Voltage-Gated Proton Channel Hv1 Controls TLR9 Activation in Plasmacytoid Dendritic Cells

Plasmacytoid dendritic cells (pDCs) show a plasma cell–like morphology, with a wide and granular cytoplasm typical of secretory cells, and initiate both adaptive and innate immune responses. They present and cross-present Ags while secreting pro- and anti-inflammatory cytokines, which promotes either T cell activation or immune tolerance (1). Moreover, pDCs are a sentinel immune cell population that detects viral and bacterial infections. Their rapid release of vast amounts of type I IFN (IFN-I), along with the expression of antiviral IFN-stimulated genes (ISGs), is an essential step for the control of viral replication (1, 2). To sense pathogenic nucleic acids, pDCs are equipped with endosomal TLRs 7 and 9, which recognize ssRNA and unmethylated dsDNA, respectively (3, 4). Synthetic oligonucleotides rich in CpG regions (CpGs) can also engage TLR9 with activation of downstream signaling. Notably, to be functionally active, TLR9 requires to be processed in the endosome by pH-dependent proteases (5, 6). This is evident as uncleaved TLR9 can bind oligonucleotides but is incapable of signal transduction (7). Under steady state, full-length TLR9 locates at the endoplasmic reticulum of pDCs in its inactive form (8) and is recruited to newly formed endosomes after CpG uptake with the help of the associated protein UNC93B1 (9). Once incorporated, the acidic endosomal environment allows TLR9 cleavage, triggering a signaling cascade that phosphorylates and activates the transcription factor IRF7, leading to a rapid induction of IFN-I expression. This cellular mechanism, intended to prevent unwanted TLR9 signaling in the absence of pathogens, relies on a tight regulation of endosomal acidification to ensure strong IFN-I responses only during infection and avoiding pDC hyperactivation and autoimmunity. However, in which way the control of endosome maturation participates in pDC activation is still incompletely understood.

Rapidly after formation, the pH of early endosomes decreases because of the recruitment of the membrane-associated vacuolar H⁺-ATPase (V-ATPase) via fusion with other endocytic vesicles. The V-ATPase pumps cytosolic H⁺ into the endosomal lumen in an ATP-dependent manner and is largely responsible for the maintenance of the low intralumenar pH of this and other cellular organelles (10). However, it is known that endosomal pH is also influenced by other independent factors, such as ion influx by chloride channels or Na⁺/K⁺ ATPases (11–13). The recently discovered voltage-gated proton channel Hv1/VSOP has also been proposed as an alternative H⁺ source involved in this process.

Hv1 consists of four transmembrane regions that act as the proton permeating pore and at the same time as the voltage-sensing domain (14). Furthermore, the voltage threshold to trigger Hv1 opening is strongly influenced by changes in H⁺ concentration across the membrane, and therefore, both membrane depolarization and pH differences across membranes control the permeability of the Hv1 channel. This special property explains its cooperative function with the NADPH oxidase complex 2 (NOX2) (15, 16). Reactive oxygen species (ROS) generation by NOX2 entails electron efflux, which results in a low intracellular pH. These conditions inhibit NOX2 but, at the same time, favor the opening of Hv1, which alleviates the accumulation of H⁺ in the cytoplasm and the electrogenic gradient between the intracellular and extracellular space. This charge compensation provided by Hv1 allows sustained ROS production by NOX2 in neutrophils and macrophages, contributing to pathogen killing (17, 18). Regarding pDCs, new evidence of a ROS-dependent enhancement of CpG-induced cytokine release (19) may suggest a potential role of Hv1 in pDC activation as well.

In the current study, we show that pDCs require Hv1 for proper endosomal acidification and consecutive TLR9 signaling. We demonstrate that pDCs from Hvcn1^−/− mice secrete lower amounts of IFN-I and show a reduced induction of ISGs compared with wild-type (WT) cells. Mechanistically, although both ROS production and endosomal pH were altered in Hv1-deficient pDCs, only the latter accounts for the observed phenotype by impairing endosomal proteolysis and phosphorylation of Irf7. Together, our results reveal a decisive function of Hv1 in the control of pDC activation and IFN-I responses.

C57BL/6J WT mice were purchased from The Jackson Laboratory, and previously described Hvcn1^−/− mice (18) were provided by D. Clapham (Boston, MA) and backcrossed to C57BL/6J mice to ensure a pure background. Animals were housed under specific pathogen-free conditions at the animal facility of the University Medical Center Hamburg-Eppendorf. We used gender- and age-matched mice between 4 and 16 wk of age for each experiment, with respective littermate controls on a C57BL/6J background. Where possible, preliminary experiments were generated to determine the requirements for sample size to minimize animal numbers. Animals were assigned randomly to experimental groups. All experiments were approved by the local ethics committee (Behörde für Soziales, Familie, Gesundheit und Verbraucherschutz in Hamburg; no. 22/13, 81/15, and Org713).

