Abstract

   Triple‐negative breast cancer (TNBC), the most aggressive subtype of
   breast cancer, has a poor prognosis and lacks effective treatment
   strategies. Here, the study discovered that TNBC shows a decreased
   expression of epithelial transcription factor ovo‐like 2 (OVOL2). The
   loss of OVOL2 promotes fatty acid oxidation (FAO), providing additional
   energy and NADPH to sustain stemness characteristics, including
   sphere‐forming capacity and tumor initiation. Mechanistically, OVOL2
   not only suppressed STAT3 phosphorylation by directly inhibiting JAK
   transcription but also recruited histone deacetylase 1 (HDAC1) to
   STAT3, thereby reducing the transcriptional activation of downstream
   genes carnitine palmitoyltransferase1 (CPT1A and CPT1B). PyVT‐Ovol2
   knockout mice develop a higher number of primary breast tumors with
   accelerated growth and increased lung‐metastases. Furthermore,
   treatment with FAO inhibitors effectively reduces stemness
   characteristics of tumor cells, breast tumor initiation, and
   metastasis, especially in OVOL2‐deficient breast tumors. The findings
   suggest that targeting JAK/STAT3 pathway and FAO is a promising
   therapeutic strategy for OVOL2‐deficient TNBC.

   Keywords: FAO, JAK/STAT3, Ovol2, stemness characteristics, TNBC
     __________________________________________________________________

   Triple‐negative breast cancer (TNBC), the most aggressive breast cancer
   subtype, exhibits downregulated expression of the epithelial
   transcription factor OVOL2. OVOL2 suppresses JAK transcription,
   recruits HDAC1 and inhibits STAT3 activation, reducing downstream gene
   (CPT1A and CPT1B) expression. OVOL2 loss enhances fatty acid oxidation
   (FAO), fueling stemness traits and tumor initiation. Targeting
   JAK/STAT3 and FAO is a promising strategy for OVOL2‐deficient TNBC.

   graphic file with name ADVS-11-2308945-g007.jpg

1. Introduction

   Triple‐negative breast cancer (TNBC), characterized by the absence of
   estrogen receptor (ER), progesterone receptor (PR), and human epidermal
   growth factor receptor 2 (HER2), is the most aggressive subtype
   accounting for 10–20% of all breast cancer cases.^[ [48]^1 ^] It has a
   poor prognosis and a high risk of recurrence. The presence of highly
   stemness characteristics in TNBC contributes to rapid oncogenic growth
   and distant metastasis.^[ [49]^2 ^] Traditional hormone‐targeted
   therapies and HER2‐targeted therapies are ineffective in treating TNBC
   due to the absence of ER, PR, and HER2 receptors. Chemotherapy remains
   the primary option, albeit with limited effectiveness.^[ [50]^3 ,
   [51]^4 ^] Therefore, it is essential to develop new strategies for TNBC
   therapy.

   Fatty acid β‐oxidation (FAO) plays a crucial role in the metabolism of
   cancer cells, providing a substantial amount of energy essential for
   sustaining rapid cell growth and tumor progression. Moreover, FAO
   generates abundant antioxidants, such as NADPH, which promotes the
   survival of tumor cells.^[ [52]^5 ^] The emerging roles of FAO in
   cancer progression, chemotherapy resistance, and the maintenance of
   cancer stem cells (CSCs) have gained significant attention. Inhibition
   of FAO has been proposed as a potential therapeutic approach to
   counteract tumor growth in breast cancer.^[ [53]^6 , [54]^7 ^] However,
   the specific regulatory mechanisms of FAO in TNBC remain unclear.

   The signal transducer and activator of transcription 3 (STAT3) is one
   of the seven members of the STAT family. Canonical STAT3 signaling
   plays a crucial role in numerous biological processes, including cell
   proliferation, survival, differentiation, and angiogenesis.^[ [55]^8 ,
   [56]^9 ^] IL‐6 and diverse cytokines exert actions via the
   transmembrane receptor gp130, triggering Janus kinases (JAKs)/STAT3
   signaling.^[ [57]^10 ^] Subsequently, STAT3 is recruited to the
   activated cytokine receptor and phosphorylated by receptor‐associated
   JAK at tyrosine. Phosphorylated STAT3 could form a dimer, translocate
   into the nucleus, and ultimately activate transcriptional of downstream
   genes.^[ [58]^11 , [59]^12 ^] Recent clinical and preclinical data
   revealed that overexpression and constitutively activation of STAT3
   were involved in the progression, proliferation, metastasis, and
   chemoresistance of breast cancer.^[ [60]^13 ^] Moreover, JAK/STAT3
   activates transcription of carnitine palmitoyl transferase 1B (CPT1B)
   facilitating FAO which is crucial for breast cancer stem cell renewal
   and chemoresistance.^[ [61]^14 ^] STAT3 is persistently activated in
   the context primarily linked to triple‐negative tumors,^[ [62]^15 ^]
   however the precise mechanistic by which sustained STAT3 activation in
   TNBC remains largely unknown.

   Ovo‐like 2 (OVOL2), a transcription repressor factor, plays a vital
   role in cell differentiation, epithelial‐mesenchymal transition (EMT),
   and tissue development. Down‐regulation of OVOL2 has been found in many
   kinds of cancers.^[ [63]^16 , [64]^17 , [65]^18 ^] We previously found
   that OVOL2 loss is associated with malignant progression in colorectal
   cancer.^[ [66]^19 ^] OVOL2 suppresses the TGFβ1‐induced EMT process in
   breast cancer.^[ [67]^20 ^] However, the role of OVOL2 in regulating
   fatty acid oxidation, particularly in TNBC, remains largely unknown.

   Here, we have observed a significant decrease in OVOL2 expression in
   TNBC, which is intricately linked to the persistent activation of
   STAT3. RNA‐seq analysis revealed that OVOL2 inhibits fatty acid
   oxidation via STAT3 signaling. Furthermore, we demonstrated that OVOL2
   directly suppresses STAT3 phosphorylation by directly inhibiting JAK
   transcription, recruits HDAC1 to STAT3, and subsequently reduces
   expressions of CPT1A and CPT1B. Consistently, OVOL2 loss promotes FAO
   and enhances the production of ATP and NADPH that support the stem
   cell‐like characteristics of TNBC. In a mouse model, PyVT‐Ovol2
   knockout mice developed a higher number of primary breast tumors with
   increased stemness properties which revealed accelerated growth and
   elevated lung‐metastases. Our findings uncover the regulatory mechanism
   by which OVOL2 represses FAO in breast cancer and propose that
   targeting JAK/STAT3‐FAO is a potential strategy for OVOL2‐absent TNBC
   therapy.

2. Result

2.1. OVOL2 Loss in TNBC Promotes Sphere Formation and Tumor Initiation

   We previously found that mRNA levels of OVOL2 decreased at a late stage
   of human breast cancer progression.^[ [68]^20 ^] To further elucidate
   the impact of OVOL2 on breast cancer carcinogenicity, we analyzed OVOL2
   expression across different molecular subtypes using data from UALCAN
   and Oncomine databases. OVOL2 was significantly downregulated in TNBC
   compared to non‐TNBC subtypes (Figure [69]1A; Figure [70]S1A,
   Supporting Information). We also analyzed mRNA levels of OVOL2 in
   breast cancer cell lines by quantitative reverse transcription‐PCR
   (RT‐qPCR). The results showed that OVOL2 mRNA levels were minimal in
   TNBC cell lines compared to those of non‐TNBC cell lines
   (Figure [71]1B). These results exhibit that TNBC has low expression of
   OVOL2.

Figure 1.

   Figure 1
   [72]Open in a new tab

   OVOL2 loss in TNBC promotes sphere formation and tumor initiation. A)
   Boxplots illustrating the expression levels of OVOL2 obtained from the
   UALCAN database. B) The expression of OVOL2 in 12 human breast cancer
   cell lines was determined by RT–qPCR. C) Cell growth was evaluated
   using a CCK8 assay. Data were analyzed 48 h after transfection. D) The
   ability of OVOL2‐overexpressing cancer cells to induce mammosphere
   formation was determined by a mammosphere formation assay. The graphs
   show the sphere numbers compared to those in the control group. E) In
   vivo tumor formation in mice injected with serial dilutions of
   MDA‐MB‐231 cells. p‐value determined by Student's t‐test. The data are
   presented as means ± SD. The symbols *, **, *** and represent p < 0.05,
   p < 0.01 and p < 0.001, respectively.

   Considering TNBC has high stemness properties and OVOL2 is critical for
   cell differentiation, we wondered whether OVOL2 might affect the
   stemness characteristics of TNBC. First, we checked the proliferation
   of cells by CCK8 assay. We overexpressed OVOL2 in TNBC cell lines
   MDA‐MB‐231, SUM159PT, and BT549 which exhibit very low OVOL2
   expression, and found that OVOL2 could reduce cell proliferation
   (Figure [73]1C; Figure [74]S1B, Supporting Information). Second, we
   analyzed sphere‐formation which is commonly utilized to study the
   stemness characteristics of cancer cells.^[ [75]^21 , [76]^22 , [77]^23
   ^] MDA‐MB‐231 cells with OVOL2 overexpression generated smaller spheres
   and the rate of sphere formation decreased by 74.7% (Figure [78]1D).
   Similarly, overexpression of OVOL2 almost blocked sphere‐formation in
   SUM159PT and BT549 cells (Figure [79]1D). In addition, we assayed the
   in vivo tumor‐initiating capacity (TIC) of limiting dilutions of
   MDA‐MB‐231 cells with or without OVOL2 overexpression. With 10^5 or
   10^4 cells injection, 100% of animals developed tumors in control
   groups while only 33% of animals formed tumors in OVOL2 expressed
   groups. Following the injection of 10^3 cells, tumor formation occurred
   in 67% of cases in the control group and no tumors were observed in
   OVOL2‐expressed groups. (Figure [80]1E). Consistently, knocked down
   OVOL2 in OVOL2‐high luminal cell lines MCF7 and T47D raised cell
   proliferation and sphere formation (Figure [81]S1C–E, Supporting
   Information). Additionally, OVOL2 expression was found to be
   significantly downregulated in sphere‐formed cells compared to adherent
   cells (Figure [82]S1F, Supporting Information). Similar results were
   observed in the Gene Expression Omnibus (GEO) database (accession
   numbers [83]GSE182532)^[ [84]^24 ^] (Figure [85]S1G, Supporting
   Information). These results highlight the crucial role of OVOL2 in
   diminishing breast cancer stemness characteristics.

