Abstract

   The increasing incidence of central nervous system lymphoma (CNSL) is
   hindered by the blood-brain barrier and costly prolonged drug
   development. To overcome these obstacles, we developed a carboplatin
   lock-designed MOF (Pt-MOFs@Glu) targeting intestinal macrophages for
   brain-directed drug transport. Single-cell RNA sequencing analyses
   revealed that intestinal macrophages migrate to the brain in response
   to chemokines. Building on this insight, Pt-MOFs@Glu was designed to
   engage these cells for precise delivery. Comprising carboplatin,
   pyrazine-quinoxaline, and β-glucan, the system induces an “avalanche
   effect” in the CNSL microenvironment, promoting tumor apoptosis and
   inhibiting metastasis. Combining pyrazine-quinoxaline to lock
   carboplatin and β-glucan to boost targeting, immunity, and oral
   absorption, the system enables ROS-triggered drug release, efficiently
   crosses the gastrointestinal tract and BBB, and synergizes
   chemo-immunotherapy to enhance therapeutic efficacy. This approach
   redefines CNSL treatment by harnessing the gut-brain axis, offering a
   transformative pathway to overcoming therapeutic barriers and improving
   patient outcomes.
     __________________________________________________________________

   Lock-designed MOF leverages gut-brain macrophage trafficking to trigger
   an avalanche effect against CNS lymphoma progression.

INTRODUCTION

   Central nervous system lymphoma (CNSL) represents a formidable
   challenge in oncology, accounting for ~4% of intracranial malignant
   tumors and with an incidence of about seven cases per million ([40]1,
   [41]2). This aggressive cancer progresses rapidly if left untreated,
   with a mean survival time of just 1.5 months, underscoring the urgent
   need for effective therapeutic strategies ([42]3, [43]4). Current
   treatment modalities—which include high-dose chemotherapy, targeted
   therapy, radiotherapy, and immunotherapy—have not substantially
   improved outcomes, as evidenced by high recurrence and mortality rates
   in clinical trials. A major obstacle to effective treatment is the
   blood-brain barrier (BBB), a selective barrier that prevents many drugs
   from entering the central nervous system (CNS) ([44]5–[45]7). In
   addition, conventional therapies often result in severe side effects
   such as leukopenia, dermatological issues, alopecia, nausea, and
   fatigue, which can notably affect the quality of life of patients.
   Therefore, overcoming the BBB to facilitate efficient drug delivery to
   CNSL sites, while minimizing side effects and preventing recurrence,
   remains a critical clinical imperative ([46]8, [47]9).

   Traditionally, CNSL treatment has relied heavily on intravenous drug
   administration. While effective in some respects, intravenous therapy
   can be invasive, increase the risk of infection, and require clinical
   supervision ([48]10, [49]11). In contrast, oral anticancer medications
   offer numerous advantages, including painless administration, a lower
   risk of contamination, the ability for self-administration, and
   cost-effectiveness, making them a preferred option for many patients
   ([50]12, [51]13). However, oral drug delivery systems face notable
   physiological challenges posed by the gastrointestinal tract (GI) and
   the BBB, making the development of a safe and effective gut-brain axis
   drug delivery strategy a key objective ([52]14, [53]15).

   The concept of the gut-brain axis, introduced in 2011, describes the
   complex and bidirectional biochemical signaling between the GI and the
   CNS ([54]16, [55]17). This system involves multifaceted biochemical
   signaling pathways that enable the gut and the brain to influence each
   other’s functions, maintaining homeostasis and responding to various
   physiological and pathological stimuli. This axis includes connections
   from the gut to the brain through the CNS, immune system, enteric
   nervous system, vagus nerve, and gut microbiota ([56]18–[57]21). Recent
   studies have highlighted the notable role of the gut-brain axis in CNS
   pathologies. Intestinal immune cells, particularly those in the
   gut-associated lymphoid tissue, can become activated by gut microbes or
   dietary antigens. Once activated, these immune cells can migrate to the
   CNS, where they can cross the BBB and potentially influence the
   progression of neurological diseases through mechanisms such as
   neuroinflammation and immune modulation ([58]22–[59]25). This suggests
   that immune cells could potentially be harnessed as vehicles to
   transport orally administered drugs to the brain. By targeting immune
   cells in the gut with drug-loaded nanoparticles or other delivery
   systems, it may be possible to facilitate the efficient delivery of
   therapeutics across the BBB, thereby enhancing treatment efficacy for
   CNS diseases.

   Building on these insights, the study initially used bioinformatics and
   single-cell RNA sequencing (scRNA-seq) techniques to analyze CNSL.
   Results revealed that the CNSL microenvironment contains a significant
   presence of gut-derived macrophages, which can migrate from the
   intestine to the brain in response to chemokines, such as Cx3cr1
   ([60]26–[61]29). Consequently, this study aims to develop a drug
   delivery platform that harnesses the gut-brain axis to bypass
   physiological barriers noninvasively, enabling macrophage-mediated
   transport from the intestine to directly target CNSL treatment. This
   approach seeks not only to eradicate tumor cells but also to activate
   the immune system synergistically, providing a comprehensive treatment
   strategy for CNSL. The proposed platform features a unique oral
   carboplatin padlock-designed metal-organic framework (MOFs), termed
   Pt-MOFs@Glu, which is designed to target and treat CNSL ([62]Fig. 1).
   This system comprises carboplatin as the metal center,
   pyrazine-quinoxaline as a molecular lock for carboplatin, and β-glucan
   as both the CNS-targeting and immune-activating component.

Fig. 1. A noninvasive oral drug delivery system for gut-to-brain transport
and its mechanism of action.

   [63]Fig. 1.
   [64]Open in a new tab

   (A) Bioinformatics and scRNA-seq analysis of human clinical CNSL
   samples and mouse CNSL samples, and the dashed marked portions are
   intestinal-derived macrophages in the tumors. (B) System relates to the
   construction and synthesis process of padlock-designed metal-organic
   frameworks (Pt-MOFs@Glu). (C) Upon oral administration, Pt-MOFs@Glu are
   taken up by intestinal M cells and subsequently endocytosed by resident
   macrophages (Mφ) in situ. These macrophages, carrying the Pt-MOFs@Glu,
   travel via the lymphatic system to the circulatory system and
   ultimately cross the blood-brain barrier (BBB) to reach the brain tumor
   site. Within the brain tumor, Pt-MOFs@Glu release carboplatin in
   response to high levels of ROS, triggering an avalanche effect to exert
   a therapeutic impact on the tumor. In addition, this system
   synergistically enhances chemotherapy and immune response, promotes
   tumor cell apoptosis, and effectively prevents tumor metastasis. Some
   elements of this figure were created in BioRender (Yang-Bao Miao, 2025;
   [65]https://BioRender.com/ef5jegu).

   The Pt-MOFs@Glu system uses a lock-and-key mechanism, using
   pyrazine-quinoxaline as a molecular lock to immobilize carboplatin,
   while reactive oxygen species (ROS) within the tumor microenvironment
   (TME) act as the “key” to trigger drug release. This dynamic
   interaction initiates an “avalanche effect,” significantly enhancing
   the therapeutic efficacy of Pt-MOFs@Glu. This mechanism enables precise
   drug delivery to the tumor site, thereby reducing systemic toxicity and
   improving treatment outcomes. Furthermore, the immune-stimulating
   damage-associated molecular pattern (DAMP) proteins released upon tumor
   cell death interact synergistically with β-glucan, a potent immune
   adjuvant, to activate the immune system ([66]30, [67]31). This combined
   chemotherapy-immunotherapy approach is finely tuned to the TME,
   offering a robust and comprehensive strategy for treating CNSL. The
   Pt-MOFs@Glu represents a notable advancement in CNSL treatment,
   leveraging the gut-brain axis to induce an avalanche effect within the
   CNSL microenvironment. By synergistically integrating the advantages of
   chemotherapy and immunotherapy and harnessing the distinctive dynamics
   of the gut-brain axis, this system heralds a groundbreaking approach to
   CNSL treatment, promising to significantly enhance patient outcomes.

RESULTS

Acquisition and analysis of scRNA-seq

   Single-cell transcriptome data for human CNSL and normal human brain
   samples were obtained as raw datasets from the ArrayExpress platform.
   The CellRanger software (version 2.0.0;
   [68]http://software.10xgenomics.com/single-cell/overview/welcome)
   ([69]32) was used to process each sample’s raw data, generating
   relevant gene expression profiles ([70]Fig. 2, A to C). Cell-type
   identification, based on distinct cell-specific markers, revealed the
   presence of B cell, dendritic cell, endothelial cell, macrophage,
   meningeal cell, cancer cells, and T and NK cell ([71]Fig. 2, B and C).
   Notably, macrophages represented the largest proportion of cell types
   when comparing normal brain samples to those of CNSL ([72]Fig. 2, D and
   I). Given their prominence as a central component of the TME,
   macrophages were isolated for detailed analysis.

Fig. 2. Bioinformatics analysis of scRNA-seq in CNSL tissue and normal human
brain, along with immunofluorescence from patients with CNSL.

