2024 Volume 47 Issue 6 Pages 1163-1171
The vital role of bile canaliculus (BC) in liver function is closely related to its morphology. Electron microscopy has contributed to understanding BC morphology; however, its invasiveness limits its use in living specimens. Here, we report non-invasive characterization of BC formation using refractive index (RI) tomography. First, we investigated and characterized the RI distribution of BCs in two-dimensional (2D) cultured HepG2 cells. BCs were identified based on their distinct morphology and functionality, as confirmed using a fluorescence-labeled bile acid analog. The RI distribution of BCs exhibited three common features: (1) luminal spaces with a low RI between adjacent hepatocytes; (2) luminal spaces surrounded by a membranous structure with a high RI; and (3) multiple microvillus structures with a high RI within the lumen. Second, we demonstrated the characterization of BC structures in a three-dimensional (3D) culture model, which is more relevant to the in vivo environment but more difficult to evaluate than 2D cultures. Various BC structures were identified inside HepG2 spheroids with the three features of RI distribution. Third, we conducted comparative analyses and found that the BC lumina of spheroids had higher circularity and lower RI standard deviation than 2D cultures. We also addressed comparison of BC and intracellular lumen-like structures within a HepG2 spheroid, and found that the BC lumina had higher RI and longer perimeter than intracellular lumen-like structures. Our demonstration of the non-destructive, label-free visualization and quantitative characterization of living BC structures will be a basis for various hepatological and pharmaceutical applications.
The bile canaliculus (BC) is the primary structure for bile duct formation, where bile acids synthesized in hepatocytes are secreted. Moreover, the BC is an important excretion route for hepatocytic metabolites, including administered drugs.1–3) The vital role of the BC in liver function is closely related to its morphology,1–5) explained as follows: (1) a luminal space is present between adjacent hepatocytes owing to their polarization; (2) the lumen is sealed by a specialized membrane that contains tight junctions, thus strictly separating the hepatocyte and BC regions, where various transporters exist to excrete metabolites from hepatocytes; and (3) the lumen-containing microvilli expand the surface area of the BC for efficient metabolite transfer.
The morphological characteristics of the BC have been exclusively identified and evaluated using electron microscopy (EM).1–5) Despite its resolution for the visualization of BC ultrastructures and contribution to the literature, the inherent invasiveness of EM restricts its use in evaluating living specimens.
The following two aspects are important in evaluating BC morphology in living samples. First, the three-dimensional (3D) architecture of the BC should be observed. High-resolution 3D imaging is effective for accurately capturing the entire BC morphology.5) Recently, multicellular liver models, including spheroids and organoids, have been used to mimic in vivo environments.4–11) However, accurately assessing the 3D architecture of the BCs within thick samples is challenging because the imaging depth is limited by multiple-scattering light originating from them.12,13) Second, intact samples should be evaluated to address the native behavior of BCs. For example, the BC undergoes cycles of expansion and contraction to secrete bile acids efficiently,14) suggesting the importance of addressing BC morphology in the context of its dynamics. Limited methodologies are currently available to address these important issues, particularly non-invasive approaches.
Refractive index (RI) tomography is a technique for the non-invasive visualization of cellular microstructures, and it has been rapidly developing in recent years.15) RI serves as a cellular intrinsic imaging contrast that reflects morphological and biophysical information.15,16) RI tomography can acquire the 3D distribution of RI with high resolution, and visualize subcellular structures, including chromosomes, mitochondria, and lipid droplets, in living cells.17,18) Although the observation target of RI tomography was basically limited to the single cell level, we recently developed an in-silico clearing approach for deep RI tomography, which suppresses multiple-scattering via corrective light propagation in a computer.13) This technique enabled the visualization of a living HepG2 spheroid with a diameter of 140 µm from inside to the surface at the organelle level.13) In summary, high-resolution RI tomography has been demonstrated mainly in 2D cultured and suspended cells.15,17,18) Recently, it has been applied to 3D cultured spheroids.12,13) However, it is not clear whether BC structures can be visualized in both 2D and 3D culture models using RI tomography.