2.2. Absence of OVOL2 Enhances Fatty Oxidation and Sensitizes TNBC Cells to
Etomoxir

   To further explore the role of OVOL2 in TNBC cells, we performed RNA
   sequencing (RNA‐Seq) of MDA‐MB‐231 cells with or without OVOL2
   expression and analyzed the data. The 1716 genes exhibited
   downregulation in OVOL2‐expressing cells. Notably, these downregulated
   genes included a substantial enrichment of genes associated with fatty
   acid degradation and metabolism. (Figure [86]2A). We next performed
   RT‐qPCR analysis and observed a significant reduction in the relative
   mRNA levels of genes associated with fatty acid oxidation with
   overexpression of OVOL2 (Figure [87]2B). To evaluate cell FAO levels,
   Seahorse Bioscience's FAO assays, which utilize palmitate‐BSA as
   substrates, were performed to measure FAO‐specific oxygen consumption
   rate (FAO‐OCR) in control and OVOL2 overexpressed MDA‐MB‐231 cells.
   Etomoxir (ETO), an inhibitor of FAO, was used to block FAO‐OCR. As
   Figure [88]2C shows, OVOL2 overexpression significantly suppressed
   FAO‐OCR. Furthermore, we employed U‐^13C‐palmitate labeling to evaluate
   the contribution of fatty acids to TCA cycle metabolites. OVOL2
   overexpression led to a significant reduction of palmitate
   incorporation into TCA cycle in TNBC cells (Figure [89]2D; Figure
   [90]S2A, Supporting Information). The uptake of U‐^13C[16] palmitate
   showed no significant difference (Figure [91]S2B, Supporting
   Information). Considering OVOL2 is highly expressed in luminal cells,
   we knocked down it and found an increase of FAO‐OCR and utilization of
   palmitate in TCA cycle (Figure [92]S2C,D, Supporting Information).
   Furthermore, we compared the incorporation ratio of palmitate to TCA
   cycle in TNBC and luminal cell lines. The results showed that
   MDA‐MB‐231, SUM159PT, and BT549 TNBC cell lines which exhibited lower
   OVOL2 expression displayed a higher incorporation ratio compared to
   OVOL2‐high expression MCF7 and T47D cells (Figure [93]2E). To gain
   insight into the significance of FAO in stem cell characteristics of
   breast cancer, we proceeded to examine sphere formation in both TNBC
   cell lines and luminal cell lines, with or without the administration
   of the FAO inhibitor ETO. The results showed that ETO significantly
   inhibited sphere formation in TNBC cell lines but had no impact in
   luminal cell lines (Figure [94]2F,G). This suggests that TNBC cell
   lines with the absence of OVOL2 exhibit a greater dependence on FAO for
   the promotion of sphere formation, and in OVOL2‐knockdown luminal
   cells, ETO could exert a blocking effect on sphere formation and cell
   proliferation (Figure [95]S2E,F, Supporting Information). Taken
   together, these findings indicate that the absence of OVOL2 enhances
   FAO activity, thereby contributing to stem cell properties.

Figure 2.

   Figure 2
   [96]Open in a new tab

   Absence of OVOL2 enhances fatty oxidation and sensitizes TNBC cells to
   Etomoxir. A) The left panel shows pie charts depicting the genes
   identified through total RNA‐seq. Upregulated and downregulated genes
   with a fold change of ≥1.5 are highlighted. The right panel displays
   the KEGG pathway enrichment analysis of the downregulated genes. B)
   RT–qPCR array analysis of lipid metabolism‐related genes in MDA‐MB‐231
   cells. C) FAO was evaluated by measuring the FAO‐OCR using
   palmitate‐BSA as a substrate. The lower panel shows the ATP‐linked
   FAO‐OCR and maximal FAO‐OCR. D,E) U‐^13C[16] palmitate incorporation
   was evaluated by GCMS. F,G) Measurement of mammosphere formation
   following treatment with 50 µm ETO in different cell lines. The graphs
   show the sphere numbers compared to those in the control group. The
   TNBC subtype cells appeared to be more vulnerable to the FAO inhibitor
   drug. p‐value determined by Student's t‐test. The data are presented as
   means ± SD. The symbols *, **, *** and represent p < 0.05, p < 0.01 and
   p < 0.001, respectively.

2.3. OVOL2 Reduces Energy Production and Antioxidant Capabilities of Cells

   Incorporation of fatty acid to TCA and followed oxidative
   phosphorylation could generate the amount of ATP as well as NADPH which
   contributes to the antioxidant capabilities of cells. Consistently,
   higher levels of ATP and lower relative reactive oxygen species (ROS)
   levels were observed in TNBC compared to that of luminal cells (Figure
   [97]3A,B). Further, overexpression of OVOL2 in TNBC cells led to a
   reduction in APT levels, NADPH/NADP^+ ratio, and GSH/GSSG ratio
   (Figure [98]3C,D). Conversely, knocking down of OVOL2 exhibited
   contrary effects on these parameters in luminal cells and ETO treatment
   could eliminate these effects (Figure [99]S3A,B, Supporting
   Information).

Figure 3.

   Figure 3
   [100]Open in a new tab

   OVOL2 suppresses cellular energy production and antioxidant capacities.
   A,B) Measurement of cellular ATP levels and ROS levels in five breast
   cancer cell lines. TNBC subtype cells exhibit higher ATP levels and
   lower ROS Levels compared to Luminal subtype cells. C) The cellular ATP
   levels in cells were measured using an ATP assay kit. D) Overexpression
   of OVOL2 in MDA‐MB‐231 cells increased the ROS level and decreased
   NADPH/NADP+ and GSH/GSSG ratios. E) Supplementation with 5 mm acetate
   for 48 h compromised the OVOL2‐induced inhibition of ATP levels. F)
   Supplementation with 5 mm acetate for 48 h compromised the
   OVOL2‐induced elevation of ROS levels and reductions in the NADPH/NADP+
   and GSH/GSSG ratios. G) Cell proliferation was assessed using a CCK8
   assay. H) The inhibition of mammosphere formation induced by OVOL2 was
   compromised upon supplementation with acetate. I) Supplementation with
   acetate impeded the decline in tumorigenicity induced by OVOL2. J) The
   OVOL2‐induced intracellular accumulation of ROS was effectively
   quenched in the presence of NAC (2 mm). K) Evaluation of the
   proliferative capacity of the indicated cells via a CCK8 assay. L) NAC
   significantly blocked OVOL2‐induced mammosphere formation properties in
   breast cancer cells. p‐value determined by a two‐way ANOVA test. The
   data are presented as means ± SD. The symbols *, **, *** and represent
   p < 0.05, p < 0.01 and p < 0.001, respectively.

   We next used FAO product analog acetate, the precursor of acetyl‐CoA,
   to treat TNBC cells with OVOL2 overexpression. The results showed that
   acetate could restore ATP generation and ROS levels (Figure [101]3E,F).
   Consequently, it can enhance cell proliferation, sphere formation, and
   tumor initiation in OVOL2‐overexpressed TNBC cells (Figure [102]3G–I).
   Moreover, to investigate the reduction of ROS levels could restore the
   diminished stem cell properties caused by OVOL2, we employed
   N‐acetylcysteine (NAC) as a scavenger of ROS scavenge. NAC treatment
   blocked ROS increase thereby facilitating cell proliferation and
   mammosphere‐forming (Figure [103]3J–L). These results indicated that
   OVOL2 abrogates FAO‐mediated energy and NADPH supply which support stem
   cell characteristics.

2.4. OVOL2 Inhibits Fatty Acid Oxidation and Stemness Characteristics by
Downregulating STAT3 Signaling Pathway

   We next wondered how OVOL2 inhibits fatty acid oxidation. As CPT1A and
   CPT1B are the rate‐limiting enzymes of fatty acid oxidation and play a
   critical role in maintaining the stemness characteristics of breast
   cancer,^[ [104]^14 , [105]^25 ^] we first checked their levels. Both
   protein and mRNA levels of CPT1A and CPT1B were significantly decreased
   in OVOL2‐overexpressing TNBC cells and elevated in OVOL2 knockdown
   luminal cells. (Figure [106]4A–D). In addition, similar to
   OVOL2‐overexpression, knockdown of CPT1A inhibited tumor initiation in
   MDA‐MB‐231 xenografts in nude mice (Figure [107]S4A, Supporting
   Information). These indicate that loss of OVOL2 promotes fatty acid
   oxidation that supports breast cancer stemness properties might though
   upregulating CPT1 expression.

Figure 4.

