An intraductal human-in-mouse transplantation model mimics the subtypes of ductal carcinoma in situ

Recipients are 6- to 10-week-old virgin female SCID-beige mice. Before transplantation, cells are resuspended as single cells in PBS and counted from DCIS.COM, SUM225, or primary human DCIS cells. A 30-gauge Hamilton syringe, 50-μl capacity, with a blunt-ended 1/2-inch needle is used to deliver the cells. The mice are anesthetized, and a Y-incision is made on the abdomen to allow the skin covering the inguinal mammary fat pads to be peeled back to expose the inguinal gland. The nipple of the inguinal gland is snipped so that the needle can be directly inserted through the nipple. Two microliters of cell-culture medium (with 0.1% trypan blue) containing cells at a concentration of 2,500 to 5,000 cells/μl are injected; the injected liquid can be visually detected in the duct. The skin flaps are repositioned normally and held together with wound clips. The primary human DCIS was chopped very finely by using a Teflon block and razor blade or scalpel followed by overnight enzymatic digestion in DMEM/F12 with antibiotics, supplemented with collagenase (1.0 mg/ml) and hyaluronidase (100 U/ml).

Animal and human experiments were conducted by following protocols approved by the Baylor College of Medicine Animal Care and Use and Human Subjects Committee. An informed consent was deemed not to be required by the Human Subjects Committee. [See Additional data file 1 for a video demonstration of the intraductal method.]

MCF10DCIS.COM and SUM-225 were purchased from Asterand, Inc. (Detroit, MI) and were maintained according to the supplier's guidelines.

Whole mount and H&E were performed as described previously [8].

Immunofluorescence (IF) was performed after tissue deparaffinization by clearance in xylene and hydration through graded ethanol series. Microwave antigen retrieval (20 min) in 10 m M sodium citrate was performed for all the antibodies used for IF. A 5% solution of bovine serum albumin in phosphate-buffered saline + 0.5% Tween 20 was used as blocking buffer. Sections were incubated with the following primary antibodies overnight at 4°C: ERα rabbit polyclonal 1:50 (Novocastra; 6F11, NCL-ER-6F11, Newcastle Upon Tyne, Tyne and Wear, UK), anti-human SMA 1:100 (1A4-Dako, M0851, Dako, Glostrup, Denmark), anti-mouse SMA 1:200 (A14; Sigma, A2547, St. Louis, MO, USA), CK5 1:50 (XM26, Vector, VP-C400, Burlingame, California, USA), CK8 1:50 (C51, Zymed, 18-0185, San Francisco, California, USA), Her-2 1:50 (SP3, Labvision, RM-9103-SO, Fermont, California, USA), Cytokeratin (AE1/3)1:50 (AE1/AE3, Dako, M3515, Glostrup, Denmark), CK-19 1:50 (Clone A53-B, Lab Vision, MS-198-P0, Fremont, California, USA), and Her-1 1:50 (Clone 31G7, Zymed Laboratories, San Francisco, California, USA). Nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI; Vector Laboratories, Burlingame, California, USA) and TO-PRO-3 iodide (Invitrogen, T3605, Carlsband, California, USA). Secondary antibodies included anti-mouse Alexa 488 and anti-rabbit Alexa 594 (Molecular Probes, Carlsband, California, USA). Confocal microscopy was performed by using a laser-scanning confocal microscope (model 510; Carl Zeiss MicroImaging, Inc., Thornwood, NY, USA). The acquisition software used was LSM image browser (Carl Zeiss MicroImaging, Inc., Thornwood, NY, USA). Phase-contrast images were captured with an inverted microscope (CK40-SLP; Olympus). The acquisition software used was Photoshop 5.0 (Adobe). [See Additional data file 2 for detailed information on primary antibodies.]

