A Novel ZIP4-HDAC4-VEGFA Axis in High-Grade Serous Ovarian Cancer

Simple Summary Despite tremendous research efforts, epithelial ovarian cancer (EOC) remains one of the most difficult cancers to detect early and treat successfully for >5-year survival. We have recently shown that ZIP4, a zinc transporter, is a novel cancer stem cell (CSC) marker and a therapeutic target for EOC. The current work focuses on developing new strategies to target ZIP4 and inhibit its CSC activities in EOC. We found that cells expressing high levels of ZIP4 were supersensitive to a group of inhibitors called HDACis. One of the major targets of these inhibitors is a protein called HDAC4. We revealed the new molecular bases for the ZIP4-HDAC4 axis and tested the efficacies of targeting this axis in the lab and in mouse models. Our study provides a new mechanistic-based targeting strategy for EOC. Abstract We have recently identified ZIP4 as a novel cancer stem cell (CSC) marker in high-grade serous ovarian cancer (HGSOC). While it converts drug-resistance to cisplatin (CDDP), we unexpectedly found that ZIP4 induced sensitization of HGSOC cells to histone deacetylase inhibitors (HDACis). Mechanistically, ZIP4 selectively upregulated HDAC IIa HDACs, with little or no effect on HDACs in other classes. HDAC4 knockdown (KD) and LMK-235 inhibited spheroid formation in vitro and tumorigenesis in vivo, with hypoxia inducible factor-1 alpha (HIF1α) and endothelial growth factor A (VEGFA) as functional downstream mediators of HDAC4. Moreover, we found that ZIP4, HDAC4, and HIF1α were involved in regulating secreted VEGFA in HGSOC cells. Furthermore, we tested our hypothesis that co-targeting CSC via the ZIP4-HDAC4 axis and non-CSC using CDDP is necessary and highly effective by comparing the effects of ZIP4-knockout/KD, HDAC4-KD, and HDACis, in the presence or absence of CDDP on tumorigenesis in mouse models. Our results showed that the co-targeting strategy was highly effective. Finally, data from human HGSOC tissues showed that ZIP4 and HDAC4 were upregulated in a subset of recurrent tumors, justifying the clinical relevance of the study. In summary, our study provides a new mechanistic-based targeting strategy for HGSOC.


Introduction
Epithelial ovarian cancer (EOC) and high-grade serous ovarian cancer (HGSOC) in particular, has a poor 5-year survival rate [1][2][3][4]. Despite tremendous research efforts, ovarian cancer remains one of the most difficult cancers to detect early and treat successfully has been conducted in linking HDAC4 and HIF1α and none of the work has linked their interaction to CSC.
HDACis are a family of synthetic and natural compounds that differ in their target specificities and activities. Several HDACis have been FDA approved for cancer treatment. In contrast to conventional chemotherapeutic agents, HDACis show strong tumor selectivity and cause less toxicity in normal tissues [51]. However, clinical trials with HDACi as single agents in solid tumors have been disappointing [52], indicating that combinational therapy is necessary for their clinical applications [30,53].
Many of the published and current studies focus on targeting CSC alone [reviews [31,[54][55][56][57]) and references herein]. However, increasing lines of evidence suggest that co-targeting CSC and non-CSC cell populations in cancer treatment is important for the treatment of EOC. Yet, this type of study [36,[57][58][59][60][61][62] is significantly underappreciated, partly due to a presumed and more rigid definition for CSCs, considering a cell population selected by a specific marker set as the only cell source initiating tumor formation. However, CSCs are highly dynamic and are interconvertible between CSCs and non-CSCs. Non-CSCs can be induced into a stem-like state enabling them to become drug-tolerant [12,[63][64][65][66][67]. Hence, co-targeting both these subpopulations effectively is essential and further investigation is urgently needed to fill in the gaps before moving towards more effective targeting of CSC in clinics.
Although combinational effects of an HDACi and CDDP in EOC cells have been tested (e.g., [36,[59][60][61][62]), and reagents co-targeting CSC and non-CSC cell populations in cancer cell lines or mouse models have been reported (e.g., [57,58]), these studies did not test the effect in clearly defined CSC and non-CSC populations and were not linked to ZIP4. In the current work, we tested our hypothesis that co-targeting the ZIP4 + CSC (via ZIP4-knockout (KO)/knockdown (KD), HDAC4/KD, or HDACis) and the bulk ZIP4 − non-CSC (using CDDP) populations is highly effective to develop new treatment modalities for HGSOC. Instead of developing specific ZIP4 inhibiting reagents in the current work, we have developed an innovative approach to selectively and effectively target ZIP4 high cells, which mainly represent CSC as we have recently shown [17]. We then conducted mechanistic studies and revealed a novel ZIP4-HDAC4-VEGFA axis functionally involved in spheroid formation in vitro and tumorigenesis in vivo in HGSOC. Taken together, our study provides a new mechanistic-based targeting strategy for HGSOC.
To our surprise, we found that ZIP4-OE significantly sensitized cells to trichostatin A (TSA), an HDACi. On the contrary, ZIP4-KO greatly increased cell resistance to TSA ( Figure 1B). These effects were consistent in several pairs of ZIP4 differentially expressing cells ( Figure 1B).

