https://orcid.org/0000-0003-4296-4054
Abstract
Peritoneal dialysis (PD) is an increasingly needed, life-maintaining kidney replacement therapy; efficient solute transport is critical for patient outcome. While the role of peritoneal perfusion on solute transport in PD has been described, the role of cellular barriers is uncertain, the mesothelium has been considered irrelevant. We calculated peritoneal blood microvascular endothelial surface area (BESA) to mesothelial surface area (MSA) ratio in human peritonea in health, chronic kidney disease, and on PD, and performed molecular transport related gene profiling and single molecule localization microscopy in two mesothelial (MC) and two endothelial cell lines (EC). Molecular-weight dependent transport was studied in-vitro, ex-vivo and in mice. Peritoneal BESA is 1-3-fold higher than MSA across age groups, and increases with PD, while the mesothelium is preserved during the first 2 years of PD. Tight junction, transmembrane and transcytotic transporter expression are cell-type specifically expressed. At nanoscale, tight junction anchoring protein Zonula occludens-1 is more abundant and more continuously expressed along the MC than the EC. Ionic conductance is 3-fold lower across the MC than human microvascular EC, as is the permeability for creatinine, 4- and 10-kDa, but not for 70-kDa dextran. MC removal from sheep peritoneum abolishes ionic barrier function. Short term intraperitoneal LPS exposure in mice selectively affects peritoneal mesothelial integrity and increases transperitoneal solute transport. We provide molecular correlates and consistent functional evidence for the mesothelium as a barrier for peritoneal solute transport, ie, essential information on peritoneal transport modeling, and for interventions to improve PD efficiency and biocompatibility, and beyond.
Introduction
For 5 decades, the peritoneum has been utilized as a dialysis membrane for removal of water and toxins. In view of the growing prevalence of end stage kidney disease, economic, staff and transplant organ shortage, the role of peritoneal dialysis (PD) as a life-maintaining therapy is increasing.1 For patient outcome adequate clearance of accumulating toxins is essential. PD function has been described by the three-pore model, which comprises ultra-small pores for selective water removal, small pores for fluid and small solute removal and large pores for large molecule transfer. This model well-predicts individual PD function.2,3 The molecular correlate for the ultra-small pores is aquaporin-1.4 At the same time, transcellular channels and transporters, endo- and exocytic vesicles, and paracellular tight junctions should represent the molecular components for the two other, hypothetical pores.
The peritoneum consists of the mesothelial cell monolayer, and the submesothelium, ie, the interstitium containing blood and lymphatic capillaries, and nerves.5,6 PD induces major peritoneal inflammation and progressive fibrosis.7 PD fluids low in glucose degradation product (GDP) content promote early angiogenesis,6 while high-GDP fluids accelerate cell death and vascular disease.8 The functional consequences of these major morphological alterations are partly described. Peritoneal perfusion, ie, vascularisation primarily defines small solute transport across the microvessel endothelial cell barrier,6 which is a leaky barrier allowing for rapid diffusion of water and small solutes.9 The healthy peritoneal interstitial space is not a barrier, but progressive submesothelial fibrosis and vasculopathy reduce osmotic conductance of glucose and thus fluid removal capacity.10 Increased peritoneal absorption occurs with increased lymphatic vascularisation.11
Based on experimental studies it has been concluded that the mesothelium does not impact on peritoneal transport function. In rodents, the thickness of the healthy peritoneum is about 10-20 µm and thus only 10% of the human peritoneum.6 PD fluid penetration depth is about 400 µm.12 Removal of the thin peritoneum in rats did not alter the osmotically induced fluid flux from the tissue into a chamber fixed on the inner abdominal wall, containing a 500 mosmolar mannitol solution. Mannitol uptake was unchanged.13 In 5 patients with peritoneal cancer, total peritonectomy resulted in a non-significant, 35% higher mitomycin C absorption than in five patients with 20%-60% peritonectomy.14 This suggested that local abdominal perfusion of the capillaries and the endothelium, but not the mesothelium control peritoneal transport function. On the other hand, the rabbit mesentery, which consists of two mesothelial layers, was demonstrated to be a permselective barrier to solute transport15,16 and the parietal pericardial mesothelium exerted most of the tissue resistance to diffusion.17 Diffusional resistance of the pericardial mesothelium to chloride, mannitol and sucrose was higher than of the pericardial connective tissue, even though the latter was approximately 35-fold thicker. A recent study in humans with normal kidney function, chronic kidney disease stage 5 (CKD5; just prior to dialysis) and on PD, demonstrated the expression of 9 paracellular permeability regulator proteins, tight junctions (TJ), and transcellular solute transporters in the parietal peritoneal mesothelium.18 These studies suggest a barrier function of peritoneal mesothelium to solute transport, which would be in contrast to the notion of solute transport in PD being only depending on the peritoneal density of perfused capillaries and resistance of the microvascular endothelium.