The aim of this study was to demonstrate the non-invasive characterization of BC morphology in living specimens using RI tomography. The HepG2 cell line was selected as the observation target owing to its widespread use. HepG2 cells tend to lose their polarity in two-dimensional (2D) cultures, but their polarity is recoverable in 3D cultures.4,6–9) Therefore, both 2D and 3D cultured HepG2 cells were prepared. First, using 2D cultured HepG2 cells, the BC loci were pre-identified with a fluorescent reagent, and the cells were observed using RI tomography without in-silico clearing to clarify the characteristics of the RI distribution of BC morphology. Next, 3D cultured HepG2 spheroids were visualized using in-silico clearing RI tomography to characterize the BC morphology in a multicellular specimen. Furthermore, based on their RI maps, we performed quantitative and statistical analyses of BCs in 2D and 3D cultured HepG2 samples.
The human hepatoma HepG2 cell line (JCRB1054)19) was obtained from the Japanese Collection of Research Bioresources (JCRB) cell bank. The HepG2 cells were cultured in Dulbecco’s modified Eagle’s medium (Gibco, U.S.A.) supplemented with 10% (v/v) fetal bovine serum (Gibco), 100 U/mL penicillin, 100 µg/mL streptomycin, and 292 µg/mL L-glutamine at 37 °C in a humidified atmosphere of 5% CO2. To visualize 2D cultured HepG2 cells, the cells were seeded on a P35G-0-14-C glass-bottom dish (MatTek, U.S.A.) and cultured for 5 d. To identify the BC loci, the cells were incubated for 45 min in Hank’s balanced salt solution with calcium and magnesium (Gibco) containing cholyl-lysyl-fluorescein (CLF; Corning, U.S.A.), a fluorescent bile acid analog, at a final concentration of 2 µM.6) For comparative analyses, dimethyl sulfoxide (DMSO; Sigma, U.S.A.) was added to the medium at a final concentration of 0.05% after culturing HepG2 cells for 5 d, and the cells were further cultured for 1 d. To form the HepG2 spheroids, 1 × 105 trypsinized cells were seeded on a 3D cell culture container, an EZSPHERE 6-well plate (AGC Techno Glass, Japan) per well, and cultured for 3 d in 2 mL of culture medium. DMSO was then added to the medium at a final concentration of 0.05%, and the cells were further cultured for 1 d.
Fluorescence and Phase-Contrast MicroscopyFluorescence and phase-contrast images were acquired using a DM IL LED inverted microscope (Leica, Germany) with a ×20/0.30 numerical-aperture (NA) objective (HI PLAN I). The excitation and emission wavelengths for fluorescence imaging were 470 and 525 ± 25 nm, respectively.
RI TomographyAn off-axis holographic microscope with illumination at various incidence angles was used to obtain the RI maps of the samples.12,13) A helium-neon laser (632.8 nm) was split into two beams: a sample beam and a reference beam path. A 2D scanning micromirror device was used to modulate the incident angles of the illumination plane waves in the conjugate plane of the sample. The plane wave illuminated the sample via a high NA objective lens (Olympus, Japan, ×60/1.0 NA, water immersion). The diffracted wave from the sample was collected using an objective lens (Olympus, ×60/1.0 NA, water immersion). The sample and reference beams were combined using a beam splitter to generate holograms on a CMOS camera (JAI, Japan, GO-5100M-USB). In total, 333 holograms were recorded for each measurement. The amplitudes and phases of the holograms were retrieved using digital holographic processing. Subsequently, the RI maps for the 2D and 3D cultured samples were retrieved without and with in-silico clearing processing, respectively. Here, the in-silico clearing method comprised partial RI reconstruction and wave backpropagation,13) thus enabling unambiguous RI reconstruction of multiple-scattering thick samples. The in-silico clearing method is essential for the clear reconstruction of 3D cultured samples because of the strong multiple-scattering caused by them.13) The thickness of all reconstituted RI distribution images in the z-direction was 400 nm.
