Mice lacking a functional cyclooxygenase-2 (COX-2) gene develop abnormal kidneys that contain hypoplastic glomeruli and reduced proximal tubular mass, and they often die of renal failure. A comparison of kidney-specific gene expression between wild-type and COX-2-deficient mice by cDNA microarrays revealed that although more than 500 mRNAs were differentially expressed between the two strains of mice depending on their ages, the genes encoding pre-pro-epidermal growth factor (pre-pro-EGF) and Tamm-Horsfall protein (THP)/uromodulin were aberrantly expressed in the kidneys of COX-2 −/− mice at all stages of their development. Downregulation of EGF could potentially affect renal development, and THP/uromodulin gene has been implicated in abnormal kidney development and end-stage renal failure in humans. We assessed in detail mechanism of defective THP/uromodulin gene expression and its potential consequences in COX-2-deficient mice. Consistent with the microarray data, the steady-state levels of THP/uromodulin mRNA were severely reduced in the COX-2 −/− kidney. Furthermore, reduced expression of renal THP/uromodulin, as assessed by Western blot and immunohistological methods, was closely corroborated by a corresponding decline in the urinary secretion of THP/uromodulin in COX-2 −/− mice. Finally, we demonstrate that the bladders of COX-2 −/− mice, in contrast to those of the wild-type mice, are highly susceptible to colonization by uropathogenic Escherichia coli.
- Escherichia coli
prostaglandins (PGs) are key regulators of renal pathophysiology. In the adult kidney, PGs modulate glomerular hemodynamics, tubular reabsorption of sodium and water, and secretion of renin (9, 19, 20, 51). Actions of PGs in the kidney are determined by cell-specific expression of enzymes that convert arachidonic acid (AA) into prostanoids and the types of eicosanoid receptors present on the target cells (19). The rate-limiting reaction in the biosynthesis of all PGs, conversion of AA to PGH2, is carried out by the cyclooxygenase (COX) and hydroperoxidase activities associated with the prostaglandin G/H synthase, PGHS (11, 14). The two isozymes, COX-1 and COX-2, are each encoded by a unique gene. Although COX-1 is constitutively expressed in most tissues, COX-2 expression is induced by proinflammatory and mitogenic stimuli (14, 45). Although both COX-1 and COX-2 catalyze the synthesis of PGH2, the two isozymes have distinct pathophysiological roles in vivo, as evident from the unique phenotypes of COX-1 and COX-2 gene knockout mice (1, 25).
Distinct cell-specific expression of COX-1 and COX-2 genes in the developing and adult kidney suggests that both COX isozymes contribute to the synthesis of renal PGs (20). COX-1 is expressed in the renal vasculature, glomeruli, and collecting tubules and is thought to be the constitutive source of PGs in the adult kidney. Inhibition of COX-1 by nonsteroidal anti-inflammatory drugs (NSAIDs) abrogates the protective effect of PGs on the blood vessels and causes renal ischemia. The highly regulated expression of COX-2 occurs mainly in the cortical juxtaglomerular apparatus and the epithelial cells of the thick ascending limb of Henle (TALH). Induction of COX-2 gene expression and biosynthesis of PGs in response to volume contraction triggers enhanced sodium reabsorption (20).
In addition to its involvement in the physiology of the adult kidney, COX-2 gene expression appears to be essential for normal renal morphogenesis. Ablation of the COX-2 gene in mice leads to anomalous development of kidneys that contain hypoplastic subcapsular glomeruli and reduced proximal tubular mass (27, 36). Due to incomplete cortical development, the remaining nephrons of COX-2 −/− kidneys are subject to increased filtration demands that cause cortical hypertrophy, glomerular sclerosis, and interstitial fibrosis, and these mice frequently die of renal failure (27, 36). The pathology of COX-2-deficient kidneys resembles the focal segmental glomerular sclerosis and chronic kidney failure seen in humans with oligomeganephronia (34). The offspring of pregnant mice treated with COX-2-selective inhibitor SC-58236 elicited renal pathology similar to that seen in COX-2 −/− mice (23).
