IL-4 Inhibits the Biogenesis of an Epigenetically Suppressive PIWI-Interacting RNA To Upregulate CD1a Molecules on Monocytes/Dendritic Cells Free

Monocytes derived from bone marrow progenitors are mononuclear phagocytes with the capacity to differentiate into macrophages and, subsequently, dendritic cells (DCs), and they play a critical role in the immune response (1). Two distinct populations of “inflammatory” and “patrolling” monocytes were defined in mice (2). In humans, based on different functional properties, CD14⁺CD16⁻ and CD14⁺CD16⁺ monocytes are classified into inflammatory subsets. In contrast, CD14^dimCD16⁺ monocytes are classified into patrolling subsets (3). Monocytes can be induced to DCs in vitro with several cytokines, including GM-CSF and IL-4 (4). DCs bridge the innate- and adaptive-immune responses by a series of steps to capture, process, and present Ags to Ag-specific T cells (5). They are divided into four subsets: conventional DCs that are derived directly from bone marrow precursors and have a short half-life, plasmacytoid DCs that can secrete a large amount of type I IFN after viral challenge, Langerhans cells, and monocyte-derived DCs (6). Many functional markers are manipulated by cytokines during the differentiation into DCs (7).

PIWI-interacting RNAs (piRNAs) are small noncoding RNAs that are predominantly expressed in germline and specifically interact with PIWI proteins (8). piRNAs are widely known for their function to silence the transposon elements and maintain the stability of the entire genome (9). However, additional evidence indicates that piRNAs not only existed in germline but also appeared in somatic cells, such as ovarian somatic cells of Drosophila, neuron cells of Aplysia, and human cancer cells (10–12). In the somatic cells of Drosophila, the piRNA/PIWI complex recruits epigenetic factors to regulate histone modification of target sites (13). In the neuron cells of Aplysia, the piRNA/PIWI complex mediates the DNA methylation of CpG islands in the CREB2 promoter to facilitate long-term memory (11). In the early embryo of Drosophila, the piRNA pathway is involved in the degradation of nos mRNA through CCR4-mediated deadenylation (14). Although some piRNAs play roles in cell proliferation and viability, the mechanism remains to be clarified in human cancer cells (12).

There is little information on how piRNA functions in the human immune system. In this study, we demonstrate that the highly expressed tRNA-derived piRNA (td-piR) in monocytes significantly repressed CD1A transcription by inducing H3K9 methylation at its promoter region. We also discovered that its biogenesis was controlled by IL-4. Our work suggests that piRNA could function as a mediator for signal transduction in highly differentiated immune cells.

PBMCs were isolated from healthy human donors through Ficoll gradient centrifugation. CD14⁺ monocytes were isolated with BD IMag anti-human CD14 magnetic particles, according to the magnetic labeling protocol supplied by the manufacturer. The cells were cultured in RPMI 1640 medium (Invitrogen) supplemented with 10% FBS (Life Technologies), 100 U/ml penicillin, and 100 U/ml streptomycin (HyClone). GM-CSF (1000 U/ml) and IL-4 (500 U/ml) were used to induce the differentiation of monocytes into DCs for ∼7 d, and the medium was changed every 3 d. Lipofectamine RNAiMAX was used to transfect various small RNAs into monocytes/DCs following the manufacturer’s instructions (Invitrogen). HEK293T cells were obtained from American Type Culture Collection and grown at 37°C in DMEM supplemented with 10% FCS (Life Technologies), 100 U/ml penicillin, and 100 U/ml streptomycin (HyClone). Lipofectamine 2000 (Invitrogen) was used to transfect plasmids and small RNAs into HEK293T cells.

The coding sequence regions of PIWIL4, SUV39H1, SETDB1, and heterochromatin protein 1 (HP1)β, tagged with hemagglutinin (HA) or FLAG, were obtained by RT-PCR, with the mRNA of human PBMCs as the template, and then inserted separately into the pcDNA3.1 vector. The accuracy of all clones was confirmed by DNA sequencing.

The T cell line CD8-2 was established as described previously, with minor modifications (15). In brief, CD8⁺ T cells were isolated from a random healthy donor with a BD Human CD8 T Lymphocytes Enrichment Set, according to the protocol supplied by the manufacturer, and cultured with an organic extract of Mycobacterium tuberculosis in the presence of an equal number of monocytes that had been treated with 500 U/ml GM-CSF and 300 U/ml IL-4 for 3 d to induce CD1 expression. Cultures were restimulated every 12 d with autologous CD1a⁺ monocytes (three times) and thereafter with allogeneic CD1a⁺ monocytes (once).

To measure mRNA levels, total RNA was reverse transcribed with oligonucleotide primer, and a reverse transcription quantitative PCR (RT-qPCR) assay was performed with primers specific to different genes, according to the supplier’s recommendations (SYBR Premix ExTaq; Takara), on a CFX96 Real-Time System (Bio-Rad). The expression of these genes was normalized to GAPDH mRNA. To measure the level of tRNA-Glu–derived piRNA [td-piR(Glu)], total RNA was reverse transcribed with a reverse transcription primer for U6 and a specific stem-loop reverse transcription primer to td-piR(Glu), and following quantitative PCR was performed with primers specific to td-piR(Glu) and U6. The expression of td-piR(Glu) was normalized to U6. Quantitative PCR data were processed using Bio-Rad CFX Manager (Version1.5).

