Objective Evaluation of the Degree of Pigmentation in Human Induced Pluripotent Stem Cell–Derived RPE

Abstract

Purpose.: For the transplantation of human induced pluripotent stem cell–derived retinal pigment epithelium (hiPSC-RPE), determination of the maturation status of these cells is essential, and the degree of pigmentation (dPG) can serve as a good indicator of this status. The aim of this study was to establish a method of objectively and quantitatively evaluating the dPG of hiPSC-RPE.

Methods.: Two observers determined the dPG subjectively by observing recorded images of hiPSC-RPE as follows: the dPG of a single cell was classified into three different pigmentation stages, and the overall dPG was compared between two cell groups to identify the group with the higher dPG. The κ statistic was applied to assess interobserver reproducibility. Next, the dPG of single cells and cell groups was objectively determined by the lightness of the hue, saturation, and value (HSL) color space, and the correlation between the subjective evaluation and time-dependent change in the objective dPG of hiPSC-RPE was investigated.

Results.: The κ statistic was 0.88 and 0.81 in the single-cell and cell-group observations, respectively. The objective dPG of single cells and cell groups was highly correlated with the subjective dPG. However, the observers were occasionally unable to subjectively determine the group with the higher dPG. The objective dPG increased in a time-dependent manner.

Conclusions.: The lightness of the HSL color space can be used to objectively and quantitatively evaluate the dPG of hiPSC-RPE in culture. The objective evaluation was consistent and was able to better identify small differences than subjective evaluation.

Introduction

Age-related macular degeneration¹ (AMD) is a leading cause of severe visual loss in developed countries, and the age-related degeneration of the retinal pigment epithelium (RPE) is considered an essential contributing factor to the disease pathogenesis. The RPE, a monolayer of pigmented cells, lies between the neural retina and the choroid and plays essential roles in the maintenance of photoreceptors² and choroidal vessels. Antivascular endothelial growth factor (VEGF) therapies have significant effects on wet AMD patients; however, recurrent cases, nonresponders, and dry AMD patients do not gain much benefit from current therapies. Retinal pigment epithelium transplantation for AMD for the purpose of replacing degenerated RPE with healthy RPE was introduced more than 2 decades ago and has attracted attention as an alternative therapy in recent years. To date, transplantation of allogeneic fetal RPE³ or autologous peripheral RPE⁴ has been reported as clinical therapy for AMD patients, although neither represents an ideal tissue source; the former has caused immune rejection, while the latter requires invasive surgery to harvest peripheral RPE. Therefore, RPE derived from human induced pluripotent stem cells (hiPSCs)⁵ has emerged as an ideal alternative cell source to overcome the disadvantages of past transplantation strategies. 

We previously reported that we could successfully differentiate functional RPE from hiPSCs.^6–8 During the induction of RPE from pluripotent stem cells, pigmented colonies containing melanin emerge, and a previous report⁹ has shown typical RPE markers to be coincident with the appearance of pigmentation. The formation of melanin in RPE is initiated early in fetal development and terminates by approximately 2 years¹⁰; therefore, the degree of pigmentation (dPG) can serve as a good indicator of maturation status in RPE derived from pluripotent stem cells. In the measurement of the dPG of RPE from pluripotent stem cells, subjective classification^9,11 of the dPG of RPE in culture has been reported; however, it neither has good reproducibility nor classifies subtle differences of the dPG. Therefore, the present study was initiated to ascertain whether we can quantify the dPG evaluation from human subjects' visual responses to the digital image of hiPSC-derived retinal pigment epithelium (hiPSC-RPE). 

The objective evaluation of the digital image, by using the hue, saturation, and value (HSV)¹² and the hue, saturation, and lightness (HSL)¹³ color spaces to rearrange the red–green–blue (RGB) values in an attempt to make them more perceptual, is widely accepted. We hypothesized that the dPG of hiPSC-RPE could be objectively quantified by using two representations. Here, we report a method to objectively and quantitatively evaluate the dPG of hiPSC-RPE and we sequentially investigated the dPG of hiPSC-RPE in culture. Additionally, we evaluated whether there was any association between the objective dPG and RPE function. 

