Paclitaxel-induced stress granules increase LINE-1 mRNA stability to promote drug resistance in breast cancer cells

Xiao Shi, Xinxin Si, Ershao Zhang, Ruochen Zang, Nan Yang, He Cheng,2, Zhihong Zhang,Beijing Pan, Yujie Sun,2,3,✉

1Key Laboratory of Human Functional Genomics of Jiangsu Province, Nanjing Medical University, Nanjing, Jiangsu 211166, China;

2Department of Cell Biology, School of Basic Medical Sciences, Nanjing Medical University, Nanjing, Jiangsu 211166,China;

3Collaborative Innovation Center for Cancer Personalized Medicine, Nanjing Medical University, Nanjing, Jiangsu 211166, China;

4Department of Pathology, the First Affiliated Hospital of Nanjing Medical University, Nanjing, Jiangsu 210036,China.

Abstract Abnormal expression of long interspersed element-1 (LINE-1) has been implicated in drug resistance, while our previous study showed that chemotherapy drug paclitaxel (PTX) increased LINE-1 level with unknown mechanism. Bioinformatics analysis suggested the regulation of LINE-1 mRNA by drug-induced stress granules(SGs). This study aimed to explore whether and how SGs are involved in drug-induced LINE-1 increase and thereby promotes drug resistance of triple negative breast cancer (TNBC) cells. We demonstrated that SGs increased LINE-1 expression by recruiting and stabilizing LINE-1 mRNA under drug stress, thereby adapting TNBC cells to chemotherapy drugs. Moreover, LINE-1 inhibitor efavirenz (EFV) could inhibit drug-induced SG to destabilize LINE-1. Our study provides the first evidence of the regulation of LINE-1 by SGs that could be an important survival mechanism for cancer cells exposed to chemotherapy drugs. The findings provide a useful clue for developing new chemotherapeutic strategies against TNBCs.

Keywords: LINE-1, stress granules, RNA stability, neoplasm drug resistance, triple negative breast cancer

Introduction

Triple negative breast cancers (TNBCs) are characterized by a lack of expression of the estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) and show high metastases and poor prognosis. Endocrine therapy and HER2 targeted therapies are usually ineffectual for TNBC patients because of high intratumor heterogeneity and a lack of molecular targets[1]. Chemotherapy is still the standard and main strategy for TNBC patients. However, 40%–80% of the patients develop chemotherapy resistance and have poor outcomes[2]. Elucidation of the molecular mechanisms underlying the drug resistance is essential for developing effective treatment strategies.

Long interspersed element 1 (LINE-1) is the only autonomously active retrotransposon in the human genome, and its total copy number accounts for approximately 20% of the genome[3]. An intact LINE-1 is about 6 kb in length and contain a 5′ untranslated region (UTR), two open reading frames (ORF1 and ORF2) that encode the ORF1p and ORF2p proteins,and a 3′UTR that ends in a polyadenosine-rich (poly(A)) tract. The ORF1p has RNA-binding and nucleic acid chaperon activities, whereas ORF2p has endonuclease and reverse transcriptase activities.LINE-1 expression is highly suppressed in normal differentiated somatic cells; however, in tumor cells,the expression of LINE-1 is abnormally activated,resulting in DNA double-stranded breaks, more homologous recombination and chromosome duplication through retrotransposition, which increase mutagenesis and genome instability[4]. Additionally,LINE-1 can also directly regulate gene expression and drug responseviaits RNA and protein[5–7]. Increased LINE-1 expression promotes cancer cell proliferation and alters the response of cancer cells to chemotherapy agents[8]. For example, LINE-1 ORF-1p can function as co-activator to promote the proliferation of breast cancer cells[9]. Inhibition of LINE-1 ORF1p enhances the response of hepatocellular carcinoma cells to epirubicin (EPI) and cisplatin (DDP)[10]. Notably, environmental stresses like chemotherapy agents, UV light and gamma radiation, can cause abnormal activation of LINE-1 expression in different cell context[11–12], which benefits cell survival under various stresses. It is reported that cells exposed to carboplatin prolong their own survival by increasing the expression of LINE-1[13]. We previously found that paclitaxel(PTX), a first line chemotherapy drug for TNBCs,caused an abnormal increase in LINE-1 expression level that significantly reduced the cell response to drugs. However, the mechanisms underlying druginduced LINE-1 expression and subsequent chemotherapy resistance are unclear.