   Figure 4
   [108]Open in a new tab

   OVOL2 inhibits fatty acid oxidation and stemness characteristics by
   downregulating STAT3 signaling pathway. A) Western blot analysis of
   CPT1A and CPT1B in cell lines stably transduced with control or
   OVOL2‐encoding vectors. B) Western blot analysis of CPT1A and CPT1B
   expression in MCF7 and T47D cells stably transduced with control or
   OVOL2 shRNA. C,D) The RNA expression levels of CPT1A and CPT1B in
   OVOL2‐overexpressing C) and OVOL2‐knockdown D) clones were determined
   by RT–qPCR. E,F) Knockdown of OVOL2 was followed by the knockdown of
   STAT3 in MCF7 cells E). Similarly, the overexpression of OVOL2 was
   carried out, followed by the overexpression of STAT3 in MDA‐MB‐231
   cells F). The expression levels of CPT1A and CPT1B were determined
   through Western blot experiments. G) U‐^13C[16] palmitate incorporation
   assay in MCF7 cells showed that STAT3 knockdown can counteract the
   effect of OVOL2 knockdown. H) STAT3 knockdown attenuated the effect of
   OVOL2 knockdown on FAO‐OCR abilities in MCF7 cells. I) Western blot
   analysis of OVOL2, pSTAT3 (Tyr705), and pSTAT3 (Ser705) in cells stably
   transduced with control or OVOL2‐encoding vectors. J) Expression of
   OVOL2 and phosphorylated STAT3 (pY705) in 12 human breast cancer cell
   lines determined by Western blotting. K) APRE‐luciferase reporter assay
   performed to detect the transcriptional activity of STAT3. L) STAT3
   compromised the changes in mammosphere formation induced by OVOL2. The
   graphs show the sphere numbers compared to those in the control group.
   p‐value determined by a two‐way ANOVA test. The data are presented as
   means ± SD. The symbols *, **, *** and represent p < 0.05, p < 0.01 and
   p < 0.001, respectively.

   Given the transcriptional repressor function of OVOL2, we hypothesized
   that it might directly modulate the transcription of CPT1. However,
   chromatin immunoprecipitation (ChIP) qPCR results showed OVOL2 did not
   bind to predicted binding sites of CPT1A and CPT1B promoters (Figure
   [109]S4B, Supporting Information). It prompts us that OVOL2 might
   downregulate CPT1 expression via other transcription factors.

   JAK/STAT3 signaling has been reported to activate FAO through the
   transcription of CPT1B in breast cancers.^[ [110]^26 ^] We observed
   that knockdown of STAT3 in MDA‐MB‐231 cells indeed reduced relative
   mRNA levels and protein levels of CPT1A and CPT1B (Figure [111]S4C,D,
   Supporting Information). ChIP‐qPCR assays showed that STAT3 could
   directly bind to the promoter regions of CPT1A and CPT1B (Figure
   [112]S4E, Supporting Information). To address whether OVOL2 inhibits
   CPT1 expression via STAT3, we double knocked down STAT3 and OVOL2 in
   MCF7 and found that loss of STAT3 suppressed shOVOL2‐induced
   enhancement of CPT1 expressions (Figure [113]4E). Overexpression of
   STAT3 could rescue the CPT1 expression in MDA‐MB‐231 cells with OVOL2
   overexpression (Figure [114]4F). Moreover, the knockdown of STAT3
   attenuated the increase in palmitate incorporation into TCA cycle and
   FAO‐mediated O[2] consumption in OVOL2‐knockdown MCF7 cells
   (Figure [115]4G,H). These data confirm that OVOL2 suppresses CPT1
   expression and FAO via STAT3.

   As phosphorylation of STAT3 at tyrosine (Tyr) 705 is crucial for STAT3
   transcription activity, we asked if OVOL2 affects its phosphorylation
   and transcriptional activity. We found OVOL2 suppressed phosphorylation
   of STAT3 in TNBC cells (Figure [116]4I). Conversely, the knockdown of
   OVOL2 in MCF7 cells resulted in increased STAT3 phosphorylation (Figure
   [117]S4F, Supporting Information). Further, high levels of STAT3 T705
   phosphorylation were observed in TNBC cells, while luminal cells
   exhibit comparatively lower levels. The results showed an inverse
   correlation between STAT3 phosphorylation and OVOL2 expression
   (Figure [118]4J). Additionally, in a STAT3‐dependent APRE‐luciferase
   reporter assay, OVOL2 dose‐dependently repressed IL‐6‐mediated STAT3
   transcriptional activity (Figure [119]4K; Figure [120]S4G,H, Supporting
   Information).

   Moreover, the mammosphere‐forming ability of MDA‐MB‐231 cells was
   partially rescued by STAT3 in OVOL2 overexpression cells while
   knockdown of STAT3 blocked sphere formation induced by shOVOL2 in MCF7
   cells. (Figure [121]4L). Importantly, neither activation nor silencing
   of STAT3 altered mRNA levels of OVOL2 (Figure [122]S4I,J, Supporting
   Information). In summary, these findings indicate that OVOL2 inhibits
   fatty acid oxidation and stemness characteristics by suppressing STAT3
   phosphorylation and activation.

2.5. OVOL2 Suppresses the STAT3 Signaling Pathway Through Multiple Mechanisms

   JAK proteins phosphorylate STAT3 at Tyr705, initiating its
   transcriptional activity. We observed that OVOL2 overexpression reduced
   the mRNA and protein levels of three JAK family members JAK1, JAK2, and
   TYK2 (Figure [123]5A,B; Figure [124]S5A,B, Supporting Information).
   Moreover, ChIP qPCR results showed that OVOL 2 directly binds to their
   promoters (Figure [125]5C) and found that JAK1 was the most significant
   contributor to STAT3 activation induced by OVOL2 knockdown (Figure 5D;
   Figure [126]S5C, Supporting Information). Consistently, JAK1
   effectively restored STAT3 phosphorylation in OVOL2‐overexpressing
   cells (Figure [127]5E,F). These results indicate that OVOL2
   downregulates STAT3 phosphorylation by suppressing transcription of
   JAK1 (Figure [128]5G).

Figure 5.

   Figure 5
   [129]Open in a new tab

   OVOL2 suppresses the JAK/STAT3 signaling pathway through multiple
   mechanisms. A,B) Western blotting and qRT‐PCR were used to measure the
   expression levels of JAK1, JAK2, and TYK2 upon OVOL2 overexpression in
   MDA‐MB‐231 cells. C) ChIP assay demonstrating the binding of OVOL2 to
   the promoters of JAK1, JAK2, and TYK2. The values were normalized to
   input DNA. D) Western blot analysis of phosphorylated STAT3 (pY705) and
   total STAT3 in OVOL2‐knockdown MCF7 cells with stable knockdown of
   JAK1, JAK2, or TYK2. E,F) Western blot showing phosphorylated STAT3
   (pY705) and total STAT3 levels in OVOL2‐overexpressing cells after JAK1
   overexpression. G) Schematic representation of the regulatory role of
   OVOL2 in inhibiting STAT3 phosphorylation through JAK1 suppression. H)
   In vivo interaction between endogenous OVOL2 and STAT3. Lysates of MCF7
   cells were subjected to immunoprecipitation using an anti‐STAT3
   antibody. I,J) Direct binding of STAT3 and OVOL2 at both the C‐terminus
   (aa 241–275) and N‐terminus (aa 1–100) of OVOL2 I), as well as at the
   linker domain (S3, aa 495–585), SH2 domain (S4, aa 586–688), and
   transactivation domain (S5, aa 689–770) of STAT3 J). Red arrows
   indicate target bands. K) Co‐IP experiments of OVOL2 and STAT3‐WT,
   STAT3C, and STAT3D reveal disrupted interaction between OVOL2 and
   unphosphorylated STAT3. Whole‐cell lysates were used for IP with an
   anti‐Flag antibody. L) Endogenous co‐IP experiments between HDAC1 and
   STAT3 in MDA‐MB‐231 cells with or without OVOL2 overexpression. M) The
   inhibitory effect of OVOL2 on expression levels of CPT1A and CPT1B was
   reversed by treatment with 2 µm tacedinaline, an HDAC1 inhibitor. (N)
   Schematic representation of how OVOL2 facilitates the recruitment of
   HDAC1 to STAT3 at CPT1A and CPT1B promoter. p‐value determined by
   Student's t‐test. The data are presented as means ± SD. The symbols *,
   **, *** and represent p < 0.05, p < 0.01 and p < 0.001, respectively.

   We further explored whether OVOL2 interacts with STAT3.
   Co‐immunoprecipitation (Co‐IP) confirmed the physical interaction
   between OVOL2 and STAT3 (Figure [130]5H; Figure [131]S5D, Supporting
   Information). Moreover, an in vitro pulldown assay demonstrated the
   direct interaction between OVOL2 and STAT3 (Figure [132]S5E, Supporting
   Information). Deletion mapping showed both the C‐terminus (aa 241–275)
   and N‐terminus (aa 1–100) of OVOL2 were crucial for direct binding to
   STAT3 (Figure [133]5I). The linker domain (S3, aa 495–585), SH2 domain
   (S4, aa 586–688), and a transactivation domain (S5, aa 689–770) of
   STAT3 could bind to OVOL2 (Figure [134]5J). We next investigated
   whether activation of STAT3 affects its binding to OVOL2, we used
   constitutively active STAT3 (STAT3C, A662C N664C) and dominant negative
   STAT3 (STAT3D, Y705F) plasmids^[ [135]^27 ^] (Figure [136]S5F,
   Supporting Information). Unlike STAT3 WT and STAT3C, STAT3D, which
   lacks activity, failed to exhibit any interactions with OVOL2
   (Figure [137]5K).

