ST8SIA1 Regulates Tumor Growth and Metastasis in TNBC by Activating the FAK–AKT–mTOR Signaling Pathway

Breast cancer stem-like cells (BCSC) comprise a small proportion of cells present in primary or metastatic breast tumors, but they are strongly tumorigenic, quiescent, chemoresistant, highly metastatic, and promote tumor relapse (1–3). To successfully develop BCSC-targeted therapy, molecular markers that are specific to BCSCs must be identified. Several cell-surface and intracellular markers have been discovered to prospectively isolate BCSCs from primary breast tumors and the CD44^highCD24^low phenotype was reported to be specific for BCSCs (4). In addition, aldehyde dehydrogenase enzyme activity has been shown to distinguish BCSCs within primary breast tumors (5). We have recently identified the ganglioside GD2 as a single cell-surface marker characterizing BCSCs in cell lines and primary tumors (6). GD2 is coexpressed on the CD44^highCD24^low fraction of breast cancer cells, and its expression has been shown to be higher in basal-like breast cancer cell lines than in their luminal counterparts (6). GD2⁺ cells are highly tumorigenic in vivo; as few as 10 GD2⁺ cells can initiate tumors in NOD/SCID mice.

GD2 is synthesized by GD2 synthase (B4GALNT1) with GD3 as precursor, whereas GD3 is synthesized by GD3 synthase (ST8SIA1) with GM3 as precursor. Using gene-expression arrays, we found that ST8SIA1 but not B4GALNT1 is highly expressed in GD2⁺ BCSCs. These findings were confirmed by other groups, which also found upregulation of ST8SIA1 in BCSCs (7, 8). Although GD2 has been shown to be highly expressed in basal-like breast cancer cell lines compared with their luminal counterparts, expression levels of ST8SIA1 in other breast cancer subtypes have not been studied. We have previously shown that knockdown of ST8SIA1 inhibited BCSC function, including anchorage-independent growth, migration, and invasion (6). In addition, induction of epithelial-to-mesenchymal transition (EMT), which induced the CD44^highCD24^low BCSC phenotype, also induced ST8SIA1 expression in breast cancer cells (6). Moreover, ST8SIA1 overexpression in breast cancer cell lines in turn induced EMT and metastasis in vivo (9). We very recently showed that ST8SIA1 expression is regulated by NFκB signaling (10). Inhibition of NFκB activation subunits such as IKKα/β inhibited ST8SIA1 expression and BCSC function in vitro and inhibited tumor growth and metastasis in vivo. However, the mechanism underlying ST8SIA1 knockdown-mediated tumor growth inhibition is unknown.

Focal adhesion kinase (FAK) signaling has been well characterized for its role in cell adhesion, migration, and metastasis (11). Inhibition of FAK signaling hinders migration and invasion of cancer cells. However, the molecular interactions between GD2 and FAK in the regulation of BCSC function including tumor initiation and metastasis have not been investigated. GD2 has been shown to bind cell-surface receptors, including integrins, in the extracellular matrix and to activate downstream FAK signaling in cancer cells (12, 13). GD2 acts as a coreceptor for other cell-surface receptors, including cMet and EGFR, to activate downstream signaling (8, 14). FAK regulates cellular activities including cell survival, adhesion, proliferation, and migration by activating several signal transducers, including PI3K, AKT, mTOR, RAS, and ERK/MAPK (15–20). These proteins further activate their downstream target genes to regulate cellular functions.

The purpose of this study was to determine the expression of ST8SIA1 in triple-negative breast cancer (TNBC) and its role in the progression of these cancers. Our results show that ST8SIA1 is highly upregulated in basal-like TNBC primary tumors and regulates breast tumor growth and metastasis. We also found signaling pathways that are highly activated in GD2⁺ BCSCs and that these signaling events are dependent on ST8SIA1 expression.

The human breast cancer cell lines SKBR3, ZR75-1, MDA361, BT474, MCF7, MDA453, T47D, SUM149, HCC38, MDA468, MDA-MB-231, SUM159, and HS578T were cultured in complete cell culture medium comprising Dulbecco modified essential medium (DMEM) containing 10% fetal bovine serum (FBS) and 1% penicillin and streptomycin with glutamine. Cells were routinely checked for mycoplasma using the PCR Mycoplasma Detection Kit (Applied Biological Materials). All the cell lines are fingerprinted routinely every 6 months at cell line core facility at MD Anderson for purity.

