Isolation and Characterization of the Cyanobacterial Macrolide Glycoside Moorenaside, an Anti-Inflammatory Analogue of Aurisides Targeting the Keap1/Nrf2 Pathway

Abstract

A new 14-membered ring brominated macrolide glycoside, named moorenaside (1), was discovered from a marine cyanobacterial sample collected from Shands Key in Florida. The structure of 1 was established by analysis of spectroscopic data including its relative configuration. The absolute configuration was inferred from optical rotation data and comparison with related compounds. The structure of 1 features an α,β-unsaturated carbonyl system, which is also found in aurisides. The presence of this motif in 1 prompted us to evaluate its effect on Keap1/Nrf2 signaling, a cytoprotective pathway culminating in the activation of antioxidant genes activated upstream by the cysteine alkylation of Keap1. Moorenaside exhibited moderate ARE luciferase activity at 32 μM. Due to the established crosstalk between Nrf2 and NF-κB pathways, we investigated the anti-inflammatory effects of 1 in LPS-induced mouse macrophages (RAW264.7 cells), a commonly used model for inflammation. Moorenaside significantly upregulated Nqo1 (Nrf2 target gene) and downregulated iNos (NF-κB target gene) at 32 μM by 5.0- and 2.5-fold, respectively, resulting in a significant reduction of nitric oxide (NO) levels. Furthermore, we performed RNA-sequencing and demonstrated the transcriptional activity of 1 on a global level and identified canonical pathways and upstream regulators involved in inflammation, immune response, and certain oxidative-stress-underlying diseases such as multiple sclerosis and chronic kidney disease.

Marine cyanobacteria are a rich and unique source of secondary metabolites that are structurally diverse and exhibit a wide spectrum of activities.^1,2 The structural diversity of cyanobacterial compounds is attributed to their large multimodular nonribosomal peptide synthetases (NRPS) and polyketide synthases (PKS) biosynthetic machinery resulting in the production of a variety of structural classes, including lipopeptides, polyketides, fatty acid amides, and macrolides.¹⁻³

Cyanobacterial macrolides are typically 14- to 18-membered, and can be larger, polyene macrocyclic lactones composed of a heavily oxygenated backbone and oftentimes glycosylated.^4,5 Based on a review of 505 new marine-derived macrolides reported from 1991–2020, marine cyanobacteria produced 4.8% of the 505 macrolides, with sponges being the predominant source with 34.3%.⁴ The therapeutic potential of marine-derived macrolides is well documented with compounds exhibiting a variety of disease-relevant effects, including cytotoxic, immunomodulatory, antimicrobial, and anti-inflammatory activities.⁶ The majority of marine microbe-derived glycosylated natural products are macrocyclic lactones (23%), with marine bacteria being the main source based on a survey of 205 new glycosides from 1997–2018.⁷

Unusual bromine-substituted 14- to 16-membered macrolide glycosides were previously discovered from marine cyanobacteria (Figure 1).^5,8 Members of this class of compounds include lyngbyaloside and 2-epi-lyngbyaloside from Lyngbya bouillonii (now known as Moorena bouillonii), lyngbyaloside B from Lyngbya sp. and its related analogues 18E and 18Z-lyngbyalosides C (Figure 1).⁹⁻¹¹ Marine cyanobacteria also produced compounds with similar scaffold but lacking the bromine in the side chain, including lyngbouilloside from Lyngbya bouillonii(12) in addition to the most recently discovered akunolides A–D¹³ and irijimasides A–E¹⁴ from the Okeania sp. (Figure 1). Polycavernosides D and E are other nonbrominated macrolide glycosides from the marine cyanobacterium Okeania sp., but the source of their previously reported analogues A–C was ascribed to the red alga Polycavernosa tsudai.^15,16 Other structurally related analogues from marine invertebrates and speculated to be from a cyanobacterial origin include aurisides, dolastatin 19, and callipeltosides.¹⁷⁻¹⁹ Aurisides A and B¹⁷ and dolastatin 19¹⁸ were isolated from the sea hare Dolabella auricularia while callipeltosides A–C¹⁹ are nonbrominated analogues characterized by their diene-yne-trans-2-chlorocyclopropane moiety at the pendent side chain from the marine sponge Callipelta sp. (Figure 1). Phorbasides isolated from the sponges of the genus Phorbas are structurally related to callipeltosides A–C, having a similar macrolide scaffold, containing the scarce motif, ene-yne-trans-2-chlorocyclopropane at the pendent side chain but with different sugar units.^20,21

Figure 1.

Chemical structures of selected previously reported macrolide glycosides from marine cyanobacteria (Lyngbya and Okeania sp.), sea hare (Dolabella auricularia), and marine sponge (Callipelta sp.). Lyngbyaloside, 2-epi-lyngbyaloside, lyngbouilloside, irijimasides, and akunolides are the proposed (originally assigned) structures. The corrected (revised) structures of lyngbyalosides B and C and dolastatin 19 following their total synthesis are shown. The structures of aurisides A and B, and callipeltoside A were confirmed by total synthesis. *Total synthesis revealed incorrect assignment of the originally proposed structure of lyngbouilloside, but no revised structure was proposed.

