Here we present the application of a broad-spectrum technique for the universal detection and differentiation of variable genotypes of the gB and UL33 genes of HCMV. All known HCMV gB and UL33 genotypes could be amplified with two sets (CMV-gBp1, CMV-UL33) of universal primate CMV primers (Table 1). Most importantly, a novel UL33 genotype (UL33-5) was discovered that differs significantly from the known genotypes (Figure 3, 4). UL33-5 could be identified in a total of 16 clinical HCMV isolates, one of which was plaque-purified (HCMV-C5).
Both the isolate HCMV-C5 and the UL33-5-positive specimen S12 contained the same gBCLS/gBN combination. While gBCLS was similar to that found in AD169, gBN was not (Figure 5). An identical gBN sequence was detected previously in an HCMV isolate of an AIDS patient , however, the CLS sequence of that isolate was not analysed. The gBCLS/gBN found in HCMV-C5 (and the UL33-5-positive specimen S12) are reported here for the first time. During the preparation of this manuscript, two HCMV isolates were deposited in GenBank with almost identical genotype sequences for UL33 and UL55 (GQ396662 and GQ221975, respectively). It is inferred that the novel UL33 and gB sequences may be not rare.
The discovery of this novel HCMV variant indicates that the primer sets are of a sufficiently broad detection range. This is underscored by our recent discoveries of novel CMV gB and UL33 sequences in chimpanzees and gorillas where these primer sets were employed (CMV-gBp1: ; CMV-gBp2 and CMV-UL33: B. Ehlers, unpublished data). We infer from these findings that the universal primers would also detect unknown, even more aberrant HCMV genotypes and variants.
The differentiation of HCMV genotypes from mixed infections was achieved by performing PCR in the presence of LNA-oligonucleotides. Previously, LNA-supplemented PCR was used for the discovery and differentiation of non-human primate herpesviruses in multi-infected samples . It was adapted here to differentiate HCMV genotypes, and proved to be equally effective. By this approach, single samples revealed up to three genotypes. The most important finding was the discovery of the novel UL33-5 genotype with LNA-supplemented consensus PCR in a clinical sample (S12) in which only the UL33-1 genotype was detected with consensus PCR alone (Tab. 3).
In principle, known HCMV genotypes can be amplified with specific primers. However, with more than ten N-terminal and CLS genotypes already known, this could become very laborious, making LNA-supplemented PCR approaches with degenerate primers as those presented here feasible alternatives. With this method, every genotype should be detectable, including unknown ones. In addition, for genotypes preferentially present in mixed infections or for the comprehensive analysis of sample material from rare patients, the LNA-based differentiation of genotypes from single samples is the method of choice.
Of note, the gBCLS-5 genotype has been detected in five HIV-positive patients . In accordance with this, we detected gBCLS-5 in two HIV-positive patients (S1, S3; Table 3). At the N-terminus, the gB sequence of S3 was identical to gBN sequences which were retrieved from two HIV-infected patients  (since these authors did not analyse the CLS, further comparisons could not be done). These findings warrant further investigation on a possible association of HCMV carrying the gBCLS-5 genotype with HIV.
For a long time, gB has been in the focus as a suitable target antigen that might be utilized for vaccination and thus prevention of maternal cytomegalovirus infection. In a phase 2 placebo controlled, randomized and double-blind trial, a gB-based vaccine was found to have a vaccination efficacy of 50% . Although the key epitopes for neutralizing antibodies are supposed to be conserved between the presently known gB genotypes , the core region of the antigenic determinant 2  is not entirely conserved (Figure 5). HCMV variants may exist that are even more diverse and only insufficiently neutralized by the antibodies elicited by the gB vaccine. The broad-spectrum PCR methodology presented here may prove useful in detecting these variants.
Whether the GPCR protein encoded by UL33 has a natural ligand or functions merely through constitutive, i.e., ligand-independent intracellular signalling is not known. All published functional tests on HCMV UL33 were undertaken with the UL33 GPCR encoded by the laboratory strain AD169 [30, 31]. When the five UL33 amino acid sequences were compared, intra-genotypic variation was most pronounced in certain sections. These correspond to the putative extracellular parts of the GPCR, whereas the intracellular parts show very low variability (Figure 4). Ligand interaction was attributed to the extracellular domain and extracellular loops (primarily loops 2 and 3) . Since these domains revealed the highest variability in UL33, it may be useful to additionally examine UL33 genotypes other than AD169. This might help to identify a ligand for UL33.
Mixed infections with several herpesvirus genotypes are frequent, and especially immunocompromised patients are affected [8, 13–19, 33–35]]. In the latter situation, infections with multiple HCMV gBCLS genotypes correlate with an increased rate of graft rejection, higher viral loads, and frequency of infection with other herpesviruses [33, 34, 36, 37]. For therapeutic purposes, it is therefore important to early identify genotypes that bear a higher virulence potential. Relationships between certain HCMV glycoprotein genotypes and organ manifestations or outcome of HCMV infections have been looked into. However, establishing correlations has frequently been difficult because of limited testing sensitivity with the majority of methods using a defined, narrow detection spectrum. Only two groups of investigators used partially degenerate primers [36, 37]. Methods enabling the detection of multiple genotypes in a single sample were applied in two studies where approximately 25% to 29% of the samples were found to be multi-infected with HCMV strains of different gB genotypes [8, 38]. In accordance with this, more than one HCMV gB genotype was detected in 32% of the samples S1-S28 in the present study. Thus, an improved technique for universally detecting and differentiating known and unknown genotypic HCMV variants in clinical specimens would enhance understanding multiple HCMV infections, the reasons for their occurrence, and the pathogenicity they cause. In contrast to detection systems used in earlier studies [1, 8, 36–39], the methodology presented here is the first that meets this requirement. It allows to comprehensively detect and differentiate cytomegaloviruses and might help to identify a correlation between a given genotype or a certain mixture of genotypes and disease.