Nephrogenesis in mouse is initiated in the embryonic day 8 (E8) stage embryo and continues until 7–10 days after birth. The reciprocal inductive transformations of the ureteric bud and the metanephric mesenchyme occur in distinct stages that include the condensation of metanephric blastema into aggregated cells, formation of renal vesicles, comma- and S-shaped bodies, and the development of the fully differentiated nephron (10, 13). Although several genes that control the process of nephrogenesis are known, the putative candidate genes that modulate the renal pathology in COX-2-ablated mice remain elusive. Therefore, a comparison of kidney-specific gene expression between wild-type (WT) and COX-2-ablated mice was done using cDNA microarrays with a goal to unravel the molecular basis of the renal pathology seen in COX-2 −/− mice. We show that the expression of 543 genes encoding regulators of metabolism, growth, and differentiation and inflammation were altered in the kidneys of COX-2 −/− mice. We also demonstrate that severely reduced renal synthesis and urinary secretion of Tamm-Horsfall protein (THP)/uromodulin in COX-2 −/− mice led to their greater susceptibility to urinary tract infections by uropathogenic Escherichia coli.
MATERIALS AND METHODS
Wild-type and PGHS-2 null (COX-2 −/−) C57BL/129-derived mice used in this study have been previously described (2, 27, 34). The mice were housed in Plexiglas cages at 25 ± 1°C and were kept on a 12:12-h light-dark cycle. At postnatal days 1, 7, 14, 21, or 29, mice were anesthetized by methoxyfluorane (Abbot Laboratories, North Chicago, IL) inhalation and then killed by CO2 suffocation according to the procedures approved by the VA Institutional Animal Care and Use Committee. Urine was collected from 3-wk-old mice by placing them individually in a collection chamber for 1 to 4 h to collect a minimum of 50 μl. The chamber consisted of a 500-ml glass jar containing two aluminum wire screens and fitted with a lid containing breathing holes. The bottom screen was convex and the upper screen was concave, which allowed the urine to be separated from the feces. The detailed methods for the treatment of mice with the COX-2-selective inhibitor SC-58236 in utero and during the perinatal period have been published (22) and were used without any modifications. Drug administration in the drinking water of the mother was initiated at E0.5 and continued for 3 wk after pups were born.
RNA and protein isolation.
The kidneys from the newborn and 1-, 7-, 14-, 21-, and 28-day-old mice were collected, homogenized in TRIzol Reagent (Gibco BRL, Rockville, MD), and stored at −80°C until RNA and protein extractions were carried out per the manufacturer's instructions. Briefly, a 20% volume of chloroform was added to the homogenate and shaken briskly. After a 15-min centrifugation (12,000 g) to separate the aqueous and phenol phases, the upper aqueous layer containing total RNA was removed to a clean tube. RNA precipitated with an equal volume of isopropanol was centrifuged for 10 min at 12,000 g. The RNA pellet was washed once with 75% ethanol, centrifuged, and after air-drying was dissolved in sterile water and stored at −80°C. A poly-A fraction from total RNA was isolated with an Oligotex mRNA Midi Kit (Qiagen, Los Angeles, CA). Purified poly-A RNA was quantified using RiboGreen dye (Molecular Probes, Eugene, OR) by a fluorescent assay, and the quality of RNA was determined by capillary electrophoresis with an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). We obtained a sufficient amount of RNA from the kidneys of individual mice at all time points except from the newborn and therefore two to three kidneys of 1-day-old mice were pooled. Although the profiles of kidney-specific gene expression of male and female mice were nearly identical (Thompson-Jaeger S, unpublished data), we used only male mice to minimize experimental variation. Microarray hybridization data were obtained from two to three replicates at each time point (Table 1).
The phenol phase and interface from the chloroform/homogenate were further processed to extract DNA and proteins. DNA from the phenol phase and interface was precipitated with ethanol (60% final volume), collected in a pellet by centrifugation at 2,000 g for 5 min, and discarded. The protein in the supernatant was precipitated by adding 3 vol of isopropanol, collected in a pallet by centrifugation; after three washes, protein pellets were air-dried and taken up in Laemmli sample buffer (24) without 2-mercaptoethanol or bromophenol blue. The amount of protein was determined by an adaptation of the Lowry method (38).
Preparation of fluorescent cDNA probes for hybridization.
The detailed protocols for the preparation of the probes and conditions of their hybridization to arrays have been previously described (54) and were used with minor modifications. Two-hundred nanograms of poly-A RNA were converted to either Cy3- or Cy5-labeled cDNAs using a labeling kit (Incyte Genomics, Fremont, CA). Each reaction contained 50 mM Tris·HCl, pH 8.3, 75 mM KCl, 15 mM MgCl2, 4 mM DTT, 2 mM dNTPs (0.5 mM each), 2 μg Cy3 or Cy5 random 9-mer (Trilink, San Diego, CA), 20 U RNase inhibitor (Ambion, Austin, TX), 200 U M-MLV RT RNase H (Promega, Madison, WI), and mRNA. Correspondingly labeled Cy3 and Cy5 cDNA products were combined and purified with a size exclusion column, concentrated by ethanol precipitation, and resuspended in hybridization buffer.
