The coexistence of closely related cp and noncp viruses represents one interesting feature of pestiviruses. A variety of cp-BVDV-specific genome alterations is responsible for the lysis of the infected cells. All these genome alterations are somehow connected with the expression of the viral nonstructural protein NS3, in most cases by induction of proteolytic processing at the amino terminus of NS3. Among the mutations leading to a cp phenotype the integration of different cellular sequences is of special interest since the presence of these sequences in the viral genome sheds light on specific aspects of the biology of their cellular counterparts, namely, their involvement in proteolytic processing. LC3, the cellular sequence identified in BVDV JaCP, was first discovered as a 16.4-kDa protein in preparations of MAPs, including MAP1 and MAP2 (26
). It was recently found that LC3 had, besides other mostly not well understood functions (23
), a role in autophagy and possibly other processes involving membrane trafficking (17
). LC3 and its yeast homologue Aut7p/Apg8 might help during the initiation of vesicle formation, vesicle elongation, vesicle transport, and/or membrane fusion.
First indications for the importance of LC3 in autophagy were obtained for the homologous yeast protein Aut7p/Apg8p. Part of the mutations that blocked autophagy in yeast affected Aut7/Apg8, and it was soon discovered that the protein encoded by this gene was associated with the membranes of autophagic vesicles that are known as autophagosomes. After we had provided evidence for the fact that LC3 inserted in the polyprotein of a naturally occurring mutant BVDV served as a proteolytic processing signal, Kirisako and coworkers (19
) found that Aut7p/Apg8p is processed at an equivalent position close to the carboxy terminus of the primary translation product. This process represents the first step in an activation cascade that finally leads to conjugation of the protein to phosphatidylethanolamine and integration of the conjugate in the membranes of autophagosomes (16
). Processing of the Aut7p/Apg8p carboxy terminus is executed by the cysteine protease Aut2p/Apg4p and represents a crucial step in autophagy since Aut2/Apg4-deficient yeast mutants are defective in this process.
Based on sequence comparison studies two different homologues of Aut2 were identified in the human genome (HsApg4A and HsApg4B) (19
). We demonstrate here for the first time that bovine cells contain mRNAs that code for proteins showing a high degree of similarity with the two human sequences. In addition, a sequence coding for a protein similar to HsApg4B was also identified in chicken cells. Only bAut2B2 and cAut2B were able to process a substrate protein in an LC3-dependent manner. It therefore can be concluded that these proteases represent the enzymes responsible for LC3 carboxy-terminal processing. Similar data have recently been published for the mouse, where a protease homologous to bAut2B2 was identified by a biochemical approach as the protease that cleaves mouse LC3 (15
Our experiments aiming at identification of a system that allowed functional tests with the candidate proteases revealed the extreme conservation of the LC3-processing system since our mammalian substrate was processed by an intrinsic protease(s) in phylogenetically very distant avian, fish, and mosquito cells. Both the conserved size of the cleavage products and the requirement of the LC3 carboxy-terminal sequence suggest that the cleavage occurs in all cases at the same site downstream of the motif TFG, which was identified first as a cleavage site in our viral substrate and later in the yeast system. We also found indications for processing of the mammalian substrate in wild-type yeast, but the respective results could not be reproduced consistently. In any case, the rescue of a yeast AUT2
deletion mutant by two of the mammalian proteases and the chicken protease cAut2B proves the functional complementation of the processing system. The rescue of the yeast aut2/apg4
mutant was also possible with two different sequences from Drosophila melanogaster
). This high degree of flexibility with regard to cross complementation is somewhat surprising since the different protease sequences are quite divergent. The obvious similarity of the substrates, especially at the cleavage site, together with a putatively conserved basic structure of the proteins is apparently sufficient to allow enzyme/substrate recognition and cleavage.