Stress granules (SGs) are assemblies of mRNAs and proteins that form in response to stressors like nutrient deprivation, heat, oxidative stress, and genotoxic stress. Their formation and dynamics significantly affect mRNA localization, translation,and stabilization, as well as the signaling pathways and stress responses of cells[14]. SGs favor cell survival during stress, while impairment of SG formation results in cell death. Importantly, several approved chemotherapeutic compounds have been reported to induce canonical or non-canonical SGs and thereby compromise treatment and promote drug tolerance[15–17]. SGs induce chemotherapy resistance by inhibiting pro-apoptotic signaling pathways, like the mTOR and MAPK pathways[18–19]or by increasing the stability of the mRNA of anti-apoptosis genes. For example, SGs formed in HeLa cells under the stress of bortezomib can stabilize the mRNA of p21, enhance the anti-apoptotic ability of cells, and develop bortezomib resistance in HeLa cells[20]. Several reports have suggested an association of LINE-1 and SGs. For example, Doucetet alhave demonstrated that LINE- 1 ribonucleoprotein complexes (RNPs), including the ORF1p component, co-localize with the SG marker G3BP1 and other SG proteins[21]. This indicates that the mRNA of LINE-1 is able to localize within SGs.A SG transcriptome study provided that mRNAs with an overall length of the coding region longer than 2.5 kb and with longer untranslated regions (UTRs) were accumulated and enriched in the SGs[22]. TheLINE-1mRNA, with a coding region of nearly 5 kb and a 1 kb 5′UTR, falls into this category. Our bioinformatics analysis using catRAPID and RPISeq website also predicted a strong interaction betweenLINE-1mRNA and T-cell intracellular antigen protein-1 (TIA1) and G3BP stress granule assembly factor 1 (G3BP1), two important SG marker proteins. This evidence suggested thatLINE-1mRNA may be translocated to SGs during drug stress.

Based on these previous studies, we proposed that the formation of SGs induced by PTX could reduce the drug response of cancer cells by stabilizingLINE-1mRNA. We tested this hypothesis by systematic experiments using a PTX-resistant TNBC cell line and a xenograft mouse model. Our results demonstrated that SG formation was significantly increased in TNBC cells under PTX stress. Furthermore, druginduced SGs recruitedLINE-1mRNA, resulting in increased stability of theLINE-1mRNA and chemotherapy resistance in TNBC cells.

Materials and methods

Cell culture and reagents

The human TNBC cell line MDA-MB-231 was obtained from ATCC. The PTX-resistant cell line MDA-MB-231R and its parental MDA-MB-231S line was established by pulse selection with 100 nmol/L PTX, as described previously[23]. The human TNBC cell line MDA-MB-468 was kindly provided by Nan Qiao (SuZhou Institute of Systems Medicine, China).LINE-1 expression plasmids PB015 and the vector plasmid PB001 with Neo resistance gene were gifts from Pasano Bojang Jr which was constructed by cloning LINE-1 sequence into a modified pGL4.15Luc2P vector[24]. We established stable LINE-1 overexpressing clones by transfecting PB015 into MDA-MB-231 cells, followed by G418 selection.The cells stably expressing LINE-1-Neo survived. A limiting dilution assay was then used to obtain a single cell incubated with conditioned medium until it grew into a clone. Clones transfected with PB001 as the control clone were named Ctr-clone. Two LINE-1 stable expressed clones named L1-clone 1 and L1-clone 2 were used for the experiments. All cells were maintained in L15 culture medium supplemented with 10% newborn calf serum, insulin (0.2 U/mL), 100 U/mL penicillin, and 100 U/mL streptomycin. All the cell lines were incubated at 37 °C in a humidified 5%CO2atmosphere. The following antibodies were used:antibody against LINE-1 (1:1000 dilution; Santa Cruz Biotechnology, USA) is a rabbit polyclonal antibody raised against amino acids 1081 to 1190 mapping near the C-terminus of human LINE-1 ORF2p; anti-ORF1p is a mouse monoclonal targets the sequence corresponding to 10 amino acids from the N-terminal region of LINE-1 ORF1p (1:1000 dilution;MilliporeSigma, USA); anti-TIA1 is a rabbit polyclonal to TIA1 C-terminal (1:1000 dilution;Abcam, UK); anti-GAPDH (1:20 000 dilution;Bioworld, China); anti-β-actin (1:4000 dilution;MilliporeSigma). Actinomycin D was obtained from GlpBio (USA). Efavirenz was purchased from ApexBio (USA). Paclitaxel, epirubicin, and cisplatin were purchased from Jiangsu Cancer Hospital(Nanjing, China).

Plasmid construction and cell transfection

In order to construct TIA1 expression plasmids, the coding sequence of TIA1 was amplified by Phanta MAX Super-Fidelity (Vazyme, China). The amplified fragments were purified and inserted intoHind Ⅲ andXhoⅠ-digested pCMV6-Entry vector. The interfere plasmids of TIA1 were obtained by inserting two different shRNA sequences into the PLKO.1 vector respectively. The two shRNA sequences targeting TIA1 were: 5′-CGATTTGGGAGGTAGTGAA-3′ and 5′-AAGCTCTAATTCTGCAACTCT-3′. All the plasmids were transfected by Lipofectamine3000 (Thermo Fisher scientific, USA).

RNA extraction, qRT-PCR and RNA stability assay

Total RNA was extracted using TRIzol reagent(Invitrogen, USA), according to the manufacturer's protocol. Before reverse transcription, a step of removing genomic DNA was performed to eliminate interference from genomic DNA. Then cDNA was prepared by reverse transcribing 500 ng of total RNA,according to the manufacturer's instructions (Vazyme,China). Real-time PCR was performed using SYBR Green I Master Mix (Takara, Japan) and 0.5 μmol/L primers and run on a Light Cycler 480 (Roche,Switzerland) with typical amplification parameters (95°C for 10 minutes, followed by 40 cycles of 95 °C for 30 seconds and 60 °C for 1 minute). The relative RNA level was determined by comparing the △CT value of each gene normalized toACTBorRPLP0for each reaction. All samples were run in at least two duplicates. For mRNA stability test, cells were treated with 2 μmol/L actinomycin D for 2, 4, or 6 hours after transfection or drug treatment. The total RNA was then extracted. After normalized toRPLP0, the group without actinomycin D treatment as initial RNA level.Primers forTIA1were: forward, 5′-CGAGATGCCC AAGACTCTATACG-3′; reverse, 5′-CCTTACCCAT TATCTTCCGTCCA-3′. Primers forLINE-1ORF1 were: forward, 5′- GAGCTACGGGAGGACATTCA-3′; reverse, 5′-CTTCCAGTTGATCGCATCGG-3′.

