Codon pair bias deoptimization of the major oncogene meq of a very virulent Marek ’ s disease virus

Codon pair bias deoptimization (CPBD) has been successfully used to attenuate several RNA viruses. CPBD involves recoding a viral protein-coding sequence to maximize the number of codon pairs that are statistically underrepresented in the host, which presumably slows protein translation and, hence, causes virus attenuation. However, since recoding preserves the amino acid composition and codon bias, attenuated and parental viruses are antigenically identical. To determine if Marek’s disease virus (MDV), a highly oncogenic herpesvirus of the chicken with a large double-stranded DNA genome, can be attenuated by CPBD of its major oncogene meq, we recoded the gene to minimize (meq-D), maximize (meq-O), or preserve (meq-R) the level of codon pairs that are overrepresented in the chicken protein-coding sequences. Unexpectedly, mutants carrying recoded genes produced comparable or increased levels of Meq in the context of viral infection in cultured cells. In addition, parental virus and mutant viruses carrying recoded meq genes replicated with comparable kinetics in vitro and in vivo, and were equally virulent in susceptible chickens. In summary, CPBD of meq failed to produce any quantifiable attenuation of MDV and confirms differences in the complexity of applying CPBD to large DNA viruses versus RNA viruses.


INTRODUCTION
Codon pair bias deoptimization (CPBD) is a highly efficient method for virus attenuation.It is based on the observation that occurrence of certain codon pairs in protein-coding sequences is significantly more (or less) frequent than their synonymous counterparts [1,2].CPBD-based attenuation involves reshuffling of available codons in a viral gene to minimize the number of codon pairs that are overrepresented in the protein-coding sequences of the host.Thus, the codon pair bias (CPB) of the recoded gene is altered, while the amino acid sequence remains identical to that of the parental protein [3].Moreover, designing CPBD-based vaccines is a matter of minutes and synthetic production a matter of weeks, making CPBD one of the most rapid methods available for virus attenuation.
The exact mechanism behind CPBD-mediated virus attenuation is currently explained by two competing, yet poorly understood and unproven, theories: (i) CPBD increases the number of naturally underrepresented codon pair combinations, which results in inefficient translation and reduced protein production [3][4][5].(ii) CPBD inadvertently increases the number of CpG and TpA dinucleotides in the sequence enabling the recoded viruses to be recognized and cleared by a yet to be identified innate immune mechanism [6,7].Nevertheless, CPBD has been successfully employed to reduce the virulence of several RNA viruses including Enterovirus C (poliovirus) [3], Human immunodeficiency virus type 1 [8], Human orthopneumovirus [9], Porcine reproductive and respiratory syndrome virus 2 [10], Indiana vesiculovirus [11], Dengue virus [12] and Zika virus [13].Furthermore, CPBD-based vaccine candidates for influenza A virus have shown up to 100000-fold attenuation compared to the virulent parental virus and were highly protective in mice and ferrets [14][15][16].
Marek's disease virus (MDV) causes Marek's disease (MD), a highly contagious, immunosuppressive and lymphoproliferative disease in chicken, with most chickens in commercial poultry operations being infected within the first days of life.Modified live virus vaccines are the only option for MD control and most commercially raised chicken are vaccinated either in ovo or immediately after hatch [17].However, vaccines against MD can only prevent disease Downloaded from www.microbiologyresearch.orgby IP: 54.70.40.11On: Sun, 02 Jun 2019 01:24:36 symptoms and do not prevent super-infection with virulent MDV strains [18].Such permissive vaccines are suspected to drive the evolution of MDV towards higher virulence [19,20] and can explain the rapid evolution from mild to virulent to very virulent MDV pathotypes that have been identified following introduction of new vaccine regimens [18,21,22].Hence, development of MD vaccines superior to the gold standard CVI988/Rispens is one of the most important goals of present-day MD research [21].As MDV's natural host is the chicken, it allows the assessment of the effects of CPBD on a large DNA virus in vivo in a unique and natural small animal model [17].Moreover, since MDV regularly causes disease in virtually all infected chickens, it is an ideal model to test vaccines [23].
Recently, we studied the effect of CPBD on an essential herpesvirus gene U L 30 which encodes the catalytic subunit of the viral DNA polymerase [24].We have shown that CPB deoptimization had a negative, while CPB optimization had a positive effect on protein expression and the biological properties of viral mutants carrying the recoded genes in cell culture.In addition, the level of CPBD correlated with the degree of attenuation.An MDV mutant that carried a fully CPB-deoptimized U L 30 was not recoverable in cell culture, a mutant that had two-thirds of U L 30 CPBdeoptimized was moderately attenuated in vivo and MDV mutants that had one-third of U L 30 CPB-deoptimized had properties of the parental virus.
The major challenge with CPBD of large DNA viruses including MDV is that recoding of the entire genome is impossible due to its sheer size.Therefore, it seems important to identify target genes for recoding for efficient attenuation.Herpesviral genes can be classified into two groups based on whether they are essential for viral replication.Since non-essential genes play an important role in herpesvirus pathogenesis and host control, we hypothesized that their deoptimization might lead to reduced virulence and possibly enhance immune response without interfering with viral growth.
Our goal was to study the effect of CPBD of meq, one of the non-essential genes of MDV, on disease progression and pathogenesis.meq is the major oncogene of MDV, an important immunogen as well as one of the few genes that are consistently expressed during lytic replication, latency and transformation [25,26].Remarkably, the vaccine strain CVI988/Rispens expresses two functional copies of meq [27].While Meq can transform human and rodent cells in vitro, it does not induce tumours in vaccinated chickens [28].
In this study, we recoded the meq gene of the very virulent RB-1B strain of MDV to obtain its CPB-deoptimized, -optimized and -randomized versions.Unexpectedly, the mutant viruses containing recoded genes produced comparable or higher levels of Meq than the parental virus in productively infected cells.In addition, the parental virus and mutant viruses replicated with comparable kinetics in vitro and in vivo, and were equally virulent in susceptible chickens.The mutant viruses retained oncogenicity in vivo, implying that meq and possibly other non-essential genes are a poor target for CPBD-based attenuation.