Flow cytometry staining of MDA-MB-231, SUM159, and their respective ST8SIA1-knockout (KO) cells was performed using a previously described method (6). In summary, cells (∼1 × 10⁶) were incubated with allophycocyanin (APC)-conjugated anti-GD2 antibody (BioLegend) for 30 minutes on ice in the dark. For measuring the expression of ceramide, GM1, and GM3, the cells were stained using anticeramide (Enzo Life Sciences) or anti-GM1 (Abcam) or anti-GM3 (Cosmo Bio USA) unconjugated antibodies. After washes, the cells were incubated with APC conjugated anti-mouse or anti-rabbit IgG secondary antibody. The cells were washed once with FACS buffer containing 4′,6-diamino-2-phenylindole (DAPI; 1 μg/mL). The cells were analyzed by an LSR II (BD) or a Galios (Beckman Coulter) flow cytometer. Cell sorting was performed on the FACSAria II (BD) as described before (10).

A specially designed gRNA targeting the ST8SIA1 gene was developed and purchased from (GenScript Biotechnologies). The sequence of the gRNA is underlined in Supplementary Fig. S3. To generate the ST8SIA1-KO cells, SUM159 and MDA-MB-231 cells were transfected with 1 μg of gRNA specific to ST8SIA1 and 1 μg of Cas9 plasmid (Systems Biosciences) using jetPRIME transfection reagent (Polyplus). After 72 hours of transfection, the cells FACS-sorted into 96-well plates (one cell per well) to generate homogeneous KO colonies. Colonies were expanded and analyzed for GD2 expression by flow cytometry.

Low-passage Cas9-Control and ST8SIA1-KO (below 5) were cultured in standard media. Once confluent, 3 × 10⁶ cells were trypsinized and pelleted for processing. DNA was then extracted using DNEasy kit from Qiagen. The gene of interest was then amplified using the Thermo Fisher Scientific PlatinumTaq DNA Polymerase kit along with custom-designed forward and reverse primers (from 5′ to 3′, respectively, TGT GGT ATG ACG GGG AGT TT and ACC TTT GCC GAA TTA TGC TG; Integrated DNA Technologies) and purified through gel electrophoresis. DNA samples of Cas9-Control and ST8SIA1-KO cells were then sequenced by Sanger-sequencing at the MD Anderson Sequencing Core facility.

RT-PCR was performed using the TaqMan gene-expression assay from Applied Biosystems as previously described (10). The primers used were as follows: GAPDH: Hs002758991_g1, B2M: Hs00187842_m1, ST8SIA1: Hs00268157_m1, BCL11A: Hs01093197_m1, PDGFRß: Hs01019589_m1, VCAM1: Hs01003372_m1, PDGFRα: Hs00998026_m1, DKK1: Hs00183740_m1, SOX2: Hs01053049_s1, FOXG1: Hs01850784_s1, CXCR4: Hs00607978_s1, CXCL12: Hs03676656_mH, and FGFR2: Hs01552926_m1. Relative mRNA expression was normalized to GAPDH.

To investigate signaling pathways that are activated in BCSCs, GD2⁺ and GD2⁻ SUM159 cells were subjected to FACS and then to proteomic profiling by Kinexus antibody array analysis. Lysates of GD2⁺ and GD2⁻ cells were labeled with fluorescent probes and applied to glass slides coated with 850 validated antibodies against various total and phosphorylated proteins per the manufacturer's instructions. After washes, the stain was developed and analyzed on a phosphoimager. The z-ratio was calculated as previously described (10), and proteins differentially expressed or phosphorylated between GD2⁺ and GD2⁻ cells were identified.

Protein lysates were prepared by pelleting cells (3 × 10⁶) and subjecting them to lysis using 200 μL of Mammalian Cell Lysis Buffer (G-Biosciences) and 1 μL of protease and phosphatase inhibitor Cocktail (Roche Diagnostics). The lysates were subjected to polyacrylamide gel electrophoresis and then transferred to PVDF membrane. The membranes were then blocked with 5% milk and incubated with primary (listed in Supplementary Table S3) and fluorescent-tagged secondary antibodies. After washes, the membranes were scanned using Odyssey fluorescent imager/scanner (LI-COR).

