DDX5 and its associated lncRNA Rmrp modulate TH17 cell effector functions.

T helper 17 (TH17) lymphocytes protect mucosal barriers from infections, but also contribute to multiple chronic inflammatory diseases. Their differentiation is controlled by RORγt, a ligand-regulated nuclear receptor. Here we identify the RNA helicase DEAD-box protein 5 (DDX5) as a RORγt partner that coordinates transcription of selective TH17 genes, and is required for TH17-mediated inflammatory pathologies. Surprisingly, the ability of DDX5 to interact with RORγt and coactivate its targets depends on intrinsic RNA helicase activity and binding of a conserved nuclear long noncoding RNA (lncRNA), Rmrp, which is mutated in patients with cartilage-hair hypoplasia. A targeted Rmrp gene mutation in mice, corresponding to a gene mutation in cartilage-hair hypoplasia patients, altered lncRNA chromatin occupancy, and reduced the DDX5-RORγt interaction and RORγt target gene transcription. Elucidation of the link between Rmrp and the DDX5-RORγt complex reveals a role for RNA helicases and lncRNAs in tissue-specific transcriptional regulation, and provides new opportunities for therapeutic intervention in TH17-dependent diseases.

T-helper 17 (Th17) cells are CD4+ lymphocytes that help protect mucosal epithelial barriers against bacterial and fungal infections [1], and that are also critically important in multiple autoimmune diseases [2-7]. In murine models, attenuation of RORγt activity results in protection from experimental autoimmune encephalomyelitis (EAE), T cell transfer-mediated colitis, and collagen-induced arthritis [2-5]. The Th17 cell differentiation program is defined by the induced expression of RORγt [2], a sterol ligand-regulated nuclear receptor that focuses the activity of a cytokine-regulated transcriptional network upon a subset of key genomic target sites, including genes encoding the signature Th17 cytokines (IL-17A, IL-17F, IL-22) as well as IL-23R, IL-1R1, and CCR6 [8]. Like other nuclear receptors, RORγt interaction with its ligands results in recruitment of co-activators at regulated genomic loci [9]. We identified two new RORγt partners in Th17 cells, an RNA helicase and a long noncoding (lnc) RNA, which together associate with RORγt to confer target locus-specific activity in enabling the T cell effector program.

The RNA helicase DEAD-box protein 5 (DDX5) functions in multiple cellular processes [10], including transcription and ribosome biogenesis [11-17] in both a helicase activity-dependent and -independent manner. The lncRNA Rmrp, RNA component of Mitochondria RNA-processing endoribonuclease (RNase MRP), is highly conserved between mouse and human and is essential for early murine development [18]. Rmrp was first identified as a component of the RNase MRP complex that cleaves mitochondrial RNAs [19]. In yeast, Rmrp contributes to ribosomal RNA processing and regulates mRNA degradation [20]. In humans, mutations located in evolutionarily conserved nucleotides at the promoter or within the transcribed region of RMRP result in cartilage-hair hypoplasia (CHH), a rare autosomal recessive disorder characterized by early childhood onset of skeletal dysplasia, hypoplastic hair, defective immunity, predisposition to lymphoma, and neuronal dysplasia of the intestine [21,22]. Immune deficiency in CHH patients is associated with recurrent infections, hematological abnormalities, and autoimmune pathologies in the joints and kidneys [23]. The precise mechanisms by which Rmrp functions in the immune system have yet to be elucidated. Here we show that DDX5, through its helicase activity, mediates Rmrp-dependent binding to RORγt and recruitment to a subset of its chromatin target sites, thus controlling the differentiation of Th17 cells at steady state and in animal models of autoimmunity.

DDX5 regulation of RORγt target genes

To identify novel interacting partners of RORγt in Th17 cells, we enriched for endogenous RORγt-containing protein complexes and subsequently determined protein composition using LC-MS/MS (workflow diagramed in Extended Data Fig. 1a). Among the top hits of RORγt-interacting proteins was the RNA helicase DDX5. We validated this interaction through conventional co-immunoprecipitation (coIP) experiments followed by immunoblot analysis (Extended Data Fig. 1b).

Extended Data Figure 1

Identification of DDX5 as a RORγt-interacting partner

a, Mass spectrometry experimental workflow. Sorted naïve CD4+ T cells from WT mice were cultured in vitro in Th17 polarizing conditions for 48h. Immunoprecipitation of endogenous RORγt was performed using RORγ/γt-specific antibodies on whole cell lysates. RORγt enrichment in pull-down was confirmed by immunoblot. Immunoprecipitated proteins were digested and analyzed by mass spectrometry. The listed DDX5 peptides were identified in the Th17 RORγt immunoprecipitate. b, Co-immunoprecipitaton of DDX5 with anti-RORγt in lysates of in vitro polarized Th17 cells. For gel source data, see Supplementary Figure 1. c, Cell surface phenotype of splenic and lymph node DAPI−CD19−CD8α−CD4+T cells from WT and DDX5-Tko animals, examined by flow cytometry. d, Immunoblot of RORγt protein expression whole cell lysate of cultured Th17 cells from WT or DDX5-Tko animals. For gel source data, see Supplementary Figure 1. e, Immunofluorescence staining of RORγt in cultured Th17 cells from WT or DDX5-Tko animals. f, Immunofluorescence staining of DDX5 revealed nuclear localization in Th17 cells.

We investigated the function of DDX5 in T cells by breeding ddx5 conditional mutant mice with CD4Cre mice to generate T cell-specific DDX5-deficient animals (Ddx5 CD4Cre mice, designated DDX5-Tko). DDX5-Tko mice were born at the expected Mendelian ratio, were fertile, and did not display any gross phenotypic abnormalities. Activation status of T cells in the periphery was similar between Ddx5+/+CD4Cre+ (designated wildtype, WT) and mutant mice (Extended Data Fig. 1c) that had no DDX5 protein in spleen and lymph node CD4+ T cells (Extended Data Fig. 1d). Sorted naïve CD4+ T cells from WT and DDX5-Tko mice did not display significant differences in polarization towards Th1, Th2, and iTreg phenotypes in vitro (Fig. 1a). In contrast, DDX5-Tko naïve T cells cultured under Th17 polarizing conditions produced substantially less IL-17A than WT cells (Fig. 1a). RORγt protein expression and nuclear localization were similar between WT and DDX5-Tko Th17-polarized cells (Extended Data Fig. 1d–e) and, like RORγt, DDX5 protein localized mainly to the nucleus (Extended Data Fig. 1f). These results suggest that DDX5 is not required for Th17 lineage commitment, but contributes to Th17 cell effector functions.

