Atractylenolide-I (AT-I, purity > 98%) and paclitaxel (purity > 98%) were purchased from Selleck Chemicals (Houston, TX, USA). A human CTGF/CCN2 DuoSet ELISA kit and recombinant human CTGF protein (rCTGF) were purchased from R&D Systems (Minneapolis, MN, USA). AT-1 was used at a dose of 0-100 μM (17). rCTGF was used at a dose of 0-20 μg/mL (38). TRIzol reagent was bought from Invitrogen (Carlsbad, CA, USA). M-MLV reverse transcriptase was bought from Promega (Madison, WI, USA). SYBR Green Real-time PCR Master Mix was bought from TOYOBO (Osaka, Osaka Prefecture, Japan). Puromycin was bought from Coolaber Technologies (Beijing, China). Anti-CTGF, anti-αSMA and anti-FAP antibodies were obtained from Abcam (Cambridge, UK). Anti-β-tubulin antibody was obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). HRP-conjugated goat anti-rabbit IgG and goat anti-mouse IgG antibodies were obtained from OriGene Technologies (Rockville, MD, USA). Immobilon Western Chemiluminescent HRP Substrate was bought from Millipore (Billerica, MA, USA).

MDA-MB-231 cells were purchased from ATCC (Manassas, VA, USA). HS578T cells were purchased from the Cell Resource Center, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences (Beijing, China). HFF1 and WI-38 cells were purchased from Shanghai Zhong Qiao Xin Zhou Biotechnology (Shanghai, China). MDA-MB-231, HS578T and WI-38 cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) (39, 40). HFF1 cells were cultured in DMEM supplemented with 15% FBS. FBS was bought from Gibco (Carlsbad, CA, USA).

Total RNA was isolated using TRIzol reagent according to the manufacturer’s protocol (41). Two microgram of total RNA was reverse transcribed using M-MLV reverse transcriptase. qRT-PCR was performed on ABI StepOne Real-Time PCR System using SYBR Green Real-time PCR Master Mix. The gene-specific primer sequences are listed in Table 1 . Gene expression levels were normalized to GAPDH. The 2^–ΔΔCt method was used for relative quantification (42).

To show changes in mRNA levels via heatmap, the normalized log2 ratios (mRNA levels with ATL-1 treatment/mRNA levels without ATL-1 treatment) of each gene were calculated. The relative mRNA levels obtained by qRT-PCR were used to calculate the normalized log2 ratios. The calculation process was as follows:

The relative mRNA levels of each gene were calculated by 2^–ΔCt method, where

Then the log2 ratios (mRNA levels with ATL-1 treatment/mRNA levels without ATL-1 treatment) were as follows (42):

where

It should be noticed that, each of three biological replicates were performed for the group with ATL-1 treatment and the group without ATL-1 treatment. One of the biological replicate in group without ATL-1 treatment was chose as the standard, then the mRNA level of the other two biological replicates without ATL-1 treatment and the three biological replicates with ATL-1 treatment were all calculated the log2 ratio to this standard. The log2 ratio of the standard to itself was zero.

At last, all of the calculated log2 ratios were dealt with Z-score normalization:

G_i means the log2 ratio of one of the replicate of gene G. Mean and SD were calculated using the six calculated log2 ratios of gene G (three in group with ATL-1 treatment and three in group without ATL-1 treatment). The normalized log2 ratio was then shown in different colors in the heatmap, which was drawn by Microsoft Excel 2010 (Redmond, WA, USA). Student’s t test was used to determine the difference between the two groups. The gene-specific primer sequences used in the heatmap are listed in Supplementary Table S2 .

Equal amounts of total protein (20 μg/lane) were separated by SDS-PAGE and transferred to PVDF membranes with a pore size of 0.45 µm (Millipore, Billerica, MA, USA). After blocking with 5% nonfat milk at room temperature for 1 hour, the membranes were incubated at 4°C overnight with primary antibodies against CTGF (1:1000, ab209780), α-SMA (1:5000, ab124964), FAP (1:1000, ab207178) and β-tubulin (1:1000, TA-10), and then with HRP-conjugated goat anti-rabbit IgG (1:2000, ZB-2301) or goat anti-mouse IgG (1:2000, ZB-2305) antibodies for 1 h at room temperature. Finally, Immobilon Western Chemiluminescent HRP Substrate was used to visualize the blots with Bio-Rad ChemiDoc XRS system (Hercules, CA, USA). Protein expression levels were quantified with ImageJ software (NIH, Bethesda, MD, USA).

