Autophagic degradation of SQSTM1 inhibits ovarian cancer motility by decreasing DICER1 and AGO2 to induce MIRLET7A-3P

ABSTRACT

The relationship between macroautophagy/autophagy and miRNA in regulating cancer cell motility is not clearly delineated. Here, we found that induction of BECN1-dependent or -independent autophagy decreased ubiquitin-binding proteins SQSTM1/p62 and CALCOCO2/NDP52. Downregulation of SQSTM1 (but not CALCOCO2) led to a decrease of the miRNA-processing enzyme DICER1 and the miRNA effector AGO2. The autophagy-mediated reduction of levels of SQSTM1, DICER1 or AGO2 resulted in increased MIRLET7A-3P (but not MIRLET7A-5P or PRE-MIRLET7A miRNA) and suppressed ovarian cancer motility. The investigation of the MIRLET7A effects on cancer cell motility showed that synthetic MIRLET7A-3P (3 nM) inhibited, whereas MIRLET7A-5P (100 nM) increased cancer cell motility. Moreover, downregulation of MIRLET7A-3P with antisense of MIRLET7A-3P miRNA (MIRLET7A-3P inhibitor; 3 nM) reversed the nutrient depletion- and rVP1-mediated suppression of ovarian cancer cell motility. In addition, restoring SQSTM1, DICER1 and AGO2 with inhibition of autophagic degradation or overexpression of DICER1 and AGO2 reversed the autophagy-associated enhancement of MIRLET7A-3P and inhibition of motility. Examination of ovarian cancer tissue microarray further showed that the levels of SQSTM1, DICER1 and AGO2 in the tumor were higher than those in the non-tumor cells and negatively correlated with the levels of autophagy and MIRLET7A-3P. Our results demonstrated that induction of autophagy to decrease SQSTM1, DICER1 and AGO2 and increase MIRLET7A-3P is a potential therapeutic strategy for suppressing ovarian cancer cell motility.

Abbreviations: ACTB: actin beta; AGO2: argonaute 2, RISC catalytic component; ATG: autophagy related; BCIP/NBT: 5-bromo-4-chloro-3-indolyl-phosphate/nitro blue tetrazolium; BECN1: beclin 1, autophagy related; CALCOCO2/NDP52: calcium binding and coiled-coil domain 2; CQ: chloroquine; DICER1: dicer 1, ribonuclease III; EBSS: Earle balanced salt solution; FBS: fetal bovine serum; HGF: hepatocyte growth factor; MAP1LC3B/LC3B: microtubule-associated protein 1 light chain 3 beta; MIRLET7A: microRNA LET-7A: MIR16: microRNA 16; MIR29C: microRNA 29C; miRNA: microRNA; MMP: matrix metallopeptidase; PRE-MIRNA: precursor microRNA; PtdIns3K: class III phosphatidylinositol 3-kinase; PtdIns3P: phosphatidylinositol-3-phosphate; RISC: RNA-induced silencing complex; rVP1: recombinant foot-and-mouth disease virus capsid protein VP1; siRNA: small interfering RNA; SQSTM1/p62: sequestosome 1; WIPI: WD repeat domain, phosphoinositide interacting.

Introduction

Autophagy is a cellular clearance system involving the formation of a distinct structure called autophagosome. The MAP1LC3B/LC3B (microtubule associated protein 1 light chain 3 beta)-related autophagosome [1,2] can be synthesized after stimulating cells with nutrient depletion via the canonical pathway involving BECN1 (beclin 1, autophagy related), the class III phosphatidylinositol 3-kinase (PtdIns3K) [3] complex and several autophagy-related (ATG) proteins including ATG5 and ATG7 [2,4]. The autophagosome can also be formed independently of BECN1 or PtdIns3K noncanonically in several cell types by specific anti-tumor stimulants such as resveratrol [5], gossypol [6], and recombinant foot-and-mouth disease virus capsid protein VP1 (rVP1) [7].

The autophagic membrane nucleation of canonical autophagy (BECN1-dependent) and noncanonical autophagy (BECN1-independent) [5–7] is dependent on 2 phosphatidylinositol-3-phosphate (PtdIns3P)-binding proteins, WIPI1 (WD repeat domain, phosphoinositide interacting 1) and WIPI2 [8–10]. Recently, Dooley and colleagues [11] have further demonstrated that the WIPI2B splice variant of WIPI2, binds to ATG16L1 to recruit the ATG12–ATG5-ATG16L1 complex and positively regulate LC3B lipidation by conjugating LC3B-I (a cytosolic form of LC3B) to phosphatidylethanolamine to form LC3B-II, a key component of the autophagosome membrane and a well-known autophagic marker [12]. The autophagosome then fuses with the lysosome and degrades the engulfed cytosolic components with the help of autophagic receptor proteins including SQSTM1/p62 (sequestosome 1) [13–15] and CALCOCO2/NDP52 (calcium binding and coiled-coil domain 2) [16,17]. Whether these autophagic receptor proteins are distinctively associated with biological effects of autophagy, however, is not clearly defined.

Autophagy protects tumor cells against therapeutic stresses on the one hand and promotes cell death of some cancer cells on the other [18]. The effects of autophagy on tumorigenesis are, therefore, still under debate. Shen and colleagues report that autophagy markers such as LC3B and BECN1 are suppressed in malignant epithelial ovarian tumors and a low level of autophagy is associated with the development of tumor malignancy [19]. Recently, we have found that BECN1-dependent autophagy inducers such as nutrient depletion and the BECN1-independent inducer rVP1 [7] suppress the growth, motility and metastasis of the epithelial ovarian cancer cell line, SKOV3 [20]. However, how autophagy inhibits motility of ovarian cancer is still poorly understood.

Potential mediators of autophagy are microRNAs (miRNAs) that exhibit pleiotropic effects on cell processes. Some miRNAs such as MIR135A, MIR21 and MIR720 promote cancer cell migration [21–23], whereas other miRNAs such as MIR218, MIR29A, MIR200C and MIRLET7A (microRNA LET-7A) suppress cancer cell migration [24–27]. The biogenesis and gene targeting of miRNAs are regulated by the miRNA-processing enzyme DICER1 (dicer 1, ribonuclease III) [28,29] and the miRNA effector AGO2 (argonaute 2, RISC catalytic component) [30]. It has also been reported that a subset of unloaded DICER1 and AGO2 may disrupt the maturation of miRNA [31,32] in an autophagy- and CALCOCO2-dependent manner [33]. Although several miRNAs have been found to modulate autophagy and cancer development [34], it is unclear whether and how BECN1-dependent and -independent autophagy regulate miRNA biogenesis and homeostasis, as well as cancer migration, invasion and metastasis.

Because MIRLET7A miRNAs are suppressors of cancer motility [35–37], we first explored in this study the effects of BECN1-dependent (via nutrient depletion) and BECN1-independent autophagy (via rVP1) on MIRLET7A miRNAs. Our results in SKOV3 cells and the clinically-related KURAMOCHI [38] ovarian cancer cells revealed that induction of autophagy (via either nutrient depletion or rVP1) induced MIRLET7A-3P but not MIRLET7A-5P miRNA to inhibit migration of ovarian cancer cells. Autophagy-mediated MIRLET7A-3P induction in ovarian cancer cells was not via autophagic degradation of CALCOCO2 but via degradation of SQSTM1. The induction of MIRLET7A-3P was associated with SQSTM1 downregulation as well as a decrease of the miRNA-processing enzyme DICER1 and the miRNA effector AGO2 in ovarian cancer cells. Moreover, negative correlations between the levels of autophagy and MIRLET7A-3P with those of SQSTM1, DICER1 and AGO2 were demonstrated in human ovarian cancer tissue microarray.

