HE4 promotes collateral resistance to cisplatin and paclitaxel in ovarian cancer cells

Cell culture

SKOV3 and OVCAR8 ovarian cancer cells were obtained from American Type Culture Collection (Manassas, VA). Cells were cultured in Dulbecco Modified Eagle Medium (DMEM, Gibco, 11965-065) with 10 % fetal bovine serum (FBS) and 1 % penicillin/streptomycin and kept in a 37 °C humidified incubator with 5 % CO².

Stable cell lines

All null vector (NV) and HE4-overexpressing stable cell lines were previously established [7]. To generate HE4-CRISPR Double Nickase stable cell lines, SKOV3-C1 cells were transfected in 6-well plates with 1 μg HE4 Double Nickase Plasmid (Santa Cruz, sc-402876-NIC) or Control Double Nickase Plasmid (Santa Cruz, sc-437281), using 5 μl Lipofectamine 2000 (Invitrogen). After 48 h, media was changed to 1 μg/ml puromycin containing media for five days, then split into larger plates and selected for an additional five days. RNA and tissue culture supernatant was collected to confirm downregulation of HE4 levels by quantitative RT-PCR (qRT-PCR) and ELISA. Cells were maintained in DMEM supplemented with 10 % FBS, 1 % penicillin/streptomycin, and 1 μg/ml puromycin.

Cell treatments

Cells were treated with the described doses of cis-diamminedichloroplatinum(II) (cisplatin, Sigma Aldrich, 1134357) or paclitaxel (Sigma Aldrich, T7402) dissolved in dimethyl sulfoxide (DMSO, Sigma Aldrich, D8418), or DMSO alone as a control. Cells were collected directly into either Trizol (Ambion, 15596018) or Cell Lysis Buffer (Cell Signaling, 9803) at the indicated time points for analysis. Cells were treated with recombinant human HE4 (rHE4, MyBioSource, MBS355616) added directly to the media to a final concentration of 20 nM. Cells were treated with recombinant human epidermal growth factor (rEGF, Calbiochem, 324831) at a concentration of 10 ng/ml.

Cell viability assays

All cells were seeded at 2 000 cells/well in 96-well plates. Cells were treated with increasing doses of cisplatin and paclitaxel as described. After 48 h, cell viability assays were performed by adding 10 μl/well of CellTiter 96® Aqueous One Solution Cell Proliferation MTS Assay (Promega, G3580), incubating at 37 °C/5 % CO² for 2 h, and reading absorbance at 492 nm. Results are displayed as percent survival of vehicle treated cells.

Microarray

SKOV3-NV and SKOV3-C1 cells were treated in triplicate at 80 % confluency with 25 μM cisplatin or DMSO vehicle. Total RNA was collected 24 h later using an RNeasy Mini Kit (Qiagen, 74104) and checked for purity by NanoDrop 2000 (Thermo Scientific). The RNA samples were randomly assigned numbers and submitted to the Brown Genomics Core Facility for Bioanalyzer (Agilent 2100) RNA quality analysis. Affymetrix HTA 2.0 Arrays were performed according to the manufacturer’s instructions at the Core Facility using 150 ng total input RNA.

DAVID analysis

Database for Annotation, Visualization, and Integrated Discovery (DAVID) v6.7 [12, 13] was used to identify the top ten enriched annotation terms among 180 genes differentially expressed (1.5-fold in either direction, p < .05) between SKOV3-NV and SKOV3-C1. Default DAVID parameters were employed as follows:

Kappa Similarity: Similarity Term Overlap – 3; Similarity Threshold – 0.5

Classification: Initial Group Membership – 3; Final Group Membership – 3; Multiple Linkage Threshold – 0.5

