Toll-like Receptor Agonist Conjugation: A Chemical Perspective

Abstract

Toll-like receptors (TLRs) are vital elements of the mammalian immune system that function by recognizing pathogen-associated molecular patterns (PAMPs), bridging innate and adaptive immunity. They have become a prominent therapeutic target for the treatment of infectious diseases, cancer, and allergies, with many TLR agonists currently in clinical trials or approved as immunostimulants. Numerous studies have shown that conjugation of TLR agonists to other molecules can beneficially influence their potency, toxicity, pharmacokinetics, or function. The functional properties of TLR agonist conjugates, however, are highly dependent on the ligation strategy employed. Here, we review the chemical structural requirements for effective functional TLR agonist conjugation. In addition, we provide similar analysis for those that have yet to be conjugated. Moreover, we discuss applications of covalent TLR agonist conjugation and their implications for clinical use.

Graphical Abstract

INTRODUCTION

The immune system is vital in protecting the body against infectious disease and cancers. Toll-like receptor (TLR) activation is the first step in stimulating the immune system to respond to disease.¹ As a consequence, TLRs are the target of several therapeutics, including vaccines,² cancer immunotherapies,³ and autoimmune disease treatments.^4,5 Activators of TLR signaling, TLR agonists, are powerful immunostimulants due to their ability to drive innate and acquired immunity and are therefore sought after as adjuvants in vaccines.⁶ However, because they elicit strong immune reactions, administration of TLR agonists can have toxic side-effects in vivo.^7–9

To mitigate these adverse effects, TLR agonists can be covalently conjugated to an antigen. This ensures simultaneous delivery of antigen and adjuvant to the same antigen presenting cell, which directs the response toward the antigen, thereby lowering the required dose for an effective immune response.^10,11 These self-adjuvanting vaccines induce stronger humoral and cellular immune responses and are less toxic than vaccines comprising coadministered antigen and adjuvant.¹⁰ TLR agonists have also been conjugated to polymers,^12–14 solid particles,^15–17 or phospholipids^18–20 to beneficially influence the compound’s drug properties.

The benefits of TLR agonist conjugation can only be achieved if the conjugation does not perturb the interaction between the agonist and the receptor. Therefore, effective strategies are required that facilitate TLR agonist conjugation with retention of, or improved, TLR agonist potency. Here, we provide a review of the chemical requirements for successful TLR agonist conjugation strategies that have been employed, and provide suggestions for alternative ligation strategies for existing TLR agonists and insight for agonists that have yet to be conjugated.

TLRs—NOMENCLATURE AND FUNCTION

Toll-like receptors are a class of transmembrane pattern recognition receptors (PRRs). Their role is to recognize conserved molecular components of pathogens called “pathogen-associated molecular patterns” (PAMPs) as well as cellular damage-associated molecular patterns (DAMPs).²¹ The receptors are predominantly expressed in immune cells and show distinct expression patterns across different cell types.²²

The human TLR family comprises 10 functional TLRs that are expressed both on the cell surface (TLR1, 2, 4, 5, and 6) and intracellularly in endosomal compartments (TLR 3, 4, 7, 8, and 9).^23,24 The cellular localization of TLRs reflects the origin of the ligand it is responsible for detecting (Figure 1). TLRs found at the cell surface respond to extracellular components of pathogens, including lipoproteins (TLR1, 2, and 6), lipopolysaccharides (TLR4), and bacterial flagellin (TLR5). Endosomal TLRs recognize components from the intracellular compartments of pathogens, such as double-stranded RNA (dsRNA, TLR3), single-stranded RNA (ssRNA, TLR7 and 8), and unmethylated cytosine-phosphate-guanine (CpG) DNA (TLR9).²⁵ Although the role and natural ligand for human TLR10 is not yet fully understood, it is thought to be a negative regulator of TLR signaling.^26–28 Binding of ligands to TLRs induces homo- or heterodimerization of TLR receptors and subsequent downstream signaling through two distinct pathways, the MyD88-dependent (TLRs 1, 2, and 4–9) and TRIF-dependent pathway (TLR3 and 4), generally mediating inflammatory and antiviral responses, respectively.^29,30

Figure 1.

Cellular location of TLRs, their ligands, and signal transduction pathways. TLRs 1, 2, 6, 4, and 5 are located on the plasma membrane and signal, together with endosomal TLRs 7, 8, and 9, in a MyD88-dependent manner. The MyD88 pathway leads to NF-κB translocation to the nucleus and, ultimately, production of inflammatory cytokines. Activated TLR4 is endocytosed and, together with endosomal TLR3, signals through the IRF pathway, leading to Type I interferon production. MD2, TRAM, and TIRAP are adaptor proteins.

TLRs are crucial for their role in the innate immune system in fighting infection. They are responsible for the initial response to pathogens and their activity drives the generation of innate and adaptive immune responses. Innate immunity begins with the recognition of PAMPs through TLRs by antigen presenting cells (APCs), mainly dendritic cells (DCs) in the peripheral tissues.²³ Binding of TLR ligands on dendritic cells initiates a downstream signaling cascade that initiates DC maturation. This process is characterized by the production of proinflammatory cytokines (TNF-α, IL-6, and IL-12), up-regulation of costimulatory molecules (CD40, CD80, and CD86), increased antigen-presenting capacity, and DC migration from the peripheral tissues to draining lymph nodes.²³ There, antigen presenting DCs stimulate naïve T-cells which begins the adaptive immune response. B-cells also express TLRs and present antigen upon activation, resulting in antibody production specific for the encountered antigen. Since T- and B-cell maturation is influenced by TLR, costimulatory, and cytokine signaling, TLR agonists play an important role in establishing the type of cellular and humoral immune response that is generated.

TLR AGONIST CONJUGATES

Due to their pivotal role in both the innate and adaptive immune system, agonists for TLRs have emerged as therapeutic agents, including as adjuvants in vaccines. Adjuvants function by activating immune cells exposed to antigen. Thus, the characteristics of efficacious adjuvants follow different design principles than traditional drugs. As reviewed by Wu,³¹ an effective adjuvant should have a higher retention time at the site of administration to promote local immune system stimulation and prevent potentially harmful, off-target inflammation.

In their review, Xu and Moyle³² state that direct conjugation of TLR agonists to antigen fulfills these requirements by ensuring codelivery of antigen and adjuvant to immune cells. This enhances the generation of antigen-specific immune responses and prevents wasted inflammation.³³ In addition, TLR agonists-conjugates have been used as chemical tools to study the role of TLR signaling in immune responses, which was recently reviewed by Oosenbrug and co-workers.³⁴ However, conjugation of TLR to antigens or other molecules requires thoughtful consideration in the choice of TLR agonist, chemical linkage strategy, and site of conjugation. Here, we review conjugation of TLR agonists from a chemistry perspective to provide insight for future applications.

CONJUGATION STRATEGIES

TLR2.

