Epigenetically regulated digital signaling defines epithelial innate immunity at the tissue level

Abstract

To prevent damage to the host or its commensal microbiota, epithelial tissues must match the intensity of the immune response to the severity of a biological threat. Toll-like receptors allow epithelial cells to identify microbe associated molecular patterns. However, the mechanisms that mitigate biological noise in single cells to ensure quantitatively appropriate responses remain unclear. Here we address this question using single cell and single molecule approaches in mammary epithelial cells and primary organoids. We find that epithelial tissues respond to bacterial microbe associated molecular patterns by activating a subset of cells in an all-or-nothing (i.e. digital) manner. The maximum fraction of responsive cells is regulated by a bimodal epigenetic switch that licenses the TLR2 promoter for transcription across multiple generations. This mechanism confers a flexible memory of inflammatory events as well as unique spatio-temporal control of epithelial tissue-level immune responses. We propose that epigenetic licensing in individual cells allows for long-term, quantitative fine-tuning of population-level responses.

Fig. 1. MAMP sensing in epithelial monolayers shows digital and analog components.

a Schematic detailing the NF-κB-dependent gene expression reporter system. An innate immune signal activates IKK which phosphorylates IκBα and sends it for degradation, freeing the fluorescently tagged NF-κB subunit (p65-mRuby) to enter the nucleus and transcribe the Venus-PEST reporter. b MCF10A NF-κB reporter cells were incubated in the presence or absence of IL-1β (10 ng/ml) and/or IKK inhibitor VII (10 µM). Representative images from five replicates taken at indicated times are shown. Scale bar, 50 μm. c Representative single-cell trace. The median intensity of the nuclear/cytoplasmic NF-κB is plotted in black, while the median nuclear + cytoplasmic intensity gene expression reporter is in green. d Schematic of innate immune receptors and inputs that signal to NF-κB. TLR5, IL1R, and TLR1/2 all signal through MyD88. e Heatmap of all cells in the MAMP screen. MCF10A monoclonal reporter cell line was imaged for 50 min then treated with the indicated input and concentration. Concentrations at 100x were 1 µg/ml (Flagellin), 200 µg/ml (Poly(I:C)), 1 µg/ml (Pam3CSK4), 1 µg/ml (TNFα), and 100 ng/ml (IL-1β). Heatmaps are ordered top to bottom from highest to lowest maximum Venus expression. f (Top) Swarmplots of the area under the curve after the first NF-κB translocation peak (AUC steady) in 200 cells randomly sampled from the indicated input and concentration. (Bottom) Swarmplots of the 95th percentile of the gene expression in 200 cells randomly sampled from the indicated input and concentration. R squared values of the linear regression model fit to responses across the log of three concentrations. g Representative images of clonal MCF10A after 8 h of incubation with indicated stimulus (1 µg/ml (Flagellin), 100 ng/ml (IL-1β), 20 µg/ml (Poly(I:C)), 100 ng/ml (TNFα), and 1 µg/ml (Pam3CSK4)). H2B-iRFP nuclear marker is shown in red and Venus gene expression reporter is shown in cyan. Scale bar, 200 µm. h MCF10A NF-κB reporter cells stimulated with 1 µg/ml Pam3CSK4. Traces of NF-κB translocation (gray) and Venus expression (green) for five non-responders and responders are shown.

Fig. 2. Pam3CSK4 triggers a capped digital response in primary epithelial cells and organoids.

