Autophagy is a highly conserved process that functions to transport cargos to the lysosome for recycling and cellular degradation in eukaryotes [1]. Autophagy not only serves to remove defective or damaged organelles and cellular components by self-digestion, as a catabolic mechanism, it recycles substrates required to sustain homeostasis when nutrients are scarce ^[1][2][1,2]. The significance of proper autophagy extends to the soundness of immune cell function [3], intercellular communication [4], regulation of tissue-resident stem cells ^[5][6][5,6], and the integrity of the tissue barrier [7]. It can be triggered by tissue remodeling, long-term nutritional deprivation, quality control of organelles, cellular stress, and immune reaction [8]. In ideal circumstances, autophagy can be cytoprotective or destructive because an immoderate self-degradation process can be damaging [9]. As the key cellular process that regulates the stress response and thus takes part in the quality control in the cells ^[10][11][10,11], autophagy has a recondite impact on human lifespan and health [4]. Consequently, numerous human illnesses, including neurodegeneration, myopathies, cancer, aging, and lung, liver and heart diseases, as well as metabolic complications such as diabetes are linked to autophagic dysfunction [12].

Autophagy can be distinguished into four categories based on how the protein is transported to the lysosome [2]—microautophagy, chaperone-mediated autophagy, macroautophagy, and selective autophagy [2]. In microautophagy, the lysosomal membrane will undergo invagination or protrusion for cargo uptake [13]. Instead of manipulating membrane structures, chaperone-mediated autophagy uses chaperones for cargo identification that carry a specific pentapeptide motif. Subsequently, each of these components are then unfolded and individually translocated through the lysosomal membrane [14]. In contrast, macroautophagy generates double-membrane vehicles (autophagosomes) for cargo sequestration [15]. Guided by specific autophagy-related genes (ATGs) and BECN1 (Beclin-1), the initiation step of macroautophagy precedes phagophore elongation, autophagosome maturation, and fusion of lysosome and autophagosome. This process is concluded by proteolytic degradation of the cargo [2]. On the other hand, the macroautophagy of a particular cellular component is known as selective autophagy. Different from macroautophagy, the key regulator of selective autophagy is PINK1 (phosphatase and tensin homolog-induced putative kinase 1) [2]. The specificity of selective autophagy is preserved by ubiquitination or labeling of each cargo. In this process, p62 is an autophagy substrate that serves as a reporter [16]. Subsequently, autophagy receptors selectively bind to the tagged cargo and proceed to the formation of autophagosome ^[17][18][17,18]. These types of autophagy are mechanically varied, but they all culminate in lysosomal degradation of unwanted substances in the cell [1]. Adding to this complexity, several key proteins of autophagy machinery are known to be regulated by long non-coding RNAs (lncRNAs). LncRNA H19 promotes autophagy via modulating the Let-7–Lin28 axis [19], whereas LC3 and beclin1 are targeted by lncRNA ROR leading to autophagy promotion [20]. ATG10 is activated by direct binding of the lncRNA AGAP2-AS1–ELAVL1 complex to its promoter region [21]. LncRNAs, including HOTAIR, TALNEC2, EGOT, ZNF649, GAS5, DANCR, OTUD6B, and NAMPT, have been reported to regulate the expression of autophagy-associated proteins and impact cancer progression ^[22][23][22,23].

Interestingly, in cancer cells autophagy contributes to both death and survival [24]. The impact of autophagy in homeostasis serves to guard the genomic integrity of quiescent and growing cells in tissues [25]. Since genome instability is one of the cancer hallmarks, fidelity of autophagy has the leverage to prevent healthy cells from becoming cancerous ^[4][26][4,26]. It has been reported that autophagy in healthy cells prevents tumorigenesis via counteracting pro-oncogene stimuli [27]. Autophagy also activates the oncogene-induced senescence program, which keeps proliferative events at bay [28]. Nevertheless, many factors, such as the stage of disease, type of cancer, and condition of the patient can interfere with the real impact of autophagy in the progression of cancer [29].

Healthy cells commonly face intrinsic and extrinsic stress that can potentially result in genomic instability and mutations, which will aid neoplasia and hyperproliferation [30]. Autophagy serves to prevent such complications by eliminating tumorigenic stressors such as oncoproteins, protein aggregates, reactive oxygen species (ROS) production, and dysfunctional mitochondria ^{[26][31][32][33][34][35][36][37][38][39]}[26,31,32,33,34,35,36,37,38,39]. Besides that, autophagy has roles in immune responses and inflammation ^{[40][41][42][43]}[40,41,42,43]. As such, the maintenance of cellular integrity and defense against neoplastic transformation are both facilitated by autophagy [40]. Owing to the cytoprotective function of autophagy, it serves to suppress tumorigenesis in terms of cancer initiation. Indeed, an elevated gene signature for autophagy is observed in healthy mammary glands, which is found to be decreased as breast cancer progresses [44]. Indeed, autophagy is a complex multistep process involving multiple proteins that participate in breast cancer initiation, growth, and metastatic progression as well as recurrence.

