Immature lung TNFR2− conventional DC 2 subpopulation activates moDCs to promote cyclic di-GMP mucosal adjuvant responses in vivo

Cyclic dinucleotides (CDNs), including cyclic di-GMP (CDG), are promising vaccine adjuvants in pre-clinical/clinical trials. The in vivo mechanisms of CDNs is not clear. Here we investigated the roles of lung DCs subsets in promoting CDG mucosal adjuvant responses in vivo. Using genetically modified mice and adoptive cell transfer, we identified lung conventional DC 2 (cDC2) as the central player in CDG mucosal responses. We further identified two functionally distinct lung cDC2 subpopulations: TNFR2+pRelB+ and TNFR2−pRelB− cDC2. The TNFR2+ cDC2 were mature and migratory upon intranasal CDG administration while the TNFR2− cDC2 were activated but not mature. Adoptive cell transfer showed that TNFR2− cDC2 mediate the antibody responses of CDG, while the TNFR2+ cDC2 generate Th1/17 responses. Mechanistically, immature TNFR2− cDC2 activate monocyte-derived DCs (moDCs), which do not take up intranasally administered CDG. moDCs promote CDG-induced generation of T follicular helper- and germinal center B-cells in the lungs. Our data revealed a previously undescribed in vivo mode of DCs action whereby an immature lung TNFR2− cDC2 subpopulation directs the non-migratory moDCs to generate CDG mucosal responses in the lung.


Introduction
Adjuvants improve vaccine safety profiles and enhance, and shape, antigen-specific immune responses. Understanding the mode of action of adjuvants is key for the development of Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use:http://www.nature.com/authors/editorial_policies/license.html#terms rationally designed modern vaccines. Recently, the small molecule cyclic dinucleotides (CDNs) have emerged as a group of promising vaccine adjuvants in preclinical and clinical trials 1 . CDNs include the bacterial second messengers cyclic di-GMP (CDG), cyclic di-AMP, 3'3'-cyclic GMP-AMP and the mammalian second messenger 2'3'-cyclic GMP-AMP 1 . CDG is the founding member and the most studied CDNs 2 . As a mucosal adjuvant, CDG does not cause acute toxicity in mice 1 . Furthermore, CDG is a more potent activator of Th1 and Th2 immune response than LPS, CpG oligonucleotides (ODN) or aluminum salt-based adjuvant in mice 3 . Last, CDG adjuvanted vaccines protect mice from H5N1 influenza 4 , Acinetobacter baumannii 5 , Staphylococcus aureus 6 , Klebsiella pneumoniae 7 and Streptococcus pneumoniae 8,9 . CDG also showed cancer vaccine adjuvant activity in animal models 10,11 .
MPYS, also known as STING and MITA, is a receptor for CDNs 12 and a critical player in sensing cytosolic DNA 13,14 . MPYS −/− mice lose CDNs adjuvant activity 15 . Additionally, TNF signaling, not type I IFN signaling, is essential for the adjuvant activity of CDG in vivo 15,16 . Notably, when administered intranasally, CDG not only induced lung production of TNF, IL-1β but also the anti-inflammatory cytokine IL-10 17 . Consequently, CDG adjuvant does not induce exaggerated inflammation responses in the lung 17 . The precise in vivo mechanism by which TNF mediates CDG adjuvant activity in vivo is unknown.
In this report, we revealed important heterogeneity in the lung cDC2 population and identified the cDC2 subpopulation that is responsible for CDG adjuvant effect. Surprisingly, the antibody responses of CDG adjuvant depends on an activated but immature TNFR2 − cDC2 subpopulation, which drive moDCs maturation to generate T follicular helper (Tfh) cells in the lung.
AM (CD11c + MHC II int ) took up most of the fluorescent CDG in vivo ( Fig S2A). Unlike DCs, AM did not increase expression of CD86 following CDG treatment (Fig S2B). To determine whether AM are required for CDG responses in vivo, we used the MPYS fl/fl LysM cre mice 17 , which deleted MPYS expression in AM (Fig S2C). The activation of lung DCs by CDG was unaltered in the MPYS fl/fl LysM cre mice ( Fig S2D). Importantly, MPYS fl/fl LysM cre mice produced similar anti-pneumococcal surface protein A (PspA) antibody as the WT upon CDG/PspA immunization ( Fig S2E).

