Serial Passage and Evolutionary Trade-Offs

Max Theiler and the One Hundred Seventy-Six Transfers of Yellow Fever 17D

Inside a liquid-nitrogen vault sits seed lot 17D-204, a viral lineage separated from a lethal West African hemorrhagic fever by thirty-two amino acid substitutions across a ten-thousand-nucleotide RNA genome. Every dose of yellow fever vaccine administered worldwide over the past eight decades—more than eight hundred million immunizations grown in embryonated chicken eggs—descends directly from a single series of tissue-culture flasks passaged in New York between 1931 and 1936.

Yellow fever is caused by an enveloped positive-sense single-stranded RNA flavivirus transmitted between primates and humans by the bite of Aedes aegypti and Haemagogus mosquitoes. In severe human infections, the virus exhibits fierce viscerotropism. After replicating in regional lymph nodes, it seeds Kupffer macrophages and hepatocytes in the liver, triggering widespread midzonal apoptotic necrosis, fulminant hepatic failure, deep jaundice, and gastrointestinal hemorrhage that kills 20 to 50 percent of symptomatic patients. In June 1927 in Lagos, Nigeria, Rockefeller Foundation investigators Adrian Stokes, Johannes Bauer, and N. Paul Hudson drew venous blood from a Ghanaian man named Asibi who was recovering from a mild febrile illness. They inoculated the sample into an Indian rhesus macaque (Macaca mulatta) and isolated the causal agent of yellow fever for the first time. The laboratory hazard was immediate and lethal. Two months later in September 1927, Adrian Stokes died in Lagos of laboratory-acquired yellow fever, becoming one of six Rockefeller Commission scientists killed by the virus during that era.

Maintaining the virulent Asibi strain by monkey-to-monkey passage was hazardous and prohibitively expensive. In 1930 at Harvard Medical School, before joining the Rockefeller Institute in New York, South African virologist Max Theiler discovered that although adult white mice resisted intraperitoneal injection of yellow fever virus, intracerebral inoculation into the mouse cortex caused fatal encephalitis. As Theiler serially passaged the virus from mouse brain to mouse brain over more than two hundred transfers, a striking evolutionary trade-off emerged. Adapting to murine neural tissue progressively stripped the virus of its viscerotropism in rhesus monkeys, so subcutaneously injected macaques no longer suffered liver necrosis. Yet the same mouse-brain selection amplified the virus's neurotropism, rendering the mouse-adapted strain unsafe for widespread human subcutaneous use because of the risk of post-vaccinal encephalitis.

To eliminate both hepatotropism and neurotropism simultaneously, Theiler and colleague Hugh Smith designed an empirical evolutionary bottleneck using in vitro tissue cultures that lacked both mammalian hepatocytes and central nervous system neurons. Starting with the virulent Asibi strain, they passaged the virus 18 times in minced whole mouse embryo tissue, followed by 58 passages in minced whole chick embryo tissue suspended in Tyrode's solution with normal monkey serum. At passage 76—establishing experimental lineage 17D—Theiler and Smith introduced the decisive anatomical modification. Before mincing each chick embryo at twelve days of incubation for the culture flasks, they surgically excised and discarded the embryonic brain and spinal cord with fine scissors. Deprived of avian neural cells, any viral variant that relied on neurotropic receptors lost its replicative advantage inside the flask.

Between the 89th and 114th subcultures of the 17D series in 1936, routine testing in rhesus macaques and mice revealed a sudden phenotypic phase transition. Monkeys injected subcutaneously with 17D supernatant showed transient, low-level viremia and zero liver pathology. Even monkeys injected intracerebrally survived without paralysis while developing robust neutralizing antibodies that withstood a subsequent challenge with lethal Asibi virus. Why does prolonged serial passage in non-native avian fibroblasts attenuate a human pathogen? Because the viral RNA-dependent RNA polymerase (NS5) lacks a $3'\to 5'$ proofreading exonuclease, it introduces roughly one point mutation per ten thousand nucleotides copied, generating a diverse quasispecies swarm in every flask. Mutations in the viral envelope glycoprotein (E) and nonstructural genes (NS2A, NS2B, NS4B) that optimized receptor binding and replication kinetics inside headless chick embryo cells imposed an antagonistic pleiotropic fitness cost in mammalian liver and neural tissue. Genomic sequencing of the 176th passage against the parental Asibi isolate confirmed 68 nucleotide mutations encoding 32 amino acid changes, twelve of which cluster in the E glycoprotein that mediates host-cell attachment and endosomal membrane fusion. Those same envelope mutations abolished the ability of 17D to infect the midgut epithelium of Aedes aegypti mosquitoes, ensuring that vaccinated humans could never spark vector-borne transmission and earning Max Theiler the 1951 Nobel Prize.