Figure

Figure. Molecular methods used to identify human adenovirus (HAdV) co-infections. A–D) Vaccinated sample 7274. A) Microarray hybridization profile. White and yellow rectangles indicate low-positive HAdV-C and HAdV-B2 targets, respectively. Spot colors denote...

By using a new 70-mer spotted microarray (18), a PCR–enzyme-linked immunosorbent assay (19), and a species-specific multiplex PCR assay (20), we generated data profiles for each of the 52 tissue culture-amplified samples; the raw data from 2 of these samples are shown as representative examples (Figure). The microarray profile of vaccinated sample 7274 detected HAdV-4 (species E), HAdV-21 (species B1), HAdV-C, and HAdV-B2 according to previously validated hybridization patterns (18) (Figure, panel A). Except for detection of an apparent low-level HAdV-C co-infectant, the results of the Adenovirus Consensus kit (HAdV-B1, HAdV-B2, and HAdV-E) (Figure, panel B), multiplex and monoplex species-specific PCR (HAdV-B and HAdV-E) (Figure, panel C), and serotype-specific PCR (HAdV-4, HAdV-21, and HAdV-B2) (Figure, panel D) confirmed the microarray-based identification of multiple adenoviral strains in sample 7274.

In contrast, the microarray profile of unvaccinated sample 10756 detected a single serotype, HAdV-4 (Figure, panel E). The microarray-based finding was verified by results of the Adenovirus Consensus kit (Figure, panel F), multiplex species-specific PCR (Figure, panel G), and HAdV-4 serotype-specific PCR (Figure, panel H). The data profiles for all 52 samples assembled and compared in this manner are shown in Table 1 and Table 2. Dual, triple, and quadruple infections were found in all 21 of the vaccinated samples and in 14 of the 31 unvaccinated samples tested (Table 1 and Table 2).

Previously vaccinated persons showed a high rate of co-infection with both species commonly associated with ARD (HAdV-B1 and HAdV-E), whereas unvaccinated persons were primarily infected with HAdV-E. Since HAdV-4 and HAdV-7 are the 2 most common ARD-associated serotypes, that they were also the most commonly paired respiratory pathogenic co-infectants detected in vaccinated persons is not surprising. When the vaccine was used, the rates of other respiratory adenoviruses were much higher than when the vaccine was not used (16). However, these isolates were chosen for study because they yielded antigenic signals consistent with either HAdV-4 or HAdV-7 and were therefore expected to contain at least 1 of these 2 viruses as the highest titer adenoviral components (Table 1) (16).

The ability of the microarray to identify to the serotype level resulted in the detection of the greatest number of co-infections, despite its inability to detect members of species B2 when a co-infecting HAdV-7 was present (hybridization pattern interference) and members of species F that were not targeted (Table 3). Microarray-based identification of multiple ARD-associated serotypes from diverse HAdV-B1 species (serotypes 3, 7, and 21) was necessary because co-infections with these serotypes would not have been indicated or resolved by methods limited to species-level identification.

Although most apparent co-infections could be verified by each of the primary methods tested and by serotype-specific PCR (e.g., single infections: 10756, 60406, 20142; co-infections: 1212, 7151, 7274), some could not be verified (e.g., 60691, CHPPM2). Those co-infectant signals that could not be verified were usually weak positives. The strains responsible for these signals appeared to be subordinate co-infectants because the predominant serotype or species signals generated for the associated samples by the microarray, Adenovirus Consensus kit, and serotype-specific PCR were corroborated in every case and matched the results obtained from the sequencing experiments previously reported (16).

The microarray and Adenovirus Consensus kit use detection and signal amplification techniques that enhance assay sensitivity and thus render them more sensitive than traditional PCR/agarose gel visualization techniques, as shown by the number of triple and quadruple co-infections detected with these techniques (Table 3). Thus, attempting to corroborate these methods with the 3 PCR-based methods used was not completely successful. Nevertheless, most of the positive results from these tests were verified by comparing the microarray and Adenovirus Consensus kit results or by comparison with the results from independent methods such as microneutralization, hexon sequence analysis, serotype-specific PCR that uses primers not used in the multiplex tests, and PCR amplicon sequencing (Table 1, Table 2, and Table 4). These results suggest that these methods can identify and corroborate HAdV co-infections and that, in general, the HAdV load in ARD patients is more complex than previously thought.