Macroautophagy is the most well-studied subtype of autophagy [9]. When cells are under stress, ULK1 is activated directly or indirectly leading to the recruitment of BECN1 and ATGs, thus allowing the assembly of molecular complexes, which subsequently lead to the initiation of phagophore formation ^[45][46][45,46]. BECN1 deficiency is observed in breast cancer [47]. Consistently, monoallelic loss of BECN1 is often observed in human breast cancer cells ^[47][48][47,48]. Further, progression of ex vivo HER2-enriched breast tumor is hindered upon overexpression of BECN1 [49]. In addition, monoallelic deletion of Becn1 in FVB/N mice results in the development of mammary tumors after parity [50]. In fact, mammary tumorigenesis in MMTV-Wnt1 mice with monoallelic Becn1 deletion is more aggressive compared with those in mice with homozygous Becn1 [50]. Among human breast cancer samples excluding HER2-enriched tumors, tumors with overexpression of WNT-related genes and a low mRNA level of BECN1 present a poorer prognosis, and they are primarily TNBC ^[50][51][50,51]. On the other hand, it has been recognized that the etiology and aggressive phenotype of TNBC is related to the activation of the Notch1 pathway [52]. Macroautophagy has roles in such oncogenic signaling as well. Of note, BECN1 can induce autophagic degradation of Notch1, which leads to a phenotype that diminishes Notch1-signaling-dependent tumorigenesis. Indeed, silencing BECN1 in TNBC cell lines leads to an enhanced clonogenicity, migration, and anchorage-independent growth [53]. These studies suggest that the expression of the autophagic indicator, BECN1, hampers the progression of breast cancer.

There are approximately twenty evolutionarily conserved ATGs that actively participate in the autophagic process. Depending on the context, some ATGs may contribute to the prevention of tumorigenesis [43]. For example, enrichment of ATG7 has a negative impact on growth and glycolysis in TNBC cells. Similarly, TNBC tumors that bear a higher level of ATG7 present better prognoses [54]. On the other hand, autophagy in human breast epithelial cells with mutationally active oncogenic Ras can be pro-tumorigenic [55]. In such mutated cells, an enhanced glycolysis capacity and proliferation is observed in autophagy-competent cells compared with autophagy-deficient cells. Additionally, more autophagy-competent cells undergo Ras-mediated adhesion-independent transformation, which suggest that autophagy has the potential to stimulate Ras-mediated tumorigenesis under certain metabolic conditions [55]. Of note, given that autophagy can stimulate Signal Transducer and Activator of Transcription 3 (STAT3), and that STAT3 is frequently activated in TNBC, modulation of autophagy influences the TNBC subtype the most [56].

As a putative tumor suppressor, Forkhead Box O (FOXO) is a transcription factor that takes part in regulating cellular homeostasis, the maintenance of stemness, and aging ^[43][57][43,57]. Downregulation of FOXO1 in breast cancer is related to a worse prognosis in breast cancer, especially in HER2-positive subtypes [58]. Similarly, nuclear localization of FOXO3 is found to be related to a reduced metastatic event in luminal-like breast cancer [59]. Interestingly, FOXO3 has the leverage to induce the expression of proteins that participate in the initiation and autophagosome formation in macroautophagy. In agreement with this, the loss of FOXO3 leads to diminished expression of several ATGs, resulting in a declined activity of autophagy ^[60][61][62][60,61,62]. Indeed, tumorigenesis can be stimulated by the absence of FOXO3, which implies suppression of FOXO3-mediated autophagy contributes to mammary carcinogenesis ^[60][61][62][60,61,62]. To add another level of complexity to this matter, FOXO3 also induces autophagy in cancer stem cells (CSCs) to preserve their well-being, and thus contributes to recurrence and metastasis ^[63][64][63,64]. Altogether, FOXO3 has the leverage to suppress tumorigenesis in healthy cells but may induce cytoprotective autophagy in cancer stem cells.

