Wnt signaling regulates MFSD2A-dependent drug delivery through endothelial transcytosis in glioma

Abstract

Background

Systemic delivery of anti-tumor therapeutic agents to brain tumors is thwarted by the blood-brain barrier (BBB), an organotypic specialization of brain endothelial cells (ECs). A failure of pharmacological compounds to cross BBB is one culprit for the dismal prognosis of glioblastoma (GBM) patients. Identification of novel vascular targets to overcome the challenges posed by the BBB in tumors for GBM treatment is urgently needed.

Methods

Temozolomide (TMZ) delivery was investigated in CT2A and PDGFB-driven RCAS/tv-a orthotopic glioma models. Transcriptome analysis was performed on ECs from murine gliomas. Mfsd2a deficient, Cav1 deficient, and Mfsd2a EC-specific inducible mice were developed to study the underlying molecular mechanisms.

Results

We demonstrated that inhibiting Wnt signaling by LGK974 could increase TMZ delivery and sensitize glioma to chemotherapy in both murine glioma models. Transcriptome analysis of ECs from murine gliomas revealed that Wnt signaling inhibition enhanced vascular transcytosis as indicated by the upregulation of PLVAP and downregulation of MFSD2A. Mfsd2a deficiency in mice enhances TMZ delivery in tumors, whereas constitutive expression of Mfsd2a in ECs suppresses the enhanced TMZ delivery induced by Wnt pathway inhibition in murine glioma. In addition, Wnt signaling inhibition enhanced caveolin-1 (Cav1)-positive caveolae-mediated transcytosis in tumor ECs. Moreover, Wnt signaling inhibitor or Mfsd2a deficiency fails to enhance TMZ penetration in tumors from Cav1-deficient mice.

Conclusions

These results demonstrated that Wnt signaling regulates MFSD2A-dependent TMZ delivery through a caveolae-mediated EC transcytosis pathway. Our findings identify Wnt signaling as a promising therapeutic target to improve drug delivery for GBM treatment.

Key Points.

• Wnt signaling inhibition leads to enhanced tumor vascular transcytosis, increasing temozolomide (TMZ) delivery and sensitizing GBM to chemotherapy in orthotopic GBM models.

• Wnt signaling regulates MFSD2A-dependent TMZ delivery through a caveolae-mediated EC transcytosis pathway. Targeting Wnt pathway is a promising therapeutic strategy to improve drug delivery and penetration for GBM treatment.

Importance of the Study.

Glioblastoma (GBM) is the most aggressive type of adult diffuse gliomas. The effectiveness of systemically delivered anti-tumor therapeutic agents is thwarted by the BBB, and the failure of pharmacological compounds to cross the BBB is one of the main reasons for the devastating prognosis of GBM patients. Here, by employing two orthotopic GBM models and several transgenic mouse strains, we reveal that Wnt signaling regulates MFSD2A-dependent TMZ delivery through a caveolae-mediated EC transcytosis pathway. Our results provide vital genetic evidence identifying Wnt signaling as a promising therapeutic target to improve drug delivery for GBM treatment, with important implications for the design of rational therapeutic regimens.

Diffuse gliomas, the most common primary brain malignancies, comprise of clinical and molecular distinct tumors, including pediatric-type low-/high-grade glioma and adult-type diffuse gliomas.¹ Glioblastoma (GBM) is the most aggressive type of adult diffuse gliomas. The median survival of GBM patients is only 14.6 months, and has not substantially improved during the past 3 decades.² The failure of pharmacological compounds to cross blood-brain barrier (BBB), a specialized brain vascular unit, is one culprit for the dismal prognosis of GBM patients.³ Recently, by using single-cell RNA sequencing of freshly isolated ECs from human GBM and paired non-malignant brain tissues, we have demonstrated that ECs in GBM have a partially intact BBB.⁴ Substantial levels of active efflux transporters and junction molecules are still present in tumor ECs, which may thwart the effectiveness of anti-tumor compounds.^4,5 Vascular targeting with bevacizumab, a recombinant humanized monoclonal antibody blocking vascular endothelial growth factor signaling, to prune and normalize tumor vasculature have been approved by FDA since 2009 but have failed to improve overall survival of GBM patients.⁶ Notably, bevacizumab administration may reduce vascular permeability and temozolomide (TMZ) delivery in human recurrent GBM.⁷ Therefore, overcoming the challenges posed by the blood-tumor barrier, a compromised form of BBB in brain tumors, to improve drug delivery and efficacy is one potential therapeutic strategy of GBM.⁸

