Clinical data were reviewed from 100 consecutive patients who had been diagnosed with early-stage BC at Kochi University Hospital, Nankoku, Japan, from January to December 2015. The study was reviewed and approved by the Ethic Committee for Clinical Research of the School of Medicine, Kochi University (approval no. 2020-123; 4th December 2020). Written consent to participate was obtained from all patients. All 100 patients had undergone surgical treatment and completed follow-up for >5 years at the hospital. All tissues obtained during surgery had been embedded in paraffin blocks following formalin fixation for preservation. The immunohistochemical evaluation of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) were also reviewed from clinical pathology reports at the time of initial diagnoses.

A total of 11 tissue samples from patients with metastasis or recurrence during 5-year follow-up were cut into 4 µm thick slices and heat-treated at 95°C for 30 min with ULTRA cell conditioning 1 retrieval solution (Ventana Automated Systems). Immunohistochemical examination was performed using a Ventana automated system with anti-fascin-1 mouse monoclonal antibody (1:100; cat. no. M3567; Dako, Agilent Technologies, Inc.). Another set of 17 consecutive tissue samples from patients without metastasis or recurrence underwent the same procedure as a control group. To evaluate immunohistochemical expression of fascin, the Allred scoring system was used (13,14). Briefly, the proportion of stained cells was categorized as negative (0), <1 (1), 1–10 (2), 11–33 (3), 34–66 (4) and >66% (5) positive. The intensity of the most predominantly stained area was categorized as no (0), weak (1), intermediate (2), or strong (3) staining (Fig. 1). Allred score (0–8 points) was calculated by adding the proportion and intensity values. Independent evaluation of immunostaining was performed by two expert pathologists (YH, IM).

Human TNBC MDA-MB-231 cells (American Type Culture Collection), were cultured at 37°C for 24 h with 5% CO₂ in DMEM (Sigma-Akdrech, Merck KGaA) with 10% fetal bovine serum (Biosera France SAS). Following recombination of the short hairpin RNA (shRNA) against fascin, the pLKO.1-puro plasmid vector (1 µg/µl; Clone ID: NM_003088.2-1699s1c; Sigma-Aldrich, Merck KGaA), containing the puromycin-resistance gene, was transfected into MDA-MB-231 cells with FuGene^®6 Transfection Reagent (Roche Diagnostics) at 37°C for 24 h, according to the manufacturer's instructions. Cells were incubated at 37°C with 2.2 µg/ml puromycin (Sigma-Akdrech, Merck KGaA) for 2 weeks. Puromycin-resistant colonies (~20) were obtained and cultured with the medium containing puromycin (2.2 µg/ml) in 100 mm-diameter dishes. When the cell confluency reached 80%, dishes were provided to carry out western blot analysis and to gain FKD MDA-MB-231 cells, respectively. Transfection of the pLKO.1-puro plasmid vector without shRNA against fascin was used to generate non-FKD MDA-MB-231 cells.

For immunocytochemical staining, non-FKD and FKD MDA-MB-231 cells were incubated with anti-fascin-1 mouse monoclonal antibody (1:100; cat. no. M3567; Dako, Agilent Technologies, Inc.) at 4°C overnight reacted with fluorescein isothiocyanate (FITC)-labeled anti-mouse IgG antibody (1:200; cat. no. F-2761; Molecular Probes, Thermo Scientific, Inc.) at room temperature for 1 h and nuclei were stained with DAPI (Sigma-Akdrech, Merck KGaA). For phalloidin staining, cells (2×10⁴/well) were cultured at 37°C for 24 h on a slide chamber (AGC Techno Co., Ltd), fixed with 100% acetone at −20°C for 20 min and stained with FITC-labeled phalloidin (Sigma-Akdrech, Merck KGaA) to bind actin filaments, according to the manufacturer's instructions. Fluorescence microscopy was performed using an Olympus BX53 (Olympus Corporation; magnification ×200) with cellSens standard (ver.1.12, Olympus Corporation).