Results

Induction of autophagy correlates with an increase of MIRLET7A-3P miRNA expression in ovarian cancer cells

To examine whether an increase of canonical and/or noncanonical autophagy could affect the level of MIRLET7A miRNA in ovarian cancer cells, we first utilized the BECN1-dependent autophagy inducer nutrient depletion (Earle balanced salt solution, EBSS) as well as the BECN1-independent autophagy inducer rVP1 to trigger autophagy in ovarian cancer cell lines including SKOV3 and KURAMOCHI. We found that treatment for 4 h with EBSS or rVP1, increased puncta formation and lipidation of LC3B in KURAMOCHI and SKOV3 cells as determined by immunofluorescence microscopy (Figure 1A) and immunoblotting (Figure 1B and S1A). Such an effect of EBSS is similar to what we have reported previously for serum starvation [7], substantiating the evidence that nutrient depletion by both serum starvation and EBSS treatment, causes activation of autophagy in ovarian cancer cells. Addition of the autophagy degradation inhibitor chloroquine (CQ) increased puncta formation and lipidation of LC3B and suppressed the EBSS- and rVP1-mediated autophagic flux as indicated by the reversal of autophagic degradation of SQSTM1 (Figure 1B and S1A).

Figure 1.

Nutrient depletion and rVP1 induce autophagy in KURAMOCHI ovarian cancer cells. (A) Cells were treated with Earle balanced salt solution (EBSS) or 3 μM rVP1 for 4 h in the presence or absence of 30 μM chloroquine (CQ) as indicated. Cells were then immunolabeled with anti-LC3B antibodies, followed by Alexa Fluor 488-conjugated anti-IgG (green). Nuclei were stained with DAPI (blue). Fluorescent images were acquired by confocal microscopy. Scale bar: 20 μm. LC3B puncta per cell are represented as means ± SEM in 50 to 100 cells/experiment in 3 independent experiments. (B) Cells were incubated with or without 30 μM CQ for 30 min and then treated with EBSS or 3 μM rVP1 for 2 to 4 h as indicated. Cell lysates were collected and analyzed by immunoblotting using anti-LC3B and anti-SQSTM1 antibodies. SQSTM1 was used as an autophagic flux marker. (C) Cells were transfected with 10 nM non-targeting (si-Cont) siRNA or siRNA targeting BECN1, LC3B, WIPI1 and WIPI2 as indicated and then treated with or without EBSS or 3 μM rVP1 for 6 h in the presence of 30 μM CQ. Cell lysates were collected and analyzed by immunoblotting using anti-BECN1, anti-WIPI1, anti-WIPI2 and anti-LC3B antibodies as indicated. ACTB was used as a loading control. Data represent means ± SEM of densitometric measurement of LC3B-II:LC3B-I and LC3B-II:ACTB from 3 independent experiments. (D) Cells were transfected with 10 nM of si-Cont siRNA or siRNA targeting BECN1 as indicated then treated with or without EBSS or 3 μM rVP1 for 4 h. Cells were collected and double-membrane autophagosome structures were observed with transmission electron microscopy and indicated by arrows. Scale bar: 0.5 μm. The autophagosome pointed by red arrows were shown in insets with scale 0.2 μm. Data represent means ± SEM of volume fraction of autophagic compartments. N.S., not significant, *P < 0.05, **P < 0.01, ****P < 0.0001.

Selective knockdown of BECN1 by BECN1 small interfering RNA (siRNA) suppressed BECN1-dependent EBSS- (Figure 1C and S1B left panel) but not BECN1-independent rVP1-mediated LC3B lipidation (Figure 1C and S1B, right panel). Electron microscopy substantiated that BECN1 siRNA inhibited EBSS-mediated but not rVP1-mediated autophagosome formation (Figure 1D). Because WIPI1, WIPI2, ATG5 and ATG7 regulate autophagy [2,4,5,7–10], we also examined the effects of WIPI1, WIPI2, ATG5 and ATG7 siRNA on nutrient depletion- and rVP1-mediated autophagy induction. Our results revealed that each siRNA was specific against its respective target as shown in Fig. S2. Nutrient depletion- and rVP1-induced LC3B lipidation was decreased by selective knockdown of ATG5, ATG7 (Fig. S3), WIPI1, WIPI2 or combination of WIPI1 and WIPI2 (WIPI1/2) (Figure 1C and S1B) with their respective siRNA as compared to knockdown effects when treated with the non-targeting siRNA control.

We then examined whether nutrient depletion or rVP1 could increase expression of MIRLET7A in ovarian cancer cells. As preliminary experiments revealed that long-term (≥ 24 h) treatment with EBSS tended to cause more SKOV3 cell death that could jeopardize the miRNA determination, we thus treated cells mainly with serum starvation instead of EBSS in our experiments involving long-term nutrient depletion. Our results showed that nutrient depletion and rVP1 increased MIRLET7A-3P (Figure 2A) but not MIRLET7A-5P (Figure 2B) in a time-dependent manner. The serum starvation-induced MIRLET7A-3P in SKOV3 cells reached plateau (about 5-fold) at 4 h then declined, whereas MIRLET7A-3P in KURAMOCHI cells kept increasing to 6-fold at 24 h after serum starvation. In comparison, rVP1 increased MIRLET7A-3P of SKOV3 and KURAMOCHI cells to about 6-fold at 24 h.

Figure 2.

Induction of autophagy increases MIRLET7A-3P miRNA expression in ovarian cancer cells. SKOV3 and KURAMOCHI cells were serum-starved or incubated with 3 μM rVP1 for various periods of time as indicated. RNA samples were collected and relative expression ratios of (A) MIRLET7A-3P (B) MIRLET7A-5P miRNA were determined using qPCR analysis. SKOV3 and KURAMOCHI cells were transfected with 10 nM (C) non-targeting (si-Cont) siRNA or siRNA targeting WIPI1, WIPI2, WIPI1 + 2 or (D) siRNA targeting ATG5, ATG7 and LC3B, as indicated. After transfection, SKOV3 cells were serum-starved for 6 h or incubated with 3 μM rVP1 for 24 h and KURAMOCHI cells were serum-starved or incubated with 3 μM rVP1 for 24 h. RNA samples were then collected and relative expression ratios of MIRLET7A-3P miRNA were determined using qPCR analysis. Data represent means ± SEM of 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Induction of autophagy increases MIRLET7A-3P expression in ovarian cancer cells

To elucidate whether the increase of MIRLET7A-3P by nutrient depletion or rVP1 is mainly via induction of autophagy, we downregulated the autophagic mediators WIPI1 and WIPI2 with their respective siRNA (Fig. S2) and then examined their effects on the expression level of MIRLET7A-3P. As preliminary experiments showed that si-WIPI1 and si-WIPI2 at 10 nM was more effective than those at lower concentrations (2.5 or 5 nM) in knocking down WIPI1 and WIPI2 (Fig. S2B), 10 nM of si-WIPI1 or si-WIPI2 was initially used. Our results showed that nutrient depletion (serum starvation) and rVP1 increased MIRLET7A-3P in the non-targeting control and such increase of MIRLET7A-3P expression was downregulated in cells transfected with WIPI1 or WIPI2 siRNA (Figure 2C). Similar findings were observed when ATG5, ATG7 and LC3B were treated with 10 nM or lower concentrations of their respective siRNA (Figure 2D and S4A). In comparison, MIRLET7A-3P precursor microRNA (PRE-MIRNA) i.e., PRE-MIRLET7A1 or PRE-MIRLET7A3 was not increased by nutrient depletion (serum starvation) and rVP1 (Fig. S4B) and treatment with LC3B siRNA did not have any effect on the level of PRE-MIRLET7A1 or PRE-MIRLET7A3 either (Fig. S4C). Moreover, northern blots for MIRLET7A-3P, MIRLET7A-5P and PRE-MIRLET7A demonstrated that serum starvation and rVP1 treatment increased mainly MIRLET7A-3P but not MIRLET7A-5P or PRE-MIRLET7A (Fig. S4D).

