Hypoxia-mediated Warburg-like metabolic changes are analogous to the Darwinian evolutionary model (23). Epithelial cells acquiring oncogenic transformation is the first step in breast carcinogenesis, whether under the stochastic or stem cell hypothesis (38). Upon reciprocal touch, non-oncogenic breast epithelial cells abide by contact inhibition and halt growing and reproducing (39). However, cells with oncogenic transformation evade contact inhibition (40), continuing to grow and proliferate in a compacted epithelial environment until apical extrusion occurs, completing the transition from monolayer to multilayer epithelium under the coordination of the CD44/COL17A1 pathway and progressively developing into pre-malignant lesions (41) which are separated from the surrounding matrix by a nearly continuous layer of myoepithelial and basement membrane proteins (42). Therefore, oxygen in the vascular matrix must traverse the basement membrane and tumor cell layer. Studies have indicated a limiting distance for oxygen diffusion (approximately 150µm), pre-malignant lesions with uncontrolled proliferation are bound to exceed this diffusion limit, resulting in a hypoxic region (23). Under normoxic conditions, hypoxia-inducible factors (HIFs) are rapidly degraded by prolyl hydroxylation of EglN family members, whereas hypoxia inhibits EglN1 activity, resulting in a massive accumulation of HIFs. In addition, glutamate secretion from TNBC causes EglN1 to undergo oxidative self-inactivation, which can also prevent HIFs from being degraded (43). Studies have shown that HIF directly or indirectly up-regulates the expression of critical glycolysis regulatory elements, resulting in metabolic adaptive transformation (OXPHOS to glycolysis), and it simultaneously stimulates angiogenesis to prepare for eventual vascular normalization. However, angiogenesis invariably lags behind, exposing tumor cells to persistent hypoxic stress (44, 45). Only glycolytic tumor cells, according to the evolutionary models, can survive in a constant hypoxic environment. Mutations or epigenetic modifications during selection lead to a glycolytic constitutive expression (manifested as the removal of oxygen concentration limitation) that is up-regulated in response to specific survival pressures, thereby conferring a survival advantage to tumor cells ( Figure 1 ) (23). The “specific survival pressures” argument is based on the metabolic plasticity doctrine, which holds that the dominant metabolism maintained by tumor cells must be elastic in response to variations in survival pressures (15). For example, TNBC, which keeps a hybrid metabolic phenotype, activates the AMPK pathway to enhance OXPHOS when exposed to glycolytic inhibitor toxicity and vice versa (21). In addition, the 3D model of TNBC developed by Liu C. et al. revealed that the tumor tissue exhibits escalating matrix stiffness from the core to the periphery, and the dominant metabolism also manifests a distribution from the core glycolysis to the peripheral OXPHOS and fatty acid metabolism (46). Microenvironmental acidification and glucose limitation as consequences of up-regulation of glycolysis will further select tumor cells with survival advantages (47–50).

