Autophagy-Mediated Clearance of Free Genomic DNA in the Cytoplasm Promoted the Growth and Survival of Breast Cancer Cells

Background: The cGAS (GMP-AMP synthase)-triggered senescence-associated secretory phenotype (SASP) in promotion of cancer progression has been extensively documented. However, the role of cGAS-mediated DNA autophagy is little evaluated in breast cancer cells. Methods: Immunouorescence, senescence associated-β-galactosidase staining (SA-β-gal) and Western blot were performed to detect gene expression, distribution and phenotypes. PCR, IP-PCR, FISH, BrdU, Comet assay, coimmunoprecipitation, sucrose density gradient centrifugation were carried out to detect possible mechanisms. Trypan blue exclusion, Live/dead staining and MTS assay were to measure the cell viability. All analyses were performed using GraphPad Prism 8. Relationships were analyzed using t-tests. A P-value of less than 0.05 was considered signicant. All statistical tests and P values were 2-sided, and the level of signicance was set at <0.05 (*), <0.01 (**), <0.001 (***), or<0.0001 (****); ns indicates no signicance. Results: Active DNA autophagy but not SASP activity could be detected in breast cancer cells with high micronucleus (MN). The selective autophagy of free genomic DNA in the cytoplasm is mediated by cGAS and usually coordinated with SQSTM1-mediated autophagy of ubiquitinated histones in breast cancer cell lines with high frequency of MN formation. Cytoplasmic DNA together with nuclear proteins derive from DNA replication-induced nuclear damage and MN collapse in breast cell lines which with severe DNA damage. The inhibition of DNA autophagy through either chemical inhibitors or genomic silencing of cGAS or SQSTM1 suppresses the growth and survival of breast cancer cells, while enhanced DNA damage increases the sensitivity to these inhibitors for more cancer cells. Human cancer cells with either high DNA autophagy or enhancement of DNA damage are sensitive to inhibition of DNA autophagy. Conclusions: Our investigation revealed DNA autophagy in breast cancer cells with high MN formation.

A total of 1x10 7 transfected cells were washed twice with PBS, 600 μl of precooled hypo-osmic buffer (25 mM Tris, pH 7.4, 85 mM KCl) was added, and the samples were incubated on ice for 30 minutes and centrifuged at 4°C at 12000 rpm for 10 minutes. The supernatant was saved and incubated with Flag antibody-conjugated agarose beads (MBL, Chiba, Japan) and gently shaken on a turntable overnight at 4°C. The beads were washed with hypo-osmic buffer containing protease inhibitor cocktail for 10 minutes; this process was repeated 4 times. Finally, the beads were dissolved in 1.5×SDS loading buffer, and 30 μl of supernatant was analyzed by Western blotting. The primary whole cytoplasmic supernatant was used as input.

IP-PCR
A total of 2×10 7 transfected cells were washed with preheated PBS at 37°C 3 times, xed with 1% formaldehyde in PBS in a 37°C incubator for 15 minutes, quickly washed with ice-cold PBS 5 times, scraped into an Ep tube, and centrifuged at 800 g at 4°C for 3 minutes. Then, the supernatant was discarded, and 500 μl of hypo-osmic buffer (25 mM Tris, pH 7.4, 85 mM KCl) was added to isolate the cytoplasmic protein. Part of the supernatant was saved as the input. The remaining part was used for IP experiments. An appropriate amount of 500 μl elution buffer (1% SDS, 0.1 M sodium bicarbonate was used to elute the protein-DNA complex on the beads for 10 minutes at room temperature. Then, RNase A was added, the samples were heated at 37°C and shaken for 2 h, Proteinase K was added, and the samples were heated at 55°C and shaken for 2~3 h. Then, the samples were heated at 65°C and shaken overnight to isolate the protein-DNA complexes. Finally, the IP DNA was extracted by the phenol and chloroform method, and the results were analyzed by PCR. Sucrose density gradient centrifugation Gradient concentrations of sucrose solution (5%-40%, the concentration interval was 5%) with protease inhibitor cocktail were established as described [19]. The cytoplasmic proteins, extracted by hypo-osmic buffer as described above, were carefully dropped on the top layer and centrifuged at 40,000 rpm (Beckman, Brea, CA, USA) for 4 hours at 4°C. After centrifugation, the samples were carefully collected from a 500 μl aliquot of each fraction, and the aliquot of each fraction was analyzed by Western blot. Cytoplasmic DNA was extracted from 100 μl of each fraction and analyzed by PCR.

