Translational Regulation of mRNAs with Distinct IRE Sequences by Iron Regulatory Proteins 1 and 2*

Iron regulatory proteins 1 and 2 (IRP-1, IRP-2) interact with iron-responsive elements (IREs) present in the 5′- or 3′-untranslated regions (UTR) of several mRNAs coding for proteins in iron metabolism. Whereas binding of IRP-1 and -2 to an IRE in the 5′-UTR inhibits mRNA translation in vitro, it has remained unknown whether either endogenous protein is sufficient to control translation in mammalian cells. We analyzed this question by taking advantage of published mutant IREs that are exclusively recognized by either IRP-1 or IRP-2 in vitro. These IREs were inserted into the 5′-UTR of a human growth hormone reporter mRNA, and translational regulation was measured in stably transfected mouse L cells. Cells cultured in iron-rich or -depleted medium were labeled with [35S]methionine, and secreted growth hormone was immunoprecipitated. IREs with loop sequences specific for IRP-1 (UAGUAC), IRP-2 (CCGAGC), or both proteins (GAGUCG and the wild-type CAGUGC sequence) all mediated translational regulation, in contrast to a control sequence (GCUCCG) that binds neither IRP-1 nor IRP-2. Control experiments excluded IRP-1 binding to the IRP-2-specific sequence in vivo. The present data demonstrate that IRP-1 and IRP-2 can independently function as translational repressors in living cells.

In the present study, we analyzed whether IRP-2 can also function as a trans-acting repressor of translation in cells. In vitro mutagenesis has revealed alternative IRE stem-loop sequences with specificity for either IRP-1 (39) or IRP-2 (40,41). Here, we inserted several specific IRE variants into the 5Ј-UTR of a human growth hormone (hGH) expression vector and analyzed the translational repression in stably transfected L cells. Translational inhibition through both IRP-1-and IRP-2specific IREs was observed, thus providing evidence for a regulatory role of IRP-2 in cells.

EXPERIMENTAL PROCEDURES
Cell Culture-Mouse thymidine kinase-deficient L cells (Ltk Ϫ ) were cultured in ␣-minimal essential medium (Life Technologies Inc.) with 10% fetal calf serum. Stably transfected Ltk ϩ cells were selected in ␣-minimal essential medium containing 100 M hypoxanthine, 0.4 M aminopterin, and 16 M thymidine (42). Iron was modulated by culturing cells for 20 h in medium with 100 M desferrioxamine (Desferal; a gift from Ciba-Geigy, Basel, Switzerland) or for an additional 4 h with 60 g/ml of ferric ammonium citrate (Sigma). In some experiments, L cells were incubated for 45 min in medium with 100 M H 2 O 2 (Fluka, Switzerland) without fetal calf serum (essentially iron-free) to prevent the toxic effect of iron combined with H 2 O 2 . Where indicated, 1 h before the addition of iron-rich medium, 100 M MG132 (peptide aldehyde) (ProScript, Cambridge, MA) was added and maintained for 4 h. * This work was supported by the Swiss National Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: TfR, transferrin receptor; hGH, human growth hormone; IRE, iron-responsive element; IRP, iron regulatory protein; UTR, untranslated region; Ltk Ϫ , thymidine kinase-deficient L cells; 2-ME, 2-mercaptoethanol. Plasmid Constructions and Stable ;transfections-IRE mutants with different loop sequences were inserted as oligonucleotides into the 5Ј-UTR of the vector pL5-GH containing a ferritin promoter and the hGH gene (43). Double-stranded oligonucleotides with the sequence GATC-CTGCTTCAANNNNNNTTGGACGGACGGATCTT (where N represents different loop sequences) were inserted between the BamHI and XbaI sites of pL5-GH, nine nucleotides downstream of the RNA cap site. Constructs with the following mutations in the IRE loop sequence were made 2 : 40), and GCUCCG (GC1 in Ref. 40), a construct also referred to as 213-GH (19). All constructs were verified by DNA sequencing. Plasmid pFer-GH (15) was kindly provided by Dr. M. Hentze, EMBL, Heidelberg. Sub-confluent Ltk Ϫ cells were co-transfected with a plasmid containing the herpes simplex thymidine kinase gene using the calcium phosphate method (44).