ECM, a substantial 3D supramolecular entity composed of hundreds of different building blocks, is an essential component of TNBC, and its changes profoundly affect the fate of TNBC cells (51, 52). Divergent matrix-remodeling strategies mediated by Mmp14/MT1-MMP define the branching morphology of developing and neoplastic epithelial cells early in breast carcinogenesis (53). This process may involve the effect of matrix remodeling on the chromatin state of mammary epithelial cells (54). On the other hand, ECM stimulates tumor growth in various ways throughout the development of TNBC. For instance, ECM compliance controls the cellular response to contact guidance by generating population transfer between round and elongated cells, hence affecting the migratory dynamics of tumor cells (55). Consequently, ECM remodeling plays a crucial role in carcinogenesis and progression. The preceding section identified hypoxia as an early event in TNBC. Interestingly, current research indicates hypoxia as a potential trigger for ECM remodeling (56, 57). Increased matrix proteins due to disruption of ECM deposition and degradation homeostasis (58), enhanced matrix cross-linking driven by lysyl oxidase (LOX) (59), and local compaction and distant stretching of matrix fibers derived from CAFs-mediated contractile forces (60, 61) are direct behaviors that cause ECM remodeling. Studies have demonstrated that hypoxia affects the aforementioned behaviors in a HIF-dependent manner (57), such as HIF-1α/miR-142-3p/LOX pathway (62) and the HIF-1α/CAF pathway (63), ultimately leading to changes in ECM biophysical and biological properties. Notably, a recent study in which ECM remodeling activated the glycolytic pathway in murine 4T1 cells suggests that ECM is involved in regulating the metabolism of TBNC cells (64). Concerning its regulation mechanism, Sullivan WJ et al. determined that ECM remodeling boosted lactate generation in TNBC cells via the Zfp36/TXNIP/Glut1 signaling pathway (65). Additionally, connective tissue growth factor (CTGF) in conjunction with ECM deposition accelerated TNBC glycolysis via integrin β3/FAK/Src/NF-B P65/Glut3 signaling (66). Moreover, ROCK isoforms differentially regulate the RhoA/ROCK1/p-MLC and RhoA/ROCK2/p-cofilin pathways through integrin β1-activated FAK signaling in an ECM stiffness-dependent manner to regulate MDA-MB-231 cell motility (67), and Rho/ROCK signaling can stimulate Glut1 translocation to the PM and enhance extra-membrane glucose uptake TNBC cells (68). In summary, hypoxia promotes matrix remodeling in various ways that further influence the metabolic pattern of TNBC cells ( Figure 2 ), which simultaneously provides suggestions for developing innovative ECM-targeting treatment approaches (69).

Growth factor stimulation of TNBC cells commences the proliferation program (70), and cell proliferation undergoes three periods: I. interphase; II. mitosis; and III. cytokinesis which is called the cell cycle. The interphase accounts for about 90% of the time required for the entire cell cycle and is divided into three phases, namely G1, S, and G2. The cell cycle is precisely modulated by checkpoints such as cyclin-cyclin-dependent kinase complex (cyclin-CDK) and CDK inhibitors (examples include P15, P21, P27), inactivation or mutation of anti-oncogenes, and overexpression or amplification of proto-oncogenes leading to cell cycle checkpoint disorders and ultimately to uncontrolled cancer cell proliferation (71, 72). Positive cell cycle regulation of the Warburg effect is generally achieved through cell cycle checkpoints (73), such as critical glycolytic kinase 6-phosphoglucose-2-kinase (PFKFB3), pyruvate kinase (PKM2) was shown to promote cell cycle (72). Mechanistically, PFKFB3 is transported to the nucleus, and its products F2 and 6BP activate CDKs. Moreover, F2 and 6BP induce degradation of the CDK-mediated G1/S transition inhibitor P27 (74, 75). PKM2 binds to β-linked proteins to form the PKM2-β-linked protein complex to up-regulate cell cycle protein D1 to drive the G1/S transition (72). A study targeting PKM2 to regulate cell cycle progression and inhibit TNBC invasion suggests that further study of the interaction between the Warburg effect and cell cycle is a concern in TNBC management (76).

Tumor metastasis starts from the early stage of the tumor and continues until the cancer is removed. It can be divided into three periods: dissemination, metastatic dormancy, and host organ colonization (77). This process is not a simple linear cascade; concurrency and overlap are more prevalent (78). During dissemination, metastasis-initiating cells (MICs), which are assumed to be derived from cancer stem cells (CSCs) or CSC-like cells, lose their epithelial features and acquire mesenchymal traits through epithelial-mesenchymal transition (EMT) (79, 80), primarily manifested by downregulation of E-calmodulin and reduced cell adhesion, in which the Warburg effect plays an instrumental role (81). For example, high expression of Y-box binding protein 1 (YBX1) was detected in TNBC, which up-regulates aerobic glycolytic flux to accelerate EMT progression (82). Immediately afterward, MICs with mesenchymal properties detach from the tumor bed as single cells or cell clusters and become circulating tumor cells (CTCs) (79). In fact, CTC clusters are more conducive to tumor metastasis (83), so primary tumors significantly up-regulate HIF-1α-mediated Desmoglein2 (DSG2) to promote the formation of CTC clusters (84).

