The use of the peritoneum as a dialysis membrane in patients with end-stage renal disease (ESRD) for the removal of solutes and fluid excess implies that the feasibility of this mode of chronic renal replacement therapy is dependent on one single membrane. Anything used intensively will decay in the long-term and the peritoneum of patients treated with peritoneal dialysis (PD) is no exception. It is however likely that the decline rate can be retarded by appropriate modifications of the PD treatment. In this review the morphological alterations and their likely causes will be discussed and their impact on peritoneal transport. Special attention will be given to glucose-induced pseudohypoxia and its effects on various growth factors and glucose-transporter-1 expression. The effects of prophylaxis and treatment, especially by adjustments of the dialysis solutions will also be examined.

The peritoneal membrane is not a single structure, but comprises the mesothelial layer and the submesothelial interstitial layer, in which blood and lymphatic vessels are dispersed. The interstitium is composed of a ground substance consisting of hyaluronan and glycosaminoglycans, and a fibrous collagen network that constitutes a scaffold for the embedded structures. Adipose cells and a few fibroblasts are the predominant cell types. The mesothelial layer offers no hindrance to peritoneal transport (Flessner, 1994), meaning that the volume of distribution of the instilled dialysis fluid includes both the peritoneal cavity and the interstitium. Consequently the peritoneal microcirculatory blood vessels are the determinants of peritoneal solute and fluid transport from the circulation to the dialysate-filled peritoneal cavity. Transport across the microvascular wall occurs through a system of pores (Pappenheimer et al., 1951). According to the generally accepted 3-pore theory solute transfer during PD occurs through inter-endothelial pores, but additionally water is also transported by the intra-endothelial water channel aquaporin-1 (AQP-1) (Rippe and Stelin, 1989; Rippe, 1993). Small inter-endothelial pores with a radius of about 40 Ǻ constitute 90% of the total number of pores and allow the transport of small solutes that all have radii < 3 Ǻ and low molecular weight proteins like β₂-microglobulin. Small clefts between endothelial cells are the most probable morphological equivalent of the small pores. Transport through these is by diffusion to the dialysate-filled interstitium. The contribution of convection is small for low molecular weight solutes, due to the low hydrostatic pressure gradient and the high diffusion rates of these solutes. The low hydrostatic pressure gradient does however induce a small amount of fluid transport, similar to trans-capillary ultrafiltration in the non-PD situation. In addition to the abundance of small pores, a limited quantity of large inter-endothelial pores is located in the venular part of the microcirculation. Their radii exceed 150 Ǻ and they allow the passage of large molecules like serum proteins. As water removal from the circulation by hydrostatic filtration is limited and counteracted by back-filtration into the circulation by colloid osmosis, PD as kidney replacement therapy is only possible by the creation of a crystalloid osmotic pressure gradient. This is usually achieved by the addition of large quantities of glucose (molecular weight 180 Da, radius < 3 Ǻ) to the dialysis solution. The small size of the glucose molecule implies that it is not very efficient as an osmotic agent, because after 4 h about 60% is reabsorbed into the circulation (Smit et al., 2003). Consequently any efficacy for fluid transport through the small pores is only present during the first few hours of a dialysis dwell (Parikova et al., 2008). The presence of the water channel AQP-1 in peritoneal capillaries and venules explains the water removal by glucose, because only water can traverse this water channel. The radius of AQP-1 is too small to enable the passage of solutes, meaning that it allows free water transport by a very efficient crystalloid osmotic gradient, irrespective of the osmole employed (Ni et al., 2006).

Already at the start of PD the inter-individual variation of peritoneal transport parameters is very large (Twardowski et al., 1987). This may partly be related to differences in the genotype of various factors or structures involved in peritoneal transport. For instance the inflammatory cytokine interleukin-6 (IL-6) is locally produced in the peritoneal cavity during PD and has a number of polymorphisms. It appeared that the CC genotype had the highest expression in peritoneal effluent and also the highest concentration of the protein, which was associated with a faster solute transport compared to the other genotypes (Gillerot et al., 2005). However follow-up to 36 months failed to show any influence of IL-6 polymorphisms on peritoneal transport (Lee et al., 2011), suggesting that local PD-related effects become more important with PD duration.