Total RNA was extracted with TRIzol reagent (Invitrogen), and the concentration was determined using a NanoDrop 2000C (Thermo).The same amount of RNA from different cells was separated on 10% Urea-PAGE gel. After transferring onto a nitrocellulose membrane, a γ-³²P–labeled oligonucleotide probe completely complementary to td-piR(Glu) was hybridized overnight at 50°C. The membrane was washed three times for 10 min with washing buffer (stringent and nonstringent buffer) at 50°C. The phosphor imaging system was used to detect the signal.

HEK293T cells were collected and lysed with immunoprecipitation lysis buffer after they were transfected with pcDNA3.1-based constructs containing interested genes using Lipofectamine 2000 (Invitrogen) for 48 h. The lyses were incubated with 40 μl of anti-HA beads (Sigma) from 4 h to 6 h at 4°C, then the bead–protein mixtures were washed three times with cold lysis buffer. The immunoprecipitated samples were analyzed by Western blotting with primary Abs (anti-HA, anti-FLAG, or anti-PIWIL4 Abs) followed by secondary Abs (goat anti–mouse IRDye 680, goat anti–rabbit IRDye 800, or HRP-conjugated anti–mouse IgG). The blotting was analyzed with the Odyssey infrared imaging system or electrochemiluminescent detection system.

RNA-binding protein immunoprecipitation (RIP) was performed with the EZ-Magna RIP Kit (Millipore), according to the manufacturer’s instructions, with modified procedures. Approximate 1 × 10⁷ human monocytes were collected and fixed in 1% formaldehyde for 10 min at room temperature, and the unreacted formaldehyde was quenched with glycine for 5 min at room temperature. After washing the fixed cells three times with cold PBS, 500 μl RIP lysis buffer containing protease inhibitor mixture was used to resuspend the cell pellet. The lysate was incubated on ice for 5 min and stored at −80°C to complete the lysis process. The magnetic beads were washed with RIP washing buffer twice and incubated with different Abs (mouse IgG, rabbit IgG, anti-human PIWIL1, and anti- human PIWIL4) for 30 min at room temperature separately. Then the magnetic beads were washed twice with RIP washing buffer. RIP buffer containing RNase inhibitor and 20 mM EDTA was add to the magnetic beads. RIP lysate was thawed and centrifuged at 14,000 rpm for 10 min at 4°C. Then, 100 μl supernatant was added to the beads-Ab complex in RIP buffer. All of the tubes were incubated with rotation for 5–6 h at 4°C. The immunoprecipitation tubes were centrifuged briefly and placed on the magnetic separator, and the supernatant was discarded. The beads-Ab-protein-RNA complex was washed with 500 μl cold RIP washing buffer three times. RNA in the immunoprecipitation complex was extracted with TRIzol reagent after the protein digestion by proteinase K. Finally, the amount of RNA in the beads-Ab-protein-RNA immunoprecipitation complex was determined by qPCR.

Chromatin immunoprecipitation (ChIP) was performed with the EZ-Magna ChIP A/G Kit (Millipore), according to the manufacturer’s instructions. For each ChIP, ∼1.5 × 10⁷ cells were collected and fixed in 1% formaldehyde for 10 min at room temperature, and glycine was added to quench the unreacted formaldehyde for 5 min at room temperature. After washing the fixed cells three times with cold PBS, cell lysis buffer containing protease inhibitor mixture was used to resuspend the cell pellet, and the suspension was incubated on ice for 15 min, followed by centrifugation at 800 × g at 4°C for 5 min. The cell pellet was resuspended in nucleic lysis buffer and sonicated to shear the DNA into 500–1000-bp fragments. After centrifugation at 10,000 × g at 4°C for 10 min to remove insoluble materials, 5 μl the supernatant was taken out as “input,” and 50 μl was used for immunoprecipitation with different Abs and protein A/G magnetic beads for ∼8 h at 4°C. The protein A/G bead–Ab-chromatin complex was washed with a low-salt/high-salt/LiCl washing buffer and TE buffer. Subsequently, the protein-DNA complex was eluted, and the free DNA was obtained through reverse cross-linking. Finally, the free DNA was purified using spin columns and subjected to qPCR.

The organic extract was purified from whole, lyophilized Mycobacterium tuberculosis strain H37Ra. To obtain the lipid Ag, we followed the protocol described previously, with minor modifications (15). For the extraction of lipid Ags from M. tuberculosis strain H37Ra, bacteria were grown in Middlebrook 7H9 medium. After 3–4 wk, the bacteria were collected and extracted with 2:1 (v/v) chloroform/methanol at 20°C for 2 h to yield solutions of mycobacterial lipids, which were dried in nitrogen. The residue was resuspended in acetone at 4°C with 2% (v/v) of 10% (m/v) MgCl₂ in methanol for 1.5 h to precipitate anionic phospholipids. The supernatants were dried, and the residue was redissolved in 100:25 hexanes/chloroform and loaded onto silica solid-phase extraction columns (Sigma). These were eluted with hexane/chloroform/isopropanol/acetic acid solutions at ratios from 100:25:0:0 to 50:75:15:1, followed by pure methanol. The antigenic activity was found in the 50:75:15:1 solutions.

