Female C57BL/6J mice were obtained from Harlan (Horst, Netherlands) and were kept under specific pathogen-free conditions at the animal facility of the Erasmus MC, Rotterdam, Netherlands. Housing, care, and experimental handling were performed in accordance with Dutch legal regulations. Ethical approval was obtained after protocol review by the independent animal experiment committee DEC Consult, and registered under permit numbers EUR1408 (128-08-05), EUR1738 (128-09-02), EMC2135 (128-10-10), and EMC2759 (128-12-07).

Monocyte-derived DC were generated by GM-CSF stimulation of bone marrow (BM) precursors as described previously (26). These cells are indicated as GM-DC. Briefly, BM cells isolated from 8 to 13-week-old C57BL/6 mice were cultured in RPMI-1640 medium (Lonza, Belgium) supplemented with 10% fetal calf serum, 2 mM glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, 50 μM 2-mercaptoethanol, and 20 ng/ml rmGM-CSF (Biosource International, Camarillo, CA, USA). Cells were kept in a humidified incubator at 37°C with 5% CO₂. At day 0, BM leukocytes were seeded at 3 × 10⁵ per ml in either 100 mm dishes (BD Biosciences), 12-wells plates (Nunc) or 96-wells round-bottom plates (Nunc). At day 3, fresh culture medium was added to the plates and at day 6, half of the medium was replaced. To induce enforced GM-DC maturation, 100 ng/ml LPS (Escherichia coli strain 055:B5, Sigma) was added on day 6. Alternatively, plasmacytoid DC (pDC) and cDC were generated by Flt3L stimulation of BM precursors essentially as described by Naik et al. with minor modifications (27). To this end, BM cells isolated from 8 to 13-week-old female C57BL/6 mice and erythrocytes were lysed by treatment for 2 min with 0.155 M NH₄Cl. Then, BM cells were extensively washed and cultured in RPMI-1640 (Lonza, Belgium) supplemented with 10% fetal calf serum, 2 mM glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, 50 μM 2-mercaptoethanol, and 200 ng/ml Flt3L (Peprotech, Rocky Hill, NJ, USA), and seeded at 4 × 10⁶ per 2 ml in six-wells plates (Nunc). At day 7, cells were washed to remove free Flt3L, and stimulated for 24 h with 10 μg/ml CpG (ODN 2395, InvivoGen, San Diego, CA, USA).

To assess the role of TNF-α converting enzyme (TACE) in M-CSFR down-regulation, TACE inhibitors TMI-1 and TMI-2 (28, 29) were used in a final concentration of 10 and 15 μM, respectively (kindly provided by Dr. B. Scholte, Erasmus, MC, Netherlands). The functional inhibition during overnight cultures was tested by determining the TACE-mediated decrease of M-CSFR expression on BM monocytes. To that end, freshly isolated BM cells (10⁶ per ml) were cultured at 37°C for 24 h in a 12-well plate in RPMI-1640 with 10% FCS and antibiotics as described before, but without additional growth factors. Inhibitors were added from the start of the culture. As TACE inducer E. coli LPS (O55 B5, Sigma) was used in a final concentration of 100 ng/ml. Cells were harvested after 24 h. Similarly, TACE activity was inhibited in GM-DC cultures by adding inhibitors during the last 24 h of a 7-day culture in combination, either or not in the presence of LPS. Expression of M-CSFR/CD115 was determined as described below.

For cell labeling, incubations were performed in staining buffer (PBS pH 7.8, 1% BSA, 0.01% sodium azide) on ice for 30 min. Reagents used were fluorescent conjugates of CD11b (M1/70), CD11c (HL3), CD86 (GL1), CD115 (anti-M-CSFR, clone AFS98), mMGL/CD301 (ER-MP23), MHC class II I-A/I-E (M5/114.15.2), SiglecH (eBio440c), and rat-anti-mouse IgG-Alexa488 and streptavidin-Alexa633. These antibodies were obtained from BD Biosciences, eBioscience, Molecular Probes or prepared as purified Ig from hybridomas created in our lab. Cells were analyzed by flow cytometry using a FACSCalibur or FACSCanto II (Becton Dickinson) and FlowJo Analysis Software (Tree Star, Ashland, OR, USA). Sorting of cells was performed using a FACSAria Cell Sorter (Becton Dickinson).