Alterations in peritoneal morphology and transport occur already during the first few years of PD. After 3 months advanced glycosylation end-products (AGEs) -formed from non-enzymatic interactions between glucose and amino acid residues and leading to irreversible cross linking of tissue proteins-can be found in the mesothelial and submesothelial layer (Yamada et al., 1994). Endothelial-to-mesenchymal transition of mesothelial cells (mesothelial-to-mesenchymal transition, MMT) is the earliest morphological change. This phenomenon has first been described in 2003 and consists of a transition of the usual epithelial phenotype to a mesenchymal one with loss of cytokeratin expression. MMT in peritoneal biopsies is characterized by the presence of cytokeratin-positive fibroblasts-like cells in the submesothelial interstitial tissue (Yáñez-Mó et al., 2003). The prevalence of MMT is highest between 1.5 and 2 years, when it is present in about one third of patients (Del Peso et al., 2008) The cytokeratin-positive fibroblasts produce large amounts of vascular endothelial growth factor (VEGF) (Aroeira et al., 2005). MMT is accompanied by high effluent concentrations of the mesothelial cell marker cancer antigen 125 (CA125) and by high small solute transfer rates. It may explain the relationship between the creatinine transport rate and both effluent CA125 and VEGF in incident PD patients who never had peritonitis, during the first year of treatment (van Esch et al., 2004).

MMT occurs during the first 2 years of PD and has been claimed to predict long-term peritoneal membrane alterations, but this is only a hypothesis in the absence of follow-up biopsies.

Vascular density in the interstitium is not different from that before the start of PD, but the wall/total surface ratio of small arteries is somewhat increased compared to non-dialyzed patients with end-stage kidney failure (Mateijsen et al., 1999). This has been called vasculopathy and was present in a mild form in 20–30% of patients during the first 2 years (Williams et al., 2002; Del Peso et al., 2008). About 40% of peritoneal capillaries was of the immature type, i.e., stained negative for α-smooth muscle actin (α-SMA) (Nakano et al., 2020).

Longitudinal follow-up of peritoneal transport either showed no effect on peritoneal solute transport and ultrafiltration (Selgas et al., 1994; Coester et al., 2014), or some increase of the dialysate/plasma (D/P) ratio of creatinine accompanied by a decrease in ultrafiltration, due to a faster disappearance of the glucose-induced osmotic gradient (Davies, 2004). Accordingly an inverse relationship was present between D/P creatinine and ultrafiltration volume during the first years on PD (Davies et al., 1993). Three possibilities could explain the reported high solute transfer rate. These include 1) the CC genotype of interleukin-6, 2) a high production of VEGF during MMT, or 3) the number of newly formed immature capillaries, because of the reported relationship with D/P creatinine (Nakano et al., 2020).

All constituents of the peritoneal dialysis membrane are affected by PD duration. Although no longitudinal studies on the development of the alterations have been published in adult PD patients treated with conventional PD solutions, a general picture can be discerned from the cross-sectional analyses, as reviewed recently (Parikova et al., 2021). Lesions that progress with PD duration include accumulation of AGEs in the subendothelial and perivascular regions. This is accompanied by loss of mesothelial cells, vasculopathy leading to subendothelial hyalinosis and narrowing of vascular lumina or even obstruction, and by an increased thickness of the submesothelial fibrous layer or even more general interstitial fibrosis of the stroma. The interstitial fibrosis consists of myofibroblasts. Vascular density is only increased in the presence of severe interstitial alterations (Mateijsen et al., 1999; Williams et al., 2002).

All longitudinal studies showed a progressive increase of peritoneal small solute transfer after 2 years (Selgas et al., 1994; Davies, 2004; Coester et al., 2014). This was accompanied by a lower ultrafiltration, which was directly related to solute transport up to 4 years of treatment. Thereafter a dissociation occurred between both parameters: ultrafiltration was markedly lower, than predicted from D/P creatinine (Davies, 2004). Analysis of ultrafiltration at 60 min after instillation of a 3.86% glucose dialysis solution showed a decrease of both small-pore fluid transport and free water transport after 4 years (Coester et al., 2014). The lower rate of small-pore fluid transport suggests a reduction of the hydrostatic pressure gradient. AGE-induced vasculopathy is the most likely cause, because the presence of a luminal stenosis likely diminishes the post-stenotic filtration pressure (Krediet, 2018; Krediet et al., 2019). Free water transport after 4 years reached its lowest values in patients who developed encapsulating peritoneal sclerosis (EPS) (Sampimon et al., 2011) and a value of less than 75 ml after a 60 min dwell was even predictive of EPS with a sensitivity of 100% and a specificity of 81% (Barreto et al., 2019).

