Research ArticleADJUVANTS

Rational design of small molecules as vaccine adjuvants

See allHide authors and affiliations

Science Translational Medicine  19 Nov 2014:
Vol. 6, Issue 263, pp. 263ra160
DOI: 10.1126/scitranslmed.3009980


Adjuvants increase vaccine potency largely by activating innate immunity and promoting inflammation. Limiting the side effects of this inflammation is a major hurdle for adjuvant use in vaccines for humans. It has been difficult to improve on adjuvant safety because of a poor understanding of adjuvant mechanism and the empirical nature of adjuvant discovery and development historically. We describe new principles for the rational optimization of small-molecule immune potentiators (SMIPs) targeting Toll-like receptor 7 as adjuvants with a predicted increase in their therapeutic indices. Unlike traditional drugs, SMIP-based adjuvants need to have limited bioavailability and remain localized for optimal efficacy. These features also lead to temporally and spatially restricted inflammation that should decrease side effects. Through medicinal and formulation chemistry and extensive immunopharmacology, we show that in vivo potency can be increased with little to no systemic exposure, localized innate immune activation and short in vivo residence times of SMIP-based adjuvants. This work provides a systematic and generalizable approach to engineering small molecules for use as vaccine adjuvants.


For decades, the strategies of drug discovery have been successful in delivering new therapies for human diseases. Today, this process is highly rational and involves the systematic optimization of low molecular weight (LMW) compounds by chemical analoging. For most drugs, conventions such as “Lipinski’s rule of 5” guide chemistry efforts toward molecules predicted to have high bioavailability and biodistribution (1). By contrast, vaccine adjuvant discovery remains empirical, and adjuvants are defined as almost anything added to a vaccine formulation that increases immunogenicity in vivo (2). Thus, there are a multitude of adjuvants that enhance vaccine potency in animal models, but how they can be optimized to be safe and well tolerated for human use remains unclear. A new class of adjuvants activates immune cells via Toll-like receptors (TLRs), and LMW compounds targeting TLR2, TLR7, and TLR8 have been identified (35). We call these LMW TLR agonists small-molecule immune potentiators (SMIPs) collectively. Imiquimod (R837) and resiquimod (R848) are clinically advanced TLR7 and/or TLR8 SMIPs used as immunotherapeutics (for example, 5% imiquimod cream, Aldara, is licensed as a topical treatment for viral and premalignant/malignant skin lesions) (6). However, these imidazoquinolines induce strong local and systemic inflammatory reactions and are poorly tolerated (711). It is unclear whether this level of inflammation is required for their efficacy.

Although R848 and R837 have been studied extensively as vaccine adjuvants in preclinical models, they were never optimized for this use and, in general, compare poorly to other preclinical and clinical adjuvant candidates (1214). Thus, their less than optimal adjuvant activity suggests that LMW compounds are ill-suited as vaccine adjuvants. Here, medicinal chemistry was used to identify and then optimize a class of TLR7 SMIPs to achieve the following goals: (i) determine the viability of drug-like molecules for vaccine adjuvant development and (ii) establish the minimal requirements for adjuvant activity to uncouple their benefits on vaccine potency from their inflammatory side effects. Other groups have studied imidazoquinolines and synthetic RNA oligonucleotides as adjuvants and have improved on their in vivo potency through direct conjugation to antigen or formulation (1517). Here, we describe a stepwise approach to the chemical analoging of TLR7-dependent SMIPs. The aim was to create a toolbox of adjuvant-active compounds for comparative in vivo testing so as to define more precisely the properties of drug-like molecules required for optimal adjuvant activity. The ultimate goal of this effort was the clinical advancement of improved adjuvants for prophylactic vaccines for which the target populations are healthy individuals, often infants. Therefore, safety, tolerability, and ease of manufacture were key drivers behind the design and optimization of compounds and formulations.

Unlike direct-acting drugs, TLR7 SMIPs activate innate immunity and initiate a cascade of immune responses that can have systemic impact and endure after the compound has been cleared from the organism. For SMIPs as vaccine adjuvants, this difference is highlighted further by the fact that two to three local injections (for example, intramuscular) can drive antigen-specific B and T cell responses at distal sites and these can provide durable protection in the form of immunologic memory. Limiting the biodistribution of the SMIP adjuvants was postulated as an approach to both increase vaccine efficacy and minimize the side effects associated with systemic and generalized inflammation. To test whether SMIPs targeting TLR7 could be optimized for adjuvant use, chemical analogs were designed to be less soluble than traditional LMW drugs to facilitate retention at the injection site. These lipophilic compounds were far more potent in vivo than the hydrophilic R848 benchmark as adjuvants and exhibited minimal systemic exposure and no systemic immune activation. Despite these highly favorable features, the first-generation SMIPs suffered from two predicted liabilities: poor solubility and long in vivo residence times (>2 weeks) at the injection site. These deficiencies were eliminated by designing soluble SMIPs with linkered phosphonates that could be adsorbed to aluminum hydroxide [Al(OH)3] via ligand exchange (fig. S1).


First generation SMIP-based adjuvants

The drug discovery strategies of in vitro screening, rational design, and chemical analoging were used to identify and optimize a class of TLR7-dependent SMIPs, benzonaphthyridines (BZNs) (1820). Through this process, hundreds of potent and TLR7-selective analogs were identified, and compounds with distinct physicochemical properties were compared in vivo for their biodistribution, impact on immune activation, and potency as adjuvants (Fig. 1, tables S1 and S3, and fig. S2, A and B). Many analogs were designed to be less soluble (for example, more lipophilic) than R848 and traditional LMW drugs. We hypothesized that lipophilic compounds would localize at the injection site, leading to minimal systemic exposure and limited generalized inflammation. The key question was, are these nontraditional immunodrugs effective adjuvants?

Fig. 1. SMIP PK/PD.

