CSNK1G2 differently sensitizes tamoxifen-induced decrease in PI3K/AKT/mTOR/S6K and ERK signaling according to the estrogen receptor existence in breast cancer cells

Tamoxifen (TAM) is a selective estrogen receptor modulator used for breast cancer patients. Prolonged use of tamoxifen is not recommended for some patients. In this study, we aimed to identify molecular targets sensitive to TAM using a genome-wide gene deletion library screening of fission yeast heterozygous mutants. From the screening, casein kinase 1 gamma 2 (CSNK1G2), a serine-/threonine protein kinase, was the most sensitive target to TAM with a significant cytotoxicity in estrogen receptor-positive (ER+) breast cancer cells but with only a slight toxicity in the case of ER- cells. In addition, tumor sphere formation and expression of breast stem cell marker genes such as CD44/CD2 were greatly inhibited by CSNK1G2 knockdown in ER+ breast cancer cells. Consistently, CSNK1G2 altered ERα activity via phosphorylation, specifically at serine (Ser)167, as well as the regulation of estrogen-responsive element (ERE) of estrogen-responsive genes such as CTSD and GREB1. However, ERα silencing almost completely blocked CSNK1G2-induced TAM sensitivity. In ER+ breast cancer cells, combined treatment with TAM and CSNK1G2 knockdown further enhanced the TAM-mediated decrease in phosphatidylinositol 3-kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR)/ribosomal protein S6 kinase (S6K) signaling but not extracellular signal-regulated kinase (ERK) signaling. Inversely, in ER- cells treated with TAM, only ERK and PI3K signaling was altered by CSNK1G2 knockdown. The CK1 inhibitor, D4476, partly mimicked the CSNK1G2 knockdown effect in ER+ breast cancer cells, but with a broader repression ranging from PI3K/AKT/mTOR/S6K to ERK signaling. Collectively, these results suggest that CSNK1G2 plays a key role in sensitizing TAM toxicity in ER+ and ER- breast cancer cells via differently regulating PI3K/AKT/mTOR/S6K and ERK signaling.

Introduction CSNK1D, CSNK1G1, CSNK1G2, CSNK1G3, and CSNK1E subfamilies. Their functions in cell survival and tumorigenesis were recently reported [7,14]. Of CSNK1 subtypes, the functional role of CSNK1A, CSNK1D, and CSNK1E in different cancer types have also been reported [15]. CSNK1A controls the mitotic spindle formation during cell division, thus regulating cell cycle during M phase in mouse oocytes [16]. CSNK1A also regulates mTOR activity by altering the stability of DEPTOR, the endogenous mTOR inhibitor [17]. However, the role of the CSNK1G isoforms, especially, CSNK1G2, has yet to be investigated in depth. Recently, the function of CSNK1G1 and CSNK1G3 in tumor necrosis factor (TNFα)-induced necroptosis, an important necrotic cell death under apoptosis-deficient conditions, has been reported in human colon cancer [7].
Fission yeast Schizosaccharomyces pombe (S. pombe) is an excellent model system to study cell morphogenesis and cell division cycles. Most genome-wide screening for drugs has been applied with a haploid gene deletion library consisting of only nonessential genes [18]. However, screening using a heterozygous gene deletion library are more powerful because they contain all the genes, including essential genes in addition to nonessential genes [19,20]. Here we constructed a genome-wide gene deletion library in fission yeast, allowing us to find sensitive or resistant target genes against TAM by the principle of drug-induced haploinsufficiency [21]. From the systematical screening, we found the fission yeast cki3 as the most sensitive target for TAM. Therefore, the main purpose of this study was to explore the potential of CSNK1G2, the human ortholog of fission yeast cki3, as a novel TAM-sensitive target. We report the function of CSNK1G2 in breast cancer cells in TAM sensitivity, specifically comparing its effects in ER + and in ERbreast cancer cells.

Genome-wide screening of heterozygous gene deletion library in fission yeast
For systematic screening of target genes sensitive to TAM, we used a previously constructed heterozygous gene deletion library [20]. Briefly, the library was constructed by homologues replacement of each gene into KanMX marker gene as a selection marker based on the parental strain of a SP286 wild-type (ade6-M210/ade6-M216, leu1-32/leu1-32, ura4-D18/ ura4-D18h + /h + ). The library was pooled and aliquoted into 100 μL vials for each screen. Vials were kept frozen at -80˚C until use. Notably, each deletion mutant had a pair (up-and downtag) for unique built-in molecular bar coded for a parallel analysis. Systematic screening of TAM target genes was performed as previously reported [19]. Briefly, a vial of frozen pool was activated in 50 ml of YES media for 24-30 h up to OD 600 = 2 (approximately 4.4 x 10 7 cells/ ml). These cells were then diluted in 50 ml of YES media to OD 600 = 0.05 and cultivated up to OD 600 = 1.6 (approximately 3.5 x 10 7 cells) with or without TAM (15 μM). This was repeated four times every 5 generations up to 20 generations. An aliquot of 3.5 x 10 7 cells was harvested every 5 generations. Genomic DNA was prepared using a ZR-Fungal/Bacterial DNA kit (Zymo Research, Irvine, CA, USA). Microarray experiment was performed using a custommade GeneChip (48 K KRIBB_SP2, Thermo Fisher Scientific) and fluorescence-labeled probe prepared by PCR for the pair of bar codes [20]. Three independent microarray experiments were performed. Repeated targets of every top 10 targets sensitive to TAM were selected as the first-line targets for TAM. Finally, 6 target strains were selected by the order of the sensitivity indicated as the relative growth fitness (RF) < 0.9 (p<0.05).