Cell viability assay

Cells were seeded at a density of 5000 cells per well in 96-well plates. On the following day, cells were treated with different drugs for another 48 hours. At the end of the culture, cell viability was measured using the MTT assay, as described previously[23].Simply, cells were treated by MTT (5 mg/mL) for 4 hours. Then discarded the supernatant and add 200 μL dimethyl sulfoxide (DMSO) into each well. After a mixing step, the absorbance at 570 nm for each well was measured by a microplate spectrophotometer(Bio-Rad, USA). All assays were carried out three times independently in triplicate.

RNA-FISH and protein immunofluorescence

Three 5'FAM modified fluorescent probes targeting LINE-1 RNA were obtained from Genepharma(China). The positions of the three probes targeting the LINE-1 RNA sequence (5′ to 3′) were: 1160–1214, 2350–2404, 5565–5621, sequences of the probes were as follows: 5′-AGAAGTGCTTAAAGGAGCT GATGGAGCTGAAAACCAAGGCTCGAGAACTA CGTGA-3′; 5′-GCACCCAGATTCATAAAGCAAG TCCTCAGTGACCTACAAAGAGACTTAGACTCC C-3′; 5′-ATGCACACGTATGTTTATTGCGGCACT ATTCACAATAGCAAAGACTTGGAACCAACC-3′. RNA-FISH was conducted according to the manufacturer's instructions. Briefly, cells were washed twice with PBS, 5 minutes each time and then fixed with 4% paraformaldehyde in PBS for 20 minutes at room temperature, followed by a 15 minutes permeabilization step in 0.5% Triton in PBS and 5 minutes in 0.1% Tween-20 in PBS. After two 5 minutes washes with PBS, 100 μL 2× buffer C was added to each well and the plates were placed in a 37 °C incubator for 30 minutes. Probes were denatured at 73 °C for 5 minutes, hybridized at a 2 μmol/L working concentration with cells in a 37 °C incubator overnight for 12 to 16 hours. On the next day, the samples were removed from the incubator,the probe mixture was aspirated, and the cells were washed once with 0.1% Tween-20, 2× Buffer C and 1× Buffer C. After blocking with 5% FBS in PBS for 1 hour, the samples were incubated at 4 °C overnight with an antibody against TIA1 (1:100). Alexa Fluor 555 donkey anti-Rabbit (Beyotime, China) was used as the secondary antibody (1:50) and incubated for 1.5 hours at room temperature. Nuclei were stained with DAPI (10 mg/mL) after washing. Images were obtained by fluorescence microscope (Olympus,Japan, FU1200). Image analysis was conducted with Image J software, V1.52a (https://imagej.nih.gov/ij/download.html).

Western blotting

Total cellular protein extracts were obtained by lysis cell in lysis buffer contained 50 mmol/L Tris-HCl, 150 mmol/L NaCl, 0.02% sodium azide, 100 ng/ml phenylmethylsulfonyl fluoride, 0.1% SDS, 1%NP-40, and complete proteinase inhibitor mixture) as described previously[23]and were separated on 10%SDS-polyacrylamide gel and transferred to PVDF membranes. After blocking in 5% skimmed milk for 1 hour, the membranes were incubated with a primary antibody overnight at 4 °C. The membranes were then washed 3 times for 10 minutes in Tris-buffered saline containing Tween-20 (TBST) and incubated with an HRP-conjugated secondary antibody (Jackson Immuno-Research, USA) for 1 hour at room temperature. After washing 3 times for 10 minutes in TBST, the membranes were developed with an ECL detection system (Bio-Rad). Image analysis was conducted with Image J software, V1.52a (https://imagej.nih.gov/ij/download.html).

Animal experiments

Female nude mice (BALB/c-null, three-weeks-old)were purchased from the Model Animal Research Center of Nanjing University (Nanjing, Jiangsu,China) and bred in special pathogen-free (SPF)condition. MDA-MB-231R cells (1×107) were subcutaneously injected into the left axilla. After the tumor diameter had reached about 5 mm, the mice were divided randomly into two groups. PTX (10 mg/kg), DMSO (4 mg/kg) or EFV (20 mg/kg) were injected every 3 days by tumorin situinjection. The body weight and tumor sizes were measured every week. Tumor volumes were calculated using the formula: volume=0.5×(length×width2). After four weeks of treatment, mice were sacrificed and their tumors were removed, weighed, and photographed.This study was carried out in strict accordance with the guidelines for the care and use of laboratory animals of the National Institutes of Health. Our protocol was approved by the Committee on the Ethics of Animal Experiments of Nanjing Medical University.

Statistical analysis

All experiments were repeated three times.Statistical differences between means were determined by the Students't-test or one-way ANOVA. Values are expressed as means±standard error of triplicate measurements.P<0.05 was considered statistically significant.

