, outgrowths with ‘normal’ phenotype Pattern A or with tumour-related phenotype Patterns B or D were disrupted by this procedure indicating
Fig. 3. Histological appearance of tumours arising in syngeneic host female mice after implantation with phenotypic outgrowths established in primary collagen gel from tumourderived organoids. (a) Tumour 1 tissue 53 days after implantation with outgrowth of Pattern D: (b) tumour 2 tissue 56 days after implantation with outgrowth of Pattern D; (c) tumour 3 tissue 75 days after implantation with outgrowth (Patterns A and A’) common to both tumour and normal mammary tissue. Tumours 1-3, which had lost pregnancy-dependence, rctaincd some degree of lobulo-alveolar and ductal organisation despite gross hyperplasia and dysplasia. x72. Fig. 4. Phenotypic patterns of outgrowth established in primary collagen culture by mammary organoids from pregnancy-independent mammary turnours. (a) Pattern B showing multicellular clusters of tumour cells with a few short. conical protrusions extending into the surrounding collagen matrix; (b,c) pattern C showing multicellular clusters of tumour cells wtth short. blunt stalks (b, arrowhead 1). The ends of these stalks swell up (b, arrowhead 2) and then bud off leaving behind satellite spheres (c. arrowhead 3) composed mainly of smgle cells x280 Fig. 5. Phenotypic pattern of outgrowth orpanoids derived from tumour 2. x200.
of Pattern
D established
in collagen
gel culture by
Fig. 6. Histological appearance of pregnancy-independent mammary tumour tissue showing no evidence of the ductal-lobular organisation of the normal contra-lateral mammary tissue, sparse stroma and no evidence of necrosis. x72.
:““.‘:
b
I
bl I[ : :
...
0
: .
f
...
: :
:
.. . ....
I
.0. I”‘i
C
. ... . .. : f : :
:i
.
‘b
7
“““:
V
i
. ..!?...... i .,.......
2i
Fig. 7. Time-lapse cinemicroscopy sequence of development of Pattern C phenotypic outgrowths by pregnancy-independent tumour organoids in primary collagen gel culture. Time (hr) from start of culture: (a) 0; (b) 21; (c) 37; (d) 41; (e) 68; (f) 85; (g) 108; (h) 132 and (i) 150. Frames a-d show three-dimensional protrusions extending into the collagen matrix. Frames e-i show the compaction of the organoid and withdrawal of cellular protrusions leaving satellite structures of one or more cells, as indicated by the dotted areas. x 116.
a high degree of adhesion matrix.
to the collagen
Discussion
Differences in growth, morphology and pathology of tumour and normal mammary epithelial tissues in uiuo are not always apparent
when tumours are disaggregated into monodispersed cells and cultured on plastic (Asch et al., 1981) or in a collagen matrix (Yang et al., 1979; Richards et al., 1983) due to the growth of peritumoural normal tissue, stroma1 tissue and eventual overgrowth with fibroblasts. Lawler et al. (1983) have shown important differences in phenotypic
MAMMARY TUMOUR PHENOTYPES
IN COLLAGEN
GEL
Fig. 8. Phenotypic outgrowths observed with stromal fibroblasts in primary showing bipolar attenuated cells extending through the matrix. x 150.
‘91
collagen
gel
culture
expression between normal and tumour mouse mammary tissue cultured in collagen matrix but did not study correlation with degree of tumour progression. By contrast, the disaggregation technique used here released intact parenchymal tumour or normal epithelial structures (organoids) from mouse mammary glands which were essentially free from fibroblasts and other stromal cell types and established distinctive threedimensional phenotypic structures when cultured in a collagen matrix. Any residual fibroblasts and other stromal material contaminating the organoids in culture were readily removed by treating the gels with collagenase and re-establishing the purified organoids in secondary culture in collagen matrix. Culture of normal mouse mammary tissue in collagen gel is a well-established technique for maintaining morphological differentiation and secretory function (see, for example, Levay-Young et al., 1990). The hydrated matrix allows the expression of
characteristic three-dimensional phenotypic structures and maintains the orientation of epithelial cells within the organoid and their interaction with the substratum as well as interactions between mammary tumours and peritumoural normal tissue (Bano et al., 1984). The distinctive phenotypic structures produced by normal mouse mammary epithelial organoids were similar to those described by Yang et al. (1979) and Guzman et al. (1982). The main phenotype (Pattern A; Fig. la) with a fine needle-like branching structure has been described by Lawler et al. (1983). which they define as Pattern B. The second characteristic phenotype (Pattern A’; Fig. lb), which does not appear to have been described previously, was only observed in the present study with normal and peritumoural normal mammary tissue from pregnant mice. Intermediate phenotypic structures, which were not investigated further, may be characteristic of the mixed populations of cells present (see Slemmer. 1974).
892
A more comprehensive comparison of phenotypes described here with published data is not possible since the concentration of collagen matrix is often not stated, although this can have a profound effect on phenotypic expression by mouse mammary epithelium (Jones and Hosick, 1986). It is of interest that the needle-like phenotype of fibroblasts displayed by material on the 300 micron filter and in the 51 micron filtrate (Fig. 8) was similar to the Pattern A described by Lawler et al. (1983). These results support the view of Yang et al. (1979) that the collagen matrix allows the three-dimensional branching tubular phenotype of mammary epithelial tissue in duo or after implantation into a mammary fat pad (Daniel et al. 1984; Durban et al. 1985) to be expressed in uitro. The importance of interactions between mammary cells in collagen culture is shown by the fact that phenotypic outgrowths produced by sublines of the RAMA series of rat mammary epithelial cells in culture depend on the proportions of epithelial and myoepithelial cells present (Bennett et al. 1981; Ormerod and Rudland, 1982). The predominant formation of a specific phenotypic pattern by normal mammary epithelial tissue from older mice (Lawler et al., 1983) also suggests that changes in the proportions of mammary cell types in uiuo can be expressed in a collagen matrix (Yang et al., 1979; Choongkittaworn et al., 1987). With tumour-derived organoids, phenotypic patterns related uniquely to the type of tumour investigated. Pregnancy-dependent tumours displayed a distinctive outgrowth (Pattern D; Fig. lc) which adhered to the matrix as did outgrowths of Patterns A and A’ produced by normal mammary tissue. By contrast, pregnancy-independent tumours established a unique structure (Pattern C; Fig. 4b, c) which did not adhere to the matrix. Disaggregation of tumour tissues to below a critical size (approximately 100 cells) that passed through the 51 micron filter destroyed this ability to form phenotypic structures in culture. This is in marked contrast to disaggregated normal mammary epithelial tissue of this size which showed a striking ability to reorganise in culture into branching tubular structures of Pattern A. This difference may explain why a specific tumour phenotype pattern may not be apparent in mixed cul-
JONES ET AL.