We now provide comprehensive evidence on the role of the peritoneal mesothelium in peritoneal membrane function. We quantified the relative peritoneal blood endothelial to mesothelial cell surface area in humans with normal kidney function, with CKD5 and during chronic PD, and demonstrate preservation of the peritoneal mesothelial cell layer during the first two years of chronic PD with the currently most widely used PD fluids with neutral pH and low GDP-content. For further understanding of the role of the peritoneal mesothelium we then analyzed the molecular machinery of paracellular barrier forming proteins in endothelial and mesothelial cells from human tissue ex-vivo 18 and in-vitro and the molecular weight-dependent transport rate across these barriers in-vitro and in mice.
Methods
For details see supplementary methods.
Human Peritoneal Tissue Analyses
Total peritoneal mesothelial, blood capillary endothelial surface area and peritoneal mesothelial coverage were calculated in peritoneal tissues from 106 patients with normal kidney function, 90 patients with CKD5 and 79 patients on maintenance PD using low-GDP fluids.6 The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Heidelberg University (S-747/2018).
Transcriptomics
Transcriptome datasets19 were analyzed for genes involved in transport processes.
Cell Cultures, Protein Quantification, and Transport Studies
Human cardiac microvascular endothelial cells (HCMEC), human umbilical vein endothelial cells (HUVEC), and immortalized mesothelial cell line (MeT-5A) were commercially obtained (HUVEC and HCMEC from PromoCell, Heidelberg, Germany; MeT-5A from LGC Standards, Wesel, Germany). Human peritoneal mesothelial cells (HPMC) were isolated from omentum of non-uremic patients undergoing abdominal surgery as described previously,20 and approved by the Institutional Review Board of Heidelberg University (S-501/2018). Validation studies on protein level were performed using immunostaining and western blotting. For measurement of transmesothelial and transendothelial resistance (TER), a measure of total ion permeability, and of solute transport across confluent polarized cell monolayers, cells were cultured on 0.4 µm polyester mesh (Transwell, Costar, USA). Creatinine transport was measured at four different time points and permeabilities for 4, 10, and 70 kDa dextran permeabilities of mesothelial and endothelial cell monolayers were calculated from 2 h and 4 h dextran transport kinetics as described in suppl. methods.
Single Molecule Localization Microscopy
Single molecule localization microscopy (SMLM) is based on reversible photo-bleaching inducing stochastic blinking of the fluorescent labeling molecules, allowing for accurate localization of single molecules (order of 10 nm).21,22,23 Data analysis was performed with an in-house developed python-based package24 to describe geometry and structural arrangements of the fluorescently labeled proteins. Here, Ripley statistics21,22 and Persistent Homology25 were applied. For further details see.26
Mouse Models
Male 7-8-week-old C57BL/6 J mice were randomly assigned to experimental groups. Control mice (n = 12) were injected intraperitoneally (i.p.) with a mixture of rhodamine B (RhodB)-conjugated 4.4 kDa dextran and fluorescein (FITC)-conjugated 70 kDa dextran (both 100 mg/kg in sterile PBS). A solution of 0.015 mg of creatinine per gram mouse body weight dissolved in PBS were injected i.p. at 100 mg/dL concentration (mean volume 300 µL). To alter mesothelial cell barrier properties, bacterial lipopolysaccharide (LPS) was given i.p. at a dose of 10 mg/kg body weight for creatinine and 10 and 50 mg/kg for dextran studies with the same amount of creatinine (n = 10 per group) and dextrans (n = 7 per group) as in the control mice. Tail blood samples were collected at time 0, 5, 15, and 30 min of LPS exposure for dextrans and after 0, 2, 7, and 15 min for creatinine, respectively. For one-path impedance spectroscopy mice parietal peritoneal tissues with and without removal of adjacent submesothelial muscle tissue were used.
