Suboptimal splice sites of equine infectious anaemia virus control Rev responsiveness

In primate lentiviruses, cis-acting repressing sequences (CRS) have been localized to various regions in the gag, pol, env, pro and vif genes and the Rev-responsive element (RRE) (Rosen et al., 1988 ; Maldarelli et al., 1991 ; Schwartz et al., 1992a , b ; Huffman & Arrigo, 1997 ). These inhibitory sequences were suggested to function by preventing the transport of unspliced and partially spliced mRNAs into the cytoplasm. Rev, as a nuclear RNA export factor, was found to relieve this suppression (reviewed in Cullen, 1998 ). The role of functional splice sites in conferring Rev dependence remains a point of intense debate (reviewed in Kingsman & Kingsman, 1996 ).

The Rev protein of equine infectious anaemia virus (EIAV), the activation domain of which is functionally interchangeable with the leucine-rich activation domains of other lentiviral Rev proteins (Fridell et al., 1993 ; Mancuso et al., 1994 ; Meyer et al., 1996 ), was also shown to augment the expression of transcripts containing introns (Martarano et al., 1994 ). Moreover, although the EIAV Rev activation domain possesses an atypical nuclear export signal that differs significantly from that of human immunodeficiency virus (HIV) Rev, it is believed to travel the same nuclear export signal-dependent export pathway (Otero et al., 1998 ). In primate lentiviruses (Malim et al., 1988 ; Rosen et al., 1988 ) and several of the ungulate lentiviruses (Schoborg & Clements, 1996 ) such as visna-maedi virus and caprine arthritis encephalitis virus, the Rev protein acts through binding the RRE located within the env gene. In the EIAV genome, however, computer analysis could not identify a stable stemloop RNA secondary structure analogous to the lentiviral RRE (Rosin-Arbesfeld et al., 1993 ; Martarano et al., 1994 ). Previously, we (Rosin-Arbesfeld et al., 1993 ) and others (Harris et al., 1998 ) showed that the 3 half of the EIAV genome sequence conferred Rev dependence on CAT-encoding transcripts. In another study, it was suggested that the responsiveness to EIAV Rev is mediated via two elements located in the env gene (Martarano et al., 1994 ). Thus, by analogy with other lentiviruses (Cullen, 1998 ), it could be assumed that the 3 half of the EIAV RNA contains sequences that confer cis-repressing dependence on EIAV unspliced and partially spliced transcripts and that this repression might be relieved by Rev. The present study suggests that Rev dependence of EIAV is mediated by several of the suboptimal, centrally located viral splice sites.

Plasmid construction.
The Rev-encoding cDNA p176 (designated pCEV176), the vector pCEV21 (designated pCEV) and the CAT reporter plasmid (pCEVCAT), containing nt 44737247 of the EIAV genome downstream of the CAT coding region, were described previously (Rosin-Arbesfeld et al., 1993 ). To construct the CAT reporter plasmids, various regions that span nt 44796162 were amplified by PCR as described previously (Rosin-Arbesfeld et al., 1998 ), using the cloned EIAV genome as a template (Fig. 2) and the oligonucleotide primers detailed in Table 1. The various PCR products, which are flanked by BglII and ApaI restriction sites, were inserted downstream of the CAT coding region within the CAT reporter plasmid pCEVCAT. pCAT/2/2 was constructed by inserting a PCR fragment that spanned exon regions from nt 406 to 7236 downstream of cat and was amplified by using pCEV/2/2 Tat cDNA (Rosin-Arbesfeld et al., 1993 ) as a template. In vitro-mutagenized CAT reporter plasmids (Fig. 3) were constructed by PCR by using mutagenizing oligonucleotides (Table 2) as described previously (Rosin-Arbesfeld et al., 1998 ). For each mutation, two separate PCRs were performed, one with a sense mutagenizing primer (Table 2) and an antisense primer, 5700(as) (Table 1), and the other with an antisense complementary mutagenizing primer (Table 2) and a sense primer, 5111(s) (Table 1). The two mutated PCR products were pooled together and served as a template for a third PCR, performed in the presence of primers 5111(s) and 5700(as). The resulting BglII/ApaI-flanked PCR products were inserted downstream of the CAT coding region.

