Design and Assessment of First-Generation Heterobifunctional PPARα/STING Modulators

Abstract

Inflammatory retinal diseases such as diabetic retinopathy (DR) and age-related macular degeneration (AMD) are prominent causes of blindness in industrialized countries. The complexity of these diseases, involving diverse cell types and pathways that give rise to a multifactorial pathogenesis, complicates drug discovery. As such, therapies exhibiting polypharmacology are expected to improve outcomes through broader disease stage coverage and beneficial spatiotemporal effects. We report herein the first dual modulator of PPARα and STING, two targets tied to disparate pathologies in retinal diseases. Recognizing structural similarities between a reported STING inhibitor SN-013 and our previously described PPARα agonist A229, we designed BH400, which agonizes PPARα (EC₅₀ = 1.2 μM) and inhibits STING (IC₅₀ = 8.1 μM). BH400 demonstrates superior protection over single-target PPARα or STING modulation in microglial and photoreceptor cells. These findings provide compelling evidence for the potential benefit of polypharmacology in common retinal diseases through dual PPARα/STING modulation, motivating further studies.

Diabetic retinopathy (DR) and age-related macular degeneration (AMD) are leading causes of blindness in developed countries. These inflammatory retinal diseases share underlying pathologies in which prolonged inflammation contributes to leukostasis and progressive degeneration of the retinal vasculature, leading to vascular leakage, retinal hypoxia, and angiogenesis.¹ The current standard of care for DR and AMD involves repeated intraocular injections of anti-VEGF (vascular endothelial growth factor) biologics.² These interventions seek to disrupt angiogenesis and slow disease progression, but the required high frequency of administration imposes a significant burden on healthcare resources and presents long-term side-effects and compliance issues.² Additionally, 40% of patients fail to respond to anti-VEGF therapies, likely due to the multifactorial nature of DR and AMD pathogenesis.² Due to these limitations, new therapeutic modalities are highly desirable, especially if they hold potential for oral bioavailability and exhibit novel non-VEGF targeted mechanisms of action.

The pathogenesis of these common retinal diseases involves degeneration of many cell types, including pericytes, endothelial cells, Müller cells, retinal pigment epithelium, and photoreceptors.³ Gene expression varies greatly between these cell types, providing an opportunity for polypharmacological approaches to reduce disease progression by extending the magnitude and scope of cytoprotection to various cell types.⁴ In fact, recent proteomic data reveals that the cellular drivers of DR change with age and disease progression, with angiogenic factors in nonproliferative disease (NPDR) expressed primarily by endothelial cells, and after progression to proliferative disease (PDR), angiogenesis is driven primarily by microglia and macrophages.⁴ As such, disease heterogeneity is hypothesized to contribute to the suboptimal patient response to existing single target treatments. For example, VEGF was determined to be expressed in PDR but not NPDR, revealing a potential explanation for the lack of improved visual acuity in NPDR patients following anti-VEGF treatment after 4 years in a recent clinical trial.^4,5 Molecules exhibiting polypharmacology are expected to not only improve the breadth of disease coverage (stages and pathological drivers) but also extend such protection spatiotemporally (various cell types).

PPARα is a known therapeutic target for DR, based on two independent, prospective, placebo-controlled clinical trials.^6,7 Recent studies have demonstrated that PPARα is a major regulator of mitochondrial function and integrity.⁸ Our previous studies showed that PPARα is downregulated in diabetic retina, which plays a pathogenic role in metabolic disturbance in retinal neuron and retinal degeneration.^9,10 Recently, compelling evidence has been reported that aberrant cGAS-STING activation leads to inflammatory and vascular pathologies that also significantly contribute to DR pathology.¹¹ An initial investigation of the interaction between the cGAS-STING and PPARα pathways in a context of retinal disease revealed that PPARα agonism protects mitochondrial function under hypoxic conditions, preventing the cytosolic release of mitochondrial DNA (mtDNA) and indirectly attenuating activation of cGAS-STING signaling.¹² Similarly, PPARα knockout increased STING expression and sensitized mice to retinal injury in oxygen-induced retinopathy.¹²

