Abstract

   Bioenergetic therapy is emerging as a promising therapeutic approach.
   However, its therapeutic effectiveness is restricted by metabolic
   plasticity, as tumor cells switch metabolic phenotypes between
   glycolysis and oxidative phosphorylation (OXPHOS) to compensate for
   energy. Herein, Metformin (MET) and BAY‐876 (BAY) co‐loaded CuFe[2]O[4]
   (CF) nanoplatform (CFMB) is developed to boost energy deprivation by
   synchronous interventions of glycolysis and OXPHOS for bioenergetic
   therapy synergetic with chemodynamic/photothermal therapy (CDT/PTT).
   The MET can simultaneously restrain glycolysis and OXPHOS by inhibiting
   hexokinase 2 (HK2) activity and damaging mitochondrial function to
   deprive energy, respectively. Besides, BAY blocks glucose uptake by
   inhibiting glucose transporter 1 (GLUT1) expression, further
   potentiating the glycolysis repression and thus achieving much more
   depletion of tumorigenic energy sources. Interestingly, the upregulated
   antioxidant glutathione (GSH) in cancer cells triggers CFMB degradation
   to release Cu^+/Fe^2+ catalyzing tumor‐overexpressed H[2]O[2] to
   hydroxyl radical (∙OH), both impairing OXPHOS and achieving
   GSH‐depletion amplified CDT. Furthermore, upon near‐infrared (NIR)
   light irradiation, CFMB has a photothermal conversion capacity to kill
   cancer cells for PTT and improve ∙OH production for enhanced CDT. In
   vivo experiments have manifested that CFMB remarkably suppressed tumor
   growth in mice without systemic toxicity. This study provides a new
   therapeutic modality paradigm to boost bioenergetic‐related therapies.

   Keywords: bioenergetic therapy; chemodynamic therapy; CuFe[2]O[4,]
   glycolysis; photothermal therapy
     __________________________________________________________________

   Metformin and BAY‐876 co‐loaded CuFe[2]O[4] (CFMB) nanoplatform is
   developed to synchronously disturb glycolysis and oxidative
   phosphorylation for boosted bioenergetic therapy synergetic with
   chemodynamic/photothermal therapy. CFMB manifests potent in vivo tumor
   growth inhibition with good biosafety.

   graphic file with name ADVS-11-2401738-g005.jpg

1. Introduction

   Energy is the basis of cell life. To meet the highly bioenergetic
   demands, indefinitely proliferative tumor cells obtain large amounts of
   energy through glucose metabolism to produce adenosine triphosphate
   (ATP).^[ [38]^1 ^] Glycolysis, also known as the Warburg effect, is the
   predominant energy production pathway used by most tumors.^[ [39]^2 ^]
   It is reported that cancer cells have at least 30‐fold and tenfold
   enhancement in glycolysis rate and glucose consumption respectively in
   comparison with normal cells.^[ [40]^3 ^] Particularly, cancer cells
   produce more than 50 times more ATP through glycolysis than normal
   cells.^[ [41]^4 ^] This feature facilitates establishing creative
   approaches to targeted glycolysis for tumor treatment. In addition,
   oxidative phosphorylation (OXPHOS), another pathway for cancer cells to
   obtain energy, also proceeds at a high rate.^[ [42]^5 ^] The OXPHOS
   metabolic pathway generates ATP by the electron transport chain (ETC)
   of mitochondria.^[ [43]^6 ^] Because of metabolic heterogeneity and
   compensation, tumor cells switch metabolic phenotypes between
   glycolysis and oxidative OXPHOS to compensate for energy.^[ [44]^7 ^]
   In this context, simultaneously intervening glycolysis as well as
   OXPHOS via cooperation mechanisms are prospective means to improve
   antitumor bioenergetic therapy.

   So far, many strategies have been developed to restrain glycolysis to
   reduce the energy supply of cancer cells, which can be generally
   categorized into two types: 1) inhibiting glycolytic key enzymes,^[
   [45]^8 ^] 2) decreasing the intracellular glucose level.^[ [46]^9 ^]
   Targeting key enzymes at different stages of glycolysis is a promising
   approach. For example, suppression of hexokinase 2 (HK2) that catalyzes
   glucose to glucose‐6‐phosphate (G6P) in the first step of glycolysis
   would lead to failure of glucose phosphorylation,^[ [47]^3 , [48]^8 ^]
   decreased ATP production, and energy deprivation. Although promising,
   inhibition of enzyme activity alone does not achieve satisfactory
   energy‐deprivation effects due to intracellular glucose accumulation
   caused by blocked glucose metabolic pathways. The other strategy
   restraining glycolysis can be recognized as glucose starvation. A
   widely used scavenger for glucose is glucose oxidase (GOX) which can
   convert glucose into hydrogen peroxide (H[2]O[2]), which has gained
   significant attention owing to its outstanding anticancer ability via
   both inhibiting energy metabolism and killing cancer cells by
   H[2]O[2].^[ [49]^9a,b ^] Unfortunately, the GOX‐mediated
   glucose‐oxidation reaction heavily depends on oxygen (O[2])
   concentration, thus the therapeutic efficiency is compromised by an
   insufficient O[2] supply of tumors. In contrast, inhibiting glucose
   transporters to block glucose conveyance into tumor cells may be a
   better alternative. Glucose transporter 1 (GLUT1), a classic example of
   a glucose transporter, is universally overexpressed in cancer cells.^[
   [50]^9b ^] Targeting GLUT1 to block glucose transport has proven
   effective in both preclinical and clinical studies.^[ [51]^9 , [52]^11
   ^] Therefore, dual‐pathway glycolysis inhibition by blocking upstream
   glucose supply while inhibiting downstream key enzyme activity might be
   a powerful and promising strategy for boosting energy deprivation of
   tumor cells.

