Stochastic cytokine expression induces mixed T helper cell States

Abstract

During eukaryotic development, the induction of a lineage-specific transcription factor typically drives differentiation of multipotent progenitor cells, while repressing that of alternative lineages. This process is often mediated by some extracellular signaling molecules, such as cytokines that can bind to cell surface receptors, leading to activation and/or repression of transcription factors. We explored the early differentiation of naive CD4 T helper (Th) cells into Th1 versus Th2 states by counting single transcripts and quantifying immunofluorescence in individual cells. Contrary to mutually exclusive expression of antagonistic transcription factors, we observed their ubiquitous co-expression in individual cells at high levels that are distinct from basal-level co-expression during lineage priming. We observed that cytokines are expressed only in a small subpopulation of cells, independent from the expression of transcription factors in these single cells. This cell-to-cell variation in the cytokine expression during the early phase of T helper cell differentiation is significantly larger than in the fully differentiated state. Upon inhibition of cytokine signaling, we observed the classic mutual exclusion of antagonistic transcription factors, thus revealing a weak intracellular network otherwise overruled by the strong signals that emanate from extracellular cytokines. These results suggest that during the early differentiation process CD4 T cells acquire a mixed Th1/Th2 state, instructed by extracellular cytokines. The interplay between extracellular and intracellular signaling components unveiled in Th1/Th2 differentiation may be a common strategy for mammalian cells to buffer against noisy cytokine expression.

Figure 1. Tbx21 and Gata3 are transcribed simultaneously in individual CD4 T cells.

(A) Current gene regulatory network proposed to govern Th1/Th2 differentiation. (B) Visualization of single transcripts of Tbx21 (red) and Gata3 (green) in individual CD4 T cells 24 h after activation. White dashed lines are boundaries of individual cells. Scale bar, 10 µm. (C) Mean counts of Tbx21 and Gata3 transcripts per cell as a function of activation time. (D) Scatter plots of Tbx21 and Gata3 transcripts in individual cells, with marginal distributions. The red line is the median line that divides data points into halves. Individual cells do not show mutual exclusion of Tbx21 and Gata3 expression. (E) Scatter plots of Tbx21 and Gata3 transcripts at 24 h in CD4 T cells treated with Th1-polarizing condition supplemented with 10 ng/ml IFNγ and IL12 and 10 µg/ml anti-IL4 antibody. (F) Scatter plots of Tbx21 and Gata3 transcripts at 24 h in CD4 T cells treated with Th2-polarizing conditions supplemented with 10 ng/ml IL4 and 10 µg/ml anti-IFNγ antibody. Error bars are s.e.m. of replicate experiments.

Figure 2. Transcript and protein levels exhibit strong positive correlations.

(A, B) Visualization of single Tbx21 transcripts by mRNA-FISH (A) simultaneously with protein levels by immunofluorescence (B) at 24 h after activation. (C, D) Visualization of single Gata3 transcripts by mRNA-FISH (C) simultaneously with protein levels by immunofluorescence (D) at 24 h after activation. All scale bars are 10 µm. (E) Scatter plot of transcript counts versus protein levels for Tbet at 24 h, with a Pearson's correlation coefficient of 0.59 (p<10⁻⁴⁴). (F) Scatter plot of transcript counts versus protein levels for Gata3 at 24 h, with a Pearson's correlation coefficient of 0.85 (p<10⁻⁸⁴).

Figure 3. Ifng and Il4 are expressed in a rare cell population and their levels show no significant correlation with Tbx21 and Gata3 expression.

(A) Visualization of single transcripts of Tbx21 and Ifng, and Gata3 and Il4 in individual CD4 T cells at 48 h. All scale bars are 10 µm. (B) Distribution of Ifng and Il4 transcripts in individual CD4 T cells, with inset diagrams to better illustrate the fraction of cells that express non-zero copies of cytokines. (C) Scatter plots show a weak positive correlation between Tbx21 and Ifng expression, or between Gata3 and Il4 expression. (D) Fraction of cells that express Ifng (defined as >20 transcripts) and that expressing Il4 (defined as >50 transcripts) as a function of activation time. Error bars are s.e.m. of replicate experiments.

Figure 4. Inhibiting IFNγ and IL4 signaling down-regulates Tbx21 and Gata3, respectively.

(A) As the concentration of anti-IFNγ antibody increases, the mean number of Tbx21 transcripts per cell decreases, while that of Gata3 transcripts remains constant. The reverse is observed upon addition of anti-IL4 antibody. (B) Conversion of Tbx21-Gata3 scatter plot into polar coordinates (r,θ), where r is the distance from the origin and computed by , where t represents Tbx21 and g represents Gata3, and θ is the angle with x-axis and computed by in the range 0≤θ≤π/2. (C) Distribution of θ for cells under non-biased condition is uniform, using the same data as Figure 1D. (D) Distribution of θ indicates that as concentration of anti-IFNγ antibody increases, the cells adopt larger θ (Th2-like state). The reverse is observed upon addition of anti-IL4 antibody. Red dashed lines show the medians of θ. All data shown are from cells at 24 h. Error bars are s.e.m. of replicate experiments.

Figure 5. Sequestration of IFNγ and IL4 leads to mutually exclusive expression of Tbx21 and Gata3.

(A) Our model of the signaling network that governs Th1/Th2 differentiation. The thickness of arrows indicates the strength of interaction. The intracellular signaling network consists of all the interactions depicted in thin arrows. (B) Illustration of the CD4 T cell population during early activation. CD4 T cells are immersed in a well-mixed cytokine milieu established by the rare cytokine-expressing cells, leading to simultaneous and ubiquitous induction of Tbx21 and Gata3 expression in individual CD4 T cells. (C) Scatter plots showing down-regulation and mutual exclusion of Tbx21 and Gata3 transcripts in individual cells treated with both anti-IFNγ and anti-IL4 antibodies. (D) Distribution of θ shows that θ of most cells is very large (close to π/2) or small (close to 0) (same data as in Figure 5C). By two-sample Kolmogorov-Smirnov goodness-of-fit test, distribution of θ for cells under IFNγ and IL4 deprivation are significantly different from cells under non-biased condition, p<10⁻¹¹ at 16 h, p<10⁻¹⁹ at 24 h, p<10⁻⁵⁴ at 48 h. Error bars are s.e.m. of replicate experiments.

Figure 6. The distributions of cells expressing transcripts of Ifng (A), Il4 (B), Tbx21 (C), and Gata3 (D).

The data are fitted to Gamma distributions.