Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Correspondence
  • Published:

Single-cell sequencing reveals dissociation-induced gene expression in tissue subpopulations

This is a preview of subscription content, access via your institution

Relevant articles

Open Access articles citing this article.

Access options

Rent or buy this article

Prices vary by article type

from$1.95

to$39.95

Prices may be subject to local taxes which are calculated during checkout

Figure 1: Widely used tissue dissociation protocol induces transcriptional changes in a subpopulation of satellite cells.

Accession codes

Accessions

Gene Expression Omnibus

References

  1. Tierney, M.T. & Sacco, A. Trends Cell Biol. 26, 434–444 (2016).

    Article  CAS  Google Scholar 

  2. Rocheteau, P., Gayraud-Morel, B., Siegl-Cachedenier, I., Blasco, M.A. & Tajbakhsh, S. Cell 148, 112–125 (2012).

    Article  CAS  Google Scholar 

  3. Price, F.D. et al. Nat. Med. 20, 1174–1181 (2014).

    Article  CAS  Google Scholar 

  4. Fukada, S. et al. Stem Cells 25, 2448–2459 (2007).

    Article  CAS  Google Scholar 

  5. Hashimshony, T., Wagner, F., Sher, N. & Yanai, I. Cell Rep. 2, 666–673 (2012).

    Article  CAS  Google Scholar 

  6. Warren, G.L. et al. J. Physiol. 582, 825–841 (2007).

    Article  CAS  Google Scholar 

  7. Muraro, M.J. et al. Cell Syst. 3, 385–394.e3 (2016).

    Article  CAS  Google Scholar 

  8. Wollny, D. et al. Dev. Cell 39, 289–301 (2016).

    Article  CAS  Google Scholar 

Download references

Acknowledgements

We thank M. Muraro and A. Lyubimova for experimental advice. We also thank the Hubrecht FACS facility, the Hubrecht Single Cell facility, the Hubrecht Imaging facility and the Utrecht Sequencing Facility (USF). This work was supported by a Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO) TOP award (NWO-CW 714.016.001) to A.v.O. and by an European Research Council grant (ERC 220-H75001EU/HSCOrigin-309361) and the UMC Utrecht “Regenerative Medicine & Stem Cells” priority research program to C.R. Pax7nGFP mice were kindly provided by S. Tajbakhsh (Institut Pasteur, Paris, France).

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Alexander van Oudenaarden.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Integrated supplementary information

Supplementary Figure 1 CEL-Seq unravels two subpopulations of SCs.

a. Schematic explanation of SC isolation protocol; Pax7 is a marker for SCs. b. SCs were sorted from a single cell suspension of 6 TA muscles isolated from 3 mice via FACS. Left panel: live cells were selected based on Hoechst staining and Forward SCatter Area (FSC-A). Right panel: SCs were selected based on GFP; percentage of GFP+ cells (from viable population) varied between 0.1 and 1.7 %. Gates are shown as black boxes. c. Histogram of number of unique transcripts per cell. Median number of transcripts detected was 556 (red line). Five cells with more than 10,000 unique transcripts detected were omitted for viewing purposes. In total, CEL-Seq was applied to 665 cells, of which the 235 cells with ≥ 700 transcripts were used for RaceID-analysis. d. Histogram of the number of genes detected per cell. Median number of genes detected per cell was 381 (red line). Five cells with more than 3,000 genes detected were omitted from the plot. e. Histogram of oversequencing per molecule. Median level of oversequencing is 4 (red line). f. t-SNE map representation of transcriptome similarities between individual cells; every dot is a cell. Red and blue numbers refer to clustering; see also Figure 1a. g. t-SNE map displaying repartition over clusters between experimental replicates. This plot shows that both clusters were identified in all three independent experiments, excluding the possibility that clustering could be due to physiological differences between mice. h. Comparison to existing literature. Shown is the p-value of overlap between our list of cluster 2 upregulated genes and fold change lists generated based on downloaded microarray data (Supplementary Methods). Our list has significant overlap with genes that are upregulated in SCs isolated from old mice compared to young mice3, with genes upregulated in quiescent versus activated SCs4 and with genes upregulated in injured versus uninjured muscles6. Dotted lines indicate p-value of 1 (no overlap).

Supplementary Figure 2 Expression of selection of genes characterizing SCs belonging to cluster 2.

a-i. Expression of selection of genes that are significantly higher expressed in cluster 2 cells than in cluster 1 cells, plotted over t-SNE map. Color-coding shows number of unique transcripts (after downsampling to 700 transcripts). Dim, t-SNE dimension. For clustering of cells in these t-SNE maps, see Supplementary Fig 1f.

