High PKCλ expression is required for ALDH1-positive cancer stem cell function and indicates a poor clinical outcome in late-stage breast cancer patients

Despite development of markers for identification of cancer stem cells, the mechanism underlying the survival and division of cancer stem cells in breast cancer remains unclear. Here we report that PKCλ expression was enriched in basal-like breast cancer, among breast cancer subtypes, and was correlated with ALDH1A3 expression (p = 0.016, χ2-test). Late stage breast cancer patients expressing PKCλhigh and ALDH1A3high had poorer disease-specific survival than those expressing PKCλlow and ALDH1A3low (p = 0.018, log rank test for Kaplan-Meier survival curves: hazard ratio 2.58, 95% CI 1.24–5.37, p = 0.011, multivariate Cox regression analysis). Functional inhibition of PKCλ through siRNA-mediated knockdown or CRISPR-Cas9-mediated knockout in ALDH1high MDA-MB 157 and MDA-MB 468 basal-like breast cancer cells led to increases in the numbers of trypan blue-positive and active-caspase 3-positive cells, as well as suppression of tumor-sphere formation and cell migration. Furthermore, the amount of CASP3 and PARP mRNA and the level of cleaved caspase-3 protein were enhanced in PKCλ-deficient ALDH1high cells. An Apoptosis inhibitor (z-VAD-FMK) suppressed the enhancement of cell death as well as the levels of cleaved caspase-3 protein in PKCλ deficient ALDH1high cells. It also altered the asymmetric/symmetric distribution ratio of ALDH1A3 protein. In addition, PKCλ knockdown led to increases in cellular ROS levels in ALDH1high cells. These results suggest that PKCλ is essential for cancer cell survival and migration, tumorigenesis, the asymmetric distribution of ALDH1A3 protein among cancer cells, and the maintenance of low ROS levels in ALDH1-positive breast cancer stem cells. This makes it a key contributor to the poorer prognosis seen in late-stage breast cancer patients.


Introduction
In numerous countries, breast cancer is the most common malignant neoplasm in women. Breast cancer is classified into at least six subtypes, normal-like, luminal A, luminal B, HER2-enriched, claudin-low, and basal-like, based on stringent patterns of gene expression [1][2][3]. Among those, basal-like breast cancer, which exhibits stem-like properties and accounts for up to 15-20% of all breast cancers, is associated with particularly poor outcomes [4]. In addition, based on their immunohistochemically determined receptor status, breast cancers have also been classified as ER and/or PgR positive type, HER2 positive type, and triple negative type (TNBC; ER and/or PgR negative, HER2 negative). About 70% of basal-like breast cancers overlap with TNBC [5][6][7][8].
Overall, breast cancer prognoses are good. However, patients with late-stage lesions (stage III or IV) have significantly shorter overall survival (OS) [9]. This is because late-stage breast cancer is often resistant to standard medical treatments such as conventional surgery, chemotherapy and radiotherapy, which makes their recurrence and metastasis much more likely [9]. Thus, new pharmacological approaches to manage late-stage cancer are needed.
Cancer stem cells (CSCs) are a small subpopulation of cancer cells exhibiting capacities for self-renewal, multipotency, and tumorigenesis. Apart from those features, CSCs also exhibit characteristic cellular properties, including cell migration, asymmetric cell division and resistance to reactive oxygen species (ROS), and most are resistant to standard chemotherapy and radiotherapy [10][11][12][13][14][15][16]. For example, CSCs make up the metastatic niche and generate bulk tumor at distant organs. It is therefore thought that CSCs are a critical factor in the metastatic cascade [12]. CD44 + /CD24 -/low CSCs, derived from breast cancer, exhibit migration potential that increases with tumor grade [13], while a human CD44 + CD24 -/low Linand mouse Thy1 + CD24 + Lin -CSC-enriched population exhibits low ROS levels and high expression of anti-ROS genes [14].
CSCs have a capacity for asymmetric propagation; that is, they have the ability to generate other stem-like cells or differentiated cells [15]. This feature is controlled by the balance of their symmetric and asymmetric cell division. Cancer cells positive for PKH26 (PKH pos ) have the capacity for asymmetric division and correlate with poorly differentiated cancers displaying a higher CSC content [16]. A detailed understanding of the mechanisms that define this property of CSCs could potentially reveal novel therapeutic targets and foster progress toward new drug development against CSCs.
In the present study, we show that in patients with stage III-IV tumors, high expression of PKCλ and ALDH1A3 contributes to poor clinical outcomes. Furthermore, PKCλ is involved in the regulation of the asymmetric distribution of ALDH1A3 among cells and the maintenance of lower ROS levels in ALDH1-positive breast CSCs. We therefore conclude that high PKCλ expression is required for ALDH1-postive cancer stem cell function and indicates a poor clinical outcome in late-stage breast cancer patients.

