Abstract
Plasmacytoid dendritic cells (pDCs) efficiently produce large amounts of type I IFN in response to TLR7 and TLR9 ligands, whereas conventional DCs (cDCs) predominantly secrete high levels of the cytokines IL-10 and IL-12. The molecular basis underlying this distinct phenotype is not well understood. In this study, we identified the MAPK phosphatase Dusp9/MKP-4 by transcriptome analysis as selectively expressed in pDCs, but not cDCs. We confirmed the constitutive expression of Dusp9 at the protein level in pDCs generated in vitro by culture with Flt3 ligand and ex vivo in sorted splenic pDCs. Dusp9 expression was low in B220− bone marrow precursors and was upregulated during pDC differentiation, concomitant with established pDC markers. Higher expression of Dusp9 in pDCs correlated with impaired phosphorylation of the MAPK ERK1/2 upon TLR9 stimulation. Notably, Dusp9 was not expressed at detectable levels in human pDCs, although these displayed similarly impaired activation of ERK1/2 MAPK compared with cDCs. Enforced retroviral expression of Dusp9 in mouse GM-CSF–induced cDCs increased the expression of TLR9-induced IL-12p40 and IFN-β, but not of IL-10. Conditional deletion of Dusp9 in pDCs was effectively achieved in Dusp9flox/flox; CD11c-Cre mice at the mRNA and protein levels. However, the lack of Dusp9 in pDC did not restore ERK1/2 activation after TLR9 stimulation and only weakly affected IFN-β and IL-12p40 production. Taken together, our results suggest that expression of Dusp9 is sufficient to impair ERK1/2 activation and enhance IFN-β expression. However, despite selective expression in pDCs, Dusp9 is not essential for high-level IFN-β production by these cells.
Introduction
In innate immune cells, recognition of microbial ligands by TLRs is a powerful activator of the MAPK cascade. The MAPKs expressed in innate immune cells (ERK1/2, p38, and JNK1/2) are activated through phosphorylation by their respective MAPKK at tyrosine and threonine residues. Phosphorylation of MAPK substrate proteins acts as a molecular switch to activate or inhibit transcription factors, other kinases, and proteins controlling mRNA stability and translation. Hence MAPKs control the macrophage and dendritic cell (DC) response to microbial encounter and are required for the expression of many inflammatory and host defense genes (1, 2).
TLR7 and TLR9 are localized in endosomal compartments and recognize nucleic acids derived from viruses or endocytosed bacteria, namely, ssRNA in the case of TLR7 and CpG-motif–containing DNA in the case of TLR9 (3). In the mouse, both TLR7 and TLR9 are expressed in a wide variety of different DC subsets and macrophages, whereas in the human system, only plasmacytoid DCs (pDCs) express significant amounts of TLR9 (4). Murine DCs can be categorized based on surface marker expression as either conventional DCs (cDCs) expressing high levels of CD11c, but not B220 or PDCA-1, or as pDCs that have intermediate levels of CD11c and stain positive for B220 and PDCA-1 (5). Both types of DCs express TLR7 and TLR9 and respond to the respective synthetic ligands (R837 and CpG oligodeoxynucleotides [ODN], respectively). However, cDCs and pDCs generate remarkably different patterns of cytokine and IFN responses, characterized by high-level release of the type I IFNs IFN-β and several IFN-α isoforms by pDCs, whereas cDCs generate only modest amounts of IFNs and produce more cytokines like IL-10, IL-12, and IL-6. The molecular basis for the capacity of pDCs to generate high amounts of IFN-β upon TLR triggering is only incompletely understood (5), but may involve constitutive expression and activation of IFN regulatory factor (IRF)7 (6), probably through mTOR activation (7). However, there is also evidence that the MAPK ERK1/2 can negatively influence the expression of IFN-β, presumably through induction of the transcription factor c-fos that acts as a negative regulator of IFN-β transcription (8).
MAPK phosphatases of the Dusp gene family bind to activated MAPK and dephosphorylate the tyrosine and threonine residues of the TxxY motif, causing reversible inactivation and thereby regulating the intensity and duration of MAPK activation (9). The 11 members of the Dusp MAPK phosphatase family differ in substrate selectivity, subcellular localization, and regulated expression during development, in specific cell types, or after stimulation (9). Several Dusp family members are highly inducible by MAPK-activating stimuli and act as negative feedback regulators. Examples with an established function in the immune system are Dusp1/MKP-1 (10–13), Dusp2/PAC1 (14), Dusp10/MKP-5 (15), and Dusp16/MKP-7 (16, 17). Investigations in the respective knockout mice have shown that a lack of these inducible Dusp phosphatases leads to alterations in the patterns of cytokines produced via the effects of dysregulated MAPK activation (for review, see Refs. 2, 18).
