Intravesicular Acidification Regulates Lipopolysaccharide Inflammation and Tolerance through TLR4 Trafficking

Innate immunity is the first line of host defense against pathogens. Innate immune cells, including macrophages and dendritic cells, detect microbial invasions through pattern recognition receptors such as TLRs, and induce inflammatory cytokines and type I IFNs that play a major role in pathogen elimination (1–3). TLR1, TLR2, TLR4, TLR5, and TLR6 are expressed on the cell surface and preferentially recognize bacterial components such as LPSs, lipopeptides, and flagellar proteins. In contrast, TLR3, TLR7, TLR8, TLR9, TLR11, TLR12, and TLR13 are exclusively expressed in intracellular acidic organelles such as endosomes and lysosomes where they encounter nucleic acids derived from viruses or bacteria engulfed through endocytic or phagocytic pathways. The intracellular localization of these TLRs is suggested to be a safeguard to avoid inappropriate contact with self-nucleic acids released extracellularly, which could otherwise lead to autoimmunity (4, 5). Notably, TLRs traffic from the endoplasmic reticulum through Golgi and are sorted to endosomes (6). Several molecules such as UNC93B1, gp96, HSP90, PRAT4A, AP2, and AP4 have been identified to regulate this process (7–11). In endosomes, the ectodomains of TLR3, TLR7, TLR8, and TLR9 are proteolytically processed by endosomal proteases such as cathepsins and asparagine endopeptidase, and TLRs mature to functional receptors that can initiate downstream signaling (12–14).

TLR4, a receptor for LPS, has a unique property by mobilizing between the plasma membrane and intracellular vesicles (15). LPS initially binds to LPS binding protein (LBP) in the extracellular fluids and then LPS–LBP complex is recognized by cluster of differentiation 14 (CD14), a GPI-anchored protein (16). The CD14 complex is transferred to MD2–TLR4 heterodimer complex, which then transmits signals from the cytoplasmic region of TLR4, namely the Toll/IL-1R (TIR) domain, by recruiting the TIR domain–containing adaptor protein (TIRAP), which is accumulated to the membrane proximal microdomain similar to lipid rafts (16, 17). CD14 recognition of LPS activates the lipid kinase phosphatidylinositol 4 phosphate 5 kinase (PIPK5) Iα and Iβ, which generates plasma membrane PI(4,5)P₂, which is responsible for recruitment of TIRAP through its lipid-binding domain (17, 18). Moreover, the small GTPase ARF6 is also reported to regulate the activation of PIPK5 and contribute to TIRAP relocation to the lipid rafts (19, 20). TIRAP subsequently recruits a protein complex called myddosome, which contains another TIR domain–containing adaptor, MyD88, and the IRAK family protein kinases, which eventually activate the transcription factor NF-κB to drive mRNA expression of an array of inflammatory cytokines (21). Following LPS recognition, TLR4–LPS complex is shown to be internalized from the plasma membrane to endosomes in a manner dependent on CD14, where it recruits another set of TIR domain–containing adaptors: TIR domain–containing adaptor inducing IFN-β (TRIF) and TRIF-related adaptor molecule (TRAM) (22). TRIF then activates signaling pathways leading to the late phase activation of NF-κB as well as the activation of the transcription factor IFN regulatory factor 3 (IRF3) that regulates induction of type I IFN (22). Thus, TLR4 signals through different cellular compartments: plasma membrane that activates TIRAP-MyD88–dependent pathway and endosome that activates TRAM-TRIF–dependent pathway, both of which are required for a robust induction of antibacterial innate immune responses (23–25). Whereas the TLR4-dependent pathway plays an essential role in LPS-triggered septic shock (26), innate immune cells from septic patients or mice treated with a sublethal dose of LPS challenge are hyporesponsive to subsequent challenge by LPS (27, 28). This hyporesponsiveness is defined as LPS tolerance and considered as a mechanism to prevent excessive inflammation and septic shock caused by LPS or bacterial infection. TLR4 internalization is therefore thought to be a mechanism that prevents uncontrolled inflammation and septic shock, whereas other mechanisms including upregulation of numerous anti-inflammatory molecules, epigenetic regulations, and miRNA-mediated suppressions are essential for induction of LPS tolerance (27, 28).

Vacuolar-type H⁺ adenosine triphosphatase (V-ATPase) is a large complex of ATP-dependent proton pumps, which localizes at endosomes, lysosomes, Golgi, secretory vesicles, or the plasma membrane, and functions to acidify intracellular compartments and regulate protein processing and degradation, ligand-receptor liberation, protein transport, and elimination of pathogens such as bacteria and viruses (29). It consists of a peripheral domain (V1) of eight subunits (A–H), which hydrolyze ATP, and an integral domain (V0) of six subunits (a, d, c, c', c'', and e) that translocate protons through the operation of a rotary mechanism (29). Although acidification in intracellular vesicles, where nucleic acid-sensing TLRs localize, is linked to their function in the context of signal transduction and ligand recognition, the physiological relevance of V-ATPase in immune responses through TLRs remains unclear.

