Chaichana, Kaisorn L.

Abstract INTRODUCTION Lung cancer is the most common tumor to metastasize to the brain, as 40% of patients develop brain metastasis (BM). Once diagnosed, survival is only 6.5-10 months. A potential target to prevent BM is circulating tumor cells (CTCs). The mechanisms that mediate a CTCs adaptation to the harsh environment of circulation and BM have not been elucidated. We have demonstrated that primary lung cancer cells (LCCs) subjected to circulation have increased migration across brain endothelial cell monolayers, increased side population, and decrease survival in rodents when injected intracardially vs plated controls. METHODS We engineered a microfluidic device capable of subjecting tumor cells to prolonged circulatory shear stress (CSS) that allows the isolation/modeling of CTC-like cells. Primary LCCs (A549, H1563) were subjected to circulation, suspension, or plated (2D) for 72h as controls. At this time point, LCCs were collected, and total mRNA was extracted. Samples were submitted for RNA sequencing. Weighted gene co-expression network analysis (WCGNA) and Ingenuity Pathway Analysis were performed. (FC >2 and p-value <0.05). RESULTS Transcriptomic analysis indicates that cells subjected to CSS activate specific gene networks involved in cell survival, migration, invasion, adhesion of tumor cells, and cytoskeletal reorganization (z-score: >2, p-value: 0.05). Canonical pathways in CTCs included ILK-signaling and Xenobiotic Metabolism. LCCs exposed to CSS had increased expression of multidrug resistance pumps ABCC2, ABCF2 (P:<0.001), and side population on flow cytometry (P: <0.05). The top differentially expressed genes in CTC-like cells (MMP1, ANGPTL4, and CEMIP) were involved in the invasion, angiogenesis, and pre-metastatic niche (PMN) formation of BM. CONCLUSION LCCs that survive CSS have increased expression of genes associated with migration, invasion, and chemoresistance compared to 2D controls. Targeting genes in CTCs that mediate survival and preparation of the PMN may be useful in preventing BM.

2071 Background: Despite advances in our understanding of brain cancer, the ability to discern between infiltrated and non-infiltrated tissue with current adjuncts, remains elusive. Isocitrate dehydrogenase (IDH) mutation is unique to IDH-mut glioma cells and can be used as a surrogate for tumor infiltration. Further, emerging evidence has shown unique signatures in brain cancer lipidomic profiling, with lipid metabolism playing a key role in brain cancer biology. Here, we aimed to evaluate the impact of desorption electrospray ionization-mass spectrometry (DESI-MS) as a near real-time method to determine infiltration status and the use of DESI-MS to perform intra-operative histological diagnosis. Methods: This is a prospective multi-institution cohort study using intra-operative DESI-MS on freshly obtained tumor samples to evaluate 2-hydroxyglutarate (2HG), with post-hoc validation by review by a board-certified neuropathologist. For tumor cell estimation and tissue diagnosis, high-throughput DESI-MS was used for rapid analysis of high-density arrays using lipidomic profiling. Analysis of the data was performed using a pre-written algorithm employing a combination of Python- and MATLAB-based custom software allowing results in less than a second per sample. Results: A total of 496 tumor biopsies from 72 patients have been analyzed. Earlier iterations focused on absolute quantitation of 2HG in 247 biopsies from 49 patients to determine IDH genotype yielding sensitivity, specificity, and accuracy values of 89, 100, and 94% for core biopsies (71). Improvement in accuracy was obtained with relative measurement of 2HG in comparison to glutamate allowing for detection of IDH-mut infiltration at the margin, resulting in 93% sensitivity, 100% specificity, and 98% accuracy with turnover times averaging 3 minutes in 183 biopsies from 23 patients. Margin assessment beyond the limits of neuro-navigation and surgeon assessment showed tumor infiltration in 88% of margin biopsies from IDH mutant tumors (30 of 34 biopsies) with 100% accuracy. Intra-operative tissue diagnosis was pursued in 66 unique tumor samples using DESI-MS lipidomic profiling, including samples of normal brain parenchyma, gliomas, meningiomas and pituitary adenomas, split into two microarray sets (36 and 40 samples). Using estimated principal components as input features, high accuracy ( > 90%; ROC AUCs ≥ 0.89) was obtained for supervised tissue classification using simple machine learning algorithms. Conclusions: We present a novel system for intra-operative evaluation of tumor infiltration at the margin in gliomas and for intra-operative tissue diagnosis of brain cancer. Our method identified tumor infiltration beyond the surgical margins and serves as a foundation for future uses employing DESI-MS to allow for molecularly guided surgery in IDH-mut gliomas.

