Chemically Tuning the Antigen Release Kinetics from Spherical Nucleic Acids Maximizes Immune Stimulation

Abstract

Cancer vaccine structure is emerging as an important design factor that offers tunable parameters to enhance the targeted immune response. We report the impact of altering the antigen release rate from spherical nucleic acid (SNA) vaccines—nanoparticles with a liposomal core and surface-anchored adjuvant DNA—on immune stimulation. Peptide antigens were incorporated into SNAs using either a nonreducible linker or one of a series of reduction-triggered traceless linkers that release the native peptide at rates controlled by their substitution pattern. Compared with a nonreducible linkage, the traceless attachment of antigens resulted in lower EC₅₀ of T cell proliferation in vitro and greater dendritic cell (DC) activation and higher T cell killing ability in vivo. Traceless linker fragmentation rates affected the rates of antigen presentation by DCs and were correlated with the in vitro potencies of SNAs. Antigen release was correlated with the ex vivo −log(EC₅₀), and more rapid antigen release resulted in an order of magnitude improvement in the EC₅₀ and earlier and greater antigen presentation over the same time-period. In vivo, increasing the rate of antigen release resulted in higher T cell activation and target killing. These findings provide fundamental insights into and underscore the importance of vaccine structure.

Short abstract

The rate of antigen release from spherical nucleic acid vaccines following a reduction trigger can be controlled through traceless linker chemistry and impacts in vitro and in vivo immune responses.

Introduction

Development of immunotherapies that activate the immune system to target cancer cells, in particular cytotoxic CD8⁺ T cells, is a major focus in the field of cancer therapeutics.¹ One prominent approach is the delivery of a peptide antigen and an adjuvant to professional antigen-presenting cells (APCs), such as dendritic cells (DCs), which causes them to mature and activate downstream immune pathways, including CD8⁺ T cell activation and proliferation.^2,3 Spherical nucleic acids (SNAs) are nanoscale architectures that have been shown to elicit such immune responses in small animals and humans.⁴⁻⁷ Immunostimulatory SNAs consist of an adjuvant DNA strand anchored to the surface of a nanoparticle core with peptide antigens incorporated into the structure. SNAs are readily taken up by DCs and can activate them toward effective T cell priming.⁸ The impacts of various structural features of SNAs—such as the identity of the surface oligonucleotides, location of antigen incorporation, or anchor strength—on immune stimulation have been systematically explored in an approach to vaccine design termed “rational vaccinology”.^5,8−10 These studies provide a firm foundation for the development of the SNA platform and exploration of its fundamental interactions with immune cells but also demonstrate the applicability of structure–function relationships to immunotherapeutic constructs in general.

The structures of immunostimulatory therapeutics are emerging as critical variables that determine their efficacy, and chemists have developed a variety of tools to access the most potent therapeutic structures.^5,8,11−16 Timing of immunotherapeutics is an important factor in therapeutic efficacy, including the rate of antigen release inside DCs, which can alter antigen processing pathways and impact downstream immune activation.^8,11,17−21 We hypothesized that the antigen release kinetics plays a critical role in the efficacy of immunostimulatory SNAs and that our ability to control these rates would provide a powerful method for increasing their therapeutic potential. While covalent disulfide-based attachment of antigens to therapeutic constructs is a common strategy, none have controlled the release rates of antigens inside cells.^5,22−29 We hypothesized that incorporating chemical modifications to the traceless linker backbone would enable control over the rate of antigen release from the SNA inside cells. In addition to impacting clinical translational potential, elucidating the relationship between the kinetics of antigen release from SNAs and their immune processing will illuminate fundamental biological interactions that could be applied to other therapeutic constructs.

Herein we study how small chemical modifications to the traceless linker structure alter the rates of chemical dissociation of antigens from DNA–peptide conjugates. A series of SNAs with reduction-labile traceless linkers and differing fragmentation kinetics were compared with one another and an SNA with a noncleavable linkage for their ability to elicit targeted immune responses. We found not only that the traceless attachment chemistry produces more potent vaccines compared with the noncleavable conjugation but also that the rates of dissociation significantly impact the biological responses. A novel traceless linker design is described that liberates the antigen at a higher rate than an unmodified linker structure and results in the most potent SNA immune stimulation. These findings present a path for the design of traceless linkages that improve the potency of constructs with covalently conjugated antigens. Finally, we show that the chemistry used to attach the peptide antigen to SNAs is relevant in vivo, where the trends observed in vitro are reproduced in DC activation and antigen-specific T cell killing of cancer cells. Taken together, these findings advance our fundamental understanding of how APCs process SNAs and present design principles for improving the efficacy of immunotherapeutics.