Statistical AnalysisThe BC luminal parameters were compared between 2D cultured HepG2 cells and HepG2 spheroids using an F test, followed by an unpaired t test using the Excel software. The luminal parameters were also compared between BC and intracellular lumen-like structures within a HepG2 spheroid using the same method. All p-values were two-sided, and results with p < 0.05 were considered significant.
CLF, a fluorescent bile acid analog with a behavior similar to that of cholyl glycine, was used in this study to identify the BC loci.6,10,20) Figure 1 shows the phase-contrast and fluorescence images of CLF-stained 2D cultured HepG2 cells. CLF accumulated in the region where luminal structures were observed between hepatocytes in the phase-contrast image, suggesting the formation of BCs (Fig. 1, arrows).
Two BC loci are indicated with arrows.
The same field of HepG2 cells was visualized using RI tomography without in-silico clearing. Figure 2A shows a representative cross-section (z = 6.4 µm), and Video S1 shows all cross-sections of the same sample as in Fig. 1. Figure 2B shows enlarged images of the dotted squares in Fig. 2A, which correspond to the two BC loci, with three representative cross-sections. The two BC structures exhibited common RI distribution features: (1) luminal spaces with a relatively low RI between adjacent hepatocytes; (2) luminal spaces surrounded by a membranous structure with a relatively high RI; and (3) multiple microvillus structures with a relatively high RI inside the lumina. In addition, the morphology of the two BC structures, identified via the RI distribution, was consistent with the ultrastructural features in electron micrograms reported in the literature,1–5) as mentioned in the Introduction.
(A) RI distribution of the representative cross-section of the same sample as in Fig. 1. Two BC loci are indicated with the dotted squares. (B) Enlarged views of the dotted squares (Areas 1 and 2) in (A). RI distribution of the three different cross-sections is provided for each area.
Next, we demonstrated morphological and biophysical analyses of a BC based on RI distribution using the image of Area 2 (z = 3.2 µm) shown in Fig. 2B (Fig. 3, Table 1). We obtained morphological parameters of the BC lumen, including area, perimeter, and circularity, along with biophysical parameters, such as mean, standard deviation, median, minimum, and maximum RI values (Table 1). We also identified microvillus structures in the lumen and estimated their numbers (Fig. 3B, Table 1). Furthermore, we obtained local RI information in the image (BC = 1.333, microvillus = 1.344, BC periphery = 1.359, and cell area = 1.344) in Fig. 3C.
Analyses of the BC of Area 2 (z = 3.2 µm) in Fig. 2. (A) Manually identified BC lumen area is surrounded by the dotted line. (B) Microvillus structures are indicated with arrows. (C) RI of the points indicated with arrows are shown. BC: bile canaliculus; MV: microvillus; peri: periphery.
| Sample | Area (µm2) | Perim (µm) | Circ | RI Mean | RI S.D. | RI Median | RI Min | RI Max | Identifiable MV# | In-silico clearing |
|---|---|---|---|---|---|---|---|---|---|---|
| Fig. 3. BC (a) | 26.2 | 19.5 | 0.86 | 1.336 | 0.004 | 1.333 | 1.333 | 1.349 | 14 | − |
| Fig. 5. BC (b) | 70.6 | 30.8 | 0.93 | 1.344 | 0.005 | 1.344 | 1.333 | 1.359 | 38 | + |
| Fig. 5. BC (c) | 73.5 | 31.5 | 0.93 | 1.344 | 0.004 | 1.344 | 1.334 | 1.357 | 34 | + |
Data related to the areas surrounded with the white dotted lines in Figs. 3 (BC a) and 5 (BC b and c) were analyzed using the ImageJ software. BC: bile canaliculus; Perim: perimeter; Circ: circularity; RI: refractive index; S.D.: standard deviation; Min: minimum; Max: maximum; MV#: microvillus number.