Enrichment Threshold: EASE – 1.0

Stringency: Medium

Quantitative PCR

RNA was collected using an RNeasy Mini Kit (Qiagen, 74104) or Trizol extraction/LiCl precipication. Total RNA (500 ng) was reverse transcribed into cDNA using the iScript cDNA Synthesis Kit (Bio-Rad, 1708890) according the manufacturer’s protocol. To validate differentially expressed genes between SKOV3-NV and SKOV3-C1 cells identified by microarray, the same RNA samples used for the microarray were employed. To validate the differential cisplatin-induced upregulation of EGR1 between SKOV3-NV and SKOV3-C1/C7, microarray RNA samples were used, as well as RNA isolated from SKOV3-C7 cells that were treated in the same manner as the cells used in the microarray. Quantitative PCR was performed in triplicate by loading 1 μl cDNA reaction, 2 μl each of 5 μM custom forward and reverse primers (Invitrogen) or 1 μM forward and reverse validated primers (realtimeprimers.com), 10 μl SYBR Green (Applied Biosciences [ABI], 4367659) and 5 μl RNAse-free water to each well. Samples were run on an ABI 7500 Fast Real-Time PCR System, and data was analyzed using the ΔΔCt method. Relative expression levels were normalized to 18 s rRNA to correct for equivalent total RNA levels. Validated MAPT, CYP1B1, and EGR1 primers were purchased from realtimeprimers.com. Custom primer sequences (Invitrogen) are as follows:

AKT3 F – AAG GGA AGA ATG GAC AGA

AKT3 R – ATG GGT TGT AGA GGC ATC

NMUR2 F – CCG TTC CAC ATT GAC CGA CT

NMUR2 R – CAC CAC ATG GAC GAG GTT GA

SEPT3 F – TTG CCC TGC TTC GAG ACT TT

SEPT3 R – CTT TCC TCT GTG TCC ACG CT

18 s rRNA F – CCG CGG TTC TAT TTT GTT GG

18 s rRNA R – GGC GCT CCC TCT TAA TCA TG

Western blot

Protein was extracted from cell pellets in Cell Lysis Buffer (Cell Signaling, 9803) with 1 mM PMSF, according to the manufacturer’s protocol. Protein concentrations were determined by DC Protein Assay (Bio-Rad Laboratories, 5000116). Western blot analysis was performed by loading equal amounts of protein boiled with Novex Sample Reducing Agent (Life Technologies, NP009) and NuPAGE LDS sample buffer (ThermoFisher Scientific, NP0007) into a 4–12 % gradient NuPAGE Novex Bis-Tris gel [Life Technologies, NP0321BOX (mini), WG1402BX10 (midi)]. Protein was transferred by semi-dry transfer to methanol-activated 0.2 μm PVDF membranes (Bio-Rad, 162-0177) at 0.12-0.2 A for 1 h 15 m. Membranes were blocked in 5 % milk in phosphate-buffered saline with 0.05 % Tween 20 (PBS-T) for 30 m at room temperature, incubated in primary antibody in 5 % milk in PBS-T overnight at 4 °C, and then in secondary antibody in 5 % milk in PBS-T for 1 h at room temperature, with PBS-T washes in between. Amersham ECL Prime Western Blot Detection System (GE Healthcare, RPN2232) was used for detection of HRP-tagged secondary antibodies. Blots were developed using x-ray film in a Kodac film developer or imaged directly in a Biorad Chemidoc MP Imaging System. GAPDH was used as a loading control. Antibodies and dilutions used are as follows:

PARP (Cell Signaling, 9532, 1:1000)

phospho-p44/42 MAPK (ERK1/2) (Cell Signaling, 4370, 1:2000)

p44/42 (ERK1/2) (Cell Signaling, 9102, 1:2000)

EGR1 (Santa Cruz, sc-110, 1:200)

p38 (Cell Signaling, 9212, 1:1000)

phospho-p38 (Cell Signaling, 9215, 1:1000)

GAPDH (Cell Signaling, 2118, 1:2000)

β-tubulin (Cell Signaling, 2146, 1:2000)

α-tubulin (Cell Signaling, 2144, 1:1000)

Densitometry

Image J was used to perform densitometry analysis of western blots. Images of blots were analyzed in 8-bit TIFF format, using the “analyze gel” function. Where no band was detected, a value of “1” was assigned. Relative band densities were normalized to a loading control, or the appropriate total protein for phospho-proteins, and then the lowest value was set to 1.

