Mystery Rays from Outer Space

In blog posts and comments I’ve seen a few people arguing that the Mexico H1N1 influenza virus must have been around for a while — 10 years? 11 years? — because it doesn’t have any really close neighbours based on sequence.  I’m not sure where this is coming from, but the precision of the “11 years” seems strange.  Has anyone knowledgeable actually said this, or is it based on guesswork?

In any case, the notion that this H1N1 has been circulating for a decade has struck me as a great underestimate of influenza variation.  This virus is more or less 96% identical — depending on which segment you choose — to other ciruclating North American or Eurasian swine flu viruses.  That seemed like a really close agreement to me and seemed like the sort of thing that could happen really fast.

But I’m not a flu guy.  So I looked for some way to compare.  What I wanted was a series of influenza sequences with known dates and phylogeny — that is, preferably from a single outbreak or series of outbreaks with a more or less common cause.  I found that in this paper¹ from Jian et al. looking at influenza outbreaks in Taiwan in the 2003-2006. I focused on the Influenza A, H1N1 type — and this represented a single flu season.

Looking at the 169 amino acid region they sequenced:

The average numbers of differences at antigenic and nonantigenic sites of H1 in 2005 and 2006 were 0.90 and 2.34, respectively.  … Moreover, we have observed that the overall average amino acid difference has steadily increased since February 2006. The average differences at antigenic and nonantigenic sites in 2005 were 0.39 and 0.67, respectively, while in the first 7 months of 2006 they were 0.98 and 2.93, respectively. ¹

I’m having a little trouble exactly mapping the sequences they deposted in GenBank with the order they were isolated in and so on.  But the bottom line is that it looks as if in a single flu season, this human strain showed up to 5% divergence at the nucleic acid level.

That’s pretty much the same divergence as we’re seeing between separate outbreaks here — that is, between known North American swine flu viruses and the new Mexican H1N1.

In other words, I believe the divergence we’re seeing between the Mexican strain and known strains is pretty much what you’d expect to see in any influenza outbreak.

1. Jian, J., Chen, G., Lai, C., Hsu, L., Chen, P., Kuo, S., Wu, H., & Shih, S. (2008). Genetic and Epidemiological Analysis of Influenza Virus Epidemics in Taiwan during 2003 to 2006 Journal of Clinical Microbiology, 46 (4), 1426-1434 DOI: 10.1128/JCM.01560-07[↩][↩]

This post is the short form of the now very long post here.  This is just the summary of what I’ve done and what I’m concluding; if you want to see why plus the grimy details, check there.

1. My question: The regular influenza vaccine this year included an H1N1 virus (A/Brisbane/59/2007 (H1N1)). Will that protect against infection with the new H1N1 strain?
2. My first question was, How similar is this H1N1 to that in the current vaccine? My answer:  Not very. The vaccine strain is only 79% identical to the present strain, which isn’t terrible but isn’t very good either.  For context, the H3 hemagglutinin in the vaccine is around 45% identical to the H1; we know there’s almost no cross-reactivity between these types.  The HA from the B/Florida/4/2006, a B rather than an A strain and expected to very quite different, is about 30% identical.  See the alignment and the phylogenetic tree in that post.
3. Are the differences between the viruses especially large at the places where protective antibodies bind?  If so, then we’d expect little cross-protection between the strains.
4. First, we have to figure out where protective antibodies bind.  A paper from 1982 (Caton et al., (1982). The antigenic structure of the influenza virus A/PR/8/34 hemagglutinin (H1 subtype) Cell, 31 (2), 417-427 DOI: 10.1016/0092-8674(82)90135-0) offered several regions of H1 where protective antibodies might bind.
5. I ran a script on a few hundred different H1 proteins from different viruses, and determined that the regions Gerhard’s group said bind to protective antibodies tend to be highly variable between viruses.  This is presumably the result of selection at that region to avoid antibody responses. So this is consistent with Gerhard’s conclusion and I’m willing to go with those regions as antibody binding sites.
6. I compared the amino acids in the regions Gerhard flagged between the vaccine strain, and the new H1N1 strain. If those regions are very similar then I would expect good cross-protection; if very different, I would expect little cross-protection.  The result (shown below) is that there are extensive differences between the putative protective antibody binding sites.