   As for the other OXPHOS energy acquisition pathway, the intervention
   strategy is mainly to induce mitochondrial function and structural
   damage,^[ [53]^12 ^] such as atovaquone (ATO) inhibiting mitochondrial
   respiration,^[ [54]^13 ^] Ca^2+ inducing mitochondrial
   Ca^2+‐overload,^[ [55]^14 ^] reactive oxygen species (ROS) reducing
   complex catalytic activity,^[ [56]^15 ^] H[2]S interrupting the ETC,^[
   [57]^16 ^] and so on.^[ [58]^17 ^] OXPHOS inhibition has been
   demonstrated to be effective in reducing energy supply.^[ [59]^18 ^]
   Recently, poly(lactic‐co‐glycolic acid) co‐encapsulated
   pyropheophorbide and syrosingopine nanoparticles (mTPPa‐Sy NPs),^[
   [60]^19 ^] aptamer‐based artificial enzyme,^[ [61]^20 ^] 3PO‐loaded
   cross‐linked sodium lipoic acid vesicle (3PO@cLANa),^[ [62]^21 ^] and
   so on^[ [63]^20 , [64]^22 ^] have been developed for targeting
   glycolysis and OXPHOS to regulate tumor metabolism, highlighting the
   effectiveness and necessity of synergetic energy deprivation. In these
   studies, the nano drugs interfere with glycolysis and OXPHOS through
   the single‐pathway action of related inhibitors and thus kill cancer
   cells only depending on the energy block. According to the concept of
   synthetic lethality,^[ [65]^23 ^] the single‐pathway suppression in
   cancer metabolism makes the other pathway essential for cell
   survival.^[ [66]^19 ^] In addition, given the heterogeneity and
   complexity of the tumor,^[ [67]^24 ^] energy deprivation alone might
   not be sufficient to eradicate the tumor. Chemodynamic therapy (CDT)
   catalyzes the decomposition of tumor‐overexpressed H[2]O[2] into
   hydroxyl radicals (∙OH), which can not only impair OXPHOS but also kill
   cancer cells directly.^[ [68]^25 ^] More importantly, photothermal
   therapy (PTT) benefits in promoting CDT catalytic reaction.^[ [69]^26
   ^] Based on these concerns, we hypothesized that synchronous
   interventions of glycolysis and OXPHOS through dual pathways, as well
   as introduction of CDT or/and PTT, would not only be lethal to energy
   deprivation but also potentiate tumor eradication.

   Herein, a versatile nano platform CuFe[2]O[4]@MET@BAY (CFMB) based on
   metformin (MET) and BAY‐876 (BAY) co‐loaded CuFe[2]O[4] (CF) was
   constructed to boost energy deprivation by synchronous interventions of
   glycolysis and OXPHOS for antitumor bioenergetic/CDT/PTT therapy.
   Scheme [70]1 illustrates that CFMB possesses multiple merits: 1) CFMB
   can efficiently deliver drugs to the tumor area, where it is
   subsequently triggered by tumor‐overexpressed antioxidant glutathione
   (GSH) to degrade and release MET and BAY. 2) MET could synchronously
   suppress glycolysis and OXPHOS by inhibiting HK2 activity and damaging
   mitochondrial function to deprive energy, respectively. 3) BAY blocks
   glucose uptake by effectively inhibiting GLUT1 expression, further
   potentiating the glycolysis repression and thus achieving much more
   depletion of tumorigenic energy sources. 4) GSH‐triggered CFMB
   degradation releases Fenton‐like/Fenton reagent Cu^+/Fe^2+ ions to
   catalyze tumor‐overexpressed H[2]O[2] to ∙OH, both impairing OXPHOS and
   achieving GSH‐depletion amplified CDT. (5) Upon 808 nm light
   irradiation, CFMB has a photothermal conversion capacity to kill cancer
   cells for PTT and improve ∙OH production for enhanced CDT. By combining
   with the above characteristics, CFMB achieves potent antitumor
   efficacies by boosting bioenergetic therapy and CDT/PTT in a mutually
   reinforcing way, providing a novel approach to designing
   bioenergetic‐based synergistic cancer treatment.

Scheme 1.

   Scheme 1
   [71]Open in a new tab

   Scheme illustrating the synthesis of CFMB and its therapeutic mechanism
   for boosting energy deprivation by synchronous interventions of
   glycolysis and OXPHOS for bioenergetic therapy synergetic with CDT/PTT.

2. Results and Discussion

2.1. Synthesis and Characterization of CFMB NPs

   As shown in Scheme [72]1, CF NPs were successfully prepared using the
   one‐pot solvothermal method. The scanning electron microscopy (SEM)
   image displays that the as‐prepared CF NPs exhibited 3D flower‐like
   spherical structures (Figure [73]1a). The detailed size distribution
   based on SEM images shows that CF NPs have an average diameter of
   200 nm (Figure [74]S1, Supporting Information). As can be seen from the
   transmission electron microscopy (TEM) image (Figure [75]1b), CF NPs
   show monodispersity and no agglomeration could be observed. A TEM image
   of individual CF nanoparticles in Figure [76]1c more clearly reveals
   the flower‐like CF was assembled from numerous nanosheets, which would
   provide lots of chemical reaction sites for catalytic reactions. X‐ray
   diffraction (XRD) analyses indicate the characteristic peaks in
   as‐prepared CF NPs could be assigned to the standard pattern (JCPDS NO.
   25–0283) of CuFe[2]O[4] (Figure [77]S2, Supporting Information). The
   high‐resolution TEM (HRTEM) of CF displays distinct lattice fringes of
   0.2956 nm (Figure [78]S3, Supporting Information), which agrees with
   the lattice spacing of (220) plane of CuFe[2]O[4], proving the spinel
   structure. Energy‐dispersive X‐ray spectroscopy (EDS) (Figure [79]1d)
   of CF demonstrates the existence of Cu, Fe, and O in these NPs and
   reveals the composition (Figure [80]S4, Supporting Information). The
   high‐resolution X‐ray photoelectron spectroscopy (XPS) analysis of Fe
   2p displayed the characteristic peaks at 710.5 and 724.7 eV belonging
   to Fe^3+, demonstrating that Fe^3+ is the only oxidation state present
   in CF NPs (Figure [81]1e). In the Cu 2p XPS spectrum, two signal peaks
   appeared at 933.9 and 953.5 eV with the satellite signals at 941.3 and
   943.8 eV could be ascribed to Cu^2+ (Figure [82]1f). The porosity of
   the CF was studied by N[2] adsorption isotherm analysis as shown in
   Figure [83]1g. The adsorption hysteresis phenomenon demonstrates the
   porous structure of CF, and its BET surface area and pore size were
   determined to be 26.7 m^2 g^−1 and 26 nm respectively, which is more
   conducive to efficient drug loading. With the advantages of mesoporous
   structure, MET and BAY drug molecules were loaded and stored in the
   pores of CF by a simple mixing method. The Zeta potential decreases
   from −13.7 to −15.8 mV after BAY and MET are loaded due to their
   negative potentials (Figure [84]1h). The new absorption band at 230 nm
   corresponds to the characteristic peaks of MET, and the absorption at
   350 nm is attributed to the characteristic peak redshift of BAY
   (Figure [85]1i). Although the absorption peak of MET in the UV region
   overlaps with that of BAY, the loading content of MET and BAY
   calculated by standard calibration curves is 4.13% and 1.63%
   respectively (Figure [86]S5, Supporting Information), further
   confirming that MET has been successfully loaded in CF NPs. The element
   mapping of CFMB illustrates the homogeneous distribution of Cu, Fe, and
   F elements among the CFMB framework (Figure [87]S6, Supporting
   Information), further proving the successful loading of BAY. The
   stability of the CFMB in H[2]O, phosphate buffer solutions (PBS), or
   Dulbecco's modified eagle medium (DMEM) was investigated by dynamic
   light scattering (DLS) (Figure [88]S7, Supporting Information), where
   CFMB shows negligible hydrodynamic diameter change upon 7 days, proving
   its high stability.