Since their discovery, marine-derived macrolide glycosides have been of interest to the synthetic community due to their unique structural features, including the large number of stereocenters. The total synthesis of the majority of the aforementioned compounds was achieved, which resulted in unambiguous confirmation of some structures and stereochemical refinements of others.^19,21−31 Despite the availability of more material, achieved by total synthesis, the biological activity of this class of compounds was limited to assessment of their cytotoxicity against a selected panel of cancer cell lines.²⁶ In general, these macrolide glycosides exhibited moderate cytotoxic activity.¹⁰ Lyngbyaloside B was reported to inhibit the proliferation of human epidermoid carcinoma KB and colon adenocarcinoma LoVo cells (IC₅₀ of 4.3 and 15 μM, respectively), while aurisides A and B displayed cytotoxicity against HeLa S3 cervical cancer cells (IC₅₀ of 0.17 and 1.2 μg/mL, respectively).^9,17 The most recent members of nonbrominated cyanobacterial macrolide glycosides, akunolides A–D, displayed moderate antiparasitic activities against Trypanosoma brucei rhodesiense (IC₅₀ of 11–14 μM),¹³ while irijimasides A–E inhibited osteoclast differentiation, thus highlighting the therapeutic potential of this class of natural products.¹⁴

Our exploration of a benthic field-collected marine cyanobacterium from the Florida Keys afforded a new member of this growing family of brominated macrolide glycosides bearing an α,β-unsaturated carbonyl group (1; Figures 1 and 2). Herein, we report its isolation, structure elucidation, and, for the first time, its Nrf2 activity. We further demonstrate its anti-inflammatory and global transcriptional activities in lipopolysaccharide (LPS)-stimulated mouse macrophages.

Figure 2.

Chemical structure of moorenaside (1) with labeled substructures (I–VI). (A) Key COSY and HMBC correlations. (B) Key NOE correlations and coupling constants of the rhamnopyranoside protons to establish the relative configuration of 1.

Results and Discussion

Samples of the benthic marine cyanobacterium VPFK22-6, identified as Moorena sp. based on morphological identification and 16S rRNA gene sequencing, collected off Shands Key in Florida, were freeze-dried and subjected to nonpolar extraction (1:1 EtOAc–MeOH) followed by subsequent solvent partitioning between EtOAc and H₂O. The EtOAc-soluble fraction was subjected to silica gel chromatography using a gradient system of increasing polarity, starting with 30% EtOAc–hexanes and ending with 50% EtOAc–MeOH. The EtOAc silica fraction was further purified by reversed-phase HPLC to afford the macrolide glycoside 1, named moorenaside (Figure 1).

The structure was elucidated using 1D and 2D NMR, acquired in CDCl₃ (Table 1; deposited in NP-MRD), and HRMS data. The HRESI spectrum of 1 in the positive mode showed a [M + Na]⁺ peak at m/z 653.1926 and an isotope peak of equal intensity at m/z 655.1909, suggesting the presence of one bromine atom. The deduced molecular formula of the compound is C₂₉H₄₃BrO₁₀ with 8 degrees of unsaturation.

Table 1. NMR Spectroscopic Data for Moorenaside (1) in CDCl₃ (600 MHz).

C/H no	δH (J in Hz)	δCa, typeb	COSY	HMBCc	NOE
1		176.3, C
2	2.62, q (7.2)	47.7, CH	H3-18	1, 3, 18	H-4a
3		98.3, C
3-OH	4.53, br.s		H-4a	4	H-7, H3-18
4a	1.11, td (11.7, 2.6)	39.2, CH2	H-5, H-4b, 3-OH	3, 5, 6
4b	2.21, dd (11.7, 4.8)		H-4a, H-5	3, 5, 6	H3-18, H-4a, H-5
5	4.11, m	70.4, CH	H2-4, H2-6		H-1′, H-4b, H-6b, H-7
6a	1.21, q (11.7)	30.8, CH2	H-5, H-6b, H-7	5, 7
6b	1.92, m		H-5, H-6a, H-7	5	H-5, H-6a, H3-19, H-1′
7	3.72, dd (11.7, 2.1)	74.0, CH	H2-6	20	3-OH, H-5, H-6b, H3-19
8		49.3, C
9		204.7, C
10	6.27, br.s (1.0)	126.1, CH		9, 11, 12	H-12b, H3-20
11		148.5, C
12a	2.27, m	47.6, CH2	H-13	10, 11
12b	2.29, dd (12.1, 10.0)		H-13	10, 11	H-10
13	5.68, ddd (10.0, 6.3, 2.3)	71.1, CH	H2-12	14	H-15, H-12a, H3-21
14	5.74, dd (15.0, 6.3)	131.2, CH	H-15	13, 16
15	6.20, dd (15.0, 10.8)	129.3, CH	H-14, H-16	13, 17
16	6.70, dd (13.5, 10.8)	136.1, CH	H-15, H-17	17
17	6.42, d (13.5)	110.5, CH	H-16	15, 16
18	1.15, d (7.2)	12.4, CH3	H-2	1, 2, 3	H-2, 3-OH
19	1.02, s	21.4, CH3		7, 8, 9, 20	H-6b, H-7, H3-20
20	1.27, s	18.5, CH3		7, 8, 9, 19	H-10
21	2.12, d (1.0)	19.33, CH3		9, 10, 11, 12	H-12a, H-13
1′	5.02, d (1.5)	94.1, CH	H-2′	2′, 3′, 5′	H-5, H-6b, H3-2′, H-2′
2′	3.41, dd (3.7, 1.5)	81.0, CH	H-1′, H-3′	4′	H-1′, H3-2′
2′-OMe	3.50, s	58.9, CH3		2′	H-1′, H-2′
3′	3.80, dd (9.5, 3.7)	71.3, CH	H-2′, H-4′		H-2′, H-4′
3′-OH	not observed
4′	2.97, t (9.5)	84.0, CH	H-3′, H-5′	4′-OMe, 5′, 6′	H-3′, H3-4′, H3-6′
4′-OMe	3.58, s	61.0, CH3		4′
5′	3.58, m	67.5 CH	H-4′, H3-6′	4′
6′	1.28, d (6.2)	17.9, CH3	H-5′	4′, 5′	H-5′

^a13C values were deduced from HSQC and HMBC spectra.

