Cutting Edge: IFN-β Expression in the Spleen Is Restricted to a Subpopulation of Plasmacytoid Dendritic Cells Exhibiting a Specific Immune Modulatory Transcriptome Signature

Early type I IFN expression is crucial for the initiation of antiviral immune responses. A primary source of type I IFNs is plasmacytoid dendritic cells (pDCs), which rapidly produce large amounts of these cytokines upon activation (1, 2) and, therefore, have been termed professional type I IFN producers. In general, pDCs are identified based on their specific surface marker expression profile as CD11c^intB220⁺mPDCA-1⁺CD11b⁻, but they represent a rather heterogeneous cell population. Functionally separate subsets of pDCs have been described based on the expression of CD9 versus CCR9. In the bone marrow (BM), CCR9⁻CD9⁺ pDCs were shown to produce high amounts of IFN-α following TLR9 activation (3, 4), whereas in peripheral lymph nodes, CCR9⁺ and CCR9⁻ pDCs were described to produce IFN-α (5). In contrast, Segura et al. (6) showed that splenic CCR9⁺ pDCs produced type I IFN. Therefore, it is a matter of debate which subset is responsible for the production of type I IFN. Following stimulation of TLR7 and TLR9, pDCs also serve as a source of other cytokines and chemokines (7). It remains to be shown whether IFN-β–producing pDCs are polyfunctional in coexpressing other immune effector molecules.

The present study clarifies that the attribute “professional type I IFN-producing cells” in the spleen following TLR9 stimulation has to be restricted to a distinct functional subset of pDCs that selectively facilitate T cell recruitment and activation.

C57BL/6N (B6), IFNAR1^−/−, IFNβ^mob/mob, and IFNAR1^−/−IFNβ^mob/mob mice were kept under pathogen-free conditions, and experiments were approved by the government of North-Rhine Westphalia. Where indicated, mice were infected i.p. with 2 × 10⁵ murine CMV (MCMV) (C3X) or injected i.v. with 10 μg CpG 1668 (TIB MOLBIOL) complexed to DOTAP (Roche) for 6 h, or as indicated.

BM-derived Flt3L-cultured pDCs were generated as previously described (8). For stimulation, CpG 1668 complexed to DOTAP (6 μg/ml; TIB MOLBIOL) was added for 6 h, or as indicated. Surface marker expression was analyzed by FACS (Supplemental Fig. 1A).

Cells were analyzed on a FACSCanto II (BD). DAPI was added for dead cell discrimination. FACS sorting was performed on a FACSAria (BD) after MACS (Miltenyi Biotec) depletion of CD3ε/CD19⁺ cells. Abs against B220, CD3ε, CD4, CD8, CD11b, CD11c, CD19, CD86, CD16/32, Ly6C, and NK1.1 were purchased from BD, SiglecH and mPDCA-1 were purchased from Miltenyi Biotec, CD9 (MZ3) was purchased from BioLegend, and CCR9 (eBioCW-1.2) was purchased from eBioscience.

Spleen sections were stained as described (8). Cytospins were performed on a Cellspin II centrifuge (Tharmac) and stained using the Hemacolor rapid staining kit (Merck). Images were taken on an LSM 780 (Zeiss), Axioskop 40 (Zeiss), or Eclipse TE 2000 (Nikon) microscope with a digital camera (CCD-1300; Vosskuehler) and processed using Adobe Photoshop and ZEN 2011/12 software.

BM-derived CpG-stimulated pDCs were sorted for IFNβ/YFP expression. Chemotaxis toward CCL19 and CCL21 (500 ng/ml; R&D Systems) was addressed using Transwell inserts with 5 μm pore size (Corning) and the CellTiter-Glo Luminescent Cell Viability Assay (Promega). A total of 5 × 10⁵ IFNβ/YFP⁺ or IFNβ/YFP⁻ pDCs was allowed to migrate for 4 h at 37°C. For the recruitment of leukocytes, B6 mice were injected i.p. with 2.5–5 × 10⁵ sorted IFNβ/YFP⁺ or IFNβ/YFP⁻ BM-derived pDCs.

RNA was isolated from ex vivo FACS-sorted pDCs using the mirVana miRNA Isolation Kit (Ambion) and hybridized to Agilent Whole Mouse Genome Oligo Microarrays (4 × 44K). Rosetta normalization and log2 transformation were performed. Pairwise comparison of IFNβ/YFP⁺ and IFNβ/YFP⁻ pDCs was performed using the Rosetta Resolver. Ratios were calculated by dividing sample signal intensity by control signal intensity. The normalization that was applied is based on the Rosetta error model (9). The p value is calculated from X dev, which takes into consideration the log(Ratio) and log(Error). X dev = Log(Ratio)/Log(Error). p value = 1− Erf [Abs (X dev)/√2].

