Time‐course quantitative mapping of caffeine within the epidermis, using high‐contrast pump–probe stimulated Raman scattering microscopy

Abstract Background An assessment of the drug penetration and distribution profiles within the skin is essential in dermatology and cosmetology. Recent advances in label‐free imaging technologies have facilitated the direct detection of unlabeled compounds in tissues, with high resolution. However, it remains challenging to provide quantitative time‐course distribution maps of drugs within the complex skin tissue. The present study aims at acquiring the real‐time quantitative skin penetration profiles of topically applied caffeine, by means of a combination of pump–probe phase‐modulated stimulated Raman scattering (PM‐SRS) and confocal reflection microscopy. The recently developed PM‐SRS microscopy is a unique imaging tool that can minimize strong background signals through a pulse‐shaping technique, while providing high‐contrast images of small molecules in tissues. Materials and methods Reconstructed human skin epidermis models were used in order to analyze caffeine penetration in tissues. The penetration profiles of caffeine in an aqueous solution, an oil‐in‐water gel, and a water‐in‐oil gel were examined by combining PM‐SRS and confocal reflection microscopy. Results The characteristic Raman signal of caffeine was directly detected in the skin model using PM‐SRS. Integrating PM‐SRS and confocal reflection microscopy allowed real‐time concentration maps of caffeine to be obtained from formulation samples, within the skin model. Compared with the conventional Raman detection method, PM‐SRS lowered the background tissue‐oriented signals and supplied high‐contrast images of caffeine. Conclusion We successfully established real‐time skin penetration profiles of caffeine from different formulations. PM‐SRS microscopy proved to be a powerful, non‐invasive, and real‐time depth‐profile imaging technique for use in quantitative studies of topically applied drugs.


INTRODUCTION
As the largest organ of the human body, the skin acts as a barrier against the environment. The outermost (upper) layer of the skin, the stratum corneum (SC), is composed of corneocytes and intercellular lipids, such as cholesterol, free fatty acids, and ceramides, and mainly provides a rate-limiting barrier, which prevents not only excessive water loss from the body, but also the invasion of toxic chemicals and microbes. 1,2 Topically applied drugs, used in treating dermatological diseases, and active ingredients for cosmetic purposes, are examples of agents that need to cross the mentioned barrier to be effective inside the skin. Therefore, understanding the kinetics of both penetration and distribution of the applied drugs is crucial for pharmaceutical and cosmetic research.
While assessing the mechanism of skin penetration, time-course quantitative maps of the target molecules are ideally required. There are several conventional methods for studying skin penetration, including the Franz diffusion cells and tape-stripping. The Franz diffusion cell system combines ex vivo skin explants or in vitro reconstructed human skin cells, assembled on diffusion cells, with the appropriate quantification methods, such as liquid scintillation spectrometry for radiation-labeled compounds 3 or high-performance liquid chromatography (HPLC). 4,5 Using these techniques, quantitative profiles can be obtained for topically applied compounds that reach the receiver chamber under the skin samples. Tape-stripping, in turn, is a minimally invasive technique that can be applied in vivo. Briefly, it consists of removing the SC through the repeated application of adhesive tapes, followed by measurement of the amount of target molecules extracted from each tape, also giving rise to concentration profiles across the SC. 6 Although these methods have been validated and widely applied, they are unable to depict with enough detail the time-course distribution maps of compounds of interest in the skin layers.
Thus far, methodologies based on Raman spectroscopy have attracted attention in the field of skin research, because they are label-free and non-invasive processes. Raman spectroscopy has been employed to define skin penetration profiles of topically applied substances, both in vitro and in vivo, [7][8][9][10][11][12][13][14][15] and in the analysis of the molecular composition of the skin samples. 16 Nevertheless, confocal Raman microscopy 7-13 does not support real-time image acquisition because of the weak signal, whereas conventional stimulated Raman scattering (SRS) microscopy 14,15 suffers from low signal-to-background ratios because of the strong background signals from tissues. These obstacles limit real-time imaging and signal detection especially at lower concentrations. In practice, a more sensitive and highly specific detection technique is needed to better understand the precise penetration kinetics of molecules of interest.
With our current work, we aimed to acquire penetration profiles of caffeine, which is widely used as a reference for skin penetration studies, in dermatology and also cosmetology, 4,8,9,10 released from different types of formulations -more exactly, an aqueous solution, a water-in-oil gel, and an oil-in-water gel. We opted for the use of the novel pump-probe phase-modulated (PM) SRS microscopy, which minimizes strong background signals with a pulse-shaping technique, thereby producing high-contrast imaging of small molecules in the tissues of interest. 17,18 As presently established, the manner in which active molecules penetrate the skin depends not only on their own physical and chemical properties, but also on the characteristics of their delivery vehicle. Thus, the formulation samples of a disparate nature were applied to the artificial epidermis models. Ultimately, this recent microscopy technology has made direct in vitro visualization of time-dependent depth profiles of caffeine in skin models, using different formulations as vehicles.