The TLR2 subfamily is unique among TLRs in that it forms heterodimers with either TLR1 or TLR6 instead of homodimers (Figure 2A). TLR2 is involved in the recognition PAMPs from a wide range of pathogens, including bacteria, viruses, and fungi.³⁵ The prototypical agonists to TLR2, however, are a class of bacterial di- or triacylated lipoproteins and lipopeptides and their synthetic analogues with N-terminal di- or tripalmitoylated cysteine residues (Pam₂Cys, Pam₃Cys, respectively) (Figure 2B). Pam₂Cys and Pam₃Cys bind TLR2 via two ester bound palmitoyl chains, whereas N-terminal acylation determines TLR1 or TLR6 specificity; an N-terminal fatty acid recruits TLR1 and a free N-terminal amine recruits TLR6 (Figure 2A).³⁶ They often contain the peptide-motif SK₄, which enhances the agonists’ adjuvanticity and water-solubility.³⁷ Isomerically pure R-epimers of Pam₂CSK₄ (EC₅₀: <1 nM) and Pam₃CSK₄ (EC₅₀: 5 nM) are more potent TLR2 agonists than their racemic mixtures (EC₅₀: 1 nM and 20 nM, respectively), due to lower activity of the S-epimers.^38,39 Because of their ability to induce robust maturation of DCs and effective humoral and cellular responses,⁴⁰ these lipopeptides have been employed as adjuvants in many vaccine studies, even predating the discovery of TLRs.⁴¹

Figure 2.

TLR2 agonists bind dimers of either TLR2-TLR1 or TLR2-TLR6. (A) Crystal structure of human TLR2-TLR1-Pam₃CSK₄ (PDB: 2Z7X). Red rectangle represents the area of the zoomed view of Pam₃CSK₄ binding to TLR2-TLR1. The analogous view from the TLR2-TLR6-Pam₂CSK₄ crystal structure is also shown (PDB: 5IJC). In both cases, the peptide C-terminus is solvent exposed. TLR2 is shown in blue, TLR1 and TLR6 are in orange, and lipopeptides are in green. TLRs in the zoomed view are shown at 40% transparency. (B) Chemical structure of Pam₂CSK₄ and Pam₃CSK₄. Green arrows highlight sites of conjugation.

Pam₂Cys and Pam₃Cys and their derivatives are the TLR2 agonists most often found in bioconjugates due to their well-defined chemical structure, ease of production, and the absence of natural endotoxins (Table 1). One exception is the TLR2-TLR6 ligand, lipoteichoic acid (LTA), a major component of the Gram-positive bacterial cell wall.⁴² LTA was conjugated to both CpG oligodeoxynucleotides (ODN) 1826,^43,44 a TLR 9 agonist, and to the cell surface of Lewis Lung Carcinoma (LLC) cells⁴⁵ via a poly(ethylene glycol) (PEG) linker. In both instances, conjugation of LTA was achieved by amide bond formation of the primary amines of side-chain D-alanines with N-hydroxysuccinimide (NHS) esters on heterobifunctional linkers. Despite its high potency, LTA has not been used in bioconjugation as extensively as Pam₂Cys and Pam₃Cys, as endotoxin free LTA is difficult to produce and is therefore expensive. Moreover, the aforementioned conjugation strategies for LTA yield a mixture of high molecular weight, structurally diverse LTA-conjugates. Conversely, Pam₂Cys and Pam₃Cys are synthetically accessible and their conjugates can be made structurally homogeneous.

Table 1.

Overview of TLR1, 2, 3, 4, 5, and 6 Agonist Conjugates

TLR	agonist	conjugation site	conjugation chemistry	conjugated to	refa
TLR1-TLR2	Pam3Cys	C-terminus cysteine	SPPS, NCL, CuAAC	Peptide, glycopeptide	132,166,49,52,54
	Pam3CSK4	C-terminus lysine	SPPS, thioether bond	Peptide, glycopeptide	39,48,50,55
		ℰ-NH2 group lysine SKKKK	CuAAC	Glycopeptide antigen	167
	Amplivant	C-terminus lysine	SPPS	Peptide	64
TLR2b	Lipoamino acids	N- and C-terminus	SPPS	Peptide and glycopeptide antigens	69,132,168–170
TLR3	Poly(I:C)	NH2 of any cytosine	Bisulfite catalyzed transamination	Polysaccharide	78
		5’ termini	Phosphoramidate coupling	Nanoparticles	82,83
TLR4	MPLA	Reducing end anomeric center	Amide, CuAAC	Oligosaccharide	98–100,102,101,171,172
		Nonreducing end 6’ position	CuAAC	Oligosaccharide	101
		Any free –OH group	Carbamate	Protein	97
	Pyrimidoindole	Carboxamide group	Amide	Polymer, TLR7/8- and 9 agonists	13,43,107
TLR5	Flagellin	N- or C-terminus, or D3 domain	Fusion proteins	Various proteins	Review32
TLR2-TLR6	Pam2Cys	C-terminus	SPPS, amide, CuAAC	Polymers, peptide, glycopeptide, TLR7- and NOD2 agonists	49,132,173–177
		N-terminus cysteine	Amide	Fluorophore, chelating agent	60
	MALP-2	C-terminus	Amide	Glycopeptide	62
	Pam2CSK4	C-terminus lysine	Amide, SPPS	TLR7 and 9 agonists	44,173,178
		ℰ-NH2 of lysine SKKKK	Amide	TLR7 agonist	173
		N-terminus cysteine	Thiourea, carbamate bond	Fluorophore, photocage	178,108
	LTA	NH2 of any d-alanine	Amide	LLC cells, TLR9 agonist	45,45

^aReferences are clustered and ordered by conjugation chemistry.

^bTLR2 ligands for which the coreceptor is unknown.

Triacylated TLR2-TLR1 agonists Pam₃Cys and Pam₃CSK₄ have been used extensively as conjugates to create self-adjuvanting vaccines. For instance, Pam₃Cys on the N-terminus of N. meningitidis lipoproteins greatly enhances the humoral response to Trumenba, an FDA approved, self-adjuvanting N. meningitidis vaccine.^46,47 Pam₃Cys and its derivatives are readily synthesized, structurally well-defined, and commercially available, and do not suffer from the drawbacks associated with LTA. Conjugation with these compounds mainly proceeds through the C-terminal carboxyl group to avoid perturbation of N-terminal lipid interactions with the receptors (Figure 2B). The incorporation of Pam₃Cys and Pam₃CSK₄ into peptide-based vaccines can easily be achieved through standard solid-phase peptide synthesis (SPPS), placing Pam₃Cys at the N-terminus of the peptide. Jung and co-workers were the first to apply this strategy for the synthesis of a peptide-based antigen-adjuvant conjugate.⁴¹ Conjugation of Pam₃Cys to peptide antigens greatly increases antigen uptake, processing, and presentation⁴⁸ and leads to enhanced humoral^41,49 and cellular⁵⁰ immune responses. Zeng et al.⁵¹ explored how the positioning of Pam₃Cys in a peptide-conjugate affects the conjugate’s pharmacodynamic and immunogenic properties. Placing Pam₃Cys in the middle of the peptide sequence dramatically increased the conjugate’s solubility and its ability to induce epitope-specific antibodies compared to terminally conjugated Pam₃Cys, Similar results were obtained for TLR2-TLR6 agonist Pam₂Cys,⁵¹ providing a powerful alternative to linear conjugation of lipopeptide TLR2 agonists.