Acid extraction of cytoplasmic histones
Cytoplasmic histones were isolated by acid extraction methods with some modi cations [20]. Brie y, cytoplasmic proteins from 1×10 7 cells of each kind, extracted by hypo-osmic buffer as described above, were slowly added to 0.4 N H 2 SO 4 (500 μl of H2SO4 to a 100 μl cytoplasmic solution) and incubated at 4°C with intermittent rotation for 2 hours. After centrifugation at 5500 rpm for 5 minutes, the supernatants were gently added to 150 μl of 100% TCA ( nal concentration of 20%) and kept on ice for at least 5 hours without agitation. After centrifugation, the pellets were washed with 500 μl of acetone+0.1% HCl, and the Quantitative real-time RT-PCR was performed on a PCR system (Applied Biosystems, Inc., Carlsbad, CA, USA) using SYBR Premix. The results were evaluated as the ratio of cytoplasmic: nuclear DNA; cytoplasmic DNA was normalized to nuclear DNA. The results were analyzed by GraphPad Prism 8.0.

FISH
The Alu probes were as follows: Alu-1 :5'-CCTGAGGTCAGGAGTTCGAGACCAGCCT-3'; Alu-2: 5'-ACGCCTGTAATCCCAGCACTTTGGGAGG-3'; Alu-3:5' -TCGCGCCACTGCACTC CAGCCTGGGCGA-3'. They were synthesized and conjugated with a single Quasar 570 molecule at the 5' end (Sango Biotech, Shanghai, China). Cells grown on cover slides were xed in 3% paraformaldehyde (pH 7.4) containing 0.1% Triton-X 100 for 30 minutes and permeabilized in 0.1% Triton X-100 for 20 minutes. After denaturation at 95°C for 5 minutes, hybridization was performed in a mixture of probes (20 ng/per slide) and 35% deionized formamide, 10% dextran sulfate, 1× Dehart's, and 2× SSC for 20 hours at 42°C. The slides were washed for 40 minutes in 2×SSC with several changes. Nuclei were stained with Hoechst 34580 (Sigma Aldrich) for 10 minutes, and the results were observed and recorded using a uorescence microscope (Model CX51, Olympus, Tokyo, Japan), and Photoshop version 7.0 (Adobe Systems, Inc. San Jose, CA, USA) was used to observe and analyze the results. At least 500 cells were evaluated, and the results were evaluated as the ratio of the intensity in cytoplasmic and nuclear DNA; cytoplasmic signals were normalized to those of the nuclei.

Comet assay
The comet assay was performed using a comet assay kit according to the manufacturer's instructions.
First, 1×10 5 cells were prepared. Neutral or alkaline electrophoresis was performed. The slides were viewed by epi uorescence microscopy using a FITC lter. The results were analyzed by Comet Score software (Version 2.0038).
The comet assay was performed using a comet assay kit (Abcam, Cambridge, UK) according to the manufacturer's instructions. First, 1×10 5 cells were prepared. Comet Agarose was pipetted onto the Comet Slide to form base layer cells, which were combined with Comet Agarose at 37°C. Cell samples were combined with Comet Agarose at a 1/10 ratio (v/v). We pipetted the agarose/cell mixture on top of the base layer. The cells were treated with lysis buffer and alkaline solution. Electrophoresis was performed under alkaline or neutral conditions. Voltage was applied to the chamber for 10-15 minutes at 1 volt/cm.
The cells were stained with DNA dye. The slides were viewed by epi uorescence microscopy using a FITC lter. The results were analyzed by Comet Score software (Version 2.0038).

BrdU incorporation assay
BrdU incorporation assays were performed as described previously (18). Brie y, BrdU (10 µM) was added to the culture medium for 30 minutes before analysis, and then, the cells were xed with 4% formaldehyde, permeabilized with 0.1% Triton X-100, and denatured with 20 mM HCl in 150 mM NaCl and 3 mM KCl for 20 minutes at 25°C. The cells were then incubated with a primary antibody mixture composed of primary antibodies (BrdU and MCM7 or Lamin B1).

Trypan blue exclusion assay
Cell growth was determined by Trypan blue exclusion assays with a Trypan Blue Staining Cell Viability Assay Kit (Beyotime, Shanghai, China). Cells (1 ×10 4 cells/per well) grown with or without treatment with CQ, ba lomycin A1, or H-151 in 96-well plates were harvested, and 50 μl of trypan blue was added to a 50 μl cell suspension according to the manufacturer's protocol. Viable cells were counted under a microscope with a hemocytometer. The assays were performed in triplicate and repeated at least three times.
