CoChat
Sign Up
Cited by 477Single-cell and spatial transcriptomicsNeuroinflammation and Neurodegeneration MechanismsMolecular Biology Techniques and Applications

Dissection of artifactual and confounding glial signatures by single-cell sequencing of mouse and human brain

Open access
Marsh, Samuel E., Walker, Alec J., Kamath, Tushar, Dissing-Olesen, Lasse, Hammond, Timothy R., de Soysa, T. Yvanka, Young, Adam M. H., Murphy, Sarah +15 more·2022·Nature Neuroscience
DOI 10.1038/s41593-022-01022-8PMC PMC11645269

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

A key aspect of nearly all single-cell sequencing experiments is dissociation of intact tissues into single-cell suspensions. While many protocols have been optimized for optimal cell yield, they have often overlooked the effects that dissociation can have on ex vivo gene expression. Here, we demonstrate that use of enzymatic dissociation on brain tissue induces an aberrant ex vivo gene expression signature, most prominently in microglia, which is prevalent in published literature and can substantially confound downstream analyses. To address this issue, we present a rigorously validated protocol that preserves both in vivo transcriptional profiles and cell-type diversity and yield across tissue types and species. We also identify a similar signature in postmortem human brain single-nucleus RNA-sequencing datasets, and show that this signature is induced in freshly isolated human tissue by exposure to elevated temperatures ex vivo. Together, our results provide a methodological solution for preventing artifactual gene expression changes during fresh tissue digestion and a reference for future deeper analysis of the potential confounding states present in postmortem human samples.

AI snapshot

Written from the paper’s full text.

TL;DR

Enzymatic dissociation of mouse brain induces an aberrant ex vivo transcriptional signature, most prominently in microglia, that is widespread in published single-cell datasets and can confound downstream analyses. The authors provide an inhibitor-based protocol that prevents this signature and show a similar signature in human postmortem microglia/astrocytes that can be induced by delayed freezing.

Key findings

  • In sorted mouse microglia, enzymatic digestion without inhibitors produced an ex vivo “activated” microglia (exAM) cluster enriched for immediate early, stress and immune genes; ENZ-NONE samples had a significant decrease in homeostatic cluster proportion and increase in exAM proportion (*FDR ≤ 0.004; ENZ-NONE versus all other groups; 19,563 cells from n = 12 mice, n = 3 per condition).
  • Adding transcriptional/translational inhibitors eliminated the artifact: DNC-NONE versus DNC-INHIB showed no significant DEGs, and in all-CNS scRNA-seq (10,166 cells, n = 2 per group) only 18 genes were significantly upregulated and none downregulated without inhibitors.
  • Reanalysis of published datasets found the exAM signature across technologies, enzymes and brain regions; overlap was 19 of 25 genes with Van Hove et al., 16 genes with mouse muscle and 11 genes with kidney dissociation signatures.
  • Human postmortem snRNA-seq (47,505 nuclei, three donors) showed enrichment in microglia and some astrocytes; LIGER factors overlapped 79% (30 of 38) for microglia and 44% (12 of 27) for astrocytes, and a 6-h freezing delay induced the microglial factor (P < 3.2 × 10−215; 31 of 38 genes) but weaker astrocyte changes (P < 0.05; 4 of 27 genes).

Methods

  • Mouse scRNA-seq compared cold mechanical dissociation with enzymatic digestion ± an inhibitor cocktail, using FACS-sorted CD45+/CD11b+ myeloid cells and unsorted all-CNS preparations; analyses used Seurat, DESeq2 on pseudobulk “Metacells,” MAST and speckle.
  • Human snRNA-seq used postmortem brain and acutely resected peri-tumor tissue, with nuclei isolation/sorting and 10X Genomics 3′ v.3; analyses used LIGER/iNMF, gene module scoring and comparison of snap-frozen versus 6-h delayed-freezing samples.
  • The study also reanalyzed published mouse and human datasets and performed RNAscope smFISH for Ccl4/Fcrls validation.

Limitations

  • Stated: the human postmortem signature was highly variable across donors and did not correlate with PMI or most available metadata; an age correlation (Spearman r = 0.3208, *P = 0.0001, n = 48) was not robust in donors aged 75+ (r = 0.1879, P = 0.2653, n = 37), and more middle-aged samples are needed.
  • Stated: current human snRNA-seq datasets are limited by microglia proportion and lower sensitivity of earlier chemistries/nuclei isolation, and the authors caution against overinterpreting biological significance of similar signatures.
  • Evident from the text: mouse scRNA-seq used male mice at P89–90 with small group sizes (n = 2–3 per group), human PBMC mock digestion used one healthy 35-year-old male, and acute human tissue was peri-tumor rather than normal brain; the authors also note that one proliferative-cluster abundance result should be interpreted cautiously because it represented only 0.1% of cells (27 of 19,563).

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41593-022-01022-8

Cites

113 of 113 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

44 of 74 cited sources checked · 23 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    A cold mechanical dissociation protocol isolates microglia for scRNA-seq with minimal ex vivo transcriptional alterations.

