Hydroxypropyl-β-Cyclodextrin Spikes Local Inflammation That Induces Th2 Cell and T Follicular Helper Cell Responses to the Coadministered Antigen

The term “adjuvant” has its origin in the Latin “adjuvare” meaning “to help,” because adjuvants enhance the effects of vaccines. Many substances can act as adjuvants, and their modes of action vary widely (1, 2). Recent studies showed that conventional adjuvants, such as alum, act via the “Ag-depot” mechanism, as well as induce multiple innate immune pathways, including the activation of inflammasomes, the production of PGE₂, and the release of damage-associated molecular patterns (DAMPs), such as DNA (3–5). In addition to enhancing the immunogenicity of vaccines, adjuvants can be responsible for their adverse effects, including local swelling, systemic fever, and a theoretical risk for autoimmunity (6). Even alum and MF59, widely used adjuvants for human vaccines, induce local and systemic adverse effects (6–8). Therefore, the high safety profile of an adjuvant, based on scientific evidence, must be demonstrated before its application to clinical practice. One approach to identifying ideal adjuvants is the selection of candidate substances from among approved drugs or excipients with very high safety profiles.

Cyclodextrin is a bucket-shaped oligosaccharide derived from starch, with a hydrophobic cavity and hydrophilic exterior (9). Natural cyclodextrins contain six (α-CD), seven (β-CD), eight (γ-CD), or more (α-1,4)-linked α-d-glucopyranose units, and many chemically modified cyclodextrins have been developed, including hydroxypropyl-β-cyclodextrin (HP-β-CD) and sulfobutyl ether-β-cyclodextrin (9). Because of their molecular structure and shape, the cyclodextrins can form an inclusion complex with hydrophobic molecules, improving the solubility of hydrophobic drugs (9), stabilizing proteins (10), and controlling the release of drugs (11). For these reasons, cyclodextrins are widely used excipients for pharmaceutical agents (9). Recent studies showed another application of cyclodextrins in the treatment of Niemann–Pick type C1 disease; they reduced the lysosomal accumulation of cholesterol and glycosphingolipids (12). However, only a few studies have investigated the usefulness of cyclodextrins as vaccine adjuvants. Mucosal vaccination with dimethyl-β-cyclodextrin was shown to enhance the immunogenicity of anti-diphtheria and anti-tetanus toxoid total IgG titers in mice (13). Sulfolipo-cyclodextrin showed improved adjuvant activity when combined with squalene for use in cattle (14). The high safety profiles of the cyclodextrins and their potential adjuvant activities demonstrated in these studies are very valuable contributions to the development of new and safer adjuvants. However, research into cyclodextrins as vaccine adjuvants has not been thorough, and the mechanism of action of HP-β-CD as a vaccine adjuvant and its immunological characteristics are not well understood. In this study, we focused on HP-β-CD, one of the cyclodextrins most commonly used in pharmaceutical preparations, because its solubility is higher and its safety profile is better than those of other cyclodextrins (9). We demonstrate its activity as an immune adjuvant and the mechanisms involved. Our results indicate that HP-β-CD is a potent and promising T follicular help (Tfh) cell adjuvant for purified protein–based vaccines.

The following were used as Ags: extremely pure LPS-free OVA protein ([Seikagaku, Tokyo, Japan]; LPS contamination was <0.1 EU/mg, as measured with the Limulus Color KY test [Wako Pure Chemical Industries, Osaka, Japan], according to the manufacturer’s protocol); influenza trivalent split vaccine (SV), containing influenza virus hemagglutinin (HA) surface Ags from three viral strains—A/California/7/2009 (H1N1), A/Victoria/210/2009 (H3N2), and B/Brisbane/60/2008; and monovalent SV, containing influenza virus HA surface Ag from New Caledonia/20/1999 (H1N1) (The Research Foundation for Microbial Diseases of Osaka University). SV was manufactured with egg-based technology (15). HP-β-CD (ISP Technologies, Assonet, MA), Good Manufacturing Practice grade K3 CpG-ODN (Gene Design, Osaka, Japan), and alum (InvivoGen, San Diego, CA) were used as adjuvants.

Mice deficient for Tlr1, Tlr2, Tlr3, Tlr4, Tlr5, Tlr6, Tlr7, Tlr9, Tlr7/9, Tlr2/4/9, Il1R, Il18R, St-2, Myd88, Tbk1, Irf3/7, Ifnar2, Asc, and Caspase 1 were generated and used for experiments previously (16). Tlr- or Myd88-knockout (KO) mice were purchased from Oriental BioService (Kyoto, Japan). Irf3/7 double-KO mice were generated by cross-breeding with Irf3-KO mice and Irf7-KO mice (17). Irf7-KO mice were provided by the RIKEN BioResource Center through the National Bio-Resource Project of the Ministry of Education, Culture, Sports, Science and Technology (Ibaraki, Japan). Bcl6^flox/+ Cd4-cre (Bcl6^f/+) and Bcl6^flox/flox Cd4-cre (Bcl6^f/f) mice were used for experiments previously (18). The other KO mice were housed and maintained at the National Institute of Biomedical Innovation. Wild-type (WT) C57BL/6 mice were purchased from CLEA Japan The mice were immunized s.c. twice with 100 μg OVA mixed with PBS (–), 30 μg K3 CpG-ODN, 1 mg alum, or 3–30% HP-β-CD [total volume: 200 μl in PBS (–)/mouse] on day 0 and day 10 after the first immunization. Sera were collected at 10, 17, 24, and 31 d after the first immunization. To evaluate the cellular immune responses, the mice were sacrificed and their spleens were collected 10 d after the second immunization. All experiments were performed under the appropriate laws and guidelines and after approval was obtained from the National Institute of Biomedical Innovation, the Animal Research Committee of the Research Institute for Microbial Diseases of Osaka University, and the Shionogi Animal Care and Use Committee.