tures of dispersed tumour and normal mammary epithelial cells (Yang et al. 1979; Richards et al. 1983) and suggests that interaction between tumour and peritumoural normal mammary tissues is required for phenotypic expression of the former in culture (Bano et al. 1984). Similarly, there are differences in vivo between different types of mammary tumours in loss of association between cuboidal epithelial and myoepithelial cells (Slemmer, 1974; Williams and Daniel, 1983). This may explain the difference in culture between the specific Pattern C (see Fig. 4b, c) produced by pregnancy-independent mammary tumours (which are essentially epithelial in structure) and the unique Pattern D (see Fig. lc) produced by pregnancy-dependent mammary tumours (which retain some degree of epithelial-myoepithelial structure). In the present study pregnancy-dependent and pregnancy-independent mammary tumours were maintained in culture for up to 20 days, which enabled phenotypic structures to be established and tumour-derived material to be recovered from the gels to produce tumours when implanted in syngeneic mice. It is of interest that pregnancydependent tumours removed during early pregnancy produced tumour-related phenotypic outgrowths even though the medium was not supplemented with hormones other than those in the foetal calf serum and the weak oestrogen-like activity of the phenol red in the medium (Berthois et al., 1986). By contrast a third pregnancy-dependent mammary tumour removed on the day of littering failed to display tumour-related outgrowths in culture (results not shown). This was due, presumably, to the tumour having started to regress in response to hormonal changes at parturition or to inadequate steroid stimulation during culture as seen, for example, with the S115 mouse mammary tumour cell line (Darbre and King, 1988). Implantation into syngeneic mice of pregnancy-dependent tumour material of Pattern D (Fig. lc) recovered from the gels gave rise to tumours which produced outgrowths in subsequent culture which correlated with the aggressiveness of each of the tumours. The most rapidly growing tumour with the most aberrant histology (tumour 1) displayed outgrowths characteristic of pregnancy-independent tumours (Pattern C; Fig. 4 b, c). By
MAMMARY
TUMOUR
PHENOTYPES
IN COLLAGEN
contrast the slowest growing tumour (tumour 3) produced the Pattern D of the original pregnancy-dependent tumour. Tumour 2, which was well differentiated, also resembled the original pregnancy-dependent tumour in producing outgrowths of Pattern D, but tumour 2 also produced a proportion of outgrowths of Pattern D characteristic of pregnancy-independent tumours. These results contrast with those of Yang et al. (1979) where in vivo transplantation of outgrowths on collagen gel resulted in mammary adenocarcinomas similar histologically to the donor tumour. The change in tumour phenotype from hormone-dependent to hormone-independent (‘tumour progression’) described here has also been reported for mammary tumours of the GR strain of mouse in vivo (Sluyser et al., 1980; Kiang et al., 1982) and is shown by many murine mammary tumours in vivo.
GEL
x93
There is also a higher incidence of metastasis associated with this change to the less differentiated state with mammary tumours of the BR6/Icrf mouse (Lee et al., 1983). In conclusion, the results demonstrate that the culture system described can be used to correlate the pathobiology of mouse mammary tumour progression in vivo with their phenotypic patterns of outgrowth in vitro. Acknowledgements We are indebted to Dr L. S. C. Pang for her help and thank Mr L. A. Rogers, Mrs Elizabeth Parsons and Mrs Carol Gomm of the Imperial Cancer Research Fund for their technical assistance and the Photographic Unit of Glaxo Group Research Limited for the photographic materials. W.J. was supported by a Studentship from the Imperial Cancer Research Fund during this work.
References Asch, B. M., Kamat, B. R. and Burstein, N. A. 1981. Interactions of normal, dysplastic, and malignant mammary epithelial cells with fibronectin in viva and in uirro Cancer Res., 41, 2115-2125. Bane, M., Lewko, W. M. and Kidwell, W. R. 1984. Characterization of rat mammary tumour cell populations. Cancer Res., 44, 3055-3062. Bennett, D. C.. Armstrong, B. L. and Okada, S. M. 1981. Reconstitution of branchmg tubules from two cloned mammary cell types in culture. Deu. Biol., 87, 19f199. Berthois, Y., Katzenellenbogen, J. A. and Katzenellenbogen, B. S. 1986. Phenol red in tissue culture media is a weak estrogen: Implications concerning the study of estrogen-responsive cells in culture. Proc. Nar[. Acad. Sci USA. 83, 24962500. Choongkittaworn. N., Hosick, H. L. and Jones, W. 1987. In oifro replication potential of serial passaged mammary parenchyma from mice with different reproductive histories. Mechanisms of Aging Deuelopment, 39, 147-175. Daniel, C. W., Berger, J. J., Strickland, P. and Garcia, R. 1984. Similar growth pattern of mouse mammary cpithclium cultivated in collagen matrix in uiuo. and in uifro. Deu. Biol., 104, 57-64. Darbre, P. D. and King. R. J. B. 1988. Steroid hormone regulation of cultured breast cancer cells. In Breast Chnwr Celhdarand Molecular Biobgy (edsM. E. Lippman and R. B. Dickson). pp.307-341. Klwer Academic Puhhshcr\. Boston. Durban, E. M.. Medina, D. and Butel, J. S. 1985. Comparative analysis of casein synthesis during mammary XII differentiation in collagen and mammary gland development in uioo. Deo. Biol., 109, 28b298. Guzman, R. C.. Osborn, R. C., Yang, J., De Ome, K. B. and Nandi. S. 1982. Transplantation of mouse mammar) epithelial cells grown in primary collagen gel cultures. Cancer Res., 42, 23762383. Hallowes, R. C., Bone, E. J. and Jones, W. 1980. A new dimension in the culture of human breast. In Tissue (‘alrare in Medical Research II (eds R. J. Richards and K. T. Rajan), pp. 213-220. Pergamon Press, Oxford. Jones, W., Hallowes, R. C., Choongkittaworn, N., Hosick, H. and Dils, R. R. 1983. Isolation of the epithelial subcomponents of the mouse mammary gland for tissue-level culture studies. J. Tissue Culture MethodJ, 8. I7 25. Jones, W. and Hosick, H. L. 1986. Collagen concentration as a significant variable for growth and morphology of mouse mammary parenchyme in collagen matrix culture. Cell Biol. Int. Reports. 10, 277-286. Kiang, D. T., King, M., Zhang, H. J., Kennedy, B. J. and Wang, N. 1982. Cyclic biological expression in mouse mammary tumours. Science, 216, 68-70. Lawler, E. M., Miller, F. R. and Heppner, G. H. 1983. Significance of three-dimensional growth patterns ol mammarl tissues in collagen gels. In Vitro, 19, 600-610. Lee, A. E. 1970. Mammary turnour development in BR6 mice: hormonal stimulation. Br. J. Cancer. 24. 56%573
894
JONES
ET AL.