The 3D cultured HepG2 spheroids were visualized using RI tomography with in-silico clearing. Figure 4A shows the representative cross-sections of two HepG2 spheroids (z = 28.8 and 66.8 µm for Spheroid 1 and z = 28.4 and 40.8 µm for Spheroid 2). All cross-sections are provided in Videos S2 (Spheroid 1) and S3 (Spheroid 2). Figure 4B shows enlarged images of the dotted squares in Fig. 4A, with three representative cross-sections. Multiple BC structures were present inside the spheroids, with characteristics identical to those of 2D cultured HepG2 cells.
(A) RI distribution of the representative cross-sections of two HepG2 spheroids. In each cross-section, the representative BC structure is surrounded by a dotted square. (B) Enlarged views of the dotted squares (Areas 1–4) in (A). RI distribution of the three different cross-sections is provided for each area.
Subsequently, we analyzed the BC structures based on RI distribution using the images of Area 1 (z = 28.8 µm) and Area 2 (z = 66.8 µm) of Spheroid 1 shown in Fig. 4 (Fig. 5, Table 1). We obtained morphological and biophysical parameters similar to the demonstration with 2D cultured HepG2 cells. These representative BC structures had a higher number of microvillus structures than the 2D cultures (Fig. 5, Table 1). In addition, a circular structure with a relatively high RI was confirmed at the center of the BC structure (Fig. 5F).
(A–C) Analyses of the BC of Area 1 (z = 28.8 µm) in Fig. 4. (D–F) Analyses of the BC of Area 2 (z = 66.8 µm) in Fig. 4. (A, D) Manually identified BC lumen areas are surrounded by dotted lines. (B, E) Microvillus structures are indicated with arrows. (C, F) RI of the points indicated with arrows are shown. BC: bile canaliculus; MV: microvillus; peri: periphery. X represents an undefined structure.
To test whether RI tomography can be used to compare BC structures, we conducted statistical analyses of BC lumina between 2D cultured HepG2 cells and HepG2 spheroids.
First, to increase the number of RI maps on BC structures in the 2D cultures, three representative BC-rich fields were identified based on phase-contrast and CLF fluorescence images (Fields 1–3, Supplementary Fig. S1). The number of BC lumina was 20 in Field 1, 16 in Field 2, and 7 in Field 3. The fields were then visualized using RI tomography without in-silico clearing. Supplementary Figs. S2A, B, and C represent RI cross-sections near the center of all identified BC structures in Fields 1, 2, and 3, respectively. We manually segmented their BC luminal areas (surrounded by white lines in right columns of Supplementary Fig. S2) and obtained their morphological and biophysical parameters based on RI distribution (Supplementary Tables S1–S3). We used them for the comparative analyses as BC lumina in 2D cultures (n = 43).
Second, we prepared additional RI maps for two HepG2 spheroids (Spheroids 3 and 4), which were visualized using in-silico clearing RI tomography. All their cross-sections are provided in Videos S4 (Spheroid 3) and S5 (Spheroid 4). The cultivation method and experimental date of these spheroids were the same as those of Spheroid 1. BC structures with the three characteristics of RI distribution were picked up from the 3D maps of Spheroids 1, 3, and 4. The number of identifiable BC structures was 20 in Spheroid 1, 31 in Spheroid 3, and 42 in Spheroid 4. Supplementary Figs. S3A, B, and C represent RI cross-sections near the center of all identifiable BC structures in Spheroids 1, 3, and 4, respectively. We manually segmented their BC luminal areas (surrounded by white lines in right columns of Supplementary Figs. S3A–C) and obtained their morphological and biophysical parameters based on RI distribution (Supplementary Tables S4 for Spheroid 1, S5 for Spheroid 3, and S6 for Spheroid 4). We used them for the comparative analyses as BC lumina in spheroids (n = 93).