Rational vaccinology recognizes that while the choice of antigen and adjuvant components is important, their structural arrangement is equally critical. The way that the antigen and adjuvant are presented to the immune system can impact processing and downstream signaling, and therefore, it is important to build structures that systematically explore how the architecture affects downstream immune activity so that it can be harnessed in the design of therapeutics. While we have used the SNA platform to explore some of these relationships, a focused examination of the effects of the antigen release kinetics from an immunostimulatory construct has yet to be reported. The findings presented herein validate the approach of rational vaccinology and demonstrate the impact that it can have in leading to the most optimal immunostimulatory constructs.

Results and Discussion

Synthesis of SNAs and Characterization of Traceless Fragmentation Kinetics

SNAs used in this study were synthesized using a 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) liposome core (∼44 nm diameter) with duplexed DNA anchored to its surface via a cholesterol moiety (Scheme 1A). The DNA duplex is composed of an adjuvant strand (class B CpG murine TLR9 ligand) with a 3′-cholesterol modification that is hybridized to a complementary strand with the antigenic peptide conjugated to its 5′ terminus (see Tables S1 and S2 for DNA and peptide sequences). The ubiquitous ovalbumin system was chosen as a model, with the MHC-I restricted epitope OVA-1 (OVA_257–264, sequence SIINFEKL) serving as the antigen. The synthesis of the conjugates was accomplished by modifying the DNA with a thiol and attaching a linker to the peptide prior to cleavage from the solid support; the two components are joined via a disulfide (reduction cleavable) or thioether (not readily cleavable) linkage in the final step of the synthesis (Scheme 1C). The constructs were purified using preparative polyacrylamide gel electrophoresis. The DNA sequences, detailed synthesis protocol, and characterization data are given in the Supporting Information (Figures S1–S7, S12, and S13 and Tables S1 and S2).

Scheme 1. (A) Schematic of an SNA Bearing Oligonucleotide Duplexes Attached to a Liposomal Nanoparticle Core through a Hydrophobic Cholesterol Anchor; (B) Schematic Showing the Chemical Structure of the Traceless Linkage between a CpG-Complement DNA (green) and a Peptide Antigen (blue); (C) Five-Step Synthesis of the Conjugates; (D) Structures of the Traceless and Nonreducible Conjugates Used in This Study (Images Not Drawn to Scale).

In (A), each duplex consists of an adjuvant CpG-motif strand (blue) that is modified with a hydrophobic cholesterol moiety on the 3′ terminus and hybridized to a complementary DNA strand (green), which is conjugated to a peptide antigen through a covalent linkage (red) on the 5′ terminus.

In (B), the disulfide bond is cleaved with rate constant k₁ under reducing conditions, while cyclization with rate constant k₂ regenerates the native peptide amine and 1 equiv of carbon dioxide and thiirane derivative.

In (C), the five steps of the synthesis include the following: (i) addition of the disulfide to the amine-containing DNA using SPDP (n = 1, R³ = H) or diMe-SPDP (n = 2, R³ = Me) via an amide bond; (ii) reduction of the disulfide to create a reactive thiol; (iii) coupling of the linker (traceless linker or BMPS) to the N-terminal amine of the solid support-bound peptide; (iv) cleavage from the support and deprotection under acidic conditions; (v) coupling of the DNA and peptide components through thiol–disulfide exchange or Michael reactions.

The DNA–antigen conjugates contain either a traceless linkage (i.e., a linkage that releases a chemically unmodified native antigen) or a nonreducible N-β-maleimidopropyloxysuccinimide ester (BMPS) linkage (which contains not readily cleavable amide and thioether bonds). While the thioether linkage derived from BMPS is not readily reduced by free thiols, it does not represent an indefinitely stable conjugate; similar linkages are known to degrade over time but much more slowly than the traceless linkages are reduced under the relevant conditions.⁴⁰ The traceless linker fragmentation occurs through two steps: disulfide bond cleavage followed by intramolecular cyclization that liberates the native peptide amine, carbon dioxide, and a thiirane derivative (Scheme 1B).³¹ The rates of disulfide cleavage and intramolecular cyclization can be modulated by the chemical substitution pattern of the traceless linker. DNA–antigen conjugates used one of four traceless linkages—HPH, HHH, HHM, or MHM—named after the substituents at positions R¹, R², and R³ (e.g., HHM bears R¹ = H, R² = H, and R³ = Me; Scheme 1D).

