|Home | About | Journals | Submit | Contact Us | Français|
Hemolytic-uremic syndrome (HUS) caused by Shiga toxin-producing Escherichia coli infection is a leading cause of pediatric acute renal failure. Bacterial toxins produced in the gut enter the circulation and cause a systemic toxemia and targeted cell damage. It had been previously shown that injection of Shiga toxin 2 (Stx2) and lipopolysaccharide (LPS) caused signs and symptoms of HUS in mice, but the mechanism leading to renal failure remained uncharacterized. The current study elucidated that murine cells of the glomerular filtration barrier were unresponsive to Stx2 because they lacked the receptor glycosphingolipid globotriaosylceramide (Gb3) in vitro and in vivo. In contrast to the analogous human cells, Stx2 did not alter inflammatory kinase activity, cytokine release, or cell viability of the murine glomerular cells. However, murine renal cortical and medullary tubular cells expressed Gb3 and responded to Stx2 by undergoing apoptosis. Stx2-induced loss of functioning collecting ducts in vivo caused production of increased dilute urine, resulted in dehydration, and contributed to renal failure. Stx2-mediated renal dysfunction was ameliorated by administration of the nonselective caspase inhibitor Q-VD-OPH in vivo. Stx2 therefore targets the murine collecting duct, and this Stx2-induced injury can be blocked by inhibitors of apoptosis in vivo.
Shiga toxin-producing Escherichia coli (STEC) is the principal etiologic agent of diarrhea-associated hemolytic-uremic syndrome (HUS) (42, 60, 66). Renal disease is thought to be due to the combined action of Shiga toxins (Shiga toxin 1 [Stx1] and Stx2), the primary virulence factors of STEC, and bacterial lipopolysaccharide (LPS) on the renal glomeruli and tubules (6, 42, 60, 66). Of these, Stx2 is most frequently associated with the development of HUS (45). Shiga toxin enters susceptible cell types after binding to the cell surface receptor glycosphingolipid globotriaosylceramide (Gb3) and specifically depurinates the 28S rRNA, thereby inhibiting protein synthesis (42, 60, 66). The damage initiates a ribotoxic stress response consisting of mitogen-activated protein (MAP) kinase activation, and this response can be associated with cytokine release and cell death (21, 22, 25-27, 61, 69, 73). This cell death is often caspase-dependent apoptosis (18, 61). Gb3 is expressed by human glomerular endothelial cells, podocytes, and multiple tubular epithelial cell types, and damage markers for these cells can be detected in urine samples from HUS patients (10-12, 15, 49, 73). Shiga toxin binds to these cells in renal sections from HUS patients, and along with the typical fibrin-rich glomerular microangiopathy, biopsy sections demonstrate apoptosis of both glomerular and tubular cell types (9, 29, 31).
Concomitant development of the most prominent features of HUS: anemia, thrombocytopenia, and renal failure, requires both Shiga toxin and LPS in the murine model (30, 33). Nevertheless, our previous work demonstrated that renal failure is mediated exclusively by Stx2 (33). While it is established that Gb3 is the unique Shiga toxin receptor (46), the current literature regarding the mechanism by which Shiga toxin causes renal dysfunction in mice is inconsistent. Even though Gb3 has been localized to some murine renal tubules and tubular damage has been observed (19, 23, 46, 53, 65, 68, 72, 74), the specific types of tubules affected have been incompletely characterized. Although multiple groups have been unable to locate the Shiga toxin receptor Gb3 in glomeruli in murine renal sections (19, 53), one group has reported that murine glomerular podocytes possess Gb3 and respond to Stx2 in vitro (40), and another group has reported that renal tubular capillaries express the Gb3 receptor (46). Furthermore, murine glomerular abnormalities, including platelet and fibrin deposition, occur in some murine HUS models (28, 30, 33, 46, 59, 63). We demonstrate here that murine glomerular endothelial cells and podocytes are unresponsive to Stx2 because they do not produce the glycosphingolipid receptor Gb3 in vitro or in vivo. Further, murine renal tubules, including collecting ducts, express Gb3 and undergo Stx2-induced apoptosis, resulting in dysfunctional urine production and dehydration.