(A) First-generation SMIP progression flowchart. (B) In vitro TLR7 activity of SMIPs, represented by IL-6 induction in mouse splenocytes. Additional information can be found in the Supplementary Materials. (C to E) Mice (3 per group) were injected intramuscularly (quadriceps) with 100 μg of compounds. (C) Compound concentrations in serum were measured at the indicated times after injection. (D) Compound concentrations in the injected muscles were measured 24 hours after injection. (E) IL-6 levels in serum were measured by Meso scale discovery (MSD) enzyme-linked immunosorbent assay (ELISA). (F) RNA extracted from the draining lymph nodes (top) and muscle (bottom) of mice (3 per group) 24 hours after injection of 25-μg compound was analyzed by whole-genome microarray (38).

Pharmacokinetic and pharmacodynamic analyses

Two BZN compounds, SMIP.7-7 and SMIP.7-8, were selected for extensive in vivo analysis, and the imidazoquinoline R848 was used as a comparator. SMIP.7-7 and SMIP.7-8 differ in their calculated polar surface area and potency as TLR7 agonists (table S1). Both compounds are less potent in vitro (Fig. 1B and fig. S2, A and B) and more lipophilic than R848, which is a pure TLR7 agonist in mice and a dual TLR7/8 agonist in humans (21). We assessed the pharmacokinetic (PK) properties of each compound by determining their in vivo biodistribution after intramuscular injection of mice (Fig. 1, C and D). Injection of R848 gave high systemic exposure, but was undetectable in the muscle at 24 hours because of its high solubility and rapid absorption into the bloodstream. SMIP.7-8 was also detected in the systemic circulation, albeit at a lower level, and showed some muscle retention. The least polar compound, SMIP.7-7, was barely detectable systemically, whereas retention in the muscle was high. Thus, the three compounds exhibited distinct PK profiles consistent with their physicochemical properties.

Serum interleukin-6 (IL-6) concentration was used as a pharmacodynamics (PD) marker to monitor the early TLR7-dependent activation of systemic inflammation (Fig. 1E). R848 and, to a lesser extent, SMIP.7-8 led to transient increases in IL-6 production systemically, whereas no IL-6 could be detected after SMIP.7-7 injection. Gene expression profiling of the draining lymph node showed substantial immune activation by R848, especially type I interferon (IFN)–responsive gene signatures, which were not observed for SMIP.7-7 or SMIP.7-8 (Fig. 1F). The differential immune activation by R848 versus the other two compounds was consistent for other PD markers in the spleens, muscles, and lymph nodes of treated mice (fig. S2C). Overall, a compound’s PD correlated well with its PK profile and physicochemical properties. These correlations were extended to other BZN SMIPs (fig. S3, A and B).

Identifying the most effective SMIP-based adjuvant

The utility of the three different TLR7 agonists as vaccine adjuvants was first determined by intramuscular immunization of BALB/c mice with ovalbumin (OVA) and measurement of the OVA-specific immunoglobulin G (IgG) response. Although less potent TLR7 agonists in vitro (Fig. 1B and fig. S2, A and B), SMIP.7-7 and SMIP.7-8 significantly outperformed R848 in boosting an anti-OVA IgG response in vivo (Fig. 2A). Only SMIP.7-7 enhanced isotype switching, as shown by a marked increase of antigen-specific IgG2a, indicating a potent CD4 T helper cell 1 (TH1) induction (22).

Fig. 2. SMIP adjuvanticity.

(A) Mice (8 per group) were immunized twice (3 weeks apart) with 10 μg of OVA simply mixed with 25 μg of compounds or vehicle. Shown are OVA-specific IgG geometric mean titers (GMTs), IgG2a isotype GMT, and IgG1 isotype GMT. (B) Mice (8 per group in two separate experiments) were immunized twice (3 weeks apart) with three recombinant MenB antigens (5 μg each) formulated with 25 μg of compounds in Al(OH)3 or MF59. SBA titers against the NZ98 strain were determined from two independent experiments. Results are from individual mice, and means ± SD are shown. (A and B) Statistical significance was determined using a paired t test.

Because anti-OVA responses are not ideal indicators of adjuvant utility in humans, a clinically relevant vaccine was examined. Neisseria meningitidis serotype B (MenB) is a deadly bacterium that is a major health threat worldwide (23). A vaccine countering MenB infections, composed of three recombinant MenB proteins (3MenB) and one bacterial outer membrane vesicle (OMV) adsorbed to aluminum hydroxide [Al(OH)3], has recently been approved by the European Medicine Agency (24). We added R848, SMIP.7-7, and SMIP.7-8 to the 3MenB antigens (without the OMV component) formulated with adjuvants used in licensed vaccines, Al(OH)3 or MF59 (25) (Fig. 2B). Each SMIP was also tested with the 3MenB antigens alone (fig. S2D). Mice were immunized intramuscularly, and MenB serum bactericidal antibody (SBA) titers were determined. SBA titers are accepted as a surrogate marker of clinical efficacy by regulatory agencies globally, and an average neutralizing titer ≥1024 in mice is considered a preclinical correlate of protection for humans (2628). The MenB “New Zealand” (NZ98) strain was selected to assess the impact of the different TLR7 vaccine formulations because recombinant protein vaccination in the absence of OMV results in poor SBA titers against this strain. Whereas all of the TLR7 agonists augmented SBA titers, SMIP.7-7 significantly outperformed R848 (Fig. 2B). SMIP.7-7 was also better at increasing SBA titers above the 1024 cut-off, achieving >80% response rates in both formulations. Similar to the OVA immunization studies, there was a hierarchy of SMIP adjuvant potency that correlated with lower systemic exposure but was not reflected by TLR7 potency in vitro (Fig. 1B and fig. S2, A and B). These results highlight how innate immune stimulation and proinflammatory activities of SMIPs can be highly focused to achieve greater efficacy with less risk of immune toxicities. Clearly, the systemic and polyclonal immune activation, which is a particular feature of R848, results in “wasted inflammation” that is, at best, unnecessary for adjuvant activity and potentially detrimental to it.

