Citations

AbstractBackgroundRecent advancements in Deep Learning (DL) methodologies have led to state‐of‐the‐art performance in a wide range of applications especially in object recognition, classification, and segmentation of medical images. However, training modern DL models requires a large amount of computation and long training times due to the complex nature of network structures and the large number of training datasets involved. Moreover, it is an intensive, repetitive manual process to select the optimized configuration of hyperparameters for a given DL network.PurposeIn this study, we present a novel approach to accelerate the training time of DL models via the progressive feeding of training datasets based on similarity measures for medical image segmentation. We term this approach Progressive Deep Learning (PDL).MethodsThe two‐stage PDL approach was tested on the auto‐segmentation task for two imaging modalities: CT and MRI. The training datasets were ranked according to similarity measures between each sample based on Mean Square Error (MSE), Peak Signal‐to‐Noise Ratio (PSNR), Structural Similarity Index (SSIM), and the Universal Quality Image Index (UQI) values. At the start of the training process, a relatively coarse sampling of training datasets with higher ranks was used to optimize the hyperparameters of the DL network. Following this, the samples with higher ranks were used in step 1 to yield accelerated loss minimization in early training epochs and the total dataset was added in step 2 for the remainder of training.ResultsOur results demonstrate that the PDL approach can reduce the training time by nearly half (∼49%) and can predict segmentations (CT U‐net/DenseNet dice coefficient: 0.9506/0.9508, MR U‐net/DenseNet dice coefficient: 0.9508/0.9510) without major statistical difference (Wilcoxon signed‐rank test) compared to the conventional DL approach. The total training times with a fixed cutoff at 0.95 DSC for the CT dataset using DenseNet and U‐Net architectures, respectively, were 17 h, 20 min and 4 h, 45 min in the conventional case compared to 8 h, 45 min and 2 h, 20 min with PDL. For the MRI dataset, the total training times using the same architectures were 2 h, 54 min and 52 min in the conventional case and 1 h, 14 min and 25 min with PDL.ConclusionThe proposed PDL training approach offers the ability to substantially reduce the training time for medical image segmentation while maintaining the performance achieved in the conventional case.

Abstract Histological and lineage immunofluorescence examination revealed that healthy conducting airways of humans and animals harbor sporadic poorly differentiated epithelial patches mostly in the dorsal noncartilage regions that remarkably manifest squamous differentiation. In vitro analysis demonstrated that this squamous phenotype is not due to intrinsic functional change in underlying airway basal cells. Rather, it is a reversible physiological response to persistent Wnt signaling stimulation during de novo differentiation. Squamous epithelial cells have elevated gene signatures of glucose uptake and cellular glycolysis. Inhibition of glycolysis or a decrease in glucose availability suppresses Wnt-induced squamous epithelial differentiation. Compared with pseudostratified airway epithelial cells, a cascade of mucosal protective functions is impaired in squamous epithelial cells, featuring increased epithelial permeability, spontaneous epithelial unjamming, and enhanced inflammatory responses. Our study raises the possibility that the squamous differentiation naturally occurring in healthy airways identified herein may represent “vulnerable spots” within the airway mucosa that are sensitive to damage and inflammation when confronted by infection or injury. Squamous metaplasia and hyperplasia are hallmarks of many airway diseases, thereby expanding these areas of vulnerability with potential pathological consequences. Thus, investigation of physiological and reversible squamous differentiation from healthy airway basal cells may provide critical knowledge to understand pathogenic squamous remodeling, which is often nonreversible, progressive, and hyperinflammatory.

