The editing enzyme ADAR1 and the mRNA surveillance hUpf1 interact in the nucleus

Lily Agranat*, Oleg Raitskin*, Joseph Sperling†, and Ruth Sperling*‡

*Department of Genetics, The Hebrew University of Jerusalem, Jerusalem 91904, Israel; and †Department of Organic Chemistry, Weizmann Institute of Science, Rehovot 76100, Israel

Communicated by Roger D. Kornberg, Stanford University School of Medicine, Stanford, CA, November 7, 2007 (received for review December 22, 2006) Posttranscriptional regulation is an important step in the regula- Another posttranscriptional regulatory process involves the tion of gene expression. In this article, we show an unexpected RNA surveillance mechanism nonsense-mediated mRNA decay connection between two that participate in different (NMD). This mechanism identifies RNA transcripts harboring processes of posttranscriptional regulation that ensures the pro- premature termination codons (PTC) and brings duction of functional mRNA molecules. Specifically, we show that about their degradation such that their potential toxic effect is the A-to-I RNA editing protein adenosine deaminase that acts on reduced (18). NMD in human cells involves the hUpf proteins RNA 1 (ADAR1) and the human Upf1 (hUpf1) protein involved in (hUpf1, hUpf2, and hUpf3), which together provide substrate RNA surveillance are found associated within nuclear RNA-splicing specificity for the recruitment of mRNA into the NMD pathway complexes. A potential functional role for this association was (19, 20). hUpf1 also participates in a mechanism of degradation, revealed by RNAi-mediated down-regulation of ADAR1, which was termed Staufen1-mediated decay (SMD), which is independent accompanied by up-regulation of a number of genes previously of hUpf2 and hUpf3. In SMD, the protein Staufen1 binds a shown to undergo A-to-I editing in Alu repeats and to be down- 3Ј-UTR bearing a stop codon and recruits hUpf1, leading to the regulated by hUpf1. This study suggests a regulatory pathway by degradation of the mRNA (21). The hUpf1 protein was also a combination of ADAR1 A-to-I editing enzyme and RNA degrada- shown to be involved in the RNA surveillance mechanism tion presumably with the aid of hUpf1. nonsense-associated alternative splicing (NAS) (20, 22). Knock- down of hUpf1 by RNAi and microarray analysis of expressed RNAi ͉ supraspliceosomes ͉ posttranscriptional regulation ͉ genes revealed a large number of genes that are down-regulated nuclear complexes ͉ cross-linking by hUpf1 (23). The present study was motivated by our finding (reported here) that hUpf1 is an integral component of the supraspliceo- NA editing catalyzed by the adenosine deaminase that acts some. This large 21-MDa nuclear ribonucleoprotein complex on RNA (ADAR) family of proteins involves the conversion R (24) has been proposed to constitute the machine where RNA of adenosines to inosines (A to I) and is one of the pre-mRNA splicing occurs in living cells. In addition to its splicing activity processing activities. Of the three identified human ADAR (25), the supraspliceosome harbors other pre-mRNA processing proteins, ADAR1 and ADAR2 are expressed ubiquitously and components including the editing enzymes ADAR1 and have different isoforms resulting from alternative splicing, ADAR2 and the A-to-I editing activity associated with them (26, whereas ADAR3 is expressed at low levels only in the brain 27). We therefore asked whether the hUpf and the ADAR (reviewed in refs. 1 and 2). ADAR1 exists in two major forms proteins, which are involved in apparently distinct RNA pro- expressed from two distinct promoters. The IFN-induced longer cessing functions, interact within the supraspliceosome. In this one is present both in the nucleus and , whereas the study, we show that ADAR1 and hUpf1 coexist in supraspliceo- constitutively expressed shorter one is present in the nucleus (3, 4). somes and in additional nuclear complexes. Our studies suggest ADARs act on double-stranded RNA and deaminate ad- a functional link between ADAR1 and hUpf1 in affecting the enosines at specific sites. Because inosines are generally base- level of a subgroup of edited RNA Pol II transcripts. paired to cytidines, specific A-to-I editing can change the coding potential within an ORF, or change splice sites and other control Results elements (1, 2). A functional significance of the specific editing hUpf1 Is Associated with Supraspliceosomes. Nuclear pre-mRNAs 2ϩ by ADARs is exemplified by the dramatic decrease of Ca together with all pre-mRNA processing components are pack- permeability of the AMPA channel by editing of the Q/R site of aged in supraspliceosomes that represent the native pre-mRNA GluR-B subunit (5). Significant changes in the G protein cou- processing machine (26–29). These complexes contain all five pling efficiency of 5-HT2CR have also been reported (6, 7). spliceosomal U small nuclear ribonucleoproteins (snRNPs) (25), However, in many other substrates, A-to-I editing was found as well as splicing factors such as the SR protein family (30), and clustered in noncoding regions mainly in Alu repetitive elements the ADAR A-to-I RNA editing enzymes (26). Because hUpf1 (8–11), as expected from the identification of a large number of protein was shown to be involved in NAS (20), we reasoned that inosines in mRNA molecules (12). The functional significance of it might be associated with supraspliceosomes. To search for the latter editing is not yet fully understood. Some of the editing such an association HeLa cells nuclear supernatant (NS) en- in noncoding regions was suggested as part of a protection riched for supraspliceosomes was fractionated in a sucrose mechanism of mRNA molecules against RNAi-like degradation (13). ADARs were also shown to bind siRNA and were thus proposed to protect mRNA molecules from RNAi-like degra- Author contributions: L.A., J.S., and R.S. designed research; L.A. and O.R. performed research; L.A. and R.S. analyzed data; and L.A., J.S., and R.S. wrote the paper. dation (14). However, double-stranded RNA molecules with The authors declare no conflict of interest. repeating U-I base pairs undergo degradation mediated by ‡To whom correspondence should be addressed at: Department of Genetics, The Hebrew Tudor, one of the RNA-induced silencing complex (RISC) University of Jerusalem, Edmond Safra Givat-Ram Campus, Jerusalem 91904, Israel. E-mail: components (15). Pan-editing by the IFN-induced ADAR1 was [email protected]. proposed as part of the antiviral protection mechanisms (3). Also This article contains supporting information online at www.pnas.org/cgi/content/full/ pan-editing by ADARs can lead to nuclear retention of the RNA 0710576105/DC1. molecule (16, 17). © 2008 by The National Academy of Sciences of the USA

