Development of Improved Vaccine Adjuvants Based on the Saponin Natural Product QS-21 through Chemical Synthesis.

Vaccines based on molecular subunit antigens are increasingly being investigated due to their improved safety and more precise targeting compared to classical whole-pathogen vaccines. However, subunit vaccines are inherently less immunogenic; thus, coadministration of an adjuvant to increase the immunogenicity of the antigen is often necessary to elicit a potent immune response. QS-21, an immunostimulatory saponin natural product, has been used as an adjuvant in conjunction with various vaccines in numerous clinical trials, but suffers from several inherent liabilities, including scarcity, chemical instability, and dose-limiting toxicity. Moreover, little is known about its mechanism of action. Over a decade-long effort, beginning at the University of Illinois at Urbana-Champaign and continuing at the Memorial Sloan Kettering Cancer Center (MSKCC), the group of Prof. David Y. Gin accomplished the total synthesis of QS-21 and developed a practical semisynthetic approach to novel variants that overcome the liabilities of the natural product. First, semisynthetic QS-21 variants were designed with stable amide linkages in the acyl chain domain that exhibited comparable in vivo adjuvant activity and lower toxicity than the natural product. Further modifications in the acyl chain domain and truncation of the linear tetrasaccharide domain led to identification of a trisaccharide variant with a simple carboxylic acid side chain that retained potent adjuvant activity, albeit with reemergence of toxicity. Conversely, an acyl chain analogue terminating in a free amine was inactive but enabled chemoselective functionalization with radiolabeled and fluorescent tags, yielding adjuvant-active saponin probes that, unlike inactive congeners, accumulated in the lymph nodes in vaccinated mice and internalized into dendritic cells. Subtle variations in length, stereochemistry, and conformational flexibility around the central glycosidic linkage provided QS-21 variants with adjuvant activities that correlated with specific conformations found in molecular dynamics simulations. Notably, deletion of the entire branched trisaccharide domain afforded potent, truncated saponin variants with negligible toxicity and improved synthetic access, facilitating subsequent investigation of the triterpene core. The triterpene C4-aldehyde substituent, previously proposed to be important for QS-21 adjuvant activity, proved to be dispensable in these truncated saponin variants, while the presence of the C16 hydroxyl group enhanced activity. Novel adjuvant conjugates incorporating the small-molecule immunopotentiator tucaresol at the acyl chain terminus afforded adjuvant-active variants but without significant synergistic enhancement of activity. Finally, a new divergent synthetic approach was developed to provide versatile and streamlined access to additional linear oligosaccharide domain variants with modified sugars and regiochemistries, opening the door to the rapid generation of diverse, synthetically accessible analogues. In this Account, we summarize these multidisciplinary studies at the interface of chemistry, immunology, and medicine, which have provided critical information on the structure-activity relationships (SAR) of this Quillaja saponin class; access to novel, potent, nontoxic adjuvants for use in subunit vaccines; and a powerful platform for investigations into the mechanisms of saponin immunopotentiation.

Introduction

Modern subunit vaccines comprising homogeneous molecular antigens are being developed to prevent and treat a variety of human diseases.[1] While these subunit vaccines allow more precise targeting and improved safety compared to classical whole-pathogen vaccines, they are poorly immunogenic and must be coadministered with an adjuvant to elicit a potent immune response. However, few adjuvants are of sufficient potency and acceptable toxicity for clinical use.[2] Aluminum salts, either alone (alum) or in proprietary mixtures (AS04),[3] and oil-in-water emulsions containing squalene (MF59, AS03)[4] have been used as adjuvants in a number of vaccines, but have relatively low potency and significant side effects, respectively. Therefore, there remains a great need for novel adjuvants to enable implementation of subunit vaccines.[5]

The saponin natural product QS-21 is one of the most potent adjuvants known. Isolated from Quillaja saponaria tree bark, it is composed of four structural domains: a branched trisaccharide, a quillaic acid triterpene, a bridging linear tetrasaccharide, and a pseudodimeric acyl chain (Figure ). QS-21 stimulates both antibody-based humoral immune responses (Th2) and cellular immunity (Th1), including production of antigen-specific cytotoxic T-lymphocytes.[6] Vaccines containing QS-21, either alone in purified form or as a major component of adjuvant mixtures (e.g., Quil A, ISCOMs, ISCOMATRIX, AS01, AS02),[7] have been investigated in clinical trials for cancers (melanoma, sarcoma, breast, prostate, ovarian, lung),[6] infectious diseases (hepatitis, HIV, malaria, tuberculosis)[8] and Alzheimer’s disease.[9]

Figure 1

Structure of QS-21 and four structural domains.