Analysis of the METABRIC dataset
The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) dataset [53,54] was downloaded from the cBioportal (https://www.cbioportal.org/) [55,56] in March, 2019, after which the downloaded data were analyzed as we described previously [31,32,57]. The clinicopathological data from patients were shown in an earlier report [32]. The median age at the time of breast cancer diagnosis was 61.1 years in this dataset (range: 21 to 90 years). The numbers of patients with the indicated PAM 50 subtypes were as follows: normal-like, 148; luminal A, 679; luminal B, 461; HER2-enriched, 220; claudin-low, 199; and basal-like, 199. The gene alteration data were obtained from cBioportal, and the mRNA expression levels were compared using the Kruskal-Wallis test with the Steel-Dwass test. To construct survival curves, we defined the optimal cutoffs for the high-and low-expression groups using receiver operator characteristic (ROC) curves plotting expression of genes versus patient disease-specific survival rate (DSS) at several tumor stage. The optimal cutoff thresholds were determined using the Youden index. Survival curves were plotted using by the Kaplan-Meier method, and curves were compared using the log-rank (Cochran-Mantel-Haenszel) test. A multivariate Cox regression model was used to evaluate the influence of gene expression and to estimate adjusted hazard ratios (HRs) with age as a confounding factor. We also defined groups based on expression of PKCλ, PKCz, and stemness markers: + (z-score > 0) and-(z-score < 0) in Table 1. The p values for the correlation between PKCλ or PKCz and stemness marker expression were calculated using the χ 2 test. For the heatmap of stemness gene and PKCλ and PKCz expression (z-score) in Fig 2A, the average value of these genes was calculated and drawn as a heatmap using R version 3.5.2 (R Foundation for Statistical Computing Vienna, Austria). p-values below 0.05 were considered to be significant ( � p < 0.05, �� p < 0.01, ��� p < 0.001). All other statistical analyses were carried out using BellCurve for Excel ver. 3.00 (SSRI, JAPAN).

Analysis of the cancer genome atlas (TCGA) dataset
TCGA breast cancer dataset [58] was downloaded from Oncomine (https://www.oncomine. org; Compendia Bioscience, Ann Arbor, MI, USA) [59] in October, 2019 and from cBioportal (https://www.cbioportal.org) [55,56] in March, 2019. The downloaded data were analyzed as we described previously [31,32,57]. Expression of PKCλ mRNA (reporter: A_23_P18392) and PKCz mRNA (A_23_P51186) was compared between normal and cancerous tissues, both of which were available from TCGA breast cancer dataset, using the Wilcoxon signed rank test. Clinicopathological data from the breast cancer patients were presented in an earlier report [32]. The median age at the time of breast cancer diagnosis was 57.9 years in this dataset (range: 26 to 90 years). The dataset contains mRNA expression data from 61 normal breast tissue samples and 532 primary breast tumor samples. The gene alteration data were obtained from cBioportal. RNA expression was displayed using paired comparison of normal tissues vs. tumor tissues from all samples using R version 3.5.2 (R Foundation for Statistical Computing Vienna, Austria). The p values were calculated using the Wilcoxon signed-rank test. To make survival curves, we defined the optimal cutoffs for the high-and low-expression groups using ROC curves based on the expression of genes vs. patient OS in each cancer. Survival curves were plotted using the Kaplan-Meier method. The p-value for each cancer was calculated using the log-rank (Cochran-Mantel-Haenszel) test. A multivariate Cox regression model was used to evaluate the influence of gene expression and to estimate the adjusted HRs with age and sex as confounding factors. p-values below 0.05 were considered to be significant ( �

Cell culture
Two human basal-like breast cancer cell lines (MDA-MB 157 and MDA-MB 468) were grown in Dulbecco's modified Eagle's medium (DMEM) medium supplemented with 10% fetal bovine serum (FBS) (Biosera) and a human normal-like (non-transformed) mammary epithelial cell line (MCF10A) were grown in mammary epithelial cell growth medium (MEGM; Lonza). These cell lines were purchased from the American Type Culture Collection (ATCC).
The cells were then cultured as described previously [31, 32].