ZIP4 Induced Upregulation of Class IIa HDACs
One of the potential mechanisms by which ZIP4 sensitizes HDACis is that ZIP4 upregulates HDACs to make cells more sensitive to their inhibitors. We tested ZIP4 s effect on representative HDACs in Class I (HDAC1, 2 and 3), Class IIa (HDAC4, 5, and 7), Class IIb (HDAC6 and 10) and Class IV (HDAC 11).
ZIP4-OE significantly increased the expression of HDAC4 and 5 and ZIP4-KO significantly decreased the expression of HDAC4 and 5 as shown in Figure 2A, as well as in Figure 2C (quantitative data summarized from > three independent experiments for ZIP4, HDAC4,5, and 7). While ZIP4-OE significantly increased the expression of HDAC7, ZIP4-KO in PE04 cells did not significantly reduce HDAC7 s expression (Figure 2A,C). It is possible that the basal level of HDAC7 is not regulated by ZIP4 in PE04 cells. We also tested ZIP4 s effect on HDAC4, 5, and 7 and found that ZIP4-KD in PEA2 cells reduced their expression, with HDAC4-KD only affecting HDAC4 itself ( Figure 2D,E). These results confirmed the regulatory effect of ZIP4 on Class IIa HDACs. In contrast to its effect on Class IIa HDACs, ZIP4 had minimal or no effect on HDACs in other classes, including Class I, Class IIb and Class IV (Figure 2A,B).

HDAC4 Played an Important Role in HDACi Sensitization in PE04 and PEA2 Cells
We decided to focus on HDAC4 for the rest of our studies due to its high clinical relevance as mentioned in the Introduction. To determine the role of HDAC4 in HDACi sensitization, we generated HDAC4-knockdown (KD) using shRNA in both PE04 and PEA2 cell lines ( Figure 3A). HDAC4-KD reduced cell numbers (measured by MTT) by 17-35% ( Figure 3B). Similar to ZIP4 KO/KD, HDAC4-KD cells increased cellular resistance to TSA, PANO, and LMK235 ( Figure 3C-E), suggesting that HDAC4 plays an important role in cellular sensitivity to HDACis.