Confirmation of primary target genes by spotting and liquid assays
Primary candidates were confirmed one by one based on individual growth analysis using a spotting assay. For the spotting assay, cells in log phase were diluted to OD 600 = 0.5 in YES media and spotted in 5-fold serial dilutions onto YES agar plates with or without 15 μM TAM. First-line targets for TAM were confirmed and subjected to gene ontology (GO) analysis using GO term finder (http://go.princeton.edu/cgi-bin/GOTermFinder).

siRNA Transfection
Oligonucleotide Silencer 1 Select siRNAs for knocking down the human genes were purchased from Ambion (Thermo Fisher Scientific). The following siRNA was used: human CSNK1G2 (siRNA ID: s3631). As a control siRNA, Silencer 1 Select Negative Control #1 siRNA (#4390843) was used.
Cells were seeded into multi-well culture plates (2×10 4 cells/well in a 48-well plate; 3.5×10 5 cells/well in a 6-well plate) and transfected with individual siRNA using Lipofectamine™ RNAi-MAX transfection reagent (Life Technologies) following the manufacturer's instructions. Twenty-four hours after transfection, silenced cells were used for analysis. The efficiency of transfection by siRNA was confirmed either by western blotting or RT-PCR.

Quantitative real-time PCR (qRT-PCR)
Total RNA was extracted from cells using TRIzol Reagent (Life Technologies) according to the manufacturer's instructions. One microgram of total RNA was reverse transcribed into cDNA using a qPCRBIO cDNA synthesis kit (#PB30.11, PCR Biosystems Inc.; London, UK). qRT-PCR reactions were then performed using 2× qPCRBIO SyGreen Blue Mix Lo-ROX (PCR Biosystems Inc.) as per the manufacturer's protocol and run on Quantstudio5 Real-time PCR log (Thermo Fisher Scientific; Massachusetts, USA). Specific primer sequences used in this study are listed in S1 Table. Each test was repeated three times, independently.

Transient transfection and dual-luciferase assay
Luciferase assay for estrogen-responsive element (ERE) activity measurement was performed using a dual-luciferase reporter assay system (Promega, WI, USA) according to the manufacturer's instructions. In brief, MCF-7 cells were seeded into 96-well plates at a density of 1 × 10 4 cells/well. Cells were transfected with (ERE) 4 -luciferase plasmid and SV40-Renilla luciferase plasmid using Lipofectamine 2000 reagent (Thermo Fisher Scientific). Transfection solutions were replaced after 24 h. Cells were then treated with and without 1 μM TAM and further incubated for 24 h prior to cell lysis and measurement of firefly-and Renilla-luciferase luminescence using a Stop-Glo reagent system (Promega; WI, USA). Each value was normalized to the Renilla luciferase control. Each data point generated represents the average of triplicate determinations.

3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay
MCF-7, T-47D, and MDA-MB-231 cell lines (2×10 4 cells/well in 48-well plates) transfected with control siRNA or each target siRNA for 24 h were treated with various drugs at indicated concentrations. Their viability was then measured using the MTT assay. Briefly, 25 μL of MTT stock solution (5 mg/mL in phosphate-buffered saline) was added to each well followed by incubation at 37˚C for 1 h to allow cell-mediated reduction of MTT. To detect the amount of reduced MTT, media were aspirated and 250 μL of DMSO was added. The absorbance was measured at 562 nm using a micro-plate reader (Epoch™ Microplate Spectrophotometer, Bio-Tek Instruments; Winooski, VT, USA).