Methods

Culture of hiPSCs and hiPSC-RPE

The hiPSC cell lines 253G1¹⁴ and 454E2,¹⁵ derived from healthy human dermal fibroblast cells by using three transcription factors (Oct3/4, Sox2, and Klf4) and dental pulp cells by using six transcription factors (Oct3/4, Sox2, Klf4, L-Myc, Lin28, and p53), were supplied from RIKEN BioResource Center (Ibaraki, Japan). The methods used for hiPSC maintenance and differentiation were as previously described.⁸ Human induced PSC-RPE was cultured in CELLstart (GIBCO, Carlsbad, CA, USA)–coated dishes in preconfluent medium (F10 [Sigma-Aldrich Corp., St. Louis, MO, USA] and 10% fetal bovine serum) before reaching confluence and in postconfluent medium (Dulbecco's modified Eagle's medium [DMEM]/F12 [7:3] supplemented with B27 [Invitrogen, Carlsbad, CA, USA], 2 mM l-glutamine [Sigma], 10 ng/mL basic fibroblast growth factor [Wako, Osaka, Osaka, Japan], and SB431542 [0.5 μM; Sigma]) after reaching confluence. The medium was changed every 2 to 3 days. 

Evaluation of the dPG in hiPSC-RPE

Human induced PSC-RPE was washed with phosphate-buffered saline, and the medium was changed to DMEM, phenol red–free (Sigma-Aldrich Corp.). Human induced PSC-RPE in culture was imaged by using a laser scanning confocal microscope (IX81; Olympus, Tokyo, Japan), charge-coupled device camera (DP10; Olympus), and digital imaging software (cellSens standard; Olympus). The bright-field RGB images (2048 × 1536 pixels) were captured under the same conditions (objective lens: Olympus UPlanFL 10× N.A. 0.30, condenser aperture diaphragm: 80%, field diaphragm: 100%, neutral ND25/ND6 filter, exposure time: 1/23 seconds, and ISO: 200) and were imported into digital imaging software (Adobe Photoshop CS2; Adobe Systems, Inc., San Jose, CA, USA). 

The dPG was subjectively determined from the imported images by two independent observers (observers I and II) as follows: the dPG in a single cell was classified into three pigmentation categories (low, medium, and high pigmentation; Fig. 1A), and the overall dPG was compared between two cell groups in a 10 × 10 field (images I and II) to determine which image indicated a higher dPG (Fig. 1B). 

Figure 1

 

The subjective classification of the dPG of hiPSC-RPE. (A) Imported bright-field image of hiPSC-RPE in DMEM (phenol red–free) in a 12-well plate (passage 4, 12-well plate). Two observers classified single cells into three different pigmentation stages (low, medium, and high pigmentation; left panels) and assessed interobserver reproducibility (right table). (B) Imported bright-field image of hiPSC-RPE in DMEM (phenol red–free) in a 12-well plate. Two observers compared two cell groups (images I and II) to identify the group with the higher dPG (left panels). Typical examples of consistent, inconsistent, and indistinguishable results are shown in image sets 1, 2, and 3, respectively (objective dPG of image set 1: image I 0.362, image II 0.227; set 2: image I 0.148, image II 0.146; set 3: image I 0.363, image II 0.353). The correlation between the subjective and objective dPG of hiPSC-RPE is shown in the right table.