Abstract Background Based on studies from animal models, growth differentiation factor-11 (GDF-11) may have rejuvenating effects in humans. GDF-11 has high sequence homology with GDF-8 (also known as myostatin); follistatin and follistatin-like protein-3 (FSTL-3) are inhibitory proteins of both GDF-8 and GDF-11. Methods Using highly specific liquid chromatography with tandem mass spectrometry assays for GDF-11 and GDF-8 and immunoassays for follistatin and FSTL-3, we quantified the association of these factors with muscle size, strength, and physical performance in 2 prospective cohort studies of community-dwelling older adults (Health, Aging, and Body Composition study [Health ABC] and Cardiovascular Health Study [CHS]). Results GDF-8 levels were positively associated with thigh muscle cross-sectional area and density in Health ABC (data not available in CHS). GDF-8 levels were positively associated with lean mass (a surrogate of muscle mass) in Health ABC but not CHS, and grip strength in CHS but not Health ABC. FSTL-3 (and perhaps follistatin) was negatively associated with lean mass and had variable associations with other variables. In contrast, GDF-11 was not significantly associated with strength or performance. Conclusions GDF-8 and its binding proteins, follistatin and FSTL-3, may constitute a counterregulatory system (chalones) to restrain age-related loss of muscle mass and strength.

Abstract Immune organ innervation by the peripheral nervous system enables autonomic function in the development and maintenance of immunity. Specifically, the role of thymic innervation in health and following neurological insults is unknown. An in depth identification of the connectome within the thymus and the extent to which nervous impulses delivered to the thymus affect T cell development and other biological processes within the thymus is needed. We sought to determine the exact nature of innervation within the thymus. Using both flow cytometry and confocal microscopy, we determined that the thymus contains both extrinsic neuronal processes (cell bodies outside of the thymus), and surprisingly, intrinsic neurons (cells bodies within the thymus). We further verified that subsets of neurons within the thymus express tyrosine hydroxylase, BIII tubulin, and/or NeuN, and are distinct from epithelial cells (EPCAM+, UEA1+, Ly51+) and tuft cells (DLCK1+). Furthermore, neuronal cell bodies found within the thymus did not express AIRE, consistent with their identity as neurons. Subsets of intrinsic neurons within the thymus express MHCII implying a potential role in T cell development. Importantly, using rabies virus retrograde and anterograde neurotracers, we mapped a novel, previously undescribed, connectome of neurons within the thymus. Finally, we determined that both epithelial cells and neurons are significantly reduced in RAG deficient mice, indicating a possible co-regulation between thymic neurons and epithelial cells. In summary, we describe a novel network of neurons within the thymus and put forward implications of the thymic connectome’s role in T cell development and establishment of thymic architecture.

Abstract Immune-mediated blood brain barrier (BBB) disruption is a prominent feature of various neurological conditions for which an emerging role of CD8 T cells is being realized. Our group has developed a unique model of CD8 T cell-mediated BBB disruption which employs Theiler’s murine encephalomyelitis virus (TMEV) infection. Upon intravenous administration of VP2121–130 viral peptide during the peak antiviral adaptive response, Db:VP2121–130 epitope specific CD8 T cells induce BBB disruption in a perforin dependent manner. In this model, we addressed the role of microglia in relation to antigen specific CD8+ T cells and permeable cerebral vessels. Using real time two-photon in vivo imaging, we demonstrate that microglia adopt distinct morphological features, including enlarged cell body and fewer ramified processes as early as 6 hours post administration of VP2 peptide. Notably, microglial population exhibit robust down-regulation of homeostatic markers including CX3CR1 during BBB disruption. This study demonstrates CD8 T cells can promote microglia activation in a perforin-dependent manner during BBB disruption. Importantly, microglial activation occurs concurrently with the onset of vascular permeability and could serve as a critical cell type in this process.

