- Email: [email protected]
A Combined Transcriptome and Bioinformatics Approach to Unilateral Ureteral Obstructive Uropathy in the Fetal Sheep Model
A Combined Transcriptome and Bioinformatics Approach to Unilateral Ureteral Obstructive Uropathy in the Fetal Sheep Model Alexander Springer, Klaus Kratochwill, Helga Bergmeister, Dagmar Csaicsich, Johann Huber, Martin Bilban, Bernd Mayer, Irmgard Mühlberger, Gabriele Amann, Ernst Horcher and Christoph Aufricht* From the Departments of Pediatric Surgery (AS, EH), Pediatrics (KK, DC, CA), Laboratory Medicine (MB) and Pathology (GA) and Division of Biomedical Research (HB), Medical University of Vienna, Section Ruminants, Education and Research Farm, University of Veterinary Medicine Vienna (JH) and Emergentec Biodevelopment GmbH (BM, IM), Vienna, Austria
Purpose: Fetal obstructive uropathy is a leading cause of loss of renal function. Characterizing the molecular fingerprint of cellular responses to obstruction in a fetal model of complete unilateral ureteral obstruction may help elucidate the activated mechanisms and suggest new therapeutic interventions. Material and Methods: Unilateral ureteral obstruction was created in 3 sheep fetuses at day 60 of gestation. For transcriptome analysis total RNA was extracted from vital renal biopsies 2 weeks after intervention from obstructed kidneys and from control kidneys of untreated twins. cDNA preparation, hybridization to the GeneChip® Bovine Genome Array and array scanning were done according to manufacturer protocols. Bioinformatics analysis was used to derive functional biological processes linked to obstructive uropathy. Quantitative reverse-transcriptase-polymerase chain reaction and immunohistochemistry were used to validate microarray results. Results: Seven biological processes were identified as significantly affected by differentially regulated features that characterize unilateral ureteral obstruction, namely protein metabolism and modification, other metabolism, neuronal activity, ligand mediated signaling, amino acid metabolism, coenzyme/ prosthetic group metabolism and rRNA metabolism. Literature mining identified 17 candidate genes previously reported as key in the context of unilateral ureteral obstruction, related pathological mechanisms or other kidney diseases. Conclusions: Combined transcriptome and bioinformatics analysis allowed the identification of enriched processes in the fetal sheep model of unilateral ureteral obstruction that are likely associated with renal damage but to our knowledge have not been previously identified. Future clarification of these molecular fingerprints may eventually provide therapeutic targets and early predictive markers involved in the pathogenesis of fetal uropathy.
Abbreviations and Acronyms PANTHER ⫽ Protein Analysis Through Evolutionary Relationships PCR ⫽ polymerase chain reaction qRT-PCR ⫽ quantitative RT-PCR RT-PCR ⫽ reverse transcriptasePCR UUO ⫽ unilateral ureteral obstruction Submitted for publication March 21, 2011. Study received approval from the Medical University of Vienna animal ethics committee and Austrian Federal Ministry of Science and Research. Supported by Grant 30000EUR from the Mayor of the City of Vienna (AS). Supplementary material for this article can be obtained at http://www.hypospadie.info/supplement_ ju/supplementary_material.pdf. * Correspondence: Department of Pediatric Nephrology, Medical University of Vienna, Waehringer Guertel 18-20, 1090 Vienna, Austria (telephone: 0043-1-40400-2908; FAX: 0043-1-40400-6812; e-mail: [email protected]).
Key Words: ureter, ureteral obstruction, gene expression, fetus, sheep CONGENITAL UUO is a leading cause of renal impairment in childhood. The cellular and molecular mechanisms of uropathy are complex and classically attributed to the triad of apoptosis,
interstitial inflammation and fibrosis.1 Fetal sheep uropathy models are generally accepted to be the best available models. The methodology is estab-
0022-5347/12/1872-0751/0 THE JOURNAL OF UROLOGY® © 2012 by AMERICAN UROLOGICAL ASSOCIATION EDUCATION
Vol. 187, 751-756, February 2012 Printed in U.S.A. DOI:10.1016/j.juro.2011.09.148
AND
RESEARCH, INC.
www.jurology.com
751
752
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
lished and comparable to humans, thus, forming an excellent basis on which to build on new technologies and opening ways for experimental surgical intervention of fetal uropathy.2 Transcriptome analysis has developed into a powerful tool to contrast gene expression in experimental and clinical settings.3 Recently cDNA microarray analysis was used to examine gene expression profiles in rodent models of postnatal UUO, allowing extensive analysis of the relative expression of a large number of protein coding genes.4 However, extracting clear, coherent hypotheses from genome wide expression data remains an important challenge.5 Bioinformatics analysis is used to transfer these vast amounts of novel insights into cellular mechanisms into biologically relevant findings. To our knowledge we present the first combined application of transcriptome and bioinformatics analysis in the fetal sheep model of UUO to characterize gene profiles. Our specific focus was on the main established pathological mechanisms but we also searched for novel molecular mechanisms with potential biological significance.
