Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Science. Author manuscript; available in PMC 2009 October 4.
Published in final edited form as:
PMCID: PMC2756524

Analysis of Drosophila Segmentation Network Identifies a JNK Pathway Factor Overexpressed in Kidney Cancer


We constructed a large-scale functional network model in Drosophila melanogaster built around two key transcription factors involved in the process of embryonic segmentation. Analysis of the model allowed the identification of a new role for the ubiquitin E3 ligase complex factor SPOP. In Drosophila, the gene encoding SPOP is a target of segmentation transcription factors. Drosophila SPOP mediates degradation of the Jun-kinase phosphatase Puckered thereby inducing TNF/Eiger dependent apoptosis. In humans we found that SPOP plays a conserved role in TNF-mediated JNK signaling and was highly expressed in 99% of clear cell renal cell carcinoma (RCC), the most prevalent form of kidney cancer. SPOP expression distinguished histological subtypes of RCC and facilitated identification of clear cell RCC as the primary tumor for metastatic lesions.

Over the last three decades, extensive molecular and genetic analyses have characterized the identity of and interactions between components of the Drosophila segmentation process (1). Maternal factors distributed in gradients along the anterior-posterior (A-P) axis activate zygotic transcription of gap genes, which encode transcription factors that activate sets of pair-rule genes including the homeobox transcription factors Even-skipped (Eve) and Fushi tarazu (Ftz). These pair rule proteins then directly regulate segment polarity genes that determine the internal A-P orientation of each segment. Many of the human homologs of these genes and their down-stream targets play critical roles in human diseases, especially cancers (2, 3). In an effort to extract new information from the Drosophila segmentation network, as well as to mine this network for novel disease related genes, we built a large-scale predictive network model around Ftz and Eve.

We analyzed gene expression changes between individual wild type embryos and embryos with null mutations in ftz and eve (4, 5) collected during a developmental time course from 2 hr until 7 hr after egg laying (AEL). By focusing on Ftz and Eve effects 2 to 3 hours AEL (early zygotic expression), we found 1310 genes differentially expressed between the ftz mutant and wild type, and 1074 genes differentially expressed between the eve mutant and wild type (false discovery rate < 0.001, table S1, S2).

Using antibodies specific for Ftz or Eve, we performed chromatin immunoprecipitation (ChIP) and mapped genome-wide transcription factor binding in cellular blastoderm embryos 2 hours AEL on custom designed high density DNA microarrays (4). We found 1286 Ftz and 1499 Eve bound probes (intensity P < 0.0001 and Z score > 1.96, see supplemental method; 21 probes on both lists map within 500bp). We analyzed several methods for probe to target gene mapping to maximize the overlap between the differentially expressed and ChIP target gene sets (see supplemental methods; fig. S1). The greatest such enrichment was obtained by designating genes within 1kb of a binding site as targets. At this threshold, we identified 969 Ftz ChIP-chip target genes and 932 Eve ChIP-chip target genes (overlap 175 genes; table S3, S4).

Genes both differentially expressed and targeted by ChIP-chip binding site mapping were considered as putative direct target genes. We thus identified 137 Ftz direct target genes (Fig. 1A) and 98 Eve direct target genes (fig. S2; overlap 9 genes). Fig. 1A (right panel) shows the locations of binding sites at Ftz or Eve direct target genes. Analysis of direct target gene annotations indicates 39 genes (21%) regulate transcription and 74 genes (40%) are involved in developmental processes (Fig. 1A, center panel); both annotation classes were significantly enriched compared to the 9.6% and 18% of Drosophila genes annotated as transcriptional or developmental regulators, respectively (p=1.05×10−6 and p=1.81×10−12; hypergeometric test). A complete target list can be found in tables S5 and S6.

Fig. 1
Drosophila segmentation network. (A) Identification of direct targets of Ftz. Heatmap at left depicts log2-fold change in gene expression, mutant vs. wild type. Columns represent time points in hours after egg laying (AEL). Rows depict individual genes, ...

To extend our Ftz-Eve network model beyond direct transcriptional regulation, we included automated literature mining methods to capture published interactions of target genes (5). We then integrated yeast two-hybrid based protein-protein interaction data (6) into our model by connecting protein interactions between existing components in the network. To limit the size of the network, we extended the protein-protein interaction only one degree from the direct targets of Ftz or Eve. The resulting Ftz-Eve regulatory network model included 4084 genes/proteins and 6648 interactions between them (fig. S3).