Total RNA was extracted using acid-phenol:chloroform (Ambion) extraction and enriched for microRNAs using a mirVana microRNA isolation kit (Ambion) according to the manufacturer’s protocols. RNA was labeled using a ULS™ aRNA labeling kit (Kreatech Diagnostics, Amsterdam). 1.5 μg of total RNA was incubated with Cy3-ULS for 30 min at 85°C and purified to remove unbound Cy3-ULS. Labeled RNA was hybridized on miRCURY LNA microRNA arrays (probe set 8.0; Exiqon, Vedbaek, Denmark) at 60°C for 16 h using a Tecan 4800 hybridization station. Slides were washed and immediately scanned using a Tecan LS Reloaded microarray laser scanner. Microarray data extraction, normalization, and data analysis were carried out as described (30). Heatmaps were generated using the TM4 microarray software suite (31). Significance Analysis of Microarrays (SAM) analysis was carried out on sorted DC populations obtained from three biological replicates, implementing a false discovery rate (FDR) ≤ 10% and a minimum 1.5-fold change in expression. All data are MIAME compliant. Raw data have been deposited in ArrayExpress and are accessible under numbers A-MEXP-2085 and E-MEXP-3311.

The full-length Csf1r-3′UTR was cloned into the XhoI/NotI site downstream the coding sequence of Renilla luciferase in the psiCHECK-2 luciferase reporter vector (Promega). Cloning primers Csf1r-3′UTR: FW 5′-GGATTCCTCGAGTCCTGCCGCT-CTCTACGT-3′ and RE 5′-GGATTCGCGGCCGCCTGGCTGTGTTAATGCTGTT-AGTT-3′. Mutant Csf1r-3′UTR constructs were generated by introducing three basepair mismatches into each seed region of the corresponding miR-22, -34a, and -155 binding sites (Csf1r-3′UTR mut all) or the miR-22 site alone (Csf1r-3′UTR 22mut) [outsourced to Genscript (Piscataway, USA)]. HEK293T cells, described by Stewart et al. (32), were plated in a 48-well plate at a density of 6 × 10⁴ cells per well and then co-transfected the next day with 10 ng psiCHECK-2 vector containing the full 3′UTR of Csf1r mRNA, together with miR-22, -34a, -155, and control over-expression oligonucleotides (Ambion) at 50 nM final concentration using Lullaby transfection reagent (Boca Scientific). Luciferase activity was measured 48 h later using the Dual Glow luciferase kit (Promega) in a TopCount NXT microplate luminescence counter (Packard Instrument Company, Connecticut, USA). Transfections were performed in duplicate and repeated three times in independent experiments. Statistical analysis was by unpaired two-tailed Student’s t test; p-values of less than 0.05 were considered significant.

3 × 10⁵ BM cells were cultured in 12-well plates to generate GM-DC. These cells were transfected on day 4 of culture with 50 nM anti-miR microRNA inhibitors (Ambion), mixed with siGLO Cy3-labeled non-targeting anti-miR oligonucleotides (Dharmacon) at a 5:1 ratio. Control samples were treated with a control non-targeting inhibitor (Dharmacon). Transient transfection was accomplished using DharmaFECT1 reagent (Dharmacon) according to the manufacturer’s protocol.

The complete ORF of mouse Csf1r was purchased as a full-length cDNA clone (Open Biosystems, IMAGE accession no. 30436119). GM-DC (8 × 10⁶ cells) were co-electroporated at day 6 of culture with 8 μg of the Csf1r cDNA clone and 2 μg pEGFP-C1 control vector (Clontech) using an Amaxa nucleofector apparatus (Lonza; program Y-01) and Amaxa mouse macrophage nucleofector kit (Lonza) according to the manufacturer’s instructions. For control experiments, cells were co-electroporated with 8 μg pCMV-SPORT6 vector and 2 μg pEGFP-C1 vector.

The quantification of mature miR-155, -34a, and -22 expression levels was carried out with 1 μg of total RNA using the miScript PCR System (Qiagen) according to the manufacturer’s instructions. ΔCt values for each microRNA were normalized to tubulin reference gene. miR-155 FW primer: 5′-TTAATGCTAATTGTGATAGGGG-3′. miR-34a FW primer: 5′-TGGCAGTGTCTTAGCTGGTTGT-3′. miR-22 FW primer: 5′-AAGCTGCCAGTTGAAGAACTGT-3′. Tubulin primers: FW 5′-CAGACCAACCACT-GCTACAT-3′ and RE 5′-AGGGAATGAAGTTGGCCAGT-3′.

