Optimized Dosing Schedule Based on Circadian Dynamics of Mouse Breast Cancer Stem Cells Improves the Antitumor Effects of Aldehyde Dehydrogenase Inhibitor

According to the World Health Organization, breast cancer is the most common cancer in patients worldwide (1). However, current treatment strategies cannot eliminate the majority of breast cancers. In particular, triple-negative breast cancer (TNBC) is highly aggressive (2). TNBC is usually resistant to chemotherapeutic drugs and has also been incriminated in recurrence after chemotherapy, radiotherapy, and resection surgery (3).

One approach for increasing the efficacy of pharmacotherapy is to administer drugs at the time of day when they are most effective and/or best tolerated. A chronopharmacologic strategy can enhance the effects of drugs and/or attenuate their toxicity (4, 5). Circadian variations in biological functions, such as gene expression and protein synthesis, are thought to be important factors affecting the efficacy of drugs. Experimental chronopharmacology studies have successfully guided the development of chronotherapy schedules with 5-fluorouracil, leucovorin, and oxaliplatin in human colorectal cancer. Chronomodulated chemotherapy regimens have also produced the highest tumor response rates and the longest survival reported in multicenter randomized trials (6–9). The circadian rhythm in the tolerability and the efficacy of docetaxel and doxorubicin in mice bearing syngeneic mammary cancer tumors derived from MA13/C cells is investigated as a prerequisite for the development of chronotherapy schedules with these drugs in human breast cancer (10). However, there is no previous study evaluating chronotherapy for treatment of TNBC.

Tumor masses are composed of heterogeneous cells, and this heterogeneity is relevant to resistance to chemotherapy and a high risk of recurrence. Cancer stem-like cells (CSC) represent a distinct proportion in cancer cells, but they play a key role in driving tumor growth, progression, and metastasis owing to their self-renewal and differentiation capacities. CSCs in TNBC tumor mass are also considered to be responsible for metastasis, recurrence, and resistance against chemotherapy and radiotherapy (11, 12). Because CSCs are characterized by specific cell surface markers (13, 14), this subpopulation of cells can be isolated from mixed tumorigenic and nontumorigenic cells using different immune selection methods (15). However, a limitation of surface marker recognition approaches is that the results are ascribed to the specific studied population (16). In addition, because the characteristics of CSCs are complex, general studies of CSCs should be more thorough and provide more data to confirm the population as CSCs (17, 18). Therefore, a more useful method involving detection of the activity of a specific protein in CSCs has been developed to identify and/or isolate CSCs in both basic research and the clinical settings (19). High aldehyde dehydrogenase (ALDH) activity is often detected in cells with stem-like properties, suggesting that this enzyme can be used as a marker to isolate CSCs (20); the ALDEFLUOR assay measures ALDH enzyme activity via cleavage of a fluorescent substrate, BODIPY-aminoacetaldehyde, and is a commonly used method to identify and isolate CSCs (21, 22).

In this study, we used the ALDEFLUOR assay to investigate the ALDH activity in mice implanted with murine TNBC 4T1 cells. The number of ALDH-positive cells (high ALDH activity cells) in a mouse breast tumor model exhibited pronounced circadian alterations, which was caused by the time-dependent release of Wingless-type mmtv integration site family 10a (WNT10a) from ALDH-negative cells. Therefore, we investigated whether antitumor and antimetastatic effects of ALDH inhibitor N,N diethylaminobenzaldehyde (DEAB) were improved by changing the dosing schedule.

