The results presented here demonstrate the potential of coronaviruses as viral vectors. Foreign genes appropriately inserted into the MHV genome and preceded by a TRS were efficiently expressed without significantly affecting the in vitro growth properties of the virus. Expression cassettes containing different, unrelated luciferase genes were successfully tested at several genomic positions, both individually and combined within a single MHV genome. The expression levels were dependent on the identity of the particular foreign gene, as well as on the insertion site, with expression being higher the closer that the gene was to the genomic 3′ end. The insertion of foreign genes was also combined with the deletion of nonessential genes and with the rearrangement of the conserved coronavirus gene order. While the deletion of nonessential genes has been shown to yield viruses that are attenuated in vivo but not in vitro (9
; Haijema and Rottier, unpublished), rearrangement of the genome organization (10
) aims to reduce the chances that recombination with virulent field virus will result in transfer of the foreign genes.
In view of their huge RNA genomes, their pleomorphic shapes, and their apparently variable sizes and of the extended helical nature of their nucleocapsids, coronaviruses are likely to accommodate large insertions of foreign sequences. In addition, the ability to delete several nonessential genes further expands this genetic space. By the combined insertion of the two luciferase gene cassettes, we showed that extension of the MHV genome by ca. 10% of its natural size (total insertion size of 2.7 kb) was indeed accepted without adverse effects on viral growth. Consistently, insertion of the GFP gene (0.7 kb) in TGEV or MHV genomes also had little effect on the in vitro replication characteristics (6
), the GFP-expressing MHVs being attenuated in their virulence (37
). Although in our system luciferase expression levels of 1 to 2 μg/106
cells were obtained, GFP production levels of more than 40 μg/106
cells were achieved by using optimized TGEV vectors (12
). These amounts are quite similar to those observed for other positive-strand RNA virus vectors, such as poliovirus and cardiovirus, but are significantly lower than those obtained with most alphavirus-based vectors, for which values of 50 to 300 μg/106
cells have been reported (see the review by Enjuanes et al. [12
As demonstrated here with MHV, foreign gene expression by coronaviruses is dependent on several conditions. One is the insertion site relative to the genomic 3′ end. As could most clearly be deduced from the RL activities expressed from different genomic positions (cf. MHV-MRLN, MHV-ERLM, and MHV-2aRLS), expression levels are generally higher when the foreign gene is inserted closer to the 3′ end of the genome. This relationship is, however, not necessarily linear since the effect is likely to be caused, at least in part, by the contribution of TRSs occurring between the expression cassette and the 3′ end rather than by the distance between the latter per se. Studies with DI-RNAs (20
) and full-length genomes (10
) have shown that downstream TRSs generally have an attenuating effect on the transcription from upstream TRSs. According to the prevailing theory TRSs function as transcriptional attenuators or terminators during negative-strand synthesis. Transcription either resumes at the same site or at the 5′ end of the positive template after a “jump” to the leader sequence (38
). Thus, the efficiency with which the RNA polymerase reaches a particular TRS decreases with the number of preceding TRSs. Accordingly, the luciferase expression levels we observed correlated inversely with the number of TRSs located downstream of the luciferase gene. Consistently, insertion of the luciferase expression cassettes in turn also had an attenuating effect on the relative expression of the viral genes located at more upstream positions.
