Abstract

   Acquired resistance to chemotherapy is one of the major causes of
   mortality in advanced nasopharyngeal carcinoma (NPC). However,
   effective strategies are limited and the underlying molecular
   mechanisms remain elusive. In this study, through transcriptomic
   profiling analysis of 23 tumor tissues, we found that NOTCH3 was
   aberrantly highly expressed in chemoresistance NPC patients, with
   NOTCH3 overexpression being positively associated with poor clinical
   outcome. Mechanistically, using an established NPC cellular model, we
   demonstrated that enhancer remodeling driven aberrant hyperactivation
   of NOTCH3 in chemoresistance NPC. We further showed that NOTCH3
   upregulates SLUG to induce chemo-resistance of NPC cells and higher
   expression of SLUG have poorer prognosis. Genetic or pharmacological
   perturbation of NOTCH3 conferred chemosensitivity of NPC in vitro and
   overexpression of NOTCH3 enhanced chemoresistance of NPC in vivo.
   Together, these data indicated that genome-wide enhancer reprogramming
   activates NOTCH3 to confer chemoresistance of NPC, suggesting that
   targeting NOTCH3 may provide a potential therapeutic strategy to
   effectively treat advanced chemoresistant NPC.

   Subject terms: Head and neck cancer, Cancer

Introduction

   Nasopharyngeal carcinoma (NPC) is a malignancy of epithelia origin with
   widespread prevalence in specific geographical distribution. It occurs
   in high frequencies in Southeast Asian countries and Southern China.
   Multiple factors contribute to its pathogenesis, including EBV
   infection, genetic predisposition and environmental factors [[50]1,
   [51]2]. Early-stage NPC is highly sensitive to radiotherapy and most of
   patients are curable. However, more than 70% newly diagnosed NPC are
   classified as advanced disease. Concurrent chemoradiotherapy is now the
   standard treatment for advanced stage NPC patients. Although high
   responsiveness is achieved initially, patients frequently develop
   acquired resistance to chemotherapy, resulting in local recurrence and
   distant metastasis, leading to high mortality [[52]3–[53]5]. Therefore,
   deciphering the resistance mechanisms of chemotherapy and exploring
   novel therapeutic strategies to overcome chemoresistance are imperative
   in the clinic.

   Aside from genetic alterations, epigenetic changes also play a
   significant role in tumorigenesis and chemoresistance [[54]6–[55]8].
   Epigenetic alterations include reversible modifications of DNA and
   histone proteins, such as DNA hypo/hypermethylation, histone
   modification, which can cause chromatin remodeling and altered gene
   expression. Genome-wide studies by the Cancer Genome Atlas project
   showed that NPC is more hypermethylated than 11 other cancer types
   [[56]9], and hypermethylation at the promoter region of tumor
   suppressors is frequently reported in NPC [[57]10–[58]12]. However,
   other types of epigenetic dysregulation, other than DNA methylations,
   which may lead to chemoresistance of NPC remain elusive. Enhancer,
   marked by acetylation of the lysine residue at position 27 of H3
   histone protein, is a short region of DNA which can be bound by
   transcription factors (TFs) to increase the transcription of an
   associated gene from far distance along the DNA. Epigenetic regulation
   of gene by enhancers is important for maintain cell identity.
   Dysregulation of enhancer has been found in many cancers, such as
   prostate cancer, lung cancer and leukemia [[59]13–[60]15] . One study
   has applied chromatin immunoprecipitation sequencing (ChIP-seq)
   analysis of NPC tissues and found that super-enhancer ETV6 correlated
   with prognosis of NPC patients [[61]16]. However, how the enhancer
   landscape evolves and regulates gene expression during chemo-resistance
   of NPC has yet to be determined.

   One of the important cellular pathways affected by epigenetic
   regulation is Notch signaling [[62]17]. Notch signaling pathway is
   highly conserved through evolution and plays an important role in cell
   proliferation, apoptosis, differentiation and maintenance of somatic
   stem cells. Deregulation of Notch pathway has been demonstrated related
   to cancer initiation, relapse, metastasis and chemoresistance
   [[63]18–[64]20]. Notch has four receptors (NOTCH1, NOTCH2, NOTCH3 and
   NOTCH4) and five ligands (DLL1, DLL3, DLL4, JAG1 and JAG2) in mammals,
   which are differentially expressed and performing different functions
   in various tissues and tumors [[65]21]. When the receptor of Notch
   binds to the ligand, it will be cleaved by γ-secretase and release the
   activated Notch intracellular domain (NICD) to the nucleus to activate
   downstream target genes. Among all four Notch receptors, NOTCH1 and
   NOTCH3 have been found highly expressed in NPC tissues compared with
   normal tissues [[66]22, [67]23]. Moreover, inhibition of NOTCH3
   increased the sensitivity of EBV related NPC cells to cisplatin. These
   points to the importance of the Notch receptors in NPC. However, the
   relevance and the regulation of NOTCH3 signaling in the chemoresistance
   of NPC have not been comprehensively investigated.

   In this study, using patient samples and cellular models, we
   demonstrated that genome-wide enhancer reprogramming activates NOTCH3
   signaling to confer chemo-resistance of NPC. Moreover, aberrant
   activation of NOTCH3-SLUG axis is correlated with chemo-resistance in
   NPC and poor prognosis of NPC patients, while targeting NOTCH3
   signaling significantly resensitized NPC cells to paclitaxel. These
   data provide a potential therapeutic strategy to treat chemoresistant
   NPC patients.