Calculation of CPB
CPB is reported in all species studied and is species-specific [3,6].The CPB score of a given protein-coding sequence is calculated as the arithmetic mean of the CPS of its constituent codon pairs, where CPS signifies the level of overrepresentation/underrepresentation of a given codon pair [3].Furthermore, a positive CPB score indicates that the ORF consists of mostly overrepresented codon pairs.Since MDV is an obligate parasite of the chicken, we assumed that it must have adapted to the chicken CPB to attain optimum translational efficiency.Therefore, we used the calculated CPS of the 3721 possible codon pair combinations (61Â61, excluding stop codons) in chicken protein-coding sequences (Gallus gallus) for our calculations [6].We calculated the CPB scores of 112 MDV genes as well as 15 762 chicken genes and plotted them against their respective lengths to visualize the distribution of CPB (Fig. 1a).The majority of the chicken genes had a positive CPB score (mean=0.0755),while the majority of the MDV genes had a negative CPB score (mean=À0.0646),suggesting that encoding of MDV genes is not substantially influenced by the host CPB.

Recoding of the MDV meq gene
The MDV genome consists of a unique long (U L ) and a unique short (U S ) region, each of which is flanked by a pair of identical but inverted repeat regions, referred to as terminal and internal repeats (TR L , IR L , IR S and TR S ).Repeats that flank the U L region (TR L and IR L ) contain only a small number of genes, including vTR, meq, LORF5a, LORF4 and vIL-8 (Fig. S1, available in the online version of this article).However, all of these genes, except LORF5a, play a significant role in MDV pathogenicity and tumour formation [26,[29][30][31][32][33].In addition, the ~4 kb region containing the meq, LORF5a, LORF4 and vIL-8 genes was shown to have a complex transcription and splicing pattern during MDV infection in vivo, giving rise to a number of different splice variants [29,32,34].
We recoded meq to deoptimize (meq-D) and optimize (meq-O) its CPB by minimizing and maximizing, respectively, the occurrence of overrepresented codon pairs.In a third recoded version, we randomized its CPB (meq-R) by reshuffling the codons such that the CPB score was comparable to that of the parental gene (meq-W).Several potential splice donor and acceptor sites have been proposed within the meq ORF, but only one, the splice donor site D2 (cacc-tacGTaagga) (Fig. S1), has been shown to be functional [30,35].This functional splice donor site was preserved during recoding.The alternative splice variants whose acceptor/donor sites lie outside of the meq ORF were unaffected during the recoding.As a result, the 15  that contains the splice donor site D2 is present in all recoded meq genes.
The initial annotation of MDV genome suggested that the ORF of the meq gene overlaps with two other ORFs that might encode functional proteins: LORF6 and an ORF that was shown to encode a 23 kDa nuclear protein in MDV-transformed lymphoblastoid cells [34].We deemed both genes non-essential for MDV pathogenesis, and, therefore, disregarded them during the recoding of meq.
We kept the Gibbs free energy (DG) of the folding of recoded RNA within a narrow range to ensure that any phenotypic changes observed are a result of alteration in CPB and not because of extensive secondary RNA structures.To confirm this, we scanned the recoded sequences through the mFold program as described [3,36].The DG profiles of all the three recoded versions were comparable to the parental gene (Fig. 1b).
The CPB scores of the parental and recoded genes are listed in Table 1.To evaluate the effect of recoding, we aligned the sequences and calculated the proportion of nucleotides and codons that occupied the same positions in the parental and the recoded sequences.For any two sequences, on average 60 % of all codons that occupied the same position were different.However, the percent nucleotide identity was higher between any two sequences since most synonymous codons share the first two nucleotides and only the third nucleotide is different.This results in the introduction of numerous silent point mutations in the recoded genes (Table 1).