(33K): Fig. 2. Localization of CRS and RRE. The second and third exons of the tat cDNA (Noiman et al., 1991 ) are presented schematically at the top of the figure. The numbers shown by arrows indicate the positions of the 5' and 3' splice sites. The bars indicate the locations of the PCR-amplified fragments present in the reporter constructs downstream of the CAT coding region. In pCAT/2/2, PCR-amplified exon regions are indicated by solid lines, while intron sequences are indicated by dotted lines. Canine cells were co-transfected with the various CAT reporter plasmids in the absence (CRS) or presence (Rev) of the Rev-expressing plasmid pCEV176. CAT activity was determined by the method of Neumann et al. (1987) as described previously (Rosin-Arbesfeld et al., 1993 ). The CAT activity of each construct was compared with that of pCEVCAT (100%). Values are the means±SD of six separate transfection experiments.

Table 1. Oligonucleotides that served as primers for PCR

(21K): Fig. 3. Structure of the various mutated CRS domains. The second and third exons of the tat cDNA (Noiman et al., 1991 ) are shown schematically at the top of the figure with the positions of the various 5' (SD) and 3' (SA) splice sites. The remainder of the figure represents the pCAT/CRS* constructs. Arrows with asterisks indicate the locations of mutated splice sites. Asterisks after construct names indicate mutation of the different splice sites. Cells were co-transfected with the various pCAT/CRS* plasmids and pCEV176 (Rev) or empty vector pCEV (CRS). CAT levels were measured and calculated as described in the legend to Fig. 2.

Table 2. Primers used for splice-site mutagenesis

Cell culture and CAT and RNase protection assays.
Canine thymus cells, CF2Th (Noiman et al., 1991 ), were co-transfected with the various CAT reporter constructs together with the Rev-encoding cDNA p176 (pCEV176) (Rosin-Arbesfeld et al., 1993 ) or pCEV21 (pCEV) as a control and pCMV/β-Gal (Clontech) as a measure of transfection efficiency. After 48 h, the levels of CAT activity were assessed by the methods of Gorman et al. (1982) or Neumann et al. (1987) , as described previously (Rosin-Arbesfeld et al., 1993 ). The RNase protection assay was performed as described previously (Rosin-Arbesfeld et al., 1998 ). CAT and actin riboprobes were synthesized by T7 RNA polymerase by using the pBluescript KS vector (Stratagene). In order to localize specific sequences in the EIAV genome that are involved in the responsiveness to Rev, CAT reporter constructs containing overlapping regions of the 3 half of the EIAV genome were analysed for Rev dependence (Fig. 1). Canine CF2Th thymus cells, which are highly permissive for EIAV replication (Noiman et al., 1991 ), were co-transfected with the various CAT reporter constructs together with the Rev-encoding cDNA p176 (Rosin-Arbesfeld et al., 1993 ) or pCEV21 as a control and the levels of CAT activity were assessed. The results showed that, in contrast to the region that spanned nt 44737247, which enabled the expression of high CAT activity in the presence or absence of Rev, the region spanning nt 44736162 reduced CAT activity when present in cis, and the supply of Rev in trans partially overcame this inhibition (Fig. 1). These data suggested that both CRS and RRE reside within nt 44736162.

(32K): Fig. 1. EIAV domains that are Rev responsive. (a) Organization of the EIAV genome and the two overlapping regions inserted into pCAT/44737247 and pCAT/44736162, downstream of the CAT coding sequences. Numbering of nucleotides is based on that of Kawakami et al. (1987) . LTR, Long terminal repeat. (b) CAT activity of the two reporter constructs pCAT/44737247 and pCAT/44736162, co-transfected with a Rev-expressing plasmid, pCEV176, or with an empty vector, pCEV, was determined by the method of Neumann et al. (1987) as described previously (Rosin-Arbesfeld et al., 1993 ). pCEVCAT served as a control for maximum (100%) CAT activity. , pCEVCAT; , pCAT/44736162 and pCEV176; , pCAT/44736162 and pCEV; , pCAT/44737242 and pCEV176; ", pCAT/44737242 and pCEV. (c) Duplicate samples were subjected to CAT analysis by determining the percentages of [14C]chloramphenicol converted to acetylated derivatives, as described previously (Rosin-Arbesfeld et al., 1993 ). The percentages obtained were: pCEVCAT (lane 1), 84·5; pCAT/44737242 and pCEV (2), 44·2; pCAT/44737242 and pCEV176 (3), 83; pCAT/44736162 and pCEV (4), 9·8; and pCAT/44736162 and pCEV176 (5), 21·4.