^bCarbon type derived from the HSQC spectrum.

^cProtons showing long-range correlation with indicated carbons.

Close inspection of the ¹H NMR spectrum and ¹³C NMR chemical shifts (deduced indirectly from HSQC and HMBC spectra) identified resonances characteristic of a highly oxygenated structure. The HMBC spectrum confirmed the presence of two carbonyl groups (C-1, δ_C 176.3; C-9, δ_C 204.7) indictive of an ester and ketone moieties, respectively, in addition to three other quaternary carbons; one oxygenated (C-3, δ_C 98.3) corresponding to a hemiketal and one olefinic (C-11, δ_C 148.5). Analysis of the HSQC spectrum identified 7 methyl groups (two oxygenated; C-2′-OMe, δ_C 58.9; C-4′-OMe, δ_C 61.0), 14 methines (8 oxygenated with one acetal type (C-1′, δ_C 94.1) and 5 olefinic), and three methylenes. Detailed analysis of ¹H–¹H COSY and TOCSY spectra revealed six spin systems (substructures I–VI) (Figure 2A), suggesting that the compound is closely related to previously isolated macrolide glycosides, aurisides (Figure 1).^{9,10,12,17−19}

Substructure I was identified as a bromine-substituted conjugated diene based on the characteristic chemical shifts (C-14, δ_C 131.2; C-15, δ_C 129.3; C-16, δ_C 136.1; C-17, δ_C 110.5) of 1 and similar moieties present in other marine-derived macrolide glycosides.^17,18 The E,E configuration of the conjugated diene was assigned on the basis of the ³J_H,H coupling constants for H-17/H-16 (13.5 Hz) and H-15/H-14 (15.0 Hz). Substructure II is a methyl-substituted conjugated enone motif, found in aurisides, and justifies the downfield chemical shift of the methyl (H₃-21; δ_H 2.12).¹⁷ Substructure III is composed of a quaternary carbon (C-8, δ_C 49.3) and two singlet methyls (H₃-19, δ_H 1.02; H₃-20, δ_H 1.27). Substructure IV was identified as a pyran ring containing a characteristic oxygenated quaternary carbon (C-3, δ_C 98.3), which corresponds to a hemiketal, based on its chemical shift. This unit was disclosed based on COSY correlations between H₂-4/H-5, H-5/H₂-6, and H₂-6/H-7. Furthermore, a ⁴J_H,H coupling (W-type; 2.6 Hz) between H-4a and 3-OH was evident. The ring was assembled based on the remaining 3 degrees of unsaturation suggesting the presence of 3 more rings. This was supported by an NOE correlation between 3-OH and H-7, which not only established the linkage between C-3 and C-7 (δ_C 74.0) via an oxygen atom but also identified the relative conformation of the pyran ring. The large coupling constants of 11.7 Hz between H-4a/H-5, H-5/H-6a, and H-6a/H-7 suggested a chair conformation of the ring with axial orientation of these protons; hence, substituents at C-5 and C-7 were equatorial. The NOE correlation between 3-OH and H-7 suggested that 3-OH was axial, as well. Substructure V was identified as a methine (C-2, δ_C 47.7/δ_H 2.62) connected to methyl (C-18, δ_C 12.4/δ_H 1.15) and ester (C-1, δ_C 176.3) based on COSY and HMBC correlations. The H₃-18 showed a NOE correlation to 3-OH, indicating that they have the same orientation. Substructure VI was identified to be a 2,4-di-O-methylrhamnopyranoside based on the chemical shifts (C-1′, δ_C 94.1; C-2′, δ_C 81.0; C-3′, δ_C 71.3; C-4′, δ_C 84.0; C-5′, δ_C 67.5) and comparison with literature data as well as COSY correlations from H-1′ through H-5′. The ring was established based on HMBC correlation between H-1′ and C-5′. The remaining two methoxy groups (δ_C/H 58.9/3.50, δ_C/H 61.0/3.58) exhibited HMBC correlations to C-2′ and C-4′, respectively. The presence of an OH group at C-3′, despite the lack of evidence by ¹H NMR, was based on the ¹³C chemical shift (δ_C 71.3) where C-3′ was clearly oxygenated. The relative configuration of the rhamnopyranoside was deduced based on coupling constants and NOE correlations (Figure 2B). The 9.5 Hz coupling between H-4′/H-5′ and H-4′/H-3′ suggested axial positioning of these protons. Furthermore, J_1′,2′ of 1.5 Hz suggested equatorial positioning of H-1′/H-2′ while J_2′,3′ of 3.7 Hz indicated eq/ax orientation of H-2′/H-3′. This data was further supported by the lack of NOE correlation between H-1′/H-3′ resulting from 1,3-diaxial interactions, confirming the equatorial orientation of the anomeric proton H-1′ and establishing the relative configuration of rhamnopyranoside to be of an α form. Furthermore, NOE correlation between H-1′/2′-OMe (Figure 2B) and the ¹J_CH coupling constant of 166 Hz between C-1′/H-1′, as in aurisides, further supported the α-anomeric structure of 1.