Rational design of more stable, scalable, and clinically viable SMIP adjuvant formulations

Despite the favorable PK, PD, and efficacy profiles of SMIP.7-7 and similar compounds [for example, SMIP.7-2 (fig. S3 and table S2)], there were key liabilities identified for highly lipophilic compounds that were predicted to hamper clinical advancement. Specifically, lipophilic SMIPs create formulation and scale-up issues because of their poor solubility in aqueous solutions at physiologic pH (that is, solutions, buffers, and formulations compatible with human use). Moreover, extended analysis of the muscle PK of lipophilic compounds showed retention at the injection site for >2 weeks. We hypothesized that this extended residence time was not minimally required for adjuvanticity and therefore represented an unnecessary feature of the first-generation compounds. The dilemma was that poor solubility in aqueous solution led to localized immune activation and potent adjuvant activity, but poor solubility created other deficiencies for the adjuvant candidates. To address this problem, a combination of medicinal and formulation chemistry was used to engineer soluble SMIPs that could be formulated with aluminum salts [specifically, aluminum hydroxide or Al(OH)3] with the aim of limiting their systemic exposure and shortening their half-lives in the muscle.

Decreased systemic exposure, generalized inflammation and in vivo muscle retention time of SMIPs adsorbed to AI(OH)3

The strongest adsorption to aluminum adjuvants is obtained through ligand exchange of hydroxyl and/or phosphate groups on the surface of aluminum hydroxide [Al(OH)3] or phosphate (AlPO4) (29). To drive the adsorption of SMIPs by the same mechanism, TLR7 agonists were functionalized with polyethylene glycol (PEG) linkers and terminal phosphonate groups (Fig. 3A). As with the first-generation compounds, candidate SMIPs were optimized for TLR7 potency and selectivity in vitro (table S2). However, in this case, solubility (>1 mg/ml) at neutral pH and stable adsorption to Al(OH)3 were key criteria for advancement to in vivo testing. PEG linker chemistry allowed for the design of SMIPs with greatly improved solubility profiles at neutral pH (Fig. 3B), and the anionic phosphonate functional groups facilitated stable and efficient adsorption to Al(OH)3 (fig. S4) (18). Exploiting the autofluorescence of BZN SMIPs, adsorption to Al(OH)3 was visualized using fluorescence microscopy and flow cytometry (Fig. 3C).

Fig. 3. Phosphonate SMIP.

(A) Second-generation SMIP progression flowchart. (B) Stepwise improvement of phosphonate SMIP solubility to achieve efficient Al(OH)3 adsorption. (C) Visualization of phosphonate SMIP adsorption to Al(OH)3 by microscopy and flow cytometry (histogram inset). Experiments were done using SMIP.7-11, which has intrinsic fluorescent property. Scale bars, 1 μm. (D and E) Mice (3 per group) were injected intramuscularly (quadriceps) with 100 μg of SMIPs alone or adsorbed to Al(OH)3. (D) Compound concentrations in serum were measured at the indicated times after injection. (E) Compound concentrations in the injected muscle were measured at the indicated times after injection.

The phosphonate SMIPs were engineered to be more soluble, which leads to increased systemic exposure and generalized inflammation. The goal was to use insoluble Al(OH)3 formulations to limit systemic exposure while simultaneously controlling muscle retention time. As predicted, unformulated phosphonate SMIPs exhibited high levels of systemic exposure when injected in mice intramuscularly, whereas the serum concentrations of the same compounds adsorbed to Al(OH)3 were markedly reduced (Fig. 3D). For the first few hours, the localization of Al(OH)3-adsorbed phosphonate SMIPs to the injection site mirrored that of their first-generation insoluble counterparts (for example, SMIP.7-7 and a related compound, SMIP.7-2, shown here and in fig. S3); however, their clearance from the muscle was more rapid with >95% of the compound eliminated within the first 24 to 48 hours (Fig. 3E). Thus, Al(OH)3 adsorption substantially altered the temporal and spatial distribution of SMIPs in vivo, demonstrating that SMIPs with physicochemical properties better suited for scalable and reproducible formulation can also elicit improved PK properties for adjuvant use.

The impact of unformulated and Al(OH)3-adsorbed SMIPs on local and systemic immune activation was consistent with their biodistribution (Fig. 4). Al(OH)3 adsorption of the soluble compounds reduced their induction of serum cytokines (Fig. 4A) and polyclonal activation of T and B cells in the draining lymph nodes (Fig. 4B) while focusing the induction of cytokine gene signatures to the muscle (Fig. 4C). The impact of Al(OH)3 adsorption of phosphonate SMIPs (for example, SMIP.7-10) on PD mirrors that seen for insoluble SMIPs (for example, SMIP.7-2); however, second-generation SMIPs adsorbed to Al(OH)3 were more effective at inducing MenB SBA titers compared to their insoluble counterparts (Fig. 5C). Hence, the iterative design approach described here yielded adjuvant formulations that elicit highly localized and short-lived inflammatory responses while simultaneously improving efficacy.

Fig. 4. Phosphonate SMIP PD.

(A and B) Mice (3 per group) were injected intramuscularly (quadriceps) with 100 μg of SMIP.7-10 alone or adsorbed to MenB/Al(OH)3. (A) Serum cytokine levels were measured by MSD ELISA at the indicated time points after injection. Statistical significance was determined using a paired t test (n = 3) for the difference in systemic cytokine levels at 3 hours after injection for Al(OH)3 formulated and unformulated SMIP.7-10 (****P = 0.007, IL-6; *** P = 0.0004, MCP-1; **P = 0.009, mKC; *P = 0.03, TNF-α). (B) Expression of CD69 marker on B cells (left) and T cells (right) isolated from the draining lymph nodes at 6 and 24 hours after injection was determined by flow cytometry. MenB/Al(OH)3 injection (no SMIP) served as the baseline and R848 without Al(OH)3 as positive control. Statistical significance was determined using a paired t test (n = 3) for the difference in CD69 expression on B and T cells at 6 hours after infection for Al(OH)3 formulated and unformulated SMIP.7-10 and SMIP.7-12 (****P = 0.01; *** P = 0.0008; **P = 0.0001; *P = 0.003). (C) RNA extracted from the injected muscles and draining lymph nodes 6 hours after injection of 25-μg compound was analyzed by whole-genome microarray (38).