Then, the dPGs of the imported images were objectively determined as follows: dots were sampled from the hiPSC-RPE of three different subjectively classified stages (Fig. 2A, black arrowhead) by hovering the computer mouse over a subjectively classified hiPSC-RPE dot and looking at Photoshop's Info Palette to examine the RGB value. We calculated these dots to rearrange the RGB value on the basis of the HSV and HSL color space according to the equation provided in the public domain by Wikipedia (http://en.wikipedia.org/wiki/HSL_and_HSV; November 17, 2014). For single-cell dPG analysis, an hiPSC-RPE was manually outlined, and the mean RGB value of the enclosed cell was used (Fig. 2B). For cell-group dPG analysis, the mean RGB value of an hiPSC-RPE image was used (Fig. 2C). The objective dPG was defined as ranging from 0 to 1 and the difference between the sample value and the blank value. In the serial objective analysis of dPG changes, four images per sample were recorded, and the mean RGB value of the four images was used. For the other cell-group dPG analysis, we measured the RGB value by using the light transmittance at different wavelengths (wavelength: blue 436 nm, green 546 nm, red 700 nm) by an absorption spectrometer (Varioskan Flash Multimode Reader; Thermo, Waltham, MA, USA), and the objective dPG was defined as ranging from 0% to 100% and blank was set as 100%. 

Figure 2

 

The dPG in hiPSC-RPE. (A) Imported bright-field RGB images of hiPSC-RPE in DMEM (phenol red–free). To examine the RGB value of the dots sampled from the hiPSC-RPE, we hovered the computer mouse over a subjectively classified dot of hiPSC-RPE and looked at Photoshop's Info Palette (red: 75, green: 52, blue 59). (B) Imported bright-field RGB images of hiPSC-RPE in DMEM (phenol red–free). To examine the mean RGB value of a single cell, we outlined a subjectively classified single cell in the hiPSC-RPE and analyzed it by using Photoshop's Info Palette (red: 89.33, green: 73.83, blue: 82.97). (C) Imported bright-field RGB images of hiPSC-RPE in DMEM (phenol red–free). To examine the mean RGB value of the cell group, we outlined the overall image of the hiPSC-RPE and assessed it by using Photoshop's Info Palette (red: 119.86, green: 101.83, blue: 91.48).

Statistical Analysis

The values were expressed as the mean ± SEM, with P < 0.05 considered statistically significant and the asterisks (*) indicating P < 0.01. The suitable temperatures for the preservation of hiPSC-RPE, the efficacy of hiPSC-RPE transplantation, the objective dPG of each hiPSC-RPE classified subjectively, and the objective dPG of hiPSC-RPE at each time point were analyzed by one-way analysis of variance followed by Scheffe's test. The RGB color value and HSL lightness of identical hiPSC-RPE images recorded with a 1-day interval were evaluated by using paired t-tests. The correlation between the subjective and objective dPG of hiPSC-RPE and time-dependent change in the objective dPG of hiPSC-RPE were assessed by Spearman's rank correlation coefficient. The interobserver reproducibility was calculated by Cohen's κ statistic. 

Results

Characterization of hiPSC-RPE In Vitro and In Vivo

Using the previously described method,⁸ we induced RPE from hiPSCs (Supplementary Figs. S1A–E). The pigmented cells had the structural characteristics of RPE and expressed typical RPE markers in vitro. To assess the function of hiPSC-RPE in vivo, we compared nontransplantation, sham surgery, and pigmented cell transplantation into the subretinal space in pink-eyed Royal College of Surgeons dystrophic rats, an animal model of inherited retinal degeneration,¹⁶ to measure the outer nuclear layer (ONL) thickness and full retinal thickness (fRT). First, we assessed the suitable temperature for preservation of the pigmented cells until the time of transplantation to measure percentage viability of the cells sequentially. Room temperature consistently yielded the best percentage viability of the cells among three tested temperatures for up to 12 hours; hence, we preserved the pigmented cells at room temperature until transplantation. The ONL thickness and fRT in the transplanted eyes were significantly greater than those in the nontransplanted and sham-surgery groups (Supplementary Figs. S2A–H). 