The region extending from nt 5135 to 5538 contains suboptimal acceptor and donor splice signals that serve to generate the alternatively spliced EIAV transcripts that encode the viral regulatory proteins, Tat and Rev (Noiman et al., 1991 ). To investigate whether the unused splice sites serve as CRS, CAT reporter plasmids containing various PCR-amplified regions that spanned nt 44796162 were assessed for CAT activity. The results showed that pCAT/44795700, which contained the region containing the five splice junctions SA5135, SA5212, SD5276, SA5437 and SD5538, exhibited a strong CRS effect, as reflected by an 85% reduction in CAT activity (Fig. 2). In contrast, pCAT/44795226, which lacked SD5276, SA5437 and SD5538, pCAT/44795173, which lacked these sites and SA5212, and pCAT/44795130, which lacked all five splice sites, displayed correspondingly decreasing levels of CRS activity (19.7, 22 and 36% CAT levels, respectively). Similarly, pCAT/44796162 displayed strong CRS activity (10% CAT activity), while CRS activity was drastically reduced (47% CAT activity) in pCAT/52216162, which lacked SA5138 and SA5212. In addition, while pCAT/51115700 displayed strong CRS activity (20% CAT activity), further deletion of splice site SA5135 in pCAT/51685700 and SA5135, SA5212 and SD5276 in pCAT/53475700 resulted in corresponding reductions in CRS activity (31 and 53% CAT levels, respectively). Similarly, the CRS activity displayed by pCAT/51115700 was much lower than that exerted by pCAT/51115394, which lacked SA5437 and SD5538 (20% compared with 55% CAT activity). Interestingly, pCAT/56956162, which did not span any splice site, did not exhibit any CRS activity (100% CAT activity). These data suggested that the suboptimal splice junctions might function as CRS. In order to confirm this notion, a reporter CAT plasmid, pCAT/2/2, was constructed by amplification of the exon regions that span nt 406 to 7236 by using pCEV/2/2 Tat cDNA (Rosin-Arbesfeld et al., 1993 ) as a template. The data showed that pCAT/2/2, which lacked intron sequences, did not display any CRS activity (Fig. 2).

We anticipated that, if the splice sites were involved in CRS activity, their elimination by in vitro mutagenesis would hinder the CRS inhibitory effect, resulting in Rev-independent CAT expression. We used mutagenizing oligonucleotides (Fig. 3 and Table 2) to disrupt the intron border dinucleotides, which are critical for the splicing process. First, we showed that disruption of all the splice sites (SA 5212, SD 5276, SA 5437 and SD 5538) in pCAT-CRS*, which lacked SA5135, resulted in complete elimination of CRS activity (Fig. 3). Next, in order to evaluate the contribution of particular splice junctions to the inhibitory activity, we mutated individual splice sites. pCAT/5212* and pCAT/5276*, in which the splice signals SA5212 and SD5276, respectively, were eliminated, showed moderate levels of CRS activity (33 and 49% CAT activity), suggesting that each of these sites contributed slightly to CRS activity. In contrast, mutations that eliminated SA5135 and SA5437 diminished CRS activity significantly (70% and 61% CAT levels, respectively). The most pronounced reduction in CRS activity was obtained when SD5538 was eliminated (85% CAT activity). These data suggested that mainly SD5538, but also SA5135 and SA5437, play a major role in CRS activity. Since disruption of SD5538 reduced CRS activity drastically, it might be suggested that the presence of this site is essential for exerting efficient CRS activity. However, because the individual elimination of each of the other splice junctions also reduced CRS activity partially, it might be suggested that SD5538 is essential, but not sufficient, for conferring full CRS activity. This notion was substantiated by further experimental data that showed that the region that spanned nt 51115394, containing SA5135, SA5212 and SD5276, as well as the region that spanned nt 53475700, containing SA5437 and SD5538, both exhibited only moderate CRS activity (5355% CAT activity) (Fig. 2). Interestingly, the region that spanned nt 44795130 also exhibited CRS activity (Fig. 2). We postulated that this activity could probably be attributed to the non-functional site that, according to sequence data, is present at nt 5095 (Noiman et al., 1990 ). To investigate this possibility, an additional mutated construct was constructed, designated pCAT/5095*, in which SA5095 within pCAT/44795130 was mutated (Table 2). The data showed that elimination of SA5095 reduced CRS activity slightly (Fig. 3; 34% CAT activity exhibited by pCAT/5095* compared with 22% exhibited by pCAT/44795130). These data suggested that, although SA5095 contributed slightly to CRS, the major inhibitory effect exerted by this region was independent of the splicing machinery.