FITC painting of mice was performed as described previously (33). Briefly, mice were painted on the shaved back with 250 μl of 1% FITC (Sigma) in 1:1 acetone:dibutylphthalate (Sigma) and draining axillary and brachial lymph nodes (LN) were collected 24 h afterward. Mesenteric LN were taken as a control. Cells were isolated by mechanical disruption of the LN, without enzymatic treatment, and stained with CD301/mMGL (ER-MP23), CD11b, CD86, anti-MHC class II I-A/I-E, and CD115 antibodies and analyzed by flow cytometry. To assess expression of M-CSFR expression in myeloid cells in the skin, ears from 8 to 13-week-old C57BL/6 mice were collected and frozen in Tissue-Tek OCT embedding medium (Sakura Finetek, Zoeterwoude, Netherlands) and cut into 6 μm-thick sections. Cryosections were fixed and stained as described earlier (33). Optimally titrated goat-anti-rat IgG-Alexa546 and streptavidin-Alexa633 were used to detect unlabeled antibodies and biotinylated antibodies, respectively. Images were acquired using a Leica TCS SP5 confocal microscope.

Statistical analysis between experimental and control groups was carried out using unpaired two-tailed Student’s t test (unless stated otherwise) with the Graphpad Prism 5 software package. P-values of less than 0.05 were considered significant. Error bars represent mean ± SEM from at least three experiments.

To investigate which microRNAs are differentially expressed during monocyte-derived DC development in vitro, we performed microRNA profiling of distinct DC maturation stages isolated from 7 days GM-CSF-stimulated BM cultures, which were either or not additionally stimulated for 16 h with LPS (i.e., forced vs. spontaneous maturation). Different populations of GM-DC were sorted based on differential expression of maturation markers CD11c, MHC class II, and CD86 (Figure 1A). We then performed profiling of 328 different microRNAs using locked nucleic acid-based microRNA arrays. In total, 14 microRNAs were found to be differentially expressed in iDC to mDC development (Significance Analysis of Microarrays criteria: FDR ≤ 10%; fold change ≥ 1.5 or ≤−1.5, Figure 1B). The changes during spontaneous or LPS-induced maturation appeared to be very similar. Of all microRNAs that were screened, miR-155 showed the most abundant increase, reaching 11- and 17-fold up-regulation in the transition of iDC to mDC or iDC to LPS-mDC, respectively. These findings support data from Ceppi et al. (34) who found high levels of miR-155 in human monocyte-derived DC upon LPS stimulation. Conversely, miR-200b and -215 levels were down-regulated approximately threefold in both mDC and LPS-mDC compared to iDC. These results demonstrate that microRNA levels change during DC development in vitro and that iDC and mDC are characterized by distinct microRNA expression profiles.

To investigate which developmental genes are regulated by microRNAs in the final maturation step from iDC to mDC we compared the list of differentially expressed microRNAs with one compiled for genes known to be involved in the development of myeloid cells (DC/macrophage/neutrophil) as listed in the KEGG (Kyoto Encyclopedia of Genes and Genomes) database (35) (KEGG entry: mmu04640; Hematopoietic cell lineage – Mus musculus). Using the Targetscan algorithm (36), we identified that 10 out of these 30 genes have predicted microRNA target sites in their 3′UTR conserved across mammals. Subsequently, we compared these 10 genes to our microRNA profiling data (Figure 1B), and found that only Csf1r, the gene encoding M-CSFR, and Kitl (SCF, stem cell factor) were potentially regulated by microRNAs differentially expressed in GM-DC. From these, M-CSFR was the most likely target for microRNA regulation in iDC to mDC transition as three out of four conserved predicted binding sites in the 3′UTR of Csf1r mRNA were targeted by differentially expressed microRNAs, i.e., miR-22, -34a, and -155 (Figure 2A). Other microRNA target prediction algorithms (incl. PicTar, EIMMo) confirmed the miR-22, -34a, and -155 binding sites in both human and mouse Csf1r-3′UTR (not shown). Interestingly, expression of all three microRNAs was up-regulated upon iDC to (LPS-) mDC transition in our array, which we could confirm by quantitative RT-PCR (Figure 2B). In accordance, we found a strong down-regulation of Csf1r mRNA levels in mDC compared to iDC (Figure 2C).

