Synergistic effects of carvacrol, α-terpinene, γ-terpinene, ρ-cymene and linalool against Gardnerella species

Bacterial vaginosis (BV) is the most common vaginal infection affecting women worldwide. This infection is characterized by the loss of the dominant Lactobacillus community in the vaginal microbiota and an increase of anaerobic bacteria, that leads to the formation of a polymicrobial biofilm, mostly composed of Gardnerella spp. Treatment of BV is normally performed using broad-spectrum antibiotics, such as metronidazole and clindamycin. However, the high levels of recurrence of infection after treatment cessation have led to a demand for new therapeutic alternatives. Thymbra capitata essential oils (EOs) are known to have a wide spectrum of biological properties, including antibacterial activity. Thus, herein, we characterized two EOs of T. capitata and tested their antimicrobial activity as well as some of their main components, aiming to assess possible synergistic effects. Our findings showed that carvacrol and ρ-cymene established a strong synergistic antimicrobial effect against planktonic cultures of Gardnerella spp. On biofilm, carvacrol and linalool at sub-MIC concentrations proved more efficient in eliminating biofilm cells, while showing no cytotoxicity observed in a reconstituted human vaginal epithelium. The antibiofilm potential of the EOs and compounds was highlighted by the fact cells were not able to recover culturability after exposure to fresh medium.


Minimal inhibitory concentration and minimal lethal concentration.
It has been shown that the biological effects of EOs are possibly due to the combined effects of their components, particularly the main ones 46 . In order to recognize the antimicrobial activity of some of these compounds, that are commercially available, the values of minimal inhibitory concentration (MIC) and minimal lethal concentration (MLC) were first determined. As shown in Fig. 1a, both the EOs had similar antimicrobial activity against the selected Gardnerella isolate. Of the tested individual components, carvacrol was by far the most potent, with a MIC value very similar to the whole EO. Despite having similar high concentrations in both EO, ρ-cymene and γ-terpinene had a fourfold difference in the MIC values, while linalool, which was present in a lower concentration, reported a similar antimicrobial activity as γ-terpinene. Interestingly, for all the active compounds, the values of MLC were almost coincident with the values of MIC.

Combined effects of individual compounds of EOs.
It has been previously shown that the combined effect of some individual compounds can enhance their antimicrobial activities 47,48 . Taking into consideration the MIC determinations, we assessed possible synergism between the individual components tested. As shown in Fig. 1b, different effects were observed wherein combinations of carvacrol and ρ-cymene resulted in the highest synergistic activity. No antagonistic effects were observed.

Activity of EOs and compounds on a Gardnerella biofilm.
Since BV is well-known to be associated with biofilm formation, mainly constituted by Gardnerella 11 , we further tested the EO and some of their components' anti-biofilm activity. As shown in Fig. 2a, α-and γ-terpinene had no effect in reducing the biomass of the biofilm, despite their antimicrobial activity assessed against planktonic cultures. Carvacrol, ρ-cymene, and linalool add a similar effect at either of the two EOs tested. Interestingly, several combinations of carvacrol and linalool, at sub-mic concentrations, revealed a significantly higher ability to reduce the biofilm biomass, as compared with the MLC concentrations of the EOs. The addition of ρ-cymene did not enhance the carvacrol and linalool synergistic effect. Interestingly, when assessing cell culturability, the total reduction of CFU's was observed for both EOs, carvacrol, linalool, or any mixture of these 2 components (Fig. 2b).
The effect of DMSO on Gardnerella sp. UM241 cells culturability was also evaluated and the results are described in Supplementary Fig. 1. DMSO did not show any negative effect on cells viability at the different concentrations tested.
Other four Gardnerella species were used to assess the activity of all components on biofilms. We have chosen the conditions that were able to eliminate the growth of Gardnerella sp. UM241, namely the EO sample from Carvoeiro, the combination between carvacrol and linalool with the lowest amount of linalool, and a combination of 3 selected compounds (at their lowest concentrations tested). The results are shown in Fig. 3. The capacity of the compounds to reduce biofilm biomass (Fig. 3a) was similar to that observed when tested in Gardnerella sp. Effect of EOs and compounds on Gardnerella biofilm structure and viability. Since we observed a total reduction of cultivable bacteria without completely removing the biofilm structure, we decided to analyze the biofilm structure after the antimicrobial challenge, by using LIVE/DEAD staining and CSLM observation. As shown in Fig. 4, the big majority of biofilm cells treated with the EO or the two most promising combinations of components, presented a damaged cell wall, similar to the dead cells from biofilm control. Nonetheless, some cells still keep an intact cell wall, but they cannot be recovered (Fig. 3b), which suggests the presence of viable but not cultivable cells (VBNC) 49 . To confirm the CLSM observation and to quantitatively determine the % of viable or dead cells in biofilms formed by the 4 species of Gardnerella, after antimicrobial challenge, we performed a fluorimeter assay, based on LIVE/DEAD staining. As shown in Fig. 5