Cells were seeded in 25 cm² culture flasks at an appropriate density, resulting in 80% confluence within 16-24 hours. When the cell confluence reached 70-80%, rinsed the cell layer with PBS and cultured the cells in 5 mL fresh serum-free medium. Supernatants were harvested 24 hours later and used for subsequent ELISA. The secreted CTGF protein levels in the medium were measured with a Human CTGF ELISA kit. The total protein concentration was also examined, and then the concentrations of secreted protein of different samples were normalized to corresponding total protein (43).

The pLV-EF1α-CTGF-CMV-Puro lentiviral plasmid (CTGF-OE), negative control lentiviral plasmid (NC), CTGF-specific shRNAs (sh-CTGF), negative control shRNAs and the according lentivirus were purchased from Shanghai GeneChem Co. (Shanghai, China). Viral titres of approximately 1 × 10⁹ infectious units/ml were obtained. Lentivirus infection was performed with polybrene according to the manufacturer’s instructions. Briefly, cells were seeded in 12-well plates (3-5 × 10⁴ cells/well) and cultured for 16-24 hours. Then, the culture medium was replaced with 480 μL fresh culture medium and 20μL polybrene (25×) per well. Lentivirus was then added to the culture medium at MOI (multiplicity of infection) of 20 (MDA-MB-231 and HS578T cells) or 10 (HFF1 and WI-38 cells).

Cells were incubated for 16 hours with lentivirus. Then the culture medium containing lentivirus was replaced with 1 mL fresh culture medium per well for the continuous culture. The stably transfected cells were selected by 1 μg/mL puromycin for 7 days. The sequences of CTGF-specific shRNAs are listed in Table S3 .

Transwell migration assays were performed using a 24-well Transwell chamber with a pore size of 8 µm (Costar, San Diego, CA, USA). Cancer cells (1 × 10⁵) were maintained in serum-free culture medium in the upper chamber. Medium containing 10% FBS was placed in the lower chamber as a chemoattractant. Cancer cells were allowed to migrate for 24 h at 37°C in a humidified atmosphere containing 5% CO₂. Subsequently, cells that had failed to migrate were removed from the upper chamber with swabs; the remaining cells on the bottom side of the basement membrane were fixed with 4% paraformaldehyde and stained with 1% crystal violet (44). The cells in the lower portion of transwell membrane were counted. A random selection of 3–5 fields were photographed and counted under the microscope (Olympus BX40, Tokyo, Japan).

Co-culture was carried out as described by Su et al. (6). Co-culture experiments were performed using 6-well or 24-well Transwell chambers with a 0.4-µm pore size (Costar, San Diego, CA, USA). Taking 6-well Transwell chambers for example, TNBC cells (1 × 10⁵) were seeded in the lower chamber with 2.5 mL culture medium, and fibroblasts (1 × 10⁵) were seeded in the upper chamber with 1.5 mL culture medium. DMEM supplemented with 10% FBS was used for HS578T and WI-38 cells co-culture experiments. DMEM supplemented with 15% FBS was used for MDA-MB-231 and HFF1 cells co-culture experiments.

Cell Counting Kit-8 was bought from Beijing Solarbio Science & Technology Co. (Beijing, China). Cells were seeded onto 96-well plates or 24-well plates at a density of 1×10⁴ cells/mL with indicated treatment. Six to eight parallel wells were assigned to each group. At different time points, the culture medium was replaced with 100 μl fresh medium per 96 well or 500 μl fresh medium per 24 well. Then, CCK-8 solution (10 μl per 96 well or 50 μl per 24 well) was added, followed by incubation for 2 h at 37 °C. A multiplate reader (Flexstation^® 3, Molecular Devices, LLC., San Jose, CA, USA) was used to measure absorbance at 450 nm.

Adherent cells were digested with trypsin into a single-cell suspension. The cell suspension was mixed with a 0.4% trypan blue solution at a ratio of 9:1. The live and dead cells were counted within 3 minutes. The proportion of live cells was then calculated.

MDA-MB-231 cells (6 × 10⁶) were implanted into the fat pads of 6-week-old Balb/c mice. Tumor volumes were measured every three days. Tumor volume was calculated using the formula Volume (mm³) = (length × height²)/2. When the tumors reached approximately 3 mm in diameter, paclitaxel treatment was started at a dose of 10 mg/kg i.p. once per week. ATL-1 was given at 50 mg/kg i.p. once daily as described by Li Y et al. (18). After 6 weeks of treatment, all the animals were sacrificed. Partial fresh primary tumors were used for primary culture of fibroblasts as described by Calvo et al. (45). Partial primary tumors and mouse organs were fixed in 10% formalin and embedded in paraffin for subsequent analysis.

