The family of herpesviruses comprises important human and many veterinary relevant pathogens. Classification of herpesviruses and separation from other virus families is predicated on virion structure, cell tropism and virus host range [1]. Their common feature is their ability to infect the host for life. There are important other reasons why herpesviruses stand out from other virus families namely their high complexity and their perfect adjustment to their host. As more and more sequence information of herpesviruses accumulate, analysis and comparison of the genomes allow the identification of core genes that are shared within different subfamilies, and unique genes that resemble adaptation to fulfill host specific prerequisites. The number of protein coding genes, also according to their genome size, vary between the subfamilies whereby α- and γ-herpesvirinae have on average around 70-80 genes and β-herpesvirinae around 160-230 [2]. There is a core set of 43 highly conserved genes, shared by all herpesviruses, and that are mainly involved in the basic and fundamental procedures of the viral life cycle as entry into the cell, DNA replication, packaging of the genome and maturation of infectious particles [3]. Although herpesviruses are extensively studied there are still genes in the list of the core genes with unknown function [4] (Figure 1). The additional species and subfamily specific genes cover areas of cellular tropism, host shut-off or anti-apoptotic processes, evasion from the immune system and maintenance of latency [5,6,7,8]. It is very likely that most of the remaining genes of unknown function will belong to one of these groups. Not all of these genes are necessary for viral replication in the host, but shape the outcome of the infection [9].

In order to identify a function of an unknown gene biological and biochemical assays of the wild-type (wt) protein or the deletion or mutation of the gene, as well as the use of DN mutants [10] and the study of their respective phenotypes and meanwhile also computational alignment to homologues genes of known function can be used [11,12].

There are different methods how a DN mutant may work, which are reviewed in detail from Veitia [35,36]. Mutations in the catalytic site of an enzyme are one way how a DN mutant can arise; in that case the substrate is bound but not converted and so the balance of the reaction is disturbed. Examples in cell biology are mutations in ATPases or GTPases [37,38,39], especially Rho GTPases have been widely used [40] (Figure 2A). Very often DN proteins work in complexes, where the inhibitory potential ranges from the block of a simple dimer, as for example membrane receptors or transcription activators, which can transmit a signal only after dimerization [41,42] (Figure 2B), up to multi-subunit complexes (Figure 2C), demonstrated nicely by Barren and Artemyev for G-protein alpha subunit complexes [43]. DN mutations in transcription factors can poison a whole pathway if they block the binding sites for active mutants, therefore, inhibiting all downstream gene expression. As for example in the case of DN mutants of the proto-oncogene p53, resulting in the loss of growth inhibition and cancer manifestation [44] (Figure 2D).

The example of a DN mutant that acts in the wt form as a homo-dimer explains why a DN mutation can cause a stronger phenotype than a deletion (in case of a diploid eukaryotic organism). By equal expression of wt and DN, homo-dimerization of the two proteins will lead to the formation of wt-wt, DN-wt, wt-DN and DN-DN complexes. Therefore only 25% of the dimeric complexes will be functional, while in a deletion it would be 50%. Overexpression of the DN protein shifts the ratio more to non-functional complexes. As (random) insertion into the eukaryotic genome is still easier to achieve than targeted deletion of both alleles, it is obvious why the DN approach is superior to deletion mutants. The problem of targeted deletion was partially overcome by the RNAi approach [45], where complementary siRNA molecules initiate the degradation of the mRNA of both alleles. A drawback of this method is still that down regulation is rarely complete and the knock down is dependent on the half life of the gene product.

Furthermore, DN proteins have one big advantage that cannot be substituted by any deletion, namely that they can arrest the complexes or pathways at different steps. Often proteins are dynamic and have several functions as they bind to many other proteins or have different localizations depending on activation status. Mutating one domain to make it a DN protein, can lead to the disturbance of only one function while leaving other domains intact. Therefore, not the most prominent phenotype due to the loss of the protein, which may in fact reflect the sum of several functions, but several arresting steps can be monitored.

An example for this postulate is the cellular protein Dynamin (Figure 3) that is involved in clathrin-coated vesicle endocytosis. It consists of five domains, a N-terminal GTP hydrolysis domain, a middle domain, a pleckstrin homology (PH) domain a GTPase effector domain (GED) and a C-terminal proline-rich domain (PRD) [46]. A well studied DN mutant is the DynaminK44A, a mutant that cannot bind GTP resulting in a block of receptor-mediated endocytosis [47,48]. Mutation of residue K535 in the PH domain, also giving rise to a DN phenotype, acts by co-oligomerizing with endogenous wt Dynamin and indirectly impairs phosphoinositide binding [49,50]. Deletion of the PRD domain by a stop codon in place of the Proline 746 of Dynamin interferes with the recruitment of Dynamin to clathrin–coated pits and inhibits in a DN fashion receptor-mediated endocytosis [51]. Together with the finding that another mutation in the GTPase domain K142A is defective in its ability to change conformation although still hydrolyzing GTP, the hypothesis was postulated that the function of Dynamin in endocytosis requires both GTP hydrolysis and a resulting conformational change before or concomitant with, vesicle scission [52]. Thus, the use of different DN mutants of the same protein can reveal more information than a deletion could have given, as they represent states that can be perceived as snap shots of highly dynamic processes.