Abstract CD8 T cell responses are essential for clearance of infections in the CNS but contribute to neuropathology. T cell interaction with antigen presenting cells (APCs) at CNS barriers is thought to allow activated antigen-specific T cells to infiltrate the CNS during neuroinflammation. However, the role of CNS-resident APCs during CD8 T cell infiltration of the CNS is unclear. To address this, we utilized Theiler’s murine encephalomyelitis virus (TMEV) and Plasmodium berghei ANKA (PbA) to assess the role of APCs during CD8 T cell responses to CNS infection in mice. During infection, we observed robust upregulation of MHC class I on microglia and brain associated macrophages (BAM). Microglia and BAMs were located in close proximity to inflamed vasculature where antigen presentation to T cells could occur. Thus, we hypothesized that microglia and BAMs are critical APCs for CNS infiltration of antigen specific CD8 T cells. To test this, we generated mice with inducible deletion of MHC class I on microglia and BAMs by crossing CX3CR1creERT2 mice with H-2Db or H-2Kb floxed mice (Db cKO or Kb cKO). We found that brain infiltrating virus-specific CD8 T cells were reduced in Db cKO mice compared to controls during TMEV infection, while peripheral priming of T cells was unaffected. Conversely, antigen-specific CD8 T cell infiltration was not impacted in Kb cKO mice infected with TMEV-OVA257–264. In Db cKO mice, brain-infiltrating CD8 T cells were increased during PbA induced experimental cerebral malaria (ECM), indicating a potential inhibitory role for CNS APCs. These results demonstrate that local antigen presentation by microglia/BAMs can both positively and negatively regulate CD8 T cell infiltration of the brain in discrete neurologic disease models.

LBA661 Background: Liposomal irinotecan administered with 5-fluorouracil/leucovorin (5-FU/LV) is approved in the USA and Europe for mPDAC following progression with gemcitabine-based therapy. A phase 1/2 study (Wainberg et al. Eur J Cancer 2021;151:14–24; NCT02551991) demonstrated promising anti-tumor activity in patients with mPDAC who received first-line liposomal irinotecan 50 mg/m2 + 5-FU 2400 mg/m2 + LV 400 mg/m2 + oxaliplatin 60 mg/m2 (NALIRIFOX). Herein, we present results from NAPOLI-3 (NCT04083235), a randomized, open-label, phase 3 study investigating the efficacy and safety of NALIRIFOX compared with nab-paclitaxel + gemcitabine as first-line therapy in patients with mPDAC. Methods: Eligible patients with histopathologically/cytologically confirmed untreated metastatic PDAC were randomized (1:1) to receive NALIRIFOX on days 1 and 15 of a 28-day cycle or nab-paclitaxel 125 mg/m2 + gemcitabine 1000 mg/m2 (Gem+NabP) on days 1, 8 and 15 of a 28-day cycle. Randomization was stratified by ECOG performance status, geographic region and presence or absence of liver metastases. The primary endpoint was overall survival (OS); secondary endpoints were progression-free survival (PFS), overall response rate (ORR) and safety. OS was evaluated when ≥ 543 events were observed using a stratified log-rank test with an overall 1-sided significance level of 0.025. Results: Overall, 770 patients (NALIRIFOX, n = 383; Gem+NabP, n = 387) were included. Baseline characteristics were well balanced between arms. At a median follow-up of 16.1 months, 544 events had occurred. The median OS was 11.1 months in the NALIFIROX arm as compared with 9.2 months in the Gem+NabP arm (HR 0.84 [95% CI 0.71–0.99]; p = 0.04); PFS was also significantly improved (7.4 months vs 5.6 months; HR 0.70 [0.59–0.84]; p = 0.0001). Grade 3/4 treatment-emergent adverse events (TEAEs) with ≥ 10% frequency in patients receiving NALIRIFOX versus Gem+NabP included diarrhea (20.3% vs 4.5%), nausea (11.9% vs 2.6%), hypokalemia (15.1% vs 4.0%), anemia (10.5% vs 17.4%) and neutropenia (14.1% vs 24.5%). Conclusions: First-line NALIRIFOX demonstrated clinically meaningful and statistically significant improvement in OS and PFS compared with Gem+NabP in treatment-naïve patients with mPDAC. The safety profile of NALIRIFOX was manageable and consistent with the profiles of the treatment components. Funding: Funded by Ipsen. Clinical trial information: NCT04083235 .[Table: see text]