4T1 mouse breast cancer cells were purchased from the American Type Culture Collection. Cells were cultured under a 5% CO₂ environment at 37°C in Roswell Park Memorial Institute (RPMI)-1640 medium supplemented with 10% FBS and 1% penicillin/streptomycin in two-dimensional CELLSTAR cell culture flasks (Greiner Bio-One) or synthetic three-dimensional (3D) scaffold biomaterials (VECELL 3D-inserts; VECELL Inc.). We confirmed that there was no microbial in this cell line using a TaKaRa PCR Mycoplasma Detection Set. We confirmed that cell lines were authenticated by each cell bank using short tandem repeat– PCR analysis, and these cell lines were used in less than 3 months from frozen stocks. We confirmed that there was no microbial growth in both cell lines using fluorochrome staining. To downregulate the Aldh3a1 gene, 4T1 cells were infected with lentiviral vectors expressing shRNA against the mouse Aldh3a1 gene (pGFP-C-shAldh3a1 Lenti Vector; Origene Technologies, Inc.). After infection of cells with lentiviral vectors, cells were maintained in a medium containing 2 μg/mL puromycin. GFP-expressing cells were selected by sorting using fluorescence-activated cell sorting (FACS; BD Biosciences). Downregulation of the Aldh3a1 gene was confirmed by reverse transcription (RT)-PCR. Aldh3a1::Luc-expressing 4T1 cells were prepared using luciferase reporter vectors under the control of the mouse Aldh3a1 promoter. The mouse Aldh3a1 promoter region spanning from −660 to +15 bp (the distance in base pairs from the putative transcription start site, +1) was amplified by PCR, and the product was ligated into the pGL4.18 luciferase reporter vector (Promega). The primer sequences used for amplification of the mouse Aldh3a1 promoter region were as follows: forward primer, 5′-ATACTCGAGACTGGCTAAACATACAGAAAGGG-3′; reverse primer, 5′- ATAAGATCTTGGAACTCCTGGAATAAGCAAG-3′. After transfection of Aldh3a1::Luc vectors into 4T1 cells, transgene-expressing cells were selected with G418 (Wako Chemicals). Individual colonies were expanded and maintained in medium containing 1 μg/mL G418. The activity of Aldh3a1::Luc was assessed using luciferase assays. To prepare Wnt10a-downregulated Aldh3a1::Luc-expressing 4T1 cells, Aldh3a1::Luc-expressing 4T1 cells were infected with lentiviral vectors expressing shRNA against the mouse Wnt10a gene (pGFP-C-shWnt10a Lenti Vector; Origene Technologies, Inc.). After infection of cells with lentiviral vectors, cells were maintained in medium containing 2 μg/mL puromycin. GFP-expressing cells were selected by sorting using FACS (BD Biosciences). Downregulation of the Wnt10a gene was confirmed by RT-PCR.

ALDH-positive (ALDH-high activity) and ALDH-negative (ALDH-low activity) cells were gated based on the ALDEFLUOR assay (StemCell Technologies) according to the manufacturer's instructions. Briefly, dissociated single cells from cell lines or tumor specimens were suspended in ALDEFLUOR assay buffer containing an ALDH substrate, BODIPY-aminoacetaldehyde, at 1.5 μmol/L; this was followed by incubation for 40 minutes at 37°C. A specific inhibitor of ALDH, DEAB (Sigma-Aldrich), was used at a 10-fold molar excess as a negative control. FACS (BD Biosciences) was performed on more than 1 × 10⁶ cells under low pressure in the absence of ultraviolet light. The data were analyzed using BD FACSDiva software V6.1.3 (BD Biosciences).

Five-week-old female BALB/c mice (Kyudo Co., Ltd.) were housed under a standardized light-dark cycle at 24 ± 1°C and 60% ± 10% humidity with food and water ad libitum. Thirty microliters medium containing 5 × 10⁴ native 4T1 cells, Aldh3a1-downregulated 4T1 cells, Aldh3a1::Luc-expressing 4T1 cells, or Wnt10a-downregulated Aldh3a1::Luc-expressing 4T1 cells was injected into the right hind footpads of the mice. Tumor volume was estimated according to the following formula: tumor volume (mm³) = 4π(XYZ)/3, where 2X, 2Y, and 2Z are the three perpendicular diameters of the tumor. After the tumor size reached 300 mm³, experiments were performed. On day 21 after implantation of 4T1 tumor cells into the mice (day 3 after the last DEAB injection), the lungs were removed, rinsed, and fixed in Bouin solution to heighten the contrast between tumor nodules and normal lung parenchyma. The numbers and sizes of metastatic tumor cells were determined under a dissecting microscope. A solution of DEAB (Sigma-Aldrich) and Wnt-C59 (Cellagen Technology) was prepared by dissolving in 5% dimethyl sulfoxide in 95% olive oil. The drugs were injected using a 30-gauge needle. All experimental procedures were performed under the approval and guidelines of Kyushu University.