The Wnt signaling pathway regulates the BBB formation and function in both the development and pathogenesis of diseases, and is essential for CNS angiogenesis and BBB formation in both embryonic and postnatal development.^9,10 Wnt signaling activation in human pluripotent stem cell-derived naïve endothelial progenitors lead to the robust acquisition of BBB phenotype.¹¹ Genetic deletion of Wnt7a/7b or GPR124 or specific deletion of β-catenin in ECs leads to embryonic lethality, and elicits several vascular defects in the CNS, including the absence of EC invasion in parenchymal, formation of hemorrhagic vascular malformation and failure of CNS specific EC differentiation.^9,10,12 Inactivation of β-catenin during the postnatal BBB maturation stage leads to downregulation of Cldn3, increased level of PLVAP and BBB breakdown.¹³ Decreased Wnt signaling as a result of conditional EC-specific knockout of Unc5B, a coactivator of Wnt signaling through phosphorylation of LRP6, in mice lead to leaky BBB with a barrier-incompetent state characterized by reduced Cldn5 and increased Plvap expression.¹⁴ Recently, Maud et al. reported an engineered BBB-specific Wnt ligands with strict specificity for the BBB’s Gpr124/Reck Wnt signaling without Wnt activation in other tissues, which could be used to restore BBB in neurological disorders.¹⁵ In adult mice, the Wnt signaling pathway is required to maintain BBB integrity in pathological conditions.^16,17 In medulloblastoma, the abundant production of WNT antagonists (eg, WIF1, DKK1) in tumors of the WNT subgroup blocks endothelial WNT signaling and induces a leaky and fenestrated vasculature permitting intra-tumoral accumulation of chemotherapeutic-drugs, indicating that WNT signaling dictates the BBB phenotype in medulloblastoma.¹⁸ In GBM, Wnt/β-catenin signaling promotes the preservation of the BBB, leading to a quiescent vascular phenotype characterized by decreased vascular density and permeability and increased mural cell coverage.^19,20 However, most of the previous studies on Wnt signaling in brain tumors have focused on BBB regulation without examining the potential effects on drug delivery, and the molecular mechanisms remain to be illuminated. Here, by using a Wnt signaling inhibitor and transgenic mouse models, we uncovered that Wnt signaling inhibition could sensitize GBM to TMZ treatment by increasing drug delivery. Mechanistically, Wnt pathway inhibition led to the downregulation of MFSD2A, which caused increased caveolae-mediated transcytosis. Thus, Wnt signaling inhibition increased chemotherapy drug delivery and promoted survival in mouse GBM models. Together, our findings indicate that Wnt signaling is a promising therapeutic target of GBM to improve drug delivery and thereby drug efficacy.

Materials and Methods

Mice

Mfsd2a ^-/- mice (Supplementary Figure S1), Cav1^-/- mice (Supplementary Figure S2), and ROSA26^{STOPfloxMfsd2a/TdTomato} mice were generated by CRISPR/Cas9-mediated genome engineering at Cyagen Biosciences, see Supplementary Materials and Methods for details.

GBM Tumor Induction and Treatment

Animal experiments were performed in accordance with the rules of Shaanxi Normal University and were approved by the local animal ethics committee. GBM was induced in mice as previously described.²¹ RCAS-producing DF-1 cells (RCAS-PDGFB-HA) were orthotopically transplanted by stereotaxic injection of 10⁵ cells into 6–8 weeks old of Gtv-a;Arf^-/- mice to induce GBM. CT2a cells (4 × 10⁴) were orthotopically injected into 6–8 weeks old of C57BL/6 mice or Mfsd2a^-/- mice or Cav1^-/- mice or Mfsd2a^-/-;Cav1^-/- mice or Cdh5-Cre^ERT2; ROSA26^{STOPfloxMfsd2a/TdTomato} (Mfsd2a-GOF) mice. The coordinates were: 0.5 mm anterior of bregma, 1.1 mm lateral, and 2.5 mm ventral. ROSA26^{STOPfloxMfsd2a/TdTomato} (Mfsd2a-GOF) mice were intraperitoneally administrated with tamoxifen (Sigma) at the dose of 2 mg/day per mouse for 5 consecutive days 1 week before tumor cell injection.

To examine the effect of LGK974 on TMZ and paclitaxel concentration in tumors, CT2A and RCAS-PDGFB tumors were implanted in mice. Seventeen days after tumor induction, mice were randomized into 2 groups, either receiving (gavage) 2.5 mg/kg LGK974 (in PBS with 0.5% tween-80 and 0.5% methylcellulose) (MedChemExpress) or control gavage (PBS with 0.5% tween-80 and 0.5% methylcellulose) for 4 consecutive days. Twenty-one days after tumor induction, mice were administrated with the peritoneal injection of 25 mg/kg temozolomide or 20 mg/kg paclitaxel. Forty-five minutes after temozolomide administration or 90 min after paclitaxel administration, mice were anaesthetized, and transcardially perfused using PBS to remove blood. Tissues were harvested for HPLC analysis. To control for the sex variable, the female/male ratio was controlled at a similar range in all the groups for the in vivo experiment.

