Heparan Sulfate Plays a Central Role in a Dynamic in Vitro Model of Protein-losing Enteropathy*

Protein-losing enteropathy (PLE), the loss of plasma proteins through the intestine, is a symptom in ostensibly unrelated diseases. Emerging commonalities indicate that genetic insufficiencies predispose for PLE and environmental insults, e.g. viral infections and inflammation, trigger PLE onset. The specific loss of heparan sulfate (HS) from the basolateral surface of intestinal epithelial cells only during episodes of PLE suggests a possible mechanistic link. In the first tissue culture model of PLE using a monolayer of intestinal epithelial HT29 cells, we proved that HS loss directly causes protein leakage and amplifies the effects of the proinflammatory cytokine tumor necrosis factor α (TNFα). Here, we extend our in vitro model to assess the individual and combined effects of HS loss, interferon γ (IFNγ), TNFα, and increased pressure, and find that HS plays a central role in the patho-mechanisms underlying PLE. Increased pressure, mimicking venous hypertension seen in post-Fontan PLE patients, substantially increased protein leakage, but HS loss, IFNγ, or TNFα alone had only minor effects. However, IFNγ up-regulated TNFR1 expression and amplified TNFα-induced protein leakage. IFNγ and TNFα compromised the integrity of the HT29 monolayer and made it more susceptible to increased pressure. HS loss itself compromises the integrity of the monolayer, amplifying the effects of pressure, but also amplifies the effects of both cytokines. In the absence of HS a combination of increased pressure, IFNγ, and TNFα caused maximum protein leakage. Soluble heparin fully compensated for HS loss, providing a reasonable explanation for patient favorable response to heparin therapy.

The intestinal epithelium blocks the entry of pathogens and their products into the body and also prevents the loss of plasma components into the intestinal lumen. In protein-losing enteropathy (PLE), 2 where plasma proteins leak into the intestine, this essential barrier is impaired (1)(2)(3)(4). PLE develops as a life-threatening complication of seemingly unrelated diseases, e.g. Crohn's disease (1), Congenital Disorders of Glycosylation (CDG) (2), or after Fontan surgery to correct congenital univentricular hearts (3,4). Emerging commonalities are beginning to suggest patho-mechanisms underlying PLE. It appears to involve a combination of genetic insufficiencies and environmental insults. The evidence for this is that patients do not suffer from PLE continuously; rather, PLE is episodic. Its onset is often associated with viral infections and a pro-inflammatory state (2,(5)(6)(7)(8), indicating that multiple factors combine to trigger PLE. Most intriguing is the loss of heparan sulfate (HS) proteoglycans (HSPG) specifically from the basolateral surface of intestinal epithelial cells only during PLE episodes (1,2,9,10) followed by its reappearance when PLE resolves (2).
The combination of multiple factors contributing to PLE is most apparent in post-Fontan patients. The Fontan procedure was developed to separate the systemic and the pulmonary circulation in patients with congenital heart malformations (11). Although the surgery vastly improves survival, it often elevates central venous pressure (11). Venous hypertension in post-Fontan patients is thought to be associated with intestinal protein leakage with the increased pressure from the basolateral side pressing plasma proteins into the intestinal lumen (11,12). Because venous pressure increases immediately after the intervention and remains elevated for the rest of the patient's life, there is no good explanation why PLE develops in only 3-10% of the patients months to years after the surgery (3,4). The most likely explanation is that genetic factors and Fontan-induced venous hypertension predispose for PLE, which precipitates upon a series of sequential or simultaneous environmental insults. Consistent with this hypothesis, seven out of eight post-Fontan patients have been diagnosed with viral infections at the onset of PLE symptoms (5), indicating that this additional insult triggered PLE. Jejunal biopsies, taken during episodes with PLE, revealed an increased IFN␥ concentration (7), most likely as a response to the viral infection and elevated levels of the pro-inflammatory cytokine TNF␣ (6). Both cytokines are known to impair the integrity of the intestinal epithelial barrier (13)(14)(15)(16)(17)(18).
