MOF-Based Nanoagent Enables NIR-Triggered Dual Damage to Mitochondria via Synergistically Reinforced Oxidative Stress and Calcium Overload

We herein developed a core-shell type antitumor nanoagent based on the synergistically reinforced oxidative stress and calcium overload to mitochondria, both of which were triggered by near-infrared (NIR) light. The folic acid (FA) moiety decorated on MOF shells enabled ecient cellular uptake of nanoagents. The upconversion nanoparticle (UCNP) core converted NIR light to ultraviolet (UV) light with the latter catalyzed Fe 3+ -to-Fe 2+ reduction and simultaneously activated the photoacid generator encapsulated in the cavities of MOFs, which enabled the release of free Fe 2+ and photoacidication of intracellular microenvironment, respectively. The overexpressed H 2 O 2 in the mitochondria, highly reactive Fe 2+ and acidic milieu synergistically reinforced Fenton reactions for producing lethal hydroxyl radicals in mitochondria. Moreover, the photoacidication of plasma induced calcium inux, leading to calcium overload in the mitochondria. The therapeutic potency of the nanoagent based on the dual mitochondrial damage has been unequivocally conrmed in cell- and patient-derived tumor xenograft models in vivo. FMUP


Introduction
Despite remarkable advancements, mainstay anticancer strategies still encounter multifaceted challenges in terms of their limited effectiveness in the clinic. For instance, chemotherapy, the standard treatment strategy for patients with advanced stages of tumor progression, mostly involves modulation of the cell cycle and typically suffers from serious side effects and drug resistance [1][2][3][4] . To date, many evidences have revealed that mitochondria play key multifunctional roles in oncogenesis, including adenosine triphosphate (ATP) generation, redox and calcium homeostasis, and metabolic signal transduction [5][6][7] . Moreover, it has been well established that mitochondria in tumor cells are characterized by distinctive properties, including hypoxia and upregulated hydrogen peroxide (H 2 O 2 ) 8,9 . Thus, an opportunity has risen to develop new antitumor agents by targeting mitochondria and utilizing their inherent properties as crucial triggering factors, which is expected to con ne the range of action locally to the tumors and therefore present anticancer capabilities with the desired speci city and e cacy.
Among emerging antitumor therapeutics, chemodynamic therapy (CDT) based on oxidative stress holds unequivocal superiority in terms of matching with the abovementioned design rationale for mitochondrial damage 10 . Speci cally, CDT on the basis of the Fenton reaction or Fenton-like reactions can be facilitated in the mitochondrial microenvironment due to the upregulation of the H 2 O 2 substrate 8,9 . Moreover, as the crucial species involved in CDT, the •OH is characterized by much stronger oxidative stress than its counterpart active species involved in photodynamic therapy (PDT), namely, 1 O 2 11,12 . Additionally, CDT circumvents the intrinsic impediments originating from the hypoxia that PDT typically encounters 13 .
Although CDT holds great promise for antitumor treatment based on mitochondrial damage, some hurdles regarding its practicality still exist. Firstly, the Fenton reaction prefers an acidic environment, with an optimal reaction pH ranging from 2.0 to 5.0 14 , and the intracellular pH of tumor predominantly exhibited ~ 7.4 15 , which is not acidic enough to enable an e cient Fenton reaction and the accumulation of su cient •OH. Furthermore, because Fe 2+ is typically more productive in generating •OH via the Fenton reaction but more unstable than Fe 3+16,17 , e cient delivery of the Fe component into the tumor cytoplasm and the on-demand release of su cient Fe 2+ in situ remain a challenge. Additionally, for improved therapeutic effectiveness via additive or even synergistic effects, the challenge of simultaneously administration of combined approaches based on different pathways involved in mitochondrial damage also remains.
We herein developed a new type of metal-organic framework (MOF)-based core-shell nanoagent with NIRtriggered potent antitumor capability based on the combination of Fenton reaction-enabled oxidative stress and an imbalance in calcium homeostasis, both of which contribute to mitochondrial damage.
Such a type of nanoagent (denoted as FMUP) was fabricated via the one-pot self-assembly of a MOF (MIL-100, denoted as M) shell around an up-conversion nanoparticle (UCNP) core (denoted as U). The MOFs shell was constructed using Fe 3+ as central metal ion while 1, 3, 5-benzenetricarboxylic acid (BTC) and a small portion of folic acid (FA, denoted as F) as bridging ligands. Additionally, the photoacid generator (4-(2-(4-hydroxy-3,5-dimethoxyphenyl)-2-oxoethoxy)-4-oxobutan-1-aminium) (pHP, denoted as P) was encapsulated into the cavities of the MOF units. FA decorated onto the MOFs is expected to facilitate accumulation of the nanoagents at tumor sites via the overexpression of folate receptors (FRs) on tumor cells. The UCNP core is capable of converting NIR photons into UV photons 18,19 and is expected to play crucial roles in two key processes: i) the generation of UV light to mediate the reduction of Fe 3+ to Fe 2+20 , and ii) the transfer of excitation energy to photoacid generators within the close proximity to consequently trigger the release of H + from the latter (Scheme 1).
