Vasculogenic mimicry is associated with trastuzumab resistance of HER2-positive breast cancer

The SKBR3, BT474, and MDA-MB-361 cell lines were purchased from the American Type Culture Collection. The JIMT-1 cell line was purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen. All cell lines were derived from HER2+ breast cancer cells. The SKBR3 cell line was maintained in McCoy’s 5A medium (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum. The BT474, MDA-MB-361, and JIMT-1 cell lines were maintained in DMEM/F12 (Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% fetal bovine serum. Human umbilical vein endothelial cells (HUVECs) were purchased from PromoCell (Heidelberg, Germany) and were maintained in Endothelial Cell Growth Medium 2 (PromoCell). For generation of Tzm-resistant cell lines, SKBR3 and BT474 cells were grown in DMEM/F12 supplemented with 5% calf serum (Sigma-Aldrich), 4 μg/mL insulin (Thermo Fisher Scientific), 0.5 μg/mL hydrocortisone (Stem Cell Technologies, Vancouver, Canada), and 1 μg/mL (for SKBR3) or 2 μg/mL (for BT474) Tzm (Herceptin; Chugai Pharmaceutical, Tokyo, Japan) for over 6 months.

SKBR3 and BT474 cells were seeded in 6-cm dishes and cultured overnight in DMEM/F12 supplemented with 5% calf serum, 4 μg/mL insulin, and 0.5 μg/mL hydrocortisone. The next day, 1 μg/mL (for SKBR3) or 2 μg/mL (for BT474) Tzm or vehicle (phosphate-buffered saline) was added to the medium, and the cells were cultured for 5, 7, 9, or 11 days. The cells were detached from the plates, stained with trypan blue and counted using a hemacytometer. For the cell death assay, detached and attached cells cultured in the presence or absence of Tzm as described above for 3 days (SKBR3) or 7 days (BT474) were collected and stained with trypan blue. Dead cells and live cells were separately counted using a Countess Cell Counter (Life Technologies, Carlsbad, CA, USA).

Cells cultured with or without Tzm for 10 days were labeled with 10 μM EdU for 2 h by using a Click-iT EdU Flow Cytometry Assay Kit (Thermo Fisher Scientific) and were analyzed with a FACSCanto II flow cytometer (BD Biosciences, San Jose, CA, USA). Dead cells were eliminated with a LIVE/DEAD Fixable Dead Cell Stain Kit (Thermo Fisher Scientific).

SKBR3 cells and BT474 cells in maintenance medium were detached from dishes with Accutase (BD Biosciences), immunostained with antibodies conjugated with a fluorochrome, and sorted using a FACSAria II cell sorter (BD Biosciences). The antibodies used are listed in Table 1. The sorted cells were seeded into the wells of 96-well plates and cultured overnight in DMEM/F12 supplemented with 5% calf serum, 4 μg/mL insulin, and 0.5 μg/mL hydrocortisone. The next day, 1 μg/mL (SKBR3) or 2 μg/mL (BT474) Tzm or vehicle was added, and the cells were further cultured for 6 days. Then, the number of cells was counted using an IN Cell Analyzer 6000 (GE Healthcare, Chicago, IL, USA). Alternatively, the viability of Tzm-resistant BT474 cells relative to control cells was estimated by using a Cell Counting Kit-8 (Dojindo, Kumamoto, Japan).

SKBR3 and BT474 cells were grown in DMEM/F12 supplemented with 5% calf serum, 4 μg/mL insulin, 0.5 μg/mL hydrocortisone, and 1 μg/mL (for SKBR3) or 2 μg/mL (for BT474) Tzm or vehicle for 13 days. The cells were detached from the dishes using Accutase and stained with antibodies provided in the Human Cell Surface Marker Screening Panel (BD Biosciences) according to the manufacturer’s protocol. The expression of each antigen was analyzed with a FACSCanto II flow cytometer and FlowJo software (FlowJo, LLC, Ashland, OR, USA). The median fluorescence intensity (MFI) and percentage of positive cells (Pos) were estimated for each antigen. Before analysis, we established the following criteria for determining which antigens were significantly upregulated or downregulated: (1) |Log₂[MFI_Tzm] − Log₂[MFI_Control]| ≥ 0.4; (2) |[Pos_Tzm] − [Pos_Control]| ≥ 2; and (3) both cell lines exhibited similar changes in the expression of an antigen. All of these criteria had to be satisfied. Heat maps were generated using web-based analysis software at http://​www.​heatmapper.​ca [15].

