FcγRIIB prevents Inflammatory Type I Interferon Production from pDCs during a Viral Memory Response

Abstract

The type I interferon (IFNα) response is crucial for viral clearance during primary viral infections. Plasmacytoid dendritic cells are important early responders during systemic viral infections and, in some cases, the sole producers of IFNα. However, their role in IFNα production during memory responses is unclear. We found that IFNα production is absent during a murine viral memory response despite colocalization of virus and pDCs to the splenic marginal zone. The absence of interferon was dependent on circulating antibody, and reversed by the transgenic expression of the activating human FcγRIIA receptor on pDCs. Furthermore, FcγRIIB was required for Sendai Virus immune complex (SeV IC) uptake by splenic pDCs in vitro and internalization via FcγRIIb prevented cargo from accessing TLR signaling endosomes. Thus, pDCs bind viral immune complexes via FcγRIIB, and prevent IFNα production in vivo during viral memory responses. This antibody-dependent, IFNα regulation maybe an important mechanism by which the potentially deleterious effects of IFNα are prevented during a secondary infection.

INTRODUCTION

The type I interferon (IFNα) response is crucial for viral clearance during a primary viral infection and modulates both innate and adaptive components (1, 2). While all cells can produce IFNα, pDCs are thought to be important early responders during primary viral infections. Indeed, early studies using synthetic TLR ligands, such as CpG, highlighted pDCs as the sole producer of Type I IFN after intravenous injections(3–5).

Results from recently developed pDC deficient mice have demonstrated that pDCs participate in the local and systemic production of IFNα after viral infections (6–8). However, these studies also suggest that the role of pDCs as exclusive producers of interferon, may be virus specific(9–12) and route dependent(6, 13). When VSV or HSV were delivered intravenously, pDC derived IFNα was evident by 6 hours(5, 6) in the circulation. The lack of IFNα production in pDC deficient mice, resulted in important phenotypic consequences, with increased viral titers and impaired adaptive immune responses. However, when MCMV or MHV were injected in the intraperitoneal cavity, a pDC specific IFNα response was delayed and less pronounced(6, 7). Taken together, these studies support a role for pDCs as important producers of Type 1 IFN and also suggest that the route of viral inoculation may dictate when and to what degree pDCs engage the immune response.

During secondary viral responses, IFNα has been traditionally thought to be less critical for viral clearance as other mechanisms provided by the adaptive memory immune response are operative and available for rapid deployment; i.e. CTL and/or antibody mediated clearance. Indeed, early studies found that clearance of secondary pox virus infections occurs comparably between WT and IFNAR KO mice^(14). More recently however in mouse models of secondary Sendai Virus (SeV) respiratory infections, IFNα was shown to be crucial in activating memory CD8+ T cells to produce granzyme B and contain viral spread(15). Yet, it is clear too that IFNα may be injurious to the host ^(16–18). pDC derived interferon has been shown to have direct negative effects on pDCs themselves resulting in apoptosis (19) and a state of over-activation⁽²⁰⁾. It would be advantageous then for pDCs to distinguish between a primary and a secondary response and appropriately regulate their interferon production to prevent unnecessary and untoward inflammatory responses.

We previously described the dominant expression of the inhibitory FcγRIIB on murine pDCs and posited that it may have important implications for interferon regulation during viral infections when circulating, virus-specific antibodies are present. In the current study, we find that pDCs are the exclusive producers of IFNα during a primary systemic SeV inoculation. However, interferon production is absent during a memory response despite the presence of virus and pDCs together in the spleen. The absence of interferon was dependent on circulating virus-specific antibody, and reversed by the transgenic expression of the activating FcγRIIA receptor on pDCs. This antibody dependent IFNα regulation may be an important mechanism by which the potential deleterious effects of IFNα are prevented during a secondary infection.

MATERIALS AND METHODS

Mice

C57BL/6, were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). Fcgr2b (FcγRIIB−/−), Fcer1g (FcγRγ−/−) and double knockout Fcgr2b (FcγRIIB−/−) Fcer1g (FcγRγ−/−), referred to in the manuscript as FcγR null mice, were purchased from Taconic Farms (Germantown, NY, USA). hFcγRIIA transgenic mice were generated as described(21) and obtained from Steve McKenzie. YFP-IFNβ were obtained from Boris Reizis with permission from Richard Locksley (UCSF). YFP-IFNβ hFcγRIIA mice were generated in our lab. All animal experiments were performed in compliance with institutional guidelines and approved by Columbia University’s Institutional Animal Care and Use Committee (New York, NY).