Following the rst process, the MOF units are expected to disintegrate owing to the change in chelation and release free Fe 2+ from the node sites of the framework 21,22 . As a result of the second process, the desired increase in the intralysosomal concentration of H + is expected. With sharp increases in intralysosomal concentration (Fe 2+ and H + ) and therefore osmotic pressure, the lysosomes are expected to burst and release Fe 2+ and H + into the cytoplasm. These NIR-triggered processes are expected to synergistically reinforce the e ciency of the Fenton reaction by presenting more productive reactants and the optimal reaction condition (an acidic environment) to therefore generate augmented oxidative stress in the H 2 O 2 -abundant mitochondria. As an accompanying effect associated with photoacidi cation, an increase in the acidity of the intratumoral cytoplasm will markedly give rise to calcium in ux causing mitochondrial calcium overload [23][24][25][26][27] . The outstanding therapeutic e cacy of FMUP based on the abovementioned NIR-triggered dual damage to mitochondria was unequivocally veri ed in various tumor cell lines in vitro, cell derived tumor xenograft (CDX) and patient derived tumor xenograft (PDX) models in vivo, proclaiming its potential as a safe and potent anticancer agent based on e ciently ampli ed mitochondrial damage. Scheme 1. Schematic Illustration of the construction of FMUP nanoagent and the underlying anticancer mechanism. (A) FMUP nanoagent containing UCNP as the core and photoacid (pHP) encapsulated in the FA-doped nanoagent shell by one-pot self-assembly. The FMUP nanoagent was synthesized by coordination of carboxyl groups on BTC and FA with Fe 3+ . The UCNP as the core was located in the shell and simultaneously the pHP was loaded in the pore of the nanoagent.  1A). During this process, the pHP generator molecules were incorporated within the cavities of the MOF (Fig. S1). The as-prepared nanoagents exhibited an average diameter of ~130 nm and a typical core-shell structure with the UCNP component located at the core cloaked by the MOF shell (Fig. 1B). This unique structure was further con rmed by the line scanning ( Fig. 1C) and energy dispersive X-ray data obtained from the elemental mapping of FMUP (Fig. S2). Owing to its partial participation in chelation with Fe 3+ ions ( Fig. S3), the FA moiety was successfully decorated onto the MOF units, which could be veri ed by the appreciable in uence on the size and surface properties (Fig. 1D). Speci cally, FMUP displayed a slightly increased size and decreased zeta potential compared to the pristine MUP, which can be attributed to the decoration of the FA moiety onto the surface of the MOF nanostructure. In terms of pHP, the loading amount was calculated to be up to 12.61%wt (Fig. S4).
With FMUP in hand, we further tested its stability. During one week of storage in cell culture medium without fetal bovine serum (FBS), FMUP did not display obvious changes in its hydrodynamic radius or zeta potential, suggesting favorable stability for intravenous injection (Fig. 1E). We also estimated the potential of such Fe-containing nanoagents as T 2 -weighted magnetic resonance imaging (MRI) contrast agents. As illustrated in Fig. 1F, the relaxation rate (1/T 2 ) of water protons markedly increased from approximately 2.7 to 49.7 s -1 , and the grey level of the T 2 -weighted images was signi cantly augmented upon increasing the concentration of Fe from 0.008 to 0.348 mM, indicating the su ciently strong magnetism of FMUP for in vivo MRI use and its potential for image-guided anticancer therapy.
Response capability of FMUP to NIR light As mentioned above, the design rationale for the nanoagent is based on the UCNP-mediated NIR-to-UV excitation energy conversion and subsequent UV light-mediated reduction of Fe 3+ to Fe 2+ and release of H + from the photoacid generator component (Fig. 1G). Upon irradiation with 980-nm laser (1.0 W/cm 2 , 5 min), we clearly observed that the FMUP-dispersed aqueous sample displayed strong emission with a peak at approximately 365 nm due to the encapsulated UCNP (Fig. S5). Moreover, the MOF-based coreshell nanostructures obviously disintegrated and aggregated in FMUP-L group (Fig. 1H), resulting in apparently marked increase in the size of the nanostructure with a much wider distribution (Fig. 1I). The above disintegration of FMUP could be assigned to the abrupt change in chelation of the central metal ions with ligands upon the NIR-to-UV triggered reduction. As shown in Fig. 1J In addition to Fe 2+ release, another outcome of the NIR-to-UV conversion was the photocatalysis of pHP to acidify the microenvironment 28,29 . For veri cation, we evaluated the acidity evolution of the FMUP suspension upon irradiation with 980-nm laser (1.0 W/cm 2 ) using a SNARF ® -1 probe. As shown in Fig.   1K, the pH of the sample rapidly decreased from weak alkalinity to weak acidity within 10 min. Together with the released Fe 2+ , such acidi cation could signi cantly accelerate the Fenton reaction. To test this aspect, we used 5, 5-dimethyl-1-pyrroline-oxide (DMPO) as a •OH trapping agent for ESR spectral characterization (Fig. 1L). The yield of •OH in the FMUP-L group was much higher than that in the FMU-L and FMUP groups, again indicating the synergistic effect of Fe 2+ and acidi cation on reinforcing •OH production. Similar results were also obtained via a bleaching experiment with methylene blue (MB) (Fig.   S6). These results together indicated the ideal response ability of FMUP to NIR light as well as the desired outcomes of •OH production.