To understand the impact of the substitution pattern on the rate constants of cleavage (k₁) and cyclization (k₂) (Scheme 1B), we measured the apparent first-order rate constants of both processes (Figures S8 and S9). The rates were quantified by incubating each traceless conjugate (15 μM) in phosphate buffer (pH 7) with a large excess of glutathione (GSH) (20 mM) and interrogating the solution by UPLC-MS every 6 min. The areas under the curve (AUCs) of the intermediate (Figure 1A, red box) and native peptide (Figure 1A, blue box) from at least three independent replicate reactions are shown. An initial inspection of the data reveals differences in both cleavage and cyclization rates, with HPH having the highest rates and MHM the lowest rates. To quantify these observations, global fits of the intermediate and native peptide AUCs were performed using an irreversible two-step kinetic model (solid red and blue lines in Figure 1B) to obtain apparent rate constants k₁ and k₂ (see the Supporting Information for equations and mathematical fit details and Figures S8–S10). The rates and 95% confidence intervals of the fits are shown in Figure 1C (see Table S3 for additional fit parameters with confidence intervals and Figure S10).

Figure 1.

Rates of cleavage of traceless linker derivatives. (A) DNA–peptide conjugates made with traceless linker derivatives were incubated with 20 mM GSH (reduced glutathione) in phosphate buffer. At least three independent reactions were carried out for each conjugate. (B) Intermediate thiol (red box and data points) and native peptide (blue box and data points) were measured over time as AUCs of the UPLC-MS traces. Solid lines show the global fits, and shading shows the 95% CI of the fit across both data sets and all replicates. (C) Fitted apparent rate constants of cleavage (k₁) and cyclization (k₂) of each conjugate. Error bars indicate the 95% CIs of the parameters.

The differences in rates can be understood on the basis of the substituents at positions R¹, R², and R³. The HHM linkage displays a rate of disulfide cleavage similar to that of unsubstituted HHH because of the lack of α-substitutions about the disulfide at R¹ and R². However, the cyclization rate of the HHM conjugate is dramatically reduced because of steric hindrance due to the methyl substitution at the electrophilic carbon R³, which is involved in the intramolecular cyclization. The MHM linkage undergoes reduction at a lower rate than HHH, and we hypothesize that this is due to the greater steric hindrance about the disulfide and the substitution at R¹, which favors unproductive GSH attack that results in a glutathione–peptide disulfide and requires a second thiol–disulfide exchange to allow for intramolecular cyclization to release the native OVA-1 peptide (Figure S11). However, no ions with masses corresponding to a peptide–glutathione conjugate that could support this hypothesis were observed by UPLC-MS. The cyclization rate of MHM is low, similar to that of HHM, because it has the same steric bulk at the electrophilic carbon (i.e., R¹ = Me). Comparing the HPH linkage to unsubstituted HHH, we found an increase in the reduction rate, which made the fit difficult to establish because only a few data points were observed prior to reaction completion, resulting in the large confidence interval (CI). We hypothesize that the bulky isopropyl substitution of HPH sterically hinders the unproductive nucleophilic attack by GSH and directs all S_N2 reactions toward productive disulfide cleavage, resulting in more rapid reduction. A much higher cyclization rate is also observed, which may be explained by steric compression (the Thorpe–Ingold effect) due to the isopropyl substituent that places the thiol in closer proximity to the electrophilic carbon.³⁰

In agreement with previous work by Sydnes and co-workers,³¹ simple alkyl substitutions on the traceless linker backbone (disulfide–ethylene–carbamate) alter the rates of disulfide cleavage and subsequent intramolecular cyclization. In view of the relatively small chemical differences between the SNAs synthesized using this set of traceless and nonreducible linkers, the differences in immunostimulation are attributable to differences in cleavage mechanism and rate. Therefore, the synthesized set of traceless linkers with various kinetic profiles presented herein is an excellent toolbox to study the effect of the rate of antigen release on immunostimulation.

