BHLHE40 Mediates Cross-Talk between Pathogenic T_H17 Cells and Myeloid Cells during Experimental Autoimmune Encephalomyelitis

The myelin oligodendrocyte glycoprotein (MOG)_35–55 peptide–induced experimental autoimmune encephalomyelitis (EAE) animal model of multiple sclerosis in C57BL/6 mice represents a CD4⁺ T cell–mediated demyelinating disease of the CNS (1). Following immunization with MOG_35–55 peptide and adjuvant, CD4⁺ T cells primed in the periphery develop into T_H1 and T_H17 cells that are both capable of mediating disease. Fate-reporter mice that make use of Il17a-driven Cre expression have been instrumental in identifying populations of CD4⁺ T cells that express both IL-17A and IFN-γ (here referred to as T_H1/17 cells) as well as IFN-γ⁺ ex-T_H17 cells following EAE induction (2), both of which have been associated with encephalitogenicity that is dependent on their GM-CSF production (2–6).

Both active immunization and adoptive transfer models of EAE have been useful for identifying important T cell–intrinsic regulators of neuroinflammation. Using adoptive transfer models of EAE, the transcription factor T-bet is needed for encephalitogenicity of both T_H17 and so-called T_HGM cells (CD4⁺ T cells cultured to enhance their GM-CSF production) (7–9); however, Il17a-Cre–mediated deletion of floxed alleles of either Tbx21 (T-bet) or Rorc (RORγt) led to only modest reductions in active EAE clinical disease (10). Other transcription factors are important for T cell pathogenicity during actively induced EAE such as Fosl2 (11) and Blimp-1 (12), although these have not been selectively deleted in only T_H17 cells to test their impact on clinical disease. Although a few transcription factors, such as JunB (13, 14) and STAT4 (15, 16), have been demonstrated as having important intrinsic roles in T_H17 cells using Il17a-Cre–mediated deletion, other transcription factors also likely contribute to T_H17 cell–intrinsic encephalitogenicity.

We and others have shown a cell-intrinsic requirement for the transcription factor basic helix-loop-helix family member e40 (BHLHE40) in CD4⁺ T cells for disease in actively induced EAE in Bhlhe40^−/− mice, through the action of BHLHE40 as a positive regulator of GM-CSF and negative regulator of IL-10 production (17–19). Furthermore, in an adoptive transfer model of EAE, BHLHE40-deficient MOG-specific T_H1 or T_H17 cells were incapable of mediating disease (20). Others have shown Bhlhe40 induction downstream of the T_H17 cell–associated transcription factors RORγt, RORα, and SATB1 (11, 21, 22). In the case of SATB1, overexpression of BHLHE40 restored GM-CSF production and pathogenicity by Satb1-deficient CD4⁺ T cells (22). Additionally, Bhlhe40 expression has been correlated with a pathogenic CD4⁺ T cell transcriptional signature during EAE (23–26). In this study, we sought to more thoroughly explore the role of BHLHE40 specifically in T_H17 cells during active EAE. Using a combination of BHLHE40 reporter and conditional knockout mice along with single-cell RNA sequencing (scRNA-seq), we identify specific roles for BHLHE40 in determining the pathogenicity of T_H17 cells during EAE.

Bhlhe40^GFP (20) and Bhlhe40^fl/fl (27) mice have been previously described. Bhlhe40^GFP mice were crossed to Il17a-Cre (The Jackson Laboratory, 016879, Il17a^{tm1.1(icre)Stck}/J) and Rosa26-TdTomato mice (The Jackson Laboratory, 007914, B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J). Bhlhe40^fl/fl mice were crossed to Cd4-Cre (022071, B6.Cg-Tg(Cd4-cre)1Cwi/BfluJ), Il17a-Cre, Lyz2-Cre (018956, B6N.129P2(B6)-Lyz2^tm1(cre)Ifo/J), and S100a8-Cre mice (021614, B6.Cg-Tg(S100A8-cre,-EGFP)1Ilw/J) (all from The Jackson Laboratory). All mice were on the C57BL/6 background and used between 8 and 20 wk of age. Animal experiments were approved by the Animal Studies Committee of Washington University in St. Louis. Littermates were used when possible, and both male and female mice were used in experiments.

EAE induction, EAE clinical scoring, MOG_35–55 peptide immunizations, and assessment of T cell responses in the draining lymph node (DLN) at day 7 postimmunization were performed as previously described (20). To achieve the best use of littermates and age-matched mice, independent experiments were performed and data were combined.