Fig. 5. Phosphonate SMIP adjuvanticity.

(A and B) Mice (10 per group) were immunized twice (2 weeks apart) with three recombinant MenB antigens formulated with 100 μg of (A) BZN SMIPs and (B) other SMIP chemotypes, either alone or adsorbed to Al(OH)3. The data represent mean SBA titers (±SD) of three replicates of pooled serum, and statistical significance was determined using a paired t test. (A) (n = 3) P = Al(OH)3 versus Al(OH)3/7.10: 0.00006; Al(OH)3 versus Al(OH)3/7.11: 0.00001; Al(OH)3 versus Al(OH)3/7.12: 0.0003; 7.10 versus Al(OH)3/7.10: 0.00004; 7.11 versus Al(OH)3/7.11: 0.000009; 7.12 versus Al(OH)3/7.12: 0.0002. (B) (n = 3) P = Al(OH)3 versus Al(OH)3/7.35: 0.002; Al(OH)3 versus Al(OH)3/7.36: 0.03; Al(OH)3 versus Al(OH)3/7.37: 0.0008; Al(OH)3 versus Al(OH)3/7.10: 0.00007; 7.35 versus Al(OH)3/7.35: 0.002; 7.36 versus Al(OH)3/7.36: 0.01; 7.37 versus Al(OH)3/7.37: 0.0002; 7.10 versus Al(OH)3/7.10: 0.00007. (C) Enhancement of SBA titers by SMIP.7-2 and Al(OH)3-adsorbed SMIP.7-10 was compared. Each data point represents mean SBA titer from each individual experiment (±SD). (D) SBA titers against 17 N. meningitidis strains were determined from the pooled serum of mice (n = 10) immunized with 3MenB/Al(OH)3/SMIP.7-10. Statistical significance was determined for the three doses of SMIP.7-10 versus the Al(OH)3 control for all strains combined (n = 17). ***P = 0.004; **P = 0.0003; *P = 0.001.

Al(OH)3 adsorption of SMIPs: A generalizable approach for the development of optimal adjuvants

Three different phosphonate SMIPs of the BZN chemotype (18) were tested as adjuvants in the presence or absence of Al(OH)3 (Fig. 5A). In all cases, adsorption to Al(OH)3 led to marked increases in SBA titers against the MenB NZ98 strain. The unformulated phosphonate SMIPs were extremely inefficient at increasing SBA titers and were usually no better than Al(OH)3 alone. However, second-generation SMIPs adsorbed to Al(OH)3 were more effective at inducing MenB SBA titers compared to first-generation compounds like SMIP.7-2 (Fig. 5C). Thus, the systematic optimization strategy ultimately led to adjuvant formulations that elicit highly localized and short-lived inflammatory responses while simultaneously improving efficacy. To extend the utility of the phosphonate chemistry to additional chemotypes, three different classes of TLR7-dependent SMIPs were modified with a phosphonate group like the BZN class and tested for their adjuvant activity in the MenB model (Fig. 5B and table S2). Again, unformulated phosphonate SMIPs were poor enhancers of SBA titers against NZ98 but were significantly improved when adsorbed to Al(OH)3. The approach of functionalizing SMIPs with linkered phosphonate groups is generally applicable to distinct chemical moieties and is likely a strategy that can be exploited for SMIPs targeting receptors other than TLR7. These findings confirm that soluble TLR7 agonists, like R848 and the unadsorbed phosphonate compounds, are weak adjuvants despite potent and generalized immune activation. Limiting the distribution and duration of inflammation not only decreases safety and tolerability concerns but can also lead to substantially increased efficacy, allowing us to define the minimal requirements of adjuvanticity more precisely.

Although increased functional antibodies against a single strain like NZ98 is an important efficacy indicator, the key hurdle for universal vaccination against N. meningitidis B is eliciting SBA antibodies capable of broad protection against the incredibly high number of circulating strains (30, 31). To understand the potential of the phosphonate BZN Al(OH)3 formulations to increase breadth of coverage, 17 MenB strains, including many of the most SBA-resistant ones, were tested for their susceptibility to killing by sera from mice immunized with 3MenB and SMIP.7-10 adsorbed to Al(OH)3 (Fig. 5D). Three doses (100, 25, and 5 μg) of SMIP.7-10 were used in the formulations and compared with the 3MenB antigens formulated in Al(OH)3 alone. At all doses and for nearly every strain, a fourfold or greater increase in SBA titers was observed (Fig. 5D). The percent of strains susceptible to killing at antibody titers ≥1024 was calculated, and 100% coverage was obtained for the 100- and 25-μg doses of SMIP.7-10, whereas the 5-μg dose led to >90% predicted efficacy. When a more rigorous cutoff was used (SBA titers ≥4096), all three doses of Al(OH)3-formulated SMIPs led to between 67 and 87% coverage. Together, these results demonstrate the potential clinical utility of SMIP-based adjuvants functionalized for adsorption to Al(OH)3.