In contrast, other Dusp genes are expressed constitutively at relatively high levels in some cell types and may therefore control the basal as well as induced MAPK phosphorylation. Given the preference of different Dusp proteins for individual MAPK substrates, the relative levels of ERK1/2, p38, and JNK activity may well be determined by constitutively expressed Dusp proteins. Cell type–selective expression of Dusp genes can therefore be envisaged to specify cell type–specific responses at the level of MAPK activation.
In this study, we mined microarray data to identify Dusp genes with differential expression in DC subsets. This approach identified Dusp9/MKP-4 as a pDC-selective gene in mice, whose expression was induced during differentiation from bone marrow progenitor cells. Functionally, higher Dusp9 mRNA and protein levels correlated with attenuated ERK activation in pDCs. As enforced expression of Dusp9 in cDCs was sufficient to increase production of IFN-β, our results suggest that Dusp9 may contribute to the characteristic differences in the pattern of cytokine and IFN production by different DC subsets.
Materials and Methods
Reagents
Abs against phospho-p38, phospho-ERK1/2, p38, ERK1/2, Nbs1, and pStat1 were purchased from Cell Signaling Technology. Anti-Grb2 Ab was from Becton-Dickinson. The polyclonal sheep anti-Dusp9 antiserum has been described previously (19). Recombinant Flt3 ligand (FL) was obtained from R&D Systems.
Mice
C57BL/6 mice were purchased from Charles River and bred at the Franz Penzoldt Zentrum of the Medical Faculty at the University Hospital Erlangen.
Mice conditionally targeted at the Dusp9 locus were generated by homologous recombination (Supplemental Fig. 3). In brief, a targeting construct in which the murine Dusp9 locus was modified by the introduction of loxP sites flanking exons 3 and 4 and the insertion of a positive selection marker (puromycin resistance) flanked by Frt sites downstream of exon 4 and immediately 5′ to the loxP site within the 3′ untranslated region was electroporated into C57BL/6 Embryonic stem cells. After puromycin selection, embryonic stem cells carrying the correctly targeted allele were identified by Southern blotting and injected into blastocysts to generate chimeric mice. Chimeric mice were then crossed with the Flp deleter strain (C57BL/6-Tg(CAG-Flpe)2Arte; TaconicArtemis), which results in recombination and removal of the puromycin selection marker to generate the conditional allele (Dusp9fl). For conditional deletion of Dusp9 in DCs, Dusp9fl/fl mice were crossed with CD11c-Cre–expressing mice (B6.Cg-Tg(Itgax-cre)1-1Reiz/J) (20). Dusp9fl/fl female and Dusp9fl/y male mice positive or negative for CD11c-Cre were matched for sex and age for hydrodynamic injection and DC sorting (see later).
Isolation of DC subsets from mouse spleen
To increase the number of DCs in vivo, we injected C57BL/6 mice with recombinant Flt3L to induce DC differentiation. On day 12 after injection, spleens were harvested, and after DC enrichment by centrifugation over OptiPrep gradient and MACS separation (using beads for isolation of PDCA-1+ and CD11c+ pDCs and cDCs, respectively), cells were stained for CD11c and B220 and purified by MoFlo cell sort.
Alternatively, in the experiments described in Figs. 4B, 4C, and Fig. 7, hydrodynamic injection of C57BL/6 mice with an FL-encoding plasmid was used to induce expansion of DCs in vivo. Mice were injected i.v. with 5 μg plasmid in 0.1 ml/g body weight Krebs-Ringer buffer within 5–7 s. The animals were sacrificed after 10 d and single-cell suspensions were prepared from the spleens using the GentleMACS dissociator (Miltenyi, Bergisch-Gladbach, Germany). After erythrocyte lysis, splenocytes were counted (yield per spleen 2–3 × 108 cells) and stained with Abs for flow-cytometric analysis and sorting (see later).
Microarray profiling of splenic DC subsets
Total RNA prepared from sorted splenic cDCs and pDCs was prepared and labeled for Affymetrix microarray analysis according to the manufacturer’s protocol. Biotinylated cRNA was hybridized to MOE430 2.0 GeneChips, developed with streptavidin-PE and scanned. Cel files containing probe-level data were normalized with GC robust multiarray average using the software Flexarray (21). Microarray data have been submitted to Gene Expression Omnibus at National Center for Biotechnology Information (accession number GSE54120; http://www.ncbi.nlm.nih.gov/genbank).