Similar to other primary diseases associated with PLE, episodes of post-Fontan PLE are characterized by a loss of HSPG specifically from the basolateral surface of intestinal epithelial cells. HSPG expression in the lamina propria is normal (10). 3 Overall intestinal architecture remains intact, and the expression of other matrix components is also normal (10). 3 The reasons why HSPGs are lost during episodes of PLE are still unknown. Recently, we established the first in vitro model of PLE and proved a direct link between HS loss and protein leakage through a monolayer of HT29 cells, a human intestinal epithelial cell line. In addition, we showed that HS loss amplifies TNF␣-induced protein leakage (19), providing the first reasonable explanation on how two of the factors associated with PLE onset synergize. Heparin compensates for the loss of cell-associated HS and abolishes the synergism between HS loss and TNF␣ (19). These results offer a potential explanation for the favorable response some PLE patients have to heparin treatment (20 -23).
HS loss and elevated TNF␣ concentrations are just two factors thought to trigger PLE. Based on the strong correlation between PLE onset and viral infections together with increased IFN␥ concentrations, we hypothesized that IFN␥ either directly or indirectly causes protein leakage. We further hypothesized that venous hypertension caused by the Fontan surgery predisposes for protein leakage. To test these hypotheses, we modified the in vitro model to mimic venous hypertension and assessed the individual and combined effects of increased pressure, HS loss, IFN␥, and TNF␣ on protein leakage. We find that HS is the central figure coordinating their synergistic effects.
Albumin Flux-Cells were grown on a semipermeable membrane in Snapwell inserts (12-mm diameter, 3.0-m pore size, Corning, NY) for 5 days until they reached confluence. Albumin flux was measured in cells treated with heparinase III (HSase), p-nitrophenyl-␤-D-xylopyranoside (␤-xyloside), heparin (Sigma), recombinant human IFN␥, or recombinant human TNF␣ (R&D, Minneapolis, MN) at final concentrations and time points as indicated for each experiment. Cells were washed twice with RPMI 1640 (with/without serum, with/without phenol red) (Invitrogen). The ring-supported membrane with the HT29 monolayer was mounted on a slider and inserted into the EasyMount Ussing Chamber System (Physiologic Instruments, San Diego, CA). Both sides of the HT29 monolayer were bathed in 4-ml Ringers solution (pH 7.4) gassed with 95% O 2 , 5% CO 2 . The chamber system was constantly heated at 37°C. To determine protein leakage, albumin-FITC (400 l/ml) was added to the Ringers solution bathing the basolateral side of the HT29 monolayer. After 3 h the concentration of albumin-FITC in the buffer bathing the apical side of the monolayer was measured using a spectrofluorometer (excitation. 485 nm; emission, 538 nm). Basolateral to apical albumin flux through the untreated monolayer is defined as 1.0. To evaluate the effects of hydrostatic pressure, a millimeter scale was tagged to the transparent chambers. The addition of 250 l of Ringers solution to one side of the chamber generates 1 mm of hydrostatic pressure.
Confocal Microscopy-HT29 cells grown on the Snapwell membrane were fixed with paraformaldehyde and incubated with a mouse monoclonal antibody against dipeptidyl peptidase IV (clone 202-36, LabVision, Fremont CA). The secondary antibody was FITC-labeled (Sigma). Cell nuclei were stained with ToPro3 (Molecular Probes, Eugene, OR). The Snapwell membrane was carefully cut out of the ring support with a razor blade and placed on a microscope slide, and the cells were visualized with a MRC 1024 SP Bio-Rad laser point scanning confocal microscope.
Flow Cytometry-HT29 cells were harvested with EDTA and incubated with a mouse monoclonal antibody against TNFR1 (clone H-5, Santa Cruz Biotechnology, Santa Cruz, CA) or with an isotype-matched control antibody to determine nonspecific binding. The secondary antibody was FITC-labeled (Sigma). Flow cytometry was performed on a BD Biosciences FACSort and the CellQuest software. Data for 10,000 events were collected. Events were considered positive if their fluorescence intensity exceeded that of 98% of the isotype-matched control antibody. The median fluorescence intensity for TNFR1(FITC) in untreated controls is defined as 1.0. The Annexin V-FITC apoptosis detection kit (BD Pharmingen) was used to quantitatively determine the percentage of cells undergoing apoptosis and cell death.
Statistical Analysis-Results are given as the means Ϯ S.D. from three independent experiments. Differences between interventions were tested by the two-tailed Student's t test. p Ͻ 0.05 is considered significant.