Effect of Linkage Type on T Cell Activation

We used the set of conjugates with a range of traceless fragmentation kinetics to investigate whether antigen release rates from SNAs affect immunostimulation. The ability of the SNAs to elicit an immune response was quantified by measuring the percent of CD8⁺ T cells proliferating within OT-1 mouse splenocytes (whose T cells are specific for the OVA-1 epitope) after treatment with SNAs in vitro. T cell proliferation in response to stimulation by APCs is indicative of a robust immune response.³² Five SNA treatment groups were compared, each made with either a traceless linker conjugate (HPH SNA, HHH SNA, HHM SNA, or MHM SNA) or the thioether conjugate (BMPS SNA). The BMPS SNA also served as a control for the disulfide cleavage mechanism and a useful reference because of previous use of similar linkers for antigen incorporation.^4,5,23,33 A simple one-to-one mixture of native OVA-1 peptide and cholesterol-modified adjuvant DNA (admix) was employed as a positive control in vitro. Because of the uniquely strong binding of OVA-1 to MHC-I, the admix is expected to outperform other formulations in vitro since free OVA-1 can displace other peptides bound to cell surface MHC-I and bypass intracellular processing, an ineffective pathway for in vivo immunostimulation.³⁴ The various treatment groups (admix, the four traceless conjugate SNAs, and the nonreducible conjugate SNA) were incubated with surface-stained (amine-reactive eFluor 450) whole splenocytes from OT-1 mice for 72 h, and following incubation, the percent of CD8⁺ T cell proliferation was quantified by flow cytometry. A range of treatment concentrations, each performed in duplicate, was used to fit a dose–response curve for each treatment (Figures 2A, S16, and S17).

Figure 2.

In vitro and in vivo immune response to SNA treatment with OVA-1 antigen. (A) Representative experiment of a dose–response curve showing the percent of T cells proliferating in response to treatment with SNAs or admix formulated using SIINFEKL antigen at a range of doses (equivalent concentration of antigen to adjuvant). Duplicates were performed at each concentration. (B) Plot of native OVA-1 released over time, with points showing individual measurements and lines showing best fit curves, all normalized to 100% by initial value from fits. The inset shows a linear regression between −log(EC₅₀) and the percent of native OVA-1 released at 30 min (dashed line). The shaded area is the 95% CI of the fit. (C) Percent bone-marrow-derived CD11c⁺ DCs displaying OVA-1–MHC-I complex on their surface at various times following a 1 h pulse treatment. Data from two independent experiments are shown. (D) Area under the curve for the data in (C). ANOVA P = 0.0392; the Tukey corrected pairwise comparison is P = 0.0343 between HPH and BMPS, with all others ≥0.05. (E) Quantification of the ability of DCs from the draining lymph node to activate OT-1 T cells 16 h following a single immunization. SFC, spot-forming cells. One-way ANOVA analysis P < 0.0001; Tukey corrected pairwise comparisons between HPH and all other groups also showed P < 0.0001. (F) Ability of isolated T cells from mice immunized with SNAs to induce killing of ovalbumin-expressing target cancer cells, as measured through cells double-positive for the apoptotic and necrotic markers annexin V and 7-AAD, respectively. P values for two-way ANOVA post hoc pairwise t tests with Tukey correction for multiple comparisons are shown. All panels include error bars that indicate standard deviations.

A clear trend in the potencies of the treatments emerged (average EC₅₀ of at least three independent experiments): admix (0.54 pM), HPH SNA (10.5 pM), HHH SNA (53.8 pM), HHM SNA (185 pM), MHM SNA (684 pM), and BMPS SNA (4820 pM) (Figure 2A). No two curves share an EC₅₀ value (P < 0.0001), demonstrating that they all result in different T cell proliferation potencies. The low potency of the nonreducible SNA further supports the importance of traceless conjugation of antigens as reported by our group and others.^5,22−25 The potency of the traceless linkers is positively correlated with the rates of cleavage and cyclization (Figure 2B). We assume that small changes in chemical structure between the conjugates do not impact SNA uptake or adjuvancy of the CpG strand, an assumption supported by previous publications.^5,8 SNAs bearing the substituted HPH linkage are over 5 times more potent than the most commonly used traceless conjugate HHH and over 60 times more potent than the traceless MHM conjugate. These results show that the chemical identity of the traceless linkage used to form the antigen–DNA conjugate has a dramatic effect on the potency of the SNA, spanning 2 orders of magnitude. The kinetics of traceless linker fragmentation clearly impacts the SNA potency, so we next quantified how the amount of released peptide is correlated with the potency as a function of conjugate type.