Subsequently, CTCs enter the TME, and the acidic TME shaped by the accumulation of glycolytic acid metabolites intensifies CTCs migration (85, 86). In addition, Glut3 overexpression in TNBC induces activation of M1 tumor-associated macrophages (M1-TAM) via lactate/C-X-C motif Chemokine ligand 8 (CXCL8), consequently creating an inflammatory TME that promotes CTCs metastasis (87). Notably, TNBC delivers Integrin beta 4 (ITGB4) to CAFs located in TME through exosomes and initiates the glycolytic phenotype of CAFs via BNIP3L-dependent mitophagy (known as the reverse Warburg effect) (88). Its product lactate shuttles between CAFs and CTCs via MCT4/MCT1 as a metabolic coupling link and activates the TGFβ1/P38 MAPK/MMP2/9 signaling axis to up-regulate CTCs’ mitochondrial activity and promote metastasis (89).Even more to the point, glycolysis is not the sole up-regulated metabolic phenotype in cancer cell migration. Other metabolic patterns, such as OXPHOS, fatty acid metabolism, and glutamine metabolism, are similarly up-regulated under certain conditions (90). And the metabolic phenotype of CTCs is dynamically remodeled at different stages of tumor metastasis to help them manage the changing microenvironment of the metastatic cascade and ensure a successful transition ( Figure 3 ) (91). Next, CTCs migrate towards the vasculature, and the Glut1/integrin β1/Src/focal adhesion kinase (FAK) signaling pathway activated by PCB29-P exposure may be involved in this process (92). CTCs entering the blood circulation are challenged by ROS, and high concentrations of ROS are cytotoxic. At this time, the advantages of CTCs clusters gradually emerge, which induce microenvironmental hypoxia and activate HIF-1α-mediated glycolysis in response to oxidative stress (93). Moreover, when CTC reaches the bloodstream and is threatened by Anoikis, only a few Anoikis-resistant tumor cells survive (94). And the phosphorylation and activation of lactate dehydrogenase (LDHA), a glycolytic critical kinase, has been discovered to boost Anoikis resistance (95).

Next is an overlap of host organ colonization and metastatic dormancy, in which the majority of tumors metastasize to specific organs, called “organotropism” (96). Studies have shown that the lung, bones, and brain are more likely to serve as TNBC target organs (97, 98). After reaching the targeted organ, CTCs occupy the favorable niche and become disseminated tumor cells (DTCs). Due to the balance between tumor cell proliferation and apoptosis, DTCs enter the reversible growth arrest of single cells located in the G0/G1 phase of the cell cycle, called cell dormancy. This persists for years or even decades, and DTCs reactivate when the balance swings toward cell proliferation, either because of proangiogenic substances or because tumor cells defy immune suppression (77, 79, 99). For instance, CD39 PD-1 CD8 T cells were demonstrated to mediate metastatic dormancy in TNBC lung-oriented DTCs, and depletion of CD8 T cells may induce DTCs reactivation (100). Activated DTCs destroy host organs and form metastases, such as up-regulation of the Warburg effect by MDA-MB-231 cells found in bone TME to produce excessive lactate, leading to MCT4/MCT1-mediated osteolytic lesions (101).

Minimal residual disease (MRD) is considered the basis of TNBC recurrence (102), and postoperative circulating tumor DNA (ctDNA) sequencing and CTCs counting have been used to detect MRD and evaluate the likelihood of patients suffering from recurrence (103). Cell dormancy is critical for MRD survival and reactivation (104, 105). Genetic heterogeneity, TME regulation, ecological niche, and immunosuppression drive MRD into a dormancy (106) characterized by cell cycle arrest, ecological niche dependence, drug resistance, immune escape, and reversibility (107). When conditions such as surgery, drugs, and metabolic remodeling intervene, MRD exits dormancy and causes recurrence (108). Glycolysis is possibly involved in the reactivation of dormant MRD. Studies have shown that dormant cells exhibit an increase in mitochondrial mass and mitochondrial ROS. Therefore, dormant cells inhibit oxidative stress-induced apoptosis via autophagy-related 7 (ATG7) mediated autophagy (109), in which up-regulated AGT7 suppresses the glycolytic phenotype (110). During the transitional period of dormant cells toward reactivation, autophagic inhibition and increased abundance of PFKFB3, a key mediator of glycolysis, facilitate dormant escape (111). The argument that a negative correlation between autophagy and the Warburg phenotype differs from Radic Shechter et al., who found that MRD retains the metabolic memory of original tumor cells and maintains high aerobic glycolytic flux to ensure survival (112). A possible explanation is that MRDs dynamically reshape the balance between autophagy and glycolysis in response to varied ecological niches (113). In addition, consistent with the primary tumor, the metabolism of the cells that undergo macroscopical metastasis after escaping the dormant state continues to be biased toward glycolysis to sustain growth (114). Remarkably, the tricarboxylic acid cycle (TCA cycle) metabolites succinate, fumarate, and 2-hydroxyglutarate (2-HG) were associated with post-treatment tumor reconstruction (115), and glycolytic product lactate functions as a substrate to support TCA cycle running (116).

