Inhibitors of phosphodiesterase 1 enhance paclitaxel cytotoxicity in a MDA-MB-231-derived cell line by promoting microtubule stabilization.

1
Introduction
Paclitaxel has been widely used as a therapeutic agent for many solid tumors, acting as a cytotoxic microtubule (MT)-targeting agent. Through its actions on MTs, paclitaxel alters multiple cellular and physiological oncogenic processes, including mitosis, angiogenesis, apoptosis, and inflammatory response [1]. However, resistance to paclitaxel occurs frequently in tumor cells via a wide variety of mechanisms, and it remains a major clinical challenge to overcome this resistance. To better understand how this obstacle may be overcome, it would be beneficial to elucidate different mechanisms by which cells can exhibit paclitaxel resistance.
Paclitaxel promotes MT polymerization and stabilization in living cells [2]. At clinically relevant concentrations, the drug suppresses the rates of MT growth and shrinkage, arrests cells in mitosis, and induces multipolar spindle formation and multipolar cell division in cultured cells. Importantly, a majority of mitotic cells in patient tumors treated with paclitaxel also exhibit multipolar spindles and multipolar division [3]. Moreover, it is known that paclitaxel-mediated cell death in patient tumors is typically due to chromosome missegregation on abnormal mitotic spindles [4]. Thus, augmentation of the MT-stabilizing and anti-mitotic effects of paclitaxel may enhance its cytotoxicity and efficacy.
To study the underlying mechanism of paclitaxel resistance, we have established a paclitaxel-resistant cell clone T50RN from the human triple-negative breast cancer (TNBC) cell line MDA-MD-231 and found that phosphodiesterase 1C (PDE1C) was substantially upregulated in this clone [5]. Phosphodiesterases (PDEs) are a class of enzymes responsible for the hydrolysis of the intracellular second messenger cyclic adenosine monophosphate (cAMP) and/or cyclic guanosine monophosphate (cGMP). These second messengers are involved in the regulation of many physiological processes, including cell proliferation and differentiation, cell cycle regulation, gene expression, inflammation, apoptosis, and metabolic function [6]. It is known that different PDEs are expressed by diverse cell types, and the PDEs regulate cellular functions via cAMP or cGMP [7]. Furthermore, modulation of PDE expression can interfere with crucial cAMP/cGMP signaling pathways and has been shown to correlate with tumorigenesis [6,8]. Among the eleven constituents within the PDE superfamily, PDE1 comprises three genes, PDE1A, PDE1B, and PDE1C, which undergo alternative splicing to generate functionally distinct isoforms with diverse catalytic and regulatory properties. It is well established that PDE1 relies on the binding of calcium/calmodulin for activation and catalyzes both cAMP and cGMP degradation [9]. Inhibition of PDE1 therefore leads to an increase in the intracellular levels of these two cyclic nucleotides, enhancing the activation of cAMP- and cGMP-dependent protein kinases. Subsequently, the activated protein kinases phosphorylate target signaling proteins and transcription factors to orchestrate a diverse array of physiological processes. Recent studies have indicated that excessive activation of PDE1 results in lower levels of cAMP and cGMP within tumor cells [[10], [11], [12]]. In addition, PDE1C has been shown to drive cell proliferation, migration and invasion in glioblastoma multiforme cells [11]. Several natural compounds, such as vinpocetine, curcumin and thymoquinone, have been shown to induce anti-proliferative activities in several cancer cell lines via targeting of PDE1 [11,13,14]. These studies indicate that PDE1 expression may be elevated in cancer cells and that PDE1 may be a promising target for cancer therapy.
TNBC is a subtype of breast cancer with poor prognosis. Due to a lack of therapeutic targets for the disease, treatment options for patients with TNBC are limited. Currently, paclitaxel-based drugs combined with other chemotherapeutics remain the first-line treatment for TNBC [15]. However, chemotherapeutic resistance often develops in the tumors, resulting in chemo-refractory disease and metastasis, with associated high relapse rates and poor survival. Since our previous study showed that PDE1C was highly expressed in T50RN cells [5] and that the selective PDE1 inhibitor ITI-214 has been shown to promote antitumor immunity and tumor growth inhibition in a mouse model of breast cancer [16], in this study, we explored whether and how PDE1C might be involved in paclitaxel resistance. Our in vitro findings indicate that inhibition of PDE1 in T50RN cells enhanced paclitaxel sensitivity by further increasing the MT-stabilizing effect of paclitaxel. This enhancement of MT stabilization possibly occurred via cAMP/EPAC signaling, suggesting that PDE1 is involved in paclitaxel resistance in T50RN cells and that inhibition of PDE1 or activation of cAMP/EPAC could be a strategy to overcome paclitaxel resistance. These results could be valuable for future preclinical in vivo researches.