Cas9-Control or ST8SIA1-KO MDA-MB-231 and SUM159 cells were plated at 1 × 10³ per well in ultra-low attachment 24-well plates (Corning) containing 1 mL per well mammosphere growth medium (Promocell). To test the effect of FAK and mTOR signaling inhibition on mammosphere formation, MDA-MB-231 and SUM159 cells were treated with PF-573228 (Sigma) or everolimus (Chemscene). The mammospheres were stained and counted as described before (10). PF-573228 structure and method of use was previously described by Slack-Davis JK and colleagues (21).

In vitro analysis tumorigenesis potential was determined by soft-agar colony assay. ST8SIA1-KO and Cas9 control SUM159 and MDA-MB-231 cells or cells treated with FAK an mTOR inhibitor were used for this assay. Cells (5 × 10³) per well were cultured in complete cell culture medium containing 1% and 0.5% low-melting agarose in a dual layer setting. The cells were cultured at 37°C with 5% CO₂ for 3 weeks. The colonies were stained and imaged as described before (10).

Migration assays were performed using 24-well transwell chambers, using membrane inserts with 8-μm pores (Corning) as described before (10). Briefly, Cas9-Control and ST8SIA1-KO SUM159 or MDA-MB-231 cells in DMEM containing 1% FBS were seeded (1 × 10⁴ per chamber) into the upper chambers. After incubating for 8 hours at 37°C in a 5% CO₂ incubator, each membrane was carefully removed from the insert, fixed with 2% formalin, and stained with DAPI. Fluorescent images were captured using the EVOS automated microscope (Thermo Fisher Scientific).

All experiments involving animals were approved by and conducted in accordance with the policies of the Institutional Animal Care and Use Committee (IACUC) of The University of Texas MD Anderson Cancer Center. Viable Cas9-Control or ST8SIA1-KO cells (3 × 10⁶) were resuspended in ECMatrix solution (EMD Millipore) and implanted into the mammary fat pads NSG mice (n = 20; 10 mice per group). Tumor volumes were measured weekly by using calipers. When the tumors reach 2 cm in diameter, the animal was humanely euthanized per IACUC guidelines. The animal survival curves were generated by the log-rank test. After euthanasia, the tumors were harvested from both groups for immunohistochemical (IHC) studies.

Cas9-Control and ST8SIA1-KO MDA-MB-231 cells were subjected to trypsinization and suspended in PBS (5 × 10⁶ cells/mL). NSG mice were injected with cells (5 × 10⁵ in 100 μL volume per mouse) via tail vein (n = 18; 9 mice per group). Beginning week 3, three mice randomly selected from each group were sacrificed weekly; body organs, including lungs, liver, and femur, were isolated and subjected to histologic analysis with hematoxylin and eosin (H&E) staining. The slides were then scanned on the EVOS automated microscope (Invitrogen).

Unless otherwise indicated, data shown represent the mean and ± standard error of the mean.

A total of 1,093 breast cancer samples with 112 normal matched tissues were investigated. Of the breast cancer samples, 109 were identified as TNBC from the clinical information provided. Based on PAM50 classification, another 232 were luminal A, 129 luminal B, 98 basal, and 58 HER2. The data for these breast cancer subtypes were also analyzed. RNA-seq data were quantile normalized to remove the sample variability and any potential batch effects. Differential expression of ST8SIA1 between groups was identified by using the Student t test or ANOVA with Tukey HSD post hoc test in the R software (version 3.2.0). Differential expression of ST8SIA1 in mutated genes versus nonmutated genes was tested by the Student t test and the resulting P values and fold changes are reported.

Samples were sequenced on the HiSeq Sequencing System by the Sequencing and Microarray Core facility at MD Anderson. Sequence reads were mapped to human genomics (build hg19) with bowtie2 aligner using RSEM software (22, 23). The EdgeR package in R was used to compare the differential expression between ST8SIA1-KO samples and controls (24). Genes with adjusted P values less than 0.05 and absolute fold changes larger than 2 were considered significant. To investigate the relationships between ST8SIA1 and the significant genes identified from our KO experiment, we used the Pearson correlation in the breast cancer expression data obtained from TCGA. A scatterplot with log₂ fold change from expression in KO and TCGA correlation coefficients was plotted.