Total RNA was extracted using RNAiso (Takara Bio Co., Ltd.) or a QIAGEN RNeasy Mini kit (Qiagen). cDNA was synthesized using a ReverTra Ace qPCR RT kit (Toyobo) and amplified by PCR. Real-time PCR analysis was performed on diluted cDNA samples using the THUNDERBIRD SYBR qPCR Mix (Toyobo) with the 7500 Real-time PCR system (Applied Biosystems). Data were normalized using 18s and β-actin mRNAs as controls because spinal expression of these MRAs is constant throughout the day. Primer sequences are listed in Supplementary Table S1.

Protein samples were prepared from ALDH-positive and ALDH-negative 4T1 cells using CelLytic MT Cell Lysis Reagent (Sigma-Aldrich) supplemented with protease inhibitor cocktail, which contained 2 μg/mL aprotinin, 2 μg/mL leupeptin, and 100 μmol/L phenylmethylsulfonyl fluoride. Then, 20 μg of the protein lysate was resolved by SDS–PAGE on 10% or 12% gels, transferred to polyvinylidene difluoride membranes, and probed with antibodies against ALDH3A1 (ab76976; Abcam), WNT10a (SAB3500393; Sigma-Aldrich), and β-ACTIN (sc-1616; Santa Cruz Biotechnology). Specific antigen–antibody complexes were visualized using horseradish peroxidase–conjugated secondary antibodies and a chemiluminescence reagent.

ALDH-positive and ALDH-negative 4T1 cells were prepared from 4T1 tumor–implanted mice at zeitgeber time (ZT)0 and ZT12 (ZT0, lights on; ZT12, lights off). Total RNA was extracted from cells using a QIAGEN RNeasy Mini Kit (Qiagen). The quality of the total RNA was checked using an Agilent 2200 TapeStation (Agilent Technologies). Then, 50 ng total RNA for each gene was used for the labeling reaction with the one-color protocol of an Agilent Low-Input QuickAmp Labeling kit (Agilent Technologies). Labeled RNA was hybridized to a 60K Agilent 60-mer SurePrint technology (SurePrint G3 Mouse Gene Expression 8 × 60K Microarray Kit version 2.0) according to the manufacturer's protocol. All hybridized microarray slides were washed and scanned using an Agilent scanner. Relative hybridization intensities and background hybridization values were calculated using Agilent Feature Extraction software (version 9.5.1.1). Raw signal intensities and flags for each probe were calculated from hybridization intensities and spot information according to procedures recommended by Agilent. The raw signal intensities of two samples were log₂-transformed and normalized using a quantile algorithm in the “preprocessCore” library package of the Bioconductor software (www.bioconductor.org). This produced a gene expression matrix consisting of 55,681 probe sets; differentially expressed genes between samples were selected using a Z-score of 2.0 or more and a ratio of 1.5-fold or more. For downregulated genes, a Z-score of –2.0 or less and a ratio of 0.75 or less were used. Functional analysis of the differentially expressed genes was performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database on the DAVID system (23). The full data have been deposited in National Center for Biotechnology Information gene expression omnibus (Accession#:GSE103598).

Cross-linked chromatin extracted from cells of the spinal cord was sonicated on ice, and nuclear fractions were obtained by centrifugation at 10,000 × g for 5 minutes. Supernatants were incubated with antibodies against β-catenin (#8480; Cell Signaling Technology), Transcription factor 7-like 2 (TCF7L2; #2569; Cell Signaling Technology), or rabbit IgG (sc66931; Santa Cruz Biotechnology). DNA was purified using a DNA purification kit (Promega) and amplified by PCR for the surrounding β-catenin/TCF binding site in the upstream region of the Aldh3a1 gene. The primer sequences for the amplification of the surrounding β-catenin/TCF binding site were as follows: 5′-CCTGGGTGATACAGAGAGGA-3′ and 5′-CACAACCTACTGGTTGGAGA-3′. The quantitative reliability of PCR was evaluated through kinetic analysis of the amplified products to ensure that signals were only derived from the exponential phase of amplification. This analysis was also performed in the absence of an antibody or in the presence of rabbit IgG as negative controls; no PCR products were detected with ethidium bromide staining in any samples.