Western blot analysis
Protein expression levels in whole cell extracts of cultured human cancer cells were measured using western blot analysis. Briefly, cells were washed with cold phosphate-buffered saline (PBS) prior to lysing with RIPA buffer (Thermo Fisher Scientific), containing protease and phosphatase inhibitor cocktail. After centrifugation, the supernatant was collected. Protein concentrations in the supernatant were measured by bicinchoninic acid (BCA) method (Thermo Fisher Scientific). For western blotting, equal amounts of proteins were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes. These membranes were blocked with Tris buffered saline with Tween 20 containing 5% skim-milk for 1 h at room temperature (RT), and incubated overnight with specific primary antibodies at 4˚C. The next day, membranes were incubated with HRP-conjugated second antibodies for 1 h at RT. Finally, immunoreactive proteins were visualized using an enhanced chemiluminescence kit (Thermo Fisher Scientific). Protein expression was quantitatively assessed by densitometric analysis of the intensity of each protein band using an Ima-geJ software. All experiments were repeated at least three times using cultures from different passages.

Tumor sphere formation assay
Prior to tumor sphere culture, MCF-7, T-47D, and MDA-MB-231 cells were grown as monolayer according to the supplier's recommendations on tissue culture-treated 10 cm dishes (STEMCELL technologies Inc., Vancouver, BC, Canada). After reaching confluence, cells were seeded onto a 6-well ultra-low adherent plate (STEMCELL Technologies Inc., Catalog #27145) at a density of 1×10 4 cells/well for individual breast cancer cells. Immediately after seeding, NC and CSNK1G2 siRNAs were transfected as mentioned previously. At 24 h after transfection, cells were treated with the drugs. Size and numbers of tumor spheres were observed daily. The number of tumor spheres over 100 μm in size or larger were used for analysis.

Statistics
All data are presented as mean (SD). For multiple comparisons among groups, one-way analysis of variance followed by a Fisher LSD post-hoc test was employed. Paired t-tests were used to analyze differences between two groups. SigmaPlot (spw8) and GraphPad Prism5 were used for statistical analysis. P-values < 0.05 were considered statistically significant.

TAM-sensitive targets and their growth inhibitory effects determined by two-rounds of microarray using S. pombe heterozygotes deletion mutant gene pools and spotting assay
Through repeated genome-wide screening against TAM, 10 target strains with RF< 0.9 (P<0.05) compared with the wild-type control were selected. Finally, five target mutant strains were confirmed by spotting and liquid assay as shown in Fig 1A and 1B. They consisted of one severe (SSS), two moderate (SS), two mild (S) targets as measured by growth rate against tamoxifen-sensitive toxicity. There were three non-essential targets and two essential targets as measured by gene dispensability. The fission yeast mutant that was most sensitive to TAM was cki3 deletion mutant (Δcki3). According to gene ontology (GO) enrichment analysis for biological process, five target genes were associated with the following processes: signal transduction specifically involved in cellular component organization (cki3, rad24), cytokinesis like septation initiation signaling (cdc2), and cellular protein metabolic process associated with protein folding (cct6, sks2) ( Fig 1C). These results suggest that the cytotoxic effect of TAM might be potentially attributed to a process that results in the biosynthesis of constituent macromolecules, assembly, arrangement of constituent parts, and/or disassembly of a cellular component in fission yeast S. pombe.
Consistently, when examined whether targets that had been identified by the fission yeast library could represent their effects in human breast cancer cells, ER + MCF-7 cells transfected with siRNAs for five individual target genes (CSNK1G2, YWHAE, CDK1, HSPA1A, and CCT6A) and tested TAM-sensitive response. Of five candidate target genes, the knockdown of CSNK1G2 showed the most sensitive tamoxifen toxicity ( Fig 1D).

CSNK1G2 knockdown differently sensitizes TAM-induced cytotoxicity in ER + and ERbreast cancer cells
The goal of this study was to find an effective target to TAM-treated human breast cancer cells. Prior to conducting the study, we first identified the types of hormone receptors expressed by each breast cancer cell line. Western blot analysis showed that MCF-7 cells were triple positive (ER + /PR + /HER2 + ), but T-47D cells expressed only ER and PR, and MDA-MB-231 cells were triple negative (ER -/PR -/HER2 -). Cytotoxicity assay following treatment with various concentrations of the ER agonist, E 2 , or the selective estrogen receptor modulator (SERM), TAM, in three different cell lines revealed that ER -MDA-MB-231 cells were the least sensitive to both Relative growths of deletion strains treated with 0 or 15 μM TAM for 9 h were measured as OD 600 using a microplate reader (mean (SD); n = 5, � P < 0.05, �� P < 0.01 vs. TAM-treated each control). (C) List of the top 5 TAM targets screened in yeast. Gene ontology (GO) in terms of the biological process was analyzed using GO term finder (http://go. Princeton.edu/cgi-bin/GOTermFinder). Human orthologs were from the HomoloGene or Ensembl database (Http://www.ncbi.nlm.nih.gov/homologene or http:// www.ensemble.org). Sensitivity was classified as follows: Severe (SSS), moderate (SS), and mild (S). (D) Bars indicate TAM-induced relative survival rates in candidate target genes-silenced breast cancer cells. MCF-7 cells were silenced with negative control or five individual target siRNAs (CNSK1G1, YWHAE, CDK1, HSPA1A, and CCT6A) treated with 1 μM tamoxifen for 24 h. Drug-mediated cytotoxicity was measured by MTT assay (mean (SD); n = 4, � P < 0.05, �� P < 0.01, # P < 0.001 vs. each represented counterpart). Since we examined the presence of hormone receptor and responsiveness of the cells to the SERM TAM, we further investigated whether these cell lines present different TAM sensitivity following CSNK1G2 knockdown. For this, transfected cells with NC or CSNK1G2 siRNA were treated with 1 μM TAM for 24 h and cell viability tests were performed. The efficiency of the gene knockdown in the breast cancer cells was confirmed by western blot analysis (S1 Fig). Interestingly, CSNK1G2 silencing significantly enhanced TAM-induced cellular toxicity in ER + cells, but not in ERcells (Fig 2D-2F). However, overexpression of ERα in ER -MDA-MB-231 cells using GFP-ERα plasmid DNA completely reproduced the CSNK1G2 silencinginduced cytotoxic effects of ER + breast cancer cells (Fig 2G).