RESULTS
In Vitro Monolayer of Intestinal Epithelial Cells-Microscopic observation revealed that the HT29 cells had grown into a confluent monolayer 5 days after seeding on Snapwell inserts. We used confocal microscopy to determine monolayer polarity. The apical surface marker dipeptidyl peptidase IV was exclusively expressed on the cell membrane facing the media (data not shown), defining the apical surface. The cell membrane attached to the porous membrane of the Snapwell did not express dipeptidyl peptidase IV, therefore defining the basolateral surface. To prove that the untreated monolayer at day 5 provides a barrier to prevent basolateral-to-apical protein leakage, we added albumin-FITC to the basolateral side and determined albumin flux through the monolayer by measuring the concentration of albumin-FITC on the apical side over time. Although albumin-FITC concentrations on both sides of an empty insert (without cells) reached equilibrium in less than 4 h, the HT29 monolayer grown on the insert prevented protein leakage and allowed only 1.2% albumin-FITC to pass from the basolateral to the apical side even after 8 h. This percentage did not further decrease when cells were grown on the Snapwell inserts for more than 5 days, confirming that the cells reached confluence at day 5 (data not shown).
Effects of Individual Factors-HS loss, IFN␥, TNF␣, and venous hypertension are thought to be involved in the molecular pathogenesis leading to protein leakage in patients with PLE. Our first aim was to evaluate the individual contributions of each of these four factors alone. To deplete cell-associated HS, we incubated the basolateral surface of the HT29 monolayer with HSase at 0.6 milliunits/ml for 1.5 h. We have reported earlier that these conditions cause maximum depletion of cellassociated HS (19). Here, albumin flux increased 2.6 Ϯ 0.1-fold compared with untreated controls (p Ͻ 0.01) (Fig. 1A). Co-incubation with 100 M ␤-xyloside, which competes out GAG chain synthesis on core proteins, did not further increase HSase-induced protein leakage (data not shown).
Incubating the cells with TNF␣ at 2 or 20 ng/ml for 12 h increased albumin flux 2.3 Ϯ 0.4 (p Ͻ 0.01)-and 4.9 Ϯ 0.5-fold (p Ͻ 0.001), respectively (Fig. 1A). These results with our Ussing chamber model are consistent with our previous data in a Transwell system showing that HS loss and TNF␣ cause protein leakage (19). IFN␥ had no effect on protein leakage when applied in low (1 ng/ml) or moderate concentrations (10 ng/ml), but high concentrations (100 ng/ml) slightly increased albumin flux 1.5 Ϯ 0.2-fold (p Ͻ 0.05) (Fig. 1A). TNF␣ or IFN␥ at concentrations and incubation times used in these experiments did not alter the percentage of annexin V-positive cells or the uptake of propidium iodine (data not shown), and we concluded that neither TNF␣nor IFN␥-induced protein leakage was caused by increased apoptosis or cell death.
To mimic venous hypertension, we raised the hydrostatic pressure on the basolateral side of the monolayer by 1.0, 2.5, and 5.0 mm H 2 O. Albumin flux increased 4.1 Ϯ 0.8-, 8.9 Ϯ 1.1-, and 15.8 Ϯ 0.7-fold, respectively (Fig. 1B). To assure that the hydrostatic pressure did not cause the cells to detach from the Snapwell membrane, we examined the monolayer by confocal microscopy after measuring albumin flux. The monolayer was still intact up to 5.0 mm H 2 O, but the cells began to detach from the membrane at 10.0 mm H 2 O (data not shown).
Effects of Paired Combinations; Synergy-After determining the effects on protein leakage for each of the four factors alone, we next did pairwise combinations. First, we combined HS loss and pressure since we hypothesized that loss of HS from the basolateral surface compromises the integrity of the monolayer and makes the monolayer more susceptible to increased pressure. Results are shown in Fig. 2A. Adding the individual effects of HS loss and 2.5 mm H 2 O pressure, we predicted a 10.5-fold increase in albumin flux. In fact, we measured a 15.7 Ϯ 0.9-fold increase in albumin flux (p Ͻ 0.001), indicating that the combined effects of HS loss and increased pressure were not only additive, but synergistic.