EDTA blood samples were obtained from healthy human donors, and PBMCs were isolated by Ficoll density gradient centrifugation. Specific immune cell subsets were isolated using a BD FACSAria II.

Spleen and lymph nodes were prepared from WT and Hvcn1^−/− mice as follows. Spleen tissue was disrupted, passed through a 40-μm cell strainer, and resuspended in ice-cold PBS before performing erythrocyte lysis in an ammonium chloride–containing buffer (150 mM NH₄Cl, 10 mM KHCO₃, and 0.1 mM Na₂EDTA in ddH₂O [pH 7.2–7.4]). Lymph nodes were shred mechanically, and tissue was digested in collagenase A–containing buffer for 30 min.

pDCs were differentiated from bone marrow cultures by the addition of 100 ng/ml Flt3 ligand (Flt3L) for 7 d. For in vitro stimulation, pDCs were sorted as CD45⁺CD317⁺CD11c^intMHC class II^lo cells using a BD Aria II (BD Biosciences). Then, 5 × 10⁴ pDCs per well were seeded on 96-well plates and stimulated with 1.25–10 μg/ml class A or class B CpG oligonucleotides (oligodeoxynucleotide [ODN] 1585 and ODN1628; InvivoGen) for 18 h at 37°C. To inhibit the activities of V-ATPase and NOX2, 100 nM bafilomycin (InvivoGen) or 5 μM diphenyleneiodonium (DPI) (Tocris Bioscience) were added to the cultures 1 h prior to CpG stimulation. Supernatants were collected to determine the concentration of IFN-α by ELISA (Mouse IFN-α Platinum ELISA Kit; eBioscience). Cells were harvested and snap frozen for RNA isolation and gene expression analysis.

Three different genomic sequences corresponding to the peaks of the binding sites of E2-2 within and close to the coding region of the Hvcn1 gene (20) were cloned into a pGL4.12_Luc vector (provided by U. Borgmeyer, Centre for Molecular Neurobiology) upstream of its minimal promotor. HEK293 cells were then cotransfected with these plasmids together with a pcDNA TCF4-expressing vector (no. 16512; Addgene) and a Renilla luciferase reporter vector using Lipofectamine LTX (Thermo Fisher Scientific). Forty-eight hours later, cells were transferred to glass-bottom 96-well plates, and luciferase activity was measured with the help of the Dual-Glo Luciferase Assay (Promega) in a Spark 10M multimode plate reader (Tecan).

Mouse pDCs were differentiated as detailed above. After 7 d, they were transfected with pcDNA3.1-Hv1-hemagglutinin (HA) using the Mouse Dendritic Cell Nucleofector Kit (VVPA-1011; Lonza), according to the manufacturer’s protocol. Subsequently, 2 × 10⁵ cells were plated on poly-d-lysine–coated wells (5 μM, catalog no. A-003-M; Sigma-Aldrich) for 24 h and were stimulated for 20 or 90 min with 5 μg/ml murine class A CpG (ODN1585 or Oregon Green 488 labeled; InvivoGen: 21-2026-3/4; Eurofins Genomics). Cells were fixed with 4% paraformaldehyde, incubated in 10% normal donkey serum containing 0.1% Triton X-100, and subsequently, immunolabeling was performed. We used primary Abs against HA (mouse, 1:1000, MMS-101R; Covance), TLR9 (rabbit, 1:200, ab37154; Abcam), EEA1 (rabbit, 1:200, 3288; Cell Signaling Technology), Rab5 (rabbit, 1:200, 3547; Cell Signaling Technology), and LAMP1 (rabbit, 1:200, 9091; Cell Signaling Technology). For visualization, we used donkey anti-rabbit Alexa Fluor (AF) 647 (1:500, ab190565; Abcam), donkey anti-mouse AF555 (1:500, ab150110; Abcam), ROTI Mount Fluor Care DAPI for nuclei, and actin stain 488 phalloidin (1:100, category no. PHDG1-A; Cytoskeleton) for cytoskeleton. Images were acquired with a confocal laser scanning microscope (LSM 700, 63× objective; ZEISS) and processed with the ImageJ 64 software.

