Interferon-β Acts Directly on T cells to Prolong Allograft Survival by Enhancing Regulatory T cell Induction through Foxp3 Acetylation

SUMMARY

Type I interferons (IFN) are pleiotropic cytokines with potent antiviral properties that also promote protective T cell and humoral immunity. Paradoxically, type I IFNs, including the widely expressed IFNβ, also have immunosuppressive properties, including promoting persistent viral infections and treating T cell-driven, remitting-relapsing multiple sclerosis. While associative evidence suggests that IFNβ mediates these immunosuppressive effects by impacting regulatory T (Treg) cells, mechanistic links remain elusive. Here, we found that IFNβ enhanced graft survival in a Treg cell-dependent murine transplant model. Genetic conditional deletion models revealed that the extended allograft survival was Treg cell-mediated and required IFNβ signaling on T cells. Using an in silico computational model and analysis of human immune cells, we found that IFNβ directly promoted Treg cell induction via STAT1- and P300-dependent Foxp3 acetylation. These findings identify a mechanistic connection between the immunosuppressive effects of IFNβ and Treg cells, with therapeutic implications for transplantation, autoimmunity, and malignancy.

Graphical Abstract

eTOC Blurb

The mechanism by which IFNβ, a pro-inflammatory cytokine, mediates its paradoxical immunosuppressive effects in autoimmunity and chronic viral infections remains elusive. Fueyo-González et al. use a transplant model to show that IFNβ synergizes with CTLA-4 Ig to prolong allograft survival. Mechanistically, IFNβ directly promotes Treg cell induction via activation of STAT1 signaling and P300-dependent Foxp3 acetylation.

INTRODUCTION

The type I interferon (IFN) system constitutes a family of cytokines that modulate both innate and adaptive immune responses by promoting an antiviral state in cells upon recognition of conserved viral and bacterial structures, and by eliciting inflammatory signals that activate dendritic cells and promote T and B cell effector functions (Crouse et al., 2015; Schoggins, 2019). The family is composed of 17(human)-18(mouse) members that bind to the type I interferon receptor (IFNAR), amongst which the interferon-alpha (IFNα) subtypes and interferon-beta (IFNβ) have been the most studied due to their wide expression (Ng et al., 2016). Strong evidence suggests that dysregulation of the type I IFN response can lead to disease (Ivashkiv and Donlin, 2014). A key feature of the type I IFN system is its rapid induction and amplification through a well-described feed-forward loop which ensures a strong immune response is mounted, but that also needs to be tightly regulated to avoid tissue damage (Hall and Rosen, 2010).

Paradoxically, IFNβ exhibits immunosuppressive properties in certain contexts, as illustrated by its successful use to treat remitting-relapsing multiple sclerosis (RRMS) (La Mantia et al., 2016; Lampl et al., 2013), and its role in establishing viral persistence in the lymphocytic choriomeningitis virus (LCMV) chronic infection mouse model (Ng et al., 2015; Teijaro et al., 2013). In both instances, evidence suggests that induction of regulatory T (Treg) cells is responsible in part for the observed immunosuppressive effects. IFNβ treatment in RRMS patients augments the frequency of peripheral blood Treg cells (de Andrés et al., 2007; Korporal et al., 2008; Namdar et al., 2010) and Treg cells are necessary for IFNβ’s protective and anti-inflammatory effects when given exogenously in murine experimental allergic encephalomyelitis (EAE) models (Axtell et al., 2010; Floris et al., 2002; Wang et al., 2016). IFNβ is also hypothesized to contribute to the observed increase in Treg cells in chronic LCMV models (Penaloza-MacMaster et al., 2014) and in peripheral blood of individuals with persistent infection with hepatitis C virus (HCV) (Ward et al., 2007) and herpes simplex virus (HSV) (Diaz and Koelle, 2006). However, the mechanism underpinning the counterintuitive link between IFNβ and Treg cells remains elusive.

Treg cells are essential modulators of central and peripheral tolerance. A reduction in Treg cell numbers or function is a hallmark of autoimmune disease in animal models (Kim et al., 2007) and in humans (Buckner, 2010), and Treg cells promote murine transplant tolerance and are associated with prolonged transplant survival in humans (Hall, 2016; Rothstein and Camirand, 2015). The transcription factor forkhead box P3 (Foxp3) is the master regulator of the Treg cells immunosuppressive program, and is responsible for Treg cell function and stability (van Loosdregt and Coffer, 2014). In Treg cells, Foxp3 expression and function are regulated through multiple molecular mechanisms including Foxp3 dimerization (Song et al., 2012), as well as Foxp3 post-translational modifications such as cleavage (de Zoeten et al., 2009), ubiquitination (van Loosdregt et al., 2013), and acetylation (Dahiya et al., 2020; Liu et al., 2012; Xiao et al., 2014).

Herein, we utilized a transplant model in which allograft survival is dependent on regulatory Treg cells (Conde et al., 2015; Purroy et al., 2017) and analysis of human immune cells to study the mechanisms through which IFNβ mediates its immunosuppressive functions. We found that IFNβ directly acts on T cells to enhance Treg cell induction both in vivo in mice and in vitro in mouse and human cells, as well as to prolong allograft survival (mouse). Guided by a mechanistic computational model we discovered that IFNβ achieves its Treg cell-enhancing effect by signal transducer of activation (STAT)1- and P300-dependent Foxp3 acetylation (van Loosdregt et al., 2010).

RESULTS

Exogenous interferon-beta synergizes with CTLA-4 Ig to prolong allograft survival.

We treated H-2^b C57BL/6 (B6) recipients of fully MHC-disparate H-2^d BALB/c heart allografts with either phosphate buffer saline (PBS), or IFNβ (10,000 U i.p. for 5 days peri-transplant and then weekly), in the presence or absence of co-stimulatory blockade with cytotoxic T-lymphocyte-Associated Antigen (CTLA)-4 Ig (160 μg on day 2 post-transplant) and monitored heart allograft survival through palpation (Fig. 1A). Consistent with previously published results (Yang et al., 2011), CTLA-4 Ig treatment prolonged graft survival in animals treated with PBS from a median survival time (MST) of 9 to 21 days. While IFNβ alone did not prolong graft survival, IFNβ synergized with CTLA-4 Ig to further prolong graft survival to an MST of 50 days (p<0.01, Fig. 1B).

Figure 1. IFNβ prolongs allograft survival in a Treg cell-dependent transplant model.

(A) Schematic of experimental design for (B)-(L). (B) Summary of the effect of IFNβ on graft survival of heterotopic cardiac allografts in the presence or absence of CTLA-4 Ig, (shown are pooled data from three independent experiments with n=7 animals/group. log-rank [Mantel-Cox] test, ** p<0.01). (C) Gene expression measured by RT-PCR from graft cells at day 14 post-transplant in control (CTLA-4 Ig + PBS) and IFNβ-treated (CTLA-4 Ig + IFNβ) mice (t-test with Bonferroni correction, *p<0.05). Shown are the color-coded z-scores of the pooled results from n=4 animals/group from two independent experiments; color represents z-score. (D-H) Representative scatter plots histograms, or ELISpot results, together with bar summaries indicating (D) the frequency of Foxp3⁺ cells in splenocytes, (E) their Foxp3 expression, (F) the number of donor-reactive T cells, (G) the number of Treg cell (CD4⁺Foxp3⁺), as well as (H) the ratio between them in mice treated with or without IFNβ at day 14 post-transplant. (I) Representative scatter plots and bar summaries of the frequency of graft infiltrating Foxp3⁺ cells and Foxp3 expression, and (J) the number of Treg cells to number of CD8⁺ T cells ratio 6 days post-transplant. (K) Representative scatter plots and (L) bar graph summaries of the frequencies of splenic NK cells 14 days post-transplant or graft infiltrating MDSCs 6 days post-transplant (mean ± S.E.M., n=4-7 per group, two independent experiments, t-test, *p<0.05; n.s. not significant). See also Figure S1.

In separate groups of transplanted mice treated with CTLA-4 Ig ± IFNβ (or PBS control) we sacrificed the recipients at day 14 post-transplant, prior to cessation of heartbeats in any of the animals, to analyze the graft tissue and peripheral immune profiles. We observed a reduction in gene expression of key inflammatory cytokines (Il6, Il1b, and Gzmb) by RT-PCR in the grafts of CTLA-4 Ig + IFNβ-treated animals when compared to CTLA-4 Ig + PBS controls (Fig. 1C). Previous studies showed that CTLA-4 Ig-induced graft survival in this model is dependent upon Treg cells that in turn suppress pathogenic donor-reactive CD4⁺ T helper (Th) 1 cells, and CD8⁺ T cells (Yang et al., 2011). We observed that IFNβ treatment increased the frequency of splenic Treg cells (CD4⁺ Foxp3⁺) (Fig. 1D and Fig. S1A for gating strategy) as well as their average Foxp3 expression (mean fluorescence intensity, MFI) (Fig. 1E). IFNβ also induced a decrease in splenic IFN-γ producing donor-reactive cells as detected by enzyme-linked immune absorbent spot (ELISpot) assays (Fig. 1F), resulting in an elevation in the Treg cell to donor-reactive cell ratio (Fig. 1G and H). At day 6 post-transplant, consistent with our findings in the periphery, we observed in the grafts of IFNβ-treated animals when compared to controls: (i) an increase in the frequency of Treg cells (CD4⁺ Foxp3⁺) and an increase in Foxp3 expression within Treg cells (Fig. 1I) and (ii) an increase in the Treg cell to CD8⁺ T cell ratio (Fig. 1J and Fig. S1B-F). Natural killer (NK) cells can recognize alloantigen (Chewning et al., 2007) and have been shown to modulate alloimmune responses to transplant (Fabritius et al., 2017; Hirohashi et al., 2012). We did not observe differences in frequencies of NK cells between groups (Fig. 1K and L). While intragraft CD11b⁺Ly6G⁻Ly6C^lo regulatory macrophages have also been shown by others to promote Treg cell-dependent allograft survival (Conde et al., 2015; Llaudo et al., 2018), our analyses did not show differences in intragraft frequencies of these myeloid cells in the presence or absence of IFNβ treatment (Fig. 1K and L). Together, these results raised the possibility that IFNβ prolongs allograft survival by directly acting on T cells.