CDG indirectly activates moDCs
Although moDCs did not take up CDG, they increased expression of CD86 in response to intranasal administration of CDG ( Fig 1C). This was independent of MPYS expression in moDCs as moDCs in MPYS fl/fl LysM cre mice had normal levels of CD86 expression ( Fig  S2D). Notably, activated moDCs did not increase CCR7 ( Fig 1D) and did not migrate to dLNs ( Fig S3B). Last, moDCs activated both RelA and RelB in response to CDG (Fig 1G). We concluded that CDG indirectly activate moDCs and activated lung moDCs were not migratory 25,33 .

cDC2 play a central role in mediating CDG adjuvant activity
We next asked which cDC subset mediates CDG adjuvant activity. IRF4 fl/fl CD11c cre mice lack cDC2 in the lung 28,29 (Fig S1C). The cDC1 and moDCs were retained in the IRF4 fl/fl CD11c cre mice ( Fig S1C). We examined CDG adjuvant activity in the IRF4 fl/fl CD11c cre mice. Mice were intranasally administered with PspA alone or with CDG. PspA-specific Ab responses were examined in the blood and bronchoalveolar lavage fluid (BALF). Unlike the WT mice, CDG did not induce anti-PspA Abs in BALF and serum ( Fig  1H, S3D-E) from immunized IRF4 fl/fl CD11c cre mice. Ex vivo recall assay in lung cells and splenocytes from immunized IRF4 fl/fl CD11c cre mice also did not show PspA-specific Th1, Th2 or Th17 responses (Fig S3F-G).
To further demonstrate that lung cDC2 mediate the adjuvant activity of CDG, we adoptively transfer (i.n.) WT lung cDC2 into IRF4 fl/fl CD11c cre mice. The recipient IRF4 fl/fl CD11c cre mice were then immunized with CDG/PspA. We found that adoptive transfer of WT cDC2 generated PspA-specific serum IgG and IgA in the IRF4 fl/fl CD11c cre mice similar to the WT mice ( Fig 1I). We concluded that lung cDC2 are critical for CDG adjuvant activity.
In contrast to the IRF4 fl/fl CD11c cre mice, Batf3 −/− mice mounted antigen-specific IgG and IgA responses in a manner comparable to the WT following CDG/PspA immunization ( Fig  1H, S3D-E). We concluded that cDC2 play a central role in mediating CDG adjuvant activity. Batf3 −/− mice had impaired Th1 responses following immunization (Fig S3F-G). Whether the defect is due to the lack of cDC1 remains to be determined since T cells also express Batf3.

RelB in DCs is required for CDG-induced cDC2 maturation in vivo.
TNFR2 on lung cDC2 is required CDG-induced cDC2 maturation in vivo ( Fig 2C). We reasoned that RelB was required for CDG-induced cDC2 maturation too. We used RelB fl/fl CD11c cre mice to ablate RelB in DCs. RelB fl/fl CD11c cre mice have normal DC populations, with all subsets intact ( Fig S5E). In the absence of RelB in DCs, cDC2 failed to upregulate CD86 in response to CDG (Fig 2H). cDC1, which do not activate RelB, upregulated CD86 (Fig S5F). We concluded that CDG-induced lung cDC2 maturation depends on the cell-intrinsic signal of TNFR2-RelB.