Results

PTX-induced LINE-1 expression reduced the sensitivity of cells to chemotherapy agents

Western blotting showed higher LINE-1 protein level in the PTX-resistant MDA-MB-231R cell line than in its parental MDA-MB-231S cell line (Fig. 1A).MDA-MB-231 cells were then treated with 100 nmol/L PTX for 48 or 72 hours, and harvested for qRT-PCR analysis. The PTX treatment caused a significant increase in the endogenousLINE-1mRNA level, which was 1.5 to 2 times higher than that in the control (Fig. 1B, left panel). A similar drug-increased LINE-1 level was also observed in the MDA-MB-468 cell line (Fig. 1B, right panel). However, PTX had no significant influence on LINE-1 promoter activity(Supplementary Fig. 1A, available online).

We then established clones stably overexpressing LINE-1. TheLINE-1mRNA levels in two LINE-1 overexpressing clones (L1-clone 1 and L1-clone 2)were nearly ten times higher than those in the control clone (Ctr-clone) (Supplementary Fig. 1B, available online). The protein levels in L1-clone 1 and L1-clone 2 were also higher than those in the Ctr-clone(Supplementary Fig. 1C, available online). The clones were subsequently used for downstream examination of drug responses. MTT results showed that compared to Ctr-clone, L1-clone 1 and L1-clone 2 were less sensitive to three chemotherapy drugs, PTX,EPI, and DDP (Fig. 1C). The half maximal inhibitory concentration (IC50) values and the resistance index(RI) of L1-clone 1 and L1-clone 2 were also increased(Table 1).

The LINE-1 inhibitor efavirenz (EFV)[25]was then used to further confirm the function of drug increased LINE-1 in drug resistance. The mRNA and protein levels of LINE-1 were significantly inhibited by EFV in MDA-MB-231R cells (Supplementary Fig. 1DandE, available online). When combined with EFV, all three anticancer drugs significantly decreased cell viability and showed lower IC50values than cells not treated with EFV (Fig. 1DandTable 2).

Table 1 The IC50 values of PTX, EPI, DDP in control and LINE-1 overexpression clones after 72 hours treatment with drugs

Table 2 IC50 values of PTX, EPI and DDP with or without EFV in MDA-MB-231R after 72 hours treatment

Inhibition of LINE-1 by EFV increased the PTX sensitivity of MDA-MB-231R xenograft tumors

We determined whether inhibition of LINE-1 by EFV sensitized the drug-resistant breast cancer cells to PTXin vivousing a xenograft tumor model. We transplanted 1×107MDA-MB-231R cells into female nude mice. When the tumors reached approximately 5 mm in diameter, the mice were randomly divided into two groups: a paclitaxel treatment group (10 mg/kg PTX plus 4 mg/kg DMSO) and a group treated with a combination of paclitaxel and EFV (10 mg/kg PTX plus 20 mg/kg EFV). Each group of mice was injected with drugs every three days. After four weeks of treatment, the mice were sacrificed and their tumors were removed, weighed and photographed. As showed inFig. 2A, the tumors of the group given the combined treatment grew more slowly than the tumors of the group that received PTX alone. Consistently,the weight and size of the tumors were significantly decreased in the combined treatment group (Fig. 2BandC).

Fig. 1 Paclitaxel-induced LINE-1 expression reduced the sensitivity of cells to chemotherapy agents. A: Whole cell lysates were analyzed by Western blotting for LINE-1 protein levels in MDA-MB-231S and MDA-MB-231R cells. Antibody against ORF2p of LINE-1 was used to detect LINE-1. B: MDA-MB-231S cells were treated with 100 nmol/L PTX for 48 or 72 hours. MDA-MB-468 cells were treated with 50 nmol/L paclitaxel (PTX) for 24 or 48 hours. LINE-1 mRNA level was normalized to ACTB mRNA level. C: LINE-1 overexpressing clones (L1-clone 1 and L1-clone 2) and control clone (Ctr-clone) cells were treated with an increasing concentration of PTX, epirubicin(EPI) or cisplatin (DDP) for 72 hours, and then cell viability was determined by the MTT assay. D: EFV reversed the PTX, EPI, and DDP resistance of MDA-MB-231R cells. MDA-MB-231R cells were treated with 45 μmol/L EFV or DMSO (vehicle control) in combination with increasing concentrations of PTX, EPI or DDP for 72 hours, as determined by cell viability assays. Mean values±SD (error bars) was shown for three independent experiments. Paired t-test was used to analyze the significance difference between the groups. *P<0.05. DMSO:dimethyl sulfoxide.

PTX stress induced the formation of SGs and decreased sensitivity of TNBC cells to chemotherapy drugs

Immunofluorescence confirmed that PTX treatment induced the formation of SGs. Treatment of cells for 24 hours with 100 nmol/L PTX resulted in SG formation in approximately 40% of the cells (Fig. 3A)and was accompanied by a 2-fold increase in the protein level of TIA1 (Supplementary Fig. 2A,available online). PTX also induced SG formation in MDA-MB-468 cells, another TNBC cell line(Supplementary Fig. 2B, available online). The number of SG-positive cells following PTX treatment was decreased from 40% to 11% by interference with TIA1 (Fig. 3B). The interference efficiency of TIA1 was shown in theSupplementary Fig. 2C(available online).

