Assessing Transcriptome Quality in Patch-Seq Datasets

Abstract

Patch-seq, combining patch-clamp electrophysiology with single-cell RNA-sequencing (scRNAseq), enables unprecedented access to a neuron's transcriptomic, electrophysiological, and morphological features. Here, we present a re-analysis of five patch-seq datasets, representing cells from ex vivo mouse brain slices and in vitro human stem-cell derived neurons. Our objective was to develop simple criteria to assess the quality of patch-seq derived single-cell transcriptomes. We evaluated patch-seq transcriptomes for the expression of marker genes of multiple cell types, benchmarking these against analogous profiles from cellular-dissociation based scRNAseq. We found an increased likelihood of off-target cell-type mRNA contamination in patch-seq cells from acute brain slices, likely due to the passage of the patch-pipette through the processes of adjacent cells. We also observed that patch-seq samples varied considerably in the amount of mRNA that could be extracted from each cell, strongly biasing the numbers of detectable genes. We developed a marker gene-based approach for scoring single-cell transcriptome quality post-hoc. Incorporating our quality metrics into downstream analyses improved the correspondence between gene expression and electrophysiological features. Our analysis suggests that technical confounds likely limit the interpretability of patch-seq based single-cell transcriptomes. However, we provide concrete recommendations for quality control steps that can be performed prior to costly RNA-sequencing to optimize the yield of high-quality samples.

Human cortical expansion involves diversification and specialization of supragranular intratelencephalic-projecting neurons

“…Our prior report describing the t-type classification used here 11 demonstrated that t-types were robust across individuals and between acute neurosurgical and postmortem frozen tissues and could be validated in independent donors with multiplex fluorescence in situ hybridization (mFISH) panels derived from these data to confirm their laminar localization. However, neurons collected via Patch-seq can exhibit contamination not seen in dissociated cells 33 , arising from adjacent neurons and/or non-neuronal cells that enter the patch pipette. Furthermore, capturing only a portion of a neuron’s content could lead to increased variability and false negatives (dropouts).…”

“…Patch-seq cells were included in this data set if they met the following criteria. All neurons: 1) had high-quality transcriptomic data, measured as the normalized summed expression of “on”-type marker genes (NMS, adapted from the single-cell quality control measures in 33 ) greater than 0.4; and 2) retained a soma through biocytin processing and imaging such that an accurate laminar association could be made. In addition, mouse neurons were: 1) located within VISp; 2) either tdTomato- or tdTomato+ from a line known to label glutamatergic neurons (i.e.…”

Transcriptomic correlates of electrophysiological and morphological diversity within and across excitatory and inhibitory neuron classes

“…We reasoned that the recently developed PatchSeq methodology, allowing morphological, electrophysiological, and transcriptomic characterization from the same single cell, presents a unique opportunity to test this possibility [12]. While these data at present are limited by relatively modest sample sizes and technical factors such as inefficient mRNA capture and potential off-target cellular mRNA contamination [27], we nonetheless sought to use these data to assess the nature of within-cell type gene-property relationships.…”

Heterogeneous somatostatin-expressing neuron population in mouse ventral tegmental area

“…The PatchSeq method uses aspirated cell content samples, making it possible for contamination to occur as the recording pipette passes through other cells and processes ( Tripathy et al, 2018 ). However, our main conclusions from the PatchSeq data regarding the mixed neurochemical phenotypes of Sst neurons are in agreement with our in situ hybridization results ( Figure 6 ).…”