2
Materials and methods
2.1
Cell culture and chemicals
MDA-MB-231 cells were purchased from American Type Culture Collection and maintained as previously described [17]. A Taxol-resistant cell line, T50RN, was established according to a previous protocol [17]. Briefly, MDA-MB-231 cells were exposed to stepwise escalating concentrations (0.5 to 50 nM) of paclitaxel (Merck, Darmstadt, Germany). The paclitaxel-exposed MDA-MB-231 cells were grown to sub-confluency and then sub-cultured for the next round of paclitaxel exposure, until the paclitaxel concentration reached 50 nM. A resistant T50RN cell line was obtained, expanded. This line could be maintained in culture medium without paclitaxel but also showed stable resistance. Small molecules used in this study included forskolin (Sigma, Saint Louis, MO, USA), ITI214, PF04822163, HA15, GSK2606414, 8‑bromo-cyclic-AMP (8-Br-cAMP), 8‑bromo-cyclic-GMP (8-Br-cGMP), and 8-CPT-2-methyl cAMP (EPAC agonist) (Cayman Chemical, Ann Arbor, MI, USA).

2.2
RNA sequencing and gene expression quantification
RNA-seq analysis was performed by the High Throughput Sequencing Core hosted in the Biodiversity Research Center at Academia Sinica (supported by grant number AS-CFII-108–114). Total RNA was extracted from MDA-MB-231 and T50RN cells with Trizol (Thermo Fisher Scientific) following the manufacturer’s protocol. RNA extraction was followed by whole transcriptome sequencing. RNA from each cell line was extracted and analyzed in independent duplicate experiments (designated as rep1 and rep2). The quality and integrity of RNA were evaluated by the Qubit RNA high sensitivity assay and the fragment analyzer RNA kits (Agilent Technologies). The sequencing library was prepared with the SMARTer stranded total RNA-Seq kit (Clontech) if RNA samples had distinct 18S and 28S ribosomal RNA peaks. The RNA libraries were sequenced on the Illumina NextSeq 2000 system (Illumina) with a per-sample target of about 20 million 100–base pair single-end reads. Low-quality bases and reads were removed by using Trimmomatic v0.39. The processed reads were mapped and aligned to the human reference genome (GRCh38) using STAR v2.7.10b The expression abundance (read counts) of each gene was estimated by using featureCounts v2.0.1; multi-mapping and chimeric reads were not counted. Genes with at least ten reads in two replicates were considered to be expressed and examined in the differential expression analysis. Cross-sample normalization was performed, and differentially expressed genes were identified using the R package DESeq2. Genes with an adjusted p-value <0.05 and a two-fold change in expression between MDA-MB-231 and T50RN cells were considered differentially expressed.

2.3
Cytotoxicity
Cytotoxicity was assessed with the trypan blue exclusion assay for cell viability [17] as previously described. Logarithmically growing cells on plates were treated for 72 h. After the treatment, cells were collected and stained with trypan blue (0.4 % w/v). The total cell number and blue dead cells were counted on a hemocytometer. The cell viability was calculated as the number of viable cells divided by the total cell number.

2.4
Detection of apoptosis
Apoptosis was detected with an annexin V-FITC/PI staining kit (BD Bioscience, New Jersey, USA) as described previously [18]. Briefly, 2  ×  105 cells were seeded in 60-mm dishes, followed by drug treatment for 72 h. Cells were then collected and washed with PBS. The second wash step was carried out with 1 × binding buffer (provided with the kit). The cells were rinsed, resuspended at 5 × 105 cells per 400 μl binding buffer, and incubated with 2 μl annexin V-FITC and propidium iodide for 15 min in the dark at room temperature. After washing with 1 × binding buffer, the cells were analyzed by flow cytometry (Attune NxT, Thermo Fisher Scientific).

2.5
Depletion of cellular PDE1C
Depletion of PDE1C was achieved by transducing T50RN cells with VSV-G-pseudotyped lentivirus-based short hairpin RNA (shRNA), as previously described [17]. The shRNAs targeting PDE1C (#1. TRCN48759 and #2. TRCN48760) were purchased from the National RNAi Core Facility (Genomic Research Center, Academia Sinica). Cells were transduced with pLKO.1- or shRNA-containing supernatants in growth medium supplemented with 10 μg/ml polybrene. At 24 h post-transduction, 2 μg/ml puromycin was added to culture medium to select for stable clones. Alternatively, cells were transduced for at least two days before depletion efficiency was verified by immunoblotting or the viability assay was conducted.

2.6
Immunofluorescence staining and analysis of mitotic spindles abnormalities
Cells seeded on glass coverslips were treated and fixed. Then, immunofluorescence staining of mitotic spindles was performed as previously described [19]. Primary antibodies included anti-α-tubulin (T5168, Sigma or GTX112141, GeneTex) and anti-γ-tubulin (T6557 or T3559, Sigma). Secondary antibodies, Alexa-Fluor 488- or 633-conjugated goat anti-mouse or anti-rabbit IgG, were purchased from Invitrogen (Carlsbad, CA, USA). Images of the immunostained samples were obtained with a confocal microscope (Zeiss LSM 980, Oberkochen, Germany). The numbers of cells with indicated spindle types were counted using a Zeiss Axioplan 2 Imaging MOT fluorescence microscope.

