Serum ferritin has been used widely in clinical medicine chiefly as an indicator of iron stores and inflammation. Circulating ferritin also can have paracrine effects. Despite the clinical significance of serum ferritin, its secretion remains an enigma. The consensus view is that serum ferritin arises from tissue ferritins--principally ferritin light--which can be glycosylated. Ferritin heavy and light chains are cytosolic proteins that form cages of 24 subunits to store intracellular iron. We show that ferritin light is secreted when its expression is increased in stable, transfected HepG2 cells or adenovirus-infected HepG2 cells. Export occurs through the classical secretory pathway and some chains are N-glycosylated. Ferritins do not need to form cages prior to secretion. Secretion is blocked specifically, effectively, and rapidly by a factor in serum. The timing of this inhibition of ferritin secretion suggests that normally cytosolic ferritin L is targeted to the secretory pathway during translation despite the absence of a conventional signal sequence. Thus, secretion of glycosylated and unglycosylated ferritin is a regulated and not a stochastic process.