Deuterated methanol (purity 99.8%) and deuterium oxide (99.9%) were supplied from Deutero GmbH (Kastellaun, Germany). Dulbecco’s Modified Eagle Medium (DMEM) with high glucose 4.5 g/L (#D5796), fetal bovine serum (FBS; #F7524), 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT; #M21281), penicillin/streptomycin/amphotericin B (10,000 IU/10 mg/25 μg; #A5955), trypsin-EDTA (#59418C), dexamethasone (DEXA; purity ≥98%; #D1756), rosmarinic acid (purity ≥98%, #R4033), trimethyl silylpropionic acid sodium salt-d4 (TSPA-d4; #11202), methanol, acetic acid of HPLC grade, Bradford reagent (#B6916), RIPA lysis buffer (#R0278), protease and phosphatase inhibitor cocktail (#PPC1010) and RNAzol RT reagent (#R4533) were from Merck KGaA (Darmstadt, Germany). The recombinant human IL-17A (#ENZ-PRT188) and IL-22 (#ENZ-PRT250) were purchased from Enzo Life Sciences AG (Lausen, Switzerland) and IFN-γ (#10067-IF) was obtained from R&D Systems (Minneapolis, MN, United States).

Buffers and chemicals used to perform electrophoresis, immunoblotting and real-time quantitative reverse transcription-polymerase chain reaction (RT-qPCR) were obtained from Bio-Rad Laboratories, Inc (Hercules, CA, United States). For the Western blotting analysis primary antibodies against the following proteins were used: AKT (#9272), JAK2 (#3230S), PI3K (#4257), STAT1 (#14994) and phospho-STAT1 (#7649) from Cell Signaling Technology (Leiden, Netherlands); tubulin (#12004166) from Bio-Rad. Secondary anti-rabbit IgG Star-Bright Blue 700 (#12004162) antibody from Bio-Rad was used for fluorescent detection. All other materials and substances were of analytical grade and were purchased from Merck KGaA (Darmstadt, Germany) unless otherwise specified.

The LV suspension culture was cultivated as previously reported on liquid Linsmayer and Skoog (LS) nutrient medium supplemented with 0.2 mg/L 2,4-dichlorophenoxyacetic acid and 30 g/L sucrose in darkness at 26°C on a rotary shaker at 100 rpm (Georgiev et al., 2006). Following cultivation, the cell suspension was freeze dried and extracted with 50% aqueous methanol under sonication at room temperature (RT) for 20 min. Further, the extract was filtrated, vacuum concentrated at 40°C, lyophilized and stored at −20°C until use.

Nuclear magnetic resonance (NMR) analysis was performed according to the protocol described by (Georgiev et al., 2015). Sample of 10 mg of the obtained LV extract was dissolved in equal amounts of 0.4 ml CD₃OD and D₂O, containing KH₂PO₄ buffer with pH 6.0 and TSPA-d4 as internal standard at final concentration of 0.005% (w/v). ¹H NMR and 2D: J-resolved, homonuclear correlation spectroscopy (COSY) and heteronuclear single quantum coherence spectroscopy (HSQC) spectra were obtained on a Bruker AVII+ 600 spectrometer (Bruker, Karlsruhe, Germany), operating at 25°C, 600.13 MHz proton frequency, with 4.07 s relaxation time and CD₃OD as an internal lock.

Quantification of the RA content in the extract was executed with high-performance liquid chromatography (HPLC). Reference RA standard was dissolved in methanol and standard solutions from 10–200 μg/ml were prepared and filtrated through 0.45 µm syringe filters for standard curve preparation. The suspension extract was prepared in 5 mg/ml solutions in 50% aqueous methanol. The analyses were performed on Waters HPLC system (Milford, MA, United States), consisting of binary pump and dual wavelength (λ) absorbance detector. A reverse-phase Kinetex^® C18, 100 Å (150 × 4.6 mm, 5 μm) core-shell column (Phenomenex, Torrance, CA, United States), operating at 26°C was used. The RA determination was based on an HPLC protocol previously used by Choi et al. (2019) with certain modifications. The mobile phases used were 0.02% aqueous acetic acid (phase A) and methanol (phase B) at flow rate of 0.8 ml/min with the following gradient: change of phase A from 88 to 85 (0–5 min), from 85 to 82 (5–6 min), from 82 to 75 (6–9.5 min), from 75 to 74 (9.5–10.5 min), from 74 to 73 (10.5–12 min), from 73 to 70 (12–20 min), from 70 to 50 (20–25 min), from 50 to 30 (25–30 min) and from 30 to 88 (30–35 min) at λ of 320 nm.