FLT3L Bone Marrow Cultures

Bone marrow cells (2 × 10⁶ cells/ml) were cultured for 7–9 days in RPMI/10%FCS supplemented with 100ng/ml murine Flt3 ligand (R&D Systems, Minneapolis MN) and typically consisted of 80% cDCs and 20% pDCs.

Virus

Sendai Virus-Cantell strain was purchased from Charles River Labs and obtained from Vincent Racaniello (Columbia University).

Murine Immune Sera and Immune Complex Formation

SeV Antisera- Mice were immunized with either live or heat killed Sendai Virus Cantell by tail vein injection. Four to six weeks post injection, mice were boosted and sera was collected from mice 1 week post-boost. The blood of 5 C57BL/6 mice were pooled and sera frozen in aliquots.

For intracellular IC trafficking studies, IgG was purified with protein A/G columns (ThermoScientific, Rockford, IL). An ELISA was developed in-house to determine the presence of mouse anti SeV IgG using SeV coated plates.

For in vitro functional assays, SeV ICs were made using 5μL of live virus stock incubated with 5μL of antisera and 0.5mls of RPMI 1640 for at least 15 minutes in a 37°C water bath. SeV ICs or SeV alone were incubated overnight with purified BMpDCs and IFNa was read by ELISA.

For in vivo experiments, 20μL of anti-SeV immune sera was injected intravenously (i.v.) either before, after or concurrent with i.v. 50μL virus, both into the tail vein.

Ovalbumin was purchased from Worthington Biochemical Industries, Lakewood, NJ. CpGB-Ovalbumin fusion molecules were kindly provided by Aliasger Salem and George Weiner (University of Iowa) and resulted in IFNα production as expected by previous published data(22) and our own data (data not shown). Rabbit anti-ovalbumin was prepared by Protein G chromatography (Repligen Co., Waltham, MA) from the sera of OVA- immunized rabbits (contracted at Covance, Princeton, NJ). Ovalbumin immune complexes were made in 1:10 or 1:5 ratios of antigen (μg/ml) to antibody (μg/ml).

Cytokine Assays

IFNα was measured by ELISA (PBL Piscataway, NJ). BD CBA Mouse Inflammatory Cytokine kit was used to measure other cytokines (BD Bioscience San Jose, CA).

Antibodies

Anti-CD11c (HL3), anti-CD19 (1D3), anti-B220 (RA3-6B2), anti-CD11b (M1/70) mAbs were obtained from BD Pharmingen, San Diego CA. Anti-PDCA1 (120G8) was a gift of Georgio Trinchieri (formerly at Schering-Plough). Chicken anti-Sendai, anti-chicken DyLight 488 or 594, sheep anti LC3 were purchased from AbCam, Cambridge MA. Anti-Chicken HRP was purchased from Thermo Scientific, Waltham, MA. Donkey anti-Rabbit alexa-fluor 568, donkey anti sheep alexa- fluor 647 were purchased from Life Technologies, Eugene, OR. Anti-MOMA1 and anti-MARCO were purchased from AbD Serotech, Kidlington, UK.

Immunofluorescence and Microscopy

Detecting SeV in the spleen- Spleens were harvested at various times post injection and either analyzed as single cell suspensions or placed into OCT compound and frozen at −80°C prior to sectioning (6–8μm continuous sections) and staining as indicated. Thawed tissue sections were stained with indicated antibodies prior to fixing with 4% Paraformaldehyde. Sendai virus was stained using Chicken anti Sendai Virus antibody and followed with an anti-chicken secondary as described above.

Intracellular IC trafficking-Splenic pDCs were isolated from spleens by depleting CD19 positive cells and then selecting for BST2 positive cells using MicroBeads purchased from Miltenyi Biotec, San Diego CA. Images were acquired by confocal microscopy using a Nikon A1 Confocal microscope with a 100X objective and analyzed by Image J. Integrated density above a certain threshold of Lamp1, LC3 and Lysotracker was measured at specific ROIs. Measurements at random ROIs were taken on images pivoted 90 degrees as described(23). Manders coefficients were determined using the Image J plugin JACoP (Just Another Colocalization Plugin) developed by Fabrice P. Cordelieres (Institut Curie, Paris, France) and Susanne Bolte (Sorbonne Universites, Paris, France).

Statistical Analysis

Differences between two groups were evaluated using Students T test. ANOVA was used to compare three or more groups with Tukeys Multiple Comparison post-test. Statistical significance is notated by asterisks (*p<0.05; **p<0.01; *** p<0.001).

