Abstract Retinol is widely used to treat skin ageing because of its effect on cell differentiation, proliferation and apoptosis. However, its potential benefits appear to be limited by its skin permeability. Herein, we investigated the transcutaneous behavior of retinol in semisolid cosmetics, in both in vitro and in vivo experiments. In vitro experiments used the modified Franz diffusion cell combined with Raman spectroscopy. In in vivo experiments, the content of retinol in rat skin and plasma was detected with HPLC. Retinol in semisolid cosmetics was mainly concentrated in the stratum corneum in the skin of the three animal models tested, and in any case did not cross the skin barrier after a 24 h dermatologic topical treatment in Franz diffusion cells tests. Similar results were obtained in live mice and rats, where retinol did not cross the skin barrier and did not enter the blood circulation. Raman spectroscopy was used to test the penetration depth of retinol in skin, which reached 16 μm out of 34 μm in pig skin, whereas the skin of mouse and rat showed too strong bakground interference. To explore epidermal transport mechanism and intradermal residence, skin transcriptomics was performed in rats, which identified 126 genes upregulated related to retinol transport and metabolism, relevant to the search terms “retinoid metabolic process” and “transporter activity”. The identity of these upregulated genes suggests that the mechanism of retinol action is linked to epidermis, skin, tissue and epithelium development. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-024-73240-y. Keywords: Retinol, Semisolid preparation, Transcutaneous permeation, Skin transcriptome, Mechanism Subject terms: Biological techniques, Computational biology and bioinformatics Introduction Due to the continuing development of society and an ageing population, the “silver hair economy” is experiencing explosive growth due to the rapid increase in the demand for skin rejuvenation and health-related treatments^[42]1. It is expected that the annual global expenditure in this area will increase from $24.6 billion to approximately $44.5 billion by 2030^[43]2. Skin ageing is having both endogenous and exogenous factors. Endogenous ageing mainly manifests as gradual slowing of cellular metabolism, gradual loss of collagen and destruction of collagen, among others. Exogenous ageing is mainly caused by factors such as environmental pollution or ultraviolet radiation. The latter results in thinning of the skin, loss of collagen and fracture of elastic fibres, thus resulting in a dry, rough, wrinkly appearance^[44]3–[45]5. The vitamin A family of compounds is widely used in cosmetics and medicines. The most popular compounds are retinoic acid, retinaldehyde, retinol and retinyl esters^[46]6. In cosmetics, vitamin A acid is used in anti-ageing and acne removal, as it can stimulate cell renewal and collagen production, thus improving skin texture and reducing wrinkles and acne^[47]7. Clinically, retinoic acid is only used in pharmaceuticals and prescription creams to treat acne, keratosis pilaris and skin cancer, but its side effects include skin irritation and damage to the liver^[48]8–[49]10. In contrast, its derivative retinol is milder and more often found in cosmetics^[50]11 where it may promote cell regeneration and production of collagen and elastin, protecting against UV damage due to its antioxidant properties^[51]12. Despite its powerful anti-ageing properties, retinol is unstable, photosensitive, prone to oxidation, and irritating, making the development of safe and effective retinol formulations an urgent needed^[52]13. In 2022, the European Union’s Scientific Committee on Consumer Safety (SCCS) revised the allowed concentration of vitamin A-derived compounds retinol, retinyl acetate and retinyl palmitate as cosmetic ingredients, with maximal concentration of 0.05% (retinol equivalents) in body lotions, and 0.3% (retinol equivalents) in hand and face creams, as well as in other leave-on or rinse-off products (e.g. sunscreens, anti-wrinkle creams and eye cream). Due these low dose limits, transdermal penetration studies of topical retinol formulations are essential to evaluate safety and efficacy^[53]14. The in vitro permeation test (IVPT) is widely used in the development of dermal delivery formulations and bioequivalence (BE) evaluation^[54]15. IVPT is performed with isolated human or animal skin to simulate the transdermal transport of ingredients in topical formulations at