Sheep Parietal Peritoneum Specimens and Ussing Chamber Measurements
Intact parietal peritoneum specimens (n = 21) were collected from the abdominal wall of female adult sheep immediately after death (warm ischemia time below 1 min) and placed in PBS at 4°C during transfer and mounted in Ussing chambers (K. Mussler Scientific Instruments, Aachen, Germany) within 30 min and analyzed. For animal tissue immunostaining see supplementary methods.
Statistics
In-vitro data were from at least 4 independent sets of experiments with 3-5 replicates. Normality was tested using Shapiro-Wilk test. In-vitro and patient data were described as means (standard deviation, SD or standard error of mean, SEM) or medians (interquartile range, IQR), unless indicated otherwise, and analyzed using t-tests and Mann-Whitney as appropriate. Multiple comparisons were conducted using one-way ANOVA or Kruskal-Wallis, followed by Holm-Sidak and Dunn’s correction. For dextran permeability experiments, two-way repeated measures ANOVA was applied. All tests were two-sided. P < 0.05 was considered statistically significant.
Results
Peritoneal Mesothelium and Blood Capillary Endothelium in Humans
Since peritoneal transport function not only depends on cell barrier characteristics but on the available surface area, we calculated the blood microvascular endothelial surface area (BESA) in the parietal submesothelium relative to the corresponding mesothelial surface area (MSA) in individuals with normal renal function, with CKD5 and on chronic PD with low-GDP fluids. Except for infants with a BESA/MSA ratio of 2 to 3, the ratio was about 1 across all age groups, ie, the endothelial surface area for transport is higher than the mesothelial area. Similar findings were obtained in patients with CKD5, but a 2-3-fold higher BESA relative to MSA in patients on PD (Figure 1A, B). Analysis of peritoneal mesothelial coverage was compromised by preparation artifacts in all groups including controls with normal kidney function, and within these limitations suggested unchanged coverage in CKD and during the first 2 years of PD, but progressive loss thereafter (CKD5 112 ± 7% versus controls, PD first year: 100 ± 9%, PD second year: 93 ± 11%, PD 3.-4. year: 66 ± 13%, >4 PD years: 48 ± 13%; mean ± SEM, P = ns/ns/ns/0.004/<0.001). Adjacent muscle was present in 29% of pediatric peritonea, and adipose tissue in the remaining pediatric and all adult samples. Muscle microvessel density was 3-fold higher than submesothelial vessel density in healthy children and children with CKD5, and similar in children on PD, due to increased submesothelial vessel density (Suppl. Fig. S1).
Peritoneal blood capillary endothelial over mesothelial surface area in individuals with normal kidney function, and in children with CKD5 and on PD. Age-dependent variation in the ratio of the human parietal peritoneal mesothelial surface area relative to the blood microvessel endothelial surface area in individuals with normal kidney function, ANOVA P = 0.0003 (A) and in children with CKD5 and on PD, ANOVA P < 0.0001 (B). Dashed line represents equal mesothelial/peritoneal and blood microvessel endothelial surface area (ratio = 1). MSA = mesothelial surface area, BESA = blood microvessel endothelial surface area, PSA = peritoneal surface area.
Human Peritoneal Mesothelial and Endothelial Transport Related Gene Expression
To understand the molecular machinery of solute transport across the peritoneum, we performed comprehensive analyses of the expression of transport related genes in HPMC, HCMEC, MeT-5A, and HUVEC. The number of cell junction, transmembrane transporter and channel, and of transcytosis related genes were similar in the four cell lines, but with differences in the cell-type specific gene expression profiles. 32% of cell junction, 41% of transmembrane transport and 23% of transcytosis genes were not shared by all four cell lines (Suppl. Fig. S2) and expression pattern varied substantially (Figure 2A). Among the non-shared transcripts, sealing TJ claudin (CLDN)1, was expressed in mesothelial cells, OCLN, CLDN7, and CLDN5 in the endothelial cells only, a finding confirmed independently for CLDN1 and 5 by confocal microscopy and western blotting (Figure 2B-D).