RESULTS

IFNα production is absent during secondary exposure to Sendai Virus

Type I Interferon’s potent innate antiviral properties also have potentially deleterious local and systemic toxic effects that could cause unnecessary cellular activation and immunopathology if left unbridled during resolving or secondary infections. This would suggest that a host would benefit by restricting interferon production to a first line of defense; i.e. before an effective adaptive immune response develops.

To investigate whether IFNα is produced during a memory response to virus delivered intravenously, a cohort of mice were immunized with SeV, and then rechallenged 14–21 days later (memory response). A similar cohort of mice received only a single viral immunization (primary response). Both groups of mice were then analyzed for circulating IFNα levels in the sera over the next two days. During a primary challenge, IFNα was detected in the sera as early as 5hr post injection and began to wane by 24hr. In contrast, the SeV memory response, failed to produce any detectable interferon throughout the 48-hour period (Fig. 1A).

Figure 1. IFNα production is absent during secondary exposure to Sendai Virus.

A) WT mice, previously immunized with SeV, were rechallenged with SeV (memory) and compared to mice receiving a single injection (naïve). IFNα in the sera was measured at the indicated timepoints by ELISA. Repeated at least 4 times with a total of N=12 mice in each group. B) YFP-IFNβ mice were injected as in A and three hours post SeV challenge spleens were analyzed by FACS for YFP positive cells in pDCs (CD11c^mid BST2+), cDCs (CD11c+CD11b+) and B cells (CD19+ CD11b−). Each symbol is a mouse obtained from pooled experiments. The experiments were repeated 5 times with between 7 and 9 mice total per group. C) Backgating analysis of B on YFP positive cells. Representative sample from B. D) CBA analysis of sera 5 hours post-SeV challenge as in A. The experiment was repeated twice with 4–6 mice in each group. E) Western blot analysis of total splenocytes of primary and memory mice as in A. Representative of one mouse as in A.

Given the importance of pDCs in the production of type I interferon during certain viral infections, we sought to determine whether pDCs were involved in IFNα production during a naïve response to SeV. We used a YFP-IFNβ reporter mouse to indirectly measure IFNβ production. YFP-IFNβ reporter mice were injected via the tail vein with live SeV and spleens were harvested 3 hours post injection. Beyond 3 hours, our ability to accurately gate on pDCs was compromised due to interferon-dependent induction of the pDC lineage marker BST2 on other cells(24). In this analysis of splenocyte populations, we found that in a naïve setting, interferon production was limited to pDCs; neither conventional DCs (CD11c⁺ BST2⁻), nor macrophages (CD11b⁺ CD11c⁻), nor B cells (CD19⁺), nor cells negative for all markers (the vast majority of which are T cells), produced detectable YFP in response to SeV (Fig. 1B). In contrast, by backgating on YFP⁺ cells, CD11c^mid, BST2⁺ pDCs were identified as the sole producers of IFNβ during a naïve viral challenge at 3 and 4 hours (Fig. 1C). Interestingly, only about 10–20% of total pDCs in the spleen responded to the viral inoculation to produce YFP. This limited percentage was not dose-dependent, as the percentage of YFP positive pDCs could not be increased by increasing the dose of virus injected (data not shown).

The induction of other inflammatory cytokines by Sendai virus were also differentially regulated between a primary and memory response. In particular, when measured in the sera 3 hours after exposure IL-6, like IFNα, was also limited to the primary response while MCP1 and IFNγ, were significantly induced in both the primary and memory response (Fig. 1D).

A lack of type 1 interferon during a viral memory challenge might be explained by an efficient and rapid clearance of virus before it is detected by pDCs. Yet Western Blot analysis of splenocytes shows that the presence of SeV is remarkably similar between a primary and memory response. While, about half of the initial inoculum is cleared in a memory versus a primary setting, the kinetics of the virus are similar in that they peak at 1 hour and persist through 48 hours compared to no treatment (Fig. 1E).

The failure of pDCs to respond to SeV during a memory response may also be explained by a lack of access to the virus. We, as others, have found pDCs in the red and white pulp in spleens of naïve mice. In the white pulp, pDCs exist largely at the marginal zone and within the T cell area. More detailed analysis of pDCs at the marginal zone, showed that pDCs were interweaved between Marco and Moma-1 positive cells, clearly delineating the zone by forming short chains of pDCs (Fig 2).

Figure 2. Splenic pDCs interact with Sendai Virus at the marginal zone.

A and B) Memory and primary mice were compared for SeV (green) colocalization at indicated timepoints with various cellular populations (red). One representative experiment from two experiments is shown. C) and D) Quantification of A and B for each cell population. Between 5 and 10 sections were quantified for each timepoint and cell population. One representative experiment of two is shown with similar patterns. D) Left Panel: Inset of Fig. 2(D) BST2 at 35 minutes is shown. Right panel: Quantification of 20 and 35 minutes is shown.

To address spatio-temporal questions and pDC accessibility to injected virus during a memory response, previously immunized mice were rechallenged with SeV and cross-sections of spleens were stained at various timepoints. We found that during a memory response, SeV was largely present at the marginal zone associated equally between Marco positive, Moma-1 positive macrophages and marginal zone B cells at 20 and 35 minutes, moving eventually to follicular B cells at 50 and 75 minutes (Fig 2A, C top panel). En route to follicular B cells, SeV colocalized with pDCs at the marginal zone between 20 and 50 minutes. Surprisingly, this analysis indicated only about 10% of SeV is present on pDCs during a memory response.

We wondered how this compared to a primary response where the interaction between pDCs and SeV is known to result in a strong IFNα response. SeV was injected as above into the tail veins of naïve mice and spleens were harvested at the indicated timepoints. Interestingly, we found that at early timepoints (20 minutes), SeV in a primary setting followed nearly the same route as in a memory setting, colocalizing with Marco+ and Moma-1+ cells. SeV trafficking differences between a memory and primary response however, became evident by 40 minutes and more apparent by 75 minutes. SeV had traveled further into the B cell area than in a memory setting and did not persist at the marginal zone (Fig 2B, C bottom panel). Interestingly, the early access to SeV by pDCs was nearly equivalent between a memory and primary setting (Fig 2D).

Taken together, these data suggests that unlike a primary response, the interaction between SeV and pDCs during a memory response at the splenic marginal zone fails to induce an interferon response.

Virus specific antibody is required for interferon suppression during a memory response

We found that 14–21 days post immunization, mouse sera contained SeV specific antibody as determined by Hemagluttination Inhibition (HI) and an SeV specific ELISA (data not shown). To investigate whether antibody was necessary for interferon suppression during a memory response, we repeated the primary and memory viral challenge using mice unable to produce virus-specific antibody. HELμMT mice express a transgenic B cell receptor specific for Hen Egg Lysozyme on an IgM deficient background. Interestingly, and in complete contrast to WT mice, HELμMT mice did not suppress the IFNα response during a memory rechallenge (Fig. 3A).

Figure 3. Virus specific antibody is required for IFNα suppression.

A) WT mice were compared to HELμMT mice for IFNα secretion in the sera by ELISA after intravenous re-exposure to Sendai virus in a memory response. B) WT naive mice were injected with SeV or pre-formed SeV ICs and IFNα concentrations were measured in the sera by ELISA. Experiment was repeated twice with a total of 6 mice. C) As in B, section of spleens were stained for SeV and BST2+ cells at 30 minutes post injections. Quantification of pooled experiments from 4 and 5 mice (SeV and SeV IC, respectively) shown from 2 experiments.

To more directly determine whether virus specific antibody was required for IFNα suppression, a cohort of naïve mice was passively immunized by transfer of immune sera, then challenged with SeV as before, IFNα was measured in the serum as described earlier. We found that immune sera was sufficient to prevent any type I interferon production in response to SeV (Fig 3B) and Flu (Supplementary Fig. 1). Lastly, colocalization of SeV and pDCs in the spleens of mice 30 minutes following injection of preformed SeV ICs suggests that the inhibition of interferon by pDCs is again not due to a lack of access by pDCs (Fig 3C). Combined, these results suggest that inhibition of interferon production by pDCs during a viral memory response requires the presence of virus specific antibody.

Transgenic expression of hFcγRIIA on pDCs permits IFNα production during a memory response

We previously described the dominant expression of FcγRIIB on murine pDCs and that the limited expression of FcγRs prevented antigen presentation of ovalbumin immune complexes to T cells(25). We designed a series of experiments to directly examine a role for FcγRIIB in the inhibition of IFNα through the use of the knockout mouse. While we show that in both a memory response and by directly injecting SeV ICs, there is an increase of IFNα in the sera in the absence of FcγRIIB, we could not determine whether pDCs were responsible for this effect because the FcγRIIB KO mouse is not pDCs specific (Supplementary Figure 2A and B). In vitro experiments co-culturing WT and FcγRIIB KO pDCs with SeV IC showed that both WT and FcγRIIB KO pDCs fail to produce IFNα, however the latter being due to the inability of FcγRIIB KO pDCs to acquire ICs (Supplementary Figure 2C and D).