To measure the OVA- or HA-specific Ab titers, 96-well plates were coated with 100 μg/ml OVA solution or 1 μg/ml SV solution overnight at 4°C. The plates were washed and incubated for 1 h with blocking buffer [PBS (–) containing 1% BSA and 0.05% Tween 20]. After blocking, the plates were washed and incubated with 3-fold serially diluted serum for 2 h. To detect the bound Ab, the plates were washed and incubated for 1 h with HRP-conjugated anti-mouse total IgG, IgG1, or IgG2c Ab (Southern Biotech, Birmingham, AL) or anti-monkey total IgG Ab (Sigma, St. Louis, MO). After the plates were washed, TMB substrate solution was added to each well to initiate the color reaction. The reaction was stopped by the addition of 2 N H₂SO₄, and the OD was measured at a wavelength of 450 nm (OD₄₅₀). The Ab titer was defined as the highest serum dilution that yielded an OD₄₅₀ > OD₄₅₀ of the negative-control serum. A DS Mouse IgE ELISA (OVA) kit (DS Pharma Biomedical, Osaka, Japan) was used to measure the concentration of anti-OVA IgE in the sera.

To evaluate the cellular immune responses, splenocytes (2 × 10⁶ cell/well) were prepared and incubated with complete RPMI 1640 medium, containing 20 μg/ml OVA class I peptide, 20 μg/ml OVA class II peptide, or 20 μg/ml OVA Ag for 48 h at 37°C under 5% CO₂. The concentrations of IFN-γ, IL-5, and IL-13 in the cell culture supernatants were measured with the DuoSet ELISA Development System (R&D Systems, Minneapolis, MN).

OVA–Alexa Fluor 647 (Invitrogen, Carlsbad, CA) and HP-β-CD–FITC (Nanodex, Kanagawa, Japan) were used to visualize the distributions of Ag and HP-β-CD in the inguinal lymph nodes (LNs). The mice were injected intradermally (i.d.) at the tail base with 100 μg OVA–Alexa Fluor 647 mixed with 3% HP-β-CD–FITC. After 30 min, 10 μl rat anti-MARCO Ab (clone ED31; Serotec, Kidlington, U.K.) was injected i.d. The inguinal LNs were excised 30 min later and visualized with two-photon excitation microscopy (FV1000MPE; Olympus, Tokyo, Japan). The imaging data were analyzed with Volocity 3D Image Analysis Software (PerkinElmer, Waltham, MA). Pearson’s correlation was calculated using Volocity colocalization analysis (PerkinElmer).

To evaluate Ag uptake, we used DQ–OVA (Invitrogen), a self-quenching conjugate of OVA that displays bright green fluorescence upon proteolytic degradation. The mice were injected s.c. with PBS (–), 100 μg DQ–OVA, or 100 μg DQ–OVA mixed with 30% HP-β-CD (total volume, 200 μl/mouse), and the LNs were excised 24 h later. The LNs were cut and incubated in CO₂-independent medium (Life Technologies, CA) containing 1 mg/ml collagenase D and 2 μg/ml DNase I (both from Roche, Penzberg, Germany) for 30 min at 37°C. The cells (2 × 10⁶ cell/sample) were prepared in FACS buffer [PBS (–) containing 1% FCS and 0.1% NaN₃]. FcγR was blocked with anti-mouse CD16/32 Ab (clone 93; eBioscience, San Diego, CA). Dead cells were stained with 7-aminoactinomycin D (eBioscience) and excluded from the analysis. To detect cells, Pacific Blue–conjugated anti-CD11c Ab (clone N418; BioLegend, San Diego, CA) was used. OVA uptake was assessed with a FACS LSR II flow cytometer (BD Biosciences, Franklin Lakes, NJ). The data were analyzed with FlowJo software (Tree Star, Ashland, OR).