IFNβ directly acts on T cells to enhance Treg cell induction and increase Foxp3 expression

To test the hypothesis that IFNβ prolongs allograft survival by directly acting on T cells, we isolated naïve splenic CD4⁺ T cells (CD44^lo CD62^hi purity >95%) from wild type B6 animals and confirmed surface expression of the type I interferon receptor (IFNAR) by imaging flow cytometry (Fig. 2A). Of note, we did not observe differences in IFNAR expression in CD4⁺Foxp3⁻ and CD4⁺Foxp3⁺ cells in transplant recipients treated with or without IFNβ at day 14 post-transplant (Fig S2A). We then cultured these naïve CD4⁺ T cells with αCD3/αCD28 activating beads under Treg cell polarizing conditions (interleukin IL-2 and tumor-growth factor TGFβ) in the presence or absence of IFNβ and quantified the frequency of Foxp3⁺ cells and their Foxp3 expression by flow cytometry 5 days later. These assays showed a dose dependent IFNβ-induced augmentation of the frequency of Foxp3⁺ cells and Foxp3 expression in cultures containing IL-2 and TGFβ (Fig. 2B-D). Control wells showed no IFNβ-induced Foxp3 expression in the absence of TGFβ (Fig. 2B-C). Addition of IFNβ did not affect the number of cells per well at the end of the culture nor did IFNβ alter viability or proliferation measured by using a dilution dye of Foxp3⁻ cells when compared with that of Foxp3⁺ T cells (data not shown). The Treg cell-enhancing effect was independent of TGFβ concentration over the range of 0.7 ng/ml to 3 ng/ml (Fig. S2B). Addition IFNα4 and IFNα2, currently FDA-approved for chronic hepatitis B, showed no effects on Treg cell induction supporting the conclusion that the effects are selective for IFNβ (Fig. S2C).

Figure 2. IFNβ directly acts on naïve T cells to enhance Treg cell induction.

(A) Imaging flow representative images and MFI summary bar graphs of murine CD4⁺ naïve T cells stained with an anti-IFNAR antibody conjugated to fluorescein isothiocyanate (FITC) or isotype control (mean ± S.E.M., n=3 per group, *p<0.05, t-test). (B) Representative scatter plots and bar summaries (C and D) of Treg cell induction cultures (CD4⁺ naïve T cells + αCD3/αCD28 + IL-2 + TGFβ) with or without IFNβ and controls. (E) Representative scatter plots and bar summaries (F) of analogous Treg cell induction cultures in Ifnar1^fl/fl x Cd4-Cre^NEG (Ifnar1^fl/fl) and Ifnar1^fl/fl x Cd4-Cre^POS (Ifnar1^ΔCD4) mice (mean ± S.E.M., n=3-6 per group, three independent experiments, ANOVA with post-hoc Tukey HSD test, *p<0.05, n.s. not significant). (G) Schematic of experimental design, (H) representative scatter plots, and (I) summary results of splenic cells 14 days after adoptive transfer of naïve CD4⁺ Foxp3-GFP⁻ in mice treated with PBS or IFNβ. (mean ± S.E.M., n=4 animals/group; two independent experiments, t-test, * p<0.05, ** p<0.01, n.s. not significant). See also Figure S2 and Figure S3.

To confirm the observed effects of IFNβ on Treg cells require IFNAR ligations on naïve T cells, we performed analogous experiments with naïve T cells obtained from Ifnar1^fl/fl x Cd4-Cre^POS (Ifnar1^ΔCD4) mice, in which a floxed sequence has been introduced with loxP sites flanking exon 3 of the type I interferon gene sequence (ifnar1) such that its removal by the Cd4-controlled Cre recombinase results in a non-functional version of IFNAR in T cells (Fig. S2D), and the corresponding Cd4-Cre^NEG controls (Ifnar1^fl/fl). These assays (Fig. 2E and F) showed that while addition of IFNβ enhanced Treg cell induction in Ifnar1^fl/fl T cells under Treg cell polarizing conditions, IFNβ had no effect on the frequency or Foxp3 expression of Ifnar1^ΔCD4 T cells Treg cell induction cultures.

To verify that Foxp3⁺CD4⁺ T cells induced in the presence of IFNβ can suppress effector T cells, we performed in vitro suppression assays. We induced alloreactive Foxp3⁺CD4⁺ T cells in the presence or absence of IFNβ by culturing CD4⁺ naïve T cells from Foxp3-GFP animals with BALB/c antigen-presenting cells (CD90.2⁻) pre-incubated with an IFNAR-blocking antibody to isolate the direct effect of IFNβ on T cells. After confirming an IFNβ-mediated increase in Foxp3⁺ cells in allostimulated Treg cells (Fig. S3A-C), we sorted Foxp3-GFP⁺ cells by flow cytometry and assessed their ability to suppress proliferation of carboxyfluorescein succinimidyl ester (CFSE)-labeled B6 conventional T (Tconv) cells stimulated in the presence of BALB/c antigen-presenting cells (APCs). These assays (Fig. S3D) demonstrated that the alloreactive Foxp3⁺CD4⁺ T cells induced in the presence of IFNβ exhibited similar suppressive capacity as those induced in the absence of IFNβ (Fig. S3D). We observed similar suppressive capacities of sorted CD4⁺CD25^hiCD127⁻ Treg cells induced in the presence or absence of IFNβ from wild type naïve CD4⁺ T cells when tested against effector T cells stimulated with either allogeneic BALB/c antigen-presenting cells (APCs) (Fig. S3E and F) or anti-CD3/CD28 (Fig. S3G and H).

To test the effects of exogenous IFNβ on Treg cell induction in vivo, we adoptively transferred naïve, GFP⁻, CD4⁺ from Foxp3-GFP mice into Rag1^−/− mice, which produce no mature T cells or B cells, and treated them with either PBS (control) or IFNβ and analyzed frequencies of Foxp3-GFP⁺ CD4⁺ T cells and their Foxp3 expression 14 days later (Fig. 2G) as we previously published (van der Touw et al., 2013). We observed a ~50% increase in the percentage of GFP⁺ Treg cells and up to 4-fold increase in the total number GFP⁺ Treg cells in the IFNβ-treated animals as well as an increase in their Foxp3 expression (Fig. 2H and I). Together, these results indicate that IFNβ promotes Treg cell induction and increases Foxp3 expression in Treg cells both in vitro and in vivo.

Prolongation of allograft survival is Treg cell-dependent and requires direct action of IFNβ on T cells.

To establish a functional connection between our observations that IFNβ prolongs heart graft survival (Fig. 1) and augments Treg cell induction (Fig. 2), we tested the hypothesis that IFNβ prolongs allograft survival via a Treg cell-dependent mechanism that requires IFNβ-IFNAR ligation on recipient T cells. We transplanted BALB/c hearts into wild type (WT) or transgenic DTR-Foxp3-GFP mice (B6 background) in which diphtheria toxin treatment depletes Treg cells, as previously demonstrated by our group (Donadei et al., 2019). We treated both WT and DTR-Foxp3-GFP mice following the same protocol for CTLA-4 Ig as in previous experiments and compared mice treated with PBS or IFNβ. Survival analyses showed accelerated rejection in BALB/c hearts in IFNβ-treated DTR-Foxp3-GFP recipients (MST 18 days) vs control WT (MST 49.5 days) (Fig. S4A). We set up analogous experiments in Ifnar1^fl/fl and Ifnar1^ΔCD4 mice. Selectively eliminating IFNAR signaling in recipient T cells prevented the IFNβ-mediated prolongation of allograft survival (Fig. 3A) and prevented the IFNβ-induced augmentation of Foxp3⁺ cells and the IFNβ-induced reduction in the number donor-reactive T cells (ELISpot and flow cytometry) observed in the control animals (Fig. 3B-G, Fig S4B-C). Together, the results support the hypothesis that in transplanted mice treated with co-stimulatory blockade, IFNβ acts directly on T cells to enhance Treg cell induction thereby prolonging allograft survival.

Figure 3. Prolongation of allograft survival is Treg cell-dependent and requires direct action of IFNβ on T cells.

(A) Comparison of survival curves of BALB/c hearts transplanted into Ifnar1^fl/fl x Cd4-Cre^NEG (Ifnar1^fl/fl) and Ifnar1^fl/fl x Cd4-Cre^POS (Ifnar1^ΔCD4) mice, treated with or without IFNβ. (n=5-7 animals/group pooled from three independent experiments, log-rank [Mantel-Cox] test * p<0.05, ** p<0.01, n.s. not significant). (B) Representative scatter plots and (C) bar summaries of the percentage and Foxp3⁺ cells and their Foxp3 expression, (D) donor-reactive cells determined by ELISpot, and (E) the ratio between them at day 14 post-transplant in recipients treated with or without IFNβ. (F) Bar summary of number of alloreactive CD4⁺ and CD8⁺ producing IFNγ or (G) TNFα determined by flow cytometry following splenocyte stimulation with BALB/c cells. (mean ± S.E.M., n=4 animals/group from two independent experiments, one-way ANOVA with post-hoc Tukey HSD test; * p<0.05, ** p<0.01, n.s. not significant). See also Figure S4.

IFNβ acts on newly generated Treg cells to prolong allograft survival

During Treg cell induction, naïve CD4⁺ T cells exposed to Treg cell polarizing conditions, in vitro or in vivo, upregulate Foxp3 expression within hours, which guides the development of the Treg cell phenotype while inhibiting Th17 cell-promoting transcription factor (TF) RAR-related orphan receptor (ROR)γt, Th2 cell-promoting TF GATA-binding (GATA)-3, and Th1 cell-promoting TF T-box expressed in T cells (T-bet) expression (Carbo et al., 2013; Tartar et al., 2010). To test the hypothesis that IFNβ facilitates this transition from CD4⁺ naïve to Treg cells we measured the percentage of Foxp3⁺ cells at day 5 in Treg cell induction cultures in which we added IFNβ only on day 0, 1, 2, 3 or 4. We observed the maximum enhancing effect when we added IFNβ at day 0, and its effect dwindled progressively when added at later days in the culture. This decrease in effect inversely correlated with the increase in the frequency of Foxp3⁺ cells in the culture supporting the idea that IFNβ’s main action occurs during the early phases of Treg cell generation (Fig. 4A).

Figure 4. IFNβ acts on newly generated Treg cells to prolong allograft survival.