   We subsequently investigated how the binding of OVOL2 regulates the
   transcriptional activity of STAT3. Given that OVOL2 has been shown to
   recruit histone deacetylases HDAC1,^[ [138]^19 ^] we hypothesized that
   the binding of OVOL2 may facilitate the recruitment of HDAC1 to STAT3,
   thereby suppressing its transcriptional activity. First, we performed
   Co‐IP assay in MDA‐MB‐231 cells and the results indicated that STAT3
   pulled down more HDAC1 in the presence of OVOL2 (Figure [139]5L). OVOL2
   facilitates the interaction between HDAC1 and STAT3 when STAT3 is
   phosphorylated and activated (Figure [140]S5G, Supporting Information).
   To assess the impact of OVOL2 on the association between HDAC1 and
   STAT3 target chromatin at the promoters of CPT1A or CPT1B, a DNA
   affinity precipitation assay (DAPA) was performed. The results showed
   that overexpression of OVOL2 enhanced the association between HDAC1 and
   STAT3 binding elements at the CPT1A or CPT1B promoters (Figure
   [141]S5H, Supporting Information). Further, ChIP Re‐IP assays revealed
   that OVOL2 overexpression enhanced the association between HDAC1 and
   the promoters of CPT1A or CPT1B that were immunoprecipitated by
   anti‐STAT3 antibodies in MDA‐MB231 cells (Figure [142]S5I,J, Supporting
   Information). Additionally, treatment with tacedinaline, an inhibitor
   of HDAC1, reversed the inhibitory effect of OVOL2 on expression levels
   of CPT1A and CPT1B (Figure [143]5M). Taken together, these results
   indicate that OVOL2 could bind to phosphorylated STAT3, facilitate the
   recruitment of HDAC1, and ultimately disrupt the transcriptional
   activation of STAT3 downstream genes like CPT1A and CPT1B
   (Figure [144]5N).

2.6. Loss of OVOL2 Promotes Breast Tumor Initiation via JAK/STAT3/CPT1
Regulated FAO in MMTV‐PyVT Mice

   To further explore the physiological role of OVOL2 in breast cancer, we
   generated MMTV‐PyVT, MMTV‐Cre Ovol2 ^f/f mice (PyVT‐Ovol2^KO) (Figure
   [145]6A; Figure [146]S6A, Supporting Information). We found that OVOL2
   knockout alone did not induce tumor development. However, in
   PyVT‐Ovol2^KO mice, mammary gland tumors appeared earlier, indicating
   accelerated tumorigenesis onset (Figure [147]6B). We also verified that
   JAK/STAT3/CPT1A pathways were upregulated in Ovol2 knockout mouse
   tumors. Western blot analysis confirmed reduced STAT3 phosphorylation
   at Tyr705 in PyVT‐Ovol2^KO tumors compared to PyVT‐WT tumors (Figure
   [148]S6B, Supporting Information). Furthermore, IHC staining revealed
   increased STAT3 phosphorylation and elevated expression of CPT1A, JAK1,
   JAK2, and TYK2 in Ovol2^KO tumors compared to control tumors
   (Figure [149]6C,D), suggesting that OVOL2 knockdown activates the
   JAK/STAT3/CPT1A pathways. We also checked FAO levels in primary breast
   tumor cells. The 21 days after the tumors became palpable, we
   euthanized the mice and isolated tumor cells to examine their metabolic
   characteristics. Ovol2^KO tumor cells showed elevated FAO‐OCRs and ATP
   levels (Figure [150]6E,F), indicating significantly enhanced FAO in the
   absence of OVOL2. Mammo sphere‐forming abilities were also remarkably
   stronger in Ovol2^KO tumor cells, which could be suppressed by ETO
   treatment (Figure [151]6G). Moreover, to further evaluate
   tumor‐initiating capacity, Ovol2^WT, and Ovol2^KO tumor cells were
   isolated and transplanted into the mammary fat pads of recipient mice
   twice. We found that Ovol2^KO cells showed robust tumor formation,
   while Ovol2^WT cells exhibited diminished tumor‐initiating capacity
   (Figure [152]6H,I). Secondary transplantation confirmed the elevated
   tumor‐initiating capacity of Ovol2^KO cells (Figure [153]6J). These
   findings demonstrate that loss of OVOL2 facilitates mammary gland tumor
   initiation via JAK/STAT3/CPT1 pathway‐regulated FAO in the murine
   model.

Figure 6.

   Figure 6
   [154]Open in a new tab

   Loss of OVOL2 promotes breast tumor initiation via JAK/STAT3/CPT1
   regulated FAO in MMTV‐PyVT mice. A) Breeding scheme illustration. B)
   Kaplan‐Meier plots showing mammary tumor‐free survival in female
   PyVT‐OVOL2^WT (n = 41) and PyVT‐OVOL2^KO (n = 54) mice. C,D) IHC
   staining of pSTAT3 and CPT1A C), as well as JAK1, JAK2, and TYK2 D) in
   mammary tumors from female PyVT and PyVT‐OVOL2^KO mice. The
   immunoreactive scores (IRSs) were calculated to analyze the IHC images.
   Representative images are shown, and the scale bars indicate 50 µm. E)
   Increased FAO‐OCR in PyVT‐OVOL2^KO mouse tumor cells (n = 5–6)
   indicating loss of OVOL2 function. FAO‐OCR measurements were conducted
   in primary cultured tumor cells isolated from PyVT mammary tumors after
   the second passage. F) Elevated ATP levels in PyVT‐OVOL2^KO mouse tumor
   cells (n = 9) due to loss of OVOL2. G) Mammosphere formation assay
   performed on primary cells isolated from PyVT mammary tumors after the
   second passage and treated with ETO. H) Schematic representation of the
   procedures for the first and second orthotopic injections. I,J)
   Enhanced tumor‐initiating ability was observed in PyVT‐OVOL2^KO mouse
   tumor cells during the first and second orthotopic injections
   (n = 10–12 mice per group) compared to normal tumor cells. The data are
   presented as means ± SEM. Statistical significance is represented as **
   (p < 0.01), *** (p < 0.001), **** (p < 0.0001), ns (not significant),
   determined by the Gehan‐Breslow‐Wilcoxon test.

   Besides, we observed that primary PyVT‐Ovol2^KO tumors exhibited faster
   growth in comparison to PyVT‐Ovol2^WT tumors (Figure [155]S6C,
   Supporting Information). Moreover, Ki67 staining indicated an increase
   in the proliferation of PyVT‐Ovol2^KO tumors (Figure [156]S6D,
   Supporting Information). Additionally, OVOL2 deletion significantly
   reduced the median survival time and life expectancy of tumor‐bearing
   PyVT‐Ovol2^KO mice (Figure [157]S6E, Supporting Information). The
   absence of OVOL2 resulted in a substantial increase in the number and
   size of metastatic nodules in the lungs (Figure [158]S6F,G, Supporting
   Information). FACS analysis further revealed an abundance of EPCAM‐low
   cells in the PyVT‐Ovol2^KO tumor population, which is rare in
   PyVT‐Ovol2^WT tumors (Figure [159]S6H,I, Supporting Information). The
   body weights of the mice indicated no significant differences between
   the experimental groups (Figure [160]S6J, Supporting Information).
   Overall, these findings establish the suppressive role of OVOL2 in
   breast cancer progression, and its loss leads to an increase in the
   malignancy of breast tumors.

2.7. Targeting FAO or JAK to Inhibit Breast Tumor Progress in OVOL2 Knockout
Mice

   Because ETO treatment could reduce the sphere formation of tumor cells
   in the above experiments, we deduced that inhibition of FAO might be an
   ideal strategy for the prevention of tumor initiation. Perhexiline, a
   CPT1 inhibitor used clinically for antianginal treatment,^[ [161]^28 ^]
   was administrated to PyVT‐Ovol2^WT and PyVT‐Ovol2^KO mice after the
   tumors were touchable (Figure [162]7A). After continuous treatment for
   21 days, some mice were kept to monitor survival and the others were
   sacrificed to isolate tumor cells which were subjected to
   mammosphere‐forming assays as well as orthotopic transplantation
   (Figure [163]7A). Overall survival was significantly improved by
   perhexiline treatment in OVOL2‐deficient mice compared to WT mice
   (Figure [164]7B). Moreover, the sphere formation of Ovol2^KO tumor
   cells was reduced by FAO inhibition in vivo with perhexiline treatment
   via gavage (Figure [165]7C). Tumor cells lacking OVOL2 exhibited higher
   tumorigenic capability, which was effectively inhibited by perhexiline
   treatment (Figure [166]7D). Our results demonstrated that OVOL2
   deletion promotes tumor initiation in MMTV‐PyVT mice via the regulation
   of FAO, thus implicating the potential clinical value of FAO inhibitors
   for breast cancer treatment.

Figure 7.

   Figure 7
   [167]Open in a new tab

   Targeting JAK or FAO to inhibit breast tumor progress in OVOL2 knockout
   mice. A) Experimental schematic for PyVT model mice treated with or
   without perhexiline (PE, 1 mg kg^−1dose by oral gavage three times
   weekly for three weeks) and subsequent applications in different
   experiments. B) Survival curves of PyVT and PyVT‐OVOL2^KO mice treated
   with or without PE as described. C) A mammosphere formation assay
   performed on primary cells isolated from PyVT or PyVT‐OVOL2^KO mice
   treated with PBS or PE (n = 5–6). D,E) Injection of 1 × 10^5 PyVT or
   PyVT‐OVOL2^KO tumor cells from PyVT model mice treated with or without
   perhexiline D) or upadacitinib E) into the mammary fat pads of FVB
   mice. After 21 days, mice were sacrificed, and tumors were collected.
   F) Schematic illustrating the effect of OVOL2 on breast cancer cells.
   The data are presented as means ± SEMs. Statistical significance is
   represented as ** (p < 0.01), *** (p < 0.001), **** (p < 0.0001), ns
   (not significant), determined by the Gehan–Breslow–Wilcoxon test.

   Furthermore, we also observed the effects of perhexiline on the
   development of primary mouse breast tumors. PyVT‐Ovol2^KO tumors with
   Perhexiline treatment exhibited significantly decreased tumor growth
   compared to the PBS‐treated groups (Figure [168]S7A,B, Supporting
   Information). Imaging and quantification of metastases showed a
   remarkable reduction in lung metastatic spread in PyVT‐Ovol2^KO mice
   treated with perhexiline (Figure [169]S7C–E, Supporting Information).
   IHC staining for the proliferation marker Ki67, the epithelial marker
   ECAD, and the EMT marker Vimentin revealed that perhexiline inhibited
   the increased cell proliferation and transformation observed in
   OVOL2‐deficient tumors compared to control tumors (Figure [170]S7F,
   Supporting Information).