MATERIALS AND METHODS Fetal Sheep Model of UUO All experimental procedures were done in accordance with the Medical University of Vienna animal ethics committee and the Austrian Federal Ministry of Science and Research. Studies were done in 3 pregnant Tyrolean Mountain sheep ewes. After premedication (0.06 mg/kg atropine sulfate, 0.01 mg/kg buprenorphine and 2 mg/kg midazolam) anesthesia was initiated with propofol 2% (5 mg/kg) and maintained by volume controlled respiration of 2.5 volume percent isoflurane. Ventilatory parameters were adjusted according to blood gas values. Intervention was performed on day 60 of gestation. After laparotomy the pregnant horn of the uterus was identified and exposed (sheep have a uterus bicornis with 1 or both horns pregnant). After palpating the lower part of the fetus a small hysterotomy was formed using monopolar cautery. Amniotic fluid was preserved. The lower body region of the fetus was carefully exposed and 5 mg ketamine were administered into the gluteal muscles. After left paravertebral incision and identification of the lower renal pole the ureter was exposed and ligated using Prolene® 6.0 in the middle section. The wound was closed using Prolene 6.0 reinforced with polytetrafluoroethylene patches. Body warm amnion fluid was restored after adding 500 mg cefuroxime. The uterus, including all membranes, was closed using Vicryl® zero polyglactin in a running single layer. Postoperatively the animal received 1 mg/kg flunixin meglumine and a fentanyl patch (100 g per hour) was applied on the front shaven leg. At 14 days after intervention the ewes again underwent laparotomy and hysterotomy was performed. The fetus was exposed and anesthetized with 10 mg ketamine
intramuscularly. The fetal kidneys were exposed and bilateral full-thickness renal biopsies were taken. Biopsies were immediately frozen in liquid nitrogen. The fetus and the ewe were sacrificed with an overdose of sodium thiopentone and potassium chloride. Histological analysis was performed on 2 m sections of formalin (4%) neutral buffered fixed, paraffin embedded kidneys after staining with hematoxylin and eosin, and periodic acid-Schiff.
Immunohistochemistry Immunohistochemistry was done using a standard procedure. Briefly, paraffin embedded sections were dewaxed and rehydrated. After thermal antigen retrieval endogenous peroxidase activity was quenched. The primary antibodies were rabbit polyclonal anti-Hmox1 (Enzo® Life Sciences) (diluted 1:100); rabbit polyclonal anti-Klk1 (Abcam®) (diluted 1:1,000), rabbit polyclonal anti-Nos2 (Thermo Fisher Scientific, Epsom, United Kingdom) (diluted 1:200), goat polyclonal anti-Kitlg (diluted 1:1,000), goat polyclonal anti-Ppara (Santa Cruz Biotechnology, Santa Cruz, California) (diluted 1:100) in 3% bovine serum albumin in phosphate buffered saline-Tween. Incubation was done after humid incubation and the addition of biotinylated anti-rabbit IgG or biotinylated anti-goat IgG (Vector Laboratories, Petersborough, United Kingdom) as the secondary antibody. Sections were counterstained with hematoxylin and mounted. Photomicroscopy was done with 10⫻ objectives.
Microarray Hybridization and Scanning For transcriptional level analyses total RNA was extracted using a standard procedure. Briefly, small frozen biopsy cubes approximately 2 mm long on the edge were homogenized in 350 l RLT buffer (Qiagen®) using a Precellys® 24 homogenizer and processed with an RNeasy® Mini Kit according to the manufacturer protocol. Total RNA was eluted in 30 l ribonuclease-free water (Qiagen). RNA yields and quality estimates were measured using an Epoch™ Take3™ NanoDrop® spectrophotometer. Samples were stored at ⫺80C until use. Total RNA was assessed for integrity by agarose gel electrophoresis using the Bioanalyzer 2100 (Agilent Technologies®). Total RNA (5 g) was then used for GeneChip analysis. Preparation of cDNA, hybridization to the Bovine Genome Array and array scanning were done according to the manufacturer protocols (https://www.affymetrix.com).
Data Processing and Statistical Analysis Microarray data preprocessing was done using the Microarray Suite 5.0 (Affymetrix®) method provided by the R affy package. This software, which is an environment for data analysis and exploration of Affymetrix oligonucleotide array probe level data, is part of the Bioconductor project.6 Preprocessing steps included background correction, normalization, perfect match correction and computation of expression values from probe intensities. To identify differentially expressed genes the t test was used and features at p ⱕ0.05 were considered for further analysis.
Bioinformatics Analysis As step 1, the list of Affymetrix Bovine Probe identifiers corresponding to genes that were identified as differen-
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
tially expressed was mapped to Probe identifiers on the Affymetrix Human Genome U133 Plus 2.0 Array using the NetAffx™ tool (http://www.affymetrix.com/analysis/ index). The set of genes was assigned to biological processes using the PANTHER classification system.7 Next to a contextual grouping of a given gene list PANTHER provides information on significant enrichment or depletion of given genes with respect to distinct processes by calculating the chi-square test based on the number of candidates belonging to a process and the total number of genes assigned to this process according to the PANTHER classification. The entire set of genes represented in PANTHER serves as the reference. Biological processes at p ⬍0.05 were considered statistically significant in terms of feature enrichment. Processes exclusively containing genes that were also in a significantly enriched parent process according to PANTHER ontology were omitted from further analysis. Biological processes involved in UUO were also analyzed based on feature overlap using VennMaster (http:// www.informatik.uni-ulm.de/ni/staff/HKestler/vennm/doc. html). Data lists of candidate names and the respective biological process were used as input. The resulting Venn/Euler diagrams, on which each circle represents a biological process, were optimized until all inconsistencies in the diagrams (missing overlaps) were eliminated. Overlapping areas in the diagrams represented candidates present in multiple biological processes. Literature mining was applied systematically to identify genes associated with UUO. This was done based on the about 19 million PubMed® articles for which the 1) MeSH® terms, 2) title and 3) abstract were considered. Key words used in the search were gene name, gene symbol, kidney, renal and obstructive uropathy. To evaluate an eventual role of candidate genes in renal disease additional literature mining was done using information provided by Information Hyperlinked Over Proteins (http:// www.ihop-net.org).
