Regulation of Drosophila Hypoxia-inducible Factor (HIF) Activity in SL2 Cells

Although hypoxia-inducible factor-α (HIFα) subunit-specific hydroxylation and proteolytic breakdown explain the binary switch between the presence (hypoxia) and absence (normoxia) of HIFs, little is known of the mechanisms that fine-tune HIF activity under constant, rather than changing, oxygen tensions. Here, we report that the Drosophila HIFα homolog, the basic helix-loop-helix/PAS protein Sima (Similar), in hypoxic cultures of SL2 cells is expressed in full-length (fl) and splice variant (sv) isoforms. The following evidence supports the role of flSima as functional HIFα and the role of SL2 HIF as a transcriptional activator or suppressor. The pO2 dependence of Sima abundance matched that of HIF activity. HIF-dependent changes in candidate target gene expression were detected through variously effective stimuli: hypoxia (strong) > iron chelation, e.g. desferrioxamine (moderate) >> transition metals, e.g. cobalt ≃ normoxia (ineffective). Sima overexpression augmented hypoxic induction or suppression of different targets. In addition to the full-length exon 1–12 transcript yielding the 1510-amino acid HIFα homolog, the sima gene also expressed, specifically under hypoxia, an exon 1–7/12 splice variant, which translated into a 426-amino acid Sima truncation termed svSima. svSima contains basic helix-loop-helix and PAS sequences identical to those of flSima, but, because of deletion of exons 8–11, lacks the oxygen-dependent degradation domain and nuclear localization signals. Overexpressed svSima failed to transactivate reporter genes. However, it attenuated HIF (Sima·Tango)-stimulated reporter expression in a dose-dependent manner. Thus, svSima has the potential to regulate Drosophila HIF function under steady and hypoxic pO2 by creating a cytosolic sink for the Sima partner protein Tango.

Oxygen homeostasis depends upon sensing variations in oxygen tension (pO 2 ) and signal transduction, leading to physio-logically appropriate changes in gene expression. Virtually all aerobic prokaryotes and eukaryotes have evolved various mechanisms for regulating genes during times of oxygen deprivation (1). Particularly in hypoxia-tolerant species, i.e. those capable of surviving and even thriving in environments with little (hypoxia) or no (anoxia) oxygen present, these changes in gene expression, together with drastically reduced steady-state ATP levels, are critical determinants of the resilience of the organism to low oxygen (2). In contrast to most endothermic vertebrates (birds and mammals), invertebrates are often hypoxia-tolerant (3)(4)(5). For example, embryos and adult stages of the fruit fly Drosophila melanogaster possess a rich and varied repertoire of survival strategies endowing them to withstand and fully recover from even hour-long exposure to anoxia (N 2 atmosphere) (6). These responses vary once oxygen levels have declined to 1.6 -3%, the critical pO 2 of the species (7,8), below which insect respiration generally ceases to be regulated (9,10). The following strategies are employed by fruit fly embryos and adults when faced with hypoxic/anoxic challenges: (a) ability to sense falling pO 2 within seconds via a nitric oxide signaling pathway (11), (b) anoxic stupor ranging from loss of coordination to complete immobility throughout the entire period of O 2 deprivation (7,12), (c) greatly reduced O 2 consumption down to 20% of normoxic values (6,12), (d) glycogen-fueled anaerobiosis with lactate as the major end product (5,13), (e) general and reversible chromatin condensation (14), and (f) cell cycle arrest (11, 14 -16). Another important oxygen-sensitive response in insects is the long known hypoxic stimulation of outgrowth and proliferation of tracheal termini for improved supply of limiting amounts of O 2 to tissues (17)(18)(19)(20). This hypoxia-induced ramification of insect breathing tubes poses intriguing parallels to the O 2 response in mammalian angiogenesis (21)(22)(23).
A wide range of animals, from mammals to fruit flies to nematodes, share a common pathway that links sensing of changes in pO 2 to transcriptional regulation. Central to hypoxia-mediated gene expression are hypoxia-inducible (transcription) factors (HIFs) 1 (24 -26), which belong to the family of basic helix-loop-helix (bHLH)/PAS (Per/ARNT/Sim) transcription factors (27)(28)(29). In mammals (30,31), Caenorhabditis elegans (32,33), and Drosophila (34 -37), HIF is a heterodimer of ␣and ␤-subunits that specifically recognizes short cis-regulatory E-box motifs called hypoxia response elements (HREs) in the promoter and/or enhancer regions of a number of genes (38,39). Functional HREs, i.e. those capable of oxygen-dependent binding of HIF proteins in vitro and transactivating reporter genes in vivo, can be summarized in the following consensus sequence: 5Ј-B(A/G)CGT-GVBBB-3Ј (where B is all bases except A and V is all bases except T) (40). The central 4-base core (underlined) is critical for the binding of any HIF complex, be it mammalian, insect, or crustacean (see accompanying article (87)).
Oxygen-dependent activation of HIF is, in cell culture, controlled primarily at the protein level through specific oxidative modifications of the ␣-subunit (41). The HIF oxygen sensor is a novel prolyl hydroxylase (42)(43)(44) that catalyzes the O 2 -dependent hydroxylation of proline residues within the oxygen-dependent degradation domain (ODD) of HIF-1␣ (45). Once hydroxylated, HIF-1␣ binds rapidly to the von Hippel-Lindau tumor suppressor product, enabling ubiquitination and rapid degradation by the proteasome (46 -49). Conversely, hypoxia and hypoxia-mimicking agents such as transition metals (e.g. Co 2ϩ ) and iron chelators (e.g. desferrioxamine) (42) inhibit proline hydroxylation of HIF-1␣, thus enabling the protein to escape proteolytic degradation and to dimerize with HIF-1␤ (ARNT). The HIF dimer then translocates into the nucleus, where it activates target genes containing HRE-binding sites. This scheme for activation of HIF extends from mammals to invertebrates (42,44). In contrast, little is known about the means utilized by hypoxic cells to regulate HIF function.
