Abstract Background The blood-brain barrier (BBB) is a major challenge to the treatment of CNS tumors, including diffuse midline glioma (DMG), a near universally fatal childhood brain cancer. Small molecule inhibitors of EZH2, a key catalytic subunit of the PRC2 complex driving DMG proliferation, have been proposed but have poor BBB penetration with additional acquired treatment resistance concerns. New proteolysis-targeting chimera (PROTAC) molecules use the ubiquitin-proteasome system to degrade oncoproteins like EZH2, disrupting both its catalytic and non-catalytic functions, and destabilize other PRC2 subunits (i.e. EED and SUZ12) to overcome the emergence of secondary mutations and treatment resistance. However, their macromolecular structures also limit BBB penetration. Methods We are advancing a clinically-compatible fucoidan (Fi) nanoparticle that can encapsulate several classes of anti-cancer drugs, homes to P-selectin on tumor vasculature after low-dose radiation (RT) to breach the BBB through active transcytosis and localize to tumor areas. We have used this strategy to Fi-encapsulate EZH2 inhibitor tazemetostat (EPZ-6438) and EZH2 PROTAC MS8815 (Fi-EPZ and Fi-MS8815, respectively). Optimal treatment regimen for tumor-specific delivery, efficacy, and side-effect profile will be determined in robust RCAS-tva DMG mouse models with a tight BBB. Results Mouse DMG models are amenable to Fi-based nanoparticle delivery as tumor vasculature exclusively expresses P-selectin, and that low-dose RT (2 Gy) enhances its expression further at 24h. Fi-EPZ and Fi-MS8815 form nanoparticles of uniform size and stability by zeta potential. We have been able to localize Fi-EPZ and Fi-MS8815 after a single RT priming dose specifically to the DMG tumor region while sparing normal brain. We hypothesize that Fi-encapsulation will improve therapeutic indices, while reducing systemic side-effects. Significance Our findings will provide the critical foundation for Fi-EPZ and Fi-MS8815 to be evaluated in clinical trials for children with DMG, and other brain tumors dependent by EZH2 and the PRC2 complex.

Abstract Breast cancer brain metastases (BCBMs) occur in 10-15% of all breast cancer patients with dismal median survival rates as low as 1 month for patients with leptomeningeal metastases (LMs). In BCBMs, EZH2, a histone methyltransferase, has been found to have non-enzymatic oncogenic function promoting metastatic proliferation and decreased overall survival. Thus, degradation versus inhibition of EZH2 in brain metastases could be a powerful strategy to treat BCBMs and LMs if a therapeutic could accumulate to therapeutic doses in the cerebrospinal fluid. In this study, we use an EZH2 proteolysis-targeting chimera (PROTAC) enabling degradation of EZH2 in BCBMs and LMs. We find that an EZH2 PROTAC administers cytotoxic efficacy in brain metastatic (Br) and leptomeningeal (LM) derivatives of breast cancer cell lines while inhibitors have no effect. Proteomics data and mitochondrial assays suggest this discrepancy can be attributed to mitochondrial complex I and electron transport chain dysfunction. PROTACs often have poor brain penetrance due to their hydrophobic nature and large molecular weight, but they can be encapsulated in P-selectin targeted nanoparticles (nanoPROTACs). We have previously shown that P-selectin can be presented selectively on primary brain tumor cells to deliver nanoparticles across the blood brain barrier in vivo. Here, we discover using humanized models of BCBMs that P-selectin is selectively presented on tumor and leptomeningeal vasculature. Only intraperitoneal delivery of P-selectin targeted EZH2 nanoPROTACs demonstrate remarkable degradation of EZH2 in LMs 48 hours after administration. This work demonstrates the ability to deliver PROTACs selectively to brain metastases and offers a therapeutic platform to extend BCBM patient survival. Citation Format: Raashed Raziuddin, Logan Hillger, Annie Ikemoto, Daniel Heller. Treating breast cancer brain metastases in the leptomeninges through p-selectin targeted delivery of EZH2 PROTACs [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1829.