Conclusion (based on lots of assumptions!):

You can see that of the five antibody-binding sites (Sa, Sb, Ca1, Ca2, and Cb), four are really very different, while one is quite similar.  Even in that one,  the single difference, from Ser to Pro, is a drastic change that would probably significantly reduce antibody binding.  So most antibodies wouldn’t bind well to the new H1N1.  However, the most similar region (”Sa”) is the one that Gerhard flagged as the most important for antibody binding, so I’m leaning to the concept that there probably will be a little bit of cross-protection, but not a lot.

But what I would really like to see is an actual experiment, testing cross-protection and cross-reactivity between the epidemic and the vaccine H1N1.

Update 5. I compared the amino acids in the putative protective antibody binding sites, and tentatively conclude that while there may be a little cross-protection between the vaccine H1N1 and the epidemic H1N1, there’s not likely to be a lot.

Update 4.5: This is getting way past tl;dr territory, so I’m going to show my work in this post but add another post to give just the summary is what I’m doing and concluding

Update 4: Looked at conserved residues to try to validate Gerhard’s epitope prediction. Conserved areas seem consistent with his conclusions

Update 3: Quick comparison of Gerhard’s prediction of antigenic sites, vs. the differences between the vaccine and the epidemic H1N1

Update 2: Added reference and figure from Walter Gerhard’s 1982 analysis of antibody epitopes of H1 hemagglutinin.)
The regular influenza vaccine this year included an H1N1 virus. Will that protect against infection with the new H1N1 strain? (See what I did there?  I didn’t say “swine flu.”)  (Oops.)

I’ve seen several statements in the media that the vaccine won’t protect, but I haven’t seen any explanation why not.  (If anyone knows more about how it was determined that the vaccine wouldn’t give much cross-protection, and who said it, please let me know.)  But it wouldn’t be surprising if there wasn’t much cross-protection; two different H1s can be significantly different.

(I’m assuming that most people who are interested in this are at least generally aware of the H and N nomenclature and so on.  If not, The Virology Blog has a brief outline of the structure of influenza here, and there are primers on influenza at The Flu Wiki here, here, here, and here.)

My first question was, How similar is this H1N1 to that in the current vaccine?  (This isn’t going to definitively answer the question of cross-reactivity, but it can give us some pointer.  This year’s vaccine included the following viruses (here’s the WHO page with a link to a PDF with more info):

• an A/Brisbane/59/2007 (H1N1)-like virus;
• an A/Brisbane/10/2007 (H3N2)-like virus;
• a B/Florida/4/2006-like virus.

So what’s the Brisbane H1N1 look like?  How close is it to the epidemic H1N1?  The answer is, Not very. I ran a quick-and-dirty alignment and phlyogenetic tree of the three vaccine virus HA proteins (not nucleic acids here because we’re interested in what antibodies might see).  The vaccine strain is only 79% identical to the present strain, which isn’t terrible but isn’t very good either.  (Quick update for context:The H3 hemagglutinin in the vaccine is around 45% identical to the H1; we know there’s almost no cross-reactivity between these types.  The HA from the B/Florida/4/2006, a B rather than an A strain and expected to very quite different, is about 30% identical.)  The phylogenetic tree shows that the present viruses are, as we already know, tightly clustered and relatively distant from the vaccine H1N1.

I’m not an antibody guy (or, as I keep pointing out, an influenza guy) but I believe there’s something known about where on HA and N the main neutralizing antibody recognition sites are.  If I find time through the day (may not happen; I have meetings in the morning, and this afternoon I’m volunteering at my son’s Grade 3 science class) I’ll try to track that down and see if we can predict neutralzing antibody cross-reactivity between the vaccine strain and the epidemic H1N1.

