An Iron Responsive Element-like Stem-Loop Regulates α-Hemoglobin-stabilizing Protein mRNA*

Hemoglobin production during erythropoiesis is mechanistically coupled to the acquisition and metabolism of iron. We discovered that iron regulates the expression of α-hemoglobin-stabilizing protein (AHSP), a molecular chaperone that binds and stabilizes free α-globin during hemoglobin synthesis. In primates, the 3′-untranslated region (UTR) of AHSP mRNA contains a nucleotide sequence resembling iron responsive elements (IREs), stem-loop structures that regulate gene expression post-transcriptionally by binding iron regulatory proteins (IRPs). The AHSP IRE-like stem-loop deviates from classical consensus sequences and binds IRPs poorly in electrophoretic mobility shift assays. However, in cytoplasmic extracts, AHSP mRNA co-immunoprecipitates with IRPs in a fashion that is dependent on the stem-loop structure and inhibited by iron. Moreover, this interaction enhances AHSP mRNA stability in erythroid and heterologous cells. Our findings demonstrate that IRPs can regulate mRNA expression through non-canonical IREs and extend the repertoire of known iron-regulated genes. In addition, we illustrate a new mechanism through which hemoglobin may be modulated according to iron status.

Hemoglobin (Hb) 2 is a heterotetramer composed of ␣and ␤-globin polypeptides, each bound to a heme prosthetic group consisting of protoporphyrin IX complexed to iron. Free globins, heme synthetic intermediates and iron are all cytotoxic. Accordingly, numerous mechanisms have evolved to coordinate Hb synthesis and limit the damage caused by its individual components. Iron is of particular concern, because it catalyzes the formation of reactive oxygen species (ROS) that chemically alter proteins, lipids, and nucleic acids.
Under normal circumstances, most circulating iron is bound by transferrin, which is internalized by specific receptors on the plasma membrane of most cells. Iron uptake in erythroid precursors is modulated by availability of transferrin-bound iron, the number of functional transferrin receptors (TFR1) on the cell surface, and in iron-overload states, nontransferrin-bound iron (1). Like most other cells, erythroid precursors regulate iron through the coordinated synthesis of TFR1 and the intracellular iron storage protein ferritin via an iron sensor system consisting of iron responsive elements (IRE) and iron regulatory proteins (IRP1 and IRP2) (reviewed in Refs. [2][3][4]. Iron responsive elements are present in mRNA UTRs and form stem-loop structures that are recognized by iron-sensing IRPs. Increased iron and/or heme levels regulate IRP activities. Insertion of an [4Fe-4S] iron-sulfur cluster into IRP1 inhibits its RNA binding activity and converts the protein to the cytosolic aconitase (5). In contrast, iron stimulates proteosomal degradation of IRP2 (6,7), though a mechanism that may involve binding to heme (8 -10). Binding of IRPs to 5Ј-UTR IREs inhibits mRNA translation while IRP/IRE interactions within the 3Ј-UTR enhance transcript stability by protecting from endonuclease degradation (11). The stability of TFR1 mRNA is regulated by multiple IREs within the 3Ј-UTR (12,13). In contrast, CDC14A and SLC11A2/DMT1 mRNAs contain single functional IREs within their 3Ј-UTRs (14,15).
In addition to utilizing the generalized IRE system, erythroid cells contain mechanisms that link globin synthesis to iron availability. For example, iron promotes globin production by regulating activity of the heme-regulated inhibitor kinase, HRI, which inactivates the general translation initiation factor elF2 (16). Heme, which is produced when iron is replete, binds and inactivates HRI, enhancing globin translation. Iron deficiency depletes heme, thereby activating HRI to inhibit globin protein synthesis. Iron also regulates globin gene transcription via the nuclear heme-binding protein, Bach1. In the absence of heme, for example during iron deficiency, Bach 1 binds cis elements within ␣-globin, ␤-globin, and other genes to repress transcription (17). Binding of heme to Bach1 induces its release from DNA and proteolytic degradation, thereby promoting globin transcription (18,19). Thus, Bach1 and HRI sense intracellular iron (via heme) and adjust globin synthesis appropriately to minimize the accumulation of free subunits.