In D. melanogaster, hypoxia-specific HIF-like DNA-binding activity was first demonstrated in so-called SL2 cells cultured from late embryos of the fruit fly (50). Subsequently, structural homologs of HIF subunits were cloned in the form of the bHLH/ PAS proteins Sima (Similar; HIF-1␣ homolog) (34) and Tango (Drosophila ARNT; HIF-1␤ homolog) (Refs. 35-37; see Ref. 27 for review). As expected for HIF proteins, Sima and Tango transcripts are ubiquitously expressed throughout embryogenesis and in nearly all tissues of Drosophila (34,36). Moreover, as in mammals, Sima stability and Drosophila HIF activity seem to be regulated in an O 2 -dependent fashion. Functionality of homologous amino acid motifs in Sima, predicted to confer normoxic instability (51), was confirmed by hypoxic inductions of luciferase reporter plasmids through Sima-Gal4 fusion constructs (52). The fact that this transactivation was observed in mammalian cells indicates that Sima is able to functionally substitute for human HIF-1␣ and provides evidence for the close conservation of these signaling pathways between mammals and insects. Recent work on transgenic flies by Lavista-Llanos et al. (53) showed that in vivo HIF activity is most pronounced in developing tracheal cells, wherein proliferation is hypoxia-stimulated (see above). Both Sima and Tango are absolutely required for regulating the transcription of HREreporter constructs within tracheae of hypoxic fly embryos.
Other ␣-like bHLH/PAS proteins known to also heterodimerize with Tango in a tissue-specific manner (54) were unable to elicit this hypoxic response (53). These include the neurogenic factor Sim (Single-minded) (55) and, most notably, Trh (Trachealess), the master regulator that drives early tracheal development (56,57). Ss (Spineless), the Drosophila aryl hydrocarbon receptor homolog, and Dys (Dysfusion), another tracheogenetic bHLH/PAS factor, are two additional partner proteins for Tango (58,59). Therefore, Tango can form functional heterodimers with five partners: Sima, Sim, Trh, Ss, and Dys. This growing list of Tango heterodimers with close or even overlapping DNA specificities (40, 58, 60 -62) begs the question as to how, in the face of potential competition for Tango (36,54,59) and/or binding site interference (59) by coexpressed Tango partners, the signaling pathways mediated by each ␣-like bHLH/PAS protein are maintained and controlled.
To address the control and maintenance of Drosophila HIF signaling, we utilized SL2 cells (Schneider cells, line 2), which were originally cultured from late 20 -24-h-old Drosophila embryos (63), a time of maximal HIF activity during Drosophila development (53). The fact that Sim and Trh, potential Tango competitors, are not expressed in this cell line (36) makes our investigations less complicated. Thus, SL2 cells are well suited as a model for investigating the function and control of Drosophila HIF.

MATERIALS AND METHODS
Cell Culture-SL2 cells were purchased from American Type Culture Collection (CRL-1963) or received as generous gifts from Drs. A. Michelson and N. Perrimon. The cells were cultured as described in the accompanying article (87). SL2 cells were grown at 22°C in air either as monolayer or suspension cultures. Standard hypoxic challenge comprised 16-h exposure at 22°C to 1 or 4% oxygen (balance N 2 ) using an Espec BNP-210 incubator (Tabai Espec Corp.). Hypoxic exposures that differed from these standard conditions are specified below. In addition, SL2 cells were exposed for 16 h to each of the known mammalian HIF-stabilizing agents CoCl 2 and desferrioxamine (DFO) at 100 M (42,64). Efforts were made to use mid-log cells for all experiments (see below).
RNA Isolation, Northern Blotting, and Reverse Transcription (RT)-PCR Experiments-Total RNA was isolated from normoxic, hypoxic (1 and 4% pO 2 ), cobalt-treated, or DFO-treated SL2 cells using TRIzol (Invitrogen) following the manufacturer's protocol. 25-g aliquots of RNA were size-separated on denaturing 6% formaldehyde and 1% agarose gels, in which the 18 S rRNA and the processed half-molecules of the 28 S rRNA (i.e. 28 Sa and 28 Sb) (65, 66) migrated as one major band between 1.7 and 2 kb in size. Next, RNA was blotted in 20ϫ SSC buffer onto positively charged nylon membranes (GeneScreen Plus, PerkinElmer Life Sciences) and hybridized for 16 h at 42°C with 20 -30 ng of the following RT-PCR-isolated and 32  (e) lactate dehydrogenase (FlyBase symbol ImpL3 and ID FBgn0001258), 514-bp ClaI/BamHI fragment between exons 1 and 2 (amino acid 51-224); (f) CG11652 (ID FBgn0036194), 624-bp XbaI/ EcoRI fragment between exons 1 and 2 (amino acids 244 -454 ϭ C terminus); and (g) DNA (cytosine 5)-methyltransferase-2 (FlyBase symbol Mt2 and ID FBgn0028707), a 785-bp BtgI fragment between amino acids 36 and 299 (the mt2 gene contains an intronless ORF). Membranes were washed the following day with 0.1ϫ SSC and 0.1% SDS solution at 55°C and exposed to phosphoimager screens. Following stripping of the membranes, the loading control was assessed through the constitutive expression of the rpL29 gene, which encodes a structural polypeptide of the large (60 S) ribosomal subunit. Relative changes in expression between normoxic (control) and hypoxic or hypoxia-mimicking (Co 2ϩ , DFO) (64) conditions were quantified by background-subtracted object-average densitometry of the signal bands (Im-ageQuant Mac Version 1.2 software, Amersham Biosciences).