Analysis of mRNAs by comparative hybridization of fluorescent probes to cDNA micorarray chips.
Fluorescent-labeled cDNA probes taken up in 20 μl of 5× SSC, 0.1% SDS, and 1 mM DTT were hybridized to microarrays at 60°C for 6 h. The microarray slides were washed after hybridization in 1× SSC, 0.1% SDS, 1 mM DTT at 45°C for 10 min, and then in 0.1× SSC, 0.2% SDS, 1 mM DTT at room temperature for 3 min. After drying by centrifugation, arrays were scanned with an Axon GenePix 4000A fluorescence reader at 535 nm for Cy3 and 625 nm for Cy5 and GenePix image-acquisition software. An image analysis algorithm in GEMTools software (Incyte Genomics) was used to quantify signal and background intensity for each target element. Hybridization of labeled cDNA was performed on three different rodent cDNA arrays made by Incyte Genomics (Rat GEM1, Rat GEM2, and Mouse GEM1). The Mouse GEM1 contained 8,734 clones representing 7,450 unique genes and ESTs. The Rat GEM1 contained 7,510 clones representing 5,462 unique genes and ESTs, and the Rat GEM2 contained 8,478 clones representing 5,990 unique genes and ESTs. A total of 1,397 unique genes and ESTs were replicated between the two rat arrays. All clones were derived from sequence-verified libraries. The mouse pairs used to prepare probes for hybridization to three different arrays are shown in Table 1.
Data acquisition and interpretation.
A gradient correction algorithm (Qualifier, Novation Biosciences, San Mateo, CA) was used to correct for improper thresholds of array signals that produce plots that are rotated off-diagonal and curvilinear in form. The second correction algorithm was used to create a two-dimensional histogram over the Cy3 signal vs. Cy5 signal plot with frequency of occurrence as the third axis that was transformed into a linear straight line and then rotated to the diagonal position.
To identify genes for further analysis, Cy3 signal vs. Cy5 signal was plotted for each array and elements demonstrating differential expression, i.e., elements visually distinct from the predominant group of genes that were not differentially expressed (slope ∼1), were identified graphically using Spotfire (Somerville, MA). Complementary DNA elements on the array that were “differentially expressed genes” in the mouse replicates and showed consistent pattern (up or down) across biological replicates were then identified. A modified Centroid method of hierarchical clustering (JMP, SAS Institute, Cary, NC) was used to group differentially expressed genes into 30 clusters. The modification requires expression precision to limit cluster identification and resolution (Novation Biosciences). Genes from each cluster demonstrating >+1.7- or <−1.7-fold differential expression ratios at any time point were then chosen for further analysis (54). Cluster analysis was used to identify a temporal relationship among genes that were differentially expressed across biological replicates.
Analysis of steady-state levels of mRNAs by Northern blotting and RT-PCR.
Twenty micrograms of total RNA extracted from the kidneys of WT or COX-2 −/− mice at different ages, as described above, were size fractionated on denaturing formaldehyde-agarose gels. RNA was transferred to nylon membranes and hybridized to 32P-labeled cDNA probes prepared from mouse-specific THP or β-actin cDNAs. The detailed methods for the Northern blot analysis have been described previously (26) and were used with minor modifications. For RT-PCR analysis, THP-specific target and competitor cDNAs were cloned with primers that were designed from the published cDNA sequences (56). Kidney- or bladder-specific mRNA was reverse transcribed, cDNAs were added to a 25-μl reaction mix containing Ready To Go PCR Beads (Amersham Pharmacia Biotech, Piscataway, NJ), and amplified in duplicate with a programmable thermal controller model PTC-100. All RT-PCR data were normalized to an internal standard as outlined in detail previously (55).