Supplementary Figure 3 Single molecule FISH analysis shows that Fos expression is induced during the SC isolation procedure.

a. Additional example of a SC in an intact Pax7nGFP muscle (GFP channel not shown) that was stained for Fos (probe library coupled to Cy5; shown in green) and Pax7 (coupled to Alexa594, shown in magenta) RNA. Fos was not detected in any of the 80 cells that were screened for Fos expression. b. smFISH on cryosection of intact muscle showing that Socs3 (coupled to Alexa Fluor 594; shown in green) expression could also not be detected in SCs in intact muscles. Socs3 was not detected in any of the 100 cells that were screened for Socs3 expression. c-d. Two more examples of SCs in dissociated muscles that express Fos (coupled to Cy5; shown in green). e. smFISH on SC that was fixed after FACS. Fos was detected in 17 out of the 28 sorted SCs imaged. f. Quantification of smFISH data. Number of Fos-expressing cells is significantly higher in dissociated SCs and in sorted SCs than in SCs in intact muscles that were obtained from the same animals (p > 0.0001). Dissociation smFISH-experiment was repeated 2 times; results from 2 replicates were pooled. g. Cryosection in which expression of Fos was detected in both a SC and in two non-SCs, showing that the response to dissociation is not specific to SCs. The SC cell shown here is the same as the one shown in Figure 1b (right panels; after dissociation), however, this supplementary image was acquired using a wider z-range and using lower gain and laser power for the Fos channel. Nuclei were stained with DAPI (blue). SC, satellite cell (identified by expression of GFP (not shown) and Pax7 RNA (magenta)); arrowheads point at background signal; arrow points at Fos transcription site; scale bar, 5 μm. All images shown were taken using confocal microscopy, except for panel b; this image was acquired using epifluorescence microscopy. In all images, GFP (not shown) was used to identify SCs, and all sections were co-stained with a smFISH probe directed against Pax7 mRNA as a positive control for our smFISH staining.

Supplementary Figure 4 Dissociation time experiment confirms that the dissociation procedure influences the transcriptome of SCs.

SCs were sorted from a single cell suspension of 8 TA muscles from 4 mice via FACS. Half of this material was treated with collagenase for 1 hour; the other half of the cells treated with collagenase for 2 hours. a. Left panel: live cells were selected based on Hoechst staining and Forward SCatter Area (FSC-A). Right panel: SCs were selected based on GFP; percentage of GFP+ cells (from viable population) varied between 0.1 and 1.7 %. Gates are shown as black boxes. b. Histogram of number of unique transcripts per cell. Median number of transcripts detected was 4,022 (red line). In total, SORT-Seq was applied to 752 cells, of which only the 495 cells with ≥ 3,000 transcripts were used for RaceID2-analysis. c. Histogram of the number of genes detected per cell. Median number of genes detected per cell was 1,912 (red line). d. Histogram of oversequencing per molecule. Median level of oversequencing is 3.8 (red line). e. Heat map showing transcriptome correlations between individual SCs. RaceID2 identified 5 clusters, as indicated by coloring on the axes. f. t-SNE representation of 272 1-hour and 223 2-hour collagenase treated SCs that were isolated from 8 TA muscles of 4 mice. Numbers represent clustering. g. t-SNE map where cells are colored according to the duration of their enzymatic dissociation. Together, panels f and g reveal that cluster 1 mostly consists of 1-hour collagenase-treated cells (see also Supplementary Fig. 4h). h. Percentages of 1-hour and 2-hour collagenase-treated cells in all five clusters show that cluster 1 mostly consists of 1-hour collagenase-treated cells. Percentages are corrected for total number of cells per collagenase-treatment as described in the Supplementary Methods. Error bars are 95 % confidence intervals generated by bootstrapping. i. Differential gene expression analysis identified 55 genes that are differentially expressed between clusters 1 (34 cells) and 4 (46 cells). For list of all significant genes, see Supplementary Table 3.

Supplementary Figure 5 Expression of selection of SC genes and of genes characterizing SCs in clusters 1 and 4 in SORT-Seq dataset.

a-d. Expression levels of SC markers plotted over t-SNE map. All markers were detected in all clusters, showing that cells that fall in clusters 1 and 4 are indeed SCs. e-h. Selection of genes that are significantly upregulated in cluster 1 cells (versus cells in clusters 2, 3 and 5). These genes are also higher expressed in 1-hour collagenase treated cells than in 2-hour collagenase treated cells (Fig. 1c). i. Expression of Hspa1b plotted over t-SNE map. Hspa1b is higher expressed in cluster 4 cells than in cluster 1 cells (Supplementary Fig. 4i). Hspa1b is also higher expressed in 2-hour collagenase-treated than in 1-hour collagenase treated cells (Fig. 1c). Color-coding shows number of unique transcripts (after downsampling to 3,000) detected in these cells. Dim, t-SNE dimension.