These techniques have been previously described (20). In brief, DLNs were collected at day 7 postimmunization to prepare single-cell suspensions. Brains and spinal cords (CNS) were processed together from either naive or EAE-induced mice at day 14 postinduction. Surface staining and intracellular cytokine staining were performed using the Abs in Supplemental Table I. For scRNA-seq experiments, CD45⁺7-aminoactinomycin D⁻ live, single cells of the CNS were sorted using a FACS-Aria II (BD Biosciences), washed, and adjusted to 10³ cells/μl in PBS + 0.04% BSA.

scRNA-seq was performed at the McDonnell Genome Institute using the Chromium single cell 3’ library kit v2 and Chromium instrument (10x Genomics, Pleasanton, CA). Sequencing was performed on an Illumina HiSeq 4000 instrument. Data were processed as described (28) for dimensionality reduction, construction of t-distributed stochastic neighbor embedding (tSNE) plots, clustering, identification of cluster-specific genes, and differential expression analysis. Differentially expressed genes (adjusted p value <0.05 and log₂ fold-change >0.35) were cross-referenced to the hallmark (29), Reactome (https://reactome.org/) (30), and KEGG (https://www.kegg.jp/) (31) gene sets in the Molecular Signatures Database (MSigDB) (https://gsea-msigdb.org/gsea/msigdb/index.jsp). Monocle 2 was used for pseudotemporal analysis of CNS myeloid cell populations (32, 33). Data have been deposited in the Gene Expression Omnibus (GSE234705).

Data were analyzed with Prism (GraphPad Software) and one-way ANOVA, Student t tests, or Mann–Whitney U tests were used as indicated in individual figure legends. For relevant comparisons where no p value is shown, the p value was >0.05. Horizontal bars represent the means, and error bars represent the SEM. Although we did not directly account for interexperiment variability in experiments in which EAE disease scores were measured, we confirmed that trends were repeated in each independent experiment.

To track BHLHE40 expression along with IL-17A fate mapping, we crossed Bhlhe40^GFP BAC transgenic mice to Il17a-Cre Rosa26-TdTomato mice. We immunized these mice with either pertussis toxin (PTX), MOG_35–55 in CFA, or MOG_35–55 in CFA with PTX injections, and on day 7 postimmunization, we analyzed the CD4⁺ T cell compartment for TdTomato and GFP (as a surrogate for BHLHE40) expression (Fig. 1A). Immunization with MOG_35–55 in CFA increased the percentage and number of TdTomato⁺ CD4⁺ cells, but only with the addition of PTX was the percentage and number of GFP⁺ and GFP⁺ TdTomato⁺ CD4⁺ T cells increased (Fig. 1B, 1C). This is consistent with our previous finding that PTX stimulates secretion of IL-1β from myeloid cells, which induces BHLHE40 expression in CD4⁺ T cells (20). To ask whether BHLHE40 was expressed in T_H1/17 and ex-T_H17 cells, we examined TdTomato⁺CD4⁺ T cells for intracellular production of IFN-γ and IL-17A. GFP was highly expressed in IFN-γ⁺IL-17A⁺ T_H1/17 cells and IFN-γ⁺IL-17A⁻ ex-T_H17 cells (Fig. 1D).

We further characterized cytokine production of IFN-γ, IL-17A, and GM-CSF from the different types of reporter cells in the DLN. GM-CSF production correlated with GFP expression in both TdTomato⁺ and TdTomato⁻ CD4⁺ T cells (Fig. 1E). In addition, the greatest frequency of multifunctional CD4⁺ T cells (i.e., simultaneously producing IL-17A, IFN-γ, and GM-CSF) was found in the GFP⁺TdTomato⁺ group. A similar trend of cytokine production was seen in the CNS on day 14 after EAE induction (Fig. 1F, Supplemental Fig. 1). Within the CNS, a notable difference was an increase in IL-17A⁻IFN-γ⁺GM-CSF⁻ and IL-17A⁻IFN-γ⁺GM-CSF⁺ cells in both TdTomato⁺GFP^– ex-T_H17 cells and TdTomato⁻GFP⁻ cells. This could indicate underreporting of BHLHE40 expression by the GFP reporter after permeabilization for intracellular cytokine staining or that some cells use BHLHE40-independent pathways to produce these cytokines. Nevertheless, the greatest fraction of total GM-CSF producers (i.e., IL-17A/IFN-γ agnostic) in both the DLN and CNS was found in GFP⁺ cells, consistent with the known role for BHLHE40 in supporting GM-CSF production.