Shiga toxin 2 was purified by immunoaffinity chromatography from cell lysates (generously provided by Alison O'Brien) of E. coli DH5α containing the Stx2-producing pJES120 plasmid (39). Lysates were processed using 11E10 antibody (48) immobilized with an AminoLink Plus kit, and endotoxin was removed using De-toxi-Gel (Pierce Biotechnology, Rockford, IL). The purity of Stx2 was assessed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and determined to be endotoxin free by a Limulus amebocyte assay, and activity was measured in a Vero cell cytotoxicity assay. E. coli O55:B5 LPS purified by gel filtration chromatography and gamma irradiation was purchased from Sigma-Aldrich (St. Louis, MO).
Human umbilical vein endothelial cells (HUVEC) were purchased from VEC Technologies (Rensselaer, NY) and grown in MCDB 131 medium (Mediatech, Herndon, VA), supplemented as described previously (62). Primary human renal proximal tubule epithelial cells (RPTEC) were purchased from Clonetics (San Diego, CA) and grown in REGM (Lonza, Walkersville, MD) (73). Conditionally immortalized human glomerular endothelial cells were grown in EBM-2 MV medium with provided supplements (Lonza) (55). Primary murine aortic endothelial cells graciously provided by Lynn Hedrick (University of Virginia, Charlottesville) were grown in Dulbecco modified Eagle medium (DMEM) (Gibco, Grand Island, NY) and supplemented as described previously (7). Simian virus 40 (SV40) large-T-antigen-immortalized murine proximal tubular epithelial cells (TKPTS) kindly provided by Mark Okusa (University of Virginia, Charlottesville) were grown in DMEM:F12 (Gibco) supplemented as described previously (73). Conditionally immortalized human podocytes were grown in RPMI 1640 (Mediatech) with appropriate supplements (54). Conditionally immortalized murine podocytes generously provided by John Sedor (Case Western Reserve University, Cleveland, OH) and conditionally immortalized murine glomerular endothelial cells provided by Michael Madaio (Temple University, Philadelphia, PA) were cultured in a 3:1 mixture of DMEM:F12 (Gibco), supplemented as described previously (2, 4, 41).
Cells were grown at 37°C with 5% CO2 and 90% humidity in Falcon 75-cm2 flasks (BD Biosciences, Bedford, MA) except for the conditionally immortalized cell lines maintained at the permissive temperature (33°C). Only the conditionally immortalized murine cell lines were given 10 U of mouse gamma interferon (Sigma) per ml at 33°C. Conditionally immortalized cell lines maintained at the permissive 33°C were considered undifferentiated. Undifferentiated cells were moved to the nonpermissive temperature (37°C) (and gamma interferon was removed) 2 weeks prior to experimental use, after which point they were considered differentiated (2, 41, 54, 55). Cells were seeded at 5 × 105 per well in 6-well plates or 2 × 104 per well in 96-well plates. All experiments were performed on 6- or 96-well plates (Corning, Corning, NY) coated with rat tail collagen I (BD Biosciences) in serum-free RPMI 1640 (Mediatech). All experiments were performed in serum-free RPMI supplemented with l-glutamine. Except for cytotoxicity assays, cells were challenged with either no toxin, Stx2, 1 μg/ml LPS, or Stx2 and LPS, with 1 pM and 1 nM Stx2 employed for human glomerular and murine glomerular cells, respectively.
Each cell type grown to confluence in one 75-cm2 flask was trypsinized, and neutral glycolipids were isolated (51). For some studies, cells were incubated with 1 μg/ml LPS (E. coli O55:B5; Sigma-Aldrich) 24 h prior to trypsinization. Gb3 content was analyzed by thin-layer chromatography (TLC) with Stx1B overlay (43). Total neutral lipids on a duplicate plate were visualized using CuSO4 along with neutral glycosphingolipid standards (Matreya LLC, Pleasant Gap, PA) (1). Images are representative of triplicate experiments.