Figure 1.

   Figure 1
   [89]Open in a new tab

   a) SEM image CF NPs. b,c) TEM images of CF NPs. d) EDS of CF NPs.
   High‐resolution e) Cu 2p, and f) Fe 2p XPS spectra. g) N[2]
   adsorption–desorption isotherm of CF. (Inset: pore‐size distribution of
   CF.) h) Zeta potentials, and i) UV–vis spectra of CF, MET, BAY, and
   CFMB.

2.2. Photothermal Performance

   As displayed in Figure [90]2a, CF NPs have a wide absorption band in
   the near‐infrared (NIR) region, which is concentration‐dependent. The
   absorbance value in Figure [91]S8 (Supporting Information) increases
   linearly as NP concentration increases. Meanwhile, the color of CF
   solutions gradually deepens, and all show good dispersion and light
   transmission (Figure [92]2a, inset). After irradiation with 808 nm
   light for 7 min, the temperature of CF NP aqueous dispersion increased
   by 7.8, 11.7, 14, 17.5, and 22.5 °C at concentrations of 32.5, 75, 150,
   200, and 300 µg mL^−1, respectively (Figure [93]2b). By contrast, the
   temperature of pure water rises by only 3.5 °C under the same
   condition. The above result suggests that CF could be a potential PTT
   reagent with excellent photothermal conversion ability. Moreover, its
   photothermal performance is laser power density‐dependent
   (Figure [94]2c). After four heating/cooling cycles for CF NPs, the
   temperature change remains stable (Figure [95]2d), demonstrating the
   excellent photostability of CF NPs. Its photothermal conversion
   efficiency (η) is as high as 35.2% (Figure [96]2e; Figure [97]S9,
   Supporting Information). The negligible change could be found in its
   UV–vis–NIR absorbance spectrum after laser irradiation, further
   demonstrating the photostability of CF NPs (Figure [98]2f). The
   excellent η value and remarkable photostability make CF a promising PTT
   reagent.

Figure 2.

   Figure 2
   [99]Open in a new tab

   a) UV–vis–NIR absorption spectra of CF aqueous dispersion at various
   concentrations. b,c) Temperature curves of CF aqueous dispersion
   exposure to 808 nm light. d) The temperature curve of CF suffered four
   light on/off cycles. e) Heating‐cooling curves of CF under 808 nm
   irradiation as well as photothermal conversion efficiency. f)
   UV–vis–NIR absorbance spectra of CF before and after 808 nm
   irradiation.

2.3. GSH‐Triggered Drugs Release and Catalytic Reactions

   As shown in Scheme [100]1, the intervention of tumor metabolism is
   initiated by GSH‐mediated degradation of CFMB NPs, and the structure
   collapsing and degradation of CFMB NPs would lead to the release of
   loaded MET and BAY, as well as a series of catalytic reactions. The
   biodegradation behavior of CFMB in different environments is directly
   demonstrated by TEM images. Phosphate buffer solutions (PBS) were used
   to simulate complex physiological environments. When immersed in PBS
   with pH 7.4 or 6.0, the structure of CFMB NPs remained intact for 24 h
   (Figure [101]3a), suggesting CFMB could keep the integrity of its
   original structure during blood circulation or at weakly acidic
   conditions. However, in the presence of GSH, the structure of CFMB
   completely collapsed at both pH 7.4 and 6.0 conditions, indicating
   GSH‐triggered degradation. The degradation of CFMB would facilitate the
   loaded BAY and MET release. Thus, drug release properties are
   subsequently studied. As shown in Figure [102]3b, ≈25% of MET was
   released at pH 7.4, 6.0, and 5.0 conditions, and once GSH was added,
   the drug release increased rapidly, and ≈90% MET was released within
   24 h. Similarly, the cumulative BAY release in the presence of GSH is
   ≈4 times higher than that without GSH (Figure [103]3c). The release of
   MET and BAY at pH 6.0 condition increased with the elevation of GSH
   concentration (Figure [104]S10, Supporting Information). The above
   result proves that CFMB possesses excellent GSH‐triggered degradation
   performance for subsequent drug release at the tumor site with GSH
   overexpression, which makes it a promising drug delivery system for
   biomedical applications. In addition, the CF NPs still retained a
   photothermal effect after descending into the ionic (Figure [105]S11,
   Supporting Information).

Figure 3.

   Figure 3
   [106]Open in a new tab

   a) TEM images of CFMB NPs subjected to different treatments for 24 h.
   The release profile of b) MET and c) BAY at various conditions.
   High‐resolution d) Cu 2p, and e) Fe 2p XPS spectra of CF NPs treated by
   GSH. f) UV–vis spectra of DTNB treated with different concentrations of
   GSH in the presence of CF. g) UV–vis spectra of TMB after different
   treatments. h) UV–vis spectra of TMB treated by diverse GSH
   concentrations. i) UV–vis spectra of TMB at 25, 48 °C, and NIR light
   irradiation.