Colonization of reconstructed human vaginal epithelium.
To assess the antimicrobial potential of the EO or carvacrol + linalool, G. vaginalis early-stage biofilms were grown on a human vaginal epithelium tissue, using a chemically defined medium that simulates genital tract secretions. The biofilms were then challenged at the EO MLC concentration or the compounds at the concentrations where synergism was observed.
As represented in Fig. 7a, G. vaginalis not exposed to the antimicrobial agents was able to form a biofilm on the vaginal epithelium, and the biofilm was drastically reduced after the antimicrobial challenge. Like the previous experiments, no culturable cells were recovered after the antimicrobial challenge (Fig. 7b). Interestingly, at this concentration, neither the EO nor the carvacrol + linalool combinations demonstrated significant cytotoxicity (Fig. 7c).

Discussion
The increasing levels of bacterial resistance associated with commonly used antimicrobial agents are one of the problems that most affect public health 50 . In BV, increased resistance to metronidazole and clindamycin have been reported [51][52][53] , and this led to seek for new alternatives of treatment 54 . As far as we are aware, the use of EOs as alternatives for the treatment of vaginal infections is still limited, however, some EOs have been tested on Candida species associated with vaginal candidiasis 55,56 , in BV 45,57,58 , trichomoniasis 59 or other vaginal infectionrelated pathogens 60 . One possible criticism of the utilization of EO as a reliable therapeutically option is the variability that is often observed between different batches of the same EO, which can affect its properties and biological activities 61 . Thus, it is extremely important to characterize the essential oils to ensure their chemical homogeneity. According to our results, the EOs analyzed were very homogeneous, characterized by high contents in carvacrol. Interestingly, the antimicrobial potential of the two EOs was also very similar. Furthermore, assessing the antimicrobial activity of single compounds or mixtures of compounds present in the EOs, has the potential to develop a more standardized product, despite having increased costs of production. In this study, we decided to study single compounds and combinations of compounds from T. capitata EO, in order to find a composition that has good antimicrobial activity against biofilms formed by Gardnerella spp. Due to the wide-ranging biological properties attributed to T. capitata EO, several studies have been conducted to assess the antibacterial activity of T. capitata EO and carvacrol, and to less extent the activity of their compounds 62 . As a result, it has been shown that the antimicrobial activity of EOs is dependent on their chemical compositions 63,64 . The high content of carvacrol, a phenolic monoterpene, in T. capitata EO is pointed to be largely responsible for the antimicrobial activity 65 . Not surprisingly, herein, carvacrol also presented the highest antimicrobial activity of the five compounds tested. It has also been described that monoterpenes have low antimicrobial activity and are usually ineffective as antimicrobial agents when used alone 66,67 , as also observed in this study. Nevertheless, components present in small amounts in EOs can still play an important role in antimicrobial activity due to the possible synergistic effects 68,69 . It has been previously shown that interactions between major and minor constituents can have a critical influence on global antimicrobial activity 27 . The most common interactions between different components include synergistic, antagonistic, additive, and indifferent effects 70 . Noticeably, the strongest effect happened in the combination of carvacrol and ρ-cymene resulting in synergism. This effect is expected to result from the interaction of ρ-cymene with the lipidic membrane of the cells causing the expansion of the membrane, which facilitates and increases the transportation of carvacrol into the cells 66,71 . Other studies already reported this synergistic effect between carvacrol and ρ-cymene, particularly against Listeria monocytogenes 72 and Escherichia coli 73 . Statistical analysis was performed using one-way ANOVA (n ≥ 3). In panel (a), differences are showed, when p < 0.05, by asterisk symbol when comparing with 5 µL/mL α-terp; hash symbol when comparing with 5 µL/ mL γ-terp; bullet symbol when comparing with 0.02 µL/mL carv + 2.5 µL/mL ρ-cym; filled square symbol when comparing with 0.02 µL/mL carv + 1.25 µL/mL linl and filled diamond symbol when comparing with 0.02 µL/ mL carv + 2.5 µL/mL ρ-cym + 0.32 µL/mL linl. All conditions are statistically different from the 48 h-biofilm control except for the cases marked (a). In panel (b), differences are represented by asterisk symbol when comparing with control 48 h and hash symbol when comparing with control 72 h, when p < 0.05. www.nature.com/scientificreports/ Interestingly, when testing the antimicrobial activity against biofilms, α-and γ-terpinene had no detectable effect, and while ρ-cymene was able to reduce ~ 30% of the total biomass, its effect on bacterial culturability was neglectable. Furthermore, despite carvacrol + ρ-cymene having a better synergistic effect against planktonic cultures, on biofilms, the most active combination was carvacrol + linalool. This last combination was more effective in reducing the total biomass of the biofilm (~ 50%) than the whole EO (~ 30%). This further demonstrates that biofilm cultures susceptibility to antimicrobials cannot be predicted from the standardized MIC determinations 74,75 . Despite not being able to completely eradicate the biofilm biomass, cell culturability was significantly affected, with no colony detected in many of the tested conditions. A similar effect was also reported with T. capitata EO against a biofilm of Candida glabrata where despite the low reduction of biofilm biomass, the cells were not metabolic active, remaining inactive or dead 41 . In another study with Staphylococcus aureus and E. coli biofilms, five different EOs also promoted the reduction of cells metabolic activity inpre-established biofilms and even the complete elimination of viability in some of the cases 76 .
To further understand the impact of the EO and some of the most promising mixtures, we also performed a resuscitation attempt 77 , whereby stimulating the cells without the antimicrobial stress, they can recover the full active metabolic state and culturability. However, in our tested conditions, no cell recovery was observed. This Statistical differences were found for all the conditions when compared with the 72 h-biofilm control, except for G. leopoldii where no differences were found (a) (t-test, n = 4). Statistical differences were found in G. leopoldii, between conditions A and C (asterisk symbol) (t-test, n = 4). (b) Effect on biofilm cells culturability. Results are represented as mean of Log (CFU.mL −1 ) with a limit of detection of Log = 3. Error bars represent s.d. Statistical analysis was performed using t-test (n = 4). Differences are represented when p < 0.05, when compared with control 72 h of G. vaginalis UM137 (asterisk symbol), G. piotii UM035 (hash symbol), G. leopoldii UGent 09.48 (bullet symbol) and G. swidsinskii GS 9838-1 (filled square symbol). www.nature.com/scientificreports/ observation suggests that these antimicrobial agents have a lasting effect on Gardnerella, and this might improve the recurrence rates currently observed after treatment with metronidazole or clindamycin 78,79 .
Herein, we also determined the feasibility of using reconstituted vaginal epithelial tissue, to establish a Gardnerella spp. biofilm. While this vaginal model has been used before to study Candida infection 80 or vaginal dryness 81 , as far as we are aware, this is the first study that reports Gardnerella spp. colonization in this model. By using a PNA-FISH specific probe, we were able to observe intact biofilms formed on the vaginal epithelium. Interestingly, after the antimicrobial challenge, most of the bacterial cells were killed, and the small clusters of cells observed were non-cultivable. Importantly, at the tested concentration used herein, we did not observe any significant cytotoxicity in the vaginal tissue.
In summary, this study revealed the potential of T. capitata EO and some of its components to be used as potential new therapeutic agents in the treatment of BV. Nonetheless, this study has some limitations and further work should be performed to test these components on multi-species biofilms, that better mimic what occurs in BV 82 . Furthermore, it would be important to assess their activity against Lactobacillus species common colonizers of the healthy vaginal microbiota. In addition, similar to what is described in other EOs 83 , it would be also important to analyze the effect of the tested components with conventional antibiotics. Apart from it, there are some limitations in the methods, namely in the biofilm biomass quantification by CV, since it is known that one of the disadvantages of this method is the lack of reproducibility 84 .  The oils components were identified by considering, concurrently: (i) the acquired retention indices on two columns with different phases SPB-1 (polydimethylsiloxane) and SupelcoWax-10 (polyethyleneglycol) determined by linear interpolation relative to the retention times of C8-C23 of n-alkanes and compared with reference data from authentic products (available in the laboratory database of the Faculty of Pharmacy, University of Coimbra) and literature data 90 ; (ii) the acquired mass spectra compared with reference data from the laboratory database, the Wiley/NIST library 91 and literature 92 . The relative amount of each component was estimated from GC peaks areas without any correction regarding FID responses.