Concentration of samples lysed in RIPA Buffer (Fujifilm Wako Pure Chemical Corp) was measured by Pierce BCA Protein Assay Kit (cat. no. 23227; Thermo Fisher Scientific, Inc.). 20 µg/lane samples were prepared for SDS-PAGE separation (10% SDS-PAGE Gel; Bio-Rad Laboratories, Inc.) and transferred to polyvinylidene difluoride membranes using the Trans-Blot Turbo Transfer System (Bio-Rad Laboratories, Inc.). Membranes were blocked with Blocking One (cat. no. 03953-93; Nacalai Tesque, Inc.) at room temperature for 1 h, and incubated at 4°C overnight with following antibodies: anti-fascin-1 mouse monoclonal (1:500; cat. no. M3567; Dako, Agilent Technologies, Inc.), E-cadherin mouse monoclonal (1:500; cat. no. M3612; Dako, Agilent Technologies, Inc.), Snail1 rabbit polyclonal antibody (1:200; cat. no. Ap205Aa; Adgent, Inc.), and GAPDH mouse monoclonal antibody (1:2000; cat. no. 20035; ProMab Biotechnologies, Inc.). Then, the membranes were incubated at room temperature for 1 h with either of following secondary antibodies: HRP-labeled anti-mouse polyclonal antibody (1:2000; cat. no. P0447; Dako, Agilent Technologies, Inc.) or anti-rabbit polyclonal antibody (1:2000; cat. no. P0399; Dako, Agilent Technologies, Inc.). Bands were visualized by ECL Prime Western Blotting Detection Reagents (Amersham, Cytiva) and observed using LAS-4000 Lumino-Image Analyzer (FUJIFILM Wako Pure Chemical Corporation). ImageJ (version no. 1.53; National Institutes of Health) was used for analysis.

Trypsinized 2×10⁴/ml parent, non-FKD and FKD cells were counted with SKC, Inc. C-Chip^™ Disposable (Thermo Fisher Scientific), disposable hemocytometer, seeded onto a 35 mm-diameter dish and cultured at 37°C for 48 h. When the cell confluence reached 90%, a scratch was made through the center of the cell layer using a 20-µl pipette tip. The cells were incubated using 10% FBS at 37°C for 20 h. Phase-contrast microscopy was performed at 0, 12, 16 and 20 h on 50 spots, which were randomly marked on 35-mm dishes containing each cell (Olympus Corporation, CKX41, original magnification ×200). To investigate cell migration ratio, the area of each scratch without cell migration was measured in the same size dimension with the same magnification, using ImageJ software, ver.1.53. The mean migration of 50 spots in each cell sheet was examined. Then, the mean value at 0 h was defined as 1.0 and relative cell migration ratio was detected at each time of each cell.

As aforementioned, a scratch was made using a 20-µl pipette tip on a layer of non-FKD and FKD cells (2×10⁴/well each) seeded onto the slide chambers (AGC Techno Co., Ltd). Immunocytochemical staining with anti-fascin-1 mouse monoclonal antibody and FITC-labeled anti-mouse IgG antibody was performed as aforementioned. Following immunofluorescence microscopy, cells were fixed with 2.5% glutaraldehyde and 1% osmic acid at 4°C for 6 h and 1h, respectively. Then, cells were stained with 1% phosphotungstic acid at room temperature for 10 min and low-vacuum scanning electron microscopy (LV-SEM) was performed. Finally, images of immunofluorescence and LV-SEM were superimposed using ImageJ software, ver.1.53.