(A) Bar summaries of the percentage of Foxp3⁺ cells at the end of Treg cell induction cultures (CD4⁺ naïve T cells + αCD3/αCD28 + IL-2 + TGFβ) to which IFNβ was added at different days (orange hue indicates the number of Foxp3⁺ cells in the culture when IFNβ was added, * p<0.05, **p<0.01, t-test). (B) Representative scatter plots and (C) bar summaries of the percentage and Foxp3⁺ cells and their Foxp3 expression in Ifnar1^fl/fl x Foxp3-YFP-Cre^NEG (Ifnar1^fl/fl) and Ifnar1^fl/fl x Foxp3-YFP-Cre^POS (Ifnar1^ΔFoxp3) Treg cell induction cultures with or without IFNβ. (D) Comparison of survival curves of BALB/c hearts transplanted into Ifnar1^fl/fl and Ifnar1^ΔCD4 mice, treated with or without IFNβ. (n=4-5 animals/group pooled from two independent experiments, log-rank [Mantel-Cox] test * p<0.05, ** p<0.01, n.s. not significant). (E) Representative scatter plots and bar summaries of the percentage distribution of total Treg cells, iTreg (Foxp3⁺Nrp1⁻) cells and nTreg (Foxp3⁺Nrp1⁺) cells in the spleen of Ifnar1^fl/fl or Ifnar1^ΔFoxp3 transplant recipients, (F) the nTreg cell / iTreg cell ratio, as well as (G) Foxp3 and (H) CD25 expression (C,E,F,G and H, mean ± S.E.M., n=4 animals/group from two independent experiments, one-way ANOVA with post-hoc Tukey HSD test; * p<0.05, ** p<0.01, n.s. not significant). See also Figure S5.

To specifically test whether the effects of IFNβ on graft rejection are Treg cell-specific, we utilized Ifnar1^fl/fl x Foxp3-YFP-Cre^POS (Ifnar1^ΔFoxp3) mice and their corresponding Ifnar1^fl/fl x Foxp3-YFP-Cre^NEG (Ifnar1^fl/fl) controls. In these animals, as Treg cells are being induced and once Foxp3 is translated, a Cre recombinase cuts the floxed exon in the Ifnar1 gene and renders the IFNAR nonfunctional in the animal (Fig. S5A). Twenty-four hours after initiating Treg cell induction cultures using naïve CD4+ (Foxp3⁻) Ifnar1^ΔFoxp3 T cells the Foxp3-YFP⁺ cells already show a decrease in IFNAR’s signaling capability through pSTAT1 (Fig. S5B). Consistent with our finding that IFNβ’s enhancing effects on Treg cell induction require IFNAR signaling as Foxp3 is being produced in CD4⁺ naïve cells, we confirmed that IFNβ did not increase the percentage of Foxp3⁺ cells, nor did it alter Foxp3 expression in Ifnar1^ΔFoxp3 Treg cell induction cultures (Fig 4B and C).

We next transplanted allogeneic BALB/c heart grafts into Ifnar1^ΔFoxp3 and Ifnar1^fl/fl recipients, treated them with or without IFNβ, and monitored allograft survival (Fig.1A). In Ifnar1^ΔFoxp3 mice, in which the IFNAR is rendered nonfunctional as Treg cells are being induced, the IFNβ-mediated prolongation of the allograft survival was abolished (Fig 4D). Flow cytometry phenotyping of the T cell compartments of non-transplanted Ifnar1^ΔFoxp3 and Ifnar1^fl/fl mice showed no differences. (Fig. S5C), indicating that IFNAR signaling in Treg cells was required for the allograft protective effect.

We used the surface marker neuropilin 1 (Nrp1) (Weiss et al., 2012; Yadav et al., 2012), to differentiate the effects of IFNβ on induced Treg (iTreg, Foxp3⁺Nrp1⁻) cells from those on natural Treg (nTreg, Foxp3⁺Nrp1⁺) cells, which in Ifnar1^ΔFoxp3 animals also lack a functional IFNAR. We confirmed in non-transplanted mice that this staining strategy was effective at identifying iTreg cells and nTreg cells, and that there were no differences in the percentages of iTreg cells and nTreg cells between Ifnar1^fl/fl and Ifnar1^ΔFoxp3 animals prior to transplant. (Fig. S5D). In Ifnar1^fl/fl allograft recipients, but not the Ifnar1^ΔFoxp3, IFNβ selectively increased the frequency of iTreg cells and their Foxp3 expression leaving that of nTreg cells unchanged (Fig. 4E-G). Of note, in all groups nTreg cells showed similar protein expression of the TF Helios (Fig. S5E), and Foxp3⁺ cells showed no differences in glucocorticoid-induced TNFR-related protein (GITR) expression, a marker associated with lower Treg cell suppression (Shimizu et al., 2002) (Fig. S5F). IFNβ also increased the expression of CD25, a marker of suppressive capacity in Foxp3⁺ cells from Ifnar1^fl/fl animals, which remained unchanged in their Ifnar1^ΔFoxp3 counterparts. (Fig 4H). Together, these results support the conclusion that IFNβ predominantly acts on newly generated (induced) Treg, cells, and not effector T cells, to prolong allograft survival.

Computational prediction of the molecular mechanism of IFNβ-enhanced Treg cell induction.

To identify potential molecular mechanisms by which exogenously administered IFNβ enhances Treg cell induction we utilized a computational mechanistic model. This approach allowed us to explore in minutes many mechanisms that would otherwise take days or weeks to test experimentally, and thus accelerate hypothesis generation.

We reused an existing ordinary differential equation (ODE) model of CD4⁺ naïve T cell polarization publicly available (Carbo et al., 2013) and expanded it to incorporate the effects of IFNβ. We included IFNAR as an additional input to the cell, and allowed it to signal phosphorylating STAT1, STAT3, STAT4, STAT5, and STAT6. Although IFNAR predominantly signals through STAT1 in the context of viral infections, it has been described to activate different STATs depending on tissue and cell type (van Boxel-Dezaire et al., 2006). This flexibility is achieved in the model by coupling the IFNAR’s signaling with variable weights to the different STATs and making those weights parameters in the model (Fig. 5A) (see also Data S1 section 1 for details on computational model building and model assumptions).

Figure 5. Computational prediction of the molecular mechanism of IFNβ-enhanced Treg cell induction.

(A) Computational strategy to study the effects of IFNβ on T cells. (B) Model calibration to mouse experimental data (top left) and computational simulations of the impact of IFNβ-IFNAR-mediated activation of the different STAT signaling pathways on the percentage of Foxp3⁺ cells simulated 5 days in culture (α₁, α₃, α₄, α₅, and α₆ are the weights that control the coupling to STAT1, STAT3, STAT4, STAT5 and STAT6 respectively; the value that maximizes the overall effect of IFNβ for each weight is depicted in color). (C) Computational simulations of the effects of pSTAT1 and pSTAT5 on Foxp3 acetylation as the IFNβ concentration increases (α1=0.983, α3=0.001, α4=0605, α5=0.858, and α6=0.098. IFNβ from 0 to 10,000 U/ml), and (D) the effects of pSTAT4 on pSTAT1 phosphorylation as IFNAR coupling to STAT4 increases. (E) Molecular mechanistic hypothesis generated by the computational model. See also Data S1.

We first tested whether the computational model recapitulated the increase in Treg cells observed experimentally. We verified that the expanded model adequately polarized in the absence of IFNβ CD4⁺ naïve T cells to Th1 (in the presence of IL-12, IL-18 and IFNγ), Th2 (with IL-4 and IL-2), Th17 (with IL-6 and TGF-β) and Treg (with IL-2 and TGF-β) cells (Data S1 section 2) We tested in silico the potential effects of IFNβ on each of these polarizations by allowing the coupling weights (α_i) of IFNAR to the different STATs to vary in the model such that they maximized the effect of IFNβ on each of the polarizations (see Data S1 for details on the optimization approach). Simulations of the optimized models for each polarization indicate a stronger and unexpected effect of IFNβ on the Treg cell polarization (Data S1 section 2), which confirms that our computational model captures the biology of the direct effects of IFNβ on T cells to enhance Treg cell polarization observed experimentally.

We next used the validated computational model to gain insight and generate a hypothesis regarding the molecular mechanisms responsible for the IFNβ-enhanced Treg cell induction. We calibrated the Treg cell optimized computational model with the experimental results from Figure 2C (Fig. 5B and Data S1 section 3). Simulation of the calibrated model allowing the ligated IFNAR only to signal through one STAT at a time, predicted that the increase of Foxp3⁺ cells is primarily driven by STAT1 activation with smaller contributions by STAT4 and STAT5 activation (Fig. 5C). A closer look at the reactions in the model revealed that the contributions of phosphorylated STAT1 (pSTAT1) and pSTAT5 result in a direct increase in the of acetylated Foxp3, which is the stable form of Foxp3 and responsible for the immunosuppressive program (van Loosdregt and Coffer, 2014). In contrast, pSTAT4 indirectly accelerates the rise of pSTAT1 thereby enhancing pSTAT1-mediated Foxp3 acetylation effects (Fig. 5C-E). Our computational model thus provided a molecular mechanistic hypothesis and postulated that IFNβ enhances Treg cell induction through the promotion of Foxp3 acetylation.

IFNβ increases Foxp3 acetylation through P300

To measure Foxp3 acetylation experimentally we utilized a proximity-ligation assay (PLA) developed by Hancock and Beier (Jiao et al., 2017) together with an image analysis Matlab toolbox developed in the lab to automate single-cell spot count associated with acetylated Foxp3 molecules. We set up Treg cell induction cultures in the presence or absence of IFNβ and measured the average number of acetylated Foxp3 counts per cell. Consistent with the model prediction, the number of spots in cultures with IFNβ (median 12) was higher than in those without (median 8), supporting the concept that IFNβ mediates an increase of Foxp3 acetylation in these cells (Fig. 6A and Fig. S6A)

Figure 6. IFNβ increases Foxp3 acetylation through P300 via STAT1 signaling.