2.7
Analysis of cell cycle progression and DNA replication in proliferating cells
Cell cycle progression was monitored using DNA flow cytometry (Attune NxT, Thermo Fisher Scientific), as previously described [17]. The percentage of cells in mitosis was measured by flow cytometry analysis of phospho-histone H3 (p-H3, the mitosis marker)-positive cells. DNA replication in proliferating cells was measured using Click-iT™ EdU Pacific Blue™ Flow Cytometry Assay Kit (ThermoFisher Sceintific). After treatment, the cells were incubated with 10 μM 5-ethynyl 2´-deoxyuridine (EdU) for 2 h before being collected for fixation and permeabilization (reagents provided by the kit). Then, the Click-iT® reaction was initiated between EdU-alkyne and pacific blue-azide. DNA was counterstained with propidium iodide. The cell cycle distribution and the percentage of cells with p-H3 or EdU-positive cells were measured by flow cytometry analysis and quantified using software from Thermo Fisher Scientific.

2.8
Immunoblotting
Cell lysis and immunoblotting were carried out as described [19]. Specific proteins were detected using antibodies against caspase-9, phospho-eIF2α at S51, cleaved PARP, phospho-PERK at Thr981, PDE1C, and acetylated tubulin (Cell Signaling Technology, Danvers, MA), GADD153 and α-tubulin (GeneTex), and de-tyrosinated tubulin (Millipore, CA, USA). GAPDH or PCNA loading controls were respectively detected with anti-GAPDH (Genetex) or anti-PCNA (Santa Cruz Biotechnology, Inc., Dallas, Texas). Quantification of proteins bands were measured using GeneTools software (Syngene, Cambridge CB4 1TF, UK). The intensity of each protein band was normalized to the intensity of the GAPDH band and then compared with that of the untreated group.

2.9
Quantification of MT polymerization
MT polymerization was measured in terms of MT polymer mass, using the “Tubeness” plugin in ImageJ software as previously described [20]. Polymerized tubulin (revealed by immunofluorescence staining of α-tubulin) was first identified and quantified using the “Tubeness” plugin on a 3D image stack. The total cellular area was determined by the “Contour” plugin. The ratio of polymerized tubulin mass to cellular area was taken as a measure of MT polymerization.

2.10
Measurement of cAMP and cGMP
Intracellular cAMP or cGMP levels were measured using the cAMP or cGMP Competitive ELISA kits (Cell Signaling Technology) according to the manufacturer’s protocols. Briefly, MDA-MB-231 or T50RN cells were untreated or treated with 20 μM forskolin or 10 μM ITI214 for 24 h. Then, 3 × 106 cells were lysed with cell lysis buffer (provided by the kit). Free cAMP or cGMP in cell lysates and a fixed amount of HRP-labeled cAMP or cGMP were loaded onto wells coated with anti-cAMP or anti-cGMP antibodies. The cAMP or cGMP in cell lysates competed with the fixed amount of HRP-linked cAMP or cGMP for binding to anti-cAMP or anti-cGMP antibodies on the plate. Following washing to remove excess sample cAMP/cGMP and HRP-linked cAMP/cGMP, the amount of bound HRP-labeled cAMP/cGMP was measured using a fluorometric HRP substrates. The magnitude of signal was inversely proportional to the quantity of cAMP/cGMP in the sample. The level of signal was quantified by comparison to a cAMP/cGMP standard curve, which was run at the same time as the tested samples.

2.11
Gene set enrichment analysis (GSEA)
GSEA was performed to identify significantly enriched gene sets associated with PDE1C expression in TNBC samples. Batch-effect-normalized mRNA expression data from 180 TNBC samples were obtained from the TCGA Pan-Cancer (PANCAN) dataset. Samples were stratified into two groups based on PDE1C expression levels: the top quartile (top 25 %, high PDE1C expression) and the bottom quartile (bottom 25 %, low PDE1C expression). GSEA was conducted using software developed by UC San Diego and the Broad Institute (https://www.gsea-msigdb.org/gsea/index.jsp). Gene sets used for enrichment analysis included the curated Canonical Pathways (C2) and Ontology (C5) gene sets available in the Molecular Signatures Database (MSigDB)(https://www.gsea-msigdb.org/gsea/msigdb).

2.12
Analysis of mitochondrial membrane potential and reactive oxygen species (ROS)
The fluorescent dye tetramethylrhodamine ethyl ester (TMRE) was used to measure mitochondrial membrane potential. In brief, T50RN cells were treated with PDE1 inhibitors with or without paclitaxel for 24 h. Then, the treated T50RN cells were collected and incubated in medium containing 50 nM TMRE for 5 min in the dark at room temperature. After thorough rinsing with PBS, the cells were immediately analyzed with a flow cytometer (Attune NxT, Thermo Fisher Scientific). ROS generation was measured with CellROX™ Green or MitoSox Red Reagents (Thermo Fisher Scientific). After treating T50RN cells with PDE1 inhibitors with or without paclitaxel for 24 h, the cells were collected and stained with 0.5 μM CellROX Green or 2 μM MitoSox Red for 30 min at 37 °C. After thorough rinsing, the cells were analyzed with flow cytometry (Attune NxT).