TH and B cell priming by AI(OH)3-adsorbed SMIPs

The data described above demonstrate that TLR7 SMIPs adsorbed to Al(OH)3 significantly enhance the magnitude and quality of antibody responses. We next explored whether improved B cell responses were the consequence of a more efficient priming of CD4 TH cells by Al(OH)3/SMIP formulations. Therefore, we compared the impact of Al(OH)3 to that of Al(OH)3/SMIP for priming of T cells using two protein antigen models: haptenated OVA and the recombinant protein MenB287-953 (fig. S5). In the OVA model mice were left untreated (naïve) or were immunized with OVA/Al(OH)3 or OVA/Al(OH)3/SMIP.7-11 (fig. S5B). Twenty days later, antigen-specific splenic CD4 T cells were quantified, and the results showed that the Al(OH)3/SMIP.7-11 formulation resulted in higher frequencies of OVA-specific T cells producing cytokines [IL-2, tumor necrosis factor–α (TNF-α), and IFN-γ] upon peptide stimulation in vitro. To test the ability of these primed T cells to drive antibody production, the mice (including the naïve group) were next immunized (day 21) with OVA conjugated with the hapten, nitrophenol (NP) alone [no Al(OH)3 or SMIP], and the anti-NP antibody titers were measured (fig. S5, D and F). With this approach, we were able to investigate the ability of the OVA-specific T cells to provide help to unprimed B cells and generate an anti-NP antibody response in the absence of TLR7 agonism. In the second model, groups of mice were left untreated (naïve) or immunized with two synthetic peptides from the MenB287-953 recombinant protein. The peptides were formulated with Al(OH)3 alone or Al(OH)3/SMIP.7-10. Again, Al(OH)3/SMIP.7-10 was able to induce more peptide-specific T cells. All mice (peptide-immunized and naïve) received an immunization (day 21) with MenB287-953 whole-protein antigen alone (no adjuvant), and antigen-specific antibodies were measured to reveal the potential of peptide-specific T cells to help naïve B cells (fig. S5, E and G), and these results correlated with the primary T cell responses to the MenB peptides (fig. S5C). In both the NP and MenB models, the T cells primed with vaccines containing Al(OH)3/SMIP drove antibody responses (total IgG and IgG2a) that were higher than those seen with Al(OH)3 alone priming.

Rapid induction of protective antibodies against Bacillus anthracis after immunization with Al(OH)3/SMIP-formulated recombinant protective antigen

The data thus far suggest that Al(OH)3/SMIP-formulated vaccines lead to faster induction of primary B and T cell responses than do vaccines formulated Al(OH)3 alone, and this feature should be important for vaccines where rapid protection is critical. Emerging infectious diseases, pandemics, and biothreats, such as anthrax, represent an area where rapid response vaccines are needed. We therefore investigated whether antibodies induced by a single immunization with recombinant protective antigen (rPA) formulated with Al(OH)3 or Al(OH)3/SMIP.7-10 could provide protection against B. anthracis (anthrax) challenge. Initial data demonstrated that Al(OH)3/SMIP formulation induced superior toxin-neutralizing antibody titers compared to Al(OH)3 alone after one or two immunizations, and the differences after a single immunization were particularly striking (Fig. 6A). To focus on protective antibodies in the challenge studies, healthy mice were immunized once with rPA formulated with Al(OH)3 alone or with Al(OH)3/SMIP.7-10, and sera were collected for passive transfer into naïve mice subsequently challenged with B. anthracis (Fig. 6B). Transfer of 100 or 25 μl of serum from mice previously immunized with rPA/Al(OH)3/SMIP.7-10 provided 100 and 80% protection from challenge, whereas the same doses of transferred serum from mice immunized with rPA/Al(OH)3 led to far less protection. Lower doses (12.5 and 6.25 μl) of serum from both groups did not confer any protective effect. Similar to the results for MenB, immunization with SMIP.7-10 alone did not yield serum with any protective benefit. These results indicate that a single immunization with the Al(OH)3/SMIP.7-10 formulation leads to rapid priming of naïve T and B cells that is sufficient to provide protection from lethal challenge with B. anthracis.

Fig. 6. Neutralizing titers to rPA-immune and serum transfer protects mice from intraperitoneal challenge with B. anthracis Sterne.

(A) Mice (10 per group) were vaccinated two times (3 weeks apart) with rPA/Al(OH)3 or rPA/Al(OH)3/SMIP.7-10, and serum was collected at day 20 (Post 1st) and day 42 (Post 2nd). Toxin neutralization assay titers were determined by the Battelle Memorial Institute, and values are reported as neutralization factor50 (NEF50) according to the methods of Omland et al. (39). *P = 0.001; **P = 0.002. (B) Survival of mice (n = 10 per group) given no serum (), 100 μl of serum (), or 25 μl of serum (– –) from mice vaccinated once with rPA/Al(OH)3/SMIP.7-10; 100 μl () or 25 μl (– –) of serum from mice vaccinated with rPA/Al(OH)3; or 100 μl () or 25 μl (– –) of serum from mice vaccinated with rPA/SMIP.7-10. A log-rank test was used to determine significance between the groups indicated on the graph (*P = 0.004; **P = 0.0005).


Innumerable preclinical adjuvants have been described in the scientific literature and patents, yet very few are approved as components of prophylactic vaccines for humans (32). With the view that adjuvant discovery is easier than adjuvant clinical use, we turned around the question of what makes an optimal adjuvant by deemphasizing potency and probing the minimal immune activation events needed for efficacy. Systematic comparison of a diverse set of tool compounds demonstrated that LMW compounds can be optimized for use as adjuvants in clinically applicable vaccine formulations. Although weaker TLR7 agonists in vitro, the first-generation insoluble SMIPs were better adjuvants than the R848 benchmark, a very potent and soluble SMIP. R848 has a poor tolerability profile when tested in humans, and common systemic side effects include fever, headache, malaise, and myalgia (911). These “flu-like symptoms” are likely due to systemic immune activation and are routinely associated with vaccine reactogenicity (33). Safety and tolerability were prioritized over potency from the outset of the BZN optimization effort. Through this unconventional approach, not only was the systemic and generalized inflammation associated with TLR stimulation eliminated but also the efficacy of adjuvant formulations was significantly enhanced. Thus, the results show that adjuvant potency and poor tolerability are not inextricably linked but can be uncoupled to open the therapeutic window.