We then conducted statistical analyses for comparison of the BC lumina between 2D cultures and spheroids (Fig. 6). Due to the limitations described in the Discussion, it is difficult to directly compare absolute RI values, such as RI mean and median, between them. We selected morphological parameters (area, perimeter, and circularity) and a relative RI parameter (RI standard deviation) for the comparative analyses. The BC luminal areas and perimeters did not show significant differences between them (Figs. 6A, B). The circularity of the BC lumina in spheroids was significantly higher than that of the 2D cultures (p < 0.001, Fig. 6C). The RI standard deviation of the BC lumina in spheroids was significantly lower than that of the 2D cultures (p < 0.01, Fig. 6D).
Comparison of BC lumina between 2D cultured HepG2 cells (2D) and HepG2 spheroids (Sph). Boxplots to compare morphological and biophysical parameters of BC lumina are shown. (A) Area, (B) perimeter, (C) circularity, and (D) RI standard deviation (S.D.). The center lines represent median values, and the cross marks indicate average values. For the comparative analyses, we selected three representative fields of the 2D cultures (Fields 1–3). The total number of BC lumina in 2D cultures that we analyzed was 43. We also selected three representative HepG2 spheroids. The total number of BC lumina in the spheroids that we analyzed was 93. The BC luminal areas of 2D cultures and spheroids were manually segmented in the cross-sections near the center of the BC structures.
We observed lumen-like structures in HepG2 spheroids that lacked the three characteristic RI distribution features of BCs. These structures were basically localized intracellularly rather than intercellularly and did not have distinct microvillus structures (Videos S2–S5). We termed these structures as “intracellular lumen-like structures,” and they were prevalent in Spheroid 3. We picked up all identifiable intracellular lumen-like structures (n = 67) based on the 3D RI map of Spheroid 3. In addition, there were 12 lumen-like structures in Spheroid 3 whose localization could not be distinguished as intercellular or intracellular. Supplementary Fig. S4 shows RI cross-sections near the center of 67 identifiable intracellular lumen-like structures in Spheroid 3. We manually segmented their lumen-like areas (surrounded by white lines in right columns of Supplementary Fig. S4) and obtained their morphological and biophysical parameters based on RI distribution (Supplementary Table S7).
We then conducted statistical analyses for the comparison of BC (n = 31) and intracellular lumen-like structures (n = 67) within Spheroid 3 (Fig. 7). The luminal perimeters were significantly longer in the BC structures than in the intracellular lumen-like structures (p < 0.01, Fig. 7B). The luminal RI (mean and median) was significantly higher than that of the intracellular lumen-like structures (p < 0.001, Figs. 7D, F). The luminal areas, circularity, and RI standard deviation were not significantly different between them (Figs. 7A, C, E).
Comparison of BC lumina (BC) and intracellular lumen-like structures (Intra) in HepG2 Spheroid 3. Boxplots to compare morphological and biophysical parameters of the lumina are shown. (A) Area, (B) perimeter, (C) circularity, (D) RI mean, (E) RI standard deviation (S.D.), and (F) RI median. The center lines represent median values, and the cross marks indicate average values. The total number of BC structures subjected to the comparative analyses was 31, and that of intracellular lumen-like structures was 67. Their luminal areas were manually segmented in the cross-sections near the center of the structures.
RI tomography has been mainly used to visualize 2D cultured and suspended cells in various studies.15,17,18) Our group recently demonstrated its applicability to 3D cultured spheroids.12,13) However, the ability of RI tomography to assess the morphological details of BCs in hepatocyte models remains to be elucidated.