Mice (n = 3) were injected i.d. or i.p. with a total volume of 200 μl 30% HP-β-CD in PBS (–). The vehicle-control mice (n = 3) were injected with the same volume of PBS (–) without adjuvant. After 6 h, the inguinal LNs, liver, and spleen were removed and stored in RNAlater RNA Stabilization Reagent (Life Technologies). Each sample in RNAlater was homogenized in Buffer RLT (QIAGEN, Hilden, Germany) by adding a 5-mm-diameter zirconium bead (AS ONE, Osaka, Japan) and shaking with a Mixer Mill 300 (QIAGEN) at 20 Hz for 5 min. The total RNA was isolated and purified from the sample homogenates with an RNeasy Mini Kit (QIAGEN), according to the manufacturer’s instructions. The gene expression profiles were determined using the 3′ IVT Express Kit and the GeneChip Mouse Genome 430 2.0 Array (both from Affymetrix, Santa Clara, CA), according to the manufacturer’s instructions. The expression values were normalized to the median value of each GeneChip readout. The resulting digital image files were preprocessed with the Affymetrix Microarray Suite version 5.0 algorithm (MAS5.0). The fold-change values were calculated as the ratio of the mean value for the treated samples/mean value for the vehicle-control samples. The presence/absence (PA) call in MAS5.0 was customized as described below. When the ratio was >1, the PA call was dependent on the treated samples, and when the ratio was ≤1, the call was dependent on the vehicle-control samples. Dominant calls (over half) were applied to a set of samples as the customized PA call, “P1” and “A0” (e.g., when the ratio < 1 and the PA call of control samples were “P,” “P,” and “A,” the customized PA call of the set was “P1”). The MAS5.0 data and PA calls were analyzed with the Bioconductor Affy package for R (http://www.bioconductor.org). The p values indicating the significance of differentially expressed genes were calculated using a t test when comparing the normalized treated samples with the normalized vehicle-control samples. For our subsequent analysis, we only included probes for which the fold change between the control and the stimulated samples was >2. We excluded a probe when it was flagged “absent” (i.e., the PA call was “A0”). The fold change of the probes with PA call “A0” were replaced with 0. Finally, we used Z-score scaling for all of the probes to generate Fig. 3. All datasets were deposited in the National Center for Biotechnology Information’s Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE63332.

Bcl6^f/+ or Bcl6^f/f mice were immunized s.c. twice with 100 μg OVA mixed with 30% HP-β-CD [total volume: 200 μl in PBS (–)/mouse] on day 0 and 10 d after the first immunization. Inguinal LNs were collected 7 d after the second immunization. To detect Tfh cells, PE anti-mouse CD4 (clone RM4-5; BioLegend), allophycocyanin anti-mouse CD279 (PD-1; clone 29F.1A12; BioLegend), and FITC anti-mouse CD185 (CXCR5; clone L138D7; BioLegend) were used. FcγR was blocked with anti-mouse CD16/32 Ab (clone 93; eBioscience). Tfh cell generation was assessed with a BD LSRFortessa (BD Biosciences). The gating strategy is shown in Fig. 4A. The data were analyzed with FlowJo software (Tree Star).

Cell viability after exposure to HP-β-CD was assessed in HEK293T cells (5 × 10⁴ cell/well). Cytotoxicity was examined with a water-soluble tetrazolium salt-8 assay using the Cell Counting Kit-8 (Dojindo Molecular Technologies, Kumamoto, Japan), according to the manufacturer’s instructions.

The mice were injected s.c. with 200 μl PBS (–) or 30% HP-β-CD. After 0.25, 0.5, 1, 3, 6, and 24 h, 400 μl PBS (–) was injected into the injection site, and the liquid in the pouch was collected. The liquid volume was adjusted to 1000 μl with PBS (–), and the acellular fraction was collected by centrifugation at 400 × g for 5 min at 4°C. The concentration of dsDNA in the acellular fraction was measured with a Qubit dsDNA HS Assay Kit (Invitrogen), according to the manufacturer’s protocol.

Mice were immunized s.c. twice with 3 μg SV mixed with PBS (–) or 3–30% HP-β-CD (total volume: 200 μl/mouse) on day 0 and 14 d after the first immunization. Sera were collected 14 and 21 d after the first immunization. Eight days after the second immunization, the mice were infected intranasally with 4 × 10⁴ 50% tissue culture infective dose (TCID₅₀) clinically isolated A/Osaka/129/2009 [A(H1N1)pdm09] influenza virus (19), which was obtained from Tetsuo Kase (Osaka Prefectural Institute of Public Health, Osaka, Japan). The body weights and survival rates of the mice were monitored for 14 d postinfection.

Twelve male 3–5-y-old cynomolgus macaques (Macaca fascicularis) were obtained from the Tsukuba Primate Research Center (TPRC) of the National Institute of Biomedical Innovation and randomly assigned to four groups (n = 3). The macaques were immunized s.c. with saline or 5 μg New Caledonia/20/1999 (H1N1) SV mixed with PBS (–), 3% HP-β-CD, or 30% HP-β-CD (total volume: 500 μl/macaque) on day 0 and 14 d after the first immunization. Sera were collected at −2, 2, 4, 6, and 8 wk after the first immunization, and HA-specific IgG titers were measured with ELISAs. The experimental protocol was approved by the Animal Welfare and Animal Care Committee of the TPRC. All macaques were housed and handled by veterinarians in accordance with the Guidelines for Laboratory Animals of the TPRC.

The statistical significance of differences between groups was determined with the Student t test. The survival curves postinfection were compared with a Kaplan–Meier analysis (log-rank test and Wilcoxon test) using the statistical analysis software SAS for Windows (version 9.2; SAS Institute).

To examine the activity of HP-β-CD as a vaccine adjuvant, we immunized C57BL/6 mice s.c. with LPS-free OVA, with or without different doses of HP-β-CD (3, 10, or 30%). Then the OVA-specific total IgG, IgG1, and IgG2c titers in the sera were determined with ELISAs. After the first immunization (day 10), the anti-OVA total IgG and IgG1 Ab titers were enhanced by all of the doses of HP-β-CD tested (Fig. 1A). In particular, the anti-OVA total IgG and IgG1 titers in the OVA/30% HP-β-CD group were significantly higher than those in the group treated with OVA alone. After boosting, the IgG2c response was also weakly induced, even by 3% HP-β-CD. Thus, all of the Ab titers in the OVA/30% HP-β-CD group were significantly increased on day 31 (21 d after boosting) relative to those in the group treated with OVA alone (Fig. 1A).