Despite its remarkable potency and extensive clinical investigation, QS-21 suffers from several limitations. First, access to homogeneous QS-21 is limited due to an exceedingly low-yielding isolation and heterogeneity of crude extracts from Quillaja saponaria.[10] Indeed, QS-21 is not a single molecule but an ≈2:1 mixture of two isomeric constituents, QS-21-Api (1) and QS-21-Xyl (2),[11] that differ at the terminal sugar of the linear tetrasaccharide domain. Second, QS-21 is associated with clinical toxicity including swelling and erythema at the injection site, and systemic flu-like symptoms.[6] Third, QS-21 undergoes spontaneous hydrolysis of the acyl chain domain ester linkages,[12] producing adjuvant-inactive and hemolytic byproducts, complicating formulation and storage. Finally, the mechanisms of action of QS-21 are poorly understood,[13] hindering rational design of improved variants and optimal matching of adjuvants with vaccine antigens based on desired immunological end points. Along these lines, it is generally agreed that QS-21 is not a ligand for Toll-like receptors that stimulate innate immunity,[14] and does not operate by a depot effect, whereby the adjuvant increases the lifetime of the antigen and its presentation to the immune system.[15] It has been hypothesized that QS-21 may facilitate antigen uptake by antigen-presenting cells by binding to cell-surface lectins through its carbohydrate domains, leading to production of specific cytokines that activate cellular and/or humoral responses.[16] The triterpene domain C4-aldehyde substituent has been proposed to form a Schiff base with amino groups on T-cell surface receptors, providing costimulation for T-cell activation.[16,17] QS-21 has been shown to activate the NLRP3 inflammasome in vitro, but this activation decreased the effects of a HIV gp120/QS-21 vaccine in vivo.[18]

The inherent liabilities of QS-21 highlight the need for improved analogues. Due to the structural complexity of and difficulty in derivatizing the natural product, only a few QS-21 analogues have been reported previously.[17,19] Of note, GPI-0100, prepared from QS-21-containing bark extracts by saponification of the acyl chain domain and installation of a dodecylamide in the branched trisaccharide domain (Figure ),[19] retained potent adjuvant activity, but required 20-fold higher doses than QS-21 for optimal efficacy in a breast cancer clinical trial, resulting in significant hepatotoxicity.[6] Together with its heterogeneity, this precluded further clinical use of GPI-0100. Recently, this heterogeneity was addressed by Wang and co-workers, who synthesized the two proposed immunoactive components of GPI-0100 and confirmed the preclinical adjuvant activity of xylose isomer 3b.[20]

Figure 2

Structures and semisynthetic modifications of main active components of GPI-0100.

Herein, we describe efforts in the laboratory of Prof. David Y. Gin to develop improved variants of QS-21 that overcome the inherent limitations of the natural product. Through extensive structure–activity relationship (SAR) studies spanning all four domains of QS-21, novel semisynthetic variants with potent adjuvant activity and low toxicity in vivo were identified. Moreover, development of structurally related pairs of adjuvant-active and -attenuated saponins enabled biodistribution and fluorescence imaging studies that provided insights into the mechanisms of saponin immunopotentiation.

Development of Improved Synthetic Saponin Adjuvants

Total Synthesis of QS-21 and Semisynthetic Approach to QS-21 Variants

Only a few efforts toward the synthesis of Quillaja saponins have been reported.[20−23] Gin and co-workers accomplished the only total syntheses of QS-21-Api (1)[24,25] and QS-21-Xyl[26] (2) (Figure ), in 76 steps, providing access to homogeneous material. In a mouse vaccination model using the weakly immunogenic glycolipid GD3 (melanoma, sarcoma, neuroblastoma antigen) conjugated to the highly immunogenic KLH carrier protein (GD3-KLH), the synthetic compounds and natural QS-21 exhibited similar adjuvant activities.[27] In studies toward the related saponin QS-7 (6),[28] Deng et al. developed a semisynthetic approach involving isolation of the branched trisaccharide–triterpene portion (prosapogenin, [QPS]-CO2H, 4) from commercial Quil A extracts followed by selective protection and functionalization (Scheme ). This methodology was then applied to a 100 mg scale synthesis of QS-21 (56 steps) for a melanoma vaccine clinical trial.[6] This semisynthetic technology also opened the door to the synthesis of diverse QS-21 variants to investigate SAR and identify improved analogues.

