Microscopy is a central method in life sciences. Many popular methods, such as antibody labeling, are used to add physical fluorescent labels to specific cellular constituents. However, these approaches have significant drawbacks, including inconsistency; limitations in the number of simultaneous labels because of spectral overlap; and necessary perturbations of the experiment, such as fixing the cells, to generate the measurement. Here, we show that a computational machine-learning approach, which we call "in silico labeling" (ISL), reliably predicts some fluorescent labels from transmitted-light images of unlabeled fixed or live biological samples. ISL predicts a range of labels, such as those for nuclei, cell type (e.g., neural), and cell state (e.g., cell death). Because prediction happens in silico, the method is consistent, is not limited by spectral overlap, and does not disturb the experiment. ISL generates biological measurements that would otherwise be problematic or impossible to acquire.

Many tumors produce platelet-derived growth factor (PDGF)-DD, which promotes cellular proliferation, epithelial-mesenchymal transition, stromal reaction, and angiogenesis through autocrine and paracrine PDGFRβ signaling. By screening a secretome library, we found that the human immunoreceptor NKp44, encoded by NCR2 and expressed on natural killer (NK) cells and innate lymphoid cells, recognizes PDGF-DD. PDGF-DD engagement of NKp44 triggered NK cell secretion of interferon gamma (IFN)-γ and tumor necrosis factor alpha (TNF-α) that induced tumor cell growth arrest. A distinctive transcriptional signature of PDGF-DD-induced cytokines and the downregulation of tumor cell-cycle genes correlated with NCR2 expression and greater survival in glioblastoma. NKp44 expression in mouse NK cells controlled the dissemination of tumors expressing PDGF-DD more effectively than control mice, an effect enhanced by blockade of the inhibitory receptor CD96 or CpG-oligonucleotide treatment. Thus, while cancer cell production of PDGF-DD supports tumor growth and stromal reaction, it concomitantly activates innate immune responses to tumor expansion.

Haploinsufficiency of transcriptional regulators causes human congenital heart disease (CHD). However, underlying CHD gene regulatory network (GRN) imbalances are unknown.Here, we define transcriptional consequences of reduced dosage of the CHD-linked transcription factor, TBX5, in individual cells during cardiomyocyte differentiation from human induced pluripotent stem cells (iPSCs). We discovered highly sensitive dysregulation of TBX5dependent pathways-including lineage decisions and genes associated with cardiomyocyte function and CHD genetics-in discrete subpopulations of cardiomyocytes. GRN analysis identified vulnerable nodes enriched for CHD genes, indicating that cardiac network stability is sensitive to TBX5 dosage. A GRN-predicted genetic interaction between Tbx5 and Mef2c was validated in mouse, manifesting as ventricular septation defects. These results demonstrate exquisite sensitivity to TBX5 dosage by diverse transcriptional responses in heterogeneous subsets of iPSC-derived cardiomyocytes. This predicts candidate GRNs for human CHDs, with implications for quantitative transcriptional regulation in disease.

Obesity is an established risk factor for severe coronavirus disease 2019 , but the contribution of overweight and/or diabetes remains unclear. In a multicenter, international study, we investigated if overweight, obesity, and diabetes were independently associated with COVID-19 severity and whether the BMIassociated risk was increased among those with diabetes. RESEARCH DESIGN AND METHODSWe retrospectively extracted data from health care records and regional databases of hospitalized adult patients with COVID-19 from 18 sites in 11 countries. We used standardized definitions and analyses to generate site-specific estimates, modeling the odds of each outcome (supplemental oxygen/noninvasive ventilatory support, invasive mechanical ventilatory support, and in-hospital mortality) by BMI category (reference, overweight, obese), adjusting for age, sex, and prespecified comorbidities. Subgroup analysis was performed on patients with preexisting diabetes. Sitespecific estimates were combined in a meta-analysis. RESULTSAmong 7,244 patients (65.6% overweight/obese), those with overweight were more likely to require oxygen/noninvasive ventilatory support (random effects adjusted odds ratio [aOR], 1.44; 95% CI 1.15-1.80) and invasive mechanical ventilatory support (aOR, 1.22; 95% CI 1.03-1.46). There was no association between overweight and in-hospital mortality (aOR, 0.88; 95% CI 0.74-1.04). Similar effects were observed in patients with obesity or diabetes. In the subgroup analysis, the aOR for any outcome was not additionally increased in those with diabetes and overweight or obesity. CONCLUSIONSIn adults hospitalized with COVID-19, overweight, obesity, and diabetes were associated with increased odds of requiring respiratory support but were not associated with death. In patients with diabetes, the odds of severe COVID-19 were not increased above the BMI-associated risk.