The T cell responses also were examined with cytokine ELISAs. Splenocytes from immunized mice were stimulated with OVA Ag or OVA-derived MHC class I– and MHC class II–restricted peptides. After 48 h, the production of IFN-γ, IL-5, and IL-13 was compared with that in splenocytes immunized with K3 CpG-ODN (a typical Th1-type adjuvant) or alum (a typical Th2-type adjuvant). Although HP-β-CD did not enhance IFN-γ production compared with K3 CpG-ODN, OVA/HP-β-CD immunization significantly increased IL-5 and IL-13 production, and the levels of these Th2-type cytokines were similar to those induced with alum (Fig. 1B). These results indicate that HP-β-CD elicits a Th2-type cell–mediated immune response when it is coadministered s.c. with OVA protein.

We then compared the IgE response induced by OVA/HP-β-CD immunization with that induced by alum. Alum is known to induce a high IgE response in animals and is commonly used to generate allergy models in mice (20). Mice were immunized s.c. twice with either OVA/30% HP-β-CD or OVA/alum. We also included K3 CpG-ODN, which is known to suppress Ag-specific IgE (21), as the negative control. Seven days after the booster immunization, the serum anti-OVA IgE concentrations were measured by ELISA. HP-β-CD induced a lower OVA-specific IgE response than did alum, whereas K3 induced virtually no IgE response (Fig. 1C). Interestingly, although HP-β-CD induced typical Th2-type cytokine (IL-5 and IL-13) responses (Fig. 1B), HP-β-CD also significantly increased the IgG2c titer, an indicator of the Th1-type immune response, in mice relative to that induced with OVA alone, although the levels of IgG2c were lower than those induced with K3. Taken together, these results suggest that HP-β-CD is a potent Th2-inducible adjuvant for OVA Ag, but it induces IgE less potently than does alum.

Additionally, we examined the adjuvanticity of HP-β-CD via several routes. When OVA/HP-β-CD was injected via the i.p. route, weaker adjuvanticity was observed compared with the s.c. route. With i.d. injection, Ab titer was elevated as seen with the s.c. route; however, skin inflammation was observed at the local injection site. No adjuvanticity of HP-β-CD was observed when given via i.m. injection. These data suggest that the s.c. route results in adjuvanticity without inducing skin inflammation (data not shown).

To further clarify the effects of HP-β-CD on the immune system, we investigated HP-β-CD–induced cytokine responses in naive mouse splenocytes in vitro. Splenocytes were stimulated with 200–5000 μg/ml (equivalent to 0.02–0.5%) HP-β-CD, and the cytokine concentrations in the supernatant were measured 24 h later using a multiplex assay. There was no significant increase in the cytokines (Supplemental Fig. 1A). Importantly, we could not stimulate the splenocytes with >0.5% HP-β-CD because higher concentrations appeared to cause immediate cell death in vitro. Therefore, we also investigated the cytokine responses in vivo. Mice were injected s.c. with 30% HP-β-CD, and their sera were collected 1, 3, and 8 h after injection. R848 also was injected into a separate group of mice as the positive control for systemic cytokine responses. Consistent with the in vitro result, we did not observe any detectable cytokine response in the sera at any time point examined when 30% HP-β-CD was injected, whereas injection of R848 induced significant responses in many cytokines (Supplemental Fig. 1B). These results indicate that HP-β-CD does not induce detectable levels of systemic cytokines either in vitro or in vivo (Supplemental Fig. 1).

Without inducing proinflammatory responses in vitro and in vivo, HP-β-CD enhanced humoral responses after immunization. This result prompted us to investigate which cells incorporate HP-β-CD and play crucial roles in the adjuvanticity. We first used flow cytometry to investigate the cells in draining inguinal LNs after immunization; however, we could not identify the cells at a dose of 3% (the maximum dose of FITC-labeled HP-β-CD applicable to cells as a result of the HP-β-CD–FITC stock concentration limitation; data not shown). Therefore, we next performed two-photon microscopic analysis, which allows imaging of tissue to a substantial depth. We focused first on Siglec-1⁺ (also called MOMA-1) macrophages, because inactivated influenza virus is captured by Siglec-1⁺ macrophages, which induces humoral immune responses (22). The mice were injected with OVA–Alexa Fluor 647/3% HP-β-CD–FITC, and the explanted inguinal LNs were analyzed 1 h later. HP-β-CD was not incorporated into Siglec-1⁺ macrophages (data not shown), but the majority of HP-β-CD was incorporated into macrophages carrying the macrophage receptor with collagenous structure (MARCO⁺ macrophages) (23), and some HP-β-CD was found in the B cell follicle covering the CD169⁺ subcapsular sinus macrophage area (Fig. 2A, Supplemental Video 1). When the distribution patterns of HP-β-CD and OVA were compared using Volocity’s colocalization analysis, injected OVA was distributed more selectively in MARCO⁺ macrophages than was HP-β-CD (Pearson correlation: HP-β-CD, +0.548, OVA, +0.756) (Fig. 2A, Supplemental Video 1), which is consistent with our recent observation (24). The recent study indicated that, although OVA and CpG (K3-SPG) were incorporated primarily into MARCO⁺ macrophages, the macrophages are dispensable for inducing adaptive immune responses by CpG (K3-SPG). In addition, the study indicated the involvement of dendritic cells (DCs) in the adjuvanticity of CpG (K3-SPG). Based on these data, we next focused on DCs in LNs. To further clarify the uptake of Ag by DCs in LNs, mice were injected s.c. with DQ–OVA or DQ–OVA/30% HP-β-CD, and the uptake of OVA Ag by CD11c⁺ cells was analyzed 24 h later using flow cytometry. Although the overall uptake was very small, OVA uptake by CD11c⁺ DCs was enhanced significantly by HP-β-CD (Fig. 2B), suggesting that HP-β-CD acts as an Ag-delivery system to CD11c⁺ cells. We also examined DC maturation after HP-β-CD injection in vivo but did not detect any upregulation of CD40 in the DCs (data not shown). It was difficult to investigate the function of MARCO⁺ macrophages by flow cytometry (25). Collectively, we found a novel immunological mechanism of HP-β-CD that enhances Ag uptake by DCs.

