Destabilization of human alpha-globin mRNA by translation anti-termination is controlled during erythroid differentiation and is paralleled by phased shortening of the poly(A) tail.

The extraordinary stability of globin mRNAs permits their accumulation to over 95% of total cellular mRNA during erythroid differentiation. The stability of human α-globin mRNA correlates with assembly of a sequence-specific ribonucleoprotein complex at its 3′-untranslated region. A naturally occurring anti-termination mutation, Constant Spring (CS), which permits ribosomes to enter the 3′-untranslated region of the α-globin mRNA, results in accelerated mRNA decay. To study the mechanism of this destabilization in vivo, we established transgenic mouse lines carrying the human αCS gene. Relative to wild-type human α-globin mRNA (αwt), αCS mRNA is destabilized in marrow erythroid cells. The poly(A) tails of both the αCS and αwt mRNAs show a periodicity of 20-25 nucleotides consistent with phased binding of poly(A) binding proteins. However, the mean size of poly(A) tails of the unstable αCS mRNA is significantly shorter than that of the αwt mRNA. Unexpectedly, the αwt and αCS mRNAs are of equal stability in peripheral reticulocytes, where their respective poly(A) tails shorten coordinately. These findings demonstrate a characteristic organization of the poly(A) tail on α-globin mRNA which is maintained during normal and accelerated decay, a correlation between poly(A) metabolism and anti-termination-mediated accelerated mRNA turnover, and a switch in the mechanism of mRNA decay during erythroid terminal differentiation.

The phenotype and function of a cell are defined by the spectrum of its cytoplasmic mRNAs. The level of a specific mRNA is determined by the balance between its rates of synthesis and degradation. Recent studies in a variety of experimental systems highlight the central role of mRNA stability in the control of gene expression. The mechanisms that control this property appear complex and encompass distinct pathways for different subgroups of mRNAs (reviews in Refs. [1][2][3]. The half-lives of mRNAs in higher eucaryotes can differ by over 1000-fold, with a range of minutes to days (3). For example, the half-lives (t1 ⁄2 ) of c-myc and c-fos mRNAs are approximately 10 -15 min (4) while the t1 ⁄2 of globin mRNA is more than 24 h (5). In some instances, mRNA stability may be a dynamic property that changes dramatically in response to developmen-tal, environmental, or metabolic signals.
The stability of an mRNA is a function of its sequence and structure and may reflect interaction(s) of cis elements with trans-acting factors. For example, the stability of the transferrin receptor mRNA is controlled by intracellular concentration of iron; at low iron concentrations the transferrin receptor mRNA is quite stable, while a rise in cytosolic iron triggers its accelerated decay (6). This control is mediated by interaction of a specific series of secondary structures in the transferrin receptor mRNA 3Ј-untranslated region (3Ј-UTR) 1 with an iron binding protein (6). In contrast to the transferrin receptor mRNA, the structural determinants that underlie the stability of most mRNAs are poorly defined.
mRNAs are degraded via a number of interrelated pathways, several of which have been defined in yeast (2). A common degradation pathway is initiated by shortening of the poly(A) tail, followed by removal of the 7-methylguanosine cap structure. The deadenylated and decapped mRNA is then degraded by a 5Ј 3 3Ј exonuclease (7,8). This pathway implies that there is communication between the 5Ј-and 3Ј-termini of the mRNA; the mechanism(s) involved in this interaction is not known (9). In higher eucaryotes, deadenylation also frequently accompanies degradation of mRNA, although the mechanism of degradation subsequent to deadenylation remains to be established (10). Two other less common degradation pathways are independent of prior deadenylation: direct 5Ј-decapping triggered by nonsense mutations (11) and site-specific endonuclease cleavage in the 3Ј-UTR (12,13). The general importance of these pathways, how each is controlled, and how they might be interrelated are questions that are presently under study.
The globin genes provide a unique model system with which to study mechanisms responsible for mRNA stability. During terminal erythroid differentiation, globin mRNAs accumulate to over 95% of total cellular mRNA (14). This remarkable enrichment of a single group of mRNAs relies on their unusually long half-lives as well as selective degradation of nonglobin mRNAs (5,(15)(16)(17). The importance of mRNA stability to globin gene expression is clearly demonstrated by the phenotype resulting from the loss of this property. For example, a UAA 3 CAA anti-termination mutation in the ␣-globin gene (Constant Spring, ␣ CS ), results in mRNA destabilization and consequent loss of over 95% expression from the affected locus (18).