In this study, we propose a non-invasive evaluation method of BC formation in hepatocyte cultures using RI tomography in both 2D and 3D hepatocyte models. We identified three characteristic RI distribution features of BCs in living 2D cultured HepG2 cells (Figs. 2, 3). Based on these three features, we identified multiple BC structures in 3D cultured HepG2 spheroids and performed their quantitative (Figs. 4, 5) and statistical analyses (Fig. 6). The RI standard deviation of the BC lumina in the spheroids was significantly lower than that in the 2D cultures (Fig. 6D), suggesting a more homogeneous environment of BC lumina in spheroids than in 2D cultures. The circularity of BC lumina in the spheroids was significantly higher than that of the 2D cultures (Fig. 6C). The circularity of BC lumina might be an indicator of BC maturation. A more circular BC lumen might be advantageous for luminal constriction for bile excretion and uniform microvillus formation. The BC luminal areas and perimeters were not significantly different between the spheroids and the 2D cultures (Figs. 6A, B). In this study, we analyzed cross-sections near the center of BC structures (Supplementary Figs. S2, S3). Developing volumetric analysis methods for BC structures in future research is important for a more comprehensive evaluation.
We also performed comparative analyses of BC and intracellular lumen-like structures in a HepG2 spheroid. The luminal perimeters were significantly longer in the BC structures than in the intracellular lumen-like structures (p < 0.01, Fig. 7B). This might include both information that BC lumina tended to have larger areas and lower circularity than the intracellular lumen-like structures (Figs. 7A, C). The luminal RI (mean and median) of the BC structures was significantly higher than that of the intracellular lumen-like structures, indicating the enrichment of luminal contents, such as microvilli, in the BC structures (Figs. 7D, F). The identity of the intracellular lumen-like structures remains unclear. One possibility is that they are precursors of BC lumina, suggesting a maturation process reflected by the higher RI and longer perimeters of BC structures. However, alternative explanations like vacuoles or other organelles cannot be ruled out. Further characterization of the intracellular lumen-like structures, such as time-lapse tracking, is essential in future research.
HepG2 spheroids, with a size of approximately 100 µm in diameter, were reported to mimic liver gene expression, such as that related to xenobiotic and lipid metabolism, compared to monolayer cells.7) In this study, we confirmed various BC structures in similar sized spheroids, and paved the way for their noninvasive and quantitative analyses (Figs. 4–6). Taken together, HepG2 spheroids, even if relatively small in size, may be useful as an in vitro liver model that can mimic in vivo conditions or artificially emphasize a certain hepatic function.
Our findings lay an important foundation for accelerating the following three applications. First, the time-course observation of BC formation process. This would be particularly effective in the field of basic hepatology to reveal the biogenesis mechanisms of BCs and their networks, as represented by the following four points: (1) primary biogenesis mechanism of BCs, including the assembly process of microvilli-lined vesicles (MLV),1) and the formation process of tight junctions between hepatocytes21); (2) maturation process of BCs, such as luminal growth and bile acid secretion21); (3) dynamics of BC contraction,14) particularly in 3D hepatocyte cultures; and (4) biogenesis mechanism of BC networks, such as examining the effects of adding taurocholic or lithocholic acids to 3D hepatocyte cultures.22–24) Rief et al. imaged and analyzed BC contraction dynamics in mouse primary cultures at 10 min intervals.14) In our optical setups, the acquisition time of raw data for one sample (angle scanned 333 interferograms) was 5 s. The reconstitution time for one 3D RI map from one set of raw data was within 40 s for a field of 2D cultures and approximately 180 s for a HepG2 spheroid in our computational settings. Once the acquired raw data are stored, their RI maps can be reconstituted later. Therefore, our technology is promising for future research on BC dynamics, including their contraction cycle.
Second, quality check of hepatocyte cultures. BC formation is used as a differentiation indicator in hepatocyte culture models.6) For various downstream applications, the non-invasiveness of RI tomography for evaluating the quality of living hepatocyte cultures could be useful. Candidate downstream applications include regenerative medicine25) and the recovery of liver metabolites secreted into BCs.26,27) Tissue-derived hepatocytes and established hepatocyte cell lines tend to lose their polarity in simple planar cultures in vitro.4,6–9) To recover their polarity, various methods have been developed, such as sandwich28–30) and 3D cultures.4–11) Using the RI maps of BC structures as a marker of polarity recovery might advance the development of such liver culture methodologies.