To further clarify the mode of action of HP-β-CD, we performed a comprehensive mRNA GeneChip analysis. In our preliminary experiments, we noted that the adjuvant activity of HP-β-CD was weak after i.p. injection compared with that after s.c. or i.d. injection. Therefore, we compared gene expression in the draining inguinal LNs and other organs, such as the spleen and liver, 6 h after i.d. injection of HP-β-CD into the tail base (local) or i.p. injection (systemic). After local i.d. injection of HP-β-CD, the expression of substantial numbers of genes was induced >2-fold in the draining LNs, spleen, and liver (Fig. 3A). In contrast, i.p. injection of the same amount of HP-β-CD induced only limited changes in gene expression in the spleen and liver (Fig. 3A). We also performed a cluster analysis and categorized the induced genes into seven major groups (Supplemental Fig. 2A). Each group of genes was examined using Gene Ontology (GO) term analysis (Supplemental Fig. 2B). This analysis showed that HP-β-CD induced the expression of genes related to the innate immune response and the inflammatory response in the LNs, spleen, and liver after i.d. injection but not after i.p. injection (Fig. 3B, Supplemental Fig. 2, group 7), consistent with its route-dependent adjuvant activity. The GO term “defense response to virus” also was called in the group 1 cluster (Supplemental Fig. 2, group 1); correspondingly, the IFN response genes also were induced in the LNs after i.d. injection of HP-β-CD (Fig. 3C). The analysis of individual genes in Supplemental Fig. 2 revealed that those associated with the Th2-type immune response, including Ptgs2 (26), Nfil3 (27), and Il33 (28), also were upregulated after i.d. injection of HP-β-CD (Fig. 3D). The induction of Il1β and Myd88 genes (Fig. 3D) also suggested that inflammasomes could be activated in vivo by HP-β-CD. These data suggest that, despite the lack of detectable levels of cytokine production in the serum (Supplemental Fig. 1B), our GeneChip analysis demonstrated that HP-β-CD induced the expression of inflammatory and immune-related genes, including type I IFNs and IL1β, after i.d. injection but not after i.p. injection. These data are consistent with the route-dependent adjuvant activity of HP-β-CD.

Our GeneChip analysis found the induction of immune-related gene expression in LNs after HP-β-CD injection. A recent study indicated that Tfh cells, a Th cell subset present in germinal centers (GCs) in LNs, facilitate the generation of long-lived plasma cells and memory B cells, which are required for long-term Ab responses (29). To investigate the involvement of Tfh cells in the adjuvanticity of HP-β-CD, we evaluated their generation in LNs from Bcl6^f/+ or Bcl6^f/f mice immunized with OVA or OVA/30% HP-β-CD. After the second immunization, the generation of Tfh cells was significantly increased by HP-β-CD in Bcl6^f/+ mice (p < 0.05), and it was significantly decreased in Bcl6^f/f mice (p < 0.0001) (Fig. 4B). We also investigated the humoral responses between Bcl6^f/+ and Bcl6^f/f mice after OVA/30% HP-β-CD immunization. Consistent with the result of flow cytometric analysis, the anti-OVA IgG1 titer of Bcl6^f/f mice was significantly lower than that of Bcl6^f/+ mice after the second immunization (p < 0.01) (Fig. 4C). However, the IgG1 titer of Bcl6^f/f mice immunized with OVA/30% HP-β-CD was not decreased to the level of that of Bcl6^f/f mice immunized with OVA alone, suggesting that the GC-independent pathway contributes, in part, to the induction of humoral responses after OVA/HP-β-CD immunization (30). These results suggested that the generation of Tfh cells is facilitated by the administration of HP-β-CD and that Tfh cells are required for the high adjuvanticity of HP-β-CD.

IL-21 secreted by Tfh cells can act in an autocrine manner to maintain Tfh cells at various stages of differentiation (29). Meanwhile, IL-21 was reported to prevent Ag-specific IgE response in mice (31). In this study, although HP-β-CD induced Th2 responses, it did not strongly induce an IgE response after immunization (Fig. 1C). To explain the weak IgE response after OVA/HP-β-CD immunization, we investigated IL-21 responses after stimulation of splenocytes from immunized Bcl6^f/+ mice; IL-21 responses were not observed from stimulated splenocytes (data not shown). Further research is necessary to explain why HP-β-CD does not induce strong IgE responses after immunization.