For herpesviruses entry via receptor-mediated endocytosis or via receptor-mediated fusion has been postulated [55]. Evidences for both cases exist and might depend on the cell line used for the study. Interestingly, use of DN versions of the focal adhesion kinase (FAK), Src-Kinases and RhoGTPase, resulted in decreased uptake of Kaposi’s sarcoma associated herpesvirus. This led to the hypothesis, that upon binding of the virus to surface receptors, signaling of integrins to FAK activates Src. This Src activation may then recruit Clathrin and Dynamin to the cell surface, allowing the bound virus to enter the cell via newly formed vesicles [56,57,58,59]. Although viral entry mechanisms are still a matter of debate, usage of defined DN mutants can help to elucidate the complex process of subsequent steps that happen at the cell membrane.

After entry, capsids are transported to nuclear pores and the viral DNA is released into the nucleus, where viral DNA replication takes place. Study of the cell cycle arrest induced by herpesvirus and the resulting apoptosis has been profiting from the huge set of available DN cyclins. Abusing their original function in cells, herpesviruses bind to cellular cyclins to control late gene expression. Advani and colleagues could demonstrate that a subset of γ2-proteins are not produced in cell lines expressing DN cdc2 [60]. Usage of cyclin D3^DN helped to understand the localization and function of ICP0 of HSV [61]. These examples are just illustrations of what can be or has been done with the help of cellular DN proteins.

Recently, we could prove the anti-apoptotic feature of M36 of murine cytomegalovirus (MCMV) by replacing the viral gene with a DN FAS-associated via death domain (FADD^DN). While the deletion virus ΔM36-MCMV was severely impaired, inhibition of the apoptosis pathway by FADD^DN rescued virus replication in vitro and in vivo. This novel approach of inserting cellular DN proteins into the virus genome has enabled us to define the biological function of M36 and might represent a strategy to evaluate other anti-apoptotic viral genes [62].

Viral DN genes are in wider use than cellular DNs [63,64,65,66,67,68,69] to analyze viral replication cycles. Instead of giving further examples of what can be done with viral DNs in terms of identifying pathways and processes on the cell level, we will address the question what can be done in the living organism.

The protein UL9 of HSV1 is the origin binding protein and, therefore, essential for viral replication [70]. After successful inhibition of HSV-1 infection in cell culture by the DN protein UL9-C535C (UL9^DN) Yao and colleagues created a strongly attenuated recombinant herpesvirus (HSV-UL9^DN) that can conditionally express the UL9^DN[71]. The induced expression of the UL9^DN abolishes the expression of γ₂-proteins almost completely as they are dependent on viral DNA replication. Co-infection of HSV-UL9^DN together with wt HSV-1 or HSV-2 inhibited replication of both viruses [72]. Intraocular infection of mice with HSV-UL9^DN did not lead to an acute herpetic infection. After challenge with the wt strain no keratitis and encephalitis was seen, due to strong neutralizing antibody titers and cell mediated immunity induced by the mutant [73]. To further increase the safety and immunogenicity of HSV-UL9^DN as a vaccine the gene of UL9 was deleted and an additional copy of a highly immunogenic glycoprotein (gD) inserted. Herpetic skin disease could be prevented using this new derivate CJ9-gD in the guinea pig model [74]. Although the DN approach in vaccine development must be carefully considered, as the addition of inhibitory proteins alone may not fulfill the same safety criteria as the deletion of essential genes. Furthermore it provides a new feature to vaccines as they can now also inhibit herpesvirus DNA replication of the wt type virus in the rare case of a co-infection of the same cell at the time point of vaccination.

We successfully used a Tn7-transposon-based mutagenesis screen on isolated viral genes (reviewed in [86]), which introduces a 5 amino acid insertion or a stop codon and reinserted the mutated genes in an ectopic position. This random screening approach proved itself to be generally applicable and subtle enough to reduce the risk of protein misfolding. As the screen is genome based, the mutants can be studied in the biological context and is in general still suitable for high-throughput analysis. These comprehensive screens allowed us to identify important sites for functionality of the protein in terms of viability of the viral progeny, localization of the protein in the cell as well as binding sites for interaction partners. In case of the nuclear egress protein M50 of MCMV the screen revealed the importance of the N-terminal part of the protein, identified the binding site to M53 therein, the crucial necessity of the transmembrane domain and the importance of a conserved proline stretch, as well as an non-essential regions which could be used to tag the protein [87].