TNFR2 + cDC2 are required for CDG-induced Th1 and Th17 responses but dispensable for the humoral responses
We next sought to determine if TNFR2 expression on cDC2 was needed for the adjuvant activity of CDG. Unexpectedly, even though adoptively transferred TNFR2-deficient cDC2 failed to upregulate CD86 in response to CDG (Fig 2C), they induced serum anti-PspA IgG and IgA when transferred into IRF4 fl/fl CD11c cre mice ( Fig 3A). In fact, the anti-PspA IgG and IgA in IRF4 fl/fl CD11c cre mice receiving TNFR2-deficient cDC2 and WT cDC2 were comparable ( Fig 3A).
We then examined CDG adjuvant activity in the RelB fl/fl CD11c cre mice, which also lack mature cDC2. Upon immunization with PspA and CDG, RelB fl/fl CD11c cre mice produced normal levels of IgG and IgA ( Fig 3B). RelB fl/fl CD11c cre mice failed to induce Th1 and Th17 responses in the lung ( Fig 3C).

TNFR2 defines two functionally distinct subpopulations of lung cDC2
cDC2 is a heterogeneous population 21,32,38,39 . We showed that lung TNFR2 + pRelB + cDC2 were mature and required for the Th1/Th17 responses but not humoral responses while the TNFR2 − cDC2 was not mature but mediates CDG-induced antibody response. We assessed whether TNFR2 expression could define functionally distinct lung cDC2 subpopulations.
Adoptively transferred TNFR2 −/− cDC2 restored CDG responses in IRF4 fl/fl CD11c cre mice ( Fig 3A). Yet, TNFR2 −/− mice had no CDG responses (Fig 2A). We reasoned that TNFR2 expression on moDCs may be important for CDG responses in vivo. Indeed, we found that CDG induced TNFR2 on moDCs in WT mice ( Fig 7E) and moDCs from TNFR2 −/− mice did not upregulate CD86 in response to CDG in vivo ( Fig 7F). Last, adoptive transfer of WT monocytes into TNFR2 −/− mice restored CDG-induced IgG and IgA responses ( Fig 7G). Notably, adoptive transfer WT cDC2 into TNFR2 −/− mice did not restore CDG-induced antibody responses (Fig 7H). We concluded that moDCs expression of TNFR2 is critical for its maturation and subsequent induction of CDG adjuvant response.

moDCs, not TNFR2 − cDC2, are very efficient in antigen processing in vivo
We next examined antigen processing in TNFR2 − cDC2 in vivo. WT mice were intranasally administered with CDG/DQ™-OVA 17 . DQ + cells were examined in lung cDC2 and moDCs (Fig 8D). DQ™-OVA is a self-quenched conjugate of OVA exhibiting bright green fluorescence upon proteolytic degradation (DQ-Green). Furthermore, high concentration of digested fragments of DQ™-OVA accumulating in organelles form excimers that exhibits bright red fluorescence (DQ-Red). We found that DQ + moDCs are mostly DQ-Red indicating a high concentration of processed antigens in moDCs. Conversely, DQ + cDC2 were DQ-Green (Fig 8D). Strikingly, comparing to the TNFR2 + cDC2 subpopulation, very few TNFR2 − cDC2 subpopulation were DQ + (Fig 8D) suggesting that the TNFR2 − cDC2 either did not take up antigen or were not efficient at antigen processing. In comparison, all DQ + moDCs were TNFR2 + cells (Fig 8D) indicating TNFR2 + moDCs were indeed mature DCs.