Intracellular fate of FMUP
Owing to the overexpressed FA receptors on the tumor membrane 30,31 , functionalization of carriers with FA moieties has been proven to be very effective for circumventing the problem of low targeting e ciency to the tumor site. To evaluate the FA-enabled targeting performance, HeLa cells were incubated with Cy5labelled FMUP (with FA) or MUP (without FA), and confocal laser scanning microscopy (CLSM) images were captured ( Fig. 2A). As expected, HeLa cells treated with FMUP presented much stronger red uorescence than cells treated with MUP. These cells were also detected by ow cytometry for quanti cation (Fig. S7), and the internalization amount of FMUP was more than 2.9 folds that of MUP, again verifying that FA decoration signi cantly improved internalization by HeLa cells.
After internalization, most foreign nanocarriers are restricted to mature lysosomes. However, ideal mitochondria damage-targeting potency is expected if the Fenton reaction occurs in close proximity to the mitochondria because of the upregulation of H 2 O 2 in the mitochondria and the extremely short diffusion distance of the •OH species 32,33 . To test whether our FMUP could break through the lysosomal compartmentalization, we evaluated the co-localization of FMUP (labelled with Cy5) and lysosomes (labelled with LysoTracker™ Green) inside HeLa cells (Fig. 2B). Speci cally, the co-localization rate of FMUP with lysosomes dramatically decreased from 70.34% to 10.38% upon irradiation with 980-nm laser (1.0 W/cm 2 , 5 min). Moreover, the TEM images provided direct evidence that FMUP resided in lysosomes, while lysosome deconstruction was observed in the FMUP-L group. The enabling factor for such e cient lysosomal escape was plausibly due to the high osmotic pressure originating from the sharply upregulation of Fe 2+ and H + in the lysosomes that FMUP caused upon NIR light stimulus.

Intracellular acidi cation and the dual mitochondrial effects
Next, the capability of FMUP nanoagents to cause intracellular acidi cation was evaluated using SNARF ® -1 as a pH probe (Fig. 2C). Owing to the e ciently generated H + from pHP upon irradiation of 980-nm laser, HeLa with internalized FMUP emitted vivid green uorescence after irradiation, indicating an acidi ed cytoplasm. By using FMU-L as a control, we observed that the uorescence of HeLa cells did not change to green due to the absence of the photoacid generator. For detailed quanti cation, the intracellular pH after different treatments was determined from the calibration curves constructed earlier from the CLSM analysis. Notably, the intracellular pH value in the FMUP-L group appreciably decreased tõ 5.5 but was still approximately neutral in the FMU-L group, further verifying the crucial role of the photoacid generator for the desired intracellular acidi cation.
Given that abnormal intracellular acidity generally leads to an imbalance in intracellular calcium ion 24,27 , we continued to investigate the evolution of intracellular Ca 2+ levels in HeLa cells using Fluo-3 as a uorescent calcium probe (Fig. 2D). When the cells were incubated with FMUP or FMU, no noticeable increase in intracellular Ca 2+ levels prior to NIR light irradiation was observed. In sharp contrast, 5-min irradiation of 980-nm laser unequivocally generated a difference in cytoplasmic Ca 2+ levels with the probe in FMUP-internalized cells displaying clearly stronger uorescence than that in FMU-internalized cells and the discrepancy in uorescence brightness in two cases appreciably increasing up to a maximum level ~25 min after the irradiation. As a result of such photoacidi cation-enabled upregulation of intracellular Ca 2+ level, increase in the mitochondrial Ca 2+ level was also observed. Speci cally, it was found that the FMUP-L group clearly displayed much higher expression level of mitochondrial calcium uniporter (MCU) [34][35][36] than that the FMU-L group (Fig. 2E).
In addition to calcium overload in the mitochondria, the observed NIR-triggered release of Fe 2+ and H + also prompted us to investigate the in uence on intracellular •OH generation by using 2′,7′dichloro uorescin diacetate (DCFH-DA) as a probe (Fig. 2F). Considering that multiple factors were probably involved in the Fenton reaction, we investigated HeLa cells upon treatments with a series of formulations (PBS, MU, MU-L, FMU-L, FMUP-L, FMUP-L+Lip-1) for comparison. Compared to the MU group with almost no fluorescence, the cells in the MU-L group exhibited weak uorescence, which veri ed the NIR light-assisted reduction of Fe 3+ to Fe 2+ for the improved Fenton reaction. Owing to the increase in internalization mediated by the FA ligand, a further increased uorescence, which indicated intracellular •OH concentration, was observed in the FMU-L group. Upon further photoacidi cation that pHP enabled, cells in the FMUP-L group indeed exhibited the brightest uorescence, indicating the generation of considerable amount of •OH species from Fenton reactions. Considering the abundant H 2 O 2 in the mitochondria 8,9 , we also evaluated the discrepancy in •OH level in mitochondria. For veri cation, we labelled mitochondria with MitoTracker® Red (red color) and evaluated beaconing lipid peroxidation with a Liper uo probe. As shown in Fig. 2G, the evolution trend of mitochondria lipid peroxidation level in various cases displayed similar characteristic to that demonstrated in the results regarding •OH production illustrated in Fig.2F, with the FMUP-L group again showing the most lipid peroxidation in the mitochondria amongst all of the groups. Note that the addition of Lip-1, serving as a ROS scavenger 37,38 , alleviated lipid peroxidation in FMUP-L group, again indicating the role of •OH production in the cell oxidative stress.