Fig. 2 Inhibition of LINE-1 by EFV increased the sensitivity of MDA-MB-231R xenograft tumors to PTX. Transplantation of 1×107 PTX-resistant MDA-MB-231R cells into the left armpit of female nude mice generated xenograft tumors. When the tumors were about 5 mm in diameter, the mice (n=4) were treated with PTX (10 mg/kg PTX plus 4 mg/kg DMSO) or a combination of PTX and EFV (10 mg/kg PTX plus 20 mg/kg EFV). A: Growth curves of the xenograft MDA-MB-231R tumors in nude mice calculated from the tumor volumes. Mean values±SD (error bars) was shown. **P<0.01, n=4. B: Average tumor weight after drug treatment. Mean values±SD (error bars) was shown.*P<0.05, n=4. C: Images of MDA-MB-231R xenograft tumors. Statistics were performed by one-way ANOVA. PTX: paclitaxel; EFV:efavirenz; DMSO: dimethyl sulfoxide.

Fig. 3 PTX-induced SGs decreased the sensitivity of TNBC cells to chemotherapy drug. A: Representative images of SGs induced by a 24 hours treatment of MDA-MB-231 cells with 100 nmol/L PTX (left panel). The percentages of SG-positive cells were calculated (right panel). The yellow arrow points to SGs. Each independent experiment included an analysis of at least 50 cells per condition. B: MDA-MB-231 cells transfected with the TIA1 interference plasmid or a control plasmid, followed by 24 hours treatment with PTX.Immunofluorescence were then performed. The left panel is a representative image of the formed SGs, and the right panel is the calculated percentage of cells with SGs. C: Cell viability was tested after transfection of MDA-MB-231S cells with pCMV6-Entry vector (Ctr) or TIA1 expression plasmids (Ex-TIA1) and exposure to different concentrations of PTX for 72 hours. D: Viability of MDA-MB-231R cells was tested after 24 hours transfection with PLKO.1 vector (Ctr) or TIA1 interference plasmids (Sh-TIA1), followed by treatment with increasing concentrations of PTX for 72 hours. E: Kaplan-Meier analysis revealed a negative correlation between TIA1 and recurrence free survival(RFS, HR=1.62, P=0.0016, left panel) or distant metastasis-free survival (DMFS, HR=1.96 P=0.034, right panel) in breast cancer patients who had undergone chemotherapy but not endocrine therapy. F: A negative correlation between G3BP1 mRNA level and RFS (HR=1.88,P=0.027). G: A negative correlation between TIAL1 mRNA level and RFS (HR=1.4, P=0.01, left panel) or DMFS (HR=1.42, P=0.043, right panel). Mean values±SD (error bars) was shown for three independent experiments. Statistics were performed by t-test. *P<0.05, **P<0.01.PTX: paclitaxel; DMSO: dimethyl sulfoxide.

We also explored whether the PTX-induced SGs would cause breast cancer cells to develop chemotherapy resistance. MDA-MB-231S cells were transfected with TIA1 expression plasmids or vectors following 48 hours of PTX treatment and then harvested for cell viability test. MTT results showed TIA1 overexpression reduced the sensitivity of cells to PTX (Fig. 3C). The overexpression efficiency of TIA1 was shown inSupplementary Fig. 2D(available online). By contrast, interference with TIA1 expression sensitized PTX-resistant MDA-MB-231R cells to PTX (Fig. 3D). At last, we analyzed the relationship between expression level of SG marker protein TIA1, G3BP1, TIAL1 and clinical outcomes using a free online survival analysis tool Kaplan-Meier Plotter[26]. Our analysis included patients who received chemotherapy only, and excluded patients who received endocrine therapy,etc. As showed inFig. 3E, breast cancer patients with high TIA1 expression had worse recurrence free survival (RFS)and distant metastasis-free survival (DMFS). A negative correlation between RFS or DMFS with another two SG marker protein TIAL1 and G3BP1 was also obtained (Fig. 3FandG). These results indicated that the induction of SGs by chemotherapy drugs could reduce the subsequent responses of breast cancer cells to chemotherapy agents.

PTX-induced SGs recruited LINE-1 mRNA in TNBC cells

The predictions from the catRAPID website[27]indicated that TIA1 had a strong interaction with LINE-1 5′UTR and a small affinity with its 3′UTR(Supplementary Fig. 3, available online). The predicted TIA1 binding region was mainly in the amino acids 111 to 162, corresponding to the RNA recognition motif 2 (RRM2) domain (its main RNA binding domain) (Supplementary Table 1, available online). The RPISeq website[28]predicted an interaction probability for LINE-1 RNA and TIA1 between 0.7 and 0.98. The interaction probability generated by RPISeq ranged from 0 to 1 and a probability >0.5 was considered positive. The prediction results for both websites indicated that TIA1 had a strong tendency to bind LINE-1 RNA.The RNA fluorescencein situhybridization and protein immunofluorescence results showed that LINE-1 RNA was localized with TIA1, and that the average co-localization was significantly higher in the PTX-treated group (36.9%) than in the control group(9.6%). By contrast, TIA1 interference reduced the percentage of cells with LINE-1 RNA localized in SGs to 21.7% in PTX-treated cells (Fig. 4A).Consistently, a high TIA1 level led to the assembly of SGs as well as a strong localization of LINE-1 RNA within the SGs (Fig. 4B). These results indicated a recruitment of LINE-1 RNA to SGs under PTX stress.