Being unable to determine the direct requirement for FcγRIIB in vivo we determined whether expressing the activating FcγRIIA would reverse the interferon suppression seen in WT responses. Transgenic mice expressing the human FcγRIIA receptor, an activating Fc receptor present only in primates, were compared to WT mice in a memory SeV response. Analysis of the pDCs from these transgenic mice showed the transgene expressed on pDCs(25). In contrast to WT mice, FcγRIIA transgenic mice responded with limited though significant IFNα concentrations in the sera (Fig 4A). To determine whether pDCs were responsible for producing the IFNα measured sera, YFP-IFNβ mice were crossed to FcγRIIA transgenic mice and the memory experiment was repeated. Importantly, we found that endowed with the presence of an activating Fcγ receptor, pDCs are capable of inducing an interferon response in a memory setting and were again the only cell involved in the response at 3 hours (Fig. 4B). Furthermore, direct injection of SeV ICs into hFcγRIIA transgenic mice resulted in IFNα production, contrasting the suppressed response in WT mice (Fig. 4C).

Figure 4. Interferon production in response to ICs is rescued by transgenic expression of FcγRIIA.

A) WT and hFcγRIIA transgenic mice were compared during a memory response for IFNα concentrations in the sera by ELISA at the indicated timepoints. Pooled experiments of 6–12 mice per group are shown. B) Dual transgenic FcγRIIA+ YFP-IFNβ reporter mice were compared to WT YFP-IFNβ reporter mice 3 hours post rechallenge. Percent YFP+ pDCs (CD11c+ BST2+) were determined by FACS. 4–8 mice per group. C) WT or hFcγRIIA transgenic mice were injected with preformed SeV ICs. Sera was analyzed by ELISA for IFNα at indicated timepoints. Pooled experiments of at least 6 mice is shown. D) Bone marrow expanded WT or hFcγRIIA expressing pDCs were incubated overnight with SeV and the supernatant was analyzed by ELISA for IFNα. Repeated twice in duplicate and triplicate. E) As in D except enriched BM pDCs were incubated with SeV ICs. F) Splenic pDCs from WT or hFcγRIIA mice incubated with pre-formed SeV ICs for 30minutes and analyzed by confocal microscopy. G) Quantification of images as in F were analyzed for the average number of SeV particles per cell per image. Pooled data from 2 independent experiments with over 50 cells analyzed per group.

To directly examine the interaction between Sendai viral IC and pDCs we turned to an in vitro system. Bone marrow expanded and isolated pDCs were enriched and incubated with Sendai ICs overnight and sampled for IFNα production. While both WT and hFcγRIIA pDCs responded equally to virus alone (Fig. 4D) only the transgenic, activating FcγR bearing pDCs produced interferon in response to viral ICs (Fig. 4E). This was not the result of an increased IC binding potential by the transgenic pDCs over the WT, as both bound SeV ICs equally (Fig. 4F and G).

Thus, taken together the data suggests that the prevention of an IFNα response to SeV IC is not because neutralizing antibody renders the complex inert, but rather it is the dominant expression of FcγRIIB on murine splenic pDCs that prevents IFNα production during viral memory responses.

Splenic pDC IC binding occurs via FcγRIIB and prevents trafficking to TLR9 signaling compartments

SeV ICs did not induced IRF-7 translocation to the nuclei in WT pDCs (data not shown), suggesting that the block to IFNα production occurred upstream, potentially by derailing trafficking of the immune complex cargo to the canonical TLR signaling endolysome. We previously reported that murine pDCs bind and internalize Ovalbumin ICs via FcγRIIB exclusively(25) and failed to reach the proteolyic endosome required for antigen processing. Therefore, we similarly assessed the intracellular trafficking of OVA-ICs and TLR ligand containing ICs (OVA-CpGB ICs), which may interact directly with endosomal TLR receptors in pDCs, and be internalized via an FcγRIIB independent manner.

First to assess the FcγR dependence of OVA-CpG IC binding, we compared the IC binding potential of WT, FcγRIIB deficient and FcγR null splenic pDCs by confocal microscopy. We found that in the absence of FcγRIIB, Ovalbumin IC binding was significantly reduced on splenic pDCs and equaled that of the FcγR null pDCs (Fig. 5A, B) confirming previous experiments (Supplementary figure 2D). The addition of a TLR9 moiety (CpG B) on the antigen did not alter the FcγRIIB dependence (Fig. 5B, right panel), strongly suggesting that WT pDCs exclusively bind ICs through the inhibitory FcγRIIB.