Data from the Human Protein Atlas reveal expression differences of PPARα and STING in nondiseased ocular cell types involved in the pathogenesis of inflammatory retinal diseases. Rod photoreceptors feature high expression of PPARα and low expression of STING, while the opposite is true for endothelial cells; microglia feature some expression of both targets (Figure 1).^13,14 All told, this suggested to us that, for a given cell type, disease pathology may be driven more significantly by one target over the other and that proper modulation of each target with a single chemotype may prove advantageous spatiotemporally. As such, we hypothesized that dual modulation will restore the equilibrium of both PPARα and STING, demonstrating simultaneous benefits in several major DR pathologies, including neurodegeneration, inflammation, vascular leakage, neovascularization, and fibrosis, a multifactorial response not achieved with single-acting agents (e.g., anti-VEGF biologics).

Figure 1.

Single-cell RNA sequencing showing PPARα, STING, and NF-κB expression in rod photoreceptors, endothelial cells, and microglia derived from nondiseased ocular cell types. Expression is quantified as normalized transcripts per million (nTPM), which is representative of the number of transcripts detected for a given gene. Rod photoreceptors exhibit high gene expression of PPARA (144.6 nTPM); endothelial cells show much lower expression (14.3 nTPM). Comparatively, STING1 expression is high in endothelial cells (99.6 nTPM) but nearly absent in rod photoreceptors (0.3 nTPM). NFKB1 is shown as a pro-inflammatory reference protein.

Upon the publication of STING inhibitors SN-011 and SN-013, we noted high structural similarity to our previously published PPARα agonist A229 (Figure 2).^15,16 Initially, due to this noted similarity and the common employment of sulfonamides as carboxylic acid isosteres,¹⁷ we speculated that the SN analogues and A229 might exhibit dual modulation of STING and PPARα. Bifunctional activity, however, was not observed at concentrations <100 μM for either chemotype in cross-comparative cell-based luciferase assays (Figure 3). Lack of PPARα agonism by the SN chemotype was suspected to arise from the appended sulfonamides being significantly larger than the carboxylic acid binding pocket has been shown to accommodate.¹⁸ This is supported by the fact that the smaller methyl sulfone exhibited by SN-013 does elicit low levels of PPARα agonism at 100 μM. The similar molecular connectivity of SN analogues and A229 implied to us that a compound exhibiting both PPARα agonism and STING inhibition at similar concentrations could be engineered. Herein, we report the design, synthesis, initial structure activity relationships, and biological confirmation of the first known bifunctional PPARα/STING modulator, BH400.

Figure 2.

(A) Structural similarity between STING inhibitory SN compounds and the PPARα agonist, A229. Noted common isosteres encircled in blue. (B) Ligand preparation, alignment, and superposition of SN-013 (lavender) and A229 (green) reveal an RMSD of zero between the maximum common substructures of each compound.

Figure 3.

Dose-response curves for SN-013 (A) and SN-011 (B) in the PPARα-dependent Gal4 hybrid luciferase reporter assay. GW7647 is a positive control full-agonist (EC₅₀ ∼ 5 nM) for PPARα. (C) Dose-response for A229 in the STING-dependent THP-1 luciferase assay. The SN-011 dose-response curve is layered for comparison as a positive control.

Two strategies were considered in the design of a heterobifunctional molecule: (1) start from the SN chemotype and design in PPARα activity or (2) start with A229 and design in STING inhibitory activity. While the first strategy would benefit from our previous successful design of PPARα agonists, we elected to pursue the second strategy, as it allowed us to leverage a more cost-effective STING inhibitory screening assay as a preliminary readout. To start, we designed a focused library of three A229 derivatives to determine whether the installation of SN substituents could impart STING inhibitory activity and retain PPARα agonism. This strategy was realized by retaining the core carbon skeleton of A229 and assessing the effects of (1) A-ring 4-hydroxy inclusion (BH400) and (2) carbonyl incorporation between the A- and B-rings (BH425 and BH436) in the presence or absence of the A-ring 4-hydroxy group. The identification of an A229 derivative with STING inhibitory effects would enable more detailed structure–activity relationships, presumably revealing opportunities to further optimize the chemotype (Figure 4).

Figure 4.

General strategy for SAR interrogation.

