a, b, Representative images (a) and quantification (b) of PGR expression of mammary epithelial cells from wild-type BALB/c mice at 4, 8, 25 and 40 weeks of age. Scale bars, 100 μm. PGR expression was reduced by 75% in 40-week-old wild-type mammary gland compared to 4-week-old mice and disappeared in primary tumours (see also Extended Data Figs 1f, 2a). n, number of ducts or glands (in early lesions and normal tissue) or visual fields in primary tumours. c, d, Representative micrographs of lesions 8 weeks after transplantation of early lesion spheres resembling DCIS (c) or less-advanced early lesions (d) displaying PGR expression (brown nuclear staining). e, Tumour growth from primary tumourspheres in young and old recipients. f, Differentiation of cells from early lesions but not of primary tumour cells in Matrigel (left) or in sphere culture (right) into acinus-like structures. Progesterone stimulation accelerated formation of acinus-like structures by early lesion cells, under mammosphere conditions. NED, no evidence of differentiation. g, Staining for CK8/18, PGR and EpCAM (epithelial cell adhesion molecule) shows expression of PGR and CK8/18 only in differentiated structures (top) compared to undifferentiated spheres (bottom). h, i, Primary tumourspheres were transplanted alone (n = 23) or co-transplanted with MM3MG–Pgr-B (n = 5) or MM3MG spheres (n = 5). DCCs numbers in bone marrow (h) and the number of mice with tumours (i) were checked 4 weeks later. Pgr-B-transduced mammary epithelia suppressed metastatic dissemination and accelerated tumour formation from primary tumourspheres. j, Pregnancy at the early lesion stage induces dissemination. A group of young BALB-NeuT mice mated (n = 5) at the early lesion stage (7-week-old) and were killed at the end of pregnancy. These mice did not form palpable tumours, but had a higher number of DCCs compared to unmated control mice (n = 6). k, Pregnancy at the advanced tumour stage. A group of BALB-NeuT mice (n = 5) were mated at the time of in situ carcinoma (15-week-old) and killed at the end of pregnancy. All pregnant mice had faster growing tumours compared to unmated control mice. l, Schematic of transplantation protocol for mammary gland or primary tumour tissue pieces into wild-type recipients. m, Example of primary tumour and macro-metastasis assessment. n, Number of metastatic foci in transplanted mice. 18 mice from the gland model and 3 mice from the tumour model were excluded from analyses owing to the fusion of metastatic lesions, which made it difficult to count individual lesions. o, Similar growth kinetics of primary tumours from gland and tumour piece models for samples from the red box in Fig. 4e. p, q, Mice from o were compared for the duration of the follow-up period after surgery. Mice from both groups were killed at the first signs of general health deterioration, which occurred earlier in gland-model mice (p). Longer follow-up time after curative surgery did not result in more metastases in recipients transplanted with primary tumour pieces (q).*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001; NS, not significant; (Student’s t-test (b, f, h, j); Fisher’s exact test (i); F-test for the slopes (c, k); Mann–Whitney U-test (n–q)). Data are shown as mean ± s.d. (b, e, f, k) or median (h, j, n–q).