Preparation of glyco(lipo)peptide conjugates of Pam₃Cys via SPPS has proven difficult. Issues with separation of the glyco(lipo)peptide conjugate after resin cleavage⁴⁹ and lack of orthogonality between carbohydrate deprotection reagents and palmitoyl esters have lead to alternative, convergent synthesis strategies. Several methods to create multicomponent vaccines that contain the tumor-associated carbohydrate antigen (TACA) MUC1, a T-helper epitope, and Pam₃Cys have been reported. The Boons lab has employed native chemical ligation (NCL)^49,52 to sequentially conjugate TACA MUC1 to a T-helper peptide epitope and Pam₃CSK₄. Other approaches involve functionalization of the C-terminus of Pam₃Cys with a functional handle, such as alkynes,^53,54 a bromo- or iodoacetyl group,⁵⁵ or a pentafluorphenyl-activated ester,^56,57 followed by conjugation to peptides and glycopeptide. Together, these examples show that Pam₃Cys retains its potency in conjugates and that its conjugation is relatively straightforward in peptide-based conjugates, but more demanding in glycopeptide conjugates.

Conjugates of Pam₃Cys can suffer from poor solubility in aqueous solutions and can therefore be difficult to dose and formulate in vaccines.⁵¹ Pam₂Cys, a synthetic analogue of the N-terminal lipopeptide motif of MALP-2, has a free amino group in place of the N-terminal palmitoyl chain in Pam₃Cys resulting in improved solubility characteristics and increased potency toward splenocytes⁵⁸ and macrophages.⁵⁹ The free amino group of Pam₂Cys has also been used as a conjugation handle. After introducing a PEG spacer at the N-terminus, Vagner and co-workers conjugated relatively bulky functionalities, such as a chelating agent and fluorescent dyes, the N-terminal amine of Pam₂Cys while retaining relatively high TLR2-TLR6 potency.⁶⁰ In contrast, conjugation of photocage 2-(2-nitrophenyl)propyl chloroformate to the N-terminal primary amine without a spacer dramatically decreases the activity of Pam₂CSK₄.⁶¹ These results suggest that the N-terminus of Pam₂Cys can be used for conjugation, but that a chemical spacer may be necessary to retain potency toward TLR2-TLR6. Similar to Pam₃Cys, conjugation of Pam₂Cys can also be accomplished by C-terminal conjugation, as was done in a conjugate of MALP-2 and MUC-1.⁶²

Many structure–activity relationship (SAR) studies into Pam₂Cys and Pam₃Cys have been performed in order to optimize their adjuvant properties. These studies have resulted in several TLR2 agonist analogues with interesting chemical and immunological properties, but some of these have not yet been used in bioconjugates. By replacing the N-terminal palmitoyl amide-linkage of Pam₃Cys for a urea-linkage, Willems and co-workers synthesized a Pam₃Cys analogue called Upam (trademarked by Isa pharmaceuticals as Amplivant) that was shown to be a stronger inducer of DC maturation than Pam₃Cys.⁶³ Analogous to Pam₃Cys, Amplivant was C-terminally conjugated to a human papillomavirus (HPV) synthetic long peptide (SLP) vaccine and enhanced HPV antigen-specific cellular immune responses in ex vivo stimulated cervical tumor material.⁶⁴ Similarly, Wu et al. have synthesized a Pam₃Cys analogue with amide-linked palmitoyl chains (SUP3) and showed it was a superior adjuvant in an in vivo tumor challenge model.⁶⁵ Both of these Pam₃Cys analogues benefit from resistance to esterases, which is hypothesized to contribute to their adjuvanticity.⁶⁵ Although SUP3 has not been used in antigen-conjugates so far, it could be an excellent adjuvant for use in self-adjuvanting vaccines and perhaps even superior to Pam₃Cys. Alternatively, Salunke et al. have reported on potent monopalmitoylated lipopeptide analogues specific to human TLR2.⁶⁶ These analogues are simpler in structure and synthetically more accessible than Pam₃Cys, but have not been used in conjugates as of yet despite their attractive properties. We suggest that any conjugation should proceed through the C-terminus, analogous to conjugations of Pam₂Cys and Pam₃Cys.

An important consideration for the conjugation of lipopeptide TLR2 agonists are the biophysical effects accompanying the hydrophobic lipid chains. PamCSK₄, Pam₂CSK₄, and Pam₃CSK₄ can self-assemble into micelles in aqueous solution.⁶⁷ These self-assembly characteristics have been exploited in self-adjuvanting vaccines where lipopeptides are used as both adjuvant and delivery vehicle. Moyle et al. created multiple group A Streptococcus (GAS) vaccines that contained Pam₂Cys, Pam₃Cys, or lipoamino acids, lipidated α-amino acids that are believed to activate TLR2.^68,69 These compounds self-assembled into nanoparticles of a suitable size (20–200 nm) for passive transport to the draining lymph nodes, which enhanced production of antigen-specific IgG antibodies in vivo.^68,69 Moreover, micellar self-assembly of TLR2 activating lipoproteins in N. meningitidis vaccine Trumenba is thought to increase the stability of the lipoprotein antigens.⁴⁶ However, the self-assembling nature of lipopeptide TLR2 agonists can differ between bioconjugates of different chemical composition. Thus, the biophysical characteristics of a specific lipopeptide-conjugate should be considered as these interactions may affect function.

An alternative to lipopeptide TLR2 agonists are a group of nonlipid TLR2-TLR1 agonists, which were discovered by Guan et al. using high-throughput screening of a chemical library.⁷⁰ Structurally distinct from lipopeptide agonists, these synthetically accessible small molecules display selective potency toward TLR2-TLR1 at concentrations as low as 30 nM, with some of them being selective for either human or mouse TLR1-TLR2. The small molecules have not yet been used in bioconjugates, nor is there any information available on SAR or binding characteristics. Thus conjugation with these agonists would be more exploratory. However, due to their small molecule nature, these TLR2-TLR1 agonists could find use in conjugate applications in which lipopeptide bulk or hydrophobicity would be problematic.

Both large and small molecule agonists have been described for TLR2. In considering a conjugation strategy, the palmitoylated peptide family offers appealing aspects of processability with the C-terminus providing a viable target for most bioconjugation methods. However, biophysical aggregation must be considered. Future work might explore novel ligands containing more simple, hydrophilic chemical motifs to improve accessibility and enhance solubility while avoiding systemic immune activation.

TLR3.

TLR3 is an endosomal TLR that senses the negatively charged backbone of double-stranded RNA (dsRNA) that is an intermediate in the replication of ssRNA viruses and makes up the genome of some dsRNA viruses (Figure 3A).⁷¹ This receptor is unique among TLRs as it signals solely through the TRIF pathway. The most commonly used TLR3 agonist is a high molecular weight synthetic dsRNA analogue, polyinosinic-polycytidylic acid (poly(I:C)).⁷² Poly(I:C) is a potent stimulator of conventional DCs (cDCs) and effector T-cell responses.⁷³ However, its development as a vaccine adjuvant has been hampered by its in vivo toxicity at high doses,⁹ which have been partially attributed to additional signaling through cytosolic PRRs MDA-5 and RIG-1 in addition to TLR3.⁷⁴ These effects could potentially be mitigated by poly(I:C) conjugation with other compounds. Alternative dsRNA TLR3 agonists IPH-3102,⁷⁵ Rintatolimod⁷⁶ (poly-IC₁₂U) and poly-ICLC⁷⁷ show reduced toxicity but have not been used as TLR-conjugates as of yet.

