Inhibition of c-erbA mRNA Splicing by a Naturally Occurring Antisense RNA*

The rat erbAa locus encodes two overlapping mRNAs, a1 and ( ~ 2 , which are identical except for their most 3‘ exons. a1 mRNA encodes a thyroid hormone receptor, while a2 encodes an altered ligand binding domain of unknown function. Previous studies have shown that the ratio of a1 to a2 is highest in cells expressing a high level of a third RNA, Rev-ErbAa mRNA, which is transcribed in the opposite direction and is complementary to a2 but not a1 mRNA. It was hypothesized that base pairing with Rev-ErbAa blocks splicing of a2 mRNA, thereby favoring formation of the non-overlapping a l . To test this model, a system was developed in which a2 pre-mRNAs were accu- rately spliced in vitro. Splicing was inhibited by the addition of a &fold excess of antisense RNAs contain- ing the 3’ end of Rev-ErbAa mRNA. Both an antisense RNA extending across the 3’ splice site and a shorter RNA complementary only to exon sequences efficiently blocked splicing. However, splicing was only inhibited by complementary RNAs. These observations are consistent with a mechanism in which base pairing with a complementary RNA regulates alternative processing of a1 and a2 mRNAs.

The c-erbAa locus also codes for a third mRNA, Rev-ErbAa, which is transcribed in the opposite direction from a1 and a2 (14,23). The 3' end of this mRNA overlaps the 3' end of a2, but not a l , mRNA (Fig. 1A). Like a1 and a2, Rev-ErbAa encodes a protein belonging to the steroid/thyroid hormone receptor superfamily (14,23,24), but of unknown function. However, previous studies have shown that in tissues where Rev-ErbAa mRNA levels are high, the ratio of a2/a1 mRNA is relatively low (16,23). Furthermore, conditions which increase transcription of Rev-ErbAa mRNA also decrease the ratio of a2/al, both by increasing a1 and decreasing a2 mRNA levels (25). These changes are due to an alteration in post-transcriptional processing of erbAa pre-mRNA, since neither transcription nor stability of these two mRNAs is altered (25).
While alternative processing of some mRNAs is known to be mediated by RNA-binding proteins (26), the unusual organization of the c-erbAa locus and the reciprocal relationship between Rev-ErbAa levels and the a2/al ratio suggested the possibility that base pairing of Rev-ErbAa mRNA with the 3' end of a 2 pre-mRNA may directly inhibit splicing of a2 mRNA, thereby favoring expression of the non-overlapping a1 mRNA (23,25). In order to test this mechanism, a model system for studying a2-specific splicing in vitro has been developed. The present work shows that erbAa2 transcripts are accurately spliced in nuclear extracts. Furthermore, RNAs derived from the 3' portion of Rev-ErbAa efficiently and specifically block splicing of exons unique to a 2 provided that complementarity is maintained, as predicted by the antisense mechanism outlined above.

MATERIALS AND METHODS
Plasmids-The construction of plasmids pHB500 and pHBS has been described previously (27). pHB-n2L was constructed by inserting the 775-nt HincIIIEcoRI fragment (A in Fig. 1A) which includes the a2-specific 3' splice site from the rat erbAn locus (16,23) into the same sites of pHBS, downstream of the @-globin insert. pHB-n2S was constructed by deleting the 473-nt HincIIISspI fragment from the a2 intron sequences of pHB-n2L. The resulting insert is labeled B in Fig. 1A. pn2-L is a minigene in which the 3' and 5' splice sites are identical to those in the rat erbAn gene. This plasmid was constructed by replacing the globin insert of pHB-n2L with the 510nt PstIIXbaI fragment (C in Fig. 1 A ) from c u l cDNA (16). This fragment contains 150 nt of exon sequence common to 01 and n2 mRNAs, including the entire length of the last common exon, and 360 n t of contiguous sequence specific for a1 mRNA (but spliced out of n2). pn2-L includes a short region of polylinker from pBluescript (Stratagene) between the Hind111 site of the vector (pGEM3, Promega) and a PstI site at the end of the n l cDNA insert as well as a short region of pGEM4 between the XbaI site of the insert and the SphI site of the vector. pn2-S was constructed by replacing fragment A in pcr2-I, with fragment R. pErhAtrP/E contains the HincIIIEcoRI fragment from pHH-cr2L ligated to the same sites in pGEM4.
