Specimens used for IHC were prepared at the Kanagawa Cancer Center Research Institute from archives of surgically removed and formalin-fixed, paraffin-embedded breast cancer tissues in the Pathology Department. With the approval of the Research Ethics Committee, these prepared specimens were used in the present study through the Kanagawa Cancer Research and Information Association, which has since been dissolved and its duties transferred to the Kanagawa Cancer Center Research Institute Biospecimen Center (approval no. 3-2009). The clinicopathological data of the patients from whom the samples were obtained are summarized in Table SI. TNM stage data is lacking 27% because the data is already anonymized. The research protocol used was also approved by the Institutional Ethics Committees of Tokyo University of Science (approval nos. 13003, 15006 and 16038), and all patients provided consent for the use of their tissue samples for research purposes.

IHC was performed as previously described (29,31–35,39,40). Briefly, 4-µm thick paraffin embedded sections were deparaffinized, rehydrated in a descending series of ethanol solutions and autoclaved (120°C for 20 min) in 10 mmol/l citrate buffer (pH 6.0) for antigen retrieval. The semi-serially prepared sections (adjacent sections) were then immersed in 0.3% hydrogen peroxide at room temperature for 30 min to quench the intrinsic peroxidase activity before incubation with a primary antibody at 4°C overnight. The antibodies used in the present study were: Mouse anti-PKCι mAb (1:250; cat. no. 610176; BD Biosciences), mouse anti-GLO 1 mAb (1:2,000; cat. no. NBP1-19015; Novus Biologicals, Inc.), mouse IgG2b κ Isotype Control (eBMG2b; 1:500; cat. no. 14-4732-82; eBioscience; Thermo Fisher Scientific, Inc.) and mouse IgG1 κ Isotype Control (P3.6.2.8.1) (1:1,000; cat. no. 14-4714-82; eBioscience; Thermo Fisher Scientific, Inc.). The labeled antigens were visualized using a Histo Fine kit (Nichirei) and DAB plus (Dako; Agilent Technologies, Inc.). The sections were counterstained with hematoxylin. The antibodies used for double staining were: Mouse anti-PKCι mAb (1:50), rabbit anti-GLO 1 pAb (1:200; cat. no. A1932; ABclonal, Inc.), mouse IgG2b κ Isotype Control and normal rabbit IgG (1:952; cat. no. PM035; MBL). The labeled antigens were visualized using a Histo Fine alkaline phosphatase kit and DAB plus. The sections were counterstained with hematoxylin.

To evaluate the expression of GLO 1 and PKCλ proteins using IHC, ImageJ version 1.51u was used (National Institutes of Health) with the IHC Profiler plugin (52). The scoring system was based on the classification calculated from the IHC Profiler (+3, high-positive; +2, positive; +1, low-positive; and 0, negative). Signal intensity of GLO 1 was classified into color density as follows; +3, High positive; +2, Positive; and +1, Low positive. Signal intensity of PKCλ was classified into color density as follows: +3, High positive; +2, Positive; +1, Low positive; and 0, Negative. Signal intensities were categorized as high (+3 or +2) or low (+1 or 0). H-scores of the scatter plot data were based on calculated values from the IHC Profiler.

Gene expression data was downloaded from cBioportal and analyzed as previously described (21,41,42,53,54). Briefly, the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) dataset (55,56) was downloaded from cBioPortal (cbioportal.org/; last entry, 25th November 2019) (57,58). The clinicopathological data of the patients are summarized in Table SII. The median age at the time of diagnosis was 61.8 years (age range, 21.9–96.3 years). Gene expression levels were classified as high if they were in the top 25% of Z-scores; or otherwise, they were classed as low.

Briefly, gene expression microarray datasets from TCGA were downloaded from Oncomine (oncomine.org; Compendia Bioscience, 28th January 2021) (59,60), and the breast cancer dataset (n=459) was obtained. Levels of GLO 1 (reporter, A_32_P53822) and PKCλ (reporter, A_23_P18392) mRNA expression are presented using log₂ median-centered ratio boxplots for normal vs. cancerous tissues.