Release of antigens from the conjugate, thus leading to the exposure of a free N-terminus, is likely a critical step for cross-presentation, since translocation from the endocytic compartment into the cytosol is dependent on size^34,35 and N-terminal modifications influence the binding of peptides to the MHC-I complex (and possibly other processes).³⁶ Previous studies have suggested that most efficient cross-presentation of MHC-I antigens via the cytosolic pathway occurs within the initial 30 min following uptake.¹⁷ Therefore, we considered whether the extent of native OVA-1 release from the various traceless linker conjugates at 30 min is correlated with immune responses (Figure 2B). The linkers produce significant differences in the proportion of native OVA-1 released at the 30 min time point (calculated on the basis of the apparent first-order rates found in Figure 1), and linear regression indeed showed a strong positive correlation (R² = 0.95, P = 0.027) between the extent of native antigen release (Figure 1A, blue box) and the potency of SNAs made with those conjugates (measured by −log(EC₅₀) from T cell proliferation measurements; Figure 2B inset). When the same analysis was performed on the extent of disulfide cleavage (Figure S18), no correlation was observed (R² = 0.57, P = 0.248), potentially because the extents of disulfide cleavage for HPH, HHH, and HHM are nearly identical at the relevant time scales. A comparison of tests for correlation between the amount of native OVA-1 released at different time points and the potency indicated that the highest correlation with the lowest P value was reached at the 40 min time point (Figure S19). The extent of native peptide released, not the extent of disulfide cleavage, may be the crucial factor that determines the immunostimulation potency.

The rate of native OVA-1 release may impact T cell priming by altering the timing and amount of peptide presented on the surface of the DCs. Therefore, we measured the amount of DCs that present OVA-1–MHC-I complexes on their surface over time after a 1 h pulse with SNAs or a simple mixture of antigen and adjuvant (Figures 2C, S14, and S15). A more rapid increase of surface OVA-1–MHC-I is observed for SNAs made using linkers with faster native OVA-1 release kinetics. The HPH SNA treatment results in the most rapid rise in OVA-1 presentation, followed by the other traceless linkers, with the BMPS SNA treatment resulting in the slowest presentation. The admix control shows an entirely different pattern, where antigen presentation remains high through the experiment because of the rapid loading of antigens directly on the surface. Furthermore, the AUC of the SNA treatments, which serves as a measure of the total amount of antigen–MHC-I complex displayed, mirrors the trend in potency (Figure 2D), with HPH exhibiting the highest total amount of antigen displayed and nonreducible BMPS exhibiting the lowest.

The linker chemistry and release rate may be especially critical when the antigens can be immediately loaded on MHC-I protein following their release without the need for further processing steps. To explore this question, we synthesized the same set of SNAs using a longer form of OVA-1 containing additional amino acids on the N-terminus (SGLEQLESIINFEKL), denoted as OVA-1L. Indeed, the differences between the SNAs made using the set of traceless linkers or the nonreducible linker vanished, and all of the treatments resulted in essentially indistinguishable potency (Figure S20A-B). However, the potency level of these SNAs was consistently lower than the those of the HHH and HPH SNAs shown in Figure 2A, suggesting that while this strategy removes the sensitivity to the linkage chemistry, it also decreases the immunostimulatory efficacy, possibly as a result of the additional required step of protease processing. The additional OVA-1L processing, which may involve endosomal cathepsin S, likely occurs prior to traceless linker fragmentation and prevents any impact of linker chemistry on antigen presentation.³⁶⁻³⁹ Thus, the SNAs made with longer OVA-1L peptides become sensitive to endosomal disruption (chloroquine treatment) and protease inhibitors (leupeptin treatment), while the OVA-1 SNAs remain largely unaffected by these agents (Figure S20C-D).