To determine whether >1 replication-competent serotype or strain was present in the samples with evidence of co-infection, we attempted to physically separate the paired co-infectants in sample 7151 by plaque purification. Of 92 plaques picked from the initial plate, all tested positive for HAdV-C by PCR and 12 of 92 also tested positive for HAdV-B. Several of the plaques that retained both HAdV-B and HAdV-C signals were replaqued, and PCR testing of these plaques yielded only HAdV-C isolates. Further efforts that used a detergent to increase separation within the original ITCF sample 7151 and applied the agarose overlay more quickly (6 hours) to prevent interplaque contamination also yielded only HAdV-C plaques (data not shown).

We demonstrate the rarely reported phenomenon of co-infections with multiple adenoviral species. Two previous studies have noted rare instances of HAdV-C dual infections in small numbers (27,28). HAdV-C, although rarely associated with pharyngitis outbreaks in recruits (10), is usually seen in children (7,8) and can produce latent infections that last into young adulthood. This fact, combined with low incidence of co-infection (27,28), has led to the assertion that multistrain adenovirus co-infections are not common (28) or clinically relevant. The results from the population tested in this study suggest otherwise. Samples from vaccinated recruits showed a high rate of co-infection with multiple species of adenovirus associated with adult ARD (HAdV-E and HAdV-B1).

Many of the identified co-infectants in this study were species not generally associated with ARD in the military (HAdV-C, HAdV-B2, and HAdV-F). Although these species were not likely the cause of ARD observed in these patients, since they are not believed to cause ARD in adults and because they have a high potential for latent carriage (1,7,8,29), their presence sheds new light on the general complexity of the human adenoviral load. In addition, they remain viable reservoirs capable of genetic complementation or recombination with upper respiratory strains.

Recombination can generate new strains with unique and stable phenotypes. Intraspecies adenovirus recombination has been demonstrated in laboratory cell-culture co-infection studies (30–32). These recombination events can generate viable hybrids with intermediate or unique immunogenic and tropic properties. Evidence suggests recombination can generate hybrids in immunocompromised patients (29,33,34), possibly as a result of co-infection with normally isolated serotypes. Recombination, particularly intraspecies, seems to play a major role in the evolution of new, virulent strains of HAdV (1,17,35,36). The currently dominant pathogenic HAdV in US military recruits, a considerably diverged variant HAdV-4 strain (16), appears to be a recent recombinant between HAdV-4 and a HAdV-B1 serotype, probably HAdV-7 (37). Given that these 2 are the most common co-infectants seen in our sample set, this finding suggests that the observed dominance of co-infections in vaccinated persons may have contributed to the emergence of the new variant. In general, the understanding and control of situations that create or promote co-infection may be important considerations.

The HAdV vaccine, an enteric-coated live-virus tablet designed to transiently and selectively infect the gastrointestinal tract with normal respiratory HAdV strains, contains viable HAdV-4 and HAdV-7. Thus, we cannot assume whether the detected co-infectants arose from the vaccine itself or from community acquisition of circulating strains. Most HAdV-4 strains in this study are not the vaccine strain but rather a highly divergent variant that has recently been dominant in military training centers (GenBank strain Z-G 95-873). This identity was shown by sequence analysis of 1,500 bp of the hexon gene from many primary infectants identified in the same sample set that was analyzed here (16). The variant HAdV-4 isolates consistently differ from the vaccine strain by 32 base substitutions, including 9 coding changes, in this region (16) (Table 1). Hexon sequence analysis showed that many HAdV-7 co-infectants are HAdV-7d2. HAdV-7d2 is distinguished from the HAdV-7 vaccine strain (HAdV-7a) by a single coding polymorphism in the hexon sequence, but this polymorphism (protein L443Q or nucleotide T1328A in GenBank [16]) is specific to HAdV-7d and HAdV-7d2 and is not found in HAdV-7a, b, c, g, or h or in the vaccine strain (16,38,39). Three of the other HAdV-7 co-infectants (1856, 4185, and 7372) were shown to be HAdV-7h by fiber gene sequencing. The fiber gene of HAdV-7h appears to have been horizontally transferred from HAdV-3 and thus is highly diverged from the usual HAdV-7 fiber gene, as found in the vaccine strain (17). Thus, sequence analyses show that most, if not all, co-infectants are currently circulating HAdV-4 and HAdV-7 strains that are distinct from the vaccine strains (16) (Table 1).

