Expression of GFP-LC3 is now in widespread use to visualize autophagy in cultured cells. Recently, Kuma et al. (Autophagy 2007; 3:323-8) highlighted some complications using GFP-LC3, demonstrating that punctate dots containing GFP-LC3 do not always represent autophagic structures. We report here that GFP-LC3 can also rapidly aggregate into autophagosome look-alike structures when cells are permeabilized with saponin before cell fixation. Treatment with saponin reduced diffuse cytosolic and nuclear GFP-LC3 but caused an increase in the number and intensity of fluorescent puncta per cell regardless of whether the cells were induced to undergo autophagy. Saponin also induced GFP-LC3 puncta in Atg5(-/-) MEF transfected with GFP-LC3, where no LC3-II is produced, demonstrating that the puncta are autophagosome-independent. The increase in GFP-LC3 puncta was not matched by an increase in endogenous LC3-II or GFP-LC3-II detected by immunoblotting when protein samples were normalized to cell number. A qualitatively similar effect was observed when cells were treated with other detergents commonly used for membrane permeabilization, such as CHAPS, Triton X-100 or digitonin. We also noted that tubulin could not be used to normalize for protein loading on blots after applying saponin as it was selectively extracted from untreated cells but not from cells treated with vinblastine. When using mild detergents to remove background fluorescence, we recommend using a membrane-associated protein such as ATP synthase beta for normalization. Thus, detergents used prior to fixation may precipitate GFP-LC3 aggregation into structures that appear autophagosomal and so should be used with caution.