CSNK1G2 silencing accelerates the suppressive effect of TAM in tumor sphere formation and expression of breast stem cell marker genes in ER + breast cancer cells
Cancer cells have stem cell-like characteristics, which may play a critical role in the heterogeneity of tumor cells in response to anti-cancer agents and drug resistance [22]. In the present study, we examined whether CSNK1G2 was associated with altered sphere-forming ability by TAM in human breast tumor cells. ER + (MCF-7) and ER -(MDA-MB-231) human breast cancer cells seeded onto a 6-well ultra-low adherent plate were treated with 1 μM TAM after silencing CSNK1G2, and gene-targeting effect was quantified by real-time PCR for CSNK1G2 (Fig 3E and 3F). For the quantitative analysis, tumor spheres over 100 μm in diameter were considered real tumor spheres with tumor-like characteristics. Consistent with the previous cytotoxicity assay results (Fig 2), the two cell lines showed different responsiveness. TAM (1 μM) significantly reduced the number of formed tumor spheres by approximately 40% in ER + breast cancer cells, whereas CSNK1G2 silencing itself did not alter the formation of ER + -MCF-7 tumor spheres (Fig 3A-3C). When MCF-7 cells were co-treated with 1 μM TAM and CSNK1G2 silencing, the number of formed tumor spheres was markedly reduced to 15%, compared to that of vehicle-treated cells (Fig 3A-3C). In ER -MDA-MB-231 cells, siRNAmediated gene targeting did not significantly affect TAM-induced decrease in the formation of tumor spheres (Fig 3B-3D).
In the same line, determination of CSNK1G2 effect on CD44/CD24 mRNA expression, which are known cancer stem cell markers, showed that CSNK1G2 silencing in MCF-7 cells significantly accelerated TAM-mediated decrease in the ratio of CD44/CD24 mRNA expression to a greater extent than in NC siRNA-transfected cells (Fig 3G). However, this CSNK1G2 silencing-induced effect was not shown in ERbreast cancer cells (Fig 3H). Since the role of CD44 and CD24 in tumor metastasis and invasiveness has previously been reported [23], CSNK1G2 may, in part, contribute to tumor progress.

TAM-mediated ERE or ER-responsive gene transcriptional activity as well as ERα activation are altered by CSNK1G2 silencing in ER + breast cancer cells
Estrogen-ERα binding is very important for genomic and non-genomic activation of cellular signaling molecules; thus, the regulation of nuclear gene expression was induced to proliferate or differentiate the breast cancer cells. Here we examined how CSNK1G2 knockdown influences E 2 or TAM-mediated phosphorylation of ERα. The phosphorylation level at different sites of ERα, specifically at serine (Ser) 118 and Ser 167 , was determined by western blot analysis. CSNK1G2 silencing significantly reduced E 2 -induced ERα phosphorylation at both sites ( Fig  4A) and reduced TAM-mediated ERα phosphorylation only at Ser 167 (Fig 4B). Hence, ERα activity was influenced by CSNK1G2, inversely proving that ERα knockdown in ER + cells eliminated CSNK1G2 effect on TAM-induced toxicity. CSNK1G2-mediated alteration of TAM toxicity was completely obstructed by ERα exclusion, implying that CSNK1G2 and ERα interact with a currently unknown mechanism (Fig 4C).
TAM can compete with estrogens for binding to ERα and can interfere with the recruitment of coactivators, required for ERα-mediated gene expression [24]. To elucidate if the target effects on tumor cell proliferation are associated with ERE activity, we assessed ERE transcriptional activity in MCF-7 cells after treatment with 10 nM E 2 , 1 μM ICI-182780, and 1 μM TAM. E 2 and ICI-182780 were used as positive and negative regulators of ERE transcriptional activity, respectively. As expected, knockdown of MCF-7 cells with CSNK1G2 significantly altered E 2 , ICI-182780, or TAM-induced ERE transcriptional activity compared to each NC siRNA transfected counterpart group (Fig 4D). These results indicated that increased TAM-sensitivity might occur following CSNK1G2 knockdown, in part, through alteration of genomic ER effect, associated with ERE site of diverse genes in ER + breast cancer cells. Realtime PCR for ER responsive genes such as CTSD and GREB1 in ER + cells confirmed the role of CNSK1G2. Low concentration of TAM-induced increase in CTSD and GREB1 expression, except PS2, was almost completely removed by CSNK1G2 siRNA transfection (Fig 4E).