Substances and formulations
Caffeine (Shizuoka Coffein Co., Ltd., Shizuoka, Japan) was utilized as the unlabeled model target for microscopic visualization. For the aqueous solution sample, caffeine was dissolved in pure water, at room temperature and 1% w/w. Regarding the water-in-oil and oil-in-water gel formulations, they each contained 1% w/w caffeine.

Oil-in-water gel
The composition of the oil-in-water gel was as follows: 0.4% w/w ammonium acryloyldimethyltaurate/beheneth-25 methacrylate crosspolymer, 38% w/w glycerine, 1% w/w squalane, 0.4% w/w phenoxyethanol, 59.2% w/w pure water, and 1% w/w caffeine. Caffeine was first dissolved in pure water. The gels were then prepared by premixing the oil-in-water and hydrophobic phases. Next, the oil-inwater and hydrophobic components were mixed by a homo mixer, at 5000 rpm for 2 min. Finally, all samples were placed on the stage of the PM-SRS microscope and images were recorded as described below.

PM-SRS imaging
The setup of our PM-SRS microscope was designed to allow switching

Raman spectra of skin model and caffeine
Prior to the penetration study, an untreated skin model sample and caffeine solution were analyzed using a self-made confocal Raman spectroscope with a 532 nm CW excitation laser and a PM-SRS microscope within the 400−1500 cm −1 range. The spontaneous Raman spectrum close to the surface of the untreated skin model showed some specific bands generated from the skin model components (Figure 2A). More precisely, the ring-breathing mode in aromatic amino acids, shown at 1004 cm −1 , the CH 2 deformation mode generated from alkyl chains in lipids, at 1314 cm −1 , and an unspecific C-H deformation, present at 1448 cm −1 , 19 were quite noticeable. By contrast, the spectrum resulting from the PM-SRS measurement, with a 1.7 ps probe delay, suppressed these background signals from the skin model components ( Figure 2B), apart from the one corresponding to the ring-breathing mode, with the origin in the aromatic amino acids as mentioned above, for 1004 cm −1 . A characteristic signal from caffeine was detected at 560 cm −1 (vibration from the O=C-N deformation 20 ), similar to the confocal Raman and PM-SRS spectra. Notwithstanding, the signal-tobackground ratio of the PM-SRS spectrum was significantly higher than that of the confocal Raman spectrum, given the long lifetime of the caffeine-specific vibration. Figure 2C shows the calibration curve of the caffeine solution, obtained using PM-SRS at 560 cm −1 . A linear fit is shown for the concentration range of 0 to 52.1 mM.