Several Abs were used for flow cytometry analysis: anti-CD40 (5C3), anti-CD80 (2D10.4), anti-CD86 (IT2.2), anti-CD209 (eBh209), anti–HLA-DR (LN3), anti-CD1a (HI149), anti-CD4 (OKT4), anti-CD8α (RPA-T8), anti-CD8β (SIDI8BEE), anti-CD28 (CD28.2), anti-TCRαβ (IP26). All of these Abs were purchased from eBioscience, with the exception of anti-CD8α, which was purchased from BD. The following Abs were used for CHIP, RIP, and Western blotting: PIWIL4 Ab (Abcam; 87939), PIWIL1 Ab (Sigma; SAB4200365), HA Ab (MBL; M180-3), FLAG Ab (MBL; PM020), β-actin Ab (CST; 4967), JNK1 Ab (Abcam; ab10664), p53 Ab (Santa Cruz; sc-126), and BRF1 Ab (Abcam; 74221).

Cells were collected and washed in PBS with 0.5% BSA, and the single-cell suspensions were labeled on ice for 30 min with various Abs. Flow cytometry was performed on a BD Fortessa and analyzed with FlowJo software.

RNAs extracted from monocytes and DCs were deep sequenced to analyze the small RNA fraction, at lengths of 18–40 nt, by the Beijing Genomics Institute (Shenzhen, China). Before the construction of the small RNA library, an Agilent 2100 Bioanalyzer (Agilent RNA 6000 Nano Kit) was used to examine sample integrity and concentration, and a NanoDrop was used to examine inorganic ions or polycarbonate contamination. This step was intended to provide a reference for library construction and later analysis. The small RNA library was constructed using a TruSeq Small RNA Sample Pre Kit (Illumina). In brief, the following steps were performed: 1) separating 18–40 nt small RNA fragment from 500 ng RNA sample using PAGE gel (ssRNA Ladder Marker; TAKARA); striped, and recycled; 2) adaptor ligation (TruSeq Small RNA Sample Kit; Illumina); 3) RT-PCR: after reverse transcription (Invitrogen), several rounds of PCR amplification, using PCR Primer Cocktail and PCR Master Mix, were performed to enrich the cDNA fragments (reaction conditions: 98°C for 30 s; 11 cycles of 98°C for 10 s, 60°C for 30 s, and 72°C for 15 s; 72°C for 10 min; 4°C hold); 4) purification of PCR products with PAGE gel and dissolution of the recycled products in EB solution; and 5) quantitation of the final library in two ways: an Agilent 2100 Bioanalyzer (Agilent DNA 1000 Reagents) was used to determine the average molecule length, and RT-qPCR (TaqMan Probe) was used to quantify the library. Then, the qualified libraries were amplified on cBot to generate the cluster on the flow cell. The amplified flow cell was sequenced with single end using the HiSEquation 2000 system. Before analysis of this sequencing result, some contaminant reads, including low-quality reads, reads with 5′ primer contaminants, reads without 3′ primer, reads without the insert tag, reads with poly A, and reads shorter than 18 nt, were removed from the fq file. Then, the length distribution of these clean reads was summarized. To avoid sequencing errors, the small RNA sequences of ultra-low reads (reads per million < 1) were not included in the bioinformatic analyses. Classification of the small RNA sequences was based on GenBank (http://www.ncbi.nlm.nih.gov/genbank) and Rfam (http://rfam.xfam.org) databases using the local Blastn program with an E value threshold of 1E-2. At this step, the sequences with more than two mismatches (including gaps) were filtered out after alignment.

All of the gene-expression profiles for PIWIL4 in mouse/human monocytes and DCs in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) Profiles database (http://www.ncbi.nlm.nih.gov/geoprofiles) were collected by searching for key words. In GEO Profiles, the values representing the expression level of all genes within one GeneChip hybridization were sorted, divided into 10 groups, and placed into percentile bins (1 to 100: lowest level to highest level; http://www.ncbi.nlm.nih.gov/geo/info/profiles.html). The key words “PIWIL4” and “dendritic” were used to retrieve the gene-expression profile data for PIWIL4 in mouse/human DCs from the NCBI GEO Profiles database, whereas the key words “PIWL4” and “monocytes” were used to retrieve the gene-expression profile data for PIWIL4 in human monocytes. Next, the GEO ID numbers of all related profiles were shown and classified according to the expression levels of the samples in each gene-expression profile research. The percentile bin into which the expression value of PIWIL4 was placed indicated the expression level of PIWIL4. For more details, please see http://www.ncbi.nlm.nih.gov/geo/info/profiles.html.