Isolation of DC subsets from human PBMCs
DCs were isolated from buffy coats obtained from healthy volunteers after written informed consent and according to institutional guidelines. PBMCs were purified from buffy coats via Ficoll density gradient centrifugation (Lucron Bioproducts, Sint Martens-Latem, Belgium). To obtain peripheral blood leukocytes, we depleted monocytes from PBMCs via adherence to plastic culture flasks. CD1c+ mDCs and CD16+ mDCs were isolated from PBMCs with a CD1c+ DC isolation kit and CD16+ monocyte isolation kit, respectively. pDCs were purified from peripheral blood leukocytes by positive selection using anti–BDCA-4–conjugated magnetic microbeads (all Miltenyi Biotec). DC purity was assessed by double staining CD11c+/CD1c+ for CD1c-mDCs (>95%), CD11c+/CD16+ for CD16-mDCs (>90%), and BDCA2/CD123 for pDCs (>95%) (all from Miltenyi Biotec). DCs were cultured in X-VIVO-15 medium (Cambrex, Verviers, Belgium) supplemented with 2% human serum. DCs were stimulated as indicated with the following TLR ligands: 4 μg/ml R848 or 5 μg/ml CpG C (M362) (both Axxora, San Diego, CA).
In vitro differentiation of DCs from bone marrow progenitors
Bone marrow cells were obtained by flushing the femurs and tibiae of C57BL/6 mice with complete RPMI (cRPMI), followed by erythrocyte lysis with NH4Cl. Cells were resuspended in cRPMI and counted. Cells were plated at a density of 1.5 × 106 cells/ml and cultured in the presence of recombinant FL at a concentration of 35 ng/ml for 8 d. The resulting mixed population was used for isolation of cDCs and pDCs by flow-cytometric cell sorting (see later). In some experiments, bone marrow cells were depleted of B220+ cells (B cells and pDCs) using MACS technology as suggested by the manufacturer (Miltenyi). The eluted B220− cells were counted and plated in the presence of FL or GM-CSF–containing supernatant of X63 cells (10% in cRPMI) for kinetic analysis of mRNA and protein expression.
Flow cytometry and cell sorting
Cells were stained in FACS buffer (PBS with 2% FBS and anti-CD16/CD32 Abs for blockade of Fc receptors) using fluorochrome-labeled Abs (CD11b-FITC, CD11c-allophycocyanin, B220-PE, PDCA-1–PE). pDCs were defined as CD11b−CD11c+B220+ or CD11b−CD11c+PDCA-1+ cells. cDCs were defined as CD11c+B220− cells. Cell sorting for pDCs and cDCs was performed using MoFlo (Beckman Coulter).
Generation of retroviral constructs and transduction of bone marrow progenitors
The full-length coding sequence of murine Dusp9 was amplified from mouse pDC cDNA and ligated into the MigR1 retroviral vector using BglII and EcoRI restriction sites. For production of retrovirus, Phoenix cells were transfected by the calcium-phosphate method with MigR1 or MigR1-Dusp9 together with the packaging plasmid pCLEco. Supernatants were collected 48 h after transfection, passed through a 0.45-μm filter, and used to infect bone marrow progenitors. To this end, 106 bone marrow cells/ml were plated in 12-well plates and cultured with cRPMI with 10% GM-CSF–containing ×63-supernatant. Two days later, spin infection was performed by centrifugation at 2000 rpm for 2 h at 30°C. After 2 h at 37°C, 5% CO2, virus-containing supernatants were gently removed and replaced by fresh cRPMI with GM-CSF. DCs were harvested on day 6 of culture and sorted based on GFP expression by MoFlo.
Stimulation of DCs with TLR ligands
Cells were plated at a density of 106/ml, rested in cRPMI for 3 h, and stimulated by addition of ligands for TLR7 (R837; Invivogen) or TLR9 (CpG ODN; TIB MOLBIOL, Berlin, Germany). The A-type CpG ODN 2216 (5′-GGGGGACGATCGTCGGGGGG-3′, only underlined bases protected by phosphorothioate backbone) and the B-type CpG ODN 1826 (5′-TCCATGACGTTCCTGACGTT-3′, full phosphorothioate backbone) were used. At the indicated time points, supernatants were harvested and the cells were lysed for analysis of proteins or RNA.