The connectivity between the substructures was established based on correlations in the HMBC spectrum (Figure 2A). In substructure I, the methylene H₂-12 showed HMBC correlation to the olefinic carbons (C-10 and C-11) in substructure II. The protons of the dimethyl groups (H₃-19 and H₃-20) at C-8 in substructure III exhibited HMBC correlations to the ketone C-9 and methine C-7, thus establishing the connection to substructures II and IV, respectively. Substructures IV and V were connected based on HMBC cross-peaks between H₃-18/H-2 and the quaternary C-3 of the pyran ring. Although the anomeric (H-1′) and methine (H-5) protons did not exhibit any HMBC correlations, substructures IV and VI were connected based on the NOE correlation between H-1′ and H-5. This linkage was further supported by the ¹³C chemical shifts (C-1′, δ_C 94.1; C-5, δ_C 70.4) which indicated oxygenation of both carbons, with C-1′ being connected to two oxygens. Based on the unsaturation number, one ring is remaining, and hence substructures I and V were linked via an ester linkage to form the 14-membered lactone. This connection was based on the ¹H and ¹³C NMR chemical shifts of the oxygenated methine (C/H-13, δ_C 71.1/δ_H 5.68) as no HMBC correlation between the methine proton H-13 and the carbonyl C-1 was evident.

1D NOE experiments as well as analysis of coupling constants (Table 1) revealed the relative configuration of the macrolide. Correlations between H-10/H-12b, H₃-19/H-6b, and H-10/H₃-20 indicated these protons were positioned at the same side of the macrolide. Furthermore, NOE correlations between H-13/H-12a, H-13/H₃-21, and H₃-21/H-12a were evident, which established the orientation of the pendent side chain (substructure I) at C-13 and the relative configuration of the macrolide (Figure 2B). The relative configuration was also established between the macrolide and rhamnose moiety based on NOE correlations (H-1′/H-5 and H-1′/H-6b) between the two moieties (Figure 2B).

The absolute configuration was inferred from a comparison of the ¹H and ¹³C NMR chemical shifts (Table S1) and optical rotation data between 1 ([α]_D²⁰ –32 (c 0.04, MeOH)) and the closely related aurisides A and B ([α]_D²⁰ –43 (c 0.05, MeOH) and [α]_D²⁰ –30 (c 0.09, MeOH), respectively). This was further supported by the accomplishment of the total synthesis of aurisides which validated their configurational assignments.²⁴

Given the reported moderate cytotoxicity of other macrolide glycosides in selected cancer cell lines, we evaluated the antiproliferative activity of 1 against HCT116 colon cancer cells. Moorenaside resulted in 55% cell viability at 32 μM (the highest concentration tested), which is comparable to lyngbyaloside and 2-epi-lyngbyaloside when evaluated against HT29 colorectal adenocarcinoma cells (IC₅₀ of 37 μM and 38 μM, respectively).¹⁰

The presence of the Michael acceptor motif in 1 prompted us to investigate its ability to activate the Kelch ECH-associated protein 1 (Keap1)/nuclear factor erythroid 2-related factor 2 (Nrf2)–antioxidant response element (ARE) pathway. This pathway provides a cytoprotective mechanism against oxidative damage and is responsible for the maintenance of cellular redox homeostasis and the regulation of inflammation. The cytoplasmic repressor of Nrf2, Keap1, is highly rich in cysteine (27 residues), and upon alkylation of specific cysteine residues (Cys151) with electrophiles via 1,4-Michael addition reaction, Nrf2 undergoes nuclear translocation resulting in its binding to ARE and activation of antioxidant (Nrf2) genes (NQO1, HMOX-1) and suppression of proinflammatory (NF-κB) genes (iNOS, COX2).³²⁻³⁴ Therefore, targeting the Keap1/Nrf2–ARE pathway is considered an attractive therapeutic strategy to combat oxidative-stress-mediated inflammatory diseases. This was further supported by having the Nrf2 activators, bearing a Michael acceptor motif, bardoxolone methyl in phase III clinical trials (NCT03550443) and dimethyl fumarate in the market for the management of diabetic kidney disease and multiple sclerosis, respectively.^32,35−37 Several cyanobacterial-derived Nrf2 activators bearing an α,β-unsaturated carbonyl were reported, including malyngamide F acetate,³⁸ honaucin A,³⁹ anaenamides C and D,⁴⁰ and more recently 7(E)-9-keto-hexadec-7-enoic acid.⁴¹ Moorenaside was tested in an ARE-luciferase reporter gene assay using human embryonic kidney cells (HEK293 cells) stably transfected with a firefly luciferase reporter gene. The compound induced Nrf2 activity 7.7- and 2.5-fold at 32 and 10 μM, respectively (Figure 3A). Cell viability assay was measured using MTT under the same experimental conditions and time points, and none of the tested concentrations of 1 were cytotoxic (Figure 3B). To further correlate the activity with the presence of the conjugate carbonyl moiety in 1, the ARE-luciferase activity for 18Z-lyngbyaloside C, a structurally related analogue lacking the Michael acceptor motif (Figure 1), was carried out (Figure S1). A significant reduction in the activity was noted in response to 18Z-lyngbyaloside C treatment compared to moorenaside (Figure S1).

Figure 3.

ARE-luciferase activity for moorenaside (1) in HEK293 cells. (A) Compound 1 activated the ARE-luc reporter at 32 μM. The activity was measured 24 h after treatment with 1. tert-Butylhydroquinone (tBHQ) (10 μM) was used as a positive control. (B) Cell viability in HEK293 cells assessed following 24 h of treatment with different concentrations of 1 using MTT. Data represent mean ± SD (n = 3).

Due to the established role of Nrf2 in the regulation of inflammation,⁴² we subsequently investigated the anti-inflammatory activity of 1 by measuring the production of nitric oxide (NO) in culture supernatant using LPS-activated mouse macrophage RAW264.7 cells, a commonly used cellular model for inflammation. The compound significantly reduced NO levels at 32 μM without any cytotoxicity, assessed in parallel using MTT (Figures 4A and B).

Figure 4.