Results

NOTCH3 is aberrantly highly expressed in chemoresistance NPC patients and
associates with poor clinical outcome

   To investigate genes related to chemoresistance of advanced
   nasopharyngeal carcinoma, transcriptome analysis was performed on a
   total of 22 patients (Supplementary Table [68]S4) treated with
   chemotherapy, of which 8 patients showed complete response (CR) and 14
   patients showed partial response (PR). Totally 644 differentially
   expressed genes were found in PR versus CR (|log2 fold change| ≥ 1,
   adjusted p-value < 0.05) (Fig. [69]1A). Gene Set Enrichment Analysis
   (GESA) analysis revealed enriched Notch signaling pathway gene clusters
   in PR patients (Fig. [70]1B). The expression levels of all four Notch
   receptors in 40 NPC patient samples (Supplementary Table [71]S5) were
   then determined by quantitative real time PCR (qRT-PCR) and only NOTCH3
   was found significantly upregulated in PR patients compared with CR
   (Fig. [72]1C). Finally, the clinical relevance of NOTCH3 was assessed
   using immunohistochemistry (IHC) staining of NOTCH3 in 234 NPC patient
   samples with chemotherapy (Supplementary Table [73]S6-[74]7).
   Consistent with the reported localization of NOTCH3 [[75]24], NOTCH3
   protein was detected on the surface and inside of the cells (Fig.
   [76]1D). The Kaplan–Meier analysis results showed that high expression
   of NOTCH3 was negatively correlated with progression-free survival and
   locoregional recurrence-free survival of the patients (Fig. [77]1D).
   Collectively, these results suggest that aberrant activation of NOTCH3
   might promote chemoresistance and malignancy of NPC.

Fig. 1. NOTCH3 is aberrantly highly expressed in chemoresistance NPC patients
and associates with poor clinical outcome.

   [78]Fig. 1
   [79]Open in a new tab

   A Volcano Plot of significantly altered genes from the RNA sequencing
   data of 22 late stage nasopharyngeal carcinoma patients, with 8
   patients completely response (CR) to chemotherapy and 14 patients
   partially response (PR) to chemotherapy. B Pathway enrichment analysis
   of genes upregulated or downregulated in PR patients compared to CR
   patients. C qRT-PCR results of all four Notch receptors in patient
   samples. Bars represent the means ± SD, n = 3. *P < 0.05. D Up,
   representative IHC images of NOTCH3 in 234 nasopharyngeal carcinoma
   patients treated with chemotherapy. Scale bars, 30 μm. Down,
   Kaplan–Meier curves showing the impact of NOTCH3 expression on the
   progression-free survival and locoregional recurrence-free survival of
   234 NPC patients. IHC staining of NOTCH3 was reviewed and scored by two
   independent pathologists blindly without knowing the clinical
   characteristics. IHC scores were calculated by multiplying the scores
   for the proportion of positively-stained tumor cells (1, <10%; 2,
   10%–50%; 3, 50%–80%; 4, >80%) and staining intensity (0, no staining;
   1, weak; 2, moderate; 3, strong) by each investigator, then averaged.
   The cutoff used in NOTCH3 survival analysis was selected by receiver
   operator characteristic curve (ROC) analysis. A score <2.5 was used to
   classify tumors with low NOTCH3 expression.

NOTCH3 overexpression promotes chemo-resistance of NPC cells

   To determine the potential role of NOTCH3 in chemo-resistance of NPC,
   we established a paclitaxel-resistant (R) cell model through long time
   culture of 5–8 F parental (5–8 F P) cells in the presence of paclitaxel
   (PTX). The morphology and sensitivity to PTX of the resistant cells
   were confirmed (Fig. [80]2A-B). Intriguingly, we observed that 5–8 F R
   cells also became resistance to other NPC chemo-drugs, such as
   carboplatin and cisplatin (Supplementary Fig. [81]S1A), which are
   commonly used in concurrent chemo-therapy of late stage NPC patients
   [[82]25, [83]26]. Among all four Notch receptors, NOTCH3 was identified
   as the most highly expressed and upregulated receptor in 5–8 F R cells
   compared with 5–8 F P cells (Supplementary Fig. [84]S1B). We then
   assessed Notch intracellular domain 3 (NICD3) expression in paired
   5–8 F P and 5–8 F R cells. The accumulation of NICD3, which is released
   into the cytosol when the NOTCH3 receptor is activated, was observed in
   5–8 F R cells, but not in 5–8 F P cells (Fig. [85]2C), which implied
   that NOTCH3 was aberrantly activated in 5–8 F R cells. Depletion of
   NOTCH3 with siRNAs conferred the chemo-sensitivity of 5–8 F R cells to
   paclitaxel in proliferation assay (Fig. [86]2D) and colony formation
   assay (Fig. [87]2E). Moreover, an inducible NOTCH3 knockdown system was
   established in 5–8 F R cells and conditional knockdown of NOTCH3 by
   doxycycline (Dox) significantly inhibited tumorsphere formation of the
   cells upon PTX treatment (Fig. [88]2F), while overexpression of NICD3
   in 5–8 F P cells or another chemosensitive NPC cell line S18 cells
   remarkably promoted the tumorsphere formation capability in the
   presence of PTX (Fig. [89]2G, H). Together, these results indicated
   that NOTCH3 upregulation contributes to paclitaxel resistance in NPC.

Fig. 2. NOTCH3 overexpression promotes chemo-resistance of NPC cells.