Scheme 1

Semisynthesis of QS-7 from Partially Purified Quillaja saponaria Extract Quil A

Development of Stable, Simplified Amide-Based Acyl Chain Variants

Adams et al. first addressed the chemical instability of QS-21 by replacing the hydrolytically labile esters in the acyl chain domain with amide linkages.[29] This required replacement of the bridging fucose residue in the linear tetrasaccharide domain with 4-amino-4-deoxygalactose. Thus, protected 4-azido-4-deoxygalactose 10 was synthesized from d-glucal (Scheme a), then coupled with trisaccharide hemiacetal 17 by dehydrative glycosylation[30] (Scheme b). The modified tetrasaccharide 20 was coupled to the protected Quillaja prosapogenin ([PQPS]-CO2H, 5)[28] via Schmidt glycosylation to provide trisaccharide–triterpene–tetrasaccharide construct 21 (Scheme a). The azide was reduced to the requisite amine 22 under mild phenylselenol conditions providing an advanced intermediate on which to append diverse acyl chains. The first and most conservative structural variant, 27 (SQS-0101), incorporated amide linkages within the elaborate QS-21 side chain, whereas the other two variants, 28 (SQS-0102) and 29 (SQS-0103), incorporated simplified acyl chains (Scheme ).

Scheme 2

Synthesis of Modified Linear Tetrasaccharide Domain via (a) Preparation of Protected 4-Azido-4-deoxygalactose 10 and (b) Carbohydrate Assembly

Scheme 3

Synthesis of Acyl Chain Domain Variants with Amide Linkages by (a) Acylation of Amine 22 with (b–d) the Corresponding Acyl Chain Analogues

The immunopotentiating activities of these saponin variants were evaluated in collaboration with Philip Livingston and Govind Ragupathi at MSKCC in a preclinical mouse vaccination model. Groups of five female C57BL/6J mice were injected subcutaneously with the GD3-KLH conjugate vaccine and the saponin of interest on days 0, 7, and 14, followed by a booster on day 65. Antibody responses to both GD3 and KLH, including both IgM (low-affinity antibodies initially produced by B cells in early, short-term, T-cell independent response) and IgG (high-affinity antibodies produced after T-helper cell-activated affinity maturation and class switching in persistent, long-term response) for GD3, were assessed by ELISA on day 72. Bisamide 27 (SQS-0101) elicited responses comparable to or higher than natural QS-21, with considerably less toxicity as assessed by median weight loss at days 1, 2, 3, and 7 after the first vaccination (Figure a–c). Arabinosyl aliphatic amide 28 (SQS-0102) and dodecanoic amide 29 (SQS-0103) also elicited similar antibody responses but the former was associated with marked weight loss while the latter was nontoxic (Figure d).[29] Subtyping of anti-GD3 IgG antibodies for the IgG1 and IgG2 subtypes, indicative of T-helper type 2 (Th2) humoral and T-helper type 1 (Th1) cell-mediated immune responses, respectively, revealed predominance of IgG2b for all three saponin variants, similar to natural QS-21. These antibodies were also able to effect binding and lysis of a GD3-positive melanoma cell line, again consistent with the activity of natural QS-21.

Figure 3

Immunological evaluation of acyl chain domain variants with amide linkages. (a–c) Antibody titers after three vaccinations and booster and (d) median weight loss after first vaccination. Mice vaccinated with GD3-KLH (10 μg) and saponin (10 μg); horizontal bars indicate median titers; statistical significance compared to no-adjuvant control assessed using two-tailed unpaired Student’s t test with 95% confidence interval, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. NQS-21 = natural QS-21.

