Some patients infected with Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) develop severe pneumonia and the acute respiratory distress syndrome (ARDS) 1 . Distinct clinical features in these patients have led to speculation that the immune response to virus in the SARS-CoV-2-infected alveolus differs from other types of pneumonia 2 . We collected bronchoalveolar lavage fluid samples from 88 patients with SARS-CoV-2-induced respiratory failure and 211 patients with known or suspected pneumonia from other pathogens and subjected them to flow cytometry and bulk transcriptomic profiling. We performed single-cell RNA-seq on 10 bronchoalveolar lavage fluid samples collected from patients with severe COVID-19 within 48 hours of intubation. In the majority of patients with SARS-CoV-2 infection, the alveolar space was persistently enriched in T cells and monocytes. Bulk and single-cell transcriptomic profiling suggested that SARS-CoV-2 infects alveolar macrophages, which in turn respond by producing T cell chemoattractants. These T cells produce interferon-gamma to induce inflammatory cytokine release from alveolar macrophages and further promote T cell activation. Collectively, our results suggest that SARS-CoV-2 causes a slowly-unfolding, spatially limited alveolitis in which alveolar macrophages harboring SARS-CoV-2 and T cells form a positive feedback loop that drives persistent alveolar inflammation.

Lung transplantation can potentially be a life-saving treatment for patients with non-resolving COVID-19-associated respiratory failure. Concerns limiting lung transplantation include recurrence of SARS-CoV-2 infection in the allograft, technical challenges imposed by viral-mediated injury to the native lung, and the potential risk for allograft infection by pathogens causing ventilator-associated pneumonia in the native lung. Importantly, the native lung might recover, resulting in long-term outcomes preferable to those of transplant. Here, we report the results of lung transplantation in three patients with non-resolving COVID-19-associated respiratory failure. We performed single molecule fluorescent in situ hybridization (smFISH) to detect both positive and negative strands of SARS-CoV-2 RNA in explanted lung tissue from the three patients and in additional control lung tissue samples. We conducted extracellular matrix imaging and single cell RNA sequencing on explanted lung tissue from the three patients who underwent transplantation and on warm post-mortem lung biopsies from two patients who had died from COVID-19-associated pneumonia. Lungs from these five patients with prolonged COVID-19 disease were free of SARS-CoV-2 as detected by smFISH, but pathology showed extensive evidence of injury and fibrosis that resembled end-stage pulmonary fibrosis. Using machine learning, we compared single cell RNA sequencing data from the lungs of patients with late stage COVID-19 to that from the lungs of patients with pulmonary fibrosis and identified similarities in gene expression across cell lineages. Our findings suggest that some patients with severe COVID-19 develop fibrotic lung disease for which lung transplantation is their only option for survival.

Ontologically distinct populations of macrophages differentially contribute to organ fibrosis through unknown mechanisms.We applied lineage tracing, single-cell RNA sequencing and single-molecule fluorescence in situ hybridisation to a spatially restricted model of asbestos-induced pulmonary fibrosis.We demonstrate that tissue-resident alveolar macrophages, tissue-resident peribronchial and perivascular interstitial macrophages, and monocyte-derived alveolar macrophages are present in the fibrotic niche. Deletion of monocyte-derived alveolar macrophages but not tissue-resident alveolar macrophages ameliorated asbestos-induced lung fibrosis. Monocyte-derived alveolar macrophages were specifically localised to fibrotic regions in the proximity of fibroblasts where they expressed molecules known to drive fibroblast proliferation, including platelet-derived growth factor subunit A. Using single-cell RNA sequencing and spatial transcriptomics in both humans and mice, we identified macrophage colony-stimulating factor receptor (M-CSFR) signalling as one of the novel druggable targets controlling self-maintenance and persistence of these pathogenic monocyte-derived alveolar macrophages. Pharmacological blockade of M-CSFR signalling led to the disappearance of monocyte-derived alveolar macrophages and ameliorated fibrosis.Our findings suggest that inhibition of M-CSFR signalling during fibrosis disrupts an essential fibrotic niche that includes monocyte-derived alveolar macrophages and fibroblasts during asbestos-induced fibrosis.

Background: Large language models such as ChatGPT can produce increasingly realistic text, with unknown information on the accuracy and integrity of using these models in scientific writing. Methods: We gathered ten research abstracts from five high impact factor medical journals (n=50) and asked ChatGPT to generate research abstracts based on their titles and journals. We evaluated the abstracts using an artificial intelligence (AI) output detector, plagiarism detector, and had blinded human reviewers try to distinguish whether abstracts were original or generated. Results: All ChatGPT-generated abstracts were written clearly but only 8% correctly followed the specific journal's formatting requirements. Most generated abstracts were detected using the AI output detector, with scores (higher meaning more likely to be generated) of median [interquartile range] of 99.98% [12.73, 99.98] compared with very low probability of AI-generated output in the original abstracts of 0.02% [0.02, 0.09]. The AUROC of the AI output detector was 0.94. Generated abstracts scored very high on originality using the plagiarism detector (100% [100, 100] originality). Generated abstracts had a similar patient cohort size as original abstracts, though the exact numbers were fabricated. When given a mixture of original and general abstracts, blinded human reviewers correctly identified 68% of generated abstracts as being generated by ChatGPT, but incorrectly identified 14% of original abstracts as being generated. Reviewers indicated that it was surprisingly difficult to differentiate between the two, but that the generated abstracts were vaguer and had a formulaic feel to the writing. Conclusion: ChatGPT writes believable scientific abstracts, though with completely generated data. These are original without any plagiarism detected but are often identifiable using an AI output detector and skeptical human reviewers. Abstract evaluation for journals and medical conferences must adapt policy and practice to maintain rigorous scientific standards; we suggest inclusion of AI output detectors in the editorial process and clear disclosure if these technologies are used. The boundaries of ethical and acceptable use of large language models to help scientific writing remain to be determined.