A. citrodora extract was obtained from a Portuguese company (Celeiro^®, batch number 03LUC1099J211S) collected in May 2021 and prepared through infusion extraction. Three concentrations of the A. citrodora extract (0.006 g/mL, 0.008 g/mL, and 0.013 g/mL) were prepared in tap water heated to 90°C using a French press. The resulting mixture was filtered and allowed to cool at room temperature (20–25°C) for 20 min before being given to animals as drinking water every two days.

The LC-DAD-ESI/MSn method (Dionex UltiMate 3000 UHPLC, Thermo Scientific, San Jose, CA, USA) was used to determine the phenolic profile. The compounds were separated and identified following the procedure described by Bessada et al. [16]. The extracts obtained were redissolved in ethanol:water (80:20, v/v) mixture at a concentration of 50 mg/mL. A double online detection was performed using a DAD with preferred wavelengths of 280, 330, and 370 nm, and a mass spectrometer (MS). The MS detection was carried out in negative mode, using a Linear Ion Trap LTQ XL MS (Finnigan, Thermo Fisher Scientific, San Jose, CA, USA) equipped with an ESI source. Phenolic compounds were identified based on their chromatographic behaviour, and UV-vis and mass spectra, by comparison with standard compounds, when available, and data reported in the literature, resulting in tentative identification. The Xcalibur^® data system (Finnigan, Thermo Fisher Scientific, San Jose, CA, USA) was used for data acquisition. For quantitative analysis, a calibration curve was constructed for each available phenolic standard based on the UV-vis signal: apigenin-7-O-glucoside (y = 10,683x – 45,794; R² = 0.996); p-coumaric acid (y = 301,950x + 6,966.7; R² = 0.9999); quercetin-3-O-glucoside (y = 34,843x – 160,173; R² = 0.9998); and verbascoside (y = 124,233x – 18,873; R² = 0.9999). For the identification of phenolic compounds without a commercial standard, quantification was performed using the calibration curve of the most similar available standard. The results were expressed as mg/g of extract.

Thirty female Mus musculus of the FVB/n strain were used in this study: twenty transgenic (HPV+) and ten wild-type (HPV–, WT) mice aged 16–24 weeks. This age group is known to develop dysplastic skin lesions that later progress into squamous cell carcinomas [8]. Female mice were selected due to their ease of maintenance, as males are known to be highly aggressive and territorial [17]. The mouse strain used in this study was generously donated by Doctors Jeffrey Arbeit and Douglas Hanahan from the University of California, through the National Cancer Institute Mouse Repository (USA). The animals were genotyped using a polymerase chain reaction technique, as previously described [18].

This experimental work was conducted at the animal facility of the University of Trás-os-Montes e Alto Douro (UTAD) with the authorization of the Ethics Committee (approval no. 852-e-CITAB-2020_A_1-e- 122 CITAB-2021) and the Portuguese Veterinary Directorate (approval no. 014139). During the experimental period, the animals were fed a standard diet (4RF21 GLP, Mucedola, Italy) ad libitum and drank tap water. The environmental temperature was maintained at 21–25°C, with a light-dark cycle of 12 hours of light and 12 hours of dark, and relative humidity of 40–60%. The experimental study involved six groups: group 1 (G1, WT, water, n = 5), group 2 (G2, HPV, water, n = 5), group 3 (G3, WT, 0.013 g/mL, n = 5), group 4 (G4, HPV, 0.006 g/mL, n = 5), group 5 (G5, HPV16, 0.008 g/mL, n = 5) and group 6 (G6, HPV16, 0.013 g/mL, n = 5). Following the guidelines of Russell and Burch regarding the use of animals for experimental purposes, particularly the 3Rs (reduce, refinement, and replacement), and aiming to minimize the number of animals used in this experiment without compromising its scientific quality, we administered the highest concentration to healthy animals [19]. We assumed that if toxicity were induced by A. citrodora, it would likely occur at the highest concentration. The humane endpoints were evaluated on a weekly basis, including body mass, hair appearance, eyes, ears and whiskers, mental status, and presence of papilloma, among other factors, as described by Silva-Reis et al., in 2021 [20]. Animals that reached a score of 4 or higher were euthanized before completing the experimental assay [21]. Throughout the experiment, the animals’ weight, water intake and food consumption were recorded on a weekly basis. At the end of the experimental trial, the animals were sacrificed using a 10:1 ratio of ketamine (100 mg/kg) and xylazine (10 mg/kg), followed by exsanguination via cardiac puncture, in accordance with the Federation of European Laboratory Animal Science Association guidelines [22]. Blood was collected for biochemical and hematological analysis, as well as for the comet assay. The heart, lungs, liver, spleen, and kidneys were collected for histological analysis and to assess oxidative stress.