Statistical Analysis

Statistical analysis was performed using unpaired Student’s t test or analysis of variance (ANOVA) analysis for experiments with 2 groups or more than 2 groups, respectively. Kaplan-Meier survival curves were generated using Prism software, and log-rank test was performed to assess the statistical significance between groups. Most of the experiments were repeated at least 2 times with similar results. Data are expressed as individual values plus mean ± SD. The following P values indicate statistical significance: *P < .05, **P < .01, ***P < .001, ****P < .0001.

Data Availability

The RNAseq raw sequencing data and also processed counts data are available in the NCBI Gene Expression Omnibus under accession number GSE203374.

Other Methods

For materials and methods related to bioinformatics analysis of single-cell sequencing, survival experiments, immunohistochemical staining, quantitative polymerase chain reaction (qPCR), western blot, lentivirus production, immunofluorescence staining, in vitro BBB assay, HPLC assay, microsphere extravasation in GBM tumors, endothelial cells isolation, GSEA and immunogold labeling for electron microscopy, see Supplementary Materials and Methods for details.

Results

Wnt Signaling Activity Correlates With BBB Status in Human Brain/GBM ECs

To examine Wnt signaling in the human brain and GBM vessels, we have performed immunohistochemical staining on the human brain and GBM tissue for LEF1, a critical component and direct feedforward target of Wnt signaling,¹⁹ and β-catenin. Vascular expression of LEF1 and nuclear staining of β-catenin in endothelial cells were observed in the majority of blood vessels from both control and tumor tissues at a similar level, suggesting a Wnt-activated state in both control and tumor vasculature. (Figure 1A–D). To further explore whether Wnt signaling is correlated with BBB status in human GBM ECs, we have analyzed single-cell transcriptome data from ECs in 4 human GBM and paired non-malignant tissues.⁴ All 1139 individual ECs were ordered by sorting points into neighborhoods (SPIN) according to a core BBB gene signature into a single one-dimensional range. SLC2A1 and ABCG2, encoding the BBB-specific transporters, peaked at the left of the SPIN range, matching the fact of the high level of SLC2A1 and ABCG2 in ECs with intact BBB. PLVAP, a vascular leaky marker indicative of BBB dysfunction, peaked at the right of SPIN range, matching the expression of PLVAP in the fenestrated vasculature (Figure 1E). Notably, Wnt targets, including APCDD1²² and MFSD2A,²³ peaked at the left side of the SPIN range and resembled a similar expression pattern to the BBB markers on the intact-leaky BBB axis, indicating that Wnt signaling activity is correlated with BBB status in human GBM ECs.

Figure 1.

Wnt signaling activity is correlated with BBB status in human brain/GBM ECs. (A–B) Co-immunofluorescence staining (A) and quantification (B) of LEF1 and CD31on human non-malignant brain, lower grade glioma and glioblastoma. (C–D) Co-immunofluorescence staining (C) and quantification (D) of β-catenin and CD31 on human non-malignant brain, lower grade glioma and glioblastoma. Endothelial cells with nuclear staining of β-catenin were indicated by white arrow. (E) Expression of core BBB genes (SLC2A1, ABCG2), BBB dysfunction gene (PLVAP) and Wnt targets (MFSD2A, APCDD1) across endothelial cells sorted by SPIN based on a core BBB gene set. Scale bar: 50 μm (A), 20 μm (C).

Wnt Singling Pathway Inhibition Increases TMZ Delivery and Sensitizes GBM to TMZ Therapy

To investigate the effect of Wnt signaling on TMZ delivery, we used a monolayer of brain endothelial cells bEND.3 as an in vitro model of the BBB.⁵ Pre-treatment of bEND.3 cells with LGK974, a potent Wnt signaling inhibitor, led to increased transport of TMZ to the abluminal side of endothelial cells (Figure 2A and B).

Figure 2.