In this study, we established a macrophage cell line lacking ATP6V0D2, a component of V-ATPase that plays an important role in the junction between the V1 domain and the V0 domain (29). ATP6V0D2 is highly expressed in osteoclasts and dendritic cells compared with ATP6V0D1, an isoform of ATP6V0D2, and ATP6V0D2-deficient mice are shown to have neutralized acidic organelles and exhibit osteopetrosis because of impaired bone resorption caused by osteoclasts (30, 31). We found that V-ATPase–dependent acidification is required for TLR3-, 7-, and 9-dependent type I IFN responses. Interestingly, ATP6V0D2 deficiency enhanced inflammatory responses and reduced type I IFN responses to LPS and resulted in defects of LPS tolerance because of prolonged TLR4 expression on the cell surface and NF-κB activation. Moreover, we provide evidence suggesting that ARF6 escorts cell surface TLR4 to intracellular compartments by binding to V-ATPase complex. Thus, ATP6V0D2-dependent acidification of intracellular compartments requires TLR4 internalization, which links to suppression of LPS-triggered inflammation and tolerance.

Wild-type (WT) C57BL/6 mice (6–8 wk of age) were obtained from CLEA Japan. All animal maintenance and experiments were performed in accordance with the guidelines of the Committee on Animal Research at Nara Institute of Science and Technology. Bone marrow–derived macrophages (BMMs) and bone marrow–derived dendritic cells (BMDCs) were obtained from mouse bone marrow cells cultured in RPMI 1640 medium (Nacalai Tesque) supplemented with 10% FBS (Life Technologies), 1% penicillin-streptomycin mixed solution (Nacalai Tesque), 100 μM 2-ME (Nacalai Tesque), and 10 ng/ml murine M-CSF (BD Biosciences) or 10 ng/ml murine granulocyte M-CSF (BD Biosciences) for 5–7 d in a 5% CO₂ incubator.

RAW264.7 cells, HEK293 cells, HEK293T cells, HEK293 cells expressing TLR4 (InvivoGen), and Plat-E cells were cultured in DMEM (Nacalai Tesque) supplemented with 10% heat-inactivated FBS in a 5% CO₂ incubator. LPS, polyinosinic polycytidylic acid (poly(I:C)), rhodamine-labeled poly(I:C), R837, and ODN1668 were purchased from InvivoGen and Alexa Fluor 488–labeled LPS was purchased from Life Technologies. IFN stimulatory DNA (ISD) was synthesized (Grainer Japan) and annealed (5′-TACAGATCTACTAGTGATCTATGACTGATCTGTACATGATCTACA-3′). Poly(I:C) and ISD were mixed with Lipofectamine 2000 (Life Technologies) at a ratio of 1:1 (μg:μl) in Opti-MEM (Life Technologies) for intracellular stimulation. Acridine orange and bafilomycin A1 (Baf A1) were purchased from Waldeck. cDNAs for ATP6V0D2, ATP6V0C, and individual ARF family members were amplified by PCR from murine brain and lung cDNAs and inserted into the pFLAG-CMV-2 Expression Vector (Sigma) pcDNA3 that contained a Myc-tag sequence (Santa Cruz Biotechnology). The reporter plasmids for Ifnb1 and NF-κB promoters were described elsewhere.

To establish frameshift mutations in ATP6V0D2, ATP6V1B2, or ARF6, guide sequences located in exon 2 of Atp6v0d2, exon 1 of Atp6v1b2, and exon 2 of Arf6 were inserted into pX330-U6-Chimeric_BB-CBh-hSpCas9 (#42230; Addgene) that expresses Cas9 and guide RNA (Atp6v0d2: sense 5′-GAAAATTCATCTCCAGACCA-3′; Atp6v1b2: sense 5′-GAACTGCCCGTGCCCACCGG-3′; Arf6: sense 5′-GCGGATCCTCATGCTGGGCC-3′). Genomic regions containing guide sequences were inserted into the pCAG EGxxFP plasmid (#50716; Addgene), which is a reporter for genome editing, and coelectroporated into RAW264.7 cells by Neon (Invitrogen). GFP-positive cells were sorted by a BD FACSAria (BD Biosciences), seeded in a 96-well plate, and cultured for 2 wk. DNA was isolated from expanded cells, and the mutations were analyzed by DNA sequencing.

RAW264.7 cells were seeded in six-well plates and treated with 5 μg/ml acridine orange for 5 min. The cells were then excited at 488 nm and emission was detected at 520 and 620 nm for the internal marker (520 nm; green) and endosomal acidification (620 nm; red) by LSM 700 (Zeiss). The fluorescence intensities were measured using a BD FACSAria for quantification.