Since the region that spanned nt 44796162 responded to Rev (Fig. 1), experiments were conducted to localize the elements that can act as RRE within this region. The data showed (Fig. 2) that all CAT reporter constructs that contained sequences spanning nt 51116162 (pCAT/51115700, pCAT/51685700, pCAT/53475700, pCAT/51115394 and pCAT/52216162) were responsive to the supply of Rev in trans, resulting in CAT levels of 7280%. This suggested that sequences residing within the region that spanned nt 51116162 can interact directly or indirectly with Rev and, thus, can be regarded as the RRE of EIAV. Interestingly, CAT reporter constructs that contained sequences that extended further upstream, up to nt 4479 (pCAT/44795700, pCAT/44795226, pCAT/44795173, pCAT/44795130 and pCAT/44796162), displayed much lower responsiveness to Rev (1943% CAT activity). These data suggested that sequences located within nt 44795111 might interfere with Rev responsiveness when present in cis adjacent to RRE.

We next investigated whether Rev responsiveness can be mediated by some of the suboptimal splice sites located within nt 51355538. The various CRS-mutated CAT reporter constructs were co-transfected into Cf2Th cells together with the Rev-expressing plasmid, pCEV176 (Rosin-Arbesfeld et al., 1993 ), or with pCEV as a control and CAT activity was measured (Fig. 3). It was expected that, in constructs in which an eliminated splice junction resulted in partial reduction of CRS activity, complete elimination of Rev activity would indicate that this site is indispensable for mediating Rev activity. The data showed that elimination of SA5212, SD5276 and SA5437 decreased Rev activity slightly, suggesting that their role in Rev responsiveness is limited. On the other hand, mutation of SD5538 eliminated Rev activity completely (85% CAT activity in the absence or presence of Rev), suggesting that this site plays a major role in mediating Rev responsiveness. Disruption of SA5135 did not inhibit Rev activity (100% CAT levels), indicating that SA5135 does not function in mediating Rev responsiveness. Obviously, mutations of all splice sites in pCAT/CRS*, which resulted in complete elimination of CRS activity, prevented a functional Rev assay. Thus, whereas no conclusions could be drawn regarding the extent of the contribution of SA5212, SD5276 and SA5437 to Rev responsiveness, our data suggested that SD5538, but not SA5135, functions in mediating Rev responsiveness. Since Rev responsiveness exhibited by pCAT/51115394 was not lower than that exhibited by pCAT/53475700 (Fig. 2), it is assumed that, although SD5538 might serve as a major element that mediates Rev responsiveness, the presence of the entire region is required to exert a full response to Rev.

In order to reveal the mechanism that mediates Rev dependence of CRS-containing transcripts, we investigated whether the inhibitory effect of CRS on CAT activity was also reflected at the mRNA level. Cf2Th cells were co-transfected with the Rev-expressing plasmid pCEV176 or pCEV empty vector and with the Rev-responsive plasmid pCAT/51115700 or the Rev-independent plasmid pCAT/56956162. After 48 h, total RNA or RNA extracted from the cytoplasm or nuclear fraction as described elsewhere (Cochrane et al., 1991 ) was subjected to the RNase protection assay. While the level of total cat mRNA was not affected by the presence of Rev, the level of cat mRNA in the nuclear fraction of pCAT/51115700 transfectants was higher in the absence of Rev than in its presence (Fig. 4a). In cytoplasmic RNA, the opposite occurred. The level of cytoplasmic cat mRNA of pCAT/51115700 transfectants was higher in the presence of Rev than in its absence. In contrast, the presence of Rev did not affect the levels of total, nuclear or cytoplasmic cat mRNA in pCAT/56956162 transfectants (Fig. 4b). These data suggested that the presence of the EIAV CRS in cis suppressed nucleocytoplasmic transport of CAT transcripts and that the supply of Rev in trans could relieve this suppression.