␣-Hemoglobin-stabilizing protein (AHSP) is an abundant erythroid protein that binds and stabilizes multiple forms of free ␣-globin (20,21). For example, AHSP binds "apo" ␣-globin before heme insertion to stabilize folding of the nascent polypeptide (22). In addition, AHSP binds heme-associated "holo" ␣-globin (␣Hb) and converts it into a more stable form, which inhibits the ability of heme iron to catalyze ROS production (23,24). In this fashion, AHSP may limit the toxicity of excess ␣Hb that accumulates at low levels during normal erythropoiesis and to a greater extent in certain hemoglobinopathies, such as ␤-thalassemia. Ahsp gene knock-out mice produce short-lived erythrocytes with Hb precipitates, indicating an important role in normal erythropoiesis (25). In addition, loss of AHSP exacerbates ␤-thalassemia in mice, and quantitative variations in AHSP expression may influence the phenotypes of some forms of human ␤-thalassemia (25,26).
Given the important functions of AHSP in normal erythropoiesis and its potential role as a quantitative trait modifier of ␤-thalassemia, it is important to define the mechanisms that govern AHSP gene expression. Previous studies demonstrated that the erythroid nuclear proteins GATA-1 and EKLF, and the more generalized factor Oct-1, positively regulate AHSP transcription (27)(28)(29)(30). In the current study, we identified a predicted stem-loop structure in the 3Ј-UTR of AHSP mRNA. This structure resembles an IRE, although it deviates significantly from canonical sequences. In RNA gel shift experiments this IRE-like sequence binds IRPs poorly compared with the classical FTH1 IRE. However, additional studies demonstrate that the AHSP mRNA stem-loop associates with IRPs in cells and modulates iron concentration-dependent mRNA stability. These findings extend the range of functional IRE-like structures that do not conform to previously defined IRE consensus sequences. In addition, they illustrate a new mechanism that links hemoglobin metabolism to cellular iron status.
Cell Lines and Culture Conditions-The K562 human erythroleukemia cell line was cultured in IMDM with 10% fetal bovine serum. HeLa/tTA cells (31) were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. HeLa IRP knockdown (KD) cells were grown with puromycin and G418 (32). To induce erythroid maturation, K562 cells were treated with 80 M hydroxyurea (Sigma) for 48 h. Cells were treated with either 1 mM ferric ammonium citrate (FAC) or 50 M hemin (hem) as iron sources. Desferrioxamine (DFO, 100 M) was used as an iron chelator. Cell transfection was performed with FuGENE (Roche Applied Sciences), according to the manufacturer's instructions. COS-7 and HeLa cells were transfected with 2 g of AHSP wt pEF1-␣ NEO or AHSP mut pEF1-␣ NEO and analyzed after 48 h. HeLa/tTA cells, used for experiments in Fig. 2, were transfected with 2 g of AHSP wt pTET or AHSP mut pTET and analyzed after 24 h. These experiments were performed in the absence of tetracycline to allow for full cDNA expression in this "tet-off" system.
Bioinformatics-We used RNAProfile to search for predicted stem-loops within AHSP mRNAs (33). Multispecies comparison of AHSP cDNAs was performed using The UCSC Genome Browser (34); the mouse and marmoset sequences were added by inspection. RNA secondary structures were evaluated further using RNAfold and mfold (35).
RNA Immunoprecipitation (IP)-Isolation of IRP-RNA complexes was performed as described with minor modifications (36). A detailed protocol is included in supplemental methods. Anti-IRP1 antibody was obtained from Santa Cruz Biotechnology. Anti-IRP2 antibody was provided by Tracey Rouault (37).