   Importantly, the XPS survey of CF NPs after reacting with GSH was
   conducted to further reveal the transformation of Cu^+/Cu^2+ and
   Fe^2+/Fe^3+ redox couple. High‐resolution Cu 2p XPS spectrum shows the
   coexistence of Cu^+ and Cu^2+ in CF NPs (Figure [107]3d), where the
   fitting peak intensity of Cu^2+ at 933.9 eV decreases significantly,
   and the characteristic peak at 932.4 eV assigned to the emerged Cu^+.
   Similarly, the fitting peaks of 709.9 eV attribute to the emerged Fe^2+
   in the Fe 2p XPS spectrum, confirming the success of the
   Fe^2+/Fe^3+redox couple conversion. According to XPS results, the
   content of Cu^+ increases from 0% to 30%, while the proportion of Fe^2+
   increases to 31.7% after the reaction of CF with GSH. Furthermore, the
   consumption of GSH was assessed by 5,5′‐Dithiobis‐(2‐nitro benzoic
   acid) (DTNB). The product of DTNB reduced by GSH has a characteristic
   absorption peak at 412 nm.^[ [108]^27 ^] As displayed in
   Figure [109]3f, the absorption value at 412 nm declines dramatically as
   CF concentration increases, demonstrating the capability of CF to
   deplete GSH is concentration‐dependent. The above results prove that
   Cu^2+ and Fe^3+ in CF react with GSH to produce new Cu^+ and Fe^2+
   ions, which may enable CF NPs to act as Fenton‐like/Fenton reagents to
   transform tumor‐overexpressed H[2]O[2] into toxic ∙OH for CDT
   applications.

   Subsequently, the ability of CF to produce ∙OH through a
   Fenton‐like/Fenton reaction was investigated by well‐developed
   3,3,5,5‐tetramethylbenzidine (TMB) oxidation experiments in H[2]O[2]
   dispersion of CF NPs with or without GSH. The product of TMB oxidized
   by ∙OH has a 652 nm absorption peak.^[ [110]^15 ^] The 652 nm peak
   shows a negligible change even after CF NPs have reacted with H[2]O[2]
   for 30 min (Figure [111]3g), indicating no ∙OH produced. However, after
   adding GSH, the peak value at 652 nm continuously increases with the
   extension of time, suggesting the reaction of CF and GSH successfully
   occurs to generate Fenton‐like/Fenton reagent Cu^+/Fe^2+ ions which
   catalyzes H[2]O[2] to ∙OH. At the fixed concentrations of CF and
   H[2]O[2], the characteristic absorbance at 652 nm increases with
   increasing GSH concentration (1–10 mm, Figure [112]3h), while the GSH
   amount further increases (15 mm), the 652 nm absorbance decreases
   significantly, which could be attributed to the excess GSH in turn
   scavenging the generated ∙OH. In addition, the promoted ∙OH generations
   by the photothermal effect have been investigated as well. The ∙OH
   yield of CF NPs at 50 °C is significantly higher than that at 25 °C
   (Figure [113]3i), implying the increased temperature can promote the
   catalytic reaction. Similarly, when CF NPs are exposed to 808 nm light,
   the ability of CF to produce ∙OH is also higher than that at 25 °C,
   confirming the boosted effect of photothermal effect on the
   Fenton‐like/Fenton reaction of CF NPs. To verify this, the steady‐state
   catalytic kinetics experiments of CF nanoparticles at 25 and 50 °C were
   conducted by changing H[2]O[2] concentrations at fixed CF nanoparticles
   and TMB concentrations, followed by measuring the generated hydroxyl
   radicals. The Michaelis–Menten constant (K [m] ) of CF nanoparticles
   was calculated to be 6.53 and 5.21 mm at 25 and 50 °C by fitting the
   kinetic data with the Michaelis–Menten equation (Figure [114]S12a,
   Supporting Information) and Lineweaver double‐reciprocal plot (Figure
   [115]S12b, Supporting Information), demonstrating the improved
   catalytic activity of CF nanoparticles at higher temperatures. All
   these results verify that the as‐prepared CF could be used as a
   GSH‐triggered Fenton‐like/Fenton reagent for amplified CDT through GSH
   consumption and photothermal effect.

2.4. In Vitro Anticancer Effect of CFMB and Related Mechanism Exploration

   The internalization of nanodrugs in cells is a prerequisite for
   therapeutic drugs to exert therapeutic effects. Therefore, cellular
   uptake of CFMB was first assessed by confocal laser scanning microscopy
   (CLSM). To be specific, 4T1 cancer cells were co‐cultured with
   fluorescein isothiocyanate (FITC) labeled CF (denoted as FITC‐CF) NPs
   and then observed using CLSM. As seen from Figure [116]S13a (Supporting
   Information), obvious green fluorescence is emitted from 4T1 cells, and
   the fluorescence reaches the strongest after the NPs were cultured with
   cells for 6 h, suggesting FITC‐CF NPs could be effectively endocytosed
   by 4T1 cells. The results of the gray value analysis further confirm
   that the amount of FITC‐CF NP endocytosis is time‐dependent before 6 h
   (Figure [117]S13b, Supporting Information). The biocompatibility of NPs
   was also evaluated before in vitro anti‐tumor experiments. HUVEC cells
   were selected in normal cells mode and cultured with various NPs (CF,
   CFM, CFB, CFMB) at the concentration of 200 µg mL^−1. CCK‐8 assays
   reveal the survival rate of HUVEC cells was no ˂85% after incubation
   for 24 h (Figure [118]S14, Supporting Information), indicating that all
   the NPs have relatively low cell toxicity. Next, the therapeutic effect
   of CFMB in vitro was assessed. After various treatments, 4T1 cancer
   cells were stained using calcein‐AM and propidium iodide (AM/PI) to
   mark the dead and living cells respectively, and then cell survival was
   observed with CLSM. As exhibited in Figure [119]4a, the NIR group
   exhibits hardly cell damage, indicating NIR light alone does not cause
   cytotoxicity. A degree of cell death is found in the CF + NIR group
   because of synergistic killing effects caused by CF itself as CDT and
   PTT reagent. Notably, as compared with the group of CFM and CFB, the
   CFMB group shows a larger proportion of dead cells, indicating that the
   combined MET and BAY synergistic intervention of energy metabolism has
   a better killing effect on cancer cells. Interestingly, it is found
   that almost all cells in the CFMB + NIR group die, confirming that the
   photothermal effect can remarkably potentiate bioenergetic therapy and
   CDT. In addition, the CCK‐8 assay showed similar antitumor results as
   the AM/PI double staining assay (Figure [120]S15, Supporting
   Information). All the above data indicates the remarkable therapeutic
   efficiency of CFMB against tumor cells.

Figure 4.