Metabolic Labeling and Immunoprecipitation-Cells were washed twice in phosphate-buffered saline, incubated in methionine-free medium for 90 min, and then labeled with 40 Ci/ml of [ 35 S]methionine (Ͼ1000 Ci/mmol, Amersham Corp.) for 2-4 h. hGH was immunoprecipitated from equal amounts of secreted 35 S-labeled protein (10 5 cpm, determined by trichloroacetic acid precipitation) with a specific rabbit polyclonal antibody (Dako, Carpinteria, CA), followed by protein A-Sepharose. Samples were analyzed on 15% SDS-polyacrylamide gels, and 35 S-labeled hGH was quantitated with a Compaq PhosphorImager equipped with Molecular Dynamics Image software.
RNA Isolation and Northern Blot Hybridization-Total RNA was prepared from 3 ϫ 10 7 cells using the RNeasy Kit of Qiagen Inc. (Chatsworth, CA). Equal amounts of RNA (8 g) were separated in 1.2% agarose gels in the presence of formaldehyde, transferred to Gene-Screen Plus nylon membranes (NEN Life Science Products), and crosslinked with ultraviolet light. Equal loading was assessed from the ethidium bromide staining of ribosomal RNA prior to the transfer. Filters were hybridized with a random-primed 32 P-labeled probe comprising the 600-bp SmaI fragment of the hGH gene (45).

RESULTS AND DISCUSSION
Alternative IRE Stem-loops Discriminate between IRP-1 and IRP-2-The iron-responsive element, first identified by phylogenetic comparison in the 5Ј-UTR of ferritin mRNA (5,46), is highly conserved in evolution. The "consensus" IRE has a loop sequence CAGUGN and an unpaired cytosine immediately 5Ј of 5 paired nucleotides forming the "upper stem" (reviewed in Refs. 3 and 47). By employing in vitro screening procedures, we have recently identified alternative IRE loop sequences that preferentially bind to IRP-1 (39) or IRP-2 (40). In this study, we analyzed different IRE mutants for their ability to control translation when placed in the 5Ј-UTR of a hGH reporter construct (Fig. 1). The mutant IRE with the loop sequence G 1 AGUC 5 G (mutant GC6) was chosen as it binds to both IRP-1 and IRP-2 in vitro (40). In contrast, the loop sequences U 1 AGUA 5 C (mutant UA34) and CC 2 GA 4 GC (mutant CG125) were tested because of their exclusive interaction with IRP-1 or IRP-2, respectively (39,40). As a positive control for translation regulation, we analyzed pFer-GH, a construct which carries the ferritin H-chain wild-type IRE (15). A mutant IRE containing the loop sequence GCUCCG (mutant GC1) served as a negative control since it binds to neither IRP-1 nor IRP-2 in vitro (40).