Colocalization was analyzed by comparing the Hv1⁺/Eea1⁺, Rab5⁺, or Lamp1⁺ double-positive area with the total Hv1 positive area (transfection and immunocytochemistry was performed as described above). First, we created a mask with the Hv1 positive area that we overlaid on the channel that recorded the respective protein of interest. Next, we measured the covered area of the protein of interest within the Hv1 mask and reported the ratio. All analysis and visualization were performed in ImageJ and the R environment.

For the analysis and sorting of immune cells by flow cytometry, cells were stained for 30 min at 4°C in the presence of Fc block (CD16/32; BioLegend). The following fluorochrome-conjugated Abs were used for the staining of human PBMCs: CD3-FITC (HIT3a; BioLegend), CD3-PerCPCy5.5 (UCHT1; BioLegend), CD4-allophycocyanin-Cy7 (RPA-T4; eBioscience), CD8-FITC (HIT8a; BioLegend), CD11c-Brilliant Violet 421 (3.9; BD Biosciences), CD14-V450 (M5E2 and RPA-T8; BD Biosciences), CD16-FITC/allophycocyanin (B73.1; BioLegend and BD Biosciences), CD19-PE-CF594/FITC (HIB19; BD Biosciences), CD20-BV711/FITC (2H7; BD Biosciences), CD56-FITC (MEM-188; BioLegend), CD123-BV605 (6H6; BioLegend), CD303-allophycocyanin (AC144; eBioscience), CD304-PE-Vio770 (AD5-17F6; Miltenyi Biotec), and HLA-DR-PE (L243; BioLegend). For the analysis and isolation of mouse immune cells, the following Abs were used: CD3-BV605 (145-2C11; BioLegend), CD4-BV786 (GK1.5; BioLegend), CD8-Pacific Blue (53-6.7; BioLegend), CD11c-PE-Cy7 (N418; BioLegend), CD19-allophycocyanin/BV605 (6D5; BioLegend), CD40-allophycocyanin (3/23; BioLegend), CD45-AF700 (30-F11; BioLegend), CD45R-PerCP-Cy5.5/BV786 (RA3-6B2; BioLegend), CD86-PerCP-Cy5.5 (GL-1; BioLegend), I-A/I-E-FITC (M5/114.15.2; BioLegend), CD317-Pacific Blue/PE (129C1 and 927; BioLegend), and PDC-TREM-PE (4A6; BioLegend). Viability was determined using the LIVE/DEAD Near-IR Fixable Stain (Thermo Fisher Scientific). If sorted from spleen, pDCs were defined as CD317⁺B220⁺CD19⁻.

Flt3L-pDCs were seeded at a density of 5 × 10⁵ cells per well on 96-well plates and stimulated with 5 μg/ml CpGs (ODN1585) labeled with the pH-insensitive dye ATTO 647N (Biomers) for 90 min at 37°C. Samples were stained with BD Viability Stain V500 as well as for their surface expression of CD45 and CD317. Mean fluorescence intensity (MFI) of ATTO 647N was measured on CD317⁺ pDCs using a BD LSR II flow cytometer (BD Biosciences).

Endosomal pH was determined using an adapted variant of an already described “pulse and chase” method (21). Biotinylated 5-nm nanoparticles (Sigma-Aldrich) were incubated with avidin-pHrodo (Thermo Fisher Scientific) or anti-biotin AF647 for 30 min at 37°C in PBS containing 1% BSA. Flt3L-pDCs were pulsed for 10 min at 37°C with the pHrodo-conjugated nanoparticles and AF647 nanoparticles, which served as an uptake control. Nanoparticle uptake and endosome maturation were stopped on ice, and cells were washed four times with ice-cold PBS/BSA to remove noninternalized particles. After washing, 1 × 10⁵ cells were chased for the indicated time points at 37°C alone or in the presence of 10 μg/ml CpGs (ODN1585). “Chase” started 10 min after particle uptake. For negative controls, incubation was performed in the presence of 100 nM bafilomycin (InvivoGen). Cells were stained with anti-CD317 Pacific Blue and anti-B220 PerCP-Cy5.5 Abs and pHrodo MFI was measured on AF647⁺ CD317⁺ B220⁺ pDCs using a BD LSR II cytometer (BD Biosciences).

To calculate endosomal pH from the recorded MFI values, a pH calibration curve (pH 4.5–7) was recorded. Flt3L–pDCs were pulsed for 10 min at 37°C with the pHrodo-conjugated nanoparticles and AF647 nanoparticles and permeabilized with 0.01% Triton X-100. Then, cells were incubated in buffers with a defined pH ranging from 4.5 to 7 for 10 min, and pHrodo MFIs for each pH value were determined by flow cytometry in the AF647⁺ gate.