RNA-labeled Probe Synthesis-Radiolabeled IRE and control probes were prepared by in vitro transcription from oligonucleotide templates using T7 RNA polymerase (Invitrogen) (38). Templates consisted of a "bottom strand" oligomer containing the 17-nucleotide T7 RNA polymerase promoter sequence fused to IRE sequences. This was hybridized to a 17-nucleotide "top strand" that is complementary to the T7 promoter sequence represented in the bottom strand (supplemental Table S1). For competition assays, FPLC-purified RNA oligonucleotides were purchased from IDT (supplemental Table S1).
Electrophoretic Mobility Shift Assay (EMSA)-Purified IRPs were prepared as described (39,40). [ 32 P]UTP-radiolabeled IRE probes (1 nM) were mixed with either 100 ng of liver cytoplasmic extract or different molar ratios of purified IRP1 or IRP2 and incubated for 10 min at room temperature. Heparin (3 mg/ml) was added, and the samples analyzed on a 6% native polyacrylamide gel as described (41).
Messenger RNA Stability Assay-Actinomycin D, 5 M, (Sigma) was added to the culture medium to inhibit transcription. Cells were collect at various time points, washed with cold phosphate-buffered saline, and lysed with Trizol reagent (Invitrogen). The relative levels of AHSP and control mRNAs were determined by real time RT-PCR.
RNA Isolation and Gene Expression Analysis-Total RNA was purified by homogenizing cells in Trizol reagent (Invitrogen). Nuclear and cytoplasmic fractions of RNA were isolated as described (42). Ten million cells were incubated in 3 volumes of lysis buffer (10 mM Tris-HCl pH 7.4, 10 mM NaCl, 3 mM MgCl 2 , 0.5% Nonidet P-40) for 5 min on ice. Nuclei were recovered by centrifugation (200 ϫ g for 5 min). The supernatant was collected as cytoplasmic fraction. Nuclear pellets were washed twice with cold phosphate-buffered saline, and RNA was extracted with Trizol reagent. RNA samples were treated with 1 unit of RNase-free DNase I (Invitrogen) for 20 min at 37°C before cDNA synthesis.
Complementary DNA (cDNA) was generated using 2 g of total RNA, oligo dT18, and SuperScript II reverse transcriptase (Invitrogen). Gene expression was quantified by real-time RT-PCR (Thermocycle 7900, Applied Biosystems) using SYBR green reagent (Invitrogen) and mRNA-specific primers (supplemental Table S2). Primers were designed for standard realtime PCR using PrimerExpress (Applied Biosystems).

Identification of a Putative Stem-Loop
Structure in the 3Ј-UTR of AHSP mRNA-Iron regulates the expression of numerous genes via IRE stem-loop structures present in the UTRs of mRNA. We investigated whether AHSP mRNA might be subject to this mode of regulation by using the program RNAProfile to search for segments of the UTRs of human AHSP mRNA that have the potential to form stem-loops. A prominent potential stem-loop (⌬G ϭ Ϫ9.50 kcal/mol) is found at the 3Ј-end of the mRNA, with the polyadenylation signal as part of the predicted stem ( Fig. 1). A similar predicted stem-loop structure is found in the homologous sequences of other simian primates, including the rhesus macaque, an Old World monkey, and marmoset, a New World monkey.
This stem-loop structure has several features in common with canonical IREs, including a 5-bp upper stem separated from a lower stem of variable length by an unpaired nucleotide in a bulge (43)(44)(45). The bulge can have a single unpaired nucleotide as in the IREs in the 3Ј-UTR of TFR1 mRNA (Fig. 1C). The apical loop of most IREs has the consensus sequence CAGUGH (where H is A, C, or U), with the C forming a base pair with the second G to generate an AGU pseudotriloop structure that is recognized by IRP1 (46). The sequence of human AHSP mRNA has a matching sequence, CAGUGC. However, the predicted stem-loop structure for human AHSP mRNA differs in several aspects from canonical IREs. One predicted structure has the CA in the CAGUGC loop consensus motif involved in base pairing in the upper stem ( Fig. 1C, structure 1). In an alternate conformation, the AGU is exposed and anchored by a G-C intraloop pair (structure 2 in Fig. 1C), but this secondary structure is expected to be less stable. The preferred nucleotide in the bulge is C in all known natural IREs, whereas the predicted structures for AHSP mRNA have an A. The A nucleotide in the bulge is part of the AAUAAA polyadenylation signal.
