Hepatoprotective Activity of Lignin-Derived Polyphenols Dereplicated Using High-Resolution Mass Spectrometry, In Vivo Experiments, and Deep Learning

Abstract

Chronic liver diseases affect more than 1 billion people worldwide and represent one of the main public health issues. Nonalcoholic fatty liver disease (NAFLD) accounts for the majority of mortal cases, while there is no currently approved therapeutics for its treatment. One of the prospective approaches to NAFLD therapy is to use a mixture of natural compounds. They showed effectiveness in alleviating NAFLD-related conditions including steatosis, fibrosis, etc. However, understanding the mechanism of action of such mixtures is important for their rational application. In this work, we propose a new dereplication workflow for deciphering the mechanism of action of the lignin-derived natural compound mixture. The workflow combines the analysis of molecular components with high-resolution mass spectrometry, selective chemical tagging and deuterium labeling, liver tissue penetration examination, assessment of biological activity in vitro, and computational chemistry tools used to generate putative structural candidates. Molecular docking was used to propose the potential mechanism of action of these structures, which was assessed by a proteomic experiment.

Keywords: NAFLD; complex mixtures; dereplication; hepatoprotective activity; labeling; mass spectrometry; polyphenolic composition; structure.

Figure 5

(a) A model of full-length KEAP1 predicted by AlphaFold2. A BTB domain (amino acids 61–179) is shown in cyan, an intervening region (IVR, amino acids 180–314) is shown in orange, and a Kelch domain (Kelch) is shown in red. N-terminal (NTR, amino acids 1–60) and C-terminal regions (CTR, amino acids 599–624) are shown in grey. Cysteine residues, which were investigated in the proteomics experiments, are shown in magenta (non-alkylated ones) and lime (alkylated). Domains are designated as in ref. [35]. (b) Docking pose of the generated compound (3,5,7-trihydroxy-2-[(4-hydroxy-3-(3-methylbut-2-en-1-yl)phenyl)methyl]-2,3-dihydro-4H-1-benzopyran-4-one), shown as ball-and-stick model) in the region close to Cys273. Positively charged amino acids (Lys312, His274, Arg269) forming the entrance to the cavity are shown. Visualization is made in VMD 1.9.3. [36].

Figure 1

The dereplication pipeline used in this work. H/D exchange was used to label lignin-derivative and find its component in liver by FTICR MS. The sample and its enriched fraction were tested on its ability to alleviate NADFL-related biochemical changes. LC-MS experiments assessing the ability of samples to interact with the KEAP-1 were then performed. Further, a recurrent neural network was trained on the natural compound structures from CoCoNut and fine-tuned on compounds possessing activity against KEAP1-Nrf2 system retrieved from ChEMBL. The network was used to sample various natural compound-like structures that were filtered by the relevance according to the experimental MS data including isotopic labeling. Finally, the ability of the filtered compounds to interact with the KEAP-1 was assessed using molecular docking simulations.

Figure 2

Van Krevelen diagrams for BP-Cx-1 fractions obtained by gradient extraction (A); summarized contribution of molecular components found in liver to the total intensity in mass-spectra of fractions (B).

Figure 3

Relative intensities of alkylated peptides containing free cysteine without (blue) and with (red) incubation with BP-Cx-1 determined by LC-MS.

Figure 4

Comparison of the chemical spaces of natural products from CoCoNut database (only CxHyOz), compounds active against Nrf2 and generated compounds using self-organizing map (SOM) (100 × 100 neurons, FragFp fingerprints, see Materials and Methods for details). (a) The SOM built on CoCoNut compounds colored by most frequent scaffolds (see Table S7). Red—benzene ring, cyan—sterone ring, orange—coumarin, blue—flavone, green—tetrahydropyran, grey—compounds bearing other scaffolds. Markers for compounds bearing frequent scaffolds were enlarged for better visibility. (b) The SOM built on CoCoNut compounds (grey) on which generated compounds (dark blue) and compounds active against Nrf2 (magenta) were projected. Markers for CoCoNut compounds and compounds active against Nrf2 were enlarged for better visibility.