As evident from recent reviews (106, 108, 117), despite the emerging roles of the Warburg phenotype in metastatic dormancy of tumor cells, there is a considerable blankness in the relevant research, including how the Warburg phenotype controls DTCs to enter a dormancy state? How does the Warburg phenotype help dormant DTCs survive? Furthermore, how the Warburg phenotype regulates dormant DTCs reactivation? Metastatic dormancy is essential for DTCs survival and reactivation as a protective mechanism (118), so filling these gaps may provide new clinical opportunities for TNBC.

From the above discussion, it can be clarified that while TNBCs adaptively modulate their metabolic phenotype following different conditions, TNBCs up-regulate the Warburg phenotype to maintain their invasion activity is still a high probability event. Accordingly, high glycolytic flux means an adequate supply of extra-membrane glucose is necessary. Increasing evidence that most immune cells present in TME ( Figure 4 ), including innate immune cells such as NK cells, macrophages, and dendritic cells; adaptive immune cells, such as T cells in the activated state, mimic the metabolic patterns of tumor cells (119, 120). Therefore, the competition between tumor cells and immune cells for metabolic substrates is inevitable (121), and the consequence is always that immune cell function is curbed due to the lack of fuel (122). Because tumor cells will induce immune cells’ glycolytic suppression, for example, IFN-γ secreted by T cells or the glycolytic phenotype of the tumor itself triggers PD-L1 expression in tumor cells. In turn, PD-L1 mediates up-regulation of the glycolytic pathway in tumor cells via the Akt/mTOR pathway, imposing a nutrient limitation on T cells. Prolonged exposure to glucose limitation will lead to irreversible functional impairment that cannot be corrected by simple nutrient supplementation (122, 123). In addition, the expression of CTL-4 associated with immune suppression and its principal ligand CD80/CD86 were detected in TNBCs (124), and a study of CTL-4 blockade by ipilimumab linked Treg instability to its glycolysis, suggesting the involvement of CTL-4 in metabolic competition between TNBCs and immune cells (125).

Lactate mediates multiple pro-tumor behaviors, including immune escape (126, 127). For instance, lactate works directly with mitochondrial antiviral signaling proteins (MAVS), attenuates retinoic acid-inducible gene I-like receptors (RLRs) signal transduction and downstream type I interferons (IFNs) production, and induces failure of immune surveillance (128). Unfortunately, this mechanism has not been characterized in TNBC cell lines, but it is a potential research direction. Furthermore, the acidic microenvironment due to lactate accumulation significantly down-regulates macrophage-mediated Programmed Cell Removal (PrCR) damage to MDA-MB-231 cells (129).

Indeed, the Warburg phenotype derived from TNBC can also evade immune surveillance by activating immunosuppressive cells. For example, tumor glycolysis regulates the molecular network of AMPK-ULK1, autophagy, and CEBPB pathways. This way, myeloid-derived suppressor cells (MDSC) are activated, and immune suppression is maintained (130). Moreover, Microenvironmental lactic acidosis induces tumor-associated macrophages (TAM) to skew toward an M2-like phenotype, resulting in the obstacle of immuno-detection of tumor cells by macrophages (129, 131). A recent study found that TNBC CTCs recruit neutrophils to form CTC-neutrophil clusters, leading to an accelerated metastatic process (132). However, further studies are needed to determine whether the Warburg phenotype is involved in neutrophil recruitment and whether CTC-neutrophil clusters contribute to immune evasion (133).

