Rationale behind our model substrate, system, and strategy to dissect the role of TMED9 in RESET

To dissect the role of TMED9 in RESET, we used a mutant variant of YFP-PrP WT, called “YFP-PrP*” as our primary misfolding model substrate. YFP-PrP WT includes the N-terminal prolactin signal sequence, which drives efficient translocation into the ER [25], and yields a homogenous population of PrP that efficiently folds and traffics to the cell surface [13,26]. YFP-PrP * is derived from YFP-PrP WT but contains a C179A mutation that disrupts the single disulfide bond in PrP. YFP-PrP * was stably expressed in the Normal Rat Kidney (NRK) cell line “YFP-PrP* NRK” at levels similar to endogenous PrP in mouse brain lysates [12,14]. Here, we verified our previous observations that in YFP-PrP * NRK cells, YFP-PrP * proteins were strongly localized to the ER and poorly localized on the plasma membrane at steady-state (S1A and S1B Fig and [12]).

GFP-CD59 (C94S) is another misfolded GPI-AP model substrate that contains a single cysteine mutation that disrupts 1 of 5 possible disulfide bonds [15,27]. Previously, we tested for the universality of the major findings that we made using YFP-PrP * as the model substrate by repeating the critical experiments with the unrelated, misfolded GPI-AP GFP-CD59 (C94S) [12,14]. Here, we show that GFP-CD59 (C94S) NRK cells expressed a heterogenous population of GFP-CD59 (C94S) that included clearly detectable ER and plasma membrane-localized populations (S1C and S1D Fig). Thus, we opted to use the more homogenous population of YFP-PrP * NRK cells as our primary model system to study RESET, and GFP-CD59 (C94S) as a secondary model system.

Misfolded GPI-APs are exported out of the ER via RESET for subsequent lysosomal turnover during steady-state conditions [12]. However, various ER-stressors including the induction of new glycoprotein expression or chemical ER-stressors enhance the flux of misfolded GPI-APs through RESET [12,14,20]. TG-treatment, in particular, induces the immediate and near-synchronized release of YFP-PrP * or GFP-CD59 (C94S) through the RESET pathway [12]. In NRK cells, the misfolded GPI-APs traffic from the ER to the Golgi within 30 min and to lysosomes for degradation within 90 min [12]. Thus, for this study, we used TG-treatment in our YFP-PrP * or GFP-CD59 (C94S) NRK cell lines as a tractable and facile system to assay for the integrity of the RESET pathway.

During RESET, misfolded GPI-APs release calnexin and increase binding to TMED9 and TMP21

We set out to validate and expand upon the previously reported finding that TMED9 associates with GFP-PrP * at steady-state and may be involved in its clearance from the ER via the RESET pathway in TG-treated cells [20]. We confirmed that during steady-state conditions, YFP-PrP * co-immunoprecipitates with TMED9, along with calnexin, TMP21, and TMED2 (S1E Fig), as previously published [12,20]. Similarly, we found that the alternate model RESET substrate, GFP-CD59 (C94S), co-immunoprecipitated with TMED9, calnexin, TMP21, and TMED2 at steady-state (S1F Fig).

To monitor YFP-PrP * protein-interactions as YFP-PrP * undergoes RESET, we synchronized the release of YFP-PrP * for RESET with TG-treatment and correlated live-cell time-lapse imaging with co-immunoprecipitation time courses. The majority of YFP-PrP * molecules were previously shown to have been cleared from the ER to the Golgi via the RESET pathway in under 30 min of TG-treatment [12]. Here, we reproduced this finding using YFP-PrP * NRK cells that were either co-transfected with Cerulean-calnexin (CER-CNX), which is a Cerulean (CER)-tagged version of the ER-resident chaperone calnexin to mark the ER, or CER-GalT to mark the Golgi. Within the first 20 min of TG-treatment, YFP-PrP * actively trafficked out of the ER into the Golgi leading to a significant decrease in the colocalization of YFP-PrP * with CER-CNX and significant increase in the colocalization of YFP-PrP * with CER-GalT by 20 min (Fig 1A–1D). Additionally, we imaged the trafficking of YFP-PrP * during RESET in cells co-transfected with CER-TMED9 and Golgi marker, FusionRed (FusRed)-SiT. Again, within 20 min of TG-treatment, YFP-PrP * moved into the FusRed-SiT-marked Golgi (Fig 1E and 1F, S1 Video). Intriguingly, TG-treatment did not alter the overall distribution of CER-TMED9 molecules over the course of the 60 min time-lapse videos. Moreover, the Pearson’s colocalization coefficient between CER-TMED9 and FusRed-SiT did not significantly change from the 0 to 20 min time point after TG-treatment, despite the dramatic shift in YFP-PrP * localization from the ER to the Golgi within that time frame (Fig 1E and 1F; S1 Video).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. During RESET, misfolded GPI-APs traffic from the ER to the Golgi.

(A) Time-lapse imaging sequence collected immediately after TG-treatment of a typical YFP-PrP * NRK cell that was co-expressing ER-marker, Cerulean-calnexin (CER-CNX). Scale bar represents 10 µm. (B) Plot of the average Pearson’s r values between YFP-PrP * and CER-CNX for 9 individual cells across the 0 and 20 min time points of time-lapses collected after addition of TG, as shown in (A). Pearson’s colocalization coefficients, r, were measured between YFP-PrP * and CER-CNX within the boundaries of the cell. For each image, the boundary of the cell was revealed by temporarily increasing the gain for CER-CNX. The Pearson’s r values are color-coded to signify each individual cell at different time points. The r values for 0 and 20 min time points were 0.742 ±  0.098 and 0.344 ±  0.108, respectively.(C) Time-lapse imaging sequence collected immediately after TG-treatment of a typical YFP-PrP * NRK cell that was co-expressing the Golgi marker, CER-GalT. Scale bar represents 10 µm. (D) Plot of the average Pearson’s r values between YFP-PrP * and CER-GalT for 10 individual cells across the 0 and 20 min time points of time-lapses collected after addition of TG, as shown in (C). Pearson’s colocalization coefficients, r, were measured between YFP-PrP * and CER-GalT within the boundaries of the cell. For each image, the boundary of the cell was revealed by temporarily maximizing the gain for YFP-PrP * . The Pearson’s r values are color-coded to signify each individual cell at different time points. The r values for 0 and 20 min time points were 0.300 ±  0.084 and 0.785 ±  0.060, respectively.(E) Time-lapse images collected immediately after TG-treatment of a typical YFP-PrP * NRK cell that was co-expressing CER-TMED9 and FusionRed (FusRed)-SiT as a Golgi-marker. Scale bar represents 10 µm. This figure panel is associated with S1 Video. (F) Plot of the average Pearson’s r values between YFP-PrP * and FusRed-SiT or CER-TMED9 and FusRed-SiT, as indicated, for 6 individual cells across the 0 and 20 min time points of time-lapses collected after addition of TG, as shown in (E). Pearson’s colocalization coefficients, r, were measured between YFP-PrP * and FusRed-SiT or CER-TMED9 and FusRed-SiT, within the boundaries of the cell. For each data point at 0 and 20 min time points, the boundary of the cell was revealed by temporarily maximizing the gain for YFP-PrP * . The Pearson’s r values are color-coded to signify each individual cell at different time points. The r values between YFP-PrP * and FusRed-SiT for 0 and 20 min time points were 0.462 ±  0.149 and 0.852 ±  0.079, respectively. The r values between CER-TMED9 and FusRed-SiT for 0 and 20 min time points were 0.662 ±  0.057 and 0.705 ±  0.077, respectively.

https://doi.org/10.1371/journal.pbio.3003084.g001

Next, we examined the dynamics of misfolded GPI-AP and TMED9 protein–protein interactions during RESET by treating YFP-PrP * NRK cells with TG, co-immunoprecipitating YFP-PrP * and analyzing the changes in YFP-PrP * ’s associations with TMED9, TMP21, and calnexin over time. During steady-state conditions, YFP-PrP * associates with calnexin, TMED9, and TMP21. However, within 20 min of TG-treatment, densitometry measurements of the western blot bands revealed that YFP-PrP * ’s association with calnexin declined by 57 ± 9% (Stdev.p, n  =  3) and concurrently YFP-PrP * ’s association with TMED9 and TMP21 increased by 335 ± 25% (Stdev.p, n  =  3) and 267 ± 91% (Stdev.p, n  =  3), respectively (Fig 2A and 2B). From 20 to 40 min after addition of TG, YFP-PrP * ’s association with TMED9 and Tmp21 declined.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. During RESET, misfolded GPI-APs dissociate from calnexin and increase association with TMED9 and TMP21.

(A) Representative western blots of GFP-tag co-immunoprecipitations from the parental untransfected NRK cells (P) or stably transfected YFP-PrP * NRK cells (n = 3 biological replicates). YFP-PrP * NRK cells were treated with TG and collected for co-immunoprecipitation at the indicated time points. In addition to using anti-GFP antibody to detect co-immunoprecipitation of YFP-PrP * , blots were probed for calnexin (CNX), TMED9, and TMP21. CNX and TMP21 were blotted on the same membrane. “L” indicates the lane that the ladder was loaded in. Under each western blot is depicted a “Stain-Free” image of the total protein in the gel. (B) Bar graph representing the mean band intensity for CNX, TMED9, and TMP21 in the eluates from three independently performed experiments as shown in (A). For each time point, CNX, TMED9, and TMP21 eluate band intensities were double normalized. First, they were normalized by the band intensity of eluate YFP-PrP * . Second, they were normalized against eluate time 0 band intensity of each protein probed. Error bars represent standard deviations of the mean of three independent experiments. Symbols were coded for each experiment. Statistics were calculated from unpaired t test with Welch’s correction with *  indicated p < 0.05 and ** indicated p < 0.01.

https://doi.org/10.1371/journal.pbio.3003084.g002

Synthesis of the data from Fig 1, S1 Video, and Fig 2 demonstrates the following points. Within the first 20 min after TG-treatment when YFP-PrP * was actively undergoing RESET (i.e., clearing out of the ER to the Golgi), YFP-PrP * dissociated from calnexin and increased its associations with TMED9 and TMP21. After 20 min of TG-treatment, during the time when most of the YFP-PrP * was moving through and out of the Golgi toward the plasma membrane, YFP-PrP * ’s associations with TMED9 and TMP21 declined.

Based on the steady-state association of TMED9 with diverse unrelated misfolded proteins shown in S1E and S1F Fig and in other reports [20,22,24], we hypothesized that TMED9 may be a PQC factor. In particular, the dynamic association between TMED9 and YFP-PrP * that is heightened during the period that YFP-PrP * was actively undergoing RESET shown in Figs 1 and 2 implicates TMED9’s involvement in the RESET pathway. Therefore, we set out to test whether TMED9 plays an essential role in coordinating RESET of misfolded GPI-APs.

The data underlying the graphs shown in Fig 1 are included in the S1 Data file.

The data underlying the graphs shown in Fig 2 are included in the S1 Data file.

siRNA depletion of TMED9 inhibits RESET and consequent access to the cell surface and degradation of misfolded GPI-APs

To test for a role of TMED9 in RESET, we used siRNA to deplete TMED9 levels. We examined the impact of TMED9 siRNA (siTMED9)-depletion on RESET at the population level using western blotting and live-cell imaging techniques, and at the individual cell level using immunofluorescence. For the below siTMED9 experiments, we exploited our previous observation that TG-treatment of YFP-PrP * NRK cells induces the rapid and synchronized ER-to-Golgi clearance of YFP-PrP * , followed by traffic to the cell surface where YFP-PrP * is internalized for near complete lysosomal degradation within 90 min [12]. Critical to the interpretation of the population-level TMED9-depletion experiments, immunofluorescence revealed that siTMED9-treatment effectively depleted TMED9 in 65%–70% of the cells, but a minority of cells continued to express TMED9 (S2A Fig).