Figure 1

Requirement for DDX5 in Th17 cytokine production in vitro and at steady state in vivo

a, Selective Th17 cell differentiation defect in DDX5-deficient T cells after polarization for 96 h. Representative of three independent experiments. b, Volcano plot of RNA-seq of cultured Th17 cells from DDX5-Tko mice and littermate controls. Black dots: differentially expressed genes (minimum fold change of two with p-value < 0.05). Blue dots: known RORγt-dependent genes. Red dots: top RORγt-DDX5 co-regulated genes. c, SFB colonization and percentage and number of RORγt+ CD4+ T cells and d, number of IL-17A-producing CD4+ T cells in ileal lamina propria (LP) of co-housed WT (n=5) and DDX5-Tko (n=5) animals. Graphs show mean ± standard deviation (s.d.) from two independent experiments, combined. N.S., not significant. ** p<0.01 (Prism, Paired t-test). e, Representative IL-17A expression in CD4+Foxp3−RORγt+ Th17 cells from ileal LP of WT and DDX5-Tko animals after restimulation.

DDX5 can function as a transcriptional coactivator [12,24,25], augmenting the activities of other nuclear receptor family members, including the estrogen and androgen receptors [12,26]. To determine if DDX5 partners with RORγt to facilitate the Th17 cell transcriptional program, we performed RNA-seq on in vitro polarized Th17 cells from WT or DDX5-Tko mice. Among the 325 genes that were significantly dysregulated in DDX5-deficient T cells 96hrs post polarization, approximately 40% had been previously identified as RORγt targets in Th17 cells (Extended Data Fig. 2a) [8]. Ingenuity Pathway Analysis of DDX5-RORγt-coregulated genes revealed enrichment in “T helper cell differentiation program” as well as “interleukin production” (Extended Data Fig. 2b). Co-regulated genes (Fig. 1b) included those for the Th17 cytokines (Il17a, Il17f, and Il22) (Extended Data Fig. 2c). Independent biological samples were used to validate a subset of RORγt target genes with and without altered expression in DDX5-deficient Th17 cells (Extended Data Fig. 2d).

Extended Data Figure 2

DDX5 co-regulates a subset of RORγt transcriptional targets in polarized Th17 cells

a, Venn diagram of distinct and overlapping genes regulated by RORγt and/or DDX5, as determined from RNA-seq studies. b, Ingenuity Pathway Analysis of DDX5- and RORγt-coregulated genes. c, IGV browser view showing biological replicate RNA-seq coverage tracks of control, DDX5-Tko, or RORγt-deficient in vitro polarized Th17 cell samples at the Il17a, Il22, Ddx5, and Rorc loci. d, Independent qRT-PCR validation of RNA-seq results confirming effects of DDX5 deletion on RORγt target gene expression. Graph shows mean ± s.d.

We used anti-DDX5 antibodies in ChIP-seq studies to identify DDX5-occupied loci genomewide. A specific subset of previously published RORγt-occupied loci, including Il17a and Il17f, were enriched for DDX5 co-localization, as determined by SeqMiner clustering analysis (Extended Data Fig. 3a–b). Conventional ChIP-qPCR was used to validate DDX5 enrichment at the Il17a and Il17f loci and its dependency on RORγt in polarized Th17 cells (Extended Data Fig. 3c). These results suggest that DDX5 overlaps with RORγt in modulating a specific subset of the Th17 cell transcriptional program.

Extended Data Figure 3

DDX5 chromatin localization in Th17 cells

a, ChIP-seq-generated heatmap of DDX5 occupancy in regions centered on 16,003 RORγt-occupied sites (+/− 2kb). KMeans linear normalization was used for clustering analysis by SeqMiner. Metagene analysis on cluster 1 depicts RORγt-occupied regions with DDX5 enrichment in wildtype but not DDX5-Tko cells; cluster 2 represents RORγt-occupied regions without DDX5 enrichment. b, IGV browser view of Il17a, Il17f, and Rorc loci with ChIP-seq enrichment for RNAPol-II, RORγt, and DDX5. c, Independent ChIP-qPCR of DDX5 in polarized Th17 cells. DDX5 occupancy at the Il17a and Il17f loci (as identified by RORγt ChIP-seq MACS peak called #32 and #39 respectively from b.) in control, DDX5-Tko, or RORγt-deficient cells. Results are representative of two independent experiments. Each experiment was performed with two technical replicates. Graph shows mean ± s.d. ** p<0.01 (Prism, t-test).

DDX5 function in vivo in Th17 cells

At steady state, cytokine-producing Th17 cells populate the small intestinal lamina propria of animals colonized with Segmented Filamentous Bacteria (SFB), a commensal microbe [27]. When colonized with SFB, DDX5-Tko and their WT littermates had similar numbers of ileal-residing Foxp3−RORγt+CD4+ Th17 cells (Fig. 1c). However, the number and proportion of IL-17A–producing cells among RORγt+CD4+ cells from DDX5-Tko animals were markedly reduced compared to WT littermate controls (Fig. 1d–e).

To evaluate the role of DDX5 in Th17-driven inflammation, we employed a T cell transfer model of colitis, in which disease severity is dependent on RORγt expression in donor T cells [3,28]. Following transfer of CD4+CD45RBhi T cells into Rag-deficient recipients, mice that received WT T cells experienced weight loss (Fig. 2a) and developed colitis (Fig 2b), whereas recipients of DDX5-Tko cells did not. Total RNA harvested from large intestine lamina propria mononuclear cells revealed a significant reduction of both IL-17A and IFNγ transcripts from recipients of DDX5-Tko cells compared to WT controls (Extended Data Fig. 4a). Interestingly, there were comparable proportions of IFNγ-producing CD4+RORγt−Tbet+ (conventional Th1) cells in recipients of either WT or DDX5-Tko cells (Extended Data Fig. 4b). However, recipients of DDX5-Tko cells displayed a significant reduction in CD4+Foxp3−RORγt+ T cells co-expressing IL-17A and IFNγ, a cardinal feature of pathogenic T cells in several inflammatory disease settings (Fig. 2c and Extended Data Fig. 4b) [2,29,30]. Consistent with a loss of pathogenic capacity, DDX5-Tko mice also exhibited attenuated disease compared to WT controls during experimental autoimmune encephalomyelitis (EAE) (Fig. 2d). Analysis of spinal cord infiltrates post immunization revealed a reduced proportion of IL-17A–producing CD4+ T cells (Fig. 2e and Extended Data Fig. 4c). In concert with our in vitro findings, these results in animals indicate that DDX5 selectively regulates the Th17 effector program, both in steady state and under inflammatory conditions.