Supplementary Figure 6 In silico purification can be used to identify and remove dissociation-affected cells from single-cell datasets.

a. Histograms showing the distribution of the percentage of transcriptome that maps to dissociation-affected reads in dissociation-affected (red) and non-dissociation affected (green) cells in our SORT-Seq dataset (see Supplementary Methods for details). Filtering out cells of which ≥ than 5.75 % (red dotted line; see Supplementary Fig. 6b for threshold optimization) of the transcriptome consists of dissociation-affected reads allows the removal of all dissociation-affected cells while removing only 6 % of the non-dissociation affected cells. With a threshold of 5.75 %, 21.2 % of the cells in this dataset are classified as being dissociation-affected. b. Threshold optimization. For all possible threshold values, the percentage of cluster 1 and 4-cells that is correctly assigned as being dissociation-affected (red), the percentage of cluster 2, 3 and 5-cells that is correctly assigned as non-dissociation affected (green) and the overall percentage of cells that are assigned correctly (black) is plotted. The overall percentage of correctly assigned cells would be optimal (96.2 %) if the threshold cutoff value would be set to 6.25 %, and with a threshold value of 6.25 %, 95 % of the dissociation-affected cells are correctly characterized as being dissociation-affected. We choose however to instead set the threshold at 5.75 % (vertical grey dotted line), because with this threshold, 100 % of the cluster 1 and 4-cells are correctly classified as dissociation-affected while only 6 % of the other cells are incorrectly classified as dissociation-affected (vertical grey dotted line). c. Cells that are identified as dissociation-affected by our in silico purification method can be marked as dissociation-affected in a t-SNE map and can, if desired, be removed from the dataset. d. Percentage of transcriptome that maps to dissociation-affected genes plotted over t-SNE map. Circles mark cells in which ≥ 5.75 % of the transcriptome maps to dissociation-affected genes. e. In silico purification method is verified by blindly applying it to CEL-Seq SC dataset (from Supplementary Fig. 1-2). Shown is a histogram of the percentage of transcriptome occupied by dissociation-affected genes for all cells in our CEL-Seq dataset. The threshold (red dotted line) is, without taking the RaceID cluster numbers into account, set to 7.5 %. With this threshold, 33.2 % of our CEL-Seq SCs are classified as being dissociation-affected. f. The blindly set threshold of 7.5 % then appeared to remove all dissociation affected (cluster 2) and 11.8 % of the non-dissociation-affected (cluster 1) cells. g. Cells of which ≥ 7.5 % of the transcriptome maps to dissociation-affected reads are encircled in the t-SNE map.

Supplementary Figure 7 Indexed FACS in combination with SORT-Seq can be used to identify dissociation-affected cells after MitoTracker staining.

Indexed FACS and SORT-seq of 119 SCs extracted from 2 TA muscles of one 4.7 months old female Pax7nGFP mouse was used to identify FACS-properties of dissociation-affected cells. a. Gating strategy to select viable cells in first experiment on MitoTracker-stained SCs. b. MitoTracker Red and GFP repartition of viable cells (Supplementary Fig. 7a) showing that some GFP-positive cells stain higher for MitoTracker then others. All GFP positive cells were sorted and their index-information was stored. c. The percentage of transcriptome occupied by dissociation-affected reads for all cells that were sorted during the MitoTracker experiment was determined via SORT-Seq (using our in silico purification method described in Supplementary Note 2). Out of the 119 cells sequenced, 14 cells had more than 5 % dissociation-affected reads and were identified as dissociation-affected cells. d. Histogram of MitoTracker levels of dissociation-affected (red) and non-dissociation affected (grey) cells showing that it is possible to remove most of the dissociation-affected cells by selecting against MitoTracker-high cells during FACS. e. Histogram of FSC-H levels of dissociation-affected (red) and non-dissociation affected (grey) cells shows that it is possible to remove dissociation-affected cells by discarding cells with high FSC-H values during FACS. Dotted lines in panels d and e refer to borders of the NOT-gate suggested in Supplementary Figure 7f. f. Scatterplot showing MitoTracker and FSC-H levels of the 119 sequenced cells. MitoTracker and FSC-H levels of all cells were retrieved from the FACS index file. Red cells represent cells that were shown to be dissociation-affected by SORT-Seq (Supplementary Fig. 7c). Shaded area represents NOT-gate that was designed on this dataset and that could be used to filter out 12 out of the 14 dissociation-affected cells during FACS.