Currently, chemotherapy based on anthracyclines and taxanes is the standard treatment for TNBC. The effectiveness of platinum compounds in neoadjuvant chemotherapy has been confirmed by evidence. Research on drugs related to targeted therapy, immunotherapy, and endocrine therapy has also collected initial results (134, 135). Nonetheless, multi-drug resistance deepens the conflict between TNBC therapy options and adverse outcomes (136). Studies have indicated that primary resistance genes pre-exist and are adaptively selected by treatment regimens. Acquired resistance gene transcriptional profiles are obtained by reprogramming in response to treatment in TNBC patients (137). Various mechanisms mediate drug resistance in TNBC (138), including the Warburg phenotype (25, 139, 140). First, the Warburg effect confers enhanced DNA repair capacity to tumor cells. Rac1 enhances nucleotide production to defend MDA-MB-231 cells from chemotherapeutic drug-induced DNA damage by triggering the aldolase A and ERK signaling cascades and upregulating aerobic glycolysis, especially its branching Pentose phosphate pathway (PPP) fluxes (141). A recent study demonstrating that overactivation of the ERK pathway promotes apoptosis in TNBC cells shows the dual role of ERK in TNBC biology, which should therefore be considered when selecting ERK as a therapeutic target (142). Second, the Warburg effect gives tumor cells a more robust antioxidant capacity. ROS up-regulation induced by chemotherapy drugs exposed tumor cells to oxidative stress, oxidative damage provoked the overexpression of the PPP rate-limiting enzyme glucose-6-phosphate dehydrogenase (G6PD), accelerated PPP enhanced the production of nicotinamide adenine dinucleotide phosphate (NADPH), thereby up-regulating the synthesis rate of the antioxidant glutathione (GSH). Ultimately, the oxidative damage caused by ROS can be inhibited (143). In addition, autophagy’s clearance of injured organelles and macromolecules is essential for inhibiting drug-induced apoptosis in TNBCs (144, 145). A recent study identified Proviral Insertion in Murine Lymphomas 2 (PIM2) directly binding to hexokinase-II (HK-2) and phosphorylating HK2 on Thr473. Phosphorylated HK2 up-regulates glycolytic flux and induces autophagy to exert chemoresistance by promoting protein stability (146). The Warburg effect also manifests resistance by inhibiting drug influx and promoting drug efflux. Glycolysis generates a large amount of H⁺ accumulation in the cytoplasm, and if it is not removed in time, the consequent acidosis threatens cell survival. The existence of a net acid extrusion system ignores such events. It leads to the rise of intracellular pH (pHi) and the reduction of extracellular pH (pHe) simultaneously. This pH gradient has been confirmed to be conducive to the formation of drug resistance (147). Elevated pHi not only inhibits the influx of alkaline drugs such as doxorubicin but also induces active drug efflux by coordinating the transporter ABCB1 (148). In addition, ABC transporters are ATP-dependent (149), and suppressing glycolysis leads to transporter inactivation (150), so it is reasonable to speculate that glycolysis provides energy support for the actions of ABC transporters. Based on the above theory, Omran Z et al. recommend that the intervention of proton pump and transporter inhibitors is vital to manage resistance (151).

Notably, with the intervention of compounds other than chemotherapy, the resistance spectrum of TNBC expands, such as lactate modulating anti-PD-1/PD-L1 drug activity (152) via activating TAMs (153). In summary, targeting the Warburg effect seems perfect for curbing resistance; however, this is not the case; silencing of GLUT1 does not ablate TNBC resistance as expected but further induces drug resistance via Akt/GSK-3β/β-catenin/survivin signaling (154). This opposite phenomenon may involve dynamic remodeling of metabolism (155) since switching metabolism to mitochondrial or lipid metabolism equally mediates resistance, depending on the molecular phenotype of the interventional drug (156, 157). Hence, careful considerations should be taken into account in sequential drug resistance control programs.