   [90]Fig. 2
   [91]Open in a new tab

   A Left, 2D colony formation assay of paired 5–8 F P and 5–8 F R cells
   under paclitaxel (PTX) treatment or not for 8 days. PTX concentration
   used was 2 nM. All the following experiments used the same
   concentrations of PTX, unless otherwise notified. Right, IC50 assay of
   5–8 F P and 5–8 F R cells under different concentration of PTX
   treatment for 96 h. Cell viability was tested by CellTiter Glo reagent.
   B Tumorsphere assay of 5–8 F P and 5–8 F R cells with PTX treatment or
   without for 10 days. C Immunoblotting analysis of NICD3 expression in
   paired 5–8 F P and 5–8 F R cells. Results shown are representative
   image from three experiments. D Chemosensitivity to PTX by
   proliferation assay of 5–8 F R cells transfected with two different
   NOTCH3 siRNAs. Cell viability was determined by CellTiter Glo reagent
   after 96 h treatment. Bars represent the means ± SD (n = 3). *P < 0.05;
   **P < 0.01; ***P < 0.001. E Colony formation assay in 5–8 F R cells
   with two different NOTCH3 siRNAs treatment. DMSO treated cells were
   served as control. F Tumorsphere formation assay of shNOTCH3 tet-on
   5–8 F R cells treated with doxycycline (Dox) in the presence or absence
   of PTX. Representative images (Left) and quantifications (Right). Bars
   represent the means ± SD (n = 3). Scar bars, 400 μm. *P < 0.05;
   **P < 0.01; ***P < 0.001. G Tumorsphere formation assay of 5–8 F P
   cells with NICD3 overexpression with or without PTX treatment.
   Representative images (Left) and quantifications (Right). Bars
   represent the means ± SD (n = 3). Scar bars, 400 μm. *P < 0.05;
   **P < 0.01; ***P < 0.001. H Tumorsphere formation assay of S18 cells
   and S18 with NICD3 overexpression cells treated with PTX or not.
   Representative images (Left) and quantifications (Right). Bars
   represent the means ± SD (n = 3). Scale bars, 400 μm. *P < 0.05;
   **P < 0.01; ***P < 0.001.

Enhancer reprogramming driven overexpression of NOTCH3 in chemo-resistant NPC
cells

   We then probed the mechanism underlying the aberrant upregulation of
   NOTCH3 in chemoresistance NPC. Transcription profiling of paired
   5–8 F P and 5–8 F R cells identified 4211 differently expressed genes,
   with 2059 genes upregulated and 2153 genes downregulated in 5–8 F R
   cells compared with 5–8 F P cells ( | log2 fold change | ≥ 1, adjusted
   p-value < 0.05) (Fig. [92]3A). To determine whether the transcriptome
   changes in chemoresistance were caused by altered epigenetic landscape,
   we performed H3K27ac (marker of open chromatin regions) and H3K4me3
   (marker of promoters) ChIP-seq in the paired cells. Promoters are
   defined as regions with both H3K27ac and H3K4me3 peaks that are within
   2.0 kb of known transcription start site (TSS) and enhancers as regions
   with H3K27ac peaks that are more than 2.0 kb away from any known TSS.
   Altogether, 23,688 gained and 39,659 lost enhancers were identified in
   5–8 F R cells (Fig. [93]3B). Similarly, 5233 gained and 4,466 lost
   promoters were found in 5–8 F R cells (Fig. [94]3C). Our integrative
   analysis of transcriptional profiles and chromatin landscapes revealed
   that differential gene expression between 5–8 F P and 5–8 F R cells
   were well correlated with changes in the H3K27ac levels (Fig. [95]3D).
   Meanwhile, gene set enrichment analysis (GESA) results indicated that
   Notch signaling pathway was aberrantly activated in 5–8 F R cells (Fig.
   [96]3E). This observation was further verified by qRT-PCR (Fig.
   [97]3F). Consistent with this result, ChIP-qPCR showed that H3K27ac
   increased at upregulated genes of Notch pathway in 5–8 F R cells (Fig.
   [98]3G). Enrichment of ChIP-seq peaks at the enhancer of several picked
   genes in chemo-resistant 5–8 F R cells also confirmed higher
   transcriptional activities (Fig. [99]3H). The upregulation of NOTCH3 is
   of importance, due to its clinical relevance. Histone Acetyltransferase
   (HAT) inhibitors have been shown to decrease H3K27ac levels and disrupt
   enhancer-driven transcriptions [[100]27–[101]29]. To further strengthen
   our hypothesis, we treated 5–8 F R cells with HAT inhibitors. NOTCH3
   expression was significantly reduced and in a dosage dependent manner
   (Supplementary Fig. [102]S2A). These results together suggest that
   enhancer reprogramming activates NOTCH3 signaling to confer
   chemoresistance of NPC.

Fig. 3. Enhancer reprogramming driven overexpression of NOTCH3 in
chemo-resistant NPC cells.

   [103]Fig. 3
   [104]Open in a new tab

   A Heatmap illustrating differentially expressed genes between 5–8 F P
   and 5–8 F R cells ( | log2 fold change | ≥ 1, adjusted p-value < 0.05).
   B, C Heatmaps of H3K27ac and H3K4me3 signals in 5–8 F P and 5–8 F R
   cells. The regions in which the H3K27ac and H3K4me3 levels gained or
   lost in 5–8 F R cells were identified in enhancer or promoter. Density
   heatmaps within ± 2.0 kb around the peak center were shown. D
   Integrative analysis of H3K27ac mark intensity and changes of
   corresponding gene expressions. Examples of genes related to Notch
   pathway have been labeled. E Gene set enrichment analysis (GSEA) shows
   enrichment of Notch signaling pathway in 5–8 F R cells. Normalized
   enrichment score (NES) and nominal P-value of GSEA are indicated. F
   qRT-PCR results of Notch pathway related genes in 5–8 F P and 5–8 F R
   cells. Bars represent the means ± SD, n = 3. *P < 0.05; **P < 0.01;
   ***P < 0.001.G Chip-qPCR analysis of H3K27ac bindings of Notch pathway
   related genes in 5–8 F P and 5–8 F R cells. Bars represent the
   means ± SD, n = 3. *P < 0.05; **P < 0.01; ***P < 0.001. H H3K27ac and
   H3K4me3 ChIP-seq peaks at FZD7, NOTCH3 and HEYL loci in 5–8 F P and
   5–8 F R cells.