Figure 3.

TLR3 agonists are dsRNAs that bind to TLR3 dimers. (A) Crystal structure of mouse TLR3-dsRNA (PDB: 3CIY). The receptors bind to the negatively charged backbone of the dsRNA molecule. TLR3 is shown in blue and orange, and dsRNA is shown in green. (B) Carton structure of dsRNA with sites that have been conjugated highlighted by green arrows and suggested sites of conjugation by blue arrows.

Despite the clinical potential for TLR3 agonist conjugates, few examples exist of poly(I:C)-conjugates (Table 1). Huang et al. conjugated commercially available poly(I:C) to a protein antigen by introducing linkers on the cytosine primary amines using sodium bisulfite catalyzed transamination reactions.⁷⁸ This strategy invariably leads to a heterogeneous mixture of conjugates, however, as there are multiple cytosines within a poly(I:C) molecule. Although they have not been directly compared in a study, we speculate that conjugation on the nucleobases of poly(I:C) is more likely to disrupt binding to TLR3 than conjugation to the termini, due to the fact that dsRNA binding to TLR3 requires a stretch of at least 40–50 unobstructed base pairs (Figure 3A).⁷⁹ To conjugate poly(I:C) with its termini, one can employ phosphoramidate chemistry on terminal 5′-phosphates (Figure 3B).^80,81 This technique was used by Shukoor et al. to conjugate poly(I:C) to γ-Fe₂O₃ nanoparticles with its 5′ terminal phosphates.^82,83

Zhang et al. recently reported⁸⁴ on the only known class of small molecule TLR3 agonists. Their lead compound, CU-CPT17e, is a weakly potent human TLR3 agonist (EC₅₀: 4.8 μM) and can also signal through human TLRs 8 (EC₅₀: 13.5 μM) and human TLR9 (EC₅₀: 5.7 μM), while signaling through murine TLRs was not investigated in this study. CU-CPT17e and its analogues could be a useful alternative to poly(I:C) in conjugates that require small molecule TLR3 agonists. However, SAR studies show that alterations to the structure of CU-CPT17e consistently leads to abrogation of TLR3 potency,⁸⁴ complicating the successful conjugation of this small molecule TLR3 agonist. Alternatively, studies have shown that nucleic acid based adjuvants can be associated with various molecules, including antigens, through electrostatic interactions with the backbone phosphates. Zhang et al. demonstrated this technique to form polyelectrolyte multilayer vaccines composed of poly(I:C) and cationic ovalbumin (OVA) peptide antigens which stimulated robust T cell responses.⁸⁵ Follow-up studies further characterized these materials and their ability to generate powerful immune responses.^86,87

In conclusion, TLR3 has a unique signaling pathway and immunological activity but has been limited in application due to a lack of agonists. Although highly potent, commercially available dsRNA TLR3 agonists are bulky, polydisperse, and difficult to characterize using conventional chemical analysis techniques, conjugation of prototypical TLR3 agonist poly(I:C) can proceed by introducing chemical handles to cytosines along its backbone. However, we suggest that the terminal ends of dsRNA analogues may be better sites for conjugation based on site availability as suggested by crystal structure analysis and the homogeneous nature of such a compound. The recently reported class of small molecule TLR3 agonists are attractive due to their small molecule nature, but they are considerably less potent toward TLR3 than poly(I:C). Thus, the field would benefit from the discovery of novel, potent small-molecule ligands that are exclusive agonists of TLR3 and can be modified to generate bioconjugates.

TLR4.

TLR4 was the first human TLR identified to bind a ligand, by Poltorak et al.⁸⁸ and Hoshino et al.,⁸⁹ and is the most studied among the TLR family. TLR4 binds ligands with the aid of myeloid differentiation factor 2 (MD2), forming a TLR4-MD2-agonist complex (Figure 4A).⁹⁰ TLR4 is unique among TLRs in that, upon ligand binding, it can signal through both the MyD88 and TRIF pathways (Figure 1), resulting in a variety of accessible immune response profiles.⁹¹ Some TLR4 agonists bias particular signaling pathways, presumably through differing conformational changes and cellular location upon receptor binding, resulting in unique immune responses.⁹² Thus, tuning the chemical structure of these agonists impacts the immunological activity of the compound. Its main natural binding ligands are lipopolysaccharides (LPS), major components of the Gram-negative bacteria cell membrane responsible for bacterial sepsis (Figure 4A).⁹⁰ Although many forms of LPS exist, the toxicity of these molecules has prevented their use in vaccines. However, several groups, notably Qureshi et al.,⁹³ developed a detoxified analog of LPS, monophosphoryl lipid A (MPLA), which is now FDA approved as an adjuvant in multiple vaccines and is currently in several clinical trials.⁹⁴ The diversity in chemical composition and immunological activity of TLR4 agonists presents an opportunity for chemical manipulation in a variety of TLR conjugate applications (Table 1).

Figure 4.

TLR4 agonists bind MD2 and TLR4. (A) Crystal structure of human TLR4-MD2-LPS (PDB: 3FXI) from two viewpoints (left), red rectangle depicts zoom view of LPS binding (middle) in which proteins are shown with 40% transparency. The analogous view from the mouse TLR4-MD2-Neoseptin-3 crystal structure is also shown (PDB: 5IJC, right). TLR4 is shown in blue and orange, MD2 is shown in gray, and LPS and Neoseptin-3 are shown in green. (B) Chemical structures of various TLR4 agonists. Areas that bind MD2 (dotted gray) and TLR4 (dotted orange) shown. Green arrows highlight previous sites of conjugation; blue arrows highlight suggested sites of conjugation.

The glycolipid moiety in LPS-derived TLR4 agonists contains several sites for modification. MPLA was the first example of an intentionally modified LPS derivative and has since been used heavily as a vaccine adjuvant. MPLA is the lipid A component of LPS, but lacking the reducing end phosphate, which decreases the toxicity over 1000-fold (Figure 4B).⁹⁵ The reduction in toxicity has been attributed to the bias of activation toward the TRIF pathway over MyD88 pathway, resulting in less pro-inflammatory cytokine production.⁹⁶ Modifications that bias activity have been further exploited to tune the immunomodulatory properties of the agonist. For example, Carter and co-workers have described glucopyranosyl lipid adjuvant (GLA) and second-generation lipid adjuvant (SLA), MPLA analogues with varying lipid chain lengths that elicit distinct immunological responses and activate both mouse and human TLR4.⁹² Although these modifications can be used to tune the immune response, they are not sites amenable to conjugation chemistry as this portion of the molecule directly binds to the MD2-TLR4 complex (Figure 4B).