The HPLC method was validated for linearity and sensitivity, including limit of detection (LOD) and limit of quantification (LOQ). An external standard calibration methodology was applied. Seven solutions with different concentrations ranging from 10 to 200 μg/ml were prepared. The analyses were performed in triplicates (for each concentration) and the calibration curve was constructed as a relationship between the peak areas and the used concentrations of the standard. The linearity of the curve was assessed according to the obtained regression equations y = ax + b, where y and x refer to peak area and the concentration of the compound (in μg/ml) and its relevant coefficient of determinations (R²). The sensitivity was determined by calculating the LOD = 3.3σ/S and the LOQ = 10σ/S, where σ is standard deviation of the response and S is the slope of the calibration curve.

Human epidermal keratinocytes (HaCaT) cell line was purchased from Cell Line Service GmbH (Eppelheim, Germany). The HaCaT cells were cultured in DMEM with 4.5 g/L glucose supplemented with 10% FBS and 1% antibiotic and anti-mycotic solution and maintained at 37°C with 5% CO₂. At 80% confluence HaCaT cells were stimulated with combination of IFN-γ/IL-17A/IL-22 (1 ng/ml each) to model psoriasis-like inflammatory milieu 1 h prior treatment with either LV (20, 40, 100 μg/ml), RA (5, 10, 25 µM), vehicle alone (0.02% DMSO) or DEXA 5 µM (as a positive control). Total RNA samples were obtained 6 h after treatment and whole cell protein lysates on the 24th hour following treatment.

To determine the effect of LV and RA at different concentrations on cell viability in HaCaT, MTT assay was conducted. Exposure to LV and RA in HaCaT cells for 24 h did not affect significantly the cell viability up to 100 μg/ml and 100 μM, respectively (Supplementary Figure S1).

Total RNA was isolated using RNAzol RT reagent and reverse transcribed using FirstStrand cDNA kit (#PR008) from Canvax (Cordoba, Spain) according to the manufacturer’s instructions. The relative expression of target genes was detected by RT-qPCR using Sso EvaGreen SuperMix (#1725204; Bio-Rad) on CFX96 detection sys-tem (Bio-Rad) and was quantified by the comparative threshold cycle (_ΔΔCT) method on the CFX Maestro software (Bio-Rad). The primer sequences are listed in Supplementary Table S1. Both GAPDH and TUBB were used as reference genes and served for normalization of all the other gene expression.

Total protein lysates were prepared using RIPA buffer freshly supplemented with 1% protease and phosphatase inhibitor cocktail for 15 min on ice, followed by centrifugation (15,000 rpm, 15 min, and 4°C) and collection of the clear supernatant. The total protein concentration was determined using Bradford assay.

Equal amounts of total protein of 50 μg per lane were separated with vertical electrophoresis on 10% TGX Stain-Free SDS-PAGE gels and transferred to nitrocellulose membranes via semi-dry transfer with Trans-Blot Turbo transfer system (Bio-Rad). Following blocking for 1 h at RT in 5% (w/v) skimmed milk in Tris buffered saline, the membranes were incubated with the primary rabbit antibodies against AKT, JAK2, PI3K, STAT1 and p-STAT1 overnight at 4°C. For fluorescent detection of the target proteins Star-Bright Blue 700 goat anti-rabbit IgG secondary antibody was used. Tubulin was detected with direct primary antibody as a housekeeping protein used for normalization. The immunodetection was visualized with ChemiDoc MP imaging system (Bio-Rad) and the acquired images were analyzed with Image Lab software 6.0.1 (Bio-Rad).