753
RESULTS Complete UUO was performed in 3 female fetuses and 3 healthy twins from the same sheep served as controls. At pregnancy termination mean ⫾ SD weight was 293 ⫾ 152 vs 290 ⫾ 188 gm and mean length was 22.00 ⫾ 3.5 and 22.33 ⫾ 3.21 cm in the UUO group vs controls. All treated kidneys showed time dependent macroscopic and microscopic evidence of obstruction, including a dilated renal pelvis, and ectasia/cysts of tubular structures and especially of glomerular structures (figs. 1 and 2). A total of 509 gene products were identified by statistical analysis as significantly different in renal biopsies from obstructed vs nonobstructed fetal sheep kidneys. Of these genes 298 were identified in PANTHER. Gene set enrichment analysis to identify significantly affected processes provided 7 parental PANTHER processes for a total of 83 members on the list of differentially regulated features. We assessed these differentially regulated genes in enriched biological processes after UUO. The 83 candidates were overrepresented in the processes of protein metabolism and modification, other metabolism, neuronal activities, ligand mediated signaling, amino acid metabolism, coenzyme/prosthetic group metabolism and rRNA metabolism. Literature mining identified 4 candidate genes (Hmox, Klk1, Camk1 and Nos2) previously reported as key in the context of obstructive uropathy in terms of inflammation, apoptosis and fibrosis. Investigating the 83 candidates for literature reported associations with Information Hyperlinked Over Proteins revealed that another 8 genes (Ccl5, Ppara, Kitlg, Mmp16, A2m, Anpep, Ripk2 and Riok3) were linked to renal disease or at least to interstitial inflammation. Their functional role could be easily delineated from other models of renal injury. Five
Quantitative RT-PCR For RT-PCR total RNA from 3 renal biopsies of UUO and 3 healthy control biopsies was extracted using the RNeasy Mini Kit. RNA yields and quality were quantified using the Epoch Take3 NanoDrop spectrophotometer. After the digestion of contaminating genomic DNA first strand cDNA was synthesized from total RNA using the QuantiTect® reverse transcription kit. Real-time PCR was performed on a CFX96 instrument (Bio-Rad®) using the QuantiTect SYBR® green PCR kit. The primer sets used were designed with the high throughput tool PrimerZ (http://www.genepipe.ngc.sinica.edu.tw/primerz) using bovine sequence information on the target genes encoding for Anpep, Camk1, Hmox1, Kitlg, Klk1, Mmp16 and Gapdh with Gapdh as the housekeeping gene.8 Samples were prepared in technical duplicates. Measurements were made with a 3-step protocol using optimal temperatures for annealing and elongation. Relative transcription was calculated from mean threshold cycle values using the ⌬⌬ threshold cycle method.
Figure 1. Hydronephrotic left fetal sheep kidney after 2 weeks of obstruction by ureteral ligation at day 60 of gestation.
754
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
obstructed kidneys predominantly in distal tubules without glomerular staining. The direction of the gene expression of Anpep, Camk1, Hmox1, Kitlg, Klk1 and Mmp16 was validated by qRT-PCR.
DISCUSSION
Figure 2. Hydronephrotic left fetal sheep kidney shows dilated tubules (arrows) and glomerular spaces (asterisk) compared to healthy control kidney.
genes (Erbb3, Fgfr1, Crhbp, Ptk7 and Fmo1) were described for kidney or renal tissue development. Figure 3 shows a Venn diagram demonstrating the involvement of candidates in 1 or multiple biological processes to visualize cellular responses to UUO. The diagram reveals consistent overlap of all enriched cellular processes except amino acid metabolism with the previously reported gene products identified by literature mining. Using immunohistochemistry and qRT-PCR we validated our microarray results. On immunohistochemistry Hmox1, Klk1, Nos2, Ppara and Kitlg showed distinct expression profiles in all obstructed and all control kidneys, eg Hmox1 expression in
To our knowledge we report the first application of transcriptomes in the fetal sheep model of UUO. Day 60 of gestation was chosen for obstruction as the earliest feasible intervention and because early mid trimester ureteral obstruction still induces fetal renal dysplasia.9 We performed renal gene expression profiles after 2 weeks of obstruction since renal damage was noted to be significant, as confirmed in our study by histological assessment, although it may still be at least partially reversible.10 The major strength of the transcriptome approach using microarrays is the global assessment of several thousand genes simultaneously. Inherent to that technology our study detected markedly more genes that were differentially regulated by UUO than described by inductive, hypothesis driven designs in earlier studies. Bioinformatics analysis of microarray data allows us to address one of the inherent limitations of “omics” technology, which is the high rate of false-positive results due to the vast number of simultaneously assessed gene expression assays in a limited number of test animals. The approach implemented in this study was gene set enrichment analysis to identify processes that are significantly populated by differentially regulated features. This procedure allows robust identification of concerted molecular processes involved
Figure 3. Overlap of enriched cellular processes in fetal sheep UUO model with previously reported gene products, as identified by literature mining.