2.13
Detection of mitochondrial Ca2+
Mitochondrial Ca2+was measured by Rhod-2 staining [21]. Cells were preloaded with Rhod-2 (10 μM) for 45 min and then treated with PDE1 inhibitors in the presence or absence of paclitaxel for 24 h. Rhod-2 intensities were measured by flow cytometry (Attune NxT).

2.14
Statistical analysis
The results are shown as mean ± standard deviation (SD). All experiments were performed at least three times. The treated cell populations were compared with vehicle controls in each experiment. The differences between groups were analyzed by Student’s t-test or by two-way analysis of variance (two-way ANOVA) and Tukey's multiple comparison test using GraphPad Prism 10.1. Statistical significance was set as p < 0.05.

3
Results
3.1
PDE1 inhibitors enhance paclitaxel cytotoxicity in T50RN cells
T50RN cells, a clonal line established from the human TNBC cell line MDA-MB-231, were significantly more resistant to paclitaxel than the parental MDA-MD-231 cells [5]. The IC50 to paclitaxel for parental MDA-MB-231 cells is 2.2 nM (calculated with Prism 10.1) and 107 nM for T50RN cells (Fig. 1A-a, compare MDA-MB-231-pLKO.1 to T50RN-pLKO.1). Our previous report has demonstrated that both the mRNA and protein expression of PDE1C was considerably elevated in T50RN cells compared to parental MDA-MB-231 cells [5]. We therefore assessed the mRNA expression level of all PDE members in this study. The expression levels of PDE4D and PDE8A were relatively high, and expression levels of the other PDEs were relatively low in parental MDA-MB-231 cells (Fig. 2B). In T50RN cells, the expression of PDE1C, PDE4D and PDE8A were high, and the other PDEs were relatively low. Of note, PDE1C was the only PDE with significant upregulation in T50RN cells (Fig. 1B, log2[fold change] = 3.82, p = 3.97E-251), as compared to the parental MDA-MB-231 cells. We thus focused on the role of PDE1C in paclitaxel resistance of T50RN cells.
The specific effects of PDE1C on paclitaxel cytotoxicity were assessed by performing experiments with shRNA-mediated depletion of PDE1C in T50RN cells. Fig. 1A (b) shows that the PDE1C protein expression was very limited in MDA-MB-231 cells and was considerably upregulated in T50RN cells. Using two specific shRNAs (#1 and #2), PDE1C was knocked down, which substantially decreased its protein expression levels in T50RN cells (Fig. 1A-b). The results indicated a modest upregulation of PDE1C protein levels in T50RN cells. This observation may be attributed to limited sensitivity or specificity of the primary antibody we applied or protein loss during the extraction process. Nonetheless, depletion of PDE1C could significantly enhance paclitaxel cytotoxicity in T50RN cells but had no effect on MDA-MB-231 cells (Fig. 1A-a), indicating that PDE1C might be involved in T50RN cell resistance to paclitaxel. We also ectopically overexpressed PDE1C in MDA-MB-231 cells (Fig. 1A-c). This overexpression could significantly increase the survival of paclitaxel-treated MDA-MB-231 cells (Fig. 1A-c), confirming that PDE1C may play a role in paclitaxel resistance. Since PDE1C was the only PDE1 upregulated in T50RN cells and the expression levels of PDE1A and PDE1B were very low (Fig. 1B), we next treated cells with PDE1-specific inhibitors to test whether PDE1C could modulate paclitaxel resistance. The inhibitors were first empirically tested to determine appropriate subtoxic treatment concentrations, which yielded 80–90 % viability by trypan blue exclusion assay. Treatment of PDE1-specific inhibitors, ITI214 and PF04822163, at sublethal concentrations showed slight additive effects on paclitaxel cytotoxicity in MDA-MB-231 cells (Fig. 1C). However, the inhibitors significantly enhanced paclitaxel cytotoxicity in T50RN cells (Fig. 1D), decreasing the IC50 of paclitaxel from ∼100 nM to <25 nM. Nevertheless, the decreased IC50 with PDE1 inhibitor treatment in T50RN cells did not reach the same IC50 level as seen in parental MDA-MB-231 cells, suggesting that other factors may be involved in mediating T50RN paclitaxel resistance.
We next examined the effects of PDE1 inhibitors on paclitaxel-induced spindle abnormalities. PDE1 inhibitors alone had no significant effects on the distribution of spindle patterns in both MDA-MB-231 and T50RN cells (Fig. 2B) and had no effect on paclitaxel-induced spindle abnormalities in MDA-MB-231 cells (Fig. 2B, left). However, the frequency of paclitaxel-induced spindle abnormalities was significantly elevated by cotreatment of T50RN cells with paclitaxel and either of the PDE1 inhibitors (Fig. 2A and B, right). The effects of PDE1 inhibitors on paclitaxel-induced cell cycle arrest in T50RN cells were also analyzed. The results showed that PDE1 inhibitors alone at the concentration applied had no apparent effect on cell cycle progression (Fig. 2C) and EdU incorporation in T50RN cells (Fig. 2D). Paclitaxel at 50 nM significantly reduced G1 and increase sub-G1 cell population (Fig. 2C). Combination of PDE1 inhibitor with paclitaxel further reduced G1, enhanced sub-G1 and increased G2/M cell population (Fig. 2C). In addition, 50 nM paclitaxel slightly reduced EdU incorporation in T50RN cells (Fig. 2D). PDE1 inhibitors could further reduce the incorporation of EdU in S phase cells in 50 nM paclitaxel-treated cells (Fig. 2D). These results suggest that PDE1 inhibition in paclitaxel-treated T50RN cells could block cell cycle progression and impair DNA replication. The effects of PDE1 inhibitors were also examined in terms of apoptosis induction in T50RN cells. The results showed that ITI214 or PF04822163 treatment alone caused <10 % apoptosis in T50RN cells, but the combination of paclitaxel and ITI214 or PF04822163 considerably enhanced T50RN cell apoptosis (Fig. 2E). Immunoblotting analysis also showed that levels of cleaved caspase-9 and PARP were significantly increased by the combined treatment of paclitaxel and ITI214 or PF04822163 (Fig. 2F). These results indicate that inhibition of PDE1 could augment paclitaxel cytotoxicity in T50RN cells. In summary, our results to this point showed that inhibition of PDE1 can sensitize T50RN cells to paclitaxel and suggested that PDE1 may be involved in paclitaxel resistance of T50RN cells.