Cells precultured with 1 μg/mL (SKBR3) or 2 μg/mL (BT474) Tzm or control human IgG for 13 days (for Fig. 3d) or Tzm-resistant cell lines and HUVECs cultured in maintenance medium (for Figs. 5, 7, and 8) were detached from dishes using Accutase. The cells were cultured in the wells of 6-, 24-, or 48-well plates coated with Matrigel (Corning, Corning, NY, USA) in Endothelial Basal Medium-2 (EBM-2; Lonza, Basel, Switzerland) supplemented with all reagents of Microvascular Endothelial Cell Growth Medium-2 SingleQuots Supplements and Growth Factors (Lonza) for up to 72 h. This medium, described after this as complete EBM-2, contained four angiogenic growth factors including epidermal growth factor (EGF), fibroblast growth factor 2 (FGF2), insulin-like growth factor 1 (IGF1), and vascular endothelial growth factor (VEGF). For the experiments in Fig. 5d, the Tzm-resistant cell lines were cultured in the wells of 24-well plates coated with growth factor-reduced Matrigel (Corning) in the presence of single angiogenic growth factors or all four growth factors. For the inhibitor experiments appearing in Figs. 7 and 8, the Tzm-resistant cell lines and HUVECs were pretreated with various inhibitors for 2 h. The pretreated cells were cultured in Matrigel-coated wells in complete EBM-2 medium with the same inhibitor for up to 72 h. For Fig. 8h and i, 1 μg/mL Rho Activator II (Cytoskeleton, Inc., Denver, CO, USA) was added 1 h prior to the addition of 0.3 μM salinomycin (Sigma-Aldrich) and continuously supplemented into the medium. Rho Activator II was designed based on the catalytic domain of bacterial cytotoxic necrotizing factors with modifications to increase cell permeability [16, 17]. At the end of the assay, photographs were obtained using a Leica DMi1 phase-contrast microscope with a × 5 objective lens (Leica Microsystems, Wetzlar, Germany). Color images were converted to grayscale, image sizes were reduced, and images were sharpened once for clarity using ImageJ software (National Institutes of Health, Bethesda, MD, USA). Tube formation was quantified by counting the number of tubes formed. When counting, unprocessed photographs were used. A tube was defined as a linear sequence of cells linking two nodes.

For the data in Fig. 3b, cells cultured in the presence of 1 μg/mL (SKBR3) or 2 μg/mL (BT474) Tzm or control human IgG (Thermo Fisher Scientific) for 13 days were lysed with RIPA buffer supplemented with protease inhibitors and phosphatase inhibitors. For the data in Fig. 5c, cells cultured on Matrigel in complete EBM-2 medium overnight were collected with Corning Cell Recovery Solution and lysed with the same RIPA buffer as above. Protein (5 μg) was loaded into gels for SDS-PAGE. The primary and secondary antibodies used are shown in Table 1. The detailed protocol was described previously [9].

Cells cultured in the presence of 1 μg/mL (SKBR3) or 2 μg/mL (BT474) Tzm or control human IgG for 13 days were lysed with TRIzol reagent (Thermo Fisher Scientific), and the lysates were subjected to RNA extraction. qRT-PCR was performed using gene-specific TaqMan probes (Thermo Fisher Scientific). The probe IDs were Hs00900055_m1 for VEGFA, Hs00153153_m1 for HIF1A, Hs01026149_m1 for HIF2A, Hs01675818_s1 for TWIST1, and Hs02758991_g1 for GAPDH. The detailed protocol was described previously [9].