An in vivo imaging system (IVIS Spectrum; Caliper Life Sciences Inc.) was used for in vivo bioluminescence monitoring (24). Mice implanted with Aldh3a1::Luc-expressing 4T1 cells or Wnt10a-downregulated Aldh3a1::Luc-expressing 4T1 cells were subcutaneously injected with 15 mg/kg D-luciferin potassium salt (Wako Chemicals) dissolved in PBS. Under isoflurane anesthesia with concentrated oxygen and a gas anesthesia system (BleaseDatum Vaporizer; Spacelabs Healthcare, Inc.), images were acquired 6 and 12 minutes after D-luciferin injection (5-second exposure time). The duration under isoflurane anesthesia was approximately 20 minutes for each experiment. After data were obtained, mice recovered from isoflurane anesthesia within 1 to 2 minutes.

The bioluminescence of cultured Aldh3a1::Luc-expressing 4T1 cells was recorded using a real-time monitoring system (Lumicycle; Actimetrics). The ALDH-positive populations of Aldh3a1::Luc-expressing 4T1 cells were cultured on VECELL 3D inserts. The 3D inserts were placed in 35-mm dishes, in which ALDH-negative 4T1 cells were seeded and stimulated with 100 nmol/L dexamethasone for synchronization of their circadian clocks. The amplitude of bioluminescence derived from Aldh3a1::Luc was calculated using Lumicycle analysis software (Actimetrics). Bioluminescence images at the cellular level were acquired using the LV200 LuminoView microscope system (Olympus).

The values presented are expressed as mean ± SEM. The significance of the 24-hour variations in each parameter was tested by one-way ANOVA. The statistical significance of differences between groups was analyzed by one-way or two-way ANOVA, followed by Tukey–Kramer post hoc tests and Dunnett test. Equal variances were not formally tested. A 5% level of probability was considered significant.

To investigate the malignancy of ALDH-positive cells, which indicated high activity of ALDH, isolated from mouse 4T1 breast cancer cell lines, we implanted ALDH-positive cells into the hind footpads of female BALB/c mice (Supplementary Fig. S1A). The ALDH-positive cells exhibited profound tumor formation as well as enhanced pulmonary metastasis (Supplementary Fig. S1B). Furthermore, the expression levels of known CSC biomarkers, SRY-related HMG-box-2 (Sox-2), POU domain, class 5, transcription factor 1 (Pou5f1), and Nanog (25–27), in ALDH-positive cells were significantly higher than those in ALDH-negative cells (Supplementary Fig. S1C), confirming that ALDH-positive cells have stem-like properties.

Next, we elucidated whether ALDH-positive cells, which indicated high activity of ALDH, in 4T1 tumor exhibited circadian characteristics. To achieve this, we assessed the number of ALDH-positive cells in 4T1 breast cancer tumor–bearing mice kept under a 12-hour light-dark cycle. After the tumor size reached 300 mm³, tumor masses were removed at six different time points and prepared as single-cell suspensions. The absence of necrosis in the suspensions was confirmed by microscopic observation. The results of ALDEFLUOR assay and FACS analysis revealed that a proportion of ALDH-positive cells in 4T1 tumor masses exhibited significant circadian variations (F_5,12 = 6.534, P = 0.004, one-way ANOVA; Fig. 1A). The number of ALDH-positive 4T1 cells increased from the late light phase to the early dark phase.

Because 17 Aldh genes have been identified in mammals (28), we next attempted to identify the Aldh gene responsible for the circadian alterations in ALDH activity in 4T1 tumor cells implanted in mice. The mRNA expression of Aldh3a1 and Aldh6a1, which showed high expression in cultured ALDH-positive cells by microarray analysis, was detected in 4T1 tumor masses. The mRNA levels for Aldh3a1, but not for Aldh6a1, in 4T1 cell–implanted mice showed a significant circadian oscillation (F_5,30 = 2.836, P = 0.033; one-way ANOVA; Fig. 1B, left; Supplementary Fig. S2). A similar rhythmic variation was also detected in the protein levels of ALDH3a1 (F_5,12 = 7.355, P = 0.002; one-way ANOVA; Fig. 1B, right), suggesting that the rhythmic expression of the Aldh3a1 gene was related to circadian alterations in ALDH activity in 4T1 tumor masses.

