Myofibroblasts are a highly secretory and contractile cell phenotype that are predominant in wound healing and fibrotic disease. Traditionally, myofibroblasts are identified by the de novo expression and assembly of alpha-smooth muscle actin stress fibers, leading to a binary classification: “activated” or “quiescent (non-activated)”. More recently, however, myofibroblast activation has been considered on a continuous spectrum, but there is no established method to quantify the position of a cell on this spectrum. To this end, we developed a strategy based on microscopy imaging and machine learning methods to quantify myofibroblast activation in vitro on a continuous scale. We first measured morphological features of over 1000 individual cardiac fibroblasts and found that these features provide sufficient information to predict activation state. We next used dimensionality reduction techniques and self-supervised machine learning to create a continuous scale of activation based on features extracted from microscopy images. Lastly, we compared our findings for mechanically activated cardiac fibroblasts to a distribution of cell phenotypes generated from transcriptomic data using single-cell RNA sequencing. Altogether, these results demonstrate a continuous spectrum of myofibroblast activation and provide an imaging-based strategy to quantify the position of a cell on that spectrum.

No abstract

As global controllers of gene expression, small RNAs represent powerful tools for engineering complex phenotypes. However, a general challenge prevents the more widespread use of sRNA engineering strategies: mechanistic analysis of these regulators in bacteria lags far behind their high-throughput search and discovery. This makes it difficult to understand how to efficiently identify useful sRNAs to engineer a phenotype of interest. To help address this, we developed a forward systems approach to identify naturally occurring sRNAs relevant to a desired phenotype: RNA-seq Examiner for Phenotype-Informed Network Engineering (REFINE). This pipeline uses existing RNAseq datasets under different growth conditions. It filters the total transcriptome to locate and rank regulatory-RNA-containing regions that can influence a metabolic phenotype of interest, without the need for previous mechanistic characterization. Application of this approach led to the uncovering of six novel sRNAs related to ethanol tolerance in non-model ethanol-producing bacterium Zymomonas mobilis. Furthermore, upon overexpressing multiple sRNA candidates predicted by REFINE, we demonstrate improved ethanol tolerance reflected by up to an approximately twofold increase in relative growth rate compared to controls not expressing these sRNAs in 7% ethanol (v/v) RMG-supplemented media. In this way, the REFINE approach informs strainengineering strategies that we expect are applicable for general strain engineering.

Myofibroblasts are a highly secretory and contractile phenotype most commonly identified by the de novo expression and assembly of alpha-smooth muscle actin stress fibers. Traditionally, this activation process has been thought of as a binary process, with cells being labeled as 'activated' or 'quiescent (non-activated)'. More recently, this view has been expanded to consider activation on a continuous spectrum. However, there is no established method to quantify the position of a cell on this spectrum, and as a result, the binary labeling system is still widely used. While transcriptomic analyses provide a continuous measure of myofibroblast markers, a faster and more facile screening method is needed. To this end, we utilized optical microscopy and machine learning methods to quantify myofibroblast activation on a spectrum. We first measured size and shape features of over 1,000 individual cardiac fibroblasts and found that these features provide enough information to predict activation state, on the binary scale, with 94% accuracy as compared to manual classification. We next performed dimensionality reduction techniques on these features to create a continuous scale of activation. Importantly, this new classification system captures a range of fibroblast activation states, but still possesses inherent bias due to choice of morphological features. Thus, we next used self-supervised machine learning to create a second continuous labeling system free from biases associated with the manually measured features. Lastly, we compared our findings for mechanically activated cardiac fibroblasts to a distribution of cell phenotypes generated from transcriptomic data using single-cell RNA sequencing. Altogether, these results demonstrate a continuous spectrum of activation from fibroblast to myofibroblast and provide a strategy to quantify the position of a cell on that spectrum.