Figure 2

Role of DDX5 in mouse models of Th17 cell-mediated autoimmune disease

a, Weight change in Rag2−/− recipients of WT or DDX5-Tko CD4+ naïve T cells in the transfer model of colitis measured on days 0, 10, 25, 37, and 45. (PBS: n=4; WT: n=9; DDX5-Tko: n=13, combined from three independent experiments). b, H&E staining and analysis of large intestine (LI) at day 45. Representative sections (black bar = 100µm) and histology scores (scale of 0–24) are shown. Scores for PBS (n=3), WT (red, n=8) and DDX5-Tko (blue, n=7) mice are from two independent experiments. c, Cytokine production defect in DDX5-Tko Th17 (RORγt+) but not Th1 (RORγt T-bet+) cells in LILP at day 45 (n=4 per group). d, EAE disease scores (scale of 0–5) in co-housed MOG-immunized littermates. WT (n=13) and DDX5-Tko (n=11) mice, combined from three independent experiments. e, Defective IL-17A production in DDX5-Tko CD4+RORγt+ cells in spinal cord of MOG-immunized mice. n=7 per group. Graphs show mean ± s.d. N.S. not significant, *p < 0.05, ** p < 0.01, *** p < 0.001 (Prism, t-test).

Extended Data Figure 4

Influence of DDX5 on T cell phenotypes in autoimmune disease models

a, At 8 weeks after T cell transfer, LILP mononuclear cells were evaluated for amounts of IL-17A and IFNγ mRNA by qRT-PCR. Results are representative of two independent experiments. Each experiment was performed using large intestines from 3 animals in each condition. qRT-PCR was performed with two technical replicates. Graph shows mean ± s.d. * p<0.03 (Prism, t-test). b, Gating strategy for analysis of Th17 and Th1 cells from large intestine of Rag2-deficient recipients of WT or DDX5-Tko naïve T cells analyzed at 8 weeks after T cell transfer. c, Representative IL-17A and IFNγ intracellular staining of Aqua−CD4+RORγt+T helper 17 cells in spinal cord of MOG immunized animals on Day 21.

Function of DDX5-associated LncRNA

RNA helicases are highly conserved enzymes that utilize the energy derived from ATP hydrolysis to unwind RNA duplexes, facilitate RNA annealing, and displace proteins from RNA. It was previously shown that DDX5 transcriptional coactivator activity for estrogen receptor, androgen receptor, and RUNX2 is independent of RNA helicase activity [12,24,26]. We tested this requirement in the context of RORγt by retrovirally transducing DDX5-deficient T cells cultured under Th17 polarizing conditions with expression constructs for WT or mutant DDX5 with an inactivated helicase domain (helicase-dead). Surprisingly, only WT DDX5 rescued IL-17A and IL-17F production in these polarized Th17 cells (Fig. 3a–b and Extended Data Fig. 5a). This result suggested that perhaps RNA substrate(s) for the helicase activity of DDX5 contribute to its transcriptional coactivator role in Th17 cells.

Figure 3

Requirement for helicase-competent DDX5 and its associated lncRNA Rmrp in induction of Th17 cell cytokines

a, Cytokine production in DDX5-Tko cells transduced with WT or helicase-mutant DDX5 and subjected to sub-optimal Th17 cell polarization. b, Results from four independent experiments shown (a). c, IGV browser view of Rmrp showing coverage of mapped RNA reads from total Th17 lysate, Ribosome TRAP-seq (described in Methods), DDX5 RIP-seq, and RORγt RIP-seq. d, Effect of mouse Rmrp-specific ASO. Results are representative of three independent experiments with two technical replicates. e, IL-17A production following Rmrp knockdown in in vitro polarized human Th17 cells. Each symbol (right panel) represents cells from a healthy donor (n=5). Graphs show mean ± s.d. CTL, control; ** p<0.01, **** p<0.0001 (Prism, t-test).

Extended Data Figure 5

ncRNAs enriched in DDX5 and RORγt RIP-seq studies

a, DDX5-Tko cells were transduced with WT or helicase-mutant DDX5 and evaluated for DDX5 expression by immunofluorescence (left) and immunoblot (right) with anti-DDX5 antibody. For gel source data, see Supplementary Figure 1. b, Venn diagram of ncRNAs detected by deep sequencing following co-immunoprecipitation (RIP-seq) of ribosome-depleted Th17 cell lysates with anti-DDX5 and anti-RORγt antibodies. c, Abundance of top ncRNAs enriched in DDX5 and RORγt immunoprecipitates from polarized Th17 cell lysates depleted of ribosomes. Top panel indicates abundance of the ncRNAs in total lysate. d, Conventional RIP-qRT-PCR experiments to compare Rmrp association with DDX5 in Th17 and developing thymocytes. Results are representative of three independent experiments. Each experiment was performed with two technical replicates. Graph shows mean ± s.d. ** p<0.001 (Prism, t-test).

We next searched for RNA molecules that might participate in DDX5-RORγt-mediated transcription in Th17 cells. To this end, we first depleted ribosome-bound mRNAs undergoing active protein synthesis. Lysates pre-cleared of ribosomes were then subjected to immunoprecipitation with antibodies specific for DDX5 or RORγt, followed by deep sequencing of the associated RNAs (RIP-seq). Among 49,893 annotated lncRNAs in the mouse RefSeq and NONCODE database, 2,533 ncRNAs were expressed in Th17 cells (FPKM>1, Extended Data Fig. 5b). Interestingly, ncSRA, previously found to be associated with DDX5 in muscle cells [15], was not enriched in DDX5-containing protein complexes in Th17 cells. Instead, we found Rmrp to be the most enriched RNA associated with DDX5 and, to a lesser degree, RORγt, in Th17 cells (Fig. 3c and Extended Data Fig. 5c). Conventional RIP-qPCR with independent biological samples confirmed enrichment of Rmrp RNA in DDX5 pull-downs from Th17 cells, but not from thymocyte lysates (Extended Data Fig. 5d).

RNA FISH revealed that Rmrp is localized in the nucleus of Th17 cells (Extended Data Fig. 6a). To evaluate the functional role of Rmrp, we transiently depleted Rmrp RNA from primary murine Th17 cells using an RNAseH-dependent antisense oligonucleotide (ASO). Similar to the DDX5-deficient Th17 cells, cells depleted for Rmrp expressed reduced IL-17A and IL-17F mRNA (Fig. 3d and Extended Data Fig. 6b). Human Th17 cells also displayed reduced cytokine production upon depletion of Rmrp or DDX5 (Fig. 3E and Extended Data Fig. 6c), suggesting that this regulatory mechanism is evolutionarily conserved. Importantly, Rmrp knockdown in DDX5-deficient murine Th17 cells did not further reduce IL-17A and IL-17F expression (Fig. 4a). Expression of RORγt-dependent, but DDX5-independent, CCR6 was unaffected by the reduction in Rmrp. Transduction of Rmrp into T cells cultured in Th1 polarization conditions had little effect on IFNγ production, but there was marked enhancement of IL-17A and IL-17F production in WT, but not DDX5-Tko, cells cultured in Th17 polarization conditions (Fig. 4b and Extended Data Fig. 7a–b). Thus, Rmrp-dependent cytokine gene expression requires the presence of DDX5.