moDCs promote CDG-induced Tfh and GC B cells generation in the lung
We next assessed how non-migratory moDCs (Fig 1D & S3B) 25,33 promote CDG-induced antibody responses. moDCs were efficient at antigen processing (Fig 8D). We first asked if they presented antigen on cell surface. We intranasally administered C57BL/6 mice with CDG and Eα-OVA, and detected I-A b /Eα + cells with the YAE mAb. Indeed, CDG increased YAE + moDCs in vivo (Fig 9A). Furthermore, the majority of YAE + moDCs upregulated CD86 (Fig 9B), indicating their potential to activate CD4 + T cells.
Tfh cells and GC B cells play central roles in promoting humoral responses. We found that 14 days after CDG/PspA immunization (i.n.), lungs from the WT mice had PD1 + CXCR5 + Bcl6 + Tfh cells and Bcl6 + GC B cells (Fig 9C-9E). In contrast, TNFR2 −/− mice were unable to generate lung Tfh or GC B cells (Fig 9C-9D). Importantly, adoptive transfer of WT monocytes into TNFR2 −/− mice restored the generation of Tfh and GC B cells in the lung (Fig 9C-9D). We concluded that moDCs promote CDG-induced Tfh and GC B cells generation in the lung. moDCs are activated by TNFR2 − cDC2. Thus restoring TNFR2 − cDC2 in the IRF4 fl/fl CD11c cre mice should restore Tfh cells. Indeed, we found that adoptive transfer of TNFR2 − cDC2, but not TNFR2 + cDC2, into IRF4 fl/fl CD11c cre mice generated Tfh cells ( Fig  9E). Together, we propose that CDG activates TNFR2 − cDC2 that matures moDCs to generate Tfh and GC B cells promoting CDG-induced antibody responses in vivo (Fig 9F).

Discussion
In this report, we examined the mechanism by which lung DCs subsets mediate the mucosal adjuvant activity of CDG. The most exciting finding in this report is the identification of new lung cDC2 subpopulations and their unusual mode of action. cDC2 is a heterogeneous population 21,32,38,39 . We found that steady-state lung cDC2 have two distinct subpopulations TNFR2 + pRelB + CX3CR1 − and TNFR2 − pRelB − CX3CR1 + . Functionally, these two cDC2 subpopulations mediate the cellular and humoral immune responses to CDG adjuvant respectively. Developmentally, they derived from pre-cDC2 and do not represent different activation states of cDC2 in vivo.
Lung DC subsets are likely influenced by lung microenvironment 43 . In a large scale of phenotypic and transcriptional profiling of human tissues DC subtypes, Heidkamp et. al., found that the phenotype of DCs is predominantly determined by ontogeny in the lymphoid organs whereas the phenotype of DCs is heavily influenced by the microenvironment in barrier tissues 43 . The lung TNFR2 − cDC2 express CX3CR1. Nakano H., et.al., showed that CX3CR1 promote pre-cDC migration to the lung at steady state 44 . The lung TNFR2 + pRelB + cDC2 have not been described before. They have constitutively activated TNFR2-RelB signaling and lack mTNF themselves. We speculate that they react to modulatory signals (e.g. mTNF) from lung microenvironment that constitutively activates TNFR2-RelB. Tussiwand, R. T. et al., previously identified a Klf4-dependent SIRP-α + CD24 + Mgl2 + cDC2 subpopulation that is required for Th2 response 39 . A subpopulation of TNFR2 + pRelB + cDC2 have high expression of Mgl2/CD301b. More studies are needed to determine if Klf4 is required for their development Intranasal administration of CDG leads to the maturation of the TNFR2 + cDC2 subpopulation, not the TNFR2 − cDC2 subpopulation. Unexpectedly, TNFR2 − cDC2 mediates the antibody responses of CDG. Why did not CDG activation mature lung TNFR2 − cDC2 in vivo? NF-κB activation is essential for DC maturation. Different from cDC1 or TNFR2 + cDC2, CDG did not activate RelA or RelB in lung TNFR2 − cDC2 in vivo. CDG did induce p-TBK1 in lung TNFR2 − cDC2. TBK1 is critical for IRF3 activation and IFNβ production 45 . However, it does not play a major role in NF-κB activation in vivo 46, 47 . Previous studies, mostly done in vitro, showed that STING/MPYS pathway activates NF-κB, in particular, RelA 15, 48 . Our results here indicated that the ability of STING/MPYS to engage NF-κB pathway is cell-type specific. In this lung resident TNFR2 − cDC2, in vivo stimulation of STING/MPYS by CDG do not activate RelA or RelB.
How does the immature TNFR2 − cDC2 promote CDG adjuvant responses in vivo? Our monocyte adoptive transfer experiment in TNFR2 −/− mice showed that moDCs are critical for CDG adjuvant responses. moDCs do not directly take up intranasally administered CDG and STING expression in moDCs is dispensable. Instead, moDCs maturation requires TNFR2 − cDC2 and the expression of TNFR2 on moDC. Though we can not rule out the possibility that mTNF on other cells interacts with TNFR2 on moDCs, we favor the model that mTNF on TNFR2 − cDC2 engages TNFR2 on moDC to induce its maturation and subsequent CDG antibody responses.
Lung moDCs are non-migratory 25,33 . We showed that moDCs promote the generation of Tfh and GC B cells in lung suggesting the formation of the inducible bronchus-associated lymphoid tissue (iBALT). DCs are required for the formation of iBALT 49, 50 . The exact lung DC subset for iBALT induction is unknown. moDCs presented antigen, expressed costimulator and activated RelA/RelB, which likely facilitate cytokine productions. We proposed that lung moDCs induce iBALT formation and promote CDG humoral responses.
In summary, we illustrated a previously unknown in vivo mode of action for CDG mucosal adjuvant whereby a new lung TNFR2 − cDC2 subpopulation activated by CDG promoting the maturation of moDCs for the generation of Tfh cells. These findings will facilitate future mucosal vaccine development and DC research.