We finally sought to determine how differences in the conjugation chemistry and rate of antigen release affect immune stimulation in vivo, specifically the initial activation of DCs and their subsequent ability to activate antigen-specific T cells over time. We focused our attention on the two traceless linkers with greatest difference in rates of antigen presentation and immunostimulatory potency (i.e., HPH and MHM) and the nonreducible linker BMPS. To explore initial DC activation, we immunized mice (C57BL/6, three per group) with a single injection of SNAs (6 nmol of antigen and 6 nmol of adjuvant) and quantified the T cell priming ability of DCs isolated from the draining lymph nodes after 16 h (Figures 2E and S21). DCs from HPH SNA treatment elicited significant antigen-specific T cell activation as measured by interferon γ (IFN-γ) secretion via the enzyme-linked immune absorbent spot (ELISpot) assay. However, the nonreducible BMPS SNA and slow-cleaving MHM SNA did not show significant improvement of T cell priming compared with naive mice in this short time frame. To measure the capacity of the constructs to prime antigen-specific T cells in vivo over time, we immunized mice (C57BL/6, five or six per group) with one injection followed by two boosters administered biweekly (6 nmol of antigen and 6 nmol of adjuvant) (Figure 2F). We sacrificed the mice 1 week following the last injection, isolated their spleens, and quantified their activation and ability of the splenic T cells to kill target cancer cells that expressed ovalbumin protein (E.G7-OVA). Splenocytes from both the HPH and MHM SNA groups boosted IFN-γ secretion compared with the BMPS SNA group (Figure S22). The most efficacious traceless linkage in vitro, HPH SNA, resulted in the strongest target cancer cell killing ex vivo at all ratios of T cells to target cells tested. The MHM SNA group exhibited stronger target cell killing ability compared with the BMPS SNA and naive groups (Figure 2F). Finally, the lack of robust killing by the BMPS SNA conjugate compared with naive mice indicates that under these conditions a nonreducible N-terminal attachment may sacrifice antigenicity of OVA-1 peptide. These findings demonstrate that traceless antigen conjugation and the rate of antigen release are critical to the function of the SNA vaccine in vivo. SNAs that rapidly release native OVA-1 antigen increase DC activation in the draining lymph nodes and the raised cell killing capacity of T cells in the spleen.

Conclusions

In this study, we explored how the structure of the linker used to attach a peptide antigen to an immunostimulatory SNA affects the construct’s efficacy at eliciting anticancer immune responses. Importantly, this work unambiguously shows that the rate of antigen release impacts immune stimulation and subsequent vaccine efficacy. Specifically, strategic chemical modifications of traceless linkers sensitive to disulfide reduction modulate the apparent rates of disulfide cleavage and traceless fragmentation by over an order of magnitude. These rate differences translated to presentation of antigens on the surface of bone marrow dendritic cells ∼6 h earlier and led to an ∼30% increase in overall presentation, demonstrating that chemical design of the linkage can be used to deliberately modulate the interaction of DCs with SNA vaccines. The differences in the rates of cleavage controlled by the linker chemistry translated to a significant impact on the SNA efficacy. The potencies of SNAs were correlated with the rates of traceless fragmentation and the proportion of native OVA-1 released at early time points; these linker properties led to a 65-fold decrease of the SNA EC₅₀ of in vitro T cell proliferation. Traceless attachment of antigens, compared with a nonreducible linkage, resulted in over 100-fold lower EC₅₀ of T cell proliferation in vitro, an order of magnitude greater antigen-specific INF-γ secretion in vivo, and 10 times more potent cancer cell killing ability of splenocytes. The differences in in vitro T cell proliferation potency disappeared when antigens extended at the N-terminus were used, indicating that additional antigen processing reduces the impact of the cleavage kinetics. The chemical structure of the traceless linker strongly impacted the in vivo efficacy: an order of magnitude increase in DC activation within draining lymph nodes and an ∼50% increase in cancer cell killing by T cells isolated from spleens were observed as a result of different kinetics of antigen release. Taken together, this work demonstrates that the conjugation chemistry used to assemble immunostimulatory SNAs can be used to tailor antigen processing and downstream immune responses and therefore guide the design of immunotherapeutics more broadly.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscentsci.1c00779.

• Experimental methods, compound characterization including NMR and MS, and additional experimental data presented as Figures S1–22 and Tables S1–S3 (PDF)

The authors declare no competing financial interest.