Development of ST8SIA1-targeted therapy for breast cancer necessitates investigation of ST8SIA1 expression in different breast cancer subtypes. To investigate the expression of ST8SIA1 in primary breast tumors, we analyzed RNA-seq data from over 1,100 primary and metastatic breast tumors and adjacent normal tissues from the TCGA database and found that ST8SIA1 was overexpressed in about 11% of these breast cancers (Fig. 1A). Interestingly, the estrogen receptor (ER)–negative and progesterone receptor (PR)–negative phenotype in these breast tumors was highly correlated with ST8SIA1 overexpression (P < 0.001 for both ER and PR), whereas HER2 overexpression was not (Fig. 1A). Basal-type tumors expressed the highest levels of ST8SIA1 of all other types, including luminal A, luminal B, or HER2-enriched tumors (all P = 0.001; Fig. 1B). Moreover, primary TNBC tumors (n = 115) had 4.63-fold higher expression levels of ST8SIA1 than non-TNBC tumors (n = 852, P < 0.001; Fig. 1C). A survival analysis by the log-rank test indicated that patients with ST8SIA1^high tumors had significantly lower rates of disease-free (P = 0.0109; Fig. 1D) and overall (P = 0.0148; Fig. 1E) survival than patients with ST8SIA1^low tumors. Indeed, in cell lines, we validated these findings in luminal or basal cell–derived breast cancer cell lines (n = 10) and found that ST8SIA1 expression was, on average, 50-fold higher in basal subtype breast cancer cell lines than in luminal subtypes (Supplementary Fig. S1). These data suggest that ST8SIA1 expression is associated with the basal-like and TNBC tumors and that high levels of ST8SIA1 in tumors may lead to shorter survival in breast cancer patients.

To investigate the relationships between ST8SIA1 expression and genes frequently mutated in breast cancer, the top 20 most frequent mutations in the TCGA breast cancer data set were analyzed for correlation with ST8SIA1 mRNA expression (Fig. 2A). Of these 20 genes, TP53 had the strongest positive correlation with ST8SIA1 expression (P < 0.00001); ST8SIA1 expression was >2-fold higher in TP53-mutated than in TP53–wild-type tumors (Fig. 2B). Another positively correlated mutation was for the nuclear envelope protein Spectrin Repeat Containing Nuclear Envelope 1 (SYNE1; P < 0.05; Fig. 2B). On the other end of the spectrum, GATA3 mutations were the most negatively correlated (P < 0.001) with ST8SIA1 expression (Fig. 2B). GATA3 plays a role in epithelial cell differentiation in the mammary gland and is a major tumor suppressor, supporting the notion that ST8SIA1 is a stem cell–associated gene.

To identify the genes that are coexpressed with ST8SIA1, we analyzed TCGA data in the breast cancer data set. Pearson correlation coefficients (Pearson r) > 0.5 were considered positively correlated, while those with Pearson r < −0.5 were considered negatively correlated. We found that 139 genes were positively and 108 genes negatively correlated with ST8SIA1 (Supplementary Fig. S2). Among these genes, those associated with BCSC function include BCL11A (ref. 25; Pearson r = 0.645) and FOXC1 (refs. 26, 27; Pearson r = 0.61), both of which were highly positively correlated with ST8SIA1 expression (Fig. 2C). In addition, the tumor suppressor gene GATA3 (ref. 28; Pearson r = −0.583) and its closely associated protein FOXA1 (ref. 29; Pearson r = −0.615) were highly negatively correlated with ST8SIA1 (Fig. 2C). These data suggest that ST8SIA1 has a strong positive correlation with p53 mutations but a negative correlation with GATA3 mutations. Finally, we found that ST8SIA1 expression was positively correlated with genes that promote tumor growth and negatively associated with genes that inhibit tumor growth.

To further investigate the mechanism of ST8SIA1-mediated regulation of BCSC function, we used a CRISPR/Cas9 gene KO approach to generate ST8SIA1-KO SUM159 and MDA-MB-231 TNBC cells. DNA sequencing of SUM159 and MDA-MB-231 cDNA identified a single-nucleotide insertion into exon 3 of the ST8SIA1 gene (Supplementary Fig. S3A and S3B). To validate that KO of ST8SIA1 was successful, GD2 expression of ST8SIA1-KO SUM159 and MDA-MB-231 cells was analyzed by flow cytometry, which was <1% for both SUM159 and MDA-MB-231 KO cells, whereas 17.3% ± 4% and 12.5% ± 3% for their respective Cas9-Control cells (Supplementary Fig. S4A and S4B). Interestingly, knockout of ST8SIA1 induced the cell-surface expression of the upstream gangliosides including its substrate GM3 as well as GM1 and ceramide in SUM159 cells (Supplementary Fig. S4C). These data further demonstrate that knockout of ST8SIA1 completely blocks GD2 expression and therefore upstream gangliosides are accumulated on cell surface. Cell growth analysis in vitro did not show any significant effects on cell proliferation between control and ST8SIA1-KO cells for the first 10 to 15 passages (Supplementary Fig. S4D). Therefore, we performed all the in vitro and in vivo functional assays involving ST8SIA1-KO cells before they reached passage 10.

