A quality control pathway that down-regulates aberrant T-cell receptor (TCR) transcripts by a mechanism requiring UPF2 and translation.

Nonsense-mediated decay (NMD) is an RNA surveillance pathway that degrades mRNAs containing premature termination codons (PTC). T-cell receptor (TCR) and immunoglobulin (Ig) transcripts, which are encoded by genes that very frequently acquire PTCs during lymphoid ontogeny, are down-regulated much more dramatically in response to PTCs than are other known transcripts. Another feature unique to TCR, Ig, and a subset of other mRNAs is that they are down-regulated in response to nonsense codons in the nuclear fraction of cells. This is paradoxical, as the only well recognized entity that recognizes nonsense codons is the cytoplasmic translation apparatus. Therefore, we investigated whether translation is responsible for this nuclear-associated mechanism. We found that the down-regulation of TCR-beta transcripts in response to nonsense codons requires several features of translation, including an initiator ATG and the ability to scan. We also found that optimal down-regulation depends on a Kozak consensus sequence surrounding the initiator ATG and that it can be initiated by an internal ribosome entry site, neither of which has been demonstrated before for any other PTC-bearing mRNA. At least a portion of this down-regulatory response is mediated by the NMD pathway as antisense hUPF2 transcripts increased the levels of PTC-bearing TCR-beta transcripts in the nuclear fraction of cells. We conclude that a hUPF2-dependent RNA surveillance pathway with translation-like features operating in the nuclear fraction of cells prevents the expression of potentially deleterious truncated proteins encoded by non-productively rearranged TCR genes.

The nonsense-mediated decay (NMD) 1 pathway is a quality control system that selectively eliminates mRNAs containing premature termination codons (PTCs) in all organisms that have been examined to date (1)(2)(3)(4)(5)(6)(7)(8). PTCs result from random mutations and biosynthetic errors (in transcription and RNA splicing) that can occur in virtually any gene. PTCs are particularly common in TCR and Ig genes, which frequently acquire frameshifts and nonsense mutations as a result of the programmed rearrangements that occur during T-and B-lymphocyte development, respectively (2). Without a surveillance pathway to detect such PTC-bearing transcripts, lymphocytes would accumulate and translate high levels of truncated proteins, some of which may possess dominant-negative or deleterious gain-of-function properties (2, 3, 8 -10).
Because the signal that triggers down-regulation is a nonsense codon, it was anticipated that it would cause the decay of mRNAs in the cytoplasm where the translation machinery is known to function. Indeed, Saccharomyces cerevisiae mRNAs harboring nonsense codons are degraded more rapidly than their wild-type counterparts in the cytoplasm (3,6,7). Nonsense codons also destabilize many mammalian and viral mRNAs in the cytoplasm fraction of mammalian cells (1,4,5,(11)(12)(13)(14). However, a surprise was the finding that TCR-␤, Ig-, and some other mammalian transcripts are down-regulated by nonsense codons in the nuclear fraction of cells (1,2,4,15,16). Two observations support such nuclear involvement. First, the decay rate of such nonsense codon-containing mRNAs is not measurably different from that of their wild-type counterparts in the cytoplasm (1,2,4). Second, reductions in the mRNA level in the nuclear fraction of cells mirror those seen in total or cytoplasmic RNA preparations (1,2,4).
The apparent involvement of the nucleus in this nonsense codon-induced event leads to the question of whether NMD requires translation or instead uses another codon-scanning mechanism. Evidence that NMD involves a mechanism with some features of translation includes its requirement for an initiator ATG and its reversal by several different protein synthesis inhibitors with different mechanisms of action (cycloheximide, emitine, anisomycin, puromycin, pactamycin, and poliovirus) (17)(18)(19)(20). However, the requirement for an ATG does not prove the involvement of conventional translation, and general protein synthesis inhibitors could reverse NMD by a mechanism not involving a blockade of mRNA scanning; for example, they could act by simply depleting cells of one or more unstable proteins necessary for NMD. Furthermore, the mechanism that down-regulates TCR and Ig transcripts in response to nonsense codons may differ from that which acts on other transcripts.