To investigate this possibility, we prepared Aldh3a1-downregulated 4T1 cells via introduction of lentivirus vectors expressing a specific shRNA (Supplementary Fig. S3A). The growth and pulmonary metastasis of tumors formed by Aldh3a1-downregulated 4T1 cells was significantly slower than those observed in mice implanted with control shRNA–expressing 4T1 cells (Supplementary Fig. S3B and S3C). Furthermore, the proportion of ALDH-positive cells in Aldh3a1-downregulated 4T1 tumors did not show significant circadian alterations (Fig. 1C). These data suggest that the Aldh3a1 gene is responsible for generating circadian alterations in ALDH activity in 4T1 tumor masses. This circadian alteration in ALDH activity appeared to be reflected in the time-dependent changes in the number of ALDH-positive cells in 4T1 tumor masses.

Next, we investigated the mechanism through which Aldh3a1 was expressed in a circadian manner in 4T1 tumor cells. In mammalian cells, circadian rhythms in gene expression are generated by a molecular oscillator driven by a transcriptional–translational feedback loop consisting of clock genes (29). Therefore, we assessed the temporal expression profiles of clock genes in both ALDH-positive and ALDH-negative populations of 4T1 tumors. The number of ALDH-positive cells in 4T1 tumors varied with the time of day; accordingly, we collected at least 10,000 ALDH-positive cells at 6 different time points and extracted their RNA. Although the mRNA levels of the main components of the circadian clock, i.e., Clock, Bmal1, Period2 (Per2), Cryptochome1 (Cry1), and Rev-erbα, did not exhibit significant circadian oscillations in ALDH-positive populations of 4T1 cells, their expression exhibited significant circadian oscillations in ALDH-negative 4T1 cells (F_5,12 = 6.278, P = 0.004 for Clock; F_5,12 = 21.714, P < 0.001 for Bmal1; F_5,12 = 5.034, P = 0.010 for Per2; F_5,12 = 5.522, P = 0.007 for Cry1; F_5,12 = 5.621, P = 0.007 for Rev-erbα; one-way ANOVA; Fig. 2A). Because ALDH-negative cells constituted the microenvironment, these findings suggest that the circadian oscillator functions in microenvironmental cells rather than CSCs in 4T1 tumors.

Considering the dysfunction of circadian machinery in CSCs, we hypothesized that the circadian expression of Aldh3a1 in ALDH-positive 4T1 cells was regulated by soluble factors released from surrounding microenvironmental cells. To identify the factors responsible for regulating the diurnal expression of Aldh3a1 in 4T1 tumor cells, we performed oligonucleotide microarray analyses using RNA isolated from ALDH-positive and ALDH-negative populations from 4T1 tumors implanted in mice at ZT0 and ZT12; at these time points, the expression of Aldh3a1 in 4T1 tumors decreased and increased, respectively (Fig. 1B). Three criteria were applied to select circadian cycle-dependent genes that regulate Aldh3a1 expression: (1) the expression of genes in ALDH-negative cells being greater than that in ALDH-positive cells, (2) the expression of genes in ALDH-negative cells at ZT0 being greater than that in ALDH-negative cells at ZT12, (3) the expression of genes in ALDH-negative cells at ZT12 being greater than that in ALDH-negative cells at ZT0. From this analysis, we identified 618 candidate circadian time-dependent genes in ALDH-negative cells (Supplementary Table S2). Functional analysis of these genes using the KEGG database (23) revealed that 19 biological pathways were enriched in a statistically significant manner (P < 0.05; Supplementary Table S3).

Among the genes involved in pathways related to cancer progression, we focused on those that encode WNT proteins, because WNT proteins are secreted molecules that act on cell- surface receptors, and also WNT signal transduction has been implicated in sustaining the stemness of CSCs (30). To determine whether WNT signaling is involved in the circadian regulation of Aldh3a1, the mRNA levels of Aldh3a1 were assessed in 4T1 tumor cells implanted in mice after the intratumoral injection of a canonical WNT signal inhibitor, Wnt-C59, at ZT2 and ZT14. Four hours after the injection of Wnt-C59 at each time point, the mRNA levels of Aldh3a1 did not exhibit a significant time-dependent variation (Fig. 2B), suggesting that WNT signaling is involved in the circadian regulation of Aldh3a1 expression in 4T1 tumor cells.