Because BHLHE40 was highly expressed in T_H1/17 and ex-T_H17 cells, we tested whether Il17a-Cre⁺Bhlhe40^fl/fl mice would have an altered course of EAE. For comparison, we also induced EAE in Cd4-Cre⁺Bhlhe40^fl/fl mice, which lack Bhlhe40 expression in all T cells and which have previously been shown to be protected from EAE (34). As expected, Cd4-Cre⁺Bhlhe40^fl/fl mice were highly protected during EAE compared with Bhlhe40^fl/fl controls (Fig. 2A). Notably, we found consistently reduced EAE severity upon Il17a-mediated deletion of Bhlhe40, although not to the same extent as Cd4-Cre–mediated deletion (Fig. 2B). We restimulated cells isolated from the CNS of naive mice and from immunized Bhlhe40^fl/fl, Cd4-Cre⁺Bhlhe40^fl/fl, or Il17a-Cre⁺Bhlhe40^fl/fl mice 14 d after immunization to examine cytokine production. In both Cd4-Cre⁺Bhlhe40^fl/fl and Il17a-Cre⁺Bhlhe40^fl/fl mice we found that the most striking reduction was in the IL-17A⁻IFN-γ⁺GM-CSF⁺ population, which was the most increased population in immunized Bhlhe40^fl/fl mice relative to naive mice (Fig. 2C, 2D). Based on our fate-mapping experiments, in Il17a-Cre⁺Bhlhe40^fl/fl mice these cells likely represent ex-T_H17 cells that had once produced IL-17A and therefore deleted Bhlhe40, reducing their potential for GM-CSF production. As BHLHE40 has been shown to negatively regulate IL-10 in CD4⁺ T cells (18), we examined IL-10 production in the context of Il17a-Cre–driven deletion of BHLHE40. At the peak of disease, Il17a-Cre⁺Bhlhe40^fl/fl CD4⁺ T cells had no statistically significant difference in IL-10 production compared with Bhlhe40^fl/fl CD4⁺ T cells, although there was a trend toward greater IL-10 production upon BHLHE40 deletion (Fig. 2E, 2F). As expected, Il17a-Cre⁺Bhlhe40^fl/fl CD4⁺ T cells had decreased total GM-CSF production. This suggests that the clinical protection seen in Il17a-Cre⁺Bhlhe40^fl/fl mice is likely not due to local IL-10 production in the CNS.

CNS-infiltrating myeloid cells are the critical responders to GM-CSF produced by encephalitogenic CD4⁺ T cells (35, 36). We therefore examined the impact of altered cytokine production from BHLHE40-deficient CD4⁺ T cells on the myeloid cells present in the CNS during EAE. Upon immunization, there was infiltration of myeloid cells into the CNS of Bhlhe40^fl/fl control mice (Fig. 2G, 2H). These infiltrating myeloid cells, as well as CNS-resident microglia, upregulated MHC class II upon activation (Fig. 2G, 2H). In both immunized Cd4-Cre⁺ and Il17a-Cre⁺Bhlhe40^fl/fl mice, there was a dramatic reduction in the number and activation of infiltrating myeloid cells and activation of microglia compared with immunized Bhlhe40^fl/fl mice (Fig. 2G, 2H). To test for the possibility of a cell-intrinsic role for BHLHE40 in infiltrating myeloid cells, we crossed Bhlhe40^fl/fl to Lyz2-Cre and actively induced EAE. We did not find a role for BHLHE40 in myeloid cells or in neutrophils using S100a8-mediated deletion (Supplemental Fig. 2). Overall, we show that altered cytokine production from BHLHE40-deficient CD4⁺ T cells resulted in a decrease in the number and activation of infiltrating myeloid cells.