Extended Data Figure 6

Rmrp and DDX5 knockdown in mouse and human Th17 cells

a, RNA FISH analysis, using probes specific for Rmrp (green) and Malat1 (red) lncRNAs, in Th17 cells at 72h following nucleofection with control (CTL) or Rmrp ASOs. b, Effect of Rmrp ASOs targeting different regions of Rmrp transcript on levels of Rmrp, IL-17F, and CCR6 RNAs in polarized Th17 cells. c, Knockdown of DDX5 reduced IL-17A production in in vitro polarized human RORγt+ Th17 cells. ** p<0.01 (Prism, t-test). Representative result in left panel. Each dot represents a different healthy donor (n=4). Graphs show mean ± s.d.

Figure 4

Analysis of DDX5-dependent Rmrp function in Th17 cell differentiation

a,b, Cytokines in WT and DDX5-Tko in vitro polarized Th17 cells upon Rmrp knockdown (a) or overexpression (b). Representative of three independent experiments. c, Sequence of Rmrp (nucleotides 258–275) from WT and Rmrp littermates. d, Rmrp-dependent expression of a RORE-directed firefly luciferase reporter nucleofected into polarized Th17 cells at 72 h. Firefly and control Renilla luciferase activities were measured 24 h later. Each dot represents the result from one nucleofection. Results from two independent experiments. e, Top RORγt targets co-regulated by DDX5 and Rmrp. f, Proportion of CD4+Foxp3− T cells expressing RORγt (left) and numbers of Th1 (IFNγ+RORγt−Tbet+), Th17 (IL-17A+RORγt+Foxp3−), and Tγδ17 (Tγδ+RORγt+) cells (right) in ileum. Symbols represent cells from one animal. Graphs show mean ± s.d. ** p<0.01, ***p<0.001, ****P < 0.0001 (Prism, t-test).

Extended Data Figure 7

Effects of WT and mutant Rmrp in T cell differentiation

a, IL-17A mRNA in cell lysates of in vitro polarized murine Th17 cells at 96 h following transduction of control vector or WT Rmrp. Results are representative of two independent experiments. b, IFNγ production in polarized murine Th1 cells at 96 h after transduction of control or Rmrp-encoding vector. Representative of two independent experiments. Each experiment was performed with two technical replicates. c, Comparison of human and mouse Rmrp sequences. Several mutations identified in CHH patients are highlighted. d, IL-17A production in polarized murine Th17 cells at 96 h after transduction of WT or mutant Rmrp vectors. Representative of two independent experiments. e, The Venn diagram depicts number of distinct and overlapping genes regulated by RORγt, DDX5, and Rmrp in in vitro-polarized Th17 cells. f, Expression of cytokine and Foxp3 mRNAs in T cells from WT or Rmrp mice cultured ex vivo in Th17, iTreg, Th1 and Th2 polarizing conditions. Results are representative of two independent experiments. Each experiment was performed with two technical replicates. *** p<0.001 (Prism, t-test). g, ChIP-qPCR experiment using anti-RORγ/γt antibodies on chromatin of Th17 cells from WT or mutant animals cultured for 48 h in vitro. Each dot represents a different biological sample. WT, n=2; Rmrp, n=2. Result is representative of three separate independent experiments. Graphs show mean ± s.d. N.S. not significant (Prism, t-test).

Th17 program in Rmrp mutant mice

In contrast to wildtype Rmrp, a mutant Rmrp carrying a single nucleotide change (270 G>T), corresponding to an allele identified in CHH patients (262 G>T), failed to potentiate IL-17A production following transduction into Th17-polarized cells (Extended Data Fig. 7c–d). To ask whether G270 of Rmrp contributes to RORγt transcriptional output in vivo, we generated mice homozygous for the Rmrp G270T point mutation, using CRISPR-Cas9 technology (Fig. 4c). These animals were born at the expected Mendelian ratios and had no gross defects. ROR element-regulated luciferase activity was reduced in transiently transfected Th17 cells from DDX5-deficient and Rmrp mice and upon ASO-mediated knockdown of Rmrp (Fig. 4d). Comparison of transcription profiles of in vitro polarized Th17 cells from WT, RORγt-deficient, DDX5-deficient, and Rmrp mice indicated that 96 RORγt-dependent Th17 cell genes were co-regulated by Rmrp together with DDX5 (Extended Data Fig. 7e and Fig. 4e). RT-qPCR analysis of independent biological samples from in vitro polarized T cells from WT and Rmrp mice confirmed reduced IL-17F mRNA expression in the latter (Extended Data Fig. 7f), despite similar amount of RORγt binding to known cis-regulatory loci (Extended Data Fig. 7g). The proportion of RORγt+Foxp3− Th17 cells among total ileal LP CD4-lineage cells was unaffected in Rmrp animals, but these cells expressed relatively little IL-17A compared to those in wildtype littermates (Fig. 4f). Transfer of Rmrp T cells into Rag−/− mice resulted in reduced colitis, as determined by weight loss and colon histology, compared to transfer of wildtype cells (Extended Fig. 8a). These phenotypes are similar to what was observed in animals with a T cell-specific Ddx5 deletion (Fig. 2a–c), which is consistent with an important role of G270 of Rmrp in executing the Th17 effector program in vivo.

Extended Data Figure 8

Effect of Ddx5 and Rmrp mutations in inflammation and thymocyte development

a, Left panel: Percent weight change in Rag2 recipients of WT (black circles) or Rmrp (gray squares) naïve CD4+ T cells in the transfer model of colitis. Animal weight was measured on day 56. (WT: n=8; Rmrp: n=8, combined from three independent experiments). Graphs show mean ± s.d. *** p<0.001 (Prism, t-test). Middle panel: histology score (scale of 0–24) (WT: n=8; and Rmrp: n=5), combined from two independent experiments. ** p<0.01 (Prism, t-test). Right panel: representative H&E staining of large intestine of Rag2−/− animals on day 56 after naïve T cell transfer. b, Mice with deletion of Ddx5 in early common lymphoid progenitors have normal thymic development. Left: immunoblot of thymocyte lysates with anti-DDX5 antibody confirmed depletion of DDX5; right: percent CD4 single positive (SP), CD8α SP, double positive (DP), and double negative (DN) cells among total thymocyte. Each bar is result from one animal (wt/het, n=9; DDX5-clpKO, n=6). For gel source data, see Supplementary Figure 1. c, Thymocyte and peripheral T cell surface phenoptypes of WT and Rmrp knock-in mice at steady state. Peripheral T cell gate: DAPI−CD19−CD8α−CD4+.

RORγ/γt perform distinct functions in diverse tissues. RORγt is critical for thymocyte development, regulating survival of double positive cells, and for lymphoid tissue inducer cell-mediated development of secondary and tertiary lymphoid organs [31]. While DDX5 and Rmrp are ubiquitously expressed, Rmrp was less enriched in thymocyte-derived than in Th17 cell-derived DDX5 immunoprecipitates (Extended Data Fig. 5d). When Ddx5 was inactivated at the common lymphoid progenitor stage, by breeding the conditional mutant mice with IL7R–Cre mice, there was no apparent defect in thymocyte subset development (Extended Data Fig. 8b). Similarly, RmrpG270T knockin animals displayed normal thymocyte subsets and, additionally, had intact secondary lymphoid organ development (Extended Data Figure 8c). Together, these results suggest that the DDX5-Rmrp complex performs Th17-specific functions.