Breast-cancer-related mortality has been increasing in the past two decades ^[43][65][43,65]. Following preliminary diagnosis, metastatic relapse accounts for 90% of breast-cancer-related deaths, which is ascribed to the resurgence of dormant breast cancer cells ^[66][67][66,67]. Many therapeutics for breast cancer, namely chemotherapy, target actively dividing cells by destructing DNA and key proteins ^{[68][69][70][71]}[68,69,70,71]. In such circumstances, autophagy serves to breakdown long-lived proteins, macromolecular waste, and damaged organelles. Residual cancer cells that survive the therapeutic assault may result in dormancy transformation. Since autophagy can be used as nutritional support, the cells will have time to repair and thus contribute to chemoresistance, relapse, and disease progression [72]. Autophagy thereby decreases drug sensitivity to breast cancer cells while protecting them. Hence, autophagy can be cytoprotective to breast cancer cells [72].

When the environment becomes unfavorable for growth, tumor cells can become quiescent, which is termed tumor dormancy [73]. It has been discussed that tumor dormancy largely contributes to metastasis, disease recurrence, and therapy resistance ^[73][74][73,74]. Dormant cancer cells can remain latent for decades before being activated to a proliferative state [75]. Autophagy not only supports the growth of dormant cells within the tumor microenvironment (TME) ^[76][77][76,77] but also participates in distant colonization and extravasation of dormant cells under environmental stress ^[66][78][66,78]. Upon inhibition of autophagy in vivo and in human preclinical models of dormant breast cancer cells with hydroxychloroquine (HCQ), a substantial decline of metastasis burden and cell viability have been observed [66]. This phenomenon is ascribed to the accumulation of damaged mitochondria and ROS, which in turns leads to cell death [72]. Further, 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (PFKFB3) acts to promote cell cycle progression while inhibiting apoptosis [43]. PFKFB3 can also be a substrate for autophagosomal degradation by interacting with the autophagy receptor p62 [79]. Interestingly, metastatic breast cancer cells bear higher levels of PFKFB3 compared with dormant cells but have less autophagic activity. Of note, low levels of autophagy stabilize PFKFB3, which leads to activation of dormant cells to a metastatic state [79]. These findings suggest that autophagy promotes dormancy in breast cancer cells [43]. The AMP-activated protein kinase (AMPK) and mammalian target of rapamycin (mTOR) pathways are essential for autophagic regulation in tumor cells [80]. Canonically, activation of the mTOR complex 1 (mTORC1) directly phosphorylates and thus sequesters transcription factor EB (TFEB) in the cytoplasm ^[81][82][81,82]. Since TFEB is a chief transcriptional regulator of lysosomal and autophagy genes, activation of mTOR suppresses induction of autophagy at the transcriptional level [82]. For AMPK-mediated regulation of autophagy, AMPK is activated upon energy depletion, which in turn inhibits the autophagy regulatory complex, thus resulting in disruption of autophagosome biogenesis ^[83][84][83,84]. Additionally, activated AMPK can phosphorylate Unc-51-like kinase (ULK1) and the TSC1/TSC2 complex, thereby inducing autophagy via suppressing the activation of mTORC1 ^[80][85][80,85]. In breast cancer, environmental stress stimulates the secretion of auto- and paracrine signaling factors, which block phosphoinositide 3-kinase (PI3K) activation and lead to the inactivation of AKT and mTOR, thereby resulting in the activation of autophagy ^[43][75][43,75]. The PI3K/AKT/mTOR pathways can be inhibited by Diras Family GTPase 3 (DIRAS3), which is found to be enriched in dormant breast cancer cells ^[75][86][87][75,86,87], suggesting that autophagy may contribute to tumor dormancy in breast cancer and thus plays a role in chemoresistance.

It is known that cancer stem cells are a significant contributor to the development of chemoresistance in breast cancer ^{[88][89][90][91]}[88,89,90,91]. Autophagy enables the survival of CSCs under hypoxia in the tumor microenvironment [92]. In fact, a subpopulation of TNBC cancer stem cells stay in an autophagic state in relation to hypoxia [93]. Indeed, environmental stress such as nutrient deprivation and hypoxia can lead to the activation of autophagy for cellular component recycling in order to sustain survival ^[94][95][96][94,95,96]. In cancer cells, hypoxia-inducible factor-1 (HIF-1) is the main regulator of hypoxic conditions ^[97][98][97,98]. Upon activation of HIF-1, stemness can be triggered via several pathways such as activation of NANOG, SOX2, SOX17, etc. ^{[98][99][100]}[98,99,100]. Importantly, hypoxia stimulates autophagy via HIF-1α [101], a subunit of HIF-1. HIF-1α is involved in the generation, differentiation, invasion, plasticity, and therapeutic resistance of CSCs [92]. In two stem-like breast cancer cell lines, induction of stemness can be performed by autophagy via the EGFR/Stat3 and TGFβ/Smad pathways in a murine model [102]. It is also reported that inhibition of autophagy in certain breast cancer cell lines results in a decreased stemness phenotype ^[56][103][56,103]. Other than that, dormant stem-cell-like breast cancer cells express autophagy markers, and upon inhibition of autophagy using 3-methyladenine (3-MA), these cells are transformed to active state [104]. Moreover, doxycycline not only inhibits EMT (epithelial–mesenchymal transition) and stemness markers in breast cancer stem cells but also causes a down-regulation of autophagy activity [105], suggesting the possibility that autophagy may play a role in stemness [106].