Similarly, we hypothesized that TNF␣ also compromises the integrity of the monolayer and enhances the effects of increased pressure. We incubated the cells with TNF␣ (2 ng/ml) for 12 h, applied different amounts of hydrostatic pressure on the basolateral side, and determined protein leakage (Fig. 2B). The measured effects on albumin flux were higher than the predicted sum, suggesting synergistic effects of TNF␣ and increased pressure. In contrast, incubating the cells with IFN␥ (10 ng/ml) instead of TNF␣ did not affect pressure-induced protein leakage (data not shown).
Next, we combined HS loss and TNF␣. We incubated the cells with HSase for 1.5 h to digest cell-associated HS. At the same time we added 100 M ␤-xyloside to inhibit GAG synthesis on newly synthesized core proteins and, thus, prevented the reappearance of cell-associated HS over time (19). Afterward, we incubated the cells with different concentrations of TNF␣ for 12 h and determined protein leakage. The effects of HS loss and TNF␣ were not only additive, but synergistic (Fig. 2C). For example, although TNF␣ (2 ng/ml) alone increased albumin flux 2.3 Ϯ 0.4-fold and HS loss alone increased albumin flux 2.6 Ϯ 0.1-fold, the combination of these two factors increased albumin flux 5.8 Ϯ 0.9-fold (p Ͻ 0.01) instead of the predicted 2.9-fold. This observation is consistent with our previously reported data showing that HS depletion ampli-fies TNF␣-induced protein leakage (19), especially at low cytokine concentrations.
We also combined HS loss with IFN␥. We either digested the cellassociated HS with HSase 1.5 h before IFN␥ treatment or incubated the cells with IFN␥ first and then digested with HSase. In both cases, IFN␥ had no effect on albumin flux beyond that of HS loss alone (data not shown).
To combine the two cytokines, we preincubated the cells with IFN␥ (10 ng/ml) for 12 h and then added TNF␣ (2 ng/ml) for another 12 h before measuring albumin flux. Although IFN␥ alone had no effect on protein leakage and TNF␣ increased albumin flux only 2.3 Ϯ 0.4-fold, incubation with IFN␥ before the addition of TNF␣ increased albumin flux 4.4 Ϯ 0.7-fold (p Ͻ 0.01), indicating that IFN␥ amplifies TNF␣induced protein leakage (Fig. 2D). In summary, results from the twofactor systems revealed synergistic effects between HS loss and pressure, TNF␣ and pressure, HS loss and TNF␣, and IFN␥ and TNF␣ but not between HS loss and IFN␥ or IFN␥ and pressure.

Effects of Triple Assaults;
Enhanced Synergy-Next we tested combinations of three factors. Fig. 3A shows the results of combining HS loss, IFN␥, and TNF␣ treatments. We incubated the cells with IFN␥ (10 ng/ml) for 12 h and added TNF␣ (2 ng/ml) for another 12 h. We depleted cell-associated HS before the addition of each cytokine. The individual effects of HS loss, IFN␥, and TNF␣ predict a 3.0-fold increase in albumin flux, but we actually measured a 11.2 Ϯ 1.8-fold increase (p Ͻ 0.001).
Depleting cell-associated HS before the addition of IFN␥ but not before TNF␣ still increased albumin flux 6.8 Ϯ 0.8-fold (p Ͻ 0.01), which was higher than predicted but significantly lower compared with HS depletion before both the addition of IFN␥ and the addition of TNF␣ (p Ͻ 0.01). We also depleted cell-associated HS before TNF␣ addition but not before IFN␥ addition, which increased albumin flux 8.1 Ϯ 0.9-fold (p Ͻ 0.01). This effect was again higher than predicted but significantly lower than the effect of HS depletion before both IFN␥ and TNF␣ addition (p Ͻ 0.05). These results show that HS loss amplified the effects of both IFN␥ and TNF␣.
To investigate the combined effects of HS loss, TNF␣, and pressure, we depleted cell-associated HS, incubated the cells with TNF␣ (2 ng/ml) for 12 h, and applied different amounts of hydrostatic pressure (Fig. 3B). With 2.5 mm H 2 O pressure, the predicted increase is 11.8-fold, but we measured a 16.5 Ϯ 0.8-fold increase in albumin flux.