Mesothelial and endothelial cell type specific expression of transport related genes, and of tight junction proteins CLDN1 and CLDN5, (A) Expression map of cell junctions, transmembrane channels and transporters, and of transcytotic carriers in human primary peritoneal mesothelial cells (HPMC), immortalized pleural mesothelial cells (MeT-5A), human umbilical vein endothelial cells (HUVEC) and human primary cardiac microvascular endothelial cells (HCMEC). Sealing tight junction claudin1 (CLDN1) is only expressed in mesothelial cells and CLDN5 only in endothelial cells. (B) Immunocytochemical staining of CLDN1 protein in HPMC and MeT-5A, and CLDN5 in HUVEC and HCMEC, together with anchoring protein ZO-1. Pearson correlation analysis of the green and red channel colocalization and RGB spectra at the cell membrane. Scale bar = 10 µm. (C) Quantification of CLDN1 and -5 relative to ZO-1 immunofluorescence in the four cell lines at the cell membrane area (z-stack spacing 0.25 µm). (D) Representative Western blot analysis of CLDN1 and CLDN5 proteins (total protein extraction).
Of the shared genes involved in transport, scaled gene expression analysis revealed lower expression rates of genes involved in paracellular and transcellular transport in MeT-5A than in the three other cell lines, and lower expression of transmembrane transporters and channels in HUVEC compared to HPMC and HCMEC (Suppl. Fig. S3A). The top 50 genes shared by all 4 cell lines with highest differences in expression levels are given in Suppl. Fig. S3B. Comparison of HPMC and HCMEC and respective network of functions are given in Suppl. Fig. S3C, D.
Cell Specific ZO-1 Membrane Content and Clustering
Since TJ function depends on the molecule abundance and spatial organisation, single molecule localization microscopy was employed for Zonula Occludens-1 (ZO-1), an anchoring molecule linking TJ with the actin cytoskeleton. The number of ZO-1 molecules was 1-2 orders of magnitude higher in mesothelial as compared to endothelial cell membranes which can be seen in the absolute Ripley distance frequency curves (Figure 3A). In both cases, the frequencies of the pairwise distances indicate a continuous distribution with a tendency of clustering below 50 nm. The frequency distribution of endpoints of the barcodes obtained by persistent homology analysis revealed mesothelial cells displaying a continuous distribution pattern (small peak at lower distance values) and endothelial cells displaying a focal pattern with larger gaps between ZO-1 clusters (broader peak including also higher distance values) (Figure 3B, C). The barcodes obtained for components and holes by persistent homology (PH) were compared for each data set of a cell type with each other and the results were averaged in second generation heatmaps. The degree of topological similarity for distance and empty (hole) spaces analyzed by PH was higher within and between the two mesothelial cell lines and lower within and between the two endothelial cell lines (Figure 3D, E); thereby the similarity between endo- and mesothelial cell was lowest. These findings demonstrate a denser and more continuous TJ barrier formation in the mesothelial cell membranes, and a discontinuous, focal TJ distribution pattern in the endothelial cells, with the latter allowing for less restricted paracellular transport.
Single ZO-1 molecules are more abundant and more clustered in the cell membrane of mesothelial cells then endothelial cells, Absolute frequency of ZO-1 at the membrane is two magnitudes higher in mesothelial then endothelial cell lines (A). Persistent homology analysis of component endpoints (0 dimension) demonstrates lower distances and higher abundance of ZO-1 in mesothelial then endothelial cell membranes, ie, a denser and more continuous distribution (B), the holes endpoints (1 dimension) are comparable between mesothelial and endothelial cells. Topological similarity of the connected components (D) and topological similarity of the holes (E) shows a higher degree of similarity within and between single cells of the two mesothelial cell lines than within and between single cells of the two endothelial cell lines. The spectrum of the color bar of the heat maps ranges from red representing low similarity between the analyzed clusters to blue depicting high similarity.
Molecular Size Dependent Transport Across Mesothelial and Endothelial Cell Barriers In-Vitro
To study size-dependent transport kinetics across confluent, polarized mesothelial, and endothelial cell monolayers, cells were grown in Transwells. Confluence was reached after day 6 for HPMC and HUVEC, day 5 for MeT-5A, and day 4 for HCMEC as demonstrated by stable TER, which reflects the general ion conductance of the cell monolayers.27 While HPMC, MeT-5A, and HUVEC demonstrated comparable TER kinetics (at the day of confluence: 17.2 ± 1.1 Ω cm², 18.1 ± 1.2 Ω cm², 17.5 ± 0.6 Ω cm²), HCMEC exhibited a 2.5-fold lower TER (7.4 ± 0.6 Ω cm², P < 0.0001 versus all), ie, a higher ion conductance (Figure 4A). In the same direction, passages of creatinine (0.11 kDa) and 4 and 10 kDa dextrans were significantly faster across HCMEC than across HUVEC, MeT-5A, and HPMC monolayers (Figure 4B, C). Two path-impedance analyses, discerning transcellular and paracellular components of TER, demonstrated that in all four cell lines paracellular resistance, which is determined by tight junction proteins, was higher than transcellular resistance, which defined them as tight epithelial/endothelial cell lines although they in general possessed low TER values (Suppl. Fig. S4).