Immunoblotting
For growth in 2D cultures, ALDH1 high cells (approximately 5 x 10 4 cells /well) were cultured in 12-well plates for 24 h in DMEM and then harvested using Trypsin. For tumor-sphere formation, ALDH1 high cells (2 x 10 4 cells /well) were cultured for 3 days in DMEM supplemented with 10% FBS and 0.6% methylcellulose (see in vitro tumor-sphere culture section) and harvested manually. ALDH1 high cells were isolated using a cell sorter. The collected cells were washed three times with 1 x PBS and lysed in RIPA buffer (50 mM Tris (pH 8.0), 150 mM NaCl, 0.5 w/v% sodium deoxycholate, 0.1 w/v% SDS, 1.0 w/v% Nonidet p-40) plus 1 x protease inhibitor (Nacalai tesque) and 1 x phosphatase inhibitor (Nacalai tesque).
To detect Caspase-3 and Cleaved caspase-3, aliquots of lysate containing approximately 25 μg of total proteins were subjected to SDS-PAGE (15% gel) and transferred (2 mA/cm 2 , 30 min) to Immobilon-P PVDF membranes (Millipore, IPVH00010). The membranes were then blocked with 5% BSA in TTBS and incubated with the primary antibodies in IMMUNOSHOT reagent 1 (CSR), after which the membranes were probed with the horseradish peroxidaseconjugated secondary antibody in IMMUNOSHOT reagent 2 (CSR). Specific signals were detected with chemiluminescence reagent Immunostar LD (Wako) or EzWestLumiOne (ATTO) using Chemi Doc MP (Bio-Rad).
The antibodies used in this study were monoclonal mouse anti-PKCι

Quantitative PCR
Total RNA was isolated using Sepasol1-RNA I Super G (Nacalai tesque) and Direct-zol TM RNA miniprep (ZYMO) according to manufacturer's instructions and reverse-transcribed using Rever Tra Ace qPCR RT Master Mix (TOYOBO). Gene expression assays were performed with THUNDERBIRD probe qPCR Mix (TOYOBO) according to the manufacturer's instructions. The reaction protocol was as follows: 95˚C for 1 min followed by 45 cycles of denaturation at 95˚C for 10 s and extension at 60˚C for 1 min. We carried out three independent experiments. 18S rRNA (ABI) was used to normalize differences in RNA input. Quantitative PCR primer and probe sequences are given in S1 Table. ALDEFLUOR assay ALDH1 high cells were isolated from MDA-MB 157 and MDA-MB 468 cells and analyzed using an ALDEFLUOR assay kit (Stem Cell Technologies) as previously described [31,32]. ALDH1high cells were sorted using FACS Aria TM II and III (BD Bioscience) and analyzed using a FACS Calibur (BD Bioscience). ALDH1 low cells were sorted as the 5% of the total population exhibiting the lowest ALDH1 activity. The data were analyzed using FlowJo 8.8.6 software (BD Bioscience).

siRNA transfection
PKCλ knockdown in breast cancer cell lines was achieved by transfection of siRNAs (SIGMA) as previously described [32]. Briefly, siRNA targeting PKCλ was transfected using OPTI-MEM (Gibco) and Lipofectamine™ RNAiMAX Transfection Reagent (Invitrogen). The cells were seeded into 6-well plates at a concentration of 3.0 x 10 5 cells/well and then treated with 10 nM (final concentration) siRNA transfection mixtures. The siRNAs used were MISSION siRNA Universal Negative Control (SIGMA) and PKCλ siRNA (5'-CAA GUG UUC UGA AGA GUU UTT-3').

in vitro tumor-sphere culture
Tumor-spheres were cultured as previously described [31,32]. ALDH1 high cells isolated from the Negative control or PKCλ KO clone were plated in ultralow attachment 96-well plates (Greiner) (1 x 10 3 cells/well) and cultured for 7 days in DMEM supplemented with 10% FBS and 0.6% methylcellulose. Images were captured with a DMIL LED (Leica). Sphere size was then measured as mean area of diameter (μm) for approximately 200 spheres using ImageJ Fiji software.

Transwell migration assays
Transwell migration assays were performed using 8.0 μm Falcon < cell culture inserts (CORN-ING) in a 24 well plate (SIGMA). ALDH1 high cells were seeded into the upper chamber at a density of 5 x 10 4 cells/well in DMEM supplemented with 5% FBS. The lower chamber contained DMEM supplemented with 10% FBS. After incubation for 24 h at 37˚C, the transwell inserts were removed from the plate, washed twice with 1xPBS, and fixed in 2% paraformaldehyde (Wako) for 2 min and methanol for 20 min. After fixation, the inserts were washed twice with 1xPBS, stained for 15 min with 0.5% crystal violet (SIGMA), and again washed twice with 1xPBS. Cells in the upper chamber were then removed using a cotton swab. Images of migrated cell on the underside of the filter were captured using a DMIL LED microscope (Leica), and the numbers of cells in different 5 fields of view were counted.