near-physiological conditions. The rate and extent of penetration of drugs through skin is used to assess the bioavailability (BA) of the drug for comparisons of self-developed topical formulations versus control formulations^[55]16,[56]17. For example, Yourick JJ studied the transdermal absorption of retinol in cosmetic formulations by applying 0.3% retinol formulations (hydroalcoholic gels and oil-in-water emulsions) to the skin of rats; retinol remained in the skin of the rats at 24 h, and its concentration decreased over 72 h. In total, 23% of the applied dose was absorbed into the skin after 24 h, and 18% was absorbed after 72 h in rats, which allowed for the formation of reservoirs^[57]18. But in cosmetics, transcutaneous permeation other than percutaneous absorption is more important, because people wish more functional components such as retinol would stay in skin instead of entering into the blood gradually. Franz diffusion studies are inexpensive and can produce fast results, but differences between animal and human skin may yield different results to a large extent; furthermore, the method does not consider skin processes in vivo^[58]19. Therefore, to elucidate the absorption process of the active substance in the skin, it is necessary to carry out in vivo animal experiments, but these experiments also convey the obstacles of high experimental costs and individual differences. In general, all technologies may have these issues, but the combined use of multiple methods with complementary advantages may mask individual limitations^[59]19. In addition to in vitro permeation experiments, attempts have been made to characterise the absorption of transdermal drugs using in vitro (diffusion cells), ex vivo (band stripping / confocal laser scanning microscopy) and in vivo (dermal microdialysis / open-flow microperfusion / assessment of plasma drug concentration and pharmacodynamics) methods^[60]20,[61]21. Since many reports have studied the underlying transport mechanism of retinol, but its mode of transport in cosmetics is not clear for the dose of retinol in cosmetics is too lower. In this study, we investigated the transdermal permeation and transport behaviour of retinol in semisolid cosmetics with 3 widely used animal skins, then preliminarily contrasted correlativity between in vivo and vitro. In addition, the transport mechanisms of retinol were evaluated using high-throughput sequencing methods. These results promote the skin quantification and practical application of Retinol Semisolid Preparations in cosmetics. Materials and methods Chemicals and reagents Bama Xiang pig skin was purchased from Linxi County Jingde Agricultural Products Sales Co. Ltd. (Hebei, China). Dibutylhydroxytoluene (BHT) was purchased from Shanghai McLean Biochemical Technology Co. Ltd. (Shanghai, China). Tween 80, methanol (AR) and hexane (AR) were purchased from Tianjin Chemo Chemical Reagent Co. Ltd. (Tianjin, China). Sodium chloride was purchased from Guangzhou Chemical Preparation Factory (Guangzhou, China). Retinoic acid was purchased from Shanghai Yuanye Biotechnology Co. Ltd. (Shanghai, China). Retinol was purchased from Shanghai Anpu Test Technology Co. Ltd. (Shanghai, China). 3 M medical breathable tape was purchased from Minnesota Mining and Manufacturing Company (Minnesota, USA). Retinol external semi-solid preparations were prepared by Jinan University - Hujia Technology Joint Laboratory (Guangzhou, China), and the formulation and physicochemical properties are shown in the Supplementary File. Animals Male (6 to 7-week-old) Sprague-Dawley (SD) rats (No. 44007200119123) and male (4 to 5-week-old) Kunming (KM) mice (No. 44007200121077) were purchased from the Guangdong Provincial Animal Center. Rats and mice were housed in a temperature- and humidity-controlled room (22 ± 1 °C, 55 ± 1%, respectively) with a 12 h dark/light cycle and free access to water and a commercial diet. The animals were allowed to adapt to the environment for one week before the experiment began. The experimental protocols used in this study were approved by the Institutional Animal Care and Use Committee of Jinan University (ethical review no. 20230302-18). All experiments were conducted according to the Guidelines for Animal Care and Use in China, and the Animal Ethics Committee of the Chinese Academy of Medical Sciences approved the experiments. All animal experiments complied with the ARRIVE guidelines for reporting animal experiments. In vitro skin permeation studies A Franz diffusion