Proteins were separated by SDS-PAGE and transferred to PVDF membranes (Millipore, Bedford, MA) in transfer buffer (0.2 M glycine, 25 mM Tris base, 20% methanol) at 90 V for 3 h with continuous cooling. Membranes were blocked in 5% defatted powder milk in TBST (20 mM Tris·HCl, pH 8.0, 150 mM NaCl, 0.05% Tween 20). Primary antibodies were diluted in 2% defatted milk in TBST and incubated at room temperature for 2 h or overnight at 4°C. Following three 5-min washes of the membrane in TBST, secondary antibody (diluted 1:2,000 to 1:10,000) was incubated in 1% nonfat milk in TBST for 1 h at room temperature with shaking. After five washes of the membrane in TBST for 10 min each, proteins were detected by with the SuperSignal Substrate from Pierce (Rockford, IL). Human anti-THP antibody (Biomedical Technologies, Stoughton, MA) was used as the primary antibody and anti-rabbit IgG conjugated with horseradish peroxidase (Santa Cruz Biotechnology) was used as a secondary antibody.
One-day-old mice were killed, and the kidneys were removed and immediately embedded in Tissue-Tek (Miles, Elkhart, CA) with OCT medium and flash-frozen in methylethylbutane on dry ice. Older mice were fixed by perfusion with 20 mM PBS followed by 4% formaldehyde in 0.1 M PBS through the left ventricle, under anesthesia. Kidneys were removed, flash-frozen, and sections (10–14 μm in thickness) were cut from tissue blocks in a cryostat, mounted on precoated glass slides, and air-dried. Sections were preincubated in blocking solution [PBS with 10% goat serum and 0.5% Triton X-100 (Sigma, St. Louis, MO)] for 1 h at room temperature, washed in PBS, and then incubated overnight with rabbit anti-human THP antibodies followed by Texas red-conjugated anti-rabbit IgG (1:200 dilution; Jackson ImmunoResearch, West Grove, PA) for 1 h at room temperature. For negative controls, sections were incubated with nonspecific IgG as a primary antibody followed by Texas red-conjugated secondary antibody. Coverslips were mounted on glass slides with anti-fade mounting medium (Molecular Probes) and slides were viewed under a Zeiss Axiophot microscope with fluorescent filter coupled to a Dell Windows 95 PC. Multiple random images were taken from each slide and the images were mainpulated with Adobe Photoshop (Adobe Systems); some tissue sections were also seen by confocal microscopy.
Inoculation of bladders with E. coli and measurement of bacterial adhesion.
The recombinant strains were constructed in E. coli strain AAEC191A, a derivative of the archetypal K-12 E. coli strain MG-1655, which has had the entire fim gene cluster deleted by allelic exchange (4). AAEC191A was transformed with plasmid pPKL114 to create strain KB18 (48). Plasmid pPKL114 (21) contains the entire fim gene cluster from the K12 strain PC31 but has a translational stop-linker inserted in the 5′-end of the fimH gene. Strain KB54 was created by transforming strain KB18 with a fimH gene cloned from uropathogenic E. coli (48). Strain KB18 does not express functional type 1 fimbriae due to its inability to express the FimH lectin. Strain KB54 expresses fully functional type 1 fimbriae, which mediate high levels of adhesion to J82 human bladder epithelial cells (46, 48).
Bacteria were cultured in Luria-Bertani (LB) broth at 37°C for 16 h, washed twice in PBS, and suspended to a density of 109 colony forming units/ml. Female WT or COX-2 −/− mice were anesthetized and inoculated via transurethral catheterization with 50 μl of the bacterial suspension as described previously (47). Six hours after inoculation, mice were killed by cervical dislocation and their bladders were aseptically removed, weighed, and homogenized in 1.0 ml of PBS containing 0.025% Triton X-100. Whenever possible, urine released at the time of death was also collected for subsequent analysis. Titers of bacteria were determined by plating serial dilutions of the bladder homogenates or urine on LB agar plates (47).
With a goal to identify candidate gene(s) associated with renal pathology in COX-2 −/− mice, we analyzed mRNAs from WT and COX-2 −/− kidneys by high throughput gene expression profiling using Incyte Genomics GEM microarrays. Regardless of the types of nucleic acids fabricated on the microarray chips (i.e., cDNA vs. oligonucleotides), the intrinsic sequence variations from one gene to another and variable degree of hybridization between replicate samples are two common problems encountered in the use of microarrays (54). An accurate assessment of gene expression by microarray-based methods may be further confounded by the intrinsic variability among the individual mice siblings (39). Therefore, we carried out independent analysis of gene expression from two to three mice at each time point. Furthermore, because hybridizations in duplicate are known to enhance the reproducibility and decrease the number of false positives while hybridizations in triplicate seemed to offer little improvement (15), we also carried out all hybridizations in duplicate. Thus our experimental strategy was optimized to efficiently use the limited amount of kidney-specific RNA available, especially from the newborn mice.