Materials and methods
Bacterial growth conditions. Various strains from the genus Gardnerella were analyzed in this study.

Determination of the minimal inhibitory concentration and minimal lethal concentration. The
MIC and MLC of both EOs, as well as five individual components of the oil, were evaluated by the macrodilution method, using the isolate Gardnerella sp. UM241, as described previously 45 with some modifications. Briefly, each component was dissolved in the same proportion (1:1) in dimethyl-sulfoxide (DMSO, Scharlau, Barcelona, Spain) to improve its solubility. Therefore, serial dilutions in glass test tubes were prepared in sBHI to obtain a range of concentrations between 0.04 and 40 µL/mL in the experiments with the compounds. Regarding the experiments with the EOs, a range of concentrations between 0.02 and 2.5 µL/mL was tested. Afterwards, 500 µL of the bacterial inoculum, adjusted to the concentration of 10 6 colony forming units (CFU)/mL, was added to each tube performing a total volume of 1 mL. Positive control with DMSO at the highest concentration used in the dilutions was included. A second positive control was used with bacterial suspension without any compounds or EOs. The negative control only included sBHI medium. After 48 h of incubation at 37 °C and 10% CO 2 , the MIC value was determined by reading the optical density (OD) of all the dilutions at 620 nm. MLC was determined by plating 10 µL of each dilution on CBA plates and defined as the lowest concentration where no growth was detected. Each experiment was repeated at least three independent times.