   Figure 4
   [121]Open in a new tab

   a) Fluorescence microscope images of AM/PI stained 4T1 cells after
   various treatments. Extracellular lactate b), glucose c), and
   intracellular ATP d) content of 4T1 cells subjected to diverse
   treatments for 12 and 24 h. e) Western blot analysis on the expressions
   of GLUT1 and HK2 in 4T1 cells. f) CLSM images of ∙OH detected by
   DCFH‐DA in 4T1 cells subjected to diverse treatments. g) Relative
   catalytic activity of complex I in 4T1 cells after different
   treatments. h) CLSM images of JC‐1 stained 4T1 cells subjected to
   diverse treatments.

   As shown in Scheme [122]1, GSH‐triggered CFMB degradation releases
   Cu^+/Fe^2+ irons, MET, and BAY. Cu^+/Fe^2+ ions‐mediated ∙OH generation
   impairs OXPHOS to decrease ATP generation. MET synchronously suppresses
   glycolysis and OXPHOS by inhibiting HK2 activity and damaging
   mitochondrial function to decrease the production of ATP and lactate.
   In addition, BAY blocks glucose transport into tumor cells by
   inhibiting GLUT1 to further potentiate glycolysis repression. Based on
   the above mechanism, the metabolism regulation of CFMB was first
   investigated by assessing the variation in lactate, glucose, and ATP at
   the cellular level. As exhibited in Figure [123]4b, in contrast to the
   control group, the extracellular lactic acid levels in the CFM and CFB
   groups decline significantly, indicating effective inhibition of
   glycolysis. The lowest lactate level is observed in the CFMB group,
   which is 37% and 23% at 12 and 24 h respectively, demonstrating the
   synergistic effect of MET and BAY on glycolytic inhibition. Similarly,
   the highest extracellular glucose level is detected in CFMB‐treated
   cells (Figure [124]4c). Intracellular ATP level was also determined
   after distinct treatments. As displayed in Figure [125]4d, the CF
   induces a slight ATP decrease, while the CFM or CFB group has a larger
   decline in ATP, suggesting that impairing OXPHOS by ∙OH, inhibiting
   glycolysis and OXPHOS by MET and ∙OH, or blocking glucose uptake alone
   would aggravate energy deprivation. Notably, the ATP level of the cells
   treated with CFMB for 24 h decreased significantly, and only 14.1% of
   the ATP was produced, confirming that metabolic intervention by
   combining ∙OH, MET, and BAY has a synergistic effect on depriving tumor
   energy. To explore the molecular mechanism of CFMB NPs in metabolic
   intervention, western blotting analysis was conducted to evaluate the
   expression of proteins associated with glycolysis in 4T1 cells.
   Figure [126]4e proves that HK2 expression in CFM and CFMB groups is
   greatly lower than that in control and CFB groups, indicating MET could
   effectively induce the down‐regulation of HK2 and thus reduce the
   glycolytic activity. Besides, GLUT1 expression in the CFB and CFMB
   groups remarkably descends in comparison with that in the control
   group, evidencing restraining effects of BAY to GLUT1. The expressions
   of GLUT1 and HK2 in 4T1 cells were further quantitatively evaluated by
   the related gray values (Figure [127]S16, Supporting Information), and
   the results were basically consistent with those observed above. These
   results suggest that CFMB can reduce glycolytic activity by inhibiting
   the expression of HK2 and GLUT1.

   As depicted in Scheme [128]1, CFMB reacts with GSH to release
   Fenton‐like/Fenton reagent Cu^+/Fe^2+ ions that catalyze
   tumor‐overexpressed H[2]O[2] to ∙OH. Therefore, 2,7‐dichlorofluorescein
   diacetate (DCFH‐DA) chemical probe was then used to detect the
   production of intracellular ∙OH, which can emit green fluorescence
   after ∙OH oxidation. The change of green fluorescence signal in NIR
   light irradiated cells is negligible (Figure [129]4f), indicating NIR
   light alone does not affect ∙OH production. Cells treated with CF or
   CFMB show stronger green fluorescence, which is associated with ∙OH
   generated by Cu^+/Fe^2+ mediated Fenton‐like/Fenton reaction. Notably,
   the strongest green fluorescence appears in the CF + NIR or CFMB + NIR
   light‐treated 4T1 cells, which is attributed to enhanced ∙OH production
   due to the combined effect of PTT, proving the increased temperature is
   conducive to H[2]O[2] being catalyzed by Cu^+/Fe^2+ to ∙OH.

   It has been reported that ROS and MET can inactivate OXPHOS‐related
   enzymes such as complex I blocking mitochondrial energy supply.^[
   [130]^28 ^] Thus, we investigated whether CFMB‐mediated ∙OH production
   and MET release affect the activity of complex I and impaired OXPHOS.
   As shown in Figure [131]4g, CF NPs resulted in an 11.6% decline in
   complex I activity compared to the control group (≈100), while CFM and
   CFMB NPs showed significant complex I inactivation with activity values
   of 50.1% and 47.8%, respectively, demonstrating the generated ∙OH and
   released MET can impede OXPHOS. Additionally, the changes in
   mitochondrial membrane potential (MMP) were further detected by a JC‐1
   probe to verify the mitochondrial damage. JC‐1 can form a J‐aggregate
   that emits red fluorescence at high MMP while remaining a monomer that
   emits green fluorescence at low MMP.^[ [132]^29 ^] As shown in
   Figure [133]4h, moderate green fluorescence is observed in the CF or
   CFB group, demonstrating Cu^+/Fe^2+ mediated ∙OH production could
   induce mitochondria damage. The stronger green fluorescence is observed
   for the CMB or CFMB group, demonstrating the synergistic effect of
   Cu^+/Fe^2+ mediated ∙OH production and MET on MMP dissipation. The
   results strongly imply that CFMB can effectively induce mitochondrial
   damage, which is promising for energy deprivation by the OXPHOS
   pathway.

