Zonula occludens‐1 distribution and barrier functions are affected by epithelial proliferation and turnover rates

Abstract Zonula occludens‐1 (ZO‐1) is a scaffolding protein of tight junctions, which seal adjacent epithelial cells, that is also expressed in adherens junctions. The distribution pattern of ZO‐1 differs among stratified squamous epithelia, including that between skin and oral buccal mucosa. However, the causes for this difference, and the mechanisms underlying ZO‐1 spatial regulation, have yet to be elucidated. In this study, we showed that epithelial turnover and proliferation are associated with ZO‐1 distribution in squamous epithelia. We tried to verify the regulation of ZO‐1 by comparing normal skin and psoriasis, known as inflammatory skin disease with rapid turnover. We as well compared buccal mucosa and oral lichen planus, known as an inflammatory oral disease with a longer turnover interval. The imiquimod (IMQ) mouse model, often used as a psoriasis model, can promote cell proliferation. On the contrary, we peritoneally injected mice mitomycin C, which reduces cell proliferation. We examined whether IMQ and mitomycin C cause changes in the distribution and appearance of ZO‐1. Human samples and mouse pharmacological models revealed that slower epithelial turnover/proliferation led to the confinement of ZO‐1 to the uppermost part of squamous epithelia. In contrast, ZO‐1 was widely distributed under conditions of faster cell turnover/proliferation. Cell culture experiments and mathematical modelling corroborated these ZO‐1 distribution patterns. These findings demonstrate that ZO‐1 distribution is affected by epithelial cell dynamics.


| INTRODUCTION
Cellular adhesion determines cell polarity and is supported by the dynamics of membrane proteins, cytoskeletons and signal transduction pathways. Multicellular organisms rely on cellular adhesion (e.g., the lining of epithelial cells) to differentiate themselves from the external environment. In vertebrates, the four intercellular junctional complexes are the following: adherens junctions (AJs), gap junctions, desmosomes and tight junctions (TJs).
The skin functions as a physical barrier to the outside world. 1 TJs are responsible for key barrier functions, such as regulating the passage of substances through paracellular spaces and controlling the diffusion of water and ions between the cells and the external environment. [2][3][4] A recent study demonstrated that TJs are found in the second layer of the stratum granulosum (SG2) in mouse ear skin 5 and human skin. 6 They are complex structures that are composed of more than 40 proteins, 7 including transmembrane proteins (i.e., occludin, 8 tricerullin, 9 angulin, 10 JAM family 11 and claudin family [12][13][14] and intracellular scaffold proteins, such as the zonula occludens (ZO) family. 15,16 The claudin family proteins bond tightly to each other and organize TJ strands. 14 During early differentiation, the TJ scaffold protein ZO-1 is also expressed in AJs. 17 We have recently shown that the distribution pattern of ZO-1 differs between the buccal mucosa and skin. 18 However, the underlying mechanism of these differences in the distribution pattern is unclear. In this study, we explored the role of epithelial cell dynamics in the regulation of ZO-1 expression. We employed human tissue samples, pharmacological treatments of murine skin and mathematical modelling to show that ZO-1 distribution is affected by the rate of epithelial proliferation and turnover.

| Animals
C57BL/6 mice were purchased from CLEA, Japan, Inc. All animal protocols were approved by the Animal Ethics Review Board of Hokkaido University (#15-0166 and #19-0164) and conform to the National Institutes of Health guidelines.

| Mitomycin C treatment
Mitomycin C treatment in mice 19 was performed with some modifications. Mitomycin C solution (Fuji Film Co., Ltd., Japan) was adjusted to a concentration of 2 mg/mL. C57BL/6J mice were intraperitoneally injected with a 5-10 mg/kg mitomycin C solution, euthanized after 10 days and the oral mucosal tissue was collected. Body weights were measured before each procedure.

| Mouse model of psoriasis
Psoriasis-like skin in mice was induced using a modified protocol described in a previous report. 20 Briefly, 5% imiquimod (IMQ) cream (Maruho Co., Ltd., Japan) was applied daily to the dorsal surface of the ears of 8-to 12-week-old male C57BL/6J mice for 4 days, after which the skin samples were collected. Ear thickness was measured daily using a pocket thickness gauge (Mitutoyo Europe GmbH, Germany), and body weight was recorded before the animals were sacrificed.

| Histology
Human and mouse tissue samples were fixed with formalin and embedded in paraffin following dehydration. For morphological analysis, deparaffinized sections were stained with haematoxylin and eosin (H&E) by conventional methods. 21 Images of H&E-stained sections were captured with a BZ-9000 microscope (Keyence, Tokyo, Japan).