First, we examined the effect of TMED9-depletion on TG-induced clearance of YFP-PrP * at the population level by western blot. In control cells (CTL) that were treated with scrambled siRNA, TG induced a 86.3 ± 6.3% (Stdev.p, n  =  3) decrease in the levels of YFP-PrP * within 90 min (Fig 3A and 3B, compare lanes 1 and 2). On the other hand, in the siTMED9-treated cells, TG induced a 34.2 ± 10.52% (Stdev.p, n  =  3) decrease in the levels of YFP-PrP * (Fig 3A and 3B, compare lanes 3 and 4), indicating a role for TMED9 in the efficient degradation of YFP-PrP * . We observed a slight and insignificant increase to 113.88 ± 17.28% (Stdev.p, n  =  3) in the steady-state levels of YFP-PrP * in siTMED9 cells when compared with CTL cells (Fig 3A and 3B, compare lanes 1 and 3). These data show that at a population level, TMED9-depletion reduces the efficiency of YFP-PrP * clearance.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. siRNA depletion of TMED9 blocks ER-export of misfolded GPI-APs but not wild-type GPI-APs.

(A) Representative western blots of control (siTMED9−) or siTMED9-treated (siTMED9+ ) YFP-PrP * NRK cells that were either untreated (TG−) or treated with TG for 90 min (TG+ ) (n = 3 biological replicates). Blots were probed for TMED9, PrP to detect YFP-PrP * and GAPDH. Each lane was numbered to facilitate cross-referencing with the quantification shown in (B). (B) Bar graph representing the mean band intensity for YFP-PrP * from three biological replicates as shown in (A). For each condition, YFP-PrP * band intensities were double normalized. First, they were normalized by the band intensity of GAPDH. Second, they were normalized against the untreated control (“siTMED9− TG−”) band intensity. Error bars represent standard deviation of the mean. Symbols were coded for each independently performed experiment. Statistics were calculated from unpaired t test with Welch’s correction with *  indicated p < 0.05 and ** indicated p < 0.01. The bars were numbered 1–4 to facilitate cross-referencing with the representative blot in (A). (C) Time-lapse images of control (CTL) or TMED9 siRNA (siTMED9)-treated YFP-PrP * NRK cells. Prior to starting the time-lapses, nuclei were stained with Hoechst. Image-collection was started immediately after the addition of TG. Scale bar represents 20 µm. (D) Percentage of cells as represented in (C) with a perinuclear Golgi-pattern of YFP-PrP * after 30 min of TG treatment. Only cells which were flat and adherent were counted. Rounded up cells undergoing mitosis were not counted. The bar graph represents three biological replicates. Symbols were coded for each independently executed experiment. The number of cells analyzed for each biological replicate is as follows (CTL: triangle 29, diamond 56, circle 51; siTMED9: triangle 69, diamond 58, circle 61). Error bars represent standard deviation of the mean. Statistics were calculated from unpaired t test with ** indicated p < 0.01. (E) Confocal images of control scrambled siRNA (CTL) treated YFP-PrP * NRK cells or parental NRK cells, or siTMED9-treated YFP-PrP * NRK cells that were incubated with TG, Alexa Fluor 647-conjugated anti-GFP antibodies and bafilomycin A1 for 90 min, fixed and stained for TMED9. An Alexa Fluor (AF) 568-conjugated secondary was used for the TMED9 immunofluorescence. Nuclei were stained with Hoechst. All images were taken with identical imaging parameters. Scale bar represents 10 µm. (F) Plot of background-subtracted mean intensities measured in arbitrary units (A.U.) for TMED9 immunofluorescence and AF647-αGFP Ab fluorescence of scrambled siRNA-treated control (CTL) parental NRK (n = 14) or CTL YFP-PrP * NRK cells (n = 17), as shown in (E). Cells were handled and images collected as shown and described in (E). Regions of interest (ROIs) around the periphery of each cell were drawn by temporarily increasing the gain in the 568 channel (TMED9-staining). We observed that in addition to strong specific TMED9 binding, a small fraction of the primary antibody against TMED9 appeared to non-specifically bind the entire cytoplasm and we took advantage of that to trace the outline of the cell and measure the mean intensities for both 568 nm (αTMED9) and 647 nm (AF647-αGFP) channels. Both the scrambled siRNA control (CTL)-treated parental NRK cells and CTL YFP-PrP * NRK cells served as positive controls for wild-type TMED9 expression. A threshold (shown as bolded vertical grid line) for wild-type TMED9 levels was placed to the left of the lowest expressing scrambled siRNA-treated (CTL) cell. CTL parental NRK served as a negative control for AF647-αGFP Ab uptake. The CTL YFP-PrP * NRK cells served as positive control for antibody uptake. A threshold (shown as bolded horizontal grid line) for antibody uptake was placed below the lowest level of AF647-αGFP Ab mean fluorescence intensity of the positive control CTL YFP-PrP * NRK, but above the highest level of AF647-αGFP Ab mean fluorescence intensity in the negative control CTL untransfected parental NRK. (G) Plot of background-subtracted mean intensities measured in A.U. for TMED9 immunofluorescence and AF647-αGFP Ab fluorescence inside of siTMED9-treated YFP-PrP * NRK cells (n = 33), as shown in (E). The data points were plotted on an identically proportioned graph to that in (F) so that direct comparisons of the data points could be made with the controls plotted in (F). Cells were handled and images collected as shown and described in (E). ROIs around the periphery of each cell were drawn and background-subtracted mean intensities measured for the 568 and 647 channels exactly as described in (F). Ten of the data points had mean TMED9 intensities within the wild-type levels and were coded with a circle symbol to represent “siTMED9 YFP-PrP* NRK knockdown failed,” and 23 of the data points had mean TMED9 below the wild-type levels and were coded with a triangle symbol to represent “siTMED9 YFP-PrP* NRK knockdown worked.” As shown in (E), in siTMED9-treated cells where knockdown worked and YFP-PrP * ER-export was blocked, the bright Golgi-patterned perinuclear staining was lost, but the low level of non-specific cytoplasmic binding persisted. (H) Immunofluorescence images of endogenous TMED9 in control siRNA or siTMED9-treated YFP-PrP * NRK cells that were depleted for TMED9. For each condition, cells were treated with TG for 0or 90 min before fixation and staining for TMED9 and ER-resident chaperone, calnexin (CNX). Nuclei were stained with Hoechst. Larger fields of views and additional data for these experiments are presented in S3A Fig. Scale bar represents 10 µm. (I) Plot of the average Pearson’s r values between YFP-PrP * and endogenous calnexin (CNX) for control (CTL) scrambled siRNA or siTMED9-treated cells at 0 or 90 min after TG addition, as shown in (H). The number of cells analyzed for each condition is listed as follows: CTL siRNA 0 min post-TG-treatment (n = 14), CTL siRNA 90 min post-TG-treatment (n = 12), siTMED9 0 min post-TG-treatment (n = 15), and siTMED9 90 min post-TG-treatment (n = 14). TMED9-depleted cells were identified by their lack of perinuclear Golgi-patterned TMED9 staining. Pearson’s colocalization coefficients, r, were measured between YFP-PrP * and CNX, within the boundaries of the cell as defined by the outline of the CNX staining. In CTL cells, the r values between YFP-PrP * and CNX for 0 and 90 min time points were 0.834 ±  0.051 and 0.538 ±  0.066, respectively. In siTMED9-treated cells, the r values between YFP-PrP * and CNX for 0 and 90 min time points were 0.797 ±  0.112 and 0.835 ±  0.100, respectively. (J) Immunofluorescence images of endogenous TMED9 in control (CTL) or TMED9 siRNA (siTMED9)-treated YFP-PrP WT NRK cells that were depleted for TMED9. Cells were fixed and stained for TMED9 and ER-resident chaperone, calnexin (CNX) at steady-state conditions. No drugs were included. Nuclei were stained with Hoechst. Scale bar represents 10 µm. (K) Plot of the average Pearson’s r values between YFP-PrP WT and endogenous calnexin (CNX) for control (CTL) scrambled siRNA or siTMED-treated cells at steady-state, as shown in (J). The number of cells analyzed for each condition is listed as follows: scrambled siRNA (CTL) (n = 22) and siTMED9 (n = 24). TMED9-depleted cells were identified by their lack of perinuclear Golgi-patterned TMED9 staining. Pearson’s colocalization coefficients, r, were measured between YFP-PrP * and CNX, within the boundaries of the cell as defined by the outline of the CNX staining. The r values between YFP-PrP WT and CNX in CTL cells and in siTMED9-treated cells were 0.325 ±  0.056 and 0.414 ±  0.066, respectively.

https://doi.org/10.1371/journal.pbio.3003084.g003

Second, we examined the impact of TMED9-depletion on RESET at the population level by live-cell time-lapse imaging. To do this, we counted the number of YFP-PrP * NRK cells that demonstrated TG-induced ER-export of YFP-PrP * by deducing the ER and Golgi localizations of YFP-PrP * based on the YFP-fluorescent patterns. The ER is well established to be the largest contiguous membrane-bound organelle that fills the cytoplasmic space and includes the nuclear envelope in mammalian cell culture [28]. In adherent cultured cells such as the NRK cells used here, the Golgi is composed of perinuclear stacks of cisternae with a characteristic appearance of a crescent-shaped ribbon that hugs part of the nucleus [29–32]. As shown previously for YFP-PrP * NRK cells in the context of organelle markers, ER localized YFP-PrP * displays a cytoplasm-filling reticulate pattern that surrounds the nucleus, while Golgi localized YFP-PrP * displays a juxtanuclear, crescent-shaped pattern that we refer to as a “perinuclear Golgi-pattern” (Figs S1A and 1, and [12]). In control (CTL) scrambled siRNA-treated YFP-PrP * NRK cells, addition of TG resulted in the rapid relocalization of YFP-PrP * from the ER to the Golgi within 30 min in 100% of the cells (n  =  136, Fig 3C and 3D). On the other hand, in siTMED9-treated YFP-PrP * NRK cells, relocalization of YFP-PrP * from the ER to the Golgi only occurred in a minor fraction of the cells (34.4%, n  =  180). In the majority (65.6%, n  =  180) of the siTMED9-treated cells, YFP-PrP * remained in the ER for at least 90 min after TG-treatment (Fig 3C and 3D). This demonstrates a block in RESET in the majority of siTMED9-treated cells.

As an orthogonal approach, we examined the impact of TMED9-depletion on RESET at the population level by antibody uptake assay. Consequent to export out of the ER to the Golgi via RESET, misfolded GPI-APs transiently access the cell surface prior to being internalized for lysosomal degradation [12,14,20]. This phenomenon occurs at steady-state or during ER stress [12,14,20]. To exploit antibody (Ab)-uptake as a visible readout for traffic of YFP-PrP * through RESET, we used Alexa Fluor 647-conjugated anti-GFP antibodies (AF647-αGFP Abs), which bind specifically to cell-surface exposed YFP-tagged misfolded GPI-APs, as published previously [12]. The AF647 continues to fluoresce inside of cells that are treated with Bafilomycin A1 (BAF), a chemical inhibitor of lysosomal acidification and degradation [12,33]. Incidentally, BAF also inhibits lysosomal degradation of YFP-PrP * , allowing it to be detected in lysosomes after RESET [12]. Thus, we incubated CTL YFP-PrP * NRK cells, CTL parental NRK cells (parent) or siTMED9-treated YFP-PrP * with TG +  BAF +  AF647-αGFP Abs and collected time-lapse images (S2B–S2D Fig, S2 Video). All of the CTL YFP-PrP * NRK cells internalized AF647-αGFP Ab (n  =  100) (S2B Fig). The untransfected CTL parental NRK cells did not internalize AF647-αGFP Ab (n  =  100) (S2C Fig). On the other hand, for the siTMED9-treated YFP-PrP * NRK cells, only 33% (n  =  100) internalized AF647-αGFP Abs (S2D Fig, S2 Video). The inhibition of Ab-uptake in the majority of siTMED9-treated YFP-PrP * NRK cells demonstrated that TMED9-expression is important for efficient ER-export via the RESET pathway.