Evaluation of the Degree of Pigmentation in hiPSC-RPE

In the bright-field images of hiPSC-RPE (253G1) in culture, two independent observers determined the dPG subjectively by observation of hiPSC-RPE images as follows: the dPG in a single cell was classified into three different stages (low, medium, and high pigmentation; Fig. 1A), and the overall dPG was compared between two cell groups (images I and II) to identify the group with the higher dPG (Fig. 1B). The distribution of the subjective classification of the dPG of single cells by the individual observers is shown in the Table (n = 128). Unanimity between the two observers was observed in 118 sample cells (both observers classified the dPG as follows: 36 low, 52 medium, and 30 high pigmentation sample cells), and disagreement was found for 10 sample cells. The κ statistic, indicating the interobserver reproducibility, was 0.88 (P < 0.01) for the single-cell observations. In the comparison of two cell groups, two independent observers subjectively chose the group with the higher dPG from either image I or II, and all the image sets were classified in a 3 × 3 table as shown in the Table (n = 50). Uniform judgment between the two observers was obtained for 44 image sets (both observers chose the higher dPG cell group as follows: 15 image I, 23 image II, and 6 indistinguishable image sets), and disagreement was found in six image sets. The κ statistic was 0.81 (P < 0.01) in the cell-group observation. Therefore, the subjective dPG evaluation was a fair indicator of the dPG classification. 

Table

 

Evaluation of the Subjective dPG in hiPSC-RPE

n*	Observer I
	Low	Medium	High
Observer II
Low	36	2	0
Medium	0	52	2
High	0	6	30
n†	Observer I
	Image I	×	Image II
Observer II
Image I	15	1	0
×	2	6	1
Image II	1	1	23
n‡	Subjective Evaluation
	A	×	B
Objective evaluation
A	17	3	1
×	0	0	0
B	0	5	24

 ×, indistinguishable.

*  The distribution of subjectively classified single-cell dPGs by individual observers. Unanimity between the two observers was observed in 118 sample cells (both observers classified the dPG as follows: 36 low, 52 medium, and 30 high pigmentation), and disagreement occurred for 10 sample cells. The κ statistic was 0.88 (95% confidence interval [CI], 0.81–0.95; P < 0.01).

†  The distribution of subjective selections of the higher dPG cell group by two independent observers. Uniform judgment between the two observers was obtained for 44 image sets (both observers chose the higher dPG cell group as follows: 15 image I, 23 image II, and 6 undistinguishable image sets), and disagreement occurred for 6 image sets. The κ statistic was 0.81 (95% CI, 0.66–0.95; P < 0.01).

‡  The distribution of the correlation between the subjective and objective dPG of cell groups. The evaluation between the subjective and objective dPG was consistent in 41 image sets, inconsistent in 1 image set, and indistinguishable in 8 image sets (κ statistic 0.68; 95% CI, 0.52–0.85; P < 0.01).

To evaluate the objective dPG, we chose dots in the hiPSC-RPE that were subjectively classified into three different stages (low, medium, and high pigmentation) and evaluated the RGB value of these dots. Subsequently, we assessed these dots to rearrange the RGB value on the basis of HSV and HSL color spaces, and we evaluated the correlation with the subjective dPG. The lightness of the HSL color space was best correlated with the subjective dPG; hence, we evaluated the objective dPG by using the lightness of the HSL color system (Fig. 3A). In addition, to investigate the reproducibility of the dPG of identical hiPSC-RPE, we recorded hiPSC-RPE images of the same sample with a 1-day interval (Fig. 3B) and compared the RGB value and the lightness of the HSL color space. No significant differences were observed in the numerical values (Fig. 3C). 

Figure 3

 

Evaluation of several objective methods for determining the dPG of hiPSC-RPE. (A) The correlation of the objective dPG value based on the HSV and HSL color spaces with the subjectively classified dots (low, medium, and high pigmentation; n = 50 for each) sampled from hiPSC-RPE. The lightness of the HSL color space is best correlated with the subjective dPG (HSV: hue 0.55, saturation 0.85, and value 0.70; HSL: hue 0.55, saturation 0.58, and lightness 0.92; Spearman rank correlation coefficient). (B) Recording of bright-field hiPSC-RPE images with a 1-day interval (D1: day 1, D2: day 2). (C) The red, green, and blue signals of the RGB color model and the lightness of the HSL color space of recorded hiPSC-RPE images with a 1-day interval (D1: day 1, D2: day 2; n = 12 for each).