Recent studies have demonstrated that ␣ CS mRNA is destabilized by the physical entry of ribosomes into its 3Ј-UTR, which is permitted by the CS anti-termination mutation. This finding suggested that ␣-globin mRNA contains a stability determinant in this region (19). Subsequent studies confirmed this hypothesis and identified the critical cis-determinant as a pyrimidine-rich motif in the 3Ј-UTR that functions in an erythroid-specific manner (20). The stability motif serves as an assembly site for a sequence-specific mRNA-protein (mRNP) complex (␣-complex, 21). The presence of this complex correlates with ␣-globin mRNA stability; mutations that block assembly of the ␣-complex in vitro also destabilize the mRNA in vivo (20,21). Although structural studies have identified one component of the ␣-complex (22), its overall composition remains to be established.
Although the importance of the ␣-complex to ␣-globin mRNA stability is apparent, the mechanism(s) of its stabilizing function remains to be defined. One approach to this problem is to follow the pathway of globin mRNA degradation triggered by a mutation that interferes with normal stabilizing functions. In the present report we describe a fully physiologic in vivo model system to facilitate such an approach. Transgenic mice that express readily detectable levels of ␣ CS mRNA were established. The stability and structure of ␣ CS mRNA in these mice can be characterized and compared with wild-type human ␣-globin mRNA (h␣ or ␣ wt ) expressed in a parallel set of transgenic lines. Using this model, we have focused our attention on the structure and dynamics of the poly(A) tail of h␣-globin mRNA in early erythroblasts in the bone marrow and in mature reticulocytes in the peripheral blood. Our findings demonstrate a phased organization of the poly(A) tail, which is independent of mRNA stability and maintained during differentiation. The data also reveal a substantial effect of the ␣ CS mutation on its poly(A) tail size. Finally, we present evidence for a switch in the mechanism of h␣-globin mRNA decay during erythroid differentiation.

EXPERIMENTAL PROCEDURES
␣ CS Gene Construction-The ␣ CS mutation was introduced by spliceoverlap extension/polymerase chain reaction (Ref. 20 and included references) into the wild-type human ␣2-globin gene carried as a 1.4-kb PstI insert in the pSP72H␣2 plasmid (23). Polymerase chain reaction amplifications were carried out using Vent polymerase (New England BioLabs) under the following conditions: 1 cycle (95°C for 5 min, 57°C for 15 s, and 73°C for 25 s), 28 cycles (92°C for 1 min, 57°C for 15 s, and 73°C for 25 s), 1 cycle (73°C for 3 min). Primer name indicates its position relative to the transcriptional start site of the human ␣2-globin gene. Primer SP72U hybridizes in the vector immediately 3Ј to the ␣-globin gene. Reaction 1 contained 10 ng of pSP72H␣2 template, 100 pmol each of primers 589 ( 5Ј CTGCACAGCTCCTAAGCCA 3Ј ) and 717Reverse ( 5Ј GAGGCTTCCAGCTTGACGGTA 3Ј ), and yielded the expected 148-bp product. Reaction 2 contained 10 ng of template and 100 pmol each of primers 717 ( 5Ј TACCGTCAAGCTGGAGCCTC 3Ј ) and SP72U ( 5Ј TAATACGACTCACTATAGGGA 3Ј ) and yielded the expected 240-bp product. 1 l each of reactions 1 and 2 were combined, heat-denatured, reannealed, and amplified with primers 589 and SP72U as above except that the extension time was increased to 30 s. The expected 368-bp splice-overlap extension product was digested with SphI and BstEII and ligated into the SphI/BstEII site of pSP72H␣2. Aliquots of the ligation mixture were electroporated into competent DH5␣ cells, and ampicillin-resistant colonies were screened by restriction analysis of plasmid DNA. Candidate clones were purified and sequenced to ensure sequence fidelity.
Generation of Transgenic Mice Expressing the ␣ CS Gene-All transgenic mice were generated by the Transgenic Mouse Core Facility at the University of Pennsylvania School of Medicine. The ␣ CS gene was released from its host pSP72 vector by EcoRI digest and introduced into the EcoRI site of the LCR/pSP72 vector (23). The 8.0-kb LCR/␣ CS fragment released by EcoRV/SalI digestion was purified from a 0.6% agarose gel over an Elutip column according to standard protocol (Schleicher & Schuell). The purified DNA was taken up in buffer (5 ng/l) for injection into fertilized mouse oocytes (23). Transgenic ␣ CS mice were identified by dot blot analysis of tail DNA using a ␤-globin LCR HSII probe. Six transgenic mouse lines carrying the normal ␣-globin transgene have been previously described (23); four additional lines constructed in the same manner were added for this study. Transgene copy numbers for ␣ wt and ␣ CS lines were in the same range. Transgene copy number was established by Southern analysis of EcoRIdigested tail DNA from F1 pups. A human ␣-globin promoter probe (0.6-kb NcoI fragment) detects a 23-kb fragment in the human genome and a 1.5-kb h␣-globin transgene fragment. The amount of DNA loaded on each lane was normalized using a probe to the 3Ј-flanking region of the endogenous mouse -globin gene (2.3-kb BamHI/EcoRI fragment), which detects a 4.5-kb fragment of the mouse genome.
RNA Preparation-Adult transgenic mice were rendered anemic by three intraperitoneal injections of acetyl-2-phenylhydrazine (Sigma, 40 mg/g of body weight) over a 36-h period. Mice were decapitated 4.5 days after the initial injection, and blood was collected into heparinized (20 units/ml) phosphate-buffered saline (PBS). Femoral and tibial marrow was flushed into heparinized PBS using a 23-gauge needle. Total RNA was isolated from each tissue by lysis in 5 M guanidine thiocyanate solution (50 mM Tris, pH 7.4, 10 mM EDTA, 700 mM ␤-mercaptoethanol, 1% sarcosyl) and centrifugation through a 5.7 M cesium chloride cushion at 42,000 rpm for 16 h in a TL55 rotor (Beckman). The RNA pellet was resuspended in 10 mM Tris, pH 7.4, 1 mM EDTA, 1% SDS, extracted twice with phenol/chloroform/isoamyl alcohol, and ethanol-precipitated. RNA purity and concentration were determined by A 260/280 measurements.
In Vitro Culture of Peripheral Blood Reticulocytes-Heparinized blood from phenylhydrazine-treated mice was collected as described above, and cells were sedimented by centrifugation for 5 min at 1000 ϫ g, washed in ice-cold PBS, resuspended in minimum essential medium ␣ supplemented with 10% fetal bovine serum, and incubated at 37°C in 5% CO 2 . During each incubation, cell viability was verified by minimal release of hemoglobin into the media. Cell aliquots harvested at 0, 24, and 48 h were washed twice in PBS, and total RNA was isolated as described (24). To determine the rate of decay of mouse ␣-globin mRNA in cultured reticulocytes, cell number for each aliquot was determined, and RNA was isolated in the presence of a trace amount of [ 35 S]UTPlabeled glyceraldehyde-3-phosphate dehydrogenase mRNA (specific activity, 700,000 cpm/g) to control for RNA extraction efficiency. The amount of m␣-globin mRNA was determined per 5 ϫ 10 5 cells by RNase protection assay.
RNase Protection Assay-Antisense RNA probes complementary to the first exon-encoded region of the human ␣-globin mRNA and the second exon-encoded region of the mouse ␣-globin mRNA were synthesized from plasmids pGEM3␣H and pGEM3␣M (23). Human ␣-globin (h␣-globin) probe was synthesized at 10 times the specific activity of the m␣-globin probe by adjusting the amount of [␣-32 P]CTP in the transcription mix (Maxiscript transcription kit, Ambion). Total RNA isolated from mouse bone marrow (500 ng of total RNA) or reticulocytes (100 ng of total RNA) was mixed with an excess of h␣-and m␣-globin RNA probes in 20 l of hybridization buffer (80% formamide, 40 mM PIPES, pH 6.4, 400 mM NaCl, 1 mM EDTA), denatured at 75°C for 15 min, and then hybridized at 52°C for 16 h. Excess probe and unhybridized RNA were then digested by addition of 4 g of RNase A and 200 ng of RNase T1 in a total of 200 l of buffer (300 mM NaCl, 10 mM Tris, pH 7.4, 5 mM EDTA) and incubated for an additional 20 min at room temperature. The reaction was terminated by the addition of 17 l of 1:4 mix of proteinase K (10 mg/ml) and SDS (10%), phenol-extracted, ethanol-precipitated, and resolved on an 8% acrylamide, 8 M urea gel. Signal intensity was quantified on a PhosphorImager with ImageQuant software (Molecular Dynamics).