Trypan blue assays
In Fig 3J and 3K, ALDH1 high cells isolated after PKCλ KD for 48 h with targeted siRNA were plated in 12-well plates (Thermo) at a density of 2 x 10 4 cells/well and incubated for 24 h. apoptosis inhibitor (using the pan-caspase inhibitor z-VAD-FMK) was purchased from PEPTIDE Institute Inc. and dissolved in 100% DMSO, making a 200 mM stock solution. In Fig 4F, ALDH1 high control cells or MDA-MB 157 PKCλ KO cells were plated in 12-well plates (Thermo) at a density of 2 x 10 4 cells/well and incubated for 48 h with 100 μM z-VAD-FMK or 0.5% DMSO as a control. After staining with 0.4 w/v% Trypan blue solution (Wako), the cells were counted manually.

Analysis of intracellular ROS levels
Sorted ALDH1 high cells were plated on 8-well Lab-Teck chamber slides (Thermo) at a density of 5 x 10 4 cells/well and cultured for 48 h. To measure cellular ROS levels, the cells were stained with H2-DCFDA (Invitrogen), which generates fluorescent signals when oxidized by ROS in the cells. Cells were incubated with prewarmed 10 μM H2-DCFDA (diluted in PBS) staining solution for 30 min at 37˚C and stained for 10 min with 0.1 μg/mL Hoechst 33342 (Invitrogen) diluted in PBS. All subsequent steps were performed in the dark. Images were captured using a 6000B microscope (Leica). Mean fluorescence values were determined using the ImageJ Fiji software.

High expression of PKCλ mRNA in basal-like breast cancers exhibiting a low frequencies of gene amplification and mutation
Earlier immunohistochemical analyses showed that PKCλ protein is overexpressed in a variety of human cancers, including breast cancer [37][38][39][40][41][42][43][44][45][46][47][48], and that amplification of its gene occurs in lung and ovarian cancers [39,44]. To assess PKCλ gene alterations in breast cancer, we used two datasets: TCGA dataset from oncomine, which includes data from normal tissues, and the METABRIC dataset from cBIoportal, which lacks data from normal tissues. We first compared PKCλ gene alterations in breast cancers to those in lung and ovarian cancers. As shown in Fig  1A, the frequency of PKCλ gene amplification was lower in both breast cancer datasets tested (2.1% or 4.6% in TCGA, 3.5% in METABRIC) than in the lung (18.0% or 33.1% in TCGA) or ovarian (19.2% or 31.4% in TCGA) cancer datasets. In addition, there are few genetic mutations (0.1% in TCGA) and no deletions (0% in TCGA and METABRIC) in the breast cancer datasets. We therefore compared PKCλ mRNA expression between breast cancer and normal tissues derived from the same patients using TCGA dataset, which revealed that PKCλ expression was significantly higher in the cancers than normal tissues ( Fig 1B). It appears, therefore, that higher PKCλ expression in breast cancers reflects increased transcription rather than gene amplification or mutation.
We next used the METABRIC dataset to examine in more detail the relationship between PKCλ overexpression and breast cancer PAM50 subtypes. As shown in Fig 1C, PKCλ expression was highest in basal-like breast cancers. This is consistent with the idea that overexpression of PKCλ protein contributes to tumorigenesis in TNBC, which is similar to the basal-like subtype [37].

Correlation between PKCλ and ALDH1A3 in basal-like breast cancer
Basal-like breast cancer is an aggressive subtype exhibiting stem-like properties [7,8,31,32]. We therefore examined expression of PKCλ and several stem cell marker genes in several breast cancer subtypes. We found that PKCλ, along with NOTCH1, MET, CD133, ALDH1A3, NOTCH3, OCT4, MYC and NANOG, was enriched in basal-like breast cancer (Fig 2A, left  panel; S1 Fig). Recently, Tokinaga-Uchiyama et al. reported that high expression of PKCλ protein associates with poor clinical outcomes in cases of stage III-IV cervical cancer [46]. In addition, we observed that PKCλ expression was highest in patients with stage III-IV basal-like tumors (S1 Fig). This prompted us to assess the relationship between expression of PKCλ and stem cell marker genes in stage III-IV breast cancer subtypes. We found that PKCλ was overexpressed in basal-like and HER2-enriched types, as was ALDH1A3, CD133 and OCT4 (Fig 2A,  right panel). Moreover, PKCλ correlated with ALDH1A1 and ALDH1A3 in basal-like breast cancer (ALDH1A1; p = 0.017; ALDH1A3; p = 0.016, χ 2 -test). Among patients with basal-like cancers, the population co-expressing both PKCλ and ALDH1A3 (43.7%; n = 87/199) was higher than that co-expressing PKCλ and ALDH1A1 (12.1%; n = 24/199) ( Table 1).

