Design, synthesis and evaluation of optimized saponin variants derived from the vaccine adjuvant QS-21

First-generation of stable amide-derived acyl chain variants

In the Gin group, the chemical instability of QS-21, which stems primarily from its hydrolytically labile acyl-fucose ester linkage, was first addressed by developing unnatural amide-derived acyl chain variants incorporating more robust amide groups instead of the unstable native esters [11]. This modification involved re-working of the linear tetrasaccharide to include a suitably protected “fucose surrogate” bearing an N-functionality at the C4-position, on which to anchor a variety of acyl chains through an amide linkage by chemoselective N-acylation. The first series of acyl chain analogs prepared by Gin and coworkers started with the most conservative variant SQS-0101 as a direct isosteric mimic of the natural acyl chain, except for the central ester-to-amide replacement (Fig. 2). More profound structural changes took the form of SQS-0102, SQS-0103, and SQS-0-0-4-5, which incorporated simplified aliphatic acyl chains, either glycosylated (SQS-0102) or non-glycosylated (SQS-0103 and SQS-0-0-4-5), the latter bearing a terminal carboxyl group to improve water solubility [12]. Immunological evaluation in mice of the stable saponin variants with the GD3–KLH melanoma conjugate vaccine demonstrated adjuvant activities comparable to QS-21, as assessed by measuring antibody responses by ELISA 1 week after booster injection (day 72). Moreover, while SQS-0102 elicited considerably increased toxicity, as judged by mouse weight loss, all other QS saponins were significantly less toxic than QS-21, with only minor (SQS-0101) or no weight loss (SQS-0103 and SQS-0-0-4-5) observed in these cases.

Fig. 2:

Structure of first-generation, chemically stable, acyl chain amide variants developed by Gin and coworkers.

Drawing inspiration from the early semisynthesis of the heterogeneous GPI-0100 adjuvant by Marciani et al. [6], Wang and co-workers recently synthesized a number of saponin variants [13], [14], in which the hydrolytically labile native QS21 acyl chain was replaced by simplified acyl chains incorporated at the branched trisaccharide glucuronic acid through an amide linkage (Fig. 3). By analogy to the Gin lab studies above, the introduction of polar groups (carboxylic acid and arabinose monosaccharide) at the terminus of the otherwise lipophilic dodecylamide chain, increased the water-solubility and adjuvant activity of the variants to levels comparable with GPI-0100 [14].

Fig. 3:

(a) Structure of immunoactive isomeric constituents of GPI-0100 and (b) terminal-functionalized acyl chain analogs of GPI-0100 synthesized by Wang et al. [14].

Carbohydrate variants via systematic truncation of the QS linear tetrasaccharide

Based on the stable, simplified carboxylic-acid acyl chain variant SQS-0-0-4-5, which showed potent adjuvant activity and low toxicity, Gin and coworkers next investigated the minimal structural requirement of the linear tetrasaccharide for adjuvant activity [12]. Thus, we prepared additional simpler structures involving systematic truncation of this carbohydrate domain, providing trisaccharide variant SQS-0-0-5-5, disaccharide variant SQS-0-0-6-5, and monosaccharide variant SQS-0-0-9-5 (Fig. 4). The synthesis of the linear trisaccharide itself devoid of the fourth apiose sugar (as trichloroacetimidate donor, 11) was accomplished in 16 steps (Scheme 1b) using a double dehydrative glycosylation procedure [15] starting from conveniently protected xylose and rhamnose monosaccharides and the previously prepared fucose surrogate, 4-azido-4-deoxygalactose 5, synthesized in five steps from Dglucal (Scheme 1a). The protected prosapogenin 12 (PPS-CO₂H), obtained by semisynthesis from QS-extracts [9], was glycosylated with the trisaccharide imidate 11 (BF₃·OEt₂), and the corresponding intermediate 13 was subjected to phenylselenol reduction to arrive at amine 14 (Scheme 2). The last steps to SQS-0-0-5-5 bearing a single sugar truncation involved acylation of the amine with dodecanedioic acid monobenzyl ester, followed by global deprotection via hydrogenolysis (H₂, Pd/C) and acid hydrolysis (TFA/H₂O) and subsequent HPLC purification. The disaccharide and monosaccharide variants (SQS-0-0-6-5 and SQS-0-0-9-5) were prepared in a similar fashion from the corresponding protected carbohydrates.