    “Previously, we optimized a cold mechanical dissociation protocol to isolate microglia for scRNA-seq with minimal ex vivo transcriptional alterations15.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [15] Hammond, T. R. et al. Single-cell RNA sequencing of microglia throughout the mouse lifespan and in the injured brain reveals complex cell-state changes. Immunity 50, 253–271.e6 (2019). source
  • Premisecitation

    Microglia are highly sensitive to perturbations in their environment.

    “As the tissue-resident macrophages of the brain, microglia are highly sensitive to perturbations in their environment13,14.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [13] Gosselin, D. et al. An environment-dependent transcriptional network specifies human microglia identity. Science 356, (2017). source
    • cites [14] Bohlen, C. J. et al. Diverse requirements for microglial survival, specification, and function revealed by defined-medium cultures. Neuron 94, 759–773.e8 (2017). source
  • Findingcitation

    Adult microglia are largely homogenous.

    “In line with our previous work15, we find that adult microglia are largely homogenous.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [15] Hammond, T. R. et al. Single-cell RNA sequencing of microglia throughout the mouse lifespan and in the injured brain reveals complex cell-state changes. Immunity 50, 253–271.e6 (2019). source
  • Findingown data

    Enzymatic digestion causes substantial loss of some cell-surface receptors.

    “Of note, we observed substantial loss of some cell-surface receptors following enzymatic digestion, in line with previous work in peripheral immunology (Supplementary Fig. 1d–f and Supplementary Note 1)3,5–9.”
    Quote in the paperSubstantial claim4 of 6 sources checked
    • cites [3] Abuzakouk, M., Feighery, C. & O’Farrelly, C. Collagenase and Dispase enzymes disrupt lymphocyte surface molecules. J. Immunol. Methods 194, 211–216 (1996). source
    • cites [5] Grange, C. et al. Phenotypic characterization and functional analysis of human tumor immune infiltration after mechanical and enzymatic disaggregation. J. Immunol. Methods 372, 119–126 (2011). source
    • cites [6] Autengruber, A., Gereke, M., Hansen, G., Hennig, C. & Bruder, D. Impact of enzymatic tissue disintegration on the level of surface molecule expression and immune cell function. Eur. J. Microbiol. Immunol. (Bp) 2, 112–120 (2012). source
    • cites [7] Botting, R. A. et al. Phenotypic and functional consequences of different isolation protocols on skin mononuclear phagocytes. J. Leukoc. Biol. 101, 1393–1403 (2017). source
    • cites [8] Schreurs, R. R. C. E. et al. Quantitative comparison of human intestinal mononuclear leukocyte isolation techniques for flow cytometric analyses. J. Immunol. Methods 445, 45–52 (2017). source
    • cites [9] Reichard, A. & Asosingh, K. Best practices for preparing a single cell suspension from solid tissues for flow cytometry. Cytometry A 95, 219–226 (2019). source

      The Tissue Disaggregation > Enzymes section says dispase can cleave specific cell-surface molecules or antigens relevant to T-cell analysis. The Enzymatic and Mechanical Dissociation section also warns that strong or concentrated enzymes may compromise surface markers. These passages support loss of some markers, but do not establish that the loss is substantial or occurs with enzymatic digestion generally. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
  • Findingown data

    There is high concordance between genes upregulated in 'digested' PBMCs and those upregulated in tissue-resident cells following tissue digestion in the Pasciuto et al. and Crinier et al. studies.

    “We also found a high concordance between genes that were upregulated in ‘digested’ PBMCs and those that were observed as upregulated in tissue-resident cells following tissue digestion in the Pasciuto et al.42 and Crinier et al.46 studies (Supplementary Fig. 13f).”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites Pasciuto et al.42 Pasciuto, E. et al. Microglia require CD4 T cells to complete the fetal-to-adult transition. Cell 182, 625–640.e24 (2020). source

      The STAR★Methods (Method Details; page unavailable) describe digestion of brain tissue to prepare single-cell suspensions, and the Results report differentially expressed genes in isolated microglia. The copy does not show that those genes were upregulated because of digestion, or compare digested with undigested tissue. The claimed high concordance is the citing authors’ own finding, not a result attributed to Pasciuto et al. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
    • cites Crinier et al.46 Crinier, A. et al. High-dimensional single-cell analysis identifies organ-specific signatures and conserved NK cell subsets in humans and mice. Immunity 49, 971–986.e5 (2018). source

      The Results report genes expressed more strongly in splenic than blood NK cells, supporting the use of Crinier et al. as a source for a tissue-cell gene signature. But the STAR Methods say the spleen specimens were mechanically disrupted, not digested; the paper therefore does not support describing those genes as upregulated following tissue digestion. The high-concordance comparison is the citing authors’ own finding, not one attributed to Crinier et al. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
  • Classified as probably not a substantial claim. 27
  • May combine two claims. 26
  • The quote only partly states this claim. 1

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Emerging results from other tissues and cell types have begun to characterize dissociation artifacts following enzymatic digestion.