Finally, we measured the combined effects of IFN␥, TNF␣, and pressure. We incubated the cells with IFN␥ for 12 h and with TNF␣ for another 12 h, applied different amounts of hydrostatic pressure, and measured albumin flux (Fig. 3C). Instead of the predicted 10.3-fold increase in albumin flux at 2.5 mm H 2 O pressure, we measured a 14.9 Ϯ 1.4-fold increase (p Ͻ 0.001).
Four Factors Produce Maximum Synergy-Finally, we combined all four factors. We depleted cell-associated HS from the basolateral surface, incubated the cells with IFN␥ and TNF␣, applied different amounts of hydrostatic pressure, and determined protein leakage (Fig.  4). Although the addition of the individual effects of the four factors predicted a 7.1-fold increase in albumin flux for a hydrostatic pressure of 1.0 mm H 2 O and a 11.9-fold increase in albumin flux for 2.5 mm H 2 O, we actually measured a 18.7 Ϯ 2.1-and 26.3 Ϯ 1.4-fold increase, respectively. Notably, the effect of only 1.0 mm H 2 O pressure in combination with the other three factors (18.7 Ϯ 2.1-fold) is higher than the effect of 5.0 mm H 2 O pressure alone (15.8 Ϯ 0.7-fold). In fact, combining a hydrostatic pressure of 5.0 mm H 2 O with the other three factors caused the cells to detach from the Snapwell membrane, which pressure alone or in combination with HS loss or TNF␣ cannot do. These results indicate that the combined effects of all four factors were highly synergistic.
Basis of HS Loss, IFN␥, and TNF␣ Synergy-Because IFN␥ alone had no effect on protein leakage but enhanced TNF␣-induced protein leakage, we asked whether this synergism is caused by an IFN␥-mediated up-regulation of the TNFR1 receptor (24). To address this question, we  To actually determine the combined effects of these three factors, IFN␥ (10 ng/ml) and TNF␣ (2 ng/ml) were added to the cells 24 and 12 h before measuring albumin flux, respectively. Digestion with HSase depletes cell-associated HS, and the addition of ␤-xylosides inhibits HS attachment to de novo synthesized HSPG core proteins (HX, co-incubation with HSase and ␤-xylosides). Cell-associated HS was depleted either 1.5 h before IFN␥ treatment (2nd bar), 1.5 h before TNF␣ treatment (3rd bar), or both (4th bar). rel., relative. B-C, albumin flux is proportional to the hydrostatic pressure applied to the basolateral side (filled diamonds). The predicted albumin flux after adding the individual effects of HS loss, TNF␣ (2 ng/ml), and hydrostatic pressure (B) or IFN␥ (10 ng/ml), TNF␣ (2 ng/ml), and hydrostatic pressure (C) are shown as open triangles and dashed lines. In both cases (B-C) the actually measured albumin flux (filled triangles) is higher than the predicted albumin flux ,indicating synergistic effects (*, p Ͻ 0.05; **, p Ͻ 0.01).
incubated the HT29 monolayer with IFN␥ (10 ng/ml) for 12 and 24 h, harvested the cells, and determined TNFR1 expression by fluorescenceactivated cell sorter analysis (Fig. 5A). Incubating the cells with IFN␥ for 12 h increased TNFR1 expression 1.9-fold compared with untreated controls. In parallel, TNF␣-induced albumin flux increased 2.0-fold when cells were preincubated with IFN␥ 12 h before the addition of TNF␣ (Fig. 5B).
Because IFN␥ binds to and may be inactivated by HS (25,26), we further hypothesized that loss of cell-associated HS increases the amount of active IFN␥ and amplifies IFN␥-mediated up-regulation of TNFR1 expression. To test this hypothesis, we digested cell-associated HS with HSase, washed the cells, and incubated them with IFN␥ for 12 or 24 h. That treatment up-regulated TNFR1 expression 2.7-fold ( Fig.  5A) and, in parallel, increased TNF␣-induced protein leakage 3.1-fold (Fig. 5B). These results support our hypothesis that IFN␥ up-regulates TNFR1 expression and, therefore, enhances TNF␣-induced protein leakage, which is further amplified in the absence of cell-associated HS. This provides a mechanistic explanation on how HS loss, IFN␥, and TNF␣ synergize.