Rmrp in RORγt-DDX5 complex formation

We next asked how Rmrp contributes to the DDX5-RORγt-regulated transcriptional circuit in Th17 cells. RORγt-DDX5 complex assembly was severely compromised in Th17 cells harboring Rmrp (Fig. 5a). Moreover, Rmrp recruitment to the RORγt protein complex was significantly reduced in Th17 cells from Rmrp mutant animals (Fig. 5b). In vitro, Rmrp binds directly to recombinant DDX5 (Extended Data Fig. 9a). Intriguingly, Rmrp was recruited to RORγt in the presence of wildtype, but not helicase-dead, DDX5. Furthermore, in vitro transcribed Rmrp RNA promoted RORγt interaction with wildtype, but not helicase-dead, DDX5 in the presence of ATP (Fig. 5c and Extended Data Fig. 9b). Mutant Rmrp was also defective in mediating DDX5-RORγt complex assembly in vitro (Extended Data Fig. 9c–d).

Figure 5

Rmrp localization at RORγt-occupied genes and role in RORγt-DDX5 assembly

a, RORγt association with immunoprecipitated (IP) DDX5 in polarized Th17 cells. IB, immunoblot. Representative of three independent experiments. b, Rmrp quantification by qRT-PCR in RORγt immunoprecipitates from polarized Th17 cells. Representative of two independent experiments with two technical replicates. c, Rmrp requirement for ATP-dependent in vitro interaction of recombinant GST-DDX5 and His-RORγt. Representative of three independent experiments. For gel source data (a,c), see Supplementary Figure 1. d, Rmrp occupancy at RORγt genomic target loci in polarized Th17 cells. Rmrp ChIRP-qPCR amplicons (bottom) are indicated in IGV browser view of RORγt ChIP at the Il17 locus (top). Data from 2–4 experiments with two technical replicates. e, Model for DDX5-Rmrp complex recruitment to RORγt-occupied chromatin in Th17 cells. Graphs show mean ± s.d. ** p<0.01 (Prism, t-test).

Extended Data Figure 9

Association of Rmrp lncRNA with DDX5 and RORγt in vitro

a, In vitro translated HA-tagged wildtype or helicase-dead DDX5 and FLAG-tagged RORγt were incubated with in vitro transcribed Rmrp. After capture on anti-HA or anti-FLAG beads, the amount of lncRNA was determined by qRT-PCR. Data are representative of two independent experiments. Each experiment was performed with two technical replicates. Graphs show mean ± s.d. *** p<0.001 (Prism, t-test). b, Helicase requirement for in vitro interaction of DDX5 and RORγt. Recombinant GST-DDX5 (wildtype or helicase-dead mutant) and His-RORγt full-length protein were synthesized in E. coli, purified, and assayed for binding with or without in vitro transcribed Rmrp RNA in the presence exogenous ATP. For gel source data, see Supplementary Figure 1. c, Association of in vitro transcribed WT and mutant Rmrp with recombinant GST-DDX5 captured on glutathione beads (left) or with recombinant GST-DDX5 and His-RORγt captured with anti-His antibody. Amounts of associated Rmrp were quantified using qRT-PCR. Data are representative of two independent experiments. Each experiment was performed with two technical replicates. Graphs show mean ± s.d. *** p<0.001 (Prism, t-test). d, Comparison of ability of in vitro transcribed WT and RmrpG270T lncRNA to promote interaction between recombinant RORγt and DDX5 in vitro. For gel source data, see Supplementary Figure 1.

To determine if Rmrp is associated with specific genomic loci, we performed chromatin isolation by RNA purification (ChIRP) followed by either locus-specific quantitative PCR or by deep sequencing (ChIRP-seq) [32]. We employed two orthogonal antisense probe sets that specifically and robustly recovered Rmrp from Th17 cells (Extended Data Fig. 10a). When combined for Rmrp ChIRP-qPCR, the probes recovered RORγt-bound regions in the Il17a and Il17f loci from Th17 polarized cells of wild type but not DDX5Tko or Rmrp mice, in an RNase-sensitive manner (Fig. 5d and Extended Data Fig. 10b). For ChIRP-seq, we focused our analysis on signals that overlapped with use of the two probe sets. HOMER motif analyses of Rmrp peak regions identified the ETS, DR2/RORE, and AP1 transcription factor (TF) motifs to be the most highly enriched (Extended Data Fig. 10c). Consistent with this, Rmrp ChIRP-seq significantly overlapped with RORγt-bound loci, but not with sites occupied by CTCF or by other Th17 TFs, such as BATF, IRF4, STAT3, and c-Maf (Extended Data Fig. 10d). There was also significant overlap with RNAPol-II- and H3K4me3-associated chromatin, which mark actively transcribed regions. Concordantly, ChIRP-seq of Rmrp in DDX5-Tko Th17 cells revealed a loss of called Rmrp peaks despite similar amount of RNA recovery (Extended Data Fig. 10e), consistent with a DDX5 contribution to Rmrp assocation with chromatin. Rmrp association with RORγt bound sites was also reduced in polarized Th17 cells from Rmrp mice (Extended Data Fig.10f). Together, these results indicate that G270 of Rmrp is critical for DDX5-RORγt complex assembly and Rmrp recruitment to RORγt-occupied loci to coordinate the Th17 effector program in trans.

Extended Data Figure 10

Rmrp chromatin localization in Th17 cells

a, ChIRP-seq sample validation of Rmrp RNA pull-down over other nuclear ncRNAs using pools of Even or Odd capture probes. Graphs show mean ± s.d. b, ChIRP-qPCR of Rmrp RNA pull-down from WT Th17 cell lysate treated with or without RNase (n=2). qPCR for each sample was performed with two technical replicates. Graph shows mean ± s.d. ** p<0.001 (Prism, t-test). c, HOMER motif analysis reveals top three DNA motifs within Rmrp-enriched peaks. d. Significance of peak overlaps between Rmrp ChIRP-seq and ChIP-seq for BATF (n=2), IRF4 (n=7), STAT3 (n=2), c-Maf (n=2), RORγt (n=2), CTCF (n=2), RNA Pol II (n=2), H3K27me3 (n=4), and H3K4me3 (n=3) in Th17 cells (hypergeometric distribution). Each dot represents a separate biological replicate of ChIP-seq experiments. e, Venn Diagram depicting changes in peaks called from Rmrp (ChIRP-seq) experiments in WT and DDX5-Tko Th17 cells. f, Comparison of Rmrp chromatin occupancy (ChIRP-seq) at known RORγt occupied loci in in vitro-polarized Th17 cells from WT and Rmrp mice.