57 Background: Metastatic colorectal cancer (mCRC) with deficient DNA mismatch repair (dMMR) or microsatellite instability-high (MSI-H) show frequent and durable responses to programmed cell death -1 (PD-1) blockade. While the majority of these tumors are sporadics and observed in elderly patients, first-line data are limited to the KEYNOTE-177 study. Here, we report a consecutive series of elderly patients with dMMR mCRC treated with pembrolizumab in routine clinical practice that includes analysis of clinical outcome based on metastatic site. Methods: Patients with dMMR mCRC received first-line pembrolizumab (200 mg every 3 weeks) (N = 41) at Mayo Clinic and Mayo Clinic Health System (2015-2022). Clinicopathological features including metastatic site and molecular data ( BRAFV600E, KRAS) were analyzed in relationship to tumor response rate (RECIST version 1.1). Furthermore, these variables were also analyzed in relationship to patient progression-free survival (PFS) using Kaplan-Meier methodology and multivariable stepwise Cox regression. Results: Among patients with dMMR mCRC, 29/41 (70.7%) were female, 30 (78.9%) harbored BRAFV600E, and 32 (80%) were sporadic. Median age at start of treatment was 80.6 years (IQR of 75.9, 85.9) and median number of treatment cycles was 9 (IQR: 4, 20). At a median follow-up of 23 months (range, 3.0 to 88.5), patient median PFS was 21.3 months (95% CI: 6.4, 38.5). An overall response was observed in 48.8% (20 of 41) of patients including 13 (31.7%) complete (CR) and 7 (17.1%) partial responses (PR). The median duration of response was 42.2 months (95%CI: 8.6, NR). Patients with metastases that included liver had significantly poorer PFS compared to those with non-liver metastases (HRadj 3.40 (95%CI: 1.27-9.13); Padj= 0. 01). In this regard, responses (CR, PR) to pembrolizumab were observed in 3 (21.4%) patients with liver metastasis compared to 17 (62.9%) of those with non-liver metastases. Treatment-related grade 3 or 4 adverse events were observed in 9 (22.0%) patients of whom two discontinued therapy, and there was one treatment-related death. Conclusions: First-line therapy with pembrolizumab for dMMR mCRC significantly prolonged survival in an elderly patient population with manageable toxicity in routine clinical practice. Moreover, the survival benefit of pembrolizumab was significantly attenuated in patients with liver vs non-liver metastases.

559 Background: Cabozantinib, a multikanase inhibitor, is approved by the Food and Drug Administration (FDA) for treatment of advanced stage hepatocellular carcinoma (HCC) following progression on sorafenib. The CELESTIAL study which demonstrated benefit from cabozantinib following sorafenib included only patients with Child Pugh class A liver function. Recently, bevacizumab plus atezolizumab has replaced sorafenib as the first line treatment for HCC. It is unclear whether there is benefit from using Cabozantinib following progression on immunotherapy as CELESTIAL trial included very few patients who received prior immunotherapy. We describe the outcomes of individuals with advanced stage HCC treated with Cabozantinib following progression on immunotherapy. Methods: This was a multicenter, retrospective analysis of patients with advanced HCC diagnosed between 2010-2021 at Mayo Clinic Arizona, Florida, and Minnesota. Progression free survival (PFS) and median overall survival analysis were performed using Kaplan-Meier method. Results: We identified a total of 26 patients with advanced stage HCC who received cabozantinib following progression on immunotherapy. The median age of patients was 61 years (range, 39 – 81) and the majority were male (85%). Eighteen (72%) patients had Child Pugh A at initiation of cabozantinib therapy. Four (15%) patients received cabozantinib as second line therapy, the rest received it as third line or later. PFS on cabozantinib was 2.1 months (95% CI: 1.3 – 3.9) and median overall survival from time of cabozantinib initiation was 7.7 months (95% CI: 5.3 – 14.9). PFS was shorter with Child Pugh Class B, 1.3 months (95% CI: 0.9 – NE) compared to those with Child Pugh Class A, 2.1 months (95% CI: 1.5 – 4); although this difference was not significant (p=0.55). Common adverse events included fatigue (50%), anorexia (35%), aspartate aminotransferase (AST) elevation (35%), and diarrhea (31%); 7 (26%) patients experienced Grade 3 or greater adverse events including hypertension, diarrhea, anorexia, and bowel obstruction. Two patients discontinued cabozantinib due to side effects including one for uncontrolled blood pressure elevations and the other due to fatigue. Conclusions: The optimal sequencing of therapy for patients with advanced HCC following progression on immunotherapy is unknown. Treatment with cabozantinib may improve clinical outcomes following progression on immunotherapy with similar adverse event profile. This study included patients with Child Pugh class B liver disease (28%) and heavily pretreated population and demonstrated mPFS and mOS similar to other work examining outcomes of cabozantinib in real world setting after progression on sorafenib. Further studies to determine optimal treatment sequencing for this patient population are needed.