Cells were incubated with toxins for 0.5 to 12 h. After incubation, cells were rinsed twice with phosphate-buffered saline and lysed in modified radioimmunoprecipitation assay buffer containing 50 mM Tris, 150 mM NaCl, 1% Igepal CA-630, 0.5% deoxycholate, 100 μM leupeptin, 1 mM phenylmethylsulfonyl fluoride, 0.15 U/ml aprotinin, and 1 mM vanadate as previously described (56). Lysate protein was quantified with a BCA protein assay (Pierce). Total cell lysate was loaded at 5 μg per well, resolved by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to a polyvinylidene difluoride membrane, and probed with the following antibodies (according to the dilutions suggested by the manufacturer): anti-p38 MAP kinase and anti-phospho-p38 antibodies (BD Biosciences), anti-β-actin (Abcam, Cambridge, MA), and anti-SV40 T antigen (Calbiochem, San Diego, CA). Anti-mouse horseradish peroxidase-tagged antibody (Amersham, Piscataway, NJ) was used as the secondary antibody. Bound horseradish peroxidase was detected by chemiluminescence (Perkin Elmer, Waltham, MA). Images are representative of triplicate experiments.
Cells were treated with Stx2 at a concentration between 1 fM and 10 nM for 24 h. CCK-8 cell viability assays were performed to determine the 50% cytotoxic dose (CD50) (Dojindo Molecular Technologies, Gaithersburg, MD). Stx2 coincubation with 1 μg/ml LPS was tested in all cell types but enhanced only HUVEC cytotoxicity (62). Caspases were inhibited with 100 μM Q-VD-OPH (MP Biochemicals, Solon, OH) suspended in dimethyl sulfoxide (DMSO) for 1 h before and after the addition of Stx2 (8). The final concentration of DMSO was 0.5%. Data are from quadruplicate experiments.
Cells were incubated with toxins for 12 h. Extracellular monocyte chemoattractant protein 1, interleukin 6 (IL-6), vascular endothelial growth factor (VEGF), human IL-8, and murine CXCL1/KC release was quantified using species-specific DuoSet enzyme-linked immunosorbent assay kits (R&D Systems, Minneapolis, MN). Because mice do not produce IL-8, CXCL1/KC was used as the murine functional homolog (35). Extracellular proteins measured in picograms per milliliter were normalized to values from control cells (grown in media only) and expressed as changes (as percentages) from those control values. Data are from triplicate experiments.
Cells were challenged with toxins for 12 h. Lysates were collected as described above and tested for caspase activity with the caspase 3/7 assay kit (Upstate, Lake Placid, NY) using the fluorogenic caspase substrate Ac-DEVD-AMC (where Ac is N-acetyl, DEVD is Asp-Glu-Val-Asp, and AMC is 7-amino-4-methylcoumarin). Data are from triplicate experiments.
Male C57BL/6 mice weighing 22 to 24 g were purchased from Charles River (Wilmington, MA). Food and water were provided ad libitum. Mice were injected intraperitoneally with 300 μg of LPS (O55:B5; Sigma-Aldrich) per kg of body weight, 225 ng of Stx2/kg, or both as described previously (33). Mice were weighed every 12 h after injection, and weight loss was expressed as percent change from the weight of the mouse at time zero. At 0, 24, 48, 60, and 72 h after injection, mice were euthanized. The kidneys and blood and urine were collected from each mouse. To prevent apoptosis in vivo, mice were intraperitoneally injected with two 18-mg/kg doses of Q-VD-OPH in 100 μl of 50% DMSO at 24 and 48 h after injection of Stx2 plus LPS (38). Q-VD-OPH forms an irreversible thioether bond with the active site cysteine of the caspase, displacing the 2,6-diflurophenol group to inhibit caspase activity. No toxicity was observed in mice receiving only Q-VD-OPH. In separate experiments, healthy mice were dehydrated for 20 h by withholding access to water. Urine samples for volume measurements were collected from mice housed in individual metabolic cages. To determine Gb3 localization in mouse kidney, C3H/HeN and CD-1 mice (Charles River), C3H/HeJ and BALB/c mice (Jackson, Bar Harbor, ME), and C57BL/6 mice were used. All animal procedures were done in accordance with University of Virginia Animal Care and Use Committee policies (Charlottesville, VA).