Numerous studies have identified RNAs with non-canonical sequences that bind IRPs (44,45,47). In addition, several naturally occurring functional IREs that deviate from canonical criteria have been described (15, 48 -53). We investigated further the functional properties of the IRE-like stem-loop element present in AHSP mRNA. We discovered that this stem-loop associates with IRE-binding proteins and modulates mRNA stability in an iron-dependent fashion. Therefore, we refer to this structure as an IRE-like stem-loop because it shares some properties with classical IREs but also exhibits significant differences in sequence, predicted secondary structure and some functional features.
An AHSP IRE-like stem-loop is not apparent in mammals outside the clade of simian primates. As expected, the AAUAAA polyadenylation signal is conserved among all mammals including the marsupial opossum, but the sequences homologous to the apical loop and the 3Ј-part of the stem show many differences outside simian primates (Fig. 1B). Many of these sequences are predicted to form stem-loops, but they differ from those expected for IREs in the length of the stems, the presence of a bulge 5 bp away from the loop, and the sequence of the loop. Thus, the phylogenetic comparisons indicate that this IRE-like structure is primate-specific. It may be an adaptation in primates to provide an additional mechanism for ironmediated regulation of AHSP.
The AHSP IRE-like Stem-Loop Associates with IRPs in Cells-We used RNA co-immunoprecipitation to study human AHSP mRNA binding to IRP1 and IRP2, the two major known IREbinding proteins (54). We introduced a full-length wt AHSP cDNA expression construct into HeLa cells and immunoprecipitated IRP1 and IRP2 complexes from cytosolic extracts. Coimmunoprecipitated RNAs were quantified by real-time RT- Within the loop, a line is drawn between the paired C-G in the CAGUG motif. The first AHSP structure gives the lowest predicted free energy (⌬G) by folding programs such as RNAfold. The second structure forces the AGU in the CAGUG motif to be unpaired, as is found in the structure of an IRE-IRP1 complex.
PCR and normalized to the same mRNA present in extracts prior to immunoprecipitation (input sample). Ectopically expressed AHSP wt mRNA was enriched significantly by immunoprecipitation with IRP1 or IRP2 antibodies, but not control IgG ( Fig. 2A). Identical results were obtained using two independent sets of PCR primers derived from different regions of AHSP cDNA (data not shown). FTH1 mRNA examined as a positive control also co-immunoprecipitated specifically with IRP1 and IRP2 ( Fig. 2A). Glyceraldehyde phosphate dehydrogenase (GAPDH) mRNA, which does not contain an IRE, was not enriched by IRP immunoprecipitation (Fig. 2A). Endogenous AHSP mRNA was also enriched in IRP1 and IRP2 complexes immunoprecipitated from cytosolic extracts of K562 erythroleukemia cells (Fig. 2B). Together, these data suggest that AHSP mRNA associates with IRP1 and IRP2 in cells.