Next, we evaluated the correlation between the subjective and objective dPG of single cells and cell groups assessed by a single observer. In the single-cell evaluation (Fig. 4A, n = 128), the objective dPG was highly correlated with the subjective dPG and the dPG values were significantly different between the subjectively classified groups. In the cell-group evaluation (Fig. 4B, n = 50), the evaluation between the subjective and objective dPG was consistent in 41 image sets, inconsistent in 1 image set, and undistinguishable in 8 image sets (Table). Specifically, the subjective and objective dPGs were 100% consistent when the objective dPG difference was above 0.031 (a typical example is seen in image set 1 of Fig. 1B). Conversely, the subjective judgment was inconsistent (set 2 of Fig. 1B) with dPG or indistinguishable (set 3 of Fig. 1B) in 9/12 of the image sets when the dPG difference was less than 0.031. These results suggest that the objective determination, based on the lightness of the HSL color space, can be used to quantitatively evaluate the dPG of hiPSC-RPE in culture, and the objective determination was basically consistent with subjective evaluation but could better distinguish between images with similar dPGs. 

Figure 4

 

The correlation between the subjective and objective dPG of hiPSC-RPE. (A) The correlation between the subjective and objective dPG of single cells. The scatter diagram (left) shows that the objective dPG was highly correlated with the subjective dPG (ρ = 0.94, P < 0.01), and the bar graph (right) shows that the dPG values were significantly different between subjectively classified groups (low: 0.13 ± 0.00; medium: 0.38 ± 0.00; high: 0.62 ± 0.01; n = 50 for each). (B) A scatterplot comparing the dPG in hiPSC-RPE of images I and II. Consistency between the subjective and objective results was observed in 41 cases (black points); inconsistency was observed in 1 case (red point); and indistinguishable results were obtained in 8 cases (orange points). The dotted line shows that the difference in the dPG between images I and II was 0.075.

Time Course of the Objective dPG in hiPSC-RPE

We investigated whether the objective dPG was associated with the maturation of hiPSC-RPE. Bright-field images of hiPSC-RPE were recorded sequentially (0, 2, 4, and 6 weeks after reaching confluence) and used to objectively assess the dPG in cell groups (Fig. 5A). The objective dPG increased in a time-dependent manner (Fig. 5E, black line). We also obtained consistent results with another hiPSC-RPE (Fig. 5B) that was differentiated from a different hiPSC line (Fig. 5E, black dotted line). Next, we evaluated the objective dPG in hiPSC-RPE grown on the insert membrane in a 12-well Transwell plate (manufacturer, city, state, country) in two different media: pre- and postconfluent medium (Fig. 5C, postconfluent medium [same as Figs. 5A, 5B; black dotted line]; Fig. 5D, we continued to culture hiPSC-RPE in preconfluent medium after confluence was reached; red line). The objective dPGs increased in a time-dependent manner (Fig. 5F). The changes in the degree of objective dPG in postconfluent medium were comparable (Fig. 5G). In contrast, the changes in the degree of objective dPG in the preconfluent medium were lower than those in the cells in postconfluent medium (Fig. 5G). There were no differences in cell proliferation status between hiPSC-RPEs in pre- and postconfluent medium; however, the expression of RPE markers showed significant differences (Supplementary Fig. S3). 

