Fig 2: Cellular mechanisms of AGR2 secretion AFormation of GFP-LC3 autophagic puncta in TMED2 overexpressing HEK239T cells as monitored using confocal microscopy (right panel). DAPI was used for nuclear staining visualization. Percentage of GFP-LC3 puncta in control (CTL) and TMED2 overexpressing HEK293T cells as quantified from three independent experiments by counting 240 GFP-LC3 positive cells for each condition (left panel). (**): P = 0.0032. The graph represents average western blot normalized signal ±SD. For statistical analyses, the Mann–Whitney test was used.BWestern blot detection (left panel) and quantification (right panel) of LC3 level in control and TMED2 overexpressing HEK293T cells treated or not with 50 µM chloroquine for 2 h. Actin (ACT) served as a loading control. (*): P = 0.0498. The graph represents average western blot normalized signal ±SD. For statistical analyses, the Mann–Whitney test was used. n = 4.CWestern blot analysis of AGR2 expression in control and TMED2 overexpressing HEK293T cells upon autophagy inhibition with 20 and 10 µM chloroquine treatment. Actin (ACT) served as a loading control.DHEK293T cells were transfected with control (ev) or TMED2 plasmid and then used in the ERMIT assay with AGR2 WT as described in Fig 1E in the presence of gradual amounts of chloroquine. The graph represents average signal ±SD. For statistical analyses, the Mann–Whitney test was used, n = 4.EWestern blot analysis of secreted AGR2 upon TMED2 overexpression. eAGR2, extracellular AGR2; iAGE2, intracellular AGR2.FRepresentative pictures of extracellular vesicles (EVs) heterogeneity purified from conditioned media of HEK293T cells transfected with control (ev) or TMED2 plasmid and analyzed by cryo-electron microscopy. Right upper panel: Western blot analysis of CD63 in total medium (left) and in the purified extracellular vesicles fraction purified from media conditioned by control and TMED2 overexpressing cells (middle) and Western blot analysis of AGR2 levels (right) in extracellular vesicles enriched from culture media conditioned by control and TMED2 overexpressing cells.GWestern blot analysis of secreted AGR2 upon TMED2 silencing eAGR2, extracellular AGR2; iAGE2, intracellular AGR2.HQuantification of extracellular AGR2 level (eAGR2) in cells silenced for TMED2 (siTMED2), compared to control (siGL2). Data are representative of three independent experiments and are presented as extracellular-to-intracellular AGR2 ratio. (**): P = 0.0029. The graph represents average western blot normalized signal ±SD. For statistical analyses, the Mann–Whitney test was used.ISecretion of AGR2 wild-type (wt), E60A, and ?45 mutant as determined by Western blot. eAGR2, extracellular AGR2; iAGE2, intracellular AGR2.

Fig 3: AGR2 status alteration impacts on client protein biogenesis AThe ERMIT-based validation of the complex formation between AGR2 and LYPD3 or CD59, but not OS9. The monomeric AGR2 E60A was used as bait, while LYPD3, CD59, and OS9 were used as prays (n = 6). The graph represents average signal normalized on reporter protein expression ±SD. For statistical analyses, the Mann–Whitney test was used.B, CExpression and subcellular localization of CD59-GFP WT and C94S.D, ECo-immunoprecipitation of CD59-GFP WT or CD59 C94S mutant with AGR2 WT (D) or AGR2 AA mutant (E) in HEK293T cells.F–IFACS analysis of CD59 WT (F and G) or CD59 C94S (H and I) total (F and H) and membrane (G and I) expression in HEK293T cells silenced for AGR2 (siAGR2) and TMED2 (siTMED2) when compared to control (siGL2) cells (F and H) or silenced for TMED2 (siTMED2) and overexpressing AGR2 WT or AGR2 AA mutant (G and I). Data are representative of three to four independent experiments. Membranous CD59 expression was detected with anti-GFP APC. ns: non-statistically significant; for (F) (*): P = 0.0385, (***): P = 0.0003 (left panel) and P = 0.0002 (right panel), respectively; for (G) (*): P = 0.0334, (**) P = 0.0036, and (***) P = 0.0003; for (H) (*): P = 0.0171 and (**): P = 0.0084; and for (I) (**) P = 0.0086 (left panel) and P = 0.0076 (right panel), respectively. The graph represents average GFP signal ±SD. For statistical analyses, the Mann–Whitney test was used.