Wnt singling pathway inhibition increases TMZ delivery and sensitizes GBM to TMZ therapy. (A) Schematic illustration of the in vitro BBB models used in the study. bEND.3 cells formed a confluent monolayer in the upper part of chamber. TMZ was added to the upper compartment to measure the BBB transport. (B) Concentration of TMZ in the bottom compartment of the chamber. (C) Immunohistochemical staining of CD31 in CT-2A glioma from mice treated with or without LGK974. (D–E) Stereological quantification of vascular area (D) and mean vascular diameters (E) in CT-2A glioma. (F–G) Co-immunofluorescence staining (F) and quantification (G) of LEF1 and CD31 in CT-2A tumors from mice treated with or without LGK974. (H-I) Co-immunofluorescence staining (H) and quantification (I) of β-catenin and CD31 in CT-2A tumors from mice treated with or without LGK974. Endothelial cells with or without nuclear staining of β-catenin were indicated by white arrow or arrowhead respectively. (J) Quantification of TMZ concentration in CT-2A glioma from mice treated with or without LGK974. (K) Immunohistochemical staining of CD31 in RCAS/tv-a glioma from mice treated with or without LGK974. (L–M) Stereological quantification of vascular area (L) and mean vascular diameters (M) in RCAS/tv-a glioma. (N–O) Co-immunofluorescence staining (N) and quantification (O) of LEF1 and CD31 in RCAS/tv-a glioma from mice treated with or without LGK974. (P–Q) Co-immunofluorescence staining (P) and quantification (Q) of β-catenin and CD31 in RCAS/tv-a glioma tumors from mice treated with or without LGK974. Endothelial cells with or without nuclear staining of β-catenin were indicated by white arrow or arrowhead respectively.(R) Quantification of TMZ concentration in RCAS/tv-a glioma from mice treated with or without LGK974. (S–T) Survival curve of CT-2A glioma- (S) or RCAS/tv-a glioma- (T) bearing mice treated with LGK974 and/or TMZ. (U–V) Quantification of paclitaxel concentration in CT-2A glioma (U) or RCAS/tv-a glioma (V) from mice treated with or without LGK974. Scale bar: 100μm (C, K), 50μm (F, N), 20μm (H, P).

To assess whether Wnt signaling inhibition affects TMZ delivery in tumors, we induced a syngeneic orthotopic GBM model by inoculating CT2A cells into the cerebral cortex of the C57BL/6 mice, followed by treatment with Wnt inhibitor LGK974 for 4 days. Treatment with LGK974 did not alter vascular density (Figure 2C–E), but led to a strong reduction of LEF1 expression in the vasculature and nuclear staining of β-catenin in endothelial cells (Figure 2F–I), indicating efficient inhibition of Wnt signaling. TMZ penetration was increased in the CT2A tumors upon LGK974 treatment (Figure 2J). By contrast, LGK974 treatment did not increase TMZ penetration in either healthy tumor-free brain or peripheral organs, including kidney, heart, and liver (Supplementary Figure S3A–D). To address whether Wnt signaling inhibition enhances TMZ penetration in other GBM models, we used another well-developed PDGFB-driven RCAS/tv-a GBM model. Similar to the CT2a model, we observed no differences in vascular density (Figure 2K–M) but reduced LEF1⁺vessels and nuclear staining of β-catenin in endothelial cells in RCAS tumors upon LGK974 treatment (Figure 2N–Q). Moreover, Wnt signaling inhibition by LGK974 resulted in increased TMZ penetration in RCAS tumors (Figure 2R).

To further explore whether enhanced TMZ penetration by Wnt signaling inhibition will lead to a therapeutic benefit, the CT2A glioma or PDGFB-driven RCAS/tv-a glioma-bearing mice were treated with LGK974 together with or without TMZ. TMZ treatment alone significantly improved mouse survival in both tumor models (Figure 2S and T). In line with the limited cytotoxicity of LGK974 to tumor cells in vitro (Supplementary Figure S3E), LGK974 treatment alone did not show survival benefit (Figure 2S and T). Notably, combined treatment of LGK974 and TMZ led to a synergistic improvement in survival, which was superior to control or any single treatment, in both glioma models (Figure 2S and T). TMZ inhibited CT2A tumor cell growth in a dose-dependent manner (Supplementary Figure S3F), and a 2-fold increase in TMZ concentration were sufficient to result in a substantial difference of tumor cell growth inhibition in vitro (Supplementary Figure S3F). Thus, survival benefits in the combination group are likely due to enhanced TMZ delivery.

Next, we investigated whether Wnt signaling inhibition could also enhance the delivery of paclitaxel, a potent cytotoxic chemotherapeutic agent for routine ovarian, breast, gastric prostate, and head-and-neck cancer treatment but with limited BBB penetration.²⁴ LGK974 treatment enhanced tumoral paclitaxel concentration to ~2.6 fold in CT2A glioma model and ~2.3 fold in RCAS/tv-a glioma model (Figure 2U and V). Taken together, these results suggest that Wnt signaling pathway inhibition increases drug delivery and sensitizes of GBM to chemotherapy.

Wnt Singling Regulates Nano-Particle Delivery

To explore whether Wnt signaling inhibition could enhance nano-drug delivery, we established a modified in vitro BBB model in which tumor cells were cultured at the bottom chamber (Figure 3A). LGK974 treatment on endothelial cells led to increased penetration of 100 nm FluoSpheres^TM nanoparticles (NPs) as determined by the concentration of NPs in the bottom chamber (Figure 3B). In line with more NPs passing through the BBB, the uptake in tumor cells was elevated as well (Figure 3C and D). We subsequently investigated the effect of Wnt inhibition on NP delivery in murine tumors. The results showed a significantly greater accumulation of NPs in tumors as determined by increased intracellular red fluorescence (Figure 3E and F). These results support further studies to investigate the synergistic effects of the combination treatment of Wnt signaling inhibitors and NP-delivered drugs.