To test whether miR-22, -34a, and -155 actually can regulate Csf1r expression through direct 3′UTR interactions, we cloned the complete 3′UTR of Csf1r into the psiCHECK-2 reporter vector downstream the coding sequence of Renilla luciferase. This vector, which also encoded firefly luciferase as an internal control, was transfected into HEK293T cells. When miR-155- or miR-34a precursor microRNAs were co-transfected we observed that Renilla luciferase expression was repressed approximately 50% compared to controls (Figure 2D). Co-transfection of miR-22 precursor inhibited Renilla luciferase expression even almost 75%. The combination of all three microRNAs did not further down-regulate luciferase expression. These findings confirm and extend the work of O’Connell et al. (37) and Lu et al. (24) who have shown that Csf1r is a validated target of miR-155. To verify whether the repressive effects were microRNA-specific, we repeated these experiments with Csf1r-3′UTR reporter constructs harboring point mutations in all three (miR-22, -34a, and -155) microRNA binding sites or in just one (miR-22) microRNA binding site. Indeed, no repressive effects of microRNAs could be observed when mutants were tested containing alterations in all binding sites, indicating the specificity of microRNA targeting (Figure 2E). Mutation of the miR-22 binding site alone (but not the miR-34a and -155 binding sites) completely impaired miR-22-mediated Csf1r-3′UTR repression, whereas the repressive effects of miR-34a and -155 remained unaffected, with similar levels of repression as in our initial experiments with the wild-type mouse Csf1r-3′UTR (Figure 2E).

Taken together, these data demonstrate that miR-22, -34a, and -155 are up-regulated in the final step of iDC to mDC maturation and that all three microRNAs can directly regulate Csf1r expression by targeting its mRNA via 3′UTR interactions.

Based on the differential expression of Csf1r-regulating microRNAs and their Csf1r target mRNA, we predicted that M-CSFR protein expression decreased during the transition of iDC to mDC. Therefore, we followed M-CSFR expression in GM-CSF-stimulated BM cultures over time by flow cytometry. To induce final maturation, part of the cultures was treated with LPS on day 6. Indeed, M-CSFR expression was rapidly down-regulated during both spontaneous and LPS-induced transition of iDC to mDC, accompanied by the acquisition of high level MHC class II and CD86 expression (Figure 3A).

Although our results suggest that microRNAs are responsible for the observed down-regulation of M-CSFR expression, we cannot exclude other mechanisms that might also be involved in this process. In particular, LPS is known to stimulate activation of the transmembrane protease TNF-α-converting enzyme (TACE), which cleaves M-CSFR from the surface (38). To study a putative contribution of TACE-mediated M-CSFR cleavage to decreasing expression during final maturation, GM-DC were incubated with and without TACE inhibitors TMI-1 and -2 (28, 29) before enforcing DC maturation with LPS during overnight culture. We found that M-CSFR levels decreased significantly in LPS-stimulated DC compared to unstimulated cells, despite inhibition of TACE (Figure 3B). TMI-1 and -2 treatment of GM-DC by itself already caused a decrease in M-CSFR expression, which was not caused by decreased cell viability (not shown). The efficacy of the TACE inhibitors under these conditions is indicated by the inhibition of spontaneous and LPS-induced decrease of membrane-bound M-CSFR on BM monocytes [compare left panels in Figure 3C to approximately 80% M-CSFR-expressing BM monocytes upon isolation (39)] (Figure 3C). Collectively, these data suggest that TACE-mediated shedding does not play a major role in reducing M-CSFR expression on maturing GM-DC.

To investigate whether down-regulation of surface M-CSFR expression is restricted to maturing GM-DC, we cultured freshly isolated BM in the presence of Flt3L to generate cDC (FL-cDC) and pDC and assessed surface M-CSFR levels by flow cytometry. We observed that only a minor subset of CD11c⁺SiglecH⁻ FL-cDC and CD11c⁺SiglecH⁺ pDC expressed M-CSFR (Figure 3D). Although enforced maturation with CpG in these cultures led to a further reduction in M-CSFR levels, we conclude that M-CSFR expression by iDC and down-regulation during their maturation is not a universal characteristic of Flt3L-generated cDC and pDC in vitro.