a p65 immunofluorescent stain of wild type MCF10A stimulated with media (top) or 1 µg/ml Pam3CSK4 (bottom) for 30 min. Scale bar, 300 µm. Histogram shows nuclear/cytoplasmic p65 intensity in logarithmic scale. Cells with nuclear/cytoplasmic p65 ratio greater than media controls were considered responders. b Dose-response curve of WT MCF10A with 30 minutes of Pam3CSK4 (0.01, 0.1, 1, 10 µg/ml), MALP (0.001, 0.01, 0.1, 1 µg/ml), or PGN (0.01, 0.1, 1, 10 µg/ml). Data represent the mean ± SD of three replicates. c Human prostate epithelial cells (RWPE-1) were cultured, treated for 30 mins with 10 µg/ml Pam3CSK4, fixed and stained as described in methods. Representative images are shown. Yellow and white arrows indicate examples of responder and non-responder cells respectively. Scale bar, 50 µm. Percent of cells in monolayer responding to increasing concentrations of Pam3CSK4. Data represent the mean ± SD of four replicates. d Primary mouse gut organoids were cultured as described in methods, treated for 30 mins with 10 µg/ml Pam3CSK4 and immunostained. Representative images are shown. Scale bar, 100 µm. Bar plot and histograms show quantification of percent responding cells in individual organoids and nuclear/cytoplasmic NF-κB amplitude of response in combined organoids. e Primary mouse mammary organoids were cultured as described in methods, treated for 30 mins with 10 µg/ml Pam3CSK4 and immunostained. Representative images of three independent replicates are shown. Scale bar, 30 µm. Bar plot and histograms show quantification as in panel d.

Fig. 3. Responder cell status is maintained over multiple generations via epigenetic mechanisms.

a WT MCF10A monolayers were grown for 24 (left) or 48 h (right), treated with 1 µg/ml Pam3CSK4, and immunostained for NF-κB. Arrows indicate clusters of responding cells. Scale bar, 100 µm. Images are representative of four replicates. b Monolayers were treated with 1 µg/ml Pam3CSK4 and immunostained for NF-κB response. Histograms show quantification of the median distance from each responder cell to the closest five responders in the monolayer versus randomized responder positions. c Schematic describing the workflow of the lineage-tracing experiment. Cell lineages were monitored for 60 h by tracking H2B-iRFP nuclear fluorescence. After the last time point cells were treated with 1 µg/ml Pam3CSK4, fixed and immunostained as described in Methods section. Responder status was measured in each terminal cell and the minimum number of status change events was assumed to reconstruct the lineage responder status. d Representative lineages obtained as described in panel c (Supplementary Fig. 5a). e Using data from lineage tracing the number of a responder cells having a sister, cousin, or extended relative cell that is a responder (blue) or non-responder (gray) was counted and the probability is reported. f Schematic to describe bistable switching rates obtained by model constructed from lineage tracing data. Responder and non-responder status is maintained with a low probability of switch per generation. See supplementary material for more information on the model. g WT MCF10A were cultured with epigenetic modifier inhibitors (HDACi (SAHA, 800 nM), HATi (A-485, 10 µM), HMTi (EPZ-6438, 5 µM), DNMTi (5-AzacytidineC, 500 nM, or 5-aza-2′-deoxycytidine (1 µM))) for 7 days prior to Pam3CSK4 treatment. Data represent the mean ± SD from four replicates. h Monolayers treated with 5-AzacytidineC as in panel g and cultured for 30 days from removal of the drug. At indicated times, the fraction of responder cells was measured as in Fig. 2a. Gray shading indicates the 95% confidence interval of the model prediction. Data represent the mean ± SD from three replicates.

Fig. 4. Bimodal TLR2 expression limits the fraction of responder cells.

a Wild-type MCF10A were stimulated with 1 µg/ml Pam3CSK4 or 100 ng/ml IL-1β. Quantification of nuclear/cytoplasmic p65 immunofluorescent stain is shown in histograms on logarithmic scale. Data represents three replicates. b Cells were transduced with TLR2 under a Tet Responsive Element 3rd Generation (TRE3G) promoter and incubated with or without Doxycycline (2 µg/ml) for 24 h. Cells were then stimulated with 1 µg/ml Pam3CSK4 and immunostained to determine responder status as in Fig. 2a. Representative images of three replicates are shown. Scale bar, 100 µm. c Immunoblot against TLR2 in cells treated with or without DNMT inhibitor (5-AzacytidineC, 500 nM) or HDACi (SAHA 800 nM). HSC70 was used as a loading control. Data represents two replicates. d Dual smRNA FISH-immunofluorescence for TLR2 and NF-κB was done as described in Methods section. Dashed line indicates cell boundaries, yellow squares highlight the TLR2 probes. Representative images of two replicates are shown. Scale bar, 10 µm. Quantification of TLR2 mRNAs in responders and non-responders determined by NF-κB nuclear translocation. N = 354 cells total (148 Responder, 206 Non-Responder), p = 3.8e-19 by χ² test with 9 degrees of freedom. e Western blot of TRE3G::TLR2 cells treated with a 4-h pulse of Doxycycline (2 µg/ml) and cultured for the indicated time. Relative amounts are normalized to HSC70 loading control. Standard deviation represents two independent experiments for each condition. Swarmplots represent technical replicates of the percentage of responding cells (top) and nucleus/cytoplasm NF-ĸB amplitude (bottom) in the same cells.