Determination of the fractional inhibitory concentration by the checkerboard method.
The checkerboard method was performed to study the interactions and the resultant effect from the combination of pairs of compounds from T. capitata against the isolate Gardnerella sp. UM241. Serial dilutions of each compound were prepared as described above in a range of concentrations appropriated to include the value of MIC for each compound. The experiments were performed on glass tubes for a final volume of 1 mL. For one of the compounds, serial dilutions were added along the lines and for the other compound were added along with the columns. Thereafter, 500 µL of a bacterial inoculum, prepared as described above and adjusted to the concentration of 10 6 CFU/mL, was added to all the tubes. Then, the tubes were incubated for 48 h at 37 °C and 10% CO 2 . To analyze the effect of the interactions between the compounds, the OD of the content of each tube was measured at 620 nm, and the value of the fractional inhibitory concentration (FIC) index was determined according to the lowest concentrations in which bacterial growth was not observed. The FIC was calculated as previously described 95  and FICB = MICB(combination) MICB(alone) 96 . The effect was considered synergistic when FIC was ≤ 0.5, partial synergistic when FIC was > 0.5 and < 1, additive when FIC = 1, indifferent when FIC was > 1 and ≤ 4, and antagonistic when FIC was > 4 97 . In each assay, the lowest value of FIC was selected, and the FIC of each combination was presented by the range of values from all assays. For each combination, experiments were performed at least two independent times.  www.nature.com/scientificreports/ (Thermo Fisher Scientific, Lenexa, KS, USA) was dispensed into each well and the OD was measured at 595 nm using a microplate reader. The reduction in biofilm biomass was determined by comparing the values of OD of each condition with the OD values of the positive control without DMSO.

Biofilm cells culturability analysis by colony-forming units.
Biofilm formation was performed as described above, with the difference that 24-well flat-bottom tissue culture plates (Orange Scientific) were used, in which a coverslip (Thermo Scientific™ Nunc™ Thermanox™ Coverslips) was distributed in each one of the wells. After that, 1 mL of 24 h-bacterial inoculum, adjusted to the concentration of 10 8 CFU/mL, was dispensed in each well and the plate was incubated at 37 °C and 10% CO 2 with replacement by the fresh medium at 24 h. After 48 h, the growth medium was removed and the coverslips were transferred to a new 24-well plate. Thereafter, 300 µL of the desirable compounds in sBHI were dispensed into the wells and the plate was incubated for more 24 h at 37 °C and 10% CO 2 . After that time, the medium was removed, and each coverslip was washed with NaCl 0.9% (w/v). 300 µL of fresh sBHI was added to the wells and the biofilm was mechanically detached from the coverslips. Then, 100 µL of each well content was removed and serial dilutions were performed in NaCl 0.9% (w/v) and 20 µL of each dilution was plated in a CBA plate and incubated for 48 h at 37 °C and 10% CO 2 . The experiments were repeated at least three independent times.
Biofilms assessed by LIVE/DEAD staining combined with confocal laser scanning microscopy. In order to evaluate the structure of the biofilm after the effect of T. capitata EOs and compounds, the biofilm was analyzed by confocal laser scanning microscopy (CLSM), as previously described 49 . Briefly, the biofilm of 4 Gardnerella species was performed in 24-well plates with coverslips, as described above and the conditions EO Car. at 0.16 µL/mL, and the combinations of 0.08 µL/mL carvacrol + 0.32 µL/mL linalool and 0.04 µL/ mL carvacrol + 2.5 µL/mL ρ-cymene + 1.25 µL/mL linalool were tested. After 72 h, the biofilm was washed with NaCl 0.9% (w/v) and then stained using a LIVE/DEAD™ BacLight™ Bacterial Viability Kit (Invitrogen, Thermo Fisher Scientific), consisting of SYTO 9 and propidium iodide (PI). After incubation, the biofilms coating the coverslips were gently washed with 1 × PBS and then, the coverslips were removed from the wells and placed on microscope glass slides (VWR, Lisbon, Portugal). Two types of controls represented by untreated and dead biofilm cells were considered for this experiment and were used to define the CLSM laser intensity threshold. The dead control was obtained by covering the coverslips with the biofilms with 200 µL of 100% (v/v) methanol (Thermo Fisher Scientific) for 30 min. Then, all the coverslips with the live, dead, and treated biofilms were covered with 100 µL of the LIVE/ DEAD staining mix, with SYTO 9 and PI used each in a concentration of 3 µL/mL. Subsequently, the coverslips were incubated for 15 min. in the dark at room temperature. Biofilm image stacks were acquired with an Olympus™ FluoView FV1000 (Olympus, Lisbon, Portugal) confocal laser scanning microscope, using a 10 × objective. SYTO9 was detected using a filter with an excitation wavelength of 485 nm and an emission filter of 498 nm. PI was detected using a filter with an excitation wavelength of 536 nm and an emission filter of 617 nm. The CLSM images were analyzed using the FV10-ASW 4.0 Viewer Software (Olympus). This experiment was performed twice with two technical duplicates.