Comparison of the rat and mouse cDNA arrays to assess gene expression in mice.
Because the Incyte Mouse GEM1 arrays contained only 7,450 unique genes and ESTs, we assessed the potential of Incyte Rat GEM1 and Rat GEM2 arrays (see materials and methods). The Rattus norvegicus and Mus musculus orthologs share more than 90% sequence identity in the coding regions and ∼90% in the 3′-untranslated regions (29). Therefore, we expect near perfect hybridization between mouse and rat cDNAs under our experimental conditions (17).
To analyze the levels of interspecific hybridization, we labeled mouse liver RNA with both Cy3 and Cy5 and hybridized these probes to the Mouse GEM1 arrays. Similarly, we labeled rat liver RNA with Cy3 and Cy5 and hybridized the fluorescent probes to the Mouse GEM1 arrays. As shown in Fig. 1A, the ranges of signal intensities for both mouse and rat RNAs on Mouse GEM1 array are very comparable. To further validate this experimental approach, mouse liver RNA labeled with Cy3 and rat liver RNA labeled with Cy5 were hybridized to Rat GEM2 arrays. The vast majority of genes align on the line with a slope of about 1 across the entire dynamic range of the signal (Fig. 1B), indicating that the level of expression of most liver-specific genes is comparable in both species. However, compared with the signals produced by rat liver cDNA probes, a subset of mouse liver cDNAs produced more intense signals on Rat GEM2 arrays. Because the rat liver cDNA sequences are identical to the sequences present on the Rat GEM1/2 arrays, more intense signals from mouse liver indicate higher expression of these transcripts in mouse liver. Similarly, greater intensity of rat liver-specific signals on rat GEM2 array may represent quantitative differences in the expression of some transcripts between rat and mouse livers. Alternately, the more intense signals from rat may indicate sequences in the rat genome that have significantly diverged from the mouse. It is noteworthy that the ratios of signal intensities of the differentially expressed genes between WT and COX-2 −/− kidneys, as we deduced from hybridization on Rat GEM2 arrays, were nearly identical from both rat and mouse RNA (Fig. 1B, red boxes). Our experimental observations of interspecific hybridizations generally support the predictions based on the genomic homology between mouse and rat. Nevertheless, we posit that even if a mouse probe produces lower signal than the rat probe on a rat GEM array, when considering mouse vs. mouse hybridization, that given sufficiently strong and reproducible signal above the background for confident detection, differential expression should be readily detectable.
To extend the rigor of our method to identify differentially expressed genes from mouse kidneys using rat cDNA GEM arrays, we compared hybridization of probes made from mouse brain vs. liver RNAs and probes from rat brain vs. liver RNAs on Rat GEM2 arrays. The differential expression ratios between rat brain and liver RNA hybridized to Rat GEM2 arrays were compared with the expression ratios of mouse brain vs. liver RNA derived from hybridization of these probes to a Rat GEM2 array (Fig. 1C). As would be predicted, distribution of most of the differentially expressed genes between brain and liver from the two species is virtually identical. Again, we note that the genetic elements that were deduced from these analyses to differ in their expression between WT and COX-2 −/− kidneys are located on the diagonal line with a slope of one (Fig. 1C, red boxes). Therefore, we conclude that it should be experimentally feasible to identify differentially expressed genes by our strategy to hybridize mouse-specific probes to rat GEM arrays.
Age-dependent regulation of kidney-specific gene expression in COX-2 −/− mice.
Replicate mRNAs from the kidneys of COX-2 −/− and WT mice were compared by hybridization to mouse and rat GEM arrays (Table 1). Differentially expressed genes identified from the three GEM types were then queried for elements that showed a consistent pattern of expression across biological replicates at every time point, i.e., always up- or downregulated. We identified 543 cDNA elements that were differentially expressed between the WT and COX-2 −/− mouse kidney; the differentially expressed elements contained 44 Incyte/public domain ESTs and 39 ESTs with similarities to known genes. We grouped a subset of differentially expressed (± 1.7-fold) 205 genes according to their biological functions (Table 2).