Synergistic reinforcement of mitochondrial damage
Having demonstrated the dual effects in terms of calcium overload and •OH production in mitochondria, we moved on to investigate their synergistic outcomes on mitochondrial damage. To provide intuitional evidence, we started the investigation with TEM to reveal the change in morphology of HeLa cells after different treatments (Fig. 3A). It was found that cells after treatment of inert MU displayed typical mitochondrial morphology similar to that of cells in PBS group. In sharp contrast, other formulations of treatment with the involvement of the abovementioned enabling factors unequivocally gave rise to distinguishable morphological changes to the mitochondrial microstructure including profound swelling, outer membrane rupture and crista dissolution with the extent of mitochondria deconstruction augmented in the sequence of MU-L, FMU-L, and FMUP-L.
Considering that mitochondrial membrane potential (MMP) is typically sensitive to both calcium overload and the accumulation of ROS in mitochondria 25,39 , we next evaluated the MMP of HeLa cells after different treatments by using JC-1 as an indicator, which typically forms aggregates on normal mitochondrial membranes with a relatively high MMP while residing as monomers on the abnormal mitochondrial membrane with a low MMP. According to the ow cytometry data shown in Fig. 3B and D, the FMUP-L group showed the lowest proportion of aggregates (28.8%) and the highest proportion of monomers (70.7%) compared to the other groups, indicating the abnormality of MMP.
In addition to MMP, mitochondrial abnormalities can also be veri ed by the translocation of the cyclophilin D (Cyp D) protein from the cytoplasm to mitochondria, which is associated with the formation of the mitochondrial permeability transition pore complex 40,41 . To this end, we labelled mitochondria with MitoTracker® Green (green color) and evaluated the translocation of immunostaining Cyp D (red color).
As expected, most of the Cyp D molecules had transferred to the mitochondria (Fig. 3C), and their colocalization rate was up to 76.71% (Fig. 3E) in the FMUP-L group, which was substantially higher than that of the other groups. Given that mitochondria serve as powerhouses in eukaryotic cells, we also detected ATP production to examine mitochondrial function. Compared with the PBS and MU groups, ATP productivity gradually decreased in the other advanced treatments, with the value decreasing to 0.1 in the FMUP-L group (Fig. 3F). Beyond doubt, these results unequivocally con rmed the most potent mitochondrial damage was induced by FMUP-L, which was ascribed to not only the improved uptake but also the synergism of calcium overload and ROS production.

In vitro evaluation of cytotoxicity
Encouraged by the abovementioned results regarding mitochondrial damage, we subsequently investigated the viability of HeLa cells in vitro after different treatments (Fig. 4A). All formulations exhibited a dose-dependent effect on the viability of HeLa cells, and the cytotoxicity increased in the following sequence: MU, MU-L, FMU-L, and FMUP-L. Taking the dose of 100 μg/mL as an example, less than 10% of cells survived in the FMUP-L group, while the cell viabilities in the other groups ranged from 40% to ~90%. Given that two pathways were involved in mitochondrial damage, the possibility of ferroptosis induced by ROS production and apoptosis induced by calcium overload thus coexisted. To this end, we next investigated the two aspects of ferroptosis and apoptosis.