   [73]Fig. 2.
   [74]Open in a new tab

   (A) Schematic representation of the experimental workflow. Some
   elements of this figure were created in BioRender (Yang-Bao Miao, 2025;
   [75]https://BioRender.com/ef5jegu). (B) UMAP plot for the analysis of
   cell subpopulations in the CNSL, which were divided into seven cell
   subsets. (C) UMAP plot for the analysis of cell subpopulations in the
   normal human brain, which were divided into seven cell subsets. (D)
   Comparison of intracellular macrophage content between CNSL tissues and
   normal brain tissues. Expression of intestinal macrophage markers in
   CNSL-associated macrophages, including CD11c (E) and CD14 (F). (G)
   Volcano plot highlighting differentially expressed genes between CNSL
   and normal brain samples, with significantly up-regulated and
   down-regulated genes marked in red and blue, respectively [P < 0.05,
   log[2](fold change) > 1]. (H) Gene Ontology (GO) annotation and KEGG
   pathway analysis of ESRP1 in CNSL using LinkedOmics. Confocal laser
   scanning microscopy (CLSM) images reveal that macrophages present in
   CNSL tissues are derived from intestinal macrophages. (l) UMAP
   visualization highlighting macrophage populations across human
   intestinal tissue and PBMCs. Although all three clusters were
   identified in both tissues, cluster 5 was predominantly found in
   PBMC-derived macrophages, whereas cluster 11 was more dominant in
   intestine-derived macrophages. (J) Annotated UMAP visualization of
   CD45^+ immune cell subsets derived from both human intestinal tissue
   and PBMCs. (K) UMAP visualization of CD45^+ immune cell subsets derived
   from either human intestinal tissue or PBMCs. (L) UMAP visualization
   highlighting macrophage populations across human intestinal tissue and
   PBMCs. (M) Heatmap showing the expression profiles of
   macrophage-related genes in human intestinal tissue and PBMCs. (N)
   Expression levels of intestinal macrophage markers in CNSL-associated
   macrophages, including LSTQ, LYZ, THBS1, HLA-DQA1, and HLA-DQB1.

   Our findings indicate a notably higher level of macrophage infiltration
   in CNSL tissue compared to normal brain tissue ([76]Fig. 2, B to D),
   suggesting a correlation between macrophage presence and CNSL
   development. Further analysis of 4346 qualified macrophage-related
   genes highlighted key markers indicative of the macrophages’ origin and
   behavior in CNSLs ([77]Fig. 2E). Genes associated with intestinal
   macrophages, including CD11c and CD14 (a key marker for human
   intestinal macrophages), exhibited increased expression in
   CNSL-associated macrophages. These results support a model in which
   intestinal macrophages are a primary source of macrophages within
   CNSLs.

   To explore the mechanisms underlying macrophage migration from the gut
   to CNSL, we performed gene set enrichment analysis to identify the top
   20 significant Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways.
   The analysis revealed that macrophage-related pathways in CNSLs are
   predominantly associated with interleukin signaling and
   immunomodulatory interactions within the innate immune system ([78]Fig.
   2, F and G). These findings indicate a potential link between
   macrophage infiltration in CNSLs and the activation of the chemokine
   pathway.

   Encouraged by the favorable bioinformatics analysis of clinical
   samples, we also collected tissue samples from patients with clinical
   neurological lymphoma for immunofluorescence analysis ([79]Fig. 2I).
   The results revealed a significant presence of intestinal
   macrophage-associated proteins, such as CD11c and CD14, in the CNSL
   tissue samples ([80]Fig. 2I). These proteins were highly expressed in
   CNSLs.

   Although CD14 is commonly used as a marker for intestinal macrophages,
   its expression is not exclusive and may also be detected in other
   immune cell types. To refine the identification of intestinal
   macrophages, we further interrogated publicly available datasets and
   used dataset [81]GSE125527, which incorporated single-cell RNA
   sequencing data (scRNA-seq) data of CD45^+ immune cells from both human
   peripheral blood and intestinal tissue ([82]Fig. 2, J to N).

   Three distinct macrophage subsets—clusters 5, 10, and 11—were
   identified from the intestine and peripheral blood immune cells
   ([83]Fig. 2, J and K). Although all three clusters were identified in
   both tissues, cluster 5 was predominantly found in Peripheral blood
   mononuclear cell (PBMC)–derived macrophages, whereas cluster 11 was
   more dominant in intestine-derived macrophages ([84]Fig. 2l).
   Transcriptomic profiling of these clusters revealed that certain genes,
   such as LYZ and LST1, were highly expressed in both cluster 5 and
   cluster 11. In contrast, certain marker genes were specifically highly
   expressed in one cluster: THBS1 showed low expression in cluster 11 but
   was markedly elevated in cluster 5, whereas HLA-DQA1 and HLA-DQB1 were
   low in cluster 5 but highly expressed in cluster 11([85]Fig. 2M).

   We then validated those marker gene expression in CNSL dataset. LYZ,
   LST1, HLA-DQA1, and HLA-DQB1 were all highly expressed in
   CNSL-associated macrophages, while THBS1 expression remained low
   ([86]Fig. 2N). This finding suggests a shared similarity between
   intestine-derived macrophages and CNSL-associated macrophages,
   supporting the hypothesis that intestinal macrophage cells may
   infiltrate or be recruited into the CNSL microenvironment. These
   findings suggest that intestinal macrophages—particularly those
   characterized by unique expression of THBS1, HLA-DQA1, and
   HLA-DQB1—have the capacity to migrate and function within the CNSL.
   This lays a theoretical foundation for exploring the potential of
   intestinal macrophages as carriers for gut-brain axis–targeted drug
   delivery in future therapeutic strategies.

   To establish an orthotopic CNSL model, we injected A20 lymphoma cells
   into mouse brains. Comparative analysis between the brains of these
   CNSL model mice and healthy controls revealed the presence of
   enteric-derived macrophages, marked by CD11c and CD64 (a key marker of
   mouse intestinal macrophages), within CNSL tissue ([87]Fig. 3, A to F).
   Furthermore, we identified elevated expression of chemokine
   receptor–associated genes in these intestinal macrophages, including
   Cx3cr1, Ccr1, Ccr2, and Ccr5 ([88]Fig. 3, G to J). These findings
   suggest that macrophages in CNSLs primarily originate from the gut and
   are recruited to the brain via chemokine receptor signaling (Cx3cr1,
   Ccr1, Ccr2, and Ccr5). In addition, CNSL tissue samples from mice were
   collected for immunofluorescence analysis ([89]Fig. 2, K and L). The
   results revealed a significant presence of intestinal
   macrophage-associated proteins, including CD11c and CD64, within the
   CNSL tissues ([90]Fig. 2K). Moreover, these intestinal macrophages
   exhibited elevated expression of chemokine receptor–related genes, such
   as Cx3cr1 ([91]Fig. 2L). This evidence lays a promising foundation for
   using intestinal macrophages to deliver nanomedicine, such as
   padlock-designed MOFs, to the brain as a potential therapeutic strategy
   for CNSLs.

Fig. 3. scRNA-seq analysis of mouse CNSL tissue and normal brain, along with
immunofluorescence findings in patients with CNSL.

   [92]Fig. 3.
   [93]Open in a new tab

   (A) Schematic representation of the experimental workflow. Some
   elements of this figure were created in BioRender (Yang-Bao Miao, 2025;
   [94]https://BioRender.com/ef5jegu). (B) UMAP plot for the analysis of
   cell subpopulations in the CNSL, which were divided into six cell
   subsets. (C) UMAP plot for the analysis of cell subpopulations in the
   normal mouse brain, which were divided into six cell subsets. (D)
   Comparative analysis of macrophage content between CNSL and normal
   brain tissues. (E and F) Expression of mesenteric macrophage markers in
   CNSL-associated macrophages, including CD11c (E) and CD64 (F). (G to J)
   Expression of chemokine receptor-related genes in CNSL-associated
   macrophages, including Cx3cr1 (G), Ccr1 (H), Ccr2 (I), and Ccr5 (J). (K
   and L) CLSM images showing high expression of enteric-derived
   macrophage markers (CD11c and CD64) (K) and the chemokine receptor
   Cx3cr1 (L) in CNSL tissue macrophages. (M) UMAP plots showing the
   distribution of cell subsets in CNSL tissue, blood, and intestine. (N)
   UMAP plots highlighting macrophage populations from CNSL tissue, blood,
   and intestine. (O) Principal components analysis (PCA) plots of
   macrophages identified by scRNA-seq and color-coded by tissue of origin
   (blood, CNSL, or intestine). (P) Heatmap depicting the intergroup
   distances between macrophages from blood, CNSL, and intestine. (Q)
   Violin plots showing the number of unique transcripts and genes
   detected per cell across the scRNA-seq dataset; each dot represents an
   individual cell. (R) Schematic illustrating the injection of
   flow-sorted intestinal macrophages into the intestine and their
   subsequent migration and homing to brain tumor sites in CNSL-bearing
   mice. Some elements of this figure were created in BioRender (Yang-Bao
   Miao, 2025; [95]https://BioRender.com/ef5jegu). (S) Schematic
   representation of the flow sorting and labeling of intestinal
   macrophages prior to injection. (T) Confocal microscopy images of CNSL
   tissue following immunohistochemical staining to detect transplanted
   intestinal macrophages. DAPI, 4′,6-diamidino-2-phenylindole; SSC, side
   scatter; FSC, forward scatter.