3.2
cAMP and EPAC are involved in PDE1 inhibition-induced paclitaxel sensitization
PDE1 is responsible for the degradation of both cAMP and cGMP. To better understand the roles of downstream effector(s) of PDE1 in paclitaxel resistance of T50RN cells, the effects of cAMP and cGMP on paclitaxel cytotoxicity were examined. The appropriate concentrations of the following used inhibitors or modulators were empirically determined, yielding 80–90 % cell viability by trypan blue exclusion assay. Treatments with forskolin (adenylate cyclase activator that increases intracellular cAMP level [22]) or 8-Br-cAMP (cell permeable cAMP analog) but not 8-Br-cGMP (cell permeable cGMP analog) significantly enhanced paclitaxel cytotoxicity (Fig. 3A). Forskolin and 8-Br-cAMP also enhanced paclitaxel-induced spindle abnormalities in T50RN cells (Fig. 3B), confirming the critical role of cAMP in this process. In addition, SQ22536 (adenylate cyclase inhibitor) had no effect on paclitaxel cytotoxicity (Fig. 3C, 400 μM SQ22536, grey bar) but could partially rescue the enhancing effect of PDE1 inhibitor ITI214 on paclitaxel cytotoxicity (Fig. 3C, ITI vs ITI+SQ, grey bar). These results suggest that cAMP but not cGMP might be the critical signaling factor to induce T50RN cell sensitization to paclitaxel. In addition, the intracellular level of cAMP was significantly lower in T50RN cells than in parental MDA-MB-231 cells (Fig. 3D, lower), suggesting that the cAMP-degrading function was enhanced in PDE1C-overexpressing T50RN cells. Treatment of the T50RN cells with ITI214, PF04822163, 8-Br-cAMP, or forskolin significantly increased the cAMP levels in (Fig. 3D, lower). The intracellular cGMP level showed no significant changes after the modulator treatment (Fig. 3D, upper), probably since there was very low expression of cGMP-generating machinery in T50RN cells. The low cGMP level also suggests that cAMP, but not cGMP, may play a critical role in paclitaxel resistance in T50RN cells.
Since PDE4D expression was as high as PDE1C in T50RN cells (Fig. 1B) and PDE4 specifically catalyzes and degrades cAMP, the role of cAMP on paclitaxel cytotoxicity was also verified by treating T50RN cells with the PDE4-specific inhibitor, piclamilast. The result showed that treatment of piclamilast at a sublethal concentration could considerably enhance paclitaxel cytotoxicity (Fig. 3A) and spindle abnormalities (Fig. 3A). However, the combination of PDE1 and PDE4 inhibitors did not further enhance paclitaxel cytotoxicity (Fig. 3C, ITI vs ITI+Pic, grey bar), suggesting that PDE1 and PDE4 might catalyze the same pool of cAMP. These results indicated that inhibition of PDE1 or activation of adenylate cyclase increased intracellular cAMP levels and concurrently enhanced paclitaxel cytotoxicity in T50RN cells. Thus, the results suggest that cAMP, but probably not cGMP, may play a pivotal role in sensitizing T50RN cells to paclitaxel. cAMP functions by binding to cAMP dependent protein kinases (PKAs) and/or cAMP-activated guanine exchange factors (EPACs) [23]. Thus, PKA and EPAC modulators were examined for their roles in cAMP-downstream effects leading to paclitaxel sensitization. The EPAC agonist 8-pCPT-2′-O-Me-cAMP significantly enhanced paclitaxel cytotoxicity (Fig. 3A) and spindle abnormalities (Fig. 3B). In contrast, the EPAC inhibitor, ESI-09, could protect T50RN cells from paclitaxel cytotoxicity (Fig. 3A). These results indicate that PDE1 inhibition might trigger cAMP/EPAC signaling to enhance paclitaxel cytotoxicity in T50RN cells. However, the PKA-selective inhibitor KT5720 also considerably enhanced paclitaxel cytotoxicity (Fig. 3A) and spindle abnormalities (Fig. 3B). Thus, PKA activation might protect cells from paclitaxel-induced damage and hamper the sensitization effect.