(A) Violin plots depicting the distribution of Foxp3-acetylation by proximity ligation assay (solid line represents the median, dotted lines indicate quartiles, N=231-245 cells per group from 3 independent experiments, only cells with at least one spot included; ** p<0.01, Mann Whitney test). (B) Gene expression measured by RT-PCR in analogous Treg cell induction cultures (CD4⁺ naïve T cells + αCD3/αCD28 + IL-2 + TGFβ) at day 5 (n=4 animals/group, * p<0.05, t-test with Bonferroni correction, color represents z-score) and (C) representative histograms and bar summaries of P300 and TIP60 protein expression. (D) Representative scatter plots, (E) bar summaries of the percentage of Foxp3⁺ cells and their Foxp3 expression, and (F) violin plots depicting acetylated Foxp3 in identical Treg cell induction cultures in the presence or absence of 1 μM P300 inhibitor C646 with or without IFNβ. (G-H) Comparison in splenic CD4⁺Foxp3⁺ cells isolated from Ifnar1^fl/fl x Cd4-Cre^NEG (Ifnar1^fl/fl) and Ifnar1^fl/fl x Cd4-Cre^POS mice (Ifnar1^ΔCD4) transplant recipients at day 14 post-transplant of (G) P300 expression and (H) average Foxp3 acetylation determined by PLA. (I) Analogous comparison of Foxp3 acetylation in Ifnar1^fl/fl x Cd4-Cre^NEG (Ifnar1^fl/fl) and Ifnar1^fl/fl x Foxp3-YFP-Cre^POS mice (Ifnar1^ΔFoxp3) transplant recipients. (J) Survival curves in BALB/c hearts transplanted into B6 mice treated with IFNβ with or without 5 mg/kg C646 given at the same time as IFNβ (n=4-6 animals/group, l log-rank [Mantel-Cox] test, ** p<0.01, n.s. not significant). (K) Representative scatter plots and (L) summary bar graphs of the percentage of Foxp3⁺ cells and (M) violin plots of Foxp3 acetylation in analogous Treg cell induction cultures with Stat1^−/− and Stat4^−/− cells. (C,E,G,H, I and L n=4-5 from at least two independent experiments, one-way ANOVA with Tukey HSD; H,I Foxp3 acetylation median values from N=105-224 cells per mouse were obtained and the averages compared; F and K Kruskal-Wallis test followed by Dunn’s multiple comparison test, F N=170-202 cells from three independent, M N= 112-129 cells from two independent experiments, *p<0.05, ** p<0.01, ***p<0.001, n.s. not significant). See also Figure S6 and Figure S7.

In the Treg cell-optimized computational model, IFNβ activated pSTAT1 which in turn induced lysine acetyltransferases (KATs) responsible for the increase in Foxp3 acetylation. We performed gene expression analyses of all KATs in Treg cell induction cultures and observed a significant increase in gene induction of 4 KATs in IFNβ-treated cultures at day 5 when compared to those without IFNβ (Fig. 6B). Among these KATs, only P300 and TIP60 have been described to acetylate Foxp3 (Xiao et al., 2014). Both the KAT P300 (encoded by Kat3b) and its cofactor (encoded by Kat2b) showed the strongest increase with IFNβ, which singled out this KAT as the most likely driver of the IFNβ-mediated increase in Foxp3 acetylation. We used flow cytometry to measure protein expression of P300 and TIP60 in the CD4⁺Foxp3⁺ compartment. Indeed, we observed increased P300 expression, and not TIP60, in IFNβ-treated cells when compared to untreated cells, an increase that was lost in the absence of a functional IFNAR (Fig. 6C). Furthermore, the P300-selective inhibitor C646 abolished the IFNβ-elicited increases in Foxp3⁺ cell frequencies, Foxp3 protein expression (Fig. 6D-E), and Foxp3 acetylation in Treg cell induction cultures (Fig. 6F) supporting the key role of this KAT in IFNβ-mediated Foxp3 acetylation.

Consistent with our in vitro results, we observed higher P300 expression (Fig. 6G and Fig. S6B) and increased acetylated Foxp3 in splenic Foxp3⁺ cells in the IFNβ-treated group in Ifnar1^fl/fl but not in Ifnar1^ΔCD4 CTLA-4 Ig-treated transplant recipients (Fig. 6H). Similarly, we observed no increase in Foxp3 acetylation by IFNβ in Ifnar1^ΔFoxp3 animals confirming that this mechanism is operating in induced Treg cells (Fig.6I) To test for a functional link between IFNβ-induced P300 upregulation and IFNβ-induced, Treg cell-dependent heart graft survival we transplanted BALB/c hearts into CTLA-4 Ig-treated recipients and administered IFNβ ± a P300 inhibitor C646 (5 mg/kg given at the same time as PBS or IFNβ). Survival analyses (Fig. 6J) showed that treatment with C646 reversed the IFNβ-mediated prolongation of allograft survival (noting that all curves were shifted to the left when compared to previous experiments potentially due to the 5% DMSO present in the vehicle and C646). Together, these results indicate that IFNβ promotes P300 expression to increase Foxp3 acetylation thereby enhancing Treg cell induction.

IFNβ-mediated Foxp3 acetylation requires STAT1 signaling and promotes a Foxp3-specific transcriptional program.

To connect the increase in Foxp3 acetylation to the signaling through the IFNAR receptor, we used flow cytometry to measure pSTATs in CD4⁺ naïve cells from splenocytes cultured with αCD3/αCD28, IL-2, TGFβ and IFNβ. The experiments indicate that in these cells and in this context, IFNβ signals through pSTAT1, pSTAT4 and pSTAT5 (enhancing IL-2-mediated signaling critical in Treg cells) and not pSTAT3 or pSTAT6 (Fig. S6C and D). Controls with IFNβ in the absence of TGFβ showed a similar pSTAT1/pSTAT4/pSTAT5 pattern of activation (data not shown). We also investigated the expression of Smad7 protein in Foxp3⁺ cells from BALB/c heart recipients at day 14 post-transplant treated with or without IFNβ and observed no differences (Fig. S6E). TGFβ-mediated Smad2 phosphorylation in Treg cells in vitro cultures remained unchanged in the presence of IFNβ further confirming pSTAT signaling as the main mediator of the Treg cell-enhancing effects elicited by IFNβ (Fig.S6F).

Both pSTAT1 and pSTAT4 have predicted binding sites in the promoter regions of the Kat3b gene (P300) (UCSC Genome Browser, Transcriptor Factor ChIP-seq Uniform Peaks from ENCODE/Analysis May 2013). To assess the relative contribution of pSTAT1 and pSTAT4 signaling to the enhancement of Treg cell induction we set up Treg cell induction cultures with CD4⁺ naïve cells from STAT1 knockout (Stat1^−/−) and STAT4 knockout (Stat4^−/−) mice. Consistent with our computational model prediction (Fig. 5), while IFNβ promoted an increase in the frequency of Foxp3⁺ cells and Foxp3 acetylation in cells from Stat4^−/− animals, these two effects were absent in Stat1^−/− cells (Fig. 6K-M). Notably, cultures with Stat1^−/− cells polarized more poorly into Treg cells (compare with Figure 2B) even in the absence of IFNβ, suggesting a wider role of pSTAT1 in controlling Foxp3 expression in the cell probably through the modulation of KATs (such as P300) expression at baseline.

Foxp3 acetylation has been associated with increased stability of the Foxp3 protein (Tao et al., 2007). We investigated whether this stability is reflected transcriptionally in CTLA-4 Ig treated Foxp3-YFP (B6 background) transplant recipients following IFNβ administration when compared to vehicle (control). We sorted splenic Foxp3-YFP⁺ cells at day 14 post-transplant and assayed their transcriptional program through RNAseq. The differentially expressed genes (DEG) between the two groups comprised only a few genes (N=26) indicating that the transcriptional program of Foxp3⁺ cells in mice treated with IFNβ and vehicle is similar.

Foxp3 regulates gene expression through binding to DNA in its homodimeric form (Bandukwala et al., 2011), but also through heteromerization with other transcription factors and chromatin modifiers (Marson et al., 2007; Song et al., 2012; Zheng et al., 2007). Foxp3 ectopic expression studies in mouse T cells have led to the identification of a gene signature of 1255 upregulated genes and 1152 downregulated specifically associated to Foxp3 in Treg cells (Xie et al., 2015). We performed gene set enrichment analyses (GSEA) between the genes with increased and decreased expression in IFNβ-treated animals (noting that Kat3b was increased log₂FC>1.5, and Foxp3 was not different, log₂FC<0.1), and the publicly available Foxp3 signature of upregulated and downregulated genes, respectively. Our analyses confirmed an enrichment of Foxp3-regulated genes in Treg cells from IFNβ-treated animals when compared to those from controls (Fig. S7A). The enriched upregulated genes associated with gene ontology (GO) terms linked to purine metabolism, whereas the enriched downregulated genes associated with GO terms linked to the Th1 cell phenotype, granzyme A and cytokine production, and IL-4 (Fig. S7B). Consistently, Treg cells from mice treated with IFNβ showed a moderate but statistically significant increase in suppressive capacity when compared to those of controls (Fig. S7C).

Together, our results support the hypothesis put forward by the computational model and indicate that IFNβ/IFNAR ligation activates pSTAT1 signaling, which in turn drives the gene and protein expression of P300 to promote Foxp3 acetylation increasing its function.

IFNβ enhances Treg cell induction in humans by promoting Foxp3 acetylation

Finally, we tested whether the enhancing effect of IFNβ on Treg cell induction applies to human cells. We set up analogous Treg cell induction cultures to those used in our mouse experiments. We used magnetic enrichment to obtain naïve CD4⁺ T cells (CD45RA⁺CD45RO⁻, >95% purity assessed by flow cytometry) from anonymous donor peripheral blood mononuclear cells (PBMCs). We assessed the expression of IFNAR in these cells (Fig. 7A) and cultured them with αCD3/αCD28, TGFβ, IL-2 with or without IFNβ. We performed additional control cultures removing TGFβ, with IFNβ instead of TGFβ, and in the presence of an IFNAR-blocking antibody. Addition of IFNβ did not affect the number of cells per well at the end of the culture, nor did IFNβ alter viability or proliferation, measured by CFSE dilution, of Foxp3⁻ cells compared with that of Foxp3⁺ T cells (data not shown). Addition of IFNβ increased the frequency of Foxp3⁺ cells in culture wells, their intracellular Foxp3 expression (MFI), as well as their surface expression of CD25. We observed these effects only in the presence of Treg cell polarizing conditions (IL-2 + TGFβ). Addition of anti-IFNAR neutralizing antibody abrogated the observed increases (Fig. 7B-D). Furthermore, sorted CD4⁺CD25^hiCD127⁻ cells from cultures treated with and without IFNβ suppressed T cell proliferation with comparable efficiency (Fig. S7D and E). Lack of IL-2 production has been associated with Treg cell suppressive function in induced human Treg cells (see (Shevach, 2018) for a review). To further characterize the potential immunosuppressive capacity of induced Foxp3⁺ human Treg cells and avoid potential artifacts due to the lack of Foxp3⁺ purity in the CD4⁺CD25^hiCD127⁻ compartment, we quantified intracellular IL-2 by flow cytometry in Foxp3⁺ cells at the end of Treg cell induction cultures with or without IFNβ. These assays showed lower IL-2 production in Foxp3⁺ cells from IFNβ-treated cultures (Fig. 7E), suggesting a potential increased immunosuppressive capacity.