    “However, emerging results from other tissues/cell types have begun to characterize dissociation artifacts following enzymatic digestion43–45,65,66.”
    Quote in the paperNot a substantial claim2 of 5 sources checked
    • cites [43] Van Hove, H. et al. A single-cell atlas of mouse brain macrophages reveals unique transcriptional identities shaped by ontogeny and tissue environment. Nat. Neurosci. 22, 1021–1035 (2019). source
    • cites [44] van den Brink, S. C. et al. Single-cell sequencing reveals dissociation-induced gene expression in tissue subpopulations. Nat. Methods 14, 935–936 (2017). source
    • cites [45] Denisenko, E. et al. Systematic assessment of tissue dissociation and storage biases in single-cell and single-nucleus RNA-seq workflows. Genome Biol. 21, 130 (2020). source
    • cites [65] Adam, M., Potter, A. S. & Potter, S. S. Psychrophilic proteases dramatically reduce single-cell RNA-seq artifacts: a molecular atlas of kidney development. Development 144, 3625–3632 (2017). source
    • cites [66] Machado, L. et al. Tissue damage induces a conserved stress response that initiates quiescent muscle stem cell activation. Cell Stem Cell 28, 1125–1135.e7 (2021). source

  • May combine two claims.

    Findingcitation

    Three genes overlapped between the microglial and astrocyte factors that were not part of consensus mouse gene lists (UBC, DDIT4 and HSPB1), and these are part of cellular stress and damage response machinery.

    “Finally, there were three genes that overlapped between the microglial and astrocyte factors that were not a part of consensus mouse gene lists (UBC, DDIT4 and HSPB1), but which unsurprisingly are also part of cellular stress and damage response machinery52–55.”
    Quote in the paperNot a substantial claim3 of 4 sources checked
    • cites [52] Concannon, C. G., Gorman, A. M. & Samali, A. On the role of Hsp27 in regulating apoptosis. Apoptosis 8, 61–70 (2003). source
    • cites [53] Wang, Z. et al. Dexamethasone-induced gene 2 (dig2) is a novel pro-survival stress gene induced rapidly by diverse apoptotic signals. J. Biol. Chem. 278, 27053–27058 (2003). source

      The Discussion identifies dig2 as a stress-response gene induced by diverse cellular stressors, supporting the stress-response characterization of the gene studied here. Our copy does not identify dig2 as DDIT4, so it does not establish that this paper supports the claim specifically about DDIT4. The microglial–astrocyte overlap is the citing authors’ finding, not a finding attributed to this source. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [54] Ryu, K. Y. et al. The mouse polyubiquitin gene UbC is essential for fetal liver development, cell-cycle progression and stress tolerance. EMBO J. 26, 2693–2706 (2007). source
    • cites [55] Bianchi, M., Crinelli, R., Arbore, V. & Magnani, M. Induction of ubiquitin C (UBC) gene transcription is mediated by HSF1: role of proteotoxic and oxidative stress. FEBS Open Bio 8, 1471–1485 (2018). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisereasoning

    Many widely used protocols for microglial isolation continue to include enzymatic digestion, likely because some experimental conditions require greater yield.

    “Some experimental conditions may dictate the need for greater yield, especially for older or less-sensitive approaches (including many older bulk RNA-sequencing (RNA-seq) studies or protein-based analyses), which is likely why many widely used and widely cited protocols for microglial isolation continue to include enzymatic digestion4,63,64.”
    Quote in the paperNot a substantial claim2 of 3 sources checked
    • cites [4] Cardona, A. E., Huang, D., Sasse, M. E. & Ransohoff, R. M. Isolation of murine microglial cells for RNA analysis or flow cytometry. Nat. Protoc. 1, 1947–1951 (2006). source
    • cites [63] Grabert, K. & McColl, B. W. Isolation and phenotyping of adult mouse microglial cells. Methods Mol. Biol. 1784, 77–86 (2018). source

      The Introduction describes a microglial-isolation protocol that uses enzymatic dissociation, and Notes (Note 3) reports that digestion improved microglial yield. Our copy does not establish that many widely used protocols include digestion, or that a need for greater yield explains their use of it, particularly for older RNA-seq or protein-based analyses. (LLM, conf=0.83)

      Experimental — accuracy not yet measured
    • cites [64] Bordt, E. A. et al. Isolation of microglia from mouse or human tissue. STAR Protoc. 1, (2020). source

      The Limitations section (‘Warm versus Cold Tissue Dissociation’; page unavailable) reports that warm enzymatic digestion yielded 2–3 times as many cells as cold mechanical dissociation. The Step-by-Step Method Details also include an enzyme digestion mix. These passages support the yield rationale and provide an example of such a protocol, but they do not establish that many widely used protocols include enzymatic digestion or that greater yield is why they do. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    A recent study proposed a modified mechanical method to isolate all CNS cell types and appeared to increase neuronal yields, but lacked biological replicates for proper analysis of cell-type abundance and reproducibility.