Total RNA was isolated with TRIzol reagent (Invitrogen) and reverse transcribed with ReverTra Ace (Toyobo) according to the manufacturers’ instructions. RT-PCR and quantitative RT-PCR (RT-qPCR) were performed with the following primers: mIfnb1, sense 5′-ATGGTGGTCCGAGCAGAGAT-3′, reverse 5′-CCACCACTCATTCTGAGGCA-3′; mIl6, sense 5′-GTAGCTATGGTACTCCAGAAGAC-3′, reverse 5′-ACGATGATGCACTTGCAGAA-3′; mIl12p40, sense 5′-AGACCCTGCCCATTGAACTG-3′, reverse 5′-GAAGCTGGTGCTGTAGTTCTCATATT-3′; mTnfa, sense 5′-CACAGAAAGCATGATCCGCGACGT-3′, reverse 5′-CGGCAGAGAGGAGGTTGACTTTCT-3′; mAtp6v0d2, sense 5′-TCAGATCTCTTCAAGGCTGTGCTG-3′, reverse 5′-GTGCCAAATGAGTTCAGAGTGATG-3′; mAtp6v1b2, sense 5′-CGAACTGTTTATGAGACTTTGGACATT-3′, reverse 5′-GGTGCTGAGGGATTCTCTTC-3′; mArf6, sense 5′-GAGCTGCACCGCATTATCAA-3′, reverse 5′-TGCTTGTTGGCGAAGATGAG-3′; m18s rRNA, sense 5′-GTAACCCGTTGAACCCCATT-3′, reverse 5′-CCATCCAATCGGTAGTAGCG-3′; hATP6V0D2, sense 5′-CTTGAGTTTGAGGCCGACAG-3′, reverse 5′-TGCCGAAGGTTGGATAGAGG-3′; hGAPDH, sense 5′-AATCCCATCACCATCTTCCA-3′, reverse 5′-TGGACTCCACGACGTACTCA-3′.

RAW264.7 cells were seeded in 24-well plates and transfected with 250 ng of an expression plasmid for ATP6V0D2 together with 10 ng pRL-TK (Promega) as an internal control, and 100 ng pGL3-IFN-β or pGL3-NF-κB reporter plasmid using Lipofectamine 2000. After 24 h of transfection, the cells were stimulated with poly(I:C), LPS, R837, or ODN1668 for 6 h and then were lysed with passive lysis buffer (Promega). Luciferase activities were measured by a TriStar² LB 942 Multidetection Microplate Reader (Berthold).

RAW264.7 cells seeded in 96-well plates were stimulated with poly(I:C), LPS, R837, ODN1668, and ISD for 2, 6, or 24 h. IL-6, IL-12p40, TNF-α, and IFN-β concentrations in the culture supernatants were measured by mouse IL-6, IL-12p40, and TNF-α Duoset (R&D Systems) and LumiKine mIFN-β (InvivoGen) according to the manufacturer’s instructions.

Cells cultured in six-well plates were lysed in lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 8, 0.5% Deoxycholate, 1% NP40, and 0.1% SDS). Following centrifugation, the supernatants were mixed with SDS sample buffer and then applied to SDS-PAGE. Immunoblot was performed with anti-IRF3 (Cell Signaling), anti–phospho-IRF3 (Cell Signaling), anti–NF-κB p65 (Cell Signaling), anti–phospho-NF-κB p65 (Cell Signaling), anti-FLAG (Sigma), anti-Myc (Sigma), anti-ARF6 (Santa Cruz Biotechnology), and anti–β-actin (Santa Cruz Biotechnology) Abs.

For the construction of ATP6V0D2-expressing retroviral vectors, cDNA for ATP6V0D2 was amplified by PCR from murine brain cDNA and inserted into the pMXs-IRES-puro (Cell Biolabs) that contained FLAG-tag and CMV promoter sequence. The retroviral vectors were transfected into Plat-E cells by Lipofectamine 2000. The virus was filtered with a 0.22-μm filter at 36 h posttransfection and infected into RAW264.7 cells. After infection for 16 h, cells were treated with 4 μg/ml puromycin as a selection, and ATP6V0D2 expression was measured by RT-qPCR and immunoblotting.

RAW264.7 cells, BMMs, and BMDCs were seeded in six-well plates and stimulated with LPS. After RAW264.7 cells, BMMs were detached by Corning Cell Lifter (Corning) and 10 mM EDTA in PBS, and the cells were washed with FACS buffer (PBS containing 1% BSA [Sigma] and 2 mM EDTA). The cells were then preincubated with an anti-mouse CD16/CD32 Ab (Fc blocking; BD Biosciences) for 30 min and then stained with the following Abs: PE-labeled anti-mouse TLR4 (BD Biosciences), FITC-labeled anti-mouse CD11b (BioLegend), Percp/Cy5.5-labeled anti-mouse F4/80 (BioLegend), or APC-labeled anti-mouse CD11c (BioLegend). Then, the cells were washed with FACS buffer and analyzed by FACS Accuri C6 (BD Biosciences) and FlowJo software (Tree Star).

