Human cytomegalovirus (HCMV; family Herpesviridae, subfamily Betaherpesvirinae, genus Cytomegalovirus, species Human herpesvirus 5) is ubiquitous and host–specific. Infection is asymptomatic for most people, but can result in serious disease in immunocompromised individuals and congenitally infected newborns. The minimally passaged Merlin strain is considered best to represent wild–type HCMV, and has a 236 kbp genome that is predicted to contain approximately 165 protein–encoding genes [Dolan et al., 2004].

Most genes are highly conserved in sequence between HCMV strains, but a number of genes predicted to encode membrane-associated or secreted proteins are characterized by a striking degree of variability, as revealed by examination of individual genes [reviewed in Pignatelli et al., 2004] and by whole genome comparisons [ Murphy et al., 2003; Dolan et al, 2004]. Various studies have attempted to make connections between the genotypes of hypervariable genes and disease outcome, and overall the conclusions reached are unclear or contradictory [reviewed in Puchhammer-Stöckl and Görzer, 2006]. The apparently random association of genotypes at different loci (that is, the absence of linkage disequilibrium), presumably as a result of recombination during HCMV evolution, limits any conclusions to the specific gene under study, except in some cases where genes are very close to each other or encode interacting hypervariable proteins [Rasmussen et al., 2002, 2003]. A further complicating factor is the occurrence of multiple HCMV genotypes in individuals, including immunocompromised patients, pregnant women and congenitally infected newborns [for recent papers, see Rasmussen et al., 2003; Hassan-Walker et al., 2004; Stanton et al., 2005; Puchhammer-Stöckl et al., 2006; Iwasenko et al., 2007].

One of the most variable HCMV genes is UL146, which encodes a chemokine designated vCXC-1. This gene is variable throughout its length [Penfold et al., 1999; Prichard et al., 2001; Hassan-Walker et al., 2004; Arav-Boger et al., 2005, 2006; Stanton et al., 2005; He et al., 2006; Lurain et al., 2006], and 14 genotypes have been catalogued [Dolan et al., 2004]. In strain Toledo, UL146 encodes a functional chemokine that is capable of neutrophil degranulation, chemotaxis and calcium mobilization. This protein contains an ELRCXC motif, which has been shown to be essential for receptor binding and IL-8 activity [Clark-Lewis et al., 1991]. vCXC-1 binds to human CXCR2 and is comparable in its activities to CXC chemokines IL-8 and gro-α [Penfold et al., 1999]. The function of vCXC-1 may be to facilitate dissemination of the virus through its ability to attract monocytes to the initial site of infection. Thus, the virus could undermine the effectiveness of antiviral immunity by manipulating the host chemokine system and, together with other virus–encoded molecular mimics, suppressing the immune system.

The most variable gene in the vicinity of UL146 is UL139, which is located 5.2 kbp distant and is predicted to encode a type I membrane glycoprotein. Variability is concentrated in a region of the ectodomain [Dolan et al., 2004]. A recent study of 26 HCMV strains isolated in China described three major genotypes, with two of these divided into subtypes [Qi et al., 2006].

The aims of the present study were to investigate whether additional UL146 genotypes exist in a large number of clinical samples obtained from a wide range of locations and clinical settings, and to define the range of UL139 genotypes in these samples. Ancillary interests were to examine the relative frequencies and geographical distribution of genotypes, to assess whether infections with more than one HCMV strain are common, and to investigate the evolution of UL146 and UL139.

This study focused on the genotype definitions, frequencies, occurrence in mixed infections, geographical distribution and evolution (in terms of linkage disequilibrium and mode of selection) of two hypervariable HCMV genes, UL146 and UL139. Totals of 182 UL146 and 183 UL139 sequences were obtained from a large panel of clinical isolates collected from Africa (South Africa and The Gambia), Asia (Hong Kong), Australia and Europe (various countries). These were used in all analyses, and were supplemented by 168 previously published UL146 and 117 UL139 sequences in analyses of genotype definitions, frequencies and mode of selection.

The UL146 sequences fell into the 14 genotypes described previously [Dolan et al., 2004], and no new genotypes were discovered. Twelve genotypes contained the ELRCXC motif, which has been shown to be essential for receptor binding and IL–8 activity [Clark-Lewis et al., 1991], and 2 contained the NGRCXC motif, which has been shown to be important for interaction with T and B cells [Baggiolini et al., 1997]. The latter genotypes (G5 and G6) are relatively rare, being present in approximately5%of samples. It is not known whether the UL146 genotypes possess different biological properties, and studies to investigate this question are required.

The UL139 sequences grouped into eight genotypes.A recent analysis of 26 clinical samples [Qi et al., 2006] described three major groups (G1, G2, and G3), two of which were divided into subgroups (G1 into G1a, G1b and G1c and G2 into G2a and G2b). Subgroups G1b and G1c in the previous study correspond to G1 in the present study, subgroup G1a corresponds to G4, subgroups G2a and G2b correspond to G6 and G2, respectively, and G3 is named identically in both studies. Thus, apart from the differences in nomenclature, the subgroups [Qi et al., 2006] correlate with a subset of the genotypes in the present study, except that the closely related subgroups G1b and G1c in the former are amalgamated as G1 in the latter. Most of the variation in UL139 is due to substitutions or deletions of variable size near theN terminus. This region is rich in S and T residues that are potentially susceptible to O–glycosylation, and also contains NXS or NXT motifs that are potentially susceptible to N–glycosylation. This suggests that selection may have focused primarily on glycosyl side chains rather than the underlying amino acid sequence. A similar feature characterizes other variable glycoprotein genes, such as UL73 (encoding glycoprotein N(gN)) and UL74 (encoding glycoprotein O(gO)) [Pignatelli et al., 2001, 2003; Mattick et al., 2004].

