1/11 Our study of pathological fibroblasts across inflammatory diseases, an incredibly collaborative effort with Kevin Wei, @MathildePohin, @baroneExpMed, @EdyKim5, @FionaPowrie, @arthritisbham, Michael Brenner, @soumya_boston, and many others! http://bit.ly/2XyAbLR 

2/11 Fibroblasts are active in multiple inflammatory diseases. We wondered if we could find the same pathological states expanded in inflamed salivary gland, lung, synovium, and intestine. To start, we generated nearly 80,000 novel scRNAseq fibroblast profiles from over 70 donors

3/11 We first analyzed each tissue separately and labeled fibroblast clusters according to previously identified states. Unfortunately, it was not clear whether these clusters were similar or distinct across tissues.

4/11 We pooled all the fibroblasts together and performed an integrated clustering analysis, building on our Harmony pipeline to now handle imbalanced numbers of cells among tissues.

5/11 We identified tissue-shared and tissue-specific cluster markers with GLMM regression, fitting effects for clusters, donors, and tissues into one model. Not all clusters had shared markers across all tissue. Here are five that did:

6/11 We next asked which clusters were expanded in inflamed tissues. To keep it simple, we defined inflammation as the percent of immune cells in each sample, normalized to a common scale.

7/11 We used logistic regression to associate the number of fibroblast subtypes in each tissue with our standardized inflammation score. Combining these results, we found that only CXCL10+CCL19+ and SPARC+COL3A1+ fibroblasts were consistently expanded in all inflamed tissues.

8/11 Pathway enrichment suggests distinct roles for these expanded fibroblasts: CXCL10+CCL19+ fibroblasts respond to inflammatory cytokines and regulate immune cell function, while SPARC+COL3A1+ fibroblasts respond to morphogens and play roles in tissue remodeling.

9/11 We wondered whether our cross-tissue atlas could be a useful tool to analyze fibroblasts from other organ systems. We used Symphony query mapping (by @joycebkang) to map skin fibroblasts, from healthy donors and patients with atopic dermatitis (aka eczema), to our atlas.

10/11 We used the same strategy to map mouse fibroblasts from inflamed joint, intestine, and lung tissue. By inferring mouse fibroblast labels with Symphony, we reasoned about which aspects of human fibroblast pathology are evident in these animal models of human disease.

11/11 We hope that our fibroblast atlas will serve as a single-cell reference atlas for all fibroblast studies and help build a common, cross-tissue taxonomy of fibroblast states. Look out for a live launch of our mappable stromal atlas, coming soon!