5028–5033 ͉ PNAS ͉ April 1, 2008 ͉ vol. 105 ͉ no. 13 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0710576105 Downloaded by guest on September 23, 2021 Fig. 2. hUpf1 is associated with nuclear complexes together with ADAR1. (A) Fig. 1. hUpf1 is associated with supraspliceosomes. (A) hUpf1 sediments Nuclear complexes sedimenting at 70S were immunoprecipitated by anti- with supraspliceosomes. Supraspliceosomes prepared from HeLa cell nuclei as hUpf1 antibodies, and the precipitated and unbound proteins were analyzed described in refs. 28 and 29 were fractionated in a sucrose gradient. The 200S by SDS/PAGE and probed by anti-ADAR1 antibodies (lanes 1 and 4, respec- fractions were refractionated in a second sucrose gradient. Aliquots from each tively). Controls: lane 2, no antibody; lane 3, no sample. (B) Nuclear complexes fraction were run on an SDS/PAGE and were Western blotted with anti-Sm, sedimenting at the top of the gradient were immunoprecipitated as described anti-hUpf1, and anti-ADAR1 antibodies. The sedimentation of 200S TMV and in A (lanes in B as in A). (C) HeLa cell NE and nuclear complexes sedimenting 70S bacterial size markers are indicated above the top row. (B) at the top of the gradient at 70S and at 200S were immunoprecipitated by a Indirect IP of hUpf1 and ADAR1 from supraspliceosomes. Supraspliceosomes nonrelevant antibody, preimmune anti-rabbit IgG, and were Western blotted were immunoprecipitated by anti-Sm antibodies, and the precipitated and by anti-ADAR1 (Left) and anti-hUpf1 (Right) antibodies. unbound proteins were analyzed by SDS/PAGE and probed by anti-hUpf1 (Left) and ADAR1 (Right) antibodies (lanes 1 and 4, respectively). As controls, we show the reaction without the antibody (lane 2) and the reaction with region of the gradient (26), where native spliceosomes sediment antibodies and buffer instead of the sample (lane 3). (25, 32). The finding that hUpf1 also appears with complexes at the 70S fraction of the first sucrose gradient (data not shown), gradient as previously described in refs. 28 and 29, and supraspli- raised the question whether hUpf1 is associated with ADAR1 in