Supplementary Figure 8 Gate designed by indexed FACS in combination with SORT-Seq on SCs stained for mitochondrial activity can be used to experimentally remove dissociation-affected cells during FACS.

The NOT-gate that was designed based on a pilot experiment (Supplementary Fig. 7) was tested on a larger number of cells (6 muscles from 3 female 6-months old Pax7nGFP mice). a. MitoTracker and GFP repartition of all viable cells in second MitoTracker experiment showing that some SCs have a higher metabolic activity then others. b. MitoTracker and FSC-H repartition of Pax7nGFP SCs during second MitoTracker experiment. Drawn is the NOT-gate that was drawn during FACS and that was designed based on a pilot experiment in which dissociation-affected cells were found to stain higher for MitoTracker and to have higher FSC-H values (Supplementary Fig. 7). All Pax7nGFP+ cells, also the cells that fell in the NOT-gate, were sequenced, however, during FACS we recorded for all cells whether they fell in or out of the NOT-gate. c. Dissociation-affected cells were identified by SORT-Seq (using our in silico purification method described in Supplementary Note 2). Out of the 284 successfully sequenced cells, 80 cells were identified as dissociation-affected cells because ≥ 5% of their transcriptome mapped to dissociation-affected genes. d. The gating strategy successfully reduces the detected expression levels of HSP and IEG genes in bulk. Shown are the average expression levels of several HSP and IEG genes in the non-gated (all Pax7nGFP cells; shown in magenta) and the gated (Pax7nGFP cells out of the NOT-gate; shown in green) population. Some of these genes (Fos, Jun and Hspa1b) are also shown in Figure 1e.

Supplementary Figure 9 Similar Fos/Jun expressing subpopulations can also be found in other single-cell datasets.

a. A Fos/Jun population described in a published mouse acinar single-cell study has an expression profile that is similar to that of dissociation-affected satellite cells. The authors of this study performed a dissociation procedure to extract acinar cells from the mouse pancreas and found two subpopulations of acinar cells, of which one expresses high levels of Fos and Jun (Fig. 6e in Wollny et al., 20168). We analyzed their data to generate a list of all genes that are differentially expressed between these two subpopulations (Supplementary methods; cells #31 and #40, the proliferative acinar cells, were excluded from the analysis). The resulting differential gene expression analysis plot is shown in this panel and reveals that the expression profile of the Fos-expressing acinar cell population is strikingly similar to that of the dissociation-affected subpopulation of satellite cells (compare this panel with Fig. 1a). The overlap with our satellite cell data suggests that this subpopulation of acinar cells might also be induced by the dissociation protocol. The authors do however not validate the in vivo existence of this Fos/Jun expressing subpopulation of acinar cells. For a full list of differentially expressed genes and overlap with our satellite cell dataset, see Supplementary Table 6. b. We detected a similar subpopulation of Fos/Jun expressing cells in a single-cell zebrafish fin dataset that was generated in our lab, suggesting that the dissociation protocol might also induce a stress response in other animals. To generate this dataset, we dissociated the fin of two fish to extract all fin cells (Supplementary methods), after which we performed SORT-Seq. We then analyzed the zebrafish fin data with RaceID2 and detected 9 clusters in the 1683 cells that survived our filtering criteria, as shown in this heath map. c. t-SNE representation of the clusters in the zebrafish fin dataset with cell-type annotation. d. All clusters were detected in all animals that were used for this experiment. e. A subpopulation of osteoblast cells (bone cells, which are harder to liberate from the tissue) expresses high levels of fosab, the zebrafish orthologue of Fos. f. Differential gene expression analysis comparing the expression profile of fosab-positive (cluster 8; 124 cells) and fosab-negative (cluster 2; 305 cells) osteoblasts shows that the expression profile of the fosab-positive cells is very similar to that of the dissociation-affected subpopulation of satellite cells (compare this panel with Fig. 1a). For a full list of differentially expressed genes and overlap with our satellite cell dataset, see Supplementary Table 7. Dim, t-SNE dimension.

Supplementary information

Supplementary Text and Figures

Supplementary Figures 1–9, Supplementary Notes 1–4, Supplementary Tables 1–7 and Supplementary Methods

Life Sciences Reporting Summary

Source data

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

van den Brink, S., Sage, F., Vértesy, Á. et al. Single-cell sequencing reveals dissociation-induced gene expression in tissue subpopulations. Nat Methods 14, 935–936 (2017). https://doi.org/10.1038/nmeth.4437

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/nmeth.4437

This article is cited by

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing