Skin evaluations were composed of three parts as follows: Study 1: basic skin research for understanding the relationship between epidermal keratinocyte architecture and visual/physiological properties of skin aging, Study 2: evaluation of SK-II’s effects on tomographic keratinocyte architecture, and Study 3: evaluation of SK-II’s effects on the visual and physiological aging parameters as well as estimated cellular architecture.

In Study 1 (Protocol CT12002), we measured epidermal keratinocyte architecture (cellular density and cellular uniformity), visual aging parameters (texture, wrinkles, pores, and L value), and physiological properties (hydration and TEWL) in 78 Asian females (age 21 - 69, mean ± SD: 43.97 ± 13.64). The epidermal architecture was assessed using in vivo two-photon stereoscopic tomography obtaining by-layer epidermal section images (DermaInspect MPTflex^TM 10; JenLab GmbH, Jena, Germany). Skin hydration was measured with a Corneometer^® CM825 (Courage + Khazaka, Courage + Khazaka Electronic GmbH, Cologne, Germany) and TEWL was assessed with a Vapometer^® (Delphine Technologies Ltd., Kuopio, Finland).

Study 2 (Protocol CT12006) was performed to evaluate whether SK-II affects the epidermal keratinocyte architecture as well as skin hydration and TEWL in 30 Asian female volunteers aged 24 - 58 (41.3 ± 11.34). SK-II was applied twice daily to the left forearm in a 3 × 3 cm square area. The right forearm was not treated during the course of the study, and served as a control in order to compare the treatment site. Skin measurements were performed in both forearms at the start of the day (baseline), on days 14 and 28.

In Study 3 (Protocol CT23002), we assessed the effects of SK-II on various skin aging parameters in facial skin. Skin hydration, visual parameters, and estimated cellular architecture were measured in 35 Asian females aged 20 - 50 (35.1 ± 6.5). SK-II was applied on the subject’s face twice daily and skin measurement was conducted at baseline (day 0) and on days 3, 7, and 14.

All study protocols were approved by the Ethical Committee of Global Product Stewardship in P&G Innovation Godo Kaisha. Written informed consent was obtained from all subjects prior to their enrollment in each study.

The keratinocyte cellular architecture was examined using in vivo two-photon tomography with a DermaInspect MPTflex^TM 10 ( Figure 1 ) [17] [18] . This instrument allows 3D imaging with high spatial resolution in the submicrometer range based on two-photon excited fluorescence and second harmonic generation. Three repeated measurements were performed at the middle of the cheek on the left side of the face, or inner forearm. A combination of intense 80-MHz femtosecond near-infrared laser pulses and flying spot technology for beam scanning allowed imaging with high spatial and temporal resolution to be obtained. Light intensities in the range from MW∙cm⁻² to GW∙cm⁻² allowed the detection of endogenous fluorophores, targeting keratinocyte architecture in the granular layer of the skin, through two-photon excitation. The laser beam scanned a skin field of interest of a size of 300 × 300 lm² (512 × 512 pixels) with a typical beam well time of <40 μs per pixel, scanning at 2 lm pitch by depth, on

the skin of the cheek or forearm. The time between two laser pulses was 12 ns. The weak fluorescence was detected by sensitive photodetectors. Image processing was performed to obtain high-resolution autofluorescence images.

To quantify the keratinocyte architecture, we collected four layers of keratinocyte cellular images at 8 μm pitch (32 μm depth) at the central field of 100 × 100 μm in the epidermis layer below the stratum corneum ( Figure 2 ). The cellular density was calculated using an image analysis algorithm. Meanwhile, the cellular uniformity was calculated using the following formula.