RNA preparation and quantitative real-time PCR
MicroRNeasy kits (Qiagen, Hilden, Germany) were used for isolation of total RNA, including an on-column DNAse digest. RNA was reverse transcribed using random-primed cDNA synthesis with the High Capacity cDNA RT kit (Applied Biosystems). Quantitative real-time PCR (qRT-PCR) was performed on an ABI 7900 SDS TaqMan System using primer-probe combinations selected from the Universal Probe Library (Roche). Semiquantitative analysis was performed using the ΔΔ cycle threshold (CT) method (22) and using Hprt and PPIA expression for normalization of murine and human data, respectively.
Cytokine and IFN detection by ELISA
The cytokines IL-10 and IL-12p40 were determined in cell culture supernatants by DuoSet ELISA Developmental Systems (R&D Systems) following the manufacturer’s protocol. Detection of IFN-β was done by sandwich ELISA using a monoclonal rat anti-mouse IFN-β capture Ab and a polyclonal rabbit anti-mouse IFN-β detection Ab (both from Tebu-bio), followed by a donkey anti-rabbit Ab coupled to HRP.
Preparation of protein extracts and Western blot analysis
Before lysis, cells were washed once with ice-cold PBS. To obtain total cell lysate, we added 100 μl ice-cold radioimmunoprecipitation assay buffer per 106 cells, and collected and stored lysates at −20°C. For the stepwise preparation of cytoplasmic and nuclear extracts, the NE-PER Nuclear and Cytoplasmic Extraction Reagents (ThermoScientific/Pierce) were used according to the manufacturer’s protocol. Lysates were sonicated for five 1-min cycles at maximum power (Bioruptor; Diagenode), 4× loading buffer including mercaptoethanol was added, and the samples were incubated at 95°C for 5 min. Proteins were separated by 12% SDS-PAGE and blotted onto Protran membranes (Neolab), blocked with 3% BSA or 5% milk, and incubated overnight with the primary Ab. After washing with TBST and incubation with secondary peroxidase-coupled Ab, the proteins were detected using Immobilon Western HRP Substrate by chemiluminescence. Where indicated, densitometry of band intensities was performed using ImageJ analysis software (National Institutes of Health).
Statistical analysis
Student t test was used to determine statistical significance: *p < 0.05, **p < 0.01, and ***p < 0.001.
Results
Transcriptome analysis identifies Dusp9 as selectively expressed in pDCs
Based on the hypothesis that differential expression of Dusp family members may be involved in cell type–specific responses of different innate immune cells to microbial stimuli, we analyzed Affymetrix microarray data generated from highly purified splenic cDCs and pDCs (Fig. 1A). Among the Dusp genes with significantly detectable expression, several were expressed more strongly in cDCs (Dusp1, Dusp2, Dusp3, Dusp16). In contrast, the three probe sets detecting Dusp9 mRNA gave 5- to 26-fold higher signals in pDCs than in cDCs. In addition, probe sets for Dusp18 and Dusp28 showed stronger signals in pDC. However, both Dusp18 and Dusp28 were expressed at markedly lower levels than Dusp9 in pDCs. Therefore, we regarded Dusp9 as a more interesting pDC-selective Dusp gene. The level of enrichment for Dusp9 was comparable with that observed for a number of established pDC-marker genes like Tlr7, Ccr9, and the transcription factors Spib, Tcf4, and Runx2 (Fig. 1B). We also mined the recently published dataset generated by the ImmGen Consortium (23) for Dusp9 expression across 678 samples from a broad range of immune cells and observed high expression of Dusp9 exclusively in the pDC samples (Fig. 1C).
Validation of Dusp9 expression in tissues and immune cell subsets
To validate the microarray data, we differentiated DCs from bone marrow cells by culture in FL, and sorted pDCs (B220+ CD11c+ CD11b−) and cDCs (B220− CD11c+ CD11b+). Confirming the microarray results obtained with splenic DCs, quantitative RT-PCR showed much higher expression of Dusp9 in pDCs, whereas Dusp1 and Dusp2 were more abundant in cDCs (Fig. 2A). Because many Dusp genes are inducibly expressed, we next asked how stimulation of sorted pDCs and cDCs with the TLR9 ligands A- and B-type CpG ODN would affect expression of selected Dusp genes, including Dusp9 (Fig. 2B). Although A-type CpG ODN 2216 did not significantly change Dusp mRNA levels in pDCs, B-type ODN 1826 strongly increased the expression of Dusp1, Dusp2, and Dusp4, but decreased Dusp6 mRNA levels. Dusp9 mRNA increased after ODN 1826 stimulation of pDCs, but not cDCs, further augmenting the differential expression between these cell types.
Next, we assessed Dusp9 protein levels in sorted immune cells by Western blot (Fig. 2C). Consistent with the mRNA data, a strong signal was detected from B220+ CD11c+ splenic pDCs, but not from cDCs, T cells, B cells, granulocytes, and monocytes. Thus, in addition to placenta and specialized renal tubular epithelial cells (24), in the immune system, pDCs selectively express significant amounts of Dusp9 mRNA and protein.