    “One recent study proposed a modified mechanical method to isolate all CNS cell types and appeared to succeed in significantly increasing neuronal cell yields during dissociation. However, the study lacked biological replicates needed to perform proper analysis of cell-type abundance and reproducibility34.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [34] Mattei, D. et al. Enzymatic dissociation induces transcriptional and proteotype bias in brain cell populations. Int. J. Mol. Sci. 21 (2020). source

      The Brain Dissociation and Cell Isolation section describes the mechanical protocol and reports a higher proportion of adult mouse hippocampal neurons than with enzymatic digestion. Our copy does not establish that the cell-type-abundance comparison lacked biological replicates or could not be assessed for reproducibility. It reports four biological replicates per group for a separate proteotype analysis (Figure 3 legend), so that result does not settle the replication claim about cell-type abundance. (LLM, conf=0.85)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    Many postmortem variables are often not provided and could have significant impacts, such as RNA integrity number (RIN), which has been shown to confound analyses of dementia-related DEGs.

    “While most studies report very common postmortem factors (for example, age at death, PMI), many variables are often not provided and could have significant impacts, such as RNA integrity number (RIN), which has been shown to confound analyses of dementia-related DEGs77.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [77] Miller, J. A. et al. Neuropathological and transcriptomic characteristics of the aged brain. eLife 6, e31126 (2017). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The MAST framework uses a two-part generalized linear model to test for DE while controlling for specified covariates.

    “The MAST framework uses a two-part generalized linear model to test for DE while controlling for specified covariates.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [90] Finak, G. et al. MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. Genome Biol. 16, 278 (2015). source

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    Pseudobulk analysis methods are among the best-performing methods for differential state analysis and for making sample-level comparisons from single-cell data, and help to overcome some of the inherent sparsity in scRNA-seq data.

    “A recent comparative analysis found that pseudobulk analysis methods are among the best-performing methods for differential state analysis and for making sample-level comparisons from single-cell data, and help to overcome some of the inherent sparsity in scRNA-seq data19,89.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [19] Crowell, H. L. et al. muscat detects subpopulation-specific state transitions from multi-sample multi-condition single-cell transcriptomics data. Nat. Commun. 11, 6077 (2020). source

      The Discussion (page not provided) says aggregation methods were fast and performed among the best in the paper’s differential-state comparison. The Methods describe pseudobulk data assembled by subpopulation and sample for sample-level inference. Although the Results describe the starting single-cell expression matrix as sparse, our copy does not state that pseudobulk analysis helps overcome that sparsity. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites [89] Lun, A. T., Bach, K. & Marioni, J. C. Pooling across cells to normalize single-cell RNA sequencing data with many zero counts. Genome Biol. 17, 75 (2016). source

      The Background says pooling counts across cells reduces problematic zeroes, supporting the sparsity point. But the paper evaluates pooled-count normalization and deconvolution, not pseudobulk methods for differential state analysis or sample-level comparisons; it does not establish that those methods are among the best-performing. (LLM, conf=0.86)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductcitation

    No explicit statistical methods were used to predetermine sample sizes, but the study's sample sizes exceed those reported in previous publications.

    “No explicit statistical methods were used to predetermine sample sizes but our sample sizes exceed those reported in previous publications32,34,43.”
    Quote in the paperNot a substantial claim1 of 3 sources checked
    • cites [32] Wu, Y. E., Pan, L., Zuo, Y., Li, X. & Hong, W. Detecting activated cell populations using single-cell RNA-seq. Neuron 96, 313–329.e6 (2017). source
    • cites [34] Mattei, D. et al. Enzymatic dissociation induces transcriptional and proteotype bias in brain cell populations. Int. J. Mol. Sci. 21 (2020). source

      The cited paper’s Introduction discusses previously published datasets, and its methods describe cell numbers and statistical analyses, but neither section compares its sample sizes with those of previous publications. The statement about not predetermining sample sizes refers to the citing authors’ own study, not to this source. Searches for sample-size comparisons and power analysis found no passage establishing the attributed comparison. (LLM, conf=0.72)

      Experimental — accuracy not yet measured
    • cites [43] Van Hove, H. et al. A single-cell atlas of mouse brain macrophages reveals unique transcriptional identities shaped by ontogeny and tissue environment. Nat. Neurosci. 22, 1021–1035 (2019). source

  • Classified as probably not a substantial claim.

    Premisecitation

    snRNA-seq has unlocked the ability to unbiasedly assess transcriptional states of the human brain at single-cell resolution.

    “The advent of snRNA-seq has unlocked the ability to unbiasedly assess transcriptional states of the human brain at single-cell resolution10,11.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [10] Lake, B. B. et al. Neuronal subtypes and diversity revealed by single-nucleus RNA sequencing of the human brain. Science 352, 1586–1590 (2016). source
    • cites [11] Habib, N. et al. Massively parallel single-nucleus RNA-seq with DroNc-seq. Nat. Methods 14, 955–958 (2017). source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductcitation

    Due to issues with statistical analysis between clusters using the same data used to generate clustering, no statistical cutoff was employed and these lists were not used in downstream module scoring or analysis.

    “Due to the issues mentioned above with statistical analysis performed between clusters using the same data used to generate clustering24,25, we did not employ any statistical cutoff or use these lists in any downstream module scoring or analysis.”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites [24] Gao, L. L., Bien, J. & Witten, D. Selective Inference for Hierarchical Clustering. Preprint at https://arxiv.org/abs/2012.02936 (2020). source
    • cites [25] Lähnemann, D. et al. Eleven grand challenges in single-cell data science. Genome Biol. 21, 31 (2020). source

  • May combine two claims.