Fig. 4:

Structures of carbohydrate-truncated variants with systematic truncation of the linear tetrasaccharide.

Scheme 1:

(a) Synthesis of fucose surrogate C4-deoxy-C4-azidogalactopyranoside 5. (b) Synthesis of modified linear trisaccharide donor (11) devoid of the fourth sugar residue.

Scheme 2:

Synthesis of the most elaborate of the carbohydrate-truncated analogs: linear trisaccharide variant SQS-0-0-5-5 with a single-sugar truncation.

Preclinical studies to evaluate the adjuvant activity of these saponins (20 μg) [12] involved mouse vaccination (groups of five mice) with the QS variant of interest and a four-component vaccine containing OVA (ovalbumin, a reliable immunogen that induces antibody responses), MUC1–KLH (MUC1: prostate, breast cancer peptide antigen, non-glycosylated tandem repeat domain), and GD3–KLH (GD3: melanoma, sarcoma glycolipid antigen, conjugated to the highly immunogenic protein KLH). Trisaccharide variant SQS-0-0-5-5 elicited high antibody titers, similar to QS-21, whereas progressively weaker responses were observed with each carbohydrate truncation in disaccharide and monosaccharide variants SQS-0-0-6-5 and SQS-0-0-9-5, respectively (Fig. 5). The toxicity of these saponins was assessed in terms of mouse weight loss. While monosaccharide SQS-0-0-9-5 was the least toxic (one dead mouse, and no weight loss observed), two mice died when administered with disaccharide variant SQS-0-0-6-5 (−8% weight change on day 1) and one mouse died with SQS-0-0-5-5, which also induced considerable weight loss (9% on day 1, 13% on day 2) (Fig. 5e). Although the toxicity issue was not completely resolved, trisaccharide variant SQS-0-0-5-5 showed the best adjuvant activity within this group, comparable to QS-21, and was also more synthetically accessible (24 steps) than tetrasaccharide variant SQS-0-0-4-5 (36 steps). This simplification opened the door to develop additional QS saponins with further structural variations, e.g. at the junction between the triterpene and linear oligosaccharide domains.

Fig. 5:

Immunological evaluation of linear tetrasaccharide truncation variants SQS-0-0-5-5, SQS-0-0-6-5, and SQS-0-0-9-5 (20 μg) with four-component vaccine of OVA (20 μg), MUC1–KLH (2.5 μg), GD3–KLH (10 μg). Adapted from Ref. [12]. Antibody titers after three biweekly vaccinations and booster against (a) KLH (IgG), (b) OVA (IgG), (c) MUC1 (IgG), and (d) GD3 (IgM). Horizontal bars represent median values of titers; statistical significance versus No Adjuvant control: *p≤0.05, **p≤0.01, ***p≤0.001. (e) Toxicity assessment based on median weight loss over 1 week after the first vaccination. NQS-21=natural QS-21; SQS-21=synthetic QS-21.

Central glycosidic linkage variants with modified linker distance, stereochemistry and flexibility

Walkowicz et al. next explored modifications of length, stereochemistry, and flexibility in the central glycosyl ester linkage, and examined their effects on adjuvant activity and toxicity with the synthesis and evaluation of several linker variants [16]: βethanolamide SQS-0-4-5-5, βcarbamate SQS-0-5-5-5, and β-thioester SQS-0-13-5-5; α-ester SQS-0-0-8-5, α-amide SQS-0-6-8-5, and α-carbamate SQS-0-5-8-5; and β-ether SQS-0-12-5-5, and β-thioether SQS-0-14-5-5, respectively (Fig. 6). The syntheses of these variants proved challenging due to the sterically demanding glycosyl bond formation required to join the prosapogenin and oligosaccharide domains together in the context of the complex saponin scaffold. The corresponding glycosyl acceptors were synthesized by functionalization of the protected prosapogenin 12 (PPS-CO₂H) at its C28-carboxylic acid (Scheme 3) and then coupled to the trisaccharide glycosyl donors using two different strategies, with the glycosyl donor acting as an electrophile in a traditional glycosylation reaction (Scheme 4a), or as a nucleophile in a polarity-reversed coupling (Scheme 4b). The resulting prosapogenin–trisaccharide intermediates were then advanced to the fully deprotected central linker variants following the synthetic sequence described above, involving reductive acylation of the azide with the previous carboxylic acid acyl chain, and global deprotection (hydrogenolysis and acid hydrolysis).