Lee, A. E., Pang, L. S. C., Rogers, L. A. and Jeffery, R. E. 1983. Metastasis of mammary turnours in mice: relationship with morphology of primary turnour, and reproductive background of host. Clin. Exp. Metastasis, 1, 223221. Levay-Young, B. K., Hamamoto, S., Imagawa, W. and Nandi, S. 1990. Casein accumulation in mouse mammary epithelial cells after growth stimulated by different hormonal and nonhormonal agents. Endocrinol., 126, 11731182. Ormerod, E. J. and Rudland, P. S. 1982. Mammary gland morphogenesis in uitro: formation of branched tubules in collagen gel by a cloned rat mammary cell line. Deu. Biol., 91, 360-375. Richards, J., Pasco, D., Yang, J., Guzman, R. and Nandi, S. 1983. Comparison of the growth of normal and neoplastic mouse mammary cells on plastic, on collagen and in collagen gels. Exp. Ce[l Res., 146, l-14. Riddle, P. N. 1990. Time-lapse cinemicroscopy. Methods in Molecular Biology, 5, 415-446. Slemmer, G. 1974. Interactions of separate types of cells during normal and neoplastic mammary gland growth. J. Invest. Dermnt., 63, 27-41. Sluyser, M., van der Valk, M. and Blitterswijk, W. J. 1980. Changes in concanavalin A mediated agglutination of hormone-dependent mouse mammary tumour cells during serial transplantation. Br. J. Cancer, 41, 341355. Stampfer. M., Hallowes, R. C. and Hackett, A. J. 1980. Growth of normal human mammary cells in culture. In Virro. 16, 415-425. Williams, J.M. and Daniel, C. W. 1983. Mammary duct elongation: differentiation of myoepithelium and basal lamina during branching morphogenesis. Deu. Biol., 97, 274-290. Yang, J., Richards, J., Bowman, P., Guzman, R., Enami, J., McCormick, K., Hamomoto, S., Pitelka, D. and Nandi, S. 1979. Sustained growth and three-dimensional organization of primary turnour epithelial ceils embedded in collagen gels. Proc. Natl. Acad. Sci. USA, 76, 3401-3405.
W. JONES*,
A. E. LEE*, P. RIDDLE*, R. R. DlLSt and R. C. HALLOWES”
PHENOTYPIC RESPONSES OF MOUSE MAMMARY TUMOURS AND NORMAL MAMMARY EPITHELIUM CULTURED IN COLLAGEN GELS: CORRELATION WITH TUMOUR TYPE AND PROGRESSION Keywords:
Mouse,
mammary
epithelium,
tumour,
phenotype,
collagen
gel
ABSTRACT. Mammary turnours in female BR6/Icrf mice and the corresponding contralatcral normal mammary glands were disaggregated with collagenase and the epithelial structures released (‘organoids’) separated from other cellular components by filtration. The organoidy were established in primary culture in a collagen matrix and the outgrowths obtained were studied by light microscopy and time-lapse cinemicroscopy. The pattern of three-dimensional outgrowths produced was found to be specific to the original tissue. Organoids from normal tissue formed a characteristic outgrowth designated Pattern A. Normal tissue from pregnant mice formed an additional characteristic outgrowth (Pattern A’) which has not been described previously. Pregnancy-dependent turnours produced a distinctive phenotypic outgrowth de\ignated Pattern D, whereas pregnancy-independent tumours gave a different distinctive Pattern B as well as a unique specific outgrowth designated Pattern C. Outgrowths of Pattern D from a pregnancy-dependent tumour were removed from culture and implanted into a syngeneic female mouse. Tumours arising in the host were found to be pregnancy-independent and showed phenotypic outgrowths in subsequent culture of pregnancy-independent Patterns B and C. The results show that the type of outgrowths in these cultures correlates with the biology of the tissue in uiuo and that changes in tumour progression in uivo are accompanied by alterations in phenotypic outgrowths in culture.
however in the phenotypic expression of a number of established mouse mammary lines and mouse mammary tissue when cultured in a collagen matrix (Lawler et al., 1983). although the interpretation of these results is complicated by the presence of stromal cells in the tissue fragments. The BR6/Icrf mouse has a relatively high incidence of hormone-dependent mammary tumours during pregnancy (Lee, 1970). These tumours regress at parturition and continue as hormone-dependent tumours for a number of pregnancies although most progress eventually to autonomously-growing tumours. Since each stage of this tumour progression is marked by distinctive pathological features we have used the tumourcontaining mammary gland of the BR6/Icrf mouse as a model to study neoplastic progression. A tissue disaggregation procedure
Introduction
Well-defined changes in cell structure and function accompany neoplastic progression. Whereas differences between populations of normal and tumour mammary epithelial cells persist in organ culture, studies with dispersed normal and tumour mammary cells cultured in a collagen matrix have failed to demonstrate differences in phenotypic behaviour (Yang et al., 1979; Richards et al., 1983). Differences have been described * Imperial Cancer Research Fund, Lincoln’s Inn Fields, London WCZA 3PX, UK and tDepartment of Biochemistry and Physiology, School of Animal and Microbial Sciences, University of Reading, P.O. Box 228, Whiteknights, Reading, RG6 2AJ. UK. Correspondence to: W. Jones, Department of Molecular Science, Glaxo Group Research Limited, Greenford Road, Greenford, Middlesex. UB6 OHE, UK. Received
21 April
1992 879
JONES ET AL.