To confirm parts of the network model topology we examined several genes that are expressed in segmental patterns (7) and validated a limited set of interactions by genetic and biochemical testing of simple predictions from our network model (fig. S4),

Analysis of the Eve-Ftz network identified 150 different genes as direct targets of Eve or Ftz that also have unambiguous human homologs. From this gene set, we identified a top candidate, CG9924 or roadkill (rdx), which ranks first in network betweenness-centrality and thus constitutes a major network hub (8) (see supplemental methods, table S8). The rdx gene encodes a BTB domain protein that has been recently shown to act to regulate Cubitus interruptus (Ci) degradation in the Hedgehog pathway (9, 10). This product of the rdx gene is 79% identical to the human protein SPOP and these proteins appear to be orthologs (fig. S5) (9, 10); we refer to the rdx gene product(s) as Drosophila SPOP (D-SPOP).

Our network model indicates that the D-SPOP gene is a direct target of Ftz at 2–3 hours AEL and that the D-SPOP protein interacts with the Jun Kinase phosphatase Puckered (Puc) (Fig. 1B). RNA in situ hybridizations for D-SPOP mRNA in ftz mutant embryos confirmed that ftz is indeed required for D-SPOP expression in parasegments that normally express Ftz (Fig. 1C). We did not observe significant mis-expression of D-SPOP mRNA in eve mutant embryos at 2–3 hours AEL (fig. S6), suggesting that the Ftz effects on D-SPOP mRNA levels occurs in advance of the Eve effect. We found that the D-SPOP protein segmental expression pattern was completely lost in eve mutant embryos 6–7 hours AEL (Fig. 1D), behaving similarly to the well-characterized Ftz and Eve target gene engrailed (11, 12). Previous studies also indicate that D-SPOP is regulated by Hedgehog (Hh) later in development, indicating another layer of D-SPOP regulation by the segment polarity system (9). Together, these data strongly indicate that D-SPOP expression is downstream of the pair-rule genes in the segmentation hierarchy.

RNAi knockdown of D-SPOP mRNA levels and P-element insertion mutagenesis of the D-SPOP gene resulted in severe and consistent disruption of both the peripheral and the central nervous system (CNS) (fig. S7). Such phenotypes are recapitulated by mutating ftz or eve and are likely due to mid-embryonic functions of D-SPOP when Ftz and Eve become active in the CNS (13, 14). Furthermore, it was recently demonstrated that the Drosophila Eiger/TNF pathway regulates embryonic neuroblast division (15). Thus we hypothesized that the function of D-SPOP in nervous tissue development may result from its interaction with Puc, which mediates a feedback loop by negatively regulating basket (Drosophila JNK) in the Drosophila Eiger/Tumor Necrosis Factor (TNF) pathway (16) (Fig. 1B).

In Drosophila, ectopic expression of Eiger in neuronal cells in the developing eye induces apoptosis through the JNK pathway, resulting in a reduced adult eye size (Fig. 2A, 2B) (17). Deletion of one wild-type copy of D-SPOP or RNAi knock down of D-SPOP mRNA partially suppresses the eye phenotype of Eiger expression (Fig. 2C, 2D). Additionally, ectopic expression of D-SPOP in the developing eyes produces a small and rough eye phenotype (Fig. 2E). Analysis of genetic interactions between the genes encoding D-SPOP and other members of the Eiger-JNK pathway (fig. S8), indicates that D-SPOP is acting downstream of dTAK1 (JNKKK) and Hep (JNKK) and upstream of Bsk (JNK) and Puc. Our experiments therefore indicate that D-SPOP functions as an essential positive regulator for Eiger triggered apoptosis, consistent with the interaction between D-SPOP and Puc predicted in the Ftz-Eve network model.

Fig. 2
D-SPOP promotes puc ubiquitination and degradation. (A) Light micrographs of Drosophila adult eyes for wild type (GMR-Gal4/+). (B) GMR>Egr triggered cell death and produced a small eye phenotype (GMR-Gal4 UAS-Egr/+). (C) Deleting one copy of D-SPOP ...

A physical interaction between D-SPOP and Puc was confirmed by both in vitro pulldown and in vivo immunoprecipitation assays (Fig. 2F, 2G). D-SPOP contains two conserved domains, a MATH domain and a BTB/POZ domain (18). MEL-26, the Caenorhabditis elegans ortholog of human SPOP, was first identified as a BTB protein that serves as an adaptor of Cul3 based ubiquitin ligase (18). Recently, human SPOP has been shown to mediate ubiquitination of death domain-associated protein (Daxx) (19), the Polycomb group protein BMI-1, the histone variant MacroH2A (20), and the transcription factor Gli (10). We found that Puc protein levels were significantly reduced when co-expressed with D-SPOP in S2 cells (Fig. 2H). Furthermore, D-SPOP promoted Puc ubiquitination in S2 cells treated with the proteasome inhibitor MG132 (Fig. 2I). Taken together, these results indicate that D-SPOP induces apoptosis in the Eiger/TNF pathway by the mediating Puc ubiquitination and degradation (Fig. 2J).

