A deep learning framework identifies dimensional representations of Alzheimer’s Disease from brain structure

Heterogeneity of brain diseases is a challenge for precision diagnosis/prognosis. We describe and validate Smile-GAN (SeMI-supervised cLustEring-Generative Adversarial Network), a semi-supervised deep-clustering method, which examines neuroanatomical heterogeneity contrasted against normal brain structure, to identify disease subtypes through neuroimaging signatures. When applied to regional volumes derived from T1-weighted MRI (two studies; 2,832 participants; 8,146 scans) including cognitively normal individuals and those with cognitive impairment and dementia, Smile-GAN identified four patterns or axes of neurodegeneration. Applying this framework to longitudinal data revealed two distinct progression pathways. Measures of expression of these patterns predicted the pathway and rate of future neurodegeneration. Pattern expression offered complementary performance to amyloid/tau in predicting clinical progression. These deep-learning derived biomarkers offer potential for precision diagnostics and targeted clinical trial recruitment.

Identifying Parkinson’s disease subtypes with motor and non-motor symptoms via model-based multi-partition clustering

Identification of Parkinson’s disease subtypes may help understand underlying disease mechanisms and provide personalized management. Although clustering methods have been previously used for subtyping, they have reported generic subtypes of limited relevance in real life practice because patients do not always fit into a single category. The aim of this study was to identify new subtypes assuming that patients could be grouped differently according to certain sets of related symptoms. To this purpose, a novel model-based multi-partition clustering method was applied on data from an international, multi-center, cross-sectional study of 402 Parkinson’s disease patients. Both motor and non-motor symptoms were considered. As a result, eight sets of related symptoms were identified. Each of them provided a different way to group patients: impulse control issues, overall non-motor symptoms, presence of dyskinesias and pyschosis, fatigue, axial symptoms and motor fluctuations, autonomic dysfunction, depression, and excessive sweating. Each of these groups could be seen as a subtype of the disease. Significant differences between subtypes (P< 0.01) were found in sex, age, age of onset, disease duration, Hoehn & Yahr stage, and treatment. Independent confirmation of these results could have implications for the clinical management of Parkinson’s disease patients.

The Genetic, Environmental, and Immunopathological Complexity of Autoantibody-Negative Rheumatoid Arthritis

Differences in clinical presentation, response to treatment, and long-term outcomes between autoantibody-positive and -negative rheumatoid arthritis (RA) highlight the need for a better comprehension of the immunopathogenic events underlying the two disease subtypes. Whilst the drivers and perpetuators of autoimmunity in autoantibody-positive RA have started to be disclosed, autoantibody-negative RA remains puzzling, also due its wide phenotypic heterogeneity and its possible misdiagnosis. Genetic susceptibility appears to mostly rely on class I HLA genes and a number of yet unidentified non-HLA loci. On the background of such variable genetic predisposition, multiple exogeneous, endogenous, and stochastic factors, some of which are not shared with autoantibody-positive RA, contribute to the onset of the inflammatory cascade. In a proportion of the patients, the immunopathology of synovitis, at least in the initial stages, appears largely myeloid driven, with abundant production of proinflammatory cytokines and only minor involvement of cells of the adaptive immune system. Better understanding of the complexity of autoantibody-negative RA is still needed in order to open new avenues for targeted intervention and improve clinical outcomes.

Genome-Wide CRISPR-Cas9 Screen Identifies Rationally Designed Combination Therapies Relevant for CRLF2-Rearranged Ph-like ALL