Quantum well defect-modified single-walled carbon nanotubes are nanomaterials with wide-ranging applications in biosensing, imaging, quantum computing, and catalysis. The most common method for covalent functionalization of nanotubes for biosensing applications involves reactions with aryl diazonium salts to generate sp3 aryl defect sites, commonly followed by wrapping with single-stranded DNA for aqueous dispersion. We describe herein a rapid aryl diazonium functionalization reaction directly compatible with DNA-wrapped nanotubes, mediated by hydrogen peroxide. The reaction uses mild aqueous conditions at physiological pH and can be easily monitored in real-time via fluorescence analysis to control the degree of functionalization. Overall, this reaction greatly simplifies the production of covalently functionalized DNA-wrapped carbon nanotubes, expanding their potential for industrial and biomedical applications.

A major limiting factor in the success of chimeric antigen receptor (CAR) T cell therapy for the treatment of solid tumors is targeting tumor antigens also found on normal tissues. CAR T cells against GD2 induced rapid, fatal neurotoxicity because of CAR recognition of GD2 + normal mouse brain tissue. To improve the selectivity of the CAR T cell, we engineered a synthetic Notch receptor that selectively expresses the CAR upon binding to P-selectin, a cell adhesion protein overexpressed in tumor neovasculature. These tumor microenvironment actuated T (MEAT) cells ameliorated T cell infiltration in the brain, preventing fatal neurotoxicity while maintaining antitumor efficacy. We found that conditional CAR expression improved the persistence of tumor-infiltrating lymphocytes because of enhanced metabolic fitness of MEAT cells and the infusion of a less differentiated product. This approach increases the repertoire of targetable solid tumor antigens by restricting CAR expression and subsequent killing to cancer cells only and provides a proof-of-concept model for other targets.

Few chemical methods controllably generate sp³ defects on single-walled carbon nanotubes (SWCNTs) in such a way as to preserve or modulate their excitonic photoluminescence. We describe herein a new substrate-catalyzed functionalization reaction that enables conjugation of diverse small molecules lacking traditional SWCNT conjugation handles, resulting in quantum well defects with tunable near-infrared emission properties. This reaction is highly versatile – we describe herein the conjugation of over 200 unique small molecules including alcohols, amines, carbonyls, phenyl groups, alkyl chains, amino acids and peptides. The optical properties of the resulting nanotubes can be precisely tuned through the choice of conjugated molecule, as the electronic structure of the attached group governs the relative configuration of defects within the graphitic lattice. These molecularly driven effects allow fine-tuning of the emission wavelength, enabling rational design of nanotube-based fluorophores across the near-infrared window.

