The Application of Artificial Intelligence to Cancer Research: A Comprehensive Guide.

Advancements in AI have notably changed cancer research, improving patient care by enhancing detection, survival prediction, and treatment efficacy. This review covers the role of Machine Learning, Soft Computing, and Deep Learning in oncology, explaining key concepts and algorithms (like SVM, Naïve Bayes, and CNN) in a clear, accessible manner. It aims to make AI advancements understandable to a broad audience, focusing on their application in diagnosing, classifying, and predicting various cancer types, thereby underlining AI's potential to better patient outcomes. Moreover, we present a tabular summary of the most significant advances from the literature, offering a time-saving resource for readers to grasp each study's main contributions. The remarkable benefits of AI-powered algorithms in cancer care underscore their potential for advancing cancer research and clinical practice. This review is a valuable resource for researchers and clinicians interested in the transformative implications of AI in cancer care.

DeeP4med: deep learning for P4 medicine to predict normal and cancer transcriptome in multiple human tissues.

BACKGROUND: P4 medicine (predict, prevent, personalize, and participate) is a new approach to diagnosing and predicting diseases on a patient-by-patient basis. For the prevention and treatment of diseases, prediction plays a fundamental role. One of the intelligent strategies is the design of deep learning models that can predict the state of the disease using gene expression data. RESULTS: We create an autoencoder deep learning model called DeeP4med, including a Classifier and a Transferor that predicts cancer's gene expression (mRNA) matrix from its matched normal sample and vice versa. The range of the F1 score of the model, depending on tissue type in the Classifier, is from 0.935 to 0.999 and in Transferor from 0.944 to 0.999. The accuracy of DeeP4med for tissue and disease classification was 0.986 and 0.992, respectively, which performed better compared to seven classic machine learning models (Support Vector Classifier, Logistic Regression, Linear Discriminant Analysis, Naive Bayes, Decision Tree, Random Forest, K Nearest Neighbors). CONCLUSIONS: Based on the idea of DeeP4med, by having the gene expression matrix of a normal tissue, we can predict its tumor gene expression matrix and, in this way, find effective genes in transforming a normal tissue into a tumor tissue. Results of Differentially Expressed Genes (DEGs) and enrichment analysis on the predicted matrices for 13 types of cancer showed a good correlation with the literature and biological databases. This led that by using the gene expression matrix, to train the model with features of each person in a normal and cancer state, this model could predict diagnosis based on gene expression data from healthy tissue and be used to identify possible therapeutic interventions for those patients.

Bevacizumab-Induced Hypertension in Glioblastoma Patients and Its Potential as a Modulator of Treatment Response.

Glioblastoma invasion is the primary mechanism responsible for its dismal prognosis and is the direct result of interactions between glioblastoma cells and the tumor vasculature. The dysregulated microvasculature in glioblastoma tumors and vessels co-opted from surrounding brain tissue support rapid tumor growth and are utilized as pathways for invasive cancer cells. Attempts to target the glioblastoma vasculature with antiangiogenic agents (eg, bevacizumab) have nonetheless shown limited and inconsistent efficacy, and the underlying causes of such heterogeneous responses remain unknown. Several studies have identified that patients with glioblastoma who develop hypertension following treatment with bevacizumab show significant improvement in overall survival compared with normotensive nonresponders. Here we review these findings and discuss the potential of hypertension as a biomarker for glioblastoma treatment response in individual patients and the role of hypertension as a modulator of interactions between tumor cells and cells in the perivascular niche. We suggest that a better understanding of the actions of bevacizumab and hypertension at the cellular level will contribute to developing more effective personalized therapies that address glioblastoma tumor cell invasion.

Distinct Assemblies of Heterodimeric Cytokine Receptors Govern Stemness Programs in Leukemia.