RAW264.7 cells and BMMs were cultured in six-well plates and stimulated with LPS (10 ng/ml) for 18 h. Then the cells were washed and restimulated with LPS (100 ng/ml) for 6 h.

Knockdown was performed with the following oligo small interfering RNAs (siRNAs): si-Control (#12935-300); si-mAtp6v0d2, sense 5′-AUUAUCUGCCACUCUCUUCAUCUGC-3′; and si-hATP6V0D2, sense 5′-CAGAUCUCUUUAAUGCCAUUCUGAU-3′ (Life Technologies). siRNA was electroporated into HEK293 cells expressing TLR4 by Neon (Life Technologies). Two days after electroporation, the cells were used for experiments.

Cells cultured in 24-well plates were stimulated with rhodamine-labeled poly(I:C) (0.5 μg/ml) or Alexa Fluor 488–labeled LPS (5 ng/ml) for 1 h. The cells were washed with FACS buffer, and the fluorescences were measured by the FACS Accuri C6.

HEK293T cells (1 × 10⁷) were transfected with the indicated plasmids (3 μg) using Lipofectamine 2000. After 24 h, the cells were lysed in 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 5 mM EDTA, and 0.2% Triton X-100, and the cell lysates were immunoprecipitated with 10 μl of anti-Myc Ab-conjugated agarose beads (Santa Cruz Biotechnology). Immunoprecipitates or whole cell lysates were immunoblotted with the indicated Abs.

All experiments were independently repeated at least three times. Statistical significance was determined by the Student unpaired t test and ANOVA with Tukey test. A p value <0.05 was considered significant.

The V-ATPase complex consists of 14 subunits and has a central role in acidification of endosomes and lysosomes (29). To understand the contribution of acidic organelles to innate immune signaling, we established a RAW264.7 murine macrophage cell line lacking ATP6V0D2, a subunit of this complex, by the CRISPR/Cas9 system (32). We obtained ATP6V0D2 knockout (KO) cells by deletion of 8 and 10 bp (Δ8/Δ10) in the Atp6v0d2 exon 2 sequence, creating frameshift mutations (Fig. 1A). We termed this cell line V0D2 KO. Expression of Atp6v0d2 mRNA was markedly suppressed in V0D2 KO cells compared with WT cells as verified by RT-PCR (Fig. 1B) and RT-qPCR (Fig. 1C).

To analyze whether pH in the intracellular acidic organelles is neutralized in V0D2 KO cells, we treated cells with acridine orange, a fluorescent pH indicator that accumulates in acidic organelles via V-ATPase and emits 620 nm fluorescence depending on the acidic pH environment (33). In addition, it binds to nucleic acids and emits 520 nm fluorescence under neutral pH. Our confocal microscopy analysis demonstrated a decrease in the endosomal 620 nm fluorescence (red) induced by acridine orange in V0D2 KO cells compared with WT cells (Fig. 1D). In contrast, nucleic 520 nm fluorescence (green) was comparable between WT and V0D2 KO cells (Fig. 1D). Moreover, treatment with Baf A1, which is an inhibitor of V-ATPase, inhibited endosomal 620 nm fluorescence induced by acridine orange in WT cells (Fig. 1D). Furthermore, we analyzed the fluorescence ratios (620/520 nm ratio) as a histogram by flow cytometry. In V0D2 KO cells and Baf A1–treated cells, the 620/520 nm ratio was lower than in WT cells (Fig. 1E). Additionally, V0D2 KO cells showed decreased 620 nm fluorescence compared with WT cells, whereas 520 nm fluorescence was comparable between WT and V0D2 KO cells (Supplemental Fig. 1). Taken together, these findings indicated that V0D2 KO cells have defects in V-ATPase–mediated endosomal acidification.