To exclude the possibility that the decrease of MIRLET7A-3P could be due to cell death, we examined the viability of the cells before and after autophagy knockdown. Our results showed that there is no significant difference in cell viability before and after knockdown of key ATG genes (Fig. S5). In view of the findings that knockdown of WIPI1, WIPI2, ATG5, ATG7 or LC3B with the same amount of their respective siRNA suppresses autophagy as determined by LC3B lipidation (Figure 1C, S1B and S3), these results collectively demonstrated that nutrient depletion- and rVP1-mediated increase of MIRLET7A-3P expression in ovarian cancer cells is most likely autophagy-dependent.

Induction of autophagy suppresses motility of ovarian cancer cell via increasing MIRLET7A-3P

Whether and how autophagy facilitates or modulates cancer migration/invasion is still under debate [39–42]. To examine the effect of autophagy induction on regulation of ovarian cancer cell migration, nutrient depletion and rVP1 were administered to SKOV3 and KURAMOCHI cells with or without knockdown of WIPI1, WIPI2, ATG5, ATG7 or LC3B. Time-lapse microscopy showed that nutrient depletion (serum starvation) and rVP1 reduced the dispersive capability (Figure 3A) and migration velocity (Figure 3B) of SKOV3 cells. The serum starvation- and rVP1-mediated inhibitory effects on migration were reversed when WIPI1, WIPI2, ATG5, ATG7 or LC3B were downregulated (Figure 3A and 3B).

Figure 3.

Induction of autophagy inhibits ovarian cancer cell migration. (A) SKOV3 cells were transfected for 48 h with 10 nM non-targeting (si-Cont) siRNA or siRNA targeting WIPI1, WIPI2, WIPI1 + 2, ATG5, ATG7 and LC3B, as indicated. After transfection, migration trajectories of cells in response to 3 μM rVP1 or serum starvation were observed by time-lapse microscopy for 24 h and displayed in diagrams drawn with the initial point of each trajectory placed at the origin of the plot. (B) Migration velocities of SKOV3 cells were counted according to serial phase-contrast images every 15 min for 24 h as described in Materials and Methods. Data represent means ± SEM of migration velocities of 20 cells with different treatments. (C) KURAMOCHI cells were transfected with 10 nM of si-Cont or siRNA targeting WIPI1, WIPI2, WIPI1 + 2, ATG5, ATG7 and LC3B for 72 h and then incubated with or without EBSS medium or 3 μM rVP1 for 24 h, as indicated. After incubation, cells were collected and their migration ability was determined by transwell migration assay. Data represent means ± SEM of 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Other than time-lapse microscopy, the motility of cancer cells can be determined by transwell migration assay [7]. As preliminary experiments showed that the motility of KURAMOCHI cells was too low to be detectable with time-lapse microscopy and KURAMOCHI cells were more sensitive to EBSS than serum starvation, we examined the effect of nutrient depletion on motility of KURAMOCHI cells mainly by using EBSS and transwell migration assay. Our results demonstrated that after knockdown of WIPI1, WIPI2, ATG5, ATG7 or LC3B, the inhibitory effect of nutrient depletion (EBSS) and rVP1 on cell migration of KURAMOCHI cells (Figure 3C) was indeed reversed, indicating that nutrient depletion and rVP1 inhibits ovarian cancer motility through increase of WIPI1-, WIPI2-, ATG5-, ATG7- and LC3B-dependent autophagy.

The MIRLET7 miRNA family members have tumor suppressive functions in a variety of cancers [35–37]. To examine whether autophagy-induced MIRLET7A-3P plays a critical role in autophagy-mediated inhibition of ovarian cancer motility, we first transfected cells with synthetic MIRLET7A-3P mimics and then measured cell trajectories and migration velocity using time-lapse microscopy. We found a significant decrease in either trajectory behavior (Figure 4A, upper-left panel) or migration velocity (Figure 4A, upper-right panel) in SKOV3 cells treated with 3, 10 or 100 nM of synthetic MIRLET7A-3P as compared to the negative control. Synthetic MIRLET7A-5P, on the other hand, increased both trajectory behavior (Figure 4A lower-left panel) and migration velocity (Figure 4A, lower-right panel) at higher concentrations (≥ 100 nM) but not lower concentrations (3 to 30 nM). In comparison, other members of the MIRLET7 family such as MIRLET7B-5P did not show any effect (Fig. S6). Similar contrast effects of MIRLET7A-3P and MIRLET7A-5P were obtained by using transwell migration assay in both SKOV3 and KURAMOCHI cells (Figure 4B).

Figure 4.

MIRLET7A-3P is required for autophagy-mediated inhibition of ovarian cancer cell migration. (A) SKOV3 cells were transfected with 3 or 100 nM Scrambled or 3 to 100 nM synthetic mimics of MIRLET7A-3P or MIRLET7A-5P. After 48 h transfection, serial phase-contrast images were taken by time-lapse microscopy every 15 min for 24 h and migration trajectories and velocities of cells were measured. Data represent means ± SEM of migration velocities of 20 cells with different treatments. (B) Scrambled (100 nM) or synthetic MIRLET7A-3P (3 nM) and MIRLET7A-5P (100 nM) were transfected into SKOV3 cells for 2 days and KURAMOCHI cells for 3 days, respectively. The transfected cells were collected and their migration ability was detected by transwell migration assay. (C) Cells were transfected with Scrambled (100 nM) or synthetic MIRLET7A-3P (3 nM) and MIRLET7A-5P (100 nM) for 3 days. Supernatants were collected and MMP2 activity was examined by gelatin zymographic analysis. Data represent means ± SEM of 3 independent experiments. (D) SKOV3 and KURAMOCHI cells were transfected with 5 nM non-targeting (si-Cont) siRNA or siRNA targeting WIPI2 and LC3B as well as Scrambled or synthetic mimics (3 nM) of MIRLET7A-3P for 48 and 72 h, respectively. After incubation, cells were collected and their migration ability was determined by transwell migration assay. Data represent means ± SEM of 3 independent experiments. (E) SKOV3 cells were transfected with Scrambled or synthetic antisense of MIRLET7A-3P miRNA (MIRLET7A-3P inhibitor) for 48 h and then treated with or without 3 μM rVP1 and serum starvation for 24 h, as indicated. The transfected cells were subjected to time-lapse microscopy and their migration behavior (upper panel) and velocities (lower panel) was detected. Data represent means ± SEM of migration velocities of 20 cells with different treatments. (F) KURAMOCHI cells were transfected with Scrambled or MIRLET7A-3P inhibitor (3 nM) for 72 h and then treated with or without 3 μM rVP1 and EBSS for 24 h, as indicated. The transfected cells were collected and their migration ability was detected by transwell migration assay. Data represent means ± SEM of 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Furthermore, as MMPs (matrix metallopeptidases) are key mediators of cancer cell migration and invasion [43,44], we analyzed whether MIRLET7A-3P mimics have any inhibitory effect on MMPs of ovarian cancer cells. Our results showed that MIRLET7A-3P (3 nM) mimics downregulated, whereas MIRLET7A-5P (100 nM) upregulated the MMP2 activity of ovarian cancer cells (Figure 4C). To further verify that MIRLET7A-3P is essential for the autophagy-mediated inhibition of cell migration, we knocked down autophagy with LC3B and WIPI2 siRNA and transfected the cells with non-targeting miRNA or MIRLET7A-3P mimics. Our results showed that MIRLET7A-3P (3 nM) inhibited cell migration even when autophagy was downregulated by WIPI2 and LC3B siRNA (Figure 4D). Moreover, 3 nM of MIRLET7A-3P inhibitor reversed the nutrient depletion- and rVP1-mediated suppression of SKOV3 and KURAMOCHI cell motility (Figure 4E and 4F). Taken together, these results demonstrated that nutrient depletion- and rVP1-mediated increase of autophagy upregulates MIRLET7A-3P to inhibit MMP2 and ovarian cancer cell migration.