Transepithelial resistance, creatinine transport (0.11 kDa), and 4-, 10-, and 70-kDa dextran permeability of polarized mesothelial and endothelial cell monolayers. (A) TER of the four cell lines with increasing cell monolayer density in Transwells, with stable TER being reached with confluence. Confluent HCMEC have a 2.5-fold lower TER, reflecting higher ionic conductance (n = 5 experiments, 4-5 replicates, P < 0.0001 for HCMEC versus all). (B) The decline in creatinine concentrations in the apical Transwell compartment relative to the initial concentration (10 mg/dL) is given on the left graph. Volume of the basolateral compartment is five times higher (1 mL). The right graph gives the creatinine appearance relative to the creatinine added to the apical compartment and corrected for the higher basolateral volume. The dashed lines indicate the expected equilibration levels between compartments. Transport of creatinine is highest across confluent HCMEC cell monolayers (P < 0.0001/0.0001 for changes in the apical and basolateral compartment with 4 cell lines, and P < 0.0001/0.0001 for HPMC versus HCMEC only; two-way repeated measure ANOVA; n = 4 experiments, 3 replicates per experiment). (C) 4, 10 and 70 kDa dextran permeabilities of mesothelial and endothelial cell monolayers were calculated from 2 h and 4 h dextran transport kinetics. Paracellular permeability is higher for the smaller macromolecular dextrans across human microvascular endothelial cells. two-way repeated measures ANOVA P < 0.0001 for dextran size, P < 0.0001 for cell type; n = 4 experiments, 3 replicates per experiment, data area mean ± SD, ** P < 0.01, **** P < 0.0001). HCMEC = human cardiac microvascular endothelial cells, HUVEC = human umbilical vein endothelial cells, HPMC = human peritoneal mesothelial cells, MeT-5A = immortalized pleural mesothelial cells.
Disruption of the Mesothelial Cell Barrier Increases Transperitoneal Solute Uptake
To study the barrier function of mesothelial cells in-vivo, mice peritoneum was exposed to LPS for up to 30 min after i.p. injection. Peritoneal uptake of creatinine was fast, with subtotal uptake after 2.5 min with and without LPS and slightly more subsequent creatinine uptake with 10 mg/kg LPS exposure for 15 min (P = 0.037 versus control, 2-way ANOVA). Absorption of 4.4 kDa dextran increased with 10 mg/kg LPS exposure concentrations after 30 min, with 50 mg/kg LPS after 15 and 30 min, and with the latter for 70 kDa dextran absorption after 30 min (Figure 5A). Immunostaining demonstrated LPS accumulation in the mesothelium, but not in the submesothelium or adjacent sub-peritoneal tissue including vessels and muscles. Serum LPS concentration were not different from saline injected mice, but 50 mg/kg LPS increased mesothelial cell thickness (Figure 5B) and altered mesothelial cell nuclei to a rounder and thicker shape (Suppl. Fig. S5A). Mesothelial cell coverage of the peritoneum was unchanged by LPS (10 mg/kg 109 ± 18% and 50 mg/kg 108 ± 31% of saline control; P = 0.518/0.645 versus saline control), as were mesothelial CLDN1 and endothelial CLDN5 abundance (Suppl. Fig. S5B). Peritoneal mesothelial and submesothelial IL-6 abundance was unchanged with LPS exposure (10 mg/kg LPS 92.2 ± 65.8%, 50 mg/kg LPS 140 ± 35.1% of saline control; P = 0.94/0.23). These findings demonstrate a significant barrier function of the peritoneal mesothelial cell monolayer to solute transport across a large molecular size range.