ceosomes sedimenting at the 200S region of the gradient were these complexes. ADAR1 (26) and hUpf1 (data not shown) were BIOCHEMISTRY collected and refractionated in a second gradient. We then checked by Western blotting for the presence of hUpf1 in fractions across the gradient. As shown in Fig. 1A, hUpf1 sediments at the 200S region of the gradient together with Sm proteins, which were previously shown to be associated with supraspliceosomes by cosedimentation analysis (25, 31) and by coimmunoprecipitation (co-IP) assays (31). ADAR1 also sedi- ments at the 200S region of this gradient as previously shown (26). Because hUpf1 was shown to be in a complex with hUpf2 and hUpf3 (19), we searched for their presence in the supraspli- ceosome region. However, neither could be detected cosedi- menting with supraspliceosomes in this gradient, whereas they could be detected in NS by the respective antibodies [supporting information (SI) Fig. 8]. To demonstrate that hUpf1 is an integral component of supraspliceosomes, we performed indirect immunoprecipitation (IP) of supraspliceosomes with monoclonal antibodies against Sm proteins (Fig. 1B). The precipitated proteins and the un- bound fraction were electrophoresed by SDS/PAGE and West- ern blotted by anti-hUpf1 antibodies (Fig. 1B Left, lanes 1 and 4, respectively). As controls, we show a mock reaction without the antibody (lane 2), and a mock reaction without the sample (lane 3). We found that hUpf1 was indeed specifically precipi- Fig. 3. ADAR1 and hUpf1 are associated in HeLa cell NE. (A) HeLa cell NE was tated by the anti-Sm antibodies (Fig. 1B Left; IP yield of 16%), immunoprecipitated by anti-ADAR1 antibodies, and the precipitated and thus confirming the association of hUpf1 with supraspliceo- unbound proteins were analyzed by SDS/PAGE and probed by anti-hUpf1 somes. Notably, probing the same blot with antibodies against antibodies (lanes 1 and 4, respectively). Controls: lane 2, no antibody; lane 3, ADAR1 revealed that ADAR1 is also associated with supraspli- no sample. (B) In a complementary experiment, we show IP of HeLa cell NE by ceosomes (Fig. 1B Right). anti-hUpf1 antibodies and Western blotting by anti-ADAR1 antibodies. (C) HeLa cell NE was subjected to RNase A treatment and was then immunopre- cipitated by anti-hUpf1 antibodies and Western blotted by ADAR1. The lanes hUpf1 and ADAR1 Coexist in Nuclear Complexes. ADAR1 was correspond to those in A.(D) The same as in C, except that the NE was treated previously shown to be associated not only with supraspliceo- with RNase V1. (E) RT-PCR of actin RNA extracted from the HeLa cell NE treated somes, but also with nuclear complexes sedimenting at the 70S and untreated with RNase A (Upper) and RNase V1 (Lower) shown in C and D.

Agranat et al. PNAS ͉ April 1, 2008 ͉ vol. 105 ͉ no. 13 ͉ 5029 Downloaded by guest on September 23, 2021 stranded RNA (Fig. 3D). Extraction of RNA followed by analysis of actin RNA by RT-PCR showed that practically all RNA was degraded by the respective RNase treatment (Fig. 3E). Because both RNase treatments did not interfere with the precipitation of ADAR1 by the hUpf1 antibodies (Fig. 3 C and D), we can conclude that RNA is not required for the ADAR1-hUpf1 association.

Association of ADAR1 and hUpf1: Cross-Linking Experiments. To further support the interaction between ADAR1 and hUpf1, we incubated a HeLa cell NE with the chemical cross-linker di- methyl suberimidate (DMS) and analyzed the protein products Fig. 4. Chemical cross-linking results in formation of oligomers containing ADAR1 and hUpf1. HeLa cell NE was cross-linked by DMS for increasing lengths by SDS/PAGE and Western blotting. Incubation of the NE with of time (5–60 min). Aliquots were Western blotted with either anti-ADAR1 the cross-linker for increasing time caused the disappearance of antibodies (A) or anti-hUpf1 antibodies (B). Non-cross-linked samples (Ϫ) are the monomeric forms of ADAR1 (Fig. 4A) and hUpf1 (Fig. 4B), shown for comparison. and the concomitant appearance of higher oligomeric forms containing ADAR1 and hUpf1. A high-molecular-mass oligo- meric form, which seems to contain both ADAR1 and hUpf1 also found at the top fraction of the gradient, where small (Fig. 4 A and B), has an apparent molecular mass of Ϸ350 kDa complexes or individual components sediment. We therefore (as determined by molecular mass markers in the range of performed additional IP experiments, this time on the 70S and 75–500 kDa). The cross-linking experiment was also performed top fractions of a first gradient, and found that in IP with after RNase treatment (either RNase A or RNase V1). In both anti-hUpf1 antibodies ADAR1 was precipitated from the 70S cases, even though the RNase treatment was effective, the fraction (14%) and to a lower extent (7%) from the top fraction cross-linking pattern did not change (SI Fig. 9). (Fig. 2 A and B, respectively). These results indicate an associ- To further characterize the high-molecular-weight cross- ation between the ADAR1 and hUpf1 proteins. Co-IP experi- linked forms that contain ADAR1 and hUpf1, we used a ments on supraspliceosomes performed with anti-hUpf1 anti- reversible chemical cross-linker and 2D gel analysis. We incu- bodies were not efficient enough to be conclusive, most probably bated a HeLa cell NE with dimethyl-3,3Ј-dithiobispropionimi- because of hindrance of the hUpf1 antigenic determinants within date (DTBP) and analyzed the sample on an SDS/PAGE. We the 21-MDa supraspliceosome. As a control for the IP experi- then reversed the cross-linking by reducing the disulfide bond of ments, we also performed IP with a nonrelevant antibody DTBP with ␤-mercaptoethanol, and electrophoresed the sample (preimmune anti-rabbit IgG) on the 200S, 70S, and top fractions on a second dimension SDS/PAGE, followed by Western blot- of a first gradient, and on a HeLa cell nuclear extract (NE). The ting with anti-ADAR1 and anti-hUpf1 antibodies. This way, results show that IP with antibodies against ADAR1, hUpf1, and proteins that are not cross-linked run the same way in both gels Sm were specific (Fig. 2C). and create a diagonal, whereas cross-linked proteins run as such To further test the association of ADAR1 and hUpf1, we on the first gel but as non-cross-linked products on the second performed IP on a HeLa cell NE by using antibodies against gel (because of the cleavage of the cross-linker) and are there- ADAR1. Fig. 3A shows that hUpf1 was precipitated with a yield fore found off the diagonal. Proteins that had been cross-linked of 23%. In a reciprocal experiment, in which antibodies against together appear off diagonal on the same vertical line. As can be hUpf1 were used for IP and anti-ADAR1 antibodies were used seen in Fig. 5, there are several off-diagonal bands corresponding for Western blotting, ADAR1 was immunoprecipitated with a to both ADAR1 (Fig. 5B) and hUpf1 (Fig. 5C), which were yield of 20% for the 110-kDa form and of 1% for the 150-kDa released from the cross-linked species by the reducing agent. For form (Fig. 3B). Both antibodies are specific and no cross- comparison, we show Western blots with ADAR1 of a first- reactivity was observed. We can therefore conclude that hUpf1 dimension gel of a non-cross-linked NE, and a cross-linked NE is associated with ADAR1. without treatment by the reducing agent (Fig. 5A). From the We next tested whether this association depends on RNA, by merge of Fig. 5 B and C (Fig. 5D), it can be seen that both analyzing the effect of RNase treatment on the association as ADAR1 and hUpf1 were cross-linked to generate a complex seen by IP. First, we incubated a HeLa cell NE with RNase A, with an apparent molecular mass of Ϸ350 kDa. On reversal of which cleaves single-stranded RNA, and then performed IP by the cross-linking, both ADAR1 and hUpf1 appear on the same using antibodies against hUpf1 (Fig. 3C). An analogous exper- vertical line, indicating that they have been released from the iment was performed with RNase V1, which cleaves double- same high-molecular-weight complex. We also show that run-