cell (effective area of 1.32 cm^2) was used to evaluate in vitro retinol permeation. The receptor chamber was filled with 15 mL of 0.5% (v/v) Tween 80 − 0.9% NaCl (pH = 7.4)^[62]22 to ensure that the sink conditions were met. Skin (pig, mouse or rat skin) was mounted between the donor and receptor compartments and approximately 0.1 g/cm^2 of ROLs were placed in the supply chamber. In the receptor chamber, rotation speed was 350 rpm and water bath temperature was 32 ± 0.5 °C. Receptor fluid (2 mL) was withdrawn from the receptor chamber after 1, 2, 4, 8, 12 and 24 h, and retinol concentration in the solution was analysed with HPLC. At the end of the 24 h experiment, residues in the supply chamber were scraped off, and the skin was rinsed three times with fresh 0.5% (v/v) Tween 80-saline. The stratum corneum of the skin was collected using the tape stripping method, by stripping the tape 10 times, and the skins were stored at -20 ℃. To quantitatively analyse the amount of retinol in skin, separate experiments were performed; that is, at 1, 2, and 4 h, the skin was removed from the Franz diffusion cell for HPLC-based quantitative analysis of the concentration of retinol. Confocal Raman spectroscopy Raman spectra were collected using a laser microconfocal Raman spectrometer (LabRAM HR Evolution, HORIBA Jobin Yvon S.A.S., France) (l[ex] = 633 nm, 17 mW). Pig skins were treated for 24 h in the Franz diffusion cells according to Sect. [63]2.4, after which they were placed on a detection platform for spectral collection. The fingerprint (400–2000 cm^−1) region was analyzed at different depths (0 to 34 μm), step size of 2 μm, and integration time of 15 s for each depth). Spectra processing including baseline correction and smoothing was performed with Origin 2021 software. In vivo skin permeation study Before the experiment, the dorsal region of each KM mouse was shaved using a hair trimmer and depilatory cream. The next day, mice were treated with ROLs applied to the back skin (at a dose of approximately 0.1 g /cm^2). After the treatment for 0, 1, 2, 4–24 h, blood samples were collected retroorbitally and centrifuged at 3000 rpm for 10 min to isolate the plasma, which was stored at -80 °C. The mice were then euthanised by intraperitoneal injection of pentobarbital (180 mg/kg). Dorsal skin samples were obtained using surgical scissors and stored at -20 °C. For rats, before the experiment, the back area of each SD rat was shaved using a hair trimmer and a hair removal cream. On the next day, ROLs were applied to the back skin of rats (the dose was about 0.1 g / cm^2, and the administration area was about (3 × 3) cm^2). Blood samples were collected at 0, 1, 2, 4, and 24 h after administration. The blood was centrifuged at 3000 rpm for 10 min, and the plasma was separated and stored at-80 ℃for later use. HPLC quantification of retinol HPLC analyses were performed using a Shimadzu LC-2040 C Performance HPLC chromatograph (Shimadzu, Kyoto, Japan) equipped with a PDA detector. Retinol separation was achieved using an inverse C18 column (Jupiter 5 μm C18 300 A, 250 mm × 4.6 mm) at a column temperature of 30 °C. The elution solvent was methanol/water (v/v = 85:15) and flow rate was 1.0 mL/min. The injection volume was 10 µL, and the detection wavelength for PDA detection was 325 nm. The methods used for validation are shown in the Appendix. Sample processing and detection Detection of retinol in skin and tape samples Skin samples of pigs, mice and rats (area of 1 to 1.32 cm^2) were minced. The skin and tape (2 layers of tape) samples were added to 2 mL of methanol (0.1% BHT), shaken overnight at 4 °C, sonicated for 20 min in an ice bath, and centrifuged, followed by filtration through a 0.45 μm filter membrane for HPLC analysis. Detection of retinol in plasma Retinol acetate (2 mg), used as internal standard (IS), was dissolved in 40 mL methanol (50 µg/mL). This was then further diluted with methanol to reach 0.9 µg/mL retinol acetate. A 100 µL alliquot of this solution was added to 100 µL plasma sample and vortexed for 30 s. N-hexane 0.1% BHT (1 mL) was added and the sample was vortexed (XW-80 A, Shanghai, China) for 1 min and centrifuged at 5,000 rpm for 5 min. BHT acts as an antioxidant, protecting the active ingredients in the solution from oxidation. After the upper layer of n-hexane was collected, the sample was again extracted with 1 mL of n-hexane, and both n-hexane layers were combined and dried with nitrogen gas. The residue was dissolved with methanol (300 µL) and