Structural determinants of human ζ-globin mRNA stability

Transgenic human ζ globin accumulates to high levels in mouse erythrocytes, consistent with the existence of cis-acting determinants that stabilize the cognate ζ-globin mRNA in transcriptionally silent progenitor cells [6, 7, 18]. Precedent analyses of both globin and non-globin mRNAs indicate that relevant stability determinants are commonly confined to the 3′ region of untranslated mRNA (3′UTR), which is not subject to disruption by translating ribosomes [27‐31]. We investigated the likely positioning of an mRNA stability element in the ζ-globin 3′UTR in a systematic manner, employing a linker-scanning strategy to identify cis-acting determinants of mRNA half-life that act in vivo in intact cells. Full-length ζ-globin genes were constructed to include either the 105-nt wild-type ζ-globin 3′UTR (ζWT); or identically sized variant ζ-globin 3′UTRs, each containing an 8-nt mutation at a different position between the UGA translational termination codon and the polyadenylate tail (Figure 1). Each gene was linked to a recombinant hybrid tetracycline response element (TRE) that promotes transcription in cultured cells expressing a hybrid tetracycline trans-activating protein (tTA), but is transcriptionally silent in the presence of tetracycline (tet) [32]. This approach permitted half-life values for wild-type and 3′UTR-variant ζ-globin mRNAs to be specified in vivo in tTA-expressing cells, using a transcriptional chase strategy that assesses the temporal reduction in the level of each variant ζ-globin mRNA relative to the steady-state level of a tet-indifferent control mRNA [33, 34].

To identify the positions of mRNA stability determinants within the ζ-globin 3′UTR, we conducted a screening assay that quantified the relative decay of wild-type and 3′UTR-variant ζ-globin mRNAs in vivo in intact cultures. Mammalian cultured-cell models for erythroid development transcribe a variety of embryonic, fetal, and adult globin mRNAs that can compete for post-transcriptional regulatory activities [18, 35, 36]; consequently, we conducted our studies in tTA-expressing HeLa cells (HeLa^tTA) that do not transcribe globin mRNAs and have previously been used to characterize post-transcriptional processes affecting both native and exogenous globin genes [16, 34]. We tested the relative decays of transiently expressed ζWT and 3′UTR-variant ζ-globin mRNAs, relative to that of a control β-globin mRNA transcribed from a co-transfected gene, using a two-point decay method of our design. One of the 3′UTR-variant mRNAs--containing a CCCCAGCC → agtgcaCa substitution at nts 57–64 (ζ57 mRNA)--reproducibly decayed over a 16-hour interval to a level that was approximately one-third that of ζWT mRNA (Figure 2A). The effect of the octanucleotide substitution was dependent upon its position rather than its content, as ζ-globin mRNAs containing a similar mutation elsewhere in the 3′UTR--including flanking nts 49–56 and 65–72 (ζ49 and ζ65 mRNAs, respectively)--decayed at the same rate as parental ζWT mRNA. The mutational effects were consistent over three or more replicate analyses, thus identifying and localizing a previously unknown post-transcriptional determinant of mRNA decay within the ζ-globin 3′UTR. For convenience, we now term the nt 57–64 region the ζ-globin m RNA r egulatory e lement (ZMRE).

We posited that the kinetics of ζ-globin mRNA decay might be similarly affected by a smaller defect in the nt 57–64 target region, according with precedent studies demonstrating the deleterious effects of site-specific single- or several-nt 3′UTR substitutions on the stabilities of both α- and β-globin mRNAs [16, 18, 38]. This hypothesis was validated by transcriptional chase analyses of transiently expressed ζ-globin mRNAs carrying tetranucleotide substitutions at either positions 1–4 or 5–8 of the ζ57 target region (ζ57.1-4 and ζ57.5-8 mRNAs, respectively; CCCCAGCC → agtgAGCC and CCCCAGCC → CCCCcaCa). While the relative decay of ζ57.5-8 mRNA was no different from ζWT mRNA (Figure 2B, left), the decay of ζ57.1-4 mRNA was significantly accelerated, and to a similar extent as that of parental ζ57 mRNA. Subsequent analyses demonstrated that smaller, dinucleotide substitutions to the nt 57–64 region did not affect mRNA decay, indicating that the ZMRE is functionally permissive for mutations of this magnitude (Figure 2B, right). Collectively, these screening data specify the importance of a site-specific tetranucleotide motif to the regulated accumulation of ζ-globin mRNA.

