The frequency of cytomegalovirus non-ELR UL146 genotypes in neonates with congenital CMV disease is comparable to strains in the background population

Cytomegalovirus (CMV) is a highly prevalent herpesvirus [1] that is commonly transmitted during early childhood [2‐4]. Congenital CMV infection (cCMV), the result of mother-to-fetus transmission, is the most common congenital infection in developed countries with an estimated incidence of ~ 0.5% of live births [5, 6]. Ten percent of infected neonates are symptomatic at birth while 15% show sequelae at later stages of their childhood [5, 6]. The symptoms include neurological developmental disorders such as non-hereditary sensorineural hearing loss (SNHL), vision loss, cerebral palsy, intellectual disability, and severe cCMV can lead to pregnancy loss, thus representing a substantial health burden [7].

Cytomegalovirus carries a genome of approximately 235 kb double-stranded DNA that encodes numerous genes used in dissemination of infection, immune evasion, and establishment of latent infection. A very high degree of genetic diversity caused by disrupted and highly variable genes has been observed between wild type strains [8, 9]—an unusual observation for DNA viruses—but its effect on the pathogenicity and virulence of individual CMV strains remains to be determined. One of the most diverse genes in the CMV genome is UL146 [8, 9], encoding a viral chemotactic cytokine (chemokine), vCXCL1, which recruits neutrophils to sites of infection through activation of the chemokine receptors CXCR1 and CXCR2 [10, 11]. The first indication that UL146 is a possible virulence gene, and thereby required for CMV infection of the human host, came from the observation that the Toledo wild type strain contained a 15 kb DNA segment containing the UL146 ORF among 22 ORFs not present in laboratory HCMV strains [12]. This indication was recently supported by the finding that recombinant MCMV expressing the UL146 gene enhanced MCMV dissemination kinetics in a mouse [13]. Previous studies have identified 14 distinct UL146 genotypes (GTs) [14] which have shown differences in receptor binding affinity, signaling efficacy, and chemotactic properties [15]. When identifying the UL146 gene product from the Toledo strain (genotype 1) as a CXCR1 and CXCR2 agonist [10], we previously noticed that the sequences of two other genotypes—genotypes 5 and 6—stood out in ways suggesting an altered molecular function likely to have biological implications. Firstly, they lack the ELR motif (amino acids Glu-Leu-Arg) in their N-terminus, a determining factor for chemokine targeting to CXCR1 and CXCR2 (Fig. 1). The ELR motif is conserved in all human chemokines targeting CXCR1 and CXCR2 [16] and is lacking in all chemokines targeting other chemokine receptors. It has been demonstrated, that introduction of an ELR motif in a non-ELR chemokine enables binding to CXCR1 and CXCR2 [17], while a change to any of the three amino acids in an ELR chemokine abolishes its signaling capabilities through CXCR1 and CXCR2 [18]. Secondly, genotypes 5 and 6 have a nine amino acids long N-terminus compared to the only four amino acid of the other UL146 genotypes (C. Berg et al. unpublished observations) (Fig. 1). The N-terminus of a chemokine is crucial for receptor binding and activation as confirmed by the solved structures of chemokine-bound chemokine receptors [19]. Thus, the chemokines encoded by UL146 genotypes 5 and 6 are distinct in ways likely to affect the receptor and host cell target, suggesting an altered biological function of these genotypes.

Previous studies have tried to establish a link between UL146 genotypes and cCMV without finding a clear association [9, 14, 20, 21]. In addition, three studies have investigated whether severe cCMV was linked to specific UL146 genotypes by comparing symptomatic to asymptomatic cCMV cases, but no association was found [22‐24]. These studies genotyped UL146 from mostly urine, but also saliva, amniotic fluid, and colon tissue, and were limited by small sample sizes. Another limitation was the lack of control groups representing CMV strains circulating in the background population of healthy individuals. In an attempt to address these problems, we first reasoned the sample size could be considerably increased by using neonatal dried blood spots (DBS) for UL146 genotyping. In many countries, a neonatal DBS is taken at birth for all newborns to screen for select hereditary diseases. These DBS are often stored in biobanks and their use in cCMV diagnosing has been suggested [25‐28]. Therefore, we developed a highly efficient extraction and amplification protocol to permit UL146 genotyping from the limited amount of blood available in DBS [29]. Furthermore, we established a control group by screening urine samples from children with a suspected bacterial urine tract infection for CMV and genotyping UL146. Thus, by using the Danish civil registration number individually coupled to diagnoses in the Danish National Patient Registry and neonatal dried blood spots taken from all Danish newborns since 1982 stored in the Danish Neonatal Screening Biobank, we here investigate whether the non-ELR UL146 genotypes affect the development of cCMV disease in 116 infected neonates by comparing to a background population of strains derived from 83 children without CMV disease.