499 Background: The impact of socioeconomic status on gastrointestinal (GI) cancer mortality in the United States (US) is not well-established. We hypothesized that socially vulnerable populations have disproportionately higher mortality rates. Hence, we assessed the association of the social vulnerability index (SVI) with GI cancer mortality across US counties. Methods: Social vulnerability indices were obtained from the agency for toxic substances and disease registry (ATSDR) from 2014-2018 to compute percentile ranking scores (PRS: ranging from 0-1) for each US county. PRS were further categorized into quartiles (Q: 1st: 0-0.25 [least vulnerable]; 4th:0.75-1.00 [most vulnerable]). The wide-ranging online data for epidemiological research (WONDER) database was queried to abstract county-level age-adjusted mortality rates (AAMR) per 100,000 person-years (PY) for populations diagnosed with GI cancers. AAMRs were then linked with quartile rankings. Rate ratios (RR) of AAMRs between 4th and 1st Q were subsequently estimated with 95% confidence intervals using population-weighted, Poisson regression. Results: A total of 3142 counties were included in this analysis. The AAMR for overall deaths (OD) and premature deaths (PD; defined as death at age <65) in the GI cancer population was 53.8 and 17.7 per 100,000 PY. A gradient increase in GI cancer-related mortality was observed from 1st Q to 4th Q (OD: 49.2 vs 59.3; PD: 14.1 vs 21.3). This stepwise increase in AAMR over the quartiles was consistent across gender, different racial/ethnic subgroups, and rural/urban categories of counties. The AAMRs for OD were significantly higher in the 4th Q as compared to the 1st Q for gastric (RR: 1.67 [95% CI, 1.51-1.86]), hepatocellular including intrahepatic biliary (1.52 [1.45-1.61]), colorectal (1.21 [1.16-1.26]) and biliary (1.17 [1.06-1.28]) cancer. Similarly, significantly higher AAMRs for PD were observed in 4th Q vs1st Q for gastric (1.81 [1.60-2.05]), hepatocellular including intrahepatic biliary (1.78 [1.58-2.00]), biliary (1.64 [1.35-1.99]), colorectal (1.29 [1.21-1.39]) and pancreatic (1.16 [1.07-1.25]) cancer. However, no significant differences were observed for esophageal and small intestinal cancers. Men with gastric (1.58 [1.41-1.78]) and hepatocellular cancer (1.54 [1.43-1.66]), non-Hispanic Black population (1.54 [1.31-1.80]) with hepatocellular cancer, and Hispanic population with colorectal cancer (1.50 [1.31-1.72]) were observed to have higher overall mortality in the 4th Q compared to the 1st Q. The findings were similar for premature mortality. Conclusions: Population-level data suggests that the US counties with higher socio-economic adversities may be at an increased risk of GI cancer-related mortality. Investigations using patient-level data are required to probe the impact of socioeconomic vulnerabilities on cancer-related mortality.