To determine if the AHSP mRNA-IRP interaction is mediated by the IRE-like stem-loop, we generated AHSP mut cDNA, which contains three mutations that disrupt the structure (Fig.  2C). This mutant cDNA was expressed normally in HeLa cells (not shown), but failed to bind IRPs 1 and 2 ( Fig. 2A). Hence, AHSP mRNA associates with IRPs 1 and 2 via the IRE-like stemloop present in the 3Ј-UTR. We also investigated the effects of iron on AHSP mRNA binding affinity to IRPs. IRP1 and IRP2 complexes were immunoprecipitated from AHSP wt mRNA-expressing HeLa cells treated with either ferric ammonium citrate (FAC) as a source of iron or desferrioxamine (DFO), an iron chelator (Fig. 2D). Treatment with DFO moderately enhanced co-immunoprecipitation between IRP2 and AHSP mRNA (p Ͻ 0.05). In contrast, FAC treatment markedly inhibited co-immunoprecipitation of AHSP mRNA with both IRPs 1 and 2 (Fig. 2D), without impairing cell viability (not shown). Overall, our data indicate that in cells, the AHSP mRNA IRE-like stem-loop associates with IRPs in an iron-regulated fashion, consistent with classical IRE/IRP interactions described for other transcripts, including FTH1, TFR1, and ALAS2 (46,55,56).
The AHSP IRE-like Stem-Loop Binds Poorly to IRPs in Vitro-We performed RNA EMSAs to further study AHSP IRE-like stem-loop-IRP interactions (Fig. 3). A [ 32 P]UTP-labeled RNA oligonucleotide representing the IRE-like structure from human AHSP mRNA was incubated with several concentrations of purified IRP1 and IRP2 and RNA-protein interactions examined by non-denaturing polyacrylamide electrophoresis (Fig. 3A). Human FTH1 IRE probe was used as a positive control for binding to IRPs. Upon incubation with the FTH1 IRE probe (1 nM), both recombinant IRP1 (1 nM) and IRP2 (1 nM) showed prominent band-shifts, indicating the formation of RNA:protein complexes (Fig. 3A, lanes 1 and 2). In contrast, the AHSP IRE-like stem-loop probe exhibited very weak IRP binding activity. Formation of RNA-IRP1 complex was visualized only at RNA/protein ratios of 1:5 and 1:10 ( Fig. 3A, lanes 4 -6). No band shift was visualized at any concentration of IRP2 (Fig. 3A,  lanes 7-9). Next, we performed gel shift competition assays. Liver cell extracts (as a source for IRP1 and IRP2) were incubated with [ 32 P]UTP-labeled FTH1 IRE probe and various unlabeled competitor RNAs, then formation of RNA-protein complexes was evaluated by EMSA (Fig. 3B). At 5-fold competitor excess, unlabeled FTH1 IRE probe efficiently competed with labeled probe, disrupting most of the [ 32 P]RNA-protein complex (Fig. 3B, lane 2 and C). In contrast, higher concentrations of the unlabeled AHSP probe were required for competition with the FTH1 IRE (Fig. 3B, lanes 6 -9 and C). The reduced FTH1 probe complexes resulting from addition of 30 -50-fold excess cold AHSP probe (lanes 8 and 9) appear to represent specific but relatively low affinity binding of the latter to IRPs, because the same concentrations of a control scrambled unlabeled probe had no effect (Fig. 3B, lanes 12-13 and C).
In summary, radiolabeled AHSP IRE-like stem-loop probe binds purified IRP1 only weakly and no binding to IRP2 was detected in EMSA experiments. This is consistent with findings that in vitro binding of IRP2 to IRE stem-loops exhibits more stringent requirements for nucleotide sequences in the bulge region (57). These EMSA results contrast with our findings that both IRP1 and IRP2 specifically co-immunoprecipitate with AHSP mRNA in cell extracts (see "Discussion").