Figure 5. Splenic pDCs Internalize ICs via FcγRIIB and traffic particles to early endosomes.

A) A representative example of splenic pDCs binding Ova ICs. B) Indicated ICs were incubated with splenic pDCs for 30 minutes and analyzed for IC binding by confocal microscopy. Ova IC and Ova CpG IC graphs were generated from at least two experiments. Each symbol represents the average number of IC particles per cell per image. C) Representative example of WT and hFcγRIIA Tg splenic pDCs incubated with Ova ICs and stained for Lysotracker® (green) and anti-rabbit Ova-ICs (red). D) Quantification of experiments provided in C, showing integrated density of Lysotracker® at IC positive ROI. Three pooled experiments are shown with over 150 IC particles analysed. E) Representative example of WT or hFcγRIIA Tg splenic pDCs incubated with Ova ICs (red) and LAMP-1 (green) and DAPI (blue). F) Quantification of experiments as in E for integrated density of LAMP-1 at IC positive ROI. Two pooled experiments are shown. G) Representative example of WT and hFcγRIIA Tg splenic pDCs incubated with Ova ICs and stained for LC3 (green) and anti-rabbit Ova-ICs (red). H) Quantification of experiments as in E for integrated density of LC3 at IC positive ROI. Two pooled experiments are shown.

We next sought to examine the intracellular trafficking pathways that may prevent viral ICs from reaching a TLR signaling compartment in pDCs. OVA ICs and OVA-CpG ICs were incubated with splenic pDCs and colocalization with various intracellular markers was assessed by immunofluorescence and confocal microscopy.

IFNα production by TLR9 ligands requires engagement of IRF7 in Lamp2 positive, Lysotracker® positive endosomes(26). We examined the trafficking of ICs in splenic pDCs and found that Ova ICs incubated with WT splenic pDCs were not colocalized with Lysotracker® as was expected from previous work(25). However, the transgenic expression of activating FcγRIIA significantly increased colocation of IC and Lysotracker (Fig 5 C, D) and results were confirmed with studies using Lamp1 (Fig 5 E, F).

Interestingly, when CpG was present in the cargo, WT pDCs delivered more ICs to Lysotracker® positive compartments, than ICs without the TLR moiety, however, these finds were not supported by the Lamp1 staining (Fig 5 E, F). In the presence of the activating FcγRIIA, however, CpG coupled Ova ICs colocalized more than three fold above WT with Lysotracker®. Furthermore, the FcγRIIA dependent increase in the colocalization of IC to acidic compartments was supported by Lamp1 (Fig 5 E, F). Thus FcγRIIB internalized immune complexes accesses a spatiotemporal endosomal pathway, distinct from activating FcγR receptors that inefficiently access the acidic, LAMP1+ IRF7+ IFNα signalosome, thus preventing IFNα production.

Recently published work has identified a non-canonical interferonogenic pathway involving the LC3-associated phagosome. In these studies using bone marrow derived murine pDCs, the investigators found that antibody-coated beads were internalized via LC3, an autophagy related protein, and that the recruitment of LC3 was Fc gamma chain-dependent(27) and both were functionally required for IFNα production. We performed a similar experiment using splenic pDCs and soluble ICs. Recruitment of LC3 was measured by calculating the integrated density of LC3 fluorescence at each IC particle (ROI, Region of Interest). We controlled for background fluorescence by measuring the same parameters and ROIs on images pivoted 90 degrees. Our experiments yielded results consistent with an LC3 mediated internalization through activating FcγRs (Fig 5 G, H). However, we found that FcγRIIB did not recruit LC3 above background when using either WT or gamma chain deficient pDCs. Indeed, only when an activating FcγR was transgenically expressed did we detect LC3 recruitment on splenic pDCs. Furthermore, the presence of a TLR moiety coupled to Ovalbumin did not change LC3 recruitment in the WT pDC supporting the conclusions of Henault et. al. that LC3 recruitment is initiated by the IgG component of the IC and not by its cargo. Our results show that ICs internalized through the inhibitory FcγRIIB does not lead to LC3 recruitment.