PTX-induced SGs increased LINE-1 mRNA stability and level

PTX treatment increased the half-life ofLINE-1mRNA as well as the mRNA level (Fig. 5A).Overexpression of TIA1 also increased the stability and level ofLINE-1mRNA (Fig. 5B), while TIA1 interference decreased theLINE-1mRNA stability following PTX exposure (Fig. 5C). The same reversal of PTX-enhancedLINE-1mRNA stability by TIA1 interference was observed in the MDA-MB-468 TNBC cell line (Supplementary Fig. 4AandB,available online). Moreover, interference of TIA1 partly inhibited the elevation ofLINE-1mRNA level induced by PTX (Fig. 5D), indicating that PTXinduced SGs aid in stabilizingLINE-1mRNA during drug stress.

EFV reduced the stability of LINE-1 mRNA by inhibiting the formation of SGs

EFV, an inhibitor of LINE-1, suppressed the increase inLINE-1mRNA expression induced by PTX. We testified whether EFV reduces the level of LINE-1 by inhibiting the formation of SGs under PTX stress. The proportion of cells containing SGs was 44% following treatment with PTX, but treatment with EFV in combination with PTX reduced this proportion to 6% (Fig. 6A), indicating that EFV treatment significantly inhibited the formation of SGs under PTX stress. RNA stability experiments further confirmed that EFV significantly reversed the stability ofLINE-1mRNA increased by PTX (Fig. 6B), indicating that EFV can reduce the stability ofLINE-1mRNA by inhibiting the PTX-induced formation of SGs.

Discussion

In the present study, we demonstrated that exposure of TNBC cells to the chemotherapy agent PTX could induce the formation of SGs and that SG formation enhanced drug resistance by stabilizingLINE-1mRNA. This work has therefore established a significant functional link between SGs, abnormal LINE-1 expression, and drug resistance. Our studies have several implications.

First, SGs are an important regulator of theLINE-1mRNA half-life during chemotherapy agent stress in TNBC cells. The expression of LINE-1 can be regulated at both the transcriptional and posttranscription levels in response to a variety of cellular environments[29–31]. RNA modifications or microRNAs were reported to regulate the half-life ofLINE-1mRNA in cancer cells. For example, LINE-1 RNA stability is negatively regulated by TUT4/7-mediated 3'uridylation[32], whereas miR-128 directly binds to the 3′UTR of LINE-1 RNA to induce mRNA degradation and further decrease its expression and retrotranposition[33–34]. However, compared with the regulation of LINE-1 transcription in cancer cells, studies on theLINE-1mRNA half-life under chemotherapy stress are very limited. Here we demonstrated that SGs induced by the chemotherapy agent PTX could recruitLINE-1mRNA, thereby increasing the mRNA halflife and expression level (Fig. 4A,Fig. 5A). The use of interference TIA1 or treatment with EFV to inhibit SG formation also reduced the stability of LINE-1(Fig. 3B,Fig. 5C,Fig. 6AandB). These results identified SG as a new factor that regulatesLINE-1mRNA stability under drug stress.

Fig. 4 PTX-induced SGs recruit LINE-1 mRNA in TNBC cells. A: Representative images showing that LINE-1 RNA colocalized with SGs (left panel) and the calculated percentages of colocalization-positive cells (right panel). MDA-MB-231 cells were transfected with PLKO.1 vector (Ctr) or TIA1 interference plasmids (Sh-TIA1) for 24 hours, followed by 24 hours treatment with DMSO (vehicle control) or 100 nmol/L PTX. Colocalization of LINE-1 RNA with SGs was determined by immunofluorescence and RNA FISH. The yellow arrow points to the colocalization of LINE-1 RNA with SGs. B: Representative images showed significant colocalization of LINE-1 RNA with SGs in cells transfected with pCMV6-Entry vector (Ctr) or TIA1 overexpression plasmids (Ex-TIA1) (left panel). The percentages of colocalized cells were calculated (right panel). The yellow arrow points to the colocalization of LINE-1 RNA with SGs. Mean values±SD (error bars)was shown for three independent experiments. Each independent experiment included an analysis of at least 50 cells per condition. Statistics were performed by t-test. *P<0.05. PTX: paclitaxel; DMSO: dimethyl sulfoxide.

Two possibilities might explain how SGs regulateLINE-1mRNA stability. One is that mRNA stability is directly regulated by the TIA1 component. Some RNA binding proteins bind to the poly(A) tails of target RNA, and this may protect the proteins from undergoing uridylation and decay[32]. TIA1 is an RNA binding protein that has three RNA RRMs[35]. TIA1 also has binding sites in the LINE-1 RNA 3′ poly (A)tail (Supplementary Fig. 3), indicating that it is capable of binding LINE-1 RNA. In partial support of this, Meyeret alfound that double knockout TIA1 and TIA1-like1 (TIAL1) decreased mRNA stability of target genes that have TIA1 binding sites located in their 3′UTRs[36]. Another possibility is the stability of LINE-1 RNA is regulated by protective mRNPs formed in SGs. Bleyet alpointed out that SGs recruited proteins and mRNAs to form protective messenger ribonucleoprotein complexes (mRNPs)that were indispensable in preventing mRNA degradation[37]. Our results confirmed the transport of LINE-1 RNA into SGs during PTX stress (Fig. 4A).The mRNA half-life of LINE-1 decreased after disruption of SG formation by interference TIA1(Fig. 3BandFig. 5C). Goodieret alfound that the LINE-1 ORF1p colocalized with a known RNA stabilizer protein YB-1 in SG loci[38]. Since the LINE-1 protein has acispreference for its RNA, a reasonable explanation is that LINE-1 RNA, in combination with YB-1, ORF1p, ORF2p, and other proteins, forms a protective mRNP in SGs. The recruitment ofLINE-1mRNA into SGs may increase its recognition by other RNA stabilizer proteins that can form protective mRNPs in SGs. Regardless of the explanation, our results highlighted that SGs can maintainLINE-1mRNA stability during chemotherapy drug stress.