   To strengthen the evidence for the intestinal origin of macrophages in
   CNSL, we carried out scRNA-seq on additional CNSL tissue, mouse
   intestinal tissue, and peripheral blood samples ([96]Fig. 3M). Diverse
   immune cell populations, particularly macrophages, were identified
   across all samples. To delineate the relationship among macrophages
   from different tissues, we extracted macrophage subsets and conducted
   uniform manifold approximation and projection (UMAP) analysis ([97]Fig.
   3N), which revealed partial overlap between intestinal and CNSL-derived
   macrophages. Principal components analysis (PCA) ([98]Fig. 3O) further
   clarified this relationship, showing substantial overlap between brain-
   and intestine-derived macrophages, while blood-derived macrophages
   remained largely distinct.

   Quantitative assessment of cellular distances ([99]Fig. 3P)
   demonstrated that CNSL macrophages were more closely related to
   intestinal macrophages than to those from blood, suggesting a common or
   highly similar origin. Comparative analysis of surface markers
   ([100]Fig. 3Q) revealed that CD74 was ubiquitously expressed across
   macrophages from all tissues, Cd4 was predominantly restricted to blood
   macrophages, and Itgax (also known as CD11c) was highly expressed in
   both intestinal and CNSL macrophages. These findings indicate that
   intestinal macrophages expressing high levels of Itgax likely
   contribute to the macrophage population within CNSL lesions.

   To functionally validate the migratory potential of intestinal
   macrophages, we isolated CD45^+ CD11c^+ macrophages from the intestine
   via flow cytometry and surgically injected them into the intestinal
   lumen of recipient mice ([101]Fig. 3, R and S). Notably, within 4 hours
   postinjection, labeled macrophages were detected in the brain
   ([102]Fig. 3T), providing direct evidence that intestinal macrophages
   can traffic to the CNS. Together, these results support the concept
   that intestinal macrophages contribute to CNSL macrophage populations
   and establish a mechanistic basis for exploiting gut-derived
   macrophages in gut-brain axis–targeted therapies

Characteristics of Pt-MOFs@glu

   To attain the desired attributes for effective cancer therapy, it is
   crucial to regulate the dimensions of nanoparticles. In this study, we
   synthesized padlock-designed Pt-MOFs through a hydrothermal process
   that involves carboplatin and pyrazinequinoxaline, with the specific
   synthesis procedure detailed in [103]Fig. 4A. The particle size was
   meticulously controlled by varying the reaction time. For enhanced
   tumor tissue penetration, the padlock-designed Pt-MOFs were synthesized
   at 180°C for 2 hours, resulting in an appropriate particle size of 130
   to 150 nm ([104]Fig. 4B). It is common knowledge that diminutive
   nanoparticles are more readily internalized by cancer cells owing to
   their size and surface characteristics. Furthermore, tumor targeting
   can be accomplished by utilizing nanoparticles within the range of 50
   to 200 nm. ([105]33)

Fig. 4. Characteristics of as-synthesized Pt-MOFs@Glu.

   [106]Fig. 4.
   [107]Open in a new tab

   (A) Composition and structure: Details of the composition and structure
   of the proposed Pt-MOFs@Glu. Pt-MOFs were formed from carboplatin and
   pyrazine-quinoxaline by solvothermal and became ROS-responsive
   drug-releasing nanoparticles by linking glucan. (B) Scanning electron
   microscopy (SEM) image of Pt-MOFs: SEM image depicting the morphology
   of Pt-MOFs. (C) Ultraviolet-visible (UV-Vis) spectra: UV-Vis spectra
   comparison of carboplatin, Pt-MOFs, and Pt-MOFs@Glu. (D and E) Spectral
   analysis: UV-Vis spectra (D) and mass spectrometry (MS) data (E) of
   Pt-MOFs@Glu after simulating the TME. (F) XPS analysis: X-ray
   photoelectron spectroscopy (XPS) data of Pt-MOFs@Glu, both pre- and
   postsimulation of the TME. (G) Fourier-transform infrared (FTIR)
   spectra: FTIR spectra of pyrazino[2,3-f] quinoxaline, carboplatin,
   Pt-MOFs, glucan, and Pt-MOFs@Glu. (H) SEM image of Pt-MOFs@Glu:
   Detailed SEM image of Pt-MOFs@Glu showcasing its surface
   characteristics. (I) TEM image of Pt-MOFs@Glu: TEM image of Pt-MOFs@Glu
   for a closer view of its internal structure. (J) Elemental mapping:
   Elemental mapping images of Pt-MOFs@Glu nanoparticles, captured using
   aberration-corrected high angle annular dark field scanning
   transmission electron microscope (AC-HAADF-STEM), providing a
   comprehensive elemental distribution analysis. (K) X-ray diffraction
   (XRD) patterns: XRD patterns of Pt-MOFs@Glu. (L) TEM images of
   Pt-MOFs@Glu at various time points (0, 10, 20, and 30 min) under ROS
   conditions within the TME. a.u., absorbance unit; m/z, mass/charge
   ratio.

   To achieve uniform nanoparticles with consistent size and morphology,
   we selected a feed ratio of 1:1 for the synthesis process using dynamic
   light scattering (DLS) detection. The optimized nanoparticles exhibited
   a uniform spherical shape ([108]Fig. 4B) with a subtly uneven surface,
   boasting a particle size of 180.06 ± 10.52 nm and a zeta potential of
   −29.48 ± 1.16 mV (n = 3; fig. S2). The morphology of the
   padlock-designed Pt-MOFs remained virtually unaltered even after 1
   month of storage in deionized water at 4°C, underscoring its remarkable
   stability during storage (fig. S3).

   To validate the efficacy of padlock-designed Pt-MOFs in harnessing the
   TME to induce the release of carboplatin, a comprehensive suite of
   analytical techniques was deployed, including DLS, ultraviolet
   (UV)–visible spectroscopy, mass spectrometry, and x-ray photoelectron
   spectroscopy (XPS).

   DLS analysis, as presented in fig. S4, revealed a notable increase in
   the particle size of padlock-designed Pt-MOFs within the TME.
   UV-visible spectroscopy results, shown in [109]Fig. 4C, demonstrated
   the absence of a characteristic absorption peak for carboplatin at 229
   nm in a normal physiological conditions [phosphate-buffered saline
   (PBS) = 7.4] solution. This is attributable to the conjugated double
   bonds in pyrazine-quinoxaline, which produce a strong absorption peak
   at this wavelength, thereby giving pyrazine-quinoxaline their
   ultraviolet absorption properties. Notably, under normal conditions,
   there was minimal carboplatin release (fig. S5). In contrast, exposure
   to the TME, characterized by elevated hydrogen peroxide (H[2]O[2])
   levels, led to the emergence of a prominent absorption peak for
   carboplatin at 229 nm, as illustrated in [110]Fig. 4D. In addition,
   evaluating a range of ROS concentrations in vitro provides a
   comprehensive assessment of the stability and responsiveness of
   Pt-MOFs@Glu under varying oxidative stress conditions (fig. S6). Our
   results demonstrate that higher ROS levels notably enhance the
   avalanche effect of Pt-MOFs@Glu, resulting in a more efficient and
   controlled release of carboplatin. Conversely, at lower ROS
   concentrations, Pt-MOFs@Glu shows limited capability to release
   carboplatin. This demonstrates a substantial release of carboplatin
   under tumor-specific conditions. The findings are further supported by
   mass spectrometry analysis and quantitative carboplatin release data,
   as illustrated in [111]Fig. 4E and fig. S7, providing robust
   confirmatory evidence.

   Concurrently, XPS was used to probe for any potential changes in the
   valence state of carboplatin within the TME ([112]Fig. 4F). The
   analysis indicated that the valence state change of divalent platinum
   (Pt) in the MOFs, designed with a platinum padlock, remained negligible
   before and after exposure to simulated TME conditions (fig. S8). This
   confirms the stability of carboplatin in these circumstances. The
   aforementioned findings comprehensively illustrate the efficacy and
   stability of the carboplatin/pyrazine-quinoxaline nanocomposite in
   facilitating the controlled release of carboplatin within the TME.

   The desired nanocomplex has been successfully obtained, yet a challenge
   remains in directing it to the brain via oral administration. In our
   prior research, we found that oral administration of
   β-glucan–conjugated drugs effectively targets intestinal macrophages
   via the intestinal epithelial barrier (IEB), capitalizing on the tumor
   homing effect to traverse the BBB and reach the site of brain tumors.
   ([113]14) To facilitate oral brain targeting of padlock-designed
   Pt-MOFs, we covalently linked them with thioethylated β-glucan and
   characterized their structure using Fourier transform infrared (FTIR)
   spectroscopy. The structural characteristics of the synthesized
   Pt-MOFs@Glu were analyzed using FTIR spectroscopy, which revealed a
   distinct absorption peak at 810 cm^−1, corresponding to the H-N-H
   stretching mode associated with cisplatin ([114]Fig. 4G). Notably, the
   spectrum of the synthesized padlock-designed MOFs exhibited a
   characteristic absorption peak of the drug at a similar wavenumber,
   along with a sulfhydryl absorption peak absent in the drug and ligand
   alone, signifying the successful synthesis of glucan padlock-designed
   Pt-MOFs (i.e., Pt-MOFs@Glu).