For RT-PCR experiments, first strand cDNA templates, derived from normoxia/hypoxia/cobalt/DFO-treated total RNAs, were amplified using gene-specific primers that mostly spanned from the 5Ј-untranslated region (UTR) sequences across the ATG start codon into the ORF (forward primer) and from the ORF across the stop codon into the 3Ј-UTR nucleotides (reverse primer) (see Supplemental Table 1 for compilation of all primers used in this study). A typical PCR protocol used ExTaq polymerase (PanVera) and continued for 30 cycles at 94°C for 1 min, 55°C for 1 min, and 72°C for up to 4 min. PCR products were TA-cloned into the pCR2.1/Topo cloning vector (Invitrogen) and sequenced for validation. Empirical determination of the formation of normoxic and hypoxic PCR products after various numbers of cycles helped define the exponential amplification phase for each cDNA individually. This approach was used as a second screen to monitor differential gene expression and will be referred to as RT-PCR exp . In general, RT-PCR exp yielded relative expression levels in excellent agreement with Northern data, yet with greater sensitivity.
Nuclear Extracts and Electrophoretic Mobility Shift Assay (EMSA)-Following 16-h exposures to normoxic, various hypoxic (0 -10% oxygen), cobalt, or DFO treatments, SL2 nuclear and cytosolic extracts were prepared and used in EMSAs as described in the accompanying article (87). We used the following wild-type HRE oligonucleotide (W18) of the 3Ј-enhancer of the human erythropoietin gene (67) as a double-stranded 32 P-labeled probe and as an unlabeled competitor for all EMSA experiments reported herein: 5Ј-AGCTTGCCC(TACGTGCT)GTCTCAG-3Ј (where the HRE is in parentheses and the core bases essential for HIF binding are underlined; only the sense strand is given). A large excess of the following M18 mutant oligonucleotide (0.75-1.5 g) was also used in these binding reactions to minimize nonspecific (but not HIF) binding to the W18 probe: 5Ј-AGCTTGCCC(TAAAAGCT)GTCTCAG-3Ј (where the mutated core bases are underlined).
Cloning of Full-length and Splice Variant Sima cDNAs into pAc5.1-Dr. S. Crews kindly donated a pAc5/Tango expression construct containing the full-length 1.9-kb cDNA (start-to-stop codon) of the Drosophila tango gene. Overexpression and cotransfection experiments of Drosophila HIF necessitated the cloning of the bHLH/PAS protein Sima as the functional fly HIF␣ homolog (52,53). Well suited for the expression in SL2 cells, the pAc5 plasmid (version A used here; Invitrogen) contains the 2.5-kb promoter of the Drosophila actin 5C gene upstream of the polylinker region for high level constitutive expression of the inserted gene of interest (68). Also, upon removal of endogenous stop codons, pAc5.1 proteins are expressed with C-terminal tags in the form of poly-His sequences and the viral protein-derived V5 epitope (69), allowing for easy product detection using a commercially available anti-V5 antibody (Invitrogen).
An ϳ4.5-kb amplicon was obtained in Sima RT-PCRs by using a start codon-spanning forward primer (Sima-F) and a stop codon-spanning reverse primer (Sima-C T ) (Supplemental Table 1). The forward primer had a KpnI site and the reverse primer an XbaI site (in replacement of the endogenous stop codon) added to their respective 5Ј-ends. In these RT-PCR experiments, an ϳ2-kb product was consistently co-amplified using the same primers, whereas additional amplicons at ϳ1.2 and ϳ3.3 kb were only irregularly observed. The 2-and 4.5-kb products were both initially cloned into the pCR2.1/Topo vector and sequenced to completion using exon-specific forward and reverse primers (Supplemental Table 1). The 4.5-kb cloned PCR product contained the fulllength start-to-stop codon cDNA of the sima gene composed of exons 1-12. It was ligated into the KpnI/XbaI sites of pAc5.1 in the form of two PCR artifact-free (i.e. no premature stop codons) subfragments from different pCR2.1 precursor clones, thus generating the expression construct pAc5.1/Sima.
On the other hand, the 2-kb product represented an exon 1-7/12 splice variant of the sima gene. Fusion of exons 7 and 12 resulted in a frame-shifted exon 12 translation frame. A new reverse PCR primer was generated (Sima2-C T ) (Supplemental Table 1) that spanned the exon 7-12 junction, including the first 19 nucleotides of exon 12, and that replaced the following consecutive stop codons, which resulted from the frame-shift, with an XbaI site for cloning. Together with the forward primer Sima-F, Sima2-C T was used to amplify the 1.3-kb translation frame of the splice variant Sima (svSima) product from 2-kb exon 1-7/12 pCR2.1 precursor templates. Following a KpnI/XbaI double digestion, 1.3 kb-PCR products were inserted into these sites of pAc5.1, generating the pAc5.1/svSima construct.