    Findingown data

    The optimized protocol preserves both in vivo transcriptional profiles and cell-type diversity and yield across tissue types and species.

    “To address this issue, we present a rigorously validated protocol that preserves both in vivo transcriptional profiles and cell-type diversity and yield across tissue types and species.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The consensus exAM gene list overlaps with mouse muscle and kidney dissociation studies by 16 and 11 genes, respectively.

    “Comparison of our consensus exAM gene list with those studies yielded overlap of 16 and 11 genes, respectively (Supplementary Table 21b,c).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The Metacell DEGs exhibited significant overlap with the previously identified DEG list from the microglia-only analysis, confirming that enzymatic digestion, not FACS, was the key factor in the induction of this aberrant response.

    “The Metacell DEGs exhibited significant overlap with the previously identified DEG list from the microglia-only analysis, confirming that enzymatic digestion, not FACS, was the key factor in the induction of this aberrant response.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Human postmortem microglia exhibited significant enrichment of the mouse dissociation signature, as did a small number of astrocytes.

    “Similar to mice, human postmortem microglia exhibited significant enrichment of this signature, as did a small number of astrocytes (Fig. 3d,e).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    ENZ-NONE samples exhibited a significant decrease in the proportion of cells in the homeostatic cluster and a concomitant increase in the proportion of cells in the exAM cluster.

    “We found that ENZ-NONE samples exhibited a significant decrease in the proportion of cells in the homeostatic cluster and a concomitant increase in proportion of cells in the exAM cluster (*false discovery rate (FDR) ≤ 0.004; ENZ-NONE versus all other groups; Methods) (Fig. 1e and Supplementary Table 3).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    There are no significant DEGs between DNC-NONE and DNC-INHIB groups, indicating that inhibitors have negligible adverse effect on gene expression.

    “Analysis of Dounce without inhibitors (DNC-NONE) versus Dounce with inhibitors (DNC-INHIB) groups found no significant DEGs between the groups (Supplementary Tables 9 and 12), indicating that the use of inhibitors in current scRNA-seq pipeline(s) appears to have negligible adverse effect on gene expression.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Microglia and some astrocytes exhibited the greatest enrichment of human orthologs of the mouse signature genes, and the percentage of nuclei above threshold per sample was highly variable and did not correlate with PMI.

    “This analysis demonstrated that microglia and some astrocytes exhibited the greatest enrichment of human orthologs of the mouse signature genes and that the percentage of nuclei above threshold per sample was both highly variable and did not correlate with PMI (Fig. 3f–k).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The exAM cluster is characterized by expression of immediate early genes, stress-induced genes, and immune-signaling genes.

    “characterized by expression of several immediate early genes (IEGs) (for example, Fos, Jun), stress-induced (Hspa1a, Dusp1) and immune-signaling genes (Ccl3, Ccl4)”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    There was a significant enrichment of the iNMF microglial factor at the 6-h timepoint compared with the 0-h timepoint (P < 3.2 × 10−215), corresponding to 31 of 38 genes from the factor being significantly different.

    “We found a significant enrichment of the iNMF microglial factor at the 6-h timepoint compared with the 0-h timepoint (P < 3.2 × 10−215) (Fig. 4d–f), corresponding to 31 of 38 genes from the factor being significantly different (Fig. 4g and Supplementary Table 23).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingcitation

    Induction of the ex vivo signature occurred only in microglia and other brain-associated myeloid cells, while other cell types may show initial signs of this response (not significant).

    “In line with previous work on the response of CNS cells to stress35, we found that induction of this ex vivo signature occurred only in microglia and other brain-associated myeloid cells, but, while not significant, we found that other cell types may show initial signs of this response, which could become significant under different conditions.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [35] Satoh, J. & Kim, S. U. HSP72 induction by heat stress in human neurons and glial cells in culture. Brain Res. 653, 243–250 (1994). source

  • May combine two claims.

    Findingreasoning

    The mechanical dissociation protocol purifies microglia but depletes the majority of other CNS cell types due to tight pestle and Percoll centrifugation.

    “The other major drawback to our mechanical dissociation protocol is that while it purifies microglia, the combination of the tight pestle during the Dounce process and Percoll centrifugation leads to depletion of the majority of other CNS cell types.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingcitation

    Induction of the aberrant ex vivo response is extremely widespread, not specific to enzyme, sequencing technology, or isolation/tissue processing protocol, and can be induced by prolonged room-temperature steps during mechanical dissociation.

    “We also confirm, through reanalysis of the CNS literature, that induction of this aberrant ex vivo response is extremely widespread, and not specific to the enzyme, sequencing technology or specifics of the isolation/tissue processing protocol, and some of the signature appears to be inducible even just by the presence of prolonged room-temperature steps during otherwise favorable mechanical dis”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Signatures in postmortem microglia and astrocytes correlated with age across the entire dataset, but the correlation was not robust across post hoc subgroup comparisons.