the solution was filtered through a 0.45 μm membrane for HPLC analysis. Transcriptomic sequencing The rats were randomly divided into three groups: two treatment groups (ROL-C1 and ROL-C2) and one control group (n = 3 each). Before the experiment, the dorsal region of each rat was shaved using a hair trimmer and depilatory cream. The next day, they were treated with the ROLs on the back skin with a 1 g dose (3 × 3 cm^2). After 24 h, the rats were euthanised by intraperitoneal injection of 180 mg/kg pentobarbital and skin samples were obtained. Total RNA was extracted from the skin tissues, and transcriptomic sequencing was performed by Gene Denovo Biotechnology Co. (Guangzhou, China). Data statistics All experimental data were obtained from at least three experiments, were statistically analysed using GraphPad Prism 8.0 software, and all data are expressed as the means ± SEMs. P values less than 0.05 are statistically significant. Raman spectra and chromatograms were plotted using Origin 2021 software. Results In vitro skin permeation studies High performance liquid chromatography chromatograms The retention times in HPLC for ROL and BHT are clearly different (11.9 min and 7.2 min (Fig. [64]1A-B), and after treatment ROL-C1 and ROL-C2, both signals were observed in skin or stratum corneum alone, but no ROL was detected in the receiver solution (Fig. [65]1C). More methodological verification results are shown in Table [66]S1. Fig. 1. [67]Fig. 1 [68]Open in a new tab HPLC chromatograms of retinol in the Franz diffusion cell method. (A) retinol (ROL) standard; (B) blank samples; (C) experimental samples. (BHT: dibutylhydroxytoluene). Franz cell diffusion assay The 24 h Franz diffusion cell experiment was performed with skin (a schematic diagram of the diffusion cell is shown in Fig. [69]2A). After treatment of Bama Xiang pig skin, ROL intradermal absorption in pig full-thickness skin increased over time and reached a plateau after 4 h. Intradermal absorption was 7.79 µg/mg for ROL-C1 and 4.87 µg/mg for ROL-C2. After this, the rate gradually decreased with time. The absorption rate in the ROL-C1 group was higher than in the ROL-C2 group at 1, 2 and 4 h (Fig. [70]2D), similar to what has been found in rats and mice skin in vitro. After 24 h, ROL penetration in full-thickness skin was 9.28 (pig), 16.53 (mouse) or 25.88 µg/mg (rat) in the ROL-C1 group. In the ROL-C2 group, these values were 2–4 times lower: 5.59 µg/mg, P < 0.05 (pig), 4.14 µg/mg, P < 0.01 (mouse) and 8.24 µg/mg, P < 0.05, (rat) (see Table [71]1). Fig. 2. [72]Fig. 2 [73]Open in a new tab Distribution of retinol in pig/mouse/rat skin was examined using the Franz diffusion cell method. (A) Schematic diagram of the vertical diffusion cell; (B) accumulated amount of retinol in the stratum corneum (SC) and skin without SC after 24-h diffusing in Franz diffusion cell. (C) (c1-c3) assessment of retinol distribution in the stratum corneum by the tape peel method in a 24-h Franz diffusion cell experiment; accumulated (D) (d1-d3) time - dependent full-thickness skin absorption of retinol. (*P < 0.05; **P < 0.01; ***P < 0.001). Table 1. Full-thickness skin retention (µg/mg). Time(hour) ROLs Pig skin Mouse skin Rat skin 1 ROL-C1 3.57 7.49 12.48 ROL-C2 4.32 2.86 2.2 2 ROL-C1 5.08 15.73 9.75 ROL-C2 4.62 8.15 2.4 4 ROL-C1 7.79 14.67 9.67 ROL-C2 4.87 8.47 1.9 24 ROL-C1 9.28 16.53 25.88 ROL-C2 5.59 4.14 8.24 [74]Open in a new tab In addition, the retinol penetration amount in the stratum corneum (SC) of pig skin for the ROL-C1 group was 7.52 µg/mg. Most of the retinol was distributed in the first and second layers (stratum corneum) of tape, add up to 5.52 µg/mg, account for 78%. In the ROL-C2 group, the total content of the 1 ~ 2 layer of tape was 2.84 µg/mg, which accounted for 62.04% of the whole SC content (4.58 µg/mg). The content decreased along with the number of layers increased. In the mouse skin group treated with ROL-C1, retinol permeation in SC was 18.19 µg/mg, and the ROL-C2 group was 3.15 µg/mg. In the rat group, retinol in SC of ROL-C1group was 19.07 µg/mg, and the ROL-C2 group was 7.85 µg/mg (Fig. [75]2C). In all three SC skin in vitro, penetration in the ROL-C1 group was higher than in the ROL-C2 group (Fig. [76]2B-C). Retinol penetration amount was rats > mice > pig skin. For the skin without SC in pig skin group, retention in the ROL-C1 and ROL-C2 groups was 1.56 and 1.39 µg/mg, respectively. Differences were not significant (P > 0.05, Fig. [77]2C c1). In mice, values were 1.42 µg/mg and 0.99 