Third, we examined the impact of TMED9-depletion on RESET at the single-cell level by immunofluorescence. To assess the effect of TMED9-depletion on Ab-uptake by YFP-PrP * in individual cells, we first set baseline mean intensities for anti-TMED9 Ab immunofluorescence and AF647-αGFP Ab fluorescence inside of scrambled siRNA-treated control (CTL) parental NRK or CTL YFP-PrP * NRK cells that were incubated with TG+BAF+AF647-αGFP Ab for 90 min prior to fixation (Fig 3E and 3F). CTL parental NRK cells served as a negative control for AF647-αGFP Ab uptake because they do not express YFP-PrP * (Fig 3E and 3F, and [12]). CTL YFP-PrP * NRK cells served as positive control for antibody uptake (Fig 3E and 3F, and [12]). Both CTL parental NRK and CTL YFP-PrP * NRK cells serve as positive controls for TMED9 expression. Of the 33 randomly selected siTMED9-treated YFP-PrP * NRK cells that we analyzed for TMED9 expression, we found that 10 of them demonstrated perinuclear Golgi-pattern TMED9 staining, indicating incomplete knockdown, while the remaining 23 were depleted for TMED9 (Fig 3E and 3G). Importantly, in siTMED9-treated YFP-PrP * NRK cells, depletion of TMED9 to levels that were quantitatively below the TMED9 levels in CTL YFP-PrP * NRK cells resulted in visible and measurable inhibition of antibody-uptake (Fig 3E–3G). This demonstrates the dependence of YFP-PrP * on TMED9 for ER-export to the cell surface on a single cell-by-cell basis.

Finally, we examined the effect of TMED9-depletion on RESET at the individual cellular level more directly by assessing ER-export of YFP-PrP * . YFP-PrP * NRK cells that were either treated with scrambled siRNA control (CTL) or siTMED9 were fixed at 0 or 90 min timepoints after TG-treatment, and co-stained for TMED9 and calnexin. Calnexin-staining served as an ER marker. In CTL YFP-PrP * NRK cells, YFP-PrP * visibly localized to the ER at steady-state and was cleared out of the ER after TG-treatment (Figs 3H and S3A). Colocalization analysis of YFP-PrP * and calnexin revealed that this correlated with a drop in Pearson’s colocalization coefficients, r, from 0.834 ± 0.05 (n  =  14 cells) at steady-state to Pearson’s r value of 0.538 ±  0.067 (n  =  12 cells) after TG-treatment (Fig 3I). In contrast, for siTMED9 YFP-PrP * NRK cells in which TMED9-levels were depleted below wild-type levels, YFP-PrP * appeared to remain localized to the calnexin-marked ER despite TG-treatment (Figs 3H and S3A). Accordingly, the Pearson’s r value was not significantly altered by TG-treatment. The Pearson’s r value between YFP-PrP * and calnexin in TMED9-depleted cells was 0.80 ±  0.11 (n  =  15 cells) at steady-state and 0.84 ±  0.10 (n  =  14 cells) in TG-treated cells (Fig 3I). This demonstrates that YFP-PrP * relies on TMED9 to leave the ER via the RESET pathway.

As with YFP-PrP * (Figs 3 and S2), siTMED9-treatment appeared to inhibit RESET of the alternate misfolded GPI-AP, GFP-CD59 (C94S) (S3B–S3E Fig). In comparison to control scrambled siRNA-treated cells, siTMED9 depleted GFP-CD59 (C94S) NRK cells exhibited (1) inefficient TG-induced clearance of GFP-CD59 (C94S), as shown through western blot analysis (S3B and S3C Fig), (2) increased steady-state levels of GFP-CD59 (C94S), as shown by western blot (S3B and S3C Fig), and (3) a significant decrease in the percentage of cells in which GFP-CD59 (C94S) underwent RESET upon TG treatment, as shown by live-cell imaging (S3D and S3E Fig). The demonstration that TMED9-depletion inhibited ER-export of two unrelated misfolded GPI-APs suggest that TMED9 could play a general role in the ER-export step of the RESET pathway for misfolded GPI-APs.

siRNA depletion of TMED9 does not prevent wild-type GPI-APs from leaving the ER and trafficking to the plasma membrane

As a control to test for general effects of TMED9-depletion on ER-export, we used YFP-PrP WT. YFP-PrP WT folds efficiently and traffics via the secretory pathway from the ER to the plasma membrane where it stably resides and is predominantly localized at steady-state [12]. Between control and siTMED9-treated YFP-PrP WT NRK cells, we observed no obvious difference in YFP-PrP WT localization in the post-ER secretory pathway compartments and the plasma membrane. To confirm whether YFP-PrP WT was exposed on the cell surface regardless of TMED9-expression, we incubated live control and siTMED9-treated, unpermeabilized cells with AF594-αGFP antibodies (Abs) for 10 min prior to fixation. The AF594-αGFP Abs would only have access to surface exposed YFP-PrP WT. Then, we fixed and stained the cells for TMED9 to detect siTMED9 knockdown efficiency (S3F–S3H Fig). Additionally, we used CTL parental NRK cells as a negative control for AF594-αGFP Ab-binding because they do not express YFP-PrP WT. CTL YFP-PrP WT NRK cells served as positive control for AF594-αGFP Ab-binding (S3F and S3G Fig). Both CTL parental NRK and CTL YFP-PrP WT NRK cells serve as positive controls for TMED9 expression. Of the 45 randomly selected siTMED9-treated YFP-PrP WT NRK cells that we analyzed for TMED9 expression, we found that 10 of them demonstrated a perinuclear Golgi-pattern for TMED9 staining, indicating no knockdown, while the remaining 35 were depleted for TMED9 (S3F and S3H Fig). Importantly, regardless of siTMED9-treatment and TMED9-expression levels, the AF594-αGFP antibodies bound to the YFP-PrP WT NRK cells at levels above the CTL parental NRK cells (S3F–S3H Fig). This experiment demonstrated that YFP-PrP WT is able to leave the ER and traffic to the cell surface independently of TMED9 expression.

As an alternate approach to analyze YFP-PrP WT localization in the presence or absence of TMED9, we fixed the cells and co-stained them for TMED9 expression and endogenous calnexin to mark the ER, which fills the cytoplasmic space. We counted the number of cells in which YFP-PrP WT localized to the plasma membrane, which surrounds the cytoplasm. In 100% of the TMED9-expressing cells (n  =  28), YFP-PrP WT visibly localized to the plasma membrane (Fig 3J). Similarly, in 100% of the TMED9-depleted cells (n  =  26), YFP-PrP WT visibly localized to the plasma membrane (Fig 3J), demonstrating that TMED9 is not required for ER-export of the wild-type variant. Further support for the idea that TMED9 is not required for ER-export of wild-type GPI-APs was provided by Pearson’s colocalization analysis of YFP-PrP WT with calnexin, which revealed similar r values of 0.325 ±  0.056 for n  =  22 control scrambled siRNA-treated (CTL) cells and 0.404 ±  0.100 for n  =  24 siTMED9-treated cells (Fig 3K). The slight increase in r value between YFP-PrP WT and calnexin in siTMED9-treated cells over CTL cells may indicate a delay in ER-export (discussed below). Critically, we show here that YFP-PrP WT localization does not require TMED9 for ER-export to post-ER compartments including the plasma membrane (Fig 3J and 3K). This is in marked contrast to the misfolded variant, YFP-PrP * , which relies on TMED9 expression for ER-export (Fig 3E–3I).

Taken together, the results from the siTMED9 knockdown experiments demonstrate that TMED9-expression is required for the ER-export of misfolded GPI-APs. On the other hand, wild-type GPI-APs traffic to the plasma membrane regardless of TMED9-expression or depletion. This indicates that the general mechanisms underlying ER-export and post-ER trafficking remained intact in the absence of TMED9 and that TMED9 is a key component of the machinery that is specifically required for the ER-export of misfolded GPI-APS.

The data underlying the graphs shown in Fig 3 are included in the S1 Data file.

Chemical inhibitor of TMED9, BRD4780, inhibits RESET of misfolded GPI-APs

The siTMED9 depletion studies strongly suggest that TMED9 expression is required for RESET. However, TMED9-depletion by knock down or knock out was reported to be associated with a decrease in the levels of other p24/TMED-family proteins, including TMP21 and TMED2 [34,35]. We replicated this phenomenon here with siTMED9-treated YFP-PrP * NRK cells (Fig 4A and 4B). As expected, siTMED9-depletion reduced TMED9 expression levels by 91.34% ± 1.54% (Stdev.p, n  =  3). However, siTMED9-depletion also resulted in a 62.18% ± 22.95% (Stdev.p, n  =  3) reduction in TMP21 and 48.26% ± 16.89% (Stdev.p, n  =  3) reduction in TMED2 levels. The concomitant reduction of known RESET factors, TMP21 and TMED2, exposes the possibility that TMED9’s involvement in RESET pathway may be indirect through the stabilization of TMP21 or TMED2. Therefore, to gain insight into whether TMED9 plays a direct role in RESET, we opted to use the recently published chemical inhibitor of TMED9, BRD4780, that has been shown to acutely and specifically destabilize TMED9 [22].

To determine the optimal parameters for acute TMED9 inhibition by BRD4780, we used live cell imaging assays to identify the shortest pre-treatment time and lowest concentration for BRD4780 that would inhibit TG-induced RESET in 100% of the YFP-PrP * NRK cells. To mark the ER, we co-transfected our YFP-PrP * NRK cells with the previously characterized ER-marker mCherry-Sec61β [36,37]. We found that a 30 min pre-treatment of 100 µ M BRD4780 was sufficient to completely block TG-induced RESET in 100% (n  =  34) of the cells. Instead of relocalizing to downstream secretory pathway compartments, YFP-PrP * remained in the ER of the YFP-PrP * NRK cells for at least 90 min after the addition of TG in BRD4780-pretreated cells (Fig 4C). Pearson’s colocalization analysis of YFP-PrP * with the ER marker, mCherry-Sec61β, revealed strong colocalization at steady-state that was reduced upon TG-treatment in control cells but maintained in BRD4780-pretreated cells (Fig 4D). Similar experiments conducted with GFP-CD59 (C94S) NRK cells revealed that in 100% of the cells that were treated with TG alone, the ER-localized population of GFP-CD59 (C94S) underwent RESET (n  =  33), while in BRD4780-pretreated cells, the ER-population of GFP-CD59 (C94S) remained trapped in the ER in 100% of the cells (n  =  32) (S4A Fig). As with YFP-PrP * (Fig 4D), colocalization analysis of GFP-CD59 (C94S) with mCherry-Sec61β supported the observation that GFP-CD59 (C94S) underwent TG-induced ER-export in the control cells but remained in the ER of the BRD4780-pretreated cells (S4B Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. BRD4780 induces TMED9-modification and inhibits RESET of misfolded GPI-APs.