Iron Destabilizes Endogenous AHSP mRNA in Erythroleukemia Cells-We next sought to investigate functional properties of the AHSP IRE-like stem-loop. Binding of IRPs to 3Ј-UTR IREs increases mRNA stability by protecting against endonucleolytic cleavage (13). The most studied example is TFR1 mRNA, which contains multiple IREs in the 3Ј-UTR (13, 58 -62). Under iron-deficient conditions, IRPs bind to TFR1 transcripts and increase their half-life. In contrast, high levels of iron cause release of IRPs, promoting mRNA degradation. We examined the effects of iron on the stability of endogenous AHSP mRNA transcripts in K562 erythroleukemia cells (Fig.  4A). Cells were treated for 16 h with FAC as an iron source or DFO to deplete iron. Actinomycin-D was added to arrest transcription and the kinetics of mRNA decay was measured by A, HeLa cells were transfected with plasmids encoding full-length human AHSP cDNA, either wt or a mutant (mut) forms containing three nucleotide substitutions that disrupt the stem-loop (C). Cytoplasmic extracts were incubated with specific antibodies for IRP1, IRP2, or control IgG. Antibody complexes were purified with protein G beads and the associated mRNAs quantified by real-time RT-PCR. The y-axis shows relative mRNA levels normalized to input (non-immunoprecipitated) RNA from each sample. Results are shown as mean values from three independent experiments with error bars indicating S.D. B, co-immunoprecipitation of IRPs and endogenous AHSP mRNA in K562 erythroleukemia cells, performed as described in A. C, stemloop structures of AHSP IRE wt (left) and AHSP IRE mut (right) with nucleotide changes underlined. D, the effects of DFO or FAC on IRP binding to AHSP wt mRNA in HeLa cells, performed as described in A. OCTOBER 3, 2008 • VOLUME 283 • NUMBER 40 real-time RT-PCR. Transcript levels were normalized to those of GAPDH. As internal controls, ␣-globin (HBA-A1) and MYC mRNAs were examined as examples of constitutively stable and unstable transcripts, respectively (63). In K562 cells treated with DFO, AHSP mRNA was relatively stable, with a half-life greater than 8 h (Fig. 4A, upper left panel). In contrast, AHSP mRNA was more rapidly degraded with a half-life of ϳ4 h in FAC-treated cells. As expected, endogenous TFR1 mRNA was destabilized by iron (13,60). In addition, HBA-A1 and MYC RNAs which do not contain IREs were unaffected by either drug. Thus, endogenous AHSP mRNA is destabilized by iron in human hematopoietic cells, consistent with the presence of a functional 3Ј-UTR IRE.

Iron Regulation of AHSP mRNA Stability
Somewhat surprisingly, DFO treatment of K562 cells decreased the steady state level of mature AHSP mRNA measured in whole cell extracts by ϳ30% (Fig. 4B, first panel), despite concurrent mRNA stabilization, as shown in Fig. 4A. To reconcile these findings, we examined AHSP mRNA in nuclear and cytoplasmic fractions of DFO-and FAC-treated cells. In the  nucleus, FAC treatment increased the level of AHSP primary transcripts, while DFO treatment produced an opposite effect (Fig. 4B, second panel). Together these results indicate that in K562 cells, iron increases AHSP gene transcription or nuclear RNA stability and concurrently destabilizes the mature mRNA. The net effect is little change in mature cytoplasmic mRNA (Fig. 4B, third panel). Hence, iron has multiple opposing effects on AHSP gene expression in K562 cells. In vivo, the relative importance of each individual effect during erythropoiesis is likely dependent on developmental stage and cellular milieu (see "Discussion").
The AHSP IRE-like Stem-Loop Mediates Iron-regulated mRNA Stability-The previous experiment demonstrated that iron levels modulate AHSP mRNA stability. To test whether this mode of regulation occurs via the 3Ј-UTR IRElike structure, we measured the effects of iron on AHSP mut RNA, in which the stem-loop sequence is mutated to impair IRP binding (Fig. 2, A and C). We expressed AHSP wt and AHSP mut mRNAs in COS cells treated with hemin (hem) as a source of iron or DFO and measured the rate of transcript decay after transcriptional arrest induced by actinomycin-D (Fig. 5). Similar to what we observed for endogenous transcripts in hematopoietic cells (Fig. 4, above), AHSP wt mRNA was stable under low-iron conditions and degradation was accelerated by iron. In contrast, the mutant mRNA was constitutively unstable, independent of added iron. The most likely explanation for these findings is that when the IRE structure is disrupted, IRPs can no longer bind the 3Ј-UTR of AHSP mut mRNA. This renders the transcript more susceptible to endonucleolytic cleavage, even when iron concentrations are low. Thus, the effects of iron on AHSP mRNA stability maps to its IRE-like stem-loop.