Live/dead viability assay A live/dead assay was performed using a live/dead cell viability assay kit (Abcam, Cambridge, UK). A total of 1×10 5 cells were seeded in 12-well or 96-well plates and incubated for 24 hours. The cells were treated with CQ or ba lomycin A1 and incubated for the indicated times. Subsequently, the cells were rinsed twice with PBS before the uorochromes were added and incubated for 45 minutes. Fluorescence images were then taken (Model CX51, Olympus, Tokyo, Japan), and live or dead cells were counted and calculated.
MTS assay MTS assays were performed using an MTS assay kit (Bestbio, Shanghai, China). A total of 1×104 cells were seeded in 96-well plates and incubated for 24 hours. The cells were treated with CQ or ba lomycin A1 and incubated for the indicated times. Ten microliters of MTS solution were added to each well and incubated for 3 hours at 37°C. The absorbance was measured at 490 nm, and cell viability was analyzed.

Results
A high frequency of MNs in breast cancer cells resulted in an autophagic phenotype MCF-7, MDA-231, and BT-549 breast cancer cells were subjected to immuno uorescence staining for Lamin B1 and NAT10 as described previously [18], and spontaneous MN formation in each kind of cell was analyzed. The rate of spontaneous MNs in the MCF-7 cells was approximately 5%, that in the MDA-231 cells was approximately 15%, and that in the BT-549 cells was approximately 35% (Fig. 1a). However, the MNs formed in three cell lines generally contained DNA and Lamin B1 or A/C, nuclear lamina proteins, and other nuclear envelope components, including LBR (Lamin B receptor), nuclear pore complex components (MAB414, nucleoporin 153), the nuclear basket protein TPR (translocated promoter region), and integral membrane components (Sun2 and nesprin2), suggesting that the nuclear membrane of MNs generally maintained structural components similar to the membrane of the main nuclei ( Supplementary Fig. S1a).
MNs are considered a major source of free DNA, which triggers the activation of the SASP through DNA binding to cGAS-STING pathway components [6]. Therefore, the MDA-231, BT-549, and MCF-7 cells were analyzed by SA-β-gal staining, and the results indicated that the MCF-7, MDA-231 cells showed 4% and 8% SA-β-gal positivity, while the BT-549 cells unexpectedly showed no SA-β-gal-positive cells even with repeated staining (Fig. 1b).
Subsequent Western blotting analysis showed the expression of the STING pSTING, IRF3, STAT6 in BT-549 cells is lower than that of MDA-231 or MCF-7 cells, indicating that the activity of SASP in BT-549 cell line was signi cantly lower than that in MDA-231 and MCF-7 cells (Fig. 1c). Staining also showed that after treatment with the STING antagonist H-151 (2 µM) for 24 hours, the positive ratio strongly decreased, con rming that the SASP phenotype is mediated via cGAS-STING in the MDA-231 cells ( Supplementary   Fig. S1b). However, treatment with the STING agonist cGAMP [1] and transfection with Flag-cGAS or Flag-STING did not induce SA-β-gal staining, indicating blockade of cGAS-STING signaling in the BT-549 cells (data not shown).
Recently, the cGAS-STING pathway was shown to mediate DNA autophagy. Therefore, autophagic activity was rst compared among the MCF-7, MDA-231 and BT-549 cells. Western blotting showed that the BT-549 cells presented high expression of LC3, SQSTM1, and LAMP2, which was not obvious in the MDA-MB-231 or MCF-7 cells, and notably, a high level of DNase II was detected in the BT-549 cells (Fig. 1d). The BT-549 cells were treated with the autophagic inhibitors CQ (10 µM) for 4 hours or ba lomycin A1 (10 nM) for 24 hours, and the levels of LC3, SQSTM1 and LAMP2 were strongly increased, con rming the autophagic activity in the BT-549 cells (Fig. 1e). Moreover, the expression of DNase II was dose-dependently enhanced by CQ treatment (10 µM) in the BT-549 cells (Fig. 1e). Furthermore, LysoTracker staining showed that lysosomes were more abundant in the BT-549 cells than in the MDA-231 cells (Fig. 1e). The results suggested the involvement of DNA autophagy in the BT-549 cells.
To further con rm the autophagic activity, we also performed electron microscopy of the MDA-MB-231 and BT-549 cells after treatment with ba lomycin A1 (10 nM) for 72 hours, and the results showed that there were many autophagic vesicles in the BT-549 cells but not the MDA-MB-231 cells (Fig. 1g). Interestingly, the distribution of cGAS was shown to be similar to that of SQSTM1 or LC3. The BT-549 cells presented accumulation of cytoplasmic cGAS granules that colocalized with SQSTM1 (~ 50%) and cytoplasmic LC3 (~ 40% cells), while the MCF-7 and MDA-231 cells were faintly stained for cGAS ( Fig. 2a, b). All three kinds of cells showed staining for cGAS in occasional MNs (3%-5%) ( Supplementary Fig. S2c).