(A) Representative western blots of either non-targeting siRNA (CTL) or TMED9 siRNA (siTMED9)-treated YFP-PrP * NRK cells. Membranes were probed for TMP21 and GAPDH or TMED2 and GAPDH. (n = 3 biological replicates). (B) Bar graph representing the mean band intensity for TMED9, TMP21, or TMED2 from three independently performed experiments as shown in (A). For each condition, TMED9, TMP21, or TMED2 band intensities were double normalized: first against the band intensity of GAPDH, and second against the CTL band intensity. Error bars represent the standard deviation of the mean. Symbols were coded for each independently performed experiment. Statistics were calculated from unpaired t test with Welch’s correction with *  indicated p < 0.05. (C) Time-lapse images of YFP-PrP * NRK cells that were co-transfected with mCh-Sec61β. Cells were either not pretreated (Control) or pretreated with BRD4780 (BRD pretreat) for 30 min. Image collection was started immediately after the addition of TG. Scale bars represent 10 µm. (D) Plot of the average Pearson’s r values between YFP-PrP * and mCh-Sec61β. For Control versus BRD4780-pretreated conditions, as described in (C), 10 time-lapses of individual cells were analyzed for the 0 and 30 min time points. Pearson’s colocalization coefficients, r, were measured between YFP-PrP * and mCh-Sec61β within the boundaries of the cell. For each data point, the boundaries of each cell at 0 and 30 min time points were revealed by temporarily maximizing the gain for YFP-PrP * . (E) Representative western blots of YFP-PrP * NRK cells that were untreated (CTL), TG-treated for 90 min (TG), or BRD4780 (30 min)-pretreated and harvested after 90 min of TG treatment (BRD+TG) (n = 3 biological replicates). Blots were probed for TMED9, PrP (YFP-PrP*), and GAPDH. (F) Average of normalized YFP-PrP * band intensity from three independently performed experiments as shown in (E). For each condition, YFP-PrP * band intensities were double normalized. First, they were normalized against the band intensity of GAPDH. Second, they were normalized against the CTL band intensity. Error bars represent standard deviations of the mean. Statistics were calculated from unpaired t test with Welch’s correction with *** indicated p < 0.001. (G) Representative western blots of YFP-PrP * NRK cells that were either untreated control (CTL) or treated with BRD4780 for 30 min (30′ BRD). Blots were probed for TMED9, TMP21, and TMED2 and their respective GAPDH (n = 3 biological replicates). TMED9 protein profile shows two distinct bands, one lower band (orange arrow, LMW), and a higher band (black arrow, HMW). (H) Average of normalized High Molecular Weight (HMW) and Low Molecular Weight (LMW) TMED9 band intensity from 3 biological replicates as shown in (G). For each condition, TMED9 band intensities were double normalized. First, they were normalized against the band intensity of GAPDH. Second, they were normalized against the CTL intensity of each protein probed. Orange bars represent the quantification of the LMW band intensity and gray bars represent the HMW band intensity. Error bars represent standard deviations of the mean. Symbols were coded for each independently performed experiment. Statistics were calculated from unpaired t test with Welch’s correction with *  indicated p < 0.05, and ***p < 0.001. (I) Representative western blots of endogenous TMED9 from YFP-PrP * NRK cells after 180 min of BRD-treatment (n = 3 biological replicates). Blots were probed for TMED9 and GAPDH. TMED9 protein profile shows two distinct bands, one LMW (orange arrow), and a HMW (black arrow). (J) Average of normalized HMW and LMW TMED9 band intensity from three biological replicates as shown in (I). For each time point, TMED9 band intensities were double normalized: first by the band intensity of GAPDH and second against the BRD4780 0 min band intensity. Orange bars representing the measurement for the LMW of TMED9 and gray, the HMW. Error bars represent standard deviations of the mean. Symbols were coded for each independently performed experiment. Statistics were calculated from unpaired t test with Welch’s correction with *  indicated p < 0.05 and **p < 0.01. (K) Representative western blots of TMP21, TMED2, and their representatives GAPDH either control (CTL) or BRD4780 (30 min)-treated YFP-PrP * NRK cells (n = 3 biological replicates). (L) Average of normalized TMP21 and TMED2 band intensities from three biological replicates as shown in (K). For each condition, TMP21 or TMED2 band intensities were double normalized: first against the band intensity of GAPDH and second against the CTL band intensity. Error bar represents standard deviation of the mean. Symbols were coded for each independently performed experiment. Statistics were calculated from unpaired t test with Welch’s correction with *  indicated p < 0.05.

https://doi.org/10.1371/journal.pbio.3003084.g004

Although BRD4780 pre-treatment appeared to completely inhibit RESET, it did not prevent the loss of YFP-PrP * . After BRD4780 pre-treatment, addition of TG induced a gradual fading of the YFP-fluorescence over the 90 min time course, without any obvious change in the localization (Fig 4C). This suggested that the YFP-PrP * was degraded from the ER by an alternate mechanism. Western blot analysis produced results that were consistent with the live-cell imaging. While TG alone induced degradation of YFP-PrP * within 90 min, the 30-min BRD4780-pretreatment partially inhibited this TG-induced clearance (Fig 4E and 4F). Altogether, our results demonstrate that although misfolded GPI-APs appear to get degraded by an alternate pathway, BRD4780 completely blocks the TG-induced clearance of YFP-PrP * via the RESET pathway.

Next, we sought to test whether the BRD4780-block on RESET was due to degradation of TMED9 in NRK cells. In untreated and TG-treated cells, our western blot results for TMED9 revealed a doublet of ~23.5 and 24 KDa just under the 25 KDa molecular weight marker (Fig 4E). For the cells pre-treated for BRD4780 for 30 min prior to the addition of TG for 90 min (BRD +  TG), we observed that the lower molecular weight (LMW) band of the doublet disappeared (Fig 4E). To gain insight into the fate of the LMW band of TMED9, we treated cells with BRD4780 alone for 30 min and observed that the overall expression level of TMED9 did not change (Fig 4G and 4H). Instead, there was a clear mobility shift from the ~23.5 KDa LMW band to the ~24 KDa higher molecular weight (HMW) band (Fig 4G and 4H). This mobility shift in TMED9 within 30 min of BRD4780-treatment suggested that the 30 min BRD4780-pretreatment induced block on RESET shown in Fig 4C was not due to degradation of TMED9. Instead, the 30-min BRD4780-treatment appeared to acutely change the physical state of TMED9.

To gain more insight into BRD4780’s effect on TMED9, we performed a BRD4780-treatment time course and probed for TMED9 by western blot. We observed the gradual loss of the LMW band from 0 to 30 min. After 30 min of BRD4780-treatment, the LMW band decreased by 75.70 ± 1.24 (Stdev.p, n  =  3) (Fig 4I and 4J), which corresponds with a complete inhibition of RESET (Fig 4C). By 60 min, there was a complete loss of the LMW band, followed by an eventual loss of the HMW band over 180 min of BRD4780-treatment. Although we cannot rule out changes to TMP21 or TMED2 that are undetectable by western blot, we show that within 30 min of BRD4780-treatment when the LMW TMED9 band disappeared and RESET was blocked, TMP21 and TMED2 demonstrated no significant change in expression levels by western blot (Fig 4K and 4L). Thus, BRD4780 provides a useful tool to inhibit TMED9’s function as a RESET factor while leaving the previously established RESET factors, TMP21 and TMED2, intact.

The results from these live-cell imaging and western blot-based BRD4780 experiments are consistent with the idea that TMED9 plays an essential role in the clearance of YFP-PrP * via RESET. The rapid 30-min BRD4780-induced blockade of RESET, BRD4780-induced molecular weight shift of TMED9, and unaltered TMP21 and TMED2 bands spotlight TMED9 as RESET factor that plays a unique role beyond simply stabilizing the essential RESET factors, TMP21 and TMED2.

The data underlying the graphs shown in Fig 4 are included in the S1 Data file.

BRD4780-treatment alters TMED9 trafficking and glycosylation

The BRD4780-treatment experiments uncovered a connection between the loss of the LMW band of TMED9 and the loss of the cell’s ability to carry out RESET (Fig 4C and 4G–4J). Intriguingly, the LMW band disappeared within 30 min of adding BRD4780, while the HMW band persisted for up to 180 min. We considered two explanations for this observation. First, the LMW band may be relatively unstable in comparison to the HMW band and thus degrade more quickly upon BRD4780-treatment. Second, the LMW band may be converted to the HMW band as a step in TMED9’s path toward its eventual degradation. A major clue was presented in Fig 4I. Between 0 and 15 min of BRD4780-treatment, there was an evident and reproducible shift between the balance of the LMW and HMW TMED9 bands that comprised the total pool of TMED9 (Fig 4I). At time 0, total TMED9 was made up of 64.83% ± 3.92% (Stdev.p, n  =  3) LMW and 34.17% ± 3.92% (Stdev.p, n  =  3) HMW pools. However, within 15 min of BRD4780-treatment, the total TMED9 pool was comprised of 28.60% ± 1.13% (Stdev.p, n  =  3) in the LMW and 71.40% ± 1.13% (Stdev.p, n  =  3) in the HMW band. Given that the total levels of TMED9 remain steady over the first 15 min, this flip in the relative distribution of the TMED9 LMW and HMW populations negates the idea that the HMW band simply persists longer than the LMW band. Instead, this flip is consistent with the idea that the LMW population converts to the HMW population as TMED9 traffics along the BRD4780-induced degradation pathway for near complete degradation by 180 min (Fig 4I and 4J).

To delineate the BRD4780-induced degradation pathway for TMED9, we started by assessing which of the two major cellular degradatory systems, lysosomes and proteasomes, are involved. Western blot analysis revealed that in cells treated with BRD4780 for 180 min, there was a reduction in total TMED9 levels (Fig 5A and 5B, compare lanes 1 and 2). Pre-treatment with the lysosomal degradation inhibitor, bafilomycin A1 (BafA1 +  BRD4780), or with the proteasomal degradation inhibitor, MG132 (MG132 +  BRD4780), revealed that bafilomycin A1 inhibited BRD4780-induced degradation of TMED9 HMW species (Fig 5A and 5B, compare lanes 2 and 3), while MG132 did not (Fig 5A and 5B, compare lanes 2 and 3). This suggests that BRD4780 induces lysosomal degradation of TMED9. We next assessed which of the two major categories of trafficking pathways between ER and lysosomes are involved: (1) the LC3-II-dependent ER-to-lysosome associated degradation pathways, including autophagy, that are inhibited by 3-methyladenine (3MA) [38] or (2) the ER-export to Golgi-dependent pathways that are inhibited with brefeldin A (BFA) [39–41]. We found that while 3-methyladenine did not affect BRD4780-induced degradation (Fig 5A and 5B, compare lanes 2 and 5), brefeldin A strongly inhibited BRD4780-induced degradation of both the LMW and HMW bands (Fig 5A and 5B, compare lanes 2 and 6). Brefeldin A causes the entire Golgi to fuse into the ER within 10 min [42]. Therefore, stabilization of both LMW and HMW bands by co-incubating cells with brefeldin A +  BRD4780 indicates that the entire population of TMED9 must traffic through the Golgi prior to degradation. Note that for this drug panel bafilomycin A1 (BAF), MG132 and brefeldin A (BFA) stocks were each prepared in DMSO and diluted 1:4,000, 1:2,000, and 1:1,000, respectively, into the medium. BRD4780 and 3-MA stocks were each prepared in water or directly included in the medium, as explained in the Materials and methods. However for the drug panel in Fig 5A and 5B, separate DMSO controls for the BAF+BRD, MG132 + BRD, or BFA+BRD treatments were not included to simplify the experiment without changing our interpretation of the results.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. BRD4780-treatment alters TMED9 trafficking and glycosylation.