Fig. 6:

Structures of central glycosidic linkage variants with modified linker length (left), stereochemistry (middle), and flexibility (right).

Scheme 3:

Synthesis of protected prosapogenin [PPS] glycosyl acceptors.

Scheme 4:

Synthesis of QS saponin linker variants with glycosyl donor acting (a) as an electrophile in a traditional glycosylation or (b) as a nucleophile in a polarity-reversed coupling.

Immunological evaluation in mice of these saponins variants with the four-antigen cocktail above revealed striking differences in and modulation of adjuvant activity, despite the seemingly minor structural perturbations. For instance, whereas longer linker length in βethanolamide SQS-0-4-5-5 and β-carbamate SQS-0-5-5-5 abrogated activity, a more modest increase in linker distance in β-thioester SQS-0-13-5-5 resulted in potent adjuvant activity, with antibody responses comparable to QS-21 and the parent β-glycosyl ester SQS-0-0-5-5 (Fig. 7). This novel variant SQS-0-13-5-5 was essentially non-toxic at the 5 μg dose, but exhibited dose-limiting toxicity at the higher 20 μg dose. The stereoisomeric variants α-ester SQS-0-0-8-5 and α-carbamate SQS-0-5-8-5 were totally inactive; in contrast, the α-glycosyl amide SQS-0-6-8-5 elicited high antibody titers but caused considerable mouse weight loss at both 5 and 20 μg doses. Finally, βether SQS-0-12-5-5 and β-thioether SQS-0-14-5-5, with increased linker flexibility exhibited modest adjuvant activities, with only the 20 μg dose being moderately effective. In general, this marked decreased in activity observed with most linker variants indicate that the central glycosidic linkage is less tolerant to structural modifications than other domains of QS-21.

Fig. 7:

Immunological evaluation of central linkage variants (β-ethanolamide SQS-0-4-5-5, β-thioester SQS-0-13-5-5, α-ester SQS-0-0-8-5, α-amide SQS-0-6-8-5, β-ether SQS-0-12-5-5, and β-thioether SQS-0-14-5-5) at 5 μg (left panels) and 20 μg (right panels) dose. Adapted from Ref. [16]. Antibody titers against (a, b) KLH (IgG), (c, d) OVA (IgG), (e, f) MUC1 (IgG).

Remarkably, molecular dynamics simulations of these central linkage variants revealed distinct conformational features that correlated with the divergent adjuvant activities [16]. Thus, unlike the inactive analogs, e.g. α-ester SQS-0-0-8-5, which exhibited rather disorganized and less-ordered conformations around the central linker, the two most active variants, βester SQS-0-5-5-5 and β-thioester SQS-0-13-5-5, as well as QS-21-api, all adopted comparatively rigid conformations, in which the linear oligosaccharide domain was extended away from the center of the triterpene and the acyl chain was folded back in parallel over the triterpene (Fig. 8). These conformational preferences and the extent of folding were quantitatively characterized by measuring the torsional angles around the central linkage, which together with the qualitative analysis above, provided a positive correlation between these specific conformational features and adjuvant activity that raises the possibility of a preferred, folded active conformation.

Fig. 8:

Conformational ensembles and torsional angle distributions of the central glycosidic linkage derived from unrestrained molecular dynamics simulations of (a) QS-21-api and (b) characteristic saponin variants: inactive α-glycosyl ester SQS-0-0-8-5, and active β-glycosyl ester SQS-0-0-5-5 and β-glycosyl thioester SQS-0-0-13-5. Adapted from Ref. [4]. The above trends also hold for the rest of the linker variants.

These results provide a molecular rationale for the striking variation in adjuvant activity despite the relatively subtle structural modifications, and yield important insights into the conformational preferences of QS saponins that correlate with activity. Thus, saponin conformation appears to be important for adjuvant activity, and may contribute to proper biodistribution, subcellular localization, and/or target binding.