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
in a given phenotype, overcoming the level of noise afflicted with lists of individual features. Of the 509 gene products 83 candidate genes were involved in such enriched processes. We validated the direction of gene expression using qRT-PCR. Also, using immunohistochemistry we validated the translation level from transcript to protein, which showed us the exact location and cell type expressing the proteins. Interstitial inflammation, tubular apoptosis and fibrosis are key mechanisms of obstructive uropathy and our literature mining protocol detected several genes involved in these processes. General transcriptional control of numerous cellular processes, such as cell cycle progression, apoptosis, cell differentiation, inflammation and extracellular matrix remodeling, are time dependent and complex. At day 80 of gestation (20 days after obstruction) the discordant regulation of gene products involved in tubular immune responses suggests proinflammatory, apoptotic and renal protective components. For the sake of brevity we concentrate on Hmox1, Nos2 and Klk1. Hmox1, a well-known anti-apoptotic factor with a key role in oxidative stress, is upregulated in obstructed kidneys.11 Also, Hmox1 upregulation provides protection against renal injury during UUO.12 Immunohistochemistry revealed distinct Hmox1 expression in renal tubules, which may reflect an attempt to prevent tubular damage. Nos2 down-regulation confirms previous reports of decreased nitric oxide production in obstructed kidneys with UUO. It is consistent with renal damage since nitric oxide derived from Nos2 has an essential role in protecting against renal fibrosis in response to UUO.13 Nos2 attenuates apoptosis and fibrosis, and enhances renal parenchymal thickness.14 Correspondingly Nos2 expression patterns showed tubulointerstitial staining. We also found significant Klk1 down-regulation. The kallikrein-kinin system, which is closely related to the renin-angiotensin-system, has an important role in the maintenance of terminal epithelial cell differentiation. It was proposed that endogenous kinins mediate developmental renal growth and differentiation.15 In adult rat renal tissue kallikrein mRNA is markedly suppressed during complete UUO.16 More genes significantly altered in our findings might be considered as likely involved in important pathogenic mechanisms of renal injury. However, their expression and local distribution in diseased kidneys are still largely undetermined. Other molecular players identified as significantly and differentially regulated in obstructed vs nonobstructed fetal kidneys were previously identified in renal tissue but to our knowledge their functions remain unknown while still other significantly altered gene
755
transcripts affected by UUO have not been previously related to the renal system at all. Global gene expression analysis allows an open approach to explorative data analysis. An integrated analysis approach is particularly well suited to investigate the complex interplay of cellular responses, such as those found in UUO. In our study combined transcriptome and bioinformatics analysis in the experimental setting of fetal ovine UUO allowed us to delineate 7 enriched biological processes that are mainly involved in cellular structural, metabolic and signaling activities. Almost all enriched biological processes included 1 or more gene products that could also be assigned by standard literature mining tools to the classic molecular mechanisms of UUO, such as interstitial inflammation, tubular atrophy and renal fibrosis. A process such as ligand mediated signaling is expected to be activated during UUO but the involvement of other processes is less obvious. However, rRNA metabolism and ribosomal synthesis are altered in renal hypertrophy.17 In renal inflammation cytokines affect the (arginine) amino acid metabolism during renal injury and repair.18 Thus, most of the latter processes likely reflect active fetal renal remodeling during UUO, also explaining the major involvement of protein metabolism and modification since this biological process forms the basis of all renal adaptations.19 Apparently trivial at first view, only recently has renal mass at birth been noted as the single best predictive factor of the longterm outcome in infants with congenital developmental nephropathy.20 From this it becomes clear that deep data integration and analysis provide significantly extended insight into the molecular processes interlinking UUO etiologies.
CONCLUSIONS UUO in the mid term causes a set of transcriptional modifications. Our results confirmed the involvement of several previously reported genes. Most previous reports and all data on fetal UUO have focused on 1 or more individual genes but the microarray technique allowed the simultaneous identification of these molecular players in our study. Thus, combined transcriptome and bioinformatics analysis allowed us to identify specific enriched processes in the fetal sheep model of UUO that are likely associated with renal damage but to our knowledge have not been identified previously. Future clarification of these molecular processes in more animals and at more time points may eventually provide molecular targets and/or early predictive markers involved in the pathogenesis of fetal uropathy. Jose Peiro, Barcelona, Spain, introduced us to ovine fetal surgery techniques. Sabine Rauscher,
756
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
Imaging Unit, Center of Translational Research, Medical University of Vienna, assisted with immu-
nohistochemistry. Elisabeth Salzer assisted with cross-species real-time PCR design.
REFERENCES 1. Chevalier RL: Pathogenesis of renal injury in obstructive uropathy. Curr Opin Pediatr 2006; 18: 153. 2. Harrison MR, Nakayama DK, Noall R et al: Correction of congenital hydronephrosis in utero II. Decompression reverses the effects of obstruction on the fetal lung and urinary tract. J Pediatr Surg 1982; 17: 965. 3. Robert C: Microarray analysis of gene expression during early development: a cautionary overview. Reproduction 2010; 140: 787. 4. Seseke F, Thelen P and Ringert RH: Characterization of an animal model of spontaneous congenital unilateral obstructive uropathy by cDNA microarray analysis. Eur Urol 2004; 45: 374. 5. Cordero F, Botta M and Calogero RA: Microarray data analysis and mining approaches. Brief Funct Genomic Proteomic 2007; 6: 265. 6. Gautier L, Cope L, Bolstad BM et al: affy— analysis of Affymetrix GeneChip data at the probe level. Bioinformatics 2004; 20: 307. 7. Thomas PD, Campbell MJ, Kejariwal A et al: PANTHER: a library of protein families and sub-
families indexed by function. Genome Res 2003; 13: 2129. 8. Tsai MF, Lin YJ, Cheng YC et al: PrimerZ: streamlined primer design for promoters, exons and human SNPs. Nucleic Acids Res 2007; 35: W63. 9. Glick PL, Harrison MR, Noall RA et al: Correction of congenital hydronephrosis in utero III. Early mid-trimester ureteral obstruction produces renal dysplasia. J Pediatr Surg 1983; 18: 681. 10. Glick PL, Harrison MR, Adzick NS et al: Correction of congenital hydronephrosis in utero IV: in utero decompression prevents renal dysplasia. J Pediatr Surg 1984; 19: 649. 11. Kawada N, Moriyama T, Ando A et al: Increased oxidative stress in mouse kidneys with unilateral ureteral obstruction. Kidney Int 1999; 56: 1004. 12. Kim JH, Yang JI, Jung MH et al: Heme oxygenase-1 protects rat kidney from ureteral obstruction via an antiapoptotic pathway. J Am Soc Nephrol 2006; 17: 1373. 13. Chang B, Mathew R, Palmer LS et al: Nitric oxide in obstructive uropathy: role of endothelial nitric oxide synthase. J Urol 2002; 168: 1801.
14. Yoo KH, Thornhill BA, Forbes MS et al: Inducible nitric oxide synthase modulates hydronephrosis following partial or complete unilateral ureteral obstruction in the neonatal mouse. Am J Physiol Renal Physiol 2010; 298: F62. 15. El-Dahr SS: Spatial expression of the kallikreinkinin system during nephrogenesis. Histol Histopathol 2004; 19: 1301. 16. el-Dahr SS, Gee J, Dipp S et al: Upregulation of renin-angiotensin system and downregulation of kallikrein in obstructive nephropathy. Am J Physiol 1993; 264: F874. 17. Ouellette AJ, Moonka R, Zelenetz AD et al: Regulation of ribosome synthesis during compensatory renal hypertrophy in mice. Am J Physiol 1987; 253: C506. 18. Ketteler M, Border WA and Noble NA: Cytokines and L-arginine in renal injury and repair. Am J Physiol 1994; 267: F197. 19. Liapis H, Barent B and Steinhardt GF: Extracellular matrix in fetal kidney after experimental obstruction. J Urol 2001; 166: 1433. 20. Winyard P and Chitty LS: Dysplastic kidneys. Semin Fetal Neonatal Med 2008; 13: 142.