To investigate the effects of ST8SIA1 deletion on cancer stem cell activity, Cas9-Control and ST8SIA1-KO SUM159 and MDA-MB-231 cells were subjected to a series of in vitro functional assays, including soft-agar colony assays (tumorigenesis), mammosphere formation assays (anchorage-independent growth), and transwell assays (cell migration). In SUM159 cells, ST8SIA1 KO inhibited in vitro tumorigenesis by >30-fold, mammosphere formation by ∼10-fold, and cell migration by 2-fold compared with Cas9-Control SUM159 cells (Supplementary Fig. S5A). In MDA-MB-231 cells, ST8SIA1 KO inhibited in vitro tumorigenesis by 3-fold, mammosphere formation by ∼12-fold, and cell migration by 3-fold compared with Cas9-Control MDA-MB-231 cells (Supplementary Fig. S5B and S5C). These data suggest that inhibition of ST8SIA1 negatively affects BCSC function in TNBC cells.

To better understand the mechanism of ST8SIA1 regulation of BCSC function, we performed RNA-seq analysis on ST8SIA1-KO and Cas9-Control SUM159 cells (triplicate samples for both cell types). After setting the parameters to P < 0.05 and absolute fold change >2, 1,502 genes were found to be downregulated and 842 genes upregulated in ST8SIA1-KO cells compared with Cas9-Control cells (Fig. 3A). To investigate the oncogenic signature affected by ST8SIA1 KO, we performed gene set enrichment analysis (GSEA) and found that many major oncogenic pathways, including AKT, EGFR, ERK, Wnt, p53, RB, KRAS, PDGF, and MEK, were downregulated in ST8SIA1-KO cells compared with Cas9-Control cells (Supplementary Table S1). Further, a correlation analysis was performed for genes that were differentially expressed in ST8SIA1-KO cells (x axis) and genes that were positively or negatively correlated with ST8SIA1 in the TCGA patient data set (y axis, Fig. 3B). Interestingly, 65% of the genes that are positively correlated with ST8SIA1 in TCGA data were also downregulated in ST8SIA1-KO cells (red box, Fig. 3B), suggesting that these genes are dependent on ST8SIA1.

To validate the inhibition of gene expression in ST8SIA1-KO cells, we performed qRT-PCR analysis and found that genes that are associated with stem cell functions, including BCL11A, PDGFR-α, PDGFRß, VCAM1, FOXC1, CXCR4, CXCL12, SOX2, FOXG1, FGFR2, and WNT5A, were indeed downregulated in ST8SIA1-KO cells compared with Cas9-Control cells (Fig. 3C). DKK1, an antagonist of WNT-ß-Catenin signaling, was upregulated in ST8SIA1-KO cells, suggesting an inhibition of stemness in KO cells (Fig. 3C). These data indicate that ST8SIA1 not only is coexpressed with BCSC-associated genes but also regulates their expression, suggesting that targeting ST8SIA1 inhibits BCSC-associated genes in TNBC tumors.

To better define the mechanism by which ST8SIA1 regulates BCSC functions, we performed proteome profiling of GD2⁺ and GD2⁻ SUM159 cells. GD2⁺ and GD2⁻ SUM159 cells were FACS sorted (Fig. 4A), and the cell lysates were analyzed using Kinexus antibody arrays. Labeled proteins were incubated on glass chips coated with 850 validated antibodies, which were further developed to identify differentially expressed or phosphorylated proteins. After validation for consistency in duplicate samples for each antibody on the chip, the proteins that were differentially upregulated, downregulated, or phosphorylated were identified (listed in Supplementary Table S2). We found that FAK and downstream proteins associated with this signaling, including Csk, 4E-BP1 (T45 and S65), and STAT3, were upregulated in GD2⁺ SUM159 cells compared with GD2⁻ cells. GD2 is known to activate FAK and FAK is known to activate PI3K–AKT–ERK–mTOR signaling. As 4E-BP1 is also activated in GD2⁺ cells, we hypothesized that FAK–AKT–ERK–mTOR signaling is activated in GD2⁺ cells.