These results demonstrated the feasibility of developing simplified saponin adjuvants with retained or enhanced activity and attenuated toxicity. However, these three variants were still suboptimal, with 27 (SQS-0101) requiring a 54-step synthesis, 28 (SQS-0102) showing increased toxicity, and 29 (SQS-0103) exhibiting poor aqueous solubility. Thus, Chea et al. investigated additional acyl chain domain variants (Figure ).[31] Dodecanedioic acyl variant 30 (SQS-0-0-4-5) was designed to improve water solubility while cholestanyl variant 31 (SQS-0-0-4-6) was devised to probe a hypothesis that interaction of QS-21 with cell membrane-bound cholesterol may be important in initiating adjuvant activity. Mouse vaccinations with GD3-KLH and the immunogenic peptide MUC1 (prostate and breast cancer antigen, nonglycosylated tandem repeat) conjugated to KLH (MUC1-KLH) revealed that 30 elicited antibody responses comparable to natural and synthetic QS-21 across all antigens, while 31 was less active (Figure a–c). Notably, no weight loss was observed with either saponin (Figure d). Thus, dodecanedioic acyl variant 30 was selected as a lead structure for additional SAR studies.

Figure 4

Structures of additional acyl chain domain variants. Four-number SQS (synthetic Quillaja saponin) codes designate structural variants in each of the four corresponding structural domains of QS-21, left to right, with 0 assigned to the natural product structure.

Figure 5

Immunological evaluation of additional acyl chain domain variants (5 μg GD3-KLH, 5 μg MUC1-KLH, 10 μg saponin). SQS-21 = synthetic QS-21 (2:1 mixture of 1 and 2).

Stepwise Truncation of Linear Tetrasaccharide Domain

Chea et al. next explored stepwise truncation of the linear tetrasaccharide domain using trisaccharide variant 32 (SQS-0-0-5-5), disaccharide variant 33 (SQS-0-0-6-5), and monosaccharide variant 34 (SQS-0-0-9-5) (Figure a).[31] The trisaccharide imidate 40 lacking the fourth (apiose/xylose) residue was synthesized in 16 steps, coupled to 5 ([PQPS]-CO2H), and the azide was reduced to afford amine 41 (Figure b). Acylation with a dodecanedioic acyl chain followed by global deprotection provided trisaccharide variant 32. The further-truncated variants 33 and 34 were prepared analogously (not shown).

Figure 6

(a) Structures of progressively truncated linear tetrasaccharide domain variants. (b) Representative synthesis of linear trisaccharide variant 32.

Immunological evaluation of these saponins was carried out with a four-component vaccine with GD3-KLH, MUC1-KLH, and ovalbumin (OVA), a reliable immunogen that induces antibody and T-cell responses. Trisaccharide variant 32 (SQS-0-0-5-5) generated antibody titers comparable to QS-21, while disaccharide 33 (SQS-0-0-6-5) and monosaccharide 34 (SQS-0-0-9-5) elicited progressively attenuated responses overall (Figure a–d). All three analogues exhibited acute toxicity, with one to two mice dying in each cohort, although weight loss decreased with truncation (Figure e). Despite this reemergence of toxicity, trisaccharide variant 32 could be synthesized more efficiently (24 steps) than the parent tetrasaccharide variant 30 (SQS-0-0-4-5) (36 steps).

Figure 7

Immunological evaluation of truncated linear tetrasaccharide domain variants (5 μg GD3-KLH, 2.5 μg MUC1-KLH, 20 μg OVA, 20 μg saponin).

Development of Functionalized Acyl Chain Domain Variants

Chea et al. next investigated the effects of ionic charge in the acyl chain domain upon adjuvant activity with cationic 6-aminohexanoyl variant 42 (SQS-0-0-5-11) (Scheme ).[31] This amine-containing saponin also enabled late-stage, chemoselective functionalization with reporter groups, as in fluorescent saponin 43 (SQS-0-0-5-12). Immunological evaluation with MUC1-KLH revealed that, while the negative charge of the dodecanedioic acyl variant 32 (SQS-0-0-5-5) was accommodated, the positive charge in the 6-aminohexanoyl variant 42 resulted in decreased activity (Figure ).[31] Strikingly, acylation of the acyl chain domain amine with fluorescein isothiocyanate restored activity in 43. In contrast, installation of other fluorophores, such as BODIPY (SQS-0-0-5-17) and Cascade Blue (SQS-0-0-5-15), resulted in attenuated adjuvant activity (not shown).

Scheme 4

Synthesis of Functionalized Acyl Chain Domain Variants

Figure 8

Immunological evaluation of functionalized acyl chain domain variants (2.5 μg MUC1-KLH, 10 μg saponin).