Third, detection of BC dysfunction in drug discovery and clinical fields. Drug-induced liver injury (DILI) is a serious pharmaceutical issue, requiring predictive methods at the preclinical stage.31–33) BC dysfunction could cause cholestasis, which accounts for a large proportion of DILI cases.34) In cholestatic DILI, BCs exhibit abnormal morphology, which is either dilated or constricted.34) Therefore, to predict DILI, it is important to develop cell-based predictive systems that preserve the BCs,29) as well as technology that can easily detect abnormalities in the BC morphology.34,35) Furthermore, certain pathological conditions could cause abnormal BC morphology.36–39) For example, in Byler disease, the “Byler bile” comprises few microvilli in the BCs.38) Additionally, in one case of cholestasis, enlargement, microvilli reduction, and blebbing in the BC was observed.39) Thus far, these abnormalities in the BC morphology have been revealed using EM. If BC dysfunction can be detected in a label-free manner using RI tomography, a new drug screening methodology for DILI and the rapid diagnosis of liver biopsy might become possible.
This study had some limitations. The first limitation is the insufficient information on the functionality of the visualized BC structures, particularly in 3D cultured HepG2 spheroids. In future research, it will be crucial to compare RI tomograms and fluorescence images of BC components, including tight junctions, transporters, and bile acids. The second limitation is the lack of morphological comparison between RI tomograms and electron micrograms. In future studies, it is necessary to link the characteristics of RI distribution of the BC structures with electron micrograms. The third limitation is the lack of experiments to clarify the universality of the characteristics of RI distribution of BC structures revealed in this study, such as their applicability during BC maturation and in cell types other than HepG2. The cultivation period in this study was relatively short compared with that in previous studies,4,6) less than 1 week for both 2D and 3D cultures, suggesting that we captured the initial stages of BC formation. Therefore, the observation of fully matured BC structures in HepG2 cultures with longer cultivation times than those used in this study should be considered in future research. Furthermore, the observation of various hepatocyte samples, including primary cultures, is an important challenge. The fourth limitation is the difficulty in directly comparing absolute RI values between various BC structures. RI distribution in a certain cross-section of a specimen is inevitably influenced by that in the z-direction. This is known as the “missing cone problem.”13) Evidently, RI distribution in a certain cross-section of a spheroid is more susceptible to that in the z-direction than a 2D culture, owing to their thickness. In addition, when obtaining RI distribution of spheroids with in-silico clearing processing, reconstitution errors accumulate as the depth increases. This is related to the imaging depth limit of the technology.13) We used different reconstitution methods for RI maps in 2D cultures and spheroids (without or with in-silico clearing, respectively), due to the differences in multiple-scattering light derived from them. Taken together, these limitations made it difficult to directly compare absolute RI values, such as RI mean and median, between 2D cultures and spheroids. Therefore, we selected morphological parameters (area, perimeter, and circularity) and a relative RI parameter (RI standard deviation) for the comparative analyses of BC lumina (Fig. 6). We also need to cautiously interpret the results of directly comparing the absolute RI values within the same spheroid (Figs. 7D, F). In further research, it is important to develop compensational methods to realize the comparison of RI values among various BC structures.
The authors thank Y. Ueda, H. Yamada, S. Ishida, T. Matsushita, and Y. Komizu for their useful comments and encouragement.
KT designed the study, conducted the biological experiments, acquired the phase-contrast and fluorescence images, and analyzed the RI tomograms. OY acquired and reconstituted the RI tomograms. All authors discussed the results and wrote the manuscript. All authors contributed to the final version of the manuscript.
All authors are employed by Hamamatsu Photonics K.K. and have submitted patent applications for the technologies used in this study.
This article contains supplementary materials.