µg/mg, respectively. In rats, the retinol in skin without SC was not detected. Overall, ROL in the skin without SC in the ROL-C1 group was slightly higher than in the ROL-C2 group, with no significant difference in skin retention (P > 0.05, Fig. [78]2B and C c2, c3). In summary, retinol penetration in pig full-thickness skin is lower than in mouse and rat skin, possibly due to a thicker and harder skin in the pig, compared to the skin of rodents^[79]23. Nevertheless, the results in three skins showed a consistent trend in that ROL-C1 penetration was more effective than that of ROL-C2. Confocal Raman spectroscopy As shown in Fig. [80]3C, the characteristic absorption peak of retinol (red arrow in the figure) was in the Raman shift interval from 1570 to 1615 cm^−1. In the Raman spectra of the blank pig skin, which did not present the characteristic ROL peak, the peak near 1450 cm^−1 corresponded to the δCH[2] of the proteins and lipids in the stratum corneum. As a biological tissue, skin composition and structure are highly complex and variable. Since Raman scattering is unique to a particular chemical functional group, the Raman spectra acquired from skin are equally complex and diverse^[81]24, and in this experimental result, it does not interfere with the results of the main peak. For example, some Raman peaks at approximately 1500 cm^−1 were observed at depths of 26 and 34 μm. Fig. 3. [82]Fig. 3 [83]Open in a new tab Raman spectra of pig skin with depth from surface 0 μm to 34 μm. (A) Raman mapping of the skin of the ROL-C1. (B) Raman mapping of the skin of the ROL-C2. (C) Raman spectra of the surface 0 μm of the skin after treatment, the active material standard and the normal skin. (D) Raman spectra of the skin after treatment of the ROL-C1. (E) Raman spectra of the skin after treatment of the ROL-C2. (Red arrow: the characteristic absorption peak; orange arrow: the δCH[2] of the proteins and lipids in the stratum corneum peak). By comparing the two Raman spectra (Fig. [84]3D and E), we speculate that after ROL-C1 treatment and ROL-C2 treatment of pig skin, ROLs can penetrate into the skin, and the Raman spectrum intensity shows a trend of attenuation with increasing skin depth. Based on the results, after ROL-C1 treatment, at the Raman shift of 1597 cm-1, it is speculated that the skin depth may reach 16 μm, while ROL-C2 to 14 μm. In addition, combined with the Raman mapping results (Fig. [85]3A and B), ROL-C1 treatment shows a stronger characteristic absorption peak than ROL-C2 treatment. In vivo permeation study We then investigated the penetration of retinol in vivo. Similar to the in vitro experiments above, absorption of retinol in full-thickness skin reached a plateau after 2 h for both treatments, was stable over 24 h and was higher for ROL-C1 (49.59 µg / mg) than for ROL-C2 (33.66 µg / mg) (Fig. [86]4B). Fig. 4. [87]Fig. 4 [88]Open in a new tab In vivo penetration studies after 24 h. (A) HPLC chromatogram of real time plasma sample, standard of retinol and blank plasma sample (IS: retinol acetate); (B) retinol retention in mouse full-thickness skin; (C) plasma retinol concentration over time in rats; (D) evolution of retinol plasma concentration in mice. For plasma concentration, as shown in Fig. [89]4A and Figure S2, the HPLC retention time of the IS was 20.0 min. The method for extracting retinol from plasma showed good specificity. More methodological verification results are shown in Table [90]S1. Retinol content did not change in plasma over the course of the experiment, since the absorption only corresponds to endogenous retinol (Fig. [91]4D). The plasma drug concentration detection method used in mouse was also used in rats. No significant changes were found in retinol concentration in groups ROL-C1 and ROL-C2 at any of the time points (0, 1, 2, 4–24 h), consistent with what was found in mice (Fig. [92]4C). This further supports that retinol in the in vitro diffusion cell did not penetrate into the receiving solution. In summary, we believe that within 24 h of treatment, retinol from treatment did not enter the blood circulation through the skin and mainly accumulated in the skin. Transcriptomic analysis of rat skin The mechanism of retinol transport was explored with high-throughput sequencing. Compared with the control group, treatment for 24 h induced less overexpressed DEGs in the ROL-C1 group (496 genes) than in the ROL-C2 group (1, 287 genes). What is more, there were 430 co-upregulated genes among the DEGs between the ROL-C1 group and