Development of acyl chain variants bearing fluorescent and iodinated probes for mechanistic studies

The previous trisaccharide amine precursor 14 was used to investigate the effect of the ionic charge on adjuvant activity with the synthesis of aminoacyl trisaccharide variant, SQS-0-0-5-11 (Scheme 5a), which incorporates a 6-aminohexanoic acyl chain with a positively charged terminal amine. In mice vaccinated with the MUC1–KLH conjugate antigen, SQS-0-0-5-11 elicited low antibody responses (Fig. 9), revealing that the positive charge in this variant attenuates adjuvant activity, as opposed to the effect observed with negatively charged dodecanedioic trisaccharide analog SQS-0-0-5-5, which retained potent adjuvant activity. Nonetheless, this amine-terminated saponin allowed late-stage, chemoselective installation of different reporter groups for the synthesis of saponin mechanistic probes. Thus, Gin and co-workers introduced fluorescent or radiolabeled tags into the QS-saponin scaffold for subsequent imaging and in vivo biodistribution studies. Commencing with SQS-0-0-5-11, we synthesized FITC-labeled SQS-0-0-5-12 [12] by reaction of the acyl chain domain amine with fluorescein isocyanate, and aryl iodide SQS-0-0-5-18 [17] by 4-iodobenzoylation of the terminal amine in SQS-0-0-5-11 (Scheme 5b).

Scheme 5:

(a) Synthesis of amine-terminated acyl chain variant SQS-0-0-5-11. (b) Synthesis of functionalized acyl chain variants bearing fluorescent (SQS-0-0-5-12) and iodinated (SQS-0-0-5-18) tags.

Fig. 9:

Immunological evaluation of functionalized acyl chain variants SQS-0-0-5-11 and fluorescent SQS-0-0-5-12 (10 μg saponin) with MUC1–KLH (2.5 μg) vaccine. Adapted from Ref. [12]. Antibody titers against (a) KLH (IgG), (b) MUC1 (IgG). (c) Toxicity assessment based on median weight loss over 1 week after the first vaccine injection.

In mouse vaccinations, these variants (10 μg SQS-0-0-5-12, 20 μg SQS-0-0-5-18) elicited comparable antibody responses to QS-21 and exhibited considerably lower weight loss (3.5% and 1.5% on day 1, 1% and 1.5% on day 2, respectively) (Figs. 9 and 10).

Fig. 10:

Immunological evaluation of aryl iodide acyl chain variant SQS-0-0-5-18 (5, 10, and 20 μg saponin, respectively) with OVA (20 μg) and MUC1–KLH (2.5 μg) vaccine. Antibody titers against (a) KLH (IgG), (b) OVA (IgG), and (c) MUC1 (IgG). (d) Toxicity assessment based on median mouse weight change over a week after first vaccine injection.

With both saponin variants established as potent adjuvants with low toxicity, the radiolabeled congener of SQS-0-0-5-18 was generated by radioiodination of the corresponding aryl tributylstannane [17], and in vivo biodistribution as well as fluorescence imaging studies were performed in mice coadministered with OVA (20 μg). Initial efforts to utilize [124I]-SQS-0-0-5-18 in positron emission tomography (PET) imaging were complicated by high signal retention at the injection site, which made quantitation difficult at nearby sites such as the lymph nodes. Instead, “cut-and-count” studies were conducted with [131I]-SQS-0-0-5-18 and a structurally related inactive variant, [131I]-SQS-0-3-7-18, devoid of the linear oligosaccharide. For the active saponin, preferential recovery of radioactivity compared to the inactive variant was observed at the injection site (17-fold higher) and the nearest lymph nodes (24-fold higher) at 24 h post-injection (Fig. 11). In mice treated with [131I]-SQS-0-0-5-18, these two sites showed also high levels of radioactivity at 72 and 96 h post-injection, in contrast to other tissues and organs where large differences were initially observed. However, for the inactive variant [131I]-SQS-0-3-7-18 radioactivity cleared from all tissues at the later time points. These data indicate that the active saponin is retained at the injection site and accumulates in the lymph nodes, but the inactive variant does not. Similar trends were observed with the radiolabeled congeners of another pair of active/attenuated variants lacking the branched trisaccharide and incorporating quillaic acid and oleanolic acid triterpenes, respectively (vide infra), which overall provides a positive correlation between this biodistribution pattern and adjuvant activity.

