Washko, G R

Abstract Rationale Interstitial lung abnormalities (ILA) are abnormal findings on chest CT scans that may represent early-stages of interstitial lung disease. The presence of fibrosis (traction bronchiectasis and/or honeycombing) identifies those with ILA at an increased risk for mortality, however, determining who will progress remains a challenge, and the role of blood-based biomarkers in ILA risk-stratification is unknown. Methods Data from four prospective cohort studies were used to identify associations between plasma proteins measured with DNA aptamers (SomaScan) and ILA, fibrosis, and mortality. The Genetic Epidemiology of COPD Study (COPDGene) was used to identify protein associations, and findings were replicated by meta-analysis of the Age, Gene/Environment Susceptibility-Reykjavik study (AGES-Reykjavik), SubPopulations and InteRmediate Outcome Measures in COPD Study (SPIROMICS), and Framingham Heart Study (FHS). Protein associations with ILA and with fibrotic ILA compared to non-fibrotic ILA were performed using multivariable logistic regression; among those with ILA, Cox proportional hazards models were used to identify protein associations with mortality. Models were adjusted for age, gender, race, BMI, smoking status, pack-year smoking history, and for mortality, COPD. Benjamini-Hochberg false discovery rate p-values <0.05 were considered significant. Results 524 (10.7%) of 4,899 COPDGene participants had ILA, of whom 57 (10.9%) had fibrosis; of 8,146 AGES-Reykjavik, SPIROMICS and FHS participants included in the meta-analysis, 666 (8.2%) had ILA, of whom 198 (29.7%) had fibrosis. In COPDGene, 230 proteins were associated with ILA; the most significant of which were SFTPD (adjusted odds ratio per standard deviation [SD] of protein [OR]=2.21, p = 3.7E-41) and SFTPB (OR = 1.82, p = 7.9E-37). In the meta-analysis, 159 of the ILA-associated proteins were replicated. Of the 230 ILA-associated proteins in COPDGene, 9 were associated with fibrosis, and 5 of these 9 proteins were associated with mortality among those with ILA. The 5 proteins associated with all three outcomes were WFDC2 (ORILA=1.90, p = 2.6E-34; ORFibrosis=2.00, p = 0.0001; adjusted hazards of mortality per SD of protein [HR]=1.59, p = 3.4E-05), GDF15 (ORILA=1.67, p = 3.6E-23; ORFibrosis=1.78, p = 0.0006; HR = 1.63, p = 1.4E-06), SCGB3A1 (ORILA=1.52, p = 4.5e-14; ORFibrosis=1.80, p = 0.0002; HR = 1.25, p = 0.02), TREM1 (ORILA=1.44, p = 1.4e-14; ORFibrosis=1.62, p = 0.001; HR = 1.25, p = 0.02) ICAM5 (ORILA=1.42, p = 1.2e-11; ORFibrosis=1.79, p = 0.0001; HR = 1.31, p = 0.002). In the meta-analysis, 6 of the 9 fibrosis-associated proteins and all 5 ILA mortality-associated proteins replicated. Conclusion This represents the first plasma proteomic screen demonstrating reproducible biomarker associations not only with ILA presence, but also with measures of severity and adverse outcomes. These results provide compelling targets for further study, including their prospective evaluation as clinical biomarkers for early disease identification and prognostication. This abstract is funded by: Multiple Sources