the ROL-C2 group (Fig. [93]5A-C). Fig. 5. [94]Fig. 5 [95]Open in a new tab Transcriptomic sequencing analysis of rat skin. (A) and (B) Volcano plots of DEGs for ROL-C1 (A) and ROL-C2 (B) versus control. FDR < 0.05, |fold change| >2; (C) the number of upregulated genes and the Venn diagram of the control vs. ROL-C1 and control vs. ROL-C2 comparisons; (D) top 10 GO terms for the 1,353 upregulated genes in the control vs. ROL-C1 and control vs. ROL-C2 comparisons; (E) KEGG pathway enrichment analysis of the top 10 genes related to “retinoid metabolic process” and “transporter activity”; (F) heatmap of genes in control, ROL-C1 and ROL-C2 groups. We identified the enriched pathways associated with the overall 1,353 upregulated genes (Fig. [96]5D). The top three Gene Ontology (GO) terms were related to skin functions, such as epidermis development, epithelium development and skin development. In addition, in the GO enrichment analysis, using “retinoid metabolic process” and “transporter activity” as search terms, 126 genes were identified, including genes related to retinol transport and metabolism (e.g., Stra6, Rdh12 and Lrat), as shown in Fig. [97]5F. KEGG pathway enrichment analysis^[98]25–[99]27 of these 126 genes revealed the top ten genes being enriched (Fig. 5E). We found the activation of the ABC transporter, Retinol metabolism, Vitamin metabolism and Absorption pathways. Discussion Members of the vitamin A family, include retinoic acid, retinol esters and retinol, are widely used in cosmetics and pharmaceuticals. Especially retinol is relatively milder, so it is more commonly used in cosmetics. External retinol preparations are a major class of anti-ageing agents, and their safety and efficacy has been a focus of investigation. In 2022, the EU SCCS revised the safety assessment for vitamin A derivatives (retinol, retinol acetate and retinol palmitate) for their use in cosmetics. The effects of cosmetics are directly related to absorption rate and amount of the active ingredients. In cosmetics research, it is important to understand the speed and range of the skin penetration of ingredients^[100]28. Therefore, appropriate detection methods are needed to evaluate and determine the absorption of active ingredients in cosmetics to ensure safety and effectiveness. Moreover, in vitro methods are used in conjunction with ex vivo and / or in vivo studies to more quickly translate newly developed topical and transdermal drug delivery systems from the laboratory to the clinic^[101]21. IVPT is an effective method that replaces equivalent clinical trials for skin drug delivery preparations and investigates the therapeutic equivalence of experimental drugs compared with reference drugs by evaluating their in vitro absorption kinetics through isolated human skin. IVPT can be used to screen drug preparations, evaluate the properties and mechanism of action of the preparation carrier system to promote skin penetration, and evaluate the systemic risk of skin exposure to chemicals by predicting the skin transport of drug molecules^[102]29. In vitro diffusion modelling is an important tool for screening the penetration ability of active ingredients in various formulations^[103]30. Therefore, in the current study, we used a modified vertical diffusion cell model, according to the relevant guidelines and literature, to perform in vitro diffusion cell experiments that simulate real skin absorption. In the evaluation of cosmetics, the experimental time of the in vitro Franz diffusion cell is related to the actual use of cosmetics. Yourick JJ et al.^[104]18. reported that the single best estimate of systemic retinol absorption from in vitro human skin studies is the 24-h receptor fluid value. Therefore, we carried out a 24 h Franz diffusion cell experiment on external retinol preparations. The results of the 24 h Franz diffusion cell experiment showed that retinol could not be detected in the receiving solution. As reported in the literature, retinol accumulates in the skin^[105]31,[106]32, possibly because the concentration of retinol is too low to reach the detection limit or because it has not penetrated into the receiving solution. Mice and rats are easy feeding and be widely used in pilot study. Literatures showed that the permeability of rats and mice were greater than that of pigskin. By comparing the amount of retinol penetration into the three skin types (mouse, rat and pig), we found that the absorption of rat in full-thickness skin was the highest, followed by mice, and finally pig skin. The amount of penetration in rats and mice was about 2 times of pig skin, which corresponds with the existing scientific literature^[107]33. Minipigs are a promising animal model for predicting percutaneous drug absorption in humans^[108]34. Moreover, Ranamukhaarachchi et al.