Statistics

In all instances where statistics are shown, they represent n ≥ 3 independent experiments, and p-values were determined by unpaired 2-tailed Student t-test. For the microarray, Affymetrix Transcriptome Analysis Console (TAC) software was used to generate fold-changes and statistical significance, and ANOVA p-values generated by TAC were used for p-value cutoffs.

HE4-overexpressing cells are more resistant to cisplatin and paclitaxel treatment

Since HE4 serum levels correlate with cisplatin resistance in ovarian cancer patients [7–10], and our previous data suggested that HE4 promotes chemotherapy resistance in vitro and in vivo [7], we set out to further define the chemotherapy response of ovarian cancer cells that overexpress HE4. We employed SKOV3 ovarian cancer cells stably expressing a null vector plasmid (SKOV3-NV) and two different clonally-selected cell lines stably expressing a pCMV6-HE4 plasmid (SKOV3-C1 and SKOV3-C7) [7]. SKOV3 cells are ideal to examine the effects of HE4 overexpression since they secrete very low levels of HE4 (<15 pM), as detected by ELISA [7]. We treated the cells with 0, 1.56, 3.125, 6.25, and 12.5 μM cisplatin or 0, 0.625, 1.25, 2.5, and 5 nM paclitaxel for 48 h, then performed MTS assays to measure cell viability. SKOV3-C1 and SKOV3-C7 cells were significantly more resistant than SKOV3-NV cells to cisplatin (Fig. 1a) at 3.125 μM (p = .029 and .024, respectively), 6.25 μM (p = .034 and 0.37, respectively), and 12.5 μM (p = .0020 and .0018, respectively). SKOV3-C1 and SKOV3-C7 cells were also more resistant than SKOV3-NV cells to paclitaxel (Fig. 1b) at 1.25 nM (p = .017 and .083, respectively), 2.5 nM (p = .0094 and .022, respectively) and 5 nM (p = .0050 and .0035, respectively).

In order to confirm that this resistance was due to inhibition in apoptosis, we analyzed the cleavage of poly ADP ribose polymerase (PARP) in SKOV3-NV, SKOV3-C1, and SKOV3-C7 cells treated with 5 μM cisplatin for 24, 48, and 72 h. Even in untreated cells, baseline levels of cleaved PARP were lower in HE4-overexpressing clones than in SKOV3-NV cells. In SKOV3-NV cells, PARP cleavage was clearly increased by 5 μM cisplatin at 24 h, and continued to remain elevated at 48 h. By 72 h, levels of cleaved PARP had diminished in SKOV3-NV cells, presumably because most of the cells had already undergone apoptosis. In HE4-overexpressing clones, uncleaved PARP did accumulate to a similar degree as in SKOV3-NV cells—suggesting that the cisplatin was able to enter the cells and cause DNA damage, thus recruiting PARP to sites of single-strand breaks. However, cleavage of PARP did not occur in HE4-overexpressing cells, indicating that the process of apoptosis was not completed (Fig. 1c-d). We also observed an accumulation in both PARP and cleaved PARP in SKOV3-NV cells treated with 1 nM paclitaxel for 48 h, but did not see this response as strongly in SKOV3-C1 or SKOV3-C7 cells (Fig. 1e-f). The fact that PARP did not accumulate robustly in HE4-overexpressing clones treated with paclitaxel—unlike what we saw with cisplatin treatment—is likely due to the vastly different mechanisms of action by which these two drugs result in DNA damage.

Preliminary dose response experiments revealed that OVCAR8 cells required high doses of cisplatin to achieve cell death. Thus, we treated OVCAR8-NV and HE4-overexpressing clones (OVCAR8-C5) with 0, 62.5, 125, 250, and 500 μM cisplatin for 24 h, and found a slight increase in resistance in the OVCAR8-C5 cells (Additional file 1A), which was also reflected by lower levels of cleaved PARP in OVCAR8-C5 cells than OVCAR8-NV cells treated with 100 μM cisplatin for 24 h (Additional file 1B-C). However, although OVCAR8 cells also secrete undetectable levels of HE4 as determined by ELISA (data not shown), they appear to already be highly resistant to cisplatin and are not an ideal model to examine the effects of HE4 on chemotherapy resistance. Likewise, OVCAR8-NV cells were also more resistant to paclitaxel than SKOV3-NV cells, and only exhibited a modest increase in resistance with HE4 overexpression (Additional file 1D). For this reason, we focused primarily on SKOV3 cells for the remainder of our studies.