Intracellular ROS was analyzed by flow cytometry using the CellROX Green Assay Kit (Thermo Fisher Scientific). Briefly, 5 × 10⁵ Flt3L–pDCs were incubated at 37°C for 90 min in the presence of 10 μg/ml CpGs (ODN1585), and CellROX Green reagent was added during the last 30 min of incubation. Fluorescently labeled anti-PDCA1 and anti-B220 Abs were added as well. Cells were washed with ice-cold PBS and analyzed using a BD LSR II device. To exclude dead cells from gating, 1 μl SYTOX was added 5 min prior to sample acquisition.

The activity of endosomal proteases was determined with the soluble Ag DQ-OVA. Then, 2 × 10⁵ Flt3L–pDCs were stimulated for 30 min at 37°C with 10 μg/ml CpGs (ODN1585). The same number of cells was incubated in parallel at 4°C in complete RPMI 1640 media supplemented with 0.1% sodium azide. DQ-OVA (100 ng/ml) was added to the culture medium for 30 min, and the reaction was stopped by adding ice-cold PBS. Cells were stained with fluorescently labeled Abs, and median fluorescence intensity of DQ-OVA was measured in CD317⁺ pDCs.

Then, 1 × 10⁶ Flt3L–pDCs were stimulated with 10 μg/ml class A oligomeric CpGs (CpGA) in complete RPMI 1640 media or left untreated for 1 h at 37°C. After incubation, cells were immediately fixed with equal volumes of Cytofix Buffer (BD Biosciences) for 10 min at 37°C and permeabilized with Perm Buffer III for 30 min on ice (BD Biosciences). Cells were stained with phospho-IRF7 Ab (1:300, D6M2I; Cell Signaling Technology) and species-specific Ab in AF555 (1:2000; Abcam). As a control, we included unstained samples and samples stained with rabbit IgG (1:300; Cell Signaling Technology) and secondary Ab. Fluorescence intensities were acquired by flow cytometry using a BD LSR II device.

WT and Hvcn1^−/− mice were injected i.v. with 10 μg CpGs (ODN 2216) conjugated with 45 μg of 1,2-dioleoyl-3-trimethylammonium propane (DOTAP). Eight hours postinjection, splenocytes were isolated as previously described and analyzed by flow cytometry. From the remaining cells, pDCs and B cells were sorted on the basis of their CD19, B220, and CD317 surface expression using a BD FACSAria II. Cell pellets were snap frozen for RNA isolation and gene expression analysis by quantitative PCR.

RNA was isolated from snap-frozen cell pellets using the RNeasy Micro Kit (QIAGEN), according to the manufacturer’s instructions, and cDNA was synthesized using the RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). mRNA expression was analyzed using TaqMan Gene Expression Master Mix and TaqMan Gene Expression Assays (Thermo Fisher Scientific) on an ABI Prism (Applied Biosystems). Tbp/TBP was used as housekeeping gene for relative quantification unless stated differently.

Data are expressed as mean ± SEM and were analyzed by unpaired Student t test or by multiple Student t tests, using the Holm–Sidak method to determine statistical significance (*p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001).

To survey the differential expression of Hv1 in immune cell subsets, we sorted defined immune cells from the spleen and lymph nodes of WT mice as well as PBMCs from human healthy donors and analyzed them by quantitative RT-PCR. We detected the highest expression levels of HVCN1/Hvcn1 in B cells, whereas expression in T cells was barely detectable (Fig. 1A, 1B). Notably, we detected a high expression of HVCN1/Hvcn1 in pDCs. In mouse pDCs, Hvcn1 expression was even higher than in conventional dendritic cells (cDCs) or migratory dendritic cells (Fig. 1A).

Because the expression of several proteins in pDCs is known to be regulated by the master transcription factor E2-2 (22), we sought to investigate whether HVCN1 expression is subject to E2-2–mediated regulation. We mined an available chromatin immunoprecipitation data set for binding sites of E2-2 within the human HVCN1 gene (20). Our analysis revealed three distinct binding sites within and close to the coding region of HVCN1 (Fig. 1C). To validate the binding of E2-2 to these regions, we cloned their individual sequences into a luciferase expression vector downstream to a minimal promotor and cotransfected HEK293 cells with each of these constructs together with a plasmid that constitutively expressed E2-2. Because the minimal promotor alone was not strong enough to induce Luc expression, we could barely detect any luciferase signal in the absence of the putative HVCN1 binding site. In contrast, the presence of the predicted E2-2 binding sites of HVCN1 resulted in a strong enhancement of luciferase activity, even exceeding the activity driven by the E2-2 binding site of IRF7, which we used as a positive control (Fig. 1D). These results indicate that E2-2 binds to these genomic regions and that HVCN1 expression is driven by the same transcription factor that determines pDC lineage.