Figure 7. IFNβ enhances Treg cell induction in humans by promoting Foxp3 acetylation.

(A) Imaging flow representative images and MFI summary bar graphs of human CD4⁺ naïve T cells stained with an anti-IFNAR antibody conjugated to FITC or isotype control (mean ± S.E.M., n=3 per group, *p<0.05, t-test). (B-D) Representative scatter plots or histograms and bar summary of the (B) frequency of Foxp3⁺ cells, (C) Foxp3 and (D) CD25 expression in Treg cell induction cultures (human CD4⁺ naïve T cells from PBMCs + αCD3/αCD28 + IL-2 + TGFβ) at day 5 with the addition of IFNβ (1000 U/ml) with control cultures without TGFβ, IFNβ replacing TGFβ, or in the presence or absence of a neutralizing anti-IFNAR antibody or isotype control (mean ± S.E.M., n=3-5 donors from two independent experiments, ANOVA with Tukey HSD t-test, * p<0.05, ** p<0.01, n.s. not significant). (E) Representative histogram of intracellular IL-2 cytokine production in Foxp3+ cells at the end of Treg cell induction cultures with or without IFNβ (t-test, *p<0.05) (F) Activation of pSTATs in human Treg cell induction cultures 40 min after addition of IFNβ to the culture (peak value). (G) KAT gene expression measured by RT-PCR in human Treg cell induction cultures at day 5 (n=4 donors/group, * p<0.05, t-test with Bonferroni correction, color represents z-score). (H) Representative scatter plots and bar summaries of Foxp3⁺ cells in Treg cell induction cultures with or without IFNβ in the presence of P300 inhibitor C646 as well as (I) violin plots depicting PLA Foxp3 acetylation in these cultures (n=3-4 donors per group pooled from two independent experiments; E mean ± S.E.M, one-way ANOVA followed by Tukey’s test; G N=227-245 pooled from 3 different donors, Kruskal-Wallis test followed by Dunn’s multiple comparison test, *p<0.05, ** p<0.01, *** p<0.001, n.s. not significant). See also Figure S7.

We explored whether IFNβ mediates the increase in Treg cell induction in human cells through the same molecular mechanism as in mice. In human CD4⁺ naïve T cells activated in Treg cell polarizing conditions (IL-2 and TGFβ), IFNβ elicited pSTAT1, pSTAT4 and pSTAT5 signaling (Fig. 7F and Fig. S7F) and induced expression of KAT3B (Fig. 7G), the gene encoding the protein P300. Furthermore, presence of the selective P300 inhibitor C646 returned the frequency of Foxp3⁺ cells in IFNβ cultures to those of cultures without (Fig. 7H). As in our mice experiments, PLA analyses established that IFNβ increased Foxp3 acetylation, and that this increase is sensitive to the presence of the P300 inhibitor (Fig. 7I). Together, these results indicate that IFNβ also enhances Treg cell induction in human cells through a pSTAT1- and P300-dependent mechanism.

DISCUSSION

We sought out to understand the mechanisms underlying the unexplained immunosuppressive effects of IFNβ. The results presented herein show that this key type I interferon acts on murine and human T cells to enhance Treg cell induction by promoting Foxp3 acetylation through a pSTAT1- and P300-dependent mechanism, and illustrate, using a Treg cell-dependent mouse transplant model, the relevance of this Treg cell-enhancing mechanism in vivo. While Foxp3 acetylation has also been previously associated with Treg cell stability (Tao et al., 2007), our results highlight an undescribed role to date for Foxp3 acetylation in Treg cell induction separate from transcriptional regulation of the Foxp3 gene.

The results presented in this work provide a molecular mechanism for the observation that treatment with human recombinant IFNβ markedly improves the frequency and suppressive function of Treg cells in patients with RRMS (Vandenbark et al., 2009), an increase that is associated with a reduction in flare-ups and disease severity (Bendtzen, 2010). These IFNβ therapeutic benefits have been proposed to be associated with an increase in IL-10 production in Tr1, a Treg cell subset (Astier et al., 2006; Rudick et al., 1996), or to the effects of IFNβ on microglial cells resulting in a reduction in their ability to present antigen, recruit and activate autoreactive encephalitogenic T cells (Teige et al., 2006). The Foxp3-acetylation mechanism described here prompts the idea that responsiveness to IFNβ therapies in RRMS patients might be correlated with Foxp3 acetylation in Treg cells.

Our results put forward an explanation for the coexistence of elevated type I interferon signatures and increased Treg cells in patients with persistent infections, which has been counterintuitive and hard to reconcile mechanistically. The best example of this coexistence is found in chronic hepatitis C virus (HCV) infection (Guidotti and Chisari, 2006; Heim, 2013; Su et al., 2002; Ward et al., 2007), although variations of these features are present in human immunodeficiency (HIV) (Hardy et al., 2013; Stylianou et al., 2000) and herpes simplex virus (HSV) infections (Diaz and Koelle, 2006; Toka et al., 2004). IFNAR signaling has been associated with key immunosuppressive features in the context of persistent infection including the increase of IL-10 (Wang et al., 2011) and T cell exhaustion of CD8⁺ and CD4⁺ T cells (other than Treg cells) (Wang et al., 2012). Of note, we did not observe an increase in program death (PD)-1 expression in CD8⁺ and CD4⁺ T cells in the IFNβ-treated animals at 14 days post-transplant (data not shown). Through a series of elegant studies with the clone of 13 of lymphocytic choriomeningitis virus (LCMV) model of persistent infection involving selective blockade of IFNα and IFNβ, the Oldstone group demonstrated that although both IFNα and IFNβ are highly induced during infection, IFNα and IFNβ have segregated functions; while IFNα thwarts the early dissemination of the virus, IFNβ is mainly responsible for promoting viral persistence (Ng et al., 2015). Our results add to these studies and potentially reveal the missing mechanism by which IFNβ controls the persistent phase of infection, through the enhancement of Treg cell induction, as well as a biological reason for the coincidence of type I IFN signatures and increased frequencies of Treg cells in peripheral blood of HCV patients. They also provide an explanation for the more efficacious antiviral effects of IFNα treatment (FDA-approved) when compared to IFNβ in HCV patients (Furusyo et al., 1999; Pérez et al., 1995) (prior to the advent of HCV antivirals) as well as a possible rationale to treat with IFNβ in infections characterized by reduced innate antiviral responses and a vigorous inflammatory response, such as SARS-CoV-1 and MERS (Channappanavar et al., 2017; Kindler et al., 2016) or more recently SARS-CoV-2 (Blanco-Melo et al., 2020; Del Valle et al., 2020), in which promoting Treg cells while preserving the antiviral properties of type I interferons would be desirable.

A significant body of data implicates the type I interferon pathway in the pathogenesis of systemic autoimmune diseases including systemic lupus erythematosus (SLE), myositis, systemic sclerosis, Sjögren’s syndrome, and rheumatoid arthritis (RA) (Hall and Rosen, 2010). A hallmark of these diseases is their self-sustained and auto-amplifying nature, in which type I IFNs might play a critical role as evidenced by: (i) the fact that circulating immune complexes can initiate type I IFN production and dendritic cell (DC) maturation (Båve et al., 2005; Vallin et al., 1999), (ii) studies showing increased IFN-regulated gene expression in patients with systemic autoimmunity (Baechler et al., 2003; Walsh et al., 2007), and (iii) the finding that polymorphisms in type I IFN pathway genes are associated with an increased risk of systemic autoimmune disease (Graham et al., 2006; Graham et al., 2007). Our results provide a likely explanation for the advantages of maintaining high levels of circulating type I interferon in these autoimmune conditions and caution us regarding the potential detrimental effects of a blanket IFNAR blockade therapeutic approach to these diseases (e.g. anifrolumab, currently in clinical trials for lupus nephritis and RA) which might result in the elimination of a critical Treg cell-promoting axis of the immune response.

Our discovery that treatment of transplant recipient mice with IFNβ preferentially increases the number of endogenous Treg cells and prolongs allograft survival in the context of CTLA-4 Ig, despite the well-established pro-inflammatory role of type I IFN in innate immune responses associated with ischemia reperfusion injury (IRI) (Christopher et al., 2002; Oberbarnscheidt et al., 2010; Zhai et al., 2004), suggests that IFNβ has therapeutic potential as adjuvant therapy to improve transplant outcomes. Identification of the Foxp3-acetylation mechanism responsible for the effects of IFNβ on Treg cells could also provide an approach for enhancing the efficacy of autologous Treg cell transfers. The strong antiviral effects of type I interferons such as IFNβ on cytomegalovirus (CMV) (Dağ et al., 2014; Holzki et al., 2015) and human polyomavirus (BKV and JCV) (Delbue et al., 2007; O'Hara and Atwood, 2008; Qin et al., 2016) infections might provide a therapeutic alternative for infected transplant recipients whose immunosuppression needs to be reduced and should be further investigated in transplant models.

The results presented herein provide a molecular mechanism by which type I interferon production in tumor cells promotes a Treg cell-rich microenvironment. In cancer, the role of type I IFN has been generally considered beneficial due to (i) their direct inhibitory effect on tumor cells limiting their proliferation as well as driving senescence and death, and (ii) their role in both promoting T cell responses and preventing metastases (Snell et al., 2017). Type I IFN therapies in cancer treatment have yielded variable outcomes. Treg cells can be coopted by cancer cells to aid in immune suppression and evasion and used to inhibit tumor specific CD8⁺ and CD4⁺ T cell effector functions through cell-to-cell contact and/or production of anti-inflammatory cytokines. Additionally, they can induce effector T cell exhaustion, which is frequently seen in the tumor microenvironment (Stewart et al., 2013). Recent work from the Azzi group demonstrated in a mouse model of multiple myeloma that cancerous cells drive Treg cell expansion and activation by secreting type I interferons. (Kawano et al., 2018). Our studies provide a probable molecular mechanism for this observation and point at IFNβ as the type I IFN whose production is beneficial for the tumor in promoting a Treg cell-rich microenvironment.

In conclusion, while type I interferons induce multiple regulatory responses, understanding which ones are important depending on time and context is critical for our understanding of the immune response. The work presented here describes a previously unrecognized cellular and molecular mechanism relevant to these immune regulatory responses induced by type I IFNs, which establishes an important additional link between innate and adaptive immunity and provides guidance for therapeutic intervention.