The obtained spectra were automatically reduced to ASCII files using AMIX soft-ware (version 3.7, Bruker), phased, base line corrected and referenced at 0.0 ppm to the internal standard TSPA-d4 using MestReNova software (version 12.0.0, Mestrelab Research, Santiago de Compostela, Spain). All signals were normalized in relation to the peak of TSPA-d4 and scaled to 1.0.

Data evaluation was performed with SigmaPlot software v11.0 (Systat Software GmbH, Erkrath, Germany) and the results are presented as mean ± standard error of the mean (SEM). Differences among groups was determined with unpaired Student’s t-test as appropriate or one-way analysis of variances (ANOVA) with Bonferroni’s post hoc test when more than two groups were compared. The significance level was defined as + p < 0.05, model group compared to non-treated control cells, *p < 0.05 and **p < 0.01, compared to IFN-γ/IL-17A/IL-22-stimulated model group.

Nuclear magnetic resonance (NMR) phytochemical profiling has a great potential for studying living organisms, owing to its non-destructive nature, i.e. it can be used for in vivo and in vitro measurements of biological processes, with no quenching of the metabolism required. The investigated extract in our study was obtained from high RA producing L. angustifolia cell suspension induced and maintained within our lab (Georgiev et al., 2006). The phytochemical characterization has been performed by 1D- and 2D-NMR metabolite profiling. According to the ¹H NMR spectral data the most abundant signals observed in the aromatic region corresponded to RA as a major secondary metabolite. Further, metabolites typical for the primary cell metabolism were identified as well. In the aliphatic region (δ 0.5–3.0 ppm) the signals of several organic acids (acetic, formic, fumaric and pyruvic acids) were detected. The identified metabolites are important intermediates from the tricarboxylic acid cycle (TCA), which is a main source of energy for cells and important part of erobic respiration. The TCA is a part of a larger carbohydrate metabolism in which the metabolites identified in the carbohydrate region (δ 3.0–5.5), such as α-, β-glucose and sucrose are oxidized to form pyruvate. The latter is an entry metabolite of the TCA cycle under the form of acetyl-CoA. The TCA cycle is also the origin of pathways leading to important structural metabolites, including the amino acids alanine, glutamine, glutamate, threonine and valine, identified in the extract (Table 1).

Rosmarinic acid [(R)-a-[[3-(3,4-dihydroxyphenyl)-1-oxo-2E-propenyl]oxy]-3,4-dihydroxy-benzenepropanoic acid] is a polyphenolic compound, an ester of caffeic acid and 3-(3,4-dihydroxyphenyl)-lactic acid. Its structure (C₁₈H₁₆O₈) comprises of two unsaturated six-membered rings (C1-C6) and (C1′-C6′), a double bond (C7′-C8′), an ester group (C9′), a carboxyl group (C9) and four hydroxyl groups (C3, C4, C3′ and C4’; Xiao et al., 2008). In the current investigation, the RA was structurally identified by its 1D and 2D NMR spectra and quantified using high performance liquid chromatography (HPLC). The signals from the proton spectrum (Table 1) of RA identified in the extract were in accordance with the signals previously reported (Pereira et al., 2013; Zhao et al., 2021).