Modification of Length, Stereochemistry, and Flexibility in Central Glycosidic Linkage

Building upon lead compound 32 (SQS-0-0-5-5), Walkowicz et al. investigated the role of the central glycosidic linkage.[32] These analogues probed the effects of linker length (β-ethanolamide 44 [SQS-0-4-5-5], β-carbamate 45 [SQS-0-5-5-5], β-thioester 46 [SQS-0-13-5-5]), stereochemistry (α-carbamate 47 [SQS-0-5-8-5], α-ester 48 [SQS-0-0-8-5], α-amide 49 [SQS-0-6-8-5]), and flexibility (β-ether 50 [SQS-0-12-5-5], β-thioether 51 [SQS-0-14-5-5]) (Figure ) on adjuvant activity and toxicity. The syntheses were accomplished using two complementary strategies in which the glycosyl donor (Scheme a) was used either as a nucleophile in a polarity-reversed coupling (Scheme b) or as an electrophile in a traditional glycosylation (Scheme c). Reduction of the resulting azides to the corresponding amines, acylation with dodecanedioic acid monobenzyl ester, and global deprotection provided the saponin variants (not shown).

Figure 9

Structures of central glycosidic linkage variants.

Scheme 5

Synthesis of Central Glycosidic Linkage Variants via (a) Preparation of Linear Trisaccharide Glycosyl Donors and (b) Installation on [PQPS]-CO2H Core (5) Using Glycosyl Acceptor as an Electrophile or (c) Nucleophile

In mice vaccinated with GD3-KLH, MUC1-KLH, and OVA, these central linker variants exhibited distinct adjuvant activities, despite their relatively conservative modifications (Figure ). Increased linker distance in β-ethanolamide 44 (SQS-0-4-5-5) and β-carbamate 45 (SQS-0-5-5-5) led to loss of activity, while a smaller increase in β-thioester 46 (SQS-0-13-5-5) was accommodated. Mice treated with thioester 46 exhibited no weight loss at a 5 μg dose, but dose-limiting toxicity at a 20 μg dose (not shown). Stereochemical inversion at the anomeric position within the glycosidic linkage abrogated adjuvant activity in α-carbamate 47 (SQS-0-5-8-5) and α-ester 48 (SQS-0-0-8-5), while α-amide 49 (SQS-0-6-8-5) retained activity but was toxic at both 5 and 20 μg doses (not shown). Increasing linker flexibility in β-ether 50 (SQS-0-12-5-5) and β-thioether 51 (SQS-0-14-5-5) resulted in attenuated activity.

Figure 10

Immunological evaluation of central linkage variants (5 μg GD3-KLH, 2.5 μg MUC1-KLH, 20 μg OVA, 5 or 20 μg saponin). Anti-GD3 responses and data for inactive α/β-carbamates (45, 47) not shown.

Strikingly, these differences in adjuvant activity correlated with specific conformational preferences identified in molecular dynamics simulations of the saponin variants.[32] The most potent analogues, β-ester 32 (SQS-0-0-5-5) and β-thioester 46 (SQS-0-13-5-5), exhibited relatively rigid conformations, similar to QS-21-Api (1), with the acyl chain folded back over the triterpene, suggesting a preferred, active conformation (Figure ). In contrast, inactive variants such as α-ester 48 (SQS-0-0-8-5) adopted distinct and less-ordered conformations. These conformational preferences were quantitatively characterized by measuring dihedral angles within the central linkage, which were similar and tightly clustered in the active saponins, in contrast to the inactive variants. This provided a molecular rationale for why subtle linker modifications lead to major changes in overall saponin conformation. Collectively, these results suggest an important role for saponin conformation in adjuvant activity, perhaps contributing to proper distribution, subcellular localization, and/or molecular recognition by a putative cellular target.

Figure 11

Conformational ensembles and central glycosidic linkage dihedral angle distributions (arbitrary axis zero-points) from unrestrained molecular dynamics simulations of (a) QS-21-Api (1) and (b) representative saponin variants, distinguishing active and inactive saponins.