In recent years, the endoscopic endonasal approach (EEA) for craniopharyngiomas has proven to be a safe option for extensive tumor resection, with minimal or no manipulation of the optic nerves and excellent visualization of the superior hypophyseal branches when compared to the Transcranial Approach (TCA). However, there is an ongoing debate regarding the criteria for selecting different approaches. To explore the current results of EEA and discuss its role in the management of craniopharyngiomas, we performed MEDLINE, Embase, and LILACS searches from 2012 to 2022. Baseline characteristics, the extent of resection, and clinical outcomes were evaluated. Statistical analysis was performed through an X2 and Fisher exact test, and a comparison between quantitative variables through a Kruskal–Wallis and verified with post hoc Bonferroni. The tumor volume was similar in both groups (EEA 11.92 cm3, -TCA 13.23 cm3). The mean follow-up in months was 39.9 for EEA and 43.94 for TCA, p = 0.76). The EEA group presented a higher visual improvement rate (41.96% vs. 25% for TCA, p < 0.0001, OR 7.7). Permanent DI was less frequent with EEA (29.20% vs. 67.40% for TCA, p < 0.0001, OR 0.2). CSF Leaks occurred more frequently with EEA (9.94% vs. 0.70% for TCA, p < 0.0001, OR 15.8). Recurrence rates were lower in the EEA group (EEA 15.50% vs. for TCA 21.20%, p = 0.04, OR 0.7). Our results demonstrate that, in selected cases, EEA for resection of craniopharyngiomas is associated with better results regarding visual preservation and extent of tumor resection. Postoperative CSF leak rates associated with EEA have improved compared to the historical series. The decision-making process should consider each person’s characteristics; however, it is noticeable that recent data regarding EEA justify its widespread application as a first-line approach in centers of excellence for skull base surgery.

Background: The complex nature and heterogeneity involving pituitary surgery results have increased interest in machine learning (ML) applications for prediction of outcomes over the last decade. This study aims to systematically review the characteristics of ML models involving pituitary surgery outcome prediction and assess their reporting quality. Methods: We searched the PubMed, Scopus, and Web of Knowledge databases for publications on the use of ML to predict pituitary surgery outcomes. We used the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) to assess report quality. Our search strategy was based on the terms “artificial intelligence”, “machine learning”, and “pituitary”. Results: 20 studies were included in this review. The principal models reported in each article were post-surgical endocrine outcomes (n = 10), tumor management (n = 3), and intra- and postoperative complications (n = 7). Overall, the included studies adhered to a median of 65% (IQR = 60–72%) of TRIPOD criteria, ranging from 43% to 83%. The median reported AUC was 0.84 (IQR = 0.80–0.91). The most popular algorithms were support vector machine (n = 5) and random forest (n = 5). Only two studies reported external validation and adherence to any reporting guideline. Calibration methods were not reported in 15 studies. No model achieved the phase of actual clinical applicability. Conclusion: Applications of ML in the prediction of pituitary outcomes are still nascent, as evidenced by the lack of any model validated for clinical practice. Although studies have demonstrated promising results, greater transparency in model development and reporting is needed to enable their use in clinical practice. Further adherence to reporting guidelines can help increase AI’s real-world utility and improve clinical practice.