We have previously shown that in HeLa cells, Shiga toxin B-fragment is transported retrogradely from the plasma membrane via endosomes and the Golgi apparatus to the endoplasmic reticulum (Johannes et al. (1997), J. Biol. Chem., 272, 19554-19561).Here we found that at 19.5 o C, B-fragment transport to the Golgi apparatus was inhibited. We then compared by confocal and immunoelectron microscopy on both fixed and living cells the transport of B-fragment from the 19.5 o C compartment to the Golgi apparatus to transport of several marker proteins of the endocytic pathway. At 19.5 o C, the B-fragment colocalized completely with transferrin in early/recycling endosomes. In contrast, the colocalization with fluid phase markers (3 kD dextran and BSA-gold) was less complete, and internalized EGF, which in association with its receptor is transported to lysosomes, was always found adjacent to B-fragment positive compartments with no overlap between both markers. These data suggest that at 19.5 o C, B-fragment is sorted to early/recycling endosomes, away from marker proteins that are destined for late endosomes/lysosomes. Upon shift to 37 o C, the B-fragment is rapidly transported to the Golgi apparatus (10-20 min). During this transport, no B-fragment specific label is found in BSA-gold containing poly-vesicular (late) endosomes, suggesting that B-fragment transport to the Golgi apparatus occurs from early/recycling endosomes. This hypothesis is further strengthened by the finding that in morphological and biochemical experiments, several drugs (bafilomycin A 1 , concanamycin B, and cytochalasin D) with established effects on vesicular transport in the late endocytic pathway did not affect B-fragment transport to the Golgi apparatus. In contrast, in the presence of brefeldin A, the B-fragment accumulated in tubular structures that also contained a marker protein of early endosomes, the transferrin receptor, while EGF was efficiently degraded in lysosomes, indicating that the transport route to late endosomes/lysosomes was open.Taken together, these results are consistent with the hypothesis that Shiga toxin B-fragment is transported directly from early/recycling endosomes to the Golgi apparatus. This novel transport route could also be used by cellular proteins, as deduced from our finding that TGN38, a protein which at steady state is concentrated in the TGN but which has been reported to cycle through the plasma membrane, colocalizes with the B-fragment on its transport to the Golgi apparatus.