Effects of recoding on protein production
To evaluate the effect of recoding on production of the Meq protein, we transfected dual eukaryotic expression plasmids into primary chicken embryo cells (CECs) and chicken epithelial 8E11 cells.The plasmids had meq-EGFP fusion genes cloned into one multiple cloning site and an mTagBFP in the other.EGFP fluorescence was localized primarily to the nucleus of transfected cells, confirming that EGFP was produced as a fusion protein of Meq (Fig. S2).We quantified the mean fluorescence intensities of TagBFP and EGFP for each sample by flow cytometry 24 h post transfection (Figs S3 and S4).The TagBFP served as a control for transfection efficiency while the EGFP provided a direct measure for protein production by the meq-EGFP fusion gene.Surprisingly, not only meq-D, but also meq-O and meq-R showed a significant reduction in Meq production compared to meq-W in CECs (72-85 % reduction; Fig. 2a) and 8E11 cells (85-90 % reduction; Fig. 2b).The reduction of Meq production expressed from meq-O and meq-R was comparable to those of meq-D, and the magnitude of reduction of EGFP production was readily observable by fluorescent microscopy (Fig. S2).These results were unexpected, because protein production of the recoded genes did not correlate with the CPB or the number of CpG dinucleotides in recoded regions (Table 1).

Construction and characterization of mutant viruses in vitro
We generated MDV mutants (pDIR L -meq-D, pDIR L -meq-O and pDIR L -meq-R) by replacing the parental meq-W gene with its CPB-deoptimized, -optimized and -randomized versions by en passant mutagenesis [37].To ensure that any phenotypic changes are caused by CPB alterations in meq and not due to unintended mutations elsewhere in the viral genome, we generated revertants (pDIR L -meq-D-rev, pDIR Lmeq-O-rev and pDIR L -meq-R-rev) from each of the mutants by restoring the parental meq.We assessed the replication properties of the reconstituted mutant viruses in CEC by multistep growth kinetics (Fig. 3a).As expected and based on the non-essential nature of meq for MDV growth in vitro, the mutants showed no significant difference in growth kinetics compared to the parental virus.Revertant viruses also exhibited growth kinetics that were virtually indistinguishable from those of the parental virus (Fig. 3a).Furthermore, the mutants as well as the revertant viruses had similar cell-to-cell spread capabilities compared to the parental virus as was evident from their similar plaque sizes (Fig. 3b).We confirmed protein production from the recoded meq variants in infected CECs by indirect immunofluorescence using the polyclonal rabbit anti-Meq antibody (Fig. S5).
To quantify the level of Meq production from the parental and recoded meq genes during productive infection of permissive cells by flow cytometry, we constructed mutant viruses vDIR L -meq-P2A-EGFP, in which the meq genes were C-terminally tagged with EGFP gene via P2A peptide of Teschovirus A. Because P2A enables highly efficient cotranslational separation of fused proteins, it can be assumed that equimolar amounts of Meq protein and EGFP are produced.In addition, the mutant viruses expressed far-red fluorescent protein mKate2, which was under the control of the early/late promoter of the U L 42 gene of a related Herpesvirus of turkeys (Meleagrid alphaherpesvirus 1).We used the red fluorescence for identification of CECs that were productively infected with mutant viruses, and the green fluorescence, which is proportional to the number of EGFP molecules produced, as a readout for the relative quantification of Meq.
The recovered viruses replicated with comparable efficiency in CECs, as they formed plaques of similar sizes (Fig. S6).Western blotting of CEC lysates infected with mutant viruses confirmed that Meq and EGFP were each produced as a separate protein (not shown).Fluorescent microscopy of CECs infected with mutant viruses verified that EGFP and mKate2 were distributed throughout the cytoplasm and nucleus of infected cells (Fig. S7).
To evaluate production of Meq in productively infected cells we infected CECs with mutant viruses and 4 days p.i., we quantified the amount of Meq in single cells by flow cytometry (Fig. 2c).The amount of Meq produced by viruses that carried meq-W and meq-R genes was similar, but significantly lower compared to those produced by the meq-O or meq-D viruses.On average, Meq levels produced by the meq-O, meq-R and meq-D viruses were 172, 95 and 145 % of the parental virus (Fig. 2c).
The results of these experiments showed that the levels of Meq produced by mutant viruses did not correlate with the levels of Meq produced by transient expression after transfection of plasmids in chicken cells.This means that quantification of protein production in transient transfections did not provide a reliable approximation of Meq protein production by mutant viruses in productively infected cells.