Total RNA was extracted using TRIzol reagent and purified further using a QIAGEN RNeasy Mini Kit, according to the manufacturer’s instructions. RNA quality was assessed by formaldehyde agarose gel electrophoresis and quantitated spectrophotometrically. An aliquot of 0.1 μg total RNA was used to synthesize double-stranded cDNA and produce biotin-tagged cRNA using the MessageAmp Premier RNA Amplification Kit. The resulting biotin-tagged cRNAs were fragmented into strands of 35–200 bp, according to the protocols from Affymetrix. The fragmented cRNA was hybridized to an Affymetrix GeneChip Human Genome U133 Plus 2.0 Array containing 47,000 transcripts. Hybridization was performed at 45°C with rotation for 16 h (Affymetrix GeneChip Hybridization Oven 640). The GeneChip arrays were washed and stained (streptavidin-PE) on an Affymetrix Fluidics Station 450, followed by scanning on a GeneChip Scanner 3000. The hybridization data were analyzed using GeneChip Operating software (GCOS 1.4). In a comparison analysis, we applied a two-class unpaired method, using Significant Analysis of Microarray software, to identify genes that had statistically significant differential expression between piRNA negative control (piNC) and td-piR(Glu), using a false discovery rate < 5% and a fold change > 1.8 as selection thresholds. The sequences presented in this article have been submitted to the NCBI GEO database under accession number GSE73945 (http://www.ncbi.nlm.nih.gov/geo/).

Statistical significance between two groups was determined by the Student t test using GraphPad Prism 6. The p values < 0.05 were considered statistically significant.

To investigate epigenetic regulation during DC differentiation, total small RNAs (20–40 nt) in human monocytes or DCs were deep sequenced. In addition to the major distribution peak of microRNAs (miRNAs), a class of small RNAs with lengths of 26–33 nt, distinguishable from miRNAs, was differentially expressed between monocytes and DCs. These small RNAs were much more abundant in monocytes than in DCs (Fig. 1A, Supplemental Fig. 1A). In addition to the existence of the annotated piRNAs, further analysis of these small RNAs sequences revealed that they could be mainly derived from small nucleolar RNAs or tRNAs (Fig. 1B), which were reported to be processed into small RNAs by different pathways (16–20). Although some of these small RNAs were derived from mitochondrial tRNAs in monocytes, most were derived from nuclear tRNAs (Supplemental Fig. 1A). The abundance of the tRNA-derived small RNA prompted us to explore the origin of these small RNAs. According to their sequences, these small RNAs were derived primarily from tRNA-Glu, although some of them were from tRNA-Gly, tRNA-Pro, and so on (Fig. 1C). We initially called these small RNAs “td-piR(X),” because their lengths were similar to those of classical piRNAs. td-piR(Glu) was one of the most abundant small RNAs in monocytes, but it was markedly reduced in DCs. The deep-sequencing of monocytes and DCs showed that the expression of td-piR(Glu) in monocytes was significantly higher than several miRNAs that were reported to play important roles in monocytes or DCs (Fig. 1D). The existence and differential expression of td-piR(Glu) were confirmed by RT-qPCR (Fig. 1E) and Northern blot (Fig. 1F). The expression of td-piR(Glu) was also much higher in monocytes than in B and T lymphocytes (Supplemental Fig. 1C). We further determined whether td-piR(Glu) had the characteristics of classical piRNAs. The results of reverse transcription at low concentrations of deoxynucleotide triphosphates, followed by PCR (21), proved that 2′-O-methylation was modified at the 3′ end of td-piR(Glu) (22) (Fig. 1G). RIP was conducted to test the interaction between PIWIL proteins and the small RNAs. To examine the expression of PIWILs in human monocytes and DCs, we assessed the mRNA levels of PIWILs by RT-qPCR; all of them were detectable, with the exception of PIWIL3 in monocytes and DCs (Supplemental Fig. 1D). The proteins levels of PIWILs in monocytes and DCs were also measured (Supplemental Fig. 1E). Although it is reported that the expression of PIWIL proteins is low in mouse immune cells, the expression of PIWIL4 is moderate to high in human monocytes and DCs based on the data that we systematically collected and analyzed from the NCBI GEO Profiles database (www.ncbi.nlm.nih.gov/geoprofiles) (Supplemental Fig. 1F, 1G). The enrichment of td-piR(Glu) by PIWIL1 or PIWIL4 proteins in monocytes was also confirmed (Fig. 1H). Collectively, td-piR(Glu) that is highly and specifically expressed in human monocytes and significantly decreased in DCs had the characteristics of a piRNA.