Leukemia stem cells (LSC) possess distinct self-renewal and arrested differentiation properties that are responsible for disease emergence, therapy failure, and recurrence in acute myeloid leukemia (AML). Despite AML displaying extensive biological and clinical heterogeneity, LSC with high interleukin-3 receptor (IL3R) levels are a constant yet puzzling feature, as this receptor lacks tyrosine kinase activity. Here, we show that the heterodimeric IL3Rα/ßc receptor assembles into hexamers and dodecamers through a unique interface in the 3D structure, where high IL3Rα/ßc ratios bias hexamer formation. Importantly, receptor stoichiometry is clinically relevant as it varies across the individual cells in the AML hierarchy, in which high IL3Rα/ßc ratios in LSCs drive hexamer-mediated stemness programs and poor patient survival, while low ratios mediate differentiation. Our study establishes a new paradigm in which alternative cytokine receptor stoichiometries differentially regulate cell fate, a signaling mechanism that may be generalizable to other transformed cellular hierarchies and of potential therapeutic significance. SIGNIFICANCE: Stemness is a hallmark of many cancers and is largely responsible for disease emergence, progression, and relapse. Our finding that clinically significant stemness programs in AML are directly regulated by different stoichiometries of cytokine receptors represents a hitherto unexplained mechanism underlying cell-fate decisions in cancer stem cell hierarchies. This article is highlighted in the In This Issue feature, p. 1749.

Scaffold geometry modulation of mechanotransduction and its influence on epigenetics.

The dynamics of cell mechanics and epigenetic signatures direct cell behaviour and fate, thus influencing regenerative outcomes. In recent years, the utilisation of 2D geometric (i.e. square, circle, hexagon, triangle or round-shaped) substrates for investigating cell mechanics in response to the extracellular microenvironment have gained increasing interest in regenerative medicine due to their tunable physicochemical properties. In contrast, there is relatively limited knowledge of cell mechanobiology and epigenetics in the context of 3D biomaterial matrices, i.e., hydrogels and scaffolds. Scaffold geometry provides biophysical signals that trigger a nucleus response (regulation of gene expression) and modulates cell behaviour and function. In this review, we explore the potential of additive manufacturing to incorporate multi length-scale geometry features on a scaffold. Then, we discuss how scaffold geometry direct cell and nuclear mechanosensing. We further discuss how cell epigenetics, particularly DNA/histone methylation and histone acetylation, are modulated by scaffold features that lead to specific gene expression and ultimately influence the outcome of tissue regeneration. Overall, we highlight that geometry of different magnitude scales can facilitate the assembly of cells and multicellular tissues into desired functional architectures through the mechanotransduction pathway. Moving forward, the challenge confronting biomedical engineers is the distillation of the vast knowledge to incorporate multiscaled geometrical features that would collectively elicit a favourable tissue regeneration response by harnessing the design flexibility of additive manufacturing. STATEMENT OF SIGNIFICANCE: It is well-established that cells sense and respond to their 2D geometric microenvironment by transmitting extracellular physiochemical forces through the cytoskeleton and biochemical signalling to the nucleus, facilitating epigenetic changes such as DNA methylation, histone acetylation, and microRNA expression. In this context, the current review presents a unique perspective and highlights the importance of 3D architectures (dimensionality and geometries) on cell and nuclear mechanics and epigenetics. Insight into current challenges around the study of mechanobiology and epigenetics utilising additively manufactured 3D scaffold geometries will progress biomaterials research in this space.

Emerging biomaterials and technologies to control stem cell fate and patterning in engineered 3D tissues and organoids.

The ability to create complex three-dimensional cellular models that can effectively replicate the structure and function of human organs and tissues in vitro has the potential to revolutionize medicine. Such models could facilitate the interrogation of developmental and disease processes underpinning fundamental discovery science, vastly accelerate drug development and screening, or even be used to create tissues for implantation into the body. Realization of this potential, however, requires the recreation of complex biochemical, biophysical, and cellular patterns of 3D tissues and remains a key challenge in the field. Recent advances are being driven by improved knowledge of tissue morphogenesis and architecture and technological developments in bioengineering and materials science that can create the multidimensional and dynamic systems required to produce complex tissue microenvironments. In this article, we discuss challenges for in vitro models of tissues and organs and summarize the current state-of-the art in biomaterials and bioengineered systems that aim to address these challenges. This includes both top-down technologies, such as 3D photopatterning, magnetism, acoustic forces, and cell origami, as well as bottom-up patterning using 3D bioprinting, microfluidics, cell sheet technology, or composite scaffolds. We illustrate the varying ways that these can be applied to suit the needs of different tissues and applications by focussing on specific examples of patterning the bone-tendon interface, kidney organoids, and brain cancer models. Finally, we discuss the challenges and future prospects in applying materials science and bioengineering to develop high-quality 3D tissue structures for in vitro studies.