CSNK1G2 contributes differently to TAM-induced PI3K/AKT/mTOR/S6K and ERK signaling in ER + and ERbreast cancer cells
Most cancer cells grow rapidly and proliferate abnormally because of altered cellular signal cascades. Presently, we focused on the phosphoinositide 3-kinase/protein kinase B (AKT)/ mammalian target of rapamycin/ribosomal S6 kinase (PI3K/AKT/mTOR/S6K) as well as ERK signaling pathways. These pathways are crucial signal transduction networks in the promotion of tumor initiation and progression.
Before we examined the CSNK1G2 silencing-induced signal changes in TAM treated cells, we primarily checked the effect of E 2 in these pathways (S2 Fig). Interestingly, E 2 treatment following CSNK1G2 silencing showed a different effect on AKT/mTOR/S6K signals between ER + and ERcells, indicating that CSNK1G2 knockdown significantly inhibited these signals. Consistent with this result, 1 μM TAM in MCF-7 cells slightly reduced the phosphorylation of PI3K but not significantly. However, CSNK1G2 knockdown significantly inhibited PI3K activity ( Fig 5A); these results were also reproduced in the ERbreast cancer cells with less significant reduction (Fig 5E). Transcriptional patterns of diverse ER-responsive genes, PS2, CTSD, and GREB1, were analyzed by RT-qPCR in CSNK1G2-silenced ER + breast cancer cells and were compared with those in NC siRNA-transfected control. All values were normalized to those in the NC siRNA-transfected vehicle group. GAPDH was used as a loading control and CSNK1G2 knockdown efficiency was also confirmed. Data are presented as mean (SD); n = 4; � P < 0.05, �� P < 0.01, # P < 0.001 vs. each represented counterpart. https://doi.org/10.1371/journal.pone.0246264.g004 Simultaneously, activity of AKT, a downstream signal of PI3K, was markedly reduced by both 1 μM TAM and target modulation (Fig 5A). However, in ER -MDA-MB-231 cells, target modulation did not alter AKT activity ( Fig 5E). As mTOR is a downstream effector in the PI3K and AKT signaling pathways, we determined whether mTOR activity was altered by modulation of CSNK1G2 in breast cancer cells. Western blots in MCF-7 cells revealed that CSNK1G2 silencing reduced ribosomal protein S6 (rpS6), the downstream signaling protein of mTOR, as well as mTOR activity (Fig 5B and 5C). Contrarily, MDA-MB-231 cells transfected with CSNK1G2 siRNA were not significantly changed in phosphor mTOR (Fig 5F) but rather increased in phosphor rpS6 (Fig 5G). In addition, CSNK1G2 did not alter ERK signal in ER + cells but slightly decreased in ERcells, another tumorigenesis-associated signal (Fig 5D-5H). ER -MDA-MB-231 cells overexpressed with GFP-ERα mimicked the results of ER + MCF-7 cells in activities of PI3K, AKT, mTOR, and rpS6 but not of ERK (Fig 5I-5K). The effect of the CSNK1G2 knockdown on the expression of the genes involved in the aforementioned signaling was consistent with the results obtained from the western blot analysis. Specifically, in ER + cells, CSNK1G2 silencing suppressed TAM-mediated activated expression in PI3R2, AKT, and PRS6 (Fig 6A), whereas in ERcells, CSNK1G2 targeting enhanced the expression of PI3R2 and PRS6 following TAM treatment (Fig 6B). During the experiments, transfection efficiency of CSNK1G2 was confirmed by real-time PCR. At this time, gene expression of other subtypes of CSNK1G such as CSNK1G1 and CSNK1G3 were also determined (S3 Fig). CSNK1G2 silencing did not alter the expression of both CSNK1G1 and CSNK1G3 in MCF-7 cells but significantly enhanced CSNK1G1 and CSNK1G3 expression in ERcells. From the results, it is possible that the lack of TAM-sensitizing effects in ERcells might be partly due to the compensation action of enhanced CSNK1G1 and CSNK1G3 expression.
Taken together, in ER + MCF-7 cells, CSNK1G2 might modulate cytotoxicity via reinforced TAM-induced reduction in PI3K/AKT/mTOR/S6K, but not in ERK signaling cascades (Fig 7,  left diagram). However, in ER -MDA-MB-231, CSNK1G2 only altered PI3K and ERK signals (Fig 7, right diagram). These results indicate that CSNK1G2 may differently modulate breast cancer cells in a cell type-dependent manner.