In addition to the well-known functions of tRNAs in translation, some small RNA fragments derived from tRNAs expand a new aspect of tRNA biological function (16, 18, 20). Because the sequences of td-piR(Glu) and the 5′ end of mature tRNA-Glu were identical, it seemed that td-piR(Glu) was cleaved from the 5′ end of mature tRNA-Glu by an endonucleolytic enzyme or 3′–5′ exonuclease (Fig. 1I). Because some stress-induced tRNA fragments, known as tRNA-derived stress-induced RNAs (tiRNAs), are also 30–40 nt in length and cleaved from the 5′ end of mature tRNAs, it was important to investigate whether td-piR was actually identical to tiRNA, which is generated from tRNA through an RNase, angiogenin (19). With small interfering RNA (siRNA)-mediated effective knockdown of angiogenin, the biogenesis of ti-piR(Glu) was not affected (Supplemental Fig. 1H). Further, to confirm that tRNA-Glu is the direct precursor of td-piR(Glu), the RNase Z and POP4, both of which are involved in tRNA biogenesis (23, 24), were knocked down by siRNAs in monocytes (Supplemental Fig. 4A). We noticed that, consistent with the reduction in mature tRNA-Glu (Fig. 1J), td-piR(Glu) was also downregulated (Fig. 1K).

Because the expression level of td-piR(Glu) in monocytes is much higher than in DCs, it could be involved in the differentiation of monocytes into DCs. We cultured CD14⁺ human monocytes with GM-CSF and IL-4 for 72 h and then transfected the chemically synthesized td-piR(Glu) into these DC-like cells at a high efficiency (Supplemental Fig. 2A). After 48 h, the change in gene expression caused by td-piR(Glu) was examined by mRNA profiling analysis. Approximately 300 genes that were differentially expressed with a fold change ≥ 1.8 or ≤ 0.56 were identified (Fig. 2A). Based on the gene ontology analysis, some of them are related to DC functions, including cell migration, Ag binding, and Ag processing (Supplemental Fig. 2B). Interestingly, we noticed that CD1a, a specific and functional marker of DCs that can present self and nonself lipid Ag to T lymphocytes (25, 26), was significantly downregulated by the overexpression of piRNA. However, the mRNA levels of other CD1 genes were not affected (Fig. 2A, 2B). To confirm the alteration in CD1a expression and investigate whether the expression of other DC markers was also altered, we performed RT-qPCR and flow cytometry analysis. By analyzing specific DC markers, including CD1a, CD40, CD80, CD86, HLA-DR, and DC-sign, which indicate the complete differentiation of DCs from monocytes (27), we found that only CD1a on the cellular surface was significantly decreased (Fig. 2C, 2D), which was consistent with the mRNA profiling data. Further, RT-qPCR showed that the reduction in CD1a mRNA by td-piR(Glu) was dose dependent (Supplemental Fig. 2C). Alternatively, we transfected the chemically synthesized antisense RNA, which is completely complementary to td-piR(Glu), into monocytes and found a significant increase in CD1a expression. The effect of td-piR(Glu) antisense was also dose dependent (Fig. 2E, 2F). To examine whether the function performed by CD1a on DCs was also affected by td-piR(Glu), importantly, to investigate the effect of td-piR(Glu) on the lipid Ag presentation activity of CD1a, we established a CD8-2 T cell line expressing the TCRαβ heterodimer restricted by CD1a (15) (Supplemental Fig. 2D). The cell line was cocultured with td-piR(Glu)– or piNC-treated DCs, with or without the Ag that was the organic extract of M. tuberculosis. The result showed that the ability of CD1a to present lipid Ag was significantly inhibited by td-piR(Glu) (Fig. 2G). However, other functions of DCs, such as the ability to secrete IL-12 and stimulate the proliferation of total CD4⁺ T cells, were not affected by td-piR(Glu) (Supplemental Fig. 2E, 2F).

Although we demonstrated that td-piR(Glu) played an important role in regulating CD1a expression, the signal transduction for regulating the expression of td-piR(Glu) needs to be unraveled. Because of the significant decrease in td-piR(Glu) during the differentiation of monocytes into DCs, it is possible that certain cytokine(s) could directly affect its expression. To this end, we treated monocytes with various cytokines, including GM-CSF, IL-4, IL-1β, IL-6, or TNF-α, which were reported to affect the differentiation of monocytes (28). td-piR(Glu) was significantly decreased by IL-4 but not by other cytokines (Fig. 3A). Importantly, we also found that IL-4 significantly increased CD1a expression (Fig. 3B). In an IL-4 dose-dependent experiment, we found a clear inverse correlation between the expression of td-piR(Glu) and CD1a (Fig. 3C). Therefore, it is highly likely that IL-4 is the major regulator that determines the downregulation of td-piR(Glu) during the differentiation of monocytes into DCs.