Purpose: Fetal obstructive uropathy is a leading cause of loss of renal function. Characterizing the molecular fingerprint of cellular responses to obstruction in a fetal model of complete unilateral ureteral obstruction may help elucidate the activated mechanisms and suggest new therapeutic interventions. Material and Methods: Unilateral ureteral obstruction was created in 3 sheep fetuses at day 60 of gestation. For transcriptome analysis total RNA was extracted from vital renal biopsies 2 weeks after intervention from obstructed kidneys and from control kidneys of untreated twins. cDNA preparation, hybridization to the GeneChip® Bovine Genome Array and array scanning were done according to manufacturer protocols. Bioinformatics analysis was used to derive functional biological processes linked to obstructive uropathy. Quantitative reverse-transcriptase-polymerase chain reaction and immunohistochemistry were used to validate microarray results. Results: Seven biological processes were identified as significantly affected by differentially regulated features that characterize unilateral ureteral obstruction, namely protein metabolism and modification, other metabolism, neuronal activity, ligand mediated signaling, amino acid metabolism, coenzyme/ prosthetic group metabolism and rRNA metabolism. Literature mining identified 17 candidate genes previously reported as key in the context of unilateral ureteral obstruction, related pathological mechanisms or other kidney diseases. Conclusions: Combined transcriptome and bioinformatics analysis allowed the identification of enriched processes in the fetal sheep model of unilateral ureteral obstruction that are likely associated with renal damage but to our knowledge have not been previously identified. Future clarification of these molecular fingerprints may eventually provide therapeutic targets and early predictive markers involved in the pathogenesis of fetal uropathy.
Abbreviations and Acronyms PANTHER ⫽ Protein Analysis Through Evolutionary Relationships PCR ⫽ polymerase chain reaction qRT-PCR ⫽ quantitative RT-PCR RT-PCR ⫽ reverse transcriptasePCR UUO ⫽ unilateral ureteral obstruction Submitted for publication March 21, 2011. Study received approval from the Medical University of Vienna animal ethics committee and Austrian Federal Ministry of Science and Research. Supported by Grant 30000EUR from the Mayor of the City of Vienna (AS). Supplementary material for this article can be obtained at http://www.hypospadie.info/supplement_ ju/supplementary_material.pdf. * Correspondence: Department of Pediatric Nephrology, Medical University of Vienna, Waehringer Guertel 18-20, 1090 Vienna, Austria (telephone: 0043-1-40400-2908; FAX: 0043-1-40400-6812; e-mail: [email protected]).
Key Words: ureter, ureteral obstruction, gene expression, fetus, sheep CONGENITAL UUO is a leading cause of renal impairment in childhood. The cellular and molecular mechanisms of uropathy are complex and classically attributed to the triad of apoptosis,
interstitial inflammation and fibrosis.1 Fetal sheep uropathy models are generally accepted to be the best available models. The methodology is estab-
0022-5347/12/1872-0751/0 THE JOURNAL OF UROLOGY® © 2012 by AMERICAN UROLOGICAL ASSOCIATION EDUCATION
Vol. 187, 751-756, February 2012 Printed in U.S.A. DOI:10.1016/j.juro.2011.09.148
AND
RESEARCH, INC.
www.jurology.com
751
752
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
lished and comparable to humans, thus, forming an excellent basis on which to build on new technologies and opening ways for experimental surgical intervention of fetal uropathy.2 Transcriptome analysis has developed into a powerful tool to contrast gene expression in experimental and clinical settings.3 Recently cDNA microarray analysis was used to examine gene expression profiles in rodent models of postnatal UUO, allowing extensive analysis of the relative expression of a large number of protein coding genes.4 However, extracting clear, coherent hypotheses from genome wide expression data remains an important challenge.5 Bioinformatics analysis is used to transfer these vast amounts of novel insights into cellular mechanisms into biologically relevant findings. To our knowledge we present the first combined application of transcriptome and bioinformatics analysis in the fetal sheep model of UUO to characterize gene profiles. Our specific focus was on the main established pathological mechanisms but we also searched for novel molecular mechanisms with potential biological significance.
MATERIALS AND METHODS Fetal Sheep Model of UUO All experimental procedures were done in accordance with the Medical University of Vienna animal ethics committee and the Austrian Federal Ministry of Science and Research. Studies were done in 3 pregnant Tyrolean Mountain sheep ewes. After premedication (0.06 mg/kg atropine sulfate, 0.01 mg/kg buprenorphine and 2 mg/kg midazolam) anesthesia was initiated with propofol 2% (5 mg/kg) and maintained by volume controlled respiration of 2.5 volume percent isoflurane. Ventilatory parameters were adjusted according to blood gas values. Intervention was performed on day 60 of gestation. After laparotomy the pregnant horn of the uterus was identified and exposed (sheep have a uterus bicornis with 1 or both horns pregnant). After palpating the lower part of the fetus a small hysterotomy was formed using monopolar cautery. Amniotic fluid was preserved. The lower body region of the fetus was carefully exposed and 5 mg ketamine were administered into the gluteal muscles. After left paravertebral incision and identification of the lower renal pole the ureter was exposed and ligated using Prolene® 6.0 in the middle section. The wound was closed using Prolene 6.0 reinforced with polytetrafluoroethylene patches. Body warm amnion fluid was restored after adding 500 mg cefuroxime. The uterus, including all membranes, was closed using Vicryl® zero polyglactin in a running single layer. Postoperatively the animal received 1 mg/kg flunixin meglumine and a fentanyl patch (100 g per hour) was applied on the front shaven leg. At 14 days after intervention the ewes again underwent laparotomy and hysterotomy was performed. The fetus was exposed and anesthetized with 10 mg ketamine
intramuscularly. The fetal kidneys were exposed and bilateral full-thickness renal biopsies were taken. Biopsies were immediately frozen in liquid nitrogen. The fetus and the ewe were sacrificed with an overdose of sodium thiopentone and potassium chloride. Histological analysis was performed on 2 m sections of formalin (4%) neutral buffered fixed, paraffin embedded kidneys after staining with hematoxylin and eosin, and periodic acid-Schiff.
