Every medical student learns that urine is yellow because of urobilin. Almost none of us learn who makes the urobilin. The answer, published in Nature Microbiology in January 2024, is that you do not. Bacteria in your colon do, using an enzyme no human cell carries.

The coverage at the time announced that scientists had finally figured out why urine is yellow. That is not quite what happened, and the paper is honest about it in the discussion. Max Jaffé identified urobilin as the yellow pigment in 1868, when he noticed that adding zinc to the urine of feverish patients produced a green fluorescence. What went missing for the next century and a half was not the pigment. It was the enzyme, and therefore the agent.

Jaffé is worth pausing on. In 1886 the same man reported that creatinine reacts with alkaline picrate to throw a red color, and clinical chemistry has been running on that reaction ever since. Many labs still measure serum creatinine by a method bearing his name, which means a substantial fraction of the eGFR values generated in this country today trace to his bench. He told us why urine is yellow and then handed nephrology the number we use to decide whether kidneys work at all.

The pathway has a known shape. Senescent red cells release heme, heme becomes bilirubin, the liver conjugates it with glucuronic acid, and bile carries it into the gut. There, glucuronidases strip the sugar off. Unconjugated bilirubin then goes one of two ways. It can be reabsorbed into the enterohepatic circulation, which raises serum bilirubin, or it can be reduced to urobilinogen and stercobilinogen, which are excretable. Urobilinogen is colorless. On contact with air it oxidizes to urobilin, which is yellow.

Only bacteria perform that reduction. No human enzyme does it. So the step that determines whether heme breakdown products leave the body or come back around is outsourced entirely to organisms that are not you, and until 2024 nobody knew which gene was doing the work.

The method is the part I admire. The team grew a panel of gut anaerobes in bilirubin-supplemented media and screened for reduction using fluorescence. Nine strains came back positive and thirteen negative. Three species had not previously been known to reduce bilirubin: Clostridium symbiosum, Clostridium sp. M62/1, and Ruminococcus gnavus CC55_001C. Every reducer fell within class Clostridia.

Then they exploited the disagreement. Some of the non-reducers were close relatives of reducers, which meant the responsible gene should be present in one set and absent in the other. Across ten genomes they found 6,256 orthogroups, of which 389 were predicted oxidoreductases, of which exactly two tracked the phenotype. One was an isoprenoid biosynthesis enzyme and made no chemical sense. The other was unannotated and homologous to 2,4-dienoyl-CoA reductase, which reduces carbon to carbon double bonds. They named it bilR.

Sufficiency came next. They cloned bilR from three species into E. coli, an organism that does not reduce bilirubin, and it did. Mass spectrometry confirmed urobilin and found no stercobilin. Mutating two active site residues to alanine killed the activity while circular dichroism showed the protein still folded, which rules out the boring explanation that they had simply broken the enzyme.

Having a gene to look for, the team went looking. In 1,801 metagenomes from healthy adults, bilR was absent in 0.1 percent. It is effectively a core function of the adult human gut. In 4,296 infant metagenomes it was frequently absent during the first months of life and mostly present by the end of the first year, which maps onto the window when neonatal jaundice is most common. In inflammatory bowel disease it was absent significantly more often than in health, in both Crohn's disease and ulcerative colitis.

The neonatal observation is the one that will make you sit up, and it is also the one to hold loosely. This is presence and absence of a gene in stool metagenomes lined up against the known epidemiology of a disease. Nobody measured serum bilirubin in these infants. The authors say as much and call for cohorts that measure bilirubin, fecal urobilinoids, and bilirubin reducer abundance together. Phototherapy is not in danger. But if colonization timing turns out to modulate how high a newborn's bilirubin climbs, that is a modifiable variable, and modifiable variables in neonatal jaundice are scarce.

Note what did not get overturned. BilR is still a bilirubin reductase, still the enzyme nobody could find for 156 years, still nearly universal in healthy adults and patchy in newborns. What got revised is the claim of completeness, which is the claim the press release leaned on hardest. The headline said mystery solved. The literature said mystery named, and then two years later added a second enzyme to it.