Circulating CMV strains exhibit a high degree of genetic diversity unparalleled by other human herpesviruses [8, 9] that affects important immunomodulators. These observations have raised the question of whether specific CMV strains have different pathogenic potential, however, current information has proven insufficient to demonstrate a clear link. UL146 is one of the most diverse CMV genes, encoding a CXC chemokine with 14 distinct genotypes that potentially have different biological functions. Using a control group of circulating CMV strains from the background population, we found no association between the two UL146 genotypes lacking the ELR-motif (genotypes 5 and 6) otherwise conserved in the 12 other genotypes and the development of cCMV disease in neonates.

We genotyped UL146 in a total of 116 CMV positive DBS samples from neonates with cCMV or SNHL-CMV and in 83 samples from children without a history of CMV disease. We observed three genotype 5 strains and no genotype 6 strains in the case group (one in each sub-group of cCMV, SNHL-CMV, and both diagnoses), while two genotype 6 and no genotype 5 strains were observed in the control group, a difference that was not found to be significant. The 2.6% frequency (3/116) of genotype 5 and 6 strains found in neonates in our study did not differ from frequencies found in other studies, which were in the range from 0 to 4.5% [9, 14, 21‐24]. This study is yet the largest among UL146 genotyping studies from neonates with cCMV [9, 14, 20‐24] and, to our knowledge, the only one with an established control group for comparison of genotype prevalence between patients and the background population. The lack of control groups in previous studies have made it difficult to determine whether the reported prevalence of non-ELR UL146 genotypes in neonates with cCMV is different from the strains circulating in the background population of the particular geographical area. However, there are potential caveats to the comparison of cases and controls. A study sampled longitudinal genomic populations from the urine and plasma of five infants with symptomatic congenital CMV infection. They found that samples from different compartments taken at the same time were more variable than samples from the same compartment taken over time. The data were consistent with models with several bottlenecks and phases of rapid expansion of the viral populations driving viral selection. Furthermore, they found that positive selection played a small role in viral evolution within a compartment in contrast to a strong role and pervasive driver of evolution associated with compartmentalization [32]. Thus, a possible difference in genotypes between cases and controls could also be interpreted as comparing sequences from viral populations drawn from different compartments, i.e. blood and urine, or from comparing sequences from compartments with a high and low viral load. However, no mutations in the UL146 gene between compartments or over time were reported. Likewise, we did not find any inter- or intra-genotypic differences in frequencies of UL146 between the cases and controls (Tables 1 and 2). Though we cannot exclude that genotypes found in children’s urine are not fully representative for strains circulating in the background population, we find it to be a very close approximation as the strains found in childhood remain throughout adult life as CMV establishes latency. Finally, we cannot rule out that non-ELR UL146 genotypes are a pathogenic factor for development of symptomatic or severe cCMV disease as no medical chart information was obtained, making it impossible to discern the severity of CMV disease between the cases. This study limitation was accepted early on as recent changes in national patient data security regulations required medical chart review to be carried out in local collaborations with doctors from each hospital having treated the patients dating back to 1982, which was not considered feasible. However, as symptomatic cCMV cases are expected to represent a significant part of neonates with a cCMV diagnosis and the prevalence of non-ELR genotypes was low (2.6%), it is not expected that chart review would have changed the conclusions of this report. Post hoc analyses for the other genotypes were performed and the overall genotype distribution was found to be similar between the patient groups and the control population with no significant differences. Furthermore, there was no observable difference in viral load between different UL146 genotypes, which to our knowledge has not been investigated previously. A second limitation is that Sanger sequencing is not suited for detection of multiple UL146 genotypes in mixed CMV infections. However, next generation sequencing has shown that mixed UL146 genotypes are found in just 2% of congenital CMV infections and that multiple CMV strains are infrequently found in congenital infection (14%) [9]. Moreover, NGS performed on enriched target material requires a substantial viral load for full genome coverage. Thus, target enriched HIV samples showed full coverage of the 10 kb HIV genome in 80% of samples with viral load above 3100 cp/ml equivalent to a target input of 1500 viral copies [33]. For comparison, the median viral load in blood in congenital CMV infection has been reported to be 2300 cp/ml [34], equivalent to a DBS target input of approximately 14 copies (from two ~ 3 μl punches), which would suggest a lower sensitivity for UL146 genotyping by NGS than seen for Sanger sequenced PCR products. Thirdly, the use of only one PCR reaction per sample can have allowed for rare polymerase introduced mutations, which could give rise to false positive intra-genotypic mutations. Although the error rate of the Pfu polymerase is low, estimated to approx. 1:1000,000 bp, and any minor random error was not likely to affect the manual genotyping of UL146. However for this reason, only deletions and non-conservative mutations appearing more than once are shown in Table 2.