^[109]35 showed that porcine stratum corneum became a closer model for human stratum corneum after freezing by measuring Young’s (elastic) modulus. In summary, for the in vitro diffusion cell, the results of pig skin are closer to the actual skin absorption. In addition, through three in vitro animal models (pig, mouse, and rat skin), we also found that ROL can penetrate the stratum corneum and accumulate in skin, but the accumulated amount of retinol in the stratum corneum is greater than that in the skin. According to amount of retinol released into the skin over time, the percutaneous permeation of retinol is time dependent, but the absorption rate of retinol in the skin decreases over time, and the amount of retinol in the skin reaches its absorption peak at 2–4 h after administration. The rapid absorption and retention of retinol in the skin is conducive to its biological activity. What is more, by calculating the slope of the curve of the experimental results for the relative amount of retention in the full-thickness skin, we found that the absorption rate of the full-thickness skin in the ROL-C1 group was faster than that in the ROL-C2 group. Therefore, it can be preliminarily explained that the in vitro percutaneous penetration effect of retinol in the ROL-C1 group was better than that in the ROL-C2 group. However, in our study, the Franz diffusion cell experiment was static, which is not the actual application of the product. To overcome this obstacle, it was been reported that a modified Franz diffusion cell method achieved significant visual improvement in both the epidermis and dermis after 4 h by applying a rolling massage to skin treated with retinol emulsion^[110]36. Therefore, the static Franz diffusion cell has many limitations and it cannot substitute for the live animal evaluation. Raman is a label-free technique, and the main advantage of confocal Raman spectroscopy is the ability to detect many labels or probes simultaneously. When labelling small bioactive compounds, fluorescent dyes often affect biological activity; however, small Raman labels can be introduced without affecting biological activity, and bioactive compounds in living cells can be visualised^[111]37, therefore we utilised this noninvasive property of Raman spectroscopy for the analysis of isolated pig skin, which allows for the detection of topical cosmetics that cannot be used for localisation detection due to the inability to add labels. As a commonly used active ingredient in the cosmetic industry, the dermal delivery of nicotinamide (NIA) was examined by Iliopoulos F et al.^[112]38 using an in vitro 24-h finite-dose Franz diffusion cell method and in vivo confocal Raman spectrometry after 60 min of treatment. They analysed the results of the two methods and found that the in vitro cumulative transmittance of NIA was correlated with the in vivo dermal uptake of NIA at 2 μm. According to the results of Raman spectroscopy, we found that at the Raman shift of 1597 cm^−1, the absorption peaks of ROL-C1 and ROL-C2 can reach a depth of more than 10 μm in the skin, showing transdermal penetration ability. In addition, the peak intensity in the ROL-C1 group gradually decayed with increasing skin depth, up to 16 μm. The intensity was higher than in the ROL-C2 group, consistent with the Franz diffusion cell experiment results. Raman mapping was not totally effective because of interference with other ingredients, especially when the concentration of the target compound in the mixture is relatively low^[113]39, but they were consistent with the Franz diffusion cell results. In vivo and in vitro correlation studies are also important components of transdermal permeation studies of topical formulations. Fan Y^[114]40 explored the differences in transdermal permeation properties between ex vivo dermal permeation experiments using Franz diffusion cells and in vivo dermal / haematological pharmacokinetic experiments facilitated by dual-site microdialysis sampling. In this study, for the homologous of mice and rat, we only employed the KM mouse model to quantitate the concentration of ROL in whole skin and plasma. In full-thickness skin (Fig. [115]4B), after treatment with ROLs, intradermal