| Immunofluorescence
Skin and buccal mucosa specimens were frozen in an optimal cutting temperature (OCT) compound (Cryomount I, 33351; Muto Pure Chemicals, Japan) and sectioned with a cryostat (Leica CM1950, Leica Biosystems, Germany). To prepare horizontal-sliced sections, specimens were embedded horizontally in the OCT compound, and

| Whole-mount staining
Whole-mount staining was carried out according to procedures

| Quantitative real-time polymerase chain reaction (qPCR)
Total RNA extraction and cDNA generation were processed as in the previous report. 24 All primers (Table S1) were obtained from Integrated DNA Technologies (IDT, Coralville, IA). Target gene expression levels were normalized to the expression level of the glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene.

| Skin equivalent model (SEM)
According to the manufacturer's protocol with minor modifications, Ker-CTs were cultured on Nunc™ Polycarbonate Cell Culture Inserts (0.4-μm pore size, 140668, Thermo Fisher Scientific) for 2-3 days with the CnT-PR medium. After 48 h, the medium was changed from Cnt-PR to CnT-PR-3D (#CnT-PR-3D, Cellntec, Switzerland), and the cells were incubated for an additional 2 days. Next, the medium in the insert was discarded, and the cell sheets were cultured for 12 days, with the medium changed three times a week. 25

| Paracellular flux measurement
Ker-CTs were cultured on Nunc™ Polycarbonate Cell Culture Inserts by the method described above. Fluorescent conjugated dextrans of various molecular weights was applied to the apical side of the insert,

| Image analysis
Image analyses were performed with ImageJ NIH (Bethesda, MD; https://imagej.nih.gov/ij/) by preparing a plug-in. In the plug-in, an enhanced image was prepared by extracting the maximum value within a distance of 10 μm from each pixel. From the resulting enhanced image, the ZO-1 intensity profile was obtained by averaging the brightness along a 40 μm line parallel to the horizontal aspects of the epidermis as a function of the depth. In the profile, the length of the region where the brightness was higher than 75% of the maximum brightness was set as ZO-1, whereas the total length corresponding to the epithelial layers was set as E. The ZO-1/E ratio was adopted to evaluate the expression of ZO-1.

| Mathematical modelling
We used a mathematical model for simulating 3D epidermal dynamics, 30 which calculated proliferation, migration and differentiation. Proliferation was controlled by a cell-division period T assigned to each cell in the stratum basale. Cell migration was computed by solving the equation of motion, where each cell was considered to be a spheroid. Differentiation into the stratum spinosum (SP), SG and SC was controlled by an inner variable S, which increased in time with the rate α. The SG cells were classified into SG1, SG2A, SG2B and SG3, 5 and defined as follows: an SP cell became an SG3 when its value of S crossed the threshold S1*, and it was in contact with at least one SG2A cell. An SG3 cell became an SG2A when it was in contact with at least one SG2B cell. The subsequent process from SG2B to SG2A and from SG2A to SG1 was controlled by the threshold values of S2* and S3*, respectively, with S1* < S2* < S3*. We regarded SG2 cells (SG2A and SG2B) as expressing ZO-1.
In the simulation used for this study, the basement membrane and the dermis were treated as a flat surface, whereas in the original model, they were deformable and undulating. 30 All other conditions for the simulation were the same as that described in the reference. 30 To investigate the effects of proliferation and turnover, we introduced an acceleration ratio λ, which changed the cell division period T to T/λ and the differentiation rate α to λα, thereby accelerating both proliferation and turnover λ times as fast when λ > 1. The thickness of the ZO-1 distribution was defined as follows: first, the horizontal region was divided into equally spaced 4 Â 4 subregions, and the maximum vertical distance among cells was computed in each subregion. These distances were subsequently averaged across all subregions.

| Statistical analysis
Statistical analysis was performed using Sigma plot ® (ver14.5, HULINKS, Japan). p-values were determined using the Student's t-test if they were homoscedastic or Welch's t-test if they were not homoscedastic. A p-value <0.05 was considered statistically significant. The values are presented as the means ± standard deviations (SD).