Because it is known that IL-4 plays a role in the differentiation of monocytes into DCs, and polymerase III transcription activity is decreased during the differentiation of mammalian cells (29, 30), we assessed the level of tRNA-Glu at different time points after IL-4 treatment. IL-4 treatment significantly decreased the levels of td-piR(Glu) and tRNA-Glu (Fig. 3D), indicating that the reduction in td-piR(Glu) could be caused by IL-4–mediated downregulation of tRNA-Glu. It was also shown that tRNA transcription is regulated by various pathways (29). To investigate the pathways that were involved in the regulation of tRNA-Glu by IL-4, the mRNA levels of various genes that participate in the regulation of polymerase III transcription were examined in monocytes. IL-4 treatment significantly increased the mRNA and protein levels of p53, whereas it significantly decreased the mRNA and protein levels of JNK1, BRF1, and TBP (Fig. 3E, 3F). To determine the roles of these proteins in the IL-4 signaling pathway, siRNAs specific for JNK1, p53, or BRF1 were transfected into monocytes (Supplemental Fig. 4A). The siRNA-mediated knockdown of JNK1 or BRF1 decreased the level of td-piR(Glu) but increased the mRNA level of CD1a, which is consistent with IL-4 treatment (Fig. 3G, 3H), whereas the knockdown of p53 counteracted the effects of IL-4 treatment on CD1a and td-piR(Glu) (Fig. 3I). Further, the expression of BRF1 was reduced after the knockdown of JNK1 in monocytes (Fig. 3G). These results were substantiated by ChIP-qPCR of TBP in monocytes (Supplemental Fig. 3A). Together, our data indicated that IL-4 repressed TFIIIB activity by inhibiting the expression of JNK1 to cause the reduction in BRF1, as well as by increasing the expression of p53, which can interact with TBP proteins and, therefore, inhibit the promoter occupancy by TFIIIB (31).

The reduction in CD1A mRNA levels caused by td-piR(Glu) could be due to the decreased transcription or facilitated posttranscriptional decay. The latter could be regulated by the small RNA-mediated mRNA degradation. To investigate the mechanism by which td-piR(Glu) repressed CD1a expression, proteins involved in miRNA- or siRNA-mediated mRNA decay, including DCP1A, DCP1B, DCP2, CNOT1, EXOSC10, RRP44, XRN1, and XRN2 (32), were knocked down by specific siRNAs (Supplemental Fig. 4C); the results showed no recovery in the repression of CD1a caused by td-piR(Glu) (Supplemental Fig. 3B). We then examined CD1A mRNA levels in nuclear and cytoplasmic fractions from DCs transfected with td-piR(Glu) and found that the reduction in CD1A mRNA began in the nucleus (Fig. 4A). To investigate the possibility of CD1A mRNA nuclear degradation, Mtr4 and Trf4 (Supplemental Fig. 4C), which are the major components of the TRAMP complex and act as major cofactors for the exosome complex to control RNA quality in the nucleus, were also knocked down (33). The result indicated that the repression of CD1a caused by td-piR(Glu) was not affected by the TRAMP complex in the nucleus (Supplemental Fig. 3C). Alternatively, it was shown that piRNAs and PIWI proteins induce transcriptional silencing of genome-wide transposon elements by recruiting HP1α and histone methyltransferase Su(Var)3–9 to facilitate H3K9 methylation (H3K9me2/3) (13, 34, 35). We then hypothesized that the decrease in CD1A mRNA mediated by td-piR(Glu) could be due to histone modification. To this end, the Abs specific for H3K4me3, H3K9me3, and H3K27me3 were used to examine the chromatic state of the CD1A gene. Eight pairs of primers located in different positions of CD1A gene loci were designed for ChIP-qPCR (Fig. 4B). After the overexpression of td-piR(Glu) in DCs, H3K9me3 at the CD1A 5′ region was significantly increased, and the level of polymerase II on the CD1A gene was also significantly decreased, indicating that CD1A transcription was repressed in td-piR(Glu)–treated DCs (Fig. 4C–F). Using the online program RNA hybrid analysis without any seed sequence setting (http://bibiserv.techfak.uni-bielefeld.de/rnahybrid/), we found two putative binding sites at positions +138 and +330 bp of transcription start site on the CD1A promoter, with minimum free energy of −25.7 and −25.3 kcal/mol, respectively (Fig. 4G).

To screen the possible proteins involved in CD1A transcriptional regulation via H3K9 methylation, we knocked down Argonaute subfamily and PIWI subfamily proteins (Supplemental Fig. 4D), which were closely related to small noncoding RNA functions (36). Only the knockdown of PIWIL4 protein significantly rescued the piRNA-induced repression of CD1a (Fig. 5A), indicating that PIWIL4, but not other proteins, was involved in the regulation of CD1a expression. Further, G9a, GLP, SUV39H1, SUV39H2, and SETDB1, all of which are H3K9 methyltransferases and involved in heterochromatin and gene repression (37–39), were knocked down by gene-specific siRNAs in td-piR(Glu)–treated DCs (Supplemental Fig. 4E). The td-piR(Glu)–mediated decrease in CD1A mRNA was affected by the knockdown of SUV39H1 or SETDB1 and not by other H3K9 methyltransferases (Fig. 5B). Furthermore, in td-piR(Glu)–treated DCs, we also knocked down heterochromatin proteins, including HP1α, HP1β, and HP1γ (Supplemental Fig. 4F), which can recognize and bind to the H3K9 methylation region in euchromatin and heterochromatin (40). Our data indicated that only HP1β was involved in the regulation of CD1a expression (Fig. 5C). Meanwhile, when siRNAs specific for PIWIL4, SUV39H1, SETDB1, and HP1β were transfected into monocytes (Supplemental Fig. 4B), CD1a expression was significantly increased, indicating that the epigenetically inhibitory mode for the CD1A gene occurred in monocytes (Fig. 5D, 5E). In addition, after blocking endogenous td-piR(Glu) in monocytes with its antisense RNA, the level of H3K9 methylation and the enrichment of PIWIL4 and HP1β proteins at the CD1A 5′ region were significantly reduced, suggesting that td-piR(Glu) specifically mediated this epigenetically suppressive mode at the CD1A gene (Fig. 5F–H). Therefore, we concluded that H3K9me3 methyltransferases (SUV39H1, SETDB1), PIWIL4, and HP1β played an important role in regulating CD1A transcription in monocytes and DCs.