Effects of recoding in vivo
Quantification of Meq production showed that none of the three recoded meq genes produced significantly less Meq than the parental virus in productively infected cells, indicating that mutant viruses and the parental virus might be equally pathogenic for its host.However, because cell culture systems that are currently available for MDV do not provide a good estimation of MDV performance in vivo, we tested pathogenicity and tumourogenicity of recoded viruses in an animal experiment.
We infected 1-day-old specific pathogen free chickens with the parental virus RB-1B vDIR L , the three mutant viruses, vDIR L -meq-D, vDIR L -meq-O and vDIR L -meq-R, and their respective revertants, vDIR L -meq-D-rev, vDIR L -meq-O-rev and vDIR L -meq-R-rev, to assess the influence of recoding.
To examine if recoding of meq affected MDV replication in vivo, we quantified viral genome copy numbers in peripheral blood of the infected chickens at 7, 14, 21 and 28 days p.i. by quantitative real time PCR (qPCR).At each time point, the viral loads in the chickens infected with different viruses were not significantly different, suggesting that all the viruses replicated with similar kinetics in vivo (Fig. 4a).The feather follicle is a unique site in which MDV replicates rapidly and produces large quantities of infectious mature virions.From infected feather follicles, MDV is spread into the environment in the form of infectious dander.It is not well understood how MDV infects the skin, but infectious particles are produced in feather follicles starting 8 days post exposure [38].Under our experimental conditions, contact chickens, which were of the same age as infected chickens, could get infected with MDV earliest when they were at least approximately 1 week old.At this age, chickens are less susceptible to infection with MDV [39].Recoding of meq did not restrict bird-to-bird transmission of mutant viruses, because viral DNA was readily detectable in the blood of contact chickens by qPCR on 14 days p.i. (not shown).In addition, contact chickens that were housed with vDIR L -meq-D-and vDIR L -meq-O-infected chickens developed MD (Fig. 4d).Chickens that were infected with the parental virus vDIR L and the revertant virus vDIR L -meq-Orev were housed together, and therefore shared a single group of contact chickens (n=11).Similarly, chickens infected with vDIR L -meq-D-rev and vDIR L -meq-R-rev were housed in a room with a single group of sentinel chickens (n=11).The MD, as well as tumour incidence for the contact group of vDIR L /vDIR L -meq-O-rev was 20 %, while the MD incidence in contact chickens housed with the vDIR Lmeq-D-rev/vDIR L -meq-R-rev-infected chickens was 27 % (Fig. 4d).These results confirmed that none of the mutants carrying recoded meq genes were attenuated in the chicken.
Owing to their large size, CPBD of entire herpesviral genomes is not possible.Therefore, identification of suitable candidate genes for recoding is an important step.In the present study, we selected MDV as our model virus.The goal of the study was to determine the suitability of meq, the major oncogene of MDV, as a target for CPBD-based attenuation.
It has been suggested that CPBD reduces the translational potential leading to attenuation of recoded viruses [3,15].Our in vitro experiment indeed showed a significant reduction in protein production by meq-D compared to meq-W (Fig. 2a).However, the CPB-optimized and -randomized versions, meq-O and meq-R, of the gene also showed reduced protein production.This was unexpected since minimizing the number of underrepresented codons (CPB optimization) should not reduce protein production.It is worth noting, however, that CPB optimization of the capsid protein L1 of Indiana vesiculovirus also leads to a similar reduction in protein production [11].
Surprisingly, quantification of Meq production by the viral recombinants showed that production of Meq, and possibly also other proteins, cannot be satisfactorily predicted by the CPBD theory, nor reliably approximated via transient transfection experiments (Fig. 2c).While all recoded meq genes produced significantly less Meq than the parental gene in transfected cells, the viral mutants carrying the same recoded genes produced the same or increased amount of Meq as the parental virus.
We tested the recoded viruses in vivo to assess if recoded meq genes might influence production of Meq protein and thus also affect MD progression and tumourigenesis.The replication kinetics of the recoded viruses as well as their respective revertants in vitro were not different from those of the parental virus (Fig. 3a).This was consistent with the non-essential nature of meq in virus replication.MD progression in the chickens infected with the recoded viruses was slightly slower, as these chickens took on an average 17 days longer to develop MD 50 compared to the parental and revertant groups (Fig. 4b).At the end of the experiment at 90 days p.i., however, there was no significant difference between the MD incidence in the chickens infected with mutant, revertant and parental viruses.The tumour incidence in the viruses infected with the recoded viruses was also lower but not significantly different than that in the chickens infected with the revertant and parental viruses (Fig. 4c).
In summary, neither CPB-optimization nor -deoptimization of the meq gene led to quantifiable changes in the biological properties of MDV.We do not have any plausible explanation why Meq production from the recoded genes did not match the theoretical predictions.In our previous work, where we studied recoded U L 30 gene variants, the CPB score of the recoded genes correlated closely with the level of protein production in transiently transfected cells as a CPB-optimized U L 30 gene produced more and CPB-deoptimized U L 30 genes produced less U L 30 [24].In addition, in the context of viral infection, the severity of MDV attenuation corresponded to the degree of CPBD of U L 30.
In a recent study that also evaluated CPBD as a method for MDV attenuation by targeting the U L 49 and U L 54 genes [40], protein production from moderately and severely CPB-deoptimized variants of these genes was quantified in transient transfection experiments.While the severely deoptimized U L 54 produced significantly less protein than the parental gene, the moderately deoptimized U L 54 produced significantly more protein.In addition, both U L 49 variants showed only moderately decreased protein production.Unexpectedly, a virus carrying a moderately deoptimized U L 54 gene was more attenuated than the mutant viruses with the severely deoptimized U L 54 in chickens.The results of this study, similar to the situation reported here, also suggest that the level of CPBD is neither a reliable predictor of protein production from recoded genes, nor attenuation of mutant viruses carrying such genes.
We show that the Meq protein levels produced ectopically in transfected cells or in the context of viral infection did not follow the predictions dictated by the CPBD theory (CPB-optimized genes produce more and CPB-deoptimized genes produce less protein).This observation and results obtained from the study of CPB-deoptimized MDV genes U L 49 and U L 54 [40] question the reliability of such predictions and indicate that molecular mechanisms that lead to attenuation by CPBD must be better understood, before this method can be fully utilized for development of efficient and safe live viral vaccines.
It is interesting to note that CPBD of the essential genes U L 30, which encodes the catalytic subunit of the viral DNA polymerase, nor U L 54, which encodes the ICP27 protein has eliminated tumourigenic potential of MDV [24,40].However, because mutant viruses showed a moderate level of attenuation (reduced mortality and reduced tumour incidence), we speculate that for large DNA viruses such as MDV, CPBD of several genes might lead to better attenuation.One of the main advantages of CPBD over alternative attenuation methods is that it enables uncomplicated modification of essential genes.High replication levels in the initial stages of infection are important for tumour formation, because high virus levels ensure establishment of a sufficient number of latently infected cells, which in turn lead to lymphomas [17].Therefore, genes that play essential roles during initial phases of viral infection are likely candidates to be examined by CPBD.
The animal experiments showed that recoding of meq did not affect functionality of Meq or its potential splice variants, because all three mutant viruses were as pathogenic as the parental virus.In addition, these experiments also confirmed that neither LORF6 nor ORF encoding 23 kDa nuclear protein played any role in replication, pathogenesis and transformation of mutant viruses, which most likely means that neither of these two hypothetical genes encodes a functional product.
While many different RNA splice variants containing coding sequences of the meq, LORF5a, LORF4 and vIL-8 genes were detected in vitro and in vivo, only a fusion protein containing the N-terminal part of Meq and C-terminal part of vIL-8 protein (exons II and III), termed Meq/vIL8, was detected in MDV-infected cells [34].This raises the question whether any of the identified meq or vIL-8 splice variants are functionally important for MDV replication or pathogenesis.Examining mutant viruses rendered unable to form a splice variant between these genes by eliminating the predicted splice donor and acceptor sites could provide answers to such questions.