    “While signatures in postmortem microglia and astrocytes correlated with age across the entire dataset, the correlation was not robust across post hoc subgroup comparisons.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The human signature was highly variable across donors and was not correlated with PMI or most other available postmortem meta variables.

    “This human signature was highly variable across donors and was not correlated with PMI or most other available postmortem meta variables.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Microglial reaction to heat stress is significantly faster than for other CNS cell types.

    “These results are in agreement with previous in vitro work which found that microglial reaction to heat stress was significantly faster than for other CNS cell types35, although that does not preclude such artifacts being present in other cell types, depending on the details of cell isolation33.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [35] Satoh, J. & Kim, S. U. HSP72 induction by heat stress in human neurons and glial cells in culture. Brain Res. 653, 243–250 (1994). source

  • May combine two claims.

    Interpretationown data

    Induction of the exAM signature can occur differentially across experimental conditions and may impact downstream analyses.

    “Our reanalyses demonstrate that induction of the exAM signature can occur differentially across experimental conditions and may impact downstream analyses.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    Biological discoveries rest on single-cell technologies being able to faithfully measure cell-type-specific expression patterns ex vivo.

    “However, these biological discoveries rest on single-cell technologies being able to faithfully measure cell-type-specific expression patterns ex vivo.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    CX3CR1 was excluded as a parameter for cell sorting due to differences in staining between isolation methods from cleavage of extracellular epitopes by enzymes.

    “CX3CR1 was excluded as a parameter used for cell sorting due to differences in staining between isolation methods as the result of cleavage of extracellular epitopes by enzymes (Supplementary Fig. 1d,e).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Lack of widespread understanding of this signature in the CNS and periphery has resulted in the pervasiveness of this signature in the current literature and can confound interpretations of results and conclusions of such studies.

    “We also demonstrate how lack of widespread understanding of this signature both in the CNS and periphery has resulted in the pervasiveness of this signature in the current literature and can confound interpretations of results and conclusions of such studies.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The majority of studies have utilized enzymatic dissociation to study all CNS populations at single-cell resolution.

  • May combine two claims.

    Interpretationreasoning

    Acute premortem factors such as agonal state, medication, comorbidities, and cause of death are not artificial, as they reflect true biology, but they confound the ability to perform analysis between desired experimental variables/groups when not properly powered/controlled.

    “It is also critical to note that acute premortem factors such as these are not ‘artificial’ as they reflect the true biology associated with agonal state, medication, comorbidities, cause of death and so on, but nonetheless these signatures do confound the ability to perform analysis between desired experimental variables/groups (for example, disease status, cognitive function and so on) when not”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductreasoning

    During sorting of the tail-vein PBS-injected dataset, a clog occurred, and the sample was transferred between room temperature and ice twice for a total time at room temperature of approximately 20 min.

    “Consequently, the sample was transferred between room temperature and ice twice for a total time at room temperature of approximately 20 min (Supplementary Note 2).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    The paper provides a method to eliminate ex vivo transcriptional activation during enzymatic digestion that is easy to implement, functional across different tissue types and species, and should require minimal effort to incorporate into existing protocols.

    “In conclusion, we provide a method by which to eliminate ex vivo transcriptional activation during enzymatic digestion that is easy to implement, is functional across different tissue types and species, and therefore should require minimal effort to incorporate into existing protocols.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Most currently available human snRNA-seq datasets are limited by both the proportion of microglia per sample and the dramatically lower sensitivity of earlier single-cell chemistries/nuclei isolation protocols.

    “Most currently available human snRNA-seq datasets are limited by both the proportion of microglia per sample81 and the dramatically lower sensitivity of earlier single-cell chemistries/nuclei isolation protocols (Supplementary Note 6).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [81] Liddelow, S. A., Marsh, S. E. & Stevens, B. Microglia and astrocytes in disease: dynamic duo or partners in crime? Trends Immunol. 41, 820–835 (2020). source

  • Classified as probably not a substantial claim. The quote only partly states this claim.

    Study design and conductnone given

    All sample processing for human acute, postmortem, and PBMC experiments was performed in random order.

    “All sample processing was performed in random order.”
    • Also stated in: § Postmortem human brain · plain
      “All sample processing was performed in random order.”
    • Also stated in: § Human PBMCs · plain
      “All sample processing was performed in random order.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    A liberal gate of any CD11b+/CD45+ cells was intentionally set so as not to bias scRNA-seq due to changes in cell-surface receptor expression.

    “We intentionally set a liberal gate of any CD11b+/CD45+ cells so as not to bias our scRNA-seq due to changes in cell-surface receptor expression (Supplementary Fig. 1b,c).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    All sample processing for mouse experiments was performed in alternating order by experimental group.

    “All sample processing was performed in alternating order by experimental group.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    smFISH experiments and imaging were performed blind.

    “smFISH experiments and imaging were performed blind.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    Experimental collection and processing of samples were not performed blind due to the conditions of the experiment.

    “Experimental collection and processing of samples were not performed blind due to the conditions of the experiment.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Datasets were quality controlled by checking for and removing doublets on the basis of dual cell-type marker gene expression.