The data from siRNA knockdown and ChIP-qPCR prompted us to further explore how piRNA/PIWIL4 complexes recruited these proteins to specifically control CD1a expression. To examine the interaction between these proteins that are involved in the regulation of CD1A transcription, including PIWIL4, SUV39H1, SETDB1, and HP1β, the coimmunoprecipitation results showed that PIWIL4 stably interacted with SUV39H1 or SETDB1 (Fig. 6A, 6B), and these three proteins interacted with HP1β (Fig. 6C, 6D). Notably, siRNA-mediated SETDB1 or SUV39H1 knockdown impaired the interaction between PIWIL4 and HP1β (Fig. 6E), indicating that PIWIL4 interacted with SETDB1 or SUV39H1 and then interacted indirectly with HP1β. To further verify the function and interaction of these proteins in monocytes, ChIP-qPCR was performed after the knockdown of PIWIL4, SETDB1, and HP1β. The knockdown of PIWIL4 decreased the amounts of SETDB1 and HP1β at the CD1A promoter region (Fig. 6F). The knockdown of SETDB1 affected the level of HP1β, but not PIWIL4, at the CD1A promoter region (Fig. 6G), whereas the amounts of PIWIL4 and SETDB1 at the CD1A promoter region were not affected by the knockdown of HP1β (Fig. 6H). Taken together, our data indicated that SETDB1 and SUV39H1 were first recruited by the td-piR(Glu)/PIWIL4 complex to the specific site at the CD1A promoter to methylate the lysine residue at the ninth position of histone 3 in this region. HP1β was then recruited by SETDB1 or SUV39H1 to this complex to stabilize histone modification and sustain the facultative heterochromatic state (Fig. 7).

The human CD1 family consists of five CD1 proteins: CD1a, CD1b, CD1c, CD1d, and CD1e. All of them are cell surface glycoproteins, with the exception of CD1e. The CD1 molecule is similar to MHC class I and can present Ag to T cells to activate immune responses to pathogens. It presents a range of Ags, including lipid, glycolipid, and lipopeptide, as well as other small molecules (25, 41). An in-depth study showed that lipid Ag-lipopeptide didehydroxymycobatin related to mycobactin metabolism can be specifically presented by the CD1a molecule to αβ T cells (15, 26). It was shown that CD1a expression can be regulated by various cytokines. In addition to GM-CSF and IL-4, IL-6, IL-1β, or TNF-α can upregulate CD1a expression on human monocytes, whereas IL-10 can downregulate it (42, 43). The activating transcription factors ATF-2 and CREB-1 are able to bind to the minimal proximal promoter region of CD1A and play a role in regulating CD1a expression in human monocytes (28). CD1a expression varies in different autoimmune disease models, including allergic bronchial asthma and psoriatic arthritis (44, 45), which raises the possibility that its expression level could be related to the intensity of the T cell immune response. However, the mechanism of CD1a regulation during the differentiation of monocytes remains to be elucidated.

By analyzing the deep-sequencing data, we found that a group of small RNAs (25–33 nt) was differentially expressed between monocytes and DCs and was primarily produced from tRNAs. Among them, abundant td-piR(Glu) could interact with PIWIL4 protein to form a complex and then recruit H3K9 methyltransferases (SETDB1 and SUV39H1) to the CD1A promoter region. Subsequently, H3K9 methylation at this region interacted with HP1β. Although PIWI proteins can interact directly with HP1α in Drosophila (46), we showed that the interaction between PIWIL4 and HP1β proteins was dependent on SETDB1 or SUV39H in human monocytes. The results of ChIP-qPCR after the knockdown of SETDB1 and HP1β confirmed this phenomenon. The discovery that the td-piR(Glu)/PIWIL4 complex induces H3K9 methylation at the promoter region provides a novel epigenetic mechanism of CD1a regulation. In monocytes, highly expressed td-piR(Glu) inhibits CD1A transcription as a result of the facultative heterochromatic state. During the differentiation of monocytes into DCs, td-piR(Glu) was downregulated, in line with the decrease in H3K9 modification at the CD1A 5′ end region. As a result, the CD1A promoter region is converted into a euchromatic state and could be recognized by different transcription factors to enhance its transcription. Although it was reported that the depletion of all three piwi genes (Miwi, Mili, Miwi2) does not affect normal hematopoiesis and, subsequently, the numbers of progenitors or committed cells in mice, it is unclear whether the functions of subsets of lymphoid and myeloid cells are affected (47). In addition, we and other investigators demonstrated that PIWIL4 is expressed in human monocytes/DCs at a significantly higher level than in mouse monocytes/DCs. Importantly, specific siRNA knockdown of PIWIL4 expression significantly decreased td-piR(Glu)–mediated CD1a repression (Fig. 5A). Collectively, therefore, our data clearly indicated that PIWIL4 specifically interacts with td-piR(Glu) and participates in the td-piR(Glu)–mediated function. A previous report indicated that a PIWI-like RNA derived from the antisense transcripts of KLD3L1 existed in NK cells to repress KLD3L1 expression by inducing DNA methylation at the promoter region (48). However, it was not verified whether this small RNA is of the features of piRNA or not, because no evidence was shown for the direct interaction with PIWI protein(s). Therefore, we showed the abundance and important function of certain piRNAs in human immune system.