Primary antibodies used were anti-human MUC-1 1:250 (Stem Cell Technologies, 01423, Vancouver, BC, Canada), CD10-APC 1:10 (BD Bioscience, 340923, San Jose, California, USA), EPCAM-FITC (Stem Cell Technologies, 10109, Vancouver, BC, Canada), Biotinylated AC133 1:5 (Miltenyi Biotec, 130-090-664, Bergisch Gladbach, Germany), PE-CY5-conjugated rat anti-human CD49f 1:50 (BD Pharmingen, 551129, San Diego, California, USA), PE-conjugated mouse anti-human CD44 1:50 (BD Pharmingen, 555479, San Diego, California, USA), PE-conjugated mouse anti-human CD29 1:50 (BD Pharmingen, 556049, San Diego, California, USA), FITC-conjugated mouse anti-human CD24 1:50 (BD Pharmingen, 555427, San Diego, California, USA), and Biotinylated Thy-1 (Abcam, ab1154, Cambridge, MA, USA). Biotinylated antibodies were labeled by using Streptavidin-PE-CY5 (Invitrogen, SA1012, Carlsband, California, USA). MUC-1 antibody was conjugated by using FITC anti-mouse Igg1 (Biolegend, 406605, San Diego, California, USA). Isotype controls used were FITC mouse Igg1 (Biolegend, 400107, San Diego, California, USA), FITC mouse Igg2a (400207, Biolegend, San Diego, California, USA), PE mouse Igg2b (Biolegend, 401207, San Diego, California, USA), PE-CY5 mouse Igg1 (BD Bioscience, 550618, San Jose, California, USA), and PE-CY5 Rat Igg2a (Biolegend, 400509, San Diego, California, USA). Cells were stained at a cell concentration of 1 × 10⁷/ml for 30 min on ice followed by washes in Hanks' Balanced Salt Solution containing 2% fetal bovine serum (HBSS; Gibco BRL, Carlsband, California, USA). Flow-cytometry analysis and acquisition were performed by using the BD LSR II flow cytometer and BD FACSDIVA based software (BD Biosciences, San Jose, California, USA). Flow-cytometry sorting was performed by using FACSAria (BD Biosciences, San Jose, California, USA). The fluorescence expression level is arbitrarily designated as low (lo) if the log¹⁰ of median fluorescence intensity (FI) is between about 0 to 2, medium (med), if the FI is between about 2.1 and 3.6, and high (hi) if the FI is higher than about 3.7.

Immuno-FISH was performed as previously described in [5].

A limiting factor for studying human DCIS is the lack of suitable in vivo models. Furthermore, sufficient amounts of primary DCIS tissue available for research purposes are often inadequate. Therefore, the main goals were to design a method for expansion of human DCIS tissue in vivo and to develop models that would represent various subtypes of human DCIS. Many investigators have tried, with limited success, to develop xenograft models of human DCIS by a variety of methods including humanized fat-pad transplantation. The reason for the limited success may be that DCIS has been treated like an invasive lesion rather than a lesion limited in its growth potential under very restrictive conditions. With the idea that DCIS initiates and grows inside the ducts, we used intraductal transplantation. This approach involves injection of human DCIS cells directly into the primary ducts (Figure 1, and video [see Additional data file 1] demonstrate the intraductal-transplantation technique). We intentionally mimicked DCIS in the mouse to the closest condition to that found in vivo in the human disease. Two established DCIS cell lines, DCIS.COM, SUM-225, and a primary human DCIS lesion were used. The primary human DCIS cells were obtained by digestion of a primary human DCIS tumor. A pathologist diagnosed the primary human DCIS lesion. The researchers were blind to any other patient identifiers. The take rate was 90% for all three cell lines. The cells were injected at 40,000 cells in 2 μl. Three replicates for DCIS.COM (ratio of growth, 9:10, 5:6, and 6:6), two replicates for SUM-225 (ratio of growth, 5:6 and 5:6), and one replicate for the primary human DCIS, FSK-H7 (ratio of growth, 9:10) were found. None of the xenografts in this study was serially transplanted. Both DCIS.COM and SUM-225 are unique because they progress through a DCIS-like stage during tumor formation. The DCIS-like lesions generated by the DCIS.COM cell line slowly progressed to invasive lesions in 10 weeks, whereas those generated by the SUM-225 did not become invasive during the 10-week study period. In one study, three (25%) of 10 mammary fat pads injected with DCIS.COM generated invasive cancers in 10 weeks. This is in contrast to other established invasive breast cancer cell lines, such as MCF-7, that develop extensive invasive cancers in as early as 4 weeks after intraductal injections in six of six fat pads (unpublished data). None of the DCIS-like lesions generated by FSK-H7 (primary human DCIS cells) showed invasion during the 6-week study period.