Development of Aryl Iodide Acyl Chain Variants and Deletion of Branched Trisaccharide Domain

With a view toward developing radiolabeled probes (subsection ), Fernández-Tejada et al. synthesized aryl iodide variant 67 (SQS-0-0-5-18) by 4-iodobenzoylation of the acyl chain domain amine in 42 (Scheme a).[33] A truncated congener, 72 (SQS-1-0-5-18), lacking the entire left-hand branched trisaccharide domain, was also prepared from quillaic acid, which was readily isolated and purified in gram scale from commercial Quil A (Scheme b).[25] Selective silylation followed by glycosylation with trisaccharide imidate 40 and reduction of the azide provided the triterpene-linear trisaccharide amine scaffold 70. Acylation with the 6-aminohexanoic acyl chain, global deprotection, and chemoselective 4-iodobenzoylation of the acyl chain domain amine in 71 then afforded truncated variant 72.

Scheme 6

Synthesis of Aryl Iodide Acyl Chain Domain Variants

Immunological evaluation with MUC1-KLH and OVA revealed that aryl iodide 67 (SQS-0-0-5-18) elicited antibody responses comparable to QS-21, with considerably lower weight loss (Figure ). Strikingly, the truncated variant 72 (SQS-1-0-5-18) lacking the branched trisaccharide domain elicited antibody responses comparable to QS-21 and 67, albeit at a higher dose. Moreover, mice treated with truncated variant 72 exhibited no weight loss at a 20 μg dose and only 1% weight loss (on day 1) at 50 μg.[33] Discovery that the branched trisaccharide domain is not required for adjuvant activity, and that its deletion reduces toxicity, provided a major structural simplification and an improved activity/toxicity profile compared to QS-21.

Figure 12

Immunological evaluation of aryl iodide acyl chain domain variants (2.5 μg MUC1-KLH, 20 μg OVA, 20 or 50 μg saponin).

Molecular Editing of Triterpene Domain

The dispensability of the branched trisaccharide domain for adjuvant activity facilitated access to additional saponin variants with specific modifications in the triterpene domain.[33] The triterpene C4-aldehyde substituent has been proposed to be important for activity based on the reduced adjuvant activity of QS-21 derivatives in which this aldehyde is replaced with an amine.[17] However, in addition to removing the aldehyde, this modification introduces a positive charge that could interfere with saponin biodistribution, subcellular localization, or noncovalent target binding (cf. inactive acyl chain domain amine variant 42 [SQS-0-0-5-11]).[31] Thus, Fernández-Tejada et al. pursued more conservative structural modifications, replacing the C4-aldehyde substituent in quillaic acid variant 72 (SQS-1-0-5-18) with a hydroxymethyl group in caullophylogenin variant 73 (SQS-1-11-5-18) or a methyl group in echinocystic acid variant 74 (SQS-1-8-5-18). The triterpene C16-hydroxyl group was also independently deleted in the corresponding gypsogenin 75 (SQS-1-9-5-18), hederagenin 76 (SQS-1-10-5-18), and oleanolic acid 77 (SQS-1-7-5-18) variants (Figure ).

Figure 13

Structures of triterpene domain variants with independent modifications of C4-aldehyde and C16-hydroxyl group.

Synthesis of the echinocystic acid (74, SQS-1-8-5-18), hederagenin (76, SQS-1-10-5-18), and oleanolic acid (77, SQS-1-7-5-18) variants started from the corresponding commercially available triterpenes, and followed a similar sequence to that described above for quillaic acid variant 72 (SQS-1-0-5-18). Gypsogenin variant 75 (SQS-1-9-5-18) was also synthesized from hederagenin while caullophylogenin variant 73 (SQS-1-11-5-18) was accessed from an advanced intermediate en route to 72 (not shown). In mouse vaccinations with GD3-KLH, MUC1-KLH, and OVA, echinocystic acid variant 74 elicited antibody titers similar to or higher than QS-21, closely followed by caullophylogenin variant 73 (Figure ). Notably, both of these variants lack the C4-aldehyde substituent, bringing the Schiff base mechanistic hypothesis into question, at least in the context of these saponin variants. Meanwhile, gypsogenin (75), hederagenin (76), and oleanolic acid (77, not shown) variants, which lack the C16-hydroxyl, induced attenuated antibody responses. This suggests an important role for this functionality in adjuvant activity, perhaps affecting saponin conformation and/or target binding. In all cases, no weight loss was observed (not shown).[33] These results reveal the dispensability of the C4-aldehyde group for potent immunostimulatory effects and point to a previously unappreciated role for the C16-hydroxyl group in enhancing activity. Indeed, other saponin adjuvants that lack a C4-aldehyde substituent but possess a C16-hydroxyl group have been reported.[34,35] Moreover, the potent adjuvant activity, non-toxicity, and improved synthetic accessibility (23 total steps) of echinocystic acid variant 74 compared to QS-21 makes it a promising candidate for further development. Despite the important structural differences between this echinocystic acid variant and QS-21, it elicited both IgG1 and IgG2b antibody subclasses, indicative of the ability to induce both Th2 and Th1 responses, respectively, a hallmark of the natural product.