NOTCH3 upregulates SLUG to induce chemo-resistance in NPC

   Previous studies reported that SLUG is a direct target of NOTCH4 and
   the overexpression of SLUG has been shown to mediate chemoresistance in
   ovarian cells [[105]30, [106]31]. Therefore, we hypothesized that
   NOTCH3 might also upregulate SLUG to induce chemo-resistance in NPC.
   Compared with 5–8 F P cells, SLUG is expressed at higher levels in
   5–8 F R cells. Knockdown of NOTCH3 in resistant cells reduced the
   expression of SLUG RNA and protein and while overexpression of NICD3 in
   parental cells resulted in significant induction of SLUG RNA and
   protein (Fig. [107]4A, Supplementary Fig. [108]S3A-D). Interestingly,
   compared with monolayer culture, tumorsphere cultured S18 and 5–8 F P
   cells have higher expression of NICD3 and SLUG (Fig. [109]4B).
   Depletion of SLUG with siRNAs resulted in reduced tumorsphere formation
   of resistant cells under PTX treatment (Fig. [110]4C), while
   overexpression of SLUG significantly enhanced the tumorsphere formation
   ability of parental cells upon PTX treatment (Fig. [111]4D), suggesting
   that SLUG is important for maintaining the stemness of cancer cells
   during drug treatment. Moreover, the clinical relevance of SLUG in 115
   nasopharyngeal carcinoma patient samples with chemotherapy
   (Supplementary Table [112]S8-[113]9) was assessed by IHC analysis in
   another cohort. The Kaplan–Meier survival analysis showed that high
   expression of SLUG is negatively correlated with the progression-free
   and metastasis-free survival of the patients (Fig. [114]4E). These
   results implied that activated NOTCH3-SLUG axis is significant for the
   chemoresistance of NPC.

Fig. 4. NOTCH3 upregulates SLUG to induce chemo-resistance in NPC.

   [115]Fig. 4
   [116]Open in a new tab

   A Immunoblotting analysis of NICD3, SLUG expression in different cells.
   Results shown are representative image from three experiments. Empty
   vector (EV), overexpression of NICD3 (OE). B Immunoblotting analysis of
   indicated protein expression in 5–8 F P and S18 cells in different
   culture conditions, monolayer or tumorsphere. Results shown are
   representative image from three experiments. C Tumorsphere formation
   assay of 5–8 F R cells treated with different SLUG siRNAs.
   Representative images (left) and quantification (right). Bars represent
   the means ± SD (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001. Scale
   bars, 400 μm. D Tumorsphere formation assay of 5–8 F P cells with SLUG
   overexpression with or without PTX treatment. Representative images
   (Left) and quantifications (Right). Bars represent the means ± SD
   (n = 3). Scar bars, 400 μm. *P < 0.05; **P < 0.01; ***P < 0.001. E
   Left, representative IHC images of SLUG in 125 nasopharyngeal carcinoma
   patients treated with chemotherapy. Scale bars, 30 μm. Right, the
   Kaplan–Meier survival analysis for progression-free survival and
   metastasis-free survival and locoregional recurrence-free survival of
   125 patients with nasopharyngeal carcinoma with high (red) or low
   (blue) SLUG expression levels. IHC staining of SLUG was blindly
   reviewed and scored by two independent pathologists without knowing the
   clinical characteristics. IHC scores were calculated by multiplying the
   scores for the proportion of positively-stained tumor cells (1, <10%;
   2, 10%–50%; 3, 50%–80%; 4, >80%) and staining intensity (0, no
   staining; 1, weak; 2, moderate; 3, strong) by each investigator, then
   averaged. The cutoff used in SLUG survival analysis was selected by
   receiver operator characteristic curve (ROC) analysis. A score < 7 was
   used to classify tumors with low SLUG expression.

Chemical inhibitor targeting Notch signaling restores chemosensitivity of NPC
cells in vitro

   To evaluate whether targeting NOTCH3 restores chemosensitivity of
   resistant NPC cells, we analyzed the chemo-sensitizing properties of
   Notch signaling inhibitor in paclitaxel resistant NPC cell line.
   Unfortunately, no specific NOTCH3 inhibitor has been developed
   currently. Drug combination index showed that the combination of
   Dibenzazepine (DBZ) and paclitaxel was synergistic (Fig. [117]5A).
   Colony formation assay and time-course proliferation assay indicated a
   dramatic increase in PTX sensitivity in 5–8 F R cells with combined
   treatment (Fig. [118]5B–C). In addition, the tumorsphere formation
   ability of 5–8 F R cell was effectively blocked by the combination
   treatment, indicating remarkable restoration of chemosensitivity (Fig.
   [119]5D). The combination treatment also remarkably increased the
   apoptosis of 5–8 F R cells (Fig. [120]5E, F). Moreover, decreased
   expression of NICD3 were observed under DBZ and combined treatment by
   western results (Fig. [121]5G). Similar results were obtained in other
   NPC cell lines (Supplementary Fig. [122]S4A–D). These data demonstrated
   that Notch signaling inhibitor restored the sensitivity to PTX in
   5–8 F R cells, inhibited cell proliferation and promoted cell death.

Fig. 5. Chemical inhibitor targeting Notch signaling restores
chemosensitivity of NPC cells in vitro.