Characterising Distinct Migratory Profiles of Infiltrating T-Cell Subsets in Human Glioblastoma.

Glioblastoma is the most common and aggressive form of primary brain cancer, with no improvements in the 5-year survival rate of 4.6% over the past three decades. T-cell-based immunotherapies such as immune-checkpoint inhibitors and chimeric antigen receptor T-cell therapy have prolonged the survival of patients with other cancers and have undergone early-phase clinical evaluation in glioblastoma patients. However, a major challenge for T-cell-based immunotherapy of glioblastoma and other solid cancers is T-cell infiltration into tumours. This process is mediated by chemokine-chemokine receptor and integrin-adhesion molecule interactions, yet the specific nature of the molecules that may facilitate T-cell homing into glioblastoma are unknown. Here, we have characterised chemokine receptor and integrin expression profiles of endogenous glioblastoma-infiltrating T cells, and the chemokine expression profile of glioblastoma-associated cells, by single-cell RNA-sequencing. Subsequently, chemokine receptors and integrins were validated at the protein level to reveal enrichment of receptors CCR2, CCR5, CXCR3, CXCR4, CXCR6, CD49a, and CD49d in glioblastoma-infiltrating T-cell populations relative to T cells in matched patient peripheral blood. Complementary chemokine ligand expression was then validated in glioblastoma biopsies and glioblastoma-derived primary cell cultures. Together, enriched expression of homing receptor-ligand pairs identified in this study implicate a potential role in mediating T-cell infiltration into glioblastoma. Importantly, our data characterising the migratory receptors on endogenous tumour-infiltrating T cells could be exploited to enhance the tumour-homing properties of future T-cell immunotherapies for glioblastoma.

Simultaneous pancreas-kidney transplantation: Early complications and long-term outcomes - a single-center experience.

INTRODUCTION: This study aimed to assess the prevalence and severity of complications after simultaneous pancreas-kidney transplantation (SPKT) and to evaluate its influence on both grafts' long-term results. METHODS: This was an observational, retrospective study including 39 consecutive SPKT cases from 2000-2018. Complications were classified into kidney-related and pancreas-related. The severity of complications was assessed using the modified Clavien-Dindo scale. Kaplan-Meier curve analysis and log-rank tests were used. Cox regression was performed for the multivariate analysis. RESULTS: All 39 recipients had long-term type I diabetes. Twenty-one (53.8%) patients suffered a Clavien-Dindo ≥IIIa complication. Most complications were pancreas-related, with 17 (43.6%) patients suffering from one. Kidney-related major complications were seen in 11 (28.2%) patients. Patient survival at one, five, and 15 years was 89.7%, 87.1%, and 83.9%, respectively; kidney survival was 87.1%, 81.4%, and 73.6%, respectively; and pancreas survival was 76.9%, 71.3%, and 72%, respectively. Pancreas graft survival was influenced by the presence of major postoperative complications; patients and kidney graft survival were not. CONCLUSIONS: Complications after SPKT influence pancreas graft survival. Despite the high rate of complications, our results suggest that patient and kidney graft survival may not be affected by complications.

Regulating life after death: how mechanical communication mediates the epithelial response to apoptosis.

It is increasingly evident that cells in tissues and organs can communicate with one another using mechanical forces. Such mechanical signalling can serve as a basis for the assembly of cellular communities. For this to occur, there must be local instabilities in tissue mechanics that are the source of the signals, and mechanisms for changes in mechanical force to be transmitted and detected within tissues. In this review, we discuss these principles using the example of cell death by apoptosis, when it occurs in epithelia. This elicits the phenomenon of apical extrusion, which can rapidly eliminate apoptotic cells by expelling them from the epithelium. Apoptotic extrusion requires that epithelial cells detect the presence of nearby apoptotic cells, something which can be elicited by the mechanotransduction of tensile instabilities caused by the apoptotic cell. We discuss the central role that adherens junctions can play in the transmission and detection of mechanical signals from apoptotic cells.

Live Imaging of Apoptotic Extrusion and Quantification of Apical Extrusion in Epithelial Cells.