Our GeneChip analysis suggested that inflammasomes can be activated by HP-β-CD in vivo (Fig. 3B). Th2-inducible adjuvants, such as alum and silica, also were shown to activate inflammasomes in macrophages, whereas the role of inflammasomes in their adjuvant activity has been questioned (5). To clarify the immunological mechanisms and pathways essential for the adjuvant activity of HP-β-CD, we examined whether it activates inflammasomes in macrophages in vitro. Peritoneal macrophages were primed with LPS and stimulated with HP-β-CD 18 h later. Compared with alum, which induced high amounts of IL-1β, 5000 μg/ml (0.5%) HP-β-CD did not enhance the production of IL-1β by macrophages (data not shown). Because HP-β-CD is cytotoxic (see Fig. 5C), we could not examine the activation of inflammasomes at a higher concentration of HP-β-CD (equivalent to the maximum dose used in the immunization experiments performed in vivo) (Fig. 1). Therefore, we examined the humoral responses in Asc^−/− mice and Caspase 1^−/− mice, which are known to be deficient in inflammasome activation (32). The humoral response after s.c. immunization with OVA/HP-β-CD was not reduced in either gene-KO mouse strain (data not shown). Although these data do not exclude the possibility that the inflammasomes were activated in vivo after HP-β-CD injection, inflammasome activation was not required for the adjuvant activity of HP-β-CD.

To identify the pathway essential for the adjuvant activity of HP-β-CD, we focused on MyD88, a critical adapter protein of TLRs and the IL-1R superfamily, which are involved in the downstream activation of NF-κB and MAPK (33). TLRs are also key players in the activation of innate immunity and recognize various adjuvants and pathogen-associated molecular patterns. Mice lacking Myd88 were immunized s.c. twice with OVA/HP-β-CD. After the second immunization, the anti-OVA IgG and IgG1 titers of Myd88^−/− mice were significantly lower than those of Myd88^+/− mice (Fig. 5A). Interestingly, Myd88^−/− mice immunized with OVA/HP-β-CD showed significantly higher total IgG and IgG1 titers than did WT mice immunized with OVA alone. These results suggest that both MyD88-dependent and -independent signaling pathways are involved in the adjuvant activity of HP-β-CD. A TLR-screening assay was performed in vitro to identify the upstream receptor of MyD88-dependent signaling. The TLR4-signaling pathway was activated slightly by HP-β-CD (Supplemental Fig. 3A), so we examined whether TLR4 is involved in the adjuvant activity of HP-β-CD in vivo; however, the humoral response was not reduced in Tlr4^−/− mice (Supplemental Fig. 3B). We then investigated other TLRs or members of the IL-1R superfamily by s.c. immunizing other Tlrs, Il1R, Il18R, and St-2 KO mice. ST-2 is a receptor that forms a complex with IL-1R accessory protein; it can bind IL-33 directly, and it mediates Th2 responses (28). All of the examined KO mice, including TLR7/9 double-deficient and TLR2/4/9 triple-deficient mice, showed anti-OVA total IgG titers that were similar to those of WT mice 7 d after the booster immunization (Supplemental Fig. 3B). Notably, after stimulation with HP-β-CD, no significant increases in proinflammatory cytokines, chemokines, or growth factors were detected in vitro (Supplemental Fig. 1A). Taken together, these results suggest that HP-β-CD itself may activate TLR4 in vitro. However, in vivo, the adjuvant activity of HP-β-CD is independent of any single TLR, TLR7/9, TLR2/4/9, or the IL-1R superfamily, although the adjuvant activity of HP-β-CD is partially dependent on MyD88 signaling.

The results described above suggest that a MyD88-independent pathway is involved in the adjuvant activity of HP-β-CD. Our GeneChip analysis suggested that HP-β-CD induces type I IFN–related gene responses in the draining LNs (Fig. 3). Therefore, we examined the involvement of IFN-αβ receptor (IFNAR), IRF3/7, and TBK1 by using mice in which the representative gene is knocked out. Tbk1^−/− mice on a TNF-deficient (Tnf^−/−) background were used because Tbk1^−/− mice die in utero, and this lethal effect can be rescued in the absence of TNF (34). The humoral responses in IFNAR2 and IRF3/7 double-KO mice were similar to those in WT mice (data not shown). However, the anti-OVA IgG titer in Tbk1^−/− mice was significantly lower than that in Tnf^−/− mice (Fig. 5B), suggesting that TBK1 also contributes to the adjuvant activity of HP-β-CD but that its activity is independent of type I IFN responses.