Fig. 5. TLR2 promoter methylation is bimodal.

a Bisulfite and nanopore sequencing data of 130 bp upstream of the TLR2 transcription start site (TSS) in MCF10A. Locations of transcription factor consensus binding sequence sites are indicated. b Histogram quantifications of the number of unmethylated CpGs per read between −110 and −60. c Nanopore methylation sequencing data of 100 bp upstream to 1000 bp downstream of TSS for indicated receptor. Percentage of unmethylated and methylated CpGs as well as percent responding cells as determined by p65 immunofluorescence following TLR5 (Flagellin), IL1R (IL-1B), and TLR2 (Pam) stimulation. d TLR2 locus targeted nanopore sequencing of genomic DNA isolated from primary mouse epithelial mammary cells. e Differentially expressed genes cosegregating with TLR2 negative and positive cells in publicly available single-cell RNA sequencing data from human breast epithelial tissues. f Percent TLR2 positive cells from ScRNA sequencing data collected from breast tissue of seven female donors. Only cells expressing luminal expression markers were considered (see Methods section for details).

Fig. 6. Oncogene-induced epigenetic modifications increase the responder percentage.

a Clonal MCF10A cell line containing TRE3G::BRAF^V600E was immunostained for p65 after 24 h of culture in the presence or absence of Doxycycline (2 µg/ml) after Pam3CSK4 (1 µg/ml) stimulation for 30 min. Representative images are shown. Scale bar, 50 μm. b Monolayers treated with Pam3CSK4 (1 µg/ml) only, or increasing duration of dox, or ERKi (5 μM ulixertinib), prior to Pam3CSK4 treatment. Data represents three replicates. c TLR2 immunoblotting of BRAF^V600E dox inducible cells as in panel b. HSC70 is loading control, relative amounts are reported with standard deviation representing two independent experiments. d TLR2 locus targeted nanopore methylation sequencing of genomic DNA isolated from human MCF10A cells with (Plus Dox, 96 h) or without (No Dox) BRAFV600E overexpression. e TRE3G::BRAF^V600E monolayers were cultured with dox or inhibitors (HATi, A-485 (10 µM) and RSKi, BI-D1870 (10 µM)) and percent Pam3CSK4 (1 µg/ml) responders was calculated by NF-κB immunofluorescence. Error bars represent three technical replicates.

Fig. 7. High steady-state NF-ĸB signaling increases the responder percentage.

a WT monolayers were treated with indicated stimulus four times over a 9-day period. Concentrations of inputs were: Poly(I:C) 1 µg/ml, TNFα 100 ng/ml, Flagellin 1 µg/ml, Pam3CSK4 1 µg/ml, IL-1β 100 ng/ml, IFNγ 5 µ/ml, and TGFβ 5 ng/ml. Error bars represent six technical replicates with n > 1000 cells per replicate. Two-sample t test *p < 0.05 and ***p < 0.001. b Schematic indicating experimental workflow for data in panel c. Monolayers were stimulated with inputs every 2 days and washed or not after 30 min. After the last stimulation (day 8), all monolayers were washed for 24 h before determining responder fraction with Pam3CSK4 as in Fig. 2a. c Monolayers treated as detailed in panel b, input concentrations as in panel a. Two biological replicates, five technical replicates for each condition, ***p < 0.001; two-sample t-test. d Summary of model. Epithelial tissue contains cells that are all-or-nothing responsive or non-responsive to lipopeptide agonists. Response status is defined by TLR2 expression, which is controlled by epigenetic modifications to the TLR2 promoter. At steady state the fraction of responders is low; however, high inflammatory signaling involving innate immune cell types (e.g. dendritic cells, macrophages), leads to changes in the rate of epigenetic switching that increase the fraction of responsive cells in the tissue.