All EPS patients had a normal expression of AQP-1 (Morelle et al., 2015). Therefore peritoneal interstitial fibrosis may be important in the observed reduction of free water transport with PD duration (Krediet et al., 2015). Two recent studies are supportive of this hypothesis. Kinetic modelling of peritoneal transport was used in a limited number of PD patients with ultrafiltration failure in the first study. The results suggested that the disappearance of glucose from the dialysate consisted of a vascular- and an interstitial component (Stachowska-Pietka et al., 2019). The tissue diffusivity of glucose appeared higher in the patients with ultrafiltration failure, than in those without this complication. This tissue diffusivity was modelled, not measured, meaning that other mechanisms like uptake in interstitial cells, cannot be excluded The second study comprised longitudinal follow-up of a large patient group treated with conventional PD solutions. A break-point in the time-course of the mass transfer area coefficient of creatinine was present after 3 years. This was also present for the percentage of the instilled glucose quantity, that had disappeared from the dialysate. (Van Diepen et al., 2020). However the slope of the increase after the breakpoint was much steeper for glucose than for creatinine. These two studies point to a contribution of interstitial fibrosis to the enhanced disappearance of glucose from the dialysis solution reducing the crystalloid osmotic gradient for AQP-1 and thereby inducing less free water transport.

A glucose molecule is too large to enter a cell by simple diffusion, but requires the presence of specific proteins in the cell membrane, known as glucose transporters. Two types can be distinguished: those that facilitate diffusion (GLUTs) and sodium glucose linked transporters (SGLTs), in which cellular glucose uptake is coupled to that of sodium (Navale and Paranjape, 2016). GLUT-1 is present on the plasma membrane of many cell types including murine fibroblasts (Ortiz and Haspel, 1993) and is especially upregulated when a high cellular influx of glucose is required, for instance during hypoxia (Airley et al., 2001). It can be hypothesized that expression of GLUT-1 by peritoneal interstitial myofibroblasts, the density of which increases in long-term PD, explains the high tissue apparent diffusivity of glucose causing its enhanced disappearance from the peritoneal cavity. Both will lead to a reduction of the crystalloid osmotic gradient for AQP-1 (Krediet, 2021). This is illustrated in Figure 1.

Twenty years ago I suggested that glucose exposure causes a status of peritoneal tissues called pseudohypoxia analogous to the pathogenesis of diabetic complications (Krediet et al., 2002). Pseudohypoxia in diabetes mellitus has been described as a consequence of hyperglycemia (Williamson et al., 1993). An impaired oxidation of cellular nicotinamide dinucleotide (NADH) to NAD⁺ causes an increase in the NADH/NAD⁺ ratio, which is characteristic of hypoxia caused by a lack of oxygen, because this blunts NADH oxidation. The consequences in diabetics include increased superoxide anion production and possibly nitric oxide formation. Pseudohypoxia in diabetes mellitus develops, because hyperglycemia increases the influx of glucose into cells, where it is subsequently metabolized in the glycolysis to pyruvate. NAD⁺ is converted to NADH during this breakdown of glucose. Pyruvate is taken-up in the mitochondria were it contributes to the formation of acetylCoA. This product is part of the Krebs circle, which participates in the respiratory chain. This oxidation requires the donation of electrons for which NADH is one of the sources, leading to NAD⁺ regain. In case of hypoxia the conversion of NADH to NAD⁺ can to some extent be accomplished by lactate dehydrogenase, which converts pyruvate to lactate. NAD⁺ is regained during this reaction. In case of a severe hyperglycemia, intracellular glucose is also metabolized in the sorbitol pathway. The formed sorbitol is subsequently metabolized to fructose. During the latter reaction NAD⁺ is converted to NADH. In consequence the NADH/NAD⁺ increases both in the glycolysis and the sorbitol pathway, but in contrast to the glycolysis the formed NADH cannot be oxidized to NAD⁺ in the sorbitol pathway. The resulting high cytosolic NADH/NAD⁺ ratio is very similar to the situation in hypoxia or ischemia and is therefore known as pseudohypoxia.

Non-diabetic PD patients have normal plasma glucose concentrations, but those in peritoneal dialysate exceed even severe hyperglycemic levels by far. Furthermore, some compensatory mechanisms for the oxidation of NADH to NAD⁺ may be impaired. For instance, mitochondrial dysfunction can occur in patients with kidney failure (Galvan et al., 2017). Also the use of high lactate concentrations as buffer substance in dialysis solutions may inhibit the lactate dehydrogenase activity, because extracellular lactate can traverse the cell membrane, leading to high cytosolic lactate concentrations (Halestrap and Wilson, 2012). A comparison between lactate–buffered and bicarbonate-buffered dialysis solutions aimed to compare the degree of pseudohypoxia, has never been published. The pathways for intracellular glucose metabolism in PD patients are illustrated in Figure 2.