A majority of the A229 analogues were prepared via reductive amination of functionalized biphenyl aldehydes with aminosalicylic acids to generate target compounds (Scheme 1). Requisite biphenyls unavailable commercially were synthesized through the coupling of the corresponding 4-bromobenzaldehydes with arylboronic acids via a Suzuki–Miyaura reaction. A229 amide derivatives were prepared through the coupling of requisite biphenyl acid chlorides with methyl aminosalicylates. The resulting methyl esters were subsequently saponified to generate the final compounds (Scheme 1).

Scheme 1. General Approach for the Synthesis of A229 Derivatives.

Some compounds are synthesized by other means; see Scheme S1. (i) Toluene, 120 °C, 2 h; THF, Na(OAc)₃BH, AcOH, RT, 16 h, 5–78%. (ii) SOCl₂, MeOH, reflux, 3 h, 80-94%. (iii) LiOH·H₂O, THF/MeOH/H₂O, 120 °C, 12 h, 22–99%.

Initially, three compounds were synthesized to impart SN analogue features onto A229, namely, the presence of the A-ring hydroxyl (BH400), the linker carbonyl (BH425), and both motifs simultaneously (Figure 5, BH436). These initial analogues were evaluated via a commercial luciferase-based STING inhibition assay leveraging THP1-Lucia ISG cells derived from the human monocytic cell line THP-1 (Invivogen, thpI-isg). Of these modifications, only addition of the A-ring hydroxyl group was required for eliciting STING inhibitory activity, demonstrated by the reduction of luciferase activity following cGAMP activation (Figure 6A). Initial computational studies of the SN chemotype reported this hydroxyl group to be important for STING inhibition due to a predicted hydrogen bond with S243. As such, our findings for its essentiality in imparting STING inhibitory activity in the A229-derived series support this hypothesis. With this key substituent effect, several further modifications were made (Figure 5): (1) substitution or repositioning of the A-ring hydroxyl group found in SN-013 (BH582, BH585), (2) substitution and repositioning of the carboxylic acid of A229 (BH555, BH621, BH623, BH680), (3) isosteric replacement of the linker between the A-ring and biphenyl motif (BH436, BH483, BH541, BH569), (4) exploration of the space around the biphenyl via a methyl walk and other substitutions (BH553, BH606, BH652, BH654, BH656, BH659), and (5) substitutions to the para position of the C-ring biphenyl (BH530, BH558, BH614, BH637, BH647, BH651). Compounds were synthesized using the chemistry presented in Scheme 1 or modifications thereof (Scheme S1). These compounds were evaluated via the same commercial luciferase-based STING inhibition assay. As shown in Figure 6A, none of the additional analogues demonstrated clear superiority to BH400 for STING inhibition and thus were not assessed for PPARα agonism. While unsuccessful in identifying a compound more potent than BH400, the SAR successfully recapitulates the trends seen from the computational docking of SN-011. The aromatic stacking interaction of the BH400 biphenyl ring between Y163 and Y167 of STING precludes sizable substitutions to the biphenyl ring, while the network of hydrogen bonds between S243, Y245, and L212 and the A-ring hydroxyl and carboxylate (sulfonamide in SN compounds) groups are disrupted by repositioning these key substituents.

Figure 5.

Compounds synthesized for SAR interrogation. Initial analogues are highlighted in gray.

Figure 6.

(A) Level of cGAS-STING activation following cGAMP (1 μM) administration to THP-1 Lucia cells in the presence of analogues at three doses (50, 25, and 12.5 μM). Low to high concentrations proceed right to left for each compound. Each concentration was conducted in triplicate, normalized to agonist response with vehicle (DMSO), and is reported as % activation relative to cGAMP-only treatment in vehicle ± SEM. Decreases in luciferase expression associated with cytotoxicity are noted with #. (B) BH400 PPARα-dependent Gal4 hybrid luciferase EC₅₀ dose-response curve. Fold of the control is RLU normalized to full agonism by GW7647 (n = 3). (C) BH400 STING-dependent THP-1 luciferase IC₅₀ dose-response curve. Fold of control is RLU normalized to the signal generated from cGAMP (1 μM) (n = 3). (D) Cell viability in THP-1 Lucia cells following BH400 administration measured via an MTS assay; no cytotoxicity is observed up to 50 μM (n = 3).