Interestingly, only 451 cases of congenital CMV disease (336 cCMV + 77 SNHL-CMV cases + 38 with both diagnoses) had been registered in Denmark since 1982 with just 267 relevant samples available—much lower than the number expected from reported incidence rates. Assuming a live birth rate of approximately 60.000 per year in Denmark (based on the numbers published by Statistics Denmark) and a cCMV incidence of 0.5% with 10% symptomatic cases, 30 symptomatic cases per year was expected, or 1080 symptomatic cases over the 36 year period, plus a larger number of asymptomatic cases. There was no change in number of diagnosed cCMV cases over this period (Fig. 4d), averaging to 10 cases per year. This suggests that cCMV disease is underreported or underdiagnosed in Denmark.

We were able to detect CMV in 45% of the 267 DBS samples upon screening with the R-gene qPCR CMV kit and found a significantly higher positive rate among samples from neonates with a SNHL-CMV diagnosis compared to those with a cCMV diagnosis. This could be a reflection of the sensitivity of the methods used to extract and detect CMV in the dried blood spots. Sensitivity is a valid concern when using DBS for pathogen detection as only ~ 6 μl dried blood was obtainable from each patient. Despite the extraction protocol being optimized for up to 100% recovery efficiency [29] and the R-gene kit having a high sensitivity (LoD_95% of 555 copies/ml and LoD_5% of 30 copy/ml), the 6 μl dried blood was diluted during DNA extraction which allowed for cases with very low viral loads to avoid detection. In previous studies, the sensitivity for detecting CMV in DBS from neonates with confirmed cCMV by urine/saliva analysis has shown variation depending on the study design. One study found 100% sensitivity for 72 cCMV cases using a relatively large sample volume [35], while another three studies reported sensitivities from 70 to 75% among 55, 70, and 103 cCMV cases [26, 36, 37] which agrees with a report that 75% of cCMV cases are viraemic at birth [38], and a fifth group found a lower sensitivity of 34% for 91 cCMV cases [27] that the authors explain as being related to DNA extraction and PCR. However, all undetectable cases are likely not accounted for by the sensitivity of the assay as there was a significant difference between the subgroups of neonates with a cCMV diagnosis and those with a SNHL-CMV diagnosis in our study. Thus, several factors are likely to influence the positive rate in the case groups. Firstly, the sensitivity of extraction and PCR, as low and non-viraemic cases could be missed. Secondly, the diagnostic precision, as cases could not be validated without chart review. And thirdly, the timing of infection, as the cCMV diagnosis may have been given due to active infection of the mother during pregnancy and not of the child at birth.

We sequenced UL146 in 116 DBS samples from neonates with cCMV disease, compared the results to 83 samples from the background population, and did not find an association between specific UL146 genotypes and cCMV disease. As studies have shown that circulating CMV strains contain disruptions of more than 25 genes and have at least 30 polymorphic genes [8, 9], each of these individual genes as well as a mix of viral and host factors should be investigated as a whole in future studies looking for an association between genetic content and clinical endpoints.