Although the hydrophobic portions of glycolipid TLR4 agonists are not ideal sites of conjugation, the exposed functional groups of the sugars have been used as conjugation handles. In one example, Schülke et al. reacted MPLA with carbonyldiimidazole (CDI) followed by OVA to form a nonspecific MPLA-OVA conjugate.⁹⁷ The conjugation resulted in boosted immune responses toward OVA compared to the mixture of unconjugated MPLA and OVA. In another example, Guo and co-workers developed a synthetic strategy to afford an MPLA derivative with an amino linker appended to the nonreducing hydroxyl group.^98–100 Conjugates of this derivative with TACAs via amide bond formation afforded self-adjuvanting, glycoconjugate cancer vaccines. Recently, Wang et al. did a SAR study on MPLA conjugation to a lipoarabinomannam oligosaccharide and found that nonreducing 6′-hydroxyl conjugates had superior immunological activity to 1-O-conjugates.¹⁰¹ The Guo lab also used 6′-hydroxyl conjugation of MPLA to produce an MPLA-Group C Meningitis glyco-antigen-conjugate vaccine.¹⁰² Although the synthetic method developed in this study resulted in 10–100’s of milligrams of MPLA derivatives amenable to a variety of conjugation strategies, the approach requires expertise in synthetic organic carbohydrate chemistry, which may not be present in many vaccine development teams. If covalent conjugation is not an option, noncovalent codelivery strategies for MPLA and antigen could be considered. Several studies have reported enhanced adjuvant effects when incorporating glycolipid TLR4 agonists into emulsions, including some FDA approved vaccines, a straightforward strategy given the hydrophobicity of the compounds.^103–105 Thus, small molecules may be more advantageous for applications involving direct chemical linkage for the greater ease in chemical manipulation and hydrophilicity.

Recently, a variety of small molecule TLR4 agonists bearing little structural resemblance to glycolipids have been developed. These include pyramido-[5,4-b]indoles, 4-aminoazoquinolines, α-aminoacyl amide “Ugi products”, and neoseptins. Given their synthetically accessible nature, these molecules are more amenable to functionalization for various applications. However, all of these molecules bind predominantly within the MD2 pocket, making it difficult to introduce chemical modifications without sacrificing agonist potency (Figure 4B). Therefore, it is critical to evaluate the site of modification before embarking on a synthetic route to obtain the desired conjugate.

The pyramido-[5,4-b]indoles were discovered by Chan and co-workers via high throughput screening in 2013.¹⁰⁶ The hit compound in this study (Compound 28, EC₅₀: 1–10 μM) was shown to stimulate high production of IP-10/CXCL10 and IL-6 in vitro. Variations of this molecule have been used in conjugation studies, including multivalent TLR conjugates, by introducing a linker on the carboxamide moiety.^43,107 In another study, Lynn and co-workers synthesized a derivative of the lead pyramido-[5,4-b]indole compound by adding an amine functionality to a carboxamide cyclohexyl group, which was then conjugated to N-isopropylacrylamide (NIPAM) polymers using a bivalent tetraethylene glycol linker.¹³ In addition, the indole nitrogen has been functionalized with nitroveratryloxycarbonyl (NVOC) to afford a light-activated TLR4 agonist, showing that this site is important for ligand binding despite docking studies, suggesting that it may be solvent accessible.¹⁰⁸ More recently, a further optimized pyramido-[5,4-b]indole TLR4 agonist (Compound 26, EC₅₀: 600 nM, Figure 4B) was shown to improve and bias human TLR4 activation over mouse TLR4 activation.¹⁰⁹ Although this molecule has yet to be implemented in conjugation applications, we expect it to behave similarly to the original hit molecule, Compound 28, based on reported docking studies and SAR studies.

The same group discovered substituted 4-aminoquinazoline human TLR4 agonists by high throughput screening in 2014.¹¹⁰ The hit compound (Compound 1a, EC₅₀: 2 μM) was predicted to bind in a similar manner as pyramido-[5,4-b]indole compounds in docking studies, despite the lack of structural similarity. SAR studies suggested that the ester moiety may be amenable to derivatization for conjugation applications as functionalization with bulky groups had favorable effects on potency (Figure 4B). However, docking studies suggested that this site may reside at the interface of TLR4 and MD-2, meaning that conjugation to this site may require a linker to prevent perturbation of TLR4-MD-2 dimerization.

Neoseptin-3 is a peptidomimetic molecule discovered by Wang et al. through screening compounds for activation of TLR4-MD2 in murine macrophages.¹¹¹ Despite the simple design, the molecule has few sites amenable to chemical functionalization due to the unique binding mechanism of the compound. Two Neoseptin-3 molecules bind asymmetrically within the same binding pocket of the TLR4-MD2 complex, a property which is thought to be unique among all studied protein–ligand interactions (Figure 4A). The Neoseptin-3 aniline moiety appears to be a potential site of conjugation, based on the reported SAR study and crystal structure evaluation (Figure 4B). However, while one of the molecule’s anilines appears to be solvent exposed, the other interacts with TLR4 residues. Hypothetically, stimulation with a 1:1 mixture of conjugated and unconjugated Neoseptin-3 molecules may avoid this issue. However, this strategy may result in little Neoseptin-3-conjugate binding if the conjugate affinity for the receptor is lower than unconjugated Neoseptin-3. Thus, Neoseptin-3 conjugation strategies would likely require empirical design and testing to determine effective sites or linkers needed. In addition, Neoseptin-3 only activates mouse TLR4, limiting its practical use.

Marshall et al. developed the human TLR4 stimulating “Ugi compounds” using a combinatorial molecule screen with Ugi multicomponent reaction products.¹¹² These molecules were highlighted for their high potency and hydrophilicity. One of the lead compounds in the study, AZ617 (Figure 4B, EC₅₀: 6.3 nM), was within an order of magnitude of LPS potency (EC₅₀: 2.4 nM) and exhibited ideal hydrophilicity characteristics. Docking studies suggest that the compound binds in a similar location to other small molecule TLR4 agonists, between MD-2 and TLR4. In this case, the free carboxyl group appears to be close to the solvent exposed area of the binding pocket. Thus, conjugation to other molecules, potentially through amide or ester bond formation, might be possible at this site. In fact, some derivatives in SAR studies contained an alkyne substituted for the carboxyl group, among other modifications, making it amenable to click chemistry. However, these compounds suffered a 10–100-fold drop in activity and reduced hydrophilicity compared to AZ617.

Neve and co-workers demonstrated that derivatives of Euodenine A, a natural product isolated from Papua New Guinea Tree, had low μM affinity for human TLR4.¹¹³ SAR studies suggest that substituting the methyl group appended to the cyclobutyl moiety in Euodenine A with functional handles for conjugation may be tolerated. However, no docking studies are reported for this molecule, making conjugation attempts to the molecule more empirical relative to other small molecule TLR4 agonists.

In summary, TLR4 agonists have great potential as immunological modulators as demonstrated by their clinical use. These applications could be expanded or improved with TLR4 agonist conjugates. Although most applications have used glycolipid based TLR4 agonists, like MPLA or GLA, these compounds are difficult to conjugate effectively. Alternatively, several small molecule TLR4 agonists exist with a range of chemical structure and potency. However, SAR studies have demonstrated that potency is sensitive to small changes in chemical structure. Thus, TLR4 agonist conjugation strategies could be advanced by developing simplified glycolipid synthesis protocols, reliable modification techniques, and by improved design of small molecules amenable to conjugation. Billod et al. recently reviewed computation based approaches for TLR4 ligand prediction and design, which could aid in the development of TLR4 agonists for conjugation.¹¹⁴

TLR5.