In addition, only a few cGAS-positive MNs (~ 3%) were positive for SQSTM1 but not LC3 (Fig. 2b, 2c). To further con rm the autophagic activity, we treated the MDA-MB-231 and BT-549 cells with CQ (10 µM) or ba lomycin A1 (10 nM) for 24 hours, and the results showed that the BT-549 cells presented marked increases in cytoplasmic cGAS, SQSTM1 and LC3, while the MDA-231 and MCF-7 cells only showed slight increases in a few cells (Fig. 2a, b). However, the MNs in the three kinds of cells presented no signi cant increase in positive staining of cGAS, SQSTM1, LC3 or Beclin-1 (data not shown). As expected, treatment with ba lomycin A1 markedly increased not only the levels of SQSTM1, LC3, and LAMP2 but also cGAS and STING in the BT-549 cells (Fig. 2d). Given that free cytosolic DNA was not membrane enclosed, coimmunoprecipitation was carried out to explore the potential interaction between cGAS or SQSTM1 and genomic DNA in the cytoplasm. Flag- Taken together, these data indicated that DNA autophagy in breast cancer cells could be selective autophagy of cytoplasmic free DNA but not nucleophagy and possibly involved cGAS, SQSTM1 and LC3.

DNA autophagy in the cytoplasm was involved in the coordination of cGAS and SQSTM1
To explore the autophagic ow in DNA autophagy, we investigated the relationship between cGAS and SQSTM1 or LC3. After knockdown of either LC3 or SQSTM1 by siRNA, the level of cGAS was obviously increased in the BT-549 cells, as shown by Western blotting (Fig. 3a, b), while the granular form of cGAS was reduced, as shown by immuno uorescence staining (Fig. 3a, b). Moreover, depletion of LC3 increased the SQSTM1 levels ( Fig. 3a). In contrast, after knockdown of cGAS by siRNA, the levels of either LC3 or SQSTM1 were decreased in the BT-549 cells, as shown by immuno uorescence and Western blotting (Fig. 3c).
After knockdown of cGAS by interfering RNAs, gel electrophoresis showed that the levels of cytoplasmic Alu-and rDNA sequences increased obviously in the BT-549 cell lines but not in the MCF-7 or MDA-231 cell lines (Fig. 3d). Similarly, knockdown of LC3 also resulted in the same ndings (Fig. 3d).
To explore the potential autophagic complex of SQSTM1, we further analyzed cGAS, LC3 and free DNA through cytoplasmic fractioning in a density gradient fraction assay, in which the detection of SQSTM1 overlapped in fractions containing cGAS, LC3 and free DNA (Fig. 3e). Moreover, the potential complex of cGAS, SQSTM1 and genomic DNA was further analyzed by a coprecipitation strategy, which showed that genomic DNA could be detected in the MDA-231 and BT-549 cells transfected with either Flag-SQSTM1 or Flag-cGAS (Fig. 2f). In addition, endogenous cGAS in the BT-549 cells coprecipitated with transfected Flag-SQSTM1 ( Supplementary Fig. S3a).
However, there is no direct interaction between SQSTM1 and cGAS or DNA binding activity, and how SQSTM1 participates in DNA autophagy needs to be addressed. Generally, SQSTM1 recognizes ubiquitinated substances during autophagy, and it has been reported that cGAS undergoes K48-linked ubiquitination at K414, leading to SQSTM1-dependent selective autophagic degradation [21]. Thus, Flag-SQSTM1 was transfected into the BT-549 cells, but K48-ubiquitinated cGAS could not be detected by either Western blotting or immuno uorescence staining (data not shown).
Since DNA is usually coated with histones or other chromatin-binding proteins, cytoplasmic free DNA was also assumed to be bound to histones. To clarify this, we isolated cytoplasmic histones from the BT-549 or MDA-231 cells by acid-based extraction [20]. Western blotting showed that cytoplasmic histones could be detected in both the BT-549 and MDA-231 cells, but the BT-549 cells presented more cytoplasmic histones than the MDA-231 cells (Fig. 3f); these structures could be detected by anti-K48 ubiquitin and FK2 antibodies (against poly-or monoubiquitinated proteins) (Fig. 3f). Moreover, immuno uorescence staining showed that more intensive staining of FK2 could be detected in the cytoplasm of the BT-549 cells than in that of the MDA-231 cells ( Supplementary Fig. S3b).