In the early stages of TNBC, hypoxia gives birth to the Warburg phenotype, and mutations or epigenetic modifications result in constitutive expression of the Warburg effect. However, the pressure on TNBC’s survival is ever-changing, necessitating that it maintain a varied metabolic pattern (15). Therefore, glycolysis inhibitors are only effective against a subset of TNBCs (34). Hence, it is necessary to identify Warburg-dependent TNBC and administer tailored treatment for it. With the advent of omics approaches, several teams have stratified TNBC based on various omics analyses and obtained experimental findings (34, 158–161) ( Table 1 ). However, the road is always tortuous; stratification schemes relying on genomics and transcriptomics failed to break through the barrier of LAR and BLIS tumors in later FUTURE experiments (162). Of note, Xiao Y et al. integrated metabolomics and genomics to classify TNBC into three subtypes: C1, with ceramide and fatty acid enrichment; C2, with overexpression of oxidative processes and glycosyl transfer-related metabolites; and C3, with low-level metabolic dysregulation, and proposed that genomics-based LAR subtypes overlap with C1 subtypes and BLIS subtypes overlap with C2 and C3 subtypes, finding a new solution for refractory TNBC (158, 163). Moreover, by investigating metabolic dysregulation in TNBC, Gong Y et al. identified the lipogenic subtype MPS1 with up-regulated lipid metabolism, the glycolytic subtype MPS2 with up-regulated carbohydrate and nucleotide metabolism, and the mixed subtype MPS3 with partially dysregulated pathways, each of MPS showed consistent sensitivity to inhibitors targeting the same metabolic pathway (34). Accordingly, stratified management of TNBC patients is crucial for minimizing inefficient medical treatment and reducing the cancer burden among patients (7).

Over the past few decades, significant progress has been made in studying compounds targeting glycolysis ( Table 2 ). Certain compounds target upstream regulators of glycolysis. For example, Cardamonin and Honokiol restricted glycolysis in MDA-MB-231 cells through the HIF-1α pathway (174, 175). Betulinic Acid inhibits glycolytic flux via GRP78/β-catenin/c-Myc or CAV-1/NF-κB/c-Myc signaling pathways (176, 177). Furthermore, some compounds target the critical regulators of the Warburg phenotype. Examples include GNE-140 and FX-11, which down-regulate glycolysis by blocking lactate dehydrogenase (LDH) (34). Elemene (β-elemene) suppresses breast cancer metastasis by obstructing pyruvate kinase M2 (PKM2) dimerization and nuclear translocation.

In addition, based on understanding the effect of ECM remodeling on TNBC invasive behavior, targeting ECM is becoming an emerging fulcrum for TNBC treatment protocols. For example, the TME-tunable BAGM therapeutic nanoplatform activates a cyclic cascade that degrades the dense ECM and converts matrix collagen into a loose state to exert antitumor effects (178). LOX inhibitor chemically linked lipid-based nanoparticles loaded with chemotherapeutic epirubicin have been demonstrated to prolong survival in patients with TNBC (179). This is probably because LOX inhibitors ablate the dense ECM and enhance chemotherapy drug permeability (62). Moreover, it has been demonstrated that inhibiting tumor development by targeting ECM to impact tumor cell metabolism is achievable (69, 180); the exact application of this technique to TNBC requires additional investigation.

The Warburg phenotype also functions as a strategy for eliminating transformed cells. RasV12-, Src-, or Erbb2-transformed cells are in a relatively rigid microenvironment formed by compacted epithelial cells, which are frequently apically extruded from epithelial tissues in search of more survival space, after which the epithelial defense against cancer (EDAC) induces epithelial protein lost in neoplasm (EPLIN)-mediated metabolic transformation (OXPHOS to glycolysis) to remove dissociative transformed cells (41, 181). This mechanism must be tested further to determine if it can be used to prevent the development of TNBC.

Worthy of note is that glycolysis suppression possibly triggers the up-regulation of compensatory metabolic pathways, such as OXPHOS (21), and drugs that simultaneously target both glycolysis and OXPHOS pathways, marizomib, and gracillin, have been successful in the mouse model (166, 182). It is necessary to clarify that while the inhibitory effect of targeted aerobic glycolysis on TNBC is undeniable, metabolic plasticity poses a new and pressing challenge for this therapeutic option (183). Perhaps targeting the metabolic switch is the key to tackling metabolic plasticity in the future (184).