Although the insoluble SMIPs had many advantages over the soluble ones, they created formulation and manufacturing hurdles. Moreover, their extended residence times (>2 weeks) in the muscle did not appear to be required for efficacy and therefore did not represent a minimal requirement for their adjuvant activity. The engineering of more soluble SMIPs that could be adsorbed to Al(OH)3 offered a solution to these problems. The second-generation compounds formulated with Al(OH)3 retained all of the key advantages of the insoluble ones but were more reliably formulated, had substantially reduced muscle half-lives, and exhibited improved efficacy in the MenB model when compared with the first-generation compounds (fig. S1). That structurally diverse TLR7 agonists could be engineered with these features demonstrates the broad utility of the linker and phosphonate chemistry used here and should be applicable to SMIPs targeting receptors and pathogen sensors other than TLR7. Aluminum-based adjuvants have been included in numerous licensed vaccines for nearly a century, and they are widely accepted to be safe and well tolerated (34). Moreover, Al(OH)3 adsorption is known to stabilize protein antigens, which allows for fully liquid, single-vial vaccine formulations (29, 34). Therefore, it is an attractive platform for the development of next-generation adjuvants, and a key discovery of these studies was engineering SMIPs for stable and efficient adsorption to Al(OH)3.

A traditional medicinal chemistry approach for drug development aims to optimize compound solubility to achieve high systemic exposure and maximal bioavailability in vivo (1, 35, 36). This, however, is not the best approach for adjuvant design. Solubility is important for formulation and scale-up and Al(OH)3 adsorption of highly soluble compounds was used to minimize the inflammatory impact of second-generation SMIPs. Through this work, we define new guidelines for small-molecule adjuvant design that emphasize the unique mechanisms and target populations of vaccines. Most drugs act directly on their targets in tissues and organs throughout the body, and therefore, sustained exposure systemically is required. By contrast, vaccine efficacy is achieved indirectly by initiating a cascade of immune activation events, leading to durable and systemic protection. Extrapolating these concepts to the present study demonstrates that the broad immune activation elicited by systemic exposure not only leads to wasted inflammation but actually decreases efficacy.

We show that it is now possible to change the fundamental characteristics of adjuvant molecules available for vaccine development. The stepwise approach for the rational design of adjuvant formulations described here should have applications beyond SMIPs targeting TLR7. Thus, by exploiting the strategies and platforms of drug discovery and development and adapting them to vaccine use, we achieved precise pharmacological control of SMIP adjuvants. We demonstrate that minimizing the proinflammatory activities of SMIPs in space and time can lead to improved efficacy, providing a framework for uncoupling the benefits of adjuvants from their inherent immunotoxicities. Combining medicinal chemistry, formulation science, and quantitative immunopharmacology enabled addressing the specific safety and efficacy requirements of vaccines. The ability to pinpoint the inflammatory activities of adjuvant compounds and fine-tune their chemical properties by iterative design provides fertile ground for greater success in adjuvant development in the coming years.


Study design

The work described represents a merger of small molecule and vaccine development in an industrial setting. The flowcharts shown in Figs. 1A and 3A outline the iterative approach used to optimize SMIPs as vaccine adjuvants. The goal was to minimize the inflammatory activity of the compounds through medicinal chemistry and formulation while maintaining efficacy. Small molecule optimization begins with in vitro screening and chemical analoging. Here, advancement of potent and selective TLR7 agonists is described for first- and second-generation SMIP adjuvants. For in vivo testing as adjuvants in mice, determining the PK and PD of the compounds allowed for investigation of candidates that differentially activated the innate immune system of the animals spatially and temporally. These candidates and the Alum/TLR7 formulations were then tested as vaccine adjuvants using the clinical correlate of protection for MenB vaccines and a challenge model for B. anthracis. Group sizes for in vivo studies were chosen based on standard practices in the industry. For PK/PD determinations 3 animals per group were used and experiments were repeated at least 3 times. For immunogenicity studies 8 to 10 animals per group allows for differences >5-fold to be statistically significant. No randomization or blinding was performed during the course of these studies.

Animal studies

For PK and PD experiments, female BALB/c mice (6 to 8 weeks old) were injected intramucularly with a total of 100 μl of the indicated formulation (50 μl in each hind leg) and then bled retro-orbitally at different time points after injection. Blood was processed into serum by centrifugation followed by protein precipitation, reversed-phase gradient elution, and multiple reaction monitoring detection via ESI+ mass spectrometry to determine SMIP serum concentration (PK). In addition, serum was also used to examine cytokines by MSD cytokine arrays (PD). In some studies, animals were sacrificed after the final blood collection (24 hours), and muscle, inguinal lymph nodes, and spleen were removed and flash frozen in liquid nitrogen for tissue PK analyses. Similarly, for cellular activation, spleen and inguinal lymph nodes were harvested and dissociated into a single-cell suspension in phosphate-buffered saline (PBS) supplemented with 10% fetal bovine serum (FBS). After red blood cell lysis, spleen and lymph node cells were stained with CD19-APC (allophycocyanin), CD86-PE (phycoerythrin), CD69-FITC (fluorescein isothiocyanate), and CD4-APC-Cy7 (Becton Dickinson) in PBS/3% FBS and subjected to flow cytometry on a BD LSRII flow cytometer. Data were analyzed using FlowJo software. For gene expression profiles, RNA was extracted from muscle or inguinal lymph nodes, labeled with Cy3, and hybridized onto Agilent 4×44k whole-mouse genome microarrays as previously described (32). One hundred sixty-three genes with fold change over vehicle ≥4 and P value ≤0.05 in triplicate were selected as significant. IFN signature gene clusters were assessed by gene annotation enrichment analysis [array data have been uploaded to the ArrayStar database of the European Bioinformatic Institute ( and the accession number is E-MTAB-2948]. For OVA (BALB/c) and 3MenB (CD-1) vaccinations, mice were immunized intramuscularly with the indicated vaccine formulations in a total volume of 100 μl (50 μl per hind leg) on day 0 of the study. On day 14 (3MenB) or day 21 (OVA), mice were boosted and blood was collected via retro-orbital eye bleeding 14 days after boost. The blood was then processed into serum to determine antibody titers by ELISA or SBA titers. For immunization studies, 8 to 10 mice per group were used, which allows statistical significance within 95% confidence limits for a fivefold increase in antibody titer. No randomization or blinding of samples was performed. All animal protocols were approved by the Novartis Institutional Animal Care and Use Committee.