Recoding of the MDV meq gene
Codon pair score (CPS) is defined as the natural logarithm of the ratio of the observed to the expected occurrences of a given codon pair [3].Chicken CPS for all 3721 possible codon pair combinations have been previously published and were calculated using 15 762 predicted chicken protein coding genes (Gallus gallus, Breed Red Jungle fowl, line UCD001, version 4.0) [6].Using the chicken CPS, we calculated the average CPScodon pair bias scores (CPB scores) for each of the 15 762 chicken genes and 112 MDV genes.(Fig. 1a) Using the calculated chicken CPS, we wrote a computer program that can recode a given protein-coding sequence to obtain a new sequence with a desired CPB score without changing the amino acid sequence [24].The program achieves this by reshuffling the available codons in the sequence, thus preserving its codon bias.We used the program to obtain three versions of the MDV meq gene: meq-D, meq-O and meq-R, with their respective CPB minimized (deoptimized), maximized (optimized) and preserved (randomized) with respect to the parental gene, meq-W.
Recoding a sequence to maximize or minimize its CPB is computationally tedious given the large number of possibilities for encoding a certain protein.A near-optimal solution to this problem can be found quickly using heuristic and meta-heuristic approaches.Similar to the algorithm used by Coleman et al., our recoding program uses a fast metaheuristic algorithm, called simulated annealing, to find a nearoptimal approximation of the actual CPB extreme [3,41].
To avoid inadvertent changes in RNA secondary structure, the program controls the predicted Gibbs free energy (DG) of folded RNA within a narrow range (Fig. 1b).This is to ensure that any changes in protein production are a result of CPB alterations and not of extensive RNA secondary structures.To confirm this, the sequences were scanned through mFold program [36], exactly as described [3].In short, an array of short sequences was generated from the coding sequences, 100 nucleotides in length with an 80nucleotide long overlap with the preceding as well as the succeeding fragments.DG is calculated for each of these fragments by the program.If any of these fragments had DG lower than À30 kcal mol À1 , we recoded the fragment to raise its DG.The final recoded sequences had a DG distribution similar to that of the parental sequence.The recoded sequences are available in the supplementary material (File S1).The recoded sequences were synthesized (BioBasic, Canada) and cloned in the cloning vector pUC57.