Fig. 5. Association of ADAR1 and hUpf1 is revealed by reversible cross-linking and analysis on 2D gels. HeLa cell NE was cross-linked by the reversible cross-linker DTBP for 20 min. (A) First-dimension SDS/PAGE analysis of untreated NE (Ϫ) and cross-linked NE (ϩ). (B and C) The gel was soaked in 3% ␤-mercaptoethanol and electrophoresed on a second identical gel. Western blotting was performed with anti-ADAR1 and anti-hUpf1 as marked on the right. (D) A merge of B and C. The arrows mark the direction of migration of the first and second dimension of the gel.

5030 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0710576105 Agranat et al. Downloaded by guest on September 23, 2021 ning a 2D gel as described above without pretreatment of the sample with DTBP does not result in off-diagonal signals (SI Fig. 10).

Knockdown of ADAR1 Results in Up-Regulation of Genes Edited by ADARs and Regulated by hUpf1. We then asked whether the apparent association between ADAR1 and hUpf1 has functional consequences. A clue for such a connection came from the observations that hUpf1 can also participate in an RNA degra- dation pathway distinct from NMD (21) and that its down- regulation results in up-regulation of the expression of many genes as detected by microarray analysis (23). We therefore searched for candidate genes that can potentially be degraded by a pathway involving both ADAR1 and hUpf1. For this aim, we compared the list of genes that are up-regulated after the knockdown of hUpf1 (23), with the list of genes that undergo editing according to bioinformatic analyses (8). Of the 13 genes that appeared in both lists, we chose the following six: MAP3K14 (mitogen-activated protein kinase kinase kinase), TXNRD2 (thioredoxin reductase), MGC10471 (hypothetical protein), PISD (phosphatidylserine decarboxylase), DAP3 (death- associated protein 3), and SARS (seryl-tRNA synthetase). RNAi of hUpf1 performed according to Mendell et al. (23), confirmed that the above six genes are down-regulated by hUpf1 (data not shown). We then performed RNAi of ADAR1 in HeLa cells, by using the siRNA duplexes described in ref. 33. As controls, we analyzed untreated cells and cells transfected with siRNA duplex against firefly luciferase, as described in ref. 34. Western blot and RT-PCR analyses (Fig. 6 A and B, respectively) show that the siRNA treatment down-regulated the level of ADAR1 protein to Ϸ20% and that of the RNA almost to completion. We used endogenous actin as an internal standard. Of the six candidate genes we found four (PISD, MGC10471, DAP3, and SARS) (Fig. 6B) that were clearly up-regulated after