Transperitoneal creatinine and dextran uptake with short term mesothelial LPS exposure in mice. (A) Increase in serum concentrations of creatinine (left panel), and of 4 and 70 kDa dextran (middle and right panel) during 15 and 30 mins following peritoneal injection of creatinine or dextran together with LPS. LPS increased peritoneal creatinine and 4- and 70 kDa dextran uptake. (creatinine: 100 mg/kg LPS *P = 0.037 versus control. 4.4 kDa dextran: *P = 0.0119 for 10 mg/kg and ***P < 0.0001 for 50 mg/kg LPS versus control. 70 kDa dextran: ***P < 0.0001 for 50 mg/kg LPS versus control; 2-way ANOVA; n = 7-12 mice/group, data are mean ± SEM). (B) Mesothelial cell thickness increased with LPS exposure (left graph upper row), LPS and DAPI immunofluorescence staining of mice parietal peritoneum demonstrates LPS deposition in the mesothelium (right graph upper row, magnification 400×), but not in the submesothelium of the peritoneum left graph lower row); serum LPS concentrations were unchanged (right graph lower row). Mesothelial claudin1 and endothelial claudin5 abundance were unchanged (Suppl. Fig. S6).
To further underpin these functional data, we studied peritoneal TER in the presence and absence of the mesothelium and the submesothelial tissue, respectively. Fresh sheep parietal peritoneum mounted in Ussing chambers had a TER with a median 14 Ω cm2 (IQR 11,17), but only 1 Ω cm² (1,1) when the mesothelium was mechanically removed (P < 0.001; Suppl. Fig. S6A). Histological analysis confirmed removal of the mesothelial cell monolayer (Suppl. Fig. S6B).
In mice peritoneum, after removal of the submesothelial muscle tissue, one-path impedance spectroscopy allowed to determine the resistance of the mesothelium and the submesothelial tissue, which both sum up to the TER. The one-path impedance spectroscopic data demonstrated that the mesothelial layer contributed significantly more to tissue resistance than the submesothelial tissue (P = 0.0005, Suppl. Fig. S6C).
Discussion
While the clinical importance of PD is increasing, the cellular and molecular counterparts for peritoneal function are hardly understood. To date, the mesothelium has been considered to be functionally irrelevant and excluded in models of peritoneal membrane function. This is striking given the physiological functions of the mesothelium, which include protection from mechanical stress and essential regulation of local fluid and solute homeostasis.15,16,28–32 We here provide consistent evidence that the mesothelium is a significant barrier for peritoneal transport.
Peritoneal dialytic solute transport implies crossing the peritoneal capillary endothelium, the interstitial space, which does not represent a barrier, unless major fibrosis has developed,33 and the mesothelium. Barrier function depends on the surface available for exchange and on physiological properties of the cell barrier. We here demonstrate that independent of the peritoneal area in contact with PD fluid, human peritoneal BESA per analyzed section area is higher than respective MSA in infants, and similar in older children and adults. The higher BESA/MSA ratio for infants is in line with the higher small solute transport in this age group.34 The ratio is not affected by CKD, but consistently 2- to 3-fold higher in patients treated with the widely used pH neutral, low-GDP PD fluids. With these fluids, peritoneal blood microvessel density doubles within four months of PD and remains elevated.6 a finding readily explaining the higher solute transport as compared with high-GDP fluids observed during the first year of PD.35,36 With high-GDP fluids, microvessel number is increased in patients with declining UF.37 Within the limitations of artifacts due to processing, we moreover demonstrate preserved peritoneal mesothelium during the first two years of PD with low-GDP fluids. This is in line with previous histological studies demonstrating superior mesothelial cell preservation as compared with high-GDP fluids8,38 and respective studies including effluent CA125 concentrations a bulk marker of mesothelial cell mass and viability.39,40 The positive BESA/MSA ratio, ie, the smaller mesothelial than endothelial surface available for transport suggests a significant barrier function of the mesothelium to peritoneal membrane transport. In line with this, effluent TJ protein abundance inversely correlate with peritoneal small solute transport,41 of which according to our data, claudin 1 is mesothelial cell derived. Superior preservation of the mesothelial barrier with low-GDP fluid usage may contribute to the observed better preservation of peritoneal membrane transporter status as with high-GDP fluids.36,42–44 These clinical findings are underpinned by our experimental barrier studies.