In the present investigation, we investigated the role of translation and ribosomes in the down-regulation of TCR-␤ transcripts harboring nonsense codons using new approaches. Our results showed not only did optimal down-regulation of TCR-␤ mRNA require an initiator ATG, it also depended on key nucleotide residues surrounding the ATG that are required for efficient translation. Such Kozak consensus sequences are a signature feature of translatable open reading frames (21), which strongly support a role for translation in TCR-␤ mRNA down-regulation. We found that TCR-␤ down-regulation in response to nonsense codons also had several other features of translation, including the ability to be down-regulated by an internal ribosome entry site (IRES), a cis element that specifically recruits ribosomes for translation (22). Lastly, to determine whether TCR-␤ down-regulation in response to nonsense codons has features of classic NMD, we examined the role of UPF2, which has been shown in S. cerevisiae to be required for NMD. Our antisense studies clearly showed that the downregulation of TCR-␤ transcripts depends on human UPF2 (hUPF2). To our knowledge, this is the first time that hUPF2 has been shown to be required for the down-regulation of any transcript in response to nonsense codons in mammalian cells.

EXPERIMENTAL PROCEDURES
Plasmids-Construct A (␤-290) contains a wild-type TCR-␤ gene with a full-length open reading frame (pAc/IF in Ref. 16). B and C (␤-367, ␤-368, respectively) are derivatives of A that contain nonsense (TAA) and missense (TAC) mutations at codon 68 in the VDJ exon generated by site-specific mutagenesis, respectively. D, E, and F (␤-658, ␤-627, and ␤-617, respectively) each contain a stem loop (⌬G Ϫ 61 kcal/mol) identical to that previously shown to impede translation (23) at a site 42 nt upstream of the initiator ATG in constructs B, C, and A, respectively. G (␤-495) is a TCR-␤ minigene (containing L, VDJ, and a C2.1-C2.4 chimeric exon) cloned between the SalI and HindIII sites of the vector pSBC-2-C4 (EV 147) (24). The TCR-␤ minigene in G is identical to the one we previously described (construct C in Ref. 19) except that the JC intron was shortened from 1021 nt to 358 nt by cutting out an internal Eco 01091 fragment. H (␤-496) is identical to G except it has a TAA nonsense codon at codon 68. I (␤-498) and J (␤-497) were prepared by inserting a 1.1-kb ClaI/NotI fragment from pSBC-1-C2 (EV-147), which contains the type I poliovirus IRES (nt 1-628 of the 5Ј-untranslated region (UTR)) (24) between the ClaI and NotI sites of G and H, respectively. K (␤-600) is identical to B except that it has mutated initiator ATGs rendered defective as previously described (construct I in Ref. 19). L (␤-639) is a derivative of B that has only the VDJ exon initiator ATG mutated as previously described (construct H in Ref. 19). M, N, O, and P (␤-626, ␤-630, ␤-647, and ␤-665, respectively) are derivatives of construct L that contain point mutations in the Kozak consensus sequences. Q (␤-760) and R (␤-756) are identical to N and P, respectively, except that they lack PTCs. S (␤-595) is a derivative of A that contains a nonsense mutation (TAG) at codon 98 (in the VDJ exon). The anti-UPF2 construct (G-407) was generated by inserting a 0.5-kb hUPF2 fragment between the SalI and BamHI sites of the pH␤ Apr-1neo (EV-107) vector (26) such that it is in the antisense orientation with respect to the ␤-actin promoter. The hUPF2 fragment was a PCR product generated with the primers MDA-720 (5Ј-CTGGGATCCCGA-GCGCTGGAGTTGGTG-3Ј) and MDA-703 (5Ј-CCTGTCGACGCTGAA-TGGATTCTTC-3Ј) using the plasmid pCMV-rent 2 (G-314) (25) as the template. G-1F was prepared by inserting the human ␤-globin gene into the HindIII and BamHI sites of EV-107. All mutations introduced in the constructs described above were generated by site-specific mutagenesis (27).