HDACis and HDAC4-KD Inhibited Spheroid Formation in HGSOC Cells
Spheroid formation is one of the standard assays for CSC [91][92][93]. ZIP4 was expressed in spheres, but not (or at lower levels) in nonsphere cells as we recently reported [17]. We tested the role of HDACis in both PE04 and PEA2 cells and found that TSA, PANO, and LMK135 dramatically inhibited spheroid formation in these cells ( Figure 4A,B), suggesting that HDACis are potent CSC inhibitors.
We have shown that ZIP4-KO inhibited cell proliferation and spheroid formation in both PE04 and PEA2 cells [16,17]. To determine the functional relationship between ZIP4 and HDAC4, we then overexpressed the HDAC4 gene with a plasmid co-expressing GFP in PE04-ZIP4-KO and PEA2-ZIP4-KD cells. HDAC4 overexpression (green cells) increased cell proliferation ( Figure 4C) and restored the spheroid formation activity in ZIP4-KO/KD HGSOC cells ( Figure 4D,E), confirming that HDAC4 is an important downstream mediator of ZIP4.
Cell proliferation measured in 2D cell culture and spheroid formation assays measured in 3D culture conditions detect partially overlapping (e.g., cell survival), yet different cellular properties. The latter reflects more CSC-related activities. We have shown previously that ZIP4-KO reduced cell numbers (measured by MTT) by 50-60% when compared to the parental cells (Figure 3 in [16]). HDAC4-KD reduced cell numbers by 17-35%, and HDAC4 overexpression increased cell numbers by 5-12%, when compared to the control cells ( Figures 3B and 4C). As shown in our previous publications [16,17] and in Figures 4 and 5. ZIP4-KO and HDAC4-KD resulted in essentially complete blockage of spheroid formation. Comparison of the results from the 2D and 3D culture conditions suggest that both ZIP4 and HDAC4 play important roles in spheroid formation and CSC properties. Their effects on cell proliferation is unlikely to account for all of the lost spheroid formation activities.    (H). LMK235 and Acri inhibited secreted VEGFA in PE04 and PEA2 cells. For quantification, cell aggregates < 100 nm and/or with irregular shapes were not counted as spheroids. For quantification, cell aggregates < 70 nm and/or with irregular shapes were not counted as spheroids. Representative pictures from ≥ three independent experiments are shown. *** p < 0.001.
HDAC4-KD significantly inhibited spheroid formation ( Figure 4C,D). To test the role of HIF1α or VEGFA in spheroid formation more specifically, we transfected HIF1α or VEGFA genes (co-expressing GFP) to HDAC4-KD PE04 and PEA2 cells. As shown in Figure 5C-F, only those cells transfected with either HIF1α or VEGFA (green cells) in (H). LMK235 and Acri inhibited secreted VEGFA in PE04 and PEA2 cells. For quantification, cell aggregates < 100 nm and/or with irregular shapes were not counted as spheroids. For quantification, cell aggregates < 70 nm and/or with irregular shapes were not counted as spheroids. Representative pictures from ≥ three independent experiments are shown. *** p < 0.001.
HDAC4-KD significantly inhibited spheroid formation ( Figure 4C,D). To test the role of HIF1α or VEGFA in spheroid formation more specifically, we transfected HIF1α or VEGFA genes (co-expressing GFP) to HDAC4-KD PE04 and PEA2 cells. As shown in Figure 5C-F, only those cells transfected with either HIF1α or VEGFA (green cells) in HDAC4-KD cell lines formed spheroids, indicating that HIF1α and VEGFA are not only involved in CSC activity, but also are downstream mediators of HDAC4.
ZIP4-KD and HDAC4-KD in PEA2 cells reduced the levels of HIF1α ( Figure S2). This may be related to HDAC4 s effect on HIF1α acetylation and stabilization as previously reported HIF1α [46,47]. In addition, we found that ZIP4/KD and HDAC4-KD in PE04 and PEA2 significantly reduced VEGFA production/secretion detected in cell supernatants ( Figure 5G). LMK-235 and acriflavine significantly reduced VEGF production in PE04 and PEA2 cells ( Figure 5H).