Skin penetration profile of caffeine
The confocal reflection XZ image of this model was acquired 15 min after applying 1% caffeine solution ( Figure 3A). Strong signals were

Distribution maps of formulation samples
The formulation samples were directly analyzed using PM-SRS, among which Figures 4A and 4B are the distribution maps of silicone oil and caffeine (560 cm −1 ) and were captured using the oil-in-water gel. In terms of the distribution maps of the formulation samples, our study suggests that caffeine density was superior in the water phase of the oil-in-water gel. By contrast, caffeine was more evenly distributed in the water and oil phases in the water-in-oil gel, which may be due to the fact that caffeine is soluble in both water and ethanol, and ethanol is also soluble in the low-viscosity dimethicone used for the gel.

Skin penetration profiles of caffeine from formulation samples
Finally, Figure 5A depicts the overlapped images of the PM-SRS and the confocal reflection modes produced 360 min after applying 1% caffeine water-in-oil gel on the skin model. The equivalent overlapped image for the oil-in-water formulation is shown in Figure 5B, 360 min after applying 1% caffeine oil-in-water gel to the tissue models. Images of the PM-SRS signal at 560 cm −1 obtained over time were used to calculate the concentration profiles of caffeine, once again within a depth of 5−15 µm from the relative zero-depth, of the water-in-oil gel and the oil-in-water gel application, using the calibration curves ( Figures 5C   and 5D). These results support the idea that caffeine clearly penetrated the skin models over time from both the water-in-oil and oil-in-water gel formulations.

DISCUSSION
In the process of confocal Raman analysis of biological tissues, Raman signal attenuation according to depth is one of the core challenges.
Since the cosmetic formulation samples usually consist of numerous ingredients and are not always homogeneous, the refractive-index difference of the samples tends to become more extensive in comparison with pure solution samples. As a result, strong scattering from the formulation samples significantly affects the optical transmission of the laser, thus disturbing the measurement of the Raman signals from the target compound. The optical scattering loss could be even more problematic in conventional confocal Raman microscopy, particularly when dealing with a 532 nm green laser setup. Despite that, the PM-SRS method could overcome this limitation and provide penetration profiles for both oil-in-water and water-in-oil samples, because of its higher sensitivity and reduced scattering, due to working with nearinfrared laser wavelengths. In the present study, we observed that the oil-in-water gel gave rise to stronger scattering and signal loss than the water-in-oil gel. The relative signal intensity, in this sense, was 0.64 for the water-in-oil gel and 0.17 for the oil-in-water gel, correspondingly, compared to 1.00 for the caffeine solution. Furthermore, PM-SRS images of the water-in-oil ( Figure 5A) and the oil-in-water ( Figure 5B) gels were obtained twice and eight times accumulation, respectively.
When tracing the skin penetration profiles of certain molecules, PM-SRS microscopy has another advantage compared with other Raman microscopy approaches. Confocal Raman microscopy and conventional amplitude-modulated SRS microscopy detect strong tissue-oriented Raman scattering, caused by components such as proteins, lipids, and water. Consecutively, high background hinders the detection of target molecules within tissues, and more so at lower concentrations. To overcome this drawback, several approaches have been proposed, including an analysis of the correlation data between the tissue-oriented signal and the signal from a molecule of interest, 10 as well as the use of a non-negative matrix algorithm of depth-resolved Raman spectra of the tissue. 13 By contrast, PM-SRS provides direct quantitative depth profiles of the target molecules, given that PM-SRS can suppress the background signals from the skin samples and trace the characteristic Raman peaks from target substances, even at lower concentrations.
However, it is impossible to determine the actual surface location of the sample or the absolute skin depth profile from the data obtained from our setup alone. Therefore, the relative zero-depth needs to be defined to discuss the depth profile. This is a limitation of the present study. For a more accurate discussion of the penetration depth, a higher depth resolution, a separation of the tissue components through multi-color Raman spectral imaging, and an estimation from the depthspecific histomorphological information will be needed. The use of multi-color imaging is also expected to further improve the separation performance of skin components and skin-like substances.

CONCLUSION
In this study, we demonstrated a label-free microscopy-based method