The hGH constructs were stably transfected into mouse L cells and proteins labeled with [ 35 S]methionine in iron-deprived (Des) or iron-rich (Fe) cell cultures. Following 35 S-labeling, cytoplasmic extracts were prepared to confirm the activation and specificity of IRP-1 and IRP-2 by gel retardation analysis (Fig. 1A). Both proteins were activated in iron-deficient cells and bound strongly to the 32 P-labeled wild-type ferritin IRE (probe CG42 of Ref. 39) as well as to the mutant IRE GC6, as expected (40). However, the IRE UA34 detected only IRP-1, while the CG125 probe identified IRP-2 specifically, as previously observed (40). The negative control, GC1, revealed no binding to either protein. In cell extracts of ironreplete cells, inactive IRP-1 was fully activated by pre-incubation with 2% 2-mercaptoethanol (2-ME) (Fig. 1A, lower panel), whereas RNA-binding activity of IRP-2 could be re-activated 3-4-fold by 2-ME only in extracts of iron-deprived, but not iron-replete cells. This finding is consistent with the degrada- 2 Bold letters indicate mutated nucleotides and superscript numbers indicate position in the loop. Each construct was stably transfected into mouse L cells. Transfected cell populations were incubated either for 24 h with the iron chelator desferrioxamine (Des) or for an additional 4 h with 60 g/ml ferric ammonium citrate (Fe) to modulate RNA-binding activities of IRP-1 and IRP-2. Proteins were labeled with [ 35 S]methionine, and hGH was immunoprecipitated from the culture supernatant (panel B). To increase the contrast, lanes 1 and 2 of panel B are 4-fold longer exposed than the other lanes of the same experiment. In parallel, cytoplasmic extracts of each cell population were tested for RNA-binding activities by gel retardation assays with the wild-type or mutant [ 32 P]IREs present in the respective transfectants (panel A). Binding reactions were done in the absence or presence of 2% 2-mercaptoethanol (2-ME). The hGH mRNA expression was analyzed on Northern blots (panel C).
IRP-specific IRE Stem-loops Regulate hGH Translation in Transfected L Cells-Translation assays were performed by immunoprecipitation of secreted 35 S-labeled hGH (Fig. 1B). In iron-deprived cells, diminished translation was observed for all IRE mutants, except for GC1, without a concomitant change in hGH mRNA levels (Fig. 1C). These results indicate that IRP-1 and/or IRP-2 activated by desferrioxamine bind to the IRE mutants and thereby diminish hGH translation as compared with the situation in iron-replete cells. In agreement with previous data (48), the control ferritin IRE-hGH construct was 10-fold repressed in translation after desferrioxamine treatment (Table I). Translational inhibition mediated by the IRE mutants GC6, UA34, and CG125 was somewhat weaker (Table  I) and seemed to correlate with lower in vitro binding affinities of mutant IREs in competition assays (39,40). The GC6 mutant, which binds both IRPs 2.5 times less efficiently than a wild-type IRE, mediated a translation inhibition of only 40%. The mutants UA34 and CG125, despite being almost equivalent to the wild-type IRE as competitors, but very specific only for one or the other IRP, mediated only a 4-or 2.5-fold reduction of hGH translation, respectively. Thus, one has to be cautious in predicting the precise effect of a given mutant on the basis of in vitro measurements. Differences in the absolute mRNA levels in different cell populations probably did not affect the extent of translational regulation. In all cell populations, an excess of unbound IRP could be measured in gel retardation assays (Fig. 1A). This excess was not strongly affected in cells with higher mRNA levels. One cannot argue, therefore, that the trans-acting IRPs were titrated out by the transfected mRNA.
Another study has reported iron-independent enhancement of translation by a 5Ј-IRE but showed that this effect was insensitive to a mutation in the IRE loop (49). Our present results are in agreement since translation efficiencies in ironrich medium varied less than 1.4-fold between different cell populations excluding a positive effect on translation due to mutations in the loop.
Most importantly, the present data show that alternative IRE loop sequences with a sub-optimal binding to IRP-1 or IRP-2 are effective as cis-acting regulatory elements in cells. Moreover, an IRE that binds exclusively to IRP-2 in vitro (CG125) appears to be sufficient to confer iron-dependent translational regulation, indicating that IRP-2 itself acts as a genuine translation regulator.