Discussion

Nuclear lncRNAs have key roles in numerous biological processes [33], including adaptive and innate immunity [34],[35], but how individual lncRNAs perform their activities and whether they contribute to immunological diseases remain unknown. We identified nuclear Rmrp as a key DDX5-associated RNA required to promote assembly and regulate the function of RORγt transcriptional complexes at a subset of critical genes implicated specifically in the Th17 effector program (model in Fig. 5e). Rmrp thus acts in trans on multiple RORγt-dependent genes, and does so only upon interaction with enzymatically active DDX5 helicase. RNA helicase-dependent functions of lncRNAs have been described, e.g. the Drosophila male cell-specific lncRNAs roX1 and roX2 that are modified by the MLE helicase to enable expression of X-chromosome genes [36,37]. In addition, DDX21 helicase activity in HEK293 cells is required for 7SK RNA regulation of polymerase pausing at ribosomal genes [38]. Our results extend the concept of RNA helicase/lncRNA function to lineage-specific regulation of transcriptional programs.

It is noteworthy that, unlike most lncRNAs, Rmrp is highly conserved among mammals. In humans, mutations of evolutionarily conserved nucleotides at the promoter or within the transcribed region of RMRP result in Cartilage-Hair Hypoplasia (CHH) [21,22]. T cells from mice carrying a single nucleotide change (270 G>T) in Rmrp, corresponding to one found in CHH patients (262 G>T), had a compromised Th17 cell effector program. CHH patients have been noted to have defective T cell-dependent immunity, which may reflect, at least in part, reduced Rmrp-dependent activity at RORγt target genes. Since forced expression of either DDX5 or Rmrp enhanced Th17 cytokine production, it is also possible that gain-of-function mutations in either of these molecules may contribute to Th17-dependent inflammatory diseases.

RORγt is an attractive therapeutic target for multiple autoimmune diseases [5,39]. However, RORγt and its closely-related isoform RORγ have multiple other functions that would likely be affected by targeting of their shared ligand-binding pocket. RORγt is required for the development of early thymocytes, lymphoid tissue inducer (LTi) cells that initiate lymphoid organogenesis [31], type 3 innate lymphoid cells (ILC3) that produce IL-22 and protect epithelial barriers, and for IL-17 production by “innate-like” T cells, including TCRγδ and NKT cells [40-43]. In the liver, RORγ contributes to regulation of metabolic functions [44]. Mechanisms by which RORγ/γt differentially regulates transcription in these diverse cell types remain poorly understood. DDX5 and Rmrp are abundantly expressed in developing T cells in the thymus and in peripheral naïve and differentiated T helper subsets (Huang, unpublished). Intriguingly, the contribution of DDX5-Rmrp to RORγt-dependent functions appears to be confined to Th17 cells, as their loss of function did not affect thymocyte or lymphoid organ development. Our results raise the prospect that tissue- or cell type-specific mechanisms exist to regulate how RNA helicases and their associated lncRNAs are assembled with distinct transcriptional complexes to promote diverse gene expression programs.

We speculate that the function of DDX5-Rmrp may be induced in response to specific tissue microenvironments in vivo. Th17 cells differentiate at mucosal barriers in response to signals from the microbiota, and upregulate their expression of IL-17A locally [45,46]. Regional signals may induce DDX5/Rmrp association with RORγt, resulting in the transcriptional activation of multiple loci that endow Th17 cells with effector functions[15]. Our finding that DDX5 was required for the differentiation of “pathogenic” Th17 cells [2,29,30] suggests that strategies to interfere with this function may be of therapeutic benefit. A better understanding of this transcriptional regulatory system may provide new approaches for therapeutic intervention in autoimmune diseases and immune deficiencies in CHH patients.

Methods

Mice

EF1a: Lox-stop-lox-GFP-L10, RORγ/γt-deficient animals, and Ddx5 mice were previously described [47-49]. Conditional mutant mice were bred to CD4-Cre transgenic animals (Taconic) and maintained on the C57BL/6 background. We bred heterozygous 6–8wk old mice to yield Ddx5+/+CD4Cre+ (subsequently referred to as WT) and Ddx5fl/flCD4Cre+ (referred to as DDX5-Tko) littermates for experiments examining DDX5 in peripheral T cell function. DDX5 conditional mutant mice were also bred to IL7R–Cre transgenic animals (Jackson Laboratory) for experiments examining DDX5 during T cell development in the thymus. Rmrp knockin mice were generated using CRISPR-Cas9 technology by the Rodent Genetic Engineering Core (RGEC) at NYULMC. Guide RNA and HDR donor template sequences are provided in Supplementary Table 1. Heterozygous crosses provided Rmrp+/+ (WT) and Rmrp littermates for in vivo studies. All animal procedures were in accordance with protocols approved by the Institutional Animal Care and Use Committee of the NYU School of Medicine (Animal Welfare Assurance Number: A3435-01).

In vivo studies

Steady state small intestines were harvested for isolation of lamina propria mononuclear cells as described [50]. For detecting SFB colonization, SFB-specific 16S primers were used [51]. Universal 16S and/or host genomic DNA were quantified simultaneously to normalize SFB colonization of each sample. All primer sequences are listed in Supplementary Information Table.

For the adoptive transfer model of colitis, 5×105 CD4+CD45RBhiCD62LhiCD44lowCD25− T cells were isolated from murine splenocytes by FACS sorting and administered intraperitoneally (i.p.) into Rag2−/− mice as previously described [52]. Animal weights were measured approximately weekly. At the time of sacrifice (between weeks seven and eight), large intestines were harvested for H&E staining and isolation of lamina propria mononuclear cells as described [50]. The H&E slides from each sample were examined in a double-blind fashion. The histology scoring (scale 0–24) was based on the evaluation of criteria described previously [53].

For induction of active experimental autoimmune encephalomyelitis (EAE), mice were immunized subcutaneously on day 0 with 70 µg of MOG 35–55 peptide, emulsified in CFA (Complete Freund’s Adjuvant supplemented with 200 mg/mL Mycobacterium tuberculosis), and injected i.p. on days 0 and 2 with 100 ng/mouse of pertussis toxin (Calbiochem). The EAE scoring system was as follows: 0-no disease, 1-limp tail; 2-weak/partially paralyzed hind legs; 3-completely paralyzed hind legs; 4-complete hind and partial front leg paralysis; 5-complete paralysis/death.

In transfer colitis and EAE experiments, animals of different genotypes were cohoused and were weighed and scored blindly. For statistical power level of 0.8, probability level of 0.05, anticipated effect size of 2, minimum sample size per group for two-tailed hypothesis is 6. Two-tailed unpaired Student’s t-test was performed using Prism (GraphPad Software). We treated a P-value of less than 0.05 as a significant difference. *; p<0.05, **; p<0.001, and ***; p<0.005. All experiments were performed at least twice.