The presence of RA in the analyzed extract was confirmed by the proton-carbon single bond correlations observed in the HSQC spectrum, shown on Figure 1A. The spectrum of the extract revealed the presence of two doublets δ _H 7.52/δ _C 145.14 and δ _H 6.32/δ _C 113.36, which on basis of the large proton-proton coupling were assigned to a pair of trans-olefinic protons (H-7′ and H-8′), indicating the presence of E-caffeic acid moiety. Further, the ABX-system of caffeic acid was confirmed by the three protons H-5′, H-6′ and H-2′, which referred at δ _H 6.88/δ _C 115.21, δ _H 7.04/δ _C 121.59 and δ _H 7.12/δ _C 113.84, respectively. The other ABX-system in the aromatic region was assigned to the 3,4-dihydroxypehenyl unit according to the three protons H-5, H-6 and H-2, which appeared at δ _H 6.78/δ _C 114.87, δ _H 6.72/δ _C 120.61 and δ _H 6.84/δ _C 116.02. The doublet in the low field region at δ _H 5.03/δ _C 75.84 suggested a proton of a CH group attached to an oxygen-bearing carbon was assigned to H-8. Two doublets at δ _H 2.94/δ _C 36.36 and δ _H 3.11/δ _C 36.36 were assigned to the two protons of 7-methine group. In addition, on Figure 1B is shown the long-rage ¹H–¹H COSY interactions. Strong cross-peaks were found between H-7’/H-8’ (δ _H 7.52/δ _C 6.32), H2’/H-5’ (δ _H 7.12/δ _C 6.88) and H2-H-5 (δ _H 6.84/δ _C 6.78; Xiao et al., 2008; Pereira et al., 2013; Zhao et al., 2021).

The HPLC method used to detect and quantify RA was validated for linearity and sensitivity. The obtained calibration curve (y = 7.96 × 10⁴x-3.31 × 10⁵) had a good linear relationship in the concentration range between 10 and 200 μg/ml, characterized with R² of 0.9918. The LOD and LOQ values, 6.9 and 14.59 μg/ml, respectively also indicated a satisfactory sensitivity.

According to the HPLC quantification, the content of RA in the cell suspension resulted to 100.16 ± 7.01 mg/g extract.

The psoriatic condition is characterized by activation of inflammatory signaling pathways. The NF-κB pathway is a key regulatory pathway during inflammation, and is considered to be a crucial mediator in the pathogenesis of psoriasis (Andres et al., 2013; Kim et al., 2014; Zhang et al., 2018). Additionally, JAK/STAT pathway is a classical signal transduction pathway in psoriasis and is triggered upon cytokine stimulation (Shi et al., 2011; Schwartz et al., 2017; Sun et al., 2019). Activation of the p38/JNK/ERK MAPKs (Shi et al., 2018; Liang et al., 2019; Sakurai et al., 2019) and PI3K/AKT (Engelman et al., 2006; Li et al., 2019) signaling pathways contributes to psoriasis exacerbation and disturbed regulation of proliferation and apoptosis in activated keratinocytes.

Stimulation with IFN-γ/IL-17A/IL-22 cytokine combination induced changes in the mRNA expression profile in human keratinocytes that mimic psoriasis-like transcriptional signature (Figure 2). Data from the hierarchical cluster analysis of the relative gene expression obtained with RT-qPCR indicated significant upregulation of genes related to NF-κB, JAK/STAT, p38/MAPK/AKT signaling pathways, as well as pro-inflammatory cytokines. The inflammatory factors IL6 and CCL2 (encoding MCP1 protein) were highly overexpressed as a result of the combined psoriasis-induction stimuli that reflects acute inflammation changes. In contrast, the anti-bacterial peptide S100A7 and the chemokine CCL20 that are frequently used as markers of psoriasis were moderately changed in the model group. The mRNA expression of NFKB1, NFKBIA, IKBKB and RELA was elevated in the model group, hence, suggesting activation of the NF-κB signaling pathway. Furthermore, overexpression was detected for the JAK2, STAT1 and STAT3 genes, which is in agreement with previous reports utilizing similar combinatorial stimulation used for modeling psoriasis in keratinocytes (Kim et al., 2014; An et al., 2018; Zhang et al., 2018; Li et al., 2020). The increase in STAT1 was about 10 folds higher that the overexpression of STAT3 upon stimulation with IFN-γ/IL-17A/IL-22, hence, pointing out STAT1 as the more influenced in our model.