Results from two-point screening analyses indicated the likelihood that the ζ57 determinant is essential for normal accumulation of ζ-globin mRNA in cell cytoplasm. We validated this hypothesis by formally assessing t_1/2 values for wild-type and informative 3′UTR-variant ζ-globin mRNAs in vivo. Analyses were conducted in triplicate in transiently transfected HeLa^tTA cells using a conventional pulse-chase strategy that assesses the levels of test mRNAs at defined intervals, relative to the level of endogenous tet-indifferent β-actin mRNA [33]. Under these conditions, the ζ57 mRNA displayed a t_1/2 value that was approximately one-fourth that of ζWT mRNA (3.7 v 15.4 hr, respectively), while ζ49 and ζ65 mRNAs--containing ZMRE-flanking mutations--displayed relatively normal t_1/2 values of 11.6 and 10.2 hr (Figure 3). These data were affirmed by subsequent analyses demonstrating the deleterious effect the ζ57.1-4 mutation (but not the ζ57.5-8 mutation) on the half-life of ζ-globin mRNA (Figure 4). In these latter studies, the t_1/2 value of control ζWT mRNA (11.2 hr) was reduced to the same extent by either full (ζ57) or partial (ζ57.1-4) substitution of the nt 57–64 target region (3.6 and 3.7 hr, respectively), confirming the centrality of the CCCC tetranucleotide to ZMRE function. Additional pulse-chase studies--conducted in cells that stably express wild-type and 3′UTR-variant ζ-globin mRNAs--revealed similar results (not shown), providing a third measure of experimental validation. Collectively, the transcriptional chase analyses agree on the positioning of a post-transcriptional regulatory element within the ζ57 region, as well on its critical importance to the constitutive half-life of ζ-globin mRNA.

Previous analyses in transgenic mice demonstrated that the stability of human α-globin mRNA was significantly reduced when its 3′UTR was exchanged for the corresponding region of ζ-globin mRNA [18]. This effect was originally attributed to a generic difference in then-unspecified determinants of α- and ζ-globin half-life [17] and, later, to sequence dissimilarities between homologous pyrimidine-rich regulatory elements within the α- and ζ-globin 3′UTRs [18]. We posited that this effect might alternately arise from a ZMRE function that was highly sensitive to its regional mRNA structural context; i.e., that the regulatory properties of the ZMRE might be reduced when repositioned near heterologous mRNA. To test this possibility, we assessed whether the half-life of a reporter coding-region mRNA contiguous with the ζ57 3′UTR would be reduced relative to an identical mRNA flanked by a ζWT 3′UTR (Figure 5A). A two-point decay analysis indicated that the stabilities of 3′UTR-chimeric β-globin mRNAs were indifferent to the presence of stabilizing (ζWT) or destabilizing (ζ57 and ζ57.1-4) 3′UTRs, validating an activity for the ZMRE that is dependent upon its RNA context (Figure 5B). While it is difficult to generalize this effect to coding regions from all mRNAs, our observations suggest that the activity of the ZMRE is defined both by its primary structure (Figures 3 and 4) as well as by undefined effects of neighboring mRNA and/or mRNA-bound factors (Figure 5).

The observation that the ζ-globin mRNA is destabilized by mutations within the ZMRE, but not by similar mutations in neighboring cytosine-rich regions, suggests a local high-order structure that favors functional interactions between the ZMRE and one or more trans-acting effector factors. An unrefined m-fold analysis of ζ-globin 3′UTR structure predicts that the ZMRE is embedded in a highly stable RNA stem, and directly participates in seven base-pair interactions (Figure 6A) [39, 40]. We were skeptical that this arrangement would support site-specific interactions with regulatory RNA-binding proteins, and consequently employed an enzymatic mapping technique to formally characterize the high-order RNA structure that encompasses the ZMRE [41]. Our analyses revealed regions of ζ-globin 3′UTR that--by virtue of their sensitivities to single- and double-strand-specific RNases--were inconsistent with the unrefined structure (Figure 6B). A subsequent m-fold analysis, constrained by unequivocal, experimentally determined regions of single- and double-stranded interactions, predicted an alternate structure with a favorable ΔG^o value (−26.3 kcal/mol) in which the nt 57–64 target region is positioned on an asymmetrical interior loop, contiguous to a 16-nt stem (Figure 6C). This arrangement--which would be expected to facilitate conformational flexibility in the regions of RNA that flank the ZMRE [42]--would also account for our observation that transitional nucleotides (bordering the base-paired and unpaired regions) are assigned double-stranded interactions but display modest sensitivities to single-strand nucleases (e.g., C57 and C58). The resistance of the cytosine-rich ZMRE to double-strand-specific RNase V may additionally suggest potential hydrophobic ‘stacking’ interactions that can serve as a target for trans-acting regulatory factors [43]. The structural dynamics of this region, combined with experimental evidence favoring its assembly, indicates that the positioning of the ZMRE may be critical to its regulatory function.