RESULTS
Generation of Mice Expressing the Human ␣ CS Transgene-In humans, the presence of the CS anti-termination mutation in the ␣-globin mRNA results in accelerated message decay. To establish an in vivo model system in which to study the mechanism(s) of this accelerated decay, transgenic mouse lines that expressed the human ␣ CS gene were generated. An ␣ CS gene was constructed by introducing the CS mutation into the ␣ wt globin gene by splice-overlap extension (see "Experimental Procedures"). The ␣ CS gene was then ligated to a DNA fragment containing all four hypersensitive sites of the human ␤-globin locus control region (27) to ensure erythroid-specific high level transgene expression (23). Five ␣ CS transgenic lines were established and the transgenes mapped in F1 progeny from each line to certify integrity and to establish copy number (Fig. 1B). For the study of ␣ wt mRNA stability, mice from previously described transgenic lines were utilized. The ␣ CS and ␣ wt transgenes are identical except for the single base substitution at the translation termination codon (Fig. 1A), and the transgene copy numbers are in the same range (23).
The CS Mutation Destabilizes ␣-Globin mRNA in Transgenic Mice-The first goal was to demonstrate that the CS mutation destabilizes the h␣-globin mRNA in mouse erythroid tissue. The relative stabilities of ␣ wt and ␣ CS mRNAs were compared using an RNase protection assay that takes advantage of the natural transcriptional silencing that occurs in bone marrow during erythroid differentiation. The ratio of transgene h␣to endogenous m␣-globin mRNA in the bone marrow and in the peripheral blood reticulocytes determines two points on an mRNA decay curve (normalized stability, ((h␣/m␣) peripheral blood / (h␣/m␣) bone marrow )). A representative study is shown in Fig. 2A, and the results from the 10 ␣ wt globin lines and the five ␣ CS lines are shown in Fig. 2B. The average value of 0.85 for ␣ wt mRNA indicates that its stability is equivalent (within experimental error) to that of the endogenous m␣-globin mRNA. The mean value of 0.2 for ␣ CS mRNA indicates that it is highly unstable relative to both m␣and h␣ wt -globin mRNAs. The 3-fold spread in the normalized stabilities among the ␣ wt lines does not adversely affect the high significance of the difference between ␣ wt and ␣ CS stability (see "Discussion"). Therefore, this transgenic model system accurately reproduces the destabilization of the mutant ␣ CS mRNA observed in humans.
The  2. The CS mutation destabilizes ␣-globin mRNA in transgenic mice. A, analysis of h␣-globin mRNA stability. Total mRNA was isolated from bone marrow (BM) and peripheral blood (PB) of phenylhydrazine-treated mice transgenic for the human ␣ wt or ␣ CS genes, and the levels of h␣and m␣-globin mRNA were determined by quantitative RNase protection assay. The position and identity of each protected fragment is indicated (h␣ and m␣). The specific lines used in this representative study are shown above the respective lanes. B, normalized stabilities of the ␣ wt and ␣ CS mRNAs. Each symbol represents the mean value for an independent line as determined from analysis of two or more mice. Solid bars represent the average normalized stabilities of the transgenic mRNAs, and the dashed line indicates the stability of endogenous m␣-globin mRNA (defined as 1.0). mRNA when compared with the stable ␣ wt mRNA. The size of the human ␣-globin mRNA poly(A) tail was determined after site-specific RNase H cleavage of total RNA from bone marrow or peripheral blood (Fig. 3A). The digested fragments were separated by denaturing polyacrylamide gel electrophoresis and the 3Ј-terminal fragment was visualized by Northern blot analysis using a h␣-globin mRNA 3Ј-UTR probe. Typical profiles for ␣ wt and ␣ CS globin poly(A) tail are shown in Fig. 3B. In a parallel reaction, oligo(dT) was also added to the RNase H digestion, in order to generate a fully deadenylated 3Ј-terminal fragment (lane dT, Fig. 3B). The densitometric profiles of single lanes on the autoradiograph reflect the spectrum of poly(A) tail lengths in each sample (Fig. 3C).
