Swine flu: why does it take so long to make a vaccine?

On Friday Department of Health and Human Services Secretary Kathleen Sebelius announced that the US was asking vaccine manufacturers to get ready to make an vaccine against this year's swine flu. Before a vaccine can be made there is a substantial amount of preparation that needs to be done, so this just gets the process started. And the first thing that is needed is a vaccine seed strain. We've been hearing from CDC that a seed strain was in preparation. So what's a vaccine seed strain? Two good news articles, one in Nature by Declan Butler, and one in ScienceInsider by Jon Cohen have some of the details. Both are recommended. Here's our version, drawn from Butler's interview with New York Medical College's Doris Bucher, whose lab is one of a handful in the world able to grow influenza seed strains. We have added some additional background.

Bucher was hired by Edwin Kilbourne, the dean of American flu scientists (and a professor of mine in medical school in the years before he moved his lab to New York Medical College and the association with Bucher). Her lab is a successor to Kilbourne's, where in 1969 he originated the technique that has been used since the early 1970s to make seed strains for producing vaccine antigen in eggs. The influenza virus has genetic instructions to make ten or so proteins but our immune system doesn't see most of them well because they are inside the virus (but there is much we don't know about which viral proteins invoke immune responses; see an earlier post on the response to the bird flu virus here). The major proteins on the surface of the virus are the hemagglutin (HA) protein and the neuriminidase protein (NA), so one of the objectives is to be able to produce these viral proteins in large quantity so they can be incorporated into a vaccine. Unfortunately viruses isolated in an outbreak in people (called the wild type) usually doesn't grow well in chicken eggs. Kilbourne figured out how to combine the HA and NA segments of the wild types with the six remaining segments from a virus that grows well in eggs to make a virus that grows well in eggs but has wild type HA and NA on its surface. That is what is called the seed strain.

Cohen's piece in ScienceInsider reports that CDC has sent five wild type isolates (i.e., viral specimens taken from cases in this outbreak) to seven laboratories for the purpose of making a seed strain. Bucher's lab is one of them, so Butler picks up the story from there. The process starts with a standard lab flu virus called PR8 (formally A/PR/8/34), originally isolated in Puerto Rico in 1934). PR8 grows well in eggs. Like the swine flu virus, PR8 is an H1N1 virus, so Bucher used a modification of PR8 that has its H1 and N1 versions of HA and NA replaced by H3 and N2. She calls this modified PR8, X157. It still grows well in eggs but how has surface proteins of the 2005 seasonal flu virus that was a subtype H3N2. The reason for doing this will be explained shortly.

The next step is to co-infect eggs with both the wild type H1N1 and the donor H3N2 (the X157). When either gets into a host cell in the chicken egg they make copies of their eight genetic segments (two of which are HA and NA) and repackage them as a whole virus. If they co-infect a cell it can happen that the segments get repackaged more or less randomly. Each of the eight slots can have a segment from either virus, so there are 28=256 combinations. Two of them are the original viruses, but the rest are various combinations of the 8 genetic segments of a flu virus. Bucher is interested in the one that has the HA and NA of the wild type but the remaining 6 from the donor, X157.

In the first step, we co-infect an egg with the swine flu target virus and the X-157 donor; we use about 20-30 eggs in every experiment. So we inject the viruses into the allantoic cavity [a fluid-filled sac surrounding the embryo] — it's in the cells of its membrane that the virus replicates. We put in enough virus to be sure of getting both strains into the same cell, as you need that to get the gene shuffling event to take place. Then we incubate them for 42 hours. The allantoic cavity contains about 10 millilitres of fluid, and once you are done, you just slurp out the liquid and proceed to the next step. (Declan Butler, Nature)

The next step is to find the combinations of the 254 that have the wild type H1 and N1 on the outside. This is where the substitution of H3 and N2 in PR8 comes in. She uses antibodies to the H3N2 target to kill all the combinations that have these surface proteins from X157. If the donor had H1 and N1 on the outside it would be harder to carry out this selection process, hence the use of X157 instead of PR8. After finding the viruses with H1 and N1 on the outside, there are 26=64 possible combinations left, each with various mixes of the wild type and X157 internal genes. Bucher then carries out 3 growth cycles of 42 hours, each time selecting the fastest growing highest yield viruses. These are the ones best adapted to growing in eggs.

After that, we amplify the virus for 42 hours, clone the reassortants, and then amplify them again, at which point we have about 80 millilitres of seed virus to ship to the CDC and other labs, who will test them with immune sera to ensure that the reassortant virus looks like H1N1. They will also infect ferrets with the reassortant to evaluate immune response. The CDC also sequences the haemagglutinin to monitor any changes that may have occurred with egg adaptation. We'll usually offer about three different candidate viruses.

According to Bucher this is going well and she expects to get CDC a seed strain on schedule by the end of the month.

But production of the seed strain is just the initial step and it's well underway. The vaccine manufacturers now will grow the seed stock in bulk and now must make it into a vaccine that contains adequate amounts of the wild type H1 and N1 to raise a protective immune response when given as an intramuscular injection (I am only covering conventional inactivated flu vaccine techniques here; there are other possibilities). Things get more complicated at this point and there are several options. Different manufacturers do things differently so each vaccine preparation must be tested for safety and efficacy.

The first task in an inactivated influenza vaccine is to inactivate the virus, i.e., make it non-infecious. When you grow it in eggs it produces an infectious virus. Classically the two agents used to inactivate the virus are formalin (formaldehyde) and β-propriolactone. During usual processing β-propriolactone is degraded and is below the level of detection in the final product. Formaldehyde is also greatly reduced but may be still detectable. It is not thought to be harmful in the residual quantities in vaccine. Formaldehyde is part of the normal metabolic economy of the body, so this claim is plausible.

Vaccines using the whole (but inactivated) virus can be very effective but they also produce more side effects. While still in use in some places, the common practice since the 1970s is to produce vaccines that contain only the most immunogenic portion of the virus, principally HA and NA. So the goal is to make a preparation with a high concentration of HA and NA but lower concentrations of less immunogenic viral proteins and of non-viral proteins from eggs or extraneous agents (e.g., avian leukosis virus that may have been in the chicken flocks that provided the eggs or other micro-organisms or cell debris). A physical process such as centrifugation over a sucrose gradient or passage of the allantoic fluid over a chromatographic column may be used, followed by dialysis or ultrafiltration. There are lots of variations, including additional processes that disrupt the viral envelope to free the HA and NA and remove some of the other viral proteins (this is called "splitting" the virus). There are different solvents that can be used for splitting, adding still more variation in commercial preparations. While the final HA content is standardized through assay, differences in manufacturing processes mean that different vaccines may have different propensities to cause side effects or have somewhat different efficacy. Hence the need for regulation and testing.

That testing involves both animal testing (usually in ferrets) and clinical trials, which also take time. The point is both to assure safety and to estimate efficacy, as measured by a preparation's ability to raise antibodies in human volunteers. Part of the efficacy testing involves the essential question of how many doses are needed and at what level of HA antigen assay. Some of the money that Secretary Sebelius announced for the swine flu effort will go to paying for the necessary clinical trials.

I've barely scratched the surface of this complicated process, concentrating on the part currently underway, the preparation of a vaccine seed strain, but I hope it's enough to indicate why it seems to take so long to do this. New technologies are in the pipeline to make some of the steps quicker, but at this point we are still dealing with an old but established technology.