   Similarly, in vivo, we aimed to validate the JAK/STAT pathway as a
   critical downstream target of OVOL2. The administration of upadacitinib
   (Upa), a JAK1 inhibitor, substantially attenuated the tumorigenic
   promotion induced by OVOL2 deficiency (Figure [171]7E).

   These findings clearly indicate the requirement of FAO and JAK/STAT3
   pathway for breast tumor progression in mice with OVOL2 deficiency.
   Moreover, these results provide compelling evidence that loss of OVOL2
   activates JAK/STAT3 pathway and enhances FAO which makes tumors more
   susceptible to perhexiline or upadacitinib treatment (Figure [172]7F).
   This suggests that targeting FAO or JAK/STAT pathway represents a
   promising therapeutic strategy for OVOL2‐deficient TNBC.

3. Discussion

   Triple‐negative breast cancer is the most aggressive subtype of breast
   cancer. The lack of ER, PR, and HER2 receptors makes it insensitive to
   hormone‐based or receptor‐targeted therapies. Thus, there is a pressing
   need to discover new characteristics of TNBC and devise novel and
   effective therapeutic approaches to combat this challenging disease. In
   this study, we found that OVOL2 expression was significantly
   downregulated in TNBC compared to other subtypes. Overexpression of
   OVOL2 in TNBC cell lines reduced breast cancer stemness characteristics
   including cell proliferation and sphere formation both in vitro and in
   vivo. Understanding how OVOL2 inhibits the development of TNBC
   progressing may provide crucial clues for the development of new
   therapeutic treatments.

   Metabolic remolding is a hallmark of cancer. Traditionally, it has been
   believed that tumor cells predominantly rely on anaerobic glycolysis, a
   phenomenon known as the “Warburg effect,” even in the presence of
   oxygen. Hypoxia could induce downregulation of OVOL2 thereby increasing
   glycolysis.^[ [173]^29 ^] However, it is important to note that cancer
   cells exhibit metabolic adaptability, which depends on the availability
   of nutrients and oxygen.^[ [174]^30 ^] This metabolic flexibility
   enables cells in diverse nutritional and oxygen environments to
   establish metabolic coupling, thereby promoting cancer cell
   proliferation and tumor growth. Substantial evidence demonstrates the
   significance of fatty acid oxidation in breast tumor progression.^[
   [175]^31 , [176]^32 , [177]^33 ^] Compared to glucose, fatty acid could
   generate more energy and antioxidants through fatty acid oxidation.^[
   [178]^5 , [179]^34 , [180]^35 , [181]^36 ^] Here, by RNA‐Seq analysis,
   we discovered that OVOL2 negatively regulated fatty acid degradation in
   TNBC cells. OVOL2 loss in TNBC resulted in elevated fatty acid
   oxidation. OVOL2 overexpression significantly inhibited
   palmitate‐mediated O[2] consumption and the incorporation of
   ^13C‐palmitate into TCA cycle thereby reducing ATP levels and
   upregulating ROS levels. The supplementation of acetate as a direct
   product of fatty acid oxidation restored the effect elicited by OVOL2
   expression. These indicate OVOL2 functions as a negative regulator in
   FAO‐mediated metabolic remodeling and antioxidant defense.

   CSCs possess the ability of self‐renewal, and tumor initiation, and are
   associated with tumor growth, metastasis, drug resistance, and
   recurrence.^[ [182]^37 ^] Recent studies have highlighted the crucial
   role of FAO in cancer stem cells.^[ [183]^38 , [184]^39 ^] Inhibiting
   FAO has been found to increase the sensitivity of breast cancer CSCs to
   chemotherapy drugs.^[ [185]^6 ^] For instance, in colorectal cancer,
   FAO‐generated acetyl‐CoA is transferred by P300 to promote H3K27
   acetylation of the Nanog promoter, thereby enhancing Nanog expression
   and inducing cellular dormancy akin to quiescent cancer stem cells.^[
   [186]^40 ^] FAO utilization also circumvents the need for amino acid
   metabolism, consequently conferring resistance to drugs like venetoclax
   or azacitidine in acute myeloid leukemia stem cells.^[ [187]^41 ^] To
   evaluate the stemness features of CSCs, we employed the sphere‐forming
   assay and in vivo tumor transplantation assay as indicators. The
   sphere‐forming assay creates a relatively deprived environment for
   cells and assesses their stemness based on their resistance to anoikis,
   a form of programmed cell death induced by detachment from the
   extracellular matrix.^[ [188]^42 ^] On the other hand, the in vivo
   tumor transplantation assay directly evaluates the tumor‐initiating
   capability of cancer cells, serving as a measure of their stemness.^[
   [189]^43 ^] Through these assessments, we made an intriguing
   observation that OVOL2 not only possesses the ability to inhibit sphere
   formation and tumor‐initiating capacity in cells cultured in vitro but
   also exerts a significant impact on the stemness characteristics of
   breast tumor cells in the PyVT mouse model. Specifically, the secondary
   transplantation experiment revealed that the loss of OVOL2 in mouse
   breast tumor cells resulted in enhanced and more stable
   tumor‐initiating capacity, further highlighting the crucial role of
   OVOL2 in regulating CSC properties in vivo. Additionally, we observed
   that impaired FAO contributed to the reduction in stemness
   characteristics in OVOL2‐overexpressing cells and acetate could restore
   the sphere‐forming abilities. In a mouse model, OVOL2 deletion resulted
   in enhanced FAO capability, increased tumor‐initiating capacity, and
   accelerated tumor growth.

   Notably, the absence of OVOL2 makes TNBC and primary breast tumors
   become more reliant on fatty acid oxidation to sustain rapid growth and
   tumor progression.

   FAO inhibitor ETO could effectively block sphere formation in OVOL2
   absent TNBC cells but not in OVOL2‐high‐expressed luminal cells.
   Besides, perhexiline, a CPT1 inhibitor used clinically for antianginal
   treatment,^[ [190]^28 ^] significantly inhibited tumor initiation and
   promoted mice survival in OVOL2 KO mice not in OVOL2 WT mice. These
   findings emphasize the promising role of OVOL2 as a predictive marker
   for sensitivity to FAO inhibitors in breast cancer. This discovery lays
   the foundation for the future precision management of breast cancer.

   We discovered that OVOL2 downregulates FAO via JAK1‐STAT3 pathway.
   STAT3 has emerged as a key regulator of lipid metabolism. In breast
   cancer, prior studies have demonstrated that STAT3 directly binds to
   the promoter of CPT1B, promoting its transcriptional regulation,
   resulting in an enhanced FAO.^[ [191]^14 ^] Consistently, we have
   demonstrated that STAT3 binds to the promoters of both CPT1A and CPT1B,
   and depletion of STAT3 leads to a decrease in the expression levels of
   CPT1A and CPT1B (Figure [192]S4C–E, Supporting Information).
   Furthermore, multiple lines of evidence in our study have demonstrated
   that the overexpression of OVOL2 impedes the expression of CPT1A and
   CPT1B, involved in FAO, by suppressing the signaling of STAT3. The
   phosphorylation status of STAT3 elevated in TNBC compared to that in
   other non‐TNBC subtypes. We discovered that OVOL2 exerts inhibitory
   effects on STAT3 by directly repressing the expression of JAK1, thereby
   restraining the phosphorylation of STAT3 at the Tyr705. Additionally,
   OVOL2 binds to phosphorylated STAT3, enhancing the recruitment of HDAC1
   which is crucial for repressing the transcriptional activity of
   STAT3.^[ [193]^44 ^] Our study reveals that the STAT3/OVOL2/HDAC1
   protein complex binds to the STAT3 binding site at the CPT1A and CPT1B
   promoters, leading to the inhibition of CPT1A and CPT1B expression.

   Moreover, OVOL2 may impact lipid catabolism through multiple
   mechanisms. The tumor necrosis factor (TNF) signaling pathway is
   closely associated with lipid metabolism.^[ [194]^45 ^] Previous
   studies have reported that TNF induces lipolysis in human adipocytes by
   down‐regulating the lipid droplet‐associated protein perilipin
   (PLIN).^[ [195]^46 ^] Additionally, TNF inhibits the accumulation of
   acetyl‐CoA carboxylase, the rate‐limiting enzyme in long‐chain fatty
   acid biosynthesis, in pre‐adipocyte cell lines, thus reducing its
   activity.^[ [196]^47 ^] Recent reports have also demonstrated that TNF
   suppresses fatty acid metabolism and oxidative phosphorylation while
   promoting the development of fatty liver.^[ [197]^48 ^] Notably, our
   RNA‐seq analysis revealed a significant enrichment of the TNF signaling
   pathway. Therefore, we cannot exclude the possibility that OVOL2 may
   affect lipid catabolism in breast cancer cells by modulating the
   activity of the TNF pathway.

   In summary, our study reveals a novel mechanism of OVOL2 as a tumor
   suppressor in breast cancer, inhibiting tumor stemness and metastasis
   through STAT3 signaling pathway and FAO inhibition. Modulating these
   molecular processes may offer new targeted therapeutic strategies for
   cancer.