AHSP mRNA Stability Is Directly Regulated by IRPs-To further study the role of IRPs in regulating AHSP mRNA stability, we used HeLa cell lines that stably express small interfering RNAs (siRNAs) against IRP1 (IRP1 KD), IRP2 (IRP2 KD), or both (IRP1/2 KD) (32). We expressed AHSP mRNA in wt and IRP KD cells and measured transcript levels before and 8 h after actinomycin-D treatment (Fig. 6). After blocking transcription AHSP mRNA was reduced by ϳ40% in wt cells. In contrast, loss of IRP1 or IRP2 decreased AHSP mRNA by ϳ50 and 70%, respectively (Fig. 6A). The effects of IRP reduction were additive, as AHSP mRNA was reduced by about 80% in IRP 1/2 KD cells. Hence, both IRP1 and IRP2 can stabilize AHSP mRNA in cells.
Three controls show that this experimental approach replicates important aspects of IRE regulation. First, endogenous TFR1 mRNA was preferentially destabilized in IRP2 KD and IRP1/2 KD cells, as previously reported (32) (Fig. 6B). Second, FTH1 mRNA levels were not affected by lack of IRPs, because IRPs regulate translation but not stability of this mRNA (64) (data not shown). Third, MYC mRNA levels rapidly declined after actinomycin-D treatment in either cell line, consistent with IRP-independent instability of this transcript (data not shown). Together, these results demonstrate that AHSP mRNA stability is regulated by iron-dependent interactions between IRP1 and IRP2 and a noncanonical 3Ј-UTR IRE-like stem loop.

DISCUSSION
Until recently, IREs were believed to be associated with relatively strict consensus sequences present in a few genes that are  directly involved in iron metabolism, such as ALAS2, TFR, and FTH1 (46,55,56). However, several lines of study indicate that IREs and related elements are more diverse in structure and distribution. Unbiased in vitro selection methods show that non-canonical IRE sequences form stem-loops that can bind IRPs (44,47). In addition, numerous mRNAs contain IREs that deviate significantly from consensus sequences (15, 48, 50 -53). The crystal structure of an IRP1-FTH1 IRE complex indicates how IRPs may recognize different sets of IRE sequences through alternate subsets of bonding amino acids (46). The 3Ј-UTR of AHSP mRNA contains an IRE-like stem-loop that deviates from consensus and recognizes IRPs poorly in vitro, yet associates with IRP1 and IRP2 in vivo to regulate AHSP mRNA stability according to iron concentration. These data show that IRPs can regulate mRNA expression through non-canonical IREs.
In cell extracts, co-immunoprecipitation of IRP1 and IRP2 with AHSP mRNA is mediated by an IRE-like stem-loop structure. However, in EMSA, a short (32 bp) oligoribonucleotide probe containing this IRE-like element bound poorly to IRP1 and no binding to IRP2 was detected. Several features of the AHSP IRE-like stem-loop are predicted to impair its interactions with IRPs. These include the shifted position of the CAGUGH loop consensus sequence and the presence of A at the stem bulge (see Fig. 1) (45,46). It is possible that additional factors not present in the EMSA reactions strengthen the binding of IRPs to the AHSP IRElike stem-loop in vivo. For example, extended sequences in the AHSP mRNA that were not present in the EMSA probe may optimize folding of an IRE-like structure (41). In addition, accessory cellular proteins absent in purified IRP preparations may enhance AHSP IRE-like stem-loop:IRP interactions. It is also possible that IRPs associate with the AHSP IRE-like stem-loop indirectly as part of higher order protein complexes. This may be true in particular for IRP2, where EMSA failed to detect any interaction with the AHSP IRElike stem-loop, despite our observations that AHSP mRNA co-immunoprecipitates with IRP2 in cell extracts and that the message is destabilized by IRP2 knockdown. Regardless of the underlying mechanisms, the RNA immunoprecipitation assay may provide a more sensitive context for detecting biologically relevant IRP associations with atypical IREs. One important general implication of our findings is that IRE-like stem-loop structures that bind IRPs weakly (or not at all) in vitro can still associate with IRPs in cells to regulate mRNA expression.