Figure 5

 

Time course of the objective dPG in hiPSC-RPE grown in a 12-well plate and a 12-well Transwell insert. (A) Bright-field images of hiPSC-RPE (passage 4, 12-well plate) were recorded sequentially. (B) Bright-field images of the other hiPSC-RPE (passage 4, 12-well plate) were recorded sequentially. (C) Bright-field images of hiPSC-RPE in maturation medium (passage 4, 12-well Transwell plate) were recorded sequentially. (D) Bright-field images of hiPSC-RPE in proliferation medium (passage 4, 12-well Transwell plate) were recorded sequentially. (E) Time course of the objective dPG of two hiPSC-RPEs grown in a 12-well plate. Both objectively determined dPG values increased in a time-dependent manner ([A] black line, ρ = 0.94, P < 0.01; [B] black dotted line, ρ = 0.97, P < 0.01). (F) Time course of the objective dPG of hiPSC-RPE grown in a 12-well Transwell insert in two different media: pre- and postconfluent medium. Both objectively determined dPG values increased in a time-dependent manner ([C] postconfluent medium, black dotted line, ρ = 0.94, P < 0.01; [D] preconfluent medium, red line, ρ = 0.85, P < 0.01). (G) The changes in the degree of objective dPG in postconfluent medium were comparable and the changes in the degree of objective dPG in proliferation medium were lower than in the others ([A] y = 0.078x; [B] y = 0.076x; [C] y = 0.073x; [D] y = 0.045x). (H) The secretion of VEGF and PEDF in two hiPSC-RPEs grown in a 12-well plate was similar ([A] black line; [B] black dotted line). (I) Vascular endothelial growth factor and PEDF secretion and TER of hiPSC-RPE grown in a 12-well Transwell insert increased in a time-dependent manner, and maturation medium induced high VEGF and PEDF secretion and TER ([C] black dotted line; [D] red line). (J) The objective dPG of hiPSC-RPE and RPE function were highly correlated ([C] black point; [D] red point, VEGF: ρ = 0.86, P < 0.01, PEDF: ρ = 0.84, P < 0.01, TER: ρ = 0.92, P < 0.01).

We evaluated whether there was any association between RPE function and color, based on the objective dPG assessment. Retinal pigment epithelium is known to secrete a range of growth factors, including VEGF and pigment epithelium–derived factor (PEDF); and barrier function can be measured by transepithelial electrical resistance¹⁷ (TER), which provides a simple and sensitive method for the detection of functional tight junctions. Using an enzyme-linked immunosorbent assay (ELISA), we measured concentrations of VEGF and PEDF in conditioned hiPSC-RPE medium for each time point. As a result, the secretion of VEGF and PEDF in both hiPSC-RPEs on the 12-well plate showed a similar change with time, although it did not exhibit a time-dependent increase and did not correlate with the objective dPG (Fig. 5H). In contrast, VEGF and PEDF secretion and the TER of hiPSC-RPE in the 12-well Transwell insert increased in a time-dependent manner, and the postconfluent medium induced a high secretion of VEGF and PEDF and a high TER (Fig. 5I). Furthermore, even under different culture media, the objective dPG of hiPSC-RPE and RPE function were highly correlated (Fig. 5J). 

Time Course of Two Different Objective dPG Analyses in hiPSC-RPE

To further improve reproducibility of the objective dPG analysis, we measured the RGB value by using the light transmittance at different wavelengths (red, green, and blue). In the objective dPG of two different hiPSC-RPEs (Fig. 6A: 253G2, Fig. 6B: 454E2), the objective dPG using image analysis increased in a time-dependent manner (Figs. 6A, 6B; right panels) and the objective dPG using the light transmittance decreased in a time-dependent manner (Figs. 6A, 6B; center panels). We also obtained consistent results with hiPSC-RPE (454E2) in the 12-well Transwell insert (Fig. 6C). Two different objective dPG analyses were highly correlated (Figs. 6A–C, right panels) and the objective dPG using the light transmittance had consistently less variation. Additionally, the objective dPG of hiPSC-RPE (454E2) using the light transmittance and RPE function were highly correlated (Figs. 6D, 6E); and the amount of melanin and Ezrin expression as the functional RPE marker¹⁸ in hiPSC-RPE increased in a time-dependent manner (Supplementary Figs. 4A, 4B). In contrast, autofluorescence in hiPSC-RPE 2w and 6w after reaching confluence did not show significant differences (Supplementary Fig. 4B). 