   Simultaneously, SEM revealed that padlock-designed Pt-MOFs exhibited a
   smoother surface upon glucan coating ([115]Fig. 4H). Moreover, the
   shell-core structure was observed via transmission electron microscopy
   (TEM; [116]Fig. 4I). The elemental profiling of STEM samples of
   presynthesized glucan-modified nanoparticles was conducted using
   energy-dispersive X-ray spectroscopy (EDS), indicating the presence of
   Pt, C, O, N, and S elements ([117]Fig. 4J). The nearly exclusive
   derivation of S elements from glucan further suggests the effective
   grafting of glucan onto padlock-designed Pt-MOFs. To confirm the
   crystal structure of padlock-designed Pt-MOFs and Pt-MOFs@Glu, PXRD
   analysis was conducted. [118]Figure 4K and fig. S9 demonstrate that the
   formation of Pt-MOFs@Glu, derived from glucan-modified padlock-designed
   Pt-MOFs, does not alter the crystal structure of padlock-designed
   Pt-MOFs.

   To further confirm its therapeutic potential, Pt-MOFs@Glu was tested
   for carboplatin release under ROS conditions within the TME, leveraging
   an avalanche effect. TEM image observations, as shown in [119]Fig. 4l,
   revealed that Pt-MOFs@Glu rapidly disintegrates under these ROS
   conditions, validating its ability to swiftly release carboplatin.
   These findings support the application of Pt-MOFs@Glu in treating CNSL
   by enabling accelerated carboplatin release through the avalanche
   effect triggered by TME ROS.

Cellular uptake in Pt-MOFs@Glu and cytotoxicity

   To achieve phagocytosis by intestinal macrophages, facilitating
   subsequent entry into brain tumor sites, efficient absorption of
   padlock-designed MOFs by these macrophages is imperative. The mouse
   macrophage cell line (RAW264.7) displays a notable ability to recognize
   trehalose via its Dectin-1 receptor ([120]Fig. 5A and fig. S10). To
   assess the uptake of Pt-MOFs@Glu by RAW264.7, these padlock-designed
   MOFs were prelabeled with Cy5.5. Laser confocal microscopy revealed a
   rapid uptake of Pt-MOFs@Glu by RAW264.7, peaking at 3 hours ([121]Fig.
   5B).

Fig. 5. In vitro characteristics of Pt-MOFs@Glu.

   [122]Fig. 5.
   [123]Open in a new tab

   (A) Transport mechanism: Pt-MOFs@Glu are transported by macrophages
   across the BBB to the brain tumor. Some elements of this figure were
   created in BioRender (Yang-Bao Miao, 2025;
   [124]https://BioRender.com/ef5jegu). (B) Cellular uptake: CLSM images
   showing Pt-MOFs@Glu uptake by RAW264.7 macrophages. (C) Cytotoxicity to
   macrophages: Evaluation of the cytotoxicity of free carboplatin and
   Pt-MOFs@Glu on RAW264.7 macrophages. (D) Cytotoxicity to A20 cells:
   Assessment of the cytotoxicity of free carboplatin and Pt-MOFs@Glu on
   A20 lymphoma cells. (E) Tumor spheroid penetration: CLSM images
   demonstrating the penetration of free Pt-MOFs@Glu and
   macrophage-hitchhiked Pt-MOFs@Glu into lymphoma spheroids. * denotes
   statistically significant (P < 0.05).

   Contrastingly, the pretreatment of RAW264.7 with a Dectin-1 receptor
   inhibitor significantly diminished the uptake of Pt-MOFs@Glu (fig.
   S10). These findings strongly imply that the efficient absorption of
   padlock-designed MOFs by macrophages is facilitated by the Dectin-1
   receptor. Furthermore, as proficient phagocytes, intestinal macrophages
   exhibit a remarkable ability to engulf a substantial number of
   padlock-designed MOFs. This underscores their potential as carriers for
   targeted CNSL drug delivery.

   Evaluating the cytotoxicity of Pt-MOFs@Glu on intestinal macrophages is
   paramount to ensure their biosafety during phagocytosis and subsequent
   transport to CNSL sites. Various concentrations of Pt-MOFs@Glu were
   incubated with intestinal macrophages, and cell viability was assessed
   using the methylthiazolyldiphenyl-tetrazolium bromide (MTT) method
   ([125]Fig. 5C). A comparable concentration of free drug (carboplatin)
   served as the control. The macrophage viability decreased significantly
   in a dose-dependent manner in the presence of free drugs. Conversely,
   macrophages treated with Pt-MOFs@Glu exhibited heightened activity,
   indicating superior biosafety during transportation.

   To further elucidate the in vitro cytotoxicity of Pt-MOFs@Glu against
   CNSL cells, diverse concentrations of these MOFs were incubated with
   the cells ([126]Fig. 5D). As a control, an equivalent concentration of
   free drug (platinum) was used. Following the cell viability experiment,
   MTT testing was conducted. Strikingly, Pt-MOFs@Glu demonstrated
   enhanced cytotoxicity against CNSL compared to free drugs. These
   findings suggest that Pt-MOFs@Glu exhibits excellent biosafety during
   transit. Once internalized by tumor cells, it can activate carboplatin
   through ROS triggers, initiating an avalanche effect that significantly
   amplifies tumor cell death ([127]Fig. 4L).

   Macrophages demonstrate the remarkable ability to infiltrate tumor
   tissues, facilitated by a cascade of chemical mediators emanating from
   hypoxic tumor environments. To assess the tumor tissue penetration of
   macrophage-conveyed Pt-MOFs@Glu, we established a tumor cell sphere
   model. In this model, free Pt-MOFs@Glu served as the control,
   predominantly adhering to the outer layer of the cell sphere upon
   co-incubation. Conversely, macrophage-facilitated Pt-MOFs@Glu exhibited
   superior penetration into the tumor cell sphere. Notably, over time,
   the Pt-MOFs@Glu gradually permeated the inner core of the cell sphere,
   as depicted in [128]Fig. 5E. These findings underscore the capacity of
   intestinal macrophages to enhance the diffusion of Pt-MOFs@Glu into the
   hypoxic regions of tumors, thereby addressing the limitations
   associated with conventional nanoparticles.

Pt-MOFs@Glu stability assessment

   As a pivotal component of oral drug delivery, Pt-MOFs@Glu must endure
   the rigors of the GI with unwavering stability. To scrutinize this
   critical aspect, in vitro experiments were conducted, wherein the
   Pt-MOFs@Glu underwent immersion in simulated gastric fluid (SGF; pH
   2.0) and simulated intestinal fluid (SIF; pH 7.0) at 37°C. As depicted
   in the fig. S11, the morphology and particle size of the Pt-MOFs@Glu
   following exposure to SGF and SIF closely mirrored those of the
   untreated counterparts. This compelling observation signifies that even
   amid the harsh milieu of the GI, the Pt-MOFs@Glu maintained their
   structural integrity, remaining steadfast and unscathed.

   Furthermore, an in vitro simulation of the TME was meticulously
   conducted using hydrogen peroxide, with glutathione and physiological
   conditions (PBS 7.0) serving as the control ([129]Fig. 4E). The results
   indicated that Pt-MOFs@Glu, in the presence of hydrogen peroxide,
   triggered an avalanche effect that rapidly released carboplatin. In
   contrast, it remained inert when exposed to glutathione (fig. S12).
   This behavior underscores the Pt-MOFs@Glu remarkable ability to
   selectively respond to hydrogen peroxide within the TME, thereby
   facilitating the targeted release of carboplatin—a pivotal advancement
   in tumor treatment strategies.

Pathways of absorption in Pt-MOFs@glu

   In this study, macrophage recruitment plays a crucial role in the
   therapeutic approach using padlock-designed MOFs for CNSL. As shown in
   [130]Fig. 2, both bioinformatics analyses and examination of two
   additional clinical CNSL samples revealed significantly higher
   macrophage infiltration in patients with CNSL compared to normal human
   tissues. This elevated macrophage presence was also observed in a CNSL
   mouse model ([131]Fig. 3). These findings underscore the potential of
   using orally administered padlock-designed MOFs, which can be
   selectively taken up by macrophages and transported to the brain, as a
   promising treatment strategy for CNSL.

   To elucidate the intricate pathway of f-Pt-MOFs@Glu from the intestine
   to tumors, we meticulously examined their biodistribution following
   oral administration. Using euthanized mice, we extracted and processed
   both villi and Peyer’s patches from the GI, alongside CNSL specimens.
   These tissues were meticulously examined ex vivo using confocal laser
   scanning microscopy (CLSM) to delineate the distribution of
   f-Pt-MOFs@Glu, thereby unraveling the intricate nanoparticle gut-brain
   axis delivery pathways.

   Our findings unveiled a targeted affinity of Pt-MOFs@Glu toward M cells
   within Peyer’s patches ([132]Fig. 6A). Subsequently, these
   nanoparticles traversed into the villi, where they were efficiently
   engulfed by local macrophages ([133]Fig. 6B). These macrophages, acting
   as efficient carriers, facilitated the transportation of nanoparticles
   through the lymphatic ducts ([134]Fig. 6C), thus providing a conduit
   into the systemic circulation. Ultimately, these nanoparticles breached
   the BBB, accumulating at the site of CNSL ([135]Fig. 6, D and E).

Fig. 6. The transport route of Pt-MOFs@Glu.