Of note, a pronounced agonist effect is seen with select compounds, most prominently with BH606, BH617, and BH621. These compounds show no agonism when administered independently of a STING agonist, and the effect persists across various agonists (e.g., poly(dA:dT), SR-717, and cGAMP). We speculate that, due to the dimeric nature of STING, cooperativity between these analogues and traditional STING agonists may produce a higher agonist response. Retention of PPARα agonism was confirmed for BH400 with a commercially available cell-based luciferase assay (Indigo, Figure 6B), and an EC₅₀ of 1.2 μM was determined. A dose-response for STING inhibition was also observed in the comparable STING luciferase assay to provide an IC₅₀ value of 8.1 μM (Figure 6C). Cytotoxicity of BH400 was assessed via a formazan-dependent cytotoxicity assay (MTS) against THP-1 cells, and no evidence of cytotoxicity was observed up to 50 μM (Figure 6D). As such, BH400 was selected for additional cell-model studies indicative of PPARα agonism and STING inhibition.

With the luciferase data for BH400 demonstrating dual target modulation, we interrogated the potential benefits of polypharmacology in more disease relevant cell-based models. A known indicator of PPARα agonism is a reduction in oxidative stress following CoCl₂ treatment. This is a commonly used cell model of hypoxia in which CoCl₂ stabilizes hypoxia-inducible factors, triggering a hypoxic response, even under normoxic conditions. In this model, the response to hypoxia is characterized by oxidative stress, which is ameliorated by mitochondrial protection provided through PPARα agonism, and this hypoxic signaling is thought to mirror the pathogenesis of DR and AMD. Pretreatment of HMC3 cells (microglial cell line) with BH400 shows dose-dependent protection against CoCl₂-induced reactive oxygen species (ROS) (Figure 7A). To further characterize the effect of BH400 on the PPARα pathway, PGC1α was quantified by Western blot following LPS treatment. PGC1α is a coactivator of PPARα with a key role in regulation of cellular metabolism and oxidative stress.¹⁹ It is well-established that PPARα agonism elicits an upregulation of PGC1α.¹⁹ As shown in Figure 7B, BH400 rescues the expression of PCG1α following the LPS treatment.

Figure 7.

(A) BH400 reduces the CoCl₂-induced production of ROS. HMC3 cells were pretreated with increasing doses of BH400 and incubated with CoCl₂ (200 μM), and then ROS production was measured by DCF fluorescence and normalized to total protein concentration (n = 6). CaCl₂ (200 μM) was used as a vehicle control that maintained the same osmolarity as the treatment. All data points are compared to DMSO by one-way ANOVA. (B) BH400 (10 μM) rescues LPS-induced reduction in PGC1α, a target gene of PPARα. HMC3 cells were pretreated with BH400; then levels of PGC1α were quantified by Western blot and normalized to β-actin (n = 3). (C) BH400 reduces LPS-induced secretion of interleukin 6 (IL-6). HMC3 cells were pretreated with increasing concentrations of BH400 and then stimulated with LPS. The medium was collected for ELISA and normalized by total protein concentration (n = 3). All data points are compared to DMSO by one-way ANOVA. (D) BH400 rescues LPS-induced reduction in pTBK1. HMC3 cells were pretreated with BH400; then levels of pTBK1 were quantified by Western blot and normalized to β-actin (n = 3). Values are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, one-way ANOVA, pairwise comparisons by t test. Control, contains no stressor (only cells); DMSO, contains no compound (only vehicle); 0 μM, untreated (no vehicle or compound).

Characterization of the STING inhibitory activity of BH400 is complicated by crosstalk between the PPARα and STING signaling pathways.¹⁴ To reduce the potential for crosstalk, IL-6 secretion is measured by ELISA as IL-6 is directly controlled by STING signaling. Although STING directly regulates IL-6 expression, some crosstalk between PPARα and NF-κB may also affect IL-6 signaling, although this relationship has yet to be fully elucidated.^14,20 As shown in Figure 7C, BH400 protects against lipopolysaccharide (LPS)-induced IL-6 secretion in a dose-dependent manner in HMC3 cells. To further characterize targets directly downstream of STING signaling, pTBK1 was quantified by Western blot, as TBK1 phosphorylation is a well-known downstream effect of STING activity. As shown in Figure 7D, BH400 significantly reduced pTBK1 levels following LPS treatment.