The MCF-10A human normal-like (non-transformed) mammary epithelial cell line and the MDA-MB-157 and MDA-MB-468 human basal-like breast cancer cell lines were obtained from the American Type Culture Collection. MCF-10A cells were grown in mammary epithelial cell growth medium (MEGM; Lonza Group, Ltd.) according to instructions from ATCC. The cancer cell lines were cultured as previously described (21,41,42,53). Mycoplasma testing was performed on all the cell lines used.

3-(1,3-Benzothiazol-2-yl)-4- (4-methoxyphenyl) but-3-enoic acid (TLSC702) was purchased from Namiki Shoji Co., Ltd. and dissolved in DMSO. Aurothiomalate (ATM) was purchased from Calbiochem (Merck KGaA) and dissolved in water.

Immunoblotting was performed as previously described (21,41,42,53). The primary antibodies used were: Mouse anti-PKCι mAb (1:5,000; cat. no. 610176; BD Biosciences), mouse anti-GLO 1 mAb (1:2,000; cat. no. sc-133144, Santa Cruz Biotechnology, Inc.) and mouse anti-β-actin mAb (1:20,000; cat. no. 60008-1-Ig, ProteinTech Group Inc.). The secondary antibody used was a goat anti-mouse IgG horseradish peroxidase-conjugate (1:5,000; cat. no. 7076S; Cell Signaling Technology, Inc.).

WST-8 assays were performed according to the manufacturer's protocol, and as previously described (21,42,53). Briefly, cells (5×10³/well) were seeded into 96-well plates (Sigma-Aldrich; Merck KGaA) and incubated for 24 h. Inhibitors were then added to the culture medium, and the cells were incubated for an additional 3 days, after which cell viability was assessed using a Cell Counting Kit-8 assay (Dojindo Molecular Technologies, Inc.). The formazan dye formed was measured using Sunrise Remote (Tecan Group, Ltd.) at 450 nm. Assays with MCF-10A cells were performed in MEGM supplemented with 10% FBS. Assays using the cancer cell lines were performed in DMEM supplemented with 10% FBS. Numerical values of the test groups were expressed relative to the control cell (no drug).

Tumor-spheres were grown as previously described (21,41,42,53). Briefly, cells (1×10³/well) were cultured in 96-well ultralow attachment plates (Greiner Bio-One) and treated with inhibitors for 6 days. Images were taken through an inverted microscope (Leica Microsystems, Inc.), and the numbers of tumor-spheres ≥50 µm in diameter were counted. Numerical values of the test groups are shown relative to the untreated cells. Cell Titer-Glo^® luminescence assays (Promega Corporation) were performed according to the manufacturer's protocol, and as previously described (21,53). Values for test groups are shown relative to cells in the absence of the drug.

For correlation between protein GLO 1 and PKCλ expression, statistical significance was calculated using the χ²-test with Yates' correction. H-scores of the scatter plot data were based on calculated values from the IHC Profiler. Spearman's rank correlation coefficients (r) and P-value are indicated. In the analysis of gene expression, Pearson's correlation coefficients (r) and P-value are indicated. P-values were calculated using a test for non-correlation. In the analysis of gene expression, survival curves were plotted using the Kaplan-Meier method, and P-values were calculated using the Gehan-Breslow generalized Wilcoxon test to weight early death points. Multiplicity was adjusted using the Holm's method for post-hoc analysis. A multivariable Cox regression model was used to evaluate the effect of gene expression and to estimate the adjusted hazard ratios (HRs) with age as a confounding factor. P-values for comparison of gene expression are presented using the Kruskal-Wallis test with the Steel-Dwass test. Statistical analysis was performed using BellCurve for Excel version 2.11 (SSRI). Data for the WST-8 assay is presented as the mean ± standard deviation of three independent experiments. Differences between groups were compared using Tukey's test. Data for the tumor-sphere assay is presented as the mean ± standard error of the mean of three independent experiments. Data for the Cell Titer-Glo^® luminescence assay is presented as the mean ± standard deviation of three independent experiments. Statistical significance was calculated using one-way ANOVA followed by Dunnett's test. For any of the analyses above, α-level was fixed at 0.05, and P<0.05 was considered to indicate a statistically significant difference.