Ferroptosis was veri ed by the expression level of glutathione peroxidase 4 (GPX4), which has been regarded as the major indicator of oxidative stress due to its reduction capacity to reactive oxygen species (ROS) 42,43 . As the western blot data show in Fig. 4B, cells in MU group displayed GPX4 level roughly similar to the counterpart in PBS group due to the inert activity of MU in cells. Upon NIR lightassisted reduction of Fe 3+ to Fe 2+ , the improved Fenton reaction and •OH production de nitely decreased the expression level of GPX4 (MU-L group) while the increase in internalization of nanoagents resulted in further suppression of GPX4 expression (FMU-L group). Particularly noteworthy was the involvement of pHP capable of photoacidi cation that signi cantly improved the e ciency of Fenton reaction and enabled highly e cient production of •OH (FMUP-L group), which substantially inhibited the expression of GPX4 and consequently presented GPX4 level merely one-tenth of that in the PBS group. Subsequently, the level of intracellular caspase-3, a typical indicator of apoptosis 44 , was evaluated via immunostaining strategy with the results illustrated in Fig. 4C. It can be clearly seen that cells in the FMUP-L group displayed much higher level of caspase-3 as compared to the counterpart level in other groups, indicating the signi cant involvement of apoptosis in the former that was assigned to the photoacidi cation effect. Such cooperation of ferroptosis and apoptosis for cytotoxicity in the FMUP-L group was also veri ed by a live/dead assay (Fig. 4D). Compared with the other treatments, dead cells (red color) dominated the cell population in the FMUP-L group, indicating that most cells had been killed due to the combination of ferroptosis and apoptosis. Similar cytotoxic results were also observed in human liver carcinoma cells (HepG2) and mouse breast cancer cells (4T1) in Fig. 4E, and FMUP-L still substantially outperformed the other counterparts

In vivo evaluation of biodistribution
It is well documented that the ultimate therapeutic e cacy of anticancer nanomedicines critically depends on their in vivo fate. Taking this into account and being inspired by the aforementioned results, we assessed the fate of the engineered nanoagents in vivo. Considering that the FA ligand is highly correlated with targeting capacity, we enlisted FMUP and MUP for comparison, which were labelled by loading Cy7 dye into the MOF cavity. As the in vivo imaging data in Fig. 5A show, after intravenous injection, the majority of the MUP signals were rapidly enriched in the liver, with a minority of the MUP signals accumulating in the tumor site. In contrast, the uorescence signal of FMUP at the tumor site increased substantially at the expense of the signal in the liver over time, indicating the pivotal role of doped FA for tumor accumulation. Despite this discrepancy, the accumulation of these two types of nanoagents at the tumor site peaked at 8 h (Fig. 5B). Upon further calculation of the signal ratio of the tumor/liver, we observed a magni ed discrepancy evolution. Speci cally, the ratio in the FMUP group increased and reached a peak value of ~1.5, while that in the MUP group continuously declined and reached 0.2 in 48 h (Fig. 5C).
To gain deeper insight, we excised the main organs, detected the uorescence, and quanti ed the signal intensity. As shown in Fig. 5D and E, the uorescence intensity of the excised tumor in the FMUP group was 2.4-fold higher than that in the MUP group, unequivocally proclaiming the FA-enabled tumortargeting capacity of FMUP for improved therapeutic performance. In other organs, the main discrepancy was found in the liver, and the FMUP group exhibited half of the accumulation the MUP group displayed, thus reducing the possibility of side effects to the liver. For higher resolution, these organs were further treated as frozen slices for microscopic observation (Fig. 5F). Compared with the MUP group, the FMUP group had substantially improved in ltration in the tumors and signi cantly suppressed accumulation in the liver. In addition to uorescence-based measurements, the T 2 signal of Fe 3+ from our nanoagents also enabled us to investigate the biodistribution via MRI (Fig. 5G). FMUP displayed remarkably stronger contrast than MUP in T 2 imaging at the tumor sites, again con rming its superior tumor targeting capacity.
In Taking the synergistically reinforced mitochondrial damage for in vitro cytotoxicity, we evaluated whether this aspect was highly correlated with the above distinct therapeutic outcomes. To estimate the ability of regulating the intratumoral acidity, SNARF ® -1 was used as a probe and administered via intravenous tail injection to map pH uctuations at the tumor sites on day 14 using an animal uorescence imaging system (Fig. 6D). Referenced by the calibration curves obtained via the imaging system, the pH value at the tumor site after FMUP-L treatment was found decreased to ~5.6, while the counterparts in other groups remained at ~6.6, highlighting the pivotal role of the photoacid generator in intratumoral acidi cation. Consequently, the intratumoral Ca 2+ level (detected by the Fluo-3 probe) in the FMUP-L group was signi cantly higher than that in the other groups without the involvement of a photoacid generator (Fig. 6E), laying the foundation for FMUP-L treatment to enable calcium overload in tumor cells.
To verify the promoting effect of intratumoral acidi cation on Fenton reaction, we evaluated the ROS production at the tumor site via two-photon uorescence microscopy using DCFH-DA as a probe. With the involvement of other promoting factors including the NIR-assisted reduction of Fe 3+ to Fe 2+ and the FAmediated improvement in internalization, the ROS productivity was found increased in the sequence of MU, MU-L, FMU-L, and FMUP-L (Fig. 6F). Due to the synergism of the above calcium overload and ROS production, the expression level of uncoupling protein 2 (UCP 2, a typical indicator of mitochondrial dysfunction 45,46 ) in the FMUP-L group was markedly stronger than that in the other groups (Fig. 6G), indicating the most substantial mitochondrial damage in tumor cells in vivo. The FMUP-L group also displayed the fewest tumor cell nuclei in the H&E slice (Fig. 6H). It also deserves mentioning that the potent antitumor effects of FMUP-L were achieved with few abnormalities to the histology of the main organs (Fig. S8), serum biochemical indicators (Fig. S9), and body weight (Fig. S10), indicating the good safety of our nanoagent.
In vivo antitumor effects in patient derived tumor xenograft (PDX) model Cell lines widely used in routine antitumor experiments typically suffer from the lack of strong clinical relevance. As practical alternative models, PDX models mostly retain the principal histological and genetic characteristics of their donor tumor and remain stable across passages, which make them more reliable for clinical outcome prediction and preclinical drug evaluation. To con rm the clinical applicability of our nanoagent, we evaluated its therapeutic effect in a PDX model, which was established by transplanting a primary tumor sample resected from a liver cancer patient into the axilla of NOD-SCID mice followed by engraftment three times for subsequent use (Fig. 7A). Based on the immunostaining and ow cytometry analysis of dissociated cells, we con rmed that the high expression of the FA receptor remained in the PDX model (Fig. 7B). Additionally, FMUP (labelled by Cy7) showed ~2.4-fold accumulation in tumors (Fig. 7C) compared with its counterpart without FA functionalization. This FAmediated targeting capacity was further echoed at the histological level, with far more nanoagents observed in the tumor slice in the FMUP group than those in MUP group (Fig. 7D).