RNAi treatment of ADAR1, compared with the level of RNA BIOCHEMISTRY expressed in control cells (Ϸ3-fold compared with the level in cells treated with RNAi of luciferase). TXNRD2 was also up-regulated after RNAi of ADAR1. At present, it is not added to the list, because a mild increase in RNA level was seen after treatment with RNAi of luciferase (Fig. 6B). MAP3K14 did not seem to be affected (Fig. 6B). As controls, we show that the RNA level of transcripts of two genes that are not known to be affected Fig. 6. Up-regulation of MGC10471, DAP3, PISD, and SARS mRNAs by RNAi by RNAi of ADAR1, namely SDHA and hUpf1, was not of ADAR1. HeLa cells were either transfected with siRNA against ADAR1, changed (Fig. 6B). These results suggest that ADAR1 partici- siRNA against firefly luciferase as a negative control or not transfected at all. After 72 h, total RNA and total proteins were prepared. (A) Western blot pates in an RNA degradation mechanism, possibly with the aid analyses of a total protein preparation extracted from the above treated cells of hUpf1. To strengthen the above results, we performed show specific down-regulation of ADAR1 (Ϸ20%). (B)(Right) RT-PCR analyses half-life experiments by using actinomycin D (Act.D). HeLa cells of actin, ADAR1, MGC10471, DAP3, PISD, SARS, MAP3K14, TXNRD2, SDHA, were treated with either siRNA against ADAR1 or mock- and hUpf1 RNAs extracted from HeLa cells treated with siRNA against ADAR1 treated. After 72 h, Act.D was added to the medium and RNA (lane 1), luciferase (lane 2), and from untreated cells (lane 3). (Left) The image was extracted at time points as indicated. RT-PCR analyses show represents 3-fold serial dilutions of RNA from untreated cells to demonstrate that the gene transcripts of PISD, MGC10471, DAP3, and that the RT-PCR is semiquantitative. SARS, which were seen to be down-regulated by ADAR1 by semiquantitative RT-PCR analyses (Fig. 6), were also degraded spectively) did not affect it. ADAR1 and hUpf1 are found more rapidly in the presence of ADAR1 than when it was down-regulated (Fig. 7). Furthermore, the transcript of associated in the nucleus within supraspliceosomes, as well as MAP3K14, which was not affected by the ADAR1 siRNA within 70S nuclear complexes, which correspond to the native treatment, did not show any difference in its degradation pattern spliceosomes described by Azubel et al. (25, 32). The cross- between control and ADAR1 knocked-down cells (Fig. 7). linking experiments (Figs. 4 and 5) show that on reversing the cross-links ADAR1 and hUpf1 are released from an oligomeric Discussion complex of an apparent molecular mass of at least 350 kDa. ADAR1 Is Associated with hUpf1 in the Nucleus. By performing Although the accurate mass of the apparent ADAR1–hUpf1 cross-linking and co-IP experiments of nuclear complexes frac- complex has yet to be determined, the fact that ADAR1 was tionated in density gradients and of HeLa cell nuclear extract, we previously shown to form dimers (35, 36), is consistent with a showed that the RNA editing enzyme ADAR1 is associated with complex containing at least one copy of hUpf1 and one of the RNA surveillance protein hUpf1. This association is not ADAR1, or possibly, an ADAR1 dimer. These experiments dependent on RNA, because treatment with either a single- or indicate that hUpf1 and ADAR1 are likely to interact with one double-stranded specific RNase (RNase A or RNase V1, re- another.

Agranat et al. PNAS ͉ April 1, 2008 ͉ vol. 105 ͉ no. 13 ͉ 5031 Downloaded by guest on September 23, 2021 Fig. 7. Increase in half-lives of MGC10471, DAP3, PISD, and SARS mRNAs after RNAi of ADAR1. HeLa cells were either mock-transfected or transfected with siRNA against ADAR1. After 72 h, Act.D was added to the cells for the indicated time. RT-PCR of actin, ADAR1, MGC10471, DAP3, SARS, PISD, and MAP3K14 RNAs was performed after RNA extraction. (A) RT-PCR analyses. (Left) ADAR1 siRNA. (Right) Mock transfection. (B) cDNA levels of the different transcripts were normalized to actin levels and then were normalized to the normalized level at time 0 (y axis). The equations of the linear functions are given, as is the R2. Results are representative of three independent experiments. The standard deviation of the slopes ranges from 0.003 to 0.007.