The next protein that we addressed with this comprehensive mutagenesis was M53, the binding partner of M50 both forming together the nuclear egress complex (NEC). Detailed analysis of the resulting mutants revealed that the very N-terminal part comprises a nuclear localization signal (NLS) that could be replaced by the NLS of SV40, the binding site to M50 could be narrowed down to a region of 31 amino acids and that there are two important domains (conserved domain 2 and 4) with other functions essential for morphogenesis of MCMV [88]. Further improvements to the transposon-based mutagenesis allowed us to cover targeted genes with insertions of on average at least every fifth amino acid (data not shown).

As DN mutations can arrest dynamic processes at different stages and therefore allow an analysis of phenotypes that cannot be resolved with deletion mutants, we were keen to find also DN mutants in the comprehensive transposon libraries of M50. In the first screening step we used a total of 104 mutants of M50 that were inserted into MCMV-ΔM50 genomes to complement the missing gene. Thereby we found 34 mutations that possessed a loss-of-function phenotype, meaning that no viruses could be reconstituted from the transfected BAC DNA. To analyze if these mutations could also be DN, the individual mutants were inserted at an ectopic position into the wt genome. Furthermore the gene expression of DN candidates was controlled via a Doxycyclin-inducible expression cassette, which also allowed a quantification of the inhibitory effect [89]. Interestingly, no mutation that destroyed the binding site to M53 possessed a DN effect. Whereas two insertion mutants, (M50ⁱ⁴⁰, M50ⁱ¹²⁵) as well as the targeted deletion mutant M50^ΔVR were able to inhibit M50^wt function after induction of gene expression and therefore leading to a reduction of virus production [90]. Semi-quantitative expression analysis by Western Blotting revealed that the DN effect was dependent on the protein amount. Further immunofluorescence analysis of the NEC complex with the DN mutants showed that there was neither mislocalization nor decreased stability of the proteins. Therefore, we concluded that the DN function affected the interaction of the NEC with another yet unidentified important protein partner. Based on sequence alignment with other homologues of M50 a motif within the deleted region of ΔVR was identified and the targeted deletion resulted in an even stronger DN effect. Electron Microscopy analysis of this mutant M50^ΔP showed the expected phenotype, as packaged capsids accumulated in the nucleus. Furthermore, the block of capsid export was accompanied by the accumulation of membrane stacks within the nucleus.

Confirmed by the positive results from the M50 screen, we next applied the DN screen to its binding partner M53. To this end we tested 46 mutations, which could not rescue the virus MCMV-ΔM53 phenotype. Unlike the M50 screen here we could find numerous DN mutants, seven insertion mutations and four stop-mutations mainly located in the conserved region 4 (CR4) and two DNs in CR3. Furthermore, seven insertion mutations in CR2 were heavily attenuated (Popa et al., manuscript in preparation). In general, all mutations in CR4 possessed a stronger inhibitory effect than mutations in CR2. All CR4 mutants blocked nuclear egress of viral capsids by interfering with the NEC, by mislocalization or destabilization. One stop mutant M53^S309 was very effective and could block virus egress completely. Surprisingly, this effect was not depended on binding to the NEC (in contrast to M50^DNs) and rather revealed a new function for the M53 protein in DNA cleavage or packaging, as imperfect cleavage of unit length genomes could be observed, which has also been shown for the positional homolog BFLF2 in EBV[91]. This leads to an accumulation of immature capsids that did not contain viral genomes. Another DN mutant M53ⁱ²⁰⁷ did not influence capsid maturation, therefore indicating that the block might be happening to a latter stage in morphogenesis. These findings suggested new functions for M53.

This screen once more proves that DN proteins can help to elucidate herpesviral morphogenesis steps, beyond the analysis of gene deletion mutants by freezing processes at different stages.

While linker-scanning transposon mutagenesis to identify DN mutants has been applied to cellular genes [92,93,94], the impact of this tool has not yet drawn full attention of the herpesvirus field. While several groups already use linker-scanning mutagenesis to identify important regions in the gene of interest, they did not screen the libraries for DN mutations [95,96]. But, Lin and Spear nicely showed in 2007 that several mutations in gB, derived from linker-scanning mutagenesis, exhibit DN features; and effect expression at the cell surface and entry activity [97]. Altogether comprehensive transposon-based screening for DN proteins can be applied to any essential viral gene to identify DN mutations. Analysis of nonessential genes might be also possible depending on the availability of read-out systems for the resulting phenotype.