Casein kinase 1 inhibitor partly mimics CSNK1G2 knockdown-mediated PI3K/AKT/mTOR/S6K signaling
Casein kinase 1 has many different subfamilies. In this study, we targeted the CSNK1G2 subfamily. The subfamilies share common characteristics within cells, but different functions are also found among the subfamilies [15]. Therefore, we examined whether the effects that we found in TAM-treated breast cancer cells were unique to CSNK1G2. Breast cancer cells were https://doi.org/10.1371/journal.pone.0246264.g006 exposed to 2 μM D4476, a CK 1 inhibitor, instead of CSNK1G2 siRNA transfection. Before examining the cellular signals, we investigated which subtype was affected by D4476. Unexpectedly, in ER + MCF-7 cells, D4476 significantly increased CSNK1G1 expression but decreased both CSNK1G2 ( # P<0.001 vs vehicle) and CSNK1G3 ( � P<0.01 vs vehicle) with different degrees of significance (Fig 8A). Similar results were produced in ER -MDA-MB-231 ( Fig 8B). Additionally, when D4476-induced cytotoxicity was compared between ER + MCF-7 and ER -MDA-MB-231, cellular viability was significantly lower upon combined treatment of D4476 together with 1 μM TAM in ER + cells than upon single treatment with 1 μM TAM ( Fig  8C). However, ERcells did not show any significant cytotoxicity following combined treatment ( Fig 8D).
Furthermore, when we examined the expression of cellular signal molecules, D4476 reproduced many of the effects of CSNK1G2 on the PI3K/AKT/mTOR signaling system in breast cancer cells. However, unlike CSNK1G2 knockdown, D4476 also altered ER phosphorylation at Ser 118 but not at Ser 167 and ERK activity in ER + cells (Fig 8E and 8F). Different from MCF-7 cells, in the ERcells, D4476 slightly reduced PI3K activity, while increasing AKT and ERK activity (Fig 8G).
Interestingly, when ER -MDA-MB-231 cells were overexpressed with GFP-ERα, MDA-MB-231 cells co-treated with D4476 and TAM showed similar pattern of TAM toxicity with the case of MCF-7 cells (Fig 9A). At the same time, activities of cellular signals such as PI3K, AKT, and rpS6 in GFP-ERα-transfected MDA-MB-231 were also similar with the MCF-7 response (Fig 9B). The findings indicated that the window of the targeting molecules by CSNK1G2 is presumably narrower than that by D4476; thus, CSNK1G2 can aim at a specific node of signal molecules. Moreover, ER existence partly contributes to CSNK1G2 effects.

