Purified, nonubiquitinated growth-associated protein of 43 kDa (GAP-43) was attacked by purified reticulocyte 20S proteasome but not by the 26S proteasome. Cleavage yielded 12 N-terminally labelled GAP-43 fragments that could be resolved by SDS/PAGE. Inhibitor experiments suggested that proteasome beta1 activity yielded the resolved bands and that proteasomebeta5 activity generated nonresolvable fragments. Processive degradation, yielding only nonresolvable fragments, therefore did not occur. Most of the resolved fragments co-migrated with fragments formed in the reticulocyte lysate translation mixture used for GAP-43 synthesis, which suggested that the fragments were also produced in the translation mixture by the endogenous reticulocyte lysate proteasome. Consistent with this idea, the addition of proteasome inhibitors to translation mixtures blocked fragment production. Ubiquitinated GAP-43 appeared to be the source of the fragments in the presence of ATP, and nonubiquitinated GAP-43 the source in the absence of ATP. The results therefore suggest that the lack of processing seen with the 20S proteasome is not an artefact arising from the way in which the 20S proteasome was purified. In one purification protocol, the GAP-43 fragments formed in translation mixtures co-purified with full-length GAP-43. These fragments were digested to nonresolvable products upon addition of purified 20S proteasome. Addition of calmodulin or G-actin blocked the consumption of both full-length GAP-43 and the co-purified GAP-43 fragments. This showed that the resolved fragments can re-enter the proteasome and be cleaved to nonresolvable products, indicating that the lack of processivity is not a result of their resistance to further proteasome attack. The difficult step therefore appears to be the transfer of the large fragments within the proteasome from the beta1 to the beta5 activity for further attack.