Structure-Based Design of Novel TLR7/8 Agonist Payloads Enabling an Immunomodulatory Conjugate Approach

Abstract

Dual activation of the TLR7 and TLR8 pathways leads to the production of type I interferon and proinflammatory cytokines, resulting in efficient antigen presentation by dendritic cells to promote T-cell priming and antitumor immunity. We developed a novel series of TLR7/8 dual agonists with varying ratios of TLR7 and TLR8 activity for use as payloads for an antibody–drug conjugate approach. The agonist-induced production of several cytokines in human whole blood confirmed their functional activity. Structure–activity relationship studies guided by structure-based drug design are described.

Toll-like receptors (TLRs) are receptors for pathogen-associated molecular patterns (PAMPs) or danger-associated molecular patterns (DAMPs) that recognize viral or bacterial signatures through pattern recognition receptors (PRRs).¹⁻³ Among 10 PRRs, TLRs 1, 2, 4–6, and 10 are expressed on the cell surface, whereas TLRs 3 and 7–9 are located in the endosomal compartment of various immune cells. The endosomal TLR7 and TLR8 have differentiated but overlapping expressions. TLR7 is expressed primarily in plasmacytoid dendritic cells (pDCs), monocytes, macrophages, and B cells. It is responsible for the induction of type 1 interferon (IFN) and IFN-induced cytokines such as IP-10.⁴ Conversely, TLR8 is expressed in conventional dendritic cells (cDCs), monocytes, macrophages, neutrophils, and mast cells. It leads to the production of cytokines (TNF-a, IL-12, IL-1b, IL-6, MCP-1, and IL-10). Several reports have disclosed small-molecule ligands for the selective activation of the TLR7⁵⁻⁷ or TLR8⁸⁻¹⁰ pathway for treating immunological disorders and cancers. Dual activation of the TLR7/8 pathway is believed to enhance the recruitment of a broader population of immune cells to the tumor microenvironment (TME), resulting in stronger antitumor immunity.¹¹⁻¹³ However, systemic activation of the TLR7/8 pathway has been associated with cytokine release syndrome (CRS), resulting in severe toxicity.¹⁴ As a result, recent efforts have been focused on tumor-specific delivery of these agents using nanoformulation¹⁵ or antibody–drug conjugate (ADC) approach.¹⁶⁻²¹

Recently, we disclosed a novel series of TLR7-selective agonists based on the pyrazolepyrimidine core.^22,23 Some of these compounds (e.g., compound 1, Figure 1) displayed modest TLR7/8 dual activity in the biochemical assay. Based on the X-ray cocrystal structure of 1 with monkey TLR8 (which bears high homology with human TLR8 that is not readily available), we realized that the pyrazole ring of these agonists sits in a hydrophobic pocket unique to TLR8 (red circle, Figure 1). We hypothesized that filling this pocket with a lipophilic group, such as a phenyl ring, may result in a productive interaction with TLR8 such that the TLR8 activity can be increased. This hypothesis prompted a systematic structure–activity relationship (SAR) campaign (Figure 2). Unfortunately, when pyrimidoindole compound 3 was synthesized by adding a phenyl ring to the core of previously identified compound 2(22) and evaluated in a reporter assay, it was inactive at both TLR7 and TLR8 up to 5 μM concentration. We understood from our previous SAR work on a related TLR7 series that a methoxy group at the 2-position of the pendant benzyl substituent was essential for activity. That was also the case in the pyrimidoindole series, where compound 4 restored some TLR7 activity but no TLR8 activity was observed. When an amine group was installed on the benzyl side chain to potentially help sequester the drug in the acidic environment of the endosome/lysosome where TLR7/8 are expressed, a dramatic improvement in the TLR7 potency and some improvement in TLR8 activity were observed in 5. Further, when the 3-aminohexanol side chain was installed at the C-7 position of the pyrimidoindole, compound 6 maintained TLR7 potency with an improvement in TLR8 agonism. However, there was a greater than 300-fold difference in activity toward these two receptors.

Figure 1.

X-ray cocrystal structure of compound 1 with monkey TLR8.

Figure 2.

Key initial SAR for TLR7/8 activity.