Strong evidence for the presence of peritoneal pseudohypoxia in long-term PD patients has been published by the group of Devuyst et al. The authors reported on the expression of the genes for nitric oxide synthase and VEGF in peritoneal tissue of patients (Combet et al., 2000). The expression of both increased with PD duration (Pseudo)hypoxia enhances these compensatory vasodilating mechanisms, as has been shown for VEGF (Ferrara, 1995), which is not only produced during MMT of mesothelial cells, but local production of the effluent VEGF protein has also been shown in a cross-sectional analysis in prevalent PD patients (Zweers et al., 1999). Furthermore, an increase of effluent VEGF has been shown during longitudinal follow-up (Zweers et al., 2001). These results are all in agreement with peritoneal hypoxia in PD patients.

Hypoxia characterized by a high NADH/NAD⁺ ratio, stimulates upregulation of the hypoxia inducible factor-1 (HIF-1), which leads to upregulation of various factors, like EPO, GLUT-1, VEGF and various profibrotic and angiogenic factors (Darby and Hewitson, 2016). The latter include TGFβ, plasminogen activator inhibitor-1 (PAI-1) and connective tissue growth factor (CTGF). Increased expression of GLUT-1 on peritoneal interstitial myofibrobasts may cause a vicious circle, in which the cellular uptake of dialysate glucose is stimulated leading to more pseudohypoxia and thereby more stimulation of GLUT-1 expression, This induces a further augmentation of cellular glucose uptake and a reduction of the interstitial glucose gradient. A progressive decline in peritoneal ultrafiltration is the clinically relevant result (Krediet, 2021).

Stimulation of TGFβ, PAI-1 and CTGF by glucose-induced pseudohypoxia occurs during peritoneal dialysis and is likely relevant for the development of the peritoneal alterations that occur in long-term PD patients (Krediet and Parikova, 2022). The upregulation of TGFβ is probably most important in the development of interstitial fibrosis (Margetts et al., 2001). However the concentration of the TGFβ protein in peritoneal effluent does not reflect its biological function, because it is bound to α₂-macroglobulin in the circulation (Zweers et al., 1999). Both PAI-1 and CTGF are downstream regulators of the TGFβ pathway. The dialysate concentrations of these pro-fibrotic proteins increase with the duration of peritoneal dialysis (Mizutani et al., 2010; Barreto et al., 2013). An effluent PAI-1 appearance rate exceeding 8.5 ng/ml had a sensitivity of 100% for a clinical diagnosis of EPS within 1 year and a specificity of 55% (Barreto et al., 2019).

It can therefore be concluded that extremely high glucose concentrations in peritoneal dialysis solutions have two effects. First they lead to the formation of AGEs, that are likely involved in the genesis of vasculopathy inducing a decline of small-pore fluid transport, and second they cause pseudohypoxia, which stimulates various genes that are involved in the peritoneal alterations that occur in long-term PD patients, of which the decline in free water transport and peritoneal fibrosis are most relevant. The severe ultrafiltration failure in long-term PD patients is explained by a combination of severely impaired small-pore fluid transport and free water transport, as illustrated in Figure 3.

Reduction of peritoneal exposure to the excessively high glucose concentrations is the cornerstone of pseudohypoxia prevention in peritoneal tissues and its consequences for peritoneal morphology and transport. This can be accomplished by modifications of the dialysis solutions and possibly by drugs that inhibit the causative factors for the development of the morphological and functional alterations. Modifications of the dialysis solutions consist of changes of osmotic agents, combinations of osmotic agents, changes in the dialysate buffer and additives to the solutions Also when the consequences of pseudohypoxia are already present, treatment consists of the discussed preventive measures to avoid further progression.

The functional and morphological alterations that occur in the peritoneal dialysis membrane after long-term PD can be ascribed to glucose-induced formation of AGEs causing vasculopathy, and to pseudohypoxia, defined as an increased cellular NADH/NAD⁺ ratio. The latter causes upregulation of HIF-1, which leads to upregulation of GLUT-1 and its expression on peritoneal cells, and to upregulation of the growth factors TGFβ, VEGF, PAI-1 and CTGF. GLUT-1 in the cell membrane increases the influx of glucose and induces a vicious circle leading to a progressive decline of free water transport. The growth factors stimulate fibrosis and angiogenesis. The obvious therapeutic approach consists of combinations of various osmotic agents, all in low concentrations. This reduces the toxicity of all individual compounds while their osmolalaties add up, maintaining their over-all osmotic effects. All other discussed options are secondary.