R N A 7'ransrripts"HHT,OO and E2-1:10 were transcribed from pHH500 with 'r7 and SI% RNA polymerases as previously described (27). Transrripts HWcr2L. HH-cr2S. tr2-I,, and cr2-S were synthesized with SP6 polymerase from the corresponding plasmids cut with I.'coRI. n2-SN was similarly transcrihed from prr2-S cut with NciI. The strurture of these transcripts is shown in Fig. 1H. Rev-470 and Rev-165 are antisense RNAs corresponding to regions D and E of Rev-ErbAcu mRNA (Fig. 1A) and were transcribed with Ti RNA polymerase from pErhAtrP/E cut with Poull and Ncil, respectively. The Ncil site at the 3' end of Rev-165 is 23 nt upstream of that at the 3' end of the non-overlapping tr2"SN RNA. Pre-mRNAs were uniformly labeled with '"P and antisense RNAs with 'H.
Splicing Hrnctions-In uifro pre-mRNA splicing in HeLa cell nuclear extract and subsequent analysis of the resulting products were carried out as previously described (27,28). Optimal splicing of erhAn pre-mRNAs required 1.6 mM MgCI, and 68 mM KCI, and optimal splicing of HR500 3.2 mM MgCI, and 4 8 mM KCI. ,"'P-Laheled DNA fragments from an Mspl-digested pRR322 were included as molecular weight markers on all gels. site of Rev-ErhAn mRNA. Two chimeric transcripts were also tested which include the first exon and 5' splice site of /jglobin pre-mRNA upstream of the n2-specific 3' splice site (Fig. 1H). Each of these transcripts was efficiently spliced in the presence of HeLa cell nuclear extract. Following incuhation of both n2-L and n2-S for 45-120 min, a prominent hand of 430 nt appeared, corresponding to the size expected for splicing of the n2 exons (Fig. 2 A , lanes 2-3 and 6-9). Additional bands corresponding to expected products and intermediates of splicing were observed. Lariat forms of the intron and intron/exon intermediate were identified on the basis of their size and characteristic shift in mobility on gels of varying polyacrylamide concentrations (results not shown). To confirm the identity of the putative spliced products, the 430-nt band was eluted and analyzed hy RNase protection assays following hybridization to an unlaheled RNA probe complementary to t h e 3' n2-specific exon and adjacent intron sequences. The 430-nt band from both n2-L and n2-S yielded a RNase-resistant fragment of 250 nt, corresponding to the length of the accurately spliced 3' exon.

Splicing
Under conditions optimized for splicing of n2 pre-mRNAs, splicing of tr2-L was substantially slower than that of n2-S. Both RNAs were spliced more slowly than a well characterized 500-nt globin pre-mRNA, HR500 (28). However, the chimeric transcripts, which include the 5' exon of HR.500 and the 3' exon of n2, were spliced more efficiently. Both HR-n2S (Fig.  2H) and HR-n2L (not shown) were spliced more rapidly than n2-S. Both yielded the expected products and intermediates, including a band which comigrated exactly with the 5' f3globin exon (lanes 2, 3 , .5, and 6, Fig. 2H). The relative rate of splicing of n2 transcripts increased with decreasing intron length. In fact, H R -d S (total intron length 118 nt) was spliced more rapidly than glohin pre-mRNA (intron length 130 nt) under conditions optimized for HR5OO splicing (compare lanes 2 and -5 in Fig. 2 H ) . This result demonstrates that of erbA mRNA Splicing t h e n2-specific 3' splice site is utilized very efficiently, even in combination with a heterologous 5' site.