4. Experimental Section

Reagents

   Carmine staining solution (Cat# G3930) was from Solarbio. B‐27 (Cat#
   17504044) was from Gibco. Agarose (Cat# 2276GR025) was from BioFroxx.
   CCK8 (Cat# k1018) was from APExBIO. Poly (deoxyinosine‐deoxycytosine)
   [poly(dI‐dC)] (Cat# 10108812001) was from Roche. CellROX Oxidative
   Stress Reagents (Cat# [198]C10444) and Pierce Glutathione Agarose (Cat#
   16101) were from Thermo Fisher. XF DMEM (Cat# 103575‐100) was were from
   Agilent. U‐^13C[16] palmitate (Cat# CLM‐409‐0.5) was from Cambridge
   Isotope Laboratories. Methoxyamine (Cat# M109434) was from Aladdin.
   Perhexiline (Cat# HY‐B1334A), Leptin (Cat# HY‐P7232), and IL‐6 (Cat#
   HY‐P7044) were from Med Chem Express. Polyethylenimine Linear MW 40000
   (Cat# MX2203) was from Maokangbio. Sodium acetate (Cat# S2889) was from
   Sigma and was dissolved in ddH2O for use in cell culture applications.
   Polybrene (Cat# H9268‐5G), MSTFA (Cat# M‐108‐5×1ML), MTBE (Cat# 34875),
   Carnitine (Cat# C0158) and palmitate (Cat# P9767) were from Sigma.
   Upadacitinib (Cat# S8162), Tacedinaline (Cat# S2818) and Y‐27632 2HCl
   (Cat# S1049) were from SELLECK. TRIzol (Cat# 15596018) was from Life
   Technologies. UltraSYBR Mixture (Cat# CW0957H) was from Cwbiotech.
   Streptavidin Magnetic Beads (Cat# P2151), RIPA buffer (Cat# P0013B),
   and DAPI (Cat# C1002) were from Beyotime. ECL detection system (Cat#
   RM00021) was from ABclonal.

Mice and Mammary Fat Pad Xenograft Model

   MMTV‐PyVT, MMTV‐Cre, OVOL2^f/+, and FVB mice were used to generate
   conditional OVOL2 knockout and control mice. Experimental mice
   (7–14‐week‐old) were treated with perhexiline (PE, 1 mg kg^−1 dose^−1)
   or PBS by oral gavage three times weekly for 3 weeks. FVB mice and nude
   mice were used to establish the mammary fat pad allograft model with
   tumor cells from MMTV‐PyVT and MMTV‐PyVT‐OVOL2^KO mice. Nude mice
   (5‐week‐old females) were purchased from Shanghai SLAC Laboratory
   Animal Center. Mouse tumor cells were dissociated and isolated from
   mammary tumors as described in the “Primary cell preparation method”
   section. Single‐cell suspensions (50 µL; diluted in PBS with 50%
   Matrigel on ice) were injected into the fat pads of 8‐week‐old FVB
   mice. The secondary orthotopically injected mice were generated by
   isolating cells from the remaining tumors and transplanting these cells
   into FVB recipient mice. The tumor size was measured with calipers, and
   the tumor volume was determined using the following equation: volume =
   length× width^2×0.5. All animals were maintained in the Laboratory
   Animal Center at Xiamen University (China) in accordance with
   institutional guidelines. All mouse protocols and experiments were
   approved by the Animal Ethics Committee of Xiamen University
   (acceptance no. XMULAC20190043).

Primary Cell Preparation Method

   Freshly harvested mammary tumors were minced into 1–2 mm pieces and
   were then placed in tumor digestion medium (RPMI‐1640 medium
   supplemented with 200 U mL collagenase Type IV, 100 U mL hyaluronidase,
   25 mm HEPES, 5% fetal bovine serum (FBS), and 1%
   penicillin–streptomycin (PS) to a 3× volume of each tissue sample) for
   up to 2 h at 37 °C in a shaker at 50 rpm. The digested tissues were
   then transferred into a 37 °C incubator overnight. The digestion status
   was assessed microscopically, and cell fractionation was initiated if
   only small tissue pieces were observed.

   The digested cell clumps were transferred into 50 mL sterile tubes and
   centrifuged at 40 × g for 1 min. The supernatants were transferred into
   new 50 mL sterile tubes and centrifuged at 200 × g for 4 min. The
   supernatants were discarded, and the pellets obtained were washed in
   cold PBS and suspended in Dulbecco's modified Eagle's medium (DMEM)
   containing 10% FBS and 10 µm Y‐27632 2HCl (a ROCK inhibitor, Selleck,
   S1049). The obtained cells were passaged and used for other
   experiments.

Cell Lines and Cell Culture/Transfection

   Cell lines, including MDA‐MB‐231, MDA‐MB‐453, HBL100, HS578T, SK‐BR‐3,
   MCF7, and HEK293T, were purchased from the American Type Culture
   Collection (ATCC; Rockville, MD) and maintained in DMEM (Biological
   Industries, cat# 01‐052‐1ACS) supplemented with 10% FBS (FBSCN500‐S,
   AusGeneX). Other cell lines, including BT549, BT474, ZR‐75‐30,
   SUM159PT, and T47D, were purchased from ATCC and cultured in RPMI‐1640
   medium (Biological Industries, cat# 01‐100‐1ACS) supplemented with 10%
   FBS. MDA‐MB‐436 cells were purchased from ATCC and maintained in L‐15
   medium (HyClone, with L‐glutamine, SH30525) supplemented with 10% FBS,
   10 µg mL^−1 insulin (Sigma–Aldrich) and GlutaMAX (Gibco, 100×,
   35050–061). All cells were cultured at 37 °C in a humidified incubator
   with 5% CO2. Transient transfection of cells was conducted using PEI
   (Polyethylenimine Linear MW 40000, cat# MX2203, Maokangbio) according
   to the manufacturer's protocol.

Lentiviral System Construction

   The pBOBi lentiviral (pLV) vector was used to overexpress target genes,
   and the lentiviral vector pLKO.1 was used to express shRNAs in human
   breast cancer cells. For lentivirus packaging, subconfluent HEK 293T
   cells cultured in 10 cm culture dishes were cotransfected with the
   lentiviral vector (8 µg), the packaging plasmid pHR (6 µg), and the
   envelope plasmid pVSV‐G (2 µg) using PEI. Eight hours after
   transfection, the transfection medium was replaced with fresh DMEM
   supplemented with 10% FBS. Viral supernatants were collected and
   filtered through a Millex‐HV 0.45 µm filter (Merck Millipore,
   SLHVR33RB). For lentiviral transduction, cells were transduced with
   lentiviral vectors in the presence of polybrene (Sigma, H9268‐5G) for
   24 h, and stable cell lines were selected with puromycin. The
   overexpression or knockdown efficiency of the target gene was validated
   by western blotting or qRT–PCR 48 h after lentiviral transduction.

   The sequences of the shRNAs targeting mRNAs are listed below (5’→3’).

   pLKO.1 human OVOL2 shRNA:

   GCTGGGATGAGCTCCCGGATGAGAA and ACATCCGCACACCAGGAGAAT

   pLKO.1 human STAT3 shRNA:

   GCACCTTCCTGCTAAGATTCA and GCACTTGTAATGGCGTCTTCA

   pLKO.1 human JAK1 shRNA:

   GCACAGAAGACGGAGGAAATG and GGTGGAAGTGATCTTCTATCT

   pLKO.1 human JAK2 shRNA:

   GCAACTTGGCAAGGGTAATTT and GCTTTGTCTTTCGTGTCATTA

   pLKO.1 human TYK2 shRNA:

   GAGATCCACCACTTTAAGAAT and CGAGCACATCATCAAGTACAA

   pLKO.1 human CPT1A shRNA:

   GCCATGAAGCTCTTAGACAAA and CGATGTTACGACAGGTGGTTT

Quantitative Reverse Transcription–Polymerase Chain Reaction (qRT–PCR)

   Total RNA from breast cancer cells was extracted using TRIzol (Life
   Technologies, 15596018) according to the manufacturer's instructions.
   cDNA was synthesized with oligo(dT) primers using 5× All‐In‐One RT
   MasterMix (abmGood, cat# G490) according to the manufacturer's
   instructions. RT–qPCR was performed using UltraSYBR Mixture (Cwbiotech,
   cat# CW0957H) according to the manufacturer's instructions with the
   RT–qPCR primers listed in Table [199]S1 (Supporting Information) in a
   CFX ConnectTM Real‐Time PCR Detection System (Bio‐Rad) with the
   following amplification conditions: 95 °C for 5 min, followed by 44
   cycles at 95 °C for 10 s and 60 °C for 1 min. 18S rRNA was used as the
   internal control. All samples were analyzed in biological and technical
   triplicate. The 2−ΔΔCT method was used to calculate fold changes in
   relative gene expression levels.

Western Blot Analysis

   Proteins from cultured cells were extracted with RIPA buffer (Beyotime,
   cat# P0013B), and protein concentrations were determined using a BCA
   Protein Assay Kit (Pierce, Thermo #23235) according to the
   manufacturer's instructions. Proteins were loaded onto 8–10% SDS–PAGE
   gels for electrophoresis and were electrotransferred to a 0.45 µm
   Immobilon‐P PVDF membrane (Millipore, MA, USA). The membrane was
   blocked with 5% bovine serum albumin (BSA) in TBST (150 mm NaCl, 20 mm
   Tris HCl, and 0.05% (v/v) Tween 20) for 1 h at room temperature and was
   then incubated with specific primary antibodies overnight at 4 °C.
   Next, the membrane was incubated with peroxidase‐conjugated secondary
   antibodies (Pierce goat anti‐rabbit IgG or Pierce goat anti‐mouse IgG)
   for 1 h at room temperature, and chemiluminescence signals were
   detected using an ECL detection system (ABclonal, cat# RM00021).
   Signals were visualized by exposure to film.