SL2 Cotransfections-Cotransfections of SL2 cells (10 ml at 0.5 ϫ 10 6 cells/ml) were conducted using a calcium phosphate kit (Invitrogen) following the manufacturer's recommendations (70,71). Cells were transfected for 16 h with a total of 22.5 g of DNA using the pUC18 plasmid as filler DNA where necessary. This amount of DNA included 1 g of HRE-luciferase reporter, 0.25 g of pAc5.1/lacZ plasmid for normalization of luciferase activities to varying transfection efficiencies, 0.25 g of HIF expression plasmids (i.e. pAc5.1/Sima, pAc5/Tango, pAc5.1/svSima), and 0.65-6.5 g of pAc5.1/svSima plasmid in titration experiments (see below). To generate a suitable HRE-luciferase reporter, a Ϫ462 base deletion product of the promoter of the hypoxiainducible hemoglobin-2 gene from the freshwater crustacean Daphnia magna, housing three HREs at positions Ϫ258, Ϫ146, and Ϫ107, two of which are able to bind human, Drosophila, or Daphnia HIF complexes (see accompanying article (87)), was cloned into the KpnI/MluI sites upstream of the luciferase gene in the pGL3-Basic luciferase plasmid (Promega) (denoted by HRE in Fig. 7). A 1:4 (0.25:1 g) ratio between expression and reporter plasmids was adopted to obtain maximal hypoxically induced transactivations of the luciferase gene as seen in pilot transfections. Negative controls included plasmid-free and insert/HREfree pGL3-Basic luciferase vector-only transfections. After transfection, culture of washed cells was continued in new medium for another 48 h. Halfway through this period, transfected cultures were split into two equal size populations for parallel normoxic (21% pO 2 at 22°C) versus hypoxic (1% pO 2 at 22°C) exposures, which occupied the final 16 h prior to harvest. Normalized luciferase activities (i.e. luciferase/␤-galactosidase activity ratios) are expressed as a percentage of the total sum of all normoxic and hypoxic activities, with the hypoxic/normoxic activity ratio expressed as -fold hypoxic induction.

RESULTS AND DISCUSSION
To initiate our investigation of the identity of SL2 HIF components, Northern blot and RT-PCR experiments were carried out to test the expression of the following bHLH/PAS transcription factors in Schneider cells: Sima, Sim, Trh, Dys, and Ss as HIF ␣-like subunit proteins (29) and Tgo (Drosophila ARNT) as a HIF ␤-like subunit protein (Fig. 1a). Although Sim and Trh, in agreement with earlier reports (36), and Ss were not expressed in SL2 cells, the Dys transcript was successfully amplified using RT-PCR (data not shown). Additionally, the Northern blot in Fig. 1b demonstrates abundant and constitutive expression of Sima and Tango, regardless of different conditions: exposure to normoxia, hypoxia (1 and 4% oxygen), or hypoxia mimetics (cobalt or desferrioxamine). The constitutive expression of Sima and Tango mRNAs in Fig. 1b suggests that HIF in Drosophila, as in mammals, is regulated by a posttranscriptional control mechanism involving the oxygen-dependent stability of the ␣-subunit. Of all Drosophila HIF␣ candidates (Fig. 1a), the Sima protein uniquely contains an ODD-like region including a highly conserved sequence around proline 850 (51) as a potential hydroxylation target (Fig. 1c). For this reason, we sought to assess the activity of Drosophila HIF along with the stability of the ␣-subunit candidate Sima as a function of pO 2 . Fig. 2a shows a typical gel shift experiment using an HRE probe (W18) derived from the human erythropoietin enhancer (67) along with nuclear extracts from normoxic (21% pO 2 ; lane 2) or hypoxic (1% pO 2 ; lanes 3 and 4) Schneider cells. SL2 HIF was found to be represented by two retarded bands (black and white arrows) that showed a marked increase in activity in hypoxic extracts (lane 3 versus lane 2). The binding specificity of Drosophila HIF to the human erythropoietin enhancer W18 probe is demonstrated in lane 4, where addition of 100-fold excess unlabeled competitor probe resulted in the disappearance of both bands. Following a 16-h exposure of mid-log phase SL2 cells to different oxygen tensions, maximal HIF activity was detected at 1% oxygen, with progressively declining activity levels toward both anoxia and higher oxygen tensions (negligible activity at pO 2 Ն 10%) (Fig. 2b). At pO 2 Ն 1%, a doublebanded gel shift was observed (lanes 3-5, black and white arrows), whereas nuclear extracts from Ͻ1% pO 2 incubations consistently contained only the lesser mobility complex (lanes 1 and 2, black arrow). Thus, SL2 HIF can assemble into different complexes in an oxygen-dependent manner. Phosphoimaging densitometry of the top complex showed that mean SL2 HIF activities were maximal at 1% oxygen (ϳ4-fold induced relative to normoxic levels), establishing it to be a hypoxic, rather than anoxic, response (Fig. 2c). In comparison with the in vivo transcriptional activity of Drosophila HIF, with a reported pO 2 maximum at 3-5% and sub-normoxic activity levels at 1% oxygen (53), the DNA binding response of SL2 HIF in Fig. 2c was markedly shifted toward more severe O 2 reductions and even included slightly elevated anoxic HRE attachment. Fig. 2d is a representative Western blot of endogenous Sima protein in SL2 nuclear extracts as a function of pO 2 . In correlation with the pO 2 response of SL2 HIF activity, the abundance of Sima protein (band at 180 kDa) also peaked at 1% oxygen, declined toward both anoxia and moderate hypoxia (2.5 and 5% pO 2 ), and was undetectable under normoxic conditions. In addition, a 98-kDa band stained with similar intensity as the Sima 180-kDa band, but displayed linearly declining levels of abundance from anoxia to normoxia. Finally, a fainter 75-kDa band, probably representing an alternative splicing product of the sima gene, was detected in hypoxic but not anoxic extracts and thus might be a constituent of the faster mobility "hypoxia-only" gel shift complex noted in Fig. 2b (white arrow). In corroboration of the notion that SL2 HIF is comprised of not one, but several species derived from alternative Sima products, supershift experiments on anoxic SL2 nuclear extracts using the above anti-Sima antiserum resulted in a band split, i.e. the generation of two new bands, one being up-shifted and the other being down-shifted in comparison with the single original complex (data not shown). Taken together, the correlation between the pO 2 dependence of HIF activity and Sima stability, both showing maxima at 1% oxygen, strongly suggests that Sima acts as a functional HIF␣ homolog in hypoxic SL2 cells.
