Cutting Edge: STAT4 Promotes Bhlhe40 Induction to Drive Protective IFN-γ from NK Cells during Viral Infection

Natural killer cells are innate lymphocytes able to respond to stress, transformed, or infected cells without prior sensitization. However, resting NK cells can become more potent effector cells through activation by germline-encoded surface receptors and proinflammatory cytokines, including IL-12, IL-18, and type I IFN produced during viral infection (1). These activating signals will induce a robust transcriptional and epigenetic response in NK cells, leading to their effector programs (2), including the production of the potent antiviral cytokine IFN-γ.

During mouse CMV (MCMV) infection, IL-12 is produced and binds to a heterodimeric receptor, leading to a signal cascade via JAK-mediated phosphorylation and homodimerization of STAT4 in the cytoplasm. In NK cells, STAT4 activation causes it to translocate to the nucleus, where it induces the transcription of a multitude of genes by targeting both promoter and nonpromoter regions (3). Previous studies have demonstrated that an IL-12/STAT4 signaling axis in NK cells can drive a plethora of effector functions, including cytokine production and proliferation (4) through transcription factors Zbtb32, IRF8, STAT5, CBFb, and Runx factors (5–8), among others. However, many additional genes targeted by STAT4 in NK cells during MCMV infection have not been characterized.

In our present study, we discovered that Bhlhe40 was highly induced in activated NK cells during MCMV infection among the family of basic helix-loop-helix (bHLH) transcription factors. Bhlhe40, also known as Bhlhb2, Dec1, and Stra13, consists of a basic DNA-binding domain; a helix-loop-helix domain, which facilitates dimerization; and a protein–protein interaction Orange domain (9, 10). Members of the bHLH family (specifically factors in the E clade, including Bhlhe40) have been shown to regulate numerous physiological processes, including differentiation, proliferation, and apoptosis in multiple cell types due to its ability to act as both transcriptional activator and repressor in a context-dependent manner (11). Although Bhlhe40 was initially described as a regulator of circadian rhythm and cellular differentiation, emerging evidence suggests that it is also acting as a regulator of the innate and adaptive immune systems (11).

In T cells, Bhlhe40 is rapidly expressed following TCR activation (12) and plays a number of critical roles to support T cell identity and effector functions. In the context of autoimmunity and aging, Bhlhe40 promotes GM-CSF expression while suppressing IL-10 production (13) and maintains regulatory T cell function/identity, respectively (14, 15). Additionally, Bhlhe40 is also required for the function of tissue-resident memory CD8⁺ and CD4⁺ T cells and tumor-infiltrating lymphocytes by maintaining their mitochondrial fitness and epigenetic states (16, 17). More recently, Bhlhe40 was shown to be a cell-intrinsic negative regulator of both activated B cells and follicular helper T cells in germinal center responses (18). Within the innate immune system, Bhlhe40 controls the self-renewal and maintenance of alveolar and large peritoneal macrophages and their proliferation in response to IL-4 (19, 20). In the present study, we investigated mechanisms of Bhlhe40 induction and consequences of its deletion on NK cell activation and effector function in response to viral infection.

Mice were bred at Memorial Sloan Kettering Cancer Center in accordance with the guidelines of the institutional animal care and use committee. The following strains were used in this study: C57BL/6 (CD45.2), B6.SJL (CD45.1), Stat4^−/−, Il12rb2^−/−, Il18r1^−/−, Bhlhe40^−/−, and Klra8^−/− (Ly49h^−/−). Experiments were conducted using age- and sex-matched mice in accordance with approved institutional protocols. Adoptive transfer studies and generation of mixed bone marrow chimeric mice were performed as previously described (21).

MCMV (Smith) was passaged through BALB/c mice three times, followed by preparation of salivary gland viral stocks using a homogenizer to dissociate the salivary glands (of BALB/c mice 3 wk postinfection [PI]). MCMV stocks were diluted prior to injection, and 10³ PFUs were delivered in adoptive transfer studies and 10⁴ PFUs in all other experiments.

Single-cell suspensions were prepared from the indicated organs, and the indicated fluorophore-conjugated Abs (BioLegend, Tonbo, eBioscience) were used to stain lymphocytes. Flow cytometry was performed on a BD Biosciences LSR II or Cytek Biosciences Aurora, and data were analyzed with FlowJo software.