The cytoprotective effects of BH400 in comparison to single-target modulators were then assessed. As oxidative stress is a key mediator of DR and AMD pathology and is modulated by both PPARα and STING, quantification of intracellular ROS by dichlorofluorescein (DCF) production is a robust metric that connects both pathways. For these comparative studies, we selected the commercially available SN derivative, SN-011, and A229 (Figure 2A) as the single-target comparators. In HMC3 microglia, STING inhibition (SN-011) shows a greater reduction in oxidative stress than PPARα agonism (A229; Figure 8A). We speculate that this is due to the higher expression levels of STING versus PPARα in immune cells (Figure 1), leading to more STING-dependent oxidative stress and thus more pronounced protection produced by STING inhibition than selective PPARα agonism in this cell line. Treatment with BH400 produced a comparable reduction in ROS compared to selective STING (SN-011) inhibition alone. It is worth reiterating that SN-011 exhibits a STING IC₅₀ of 76 nM, nearly 100-fold more potent than that of BH400. As such, the ability for BH400 to elicit the same level of response at only 1.5 times the STING IC₅₀ is indicative of contributions from both STING inhibition and PPARα agonism. In 661W photoreceptor cells (Figure 8B and C), the difference between SN-011 and A229 is less pronounced, likely due to higher PPARα expression in these cells and the ability of PPARα agonists to indirectly attenuate STING signaling.¹² Dual modulation with BH400 in this cell type elicits a statistically significant improvement over single-target modulation in the H₂O₂-induced model (Figure 8C) and comparable effects to single modulation in the 4-HNE-induced model (Figure 8B) at the concentration tested (12.5 μM). Once again, the ability of the less potent BH400 to attain the same level or more pronounced effect in these studies than significantly more potent single target comparators is indicative of multimodal impacts.

Figure 8.

(A) Reduction in the TBHP-induced production of ROS. HMC3 cells are treated with 12.5 μM test compounds and 200 μM TBHP. ROS production was measured by DCF fluorescence. (B) Reduction in 4-HNE-induced production of ROS in 661W cells. 661W cells were pretreated with SN-011, A229, or BH400 (12.5 μM) for 2 h and 4-HNE (10 μM) for another 24 h. ROS production was measured using H2DCFDA and normalized by total protein concentration (n = 6). (C) Reduction in H₂O₂-induced ROS production in 661W cells. 661W cells were pretreated with SN-011, A229, and BH400 (12.5 μM) for 2 h and then exposed to H₂O₂ (500 μM) for another 24 h. ROS production was measured using H2DCFDA and normalized by total protein concentration (n = 6). (D) Improvement in cell viability in 661W cells measured by trypan blue exclusion after administration of 4-HNE. 661W cells were treated with compounds SN-011 and BH400 at concentrations of 1 and 10 μM for 2 h and then exposed to 4-HNE (10 μM) for another 24 h. Cell viability was counted with trypan blue exclusion (n = 3). (E) Improvement in cell viability measured by trypan blue exclusion after exposure to H₂O₂. 661W cells were treated with compounds SN-011 and BH400 at concentrations of 1 and 10 μM for 2 h and then stressed with H₂O₂ (500 μM) for another 24 h. Cell viability was counted with trypan blue exclusion (n = 3). (F) Reduction in 4-HNE-induced production of ROS in 661W cells. 661W cells were pretreated with SN-011, A229, and/or BH400 at various concentrations for 2 h and 4-HNE (10 μM) for another 24 h. ROS production was measured using H2DCFDA and normalized by total protein concentration (n = 6). Values are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. (A–C) Statistical significance is represented above each column in comparison to DMSO + positive control by one-way ANOVA, pairwise comparisons by t-test. (D, E) Data points above each column in comparison to positive control by one-way ANOVA, pairwise comparisons by t test.

We then compared the activity of BH400 against the coadministration of both single-target agents. As shown in Figure 8F, BH400 (12.5 μM) outperformed the coadministration of A229 and SN-011 (12.5 μM each), while BH400 (1 μM) demonstrated a statistically comparable effect compared to coadministration of A229 and SN-011 (1 μM each) and outperformed a co-dose of 0.5 μM for each single target ligand. While many variables (e.g., cell penetration rates/efficiency) complicate the ability to compare single molecule polypharmacological effects against coadministration, these results provide further evidence of dual target modulation.