A total of 1×10⁴/ml non-FKD and FKD MDA-MB-231 cells were counted as aforementioned and Cellmatrix (Nitta Gelatin Inc.) was prepared according to the manufacturer's instructions. Subsequently, each group of cells and 500 µl Cellmatrix^® were poured into a 35-mm-diameter dish (coated with Cellmatrix at 37°C for 30 min), covered with 200 µl DMEM and cultured at 37°C for 10 days. The gels were fixed in 20% buffered formaldehyde at room temperature overnight and embedded in paraffin. Following HE stain (Mayer's hematoxylin and 1% eosin staining at room temperature for 10 and 5 min, respectively), immunohistochemical staining for E-cadherin, Snail1 and vimentin was performed as follows. 4 µm-thick tissue samples were immersed in 0.01 M citrate buffer (pH 7.0) for antigen retrieval (98°C, 30 min). Sections were immersed in 0.3% hydrogen peroxide/methanol at room temperature for 10 min to remove endogenous peroxidase. Then, the sections were blocked using Blocking One (cat. no. 03953-93; Nacalai Tesque, Inc.) at room temperature for 1 h and incubated at 4°C overnight with following antibodies: anti-E-cadherin mouse monoclonal antibody (1:100; cat. no. M3612; Dako, Agilent Technologies, Inc.), Snail1 rabbit polyclonal antibody (1:100; cat. no. Ap2054a; Adbent, Inc.) and vimentin mouse monoclonal antibody (1:200; cat. no. M725; Dako, Agilent Technologies, Inc.). After washing with PBS, sections were incubated in N-Histofine Simple Stain MAX PO (MULTI; cat. no. 424151; Nichirei Biosciences Inc.) at room temperature for 1 h and washed again with PBS. Finally, the sections were immersed in DAB substrate solution (tablet/15 ml distilled water; SIGMAFAST 3,3′-Diaminobenzin tablets; D4418; Sigma-Aldrich, Merck KGaA) and the nucleus was stained with Mayer Hematoxylin at room temperature for 1 min. Optical microscope images were performed using an Olympus BX53 with cellSens (Olympus Corporation). An additional set of gels was fixed in 2.5% glutaraldehyde at 4°C for 6 h for LV-SEM.

A total of 1×10⁴/well parent, non-FKD and FKD MDA-MB-231 cells were counted as aforementioned and seeded onto PrimeSurface^® (Sumitomo Bakelite Co., Ltd.), 96-well plate with ultra-low adhesion round bottom dishes. Following incubation at 37°C with 5% CO₂ for 5 days, multicellular spheroids were generated. As aforementioned, fixation of the samples, HE and immunohistochemical staining of fascin and E-cadherin for microscopy (Olympus BX53), and LV-SEM were completed. A total of 10 spheroids was collected in a 35-mm-diameter dish (coated with Cellmatrix at 37°C for 30 min) with 500 µl Cellmatrix, covered with 200 µl DMEM and cultured at 37°C for 3 days. After fixation in 20% buffered formaldehyde at room temperature for 6 h, the samples were observed under stereoscopic microscope MZ16FA (Leica Microsystems Tokyo, Japan, original magnification ×200).

Immunohistochemical and immunofluorescent analyses were performed as previously described (9). The antibodies and chemical agents are shown in Table I.

Samples containing cells or spheroids were fixed using 2.5% glutaraldehyde in 0.1M phosphate buffer (PB, pH 7.4) at 4°C for 4 h and postfixed with 1% osmium tetroxide in PB at 4°C for 1 h. Then, each block was washed with distilled water for 30 min and stained with 1% phosphotungstic acid solution at room temperature for 10 min. Following a final wash with distilled water for 30 min, the block was dried on an electrically conductive tape (Nisshin EM Co., Ltd.) and observed using a Miniscope^® TM3030 (Hitachi Ltd.).

χ² test was performed to detect the association between a TNBC subtype and the incidence of metastasis or recurrence. Cochran-Armitage test was performed to detect the relationship between the Allred score and the incidence of metastasis or recurrence. χ² test was performed to analyze the relationship between E-cadherin expression in FKD and non-FKD cells, and Snail1 expression as well. Tukey-Kramer test was used to compare the mean values of cell migration ratios in wound healing assay. Two-side test was applied for all analyses except Cochran-Armitage test. JMP (ver. 14.3.0, SAS Institute inc.) was used for statistical analyses.

A total of 11 out of 100 consecutive patients with BC developed metastasis or recurrence within five years (Table II). Fresh fascin immunostaining was performed on tissue samples from these patients as well as 17 patients without metastasis or recurrence. χ² test result showed a significant association between TNBC subtype and the incidence of metastasis or recurrence (P<0.05, Table II). Cochran-Armitage test showed a significant association between the Allred score of fascin and the incidence of metastasis or recurrence (P<0.05, Table II). However, Cases #2 and #6 developed poor prognosis regardless of negative or slightly positive fascin expression (0 and 2, respectively). Cases #14 and #22 did not develop metastasis or recurrence during follow-up periods, although they showed high Allred scores (6 and 8, respectively).