Discussion
Target screening using a heterozygous gene deletion mutant library of S. pombe in this study reveals that nonessential genes (cki3, rad24, sks2) presently induce more sensitized cell growth than essential genes (cdc2 and cct6). In the fission yeast genome, essential and nonessential genes are distributed evenly. Essential genes are significantly enriched for core cellular processes such as macromolecular (DNA, RNA, protein, and lipid) metabolism [20,25]. In contrast, nonessential genes noticeably enriched for regulatory functions including control of gene expression, cell communication, and conditional or life-cycle specific processes such as stress response, transmembrane transport, meiosis, and sexual reproduction [26]. Therefore, nonessential genes like cki3 would be better drug targets for drug delivery because most genes indispensable for cellular survival should be protected for the viability of normal cells. Breast cancer cells have heterogeneous characteristics. Therefore, ascertaining of the types of steroid receptors as well as assessment of multigene profiles that might influence a patient's response to specific therapy should be first considered. ER signaling develops in various ways. ER activity in cells is largely due to the direct binding of dimeric ER to DNA sequences, EREs [27], which can accelerate transcriptional activity and mediate the expressions of various genes associated with protein synthesis, cell proliferation, and survival [28]. In addition, ER can indirectly associate with promoters via protein-protein interactions with other DNA-binding transcription factors. These processes are collectively termed genomic actions of estrogen [28]. Alternately, for non-genomic actions, ER associated with diverse receptors on plasma membrane such as G-protein coupled receptor, insulin growth factor receptor, and epidermal growth factor receptor can activate subsequent intracellular secondary messengers such as cAMP and other signal molecules, and initiate cellular signal cascades, ultimately resulting in indirect changes in gene expression [28].
TAM targets ER and attenuates the transcriptional activity in an ERE-dependent or non-ERE-dependent manner, eventually disrupting signaling pathways involved in tumor growth and migration [27,29]. In line with the drug action point of TAM, our experiment for TAM sensitivity in different breast cancer cell lines proved that ER + cells responded sensitively to TAM, but ERcells showed poor reactivity. TAM can have antagonistic and agonistic effects against ER depending on the intracellular contents of the cells, thus it is important to discover new breast cancer targets that can reduce the adverse effects of TAM, while maintaining cytotoxicity to tumor cells [27]. In this study, the effect of CSNK1G2 as a supplement of TAM was confirmed. The ERE activity as assessed by the reporter assay and analysis of the several genes such as CTSD and GREB1 in MCF-7 cells raised the possibility that CSNK1G2 partially contributes to the transcriptional activity of the ER-responsive genes through ERE regulation.
It has also been documented that CTSD and GREB1 were increased in breast cancer patients [30,31]. In particular, CTSD expression is used as a prognostic marker, since the patients with high CTSD level resulted in poor overall survival and disease-free survival [30,31]. CTSD is active in low pH environments, stimulating tumor cells for pro-invasive and pro-metastatic capacities [32]. In hormone-responsive breast cancer cells, transcription of CTSD is also induced by estradiol but blocked by TAM [33]. Real-time PCR results in MCF-7 cells showed that the slightly upregulated expression of estrogen-responsive genes following treatment with low concentration of TAM was almost completely returned to the normal level following target gene knockdown. However, it is not yet clear how CSNK1G2 is involved in a series of processes such as ERα activation and subsequent activation of ER-responsive genes.
CSNK1G is ubiquitously expressed in a wide range of eukaryotes ranging from yeasts to humans [34]. In contrast to yeast CSNK1 isoforms, the functions of mammalian CSNK1 isoforms remain poorly characterized. CSNK1 can interact with a large number of substrates involved in multiple cellular functions, including regulation of membrane trafficking, vesicular transport, cytokinesis, ribosome biogenesis, DNA repair, signal transduction pathways, and circadian rhythm [13]. Many reports have described the regulation of CSNK1 isoforms, including MDM2 stability of p53 by CSNK1A, and control of Wnt/β-catenin and hedgehog signaling by CSNK1E and CSNK1D [35,36]. However, the roles of CSNK1G isoforms have yet to be extensively investigated and remain unclear. Only enhanced expression of CSNK1G3 in renal cell carcinoma has been reported [37]. Analysis of the actual datasets of the cancer genome atlas (TCGA), available at the cBioPortal for Cancer Genomics [38,39], revealed the mutation pattern and copy number alterations of CSNK1 subtypes in 24 different tumor types. The analysis focused only on CSNK1A1, CSNK1D, and CSNK1E, not on CSNK1G; CSNK1A1 alterations are dominant in clear cell renal cell carcinoma, and pancreas cancer, CSNK1D in liver and sarcoma, and CSNK1E in melanoma and liver.
As presented by the expression of CD44/CD24 and tumor sphere formation in MCF-7 and MDA-MB-231, CSNK1G2 also affected the expression of breast stem cell-markers only in ER + breast cancer cells. The combination of CD44 and CD24 or aldehyde dehydrogenase 1 (ALDH1) is a representative cancer stem cell marker in breast cancer [40,41]. Breast stem cells potentially have the ability of replication and tumor formation, thus leading to metastasis from one organ to another [40,41]. Indeed, a xenograft study using nude mice and other in vivo experiments have reported that disseminated tumor cells in the bone marrow of breast cancer patients showed a high CD44 + /CD24ratio [42] and the elevated expression of CD44 + /CD24was closely associated with lung metastasis [41].