| Lower dispersal of ZO-1 with slower proliferation and turnover in inflammatory oral mucosa
ZO-1 is more dispersed in buccal mucosa than in skin 18 (Figures S1   and S2). To further elucidate distinct ZO-1 dispersal patterns, we focused on the cell dynamics in each type of epithelium. The turnover period in the buccal mucosa is typically 3.2-5.8 days, 31 while that in the skin is 40-56 days. 32 We hypothesized that this difference in the tissue turnover period could account for the distinct ZO-1 distribution patterns observed in each epithelium type. To prove this hypothesis, we first investigated human buccal mucosa samples of OLP, in which the turnover period is pathologically prolonged due to the damage caused to basal keratinocytes by inflammation. 33 The thickness of the mucosal epithelium was reduced in OLP samples ( Figure 1D) compared with that of normal buccal mucosa ( Figure 1A) (664 vs. 195 μm on average). In accordance with the thinner epithelium, the number Ki-67-positive cells was reduced in OLP ( Figure 1E,G) compared to that in normal buccal mucosa ( Figure 1B).
We then compared the ZO-1 distribution pattern in OLP and normal buccal mucosa. ZO-1 was dispersed in the upper half of the epithelium in normal human buccal mucosa ( Figure 1C,H) and in normal mouse mucosa (Figures S1b and S2). In contrast, ZO-1 was confined to the uppermost part of the epithelium in OLP ( Figure 1F,F' and 1H), which recapitulates skin ZO-1 ( Figure S1a).

| Lower dispersal of ZO-1 in mitomycin C-treated oral mucosa
To exclude the effects of inflammation on ZO-1 distribution, we intra-  (10 mg/mL; Figure 2J-L). These data demonstrate that slower proliferation and turnover lead to less dispersed ZO-1 distribution in the epithelium.

| Dispersed ZO-1 in human psoriatic skin
Based on the results in mouse skin, we speculated whether faster proliferation and turnover altered ZO-1 distributions in human psoriatic skin. Psoriasis is an inflammatory skin disease that is hyperproliferative in the epidermis and shortens the epidermal turnover period to 4-5 days, 34  previously. 18 In contrast, ZO-1 was distributed broadly in the psoriatic epidermis ( Figure 3F,H), simulating the distribution pattern found in normal human buccal mucosa ( Figure 1C,H). Similar to ZO-1, occludin was present in multiple epithelial layers in the psoriatic epidermis. In contrast, claudin-1 was distributed through the epidermis in both the normal human skin and the psoriatic skin ( Figure S3).

| Mathematical modelling
The results from the human, mouse and cultured cell experiments led us to speculate whether epithelial cell proliferation and turnover are sufficient to determine ZO-1 distribution. To answer this question, we employed mathematical modelling and a simulator of the epidermis 30 to compute ZO-1 distribution by varying the periods of proliferation and turnover. Compared to the control ( Figure 5A), ZO-1 distribution was broader when proliferation and turnover were accelerated ( Figure 5B). Moreover, the thickness of the ZO-1 distribution increased as the acceleration rate increased ( Figure 5C). These results suggest that the acceleration of cell proliferation and turnover can lead to a broadening of ZO-1 distribution.

| Relative decrease in ZO-1 and attenuation of barrier function upon acceleration of keratinocyte proliferation
We subsequently explored whether ZO-1 was quantitatively affected by cellular dynamics. EGF is a strong inducer of cultured keratinocyte proliferation 35 (Figure 6A-C). In contrast, gene expression of TJP1, encoding ZO-1, was reduced after EGF treatment ( Figure 6D). ZO-1 protein was also decreased in the treated epidermal cell sheet ( Figure 6E). Transepidermal water loss (TEWL) measurement and biotin tracer assay revealed a barrier dysfunction in the IMQ mice(- Figure 6F-H). Similarly, the SEM treated with EGF showed the apical leakage of biotin tracer as well as the reduction in ZO-1 expression ( Figure S4a-f). We further evaluated the outside-in barrier by applying fluorescein and 4-kDa FITC dextran to the monolayer cell sheets (N = 6). Dye leakage from the apical side to the basal side was estimated. Fluorescein leaked more dye when EGF was added, whereas 4-kDa FITC-dextran showed no significant dye leakage in the EGFtreated cells (Figure 6I,J). Although no significant differences were seen between the EGF-treated groups and the controls at 24 h after incubation, TER was significantly higher with EGF administration after 2 weeks when multi-layered epidermis was present (data not shown).
These data indicate that ZO-1 expression correlates negatively with cell proliferation and that the barrier function is attenuated in hyperproliferative epithelia, as observed in normal buccal mucosa and psoriatic skin (Figures 1A-C, 3D-F, and 4H-I).