Protocol for acquisition of pH-dependent solubility

The solubility-pH profiles of SMIP.7-10, SMIP.7-12, and SMIP.7-6 were determined by potentiometric titration using Sirius T3 titrator (Sirius Analytical, East Sussex, UK), whereas the solubility-pH profiles of SMIP.7-14 and SMIP.7-11 were obtained by saturation shake-flask solubility assay because the solubility of these two compounds was poor and no aqueous pKa values were able to be obtained. For the potentiometric titration approach, the pKa values of each compound were determined using the same Sirius T3 titrator before solubility determination using methanol as the co-solvent, and data were processed with T3 accompanying software. In the potentiometric solubility assay, about 2 mg of the solid sample was weighed into a T3 vial, and titration was performed between pH 10-1.5. Precipitation occurred during the titration range, causing pKa values to shift. The intrinsic solubility was calculated according to the pKa shift on the basis of Noyes-Whitney theory using T3 accompanying software, and solubility-pH profile was extrapolated on the basis of the Henderson-Hasselbalch equation. In the saturated shake-flask approach, saturated solubility samples were prepared at different pHs and incubated for 24 hours. The samples were then filtered through 0.22-μm polyvinylidene difluoride membranes, and the filtrates were analyzed and quantified against a standard calibration curve of each compound. The solubility-pH profiles were constructed by connecting solubility data at different pH values.

Measurement of SBA titers against N. meningitides B

Where indicated, SBA activity against N. meningitidis strains was evaluated using a standard method (30) or an automated assay (37).

Cell-based assays

Assays of SMIP activity in vitro on TLR7 or TLR8 stable transfectants, murine splenocytes, and human peripheral blood mononuclear cells were performed as described (21).

Analysis of SMIP adsorption to Al(OH)3

SMIP was adsorbed to Al(OH)3 (3 mg/ml) in 10 mM histidine buffer (pH 6.5). NaCl was added to adjust the osmolality to 300 ± 60 mosmol/kg. Formulations were left under gentle agitation for 3 hours at room temperature to allow SMIP adsorption to occur. To quantify the amount of SMIP adsorbed to Al(OH)3, the concentration-free SMIP in the supernatant was measured. Al(OH)3 was pelleted for 15 min at 13,000g, and the supernatant was evaluated by high-performance liquid chromatography (HPLC) (Alliance, Waters). Supernatant (0.010 ml) was loaded on a Waters Atlantis T3, 5 μm, 2.1 × 50–mm column, and SMIP was eluted with a 4-min gradient of 10 to 100% of 100% acetonitrile in 0.1% trifluoroacetic acid at a flow rate of 1.25 ml/min, using a Diode Array UV-Vis detector. Analysis was run at 270 nm, corresponding to the wavelength of maximum absorption of SMIP. A standard curve and standards were prepared by dissolving an appropriate amount of sodium salt of the compound to reach the final concentration of 0.1 mg/ml. Standards (50, 25, 10, and 5%) were prepared by dilution from the first standard. To demonstrate SMIP recovery from Al(OH)3, the formulations were centrifuged for 15 min at 13,000g, and the Al(OH)3/SMIP pellet was resuspended in 0.5 M KH2PO4 pH 9 and left at 37°C overnight in gentle agitation. Al(OH)3 was then repelleted by 15 min at 13,000g, and the SMIP-containing supernatant was analyzed by HPLC at the same conditions described above. For imaging, a 20-μl drop of the Al(OH)3/SMIP suspension was placed on a glass microscope slide. A glass coverslip was placed over the sample, which was then imaged under ×20 magnification on a Carl Zeiss LSM700 confocal microsope with an excitation of 405 nm and an emission of 605 nm with a 70-μm pinhole diameter. The Al(OH)3/SMIP formulation was prepared as described above, and fluorescence was acquired on a FACSCanto II (BD Biosciences) with an AmCyan channel (525/50 nm) filter. Al(OH)3 alone was used as a negative control. For each sample, at least 10,000 events were acquired, with a flow rate of 1000 events/s. The morphological analysis of Al(OH)3 particles was the first step used to define the Al(OH)3 population that subsequently was used to evaluate the intensity of fluorescence.

Trivalent N. meningitidis B vaccine antigen formulation characterization

For immunization purposes, the three recombinant antigens for N. meningitidis serotype B were added at a concentration of 0.1 mg/ml each. Antigen adsorption was monitored via SDS–polyacrylamide gel electrophoresis (SDS-PAGE), and antigen integrity via Western blot. Formulations (0.25 ml) were centrifuged for 15 min at 13,000g. Supernatant (0.02 ml) was denatured for 10 min at 95°C in SDS-PAGE sample buffer containing 2% (w/v) SDS and loaded onto 4 to 12% (w/v) polyacrylamide gels (Invitrogen). Supernatant (0.2 ml) was precipitated with 10% trichloroacetic acid/0.04 % deoxycholate and loaded onto 4 to 12% (w/v) polyacrylamide gels (Invitrogen). Antigens adsorbed onto the Al(OH)3 pellet were desorbed in 0.5 M KH2PO4 (pH 6.5), 2% SDS, 5 mM dithiothreitol, 6.25% glycerol for 10 min at 95°C and loaded onto 4 to 12% (w/v) polyacrylamide gels (Invitrogen). Gels were run in Mops buffer (Invitrogen) and stained with SimplyBlue SafeStain (Invitrogen).

Western blot analysis was performed as follows: After protein transfer to nitrocellulose filters (using semidry iBlot blotting system and kit, Invitrogen), protein integrity was visualized using specific rabbit antibodies and a peroxidase-conjugated goat anti-rabbit IgG polyclonal antibody.