In addition to ROS production, we also determined the overall impact on cell viability. Photoreceptor degeneration is a direct cause of vision loss in DR and AMD, and improving photoreceptor viability under conditions of oxidative stress is thought to delay disease progression.²¹ At a 1 μM concentration, SN-011 and BH400 exhibited similar cytoprotection of 661W cells after exposure to 4-HNE or H₂O₂. While chemical oxidants like TBHP and H₂O₂ are directly causative of oxidative stress, 4-HNE is a product of lipid peroxidation that mediates pathological responses to oxidative stress and a commonly used diabetic stressor.²² At 10 μM, against either 4-HNE or H₂O₂, BH400 significantly improved viability compared to SN-011, once again demonstrating a benefit for dual modulation (Figure 8D and E). On the contrary, SN-011 fails to show improvement upon dose escalation. It can therefore be postulated that this increased viability produced by BH400 is due to additive effects captured through the dual modulation of PPARα and STING.

In summary, BH400 exhibits (1) PPARα agonistic and STING inhibitory dual modulation activity, (2) protective effects against oxidative stress that in some contexts exceed the effects of selective PPARα agonism or STING inhibition, (3) comparable to improved protective effects over coadministration of single target agents, and (4) superior protection of cell viability over selective STING inhibition. Spatiotemporally, BH400 exhibits superior protection against ROS production in microglial cells (HMC3) compared with selective PPARα agonism and an improvement in photoreceptor (661W) cell viability beyond that observed for selective STING inhibition. All told, this provides compelling evidence that the bifunctional activity of BH400 exhibits polypharmacological and spatiotemporal benefits not attained by single target modulation.

Heterogeneity of disease pathogenesis in DR and AMD diminishes the disease modifying impact of single-target therapeutic agents. The large number of affected cell types and change in pathogenic drivers with disease progression lead to a significant population unresponsive to current therapeutic modalities. With dual modulation, cytoprotection is expanded beyond cell types that predominantly express one target and the cytoprotective effect is increased in cell types that express both targets. In microglia and photoreceptor cells, which are both implicated in disease progression, the STING/PPARα modulator BH400 shows benefits beyond selective STING inhibition or PPARα agonism in the assessed cell-based disease models. Given the expansion of the protective effect across additional cell types, a combination of PPARα and STING modulation in a single chemotype is expected to not only improve the magnitude of the protective effect but also extend such protection spatiotemporally. While the SAR for this scaffold suggests a limited possibility of improving potency for both targets further, it provides evidence that such ligands can be designed, opening the door to explore other chemotypes. These promising foundational results support continued development of multitarget modulators in inflammatory retinal diseases and provide compelling evidence to assess dual PPARα/STING modulator efficacy in in vivo models.

Glossary

Abbreviations

4-HNE

4-hydroxynonenal

AMD

age-related macular degeneration

cGAMP

cyclic guanosine monophosphate-adenosine monophosphate

cGAS-STING

cyclic GMP-AMP synthase (cGAS)-stimulator of interferon genes (STING)

DCF

dichlorofluorescein

DR

diabetic retinopathy

EC₅₀

effective concentration for half-maximal activity

ELISA

enzyme-linked immunosorbent assay

IC₅₀

inhibitory concentration for half-maximal inhibition

mtDNA

mitochondrial DNA

PPAR

peroxisome proliferator-activated receptor

ROS

reactive oxygen species

SEM

standard error of the mean

VEGF

vascular endothelial growth factor

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.4c00153.

• Biological assay protocols, supplemental synthetic scheme, synthetic procedures, NMR spectra for all final compounds, and HPLC trace for BH400 (PDF)

Author Contributions

B.H. and A.S.D. designed all compounds, and B.H. synthesized all molecules. J.J.L. and B.H. conducted the cell-based luciferase assessments. Y.C. performed all biology outside of the luciferase evaluation. The manuscript was written by B.H. and A.S.D. and edited and approved by all authors.

The authors declare the following competing financial interest(s): A.S.D. and J.X.M. are co-founders of Excitant Therapeutics.