Treatment with LV extract dose-dependently downregulated IL6, CCL2 and CCL20. Further, the inhibition of mRNA expression of genes from the NF-κB signaling (CHUK, NFKB1, NFKBIA, IKBKB, RELA) at the highest treatment concentration of LV (100 μg/ml) exceeded that achieved with the positive control DEXA 5 μM. The profound influence of the LV extract on the NF-κB-related genes suggest its potent anti-inflammatory action in psoriatic keratinocytes. Additionally, the key genes STAT3, AKT and MAPK8 (encoding the JNK protein) were found to be significantly inhibited upon LV application which suggest that the extract interfere with the respective JAK/STAT, MAPK and PI3K/AKT signaling pathways.

Correspondingly, treatment with pure RA in concentrations matching those found in the respective doses of LV extract resulted in similar transcriptional changes. Dose-dependent decrease in the gene expression was registered for CHUK, NFKB1, NFKBIA, IKBKB, RELA upon RA treatment which was higher than the one achieved by DEXA, but not by LV extract. These results hint that RA is only partly responsible for the LV extract inhibition in NF-κB-related genes. The inflammatory mediators IL6 and CCL20 mRNA expression was significantly downregulated by RA in a manner exceeding the effect of LV treatment. Similarly to the LV extract treatment, the gene expression of STAT3 and AKT was influenced by RA, but to a lesser extent. In contrast, RA application did not affected CCL2 expression. Additionally, the MAPK8 (encoding JNK) and MAPK1 (encoding ERK1/2) genes were dose-dependently decreased as a result of RA application. Interestingly, RA treatment in stimulated keratinocytes resulted in significant downregulation of JAK2 and STAT1, which was not observed upon the LV extract application.

Together, these findings indicated that LV extract affects mainly genes from the NF-κB signaling in activated keratinocytes, while RA interfere with JAK/STAT signaling as well.

Cytokines such as IFN-γ, IL-17A or IL-22 bind to cell surface receptors and the signal transduction occurs through JAKs, which in turn promotes STAT1 and/or STAT3 phosphorylation and activation in psoriasis (Wolf et al., 2010; Shi et al., 2011; Kim et al., 2014; Zhang et al., 2018; Itoh et al., 2019). As STAT1 expression was activated predominantly over STAT3 by cytokine stimulation in our study, we then evaluated whether RA could antagonize STAT1 activation induced by IFN-γ/IL-17A/IL-22 in keratinocytes at a protein level. The Western blot analysis (Figure 3) showed that the inhibitory effects of both LV extract and RA on the total protein (Figure 3A) and phosphorylation levels of STAT1 (Figure 3B) were in a concentration-dependent manner in HaCaT cells. Further, at the highest concentration used the decrease in STAT1 and p-STAT1 levels were greater than achieved upon DEXA treatment. Moreover, LV extract and RA significantly decreased the total protein levels of JAK2 (Figure 3C) in a concentration dependent manner. The RT-qPCR analysis showed that the JAK2 and STAT1 transcriptional activity was significantly decreased by RA (Figure 2), which corresponded to the changes observed at a protein level. In contrast, JAK2 and STAT1 mRNA expression levels (Figure 2) were not influenced upon LV treatment, however, their respective protein levels were significantly diminished. The discrepancy between gene expression and protein abundance could be due to either posttranslational modifications or stimulated protein degradation of JAK2 and/or STAT1 induced by the LV extract.

The PI3K/AKT signaling is involved in cell proliferation and is a hallmark of psoriasis in regard to hyper-keratinization (Engelman et al., 2006; Li et al., 2019). The gene expression analysis revealed that both LV and RA downregulated AKT (Figure 2) to an extent similar to that of the DEXA treatment. In order to acquire further clarification of the mechanism that LV and RA inhibits psoriasis-like keratinocyte activation PI3K/AKT pathway was examined at a protein level. The results indicated that both LV and RA treatments at the highest concentrations used resulted in a moderate decrease in AKT (Figure 3D) while PI3K (Figure 3E) was influenced only by the LV extract.

Together, these results indicated that RA interfere with JAK2/STAT1 signaling in keratinocytes. Further, the biotechnologically produced LV extract rich in RA influence both JAK2/STAT1 and PI3K/AKT signaling pathways.