IRP-2 Is Sufficient to Repress Translation of IRE-containing mRNAs in Cells-The results of Fig. 1 do not exclude the possibility that activated IRP-1 in cells may cross-react with the IRP-2-specific IRE. Therefore, experiments were performed under conditions in which only IRP-1 was activated. Transfected L cells were pre-treated for 2 h with ferric ammonium citrate to reduce IRP RNA-binding activities and then incubated for 45 min with 100 M H 2 O 2 , which is known to specifically and rapidly induce IRP-1 (28,30). Indeed, in gel retardation assays with extracts from H 2 O 2 -treated cells, we detected only complexes of IRP-1 with the wild-type ferritin and the UA34 IRE probes ( Fig. 2A). No complexes were visible for IRP-2, either with the wild-type IRE or with the IRP-2specific probe CG125 (Fig. 2A, lane 6), indicating that IRP-2 was not activated by H 2 O 2 . Immunoprecipitation of the secreted 35 S-labeled hGH revealed that H 2 O 2 -activated IRP-1 prevents translation when bound to the ferritin IRE (Fig. 2B). This agrees with findings by others (30). A clear inhibition was also observed with the IRP-1-specific UA34 construct (Fig. 2B), indicating that IRP-1 activation alone is sufficient to repress translation. However, despite a strongly induced IRP-1 activity, no regulation was observed with the IRP-2-specific IRE CG125 and the negative mutant IRE GC1. These findings indicate that IRP-1 does not bind to the IRP-2-specific IRE in cells, and confirms that the regulation mediated by CG125 in iron-deficient cells (Fig. 1) cannot be attributed to IRP-1, but depends only on IRP-2.
Attempts were also made to study the question by selectively inactivating cellular IRP-1 but not IRP-2. Ltk Ϫ cells were preincubated with 100 M MG132, a proteasome inhibitor thought  (39,40). The values indicate the inverse of the concentration at which a given IRE competes half-maximally the interaction between the radio-labeled wild-type IRE and the respective IRP. All values were normalized to the wild-type IRE, which was arbitrarily set at 1.  (Fig. 1) was selectively induced by 100 M hydrogen peroxide (H 2 O 2 ) or inactivated with 60 g/ml ferric ammonium citrate (Fe). IRP-1 activity was measured in extracts, without or with pre-incubation in 2% 2-ME, by performing gel retardation assays with the corresponding 32 P-labeled IRE mutants as probes (panel A). In parallel, translational regulation was analyzed by immunoprecipitation of [ 35 S]methionine-labeled secreted hGH in the culture medium (panel B). Results are shown for constructs with a ferritin IRE (Fer-GH) and mutant IREs UA34, CG125, and GC1. to prevent IRP-2 degradation specifically (37,38), and then exposed to 60 g/ml ferric ammonium citrate to fully inactivate IRP-1. We found that not only was IRP-2 degradation inhibited, but the inactivation of IRP-1 was also partly prevented. Its RNA-binding activity in the presence of MG132 for 4 h remained 1.6-fold higher than without MG132 (data not shown). It was not possible, therefore, to obtain cells with active IRP-2 alone.
Conclusions-In this study, we demonstrate for the first time that both endogenous IRP-1 and IRP-2 function as translation repressors in mammalian cells. This finding extends those of previous studies demonstrating that exogenously expressed IRP-1 acts as a translational repressor (50,51) and that recombinant IRP-1 and IRP-2 repress ferritin mRNA translation in reticulocyte lysates (52) and wheat germ extracts (33) in vitro. We can conclude, therefore, that both IRPs, when simultaneously activated by iron-deprivation or NO-synthase induction, exert a combined effect in cells. As the two proteins have a relatively early evolutionary origin and were detected both in insects and mammals (53), it remains to be explained why, during evolution, organisms have acquired and maintained such a duplication of function. One possible reason could relate to a potentially broader and more differentiated response to distinct signals. Hydrogen peroxide was reported to induce only IRP-1 activity, whereas IRP-2 could be detected with preference in certain cells or tissues (32,35) and under certain inductive conditions in erythroid cells (54) and rat liver (55). Another reason for the maintenance of two IRE-binding proteins may relate to the potential of specific targets. Although no naturally occurring IRP-1-or IRP-2-specific IRE sequences have yet been identified, there is a clear potential for the existence of mRNAs with such sequences. The results of the present study, which show that sub-optimal affinities of IRE-IRP interactions are sufficient for translational control, therefore suggest these RNA-binding proteins may mediate other novel functions within cells.