Dusp9 expression is induced during differentiation of pDCs
Next, we sought to determine whether Dusp9 is expressed in DC precursors and downregulated during differentiation of cDCs, or instead upregulated during the development of pDCs. To address this question, we depleted B220+ cells from bone marrow, used the B220− fraction to differentiate DCs in the presence of FL or GM-CSF, and monitored the appearance of pDCs and cDCs by FACS and Dusp9 expression by qRT-PCR. As expected, FL culture gave rise to a mixed population of CD11c+ DCs, containing B220−/PDCA-1− cDCs and B220−/PDCA-1+ pDCs, which were first detectable on day 4 and increased in percentage to day 8, whereas only a small fraction of pDCs was generated in GM-CSF cultures (Fig. 3A, 3B). Ex vivo, Dusp9 expression was much lower in B220− compared with B220+ bone marrow cells, consistent with the presence of pDCs in the B220+ bone marrow cell population. Dusp9 expression increased in FL-treated cultures in parallel to the appearance of pDCs (Fig. 3C). The expression of the established pDC markers Tcf4 and Irf7 followed a similar pattern as Dusp9 in FL cultures (Fig. 3C). In contrast, DC differentiation with GM-CSF did not induce substantial induction of Dusp9, Tcf4, or Irf7 mRNA expression. Furthermore, mining of the ImmGen data for Dusp9 showed that expression is low in progenitor populations (common myeloid progenitor, common DC progenitor, granulocyte monocyte progenitor, macrophage DC progenitor) and in stem cells (Supplemental Fig. 1). We conclude that Dusp9 is expressed at low levels in DC precursors and is upregulated during pDC differentiation.
Dusp9 expression in pDC correlates with attenuated ERK activation
Because Dusp9 has been shown to preferentially dephosphorylate ERK1/2 and p38 MAPK (19), we determined the activation of these MAPKs in pDCs and cDCs after stimulation with CpG ODN. We observed strongly reduced basal and CpG ODN-induced phosphorylation of ERK1/2 in total cell lysates of FACS-sorted pDCs compared with cDCs, which correlated to the high protein expression levels of Dusp9 in pDCs (Fig. 4A). Despite a reduced basal phosphorylation of p38 in pDCs, we still observed a significant increase after stimulation. Total levels of ERK1/2 and p38 were comparable between pDCs and cDCs (Supplemental Fig. 2A).
To corroborate and extend these findings, we sorted splenic DC populations from mice expressing FL from hepatocytes after hydrodynamic plasmid injection, which greatly expands the number of splenic DCs. Sorted pDCs and cDCs, as well as unsorted splenocytes of FL-expressing mice, were stimulated with CpG ODN and with the TLR7 ligand R837. As expected, pDCs produced much higher levels of IFN-β in response to TLR7 or TLR9 stimulation than unsorted splenocytes or cDCs (Fig. 4B). In contrast, IL-10 production to the same stimuli was lower in pDCs, whereas all populations generated large amounts of IL-12p40 (Fig. 4B). The relatively large number of pDCs and cDCs obtained using this approach allowed subcellular fractionation. Analysis of nuclear and cytoplasmic fractions (Fig. 4C) revealed strong and constitutive cytoplasmic expression of Dusp9 in pDCs but not cDCs, confirming the differential expression previously observed after in vitro DC differentiation. High Dusp9 protein levels in pDCs corresponded to very low activation of ERK in the cytoplasm and no detectable phosphorylated and very little total ERK in the nucleus (Fig. 4C, Supplemental Fig. 2B). However, pDCs clearly responded to CpG 1826 because strong tyrosine phosphorylation of Stat1 was detected at the 90-min time point. Stat1 activation is likely due to the autocrine action of IFN-β produced by pDCs (Fig. 4B). In contrast, a strong increase in the cytoplasmic and nuclear levels of pERK was observed in cDCs. Together, the presence of cytoplasmic Dusp9 protein in pDCs correlated strongly with attenuated activation and nuclear translocation of ERK1/2 and robust IFN-β production compared with cDCs.