TLR5 resides on the cell membrane and responds to flagellin.^115,116 Flagellin is a bacterial protein (~50 kDa) that self-assembles to form the flagellum organelle, providing motility for the organism. Recombinant flagellin and modified variants are also potent activators of TLR5.^117–119 TLR5 agonist conjugates have largely been used as recombinant fusion proteins with disease related antigens to generate single molecule vaccines, several of which are in clinical trials (Table 1).^120–122 Recombinant expression of antigens with fused flagellin may not work in every application. For example, the fused protein may have poor solubility characteristics, compromise antigen expression, block antibody epitopes, or generate immune responses that are not well suited for specific applications. To overcome some of the limitations of flagellin-antigen fusion proteins, several studies have shown that codelivery of flagellin-coated nanoparticles and antigen can also be used as effective vaccines.^123–125 Flagellin has also been conjugated to other compounds through non-site-specific cross-linking chemistries with surface exposed lysines.^{118,126–128}

In these studies, authors have noted that the conjugation approach could be improved by generating flagellin variants which are designed to facilitate conjugation in portions of the protein that are not involved with TLR5 binding.

The only synthetic small molecule agonists for TLR5 are flagellin derived peptides.¹²⁹ These peptides share homology to flagellin’s highly conserved D1 binding domain and can be as small as 13 residues in size. To the best of our knowledge, conjugates of flagellin derived peptides have yet to be synthesized for vaccine applications. The flagellin peptide should be amenable to conjugation with other peptides or molecules via traditional SPPS approaches. However, we observed no in vitro activity of TLR5 peptides or TLR5 peptides conjugated with small molecules in initial studies (unpublished data).

TLR7 and TLR8.

TLR7 and TLR8 agonists are among the most studied of the TLR family. Similar to TLR3 and TLR9, the receptors are located in the endosome and bind nucleic acid-based agonists (Figure 5A).¹³⁰ However, unlike TLR3 and 9, TLR7 and 8 are expressed across most human immune cell types. This feature has made TLR 7 and 8 targeting adjuvants sought after to generate efficacious T cell responses in human vaccines.¹³¹ In addition, most TLR7/8 agonists are effective across many species.

Figure 5.

TLR7/8 agonists bind TLR7 and/or TLR8 homodimers. (A) Crystal structure of monkey TLR7-Resiquimod (PDB: 3FXI), red rectangle depicts view of Resiquimod binding (right) in which proteins are shown at 20% transparency. The molecule bridges the receptors at the interface. TLR7 is shown in blue and orange, and Resiquimod is shown in green. (B) Chemical structures of various TLR7/8 agonists. Green arrows highlight recommended sites of conjugation.

The natural ligands for TLR7 and TLR8 are ssRNAs of viral origin.¹³⁰ Due to the poor drug properties of RNA, several small molecule agonists have been developed, including imidazoquinolines, purine derivatives, and guanosine analogues.⁷⁵ Many of these heteroaromatic molecules are agonists for both TLR7 and TLR8.¹³² However, empirical testing, crystal structure examination, and SAR studies have shed light on the structural requirements for TLR7 or TLR8 agonist specificity.^133–135 These small molecule agonists have been effectively utilized as immunostimulants in a variety of applications. One synthetic small molecule imidazoquinoline TLR7 agonist, Imiquimod, is approved for clinical use in topical treatment of certain skin cancers, genital warts, and skin conditions.¹³⁶ However, promising results in vitro are often difficult to recapitulate, and even potentially toxic, in vivo.^8,137 This observation has been attributed to the molecules’ pharmacokinetic properties as they rapidly diffuse to the bloodstream, resulting in systemic cytokine production and inflammation.³¹ Due to the practical issues, yet clinical promise, of TLR7/8 agonists, they have received considerable attention as TLR conjugates (Table 2). The basic chemical structure of these heteroaromatics has multiple sites amenable for functionalization; both a primary and secondary amine are present for conjugation, as well as an aromatic ring system capable of bearing other functional groups. Several conjugation strategies have been successfully employed, but the potency and biological effects of the conjugates varies greatly.

Table 2.

Overview of TLR 7, 8, and 9 Agonist Conjugates

TLR	agonist	Conjugation site	conjugation chemistry	conjugated to	refa
TLR7	SM360320	N1-benzyl linkage	Amide, hydrazone	Polymer, protein, phospholipid	14,18,139
		N1-alkyl linkage	Amide	Protein	179
	2-alkoxy-8-hydroxyadenyl derivative	C-2 linkage	CuAAC	Peptide	140
TLR7/8	3M-012	N1 linkage	thiol-maleimide, UV light induced coupling	Peptide, protein	150,149
	Resiquimod	C-4 linkage	Carbamate linkage	Photocage	147
	N1-benzyl imidazoquinoline	N1-benzyl linkage	Thiourea, reductive amination, amide	Fluorophore, protein, TLR7/8 agonist	142,141,148
		C-4, N1-benzyl, N1-alkyl linkage	Amide	Polymer	13
	Imidazoquinoline amino acid	N- and C- terminus	SPPS	Peptide	155
	Loxoribine	C5’ OH of ribose	CuAAC	TLR4- and 9 agonist
TLR9	Class A CpG	5’ Phosphate	Phosphoramidite	Phospholipid	20
	Class B CpG	5’ Phosphate	Thioether, Thiol-maleimide, phophoramidite, CuAAC, bis-aryl hydrazone, reductive amination, disulfide	TLR4-, 7/8 and 9 agonists, polymers, biotin, fluorophore phospholipid, protein	163,19,107,180,20,181,165,12,15
	Class B CpG	3’ Phosphate	Amide, thiol-maleimide, bis-aryl hydrazone	LLC cells, TLR2/6- and 4 agonists, peptide, protein	45,44,48,165
	Class C CpG	5’ Phosphate	Disulfide coupling, phosphoramidite	Nanoparticles, phospholipid	15,20

^aReferences are clustered and ordered by conjugation chemistry.

Most adenine analogues found in conjugate applications are tethered through a substituted N¹-benzyl linker, based on the potent TLR7 agonist 9-benzyl-8-hydroxy-2-(2-methoxyethoxy) adenine (SM360320).¹³⁸ Wu and co-workers were the first to synthesize a SM360320 derivative for conjugation by functionalizing the N¹-benzyl with an aldehyde moiety.¹³⁹ The resulting TLR7 agonist, UC-1 V150 (Figure 5B), was ligated to hydrazine functionalized mouse serum albumin (MSA), while retaining its TLR7 agonist activity. Another conjugation method, from the same lab, involved using an N¹-benzylic carboxylic acid UC-1 V150 analogue, which was conjugated to phospholipids, with or without PEG spacers.¹⁸ Adenine derivatives have also been ligated through N¹-alkyl linkages, although this resulted in less potent conjugates. Alternatively, Weterings and colleagues have described a method to covalently link a 2-alkoxy-8-hydroxyadenine derivative through an ethylene glycol linked azide on its C-2 position to OVA peptide epitopes.¹⁴⁰ This adenine derivative was conjugated to a peptide epitope through click chemistry, creating a self-adjuvanting vaccine. Although conjugation of the TLR7 agonist increased presentation of the antigen, it also abolished the induction of DC activation. The authors suggest steric hindrance of the peptide moiety may influence binding of the conjugate to the TLR.