Heparin Abolishes Synergism between HS Loss, IFN␥, and TNF␣-In our previous study we showed that soluble HS and heparin compensate for the loss of cell-associated HS and abolish synergism between HS loss and TNF␣ (19). We now asked whether heparin also abolishes the synergistic effects between IFN␥ and TNF␣ based on the report that binding to heparin inactivates IFN␥ (25,26). To answer these questions, we incubated the cells with IFN␥ (10 ng/ml) for 12 h, washed them, added TNF␣ (2 ng/ml) for another 12 h, and measured albumin flux. We depleted cell-associated HS 1.5 h before both IFN␥ addition and TNF␣ addition (Fig. 6A). This combination of HS loss, IFN␥, and TNF␣ increased albumin flux 11.4 Ϯ 1.3-fold. To determine the effect of heparin on IFN␥, we used the same incubation scheme as before but coincubated IFN␥ together with heparin (2.5 g/ml). Now, albumin flux was increased only 8.5 Ϯ 1.3-fold, which was significantly lower than the effects without heparin (p Ͻ 0.05). Similarly, co-incubating TNF␣ with heparin also reduced albumin flux, which confirmed our previous results. Finally, we co-incubated both IFN␥ and TNF␣ with heparin, and albumin flux increased only 3.7 Ϯ 0.9-fold. Notably, albumin flux after co-incubation of both IFN␥ and TNF␣ with heparin in the absence of cell-associated HS is not significantly different from albumin flux after incubation with IFN␥ and TNF␣ without heparin but in the presence of HS (4.4 Ϯ 0.7-fold increase). Heparin at 10-fold higher concentrations reduced albumin flux even below these levels (only a 2.6 Ϯ 0.3-fold increase). These results indicate that a low concentration of heparin fully compensates for HS loss. At a higher concentration heparin even quenches more cytokine activity than cell-associated HS would. However, heparin had no effect on pressure-induced protein leakage (Fig. 6B), suggesting that that these soluble glycans cannot compensate for the mechanical barrier function of cell-associated HS.
Albumin Flux Is Bidirectional-So far we only determined the effects on basolateral-to-apical protein leakage. However, we hypothesized that the interventions not only cause leakage in one direction but impair barrier function in general and also induce apical-to-basolateral leakage. To test this hypothesis, we repeated key experiments and now added albumin-FITC to the apical side and determined apical-to-basolateral albumin flux by measuring the concentration of albumin-FITC on the basolateral side. There was no difference between basolateral-to-apical and apical-to-basolateral albumin flux in any of the interventions tested (Fig. 7, A and B).

DISCUSSION
The Fontan procedure vastly improves survival of patients born with univentricular hearts but elevates venous pressure (11). We hypothesized that venous hypertension is one of the key players involved in the patho-mechanisms underlying PLE as a late complication of the Fontan surgery. To mimic venous hypertension in vitro, we established a tissue culture model of PLE using modified Ussing chambers that allow us to apply hydrostatic pressure to the basolateral surface of intestinal epithelial cells. The Fontan procedure elevates right atrial pressure by several mm Hg (mean, 10.9 mm Hg; normal, Ͻ3-5 mm Hg) (12), but the exact pressure in subepithelial capillaries is hard to assess. In our in vitro model, an increased hydrostatic pressure in the magnitude of a few mm H 2 O ( 1 ⁄ 13 mm Hg) already causes significant protein leakage (Fig. 1B). The cells even begin to detach from the Snapwell membrane at 10.0 mm H 2 O pressure, which is less than 1 mm Hg. The epithelial monolayer grown on the semipermeable Snapwell membrane in our in vitro model lacks a stable extracellular matrix and the context of subepithelial cells, which makes it more susceptible to increased pressure and limits the ability to directly translate in vivo venous pressure to the amounts of hydrostatic pressure applied in our in vitro model. However, results with mouse mucosal explants stripped of seromuscular layers show that increasing basolateral pressure to 5.0 mm H 2 O causes similar protein leakage ex vivo compared with our in vitro model (27).