   To further explore the therapeutic mechanism of CFMB, genome‐wide
   RNA‐sequencing was performed on the 4T1 cells treated with CFMB. As
   displayed in Figure [134]5a, volcano plots present that 7551
   differentially expressed genes (DEGs) are detected after treatment with
   CFMB, with 4494 up‐regulated genes (red dots) and 3057 down‐regulated
   genes (blue dots). DEGs enrichment results imply that multiple
   signaling pathways associated with the proliferation and death of 4T1
   cancer cells are determined to be significantly changed after treatment
   of CFMB NPs, such as the apoptosis signaling pathway, the TNF signaling
   pathway, the p53 signaling pathway, the MAPK signaling pathway, the
   MTOR signaling pathway, cell cycle signaling pathway and so on
   (Figure [135]5b). These results indicate that these biological events
   may be attributable to effective anticancer treatments resulting from
   CFMB. It is worth noting that multiple tumor metabolism‐related gene
   signaling pathways are affected including the PI3K‐Akt signaling
   pathway, the AMPK signaling pathway glycolysis/gluconeogenesis
   signaling pathway, and so on. For example, the PI3K‐Akt signaling
   pathway plays key roles in the uptake and utilization of various
   nutrients involving glutamine, glucose, and nucleotides.^[ [136]^30 ^]
   Specifically, a number of DEGs significantly affect these pathways
   (Figure [137]5c,d). For example, down‐regulation of LDHA, LDHB, and
   PFKL genes appears to result in the inhibition of glycolysis,^[
   [138]^31 ^] the overexpression of positively related genes STK11 and
   MAP3K7 accounts for the activated AMPK signaling pathway.^[ [139]^32 ^]
   In addition, DEGs involved in mitochondrial damage are regulated in the
   CFMB‐treated cells. Both CHCHD6 and Ndufv1, which are associated with
   the integrity of the mitochondrial ridge structure and the
   mitochondrial complex I,^[ [140]^33 ^] are significantly
   down‐regulated, resulting in affected mitochondrial function and
   reduced ATP production. The data demonstrates that CFMB can
   simultaneously restrain glycolysis and damage mitochondrial function,
   thus depriving the energy supply of the tumor. Furthermore, the
   ferroptosis signaling pathway and down‐regulated GPX4 suggest that CFMB
   has anticancer effects on 4T1 cancer cells through GSH depletion and
   ROS elevation.^[ [141]^34 ^] Overall, all these results further reveal
   the anti‐tumor mechanism of CFMB is related to glycolysis and
   mitochondrial damage‐mediated energy metabolic intervention and
   GSH‐enhanced CDT.

Figure 5.

   Figure 5
   [142]Open in a new tab

   a) Volcano plot of DEGs for the control and CFMB group. b) KEGG pathway
   enrichment analysis after treatments. c) PPI network including these
   representative DEGs. d) The DEGs of interest between different groups
   were displayed by a heat map.

   All the above data verify the therapeutic mechanism illustrated in
   Scheme [143]1. Tumor‐overexpressed GSH triggers CFMB degradation to
   release Cu^+/Fe^2+ irons, MET, and BAY. Cu^+/Fe^2+ ions‐mediated ∙OH
   generation impairs OXPHOS by reducing complex I activity. The MET
   simultaneously restrains glycolysis and OXPHOS by inhibiting HK2 and
   damaging mitochondrial function, respectively. BAY blocks glucose
   uptake by inhibiting GLUT1 expression to further potentiate the
   glycolysis repression. Therefore, synchronous interventions of
   glycolysis (targets of HK2 and GLUT1) and OXPHOS (inhibitors of ∙OH and
   MET) through dual pathways were achieved to boost energy deprivation of
   tumor cells. Moreover, the inherent physicochemical properties of CF
   NPs enable amplified CDT with concurrent GSH depletion and photothermal
   effects. As a result, CFMB achieves potent antitumor efficacies by
   boosting bioenergetic therapy and CDT/PTT in a mutually reinforcing
   way.

2.5. In Vivo Antitumor Effects of CFMB NPs

   The prerequisite for the therapeutic effects of nanodrugs is their
   effective accumulation at the tumor site. Therefore, we first evaluated
   the biodistribution of CFMB NPs in tumor‐bearing mice before in vivo
   therapy experiments. The CFMB was administered to mice via the tail
   vein and then mice were sacrificed Cu level in major organs and tumor
   tissue was detected by ICP‐MS at scheduled time. As exhibited in Figure
   [144]6a, the CFMB distribution in tumors reaches its maximum 24 h after
   administration, demonstrating that CFMB might assemble in tumor tissue
   via enhanced permeability and retention effect (EPR) due to the
   applicable size. Meanwhile, it is worth noting that CFMB distribution
   in major organs shows a first increasing and then declining trend with
   time extension. As time extends to 7 days, CFMB NPs could be excreted
   from mice, manifesting the clearance process is time‐dependent, which
   is similar to the scavenging behavior of some reported nano‐drugs.

Figure 6.

   Figure 6
   [145]Open in a new tab

   a) Biodistribution of Cu in mice at day 1, 3, and 7 days after
   administration. b) The relative 4T1 tumor volume after different
   treatments. c) The average tumor weight in various groups. The relative
   d) lactate and e) ATP levels in the harvested tumors after different
   treatments. f) Western blot analysis on the expressions of GLUT1 and
   HK2 in the harvested tumors. g) Body weight variation of mice from
   different groups. h) Hematological parameters of mice after being
   treated with CFMB plus NIR light. i) H&E staining images of major
   organs after various treatments.

   Based on the efficient in vitro therapeutic capability and tumor
   accumulation of CFMB NPs, the in vivo antitumor effect of CFMB was
   further assessed in a 4T1 tumor xenograft mouse model. All
   tumor‐bearing mice were divided into 7 groups: PBS, NIR, CF + NIR, CFM,
   CFB, CFMB, and CFMB + NIR. During the whole treatment course, the
   corresponding nano‐drug was administered to mice via the tail vein once
   (10 mg kg^−1). For mice requiring laser treatment, the tumor sites were
   irradiated once 24 h after administration. Their body weights as well
   as tumor volumes were monitored every other day during treatment.
   Figure [146]6b proves that NIR light irradiation alone has no antitumor
   effect. CFMB NPs exhibit a better tumor suppression effect than either
   CFM or CFB NPs due to the much more tumor energy deprivation caused by
   the synergistic metabolic intervention of MET and BAY. Besides, CF +
   NIR light irradiation shows an almost identical inhibitory effect on
   tumor growth as CFMB, which is related to PTT and Cu^+/Fe^2+ mediated
   CDT. Notably, CFMB + NIR light exhibits the best effect on the
   suppression of tumor growth, revealing the positive role of PTT in
   synergistic bioenergetics therapy and CDT. At the end of treatment, the
   digital photography of the dissected representative tumor in each group
   basically corresponds with the result of tumor growth inhibition
   (Figure [147]S17, Supporting Information). The antitumor effect is
   further validated by measuring the average weight of mouse tumors
   dissected after treatment. The average tumor weight of CFMB + NIR
   light‐treated mice is the lowest (Figure [148]6c), agreeing with tumor
   growth suppression. In order to further explore the relevant mechanisms
   of CFMB NPs in metabolic intervention, the metabolic changes of tumors
   in vivo were investigated. As displayed in Figure [149]6d and
   Figure [150]6e, the lactic acid and ATP levels in the tumors dissected
   from mice treated with CFMB NPs were significantly decreased, which is
   basically consistent with the results of in vitro experiments. Western
   blotting analysis of HK2 and GLUT1 of dissected tumors was conducted
   (Figure [151]6f). Compared with the control group, both HK2 and GLUT1
   expressions were down‐regulated in the CFMB group, demonstrating the
   inhibitory effect of CFMB on glycolysis.