   Antibodies used for western blotting were as follows: anti‐STAT3 (Cat#
   9139, CST), anti‐phospho‐STAT3‐Tyr705 (Cat# 9145, CST),
   anti‐phospho‐STAT3‐Ser727 (Cat# AP0474, ABclonal), anti‐CPT1A (Cat#
   A5307, ABclonal), anti‐CPT1B (Cat# A6796, ABclonal), anti‐JAK1 (Cat#
   66466‐1‐Ig, Proteintech), anti‐JAK2 (Cat# 3230, CST), anti‐TYK2 (Cat#
   A2128, ABclonal), anti‐HDAC1 (Cat# 5356, CST), anti‐HDAC1 (Cat#
   ab280198, Abcam), anti‐E‐cadherin (Cat# 3195, CST), anti‐Vimentin (Cat#
   5741, CST), anti‐GST (Cat# ab9085, Abcam), anti‐OVOL2 (Cat# NBP2‐42907,
   Novus), anti‐OVOL2 (Cat# sc‐515001, Santa Cruz), anti‐Flag‐tag (Cat#
   14793S, CST), anti‐HA‐tag (Cat# H6908, Sigma), anti‐Myc‐tag (Cat#
   C3956, Sigma), anti‐His tag (Cat# AE003, ABclonal), anti Ki67 (Cat#
   GTX16667, GeneTex), anti‐β‐actin (Cat# A1978, Sigma).

Flow Cytometry

   To analyze the cellular composition of mammary tumors, digested tumor
   cells were prepared as described above (“Primary cell preparation
   method” section). In addition, the digested cell clumps were
   transferred into 50 mL sterile tubes and centrifuged at 450 × g for
   10 min. The pellets were resuspended in a Red Blood Cell Lysis Buffer
   (154 mm NH[4]Cl, 10 mm KHCO[3], and 0.1 mm EDTA; pH 8.0) with gentle
   pipetting and sequentially incubated with prewarmed 0.25% trypsin‐EDTA
   and 0.5 mg mL^−1 DNase I (NEB, #M0303S) for 5 min and centrifuged at
   450 × g for 10 min. The pellets obtained were washed in cold PBS and
   filtered through 100 µm cell strainers. Single‐cell suspensions were
   maintained in MACS buffer (PBS supplemented with 2% FBS, 1% PS, and
   1 mm EDTA) and used for flow cytometric staining with fluorescently
   labeled antibodies.

Immunohistochemical (IHC) and Hematoxylin and Eosin (H&E) Staining

   After overnight fixation with formalin at 4 °C and paraffin embedding,
   mammary tumor tissues were deparaffinized with 100% xylene and a
   descending ethanol series for further peroxidase IHC staining using an
   UltraSensitiveTM SP (Mouse/Rabbit) IHC Kit (MXB Biotechnologies,
   KIT‐9710) and a DAB Detection Kit (MXB Biotechnologies, DAB‐2031).
   Paraffin or frozen tissue sections were incubated with a primary
   antibody diluted in antibody diluent (0.1% Triton X‐100 and 2% BSA in
   PBS) overnight at 4 °C. After incubation with a secondary antibody,
   5‐µm‐thick sections were stained using DAB and costained with
   hematoxylin to visualize the staining of the protein of interest. Mouse
   tissue sections were stained with anti‐JAK1 (1:100), anti‐JAK2 (1:100),
   anti‐TYK2 (1:100), anti‐OVOL2 (1:100), anti‐phospho‐Stat3 (Tyr705)
   (1:100), anti‐CPT1A (1:200), anti‐ECAD (1:200), anti‐Vimentin (1:200)
   and anti‐Ki67 (1:200) primary antibodies. For H&E staining, tissues
   were fixed with 4% formaldehyde and embedded in paraffin. The
   5‐µm‐thick paraffin sections were dried at 65 °C for 2 h. The tissues
   were deparaffinized using 100% xylene and a descending ethanol series
   (100%, 95%, 80%, and 70%) and were then stained with H&E
   (Bioengineering Institute of Jiancheng, Nanjing, China, cat# D006)
   according to the manufacturer's instructions.

Oncomine Analysis and Gene Expression Omnibus (GEO) Database Analysis

   The UALCAN database (ualcan.path.uab.edu) and Oncomine database
   ([200]www.oncomine.org) were used to analyze differences in OVOL2 mRNA
   expression levels between TNBC and non‐TNBC samples based on the
   following thresholds: p value <0.0001, minimum two‐fold change in gene
   expression and gene rank in the top 10%.

RNA‐Seq Analysis

   Cells were collected, and total RNA was isolated using a TRIzol Kit
   following the manufacturer's instructions. Then, total RNA was treated
   with RNase‐free DNase I and enriched for poly(A) RNA using an NEBNext
   Poly(A) mRNA Magnetic Isolation Module (New England Biolabs). All RNA
   was then fragmented by adding a fragmentation buffer. Next, cDNA was
   synthesized, and PCR was performed using a KAPA Stranded RNA‐Seq
   Library Prep Kit (Illumina) to construct the final cDNA library.
   RNA‐seq libraries were subjected to QC analyses and were sequenced
   using an Agilent 2100 Bioanalyzer. A Perl program was written to select
   clean reads by removing low‐quality sequences and reads containing
   adaptor sequences. Raw sequencing data (fastq) were quality‐checked
   with FastQC, aligned to human genomes, and further used to calculate
   FPKM values using quartile normalization. Differentially expressed
   genes between the control and OVOL2 overexpression groups were
   identified using quantile normalization and the per‐condition
   dispersion method. Genes with FPKM>1 were retained for further analysis
   and were used to generate the pre‐ranked gene list. Gene set enrichment
   analysis (GSEA) was performed on the preranked gene list. The raw data
   have been submitted to the NCBI Sequence Read Archive (SRA) database,
   and the accession number is PRJNA848624.

Mammosphere Formation Assay

   Single‐cell suspensions of cell lines (10 000 cells mL^−1) were
   cultured in 24‐well ultra‐low attachment plates (Corning, 3473) in
   DMEM/F12 (Invitrogen, Waltham, MA, USA) containing B‐27 (1:50,
   Invitrogen), 20 ng mL^−1 epidermal growth factor (EGF; BD Biosciences,
   San Jose, CA, USA), 20 ng mL^−1 basic fibroblast growth factor (bFGF;
   BD Biosciences), 4 µg mL^−1 insulin (Sigma–Aldrich) and 1%
   methylcellulose. Mammosphere cultures were fed every 3 days, and
   mammospheres were counted after 2–3 weeks.

ATP Measurement

   Cellular ATP levels were measured with a luminometric method using an
   ENLITEN ATP assay system (Promega). In brief, cells were seeded into a
   24‐well plate, and cellular ATP was extracted by incubation with 100 µL
   of trichloroacetic acid (2%) for 2 min followed by centrifugation at
   20 000 × g at 4 °C for 3 min. The supernatants were neutralized using a
   neutralizer solution (20× 0.33m KOH with 0.1m Tris‐acetate; pH 7.75),
   and 40 µL of each supernatant was then added to the same volume of rL/L
   reagent for measurement of ATP‐driven chemiluminescence signals in a
   luminometer.

Seahorse Assay

   A Seahorse XF96 Extracellular Flux Analyzer (Seahorse Bioscience) was
   used to evaluate the oxygen consumption rate (OCR) and FAO capacity in
   breast cancer cells according to the manufacturer's instructions. All
   cancer cells and murine primary mammary tumor cells were obtained as
   described above. Cells (10 000 live cells per well) were plated in an
   XF96 cell culture plate at 90% confluence. For determination of the
   OCR, the cell culture medium was replaced with XF DMEM (Agilent,
   #103575‐100) supplemented with 25 mm glucose, 1 mm sodium pyruvate, and
   2 mm glutamine and incubated for 1 h at 37 °C prior to the assay.
   Mitochondrial stress was induced with a Mito Stress Test Kit (Agilent,
   103015–100) by sequential injection of 1 µm oligomycin, 1 µm FCCP, and
   0.5 µm rotenone and antimycin A. For measurement of the FAO rate, the
   presence of palmitate‐BSA in an assay medium in combination with
   etomoxir (ETO) can drive cells to oxidize fatty acids (FAs) and can
   reveal the portion of the OCR signal generated by FAO. As described in
   detail below, when cells were 80–90% confluent, the culture medium was
   replaced with a substrate‐limited medium (XF DMEM supplemented with
   0.5 mm glucose, 1.0 mm glutamine, 0.5 mm carnitine, and 1% FBS) for 4 h
   at 37 °C. The substrate‐limited medium was then replaced with FAO assay
   KHB buffer (111 mm NaCl, 4.7 mm KCl, 1.25 mm CaCl[2], 2.0 mm MgSO[4],
   1.2 mm NaH[2]PO[4]) supplemented with 2.5 mm glucose, 0.5 mm carnitine,
   5 mm HEPES and adjusted to pH 7.4 for 45 min prior to the assay. To
   quantify FAO, the assay medium was supplemented with 50 µm ETO 15 min
   before the assay. BSA control (0.17 mm BSA (Low Free Fatty Acid, MP
   Biomedicals, cat# 199899) and 150 mm NaCl; pH 7.2) or palmitate (Sigma,
   cat# P9767) conjugated to BSA (6:1) was added at a final concentration
   of 175 µm as exogenous fatty acids immediately before the measurement.
   Mitochondrial stress was induced as described above. The maximal OCR
   was calculated by subtracting the nonmitochondrial OCR from the OCR
   after FCCP treatment, while the ATP‐linked OCR was calculated by
   subtracting the OCR after oligomycin treatment from the basal OCR. The
   FAO‐associated OCR was measured according to the instructions of the
   Agilent Seahorse XF Palmitate‐BSA FAO Substrate Quickstart Guide. Cells
   were counted after completion of the experiment to normalize the rates
   in the individual wells, and the data were analyzed with Agilent Wave
   software v.2.6.