Transfection, RNA Isolation, and RNase Protection Analysis-DNA constructs were transiently transfected into HeLa cells using Lipo-fectAMINE according to the manufacturer's instructions (Invitrogen). Total and nuclear RNA were isolated as described previously (28). TCR-␤ mRNA levels were determined using a direct radioactivity scanner (Instant Imager; Packard Instruments, Downers Grove, IL). RNase protection analysis (RPA) and the riboprobes (TCR-␤, neomycin, ␤-actin) used for this analysis were described previously (28). The ␤Ϫglobin riboprobe template is a 250-nt PCR fragment containing 50 nt of the 3Ј end of human ␤Ϫglobin intron 2 and 200 nt of exon 3. The hUPF2 riboprobe template is a 189-nt hUPF2 3Ј-cDNA fragment. We determined that our RPA assay was quantitative by performing titration experiments; increasing the amount of input RNA linearly increased the level of the protected TCR-␤ bands, whereas increasing the amount of riboprobe had no effect on the protected bands, indicating that excess probe was present in the annealing reaction (data not shown).

A Stem Loop Reverses TCR-␤ Down-regulation in Response to
a Nonsense Codon-To examine the role of translation in the down-regulation of TCR-␤ mRNA in response to nonsense codons, we transiently transfected TCR-␤ constructs into HeLa cells. We used HeLa cells because they reproduce all aspects of the PTC-mediated down-regulation of TCR-␤ transcripts that we have observed in stably transfected T cells (16,18,28). A major advantage of HeLa cells over T cells is they can be transiently transfected with sufficient efficiency to permit analysis of the RNA products by RPA.
To determine whether the down-regulation of TCR-␤ transcripts in response to nonsense codons depends on a scanning process, we introduced a stem loop in the 5Ј-UTR of a functionally rearranged V ␤8.1 D ␤2 J ␤2.3 C ␤2 gene that we have used in past studies to examine the response to nonsense codons (16, 18, 28 -30). The stem loop that we introduced was identical to that which has been shown in past studies to efficiently inhibit 5Ј cap-dependent translation (23). Introduction of this stem loop almost completely abolished the down-regulation of TCR-␤ mRNA in response to a PTC (Fig. 1). Down-regulation was Ͼ25-fold without the stem loop (compare construct B with A) and was reduced to ϳ2-fold after addition of the stem loop (compare construct D with F). The stem loop specifically pre-FIG. 1. A stem loop reverses TCR-␤ down-regulation in response to a nonsense codon. RPA of total cellular RNA (10 g) isolated from HeLa cells transiently transfected with the constructs is shown. Construct A consists of the L and VDJ exons of a functionally rearranged V ␤8.1 D ␤2 J ␤2.3 gene (labeled L ␤ and VDJ ␤ , respectively) and four C region exons (labeled C ␤2.1 , C ␤2.2 , C ␤2.3 , and C ␤2.4 ). Constructs B and C contain nonsense and missense mutations, respectively, at codon 68 but otherwise are identical to construct A. Constructs D, E, and F are identical to B, C, and A, respectively, except that they contain a stem loop in the 5Ј-UTR. Because all constructs also contain an independent transcription unit encoding neomycin, neomycin (Neo) mRNA levels were used as a measure of transfection efficiency. The TCR-␤ mRNA band protected by the TCR-␤ probe is ϳ72 nt, which is the size expected based on the positions of the splice sites in TCR-␤ mRNA. Similar results were obtained in three independent transfection experiments.
vented the down-regulatory response to a nonsense mutation, as it had no effect on the level of mRNA that had a silent mutation at the same site as the nonsense codon (constructs C and E). We conclude that either translation or some other scanning mechanism that reads across the 5Ј-UTR is required to trigger TCR-␤ mRNA down-regulation in response to a nonsense codon.
TCR-␤ mRNA Down-regulation in Response to a Nonsense Codon Can Be Initiated by an IRES-To test whether ribosomes are involved in the scanning event, we examined whether an IRES permitted TCR-␤ down-regulation in response to a nonsense codon. To assess this, we generated TCR-␤ minigene constructs with or without a type I poliovirus IRES (31,32). We found that introduction of a nonsense codon caused 3-fold down-regulation of IRES-containing TCR-␤ transcripts (compare constructs I with J in Fig. 2), which is comparable with the amount of down-regulation of many non-IRES-containing transcripts, including those encoding triosephosphate isomerase and ␤-globin (18,23). Introduction of a nonsense codon in the IRES-lacking (control) construct down-regulated TCR-␤ mRNA expression by 10-fold (compare construct G with H in Fig. 2). The ϳ3-fold higher degree of down-regulation exhibited by the IRES-lacking TCR-␤ mRNA is consistent with previous studies showing that cap-dependent translation is ϳ3-fold more efficient than poliovirus IRESmediated translation (24,33).