   [123]Fig. 5
   [124]Open in a new tab

   A Combination index (CI) of Dibenzazepine (DBZ) and paclitaxel (PTX) in
   5–8 F R cells. The synergistic effect is demonstrated by CI
   value < 1.0. CI values were calculated with CalcuSyn software. Cells
   treated with indicated concentration of DBZ or PTX, single or in
   combination for 96 h. Cell viability was determined by CellTiter Glo
   reagent. B Colony formation assay of 5–8 F R cells treated with 25 μM
   DBZ in the presence or absence of 3.0 nM PTX for 8 days. All the
   following experiments used the same concentrations of DBZ or PTX,
   unless otherwise notified. PTX + DBZ (P + D). C Cell relative
   proliferation curve of 5–8 F R cells with different treatments. Data
   are shown as means ± SD (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001. D
   Tumorsphere formation assay of 5–8 F R cells treated with Notch
   signaling inhibitor in the presence or absence of PTX for 12 days.
   Scale bar, 400 μm. E Sub-G1 population analysis in 5–8 F R cells
   treated with Notch signaling inhibitor, PTX or both for 72 h. Data are
   shown as means ± SD (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001. F
   Annexin-V-PI staining to assess the apoptotic ratios of 5–8 F R cells
   under single or combined treatment for 72 h. Left, representative
   images of FACS analysis; Right, percentages of apoptotic cells in
   treatment groups as observed with flow cytometry. Bars represent the
   means ± SD, n = 3. *P < 0.05; **P < 0.01; ***P < 0.001. G
   Immunoblotting analysis of NICD3 under different treatments. Cells were
   collected after 72 h treatment. Results shown are representative image
   from three experiments.

NOTCH3 overexpression enhances chemoresistance of NPC in vivo

   Due to that 5–8 F R cells had a very low tumorigenesis rate, to further
   verify the role of NOTCH3 in chemoresistance in vivo, S18 cells and S18
   with stable overexpression of NICD3 (S18 OE) cells were engrafted
   subcutaneously to nude mice. Mice with S18 and S18 OE xenograft tumors
   were treated with or without paclitaxel. Notably, PTX treatment reduced
   the tumor sizes of S18 cells significantly but not S18 OE cells,
   indicating that enhanced NOTCH3 activation and signaling confers
   chemoresistance of NPC in vivo (Fig. [125]6A–C, Supplementary Fig.
   [126]S5A–B). IHC results demonstrated that cell proliferation
   (represented by KI-67 expression) was remarkably reduced and apoptosis
   (represented by cleaved caspase-3 expression) was increased in S18
   xenograft tumors treated with PTX compared with S18 OE xenograft (Fig.
   [127]6D). Moreover, the overexpression of NOTCH3 in S18 OE groups were
   confirmed by IHC (Fig. [128]6D). These data certified that NOTCH3
   overexpression plays an important role in the chemoresistance of NPC in
   vivo.

Fig. 6. NOTCH3 overexpression enhances chemoresistance of NPC in vivo.

   [129]Fig. 6
   [130]Open in a new tab

   A Xenograft tumor growth curve of S18 cells and S18 with stable
   overexpression of NICD3 (S18 OE) cells under different treatments. S18
   and S18 OE cells were treated with 5 mg/kg paclitaxel or without for 16
   days, as descried in Material and Methods. Error bars, mean ± SEM
   (n = 6 per group). *P < 0.05; **P < 0.01 (independent t-test). B Tumors
   were presented by photographs. Each group consists of 6 mice. C Bar
   graph shows the tumor weight at the 16th day post-treatment. Data are
   presented as the mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001
   (independent t-test). D HE, KI-67, cleaved caspase-3 and NOTCH3 IHC
   staining of xenograft tumor. Up: Representative images. Scale bars 20
   μm. Down: Z-score of KI-67, cleaved caspase-3, NOTCH3 in tumor sections
   with different treatments. IHC scores were calculated by multiply the
   scores for the proportion of positively-stained tumor cells (1, <10%;
   2, 10%–50%; 3, 50%–80%; 4, >80%) and staining intensity (0, no
   staining; 1, weak; 2, moderate; 3, strong) by each investigator, then
   averaged. Z-score of IHC was defined by using the formula z = (x − μ)/
   σ, where x is the raw IHC score, μ is the population mean, and σ is the
   population SD. Two-tailed t-test was used to calculate p-values.
   *P < 0.05; **P < 0.01; ***P < 0.001.

Discussion

   The incidence of advanced NPC is rising in Southeast Asian countries
   and in Southern China, yet therapeutic strategies remain limited and
   ineffective, resulting in high mortality. Although the combined
   treatment of chemo-radiotherapy produces a satisfying survival rate
   initially, metastasis and chemo-resistance inevitable develops. In the
   present study, we seek to unravel the molecular mechanism of
   chemoresistance in NPC, which may provide potential therapeutic targets
   to improve clinical outcomes. Through analysis of NPC tumor tissues and
   cellular models, we reported that enhancer remodeling activates NOTCH3
   to confer chemo-resistance in NPC. Moreover, genetic or pharmacological
   perturbation of NOTCH3 could potentiate the sensitivity to paclitaxel
   in NPC. Thus, we propose that targeting NOTCH3 signaling will help
   overcome chemoresistance in NPC.