We stimulated WT and V0D2 KO cells with poly(I:C), R837, or ODN1668, agonists of TLR3, TLR7, and TLR9, respectively, and measured Ifnb1 mRNA by RT-qPCR. Induction of Ifnb1 was significantly reduced in V0D2 KO cells compared with WT cells (Fig. 2A). Moreover, production of IFN-β and TNF-α was also reduced in V0D2 KO cells as measured by ELISA (Fig. 2B). Next, we transiently transfected ATP6V0D2 expression plasmids into V0D2 KO cells and measured Ifnb1 promoter activity by a luciferase reporter assay after stimulation with poly(I:C). Ifnb1 promoter activity was significantly higher in ATP6V0D2-expressing V0D2 KO cells than in V0D2 KO cells (Fig. 2C), suggesting that ATP6V0D2 expression is sufficient for restoration of the responses to TLR3 in V0D2 KO cells. Moreover, NF-κB promoter activity induced by TLR3, TLR7, and TLR9 ligands was reduced in V0D2 KO cells compared with WT cells, and IRF3 phosphorylation after poly(I:C) stimulation was impaired in V0D2 KO cells (Fig. 2D, 2E). We established RAW264.7 cells lacking another component of V-ATPase, ATP6V1B2 KO (V1B2 KO, Supplemental Fig. 2A, 2B), and found that these cells also exhibited impaired Ifnb1 induction in response to TLR3, TLR7, and TLR9 (Supplemental Fig. 2C). Taken together, these results suggest that endosomal acidification plays an important role in type I IFN and inflammatory cytokine induction in response to agonists of endosomal TLRs.

RIG-I-like receptors and multiple DNA sensors (AIM2, DAI, DDX41, LRRFIP1, MRE11, and cGAS) have been identified as cytosolic RNA and DNA sensors, respectively (34–36). We therefore evaluated the contribution of endosomal acidification to cytosolic nucleic acid sensor-mediated innate immune responses. Expression of Ifnb1 and Il6 mRNA and production of IL-6 in response to poly(I:C) or ISD transfection were comparable between WT and V0D2 KO cells (Supplemental Fig. 3A, 3B). Moreover, poly(I:C) or ISD transfection-mediated IRF3 phosphorylation was also similarly induced in both WT and V0D2 KO cells (Supplemental Fig. 3C, 3D). These observations suggest that endosomal acidification is dispensable for production of antiviral and inflammatory cytokines in response to cytosolic RNA and DNA.

To examine whether ATP6V0D2 deficiency influences TLR4-mediated innate immune responses, we stimulated WT and V0D2 KO cells with LPS and then measured expression of Il6 and Ifnb1 mRNAs by RT-qPCR. Il6 mRNA expression was markedly increased in V0D2 KO cells compared with WT cells, whereas Ifnb1 mRNA expression was decreased in V0D2 KO cells (Fig. 3A). Moreover, time course analysis indicated that IL-6, IL-12p40, and TNF-α production after LPS stimulation was higher in V0D2 KO cells than in WT cells post-LPS stimulation, whereas IL-6 production after stimulation with macrophage-activating lipopeptide 2 (MALP2), a TLR2 ligand, was comparable in both cell lines (Fig. 3B). V0D2 KO cells showed reduced IFN-β production by LPS stimulation, but showed normal production by MALP2 stimulation (Fig. 3B). ATP6V1B2 deficiency also resulted in increased IL-6 production and reduced Ifnb1 expression after LPS stimulation (Supplemental Fig. 2D, 2E). Next, to evaluate the role of ATP6V0D2 in primary innate immune cells, we knocked down ATP6V0D2 expression in BMDCs (V0D2 KD). Knockdown of Atp6v0d2 mRNA was confirmed by RT-qPCR (Fig. 3C). Induction of Il6 and Tnfa mRNA in response to LPS was higher in V0D2 KD cells than WT cells whereas Ifnb1 mRNA was decreased in V0D2 KD cells (Fig. 3C).

We then examined NF-κB and IRF3 activation. Phosphorylation of the NF-κB p65 subunit induced by LPS stimulation was enhanced in V0D2 KO cells in comparison with WT cells (Fig. 3D), whereas IRF3 phosphorylation was decreased in V0D2 KO cells (Fig. 3D). These findings suggest that suppression of TLR4-mediated NF-κB activation requires neutralization of intracellular acidic organelles.

To investigate whether the phenotype of V0D2 KO cells in terms of cytokine expression was due to a loss of V0D2 or not, we restored ATP6V0D2 expression into V0D2 KO cells by transfecting FLAG-tagged ATP6V0D2 expression plasmid and measured Il6 and Ifnb1 mRNA level. The expression of ATP6V0D2 in V0D2 KO cells was evaluated by immunoblotting and RT-qPCR (Fig. 3E, 3F). Il6 induction in response to LPS was lower in ATP6V0D2-expressing V0D2 KO cells than in V0D2 KO cells (Fig. 3G). On the other hand, Ifnb1 mRNA induction was increased in ATP6V0D2-expressing V0D2 KO cells compared with V0D2 KO cells (Fig. 3G). Therefore, ATP6V0D2 expression was able to partly rescue TLR4-mediated cytokine induction in V0D2 KO cells.