Antisense Inhibition of erbAn2 Splicing by Rev-ErbAn RNA-Having demonstrated accurate splicing of the n2 pre-mRNA in vitro, the proposed role of Rev-ErbAn mRNA in blocking this process was explored. Fig. 3 shows that splicing of the n2-L pre-mRNA was strongly inhibited in the presence of Rev-470 RNA, which consists of 470 nt from the 3' end of the Rev-ErbAn mRNA; 246 nt at t h e 5' end of Rev-470 are complementary to the 3' n2 exon, and the 3' end is 10 n t upstream of the Rev-ErbAn polyadenylation site. Of note, a 5or 10-fold molar excess of Rev-470 blocked the n2-L splicing by more than 70% (compare lanes 5 and 6 with lane 2 in Fig. 3). In contrast, the same molar excess of an RNA complementary to the @-globin pre-mRNA (E2-130) had no effect on n2-L splicing (lanes 2 and 3 ) but did block splicing of HB500 to which it is complementary (compare lanes 8 and 9  with lane 7). Importantly, however, the inhibitory effect of Rev-470 was specific for the n2-containing pre-mRNA. It had n o effect on splicing of @-globin pre-mRNA (lanes 11 and 12).
T h e region of complementarity between Rev-470 RNA and n2-L includes the 3' splice site of the a2 pre-mRNA. In order to determine whether base pairing in this region is required for inhibition of splicing, A shorter RNA, Rev-165, was syn- thesized which is complementary to 165 nt at the 3' end in the n2 pre-mRNA. This RNA does not include a sequence which hybridizes to t h e 3' splice site. Fig. 4 shows that Rev-165 is approximately equal to Rev-470 in its ability to inhibit the splicing of n2-S (lanm 4 and ii), indicating that inhibition of 02 mRNA splicing does not require base pairing with the 3' splice site. However, the complementary nature of the Rev-165 is necessary for its inhibitory properties in this system, because Rev-165 has no effect on the splicing of n2-SN, a truncated n2 pre-mRNA lacking sequences complementary to Rev-165 (lane 9 ) while Rev-470, which contains sequences complementary to this pre-mRNA, does efficiently inhihit its splicing (lane IO). DISCIJSSION The present studies show that a naturally occurring RNA, Rev-ErbAn mRNA, can inhibit splicing of n2 pre-mRNA to which it is complementary. These results are formally similar to those of earlier experiments in which antisense RNAs were observed to block splicing of glohin pre-mRNA ( X ) , although in the latter case complementary mKNAs do not exist in viLn Results demonstrate that inhibition of globin mRNA splicing is associated with base pairing between complementary regions of the pre-mRNA and antisense RNA molecules. Inhihition of globin mRNA was observed with RNAs which anneal either directly to splice sites or exclusively to exon sequences as far as 170 nt from the nearest splice site (27). Studies of globin mRNA splicing also demonstrated that a number of proteins facilitate or disrupt base pairing in HeLa cell nuclear extract (27). The potential involvement in antisense inhibition of factors which modulate RNA-RNA base pairing remains to be established.
Previously it was suggested that levels of the alternatively spliced a1 and a2 mRNAs may be regulated at the level of splicing by a novel antisense mechanism in which Rev-ErbAa transcript base pairs with the complementary pre-mRNA and blocks formation of a2 but not a1 mRNA (23,25). The results presented here are consistent with this model. Transcripts containing sequences from the 3' end of Rev-ErbAa mRNA efficiently inhibit splicing of a 2 transcripts in HeLa cell nuclear extract (Figs. 3 and 4). This result was observed with Rev-470 which, like endogenous Rev-ErbAa mRNA, extends across the 3' a2-specific splice site, but also with a shorter antisense RNA complementary only to the 3' exon of a 2 mRNA, indicating that splicing is sensitive to relatively limited base-pairing interactions. The very efficient splicing of a 2 mRNA i n uitro raises the possibility that formation of a2 mRNA is favored over a1 in the absence of specific regulation. These results, taken together with those of other studies, suggest that inhibition of splicing by a naturally occurring antisense RNA may play a physiologically important role. Since the a1 and a2 proteins are functionally antagonistic (20,21) relatively modest changes in splice site selection could cause major changes in cellular T3 responsiveness. However, it is possible that other mechanisms acting at the level of polyadenylation or transcriptional termination, as well as splicing, may regulate alternative processing of a1 and a2 mRNAs. Additional genetic and biochemical approaches will be required to ascertain that Rev-ErbAa functions as a physiologically significant antisense regulator i n uiuo.