   The current work focused on advanced chemo-resistant NPC and epigenetic
   regulation. To our knowledge, this is the first report demonstrated
   that enhancer remodeling driven aberrant activation of NOTCH3 to
   enhance the expression of SLUG, which confers chemoresistance of NPC.
   Epigenetic signature changes are dynamic and reversible, making
   chemoresistance-specific enhancers ideal therapeutic targets since they
   wouldn’t affect normal tissues. At therapeutic perspective, NOTCH3
   enhancer could be a best potential target in advanced NPC. Gamma
   secretase inhibitor (GSI) gained attention in cancer treatment during
   the past decade [[131]32–[132]34]. It acts by inhibiting the cleavage
   of γ-secretase, which results in blocking Notch signaling pathway. DBZ
   is a type of GSI, our data showed that DBZ treatment reduced the
   expression level of NICD3 in 5–8 F R cells and sensitized resistance
   cells to PTX. Unfortunately, it is not a specific inhibitor for NOTCH3
   and the usage of GSI is limited due to its toxicity [[133]35], better
   inhibitor should be developed in the future.

   SLUG is a member of SNAIL family of zinc-finger transcription factors
   and its deregulation has been found in several types of human cancer
   [[134]36–[135]38]. In addition to regulating epithelia-mesenchymal
   transition, researchers have shown that SLUG also plays a role in
   maintenance of the stemness properties of tumor cells. The stem-like
   phenotype of cancer cells has been proven related with resistance to
   chemotherapy and radiotherapy in various cancers [[136]39–[137]41]. Our
   study showed that SLUG is a downstream target of NOTCH3 with clinical
   relevance in NPC. Moreover, SLUG overexpression was correlated with the
   increased tumorsphere formation ability of NPC cells upon PTX
   treatment, which suggested that SLUG maintained the stemness of NPC
   cells in the presence of PTX. Similar with our results, SLUG
   upregulation has been implicated in radiotherapy resistance in NPC
   previously [[138]42–[139]44]. Taken together, our data firstly showed
   that NOTCH3-SLUG axis is correlated with chemo-resistance in NPC and
   poor prognosis of NPC patients.

   The present study also has limitations, including the inability to find
   another NPC cell line with NOTCH3 overexpression. Actually NPC is
   consistently associated with Epstein-Barr virus, however, most if not
   all the reported NPC cell lines have lost their EBV episomes and became
   EBV-negative upon in vitro propagation. So it’s hard using in vitro
   models to mimic clinical situation.

   In summary, our study demonstrated that genome-wide enhancer
   reprogramming activates NOTCH3 signaling to confer chemo-resistance of
   NPC (Fig. [140]7). The overexpression of NOTCH3 and SLUG were
   correlated with chemoresistance of NPC and poorer prognosis of NPC
   patients. Importantly, genetic depletion or pharmacological inhibition
   of NOTCH3 restores the chemosensitivity of NPC, which may provide a
   potential therapeutic strategy for advanced NPC patients.

Fig. 7. A schematic summary of the findings of this work.

   [141]Fig. 7
   [142]Open in a new tab

   Genome-wide enhancer reprogramming activates NOTCH3 signaling to confer
   chemoresistance of NPC, suggesting that targeting NOTCH3 could provide
   a potential therapeutic strategy to effectively treat chemoresistant
   NPC.

Materials and methods

Patient samples, cell culture, and reagent

   This study was approved by the Institutional Ethical Review Board of
   Guangdong Provincial People’s Hospital (KYZ202157201). Written informed
   consents were obtained from all patients. Human nasopharyngeal
   carcinoma cell lines 5–8 F, 6–10B, CNE2, S26, S18 and HEK293T cells
   were grown in RPMI1640 (Invitrogen) or DMEM (Invitrogen) supplemented
   with 10% FBS and 1% penicillin–streptomycin (Gibco BRL). Cells were
   maintained at 37 °C in a 5% CO[2] incubator and tested for mycoplasma
   contamination every month. All the cell lines used in this study had
   been authenticated and were generously provided by Dr. M. Zeng (Sun
   Yat-sen Univeristy Cancer Centre). To establish paclitaxel-resistant
   cell lines, 5–8 F cells were treated with paclitaxel for 3-months by
   escalating concentration, until certain target concentration was
   achieved. Resistant cells were maintained under 2.5 nM paclitaxel
   exposure until experiments were performed. Dibenzazepine (DBZ),
   SGC-CBP30 and dCBP1 were purchased from TargetMol.

RNA sequencing (RNA-seq) and data analysis

   Total RNA was extracted from NPC patient samples or 5–8 F P and 5–8 F R
   cells by Trizol (Invitrogen), followed by purification using RNeasy
   Mini Kit (Qiagen). RNA-seq library were constructed using Truseq
   Stranded Total RNA Gold library kit (Illumina) according to
   manufacturer’s protocol. The libraries were sequenced using an Illumina
   HiSeq2500 platform with paired-end reads of 150 bases. For data
   analysis, clean reads gained with fastp software (version 0.12.5)
   [[143]45] were aligned to human reference genome (GRCh38, hg38 with
   STAR (version: 2.7.0 f)) [[144]46]. RSEM (version: 1.3.3) was used to
   calculate the transcripts value for each gene [[145]47]. Differentially
   expressed genes were called by DESeq2 R package (version:1.28.1) with
   | log[2] fold change | ≥ 1 and adjusted p-value < 0.05. Gene Set
   Enrichment Analysis was assessed by a GSEA program [[146]48].

Quantitative real-time PCR (qRT-PCR)

   For qRT-PCR, total RNA was extracted as described above and
   reverse-transcribed into cDNA using TransScript All-in-one First-Stand
   cDNA Synthesis SuperMix for qPCR kit (Transgene Biotech). Quantitative
   real-time PCR was performed with SYBR Green(Vazyme) on a Bio-Rad PCR
   Detection System. 18 S was used as internal control. All experiments
   were performed in biological triplicates unless stated otherwise.
   Sequences of qRT-PCR primers are listed in Supplementary Table [147]S1.