A region of sequence identity (SETTTGTSSNSS in Fig. 1A) has been noted between the HCMV UL139 protein and CD24, a cellular glycosyl phosphatidylinositol–linked glycoprotein that is involved in B cell activation [Qi et al., 2006]. It is difficult to assess the significance of this observation, especially since 9 of the 12 residues are S or T and the region is not conserved in CD24 orthologues from other mammals. However, the sequence is present in all of the UL139 genotypes identified in the present study, except for G5, and also in the homologous protein from chimpanzee cytomegalovirus (CCMV) (Fig. 1A). Variation in glycosylation has been observed in CD24 and has been linked to differences in cell and tissue specificity [Poncet et al., 1996]. Additional roles for CD24 in apoptosis and cell adhesion have also been suggested, and more recently in regulating the responsiveness of a chemokine receptor, CXCR4 [Schabath et al., 2006; Smith et al., 2006]. The possibility that UL139 may be a CD24 homologue remains intriguing, but, in the absence of functional data, unproven.

Studies of HCMV genotype frequency, including the present one, are usually based on the use of conserved PCR primers, and face limitations as a result. Firstly, there is no guarantee that all genotypes will be detected, since primers are chosen on the basis of alignments of available sequences. Secondly, samples containing more than one strain yield mixed sequences,which when cloned are recovered approximately in proportion to their abundance (although stochastic processes may introduce bias during PCR). Therefore, the absence of a genotype from a particular sample cannot be assured. If anyUL146 or UL139 genotypes have escaped recognition, they may emerge from future studies involving different primers or from whole genome sequencing exercises.

As found in previous studies [reviewed in Puchhammer-Stöckl and Görzer, 2006], mixed infections with different HCMV strains were common. In some samples, a single UL139 genotype and multiple UL146 genotypes, or vice versa, were detected. This could be due to different strains happening to contain the same genotype at one locus but not at the other, or to the limitations of amplifying sequences present as mixtures in unequal proportions. Some samples tested more than once were found to contain additional genotypes not detected in the first experiment, suggesting that the number of mixed infections was underestimated by the methodology used. Mixed infections were more frequently detected from certain regions, namely Hong Kong, South Africa and, to a lesser extent, The Gambia. It is possible that this is a result of higher transmission frequencies. In one study [Beyari et al., 2005], a higher seroprevalence of HCMV in children in Malawi compared to European countries and the USA was taken as possibly reflecting greater opportunities for transmission, although multiple genotypes were detected in only a small number of samples.

The occurrence of mixed infections is being recognized increasingly as potentially significant to the biology of HCMV. This feature adds to the limitations inherent in studies of whether particular genotypes are associated with disease outcome; other features include the number, origin and pathological categorization of samples, the choice of gene, the absence of linkage disequilibrium, and host factors. In light of these limitations, our opinion is that robust evidence in favor of any association between genotype and pathology has proved elusive in the literature. Further work utilizing genotype–specific approaches is required to explore the true frequency of mixed infections, both to validate studies of this type and to determine whether mixed infections have geographical or biological correlates.

Similar to the conclusions drawn from a study on UL73 (encoding gN) [Pignatelli et al., 2003], no statistically significant association of UL146 or UL139 genotypes with geographical origin arose from the analysis. However, this may reflect low sample numbers (albeit much larger than those utilized in previous studies on UL146 and UL139) and the lack of detailed information on ethnic origin. Likewise, investigation of linkage disequilibrium between UL146 and UL139 genotypes was compromised by the small sample number (60) relative to the large number of possible genotype combinations (112). However, no evidence for linkage disequilibrium was obtained, indicating the involvement of recombination in HCMV evolution since the genotypes arose. Taking into account the size of the HCMV gene complement, we agree with the view that very many strains are likely to be circulating in the world [Rasmussen et al., 2003].

The extensive divergence between genotypes and the consequent inability to produce reliable sequence alignments for both UL146 and UL139 in the hypervariable regions compromised assessments of the role of positive selection in generating the genotypes. In contrast, variation within genotypes is low, and identical nucleotide sequences were obtained from geographically distant individuals. The analysis suggests that constraint has been the predominant factor in evolution within genotypes, with positive selection detected only marginally. A previous study [Arav-Boger et al., 2005] involving 30 sequences also concluded that UL146 has evolved under constraint. The sequences of hypervariable genes are stable on short timescales in patients [Hassan-Walker et al., 2004; Stanton et al., 2005] and cell culture [Lurain et al., 2006], consistent with the perception of herpesvirus genomes as relatively slowly evolving [McGeoch et al., 2006]. The most likely scenario for the evolution of HCMV emerging from the literature and from the present study is that the genotypes developed in early human populations (or even earlier), becoming established via founder or bottleneck effects, and have spread, recombined and mixed worldwide in more recent times, with mixed infections being common.