NK cells were enriched ex vivo using a no-touch bead enrichment protocol as previously described (21) and plated at 10⁵ per well with mouse IL-2, IL-12, or IL-15 (R&D Systems) at 20 ng/ml; with IL-18 (MBL) at 10 ng/ml; or with IFN-α (R&D Systems) at 100 IU. Intracellular staining for IFN-γ was performed as previously described (21).

The following datasets and analyses were performed on NK cells as previously described by our laboratory and can be found in the GEO database as a SuperSeries under accession numbers GSE106139 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE106139) and GSE140044 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE140044): H3K4me3 chromatin immunoprecipitation sequencing (ChIP-seq) and assay for transposase-accessible chromatin with sequencing (ATAC-seq) following cytokine stimulation; RNA sequencing (RNA-seq) from MCMV-infected Il12rb2^−/−, Stat4^−/− mice; RNA-seq on days 0, 2, 4, 7, and 35 after MCMV infection; and RNA-seq on 3-h and overnight in vitro cytokine stimulations (2, 3, 6, 22, 23).

For new datasets generated in this study, RNA-seq was performed as previously described (2). Differential analysis was executed with DESeq2 (version 1.22.2). For all differential analyses, genes were considered differential if they showed a false discovery rate-adjusted p value (FDR) ≤0.05 by contrast using the DESeq2 model. For all sequencing counts shown, values were normalized using sizeFactors calculated by the DESeq2 software or Saccharomyces cerevisiae spike-in DNA. Paired analysis was performed between wild-type (WT) versus Bhlhe40^−/− NK cells for day 0 versus day 2 PI, where calculations for generalized linear models were fitted with a design that controlled for each group. RNA-seq data comparing WT with Bhlhe40^−/− NK cells before and after MCMV infection are available under GEO accession number GSE235608 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE235608) and RNA-seq data comparing WT with Il18r1^−/− NK cells before and after MCMV infection are available under GEO accession number GSE244300 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE244300).

For the Stat4 and Bhlhe40 cleavage under targets and release using nuclease (CUT&RUN), 500,000 sorted NK cells were washed three times with PBS, resuspended in buffer 1 (1× eBioscience Perm/Wash Buffer, 1× Roche cOmplete EDTA-free protease inhibitor, 0.5 μM spermidine in H₂O) and incubated with BHLHE40 (anti-DEC1; Novus, NB100-1800) or STAT4 (Invitrogen, 71-4500) Abs at 1:100 dilution in Ab buffer (buffer 1 + 2 μM EDTA) in a 96-well V-bottomed plate at 4°C overnight. Upon Ab incubation, cells were washed twice with buffer 1 and resuspended in 50 μl of buffer 1 + 1× pA/G-MNase (Cell Signaling Technology, 57813) and incubated on ice for 1 h and washed with buffer 2 (0.05% w/v saponin, 1× Roche cOmplete EDTA-free protease inhibitor, 0.5 μM spermidine in 1× PBS) three times. After washing, calcium buffer (buffer 2 + 2 μM CaCl₂) was used to resuspend the cells for 30 min on ice to activate the pA/G-MNase reaction, and an equal volume of 2× STOP Buffer (buffer 2 + 20 μM EDTA + 4 μM EGTA) was added along with 1 pg Saccharomyces cerevisiae spike-in DNA (Cell Signaling Technology, 29987). Samples were incubated for 15 min at 37°C, and DNA was isolated and purified using the Qiagen MinElute Kit according to the manufacturer’s protocol and subjected to library amplification.

Immunoprecipitated DNA was quantified by PicoGreen, and the size was evaluated using an Agilent BioAnalyzer. When possible, fragments between 100 and 600 bp were size selected using aMPure XP beads (Beckman Coulter, A63882), and Illumina sequencing libraries were prepared using the KAPA HTP Library Preparation Kit (Kapa Biosystems, KK8234) according to the manufacturer’s instructions with 0.001–0.5 ng input DNA and 14 cycles of PCR. Barcoded libraries were run on the NovaSeq 6000 in a PE100 run using S4 kit version 1.5 with XP mode (Bhlhe40 CUT&RUN) or NovaSeq X in a PE100 run using the 10B reagent kit (200 cycles; Illumina, STAT4 CUT&RUN). An average of 23 million paired reads were generated per sample. CUT&RUN data are available under GEO accession numbers GSE242900 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE242900) and GSE242911 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE242911).