LIVE/DEAD quantification of biofilm cells by fluorometry. To analyze the biofilm cells viability after
the action of the compounds and EO, a quantitative LIVE/DEAD analysis was performed of the 4 Gardnerella species biofilm challenged with EO Car. at 0.16 µL/mL, and the combinations of 0.08 µL/mL carvacrol + 0.32 µL/ mL linalool and 0.04 µL/mL carvacrol + 2.5 µL/mL ρ-cymene + 1.25 µL/mL linalool. After 72 h of biofilm formation as described above, the medium was removed, and the biofilm was washed with NaCl 0.9% (w/v) and then detached from the 24 well-plate using NaCl 0.85% (w/v) and centrifuged for 5 min at maximum velocity. Subsequently, the concentration of biofilm suspension was adjusted to 10 7 CFU/mL and the LIVE/DEAD staining was added to biofilms as described in the CLSM section. Afterwards, the absorbance was measured using a multilabel microplate reader (Cytation 3, Bio Tek, Maine, USA) using the wavelengths of 485/530 nm and 485/630 nm (excitation/emission wavelength). Of note that before these experiments, a calibration curve was performed for each Gardnerella spp. according to the manufacturer's instructions in each independent assay. In brief, from a fresh bacterial suspension collected from a CBA plate, suspensions with live and dead cells were prepared, as well as artificial dilutions of live/dead cells (33% live + 67% dead cells; 50% live + 50% dead cells; 67% live + 33% dead cells). A linear equation was obtained from the relation between the percentage of live/dead bacteria from the suspension and the green/red fluorescence ratio. Finally, for each biofilm assay, the obtained LIVE/DEAD ratios were then corrected according to the equation obtained from the calibration of the respective experiment. The percentage of reduction in cell integrity was obtained by comparison to the live control. Three independent assays were performed with two technical replicates.

Biofilm cells recovery after removing the effect of EO and compounds from T. capitata.
In an attempt to verify if biofilm cells can recover culturability after the activity of EOs and compounds, an experiment was performed as described for quantification of biofilm cells viability, in 24-well plates, without coverslips. After 48 h of biofilm formation, the medium was removed, and 1 mL of compounds was added. After 24 h of action of compounds and EOs on the biofilm, the medium from the biofilm was removed and the biofilm was washed with 1 mL of NaCl 0.9% (w/v). Then, 1 mL of fresh sBHI or sBHI supplemented with 0.25% (w/v) of maltose (Fisher Bioreagents, Fair Lawn, New Jersey, USA) was added to each well, and the plate was incubated for more 24 100 . After that, the OD was adjusted to a concentration of 10 7 CFU/mL and 1 mL was dispensed on the tissues for colonization of the vaginal epithelium. The tissues were then incubated at 37˚C and 10% CO 2 , for 9 h. Then, mGTS with either EO Car. at a concentration of 0.16 μL/mL or a combination of 0.08 μL/mL carvacrol + 0.32 μL/mL linalool was added to each tissue, after removal of the spent media, and incubated for a further 14 h. For microscopic analysis, the tissues were placed in 4% paraformaldehyde (Thermo Fisher Scientific) and then embedded in paraffin (Leica TP1020, Leica Biosystems, Nussloch, Germany). Paraffin tissue blocks were prepared (Leica EG 1140 H, Leica Biosystems, Nussloch, Germany) and 3-μm-thick sections were obtained using a microtome (Microm HM 325, Thermo Fisher Scientific, Walldorf, Germany). For the deparaffinization step, sections were placed in xylene (Thermo Fisher Scientific) twice for 5 min, followed by a hydration step with 100% and 50% of ethanol (Thermo Fisher Scientific) for 5 min each and a final step in distilled water for 5 min. The samples were allowed to air-dry and the PNA-FISH procedure was performed using a previously developed PNA probe for Gardnerella 101 , with a hybridization step at 60 °C for 90 min. The samples were analyzed using an Olympus BX51 epifluorescence microscope (Olympus, Lisbon, Portugal) equipped with a TRITC filter (BP 530-550, FT 570, LP 591 sensitive to the Alexa Fluor 594 molecule). To perform the culturability assays the tissues were washed once with NaCl 0.9% (w/v) and 500 µL of mGTS was added. A cycle of sonication was performed to displace the cells from the tissues, for 10 s with an amplitude of 33%. Serial dilutions in NaCl 0.9% (w/v) were then performed in duplicates from the content of each well, and each dilution was plated on CBA plates and incubated at 37ºC and 10% CO 2 , for 72 h. The experiment was performed twice, with technical duplicates.