3.3
cAMP/EPAC enhances paclitaxel-induced MT stabilization
Paclitaxel induces MT stabilization to disrupt spindle assembly and induce cell death [1]. We therefore examined whether cAMP/EPAC could alter paclitaxel-induced MT stability in T50RN cells. Tubulin acetylation occurs preferentially on polymeric tubulins [24] and reduces the conformational flexibility of the acetylation loop [25]. Upon depolymerization, the detyrosinated tubulin is quickly retyrosinated [26], and this detyrosinated-tubulin associates with stable cellular MTs [27]. Thus, tubulin acetylation and detyrosination are often associated with MT stability. The results in Fig. 4A show that PDE1 inhibitors, ITI214 and PF04822163, could enhance the expression levels of acetylated- and detyrosinated-tubulin in paclitaxel-treated T50RN cells. This finding indicates that MTs in T50RN cells might be more stabilized after cotreatment with paclitaxel and ITI214 or PF04822163 than after paclitaxel treatment alone. MT stability was also examined by analyzing the MT polymer mass, according to immunofluorescence staining of α-tubulin. The results showed that paclitaxel alone significantly increased MT polymer mass in T50RN cells (Fig. 4B right, T), and ITI214 or PF04822163 alone slightly but significantly increased the MT polymer mass (Fig. 4B right, ITI and PF). Importantly, cotreatment of ITI214 or PF04822163 with paclitaxel markedly enhanced the MT polymer-stabilizing effect (Fig. 4B right, T+ITI and T+PF). These results indicate that inhibition of PDE1 may enhance MT stabilization in paclitaxel-treated T50RN cells. In addition, forskolin and the EPAC agonist (CPT-methy-cAMP) further enhanced the expression of acetylated- and detyrosinated-tubulin in paclitaxel-treated T50RN cells (Fig. 4C). The MT polymer mass in paclitaxel-treated T50RN cells was also significantly enhanced by cotreatment with forskolin or the EPAC activator (Fig. 4D; T+FK and T+EPACa). These results indicate that PDE1 inhibition may trigger cAMP/EPAC to further enhance MT stability in paclitaxel-treated cells. Since paclitaxel induces cytotoxicity by stabilizing MTs, PDE1 inhibition might augment paclitaxel cytotoxicity by further enhancing MT stabilization in T50RN cells.

3.4
PDE1 inhibition induces mitochondrial dysfunction and ER stress in paclitaxel-treated T50RN cells
We next examined how PDE1 inhibition might enhance paclitaxel-induced cell death. To understand the downstream effects of inhibition of PDE1C, we assessed gene expression in TNBC samples (total 180 patients) with PDE1C mRNA expression levels in the highest quartile (top 25 %) and the lowest quartile (bottom 25 %). The mRNA expression dataset was retrieved from the TCGA database and evaluated with Gene Set Enrichment Analysis (GSEA). The results showed no significant enrichment in gene sets for tumors in the top 25 % expression level group. However, the GSEA analysis of TNBC samples with the lowest 25 % of PDE1C expression showed several enriched pathways. Notably, five gene sets among the top 10 enriched ontology or pathway gene sets were mitochondrial- or electron transport chain-related genes (Fig. 5A and B, arrow heads), while others were lactate metabolism- or protein modification-related gene sets. Since lactate metabolism and protein modifications are closely related to ER function [28,29], these results suggest that mitochondrial and/or ER function might be altered in TNBC samples with low PDE1C expression.
We next assessed whether ER stress was induced in the treated cells. Markers of ER stress, phospho-PERK at T981, phospho-eIF2α at Ser51 and GADD153 were all elevated in T50RN cells cotreated with paclitaxel and ITI214 or PF04822163 (Fig. 5C and D), as compared to each treatment alone. These results indicated that cells receiving combined treatment were under ER stress. We then used modulators of ER stress, HA15 (GRP78 inhibitor), GSK2606414 (PERK activation inhibitor), or Sal003 (eIF2α dephosphorylation inhibitor) to assess the possible role of ER stress on the PDE1 inhibition-mediated enhancement of paclitaxel cytotoxicity. The results showed that treatments of these inhibitors at sublethal concentrations had little effect on the cytotoxicity induced by ITI214 or PF04822163 (Fig. 5E). These inhibitors, like PDE1 inhibitors ITI214 and PF04822163, also significantly enhanced paclitaxel cytotoxicity (Fig. 5E), indicating that GRP78-mediated or PERK-eIF2α-mediated pathways could protect paclitaxel-induced cell death. Interestingly, HA15 and GSK2606414 slightly but significantly protected T50RN cells from the cytotoxicity-enhancing effects of PDE1 inhibitors on paclitaxel (Fig. 5E), indicating that GRP78- or PERK-mediated pathways might increase PDE1 inhibitor-mediated enhancement of paclitaxel-induced cell death. Sal003 slightly enhanced PDE1 inhibitor-mediated increase in paclitaxel cytotoxicity (Fig. 5E). This result implied that persistent eIF2α activation might enhance cell death induced by cotreatment of paclitaxel plus PDE1 inhibitors. These results suggest that activation of PERK-eIF2α might mediate the enhancing effects of PDE1 inhibitors on paclitaxel-induced cytotoxicity.
Since mitochondria interact closely with the ER and play critical roles in apoptosis induction, we next tested the effects of PDE1 inhibitors on mitochondrial function. The data in Fig. 6A show that inhibition of PDE1 in T50RN cells treated with ITI214 or PF04822163 either with or without paclitaxel cotreatment induced higher TMRE intensities, indicating that PDE1 inhibition may induce hyperpolarization of mitochondria. The levels of mitochondrial superoxide (Fig. 6B) and intracellular peroxide (Fig. 6C) were also significantly increased by treatment with ITI214 or PF04822163 in T50RN cells, with or without paclitaxel treatment. In addition, mitochondrial calcium (as measured by the mitochondria-specific calcium dye Rhod-2) was also significantly increased by treatment with ITI214 or PF04822163 in T50RN cells with or without paclitaxel treatment. (Fig. 6D). Taken together, these results suggest that PDE1 inhibition in T50RN cells can induce mitochondrial dysfunction, calcium overload and excess ROS generation.