Recent reports suggest that activation of cGAS upon binding to DNA could trigger activation of STING, leading to either SASP or autophagy, and MCF-7, MDA-231 and BT-549 cells presented different endogenous levels of STING or phosphorylated STING, which was consistent with their SASP and autophagic phenotypes (Fig. 1b, c). Moreover, endogenous or exogenous STING was present in a few cytoplasmic vesicles in the MCF-7 and MDA-231 cells but was distributed in the Golgi apparatus in the BT-549 cells and could be disrupted by brefeldin A (BFA) (Supplementary Fig. S3c). In the BT-549 cells, the level of STING markedly increased in the presence of ba lomycin A1 (10 nM) (Fig. 2d) or with downregulation of DNase II (Supplementary Fig. S3d). However, after treatment of the BT-549 cells with either the STING antagonist H-151 (2 µM) or agonist cGAMP (300 nM) for 24 hours, SQSTM1 or LC3 presented no change in immuno uorescence staining (data not shown). Thus, degradation of STING may be involved in the autophagic process, while its activity in the regulation of autophagy in BT-549 cells requires further exploration (see Discussion).
Taken together, the results suggested that free cytoplasmic DNA autophagy could be mediated in a complicated process involving cGAS binding to DNA and recognition of ubiquitinated histones by SQSTM1, consequently resulting in activation of LC3.
Genomic DNA in the cytoplasm could be derived from either damaged nuclei or MNs Since the cytoplasmic DNA undergoing autophagy was genomic DNA from the nuclei and nuclear membrane in the breast cancer cells were not generally broken ( Supplementary Fig. S1a), DNA damage was assumed to be involved. The level of DNA damage in the BT-549, MCF-7 and MDA-231 cells was analyzed by comet assays. Table 1 shows that the tail length of the BT-549 cells was substantially longer than that of the MDA-231 (76.83, P < 0.0001) and MCF-7 cells (70.59, P < 0.0001), the tail intensity of the BT-549 cells was greater than that of the MDA-231 (70319, P < 0.01), and MCF-7 cells (75513, P < 0.0001), and the tail movement of the BT-549 cells was also obviously higher than that of the MDA-231 (13.09) and MCF-7 cells (15.66, P < 0.0001). These results indicated that the DNA damage was the most severe in the BT-549 cells (Table 1, Fig. 4a).  Fig. 4a). Similarly, knocking down cGAS and LC3 in BT-549 cells by siRNA also increased the tail values in the comet assay (  Fig. 4a).
These results indicated that inhibition of autophagy could in uence the status of DNA damage.
Moreover, TUNEL assays were performed to directly measure DNA breaks in breast cancer cells. BT-549 cells presented more TUNEL-positive cells than those of MCF-7, MDA-231, and the percentage of TUNELpositive BT-549 cells was approximately 35%, and the percentage of MDA-231 cells was 5% (P < 0.05) (Fig. 4b). Furthermore, the expression of ATM, ATR, and γ-H2AX in the MDA-231 and BT-549 cells further con rmed that the BT-549 cells generally had lower levels of ATM, ATR, and γ-H2AX, indicating a failure in the DNA damage response (Fig. 4b).
Interestingly, the TUNEL assay also showed that the BT-549 cells presented TUNEL positivity not only in the nuclei but also in 30% of the MNs (Fig. 4b), while there were no differences in 53BP1 or γ-H2AX staining in the MNs among the BT-549, MDA-231 and/or MCF-7 cells (Supplementary Fig. S4a). However, in further DNA damage analysis, RPA2 (replication protein A2) and PICH (PLK1-interacting checkpoint helicase), the factors involved in DNA replication, were stained in breast cancer cells. Some MNs in the BT-549 cells, but not MDA-231 and/or MCF-7 cells, exhibited bright foci with RPA2 or PICH staining (Fig. 4d, e), indicating DNA single or double breaks owing to DNA replication. Some MNs could be separately labeled by BrdU, indicating unscheduled DNA replication in the MNs (Fig. 4f). More interestingly, immuno uorescence staining of pHH3 (phosphorylated histone 3), a marker of chromatin condensation, could be detected in some MNs, especially in the BT-549 cells, similar to mitotic cells, and pHH3-stained MNs were usually Lamin B1 negative (Fig. 4f), which indicated that this DNA damage might be caused by a process similar to apoptosis in the MNs [22]. The results indicated that at least a portion of cytoplasmic DNA possibly came from the collapse or degradation of MNs.
Taken together, the results suggested that free cytoplasmic DNA was involved in the DNA damage response failure and its consequential MN formation, part of which underwent collapse owing to replication and DNA damage.

Inhibition of DNA autophagy induced growth arrest or cell death of cancer cells
The above results demonstrated that breast cancer cells presented high DNA autophagic activity, which raised the question of whether autophagy in uences the biological activity of this kind of cancer cell.