Update 2. Walter Gerhard published a paper in 1982¹ looking at major binding sites on H1.  Here’s the most relevant figure from his paper — later today I’ll look at how this compares to the presence H1N1 virus protein.

Update 3: Gerhard says the most important antibody binding sites on H1 are what he defined as the Sa, Sb, Ca1, Ca2, and Cb sites, with the Sa site being the most important. Here’s the relevant residues. They refer to the PR8 influenza strain, so to see how our H1s compare I will have to dig the sequence of that one up (I have it somewhere already, since I use the hemagglutinin in the lab as a marker quite a bit, but I won’t have time for that for a couple of hours):

Update 4: To look at Gerhard’s conclusions from a different viewpoint, I looked at conserved residues within influenza H1s. My reasoning here is that areas in the hemagglutinin that are vulnerable to antibodies will be rapidly selected against, and therefore will be lesat conserved. In other words, the amino acids that are most different from each other, in different flu strains, are most likely to be antibody epitopes. I downloaded a thousand or so H1 protein sequences, and ran a python script to determine amino acid frequency at each position, then plotted it out. Here is frequency (Y axis) vs. position (X axis) for the first third of H1. A completely conserved amino acid will have a frequency of 100, less than that means variable. I’ve boxed the Sa region that Gerhard says is likely the most important antibody target. As you can probably see (click for a larger version), this region is also highly non-conserved — presumably the result of selection at that region to avoid antibody responses. So this is consistent with Gerhard’s conclusion and I’m willing to go with those regions as antibody binding sites.

Update 5. Assuming that those sites are actually the important antigenic sites — and also assuming I didn’t lose track of numbering and forget what I was doing ( this was run in between volunteering at one son’s Grade 3 science class, getting dinner ready, taking son 1 to soccer practice, picking up son 2, picking up son 1 from soccer practice, working on dinner, and, oh yes, doing the work I actually have to do) — the result isn’t completely clear — but my guess would be that the vaccine H1N1 would confer some, but not a lot, of protection against the epidemic H1N1.

Here’s the table showing the vaccine amino acids (“Bris”) vs the epidemic amino acids (“Tex”) in each of the putative protective antibody-binding sites.

I’ve highlighted differences in pink, identities in green.  You can see that of the five antibody-binding sites, four are really very different, while one is quite similar.  Even in that one,  the single difference, from Ser to Pro, is a drastic change that would probably significantly reduce antibody binding.  So most antibodies wouldn’t bind well to the new H1N1.  However, the most similar region (“Sa”) is the one that Gerhard flagged as the most important for antibody binding, so I’m leaning to the concept that there probably will be a little bit of cross-protection, but not a lot.

The phylogenetic tree includes three of the present H1N1 strain, plus the three components of this year’s vaccine (click for a larger version):

Here’s the alignment, the A/Texas/05/2009(H1N1) isolate from the present outbreak and the  A/Brisbane/59/2007(H1N1) from the vaccine (let’s see if I can make the fonts work out properly so I don’t have to paste in an image here; you may need to play with your browser window to stop the lines from wraping and let them line up .. yeah, that’s ugly, but it’s the best I can do for now):

A/Texas/05/2009(H1N1) vs. A/Brisbane/59/2007(H1N1)
Identities = 449/566 (79%), Positives = 497/566 (87%), Gaps = 1/566 (0%)

Texas  1    MKAILVVLLYTFATANADTLCIGYHANNSTDTVDTVLEKNVTVTHSVNLLEDKHNGKLCK  60
            MK  L+VLL TF    ADT+CIGYHANNSTDTVDTVLEKNVTVTHSVNLLE+ HNGKLC
Brisb  1    MKVKLLVLLCTFTATYADTICIGYHANNSTDTVDTVLEKNVTVTHSVNLLENSHNGKLCL  60

Texas  61   LRGVAPLHLGKCNIAGWILGNPECESLSTASSWSYIVETSSSDNGTCYPGDFIDYEELRE  120
            L+G+APL LG C++AGWILGNPECE L +  SWSYIVE  + +NGTCYPG F DYEELRE
Brisb  61   LKGIAPLQLGNCSVAGWILGNPECELLISKESWSYIVEKPNPENGTCYPGHFADYEELRE  120