Figure 3.

Wnt signaling inhibition enhances nano-particles BBB delivery. (A) Schematic illustration of the in vitro BBB models with CT-2A glioma cells cultured in the bottom part of the chamber. Nano-particles (NPs) was added to the upper compartment to measure the BBB transport. (B) Concentration of NPs in the bottom compartment of the chamber. (C–D) Fluorescence image (C) and quantification (D) of NPs uptake by tumor cells in the bottom compartment of the chamber. (E–F) Fluorescence image (E) and quantification (F) of NPs in CT-2A glioma from mice treated with or without LGK974. Scale bar: 50 μm (C) and 200 μm (E).

Wnt Signaling Inhibition Results in Upregulation of Genes Associated With Transcytosis in Tumor ECs

To understand how Wnt signaling inhibition affects TMZ delivery across the BBB, we analyzed tumor endothelial transcripts by RNA sequencing (Figure 4A). Endothelial cells were isolated by magnetic-activated cell sorting (MACS) based on the expression of CD31 and CD45 (CD45^-/CD31⁺), from tumors in control or LGK974-treated mice (Figure 4A). The CT2A cells were engineered to constitutively express a green fluorescent protein, allowing tumor visualization and dissection from the brain. Transcriptome analysis uncovered many Wnt targeted genes (Apcdd1, Axin2, Dkk2, Lef1, and Mfsd2a) with significantly decreased expression in tumor ECs upon LGK974 treatment, indicating a strong suppression of Wnt signaling in tumor vessels (Figure 4B, Supplementary Figure S4A–E). In accordance with the recent finding that aberrant Wnt signaling triggers CNS inflammation in multiple sclerosis,²⁵ Wnt signaling inhibition increased expression of inflammatory and cytokine genes such as Il6, Sele, Selp, Cxcl2, Lgals3, Ccl11, and Ptgs1 (Figure 4B, Supplementary Figure S4F–L). Alteration of Wnt and inflammatory signaling was further confirmed by functional annotation of 205 differentially expressed genes (threshold: fold change > 1.5 and P < 0.05) (Supplementary Table S1) in tumor ECs between the control group and LGK974 treated group (Figure 4C). GO terms including “neutrophil/granulocyte/leukocyte chemotaxis” and “neutrophil/granulocyte/leukocyte migration” were significantly enriched in tumor ECs upon Wnt inhibition, whereas “negative regulation of Wnt signaling pathway”, the feedback response to the Wnt signaling pathway indicating Wnt signaling activity, was the most enriched GO term in tumor ECs in the control group (Figure 4C).

Figure 4.

Wnt regulates genes associated with inflammatory and transcytosis in tumor ECs. (A) Schematic overview of the transcriptome study design. (B) Expression level of selected genes in tumor ECs isolated from control or LGK974 treated tumor-bearing mice. (C) GSEA analysis of up- and down-regulated gene sets in tumor ECs isolated from control or LGK974 treated tumor-bearing mice. (D–E) Immunofluorescence (D) and quantification (E) of PLVAP in CT-2A glioma from mice treated with or without LGK974. (F–G) Immuno-staining (F) and quantification (G) of Mfsd2a in CT-2A glioma from mice treated with or without LGK974. (H–I) Immunofluorescence (H) and quantification (I) of Cav1 in CT-2A glioma from mice treated with or without LGK974. Scale bar: 100 μm.

Considering the key role of Wnt signaling pathway in BBB regulation, we subsequently investigated whether LGK974 treatment affected expression of key BBB gene expression. Surprisingly, Wnt inhibition did not produce a general alteration of key BBB genes (Figure 4B). Except for Slc7a1, expression of most junction molecules and BBB transporters, including Cdh5, Cldn5, Esam, Ocln, Tjp1, Tjp2, Abcb1a, Abcb1b, Abcg2, Slc2a1, and Slc7a5, were similar in tumor ECs from control or LGK974 treated mice (Figure 4B, Supplementary Figure S4M–X). However, Plvap, a marker of leaky vasculature, was one of the top-10 up-regulated genes in tumor EC upon LGK974 treatment, suggesting that Wnt signaling inhibition promoted vascular permeability in GBM (Figure 4B, D and E, Supplementary Figure S4Y, Table S1). Notably, in consistent with enhanced vascular permeability in tumors upon LGK974 treatment, Mfsd2a, a key BBB gene that suppresses vascular permeability by controlling caveolae-mediated transcytosis,²⁶ was significantly downregulated in tumor ECs in response to LGK974 treatment (Figure 4B, F and G, Supplementary Table S1), indicating an elevated level of endothelial transcytosis in tumor vessels following Wnt inhibition. Intriguingly, Cav1, which is a molecule obligatory for caveolae structure formation as well as caveolae-mediated transcytosis,²⁷ was increased in ECs from LGK974 treated tumors (Figure 4B, H and I, Supplementary Figure S4Z). Taken together, Wnt signaling inhibition results in the upregulation of genes associated with endothelial transcytosis in tumor vessels.

