Using horseradish peroxidase (HRP) as a fluid-phase endocytic tracer, we observed through the electron microscope numerous tubular endosomes with a diameter of 30-50 nm and lengths of greater than 2 microns in thick sections (0.2-0.5 microns) of AtT20 cells. These tubular endosomes are multibranching and form local networks but not a single reticulum throughout the cytoplasm. They are sometimes in continuity with vesicular endosomal structures but have not been observed in continuity with AtT20 cell late endosomes. Tubular endosomal networks are not uniformly distributed throughout the cytoplasm, but are particularly abundant in growth cones, in patches below the plasma membrane of the cell body, and surrounding the centrioles and microtubule organizing center (MTOC). Tubular endosomes at all these locations receive HRP within the first 5 min of endocytosis but approximately 30 min of endocytosis are required to load the tubular endosomal networks with HRP so that their full extent can be visualized in the electron microscope. After 10 min of endocytosis, complete unloading occurs within 30 min of chase, but between 30 and 60 min are required to chase out all the tracer from the tubular endosomes loaded to steady state during 60 min endocytosis of 10 mg/ml HRP. In interphase cells, neither the loading nor unloading of tubular endosomes depends on microtubules but in cells blocked in mitosis by depolymerization of the mitotic spindle with nocodazole, HRP does not chase out of tubular endosomes. The thread-like shape of tubular endosomes is not dependent on microtubules. Furthermore, HRP is delivered to AtT20 tubular endosomes at 20 degrees C. All these properties indicate that AtT20 cell tubular endosomes are an early endocytic compartment distinct from late endosomes. Tubular endosomes like those in AtT20 cells have been seen in cells of the following lines: PC12, HeLa, Hep2, Vero, MDCK I and II, CCL64, RK13, and NRK; they are particularly abundant in the first three lines. In contrast, tubular endosomes are sparse in 3T3 and BHK21 cells. The tubular endosomes we have observed appear to be identical to the endosomal reticulum observed in the living Hep2 cells by Hopkins, C. R., A. Gibson, H. Shipman, and K. Miller. 1990.