Having con rmed the targeting capacity of FMUP in the PDX model, we continued to investigate the therapeutic outcomes of FMUP-L treatment. To this end, patient-derived tumor-bearing mice were irradiated using a 980-nm laser (1.0 W/cm 2 , 5 min) twice, at 8 h and 32 h, respectively, after a single intravenous administration of FMUP nanoagent. Both in vivo imaging and two-photon microscopy imaging of tumors with pretreatment of SNARF ® -1 probe unequivocally con rmed that the pH value at the tumor site in the FMUP-L group was lower than that of the FMUP group ( Fig. 7E and F). Downstream of intratumoral acidi cation was further evaluated by ow cytometry and frozen section staining imaging. Speci cally, the FMUP-L group showed substantially shifted beaconing signal of DCFH-DA probe (Fig. 7G) and stronger signal of Fluo-3 (Fig. 7H) as compared to those in FMUP group, therefore strengthening the argument that intratumoral acidi cation could induce both calcium overload and e cient ROS production. Owing to the combination of these two enabling factors, signi cantly elevated expression of UCP 2 was observed on FMUP-L-treated tumor slices (Fig. 7I). Such potent mitochondrial damage resulted in prominent suppression of tumor proliferation (Fig. 7J) and almost complete inhibition of tumor growth (Fig. 7K). Correspondingly, the FMUP-L group showed a 100% survival rate after 60 days, compared with all untreated mice that died within 25 days. These results together suggested that our FMUP-L nanoagent offered speci c and highly effective antitumor therapy in the PDX model, thus shedding light on its potential clinical e cacy.

Discussion
The salient features of our FMUP nanoagent can be summarized in terms of the ingenious structureactivity relationships. The FMUP nanoagent is characterized with unique core-shell structures with UCNP as the core and MOF as the shell while photoacid generator component incorporated in the cavities of the MOF units. The UCNP core capable of upconverting NIR to UV light markedly circumvents the inherent impediments that UV light typically encounters in biological applications regarding phototoxicity to normal tissues on the optical path and the restricted penetration depth. Moreover, full exploitation of central metal ions, FA trims, and cavities in the MOF shell for antitumor action was realized for the Fenton reaction, targeting capacity, and pHP loading, respectively.
On the basis of the abovementioned design rationale, we succeeded in implementing synergistic damage to mitochondria with FMUP nanoagent. Speci cally, 980-nm light mediated reduction of Fe 3+ to Fe 2+ and simultaneously activated the photoacid generator for acidifying the intracellular microenvironment.
Together with the released Fe 2+ , the acidic intracellular microenvironment, on the one hand, improved the e ciency of Fenton reactions for production of lethal •OH. Of particular note is that, as abovementioned, the upregulated H 2 O 2 in mitochondria of tumor cells con ned the •OH-based oxidative stress within the close proximity of mitochondria. On the other hand, intracellular acidi cation gave rise to calcium overload in mitochondria. The combination of the FA-enabled tumor targeting ability and the targetspeci c damage based on the abovementioned two mechanisms tactfully con ned the range of antitumor action locally to the tumors and therefore presented the desired antitumor speci city and e cacy. Based on the above synergistic damage to mitochondria, the FMUP nanoagent unequivocally displayed a potent killing effect on various tumor cell lines, including HeLa, 4T1, and HepG2. The in vivo experiments also provided unequivocal evidence that a single administration of FMUP with twice irradiation of 980-nm light could completely suppress tumor development in both CDX and PDX models, supporting FMUP-L as a promising modality to ght against cancer.
To the best of our knowledge, this is the rst paradigm where a photoacid generator incorporated in MOF nanostructure was used for triggering synergistic damage to mitochondria based on the intratumoral photoacidi cation process and the consequent oxidative stress and calcium overload. This proof-ofconcept implementation opens up a way of exploring optically manipulating intracellular acidity/alkalinity for combined/synergistic antitumor therapeutics. For instance, replacing photoacid with photoalkali components (e.g., malachite green) and exploring the potential effects on cell viability and synergism with other therapeutic modalities could plausibly provide promising opportunities to antitumor applications. Further opportunities also exist in imparting additional and more potent therapeutic activities via the way of optimizing the functional components of MOFs. For example, replacing Fe 3+ with Cu 2+ in constructing the MOF shells is expected to enable Cu + -mediated pyroptosis pathway and chemodynamic therapy based on the photocatalyzed reduction of Cu 2+ to highly active Cu + for Fentonlike reaction. Additionally, integrating drug active ingredients into the organic linkers could be a way for introducing additional antitumor activity for further enhancement of therapeutic effects. DLS characterization of FMUP. The size and zeta potential of FMUP or MUP were determined using a Malvern laser particle size analyzer; (NANO ZS, England). In order to determine stability of FMUP, the FMUP were stored in cell culture medium without fetal bovine serum (FBS). Then its size and zeta potential were detected every day.