To further determine the role of autophagy in breast cancer cells, we grew MCF-7, MDA-MB-231 and BT-549 cells in the presence of ba lomycin A1 at various concentrations (0 nM, 1 nM, 5 nM, 10 nM) for 72 hours, and the results showed that the growth of both the MDA-MB-231 and BT-549 cells was inhibited in a dose-dependent manner (Fig. 5a). However, the BT-549 cells showed cytotoxicity in the presence of 10 nM ba lomycin A1. This nding was con rmed by time course analysis, in which the BT-549 cells with high autophagic activity died after treatment with 10 nM ba lomycin A1 for 72 hours, while the MDA-MB-231 cells only showed growth inhibition (Fig. 5b).
The effects of inhibition of DNA autophagy on cell growth and survival of breast cancer cells were also clari ed through silencing of either cGAS or SQSTM1 in the MCF-7, MDA-231 and BT-549 cells. The viability of the BT-549 cells was markedly reduced after either si-cGAS or si-SQSTM1 silencing. However, the MCF-7, MDA-231 and HeLa cells were not affected signi cantly (Fig. 5c). Similarly, the BT-549 cells, but not MDA-231 cells, were inhibited by treatment with the STING antagonist H-151 (2 µM, 10 µM, 20 M) for 24 hours (Supplementary Fig. S5a).
To further expand the observation of DNA autophagic inhibition to cell activity, we screened a series of human cancer cells with immuno uorescence double-staining of cGAS and SQSTM1, and HCT116, 786-0, PC3M, and DU145 cells were screened. Three categories of cancer cells-cells with high cGAS and SQSTM1 (786-0 and PC3M), low cGAS and SQSTM1 (HeLa and HCT116), and high SQSTM1 but low cGAS (DU145)-were selected (Fig. 5e). We treated the 786-0 cells with strong cGAS and SQSTM1 expression levels with CQ (50 µM) or ba lomycin A1 (10 nM) (Fig. 5f). In contrast, the cells with low cGAS and SQSTM1 levels were treated with CQ (50 µM) or ba lomycin A1 (10 nM). Moreover, the cells with high SQSTM1 but low cGAS levels were treated with CQ (50 µM) or ba lomycin A1 (10 nM) (Fig. 5g). The results proved that the growth or survival of cancer cells with high DNA autophagy was sensitive to autophagic inhibition.
To explore the cell death induced by DNA autophagic inhibition, we analyzed the CQ-or ba lomycin A1treated MCF-7, MDA-MB-231 and BT-549 cells, and the active form of caspase-3 was not detected by staining (Fig. 5h). Therefore, caspase-independent cell death (CICD) was investigated. Autophagy-related CICD could be achieved through lysosomal membrane permeabilization (LMP), but because CQ is an LMP inducer and ba lomycin A1 is an LMP inhibitor (data not shown), LMP-induced cell death could be ruled out. Alternatively, CICD is usually mediated by increasing ROS (reactive oxygen species) owing to mitochondrial outer-membrane permeabilization (MOMP). The production of ROS was measured in the presence of CQ and ba lomycin A1 by dihydroethidium (1 mM), and the results showed that either CQ or ba lomycin A1 could increase ethidium-stained BT-549 cells but only slightly affected the MDA-231 or MCF-7 cells (Fig. 5i), con rming that CQ or ba lomycin A1 treatment could induce caspase-independent cell death in the BT-549 cells.
The results demonstrated that DNA autophagy could be necessary for the survival of cancer cells by clearing cytoplasmic free DNA to protect against cell death.

Discussion
Autophagy is an important adaptive process to recycle substances or clear damaged organelles. For decades, autophagy has been thoroughly elucidated in terms of its process, forms, regulation and biological roles. DNA autophagy has also been identi ed, especially as an mechanism against exogenous invasion of organisms, and is considered an innate immune mechanism [23]. However, apparently, unlike that of other substances, autophagy of genomic DNA, the cellular genetic materials, is di cult to be considered for multicellular organisms. Nevertheless, in unicellular lower eukaryotes such as yeasts, DNA autophagy, in which cell nuclei can undergo autophagy by well-established regulatory pathways for nucleophagy through either piecemeal microautophagy of the nucleus (PMN) or late nucleophagy, has been identi ed [23]. Even entire nuclei could be degraded by macroautophagy in lamentous fungi. For mammalian cells, MN-related nucleophagy has been described occasionally, but mechanistically, its detailed process is still elusive. Mammalian cells have a nuclear lamina structure, which is different from that of lower eukaryotes, such as yeasts, with no comparable lamins or a brous nuclear envelope scaffold [23]. Nevertheless, nucleophagy is considered to play an important role in maintaining cellular genomic stability, detecting DNA damage, and regulating cellular apoptosis, as well as cellular senescence [24,25]. MN assays showed increased MN frequencies in breast cancer lymphocytes, which were correlated with the progression of breast cancer [26]. MNs are abnormal components that exists in the cytoplasm, independent of the nuclear nucleus. Autophagy also contributes to the elimination of MNs [23].