Mice
Eight to sixteen-week old mice, both males, and females, were used for experiments. All mice are on a C57BL/6 background. A detailed description of the lines can be found in the Supplemental Methods. Mice were housed and bred in the Animal Research Facility at the University of Florida. All experiments with mice were performed by the regulations and approval of the Institutional Animal Care and Use Committee from the University of Florida.

Intranasal CDG Immunization
A detailed description of intranasal vaccination and reagents can be found in the Supplemental Methods. Sera were collected 14 days after the last immunization. The PspAspecific Abs were determined by ELISA. To determine Ag-specific Th response, splenocytes and lung cells from PspA or CDG + PspA immunized mice were stimulated with 5µg/ml PspA for four days in culture. Th1, Th2, and Th17 cytokines were measured in the supernatant by ELISA.

Detection of Lung Cytokine Production
Mice were intranasally administered 5µg CDG, then sacrificed after 5hrs by CO2 asphyxiation 17 . Lungs were harvested and lung cytokines was determined in lung homogenates. A detailed description of lung homogenates preparation can be found in the Supplemental Methods.

Isolation of lung cells
Mice were intranasally administered with or without CDG (5µg, vaccine-grade). After 20hrs, the lungs were lavaged, perfused with ice-cold PBS and harvested. A detailed description of lung cell isolation can be found in the Supplemental Methods.

Flow Cytometry and cell sorting
A detailed description of Flow antibodies used can be found in the Supplemental Methods. Cell sorting was performed on the BD FACSAriaIII Flow Cytometer and Cell Sorter. After sorting, dendritic cells were CFSE labeled, according to the protocol from the manufacturer (Invitrogen).

Intracellular staining
The intracellular cytokine staining was performed using the Cytofix/Cytoperm™ kit from BD Biosciences (cat#555028). Briefly, mice were intranasally administered saline or cyclic di-GMP (5µg, vaccine-grade). The single lung cell suspension was fixed in Cytofix/perm buffer (BD Biosciences) in the dark for 20min at RT. Fixed cells were then washed and kept in Perm/wash buffer at 4°C. Golgi-plug was present during every step before fixation.

Mouse cDC2 and monocyte purification
Primary mouse cDC2 (cat#18970A, Stemcell Technologies; cat# 480097, Biolegend) were purified from lungs of naïve mice following the protocol according to the manufacturer. Mouse monocytes (cat#19861, Stemcell Technologies) were purified from the bone marrow of naïve mice following the protocol according to the manufacturer.