Apoptotic cell death eliminates unhealthy cells and maintains homeostatic cell numbers within tissues. Epithelia, which serve as fundamental tissue barriers for the body, depend on a physical expulsion of dying cells (apoptotic cell extrusion) to remain sealed and intact. Apoptotic cell extrusion has been widely studied over recent years, with researchers using various approaches to induce apoptotic cell death. Unfortunately, the majority of chemical-based approaches for cell death induction rely on sporadically occurring apoptosis and extrusion, making imagining lengthy, often unsuccessful, and difficult to capture in high-quality images because of the frequent frame sampling needed to visualise the key molecular processes that drive extrusion. Here, we present a protocol that describes steps needed for laser-mediated induction of apoptosis in a cell of choice, followed by imaging of apoptotic extrusion in confluent monolayers of epithelial cells. Moreover, we provide the description of a new approach involving the mixing of labelled and unlabelled cells. In particular, this protocol characterises how cells surrounding apoptotic cells behave, with high spatial and temporal resolution. This can be achieved without the optical interference that apoptotic cells cause as they are physically expelled from the monolayer and move out of focus for imaging. Finally, the protocol is accompanied by detailed procedures describing cell preparation for apoptotic extrusion experiments, as well as post-acquisition analysis required to evaluate rates of successful extrusion.

Enhanced RhoA signalling stabilizes E-cadherin in migrating epithelial monolayers.

Epithelia migrate as physically coherent populations of cells. Previous studies have revealed that mechanical stress accumulates in these cellular layers as they move. These stresses are characteristically tensile in nature and have often been inferred to arise when moving cells pull upon the cell-cell adhesions that hold them together. We now report that epithelial tension at adherens junctions between migrating cells also increases due to an increase in RhoA-mediated junctional contractility. We found that active RhoA levels were stimulated by p114 RhoGEF (also known as ARHGEF18) at the junctions between migrating MCF-7 monolayers, and this was accompanied by increased levels of actomyosin and mechanical tension. Applying a strategy to restore active RhoA specifically at adherens junctions by manipulating its scaffold, anillin, we found that this junctional RhoA signal was necessary to stabilize junctional E-cadherin (CDH1) during epithelial migration and promoted orderly collective movement. We suggest that stabilization of E-cadherin by RhoA serves to increase cell-cell adhesion to protect against the mechanical stresses of migration. This article has an associated First Person interview with the first author of the paper.

Cavin3 released from caveolae interacts with BRCA1 to regulate the cellular stress response.

Caveolae-associated protein 3 (cavin3) is inactivated in most cancers. We characterized how cavin3 affects the cellular proteome using genome-edited cells together with label-free quantitative proteomics. These studies revealed a prominent role for cavin3 in DNA repair, with BRCA1 and BRCA1 A-complex components being downregulated on cavin3 deletion. Cellular and cell-free expression assays revealed a direct interaction between BRCA1 and cavin3 that occurs when cavin3 is released from caveolae that are disassembled in response to UV and mechanical stress. Overexpression and RNAi-depletion revealed that cavin3 sensitized various cancer cells to UV-induced apoptosis. Supporting a role in DNA repair, cavin3-deficient cells were sensitive to PARP inhibition, where concomitant depletion of 53BP1 restored BRCA1-dependent sensitivity to PARP inhibition. We conclude that cavin3 functions together with BRCA1 in multiple cancer-related pathways. The loss of cavin3 function may provide tumor cell survival by attenuating apoptotic sensitivity and hindering DNA repair under chronic stress conditions.

When cells become cancerous they often stop making certain proteins. This includes a protein known as cavin3 which resides in bulb-shaped pits of the membrane that surrounds the cell called caveolae. These structures work like stress detectors, picking up changes in the membrane and releasing proteins, such as cavin3, into the cell's interior. Past studies suggest that cavin3 might interact with a protein called BRCA1 that suppresses the formation of tumors. Cells with mutations in the gene for BRCA1 struggle to fix damage in their DNA, and have to rely on other repair proteins, such as PARPs (short for poly (ADP-ribose) polymerases). Blocking PARP proteins with drugs can kill cancer cells with problems in their BRCA1 proteins. However, it was unclear what role cavin3 plays in this mechanism. To investigate this, McMahon et al. exposed cells grown in the laboratory to DNA-damaging UV light to stimulate the release of cavin3 from caveolae. This revealed that cavin3 interacts with BRCA1 when cells are under stress, and helps stabilize the protein so it can perform DNA repairs. Cells without cavin3 showed decreased levels of the BRCA1 protein, but compensated for the loss of BRCA1 by increasing the levels of their PARP proteins. These cells also had increased DNA damage following treatment with drugs that block PARPs, similar to cancer cells carrying mutations in the gene for BRCA1. These findings suggest that cavin3 helps BRCA1 to suppress the formation of tumors, and therefore should be considered when developing new anti-cancer treatments.