880
(Jones et al., 1983) has been used which preserves the structural integrity of mammary parenchymal units whilst removing virtually all stromal components including fibroblasts. The defined populations of epithelial components produced (i.e. organoids) were then cultured in a collagen matrix to investigate differences in the patterns of phenotypic behaviour of atypic and neoplastic mammary epithelia compared with control mammary epithelia. We report that primary mammary tumours of the BR6/Icrf mouse cultured in a collagen matrix display characteristic phenotypic patterns of outgrowth which differ from the phenotypic outgrowths observed with normal mammary tissue and which can be used to determine the pathobiology of the mammary tumours.
Materials and Methods Mammary tumours and control tissue The mammary gland containing a tumour in the left vulva1 region which had shown hormone dependence during the previous pregnancy was removed from two 8- to 9-day pregnant BR6/Icrf mice (designated tumour A and tumour B). For comparison, pregnancy-independent tumours in the mammary fat pad of the left vulva1 region were removed from two virgin female BR6/Icrf mice (designated tumours C and D). With each of the four mice, the complimentary non-involved mammary gland was used as the control. Tissue disaggregation
Each mammary gland was disaggregated by limited digestion with collagenase as described by Jones et al. (1983). The epithelial structures released (organoids) were washed for three 30-min periods on a vertical rotary mixer at 37°C in Medium RPMI-1640 supplemented with 5% (v/v) foetal calf serum (Gibco) and filtered sequentially through 300, 95 and 51 micron pore-size Nylon monofilament cloth (H. Simon, Stockport, Cheshire). The organoid fractions collected on each filter, as well as the cellular material in the filtrate from the 51 micron filter, were suspended in warm Medium RPMI-1640 supplemented with 20% (v/v) foetal calf serum and lOpg/ml insulin (Sigma) and cultured separately in a collagen gel matrix.
Primary culture in collagen matrix
Each organoid fraction as well as the cellular material in the filtrate from the 51 micron filter was separately pelleted and mixed at 4°C with rat tail collagen at 2 mg/ml (Hallowes et al., 1980; Jones and Hosick, 1986). Portions (1 ml) of the mixture were layered over preformed collagen gel (0.5ml) in 35 mm diameter culture dishes and the dishes warmed to 37°C for 10min for the collagen gel to set. The cultures were overlayed with 2.0 ml of Medium RPMI-1640 supplemented with 5% (v/v) foetal calf serum and insulin (10 pg/ml) and maintained in a humidified atmosphere of 10% CO2 in air at 37°C for up to 20 days. The medium was changed after 24 hr and subsequently at 48-hr intervals. Secondary culture in collagen matrix
Once the characteristic phenotypic outgrowths from organoids had been established during primary culture in the collagen matrix, the individual gels were digested for 30 min at 37°C with 20ml of collagenase (Type lA; 50 u/ml; Sigma) in Medium RPMI-1640 supplemented with 10% (v/v) foetal calf serum on a vertical rotary mixer. The digested material was then allowed to settle under gravity for 10 min and the supernatant discarded. The sediment was washed for three 30-min periods in Medium RPMI-1640 supplemented with 10% (v/v) foetal calf serum at room temperature on a vertical rotary mixer. After each washing, the sediment was allowed to settle under gravity and the supernatant discarded. The final sediment was mixed with rat tail collagen at 4°C and again cultured in collagen gel as described above for primary culture. Implantation in vivo of specific phenotypic outgrowths and subsequent culture in collagen gel of the mammary tumours that developed
Once the characteristic phenotypic outgrowths from organoids prepared from tumour A had been established in primary culture, specific phenotypic outgrowths were removed individually from the gel under a dissecting microscope by cutting the gel around the structures with a pair of l&gauge hypodermic needles attached to 1 ml syringes. Fragments of collagen devoid of mammary structures were removed as controls.
D
tumour
t Normal
C
tumour
phenotype
Pregnancy-independent
patterns,
representing
Pattern
Pattern
Pattern
t Pattern
::::::::,>
Pregnancy-dependent tumour B
b{
pcri-tumoral
C
B
A
D
D
Phenotypic outgrowths in primary culture
and summary of results
b Pattern
epithelium
1. Design of experiment
Pregnancy-dependent tumour A
Normal
Tissue
Table
t
normal
2
1
from culture
2: tumour
1: tumour
selected
Mouse
Mouse
::::::1::::::
tissue,
*{
After implanting into pairs of syngeneic mice
of organoids
C
B
A
C
D
from pregnancy-dependent
* Pattern
* Pattern
+ Pattern
*Pattern
rPattern
Phenotypic outgrowths in secondary culture
tumour
A.
developed
b tumours
Mammary
After implanting into pairs of syngenic mice
?J
Y
z
8 F
z
E
!? 0 ::
z
JONES ET AL.
The samples were implanted individually into the right flank of pairs of syngeneic nonpregnant host female BR6/Icrf mice which were palpated regularly for evidence of tumour formation in the implanted areas. Mice which developed tumours in the implanted areas were killed, and the mammary tumours (tumours l-3) were disaggregated and cultured in collagen gel as described above for the original tumours. A small piece of each tumour was processed for histological examination. Htitology
A fragment of each of the primary mammary tumours (tumours A-D), as well as of tumours l-3 arising from implantation of specific phenotypic outgrowths, was removed and fixed in formol-saline, dehydrated through a graded series of alcohols and embedded in paraffin wax. Sections (5 microns) were stained with haematoxylin and eosin and examined by light microscopy. Time-lapse cinemicroscopy
Representative cultures were filmed by timelapse cinemicroscopy using 2-set exposures at intervals of 4 min (Riddle, 1990). Tissue-matrix adhesion
Adhesion of mammary outgrowths to the collagen matrix was estimated by disrupting the gel in the vicinity of the structures with a pair of hypodermic needles. A dissecting microscope was used to observe whether the structures were released intact and free from adhering collagen, or were disrupted by the procedure. Results For convenience, Table 1 outlines the design of the experiments and summarises the results obtained. A. Phenotypic outgrowths of normal mammary epithelium
Organoids derived from preparations of normal and peritumoural mammary epithelium tissue produced three-dimensional branching tubular structures after 24 hr in culture. The prevalent phenotype (Pattern A; Fig. la) (see Jones and Hosick, 1986) consisted of fine pointed multicellular tubules protruding into the gel which became disorganised after 8 days in culture. The filtrate from the 51
micron filter, which contained very small clusters of cells and monodispersed cells, demonstrated a high degree of reorganisation of these cells in culture and produced phenotypic pattern A after about 48 hr. The second phenotypic pattern of outgrowth (Pattern A’; Fig. lb) which was only observed with normal and peritumoural tissue from pregnant mice consisted of a few broad, blunt projections extending from the organoids into the collagen matrix which persisted for up to 20 days in culture. A small proportion of the organoids showed growth patterns intermediate between Patterns A and A’ but these were not distinct enough to enable a separate phenotype pattern to be assigned. B. Pregnancy-dependent
tumours
Histological examination showed that the pregnancy-dependent tumours were well differentiated and had maintained the ductallobular organisation of the normal differentiated mammary gland in pregnancy although there was some epithelial hyperplasia (Fig.2a). Connective tissue stroma surrounded the epithelial units and the tumours were not invasive. There was some secretory material in the acini but much less than in the distended lumina of the acini of the complementary control gland (Fig. 2b). Multicellular epithelial organoids of various sizes were released from the tumour tissue on disaggregation. Although filtration efficiently separated these structures by size, it was not possible to distinguish between organoids from tumour and peritumoural normal tissue by shape or size. The filtrate from the 51 micron filter consisted mainly of monodispersed and small clusters of epithelial cells and stromal fibroblasts. Hence this filtration procedure resulted in the organoids collected on the 95 and 51 micron filters consisting predominantly of epithelial structures essentially devoid of contamination with fibroblasts and other stromal components. Phenotypic outgrowths of pregnancy-dependent tumours in primary culture. Limited
three-dimensional outgrowths from tumourderived organoids were observed after 24 hr, but phenotypic differences between tumour and complementary normal tissue did not become apparent until after 4 days.