Although a few hypoxia-responsive fly genes have been reported (see Ref. 72 for summary), we have very limited understanding of the biological scope of Drosophila HIF or any other invertebrate HIF. As a first step toward this goal, we conducted a genome-wide computational survey for potential HIF target genes in both Drosophila and C. elegans as characterized by the presence of reportedly functional HRE sequences within 600 nucleotides upstream from start codons and downstream from stop codons (72). One of the genes we retrieved as a multi-HREflanked ORF was glob1, the recently discovered (73) singlecopy gene for Drosophila globin that generates a high O 2 affinity (P 50 ϳ 0.1-0.2 torr) hemoglobin in association with the tracheal system and the fat body of fly embryos, larvae, and adults (74). Fig. 3a shows a diagram of the globin gene, including TATA box-like transcription start sites and potential HIFbinding HREs in 5Ј-and 3Ј-flanks as well as the coding region. At least four different globin transcripts from two different promoters are produced in vivo (74). All four mRNAs share the same coding region (exons 1-3) (Fig. 3a) and 3Ј-UTRs, but differ in their 5Ј-ends. Fly embryos express two globin transcripts, termed A (predicted ϳ1.0-kb size) and C (predicted ϳ1.3-kb size), from different promoters upstream of a respective 5Ј-UTR exon (Fig. 3a, hatched boxes labeled 5ЈA and 5ЈC, respectively) (74).
The wealth of flanking HREs, together with the fact that expression of several other invertebrate globin genes is known to be induced under low oxygen (see accompanying article (87)), led us to evaluate, by Northern blotting and exponential phase (RT-PCR exp ), the response of globin A and C mRNAs to graded hypoxia in Schneider cells (Fig. 3b). To our surprise, a progressive down-regulation, rather than up-regulation, of both globin messages was observed (Globin panels). This suppression, already evident after 2 h at 1% oxygen, continuously increased during 16 h of hypoxia, when levels of globin transcripts declined by ϳ80 -85% compared with normoxic controls. Completely unaffected by this hypoxia time course, Sima and Tango transcripts are presented here as loading controls. The Northern blot results (lanes 1-5) were in good agreement with the exponential phase RT-PCR results (lanes 6 -10).
We next addressed whether SL2 HIF mediates the hypoxic down-regulation of globin, which would, if true, represent the first description of a transinactivator role conferred by an invertebrate HIF. A few cases of gene silencing are known to be directly triggered by human HIF signaling (75,76). In normoxic mammalian cells, HIF can be fully activated by so-called hypoxia mimetics such as iron chelators (e.g. DFO) and transition metals (e.g. Co 2ϩ , Ni 2ϩ , and Mn 2ϩ ions) (64), all of which inhibit proline hydroxylation of the ␣-subunit, thus stabilizing HIF even under high pO 2 (42). In contrast, and in agreement with previous reports (50), SL2 HIF was predominantly active in hypoxia, weakly so following DFO treatment of cells, and refractory to Co 2ϩ ions (Fig. 3c), in accord with the following scheme of relative activities: hypoxia Ͼ DFO Ͼ Ͼ normoxia Ӎ cobalt. As demonstrated in the representative Western blot in Fig. 3c, Sima abundance correlated with the above HIF activity ranking (i.e. exposure to hypoxia and DFO, but not cobalt, led to measurable Sima accumulation), once again underscoring Sima protein as the determining factor in controlling fly HIF function. The effect of these treatments on the regulation of potential HIF target genes is summarized in Fig. 3d. Northern blot and exponential phase RT-PCR analyses demonstrated that, compared with normoxic controls, treatment for 16 h at 1% oxygen or with 100 M DFO caused maximal and moderate suppression of the globin A and C transcripts, respectively, whereas cobalt treatment had no significant effect on globin mRNAs (Globin panels). To confirm that this differential expression pattern is HIF-derived, rather than from an iron chelation/DFO-mediated inhibition of heme biosynthesis, which, in turn, could lead to moderate down-regulation of fly globin transcription, it was necessary to look for similar responses in other HRE-flanked genes encoding heme-free products. These genes included Drosophila lactate dehydrogenase (ImpL3, CG10160 (77), whose mRNA levels are hypoxia-inducible and, in addition, markedly suppressed by Sima RNA interference (44); CG11652, a cell cycle-regulating homolog of the human candidate tumor suppressor protein OVCA1 (78); and Drosophila single-copy DNA (cytosine 5)-methyltransferase (Mt2), to determine the pO 2 dependence of SL2 HIF activity (Drosophila HIF bands indicated by the black and white arrows) in response to seven different oxygen tensions (indicated above each lane). N, normoxia at 21% oxygen. c, mean HIF activity (top complex, black arrow in b) as a function of pO 2 (i.e. 0, 0.5, 1, 2.5, 5, 10, N ϭ 21% oxygen) obtained by phosphoimaging densitometry on three EMSA series. The mean Ϯ S.E. of the activity for each pO 2 is shown as -fold induction relative to normoxic activity set to 1.0. d, Western blot of endogenous Sima protein (180 kDa) present in SL2 nuclear extracts as a function of pO 2 (indicated above each lane). Rat anti-Sima antiserum was used for Sima detection. Cross-reactivity with 98-and 75-kDa bands (?) is discussed in "Results and Discussion." which reportedly is both necessary and sufficient for DNA methylation in fruit fly embryos (79 -83).