A Potential Functional Role for the Association of ADAR1 and hUpf1. In conclusion, we show a connection between ADAR1 and The finding that ADAR1 and hUpf1 coexist in nuclear com- hUpf1 and a down-regulation of genes by ADAR1. Our finding plexes is interesting in the sense that, although both ADARs and thus demonstrates once more how closely integrated are the hUpfs are known to be involved in the processing of pre-mRNAs apparently independent pathways that ensure the production of to functional mRNA molecules, the pathways in which they mRNA molecules that are adequate for translation of proteins. participate have not been connected so far. A clue for a possible functional role of the association of Materials and Methods ADAR1 and hUpf1 emerges from the observations that hUpf1 Preparation of Splicing Complexes and Supraspliceosomes. Supraspliceosomes is responsible for the down-regulation of many mRNA molecules were prepared from HeLa cells (CILBIOTECH) as described in refs. 28 and 29. For (23), and that it is also involved in an mRNA degradation details, see SI Materials and Methods. pathway independent of NMD (21). We have thus reasoned that Western Blot Analysis. For Western blot analysis, aliquots were applied directly ADAR1 and hUpf1 could participate in concert in a mechanism to the wells ofa8or6%SDS/PAGE gel and analyzed as described in ref. 38, of RNA degradation. To test this possibility, we searched for using antibodies against hUpf1 [kindly provided by J. Lykke-Andersen (Uni- genes that are both down-regulated by hUpf1 (23) and undergo versity of Colorado, Boulder, CO), and J. Mendell (Johns Hopkins University, A-to-I editing by ADAR enzymes (8). We found 13 such genes Baltimore, MD)], ADAR1 [mAb 15.8.6 kindly provided by K. Nishikura (Univer- [6.6% of the smaller list (23)] and tested the effect of RNAi of sity of Pennsylvania, Philadelphia, PA)], Sm proteins (mAb Y12) (39), and actin ADAR1 on the expression of six of them: DAP3, PISD, (I19; Santa Cruz). For details, see SI Materials and Methods. MGC10471, SARS, MAP3K14, and TXNRD2. These genes undergo editing in introns or in UTRs in Alu elements (8), short Chemical Cross-Linking. HeLa cell NE (CILBIOTECH) was mixed with aliquots of Ϸ the cross-linker dimethyl suberimidate dihychloride (DMS) (Pierce) to final interspersed repetitive elements of 280 bp, widespread in the concentrations of 13% vol/vol and 10 mM, respectively, in 0.2 M triethanol- Ϸ 6 human genome ( 10 copies per haploid genome) and mainly amine, pH 8, and incubated at 20°C for 5, 10, 20, 40, and 60 min. The reaction found in introns (37). Our results showed that DAP3, PISD, was stopped by the addition of 1.5 vol of SDS/PAGE sample buffer (50 mM SARS, and MGC10471 were up-regulated and had longer half- Tris⅐HCl, pH 6.8, 200 mM DTT, 4% SDS, 0.1% bromophenol blue, and 10% lives after RNAi of ADAR1, whereas MAP3K14 was not glycerol) and incubation on ice. The samples were then boiled for 5 min and affected by down-regulation of ADAR1, neither in its expression run on an SDS/PAGE. The molecular mass markers (75–500 kDa) used were as level nor in its half-life. These results are therefore in support of follows: Rainbow RPN 800 (Amersham Biosciences), HiMark LC5699 (Invitro- a mechanism by which a subgroup of A-to-I edited genes are gen), and PageRuler SM0671 (Fermentas). down-regulated by ADAR1, possibly with the aid of hUpf1. Reversible Cross-Linking and 2D Gel Analysis. To HeLa cell NE (95% vol/vol final Interestingly, testing target genes by using the AceView gene concentration), aliquots of the solution of the reversible cross-linker DTBP models with alternative splicing database (National Center for (Pierce) in 0.2 M triethanolamine, pH 8, were added to a final concentration Biotechnology Information), we found that the edited sites of of 5 mM, and the solution was incubated at 20°C for 20 min. The reaction was DAP3, PISD, and MGC10471 transcripts are located within, or stopped by the addition of 1.1 vol of gel sample buffer (as above, but lacking next to, a putative alternatively spliced RNA with an ORF having DTT) and incubation on ice. The sample was then boiled for 5 min and run on a stop codon. We did not find such a potential stop codon in an SDS/PAGE. For reversal of the cross-linking, by the reduction of the disulfide MAP3K14. This observation, by analogy to NMD and SMD, bond of DTBP, the analyzed lane was cut from the rest of the gel, and incubated for 15 min at 65°C with 3% (vol/vol) ␤-mercaptoethanol in 50 ml of may provide a possible explanation for the down-regulation by the SDS/PAGE running buffer (25 mM Tris, 250 mM glycine, 0.1% SDS) adjusted ADAR1 of only a subset of ADAR substrates. Further analyses to pH 6.8. Next, the treated lane was incubated for 30 min at 20°C in 50 ml of are required to test this hypothesis, and other explanations the SDS/PAGE running buffer adjusted to pH 6.8, placed horizontally on top of cannot be excluded at this stage. a second gel, and run on an SDS/PAGE for the second dimension.