Blood was collected into lithium heparin tubes and centrifuged at 20,000 rpm for 15 min. The plasma supernatant was then stored at –80°C. For microhematocrit evaluation, blood was centrifuged in capillary tubes at 12,000 rpm for 5 min. Biochemical parameters were studied using previously frozen plasma to determine concentrations of albumin, total protein, creatinine, urea, and alanine aminotransferase in an autoanalyzer (PRESTIGE 24I, PZ CORMAY). To measure blood glucose levels, the animals fasted for 12 hours before a blood sample was taken and analyzed using a GlucoMen Areo 2K glucometer with Glucomen Areo Sensor brand strips.

The alkaline comet assay (pH > 13) was performed using previously established methods [23, 24]. Normal melting point agarose (1%) was used to cover the slides. Four slides were prepared for each animal, with two slides used for the assay with the enzyme and the other two for the assay without the enzyme. To create a cell suspension, 20 µL of blood was diluted in 200 µL of ice-cold phosphate-buffered saline (1×) in a 0.5 mL microtube. Sixty µL of cell suspension were mixed with 600 µL of low melting point agarose (1%) and placed on four pre-coated slides (two replicates per slide), with 70 µL gels. The slides were then immersed in a lysis solution and rinsed in a washing buffer three times. To measure oxidatively damaged deoxyribonucleic acid (DNA) precisely, namely 8-oxoguanines and other altered purines, two slides per animal were incubated with formamidopyrimidine DNA glycosylase (FPG) for 30 min. FPG is a DNA lesion-specific enzyme that converts oxidized purines into DNA single-strand breaks. The enzyme was donated by Professor Andrew Collins from the University of Oslo, Norway. To allow DNA unwinding, slides with and without FPG treatment were gently immersed in a freshly prepared alkaline electrophoresis solution. The cells underwent electrophoresis in the same solution for 20 min at 25 V and a current of 300 mA. Subsequently, the cells were immersed in phosphate-buffered saline, followed by distilled water, dehydrated, and air-dried. To enable visual scoring, DNA was stained with a 1 µg/mL solution of 4,6-diamidino-2-phenylindole (Sigma-Aldrich Chemical Company, Spain) and observed using an Olympus BX41 fluorescence microscope at 400×. The nuclei were visually classified into five classes ranging from 0 (no tail) to 4 (almost all DNA in the tail) [25]. One hundred nuclei were observed per replicate (200 per case) and the genetic damage index (GDI) was expressed on an arbitrary scale of 0 to 400 (arbitrary units, AU) using the following formula:

GDI = (n class 0 × 0) + (n class 1 × 1) + (n class 2 × 2) + (n class 3 × 3) + (n class 4 × 4)

The tissue samples were fixed in 10% neutral buffered formalin and processed for histological analysis. They were then stained with hematoxylin and eosin. All slides were examined under light microscopy using a Zeiss Axioplan 2 microscope. Image processing was performed using the LAS Advanced Analysis Software Bundle (No: 12730448). Lesions found in the ear pavilion and chest skin were classified as epidermal hyperplasia, dysplasia, papilloma, or carcinoma, as previously described [15]. The lung lesions were classified as mononuclear inflammation, atelectasis, emphysema, bleeding, airway desquamation, congestion, hyalinization, and hemorrhage. The liver lesions were classified as vacuolar degeneration, inflammation, and congestion, while the spleen showed hyperplasia and congestion [13, 15]. The heart lesions were classified as myofiber hyalinization, congestion, hemorrhage, and vascular ectasia. The kidney lesions were classified as inflammatory infiltrate, hydronephrosis, congestion, hyperemia, and vascular ectasia, as previously described [15, 26–29].