Northern blot and RT-PCR exp experiments revealed similar expression changes for the lactate dehydrogenase and CG11652 transcripts (Fig. 3d, LDH and CG11652 panels), with maximal induction upon hypoxic treatment (ϳ8-fold for lactate dehydrogenase and ϳ3-fold for CG11652), milder up-regulation triggered by DFO, and unaltered low level expression under normoxic or cobalt conditions. In contrast, Mt2 expression (Mt2 panels), like that of globin, was strongly inhibited by hypoxia and weakly inhibited by DFO and was not impacted by cobalt treatment compared with normoxia. As noted above (Fig. 1b), none of these treatments had any effect on Sima transcript levels (Sima panels; with the top band shown as a loading control). Therefore, the observed correlation between normoxia/hypoxia/cobalt/DFO-induced HIF activities or Sima abundance and the differential lactate dehydrogenase, CG11652, globin, or Mt2 expression levels demonstrates that SL2/fly HIF is able to confer both positive or negative regulations on HRE-flanked target genes.
During our assessment of Sima expression by Northern blot and RT-PCR analyses, we consistently observed one smaller band in particular (Fig. 3, b and d, Sima panels, white arrows).
In striking contrast to full-length Sima mRNA, this smaller transcript was not constitutively expressed. Rather, its abundance increased linearly throughout the 2-16-h hypoxia time course (Fig. 3b) until, after 16 h of exposure at 1% oxygen, the levels of this transcript were 2-3-fold induced over normoxic controls (Fig. 3, b and d). Yet even at this peak abundance, the relative amount of the smaller transcript never exceeded ϳ13% of full-length Sima mRNA. Since the same set of PCR primers, which annealed against the start codon region of the first exon and the stop codon region of the last exon of Sima, were able to generate both the full-length (ϳ4.5 kb) and smaller (ϳ2.0 kb) amplicons in these competitive PCRs, both transcripts had at least these exons in common. Thus, we set out to clone and sequence both Sima amplicons in parallel. As shown in Fig. 4a, sequencing of both PCR products identified them as Sima transcripts, generated by alternative splicing. The 4.5-kb PCR product corresponded to the 12-exon fulllength Sima (flSima) transcript. In contrast, the 2-kb amplicon was obtained from a transcript that was composed of exons 1-7 plus exon 12. Lacking exons 8 -11 in the variant mRNA agrees with the observed 2.5-kb reduction in amplicon size compared with full-length mRNA. Therefore, the smaller hypoxia-induced transcript encodes a natural ⌬ODD Sima splice variant (svSima), which also lacks all polyglutamine repeats, whereas it includes the entire N-terminal bHLH and PAS region (exons 1-7). In svSima, fusion of exon 7 with exon 12 resulted in a frame-shift in which only the first 19 nucleotides of exon 12 could be translated, thus generating a novel 7-amino acid C terminus (i.e. RMTIPKQ) (Fig. 4, b and c). Supplemental Fig. 1 presents the amino acid sequences predicted for flSima and svSima in alignment with Sima cDNA (34). From this alignment, one can see that, with few PCR or sequencing inaccuracies, flSima and svSima polypeptides are identical in sequence throughout amino acids 1-419 and that the end of the PAS region at amino acid codon 420 maps precisely to the exon 7/8 (flSima) or 7/12 (svSima) junction and thus to the end of the svSima reading frame prior to its unique 7-amino acid C terminus. Therefore, svSima mRNA is predicted to generate a 426-amino acid bHLH/PAS "mini" Sima protein that lacks the ODD and all polyglutamine motifs. Fig. 4c summarizes differences in the amino acid sequences between Sima cDNA, flSima, and svSima expression constructs, confirmed by DNA sequencing of at least two flSima or svSima clones. These substitutions probably represent polymorphic codons of the sima gene.
Obvious parallels between IPAS, the mammalian dominantnegative feedback regulator of HIF signaling (85,86), and svSima are apparent. Both are hypoxia-induced alternative splice variants of HIF␣ genes that generate bHLH/PAS truncations of similar (IPAS) or identical (svSima) sequences compared with their full-length equivalents. One might expect that, being devoid of the ODD, svSima would be a constitutively stable slow turnover HIF␣ protein in contrast to flSima. The anti-Sima antiserum used in Fig. 2d was directed against amino acids 284 -498, of which only amino acids 284 -419 are retained in svSima, which probably explains our inability to detect endogenous svSima protein in nuclear or cytosolic extracts from SL2 cells. This conclusion is supported by the failure of this anti-Sima antiserum to detect svSima overexpressed in transfected SL2 cells (data not shown).