Anti-inflammatory activity of moorenaside (1) in RAW264.7 cells. (A) NO production 24 h post-treatment measured using Griess reagent. RAW264.7 cells were pretreated with different concentrations of 1 for 1 h followed by LPS stimulation (1 μg/mL). (B) Cell viability in RAW264.7 was assessed using MTT following 24 h treatment with different concentrations of 1 and LPS (1 μg/mL). (C,D) Relative transcript levels of iNos and Nqo1 measured by TaqMan analysis, using actin as an endogenous control. RAW264.7 cells were pretreated with different concentrations of compound 1 for 1 h followed by LPS stimulation (1 μg/mL) and a 12 h incubation. Data represent mean ± SD (n = 3). The asterisks denote significance of p < 0.05 relative to DMSO + LPS using two-tailed unpaired t-test (** denotes p ≤ 0.01, *** denotes p ≤ 0.001, **** denotes p ≤ 0.0001).

The effect of 1 on the transcriptional regulation of NO levels was assessed in LPS-stimulated RAW264.7 cells, by measuring the relative transcript levels of the inducible nitric oxide synthase (iNOS) and NAD(P)H:quinone oxidoreductase 1 (Nqo1), well-known target genes of NF-κB and Nrf2, respectively. The compound significantly suppressed iNos transcript levels 2.5-fold at 32 μM and concomitantly induced Nqo1 transcript levels 5.0-fold at the same concentration, suggesting that the anti-inflammatory activity of 1 may be functionally linked to Nrf2-ARE activation or that both effects occur independently at the same concentration (Figures 4C and D).

To further capture the effects of 1 on the global network of genes and explore additional upstream regulators and canonical pathways that might be modulated in RAW264.7 cells, RNA sequencing was carried out at two concentrations (32 and 10 μM) followed by Ingenuity Pathway Analysis (IPA) (Figure 5). In order to focus on biologically and statistically significant genes, we used a cutoff criterion of 1.5-fold change (exp log ratio 0.586) and p-value of 0.05. Using these criteria, the data set at 32 μM was highly enriched with 421 differentially expressed genes as opposed to the data set at 10 μM (76 differently expressed genes).

Figure 5.

RNA sequencing data of moorenaside (1) at 32 and 10 μM (12 h) in RAW264.7 cells. (A) Volcano plots of the differentially regulated genes (cutoff values of log2 fold change >1 and p-value <0.05) in response to treatment with 1. (B) Heatmap of differentially regulated transcripts by LPS with and without pretreatment with 1 at 32 μM.

The analysis at 32 μM identified Nrf2-mediated oxidative stress response and xenobiotic metabolism general signaling pathways, both being upregulated with z-scores of 2.236 and 2.646, respectively. Moreover, the role of hypercytokinemia/hyperchemokinemia in the pathogenesis of influenza, pathogen-induced cytokine storm, macrophage classical activation, and multiple sclerosis signaling pathways as well as pattern recognition receptors are among the top five downregulated canonical pathways identified in the analysis with negative z-scores. Additional canonical pathways involved in inflammation and immune response were also identified and predicted to be downregulated (Table 2).

Table 2. Selected Inflammatory and Immune Canonical Pathways Modulated by 1 at 32 μM Based on IPA (1.5-Fold Cutoff, p-Value <0.05).

Canonical Pathway	Activation z-Score	p-Value	No. of Molecules	Selected Genes
Role of hypercytokinemia/hyperchemokinemia in the pathogenesis of Influenza	–3.441	4.64 × 10–17	19	CXCL10, IFNB1, IL1A, IL1B, IL6, STAT1
Pathogen-induced cytokine storm signaling pathway	–2.335	2.26 × 10–14	31	CSF2, CXCL10, IFNB1, NOS2, IL6, IL1A, IL1B
Macrophage classical activation signaling pathway	–2.683	1.47 × 10–11	20	CSF2, IFNB1, IL1A, IL1B, NOS2, STAT1
Multiple sclerosis signaling pathway	–1.789	2.71 × 10–10	20	CSF2, IL1A, IL1B, STAT1, IL6
Role of pattern recognition receptors in recognition of bacteria and viruses	–1.890	3.18 × 10–10	17	CSF2, IFNB1, IL6, IL1A, IL1B
Interferon signaling	–3.000	2.81 × 10–9	9	IFNB1, STAT1, ISG15
NOD1/2 signaling pathway	–1.500	4.09 × 10–8	16	IL1A, IFNB1, RIGI, CSF2, NOS2
HMGB1 signaling	–1.897	4.95 × 10–8	15	CSF2, IL6, IL1B, IL1A
TREM1 signaling	–1.000	2.76 × 10–6	9	IL1B, IL6, IL18, CSF2
IL-17 signaling	–2.496	6.91 × 10–6	13	IL1A, IL1B, MMP13, CSF2, NOS2
Xenobiotic metabolism general signaling pathwaya	2.646	6.69 × 10–3	7	GSTA5, GSTA3, NQO1, HMOX1, RXRA, MAP3K9
NRF2-mediated oxidative stress responsea	2.236	3.06 × 10–2	8	GSTA5, GSTA3, NQO1, HMOX1

^aThe genes involved in these pathways were upregulated.

The analysis also revealed immunological diseases, organismal injury, inflammatory diseases/response, and immune cell trafficking among the top five diseases and functions, which were predicted to be downregulated based on their negative z-scores. The data supports beneficial activity of 1 as the activation status of selected diseases in the aforementioned categories were predicted to be decreased in response to treatment with 1 (Table S2).