In vitro T cell culture and phenotypic analysis

Mouse T cells were purified from lymph nodes and spleens of six to eight week old mice, by sorting live (DAPI−), CD8−CD19−CD4+CD25−CD62L+CD44low/Int naïve T cells using a FACSAria (BD). Detailed antibody information is provided in Supplementary Table 1. Cells were cultured in IMDM (Sigma) supplemented with 10% heat-inactivated FBS (Hyclone), 50 U penicillin-streptomycin (Invitrogen), 4 mM glutamine, and 50 µM β-mercaptoethanol. For T cell polarization, 200 µl of cells was seeded at 0.3 × 10^5 cells per mL in 96-well plates pre-coated with goat anti-hamster IgG at a 1:20 dilution of stock (1mg/ml, MP Biomedicals Catalog # 55398). Naïve T cells were activated with anti-CD3ε (2.5 µg/mL) and anti-CD28 (10 µg/mL). Cells were cultured for 4−5 days under Th17 polarizing conditions (0.1–0.3 ng/mL TGF-β, 20 ng/mL IL-6), Th1 (10 ng/mL IL-12, 10 U/mL IL-2, and 2 ng/mL anti-IL-4), Th2 (10 ng/mL IL-4), or Treg conditions (5 ng/mL TGF-β, 10 U/mL IL-2).

Human T cells were isolated from peripheral blood of healthy donors using anti-human CD4 MACS beads (Miltenyi). Human CD4 T cells were cultured in 96-well U bottom plates in 10 U/ml of IL-2, 10 ng/mL of IL-1β, 10 ng/ml of IL-23, 1 µg/ml of anti-IL-4, 1 µg/mL of anti-IFNγ and anti-CD3/CD28 activation beads (LifeTechnologies) at a ratio of 1 bead per cell, as previously described [54].

For cytokine analysis, cells were incubated for 5 h with phorbol PMA (50 ng/mL; Sigma), ionomycin (500 ng/mL; Sigma) and GolgiStop (BD). Intracellular cytokine staining was performed according to the manufacturer’s protocol (Cytofix/Cytoperm buffer set from BD Biosciences and FoxP3 staining buffer set from eBioscience). A LSR II flow cytometer (BD Biosciences) and FlowJo (Tree Star) software were used for flow cytometry and analysis. Dead cells were excluded using the Live/Dead fixable aqua dead cell stain kit (Invitrogen).

Nucleic acid reagents and T cell transduction

Custom Rmrp and predesigned Malat1 Stellaris RNA FISH probes were purchased from BiosearchTech and used to label mRmrp and mMalat1 RNA in cultured Th17 cells according to the manufacturer’s protocol. Control and human DDX5 specific siRNAs (#8626) were obtained from Cell Signaling. Synthesis of ASOs was performed as previously described [55]. All ASOs were 20 nts in length and had a phosphorothioate backbone. The ASOs had five nucleotides at the 5’ and 3’ ends modified with 2’-O-methoxyethyl (2’-MOE) for increased stability. ASOs and siRNA sequences are provided in Supplementary Table 1. siRNA and ASOs were introduced into murine Th17 cells by Amaxa nucleofection as previously described [56].

WT and helicase dead mutant DDX5 were described previously [57]. DDX5 and Rmrp were subcloned into the MSCV Thy1.1 vectors for retroviral overexpression and rescue assays in T cells. Retrovirus production was carried out in PlatE cells as described [58]. Spin transduction was performed 24h after in vitro T cell activation by centrifugation in a Sorvall Legend RT at 2000rpm for 90min at 32° C. Aqua−Thy1.1+ live and transduced cells were analyzed by flow cytometry after 5 days of culture in Th17 polarizing conditions.

RORγt Transcriptional Activity in Polarized Th17 Cells

A ROR luciferase reporter was constructed with 4 RORE sites replacing the Gal4 (UAS) sites from the pGL4.31 vector (luc2P/GAL4 UAS/Hygro) from Promega (C935A) as described in [59]. Naïve CD4+ T cells were cultured in Th17 polarizing conditions for 72h. Nucleofection (Amaxa Nucleofector 4D, Lonza) was then used to introduce 1 µg RORE-firefly luciferase reporter construct and 1 µg control renilla luciferase construct according to the manufacturer’s instructions. Luciferase activity was measured using the dual luciferase reporter kit (Promega) at 24h after transfection. Relative luciferase units (RLU) were calculated as a function of firefly luciferase reads over those of renilla luciferase.

Co-immunoprecipitation and mass spectrometry

100×106 cultured Th17 cells were lysed in 25mM Tris (pH 8.0), 100mM NaCl, 0.5% NP40, 10mM MgCl2, 10% glycerol, 1X protease inhibitor and PhosphoSTOP (Roche) on ice for 30min, followed by homogenization with a 25g needle. The RORγ/γt-specific antibody used for pull down assays was previously described [56]. Co-immunoprecipitated complexes were harvested with Protein G Dynabeads (Dynal, Invitrogen). Detailed antibody information is provided in Supplementary Table 1. Mass spectrometry and the Mascot database search to identify protein complex composition were both performed by the Central Proteomics Facility at the Dunn School of Pathology, Oxford, UK.

Ribosome TRAP-seq, RIP-seq, and RNA-seq

20 million cells cultured in Th17 polarizing conditions for 48 h were lysed in 10 mM HEPES (pH 7.4), 150 mM KCl, 0.5 mM dithiothreitol (DTT), 100 ug/ml cycloheximide, 1% NP40, 30 mM DHPC, 1X protease inhibitor, and PhosphoSTOP (Roche). Ribosome-TRAP IP was first performed using 2 µg of anti-GFP antibody (Invitrogen) and harvested in 20 µl of Protein G magnetic Dyna beads. The supernatant was removed for subsequent RIP pull down using anti-DDX5 (Abcam) or anti-RORγt antibodies and harvested with Protein G Dyna beads. TRAPseq samples were washed with high-salt wash buffer (10 mM HEPES pH 7.4, 350 mM KCl, 5 mM MgCl2, 1% NP-40, 0.5 mM dithiothreitol (DTT), and 100 ug/ml cycloheximide). RIPseq samples were washed three times with 25mM Tris (pH 8.0), 100mM NaCl, 0.5% NP40, 10mM MgCl2, 10% glycerol, 1X protease inhibitor, and PhosphoSTOP (Roche). Enrichment of target proteins was confirmed by immunoblot analysis. Complementary DNAs (cDNAs) were synthesized from TRIzol (Invitrogen) isolated RNA, using Superscript III kits (Invitrogen). RNA-seq libraries were prepared and sequenced at Genome Services Laboratory, HudsonAlpha. Sequencing reads were mapped by Tophat and transcripts called by Cufflinks. Pulldown enrichment was calculated for each transcript as a ratio of FPKM recovered from TRAP-seq and RIP-seq samples compared to those from 5% input.