4
Discussion
Anticancer drug resistance remains a major obstacle to curing cancer. In this study, our in vitro data showed that elevated expression of PDE1C plays a critical role in conferring paclitaxel resistance to a human TNBC cell line-derived cell clone, T50RN. We found that inhibition of PDE1 increased cAMP, activated EPAC, and augmented MT stabilization in paclitaxel-treated T50RN cells, sensitizing cells to paclitaxel. Since paclitaxel stabilizes MTs to induce spindle abnormalities and cytotoxicity, elevated expression of PDE1C in T50RN cells may lower the levels of cAMP and counteract the MT-stabilizing effect of paclitaxel, thereby causing drug resistance in the cells. Since high levels of intracellular cAMP have been demonstrated to inhibit tumor growth or sensitize cancer cells to anticancer drugs, serving as a potential protective mechanism against tumor progression [[30], [31], [32]], our results suggest that the use of specific PDE inhibitors may be one of the viable approaches for overcoming paclitaxel resistance.
T50RN paclitaxel resistant cell was cloned by continuously culturing MDA-MB-231 cells in stepwise escalating concentration of paclitaxel and then cloned and expanded. Thus, PDE1C upregulation might be unique in T50RN. However, the PDE1C has been shown to drive cell proliferation, migration and invasion in glioblastoma multiforme, indicating that PDE1C plays a critical role in tumor progression [11]. In addition, the emergence of paclitaxel resistance remains a persistent challenge in cancer therapeutics and continues to limit clinical efficacy. Our study identifies PDE1C as a possible mediator of paclitaxel resistance, providing new insights into the alternative mechanisms governing cellular drug response.
Our results showed that PDE1C expression was low in parental MDA-MB-231 cells but significantly upregulated in the drug-resistant T50RN cells. How PDE1C upregulation is induced in T50RN cells is currently unclear. PDE1 isoforms are differently distributed throughout the body and may have different functional roles in cardiovascular, nervous, reproductive and immune systems [33]. As such, it remains unclear whether PDE1C-overexpressing cells are present as a pre-existing subset of the MDA-MB-231 cell population, or if PDE1C expression is induced by paclitaxel treatment. This question warrants further investigation in future work. Nonetheless, the cAMP level was significantly lower in T50RN cells as compared to parental MDA-MB-231 cells, and it could be increased by treating T50RN cells with PDE1 inhibitors. These results confirmed that the elevated expression of PDE1 led to functional effects in T50RN cells. However, the cGMP level remained low in MDA-MB-231 and T50RN cells, either untreated or treated with PDE1 inhibitor. One possible reason for this difference might be that the machinery for cGMP generation was not actively expressed in the parental or resistant cells, so inhibition of cGMP destruction could not induce an increase. Thus, we conclude that cGMP is not likely involved in PDE1C-mediated paclitaxel resistance. Our finding that the cGMP analog had no effect on paclitaxel cytotoxicity was also consistent with this notion. Together, our results confirmed that PDE1C was upregulated and cAMP level was decreased in T50RN cells and that PDE1C and cAMP may play critical roles in paclitaxel resistance.
Our findings indicate that the depletion or pharmacological inhibition of PDE1C enhances the sensitivity of T50RN cells to paclitaxel. However, the resulting IC50 remained higher than that observed in parental MDA-MB-231 cells. Given our previous evidence highlighting the roles of SYK and MDR1 in T50RN chemoresistance [5,34], these data suggest that paclitaxel resistance in this cell line is multifactorial, with PDE1C representing one of several contributing factors.
The two primary cAMP effectors, EPAC and PKA, can function either antagonistically, independently or synergistically to modulate cancer cell proliferation, apoptosis, migration, invasion, and drug resistance. Our results showed that like PDE1 inhibitors, forskolin, a cAMP analog, and an EPAC agonist could enhance paclitaxel cytotoxicity and spindle abnormalities in T50RN cells. The EPAC inhibitor could further protect T50RN cells from paclitaxel-induced cytotoxicity, indicating that PDE1 inhibition-induced increases in cAMP might trigger EPAC signaling, which may then sensitize T50RN cells to paclitaxel. However, the PKA inhibitor further enhanced paclitaxel cytotoxicity, suggesting PKA activation might protect cells from paclitaxel and that EPAC and PKA might act independently or antagonistically.