Homologs of several Ftz and Eve targets have been shown to be involved in human cancers (21), a large body of experimental and clinical data indicates that defects in ubiquitin signaling pathways have roles in the genesis of different tumor types (22), and JNK activation is required for cellular transformation induced by RAS, an oncogene mutated in 30% of human cancers (23). To determine whether human SPOP’s role in modulating TNF stimulated JNK signaling is conserved, we treated HEK293 cells over-expressing SPOP with TNF-α, then analyzed phosphorylated JNK (P-JNK) and phosphorylated c-Jun (P-c-Jun) levels. Consistent with its role in Drosophila as an activator of the pathway, overexpression of SPOP increases the level of P-JNK and P-c-Jun, indicating conservation of its function in modulating the JNK pathway (Fig. 3A).

Fig. 3
Function of SPOP in the mammalian TNF pathway and over-expression in renal cell carcinoma. (A) Over-expression of SPOP increases the level of P-JNK and P-c-Jun. Human embryonic kidney (HEK293) cells were transfected with SPOP, treated with 50ng/ml TNF ...

To test whether SPOP is associated with human cancers, SPOP protein expression levels were screened with tissue microarrays that contained 20 tumors from each of 18 different organs. We found that 85% of renal cell carcinomas (RCC) showed high expression of SPOP, while normal kidney tissue was uniformly negative (Fig 3B; Table 1). To further investigate the potential of SPOP as a marker, we designed a large tissue array containing more than 300 RCC samples. 77% of the tumor samples were positive for SPOP staining; normal kidney samples were all negative (Table 2).

Table 1
Tissue micro-array screening for SPOP expression in 18 cancer types from different organs.
Table 2
SPOP expression in RCCs. RCC tissue sections were analyzed by staining with SPOP specific monoclonal antibody (SPOP-5G). Patient samples are classified into different categories depending on cell type.

RCC is a heterogeneous group of tumors with distinct histological subtypes, including clear cell, papillary, chromophobe and other rare subtypes in addition to oncocytoma, which is a benign solid renal tumor (24). The majority of RCC is of clear cell type, comprising up to 75%. While the majority of RCCs can be subtyped by hematoxylin + eosin staining morphology, diagnostic difficulties arise when clear cell RCCs display morphologic features that overlap with other RCC subtypes and non-renal tumors (2527). Currently, a panel of immunohistochemical markers is used to differentiate the major subtypes of RCC in difficult cases (26, 27). Unfortunately, these panels lack a specific and sensitive marker that is positive in clear cell RCC (26, 27). Recently Carbonic Anhydrase IX has been proposed as a positive marker for clear cell RCC, but it is positive in other RCCs and several other tumor types as well (2830). Patient tumor samples in our studies were classified into different types according to the recent World Health Organization (WHO) classification system. We found that 99% of the clear cell RCC and 86% of the chromophobe RCC showed positive staining for SPOP, but only 22% of papillary-type RCC were SPOP positive. Four out of 31 papillary RCCs from the general pathology reports were shown to be misdiagnosed as clear cell RCCs when the tumor biopsies were re-analyzed by urological pathologists. All four of these misdiagnosed RCCs have papillary architecture and were subsequently shown to stain positive for SPOP. Our tissue array also included benign oncocytomas, which can mimic renal cell carcinoma both clinically and pathologically, in turn potentially subjecting patients to unnecessary surgeries and additional morbidities. Only 6% of oncocytomas showed weak positive staining. These results indicate that SPOP is a highly sensitive and specific diagnostic biomarker for clear cell RCC and can help distinguish histological subtypes of RCC.

Up to 30% of RCC patients present with metastases; half of the rest will develop metastases later in their course, 90% of which are clear cell RCCs. Accordingly, we further screened for SPOP staining in confirmed metastases from RCC and found that 97% of them were positive (Table 3), indicating that SPOP may be a useful biomarker to identify clear cell RCC as the site of the primary tumors in cases of metastases from unknown origin. Taken together, our results demonstrate that novel functions for conserved molecules can readily be extracted from data mining of large scale networks in Drosophila, and provide a strategy for rapid identification of factors that may have clinical relevance as biomarkers or drug targets for human diseases.