Extracellular WNT stimulates several intracellular signal transduction cascades, resulting in the activation or repression of a variety of genes (31). The major effector of these transduction cascades is a bipartite transcription factor formed by β-catenin and a member of the TCF protein family, such as TCF7L2 (32). A consensus DNA sequence of the β-catenin/TCF binding site CTTTGA is located between −532 and −538 bp from the transcription start site of the mouse Aldh3a1 gene (Fig. 2C, left). The DNA sequence of the β-catenin/TCF binding site has also been found at a similar location in all mammals examined, including mice, rats, monkeys, and humans (Supplementary Fig. S4A). Thus, the luciferase reporter of the mouse Aldh3a1 promoter containing the motif CTTTGA (Aldh3a1::Luc) responded to β-catenin and TCF7L2 (Supplementary Fig. S4B). The results of chromatin immunoprecipitation also revealed that both β-catenin and TCF7L2 bound to the promoter region of the Aldh3a1 gene in 4T1 tumors, and the amount of bound protein was increased at ZT14 compared with that at ZT2 (Fig. 2C, right).

To determine whether the upstream region containing the β-catenin/TCF binding site was responsible for the circadian expression of Aldh3a1, we prepared 4T1 cells that stably expressed Aldh3a1::Luc (Supplementary Fig. S5A). After confirming that the luciferase activity of the Aldh3a1::Luc-expressing cells was mainly driven by ALDH-positive populations (Supplementary Fig. S5B), we implanted these cells into the hind footpads of mice. In vivo imaging analysis results revealed that the bioluminescence from tumors formed by Aldh3a1::Luc-expressing 4T1 cells also showed significant circadian oscillation, with peak levels during the early dark phase (F_5,34 = 15.427, P < 0.001, one-way ANOVA; Fig. 2D). The rhythmic pattern of the bioluminescence from Aldh3a1::Luc-expressing 4T1 tumors resembled the overall rhythm of the expression of ALHD3a1 (Fig. 1B).

Several genes encoding WNT ligands were highly expressed in ALDH-negative 4T1 cells (Supplementary Fig. S6). Among these, the expression of Wnt10a mRNA exhibited profound circadian oscillation only in ALDH-negative cells (F_5,12 = 17.117, P < 0.001, one-way ANOVA; Fig. 3A). In vitro promoter analysis revealed that the transcription of Wnt10a was controlled by the main components of the circadian clock (Supplementary Fig. S7). The CLOCK/BMAL1-mediated transactivation of Wnt10a was repressed by PER2 and CRY1, suggesting that PER and CRY proteins periodically repress the CLOCK/BMAL1-mediated transactivation of the Wnt10a gene. Positive and negative regulation by the products of circadian clock genes appeared to generate a circadian rhythm in the mRNA and protein expression of Wnt10a. Although the expression of the WNT10a protein in ALDH-positive 4T1 cells was not detected by Western blot analysis, its protein levels in ALDH-negative populations of 4T1 tumors exhibited a significant time-dependent variation (F_1.14 = 9.699, P = 0.036, two-way ANOVA; Fig. 3B). In contrast, the ALDH3a1 protein was difficult to detect in ALDH-negative populations of 4T1 tumors; however, the protein levels showed a significant time-dependent variation in ALDH-positive 4T1 cells (F_1,4 = 20.415, P = 0.011, two-way ANOVA; Fig. 3B). These results suggest the possibility that extracellularly produced WNT10a from microenvironmental cells act on ALDH-positive CSCs to induce Aldh3a1 expression.

To investigate this possibility, we prepared Wnt10a-downregulated Aldh3a1::Luc-expressing 4T1 cells (Supplementary Figs. S8A and S8B, and S9) and implanted them into the hind footpads of mice. The bioluminescence from tumors formed by control Aldh3a1::Luc-expressing 4T1 cells (expressing the control shRNA) showed a significant time-dependent variation (P < 0.01, Fig. 3C). However, the variation in the bioluminescence from Aldh3a1::Luc-expressing 4T1 tumors was dampened by the downregulation of Wnt10a. The intensity of the bioluminescence from Wnt10a-downregulated 4T1 tumors remained low at both the light and dark phases. Consistent with these observations, the downregulation of Wnt10a in 4T1 cells decreased the number of ALDH-positive cells and dampened their circadian oscillation (Fig. 3D), supporting the notion that WNT10a is a major regulator of the circadian expression of the Aldh3a1 gene in 4T1 tumor cells.