Fig 4: siRNA screen for identifying regulators of AGR2 dimer formation AThe first screen was carried out using HEK293T cells transfected with the ERMIT bait/prey, system, the XBP1 splicing reporter, and the siRNA library (n = 5). The counter screen was carried out using the same experimental setup without the ERMIT bait/prey system (n = 4).BVolcano plot representing the statistical analysis of the data from the screen, where the y axis represents statistical criteria and the x axis the fold change in luminescence units.CNumber of genes associated with the indicated processes, as revealed by functional analysis based on Gene Ontology and Reactome terms, (*): Apoptotic signaling pathway (P = 0.0083), ER-associated degradation (P = 0.0076), and lipid biosynthetic process (P = 0.0059), respectively; (**): ER stress (P = 0.0044), Protein folding (P = 0.003), Calcium homeostasis (P = 0.0022), N-glycosylation (P = 0.0018), respectively.D, EEnrichment of AGR2 homodimer enhancers in functions associated with ER homeostasis (green) and of AGR2 homodimer inhibitors in functions associated with ER homeostasis imbalance (red).

Fig 5: AGR2 dimerization with ERMIT assay, principles, and validation of the methodThe ERMIT assay relies on the signaling properties of IRE1, one of the three ER stress sensors and reports for a dimerization event occurring in the lumen of the ER. The assay can be applied to heterodimerization or homodimerization events. AUpper panel: NMR structure of the non-covalent dimer of AGR2 (PDB ID: 2LNS). The dimer domain is highlighted with a yellow circle (residue 54–70 of each monomer). Monomer A is colored green and monomer B in pink. The figure was generated in Chimera. Lower panel: a close-up illustration of the dimer domain, showing that the dimer is stabilized through two salt bridges between E60 and K64 of each monomer.BMolecular modeling showing root-mean-square deviation (RMSD) plot for the MD simulation of the AGR2 WT and E60A mutant. The interaction energies are defined as the sum of the short-range Coulomb interactions and the short-range Lennard-Jones potential between monomer A and B.C, DWestern blot showing the expression of AGR2 dimers (D) and monomers (M) in HEK293T subjected to DSP-mediated cross-linking and that were previously transfected with either a control siRNA (siCTL) or a siRNA targeting AGR2 (siAGR2) for 24 h (C) or treated or not with tunicamycin (Tun) prior to cross-linking (D). Reduced or non-reduced samples were resolved by SDS–PAGE and immunoblotted using anti-AGR2, anti-ERK1, or anti-calnexin (CANX) antibodies (for loading control).EPrinciples of the ERMIT assay. A wild-type IRE1 bait is used to report for dimerization, whereas a kinase catalytic mutant (K599A) is used as a control to prove that signal observed with the wild-type form is due to IRE1 activation.FAGR2 dimerization was monitored with ERMIT [wild-type (WT)]. The luminescence signal was abrogated when using the kinase dead (KD) constructs and by the constructs exhibiting mutations in the dimerization domain [E60A or C81S or double mutant (DM)]. The graph represents average signal normalized on reporter protein expression ± SD [n = 5; **: WT/KD (P = 0.002), E60/WT (P = 0.0034), E60/KD (P = 0.001), C81/KD (P = 0.004), DM/WT (P = 0.0021), and DM/KD (P = 0.0017), respectively; *: C81/WT (P = 0.0098)]. The Mann–Whitney statistical test was used.GCells expressing various bait and prey constructs and the XBP1 splicing reporter or the XBP1 splicing reporter alone were exposed to increasing concentrations of DTT. The ERMIT signals obtained with both baits were then normalized to that of XBP1s to obtain results independent of the activation of endogenous IRE1 by the use of chemical ER stressors (n = 4). The graph represents average signal normalized on reporter protein expression ±SD.