As the function of Hvcn1 in pDCs was unknown, we next investigated whether Hvcn1 expression levels in mouse pDCs were modulated during pDC activation by TLR9 engagement. Depending on the oligonucleotide sequence and the endosome maturation state, this can trigger two different signaling pathways. CpGA bind to TLR9 in early endosomes and mainly mediate IFN-I secretion, whereas activation by class B monomeric CpGs (CpGB) occurs in late endosomes and enhances IL-6 and TNF-α production and Ag presentation (23). Stimulating pDCs sorted from Flt3L–bone marrow–derived dendritic cell cultures (Flt3L–pDCs) with CpGA resulted in a 2-fold induction of Hvcn1 transcripts, whereas stimulation with CpGB had no effect on Hvcn1 expression (Fig. 1E). Together, Hv1 expression is controlled by E2-2 and induced by CpGA, implying that Hv1 might participate in CpGA-mediated IFN-I responses in pDCs.

As Hvcn1 expression increases with CpGA stimulation, we hypothesized a role of voltage-gated proton channels in pDC activation. To investigate that, we analyzed IFN-α release by Flt3L–pDCs after treatment with increasing concentrations of CpGA. Although WT pDCs showed a robust dose-dependent IFN-α response, IFN-α secretion by Hv1-deficient pDCs was markedly reduced (Fig. 2A). Additionally, we analyzed the expression of the ISGs Irf7, Mx1, and Oas3 in CpGA-treated pDCs (24). Consistent with the lower concentration of IFN-α in the supernatant, the induction of Irf7, Mx1, and Oas3 in Hv1-deficient pDCs was significantly reduced compared with WT pDCs (Fig. 2B–D). Importantly, this was not due to an altered autocrine feedback response to IFN-I itself because Hv1-deficient pDCs showed a comparable ISG expression when stimulated with rIFN-α in vitro (Supplemental Fig. 1A). Moreover, the extent of stimulation-induced cell death was similar in WT and Hvcn1^−/− samples (Supplemental Fig. 1B). To investigate whether other TLR9-induced activation mechanisms were also Hv1 dependent, we stimulated Flt3L–pDCs with increasing concentrations of CpGBs. TNF-α and IL-6 concentrations in the culture medium and surface expression of the costimulatory receptor CD40 were lower in Hv1-deficient pDCs compared with WT control pDCs, whereas CD86 expression levels remained unaltered (Supplemental Fig. 1C, 1D). These results show that Hv1 is necessary for the strong IFN-I induction and proinflammatory response of pDCs after activation with TLR9 ligands.

As Hvcn1^−/− pDCs show deficient IFN-I secretion, we next investigated whether Hvcn1^−/− mice also display deficient antiviral pDC responses in vivo. We mimicked virus infection by injecting CpGA conjugated with DOTAP, a liposomal transfection reagent that facilitates DNA uptake. Eight hours after injection we detected a significantly reduced proportion of CD317⁺ pDCs in the spleen of Hvcn1^−/− mice compared with their WT littermates (Fig. 3A). Similarly, surface expression of PDC-TREM, an activation marker for IFN-α–producing pDCs (25), was reduced on Hvcn1^−/− pDCs (Fig. 3B), indicating that Hv1-deficient pDCs were less activated by CpGA than WT pDCs in vivo. Corroborating this finding, expression analysis of sorted pDCs from the spleens of the same mice revealed a significantly reduced induction of Irf7 and the antiviral gene Oas3 and a tendency to reduced Mx1 expression (Fig. 3C). Together, our results show that pDCs in Hvcn1^−/− mice are unable to mount a complete IFN-I response after CpGA challenge.

To mechanistically understand the deficient IFN-I response in the absence of Hv1, we next investigated the intracellular localization of Hv1 in pDCs. Therefore, we transfected Flt3L–pDCs with a Hvcn1-expressing construct that revealed in immunocytochemistry stainings a Hv1-specific signal in the cytosol of transfected pDCs. Notably, the Hv1 signal clustered in intracellular vesicular structures 20 and 90 min after CpGA stimulation and colocalized with the early endosomal markers EEA1 and Rab5 (Fig. 4A, 4B). In contrast, staining with the late endosomal marker LAMP1 showed only limited colocalization with Hv1 20 min after stimulation that significantly increased after 90 min (Fig. 4C). Hence, our analysis suggested that in pDCs, Hv1 is recruited to endosomes after CpGA uptake and consecutive activation and is retained there during endosomal maturation.