To test our hypothesis, we performed Western blot analysis using antibodies associated with FAK–AKT–ERK–mTOR signaling in lysates of GD2⁺ and GD2⁻ SUM159 and MDA-MB-231 cells (Fig. 4A). Consistent with the protein–antibody array data, we found that FAK (Y861) and 4E-BP1 (S65) were significantly more phosphorylated in GD2⁺ than in GD2⁻ cells (Fig. 4B). We also found that AKT (S473), ERK (T202/Y204), and mTOR (S2448) were significantly more phosphorylated in GD2⁺ than in GD2⁻ SUM159 and MDA231 cells, while total protein levels were relatively unchanged, suggesting that FAK–AKT–ERK–mTOR signaling is activated in GD2⁺ cells (Fig. 4B).

As GD2 expression is regulated by ST8SIA1, we hypothesized that ST8SIA1 regulates cell signaling that is downstream of GD2 in TNBC cells. We analyzed expression and activity of FAK–AKT–ERK–mTOR signaling proteins in ST8SIA1-KO and Cas9-Control SUM159 cells by Western blotting and confirmed that pFAK (Y861), pAKT (S473), pERK (T202/Y204), pmTOR (S2448), pStat3, and p4E-BP1 (S65) were downregulated in ST8SIA1-KO cells compared with Cas9-Controls, indicating that FAK–AKT–ERK–mTOR signaling is regulated by ST8SIA1 in BCSCs (Fig. 4C).

To investigate the role of FAK–mTOR signaling in BCSC function, in vitro tumorigenesis and mammosphere formation (i.e., anchorage-independent growth) was assayed in SUM159 and MDA-MB-231 cells in the presence or absence of the FAK inhibitor PF-573228 (Fig. 5A) or FDA-approved mTOR inhibitor everolimus (Fig. 5B). Soft-agar colony formation by both SUM159 and MDA-MB-231 cells was inhibited 5- and 30-fold, respectively, by PF-573228 in a concentration-dependent manner (1, 10, 100, 1,000 nmol/L; Fig. 5A), and mammosphere formation was inhibited by 3- to 5-fold in both MDA-MB-231 and SUM159 cells when treated with PF-573228 in a concentration-dependent manner (Fig. 5B). To investigate the role of mTOR signaling in BCSC function, we treated SUM159 and MDA-MB-231 cells with the FDA-approved mTOR inhibitor everolimus and found that everolimus inhibited soft-agar colony formation 10-to 15-fold compared with untreated cells in a concentration-dependent manner in both cell types (Fig. 5C). Together, these data indicate that signaling pathways downstream of FAK and mTOR regulate BCSC function in TNBC cells.

Finally, to investigate the effects of ST8SIA1inhibition on in vivo tumor growth, Cas9-Control or ST8SIA1-KO SUM159 cells (3 × 10⁶ per mouse) were implanted into the mammary fat pads of NSG mice, and tumor growth was monitored weekly. The experiment was terminated when the control tumors reached ∼2 cm in diameter as per institutional IACUC regulations. Control tumors reached 2 cm in diameter within 4 to 5 weeks after implantation, whereas no tumor growth was observed in mice implanted with ST8SIA1-KO cells, even 10 weeks after implantation (Fig. 6A). In the mice implanted with ST8SIA1-KO cells, surgical dissection found only the matrigel plug that was implanted at the site of injection (Fig. 6B). Histologic analysis revealed typical xenograft tumor histology in the control tumors but only a few randomly dispersed cells in matrigel derived from ST8SIA1-KO cells (Fig. 6C). Cas9-Control–implanted mice were dead by day 40 (n = 10), whereas 100% of the ST8SIA1-KO–implanted mice survived at least 100 days after tumor cell implantation (n = 7; Fig. 6D). These data suggest that the ST8SIA1-KO cells lost their ability to initiate tumors in vivo.