The cellular level of MIRLET7A-3P is much lower than MIRLET7A-5P

As a passenger strand of MIRLET7A, MIRLET7A-3P is complementary to guide strand MIRLET7A-5P and may exist at levels different from MIRLET7A-5P [45]. We thus examined the relative abundance of these 2 strands in the ovarian cancer cells. We found that the level of MIRLET7A-5P was a hundred fold higher than MIRLET7A-3P in SKOV3 and KURAMOCHI cells when their autophagy was not activated (Table 1). Even in the presence of nutrient depletion- or rVP1-mediated autophagy activation when the level of MIRLET7A-3P was elevated, the relative ratio of MIRLET7A-5P to MIRLET7A-3P was still higher than 20 fold (Table 1).

Table 1.

Effects of serum starvation and rVP1 on relative expression ratios of MIRLET7A-5P to MIRLET7A-3P in SKOV3 and KURAMOCHI cells.

Treatment Cell	Control	Serum starvation	rVP1
SKOV3	426.6 ± 101	41.2 ± 10	135.3 ± 29
KURAMOCHI	101.0 ± 18	32.0 ± 2	20.6 ± 3

SKOV3 and KURAMOCHI cells were treated with or without serum starvation or 3 μM rVP1 for 24 h. RNA of cells were collected and cycle thresholds (CT) of MIRLET7A-5P and MIRLET7A-3P were determined by quantitative real-time PCR. Relative expression ratio of MIRLET7A-5P to MIRLET7A-3P was calculated and shown as 2^−ΔCT. ΔCT = CT of MIRLET7A-5P – CT of MIRLET7A-3P.

Autophagy-mediated increase of MIRLET7A-3P is via downregulation of SQSTM1 instead of CALCOCO2

Potential mediators that could be associated with autophagy-mediated increase of MIRLET7A-3P are autophagic receptors such as SQSTM1 [13–15,46] and CALCOCO2 [16,17]. We found previously that rVP1 and serum starvation increased autophagy and downregulated SQSTM1 in SKOV3 cells [7]. Here, we further found by immunofluorescence microscopy (Figure 5A) that rVP1 and EBSS downregulated puncta of not only SQSTM1 but also CALCOCO2, an ubiquitine-binding protein known to regulate homeostasis of miRNAs of HeLa cells via targeting the miRNA-processing enzyme DICER1 and the major miRNA effector AGO2 [33]. As expected, rVP1- and EBSS-mediated downregulation of SQSTM1 and CALCOCO2 was reversed by the autophagic degradation inhibitor CQ (Figure 5A). Interestingly, knockdown of SQSTM1 but not CALCOCO2 in ovarian cancer cells (Figure 5B) resulted in increase of MIRLET7A-3P (Figure 5C) and suppression of cell migration ability (Figure 5D). These effects of SQSTM1 knockdown were reversed by CQ (Figure 5E and 5F) at a concentration (10 μM) that was sufficient to affect EBSS- and rVP1-mediated autophagy of SKOV3 and KURAMOCHI ovarian cancer cells (Fig. S7). Moreover, overexpression of SQSTM1 reversed the rVP1- and EBSS-inhibited cell migration (Fig. S8). Taken together, these results demonstrated that induction of autophagy promotes degradation of SQSTM1 leading to increase in MIRLET7A-3P and inhibition of cancer cell motility.

Figure 5.

Downregulation of SQSTM1 but not CALCOCO2 is required for autophagy-mediated increase of MIRLET7A-3P and inhibition of migration in ovarian cancer cells. (A) KURAMOCHI cells were treated with EBSS or 3 μM rVP1 for 4 h in the presence or absence of 30 μM CQ. Cells were then immunolabeled with anti-SQSTM1 or anti-CALCOCO2 antibodies followed by rhodamine-conjugated anti-IgG (red) and Alexa Fluor 488-conjugated anti-IgG (green), respectively. Nuclei were stained with DAPI (blue). Fluorescent images were acquired by confocal microscopy. Scale bar: 20 μm. SQSTM1 and CALCOCO2 puncta per cell are represented as means ± SEM in 50 to 100 cells/experiment in 3 independent experiments. (B) Cells were transfected with 10 nM of non-targeting (si-Cont) siRNA or siRNA targeting SQSTM1 or CALCOCO2 as indicated. Cell lysates were collected and analyzed by immunoblotting. Blots are representative of 3 independent experiments. (C) RNA samples were collected and relative expression ratios of MIRLET7A-3P miRNA were determined using qPCR analysis. Data represent means ± SEM of 3 independent experiments. (D) KURAMOCHI cells were transfected with 10 nM of si-Cont or siRNA targeting SQSTM1 or CALCOCO2 as indicated for 72 h. The transfected cells were collected and their migration ability was detected by transwell migration assay. (E) Cells were treated with EBSS or 3 μM rVP1 for 6 h in the presence or absence of 10 μM CQ as indicated. Relative expression ratios of MIRLET7A-3P miRNA were determined using qPCR analysis. (F) KURAMOCHI cells were treated with EBSS or 3 μM rVP1 for 24 h in the presence or absence of 10 μM CQ as indicated and cell migration ability was detected by transwell migration assay. Data represent means ± SEM of 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Downregulation of SQSTM1 decreases DICER1 and AGO2 to increase MIRLET7A-3P and inhibit ovarian cancer motility

To evaluate how autophagy-mediated downregulation of SQSTM1 could increase passenger strand MIRLET7A-3P, we examined the relationship between SQSTM1 and the miRNA-processing enzyme DICER1 [28,29] as well as the major miRNA effector AGO2 [30] that may incorporate the guide strand miRNA (such as MIRLET7A-5P) into RISC (RNA-Induced Silencing Complex) and leave the passenger strand (such as MIRLET7A-3P) degraded [47]. We found that downregulation of SQSTM1 decreased DICER1 and AGO2 expression in ovarian cancer cells (Figure 6A). Upregulation of autophagy by nutrient depletion (EBSS) or rVP1 resulted in decrease of not only SQSTM1 but also DICER1 and AGO2 and such effects were reversed by autophagy degradation inhibitor CQ (Figure 6B). Moreover, knockdown of DICER1 and AGO2 (Figure 6C), like knockdown of SQSTM1, increased MIRLET7A-3P (Figure 6D) and inhibited motility of ovarian cancer cells (Figure 6E and 6F).

Figure 6.