Figure 6

 

Time course of two different objective dPG evaluations in hiPSC-RPE. (A–C) Time course of two different objective dPG evaluations of hiPSC-RPE in a 12-well plate ([A] 253GI; [B] 454E2, passage 4) and in a 12-well Transwell insert ([C] 454E2, passage 4). All objectively determined dPG values changed in a time-dependent manner (left panels: image analysis; [A] ρ = 0.96, P < 0.01; [B] ρ = 0.89, P < 0.01; [C] ρ = 0.96, P < 0.01) (center panels: transmittance; [A] ρ = 0.97, P < 0.01; [B] ρ = 0.93, P < 0.01; [C] ρ = 0.97, P < 0.01) and were highly correlated (right panels; [A] ρ = 0.94, P < 0.01; [B] ρ = 0.94, P < 0.01; [C] ρ = 0.91, P < 0.01). (D) Vascular endothelial growth factor and PEDF secretion and TER of hiPSC-RPE (454E2) grown in a 12-well Transwell insert increased in a time-dependent manner. (E) The transmittance of hiPSC-RPE (454E2) and RPE function were highly correlated (VEGF: ρ = 0.84, P < 0.01; PEDF: ρ = 0.89, P < 0.01; TER: ρ = 0.93, P < 0.01).

Discussion

Current therapies for AMD show only limited efficacy; in recent years, increasing attention has been given to RPE transplantation using pluripotent stem cells, and transplantation of human embryonic stem cell-derived RPE (hESC-RPE) suspension in patients with dry AMD and Stargardt's disease is being tested in a clinical trial.¹⁹ Despite the considerable concern about use of hESCs or hiPSCs as a cell resource, one reason that transplantation of RPE derived from pluripotent stem cells has been placed on the fast track for clinical application is the appearance of pigmentation in RPE during differentiation, which is helpful in the identification and purification of these cells. The melanin granules representing RPE pigmentation have significant functions, such as neural retina protection by scavenging free radicals (from UV and biochemical sources²⁰) and absorption of visible light to aid in minimizing light reflection and scattering.²¹ In embryology, the formation of melanin in the RPE is initiated early in fetal development and terminates by approximately 2 years after birth. During the induction of RPE from pluripotent stem cells, the appearance of pigmentation in RPE coincides with the expression of typical RPE markers in previous reports,^9,11 and we showed that the amount of melanin and expression of typical RPE markers in hiPSC-RPE increases in a time-dependent manner. Therefore, the dPG of RPE can serve as a good indicator of maturation status in RPE derived from pluripotent stem cells. In this study, we first confirmed that the hiPSC-RPE showed characteristics similar to native RPE in vitro and in vivo and developed a method of objectively and quantitatively evaluating the dPG of differentiated hiPSC-RPE. We were able to systematically quantify human subjective visual responses to the dPG of hiPSC-RPE and evaluate the dPG of hiPSC-RPE in culture sequentially on the basis of lightness of the HSL color space. Our results suggest that this method is a useful preoperative evaluation method before hiPSC-RPE transplantation. 