(A) Representative western blots of YFP-PrP * NRK cells that were treated as listed below (n = 3 biological replicates). Untreated, treated with BRD4780 for 180 min (BRD), pretreated with bafilomycin A1 (BAF) for 180 min and co-treated with BRD for 180 min (BAF+BRD), pretreated with MG132 for 180 min and co-treated with BRD for 180 min (MG132 + BRD), pretreated with 3-methyladenine (3MA) for 180 min and co-treated with BRD for 180 min (3MA+BRD), or pretreated with BFA for 180 min and co-treated with BRD for 180 min (BFA+BRD). Bafilomycin A1 (BAF), MG132, and brefeldin A (BFA) stocks were each prepared in DMSO and diluted 1:4,000, 1:2,000, and 1:1,000 into the medium. For this panel, separate DMSO controls for the BAF+BRD, MG132 + BRD, or BFA+BRD treatments were not included. Blots are probed for endogenous TMED9 and GAPDH. (B) Bar graph representing the mean band intensity for TMED9 from three biological replicates as shown in (A). For each condition, TMED9 band intensities were double normalized first by the band intensity of GAPDH and second within individual experiments using control band intensity. Orange bars represent the measurement for the low molecular weight (LMW) band of TMED9 and gray bars represent the measurement of the high molecular weight (HMW) band of TMED9. Error bars represent standard deviations of the mean. Symbols were coded for individual experiments. Statistics were calculated from unpaired t test with Welch’s correction with *  indicated p < 0.05 and **p < 0.01. (C) Western blots of digested protein lysates of stably transfected YFP-PrP * NRK cells either in control condition (Control) or after a pre-treatment of 2 h of bafilomycin followed by co-incubation with 2 h of BRD4780 (BAF+BRD). Protein lysates undigested (water) or digested with Endoglycosidase H (Endo H), Neuraminidase + O-Glycosidase (NA + O-Glyc.), Peptide-N-glycosidase F (PNG), or NA + O-Glyc+PNGase (NA + O-G + PNG) for 3 h at 37^oC (n = 1 experiment). Blots were probed for TMED9 and calnexin (CNX). (D) Western blot band graphics obtained by first analyzing the plot lanes and second generating the line graphs of pre-selected ROIs from the western blot for control samples (left) or BAF+BRD samples (right). Black arrow represents the HMW form, orange arrow represents the LMW and purple, the unglycosylated form of TMED9. (E) (left) Fluorescence image of a typical untreated YFP-PrP * NRK cell that was co-transfected with CER-TMED9 and ER-marker, mCherry-Sec61β. Scale bar represents 10 µm. (right) Magnified panel of the stroked region on the right. Scale bar represents 1 µm. (F) (left) Fluorescence image of a typical untreated YFP-PrP * NRK cell that was co-transfected with CER-TMED9 and Golgi marker, FusionRed-SiT. Scale bar represents 10 µm. (right) Magnified panel of the stroked region on the right. Scale bar represents 1 µm. (G) Time-lapse imaging sequence of an NRK cell that was co-transfected with CER-TMED9 and ER marker, mCherry-Sec61β. Image-collection was started immediately after the addition of 100 µM BRD4780 treatment. Scale bar is 10 µm. (H) Example of an ER mask created from the ER-marker, mCherry-Sec61β. The ER mask was used to measure the CER-TMED9 fluorescence intensity within the ER at 0 or 30 min time points after the addition of BRD4780 from 9 individual time-lapses as shown in (G) and quantified in (I). (I) Bar graph representing the mean of ER-localized CER-TMED9 fluorescence-intensity measurements of 9 individual time-lapses of CER-TMED9 and mCherry-Sec61β-expressing cells, as shown in (G). The ER mask was generated based on and as analyzed in (H). Statistics were calculated from unpaired t test with Welch’s correction with ** indicated p < 0.01. (J) Representative immunofluorescence images of YFP-PrP* NRK cells that were transfected with CER-TMED9 under control untreated (CTL) or 100 µM BRD4780 (30 min)-treated (BRD) conditions and stained for GM130. Scale bar is 20 µm. (K) Bar graph representing the mean of Pearson’s correlation coefficient r to measure colocalization between CER-TMED9 and GM130 under untreated (CTL) and BRD4780 (30 min)-treated (BRD) conditions, as exemplified in (J). Colocalization for 18 cells was measured for control conditions and 11 cells were measured for BRD4780 (30 min)-treated cells. The r values between CER-TMED9 and GM130 for 0 and 30 min time points were 0.589 ± 0.099 and 0.739 ± 0.086, respectively. Statistics were calculated from unpaired t test Welch’s correction with *** indicating p < 0.001. .001.

https://doi.org/10.1371/journal.pbio.3003084.g005

A possible explanation for the BRD4780-induced shift in TMED9’s molecular weight is that it reflects the maturation of glycosylation as TMED9 moves from the ER and ERGIC through the Golgi. In eukaryotic cells, the organization of glycosyltransferases across the ER and Golgi allows for predictable sequential glycosylation reactions to occur as secretory pathway proteins traffic through these compartments. Thus, glycosidase digests are a standard procedure to monitor glycoprotein movement through the secretory pathway [43,44]. Upon entry into the ER, most secretory pathway proteins are N-linked glycosylated [45]. PNGase F cleaves all N-linked glycosylations. As glycoproteins move from the cis to the medial cisternae, they encounter Golgi-resident glycosyltransferases that modify the existing N-linked glycosylations to Endo H resistant forms [43,44]. As they move through medial and trans cisternae, they may encounter O-linked glycosyltransferases and sialyltransferases, which conjugate sialic acids to either N- or O-linked glycosylations [43,46,47]. With this in mind, we hypothesized that the steady-state pool of LMW and HMW TMED9 bands may be a reflection of different glycosylation-states of TMED9 whose ratios are affected by BRD4780-induced of TMED9 ER-to-Golgi transport. We therefore took an open approach to test our hypothesis with a panel of glycosidase digests.

PNGase F and Endo H digests of the YFP-PrP * NRK lysates confirmed that at steady-state, TMED9 is N-linked glycosylated with Endo H resistant and sensitive populations (Figs 5C, 5D, S5A, and S5B). The Endo H resistant populations aligned with the LMW and HMW bands, and the Endo H sensitive population aligned with the PNGase F digested TMED9 core protein. The neuraminidase +  O-glycosidase digest appeared to collapse the HMW and LMW doublet into single band with a similar mobility to the LMW band, indicating that the HMW band corresponded with the subset of TMED9 that accessed the trans cisternae. After BRD4780-treatment for 2 h in cells that were pretreated with bafilomycin to prevent TMED9 degradation, the entire TMED9 population ran as a single HMW band that was entirely Endo H resistant and neuraminidase +  O-glycosidase sensitive (Fig 5C and 5D). The following two results suggest that TMED9 does not get O-linked glycosylated, but instead its N-linked glycosylation gets sialylated: (1) PNGase F and PNGase F +  neuraminidase +  O-glycosidase digests each produced a single band for TMED9 that ran at the exact same molecular weight as the core TMED9 protein (Fig 5C and 5D) and (2) neuraminidase and neuraminidase +  O-glycosidase each produced a single band for TMED9 that ran at the exact same molecular weight as LMW TMED9 in untreated and BRD4780 (30 min)-treated cells (S5A and S5B Fig). Our results match previously published reports that at steady-state TMED9 obtains an N-linked glycosylation that can be modified with Endo H resistance and neuraminidase sensitivity [17,48,49] but does not get O-linked glycosylated [17]. Taken together, these glycosidase digests demonstrate that TMED9 localization shifts from the ER to the Golgi upon BRD4780-treatment in YFP-PrP * NRK cells.

To test whether BRD4780-treatment induces a shift in localization of TMED9 from the ER to the Golgi, we performed live-cell imaging in YFP-PrP * NRK cells transfected with CER-TMED9. This allowed us to image and measure CER-TMED9 distribution before and after BRD4780-treatment in the same cell. We co-transfected the cells with either the ER marker, mCherry-Sec61β, or the Golgi marker, FusRed-SiT. At steady-state, the CER-TMED9 appeared to colocalize with the ER marker, mCherry-Sec61β (Fig 5E), and the Golgi marker, FusRed-SiT (Fig 5F). Upon BRD4780-treatment, the ER-localized population of CER-TMED9 appeared to move out of the ER and through the Golgi (Fig 5G–5K). In YFP-PrP * NRK cells that were co-transfected with CER-TMED9 and mCherry-Sec61β, CER-TMED9 visibly moved out of the mCherry-Sec61β marked ER (Fig 5G). For the 0 and 30 min timepoints of these time-courses, we quantified the decrease of cumulative CER-TMED9 fluorescence intensity in the ER by establishing an ER mask by segmentation of the mCherry-Sec61β image (Fig 5H), and measuring the cumulative, background-subtracted pixel intensity of CER-TMED9 within the ER (Fig 5I). BRD4780 (30 min)-treatment induced a 55.17% ± 14.26% (Stdev.p, n  =  5) decrease in the ER pool of CER-TMED9 (Fig 5I). The decrease in the ER-pool of TMED9 appeared to correlate with an increase in the perinuclear pool of TMED9 that formed the typical pattern of the Golgi (Fig 5G). To verify a shift in TMED9-localization from the ER to the Golgi within 30 min of BRD4780-treatment, we stained control or BRD4780-treated CER-TMED9-expressing cells for the Golgi-marker, GM130, and imaged them by immunofluorescence. GM130 is a cytosolic Golgi-matrix protein that is directly targeted to the Golgi membrane [50]. BRD4780-treated cells demonstrate a visible increase in CER-TMED9 in the Golgi, and an increase in the Pearson’s colocalization coefficient r between CER-TMED9 and GM130-stained compared with control (Fig 5J and 5K).

The results in Fig 5 suggest that BRD4780-treatment changes TMED9’s distribution within the secretory pathway. The glycosidase digests panel for endogenous TMED9 and live-cell imaging results using exogenously expressed CER-TMED9 each represent TMED9 populations that span across the ER, Golgi and intermediate compartments at steady-state, but shift from the ER toward the Golgi and downstream compartments within 30 min of BRD4780-treatment. Altogether, the above results demonstrate that RESET inhibition by BRD4780 is associated with alterations in TMED9 trafficking and glycosylation.

The data underlying the graphs shown in Fig 5 are included in the S1 Data file.

TMED9’s glycosylation state does not impact its function in RESET

The BRD4780 (30 min)-induced inhibition of RESET coincided with conversion of TMED9 from the LMW form to the fully Endo H resistant, sialylated HMW form and relocalization of CER-TMED9 from the ER to the Golgi (Fig 5). These observations could be explained by two possible scenarios: (1) Only the LMW population of TMED9 is able to carry out RESET or (2) TMED9 must be localized to the ER and/or immediately ER-accessible compartments to carry out RESET. If the first scenario were correct, then TMED9 that was converted to HMW would not be able to support RESET. If the second scenario were correct, then TMED9 that was relocated into compartments downstream of the cis-Golgi would not be able to carry out RESET.

To test the possible explanations, we performed a series of identical incubations in two separate but identical dishes of YFP-PrP * NRK cells. One dish culminated in the collection of the lysates to examine the LMW/HMW states of TMED9. The other dish was assessed by imaging to test for the cells’ ability to carry out TG-induced RESET of YFP-PrP * based on the relocalization of the YFP-PrP * signal from ER (cytoplasm-filling reticular pattern excluding the nucleus) to Golgi (perinuclear crescents). First, we replicated our previous finding shown in Fig 3G. In untreated cells, the majority of TMED9 was in the LMW state and 100% of the cells were able to undergo RESET (Fig 6A, lane 1 and Fig 6B, panel 1). Upon a BRD4780 (30 min)-treatment, TMED9 converted to the HMW population and all of the cells lost their ability to undergo RESET (Fig 6A, compare lanes 1 and 2, and Fig 6B, panels 1 and 2). However, when we introduced a 30 min wash-out step immediately after the BRD4780 (30 min)-treatment, TMED9 remained in the HMW form but the majority of the cells (72%, n  =  50) regained competency to carry out RESET (Fig 6A, lane 3 and Fig 6B, panel 3). Thus, acute inhibition of RESET with BRD4780-treatment is reversible, despite the glycosylation state of TMED9. As shown previously (Fig 3I and 3J), long-term treatment with BRD4780 for 180 min induced near total TMED9-degradation (Fig 6A, lane 4) and blocked RESET (Fig 6B, panel 4) and was not reversible upon BRD4780 wash-out (Fig 6A, lane 5, Fig 6B, panel 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. TMED9’s function in RESET is not related to its glycosylation state.