CRISPR/Cas-mediated HE4 downregulation reduces chemoresistance in SKOV3-C1 cells

In order to confirm that overexpression of HE4 was responsible for the increase in cisplatin resistance observed, we knocked down HE4 in SKOV3-C1 cells by stable transfection of a human HE4 CRISPR/Cas Double Nickase plasmid. RNA from HE4-targeted cells (CRISPR-HE4) and control plasmid transfected cells (CRISPR-Ctrl) was collected to confirm effective targeting of the HE4 gene. Cell culture supernatant was also collected and subjected to ELISA to determine levels of secreted HE4 in the media. CRISPR-HE4 transfected cells displayed 70.3 % knockdown (p = .02) in HE4 mRNA levels compared to control cells. Furthermore, CRISPR-HE4 media had undetectable levels (<15 pM) of HE4, which is in agreement with what we previously observed in SKOV3 wild type and SKOV3-NV cells. HE4 concentration in CRISPR-Ctrl media measured 307 pM, which was comparable to our previously reported levels for SKOV3-C1 and SKOV3-C7 cells (250 pM and 362 pM, respectively) [7] (Fig. 2a). Next, we measured survival of CRISPR-Ctrl and CRISPR-HE4 cells in response to 25 μM cisplatin treatment (24 h), and found that CRISPR-Ctrl cells exhibited 92.2 % cell survival compared to vehicle-treated cells, while CRISPR-HE4 cells displayed 81.9 % survival compared to vehicle-treated cells (p = .005; Fig. 2b). Likewise, CRISPR-Ctrl cells exhibited 73.9 % survival in response to paclitaxel versus 65.9 % survival of CRISPR-HE4 cells (p = .005) (Fig. 2c).

Microarray analysis reveals genes differentially regulated by HE4 in ovarian cancer cells

In order to gain a better understanding of the differential regulation of transcription between SKOV3-NV and SKOV3-C1 cells in the presence and absence of cisplatin, we performed a microarray analysis of total RNA collected from SKOV3-NV and SKOV3-C1 cells treated with either vehicle [dimethylsulfoxide (DMSO)] or 25 μM cisplatin for 24 h (n = 3/group). All transcripts differentially regulated between groups 1.5-fold in either direction (ANOVA p-value < .05) are available in Additional file 2. The top fifteen annotated, protein-coding genes that were differentially regulated in vehicle-treated SKOV3-C1 cells compared to SKOV3-NV (p < .05) are shown in Table 1. The Database for Annotation, Visualization and Integrated Discovery (DAVID) [12, 13] was used to perform gene ontology analysis of all genes differentially expressed 1.5-fold in either direction (p < .05) in SKOV3-C1 cells versus SKOV3-NV. This analysis revealed enrichment for the terms “secreted glycoproteins”, “transmembrane domains”, “pregnancy and immunoglobulin-like domains”, “cell adhesion”, “plasma membrane”, “fibronectin”, “insulin-like growth factor binding”, “membrane fraction”, “extracellular matrix”, and “EGF-like domains” (Additional file 3). We validated four of the genes from the list of most differentially expressed genes by quantitative reverse-transcription PCR (qRT-PCR), as follows: V-Akt Murine Thymoma Viral Oncogene Homolog 3 (AKT3); septin 3 (SEPT3); cytochrome p450, family 1, subfamily B, polypeptide 1 (CYP1B1); and neuromedin 2 (NMUR2) (Fig. 3a-b).