Mutations affecting the RNA splicing factors SF3B1, U2AF1, and SRSF2 are the most common class of genetic alterations in MDS. Patients with high-risk MDS have a median survival of 1-2 years and there is great need to develop novel therapies for MDS. Recent studies have engineered synthetic RNAs responsive to mutations in SF3B1 to express therapeutic payloads specifically in SF3B1 mutant cells and preferentially eliminate them. However, this approach relied on intronic sequences recognized by mutant SF3B1 and it is unclear if this strategy can be applied to higher risk mutations such as those in U2AF1, as U2AF1 is affected by both intronic and exonic sequence. To develop synthetic RNAs responsive to U2AF1 mutations, we identified endogenous genes that respond to the most common U2AF1mutations (those at the S34 residue). Analysis of RNA-seq data from 263 MDS/AML patients with U2AF1 S34 mutations or without any splicing factor mutations as well as isogenic AML cells with or without knockin of U2AF1 S34F or S34Y mutations, revealed a series of 3' splice sites consistently aberrantly spliced in U2AF1 mutant cells. This change in 3' splice site usage was validated in three mRNAs- MTA1, PRUNE1, and TRMT13- where U2AF1S34F/Y mutant MDS cells utilized a distal 3' splice site compared to U2AF1 wild-type (WT) cells. The entire endogenous intron of each of these three mRNAs and its alternative 3' splice sites were inserted to interrupt the cDNA sequence encoding the suicide gene HSV-TK. HSV-TK is inert in cells until the prodrug ganciclovir is provided which leads to cytotoxic metabolite production. Expression of the MTA1-HSV-TK construct in U2AF1S34F/Y mutant cells yielded full-length HSV-TK mRNA and protein. In contrast, U2AF1 WT cells expressed an incompletely spliced, non-functional mRNA and, as such, only U2AF1 mutant cells were killed with ganciclovir treatment while U2AF1 WT cells were unaffected. To enable massively parallel screening, optimize mutant selectivity, and allow for in vivo delivery, we next reduced the size of the natural 315 base pair MTA1 intron by ligating the minimal sequences at the 5' and 3' end to form a new 208 base pair synthetic intron. To improve mutant selectivity, we utilized SpliceAI, a deep learning model for splicing prediction. We evaluated 624 variants in silico using SpliceAI and identified intron variants that strengthen the intron-proximal 3' splice site in WT cells while reducing usage of the distal 3' splice site. Experimental validation identified a synthetic MTA1 intron allowing preferential killing of U2AF1S34F/WT and U2AF1S34Y/WT cells compared to WT cells. To further improve efficacy of the synthetic intron in mutant cells, we performed a massively parallel functional screen with 12,000 variant introns as a pool in U2AF1WT and U2AF1S34Y mutant K562 cells for unbiased identification of variants that eliminate mutant cells while leaving WT cells unaffected. We evaluated 6 of the identified introns and found one which yielded optimal ganciclovir-mediated killing in U2AF1 mutant cells while sparing WT cells. When isogenic WT or U2AF1 mutant AML cells bearing this construct were introduced into NSG mice, ganciclovir treatment improved survival of animals engrafted with U2AF1 mutant cells by a median of 20 days relative to the untreated group (p=0.0015; ganciclovir did not alter survival of mice engrafted with U2AF1 WT cells). We next evaluated lipid nanoparticles (LNPs) for in vivo delivery of U2AF1 mutant-selective synthetic introns. While LNPs are a clinically advanced RNA delivery vehicle, they yield minimal import of RNA cargo into the nucleus (the site of RNA splicing). We therefore tested delivery of plasmid DNA (allowing for nuclear import) in LNP formulations incorporating lipids that suppress STING-mediated inflammation which previously limited plasmid DNA delivery. This approach enabled nuclear expression in myeloid leukemia cells and is now being tested in vivo. This study identified a novel precision gene therapy targeting cells with MDS-associated mutations in U2AF1 by harnessing the neomorphic splicing activity of U2AF1 mutations. Moreover, the synthetic introns responsive to U2AF1 mutations created here enable future mechanistic studies to identify cis and trans factors required for mutant U2AF1 function. Finally, we investigate means for in vivo delivery of plasmid DNAs to myeloid cells which could have broader applications in gene therapy for myeloid neoplasms.

Quantum-well defects (QWDs) on single-walled carbon nanotubes (SWCNTs) enable expanded analyte sensitization and functionalization for biosensing applications. We are developing these materials into quantum well nanosensors (QWNS) to enable new research and diagnostic methods for cancers and other indications. We are finding new synthetic methods for QWD-modified SWCNTs (color center nanotubes, CCNTs, or quantum-well defect nanotubes (QWNTs)) in order to improve yield and simplify processing, as well as to diversify the sensitivities of QWNS. We are developing arrays of QWNS for the investigation and detection of cancers and other conditions using AI-enabled methods. Via machine learning algorithms, we built prediction models of nanosensor responses that can reliably identify several cancer and non-cancer conditions, and we have expanded this approach to indications without known biomarkers, providing a general method to identify clinical indications.

Single-stranded DNA-wrapped single-walled carbon nanotube (DNA-SWCNT) biosensors have emerged as highly sensitive tools for detecting biomolecular interactions, offering significant promise for disease and biomarker detection. However, the high costs of testing these nanosensors and the limited availability of real patient data highlight the need for a cost-effective optimization strategy. This study introduces a novel framework leveraging amino acids—key protein building blocks and metabolic intermediates—as proxies to optimize biosensor arrays. By incorporating cancer-related or hydrophobic properties of amino acids, we employ feature selection methods such as maximum relevance minimum redundancy (mRMR) and supervised learning algorithms with recursive feature elimination (RFE) to identify the optimal nanosensor array. This amino acid-based approach enables efficient identification of biologically relevant features that influence biosensor responses. The insights derived are further validated using a brain tumor biosensor dataset from real patients, where supervised learning and K-best feature importance rankings are applied to refine sensor selection for disease prediction. This framework provides a cost-effective method for optimizing nanosensor arrays in the absence of extensive patient-derived datasets and enhances interpretability by elucidating the role of specific amino acid properties in driving sensor responses.