The Combinational Targeting CSC via the ZIP4-HDAC4 Axis and Non-CSC Using CDDP Was Highly Effective in Blocking Tumorigenesis
To test our overall hypothesis that co-targeting the ZIP4 + CSC, via ZIP4-KO/KD or HDAC4/KD, and the bulk ZIP4 − non-CSC (using CDDP) populations is crucial to develop more effective treatment modalities for blocking development of HGSOC, we used the PE04 xenograft mouse model and compared the tumor development and mouse survival times in five pairs of groups in the absence or presence of CDDP: PE04 control cells, PE04-ZIP4-KO cells, PE04-HDAC4-KD cells, PE04 control cells treated with PANO, and PE04 control cells treated with LMK235. The results are summarized in Table 1. In all mice, 5 × 10 6 cells/mouse were i.p. injected with n = 5-6 in each group. Both genetic manipulations (such as ZIP4 or HDAC4 KO/KD) and HDACi (PANO or LMK-235) were mainly aimed at targeting ZIP4-dependent CSCs and CDDP were mainly aimed at targeting the bulk tumor cells.
As shown in Table 1 and Figure 6A, genetically KO or KD ZIP4 and HDAC4, extended the mouse survival time from~40 days to~96 to 124 days, respectively. These data suggest that although ZIP4 was an upstream regulator of HDAC4, the latter is likely to be regulated by other factors and is likely to play a more important role in tumor development in vivo.
We found that single reagent treatments, including CDDP, PANO, and LMK-235 of PE04 cells, extended mouse survival days from~40 days to 86-95 days (Table 1 and Figure 6A). At the endpoint, all three reagents reduced overall tumor volumes in mice, with LMK-235 being more effective. In addition, all three drugs reduced the volumes of ascites compared to the control group ( Table 1).
The effects of combinational treatments with one arm on the ZIP4-HDAC4 axis and the other on CDDP were tested and compared. As expected, addition of CDDP significantly extended mouse survival times from the corresponding single reagent treatment group, by 85 (the ZIP4-KO group), 96 (the LMK235+CDDP group), 46 (the PANO+CDDP group), and 63 days (the HDAC4-KD) ( Table 1 and Figure 6A), supporting the concept of the co-targeting strategy. The longest survival times were observed in the LMK-235+CDDP and the HDAC4-KD+CDDP groups (187-191 days), suggesting blocking HDAC4 is very critical and effective.
All of the drug treatments lasted for 4 weeks (starting 7 days after tumor cell injection). There was no drug delivered during the rest of the days when mice survived. While the mouse survival times were significantly extended with treatments, the tumors in long-lived mice grew back to even bigger sizes in some of the groups in a tissue location-specific manner (Table 1 and Figure 6C). For example, addition of CDDP increased tumor volumes on the ovaries and fallopian tubes as shown in the representative photos in Figure 6C. More dramatically, HDAC4-KD resulted in large tumors (10-15 mm in diameter) in the absence or presence of CDDP, around ovaries and fallopian tubes ( Figure 6C), which was not observed in the LMK-235 groups, suggesting that there are differences between genetic depletion of the HDAC4 gene and pharmacological inhibition of HDAC4 and/or HDAC5. In addition, compared to control and ZIP4-KO mice, the mice in the groups treated with the two HDACis (PANO and LMK-235) and HDAC4-KD shared an interesting commonality: these reagents significantly blocked tumor metastases on the peritoneal walls ( Figure 6D). These results suggest that HDAC4 may have an inhibitory role for large tumor development on the ovaries and fallopian tubes but facilitates peritoneal dissemination.
As shown in Figure 6D, these three CSC-related proteins were expressed in PE04 control cell-derived tumors. ZIP4-KO resulted in greatly diminished expression of ZIP4 and HDAC4 in tumor sections, with reduced ALDH1A expression. HDAC4-KD and LMK235, on the other hand, dramatically reduced HDAC4 expression, without significantly affecting ZIP4 and ALDH1A expression, supporting that ZIP4 is upstream of HDAC4. In general, when CDDP was added in combination with one of the ZIP4-HDAC4 axis modulators, it enhanced ZIP4, HDAC4, or ALDA1A expression in tumors from different mouse groups, supporting the idea that CDDP has the ability to induce CSCs [95]. We analyzed human VEGF in ascites and found that human VEGFA was downregulated in the PE04-ZIP4-KO and -HDAC4-KD groups when compared to the control PE04 group ( Figure 6B).
The H&E staining of tumors from the ovaries ( Figure 7A) or from other organs ( Figure 7B) showed HGSOC histology confirmed by our pathologist. Immunohistochemistry (IHC) staining for ZIP4, HDAC4, and ALDA1A were conducted in tumors from different groups of mice.

ZIP4 and HDAC4 Were Upregulated in a Subset of Recurrent vs. Primary Human HGSOC Samples
Our published data, along with TCGA and Oncomine data have shown and confirmed that ZIP4 is upregulated in human EOC tissues [13,16,96,97], justifying its clinical relevance to EOC. We tested ZIP4 and HDAC4 expressions in 10 pairs of primary and recurrent (platinum resistant) human HGSOC samples (each pair was from the same patient). Three of 10 these pairs overexpressed ZIP4 (more brown colored) in the recurrent vs. primary tumor tissues ( Figure 8A). Among them, two of three of the ZIP4 overexpression pairs also had HDAC4 overexpression ( Figure 8B). HDAC4 alone was overexpressed in 37% (7 of 19) of recurrent paired samples tested.