We showed that alum induces the release of dsDNA from host cells, resulting in TBK1/IRF3-dependent, but IFNAR2-independent, Th2 responses (16). The results described above indicating TBK1 involvement prompted us to examine the release of dsDNA from host cells after the local injection of HP-β-CD. We first examined the cytotoxicity of HP-β-CD against HEK293T cells in vitro. HP-β-CD started killing HEK293 cells at a concentration of 0.5%; at 1.5%, almost all of the cells were killed. However, the overall cytotoxicity of HP-β-CD was 1,000,000 times lower than that of mitomycin C, which damages cells by DNA cross-linking (35) (Fig. 5C). In an in vivo study, we observed the adjuvant activity of 3% HP-β-CD in mice (Fig. 1A). These data suggest that HP-β-CD potentially causes the release of dsDNA from host cells in vivo, similar to alum. To examine this possibility, we tested whether HP-β-CD induced the release of host DNA in vivo. HP-β-CD was injected s.c. into mice, and the concentration of host DNA at the injection site, which formed a liquid pouch under the skin, was measured. Increasing amounts of dsDNA, triggered by HP-β-CD, were detected until 6 h, after which they decreased and returned to background levels at 24 h after injection (Fig. 5D). This result suggests that HP-β-CD triggers temporary host dsDNA release at its injection site. Furthermore, DNase I treatment at the time of immunization significantly reduced the total IgG, IgG1, and IgE titers but enhanced the IgG2c response (Fig. 5E). This result strongly suggests that HP-β-CD triggers dsDNA release, which acts as a DAMP, triggering a signal through the TBK1-dependent pathway and resulting in an enhanced Th2-type immune response. However, at the same time, HP-β-CD induces a MyD88-dependent response, which may modulate the IgE response upregulated by the dsDNA/TBK1 axis. Further study is required to identify the critical upstream/downstream pathways of MyD88 and TBK1 induced by the local administration of HP-β-CD.

Although HP-β-CD temporarily triggered the release of dsDNA (Fig. 5D), it did not elicit strong IgE responses (Fig. 1C), and it induced almost no systemic cytokine responses in mice (Supplemental Fig. 1B). This suggests that HP-β-CD is a suitable adjuvant for a current split influenza vaccine. Most influenza vaccines are used to immunize healthy individuals, so minimizing the toxicity of the vaccine is an absolute requirement for the adjuvant. We evaluated the adjuvant activity of HP-β-CD using an influenza SV immunization model in mice and macaques. Mice were immunized s.c. twice with SV only or with SV/3–30% HP-β-CD, and their total IgG titers were measured with an ELISA. After the booster immunization, the anti-HA total IgG titer was significantly enhanced by all doses of HP-β-CD tested (Fig. 6A). After the immunized mice were challenged with a lethal dose of A/Osaka/129/2009 [A(H1N1)pdm09] influenza virus, their body weights and survival rates were determined for 14 d. The survival and body weight parameters of the SV/HP-β-CD–immunized groups were significantly better than were those of the group immunized with SV only (Fig. 6B, 6C).

To further estimate the adjuvant activity of HP-β-CD in humans, cynomolgus macaques were immunized s.c. twice with saline, SV, SV/3% HP-β-CD, or SV/30% HP-β-CD, and their anti-HA IgG titers were evaluated every 2 wk. The anti-HA IgG titers of all macaques in the SV/30% HP-β-CD group were increased 2 wk after the first immunization (Fig. 7A). A boosting effect of HP-β-CD also was observed in this group (Fig. 7A). The anti-HA IgG titers of the SV group varied greatly among individuals (Fig. 7A). When we compared the geometric mean titer of the anti-HA IgG titers of the SV/30% HP-β-CD group and the SV group, the anti-HA IgG titer of the SV/30% HP-β-CD group was 3-fold higher than that of the SV group 2 wk after the first immunization, and it was 2-fold higher than that of the SV group 6 wk after the second immunization (Fig. 7B). Unlike the results for mice, the anti-HA IgG titers in the SV/3% HP-β-CD group were not better than those in the SV group (Fig. 7). These results demonstrate that 30% HP-β-CD displays adjuvant activity with the influenza SV vaccine in both mice and cynomolgus macaques.

In addition to their efficacy, one of the most important characteristics of vaccine adjuvants is their safety, because many vaccines are intended for use in healthy individuals. HP-β-CD has been used safely as an excipient for pharmaceutical agents for decades (9). However, no extensive immunological characterization of cyclodextrins as adjuvants has been conducted, so the molecular and cellular mechanisms of their actions are largely unknown. Our characterization of HP-β-CD as a vaccine adjuvant revealed that it induces unique Th2 responses, enhancing Ag-specific IgG titers, including IgG1 and IgG2c titers, by s.c. injection. Furthermore, the adjuvanticity is largely dependent on Tfh cells. Unlike alum, a commonly used adjuvant for many human vaccines throughout the world, HP-β-CD induced little IgE production (Fig. 1A, 1C). A T cell analysis clearly indicated that HP-β-CD preferentially activates Th2 cells over Th1 cells (Fig. 1B). Traditionally, it has been considered that the Th2 and IgE responses are sequential, but our current data indicate that Th2 induction does not always induce strong IgE production. The ability of an adjuvant to induce IgE may be a crucial risk factor affecting the allergenic potential of vaccines. Therefore, HP-β-CD may be a safer adjuvant than alum because it entails less risk for inducing potentially allergic IgE responses. Injection of HP-β-CD obviously induced some acute inflammatory, type I IFN–related and Th2-associated gene responses in LNs. However, no systemic proinflammatory cytokine responses were observed after its injection into mice (Supplemental Fig. 1B), suggesting that the gene expression induced in LNs does not induce systemic proinflammatory cytokine responses, which can be detrimental to the host. Without inducing systemic proinflammatory cytokine responses, HP-β-CD exhibited enough adjuvanticity to improve the immunogenicity and protective efficacy of SV in mice and cynomolgus macaques. Hence, we conclude that HP-β-CD is a well-balanced adjuvant possessing high adjuvant activity and little risk for inducing allergic potential IgE responses and systemic inflammation.