Analysis of ␣ wt mRNA revealed that poly(A) tail lengths have a periodicity representing increments of 20 -25 nucleotides (Fig. 3B). In the bone marrow, the ␣ wt mRNA contains a dominant A 60 peak flanked by peaks of lower intensity that correspond to A 85 and A 40 (Fig. 3B). By comparison, the poly(A) tail of the h␣-globin mRNA in peripheral blood reticulocytes is partitioned into four peaks with an overall shift to a smaller mean size; although A 60 is still the dominant peak, A 85 decreases in intensity, and a significant A 20 peak appears. Of note, we do not detect any evidence of fully deadenylated mRNA. These analyses of the ␣ wt globin mRNA indicate that the poly(A) tail is organized in a modular fashion and that the mean poly(A) tail size decreases as bone marrow erythroid cells mature into peripheral reticulocytes. Therefore, ␣-globin mRNA poly(A) tail phasing is maintained despite its age-related shortening in differentiating erythroid cells.
The CS Mutation Is Linked to an Accelerated and Phased Shortening of Poly(A) Tail-The size distribution of the ␣ CS mRNA poly(A) tail was determined using the RNase H assay. The fully deadenylated ␣ CS fragment migrated at the same position as the ␣ wt fragment (data not shown), indicating that the 3Ј end of both mRNAs were processed identically. In bone marrow, the ␣ CS mRNA poly(A) tail was significantly shorter than the ␣ wt and was present in two major peaks of A 60 and A 40 (Fig. 3, B and C). In reticulocytes, the ␣ CS A 40 peak became more prominent. In contrast to the ␣ wt globin mRNA, there was no significant A 20 signal. Thus, like the ␣ wt mRNA, the ␣ CS mRNA shows age-related, phased shortening of its poly(A) tail. However, the distribution of the poly(A) tail size on the ␣ CS mRNA is significantly different than on the ␣ wt mRNA in both the bone marrow and in peripheral reticulocytes.
␣ wt and ␣ CS mRNAs Are of Equal Stability in Late Stage Reticulocytes-Under normal conditions, all erythroid mRNAs undergo final degradation over a 2-3-day period in peripheral blood reticulocytes. This terminal event in erythroid cell maturation may involve stability mechanisms and determinants distinct from those that favor rapid accumulation of globin mRNAs during the earlier phases of erythroid maturation. To directly assess this possibility, we followed mRNA decay during a 48-h ex vivo incubation of peripheral blood reticulocytes. Reticulocyte viability appeared fully maintained during incubation. Endogenous m␣-globin mRNA levels fell by 70% during the 48-h incubation (Fig. 4A), indicating that this ex vivo incubation reproduces the reticulocyte mRNA clearance observed in vivo (28). The levels of transgenic ␣ wt and ␣ CS mRNAs were measured relative to endogenous mouse ␣-globin mRNA in ex vivo incubated reticulocytes over a 48-h period. A typical RNase protection assay is shown in Fig. 4B, and average stabilities from studies of four ␣ wt and three ␣ CS lines are plotted in Fig. 4C. Consistent with the accelerated decay of the ␣ CS mRNA in bone marrow erythroid cells, the initial levels of ␣ CS mRNA were considerably lower than ␣ wt mRNA levels. Surprisingly, however, the stabilities of the ␣ CS and the ␣ wt mRNAs were equivalent. Thus the ␣ CS mRNA is unstable relative to the ␣ wt mRNA in the bone marrow, but the ␣ wt and ␣ CS mRNAs decay at comparable rates in peripheral reticulocytes.

Poly(A) Tails of ␣ wt and ␣ CS Globin mRNAs Shorten Coordinately in Peripheral Reticulocytes-The kinetics of poly(A) tail
shortening were determined in the ex vivo incubated reticulocytes (Fig. 5). The ␣ wt globin mRNA shortened over the 48-h incubation from an initial multipeak distribution to a single intense peak of A 20 . At the beginning of the incubation period, ␣ CS mRNA was distributed into two peaks, A 60 and A 40 . Like the ␣ wt mRNA poly(A) tails, the ␣ CS poly(A) tails that remain after an additional 48 h were shortened to a single A 20 peak. Thus for both mRNAs there is a progressive shortening of the poly(A) tail to a minimal size of A 20 . Whether deadenylation precedes or parallels mRNA degradation during late reticulocyte differentiation is unknown. The appearance of the A 20 peak in the ␣ CS mRNA poly(A) tail, in coordination with a equivalent rate of decay for both ␣ wt and ␣ CS mRNA, suggests a switch in the degradation pathway of globin mRNA in the late stages of reticulocyte differentiation.