We assessed, by RT-PCR and Northern blot analysis, 25 genes to validate microarray data and found excellent correspondence between microarray and RT-PCR data (Dou W, unpublished data). We began our analysis with the assumption that the genes affected in the newborn and 7-day-old COX-2 −/− mice may be directly involved in the renal pathology (Fig. 2), whereas many more genes may be secondarily affected as the renal pathology evolved with age (Table 2). Consistent with this expectation, expression of only four genes, THP/uromodulin, kidney androgen-regulated protein (KAP), dipeptidase (DPEP1), and phenylalanine hydroxylase (PH), was significantly altered in the kidneys of the newborn COX-2 −/− mice. One factor that links the aberrantly expressed genes at day 1 is the metabolism of thyroid hormone. Thus mRNA for PH that catalyzes the synthesis of tyrosine (a precursor of thyroid hormones T3/T4) from phenylalanine is lower in COX-2 −/− mice kidneys at all ages, but most dramatically on days 1, 7, and 14. Similarly, DPEP1, involved in the processing of T3/T4 from thyroglobulin, is lower in COX-2 −/− mice. Finally, mRNA encoding transthyretin, implicated in the transport of T3/T4 (44), is elevated in the kidneys of the newborn COX-2 −/− mice (Fig. 2), perhaps reflecting a compensatory expression in response to altered metabolism of thyroid hormone.
Three growth and differentiation factors, pleiotrophin/heparin-binding growth-associated molecule, epidermal growth factor (EGF), and β-catenin, were also abnormally expressed in COX-2 −/− mice (Fig. 2). The altered expression of pleiotrophin, a key regulator of branching morphogenesis of the ureteric bud (42), may reflect the retarded glomerulogenesis seen in COX-2 −/− kidneys. The reduced expression of EGF in the COX-2-ablated kidneys is similar to that seen in renal injury caused by ischemia (18). In light of the known interactions between EGF and PGs (40), we validated the microarray data and found reduced expression of pre-pro-EGF in the COX-2 −/− kidneys (Dou W, unpublished data). The reduced expression pre-pro-EGF and β-catenin continued to occur in COX-2 −/− kidneys with age as judged by RT-PCR, Western blot, and immunohistological analyses (Dou W and Thompson-Jaeger S, unpublished data).
PGs are known regulators of immunity and inflammation (1, 41). We found that a number of genes involved in immunity and inflammation such as class II MHC proteins, KAP, osteopontin, and DNase I were abnormally regulated in COX-2 −/− mice. KAP is abundantly expressed in the proximal tubules (16) and is strongly downregulated in the kidneys of COX-2 −/− mice. KAP is known to bind to the immunosuppressive drug cyclosporine A (CsA)-binding protein and overexpression of KAP was shown to reduce the toxicity of CsA in the proximal tubule cells in vitro (7). The enhanced expression of OPN in COX −/− mice is particularly noteworthy. OPN is mainly present in the loop of Henle and distal nephrons and its synthesis in the kidney is enhanced during reperfusion following renal ischemia (35, 52). Consistent with its proposed role in inflammation (8), OPN-ablated mice elicit reduced toxicity to CsA (32).
The functional profile of genes affected in the kidneys of 21- and 29-day-old mice reveals the evolving renal pathology of COX-2-ablated mice as they age. Thus genes encoding collagen types I, III, and IV, fibronectin, and matrix Gla protein were highly expressed in the kidneys of older animals. Increased synthesis and deposition of fibronectin, laminin, and interstitial collagens are a hallmark of various nephropathies in experimental animals and humans (30). Therefore, enhanced renal expression of extracellular matrix-associated proteins is predictable as the majority of the surviving COX −/− mice develop renal fibrosis with age (28).
Because PGs are vasodilatory in the kidney, it is conceivable that COX-2-deficient mice elicit diminished renal vascular tone and consequent hypoxia. It is significant therefore that the expression of a number of mitochondrial genes involved in the oxidation of amino acids, lipids, and carbohydrates and critical for the generation of ATP was severely reduced in COX-2 −/− mice (Table 2). Because gluconeogenesis is of critical importance for renal function, it was significant to note that the levels of fructose 1,6 bisphosphatase and phosphoenol pyruvate carboxy kinase (PEPCK), key components of gluconeogenesis, were severely reduced in the kidneys of COX-2 −/− animals. It is known that treatment with NSAIDs can lead to uncoupling of oxidative phosphorylation (5, 31) and a decline in the rate of ATP synthesis (50).
Aberrant expression of THP/uromodulin in COX-2 −/− kidneys.
