Toll-like receptors (TLRs) are pattern recognition receptors that recognize specific molecular patterns present in molecules that are broadly shared by pathogens, but are structurally distinct from host molecules [1], [2]. The activation of TLRs by their cognate ligands leads to activation of innate immune effector mechanisms, including the production of pro-inflammatory cytokines, and up-regulation of major histocompatibility complex (MHC) molecules and co-stimulatory signals in antigen-presenting cells. The activation of the innate immune system serves to mobilize and amplify subsequent specific adaptive immune responses involving both T- and B-cell effector functions [3]–[6]. Thus, TLR stimuli serve to link innate and adaptive immunity [4] and can therefore be exploited as powerful adjuvants in eliciting both primary and anamnestic immune responses.

At least 10 TLRs are encoded in the human genome [2]. The ligands for these receptors are highly conserved microbial molecules such as lipopolysaccharides (LPS) (recognized by TLR4), lipopeptides (TLR2 in combination with TLR1 or TLR6), flagellin (TLR5), single stranded RNA (TLR7 and TLR8), double stranded RNA (TLR3), CpG motif-containing DNA (recognized by TLR9), and profilin present on uropathogenic bacteria (TLR 11) [7]. TLR1, -2, -4, -5, and -6 respond to extracellular stimuli, while TLR3, -7, -8 and -9 respond to intracytoplasmic pathogen-associated molecular patterns (PAMPs), being associated with the endolysosomal compartment [2].

In evaluating representative members of virtually all known TLR agonists in a series of hierarchical assays including primary TLR-reporter assays, secondary indices of immune activation such as cytokine induction and activation of lymphocytic subsets in whole human blood, and tertiary screens characterizing transcriptomal activation patterns with a view to identifying optimal immunostimulatory chemotypes [8], we found that TLR7 agonists, represented by the imidazoquinoline chemotype (Compound 2, Fig. 1), were extraordinarily immunostimulatory. Extensive structure-activity relationship studies [9]–[13] led to the synthesis of a highly potent TLR7/TLR8 dual-agonistic 1-(4-(aminomethyl)benzyl)-2-butyl-1H-imidazo[4,5-c]quinolin-4-amine (Compound 4, Fig. 2), whose free primary amine on the N¹-benzyl substituent proved a convenient and versatile handle for the attachment of fluorophores [11] as well as a precursor for model self-adjuvanting subunit vaccine constructs with the imidazoquinoline covalently attached to the antigen [13].

Initiation of signaling by TLRs involve homotypic or heterotypic dimerization [14], [15], and we have previously explored the activities of dimeric imidazoquinoline constructs [12]. We asked if a multimeric (dendrimeric) construct of 4 would alter its activity profile. We therefore synthesized a dendrimeric molecule bearing six units of the TLR7/TLR8 dual-agonistic imidazoquinoline 4. Upon dendrimerization, however, we noted a complete loss of TLR8-stimulatory activity, with selective retention of the TLR7-agonistic activity of its parent monomer in primary screens employing TLR-specific reporter gene assays. This was reflected by a complete absence of TLR8-driven proinflammatory cytokine and interferon (IFN)-γ induction in human PBMCs, but with preservation of TLR7-driven IFN-α induction. The dendrimer was found to be superior to the imidazoquinoline monomer in inducing high titers of high-affinity antibodies. Additionally, epitope mapping experiments showed that the dendrimer induced immunoreactivity to more contiguous peptide epitopes along the amino acid sequence of bovine α-lactalbumin.

All of the solvents and reagents used were obtained commercially and used as such unless noted otherwise. Moisture- or air-sensitive reactions were conducted under nitrogen atmosphere in oven-dried glass apparatus. The solvents were removed under reduced pressure using standard rotary evaporators. Flash column chromatography was carried out using RediSep R_f ‘Gold’ high performance silica columns on CombiFlash R_f instrument unless otherwise mentioned, while thin-layer chromatography was carried out on silica gel CCM pre-coated aluminum sheets. All intermediates and final target compounds were characterized by ¹H and ¹³C NMR; spectra were verified to be consistent with their structure. Purity for all final compounds was confirmed to be greater than 97% by LC-MS using a Zorbax Eclipse Plus 4.6 mm × 150 mm, 5 µm analytical reverse phase C₁₈ column with H₂O-isopropanol or H₂O-CH₃CN gradients and an Agilent ESI-TOF mass spectrometer (mass accuracy of 10 ppm) operating in the positive ion acquisition mode.

As mentioned earlier, we were desirous in examining dendrimeric constructs of the imidazoquinoline with to view to testing the hypothesis that such larger constructs may enhance adjuvanticity. We did not, however, know a priori if such molecules would even retain TLR7/8 agonistic activity. Our first attempt at evaluating such dendrimeric molecules began with the synthesis of 3 (Fig. 1) which was accomplished by reacting an isothiocyanate derivative 2 that we had used earlier [11], [13] with the triamine 1. We were gratified that this trimeric molecule 3 retained, in large measure, both the TLR7- and TLR8-agonistic activities of the parent monomeric imidazoquinoline (Fig. 1). We next extended our efforts in synthesizing a hexameric construct 8, which was obtained via a ‘click’ reaction of the azide-bearing 6 with the hexa-alkyne 7 (Fig. 2). To our surprise, we noted that while the dendrimer 8 retained substantial TLR7-agonistic activity, its TLR8-stimulating properties appeared entirely abrogated in primary screens employing reporter gene assays (Fig. 3).