Immunohistochemistry Immunohistochemistry was done using a standard procedure. Briefly, paraffin embedded sections were dewaxed and rehydrated. After thermal antigen retrieval endogenous peroxidase activity was quenched. The primary antibodies were rabbit polyclonal anti-Hmox1 (Enzo® Life Sciences) (diluted 1:100); rabbit polyclonal anti-Klk1 (Abcam®) (diluted 1:1,000), rabbit polyclonal anti-Nos2 (Thermo Fisher Scientific, Epsom, United Kingdom) (diluted 1:200), goat polyclonal anti-Kitlg (diluted 1:1,000), goat polyclonal anti-Ppara (Santa Cruz Biotechnology, Santa Cruz, California) (diluted 1:100) in 3% bovine serum albumin in phosphate buffered saline-Tween. Incubation was done after humid incubation and the addition of biotinylated anti-rabbit IgG or biotinylated anti-goat IgG (Vector Laboratories, Petersborough, United Kingdom) as the secondary antibody. Sections were counterstained with hematoxylin and mounted. Photomicroscopy was done with 10⫻ objectives.
Microarray Hybridization and Scanning For transcriptional level analyses total RNA was extracted using a standard procedure. Briefly, small frozen biopsy cubes approximately 2 mm long on the edge were homogenized in 350 l RLT buffer (Qiagen®) using a Precellys® 24 homogenizer and processed with an RNeasy® Mini Kit according to the manufacturer protocol. Total RNA was eluted in 30 l ribonuclease-free water (Qiagen). RNA yields and quality estimates were measured using an Epoch™ Take3™ NanoDrop® spectrophotometer. Samples were stored at ⫺80C until use. Total RNA was assessed for integrity by agarose gel electrophoresis using the Bioanalyzer 2100 (Agilent Technologies®). Total RNA (5 g) was then used for GeneChip analysis. Preparation of cDNA, hybridization to the Bovine Genome Array and array scanning were done according to the manufacturer protocols (https://www.affymetrix.com).
Data Processing and Statistical Analysis Microarray data preprocessing was done using the Microarray Suite 5.0 (Affymetrix®) method provided by the R affy package. This software, which is an environment for data analysis and exploration of Affymetrix oligonucleotide array probe level data, is part of the Bioconductor project.6 Preprocessing steps included background correction, normalization, perfect match correction and computation of expression values from probe intensities. To identify differentially expressed genes the t test was used and features at p ⱕ0.05 were considered for further analysis.
Bioinformatics Analysis As step 1, the list of Affymetrix Bovine Probe identifiers corresponding to genes that were identified as differen-
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
tially expressed was mapped to Probe identifiers on the Affymetrix Human Genome U133 Plus 2.0 Array using the NetAffx™ tool (http://www.affymetrix.com/analysis/ index). The set of genes was assigned to biological processes using the PANTHER classification system.7 Next to a contextual grouping of a given gene list PANTHER provides information on significant enrichment or depletion of given genes with respect to distinct processes by calculating the chi-square test based on the number of candidates belonging to a process and the total number of genes assigned to this process according to the PANTHER classification. The entire set of genes represented in PANTHER serves as the reference. Biological processes at p ⬍0.05 were considered statistically significant in terms of feature enrichment. Processes exclusively containing genes that were also in a significantly enriched parent process according to PANTHER ontology were omitted from further analysis. Biological processes involved in UUO were also analyzed based on feature overlap using VennMaster (http:// www.informatik.uni-ulm.de/ni/staff/HKestler/vennm/doc. html). Data lists of candidate names and the respective biological process were used as input. The resulting Venn/Euler diagrams, on which each circle represents a biological process, were optimized until all inconsistencies in the diagrams (missing overlaps) were eliminated. Overlapping areas in the diagrams represented candidates present in multiple biological processes. Literature mining was applied systematically to identify genes associated with UUO. This was done based on the about 19 million PubMed® articles for which the 1) MeSH® terms, 2) title and 3) abstract were considered. Key words used in the search were gene name, gene symbol, kidney, renal and obstructive uropathy. To evaluate an eventual role of candidate genes in renal disease additional literature mining was done using information provided by Information Hyperlinked Over Proteins (http:// www.ihop-net.org).