Enhanced TMZ Delivery by Wnt Signaling Inhibition is Dependent on Mfsd2a

To evaluate the role of Mfsd2a in affecting TMZ delivery in tumors, we established CT2A tumors in Mfsd2a knockout mice and their littermates (Supplementary Figure S1). Mfsd2a knockout in mice resulted in an upregulation of TMZ delivery in tumors (Figure 5A). Mfsd2a knockout in mice did not affect the survival of CT2A tumor-bearing mice as compared to the control mice (Figure 5B), but significantly improved the survival of tumor-bearing mice treated with TMZ (Figure 5B). Interestingly, Mfsd2a knockout in mice did not alter nano-particle delivery in CT2A tumors (Supplementary Figure S5A and B), suggesting that TMZ delivery, but not nano-particle penetration, is regulated by Mfsd2a. Considering the key role of Mfsd2a in suppressing brain EC transcytosis and the fact that Mfsd2a is a direct target of Wnt/β-catenin signaling,²³ we then assessed whether enhanced TMZ delivery by Wnt signaling inhibition is dependent on Mfsd2a. The bEND.3 cells were transduced by lentivirus to overexpress Mfsd2a (Supplementary Figure S6A and B), and then used for in vitro BBB assay. In contrast to increased TMZ delivery crossing BBB with scramble transduced ECs, enhanced TMZ delivery by Wnt inhibition was attenuated by Mfsd2a overexpression (Figure 5C). To further determine whether enhanced TMZ delivery by Wnt inhibition is dependent on Mfsd2a in vivo, we generated ROSA26^{STOPfloxMfsd2a/TdTomato} mice (Figure 5D), which the stop cassette will be removed under Cre-mediated recombination allowing constitutive expression of Mfsd2a and TdTomato reporter. By crossing ROSA26^{STOPfloxMfsd2a/TdTomato} with Cdh5-Cre^ERT2 mice, yielding Cdh5-Cre^ERT2; ROSA26^{STOPfloxMfsd2a/TdTomato} (Mfsd2a-GOF) mice, MFSD2A is constitutively expressed in endothelial cells after tamoxifen induction in Mfsd2a-GOF mice (Figure 5D–F). Quantification of CD31 staining and endogenous TdTomato reporter revealed more than 90% cre recombination rate (Figure 5D). Notably, TMZ delivery enhancement by Wnt inhibition was not observed in Mfsd2a-GOF mice (Figure 5G). Importantly, the survival benefit by combination treatment of LGK974 and TMZ treatment was reversed in Mfsd2a-GOF mice (Figure 5H). Taken together, these results suggest that Wnt-mediated regulation of TMZ delivery is dependent on Mfsd2a.

Figure 5.

Enhanced TMZ delivery by Wnt signaling inhibition is dependent on Mfsd2a. (A) Quantification of TMZ concentration in CT-2A glioma from wild type or Mfsd2a-ko mice. (B) Survival curve of CT-2A glioma form wild type or Mfsd2a ko tumor-bearing mice treated with or without LGK974. (C) Concentration of TMZ passing monolayer of control or Mfsd2a overexpressing bEND.3 cells with or without LGK974 treatment. (D) Technical principle for construction of ROSA26^{STOPfloxMfsd2a/TdTomato} mice and schematic overview to generate Mfsd2a-GOF mice. (E–F) Immunofluorescence (E) and quantification (F) of CD31 and Mfsd2a in CT-2A glioma from tamoxifen induced or non-induced Mfsd2a-GOF mice treated with or without LGK974. (G) Quantification of TMZ concentration in CT-2A glioma from tamoxifen induced or non-induced Mfsd2a-GOF mice treated with or without LGK974. (H) Survival curve of CT-2A glioma form tamoxifen induced or non-induced Mfsd2a-GOF tumor-bearing mice treated with or without LGK974. Scale bar: 100 μm.

Wnt Signaling Inhibition Increases TMZ Delivery Through Enhancement of Mfsd2a-Mediated EC Caveolae Transcytosis

Mfsd2a could maintain BBB homeostasis by suppressing caveolae-mediated transcytosis,²⁷ raising the possibility that increased caveolae-mediated transcytosis accounts for the enhanced TMZ delivery in tumors in Mfsd2a^-/- mice. To test this hypothesis, we generated Cav1^-/- mice (Supplementary Figure S2A–C), and created Mfsd2a and Cav1 double knockout transgenic mouse line by crossing Mfsd2a^-/- with Cav1^-/- mice. Cav1 is essential for caveolae vesicle formation and caveolae-mediated transcytosis.²⁷ Double knockout of Mfsd2a and Cav1 in mice attenuated enhanced tumor TMZ delivery in Mfsd2a knockout mice (Figure 6A), suggesting Mfsd2a regulates TMZ delivery in tumors through regulation of caveolae-mediated transcytosis.