Discussion
Targeting CSC is a well-accepted concept for cancer treatment, especially for overcoming drug-resistance. Many studies have been conducted to target CSC in various cancers, including EOC [12,14]. Most of these studies focus on developing direct depleting or inhibiting the CSC marker. For example, several inhibitors have been developed and/or tested to inhibit ALDH isoforms in cancer CSC studies a very active field [9,[98][99][100][101]. In addition, many of these studies have tested the effect of targeting CSC only (reviews [31,[54][55][56][57] and references herein).
In the current work, however, we have challenged the current paradigm in developing anticancer strategies in several aspects. Firstly, CSCs have been considered mainly to display MDR in general. However, our data demonstrate that they may also exhibit high selectivity in drug response and sensitize to specific drugs. We developed and tested the strategy by taking the advantage of the preferential response to HDACis in ZIP4 + cells to selectively target CSCs, as a contrary to inhibit the CSC marker ZIP4 directly. We showed here that ZIP4 + CSCs were highly sensitive to HDACis, especially HDACis targeting IIa HDACs, providing feasible tools to target these cells selectively and efficiently. HDACis, and LMK-235 in particular, indeed displayed strong antitumor and anti-CSC activities. At the current stage, genetic manipulation is still relatively difficult to apply to humans directly and a selective inhibitor of ZIP4 has yet to be developed. On the contrary, several HDACis have been FDA approved and are in clinical trials.
Secondly, in the present work, we have tested our hypothesis that co-targeting CSC and non-CSC cell populations in cancer treatment is necessary and/or pivotally important for the treatment of EOC. While co-targeting the bulk and CSC populations in cancer have been reported (e.g., [57,58,[102][103][104][105]), it is significantly underappreciated, partly due to a presumed and more rigid definition for CSCs, considering a cell population selected by a specific marker set is the only cell source initiating tumor formation. However, when compared to the developmental stem cell concept, CSC is more of an operational concept, mainly defined by their ability to initiate tumors and mainly identified by surface makers [8], with high translational/clinical application potential. The CSC phenotype varies substantially and may undergo reversible phenotypic changes. Non-CSCs can be induced into a stem-like state enabling them to become drug-tolerant [12,[63][64][65][66][67]95]. Hence, it is important to co-target both subpopulations of tumor cells. In our in vivo mouse models, CDDP is used to target mainly the highly proliferative non-CSCs and inhibition of the novel ZIP4-HDAC4 axis is the major strategy for targeting CSCs. We used five sets of different mouse models to test and compare the efficacies: PE04 control cells ± CDDP; ZIP4-KO ± CDDP, HDAC4-KD ± CDDP, a pan-HDACi (PANO) ± CDDP, and a selective HDAC4/5 inhibitor, LMK-235 ± CDDP. In all five pairs of comparisons, even though single gene KO or KD or single HDACi significant increased mouse survival times, combination with CDDP clearly had significant additional benefit, extending additional mouse survival times ranging from 45 to 96 days (~4.5-9.6 human years [106]), supporting our overall hypothesis. It is important to note that CDDP treatment eventually enhanced expression of ZIP4, HDAC4, or ALDA1 in tumors from different mouse groups, supporting the idea that CDDP has the ability to induce CSCs [95]. These data suggest that a longer and continuous CSC-targeted treatment modality should be seriously considered.
Thirdly, although HDACis as a therapeutic strategy against CSCs has been studied using different CSC markers [107][108][109][110][111], the mechanisms have only been minimally or indirectly studied [112][113][114][115]. In addition, although combinational effects of an HDACi and CDDP in EOC cells have been tested (e.g., [36,[60][61][62]) and reagents co-targeting CSC and non-CSC cell populations in cancer cell lines or mouse models have been reported (e.g., [57,58]), these studies did not test the effect in clearly defined CSC and non-CSC populations or were not linked to ZIP4. Importantly, most, if not all, of these studies were conducted in vitro and were focused on Class I HDACis. In fact, the majority of HDACi studies, including preclinical studies and clinical trials are mainly focused on Class I HDACs, since they are the main nuclear HDACs with high enzymatic activities and important physiological functions. However, in vivo HDACi toxicity observed up to date are also mainly caused by inhibiting HDAC1-3, the core nuclear