Figure 14

Immunological evaluation of triterpene domain variants (5 μg GD3-KLH, 2.5 μg MUC1-KLH, 20 μg OVA, 20 or 50 μg saponin). Data for attenuated oleanolic acid variant 77 (SQS-1-7-5-18) not shown.

Mechanistic Studies with Radiolabeled and Fluorescent Saponin Probes

Having established the potent adjuvant activity and low toxicity of aryl iodide variants 67 (SQS-0-0-5-18) and 72 (SQS-1-0-5-18), [131I]-radiolabeled congeners were generated (not shown)[33] in collaboration with Jason Lewis and Nagavarakishore Pillarsetty at MSKCC. Biodistribution studies were carried out in mice coadministered with OVA. The active variant [I]-67 ([131I]-SQS-0-0-5-18) was recovered at significantly higher levels at the injection site and nearest draining lymph nodes at 24 h post injection compared to a structurally similar but inactive variant, [131I]-SQS-0-3-7-18 (not shown), which lacks the linear tetrasaccharide domain (Figure a). Radioactivity was also retained at these two sites at 72 and 96 h for [I]-67, but not in other tissues and organs (where large differences were also initially observed at 24 h). In contrast, for the inactive saponin, [131I]-SQS-0-3-7-18, radioactivity cleared from all tissues by 72 and 96 h post injection. Similar biodistribution results were obtained using two truncated variants that lack the branched trisaccharide domain, active quillaic acid variant [I]-72 ([131I]-SQS-1-0-5-18) and attenuated oleanolic acid variant [I]-77 ([131I]-SQS-1-7-5-18) (not shown). This provided a positive correlation between adjuvant activity and this specific biodistribution profile.

Figure 15

Mechanistic studies of saponin variants. (a) Biodistribution of active and inactive radioiodinated saponins in mice (24 h post injection; 20 μg OVA, 20 μg unlabeled saponin, ≈25 μCi radiolabeled saponin); statistical significance shown only for injection site and lymph nodes. (b) In vivo fluorescence imaging of active fluorescein-labeled saponin 43 and inactive nonfluorescent precursor 42, each coadministered with Alexa-467-OVA (A647-OVA), in mice (24 h post injection; 20 μg A647-OVA, 10 μg saponin) and (c) fluorescence imaging of dissected lymph nodes (mice injected in left flank; right lymph nodes are negative controls). (d) Confocal microscopy imaging of subcellular localization of active and inactive fluorescent saponins and fluorescent glycine methyl ester (GlyOMe) negative controls in immature dendritic cells.

Other Quillaja saponin-containing adjuvant mixtures have been reported to impact the biodistribution of coadministered antigens.[36,37] To probe this possibility, biodistribution of [131I]-OVA in the presence or absence of active aryl iodide variant 67 (SQS-0-0-5-18) was also investigated. However, high thyroid uptake of radioactivity was observed, indicative of rapid deiodination of OVA. Thus, in a complementary approach, in vivo fluorescence imaging studies were performed with fluorescein-labeled active adjuvant 43 (SQS-0-0-5-12), in collaboration with Jeffrey Gardner at MSKCC. In mice injected with 43 and Alexa-647-labeled OVA (A647-OVA), both the adjuvant and antigen localized at the injection site and nearest draining lymph node at 24 h post injection (Figure b,c). In contrast, when coadministered with the inactive, nonfluorescent precursor 42 (SQS-0-0-5-11), A647-OVA was retained only at the injection site with no accumulation in the lymph nodes. Immunohistochemical analysis of dissected lymph nodes indicated subnodal localization of the active adjuvant 43 (SQS-0-0-5-12) to the cortex of the draining lymph nodes, and flow cytometric analysis revealed that internalization of 43 (SQS-0-0-5-12) was specific to dendritic cells. In an earlier confocal microscopy study, Chea et al. showed internalization of 43 to a distinct cellular compartment of immature dendritic cells, in contrast to related attenuated variants bearing BODIPY (SQS-0-0-5-17) and Cascade Blue (SQS-0-0-5-15) fluorophores, and other fluorescent negative controls (Figure d).[31] Taken together, these results suggest a role for active adjuvants in the trafficking of OVA by antigen-presenting cells to the draining lymph nodes, a known site of immune cell maturation, and provide early insights into the mechanisms of saponin immunopotentation.