3D-printed microplate inserts for long term high-resolution imaging of live brain organoids.

BACKGROUND: Organoids are a reliable model used in the study of human brain development and under pathological conditions. However, current methods for brain organoid culture generate tissues that range from 0.5 to 2 mm of size, which need to be constantly agitated to allow proper oxygenation. The culture conditions are, therefore, not suitable for whole-brain organoid live imaging, required to study developmental processes and disease progression within physiologically relevant time frames (i.e. days, weeks, months). RESULTS: Here we designed 3D-printed microplate inserts adaptable to standard 24 multi-well plates, which allow the growth of multiple organoids in pre-defined and fixed XYZ coordinates. This innovation facilitates high-resolution imaging of whole-cerebral organoids, allowing precise assessment of organoid growth and morphology, as well as cell tracking within the organoids, over long periods. We applied this technology to track neocortex development through neuronal progenitors in brain organoids, as well as the movement of patient-derived glioblastoma stem cells within healthy brain organoids. CONCLUSIONS: This new bioengineering platform constitutes a significant advance that permits long term detailed analysis of whole-brain organoids using multimodal inverted fluorescence microscopy.

A Drug Screening Pipeline Using 2D and 3D Patient-Derived In Vitro Models for Pre-Clinical Analysis of Therapy Response in Glioblastoma.

Glioblastoma is one of the most common and lethal types of primary brain tumor. Despite aggressive treatment with chemotherapy and radiotherapy, tumor recurrence within 6-9 months is common. To overcome this, more effective therapies targeting cancer cell stemness, invasion, metabolism, cell death resistance and the interactions of tumor cells with their surrounding microenvironment are required. In this study, we performed a systematic review of the molecular mechanisms that drive glioblastoma progression, which led to the identification of 65 drugs/inhibitors that we screened for their efficacy to kill patient-derived glioma stem cells in two dimensional (2D) cultures and patient-derived three dimensional (3D) glioblastoma explant organoids (GBOs). From the screening, we found a group of drugs that presented different selectivity on different patient-derived in vitro models. Moreover, we found that Costunolide, a TERT inhibitor, was effective in reducing the cell viability in vitro of both primary tumor models as well as tumor models pre-treated with chemotherapy and radiotherapy. These results present a novel workflow for screening a relatively large groups of drugs, whose results could lead to the identification of more personalized and effective treatment for recurrent glioblastoma.

A deep convolutional neural network for segmentation of whole-slide pathology images identifies novel tumour cell-perivascular niche interactions that are associated with poor survival in glioblastoma.

BACKGROUND: Glioblastoma is the most aggressive type of brain cancer with high-levels of intra- and inter-tumour heterogeneity that contribute to its rapid growth and invasion within the brain. However, a spatial characterisation of gene signatures and the cell types expressing these in different tumour locations is still lacking. METHODS: We have used a deep convolutional neural network (DCNN) as a semantic segmentation model to segment seven different tumour regions including leading edge (LE), infiltrating tumour (IT), cellular tumour (CT), cellular tumour microvascular proliferation (CTmvp), cellular tumour pseudopalisading region around necrosis (CTpan), cellular tumour perinecrotic zones (CTpnz) and cellular tumour necrosis (CTne) in digitised glioblastoma histopathological slides from The Cancer Genome Atlas (TCGA). Correlation analysis between segmentation results from tumour images together with matched RNA expression data was performed to identify genetic signatures that are specific to different tumour regions. RESULTS: We found that spatially resolved gene signatures were strongly correlated with survival in patients with defined genetic mutations. Further in silico cell ontology analysis along with single-cell RNA sequencing data from resected glioblastoma tissue samples showed that these tumour regions had different gene signatures, whose expression was driven by different cell types in the regional tumour microenvironment. Our results further pointed to a key role for interactions between microglia/pericytes/monocytes and tumour cells that occur in the IT and CTmvp regions, which may contribute to poor patient survival. CONCLUSIONS: This work identified key histopathological features that correlate with patient survival and detected spatially associated genetic signatures that contribute to tumour-stroma interactions and which should be investigated as new targets in glioblastoma. The source codes and datasets used are available in GitHub: https://github.com/amin20/GBM_WSSM .