A cohort of all HER2+ breast cancer patients who primarily underwent surgery between January 2016 and August 2017 (N = 25) and all HER2+ breast cancer patients intending to receive neoadjuvant chemotherapy (NAC; 12 weekly cycles of 80 mg/m² paclitaxel followed by 4 triweekly cycles of 500 mg/m² 5-fluorouracil, 75 mg/m² epirubicin, and 500 mg/m² cyclophosphamide [FEC]) at our institution (N = 110) were initially targeted for retrospective analysis. Patients having histories of any preceding malignancies were excluded. The first administration of paclitaxel to the NAC groups with and without Tzm occurred between May 2008 and July 2017. Of the 110 HER2+ breast cancer patients, 27 patients received NAC without Tzm and 78 patients received NAC with Tzm (the first dose of 4 mg/kg followed by 2 mg/kg doses in 12 cycles) concurrently with paclitaxel. In patients who had received complete doses of the planned drugs, we analyzed specimens that contained at least several clusters of invasive cancer cells, in order to observe if VM was present in this remaining cluster of cancer cells, which were considered to be Tzm-resistant. A schematic of the process of patient selection is shown in Fig. 6a. Pathologists at our institution independently determined the extent of remaining cancer cell clusters according to criteria described elsewhere [18], and we analyzed cases of breast cancer for which the postchemotherapeutic effect was estimated as grade 0, 1a, 1b, or 2a. To compare the number of VM channels in NAC-pretreated tumors and untreated tumors, we further chose cases with available untreated tumor samples that were biopsied before NAC (Fig. 6a). Prior to NAC, we performed breast tumor biopsy by using a vacuum-assisted biopsy system (Mammotome; Devicor Medical, Tokyo, Japan) with an 8-gauge biopsy needle. Tumor biopsy samples obtained before NAC were unavailable (2 cases in the NAC without Tzm group and 4 cases in the NAC with Tzm group) in cases where patients were subjected to a biopsy elsewhere before the first visit to our institution. For immunohistochemical (IHC) CD31 and periodic acid-Schiff (PAS) staining, 4-μm-thick slices from formalin-fixed, paraffin-embedded tumor blocks prepared from surgically removed tumors and biopsied tumor samples were deparaffinized and subjected to heat-induced antigen retrieval using Target Retrieval Solution, pH 9.0 (Dako, Santa Clara, CA, USA). The specimens were immunostained with mouse anti-CD31 antibody (JC70A; Dako) and visualized with 3,3′-diaminobenzidine (DAB; Wako, Tokyo, Japan). PAS staining was performed using a Periodic Acid Schiff Stain Kit (ScyTek Laboratories, West Logan, UT, USA) according to the manufacturer’s instructions. After dehydration, clearing, and mounting by standard methods, the specimens were observed using a BX51 light microscope (Olympus, Tokyo, Japan). Clinicopathological evaluation was performed as described previously [9]. This study was approved by the institutional review board of Osaka University Hospital, and written informed consent was obtained from each patient.

Cells were precultured in maintenance medium supplemented with 0 or 4 μM salinomycin for 2 h. Then, the cells were collected using Accutase and seeded into 35-mm dishes coated with Matrigel. The cells were cultured in complete EBM-2 medium with 0 or 4 μM salinomycin under an IX83 inverted microscope (Olympus) equipped with an incubator at 37 °C in 5% CO₂/95% air. Phase-contrast images were acquired beginning 15 min after seeding at time intervals of 2 min 30 s up to 14 h.

Tzm-resistant SKBR3 cells were seeded and incubated on Matrigel-coated 4-well chamber slides (Thermo Fisher Scientific) in complete EBM-2 medium for 30 min. Then, the medium was replaced with Hank’s balanced salt solution supplemented with 0 or 4 μM salinomycin, and the cells were further incubated for 2 h. The cells were fixed with 4% paraformaldehyde for 10 min at room temperature. After permeabilization with 0.2% Triton X-100 for 2 min, filamentous actin (F-actin) was stained with ActinGreen 488 Ready Probe (Thermo Fisher Scientific) for 30 min. Nuclei were counterstained with DAPI, and confocal images were obtained using an FV10i confocal laser scanning microscope (Olympus). The amount of F-actin in a cell was quantified using ImageJ software and was represented as integrated density.