Fig. 11:

Twenty four hours in vivo biodistribution of radiolabeled saponins (25 μCi) SQS-0-0-5-18 (active) and SQS-0-3-7-18 (inactive) in mice co-injected with 20 μg OVA and 20 μg of the corresponding cold saponin. Adapted from Ref. [4]. Statistical significance of the difference in recovery of radioactivity (%ID/g) only shown for lymph nodes and injection site.

As other QS-based adjuvant mixtures have been shown to affect antigen biodistribution [18], [19], the biodistribution of [131I]OVA in the presence and absence of active aryl iodide SQS-0-0-5-18 was also explored. Unfortunately, elevated radioiodine uptake was observed in the thyroid tissue, suggesting rapid deiodination of [131I]OVA. Thus, in vivo mouse fluorescence imaging studies with FITC-labeled active variant SQS-0-0-5-12 and Alexa-647-conjugated OVA (OVA-A647) were performed as an alternative strategy. At 24 h post-injection, OVA-A647 colocalized at the lymph nodes and the injection site when coadministered with active SQS-0-0-5-12 (Fig. 12a,b); however, in mice coinjected with the inactive variant SQS-0-0-5-11, OVA-A647 was only observed at the injection site and did not accumulate in the lymph nodes. The active fluorescent saponin SQS-0-0-5-12 was retained at both sites, consistent with the previous biodistribution results, and sublocalized to the cortex of the draining inguinal node, as assessed by immunohistochemistry of dissected lymph nodes. Flow cytometric analysis demonstrated that internalization of SQS-0-0-5-12 was specific to dendritic cells within the lymph nodes, in agreement with subcellular localization studies in immature dendritic cells by confocal microscopy [12] that showed internalization of active SQS-0-0-5-12 to a discrete cellular compartment, but not of related attenuated saponins bearing BODIPY (SQS-0-0-5-17) and Cascade Blue (SQS-0-0-5-15) fluorophores, or other fluorescent negative controls (Fig. 12c). Collectively, these studies suggest that active variants help direct antigen-presenting cells trafficking OVA to the lymph nodes for presentation of the antigen to the immune system, and provide preliminary mechanistic understanding of how these saponin adjuvants may potentiate the immune response.

Fig. 12:

Twenty four hours in vivo fluorescence imaging studies in mice injected with active fluorescein-conjugated saponin SQS-0-0-5-12 or inactive non-fluorescent variant SQS-0-0-5-11 (10 μg each), as well as OVA-Alexa647 (20 μg). Adapted from Ref. [4]. (a) Imaging of dissected lymph nodes (mice injected in left nodes, right nodes are negative controls). (b) Whole-body imaging at the injection site. (c) Subcellular localization of fluorescein-labeled active saponin SQS-0-0-5-12 and inactive BODIPY (SQS-0-0-5-17) and Cascade Blue (SQS-0-0-5-15) saponins in immature dendritic cells imaged by confocal microscopy (upper panel); the corresponding glycine methyl esters (GlyOMe) labeled with each corresponding fluorophore were administered as negative controls (lower panel).

Carbohydrate variants via deletion of the QS branched trisaccharide

In our next investigations, we prepared a truncated saponin variant, SQS-1-0-5-18, derived from SQS-0-0-5-18 but devoid of the complete branched trisaccharide, to determine if this domain, as a whole, is required for adjuvant activity. Starting from quillaic acid, isolated and purified in gram quantities from commercial QS-extracts [7], [20], selective silylation of the hydroxyl groups followed by glycosylation of the triterpene C28-carboxylate with the aza-trisaccharide imidate 11 gave the triterpene–linear trisaccharide azide intermediate 36 (Scheme 6). The remaining steps towards SQS-1-0-5-18 involved reduction of the azide, acylation with the 6-aminohexanoic acyl chain, global deprotection via hydrogenolysis and acid hydrolysis, and late-stage chemoselective acylation of the side chain terminal amine with the aryl iodide succinimidyl ester 33.

Scheme 6:

Synthesis of truncated aryl iodide acyl variant SQS-1-0-5-18 devoid of the branched trisaccharide.