The Level of an Essential Subunit of the ␣-Globin Stabilizing Complex Decreases During Erythroid Terminal Differentiation-␣ CS is degraded faster than ␣ wt mRNA in marrow eryth- roid cells. The same accelerated decay of ␣ CS mRNA has been noted in cultured MEL cells, which represent an early stage in terminal erythroid differentiation (19). By comparison, both mRNAs are equally stable in peripheral reticulocytes. The stability of ␣ wt mRNA in early erythroblasts (MEL cells) is dependent upon formation of an RNP complex on its 3Ј-UTR (21). The convergence of ␣ wt and ␣ CS mRNA stabilities during terminal erythroid differentiation might therefore reflect a decrease in the ability of mature cells to assemble the ␣-complex and selectively protect the ␣ wt mRNA. To test this hypothesis, we compared the levels of ␣CP, a 39-kDa RNA-binding protein that is an essential subunit of the ␣-complex (22), in MEL cells and in peripheral reticulocytes from phenylhydrazine-treated mice (Fig. 6, A and B). The limiting amount of mouse bone marrow and the necessity to separate erythroid from nonerythroid cells prior to assay (␣CP is expressed ubiquituously (21)) precluded use of bone marrow cells as a source of early erythroid cells for this purpose. The signal observed by Western analysis in 450 g of reticulocyte proteins is equivalent to that of 15 g of MEL cell proteins. A slightly slower migration for ␣CP is observed in reticulocyte extract. Whether this is a result of a posttranslational modification or an aberrant migration pattern due to sample overloading is unknown. The ␣CP in the reticulocyte extracts retains the ability to bind to polyribocytidine (data not shown). The capacity of protein extracts from reticulocytes and MEL cells to assemble the ␣-complex was tested in parallel (Fig. 6C). The reticulocyte extract can assemble a normal ␣-complex, as judged by its comigration with the ␣-complex from MEL extract and its poly(C) sensitivity. However, the efficiency of complex formation is markedly reduced in comparison with the MEL cell extract. Thus, levels of ␣CP and ␣-complex decrease sharply in erythroid cells undergoing terminal differentiation.

DISCUSSION
The ␣-globin gene encodes an mRNA whose extraordinary stability has been demonstrated in multiple experimental systems (5,14,15,17,20). This stability is dependent upon the assembly of an mRNP complex on its 3Ј-UTR (21). Although the cis and trans elements of the ␣-complex are under study, the mechanisms by which the mRNA is stabilized in the developing erythroblast and subsequently cleared from the terminally differentiating peripheral reticulocyte are poorly understood. The generation of transgenic mice that express the h␣-globin (␣ wt ) mRNA or the mutant ␣ CS permits detailed study of these pathways in vivo.
The ␣ CS Transgenic Mouse Model-Of the one hundred or more mutations of the ␣-globin gene documented in ␣-thalassemic individuals, the CS defect is unique in its direct destabilization of the mRNA (29,30). The accelerated decay of ␣ CS mRNA in mice appears to recapitulate its observed instability FIG. 4. ␣ wt and ␣ CS mRNA decay at equivalent rates in ex vivo cultured reticulocytes. Peripheral blood was collected from phenylhydrazine-treated mice and cultured ex vivo for 48 h. Total RNA was isolated from aliquots harvested at 0, 24, and 48 h. A, murine ␣-globin mRNA content per cell aliquot as quantified by RNase protection assay. Values for 24 and 48 h incubations were expressed as a fraction of the t 0 value. The mean and standard error is derived from the study of reticulocytes from four nontransgenic mice. B, RNase protection assay of ex vivo incubated reticulocytes from transgenic mice. RNA isolated from ␣ wt and ␣ CS reticulocytes was analyzed as described in Fig. 2. The positions of m␣and h␣-protected fragments are noted. The band migrating between m␣ and h␣ bands in ␣ CS lanes, also present in the reticulocyte control (right lane), is due to protection of an unidentified endogenous mouse RNA. C, signals from the gel (B) were quantitated on PhosphorImager. Levels of transgenic h␣-globin mRNA (␣ wt or ␣ CS ) were estimated relative to endogenous m␣-globin mRNA; values for 24 and 48 h incubation were expressed as a fraction of the t 0 value. The data shown are averaged from studies of four ␣ wt transgenic lines and three ␣ CS lines.

FIG. 5. Poly(A) tails of ␣ wt and ␣ CS -globin mRNAs shorten coordinately in ex vivo cultured reticulocytes.