5032 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0710576105 Agranat et al. Downloaded by guest on September 23, 2021 Immunoprecipitations. Indirect IPs were performed as described in ref. 30 by experiments. The percentage of ADAR1 protein and the effect of ADAR1 using anti-ADAR1 WI9 antibodies (kindly provided by K. Nishikura), anti- siRNA on the levels of the tested RNAs was calculated after normalization to hUpf1 antibodies, or anti-Sm mAb Y12 or a nonrelevant antibody, preimmune the level of actin and then to the levels in the siRNA luciferase-transfection anti-rabbit IgG (see SI Materials and Methods). IP yield is expressed as the ratio lane. Band intensities were measured by using the Image Gauge program. between the intensity of the immunoprecipitated band [minus background of control lanes (lanes 2 and 3)] and the sum of this intensity and that of the RT-PCR. RT-PCR was performed as described in ref. 40. Details and primer list corresponding band in the unbound lane. Intensities were measured by the are published as SI Materials and Methods. Image Gauge program. Half-Life Experiments. HeLa cells were either treated with siRNA against ␮ RNase Assays. HeLa cell NE was incubated with either RNase A at 2 g/ml for ADAR1 or mock-treated as described above. After 72 h, Act.D was added to 1 ␮ 1 h at 4°C, or with RNase V1 (Ambion) at 0.01 units/ l for 18 h at 4°C. RNA was ␮g/ml medium. RNA was extracted at time points as indicated. RNA was also ␮ extracted from 10- l aliquots of treated and untreated NEs, and the remain- extracted from cells that were not treated with Act.D. RT-PCR analyses were ing reaction mixtures were taken either for cross-linking experiments or for performed as described above. Band intensities were measured by using the immunoprecipitation as described above (cross-linking time was 20 min). Image Gauge program.

RNAi of ADAR1. siRNAs targeted to (i) the long and the short forms of ADAR1 ACKNOWLEDGMENTS. We thank Aviva Pecho for excellent technical assis- (33), and (ii) to the firefly’s luciferase mRNA (34) were purchased from Dhar- tance; Erez Levanon for helpful information; and Drs. Joshua Mendell, Kazuko macon and transfected into HeLa cells as described in ref. 40, with some Nishikura, and Jens Lykke-Andersen for antibodies. This work was supported modifications. Cells grown in 6-cm plates were transfected with 125 nM siRNA in part by grants from the Israel Science Foundation (to R.S.) and from The by using 0.23% oligofectamine (Invitrogen). After 72 h, total proteins and Helen and Milton Kimmelman Center for and Assem- RNA were extracted. The results are representative of three independent bly at The Weizmann Institute of Science (to J.S.).