   The in vivo biocompatibility of CFMB NPs was systematically assessed.
   Firstly, during the whole treatment period, the mice in all
   experimental groups maintained a slight trend of weight gain, and no
   weight loss was observed, implying the adopted therapy did not induce
   serious systemic toxicity (Figure [152]6g). In addition, the blood
   samples of the mice 14 days after intravenous injection were analyzed.
   As shown in Figure [153]6h, the eight indexes representing liver
   function and three indexes representing renal function have no abnormal
   difference from the control group, further proving the high
   hemocompatibility, ignorable nephrotoxicity, and low hepatotoxicity of
   CFMB NPs. Furthermore, hematoxylin and eosin (H&E) staining of major
   organs including heart, liver, lung, spleen, and kidney presents
   negligible histological changes and normal cellular morphology
   (Figure [154]6i). All the above results demonstrate that the designed
   CFMB nano‐drug possesses decent in vivo biocompatibility and biosafety.

3. Conclusion

   In summary, we developed a CFMB nanoplatform to boost energy
   deprivation by synchronous interventions of glycolysis and OXPHOS for
   antitumor bioenergetic therapy synergetic with CDT/PTT. CFMB can
   efficiently deliver drugs to the tumor area, where it is subsequently
   triggered by tumor‐overexpressed antioxidant GSH to degrade and release
   MET and BAY. The released MET simultaneously restrains glycolysis and
   OXPHOS by inhibiting HK2 activity and damaging mitochondrial function
   to deprive energy, respectively. Meanwhile, BAY blocks glucose uptake
   by inhibiting GLUT1, further potentiating the glycolysis repression.
   Moreover, GSH‐triggered CFMB degradation releases Cu^+/Fe^2+ ions to
   catalyze tumor‐overexpressed H[2]O[2] to ∙OH, which not only achieves
   GSH‐depletion amplified CDT but also impairs OXPHOS. As a result, the
   multiple pathways of metabolism interference significantly boosts
   energy deprivation of tumorigenic energy sources. Furthermore, the
   photothermal effect of CFMB could both directly kill cancer cells and
   promote ∙OH generation. In vitro and in vivo experiments concurrently
   manifest that CFMB achieves potent antitumor efficacies against 4T1
   breast tumors without causing systemic toxicity. Therefore, this study
   might propose a valid strategy for boosting bioenergetic‐based cancer
   therapy via integrating multiple pathways of metabolic intervention and
   PTT/CDT therapy.

4. Experimental Section

Materials

   H[2]O[2] (30%), ethanol, metformin hydrochloride (MET),
   5,5′‐Dithiobis‐(2‐nitrobenzoicacid) (DTNB), Bay‐876 (BAY),
   3,3,5,5‐tetramethylbenzidine (TMB) and cupric chloride dihydrate
   (CuCl[2]∙2H[2]O) were ordered from Aladdin Reagent Co., Ltd. (China).
   Calcein acetoxymethyl ester (calcein‐AM), 2,7‐Dichlorofluorescein
   diacetate (DCFH‐DA), glutathione (GSH), propidium iodide (PI) and
   4′,6‐diamidino‐2‐phenylindole (DAPI) were ordered from Sigma‐Aldrich.
   (St. Louis, MO). Iron chloride hexahydrate (FeCl[3]∙6H[2]O), urea
   (CH[4]N[2]O), and polyacrylamide ([CH[2]CH(CONH[2])]n) were supplied by
   Macklin Biochemical Technology Co., Ltd. (China)., ATP assay kit and
   glucose assay kit were purchased from Beyotime Biotech. Inc. (China).
   Lactate assay kit bought from Nanjing Jiancheng Bioengineering
   Institute (China). GLUT1/SLC2A1 Rabbit pAb, Hexokinase II Rabbit pAb,
   and GAPDH Rabbit pAb were purchased from Abclonal Technology.

Characterizations

   The zeta potential and size of NPs were tested via dynamic light
   scattering (DLS, NanoZS90, Malvern). Their morphologies were
   characterized by transmission electron microscopy (TEM, Tecnai F20, US)
   and scanning electron microscopy (SEM, US‐70, Hitachi). Their crystal
   was recorded via an X‐ray diffractometer (XRD, BRUKERD8 Advance). The
   UV–vis–NIR absorption spectrum of NPs was recorded by UV–vis–NIR
   spectrophotometer (UV–vis, UV‐1300, Shimadzu). The valence state of the
   element was analyzed by X‐ray Photoelectron Spectroscopy (XPS, AXIS
   Ultra DLD, Shimadzu).

Preparation of CF NPs

   The CF NPs were synthesized using the hydrothermal method.^[ [155]^35
   ^] Firstly, CuCl[2]·2H[2]O (0.25 mmol) and FeCl[3]·6H[2]O (0.83 mmol)
   were stirred in deionized water (15 mL). Subsequently, trisodium
   citrate dihydrate (1.5 mmol), urea (2.25 mmol), and polyacrylamide
   (225 mg) were successively added to the above solution for further 1 h
   stirring. After a 12 h hydrothermal reaction at 473 K, the obtained CF
   was magnetically separated and washed using ethanol three times.

Preparation of CFMB NPs

   MET (3 mg) and BAY (1 mg) were successively added to the methanol
   dispersion of CF NPs (20 mg) for 24 h stirring at 25 °C. Next, prepared
   CFMB NPs were isolated via magnetism with distilled water washing 3
   times, and then freeze‐dried. According to the difference between the
   supernatant collected after centrifugation and the initial solution at
   290 nm, the loading rate (LC) of BAY was determined by the standard
   calibration curve. The LC of MET was determined as follows: CFMB NPs
   were dissolved in GSH (1 mm) to induce the degradation of CFMB and the
   complete release of MET. After the determination of their 230 nm
   absorption value, the LC of MET was obtained according to MET standard
   curves.