LIMITATIONS OF THE STUDY

We acknowledge that the differences between the effects of systemic and local IFNβ remains a key open question regarding the functional connection between IFNβ and Treg cells. We postulate that local IFNβ might have a predominant effect on activating dendritic cells, whereas the direct effect on T cells might be prevalent in the presence of systemic IFNβ. We also recognize that the increase in suppressive function when compared to controls of Foxp3⁺ cells isolated from IFNβ-treated allograft recipients observed in the in vitro suppression assays might not be representative of their immunosuppressive function in the graft. We also understand that some induced Treg cells could also be Nrp1⁺, but the high expression of TF Helios observed in this population in our transplant model supports the conclusion that it is mostly comprised of nTreg cells.

STAR METHODS

RESOURCE AVAILABILITY

Materials availability

We will provide Ifnar1^fl/fl x CD4-Cre^POS and Ifnar1^fl/fl x Foxp3-YFP-Cre^POS mice to requesting labs that have received approval for the use of these mice via an MTA with the Icahn School of Medicine at Mount Sinai.

Data and code availability

• RNA-seq data have been deposited at GEO and are publicly available as of the data of publication. The accession number is listed in the key resources table. Microscopy data reported in this paper will be shared by the lead contact upon request.

• All original code and computational has been deposited either in Matlab Central or in GitLab, respectively. DOIs are listed in the key resources table.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
PacBlue anti-mouse CD8 antibody	BioLegend	Cat# 100725 RRID:AB_493425
PE-Cy7 anti-mouse CD4 antibody antibody	eBiosciences	Cat# 25-0041-82 RRID:AB_469576
PerCP Cy5.5 anti-mouse CD45 antibody	Invitrogen	Cat# 45-0451-82 RRID:AB_1107002
FITC anti-mouse CD3 antibody	Invitrogen	Cat# 11-0031-85 RRID:AB_464883
BV450 anti-human Foxp3	eBiosciences	Cat# 48-4776-42 RRID:AB_1834364
eFluor 450 anti-mouse Foxp3	eBiosciences	Cat# 48-5773-82 RRID:AB_1518812
FITC anti-mouse H-2Kd antibody	eBiosciences	Cat# 11-5998-82 RRID:AB_465358
PerCP Cy5.5 anti-mouse IFN-gamma antibody	BD Biosciences	Cat# 560660 RRID:AB_1727533
PE anti-mouse TNF-alpha antibody	Invitrogen	Cat# 12-7321-82 RRID:AB_466199
eFluor450 anti-mouse CD11b antibody	Themofisher	Cat# 48-0112-82 RRID:AB_1582236
BV510 anti-mouse NK1.1 antibody	BioLegend	Cat# 108737 RRID:AB_2562216
PE-Cy7 anti-mouse NKp46 antibody	BioLegend	Cat# 137617 RRID:AB_11218594
PE-Cy7 anti-mouse Ly6C antibody	Invitrogen	Cat# 25-5931-81 RRID:AB_469662
APC anti-mouse Ly6G antibody	Invitrogen	Cat# 17-5931-82 RRID:AB_469476
Purified rat anti-Mouse IFN-gamma antibody (ELISPOT capture)	BD Biosciences	Cat# 551216 RRID:AB_394094
Biotinylated rat anti-Mouse IFN-gamma antibody (ELISPOT detection 1)	BD Biosciences	Cat# 554410 RRID:AB_395374
Goat anti-Biotin, alkaline phosphatase conjugated antibody (ELISPOT detection 2)	Vector Laboratories	Cat# SP-3020 RRID:AB_2336088
eFluor450 anti-mouse CD44 antibody	eBiosciences	Cat# 48-0441-82 RRID:AB_1272246
PE-Cy7 anti mouse CD62L antibody	BioLegend	Cat# 100434 RRID:AB_893324
BV510 anti-human CD45RA antibody	BD Biosciences	Cat# 563031 RRID:AB_2722499
APC anti-human CD45RO antibody	BD Biosciences	Cat# 559865 RRID:AB_398673
PE anti-mouse Nrp1 antibody	Invitrogen	Cat# 12-3041-82 RRID:AB_2572603
PE-Cy7 anti-mouse GITR antibody	Invitrogen	Cat# 25-5874-82 RRID:AB 10548516
FITC anti-mouse Helios antibody	eBioscience	Cat# 11-9883-82 RRID:AB_11041115
FITC anti-mouse CD25 antibody	BD Biosciences	Cat# 553072 RRID:AB 394604
IFNAR1 blocking antibody mouse	Bio X Cell	Cat# BE0241 RRID:AB_2687723
IFNAR1 blocking antibody human	Abcam	Cat# ab97701 RRID:AB_10679610
IFNAR1 blocking antibody isotype control	Abcam	Cat# ab172730 RRID:AB_2687931
anti-mouse anti-human IFNAR1 (surface expression Imaging Flow) antibody	Abcam	Cat# ab124764 RRID:AB_10972855
anti-mouse anti-human IFNAR1 (surface expression Imaging Flow) antibody control	Abcam	Cat# ab172730 RRID:AB_2687931
Anti-Mouse IFNAR1 - DyLight® 488 antibody	Leinco Technologies	Cat# ab1014 RRID:AB 2830347
Mouse anti-mouse anti-human Foxp3 antibody PLA assay	eBioscience	Cat# 14-7979-80 RRID:AB_468499
Rabbit anti-mouse anti-human acetylated lysines PLA assay	CellSignaling	Cat# 9441 RRID:AB_331805
aCD3 mouse	BioLegend	Cat# 100201 RRID:AB_312658
aCD3 human	BD Biosciences	Cat# 566685
PE anti-human anti-mouse pSTAT1 (pY701) antibody	BD Biosciences	Cat# 612564 RRID:AB_399855
PacBlue anti-human anti-mouse pSTAT3 (pY705) antibody	BD Biosciences	Cat# 560312 RRID:AB_1645327
AF488 anti-human anti-mouse pSTAT4 (pY693) antibody	BD Biosciences	Cat# 558136 RRID:AB_397051
PE-Cy7 anti-human anti-mouse pSTAT5 (pY694) antibody	BD Biosciences	Cat# 560117 RRID:AB_1645546
V450 anti-human anti-mouse pSTAT6 (pY641) antibody	BD Biosciences	Cat# 561203 RRID:AB_10565979
Alexa Fluor 647 anti-mouse Smad2 (pS465/pS467)/Smad3 (pS423/pS425) antibody	BD Biosciences	Cat# 562696 RRID:AB 2716578
Alexa Fluor 647 anti-mouse Smad7 antibody	Santa Cruz Biotechnologies	Cat# sc-365846 RRID:AB 10859551
PerCP Cy5.5 anti-human CD4 antibody	BD Biosciences	Cat# 552838 RRID:AB_394488
APC anti-human CD25 antibody	eBiosciences	Cat# 17-0259-42 RRID:AB_1582219
PE anti-human CD127 antibody	BD Biosciences	Cat# 557938 RRID:AB_2296056
PE anti-mouse CD127 antibody	eBiosciences	Cat# 12-1271-82 RRID:AB_465844
eFluor647 anti-mouse TIP60 antibody	Santa Cruz Biotechnologies	Cat# sc-166323 RRID:AB_2296327
PE anti-human IL2 antibody	Invitrogen	Cat# 12-7029-42 RRID:AB 30300581
FITC anti-human CD4 antibody	eBiosciences	Cat# 11-0042-86 RRID:AB_464898
Chemical, Peptides and Recombinant Proteins
Mouse Interferon-Beta Mammalian	PBL Biomedical	Cat# 12405-1
Mouse Interferon-Alpha4 Mammalian	PBL Biomedical	Cat# 12115-1
Recombinant Interferon-Alpha2	PBL Biomedical	Cat# 110591
InVivo MAb recombinant (hum/hum) CTLA-4 Ig	Bio X Cell	Cat# BE0099 RRID: AB_10949064
Recombinant murine IL-2	Peprotech	Cat# 212-12
Recombinant TGFβ1	Peprotech	Cat# 100-21C
Recombinant human IL-2	BD Pharmingen	Cat# 554603
Human IFN Beta (Beta 1a, Mammalian Expressed)	PBL Biomedical	Cat# 11415-1
Diphtheria toxin	SigmaAldrich	Cat# D0564-1MG
PMA	SigmaAldrich	Cat# P-8139
Ionomycin	SigmaAldrich	Cat# I-0634
1-step NBT/BCIP Substrate	Thermofisher	Cat# 34042
CellTrace CFSE Cell Proliferation	Thermofisher	Cat# C34554
CellTrace Far Red Cell Proliferation	Thermofisher	Cat# C45571
Golgi plug	BD Biosciences	Cat# 5550291
C646	SigmaAldrich	Cat# SML0002
Collagenase type IV from Clostridium histolyticum	SigmaAldrich	Cat# C5138
eFluor 780 Fixable viability dye	eBioscience	Cat# 65-0865-14
ACK lysis buffer	Roche	Cat# 11814389001
AffinityScript MultiTemp RT	Agilent	Cat# 600105
PlatinumTaq DNA	Thermofisher	Cat# 14966001
SYBR-green DNA-binding dye	Thermofisher	Cat# S33102
Histopaque	SigmaAldrich	Cat# 10771
Critical Commercial Assays
EasySep™ Mouse Naïve CD4+ T Cell Isolation Kit	STEMCELL Technologies	Cat# 19765
EasySep™ Human Naïve CD4+ T Cell Isolation Kit	STEMCELL Technologies	Cat# 19555
EasySep™ Mouse T Cell Isolation Kit	STEMCELL Technologies	Cat# 19851
EasySep™ Mouse CD90.2 Positive Selection Kit II	STEMCELL Technologies	Cat# 18951
Zenon™ Alexa Fluor™ 488 Rabbit IgG Labeling Kit	Thermofisher	Cat# Z25302
aCD3/aCD28 stimulating beads murine	Gibco	Cat# 11-161D
aCD3/aCD28 stimulating beads human	Gibco	Cat# 11-456D
Duolink™ In Situ PLA® Probe Anti-Mouse PLUS	Sigma Aldrich	Cat# DUO92001
Duolink™ In Situ PLA® Probe Anti-Rabbit MINUS	Sigma Aldrich	Cat# DUO92005
Duolink™ In Situ Detection Reagents Red	Sigma Aldrich	Cat# DUO92008
Duolink™ In Situ Mounting Medium with DAPI	Sigma Aldrich	Cat# DUO82040
RNeasy Minikit	QIAGEN	Cat# 74104
Intracellular/transcription factor staining buffer kit	eBiosciences	Cat# 00-5523-00
Deposited Data
RNAseq data	This manuscript	NCBI Gene Expression Omnibus (GEO) accession # GSE 191154
Experimental Models: Organisms/Strains
C57BL/6J (B6)	The Jackson Laboratory	Stock No: 000664
BALB/cJ (BALB/c)	The Jackson Laboratory	Stock No: 000651
B6(Cg)-Ifnar1tm1.1Ees/J (Ifnar1fl/fl)	The Jackson Laboratory	Stock No: 028256
B6.Cg-Tg(Cd4-cre)1Cwi/BfluJ (Cd4-Cre)	The Jackson Laboratory	Stock No: 022071
C57BL/6-Tg(Foxp3-DTR/EGFP)23.2Spar/Mmjax (DTR-Foxp3-GFP)	The Jackson Laboratory	Stock No: 032050
B6.129(Cg)-Foxp3tm4(YFP/icre)Ayr/J (Foxp3-YFP-Cre) (Foxp3-YFP-Cre)	The Jackson Laboratory	Stock No. 016959
C57BL/6-Tg(Foxp3-GFP)90Pkraj/J (Foxp3-GFP)	The Jackson Laboratory	Stock No: 023800
B6.129S7-Rag1tm1Mom/J (Rag1−/−)	The Jackson Laboratory	Stock No: 002216
B6.129S(Cg)-Stat1tm1Dlv/J (Stat1−/−)	The Jackson Laboratory	Stock No: 012606
C.129S2-Stat4tm1Gru/J (Stat4−/−)	The Jackson Laboratory	Stock No: 002826
Ifnar1fl/fl x Cd4-CrePOS (bred from Ifnar1fl/fl and Cd4-Cre)	Fribourg Lab (this paper)	N/A
Ifnar1fl/fl x Cd4-CreNEG (bred from Ifnar1fl/fl and Cd4-Cre)	Fribourg Lab (this paper)	N/A
Ifnar1fl/fl x Foxp3-YFP-CrePOS (bred from Ifnar1fl/fl and Foxp3-YFP-Cre)	Fribourg Lab (this paper)	N/A
Ifnar1fl/fl x Cd4-CreNEG (bred from Ifnar1fl/fl and Foxp3-YFP-Cre)	Fribourg Lab (this paper)	N/A
Software and Algorithms
BIOMD0000000451_url.xml (CD4+ naïve polarization model)	(Carbo et al., 2013)	https://www.ebi.ac.uk/biomodels/BIOMD0000000451
Optimized models for Th1, Th2, Th17 and Treg polarization	Fribourg Lab (this manuscript)	doi:10.5281/zenodo.5794186
FCS Express 7	De Novo Software	N/A
IDEAS	Amnis Corporation/Millipore	N/A
SpotCountPLA Matlab Toolbox	Fribourg Lab (this manuscript)	doi:10.5281/zenodo.5794212
Matlab	Mathworks	N/A
COPASI: Biochemical System Simulator	copasi.org	N/A
Rstudio	Rstudio.com	N/A