Blood was collected with heparinized capillary tubes (Fisher, Pittsburgh, PA) by retro-orbital bleed and centrifuged at 840 × g for 15 min at 4°C to collect the plasma layer. The level of blood urea nitrogen (BUN) was determined spectrophotometrically with VetScan (Idexx Corporation, Westbrook, ME). Urine was collected by direct bladder puncture, and osmolality was determined with the Vapro vapor pressure osmometer (Wescor, Logan, UT). Each data point represents the average for eight mice.
C57BL/6, C3H/HeN, and CD-1 male mice were from Charles River (Wilmington, MA). C3H/HeJ and BALB/c mice were from Jackson (Bar Harbor, ME). Control C57BL/6 mice, C57BL/6 mice challenged with Stx2 plus LPS, and untreated CD-1, BALB/c, C3H/HeN, and C3H/HeJ mice were used. Kidneys were fixed in 4% paraformaldehyde, processed in acetone for Gb3 or ethanol for terminal deoxynucleotidyltransferase biotin-dUTP nick end labeling (TUNEL) staining, and embedded in paraffin as described previously (34). Ethanol has been previously demonstrated to both remove endogenous Gb3 and cause false-positive Gb3 staining (34). Sections were incubated with anti-Gb3/CD77 immunoglobulin M (IgM) (Beckman Coulter, Fullerton, CA) at a dilution of 1:40, isotype-matched rat IgM (Millipore, Billerica, MA), anti-activated caspase 3 (Cell Signaling, Danvers, MA) at a dilution of 1:200 or isotype-matched rabbit IgG (Chemicon), and anti-rat IgM biotin conjugate (American Qualex, San Clemente, CA) at a dilution of 1:500. TUNEL was performed with the ApopTag peroxidase in situ apoptosis detection kit (Chemicon, Temecula, CA), with postweaning female rat mammary gland as a positive control. Immunoreactivity was detected using Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA). Hematoxylin was the counterstain. Renal apoptosis was quantified by counting the number of apoptotic cell nuclei per 16 fields (at a magnification of ×200) spread among the cortex and outer and inner medulla. Renal apoptosis was expressed as the average number of apoptotic nuclei per kidney. Each data point represents the average for eight mice.
Kidney tissue was isolated following perfusion fixation in 4% paraformaldehyde. Sections were blocked with anti-rat IgM or secondary antibody-matched normal sera, incubated with anti-Gb3/CD77 at a dilution of 1:100, anti-CD31 (BD Pharmingen) at 1:50, anti-aquaporin-1 (AQP-1; Chemicon) at 1:1,000, anti-aquaporin-2 (AQP-2; Chemicon) at 1:800, or isotype control antibody (Chemicon) at the equivalent concentration in normal goat serum, washed, and incubated with fluorescent secondary antibodies (Invitrogen, Carlsbad, CA). Specimens were examined with an LSM 510 microscope (Zeiss, Thornwood, NY) and analyzed using LSM Image Browser software (Zeiss).
All data are expressed as means ± standard deviations. Statistics were performed using two-sample Student's t test assuming unequal variances. A P of <0.05 was considered significant.
Human HUS is thought to result from glomerular damage by Shiga toxin (10, 66). We began investigating the mechanism of renal failure in the mouse model by determining the sensitivity of murine glomerular filtration barrier cells to Stx2. Conditionally immortalized human and murine glomerular endothelial cells and podocytes have been described in detail (2, 41, 54, 55). These cells grow indefinitely at 33°C. When differentiated at 37°C, these cells express cell type-specific markers and slow their proliferation. Whole-cell lysates were derived from cells grown at permissive and nonpermissive temperatures and analyzed by immunoblotting (Fig. (Fig.1A).1A). These cells were compared to HUVEC that do not express the temperature-sensitive SV40 large T antigen. When incubated at the nonpermissive temperature (37°C), all conditionally immortalized cell types appropriately degraded the transgene and slowed or stopped proliferation. Differentiated human and murine podocytes also developed typical arborizations (data not shown) (41, 54).