MAMMARY
TUMOUR
PHENOTYPES
IN COLLAGEN
GEL
883
Fig. 1. Phenotypic patterns of three-dimensional outgrowths established in primary collagen culture by mammary organoids. (a) Characteristic fine needle-like structures (Pattern A) observed with normal and peritumoural normal tissue; (b) amoeboid-like structures (Pattern A’) only observed with normal and peritumoural tissue from pregnant mice; (c) lichen-like wwtures (Pattern D) only observed with pregnancy-dependent turnours. The cells are spread as a sheet at the interface of the upper and lower layers of the collagen gel (see arrowheads): (d) spherical multicellular structures (see arrowheads) only observed with cellular material from the 51 micron filtrate of disaggregated pregnancy-dependent turnours. Fig. la x 130; Fig. lb x260; Fig. lc x130; Fig. Id x260 (phase contrastl.
Although ‘normal’ Patterns A and A’ of phenotype outgrowth (see Fig. la, b) occurred with organoids from the pregnancydependent tumour tissue, this was probably due to the presence of peritumoural normal tissue. In addition, a unique phenotype (Pattern D) with a lichen-like appearance (Fig. lc) was observed with all cultures of organoids prepared from the pregnancydependent turnours. This structure increased in size throughout the 20 days culture due, in part, to distension of the lumen. The filtrate from the 51 micron filter produced small
rounded clusters of cells by day 4 in culture which had a rather dark appearance under phase-contrast (Fig. Id). Tumourigenicpotential of specificphenotypic outgrowths in primary collagen gei after implantation into syngeneic mice. The phenotypic patterns of outgrowth established by organoids derived from tumour A were analysed further as follows. Phenotypic patterns of outgrowths common to both tumour and non-involved glands (Pattern A, Fig. la and Pattern A’, Fig. lb) and of outgrowths uni-
JONES ET AL.
884
que to the primary mammary tumours (Pattern D, Fig. lc) were dissected from the gels as well as areas of tissue-free gel as controls. Samples of each type of outgrowth were then implanted into pairs of syngeneic non-pregnant female BR6/Icrf mice. The two mice implanted with tissue that displayed phenotype Pattern D in primary culture developed palpable tumours at day 21 (mouse 1) and at day 32 (mouse 2) after implantation, i.e. these tumours arose independently of pregnancy. Mouse 1 was then mated at day 27 post-implantation and littered 21 days later. The mammary tumour carried by this mouse (tumour 1) did not regress after parturition and was removed at day 53 post-implantation. Mouse 2 was not mated and the mammary tumour (tumour 2) was removed on day 56 post-implantation. Both of these tumours, which were classified histologically as Type B mammary adenocarcinomas had maintained a degree of tissue organisation (Fig. 3a, b) despite gross hyperplasia and dysplasia (see Lee et al., 1983). This loss of pregnancy-dependence, together with the increasing degree of tissue disorganisation observed by histological examination, indicated that both tumours had progressed to a more autonomous form. One of the pair of mice implanted with fragments of tissue from control animals that displayed the phenotypic outgrowths common to both tumour and normal tissue (Patterns A and A’) in primary culture developed a tumour (tumour 3; Fig. 3c) 28 days after implantation. This tumour grew more slowly in vivo than tumours 2 and 3 and, when removed on day 75 post-implantation, had a very necrotic core. Neither the other mouse of this pair nor the pair of mice implanted with organoid- free areas of collagen gel produced tumours during the 103 days after implantation that they were observed. Phenotypic outgrowths in collagen gel of tumours l-3. Tumours l-3 were disag-
gregated and the organoids were recultured separately in collagen gels as described previously. Tumour 1, derived originally from material showing phenotype Pattern D in primary culture, gave rise within 12 days to outgrowths mainly displaying phenotype Pattern C in collagen gel culture. This pattern of outgrowth (see Fig. 4 b, c) is characteristic of
pregnancy-independent mouse mammary tumours in culture (see below). The main phenotype observed with organoids prepared from tumour 2 were amoeboid shaped and consisted of a large rounded body with up to 5 large multicellular protrusions with rounded tips (Pattern B; Fig. 4a). Tumour 2 also gave rise to a small proportion of the lichen-like phenotype Pattern D (Fig. 5) seen with the original tumour organoids in primary culture (cf. Fig. lc). A further proportion of organoids displayed sheet-like spreading of cells (data not shown) which was also seen with some of the organoids displaying phenotype Pattern D. This represents the growth of a sheet of cells along the interface between the upper and lower layers of the collagen gel where the organoids had settled. These phenotypic outgrowths were fully established within 9 days in culture. Tumour 3, which arose from implanted material displaying the ‘normal’ phenotype Patterns A and A’ in culture, gave rise predominantly to the lichen-like Pattern D associated with outgrowths in culture from pregnancy-dependent tumours (Fig. lc). Again this phenotypic outgrowth was fully established within 9 days in culture. With each of these tumours a proportion of the organoids displayed phenotypes in culture which were not classified further. Tumourigenicity of outgrowths in culture from tumour 2. Phenotypic outgrowths (Pat-
terns B and D) from cultures of organoids derived from tumour 2, as well as the remaining fragments of indeterminate phenotype, were dissected from the collagen gels and each implanted into pairs of syngenic mice as described previously. Mammary tumours developed in all of these mice within 59 days of implantation. Gross metastatic nodules were present in the lungs of the pair of mice implanted with material of phenotype Patterns B or D. One of the mice implanted with fragments of material of indeterminate phenotype showed micro-metastases in the lungs. C. Pregnancy-independent Histological examination that both tumour C and typical of adenocarcinomas this strain of mouse, i.e.,
tumours
(Fig. 6) showed tumour D were associated with they showed no
MAMMARY TUMOUR PHENOTYPES
IN COLLAGEN
GEL
Fig. 2. Histological appearance of (a) pregnancy-dependent tumour in the mammary gland showing maintained lobule-alveolar development but varying degrees of epithelial hyperplasia and less secretory material than in (b); (b) contralateral normal gland at day 9 of pregnancy showing extensive ductal lobule-alveolar development and dilated lumen containing secretory material x72.