The cDNA microarray hybridization data revealed that mRNA encoding THP/uromodulin was severely reduced in COX-2 null kidneys, regardless of their age (Fig. 2). Because THP/uromodulin has been implicated in a variety of immune and inflammatory reactions, we compared its regulation in WT and COX-2 −/− mice in greater detail. The total kidney-specific mRNAs from WT or COX-2 −/− mice were hybridized to THP/uromodulin-specific cDNA probes on Northern blots (Fig. 3A). The steady-state levels of THP/uromodulin mRNA were lower in the kidneys of the newborn COX-2 −/− mice. Although not shown in Fig. 3A, a 15- to 20-fold decline in THP/uromodulin mRNA in COX-2 −/− kidneys was consistently observed. Furthermore, the reduced levels of THP/uromodulin mRNA were highly specific as the steady-state levels of β-actin mRNA remained similar in the kidneys of WT and COX-2 −/− mice (Fig. 3A).
To determine whether synthesis of THP/uromodulin protein was also defective in COX-2 −/− kidneys, we assessed kidney-specific accumulation of THP by Western blot analysis and by staining sections of WT and COX-2-ablated mouse kidneys. A comparison of two WT and two COX-2 −/− renal homogenates is shown in Fig. 3B; the anti-THP/uromodulin antibody readily detected a 90-kDa THP-specific polypeptide band in the two samples from the WT mice, whereas a 90-kDa protein was barley detectable in the COX-2 −/− samples (Fig. 3B). A 15- to 20-fold decline in the immunodetectable THP/uromodulin in COX-2 −/− renal homogenates was observed in four independent experiments (Dou W, data not shown). THP/uromodulin is known to localize mainly in the TALH where it is inserted into the luminal surface of the epithelial cells via a GPI anchor that is cleaved by proteases to release THP in the urine (6). Consistent with the published data (6), the THP/uromodulin-specific immunostaining was readily detected in the renal tubules concentrated in the deep cortex and medulla of the WT kidneys (Fig. 4). In contrast, the staining for immunoreactive THP/uromodulin in the kidneys of COX-2 −/− mice (Fig. 4) was not significantly above the background observed with nonspecific IgG as the primary antibody (Dou W, unpublished observations). Thus both the Western blotting and immunohistological data corroborate the levels of THP-specific mRNA in WT and COX-2 −/− mice kidneys, as assessed by either cDNA microarrays or by Northern blot analysis.
Consistent with the published data showing that THP/uromodulin is a major urinary protein (56), an intense THP/uromodulin-specific 90-kDa polypeptide band could be readily detected by Western blot in the samples of WT urine (Fig. 5). In contrast, the urine of COX-2 −/− mice was nearly completely devoid of THP/uromodulin (Fig. 5). We also found that the steady-state levels of urinary THP/uromodulin were not significantly different between COX-1 −/− and WT mice (Fig. 5) (Dou W, unpublished data). These data indicate that the steady-state levels of THP/uromodulin mRNA and its translation in the kidney and the excretion of THP/uromodulin in the urine are coordinately regulated in the WT and COX-2 −/− mice.
Because THP/uromodulin has been implicated in the mechanism of adhesion of type I fimbriated E. coli to the bladder epithelium (37), we assessed whether the severely reduced biosynthesis of THP in the kidneys of the COX-2 −/− mice and its diminished levels in the urine resulted in an altered response of bladders to colonization by E. coli. The results of a representative experiment shown in Fig. 6 reveal that the bladders of COX-2 −/− mice elicited about fivefold greater susceptibility to colonization by KB54, a uropathogenic strain of E. coli. In contrast, the colonization of WT and COX-2 −/− bladders by KB18, a strain of E. coli that lacks the FimH adhesin implicated in THP/uromodulin epithelial cell interaction, was not statistically significant (Fig. 6). The binding of the FimH adhesin to the epithelial cell is mediated via oligosaccharides present on THP/uromodulin (37). Consistent with the proposed biochemical mechanism of bacterial adhesion mediated by THP/uromodulin, the colonization of bladders by KB54 strain of E. coli was dramatically reduced if 50 mM methylmannoside was included in the bacterial inoculum. Furthermore, consistent with the predicted mechanism of THP/uromodulin-mediated binding of E. coli to the epithelial cells, the presence of 50 mM methylmannoside had no effect on the binding of the FimH− strain KB18 to the bladder (Dou W, unpublished data).