An Initiator ATG and Surrounding Kozak Consensus Nucleotides Are Essential for Optimal Down-regulation of TCR-␤ Transcripts in Response to a Nonsense Codon-We previously showed that an initiator ATG is essential for the down-regulation of transcripts derived from a TCR-␤ minigene containing a PTC (19). We showed that there are two ATGs that could trigger down-regulation in this minigene: the normal initiator in the leader (L) exon and a downstream ATG in the 5Ј end of the variable region (VDJ) exon that is in frame with the L exon ATG. Only mutation of both ATGs reversed TCR-␤ down-regulation in response to a PTC (19). Here, we first examined whether an ATG was needed for the down-regulation of transcripts from a full-length version of this TCR-␤ gene. Mutation of the VDJ ATG had no appreciable effect on TCR-␤ downregulation (data not shown). However, when the L ATG was also mutated (construct K), down-regulation was almost completely reversed (compare construct B that has both the L and VDJ ATGs intact with construct K that has both the L and VDJ ATGs mutated) (Fig. 3A). We conclude that the normal initiator ATG in the L exon is essential for the robust down-regulation of full-length TCR-␤ transcripts in response to a PTC when the alternative ATG in the VDJ exon is also mutated so that it cannot be used as an alternative initiation site.
If conventional translation is responsible for TCR-␤ downregulation in response to nonsense codons, then both the initiator ATG and the surrounding nucleotides known to be essential for efficient translational initiation should be required for optimal down-regulation (21). This issue has never been addressed before for any gene. To test this, we mutated the nucleotides surrounding the L exon ATG in constructs that had the VDJ exon ATG mutated to prevent alternative initiation FIG. 2. TCR-␤ mRNA down-regulation in response to a nonsense codon can be elicited by an internal ribosome entry site (IRES). RPA of total cellular RNA (10 g) isolated from HeLa cells transiently transfected with the constructs is shown. Construct G is a TCR-␤ minigene composed of three exons, including a chimeric C exon (labeled C ␤ ) that has portions of the C ␤2.1 and C ␤2.4 exons, driven by an MT7 retroviral promoter (24). Construct H is identical to G except that it harbors a PTC at codon 68. Constructs I and J are identical to G and H, respectively, except that an IRES was inserted before the ATG start codon in the L exon. A plasmid expressing the ␤-globin gene (G1-F) was cotransfected with the TCR-␤ constructs to permit measurement of transfection efficiency. Similar results were obtained in at least three independent transfection experiments.

FIG.
3. An initiator ATG and surrounding Kozak consensus nucleotides are essential for optimal down-regulation of TCR-␤ transcripts in response to a nonsense codon. RPA of total cellular RNA (10 g) isolated from HeLa cells transiently transfected with the constructs is shown. mRNA levels were quantitated as described in Fig.  1. Similar results were obtained in at least three independent transfection experiments. A, mutation of both initiator ATGs reversed the down-regulation of TCR-␤ transcripts bearing a PTC. The ATG in the L exon is the normal translation start site; the ATG in the VDJ exon is in the same reading frame as the L exon start codon. Construct K has both ATGs mutated and has a PTC at codon 68. B, the Kozak consensus sequence around the initiator ATG is required for optimal down-regulation in response to a nonsense codon. Constructs N and P have a PTC at codon 68, whereas constructs A, Q, and R have the normal sense codon at this position. All constructs have a normal L exon ATG and a mutated VDJ exon ATG. The sequences surrounding the L exon ATG in each construct are indicated in Table I. from this second initiator site. We found that mutation of the two most critical Kozak consensus sequences required for translation (purines at the Ϫ3 and ϩ4 positions with respect to the ATG) prevented a full down-regulatory response (compare mutated construct P with the wild-type construct L; Table I).