Phosphorylation of ERα is essential for its activation after stimulation with various ligands and nonsteroidal growth factors. Phosphorylation of ERα at Serine (Ser) 118 in the NH 2 terminus within the AF-1 region is mediated by the CDK7 [43] and Ras/MAPK pathways [44,45], which enable ligand-independent transactivation of ERα by interaction with the transcriptional coactivators, CBP and SRC-1. However, phosphorylation at Ser 167 in AF-1 is associated with several protein kinases including casein kinase II, p90 RSK (p90 ribosomal S6 kinase), AKT, and p70 RSK, thereby enhancing DNA binding and transcriptional activity [46,47]. In our study, CSNK1G2 silencing significantly suppressed E 2 -triggered phosphorylation of ERα at both Ser 118 and Ser 167 , but it only reduced TAM-mediated ERα phosphorylation at Ser 167 . This is very important because phosphorylation of ERα at Ser 167 is indicative of prolonged disease-free and overall survival in breast cancer patients [48]. Furthermore, when the activity was lost by ERα siRNA, CSNK1G2 silencing-induced TAM sensitivity was completely blocked. However, it is not clear whether the effect of CSNK1G2 on ERα activity occurs via direct interaction to ERα or via any connection with other substrates such as AKT, RSK, or S6K.
Representative signaling pathways known to be involved in tumor growth include PI3K/ AKT/mTOR, and Ras/Raf/MAPK [49], which are associated with stimulation of proliferation and protection from apoptosis [50]. Hyperactivation of PI3K/AKT/mTOR pathway drives tumorigenesis of ER + breast cancer [51,52] and resistance to endocrine therapy [53,54]. Therefore, new approaches with various inhibitors of the PI3K/AKT/mTOR pathway, such as everolimus and exemestane, augment the benefit of existing endocrine therapy by extending time to disease progression and preventing or overcoming resistance to endocrine treatment [55]; thus, many preclinical and clinical studies are currently focusing on this. AKT has been focused on as a therapeutic target for cancer as it works as a nodal point linking cell growth, apoptosis, ribosomal biogenesis, and cellular metabolism [56]. The AKT3 targeting strategy is specifically effective for ER-negative, androgen receptor (AR)-negative breast or prostate cancer cells [57]. Interestingly, a previous proteomic study demonstrated that breast cancer cell line-specific associations with specific signals exist. For example, luminal B subtype of human breast cancer hyperactivates PI3K signaling associated with lower ER levels, but PI3K blockade increases ER level. Therefore, dual targeting of the PI3K and ER signaling pathways may be useful in patients with ER + tumors [58]. Conversely, basal-like breast cancer displays greater sensitivity to selective MEK inhibition than luminal or HER2 + subtypes [59]. Together with PI3K pathways, Raf/MEK/ERK pathway is highly conserved and transmits signals from the cell surface receptors to nuclear transcription factors. Of the three, ERK is specially focused on crosstalk with PI3K signals. ERK1/2 converges on mTORC1 via p90S6K, thus influencing cell proliferation, survival, motility, and angiogenesis [12].
Recently, treatment-predictive role and prognostic value of the S6K has been also reported. Aberrant expression and localization of S6K are characterized as a poor sign for ER + patients who require TAM treatment [56]. In accordance with this, S6K1 has been shown to phosphorylate the ER at Ser 167 , inducing conformational changes in the receptor and making it less responsive to TAM [60]. We here observe that, in breast cancer cells, CSNK1G2-silenced cells lead to the suppression of not only PI3K/AKT/mTOR/S6K but also ERK. Although attenuation of ERK signal was only seen in ERbreast cancer cells, it is interesting that CSNK1G2 can target both PI3K/AKT/mTOR/S6K and ERK signal pathways depending on the characteristics of hormone receptor subtypes. CSNK1G2 selectively manipulates the tumor progress, sometimes in a nuclear and genomic manner via ERE inhibition of specific genes associated with tumor growth, but sometimes by non-genomic manner via the AKT/mTOR/S6K pathway, followed by p-ER 167 activity. However, these effects can occur simultaneously, thus leading to strong suppression.
Many chemotherapies that utilize selective ER modulators suppress cell growth-associated signals, but their effects do not last, owing to the activation of multiple compensatory mechanisms [61]. This leads to adverse effects of cancer therapy. Nonetheless, our results indicate that CSNK1G2 induces complete reduction in most signals ranging from PI3K to S6K; even these inhibitory effects on cell growth were ER-dependent. In addition, we presently observed that S6 kinase signaling, the downstream of AKT-mTOR, was the most strongly reduced by the combined treatment with TAM and CSNK1G2 silencing. Interestingly, however, many reports demonstrated that the continuous suppression of S6 kinase can activate upstream signals such as AKT and PI3K signaling, thereby hindering the inhibition of cancer progress [56]. This is because there are multiple crosstalk and interconnections present in the PI3K/AKT/ mTOR/S6K and ERK signaling pathways; thus, suppression of the specific signals may lead to activation in negative feedback regulatory mechanisms. Consequently, this may provide unwanted activation of signaling molecules such as AKT and ERK, upon inhibition [12,62]. However, we observed that the reduction of S6 kinase by CSNK1G2 in ER + cells did not activate upstream signaling molecules such as PI3K, AKT, and mTOR, as well as ERK, indicating the possible role of CSNK1G2 in specific cellular signaling control without interruption of other signals. Moreover, the study using D4476 also proved that CSNK1G2 does not follow all the effects of CK1, and instead displays its own effects via different targeting points from D4476.
Taken together, our results reveal a unique action mechanism of CSNK1G2 on PI3K/AKT/ mTOR/S6K and ERK signaling pathways in breast cancer cells and provide insight into how the target selectively controls and contributes to cell survival in different human breast cancer cell types. Furthermore, these results provide a rationale for collectively assessing cellular signals for improved prognosis in clinical trials of breast cancers treated with inhibitors of PI3K and MAPK pathways. GAPDH was used as a loading control. Data are expressed as mean (SD); n = 4; � P < 0.05, �� P < 0.01, # P < 0.001 vs. each represented counterpart. (TIF) S1