The activation of immune responses in LNs is a major event to establish adaptive immunity (29). Our cellular analysis found enhanced Ag uptake by DCs and generation of Tfh cells in LNs after OVA/HP-β-CD injection (Figs. 2B, 4B). In addition, Tfh cells were required for the high adjuvanticity of HP-β-CD (Fig. 4C). DCs provide signals that upregulate CXCR5 and downregulate CCR7 on naive CD4⁺ T cells, allowing them to migrate to B cell follicles and, consequently, facilitate the generation of Tfh cells (29). Tfh cells are recognized as the key player required for the formation of GCs and the generation of long-lived serological memory (30). Therefore, these results explained why humoral responses were enhanced and maintained by HP-β-CD.

The biological effects of HP-β-CD are reported to include the binding and sequestration of plasma membrane cholesterol and, consequently, the dispersal of lipid rafts, which include lipids and associated plasma membrane–spanning proteins, such as sphingolipids (36). Such treatments are considered to affect the lateral diffusion, aggregation, and function of receptor complexes (37). Bioactive lipids, such as ceramides and sphingosine 1-phosphate, are known to regulate inflammation in cells and to activate several signaling pathways (38). Therefore, it is reasonable to infer that HP-β-CD activates the innate immune response and exerts its adjuvant activity by modifying these lipid-dependent biological processes. Methyl-β-cyclodextrin is reported to disperse lipid rafts, thereby stimulating MyD88-dependent NF-κB activation in immature B cells, and it partially stimulates the TLR4-signaling pathway in vitro (39). In this study, the results of an in vitro TLR-screening assay supported those findings (Supplemental Fig. 3A). However, analysis of gene-KO mice indicated that MyD88, but not TLR4, contributes, in part, to the adjuvant activity of HP-β-CD (Supplemental Fig. 3B), suggesting that signaling pathways other than TLR pathways, such as lipid raft dispersal/MyD88/NF-κB activation, contribute to the MyD88-dependent mechanism underlying the adjuvant activity of HP-β-CD (Supplemental Fig. 4). This is an important issue for future research.

We also demonstrated that TBK1 contributes, in part, to the adjuvant activity of HP-β-CD (Fig. 5B). HP-β-CD causes the release of several factors, including lipids and dsDNA, that may function as DAMPs to activate the innate immune response. In this study, HP-β-CD induced the temporary release of dsDNA at its injection site. These facts strongly suggest that the adjuvant activity of HP-β-CD is also mediated by host-derived factors, such as DAMPs. Therefore, it is reasonable that the adjuvant activity of HP-β-CD was not diminished in single Myd88- or Tbk1-deficient mice. Our findings indicate that the activation of several signaling pathways as a result of HP-β-CD injection contributes to its adjuvant activity (Supplemental Fig. 4).

It is important to exploit the safety profile of HP-β-CD for the development of adjuvanted vaccines. Vaccination is the primary strategy used to prevent influenza infection. The efficacy of influenza vaccines in young and healthy adults is estimated to be 70–90%, but it is <17–53% in the elderly (40). The use of an adjuvant to improve the efficacy of a vaccine has been investigated, but minimizing adjuvant toxicity remains one of the major challenges in adjuvant research (41). To exploit the safety profile of HP-β-CD, we evaluated the combined effects of HP-β-CD and SV and demonstrated that HP-β-CD enhanced the protective efficacy of SV against influenza virus infection in mice (Fig. 6). Its adjuvant activity also was examined in cynomolgus macaques (Fig. 7). In the macaque model, the anti-HA IgG titers were not strongly elevated in any of macaques in the SV group. In contrast, immunization with SV/30% HP-β-CD strongly enhanced the anti-HA IgG titers of all of the macaques in the group. To improve the efficacy of SV, it is important to overcome the poor immunogenicity and enhance the humoral responses of weak responders with an adjuvant. In contrast to the results with s.c. injection, we did not find adjuvanticity for HP-β-CD via the i.m. route. The i.m. route is a major route for human vaccines; however, all seasonal influenza SVs are injected s.c. in Japan. Outside of Japan, some influenza vaccines, such as Fluvax (42) and VAXIGRIP (43), are acceptable to inject s.c. Therefore, the lack of adjuvanticity of HP-β-CD via i.m. injection is not a major concern. Collectively, our results suggest that HP-β-CD is a promising adjuvant for enhancing the immunogenicity of s.c. SV injections in humans.

This study has important implications, demonstrating that HP-β-CD is safe and can be used effectively as an adjuvant for SV in humans. It also contributes novel immunological findings that advance cyclodextrin research and promote the innovation of a rationally designed safer adjuvant.

We thank Tetsuo Kase for providing the A/Osaka/129/2009 virus. We also thank Akiko Okabe, Mariko Nakamura, Yuko Fujita, Aki Konishi, Kousaku Murase, and Yasunari Haseda for excellent technical assistance with animal husbandry and genotyping, as well as the members of K.J.I.’s and C.C.’s laboratories for valuable comments and help.

The authors have no financial conflicts of interest.