885
886
JONES ET Al
MAMMARY
TUMOUR
PHENOTYPES
IN COLLAGEN
GEL
X87
888
JONES ET AL.
MAMMARY
TLMOUR
PHENOTYPES
IN COLLAGEN
evidence of the ductal-lobular organisation of the normal gland, the stroma was sparce and there was no evidence of necrosis. The complimentary control mammary gland of these mice showed no obvious pathological changes. Phenotype
outgrowths
in primary
culture.
After 2 days in culture the tumour-derived organoids from each of the three filters showed detectable outgrowths of both Pattern A (Fig. la) and Pattern B (Fig. 4a), as well as the unique Pattern C (see Fig. 4 b, C). The tumour-derived filtrate from the 51 micron filter produced a distinctive phenotype pattern consisting of a few small refractive clusters of cells which persisted throughout the 16day culture period. The development of the unique phenotype Pattern C was analysed by time-lapse cinemicroscopy (Fig. 7) for a period of 1.50hr. Representative frames from the film showed that the main body of the cluster of tumour cells maintained its spherical shape with minimum interaction with the collagen matrix. An occasional columnar protusion arose from the surface of the structure and extended some distance into the matrix as a stalk with a rounded tip (Fig. 7f) which usually regressed with time. The tips swelled and pinched off by budding, forming satellite spheres in the vicinity of the parent organoid
GEL
WY
(Fig. 7i) which appeared to consist of single cells or clusters of cells. Phenotypic outgrowth patterns in secondar? and tertiary culture. Multicellular structures
displaying normal phenotype outgrowths of Pattern A (Fig. la) or tumour-derived Patterns B or C (Fig. 4) in primary culture were released intact from the gels by digestion with collagenase. These structures were settled under gravity and the supernatant, which contained contaminating fibroblasts and monodispersed and small clusters of stromal cells, was discarded. The multicellular structures were re-established in collagen gel and secondary culture continued for 18 days. The gels were again treated with collagenase, contaminating fibroblasts and stromal cells removed as described previously and the multicellular structures again established in collagen gel (tertiary culture). After 3 days in secondary or in tertiary culture, each preparation had established its characteristic phenotype of outgrowth, i.e.. Pattern A, B or C. Tissue-matrix adhesion. Microdissection at the end of culture released structures of Pattern C intact and essentially free from collagen gel. By contrast, outgrowths with ‘normal’ phenotype Pattern A or with tumour-related phenotype Patterns B or D were disrupted by this procedure indicating
Fig. 3. Histological appearance of tumours arising in syngeneic host female mice after implantation with phenotypic outgrowths established in primary collagen gel from tumourderived organoids. (a) Tumour 1 tissue 53 days after implantation with outgrowth of Pattern D: (b) tumour 2 tissue 56 days after implantation with outgrowth of Pattern D; (c) tumour 3 tissue 75 days after implantation with outgrowth (Patterns A and A’) common to both tumour and normal mammary tissue. Tumours 1-3, which had lost pregnancy-dependence, rctaincd some degree of lobulo-alveolar and ductal organisation despite gross hyperplasia and dysplasia. x72. Fig. 4. Phenotypic patterns of outgrowth established in primary collagen culture by mammary organoids from pregnancy-independent mammary turnours. (a) Pattern B showing multicellular clusters of tumour cells with a few short. conical protrusions extending into the surrounding collagen matrix; (b,c) pattern C showing multicellular clusters of tumour cells wtth short. blunt stalks (b, arrowhead 1). The ends of these stalks swell up (b, arrowhead 2) and then bud off leaving behind satellite spheres (c. arrowhead 3) composed mainly of smgle cells x280 Fig. 5. Phenotypic pattern of outgrowth orpanoids derived from tumour 2. x200.
of Pattern
D established
in collagen
gel culture by
Fig. 6. Histological appearance of pregnancy-independent mammary tumour tissue showing no evidence of the ductal-lobular organisation of the normal contra-lateral mammary tissue, sparse stroma and no evidence of necrosis. x72.
:““.‘:
b
I
bl I[ : :
...
0
: .
f
...
: :
:
.. . ....
I
.0. I”‘i
C
. ... . .. : f : :
:i
.
‘b
7
“““:
V
i
. ..!?...... i .,.......
2i
Fig. 7. Time-lapse cinemicroscopy sequence of development of Pattern C phenotypic outgrowths by pregnancy-independent tumour organoids in primary collagen gel culture. Time (hr) from start of culture: (a) 0; (b) 21; (c) 37; (d) 41; (e) 68; (f) 85; (g) 108; (h) 132 and (i) 150. Frames a-d show three-dimensional protrusions extending into the collagen matrix. Frames e-i show the compaction of the organoid and withdrawal of cellular protrusions leaving satellite structures of one or more cells, as indicated by the dotted areas. x 116.
a high degree of adhesion matrix.
to the collagen
Discussion
Differences in growth, morphology and pathology of tumour and normal mammary epithelial tissues in uiuo are not always apparent
when tumours are disaggregated into monodispersed cells and cultured on plastic (Asch et al., 1981) or in a collagen matrix (Yang et al., 1979; Richards et al., 1983) due to the growth of peritumoural normal tissue, stroma1 tissue and eventual overgrowth with fibroblasts. Lawler et al. (1983) have shown important differences in phenotypic
MAMMARY TUMOUR PHENOTYPES
IN COLLAGEN
GEL
Fig. 8. Phenotypic outgrowths observed with stromal fibroblasts in primary showing bipolar attenuated cells extending through the matrix. x 150.