Venous pressure is elevated immediately after the Fontan procedure, and post-Fontan patients already show a subtle increase in enteric protein loss determined by fecal ␣1-antitrypsin concentrations (12). Heart transplantation decreases pressure and normalizes protein loss (28), indicating a direct correlation between increased pressure and protein leakage. However, PLE becomes manifest only months to years after the surgery (3,4), suggesting that multiple sequential or simultaneous factors finally trigger PLE. We now identify increased pressure together with HS loss, IFN␥, and TNF␣ as FIGURE 5. IFN␥ up-regulates TNFR1 expression (A) and enhances TNF␣-induced albumin flux (B), which is further amplified in the absence of cell-associated HS. A, fluorescence-activated cell sorter analysis shows that incubating cells with IFN␥ (10 ng/ml) for 12 h (light gray bars) increases TNFR1 expression 1.9-fold compared with untreated controls. IFN␥-mediated TNFR1 up-regulation is further enhanced when cell-associated HS is digested with HSase before the addition of IFN␥. TNFR1 expression reaches base-line levels after 24 h (dark gray bars). B, incubating cells with IFN␥ for 12 h before the addition of TNF␣ increases albumin flux 2.0-fold compared with cells that are incubated with TNF␣ without prior addition of IFN␥ (defined as 1.0). Digesting cell-associated HS before the addition of IFN␥ further amplifies TNF␣-mediated albumin flux. rel., relative.
key players in the pathogenesis of post-Fontan PLE, evaluate their individual contributions and their combined effects, and provide mechanistic explanations on how they synergize.
We confirm previous results (19) showing that both HS loss and TNF␣ alone increase protein leakage (Fig. 1A) and that they act synergistically (Fig. 2C). In addition, we show that HS loss and TNF␣ make the HT29 monolayer more susceptible to increased pressure (Figs. 2, A and B, and 3B), presumably through different mechanisms. While cellassociated HS connects the cell to the extracellular matrix and to neighboring cells (29), TNF␣ regulates tight junctions (30) and shedding of membrane proteins that connect to the extracellular matrix (31). Results from our previous study in HT29 cells show that TNF␣-induced protein leakage is mediated by tyrosine kinases and proteinase A and partially involves matrix metalloproteases (19). HS loss and TNF␣ may, therefore, impair the mechanical integrity of the monolayer and increase its susceptibility to increased pressure. Onset of post-Fontan PLE is often associated with viral infections (5) and one of the predominant cytokines responding to viral infections is IFN␥, which is indeed elevated in post-Fontan PLE patients (7). Incubating T84 cells, another human intestinal epithelial cell line, with IFN␥ (100 ng/ml) reduces their barrier function after 72 h but not after 24 or 48 h (18,32). Although we did not test the long term effects of IFN␥, the results after 24 h of incubation are consistent with our observations in HT29 cells where IFN␥ induced only minor albumin flux (Fig. 1A). However, in concert with the other three factors IFN␥ is highly deleterious in our model. IFN␥ enhances the expression of TNFR1 (Fig. 5A) to the same extent as it increases TNF␣-induced protein leakage (Fig. 5B). IFN␥-mediated up-regulation of TNFR1 provides a reasonable explanation for the mechanisms of IFN␥ and TNF␣ synergy.
Loss of cell-associated HS further amplifies the effects of both IFN␥ and TNF␣ (Fig. 3A). HS binds to both IFN␥ (25,33) and TNF␣ (34,35). In most cases, cytokines as well as growth factors employ cell-associated HS to enhance their signal, as previously shown for TNF␣ (34). In contrast, we report that binding to HS seems to lower the concentration of available TNF␣ as well as of IFN␥, whereas loss of cell-associated HS enhances their activities. Why cell-associated HS inactivates IFN␥ in some cases but activates it in others is unknown. We speculate that these cell-and tissue-specific differences stem from the differential expression (36,37) and Golgi organization (38) of HSmodifying enzymes such as sulfotransferases or sulfatases, which generate specific HS epitopes that bind cytokines and change their conformation to an active or inactive state (25).
The loss of cell-associated HS can be fully compensated by the addi- . Heparin compensates for HS loss and alleviates IFN␥-amplified and TNF␣-induced albumin flux (A) but has no effect on pressure-induced albumin flux (B). A, IFN␥ (10 ng/ml) and TNF␣ (2 ng/ml) were added to the cells 24 and 12 h before measuring albumin flux, respectively. 1.5 h before the addition of IFN␥ and 1.5 h before the addition of TNF␣, cell-associated HS was depleted, and its de novo synthesis inhibited by digestion with HSase and incubation with ␤-xyloside (HX). Heparin in low (hp, 2.5 g/ml) and 10-fold higher (HP, 25 g/ml) concentration was added to the medium either together with IFN␥, together with TNF␣, or both. B, heparin, added 1.5 h before measuring albumin flux, had no effect on pressure induced albumin flux. (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001).