Abstract Rationale Alpha-1 antitrypsin deficiency (AATD) continues to be underdiagnosed. The estimated prevalence of the most common severe deficient genotype, PI*ZZ, is ∼1:3500 or 100,000 Americans but only ∼10,000 such individuals have been recognized. Diagnostic delays persist and these delays are associated with worse disease. High-resolution CT (HRCT) imaging may contain latent phenotypic AATD signatures that are not visibly apparent to clinicians. We hypothesized a deep learning model could identify imaging patterns predictive of AATD from inspiratory HRCT scans. Methods Confirmed genotypes and HRCT scans were pooled from COPDGene, SPIROMICS, and GRADS, yielding 350 MM, 102 MZ, and 69 ZZ participants. Only scans reconstructed with smooth kernels (GE STANDARD, Philips B, Siemens B31f/B35f) and slice thickness < =1.25mm were used. The cohort was selected to have similar distributions of quantitative emphysema, sex, and age by matching five MZ and two MM subjects to each ZZ subject (<10-year age difference, same sex, and <3% difference in %LAA). Tobacco smoke-exposure was more prevalent among the MM/MZ groups given these were predominantly from the tobacco smoke-enriched COPDGene and SPIROMICS cohorts. A DenseNet121 convolutional neural network (CNN) was trained using the MONAI platform to predict genotype by outputting probabilities for MM, MZ, and ZZ. Input to the CNN was a three-channel LAA-910, LAA-950, and LAA-970 map downsampled to 2.0mm3 resolution to remove bias due to differences in scanner and image reconstruction parameters. Five-fold cross-validation was performed. Area under the curve (AUC) was calculated for classification of ZZ vs other genotypes (MM/MZ) and MZ vs MM, with statistical significance assessed at p < 0.05. Results The three groups were similar across age (mean±SD years: MM 57.9±8.8, MZ: 57.5±8.9, ZZ: 55.0±10.7), sex (% female: MM: 48.2%, MZ: 46.7%, ZZ: 47.6), and emphysema (LAA-950%: MM: 13.0±12.2, MZ: 12.4±11.8, ZZ: 13.1±11.9). There was a higher prevalence of ever smokers among MM (100%) and MZ (91.7%), compared to ZZ (47.6%). The model achieved an AUC of 0.806 (p < 0.001) for classification of ZZ vs other genotypes, demonstrating robust discriminatory ability for homozygous AATD. Classification of MZ vs other genotypes did not achieve an AUC significantly greater than 0.5. Conclusions This proof-of-concept study demonstrates that deep learning applied to routine inspiratory HRCT can accurately identify homozygous AATD. These findings support the potential role of imaging-based machine learning as a non-invasive screening tool for undiagnosed ZZ individuals. Future work will focus on cohort enrichment, integration with clinical parameters, and explainability of model-derived imaging features. This abstract is funded by: NIH NHLBI K23HL173570

Abstract Background Muscle loss in COPD results from both disuse atrophy and systemic inflammation. We sought to identify inflammatory biomarkers associated with loss of pectoralis muscle area in COPD using serial CT scans and peripheral blood proteomics. Design We analyzed five-year longitudinal data from COPDGene, an ongoing observational cohort of current and former tobacco smokers with or without airflow obstruction. We quantified pectoralis major muscle (PMM) area on chest CT scans using a 3D convolutional neural network (CNN) trained on CT volumes and annotations provided by the TotalSegmentator project. We compared participants with CT PMM data and plasma proteins by SOMAScan (1K platform) from Phase 2 and 3 (paired n = 1,191). The correlation coefficient (r) between longitudinal changes in protein and in PMM area was plotted against the false discovery rate (FDR)-adjusted p-values. Proteins significantly associated with muscle loss (blue) or gain (red) were visualized using volcano plots. We then analyzed a subset (paired n = 500) of individuals with BMI>25 in Phase 2 whose BMI declined in Phase 3. Proteomic models were adjusted for age and clinical center. Pathway analysis utilized G:Profiler. Results Of the participants with paired proteomic and PMM data, mean PMM area was 21.7 ± 17.3 cm² at Phase 2 and 21.3 ± 17.9 cm² at Phase 3, with an annual change in PMM area of -0.31 + 0.99 cm². Multiple proteins were associated with muscle loss from Phase 2 to 3 (including leptin, FABP, and growth hormone receptor) and muscle gain (including F172A, GMEB2, and RNF41) (Fig 1a). Significant pathways were related to cancer, growth factor receptor dysregulation, and PI3K/AKT signaling. When restricting to those with BMI>25 who experienced BMI decline at Phase 3, muscle loss was associated with increased leptin and PACAP, versus association of gain with increased GDF118, EF1β, and FJF6 (Fig 1b). In the entire group, leptin declined between Phase 2 to Phase 3 in n = 571 participants and increased or was unchanged in n = 620. Compared with leptin decline, unchanged or increased leptin was associated with significantly greater interval PMM loss (−2.25 ± 9.75 cm² vs + 0.98 ± 9.43 cm²; p < 0.001) (Fig 1c). Conclusion Loss of PMM over five years was associated with several plasma proteins, including FABP and growth hormone receptor. The top biomarker, leptin, is a pleiotropic adipokine that regulates energy balance and inflammation. Integrating imaging and proteomics may identify future therapeutic targets to mitigate skeletal muscle loss in COPD. This abstract is funded by: This work was supported by NHLBI grants U01 HL089897 and U01 HL089856 and by NIH contract 75N92023D00011