In addition to analysis of the 4T1 tumor mass, we also detected significant circadian accumulation of the WNT10a protein in the culture medium of ALDH-negative 4T1 cells after synchronizing their circadian clocks by dexamethasone treatment (F_13,28 = 6.325, P < 0.001, one-way ANOVA; Fig. 4A). This finding suggests that tumor microenvironmental cells enhance the release of WNT10a in a circadian fashion. The rapid degradation of WNT protein has been reported previously (33). The half-life of the WNT10a protein in the medium was approximately 4 hours (Fig. 4B). Consequently, circadian accumulation of WNT10a in the culture medium of ALDH-negative 4T1 cells may be associated with both its time-dependent synthesis and rapid degradation.

To determine whether the time-dependent release of WNT10a from tumor microenvironmental cells causes circadian expression of Aldh3a1 in ALDH-positive 4T1 cells, dexamethasone-treated ALDH-negative 4T1 cells were spatially cocultured with ALDH-positive cells that were isolated from Aldh3a1::Luc-expressing 4T1 tumors. Because CSC stemness decreases under normal culture conditions, Aldh3a1::Luc-expressing ALDH-positive 4T1 cells were maintained in transwell inserts (chambers) containing a 3D scaffold (34). The 3D scaffold chambers were inserted into wells in which ALDH-negative 4T1 cells were seeded on the bottom (Fig. 4C, left). Aldh3a1::Luc-expressing ALDH-positive 4T1 cells showed a significant time-dependent oscillation of bioluminescence when cocultured with dexamethasone-treated ALDH-negative 4T1 cells (Fig. 4C, right; Supplementary Movie S1); however, this oscillation did not occur when they were cocultured with dexamethasone-untreated ALDH-negative 4T1 cells (Supplementary Movie S2). These results suggest that the time-dependent enhancement of WNT10a released from tumor microenvironmental cells regulates the circadian expression of the Aldh3a1 gene in ALDH-positive 4T1 cells.

Because the number of ALDH-positive cells showed significant circadian variation in 4T1 tumor–bearing mice (Fig. 1), we investigated whether antitumor and antimetastatic effects of ab ALDH inhibitor were changed by optimizing dosing schedule. DEAB is commonly used as a selective inhibitor of ALDH in CSCs (35). The intraperitoneal administration of DEAB (50 mg/kg) at ZT14, when the number of ALDH-positive cells was abundant, significantly suppressed the growth and pulmonary metastasis of 4T1 tumor cells in mice (Fig. 5A and B). In contrast, administration of the same dosage of DEAB at ZT2 did not result in significant antitumor as well as antimetastatic activity. These data reveal a significant relationship between the circadian alterations in the number of ALDH-positive 4T1 cells and their malignant characteristics. Choosing the most appropriate dosing time can improve the antitumor and antimetastatic effects of ALDH inhibitor on malignant TNBC.

Although hormone replacement therapy is effective in patients with predominantly estrogen receptor–positive breast cancers, patients with TNBC have poor prognosis (2). The intratumoral heterogeneity consisting of CSCs and their environment cells could be explained for malignancies and high recurrence of TNBC. Our results suggest that circadian variation in the number of ALDH-positive cells in 4T1 tumor–bearing mice plays an important role affecting the antitumor and antimetastatic effects of ALDH inhibitor. The activity of ALDH was regulated by time-dependent variations in WNT10a released from ALDH-negative cells (Fig. 5C). This circadian interaction between ALDH-positive and ALDH-negative cells suggested a potential therapeutic target as chronotherapy for treatment of TNBC.

ALDHs exhibit a wide taxonomic distribution from bacteria to humans. They catalyze the conversion of aldehydes to corresponding acids via an NAD(P)⁺-dependent irreversible reaction (36). The family of ALDH contributes to sustain the stemness of cancer cells; therefore, their activity is used as a marker for CSCs (20). Implantation of ALDH-negative cells to mice showed slow tumor growth and poor metastasis. Furthermore, inhibition of ALDH activity resulted not only in the disruption of circadian variations in the number of ALDH-positive cells, but also in the prevention of tumor growth and metastasis. ALDH has the ability to ameliorate oxidative stress in tumor cells (37). Therefore, development of method selectively inhibiting ALDH activity in CSCs would be useful for treatment of malignant cancers including TNBC.