Late-stage breast cancer patients exhibiting correlated expression of PKCλ and ALDH1A3 had poorer clinical outcomes
To evaluate the prognosis of patients showing higher expression of PKCλ and/or ALDH1A3, we performed a Kaplan-Meier analysis of DSS. At all disease stages, patients expressing only PKCλ high had a poorer prognosis (all stage, p = 0.0023; stage 0-II, p = 0.020; stage III-IV, p = 0.0094, log-rank test). By contrast, patients expressing ALDH1A3 high had a poorer prognosis only at stages III-IV (p = 0.022, log-rank test) (Fig 2B). Stage III-IV patients expressing both PKCλ high and ALDH1A3 high also had poorer clinical outcomes (Fig 2B, p = 0.018, log-rank test). Multivariate analysis of DSS showed that patients expressing only PKCλ high (HR 1.91, 95% CI 1.18-3.08, p = 0.0079) or both PKCλ high and ALDH1A3 high (HR 2.58, 95% CI 1.24-5.37, p = 0.011) had poorer prognoses at stages III-IV (Fig 2C). Among breast cancer subtypes,  tumors expressing PKCλ high and ALDH1A3 high at late stages included a higher fraction of basal-like breast cancers (26.5%) than did all stages (20.6%) or early-stages (16.1%) (S2 Fig). These results suggest that PKCλ may be involved in cancerous progression and may contribute to poor clinical outcomes when expressed in ALDH1-positive CSCs in basal-like breast cancers at late tumor stages.

PKCλ contributes to in vitro tumor-sphere formation by ALDH1 high cells
Using the MDA-MB 157 and MDA-MB 468 human basal-like breast cancer cell lines, we observed that levels of PKCλ and ALDH1A3 expression were higher in both breast cancer cell lines than in human normal-like (non-transformed) MCF10A cells (Fig 3A). In addition, protein levels of both ALDH1A3 and PKCλ were higher in ALDH1 high than ALDH1 low cells isolated from among MDA-MB 157 and MDA-MB 468 cells (Fig 3B and [32]). The ALDH1 high cells also exhibited such CSC features as self-renewal, multidifferentiation and tumorigenesis to a greater degree than ALDH1 low cells [31]. Therefore, to further investigate the role of PKCλ in ALDH1-positive CSCs, we used two methods to inhibit the enzyme: siRNA-mediated knockdown (KD) and CRISPR-Cas9-mediated knockout (KO). As shown in Fig 3B and 3C, PKCλ KD or KO did not significantly affect levels of ALDH1A3 protein in ALDH1 high MDA-MB 157 cells, though PKCλ KD did elicit a decrease in ALDH1A3 levels in ALDH1 high MDA-MB 468 cells. In addition, ALDEFLUOR assays showed that the number of ALDH1 high cells was reduced in PKCλ-depleted cancer cells (Fig 3D and 3E). To assess the function of PKCλ in ALDH1-positive basal-like breast CSCs, we performed in vitro tumor-sphere assays using PKCλ-deficient ALDH1 high cells. PKCλ depletion in ALDH1 high cells led to decreases in both the number and size of tumor-spheres (Fig 3F-3H). These results suggest that PKCλ is involved in cell proliferation and/or survival and contributes to tumor-sphere formation by ALDH1-positive CSCs in basal-like breast cancer.

PKCλ contributes to ALDH1 high cell migration
It is known that migration of the breast CSC population (CD44 + CD24 -/low ) gradually increases with tumor stage progression [13] and that PKCλ KD decreases the migration potential of multiple TNBC cell lines [37]. That suggests PKCλ plays an important role in the migration of ALDH1-positive CSCs. To further test that idea, we assessed the effect of PKCλ depletion on ALDH1 high cell migration. We found that PKCλ depletion in ALDH1 high cells led to a decrease in the number of migrated cells (Fig 3I), which suggests that PKCλ is required for ALDH1-positive basal-like breast CSC migration.

PKCλ contributes to ALDH1 high cell survival
To determine the reason why PKCλ depletion led to decreases in the number of ALDH1 high cells and in tumor-sphere formation by ALDH1 high cells, we performed trypan blue assays with PKCλ-depleted ALDH1 high cells. We found that PKCλ depletion led to increases in the numbers of trypan blue-positive cells among ALDH1 high MDA-MB 157 and MDA-MB 468 cells (Fig 3J and 3K). We also examined the Akt and p44/42 MAPK phosphorylation status in PKCλ-depleted ALDH1 high cells. The levels of phospho-Akt (S473) were not significantly changed in PKCλ-deficient ALDH1 high cells. The levels of phospho-Akt (T308) were differed in between ALDH1 high PKCλ KD MDA-MB468 cells and PKCλ KO cells (S3 Fig). On the other hand, levels of phospho-44/42 MAPK were slightly enhanced in PKCλ-deficient ALDH1 high cells (S3 Fig). These results suggest that PKCλ is essential for the survival of ALDH1-positive basal-like breast CSCs with Akt and MAPK independent manner.