histone deacetylase [116][117][118]. On the other hand, the concept that HDAC4 is a highly effective target is supported by the data presented here and it is not only more innovative, but also has the potential to significantly reduce toxicity. Comparison of our results from PANO and LMK235 suggests that developing more specific and selective HDACis is likely to be a promising direction to overcome the toxicity issues related to HDACIs. Fourthly, we have previously shown that ZIP4 is an upstream regulator of several CSC markers, including ALDH1, OCT4, and SOX9 and NOTCH3 [16,17,119,120]. Here, we showed here that ZIP4 regulated HDAC4/5 and NOTCH3 may also be an upstream regulator of HDAC4/5, since NOTCH3-KO in PE04 cells also reduced the expression of HDAC4/5 ( Figure 2). It is interesting to note that while ZIP4-KO extended mouse survival times from~40 days to~96 days, HDAC4-KD had an even longer survival time (~124 days). These data suggest that HDAC4 may also be regulated by other factors and may represent a better target for CSCs. In addition, HDAC4 may have organ-specific effects on tumor metastasis as shown in Table 1 and Figure 6. The larger, but more localized tumors resulting from HDAC4-KD may have clinical significance: these tumors may be easier to remove surgically than numerous microscopic metastases, a major source of recurrence. We also found that treatment with LMK-235, a selective inhibitor of HDAC4/5, did not result in large tumors in the ovaries and fallopian tubes, suggesting that this effect is more HDAC4specific and/or not directly related to the inhibitory effect of LMK-235 on HDAC4. These concepts warrant further investigation in more HGSOC cell lines/PDXs.
Finally, it is almost certain that not all the functions of ZIP4 are mediated by HDAC4. We have already shown that ZIP4 also upregulates HDAC5, and possibly HDAC7 and/or HDAC9. The roles of HDAC5, 7 and 9 in CSC of HGSOC need to be further investigated. In addition, how ZIP4 regulates HDAC4 and other HDACIIa HDACs remains to be further investigated. Similarly, it is very likely that there are multiple downstream mediators of HDAC4, which play important roles in the overall effects of HDAC4 in cancer development in vivo. We focused on HIF1α and VEGFA in the current work, with other downstream mediators warranting further investigation. HGSOC is a highly heterogenic disease. Other than TP53, any other single gene occurs in 0.5 to 25% cases, with one exception (CDKN2A, 32%) [13]. Detecting 20-30% ZIP4-HDAC4 co-overexpression in human recurrent HGSOC (Figure 8) is highly significant to justifying the clinical significance. The cancer cell lines used in this work were selected based on that (1) they are HGSOC cell lines; (2) the two pairs of cell lines were derived from two individual patients before (PE01 or PEA1) and after (PE04 or PEA2) the onset of multidrug resistance [88,89]; hence they are the most clinical relevant ovarian cancer cell lines to study drug resistance in EOC; and (3) we have recently shown that endogenous ZIP4 is expressed in all HGSOC cell lines examined, including PE04 and PEA2 cells, In contrast, several nontumorigenic and/or immortalized cell lines, including NIH3T3, Cos7, human ovarian surface epithelial (HOSE) cell lines, T29, T80, and a human fallopian tube cell line FT194 expressed low or undetectable ZIP4 (Figure 1, ref. [17]). In this work, we chose T80 for ZIP4 overexpression, since the endogenous ZIP4 expression in T80 cells was low. ZIP4 or HDAC4 genes were either knocked out using the CRISPR technology, knocked down using shRNA, or transfected with the human HDAC4 gene for overexpression, so that both gain-and loss-of-functional studies can be carried out. In addition, a PE04-NOTCH3-KO cell line was included, since we have shown that NOTCH3 is a functional downstream mediator of ZIP4 [17]. The pair of PE01/PE04 cell lines was from Dr. Daniela Matei (Northwestern University); the pairs of PEA1/PEA2 cells were from Sigma (St Louis, MO, USA). The T29 human ovarian surface epithelial cell line was from Dr. Jinsong Liu (M.D., Anderson). PE04-ZIP4-KO cell lines were generated using CRISPR as we described previously [16]. All cell lines were maintained in a humidified atmosphere at 37 • C with 5% CO 2 . PE01/PE04 cells were cultured in RPMI 1640 with glutamine, 10% fetal bovine serum (FBS), and 100 µg/mL Penicillin-Streptomycin-Amphotericin B. PEA1 and PEA2 were cultured in RPMI 1640 with 2 mM Glutamine, 2 mM sodium pyruvate, 10% FBS and 100 µg/mL penicillin incubation (37 • C, 5% CO 2 ). After enzymatic disaggregation, cells were harvested for experiments. Human HGSOC tissues were obtained from CHTN as we described in our previous studies through Dr. Xu's IRB [121].