Downregulation of SQSTM1 increases MIRLET7A-3P and inhibits migration of ovarian cancer cells via downregulating DICER1 and AOG2. (A) SKOV3 and KURAMOCHI cells were transfected with 10 nM of non-targeting (si-Cont) siRNA or siRNA targeting SQSTM1 as indicated. Cell lysates were collected and analyzed by immunoblotting using anti-SQSTM1, anti-DICER1 and anti-AGO2 antibodies. ACTB was used as loading control. (B) Cells were treated with 3 μM rVP1 or EBSS for 4 h in the presence or absence of 30 μM CQ as indicated. Cell lysates were collected and analyzed by immunoblotting. (C) Cells were transfected with 10 nM of si-Cont or siRNA targeting DICER1 and AGO2 as indicated. Transfection specificity was determined by immunoblotting using anti-DICER1 and anti-AGO2. (D) The relative expression ratios of MIRLET7A-3P were determined using qPCR analysis. Cells transfected with 10 nM of si-Cont or siRNA targeting DICER1 or AGO2 were collected and subject to (E) transwell migration assay for determining the motility of KURAMOCHI cells and (F) time-lapse microscopy for monitoring migration behavior (upper panel) and velocities (lower panel) of SKOV3 cells with or without siRNA transfection. SKOV3 and KURAMOCHI cells were transfected with AGO2 plasmid (pAGO2), DICER1 plasmid (pDICER1) or vector control (VC) as indicated. After 48 h transfection, (G) protein samples were collected and subjected to immunoblotting using anti-DICER1 and anti-AGO2 antibodies. (H) RNA samples were collected and relative expression ratios of MIRLET7A-3P miRNA were determined using qPCR analysis. (I) SKOV3 cells were serum-starved or treated with 3 μM of rVP1 and migration behavior (upper panel) and velocities (lower panel) of cells with VC, pAGO2 or pDICER1 transfection were determined by time-lapse microscopy. Data represent means ± SEM of 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001.

Overexpression of DICER1 and AGO2 (Figure 6G), on the other hand, suppressed nutrient depletion and rVP1-mediated increase of MIRLET7A-3P (Figure 6H) and reversed their inhibitory effects on the motility of ovarian cancer cells (Figure 6I). Collectively, these results demonstrated that autophagic degradation of SQSTM1 increases MIRLET7A passenger strand (MIRLET7A-3P) to suppress cancer cell motility via downregulation of DICER1 and AGO2.

High level of DICER1 and AGO2 correlates negatively with autophagy and MIRLET7A-3P miRNA in ovarian cancer

We then examined the clinical correlation between DICER1, AGO2, SQSTM1, autophagy and MIRLET7A-3P in a set of ovarian tissue microarrays including 5 normal, 5 cancer adjacent normal, 20 serous papillary adenocarcinoma and 3 serous adenocarcinoma. Patient specifications and tumor types are listed in Table 2. Our results revealed that control IgGs did not show any immunohistochemical staining of puncta (Fig. S9A), whereas the anti-LC3B, anti-WIPI2 and anti-SQSTM1 antibodies not only bound specifically to LC3B, WIPI2 and SQSTM1 respectively in immunoblotting (Fig. S9B) but also stained puncta of LC3B, WIPI2 and level of SQSTM1 immunohistochemically (Figure 7A, upper panel). Analysis of the immunohistochemistry staining results showed that the puncta of LC3B and WIPI2 were both lower whereas level of SQSTM1, an autophagy receptor known to be degraded by autophagy, was significantly higher in ovarian adenocarcinoma specimens as compared to non-tumor tissues (Figure 7A, lower panel).

Table 2.

Patient specifications of human ovarian cancer tissue array.

No.	Sex	Age	Organ	Pathology diagnosis	Type
1	F	68	Ovary	Serous papillary adenocarcinoma	Malignant
2	F	30	Ovary	Serous papillary adenocarcinoma	Malignant
3	F	40	Ovary	Serous papillary adenocarcinoma	Malignant
4	F	46	Ovary	Serous papillary adenocarcinoma	Malignant
5	F	34	Ovary	Serous papillary adenocarcinoma	Malignant
6	F	38	Ovary	Serous papillary adenocarcinoma	Malignant
7	F	34	Ovary	Serous papillary adenocarcinoma	Malignant
8	F	60	Ovary	Serous papillary adenocarcinoma	Malignant
9	F	47	Ovary	Serous papillary adenocarcinoma	Malignant
10	F	56	Ovary	Serous papillary adenocarcinoma	Malignant
11	F	57	Ovary	Serous papillary adenocarcinoma	Malignant
12	F	54	Ovary	Serous papillary adenocarcinoma	Malignant
13	F	43	Ovary	Serous papillary adenocarcinoma	Malignant
14	F	51	Ovary	Serous papillary adenocarcinoma	Malignant
15	F	40	Ovary	Serous adenocarcinoma/Serous papillary adenocarcinoma	Malignant
16	F	50	Ovary	Serous papillary adenocarcinoma	Malignant
17	F	49	Ovary	Serous papillary adenocarcinoma	Malignant
18	F	39	Ovary	Serous papillary adenocarcinoma	Malignant
19	F	60	Ovary	Serous papillary adenocarcinoma	Malignant
20	F	53	Ovary	Serous papillary adenocarcinoma	Malignant
21	F	47	Ovary	Serous adenocarcinoma	Malignant
22	F	54	Ovary	Serous adenocarcinoma	Malignant
23	F	39	Ovary	Serous adenocarcinoma	Malignant
24	F	20	Ovary	Normal ovarial tissue	Normal
25	F	15	Ovary	Normal ovarial tissue	Normal
26	F	18	Ovary	Normal ovarial tissue	Normal
27	F	41	Ovary	Normal ovarial tissue	Normal
28	F	40	Ovary	Normal ovarial tissue	Normal
29	F	30	Ovary	Cancer adjacent normal ovarial tissue	NAT
30	F	63	Ovary	Cancer adjacent normal ovarial tissue	NAT
31	F	39	Ovary	Cancer adjacent normal ovarial tissue (fibrous tissue and blood vessel)	NAT
32	F	40	Ovary	Cancer adjacent normal ovarial tissue	NAT
33	F	29	Ovary	Cancer adjacent normal ovarial tissue (fibrous tissue, blood vessel and mesothelial tissue)	NAT

NAT: normal tumor-adjacent tissue

Figure 7.

The level of autophagy correlates negatively with the levels of DICER1 and AGO2 in ovarian specimens. Representative immunohistochemistry of tumor specimens (patient 1: serous papillary adenocarcinoma; patient 2: serous adenocarcinoma) and non-tumor (Normal: normal tissue; NAT: normal tumor-adjacent tissue) were stained with (A) anti-LC3B, anti-WIPI2 and anti-SQSTM1, (B) anti-AGO2 and anti-DICER1 antibodies, respectively. Scale bar: 20 μm. The levels of LC3 and WIPI2 were determined by counting the puncta of LC3B and WIPI2, whereas the levels of SQSTM1, AGO2 and DICER1 were analyzed by counting the positive area of SQSTM1, AGO2 and DICER1 in non-tumor (10 specimens) and tumor (23 specimens) tissues using Image Scope software. (C) Correlation of AGO2 and DICER1 with LC3B, WIPI2 and SQSTM1 were analyzed by using GraphPad Prism 5 (GraphPad Software Inc.) and linear regression coefficients as well as statistical significances (P values) are indicated.

We further examined the level of DICER1 and AGO2 in ovarian cancer tissue microarrays. Our results revealed that the expression of DICER1 and AGO2 were both significantly higher in ovarian adenocarcinoma specimens as compared to non-tumor tissues (Figure 7B). We then analyzed the correlation between autophagy markers with DICER1 and AGO2 in the tissue microarray. We found that the levels of DICER1 and AGO2 correlated positively with SQSTM1 and negatively with autophagy markers such as LC3B and WIPI2 (Figure 7C).

Moreover, we used in situ hybridization to stain MIRLET7A-3P and MIRLET7A-5P within ovarian specimens and found that MIRLET7A-3P declined whereas MIRLET7A-5P increased significantly in ovarian tumor specimens as compared to non-tumorous tissues (Figure 8A). We then analyzed the correlation between MIRLET7A-3P and MIRLET7A-5P with either autophagy or DICER1 and AGO2 in the tissue microarray. Our results showed that the level of MIRLET7A-3P within ovarian specimens correlated positively with LC3B, WIPI2 and negatively with SQSTM1, AGO2 and DICER1 (Figure 8B). In contrast, there was no correlation between MIRLET7A-5P and either LC3B, WIPI2 or SQSTM1 (P = 0.92, 0.54 or 0.34) (Figure 8C).