The stabilities of many mRNAs--including those encoding α globin [27], β globin [16], α(I) collagen [28], tyrosine hydroxylase [29], histone [30], and the transferrin receptor [31]--require the assembly of defined mRNP effector complexes on specific determinants within their 3′UTRs. To identify candidate trans-acting factors that might functionally interact with the ZMRE, we conducted affinity chromatography analyses on cell extracts from both non-erythroid HeLa and erythroid K562 cells, using three 32-nt single-stranded (ss) DNAs corresponding to comparable regions of ζ57, control ζWT, and control ζ65 3′UTRs. Both the control ζWT and ζ65 ssDNAs retained fewer than 10 distinct factors--two of which were not retained by the ζ57 ssDNA--suggesting site-specific trans-acting interactions and potential regulatory functions (Figure 7A). Excised bands were subjected to nanoLC/MS/MS and provisionally identified as isoforms of AUF1 (hnRNP D), a protein that is ubiquitously expressed [44], displays known mRNA-binding characteristics [45, 46], participates in post-transcriptional regulatory events that effect both mRNA stabilization [47] and destabilization [48], and has been implicated as a determinant of β-globin mRNA half-life [16]. The identity of AUF1--and its relative affinities for ζWT, ζ57, and specificity-control ζ65 3′UTRs--was subsequently validated by immunoblot analyses of the three retentates (Figure 7B). Additional analyses demonstrated that AUF1 immunoprecipitate of extract prepared from ζWT-expressing HeLa^tTA cells is enriched for both ζWT and positive control c-myc mRNAs (Figure 7C), demonstrating the assembly of ζ-globin/AUF1 mRNPs in vivo. These results document the interaction of AUF1 with the ZMRE and provide an attractive mechanistic link between the structure of the ζ-globin 3′UTR and its mRNA-stabilizing functions. The importance of AUF1 to this process also suggests that the half-life of ζ-globin mRNA may be regulated in definitive erythropoiesis through post-transcriptional programs that act on other mRNAs, including β-globin mRNA, that are central to normal red cell development.

Parental pTRE2Aζ was derived from pTRE2 (Clontech) by inserting the full-length human ζ-globin gene (including ~100 nt of contiguous 5′- and 3′-flanking regions) into the polylinker Sac II-Bam HI site. A ~1.0 kb Nco I-Sap I vector fragment was subsequently deleted to eliminate a default polyadenylation signal that competes with the native ζ-globin poly(A) signal. 3′UTR-derivative ζ-globin genes were constructed from pTRE2Aζ by exchanging the exon III Bst EII-polylinker Bam HI fragment with 330-bp synthetic DNAs (Genscript), each encompassing a mutated ζ-globin 3′UTR. The construction of a corresponding full-length β-globin gene (pTRE2Aβ) is described elsewhere [15]. Genes encoding chimeric βζ-globin mRNAs were generated from pTRE2Aβ by exchanging the parental Eco RI-Eco NI DNA fragment (encompassing β-globin exon III and contiguous 3′-flanking region) for a corresponding synthetic DNA containing the desired wild-type or derivative ζ-globin 3′UTR. All recombinant DNAs were validated by automated sequencing.

HeLa^tTA cells expressing the tetracycline transactivator fusion protein (tTA; Clontech) were maintained in DMEM supplemented with 10% FBS and antibiotics, at 37°C in a humidified 5% CO₂ environment. Wild-type and 3′UTR-derivative pTRE2Aζ vectors used for stable transfections were modified by inserting a 1.5-kb DNA fragment encoding hygromycin (hyg) resistance into a unique vector Xho I site. Approximately 6×10⁵ cells were transfected with 5.0 μg of DNA using Superfect reagent under conditions recommended by the manufacturer (Qiagen), and selected with 400 μg/mL hyg. Hyg-resistant clones were tested by RT-qPCR for levels of ζ-globin mRNA and control endogenous β-actin mRNA. Tetracycline (tet) response was assessed by quantifying the level of ζ-globin mRNA, relative to the level of tet-indifferent β-actin mRNA, following a 48-hr incubation in tet-supplemented media (2 μg/mL).