Chromatin immunoprecipitation DNA-sequencing (ChIP-seq) and data analysis

   5–8 F P and 5–8 F R cells were seeded into T75 flask for 72 h before
   collected for ChIP-sequencing. 1 × 10^6 cells were used for each
   ChIP-sequencing experiment and the cells were crosslinked for 10 min
   with 1% formaldehyde at room temperature. The cells were then lysed
   with 1% SDS lysis buffer (50 mM Tris-HCl pH 8.0, 10 mM EDTA, 1% SDS)
   and sonicated on ice using Bioruptor (Diagenode) for 16 cycles (on and
   off session at 30 s each). Supernatant was collected and precleared
   with protein G Dynabeads (Invitrogen) for 2 h at 4 °C. Then the
   chromatin extract was incubated with specific antibody-binding magnetic
   beads overnight at 4 °C. The beads were washed with low and high salt
   buffer and chromatin was eluted from the beads with elution buffer
   (50 mM Tris-HCl pH 8.0, 10 mM EDTA, 1%SDS) after incubating for 40 min
   at 68 °C with agitation. Precleared chromatin without incubation with
   antibody-binding beads was served as input. Reversed cross-link was
   performed and DNA were isolated using Phenol-Chloroform-Isoamyl alcohol
   (Thermo Fisher Scientific) and ethanol precipitation. ChIP-DNA obtained
   were used for ChIP-qPCR and sequences of the primers used were shown in
   Supplementary Table [148]S2.

   Input and ChIP DNA were amplified using GenomePlex® Single Cell Whole
   Genome Amplification kit. After amplification, DNA was purified by PCR
   product purification kit (Qiagen) and the concentration of amplified
   DNA was determined by PicoGreen. After digestion by BpmI restriction
   enzyme, 30 ng DNA was used for library preparation using NEBNext®
   ChIP-seq Library Prep Master Mix Set for Illumina® kit. KAPA library
   Quantification and qPCR were performed for specific genes, as quality
   control. Libraries were sequenced using an Illumina HiSeq2500 platform
   with paired-end reads of 150 bases.

   For ChIP-seq data analysis, reads were aligned to the human reference
   genome (GRCh38, hg 38) using Bowtie2 (version 2.3.2) [[149]49]. MACS2
   (version: 2.1.2) was employed to call narrow peaks of H3K27ac and
   H3K4me3 with default settings (-q 0.01) [[150]50]. ChIPseeker (version
   1.14.2) [[151]51] was used to annotate the peaks and visualization of
   the data was performed by deepTools [[152]52]. Promoters are defined as
   regions with both H3K27ac and H3K4me3 peaks that are within 2.0 kb of
   known transcription start site (TSS) and enhancers as regions with
   H3K27ac peaks that are >2.0 kb away from any known TSS. Alignments for
   specific gene were visualized using Integrative Genomics Viewer
   (version 2.4.19) [[153]53].

Colony formation, tumorsphere formation, IC50 and cell proliferation assay

   For colony formation assay, 1 × 10^4 cells were plated in 6-well plate
   24 h. before the indicated reagents were added. When the control group
   reached 100% confluence, colonies were fixed with methanol and stained
   with Crystal Violet (Sigma) for colony visualization. For tumorsphere
   formation assay, cells were trypsinized and passed through 0.4 μm cell
   strainer to achieve single-cell suspension. Cells was cultured
   (1 × 10^4 cells/well) in 6-well ultra-low-attachment plates (Corning)
   with MammoCult Medium (STEMCELL Technologies, Vancouver, BC)
   supplemented with heparin (1:500) and fresh hydrocortisone (0.5 μg/ml).
   Tumorspheres were cultured for 10–12 days and stained with
   2-(4-iodo-phenyl)-3-(4-nitrophenyl)-5-phenyl-2H-tetrazolium chloride
   (INT) (Sigma-Aldrich) and photographed. Images were taken with Olympus
   microscope IX71 and colony quantification were performed using Olympus
   CellSens Dimension software. For IC50 assay, 2 × 10^3 cells per well
   were seeded in 96-well plate 24 h before treatment with different
   dosages of PTX (0.01, 0.05, 1, 5, 10, 50, 100 and 500 nM). 96 h later,
   cells were lysed with CellTiter Glo and chemiluminescent signal was
   detected with a microplate reader. For proliferation assay, 2 × 10^3
   cells per well were seeded in 96-well plate. 24 h later, cells were
   treated with the indicated reagents. The number of viable cells were
   measured by CellTiter Glo reagent at the indicated time points.

Combination index assay

   Synergistic effects of combined drug treatment were assessed by
   combination index (CI) assay, where CI < 1.0 represents synergy.
   Briefly, 5–8 F R cells were exposed to indicated concentrations of
   Notch signaling inhibitor with or without paclitaxel for 96 h. CI
   values were determined by the inhibition rate of the cells calculated
   with CalcuSyn software.

Immunoblotting analysis

   Cells were lysed in RIPA buffer with protease inhibitors. Protein
   concentration was determined by a Bradford Assay kit (Bio-rad). Protein
   was separated by 8% or 12% SDS-PAGE and transferred onto a PVDF
   membrane. The membrane was blocked in 5% skim milk or 5% BSA, followed
   by primary antibody incubation overnight. The primary antibodies used
   were: NOTCH3 (CST, 5276), SLUG (CST, 9585), Actin (CST, 4359) and GAPDH
   (CST, 2118). The secondary antibodies used were horseradish
   peroxidase-conjugated anti-rabbit and anti-mouse (GE Healthcare Life
   Science, NA934 and NA931). The bands were detected using
   chemiluminescence with ChemiDox^TM MP Imaging System (Bio-Rad).