The study of small noncoding RNAs revealed that tRNAs are a source of small RNAs that are derived from their 5′ or 3′ end (16). It was proposed that these tRNA-derived small RNA fragments (tRFs) are not generated randomly as by-products of tRNA degradation or decay but rather by some precise mechanisms that are involved in dicer or RNase Z (18). The length of these tRFs usually ranges from 18 to 23 nt. Further studies suggested that these tRFs have important biological functions in cell proliferation and global gene silencing via association with different Argonaute proteins (16, 18). Alternatively, it was reported that tRNAs can be cleaved within the anticodon loop by a secreted RNase, angiogenin, under the conditions of arsenite, heat shock, or UV radiation (19). These stress-induced and tRNA-derived small RNAs are 30–40 nt in length and are known as tiRNAs. Some reports indicated that they directly inhibit protein synthesis (19, 20). In this study, we identified a class of small RNAs in human monocytes that was generated from the 5′ end of tRNAs and that had characteristics of classical piRNAs, such as 2′-O-methylation at the 3′ end and specific interaction with PIWIL proteins. The level of td-piR(Glu) was not affected by siRNA-mediated knockdown of dicer or angiogenin, which suggested that the generation of td-piR(Glu) may be through a novel pathway. Furthermore, tRFs have a relative preference for Argonaute3–4 proteins (18), tiRNA is only associated with YB-1 (20), and td-piR(Glu) interacts with PIWIL1/4 proteins but not Argonaute proteins. Moreover, tRFs are 5′-phosphorylated and 3′-hydroxylated (18), tiRNAs are likely to be 2′-, 3′-cyclic phosphorylated at the 3′ end (19), and td-piR(Glu) harbors a 2′-O-methylation at its 3′ end. Taken together, td-piR(Glu) with characteristics of piRNAs is different from tRFs or tiRNAs.

Our data showed that the level of td-piR(Glu) was significantly reduced during differentiation, and this was attributed to IL-4 treatment that decreased TFIIIB activity to repress the transcription of tRNA-Glu. Although tRNAs play roles in protein translation and have many other functions, including the regulation of apoptosis in mammalian cells, use in retroviral transcription, and involvement in signal transduction pathways responding to nutrient deprivation (49), our work demonstrates that tRNAs can act as a precursor of the signaling molecule td-piR(Glu) and participate in the regulation of gene expression during the differentiation of immune cells. Nevertheless, it is possible that the level of tRNA-Glu is also controlled by other transduction pathways that could regulate exonuclease or endonuclease activity. As a multifunctional cytokine, IL-4 activates several signaling pathways and plays an important role in the immune response. In this study, we showed that IL-4 potently downregulated tRNA-Glu expression through the JNK1 and p53 signaling pathways. Although it is well-known that IL-4 can guide the differentiation of monocytes into DCs but not macrophages (4), there is no evidence with regard to the underlying mechanism. In this study, we found that IL-4 increases the transcription of CD1A by downregulating the level of its inhibitor, td-piR(Glu). Therefore, our work demonstrates that piRNAs can function as a transduction molecule to transfer the signal from IL-4 to the specific expression of CD1a. This discovery suggested that IL-4 regulates the levels of a variety of small noncoding RNAs through different pathways to orchestrate the differentiation of monocytes.

In this study, we uncovered the role of td-piR(Glu)/PIWIL4 in regulating CD1a expression; we also revealed that td-piR(Glu) can be regulated by IL-4 through the JNK1 and p53 signaling pathways and acts as a functional signaling molecule (Fig. 7). These results should prompt further exploration of the function, biogenesis, and regulation of piRNAs in the mammalian immune system. In addition, the novel mechanism that we propose in this article provides a feasible and specific method to control CD1a expression to treat some autoimmune diseases, as well as to avoid immune rejection of transplants and immune evasion of tumors caused by aberrant CD1a expression.

We thank Dr. Chuan Bai (Sun Yat-sen University) for guiding the preparation of lipid Ag and Dr. Xiaomin Lai (Sun Yat-sen University) for providing M. tuberculosis strain H37Ra.

The authors have no financial conflicts of interest.