However, for most cancer cells, MNs can persist, indicating the general elimination of MNs by nucleophagy. In addition, similar to other studies, our investigations showed that cGAS and/or SQSTM1 could be detected in some MNs. However, the percentage of MNs with colocalization was low even in the highly abundant MNs in the BT-549 cells. Nucleophagy was not a prevalent event in cancer cells. A previous report showed the interaction between Lamin B1 and LC3 and suggesting that it is a nucleophagic mechanism (23). However, we still did not observe such interactions in our experiments, for instance, in some breast cell lines, such as BT-549, MDA-231, which usually showed a relatively high frequency of MN formation. Another possibility could not be ruled out: MNs undergoing nucleophagy could directly fuse with lysozyme, but this hypothesis needs to be further explored.
Selective DNA autophagy has been well clari ed in mammalian cells. It has been demonstrated that free DNA or RNA could directly mediate microautophagy via LAMP2 (lysosome-associated membrane protein 2) without the need for LC3 or other autophagic factors, but the nucleic acid transporter SIDT2 (SID1 transmembrane family member 2, SIDT2) is an integral lysosomal membrane protein for translocation into the lysosomal lumen. Nevertheless, speci c DNA sensors, such as cGAS-STING, have also been revealed to participate in DNA autophagic initiation. cGAS binds to DNA (in MNs or free) and recruits Beclin-1 and STING, promoting autophagy. cGAS generates cGAMP and stimulates the STING-Golgi apparatus, but some studies have also shown that STING itself could mediate autophagy after its binding to DNA, and the intrinsic domains of STING could directly interact with LC3 [8]. In a recent report, exogenous plasmids were rst recognized by DAI/ZBP1 (DNA-dependent activator of interferon regulatory factors/Z-DNA binding protein 1) but not cGAS [27]. However, in our study, upon introducing genomic DNA into the cancer cells, we found that cytosolic inclusions of various sizes were positive for cGAS, LC3 and lysosomes, con rming DNA autophagy existed in cancer cells (Supplementary Fig. S6). However, few inclusions were Beclin-1 positive (Supplementary Fig. S6). Therefore, it seemed that cytoplasmic DNA autophagy in cancer cells might involve different forms owing to the source of DNA and especially the DNA status and its level, nucleophagy, and selective autophagy. Apparently, the molecular mechanism could be diverse, such as LC3-dependent or LC3-independent DNA sensors. In actual conditions, free DNA is not naked but is instead usually bound to nuclear proteins. Therefore, genomic DNA from MN collapse or nuclear release might trigger a more complicated autophagic reaction to both DNA and proteins, especially ubiquitination (see below).
In recent years, extensive research has demonstrated that SASP is activated by the cGAS-STING pathway, and its proin ammatory role has been demonstrated to be crucial for the occurrence of autoin ammatory disorders, age-related diseases and even cancer progression [4]. However, either cGAS or STING alone or their combination could mediate DNA autophagy [8]. The described data also indicated that cGAS could play an important role in autophagy. These ndings raise an important question of how the decision to choose SASP or autophagy is made in cells. Our research showed that autophagy was usually found in breast cells with profound DNA damage such as the BT-549, while increased DNA damage in other cancer cells could induce DNA autophagic activity, indicating that the extent of DNA damage could be a factor in uencing this determination. cGAS-mediated SASP and autophagy could respond to DNA damage.
Severe DNA damage should be cleared by autophagy, but relatively less severe damage triggers SASP.
Apparently, the cellular ability of DNA damage repair could also be a factor. These ndings raise the question of whether there is any difference between cGAS in mediating SASP and autophagy. The details of cGAS recognition of DNA to induce autophagy or SASP are still unclear. However, more importantly, released genomic DNA from the cytoplasm is not protein-free but is instead bound with histones or other nuclear proteins [28]. The complex of DNA and protein could more easily activate autophagic activity since histones are usually ubiquitinated and generally recognizable by SQSTM1. To date, many studies have observed that SQSTM1 could be detected in the cell nucleus by either tagged SQSTM1 or immunohistochemistry [21]. In fact, the nuclear localization signal of SQSTM1 has been revealed, and its nuclear translocation has been demonstrated to be involved in the DNA damage response [29,30]. Therefore, the cytoplasmic or nuclear distribution of SQSTM1 could be similar. Thus, it is more likely that with severe DNA damage or repair failure, a relative amount of genomic DNA with coated proteins is released into the cytoplasm to trigger an autophagy-mediated clearance response.