Remarkably, in mice vaccinated with MUC1–KLH and OVA, this truncated variant SQS-1-0-5-18 elicited antibody responses comparable to QS-21 and the branched trisaccharide-containing SQS-0-0-5-18, although at a 50 μg dose (Fig. 13). Moreover, SQS-1-0-5-18 was barely toxic, causing no mouse weight loss at 20 μg (+1.8%, day 1) and only −1% (day 1) at 50 μg [17]. These results revealed that deletion of the branched trisaccharide does not impair adjuvant activity and also lowers toxicity, yielding a greatly simplified and potent QS-21 variant that is more synthetically accessible and less toxic than the natural product.

Fig. 13:

Immunological evaluation of truncated aryl iodide variant SQS-1-0-5-18 (10 and 20 μg saponin) with OVA (20 μg) and MUC1–KLH (2.5 μg) vaccine. Antibody titers against (a) KLH (IgG), (b) OVA (IgG), and (c) MUC1 (IgG). (d) Toxicity assessment based on median mouse weight change over a week after first vaccination.

Triterpene variants via specific modifications at C4-aldehyde and C16hydroxyl group

The discovery that the branched trisaccharide is not necessary for activity allowed us to carry out detailed structure-activity relationship studies on the triterpene, particularly at its C4-aldehyde substituent and C16-hydroxyl group, by synthesizing several saponin variants with targeted modifications at these positions using other readily available triterpenes as starting materials (Fig. 14) [17]. The potential role of the aldehyde in the adjuvant’s mode of action is only speculative, but it has been proposed to participate in Schiff base formation with a presumed Tcell surface receptor target [21], [22] based on the finding that QS-21 derivatives that compromise the C4-aldehyde substituent by reductive amination with exogenous amines exhibited significantly attenuated adjuvant activity [21]. However, these derivatives not only were devoid of the aldehyde (as intended), but also incorporated a positively-charged amino group that may be the origin of its decreased activity, analogously to the inactive acyl chain terminal amine variant SQS-0-0-5-11 [12]. Thus, we introduced more subtle structural changes at the triterpene C4 position with the synthesis of caullophylogenin variant SQS-1-11-5-18 and echinocystic acid variant SQS-1-8-5-18, in which the C4-aldehyde in quillaic acid variant SQS-1-0-5-18 was replaced with a hydroxymethyl and a methyl group, respectively. Moreover, the related gypsogenin SQS-1-9-5-18, hederagenin SQS-1-10-5-18, and oleanolic acid SQS-1-7-5-18 variants, in which the C16-hydroxyl group of the respective triterpene cores was orthogonally replaced by a proton, were also pursued (Fig. 14). The preparation of these new triterpene variants started from commercially available echinocystic acid, hederagenin, and oleanolic acid triterpenes, following a similar synthetic sequence to that of quillaic acid variant SQS-1-0-5-18, with the exception of caullophylogenin variant SQS-1-11-5-18, which was prepared from an advanced protected intermediate en route to SQS-1-0-5-18.

Fig. 14:

Structures of triterpene variants via molecular editing at C4 (aldehyde) and C16 (hydroxyl group) positions of the triterpene.

In mice vaccinated with the four-antigen formulation (GD3–KLH, MUC1–KLH, and OVA), both variants retaining the C16-hydroxyl group, echinocystic acid variant SQS-1-8-5-18 and caullophylogenin variant SQS-1-11-5-18, showed potent adjuvant activity, higher than or similar to QS-21, respectively (Fig. 15). Remarkably, while these variants contain the C16-hydroxyl group, neither of them bear the C-4 aldehyde substituent, which casts doubt on the Schiff-base hypothesis, at least within these truncated variants. On the other hand, gypsogenin SQS-1-9-5-18, hederagenin SQS-1-10-5-18 and oleanolic acid SQS-1-7-5-18 (not shown) variants, all lacking the C16-hydroxyl, exhibited attenuated adjuvant activities across all antigens, suggesting an important role for this functionality in adjuvant activity. Notably, none of these variants showed any appreciable toxicity as assessed by mouse weight loss (Fig. 15f). These results indicate that potent adjuvant activity requires the C16-hydroxyl group but not the C4-aldehyde substituent, which is consistent with the identification of other saponin adjuvants lacking the C4-aldehyde and bearing the C16-hydroxyl [23], [24], [25]. The discovery of this potent, non-toxic and synthetically accessible (23 steps total) echinocystic acid variant SQS-1-8-5-18 provides an attractive lead for further advancement and mechanistic studies in the future.