Our transcriptome studies demonstrate that mesothelial and endothelial cells exhibit cell-type specific expression patterns of cell junction proteins such as claudins, but also transmembrane transporters and channels, and transcytosis related genes. TJ proteins aggregate at the apical cell membrane and interact with the TJ proteins of the neighboring cell regulating paracellular solute and water permeability. Our two path-impedance analysis demonstrates a relatively higher contribution of paracellular than transcellular resistance to the barrier properties, ie, a relatively less permeable paracellular pathway than the transcellular route. To compare abundance and spatial organisation of TJ protein complexes in polarized endothelial and mesothelial cell membranes, a reflection of TJ function,45,46 we moreover performed single molecular localisation microscopy of ZO-1, which links TJ proteins to the actin cytoskeleton.47 ZO-1 abundance was significantly higher and had a greater continuity in mesothelial cell membranes, but was discontinuous in endothelial cell membranes, a finding which is in line with less restricted paracellular solute transport across the endothelial cell barrier. This finding reflects the previously described capillary endothelial intercellular clefts.48,49 In line with this, transport of creatinine as a small molecule and permeabilities for 4 and 10 kDa, as middle molecules below the general cut-off for transport, were higher across HCMEC than across mesothelial cells. Differences in creatinine transport could also be related to transcellular transport,50 with creatinine transporters OAT and OCT being expressed in both the mesothelial and endothelial cell lines; therefore we did not compare the creatinine transport rates directly with the permeabilities for the well-established paracellular flux markers we used27 Likewise, TER, an inverse measure of total ion permeability was significantly lower in HCMEC than in the other cell lines. TER of HUVEC was comparable to mesothelial cells. HUVEC, lining the umbilical vein, exert different physiological functions in humans, ie, nutrient and oxygen transport to the fetus. Conclusions on microvascular endothelial and peritoneal membrane transport properties51 based on HUVEC are difficult to draw as the exchange of water and solutes occurs in the capillary bed where endothelial cells are the most diverse in terms of gene expression.52 Likewise, MeT-5A substantially differ in expression patterns of genes involved in many essential biological processes including cell adhesion, and migration19 and in transport related genes (Figure 2). While the immortalized cell line is less challenging in experimental settings, findings should be reconfirmed with HPMC. A limitation of our in vitro transport regards the direction of solute transport. While it was in line with the in vivo setting in PD across endothelial cells, ie, from blood to the subendothelial space, in vitro transport across mesothelial cells was in the opposite direction to the in vivo setting, where it occurs from the basal to the apical side of the cells. However, the direction of transport may only matter for transcellular transport processes, when apical and basolateral localization of the involved transporters is different, while paracellular is a pure passive transport and thus only depending on gradients. In order to compare mesothelial against endothelial barrier/transport function, we refrained from adding solutes to the 5-fold larger lower compartment, implying large differences in distribution volumes and concentration gradients, respectively. We then demonstrated the impact of the mesothelial barrier in-vivo. Intraperitoneal exposure of mice to LPS affected the peritoneal mesothelium directly, given that its size largely excludes paracellular transport to the submesothelial compartment and which was reconfirmed by LPS and IL6 staining. LPS led to preservation of TJ proteins abundance (Claudin-1 and Claudin-5), however, mesothelial cell shape was altered, reflecting the LPS induced local reaction.53 LPS disrupts TJ function54 and thus the paracellular barrier. In line with this, LPS exposure increased the transport of creatinine, 4 and 70 kDa dextran, with transport rates higher as in humans, in whom 70 kDa dextran is used as intraperitoneal volume marker.55,56 In our ex-vivo experiments using mouse parietal peritoneum, the mesothelial resistance contributed more to total peritoneal resistance than the submesothelial tissue. In sheep parietal peritoneal tissue that has similar ion barrier properties with the human peritoneum,31,57 and that is also in the range of the polarized mesothelial cells in-vitro presented here, its spontaneous transmesothelial resistance was abolished after the mechanical removal of the mesothelial layer. This was done similar as previously with the parietal pericardial mesothelium, which also forms an ion barrier and exerts specific solute transport function.17 These findings consistently demonstrate that the mesothelium represents a barrier to solute transport.
