Moving fast meant navigating significant uncertainties. Dosing was a particularly fraught issue, and the prospects for producing a successful mRNA drug or vaccine hinged on getting it right. A smaller dose would be easier to manufacture and less likely to produce side effects. At the same time, previous experimental mRNA vaccines had not been shown to induce the kind of long-lasting cellular immunity one could get from, say, an adenovirus vector vaccine, such as Johnson & Johnson’s. Back in 2019, Moderna published data from a Phase 1 trial of two mRNA-based bird-flu vaccines: The results had looked solid in the first month or two, but antibody levels dropped back toward baseline by month six. The two doses of those vaccines had been spaced just three weeks apart, which may have limited the body’s immune memory. John Mascola, the head of the Vaccine Research Center at the NIH, told me that durability was going to be a big unknown with all of the COVID-19 vaccines, and the Moderna team “wanted to be conservative” in selecting sufficiently large doses and spacing those doses at four weeks. They knew from early-stage trials that with just 25 micrograms, the immune response declines by one-fourth after a month. A 250-microgram dose seemed, conversely, was clearly too high. In the end, they settled on 100 micrograms.

In the meantime, Pfizer and BioNTech were still scrambling to choose among four possible mRNA-vaccine candidates. At first, the internal favorite of the scientific team was one named BNT162b1, which consisted of just a fragment of the coronavirus spike protein, known as the receptor-binding domain. (Moderna was using the full spike for its vaccine.) As was the case for Moderna, the Pfizer-BioNTech team had to figure out the right dose. Across Phase 1 trials in Germany and the U.S., the companies had tested that candidate at doses of 10, 20, 30, and 100 micrograms, injected in volunteers at just three weeks apart, compared with Moderna’s four. The highest dose produced such severe side effects, including fever and chills, that it was dropped from the trial. That’s what vaccine makers call a “hot” reaction, and it’s something Dolsten’s team wanted to steer well clear of.

Then Pfizer and BioNTech tested their own version of the full-length spike vaccine, BNT162b2—this time going up to only 30 micrograms. Because the full-length spike’s gene sequence was about five times as long as the fragment’s, each microgram of vaccine contained one-fifth the number of copies. It was immediately obvious that the side effects were less intense as a result, but the antibody response might end up being smaller too. That would take several weeks to assess—and the clock was ticking.