    “Datasets were additionally quality controlled at this step by checking for and removing doublets on the basis of dual cell-type marker gene expression.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    To determine thresholds for enrichment, datasets were downsampled so that all major cell classes had the same number of cells, so that cell number/proportion did not influence outlier detection.

    “We then downsampled each dataset so that all of the major cell classes had the same number of cells, so that cell number/proportion did not influence outlier detection.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    scRNA-seq/snRNA-seq were not performed blind but using automated and appropriate field standard analysis techniques that require metadata to be present.

    “scRNA-seq/snRNA-seq were not performed blind but using automated and appropriate field standard analysis techniques that require metadata to be present.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Some datasets required additional rounds of analysis and subclustering to completely remove likely doublet nuclei that were not found during the first round.

    “Some datasets required additional rounds of analysis and subclustering to completely remove likely doublet nuclei that were not found during the first round.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The only datapoints excluded from this analysis were cells defined as low-quality or doublets.

    “The only datapoints that were excluded from this analysis were those cells defined as low-quality (see above scRNA-seq/snRNA-seq analyses sections and GitHub code repository) or those that were defined as doublets using manual annotation of doublet cells by the experimenter”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    To detect and filter doublets and/or low-quality cells in snRNA-seq datasets, iterative rounds of clustering using LIGER were performed.

    “To detect and filter doublets and/or low-quality cells in snRNA-seq datasets, we performed iterative rounds of clustering using LIGER.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    For datasets in Supplementary Fig. 9 composed of different genotypes or treatments, only cells from WT or control mice were included in downstream analysis.

    “For the datasets in Supplementary Fig. 9 that were composed of different genotypes or treatments, only cells from WT or control mice were included in downstream analysis.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    To enhance confidence in subclustering analysis, only cell classes with greater than 80 cells per replicate (>300 cells total) were considered for subclustering.

    “To enhance confidence in subclustering analysis, only cell classes with greater than 80 cells per replicate (>300 cells total) were considered for subclustering.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    All datasets were further subclustered to manually examine and remove any remaining doublets.

    “All datasets were then further subclustered to manually examine and remove any remaining doublets.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    All filtering was performed using field standard techniques.

    “All filtering was performed using field standard techniques as described in the analysis subsections above.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Doublet identification was performed using a combination of marker gene expression and shared LIGER factors to identify nuclei expressing markers exclusive to two different cell classes.

    “Doublet identification was then performed using a combination of marker gene expression and shared LIGER ‘factors’ to identify nuclei that expressed markers or combinations of markers that are exclusive to two different cell classes.”
    Quote in the paperSubstantial claim
49 of 49

  • The paper relies on this without support.

    No prior study has examined the human brain for cell-type-specific differences in transcriptional states using snRNA-seq.

    “and to our knowledge none have done so in a cell-type-specific manner using snRNA-seq”

  • The paper relies on this without support.

    The majority of studies have utilized enzymatic dissociation to study all CNS populations at single-cell resolution.

    “Consequently, the majority of studies have utilized enzymatic dissociation to study all CNS populations at single-cell resolution.”
2 of 2

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Conventional cell dissociation protocols often compromise neuronal integrity through mechanical shearing or induce glial activation.

    “For example, conventional cell dissociation protocols often compromise the integrity of neurons through mechanical shearing or induce the activation of glia[6, 8–12].”

    From Comparative analysis of nuclei isolation methods for brain single-nucleus RNA sequencing (2026) · cites this paper as [10]

    The Discussion reports that typical enzymatic digestion of mouse brain tissue induces ex vivo transcriptional changes, preferentially in microglia and other brain myeloid cells, supporting the glial-activation part. It also reports depletion of other CNS cell types with a mechanical Dounce-and-Percoll protocol, but does not establish that mechanical shearing compromises neuronal integrity. Page numbers were not provided in our copy. (LLM, conf=0.87)

    Experimental — accuracy not yet measured

  • All three methods typically involve enzymatic digestion, which can potentially activate certain cell types, introducing a common source of potential bias.

    “Notably, all three methods typically involve enzymatic digestion, which can potentially activate certain cell types, introducing a common source of potential bias[10].”

    From Comparative analysis of nuclei isolation methods for brain single-nucleus RNA sequencing (2026) · cites this paper as [10]

    The Discussion reports that typical enzymatic digestion of mouse brain tissue induced transcriptional changes, preferentially in microglia and other brain myeloid cells, and could confound downstream analysis. It says enzymatic dissociation is used in the majority of single-cell studies of CNS populations, but does not establish that each of the citing sentence’s three unspecified methods typically involves it. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

  • Many cell types, such as microglia, are especially sensitive to dissociation-induced activation.

    “Many cell types, such as microglia, are especially sensitive to dissociation-induced activation, which is a known challenge in single-cell studies[9, 10].”

    From Comparative analysis of nuclei isolation methods for brain single-nucleus RNA sequencing (2026) · cites this paper as [10]

    The Results section ('Mouse microglia are especially sensitive to ex vivo alterations'; page not provided) shows that enzymatic dissociation induced an artifactual activation signature in microglia. But it reports that microglia/myeloid cells were the only major CNS cell class with a clear inhibitor-dependent clustering response; other cell types showed only slight changes. Thus the source supports the microglia example and the single-cell-study challenge, not the claim that many cell types are especially sensitive. (LLM, conf=0.86)

    Experimental — accuracy not yet measured

  • 5 μM Actinomycin D was kept during collection and enzymatic dissociation of the tissue to preserve transcriptional state and prevent artificial induction of stress-response genes during tissue dissociation.