Philadelphia chromosome-like acute lymphoblastic leukemia (Ph-like ALL, also known as BCR-ABL1-like ALL) is a disease entity of B-cell ALL that exhibits a gene expression profile, similar to that of Philadelphia chromosome-positive ALL. Ph-like ALL is categorized into two disease subtypes: "ABL-class"- and "CRLF2/JAK pathway"-types, both of which harbor gene alterations that constitutively activate cytokine/growth factor-related signals. Ph-like ALL with the CRLF2 rearrangement exhibits poor clinical outcomes and is the most common subtype of Ph-like ALL. Tyrosine kinase inhibitor (TKI)-based treatment regimens are effective in treating ABL-class type Ph-like ALL; however, no standard regimen has been established for the CRLF2/JAK pathway type. While multiple chemotherapeutic regimens, including Ruxolitinib monotherapy and/or its combination with chemotherapy, are being tested, their efficacy is reportedly limited. Thus, novel approaches are needed to treat CRLF2/JAK pathway type Ph-like ALL, in particular for CRLF2-rearranged (CRLF2-r) ALL. To identify molecules/pathways relevant for CRLF2-r ALL pathogenesis, we performed genome-wide CRISPR-Cas9 dropout screens in the presence or absence of Ruxolitinib using two IgH-CRLF2-r ALL lines (MUTZ5 and KOPN49) that differ in RAS mutational status. To do so, we employed a baboon envelope pseudotyped lentiviral vector system, which, for the first time, enabled highly efficient transduction of human B cell amenable for genome-wide CRSPR/Cas9 screens. While sgRNAs targeting CRLF2, IL7RA or JAK1/2 significantly affected cell fitness in both lines, those targeting STAT5A, STAT5B or STAT3 did not, suggesting that the JAK-STAT axis is largely dispensable for IgH-CRLF2-r ALL cell survival. We show that regulators of RAS signaling are critical for cell fitness and Ruxolitinib sensitivity and that CRKL and FLT3 depletion enhances Ruxolitinib sensitivity in RAS wild-type (WT) cells. Gilteritinib, a pan-tyrosine kinase inhibitor that reduces CRKL activity, effectively killed RAS WT IgH-CRLF2-r ALL cells in vitro and in vivo, either alone or combined with Ruxolitinib. We further show that combining Gilteritinib with Trametinib, a MEK1/2 inhibitor, is an effective means to target IgH-CRLF2-r ALL cells regardless of RAS mutational status. Our study delineates molecules/pathways relevant for CRLF2-r ALL pathogenesis and could suggest rationally designed combination therapies appropriate for disease subtypes. Disclosures No relevant conflicts of interest to declare.

915 Molecular characterization of AXL in solid tumor malignancies using real-world data

BackgroundThe receptor tyrosine kinase AXL is aberrantly expressed in many cancer types and associated with epithelial-to-mesenchymal transition (EMT), poor prognosis, and therapy resistance. To better understand the expression of this gene across specific disease subtypes, correlated pathways, and how certain therapies potentially modulate AXL expression, we investigated real-world clinical and molecular data across five solid tumor types before and after chemotherapy or immune checkpoint inhibitor (CPI) therapy.MethodsWhole transcriptome and exome sequencing were derived from patient tumor specimens obtained either prior to treatment or following chemotherapy or CPI therapies. RNA reads were mapped using STAR and data was normalized using transcripts per million. DNA reads were mapped using Novoalign and variants were called using Freebayes and Pindel. Clinical data was curated from multiple sources, QC’d and deidentified according to standard protocols. Five diseases were included: non-small cell lung cancer (NSCLC, n=1181), ovarian cancer (OV, n=300), urothelial carcinoma (UC, n=140), pancreatic ductal adenocarcinoma (PDAC, n=942), and skin cutaneous melanoma (SKCM, n=157). PD-L1 positivity was defined as ≥1% tumor cells with PD-L1 immunohistochemical staining at any intensity.ResultsAXL mRNA levels were highest in PDAC followed by NSCLC, SKCM, UC and OV. Within OV, AXL expression levels were higher in tumors pre-treated with chemotherapy relative to untreated. For other tumor types, chemotherapy or CPI pre-treated tumors had AXL mRNA levels comparable to untreated tumors. Copy number amplifications of AXL were rare across all tumor types (<3%) and did not associate with mRNA expression. Distinct molecular subtypes in several cancers displayed relatively high AXL mRNA levels, including the mesenchymal subtype in OV and the stromal rich subtypes in PDAC. AXL levels also correlated with an EMT mRNA signature across all tumors (rho=0.67). Further, higher AXL expression was associated with PD-L1 positivity in NSCLC, UC and PDAC (p<0.01), but not OV where only a few tumors were PD-L1 positive.Oncogenic KRAS mutations were associated with higher AXL expression in NSCLC and PDAC (p<0.001) and lower AXL expression in OV (p=0.01). Loss of KDM6A, known to induce tumorigenesis in PDAC, was associated with higher AXL expression in PDAC (p<0.01). Loss-of-function mutations in ARID1A, previously implicated as CPI sensitizing, were associated with lower AXL mRNA levels in OV tumors (p<0.001).ConclusionsAnalyses of real-world mRNA datasets showed that AXL was upregulated in specific tumor and treatment settings as well as in patient populations with specific mutations and disease subtypes. Findings here should be validated with independent datasets.