It is interesting to consider the AHSP IRE-like structure in context of the general evolution of the IRE regulatory module (65). Some IREs, such as those present in ferritin genes and TRF1, occur in most metazoan species. Others, such as those found in the divalent metal transporter 1 (DMT1) gene, are largely restricted to mammals. The IRE-like stem-loop involved in modulating the response of AHSP to iron has an even more restricted phylogenetic distribution. It is only found in simian primates and thus appears to be an adaptation in this lineage. If so, the primate IRE-like element was generated from sequences that include a polyadenylation signal. Thus, these sequences in primates are under constraint for two functions and that may account for some of the non-canonical features of this IRE-like element. For example, it is possible that the constraint to maintain an AAUAAA polyadenylation signal outweighs the constraint for a C in the bulge in the stem-loop of the IRE-like structure.
The AHSP gene has been identified to date only in mammals, including eutherians and marsupials. No evidence for a homolog to AHSP has been observed in other tetrapods such as chickens, lizards or amphibians. This suggests that the AHSP protein has appeared relatively recently in evolution as a mechanism to fine-tune hemoglobin synthesis and homeostasis (66). Acquisition of the IRE-like structure in primates may represent a new level of control through which the stability of a hemoglobin subunit is modulated according to iron concentration within developing erythroid cells.
Iron destabilized endogenous AHSP mRNA in K562 human erythroleukemia cells, most likely by causing release of IRPs from the IRE-like structure in the 3Ј-UTR. However, iron simultaneously increased levels of nascent primary transcripts (Fig. 5). Our preliminary unpublished data indicate that the transcriptional repressor Bach1 binds the AHSP gene (MJW and COS). Thus, iron may activate transcription by inducing the synthesis of heme, which binds Bach1 causing its release from DNA and replacement by a positively acting transcription factor (17). These opposing effects reflect multilayered mechanisms of gene regulation for AHSP and are consistent with findings that the effects of iron on expression of IRE-regulated genes vary according to cell type and developmental stage (15,67). In vivo, the net effects of IRP interactions with the AHSP mRNA IRE-like structure may predominate during later stages of erythropoiesis as transcription subsides.
Several biological implications for the AHSP IRE-like stemloop are possible based on known functions of the protein. For example, in iron deficiency apo globins accumulate due to reduced heme synthesis. Apo ␣-globin is particularly unstable, as evidenced by its reduced secondary structure in solution (68). Moreover, ␣-globin precipitation is relatively increased in iron deficiency (69). AHSP binds newly synthesized apo ␣-globin to enhance its folding and solubility (22). Induction of AHSP production via IRP interactions at the 3Ј-UTR may provide a compensatory mechanism to stabilize excess apo ␣-globin during iron deficiency. Another function of AHSP is to bind and detoxify excess free ␣Hb, which is present at high levels in ␤-thalassemia (24,25). This disorder is frequently accompanied by iron overload from increased dietary absorption and red blood cell transfusions. Free iron causes tissue damage by catalyzing the production of ROS. In addition, the current study indicates that iron overload promotes degradation of AHSP mRNA by reducing IRP/IRE interactions. In this case, consequent reduction of AHSP protein could further destabilize accumulated ␣Hb, enhancing its toxicity in erythroid cells of individuals with ␤-thalassemia.