It has been shown that TLR4 internalizes into endosomes and lysosomes from the cell surface upon LPS engagement (15). Therefore, we analyzed cell surface expression of TLR4 by flow cytometry. In the unstimulated condition, TLR4 surface expression was comparable between WT and V0D2 KO cells (Fig. 4A, 4B). After 1 h of stimulation, surface expression of TLR4 was downregulated in WT cells (Fig. 4A, 4B). However, in V0D2 KO cells, some TLR4 remained on the cell surface (Fig. 4A, 4B). Furthermore, surface expression of TLR4 was found at 3 and 6 h after LPS stimulation in V0D2 KO cells (Fig. 4A, 4B). Similar to TLR4 surface expression in V0D2 KO cells, BMMs and BMDCs also showed TLR4 downregulation following LPS stimulation, which was suppressed by Baf A1 treatment and knockdown of ATP6V0D2, respectively (Fig. 4C, 4D). Next, we examined whether LPS is also uptaken by cells or not. Alexa Fluor 488–labeled LPS was taken up by WT RAW264.7 cells, whereas it was reduced in V0D2 KO cells (Fig. 4E). By contrast, uptake of rhodamine-labeled poly(I:C) was comparable between WT and V0D2 KO cells (Fig. 4F). These findings suggest that blockade of intracellular acidification dampens internalization of TLR4 and LPS.

Previous studies have shown that pretreatment of macrophages with LPS prevents inflammatory cytokine production upon secondary LPS treatment (37). This mechanism is regarded as LPS tolerance or endotoxin tolerance. There are multiple mechanisms for LPS tolerance, which include upregulation of numerous anti-inflammatory molecules, epigenetic regulations, miRNA-mediated suppressions of inflammatory genes, and TLR4 downregulation (27, 28). In this regard, we pretreated WT and V0D2 KO cells with or without a low dose of LPS (10 ng/ml). After removing the culture supernatant, we treated the cells with a high dose of LPS (100 ng/ml) and measured cytokine expression. V0D2 KO cells in the absence of LPS pretreatment produced higher levels of Il6, Il12p40, and Tnfa mRNAs and proteins than WT cells after LPS stimulation (Figs. 3A, 3B, 5A, 5B). In WT cells, mRNA expression of Il12p40 and Tnfa after secondary LPS stimulation was reduced to the basal level (Fig. 5A). In V0D2 KO cells, expression of these genes after secondary LPS stimulation was also reduced, but expression of these genes was significantly higher than the basal level. Moreover, whereas Il6 mRNA level after the secondary LPS stimulation was reduced when compared with LPS stimulation alone in both WT and V0D2 KO cells, reduction of Il6 mRNA after secondary LPS stimulation in V0D2 KO cells was lower than WT cells (Fig. 5A). These findings suggest that V0D2 KO cells are capable of upregulating these genes after secondary LPS stimulation and therefore have partial defects in inducing LPS tolerance. Then, we measured cytokine production by ELISA and found that IL-12p40 production was significantly increased after secondary LPS stimulation in V0D2 KO cells whereas it was not induced in WT cells (Fig. 5B). IL-6 production was also higher in V0D2 KO cells than WT cells following secondary LPS stimulation (Fig. 5B). These results also suggest that ATP6V0D2 deficiency partially abrogated LPS tolerance. Furthermore, similar to V0D2 KO cells, LPS-pretreated V1B2 KO cells showed increased Il6 mRNA induction in response to secondary LPS stimulation (Supplemental Fig. 2F). We next addressed NF-κB activation. LPS stimulation caused p65 phosphorylation in WT cells that had not been pretreated with LPS, and this was decreased in LPS-pretreated WT cells (Fig. 5C). NF-κB activation in response to secondary LPS stimulation in LPS-pretreated V0D2 KO cells was lower than that in response to LPS stimulation in V0D2 KO cells that had not been pretreated with LPS (Fig. 5C). However, the level of NF-κB activation in both conditions was higher than WT cells (Fig. 5C). Taken together, these results suggest that LPS tolerance was partially impaired in V0D2 KO cells, presumably because of abrogated downregulation of cell surface TLR4.