Figure 6.

Wnt signaling inhibition increases TMZ delivery through enhancing Mfsd2a-mediated EC caveolae transcytosis. (A) Quantification of TMZ concentration in CT-2A glioma from wild type, Mfsd2a^-/-, Cav1^-/- or Mfsd2a^-/-/Cav1^-/- mice. (B) EM analysis with immune-gold labeling Cav1 in CT-2A tumors from mice with or without LGK974 treatment. (C) Quantification of Cav-1 positive vesicles in tumor vessels. n = 3 per group. At least 6 vessels per animal were analyzed. (D) Quantification of TMZ concentration in CT-2A glioma from wild type or Cav1^-/- mice treated with or without LGK974. (E–F) TEM images (E) and quantification (F) of endothelial transcytotic vesicles in CT-2A tumors. (G) Survival curve of CT-2A glioma form wild type or Cav1^-/- tumor-bearing mice treated with or without LGK974. (H) Scheme illustration on the role of Wnt signaling in controlling TMZ delivery. Scale bar: 1 μm.

Wnt signaling regulates endothelial caveolae-mediated transcytosis in blood-retinal barrier.²³ To further explore whether Wnt inhibition increases caveolae-mediated transcytosis in brain tumor ECs, we performed immuno-electronic microscope labeling of Cav1 in tumors from control or LGK974-treated mice (Figure 6B and C). Electron-dense gold particle labeled Cav1 were clearly seen in the vasculature associated with vesicle structures (Figure 6B). LGK974 treatment led to a remarkable increase of CAV1-associated vesicles in tumor endothelium (Figure 6B and C). This suggests that Wnt signaling modulation affects EC caveolae transcytosis in brain tumors.

To test the hypothesis that Wnt inhibition increases TMZ delivery through enhanced caveolae-mediated transcytosis, we transplanted CT2A tumors into Cav1 knockout mice and treated the mice with LGK974. LGK974 treatment led to increased TMZ delivery in control mice but not in Cav1^-/- mice (Figure 6D). Moreover, LGK974 treatment resulted in an increase of EC transcytosis vesicles in tumors from control mice, but not in tumors from Cav1^-/- mice (Figure 6E and F). The results supported the hypothesis that elevated TMZ penetration in tumors by Wnt inhibition was mediated through the enhancement of caveolae-mediated transcytosis. In addition, in agreement with the key role of caveolae-mediated transcytosis on Wnt-mediated regulation of TMZ delivery and EC transcytosis, the synergistic effect of TMZ and LGK974 treatment on survival improvement in control tumor-bearing mice was reversed in Cav1^-/- tumor-bearing mice (Figure 6G). Interestingly, in line with results showing that nano-particle penetration is not regulated by Mfsd2a in tumors (Supplementary Figure S5A and B), enhanced nano-particle delivery by LGK974 could not be reversed in tumors from Cav1^-/- mice (Supplementary Figure S7A and B), indicating a caveolae-independent mechanism. Taken together, these results suggest that Wnt inhibition increases TMZ delivery through caveolae-mediated transcytosis.

Discussion

In contrast to tremendous progress in improving the outcome for most other non-CNS tumors, the prognosis for glioblastoma remains dire.³ The BBB is one of the major rate-limiting factors for effective GBM therapy.³ Since Wnt signaling is the key pathway regulating BBB formation and function, we reasoned that the Wnt pathway could serve as a therapeutic target to improve drug delivery and penetration in glioblastoma. Using Wnt signaling inhibitor and transgenic mouse models, we uncovered that Wnt signaling inhibition could increase TMZ delivery through down-regulation of Mfsd2a, thus enhancing caveolae-mediated transcytosis (Figure 6H). We showed that LGK974 treatment increased TMZ delivery and nano-particle penetration in the tumors but not in healthy brains, indicating that the homeostatic BBB in the adult brain may be less sensitive to Wnt signaling perturbation. Indeed, decreased Wnt signaling as a result of conditional knockout of Gpr124, a coactivator of Wnt7a/7b-stimulated canonical Wnt signaling, in adult mice did not affect physiological BBB integrity, but led to profound BBB disruption in pathological conditions including ischemic stroke and glioblastoma.¹⁶ Therefore, manipulation of Wnt signaling to a decreased level could be a promising strategy to improve chemotherapy or nano- drugs delivery with minimal adverse effects.