Relapsing-remitting multiple sclerosis was a top hit in the list of diseases (Table S2) with decreased activation status in response to 1 (z-score −2.219; p-value 3.12 × 10^–7) (Figure 6A-left). Analysis of machine learning (ML) disease pathways, which links key genes in the data set to a single disease and its associated phenotypes, identified multiple sclerosis as the top disease pathway with a negative z-score. This analysis also identified oxidative stress (z-score −1.342; p-value 3.95 × 10^–4) predicted to be decreased and associated with five key genes in the data set (upregulated: GCLM, HMOX1, Mt1, Mt2; downregulated: NR4A3) (Figure S2). The list of functions revealed functions related to free radical scavenging with predicted decrease in the quantity of reactive oxygen species (ROS) (z-score −2.151; p-value 5.56 × 10^–10) (Figure 6A-right). Taken together, the analysis of diseases and functions supports previous findings related to the role of oxidative stress and ROS in the pathogenesis of multiple sclerosis.³⁶ Also, in line with the established crosstalk between oxidative stress and inflammation, the oxidative stress network (Figure S2) shows and predicts Nrf2 activation linked to the upregulated HMOX1 in the data set (exp log ratio 1.031), a key player involved in the inhibition of NF-κB proinflammatory genes.^43,44

Figure 6.

IPA analysis of RNA sequencing data of 1 at 32 μM. (A) Diseases (left) and functions (right) predicted to be inhibited and genes with measurement direction supporting their inhibition. (B) Upstream regulator bardoxolone methyl with key genes supporting its predicted activation status resulting in anti-inflammatory effect. (C–E) Regulator effect networks of IKKβ, fingolimod, and dimethyl fumarate, linking upstream regulator through genes in the data set to a phenotype.

Upstream analysis identified LPS as the top upstream regulator and predicted it to be inhibited based on its activation z-score (−7.075). As expected, the upstream regulator NFE2L2 (Nrf2) was activated with 33 of 42 genes having a measurement direction consistent with its activation (z-score 4.638). Interestingly, the steroidal drug dexamethasone, known for its anti-inflammatory activity, was identified in the analysis with 70 of 125 genes having measurement direction consistent with its activation (z-score 3.307), suggesting that 1 act similarly to dexamethasone. Several studies had reported the inhibitory effects of dexamethasone on iNos expression and NO production.^45,46 Additionally, the analysis also identified bardoxolone methyl, a natural product-derived Nrf2 activator in phase III clinical trials for the treatment of diabetic kidney disease, predicted to be activated (z-score 2.588) with 7 of 7 genes (upregulated: HMOX1, GCLM; downregulated: IL6, Ccl2, MYC, IL1β, NOS2) in the data set having measurement direction consistent with its activation (Figure 6B).³⁵

Regulator effect networks, which link an upstream regulator through differentially expressed genes in the data set to a phenotype, revealed downregulation of the inhibitor of nuclear factor kappa B kinase subunit beta (IKKβ) among the top three regulators identified in the analysis (Figure 6C). The predicted inhibition of IKKβ (z-score 3.425) is linked to the activation of HMOX1 and inhibition of iNOS with subsequent targeting of inflammation (Figure 6C). The analysis also revealed fingolimod and dimethyl fumarate, which are ARE activators used for the treatment of relapsing-remitting multiple sclerosis, to have effects similar to those of 1 (Figure 6D,E). Several studies have supported their anti-inflammatory and neuroprotective effects through suppression of iNOS and other pro-inflammatory cytokines.⁴⁷⁻⁵⁵ The naturally occurring epigallocatechin and 15-deoxy-Δ^12,14-prostaglandin J2 were also identified and predicted to act similarly as 1 (Figure S3). Epigallocatechin is a polyphenolic compound commonly extracted from green tea, known for its antioxidant and anti-inflammatory effects, has just completed phase II trials for inflammation (NCT04553666).^56,57 It has been reported to modulate the NF-κB signaling pathway and subsequently suppress LPS-induced inflammation in RAW264.7 mouse macrophages via suppression of NO and ROS production and inhibition of IκB and p65 phosphorolation.⁵⁸ 15-deoxy-Δ^12,14-prostaglandin J2 is an endogenous agonist of PPARδ, structurally characterized by the presence of a Michael acceptor motif, has been reported to negatively regulate ROS, iNOS, proinflammatory mediators such as IL-1β and TNF-α in addition to direct inhibition of IκB and DNA binding of NF-κB.^59,60

RNA sequencing was also carried out at two concentrations (32 and 10 μM) of moorenaside (1) in HCT116; however, no differential expressions between the treated samples and the control were observed, hence highlighting the potential selectivity of 1 toward RAW264.7 cells.

Inflammation is a pervasive mechanism to disturbances in tissue homeostasis such as injury or oxidative stress resulting from high levels of ROS.⁴² It is considered the most significant cause of death as it underlies the pathogenesis of several diseases including chronic obstructive pulmonary disorder (COPD), diabetes, chronic kidney disease, neurodegenerative and cardiovascular diseases, and can even progress to cancer.⁶¹ Our analysis of global LPS-induced transcript changes in response to 1 identified canonical pathways and regulator effect networks that support the role of Nrf2 activation in inflammation, the immune response, and certain underlying diseases such as multiple sclerosis and chronic kidney disease.

In multiple sclerosis, chronic cycles of oxidative stress and inflammation result in mitochondrial dysfunction and dysregulated immune response, leading to the hallmark features of the disease, demyelination and axonal degeneration.³⁷ It was demonstrated that activated immune cells induce kinases and transcriptional factors such as MAPKs, AP1, NF-κB as well as genes involved in the pathogenesis of multiple sclerosis, including TNFα, iNos, and IL1α/IL1β.³⁶ All of these were found to be downregulated in our transcriptomic analysis in response to treatment with 1. Mounting evidence supports the role of Nrf2 in multiple sclerosis pathogenesis and as a potential therapeutic target.^36,37,62 Nrf2^–/– microglia was reported to enhance the expression of IL-6, IL-1β, and iNos but fail to express HO-1 and Nqo1.⁶² Interestingly, our analysis identified relapsing-remitting multiple sclerosis as a top hit in the list of diseases with decreased activation statuses in response to 1. The analysis also predicted 1 to act similarly as dimethyl fumarate and fingolimod, two FDA approved Nrf2 activators for the management of relapsing-remitting multiple sclerosis, further supporting the role of Nrf2 activation as a potential therapeutic target.