Human pDCs do not express Dusp9 but fail to phosphorylate ERK after TLR7/9 stimulation
Human Dusp9 differs from the mouse protein by the lack of several AEAK-repeats in the N-terminal domain, which explains the difference in the molecular mass (45 kDa for human versus 62 kDa for murine Dusp9). However, in both species, the highest expression of Dusp9 has been reported in placenta and kidney (19, 25). To determine whether the Dusp9 expression in pDC is conserved between mouse and human, we sorted pDCs and CD1c+ and CD16+ cDCs from PBMCs of healthy donors and determined expression of selected Dusp genes by qRT-PCR. In marked contrast with murine pDCs, expression of Dusp9 was barely detectable by qRT-PCR in resting human pDCs and lower than in CD1c+ mDCs (Fig. 5A). In addition, Dusp9 mRNA was not inducible in pDCs or cDCs after stimulation with the TLR7/8 ligand R848 or with a C-type CpG ODN (Fig. 5B). As expected, human pDCs responded to stimulation with strong expression of IFN-β and IFN-α mRNA, exceeding the levels observed in CD1c+ or CD16+ cDCs (Fig. 5C). Thus, although the pDC-typical robust type I IFN response is conserved between mouse and human, the selective expression of Dusp9 in pDCs is specific to mouse. However, analysis of ERK1/2 phosphorylation in stimulated human DC subsets revealed a much weaker response of pDCs compared with both cDC populations (Fig. 5D), which recapitulates the attenuated ERK phosphorylation in murine pDCs. Of note, expression of Dusp4, Dusp5, and Dusp6 was relatively high in human pDCs, and in the case of Dusp4 and Dusp5 showed a tendency to higher levels than in cDCs (Fig. 5A). Because these Dusp family members preferentially dephosphorylate ERK1/2, they may functionally replace the lack of Dusp9 in human pDCs.
Overexpression of Dusp9 in cDCs regulates ERK activation and increases IFN-β expression
To test whether Dusp9 expression is sufficient to regulate MAPK activation and production of cytokines and IFNs in response to TLR9 stimulation in DCs, we overexpressed Dusp9 in GM-CSF–derived cDCs. Bone marrow cells were transduced with a retrovirus conferring bicistronic expression of Dusp9 and GFP on day 2 of culture with GM-CSF and sorted on day 6 based on GFP expression. Dusp9-MigR1 conferred robust Dusp9 expression in GM-CSF–derived cDCs. Retrovirally encoded Dusp9 was predominantly found in the cytoplasm of cDCs (Fig. 6A), as was the endogenous Dusp9 protein in pDCs (Fig. 4C). The level and cytoplasmic localization of Dusp9 was not affected by stimulation with CpG ODN. Sorted GM-CSF DCs showed relatively high basal levels of ERK phosphorylation, which were only moderately increased by CpG stimulation. However, Dusp9-MigR1–transduced DCs exhibited less cytoplasmic phospho-ERK under basal conditions and 3 h after stimulation (Fig. 6A). The phosphorylation of p38 MAPK was not significantly altered in cDCs overexpressing Dusp9 (Fig. 6A).
The impact of enforced Dusp9 expression in cDCs on TLR9-induced production of IFN-β, IL-12p40, and IL-10 was examined at the mRNA and protein levels (Fig. 6B, 6C). Sorted Dusp9-expressing GM-CSF–derived DCs produced significantly greater levels of IFN-β and IL-12p40 in response to CpG 1826, whereas IL-10 production was not significantly altered compared with MigR1-control transduced DCs. To gain insight into how Dusp9 overexpression may increase IFN-β and IL-12 expression, we determined its effect on the expression of the transcription factors Fos and IRF1. ERK-dependent expression of Fos inhibits IFN-β expression (8), whereas IRF-1 promotes its expression in response to CpG (26). Indeed, retroviral Dusp9 downregulated Fos mRNA but increased the expression of IRF-1 in cDCs (Fig. 6D). Taken together, the results from the retroviral transduction experiments showed that increasing the levels of Dusp9 in GM-CSF–derived DCs was sufficient to modulate the output of TLR9-triggered activation toward a more pDC-like profile of robust IFN-β production.