While compound 6 displayed remarkably potent TLR7 activity, we believed there was room for further improvement in TLR8 activity, for which the structural features of compound 1 were analyzed for application toward the pyrimidoindole series. We realized that, unlike the pyrimidoindole series, compound 1 contained a unique trans-4-amino-tetrahydrofuran-3-ol moiety at the meta position relative to the benzyl linker. Additionally, this compound displayed more balanced TLR7/8 activity (TLR7:8 ratio of ∼1:5). The pyrimidoindole analog 6 afforded the most potent TLR7 activity. In contrast, the pyrazolepyrimidine series displayed balanced TLR7/8 potency; we envisioned that combining these two series through a scaffold hopping approach may increase TLR8 potency with more balanced TLR7/8 activity. When the structural features of these series were combined in compound 7, a 4-fold improvement in TLR8 potency was observed, albeit with loss of some TLR7 activity (Figure 3) compared to compound 6. However, a more balanced profile was observed with about a 13-fold difference between the TLR7 and TLR8 activities. A direct comparison of compounds 1 and 7 showed a remarkable ∼30-fold increase of TLR8 activity in compound 7. This result validated our original hypothesis that filling the hydrophobic pocket in Figure 1 may boost the TLR8 activity. We were also excited to see an increase in TLR7 potency that can be attributed to filling a similar hydrophobic pocket in the TLR7 structure (vide infra). When we cocrystallized compound 7 with monkey TLR8, it was confirmed that the lipophilic pocket (red circle) was filled with a fused phenyl ring of pyrimidoindole 7, which is hypothesized to result in the improvement in TLR8 potency.

Figure 3.

X-ray of compound 7 with monkey TLR8.

Since the program aimed to identify a TLR7/8 dual agonist payload for targeted delivery of these immunomodulatory agents via conjugation to an antibody as immunomodulatory conjugate (IMC) approach, we were excited by the prospect of using compound 7 as a tool payload. The secondary amine in 7 not only served the purpose of lysosomal sequestration for prolonged target engagement with TLR7/8 but also provided a handle for attaching a linker through the formation of a carbamate bond using the valine-citrulline para-aminobenzyl group. Unfortunately, our attempts to attach a linker through a carbamate bond were unsuccessful because of the amine’s diminished reactivity on the hydroxy tetrahydrofuran (THF) group that was both sterically encumbered and electronically deactivated. Other linker attachment points were considered, such as alcohol on the C-7 side chain or the amine on the 2-aminopyrimidine core. These explorations were met with much less success due to the presence of a regioisomeric mixture of products when alcohols were used as a handle to attach a linker or an unreactive 2-aminopyrimidine. This led us to explore other, less sterically-hindered amines that would be more amenable for linker work to generate IMC derivatives. Additionally, we wanted to understand more if there was anything unique about this hydroxy THF moiety that provided the high TLR7/8 potency observed. Systematic SAR was developed with the capping amines where a cis-hydroxy THF 8 displayed ∼2-fold potency loss compared to the trans isomer 7 (Table 1). Enantiomers of both cis and trans isomers had similar potency levels (data not shown). Removing the hydroxy group in 9 maintained potencies in the TLR7 and TLR8 reporter assays. The R-isomer 10 was found to be slightly less potent than the S-isomer 9. Since secondary amines resulting from the direct attachment of THF amines are sterically hindered, we introduced a methylene spacer between the amine and THF ring to reduce steric hindrance and increase amine reactivity (compound 11). However, it was much less potent in TLR7 and TLR8. Finally, we also explored a variety of amines (data not shown), including aliphatic and aromatic groups such as in compounds 12 and 13, respectively, which exhibited a significant loss of potency toward TLR8. Through this SAR work, we identified less sterically encumbered amines, which we believed would be more reactive for linker chemistry. While our primary focus was to generate potent activity in the reporter assay, we also monitored other parameters, such as metabolic stability, T-cell toxicity (PHA blast assay), CYP, and hERG inhibition.

Table 1. SAR Studies on the Terminal Amine Group.