We uncovered that Mfsd2a is sharply down-regulated by LGK974 treatment in tumor ECs, which is supported by a recent study showing that Mfsd2a transcription is directly regulated by Wnt signaling through the binding of β-catenin to its promoter region site.²³ Moreover, Mfsd2a deficiency phenocopy Wnt inhibition mediated upregulation of TMZ delivery, together with the fact that enhanced TMZ delivery was completely reversed by constitutive expression of Mfsd2a in ECs, indicating that Wnt-mediated regulation of TMZ delivery is dependent on Mfsd2a. Mfsd2a, a docosahexaenoic acid transporter, is specifically expressed in CNS endothelial cells.²⁸Mfsd2a maintains BBB and blood-retina barrier (BRB) homeostasis by establishing a unique lipid environment that suppresses caveolae-mediated EC transcytosis.^23,27

In accordance with the role of Mfsd2a in suppressing EC transcytosis, Wnt signaling inhibition by LGK974 treatment induced robust upregulation of Plvap, which is a marker of high level of vascular transcytosis.²⁹ Plvap promotes endothelial transcytosis and vascular permeability by forming cap-like diaphragms that bridge the opening of fenestrae, transendothelial channels and caveolae.²⁹ Accordingly, we demonstrated that LGK974 treatment led to an elevated caveolae-mediated EC transcytosis. In addition, enhanced TMZ delivery in tumors in Mfsd2a knockout mice or by Wnt inhibition was reversed in Cav1 knockout mice, indicating regulation of Wnt signaling and Mfsd2a on TMZ delivery is dependent on caveolae-mediated EC transcytosis. Caveolae is formed by the invagination of lipid rafts, which could non-specifically shuttle proteins or other substrates.³⁰ Caveolae-mediated transcytosis is active in the ECs from peripheral organs, including lung and heart, but suppressed in the CNS ECs.²⁸ Our genetic experiments demonstrate that elevated caveolae transcytosis accounts for the enhanced TMZ delivery in tumors upon Wnt inhibition or in the absence of Mfsd2a.

Several types of immunotherapy, including checkpoint inhibition, adoptive immune cell transfer and CAR-T, have been evaluated in clinical trials for GBM, but resulted in very little benefit. Failure to traffic and recruit tumor-targeted immune cells is likely one of the major reasons for the limited survival improvement of immunotherapy.³¹ We observed that Wnt inhibition also led to the up-regulated expression of inflammatory genes which may indicate regulations of immune cell recruitment to the tumor. Therefore, testing whether Wnt inhibition enhances the efficiency of immunotherapy in GBM will be worth exploring in future studies.

In addition to regulation of vascular transcytosis, Wnt/β-catenin signaling, activated by tumor cell-derived HGF, induces stemness-like transformation and mesenchymalization of endothelial cells, leading to endothelial cell chemo-resistance in GBM.³² Moreover, Wnt signaling contributes to single tumor cell dissemination and invasion.^19,33,34Wnt5a promote glioblastoma stem-like cell invasion,³³ and induces trans-differentiation of glioblastoma stem-like cell to endothelial cells, which promote peri-tumoral satellite lesions and supports the growth of invasive tumor cells away from the primary tumor.³⁴Wnt7, derived from glioma cells, is essential for Olig2 + oligodendrocyte precursor-like glioma cells invasion by single-cell vessel co-option.¹⁹

In summary, our study provides mechanistic insights into the regulation of chemotherapy drug delivery by Wnt signaling. Wnt inhibition enhances TMZ delivery through down-regulation of Mfsd2a, then enhancing caveolae-mediated transcytosis. The fact that Wnt signaling inhibition sensitizes tumors to TMZ treatment gives important therapeutic implications, and supports further investigation of the synergistic effect of Wnt inhibitors and TMZ in clinical trials. These and other studies showing the key role of Wnt signaling on tumor cell dissemination and stemness-like transformation of tumor ECs suggest that targeting Wnt signaling may offer promising opportunities for successful GBM treatment.

Funding

This work was supported by the National Natural Science Foundation of China (NSFC)/the Swedish Foundation for International Cooperation in Research and Higher Education (STINT) Mobility Program (No. 81911530166), the NSFC (No. 81702489, 82002659, 81870978, 81772661), the Swedish Cancer Foundation and Swedish Science Council, the Natural Science Foundation of Shaanxi Province (No. 2021KW-46, 2020JZ-30), the Natural Science Foundation of Hunan Province (No. 2022JJ50300).

Authorship statement

Y.X., L.H., Y.Z., H.H. designed and performed research, collected, analyzed, and interpreted data, and wrote the manuscript; F.Y., J.W performed bioinformatic analysis; M.C., H.C. and Y.L. performed research, collected data; Lingxue Z. performed electron microscopy analysis and interpreted data; L.X., B.X., X.S. and X.Z. performed animal experiments; L.U. provided RCAS glioma model and interpreted data; J.T. and A.D. interpreted data; L.W. and Lei Z. designed research, interpreted data, wrote the manuscript, and supervised the study.