As expected, polysome analysis demonstrated that this double mutation significantly inhibited translation (data not shown). More subtle mutations that only substituted one (not both) of the critical purines (constructs M and O), and thus would be expected to only minimally inhibit translation, had either no or only a modest inhibitory effect on down-regulation (Table I).
Inversely, a mutation that better matched the TCR-␤ sequence to the Kozak consensus sequence (construct N) triggered a stronger down-regulatory response ( Table I).
The nonsense codon-bearing constructs with the best-and least-matched Kozak sequences (constructs N and P, respectively) differed in their mRNA levels by almost 3-fold (Fig. 3B). In contrast, there was no significant difference in the level of mRNA expressed from constructs that had these same mutations but lacked a nonsense codon (constructs Q and R) (Fig.  3B). Thus, these Kozak consensus sequence mutations do not have a general effect on mRNA metabolism. We conclude that an optimal down-regulatory response to a nonsense codon requires Kozak consensus nucleotides surrounding an intact initiator ATG. Our finding that the match to the Kozak consensus sequence correlated with the ability to down-regulate TCR-␤ transcripts strongly supports the notion that the RNA surveillance pathway responsible for this down-regulatory response involves translation.
The NMD Factor UPF2 Plays a Role in the Down-regulation of TCR-␤ Transcripts in the Nuclear Fraction of Cells-Once we had obtained several lines of evidence that translation is required for PTC-induced down-regulation of TCR-␤ transcripts, we next assessed whether this down-regulation is exerted using the NMD RNA surveillance pathway. We considered the possibility that TCR and Ig transcripts use a pathway distinct from NMD, as these mRNAs are transcribed from genes that acquire PTCs much more frequently than do other genes, are down-regulated more strongly in response to nonsense codons than are other known transcripts, and may require a unique second signal to be down-regulated (2,29,30). To examine the role of NMD, we assessed whether TCR-␤ down-regulation required hUPF2. Although human UPF2 has not been proven to be involved in mammalian NMD, its orthologues in S. cerevisiae (Upf2) and Caenorhabditis elegans (Smg3) have been shown to be essential for NMD (3,6,7). Furthermore, tethering a hUPF2/MS2 fusion protein downstream of the stop codon in ␤-globin mRNA triggers an NMD-like response in HeLa cells (34).
To determine the role of hUPF2 in TCR-␤ down-regulation, we generated an expression plasmid that transcribes antisense hUPF2 mRNA. This anti-hUPF2 plasmid was cotransfected with TCR-␤ plasmids into HeLa cells, followed by RPA of nuclear RNA from these cells. Cotransfection of the anti-hUPF2 plasmid partially reversed the down-regulation of PTC-bearing TCR-␤ transcripts in the nuclear fraction (Fig. 4A). This reversal was specific for the transcript harboring the nonsense codon (construct S) as its level was increased 4-fold, whereas the levels of the wild-type transcript (construct A) were not altered (Fig. 4A). Cotransfection of a control vector plasmid that lacked antisense hUPF2 sequences had no effect on TCR-␤ mRNA down-regulation (data not shown). To determine the degree of inhibition caused by anti-hUPF2, we examined hUPF2 mRNA levels by RPA. We found that cotransfection of the antisense-hUPF2 construct significantly decreased the level of endoge-  (R ϭ A, G) and G are the most critical residues for start-site recognition. The VDJ exon ATG in each construct was disrupted by mutation, whereas the ATG in the L exon (shown) is intact. mRNA levels were determined by RPA and normalization against neomycin mRNA levels (as in Fig. 1). The values obtained reflect the average and standard error from three to four experiments.

FIG. 4. hUPF2 plays a role in nonsense codon-induced downregulation of TCR-␤ transcripts in the nuclear fraction of cells.