In recognition that bioactivity readouts using cell-culture systems may not always reflect with fidelity in vivo behavior owing to a variety of reasons, including differential plasma protein binding (that we ourselves have observed and characterized) [19], it was important to verify that the activity profiles observed in TLR-specific reporter gene assays was also seen in primary human cells. It is to be noted that we [10], [16], [20] and others [21]–[23] have shown that pure TLR7 agonists selectively induce Type I interferons, while the production of IFN-γ and proinflammatory cytokines is attributable primarily to TLR8 activation. We confirmed that the dendrimer was selectively bereft of TLR8-agonistic properties in secondary screens using human PBMCs in which the dendrimer 8, unlike its parent monomer 4, showed selective induction of TLR7-driven IFN-α, but not of IFN-γ (Fig. 3), or proinflammatory cytokines such as TNF-α, IL-1β, and IL-6 (Fig. 4).

It is to be noted that both in interferon (Fig. 3) and in cytokine-release assays (Fig. 4), a bimodal dose-response profile characterized by an initial dose-dependent increase in analyte concentration followed by an apparent suppression in interferon/cytokine production was observed as we have previously reported [10], [13], [16]. The origin and cause of this bimodal response is not clear, but we verified that this is not due to cytotoxicity using standard XTT [24] and resazurin [25] assays.

Immunization of rabbits with bovine α-lactalbumin with either the monomeric compound 4 or the dendrimer 8 elicited strong humoral responses (Fig. 5); however, although not statistically significant owing to the small sample sizes, anti-α-lactalbumin IgG titers in rabbits adjuvanted with 8 were higher than titers elicited by 4, or a high-potency, pure TLR7-agonistic imidazoquinoline (1-benzyl-2-butyl-1H-imidazo[4,5-c]quinolin-4-amine; reported as Compound 31 in Ref. 9) (Fig. 5). Affinity measures of antibody quality using chaotropic ELISAs also indicated that the use of 8 as adjuvant resulted in significantly higher-affinity antibodies (Fig. 5). Linear epitope mapping clearly showed that the dendrimer 8 induced immunoreactivity to more contiguous peptide epitopes along the amino acid sequence of bovine α-lactalbumin. First, a pronounced immunoreactivity to a peptide sequence centered around amino acids 42–54 (VQNNDSTEYGLFQ) was observed in animals adjuvanted with 8; second, higher immunoreactivity to peptide sequences centered on amino acids 6–18 (CEVFRELKDLKGY) and 108–120 (KALCSEKLDQWLC) were also observed (Fig. 6). These data clearly show that immunization with 8 as adjuvant extends the immunoreactivity profile of the humoral response to encompass longer stretches of the amino acid sequence of the protein antigen. These results were somewhat unexpected because an examination of the amino acid sequences of bovine and rabbit α-lactalbumin shows differences throughput the length of the protein (Fig. 7; Waterman-Eggert score: 579; 61.2% identity; 85.1% similar), and we had surmised, perhaps naïvely, that the immunoreactivity patterns of linear epitopes in the monomer- and dendrimer-adjuvanted samples would be very similar, with the dendrimer evoking stronger responses. The mechanism(s) underlying the superior adjuvanticity of 8 are unknown; we hypothesize that this may be related to the lack of induction of counter-regulatory IL-10.

The availability of experimentally determined linear epitope mapping data provided an opportunity to test and benchmark known linear epitope prediction algorithms [26]. We compared six methods that are implemented in Immune Epitope Database Analysis Resource (www.immuneepitope.org): (a) Chou-Fasman β-turn prediction propensity [27]; (b) Emini surface accessibility parameter [28]; (c) Karplus and Schulz flexibility prediction [29]; (d) Kolaskar and Tongaonkar method [30]; (e) Parker hydrophilicity prediction [31]; (f) Bepipred linear epitope prediction method [32]. The Parker hydrophilicity predicition algorithm yielded a profile that correlated best with the experimentally-determined linear epitope immunoreactivity patterns (Fig. 8).

It would also be pertinent to note two additional points: First, the absolute concentration of 100 µg/dose of 4 and 8 are, respectively, 278 nMoles and 33 nMoles; therefore, lower concentrations of 8 exhibits higher adjuvantic activity than 4, inducing higher titers of antibodies, which are also of higher quality as adjudged by chaotropic ELISA and epitope mapping. Second, although no apparent local or systemic adverse effects were reported for any of the rabbits, the dendrimer 8, unlike 4, is a pure TLR7 agonist, inducing only Type I IFN, and no proinflammatory cytokines. Based on current paradigms, one would predict far lower reactogenicity for 8. Accordingly, a detailed comparison of local and systemic reactogenicity with biomarker profiling is being planned.