MET and BAY Release from CFMB NPs

   Phosphate buffer solution (PBS) was used to simulate a complex
   physiological environment. 10 mg of CFMB NPs were dispersed into 2 mL
   of PBS with a simulated TME condition (pH 6.5, GSH 10 mm) and kept
   continuous shaking. At the predetermined time point, 100 µL of solution
   was taken out for detection and filled with the same volume of fresh
   PBS. The cumulative quantity of MET and BAY released from NPs was
   calculated according to the UV–vis absorption intensity of MET and BAY
   at 230 and 290 nm, respectively.

Photothermal Performance Evaluation

   CFMB NP aqueous solution with various concentrations was illuminated by
   808 nm light, and its temperature variation was detected by a
   temperature sensor. Additionally, the photothermal stability is
   assessed by 808 nm light on/off cycle. Moreover, CFMB NPs were 808 nm
   light irradiated and then cooled down at room temperature to estimate
   the photothermal conversion efficiency based on the reported method.^[
   [156]^36 ^]

∙OH Detection

   A total of 2 mL of CF (50 µg mL^−1) aqueous solution containing GSH
   (10 mm) or (0 mm) were shaken for 10 min, and then TMB (4.17 mm) and
   H[2]O[2] (1 mm) were added, followed by testing the absorbance at
   652 nm. To research GSH influences on ∙OH generation, 652 nm adsorption
   variations of TMB in 2 mL of CF (50 µg mL^−1) aqueous solution
   containing TMB (4.17 mm), H[2]O[2] (1 mm) and various concentrations of
   GSH (1, 2, 4, 8, 10 and 15 mm) were tested. For estimating the role of
   temperature on ∙OH generation, the CF dispersion containing GSH,
   H[2]O[2,] and TMB was put in a 50 °C water bath or exposed to 808 nm
   light and then centrifuged to test the UV–vis absorption of the
   supernatant.

In Vitro Therapy Effect

   Live/dead cell staining experiments were carried out to evaluate the in
   vitro therapy effect. To be specific, 4T1 cells were put in 96‐well
   culture plates to grow overnight. Next, various treatments including
   NIR light, CF + NIR light, CFM, CFB, CFMB, or CFMB + NIR light
   respectively were applied. For experiments requiring NIR light, the
   1.25 W cm^−2 808 nm laser would be used to illuminate cells within
   10 min. After 24 h of treatment, calcein‐AM (2 µm) and PI (6 µm) were
   added to dye cells. The survival and death of cells were observed by an
   inverted fluorescence microscope.

Detection of Glucose, Lactic Acid, and ATP

   4T1 cancer cells were added to 6‐well plates and cultured overnight.
   Then various treatments including CF, CFM, CFM, or CFMB for 12 or 24 h
   respectively were applied. Following that, the cell culture medium was
   measured by the Lactate Assay Kit and Glucose Assay Kit following the
   instructions, respectively. For ATP, after different treatments, cells
   were washed using PBS and collected through trypsin digestion. The
   intracellular ATP content of different groups was determined by the ATP
   Assay Kit.

Western Blot Analysis

   The 4T1 cells were added to 6‐well plates and subjected to various
   treatments including CF, CFM, CFM, or CFMB for 12 or 24 h,
   respectively. Next, cells are digested and collected to extract
   proteins. A BCA protein assay kit has been employed in measuring the
   extracted protein concentrations. Subsequently, 15 µg isolated protein
   was blocked into the PVDF film. After that, PVDF film was cultured with
   the corresponding primary antibody and HRP‐conjugated secondary
   antibody. Last, an improved chemiluminescence system has been employed
   to visualize protein bands. The protein expressions were evaluated by
   analyzing the gray level.

Detection of Intracellular ∙OH

   The 4T1 cells were added to 6‐well culture plates to grow overnight.
   Next, NIR light, CF, CFMB, CF + NIR light, as well as CFMB + NIR light
   for 6 h respectively were applied. Finally, cells were stained by
   DCFH‐DA (5 µm) for 30 min and imaged with CLSM.

In Vivo Antitumor Study

   Balb/c mice have been ordered by Zhejiang Wei Tong Li Hua Biotechnology
   Co., Ltd. Animal‐related experiments have been conducted according to
   the protocols sanctified by the Animal Experiment Ethics Committee of
   Ningbo University. To establish a tumor model, 5 × 10^6 of 4T1 cells
   were injected subcutaneously into the right hind of each mouse. As
   tumor volume grew about 100 mm^3, 7 groups (5 mice for each group) were
   randomly divided and received diverse treatments as follows: 1) PBS, 2)
   PBS + NIR light, 3) CF + NIR light, 4) CFM, 5) CFB, 6) CFMB, and 7)
   CFMB + NIR light. The corresponding nanodrug in 0.1 mL of PBS was
   injected into each mouse through the tail vein. The time lag between
   material injection and the initiation of NIR light (0.5 W cm^−2) in
   animal experiments is 24 h. Their body weights as well as tumor volumes
   were monitored every other day during treatments.

RNA‐Sequencing Analysis

   The 4T1 cancer cells were added to 6‐well plates and cultured
   overnight. CFMB NPs were next added. After another 24 h treatment,
   total RNA was extracted for RNA‐sequencing analysis (OEbiotech, China).
   An R package, DESeq, was conducted to determine the differentially
   expressed genes (DEGs), using a cut‐off of P < 0.05. The heatmaps were
   built with R programming. The DEGs with | log2 (foldchange) | > 1 and P
   < 0.05 were used for Gene Set Enrichment Analysis (GSEA) via the
   Pathway enrichment analysis package in R software.

Statistical Analysis

   All data results were exhibited as the mean ± standard deviation by
   three parallel experiments. The statistical comparisons between
   experimental groups were determined by the Student's t‐tests. ^* p <
   0.05, ^** p < 0.01, and ^*** p < 0.001 were supposed to significant
   differences.

Conflict of Interest

   The authors declare no conflict of interest.

Supporting information

   Supporting Information
   [157]ADVS-11-2401738-s001.pdf^ (1.1MB, pdf)

Acknowledgements