As we found Hv1 in endosomes, we reasoned that Hv1 might act together with other protein complexes involved in endosomal formation and maturation, namely the V-ATPase and NOX2 (Fig. 5A). Therefore, we next investigated how the absence of Hv1 would affect endosomal homeostasis and analyzed each one of these elements in more detail.

First, we ruled out that the observed differences in pDC activation were due to an impaired CpG uptake. We monitored the endocytosis of ATTO 647–labeled CpGs by flow cytometry and found that the percentage of ATTO 647–positive pDCs and their ATTO 647 MFI were comparable between WT and Hvcn1^−/− Flt3L–pDCs (Fig. 5B, Supplemental Fig. 2A). We confirmed the internalization of CpGA and the intracellular colocalization with Hv1 in Hv1-HA transfected Flt3L–pDCs, which were stimulated with labeled CpGAs (Supplemental Fig. 2B).

Because regulation of pH by Hv1 had previously been postulated for other immune cell types (17), we investigated whether Hv1 expression was required for endosomal acidification in pDCs. To determine their endosomal pH, we pulsed WT and Hv1-deficient Flt3L–pDCs for 10 min with nanoparticles coupled to the pH-sensitive dye pHrodo, which because of their size and lack of specific receptors are also internalized via endopinocytosis in a nonspecific fashion (26). Nanoparticles coupled to the pH-insensitive dye AF647 served as an uptake control. By extrapolation of the recorded pHrodo intensities from a pH standard curve (Supplemental Fig. 2C, 2D), we monitored the endosomal pH at different time points after uptake and observed higher pH in Hv1-deficient pDCs compared with WT pDCs 10 and 40 min after particle uptake (both p < 0.001), with approximation of pH values from both genotypes at later time points (Fig. 5C). These data indicate that early endosomal acidification is impaired in the absence of Hv1. CpGA stimulation promoted further endosomal acidification in both genotypes, yet the pH remained significantly higher in Hvcn1^−/− cells (pH 5.88 ± 0.07) compared with WT pDCs (pH 5.60 ± 0.07; Fig. 5D) at the 20-min chase time. Pretreatment with the V-ATPase inhibitor bafilomycin A, in contrast, raised the endosomal pH in both WT and Hvcn1^−/− pDCs and equalized the differences between both genotypes. pHrodo measurements have been reported to be influenced by the redox environment of the phagosomes (27). Although in the pH range of pDCs endosomes these differences should be minimal, we complemented our pH measurements by using the alternative pH-sensitive fluorophore Oregon Green succinimidyl ester, which presents a stable fluorescence that is independent of NOX2 activity (27). Using this dye, we confirmed that the acidification of Hv1-deficient endosomes was impaired in comparison with WT endosomes (Supplemental Fig. 2E).

Notably, the fact that acidification in WT as well as in Hv1-deficient cells is completely abolished by treatment with bafilomycin A, a specific inhibitor of the V-ATPase (Fig. 5D), highlights that V-ATPase is strictly required to lower the pH in the endosome in pDCs. We analyzed whether Hv1 deficiency was altering the expression of the V-ATPase in pDCs but could not find any differences in Atpv6e1e expression compared with WT pDCs (Fig. 5E). Thus, our results show that endosomal acidification is impaired in the absence of Hv1 without affecting V-ATPase expression, indicating that Hv1 and V-ATPase may serve different but nevertheless synergistic roles in endosomal acidification.

It is known that sustained ROS production by NOX2 often depends on Hv1 dissipation of H⁺ from the intracellular compartment (16, 18). We reasoned that ROS generation in pDCs could also be affected by the deletion of Hv1 and analyzed the amount of intracellular ROS in Flt3L–pDCs. Of note, Hv1-deficient pDCs produced higher amounts of ROS than WT pDCs after CpGA stimulation, even though Cybb, the gene encoding NOX2, was comparably expressed in pDCs of both genotypes (Fig. 5F, 5G). Pretreatment with the NOX2 inhibitor DPI blocked ROS generation in WT and Hvcn1^−/− pDCs (Fig. 5F). These results indicate that Hv1 restricts ROS production in pDCs.