Paired reads were trimmed for adaptors and removal of low-quality reads by using Trimmomatic (version 0.39) and aligned to the mm10 reference genome using Bowtie 2 (version 2.4.1). Upon alignment, peaks were called using MACS2 (version 2.2.7.1) with input samples as a control using narrow peak parameters --cutoff-analysis -p 1e-5 --keep-dup all -B --SPMR. For each sample group, irreproducible discovery rate calculations were performed using scripts provided by the ENCODE project (https://www.encodeproject.org/software/idr/; version 2.0.4.2). Reproducible peaks showing an irreproducible discovery rate value of 0.05 or less in each condition were aggregated across the experiment and merged via union to the peak atlas, annotated with the UCSC Known Gene model. The peak atlas consisted of 316 peaks for BHLHE40 CUT&RUN and 2654 for STAT4 CUT&RUN.

Data are shown as mean ± SEM in all graphs, and statistical differences, and unless noted otherwise, were calculated using a two-tailed unpaired Student t test. A p value ≤0.05 was considered significant. In these figures, p values are denoted as *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, and ****p ≤ 0.0001. For data showing a summary of normalized counts from RNA-seq and differential analysis from various sequencing experiments, statistics were calculated using the DESeq2 method as noted above. For these data, FDR adjusted p values ≤0.05 were considered statistically significant. In these figures, *p ≤ 0.05. All statistical analyses and plots were produced in GraphPad Prism or R.

bHLH transcription factors are widely expressed in different immune cells (11). To assess the regulation of bHLH transcription factors in NK cells during MCMV infection, we performed comparative transcriptomic analysis by RNA-seq during their immune response. At day 2 PI, a time point when NK cells are robustly exhibiting their innate immune functions (1), the Bhlhe40 transcript was the most significantly upregulated among the E family bHLH transcription factors in NK cells (Fig. 1A, 1B).

Given that homeostatic and proinflammatory cytokines (IL-2, IL-12, IL-15, IL-18, or type I IFN) and their downstream STAT molecules (STAT1, STAT4, or STAT5) are strongly signaling in NK cells at day 2 PI (3), we analyzed Bhlhe40 expression in NK cells stimulated with various cytokines for 3 h or overnight. We found that NK cells stimulated with IL-12 + IL-18, but not other cytokines, showed the greatest induction of Bhlhe40 (Fig. 1C, 1D). Consistent with this observation, the induction of Bhlhe40 was largely or partially ablated in NK cells deficient for the IL-12 receptor or IL-18 receptor, respectively, during MCMV infection (Fig. 1E, 1F). Together, these data suggest that Bhlhe40 induction in NK cells early after viral infection is dependent mainly on IL-12 and to a lesser extent on IL-18 signals.

Given that IL-12 signals via the transcription factor STAT4, we investigated whether STAT4 controls transcription of Bhlhe40 in NK cells during MCMV infection. Indeed, STAT4 deficiency resulted in a significant loss of Bhlhe40 induction (Fig. 2A), regardless of activating receptor Ly49H expression, suggesting that Bhlhe40 is critically dependent on STAT4 in activated NK cells. To determine whether STAT4 directly targets the Bhlhe40 loci, we performed STAT4 CUT&RUN on NK cells stimulated with IL-12 + IL-18 (2). Significantly increased differential STAT4 binding was detected at the promoter of Bhlhe40 in stimulated NK cells when contrasted with unstimulated or cytokine-stimulated Stat4^−/− NK cells (log₂ fold changes of 3.17 and 2.26 and FDRs of 8.95e-40 and 1.74e-17, respectively) (Fig. 2B). Although we did not observe large changes in chromatin accessibility (by ATAC-seq) at the promoter region of Bhlhe40 in NK cells stimulated with IL-12 + IL-18 (Fig. 2C), we observed a significant increase in the permissive histone mark H3K4me3 following stimulation (Fig. 2D). Furthermore, we determined that the increase in H3K4me3 at the Bhlhe40 locus in stimulated NK cells is dependent on an IL-12/STAT4 signaling axis (Fig. 2E). Altogether, we provide evidence that STAT4 regulates the expression of Bhlhe40 in activated NK cells during MCMV infection via deposition of the permissive histone mark H3K4me3.