EXPERIMENTAL MODEL

Mice

Male and female C57BL/6 (H-2K^b) and BALB/c (H-2K^d) were purchased from The Jackson Laboratory (stocks #000664 and #000651, respectively) and bred at mouse facilities of the Icahn School of Medicine at Mount Sinai. Ifnar1^fl/fl and Cd4-Cre mice were purchased from The Jackson Laboratory (stocks # 028256 and # 022071, respectively) and crossed to obtain Ifnar1^fl/flxCd4-Cre^POS and Ifnar1^fl/flxCd4-Cre^NEG used in Figure 2, Figure 3 and Figure 6. Ifnar1^fl/fl mice were also crossed with Foxp3-YFP-Cre mice purchased from The Jackson Laboratory (stock # 023800) and crossed to obtain Ifnar1^fl/flxFoxp3-YFP-Cre^POS and Ifnar1^fl/flxFoxp3-YFP-Cre^NEG used in Figure 4 and Figure 6. Foxp3-GFP and Rag1^−/− mice (Fig. 2), DTR-Foxp3-GFP (Fig. S4) mice, and Stat1^−/− and Stat4^−/− mice (Fig. 6) mice were purchased from the Jackson Laboratory (stocks #023800, #002216, #032050, #012606, and #002826, respectively). Foxp3-YFP mice (stock # 023800) were used for the transcriptional analyses in Figure S7. Mice were aged 8 weeks at the start of the experiment. Data depicted in the figures include male and female mice as we used both male and female mice as transplant donors and recipients, for adoptive transfer experiments, and for Treg cell induction experiments. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals and under the protocol (IACUC-2018-2) approved by the Institutional Animal Care and Use Committee of the Icahn School of Medicine at Mount Sinai. All surgery was performed under general anesthesia, and post-transplant analgesia was provided. All animals were housed in the animal facilities at the Icahn School of Medicine at Mount Sinai.

Human subjects

For experiments in human cells (Fig.7 and Fig. S7) we used peripheral blood mononuclear cells (PBMCs) from buffy coats obtained from anonymous donors to the New York Blood Bank, whose sex and age (all older than 18 years old) is unknown to the investigators. This work was deemed not Human Subject Research by the Icahn School of Medicine’s Institutional Review Board (IRB). The number of donors used for every experiment is indicated in the figure legends, and data points for each donor are indicated in the bar graphs except for Figure S7E for clarity.

METHOD DETAILS

Heterotopic heart transplants

Donor and recipient animals were anesthetized with ketamine/xylazine. Donor hearts were arrested by infusion of 0.5 ml of cold University of Wisconsin (UW) solution into the inferior vena cava. The chest was then opened and after ligation of the aortic arch proximal to the right subclavian artery the aortic root was flushed. The heart was removed from the chest and placed in chilled (4°C) sterile UW solution until transplanted. The recipient’s infra-renal abdominal aorta and inferior vena cava were isolated and obstructed both proximally and distally with two microvascular clamps. Incisions were made on both vessels. The graft was rinsed with cold lactated Ringer’s solution. The donor aorta and pulmonary artery were joined end-to-side to the recipient aorta and vena cava, respectively, and after completion of the anastomosis, the proximal and distal clamps released. When the graft began beating the abdomen is closed. Recipients were treated with either CTLA-4 Ig (Bio X Cell, 160 μg i.p. on day 2 post-transplant), interferon beta (PBL Biomedical, 10,000 U/ml i.p. on days −1,1,2,3 and 4 post-transplant and then weekly), diphtheria toxin (Sigma Aldrich, 10 μg/kg i.p. on days 13,15,17,20,22,24, and 27 post-transplant), or C646 (Sigma Aldrich, 5 mg/kg on days −1,1,2,3 and 4 post-transplant and then weekly). Graft survival was monitored twice a week through palpation with a scoring system of 0 to 2 by a researcher blinded to the experiment. Rejection was defined as the day on which a palpable heartbeat was no longer detectable (score 0) and was confirmed by direct visualization at laparotomy.

Lymphocyte isolation from the spleen and graft.

Spleens were harvested in PBS and mechanically disaggregated through a mesh strainer (70 μm) with the aid of the back of a syringe plunger. PBS was decanted after centrifugation and red blood cells lysed (ACK lysis buffer, 4 min at room temperature, Roche). Cells were then resuspended in PBS and filtered again through a 70 μm nylon mesh. The graft heart tissue was recovered, minced and washed three times with PBS/EDTA, followed by collagenase digestion (0.2 mg/mL collagenase type IV Sigma Aldrich) diluted in RPMI supplemented with 10% fetal bovine serum (FBS) for 2.5 h at 37 °C. The digested tissue was then dissociated through a 19 g needle and filtered through a 70 μm nylon mesh. The isolated single-cell suspension of enriched either spleen or tissue-derived leukocytes was processed for different assays, or used to isolate naïve CD4⁺T cells naïve T cells using magnetic separation (EasySep™ Mouse Naïve CD4+ T Cell Isolation Kit, STEMCELL Kit) for Treg cell induction cultures.

RT-PCR

RNA was extracted from grafts or mouse and human culture cells (RNeasy Plus Minikit, QIAGEN). cDNA was synthesized from total RNA using AffinityScript MultiTemp RT (Agilent) with an oligo(dT)18 primer. Real-time PCR was performed using PlatinumTaq DNA polymerase (Thermofisher) and SYBR-green (Thermofisher) on an ABI7900HT thermal cycler (Applied Biosystems). A robust global normalization algorithm, using expression of the housekeeping genes ribosomal protein S11 (Rps11), β-actin (Actb), and α-tubulin (Tuba), was used for all experiments, as described elsewhere (Bordería et al., 2008). In brief, all crossing threshold (Ct) values were first adjusted by median difference of all samples from Actb. Each individual sample was then further corrected by the median Ct value of the three corrected housekeeping controls for that sample. Nominal copy numbers were calculated by assuming 2500 molecules of Actb mRNA per cell, and an amplification efficiency of 93% using the difference in Ct value (ΔCt method).

Flow cytometry, transcription factor and KAT staining, and intracellular cytokine staining.

Cells were evaluated for surface antigen expression following incubation with fluorescently conjugated antibodies in PBS or a buffer consisting of 2% rat serum 2 mM EDTA according to the manufacturer’s instructions. For Foxp3, P300, and TIP60 staining, cells were permeabilized using an intracellular/transcription factor staining kit (eBiosciences) and stained with Foxp3, P300, and TIP60. For cytokine staining 100,000 mouse splenocytes or cells at the end of human Treg cell induction culture were stimulated ex vivo with 100,000 BALB/c splenocytes for 24h, adding a Golgi Plug (BD Biosciences) for the last 4 hours. Cells were then processed and stained using an intracellular staining kit (eBiosciences) according to manufacturer’s instructions with TNFα, IFNγ, or IL-2 antibodies. For phospho-STAT activation staining, cells were stimulated and fixed after 40 min post-stimulation (1.6% paraformaldehyde, 10 min at 37°C), stained for surface CD4, permeabilized with cold methanol (10 min 4°C) and then stained with pSTAT1, pSTAT3, pSTAT5, pSTAT5, and pSTAT6 antibodies. All samples were collected using a FACS Canto II flow cytometer (BD Biosciences) and analyzed using FCS Express 7 (De Novo Software).

