The matrix gene real-time PCR is widely used to estimate the prevalence of AI in wild bird populations, and the assay is sometimes used to screen samples for subsequent virus isolation [5
]. Matrix Ct values are inversely related to the amount of viral RNA present in a sample-samples with low CT values have more viral RNA while samples with less RNA must undergo additional cycles to reach the threshold response. Although the maximum limit of detection for the matrix PCR assay is 45, many studies have shown that virus isolation success is much higher for samples having lower (<35) Ct counts [7
]. In the present study, viruses were isolated from 39% of birds indicating that these birds were shedding infectious virus ( and ). Virus was isolated from a single swab for 11 of 31 birds, most often corresponding to VTM samples that had high matrix Ct values (=low amount of RNA). Eight VTM samples with Ct values >35 grew in eggs, including two with values above detectable limits (>45), while three samples with Ct values <35 did not grow. Thus, as shown in previous studies [5
] the VTM matrix Ct value is a good, but imperfect predictor for virus growth in ECE, and several isolates would have been missed if a low matrix Ct value were used as a prerequisite for inoculation into eggs.
We amplified and sequenced a portion of the HA gene to provide an independent assessment (other than matrix assay) of AI infection in samples prior to egg inoculation. As with the matrix assay, the 640PCR test only detects viral RNA and does not indicate whether infectious virus is present. Thirty-nine percent of birds (31/79) were HA positive on one or both of their VTM samples (), and six HA subtypes were detected (H3, H4, H5, H6, H11, H12). Mixed infections (two HA subtypes) were detected in 11 swab samples from a total of seven birds. All of the swab samples with matrix Ct values <35 were able to be HA subtyped, likely because these samples had the most viral RNA template for both the matrix and HA genes. However, 17 additional HA-subtyped samples were obtained by sequencing samples with matrix Ct values >35, demonstrating that samples with high matrix Ct values can provide important genotype information prior to virus culture and amplification ( and ).
The detection of HA sequences in VTM samples was a good predictor of virus growth in ECE, and >90% of HA positive samples (46/51) yielded virus isolates. Conversely, growth in eggs was a good indicator that HA sequences could be obtained from the original VTM sample. Of the 51 VTM samples that yielded virus isolates, only five could not be HA subtyped prior to inoculation in eggs. Although the overall frequency of HA positive results did not differ between paired swab samples, 35% (11/31) of the HA positive birds had only a single positive swab pre-inoculation. As with the matrix gene PCR test, combining the results of the two samples from each bird increased the estimated prevalence of AI within the population, and it provided a more accurate view of which birds were shedding infectious virus. These results are consistent with other studies that showed a combination of methods improved the accuracy of detection [6
It is widely recognized that virus culture in ECE is a necessary, but imperfect method for isolating influenza viruses infecting wild birds [5
], and comparison of HA subtypes obtained before and after egg inoculation showed that egg culture altered the occurrence and detection of viral strains. Although the same six HA subtypes were detected in birds before and after egg inoculation, the detection of H3 and H4 subtypes together increased in frequency from 58% to 77% post-inoculation, while H5 detections decreased by 50%. Mixed infections were also detected less frequently, as has been suggested by other direct studies of VTM samples [14
], occurring in seven birds pre-inoculation but in only three birds post-inoculation. These changes likely reflect low amounts or a lack of viable virus in the original sample, variable fitness of different strains to grow in ECE, and/or virus competition in mixed infections [7
]. Most importantly, these results show that culture in eggs provides a biased view of viruses present in original or primary samples. The ideal surveillance assay would yield full genome sequences from original samples without culture. Based on the results of the present study, we now perform both 640PCR and M-RTPCR assays to obtain partial and even full genome sequences from primary VTM samples. However, the low amount of RNA present in many samples poses a major obstacle, and more sensitive methods are needed.
The multi-segment M-RTPCR and the 640PCR gave essentially identical results for subtyping virus isolates in ALF. Although M-RTPCR amplification of a subset of HA and NA segments was inefficient in some cases, subsequent NGS resulted in complete genome sequences. This indicated that all eight RNA segments were present in the ALF samples as expected, but that the HA/NA segments were not efficiently amplified during the M-RTPCR reaction. Interestingly, the poor amplification was restricted to H4 and N9 in our sample set, perhaps because RNA secondary structure inhibited efficient transcription, or the PCR conditions were not optimal for these subtypes. Decreased efficiency of H4 amplification compared to other HA subtypes has also been reported in a study using a different PCR protocol that also used primers based on the conserved 5' and 3' ends of the viral segments [18
Virus selection and possibly mutation in ECE was observed by comparing the sequences obtained from duplicate AI isolates (two swab samples from the same bird). Nucleotide differences were observed in 29% of the paired segments (n = 20 birds; 160 pairwise comparisons), with identities as low as 91.63%. Variability between paired samples with high identity (>99.9%) may have been caused by the high error rate (≤0.1%) of virus RNA polymerase [1
], as well as PCR and DNA sequencing polymerase errors [19
]. Paired samples that showed <99.9% identity among one or more segments (14% of total, n = 23 segment pairs) were from birds (8/20) with mixed infections, i.e.
, two HA subtypes were detected in the ALF sample, or, more commonly, there were differing HA subtypes detected in the VTM sample versus
the corresponding ALF. However, not all segments in these mixed samples showed variability, likely because the different virus strains shared common segment sequences as observed previously [3
]. Another possibility is that during growth in the ECE, certain virus segments from the mix were retained and others “discarded” during reassortment and virus packaging. Non-random reassortment of segments in mixed infections has been observed in other studies, and it has been postulated that strain-specific signals in certain genomic segments result in selective packaging in progeny virus particles [17
This study showed that mixed infections and naturally occurring viruses can be overlooked or altered by laboratory methods, and that analysis of primary samples prior to culture provides a more accurate view of viruses occurring in vivo
. However, despite the use of increasingly sophisticated tools, our attempts to identify and characterize viral diversity will continue to be challenged by the dynamic, quasispecies nature of influenza viruses [22
]. The diversity of subtypes and mixed infections we detected prior to culture, and the nucleotide differences we observed in sequences from paired swabs from the same bird, hint at the underlying complexity of infection likely occurring at the cellular level. However, consensus sequences oversimplify the genetic diversity and complexity of influenza virus infection in vivo
, and they fail to fully convey a given isolate’s potential for cross-species transmission and pathogenicity [24
]. Deep sequencing efforts that focus on characterizing the full diversity and interactions among virus subpopulations at the molecular and cellular levels will be essential going forward, particularly if they are applied to primary samples not altered by sample handling and laboratory methods [25