    “Note that 5 μM ActD was kept during collection and enzymatic dissociation of the tissue to preserve transcriptional state and prevent artificial induction stress-response genes during tissue dissociation (Marsh et al, 2022; Perocchi et al, 2007).”

    From APOE from astrocytes restores Alzheimer’s Aβ-pathology and DAM-like responses in APOE deficient microglia (2024) · cites this paper as Marsh et al, 2022

    The Methods section, “Inhibitor cocktail to prevent activation” (page not provided), specifies actinomycin D in both dissection and digestion buffers, and the Main section says the inhibitor protocol prevents artifactual transcription during dissociation. The specified concentration is 5 μg ml−1, not 5 μM, and the protocol uses actinomycin D alongside other inhibitors rather than establishing that actinomycin D alone has this effect. (LLM, conf=0.90)

    Experimental — accuracy not yet measured

  • Potential ex vivo activation was inhibited as described in a previous study.

    “Potential ex vivo activation was inhibited as described in a previous study25 (Supplementary Fig. 7c).”

    From Blockade of STING Activation Alleviates Microglial Dysfunction and a Broad Spectrum of Alzheimer's Disease Pathologies (2024) · cites this paper as [25]

Cites

Flagged first, then by citations

Research works printed but not linked: 1

  1. Doberentz, E. & Madea, B. in Regulation of Heat Shock Protein Responses: Heat Shock Proteins (eds Asea, A. A. A. & Kaur, P.) Ch. 18 (Springer International Publishing, 2018).

Cited by

Sorted by citations

Related

Authors
Marsh, S. E., Walker, A. J., Kamath, T., Dissing-Olesen, L., Hammond, T. R., de Soysa, T. Y., Young, A. M. H., Murphy, S., Abdulraouf, A., Nadaf, N., Dufort, C., Walker, A. C., Lucca, L. E., Kozareva, V., Vanderburg, C., Hong, S., Bulstrode, H., Hutchinson, P. J., Gaffney, D. J., Hafler, D. A., Franklin, R. J. M., Macosko, E. Z., & Stevens, B.
ORCID
Venue
Nature Neuroscience
Publisher
Springer Science and Business Media LLC
Published
2022-03
Volume
25
Issue
3
Pages
306-316
ISSN
  • 1097-6256 (ISSN-L)
  • 1546-1726
DOI
10.1038/s41593-022-01022-8
PMC
PMC11645269
OpenAlex
W4220661922
Cited by
432
References
108
Citations by year
2026: 85 · 2025: 73 · 2024: 102 · 2023: 93 · 2022: 66 · 2021: 12 · 2012: 1
FWCI
26.20
Type
journal article
Language
en
Update notices
correction (2024-12-13)
Open access
Licence
Locations
  • Nature Neuroscience (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by-nd)
  • PubMed (publishedVersion)
  • PubMed Central (submittedVersion, cc-by)
  • UCL Discovery (University College London) (submittedVersion)
Funders
  • Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (203151/Z/16/Z)
  • Cure Alzheimer's Fund
  • Rush University (P30AG10161, R01AG36042, R01 AG15819, R01AG30146)
  • Wellcome Trust (203151/Z/16/Z, 203151/Z/16, 203151)
  • Cancer Research UK (24423)
  • National Institute for Health and Care Research (R01AG15819, CL-2016-14-005, P30AG10161)
  • Royal College of Surgeons of England
  • Multiple Sclerosis Society
  • Academy of Medical Sciences (SGL017\1009)
  • Lundbeckfonden (R182-2014-2844, R193-2015-4103)
  • H. Lundbeck A/S
  • University of California, Irvine
  • Medical Research Council (203151/Z/16/Z)
  • National Institute on Aging (U01‐AG032984, R01AG36042, U01AG46161, P30-AG 10161, U01 AG32984, R01 AG036042, RC2AG036547, R01 AG30146, R01‐AG17917, R01 AG15819, U01AG061356, R01-AG36836, U01‐AG61356, U01AG046152, RF1 AG057473, R01-AG030146, R01 AG017917, R01AG48015, R01AG048015, U01AG046161, RF1AG57473, P30AG010161, U01‐AG46152, R01AG036836, R01-AG015819)
  • National Institute of General Medical Sciences (T32GM008313)
  • National Institute of Neurological Disorders and Stroke (T32NS007473)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
  • Life Sciences › Neuroscience › Neurology › Neuroinflammation and Neurodegeneration Mechanisms
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Molecular Biology Techniques and Applications
Keywords
tissue dissociation, enzymatic tissue dissociation, microglia, single-nucleus RNA sequencing, postmortem brain, single-cell RNA sequencing, brain cell types, mouse brain, human brain
MeSH terms
Brain, Humans, Neuroglia, Sequence Analysis, RNA, Microglia, Single-Cell Analysis, Transcriptome

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review