GLO 1 protein expression in both the cytosol and nucleus of breast cancer cells was detected, and the results were consistent with earlier observations (Fig. 1A) (20). There were no 0 image samples for GLO 1 protein. GLO 1 protein expression was detected in all breast cancer samples, and the results were consistent with a previous report (20). PKCλ was also localized in the cytosol and nucleus of breast cancer cells (Fig. 1B) as reported in our previous study (29). Double IHC staining showed that GLO 1 and PKCλ were colocalized in breast cancer cells (Fig. 1C). To evaluate the relationship between GLO 1 and PKCλ, their signal intensities were quantified (Fig. 1A and B). High expression of GLO 1 was significantly correlated with high expression of PKCλ in breast cancer tissue (Fig. 1D, P<0.01, χ²-test; Fig. 1E, r=0.71, P<0.01).

To further examine the relationship between GLO 1 and PKCλ at the mRNA level, mRNA expression data from the METABRIC breast cancer dataset (n=1,904) was downloaded from cBioPortal (Fig. 2A). In our previous study, it was shown that GLO 1 expression is higher in breast cancer compared with normal tissue samples (21). In addition, GLO 1 gene expression was enhanced at all breast cancer tumor stages compared with normal tissue samples. Similarly, PKCλ expression was significantly higher in breast cancer tissues and at all tumor stages compared with the normal tissues (41) (Fig. S1). Scatter plot analysis indicated that GLO 1 expression was weakly correlated with PKCλ expression in all patients and in patients with early-stage (stage 0-II) tumors (Fig. 2A), but was not correlated with PKCλ expression in patients with late-stage (stage III–IV) tumors (Fig. 2A). As shown in Table SIII, none of the clinicopathological parameters were correlated with high GLO 1 and PKCλ gene expression. Nonetheless, these results, together with the IHC findings, suggest that GLO 1 and PKCλ may be cooperatively involved in certain cases of breast cancer.

Next, we examined weather high GLO 1 and PKCλ expression is correlated with prognosis (Figs. 2B-D and S2; Tables I and SIV). Examination of the prognosis of patients with GLO 1^high and PKCλ^high breast cancer revealed that GLO 1^high was not associated with a poorer OS amongst all patients (Fig. 2B) (21). It was previously shown that overexpression of PKCλ and its signaling promotes TNBC growth and metastasis (30). Consistent with this finding, patients classed as PKCλ^high had worse OS (Fig. 2B; P<0.01) (42). In our previous study, it was shown that patients with stage III or IV cervical cancer with high PKCλ expression had a worse clinical outcome (32). Thus, tumors were classified into early-stage (stage 0-II) and late-stage (stage III–IV) tumors, and the difference in survival was assessed using Kaplan-Meier analysis. Patients with early-stage tumors (GLO 1^high, PKCλ^high or GLO 1^high PKCλ^high) exhibited similar clinical outcomes to those expressing lower levels of these genes (Fig. 2C), as did those with late-stage tumors (GLO 1^high; Fig. 2D). However, patients with late-stage tumors (PKCλ^high) had a worse OS (Fig. 2D) (42). Furthermore, patients with GLO 1^high PKCλ^high tumors had poorer prognoses than patients with GLO 1^high PKCλ^low (adjusted by Holm's method; P=0.015) and GLO 1^low PKCλ^low (P=0.040), but not GLO 1^low PKCλ^high (P=0.83) tumors. Multivariable Cox regression analysis revealed that GLO 1^high PKCλ^high was associated with poorer clinical outcomes than GLO 1^low PKCλ^low (Table I; HR=2.36; 95% CI, 1.08–5.16; P=0.03) and GLO 1^high PKCλ^low (Table I; HR=3.25, 95% CI, 1.26–8.35; P=0.01) in those with late-stage tumors. Furthermore, normal-like breast cancer patients classified as GLO 1^high PKCλ^high also exhibited a worse prognosis (Fig. S2; adjusted by Holm's method; P=0.016 and Table SIV; HR=9.11, 95% CI, 2.07–40.15; P<0.01). Thus, high PKCλ expression, regardless of GLO 1 expression level, contributes to poor clinical outcome. Of note, GLO 1 protein (Fig. 1) and mRNA (Fig. S1) expression was considerably higher in breast cancer, reflecting enhanced glycolysis. Therefore, GLO 1 and PKCλ may function cooperatively to promote cancer progression, and contribute to worse clinical outcomes in patients with late-stage breast cancer.