To test the murine glomerular cells for the presence of the Stx2 receptor Gb3, cellular lipids were isolated and separated by TLC. Total neutral lipids were visualized on the TLC plate by staining with cupric sulfate (Fig. (Fig.1B,1B, bottom panel), and Gb3 was specifically identified by overlay with Stx1B (Fig. (Fig.1B,1B, top panel). Stx1B bound to no other glycolipid bands on these plates. Although the lipid profiles differed for the various cell types, similar loads were verified by the slowly migrating glycolipid band at the bottom of the TLC plate. Human glomerular endothelial cells and podocytes expressed very high levels of Gb3, and human RPTEC and HUVEC produced moderate levels, consistent with previous reports (15, 49). In contrast, murine glomerular endothelial cells, podocytes, and primary cells failed to express detectable Gb3. When these cells were incubated with LPS for 24 h to test for cytokine-mediated Gb3 upregulation, only HUVEC produced more Gb3 (data not shown) (62).
Stx2 has been demonstrated to have multiple effects on susceptible cell types; these effects include initiation of inflammatory intracellular signaling, cytokine release, and cellular apoptosis (21, 22, 25-27, 61, 69). Table Table11 summarizes the immortalization status and CD50 data for the human and mouse cells used in this study after the cells were treated with Stx2 for 24 h. Human glomerular cells were extremely sensitive to the cytotoxic effects of Stx2, while murine glomerular cells were insensitive to Stx2, even at a dose of 10 nM. Primary cells from both species were used as controls for the conditional immortalization. Primary HUVEC and RPTEC were sensitive, as previously reported, while primary murine aortic endothelial cells were not (62, 73). The sensitivities of cells to the cytotoxic effect of Stx2 generally correlated with their level of Gb3 expression (Table (Table11 and Fig. Fig.1B1B).
Stx2 increases inflammatory mediator release and MAP kinase activation in some cell types, even in the absence of cytotoxicity (25, 61, 69). Even though the murine cells were not sensitive to the cytotoxic effects of Stx2, they might respond by activating intracellular kinases or by releasing extracellular signaling molecules. To test this, human and murine glomerular cells were treated with Stx2 or LPS over a 12-h time course, and lysates were immunoblotted for total and activated p38 MAP kinase (Fig. (Fig.2)2) and JNK (Jun N-terminal protein kinase) (data not shown). p38 was activated in the human cells by both LPS and 1 pM Stx2, though the kinetics of activation differed. LPS caused a rapid increase in p38 activation from 1 to 2 h, while Stx2 began to exert its effect by 2 to 6 h and continued throughout the experiment. In contrast, the murine glomerular cells responded only to LPS, even at a Stx2 concentration of 1 nM (Fig. (Fig.2).2). The murine glomerular cells were more sensitive than the human glomerular cells to the effects of LPS, showing a more rapid response between 0.5 and 2 h, but no significant increase in p38 phosphorylation was observed in response to Stx2. The qualitative results of JNK activation due to Stx2 and LPS were similar to those of p38 activation in all cell types (data not shown).
Table Table22 presents the relative changes in extracellular signaling molecule release by human and murine glomerular cells after 12 h with Stx2 and LPS. LPS increased cytokine release into the supernatant by these cell types. In contrast, Stx2 affected only the Gb3-expressing human cells and did so by decreasing the signaling molecules released. Cytokines upregulated by LPS were reduced by Stx2, as was basal VEGF secretion by human podocytes. Stx2 mediated 15% ± 3% and 25% ± 5% cytotoxicity of human podocytes and endothelial cells, respectively, at the 24-h time point. Human glomerular cells incubated with the same dose of LPS but a 10-fold-lower dose of Stx2 (100 fM) did not exhibit significantly decreased cell viability or LPS-induced cytokine release by 12 h (data not shown). This suggested that the Stx2 inhibition of inflammatory mediator release was secondary to Stx2-mediated cell death.