Cells were seeded into a 35-mm μ-Dish with a 2-well culture insert (Ibidi, Martinsried, Germany) and cultured overnight in complete EBM-2 medium. The next day, DMSO or 1 μM salinomycin was added to the medium, and the cells were cultured for another 2 h. For the data in Fig. 8g, 2 μg/mL Rho Activator II was added 30 min prior to the addition of 0.5 μM salinomycin. Then, the inserts were removed, and phase-contrast images were obtained several times during a period of up to 36 h using a Leica DMi1 phase-contrast microscope with a × 5 objective lens.

JIMT-1 cells were cultured on Matrigel in complete EBM-2 medium. After the medium was replaced by Hank’s balanced salt solution with DMSO or 0.5 μM salinomycin, the cells were cultured for another 2 h. Occasionally, 2 μg/mL Rho Activator II was added 30 min prior to the addition of 0.5 μM salinomycin. Cell lysates were prepared and subjected to GTP-bound Rho pulldown assays using an Active Rho Detection Kit (Cell Signaling Technology, Danvers, MA, USA) under the manufacturer’s instructions. RhoA was detected using rabbit anti-RhoA antibody (Cell Signaling Technology, #2117).

Statistical analysis was performed using GraphPad Prism 6 (GraphPad Software, Inc., San Diego, CA, USA) and SPSS software (IBM, Armonk, NY, USA). For parametric analysis, Student’s t test was applied unless otherwise specified. For nonparametric analysis, Mann-Whitney U test was applied. Dunn’s multiple comparison test was performed for the data shown in Fig. 6d. For the pairwise comparisons in Fig. 6e, the Wilcoxon matched-pairs signed-rank test was used. For contingency analysis, the chi-square test was applied. A P value of less than 0.05 was considered significant. All tests were described as two-tailed.

Initially, we assumed that a phenotypic change as a result of short-term loading of Tzm might represent a primary change, although it might be subtle; on the other hand, phenotypic changes associated with long-term loading of Tzm might include many easily detectable secondary changes that mask a significant primary change. Thus, we sought to compare HER2+ breast cancer cell lines cultured with Tzm for a short period (13 days) with those cultured without Tzm. We used two HER2+ cell lines, SKBR3 and BT474, which are estrogen receptor (ER)-negative and ER-positive, respectively. The optimal Tzm concentrations in the culture system were determined to be 1 μg/mL for SKBR3 and 2 μg/mL for BT474; these concentrations were the minimum concentrations producing a maximum effect in each dose-response curve [9] and were used to prevent overloading of Tzm. First, we clarified the effect of short-term Tzm loading on the growth of HER2+ breast cancer cell lines. During short-term culture with Tzm, the growth of both cell lines was suppressed (Fig. 1a). This suppressed growth was due to both induction of cell death and reductions in cellular proliferation (Fig. 1b, c). Although statistically significant, the differences in cell death and proliferation between the two groups were so small that cells surviving Tzm loading for 13 days were considered not to have represented a minor fraction of the original population for either group.