In patient-derived xenografts, suppression of autophagy via inhibition of BECN1 leads to re-sensitization of chemoresistant cells to therapy [93], which emphasizes the role of autophagy in the development of chemoresistance. In luminal and HER2-enriched subtypes of breast cancer, similar results are observed, as chemoresistant cells not only have an elevated autophagic activity compared with their drug-sensitive counterparts but inhibition of autophagy also results in the restoration of chemosensitivity ^{[107][108][109][110]}[107,108,109,110]. Further, expression of mesenchymal markers, vimentin, and the stem cell marker CD44 is increased upon autophagic activity in CSCs [106]. Additionally, self-renewal of a hormone-independent murine breast cancer cell line LM38-LP requires autophagy [111]. Interestingly, disruption of circadian rhythm is related to the acquirement of chemoresistance [112]. By itself, melatonin suppresses the development of chemoresistance in breast cancer by interfering with tumor metabolism ^[112][113][112,113]. However, the combination of dim light at night (dLAN) results in the activation of STAT3, which is often overexpressed in paclitaxel-resistant breast cancer [112]. Under the synergetic influence of dLAN and melatonin, the activated STAT3 inhibits DIRAS3 in an epigenetic manner, resulting in decreased autophagic activity and increased resistance of breast cancer to paclitaxel [112]. This finding implies that DIRAS3 can be a regulator for the development of chemoresistance via autophagy. On the other hand, it has been demonstrated that chemotherapeutics can trigger autophagy, which enhances the survival of CSCs [114]. In TNBC, inhibiting autophagy with chloroquine (CQ) causes the accumulation of dysfunctional mitochondria and ROS in CSCs, which results in cell death [115]. Interference of autophagy also disrupts the preservation of breast CSCs. When combining autophagy inhibitors and chemotherapy, a decreased expression of stemness markers is observed along with an increased sensitivity to chemotherapeutics and a decreased cancer cell viability and metastasis ^{[105][111][115][116]}[105,111,115,116]. Indeed, autophagy can be cytoprotective by contributing to induced chemosensitivity in breast cancer cells [72].

Autophagy flux can impact how breast cancer cells respond to treatment. Unsurprisingly, in-depth inquiries about autophagy have discovered that intermediate regulation of autophagy flux can impact the influence of autophagy on therapeutic resistance [72]. For example, the reporter in selective autophagy, p62, also has roles in the proteolytic system [16]. It not only functions to deliver ubiquitinated proteins to the proteasome for breakdown but also governs protein quality by binding with ubiquitinated cargoes while shuttling between the nucleus and cytoplasm [16]. Upon administration of bortezomib (a proteasome inhibitor), an elevated p62 expression is observed, which implies the failure in the turnover of autolysosomal protein, demonstrating the interplay between proteasomal degradation and selective autophagy [117]. Breast cancer cells fail to restore metabolic homeostasis via autophagy upon inhibition of autophagosomal degradation using obatoclax [118]. Further, the combined treatment of bortezomib and obatoclax in antiestrogen-resistant breast cancer cells results in a hindered autolysosomal function without preventing the formation of autophagosomes ^[117][118][117,118]. Interestingly, this sensitizes antiestrogen-resistant breast cancer cells to tamoxifen ^[76][118][76,118], which indicates that the indirect influence of the proteasome pathway can contribute to drug resistance. On the other hand, lysosomal-associated protein transmembrane 4β (LAPTM4B) is an essential maturation step in autophagy but also plays a significant role in lysosomal activities [119]. Expression of LAPTM4B is positively related to chemoresistance in breast cancer [119]. From the mechanistic standpoint, deficiency of LAPTM4B leads to an enhanced permeability of the lysosomal and the autolysosomal membranes [120]. Consequently, drugs can enter the nucleus more readily. The increased permeability of cathepsin also causes cathepsin to be released, which triggers lysosomal-mediated programmed cell death [120]. LAPTM4B deficiency significantly hinders the fusion of lysosomes and autophagic bodies, which begets the accumulation of autophagosomes, resulting in cell death [121] and thus diminishing therapeutic resistance. In summary, simply increasing or decreasing autophagic activity is not a wise route to regulate autophagy-mediated therapeutic resistance in breast cancer. It is also important to inquire how autophagy regulates the sensitivity of breast cancer to therapeutics [72].