Investigation of Saponin–Tucaresol Conjugates

By analogy to aldehyde-containing adjuvants such as tucaresol,[38] QS-21 has been suggested to interact with putative T-cell surface receptors through its C4-aldehyde substituent, providing a costimulatory signal leading to T-cell activation.[16,17] To investigate potential synergies between QS-21 and tucaresol, Fernández-Tejada et al. developed saponin–tucaresol conjugate 78 (SQS-0-0-5-19) and its truncated congener 79 (SQS-1-0-5-19) (Scheme ).[39] However, in mouse vaccinations with MUC1-KLH (2.5 μg) and OVA (20 μg), incorporation of tucaresol did not significantly enhance the adjuvant activity of these saponins (20 and 50 μg) compared to aryl iodide variants 67 (SQS-0-0-5-18) and 72 (SQS-1-0-5-18) (not shown). Aryl iodide variant 67 also exhibited similar adjuvant activity with or without equimolar tucaresol.

Scheme 7

Synthesis of Saponin–Tucaresol Conjugates

Streamlined, Divergent Synthesis of Linear Oligosaccharide Domain Variants

To develop more streamlined synthetic access to saponin variants, Fernández-Tejada et al. designed additional linear oligosaccharide domain variants, based on aryl iodide variant 72 (SQS-1-0-5-18), using readily available carbohydrate precursors. Dirhamnose variant 80 (SQS-1-0-10-18) and lactose variant 81 (SQS-1-0-11-18) incorporated modifications of individual sugars and linkages while 2-galactosamine variant 82 (SQS-1-0-12-18) altered the regiochemistry at the bridging monosaccharide (Figure ).[40] While the trisaccharide was assembled then coupled en bloc to the triterpene domain in the parent saponin 72 (SQS-1-0-5-18; 23 total steps, 16 for the linear trisaccharide) and dirhamnose variant 80 (22 total steps), a divergent strategy was developed for lactose variant 81 (16 total steps) and 2-galactosamine variant 82 (19 total steps), involving stepwise monoglycosylation of the triterpene domain with a single bridging sugar followed by elongation with the desired terminal disaccharide. This late-stage diversification should facilitate rapid preparation of diverse linear oligosaccharide variants to identify synthetically streamlined saponin adjuvants in the future.

Figure 16

Streamlined synthetic access using (a) linear oligosaccharide domain variants derived from readily available carbohydrate precursors and (b) a divergent synthetic strategy.

Conclusions

Despite being one of the most promising investigational immunoadjuvants, clinical advancement of QS-21 has been constrained due to its inherent liabilities, including scarcity, chemical instability, and toxicity. To realize the full potential of subunit vaccines, improved adjuvants that overcome these liabilities are required. Toward this end, the Gin lab first developed synthetic technologies to access QS-21, then leveraged them to prepare nearly 50 saponin analogues, providing detailed SAR (Figure ) and identifying variants with potent adjuvant activity, increased stability, and decreased toxicity. Several are promising candidates for further preclinical development, such as echinocystic acid variant 74 (SQS-1-8-5-18). Development of radiolabeled and fluorescent probes also enabled investigations into the enigmatic mechanisms of action of these saponin adjuvants. Correlation of three-dimensional structure with adjuvant activity and the pronounced SAR within this family suggest that these saponins may act by interacting with discrete molecular targets, in contrast to nonspecific inflammatory mechanisms attributed to other adjuvants.[3] Although the Gin lab research program has now concluded, it has provided efficient synthetic approaches and valuable mechanistic insights that should facilitate future development of improved adjuvants for use in subunit vaccines to address a variety of human diseases.

Figure 17

Summary of saponin structure–adjuvant activity relationships.