The quantum well defect (QWD) modification of single-walled carbon nanotubes (SWCNTs), to produce color center nanotubes (CCNTs, or quantum-well defect nanotubes (QWNTs)) confer novel properties to SWCNT-based sensors to enable new applications. Quantum well nanosensors (QWNS) can be highly chemically diversified by a combination of both covalent (QWD) modification as well as non-covalent wrappings. We developed QWNS to detect specific protein analytes, to monitor disease processes in live cells, and for vitro diagnostics. For the latter, we developed a machine perception liquid biopsy (MPLB or MaPeLBx) platform consisting of an array of QWNS which respond to patient blood samples via photoluminescence modulations that are used to train machine learning algorithms. We have applied MPLB to detect both cancer and non-cancer indications.

BRAF remains one of the most important therapeutic targets in cancer, but BRAF inhibitors can cause "paradoxical" pathway activation and drug resistance through RAF dimerization. A clinical "paradox breaker" inhibitor of BRAF monomers and dimers can potentially evade drug resistance. However, patients are required to receive a high oral daily drug dose to achieve the target therapeutic window. Co-administration of a cytochrome P450 blocker can improve drug exposure, but the combination can lead to drug-drug interactions. We investigated delivery via fucoidan-based nanocarriers to improve pharmacologic properties. We found that the nanoparticles extended BRAF inhibition in cancer cells due to sequestration into endolysosomes, followed by controlled release from a lysosomal depot. Following intraperitoneal administration, nanoparticles improved drug pharmacokinetics in vivo without inhibiting cytochrome P450 and also resulted in substantial improvements in antitumor efficacy. This work describes a general nanotherapeutic strategy to improve the pharmacologic properties of drugs via intracellular depot formation.

The efficacy of fluorescence-guided surgery in facilitating the real-time delineation of tumours depends on the optical contrast of tumour tissue over healthy tissue. Here we show that CJ215-a commercially available, renally cleared carbocyanine dye sensitive to apoptosis, and with an absorption and emission spectra suitable for near-infrared fluorescence imaging (wavelengths of 650-900 nm) and shortwave infrared (SWIR) fluorescence imaging (900-1,700 nm)-can facilitate fluorescence-guided tumour screening, tumour resection and the assessment of wound healing. In tumour models of either murine or human-derived breast, prostate and colon cancers and of fibrosarcoma, and in a model of intraperitoneal carcinomatosis, imaging of CJ215 with ambient light allowed for the delineation of nearly all tumours within 24 h after intravenous injection of the dye, which was minimally taken up by healthy organs. At later timepoints, CJ215 provided tumour-to-muscle contrast ratios up to 100 and tumour-to-liver contrast ratios up to 18. SWIR fluorescence imaging with the dye also allowed for quantifiable non-contact wound monitoring through commercial bandages. CJ215 may be compatible with existing and emerging clinical solutions. A commercial near-infrared dye that is sensitive to apoptosis and that provides high tumour-to-muscle and tumour-to-liver contrast ratios facilitates fluorescence-guided tumour screening, tumour resection and the assessment of wound healing.