   [136]Fig. 6.
   [137]Open in a new tab

   (A to C) Colocalization in the intestinal tract: Schematic depictions
   and CLSM images showing the colocalization of Pt-MOFs@Glu with M cells
   (A), lymphatic vessels (B), and resident macrophages (C) in the
   intestinal tract. (D) Transport to brain tumors: Pt-MOFs@Glu are
   carried by macrophages into brain tumors. Some elements of this figure
   were created in BioRender (Yang-Bao Miao, 2025;
   [138]https://BioRender.com/ef5jegu). (E and F) Brain tumor imaging:
   CLSM images of brain tumors without (E) and with macrophage blockers
   (F). (G) Ex vivo IVIS imaging: Ex vivo images from the IVIS showing the
   accumulation of Pt-MOFs@Glu. (H) Real-time IVIS imaging: Real-time IVIS
   images depicting the accumulation of Pt-MOFs@Glu in brain tumors at
   specified times following oral administration in mice. (I) Radiant
   efficiency analysis: Corresponding total radiant efficiencies from ex
   vivo IVIS images showing the accumulation of Pt-MOFs@Glu. (J)
   Carboplatin content in brain tissue: Measurement of carboplatin levels
   in brain tissue across different groups: Control, carboplatin, Pt-MOFs
   + glucan, Pt-MOFs, and Pt-MOFs@Glu.

   Moreover, our study used intravenous injection of chlorophosphate
   liposomes to induce macrophage depletion, thereby inhibiting the uptake
   of Pt-MOFs@Glu by macrophages. This intervention served to corroborate
   the absorption pathway of Pt-MOFs@Glu (fig. S13). The absence of
   fluorescence observed in the brain ([139]Fig. 6F and fig. S14)
   confirmed that Pt-MOFs@Glu, upon engulfment by macrophages, indeed
   traversed through the lymphatic system into the bloodstream.
   Consequently, they successfully crossed the BBB, thereby penetrating
   the brain to combat CNSL.

   To further elucidate the mechanism governing macrophage migration, we
   investigated the role of the chemokine receptor Cx3Cr1, a key regulator
   of macrophage trafficking. To selectively inhibit Cx3Cr1 in intestinal
   macrophages, we administered JMS-17, a specific Cx3Cr1 inhibitor, to
   CNSL model mice. Following this intervention, we evaluated the targeted
   delivery of Pt-MOFs@Glu to brain tumor sites (fig. S14). Our findings
   revealed that Cx3Cr1 inhibition completely abrogated the accumulation
   of Pt-MOFs@Glu at the tumor site, confirming the critical role of
   Cx3Cr1 in mediating macrophage homing and recruitment from the
   intestine to CNSL. These results provide compelling evidence that CNSLs
   harbor a substantial population of gut-derived macrophages, which
   infiltrate the TME through Cx3Cr1-dependent pathways. This study
   underscores the significance of intestinal-immune interactions in CNSL
   pathogenesis and identifies Cx3Cr1 as a promising therapeutic target
   for modulating macrophage infiltration.

In vivo distribution analysis of Pt-MOFs@Glu

   We explored the potential of Pt-MOFs@Glu as a gut-brain axis
   nanodelivery system using an orthotopic CNSL mouse model. The
   experimental group received orally administered Pt-MOFs@Glu, while the
   control group received oral deionized water. Using an in vivo imaging
   system (IVIS), we tracked the distribution of Pt-MOFs@Glu
   postadministration ([140]Fig. 6, G to I). Notably, no fluorescent
   signals were detected in brain tissues of the control group. In
   contrast, in the study, CNSL mice that were orally administered the
   platinum padlock nanodrug exhibited detectable fluorescence signals in
   the brain as early as 4 hours postadministration. Ex vivo imaging of
   tissues was performed 24 hours postadministration, with quantification
   of fluorescence intensity in each tissue ([141]Fig. 6G). These signals
   progressively intensified, culminating in peak fluorescence levels at
   12 hours. Consequently, 12 hours postadministration was selected as the
   optimal time point for subsequent in vitro brain experiments. Ex vivo
   fluorescence analysis revealed the predominant accumulation of
   Pt-MOFs@Glu in brain tumors, liver, and kidneys following oral
   administration. These findings suggest that Pt-MOFs@Glu can surmount
   the formidable barriers to oral drug delivery, facilitated by
   intestinal macrophages traversing the IEB and the BBB, ultimately
   reaching the site of CNSL.

   To precisely assess the delivery efficiency of carboplatin in
   Pt-MOFs@Glu, we used inductively coupled plasma mass spectrometry
   (ICP-MS) to analyze carboplatin concentrations in CNSL. This analysis
   aimed to demonstrate the efficacy of carboplatin drugs in reaching CNSL
   following oral administration ([142]Fig. 6J). The experimental group
   received Pt-MOFs@Glu orally, while the control group received an
   equivalent dosage of free cisplatin. Results revealed that the
   concentration of Pt-MOFs@Glu at the brain tumor site in the
   experimental group exceeded that of the control group by 15-fold. This
   unequivocally confirms the capacity of Pt-MOFs@Glu to effectively
   leverage intestinal macrophages for targeted delivery to the CNSL site.

Antitumor effectiveness

   Motivated by the exceptional performance demonstrated by Pt-MOFs@Glu,
   we proceeded to assess their anticancer efficacy in vivo. Primary CNSL
   was meticulously established through intracranial injection of A20
   lymphoma cells into Balb/c mice. The resultant CNSLs were subsequently
   stratified into five distinct experimental groups for comprehensive
   investigation: an untreated control group receiving deionized water, a
   group receiving free carboplatin, a group with carboplatin equivalent
   to Pt-MOFs, a group administered free glucan, and a final group treated
   with Pt-MOFs@Glu ([143]Fig. 7A).

Fig. 7. Antitumor efficacy of Pt-MOFs@Glu in mice bearing CNSL.

   [144]Fig. 7.
   [145]Open in a new tab

   (A) Study design: Schematic representation of the time course for the
   establishment of an orthotopic brain tumor model in mice and the
   treatment regimen. Treatment was administered on day 7 after modeling
   and the trial cycle ended on day 35. Some elements of this figure were
   created in BioRender (Yang-Bao Miao, 2025;
   [146]https://BioRender.com/ef5jegu). (B) Health metrics: Body weights
   and survival rates of mice following various treatments. (C) Brain
   pathology: Pathological sections of brain tissue from mice treated for
   CNSL. (D and E) Tumor analysis: TUNEL and Ki67 staining of tumor
   tissues showing pathological changes across different treatment groups.
   * denotes statistically significant (P < 0.05).

   Throughout the treatment period, we meticulously documented changes in
   body weight ([147]Fig. 7B) and survival rates ([148]Fig. 7B) across all
   groups. Upon completion of the experiment, the tumors were excised and
   subjected to hematoxylin and eosin (H&E) staining to evaluate the
   changes in tumor size posttreatment, as depicted in [149]Fig. 7C. While
   treatment with free carboplatin and free glucan exhibited some degree
   of antitumor activity, it was accompanied by significant reductions in
   body weight and poor survival rates, indicative of insufficient tumor
   growth inhibition.

   Notably, Pt-MOFs@Glu, when orally administered, traverses the body via
   M cells, undergoes phagocytosis, and is then transported by macrophages
   into the brain, where they accumulate at the CNSL site. Subsequently,
   the substantial quantity of ROS present in CNSL instigates an avalanche
   effect within the Pt-MOFs@Glu, resulting in the release of a
   significant amount of carboplatin, which effectively eradicates the
   tumor cells. Pt-MOFs@Glu therapy demonstrated significantly heightened
   antitumor efficacy compared to the control group, as evidenced by
   notable improvements in body weight, survival rates, and reductions in
   both tumor volume (P < 0.05). Furthermore, the oral administration of
   Pt-MOFs@Glu exhibited superior tumor cell eradication (H&E staining;
   [150]Fig. 7C), an increased presence of apoptotic cells [terminal
   deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick
   end labeling (TUNEL) measurement; [151]Fig. 7D], and a lower proportion
   of proliferating cells (Ki67 staining; [152]Fig. 7E) compared to other
   control groups. These findings underscore the potency of Pt-MOFs@Glu,
   featuring multiple platinum padlocks, in effectively combating CNSL,
   offering promising prospects in cancer therapeutics.

   Inspired by the exceptional efficacy of Pt-MOFs@Glu in the treatment of
   CNSL in vivo, we sought to elucidate the underlying mechanisms
   contributing to its remarkable performance. We hypothesized that the
   effectiveness of Pt-MOFs@Glu stems from its dual action: The release of
   a substantial amount of carboplatin through an ROS-induced avalanche
   effect and the release of glucan, a potent immune adjuvant, which has
   been previously validated in cellular experiments (as shown in the fig.
   S15). Glucan is known to activate the body’s immune response, thereby
   aiding in the eradication of tumors and preventing recurrence
   postchemotherapy ([153]31). To test this hypothesis, we assessed the
   immune activation potential of Pt-MOFs@Glu in a CNSL mouse model
   ([154]Fig. 8A). Compared to the untreated CNSL control group, mice
   treated with Pt-MOFs@Glu for 2 weeks exhibited a marked increase in
   brain-infiltrating immune cells, including CD4^+ T cells ([155]Fig.
   8B), CD8^+ T cells ([156]Fig. 8C), natural killer (NK) T cells (fig.
   S16), and effector T cells (fig. S17).