Cells and viruses
Chicken cells were grown and maintained at 37 C in a 5 % CO 2 environment.Primary CECs were grown in minimal essential medium with Earle's salts supplemented with 1-10 % FBS, 100 U ml À1 penicillin and 100 µg ml À1 streptomycin.Chicken epithelial cells 8E11 (MicroMol, Germany) were grown in Dulbecco's modified Eagle's medium supplemented with 10 % FBS, 100 U ml À1 penicillin and 100 µg ml À1 streptomycin.
For reconstitution of viruses, CECs were transfected with purified bacterial artificial chromosome (BAC) DNA (see Generation of recombinant viruses) along with a Cre recombinase expression vector using polyethyleneimine [42].The Cre recombinase ensured removal of the BAC cassette flanked by loxP sites from the genomes of the recovered infectious virus.Removal of the BAC cassette was confirmed by PCR as previously described [33].Virus was grown on CECs and titrated stocks of infected cells were stored in liquid nitrogen.

Generation of recombinant viruses
The meq mutant viruses were generated from pDIR L , an infectious BAC clone of the highly oncogenic RB-1B strain of MDV in which the inverted repeat long (IR L ) has been deleted [43].Thus, the BAC possesses only one copy of meq, circumventing the possibility of viral clones with two distinct versions of meq at the two loci.Moreover, the reconstituted virus has been shown to restore the deleted IR L as early as two passages in cell culture [43].The mutant BACs were generated in two steps as follows: (i) the parental meq gene was replaced by homologous recombination with an ampicillin resistance (amp r ) gene.(ii) amp r was then replaced by one of the recoded meq genes by two-step Red-mediated en passant mutagenesis as previously described [37].The initial deletion of meq was necessary to avoid undesired recombination events during mutagenesis owing to the high nucleotide similarity (Table 1) between the recoded and parental meq genes.Revertant BACs were similarly generated from the mutants by replacing the recoded meq gene with amp r , which in turn was replaced by the parental meq-W gene.
To enable quantification of Meq production in the viral context, we constructed mutant viruses that expressed green -EGFP and red -mKate2 (Evrogen) fluorescent proteins.The parental and recoded meq genes were C-terminally tagged with EGFP via the P2A peptide.mKate2 was placed under the control of the early promoter of the U L 42 gene, encoding DNA polymerase processivity subunit, of Meleagrid alphaherpesvirus 1 (MeHV-1; GenBank accession: NC_002641.1,nucleotide position: 89 433-90 055) and early mRNA polyadenylation signals of Macaca mulatta polyomavirus 1 (SV40; GenBank accession: NC_001669.1,nucleotide position: 2775-2533).We determined that the MeHV-1 U L 42promoter provides high-level expression of transgenes during productive infection of CECs with MDV (unpublished).The mKate2 expression cassette (MeHV-1 U L 24 promoter -mKate2 -SV40 polyadenylation signals) was inserted between genes encoding chloramphenicol acetyltransferase and redF of the BAC vector.
All mutant and revertant BAC clones were analysed by RFLP for integrity of the genome.Complete and correct insertion of the target meq genes was confirmed by PCR and sequencing using primers listed in the supplementary material (Table S1).

Construction of expression plasmids
To assay protein production by the recoded and parental meq genes, we used a eukaryotic dual expression vector, pVitro2-MCS (InvivoGen).An EGFP gene was fused in frame to the 3¢ end of each meq ORF and cloned under the influence of human ferritin H/mouse elongation factor 1 promoter.An mTagBFP gene (Evrogen) was cloned into the second cloning site under the influence of the human ferritin L/chimpanzee elongation factor 1 promoter.