Materials And Methods
TEM characterization of FMUP. The FMUP was dropped on the copper grid for three replicates. The morphology of the FMUP was observed by TEM at 100 kV (JEOL JEM-1400, Japan). In order to analyze the elements proportion of component in FMUP, a mapping scan analysis was performed.
BET characterization for different formulations. The Brunauer-Emmett-Teller (BET) surface area and pore size of FMU were measured using ASAP2050 system (England Fe 2+ can form a stable complex with 1,10-phenanthroline which has a maximum absorbance at 510 nm. The absorbance of the complex was monitored using automatic microplate reader (Tecan In nite M200, Switzerland) at 25 °C.  (1000 rpm, 5min). Finally, the uptake amount upon different treatments was determined by ow cytometer (FCM). Data were obtained from endocytosis, the same method was executed within different internalized times. Assessment of NIR light-triggered acidi cation of HeLa cells by CLSM. The cultural conditions for HeLa cells were described above and FMUP were added and incubated into the cells at 0.05 mg/mL for 12 h.
Before NIR irradiation (980 nm, 1.0 W/cm 2 , 5 min), the cells were incubated with pH indicator (SNARF ® -1) at 50 nM for 30 min. SNARF ® -1 was excited at 488 nm. The corresponding fluorescent images at 560-590 nm and 620-650 nm were taken by CLSM. And the uorescence ratio of intensity 580 nm and intensity 640 nm were calculated by NIKON analyzer. Subsequently, the pH value was calculated by using the calibration curve.
In vitro assessment of NIR light-triggered acidi cation on calcium in ux by CLSM. The cultural conditions for HeLa cells were described above and FMUP or FMU were added at 0.05 mg/mL for 12 h incubation.
Before NIR irradiation, the cells were incubated with Ca 2+ indicator (Fluo-3) at 5 μM for 30 min. Then, the excess Fluo-3 and MOFs were removed and washed by cold PBS in dish. Finally, the fresh DMEM media were added into the dish. Next, we employed the CLSM to monitor the distribution of Ca 2+ in HeLa cells In vitro assessment of NIR light-triggered acidi cation on calcium overload in mitochondria by CLSM. Different groups' cell was treated and cultured with the described condition in 35 mm glass-bottom dishes. After NIR irradiation (980 nm, 1.0 W/cm 2 , 5 min), mitochondria were stained by MitoTracker™ Green at 200 nM for 30 min in 37 °C. Then cells in different groups were washed by cold PBS and were xed by 4% paraformaldehyde for 30 min in 37 °C. Subsequently, cells were incubated with 0.2% Triton X-100 in PBS for 10 min to achieve good permeabilization. Immediately, the cells were blocked with blocking buffer for 2 h at room temperature. Primary antibody to the mitochondrial calcium uniporter (MCU) was incubated with the cells in 4 °C overnight. After the cells were washed three times by cold PBS, they were incubated with goat anti-rabbit IgG-Alexa 647 secondary antibody for 2 h in room temperature. Red MCU was exited at 633 nm. Green mitochondrial was exited at 488 nm. The corresponding fluorescent images were obtained by CLSM. And the expression level of MCU was calculated by NIKON analyzer.
In vitro evaluation of NIR light-triggered acidi cation on ROS generation by CLSM. The cultural conditions for HeLa cells were described above, and different formulations were added at 0.05 mg/mL overnight.
The ROS were reacted with DCFH-DA for 20 min in 37°C before NIR irradiation. Based on the Fenton reaction, Fe 2+ could react with H 2 O 2 to produce ROS more strongly and actively in acidic environment than that in weak alkaline environment after NIR irradiation (980 nm, 1.0 W/cm 2 , 5 min). Then the cellular ROS oxidized DCF can be used as indicator for ROS production. DCF was excited at 488 nm. The corresponding fluorescent images of cellular DCF at excitation wavelength of 510-555 nm were taken by CLSM.
In In vitro evaluation of NIR light-triggered acidi cation on mitochondrial membrane potential by FCM. In short, HeLa cells were seeded in 48-well plates at a density of 1 × 10 6 cells per well for attaching overnight. Then the wells were added to different formulations (equal concentration 50 μg/mL) for incubating 12 h. After NIR irradiation (980 nm, 1.0 W/cm 2 , 5 min), the probe (JC-1) for measuring mitochondrial membrane potential (10 μg/mL) incubated with cells for 10 min in 37 °C. Next, the cells were washed by cold PBS with three times before collecting cells. And the mitochondrial membrane potential was determined on ow cytometer. The receiving emission wavelength of the JC-1 monomer is 529 nm and the receiving emission wavelength of the JC-1 aggregate is 590 nm. Data were obtained from 15000 cells for each sample. Assessment of NIR light-triggered acidi cation on apoptosis by CLSM. Different groups' cell was treated with the described condition in 35 mm glass-bottom dishes. After NIR irradiation, nuclei were stained by Hoechst 33342 with 20 min in 37°C. Then cells of different groups were washed by cold PBS and were xed by 4% paraformaldehyde for 30 min in 37°C. Subsequently, Cells were incubated with 0.2% Triton X-100 in PBS for 10 min to achieve good permeabilization. Immediately, the cells were blocked with blocking buffer for 2 h at room temperature. Primary antibody to the caspase-3 was incubated with the cells at 4°C overnight. After the cells were washed three times by cold PBS, they were incubated with goat anti-rabbit IgG-Alexa 647 secondary antibody for 2 h at room temperature. Blue nuclei were excited at 405 nm. Pink caspase-3 was exited at 633 nm. The corresponding fluorescent images were taken by CLSM. And the expression level of caspase-3 was calculated by NIKON analyzer.