The expression of fascin (green) was strongly positive in non-FKD cells and effectively suppressed in FKD cells (Fig. 2A). Numerous filopodia, including actin filaments (arrows), were observed on the membrane of non-FKD cells, however, these filopodia were decreased and actin positive granules (arrowheads) were observed on the membrane of FKD cells. Fascin was strongly positive in non-FKD cells and was suppressed in FKD cells (Fig. 2B). Snail1 expression was also decreased in FKD cells. By contrast, E-cadherin expression increased in FKD cells.

2D LV-SEM were performed following 3 day cultivation of non-FKD and FKD MDA-MB-231 cells. Non-FKD cells exhibited loose cell-cell connections (Fig. 3A), however, bundles of extremely thin microfibrils on the cell surface were observed (arrows; Fig. 3B). FKD cells exhibited cell-cell adhesion (Fig. 3C) and granular nodules of various sizes were observed on the cell surface (arrowheads; Fig. 3D).

Following 20 h incubation, the mean scratch areas in parent and non-FKD cell sheets decreased notably due to the cell migration, however, the mean scratch area in FKD cell sheets decreased only slightly (Fig. 4A and B). FKD cell migration showed a statistical difference from parent and non-FKD cells (Tukey-Kramer test, P<0.01).

In 2D LV-SEM observation, non-FKD cell sheet exhibited loose cell-cell connections (Fig. 5A), whereas the cells of FKD cell sheet were observed as groups with tight cell-cell connections (Fig. 5B). Bundles of extremely thin microfibrils (filopodia) existed on the surface of leading migratory non-FKD cells (Fig. 5C). By contrast, globular-shaped lamps of different sizes were observed on the surface of FKD cells (Fig. 5D).

Fascin expression (green; Fig. 6) was strongly positive in non-FKD cells, which exhibited thin filopodia on the cell membrane (arrows). Meanwhile, fascin expression was effectively suppressed in FKD cells, and globular-shaped lamps were observed on the cell membrane (arrowheads). In LV-SEM, numerous lamps were recognized as whitish spots on the surface of FKD cells (arrowheads). Thus, fascin-positive spots observed by optical microscope were demonstrated to locate at the whitish spots in the bulbous-shaped protrusions on the FKD cell surface in LV-SEM.

Non-FKD and FKD cells were cultured in Cellmatrix^® for 10 days. Following HE staining, non-FKD cells were observed to have loose cell-cell adhesions, whereas clusters with cell-cell connections were observed in FKD cells (Fig. 7). E-cadherin immunohistochemical staining was negative in non-FKD cells but positive in FKD cells. Snail1 expression (red) was observed in the nucleus of non-FKD cells but was decreased in FKD cells; however, vimentin was strongly positive in both cell lines (green). E-cadherin expression was significantly different between non-FKD and FKD cells, and Snail1 expression as well (P<0.01).

Numerous thin cell protrusions (filopodia) were observed on the surface of non-FKD cells (Fig. 8). These cells exhibited loose cell-cell adhesions. By contrast, bulbous-shaped protrusions of varied sizes were detected on the surface of FKD cells, which exhibited partial cell-cell adhesions.

Densely aggregated tumor cells with irregularly shaped nuclei were observed in both non-FKD and FKD cells (Fig. 9). Fascin expression was strongly positive in non-FKD cells. Arrows indicate E-cadherin positive-stained membranes observed in FKD cells.

In 3D LV-SEM observation, non-FKD spheroid cells gathered densely forming a spherical spheroid block, by contrast, cells of the FKD spheroid gathered sporadically and eventually formed an irregularly shaped spheroid block (Fig. 10). There were numerous cell protrusions (filopodia) on the surface of non-FKD spheroid cells, in contrast, globular-shaped protrusions of varied sizes were observed on the surface of FKD spheroid cells. The morphological appearances of parent cells were similar to non-FKD cells.

In stereoscopic microscope observation, homogeneous filamentous cells were observed infiltrating into the surrounding gel from the non-FKD spheroid (Fig. 11A), while irregular heterogeneous fascicular cell invasion into the gel was observed from the FKD spheroid (Fig. 11B). HE staining revealed fibrous spindle-shaped tumor cells invading the surrounding gel from the non-FKD spheroid (Fig. 11C), while clusters of tumor cells (arrows; Fig. 11D) originating from the FKD spheroid were observed invading the gel.