Attenuation of dengue (and other RNA viruses) with codon pair recoding can be explained by increased CpG/UpA dinucleotide frequencies

Shen et al. (1) describe the modification of three regions of the Dengue virus type 2 (DENV-2) genome to match the codon pair use of insect genes rather than those of mammals. We have previously shown that such recoding also modifies frequencies of CpG and UpA dinucleotides and have proposed it is this, rather than codon pair use, that restricts replication in mammalian cells (2). We argue that DENV-2 mutants are attenuated for cell culture and in vivo replication in mice through the same mechanism. Our conclusions are based on the following observations.

First, recoding for insect codon pair use increased frequencies of CpG and UpA dinucleotides by ∼3- and 1.5-fold, respectively, in all three modified regions (Table 1). The reduction in replication was proportionate to the numbers of dinucleotides added, consistent with degrees of attenuation of previously constructed virus mutants modified for dinucleotide frequencies or codon pair use (2).

Table 1.

Alteration in CpG and UpA dinucleotide frequencies on optimization for insect cell codon pair use

Sequence	Region	Position	Length (%)*	CpG (O/E)†	UpA (O/E)†
Wild-type	E	940–2421	1,482 (14.0%)	20 (0.268)	55 (0.528)
Recoded				87 (1.164)	86 (0.817)
Wild-type	NS3	4528–6042	1,515 (14.1%)	32 (0.404)	68 (0.652)
Recoded				99 (1.252)	111 (1.062)
Wild-type	NS5	7576–9942	2,367 (22.1%)	62 (0.516)	91 (0.566)
Recoded				147 (1.229)	134 (0.832)

^*Length and proportion of genome modified.

^†Numbers of CpG and UpA dinucleotides (observed/expected ratio in parentheses); increased numbers indicated in bold.

Second, although we welcome the acknowledgment by Shen et al. (1) that codon pair use is related to dinucleotide frequencies (see the second paragraph in the introduction to ref. 1), the text (the third paragraph in the Discussion of ref. 1) misinterprets the findings of our previous study, which separated the effects of these two compositional variables on virus replication (2). Briefly, the “Min” series of mutants we constructed incorporated disfavored codon pairs, while at the same time keeping CpG and UpA frequencies at wild-type (Min-E) or at increased frequencies (Min-U, Min-H). In both replication kinetic and competition assays, Min-E was equally fit as the wild-type virus despite codon pair deoptimization, whereas Min-U and Min-H were attenuated to degrees proportional to numbers of additional CpG and UpA dinucleotides incorporated (2). Thus, we successfully disentangled the effects of these two modifications and showed that dinucleotide frequencies, and not consequent codon pair bias, determine replication fitness in mammalian cells.

Third, if the effects of codon pair deoptimization are mediated through viral translation, then Shen et al. (1) should account for the variability in attenuation of recoded mutants between mammalian cell-lines (figures 2 C and D, and 3D in ref. 1). The authors suggest that greater sensitivity of BHK cells to DENV-2 infection masked replication differences, but this is not supported by the similarity in replication kinetics in the two cell-lines (y axis in figure 2 of ref. 1). These data (plus those involving the Jak inhibitor) brings Shen et al.’s conclusions into question because neither cell-line nor innate immune pathways should influence attenuation if it originates from differences in translation efficiency.

In summary, we do not believe that Shen et al. (1) have demonstrated that attenuation of DENV-2 in mammalian cells and mice is the result of codon pair deoptimization. Resolving this issue is critical if such recoded viruses are to be used as attenuated vaccines. If the viruses are not intrinsically defective but are simply more visible to the host defense, then they cannot be regarded as safe for use in individuals with immunodeficiency or who show any natural genetic variability in recognition pathways for viruses with high CpG/UpA compositions. That such variability does exist is attested by the behavior of BHK cells in the present study (1).