Fig. 15:

Immunological evaluation of triterpene variants SQS-1-8-5-18, SQS-1-9-5-18, SQS-1-10-5-18, and SQS-1-11-5-18 (50 μg saponin) with OVA (20 μg), MUC1–KLH (2.5 μg) and GD3–KLH (5 μg) vaccine. Antibody titers against (a) KLH (IgG), (b) OVA (IgG), (c) MUC1 (IgG), (d) GD3 (IgM), and (e) GD3 (IgG). (f) Toxicity assessment based on median mouse weight change over a week after first vaccination.

Saponin–tucaresol conjugates to explore the effect of aldehyde-based adjuvant combinations on adjuvant activity

The previous hypothesis that QS-21 may interact with putative T-cell surface receptors via Schiff base formation through its C4-aldehyde substituent [21], [22] had its origin on the identification of aldehyde-containing adjuvants, e.g. tucaresol, that react forming imines with amino groups on the surface of T cells, leading to T-cell costimulation and activation, and Th1 immune responses [26], [27], [28]. On this basis, and given the increasing interest in using combinations of adjuvants to achieve optimal vaccine efficacy [29], we next designed and synthesized novel saponin−tucaresol conjugates, i.e. branched-trisaccharide containing SQS-0-0-5-19 and its truncated congener SQS-1-0-5-19, to investigate if the aldehyde-based activity of tucaresol might further enhance the adjuvant activity of QS saponin variants [30].

These new variants were synthesized by chemoselective acylation with tucaresol succinimidyl ester at the acyl chain terminal amine of the corresponding saponin precursors (Scheme 7), and were evaluated in mice together with equimolar combinations of related saponin variants (SQS-0-0-5-18 and SQS-1-0-5-18) and tucaresol. While the saponin–tucaresol conjugates retained potent adjuvant activity, similar to their aryl iodide saponin congeners, the presence of tucaresol, either in combination (with SQS-0-0-5-18 or SQS-1-0-5-18), or covalently incorporated into the acyl chain, as in SQS-0-0-5-19 and SQS-1-0-5-19, did not significantly enhance the adjuvant activity of our QS saponins [30].

Scheme 7:

(a) Synthesis of saponin–tucaresol conjugate SQS-0-0-5-19 incorporating the branched trisaccharide. (b) Synthesis of truncated saponin–tucaresol conjugate SQS-1-0-5-19 devoid of the left-hand carbohydrate.

Linear trisaccharide variants via versatile divergent synthesis

In considering further specific molecular modifications of the QS saponin scaffold, our goal was to provide access to saponin variants that can be prepared efficiently in a streamlined and versatile synthetic route. Thus, we developed a novel divergent strategy and synthesized three saponin variants with variation of individual sugars and connectivities within the linear trisaccharide (Fig. 16) that were prepared in less steps than the previous lead compound SQS-1-0-5-18 [31]. The first analog, rhamnose–rhamnose variant SQS-1-0-10-18 (22 total steps), incorporated a rather conservative modification with a single sugar substitution at the third, terminal residue, and was synthesized following the same linear, convergent approach used for the parent saponin SQS-1-0-5-18 bearing the original linear trisaccharide. More profound structural changes took the form of lactose variant SQS-1-0-11-18 (16 total steps), in which the entire terminal disaccharide was replaced by commercially available lactose, and regioisomeric variant SQS-1-0-12-18 (19 total steps), in which the original terminal disaccharide and the acyl chain exchanged positions within a more readily accessible 2-azidogalactose bridging sugar. For the synthesis of the latter two variants, a new synthetic approach was designed consisting of stepwise glycosylation of the triterpene with a bridging monosaccharide residue followed by installation of the desired terminal disaccharides, enabling expedited preparation of the lactose variant SQS-1-0-11-18 in only 16 steps. Moreover, through its versatile triterpene–monosaccharide late-stage diversification point, this divergent strategy sets the stage for more efficient access to a range of linear trisaccharide variants for the rapid identification of new, potentially improved saponin adjuvants that are synthetically accessible and easily scalable for further clinical development in vaccines.

Fig. 16:

(a) Versatile divergent strategy towards linear oligosaccharide saponin variants. (b) Structures of synthetically streamlined linear oligosaccharide variants with modified, readily accessible individual sugars.