Fig. 5 PTX-induced SGs increased LINE-1 mRNA stability and level. A: MDA-MB-231 cells were treated with DMSO or 100 nmol/L PTX for 48 hours, followed by 2, 4, or 6 hours treatment with 2 mol/L actinomycin D to inhibit transcription. qRT-PCR was used to analyze the decay of mRNAs (left panel). The relative level of LINE-1 mRNA was calculated in cells without actinomycin D treatment (right panel).B: Cells were transfected with TIA1 overexpression (Ex-TIA1) or pCMV6-Entry (Ctr) plasmids for 24 hours, followed by transcription inhibition with 2, 4, or 6 hours treatment with 2 μmol/L actinomycin D. Total RNA was then extracted for qRT-PCR (normalized to ACTB mRNA levels). The left panel shows the relative stability of LINE-1 mRNA; the right panel showed the level of LINE-1 mRNA. C: Cells were transfected with PLKO.1 (Ctr) or TIA1 interference (Sh-TIA1) plasmids for 24 hours, followed by 48 hours exposure to PTX and then exposure to actinomycin D for different times. qRT-PCR was used to analyze the decay of mRNAs (normalized to ACTB mRNA levels).D: Cells were transfected with PLKO.1 (Ctr) or TIA1 interference plasmids for 24 hours and then treated with DMSO or 100 nmol/L PTX for another 48 hours. The relative mRNA levels of LINE-1 were then measured by RT-PCT (normalized to ACTB mRNA levels). Mean values±SD (error bars) was shown for three independent experiments. Statistics were performed by t-test. *P<0.05, **P<0.01. PTX: paclitaxel;DMSO: dimethyl sulfoxide.

Fig. 6 EFV reduced the stability of LINE-1 mRNA by inhibiting the formation of SGs. A: MDA-MB-231 cells were treated with DMSO (vehicle control) or 100 nmol/L PTX with or without 45 μmol/L EFV for 24 hours, followed by fixation for immunofluorescence.Representative images of the formed SGs (left panel). The percentages of SG-positive cells were calculated (right panel). B: Cells were treated with DMSO, PTX (100 nmol/L), or a combination of PTX and EFV for 24 hours, followed by treatment with 2 μmol/L actinomycin D for 2, 4, or 6 hours to inhibit transcription. qRT-PCR was used to analyze the decay of mRNAs (normalized to RPLP0 mRNA levels).Mean values±SD (error bars) was shown for three independent experiments. Statistics were performed by t-test. *P<0.05, **P<0.01. PTX:paclitaxel; EFV: efavirenz; SGs: stress granules; DMSO: dimethyl sulfoxide.

A second implication is that SGs induced by chemotherapeutic drugs contribute to in breast cancer cell drug resistance. Although some chemotherapeutics, such as sorafenib, can induce the production of SGs in cancer cells, our finding that the first-line breast cancer chemotherapeutic, PTX, can induce SGs in TNBCs has not been previously reported. Our study proved that PTX can induce the production of SGs in TNBC cells (Fig. 3A,Supplementary Fig. 2B, available online). The formation of SGs is usually triggered by phosphorylation of the translation initiation factor eIF2α[14]. The enzymes that phosphorylate eIF2α differ depending on the chemotherapy agent. For example,sorafenib functions by PERK-mediated phosphorylation, while bortezomib functions by HRI-mediated phosphorylation, of eIF2α to induce SGs[39].

Changes in other translation complexes also trigger the formation of SGs. For example, hydrogen peroxide inhibits translation initiation by disrupting the interaction between eIF4E and eIF4G, thereby initiating the assembly of SGs[15]. Similarly, melamine A interferes with translation initiation by blocking the eIF4A helicase required by the ribosome to start the assembly of SGs[40]. In our study, we did not find significantly changed phosphorylation levels for eIF2α after the cells were exposed to 100 nmol/L PTX at different time points (data not shown). This suggested that PTX may induce the assembly of SGs by affecting other translation components, and further experimental investigation is needed.