ELISpot Assays

Ninety-six well ELISPOT plates were coated with capture antibody for IFNγ (1 μg/mL, BD Biosciences) in PBS at 4°C overnight. The plates were then blocked with PBS/0.1% BSA and washed with PBS. Three hundred thousand stimulating BALB/c splenocytes were added to each well in 100 ul of complete RPMI medium, together with multiple dilutions of splenocytes from transplanted mice (B6) with a maximum starting cell count of five hundred thousand cells. Control wells contained responder cells plus medium alone or cells stimulated with PMA (5ng/ml) ionomycin (100ng/ml) (Sigma Aldrich). After 24h, the plates were washed and biotinylated detection IFNγ antibody (2μg/ml, BD Biosciences) added to the wells overnight at 4°C. After washing, an alkaline phosphatase-conjugated anti-biotin antibody (Vector Laboratories) diluted 1:2000 in PBS supplemented with 0.1% Tween and 1% bovine serum albumin (BSA) was added for 90 min, the plates were developed by addition of 1-Step NBT/BCIP substrate (Thermofisher), and the resulting spots were counted on an ImmunoSpot Series 3 Analyzer (Cellular Technology Ltd).

Imaging Flow Analysis

Anti-murine and anti-human IFNAR1 and its corresponding isotype control were conjugated to AF488 using the Zenon Labeling kit (Thermofisher) following the manufacturer’s instructions. Samples were fixed (1.6% paraformaldehyde 10 min 37°C) to avoid receptor internalization, and surfaces stained at 1:3 dilution from the obtained conjugated products product. Following staining, images of each cell were acquired at a magnification of 60x on an ImageStream flow cytometer (Amnis) and analyzed with the IDEAS analysis software. Single-color controls were used for the creation of a compensation matrix that was applied to all files to correct for spectral crosstalk.

In vitro Treg cell induction

Human peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats by Ficoll density gradient centrifugation (Histopaque, SigmaAldrich) at 490g. Naïve CD4⁺ T cells were enriched from murine splenocytes (CD44^loCD62^hi) or human peripheral blood mononuclear cells (CD45RA⁺CD45RO⁻) (EasySep™ Mouse Naïve CD4⁺ T Cell Isolation Kit, EasySep™ Human Naïve CD4⁺ T Cell Isolation Kit, respectively, STEMCELL Technologies) and their purity checked by flow cytometry. Mouse cultures: 200,000 naïve CD4⁺ T cells were incubated with IL-2 (2.75 ng/ml, Peprotech), TGFβ (0.7 ng/ml, Peprotech) ± IFNβ (100 U/ml or 1,000 U/ml, PBL Biosciences) or IFNα2 (1000 U/ml) or IFNα4 (1000 U/ml and stimulated with either αCD3/αCD28 (15 μl/million cells, Gibco) (polyclonal), or 200,000 BALB/c APCs (EasySep™ Mouse CD90.2 Positive Selection Kit II, STEMCELL Technologies) cells previously incubated with anti-IFNAR1 antibody for 1 h (10ug/ml, Bio X Cell) (allostimulation). Human cultures: 200,000 naïve CD4⁺ T cells were, in certain cases previously incubated with anti-IFNAR1 antibody or isotype for 30 min (1:1000 dilution, Abcam), and then cultured with IL-2 (100 U/ml, BD Pharmingen), TGFβ (3 ng/ml, Peprotech) ± IFNβ (100 U/ml, or 1000 U/ml, PBL Biosciences) and stimulated with aCD3/aCD28 (15 μl/million cells, Gibco).

Treg cell suppression assays

Mice. 200,000 conventional B6 T cells (EASYSEP) labeled with a cell tracker (CFSE, or CellTrace Far Red) were stimulated by 50,000 BALB/c APCs (STEM cell) + αCD3 (1ng/ml, Biolegend) in the presence of different amounts of induced Treg, either with APCs (allostimulus) or with aCD3/aCD28 (polyclonal), previously sorted as Foxp3-GFP⁺ (Foxp3-GFP mice) or CD4⁺CD25^hiCD127⁻ (wild-type). Human. 200,000 conventional T cells labeled with CFSE were stimulated with aCD3 (1ng/ml, BD Biosciences) in the presence of different amounts of polyclonally induced Treg cells, previously sorted as CD4⁺CD25^HICD127⁻, or CD4⁺CD25^LOCD127- as controls. Both for mouse and human suppression assays, the percentage of proliferating cells was determined by flow cytometry based on cell tracker dilution 5 days after stimulation. The suppression was calculated as (Proliferation without Treg cells - Proliferation with Treg cells)/Proliferation without Treg cells.

In vivo Treg cell induction

Naïve CD4⁺ T cells (CD44^loCD62^hi) were enriched from murine splenocytes of Foxp3-GFP and sorted for Foxp3-GFP⁻ cells. Each Rag1^−/− animal received an adoptive transfer of 20,000 Foxp3-GFP⁻ naïve cells through retroorbital injection. Mice were then treated with PBS or IFNβ on days 0,1,2,3, 4, 11 and 14 post-transfer. On day 14 the spleens of the animals were harvested and the frequency and number of Foxp3⁺ cells assessed by flow cytometry.

Proximity ligation assay (PLA) for acetylated Foxp3.

Mouse Treg cells from induction cultures, human Treg cells from induction cultures, or CD4⁺ enriched cells splenocytes from mice were harvested and cultured for 2 hours at 37°C in 50-100 ul drops containing 50,000 cells on top of coverslips previously coated with poly-D-lysine and laminin. Coverslips were then washed and the cells attached to it fixed by immersion in 3.7% PFA for 10 min at room temperature, and permeabilized by immersion in 70% ethanol for 20 min. The proximity ligation assay was then performed following the manufacturer’s recommendations and the particularizations described in the protocol from Hancock and Beier (Jiao et al., 2017) using primary antibodies against Foxp3 (raised in mouse, eBiosciences) and acetylated lysines (raised in rabbit, Cell Signaling), Duolink™ In Situ PLA® Probe Anti-Mouse PLUS Reagents (Sigma Aldrich) together with Anti-Rabbit MINUS (Sigma Aldrich) and Duolink™ In Situ Detection Reagents Red. Finally, the coverslips were mounted on slides using a DAPI-containing mounting medium.

PLA image acquisition

Cells were imaged with and upright epifluorescence microscope, Axioplan 2 (Zeiss), with a 100x oil immersion objective and a numerical aperture of 1.3. The microscope was controlled by AxioVision software and was equipped with a Zeiss AxioCam MRm camera. Images were acquired in brightfield. The laser exposure time for the red PLA probes was adjusted for every experiment and ranged between 1 and 2 seconds. 20-25 fields were obtained from each coverslip with 2-10 per field. Multiple coverslip per mouse and per treatment were analyzed per experiment. Control experiments were performed with only one primary antibody (either Foxp3 or acetylated lysines) to establish background noise and specificity of the signal.

PLA image analysis

The nuclear image from the blue channel (DAPI) and the spot image from the red channel (PLA) are processed separately. Briefly, both the nuclear image and the spot image are (i) normalized to maximum intensity, (ii) contrast-enhanced using adaptive histogram equalization, (iii) and a threshold applied to them to eliminate background. To segment the cells, we utilized an extended-maxima transform followed by a watershed transform to define a region of influence for each cell. The spot image was sharpened using the unsharp masking technique, and the spots identified by labeling connected components in a binary version of the image. The images with the regions of influence and with the spots were overlaid and each spot assigned to the corresponding cell provided it was at a distance shorter than 150 pixels from the nucleus of the cell. To establish fair comparisons of Foxp3 acetylation between treatments, cells with zero spots were excluded from each group reasoning that they were not Foxp3⁺. In Treg cell induction experiments, at least 40% of the cells had at least one spot. In CD4⁺ enriched cells from mouse spleens

Treg cell polarization computational model

The computational model developed for this work constitutes an expansion of a publicly available (https://www.ebi.ac.uk/biomodels/BIOMD0000000451) ordinary differential equation (ODE) model developed at the Virginia Bioinformatics Institute (Carbo et al., 2013). The model includes 60 differential equations representing 52 reactions and 93 species and was implemented and simulated in COPASI (Hoops et al., 2006). To expand the model to include the effects of IFNβ on T cell polarization we introduced two additional species: the concentration of IFNAR (type I interferon receptor in the membrane), and the concentration of IFNβ as one of the cytokines in the environment as a new input to the cell. Full details on how the model was expanded to include and model calibration can be found in Data S1.

RNA-Seq

RNA from CD4⁺Foxp3-YFP⁺ T cells, sorted from the spleen of BALB/c transplant recipients at day 14 post-transplant was isolated using the RNeasy Plus Micro kit (QIAGEN). Quality control was performed by bioanalyzer (Agilent), and RNA samples with an RNA integrity number (RIN) > 8 were processed for library preparation using SMARTer Ultra Low Input reagent (Takara) and Nextera XT DNA (Illumina) library preparation kits. Libraries were sequenced with paired-end reads of 126bp on a HiSeq2500 sequencer (Illumina) to reach 50 million read pairs per sample. For each group, IFNβ or vehicle, and genotype, data are derived from three distinct mice, with separate processing from cell sorting to sequencing. Raw RNaseq fastq reads were trimmed with bbtools and aligned to mouse genome (mm10) using STAR (v. 2.4.0h). Gene-assignment and count of RNA reads were performed with HTseq. Further analyses were performed with R software. Differentially expressed genes were identified using the R package DESeq2 (v. 1.22.1) using the Wald test (FDR < 0.01). Gene set enrichment analyses were performed using the R package FGSEA (v.1.8.0). Gene Ontology enrichment analyses were performed using the R packages Go.db (v.3.7.0) and GOstats (2.48.0).

QUANTIFICATION AND STATISTICAL ANALYSIS

Graft survival data were plotted on Kaplan-Meier curves and compared with a log-rank (Mantel-Cox) test. Comparisons in RT-PCR gene expression experiments were performed with a t-test with a Bonferroni correction for multiple comparisons. Foxp3 acetylation (spot count, no normal distribution of the data assumed) were compared between groups using a Mann Whitney test (one group) or Kruskal-Wallis test followed by Dunn’s multiple comparison test (multiple groups). Suppression data were fitted to a one-site total binding curve followed by an extra-sum-of-squares F test to determine if the curves are different. For the rest of the comparisons we applied unpaired t-tests (two groups) or one-way ANOVA followed by Tukey’s post-hoc HSD test (multiple groups). All analyses were done using Prism (GraphPad Software). In all legends and figures, mean ± S.E.M. is shown, and * p<0.05, **p<0.01,*** p<0.001 and n.s., non-significant.

Supplementary Material

1

Supplementary Data S1. Computational Model Details

HIGHLIGHTS.

• IFNβ synergizes with CTLA-4 Ig to prolong allograft survival by promoting Treg cells

• IFNβ directly acts on T cells to enhance Treg cell induction and increase Foxp3 protein expression

• IFNβ activates STAT1 signaling to promote P300 expression and increase Foxp3 acetylation