Figure 8.

The level of MIRLET7A-3P correlates positively with autophagy and negatively with DICER1 and AGO2 in ovarian specimens. (A) Representative expression of MIRLET7A-3P and MIRLET7A-5P in tumors (23 specimens) and non-tumor tissues (10 specimens) were examined by miRNA in situ hybridization and analyzed by defining percentage of positive area using ImageScope software. Scale bar: 20 μm. Correlation of (B) MIRLET7A-3P and (C) MIRLET7A-5P with puncta numbers of LC3B and WIPI2 or positive area of SQSTM1, AGO2 and DICER1 and linear regression coefficients as well as statistical significances (P values) are analyzed by using GraphPad Prism 5 (GraphPad Software Inc.) .

Discussion

MicroRNAs have pleotropic effects on a variety of cellular processes including autophagy [34,48,49], whereas, autophagy regulates the homeostasis of some miRNAs [33,50]. By studying the effect of autophagy on the complex formed by the miRNA-processing enzyme DICER1 and the miRNA effector AGO2, Gibbings and colleagues [33] report that autophagy-associated degradation of the autophagic receptor CALCOCO2 but not SQSTM1 causes the removal of unloaded DICER1 and AGO2 leading to increase in MIRLET7A and MIR16 (microRNA 16). In our hand, however, we found that increase in autophagic degradation of SQSTM1 instead of CALCOCO2 downregulated DICER1 and AGO2 to increase MIRLET7A-3P and inhibit cell motility of ovarian cancer cells (Figures 5 and 6). Since upregulation of SQSTM1 has also been shown to increase cancer cell proliferation and tumorigenesis via stabilizing oncogenic factor TWIST1 [51], whether SQSTM1 is distinguished from CALCOCO2 in regulating DICER1 and AGO2 in different cell lines or subcellular localizations to play distinct role in mediating miRNA expression remains to be elucidated.

The interrelationship and crosstalk between autophagy and miRNAs in cancer cells is not clearly defined yet. Here, we demonstrated that BECN1-dependent and -independent autophagy (Figure 1 and S1) increased MIRLET7A-3P miRNA expression in ovarian cancer cells (Figure 2A). Since MIRLET7A-3P mimics did not increase expression of autophagy markers such as LC3B-II (Fig. S10) and downregulating MIRLET7A-3P did not significantly affect rVP1- and nutrient depletion-mediated autophagy (Fig. S11), it is likely that MIRLET7A-3P is a downstream effector of autophagy and does not exhibit any apparent feedback effect on autophagy.

Mature guide and passenger strands need DICER1 to process from PRE-MIRNA as well as AGO2 to stabilize [52,53]. Our findings that autophagy increased MIRLET7A-3P despite the loss of the enzyme required for its production and the protein required for its stabilization suggesting 2 possibilities. One possibility is that the so called ‘autophagy-induced MIRLET7A-3P’ had been mainly due to an increase in MIRLET7A-3P from PRE-MIRNA, i.e., PRE-MIRLET7A1 and PRE-MIRLET7A3. However, nutrient depletion (serum starvation) and rVP1 did not increase PRE-MIRLET7A1 and PRE-MIRLET7A3 (Fig. S4B) and knockdown of LC3B did not have any effect on the level of PRE-MIRLET7A1 or PRE-MIRLET7A3 either (Fig. S4C). In addition, northern blots for MIRLET7A-3P, MIRLET7A-5P and PRE-MIRLET7A demonstrated that a 21 nucleotide passenger strand, i.e., MIRLET7A-3P but not MIRLET7A-5P or PRE-MIRLET7A, is increased by serum starvation and rVP1 (Fig. S4D). Another possibility is that the levels of DICER1 and AGO2 in SKOV3 and KURAMOCHI ovarian cancer cells are positively dependent on SQSTM1, as siRNA knockdown or autophagy-mediated degradation of SQSTM1 results in downregulation of DICER1 and AGO2 (Figure 6A and 6B) and increase in the level of MIRLET7A-3P (Figure 5C). To elucidate whether DICER1 or AGO2, like SQSTM1, could also be degraded within autophagosomes, we examined the localization of DICER1 and AGO2 within LC3B-positive vacuoles under treatment with CQ. Immunofluorescence staining experiments showed that DICER1 and SQSTM1 but not AGO2 localizes in nutrient depletion- and rVP1-induced LC3B-positive vacuoles (Fig. S12 and S13), suggesting that AGO2 does not interact with SQSTM1 to be sequestered into autophagosome. In addition, even though DICER1 localizes in autophagosomes, its sequestering into autophagosome might not be due to SQSTM1 as downregulation of SQSTM1 did not increase DICER1 (Figure 6A). The interrelationships between autophagy, SQSTM1, DICER1, AGO2 and MIRLET7A-3P are depicted as shown in the schematic diagram (Fig. S14). More experiments are currently underway to elucidate not only how autophagy-mediated degradation of SQSTM1 could downregulate DICER1 and AGO2 but also whether downregulation of DICER1 and AGO2 decreases a factor(s) responsible for MIRLET7A-3P degradation or increases a factor(s) responsible for MIRLET7A-3P formation.

Of note, Gibbings and colleagues propose that autophagy degrades inactive DICER1-AGO2 to facilitate the loading of the miRNA-miRNA* duplex to active DICER1-AGO2 complexes and increase the yield of miRNA [33]. Other investigators, however, found that autophagy decreases rather than increases miRNAs such as MIR224 in hepatitis B virus-associated cell responses [50] and the effects of autophagy may thus be cell type-dependent. Here, we reported that the levels of LC3B, WIPI2 and MIRLET7A-3P (but not MIRLET7A-5P) were all lower and exhibited a positive correlation in human ovarian cancer specimens (Figs. 7 and 8). Our results are consistent with those of Gibbings and colleagues who find that autophagy is positively correlated with MIRLET7A in HeLa cells [33].

Besides MIRLET7A-3P, we also examined whether the levels of other passenger and PRE-MIRNA pairs of miRNAs, notably, MIR16-1, MIR16-2 and microRNA 29C (MIR29C) were increased by autophagy activation. Our results showed that even though the levels of passenger strand of MIR16-2 and MIR29C (Fig. S15) as well as PRE-MIR16-1 and MIR16-5P (Fig. S16) were not altered, MIR16-1-3P was increased by nutrient depletion and rVP1 in an autophagy-dependent manner (Fig. S15). It is thus likely that activation of autophagy may selectively increase not only MIRLET7A-3P but also passenger strands of multiple miRNAs. Whether other passenger strands of miRNA also contribute to the autophagy-mediated inhibition of cancer cell motility remains to be determined.

In summary, our results demonstrated that induction of BECN1-dependent or -independent autophagy increases MIRLET7A-3P to inhibit ovarian cancer motility via downregulating SQSTM1 to decrease DICER1 and AGO2. This identification of relationship among autophagy, SQSTM1, DICER1, AGO2 and MIRLET7A-3P in ovarian cancer migration may facilitate the potential use of these mediators and effectors not only as biomarkers for cancer motility but also as targets for modulating ovarian cancer migration, invasion and metastasis.