Quantification of protein production by flow cytometry
To quantify ectopic Meq production, expression plasmids pVITRO2-mTagBFP-meq-EGFP or control plasmids pEGFP-N1 and pVITRO2-mTagBFP were transfected into subconfluent CECs, or 8E11 cells grown in six-well plates with Lipofectamine 3000 (Invitrogen) in triplicates.After 24 h, the cells were trypsinized, washed with PBS and analysed by flow cytometry.To quantify Meq production in the context of viral infection, 1Â10 6 CECs seeded in a well of a six-well plate were infected with 1000 p.f.u. of mutant viruses expressing Meq-P2A-EGFP and mKate2.After 4 days the cells were trypsinized, washed and analysed by flow cytometry.
The cell suspensions were analysed using a CytoFLEX flow cytometer (Beckmann Coulter) equipped with 405, 488 and 561 nm lasers, and the following bandpass filters: 450/45 nm for TagBFP, 525/40 nm for EGFP, and 610/20 nm for mKate2.A minimum of 50 000 events were collected for each sample.Data were analysed using CytoFLEX CytExpert Software version 1.2.11.0.In transient transfections, EGFP provided a measurable marker for Meq-EGFP protein production while TagBFP served as a control for transfection efficiency.The mean fluorescent intensity of EGFP was normalized using the mean fluorescence intensity of TagBFP in the respective samples.To quantify Meq production in cells infected by viruses, the mean fluorescent intensity of EGFP was normalized against the mean fluorescence intensity of mKate2.Relative protein production (percentage) was calculated with respect to the parental protein, Meq-W.Artifacts due to the overlap of the emission spectra of TagBFP/EGFP and EGFP/mKate2 were eliminated by colour compensation using positive controls.Duplets and cell debris were excluded from analyses.

Immunofluorescence, plaque size assay and multistep growth kinetics
To assess the cell-to-cell spread of the virus, plaque size assay was performed in three independent double-blind experiments as previously described [44].Briefly, CECs (1Â10 6 per well) were mixed with 50 p.f.u. of each virus and seeded onto six-well plates.Plaques were visualized using indirect immunofluorescence 6 days p.i. as follows: cells were fixed with 2 % paraformaldehyde, permeabilized with 0.1 % Triton-X 100 and blocked with 3 % BSA in PBS.The cells were then incubated with a chicken anti-MDV serum (dilution 1 : 2000) for 1 h followed by washing with PBS and incubation with anti-chicken IgG-Alexa Fluor 488 (dilution 1 : 2000; Invitrogen) for 45 min.Following a second wash with PBS, plaques were visualized and images were taken at 100-fold magnification using an inverted fluorescence microscope (Axiovert S100, Zeiss).ImageJ software version 1.48 v [45] was used to measure the plaque areas from which plaque diameters were calculated.
To assess the replication properties of the reconstituted viruses, multistep growth kinetics were performed as described previously [44].Briefly, CECs were infected 100 p.f.u. of each virus in duplicates for each time point.Cells were trypsinized every day from day 1 through 6 p.i. and serial 10-fold dilutions were inoculated onto fresh CECs.After 6 days, plaques were visualized by indirect immunofluorescence (see above), counted and titres were determined.

Animal experiments
One-day-old VALO-specific pathogen-free layer chickens (Lohmann Tierzucht, Germany) were subcutaneously infected in a blind fashion with 5000 p.f.u. of the parental virus, vDIR L ; the mutants: vDIR L -meq-D, vDIR L -meq-O, vDIR L -meq-R; or their respective revertants: vDIR L -meq-Drev, vDIR L -meq-O-rev, vDIR L -meq-R-rev.The infected chickens were housed with uninfected contact chickens to assess the spread of the virus through shedding.Food and water were provided ad libitum.Animals were observed for MD symptoms.To examine tumour formation, necropsies were performed either after manifestation of clinical symptoms or at the end of the experiment 90 days p.i.

Quantification of MDV genome copies in chicken blood
Blood samples (40 µl) were collected from wing veins of the infected chickens in a 100 mM EDTA solution (20 µl) in deep well plates.Blood was collected from the same eight chickens per infected group on days 7, 14, 21 and 28 p.i. Dead animals were replaced.DNA was isolated using E-Z96 96-well blood DNA isolation kit (Omega Biotek).MDV genome copies in the chicken blood were quantified by qPCR using specific primers and probes for the MDV gene ICP4 [29].Copy numbers of the cellular nitric oxide synthase (iNOS) were used for normalization [29].

Statistics
Statistical analysis was performed using GraphPad Prism 7.02.Data were first tested for normal distribution.The data for protein production and plaque size assays were analysed for statistical significance by one-way ANOVA, with Bonferroni correction for multiple comparisons.Growth curves and qPCR data were analysed using the Kruskal-Wallis non-parametric test for significance.Tumour incidence data were analysed by Fisher's exact test with Bonferroni correction for multiple comparisons.