Having observed that both ROS generation and endosomal acidification were altered in Hvcn1^−/− pDCs, we evaluated whether inhibition of NOX2 or the V-ATPase would directly affect pDC activation. Inhibition of the oxidative burst by DPI during CpGA stimulation reduced the expression of Mx1 to ∼50% in WT pDCs and up to 25% in Hvcn1^−/− pDCs (Fig. 5H), showing that ROS contribute to IFN-I signaling and ISG expression in pDCs. Hence, the enhanced ROS production in Hv1-deficient pDCs cannot account for their reduced activation. In contrast, Mx1 expression was almost completely abolished when acidification was blocked by bafilomycin (Fig. 5H). This illustrates that the activity of the V-ATPase is crucial for pDC responses and that the higher endosomal pH of Hvcn1^−/− pDCs probably limits their IFN-I response. Together, our results suggest that Hv1 participates in endosomal homeostasis by assisting the V-ATPase during immediate acidification and restricting ROS production by NOX2.

As endosomal acidification appears to be the pivotal factor for pDC activation via TLR9, the endosomal acidification defect in Hvcn1^−/− pDCs is suited to explain the observed IFN-I response defect. Finally, we explored the mechanistic consequences of Hv1 deletion for TLR9 activity.

First, we investigated the cellular localization of Hv1 and TLR9. By transfecting HA-tagged Hv1 into Flt3L–pDCs, we observed that Hv1 and TLR9 partially colocalized in steady-state conditions as well as after stimulation with CpGA (Fig. 6A). Tlr9 expression itself was not affected by Hv1 deficiency in pDCs (Fig. 6B), but phosphorylation of the transcription factor IRF7 was significantly reduced in Hvcn1^−/− pDCs before and after CpGA stimulation (Fig. 6C). These results indicate that endosomal Hv1 modulates TLR9 activation and its downstream signaling.

Impaired endosomal cleavage of TLRs by acidic proteases may be a consequence of a suboptimal endosomal pH. Hence, we next analyzed general endosomal protease activity by loading the cells with DQ-OVA, a self-quenched fluorescent conjugate of OVA that exhibits fluorescent emission after proteolytic cleavage. We detected that Hvcn1^−/− pDCs display a lower fluorescent emission (i.e., endosomal proteolytic activity) than WT cells (Fig. 6D). Together, these observations indicate that the activity of the proton channel Hv1 is required for endosomal acidification, which enables activation of endosomal proteases and consecutive TLR9 signaling.

Because of their multifaceted biology combining the release of a wide variety of cytokines and low-intensity, yet efficient, Ag presentation, pDCs are crucial for the early control of viral replication, formation of immunological memory, and maintenance of tolerance (28). Most immune cells express several TLRs for pathogen recognition, but pDCs primarily express the endosomal TLRs 7/8 and 9, which are mainly responsible for activation (3, 4). Therefore, the efficacy and extent of pDC responses is dependent on a tightly balanced endosomal homeostasis. Our study reveals that Hv1 expression is intrinsically linked to the pDC lineage and induced during pDC activation. We also confirmed that pDCs require Hv1 to achieve their full activation potential as Hv1-deficient pDCs displayed reduced IFN-I responses after exposure to CpGA in vitro and in vivo. After CpGB stimulation, activation of Hv1-deficient pDCs was also reduced in vitro. This observed phenotype of Hv1-deficient pDCs could be explained by 1) a reduced nucleic acid uptake in endocytic vesicles, 2) altered ROS production, or 3) changes in endosomal pH.

Although receptor-mediated internalization of DNA oligonucleotides has been debated, there is evidence supporting that CpG uptake occurs via nonspecific endocytosis (29), and we did not detect differences between WT and Hv1-deficient cells regarding this mechanism. After uptake, two critical factors that determine the fate of signaling from endosomes in pDCs are the redox and the pH environment, and both of them have been shown to be finely regulated by Hv1. We observed increased intracellular ROS in Hv1-deficient pDCs after CpGA challenge. This is in contrast to findings described for other immune cells. In neutrophils, for example, phagosomal ROS production is markedly reduced during the respiratory burst in the absence of Hv1 (18). A similar contribution of Hv1 to NOX2-dependent ROS generation has been shown in macrophages, eosinophils, B cells, and cDCs (17, 27, 30, 31). In microglia, two different studies came up with conflicting results. The first showed that, similarly to the above-mentioned immune cell types, Hv1 deletion diminished microglial ROS production and was protective against cerebral stroke (32). By contrast, another study observed that primary microglia from Hvcn1^−/− mice produced higher levels of ROS in steady state and after activation (33, 34). However, compared with neutrophils or microglia, in which high amounts of ROS are produced with a microbicidal purpose, in pDCs, ROS are present at very low levels and probably act as signaling molecules (35). Of note, ROS has recently been related to IFN-I production by pDC. Mitochondrial ROS was reported to suppress CpGA-induced production of IFN-α in the human pDC cell line Gen2.2 (36). By contrast, enhanced ROS generation increased IFN-α production in mouse pDCs activated by TLR7 ligands, regardless of whether its source was mitochondrial or the NOX2 complex (19). In line with this latter finding, we observed that inhibition of ROS production diminished ISG expression. However, in the context of our study, this seems to be an independent effect not related to the reduced IFN-I response of Hv1-deficient pDCs and is likely counteracted by the endosomal pH alterations that we found in the absence of Hv1.