Materials and methods

Materials

rVP1 was produced as described previously [54]. Human ovarian cancer tissue microarray was purchased from US Biomax (OV807). The human tissues were collected from certified hospitals under the highest ethical standards, with the donor and their relatives being fully informed. All human tissues were collected in accordance with the Anatomical Gift Act. The collection protocol was complied with all federal, state and local laws. The miRNA primer sets of PRE-MIRLET7A1 (design ID: 361,805), PRE-MIRLET7A3 (design ID:361,809), MIRLET7A-3P (206,084), MIRLET7A-5P (205,727) and U6 (203,907); in situ hybridization detection probe of non-targeting-MIR (99,004–15), MIRLET7A-3P (38,771–15) and MIRLET7A-5P (18,000–15) were all purchased from EXIQON. The miRNA primer sets of PRE-MIR16-1 (YCP0032270), MIR16-1-3P (YP00206012), MIR16-2-3P (YP00204309), MIR16-5P (YP00205702) and MIR29C-3P (YP00204729) were purchased from QIAGEN. IsHyb in situ hybridization kit (K2191020) was obtained from BioChain. Anti-LC3B antibodies (NB600-1384 for immunoblot; NBP2-46,892 for immunohistochemistry) were obtained from Novus. Anti-BECN1 antibody (3738) was obtained from Cell Signaling Technology. Anti-WIPI1 antibody (ALX-210–955) was obtained from Enzo Life Science. Anti-WIPI2 (ab101985), anti-AGO2 (ab186733), mouse IgG2β; isotype control (ab18415) and rabbit IgG isotype control (ab27478) antibodies were obtained from Abcam. Anti-CALCOCO2 (GTX115378) and anti-SQSTM1 (GTX100685 for colocalization immunofluorescence staining) antibodies were obtained from GeneTex. Anti-SQSTM1 antibody (H00008878-M01) was obtained from Abnova. Anti-ATG5 antibody (3167-S) was obtained from EPITOMICS. Anti-ACTB (sc-47,778), anti-ATG7 (sc-8668) and anti-LC3B (sc-271,625 for colocalization immunofluorescence staining) antibodies; non-targeting siRNA (sc-37,007), BECN1 siRNA (sc-29,797), LC3B siRNA (sc-43,390), ATG5 siRNA (sc-41,445), ATG7 siRNA (sc-41,447), SQSTM1 siRNA (sc-29,679), WIPI1 siRNA (sc-72,210) and WIPI2 siRNA (sc-72,212) were obtained from Santa Cruz Biotechnology. Anti-DICER1 (AMAb90737) antibody and chloroquine (C6628) were obtained from Sigma-Aldrich. EBSS (14,155,063), non-targeting (4,464,058) miRNA mimic, MIRLET7A-3P (MC12169) miRNA mimic, MIRLET7A-5P (MC11050) miRNA mimic, non-targeting (4,464,076) miRNA inhibitor (negative control), MIRLET7A-3P (MH12169) miRNA inhibitor and Lipo3000 transfection reagent (L3000015) were all obtained from ThermoFisher Scientific. Water-soluble tetrazolium (WST) cell proliferation kit (K301-2500) was obtained from BioVision. Plasmids of pENTER-AGO2 (CH859024) and pENTER-DICER1 (CH807331) were constructed and purchased from ViGene Biosciences. Plasmid of pHA-SQSTM1 (28,027; deposited by Qing Zhong) was obtained from Addgene.

Cell culture

Human ovarian cancer cell line SKOV3 (HTB-77) and KURAMOCHI (JCRB0098) were purchased from American Type Culture Collection and the Japanese Collection of Research Bioresources Cell Bank, respectively. SKOV3 was maintained in McCoy 5A medium (GIBCO, 16,600–082); KURAMOCHI was maintained in RPMI1640 medium (GIBCO, 22,400–089) supplemented with 10% fetal bovine serum (FBS), 2 mM L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin in a 37°C incubator with a humidified atmosphere containing 5% CO₂.

Immunohistochemistry

Immunohistochemistry staining of LC3B, WIPI2, SQSTM1, AGO2 and DICER1 were conducted as described previously [20]. Briefly, the slides of human ovarian cancer tissue microarray were deparaffinized in xylene and rehydrated through graded ethanol. The slides were incubated with antigen retrieval buffer in a heated oven for 20 min and then hybridized with specific primary antibodies overnight at 4°C. After blocking with normal goat serum, slides were stained with biotinylated secondary antibodies. The signals were amplified by treating with avidin and biotinylated enzyme complex (VECTOR, PK-7100) and developed with 3,3′-diaminobenzidine substrate solutions (VECTOR, SK-4100). All slides were counterstained with haematoxylin. The puncta and percentage of positive area from 5 fields of each tissue sample were quantified using positive pixel count and color deconvolution algorithm of Image-Scope software (Aperio Technologies) respectively. The specificity of each primary antibody was verified by immunoblottings and gene knockdown to show that the antibody primarily binds to its target protein and does not cross-react with other proteins. To make sure that the IHC staining is due to the specificity of the primary antibodies, we also used control IgGs as primary antibodies to demonstrate that the appearance of signals in IHC is due to specific binding of primary antibodies rather than nonspecific binding of secondary antibodies.

MiRNA in situ hybridization

The miRNA in situ hybridization was performed using the IsHyb In Situ Hybridization Kit (BioChain, K2191020) according to the manufacturer’s instructions. In brief, the slides of human ovarian cancer tissue microarray were deparaffinized in xylene, rehydrated through graded ethanol and then fixed with 4% paraformaldehyde in DEPC-PBS. The fixed slides were incubated with pre-hybridization solution for 4 h at 50°C and then hybridized with 80 nM 5′-digoxigenin (DIG)-LAN-detection probes of non-targeting-MIR, MIRLET7A-3P and MIRLET7A-5P for 16 h at 45°C. After washing and blocking, slides were incubated with alkaline phosphatase-conjugated anti-DIG antibody overnight at 4°C. The miRNA signal was amplified by incubating with BCIP/NBT (5-bromo-4-chloro-3-indolyl-phosphate/nitro blue tetrazolium) color development substrate solution for an appropriate time. All slides were counterstained with methyl green. The percentage of positive area was analyzed using the color deconvolution algorithm of Image-Scope software from 5 fields of each tissue sample.

Immunofluorescence staining

The puncta formation or localization of DICER1 and AGO2 with SQSTM1 in LC3B-positive vacuoles were observed by immunofluorescence staining. Briefly, cells were seeded onto the slide, fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 (Sigma-Aldrich, T8787) and blocked with 1% BSA (Sigma-Aldrich, A2153) at room temperature. Slides were further incubated with primary antibodies at 4°C overnight and treated with fluorescent-labeled-secondary antibodies (Invitrogen, A11008 and R6393) for 30 min at room temperature. Slides were finally mounted by adding DAPI (4′,6-diamidino-2-phenylindole) Fluoromount G mounting buffer (SouthernBiotech, 0100–20). Images were obtained using LSM 780 confocal microscopy (Carl Zeiss AG, Carl-Zeiss-Promenade 10, 07745 Jena, Germany). The puncta per cell were quantified by examining 50 to 100 cells in each experiment.

Immunoblot analysis

Proteins were extracted from cell lysates of SKOV3 and KURAMOCHI by using protein extraction buffer (Santa Cruz Biotechnology, sc-24,948) containing protease inhibitors (Roche, 04906845001). They were then resolved in 4 × SDS sample buffer containing 10% SDS, 0.5 μM Tris-HCl, 4% glycerol, 1 M β-mercaptoethanol and 0.1% bromophenol blue. After denaturing at 95°C for 5 min, protein samples were subjected to SDS-PAGE and subsequently transferred to a polyvinylidene difluoride membrane, which was then incubated in freshly prepared 150 mM NaCl, 20 mM Tris-HCl, pH 7.5, 0.1% Tween 20 buffer containing 5% skim milk. Protein expression levels were detected with specific primary antibodies, followed by horseradish peroxidase-conjugated secondary antibodies (Santa Cruz Biotechnology, sc-2004 and sc-2005). Immunoreactivity was detected on Biomax ML films using Pico Chemiluminescent Substrate (Pierce, 34,079) according to the manufacturer’s instructions. Quantitative data of immunoreactivity was obtained using densitometric analysis with ImageJ software.