$U_{shape}=\left({\displaystyle {\sum }_{i=1}^m{\left(X_m-\frac{{\bar{X}}_i+{\bar{Y}}_i}{2}\right)}^2}+{\displaystyle {\sum }_{i=1}^m{\left(Y_m-\frac{{\bar{X}}_i+{\bar{Y}}_i}{2}\right)}^2}\right)/m$

$U_{size}=\left({\displaystyle {\sum }_{i=1}^m{\left(X_m-{\bar{X}}_i\right)}^2}+{\displaystyle {\sum }_{i=1}^m{\left(Y_m-{\bar{Y}}_i\right)}^2}\right)/m$

$\text{Cellular}\text{Uniformity}=U_{shape}+U_{Size}$

m: cell count, xi: short diameter of the cell, Y_i: long diameter of the cell

The subjects washed their faces using the prescribed cleansing foam and then spent 20 min acclimating to the environment of the measurement room. The examination room was maintained at a constant temperature and humidity (room temperature 20˚C ± 2˚C, relative humidity 50% ± 5%). None of the subjects underwent any type of esthetic treatment such as laser cosmetic procedures during the study period. Each subject’s face was photographed using a portable image capture system (Magic Scan) illuminated by a number of 5600-K light-emitting diodes mounted at both left and right sides of the system ( Figure 3 ) [9] . A high-resolution complementary-symmetry metal–oxide–semiconductor digital camera, capable of generating 1980 (vertical) × 1024 (horizontal) effective picture elements (pixels), was also mounted in the imaging module. A series of deep-learning-based post-calibration algorithms was carried out to adjust to the consistent brightness, hue, and saturation after capturing the facial images. This

enabled the captured images to be controlled to ensure reproducible collection under the different external optical conditions. A neutral 8.0 gray color board (GretagMacbeth GmbH, Munich, Germany) was used for white balancing of the camera.

The region of interest (ROI) of the images was from the outer edge of the eye to the cheek, and the following characteristic objects were extracted by measuring the contrast in the shape and pixels using an image analysis algorithm [15] [16] . Wrinkles were defined as ≥5 mm in length, perimeter/length ratio ≤ 2.5, and circularity (perimeter²/area) ≥ 34. Total wrinkle area fraction was quantified as follows: total wrinkle area (pixels)/ROI (pixels). As an index of texture, total texture area fraction [total texture area (pixels)/ROI (pixels)] was quantified as ≤ 3 mm² in area, aspect ratio ranging from 0.5 to 2, and color contrast deltaE ≥ 1.5. Facial skin tone (L-value) was also measured in the ROI. Pores were identified in an image analysis algorithm by detecting circular pore-like shapes via edge-enhanced binary images at the cheek region. The total area of detected pores was then calculated.

All statistical analyses were performed using JMP^® Pro 16.1.0 (SAS Institute Inc., Cary, NC, USA). For each outcome variable, data were analyzed by fitting a linear mixed-effects model with timepoint as the fixed effect and subject as the random effect to account for within-subject variation. Pearson’s correlation coefficients (r) between the following variables were examined: 1: between skin cellular density and chronological age, skin hydration, TEWL, and four skin visual parameters (texture, wrinkles, pores, and L value) in Study 1. Quantitative comparisons were also performed for the following variables: 1: cellular density, cellular uniformity, hydration, TEWL, and visual parameters (texture, wrinkles, pores, and L value) at the baseline versus at week 2 and week 4, upon treatment with SK-II in Study 2; and 2: those skin parameters at the baseline versus on days 3, 7, and 14 after the SK-II treatment in Study 3 using two-way ANOVA. Logistic regression analysis was performed to model the association of cell density and uniformity as response variables, and signs of skin aging parameters [texture, pores, wrinkles, and skin tone (L-value)] as exploratory variables. A P-value less than 0.05 was considered to be statistically significant.

We measured tomographic architecture, visual, and physiological skin aging parameters in 78 women of various ages. A significant decrease in skin cellular density was observed in association with increasing age (r = −0.289, P < 0.05) ( Figure 4(A) ), while the cellular uniformity significantly increased with age (r = 0.190, P < 0.05) ( Figure 4(B) ). Facial skin aging was also significantly correlated with decreased hydration ( Figure 4(C) ) and increased TEWL ( Figure 4(D) ). Example images of facial skin and tomographic keratinocyte architecture at younger, middle, and older ages are shown in Figure 5 .