Recently, due to the unmet therapeutic needs, there has been increasing interest in Nrf2 activators for the prevention/management of chronic kidney disease, where oxidative stress is a key player involved in its pathogenesis and progression. Chronic kidney disease is characterized by increased levels of ROS, inflammation, and fibrosis, leading to progressive loss of renal function.⁶³ Acute kidney disease can also promote progression to chronic disease. Beneficial effects of various Nrf2 activators, such as curcumin, resveratrol, epigallocatechin-3-gallate, and bardoxolone methyl, were reported in various preclinical experimental models of acute as well as chronic kidney diseases.⁶³ In several models of acute kidney disease, bardoxolone methyl treatment resulted in an increase in the expression levels of Nrf2, HO-1, and Nqo1, all of which were upregulated in our data set.⁶³ Interestingly, renal impairment and chronic kidney disease were predicted to be inhibited in response to 1 in our IPA analysis of RNA sequencing data (Table S2). Bardoxolone methyl was also identified in the analysis of regulator effect networks and predicted to act similarly as 1, modulating a similar set of genes. Bardoxolone methyl was among the most extensively studied Nrf2 activators in clinical trials, but despite promising effects in improving renal function, concerns related to its cardiovascular side effects were reported.⁶³ Therefore, pleiotropic effects of Nrf2 activators bearing conjugated enone need to be taken into consideration, as they contribute to both therapeutic efficacy and toxicity.

In conclusion, we report the discovery of a new anti-inflammatory macrolide glycoside (Figure 1) containing an α,β-unsaturated carbonyl (1) from a benthic marine cyanobacterium Moorena sp. To the best of our knowledge, this is the first report of the anti-inflammatory activity of a cyanobacterial-derived macrolide glycoside belonging to the auriside class of natural products, targeting the Keap1/Nrf2–ARE pathway. Our data support the therapeutic potential of this class of compounds, containing a Michael acceptor motif, in oxidative stress and inflammatory-mediated diseases where the Keap1/Nrf2–ARE pathway serves as a target.

Experimental Section

General Experimental Procedures

The optical rotation was measured using a JASCO P-2000 polarimeter. ¹H and 2D NMR spectra for 1 in CDCl₃ and C₆D₆ were recorded on a Cryo 600-MHz/54 mm Bruker Avance III HD (Bruker Biospin Corporation, Billerica, MA, USA). The spectra were referenced using the residual solvent signal (δ_H/C 7.26/77.0 (CDCl₃); δ_H/C 7.16/128.0 (C₆D₆)). The HRESIMS data were acquired in positive mode using a high-resolution LC Q-Exactive orbitrap mass spectrometer (ThermoSci, Waltham, MA, USA) equipped with the APCI/ESI multimode ion source detector.

Biological Material

Samples of benthic cyanobacterium VPFK22-6, Moorena sp., were collected off Shands Key in Florida on May 21, 2022. The 16S rRNA gene sequence has been deposited in GenBank (accession #PP555179).

Extraction and Isolation

The freeze-dried samples of VPFK22-6 Moorena cyanobacterium were subjected to nonpolar and polar extractions using EtOAc–MeOH (1:1) and 30% aq EtOH, respectively. The nonpolar extract (1.8 g) was partitioned between EtOAc and H₂O. The EtOAc soluble fraction (176.6 mg) was concentrated and subsequently subjected to silica gel chromatography using a solvent gradient of 30% EtOAc–Hex, 100% EtOAc, 10% MeOH–EtOAc, 1:1 EtOAc–MeOH, and finally 100% MeOH. The EtOAc silica fraction (26 mg) was further chromatographed by reversed-phase HPLC [Synergi Hydro 4u-RP, 250 × 10.0 mm; flow rate, 2.0 mL/min; PDA detection 200–800 nm] using a linear MeOH–H₂O gradient (20–100% MeOH over 15 min, 100% MeOH for 20 min) and generating five peaks eluting with 100% MeOH. These peaks were further resolved by HPLC using an analytical column [Synergi 4u Hydro-RP 80A, 150 × 4.60 mm, 4 μm; flow rate 0.5 mL/min, PDA detection 200–800 nm] using a linear ACN–H₂O gradient (60–100% ACN for 10 min, 100% ACN for 15 min) to afford 1 (0.9 mg) eluting at t_R 10.6 min. Compound 1 was repurified by HPLC using an analytical column [Luna 5 μm Phenyl-Hexyl 100A, 250 × 4.60 mm; flow rate 0.5 mL/min, PDA detection 200–800 nm] using a linear ACN–H₂O gradient (70–100% ACN for 10 min, 100% ACN for 10 min) to afford 1 (0.5 mg; 0.03% yield) eluting at t_R 11.4 min.

Moorenaside (1)

Colorless, amorphous powder; [α] _D²⁰ –32 (c 0.04, MeOH); NMR data (CDCl₃ and C₆D₆), Tables 1 and S1; HRESIMS m/z 653.1926 [M + Na]⁺ (calcd. for C₂₉H₄₃⁷⁹BrO₁₀Na, 653.1937) and m/z 655.1909 [M + Na]⁺ (calcd. for C₂₉H₄₃⁸¹BrO₁₀Na, 655.1917) (1:1 [M + Na]⁺ ion cluster).