Microarray data are available under Gene Expression Omnibus Series accession number GSE68788 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE68788). Significant (p ≤ 0.01) and >2-fold differentially regulated genes were subjected to multilevel ontogeny analysis (MONA) within the section “biological process” of this online tool and summarized to superordinate terms (http://mips.helmholtz-muenchen.de/mona/index.aspx). Quantitative RT-PCR (qRT-PCR) was performed using the TaqMan Master Kit with the Universal Probe Library Set (Roche) or the MESA GREEN qPCR MasterMix Plus (Eurogentec) on an iQ5 or CFX96 (Bio-Rad). The primer sequences used are shown in Supplemental Table I.

Data are given as mean + SD, Student t tests were used, and one representative of at least independent experiments with at least three mice/group is shown, unless stated otherwise.

Because the identity of IFN-β producers within the pDC population remains controversial, we sought to determine which specific subset of splenic pDCs is responsible for the early IFN-β production. For this we used the unbiased in vivo approach of IFNβ/YFP reporter mice in MCMV infection or stimulation with the synthetic TLR9 ligand CpG (8). The majority of IFNβ/YFP⁺ cells within the spleen expressed the cell surface markers CD11c, mPDCA-1, and B220 (Fig. 1A), in line with earlier reports that pDCs are the primary source of IFN-β after MCMV infection or TLR9 activation (8, 10, 11). IFNβ/YFP expression was restricted to ∼5% (4.62 ± 0.7% [mean ± SEM]) of all pDCs after TLR9 activation, as determined by flow cytometry (Fig. 1B). Dose-titration experiments indicated that the low frequencies of IFNβ/YFP⁺cells are not a linear reflection of the stimulus strength (Supplemental Fig. 1B, 1C). CD86 was upregulated on IFNβ/YFP⁺ and IFNβ/YFP⁻ pDCs following TLR9 stimulation or MCMV infection, indicating that both cell subsets are capable of being activated (Fig. 1C, Supplemental Fig. 1D, 1E). Therefore, IFN-β expression is restricted to only a small fraction of pDCs in vivo, despite similar activation of the overall pDC population. Differences in activation marker expression observed after MCMV infection (Supplemental Fig. 1E) point to a specific influence of the virus on CD86 expression on IFNβ/YFP⁻ pDCs. Also, insufficient activation of IFNβ/YFP⁻ pDCs cannot be excluded in the case of the more complex virus infection. IFNβ/YFP⁺ and IFNβ/YFP⁻ pDCs exhibited the typical plasmacytoid morphology with a round and smooth cellular body and excentered nucleus (Fig. 1D). Splenic IFNβ/YFP⁺ pDCs expressed the pDC markers mPDCA-1, B220, SiglecH, and Ly6C (Fig. 1A, 1E). Furthermore, the majority of IFN-β–producing pDCs stably expressed CCR9, but not CD9, for ≥24 h after stimulation (Fig. 1E). Taken together, although IFN-β–producing and nonproducing pDCs in the spleen upregulate surface activation markers to the same extent after CpG stimulation, IFN-β expression is restricted to a low-frequent subset of all pDCs. Furthermore, this subset expresses CCR9 but not CD9. Our data are supported by earlier findings demonstrating that CCR9⁻ cells within the spleen represent precursors of conventional dendritic cells rather than bona fide pDCs (6). Controversial reports on CCR9 and CD9 expression levels on type I IFN–producing pDCs point to organ-specific differences in pDC populations (3, 4, 6). Alternatively, the capacity to produce type I IFN might be differentially regulated, depending on the developmental stage of pDCs (4).

We investigated whether IFN-β–producing pDCs localize to specific anatomical sites within the splenic microarchitecture. In naive mice, ∼60% of all pDCs were located within the marginal zone (Fig. 2A, 2C). Following CpG stimulation, pDCs formed clusters, as reported previously (12, 13), and were distributed equally between the T and B cell zone and the marginal zone (Fig. 2A, 2D). In marked contrast, >97% of IFNβ/YFP⁺ pDCs were located within T and B cell zones (Fig. 2B, 2E). Our data demonstrate that IFN-β–producing pDCs show a specific distribution pattern within the spleen rather than being randomly interspersed within the overall splenic pDC population.