Consistent with studies on other cancers[17–18], we also confirmed that the SGs induced by chemotherapeutic agents also promoted drug resistance in TNBC cells. We found that interference TIA1 reduced the formation of SGs and improved the sensitivity of drug-resistant breast cancer cells to PTX (Fig. 3BandD). In particular, we found that the PTX-induced SGs promoted breast cancer resistance by improving the stability and level ofLINE-1mRNA. Two possible mechanisms might explain how the SG stabilization of LINE-1 promoted drug resistance. One is thatLINE-1mRNA was recruited into the SGs to provide transient protection. The stabilized mRNA then was exchanged quickly and translated to promote cell survival after the relief of the drug stress. This was supported by the significantly increased LINE-1 protein level after PTX was removed 8 hours later (data not shown).Chemotherapy is a cyclical administration process,which means that the stress is switched on and off periodically, so this would cause a gradual accumulation of LINE-1 levels. The elevated LINE-1 protein level in MDA-MB-231R cells (Fig. 1A),which were established by mimic clinical chemotherapy, supports this hypothesis. Another possible mechanism is that the LINE-1 RNA may help to maintain SG structures by forming RNA guanine quadruplexes (G-quadruplexes). The mammalian LINE-1 3′UTR contains conserved guanine-rich sequences that can fold into G-quadruplexes[41]. In addition to protein-protein and protein-RNA interactions, RNA-RNA interactions are also an important components of SG structures. Gquadruplexes are noncanonical secondary structures formed by guanine-rich nucleic acids and stabilized by the stacking of guanine tetrads held together by Hoogsteen base pairing. Guanosine can transform into hydrogel-like state structures[42–43]that aid in SG condensation[44]. SG proteins like YB1, hnRNPA1 and FUS have been reported to interact with G quadruplexes[45]. LINE-1 RNA was transported to SGs under stress (Fig. 4A), and high concentrations of RNA would promote their self-stacking into Gquadruplexes by the guanine-rich 3′UTR. This structure not only stabilizes itself but it also helps to maintain the SG structure. However, this is just a reasonable hypothesis and still needs further experiments for verification.

A third implication is that EFV can also inhibit the formation of drug-induced SGs. EFV is a nonnucleoside reverse transcriptase inhibitor that functions by directly binding to reverse transcriptase to inhibit its activity. In addition to being used as an inhibitor of HIV-1 encoding RT, it can also effectively inhibit the RT activity of the LINE-1 ORF2p, thereby inhibiting the transposition of LINE-1 and reducing the proliferation of cancer cells[46–47].Invivoexperiments and clinical studies have shown that EFV can reduce the growth of tumor cells[25,48]; however,whether it also enhances the effect of chemotherapeutics and reverses drug resistance is not clear. Our results proved that EFV can significantly enhance the effects of multiple chemotherapeutics, including paclitaxel, cisplatin, and epirubicin (Fig. 1D), which compensates the deficiencies of previous studies. The combination of EFV and chemotherapeutics also appears to be a promising strategy for reversing the cell resistance phenotype. In addition, our research also revealed a new mechanism by which EFV may inhibit LINE-1 levels and exert an anti-tumor effect,as EFV can reduce the mRNA stability of LINE-1 by inhibiting the SGs induced by chemotherapeutics(Fig. 6AandB).

The possible mechanism used by EFV to inhibit the formation of SGs is autophagy induction. SGs are disassembled by translation recovery, chaperonemediated disaggregation, and autophagy[14].Autophagy eliminates SGs through a process called granulophagy, which is mainly mediated by the ATPase ACP[49]. EFV can induce autophagy under certain experimental conditions and can show time and concentration dependence. For example, Bellisaiet alfound that 20 μmol/L EFV induced autophagy in the PC3 prostate cancer cell line after 96 hours of treatments, but it did not induce autophagy in PNT2 normal prostate epithelial cells[50]. Phillipet alfound that 5 μmol/L EFV treatment for 4 hours and 24 hours induced autophagy in SH-SY5Y neuroblastoma cells[51], while Apostolovaet alfound that after 24 hours treatment with 10 to 50 μmol/L EFV induced autophagy in a concentration-dependent manner in Hep3B the human liver cancer cell line[52]. The NDP52 and p62 autophagy receptors can mediate the degradation of LINE-1 RNA, and this may further support the possibility that EFV can inhibit SG stabilization of LINE-1 by inducing autophagy[53].Whether EFV can inhibit SGs by inducing autophagy in breast cancer cells, and whether this inhibitory effect has tissue and cell specificity, is worth further study.

The final implication is that SGs may also regulate LINE-1 in the tumorigenesis and cancer progression.Tumor cells are well known to undergo challenges like hypoxia, hypertonicity, and nutritional deficiencies due to high replication rates[54]. SGs formation is helpful for the adaption and survival of cancer cells in these unfavorable environments[40,55].We found thatLINE-1mRNA was stabilized by the SGs induced by chemotherapy agent PTX. This mechanism may also be applicable in SGs triggered by hypoxia, hypertonicity, and nutritional deficiencies during tumor initiation and progression. Further investigation and verification of the relationship between LINE-1 and SGs in different tumor stages or subtypes would be meaningful for understanding the function of LINE-1 and the pathology of cancers. The role of SGs in LINE-1 post-transcriptional regulation in different cancers and contexts is also worthy of attention.

In conclusion, our results highlighted a new function for SGs in the post-transitional regulation ofLINE-1mRNA that could be an important survival mechanism for cancer cells exposed to chemotherapy drugs. The findings also enhance our understanding of the mechanisms of chemotherapy resistance and provide evidence of developing new chemotherapeutic strategies. We believe that the crosstalk between LINE-1 and SGs represents a promising and understudied field.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (Grant No. 82072580 and No. 81572789). We thank all of our colleagues in the laboratory for their assistance during the progression of this study.