Mechanotransduction activates RhoA in the neighbors of apoptotic epithelial cells to engage apical extrusion.

Epithelia must eliminate apoptotic cells to preserve tissue barriers and prevent inflammation.1 Several different mechanisms exist for apoptotic clearance, including efferocytosis2,3 and apical extrusion.4,5 We found that extrusion was the first-line response to apoptosis in cultured monolayers and in zebrafish epidermis. During extrusion, the apoptotic cell elicited active lamellipodial protrusions and assembly of a contractile extrusion ring in its neighbors. Depleting E-cadherin compromised both the contractile ring and extrusion, implying that a cadherin-dependent pathway allows apoptotic cells to engage their neighbors for extrusion. We identify RhoA as the cadherin-dependent signal in the neighbor cells and show that it is activated in response to contractile tension from the apoptotic cell. This mechanical stimulus is conveyed by a myosin-VI-dependent mechanotransduction pathway that is necessary both for extrusion and to preserve the epithelial barrier when apoptosis was stimulated. Earlier studies suggested that release of sphingosine-1-phosphate (S1P) from apoptotic cells might define where RhoA was activated. However, we found that, although S1P is necessary for extrusion, its contribution does not require a localized source of S1P in the epithelium. We therefore propose a unified view of how RhoA is stimulated to engage neighbor cells for apoptotic extrusion. Here, tension-sensitive mechanotransduction is the proximate mechanism that activates RhoA specifically in the immediate neighbors of apoptotic cells, but this also must be primed by S1P in the tissue environment. Together, these elements provide a coincidence detection system that confers robustness on the extrusion response.

The Application of Deep Convolutional Neural Networks to Brain Cancer Images: A Survey.

In recent years, improved deep learning techniques have been applied to biomedical image processing for the classification and segmentation of different tumors based on magnetic resonance imaging (MRI) and histopathological imaging (H&E) clinical information. Deep Convolutional Neural Networks (DCNNs) architectures include tens to hundreds of processing layers that can extract multiple levels of features in image-based data, which would be otherwise very difficult and time-consuming to be recognized and extracted by experts for classification of tumors into different tumor types, as well as segmentation of tumor images. This article summarizes the latest studies of deep learning techniques applied to three different kinds of brain cancer medical images (histology, magnetic resonance, and computed tomography) and highlights current challenges in the field for the broader applicability of DCNN in personalized brain cancer care by focusing on two main applications of DCNNs: classification and segmentation of brain cancer tumors images.

Endothelial, pericyte and tumor cell expression in glioblastoma identifies fibroblast activation protein (FAP) as an excellent target for immunotherapy.

OBJECTIVES: Targeted immunotherapies such as chimeric antigen receptor (CAR)-T cells are emerging as attractive treatment options for glioblastoma, but rely on identification of a suitable tumor antigen. We validated a new target antigen for glioblastoma, fibroblast activation protein (FAP), by undertaking a detailed expression study of human samples. METHODS: Glioblastoma and normal tissues were assessed using immunostaining, supported by analyses of published transcriptomic datasets. Short-term cultures of glioma neural stem (GNS) cells were compared to cultures of healthy astrocytes and neurons using flow cytometry. Glioblastoma tissues were dissociated and analysed by high-parameter flow cytometry and single-cell transcriptomics (scRNAseq). RESULTS: Compared to normal brain, FAP was overexpressed at the gene and protein level in a large percentage of glioblastoma tissues, with highest levels of expression associated with poorer prognosis. FAP was also overexpressed in several paediatric brain cancers. FAP was commonly expressed by cultured GNS cells but absent from normal neurons and astrocytes. Within glioblastoma tissues, the strongest expression of FAP was around blood vessels. In fact, almost every tumor vessel was highlighted by FAP expression, whereas normal tissue vessels and cultured endothelial cells (ECs) lacked expression. Single-cell analyses of dissociated tumors facilitated a detailed characterisation of the main cellular components of the glioblastoma microenvironment and revealed that vessel-localised FAP is because of expression on both ECs and pericytes. CONCLUSION: Fibroblast activation protein is expressed by multiple cell types within glioblastoma, highlighting it as an ideal immunotherapy antigen to target destruction of both tumor cells and their supporting vascular network.