‘91
collagen
gel
culture
expression between normal and tumour mouse mammary tissue cultured in collagen matrix but did not study correlation with degree of tumour progression. By contrast, the disaggregation technique used here released intact parenchymal tumour or normal epithelial structures (organoids) from mouse mammary glands which were essentially free from fibroblasts and other stromal cell types and established distinctive threedimensional phenotypic structures when cultured in a collagen matrix. Any residual fibroblasts and other stromal material contaminating the organoids in culture were readily removed by treating the gels with collagenase and re-establishing the purified organoids in secondary culture in collagen matrix. Culture of normal mouse mammary tissue in collagen gel is a well-established technique for maintaining morphological differentiation and secretory function (see, for example, Levay-Young et al., 1990). The hydrated matrix allows the expression of
characteristic three-dimensional phenotypic structures and maintains the orientation of epithelial cells within the organoid and their interaction with the substratum as well as interactions between mammary tumours and peritumoural normal tissue (Bano et al., 1984). The distinctive phenotypic structures produced by normal mouse mammary epithelial organoids were similar to those described by Yang et al. (1979) and Guzman et al. (1982). The main phenotype (Pattern A; Fig. la) with a fine needle-like branching structure has been described by Lawler et al. (1983). which they define as Pattern B. The second characteristic phenotype (Pattern A’; Fig. lb), which does not appear to have been described previously, was only observed in the present study with normal and peritumoural normal mammary tissue from pregnant mice. Intermediate phenotypic structures, which were not investigated further, may be characteristic of the mixed populations of cells present (see Slemmer. 1974).
892
A more comprehensive comparison of phenotypes described here with published data is not possible since the concentration of collagen matrix is often not stated, although this can have a profound effect on phenotypic expression by mouse mammary epithelium (Jones and Hosick, 1986). It is of interest that the needle-like phenotype of fibroblasts displayed by material on the 300 micron filter and in the 51 micron filtrate (Fig. 8) was similar to the Pattern A described by Lawler et al. (1983). These results support the view of Yang et al. (1979) that the collagen matrix allows the three-dimensional branching tubular phenotype of mammary epithelial tissue in duo or after implantation into a mammary fat pad (Daniel et al. 1984; Durban et al. 1985) to be expressed in uitro. The importance of interactions between mammary cells in collagen culture is shown by the fact that phenotypic outgrowths produced by sublines of the RAMA series of rat mammary epithelial cells in culture depend on the proportions of epithelial and myoepithelial cells present (Bennett et al. 1981; Ormerod and Rudland, 1982). The predominant formation of a specific phenotypic pattern by normal mammary epithelial tissue from older mice (Lawler et al., 1983) also suggests that changes in the proportions of mammary cell types in uiuo can be expressed in a collagen matrix (Yang et al., 1979; Choongkittaworn et al., 1987). With tumour-derived organoids, phenotypic patterns related uniquely to the type of tumour investigated. Pregnancy-dependent tumours displayed a distinctive outgrowth (Pattern D; Fig. lc) which adhered to the matrix as did outgrowths of Patterns A and A’ produced by normal mammary tissue. By contrast, pregnancy-independent tumours established a unique structure (Pattern C; Fig. 4b, c) which did not adhere to the matrix. Disaggregation of tumour tissues to below a critical size (approximately 100 cells) that passed through the 51 micron filter destroyed this ability to form phenotypic structures in culture. This is in marked contrast to disaggregated normal mammary epithelial tissue of this size which showed a striking ability to reorganise in culture into branching tubular structures of Pattern A. This difference may explain why a specific tumour phenotype pattern may not be apparent in mixed cul-
JONES ET AL.
tures of dispersed tumour and normal mammary epithelial cells (Yang et al. 1979; Richards et al. 1983) and suggests that interaction between tumour and peritumoural normal mammary tissues is required for phenotypic expression of the former in culture (Bano et al. 1984). Similarly, there are differences in vivo between different types of mammary tumours in loss of association between cuboidal epithelial and myoepithelial cells (Slemmer, 1974; Williams and Daniel, 1983). This may explain the difference in culture between the specific Pattern C (see Fig. 4b, c) produced by pregnancy-independent mammary tumours (which are essentially epithelial in structure) and the unique Pattern D (see Fig. lc) produced by pregnancy-dependent mammary tumours (which retain some degree of epithelial-myoepithelial structure). In the present study pregnancy-dependent and pregnancy-independent mammary tumours were maintained in culture for up to 20 days, which enabled phenotypic structures to be established and tumour-derived material to be recovered from the gels to produce tumours when implanted in syngeneic mice. It is of interest that pregnancydependent tumours removed during early pregnancy produced tumour-related phenotypic outgrowths even though the medium was not supplemented with hormones other than those in the foetal calf serum and the weak oestrogen-like activity of the phenol red in the medium (Berthois et al., 1986). By contrast a third pregnancy-dependent mammary tumour removed on the day of littering failed to display tumour-related outgrowths in culture (results not shown). This was due, presumably, to the tumour having started to regress in response to hormonal changes at parturition or to inadequate steroid stimulation during culture as seen, for example, with the S115 mouse mammary tumour cell line (Darbre and King, 1988). Implantation into syngeneic mice of pregnancy-dependent tumour material of Pattern D (Fig. lc) recovered from the gels gave rise to tumours which produced outgrowths in subsequent culture which correlated with the aggressiveness of each of the tumours. The most rapidly growing tumour with the most aberrant histology (tumour 1) displayed outgrowths characteristic of pregnancy-independent tumours (Pattern C; Fig. 4 b, c). By
MAMMARY
TUMOUR
PHENOTYPES
IN COLLAGEN
contrast the slowest growing tumour (tumour 3) produced the Pattern D of the original pregnancy-dependent tumour. Tumour 2, which was well differentiated, also resembled the original pregnancy-dependent tumour in producing outgrowths of Pattern D, but tumour 2 also produced a proportion of outgrowths of Pattern D characteristic of pregnancy-independent tumours. These results contrast with those of Yang et al. (1979) where in vivo transplantation of outgrowths on collagen gel resulted in mammary adenocarcinomas similar histologically to the donor tumour. The change in tumour phenotype from hormone-dependent to hormone-independent (‘tumour progression’) described here has also been reported for mammary tumours of the GR strain of mouse in vivo (Sluyser et al., 1980; Kiang et al., 1982) and is shown by many murine mammary tumours in vivo.