We next asked whether these IFN-β–producing pDCs harbor further specialized functional properties associated with a specific gene-expression profile. Splenic CD11c^intmPDCA-1⁺CD11b⁻ pDCs were sorted by FACS 6 h after CpG administration into IFNβ/YFP⁻ and IFNβ/YFP⁺ populations (Supplemental Fig. 1G). Microarray analyses defined 1446 genes as significantly (p ≤ 0.01) and >2-fold over- or underrepresented in IFNβ/YFP⁺ pDCs versus IFNβ/YFP⁻ pDCs (Fig. 3A). The classical pDC surface marker mPDCA-1 and hallmark transcription factor E2-2 (2) did not exhibit significantly different expression levels in IFNβ/YFP⁺ and IFNβ/YFP⁻ pDCs. Several genes associated with TLR9 stimulation were expressed similarly in IFNβ/YFP⁺ and IFNβ/YFP⁻ pDCs compared with naive pDCs (Supplemental Fig. 1H). In contrast, IFN-β, other type I IFNs, the type III IFN Il28b, Th cell–differentiation cytokines (e.g., Il12b and Il12a), and genes involved in chemotaxis (e.g., CCL3, CCL5, and CCR7) were among the highest overrepresented genes in IFNβ/YFP⁺ pDCs. This was confirmed by independent qRT-PCR analyses (Supplemental Fig. 1I). Of note, expression changes in these genes were observed in IFNβ/YFP⁺ pDCs versus IFNβ/YFP⁻ pDCs, as well as in IFNβ/YFP⁺ pDCs and IFNβ/YFP⁻ pDCs versus naive pDCs (Supplemental Fig. 1J). To fully clarify the relative differentiation and activation status of IFNβ/YFP⁺ pDCs and IFNβ/YFP⁻ pDCs versus naive pDCs, future experiments (e.g., direct comparisons of the transcriptome and proteome of these three groups) will be needed. MONA identified genes involved in immune effector processes and regulation of localization as active specifically in IFN-β–producing pDCs (Fig. 3B). In line with the differential expression of CCR7, IFNβ/YFP⁺ pDCs exhibited a greater migration toward CCL19 and CCL21 compared with IFNβ/YFP⁻ pDCs (Fig. 3C), which correlated with their specific localization within the splenic white pulp. This further indicates that IFNβ/YFP⁺ pDCs are a specialized subset that is highly capable of efficiently producing IFN-β.

Following TLR9 activation of IFNAR1^−/− mice, the majority of IFN-β–producing pDCs were located within the splenic T and B cell zones, similar to their localization in IFNAR1^+/+ animals (Figs. 2A, 2B, 4A, 4B). Furthermore, no significant differences in the percentages or absolute numbers of total and IFNβ/YFP⁺ pDCs were observed between IFNAR1^+/+ and IFNAR1^−/− mice after CpG stimulation (Fig. 4C, 4D), in line with previous studies demonstrating that type I IFN production by pDCs occurred independently of the IFNAR-mediated feedback loop (14). Also, IFN-β serum levels were similar in IFNAR1^+/+ and IFNAR1^−/− mice, whereas IFN-α levels were reduced in the absence of IFNAR1 (Supplemental Fig. 1K, 1L). To elucidate a possible involvement of type I IFN signaling in the specific gene-expression signature in IFN-β–producing pDCs, we isolated IFNβ/YFP⁺ or IFNβ/YFP⁻ pDCs from IFNAR1^−/−IFNβ^mob/mob mice by FACS and subsequently performed qRT-PCR on selected genes (Fig. 4E). The expression profile of selected immune-response genes was comparable to that of IFN-β–producing pDCs from IFNAR1^+/+ mice, demonstrating that, independent of IFNAR-mediated signaling, IFN-β–producing pDCs are equipped with an intrinsic and specialized gene-expression profile for the effective orchestration of cellular immune responses.

The gene-expression signature of IFN-β–producing pDCs demonstrates that these cells produce a broad range of type I IFNs and proinflammatory cytokines, as well as express CCL3 and CCL5 capable of promoting the recruitment of NK cells and CD4⁺ and CD8⁺ T cells. In an in vivo model in which naive untreated pDCs or sorted CpG-stimulated IFNβ/YFP⁺ or IFNβ/YFP⁻ pDCs were injected i.p. into wild-type mice, IFNβ/YFP⁺, but not IFNβ/YFP⁻, pDCs promoted the influx of greater absolute numbers of total cells into the peritoneal cavity compared with naive pDCs (Fig. 5). This differential recruitment ability pertained to CD8⁺ and CD4⁺ T cells, whereas no significant differences were noted with regard to NK cells (Fig. 5).

In this study, we identified a functionally distinct subset of IFN-β–producing pDCs with enhanced migratory capacity capable of mediating cellular immune responses. Our results indicate that, after systemic activation of TLR9, only a small subset of pDCs was responsible for the initial expression of type I IFNs in the antiviral response. The IFN-β–producing pDCs were strategically located within the T cell zone of the splenic white pulp and were equipped with a specific gene signature profile enabling them to control leukocyte recruitment and coordinate early cellular immune responses. These findings cast doubt on current models suggesting a stochastic expression of IFN-β (15, 16). A possible explanation for the differential regulation of IFN-β expression is an epigenetic blockade of the IFN-β promoter in non-IFN-β–producing pDCs, as described for mouse embryonic fibroblasts (17). None of the functions described above were dependent on IFNAR-mediated signaling but represented cell-intrinsic properties of these natural type I IFN–producing cells. These findings have to be taken into consideration in the design and interpretation of vaccination and therapy approaches involving pDC-based strategies in infectious and antitumor immunology, as well as autoimmunity.

We thank S. Kropp for technical assistance and T. Klitz for animal husbandry.

The authors have no financial conflicts of interest.