Abstract The blood-brain barrier (BBB) is one of the greatest barriers for the effective treatment of brain tumors, including H3K27-altered diffuse midline glioma (DMG), a near universally fatal childhood brain cancer. The BBB typically requires drugs to be given at maximally tolerated doses that are limited by systemic toxicities, particularly in settings of combination therapy. We have developed a clinically compatible fucoidan nanoparticle (Fi-NP) that homes to P-selectin on tumor vasculature after low-dose radiation (RT) to breach the BBB through an active caveolin-1-dependent mechanism and deliver several classes of targeted therapies. In non-CNS cancer xenograft models and a transgenic mouse model of SHH-driven medulloblastoma with an intact BBB, this approach improved survival while eliminating on-target systemic toxicities. We have now applied this approach to both brainstem and non-brainstem RCAS-TVA mouse models of DMG with an intact BBB. Specifically, we have found that DMG tumor vasculature expresses P-selectin, and that a single low-dose 2 Gy fraction of ionizing radiation further enhances it in a time-dependent manner up to 24 hours post-treatment. Importantly, we have also found that this P-selectin targeted drug delivery approach facilitates DMG tumor localization of Fi-NP encapsulated EZH2 inhibitor tazemetostat (EPZ-6438) as well as larger macromolecules including the PROTAC BET degrader dBET6, two promising targeted therapies for DMG limited by extremely poor BBB-penetration, while sparing drug delivery to non-tumor healthy brain regions. Work to measure PK, biodistribution, survival benefit, and drug target inhibition is ongoing. Our findings will provide the foundation for this P-selectin targeted Fi-NP approach to be evaluated in clinical trials for children with this lethal cancer.

While cancer research and care have benefited from revolutionary advances in the ability to manipulate and study living systems, the field is limited by a lack of synergy to leverage the power of engineering approaches. Cancer engineering is an emerging subfield of biomedical engineering that unifies engineering and cancer biology to better understand, diagnose, and treat cancer. We highlight cancer engineering’s unique challenges, the importance of creating dedicated centers and departments that enable translational collaboration, and educational approaches to arm a new generation of scientists with engineering expertise and a fundamental understanding of cancer biology to transform clinical cancer care.

Semiconducting single-walled carbon nanotubes (SWCNTs) transduce subtle changes in the microenvironment via modulation of their near-infrared fluorescence with potential applications in sensing and imaging. Controlled liquid filling in an endohedral volume of SWCNTs modifies the optical properties of SWCNTs. However, the effects of endohedral solvent dielectric on the nanotube fluorescence and their structure property relationship for biomedical applications are largely unexplored. In this talk, we discuss the environmental effects on SWCNT fluorescence to improve the spectral response of SWCNT-based imaging and sensing agents. We analyze a large dataset of hyperspectral single-tube images of diverse SWCNT structures with controlled endohedral filling. We investigate the correlations between SWCNT structure and molecular descriptors of filled molecules to explain spectral homogeneity at a single tube level as well as at an ensemble.

Given the tubal origin of high-grade serous ovarian cancer (HGSC), we sought to investigate intrauterine lavage (IUL) as a novel method of biomarker detection. IUL and serum samples were collected from patients with HGSC or benign pathology. Although CA-125 and HE4 concentrations were significantly higher in IUL samples compared to serum, they were similar between IUL samples from patients with HGSC vs benign conditions. In contrast, CA-125 and HE4 serum concentrations differed between HGSC and benign pathology (P =.002 for both). IUL and tumor samples from patients with HGSC were subjected to targeted panel sequencing and droplet digital PCR (ddPCR). Tumor mutations were found in 75 % of matched IUL samples. Serum CA-125 and HE4 biomarker levels allowed for better differentiation of HGSC and benign pathology compared to IUL samples. We believe using IUL for early detection of HGSC requires optimization, and current strategies should focus on prevention until early detection strategies improve.