To search for molecules that regulate TLR4 trafficking, we focused on ARF family members which are the small GTPase proteins and reportedly involved in the regulation of innate immune responses via intracellular vesicular trafficking (38–40). We found that Alexa Fluor 488–labeled LPS uptake was induced in HEK293 cells expressing TLR4 (Fig. 6A). In contrast, HEK293 cells (which do not express TLR4) showed LPS uptake, albeit at a low level, suggesting that LPS uptake largely depends on TLR4 (Fig. 6A). Therefore, we used HEK293 cells expressing TLR4 and assessed the contribution of ARF family members to LPS uptake. In this regard, we overexpressed a series of ARF proteins (ARF1–6) into HEK293 cells expressing TLR4 and uptake of Alexa Fluor 488–labeled LPS was measured by FACS analysis (Fig. 6B). Among ARF members tested, ARF6 overexpression markedly increased LPS uptake (Fig. 6B). Moreover, the ARF6 overexpression–mediated increase in LPS uptake was suppressed by Baf A1 treatment (Fig. 6C). In addition, overexpression of ARF6 T27N mutant with a defective GTP binding ability (41) failed to increase LPS uptake (Fig. 6C). These results suggest that ARF6 regulates LPS uptake presumably via TLR4 in HEK293 cells expressing TLR4. Next, to address whether ARF6-mediated LPS uptake links to ATP6V0D2 function or not, we knocked down ATP6V0D2 in HEK293T cells. Knockdown of ATP6V0D2 mRNA was confirmed by RT-qPCR (Fig. 6D). We found that ARF6 overexpression-dependent LPS uptake was suppressed by ATP6V0D2 knockdown (Fig. 6E). Furthermore, we then examined interactions between ARF6 and V-ATPase complex. We overexpressed HEK293T cells with Myc-tagged ARF6 together with FLAG-tagged ATP6V0D2 or ATP6V0C. Immunoblot analysis using anti-FLAG Ab revealed that FLAG-ATP6V0C but not FLAG-ATP6V0D2 coprecipitated with Myc-Arf6, indicating that Arf6 interacts with ATP6V0C but not with ATP6V0D2 (Fig. 6F). Collectively, these findings suggest that ARF6 is required for LPS-TLR4 internalization presumably depending on V-ATPase.

To understand the physiological function of ARF6 in terms of TLR4 regulation, we generated ARF6-deficient RAW264.7 cells by CRISPR/Cas9 system (Supplemental Fig. 4) (32), and verified the surface expression level of TLR4 by FACS analysis. In ARF6 KO cells, surface TLR4 level was increased in comparison with WT cells in unstimulated condition. After LPS treatment, WT cells showed downregulation of TLR4, whereas ARF6 KO cells exhibited remaining surface expression of TLR4 even after 6 h treatment (Fig. 7A, 7B). Next, we measured cytokine expression and activation of IRF3 and NF-κB in ARF6-deficient cells. The expression of Il6 and Ifnb1 mRNA in response to LPS was markedly reduced in ARF6 KO cells in comparison with WT cells (Fig. 7C). ARF6 KO cells showed reduced expression of Il6 expression in response to MALP2 (Fig. 7C). Furthermore, LPS-induced activation of IRF3 and NF-κB p65 was also reduced in ARF6 KO cells compared with WT cells (Fig. 7D, 7E). These findings suggest that ARF6 is required for TLR4 internalization as well as for activation of IRF3 and NF-κB, which was in contrast to ATP6V0D2 KO cells that showed increased NF-κB but decreased IRF3 activation.

It has been demonstrated that cellular localization of TLRs is tightly associated with their functions. In particular, viruses or bacteria engulfed by innate immune cells are transported to endosomes or lysosomes before their nucleic acids are released and gain access to endosomal TLRs (42). In the acidic condition, endosomal TLRs are processed by acid proteases to their mature form (12–14). Our data indicated that ATP6V0D2 deficiency resulted in neutralization of pH in endosomes or lysosomes as well as abrogation of cytokine induction after ligand stimulation of TLR3, 7, and 9. It was assumed that ATP6V0D2-deficient cells failed to activate the proteases and generate mature forms of endosomal TLRs. Although it is possible that these cells were unable to uptake nucleic acid ligands, we found that incorporation of rhodamine-labeled poly(I:C) was not impaired in ATP6V0D2-deficient cells as determined by flow cytometric analysis, indicating that nucleic acid incorporation occurs in a ATP6V0D2-independent manner.