It has been shown that MT-associated protein 2 binds, stabilizes, and regulates dynamics of MTs in a cAMP-dependent protein kinase-dependent manner [35]. cAMP-dependent signaling can increase tubulin post-translational modifications related to MT stability [36] and downregulate the MT-destabilizing activity [37]. cAMP signalings also play a crucial role of dynamic MTs in leading-edge protrusion for cell chemotaxis [38]. These studies reveal that cAMP signaling regulates MT dynamics and that inhibition of PDE1 may disrupt MT dynamics by increasing intracellular cAMP level. EPAC has been shown to associate with α/β tubulin dimers in cells and promote MT formation [39]. cAMP analog-induced EPAC-dependent signaling has also been shown to promote MT stability in neuroblastoma differentiation [36]. EPAC was also required for Golgi-derived MT formation [40]. These studies highlight the known role of EPAC in promoting MT formation and stabilization. Our results further showed that inhibition of PDE1 in T50RN cells may trigger cAMP-EPAC signaling to induce augmentation of paclitaxel-induced MT stabilization. Since paclitaxel induces toxicity by targeting and stabilizing MTs, this augmentation effect may consequently enhance T50RN cell sensitivity to paclitaxel.
MTs and the ER are highly interdependent structures [41,42]. As such, disruption of the MT network has been shown to alter ER structure and induce ER stress [41,43]. Our results also show that PDE1 inhibitors enhanced the MT-stabilizing effects of paclitaxel and concomitantly stimulated the accumulation of ER stress sensors and downstream mediators. Our results showed that HA15 and GSK2606414 could marginally protect from cytotoxicity, but Sal003 slightly enhanced cell death upon treatment of PDE1 inhibitor plus paclitaxel. HA15 binds and inhibits GRP78, thus preventing the association of GRP78 with its ER partners, PERK, IRE1α and ATF6 [44]. Meanwhile, GSK2606414 inhibits PERK phosphorylation and activation [45], and Sal003 is a selective inhibitor of eIF2α dephosphorylation that can stimulate eIF2α activation [46]. Notably, GSK2606414 has been shown to suppress PERK/eIF2α/CHOP pathway-mediated apoptosis [47] and Sal003 has been shown to exhibit anti-tumor effects and sensitizes resistant tumors to targeted therapy in a mouse breast cancer model [48]. Thus, our finding suggests that the PERK-eIF2α ER stress pathway might be involved in cytotoxicity caused by PDE1 inhibitor plus paclitaxel in our study setting.
The MT cytoskeleton is a critical signaling platform, and its disruption may impair cellular homeostasis and induce broad cellular stress. For instance, taxanes are known to induce ER stress [[49], [50], [51]]. It is therefore plausible that the enhanced MT stabilization induced by combined treatment with PDE1 inhibitor and paclitaxel would induce ER stress. It is known that activation of the unfolded protein response and ER stress can trigger changes in mitochondrial function, exacerbating ER stress [52]. It is also known that stress-induced disruption of ER homeostasis leads to alterations of mitochondrial function and apoptosis [53]. Our results showed that treatment of paclitaxel alone in T50RN cells might induce a protective ER stress and that PDE1 inhibition induced a concurrent mitochondrial hyperpolarization, ROS generation, and calcium overload, potentially changing the protective ER stress into a cell death-inducing response in paclitaxel-treated cells. Thus, it might be likely that the combined treatment in our study triggered enhanced MT stabilization, induced ER stress and mitochondria dysfunction, ultimately leading to apoptosis.
In conclusions, our findings in this study provide evidence that PDE1C upregulation may be involved in paclitaxel resistance in a human TNBC-derived cell clone, T50RN. Inhibition of PDE1 or activation of cAMP/EPAC may enhance paclitaxel cytotoxicity and overcome this resistance. These results could be valuable for future in vivo xenograft and/or preclinical researches.

Funding
This work was supported by Academia Sinica, Taiwan.

Ethics statement
N/A.

CRediT authorship contribution statement
Hsiao-Hui Kuo: Investigation, Data curation. Chien-Wei Huang: Investigation. Tsai-Ming Lu: Software, Investigation, Data curation. Wei-Rou Chiang: Investigation. Ling-Huei Yih: Writing – review & editing, Writing – original draft, Validation, Supervision, Project administration, Funding acquisition, Formal analysis, Data curation.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.