| Expression of other epithelial and apoptosis markers in the epithelia with accelerated and slower cell proliferation/turnover
Our findings demonstrate that ZO-1 distribution correlates with cell proliferation/turnover, and yet, other epithelial markers may exhibit similar expression patterns. To test this hypothesis, we utilized keratin  Figure 7D,H,l). In contrast, loricrin was absent in OLP ( Figure 7P). These data suggest that the expression of major epithelial markers may partially reflect epithelial proliferation/turnover rate, and yet, ZO-1 may serve as a superior marker in this regard. We further excluded the involvement of apoptosis in ZO-1 distribution through the assessment of cleaved caspase-3 (CCP3) labelling, which showed that CCP3-positive cells were present in the psoriatic samples, 36 while NHS, BM and OLP had limited or no presence of such cells ( Figure S5).

| DISCUSSION
This study examined the role of cell proliferation and turnover in determining the distribution patterns of ZO-1. Slower epithelial proliferation and turnover (e.g., normal skin, OLP and MMC-treated mucosa) resulted in a restricted distribution ZO-1, whereas faster proliferation and turnover (e.g., normal buccal mucosa and psoriasis) increased the distribution of ZO-1 in the epithelium. It is noteworthy that this altered distribution indirectly relates to cell proliferation rates, as proliferating cells do not express ZO-1.
Cell proliferation and turnover are closely linked processes and are often distorted into a faster state, such as in psoriasis 37,38 or a slower state, as in OLP. 33 One caveat of our study is that we could not solely manipulate proliferation or turnover without affecting the other. Further studies are needed to elucidate the contribution of each of these factors to the dynamics of ZO-1. In murine ear skin, TJs are located in the SG2 layer where they form a flat barrier, and their distribution is strictly controlled by dynamic remodelling. Based on our results, epithelia with faster proliferation and turnover form continuous ZO-1 over several layers ( Figure S2). In epithelia with these characteristics, the cells may start expressing ZO-1 earlier in the differentiation step to functionally compensate for immature junctional complexes. In line with this scenario, when we investigated a single cell, ZO-1 did not cover the whole cell periphery in the epithelia with faster proliferation and turnover ( Figure S2). However, it is still possible that the production and degradation of ZO-1 protein (protein turnover) are altered with the cell proliferation status. Longterm live imaging may solve this issue in future studies.
Previous studies have shown that TEWL is higher in psoriatic skin than in normal skin, 39,40 which is consistent with our data ( Figure 6F).
This barrier dysfunction has been attributed to the disruption of epidermal tight, gap and adhesion-binding proteins. 41 In normal epidermis, TJ proteins are expressed only in the granular and upper spinous layers. However, in psoriatic lesions, the TJ proteins, including ZO-1, occludin and claudin-4, are highly expressed outside the granular layer 42,43 ( Figure S3d-f) and are downregulated, suggesting an overall impairment of the TJ function. 44,45 This impairment is likely to contribute to the increased TEWL and decreased hydration seen in psoriatic lesions. 46 IL-17 and IL-1β, which are implicated in psoriasis pathogenesis, are also known to downregulate keratinocyte adhesion proteins, at least partially explaining the dysregulation of the TJ function in psoriasis. 47,48 Our study confirms the barrier dysfunction in the psoriatic epidermis and suggests that proliferation and turnover may be responsible for this dysfunction.
Our study took advantage of OLP for hypoproliferative conditions in the mucosal epithelium. Previous studies have shown the expression of TJ proteins in cutaneous lichen planus (CLP) and OLP 49,50 ; although to our knowledge, ZO-1 labelling in OLP has not been investigated. ZO-1 labelling in CLP was described in one paper as being as extensive as in psoriatic epidermis. 49 However, upon reexamination of the original figure, it appears that ZO-1 immunoreactivity in CLP was weak and its distribution was restricted. 49 In conclusion, the confinement of ZO-1 within a single layer of squamous epithelia is indicative of slow epithelial proliferation and turnover rates. Our findings do not fully elucidate the functional significance of ZO-1 distribution, but our study does suggest that ZO-1 reflects the proliferation and turnover of epithelia, which could have potential use as a marker of disease severity in psoriasis and OLP.
Among the TJ proteins, claudin-1 was unable to discriminate between epidermis from normal human subjects and psoriasis patients ( Figure S3). Occludin exhibited a similar expression pattern to ZO-1 ( Figure S3), but occludin signals appeared as discrete dots in the images. Therefore, ZO-1 labelling may be a more suitable proxy for this purpose. Further studies utilizing clinical samples and live imaging are needed to confirm our hypothesis. Kobayashi.