To detect the phenotypic changes in HER2+ breast cancer cells loaded with Tzm, we conducted comprehensive immunophenotyping of Tzm-loaded cells and control cells using a BD Lyoplate Human Cell Surface Marker Screening Panel. We used this method because the detection of marker expression with flow cytometry is sensitive and quantitative. We evaluated the MFI and Pos values for 242 cell surface markers in Tzm-loaded cells and control cells (Fig. 2a and Additional file 1), and the results indicated that the expression of most, though not all, cell surface markers in Tzm-loaded cells did not substantially differ from that in control cells. The criteria we had determined a priori identified 9 antigens as significantly upregulated and 3 antigens as significantly downregulated by Tzm loading (Fig. 2b). Notably, 7 upregulated antigens, including CD144/vascular endothelial-cadherin (VE-cadherin), were endothelial cell markers, and 2 upregulated antigens, including CD44 and SSEA-1, were stemness markers. The heat map of 75 cell surface antigens expressed in endothelial cells among the 242 antigens evaluated (as determined by the Human Cell Differentiation Molecules organization) reveals that most of the antigens were upregulated in Tzm-loaded cells compared with their control counterparts (Fig. 2c). These findings allow us to hypothesize that Tzm loading might be related to vasculogenic mimicry (VM); VM is an example of tumor cell plasticity in which tumor cells form de novo perfusable vascular-like networks, and genes promoting VM are associated with endothelial, stemness, and hypoxia signaling pathways [19].

To determine whether the upregulation of each cell surface marker was associated with Tzm sensitivity, the two cell lines grown in maintenance medium without Tzm loading were FACS-sorted by each marker and were subjected to Tzm sensitivity assays. CD142, CD144, CD171, and SSEA-1 were significantly associated with Tzm resistance in SKBR3 cells, and CD144 was significantly associated with Tzm resistance in BT474 cells (Fig. 3a).

Next, we examined the expression of various VM markers in Tzm-loaded cells. As expected, COX2, MMP2, MMP14, and phosphorylated SMAD2/3 were upregulated in Tzm-loaded cells compared with control cells, as were VEGFA, HIF1A, HIF2A, and TWIST1 mRNA (Fig. 3b, c). We then performed a tube formation assay to examine whether cells formed tube-like structures on Matrigel, which is considered a hallmark process of VM. Tzm-loaded SKBR3 cells and BT474 cells did not exhibit tube formation on Matrigel in the presence of four angiogenic growth factors, including VEGF, IGF1, FGF2, and EGF (Fig. 3d). To determine why tubes did not form, we explored the expression of several angiogenic receptors: phosphorylation of both EPH receptor A2 and its functional counterpart (EPH receptor A4) was suppressed, presumably because phosphorylation of EPH receptors mediated by the HER2 tyrosine kinase was inhibited by Tzm [20]. Furthermore, the expressions of FGFR1, VEGFR1, and VEGFR2, all of which play a pivotal role in VM, were not observed in any cells (Fig. 3b). These results indicate that short-term Tzm loading is partially associated with vascular phenotypes in HER2+ breast cancer cells.

Taking the results described above into account, we assumed that long-term Tzm loading, which gives rise to Tzm resistance, might be associated with a complete VM phenotype in HER2+ breast cancer cells. We cultured SKBR3 cells and BT474 cells in the presence of Tzm for more than 6 months. BT474 cells loaded with Tzm became completely unattached to culture vessels and grew to form spheroid-like structures, while SKBR3 cells loaded with Tzm remained adherent (Fig. 4a). Both cell lines acquired resistance to Tzm (Fig. 4b). The expression of CD144 in both Tzm-resistant cell lines was upregulated compared with that in the parental cell lines, confirming the relevance of this molecule to Tzm resistance (Fig. 4c). We did not use Tzm-resistant BT474 cells for further experiments because the cells were not adherent to any surfaces; instead, we used two Tzm-resistant HER2+ cell lines, JIMT-1 and MDA-MB-361, in addition to Tzm-resistant SKBR3 cells (designated SKBR3-TR). The JIMT-1 cell line is derived from a patient who received Tzm treatment [21], and the MDA-MB-361 cell line is primarily resistant to Tzm [22]. When these cell lines were subjected to tube formation assays performed in the presence of VEGF, IGF1, FGF2, and EGF, they all exhibited tube formation (Fig. 5a, b).