Shiga toxin mediates caspase-dependent apoptotic cell death in certain cell types (18, 61). Caspases are cytoplasmic cysteine proteases essential to the destructive phase of apoptosis (67). Thus, activity of the major effector caspase 3 was measured in lysates from cells treated for 12 h with Stx2, LPS, or both. Caspase 3 activity was increased in human glomerular endothelial cells (Fig. (Fig.3A)3A) and podocytes (Fig. (Fig.3B)3B) in response to Stx2, but not in murine glomerular cells (data not shown). LPS did not have a significant impact on caspase 3 activity in any cell type (Fig. 3A and B). In human cells, the nonselective caspase inhibitor Q-VD-OPH rescued 80% of the cytotoxic effect of Stx2 on endothelial cells (Fig. (Fig.3C)3C) and 100% of the effect on podocytes (Fig. (Fig.3D).3D). Although higher and lower concentrations of the caspase inhibitor were tested, 100 μM provided the maximum nontoxic effect (data not shown).
It has previously been shown that Stx2 causes renal failure in mice (3, 33, 47). Having demonstrated that Stx2 does not directly affect the murine renal glomerular filtration barrier cells in vitro, we sought to determine the Stx2 target cells in the mouse kidney. Healthy mouse renal tissue subjected to immunohistochemistry with anti-Gb3 antibody demonstrated Gb3 only on cortical and medullary tubular cells and not in glomeruli or blood vessels (Fig. (Fig.4).4). Similar qualitative staining was observed in all mouse strains tested, including C57BL/6, CD-1, BALB/c, C3H/HeN, and C3H/HeJ mice (data not shown), as noted previously (19, 34). Morphologically, Gb3 appeared to localize to specific cell types in the three different areas of the murine kidney (Fig. (Fig.4B,4B, C, and D).
To identify the most abundant Gb3-positive cell types, immunofluorescence colocalization for Gb3 and aquaporin-1 (AQP-1) and AQP-2 were performed (16). Aquaporins are cytoplasmic and membrane proteins that mediate water reabsorption from the renal tubular lumen (16). Proximal tubules in the cortex and the thin descending loops of Henle in the medulla specifically express AQP-1, while collecting ducts in the cortex, medulla, and papilla express AQP-2 (16). Gb3 was expressed on some AQP-1-producing cortical proximal tubules (Fig. (Fig.5A),5A), although expression of the two markers appeared to be mostly localized to different sites within the same cell. It was noted that not every AQP1-positive proximal tubule expressed Gb3 (Fig. (Fig.5A),5A), and no medullary AQP-1-positive loops of Henle were Gb3 positive (Fig. (Fig.5B).5B). In contrast, some AQP-1-negative cortical tubules robustly expressed Gb3 (Fig. (Fig.5A).5A). High-power cortical and medullary images of Gb3 and AQP-2 showed colocalized staining: Gb3 was found on the outer membrane, consistent with antibody binding the outer carbohydrate moiety, and AQP-2 was distributed in both the membrane and cytoplasm (Fig. (Fig.6A).6A). Low-power images of the murine renal medulla stained for Gb3 and AQP-2 (Fig. (Fig.6B)6B) showed that almost all Gb3-expressing medullary tubules were collecting ducts. Gb3 staining did not colocalize with the endothelial marker CD31 in the mouse kidney (data not shown and reference 34).
To determine whether the cells found to produce Gb3 in vivo undergo apoptosis, TUNEL staining was performed on renal sections from mice 72 h after the mice were injected with Stx2 plus LPS. Positive TUNEL stain was found only in tubular cell nuclei (Fig. (Fig.7A,7A, inset), and only rare TUNEL-positive cells were observed in renal sections from healthy mice. Consistent with the absence of Gb3, there were no apoptotic cells visualized in the glomeruli or renal vasculature of mice challenged with Stx2 plus LPS. The qualitative immunohistochemical results for activated caspase 3 were similar to the qualitative TUNEL results (data not shown). Analysis of kidneys from mice between 0 and 72 h after Stx2-plus-LPS injection demonstrated increased TUNEL-positive cells at 60 and 72 h postinjection (Fig. (Fig.7A).7A). LPS alone did not increase renal apoptosis above the baseline level (data not shown).