Figure 2 shows whole-mount and hematoxylin and eosin (H&E) stains of intraductal lesions generated from DCIS.COM (A, B), SUM-225 (C, D), and the primary human DCIS, FSK-H7 (E, F) 6 weeks after injection. SUM-225 exhibited a comedo pattern, whereas DCIS.COM exhibited a cribriform pattern. Previously, it was reported that DCIS.COM generates a comedo DCIS when injected subcutaneously [3]. This difference, most likely, is due to the subcutaneous versus intraductal sites of injections. Injection of the primary human DCIS cells (FSK-H7) generated apocrine-like DCIS lesions. Interestingly, the whole mounts show that many of the growths appear to be at the periphery and separated from the point of injection. This may be because the cells were injected when the SCID-beige mice were 6 to 8 weeks old, when the ducts are expanding. At the time of ductal expansion, the terminal ducts may provide a favorable environment for cancer cell growth. The reason for this effect is not known.

To assess the human and mouse origin of cells within the intraductal lesions, immuno-fish (FISH) was performed by using anti-mouse smooth muscle actin (SMA) (Figure 3A-C) and anti-human pan-cytokeratin antibodies (D-E), fluorescently labeled human (green) (F-J), and mouse (red) Cot1 DNA probes for FISH (K-O). The data show that human cells, indicated by green human Cot1 probe staining, are surrounded by mouse myoepithelial cells and myofibroblasts, indicated by a red mouse Cot1 probe. Therefore, the intraductal model recapitulates human DCIS by allowing the cancer cells to grow within the boundaries of the mouse basement membrane and myoepithelial cells. Furthermore, the model provides a unique opportunity for studying the factors that influence the process of cancer cell invasion into the stroma.

The transitional process of in situ breast lesion to invasive carcinoma is currently poorly understood. It has been proposed that the loss of myoepithelial cells concurrent with degradation of the basement membrane may play a major role in DCIS invasive progression. A recent report described the role of myoepithelial cells in this process in detail [5]. By subcutaneous transplantation of DCIS.COM, this group demonstrated that myoepithelial cells suppress, whereas fibroblasts enhance DCIS invasive progression. It was further demonstrated that the loss of myoepithelial expression of TGFβ R2 and SMAD-4, Gli-2, and MMP14 play a role in the invasive progression of DCIS [5]. Thus, the current model allows the study of many early processes of breast cancer cell invasion into the stroma, including the interactions of cancer cells with the surrounding normal luminal and myoepithelial cells as well as the basement membrane.

The long-term goal of our studies is to develop subtype-specific stable xenograft models of human DCIS by intraductal transplantation. The xenograft lines are termed stable if their morphology, subtype specificity, and invasive properties remain unchanged with repeated transplantation. Subtype-specific xenograft lines should allow one to study the distinct molecular and biologic mechanisms underlying the invasive progression of subtypes of human DCIS. We have identified two cell lines, DCIS.COM and SUM-225, that generate stable basal-like and Her-2-overexpressing DCIS-like lesions, respectively. Intraductal xenografts of a human primary DCIS, FSK-H7, generated Her-2-overexpressing lesions.

