Research ArticleImmunology

Signals trigger state-specific transcriptional programs to support diversity and homeostasis in immune cells

See allHide authors and affiliations

Science Signaling  14 May 2019:
Vol. 12, Issue 581, eaao5820
DOI: 10.1126/scisignal.aao5820
  • Fig. 1 ICA-based separation of signal-induced cell transcriptional states.

    (A) ICA of scRNA-seq gene expression data. Colors indicate identified cell states of LPS-treated [n = 83 cells; LPS, 22 hours phorbol 12-myristate 13-acetate (PMA) + 2 hours LPS + PMA], PAL-treated (n = 74 cells; PAL, 24 hours PAL + PMA), and untreated (UN) (n = 72 cells; UN, 24 hours PMA) THP-1 macrophages. (B) Average expression of treatment-induced differentially expressed genes, used as an ICA input, in identified cell states (n = 1001 genes for LPS treatment and n = 266 for PAL treatment). Statistical analysis was performed by one-way analysis of variance (ANOVA) with Tukey’s honestly significant difference (HSD) test. **P < 0.01; ***P < 0.001. n.s., not significant; FPKM, fragments per kilobase of per million reads. (C) Definition of macrophage states after activation with LPS or PAL.

  • Fig. 2 ICA-based representation of individual genes and pathways in single cells and identified cell states.

    (A) Expression of key pro- and anti-inflammatory genes in LPS- and PAL-induced cell states. Each dot represents the expression value in an individual cell from the corresponding cell state. (B) Expression of selected pro- and anti-inflammatory genes in individual cells in ICA-based two-dimensional space. Each point corresponds to an individual cell. (C) Differentially expressed key regulatory pathways in identified cell states. (D) ICA-based representation of key differentially expressed regulatory pathways. Each dot represents an average expression of the genes from the corresponding pathway in individual cells from the specified cell state. Statistical analysis for (A) and (C) was performed by one-way ANOVA with Tukey’s HSD test for LPS-treated (n = 83 cells), PAL-treated (n = 74 cells), and untreated THP-1 macrophages (n = 72 cells). *P < 0.05; **P < 0.01; ***P < 0.001; n.s., not significant.

  • Fig. 3 Cell morphology can be linked to transcriptional states.

    (A) Representative images from RNA FISH analysis of THP-1 macrophages treated with LPS (100 ng/ml) for IL1B (red) and NR3C1 (green) transcripts, as described in Materials and Methods. Bottom: Merged image of fluorescence channels and differential interference contrast images. (B) Quantification of cell size (arbitrary units) and eccentricity (0 = circle, 1 = ellipse) for cells with high expression of the indicated genes, as described in Materials and Methods. N indicates the number of cells analyzed. Data were acquired from two independent experiments. The red box plots represent data from IL1B-positive cells. White box plots represent data from cells with high expression (top 50%) of the genes indicated at the top of the panel. Statistical analysis was done by one-way ANOVA followed by Dunn’s multiple comparisons test. **P < 0.01; ***P < 0.001. a.u., arbitrary units; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; n.s., not significant. (C) Bright-field microscopy images were acquired at 1-hour intervals at constant positions. THP-1 macrophages were treated with LPS (100 ng/ml) alone or together with 1 μM dexamethasone (Dex). Cells in boxes at the 36-hour time point are shown in magnified view. Experiments were repeated three times. Representative images shown here were taken from one experiment. (D) Morphological properties of untreated THP-1 macrophages that had been differentiated for 72 hours and cells that were treated for 48 hours with LPS alone or in the presence of dexamethasone and differentiated THP1-XBlue–defMyD cells. Experiments were repeated three times. Representative images are shown. (E) Representative bright-field, live cell imaging results for cells with changing morphological properties. Image acquisition was performed every 2 min in Z-stack mode for a time period of 16 hours. Individual cells were tracked manually. Every line shows the same cells recorded at different times. Red color bars illustrate (nonquantitatively) big/flat M1-like cells, whereas blue color bars illustrate small/round M2-like cells. Experiments were repeated three times. Representative image regions are shown.

  • Fig. 4 The key antagonistic transcriptional regulator genes NR3C1/IL1B and ATF3/IL1B and their proteins are expressed in distinct transient cell states of LPS- and PAL-stimulated macrophages.

    (A) scRNA-seq and sc-qPCR analyses of THP-1 macrophages indicate trends for mutually exclusive expression of the LPS-induced genes NR3C1 and IL1B and the PAL-induced genes ATF3 and IL1B. For the scRNA-seq experiments, n = 83 cells for LPS treatment and n = 74 cells for PAL treatment; for the sc-qPCR experiments, n = 88 cells for LPS treatment and n = 168 cells for PAL treatment. (B) Independent immunofluorescence detection of antagonistic proteins GR and IL-1β (for LPS treatment) and ATF3 and IL-1β (for PAL treatment) compared to the highly correlating protein pairs IL-8 and IL-1β (for LPS) and PPARG and IL-1β (for PAL) in primary macrophages. Each image represents an individual specimen for the detection of the indicated pair of proteins. Alexa Fluor 488– or Alexa Fluor 594–conjugated secondary antibodies were used for protein detection. Nuclei were detected with 4′,6-diamidino-2-phenylindole (DAPI) (blue staining, shown nuclei or nuclei borders). Cell borders defined by segmentation are shown in pink. Scale bar, 20 μm. (C) Quantification of immunofluorescence staining as shown in (B). N indicates the number of cells analyzed. R represents the Spearman’s rank correlation coefficient in (A) to (C).