As murine cortical tubular damage (23, 46, 53, 65, 68, 72, 74) and proximal tubular physiologic dysfunction manifest by glucosuria had been previously described (47), we determined the effect of collecting duct dysfunction. BUN values after 24 h in mice injected with Stx2 plus LPS increased with a time course similar to that of tubular apoptosis (Fig. (Fig.7B).7B). Previous studies demonstrated that LPS mediated the initial 24-h weight loss and Stx2 mediated the later weight loss in these mice (33). Because the collecting ducts are responsible for water reabsorption, we tested whether these mice had a defect in urine concentration that might cause dehydration and weight loss (16). Mice challenged with Stx2 plus LPS developed brief polyuria (increased urine volume) during the first 12 h that led to a compensatory increase in urine osmolality when measured at 24 h postinjection (Fig. (Fig.7C).7C). The initial increase in urine osmolality was reproduced by injection with LPS alone (data not shown), as published previously (17). Mice given Stx2 plus LPS or Stx2 alone developed polyuria and osmotically dilute urine between 48 and 72 h postinjection (Fig. (Fig.7C).7C). Stx2-plus-LPS-challenged mice produced 2.7 ± 0.8 ml of urine between 48 and 72 h compared to 1.1 ± 0.5 ml of urine from nonchallenged controls (P < 0.05; n = 8). Polyuria from 48 to 72 h correlated with the weight loss and observed signs of dehydration in mice given Stx2 plus LPS (Fig. (Fig.7D).7D). This decreased urine osmolality contrasts with 20-h dehydration of normal mice by water restriction. Water-restricted mice lost 10% ± 0.5% of their body weight and produced a minimal volume of urine, and all of the urine had a high osmolality of 3,485 ± 762 mmol/kg (n = 4).
To test whether tubular apoptosis caused collecting duct dysfunction, dehydration, and renal failure, mice injected with Stx2 plus LPS were given two divided doses of the nonselective apoptosis inhibitor Q-VD-OPH at 24 and 48 h. At 72 h after injection with Stx2 plus LPS, these mice demonstrated significantly decreased numbers of renal tubular apoptotic cells, increased urine osmolality, and decreased BUN levels (Fig. 8A to C). Although these mice developed the initial polyuria and weight loss mediated by LPS, they exhibited a significant recovery in body weight during the time when Stx2 normally caused a loss in body weight (Fig. (Fig.8D8D).
Previous studies examining the location of murine renal Gb3 have provided conflicting results (19, 34, 46, 53, 68). Our data support the conclusion that the primary Gb3-producing structure and Stx2 target in the murine kidney is the tubular system. We did not detect Gb3 expression by murine endothelial cells, and it is noteworthy that the previous study that reported murine renal endothelial production of Gb3 did not perform direct colocalization (46). Even though not all collecting duct cells appeared TUNEL positive at any single time point after Stx2 challenge, it is likely that more cells died than were visualized because apoptotic cells stain TUNEL positive for only 3 hours (20). Collecting duct dysfunction is in agreement with findings for other murine models of Shiga toxin-mediated injury and with microarray analysis in this model, which revealed Stx2-mediated downregulation of collecting duct-specific transcripts (9, 31, 33, 53, 58, 64). LPS has been previously shown not to cause tubular damage when administered at similar doses over this time course (23, 33). Although functional collecting duct damage in response to Shiga toxin was postulated in prior reports (47, 53), it was probably not observed because little morphological change occurs. In support of our findings, production of dilute urine has recently been reported for mice inoculated with STEC (13). The increased murine BUN level in response to Stx2 challenge may be secondary to dehydration caused by collecting duct dysfunction. Significant dehydration can decrease renal perfusion and raise the BUN value.