Hv1 contribution to pH modulation strongly differs between cell types. Neutrophils require Hv1 to maintain a neutral phagosomal pH by compensating the negative charges produced during the oxidative burst, whereas in macrophages, with a milder phagosomal oxidative burst, Hv1 ensures a rapid acidification, essential for the digestion of the engulfed pathogens (17, 36). Moreover, the extent of acidification determines whether a specific vesicular compartment is more adapted to kill or to process engulfed material more gently for Ag presentation as excessive Ag degradation may destroy potential T cell epitopes (37). For instance, cDCs, as professional APCs, have either a neutral (37) or only slightly acidic phagosomal lumen (26), and Hv1 does not appear to play a major role modulating phagosomal pH in this cell type (26). Similarly, the pH of pDCs’ phagosomes is neutral in the steady state and turns alkaline upon TLR7 stimulation via a ROS-dependent mechanism, which is essential for cross-presentation of Ags to CD8⁺ T cells (19). Because of the higher ROS levels that we found in Hv1-deficient pDCs, we consider it unlikely that Hv1 channels participate in Ag presentation by pDCs. In contrast, we observed that pDCs’ endosomes quickly acidify after formation with the help of Hv1, reaching pH values around 6 for unstimulated cells and 5.5 after stimulation with TLR9 ligands. A recent study shows that the proteolytic activity of endosomes in pDCs is almost twice as high as in phagosomes (19), supporting the hypothesis that pDCs’ phagosomes are more prone to present Ags, whereas the more acidic endosomes ensure a more degradative environment for TLR signaling. Thus, we believe that the defective acidification that we found in Hvcn1^−/− pDCs after CpGA stimulation correlates with their reduced protease activity and is most likely responsible for their defective IFN-I response.

Integrating our data, the endosomal maturation defect that we observed can be explained by 1) the lack of proton export from the cytosol into the endosome by Hv1, 2) the higher amount of ROS, which would consume protons from the endosomal lumen to dismutate the produced oxygen radicals, and 3) a potential modulation of the V-ATPase by Hv1 or the elevated ROS production. An inhibition of V-ATPase recruitment to phagosomes by ROS has been suggested in the past, which would be a plausible explanation for the effect we observed in Hv1-deficient pDCs (17). Clearly, V-ATPases seem to be dominant drivers for acidification in pDCs, whereas Hv1 would act as a regulator of endosomal formation and maturation. However, modulation at this range of pH is of great importance for endosomal TLR signaling because it corresponds to the activation threshold of several acidic proteases present in the endosome, such as cathepsin L and asparagine endopeptidase (38, 39). These two enzymes participate in the sequential cleavage of TLR9 into its active form (5–7). Moreover, TLR9 binding affinity to DNA has been shown to be stronger at the pH between 5.5 and 6 (40), and small ssDNA degradation products only form at acidic pH (41). Consequently, the blockade of endosomal acidification inhibits TLR9 cleavage and signaling (42), and coherently, we observed that the total proteolytic activity and Irf7 phosphorylation in Hv1-deficient pDCs were significantly reduced.

Together, our data demonstrate that Hv1 works cooperatively with V-ATPases and NOX2 in pDCs, acting as a fine modulator of cellular processes in which charge compensation is needed, in this case, endosomal homeostasis. Based on this, we propose a model for pDCs in which Hv1 is required to ensure quick endosomal acidification after foreign DNA uptake, providing sufficient protein degradation to generate active TLR9 able to sense the viral nucleic acids and initiate IFN-I release (see graphical summary). Our findings describe an unrecognized function of Hv1 in early innate immunity and provide insight into the molecular mechanisms that coordinate pDC activation and antiviral IFN-I responses. In view of these results, Hv1 targeting could be an interesting approach to tune IFN-I responses in autoimmunity.

We thank David Clapham for providing Hvcn1 knockout mice and Uwe Borgmeyer for providing the luciferase plasmids and for technical advice.

The authors have no financial conflicts of interest.