Table 3
SPOP expression in metastatic lesions where RCC were the primary tumors. Metastatic tissues were analyzed by staining with SPOP specific monoclonal antibody (SPOP-5G). Patient samples are classified into different categories depending on cell type of ...

Supplementary Material


References and Notes

1. Nasiadka A, Dietrich BH, Krause HM. Anterior-Posterior Patterning in the Drosophila Embryo. In: DePamphilis M, editor. Advances in Developmental Biology and Biochemistry. Vol. 12. Elsevier Science B.V; 2002. p. 155.
2. Hahn SA, et al. Science. 1996 Jan 19;271:350. [PubMed]
3. Johnson RL, et al. Science. 1996 Jun 14;272:1668. [PubMed]
4. Stolc V, et al. Science. 2004 Oct 22;306:655. [PubMed]
5. Rzhetsky A, et al. J Biomed Inform. 2004 Feb;37:43. [PubMed]
6. Giot L, et al. Science. 2003 Dec 5;302:1727. [PubMed]
7. Tomancak P, et al. Genome Biol. 2007;8:R145. [PMC free article] [PubMed]
8. Jeong H, Mason SP, Barabasi AL, Oltvai ZN. Nature. 2001 May 3;411:41. [PubMed]
9. Kent D, Bush EW, Hooper JE. Development. 2006 May;133:2001. [PubMed]
10. Zhang Q, et al. Dev Cell. 2006 Jun;10:719. [PubMed]
11. Harding K, Rushlow C, Doyle HJ, Hoey T, Levine M. Science. 1986 Aug 29;233:953. [PubMed]
12. Jaynes JB, Fujioka M. Dev Biol. 2004 May 15;269:609. [PMC free article] [PubMed]
13. Doe CQ, Hiromi Y, Gehring WJ, Goodman CS. Science. 1988 Jan 8;239:170. [PubMed]
14. Broadus J, et al. Mech Dev. 1995 Nov;53:393. [PubMed]
15. Wang H, Cai Y, Chia W, Yang X. Embo J. 2006 Dec 13;25:5783. [PubMed]
16. Martin-Blanco E, et al. Genes Dev. 1998 Feb 15;12:557. [PubMed]
17. Igaki T, et al. Embo J. 2002 Jun 17;21:3009. [PubMed]
18. Xu L, et al. Nature. 2003 Sep 18;425:316. [PubMed]
19. Kwon JE, et al. J Biol Chem. 2006 May 5;281:12664. [PubMed]
20. Hernandez-Munoz I, et al. Proc Natl Acad Sci U S A. 2005 May 24;102:7635. [PubMed]
21. Potter CJ, Turenchalk GS, Xu T. Trends Genet. 2000 Jan;16:33. [PubMed]
22. Hoeller D, Hecker CM, Dikic I. Nat Rev Cancer. 2006 Oct;6:776. [PubMed]
23. Weiss C, Bohmann D. Cell Cycle. 2004 Feb;3:111. [PubMed]
24. Cohen HT, McGovern FJ. N Engl J Med. 2005 Dec 8;353:2477. [PubMed]
25. Rosai J. Rosai and Ackerman’s Surgical Pathology. Mosby; New York: 2004.
26. Skinnider BF, Amin MB. Semin Diagn Pathol. 2005 Feb;22:51. [PubMed]
27. Zhou M, Roma A, Magi-Galluzzi C. Clin Lab Med. 2005 Jun;25:247. [PubMed]
28. Al-Ahmadie HA, et al. Am J Surg Pathol. 2008 Mar;32:377. [PubMed]
29. Leibovich BC, et al. J Clin Oncol. 2007 Oct 20;25:4757. [PubMed]
30. Potter C, Harris AL. Cell Cycle. 2004 Feb;3:164. [PubMed]
31. We thank J. Jiang, M. Van Lohuizen, C. Chung, D. McEwen for providing expression vectors. Microarray data described in this paper is has been deposited in the NCBI Gene Expression Omnibus (GEO) under accession code GSE14086 (expression data) and GSE14289 (ChIP data). M.G. was supported by Vaadia-BARD Postdoctoral Fellowship Award No. FI-315-2001 from BARD, The United States - Israel Binational Agricultural Research and Development Fund. C.D.B was supported by a Lilly Life Science Research Fellowship. This work was supported by grants from the W. M. Keck Foundation, the Arnold and Mabel Beckman Foundation, and the Searle Funds at The Chicago Community Trust from the Chicago Biomedical Consortium to K.P.W.