For RNA-seq analysis, volcano scores for WT, DDX5-Tko, and RORγt knockout Th17 cells were calculated for each transcript as a function of its p-value and fold-change between mutant and WT controls. BAM files were converted to tdf format for viewing with the IGV Browser Tool. Ingenuity Pathway Analysis was used to identify enriched Gene Ontology (GO) terms in the DDX5-RORγt co-regulated gene set.

ChIRP-seq and ChIRP-qPCR

The ChIRP-seq assay was performed largely as described previously [60]. Mouse Th17 cells were cultured as above and in vivo RNA-chromatin interactions were fixed with 1% glutaraldehyde for 10 minutes at 25°C. Anti-sense DNA probes (designated odd or even) against Rmrp were designed by Biosearch Probe Designer (#1: taggaaacaggccttcagag, #2: aacatgtccctcgtatgtag, #3: cccctaggcgaaaggataag, #4: aacagtgacttgcgggggaa, #5: ctatgtgagctgacggatga). Probes modified with BiotinTEG at the 3-prime end were synthesized by Integrated DNA Technologies (IDT). Isolated RNA was used in qRT-PCR analysis (Stratagene) to quantify enrichment of RMRP and depletion of other cellular RNAs. Isolated DNA was used for qPCR analysis or to make deep sequencing libraries with the NEBNext DNA Library Prep Master Mix Set for Illumina (NEB). Library DNA was quantified on the High Sensitivity Bioanalyzer (Agilent) and sequenced from a single end for 75 cycles on an Illumina NetSeq 500.

Sequencing reads were first trimmed of adaptors (FASTX Toolkit) and then mapped using Bowtie to a custom bowtie index containing single-copy loci of repetitive RNA elements (rRNA, snRNAs, and y-RNAs [61]). Reads that did not map to the custom index were then mapped to mm9. Mapped reads were separately shifted towards the 3’ end using MACS and normalized to a total of 10 million reads. Even and Odd replicates were merged as described previously[60] by taking the lower of the two read density values at each nucleotide across the entire genome. These processing steps take raw FASTQ files and yield processed files that contain genome-wide RMRP-chromatin association maps, where each nucleotide in the genome has a value that represents the relative binding level of the RMRP RNA. MACS parameters were as follows: band width = 300, model fold = 10,30, p-value cutoff = 1.00e-05. The full pipeline is available at https://github.com/bdo311/chirpseq-analysis.

ChIRP-qPCR was performed on DNA purified after treatment with RNase (60 min, 37°C) and proteinase K (45min, 65°C). The primers used for qPCR can be found in Supplementary Information Table 1. For enrichment analysis, we tested for the enrichment of Rmrp ChIRP peaks among ChIP peak sets for key Th17 transcription factors, CTCF, RNA pol II, and several histone marks (Ciofani et al., 2012). ATACseq, according to published protocol[62], was performed on cultured Th17-polarized cells in vitro for 48h (unpublished). Because of differences in ChIP antibody affinities and the bias in the selection of ChIP and ChIRP factors, we used peaks generated from ATACseq data as a background set for the enrichment analysis. In our analysis, we considered all ChIRP and ChIP peaks that fell within +/−500bp of ATACseq peaks, and then calculated the overlap among the ChIRP and ChIP sets, using the hypergeometric distribution to estimate significance.

In vitro binding assay

For in vitro binding assays, pcDNA3.1-Rmrp vectors were used for T7 polymerase-driven in vitro transcription reactions (Promega). HA-DDX5 and FLAG-RORγt were in vitro transcribed and translated using an in vitro transcription and translation (TNT) system according to the manufacturer’s protocol (Promega). Alternatively, pGEX4.1-DDX5 (wildtype and helicase-dead mutant) constructs were transformed into BL21 to synthesize recombinant full-length GST-hDDX5 proteins. Full-length His-tagged human RORγt was purified in three steps through Ni-Resin, S column, and gel-filtration on AKTA. 0.5µg of each recombinant protein was incubated in the presence or absence of 200 µM ATP, 300ng in vitro transcribed Rmrp in coIP buffer containing 25mM Tris (pH 8.0), 100mM NaCl, 0.5% NP40, 10mM MgCl2, 10% glycerol, 1X protease inhibitor, RNaseInhibitor (Invitrogen), and PhosphoSTOP (Roche). GST-DDX5 was enriched on glutathione beads (GE), HA-DDX5, FLAG-RORγt, and His-RORγt were enriched using anti-HA, (Covance), anti-FLAG (Sigma), and anti-His antibodies (Santa Cruz Bio) coupled to Anti-Mouse immunoglobulin Dynabeads (Dynal, Invitrogen).

Microscopy

Th17 cells were cultured on glass coverslips for 48 h and fixed in 4% paraformaldehyde in PBS for 5min in room temperature. Fixed cells were permeabilized with 0.1% BSA, 0.1% Triton, 10% normal serum in PBS for 1 h. Cells were then incubated with primary antibodies (DDX5, Abcam or RORγt, eBiosciences) in 0.1% BSA, 0.2% Triton PBS overnight at 4°C. Secondary antibodies (anti-goat-Alexa 488 or anti-rat-Alexa647, Molecular Probe) were incubated in 4°C for 1 h. Stained cells were washed three times with 0.5% tween, 0.1% BSA in PBS. DAPI was used to stain for DNA inside the nucleus. Immunofluorescence images were captured on a Zeiss 510 microscope at 40x.

ChIP and RT-qPCR analysis

Th17 polarized cells were crosslinked with 1% paraformaldehyde (EMS) and incubated with rotation at RT. Crosslinking was stopped after 10 min with glycine to a final concentration of 0.125M and incubated 5 min further with rotation. Cells were washed with ice cold PBS 3X and pellets were either flash frozen in liquid N2 or immediately resuspended in Farnham Lysis buffer (5mM PIPES, 85mM KCl, 0.5% NP-40). Hypotonic lysis continued for 10 min on ice before cells were spun down and resuspended in RIPA buffer (1X PBS, 1% NP-40, 0.5% SDS, 0.5% Na-deoxycholate), transferred into TPX microtubes, and lysed on ice for 30 minutes. Nuclear lysates were sonicated for 40 cycles of 30 sec ON and 30 sec OFF in 10 cycle increments using a Biorupter (Diadenode) at high setting. After pelleting debris, chromatin was precleared with protein G dynabeads (Dynabeads, TFS) for 2h with rotation at 4°C. For immunoprecipitation, precleared chromatin was incubated with anti-RORγt antibodies (1µg per 2 million cells) overnight with rotation at 4°C and protein G was added for the final 2h of incubation. Beads were washed and bound chromatin was eluted. ChIP-qPCR was performed on DNA purified after treatment with RNase (30 min, 37°C) and proteinase K (2h, 55°C) followed by reversal of crosslinks (8–12h, 65°C). The primers used for qPCR were described in [63].

For analysis of mRNA transcripts, gene specific values were normalized to the GAPDH housekeeping gene for each sample. All primer sequences are listed in Supplementary Information Table.