Because Bhlhe40 is strongly induced in NK cells early after viral infection, we investigated whether Bhlhe40 may be driving the specific NK cell effector functions of cytokine secretion, killing, and proliferation. To test the latter using a well-established adoptive transfer model, we injected an equal number of WT (CD45.1) and Bhlhe40^−/− (CD45.2) NK cells into Ly49H-deficient recipient mice infected with MCMV and tracked the Ly49H⁺ NK cell responses in the recipient. No detectable difference was observed in the magnitude of NK cell expansion following MCMV infection, because WT and Bhlhe40^−/− counterparts not only proliferated and contracted similarly but also generated a comparable pool of memory cells (Supplemental Fig. 1A, 1B), suggesting that NK cell proliferation and “adaptive” features are not impacted by loss of Bhlhe40. Additionally, their maturation/activation profiles were similar and commensurate with effector NK cells at day 7 PI (Supplemental Fig. 1C). Furthermore, WT and Bhlhe40^−/− NK cells were indistinguishable in their ability to produce granzyme B and upregulate CD69 (Supplemental Fig. 1D), suggesting that cytolytic capacity and activation of NK cells remained intact in the absence of Bhlhe40.

In transcriptomic analysis, among the genes dysregulated in Bhlhe40^−/− NK cells compared with WT during MCMV infection were Ifng, Cxcr4, and Bhlhe40 (lower in Bhlhe40^−/−) and Il21r, Miga1, and Cnbd2 (elevated in Bhlhe40^−/−) (Fig. 3A). Confirming that Ifng mRNA was downregulated, Bhlhe40^−/− NK cells produced less IFN-γ protein than did WT NK cells following infection (Fig. 3B). Moreover, Bhlhe40^−/− NK cells similarly produced diminished IFN-γ compared with their WT counterparts following in vitro stimulation with proinflammatory cytokines (Fig. 3C). Finally, we observed that Bhlhe40^−/− mice were more susceptible than WT mice to MCMV challenge, and most Bhlhe40^−/− mice succumbed at a similar rate and with similar kinetics to NK cell-depleted WT mice (Fig. 3D).

To determine the direct binding sites of Bhlhe40, we performed CUT&RUN on NK cells stimulated with IL-12 + IL-18. We identified a total of 316 Bhlhe40 binding sites, where most of these peaks were called in the stimulated condition (n = 279; Fig. 4A, 4B). Through global differential binding analysis, we found that the majority of these sites were significantly enriched upon IL-12 + IL-18 stimulation (Fig. 4C). Notably, the top differential binding site of Bhlhe40 was within the Ifng locus (Fig. 4D). Among the significant differentially binding sites, most were annotated as promoter peaks (74%, 208 of 281; Fig. 4E). Motif analysis with differentially binding peaks revealed de novo motifs that suggest potential interactions with families including ATF1 and IRF6, and because the bHLH family is present within the motif database used, the top match includes TFEC, a bHLH family member (Fig. 4F). Taken together, our data indicate that Bhlhe40 is critical for host protection against viral infection by acting as a direct driver of the proinflammatory/immunoregulatory cytokine IFN-γ in NK cells.

Previous studies have identified that IL-12 and STAT4 signaling can drive multiple effector functions in NK cells during MCMV infection, including the secretion of IFN-γ and the prolific expansion of antiviral “adaptive” NK cells (1). How these different processes are controlled by a single transcription factor, STAT4, and whether these specific functions can be segregated in their regulation remained incompletely understood. We previously identified that STAT4-mediated expression of Zbtb32, Runx1, Runx3, and IRF8 could specifically impact clonal proliferation but not IFN-γ secretion by antiviral NK cells (5, 6, 8), whereas STAT4-mediated upregulation of STAT5 could impact both processes (7). In the present study, we have discovered a cell-intrinsic role for the transcription factor Bhlhe40 in specifically controlling IFN-γ production but not proliferation of NK cells following viral infection, highlighting a mechanism that decouples STAT4-driven innate functions from STAT4-driven adaptive features.

Control of cytokine secretion by Bhlhe40 has also been observed in other cell types, but whether Bhlhe40 was directly regulated by STAT transcription factors in these settings was not elucidated (12, 13, 24–27). Through epigenetic studies, we found that STAT4 is inducing the permissive histone H3K4me3 mark at the Bhlhe40 locus rather than mediating changes in chromatin accessibility, in contrast to many other genetic loci targeted by STAT4 (3). This distinct role for STAT4 in modulating the chromatin landscape at the Bhlhe40 locus critically drives its expression and subsequently protective IFN-γ from NK cells during viral infection. Thus, the mode of STAT4 activity can potentially show specificity dependent upon the range of cotranscription factors it induces in activated NK cells (given that STAT4 also binds to the Ifng loci [28]), and understanding such underlying mechanisms can aid in the goal of harnessing these potent effector cells in our battle against infectious diseases.

The authors have no financial conflicts of interest.

We thank members of the Sun Lab for experimental assistance and helpful discussions.