PKCλ suppresses apoptosis of ALDH1 high cells
To determine more specifically how PKCλ depletion led to an increase in ALDH1 high cell death, we next considered whether PKCλ contributes to apoptosis in ALDH1 high cells. Fig 4A  shows that PKCλ KD led to increased numbers of cleaved (active) caspase-3-positive ALDH1high cells derived from both of MDA-MB 157 and MDA-MB 468 cells. PKCλ depletion also led to higher levels of cleaved caspase-3 protein (Fig 4B) as well as higher levels Casp3 and PARP mRNA in ALDH1 high cells (Fig 4C and 4D). In addition, PKCλ KO led to increases in the levels of cleaved casepase-3 protein within 2D culture condition and tumor-spheres formed by ALDH1 high cells (Fig 4E). Treating ALDH1 high cells with the apoptosis inhibitor (z-VAD-FMK) suppressed the cell death and reduced the levels of cleaved caspase-3 otherwise seen with PKCλ depletion (Fig 4F and 4G). These results suggest that PKCλ suppresses apoptosis and promotes cell survival among ALDH1-positive basal-like breast CSCs.

PKCλ is required for asymmetric and symmetric ALDH1A3 protein distribution in paired cells
An important and characteristic property of CSCs is their capacity for asymmetric cell division to propagate cancer stem-like cells or generate differentiated cells [15]. Because of the involvement of PKCλ in asymmetric cell division of both normal stem/progenitor cells [35,36,60] and CSCs [49], we hypothesized that PKCλ controls the balance between symmetric and asymmetric division of ALDH1-positive breast CSCs. To test that idea, we examined the asymmetric/symmetric distribution of PKCλ and ALDH1A3 proteins in ALDH1 high cells by immunofluorescently staining corresponding cell pairs for PKCλ and ALDH1A3. PKCλ was detected in both the asymmetric and symmetric fractions (High/Low, 19.4%; High/High, 59.1%; Low/Low, 21.5%) (Fig 5A and 5B). ALDH1A3 was also detected symmetric and asymmetric fractions (High/Low, 23.3%; High/High, 49.5%; Low/Low, 27.2%), and exhibited a similar expression distribution among cells (Fig 5A and 5C). Within cells, moreover, PKCλ largely colocalized with ALDH1A3 ( Fig 5A). Within the ALDH1A3 asymmetric distribution (23.3% of total paired cells), higher PKCλ colocalized with higher ALDH1A3 levels, while lower PKCλ colocalized with lower ALDH1A3 levels (60.5%). Within the symmetric distribution (76.7% of total paired cells), higher ALDH1A3 colocalized with higher PKCλ in 85.7% of cells, while lower ALDH1A3 colocalized with lower PKCλ in 59.5% of cells. Higher PKCλ colocalized with lower ALDH1A3 in 37.2% of cells (Fig 5C and 5D). Interestingly, PKCλ depletion caused a decrease in the asymmetric distribution of ALDH1A3 (High/Low; 17.3% to 8.4%) and an increase in the symmetric distribution (High/High, 35.7% to 45.8%; Low/Low, 36.9% to 55.9%) (Fig 5E). These results suggest that PKCλ is required for both asymmetric and symmetric cell division of ALDH1-positive basal-like breast CSCs.

PKCλ suppresses ROS accumulation in ALDH1 high cells
ROS levels are reportedly lower in CSCs than non-CSCs in cancer [14]. A previous report suggested that increases in ROS can induce caspase-3-induced apoptosis [61]. We therefore assessed the contribution made by PKCλ to the maintenance of low ROS levels in ALDH1-positive CSCs. Fig 6 shows  ALDH1 low cells. PKCλKD led to accumulation of ROS in ALDH1 high cells derived from MDA-MB 157 and MDA-MB 468, but led to reductions in ROS levels in ALDH1 low MDA-MB 157 cells but not MDA-MB 468 cells (Fig 6B and 6D). In addition, we proposed that depletion PKCλ may lead to enhanced expression of ROS defense genes, including SOD1, SOD2 and Gpx1, in response to ROS accumulation. Consistent with that idea, PKCλ depletion led to increased expression of SOD1, SOD2 and Gpx1 mRNA in response to ROS accumulation in ALDH1 high cells derived from both MDA-MB 157 and MDA-MB 468 cells (Fig 6C and 6E). These results suggest that PKCλ plays different roles in the regulation of ROS in CSCs and non-CSCs, and that PKCλ is at least involved in the maintenance of lower ROS-levels to protect ALDH1-positive breast CSCs from ROS damage.