753
RESULTS Complete UUO was performed in 3 female fetuses and 3 healthy twins from the same sheep served as controls. At pregnancy termination mean ⫾ SD weight was 293 ⫾ 152 vs 290 ⫾ 188 gm and mean length was 22.00 ⫾ 3.5 and 22.33 ⫾ 3.21 cm in the UUO group vs controls. All treated kidneys showed time dependent macroscopic and microscopic evidence of obstruction, including a dilated renal pelvis, and ectasia/cysts of tubular structures and especially of glomerular structures (figs. 1 and 2). A total of 509 gene products were identified by statistical analysis as significantly different in renal biopsies from obstructed vs nonobstructed fetal sheep kidneys. Of these genes 298 were identified in PANTHER. Gene set enrichment analysis to identify significantly affected processes provided 7 parental PANTHER processes for a total of 83 members on the list of differentially regulated features. We assessed these differentially regulated genes in enriched biological processes after UUO. The 83 candidates were overrepresented in the processes of protein metabolism and modification, other metabolism, neuronal activities, ligand mediated signaling, amino acid metabolism, coenzyme/prosthetic group metabolism and rRNA metabolism. Literature mining identified 4 candidate genes (Hmox, Klk1, Camk1 and Nos2) previously reported as key in the context of obstructive uropathy in terms of inflammation, apoptosis and fibrosis. Investigating the 83 candidates for literature reported associations with Information Hyperlinked Over Proteins revealed that another 8 genes (Ccl5, Ppara, Kitlg, Mmp16, A2m, Anpep, Ripk2 and Riok3) were linked to renal disease or at least to interstitial inflammation. Their functional role could be easily delineated from other models of renal injury. Five
Quantitative RT-PCR For RT-PCR total RNA from 3 renal biopsies of UUO and 3 healthy control biopsies was extracted using the RNeasy Mini Kit. RNA yields and quality were quantified using the Epoch Take3 NanoDrop spectrophotometer. After the digestion of contaminating genomic DNA first strand cDNA was synthesized from total RNA using the QuantiTect® reverse transcription kit. Real-time PCR was performed on a CFX96 instrument (Bio-Rad®) using the QuantiTect SYBR® green PCR kit. The primer sets used were designed with the high throughput tool PrimerZ (http://www.genepipe.ngc.sinica.edu.tw/primerz) using bovine sequence information on the target genes encoding for Anpep, Camk1, Hmox1, Kitlg, Klk1, Mmp16 and Gapdh with Gapdh as the housekeeping gene.8 Samples were prepared in technical duplicates. Measurements were made with a 3-step protocol using optimal temperatures for annealing and elongation. Relative transcription was calculated from mean threshold cycle values using the ⌬⌬ threshold cycle method.
Figure 1. Hydronephrotic left fetal sheep kidney after 2 weeks of obstruction by ureteral ligation at day 60 of gestation.
754
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
obstructed kidneys predominantly in distal tubules without glomerular staining. The direction of the gene expression of Anpep, Camk1, Hmox1, Kitlg, Klk1 and Mmp16 was validated by qRT-PCR.
DISCUSSION
Figure 2. Hydronephrotic left fetal sheep kidney shows dilated tubules (arrows) and glomerular spaces (asterisk) compared to healthy control kidney.
genes (Erbb3, Fgfr1, Crhbp, Ptk7 and Fmo1) were described for kidney or renal tissue development. Figure 3 shows a Venn diagram demonstrating the involvement of candidates in 1 or multiple biological processes to visualize cellular responses to UUO. The diagram reveals consistent overlap of all enriched cellular processes except amino acid metabolism with the previously reported gene products identified by literature mining. Using immunohistochemistry and qRT-PCR we validated our microarray results. On immunohistochemistry Hmox1, Klk1, Nos2, Ppara and Kitlg showed distinct expression profiles in all obstructed and all control kidneys, eg Hmox1 expression in
To our knowledge we report the first application of transcriptomes in the fetal sheep model of UUO. Day 60 of gestation was chosen for obstruction as the earliest feasible intervention and because early mid trimester ureteral obstruction still induces fetal renal dysplasia.9 We performed renal gene expression profiles after 2 weeks of obstruction since renal damage was noted to be significant, as confirmed in our study by histological assessment, although it may still be at least partially reversible.10 The major strength of the transcriptome approach using microarrays is the global assessment of several thousand genes simultaneously. Inherent to that technology our study detected markedly more genes that were differentially regulated by UUO than described by inductive, hypothesis driven designs in earlier studies. Bioinformatics analysis of microarray data allows us to address one of the inherent limitations of “omics” technology, which is the high rate of false-positive results due to the vast number of simultaneously assessed gene expression assays in a limited number of test animals. The approach implemented in this study was gene set enrichment analysis to identify processes that are significantly populated by differentially regulated features. This procedure allows robust identification of concerted molecular processes involved
Figure 3. Overlap of enriched cellular processes in fetal sheep UUO model with previously reported gene products, as identified by literature mining.
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
in a given phenotype, overcoming the level of noise afflicted with lists of individual features. Of the 509 gene products 83 candidate genes were involved in such enriched processes. We validated the direction of gene expression using qRT-PCR. Also, using immunohistochemistry we validated the translation level from transcript to protein, which showed us the exact location and cell type expressing the proteins. Interstitial inflammation, tubular apoptosis and fibrosis are key mechanisms of obstructive uropathy and our literature mining protocol detected several genes involved in these processes. General transcriptional control of numerous cellular processes, such as cell cycle progression, apoptosis, cell differentiation, inflammation and extracellular matrix remodeling, are time dependent and complex. At day 80 of gestation (20 days after obstruction) the discordant regulation of gene products involved in tubular immune responses suggests proinflammatory, apoptotic and renal protective components. For the sake of brevity we concentrate on Hmox1, Nos2 and Klk1. Hmox1, a well-known anti-apoptotic factor with a key role in oxidative stress, is upregulated in obstructed kidneys.11 Also, Hmox1 upregulation provides protection against renal injury during UUO.12 Immunohistochemistry revealed distinct Hmox1 expression in renal tubules, which may reflect an attempt to prevent tubular damage. Nos2 down-regulation confirms previous reports of decreased nitric oxide production in obstructed kidneys with UUO. It is consistent with renal damage since nitric oxide derived from Nos2 has an essential role in protecting against renal fibrosis in response to UUO.13 Nos2 attenuates apoptosis and fibrosis, and enhances renal parenchymal thickness.14 Correspondingly Nos2 expression patterns showed tubulointerstitial staining. We also found significant Klk1 down-regulation. The kallikrein-kinin system, which is closely related to the renin-angiotensin-system, has an important role in the maintenance of terminal epithelial cell differentiation. It was proposed that endogenous kinins mediate developmental renal growth and differentiation.15 In adult rat renal tissue kallikrein mRNA is markedly suppressed during complete UUO.16 More genes significantly altered in our findings might be considered as likely involved in important pathogenic mechanisms of renal injury. However, their expression and local distribution in diseased kidneys are still largely undetermined. Other molecular players identified as significantly and differentially regulated in obstructed vs nonobstructed fetal kidneys were previously identified in renal tissue but to our knowledge their functions remain unknown while still other significantly altered gene
755
transcripts affected by UUO have not been previously related to the renal system at all. Global gene expression analysis allows an open approach to explorative data analysis. An integrated analysis approach is particularly well suited to investigate the complex interplay of cellular responses, such as those found in UUO. In our study combined transcriptome and bioinformatics analysis in the experimental setting of fetal ovine UUO allowed us to delineate 7 enriched biological processes that are mainly involved in cellular structural, metabolic and signaling activities. Almost all enriched biological processes included 1 or more gene products that could also be assigned by standard literature mining tools to the classic molecular mechanisms of UUO, such as interstitial inflammation, tubular atrophy and renal fibrosis. A process such as ligand mediated signaling is expected to be activated during UUO but the involvement of other processes is less obvious. However, rRNA metabolism and ribosomal synthesis are altered in renal hypertrophy.17 In renal inflammation cytokines affect the (arginine) amino acid metabolism during renal injury and repair.18 Thus, most of the latter processes likely reflect active fetal renal remodeling during UUO, also explaining the major involvement of protein metabolism and modification since this biological process forms the basis of all renal adaptations.19 Apparently trivial at first view, only recently has renal mass at birth been noted as the single best predictive factor of the longterm outcome in infants with congenital developmental nephropathy.20 From this it becomes clear that deep data integration and analysis provide significantly extended insight into the molecular processes interlinking UUO etiologies.