T and B cell priming studies

Pooled spleen cells were prepared, stimulated with antigenic peptides, and stained for cell surface CD4 and CD8 and intracellular cytokines as described by Geall et al. (see Supplementary Materials). OVA-specific CD4 T cells were stimulated with the OVA323–339 synthetic peptide (10 μg/ml) (sequence ISQAVHAAHAEINEAGR). MenB-specific CD4 T cells were stimulated with a mixture of two synthetic peptides with sequences YKPKPTSFARFRRSARSRR and VNVGMTKSVRIDIQIEAAKQ (10 μg/ml each). ELISA assays were performed as described by Geall et al., modified for the antigens of interest. Briefly, mouse sera were serially diluted and added to wells of 96-well plates that had been coated with antigen, and then blocked with Superblock blocking buffer (Thermo Scientific). Antigen-specific serum antibodies bound to the antigen-coated wells were detected using either goat anti-mouse IgG or goat anti-mouse IgG2a (both horseradish peroxidase conjugates; Southern Biotech), and tetramethylbenzidine peroxidase substrate solution (KPL) ELISA plates were coated with 0.5 μg per well NP4–bovine serum albumin (Biosearch Technologies). ELISA plates were coated with MenB287-953 protein (0.1 μg per well).

Vaccination with rPA and protection from anthrax challenge after passive serum transfer

Female A/J mice were vaccinated with rPA (1 μg; List Laboratories) formulated with Al(OH)3 (200 μg), SMIP-7.10 (25 μg), or both. Additional mice were vaccinated with irrelevant antigen (chicken OVA, 1 μg) formulated with Al(OH)3/SMIP-7.10. Sera were collected 28 days later and pooled. Naïve A/J mice were then given intraperitoneal injections of pooled serum in the amounts of 100, 25, 6.25, or 0.78 μl (diluted with PBS to a total volume of 200 μL). There were 10 mice per group. Control mice were not given injections. The next day, mice were challenged with 1 × 105 colony-forming units of B. anthracis Sterne spores injected intraperitoneally. Survival was monitored for 28 days after challenge. Survival differences between groups were determined by the log-rank test (GraphPad Prism).

Statistical methods

Unless otherwise stated, all PK and PD studies were performed using 3 mice per group and all experiments were repeated 3 to 10 times, yielding consistent results. For immunization studies, 10 mice per group were used routinely except when indicated in the figure legends. To compare SBA titers between groups, two approaches were employed. In some experiments, sera from individual mice were assayed and the data are presented individually. In others, sera was pooled from each group of animals and assayed in triplicate. For all studies except the B. anthracis challenge studies (Fig. 6B) t tests were used to determine statistical significance. One-tailed tests were employed for comparison within a single experiment between groups and controls, whereas, 2-tailed tests were used for comparison across different experiments. For the challenge studies, a log-rank test was used to determine significance. Statistical analysis was performed using GraphPad Prism software.


Fig. S1. Adjuvants by design.

Fig. S2. Cellular activation in draining lymph nodes by first generation SMIPs.

Fig. S3. First generation SMIPs localized by physical chemical properties.

Fig. S4. 3MenB/Al(OH)3/SMIP formulation characterization.

Fig. S5. Rapid priming of TH cells by Al(OH)3 adsorbed SMIPs.

Table S1. First-generation SMIP in vitro activity.

Table S2. Second-generation SMIP in vitro activity.

Table S3. SMIP in vitro cytokine profile.

Flow cytometry gating strategy and isotype controls (Fig. 4B)

Source data Fig. 1

Source data Fig. 2

Source data Fig. 3

Source data Fig. 4

Source data Fig. 5

Source data Fig. 6


  1. Acknowledgments: We wish to acknowledge R. Rappuoli, D. Piccioli, E. Soldaini, and A. Seubert for helpful discussions. We also want to thank L. Boase, D. Hall, J. Hampton, M. Hedden, M. Ibanez, J. Janes, A. Mak, K.-A. Masterman, P. Mishra, J. Rojek, G. Santos, M. Shapiro, B. Wacknov, M. Young, K. Yue, X. Zhang, L. Zhou, and Y. Zou for important technical contributions to the work and V. Pasquetto and B. Owens for project management support. Funding: This work was supported in part by the Transformational Medical Technologies program contract HDTRA1-07-9-0001 from the Department of Defense Chemical and Biological Defense program through the Defense Threat Reduction Agency. Author contributions: T.Y.-H.W. led chemistry efforts; M.S. led formulation efforts; A.T.M. led pharmacology efforts; E.D.G. led Siena team and conceived the project; F.D. generated and characterized Al(OH)3/TLR7 formulations; U.D. performed in vitro and in vivo characterization of SMIPs; D.A.G.S. helped conceive second-generation SMIPs; S. Bufali generated and characterized Al(OH)3/TLR7 formulations; M.L.M. is a former SMIP team leader; A.E.H. performed in vivo pharmacology; A.C. and Y.L. designed and synthesized SMIPs; B.P.N. performed in vitro characterization of SMIPs; E.T. performed and analyzed microarray experiments; C.M.F. performed in vivo pharmacology; G.R.O. performed in vitro and in vivo characterization of SMIPs and statistical analysis; L.A.B. generated and characterized Al(OH)3/TLR7 formulations; E.M. performed and analyzed microarray experiments; C.L. performed SMIP PKs; S.A. performed in vitro and in vivo characterization of SMIPs; S.V. performed in vitro and in vivo characterization of SMIPs; S.C. performed in vitro and in vivo characterization of SMIPs; D.L. generated and characterized Al(OH)3/TLR7 formulations; B.B. performed MenB SBA assays; E.C. performed MenB SBA assays; P.M. formulated Alum/TLR7/rPA; S. Bavari, R.G.P., and T.K.W. coordinated B. anthracis challenge studies; D.T.O. led adjuvant team and conceived the project; M.P.C. led San Diego team and conceived the project; N.M.V. is a global SMIP team founder, statistical analysis, and conceived the project. Competing interests: The authors declare competing interests as current or former employees of Novartis AG. Data and materials availability: Microarray data has been deposited with ArrayExpress (accession number, E-MTAB-2948). No material used in this study was obtained from external parties under material transfer agreement.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article