Authors’ Reply

• RNA-sequencing
• Cell dissociation
• fibrosis

We thank Ferenbach and colleagues for this important comment on the biases introduced by different single-cell dissociation methods during the single-cell RNA sequencing (scRNA-seq) workflow. An ideal dissociation protocol would generate a cell suspension that accurately reflects the abundance of each cell type in vivo, minimizes artifactual transcriptional responses induced by the dissociation procedure, avoids RNA degradation, and can be applied to cryopreserved tissue. Currently no single protocol accomplishes all of these goals.

Although we agree that nuclear preparations underrepresent leukocytes, enzymatic cell dissociation protocols overrepresent them. In our own human biopsy scRNA-seq dataset relying on enzymatic cell dissociation, 41.4% of the cells were leukocytes.¹ But it is unlikely that 40% (in some publications up to 60%) of the cells in a unilateral ureteral obstruction (UUO) kidney are leukocytes because this is typically measured by FACS, which relies on the same enzymatic dissociation protocols that bias against parenchymal cell representation. Indeed, Figure 2 from Eis et al.² used a histologic examination of F4/80-positive cells, and reported only 14 F4/80-positive cells per high power field, which is about 10% of the cells in that field and four-fold fewer than measured by FACS.

We speculate that nuclear preparations are not simply better at resolving glomerular cell types, but all parenchymal cell types that contain tight junctions and are surrounded by basement membrane. Enzymatic dissociation may lead these epithelia to form clumps that get filtered out of single-cell preparations. This is supported by the observation that our UUO³ dataset of comparable size identified 12 epithelial clusters compared with three epithelial clusters identified in the UUO dataset provided by O’Sullivan et al.⁴ We also point out the importance of artifactual stress responses induced by dissociation at 37°C when analyzing diseased kidney. In this case deconvoluting artifactual gene expression from appropriate cellular injury responses may be challenging. We agree that applying different dissociation strategies to the same tissue may help resolve the biases intrinsic to scRNA-seq versus single-nucleus RNA sequencing.

• Copyright © 2019 by the American Society of Nephrology