To elucidate the molecular basis of VM in these cell lines, the expression of VM markers and growth factor receptors involved in angiogenesis, including ERBB1, FGFR1/2, IGF1R, and VEGFR1/2, was examined. Multiple angiogenic growth factor receptors as well as several VM markers were expressed in these cell lines, though the extent and pattern of expression were different among the cell lines (Fig. 5c). Notably, VEGFR2, which plays a central role in VM, was present in all the cell lines. To examine the effect of each angiogenic growth factor on VM in these cell lines, we subjected the cells to tube formation assays in the presence of one of four growth factors including VEGF, IGF1, FGF2, and EGF. VEGF and IGF1 promoted tube formation in all the cell lines, while FGF2 and EGF promoted tube formation in one or two cell lines (Fig. 5d). These results indicate that Tzm-resistant HER2+ breast cancer cells that express angiogenic growth factor receptors exhibit VM in response to multiple angiogenic growth factors. Thus, one can assume that a drug targeting an angiogenic growth factor or its receptor may not effectively inhibit VM, because other growth factor pathways can quickly compensate for the blocked VM pathways.

The results of our in vitro study indicate that VM is associated with Tzm resistance. To validate this finding in breast cancer patients, we examined VM in tumor specimens surgically removed from HER2+ breast cancer patients, who had received or not received NAC with or without Tzm. The detailed selection process for patients receiving NAC is described in Fig. 6a. Twenty-five patients who had not received neoadjuvant therapy (primary surgery group), 14 patients who had received neoadjuvant paclitaxel without Tzm (NAC without Tzm group), and 24 patients who had received neoadjuvant paclitaxel with concurrent Tzm (NAC with Tzm group) were analyzed. Menopausal status, tumor size, and lymph node metastasis were statistically different among the three groups (Table 2), suggesting that NAC was chosen more frequently as a primary therapy for younger women with more advanced breast cancer. We recognized VM in tumor tissue as a channel consisting of PAS-positive, CD31-negative cells [23] (Fig. 6b, c). Number of VM channels per unit area was significantly larger in surgically removed tumors of the NAC with Tzm group than that in surgically removed tumors of the primary surgery group (P = 0.004; Fig. 6d). For paired tumor samples obtained before and after NAC, the number of VM channels was significantly increased after NAC in the NAC with Tzm group, while the difference in the number of VM channels was not significant in the NAC without Tzm group (Fig. 6e). These results indicate that the in vitro findings are also applicable to HER2+ breast cancer patient samples.

Finally, we searched for a drug that might effectively inhibit VM in Tzm-resistant cells. Among 9 small molecule inhibitors screened, we found that salinomycin, a drug that selectively targets breast CSCs [24], strongly inhibited tube formation in three Tzm-resistant cell lines (Fig. 7a). The validation experiments showed that 4 μM salinomycin completely inhibited tube formation in the three Tzm-resistant cell lines, while the same concentration of salinomycin did not inhibit tube formation in HUVECs (Fig. 7b, c). Time-lapse microscopy of tube formation in MDA-MB-361 cells confirmed the inhibition of tube formation by salinomycin, and it appeared that cell shape and motility were affected (Fig. 7d and Additional files 2 and 3).

These results prompted us to examine if salinomycin affected cytoskeletal integrity, which might cause the observed changes in cell shape and reductions in cell motility. As expected, SKBR3-TR cells treated with salinomycin showed reduced cell sizes and increased roundness (Fig. 8a–c). Phalloidin staining demonstrated that the amount of F-actin in each cell was decreased, suggesting that salinomycin affected the integrity of the actin cytoskeleton (Fig. 8a, d). Further, salinomycin significantly inhibited cell migration, consistent with the results of time-lapse microscopy (Fig. 8e). We next examined whether the Rho-GTPases, which are master regulators of the actin cytoskeleton, were involved in the inhibition of VM by salinomycin. GTP-bound Rho was reduced in JIMT-1 cells treated with salinomycin but restored by the Rho activator (Fig. 8f), indicating that salinomycin was able to inactivate Rho-GTPases. Accordingly, activation of Rho restored the reduction in cell motility (Fig. 8g) and the inhibition of VM (Fig. 8h, i) caused by salinomycin. These results indicate that salinomycin inhibits VM through inactivation of Rho-GTPases that regulate the integrity of the actin cytoskeleton.