Cisplatin-induced upregulation of EGR1 is suppressed in HE4-overexpressing cells

The top fifteen annotated, protein-coding genes that were differentially regulated between SKOV3-NV and SKOV3-C1 cells in the presence of cisplatin are listed in Table 2. This list excludes genes that were already differentially regulated between SKOV3-NV and SKOV3-C1 vehicle treated cells. Of these genes, early growth response 1 (EGR1), which is a key gene involved in regulating growth and apoptosis in a wide variety of tissues, presented as the most differentially expressed, with 4.07-fold higher expression in SKOV3-NV than SKOV3-C1 (p = .000201). Silencing of EGR1 has been shown to promote resistance to cisplatin in esophageal cancer cells [14], while overexpression of this gene sensitizes ovarian cancer cells to cisplatin-induced apoptosis [15]. Thus, we speculate that suppression of EGR1 upregulation serves as a downstream effector of HE4-mediated cisplatin resistance.

To further analyze genome-wide differences in cisplatin response between SKOV3-NV and SKOV3-C1 cells, we also conducted the following comparisons: SKOV3-NV (vehicle vs. cisplatin) and SKOV3-C1 (vehicle vs. cisplatin). The top fifteen differentially expressed annotated, protein-coding genes in either direction are displayed in Table 3. We also included EGR1 in this table, since it just missed the cutoff as the sixteenth gene upregulated in SKOV3-NV cells by cisplatin, but not in SKOV3-C1. Several other genes are represented in both Tables 2 and 3, including snail family zinc finger 1, (SNAI1), DNA-damage-inducible transcript 3 (DDIT3), ADP-ribosylation factor-like 14 (ARL14), and tumor necrosis factor (TNF). Moreover, while there were several genes that were cisplatin-regulated in both SKOV3-NV and SKOV3-C1, the overall number of all transcripts regulated by cisplatin (+/- 1.5-fold, p < .05) in SKOV3-NV was much greater than in SKOV3-C1 cells (5935 vs. 1035), as represented by comparative scatter plots (Additional file 4). Even among those genes that were regulated in both SKOV3-NV and SKOV3-C1 in Table 3, there tended to be a greater degree of change in SKOV3-NV cells, especially for those genes that were upregulated by cisplatin. These observations are in agreement with the stronger cisplatin-response in SKOV3-NV cells.

Because EGR1 was most the differentially expressed gene between SKOV3-NV and SKOV3-C1 treated with cisplatin, and has known roles in cisplatin response, we validated its differential regulation by qRT-PCR using SKOV3-NV, SKOV3-C1, and SKOV3-C7 RNA exposed to vehicle (DMSO) or 25 μM cisplatin for 24 h (Fig. 4a). Next, because EGR1 is an early response gene reported to be upregulated as early as 15–30 min following exposure to a stimulant [16, 17], we performed a time course analysis of EGR1 expression following cisplatin treatment of SKOV3-NV and SKOV3-C1 cells. Interestingly, an early upregulation of EGR1 occurred in both SKOV3-NV and SKOV3-C1 cells beginning at 15 min after cisplatin treatment, which peaked between 1 and 3 h post-treatment and diminished by 6 h. However, the second more robust increase in EGR1 expression observed at 24 h in SKOV3-NV cells was dramatically suppressed in SKOV3-C1 cells (Fig. 4b). A similar trend was noted in EGR1 protein levels. While EGR1 protein was increased at 6 h in both SKOV3-NV and SKOV3-C1 cells treated with 25 μM cisplatin, only SKOV3-NV cells maintained higher levels of EGR1 by 24 h (Fig. 4c-d). To confirm the intrinsic ability of SKOV3-C1 cells to upregulate EGR1, both SKOV3-NV and SKOV-C1 cells were treated with 10 ng/ml human recombinant epidermal growth factor (rEGF) to activate the EGFR signaling pathway, a known positive regulator of EGR1 gene expression [18, 19]. We observed a robust increase in EGR1 at 1 h after treatment in both SKOV3-NV and SKOV3-C1 cells (Fig. 4c-d), which is in agreement with a previous study using the same dose and time point [18]. These data indicate that HE4-overexpressing cells are capable of upregulating EGR1, but some mechanism is activated to suppress its later upregulation in response to cisplatin. Activation of ERK in response to cisplatin and rEGF was comparable between SKOV3-NV and SKOV3-C1 cells (Fig. 4c and e).