Treatment of mice with COX-2-selective inhibitor SC-58236 during gestation and perinatal growth is known to induce renal histological changes that are similar to what occurs in mice lacking a functional COX-2 (22). A detailed assessment of renal function in SC-58236-treated mice has not been done. To test whether THP expression was also altered in the mice treated with COX-2-selective inhibitor, we quantified the steady-state levels of urinary THP by Western blot followed by densitometry. Proteins from an equivalent volume of urine from control and SC-58236-treated mice, normalized against urinary creatinine, were size-fractionated by SDS-PAGE and analyzed. Parallel quantifications of THP-specific mRNA from kidneys of untreated or SC-58236-treated mice were also carried out by RT-PCR. The levels of excreted THP in the untreated and drug-treated animals were almost indistinguishable (Table 3). In contrast to the severely reduced expression of THP mRNA seen in the COX-2-ablated kidneys, the levels of THP-specific mRNA in the kidneys of SC-58236-treated mice were only modestly reduced; a 20–30% reduction was seen compared with controls (Table 3).
A congenital lack of the inducible COX-2 gene expression leads to defective renal development and pathology in mice. We undertook this study to test the hypothesis that the aberrant renal development seen in COX-2-ablated mice is likely due to altered gene expression as a result of aberrant metabolism of AA in the absence of the inducible isozyme of COX. A comparison of the transcriptomes of the control and COX-2 −/− kidneys by high throughput cDNA microarray hybridization revealed two aspects of the renal pathology at the molecular level. First, our data suggest that the reduced glomerulogenesis and concomitant accumulation of undifferentiated mesenchyme seen in the kidneys of COX-2 −/− mice may have resulted from defective expression of pre-pro-EGF and β-catenin, two proteins thought to be critical in the development of the kidney. Although our experiments shed little light on the potential mechanism by which lack of COX-2 expression results in altered regulation of pre-pro-EGF and β-catenin, we speculate that altered AA metabolites elicited in the absence of a functional COX-2 isozyme may directly or indirectly regulate these two developmental morphogens in the developing kidney.
Second, our cDNA microarray hybridization analysis revealed that the kidneys of the COX-2 knockout mice elicited defective expression of THP gene and reduced urinary excretion of THP. As a direct result of the reduced expression of THP, the bladders and kidneys of the COX-2-ablated mice were more susceptible to colonization by uropathogenic E. coli. Our data raised the possibility that COX-2-selective NSAIDs may also lead to aberrant expression of THP and concomitant susceptibility of mice to uropathogenic infections. It has been demonstrated previously that mice exposed to the COX-2-selective inhibitors during gestation and in the first 2–3 wk of perinatal growth developed kidney lesions reminiscent of the pathology seen in the COX-2 −/− mice (22). Therefore, we were surprised to learn that mice exposed to COX-2-selective inhibitor in utero showed only modest decrease in the expression of THP-specific mRNA in the kidneys and negligible reduction in the level of urinary excretion of THP. At present, we can only speculate as to why the renal consequences of the congenital lack of COX-2 and its inhibition by SC-58236 are not identical.
DNA microarray-based strategies have been successfully used to elucidate global gene expression in the developing kidney (49) and to unravel altered gene expressions that accompany the pathophysiology of the Alport's syndrome (43) and renal ischemia-reperfusion injury (53). We are first to demonstrate that the renal development of mice lacking a functional COX-2 gene may be initiated by an aberrant regulation of pre-pro-EGF and THP/uromodulin genes that lead to a cascade of age-dependent changes as the renal pathology evolves in the COX-2 −/− mice. A direct role for THP-uromodulin in urinary tract infection has been elegantly demonstrated by two independent studies of THP knockout mice (3, 33). Interestingly, mice exposed to selective COX-2 inhibitors during the first 2 wk after birth develop renal deficits similar to those seen in COX-2-deficient mice (22) but show only modest effect on the expression of THP. We posit that more thorough investigation of the effects of COX-2-selective NSAIDs on the morphogenesis of kidney and its homeostasis in the humans may be warranted.
These studies were supported by grants from the Department of Veterans Affairs (DVA), National Institutes of Health, the National Kidney Foundation, and the Center of Excellence for the Diseases of Connective Tissue of the University of Tennessee Health Science Center. R. Raghow and D. L. Hasty are Senior Research Career Scientists of DVA.
We thank J. M. Minor of Novation Biosciences, who assisted in removing nonbiological and manufacturing patterns from the microarray data. We appreciate S. Goorha for help in maintaining the breeding colony of COX-deficient mice, J. Fountain for secretarial help, and T. Higgins for help with illustrations.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2005 the American Physiological Society