(ATable presenting the different treatments and their effects on TMED9’s molecular weight. As depicted in the treatment time-course, YFP-PrP * NRK cells were pre-treated with BRD4780 (BRD) or cycloheximide +  BRD4780 (CHX BRD) for different periods of time with or without a 30 min BRD4780 wash-out (wash). “CHX wash” indicates that after pre-treatment with CHX +  BRD, the BRD4780 wash-out included cycloheximide. The numbers below the treatment time-course diagram indicate the percentage of cells that were competent for RESET after each treatment course. The percentages of RESET competent cells were obtained by dividing the number of cells in which YFP-PrP * re-localized from the ER to the Golgi within 30 min of adding TG by the total number of cells counted. The n values (i.e., total number of cells counted per treatment course) were included directly under the percentage. After the pre-treatments, “ER” versus “Golgi” localization was assessed based on the pattern of YFP fluorescence as observed through live-cell confocal imaging. Cells were counted as RESET competent if the YFP-PrP * fluorescence pattern shifted from an ER pattern to a Golgi pattern. At steady-state, YFP-PrP * fluorescence is dispersed throughout the cell in a reticulo-tubular pattern with a sharp outline of nucleus that is consistent with ER and the nuclear envelope subdomain of the ER. During RESET, YFP-PrP * fluorescence changes from disperse to a perinuclear localization that is consistent with the Golgi. Aligned directly below the treatment courses and percentages are the TMED9 and GAPDH western blots. Each lane encompassing the treatment, % RESET competent cells, TMED9 molecular weight, and GAPDH loading control, was numbered on the bottom. There were 10 individual treatment conditions. (B) Representative live-cell images of YFP-PrP * NRK cells after each indicated treatment conditions (1–10). The brightness and contrast settings were increased in panels 4, 5, 9, and 10 with respect to the other panels in order to more easily assess ER versus Golgi localization patterns. The nuclei were stained with Hoechst.

https://doi.org/10.1371/journal.pbio.3003084.g006

To eliminate the possibility that newly synthesized TMED9 could be contributing to the pool of TMED9 involved in RESET, we performed the above experiments in the presence of the protein translation inhibitor, cycloheximide (CHX). Again, in the presence of CHX, BRD4780 (30 min)-treatment induced the conversion of TMED9 to the HMW population and blocked RESET (Fig 6A, compare lanes 6 and 7, Fig 6B, compare panels 6 and 7). Following the CHX +  BRD4780 (30 min)-co-treatment, BRD4780 wash-out using medium containing only CHX restored the cell’s ability to carry out RESET in the majority of the cells (67%, n  =  99) but did not alter the HMW state of TMED9 (Fig 6A, lane 8, Fig 6B, panel 8). The finding that simply washing out BRD4780 restored RESET in the presence of a protein translation inhibitor disconnects the glycosylation state of TMED9 from its ability to support RESET. Finally, near complete depletion of the existing pool of TMED9 with CHX +  BRD4780 (180 min) co-treatment did not allow for the restoral of RESET with BRD4780 wash-out (Fig 6A, lanes 9 and 10, Fig 6B, panels 9 and 10).

Collectively, these experiments reiterate the requirement for TMED9’s presence to carry out RESET. These experiments demonstrate that TMED9’s glycosylation state does not impact its function in RESET.

TMED9 must have access to misfolded GPI-APs in the ER to coordinate RESET

To gain insight into whether TMED9’s localization was involved in its ability to execute RESET, we again used exogenously expressed CER-TMED9 as a proxy for TMED9. We co-transfected NRK cells with YFP-PrP * , CER-TMED9, and mCherry-Sec61β. We co-incubated these cells with cycloheximide (CHX) +  BRD4780 and then performed BRD4780 wash-out using medium with only CHX (Fig 7A). Live cell imaging revealed that CHX +  BRD4780 (30 min) induced export of CER-TMED9 out of the ER, as demonstrated earlier. However during the CHX-only wash-out step, CER-TMED9 appeared to move back into the ER and the cells regained their ability to carry out RESET (Fig 7A, S3 Video).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. TMED9 must have access to misfolded GPI-APs in the ER to coordinate RESET.

(A) Still frames of a typical YFP-PrP * NRK cell that was co-transfected with CER-TMED9 and mCherry-Sec61β and treated as follows. The first image is of the cell before treatment began (Untreated). The second image is of the cell after 30 min with 50 µg/mL of CHX +  BRD4780 (CHX +  BRD4780 30 min). The third image is of the cell 30 min after washing with CHX-only medium (CHX wash 30 min). The fourth image is of the cell 30 min after treatment with 50 µg/mL of CHX +  0.1 µM TG (CHX +  TG 30 min). Juxtaposed with each image is an area zoom of the region within the stroked frame. Scale bar represents 10 µm in the larger field of view and in the magnified panel of the stroked region. These select images are also provided in S2 Video. (B) Representative western blots of GFP-tag co-immunoprecipitations (co-IPs) from the parental untransfected NRK cells (P) or YFP-PrP * NRK cells (n = 3 biological replicates). Quantified in (C). Cells were harvested at the indicated time points after BRD4780-treatment. Blots were probed for PrP to detect YFP-PrP *  or for endogenous calnexin (CNX) and TMED9. Under each western blot is depicted a “Stain-Free” image of the total protein in the gel. (C) Bar graph displaying the normalized mean band intensities for the CNX and TMED9 that co-eluted with YFP-PrP * from three biological replicates as shown in (B). For each time point, CNX and TMED9 eluate band intensities were double normalized. First, they were normalized by the band intensity of the eluted YFP-PrP * band. Second, they were normalized by the time 0 eluate band intensity for each protein probed. Error bars represent standard deviations of the mean. Symbols were coded for individual experiments. Statistics were calculated from unpaired t test with Welch’s correction with and ** indicated p < 0.01. (D) Bar graph displaying the normalized mean band intensities for the CNX and TMED9 that co-eluted with YFP-PrP * from three biological replicates as shown in (B). This graph represents the co-eluted CNX and TMED9 normalized against the amount of CNX and TMED9 in the inputs. Thus, for each time point, CNX and TMED9 eluate band intensities were triple normalized. First, they were normalized against the band intensities in the input. Second, they were normalized by the band intensity of the eluted YFP-PrP * band. Third, they were normalized by the time 0 eluate band intensity for each protein probed. Error bars represent standard deviations of the mean. Symbols were coded for individual experiments. Statistics were calculated from unpaired t test with Welch’s correction with and ** indicated p < 0.01. (E) Western blots of GFP-tag co-immunoprecipitations (co-IPs) from the parental untransfected NRK cells (P) or YFP-PrP * NRK cells (n = 1 experiment). Cells were harvested after the indicated treatments. Blots were probed for PrP to detect YFP-PrP * , or for endogenous calnexin (CNX) and TMED9. Under each western blot is depicted a “Stain-Free” image of the total protein in the gel.

https://doi.org/10.1371/journal.pbio.3003084.g007

Based on our combined western blot and live-cell imaging results for siRNA depletion or BRD4780-inhibition and depletion of TMED9 (Figs 3, 4, 6, 7A, S2, and S3), we concluded that TMED9 expression was required for RESET, regardless of whether TMED9 was in the LMW or HMW form. The acute inhibition of RESET by BRD4780-treatment was concomitant with the relocalization of TMED9 out of the ER, while restoration of RESET was concomitant with the return of CER-TMED9 to the ER upon BRD4780 wash-out (Figs 4, 6, and 7A). Importantly, we observed that YFP-PrP * did not co-traffic with CER-TMED9 out of the ER upon BRD4780-treatment (Figs 7A and S6). Instead, YFP-PrP * appeared to be left behind.

The above observations are consistent with the idea that in order to coordinate RESET, TMED9 requires access to the ER where RESET substrates are largely localized in association with calnexin (Figs S1 and 2, [12,14]). To directly test this idea, we performed a co-immunoprecipitation time course, pulling down YFP-PrP * from untreated or BRD4780-treated YFP-PrP * NRK cells. In untreated cells, YFP-PrP * co-purified with calnexin and TMED9 (Fig 7B, t = 0). However, upon BRD4780-treatment, we observed the rapid dissociation of TMED9 from YFP-PrP * between 5 and 30 min (Fig 7B). We quantified the results by two different methods (Fig 7C and 7D). In Fig 7C, we double normalized mean band intensities for the calnexin and TMED9 that co-eluted with YFP-PrP * from three biological replicates, as shown in Fig 7B, first by the band intensity of the eluted YFP-PrP * band and second by the time 0 eluate band intensity for each protein probed. This provided a measurement of the change in how much calnexin or TMED9 co-immunoprecipitated with YFP-PrP * over the BRD4780-treatment time course. In Fig 7D, we triple normalized the same data by additionally taking into account the amount of starting material in the inputs. First, the eluate bands were normalized against the band intensities in the input. Second, they were normalized by the band intensity of the eluted YFP-PrP * band. Third, they were normalized by the time 0 eluate band intensity for each protein probed. Regardless of the quantification scheme, the data revealed that upon BRD4780-treatment, TMED9 nearly completely dissociated from YFP-PrP * (Fig 7C and 7D). Additionally, TMP21 and TMED2 appeared to dissociate from YFP-PrP * within the same time frame (S7 Fig). Intriguingly, YFP-PrP * steadily maintained association with calnexin, even while the p24-family members rapidly disengaged (Fig 7B–7D), which is consistent with the imaging results showing that YFP-PrP * remained in the ER upon BRD4780-treatment (Figs 7A and S6).

Finally, to test our hypothesis that BRD4780 wash-out restores RESET by allowing TMED9 to return to the ER where it can interact with the RESET substrate, we performed co-immunoprecipitations of YFP-PrP * with anti-GFP antibody conjugated beads under the BRD4780-treatment and wash-out conditions (Fig 7D). We prevented the involvement of newly synthesized TMED9 by including cycloheximide (CHX) in all of the incubations. As a negative control for the co-immunoprecipitation, we included the parental cells with CHX. We treated YFP-PrP * NRK cells as follows: (1) CHX for 30 min, (2) CHX +  BRD4780 for 30 min, and (3) CHX +  BRD4780 for 30 min followed by a 30 min wash-out in medium containing only CHX. Again, we observed the TMED9 band shift from LMW to HMW upon the CHX +  BRD4780 (30 min)-treatment in the input lanes. In the eluate lanes, the CHX (30 min) treatment allowed YFP-PrP * to pull down both calnexin and TMED9, while the CHX +  BRD4780 (30 min)-treatment allowed YFP-PrP * to pull down calnexin but not TMED9. Excitingly, in the eluate lane for CHX +  BRD4780 (30 min) followed by the CHX-only wash-out, pull-down of YFP-PrP * co-eluted calnexin and TMED9. After the wash-out, the TMED9 that co-eluted with YFP-PrP * ran at the higher molecular weight, indicating that it had trafficked to downstream compartments of the Golgi before returning to the ER where it could associate with YFP-PrP * .

Taken together, our results demonstrate that TMED9’s location in the ER is essential for it to carry out RESET. We showed that TMED9 relocalizes from the ER to the Golgi upon BRD4780 (30 min)-treatment. Upon relocalization out of the ER, TMED9 loses contact with the RESET substrate and its ability to conduct RESET. However if TMED9 is restored to the ER upon BRD4780 wash-out, then TMED9 is able to associate with the RESET substrate and perform its essential function in the RESET pathway. These results presented here are consistent with a model in which TMED9, in conjunction with other p24-family members, is a key constituent of the essential ER-export machinery that captures misfolded GPI-APs released by calnexin and conveys them from the ER to the Golgi (S8 Fig).

The data underlying the graphs shown in Fig 7 are included in the S1 Data file.

Cell lines

The isolation of the clonal line of NRK cells that we refer to here as parental NRK cells here was described previously [74]. YFP-PrP * NRK, YFP-PrP WT NRK, and GFP-CD59 (C94S) NRK cells are each derived from the NRK cells and their creation were described previously [12,14]. In short, each of the cell lines were created by transiently transfecting a 10-cm dish with expression plasmids, expanding the transfected cells and selecting them for the stable integrants by using antibiotic resistance, and sorting them for by their fluorescence intensity. As described previously, YFP-PrP * NRK cells were sorted for expression of YFP-PrP * at similar levels to endogenous PrP in mouse brain lysate and YFP-PrP WT NRK and GFP-CD59 (C94S) NRK each express YFP-PrP WT and GFP-CD59 at similar levels to YFP-PrP * [12,14].

Plasmids

The construction of YFP-PrP WT, YFP-PrP *, and GFP-CD59 (C94S) was described in detail [12]. Briefly, YFP-PrP WT includes the N-terminal prolactin signal sequence, which drives efficient translocation into the ER [25], and YFP-PrP * is derived from YFP-PrP WT but contains a C179A mutation that disrupts the single disulfide bond in PrP. GFP-CD59 (C94S) includes the N-terminal rabbit lactase phlorizin-hydrolase “lactase” signal sequence, which drives efficient translocation into the ER [75–79].

Cerulean (CER)-TMED9 includes the N-terminal rabbit lactase phlorizin-hydrolase signal sequence and its construction was described in detail [24].