To further probe how deletion of BHLHE40 in T cells or IL-17A–producing cells impacts on the immune cells present in the CNS during EAE, we performed scRNA-seq. Immune cells (CD45⁺) were sorted from the CNS of naive Bhlhe40^fl/fl controls or from immunized Bhlhe40^fl/fl, Cd4-Cre⁺Bhlhe40^fl/fl, or Il17a-Cre⁺Bhlhe40^fl/fl mice on day 14 of EAE. After dimensionality reduction of the scRNA-seq data, immune cells partitioned into 17 unique clusters (Fig. 3A). Clusters were identified based on gene expression of common lineage-specific markers (Fig. 3B, Supplemental Fig. 3A). Immunization of Bhlhe40^fl/fl mice resulted in an increased fraction of CD4⁺ T cells (cluster 1) and myeloid cells (clusters 5 and 7) infiltrating into the CNS compared with naive mice (Fig. 3C, Supplemental Fig. 3B). This increase was not seen in the CNS of immunized Cd4-Cre⁺Bhlhe40^fl/fl mice, which largely resembled the naive CNS with the exception of an increased frequency of monocytes (cluster 2) and neutrophils (clusters 4 and 11) (Fig. 3C). The CNS of immunized Il17a-Cre⁺Bhlhe40^fl/fl mice largely resembled those of immunized Bhlhe40^fl/fl mice, with a high percentage of CD4⁺ T cells (cluster 1) and infiltrating myeloid cells (the collection of clusters 5, 7, and 9) (Fig. 3C, Supplemental Fig. 3B). Based on our sorting strategy using CD45 expression, few microglia (cluster 14) were sequenced from our various samples. In general, microglia resembled myeloid cluster 7 in terms of gene expression based on their proximity on the tSNE plot, but due to their low abundance in individual samples, comparisons between genotypes were not made.

To determine the different pathways upregulated or downregulated in CD4⁺ T cells from immunized Bhlhe40^fl/fl and Il17a-Cre⁺Bhlhe40^fl/fl mice, we cross-referenced the list of top differentially expressed genes to gene signatures within the MSigDB. CD4⁺ T cells from the immunized Bhlhe40^fl/fl CNS upregulated genes in pathways involved in the innate immune system and complement (Supplemental Fig. 3C). The pathways most enriched in CD4⁺ T cells from immunized Il17a-Cre⁺Bhlhe40^fl/fl CNS included both type 1 and type 2 IFN responses (Supplemental Fig. 3D).

We further examined several clusters of infiltrating myeloid cells (clusters 5, 7, and 9). Cluster 5 was present in immunized Bhlhe40^fl/fl and Il17a-Cre⁺Bhlhe40^fl/fl CNS in roughly equal proportions (Fig. 4A). Cluster 7 was dramatically higher in immunized Bhlhe40^fl/fl CNS, compared with its near absence in the other samples. Similarly, cluster 9 was largely specific for the Il17a-Cre⁺Bhlhe40^fl/fl CNS. Cluster 5 was enriched in inflammatory pathways, such as IFN-γ response, TNF-α signaling, inflammatory response, and cytokine signaling (Fig. 4B, 4E). Cluster 7 appeared to be a highly phagocytic population, with enrichment in pathways such as neutrophil degranulation, lysosome, and complement (Fig. 4C, 4F). Cluster 9 was enriched in cytokine signaling and type 1 IFN responses (Fig. 4D, 4G), similar to the CD4⁺ T cells from immunized Il17a-Cre⁺Bhlhe40^fl/fl CNS. The emergence of a type 1 IFN signature in cluster 9, which was almost exclusively present in the Il17a-Cre⁺Bhlhe40^fl/fl CNS, and the CD4⁺ T cells from this same sample was notable. In summary, deletion of Bhlhe40 from IL-17A–producing cells resulted in CNS-infiltrating myeloid cells with altered phenotypes during EAE.

To examine the potential for developmental relationships between the myeloid clusters, we employed the Monocle 2 algorithm (32, 33) to assign a pseudotime trajectory to the myeloid cells in clusters 2, 5, 7, and 9 (Fig. 4H). We included cluster 2 (monocytes) in this analysis, as these cells likely represent the precursor population that gives rise to the other myeloid clusters (36–39). When the clusters were mapped onto the pseudotime plot, monocytes (cluster 2) were predicted to be the precursor that progresses through one of two differentiation pathways (Fig. 4I, 4J). The first branch point (branch point 1) was where most of cluster 9 resided, and the second branch point (branch point 2) involved a differentiation from cluster 5 to cluster 7. Part of cluster 5 appeared to be a distinct cell fate, while the rest of cluster 5 appeared mostly in the transition between clusters 9 and 7. This is consistent with the fact that some of the enriched pathways present in cluster 5 are also shared pathways with cluster 9 (Fig. 4B, 4D). It appeared that cluster 9, an IFN-responsive population, is a cell fate that was largely unlocked by the absence of Bhlhe40 from IL-17A–producing cells. Overall, scRNA-seq revealed regulation of myeloid cell gene expression by pathogenic CD4⁺ T cells in a BHLHE40-dependent manner.