Deletion of Dusp9 in pDCs does not alter ERK activation and IFN-β production
To investigate the function of Dusp9 in pDCs, we crossed newly established Dusp9fl/fl mice with CD11c-Cre mice to generate a conditional deletion of Dusp9 in DCs (Supplemental Fig. 3A, 3B). FACS analysis of splenocytes after hydrodynamic injection of FL showed similar frequencies of pDCs and cDCs independent of CD11c-Cre expression (Supplemental Fig. 4A). FACS-sorted pDCs from Dusp9fl/fl; CD11c-Cre+ females and Dusp9fl/y; CD11c-Cre+ males showed efficient deletion of the loxP-flanked exons (Supplemental Fig. 4B), leading to, on average, 90% reduction of Dusp9 mRNA expression (Fig. 7A). Dusp9-deficient pDCs responded largely normal to stimulation with CpG ODN, as induction of mRNA expression for Ifnb was not significantly altered and Il12b was slightly reduced (Fig. 7A). There was no significant difference in the CpG-induced expression of other type I IFN genes (Ifna4, Ifna5) and of Il10 (data not shown). Western blot analysis of conditional Dusp9 knockout DCs showed a nearly complete absence of Dusp9 protein (Fig. 7B), confirming the mRNA data. However, the phosphorylation of ERK1/2 and p38 MAPK in pDCs was not increased in the absence of Dusp9, and still much less compared with that observed in cDCs after stimulation with CpG ODN (Fig. 7B). Thus, Dusp9 was not required for impairment of ERK and p38 activation, and its absence only weakly reduced high-level Ifnb and Il12b expression in pDCs.
Discussion
There is ample evidence that individual MAPK can differentially control the balance of cytokines by macrophages and DCs after TLR stimulation (27, 28). Based on our working hypothesis that Dusp MAPK phosphatases specify cell type–specific transcriptional responses by controlling the intensity and duration of MAPK phosphorylation, we have sought to identify Dusp genes that are constitutively present in certain immune cell types and may control the initial response type (18). In this article, we show that Dusp9/MKP4 is strongly and selectively expressed in mouse pDCs. We further provide evidence that Dusp9 may be involved in specifying the robust IFN-β production of pDCs because high Dusp9 expression in pDCs correlates with impaired TLR9-induced ERK1/2 activation, and enforced expression of Dusp9 in cDCs leads to increased IFN-β production. However, the results of experiments using the newly established conditional Dusp9 knockout mice revealed that this MAPK phosphatase is not essentially required for high-level IFN-β expression and the attenuation of ERK1/2 phosphorylation in pDCs. The unaltered weakness of ERK1/2 phosphorylation in Dusp9-deficient pDC could be due to: 1) functional compensation by other Dusp phosphatases; 2) the activity of unrelated phosphatases; or 3) an intrinsic absence of MAPK activation in pDCs due to spatiotemporal differences in signaling, for example, controlled by differential trafficking of TLR7/9 to the lysosomal compartment (29). A number of ERK1/2-selective phosphatases are expressed in pDCs in addition to Dusp9, including Dusp4, Dusp5, and Dusp6 (Figs. 1, 2), which may compensate in the absence of Dusp9. However, we have not explored yet whether these Dusp genes are overexpressed in Dusp9-deficient pDCs. Testing the role of these Dusp family members in pDCs is in principle possible, because knockout mice for Dusp4 and Dusp6 have been described (30–32), but may only be informative after crossing the mice to delete several Dusp genes simultaneously.
Dusp9 expression was initially described in the kidney and the placenta (19, 25); its expression in the placenta is essential during development (24). More recently, Dusp9 has been linked to adipocyte differentiation and insulin resistance in mouse models (33) and by human GWAS studies (34–36). In addition, Dusp9 expression is induced by BMP4 signaling and attenuates ERK activity in embryonic stem cells (37). To the best of our knowledge, we report for the first time on a role for Dusp9 in immune cells. However, mining other transcriptome or proteome datasets, which became available during the course of this study, yielded information supporting our identification of pDC-selective Dusp9 expression. First, the impressively comprehensive transcriptomic analysis of murine immune cell subsets, which was recently published by the ImmGen Consortium (23), confirms our results of high Dusp9 expression in pDCs, but not cDCs or any other immune cell type (Fig. 1C). Second, the dataset from a proteomic comparison of murine DC subsets, performed by mass spectrometry by Luber et al. (38), shows that Dusp9 protein is found at significant levels in pDCs, but not in CD4+ or CD8+ cDCs. Together, these independent data from a variety of pDC sources, including pDCs sorted from steady-state mouse organs, show that the selective Dusp9 expression in pDCs observed by us is not caused by in vitro culture conditions or the presence of unphysiological amounts of FL in vivo in our experiments.
The factors controlling Dusp9 expression in a cell type–restricted manner are currently not well defined. Only in the case of embryonic stem cells, the induction of Dusp9 by BMP4 signaling was shown to require binding of SMAD1 and SMAD4 to a conserved response element in the Dusp9 promoter (37). Whether the BMP4-SMAD axis also plays a role in Dusp9 expression in placenta, kidney, and pDC remains to be elucidated. Our kinetic analysis during differentiation of B220− bone marrow cells in culture with FL clearly showed that Dusp9 expression is induced in parallel with the pDC-specific transcription factors Tcf4 and Irf7, and increases concomitant with the percentage of pDC in the cultures. FL-induced transcription factors are therefore candidate-inducers of Dusp9 expression.