FIGURE 7. Albumin flux is induced in both directions, basolateral to apical as well as apical to basolateral. A-B, albumin-FITC was either added to the basolateral side, and its appearance was measured on the apical side (basolateral to apical flux (white bar)), or albumin-FITC was added to the apical side, and its appearance was measured on the basolateral side (apical to basolateral flux (gray bar)). Basolateral to apical albumin flux in untreated controls is defined as 1.0. Experiments are performed for key interventions showing that there is no difference in the direction of albumin flux. rel., relative.
tion of low concentrations of soluble heparin (2.5 g/ml). Heparin binds to both IFN␥ (25,33) and TNF␣ (34,35) and inactivates them (25,34), which alleviates protein leakage in our model (Fig. 6A). At higher concentrations (25 g/ml), heparin not only compensates for HS loss but reduces protein leakage even further, suggesting that heparin inactivates more IFN␥ and TNF␣ than would cell-associated HS. These results provide an explanation of why some post-Fontan patients with PLE respond to heparin treatment, especially when given early after PLE onset (20 -23) at a time when the vicious circle of HS loss, IFN␥, and TNF␣ can still be stopped.
In conclusion, HS loss increases the local availability and activity of IFN␥ and TNF␣. IFN␥ up-regulates TNFR1 and, therefore, amplifies the effects of TNF␣; TNF␣ induces HSPG shedding and HS loss (19,31) and down-regulates HSPG expression (31). This closes the vicious circle that impairs the integrity of the intestinal epithelium and reduces its resilience to increased pressure. The combination of all four factors produces maximum synergy with more than a 25-fold increase in protein leakage (Fig. 4). In post-Fontan patients, episodes of PLE are triggered by the combination of venous hypertension, HS loss, viral infections, and inflammation. These multiple hits increase intestinal protein leakage more than 20-fold (5).
However, reasons why HS is lost specifically from the basolateral surface of intestinal epithelial cells at the onset of PLE remain unknown. Insufficiencies in genes involved in HS or HSPG synthesis, trafficking, or degradation may predispose for PLE, which then precipitates upon multiple additional hits. This becomes most apparent in CDG-Ib and CDG-Ic patients. Their genetically insufficient N-glycosylation might predispose HSPG loss since HSPG carry N-glycans, which can be important for synthesis, intracellular trafficking, or degradation. In fact, biopsies taken from CDG-Ic patients during episodes of PLE reveal a punctate intracellular GAG staining, indicating a disruption of HSPG trafficking. GAGs reappear on the basolateral surface when PLE resolves (2). In other primary diseases associated with PLE, other genes may be involved that predispose for HS loss and PLE. Understanding the reasons and mechanisms for HS loss may identify candidates likely to develop PLE before symptoms appear. Early treatment of these high risk candidates with heparin or heparin-like compounds may prevent PLE. Thus, the central role of HS in the patho-mechanisms of PLE has direct implications for patients with PLE.
Our results may have a broader implication beyond their importance for PLE. In our in vitro model, proteins leak from the basolateral to the apical but also just as well from the apical to the basolateral side (Fig. 7). These results suggest that the synergism of the identified key factors not only cause inside-out protein leakage but may also impair the intestinal barrier function as a first line of defense against outside-in translocation of bacteria and bacterial products such as LPS. Therefore, our findings may also have implications for intestinal inflammation or sepsis. In fact, intestinal biopsies from patients with inflammatory bowel diseases reveal the very same loss of HS specifically from the basolateral surface of the epithelial cells (39). The prospect of heparin treatment strongly warrants further investigations in this field.
We report the individual contributions and interactions of key factors involved in protein leakage in an in vitro model. However, animal models will be required to prove the importance of these four factors for the in vivo situation. Preliminary data with mice deficient in syndecan-1, the predominant HSPG in intestinal epithelial cells, indicate that HS loss elevates intestinal protein leakage. 4 In the future we will study whether these mice are more susceptible to viral infections, inflammation, and increased venous pressure and lead to PLE.