We sought to interrogate the expression of and role for BHLHE40 in T_H17 cells and T_H17-cell derived populations (T_H1/17 and ex-T_H17 cells) during active EAE. Using our dual reporter system we demonstrate that after EAE induction, BHLHE40 is expressed in both T_H1/17 and ex-T_H17 cells and that BHLHE40 expression correlates with GM-CSF production. Deletion of BHLHE40 either in all T cells (Cd4-Cre) or in T_H17 cells (Il17a-Cre) was protective from clinical EAE disease. The CNS of immunized Il17a-Cre⁺Bhlhe40^fl/fl mice contained fewer IL-17A⁻IFN-γ⁺GM-CSF⁺CD4⁺ T cells compared with immunized Bhlhe40^fl/fl control mice, suggesting that BHLHE40 regulates pathogenicity at a step subsequent to T_H17 differentiation, likely as these cells convert into ex-T_H17 cells. As a consequence of altered CD4⁺ T cell cytokine production, there were fewer and less activated infiltrating myeloid cells in the CNS at the peak of EAE clinical disease of Il17a-Cre⁺Bhlhe40^fl/fl mice.

Using scRNA-seq we show how populations of immune cells in the CNS change with the presence or absence of BHLHE40 in our different deletion systems. The Cd4-Cre⁺Bhlhe40^fl/fl CNS largely resembled the CNS of naive mice, with low percentages of pathogenic CD4⁺ T cells and few infiltrating myeloid cells compared with the increases seen in the CNS of immunized Bhlhe40^fl/fl and Il17a-Cre⁺Bhlhe40^fl/fl mice. Both CD4⁺ T cells (cluster 1) and myeloid cluster 9 (abundant in the Il17a-Cre⁺Bhlhe40^fl/fl CNS) were enriched for IFN response pathways, highlighting a potential mechanism of protection upon deletion of BHLHE40 in T_H17 cells, potentially downstream of decreased GM-CSF and IFN-γ production by BHLHE40-deficient CD4⁺ T cells.

GM-CSF production from encephalitogenic T cells is essential for infiltrating myeloid cell activation in the CNS during EAE (6, 35, 36, 38, 40, 41). We hypothesize that in the setting of Cd4-Cre–mediated deletion of Bhlhe40, there is not enough GM-CSF produced from CD4⁺ T cells to differentiate infiltrating monocytes in the CNS, potentially explaining why this scRNA-seq sample lacks monocyte-derived cells (clusters 5, 7, and 9). In the case of Il17a-Cre–mediated deletion of Bhlhe40, there may still be enough GM-CSF production from unaffected T_H1 cells to activate infiltrating monocytes, allowing these cells to differentiate into unique cell fates. The loss of Bhlhe40 in T_H1/17 and ex-T_H17 cells in Il17a-Cre⁺Bhlhe40^fl/fl mice allows for a new type I IFN–responsive myeloid cell fate to emerge (cluster 9). It is interesting that both the protective effects of endogenous type I IFNs and the pathogenic effects of GM-CSF act through monocyte-derived cells that directly respond to these cytokines during active EAE (42). GM-CSF may directly inhibit a monocyte-derived cell’s ability to effectively respond to type I IFNs, which has been suggested in vitro (43) and recently in vivo during EAE (36).

Additionally, it has been suggested that one mechanism of protection by type I IFNs is to limit IL-1β production from CNS-infiltrating monocytes, resulting in less IL-1β–dependent GM-CSF production from CD4⁺ T cells (44). BHLHE40 deficiency in IL-17A–producing CD4⁺ T cells could disrupt the normal GM-CSF/IL-1β–dependent positive amplification loops shown to take place during EAE (20, 36, 44–46). A more comprehensive understanding of cytokine production by autoreactive CD4⁺ T cells and cytokine feedback loops active in infiltrating monocytes in EAE could inform future therapeutic discovery efforts in multiple sclerosis. Recombinant IFN-β was the first Food and Drug Administration–approved therapy for multiple sclerosis, although its clinical efficacy has been surpassed by other therapeutics. Perhaps a dual approach of altering myeloid cells’ responsiveness to pathogenic signals (GM-CSF) and enhancing their responsiveness to protective signals (type I IFNs) may be worthy of future investigation.

The authors have no financial conflicts of interest.

We thank the Genome Technology Access Center at the McDonnell Genome Institute at Washington University School of Medicine for help with genomic analysis. We also thank E. Lantelme and A. Cullen for help with cell sorting.