Next, we focused on modifying the C-7 side chain, which has been shown to impact the potency and selectivity of TLR7 agonists (Table 2).²² Truncation of C-7 alcohol to a hydroxymethyl group provided compound 14, which was equipotent for TLR7 and TLR8. This observation was especially remarkable because it provided access to an agonist equipotent for TLR7 and TLR8 that was not achieved through benzylamine modification. Interestingly, the R-isomer 15 was significantly less potent for both receptors than the S-isomer, indicating the importance of stereochemistry in this region of the molecule. Branching or shortening of the alkyl side chain, as in compounds 16 and 17, led to a significant loss of TLR7 and complete loss of TLR8 activity, indicating a narrow hydrophobic pocket optimal for an n-butyl chain. Finally, the transposition of the hydroxyethyl group to nitrogen to generate a tertiary amine was not tolerated because it was unable to form an intramolecular H-bonding interaction with the methoxy group observed in the X-ray cocrystal structure.

Table 2. Modification of Substituents at the C-7 Position.

Since the newly identified C-7 side chain in analog 14 provided the most balanced TLR7/8 biochemical activity, we explored this series further by modifying the benzylamine part (Table 3). Once again, a similar trend was observed when a simple THF amine was used, where compound 19 provided balanced TLR7/8 activity, albeit with a slight loss of potency. However, we were pleasantly surprised when a nitrogen atom was introduced in the phenyl ring in the proximity of the methoxy group in compound 20. This effort was meant to lower the overall lipophilicity, which was found to correlate with T-cell cytotoxicity.²³ As the project aimed to achieve T-cell-mediated tumor cell killing, we intended to minimize any T-cell cytotoxicity measured by PHA blast activity. However, this modification led to a reversal in TLR7/8 activity, where more potent TLR8 activity was observed. Although we could not rationalize this behavior from a structural point of view, it certainly provided access to a new series of compounds with a more potent TLR8 activity. Further expansion of SAR on this series by incorporating various amines held a similar trend, with more potent TLR8 activity. It turned out that this trend was only observed in a series of benzylamine analogs. In a new series where the benzylamine is replaced by unsaturated piperidines attached to the aryl ring, the TLR8 selectivity was diminished, and these analogs were slightly more potent in the TLR7 reporter assay, although much less potent overall.

Table 3. SAR Leading to Diverse TLR7/8 Ratio.

With the identification of a panel of agonists displaying a spectrum of TLR7/8 agonist activity and selectivity, we obtained another X-ray cocrystal structure. A close structural analog of early hit 7 without the sterically demanding hydroxy group (compound 9) was equipotent. Still, we were curious to study if the absence of the hydroxy group resulted in a different binding mode between TLR7 and TLR8. Compound 9 was cocrystallized with human TLR8 and monkey TLR7 (Figure 4). In the TLR8 cocrystal structure, two π–π interactions were observed between (a) the tricyclic core and Phe405 and (b) the benzylamine side chain and Tyr353. The 2-aminopyrimidine group was involved in a bifurcated H-bonding interaction with Asp543 and Thr574. The n-propyl group on the C-7 side chain sits in a deep hydrophobic pocket, whereas the hydroxy group interacts with Tyr348. Additionally, the N–H group on C-7 is a hydrogen bond donor for an intramolecular H-bond with the methoxy group on the benzyl side chain. The terminal THF amine group points toward the solvent-exposed region. This can explain why the lack of hydroxy group on THF of 9 did not lead to any deterioration of potency. It is interesting to note that the binding modes of compounds 7 and 9 are identical in the TLR8 structure, which can explain their similar potency in a reporter assay. Except for the difference in amino acid sequence, all these key interactions were also preserved in the TLR7 crystal structure, confirming structural engagement with TLR7 and TLR8.

Figure 4.

Co-crystal structure of 9 with human TLR8 (a) and monkey TLR7 (b).