absorption rate reached a plateau after 4 h and decreased after that. Although ROL was detected in skin after 24 h, exogenous ROL was not detected in plasma (Fig. [116]4D). ROL was not detected in SD rat plasma either (Fig. [117]4C). It is interesting that ROL-C1 with lower ROL concentration has more penetration than in ROL-C2. The appearance of ROL-C1, a lecithin organogel, was clear. This preparaion enhances permeation and skin deposition^[118]41,[119]42. A lecithin organogel may improve permeation at least 10 times^[120]43. Particles of ROL-C1 were smoother and spherical from the images of SEM and Polarizing microscope. In contrast, the surface of particles in the ROL-C2 were flaky, which prevented mobility (see more details in supplement). The preparation protocol for these samples maybe also important. In addition, by comparing the in vitro and in vivo data of mice, we found that the absorption of the full-thickness skin in vivo was about three times of the ex vivo skin. We speculated this might be related to some transport genes in the in vivo skin, which would be removed or deactivated during the process of ex vivo skin, because the skin would be removed the cells and the cell debris by some extraordinary methods. So we performed transcriptome sequencing to study the transport and absorption mechanism of retinol. The transcriptome changed after treatment with ROLs. The cellular and molecular changes were found in epidermis and dermis similarly as previous studies of retinol topical application^[121]44. Since the transcutaneous absorption and transport of retinol involves pathway activation, we focused on up-regulating differential genes. In our study, the development of the cuticle, epidermis and dermis in the skin was significantly activated after treatment with the ROLs, reflecting the biological function of the external retinol preparations. KEGG pathway enrichment analysis revealed that the ABC transporter and the retinol metabolism pathways being activated. The relationship between retinol transport and activation of the ABC transporter pathway is an interesting topic for future research. Among the 126 DEGs identified, we found genes related to retinol metabolism and transport. One of these, Stra6, encodes a transmembrane protein with unique structural characteristics that can catalyse the two-way transport of retinol and isolate retinol from target cells^[122]45. In addition, it has been reported that Stra6 acts as both an RBP receptor and a transporter for vitamin A uptake^[123]46. These findings suggest that retinol may be transported into cells through the Stra6 receptor after skin treatment with retinol topical preparations. However, the skin acts as a microretinol reservoir, and retinol as a fat-soluble molecule may also enter the skin cells through passive diffusion. Therefore, the next study on the process of retinol transport into cells in topical preparations will be interesting. After transported into the cell, the metabolic pathway of retinol is activated, resulting in metabolic processes, such as the Lrat gene, which catalyzes the esterification of all-trans retinol to all-trans retinol ester; the Rdh12 and Aldh1a3 genes catalyze the metabolism of retinol to retinoic acid, which in turn exerts biological functions. In summary, after ROLs treatment, the metabolism and transport pathways of retinol in the skin are activated, and there is an absorption and transport process that occurs in the skin. Retinol may enter skin cells through upregulated expression of the Stra6 gene and then be retained in the skin and play a role in skin development. However, RNA-seq analysis has several limitations; it is only a reference and needs quantitative verification. Thus, further experimental verification is needed. Conclusion We report an in vitro/in vivo correlation of transdermal absorption of two external retinol cosmetics. Absorption of full-thickness skin in vivo was about three times higher than in ex vivo skin. Transcriptomic sequencing showed that Stra6 receptor and Aldh1a3, are relevant to the search terms “retinoid metabolic process” and “transporter activity”. ROL did not enter blood circulation, which suggests its application as long-term daily use. In summary, this study attempted to explore the in vitro/in vivo correlation of transdermal absorption behaviour of retinol cosmetics. Electronic supplementary material Below is the link to the electronic supplementary material. [124]Supplementary Material 1.^ (3.1MB, docx) Acknowledgements