Similar to adenine analogues, conjugation of imidazoquinolines mainly proceeds through benzylic linkages. Shukla and co-workers conjugated an N¹-(4-aminomethyl)benzyl substituted imidazoquinoline (analogue 5d, Figure 5B) to three fluorescent dyes, fluorescein, rhodamine B, and BODIPY-TR-cavarine, either directly or by first converting the free amine to an isothiocyanate.¹⁴¹ In doing so, they created potent, fluorescent probes for TLR7 activation. The same group also ligated N¹-(4-aminomethyl)benzyl substituted imidazoquinolines under more mild conditions. By converting the free amine to an isothiocyanate- and maleimide-group, they conjugated the TLR agonist to both a model peptide and reduced glutathione, in aqueous buffer, without additional reagents.¹⁴² The amino group was also reacted directly with a model oligosaccharide under reductive amination conditions. All resulting conjugates were obtained in high yield and preserved their agonistic activity.

Lynn et al. demonstrated ligation of N¹-alkyl and N¹-(4-aminomethyl)benzyl substituted (analogue 5d reported previously) TLR 7 agonists.^13,143 They showed that the benzylic linker (Figure 5B, EC₅₀: 0.1 μM) increased potency of the conjugates 20-fold compared to the alkyl linker (EC₅₀: 2 μM) inspired by previously reported SAR studies.¹⁴⁴ The benzylic imidazoquinoline compound was conjugated to an N-(2-hydroxypropyl)methacrylamide (HPMA) polymer which enhanced its immunological activity in vivo. This effect was achieved by causing accumulation in the draining lymph node instead of diffusion to the bloodstream.

Conjugation of imidazoquinolines through the C-4 amine resulted in a complete loss of agonist activity according to SAR studies¹⁴⁵ and crystal structures¹⁴⁶ of TLR8. In one example, a method to gain spatial and temporal control over immune cell activation was demonstrated by photocaging the C-4 amine of Imiquimod and Resiquimod.¹⁴⁷ The free C-4 amine of the imidazoquinolines was functionalized with 2-(2-nitrophenyl)-propyl chloroformate (NPPOC-Cl), creating a light-responsive TLR agonist. TLR7/8 agonist dimers linked through their C-4 amine also display sharply decreased TLR7 agonist activity compared to N¹-benzyl linked dimers.¹⁴⁸

Willie-Reece and co-workers demonstrated the promise for TLR7/8 agonist-antigen conjugates as a subunit HIV vaccine.¹⁴⁹ An amine functionalized resiquimod (3M-012) was photo-cross-linked with HIV Gag protein. The conjugate was shown to potently stimulate T and B cell responses, in mice and nonhuman primates, toward the antigen, whereas the unconjugated mixture did not. A thiol-functionalized 3M-012 was also cross-linked via thiol-maleimide chemistry to an HIV envelope protein.¹⁵⁰ Despite high immunostimulatory activity, the conjugation blocked antibody responses toward certain antigen epitopes. Recently, the same lab demonstrated that protective T cell responses via TLR7/8-antigen conjugates required aggregation of the conjugate.¹⁵¹ In addition, Holbrook and co-workers conjugated an amine functionalized Resiquimod to influenza virus via a bifunctional NHS-maleimide PEG linker.¹⁵² This and follow-up studies demonstrated the Resiquimod-influenza conjugate vaccine generated efficacious influenza immunity in nonhuman primate neonates, a difficult vaccination model.^153,154 These studies highlight both the clinical potential of TLR7/8 agonist conjugates and the need to development site-specific methods for agonist conjugation to improve antigen-specific immune responses while preserving antigen recognition.

Further broadening the scope of conjugation strategies for TLR7/8 agonists, Fujita and co-workers synthesized an imidazoquinoline-like amino acid, 6-(4-amino-2-butyl-imidazoquinolyl)-norleucine.¹⁵⁵ Although weakly potent, the novel amino acid provides a way to easily introduce imidazoquinolines into peptides via SPPS. To create a more potent TLR7/8 agonistic amino acid we propose substituting norleucine for phenyl alanine to create 4-((4-amino-2-butyl-imidazoquinolyl)-methyl) phenylalanine, which contains the N¹-benzyl substituent associated with high TLR7/8 agonist activity. Such a highly potent imidazoquinoline-like amino acid would enable facile synthesis of powerful self-adjuvanting vaccines targeting TLR7/8 through SPPS.

Wu and co-workers have developed a guide to effectively modify imidazoquinoline-like TLR7/8 agonists to create small molecule immunopotentiators (SMIPs).³³ They showed that creating an antigen depot effect was key for in vivo adjuvant effectiveness by limiting “wasted inflammation” and directing immune responses toward intended targets. This effect was achieved by making the molecules more lipophilic through conjugation with hydrophobic moieties. However, they found these conjugates were impractical due to scale and formulation issues caused by poor conjugate solubility. Instead, the group found that by conjugating SMIPs to phosphate-functionalized PEG linkers followed by noncovalent absorption with alum, the depot effect was achieved, yielding effective adjuvants.

Loxoribine is the only guanosine analogue to yet be conjugated. In an effort to synthesize a trivalent TLR agonist conjugate, a Loxoribine derivative amenable to conjugation was synthesized by substituting the 5′ primary alcohol for an azide (Figure 5B).¹⁰⁷ The Loxoribine-azide was conjugated to an alkyne-functionalized core molecule through copper(I)-catalyzed alkyne–azide cycloaddition (CuAAC). However, whether this conjugation strategy preserves TLR7/8 agonist potency remains uncertain. Conjugation of Loxoribine to a diagonist of CpG ODN (TLR9) and indole (TLR4) did not increase the conjugate’s potency, which we attribute to the relatively low potency of Loxoribine.

In summary, TLR7/8 agonists are synthetically accessible, relatively potent, and highly amenable to conjugation. Conjugation of TLR7/8 agonists to phospholipids and polymers has proven to be an effective strategy in improving their pharmacodynamics/pharmacokinetic properties, and their conjugation to protein and peptide antigens creates highly immunogenic self-adjuvanting vaccines. In addition, the synthesis of TLR7/8 agonist conjugates is effective as the compounds are more amenable to synthetic methods than lipidated TLR4 agonists. Together, these qualities make TLR7/8 agonists appealing immunostimulants for bioconjugation in many applications.

TLR9.

TLR9 is located in the endosome and senses DNA rich in unmethylated CpG dinucleotides. TLR9 also binds synthetic CpG oligodeoxynuceotides (ODNs), which are short, single-stranded DNA (ssDNA) oligomers with unmethylated CpG motifs (Figure 6A). CpG ODNs induce strong cellular and humoral responses against coadministered antigens, making them highly attractive vaccine adjuvants.¹⁵⁶ The sequence of the CpG ODN dictates the species specificity of the agonist, with many commercially available varieties developed for human, murine, bovine, and other organisms.

Figure 6.