To explore the functional differences between flSima and svSima, both proteins were tagged and overexpressed in SL2 cells. For this purpose, the 4.5-kb flSima and 1.3-kb svSima translation frames (see "Materials and Methods") were subsequently subcloned into an insect-specific pAc5.1 expression vector, enabling the addition of C-terminal V5 epitope and poly-His tags (Fig. 4b). The Western blot in Fig. 5 documents, for overexpressed flSima, the expected ϳ180-kDa band (Fig.  2d) in nuclear extracts with enhanced concentration in hypoxic samples (lanes 2 and 4) and, for overexpressed Tango protein, a more or less constitutively stable ϳ83-kDa band that also predominates in nuclear extracts (lanes 1-4). In vivo, Tango protein specifically translocates from the cytosol to the nucleus only in the presence of an ␣-like partner protein such as Sim or Trh (54,58,59). To find it within the nuclei of normoxic SL2 cells (lanes 1 and 3) suggests the presence of a Tango partner protein other than endogenous Sima, Sim, or Trh (see above). In contrast, the ϳ55-kDa svSima band was observed exclusively in the cytosol (lanes 6 -8). Removal of exons 8 -11 apparently deleted sequences that serve as nuclear localization signals of the Sima polypeptide. The Sima exon 1-7/12 spliced mRNA thus creates a ⌬ODD HIF␣ polypeptide that also seems to be devoid of nuclear localization signals. In addition, and to our surprise, svSima protein, rather than being constitutively stable, was far more abundant in hypoxic compared with To better understand the markedly enhanced hypoxia stability of a ⌬ODD HIF␣ protein, Sima Northern blotting of SL2 cells transfected with flSima and svSima was conducted to determine whether the predominant occurrence of svSima protein under hypoxic conditions could be explained by differential oxygen-dependent stabilities of its overexpressed transcript. In these Northern analyses, we indeed found marked degradation of transfected svSima and, to a lesser extent, flSima transcripts particularly at normoxic pO 2 (data not shown). This suggests that the augmented normoxic degradation of overexpressed svSima mRNA might at least contribute to the seemingly greater abundance of this splice variant protein in hypoxic cells.
To further investigate the potential biological role of svSima, we carried out Northern blot analyses of candidate HIF target genes (Fig. 3d) following transient SL2 transfection of increasing amounts of either the pAc5.1/flSima or pAc5.1/svSima construct. As shown in Fig. 6 6). Apparently, svSima functions gene-specifically and may be more effective in reversing Sima-controlled gene suppression, rather than induction, thus pointing to a selective mode of interference. Steady-state levels of the loading control transcript RpL29 were unaffected by flSima or svSima overexpression (RpL29 panel).
The fact that overexpressed flSima augmented both downregulated (globin and Mt2) and up-regulated (lactate dehydrogenase and CG11652) hypoxic transcription supports the above-mentioned correlation between differential HIF activity or Sima abundance and gene expression levels (Fig. 3, c and d). From these data, it is clear, that Sima (HIF) functions genespecifically and highly economically as a hypoxic transactivator (lactate dehydrogenase and CG11652) and transinactivator (globin and Mt2) complex, respectively. A similar duality in function has recently been established for human HIF and led to the proposed existence of activating versus silencing HIF/ HRE⅐coactivator complexes (75,76). The above observations document that far ranging physiological functions in Drosophila such as oxygen transport/scavenging (globin), DNA methylation (Mt2), glycolytic reactions (lactate dehydrogenase), and cell cycle control (CG11652) are all oxygen-dependent and impacted by fly Sima/HIF. Moreover, since transiently expressed svSima was able to antagonize HIF-driven suppression (globin and Mt2) far better than induction (lactate dehydrogenase and CG11652), the data presented in Fig. 6 suggest different susceptibilities to svSima interference of suppressive versus inductive HIF⅐coactivator complexes.