Fig. 1 .
Fig. 1.Distribution of CPB scores and free-folding energy (DG) of parental and recoded MDV meq genes.(a) Distribution of calculated CPB scores of 15 762 predicted chicken, 112 MDV and recoded MDV meq genes.Each light blue circle represents the calculated CPB score of a single chicken-protein-coding gene plotted against its protein length (amino acids).The mean CPB score of 15 762 chicken genes is 0.0755.Dark blue circles represent 112 predicted MDV-protein-coding genes, with an average CPB score of À0.0646.The pink circle represents the parental MDV meq gene (meq-W).The black cross, which overlaps with meq-W, represents the CPB randomized meq gene (meq-R).The red diamond and the green square represent CPB deoptimized (meq-D) and optimized (meq-O) meq genes, respectively.(b) DG of the RNA encoded by meq-D, meq-O and meq-R is similar to that of the RNA encoded by meq-W.DG of any 100base-pair fragment derived from the three recoded genes is not lower than À30 kcal mol À1 .The parental and recoded sequences have a similar mean DG: meq-W=À20.35,meq-D=À20.33,meq-O=À20.89and meq-R=À20.13kcal mol À1 .
Furthermore, MD incidence was not significantly different among the chickens infected with different viruses.The time until 50 % of the infected chickens developed MD (MD 50 ) was slightly increased in the case of mutant viruses.MD 50 was reached at 67, 71 and 72 days p.i. in the groups infected with vDIR L -meq-D, vDIR L -meq-O and vDIR L -meq-R, respectively, while it took 59 days in chickens infected with parental vDIR L , and 48, 56 and 50 days in the case of the respective revertant viruses, vDIR L -meq-D-rev, vDIR Lmeq-O-rev and vDIR L -meq-R-rev.(Fig.4b).

Fig. 2 .
Fig. 2. Quantification of Meq production from the recoded meq genes by flow cytometry.(a) CECs and (b) 8E11 were transiently transfected with dual expression plasmids pVITRO2-mTagBFP-meq-EGFP, which express meq-EGFP and mTagBFP from independent promoters.Fluorescence intensity of Meq-EGFP fusion protein was used as a measure of Meq expression levels.The mean fluorescence intensity of EGFP was normalized against that of TagBFP.Percent relative expression was calculated with respect to the normalized mean fluorescence intensity of Meq-W-EGFP.Measurements were obtained in six independent experiments.Protein production from all recoded meq genes was significantly lower in comparison to production from the meq-W gene (n=6, one-way ANOVA, * indicates P<0.016).(c) Quantification of Meq production by mutant viruses in productively infected CECs.Cells were infected with mutant viruses that harboured mKate2 and meq-P2A-EGFP genes.The mean fluorescence intensity of EGFP was normalized

Fig. 3 .
Fig. 3. Characterization of mutant MDV with recoded meq genes.(a) Multistep growth curves of the parental (vDIR L ), mutant (vDIR Lmeq-D, vDIR L -meq-O and vDIR L -meq-R) and revertant viruses (vDIR L -meq-D-rev, vDIR L -meq-O-rev and vDIR L -meq-R-rev).Viral titres are shown as geometric means with SEM.(n=4, P>0.01, Kruskal-Wallis H test).(b) Diameters of plaques formed by the parental, mutant and revertant viruses were not significantly different 6 days post infection (p.i.) in CECs (n=153, one-way ANOVA, P>0.01).Relative plaque diameters were calculated with respect to the median plaque diameter of the parental virus and their distribution is represented as box plots.

Fig. 4 .
Fig. 4. In vivo effects of CPBD of the MDV meq gene.(a) Viral replication in vivo.Blood samples were collected from chickens infected with the indicated viruses on days 7, 14, 21 and 28 p.i. Viral titres in the blood of eight chickens per group are represented as MDV genome copy number per 1Â10 6 cells (Kruskal-Wallis H test, P>0.01).(b) MD incidence in chickens infected with the parental (vDIR L ), mutant (vDIR L -meq-D, vDIR L -meq-O, vDIR L -meq-R) and revertant (vDIR L -meq-D-rev, vDIR L -meq-O-rev, vDIR L -meq-R-rev) viruses (Mantel-Cox test, P>0.01).(c) MD and gross lymphomas in chickens infected with the indicated viruses (Fisher's exact test, two-sided, P>0.01).(d) MD in contact chickens that were housed with the chickens infected with the indicated virus (Fisher's exact test, two-sided, P>0.01).MD incidence is shown as percentage of chickens per group.
nt sequence

Table 1 .
Characteristics of the parental and recoded meq genesAll sequences comprise of the same codons, thus preserving the codon bias, but the order of the codons in individual sequences is different.CPB score, codon pair bias score; CpG, number of CpG dinucleotides.