  • Fig. 5 Changes in cell state composition.

    (A) Top: Increasing concentrations of LPS resulted in increases in the percentage of cells (THP-1 macrophages) that showed increased gene expression (right), whereas the increase in the number of transcripts per cell was largely constant (right). Gene expression was assessed using sc-qPCR for 172 cells treated with different concentrations of LPS. The percentage of (responsive) cells was calculated for every gene under investigation. All cells that showed expression of at least one gene of interest constituted the total quantity of cells (positive or responsive cells). For all positive cells, the numbers of transcripts per cell were calculated by absolute quantification relative to a DNA standard of known concentration and estimated molecules per volume. Data were derived from two independent experiments. **P < 0.01; ***P < 0.001; otherwise not significant. Statistical analysis was performed by two-way ANOVA with Bonferroni posttest; mean values (bars) and SDs (error bars) are indicated. Bottom: Density plots show the distribution of gene expression (transcripts per cell) as detected by sc-qPCR. Data were derived from two independent experiments. (B) Model for the inhibitory interaction of transcriptional regulators NR3C1 and ATF3. Multiple studies have shown the (competitive) inhibitory effects of GR and ATF3 on production of inflammatory mediators, such as IL-1β (51, 52). Our analyses of single cells indicated that macrophages expressing genes encoding GR or ATF3 were almost devoid of IL1B expression and of IL-1β protein, whereas cells that did not contain GR or ATF3 showed strong expression of IL1B.

  • Fig. 6 Analysis of SE reveals the organization of intracellular transcriptional responses to the signal molecules LPS and PAL.

    (A) SE network model. High SE is suggestive of high uncertainty, thus promiscuous, inefficient usage of network components, whereas lower SE indicates lower uncertainty and a more deterministic, coordinated usage of functional networks of cells. (B) Normalized global transcriptional SE for macrophage states under the indicated conditions for LPS-treated (n = 83 cells), PAL-treated (n = 74 cells), and untreated THP-1 macrophages (n = 72 cells). (C) SE analysis for the TNF-α and NF-κB signaling pathway (n = 519 genes). Statistical analysis for B and C was performed by one-way ANOVA with Tukey’s HSD test. *P < 0.05; **P < 0.01; ***P < 0.001; n.s., not significant. (D) Summary of the defined macrophage states and associated entropy states after treatment with PAL or LPS.

Supplementary Materials

  • stke.sciencemag.org/cgi/content/full/12/581/eaao5820/DC1

    Fig. S1. Experimental approaches and technical background data.

    Fig. S2. ICA of untreated, LPS-treated, and PAL-treated cells and WGCNA gene coexpression module construction.

    Fig. S3. Characterization of WGCNA-defined gene coexpression modules and corresponding pathways.

    Fig. S4. The SOM-based approach uses whole transcriptome data to determine cell transcriptional states in response to LPS and PAL.

    Fig. S5. SOM-based analysis of whole transcriptome single-cell profiles.

    Fig. S6. Single-cell gene expression analysis of IL10, STAT3, and IL1B.

    Fig. S7. Differential expression analysis of THP-1 cells deficient in MyD88.

    Fig. S8. Key antagonistic transcriptional regulators are expressed in distinct transient cell populations.

    Fig. S9. Cumulative expression proportion counted for each cell in the different treatments and states.

    Fig. S10. Variance estimation of gene expression data derived from single cells: Part I.

    Fig. S11. Variance estimation of gene expression data derived from single cells: Part II.

    Fig. S12. Quality control of single-cell sequencing data.

    Fig. S13. Microfluidic IFC microchamber screening.

    Table S1. WGCNA-based analysis of pathways and genes.

    Table S2. Output list of SOM-based analysis of pathways and genes.

    Table S3. A high number of genes specific for the LPS-induced, proinflammatory state that could be repressed by stable knockdown of MyD88.

    Table S4. List of primers.

  • The PDF file includes:

    • Fig. S1. Experimental approaches and technical background data.
    • Fig. S2. ICA of untreated, LPS-treated, and PAL-treated cells and WGCNA gene coexpression module construction.
    • Fig. S3. Characterization of WGCNA-defined gene coexpression modules and corresponding pathways.
    • Fig. S4. The SOM-based approach uses whole transcriptome data to determine cell transcriptional states in response to LPS and PAL.
    • Fig. S5. SOM-based analysis of whole transcriptome single-cell profiles.
    • Fig. S6. Single-cell gene expression analysis of IL10, STAT3, and IL1B.
    • Fig. S7. Differential expression analysis of THP-1 cells deficient in MyD88.
    • Fig. S8. Key antagonistic transcriptional regulators are expressed in distinct transient cell populations.
    • Fig. S9. Cumulative expression proportion counted for each cell in the different treatments and states.
    • Fig. S10. Variance estimation of gene expression data derived from single cells: Part I.
    • Fig. S11. Variance estimation of gene expression data derived from single cells: Part II.
    • Fig. S12. Quality control of single-cell sequencing data.
    • Fig. S13. Microfluidic IFC microchamber screening.

    [Download PDF]

    Other Supplementary Material for this manuscript includes the following:

    • Table S1 (Microsoft Excel format). WGCNA-based analysis of pathways and genes.
    • Table S2 (Microsoft Excel format). Output list of SOM-based analysis of pathways and genes.
    • Table S3 (Microsoft Excel format). A high number of genes specific for the LPS-induced, proinflammatory state that could be repressed by stable knockdown of MyD88.
    • Table S4 (Microsoft Excel format). List of primers.

Stay Connected to Science Signaling

Navigate This Article