Gas Chromatography–Mass Spectrometry (GC–MS) Analysis

   For tricarboxylic acid (TCA) metabolite extraction in U‐^13C[16]
   palmitate and U‐^13C[2] acetate cellular labeling experiments, breast
   cancer cells cultured to ≈80–90% confluence were treated with 0.5 mm
   U‐^13C[16] palmitate (Cambridge Isotope Laboratories, Inc.; cat#
   CLM‐409‐0.5) and 1 mm carnitine (Sigma, cat# C0158) for 2 h prior to
   harvesting. Cells were washed three times in precooled PBS and quenched
   by bringing the bottom (outer wall) of the culture dish in contact with
   liquid nitrogen. Metabolites were extracted using ice‐cold 80% methanol
   (methanol/water, v/v) and transferred to a new tube using a cell
   scraper. Samples were centrifuged at 13 500 rpm for 10 min at 4 °C to
   remove cell debris. Supernatants were transferred into 1.5 mL Eppendorf
   tubes and dried in a vacuum centrifugal enrichment system at 4 °C
   (Labconco Corporation), and 50 µL of methoxyamine (Aladdin, cat#
   M109434) in pyridine solution (20 mg mL^−1) was added. After vortexing
   for 60 s and ultrasonication for 5 min in an ice bath, the oximation
   reaction was conducted at 37 °C in a water bath for 1.5 h.
   Subsequently, 50 µL of N‐methyl‐N‐(trimethylsilyl)trifluoroacetamide
   (MSTFA; Sigma, cat#M‐108‐5×1 ML) was added, and the final silylation
   derivatization was performed at 37 °C for 1.5 h. The mixture was
   centrifuged at 15 000 × g for 10 min at 4 °C, and 60 µL of each sample
   was transferred into injection vials (CNW Technologies, P/N:
   VDAP‐4025BS‐629‐100, P/N: VAAP‐32009E‐1232A‐100, P/N:
   VEAP‐5394‐09FRB‐100) and analyzed by GC–MS within 24 h. To quantify
   U‐^13C[16] palmitate in medium supernatants, 200 µL of homogenized
   medium samples were extracted with 400 µL of methyl tert‐butyl ether
   (MTBE; Sigma, cat# 34875) and 80 µL of methanol, followed by vortexing
   for 30 s and ultrasonication for 10 min in an ice bath. After
   centrifugation at 3000 rpm for 15 min at 4 °C, 300 µL of each
   supernatant was transferred to a 1.5 mL Eppendorf tube, dried under
   nitrogen gas, and dissolved in 100 µL of MSTFA for subsequent GC–MS
   analysis (Agilent Technologies; Agilent 5977B).

Immunoprecipitation (IP)

   Cells were lysed in lysis buffer (150 mm NaCl, 20 mm Tris‐HCl, 1 mm
   EDTA, 1 mm EGTA, and 1% Triton X‐100) supplemented with 1 mm PMSF. The
   cell lysates were precleared with protein A/G beads and were then
   immunoprecipitated with the corresponding specific antibody and protein
   A/G beads at 4 °C overnight with gentle agitation by vertical rotation.
   The immunoreactive products were washed three times and boiled in
   protein loading buffer for 10 min prior to SDS–PAGE. Bound proteins
   were detected using Western blotting as described above.

GST Pulldown Assay

   GST fusion proteins or His fusion proteins were expressed in E. coli
   BL21. To purify GST or GST‐tagged proteins, cells were lysed by
   sonication in lysis buffer (137 mm NaCl, 20 mm Tris‐HCl, 2 mm EDTA, 10%
   glycerol, and 1% Nonidet P‐40) supplemented with 1 mm PMSF and were
   then purified using Pierce Glutathione Agarose (Thermo, cat# 16101) at
   4 °C for 1 h. The bead‐bound GST fusion proteins were washed with lysis
   buffer and were then incubated with His‐tagged protein extraction
   solution at 4 °C for 3 h with gentle agitation by vertical rotation.
   After unbound proteins were removed with washing buffer, bound proteins
   were eluted by boiling and separated via SDS–PAGE.

Luciferase Reporter Assay

   Cells were seeded in 24‐well plates and cotransfected with the STAT3
   luciferase reporter hSIE/APRE (containing a STAT3‐responsive element),
   β‐galactosidase (β‐gal) expression vectors, and other relevant
   plasmids. After transfection, cells were treated with or without IL‐6
   (25 ng mL, 6 h) for STAT3 activation. A luciferase assay was performed
   as described previously^[ [201]^49 ^], and luciferase activity was
   normalized to the corresponding β‐gal level.

Chromatin Immunoprecipitation (ChIP) Assay

   The ChIP assay was performed using a ChIP Assay Kit (Beyotime, cat#
   P2078) according to the manufacturer's instructions. In brief, cells
   were plated in 10‐cm dishes and fixed with 1% paraformaldehyde at
   80–90% confluence. After the cells were lysed with SDS lysis buffer,
   the lysates were sonicated to generate 400–1000‐bp DNA fragments and
   immunoprecipitated using specific antibodies (4 µg of an anti‐STAT3
   antibody and 8 µg of an anti‐OVOL2 antibody). Transcription factor
   binding‐ sites were predicted using the online JASPAR database
   ([202]http://jaspardev.genereg.net). The STAT3 binding site sequences
   in CPT1A is: 5’‐cggtttttcctggaataaacta‐3’. The STAT3 binding site
   sequences in CPT1B is: 5’‐gctcacttccgggtaggagccc‐3’. The OVOL2 binding
   site sequence in JAK1 is 5’‐tattatagccgttacaattatt‐3’. The OVOL2
   binding site sequence in JAK2 is 5’‐tccacaaaaactgctaggattt‐3’. The
   OVOL2 binding site sequence in TYK2 is 5’‐taacaattatgttattcttgtt‐3’.
   DNA products were analyzed using real‐time PCR and normalized to input
   DNA.

DNA Affinity Precipitation Assay

   Nuclear extracts (200 µg) were combined with 500 µL of DNA affinity
   precipitation assay buffer (25 mmol/l HEPES, pH 7.6, 60 mmol L^−1 KCl,
   5 mmol L^−1 MgCl[2], 7.5% glycerol, 0.1 mmol L^−1 EDTA, 1 mmol L^−1
   dithiothreitol, and 0.25% Triton X‐100). This mixture included
   0.2 nmol L^−1 biotin‐labeled DNA probe and 10 µg poly
   (deoxyinosine‐deoxycytosine) [poly(Di‐Dc)] (Roche, cat# 10108812001)
   and was incubated overnight at 4 °C. Following this, 20 microliters of
   streptavidin magnetic beads (Beyotime, cat# P2151‐1 mL) were added, and
   the combination was further incubated for 2 h. After three washes with
   DNA affinity precipitation assay buffer, the resulting precipitates
   were subjected to Western blot analysis.

   The sequence information for the probes used in this study is as
   follows:

   CPT1A Probe:

   ACATAAGGGATAGTTTATTCCAGGAAAAACCGAGCTTC

   CPT1B Probe:

   GGCTCCTACCCGGAAGTGAGCCAG

   CON Probe:

   TTCCCCGGGGAAAAATTTTCCCCGG

ChIP Re‐IP Assay

   Re‐Chromatin immunoprecipitation was conducted to confirm the
   coexistence of proteins and DNA within the same complex. Following an
   initial overnight immunoprecipitation using STAT3 antibody (1:100,
   CST), the protein‐DNA‐bead complexes underwent three washes with
   chromatin immunoprecipitation wash buffer, followed by a double wash
   with 1x TE buffer. Subsequently, the washed immunoprecipitated
   protein‐DNA complexes were eluted by incubating for 30 min at 37 °C in
   75 µL of re‐chromatin immunoprecipitation elution buffer (2% sodium
   dodecyl sulfate, 0.1 mol L^−1 NaHCO[3]), followed by centrifugation for
   3 min at 600 g in a microcentrifuge to collect the beads at the bottom
   of the Eppendorf tube. After centrifugation, the supernatant was
   isolated and diluted 20 times (to a final volume of 1.5 mL) with
   chromatin immunoprecipitation dilution buffer supplemented with 50 µg
   of bovine serum albumin and protease inhibitors. A second
   immunoprecipitation reaction was then carried out using antibodies
   against HDAC1(1:200, Abcam) and OVOL2 (1:100, Novus).

Statistical Analysis

   GraphPad Prism 8.0 software (GraphPad Software Inc., CA, USA) was used
   to evaluate the data. The Kaplan–Meier method was used for tumor‐free
   survival and overall survival analysis. The two‐tailed Student's t‐test
   was performed to compare differences between groups, while analysis of
   variance (ANOVA) was used for comparisons among multiple groups.
   p < 0.05 was considered statistically significant.

Ethics Approval and Consent to Participate

   All animals were maintained in the Laboratory Animal Center at Xiamen
   University (China) in accordance with institutional guidelines. All
   mouse protocols and experiments were approved by the Animal Ethics
   Committee of Xiamen University (acceptance no. XMULAC20190043). All
   samples were collected between 15:00 and 17:00 to avoid circadian
   variability.

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   R.‐P.L. performed conceptualization, methodology, validation, formal
   analysis, data curation, visualization, and wrote—original draft.
   J.‐J.H. performed resources, investigation, data curation, and
   validation. T.F. performed resources, investigation, data curation, and
   validation. Y.Z. performed Investigation. R.‐Q.T. performed resources,
   investigation, and data curation. D.A. performed resources and
   investigation. S.W. performed an Investigation. Q.‐S.H. performed
   investigation. C.C. performed supervision. Z.‐M.Z. performed
   supervision, project administration, and funding acquisition. R.Z.
   performed supervision. H.G. performed conceptualization,
   wrote—reviewed, and edited, supervision, project administration, and
   funding acquisition. B.‐A.L. performed conceptualization,
   wrote—reviewed, and edited, supervision, project administration, and
   funding acquisition. All authors reviewed the manuscript. Further
   information and requests for resources and reagents should be directed
   to and will be fulfilled by the lead contact, Bo‐An Li
   (bali@xmu.edu.cn.)

Supporting information

   Supporting Information
   [203]ADVS-11-2308945-s001.pdf^ (1.2MB, pdf)

Acknowledgements