Two-point decay analyses were conducted on 5×10⁵ preplated HeLa^tTA cells maintained in tet-supplemented medium (0.5 μg/mL). Cells were transfected with 5.0 μg DNA comprising equal quantities of wild-type (or 3′UTR-derivative) TRE2Aζ and control pTRE2Aβ using Superfect reagent, and replated as two aliquots for overnight growth in tet-supplemented medium (0.03 μg/mL). PBS-washed aliquots were then incubated for five hr in tet-free medium; one aliquot was sacrificed immediately (t = 0) and a second aliquot after an additional 16-hr incubation in tet-supplemented medium (2.0 μg/mL). Conventional half-life analyses were conducted on 2×10⁶ HeLa^tTA cells maintained in tet-supplemented medium (0.5 μg/mL), transfected with 10 μg DNA using Superfect reagent, and then replated in aliquots in tet-supplemented medium (0.03 μg/mL) for overnight growth. PBS-washed aliquots were then incubated in tet-free medium for five hr, supplemented with tet (2.0 μg/mL) and sacrificed at defined intervals. For both transient and stable analyses, cells were sacrificed by immersion in Trizol, and RNA prepared as recommended by the manufacturer (Invitrogen). Purified RNA was resuspended at 10 μg/mL in H₂O, and stored at −80°C.

Purified RNA (50 ng) was assayed using Taqman One-Step reagents on a model 7500 real-time PCR system, using protocols recommended by the manufacturer (Applied Biosystems; AB). Analyses were conducted using assays for human ζ globin (AB catalogue HS00923579_m1), β globin (HS00747223_g1), and β actin (HS99999903_m1), and quantified by ΔΔCt methodology that we describe elsewhere [34]. Relative decay values were calculated as the quantity of each ζ-globin mRNA remaining after a 16-hr transcriptional chase interval, relative to transiently co-expressed β-globin mRNA; the relative decay of ζWT mRNA was arbitrarily assigned unit value. p values--where indicated--were calculated using standard Student t-test methods.

Custom 5′-terminal biotinylated single-strand (ss) DNAs corresponding to 3′UTRs from ζ^WT (5′TGGAGGTTCCCCAGCCCCACTTACCGCGTAAT3′), ζ⁵⁷ (5′TGGAG GTTAGTGCACACCACTTACCGCGTAAT3′), and ζ⁶⁵ (5′TGGAGGTTCCCCAGCCAGTGCACACGCGTAAT3′) mRNAs were commercially sourced (IDT). Approximately 1 μg of each ssDNA was incubated overnight at 4°C in 500 μL cytoplasmic extract supplemented with 50 μL ImmunoPure Avidin Agarose beads (Pierce). The pelleted beads were washed 2× with PBS + Triton-×100 (0.05%) and 2× with PBS + Triton-×100 (1.0%), resuspended in 10 μL loading buffer, and resolved on a precast 4-12% gradient SDS-polyacrylamide gel (Invitrogen). Cytoplasmic extracts were prepared from ~1×10⁷ cells lysed in buffer (1 mM Hepes pH 7.9, 0.15 mM MgCl₂, 1 mM KCl) and clarified by centrifugation at 13000xg for five minutes at 4°C; extract was stored in aliquots at −80°C.

Analyses were conducted by the University of Pennsylvania Proteomics Core Facility. Tryptic digests were studied by nanoLC/MS/MS using Thermo LTQ and Eksigebt nano LC-2 Da instruments, and data analyzed from the Uniprot_Sprot database using Sequest and Scaffold software packages. Statistical p-value cutoffs of 95.0% and 99.0% were applied for peptides and proteins, respectively.

A DNA template for in vitro transcription of the ζ^WT 3′UTR, containing an 18-nt polyadenylate tail, was directionally inserted into the Xho I-Bgl II polylinker site of pSP72. The Bgl II-linearized plasmid was transcribed in vitro as previously described [41], and the purified RNA then 5′-end labeled with [γ-³²P]ATP using a Kinase Max kit (Ambion). The [³²P]-labeled RNA was digested with RNase A (100, 10, 1.0, and 0.1 ng/mL), RNase T1 (100, 10, 1.0, and 0.1 mU/μL), or RNase V1 (10, 1.0, 0.1, and 0.01 mU/μL) as directed by the supplier (Ambion). RNases A and T1 exhibit cleavage specificities for pyrimidine and guanosine bases in single-stranded regions of RNA, respectively, while RNase V1 cleaves 3′ to nucleotides within double-stranded regions of RNA. Reaction products were resolved on a 33 cm × 40 cm 6%:8 M acrylamide:urea gel, in parallel with a migration-control ‘ladder’ generated by alkaline hydrolysis of [³²P]-labeled transcripts [41].