Previous studies have shown that immunoassays may be used to predict the tumor subtypes identified with microarray analyses with a high degree of specificity [see Additional data file 3) [9, 10]. Accordingly, immunofluorescence studies using anti-human cytokeratin-8 (CK-8), CK-19, CK-5, estrogen receptor (ER), Her-1, and Her-2 antibodies were performed. Figure 4 shows that SUM-225 and FSK-H7 lesions uniformly express CK-19 (A, C), whereas DCIS.COM lesions are CK-19 negative (B). SUM-225 and FSK-H7 express Her-2 (D-F), whereas CK-5 is exclusively expressed by DCIS.COM (G-L). All three intraductal lesions express CK-8; however, none of the lesions expresses ER (data not shown). These results suggest that DCIS.COM generates basal-type lesions, whereas SUM-225 and FSK-H7 generate Her-2-overexpressing lesions ([see Additional data file 1], and Figure 4).

We further used the intraductal model to test the hypothesis that xenografts representing two distinct subtypes of DCIS-like lesions may contain distinct tumor-initiating subpopulations. The two DCIS cell lines were used for this study. The cell lines were analyzed for the expression of human mammary basal and luminal surface markers. Figure 5 is a histogram depicting the relative expression level of the individual markers in DCIS.COM and SUM-225. The expression level is arbitrarily designated low (lo) if the log¹⁰ of the median fluorescence intensity (FI) is between about 0 and 2, medium (med) if the FI is between about 2.1 and 3.6, and high (hi) if the FI is higher than about 3.7. The data in Figure 5 show that DCIS.COM expressed a hi level of the basal markers (CD44 and CD49f) and lo levels of the luminal markers (CD24 and MUC-1). In contrast, SUM-225 expressed medium levels of the basal markers and hi levels of the luminal markers. Various subpopulations expressing putative markers of mammary stem and progenitor cells were sorted with flow cytometry and transplanted. These subpopulations (Tables 1 and 2; column 2) were chosen based on previous reports of markers expressed by human breast stem/progenitors, mouse mammary epithelial stem cells, and human breast tumor-initiating cells [10, 11]. The gates for the subpopulations sorted and transplanted are shown in the supplementary data [see Additional data file 4]. Fisher's Exact test was used to compare the in vivo growth potentials between the various subpopulations. Table 1 shows transplant results for two replicate experiments per sorted subpopulation per cell line. DCIS-like lesions are referred to as in vivo or intraductal growth. The number of fat pads containing positive intraductal growth over the total number of fat pads transplanted, growth fractions, and the total number of ducts containing intraductal growth are listed in Table 1, columns 3, 4, and 5, respectively. [See Additional data file 5.] Table 2 shows the results of Fisher's Exact test. As depicted in Table 1 in DCIS.COM, subpopulations expressing high levels of the basal markers and low levels of the luminal markers (CD49f^hi/CD24^lo, CD49f^hi/MUC-1^lo, and CD44^hi/CD24^lo) demonstrated higher in vivo growth fractions and generated more ducts containing lesions. However, as shown in Table 2 with Fisher's Exact test, none of the subpopulations in DCIS.COM achieved significantly higher growth fractions when compared with the random sort. Therefore, the majority of cells in DCIS.COM possessed tumor-initiating properties. In contrast, in SUM-225 cells, two subpopulations, CD49f^med/CD24^hi and CD49f^medMUC-1^hi, achieved significantly higher growth fractions compared with the random sort (Table 2). These subpopulations also generated the largest number of ducts with lesions (Table 1). Because the tumor-initiating cells in DCIS.COM and SUM-225 cells expressed distinct surface markers, we postulate that the tumor-initiating cells may arise from distinct cell types. However, lineage-tracing experiments are required to establish whether different types of tumors arise from distinct cell types. Notably, although tumor formation was not an end point for this experiment, a few palpable tumors were found in the random sort as well as the CD44^hi/CD24^lo groups in DCIS.COM. None of the SUM-225 subpopulations generated a tumor during the time course of our study (6 weeks).