Characterization of the 1918 “Spanish” Influenza Virus Matrix Gene Segment

MATERIALS AND METHODS

RESULTS

DISCUSSION

Several model systems have indicated that small changes in the matrix proteins can have large effects on phenotype. Replacing the matrix segment of human viruses with that of PR34 consistently results in a higher yield of the strain for vaccine development (2). The PR34 M1 gene has been shown to confer high-growth characteristics in single-gene reassortant strains (55). The PR34 M1 protein, however, differs by fewer than 10 amino acids from modern human strains. Other studies indicate that the matrix segment of WSN33 (like PR34, a repeatedly passaged derivative of an early human strain) can also confer a high-yield phenotype (55). However, between them, the M1 proteins of WSN33 and PR34 share only a few amino acids that differ consistently from modern human strains and each has additional, unique differences. Therefore, it is not clear which amino acid changes are responsible for the high-yield phenotype or even if the same amino acids are responsible in both PR34 and WSN33. At those M1 sites where PR34 and WSN33 share amino acids found in few other strains, the Brevig18 sequence instead matches the human consensus. Only at M1 site 218 does Brevig18 share a characteristic with PR34 and WSN33 that distinguishes them from modern human strains. PR34, WSN33, and Brevig18 all retain the threonine that is found in avian strains, whereas all subsequent human strains have replaced the avian threonine with an alanine or valine.

While the PR34 and WSN33 matrix segments contribute to a high-growth phenotype when reassortant human viruses are grown on chicken eggs, other experiments suggest that minor changes in the matrix proteins also contribute to higher growth in mouse lung. Passage of the FM47 virus through the mouse lung resulted in a single amino acid change in M1, T139A, that was shown in single gene reassortment assays to be responsible for higher yield in mouse lung (42, 43). Adaptation of another human virus (HK68) to mouse resulted in high-yield variants with one amino acid change in M1 (T167A) and one in M2 (D44N) (6). Govorkova et al. adapted one avian (A/black duck/NJ/1580/78 [H2N3]) and one human (A/Japan × Bellamy/57 [H2N1]) virus to the mouse lung and again found that high-growth variants had only 1 or 2 amino acid changes in the matrix proteins (13). Changes making viruses more virulent in the mouse lung also appear to increase growth rates in embryonated chicken eggs and MDCK cells, suggesting that the changes may have functional significance beyond mouse adaptation (13, 56). None of these changes were found in Brevig18.

Several closely related H5N1 viruses were isolated from humans in 1997. When the complete genomes of strains with a high-pathogenicity phenotype were compared to those of low pathogenicity, a total of 5 amino acid differences were found to correlate with the high-pathogenicity phenotype. One change was in M1 (V15I), a change that has also been found in PR34 but not in Brevig18 (24).

The amino acid changes found to increase virulence in these studies are distributed over the M1 gene from amino acids 15 to 227 and often fall in areas of no known function. Almost all of the changes resulting in increased yield or virulence are not found in any other strain. Overall, the studies suggest that matrix sequence variation plays an important role in virulence but that virulence cannot be predicted directly from the matrix sequence. The 1918 matrix sequence does not code for any of the single amino acid changes that correlate with virulence in these studies.

While the PR34 matrix segment appears to confer a high-growth phenotype on a wide range of human viruses, other studies suggest that the phenotypic effect of the matrix segment depends on viral context. Murphy et al. showed that replacing the matrix segment of a human H3N2 virus (A/Udorn/307/72 [H3N2]) with that of A/mallard/NY/6750/78 (H2N2) resulted in severe growth restriction in the squirrel monkey respiratory tree, while replacement with the matrix segment of A/pintail/Alberta/119/79 (H4N6) caused no growth restriction (33). The two avian segments differed at only 3 amino acids, 2 in M1 and 1 in M2. Similarly, Subbarao et al. showed that the matrix segment of A/Ann Arbor/6/60 (H2N2) attenuated A/Korea/82 (H3N2) but not A/Udorn/72 (H3N2) (44).

The Brevig18 M1 and M2 sequences do not code for any of the unique amino acid changes that correlate with the high-yield phenotype in different model systems. The amino acids that are unique to the PR34 matrix segment, which may contribute to its ability to confer a high-yield phenotype on many human viruses, are not shared by Brevig18. There is not a single amino acid position in M1 or M2 that is unique to Brevig18, suggesting that if the matrix genes were involved specifically in viral lethality in 1918, a combination of substitutions may have played a role. There is, however, substantial evidence that small changes in matrix sequence, in a particular viral context, can result in a high-yield phenotype. It is possible that the unique Brevig18 matrix sequence would result in a high-yield phenotype in any human strain, but the possibility that it would contribute to lethality only in the context of the other 1918 viral proteins cannot be ruled out. It may well be that a particular constellation of amino acids in several proteins contributed to the virulence of the 1918 strain.

Influenza viruses must have at least a novel HA to cause a pandemic (26). However, they must also be readily transmissible between humans and capable of replicating in the human lung. In the pandemics of 1957 and 1968, all of the gene segments encoding internal proteins (with the exception of PB1) were retained from the strain previously circulating in humans (25). Determining whether the matrix segment was similarly retained in 1918 is complicated because the matrix proteins are very highly conserved. Only about 5% of M1 amino acids differ among the most divergent strains. While almost 20% of M2 amino acids vary, they tend to vary more between species than within them and do not appear to change progressively over time like the surface proteins. Phylogenetic analyses suggest that the matrix proteins are not under continuous positive selective pressure in humans; however, there are certain amino acid residues that distinguish the human clades from those of swine and birds. In M1, 3 residues (115, 121, and 137) distinguish the human clade from the 1930s to the present. Brevig18 matches the human consensus at one of the three, retaining the amino acid characteristic of avian strains at the other two. By contrast, in M2, 7 amino acids are found in all human adapted strains; Brevig18 matches the human consensus at 5 of them (Fig. 3). Of interest is the fact that 4 of these 5 amino acids are coded for by the short stretch of nucleotides that are used in overlapping reading frames in both M1 and M2. Since the corresponding amino acids in M1 are extremely conserved, change in M2 in this area is highly constrained. If Brevig18 was acquired directly from an avian source, these four concerted changes had to have been made very quickly. While this observation might suggest that the 1918 matrix segment was retained from the previously circulating human strain, overall phylogenetic analyses suggest an intermediate position for Brevig18 between avian and subsequent mammalian strains, similar to the phylogenetic position of previously analyzed 1918 gene segments. It will probably not be possible to determine whether or not the 1918 matrix segment shifted immediately before the pandemic without samples of the influenza virus strain circulating in humans before 1918.