Primary culture was carried out as described by Calvo et al. (45), with some modifications. Tumor samples were cut into small pieces and digested by collagenase/dispase. Then, 44 μm nylon meshes were used to remove undigested tissue. After centrifuging at 1000 rpm/min for 10 min, the filtered cells were collected and then re-suspended in DMEM with 10% FBS and cultured in the culture dish. Thirty minutes later, the fibroblasts had already adhered to the culture dish, while other cell types remained in suspension (45). Afterward, the culture media were changed and fibroblasts were cultured for subsequent experiments.

Data were assessed using SPSS version 20.0 (SPSS Inc., Chicago, IL, USA). Adobe Illustrator CC 2018 and GraphPad Prism 5.0 (GraphPad Software Inc., San Diego, CA, USA) were used to represent the data. Student’s t test was used to determine the difference between each two groups. Error bars in the experiments indicate standard deviation (SD). Any values of P < 0.05 were considered statistically significant.

As shown in Figures 1A, B , we found that 50 μM and 100 μM ATL-1 significantly inhibited the migration of MDA-MB-231 and HS578T TNBC cells in wound healing assays and Transwell migration assays. We chose 50μM as the ATL-1 concentration for subsequent experiments. To mimic the tumor microenvironment, we co-cultured TNBC cells and fibroblasts (HFF1 and WI38 cells), and then examined the growth effects of ATL-1 on MDA-MB-231 and HS578T cells by CCK8 assays. We found 50 μM ATL-1 alone had no growth-inhibiting effect on TNBC cells cultured alone (–) or co-cultured with fibroblasts ( Figures 1C , S1A ). We then examined the sensitivity of the tumor cells to paclitaxel by CCK8 assays and trypan blue exclusion assays. We found that when TNBC cells were challenged with paclitaxel for 48 hrs, the inhibition rate of TNBC cells in co-cultured systems was lower than that of TNBC cells cultured alone (-)( Figures 1D , S1B ), and the proportion of survived TNBC cells in co-cultured systems was higher than that of TNBC cells cultured alone ( Figures 1E , S1C ). However, we found that when TNBC cells were challenged with paclitaxel in co-cultured systems, the inhibition rate of TNBC cells with 50 μM ATL-1 treatment was higher than that of TNBC cells without ATL-1 treatment ( Figures 1D , S1B ), and the proportion of survived TNBC cells with ATL-1 treatment was lower than that of TNBC cells without ATL-1 treatment ( Figures 1E , S1C ). These data suggested that ATL-1 inhibited tumor cell migration and increased the inhibitory effects of paclitaxel on TNBC cells in co-culture systems.

To reveal the potential mechanism by which ATL-1 inhibits TNBC cell migration, we examined differentially expressed migration-related genes in MDA-MB-231 TNBC cells by qRT-PCR after ATL-1 treatment. The normalized log2 ratio and p value for significantly downregulated genes were shown in Table S1 . We found that the CTGF mRNA level was significantly decreased after treatment with 50 μM ATL-1 for 24 hrs ( Figure 2A ). qRT-PCR and Western blotting verified the downregulation of CTGF at the mRNA and protein levels in MDA-MB-231 and HS578T TNBC cells following ATL-1 treatment ( Figures 2B, C ). ELISA verified the downregulation of the secreted level of CTGF ( Figure 2D ).

We wondered whether ATL-1 inhibits TNBC cell migration via CTGF. We found that recombinant CTGF (rCTGF) could rescue the downregulation of cancer cell migration by ATL-1 in wound healing assays and Transwell migration assays ( Figures 3A, B ). To further determine the role of CTGF, we overexpressed the expression of CTGF with CTGF-OE lentivirus in MDA-MB-231 and HS578T cells (CTGF-OE), and used the NC lentivirus as the negative control (NC) ( Figure 3C ). After being overexpressed, there was no difference in CTGF expression between with and without ATL-1 treatment ( Figure 3C ). We found that the reduction in migration induced by ATL-1 could be attenuated in CTGF-OE cells ( Figures 3D, E ). We also knocked down the expression of CTGF with shRNAs in MDA-MB-231 and HS578T cells ( Figure S2A ). We found that the reduction in migration induced by ATL-1 could be attenuated by CTGF-specific shRNAs ( Figures S2B, C ). These data further confirmed that CTGF mediated the inhibitory effect of ATL-1 on tumor cell migration.