Liver and kidney samples were collected and stored at –80°C. The samples were homogenized using Tissuelyser II QIAGEN in a cold buffer solution (0.32 mM sucrose, 20 mM HEPES, 1 mM MgCl₂, and 0.5 mM phenylmethyl sulfonyl fluoride, pH 7.4). The homogenate was then centrifuged at 15,000 × g for 20 min at 4°C (Prism R, Labnet International, USA), and the resulting supernatants were collected for further analysis. Superoxide dismutase (Cu/Zn-SOD) activity was measured by reducing nitroblue tetrazolium generated by the xanthine/xanthine oxidase system at 560 nm [30]. A standard curve was created using SOD (Sigma, S7446) from bovine erythrocytes (0–600 U·mL^-1). Catalase (CAT) activity was determined at 240 nm using a previously published method [31] and calculated using bovine CAT (Sigma, C40) as a standard (0–696 U·mL^-1). Glutathione peroxidase (GPx) activity was determined at 340 nm [32] using an extinction coefficient of 6.22 mM^-1·cm^-1. Glutathione reductase (GR) activity was determined by measuring the increase in absorbance of NADPH at 340 nm [33] using a molar extinction coefficient of 6.22 mM^-1·cm^-1. Glutathione S-Transferase (GST) activity was determined by measuring the increase in absorbance at 340 nm resulting from the conjugation of the thiol group of glutathione to the 1-Chloro-2,4-dinitrobenzene (CDNB) substrate [34], using a molar extinction coefficient of 9.60 mM^-1·cm^-1. The levels of glutathione were determined by measuring its reduced (GSH) and oxidized states (GSSG) using the fluorochrome ortho-phthalaldehyde with excitation and emission wavelengths of 320 nm and 420 nm, respectively [35]. Concentrations were estimated based on standard curves of GSH (Sigma, G4251) and GSSG (Sigma, G4626) (0–1 mM). The oxidative stress index (OSI) was calculated as the ratio between GSH and GSSG. The synthesis of reactive oxygen species (ROS) was estimated using the fluorescent probe 2,7-dichlorofluorescein diacetate, according to a previously published method [36], with excitation and emission wavelengths of 485 nm and 530 nm, respectively. A standard curve was constructed using DCF (Sigma, 35848) (0–500 µM). The levels of malondialdehyde (MDA), an indicator of lipid peroxidation (LPO), were measured at 535 nm (MDA-thiobarbituric acid adducts) and 600 nm (nonspecific adducts) wavelengths using the thiobarbituric acid method described elsewhere [37]. MDA levels (Sigma, 31640) were estimated based on a standard curve (0–1,000 µM) of malonaldehyde bis(dimethyl acetal). Carbonyls (CO) were estimated at 450 nm using the methodology described by Mesquita et al. [38].

Statistical analysis was performed using IBM SPSS Statistics for Windows, version 26 (IBM Corp., Armonk, NY, USA) and GraphPad Prism version 9. The normal distribution was analyzed considering the Shapiro-Wilks test and the analysis of variances considering the Levene test. For results with normal distribution, we performed an ANOVA followed by the Bonferroni test. For results without normal distribution, we use non-parametric tests such as the Kruskal-Wallis test. We considered values of P < 0.05 to be statistically significant. The parameters that apply this rule are body weight, food and drink consumption, organ mass, humane endpoints, hematologic, and oxidative stress parameters. A Chi-square test was performed to analyze the association between histological lesions and groups.

Humane endpoints = (sum score per animal) / (number of animals per cage)

The phenolic profile of A. citrodora extract was analyzed and several compounds were identified and quantified (Table 1). Most of the identified molecules were previously described in the literature, as mentioned in Table 1. Verbascoside was the main compound present, followed by luteolin 7-O-diglucuronide.

No behavioral or phenotypic changes were observed throughout the trial. Additionally, no mortality occurred during the experimental period. Regarding humane endpoints, statistically significant differences were found between group 2 (HPV, water) and group 4 (HPV, 0.006 g/mL), as well as between group 4 (HPV, 0.006 g/mL) and group 5 (HPV, 0.008 g/mL) (P < 0.05) (Table 2). All animals gained weight during the experiment. No animal reached the threshold for euthanasia (score 4). The body weight increased throughout the experimental assay, but there were no statistically significant differences between groups. In terms of weight gain, there were also no statistically significant differences between groups (Table 2). Regarding to food and water consumption per animal, HPV16 animals exhibited higher consumption levels (Table 2). Finally, although there were no statistically significant differences between groups, the weight of all organs was higher in the HPV16 groups compared to the WT groups (Table 3).

No concentration-dependent changes were observed for any parameters (Table 4). However, statistically significant differences were found between groups in terms of glucose levels: G2 (HPV, water) showed a larger increase compared to G4 (HPV, 0.006 g/mL) and G6 (HPV, 0.013 g/mL). The same trend was observed for the parameter total protein, which showed a statistically significant difference between G3 (WT, 0.013 g/mL) and G6 (HPV, 0.013 g/mL), with G6 showing a higher value.

The comet assay did not reveal any significant genotoxic blood damage in association with the HPV16 transgenes. There was a tendency to increase the total GDI in HPV animals that ingested the extract when compared to their control. The same applies to oxidation damage revealed by GDI_Fpg (Figure 1), but these observations did not reach the level of statistical significance (P > 0.05).

In this model, A. citrodora decrease the incidence of dysplastic skin lesions caused by HPV16. This suggests that dietary supplementation at concentrations of 0.008 g/mL and 0.013 g/mL has chemopreventive effects. Furthermore, these results show that the tested concentrations of A. citrodora were safe and did not induce toxicity under the current experimental conditions. Moving forward, it would be crucial to test concentrations higher than 0.013 g/mL to ascertain if we can achieve more promising outcomes against HPV16 lesions in this strain.