Transmission electron microscopy

Autophagic structures were observed by transmission electron microscopy as described previously [7]. Briefly, cells were transfected with non-targeting siRNA or siRNA targeting BECN1 for 48 h and then treated with EBSS or 3 μM rVP1 for another 4 h. After treatment, cells were fixed and embedded in 2% agarose and dissected into 1 mm cubes and further fixed overnight. Cells were then fixed in 0.1 M cacodylate buffer containing 1% OsO₄ and subsequently dehydrated and embedded in Spurr’s resin (Electron Microscopy Sciences, 14,300), that was then sliced into 60-nm ultrathin sections and examined with a Hitachi-H7000 transmission electron microscope (S/N 747–32-03; Rapid City, SD, USA). Random sampling was applied to quantification of autophagic compartment volume fraction in each section [55]. At least 3 squares were selected from grid squares. Multilayered autophagosomal structures with clear cytosolic components inside were point counted at 40,000 × magnification, whereas, cell areas were obtained at 2,000 × magnification.

Quantitative real-time PCR analysis

The miRNA level was determined using quantitative real-time PCR analysis. In brief, RNA samples were isolated using miRCURY RNA Isolation Kit (EXIQON, 300,110) according to the instructions of the manufacturer. The isolated RNA samples were then subjected to cDNA synthesis by miRCURY LNA Universal RT kit (EXIQON, 203,301). The real-time PCR reactions were subsequently performed by using miRNA LNA PCR primer sets on a 7500 system (LS4351105, Applied Biosystems, Foster City, CA, USA) with a SYBR green qPCR kit (EXIQON, 203,402).

Cell viability assay

Cell viability was measured by WST-1 assay as described previously [20]. Briefly, SKOV3 and KURAMOCHI cells were seeded onto 96-well plates and transfected with specific siRNAs for 72 h in a 37°C incubator with a humidified atmosphere containing 5% CO₂. After transfection, medium were removed from the well and 10 μL WST-1 reagent in 100 μL serum free medium was added to each well for 2 h. The ratios of surviving cells were measured with an ELISA reader (ELx800^TM, BioTek, Winooski, VT, USA) at a wavelength of 450 nm.

Transient transfection

The siRNAs of BECN1, LC3B, WIPI1, WIPI2, SQSTM1, CALCOCO2, AGO2 and DICER1; plasmid of pENTER-AGO2, pENTER-DICER1 and pHA-SQSTM1; miRNA mimic of MIRLET7A-3P and MIRLET7A-5P and synthetic antisense of MIRLET7A-3P (MIRLET7A-3P inhibitor) were transfected into SKOV3 or KURAMOCHI cells with Lipo3000 transfection reagent according to the manufacturer’s instructions. Briefly, suitable concentration of siRNAs or 1 μg of plasmids were transfected into cells. Suitable concentrations of rVP1, serum starvation or EBSS medium were added after transfection if required. For miRNA mimic or inhibitor transfection, 3 to 100 nM of MIRLET7A-3P and MIRLET7A-5P miRNA mimics or 3 to 30 nM of MIRLET7A-3P miRNA inhibitors were transfected into cells. Suitable concentrations of rVP1, serum starvation or EBSS medium were then added and the migration ability was determined.

Northern blot analysis

MiRNA expression was determined using highly sensitive miRNA northern blot assay kit (Signosis, NB-1001) according to the instructions of the manufacturer. In brief, RNA samples were isolated, heated at 70°C for 5 min, loaded into the 15% pre-run urea-polyacrylamide gel and then transferred to the NB membrane. The membrane was subsequently hybridized with MIRLET7A-3P (Signosis, Cus-HP-1), MIRLET7A-5P (Signosis, HP-0001) and RNU6-1 (Signosis, HP-1001) probes respectively at 42°C overnight and incubated with Amplifier at 42°C for 2 h. The membrane was then incubated with blocking buffer for 30 min and treated with streptavidin-HRP conjugate for another 45 min. The membrane was finally incubated with biotin substrate and the chemiluminescent signal was detected on Biomax ML films.

Cell migration assay

Cell migration assay was determined by time-lapse microscopy as described previously [7]. In brief, SKOV3 cells transfected with specific miRNA mimics, miRNA inhibitors, siRNAs or plasmids were seeded into a 12-well plate at a density of 2 × 10⁴ cells/cm². Forty-eight h after transfection, cells were treated with 3 μM rVP1 or incubated with serum starvation medium if needed. The cells were then subjected to time-lapse microscopy experiments to obtain serial phase-contrast images every 15 min for 24 h using an inverted LSM 510 META confocal microscope (Carl Zeiss AG, Jena, Germany) equipped with a humidified 37°C chamber in 5% CO₂. The cell trajectories and migration velocities were recorded for 24 h using the tracking point tool of Metamorph software (Molecular Devices). The average of cell migration velocities was defined as the average of 20 subsequent cell centroid displacements/one time interval between 2 images.

Transwell migration assay

Cell migration ability was determined by transwell migration assay as described previously [20]. Briefly, transwell inserts (8-µm pore; Costar, 3422) were placed into the wells before seeding cells into the upper well. After transfections or treatments, cells were trypsinized and a total of 5 × 10⁴ to 2 × 10⁵ KURAMOCHI or 3 × 10⁴ SKOV3 cells in serum-free medium were placed in the upper chamber, whereas medium with 10% FBS was added in the lower chamber. To facilitate migration of KURAMOCHI cells, 10% FBS-RPMI1640 medium containing 10 ng/ml HGF (hepatocyte growth factor; Gibco, PHG0324) was added in the lower chamber. After incubation at 37°C in a humidified 5% CO₂ atmosphere for 24 h, the cells that had migrated to the other side of the membrane were fixed with methanol and the non-migrated cells were mechanically removed with a cotton swab. Cells adherent on the membrane were stained with Liu stain (ASK, 03R011/03R021). Cell numbers were examined under light microscopy at 200 × magnification (Axiovert 200, Carl Zeiss AG, Jena, Germany).

Gelatin zymographic analysis

Gelatin zymographic analysis was conducted as described previously [7]. Supernatants were collected from cells after miRNA mimic transfection for 3 days and then concentrated with Vivaspin 6 centrifugal concentrators (Vivascience, VS2002). Five micrograms of concentrated supernatants were loaded to 10% SDS-polyacrylamide gels containing 1 mg/ml gelatin under nonreducing conditions. After electrophoresis, the gels were washed by 2.5% Triton X-100 and developed for 20 h at 37°C with Tris-HCl (pH 8.0) buffer containing 0.01 M CaCl₂. Finally, gels were fixed in 50% methanol containing 10% acetic acid and stained with 0.5% Coomassie brilliant blue R-250, then destained in 10% acetic acid until the bands turned clear (zones of gelatin degradation) against the blue background of stained gelatin. The bands were quantified with ImageJ software.

Statistical analysis

Statistical significance and P values were determined by two-tailed Student’s t test and related values are shown as means ± SEM.

Acknowledgments

We thank Shu-Chen Shen (Scientific Instrument Center, Academia Sinica, Taiwan) for instruction and assistance in confocal microscopy analysis; Yao-Kuan Huang of the electron microscope core facility, Academia Sinica, Taiwan for sample preparation and images acquiring; Miranda Loney (Agricultural Biotechnology Research Center Editors’ Office, Academia Sinica, Taiwan) for English editorial assistance; and Academia Sinica for grant support (to S.-M. Liang and C.-M. Liang).

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.