Fig. 8. Pt-MOFs@Glu prevents relapse postchemotherapy through immune
activation.

   [157]Fig. 8.
   [158]Open in a new tab

   (A) Immune activation mechanism: Pt-MOFs@Glu promotes the maturation of
   dendritic cells (DCs) and subsequently acts on lymph nodes to activate
   T cells. This activation leads to tumor cell apoptosis and suppression
   of tumor growth. Some elements of this figure were created in BioRender
   (Yang-Bao Miao, 2025; [159]https://BioRender.com/ef5jegu). (B and C) T
   cell analysis: Evaluation of intratumoral CD4^+ and CD8^+ T cell
   populations in CNSL mice following treatment with different therapeutic
   strategies. (D to G) Cytokine and immune marker analysis:
   Quantification of immune-related markers and cytokines in tumor
   tissues, including HMGB1 (high mobility group box 1) (D), CRT
   (calreticulin) (E), IL-12 (interleukin-12) (F), and TNF-α (tumor
   necrosis factor–α) (G) using ELISA kits. * denotes statistically
   significant (P < 0.05).

   To further substantiate our hypothesis regarding the immune activation
   mechanism induced by Pt-MOFs@Glu, we used immunofluorescence analysis.
   We conducted a comprehensive examination of high mobility group box 1
   (HMGB1) ([160]Fig. 8D) and calreticulin (CRT) ([161]Fig. 8E) of DAMP
   proteins in the brains of CNSL mice subjected to various treatments:
   Pt-MOFs@Glu, Pt-MOFs with equivalent carboplatin content, free glucan,
   carboplatin, and untreated control groups.

   Our findings demonstrate that Pt-MOFs@Glu significantly up-regulates
   CRT, HMGB1, and other DAMP proteins in CNSL cells. Additional analysis
   explored whether the immunoadjuvant glucan, released upon tumor cell
   apoptosis via the avalanche effect, enhances the release of
   immunoinflammatory mediators, such as interferon-α (IFN-α) and
   interleukin-12 (IL-12), which collectively promote immune activation
   ([162]Fig. 8, F and G). These results indicate that Pt-MOFs@Glu
   treatment effectively stimulates the release of key immunoinflammatory
   mediators following tumor cell death.

   The preceding investigation used a primary CNSL mouse model; however,
   clinically, the majority of CNSL cases are secondary. Thus, we
   established a secondary CNSL model for further study ([163]Fig. 9A).
   This model involves subcutaneous lymphoma establishment followed by
   inoculation of secondary CNSL into the mice’s brains on the 14th day
   postestablishment. The results presented in [164]Fig. 9 (B to D)
   demonstrate that Pt-MOFs@Glu therapy exhibited superior antitumor
   efficacy in secondary CNSL compared to the control group. Notably,
   improvements in body weight, survival rate, and tumor volume were
   statistically significant (P < 0.05).

Fig. 9. Efficacy of Pt-MOFs@Glu in the treatment of secondary CNSL.

   [165]Fig. 9.
   [166]Open in a new tab

   (A) Study design: Schematic representation of the time course for
   establishing a brain tumor model in mice and the treatment regimen.
   Treatment was performed on the fourth day after initial tumor
   inoculation, and secondary tumour modelling was performed on the 14th
   day, ending the experiment on the 21st day. Some elements of this
   figure were created in BioRender (Yang-Bao Miao, 2025;
   [167]https://BioRender.com/ef5jegu). (B) Measurement of mouse weight
   and survival rate: Changes in mouse weight and survival rate over time
   were observed following various treatments. (C) Photographs of resected
   tumors with H&E staining, along with (D) a quantitative chart
   illustrating the corresponding tumor volumes. (E and F)
   Histopathological analysis: TUNEL and Ki67 staining revealed
   pathological changes in tumor tissues across different treatment
   groups. (G) T cell analysis: Assessment of intratumoral CD4^+ and CD8^+
   T cell populations in CNSL mice following various therapeutic
   interventions. * denotes statistically significant (P < 0.05).

   Moreover, the oral administration of Pt-MOFs@Glu significantly inhibits
   the growth of CNSL tumor cells ([168]Fig. 9, C and D), a reduced
   proportion of proliferating cells (as indicated by Ki67 staining,
   [169]Fig. 9E), and an increase in apoptotic cells (as observed in TUNEL
   measurements, [170]Fig. 9F), in contrast to other control groups. These
   findings underscore the potent efficacy of Pt-MOFs@Glu in effectively
   combatting CNSL. This promising efficacy holds notable potential for
   advancing cancer treatment strategies.

   Simultaneously, we validated the immune activation potential of
   Pt-MOFs@Glu in a secondary CNSL model. Notably, mice treated with
   Pt-MOFs@Glu for 2 weeks exhibited a significant increase in
   brain-infiltrating CD4^+ ([171]Fig. 9G) and CD8^+ T cells ([172]Fig.
   9G) compared to the untreated CNSL control group. This result
   demonstrates that Pt-MOFs@Glu can effectively release carboplatin to
   eradicate tumors while simultaneously activating the immune system
   through the action of glucan and DAMP proteins ([173]Fig. 8, D and E).
   This dual mechanism of action substantially inhibits secondary CNSL.

Unveiling the mechanisms behind the antitumor potential of Pt-MOFs@Glu

   To investigate the mechanism by which oral carboplatin lock-designed
   MOFs (Pt-MOFs@Glu) triggers an avalanche effect to promote CNSL cell
   death in the CNSL microenvironment via the gut-brain axis, we collected
   tissue samples from both treated (post-oral Pt-MOFs@Glu) and untreated
   CNSL cases. RNA was extracted from these samples for analysis. Venn
   analysis of gene expression (transcripts per million ≥ 1) revealed
   57,180 coexpressed genes across the three groups. The data indicated
   that, following Pt-MOFs@Glu intervention, 1283 death-related genes were
   up-regulated and 1451 were down-regulated ([174]Fig. 10, A and B). A
   total of 2734 differential genes (both up- and down-regulated) were
   selected to identify the signaling pathways through which oral
   Pt-MOFs@Glu induces CNSL cell death.

Fig. 10. Mechanism behind the antitumor potential of Pt-MOFs@Glu.

   [175]Fig. 10.
   [176]Open in a new tab

   (A) Thermogram of differentially expressed gene expression before and
   after treatment with Pt-MOFs@Glu. (B) Volcano plots highlighting
   significantly up-regulated (red) and down-regulated (blue) genes. (C)
   KEGG pathway enrichment analysis. (D) GO enrichment analysis, where the
   x axis represents the enrichment fraction, and the y axis corresponds
   to the pathway information. Larger bubbles indicate greater
   differences, while the color gradient from blue to red signifies a
   decreasing P value, indicating higher statistical significance. (E)
   Differential gene GO enrichment analysis between control and treatment
   groups. (F) Network mapping of functionally linked, differentially
   expressed antitumor potential associated proteins. (G) Diagram
   illustrating the antitumor mechanism based on gene sequencing results.
   JAK, Janus kinase; STAT, signal transducers and activators of
   transcription; PI3K, phosphatidylinositol 3-kinase.

   The Pt-MOFs@Glu promoted apoptosis in CNSL cells, with most of the
   related genes associated with immune function ([177]Fig. 10C). The
   results indirectly indicate that Pt-MOFs@Glu, traveling from the gut to
   the brain, initiates an avalanche effect in the CNSL microenvironment
   via ROS, thereby facilitating the targeted release of carboplatin to
   CNSL cells. The release of carboplatin directly kills CNSL cells,
   triggering the secretion of tumor necrosis factor. This binds with
   glucan, an immune adjuvant released during the avalanche effect, to
   activate a tumor-targeted immune response. Together, these actions
   robustly enhance the immune response against CNSL. KEGG pathway
   analysis revealed that many cross-signaling genes are involved in
   pathways related to tumor apoptosis and T cell activation ([178]Fig.
   10, D and E). These genes are enriched in apoptosis pathways, the p53
   signaling pathway, T helper 1 (T[H]1) and T[H]2 cell differentiation,
   and the IL-17 signaling pathway, underscoring the role of Pt-MOFs@Glu
   in inducing tumor cell apoptosis, inflammation, and immune activation
   following oral therapy for CNSL.

   Further analysis of protein-protein interactions identified 11 key hub
   proteins encoded by 9 up-regulated and 2 down-regulated genes, all
   implicated in critical antitumor processes ([179]Fig. 10E). Tyrosine
   protein kinases, in particular, were associated with 37 proteins
   encoded by differential genes and play an essential role in
   phosphorylating tyrosine residues in key signaling molecules. These
   kinases facilitate tumor cell apoptosis and stimulate antigen release,
   thereby activating the immune response against tumor cells. In
   addition, the network analysis of differentially expressed,
   functionally associated proteins revealed that Pt-MOFs@Glu primarily
   exerted its antitumor activity through mitochondrial
   dysfunction-induced oxidative stress ([180]Fig. 10F). Central nodes in
   the interaction network—including mt-Co1, mt-Co2, mt-Nd1, mt-Nd2,
   Ndufa1, Ndufb3, and Uqcrc1—were predominantly enriched in mitochondrial
   electron transport chain components. This suggests that Pt-MOFs@Glu
   substantially impaired mitochondrial function, leading to excessive
   generation of ROS, which in turn initiated intrinsic apoptosis via
   cytochrome c release and activation of proapoptotic proteins such as
   Bax and Bak.