Discussion
Our findings show that breast cancer patients expressing both PKCλ high and ALDH1A3 high exhibit poorer clinical outcomes at late-stage than those expressing PKCλ low and ALDH1A3 low ( Fig 2B and 2C). Moreover, among breast cancer subtypes, tumors showing PKCλ high and ALDH1A3 high at late stages include a higher frequency of basal-like breast cancers (26.5%) than all-stage (20.6%) or early-stage cancers (16.1%) (S2 Fig). Earlier studies indicate that it is ALDH1A3, not ALDH1A1, that contributes to ALDH activity in basal-like breast cancer cell lines [28][29][30][31][32], and ALDH1A3 is reported to positively associate with tumor grade, stage and metastasis [27]. Furthermore, Opdenaker et al. showed that there is a positive correlation between ALDH1A3 expression and tumor stage in TNBC patients [27], and we previously reported that ALDH1A3 is highly expressed in basal-like breast cancers (Fig 3A and [32]). Given the correlation between PKCλ and ALDH1A3 expression (Table 1), we suggest that PKCλ contributes to the cancerous progression of ALDH1-positive breast CSCs. Note that basal-like and late-stage breast cancers exhibit stemness [6][7][8][9], and that ALDH1 is involved in mediating chemoresistance through drug metabolism and detoxification of cellular aldehydes [62]. Consequently, a new pharmacological approach that targets PKCλ-dependent cellular regulation of ALDH1-positive CSCs is needed for the treatment of late-stage basal-like breast cancer.
One recent study suggests that PKCλ KD leads to a decrease in asymmetric cell division to generate CD133-positive and CD133-negative daughter cells in lung adenocarcinoma oncospheres [49]. In the present study, we revealed that PKCλ KD in MDA-MB 468 cells also leads to decreases in the numbers of CD133-postive cells (S4 Fig), suggesting PKCλ is involved in the stemness of CD133-postive breast CSCs. We further observed that CD133-poistive cells account for large fractions of both MDA-MB157 (46.1 ± 8.7%) and MDA-MB 468 cells (20.1 ± 6.7%) (S4 Fig). As a result, we deemed CD133 positivity not to be an effective indicator for distinguishing CSCs from non-CSCs MDA-MB 157 and MDA-MB 468 cells. This is consistent with an earlier report that in pancreatic cancer, CD133-positive cells are not significantly enriched among CSCs, but ALDH1 high cells are [22]. We therefore focused on examining the role of PKCλ in ALDH1-positive breast CSCs.
As with PKCλ, the aPKC subfamily member PKCz exhibited low frequencies of gene amplification and mutation (S6A Fig), and expression of PKCz was significantly higher in breast cancers than in normal tissues from the same patients (S6B Fig). Unlike PKCλ, however, PKCz did not correlate with ALDH1A3 in basal-like breast cancer (p = 0.191, χ 2 -test) ( Table 1). We therefore limited our focus to the role of PKCλ in ALDH1-positive breast CSCs. Nonetheless, late-stage breast cancer patients expressing PKCz high and ALDH1A3 high had poorer clinical outcomes than patients expressing PKCz low and ALDH1A3 high (S6 Fig). The role of PKCz in ALDH1-positive breast CSCs will need to be addressed in future studies.
An earlier report suggests that PKCλ activates NOTCH3 expression via ELF3 phosphorylation, and this axis maintains the highly tumorigenic TIC phenotype in KRAS-mediated lung adenocarcinoma [49]. Moreover, the PKCι(λ)-NOTCH1 pathway contributes to the survival of glioblastoma CSCs [50]. As shown in Fig 2A, both NOTCH3 and PKCλ are highly expressed in basal-like cancers at all stages. Furthermore, patients expressing PKCλ high and NOTCH3 high account for a large percentage of basal-like breast cancer patients (46.2%, n = 92 / 199) ( Table 1). We therefore suggest that PKCλ plays a key role in determining the cellular properties of ALDH1-positive breast CSCs via NOTCH3 signaling. Deletion of PKCλ suppresses migration of MDA-MB 231, a TNBC cell line [37]. In Fig 3I, we show that PKCλ depletion suppresses migration of ALDH1 high cells. It may be that PKCλ is required for cell migration of ALDH1-positive breast CSCs. On the other hand, the results summarized in Figs 3 and 4 suggest that the suppression of ALDH1 high cell migration reflects increased cell death related to the PKCλ depletion.
It was previously suggested that activation of the PI3K / Akt pathway enriches the tumorigenic stem/progenitor cell population in breast cancer cell lines and tumor xenografts [65]. In addition, PI3K activates PKCλ [66]. In the present study, we found that the levels of phosphorylated Akt differed and the enhancement of phosphorylated MAPK in PKCλ-depleted ALDH1 high cells derived from breast cancer cell lines. This suggests that PKCλ is not essentially involved in the activation for Akt and MAPK in ALDH1-positive breast CSCs. That finding is consistent with the earlier report indicating that whether or not PKCλ-dependent activation of Akt and MAPK occurs depends on the cancer cell type [67]. Our results thus strongly suggest that PKCλ is essential for the survival of ALDH1-positive breast CSCs with independent manner of the activation of Akt and MAPK.