Golgi markers include CER-GalT (gift from Jennifer Lippincott-Schwartz, Addgene; plasmid# 11930 [29]) and FusionRed-SiT-15 (gift from Michael W. Davidson, Addgene; plasmid# 56133 [80]).

ER markers include mCherry (mCh)-Sec61β (gift from Jennifer Lippincott-Schwartz, Addgene; plasmid# 90994 [37]) and Cerulean (CER)-calnexin, which we constructed for this study. For CER-calnexin, we used the same vector backbone as the SS(lactase)-CER-TMED9 described previously [24]. However, instead of the mature domain of TMED9, we PCR amplified the mature domain of calnexin (human) from a plasmid template obtained from DNASU (DNASU, HsCD00042644). “SS(lactase)” refers to rabbit lactase-phlorizin hydrolase signal sequence. “Mature domain” refers to the ORF that encodes the processed type I transmembrane protein after the N-terminal signal sequence has been cleaved by the signal peptidase. For assembly we used NEBuilder (NEB E5520S). The final ORF was verified to be the following through sequencing analysis with SS(lactase) in all capital letters, followed by linker and CER in lowercase and by calnexin mature domain in capital letters:

ATG GAG CTC TTT TGG AGT ATA GTC TTT ACT GTC CTC CTG AGT TTC TCC TGC CGG GGG TCA GAC TGG GAA TCT CTG CAG TCG ACG GTA CCG CGG GCC CGG GAT CCA CCG GTC GCC ACC atg gtg agc aag ggc gag gag ctg ttc acc ggg gtg gtg ccc atc ctg gtc gag ctg gac ggc gac gta aac ggc cac aag ttc agc gtg tcc ggc gag ggc gag ggc gat gcc acc tac ggc aag ctg acc ctg aag ttc atc tgc acc acc ggc aag ctg ccc gtg ccc tgg ccc acc ctc gtg acc acc ctg acc tgg ggc gtg cag tgc ttc gcc cgc tac ccc gac cac atg aag cag cac gac ttc ttc aag tcc gcc atg ccc gaa ggc tac gtc cag gag cgc acc atc ttc ttc aag gac gac ggc aac tac aag acc cgc gcc gag gtg aag ttc gag ggc gac acc ctg gtg aac cgc atc gag ctg aag ggc atc gac ttc aag gag gac ggc aac atc ctg ggg cac aag ctg gag tac aac gcc atc agc gac aac gtc tat atc acc gcc gac aag cag aag aac ggc atc aag gcc aac ttc aag atc cgc cac aac atc gag gac ggc agc gtg cag ctc gcc gac cac tac cag cag aac acc ccc atc ggc gac ggc ccc gtg ctg ctg ccc gac aac cac tac ctg agc acc cag tcc aag ctg agc aaa gac ccc aac gag aag cgc gat cac atg gtc ctg ctg gag ttc gtg acc gcc gcc ggg atc act ctc ggc atg gac gag ctg tac aag gca gga ggc agc CAT GAT GGA CAT GAT GAT GAT GTG ATT GAT ATT GAG GAT GAC CTT GAC GAT GTC ATT GAA GAG GTA GAA GAC TCA AAA CCA GAT ACC ACT GCT CCT CCT TCA TCT CCC AAG GTT ACT TAC AAA GCT CCA GTT CCA ACA GGG GAA GTA TAT TTT GCT GAT TCT TTT GAC AGA GGA ACT CTG TCA GGG TGG ATT TTA TCC AAA GCC AAG AAA GAC GAT ACC GAT GAT GAA ATT GCC AAA TAT GAT GGA AAG TGG GAG GTA GAG GAA ATG AAG GAG TCA AAG CTT CCA GGT GAT AAA GGA CTT GTG TTG ATG TCT CGG GCC AAG CAT CAT GCC ATC TCT GCT AAA CTG AAC AAG CCC TTC CTG TTT GAC ACC AAG CCT CTC ATT GTT CAG TAT GAG GTT AAT TTC CAA AAT GGA ATA GAA TGT GGT GGT GCC TAT GTG AAA CTG CTT TCT AAA ACA CCA GAA CTC AAC CTG GAT CAG TTC CAT GAC AAG ACC CCT TAT ACG ATT ATG TTT GGT CCA GAT AAA TGT GGA GAG GAC TAT AAA CTG CAC TTC ATC TTC CGA CAC AAA AAC CCC AAA ACG GGT ATC TAT GAA GAA AAA CAT GCT AAG AGG CCA GAT GCA GAT CTG AAG ACC TAT TTT ACT GAT AAG AAA ACA CAT CTT TAC ACA CTA ATC TTG AAT CCA GAT AAT AGT TTT GAA ATA CTG GTT GAC CAA TCT GTG GTG AAT AGT GGA AAT CTG CTC AAT GAC ATG ACT CCT CCT GTA AAT CCT TCA CGT GAA ATT GAG GAC CCA GAA GAC CGG AAG CCC GAG GAT TGG GAT GAA AGA CCA AAA ATC CCA GAT CCA GAA GCT GTC AAG CCA GAT GAC TGG GAT GAA GAT GCC CCT GCT AAG ATT CCA GAT GAA GAG GCC ACA AAA CCC GAA GGC TGG TTA GAT GAT GAG CCT GAG TAC GTA CCT GAT CCA GAC GCA GAG AAA CCT GAG GAT TGG GAT GAA GAC ATG GAT GGA GAA TGG GAG GCT CCT CAG ATT GCC AAC CCT AGA TGT GAG TCA GCT CCT GGA TGT GGT GTC TGG CAG CGA CCT GTG ATT GAC AAC CCC AAT TAT AAA GGC AAA TGG AAG CCT CCT ATG ATT GAC AAT CCC AGT TAC CAG GGA ATC TGG AAA CCC AGG AAA ATA CCA AAT CCA GAT TTC TTT GAA GAT CTG GAA CCT TTC AGA ATG ACT CCT TTT AGT GCT ATT GGT TTG GAG CTG TGG TCC ATG ACC TCT GAC ATT TTT TTT GAC AAC TTT ATC ATT TGT GCT GAT CGA AGA ATA GTT GAT GAT TGG GCC AAT GAT GGA TGG GGC CTG AAG AAA GCT GCT GAT GGG GCT GCT GAG CCA GGC GTT GTG GGG CAG ATG ATC GAG GCA GCT GAA GAG CGC CCG TGG CTG TGG GTA GTC TAT ATT CTA ACT GTA GCC CTT CCT GTG TTC CTG GTT ATC CTC TTC TGC TGT TCT GGA AAG AAA CAG ACC AGT GGT ATG GAG TAT AAG AAA ACT GAT GCA CCT CAA CCG GAT GTG AAG GAA GAG GAA GAA GAG AAG GAA GAG GAA AAG GAC AAG GGA GAT GAG GAG GAG GAA GGA GAA GAG AAA CTT GAA GAG AAA CAG AAA AGT GAT GCT GAA GAA GAT GGT GGC ACT GTC AGT CAA GAG GAG GAA GAC AGA AAA CCT AAA GCA GAG GAG GAT GAA ATT TTG AAC AGA TCA CCA AGA AAC AGA AAG CCA CGA AGA GAG TAG.

Cell culture, transfection, siRNA knock down, and chemical inhibitors

Cell culture conditions included incubation at 37°C, 5% CO2, and 90% humidity in Dulbecco’s Modified Eagle Medium (DMEM, Corning, 17–205-CV) containing 10% fetal bovine serum (Corning 35–011-CV), 1×L-glutamine (Corning, 25–005-CL). Cells used in experiments were never allowed to grow beyond 80% confluency and regularly split every 2 or 3 d using 0.05% Trypsin/0.53 mM EDTA in HBSS (Corning, 25–051-CL). Cells were typically cultured on tissue culture treated plastic (SCBT, sc-251460) and seeded onto #1.5 coverslip glass-bottom dishes (Cell Vis, D35C4-20-1.5-N) for imaging.

Transfections of YFP-PrP * NRK cells were performed at 70% confluency using PolyJet In Vitro DNA Transfection Reagent (SignaGen Laboratories, SL100688). For all experiments involving transient transfections, we allowed time for cells that were overexpressing the plasmid-encoded protein to die. We split transiently transfected cells 24 h after transfection onto #1.5 glass-bottom coverslips (Cell Vis, D35C4-20-1.5-N) for imaging experiments. Imaging experiments were performed 48–72 h after transfection.

siRNA knockdown of TMED9 was carried out by two sequential transfections 48 h apart of the ON-TARGETplus SMARTpool Rat TMED9 siRNA (Dharmacon, L-092937-01-0005) with Dharmafect 2 transfection reagent (Dharmacon, T-2002-02) exactly as recommended by the manufacturer. For imaging experiments, immediately after the second siRNA transfection, cells were transferred to #1.5 coverslips (Cell Vis, D35C4-20-1.5-N), and 24 h later, the imaging experiments were performed. For western blots, immediately after the second siRNA transfection, cells were transferred to 6- or 12-well tissue culture-treated dishes (Genesee, 25–105MP or 25–106MP).

Chemical inhibitors were added to cell culture in complete DMEM at the following working concentrations: 0.1 µM TG (diluted from 10-mM stock prepared in DMSO; Calbiochem, 586006), 100 µM BRD4780 (also called AGN 192403 hydrochloride; diluted from 50-mM stock prepared in water; Tocris, 1072), 10 µM of MG132 (diluted from 20-mM stock prepared in DMSO; Calbiochem, 474790), 250 nM bafilomycin A1 (diluted from 1-mM stock prepared in DMSO; SCBT, SC-201550B), 10 mM 3-methyladenine (3MA, diluted directly into cell culture medium at 10 mM; Tocris, 3977), 2.5 µM brefeldin A (diluted from 25-mM stock prepared in DMSO; LC Laboratories, B-8500), and 50 µg/mL cycloheximide (diluted from 10-mg/mL stock prepared in water; Calbiochem, 23-97-64). As a control for either TG or bafilomycin-treatments, the equivalent concentrations of DMSO were added to cells. For BRD4780, 3MA, and CHX, since they were originally diluted in water, control conditions are simply the cell culture medium without the drug. The single exception to this was in the drug panel presented in Fig 5A, where as explicitly stated in the Figure Legend, DMSO controls were not included.

Co-immunoprecipitation

Cells were cultured in 10-cm dishes to ~70% confluency and treated as described. The entire procedure prior to elution was performed in a 4°C cold room and on ice to stabilize interactions. First, cells were washed twice with 5 mL of cold Low Salt Buffer (50 mM HEPES, pH7.4, 100 mM NaCl, 2 mM MgCl₂). Cells were lysed with 1 mL of IP buffer (1% CHAPS, 50 mM HEPES, pH 7.4, 100 mM NaCl, and 2 mM CaCl2) and transferred to a microfuge tube. The lysate was pipetted up and down 20 times through a 200-µL pipet tip to disrupt the cell membranes. To clear the lysates of debris that may clog the column, lysates were subjected to two rounds of centrifugation for 5 min at 16,000 g at 4°C, and supernatants were transferred to new 1.5-mL microfuge tube after each centrifugation. Ninety microliter of the cleared lysate were set aside to load as “inputs” and the remaining 900 μL were incubated on ice for 15 min with 60 μL of super-paramagnetic eGFP µ MACS MicroBead from the μMACS GFP Isolation Kit (Miltenyi, 130-091-288). In parallel, µ Columns (Miltenyi, 130-042-701) were placed in the µ Macs Separator (Miltenyi, 130-042-602) and equilibrated with 200 µ L of IP buffer. At the end of the incubation period, the bead/lysate mixtures were loaded into the columns. Columns were then rinsed five times with 200 μL of IP buffer and washed once with 20 mM Tris Buffer, pH 7.5. After the washes, the experiment was transferred to room temperature and elution was performed by adding 40 μL of boiling elution buffer made of 2× Laemmli Sample Buffer (Bio-Rad, 1610737) and 20 mM TCEP (tris(2-carboxyethyl)phosphine), Sigma, 646547). The input samples were denatured with 35 μL of 4× sample buffer (Bio-Rad, 1610747) with 40 mM TCEP. The input and eluate samples were further analyzed by western blot.