Next, we evaluated a few selected examples of agonists with differentiated profiles in a functional assay in human whole blood by measuring the induction of various cytokines (Figure 5). Agonists with various degrees of potency and selectivity were incubated with fresh human whole blood for 4 h, and multiple cytokines were measured. We primarily focused on cytokines that are downstream markers of either TLR7 (e.g., IFNa) or TLR8 (e.g., TNFa). All representative compounds displayed potent induction of cytokines in human whole blood. There was some trend between TLR7/8 selectivity in reporter activity and functional activity; i.e., those more active in TLR7 tended to induce higher levels of IFNa (e.g., compound 13). In contrast, analogs with more potent TLR8 activity induced higher levels of TNFa (e.g., compound 23). Although a correlation could not be generated due to the difference in absolute potency in reporter assay and overlapping pattern of cytokines elicited by TLR7 and TLR8 pathway, as well as a level of donor-to-donor variability, induction of cytokines confirmed the functional activity of these agonists.

Figure 5.

Selected agonists for functional activity (a) in human whole blood: IFNα (b) and TNFα (c).

With the identification of potent TLR7/8 dual agonists, we revisited our effort to attach a linker to some of these payloads to generate immunomodulatory conjugates (IMCs) (Scheme 1). Commercial 2-cyanoaniline was alkylated to yield compound 27, which was then cyclized under basic conditions to provide substituted indole 28. The primary amine was then converted into a guanidine intermediate, which was cyclized to generate a tricyclic core containing a 2-aminopyrimidine pharmacophore (compound 30). Conversion of phenol to chloride followed by alkylation of indole nitrogen with substituted benzyl bromide resulted in a benzyl side chain installation. The ester group on 34 was selectively reduced using lithium diisobutyl-tert-butoxyaluminum hydride (LDBBA) to generate benzyl alcohol 35. Compound 35 served as a common intermediate for diversification at the C-7 position or at the benzyl group. For example, the chloride was displaced with amine 36, and after deprotection of the TBDPS group, a bis-alcohol 37 was obtained. Selective oxidation of benzylic alcohol followed by reductive amination provided compound 9. This sequence allowed late-stage diversification of the benzyl group by using various amines. A similar sequence with different C-7 amines was used to generate the analogs shown in Tables 2 and 3. For the installation of the linker, unlike earlier compound 7, where we had difficulty in forming a carbamate bond between the THF amine and linker, a des-THF hydroxy analog 9 was much more reactive, and a linker-payload 39 was generated with a modified valine-citrulline linker (details in the Supporting Information). The linker-payload was then conjugated to the LIV1 antibody using cysteine-maleimide chemistry, utilizing eight thiols generated after the reduction of interchain disulfide bonds to provide LIV1-IMC with eight drugs per antibody. The IMCs were evaluated in tumor-monocyte coculture studies using LIV1-positive MCF7 tumor cells to produce cytokines (Figure 6). The LIV1-IMCs potentiated induction of a downstream marker of the TLR7/8 pathway (e.g., PD-L1) compared to the isotype control (KLH IMCs), confirming LIV1-dependent functional activity.

Scheme 1. Synthesis of Linker-Payload and LIV-1 IMC.

Figure 6.

In vitro characterization of LIV1-IMC.

In summary, using a structure-based drug design approach, we developed a novel series of TLR7/8 dual agonists with various potency levels and TLR7/8 selectivity. The induction of cytokines in the human blood confirmed these agonists’ functional activity. Furthermore, cocrystals of selected compounds with TLR7 and TLR8 verified structural target engagement. Selected analogs were used to attach linkers and generate LIV1-IMCs, which displayed LIV1-dependent in vitro activity. Further characterizations of lead IMCs for cleavability, stability, in vitro activity, in vivo efficacy, and pharmacokinetic/toxicology studies in cynomolgus monkeys will be reported separately in a full disclosure.

Safety Information

The compounds prepared herein are potent immunostimulatory agents and should be handled with the appropriate personal protective equipment and engineering controls.

Glossary

Abbreviations

ADC

antibody–drug conjugate

IMC

immunomodulatory conjugate

PAMP

pathogen-associated molecular pattern

DAMP

danger-associated molecular pattern

pDC

plasmacytoid dendritic cell

TME

tumor microenvironment

CRS

cytokine release syndrome

LDBBA

lithium diisobutyl-tert-butoxyaluminum hydride

DIPEA

diisopropylamine

NMP

N-methylpyrrolidone

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.4c00463.

• Full experimental details for key compounds, biological assay protocols, and crystallography details (PDF)

The authors declare no competing financial interest.