A, poly(A) tail analysis of mRNA from ␣ wt and ␣ CS transgenic mice reticulocytes incubated ex vivo for 48 h. Analysis was performed as described in Fig. 3. B, scanning densitometry of single lanes from gel (A). The position of each peak is indicated, as is the position of the deadenylated 3Ј-terminal fragment. The direction of electrophoretic migration is noted below the scan.
in humans. The two-point decay curve that was determined in vivo in transgenic mice demonstrated a 4 -5-fold lower stability for ␣ CS mRNA than for ␣ wt -globin mRNA (Fig. 2). Cells from the bone marrow compartment, which include transcriptionally active erythroblasts as well as transcriptionally inactive normoblasts and reticulocytes, were compared in this analysis with a population of purely posttranscriptional, nonnucleated reticulocytes from the peripheral circulation. The marked fall in the ␣ CS mRNA levels between these two compartments parallels observations in affected humans. In the bone marrow of individuals heterozygous for the ␣ CS mutation, ␣ CS and ␣ wt mRNA levels are equivalent, whereas in peripheral reticulocytes ␣ CS mRNA is nearly undetectable despite the continued presence of ␣ wt mRNA (29).
The measurement of mRNA stability in the intact mouse as described in this report is noteworthy because the environment is entirely physiologic. Our approach avoids the use of transcriptional inhibitors such as actinomycin D, which can have a severe impact on cellular physiology and has been shown, in some cases, to paradoxically stabilize mRNA (31). The intraassay variability in our study (multiple assays on an individual sample) is small, as is the variability among mice from the same line (12%, data not shown). However, we observed a 3-fold spread in the stability of ␣ wt mRNA among individual transgenic lines (Fig. 2B). This variability does not appear to relate to transgene copy number, age, or sex of the mice. Therefore, although the assay itself is highly reliable, an additional biologic variable that we do not yet understand contributes to the stability of h␣-globin mRNA in the transgenic setting. It will be of great interest to identify this variable(s) in future studies. The ␣ CS transgenic lines do not show such a variability in their normalized stability. Despite the variability between different ␣ wt transgenic lines, the overall difference between the stability of ␣ wt and ␣ CS mRNA remains highly significant.

Poly(A) Tails of ␣-Globin mRNA Display a Discontinuous Size Distribution That May Reflect a Phased Array of Poly(A)
Binding Proteins-Analysis of the ␣ wt mRNA revealed that the size distribution of its poly(A) tails is not continuous but rather peaks with a periodicity of 20 -25 nucleotides. This spacing suggests a role for the poly(A) binding protein (PABP) in the protection and/or organization of the poly(A) tail. Mammalian PABP is a 70-kDa protein that is involved in mRNA stability and translation (9,(32)(33)(34). PABP monomers bind poly(A) with a footprint of 20 -25 residues (35,36). The observed poly(A) tail profile for the ␣-globin mRNA (Fig. 3) indicates than the PABPs are "anchored" at fixed positions relative to the mRNA, i.e. they are phased with regard to the terminus of the 3Ј-UTR.
The demonstration of multiple peaks of poly(A) tail size is scarce in the literature. Previous studies on globin mRNA of other species (37,38) have shown a similar pattern. One other report shows discrete poly(A) tail lengths for long-lived ribulose-bisphosphate carboxylase/oxygenase small subunit 2 mRNA in Chlamydomonas (34). Poly(A) tail phasing may be important for maintaining stability of long-lived mRNAs. Using an in vivo system, we have demonstrated that the poly(A) phasing pattern is dynamic. Specifically, we show that the 20 -25 nucleotide spacing in poly(A) tail size was maintained during erythroid differentiation despite the overall poly(A) tail shortening of the human ␣-globin. This observation suggests that PABP associates with the poly(A) tail throughout terminal erythroid differentiation, and its shortening by integral 20 -25 nucleotide units may reflect rapid degradation of deprotected segments as PABP monomers are released from the 3Ј-terminus.
Despite the accelerated shortening of its poly(A) tail, phasing at intervals of 20 -25 nucleotides is maintained for the ␣ CS mRNA. Because accelerated decay of the ␣ CS mRNA is likely to reflect the displacement of the ␣-complex by the translating ribosome (19), PABP phasing would not appear to be dependent on the presence of the ␣-complex. However, the ␣-complex might serve to stabilize the poly(A) tail, perhaps by direct physical interaction with PABP or by strengthening the interaction between PABP and the poly(A) tail.