Materials and Methods
Acriflavine

Western Blot Analysis
Western blot analyses were conducted using standard procedures and proteins were detected using primary antibodies and fluorescent secondary antibodies (IRDye 800CWconjugated or IRDye 680-conjugated antispecies IgG, Li-Cor Biosciences, Lincoln, NE, USA) as we described previously [5].

Fluorescence-Activated Cell Sorting (FACS) of ZIP4 + Cells
FACS-based sorting and analysis of marker ZIP4 + cells was conducted using the BD FACSAria cell sorter system (Becton-Dickinson, Franklin Lakes, NJ, USA) and BD LSR Fortessa Analyser (BD Biosciences, San Jose, CA, USA), and data analyzed by FlowJo V10 (BD Biosciences, San Jose, CA, USA) as we described in [17]. shRNA lentivirus vectors were co-transfected with packaging vectors to 293T cells in OptiMEM media. PE04 or PEA2 cells were infected by viruses three times and stable clones were selected by puromycin (0.5 µg/mL) for 10 days.

Spheroids-Formation Assays
As described in detail in our recent publication [17].

VEGFA ELISA Assays
Cells (1 × 10 4 /well) plated in six-well plates were grown in 2 mL of RPMI-1640 medium with 10% FBS for 24 h and then cultured in serum-free medium for 48 h. The culture supernatants were collected and their human VEGFA concentration were measured using ELISA kits (R&D). Ascitic fluids were collected from groups of treated and control female NSG mice (groups: PE04 with or without CDDP treatment; PANO with or without CDDP, LMK235 with or without CDDP; and HDAC4-KD with or without CDDP) and stored at −80 • C until ELISA assays were conducted. Ascites was diluted 8-to 16-fold in 1× reagent diluent for the ELISA assays. Both human VEGFA and mouse VEGF ELISA assays were performed using the DuoSet VEGFA ELISA kits from R&D Systems (Minneapolis, MN, USA) in triplicate wells according to the manufacturer's instructions. The optical density at 450 nm was measured on an automated plate reader (PerkinElmer, Santa Clara, CA, USA). Experiments were repeated three times.

HDACi Cytotoxicity Assay
Cell viability under HDACs inhibitions were performed using different concentrations of pan-HDAC inhibitors TSA and PANO, as well as LMK-235, a selective HDAC5/HDAC4 inhibitor. All of them were dissolved in 100% dimethyl sulfoxide (DMSO) according to the manufacturer's protocol to a stock concentration of 20 mM and stored at −20 • C. Cells were seeded at a density of 2000 cells per well in 96-well plates. After culturing for 6 h, various doses of TSA (Sigma), panobinostat (PANO, VWR) or LMK-235 (VWR) were added. After cells were exposed to drugs for the indicated times (24/48/72/h), cell viability was determined using the MTT assay as described previously (81). Engrafted mice were inspected daily for tumor/ascites appearance by visual observation, palpation, and tumor latencies. Mice were sacrificed by cervical dislocation at obvious distended abdomens or at >200 days post-transplantation. Tumors were counted at each metastatic location and tumor diameters were measured. Data are represented as mean ± SD (n = 5). The curves represented the trend of the surviving days increased after co-treatments. Animal protocols were approved by the Indiana University School of Medicine Animal Care and Use Committee (#11345). Xenograft tumors were resected, fixed in 10% neutral, buffered formalin, and embedded in paraffin for sectioning (5 µm) on a rotary microtome, followed by slide mounting, H&E staining, and histologic assessment by Dr. Robert Emerson, a pathologist at the Indiana University School of Medicine.

Human HGSOC IHC
Formalin-fixed and paraffin-embedded primary ovarian cancer tissues (all from HG-SOC patients) were obtained from the Department of Pathology at the Johns Hopkins Hospital, Baltimore, Maryland. Specimens with tumor cell population > 50% and minimal or no necrosis were included in this study. Specimens were arranged in tissue microarrays to facilitate IHC and to ensure that the tissues were stained under the same conditions. The study was approved by the Johns Hopkins University School of Medicine Institutional Review Board. The scoring of IHC is the following method published [122].

Statistical Analyses
The Student's t-test was utilized to assess the statistical significance of the difference between two treatments. The asterisk rating system as well as quoting the p value in this study was * p < 0.05; ** p < 0.01; and *** p < 0.001. A p value of less than 0.05 was considered significant.

Conclusions
Our studies have shown that co-targeting the ZIP4-HDAC4-VEGFA axis with a common chemotherapy represents new strategies to circumvent drug resistance in vitro and in vivo for HGSOC, a deadly disease. These studies significantly expand our understanding of the roles, the molecular mechanisms, and interactions of ZIP4 with other signaling molecules, as well as its potential clinical targeting value in EOC.