Cell Culture

HEK293 ARE-luc cells (SL-0042-NP; Signosis) and RAW264.7 cells (ATCC) were propagated and maintained using Dulbecco’s Modified Eagle’s Medium (DMEM; Invitrogen, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS; Sigma, St. Louis, MO, USA), and 1% antibiotic-antimycotic (Invitrogen) at 37 °C humidified air and 5% CO₂.

Cell Viability Assay

HEK293 ARE-luc cells and RAW264.7 cells were seeded (10,000 cells/well and 20,000 cells/well, respectively) in 96-well plates. Following overnight incubation, the cells were treated with half-log serial dilutions of 1 starting at 32 μM, positive control tBHQ (10 μM), and 0.5% DMSO as the solvent control. After 24 h incubation, cell viability was measured using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazoliumbromide according to the manufacturer’s instructions (Promega, Madison, WI, USA). For cytotoxicity against human cells, HCT116 cells were seeded (10,000 cells/well) and incubated overnight. The cells were then treated with half-log serial dilutions of 1 starting at 32 μM and 0.5% DMSO as a solvent control. Following 48 h of incubation, cell viability was measured.

ARE-luciferase Reporter Assay

HEK293 ARE-luc cells were seeded (10,000 cells/well) in 96-well white solid plates (Costar; COS3917) and allowed to attach overnight. Following overnight incubation, the cells were treated with half-log serial dilutions of 1 starting at 32 μM, positive control tBHQ (10 μM), and 0.5% DMSO as the solvent control. After 24 h of incubation, BriteLite detection reagent (PerkinElmer) was added to each well according to the manufacturer’s protocol and the luciferase activity was measured using Envision plate reader (PerkinElmer, Waltham, MA, USA).

NO Assay

RAW264.7 cells were seeded (20,000 cells/well) in 96-well plates and allowed to attach overnight. Following overnight incubation, the cells were pretreated for 1 h with half-log serial dilutions of 1 starting at 32 μM and 0.5% DMSO as a solvent control. After 1 h incubation, the cells were stimulated with 1 μg/mL LPS. Non LPS treated cells (nonstimulated) were also tested simultaneously. After 24 h of incubation, 50 μL of culture supernatant was mixed with 50 μL of Griess reagent (Promega) according to the manufacturer’s protocol, and the absorbance was measured at 540 nm using SpectraMax M5 plate reader (Molecular Devices, SAN Jose, CA, USA). A standard curve was generated using the nitrile standard provided, and NO production was calculated.

RNA Isolation, Reverse Transcription, and Real-Time Quantitative Polymerase Chain Reaction (RT-qPCR)

RAW264.7 cells were seeded (175,000 cells/well) in 12-well plates and incubated overnight. Following overnight incubation, the media were replaced with fresh media, and the cells were treated for 1 h with half-log serial dilutions of compound 1 starting at 32 μM and solvent control (0.5% DMSO). After 1 h incubation, the cells were stimulated with 1 μg/mL LPS. Non-LPS treated cells (nonstimulated) were set up simultaneously. Following 12 h incubation, RNeasy mini kit (Qiagen) was used to isolated RNA according to the manufacturer’s instructions. cDNA synthesis was performed using SuperScript II reverse transcriptase and oligo (dT) (Invitrogen). Following reverse transcription, qPCR was performed on a reaction solution (25 μL) prepared using a 1 μL aliquot of cDNA, 12.5 μL of TaqMan gene expression assay mix, 1.25 μL of 20× TaqMan gene expression assay mix, and 10.25 μL of RNase-free water. The qPCR experiment was carried out using an ABI 7300 sequence detection system. The thermocycler program used was: 2 min at 50 °C, 10 min at 95 °C, 15 s at 95 °C (40 cycles), and 1 min at 60 °C. iNos (Mm00440502_m1) and NQO1 (Mm01253561_m1) were used as target genes, and mouse ACTB (no. 4352933, Applied Biosystems) was used as endogenous control.

Library Construction and Illumina NovaSeq6000 Sequencing

The QUBIT fluorescent method (Invitrogen) and Agilent Bioanalyzer were used to measure the RNA samples. The library construction was carried out as described using 500 ng high-quality total RNA with a RIN of 9.6.⁴¹ RNA-seq library was performed at UF ICBR Gene Expression Core (accessed on 18th of July 2023; https://biotech.ufl.edu/gene-expression-genotyping/).

Illumina NovaSeq6000 Sequencing

The Illumina NovaSeq 6000 was used to sequence the libraries for 2 × 150 cycles as described.⁴¹ Sequencing was performed at the ICBR NextGen Sequencing Core (https://biotech.ufl.edu/next-gen-dna/). The data are deposited with GEO accession number: GSE262850.

Data Availability Statement

The 16S rRNA gene sequence was deposited in GenBank (accession #PP555179). The RNAseq data were deposited to Gene Expression Omnibus (GEO) (accession number: GSE262850). The NMR data for moorenaside has been deposited in the Natural Products Magnetic Resonance Database (NP-MRD; https://np-mrd.org/) and can be found at NP0333275.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jnatprod.4c00420.

• ¹H NMR, HSQC, COSY, TOCSY, HMBC (CDCl₃ and C₆D₆), and 1D NOESY in CDCl₃ for moorenaside (1); ARE-luciferase activity of 18Z-lyngbyaloside C in HEK293 cells (Figure S1); Oxidative stress network (Figure S2); Regulator effect networks of epigallocatechin and 15-deoxy-Δ^12,14-prostaglandin J2 (Figure S3); ¹H and ¹³C NMR chemical shifts of 1, aurisides A and B (Table S1); Selected diseases predicted to be decreased in response to 1 at 32 μM based on IPA (Table S2) (PDF)

The authors declare no competing financial interest.