HE4 promotes deregulation of mitogen activated protein kinase (MAPK) signaling

Because MAPKs transcriptionally activate EGR1 expression by phosphorylating the EGR1 transcription factor ELK-1 [19–21], we assessed levels of phosphorylated extracellular signal-regulated kinase (ERK) in SKOV3-NV, SKOV3-C1, SKOV3-C7, OVCAR8-NV, and OVCAR8-C5 cell lines, and observed that levels of phospho-ERK were higher in SKOV3 and OVCAR8 HE4-overexpressing clones than in SKOV3-NV cells (Fig. 5a-b). In addition, when SKOV3 wild type (WT) and OVCAR8-WT cells were treated with 20 nM human recombinant HE4 (rHE4) protein, levels of phospho-ERK were dramatically affected (Fig. 5c-d). In SKOV3-WT cells, phospho-ERK was almost completely obliterated as early as 1 h after treatment, but levels were restored to above baseline by 24 h. Conversely, phospho-ERK continuously increased from 1 h to 24 h in OVCAR8-WT cells. These differential results between SKOV3 and OVCAR8 cells suggest time-dependent and cell type-specific effects of HE4 on growth factor signaling, but highlight the fact that HE4 has a rapid and dramatic effect on kinase activation, which could in turn affect regulation of EGR1 in response to cisplatin.

Although we see a significant effect of rHE4 on phospho-ERK levels, we were unable to detect differences in ERK activation between SKOV3-NV and SKOV3-C1 in response to 6 h cisplatin treatment, as previously shown. Moreover, by 24 h (the timepoint where we see the greatest rise in EGR1), phospho-ERK levels are diminished in both SKOV3-NV and SKOV3-C1 cells (Fig. 4c and e). Since p38—another MAPK known to regulate EGR1 transcription [20]—is also a key player in promoting apoptosis in response to cisplatin, we examined levels of phospho-p38 in SKOV3-NV and SKOV3-C1 clones following treatment with 25 μM cisplatin for 1, 6, and 24 h. Interestingly, p38 was robustly activated at 24 h in SKOV3-NV cells following cisplatin treatment, but this response was blunted in SKOV3-C1 cells (Fig. 5e-f). We also observed similar results comparing p38 activation in SKOV3-NV, SKOV3-C1, and SKOV3-C7 cells using a higher dose of cisplatin (80 μM), with robust activation noted at 6 h only in SKOV3-NV cells (Additional file 5A and B). We did observe a small suppression of ERK activation beginning at 1 h in HE4-overexpressing SKOV3 clones treated with this high dose of cisplatin, which was most apparent at 6 h after treatment, suggesting there may be subtle differences in how ERK is activated in HE4-overexpressing clones that only become apparent with high dose treatment (Additional file 5A and C).

HE4 modulates tubulin levels and MAPT gene expression

One of the genes we found to be differentially expressed between SKOV3-NV and SKOV3-C1 (vehicle treated) in our microarray analysis was microtubule associated protein tau (MAPT). MAPT, which is associated with paclitaxel resistance in several cancers including EOC [22–28], was more highly expressed in SKOV3-C1 cells than in SKOV3-NV cells (2.47-fold, p = .00702) (Additional file 2). We validated these results by qPCR, finding 4.42-fold higher levels in SKOV3-C1 than SKOV3-NV (p = .00446) (Fig. 6a). Since higher levels of MAPT together with β-tubulin correlate with paclitaxel resistance in gastric cancer [23], we sought to determine if HE4 affects β-tubulin levels and microtubule stability. Treatment of SKOV3-WT and OVCAR8-WT cells with 20 nM rHE4 produced a continuous increase in β-tubulin levels between 1 h and 48 h after treatment (Fig. 6b and c). Levels of α-tubulin also increased over time, similar to what was seen with β-tubulin (Fig. 6b and d). Collectively, these data suggest that HE4 may in part promote paclitaxel resistance by affecting microtubule levels or stability.