Improper H5N1 Surveillance
Recombinomics Commentary 13:50
March 31, 2008
Guan, a professor at the University of Hong Kong, knows just how backbreaking and mundane surveillance work can be. He and his researchers have tested more than 200,000 stool samples of chickens, aquatic and wild birds collected from various parts of China since 2000, screening them for the H5N1 virus
"For disease control, surveillance must be a long term effort. You know where it is and you know it is coming, like a spark of fire you can extinguish it," Guan said. "If not for all this surveillance and detection ability, the pandemic would probably have already come." Drawing from what is known of past pandemics, Guan believes that surveillance and strict control measures are the answers. "Pandemics don't happen suddenly, they have an early phase, mature phase, outbreak phase. The virus changes step by step, it takes a long cooking time," he said. "If a virus gets into humans in the early phase, the transmission ability is very low. At most, they infect their families, but it can't go further into the community. "This phase is the golden point to control. Once it matures and becomes (efficient in) human-to-human (transmission), it will be too late."
The above comments on surveillance fail to recognize a host of shortfalls in the current surveillance and associated sequencing, which significantly limits current analysis. One of the most glaring aspects begins with the collection of samples. Recent data on H5N1 in wild birds west of China strongly suggest that collection of stool samples or cloacal swabs will frequently miss clade 2.2 (Qinghai strain), the only H5N1 sub-clade identified in wild bird, domestic poultry, or human infections west of China. The presence of this virus was not detected in the intestines of twenty three ducks experimentally infected with clade 2.2. The virus was rarely detected in cloacal swabs, although it was universally detected in pharyngeal swabs. However, even in pharyngeal swabs the virus was only detectable for one day, even under carefully controlled lab conditions. Thus, even if all birds were infected once a year, on average only 1 in every 365 throat pharyngeal samples would be positive.
This low level of clade 2.2 in wild birds is the likely explanation for the lack of detection in stool samples from southern China. Only one of 404 HA samples from southern China was positive for clade 2.2. The others were clade 2.3 and were primarily the Fujian strain. These samples were collected between the middle of 2005 and middle of 2006, when clade 2.2 was spreading throughout Europe, the Middle East, and Africa.
The concentration of the Fujian strain in southern China and southeastern Asia led to the prediction that the Fujian strain would represent a third wave of H5N1, but the prediction has not been realized. Clade 2.2, was widely reported in 2007 and 2008 in countries to the west of China. In contrast, clade 2.3 has been limited to southern China and southeast Asia. Thus, the predicted movement of H5N1 was influenced by a strong bias introduced by the collection method, which was limited to stool samples and cloacal swabs leading to the under representation of clade 2.2.
These limitations have been extended by various conservation groups which have assayed 350,000 stool samples and have rarely found H5N1, including areas where detection of H5N1 in dead or dying birds was common. Thus, the assay alone can seriously impact surveillance over a wide area, including Europe, the Middle East, Africa, and North America.
These negative data also lead to false predictions at the Options VI meeting in Toronato in May, 2007. Researchers in Europe relied to the negative data on the stool samples to declare wild birds in Europe free from H5N1. Within minutes of the talk, the Czech Republic announced H5N1 infections in poultry. That was quickly followed by reports of H5N1 in wild birds in the Czech Republic, Germany, and France. In all cases the sequences linked back to the wild bird outbreak at Uvs Lake in Mongolia in the summer of 2006. The outbreak in Mongolia was massive, raising concerns that the clade 2.2.3 that had been limited to areas in eastern Europe and south Asia would then spread as region.
The Uvs Lake was then reported in the outbreak in South Korea at the end of 2006, followed by an outbreak in Kuwait in early 2007. The appearance in multiple countries in Europe in the summer of 2007 was followed by outbreaks throughout Europe and the Middle East. The European outbreaks extended to resident mute swans in England at the end of 2007 and beginning of 2008. This outbreak lasted for two months and although the number of dead birds that tested positive was limited, all were the Uvs Lake strain and had considerable genetic diversity, indicating infections had not been recent. Attempts to detect the H5N1 in live birds failed, as expected from the wild bird experimental studies which reported detection was limied to a single day post infection, suggesting testing of 60 swans would have produced negatives, even if all 60 were infected during the testing period. The use of limited testing to declare regions H5N1 free is widespread and seriously limits surveillance.
Further limits are introduced by biases in the sequences generated. Most of the sequences from the H5N1 collected by the University of Hong Kong have been partial sequences, especially in the longer internal genes (PB2, PB1, PA, NP). The partial sequences allow for reassortment analysis, but severely limit recombination analysis to the tracing of individual polymorphisms. Recent data has identified statically significant homologous recombination over short stretches in multiple genes in human influenza. However, the analysis of the human sequences, as well as swine and avian sequences, is significantly inhibited by the flooding of the database with partial sequences.
However, in spite of these limitations, individual polymorphisms can be tracked through the database to identify future and changes and combinations, which are important on vaccine target design. Remarkably, acquisition of polymorphisms onto one genetic background leads to the same change on different genetic backgrounds. This has been demonstrated conclusively for the NA polymorphism, G743A. Although this change is silent, it has moved from a well defined geographical and genetic distribution in 2006, which was almost exclusively limited to similar isolates in southern Germany, Switzerland and France, to a widespread distribution in 2007, including the concurrent acquisition onto multiple genetic backgrounds in Egypt, Russia, Kuwait, Ghana, and Nigeria. The movement onto clade 2.2.3 in Kuwait in early 2007 has led to widespread distribution throughout Europe and the Middle East. All published 2007 NA sequences which trace to Uva Lake have had G743A.
However, this type of broader distribution has not been limited to silent changes. In 2005, the H5N1 isolates in Mongolia had a novel cleavage site. In early 2007, the same cleavage site was found in Egypt and Nigeria on clade 2.2 genetic backgrounds that were distinct from each other, as well as the original Mongolian sequence. Similarly, the receptor binding domain change V223I which was also present in 2005 in isolates in Mongolia appeared in Egypt in late 2006, followed by appearance ion Togo later in 2007. These types of tracing of individual polymorphisms can be used to predict new vaccine targets, which can be used to help control the spread of H5N1.
Current approaches use older targets, which have led to vaccine failures and more rapid evolution in targeted areas, such as Egypt, where the H5N1 that has emerged from vaccinated stocks has a large number of non-synonymous changes. Many of these changes were accumulated from earlier Egyptian isolates or new clade 2.2.3 acquisitions.
Thus, while surveillance can help control H5N1, the current approaches which use fatally flawed surveillance approaches, followed by limited sequence data or hoarding of sequence data, greatly limits the utility of such surveillance and repeated inaccurate predictions based on negative data.
Media Links
Recombinomics Presentations
Recombinomics Publications
Recombinomics Paper at Nature Precedings