RPA of total cellular RNA (10 g) isolated from HeLa cells transiently transfected with the constructs shown. A, antisense hUPF2 reverses down-regulation of TCR-␤ transcripts in response to a nonsense codon. ␤-globin mRNA levels were used as a measure of transfection efficiency as described in Fig. 2. Construct S is identical to construct A except that the nonsense mutation (TAG) is at codon 98. The anti-hUPF2 plasmid was cotransfected with constructs A and S in the lanes shown. Similar results were obtained in three independent transfection experiments. B, antisense hUPF2 decreases the level of endogenous hUPF2 mRNA. The hUPF2 mRNA band protected by the hUPF2 probe is ϳ190 nt. hUPF2 mRNA levels were determined by normalizing against the level of endogenous ␤-actin transcripts. Constructs A and S were cotransfected with either the anti-hUPF2 plasmid or a control vector-only plasmid. Similar results were obtained in at least two independent transfection experiments. nous hUPF2 mRNA (ϳ3-fold) compared with its level after cotransfection with a control vector-only plasmid (Fig. 4B). We therefore conclude that hUPF2 participates in the down-regulation of TCR-␤ transcripts in response to nonsense codons. To our knowledge, this is the first demonstration that human UPF2 is essential for the NMD response in mammalian cells. DISCUSSION We have provided several lines of new evidence that the down-regulation of TCR-␤ transcripts in response to nonsense codons requires translation even though it occurs in the nuclear fraction of cells. We demonstrated that this down-regulation is mediated by an initiator ATG-dependent scanning process that requires the Kozak consensus sequence for optimal down-regulation ( Figs. 1 and 3). This, along with our observation that the down-regulation can be triggered by an IRES (Fig.  2) and that it is reversed by specific suppressor tRNAs (19), strongly suggests that nonsense codons down-regulate TCR-␤ transcripts by a mechanism that depends on translation.
The simplest interpretation of our IRES data is that the NMD down-regulatory response requires a ribosome, although we cannot rule out that a non-ribosomal entity was recruited by the IRES in our experiments. To our knowledge, this is the first time that the down-regulatory response to a nonsense codon has been shown to be elicited by a 5Ј cap-independent mechanism. Several lines of evidence support the view that the IRES we used drives cap-independent translation and blocks cap-dependent translation. First, dicistronic mRNAs that contain the poliovirus IRES inserted between two reporter cistrons efficiently translate the second cistron without a requirement for ribosomes to traverse the first cistron (24). Second, introduction of small lesions throughout the central portion of the poliovirus IRES (region P) block cap-independent initiation (35). Third, when the poliovirus IRES situated upstream of a reporter gene was debilitated by deletion or point mutation, translation of the reporter gene was extinguished, indicating that this IRES does not permit cap-dependent translation, probably because of its strong secondary structure (36). Fourth, when the extensive stem loop-rich region of the poliovirus IRES was deleted, cap-dependent translation was now permitted (37).
Collectively, our data strongly suggests that the translation apparatus is responsible for scanning TCR-␤ mRNAs for nonsense codons. This codon scanning could occur in the nucleus proper as recent evidence suggests that ϳ10% of mammalian cell translation is coupled with transcription in the nucleus (38). This notion is further supported by evidence for coupled transcription and translation in lower eukaryotes and the finding that charged tRNAs and some translation initiation and elongation factors accumulate in the nucleus of mammalian cells (38 -40). Alternatively, nonsense codon scanning may be mediated by cytoplasmic ribosomes that biochemically cofractionate with the nucleus. Even though we purified nuclei using two detergents (Nonidet P-40 and sodium deoxycholate) by a protocol that removes Ͼ97% of cytoplasmic mRNA (16), it is possible that NMD occurs near the nuclear membrane and thus copurifies with the nucleus (1,41). This cytoplasmic scanning model appears to apply to ␤-globin mRNA, as it has been shown that specific blockade of cytoplasmic scanning by the cytoplasmic RNA-binding protein aconitase reverses ␤-globin NMD (42).
Our demonstration that nonsense codons down-regulate TCR-␤ transcripts by a UPF2-dependent mechanism (Fig. 4) demonstrates the involvement of the classic NMD RNA surveillance pathway in this down-regulatory response. However, we were not able to completely reverse TCR-␤ mRNA downregulation, even when we cotransfected high concentrations of anti-hUPF2 plasmid (data not shown). This may reflect the limitations of antisense technology, but it could also indicate that TCR-␤ transcripts are down-regulated by two mechanisms, one that is hUPF2-dependent (the NMD RNA surveillance pathway) and another that is hUPF2-independent.