Next, we asked whether M-CSFR expression is regulated in DC in vivo as well. To that end, we focused on DC present in the skin, and assessed their M-CSFR expression in situ and after stimulated migration to skin-draining LN. The skin harbors various cells of myeloid origin, such as Langerhans cells in the epidermis, and macrophages/iDC in the dermis (1, 33, 40). Both epidermal and dermal populations of cells are able to incorporate skin-applied antigens and migrate subsequently to the draining LN, where they appear as phenotypically mDC (33, 41). Although flow cytometric analysis is the preferred method to quantify protein expression in mixed cell populations, we observed that the sample preparation of mouse skin tissue into single-cell suspensions using enzymatic digestion did not allow the use of flow cytometry due to enzymatic cleavage of surface-bound M-CSFR (not shown). Therefore, we analyzed M-CSFR expression by skin mononuclear phagocytes in situ by less sensitive confocal microscopy of tissue sections. We labeled mouse ear skin cryosections with CD115 – an antibody directed against M-CSFR – together with mMGL (ER-MP23) and CD11b antibodies to identify all dermal macrophages and iDC of the partially M-CSFR-dependent CD103^neg subset (7) and analyzed expression of these markers (Figure 3E). Cells that stain positively for the C-type lectin mMGL (CD301) are dermal macrophages and iDC that migrate to draining LN upon antigen uptake (33, 42). The images demonstrate that CD115 co-localizes with CD11b in the skin epidermis as well as with CD11b and mMGL in the dermis of the mouse ear. Similar observations were made in sections taken from mouse back skin (not shown). This indicates that both epidermal Langerhans cells and dermal mononuclear phagocytes uniformly express M-CSFR in situ, in agreement with previous findings by Hume et al. at the mRNA level (43).

To assess whether these CD115-positive mononuclear phagocytes in the skin down-regulate expression upon maturation during migration to draining LN, we skin-painted FITC onto shaved back skin of mice and analyzed the phenotype of the FITC⁺ cells that had emigrated from the skin to skin-draining LN 24 h later (Figure 3F). The majority of LN-immigrating, FITC-transporting cells from the skin express mMGL (ER-MP23) at a high level, reminiscent of their dermal origin. These findings are in agreement with our previous results (33) and with those of Irimura and colleagues (42). The skin-derived cells, expressing CD115 in situ, have lost CD115 expression upon LN arrival. Furthermore, FITC-positive LN cells have retained CD11b at an intermediate or low level, related to a DC nature, rather than the high level CD11b⁺ myelomonocytic cells, which are FITC-negative and partially M-CSFR⁺ (not shown). In agreement with their DC identity, gated FITC-positive LN cells show high levels of MHC class II and CD86 expression.

Together, our data demonstrate that M-CSFR protein is expressed by DC at the precursor and iDC stage, but down-regulated in the final stage of DC maturation. We show this during in vitro maturation of DC derived from BM precursors stimulated with GM-CSF, and our in vivo findings on maturation of connective tissue DC that migrate to skin-draining LN support this notion. In agreement with this, Cheong et al. have shown that monocytes down-regulate M-CSFR expression upon maturation to inflammatory LN DC in vivo (44).

At this point, it is reasonable to hypothesize that up-regulation of miR-22, -34a, and -155 is causally involved with Csf1r down-regulation. To substantiate this, we investigated whether M-CSFR protein down-regulation in maturing DC in vitro could be prevented by inhibiting miR-22, -34a, and -155 using microRNA inhibitor oligonucleotides. First, we used siGLO Cy3-labeled non-targeting anti-miR oligonucleotides to evaluate transfection efficiency, generally varying between 30 and 40% (Figure 4A). Then we inhibited microRNA-22, -34a, or -155 on day 4 of GM-DC culture and observed that M-CSFR down-regulation upon LPS-induced DC maturation at day 7 was significantly prevented (Figure 4B), demonstrating that miR-22, -34a, and -155 regulate M-CSFR expression in DC in the final stage of maturation. Interestingly, in these experiments we observed that not only down-regulation of M-CSFR was reduced on cells treated miR-22, -34a, or -155 inhibitors, compared to the cells transfected with control inhibitor, but that also the frequency of cells with a mDC phenotype (CD11c⁺MHCII^hiCD86^hi) was significantly lower (p < 0.0001) (Figure 4C). This suggests that these microRNAs are directly involved with the final step in DC maturation. To investigate whether microRNA-mediated M-CSFR down-regulation was functionally involved in this step, we enforced Csf1r expression in developing GM-DC using an expression vector containing the complete ORF of mouse Csf1r. Cells were co-transfected with an EGFP expression vector at a 1:4 ratio (EGFP: Csfr1) to enable selection for successfully transfected cells (Figure 4D). In these experiments we found that enforced Csf1r expression significantly (p = 0.0013) impaired up-regulation of DC maturation markers MHC class II and CD86 upon stimulation with LPS (Figures 4E,F).

In summary, these results show that in vitro inhibition of miR-22, -34a, and -155 reduces LPS-induced M-CSFR down-regulation in GM-DC. Moreover, our findings suggest that down-regulation of M-CSFR in DC is required for full DC maturation.