(40K): Fig. 4. Effect of Rev on the subcellular distribution of CAT mRNAs. After transfection with pCAT/5111-5700 (a) or pCAT/5695-6162 (b) reporter plasmids, in the presence of pCEV176 Rev-expressing plasmid (+, filled bars) or pCEV empty vector as a control (-, hatched bars), total, nuclear (Nucl) and cytoplasmic (Cyt) RNAs were subjected to RNase protection analysis. To ensure the same efficiency of transfection, the level of CAT DNA was measured by PCR (not shown). RNA samples were obtained from a pool of three independent transfections and the assay was carried out as described previously (Rosin-Arbesfeld et al., 1998 ). The upper panels show representative RNase protection assays. The lower panels show the means of densitometric measurements of the RNase-protected bands obtained from two independent assays.

Similar to primate lentiviruses, the suboptimal splice acceptor and donor splice sites located towards the centre of the EIAV genomic RNA serve to generate the alternatively multispliced transcripts that encode the virus regulatory proteins. In contrast, the structural polyproteins of EIAV, Gag, Pol and Env, are encoded by unspliced and partially spliced messages, the nucleocytoplasmic transport of which is controlled by Rev (Martarano et al., 1994 ). Several studies have suggested a role for HIV splice sites in Rev dependence (reviewed in Borg et al., 1997 ). Also, the major splice donor site of EIAV, SD459, was shown to confer Rev-dependence when unused (Tan et al., 1996 ). In the present report, we identified several of the centrally located suboptimal splice junctions that play a role in Rev dependence. Our data suggest that the main mediator of both CRS and RRE activities is SD5538. Interestingly, SD5538 was presumably involved in exon skipping induced by the EIAV Rev, possibly by interacting with splicing factors (Gontarek & Derse, 1996 ). Of note, our data suggest that SA5212 does not play a major role in CRS activity. SA5212 is a seldom-used splice site that is used to generate the infrequent rev transcript represented by cDNA p176 (Rosin-Arbesfeld et al., 1993 ) and a rare env transcript (Martarano et al., 1994 ). This splice acceptor site remains unused while all of the other multiply spliced regulatory transcripts of EIAV are generated. Thus, our data suggesting that SA5212 is not a potent CRS element are consistent with its presence in most of the tat and tat-rev transcripts whose nuclear export is not controlled by Rev.

The region that spanned nt 44795111 exhibited an inhibitory effect on CAT expression, most of which was independent of splice junctions. Moreover, Rev could barely overcome this inhibitory effect. Although the molecular mechanism of this negative effect is still obscure, this observation is consistent with the absence of this region in env transcripts, which are Rev dependent. Since this suppressing region is present in the unspliced genomic RNA, it might be assumed that its effect is diminished by the presence of the unused major splice donor, SD459, which is Rev responsive (Tan et al., 1996 ).

In the present study, we showed that the centrally located suboptimal EIAV splice sites confer Rev dependence on the CRS-containing transcripts. These sequences presumably act in concert with the major 5 splice site (Tan et al., 1996 ) and with the 3 end-located elements (Belshan et al., 1998 ; Mancuso et al., 1994 ) to exert Rev responsiveness. Additional regulatory elements reside within the EIAV genome that control EIAV gene expression positively and negatively independently of Rev. Thus, it is noteworthy that the region spanning nt 44737247 enabled efficient expression of CAT activity in the presence or absence of Rev (Fig. 1) due to stimulatory elements residing in the region nt 61627247 (our unpublished results). Also, the inhibitory effect conferred by elements located within nt 44796162 could not be overcome by Rev (Fig. 2), presumably due to the presence of an inhibitory element the repression effect of which was independent of Rev.

In conclusion, although the use of a heterologous CAT reporter construct enabled the identification of EIAV suboptimal splice sites as conferring Rev dependence, their activity should be evaluated in the context of the complete EIAV genome. Moreover, the lentiviral Rev protein plays a role in the switch between the early and late phases of virus infection (reviewed in Pollard & Malim, 1998 ). Thus, elucidation of the biological relevance of the unused EIAV splice sites to virus replication, persistence and pathogenesis should await further studies in the context of EIAV infection in vivo.

This work is in partial fulfilment of the requirements for the PhD degree of R.R.-A. from the Sackler School of Medicine of Tel Aviv University. This research was supported by a project grant from the USIsrael Binational Science Foundation (BSF) and the German Israeli Foundation for Scientific Research and Development (GIF).