Vaginal irritation test.
A toxicity study was performed using the Reconstructed Human Vaginal Epithelium (HVE-SkinEthic®, Episkin, France) model. This reconstructed epithelium reveals a strong histologic resemblance with human vaginal tissue. Upon arrival, the tissues were incubated overnight, at 37 °C, ≥ 90% humidity, 5% CO 2 (Binder APT.lineTM C150E2 Incubator, Binder, NY, USA), in 6-well plates (VWR) using 1 mL of Maintenance Medium (provided by Episkin together with the tissues). Tissues were then transferred to new 24-well plates (one plate per condition) containing 300 μL of Maintenance Medium, and then 30 μL of the test substances or controls were gently dispersed over the entire tissue surface. EO Car. at 0.32 μL/mL and 0.08 μL/mL carvacrol + 0.32 μL/mL linalool tested concentrations were prepared in sesame oil (Ph Eur. Grade, Sigma), since this is the oily vehicle recommended for skin irritation testing according to ISO 10993-23:2021. Phosphate Buffer Saline without Ca 2+ and Mg 2+ (DPBS, VWR) and Sodium Lauryl Sulfate 1% w/v, were used as negative and positive controls, respectively. Solvent control (sesame oil) was also included. A different plate was used for each study substance (n = 3 tissues per substance). After 24 h of incubation at 37 °C, ≥ 90% humidity, 5% CO 2 , tissue integrity was assessed by the MTT assay, as previously described 102 . Briefly, the tissues were washed with PBS and gently dried. Then, the tissues were transferred to a 24-well plate containing 300 µL per well of a 0.5 mg/mL MTT (Alfa Aeser) solution (in PBS, VWR) and incubated for 3 h, at 37 °C, ≥ 90% humidity, 5% CO 2 , protected from light. After this period, the tissues were transferred to a single 24-well plate containing 750 µL of isopropyl alcohol and more 750 µL were further added at the top of each tissue to allow for the extraction of formazan for ≥ 2 h, in a sealed plastic bag, under agitation in a plate stirrer. Absorbance was then measured at 570 nm without a reference filter, according to the tissues provider instructions (Promega GloMax® Explorer System, USA). The background was deducted from all measured absorbance using isopropyl alcohol in free wells of the same reading plate. The absorbance of the negative control was considered 100% viability reference for products toxicity calculation.
Statistical analysis. Data were analyzed with a one-way analysis of variance (ANOVA) and t-test with an alpha level of 0.05, using GraphPad Prism version 8 (GraphPad Software, California, USA). Results are presented as the mean + standard deviation (s.d.) or mean + standard error of the mean (s.e.m.). Statistical differences were considered significant when p-values were less than 0.05.

Data availability
All data generated or analyzed during this study are included in this published article (and its Supplementary Information files).