CSCs have a capacity for asymmetric cell division to propagate CSCs or generate differentiated cells [15]. This capacity for asymmetric cell division is thought to underlie the cellular heterogeneity of tumors. PKCλ is a master regulator of asymmetric cell division in multicellular organisms, including C. elegance and Drosophila melanogaster [35,36]. PKCλ KD decreases asymmetric cell division to generate CD133-positive and CD133-negative daughter cells in lung adenocarcinoma oncospheres [49]. In present study, we observed that PKCλ KD alters the distribution of ALDH1A3 protein. Importantly, we detected 10 propagated patterns for ALDH1A3 and PKCλ (Fig 5C and 5D). This suggests PKCλ may be required for symmetric and asymmetric cell propagation of ALDH1A3-positive cells and may be a key contributor to the cellular heterogeneity seen in breast cancer.
Several recent studies indicate that CSCs contain less intracellular ROS than non-CSCs [14]. These lower ROS levels are reportedly associated with increased expression of free radical scavenging systems and contribute to radiotherapy resistance [14]. However, the mechanism by which the lower ROS levels are maintained is poorly understood. Our results suggest that PKCλ contributes to the maintenance of low ROS levels in ALDH1-positive breast CSCs ( Fig  6). Inhibition of ALDH1 in breast cancer cells is associated with increased ROS levels [68,69]. Furthermore, levels of ROS and ALDH1 activity are inversely related in melanoma [70]. It thus appears that ALDH1 activity mediates scavenging of ROS in CSCs. Human CD44 + CD24 -/low Linand mouse Thy1 + CD24 + Lin -CSC-enriched populations exhibit low ROS levels and higher expression of anti-ROS genes, including Foxo1, than a non-CSCs population [14]. Moreover, Foxo1 phosphorylation by PKCλ contributes to cell proliferation in angiosarcoma [71]. In addition, it has been reported that PKCz, another PKC isoform, regulates glucose-6-phosphate dehydrogenase (G6PD) gene expression [72]. G6PD is the rate limiting enzyme in the pentose phosphate pathway (PPP). This is the major pathway for nicotinamide adenine dinucleotide phosphate (NADPH) generation [73], which is an essential cofactor for maintenance of redox balance within cells [74,75]. Mele et al. reported that lapatinib, a tyrosine kinase inhibitor widely used for the treatment of breast cancer, and polydatin, a G6PD inhibitor, exerted a synergic effect on MCF7 (MCF7 mock ) cell viability, but had no effect on G6PD-overexpressing MCF7 (MCF7 G6PD+ ) cells [76]. It was concluded that both PKCz and PKCλ regulate G6PD gene expression and maintenance of NADPH levels in cells. We therefore suggest that one mechanism to maintain low ROS levels in CSCs may involve regulating PKCλ-dependent phosphorylation of Foxo1 or G6PD gene expression. On the other hand, PKCλ knockdown suppressed ROS levels in MDA-MB157 ALDH1 low cells, but did not change ROS levels in MDA-MB 468 ALDH1 low cells (Fig 6). Thus, the role of PKCλ in the regulation of ROS levels appears to differ between CSCs and non-CSCs.
ALDH1 is involved in the detoxification of toxic aldehyde intermediates, which are generated by ROS-induced peroxidation of intracellular lipids [77]. PKCλ depletion in ALDH1 high cells leads to increases in the numbers of apoptotic and dead cells (Fig 3J, 3K and Fig 4) and a corresponding increase in intracellular ROS (Fig 6). This suggests PKCλ may be involved in preventing apoptosis by maintaining lower intracellular ROS levels in ALDH1-positive breast CSCs. Because cancer cells are continuously exposed to ROS, they have developed protective mechanisms that enable proliferation, survival, migration and tumorigenesis, despite the presence of ROS in CSCs. The PKCλ-dependent maintenance of low ROS levels in ALDH1 high cells may be one of those mechanisms. However, the complex relationship between ROS levels and the cellular properties of CSCs remain unclear.

Conclusions
In this study, we have shown that patients with stage III-IV breast cancer expressing PKCλ high and ALDH1A3 high have poorer clinical outcomes than those expressing PKCλ low and ALD-H1A3 low . Furthermore, PKCλ deficiency led to suppression of cell survival, tumor-sphere formation, migration, and asymmetric cell propagation, as well as intracellular accumulation of ROS in ALDH1-positive breast CSCs. PKCλ thus appears to be essential for the regulation for these cellular properties of ALDH1-positive breast CSCs. We therefore conclude that high PKCλ expression is required for ALDH1-postive cancer stem cell function and indicates a poor clinical outcome in late-stage breast cancer patients.