Abstract Introduction: Glioblastoma multiforme (GBM) is the most common primary malignant brain tumor in adults. It is highly resistant to its current standard-of-care regimen, which includes surgical resection followed by adjuvant ionizing radiation and temozolomide. Therapy resistance can be attributed to the abundance of genomic alterations in GBM tumors that evolve over time, which contribute to intra-and inter-tumoral heterogeneities that render targeted therapies as inadequate treatment options. Thus, there is an urgent unmet need for effective therapies targeting primary and recurrent GBM. Here, we investigated the impact of ionizing radiation on the transcriptome of patient derived GBM tumor spheres to identify key alterations and biological pathways involved in therapy resistance. Materials and methods: We acquired four independent human GBM tumor sphere cell lines, which were derived from patients and cultured in serum-free media. Tumor spheres were treated with either a sham radiation or a single dose of 10Gy. Using poly-A enrichment whole transcriptome sequencing, we evaluated transcriptional alterations occurring in GBM tumors spheres at 96h post-radiation. To do so, we performed differential expression analysis and gene set enrichment analysis (GSEA) to identify top differentially expressed genes and significant changes in biological phenomena. Additionally, we evaluated the cell viability response of these GBM tumor spheres to radiation. Results and discussion: PCA analysis of the four human GBM tumor sphere cell lines demonstrates that cell type has a substantial impact on variation among samples compared to non-irradiated and irradiated treatment conditions. Upon analysis of genes in common that are either upregulated or downregulated followed by GSEA, we identified enrichment of genes associated with inflammatory responses and ferroptosis repression (e.g., NUPR1, PTGS1, AOX1), and depletion of genes involved in fatty acid metabolism (e.g., INSIG1, PLA2G3, ACAT2). Furthermore, the GBM tumor sphere cell viability responses allowed us to stratify our models into radiosensitive and radioresistant subgroups. Conclusion: Radiosensitive human GBM tumor spheres exhibit transcriptional alterations in genes linked to inflammation and fatty acid metabolism to a greater extent compared to radioresistant tumor spheres. Our study characterizes the radiation response of patient derived GBM models, which is to be leveraged for combating therapeutic resistance. Citation Format: Arianna Richelle Izawa-Ishiguro, Subhiksha Nandakumar, Nicholas Carbone, Annalisa V. Ferrotta, Raashed Raziuddin, Kristen C. Vogt, Daniel A. Heller, Ingo K. Mellinghoff. Ionizing radiation induces lipid metabolism-associated vulnerabilities in glioblastoma multiforme [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 387.

Abstract Serum biomarkers are often insufficiently sensitive or specific to facilitate cancer screening or diagnostic testing. In many cancers, including high-grade serous carcinoma (HGSC), biomarkers fail to detect early-stage cancer detection or to substantially impact mortality rates. We developed a perception-based sensing method that captures a biomarker-agnostic ‘disease fingerprint’ of HGSC using serum. Instead of measuring individual biomarkers, the method collects large data sets of molecular binding interactions to a diverse array of moderately-selective sensors, which were used to train machine learning algorithms. In an initial study using serum from 264 patients, we built a prediction model of nanosensor responses that reliably identified HGSC substantially better than CA125. We believe that the improvement of such a method, using larger nanosensor arrays, more sophisticated AI algorithms, and larger patient cohorts, could improve the detection of HGSC and facilitate biomarker discovery efforts. Citation Format: Daniel A. Heller, Mijin Kim, Lakshmi Ramanathan, Kara Long Roche. A liquid biopsy fingerprint of disease via nanoengineering [abstract]. In: Proceedings of the AACR Special Conference on Ovarian Cancer; 2023 Oct 5-7; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(5 Suppl_2):Abstract nr IA017.

Power cavitation imaging (PCI) is an emerging strategy for monitoring blood–brain barrier (BBB) opening, enabling spatial discrimination of cavitation intensity through power Doppler-analogous processing of passive cavitation images. Single-element cavitation detectors, while beneficial in signal monitoring for cavitation controllers, provide limited spatial information, conferring the need for an image-guided approach to ensure accurate cavitation localization and regulation. This study aims to evaluate the capability of PCI to spatially correlate real-time acoustic cavitation emissions with mechanical bioeffects as a predictor of P-selectin-targeted nanocarrier delivery. Preliminary sonoporation experiments were performed in vitro with brain tissue derived mouse endothelial cells and 3 kDa tetramethylrhodamine (TMR)-dextran. A focused ultrasound transducer (0.5 MHz, 5000 cycles/pulse, 5 Hz PRF, 30 second treatment duration) was used to sonicate cells in small volume (25 μl) cell suspension. Homogenous distribution of TMR-dextran in the cytosol and nucleus was observed via fluorescence microscopy at lower pressures (7% ± 2% fluorescent cells), while higher pressures had no significant differences above control (3% ± 2% vs 1% nontreated) due to microbubble destruction. Future experiments will correlate sonoporation with P-selectin expression and validate a linear array PCI system against contrast-enhanced MRI as a guide for P-selectin-targeted nanocarrier delivery in mouse models of medulloblastoma.