CONCLUSIONS UUO in the mid term causes a set of transcriptional modifications. Our results confirmed the involvement of several previously reported genes. Most previous reports and all data on fetal UUO have focused on 1 or more individual genes but the microarray technique allowed the simultaneous identification of these molecular players in our study. Thus, combined transcriptome and bioinformatics analysis allowed us to identify specific enriched processes in the fetal sheep model of UUO that are likely associated with renal damage but to our knowledge have not been identified previously. Future clarification of these molecular processes in more animals and at more time points may eventually provide molecular targets and/or early predictive markers involved in the pathogenesis of fetal uropathy. Jose Peiro, Barcelona, Spain, introduced us to ovine fetal surgery techniques. Sabine Rauscher,
756
TRANSCRIPTOME AND BIOINFORMATICS APPROACH TO FETAL URETERAL OBSTRUCTIVE UROPATHY
Imaging Unit, Center of Translational Research, Medical University of Vienna, assisted with immu-
nohistochemistry. Elisabeth Salzer assisted with cross-species real-time PCR design.
REFERENCES 1. Chevalier RL: Pathogenesis of renal injury in obstructive uropathy. Curr Opin Pediatr 2006; 18: 153. 2. Harrison MR, Nakayama DK, Noall R et al: Correction of congenital hydronephrosis in utero II. Decompression reverses the effects of obstruction on the fetal lung and urinary tract. J Pediatr Surg 1982; 17: 965. 3. Robert C: Microarray analysis of gene expression during early development: a cautionary overview. Reproduction 2010; 140: 787. 4. Seseke F, Thelen P and Ringert RH: Characterization of an animal model of spontaneous congenital unilateral obstructive uropathy by cDNA microarray analysis. Eur Urol 2004; 45: 374. 5. Cordero F, Botta M and Calogero RA: Microarray data analysis and mining approaches. Brief Funct Genomic Proteomic 2007; 6: 265. 6. Gautier L, Cope L, Bolstad BM et al: affy— analysis of Affymetrix GeneChip data at the probe level. Bioinformatics 2004; 20: 307. 7. Thomas PD, Campbell MJ, Kejariwal A et al: PANTHER: a library of protein families and sub-
families indexed by function. Genome Res 2003; 13: 2129. 8. Tsai MF, Lin YJ, Cheng YC et al: PrimerZ: streamlined primer design for promoters, exons and human SNPs. Nucleic Acids Res 2007; 35: W63. 9. Glick PL, Harrison MR, Noall RA et al: Correction of congenital hydronephrosis in utero III. Early mid-trimester ureteral obstruction produces renal dysplasia. J Pediatr Surg 1983; 18: 681. 10. Glick PL, Harrison MR, Adzick NS et al: Correction of congenital hydronephrosis in utero IV: in utero decompression prevents renal dysplasia. J Pediatr Surg 1984; 19: 649. 11. Kawada N, Moriyama T, Ando A et al: Increased oxidative stress in mouse kidneys with unilateral ureteral obstruction. Kidney Int 1999; 56: 1004. 12. Kim JH, Yang JI, Jung MH et al: Heme oxygenase-1 protects rat kidney from ureteral obstruction via an antiapoptotic pathway. J Am Soc Nephrol 2006; 17: 1373. 13. Chang B, Mathew R, Palmer LS et al: Nitric oxide in obstructive uropathy: role of endothelial nitric oxide synthase. J Urol 2002; 168: 1801.
14. Yoo KH, Thornhill BA, Forbes MS et al: Inducible nitric oxide synthase modulates hydronephrosis following partial or complete unilateral ureteral obstruction in the neonatal mouse. Am J Physiol Renal Physiol 2010; 298: F62. 15. El-Dahr SS: Spatial expression of the kallikreinkinin system during nephrogenesis. Histol Histopathol 2004; 19: 1301. 16. el-Dahr SS, Gee J, Dipp S et al: Upregulation of renin-angiotensin system and downregulation of kallikrein in obstructive nephropathy. Am J Physiol 1993; 264: F874. 17. Ouellette AJ, Moonka R, Zelenetz AD et al: Regulation of ribosome synthesis during compensatory renal hypertrophy in mice. Am J Physiol 1987; 253: C506. 18. Ketteler M, Border WA and Noble NA: Cytokines and L-arginine in renal injury and repair. Am J Physiol 1994; 267: F197. 19. Liapis H, Barent B and Steinhardt GF: Extracellular matrix in fetal kidney after experimental obstruction. J Urol 2001; 166: 1433. 20. Winyard P and Chitty LS: Dysplastic kidneys. Semin Fetal Neonatal Med 2008; 13: 142.