Amongst patients with basal-like breast tumors, the proportion of patients classed as GLO 1^high PKCλ^high was larger than the proportion classed as GLO 1^low PKCλ^low, irrespective of whether they had early- or late-stage tumors (Fig. S3). All patients with breast cancer classed as GLO 1^high PKCλ^high primarily had luminal B type breast cancer (43.1%, 59/137), whereas those with GLO 1^low PKCλ^low primarily had luminal A type breast cancer (42.3%, 459/1,084) (Fig. S3). Similarly, amongst patients with early-stage lesions, those classed as GLO 1^high PKCλ^high primarily had luminal B breast cancer (48.4%, 46/95), and those with GLO 1^low PKCλ^low primarily had luminal A breast cancer (45.2%, 333/737). However, amongst patients with late-stage lesions, those classed as GLO 1^high PKCλ^high exhibited higher incidences of luminal A or basal-like type breast cancer (luminal A, 33.3%, 3/9; basal-like, 33.3%, 3/9) compared with patients classed as GLO 1^low PKCλ^low (luminal A, 28.6%, 18/63; basal-like, 6.3%, 4/63). As shown in Fig. S4, 38% (76/199) of patients with basal-like type cancer were classed as GLO 1^high PKCλ^high. In our previous study, it was shown that GLO 1 expression is upregulated in basal-like breast cancer, and that inhibition of GLO 1 suppresses basal-like breast cancer cell viability (21). To further clarify the effects of inhibiting GLO 1 and PKCλ in basal-like breast cancer cells, MCF-10A human normal-like (non-transformed) mammary epithelial cells were compared with MDA-MB-157 and MDA-MB-468 human basal-like breast cancer cells (Fig. 3) (21). Expression levels of both GLO 1 and PKCλ were higher in MDA-MB-157 and MDA-MB-468 cells than in MCF-10A cells (Fig. 3A). TLSC702 has previously been shown to inhibit GLO 1 activity and induce MG accumulation and apoptosis in cancer cells (21,61). Based on the inhibitory effects of TLSC702 on the viability of MCF-10A, MDA-MB-157 and MDA-MB-468 cells, concentrations of 50, 75 and 100 µM TLSC702 were used in the present study (21). To inhibit PKCλ, ATM was used (5 and 10 µM), which interferes with the PB1-PB1 domain interactions between PKCλ and Par6 and induces apoptosis (62,63). As a result of the inhibitory effect of 0.5, 1, 5 and 10 µM ATM on the colony formation ability of MDA-MB-157 cells, it was found that 5 and 10 µM ATM markedly suppressed colony formation compared with the untreated control group (unpublished data). The inhibitor concentrations used in the present study were based on these findings. WST-8 assays showed that MCF-10A cells were less sensitive to GLO 1 inhibition than the two cancer cell lines, consistent with our previous study (Fig. 3B; TLSC702 50 and 75 µM) (21). By contrast, inhibition of PKCλ using ATM did not significantly affect the cell viability of any of the three cell types assessed (Fig. 3B). In addition, the combination of TLSC702 and ATM decreased the viability of the two cancer cell lines, which was reduced to a greater degree than that of the control cells (TLSC702/ATM, 50/10, 75/5, 75/10, 100/5 and 100/10 µM; Fig. 3B).

In vitro tumor-sphere formation was assessed using MDA-MB-157 and MDA-MB-468 cells to determine the roles of GLO 1 and PKCλ in the tumorigenicity of basal-like breast cancer cells. As shown in Fig. 4A and B, treatment with TLSC702 or ATM reduced the relative numbers of tumor-spheres ≥50 µm in diameter. Furthermore, a combination of TLSC702 and ATM inhibited the formation of tumor-spheres ≥50 µm in diameter (Fig. 4A and B; P<0.05). Cell Titer-Glo^® assays also showed that the combination of TLSC702 and ATM suppressed the viability of MDA-MB-157 and MDA-MB-468 cells (Fig. 4C and D; P<0.05). Taken together, these results along with the results of the viability assays suggest that GLO 1 and PKCλ are cooperatively involved in cancer progression and survival of basal-like breast cancer cells.