Interestingly, ATP6V0D2 deficiency exacerbated LPS-induced Il6, Tnfa, and Il12p40 expression, but reduced Ifnb1 expression. Our results demonstrated prolonged expression of TLR4 on the surface of ATP6V0D2-deficient cells, which may have been responsible for the increased NF-κB activation. Although the reason why TLR4 internalization requires acidification of intracellular compartments remains elusive, recent studies have suggested that V-ATPase directly regulates vesicular trafficking. ARF6 and its GDP-GTP exchange factor ARNO bind to c and a2 s of V-ATPase, respectively, and trigger clathrin-dynamin-AP2–dependent endocytosis in an acidic pH-dependent manner (43). Moreover, Rab-interacting lysosomal protein, RILP, localizes at endosomes and binds G1 subunit of V-ATPase and controls endocytosis through interaction with the Rab family (44). Therefore, it is likely that TLR4 internalization is controlled by V-ATPase–associated molecules. We found that ARF6 did not directly interact with ATP6V0D2 but interacted with ATP6V0C, suggesting that ARF6 binds to V-ATPase complex via ATP6V0C. Hence, it is considered that ATP6V0D2 regulates ARF6-dependent LPS uptake through interacting with ATP6V0C. Furthermore, ARF6 overexpression caused LPS uptake, which was suppressed by Baf A1 and ATP6V0D2 knockdown, and ARF6-deficient cells displayed prolonged expression of TLR4 on the cell surface after LPS treatment. These findings suggest that ARF6 mediates internalization of LPS–TLR4 complex and delivers it to ATP6V0D2-containing intracellular vesicles where it activates TRAM-TRIF–dependent pathway. A previous report indicated that ARF6 is required for activation of both plasma membrane TLR4-mediated TIRAP-MyD88–dependent pathway and endosomal TLR4-mediated TRAM-TRIF–dependent pathway (20, 39). Consistently, our results with ARF6-deficient cells showed that LPS-induced activation of NF-κB and IRF3 was suppressed. Previously, it was reported that ARF6-mediated activation of PIPK5 leads to synthesis of PI(4,5)P₂, which functions as a hub to recruit TIRAP to the lipid rafts where it activates the MyD88-dependent pathway (19, 20). As we found that ARF6 overexpression caused TLR4 internalization, cells overexpressing ARF6 insufficiently recognized LPS and therefore failed to activate NF-κB. Moreover, ARF6-deficient cells constitutively expressed TLR4 on the cell surface even after LPS stimulation, but they were unable to activate NF-κB and IRF3. In this case, TLR4 may fail to recruit TIRAP to lipid rafts because of abrogated ARF6 function, and therefore ARF6-deficient cells have impaired activation of both TIRAP-MyD88– and TRAM-TRIF–dependent pathways (Fig. 7F). It is shown that TLR4 signaling activation is not required for TLR4 internalization; rather, CD14 mediates TLR4-LPS uptake. Therefore, ARF6 may have a role to directly control CD14 function, which is responsible for TLR4 internalization. Furthermore, we also found that IL-6 and TNF-α induction after stimulation with MALP2 was comparable between WT and V0D2 KO cells, although IL-6 induction was reduced in ARF6 KO cells. Previously, it was reported that TLR2 has an ability to induce type I IFN via TRIF in inflammatory monocytes, whereas TLR2 fails to induce type I IFN in macrophages and dendritic cells (45). Indeed, IFN-β was not upregulated in response to MALP2 in RAW264.7 cells (Fig. 3B). Thus, it will be required to clarify whether ATP6V0D2 is involved in TLR2-mediated type I IFN–inducing pathway in inflammatory monocytes in the future.

Previous reports suggest TLR4-mediated inflammatory responses through the MyD88-dependent pathway during LPS tolerance is suppressed, whereas the TRIF-dependent pathway remains intact or rather activated (27, 28, 37). We found that ATP6V0D2-deficient cells produced higher amounts of IL-6 and TNF-α than WT cells in response to the secondary LPS stimulation and showed reduced internalization of surface TLR4 after the first LPS stimulation. However, LPS tolerance was modestly induced in ATP6V0D2-deficient cells, suggesting the existence of additional mechanisms controlling LPS tolerance other than TLR4 internalization. It has been shown that tolerant cells highly express the anti-inflammatory cytokines IL-10 and TGF-β, as well as the NF-κB p50/p50 homodimer, a transcriptionally inactive form that competes with the transcriptionally active NF-κB p65/p50 heterodimer (27, 46, 47). Moreover, numerous molecules such as IL-1RA, IRAK-M, A20, ST2, MKP1, and FLN29 are shown to suppress TLR4-mediated signaling pathways during LPS tolerance, and therefore mice deficient for these molecules cause enhanced susceptibility to endotoxin shock (27, 28). Furthermore, epigenetic regulations including histone modification and nucleosome remodeling, and miRNA-mediated suppression of TLR4, TLR signaling molecules, and inflammatory cytokines have been suggested to participate in LPS tolerance (28, 48–50). Transcription factors ATF7 and RelB are reported to suppress IL-1β and TNF-α promoter activities by H3K9 dimethylation through recruitment of H3K9 methyltransferase G9a in naive LPS response and LPS tolerance, respectively (51, 52). Collectively, although it appears that V-ATPase–mediated acidification is involved in eliciting endotoxin tolerance, other pathways are also presumably involved in ATP6V0D2-deficient cells.

Accumulating evidence has suggested that endosomal acidification is related to the entry, replication, and infection phases of several viruses such as Ebola virus, dengue virus, influenza A virus, hepatitis C virus, and HIV (53). Intracellular pathogens including Mycobacterium tuberculosis and Legionella pneumophila inhibit vesicular acidification through interaction with V-ATPase to increase their survival (54, 55). Moreover, dysfunction of endosomal TLRs is closely associated with autoimmune disorders (4, 5, 47), and TLR4-mediated inflammatory responses are responsible for septic shock that is life threatening (26). Thus, this complex might be a target to control these therapies.

We thank K. Abe for secretarial assistance and W. Monwan, M. Nagayama, and T. Deguchi for technical support.

The authors have no financial conflicts of interest.