GEL
x93
There is also a higher incidence of metastasis associated with this change to the less differentiated state with mammary tumours of the BR6/Icrf mouse (Lee et al., 1983). In conclusion, the results demonstrate that the culture system described can be used to correlate the pathobiology of mouse mammary tumour progression in vivo with their phenotypic patterns of outgrowth in vitro. Acknowledgements We are indebted to Dr L. S. C. Pang for her help and thank Mr L. A. Rogers, Mrs Elizabeth Parsons and Mrs Carol Gomm of the Imperial Cancer Research Fund for their technical assistance and the Photographic Unit of Glaxo Group Research Limited for the photographic materials. W.J. was supported by a Studentship from the Imperial Cancer Research Fund during this work.
References Asch, B. M., Kamat, B. R. and Burstein, N. A. 1981. Interactions of normal, dysplastic, and malignant mammary epithelial cells with fibronectin in viva and in uirro Cancer Res., 41, 2115-2125. Bane, M., Lewko, W. M. and Kidwell, W. R. 1984. Characterization of rat mammary tumour cell populations. Cancer Res., 44, 3055-3062. Bennett, D. C.. Armstrong, B. L. and Okada, S. M. 1981. Reconstitution of branchmg tubules from two cloned mammary cell types in culture. Deu. Biol., 87, 19f199. Berthois, Y., Katzenellenbogen, J. A. and Katzenellenbogen, B. S. 1986. Phenol red in tissue culture media is a weak estrogen: Implications concerning the study of estrogen-responsive cells in culture. Proc. Nar[. Acad. Sci USA. 83, 24962500. Choongkittaworn. N., Hosick, H. L. and Jones, W. 1987. In oifro replication potential of serial passaged mammary parenchyma from mice with different reproductive histories. Mechanisms of Aging Deuelopment, 39, 147-175. Daniel, C. W., Berger, J. J., Strickland, P. and Garcia, R. 1984. Similar growth pattern of mouse mammary cpithclium cultivated in collagen matrix in uiuo. and in uifro. Deu. Biol., 104, 57-64. Darbre, P. D. and King. R. J. B. 1988. Steroid hormone regulation of cultured breast cancer cells. In Breast Chnwr Celhdarand Molecular Biobgy (edsM. E. Lippman and R. B. Dickson). pp.307-341. Klwer Academic Puhhshcr\. Boston. Durban, E. M.. Medina, D. and Butel, J. S. 1985. Comparative analysis of casein synthesis during mammary XII differentiation in collagen and mammary gland development in uioo. Deo. Biol., 109, 28b298. Guzman, R. C.. Osborn, R. C., Yang, J., De Ome, K. B. and Nandi. S. 1982. Transplantation of mouse mammar) epithelial cells grown in primary collagen gel cultures. Cancer Res., 42, 23762383. Hallowes, R. C., Bone, E. J. and Jones, W. 1980. A new dimension in the culture of human breast. In Tissue (‘alrare in Medical Research II (eds R. J. Richards and K. T. Rajan), pp. 213-220. Pergamon Press, Oxford. Jones, W., Hallowes, R. C., Choongkittaworn, N., Hosick, H. and Dils, R. R. 1983. Isolation of the epithelial subcomponents of the mouse mammary gland for tissue-level culture studies. J. Tissue Culture MethodJ, 8. I7 25. Jones, W. and Hosick, H. L. 1986. Collagen concentration as a significant variable for growth and morphology of mouse mammary parenchyme in collagen matrix culture. Cell Biol. Int. Reports. 10, 277-286. Kiang, D. T., King, M., Zhang, H. J., Kennedy, B. J. and Wang, N. 1982. Cyclic biological expression in mouse mammary tumours. Science, 216, 68-70. Lawler, E. M., Miller, F. R. and Heppner, G. H. 1983. Significance of three-dimensional growth patterns ol mammarl tissues in collagen gels. In Vitro, 19, 600-610. Lee, A. E. 1970. Mammary turnour development in BR6 mice: hormonal stimulation. Br. J. Cancer. 24. 56%573
894
JONES
ET AL.
Lee, A. E., Pang, L. S. C., Rogers, L. A. and Jeffery, R. E. 1983. Metastasis of mammary turnours in mice: relationship with morphology of primary turnour, and reproductive background of host. Clin. Exp. Metastasis, 1, 223221. Levay-Young, B. K., Hamamoto, S., Imagawa, W. and Nandi, S. 1990. Casein accumulation in mouse mammary epithelial cells after growth stimulated by different hormonal and nonhormonal agents. Endocrinol., 126, 11731182. Ormerod, E. J. and Rudland, P. S. 1982. Mammary gland morphogenesis in uitro: formation of branched tubules in collagen gel by a cloned rat mammary cell line. Deu. Biol., 91, 360-375. Richards, J., Pasco, D., Yang, J., Guzman, R. and Nandi, S. 1983. Comparison of the growth of normal and neoplastic mouse mammary cells on plastic, on collagen and in collagen gels. Exp. Ce[l Res., 146, l-14. Riddle, P. N. 1990. Time-lapse cinemicroscopy. Methods in Molecular Biology, 5, 415-446. Slemmer, G. 1974. Interactions of separate types of cells during normal and neoplastic mammary gland growth. J. Invest. Dermnt., 63, 27-41. Sluyser, M., van der Valk, M. and Blitterswijk, W. J. 1980. Changes in concanavalin A mediated agglutination of hormone-dependent mouse mammary tumour cells during serial transplantation. Br. J. Cancer, 41, 341355. Stampfer. M., Hallowes, R. C. and Hackett, A. J. 1980. Growth of normal human mammary cells in culture. In Virro. 16, 415-425. Williams, J.M. and Daniel, C. W. 1983. Mammary duct elongation: differentiation of myoepithelium and basal lamina during branching morphogenesis. Deu. Biol., 97, 274-290. Yang, J., Richards, J., Bowman, P., Guzman, R., Enami, J., McCormick, K., Hamomoto, S., Pitelka, D. and Nandi, S. 1979. Sustained growth and three-dimensional organization of primary turnour epithelial ceils embedded in collagen gels. Proc. Natl. Acad. Sci. USA, 76, 3401-3405.