In vitro evaluation of cellular apoptosis by live/dead assay. HeLa cells were cultured as described above, where L is the longest and W is the shortest tumor diameter (mm).
Antitumor study in humanized PDX model. The PDX tumor samples were s.c. transplanted into the axilla of the NTG mice to establish the PDX model and then these mice were randomly divided into two groups: PBS group and FMUP+L group. Two weeks later, when the tumors were visible (~150 mm 3 ), the mice of these groups were i.v. injected with PBS and FMUP, respectively. Eight hour later, the mice in FMUP+L group were anaesthetized and the tumors of them were illuminated with a 980-nm laser (1.0 W/cm 2 , 5 min). Note that mice in the groups with the involvement of light irradiation underwent treatment of irradiation (980 nm, 1.0 W/cm 2 , 5 min) twice, at 8 h and 32 h, respectively, after a single intravenous administration. Finally, the tumor sizes and survival percent were measured every two days, and the experimental endpoint was de ned as either death or the tumor size greater than 1500 mm 3 . The biodistribution study was administered in the same ways as we used in the corresponding studies. And the acidi cation study was investigated by the multiphoton laser confocal microscopy. The evaluation of ROS generation in tumor was measured by FCM. Besides, the investigation for calcium in ux in tumor was detected by frozen section staining using Fluo-3. Moreover, the evaluation of mitochondria damage was measured via immunohistochemical staining. The haematoxylin and eosin (H&E) staining of tumor sections for the study was administered in the same way. The study was approved by Shanghai Tongren Evaluation of acidi cation on tumor by animal imaging system. Before measuring the pH values in tumor sites, the standard solutions with different pH values (5.0, 5.5, 6.0, 6.5, 7.0, and 7.5) were mixed with SNARF ® -1 at a volume ratio of 1:100, respectively. And the uorescence intensities at 580 nm and 640 nm of various standard solutions were determined by animal imaging system. As a result, the calibration curve was obtained. To estimate the pH reversal in HeLa the tumor site, the mice with different treatments (PBS, MU, MU-L, FMU-L, and FMUP-L) were intravenously injected with a pH-sensitive uorescent dye (SNARF®-1) at 20 min before imaging. After that, the mice received NIR irradiation in the corresponding groups. Finally, the pH values in tumor sites were calculated by the ratio of the uorescence at 580 nm and 640 nm.
Evaluation of acidi cation on ROS production in tumor. In order to investigate the generation of ROS in situ, DCFH-DA was intratumoral administered before imaging by multiphoton laser confocal scanning microscopy. And then mice's tumor was irradiated by NIR (980 nm, 1.0 W/cm 2 , 5min). Subsequently, the green ROS was detected and measured by multiphoton laser confocal scanning microscopy and further quantitative analysis for different treatments.
Evaluation of the synergistic effect on mitochondria by immunohistochemical section. After the completion administering, one mouse was taken from each group for immuno uorescence analysis. The expression of uncoupling protein 2 (UCP 2) in mitochondrial membrane induced by the synergistic effect was also veri ed by immunohistochemistry.
Evaluation of the synergistic effect on tumor inhibition by immunohistochemical section. After the completion administering, one mouse was taken from each group for immuno uorescence analysis.
Proliferating cell nuclear antigen-K i 67 was used for determining nucleus proliferating of tumor tissue.
Safety evaluation of different treatments. For further investigate the safety of different formulations in vivo, the body weight of mice in each groups were recorded until 60 days. Besides, the serum levels of urea nitrogen (BUN), lactate dehydrogenase (LDH), alanine aminotransferase (ALT), aspartate transaminase (AST), and alkaline phosphatase (ALP) were analyzed by using an automated analyzer (Hitachi Ltd Hitachi-917, Japan). The main organs (heart, liver, spleen, lung, and kidney) were sliced and stained by hematoxylin-eosin (H&E) staining.
Animal care. Balb/c-nu mice and NTG mice, 4-6 weeks of age, were obtained from Vital River laboratories Statistical analysis. All the data are presented as the mean ± SD. Statistical analysis was performed with Prism 8.0 software (GraphPad Software) by an unpaired Student's t-test, log-rank test, one-way ANOVA.

Declarations
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The authors declare that all relevant data are included in the paper and supplementary information les.
Source data are provided with this paper.   represent the means ± SD (n=3). Statistical signi cance between groups in (C) was calculated using twotailed unpaired Student's t-test. *P<0.05, **P<0.01.