A still open question is why more than one Aut2p/Apg4p homologue is expressed in mammalian and insect cells. It is expected that the different proteases have clearly distinct functions. The deficiency of bAut2B1 for processing LC3 in all tested systems and its inability to complement the defect in aut2/apg4
-deficient yeast cells indicate that this protein represents no active protease or processes a substrate different from LC3. The latter conclusion is even more likely for bAut2A. It has to be considered in this context that mammalian cells contain at least two LC3-homologous proteins, namely, the Golgi-associated ATPase Enhancer of 16 kDa (GATE-16) (41
) and γ-aminobutyric acid A receptor-associated protein (GABARAP) (49
). GATE-16 was first described as a membrane transport modulator in the constitutive secretory pathway (24
). It is localized in the Golgi apparatus, interacts with different proteins engaged in vesicle transport, and might assist in vesicle-membrane fusion. Interestingly, Aut7p/Apg8p seems to be the functional homologue of GATE-16 in yeast and can replace GATE-16 in assays mimicking mammalian intra-Golgi transport in vitro (23
). In contrast, it was proposed that GABARAP plays a role in receptor anchoring and clustering via its ability to interact with the microtubule and microfilaments (49
). Both proteins have considerable sequence similarity with LC3, including glycine 120. It was shown that these two proteins are also processed at their carboxy termini (45
), and sequences coding for fragments of both proteins have recently been identified in the genomes of cp pestiviruses (5
). HsApg4A expressed in bacteria was shown to cleave GATE-16, so future experiments might show that this protease is responsible for processing this protein in vivo (43
It is likely that the system of mammalian homologues of Aut7p/Apg8p and Aut2p/Apg4p is even more complex than would be expected on the basis of the above-described data. A recent publication reported that human and mouse cells contain sequences able to encode two additional homologues of Aut2p/Apg4p. One of these new sequences was able to complement an Aut2/Apg4 defect in yeast (27
). Taken together, it seems that in higher-eukaryote cells the yeast Aut7p/Apg8p system has been differentiated so that at least three Aut7p/Apg8 homologues with different functions exist. Each of these proteins might be processed by its own specific protease. Moreover, the existence of multiple Aut2p/Apg4p-homologous proteases in higher eukaryotes could also be necessary for differentiation with regard to deconjugation of the Aut7p/Apg8p homologues attached to their different targets. It is an interesting result in this context that bAut2A, although not able to process LC3, can substitute for Aut2p/Apg4p, so that the systems seem to be more interchangeable between yeast and bovine cells than between the different substrates within one species. Further analyses are necessary to verify whether the existence of more than one Aut2p/Apg4p-homologous protease is indeed a consequence of this differentiation with, e.g., each protease responsible for activation cleavage of one individual Aut7p/Apg8p homologue in vivo. It is also possible, however, that these proteases have totally different functions since it has been reported recently that four different mouse homologues of Aut7p/Apg8p, including LC3, GATE-16, and GABARAP, are all cleaved by a single protease homologous to Aut7p/Apg8p (15
Viruses are known to interact with their host cells at different levels, including regulation of gene expression. It therefore is possible that BVDV infection specifically influences the expression of sequences that serve as partners for RNA recombination or code for proteins important for the processing of a recombinant polyprotein from a cp BVDV. However, we show here that, at least with regard to LC3 and its corresponding protease, no significant changes of the expression levels can be detected. It therefore seems reasonable to conclude that the virus by chance picked up a sequence that is not induced by viral replication. It is likely that neither LC3 nor its protease belongs to a set of cellular proteins recruited by the virus for its normal replication; instead, LC3 was probably specifically selected for its ability to serve as a processing signal in the context of virus cytopathogenicity.
The genome of BVDV JaCP contains a duplication of viral NS2-3/NS4-coding sequences flanking the LC3 insertion, and we have shown before that a viral genome containing LC3 but no duplication cannot direct the generation of infectious progeny virus but can be propagated in the presence of a helper virus (29
). As shown here, mutations considerably reducing the processing at the LC3/NS3 site prevent recovery of the defective viruses even after several blind passages, so it can be concluded that replication of these RNAs is (nearly) blocked. It has to be stressed that the mutations tested here all are located within the LC3 sequence and therefore do not change a viral protein with a distinct function in viral replication. This stands in marked contrast to the data published before where mutations within the NS2 gene were tested (20
). We therefore can conclude now that interference with the cp-BVDV-specific processing at the amino terminus of NS3 has a prominent effect on virus viability. This can either be due to an increase of the concentration of the aberrant fusion protein NS2/LC3/NS3 or can reflect a dependency of the virus on a certain level of processing, as proposed before (20
). As shown in the experiments with the RNAs mimicking the JaCP genome, impairment of processing at the NS3 amino terminus cannot be overcome by a second copy of the NS3 gene in a genome containing a duplication. It was shown before that NS3 and most other nonstructural pestivirus proteins cannot be provided in trans
to support viral RNA replication (11
). Obviously, the duplication and the cellular insertion lead to a “quasi-trans”
arrangement of the functional copy of the gene located upstream of the insertion. Nevertheless, low-level RNA replication was apparently occurring for the CysP1 construct with duplication since the mutated viral genome was able to delete the defective processing signal together with the duplicated viral sequences by recombination. In any case, it has to be stated that a BVDV containing an LC3 coding sequence insertion in its genome is absolutely dependent on the cellular LC3-specific processing in order to replicate its RNA efficiently and generate infectious progeny virus.
The LC3-Aut7p/Apg8p system is reminiscent of the ubiquitin system. The fundamental processes of proteolytic processing at the carboxy terminus, activation, and conjugation are found in both systems (see references 35
, and 50
for recent reviews). Moreover, ubiquitin and LC3 seem to be multifunctional, with one function being to provide a kind of covalently bound label for cellular molecules or structures that marks the target for further processes. This functional similarity is also represented in the structure of the proteins since a ubiquitin-like fold was identified in the three-dimensional structure of the LC3 homologue GATE-16 (36
). Apparently, the processes involving conjugation of small proteins after carboxy-terminal cleavage and activation are highly important for living cells, so that the processes themselves and the interacting components are among the most conserved systems of biology. Pestiviruses have learned to exploit these systems for their own purposes. Nearly all members of the respective families of cellular sequences can be found as insertions in cp pestivirus genomes. It seems to be worthwhile to keep an eye on new cp pestivirus isolates since they might have found further interesting candidate proteins with not yet known cellular functions.