As has been reported previously [3-9], facial skin aging was significantly associated with rough texture ( Figure 6(A) ), wrinkles ( Figure 6(B) ), pore dilation ( Figure 6(C) ), and decreased L value ( Figure 6(D) ).

Notably, the cellular density showed significant correlations with skin hydration ( Figure 7(A) ), TEWL ( Figure 7(B) ), texture ( Figure 7(C) ), wrinkles ( Figure 7(D) ), pores ( Figure 7(E) ), and L value ( Figure 7(F) ).

The cellular uniformity was also significantly associated with each aging parameter, namely, skin hydration ( Figure 8(A) ), TEWL ( Figure 8(B) ), texture ( Figure 8(C) ), wrinkles ( Figure 8(D) ), pores ( Figure 8(E) ), and L value ( Figure 8(F) ).

The highly significant relationship among visual aging parameters and tomographic architecture allowed us to estimate the in vivo cellular architecture from

visual parameters. Using regression analysis, we established the following formulae:

Estimated keratinocyte density (Number of cells/space) = −13.3002 × Texture Score + 363.6159 × Pore Score − 78.3089 × Wrinkle score − 0.005611 × L (lightness) score + 223.5498

Estimated keratinocyte uniformity = 0.1444 × Texture Score + 3.7958 × Pore Score + 7.4499 × Wrinkle score − 0.03147 × L (lightness) score + 2.6077

According to the formulae, the estimated keratinocyte density and estimated keratinocyte uniformity showed highly significant correlations with architectural density and uniformity, respectively ( Figure 9 ).

Skin cell interconnectivity parameter

Cell density and cell uniformity were also merged into a single parameter that represents tomographic architecture, which we call skin cell interconnectivity. This parameter was calculated by adding the fit-scale from 0 to 1 of cell density and the reversed fit-scale from 0 to 1 of cell uniformity from the database established in this study. In this way, a higher score of the skin cell interconnectivity parameter represents higher cell density and more uniform cells, in terms of the tomographic cell architecture. There was a high positive correlation between cell density and cell interconnectivity, and a high negative correlation between cell uniformity and cell interconnectivity ( Figure 10(A) , Figure 10(B) ). Skin cell interconnectivity was also estimated from visible skin aging parameters by using regression analysis, which demonstrated a highly significant correlation ( Figure 10(C) ).

We next examined whether treatment with SK-II improves the architectural aging. Compared with that in the untreated control area in the right forearm, the application of SK-II on the left forearm significantly improved skin hydration ( Figure 11(A) ) and TEWL ( Figure 11(B) ), even after 2 weeks of treatment. In parallel with the improvement of skin moisture, SK-II significantly increased the cellular density and decreased the cellular uniformity after 2 and 4 weeks of application, compared with the findings in the untreated control area ( Figure 11(C) , Figure 11(D) , and Figure 12 ).

As SK-II rejuvenated the architectural keratinocyte aging in the forearm as early as at 2 weeks of daily application, we next examined whether its treatment reverses the facial aging parameters earlier than 2 weeks. SK-II was applied twice daily on the face in 35 women and the facial aging parameters were assessed at baseline (day 0) and on 3, 7, and 14 days after its application. As shown in Table 1 , the treatment with SK-II significantly improved all facial skin aging parameters: hydration, TEWL, texture, wrinkles, pores, and L value, compared with the findings at baseline, as early as at day 3. The rejuvenating effects of SK-II became even more pronounced at days 7 and 14 compared with those at day 3 ( Table 1 ).

In addition, the estimated keratinocyte density, estimated keratinocyte uniformity and estimated skin cell interconnectivity were significantly and time-dependently rejuvenated by daily treatment with SK-II ( Figure 13 and Figure 14 ).

*: Significant difference from baseline (P < 0.05), †: Significant difference from day 3 (P < 0.05).