   Concomitantly, Pt-MOFs@Glu triggered extensive immunomodulatory effects
   within the TME. The up-regulation of immune signaling regulators—such
   as Foxo3, Il17d, Jak3, Tyk2, Eif5a2, and Gbp6—indicated activation of
   IFN-related pathways, T[H]17 differentiation, and the Janus
   kinase/signal transducers and activators of transcription
   axis—hallmarks of enhanced antitumor immune responses. The engagement
   of immune-enhancing chemokines and cytokine regulators—including Ccl5,
   Cxcl, Wnt6, and Ptpn11—further supports the recruitment and activation
   of cytotoxic T lymphocytes and NK cells, thereby augmenting tumor
   immunosurveillance.

   Moreover, Pt-MOFs@Glu modulated key immunological and structural
   proteins associated with cell adhesion and immune evasion. The
   down-regulation of Spp1 and disruption of Lam5-integrin signaling
   contributed to the loss of cell-matrix interactions, promoting
   apoptosis through anoikis. Simultaneously, the presence of
   apoptosis-regulatory factors such as Trp^73 and Tp^53 within the
   network, alongside changes in metabolic regulators (e.g., Hmgcs1,
   Acat2, and Cyp51), suggested that Pt-MOFs@Glu reprogrammed tumor cell
   metabolism and reinforced proapoptotic signaling. These findings
   underscore the dual therapeutic capacity of Pt-MOFs@Glu in mediating
   both direct cytotoxicity through mitochondrial damage and ROS
   generation and indirect antitumor effects via immune activation and TME
   remodeling. This comprehensive mechanism highlights the effectiveness
   of Pt-MOFs@Glu in inducing an avalanche effect, where carboplatin
   release synergizes with the immunoadjuvant action of glucan to
   effectively eradicate CNSL.

Biosafety evaluation of Pt-MOFs@Glu

   To assess the potential toxicity of Pt-MOFs@Glu, a comprehensive series
   of in vivo experiments were conducted, including examination of tissue
   sections from major organs, blood chemistry analysis, and hematological
   analysis. Methotrexate, a first-line clinical drug, was used as a
   positive control. As shown in figs. S18 to S20, the results revealed
   that methotrexate treatment induced significant pathological changes in
   the liver and kidneys of the mice, including cellular necrosis and
   inflammation, consistent with the well-established toxic effects of the
   drug. Biochemical analysis revealed a notable increase in blood urea
   nitrogen (BUN) levels, indicative of kidney dysfunction, while
   hematological analysis demonstrated a significant decrease in white
   blood cell (WBC) count, neutrophils (Neu), and platelet count (PLT),
   reflecting the hematopoietic toxicity typically associated with
   methotrexate therapy.

   In contrast, the group treated with Pt-MOFs@Glu exhibited no such
   abnormalities in organ histology or hematological parameters,
   highlighting the biocompatibility and lower toxicity of Pt-MOFs@Glu.
   The absence of these adverse effects suggests that Pt-MOFs@Glu not only
   provides effective therapeutic potential but also offers a superior
   safety profile compared to conventional treatments such as
   methotrexate. These findings emphasize the promise of Pt-MOFs@Glu as a
   safer alternative, potentially reducing the systemic toxicity commonly
   seen with current first-line CNSL treatments. This evidence strengthens
   the case for further clinical investigation of Pt-MOFs@Glu, with the
   goal of improving patient outcomes by minimizing off-target effects and
   enhancing the overall therapeutic index.

DISCUSSION

   This study introduces a pioneering therapeutic strategy for the
   treatment of CNSL by using a platinum-based MOF (Pt-MOFs@Glu) with a
   padlock design to exploit the gut-brain axis for targeted drug
   delivery. CNSL treatment is inherently challenging due to the BBB,
   which notably hinders the effective delivery of therapeutic agents to
   brain tumors ([181]34). In addressing this limitation, our approach
   leverages scRNA-seq data to identify intestinal macrophages as
   potential drug carriers that can migrate from the gut to the brain via
   chemokine-mediated signaling. By designing Pt-MOFs@Glu, which
   incorporates carboplatin, pyrazine-quinoxaline, and β-glucan, we
   harness this migratory pathway, ensuring precise delivery of the drug
   directly to CNSL sites through macrophage-mediated transport.

   Key findings of this study include the identification of Cx3cr1
   signaling as a critical pathway in macrophage migration, with
   experimental results showing that intestinal macrophages efficiently
   internalize Pt-MOFs@Glu through Dectin-1 receptors. This provides a
   robust transport mechanism for delivering the drug to CNSL sites.
   Cytotoxicity assays confirmed the favorable biosafety profile of
   Pt-MOFs@Glu, with enhanced cell viability compared to free carboplatin
   and effective, controlled drug release upon ROS-triggering. The induced
   avalanche effect resulted in significant tumor cell apoptosis,
   substantially enhancing therapeutic efficacy. In vivo studies further
   validated the oral administration route of Pt-MOFs@Glu, demonstrating
   its ability to traverse the GI, cross the BBB, and accumulate at CNSL
   sites. Imaging studies confirmed sustained brain accumulation, and
   immune profiling revealed that Pt-MOFs@Glu activated CD^4+, CD^8+ T
   cells, killer T cells, and effector T cells and enhanced
   proinflammatory cytokine expression, indicating an additional
   immune-mediated antitumor effect.

   Despite these promising results, several challenges must be addressed
   before clinical translation can be realized. First, while our scRNA-seq
   data and animal models have provided valuable insights into macrophage
   migration and Cx3cr1 signaling, the precise role of these pathways in
   human CNSL requires further investigation. Species-specific differences
   in the gut-brain axis dynamics present a critical challenge, as the
   mechanisms observed in animal models may not entirely reflect those in
   humans. Humanized models or early-phase clinical trials will be
   essential to assess the pharmacokinetics, biodistribution, and efficacy
   of Pt-MOFs@Glu in humans. Understanding interspecies variability in
   drug transport and macrophage recruitment will be critical for ensuring
   the clinical relevance of these findings.

   Second, while the ROS-triggered avalanche effect has shown efficacy in
   preclinical models, the consistency of ROS production within the human
   CNSL microenvironment remains uncertain. ROS generation may vary across
   different TMEs and between individuals. Therefore, strategies to
   modulate ROS levels—such as combining Pt-MOFs@Glu with pro-oxidant
   drugs or photothermal therapies—could improve the reliability and
   effectiveness of this activation mechanism. However, these strategies
   will require extensive optimization and validation in human models to
   ensure their applicability and clinical success ([182]35–[183]37).

   Last, although the Pt-MOFs@Glu system demonstrated stability under
   gastrointestinal conditions, its long-term biocompatibility and
   potential systemic toxicity must be rigorously evaluated in preclinical
   and clinical studies. It is crucial to assess both the acute and
   chronic safety profiles of Pt-MOFs@Glu, including its long-term effects
   on organ function and overall health. Special attention must be given
   to immune activation and inflammation, as excessive immune responses
   could undermine therapeutic efficacy and induce adverse side effects.
   While the Pt-MOFs@Glu system leverages macrophages for targeted drug
   delivery, macrophages also play a crucial role in systemic immune
   surveillance, and their activation could potentially lead to off-target
   effects. These unintended interactions may result in immune responses
   in nontumor tissues, affecting the overall treatment efficacy and
   safety. Potential off-target effects should therefore be carefully
   considered, as the Pt-MOFs@Glu system may interact with unintended
   tissues or cells, leading to unforeseen consequences. These off-target
   interactions could affect nontumor tissues, potentially causing
   unintended toxicity or immunological responses that could limit the
   therapeutic efficacy and safety of this system.

   To address these concerns, potential solutions include incorporating
   macrophage-targeting strategies with enhanced selectivity for
   tumor-associated macrophages (TAMs) rather than those involved in
   systemic immune surveillance. This could be achieved by designing
   Pt-MOFs@Glu to preferentially bind to specific surface markers
   expressed by TAMs, reducing the risk of off-target immune activation.
   In addition, incorporating “shielding” strategies, such as coating
   nanoparticles with immunologically inert or anti-inflammatory agents,
   could help mitigate undesired immune responses. Advanced in vivo
   imaging and profiling techniques will be critical for monitoring
   macrophage activity in real time and guiding the refinement of these
   delivery systems. Comprehensive studies focused on identifying and
   mitigating these off-target effects will be essential to ensure the
   clinical feasibility of this therapeutic platform.

   In summary, this study proposes a promising gut-to-brain delivery
   system using Pt-MOFs@Glu, offering notable potential for overcoming the
   treatment barriers of CNSL. Although the preclinical data are
   compelling, several critical challenges remain to be addressed before
   advancing to clinical application. Future research should focus on
   validating these findings in humanized models, optimizing the system’s
   pharmacokinetics, and evaluating long-term safety and efficacy in
   clinical trials. The development of this novel therapeutic strategy
   holds great promise for improving CNSL treatment and potentially other
   brain-related diseases.