Adoptive transfer
Lung TNFR2 + and TNFR2 − cDC2 were sorted from the lungs of naïve donor mice with a FACSAriaIII flow cytometer. After sorting, dendritic cells were CFSE labeled, according to the protocol from the manufacturer (Invitrogen). Cells were administered intranasally into recipient mice. 24 hours later of transfer, recipient mice were intranasally vaccinated with CDG (5µg, Invivogen, cat# vac-cdg) adjuvanted PspA (2µg, BEI Resources) or PspA alone 17 . Recipient mice received two doses of transferred cells and were immunized at 14 days interval.

Statistical Analysis
All data are expressed as means ± SEM. Statistical significance was evaluated using Prism 5.0 software to perform a Student's t-test (unpaired, two-tailed) for comparison between mean values.

Lung digestion
The lungs were lavaged, perfused, and harvested at 5hr post-treatment. Excised lungs were washed in PBS and digested in DMEM containing 200µg/ml DNase I (Roche, 10104159001), 25µg/ml Liberase TM (Roche, 05401119001), at 37°C for 3hrs. Red blood cells were then lysed and a single cell suspension was prepared and analyzed by BD™ LSR II and FACScan flow cytometry.

Measure cytokines in lung homogenates
Lungs were perfused with cold PBS. The harvested lungs were washed with PBS once, then stored in 0.7ml Tissue protein extraction reagent (T-PER) (Thermo Scientific, cat#78510) containing protease inhibitors (Roche, cat#11836153001) at −80°C. Later, the lung was thawed on ice and homogenized with Minilys® (Precellys, 5,000 RPM for 30sec) using Precellys lysing kit (Precellys, cat# KT03961). Lung homogenates were transferred to a 1.5ml tube and spun at 14,000g for 30min at 4°C. The supernatant was collected and analyzed for TNF production by ELISA (eBioscience, cat#88-7324).

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.       A-B. Indicated mice were administered (i.n.) with saline or CDG for 5hrs. TNF production was measured in lung homogenates by ELISA. n>3. C. WT and TNFR2 −/− mice were treated with saline or CDG for 16 hours. TNF in lung cDC2 was determined by an intracellular cytokine stain. n=3. D. TBK1 fl/fl and TBK1 fl/fl Vav cre mice were administered (i.n.) with saline or CDG for 5hrs. TNF production was measured in lung homogenates by ELISA. n=3. E. Flow cytometry analysis of p-TBK1 expression in lung cDC2 from WT and TNFR2 −/− mice treated with saline or CDG for 16 hours. n=3. Graphs represent means ± standard error from three independent experiments. The significance is represented by and asterisk (*) where p<0.05 (unpaired Student's t test).
into TNFR2 −/− mice were immunized with CDG/PspA. Serum anti-PspA IgG were determined by ELISA. n=3. Graphs represent means ± standard error from three independent experiments. The significance is represented by and asterisk (*) where p<0.05 (unpaired Student's t test). A. WT mice were administered with saline or CDG (5µg) for 16hrs. mTNF expression was determined by Flow cytometry using mouse TNFR2-Fc recombinant protein. n=3. B-C. Flow cytometry analysis of mTNF expression in lung DCs (B) and cDC2 (C) n=3. D. Flow cytometry analysis of antigen uptake and processing in lung CD11b + DC from WT mice treated (i.n.) with DQ-OVA (20ug) and CDG (5ug) for 16 hours. n=3. Graphs represent means ± standard error from three independent experiments. The significance is represented by and asterisk (*) where p<0.05 (unpaired Student's t test). A-B. WT mice were administered with Ea-OVA (10µg) or Ea-OVA/CDG (5µg) for 16hrs. YAE + moDCs (A) and CD86 + YAE + moDCs were determined by Flow cytometry. n=3. C-D.