Taken together, our data demonstrates that FcγRIIB on splenic murine pDCs is required to internalize ICs and that this internalization occurs outside of LC3 positive compartments and results in IC trafficking away from the Lysotracker positive endolysomes where IRF7 and TLR9 signal. In contrast, ICs internalized through activating FcγRs access an LC3-mediated internalization pathway that leads to LAMP1+, Lysotracker+ endolysomes, enabling MYD88 signaling and IRF7 nuclear translocation. In human pDCs, in which both activating and inhibitory FcγRs are co-expressed, as modeled here, the amplitude of the IFNα signal would be balanced by the relative functional expression of these opposing receptors. This situation may be relevant to primates, including humans, in which the activating FcγRIIA is dominant. In mice, the exclusive expression of FcγRIIB should prevent untoward IFNα production during recurrent or chronic exposure to a virus, in which the presence of adequate functional viral-specific neutralizing antibodies enable benign clearance and limited viral replication and systemic spread.

Discussion

In the present study, we have described one mechanism by which the innate IFNα response is averted upon secondary exposure to virus. We have determined that the presence of circulating virus-specific antibody and the dominance of the inhibitory Fcγ receptor on pDCs are required for preventing the potentially toxic side effects of systemic interferon. Thus, the humoral immune memory response functions to temper the innate response upon reexposure to virus.

Our work also describes a molecular pathway that differentiates activating and inhibitory FcγRs in their ability to traffic immune complexes for TLR sensing. We have taken several approaches, most of which have been technically challenging, including using TLR9 antibodies to demonstrate that OVA-CpG IC internalized by WT pDCs failed to colocalize with TLR9, while hFcγRIIA expressing pDCs showed remarkable colocalization (data not shown). Staining with TLR9 specific antibodies, however, was too weak for rigorous quantitative microscopy. In a second approach we reasoned that pulsing pDCs with CpG-FITC (purchased from Invivogen, San Diego CA) would delineate a pathway known to engage TLR9 and lead to IFNα production. However, these experiments were also fraught with unreliable staining. Thus, we took advantage of well-described studies that reveal the requirement of acidification for TLR9 activation and IRF7 recruitment(26, 28) and capitalized on the highly reproducible staining we achieved with Lysotracker which reliably stains the acidic lysosomal organelles. Using this approach we found that FcγRIIB on splenic pDCs was required for binding viral immune complexes and that subsequent internalization failed to deliver ICs to Lysotracker positive compartments. These results were further confirmed by similar findings with Lamp-1 and 2. In contrast, the transgenic expression of an activating FcγR was sufficient to restore IFNα production and delivered ICs to Lysotracker+ Lamp-1 and Lamp-2 + endolysosomes.

A recent study also described a non-conical, autophagy related pathway that is alternatively used by activating FcγRs for IFNα induction by DNA containing ICs(27). This pathway recruits LC3 to the IC in an IgG dependent manner. Our studies showed that LC3 was not recruited by WT splenic pDCs, nor by gamma chain KO pDCs, but only when the activating FcγR was transgenically expressed. Taken together, these studies suggest that activating and inhibitory FcγRs are discriminating pathways that modulate TLR sensing by directing ICs either towards or away from TLR sensing compartments. Furthermore, pDCs, by dominantly expressing the inhibitory FcγR, are prone to suppress the IFNα response to ICs.

Our work supports the findings of Henault et. al. that LC3 is utilized by activating FcγRs to delivery cargo to TLR sensing compartments. However, they show that bone marrow expanded WT pDCs do recruit LC3 to internalized ICs in a gamma chain dependent manner. Our studies have consistently identified a lack of activating FcγRs on murine splenic pDCs. This apparent conflict could be due to the source of pDCs and contaminating cDCs. We found that expanding bone marrow with FLT3L to obtain pDCs, inconsistently altered the Fcγ receptor expression compared to splenic pDCs. B220 and CD11b, markers used in the literature to isolate pDCs and cDCs from FLT3L cultures, were often not mutually exclusive or were suspected of containing a large percentage of cDC precursors, as has been reported(29). By using primary splenic pDCs to investigate the FcγRIIB dependent requirement for binding, internalizing and intracellular trafficking of ICs, we have circumvented the problems that might arise by exposing dendritic cells to fetal calf serum and an inflammatory cytokine such as FLT3L.

Our major finding that IFNα production by splenic pDCs was prevented during a memory response, might have been explained by a lack of access to the virus during transport through the spleen in a memory setting. However, we found that in both primary and memory responses, pDCs had comparable access to the virus. Indeed, it may be possible that the same molecular mechanisms may be directing traffic in both settings, i.e. complement by acting with natural IgG and IgM antibody in the primary response and with virus specific IgG antibody in the memory(30–33) which ultimately delivers virus to follicular B cells for trafficking and presentation by follicular dendritic cells.