To investigate whether ST8SIA1 regulates TNBC metastasis, NSG mice were injected in the tail vein with Cas9-Control or ST8SIA1-KO SUM159 cells (5 × 10⁵ per mouse). Beginning at week 3, three mice were sacrificed weekly, and their lungs were isolated, fixed in formalin, and embedded in paraffin for histologic studies. The H&E-stained sections showed distinct lung metastasis in the mice injected with the control SUM159 cells 4 weeks after implantation. In contrast, the lungs of mice injected with KO cells were completely free of metastases even after week 5 (Fig. 6E and F). These data suggest that ST8SIA1 regulates tumor growth and metastasis and is a potential therapeutic target in TNBC.

In this report, we conclude that ST8SIA1 is a potential target in TNBC and that inhibition of this enzyme will reduce tumor growth and metastasis by eliminating GD2⁺ BCSCs. ST8SIA1 is highly upregulated in basal-like TNBC cell lines and patient samples compared with other breast cancer subtypes. In addition, ST8SIA1 is a prognostic factor in TNBC and its expression is highly correlated with the presence of p53 mutations. ST8SIA1 expression is positively correlated with genes that are closely associated with BCSCs. Knockout of ST8SIA1 dramatically reduced in vitro tumorigenesis and mammosphere formation and completely eradicated tumor growth and metastasis in vivo. In mechanistic studies, knockout of ST8SIA expression reduced activation of several oncogenic pathways, including RAS, AKT, and MAPK. Proteomics analysis revealed that FAK and its downstream signaling are highly active in GD2⁺ BCSCs, and that this activity is highly dependent on ST8SIA1. Knockout of ST8SIA1 inhibited FAK–AKT–ERK–mTOR signaling, suggesting that ST8SIA1-regulated stem cell activity is mediated through this pathway.

Although ST8SIA1 is not directly involved in the biosynthesis of GD2, it is highly upregulated in GD2⁺ BCSCs. ST8SIA1 is the key enzyme in the biosynthesis of b- and c-series gangliosides, including GD3, GD2, GD1b, GT1b, and GQ1b in the b-series and GT3, GT2, GT1c, GQ1c, and GP1c in the c-series. Among these gangliosides, GD3 and GD2 are known to be upregulated in several different cancer types, including malignant melanoma (30, 31), neuroblastoma (32), glioma (33, 34), small cell lung cancer (35), and osteosarcoma (12, 36). However, only a few normal tissues, such as brain and skin, express these gangliosides. Our data suggest that ST8SIA1 is highly upregulated in basal-like cell breast cancers compared with the luminal types. It is interesting to note that GD2 expression in BCSCs is also associated mostly with basal-like cell lines, further indicating the association between ST8SIA1 and GD2. Analysis of TCGA data suggests that ST8SIA1 is overexpressed in breast tumors with the ER-negative and PR-negative phenotype. In addition, ST8SIA1 expression was 4.6-fold higher in TNBC than in non-TNBC breast cancer types. Furthermore, the mesenchymal subtype of TNBC has the highest ST8SIA1 expression of all subtypes. These findings are particularly interesting because the mesenchymal subtype is one of the most aggressive subtypes of TNBC, with the lowest overall survival among all TNBC subtypes (37). Although the role of ST8SIA1 in normal stem cells is not clear, ST8SIA1 knockout was found not to affect behavior, total life span, or reproductive capacity in mice (38, 39). These data suggest that loss of ST8SIA1 may not affect normal stem cells but may be deleterious for cancer stem cells, making ST8SIA1 an ideal target for inhibiting BCSC function in TNBC. These data suggest that ST8SIA1 expression is a specific marker for TNBC and could potentially be a therapeutic target in these cancers.

The high correlation between ST8SIA1 expression and mutations in the TP53 gene is another indication that ST8SIA1 is associated with the TNBC subtype, as over 80% of TNBC tumors have a mutation in this gene. Recent data suggest that mutations in TP53 disrupt mammary tissue architecture and are associated with a cancer stem cell phenotype (40, 41). Considering that mutations in the TP53 gene occur in many different types of cancer, ST8SIA1 could be a novel downstream target of mutated TP53 that supports oncogenic potential. In addition, ST8SIA1 expression is positively correlated with genes that are associated with BCSC function, including BCL11A, FOXC1, CXCR4, CXCL12, and VGLL1, suggesting the tumorigenic nature of cells expressing ST8SIA1. These observations further support our finding that ST8SIA1 is associated with BCSCs in TNBC tumors.