The circadian clock machinery in immature cells, e.g., ES cells and stem-like cells, is functional after differentiation (38, 39). Low-level expression of clock genes in those immature cells is thought to be important for sustaining their stemness (11). The expressions of clock gene in ALDH-positive cells were lower than those in ALDH-negative cells, suggesting that dysfunction of circadian machinery is also required for sustaining the stemness of CSCs in 4T1 tumors. Overexpression of Aldh3a1 in 4T1 cells slightly increased the mRNA levels of Bmal1 and Rev-Erbα, but in contrary decreased the expression of Clock (Supplementary Fig. S10). Although we were unable to clarify whether ALDH was indispensable for suppression of circadian clock machinery in CSCs, low-level expression of these clock genes in ALDH-positive cells was unlikely due to the elevation of ALDH enzymatic activity. Recent study has demonstrated that pharmacologic activation of REV-ERBs is specifically lethal in cancer cells (40). However, such strategy may be ineffective to CSCs in 4T1 tumors because of low expression of clock genes including Rev-erbα.

Circadian expression of Aldh3a1 in ALDH-positive cells was dependent on Wnt/β-catenin signaling. The expression of WNT10a in ALDH-negative cells was controlled by the components of circadian clock. The results of coculture experiment supported the notion that the time-dependent enhancement of WNT10a release from ALDH-negative cells caused the circadian expression of Aldh3a1 in the ALDH-positive CSCs. In mammals, WNT/β-catenin signaling is prominent in stem cells and cancer cells (41). WNT proteins act as critical microenvironmental factors for sustaining the stem cells in a self-renewing state. In 4T1 tumor masses, ALDH-negative cells were distinct from ALDH-positive cells, confirming by the difference in the expression levels of typical CSC markers. Although ALDH-positive and ALDH-negative cells were somewhat identical to each other, there were obvious genetic differences between those cell populations.

It has been well known that microenvironment surrounding CSCs is composed by tumor-associated fibroblasts, macrophages, myeloid-derived suppressor cells, and/or regulatory T cells (42). In addition, stromal fibroblasts are also involved in microenvironmental constitutive cells, which is capable of releasing WNT proteins (43, 44). Because gain of β-catenin activity allows stem cell overpopulation and cancer development (45), disruption of the circadian machinery in microenvironment cells may lead to arrhythmic expression of WNT10a, therefore enhancing malignancies of TNBC. This notion is also supported by previous findings that circadian host disruption accelerates the progression of many types of cancer, including those without any clock or barely detectable Bmal1 expression and shorten survival (46, 47). Disruption of circadian rhythms in host animal also accelerates tumor growth and angiogenesis/stromagenesis through the mediation of Wnt signaling pathway (48). In contrast, the reinforcement of the host circadian clock or that of cancer tissues slows proliferation of tumor cells in relation to the timing of meals or kinase inhibitor administration (49, 50). Consequently, the maintenance of circadian clock function in microenvironment may be important to suppress overpopulation of CSCs.

Our present findings suggest that the effectiveness of anticancer drugs varies with the circadian dynamics of CSCs, which are regulated by the tumor microenvironmental factors. However, many drugs are still administered without regard to the time of day. Identification of rhythmic markers for detection of the circadian dynamics of CSCs in tumors should enable their use in chronotherapy, in which chemoradiation and/or high-dosage treatments are administered at a time of day when CSCs are most vulnerable. Furthermore, circadian machinery in tumor microenvironment may be a therapeutic target of TNBC, because the machinery was participated in the regulation of tumor malignancies.

No potential conflicts of interest were disclosed.

Conception and design: N. Matsunaga, S. Koyanagi, S. Ohdo

Development of methodology: N. Matsunaga

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Matsunaga, T. Ogino, Y. Hara, T. Tanaka

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Matsunaga, S. Koyanagi

Writing, review, and/or revision of the manuscript: N. Matsunaga, S. Koyanagi, S. Ohdo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Matsunaga

Study supervision: N. Matsunaga, S. Ohdo

This work was supported in part by a Grant-in-Aid for Scientific Research A (16H02636 to S. Ohdo), Grant-in-Aid for Challenging Exploratory Research (17H06262 to S. Ohdo), and Scientific Research C (15K08098 to N. Matsunaga) from Japan for the Promotion of Science. This research was supported by Platform Project for Supporting Drug Discovery and Life Science Research [Basis for Supporting Innovative Drug Discovery and Life Science Research (BINDS)] from AMED under Grant Number JP18am0101091. VECELL 3D plates were a gift from Makoto Kodama, PhD (VECELL, Inc.).

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.