Western blots and antibodies used for western blots

After treatments, cells were washed with cold 1× phosphate-buffered saline (PBS), lysed in a small volume of cold radioimmunoprecipitation assay buffer (RIPA) buffer (50 mM Tris HCl, pH 7.6, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% SDS). Protein lysates were centrifuged at 4°C for 10 min at 15,000 rpm on a microcentrifuge. Supernatants were stored at −20°C and processed as follows. Protein lysate concentrations were determined with Pierce Micro BCA Protein Assay kit (ThermoFisher, 23235). Equal total quantities of protein were denatured using Laemmli sample buffer supplemented with tris(2-carboxyethyl)phosphine (TCEP) and boiled for 5 min before loading into a 12% or 4%–20% acrylamide pre-cast SDS-PAGE gels (Bio-Rad, 4561045) for electrophoresis. Where indicated, we ran samples on Bio-Rad 4%–20% Mini-PROTEAN TGX Stain-Free Protein Gels (Bio-Rad, 4568094) to obtain images of total protein inputted into the gel using Bio-Rad’s Stain-Free Imaging system on the ChemiDoc Touch Imaging System (Bio-Rad, 17001401).

Proteins were transferred to a PVDF (polyvinylidene difluoride, Bio-Rad, 1704275) membrane using a Trans-Blot Turbo Transfer System (Bio-Rad, 1704150). Membranes were blocked with 5% non-fat milk in Tris-buffered saline with 0.1% Tween-20 (TBS-T +  5% Milk) for 1 h at room temperature prior to probing with primary antibodies.

The following primary antibodies were added to the TBS-T +  2.5% milk solution at the following concentrations: mouse monoclonal αGFP (Proteintech, 66002–1-IG) at 1:2,000, rabbit polyclonal αTMED9 (Proteintech, 21620–1-AP) at 1:2,500, rabbit polyclonal αTMP21 (homemade and described previously [12]) at 1:2,000, rabbit polyclonal αTMED2 (Proteintech, 11981–1-AP) at 1:2,000, mouse monoclonal αCalnexin (Proteintech, 66903–1) at 1:2,500, rabbit polyclonal αPrP (Proteintech, 12555–1-AP) at 1:2,500, and mouse monoclonal αGAPDH antibody (ThermoFisher, MA5–15738) at 1:4,000, for 2 h at room temperature or overnight at 4°C.

Membranes were washed three times with TBS-T and incubated in TBS-T +  5% milk for 1 h at room temperature with either of the following secondary antibodies at 1:5,000: HRP-conjugated Donkey anti-Rabbit (Sigma Aldrich, NA934−1ML) or HRP-conjugated Goat anti-Mouse (ThermoFisher, A28177). Membranes were washed three times and briefly incubated in Pierce ECL Western Blotting Substrate (ThermoFisher, 32209) or SuperSignal West Femto Maximum Sensitivity Substrate (ThermoFisher, 34096) prior to imaging the chemiluminescence on the ChemiDoc Touch Imaging System (Bio-Rad, 17001401).

FIJI (ImageJ 1.52p, NIH USA) was used to measure relative pixel densities of bands on unsaturated images collected as eight-bit.tif files from the Chemi-Doc (Bio-Rad). A rectangular ROI sized to enclose the band with the largest area was created and the same ROI was used to measure the integrated density of the rest of the bands. Integrated density values were background subtracted and normalized against “time 0,” “untreated,” or “control” bands to obtain relative pixel densities. Relative pixel density values from three independently performed experiments were presented as mean ±  SD (n =  3), as indicated in the figure legends. Means and SD were calculated and plotted using Microsoft Excel. Quantification details for specific experiments are provided in the adjoining figure legends.

Glycosidase digests

Cell lysates were divided equally into multiple samples that were treated individually with water (no digested control) or with Endoglycosidase H (Endo H, New England Biolabs, P0702), with a2-3,6,8,9 Neuraminidase A (neuraminidase, New England Biolabs, P0722), with a combination of neuraminidase and O-Glycosidase (NA and O-G, New England Biolabs, P0733), with Peptide-N-glycosidase F (PNGase F, New England Biolabs, P0704) or with a combination of O-Glyc+PNGase (O-G +  PNGase F) for 3 h at 37°C according to the manufacturer’s instructions.

To draw the band density histograms associated with the glycosidase digests, FIJI (ImageJ 1.52p, NIH USA) was used. Since the Endo H digest in the untreated lysate produced bands spanned the greatest distance from highest molecular weight band to lowest molecular weight band, a rectangular ROI sized to enclose the Endo H bands was created. This same ROI was dragged from lane to lane and the plot graph was generated by selecting the “Analyze,” “Gel,” and “Plot lanes” menu items in Image J and graphs representing the densitometry intensity versus pixel distance were generated by selecting the “Analyze,” “Tools” menu items in Image J to run the Analyze Line Graph.

Live-cell imaging, immunofluorescence and antibodies, nuclear stain, and the microscope

Cells were seeded on a #1.5 glass bottom dish (Cell Vis, D35-20-1.5-N or D35C4-20-1.5-N) 24–48 h prior to the imaging experiment. Cells were imaged at 60%–80% confluency. Treatment conditions (e.g., untreated, BRD4780, or TG-treatment) were specified in the figure legend.

For live-cell experiments and time-lapse imaging, cells were imaged in their normal DMEM medium inside of a pre-equilibrated stage top incubator chamber adjusted to 37°C, 5% CO2, and 90% humidity. To stain nuclei for live cell imaging, cells were incubated for 10 min with NucBlue Live ReadyProbes Reagent Hoechst 33342 (Thermofisher, R37605).

For immunofluorescence against calnexin (CNX), TMED9, and GM130, cells were washed once with PBS and then fixed and permeabilized with methanol/acetone (3:1) solution for 10 min at −20°C. For immunofluorescence against cell surface GFP without additional staining, cells were washed once with PBS and then fixed with 4% paraformaldehyde (PFA) in PBS for 15 min at room temperature, but not permeabilized. After the fixation step, cells were washed three times with PBS and blocked for 30 min at room temperature with UltraCruz blocking reagent (SantaCruz; SC-516214) to stain GFP, calnexin, or GM130, or 5% normal goat serum (ThermoFisher, 31872) diluted into PBS to stain TMED9. Primary antibodies were diluted 1:200 in either UltraCruz blocking reagent or 2.5% normal goat serum in PBS buffer and added to the cells for 2 h at room temperature. The following primary antibodies were used: mouse monoclonal αGFP (Proteintech, 66002–1-IG), rabbit polyclonal αTMED9 (Novus, NBP1–31650, lot# 40121), mouse monoclonal αCalnexin (Proteintech, 66903–1), or rabbit polyclonal αGM130 (Novus Bio, NBP2–53420). Cells were washed three times with PBS. Secondary antibodies, Goat anti-Rabbit IgG, Alexa Fluor 660 (ThermoFisher Scientific, A21074), or Goat anti-Mouse IgG, Alexa Fluor 568 (ThermoFisher Scientific, A1103), were diluted at 1:2,500 in either UltraCruz blocking reagent or 2.5% normal goat serum in PBS buffer and added to the cells for 1 h at room temperature. Cells were washed three times with PBS and incubated with fresh PBS with NucBlue nuclear stain (ThermoFisher, 33342) diluted at 1:20. Images were captured using the Nikon spinning disk (described below).

For immunofluorescence against cell surface GFP followed by immunofluorescence of TMED9, live unpermeabilized cells were first incubated with Alexa Fluor 594 (AF594)-conjugated Rabbit αGFP antibody (Thermofisher, A-21312) for 10 min. The AF594-αGFP antibody was diluted at 1:200 directly into the cell culture medium. Next, the cells were fixed and permeabilized with methanol/acetone (3:1) solution for 10 min at −20°C. Cells were blocked with 5% normal goat serum in PBS (Thermofisher, 31872) and stained with mouse monoclonal αTMED9 (Novus, NBP3–33028) primary antibody at 1:500 followed by Goat anti-Mouse IgG, Alexa Fluor 660 (ThermoFisher Scientific, A-21055) secondary antibody diluted at 1:2,000.

Images were acquired using a Nikon inverted spinning disk confocal microscope equipped with a Yokogawa CSU-X1 Spinning Disk, EMCCD camera, 60× Plan Apo 1.40 NA oil/ 0.13-mm WD, Tokai live-cell incubator, Perfect Focus system to prevent focus drift, lasers (405, 445, 488, 514, 561, and 647 nm), emission filter sets (455/50, 480/40, 525/36, 540/30, 605/70, and 700/75), and dichroic mirror sets (405/488/561/640 and 440/514/561). This system is set up to prevent bleed-through between the following sets of fluorophores or fluorescent proteins: Cerulean, YFP, FusionRed/mCherry/AF568/AF594 and AF647/AF660, or Hoechst/DAPI, YFP, FusionRed/mCherry/AF568/AF594 and AF647/AF660, or Hoechst/DAPI, GFP, and FusionRed/mCherry/AF568/AF594 and AF647/AF660. The Nikon microscope and image acquisition was controlled by the NIS-Elements software.

Antibody uptake

Cells were seeded in 4-well #1.5 glass bottom dish (Cell Vis D35C4-20-1.5-N) 24 h prior to the live-cell imaging or immunofluorescence experiment. The following were added to the cell culture medium: a 1:5,000 dilution of Alexa Fluor 647-conjugated Rabbit anti-GFP antibodies (Thermofisher, A-31852), 0.1 µ M TG (Calbiochem, 586006) to induce RESET, and 250 nM balifomycin A1 (SCBT, SC-201550B) to inhibit lysosomal degradation. Cells were imaged live at 37°C, 5% CO2, and 90% humidity for 90 min or incubated in incubator for 90 min prior to fixation. Cells were fixed in methanol/acetone (3:1) solution for 10 min at −20°C and stained for TMED9 using a mouse monoclonal anti-TMED9 (Novus NBP3–33028).

Image analysis

Pearson colocalization coefficients r were obtained by the following steps. In cells co-expressing or stained for two non-cross-talking fluorescent proteins or dyes, multichannel images were captured and cropped tightly around a single cell. The cell edge was identified by temporarily increasing the contrast of the image and ROIs were manually drawn around the edges of the cell and saved using ROI manager. Within the ROI for each set of 2 channels, Pearson’s correlation coefficient value r was measured using the Coloc2 plugin by selecting Analyze, Colocalization, and Coloc 2 commands in FIJI (ImageJ 1.52p, NIH USA).

To measure the fluorescence intensity of ER-localized CER-TMED9 within an imaging generate ER mask, cells were co-transfected with mCh-Sec61β to mark the ER. Multichannel images were collected before and after treatment in RFP (for mCh-Sec61β) and CFP (for CER-TMED9) channels and cropped tightly around a single cell. For each cell, the cell edge was identified by temporarily increasing the contrast of the image and ROIs were manually drawn around the edges of the cell and saved using ROI manager. ER masks were created from Sec61-mCherry images using Fiji (NIH Image J, 1.52p) by setting the threshold to cover a maximum of the ER compartment by selecting the following Image, Adjust from the menu, and selecting the Threshold command. The threshold setting was used to create and save the ER-mask by selecting Edit, Selection from the menu and selecting the Create Mask command. ER-localized CER-TMED9 fluorescence intensity was extrapolated using the ER mask by selecting Process from the menu and the Image Calculator command with Image 1 set as “ER mark,” operation set as “AND” and Image 2 set as “CER-TMED9.” The ER-localized CER-TMED9 fluorescence intensity was then measured using Measurement command.

Statistics

Statistical analyses were performed using Prism 9 (GraphPad). Statistical significance was calculated from unpaired t test with Welch’s correction on the mean value of at least three independent experiments. The p values that were equal or less than 0.05 were annotated with one star (*); equal or less than 0.01 were annotated with two stars (**); equal or less than 0.001 were annotated with three stars (***); and equal or less than 0.0001 were annotated with four stars (****).