Knockout of ST8SIA1 completely blocked BCSC function, including anchorage-independent growth, mammosphere formation, and migration in vitro. In addition, knockout of ST8SIA1 completely inhibited tumor initiation and metastasis, which are characteristic features of BCSCs. These data support ST8SIA1 as a major target in TNBC and suggest that inhibitors targeting this enzyme could have high therapeutic potential in TNBC expressing ST8SIA1. Our findings suggest that the mechanism of ST8SIA1-mediated regulation of BCSCs involves the regulation of several genes associated with cancer progression and, more importantly, BCSC function. These genes include BCL11A, PDGFRβ, VCAM1, PDGFRα, SOX2, FOXC1, CXCR4, CXCL12, and FGFR2. BCL11A and FOXC1 have been reported to be upregulated in BCSCs and to regulate their function (25, 27). We have previously shown that induction of EMT enhances PDGFRβ expression in breast epithelial cells (42). Considering that EMT induces the stem cell phenotype in breast cancer cells (43), ST8SIA1 regulation of PDGFRβ suggests that this receptor has a role in the regulation of BCSC function. SOX2 and VCAM1 are other important proteins implicated in maintenance of BCSC function that are regulated by ST8SIA1 (44, 45). We reported that CXCR4, the receptor for CXCL12, has a crucial role in the regulation of breast cancer metastasis (46). GSEA analysis of the genes differentially regulated in ST8SIA1-KO and control cells suggests that several signaling pathways that regulate cell survival, proliferation, and migration, including AKT, ERK, RAS, p53, EGFR, and PDGF, are downregulated in the absence of ST8SIA1 in TNBC. These findings indicate that ST8SIA1 regulates BCSC function by regulating expression of several oncogenes involved in tumor growth and metastasis.

GD2 has been shown to interact with cell-surface receptors and to activate signaling downstream in several cancer types (8). We are currently developing a clinical protocol to target GD2⁺ BCSCs using the FDA-approved chimeric anti-GD2 monoclonal antibody dinutuximab in patients with TNBC. Our proteomics finding that FAK and mTOR signaling is activated in GD2⁺ compared with GD2⁻ cells, along with further analysis, provides important clues about downstream signal transducers associated with this signaling pathway. AKT and ERK, key proteins in the regulation of cell survival and proliferation, are highly upregulated in GD2⁺ TNBC cells compared with GD2⁻ cells. In addition, phosphorylation of these proteins was inhibited by ST8SIA1 knockout, suggesting that ST8SIA1 regulates activation of these signaling proteins in TNBC cells. Furthermore, we found that 4E-BP1, a downstream target of the mTORC1 complex, is highly phosphorylated in GD2⁺ cells. Western blot analysis confirmed downregulation of phosphorylated mTOR and 4E-BP1 in ST8SIA1-KO cells. Thus, our data demonstrate, for the first time, that ST8SIA1 regulates the FAK–AKT–ERK–mTOR signaling pathway in BCSCs.

In conclusion, our findings elucidate some of the mechanisms by which ST8SIA1 inhibition could suppress tumor growth/initiation and metastasis in TNBC. Targeting ST8SIA1 could be future therapeutic strategy that could reduce cancer recurrence and extend survival in TNBC patients.

No potential conflicts of interest were disclosed.

Conception and design: K. Nguyen, J. Sun, M. Andreeff, V.L. Battula

Development of methodology: K. Nguyen, A. Dasgupta, J. Sun, V.L. Battula

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Nguyen, B. Yuan, A. Dasgupta, J. Sun, V.L. Battula

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Nguyen, Y. Yan, B. Yuan, A. Dasgupta, J. Sun, K.-A. Do, N.T. Ueno, M. Andreeff, V.L. Battula

Writing, review, and/or revision of the manuscript: K. Nguyen, Y. Yan, K.-A. Do, N.T. Ueno, M. Andreeff, V.L. Battula

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Nguyen, Y. Yan, B. Yuan, A. Dasgupta, J. Sun, H. Mu, K.-A. Do, N.T. Ueno, M. Andreeff, V.L. Battula

Study supervision: M. Andreeff, V.L. Battula

Other (funding): M. Andreeff, V.L. Battula

The authors thank Dr. Arvind Rao at MD Anderson Cancer Center for his suggestions concerning the TCGA data analysis.

This work is supported by grants from the Breast Cancer Research Foundation (BCRF), the MD Anderson Cancer Center Support Grant (CA016672), the Paul and Mary Haas Chair in Genetics (to M. Andreeff), and the Berdon Lawrence Research Award from Bone Disease Program of Texas (to V.L. Battula).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.