Flow cytometric analysis

   PI staining was done to quantify the sub-G1 population, which can
   reflect the extent of cell death. Cells were seeded in 6-well plates
   24 h prior to treatment with different compounds. After 72 h, cells
   then collected, fixed and stained with PI (50 μg/mL). The stained cells
   were analyzed by SP6800 Spectral Cell Analyzer (Sony) and quantified by
   using the FlowJo software.

   Cell death was also detected by flow cytometry using Annexin V-FITC/PI
   double staining (BD Biosciences) using the manufacturer’s protocol. The
   stained cells were immediately analyzed by flow cytometric analysis.

siRNA, shRNA and plasmid transfection

   Paclitaxel-resistant cells were transiently transfected with siRNAs
   targeting NOTCH3, SLUG or with negative control siRNA obtained from
   Sangon Biotech Company, using lipofectamine RNAimax reagents (Thermo
   Fisher Scientific). Transfection was performed according to the
   manufacturer’s instructions. The sequences of the siRNAs are shown
   Supplementary Table [154]S3.

   To generate NOTCH3 inducible knockdown cell lines, an shRNA from the
   TRIPZ Dox-inducing lentiviral shRNA system was used. The lentivirus
   vectors were co-transfected with two packaging vectors psPAX2 and
   pMD2.G at a ratio of 2 μg: 1 μg: 0.5 μg into packaging cell HEK293T.
   After 48 h, supernatant containing virus were collected and filtered.
   The filtered virus was used to infect the host cells. 48 h after
   infection, cells were selected by puromycin (1.0 μg/mL). To induce
   NOTCH3 knockdown, 1.0 μg/mL doxycycline was added. The sequence of
   shRNA targeting NOTCH3 is shown in Supplementary Table [155]S3. The
   expression plasmid of NICD3 is a kind gift from Prof. Suling Liu (Fudan
   University Shanghai Cancer Center). Briefly, cDNA of NOTCH3
   intracellular domain (amino acids 1663–2321) was synthesized at Sangon
   Biotech (Shanghai, China) and cloned into pLVX-puro vector. Similarly,
   the coding sequences for SLUG were amplified from cDNA obtained from
   HEK293T cells and then inserted into pLVX-puro plasmid. The empty
   vector (pLVX-puro) and expression plasmids were transfected with
   Lipofectamine 2000 (Thermo Fisher Scientific) reagents into 5–8 F R or
   S18 cells.

Immunohistochemistry (IHC) analysis

   The procedure of IHC was performed as previously described [[156]54].
   Briefly, the slides were incubated overnight at 4 °C with primary
   antibodies and followed by 60-minnute incubation with rabbit polymer
   detection system kit or mouse polymer detection system kit
   (ZsbioCommerce Store, PV-6001, PV-6002). IHC staining was reviewed and
   scored by two independent pathologists blindly without knowing the
   clinical characteristics. IHC scores were calculated by multiplying the
   scores for the proportion of positively-stained tumor cells (1, <10%;
   2, 10%–50%; 3, 50%–80%; 4, >80%) and staining intensity (0, no
   staining; 1, weak; 2, moderate; 3, strong) by each investigator, then
   averaged. The cutoff used in NOTCH3 and SLUG survival analysis was
   selected by receiver operator characteristic curve (ROC) analysis. A
   score <2.5 was used to classify tumors with low NOTCH3 expression and a
   score <7 was used to classify tumors with low SLUG expression. Z-score
   of IHC was defined by using the formula z = (x-μ) / σ, where x is the
   raw IHC score, μ is the population mean, and σ is the population SD.
   The antibody used were KI-67 (Zsbio Commerce Stor, ZA-0502),
   cleaved-caspase 3 (CST, 9661), SLUG (Novus, NBP2-03886) and NOTCH3
   (Abcam, ab23426).

Animal studies

   Female BALB/c nude mice (5–6 weeks) were purchased from Beijing Vital
   River Laboratory Animal Technology Company. All experiments were
   conducted in compliance with animal protocols approved by the Ethics of
   Animal Experiments of Guangdong Provincial People’s Hospital. Tumors
   were measured by Vernier caliper every two days. And tumor volume was
   calculated with the following formula: tumor volume
   (V) = width × width × length × 0.537. For S18 xenograft model, 2 × 10^6
   cells were subcutaneously injected into the left flank of the mice.
   After the tumors volume reached 100 mm^3, the mice were randomly
   divided into two groups: vehicle-treated group (n = 6) and
   paclitaxel-treated (5 mg/kg) group (n = 6). For S18 cells with NICD3
   overexpression (S18 OE), same cells number as S18 were implanted
   subcutaneously at the same time. And mice were divided randomly into
   two groups for drug treatment simultaneously: vehicle alone (n = 6),
   paclitaxel treatment alone (n = 6). Paclitaxel was given every two days
   by intraperitoneal injection. Mice were sacrificed by CO[2] inhalation
   after the tumor volume of S18 control group reached 1000 mm^3 and
   tumors were harvested for further analysis. No statistical method was
   used to predetermine the sample size for this experiment. No data was
   excluded and the investigator was blinded to the group allocation
   during the experiment.

Statistical analysis

   The two-tailed student’s t-test was used to calculate differences
   between experimental groups. P < 0.05 was considered statistically
   significant unless stated otherwise. All statistical analyses were done
   using GraphPad Prism version 7.0 and all in vitro experiments were
   repeated at least 3 times. Protein expression levels were quantified
   using Image J software, and the relative quantity of protein was
   calculated.

Supplementary information

   [157]Supplementary Figures S1-S5^ (359.9KB, pdf)
   [158]Supplementary Tables S1-S9^ (499.6KB, pdf)
   [159]Western uncut^ (435.8KB, pdf)

Acknowledgements