In the classification of the dPG of RPE, the subjective dPG of RPE classified into two (nonpigmentation or pigmentation) or three (low, medium, or high pigmentation) different groups has been reported,^9,11 which we also confirmed in our study; the interobserver reproducibility of the subjective dPG of hiPSC-RPE was excellent, demonstrating that the subjective dPG of hiPSC-RPE was a good indicator. However, in an average interval of 2 weeks, the average difference in the objective dPG of hiPSC-RPE in postconfluent medium was 0.075 (Fig. 4B, dotted line), and the inconsistent or indistinguishable ratio was 9/17 (53%), indicating that the subjective judgment was inconsistent when distinguishing between two cell groups with a similar dPG; therefore, we attempted to objectively and quantitatively evaluate the digital image of dPG of hiPSC-RPE. It is widely accepted that the HSL and HSV color spaces, in which the RGB values are rearranged in an attempt to make them more perceptual, can serve as an objective evaluation method for human subjective visual responses to digital image color. The HSL and HSV color spaces use three color dimensions, namely, hue, saturation, and lightness or value. Hue defines the pure color or mixtures of two pure colors such as “red” or “red–yellow” and ranges from 0° to 359° (red–yellow–green–cyan–blue–magenta–red). Saturation defines the purity of a color and ranges from 0 (gray) to 1 (pure color). Lightness or value defines the brightness of a color and ranges from 0 (black) to 1 (white or pure color). These dimensions are derived from the Munsell color system,²² which defines objective values of human subjects' visual responses to color on the basis of scientific evidence. We assessed each hiPSC-RPE that was subjectively classified by the HSL and HSV color spaces and showed that the lightness of the HSL color space is best correlated with the subjective dPG of hiPSC-RPE; hence, we applied the lightness of the HSL color system to the evaluation method of the objective dPG of hiPSC-RPE. The lightness of the HSL color system varies from 0 (black) to 1 (white) and hiPSC-RPE color changes from white to black (dark brown); therefore, we considered this method as reasonable. In this study, to minimize the effect of the state of the medium and the color of the dish itself, we used phenol red–free medium and defined the objective dPG of hiPSC-RPE as the difference between the sample value and the blank value. Consequently, the objective dPG of hiPSC-RPE increased as the hiPSC-RPE became blacker, while the lightness of the HSL color space decreased as the color became darker. We lastly evaluated the objective dPG of hiPSC-RPE to measure the RGB values not from image analysis but from light transmittance at different wavelengths by absorption spectrometry, which may be easily reproduced in other laboratories. This other analysis was highly correlated with image analysis in addition to being able to measure easier, shorter, and less variation. These two evaluation methods could assess hiPSC-RPE as graft in culture; therefore, it can be used to determine the optimal timing of hiPSC-RPE transplantation by assessing the transplantation efficacy, such as neuroprotection and survival rate, after transplantation. We showed the objective dPG evaluation from the image analysis with only a single microscopic system; hence, it cannot be immediately asserted that this image analysis is applicable to all other labs. We hope that the evaluation method using light transmittance will be applicable in other laboratories. 

One major interest in this field is whether the dPG correlates with RPE function. The dPG of both hiPSC-RPEs grown on a Transwell insert, which separates the apical and basal sides, was considered close to the physiological RPE condition and was correlated with VEGF and PEDF secretion and the barrier function. A lower dPG was associated with a lower secretion of these factors. This result may imply that the depigmentation of RPE in the fundus, which is an age-related maculopathy as a disorder of the macular area, may also be correlated with the secretion of VEGF or PEDF and the barrier function by those cells in vivo. This further indicates that if advances in clinical imaging techniques—such as two-photon microscopy to image the RPE itself directly and noninvasively²³—can enable us to compare dPG in vivo, we could estimate the function of the RPE as well. 

In summary, we established a method of objectively and quantitatively evaluating the dPG of hiPSC-RPE. With our method, it is possible to quantify the dPG of hiPSC-RPE in culture sequentially without consuming the sample and to evaluate the graft itself before RPE transplantation. This method is also simple, rapid, and inexpensive. We hope that this evaluation method will provide additional information on the correlation between RPE pigmentation and function in future studies. 

Acknowledgments

Supported by Grant 24GS-6 from Kawasaki Medical School, the KAWASAKI Foundation for Medical Science & Medical Welfare, and an academic contribution from Pfizer Japan, Inc. 

Disclosure: H. Kamao, None; M. Mandai, None; S. Wakamiya, None; J. Ishida, None; K. Goto, None; T. Ono, None; T. Suda, None; M. Takahashi, None; J. Kiryu, None