Cytoplasmic DNA can easily be derived from mitochondria, organelles in the cytoplasm, but the mechanism by which genomic DNA accumulates in the cytoplasm is still unclear. MN formation is believed to be a major source since various reports have shown that MNs from some cancer cells are not intact in their nuclear lamina due to RB de ciency. Indeed, defects in nuclear membrane assembly, either Chromothripsis has been demonstrated to be a consequence of DNA damage in MNs [10]. This research also showed that unscheduled DNA replication and DNA damage could be detected in a portion of the MNs. Interestingly, some BT-549 cells with abundant MNs more frequently presented condensed focal staining of RPA2 or PICH, both of which participate in the DNA damage response; the former usually binds to single-stranded DNA, and the latter binds to double-stranded DNA [33,34]. More importantly, some MNs were frequently observed in pHH3-positive cells, similar to BT-549 cells (Fig. 4f). It has been acknowledged that pHH3 is mainly found during chromatin condensation in mitosis and in apoptotic nuclei of cells [35,36]. For determination of whether pHH3-positive MNs undergo mitosis, cells were stained for Hec1, a protein involved in kinetochore assembly. The results proved that no MNs were positive, but only mitotic and apoptotic cells were stained (data not shown). In a cell-free apoptotic model, nuclear condensation could sequentially proceed with condensation, a nuclear necklace, collapse or disassembly [22,37].
Similarly, pHH3-stained MNs also presented these morphological changes (Fig. 4f), indicating that MNs could undergo collapse owing to replication and damage in breast cancer cells with DNA autopagy.
Accordingly, a portion of the MNs in the BT-549 cells showed positive staining of RPA2 and PICH, especially in the foci-staining pattern (Fig. 4d, e), indicating severe DNA damage in these MNs. Apparently, the more MNs formed, the more frequently breakage was detected. In addition, the release of nuclear eccrDNA (extrachromosomal circular rDNA) should be another source of cytoplasm (see below).
Since MN formation can be generally induced by a variety of genotoxic agents, DNA damage is reasonably considered a key process, and aberrant mitosis is widely accepted [6,12].However, recent studies have suggested that for cancer cells, DNA replication stress could be a likely common mechanism [33,38,39]. Oncogenic mutations induce accelerated DNA replication and trigger replication stress. The so-called common fragile sites in the genome, such as rDNA, are di cult to replicate, and stalled or collapsed replication forks usually induce the formation of UFBs (ultra ne bridges) or lagging chromosomes to result in MN formation or to generate free DNA fragments, such as eccrDNA, which are hard to enclose in the late phase of mitosis during nuclear membrane assembly and are consequently released to the cytoplasm [16,34]. In our investigation, in addition to MNs, cytoplasmic DNA from the genome, including Alu-repeated sequences and rDNA, was easily detected, indicating a replication stressrelated mechanism. Moreover, intra-S phase checkpoints mediated by ATR and ATM kinases are crucial to replication stress, and their de ciency causes replication stress-related DNA damage [40]. The BT-549 cells generally had low levels of pATM and pATR and high formation of MNs and cytoplasmic DNA as well as strong TUNEL staining, suggesting relationships between these factors.
Autophagy is a form of cellular activity that adapts to endogenous and environmental changes. Although gene mutations related to the regulation of autophagy have been clari ed in tumorigenesis, for some kinds of cancer cells, inhibition of autophagy could promote their growth and survival, indicating that autophagic activity could be necessary for these cancer cells [41]. However, how to determine the sensitivity of cancers to autophagic inhibition is still not established. This investigation mainly showed that inhibition of DNA autophagy could decrease cell viability in breast cancer cells even more cancer cells, suggesting its potential therapeutic utility in cancer treatment, especially for cancer cells de cient in DNA repair. In recent years, DNA damage repair de ciency has been successfully used in cancer therapy; for instance, cancers with MSI (microsatellite instability) can be treated with immune checkpoint blockadebased immunotherapies, while genomic mutations of BRCA1/2 or HRR (homologous recombination repair) are targets of PARP inhibitors. DDR de ciency has been considered a promising anticancer target [42][43][44]. Targeting autophagy could be another approach to treat DDR-de cient cancers.

Conclusion
In summary (Fig. 6) Table 1. b TUNEL analysis of breast cancer cells. A TUNEL assay was