The species-specificity in robust Dusp9 expression for mouse but not human pDCs we observed is surprising in view of the conserved expression of the gene in the placenta and the kidney (19, 25), as well as its inducibility by BMP4 in mouse and human embryonic stem cells (37). One possible explanation could be differences in the source of cells, as we analyzed pDCs isolated from human peripheral blood, whereas in the murine experiments, DCs were isolated from the spleens or differentiated with FL in vitro. However, in preliminary experiments, we also differentiated human pDCs by culture of CD14+ monocytes in FL according to a recently published protocol (39), and again failed to observe upregulation of Dusp9 relative to cDCs generated with GM-CSF plus IL-4 (data not shown). In addition, a lack of Dusp9 expression in human pDCs is also evident when microarray data from DC subsets isolated from tonsils are mined (40) (dataset E-TABM-34 in ArrayExpress database at: www.ebi.ac.uk/miamexpress). Furthermore, evidence for a species difference in pDC-selective Dusp9 expression can be found in the microarray datasets generated by Robbins et al. (41), where Dusp9 expression is high in mouse pDCs but not detected in human DCs. The molecular basis for this species-specific expression in mouse pDCs is at present unknown. Analysis of Dusp9 expression in pDCs from additional species in combination with comparative promoter analysis should be helpful in defining the critical promoter regions controlling expression.
Our experiments revealed a strong correlation of weak ERK phosphorylation with high IFN-β production in pDCs. Overexpression of Dusp9 in cDCs significantly increased IFN-β mRNA and protein levels. The phenotype of macrophages and cDCs deficient in Tpl-2, a MAP3K, is reminiscent of the effect of Dusp9 overexpression in cDC: Tpl-2 is required for ERK activation in response to CpG stimulation, and Tpl-2 deficiency leads to enhanced production of IFN-β and IL-12 from macrophages and cDCs (8). In contrast, pDCs lacking Tpl-2 produced less IFN-β, indicating that the impact of ERK activity may differ between cell types (8). Kaiser et al. (8) identified the transcription factor c-fos as an ERK-dependent target gene of TLR stimulation in macrophages and observed reduced expression and DNA-binding activity in Tpl2-deficient macrophages. Furthermore, retroviral expression of Fos suppressed IFN-β and IL-12 expression in Tpl-2–deficient BMDCs. In previous work, we have described that the IFN regulatory factor IRF-1 promotes the expression of IFN-β in GM-CSF DCs stimulated with CpG ODN (26). Our results from retrovirally transduced GM-CSF DCs show that Dusp9 indeed reciprocally altered the expression of the transcription factors Fos or IRF-1 (Fig. 6D), which may mechanistically explain the regulation of IFN-β and IL-12 expression.
In conclusion, we have identified in this study the MAPK phosphatase Dusp9 as selectively expressed in mouse pDCs and shown by retroviral overexpression in cDCs that it can contribute to the specification of IFN-β production by DCs in response to TLR9 stimulation. However, the lack of a strong phenotype in Dusp9-deficient pDCs shows that it is not essential for high-level IFN-β production by these cells. Still, our results suggest that Dusp9 joins the group of MAPK phosphatases with a potential role in innate immune regulation. In contrast with the inducible MAPK phosphatases (e.g., Dusp1, Dusp2, or Dusp16), which function as classical negative feedback regulators, the constitutive yet cell type–specific expression of Dusp9 indicates that it may contribute to control already of the initial MAPK response to TLR stimulation, and thereby could explain at least partially why different DC subsets respond with such divergent cytokine and IFN output to triggering of identical TLRs.
Acknowledgements
We thank Harald Dietrich and Barbara Bodendorfer for expert technical assistance, Dr. Christian Bogdan for support, and Drs. Barbara Schmidt and Diana Dudziak for sharing unpublished data.
Footnotes
This work was supported by Deutsche Forschungsgemeinschaft Grant SFB 643, TP A10 (to R.L.), a fellowship from Bayerische Forschungsstiftung (to M.N.), Cancer Research UK Stress Response Laboratory Program Grant C8227/A12053 (to S.M.K. and S.M.). Work in the Figdor Laboratory was supported by Netherlands Institute for Regenerative Medicine Grant FES0908 and European Union/European Research Council Grant 299019.
The microarray data presented in this article have been submitted to the National Center for Biotechnology Information’s Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/genbank) under accession number GSE54120.
The online version of this article contains supplemental material.
References
Disclosures
The authors have no financial conflicts of interest.