Because CTGF is associated with the transformation of fibroblasts into CAFs and chemotherapy resistance (35), we examined the effect of ATL-1 on CTGF expression in fibroblasts. We found that CTGF mRNA and protein levels were significantly decreased by ATL-1 treatment in HFF1 and WI-38 fibroblasts ( Figures 4A, B ). ELISA verified the downregulation of the secreted level of CTGF ( Figure 4C ). To mimic fibroblast transformation in the tumor microenvironment, we co-cultured TNBC cells and fibroblasts for 2 days and then examined the mRNA and protein levels of the CAF markers FAP and α-SMA. Firstly, we examined the effect of recombinant CTGF (rCTGF) on the transformation of fibroblasts into CAFs in co-culture systems. We found that the treatment with rCTGF (higher than 5 μg/mL) for 2 days could significantly increase the mRNA levels of FAP and α-SMA in co-cultured systems, and we chose 10 μg/mL for the subsequence experiments ( Supplementary Figures S3A, B ). Then, we examined the effects of ATL-1 on fibroblasts transformation. We found that treatment with 50 μM ATL-1 for 2 days dramatically reduced the mRNA and protein levels of FAP and α-SMA in co-cultured systems, which could be rescued by rCTGF ( Figures 4D, E ; Supplementary Figure S3C ).

To study the role of CTGF in the sensitization of TNBC cells to chemotherapy, we co-cultured TNBC cells and fibroblasts and then examined the effects of ATL-1 and rCTGF on the sensitivity of the tumor cells to paclitaxel by CCK8 assays. All treatments lasted 48 hrs. We found that the addition of ATL-1 could obviously increase the inhibition rates of TNBC cells in the co-culture systems, which could be attenuated by rCTGF ( Figures 5A , S4A ). We then overexpressed CTGF expression with CTGF-OE lentivirus in HFF1 and WI-38 fibroblasts (CTGF-OE), and used the NC lentivirus as the negative control (NC) ( Figure 5B ). After being overexpressed, there was no difference in CTGF expression between with and without ATL-1 treatment ( Figure 5B ). We examined the sensitivity of tumor cells to paclitaxel with CCK8 assays and trypan blue exclusion assays. All treatments lasted 48 hrs. We found that when TNBC cells were co-cultured with NC fibroblasts, the IC50 of paclitaxel and the proportion of survived TNBC cells after paclitaxel treatment were reduced by ATL-1 treatment. While, when TNBC cells were co-cultured with CTGF-OE fibroblasts, the IC50 of paclitaxel and the proportion of survived TNBC cells after paclitaxel treatment were no longer reduced by ATL-1 treatment ( Figures 5C, D ; Figures S4B, C ). We also knocked down CTGF expression with shRNAs in HFF1 and WI-38 fibroblasts ( Figure S5A ) and found that the ATL-1-driven chemosensitization in co-cultured systems could be attenuated by CTGF-specific shRNAs acting on fibroblasts ( Figures S5B, C ). These data suggest that CTGF secreted by fibroblasts mediates the chemosensitizing effect of ATL-1 on TNBC cells.

Finally, we analyzed the effect of ATL-1 on chemosensitivity in vivo. We found that in model mice bearing subcutaneously inoculated MDA-MB-231 cells, paclitaxel treatment inhibited tumor growth and lung and liver micrometastases. ATL-1 dramatically enhanced these inhibitory effects ( Figures 6A, B ). However, ATL-1 alone had no effect on tumor growth and lung and liver micrometastases ( Figures 6A, B ). Western blotting showed that ATL-1 could reduce CTGF protein levels in primary tumors ( Figure 6C ). In addition, qRT-PCR revealed that primary cultured fibroblasts derived from tumors treated with ATL-1 alone (ATL-1) expressed lower mRNA levels of FAP and α-SMA than fibroblasts from control tumors (control) ( Figure 6D ). And primary cultured fibroblasts derived from tumors treated with ATL-1 in combination with paclitaxel (paclitaxel+ATL-1) expressed lower mRNA levels of FAP and α-SMA than fibroblasts from tumors treated with paclitaxel alone (paclitaxel) ( Figure 6D ).