Selective Degradation of ␣ CS mRNA Occurs Early but Not Late in Erythrocyte Differentiation-An unexpected observation in the present study was that the ␣ CS and ␣ wt mRNAs were degraded at equivalent rates in peripheral reticulocytes. This contrasts with the accelerated decay of ␣ CS mRNA compared with ␣ wt mRNA in the less differentiated erythroid cells both in vivo (this report) and in cultured mouse erythroblast cells (MEL,Ref. 20). The 60 h t1 ⁄2 of the ␣-globin mRNA in MEL cells (39) is shortened to 20 h by the CS mutation. 3 Since the rate of decay of both ␣ CS and ␣ wt mRNAs paralleled that of the endogenous mouse ␣-globin mRNA (Fig. 4), we estimated their t1 ⁄2 in reticulocytes to be approximately 24 h. This change in the relative stability of the ␣ CS and ␣ wt mRNA may reflect a difference between the mechanism that determines ␣ wt mRNA 3  stability in early and late erythroid differentiation. In early erythroid cells, globin mRNAs are preferentially stabilized to facilitate their rapid accumulation and permit synthesis of high levels of globin protein. In contrast, in reticulocytes all cellular organelles, membrane remnants, and RNAs, including ␣-globin mRNAs, are rapidly cleared to generate a maximally deformable and functional erythrocyte. The estimated half-lives of the globin mRNAs are consistent with the time frame required for this clearance. We therefore speculate that the mechanism necessary to selectively accumulate globin mRNA is operative only in early stages of differentiation and that a more general mechanism is responsible for clearance of RNA from peripheral reticulocytes. To support this hypothesis, we determined that the concentration of ␣CP, one component of the RNP complex necessary for stabilization of ␣ wt in early erythroblasts, is markedly diminished in circulating reticulocytes. In parallel, the ability of cell extracts to assemble the ␣-complex decreases from early erythroblasts to reticulocytes (Fig. 6).
A Model for the Regulation of ␣-Globin Stability in Differentiating Erythroid Cells-The stability of the ␣-globin mRNA is determined by the assembly of an mRNP complex on the 3Ј-UTR (20,21). Based upon our data, we propose a model for the regulation of ␣-globin mRNA stability in differentiating erythroid cells (Fig. 7). mRNA bound to the ␣-complex is stabilized in early erythroid cells. ␤-Globin mRNA and other mRNAs that are present at the reticulocyte stage may be stabilized by a related or different mechanism. Other mRNAs, whose decay is known to be accelerated during terminal erythroid differentiation (16,17), are hypothesized to lack a stabilizing complex. As modeled by the ␣ CS mRNA, accelerated decay is accompanied by accelerated shortening of the poly(A) tail. In the terminal stages of reticulocyte maturation, selectivity of mRNA decay is lost, and all mRNAs, including ␣ wt mRNA, are rapidly cleared from the cell. This may reflect either a switch in the degradation pathway or a change in the manner in which specific mRNAs are protected. For instance, terminal decay might be accelerated by the loss of one or more components of the ␣-complex as detailed above and demonstrated in Fig. 6. Our model suggests that the ␣-complex no longer protects ␣-globin mRNA in terminally differentiated reticulocytes. Whether this is a solely passive effect of the decrease in the components of the ␣-complex or reflects a more active control of ␣-complex assembly remains to be determined. FIG. 7. Regulation of human ␣-globin stability during erythroid cell maturation: a model. In early erythroblasts the globin genes are actively transcribed. At this stage the ␣-complex forms on the 3Ј-UTR of the h␣-globin mRNA, stabilizing the mRNA, and facilitating its accumulation to high concentrations. This stabilization may occur via an interaction between the ␣-complex and the poly(A)-PABP complex, protecting the poly(A) tail from degradation. Mutations (for example CS) that prevent formation or function of the ␣-complex reduce mRNA stability and expose the poly(A) tail to accelerated shortening. As the erythroblast differentiates and accumulates a critical level of globin mRNA, all transcription is silenced. At this point the levels of ␣CP fall and the ␣-complex dissociates from the 3Ј-UTR. Both the normal (wt) and mutant (CS) ␣-globin mRNAs now lack the ␣-complex and have equivalent stability. The poly(A) tails of both are now shortened at an equal rate with consequent clearance of the mRNA.