Fig 2: TMEM33 promotes the formation of RTN4 noncanonical autophagosomes.a ATG16L1–/– and ATG5–/– HeLa cells with or without TMEM33 knockdown were transfected with RTN4B-mCherry, after which the number of RTN4B-mCherry vesicles was counted. Scale bar, 10 μm. The data are presented as means ± SEMs. n > 15 cells from three independent experiments. p values were calculated by Student’s t-test. ****p < 0.0001. b HeLa cells were transfected with Vector or mCherry-TMEM33, and the number of endogenous RTN4B noncanonical autophagosomes colocalized with endogenous LC3 denoted by white arrows was quantified. Scale bar, 10 μm. The data are presented as means ± SEMs. n = 40, 45 cells from three independent experiments. p values were calculated by Student’s t-test. ****p < 0.0001. c Colocalization analysis of GFP-TMEM33 and endogenous RAB22A on RTN4B-mCherry noncanonical autophagosomes (upper lane) and RTN4B-mCherry vesicles (lower lane) in WT and ATG5–/– HeLa cells co-transfected with RTN4B-mCherry and GFP-TMEM33. Scale bar, 10 μm. Quantification of colocalization was presented as Pearson’s correlation coefficient (r). n = 32, 30 cells from three independent experiments. d Working model for the in vitro fusion assay between GFP-labeled RAB22AQ64L early endosomes and mCherry-labeled RTN4B noncanonical autophagosomes. e Fusion puncta (yellow), which indicate Rafeesomes, were quantified after captured using a super-resolution confocal microscopy. The represented merged puncta were shown on the xy, xz and yz axis. Scale bar, 10 μm. The data are presented as means ± SEMs. n = 11, 19 fields from three independent experiments. p values were calculated by Student’s t-test. ****p < 0.0001.

Fig 3: The ER membrane proteins TMEM33 and RTN4B are both enriched in Rafeesome.a Schematic diagram of the purification of Rafeesomes in HeLa cells stably expressing SFB-RAB22AWT or SFB-RAB22AQ64L using streptavidin-Sepharose followed by qualitative mass spectrometry analysis. b The colocalization of endogenous TMEM33 (red) and endogenous LC3 (magenta) with GFP-Vector, GFP-RAB22AWT or GFP-RAB22AQ64L was detected by immunofluorescence. Scale bar, 10 μm. Quantification of TMEM33 colocalization with LC3 was presented as Pearson’s correlation coefficient (r). The data are presented as means ± SEMs. n = 40, 41, 47 cells from three independent experiments. p values were calculated by Student’s t-test. ****p < 0.0001. The ratio of Rafeesomes induced by GFP-RAB22AQ64L containing endogenous TMEM33 and LC3 was quantified and calculated relative to total RAB22A-positive vesicles. c The enrichment of the tubular ER markers RTN4B, ATL3 or the sheet ER marker Climp63 in Rafeesomes was examined using the indicated endogenous antibodies in HeLa cells stably expressing GFP-RAB22AQ64L. The left lane shows the distribution of the indicated proteins under steady-state conditions in WT HeLa cells. Scale bar, 10 μm. The data are presented as means ± SEMs. n = 51, 50, 52 cells from three independent experiments. p values were calculated by Student’s t-test. ***p < 0.0001. The number of RAB22A-positive vesicles containing corresponding ER marker was counted. d The colocalization of endogenous TMEM33 (red) with RTN4B (magenta) within Rafeesomes was assessed in HeLa cells stably expressing GFP-RAB22AQ64L. Scale bar, 10 μm. Quantification of TMEM33 colocalization with RTN4B was presented as Pearson’s correlation coefficient (r). n = 29 cells from three independent experiments. The ratio of Rafeesomes containing endogenous TMEM33 and RTN4B was quantified and calculated relative to total RAB22A-positive vesicles. e The distribution patterns of endogenous LC3, TMEM33 and RTN4B within Rafeesomes were determined using Structure Illumination Microscopy (SIM) imaging. Scale bar, 1 μm. f Electron microscopy (EM) of the APEX2-labeled LC3, TMEM33 and RTN4B in HeLa cells stably expressing GFP-RAB22AQ64L. Scale bar, 200 nm. g Extracellular vesicle particles (EVPs) derived from HeLa cells stably expressing vector, FLAG-RAB22AWT or FLAG-RAB22AQ64L were isolated using differential ultracentrifugation, and whole cell lysates (WCLs) and EVPs were subjected to western blotting using the indicated antibodies. The data are presented as means ± SEMs. n = 3. p values were calculated by Student’s t-test. *p < 0.05.

Fig 4: The TM2 domain of RTN4 is required for its oligomerization and interaction with TMEM33.a A sketch map of the RTN4B transmembrane topology structure in the ER membrane. b HeLa cells were transiently transfected with RTN4B-truncation mutants according to a, and RTN4B oligomerization was assessed via western blotting. c HEK293T cells were co-transfected with FLAG-tagged RTN4B and V5-tagged RTN4B-truncation mutants, then the cell lysates were incubated with V5 agarose for 6 h at 4 °C, after which western blotting was performed to assess their interaction. d HEK293T cells were co-transfected with FLAG-TMEM33 and RTN4B-truncation mutants, and immunoprecipitation was subsequently performed using FLAG agarose. e Schematic representation of the TMEM33 transmembrane topology structure in the ER membrane. f Cell lysates extracted from HEK293T cells co-transfected with RTN4B-V5 and the indicated HA-tagged TMEM33-truncation mutants were incubated with HA agarose and were then subjected to western blotting.

Fig 5: A proposed model deciphering a secretory ER-phagy pathway initiated by the assembly of RAB22A/TMEM33/RTN4.on the tubular ER membrane, RAB22A/TMEM33/RTN4 assembly promotes RTN4 oligomerization, which generates RTN4 cluster-enriched microdomains. ER remodeling resulted from RTN4 microdomains induces the bud scission of membrane-bound RTN4 vesicles. ATG9A transports RTN4 vesicles for homotypic fusion, leading to the formation of the early IM, which serves as phagophore of RTN4 noncanonical autophagosomes. Finally, the secretory ER-phagy-derived RTN4 noncanonical autophagosomes are secreted via Rafeesomes as R-EVs marked by TMEM33.