Proximity of the heterologous gene to the 3′ end of the genome is, however, not the sole factor determining the efficiency of expression. The sequence context of the expression cassette is another important aspect. This was demonstrated most clearly when we compared the RL expression levels of MHV-ERLM and MHV-ERLMSmN. In these viruses the RL gene context is identical: the cassette is flanked by the M gene and by the gene cluster 4/5a/E at its 3′ and 5′ sides, respectively. However, the distance of the foreign gene with respect to the genomic 3′ end is very different (3.3 kb versus 7.4 kb). Nevertheless, the relative levels of luciferase activity expressed from the viruses appeared to be quite similar. Thus, the performance of MHV-ERLMSmN was not essentially affected by the positioning of the RL gene farther away from the 3′ end or by the consequent presence of an additional downstream TRS. The findings further show that the drastic genomic rearrangement realized in this vector is without severe consequences for the viability of the coronavirus, a finding consistent with our earlier observations (10
Another important factor determining the expression level of an inserted gene cassette is the sequence context around the TRS. As was demonstrated in many studies with DI-RNAs (2
), flanking sequences can affect the transcriptional activity of a certain TRS, presumably by changing the local secondary or tertiary structure or by modulating RNA-protein interactions. Except when used in its natural genomic position the TRS of an inserted expression cassette will find itself in an artificial environment. The impact of the sequences upstream of the TRS will depend on the particular construct. Sequences flanking the core TRS (5′-AAUCUAAAC-3′) will, for instance, affect the complementarity with the leader sequence and thereby the transcriptional activity. The effects of upstream sequences may be negative or positive. This was demonstrated by a recombinant MHV in which transcription of the M gene appeared to be reduced by placing it immediately downstream of the S gene (10
) and by another MHV in which three nucleotide changes introduced just upstream of the gene 4 TRS led to a strong increase in its transcriptional activity (9
). Consistently, insertion of foreign sequences may also alter the transcription of viral genes located immediately downstream of the insertion site. Examples thereof were observed here showing that the expression of the M and the S gene was decreased by the upstream presence of the luciferase gene in the cases of MHV-EFLM and MHV-2aRLS, respectively.
Also, sequences located downstream of a TRS can modify its transcriptional activity. Again, nucleotides immediately flanking the core sequence will do so by their direct effect on the complementarity with the leader sequence. However, the foreign sequences also contained in an expression cassette can affect their own expression by somehow influencing the TRS. This became clear when we compared the transcription patterns of the viruses MHV-ERLM and MHV-EFLM, which differ only in the identity of their luciferase genes. Dramatic differences were observed in the transcription levels of the mRNAs specifying the foreign genes, whereas the transcription of the viral mRNAs remained quite similar for the two viruses. How these foreign sequences exert these specific effects is not yet clear. It will, however, be important to gain more insight into this phenomenon since it should eventually allow prediction of foreign gene expression efficiencies, thereby enabling the rational design of future coronaviral vectors.
Besides the synthesis of the intended sgRNAs, insertion of foreign gene expression cassettes into coronavirus genomes has thus far almost invariably revealed the appearance of additional, smaller sgRNAs (13
) (the present study). These are the products of aberrant leader-to-body fusions that occur due to the occurrence of TRS-like sequences in the inserted gene. The unintended leader-to-body fusion sites in the FL and RL genes contained stretches of 6 or 4 nt homologous to the canonical TRS, respectively. Interestingly, identical stretches of sequence present in the MHV genome (21
) and in a DI-RNA (29
) have previously been shown not to direct the synthesis of sgRNAs. Apparently, the sequences flanking the unconventional leader-to-body fusion sites in the foreign genes have a positive effect (or lack an inhibitory effect) on the functioning of the TRS-like sequence. Alternatively, long-range RNA effects or processes involving the ribonucleoprotein interactions, as suggested by Fischer et al. (13
), might play a role in determining the site of transcription initiation. Elimination of the TRS-like sequences, by mutations not affecting the encoded protein, might result in higher foreign gene expression levels as this is likely to reduce the attenuating effect of downstream TRS(-like) sequences on the expression from upstream TRSs.
Recently, genetic engineering of coronaviruses has become feasible by the availability of infectious cDNA clones (1
) and by the development of efficient RNA recombination systems (14
). This allows us now to study the complex interplay of factors determining sgRNA synthesis in the context of the complete genome. This is important because, whereas most data on coronavirus transcription have been generated by using DI-RNAs, these results may not always be directly applicable to full-length viral genomes (25
). For the further evaluation of coronaviruses as vectors, an important issue will be the genetic stability of the viruses carrying foreign gene cassettes, since this will eventually determine the applicability of these vectors. While expression of the GFP gene was found to be stable in several coronaviruses (6
), our current studies with RL- and FL-expressing viruses indicate that the stable maintenance of foreign genes is dependent on the nature of the heterologous gene, on the particular coronavirus vector used, and on the particular site of gene insertion (C. A. M. de Haan, B. J. Haijema, and P. J. M. Rottier, unpublished results). Thus, while the RL gene was stably expressed from various genomic positions, the FL gene was much less stable, its expression being lost gradually through deletions in the FL expression cassette. Interestingly, moving the cassette to a more upstream position or inserting it in the FIPV genome resulted in a more stable phenotype.