1. Bass BL (2002) RNA editing by adenosine deaminases that act on RNA. Annu Rev 22. Wang J, Hamilton JI, Carter MS, Li S, Wilkinson MF (2002) Alternatively spliced TCR Biochem 71:817–846. mRNA induced by disruption of reading frame. Science 297:108–110. 2. Valente L, Nishikura K (2005) ADAR gene family and A-to-I RNA editing: Diverse roles 23. Mendell JT, Sharifi NA, Meyers JL, Martinez-Murillo F, Dietz HC (2004) Nonsense in posttranscriptional gene regulation. Prog Nucleic Acid Res Mol Biol 79:299–338. surveillance regulates expression of diverse classes of mammalian transcripts and 3. Patterson JB, Samuel CE (1995) Expression and regulation by interferon of a double- mutes genomic noise. Nat Genet 36:1073–1078. stranded-RNA-specific adenosine deaminase from human cells: Evidence for two forms 24. Muller S, Wolpensinger B, Angenitzki M, Engel A, Sperling J, et al. (1998) A supraspli- of the deaminase. Mol Cell Biol 15:5376–5388. ceosome model for large nuclear ribonucleoprotein particles based on mass determi- 4. George CX, Samuel CE (1999) Human RNA-specific adenosine deaminase ADAR1 nations by scanning transmission electron microscopy. J Mol Biol 283:383–394. transcripts possess alternative exon 1 structures that initiate from different promoters, 25. Azubel M, Habib N, Sperling J, Sperling R (2006) Native spliceosomes assemble with one constitutively active and the other interferon inducible. Proc Natl Acad Sci USA pre-mRNA to form supraspliceosomes. J Mol Biol 356:955–966. 96:4621–4626. 26. Raitskin O, Cho DS, Sperling J, Nishikura K, Sperling R (2001) RNA editing activity is 5. Sommer B, Kohler M, Sprengel R, Seeburg PH (1991) RNA editing in brain controls a associated with splicing factors in lnRNP particles: The nuclear pre-mRNA processing determinant of ion flow in glutamate-gated channels. Cell 67:11–19. machinery. Proc Natl Acad Sci USA 98:6571–6576. 6. Burns CM, et al. (1997) Regulation of serotonin-2C receptor G-protein coupling by RNA 27. Raitskin O, Angenitzki M, Sperling J, Sperling R (2002) Large nuclear RNP particles—the editing. Nature 387:303–308. nuclear pre-mRNA processing machine. J Struct Biol 140:123–130. 7. Wang Q, et al. (2000) Altered G protein-coupling functions of RNA editing isoform and 28. Sperling R, Sperling J (1998) in RNP Particles, Splicing and Autoimmune Diseases,ed splicing variant serotonin 2C receptors. J Neurochem 74:1290–1300. Schenkel J (Springer, Berlin), pp 29–47. 8. Levanon EY, et al. (2004) Systematic identification of abundant A-to-I editing sites in 29. Spann P, Feinerman M, Sperling J, Sperling R (1989) Isolation and visualization of large the human transcriptome. Nat Biotechnol 22:1001–1005. compact ribonucleoprotein particles of specific nuclear RNAs. Proc Natl Acad Sci USA BIOCHEMISTRY 9. Athanasiadis A, Rich A, Maas S (2004) Widespread A-to-I RNA editing of Alu-containing 86:466–470. mRNAs in the human transcriptome. PLoS Biol 2:e391. 30. Yitzhaki S, Miriami E, Sperling R, Sperling J (1996) Phosphorylated Ser/Arg-rich pro- 10. Kim DD, et al. (2004) Widespread RNA editing of embedded alu elements in the human teins: Limiting factors in the assembly of 200S large nuclear ribonucleoprotein parti- transcriptome. Genome Res 14:1719–1725. cles. Proc Natl Acad Sci USA 93:8830–8835. 11. Morse DP, Aruscavage PJ, Bass BL (2002) RNA hairpins in noncoding regions of human 31. Sperling R, Spann P, Offen D, Sperling J (1986) U1, U2, and U6 small nuclear ribonucleo- brain and mRNA are edited by adenosine deaminases that act proteins (snRNPs) are associated with large nuclear RNP particles containing transcripts on RNA. Proc Natl Acad Sci USA 99:7906–7911. of an amplified gene in vivo. Proc Natl Acad Sci USA 83:6721–6725. 12. Paul MS, Bass BL (1998) Inosine exists in mRNA at tissue-specific levels and is most 32. Azubel M, Wolf SG, Sperling J, Sperling R (2004) Three-dimensional structure of the abundant in brain mRNA. EMBO J 17:1120–1127. native spliceosome by cryo-electron microscopy. Mol Cell 15:833–839. 13. Knight SW, Bass BL (2002) The role of RNA editing by ADARs in RNAi. Mol Cell 33. Wong SK, Lazinski DW (2002) Replicating hepatitis delta virus RNA is edited in the 10:809–817. nucleus by the small form of ADAR1. Proc Natl Acad Sci USA 99:15118–15123. 14. Yang W, et al. (2005) ADAR1 RNA deaminase limits short interfering RNA efficacy in 34. Elbashir SM, et al. (2001) Duplexes of 21- RNAs mediate RNA interference in mammalian cells. J Biol Chem 280:3946–3953. cultured mammalian cells. Nature 411:494–498. 15. Scadden AD (2005) The RISC subunit Tudor-SN binds to hyper-edited double-stranded 35. Cho DS, Yang W, Lee JT, Shiekhattar R, Murray JM, et al. (2003) Requirement of RNA, promotes its cleavage. Nat Struct Mol Biol 12:489–496. dimerization for RNA editing activity of adenosine deaminases acting on RNA. J Biol 16. Zhang Z, Carmichael GG (2001) The fate of dsRNA in the nucleus: A p54(nrb)-containing Chem 278:17093–17102. complex mediates the nuclear retention of promiscuously A-to-I edited RNAs. Cell 36. Gallo A, Keegan LP, Ring GM, O’Connell MA (2003) An ADAR that edits transcripts 106:465–475. encoding ion channel subunits functions as a dimer. EMBO J 22:3421–3430. 17. Prasanth KV, et al. (2005) Regulating gene expression through RNA nuclear retention. 37. Jurka J (2004) Evolutionary impact of human Alu repetitive elements. Curr Opin Genet Cell 123:249–263. Dev 14:603–608. 18. Maquat LE (2002) Nonsense-mediated mRNA decay. Curr Biol 12:R196–R197. 38. Yitzhaki S, Sperling J (1998) Concentrating dilute protein solutions for gel electro- 19. Lykke-Andersen J, Shu MD, Steitz JA (2000) Human Upf proteins target an mRNA for phoresis. Biotechniques 24:762–764,766. nonsense-mediated decay when bound downstream of a termination codon. Cell 39. Lerner EA, Lerner MR, Janeway CA, Jr, Steitz JA (1981) Monoclonal antibodies to 103:1121–1131. nucleic acid-containing cellular constituents: probes for molecular biology and auto- 20. Mendell JT, ap Rhys CM, Dietz HC (2002) Separable roles for rent1/hUpf1 in altered immune disease. Proc Natl Acad Sci USA 78:2737–2741. splicing and decay of nonsense transcripts. Science 298:419–422. 40. Wachtel C, Li B, Sperling J, Sperling R (2004) Stop codon-mediated suppression of 21. Kim YK, Furic L, Desgroseillers L, Maquat LE (2005) Mammalian Staufen1 recruits Upf1 splicing is a novel nuclear scanning mechanism not affected by elements of protein to specific mRNA 3ЈUTRs so as to elicit mRNA decay. Cell 120:195–208. synthesis and NMD. RNA 10:1740–1750.

Agranat et al. PNAS ͉ April 1, 2008 ͉ vol. 105 ͉ no. 13 ͉ 5033 Downloaded by guest on September 23, 2021