The present study confirmed that murine glomerular podocytes lack Gb3 and are insensitive to Stx2. Even though identical conditionally immortalized mouse podocytes were previously reported to produce Gb3 and respond to Stx2, we failed to detect Gb3 by a more specific method or to demonstrate a response to Stx2, even at 500 times the reported dose (40, 41). These cells are known to express TLR4 (Toll-like receptor 4) and release cytokines in response to LPS (5), and our cells responded to LPS by activating p38 in a time course similar to that detailed for Stx2 (40). Additionally, we have demonstrated that these cells lack Gb3 in vivo. Therefore, the effects previously ascribed to Stx2 in murine podocytes may be due to a small contaminating dose of LPS. Furthermore, the murine glomerular endothelial cells displayed similar responses to LPS and insensitivity to Stx2, suggesting that murine models reporting glomerular damage are likely due to LPS or indirect effects of Stx2; only those models that use live STEC or inject mice with Shiga toxin plus LPS observe glomerular defects (13, 28, 33, 59, 63). Although the Stx2-induced HUS mouse model lacks glomerular damage, we believe this difference from human disease does not preclude the utility of this system. Challenging mice with Stx2 plus LPS results in anemia, leukocytosis, thrombocytopenia, and cytokine-dependent fibrin deposition, and their relationships to HUS patient findings remain to be investigated (32, 33).
The human glomerular cells studied here were exquisitely sensitive to the cytotoxic effects of Stx2. Whereas it was previously reported that human glomerular epithelial cells were sensitive to Shiga toxin in vitro only at a much higher dose (27), the cells used prior were likely to be glomerular parietal epithelial cells rather than podocytes. This supposition is supported by their isolation using a sieving procedure shown to create cultures of nonpodocyte glomerular epithelial cells, their adoption of cobblestone as opposed to arborized morphology, and their lack of expression of the podocyte marker WT-1 (41, 75, 76).
Human tubular damage does occur in HUS patients, though the glomerular dysfunction may be predominant (10, 31). We showed that Stx2 is more toxic to human glomerular cells than to tubular cells. This supports studies that have failed to find cases of renal disease in the absence of microvascular and hemolytic symptoms following bloody diarrhea caused by STEC (37, 52). In contrast to the polyuria and dilute urine of the mice challenged with Stx2 plus LPS, most HUS patients are oligoanuric (66). However, two case reports detail Shiga toxin-mediated HUS associated with polyuria and persistent production of isosmotic urine (29, 57). Thus, direct tubular insult by Stx2 may participate in HUS-associated renal failure, and we hypothesize that collecting duct damage may facilitate dehydration that contributes to worse outcomes in some patients (24, 44). Although not without technical difficulty (66), testing prodromal HUS patients for urine-concentrating defects may identify those with severe disease and at greater risk for dehydration with a worse outcome.
The findings reported here have specific implications for understanding and treating human HUS. In contrast to the other human endothelial and epithelial cells described previously (22, 25-27, 70), the response of the human glomerular filtration barrier to Stx2 appeared distinctly noninflammatory. Despite causing a ribotoxic stress response in the human glomerular cells, Shiga toxin did not increase release of the inflammatory mediators tested. This may explain why HUS patients often report a fever during the diarrheal prodrome, presumably due to increased circulating inflammatory mediators, but are afebrile upon HUS presentation (36, 50, 66, 69, 70). However, Stx2 mediated a decline in human podocyte VEGF release. As decreased podocyte VEGF has been demonstrated to cause renal glomerular thrombotic microangiopathy in mice and in human patients, this mechanism of Shiga toxin-mediated reduction in VEGF may contribute to HUS clinically (14). Finally, we have also described how blocking apoptosis can rescue direct Stx2 renal insult in vivo and how Stx2-induced human glomerular endothelial and podocyte apoptosis can be inhibited by the same antiapoptotic agent in vitro. Thus, a clinically approved caspase inhibitor may be able to block Shiga toxin-mediated apoptosis in patients (9, 31, 71).
This research was supported by U.S. Public Health Service grants AI024431 and AI075778 (T.G.O.) and Wellcome Trust Fellowship 075731 (S.C.S.).
We thank Sanford Feldman for expert animal advice.
Editor: V. J. DiRita
Published ahead of print on 5 January 2009.