Texas  121  QLSSVSSFERFEIFPKTSSWPNHDSNKGVTAACPHAGAKSFYKNLIWLVKKGNSYPKLSK  180
            QLSSVSSFERFEIFPK SSWPNH +  GV+A+C H G  SFY+NL+WL  K   YP LSK
Brisb  121  QLSSVSSFERFEIFPKESSWPNH-TVTGVSASCSHNGESSFYRNLLWLTGKNGLYPNLSK  179

Texas  181  SYINDKGKEVLVLWGIHHPSTSADQQSLYQNADAYVFVGSSRYSKKFKPEIAIRPKVRDQ  240
            SY N+K KEVLVLWG+HHP    DQ++LY   +AYV V SS YS+KF PEIA RPKVRDQ
Brisb  180  SYANNKEKEVLVLWGVHHPPNIGDQKALYHTENAYVSVVSSHYSRKFTPEIAKRPKVRDQ  239

Texas  241  EGRMNYYWTLVEPGDKITFEATGNLVVPRYAFAMERNAGSGIIISDTPVHDCNTTCQTPK  300
            EGR+NYYWTL+EPGD I FEA GNL+ PRYAFA+ R  GSGII S+ P+  C+  CQTP+
Brisb  240  EGRINYYWTLLEPGDTIIFEANGNLIAPRYAFALSRGFGSGIINSNAPMDKCDAKCQTPQ  299

Texas  301  GAINTSLPFQNIHPITIGKCPKYVKSTKLRLATGLRNVPSIQSRGLFGAIAGFIEGGWTG  360
            GAIN+SLPFQN+HP+TIG+CPKYV+S KLR+ TGLRN+PSIQSRGLFGAIAGFIEGGWTG
Brisb  300  GAINSSLPFQNVHPVTIGECPKYVRSAKLRMVTGLRNIPSIQSRGLFGAIAGFIEGGWTG  359

Texas  361  MVDGWYGYHHQNEQGSGYAADLKSTQNAIDEITNKVNSVIEKMNTQFTAVGKEFNHLEKR  420
            MVDGWYGYHHQNEQGSGYAAD KSTQNAI+ ITNKVNSVIEKMNTQFTAVGKEFN LE+R
Brisb  360  MVDGWYGYHHQNEQGSGYAADQKSTQNAINGITNKVNSVIEKMNTQFTAVGKEFNKLERR  419

Texas  421  IENLNKKVDDGFLDIWTYNAELLVLLENERTLDYHDSNVKNLYEKVRSQLKNNAKEIGNG  480
            +ENLNKKVDDGF+DIWTYNAELLVLLENERTLD+HDSNVKNLYEKV+SQLKNNAKEIGNG
Brisb  420  MENLNKKVDDGFIDIWTYNAELLVLLENERTLDFHDSNVKNLYEKVKSQLKNNAKEIGNG  479

Texas  481  CFEFYHKCDNTCMESVKNGTYDYPKYSEEAKLNREEIDGVKLESTRIYQILAIYSTVASS  540
            CFEFYHKC++ CMESVKNGTYDYPKYSEE+KLNRE+IDGVKLES  +YQILAIYSTVASS
Brisb  480  CFEFYHKCNDECMESVKNGTYDYPKYSEESKLNREKIDGVKLESMGVYQILAIYSTVASS  539

Texas  541  LVLVVSLGAISFWMCSNGSLQCRICI  566
            LVL+VSLGAISFWMCSNGSLQCRICI
Brisb  540  LVLLVSLGAISFWMCSNGSLQCRICI  565

1. CATON, A. (1982). The antigenic structure of the influenza virus A/PR/8/34 hemagglutinin (H1 subtype) Cell, 31 (2), 417-427 DOI: 10.1016/0092-8674(82)90135-0[↩]