For additional studies of the functional differences between fly HIF proteins, SL2 cells were cotransfected with HRE-luciferase reporter constructs along with flSima, svSima, and Tango expression plasmids and incubated under normoxic or hypoxic pO 2 . The luciferase reporter contained a triple-HRE portion of the promoter of the hypoxia-inducible hemoglobin-2 gene from the freshwater crustacean Daphnia magna that was cloned upstream of the luciferase gene in the pGL3-Basic plasmid (Fig. 7, inset). Two of these three HREs are able to bind human, Drosophila, or Daphnia HIF proteins (see accompanying article (87)). As shown in Fig. 7 (sections 1-3), all negative controls (i.e. plasmid-free and insert-free reporter plasmid transfections, with and without cotransfected flSima) failed to show any significant luciferase activities. Similarly, transfec-  1.3kb; lower). Identical sequences (gray backgrounds), amino acid codons (light-blue lines), and phase 2 exon 7/8 (Sima cDNA and flSima) and 7/12 (svSima) junctions (yellow line) are indicated in the nucleotide sequence alignment. In the Sima 1.3kb sequence, exon 12 (red), the XbaI restriction site (brown), and pAc5.1 expression vector DNA (blue), including the V5 epitope, poly-His motif, and vector-provided stop codon (gold) are indicated. c, summary of all amino acid differences between Sima cDNA (upper) (34) and the full-length (4.5 kbfl; middle) and splice variant (1.3 kbsv; lower) expression constructs. cDNA 7 flSima or svSima substitutions were confirmed in at least two independent Sima clones and, therefore, probably represent polymorphic codons of the sima gene. Note the increased substitution frequency toward C termini between Sima cDNA and flSima polypeptides, particularly within exon 12 (positions 1361, 1447, 1454, and 1497), possibly in reflection of region dispensability for some Sima protein products (svSima). The unique svSima C terminus is shown in red. Q 2 and Q 3 indicated two and three consecutive glutamine mismatches, respectively. tion of the HRE-reporter alone yielded only base-line activation of the luciferase gene, suggesting that low levels of active endogenous HIF complexes are present in the SL2 batches used for these transfections (section 4). However, cotransfecting the HRE-reporter with the flSima construct alone (section 6a) transactivated specifically the hypoxic reporter expression and resulted in an ϳ8.5-fold induction under 1% oxygen, in good agreement with previous Sima-Gal4 fusion construct UAS/luciferase assays in Hep3B cells (52) and with the predominant occurrence of the protein in hypoxic nuclei of SL2 cells (Fig. 5). In marked contrast, neither Tango (section 7a) nor svSima (section 5) cotransfections were able to stimulate the reporter under any (normoxic or hypoxic) condition above the vector controls used. Since Sima is known to interact only with either Tango or human ARNT to generate functional HIF complexes (36,52) and since no evidence for Sima⅐Sima homodimers has been reported, it is likely that the hypoxic induction of reporters in response to Sima cotransfections (Fig. 7) (52) depends on heterodimers formed between overexpressed Sima and endogenous Tango/ARNT polypeptides. Failure of overexpressed Tango to activate the HRE-reporter also indicates that endogenous Sima is limiting for HIF-driven gene regulation. The failure of svSima to confer stimulated luciferase expression agrees well with its exclusive cytosolic residence (Fig. 5) and its deletion of a transactivation domain within the ODD. In comparison with flSima alone, flSimaϩTango (i.e. HIF) double transfections (section 8a) resulted in no significant change in hypoxic transcription, which again supports the presence of excess endogenous Tango in SL2 cells.
To further assess the influence of svSima on HIF function, we titrated the activity of the HRE-reporter mediated by flSima (Fig. 7, section 6b), Tango (section 7b), and flSimaϩTango (section 8b) by cotransfection with increasing amounts of the svSima expression construct. svSima decreased Sima-and SimaϩTango-driven reporter expression in a dosedependent manner while maintaining their hypoxic induction across the whole range of svSima concentrations used. Note, however, that low doses of the svSima expression construct (i.e. 0.65 g in sections 6b and 8b) primarily diminished hypoxic, rather than normoxic, reporter activities, in support of a predominantly hypoxic role of the splice variant product.
This induction-maintaining antagonism of Sima (Fig. 7, section 6) or HIF (section 8) function by svSima suggests that Tango, i.e. the HIF constitutive component, is captured by the splice variant, thus forming an unproductive cytosolic heterodimer at the expense of flSima⅐Tango complexes. This hy- FIG. 6. Agonistic and antagonistic effects of flSima or svSima overexpression on target genes. Shown are the results from Northern blot screening of lactate dehydrogenase (LDH; ImpL3), CG11652, globin, Mt2, and RpL29 (loading control) in response to mock (vec; 20 g of pAc5.1 vector), flSima (2 and 10 g of construct; filled to 20 g of plasmid with pAc5.1), or svSima (2 and 10 g of construct; filled to 20 g of plasmid with pAc5.1) transfection. 32 h post-transfection, SL2 transfectants were exposed to normoxic (N) or hypoxic (H; 1% oxygen) pO 2 values for 16 h. Double arrows highlight the most important alterations in mRNA levels of candidate HIF target genes (discussed in "Results and Discussion"). pothesis is in line with results from earlier yeast two-hybrid analyses reporting avid and specific interactions between Tango and the N-terminal portion of Sima used as the bait construct (i.e. Sima amino acids 32-488 (36); compare with Sima amino acids 1-419 present in svSima). On the other hand, mammalian IPAS forms abortive complexes in physical contact with HIF␣ proteins and consequentially reduces hypoxic induction dose-dependently in similar transfection titration experiments (85).
On the basis of our svSima overexpression results, we propose a model in which a cytosolic sima gene splice variant (svSima) serves, specifically under hypoxic conditions, to control access of Tango protein to flSima and other ␣-like bHLH/PAS partners in Drosophila cells: when svSima is present at low levels (normoxia), Tango can "freely" bind to partner proteins (e.g. Dys) and translocate into the nucleus. Once its levels begin to increase (hypoxia), svSima can more effectively sequester Tango into the cytosol and keep it away from nuclear partners (Sima and Dys), thus serving to regulate the relative activities of Tango complexes and eventually even terminating Sima/HIF-controlled gene regulation. Therefore, the sima gene expresses its own agonistic and antagonistic or regulatory protein products.
Further work is needed to extend the findings reported here to in vivo models to assess potential tissue-specific expression of svSima protein as has been reported for IPAS (85) and changes in svSima production at different developmental stages. These studies will help to determine whether svSima is involved in the fine-tuning of HIF-dependent gene expression. It will also be of interest to work out the mechanism responsible for the oxygen-dependent splicing of Sima transcripts. Finally, we need to assess the quantitative levels of the different ␣-like bHLH/PAS proteins in cells and tissues as well as their relative affinities for Tango to elucidate the impact of regulatory factors such as svSima on various Tango complexes and on bHLH/PAS protein-mediated signaling pathways.