   In conclusion, the padlock-designed Pt-MOFs@Glu system offers a highly
   promising approach to overcoming the challenges of CNSL treatment. By
   harnessing the gut-brain axis and a lock-and-key mechanism, this
   platform enables targeted, noninvasive drug delivery across the BBB.
   The combination of carboplatin, pyrazine-quinoxaline, and β-glucan
   ensures precise tumor targeting, ROS-triggered drug release, and immune
   activation, resulting in an avalanche effect that not only enhances
   therapeutic efficacy but also minimizes systemic toxicity. While the
   findings are promising, it is important to acknowledge that this study
   is preclinical, and further research is required to address crucial
   factors such as long-term safety, pharmacokinetics in humans, and the
   potential immunogenicity of the MOF formulation. Nevertheless, this
   work lays the foundation for the future of precision therapies, and
   with continued advancements, the promise of transforming the treatment
   landscape for CNSL and other CNS diseases becomes increasingly
   tangible.

MATERIALS AND METHODS

Materials

   The following compounds were sourced from Sigma-Aldrich in St. Louis,
   Missouri, USA: pyrazine-quinoxaline, carboplatin, hydrogen peroxide
   (30%, w/w), yeast, triphenylphosphine (PPh[3]), and diisopropyl
   azodicarboxylate. The Alexa Fluor 633 N-hydroxysuccinimide ester was
   obtained from Thermo Fisher Scientific in Waltham, Massachusetts, USA.
   The A20 and RAW264.7 cells used in this study were provided by the
   American Type Culture Collection in Manassas, VA. Cell culture reagents
   were supplied by Gibco, located in Grand Island, NY, USA. All compounds
   and reagents used in this research were of analytical grade.

scRNA-seq datasets processing

   Quantitative parameters and cell screening criteria were applied
   following previously published literature. Briefly, the gene barcode
   counting matrix was analyzed using the Seurat R package (version
   4.0.2). Cells with more than 200 genes and less than 10% mitochondrial
   gene content were selected for downstream analysis. After converting
   samples into Seurat objects, normalization and scaling were performed
   by regressing unique molecular identifier counts and mitochondrial gene
   percentages. For dimensionality reduction, the FindVariableGenes
   function identified the most variable genes, followed by PCA for
   dimensional reduction. The RunTSNE function in Seurat (version 3.1.3)
   was then used to generate the TSNE plot.

Preparation and characterization of Pt-MOFs@Glu

   The padlock-designed Pt-MOFs were synthesized through a solvothermal
   reaction to achieve a highly efficient and structurally robust
   nanocarrier. In summary, a mixture of carboplatin (15 mg),
   pyrazino[2,3-f]quinoxaline (15 mg), and acetic acid (100 μl) was
   dissolved in N,N′-dimethylformamide (15 ml). The solution was then
   transferred into a 50-ml round-bottom flask, which was subsequently
   heated to 180°C for 4 hours to facilitate the solvothermal reaction.
   After the reaction was completed, the resulting mixture was centrifuged
   at 10,000 rpm for 10 min to separate the solid products from the
   solution. The solid nanoparticles were then washed sequentially with
   30 ml of deionized (DI) water and ethanol to remove any unreacted
   precursors or by-products.

   Following the purification, the Pt-MOFs were dispersed in DI water (0.5
   mg/ml) to obtain a stable nanoparticle suspension. To further
   functionalize the nanoparticles, mercaptoethylated β-glucan (0.5 mg/ml)
   was added to the dispersion, and the mixture was stirred overnight at
   room temperature. Last, to ensure complete the removal of any residual
   reagents and unreacted β-glucan, the nanoparticles underwent an
   additional washing step. The purified Pt-MOFs were then ready for
   further characterization and application in drug delivery and
   therapeutic studies.

   The morphology and structure of the nanoparticles were analyzed using
   scanning electron microscopy, transmission electron microscopy, FTIR,
   and ICP-MS. In addition, the particle size and surface charge of the
   Pt-MOFs@Glu were determined using DLS. To assess their long-term
   stability and gastrointestinal resilience, the particles were subjected
   to 4°C DI water and simulated gastric and intestinal fluids, with
   samples withdrawn at specific intervals for particle size analysis.

   Furthermore, to evaluate the responsiveness of the Pt-MOFs@Glu to ROS
   within the TME, a solution of 100 μM H[2]O[2] was prepared in PBS for
   simulation purposes. Drug release studies were conducted at 37°C, with
   supernatant samples collected at designated time intervals. The
   concentration of carboplatin in the supernatant was quantified by
   measuring absorbance values at 229 nm using a UV-visible
   spectrophotometer.

Cell viability assay

   The cytotoxicity or cytocompatibility of the prepared nanoparticles was
   assessed by incubating them at different concentrations (0 to 400
   μg/ml) with A20 cells or RAW264.7 macrophages (5 × 10^3 cells per well)
   in 96-well plates. Carboplatin at an equivalent concentration was used
   as a positive control. Following 48 hours of incubation, cell viability
   was determined using an MTT cell viability assay kit.

Uptake by macrophage

   To visualize the uptake of Pt-MOFs@Glu by macrophage cells in vitro,
   RAW264.7 cells (1 × 10^6 cells/ml) were cultured with
   fluorescence-labeled Pt-MOFs@Glu (100 μg/ml), synthesized using Cy5.5
   according to the manufacturer’s protocol. After incubation for
   predetermined time intervals (0, 1, 2, 4, or 6 hours), the cells were
   collected, washed with PBS, stained with 4′,6-diamidino-2-phenylindole,
   and subsequently examined using CLSM.

Penetration of CNSL by Pt-MOFs@Glu

   Three-dimensional CNSL spheroids composed of A20 cells were generated.
   These spheroids were then incubated with Pt-MOFs@Glu or macrophages
   following predetermined co-incubation periods (3, 6, or 12 hours).
   Subsequently, they were rinsed with PBS, fixed in 4% paraformaldehyde
   (PFA), immunofluorescently stained with F4/80 antibody, transferred
   onto a confocal dish, and examined using CLSM.

Animal study

   BALB/c mice (8 weeks old, weighing 18 to 20 g) were procured from
   SiPeiFu. All animal experiments were conducted in accordance with the
   guidelines outlined in the “Guide for the Care and Use of Laboratory
   Animals,” formulated by the Institute of Laboratory Animal Resources,
   National Research Council and published by the National Academy Press
   in 2011. Animal study protocols were approved by the Institutional
   Animal Care and Use Committee of Sichuan Provincial People’s Hospital
   (protocol no. 2024416). To establish the brain tumor model, test mice
   were intracranially inoculated with 1 × 10^7 A20 cells suspended in
   5 μl of PBS.

Biodistribution of Pt-MOFs@Glu

   In the biodistribution study, mice bearing tumors and fasted overnight
   were orally administered Pt-MOFs@Glu. At predetermined intervals (0, 2,
   4, 6, or 8 hours) posttreatment, the accumulation of Pt-MOFs@Glu in the
   brain tumor and major organs—including the heart, lungs, liver, spleen,
   and kidneys—was quantified using IVIS and ICP-MS, respectively.
   Untreated mice and those receiving an equivalent dose of free cisplatin
   served as controls.

Transport route of Pt-MOFs@Glu

   To investigate the transport route of Pt-MOFs@Glu in vivo, mice fasted
   overnight were orally administered Pt-MOFs@Glu (n = 3). In the
   inhibitor group, mice were intraperitoneally injected with
   cycloheximide (3 mg/kg, n = 3). After six hours, small intestines
   (villi, Peyer’s patches, and lymphatic vessels) and brain tumors were
   harvested from euthanized mice. The tissues were cryosectioned and
   immunofluorescently stained with anti-M cell antibody, F4/80 antibody,
   and anti-LYVE-1 antibody to identify M cells, intestinal macrophages,
   and lymphatic vessels, respectively. CLSM was used to detect the
   colocalization of Pt-MOFs@Glu with M cells, intestinal macrophages, and
   lymphatic vessels.

In vivo antitumor efficacy

   To evaluate the antitumor efficacy of each formulation, tumor-bearing
   mice fasted overnight were randomly allocated to five test groups:
   untreated control, free drug (carboplatin), and Pt-MOFs@Glu. Each mouse
   in the treatment groups received an equivalent dose of carboplatin (20
   mg/kg). The body weights of tumor-bearing mice were monitored every
   other day throughout the study. At the study’s conclusion, mice were
   euthanized, and tumor tissues were retrieved, fixed with 4% PFA,
   embedded in paraffin wax, and then sectioned and stained with H&E, Ki67
   monoclonal antibody, or TUNEL.

In vivo toxicity evaluation

   To assess in vivo toxicity, vital organs from each study group were
   retrieved at the study’s conclusion, fixed in 4% PFA, embedded in
   paraffin blocks, and stained with H&E for histopathological
   examination. Organs obtained from healthy mice served as controls.
   Serum samples collected from mice were used to measure the levels of
   aspartate aminotransferase and BUN to evaluate potential liver and
   kidney toxicities.

Statistical analysis

   All data are presented as the means ± SD, unless otherwise specified.
   Each experiment was conducted with a minimum of three repetitions.
   Statistical comparisons were performed using one-way or two-way
   analysis of variance (ANOVA), followed by Bonferroni multiple
   comparisons posttest. The survival curves were compared using the
   log-rank test (Mantel-Cox). Statistical significance was set at
   *P < 0.05 for all analyses. All statistical calculations were conducted
   using Prism 8 (GraphPad Software).

Acknowledgments