TLR9 agonists are ssDNAs that bind TLR9 dimers. (A) Two views of the horse TLR9-ssDNA (CpG 1668) crystal structure (PDB: 3WPC). Shown below each view is the zoomed region depicted by the red rectangle where TLR9 is shown at 40% transparency. The receptors bind to the CpG backbone and several nucleobases of the molecule. TLR9 is shown in blue and orange, and ssDNA is shown in green. (B) Carton structure of CpG. Green arrows highlight sites of conjugation.

CpG ODNs have been used in bioconjugates for a variety of applications; CpG ODNs have been conjugated to peptide antigens,⁴⁸ antibodies,¹⁵⁷ other TLR agonists,^43,44,107 nanoparticles,^15,158 cancer cells,⁴⁵ and lipids^19,20 (Table 2). Conjugation of CpG ODNs to nanoparticles and lipids can improve pharmacokinetics, pharmacodynamics, and targeting of the adjuvant to draining lymph nodes.^12,20 Conjugation of CpG ODNs to antigens can improve antigen uptake, antigen presentation,⁴⁸ and cross-priming¹⁵⁹ of cytotoxic T-lymphocytes. However, some have noted that promising results in animal models may not translate to humans, as TLR9 expression across human immune cell types are more limited.¹⁶⁰ Exciting for the field, the first vaccine with unconjugated CpG (Heplisav) was just approved for use in humans indicating the vast potential of this class of compounds.

Many studies report conjugates of CpG ODN linked through either the 3′ or 5′ end (Figure 6B). However, the optimal orientation for CpG ODN in conjugates remains controversial. The crystal structure of the extracellular domain of horse TLR9 ligated to CpG suggests that the 5′ end may be more accessible for conjugation (Figure 6B). It should be noted however that the 5′ end is positioned at the bottom of the TLR9 complex, toward the endosomal membrane, which could inflict steric constraints on 5′ end CpG conjugates. Crystal structure analysis has been insufficient to determine optimal binding sites as suggested by SAR studies. Initial reports by Agrawal and co-workers suggested that the 3′ end was ideal for conjugation and that a free 5′ end was essential for agonist potency.¹⁶¹ For example, CpG dimers that were conjugated 5′-to-5′ were significantly less immunogenic than 3′-to-3′ tethered CpG ODNs, measured by the in vitro induction of spleen lymphocyte proliferation and splenomegaly.¹⁶¹ In a follow-up study, this decrease in TLR9 potency for 5′ conjugation was shown to be dependent on the size of the attached moiety and independent of cellular uptake.¹⁶² In vivo, 5′ conjugation of CpG ODNs to a bulky amyloid-β peptide antigen sharply decreased IL-12 induction, while 3′ conjugates elicited strong IL-12 responses in C57BL/6 mice.¹⁶³

However, some studies contradict the notion that 5′ conjugation of CpG ODNs leads to reduced TLR9 potency compared to 3′ conjugation. Zhang et al. showed that 5′ conjugation of class B CpG ODN 1826 to bulky dextran polymers did not alter NF-κB activity in TLR9-transfected HEK293 cells.¹² It is, however, important to note that an acid labile imine bond¹⁶⁴ was used to conjugate CpG to dextran in this study, which could facilitate release of CpG after endosomal uptake. Acid hydrolysis of the imine bond would free the 5′-end of CpG and allow unobstructed binding to TLR9, which would explain why CpG conjugated to dextran retained TLR9 potency. We have, however, also observed (2-fold) decreased TLR9 activity upon 3′ CpG conjugation via thiol-maleimide chemistry with a triazine linker, while activity was maintained when conjugated through the 5′ end (unpublished data). In an effort to end the controversy, Kramer et al. recently compared conjugation of CpG ODN 1668 to an ovalbumin (OVA) protein antigen through both 3′ and 5′ ends using an acid-stable bis-aryl hydrazone linker (Solulink Inc.).¹⁶⁵ Interestingly, no statistically significant difference was found in the induction of proliferation of CD4⁺ and CD8⁺ T- cells and IFN-γ production in T-cells from OT-I and OT-II mice between the 3′ and 5′ CpG-OVA conjugates.¹⁶⁵ This study suggests that both 3′ and 5′ conjugation of CpG ODNs yield conjugates with comparable potency for induction of cellular T-cell responses, and therefore, in some cases 5′ conjugation does not lead to reduced TLR9 potency of CpG.

Conjugation of CpG ODNs is easily achieved by synthesizing (or ordering) custom CpG ODNs containing reactive groups on the 3′ or 5′ termini via automated phosphoramidite DNA synthesis. In terms of structural requirements for CpG ODN conjugation, both 3′ and 5′ conjugation of CpG ODNs can yield functional conjugates, albeit with reduced potency for some reported 5′ conjugates. Recent studies show that 5′ conjugation of CpG does not necessarily lead to decreased potency, although other studies suggest that a free 5′ end of CpG is vital for immune stimulation. Thus, either 3′ or 5′ linkage strategies are appropriate for most conjugates. However, TLR9 activity of conjugates should be confirmed prior to application.

CONCLUSIONS

TLR agonists have demonstrated their therapeutic potential in a variety of applications, including approved vaccine adjuvants and cancer immunotherapies. However, off-target effects, like systemic inflammation, have limited their practicality in the clinic. Covalent conjugation of TLR agonists to other functional compounds can circumvent these issues by generating effective, directed immune responses. Although only one TLR agonist conjugate, Trumenba, has been FDA approved, these compounds have shown immense promise and additional studies will advance their implementation.

Production of TLR agonist requires careful design to ensure efficacy. Through case studies with SAR analysis, TLR-agonist crystal structure evaluation, and molecular modeling, we have learned important considerations when designing TLR agonist conjugates. These considerations include the choice of agonists, choice of linkage chemistry, selection of conjugation sites that maintain agonist activity, and scalability considerations. TLR agonist discovery and chemical insights in effective TLR agonist conjugation will accelerate the ongoing developments at the forefront of vaccine and immunotherapy development.

ABBREVIATIONS

TLR

toll-like receptor

PRR

pattern recognition receptor

PAMP

pathogen-associated molecular pattern

DAMP

damage-associated molecular pattern

dsRNA

double-stranded RNA

ssRNA

single-stranded RNA

CpG

cytosine-phosphate-guanosine

APC

antigen presenting cell

DC

dendritic cell

LTA

lipoteichoic acid

ODN

oligodeoxynucleotides

LLC

Lewis lung carcinoma

NHS

N-hydroxysuccinimide

SPPS

solid phase peptide synthesis

TACA

tumor associated carbohydrate antigen

NCL

native chemical ligation

SAR

structure activity relationship

HPV

human papillomavirus

poly(I:C)

polyinosinic-polycytidylic acid

OVA

ovalbumin

cDC

conventional DC

MD2

myeloid differentiation factor 2

LPS

lipopolysaccharide

MPLA

monophosphoryl lipid A

GLA

glucopyranosyl lipid adjuvant

SLA

second-generation lipid adjuvant

CDI

carbonyldiimidazole

NIPAM

N-isopropylacrylamide

NVOC

nitroveratryloxycarbonyl

NPPOC

2-(2-nitrophenyl)propyl chloroformate

SMIP

small molecule immunopotentiators

CuAAC

copper(I)-catalyzed alkyne–azide cycloaddition