Other mechanisms that may have contributed to the prevention of an interferon response during a memory setting may include pDC ‘exhaustion’ due to low-level viral exposure(34). We found by western blot that Sendai virus did not persist in the spleens of mice by day 14 (Figure 1E) and in naïve mice we found weight was recovered by day 7 post Sendai injection, suggesting that they had cleared the virus by day 14 and 21 (data not shown). However, during the course of our studies, we found that there was a tendency of previously SeV immunized mice to respond less robustly to unrelated stimuli (not statistically significant and data not shown). This tendency may have contributed to the absence of a response to SeV in memory mice. However, we also found that passive immunization with antisera was sufficient to suppress the interferon response in naïve mice (Figure 3B) and that the transgenic expression of activating FcγRs could reverse the interferon suppression in a memory setting with presumably equally exhausted mice (Figure 4).

Our current study highlights the dominance of FcγRIIB on pDCs as a mechanism to prevent IFNα production by TLR bearing ICs. Systemic infections elicit IFNα from pDCs and provide the host a protective, potent and immediate source of IFNα. In the setting of a well-orchestrated adaptive memory response, this innate production of IFNα is mitigated by circulating virus specific antibodies and through inhibitory Fcγ receptors on pDCs. Would pDCs express an FcγR repertoire equal to that of their conventional cDC counterparts, secondary infections would lead to unnecessary IFNα production and elicit an exacerbated immune response that could lead to untoward consequences of injury and or autoimmmunity.

Alternatively, a beautiful study by Honke et. al. highlights another possible mechanism for halting IFNα production in a memory setting. They found that IFNα resistance by MOMA-1 positive macrophages during primary viral responses is required for the mounting of a protective adaptive immune response. By over-expressing Usp18, an inhibitor of IFNAR signaling, MOMA-1+ macrophages at the marginal zone capture virus and allow viral replication for presentation to T and B cells(33). Thus, during memory responses, a synergistic action to improve upon the memory response could be aided by viral immune complex presentation and the viral replication allowed by the absence of interferon.

The FcγRIIB specific means of controlling pDC derived IFNα may be important as well during chronic infections when circulating viral immune complexes may be present. In a classic model of chronic LCMV infection, Zuniga et. al. (34) found that pDCs were prevented from making interferon in response to CpG 30 days post infection with LCMV. This timepoint correlates with the presence of neutralizing antibodies as gauged by Bergthaler et. al.(35). It is possible then that in this model of chronic infection, IFNα production is suppressed as a result of circulating viral immune complexes being sensed through FcγRIIB on pDCs.

Human pDCs, have been shown to dominantly express the activating receptor FcγRIIA(36). Nevertheless, evidence that human pDCs, like mice, exhibit an Fc-dependent inhibition of IFNα production has been previously described(37, 38) and may have important implications for vaccine development. In one prominent example, a Merck designed Adenovirus based HIV vaccine (Ad5/HIV) showed an increase in HIV susceptibility in patients with preexisting antibodies to Adenovirus(39). A separate study, using a systems analysis approach, identified the production of Type 1 Interferon as the most significant difference between those with and without preexisting antibodies to adenovirus(40). In another example, the whole inactivated influenza vaccine elicits a strong IFNα production and a potent adaptive immune response(41) whereas, the split vaccine mediates immunity through Fc receptors and prevents an IFNα response(41, 42). Not surprisingly the latter is far less immunogenic, and children, without preexisting flu antibodies, require two vaccinations for efficacy. These studies point to the perils involved in mounting an adaptive immune response in the absence of an innate immune response. In the case of the Ad5/HIV clinical trial, the lack of interferon during immune priming could have resulted in the production of non-neutralizing antibodies, through a lack of optimal B cell activation, and subsequent antibody enhanced infection of HIV(43).

In our own unpublished results and in those of others(37, 44), we found that both humans and mice have mechanisms to prevent IFNα production from pDCs when exposed to immune complexes. In humans, this mechanism is mediated my monocytes responding to immune complexes(44, 45). In the mouse, we have presently described a mechanism mediated through FcγRIIB directly on pDCs. Understanding whether the inhibition in the human is similarly controlled by the inhibitory FcγR on monocytes is difficult without a commercially available antibody specifically recognizing FcγRIIB. However, future work should include devising vaccine strategies including adjuvants that block this mechanism to restore the interferon response and elicit better immune protection.