Four lines of evidence support the idea that most of the apparent genetic complexity in the throat swab samples comes from multiple strains, as opposed to recombinants with mixed genetic characteristics. The first comes from the microarray data. The microarray tests for hybridization of 6 independent probes designed to match serotype-specific sequences in 3 genes (18). Since different species do not cross-react among the microarray probes, hybridization of genes from 1 species to the identifying probes for 2 species would require redundant presence of 2 different alleles in all 3 genes. Since both natural recombination in hosts (17) and artificially encouraged recombination in cell culture (30,32) strongly favor homologous recombination and generation of nonredundant hybrid strains, redundant characterization of paired, divergent alleles is inconsistent with a single recombinant genome.

The second line of evidence supporting co-infection with independent genomes comes from comparisons of relative co-infectant titers before and after potentially selective events, such as growth in tissue culture. PCR amplification of fiber gene sequences using species B- and E-specific primers was performed on serial dilutions of vaccinated sample 7274 before and after passage of the original ITCF through 2 additional cycles of growth in A549 cells. In this instance, the relative titers of HAdV-4 and HAdV-7, as measured by serial-dilution PCR, changed by 2 orders of magnitude (data not shown). The rapid drift in relative concentrations of PCR targets from paired co-infecting strains strongly suggests that the co-infectants' genomes are replicating independently and thus likely to be physically separate entities.

The third line of evidence supporting co-infectant independence comes from whole-genome sequencing efforts. Several molecular methods indicated that vaccinated sample 7151 harbored an HAdV-5/HAdV-21 co-infection (Table 1 and Table 2). The genome of the HAdV-5 co-infecting strain was sequenced and assembled into a contiguous sequence (GenBank no. AY601635) consistent with a published HAdV-5 genome (GenBank accession no. AY339865) (40), which suggested no recombination of foreign DNA. However, this effort also generated several orphan sequences that did not fit into the assembled sequence and were subsequently identified as genetically redundant HAdV-21 regions. Further amplification and sequencing of several genetically distant fragments from the same sample using HAdV-21-specific primers yielded ≈2 kb of HAdV-21 sequence. On the basis of the entire genome and partial PCR sequencing analyses, >2 co-infecting HAdV genomes are contained in sample 7151.

The fourth line of evidence comes from our attempts to physically separate paired co-infectants by plaque purification. Sample 7151, which contained the HAdV-5/HAdV-21 co-infection, was used initially because it contained relatively equal titers of both co-infectants. Although most of the plaques tested contained HAdV-5, some contained both HAdV-5 and HAdV-21. Although we were unable to identify plaques that contained only HAdV-21, our results demonstrate the physical independence of the co-infecting entities and the functional independence of HAdV-5. Our results also suggest that either the HAdV-21 co-infectant is functionally dependent on HAdV-5 or is effectively outgrown by HAdV-5 to a degree that prevents independent isolation. Similar attempts were made with a few samples that had HAdV-4/HAdV-7 co-infections, but these were generally biased in titer (10⁴ in favor of HAdV-7) and, as expected, yielded only HAdV-7 in >300 plaques tested. The data demonstrated the functional independence of 1 co-infectant (HAdV-7) and physical independence of the co-infecting entities but could not conclusively demonstrate functional independence of the minor co-infectant.

Conventional clinical microbiologic methods, including microneutralization and hemagglutination inhibition, are comparative and designed to identify the primary HAdV serotype (or species) in a sample. Secondary infections are masked in these methods by the tests (e.g., microneutralization is reported as the strongest reaction, not the spectrum of reactions across all serotypes). Likewise, direct sequencing (16) may restrict identification to a single strain, particularly if 1 co-infectant is dominant. Restriction enzyme analysis methods are capable of resolving HAdV-C dual infections in which both serotypes are present in similar numbers (27). In contrast, when using sensitive molecular methods that can yield measurable signals from secondary (less numerous) co-infectants against the background of stronger signals produced by primary infecting strains, these methods may identify co-infections more than do conventional methods. In the case of respiratory infections, this finding has previously been documented (41).

Finally, each of the methods designed to test for multiple species or serotypes showed a higher number of HAdV (and accepted virulent HAdV species and serotypes) in vaccinated persons than in unvaccinated persons. HAdV vaccine was administered routinely to all trainees until supplies were exhausted, at which point adenovirus vaccination was stopped. Since trainees were vaccinated systematically, persons tended to be sampled at times when either all or no recruits were being vaccinated. Therefore, vaccinated samples collected and tested (from 1996 to 1998) are not concurrent with unvaccinated samples (collected from 1998 to 2000). Because of this sampling limitation, we could not confidently correlate HAdV co-infection with breakthrough infections in previously vaccinated persons. Thus, although this study highlights the previously underappreciated phenomenon of adenoviral co-infection, the conclusive examination of its relationship to vaccination must await reintroduction of HAdV vaccine (15).