We did not apply PD fluids in our in-vitro and in-vivo studies, since these induce complex transport alterations across cellular barriers in-vitro and have major impact on peritoneal function in-vivo.6,38,40 In rodents, saline and PD fluids are rapidly absorbed and differentially affect the peritoneum, e.g., peritoneal perfusion58 and thus physiological transport properties. Ex-vivo human peritoneal tissue studies demonstrated only few alterations in peritoneal transporter, channels and TJ abundance with chronic PD.18 comparative PET studies in children demonstrated similar transport kinetics with acidic, high-GDP and pH neutral, low-GDP fluids.39,55 We therefore studied peritoneal absorption of molecules, differing in size and transport routes and which do not exert significant local and systemic effects.59 Using single molecules dissolved in small amount of saline instead of PD fluids, however, precluded studies on glucose transport, present in PD fluids at far supraphysiological concentrations for fluid removal, and of water transport. Recent studies in humans demonstrated expression of SGLT-1 and -2 mainly in the peritoneal mesothelium.18,60 Blockade of these transporters reduced peritoneal glucose uptake and increased fluid removal (ultrafiltration)61,62 suggesting a respective impact of the mesothelium, with active transporters facilitating glucose uptake. In-vitro, calculated hydraulic conductivity of the mesothelium and the endothelium are comparable.63 Based on the present findings of the mesothelium representing a barrier for transport, it should moreover be of interest to correlate peritoneal mesothelial integrity and surface coverage with peritoneal solute transport function, and its contribution relative to peritoneal perfusion.
Altogether, we provide consistent in-vitro and ex-vivo evidence that the peritoneal mesothelium exerts barrier functions for small, middle and large molecules. Our findings inform PD related studies and kinetic modeling. They are of interest beyond PD, eg, in patients with ascites and with intraperitoneal drug delivery. Studies of the molecular correlates of mesothelial barrier function, eg, by gene knock-out will provide mechanistic understanding and identify potential therapeutic targets, eg, for improving PD efficacy and sustainability.
Acknowledgements
The authors are grateful to The Tissue Bank of the National Center for Tumor Diseases (NCT, Heidelberg, Germany) and Institute of Pathology (Heidelberg University Hospital) for technical assistance. The support of Kerpel-Fronius Talent Programme at the Semmelweis University and the Research Laboratory of the Semmelweis University Pediatric Center is gratefully acknowledged (E.L.).
Author Contributions
I.M. and M.B. contributed to the concept of the study, performed in-vitro experiments and human tissue studies, analyzed the data, and wrote the manuscript. E.L., R.H., A.J.S., and D.P. performed the mice transport studies, analyzed the data. A.S. and M.H. performed and analyzed the single molecule localization microscopy studies and wrote the respective part of the manuscript. Z.D., C.Z., and J.M.S. performed analyses regarding tissue studies and RNAseq studies. E.P. and S.S. performed sheep experiments and analyzed data. I.D. performed statistical analyses. D.K. performed microscopy experiments and analyzed data. T.L. and M.A. collected tissues and analyzed data. K.K. supervised RNAseq analysis and creatinine mice experiments. S.M.K. contributed to the concept of the study, performed impedance experiments, and analyzed data and contributed to the manuscript. S.G.Z. conceptualized the study, obtained funding, performed experiments, analyzed data, and supervised the sheep studies. C.P.S. obtained funding, conceptualized and supervised the study, analyzed data. and wrote the manuscript. All authors reviewed the manuscript.
Funding
This work has been part of the IMPROVE-PD project, that has received funding from the European Union’s Horizon 2020 Research and Innovation Programs under the Marie Sklodowska-Curie Grant Agreement Number 812699 (I.M., C.P.S.). S.G.Z. acknowledges the Alexander von Humboldt Stiftung/Foundation for an Experienced Researcher Fellowship (2019–2021) and the International Peritoneal Dialysis Society (ISPD) for an International Cooperation Research Grant (2019–2021). M.B. was funded by the German Research Foundation (Deutsche Forschungsgemeinschaft, project number 419826430) and Olympia Morata Fellowship from Heidelberg University and acknowledges Baden-Württemberg Stiftung for the financial support by the “Eliteprogramme für Postdocs.” C.Z. and Z.D. are supported by the China Scholarship Council (CSC) (Grant Number: 201908080237 and 202108320064). E.L. acknowledges the Jellinek-Harry Scholarship for the financial support. D.P. acknowledges Hungarian Academy of Sciences for the János Bolyai Research Scholarship. The financial support by the Austrian Federal Ministry of Science, Research and Economy and the National Foundation for Research, Technology and Development is gratefully acknowledged. C.P.S. has obtained funding from European Nephrology and Dialysis Institute (ENDI).
Conflict of Interest
Authors disclose no conflict of interests.
Data Availability
The raw data from the RNAseq from endothelial and mesothelial cell lines were uploaded to ArrayExpress and can be accessed under E-MTAB-12021.
