The role of the AP-1 adaptor complex in outgoing and incoming membrane traffic

Abstract

The AP-1 adaptor complex is found in all eukaryotes, but it has been implicated in different pathways in different organisms. To look directly at AP-1 function, we generated stably transduced HeLa cells coexpressing tagged AP-1 and various tagged membrane proteins. Live cell imaging showed that AP-1 is recruited onto tubular carriers trafficking from the Golgi apparatus to the plasma membrane, as well as onto transferrin-containing early/recycling endosomes. Analysis of single AP-1 vesicles showed that they are a heterogeneous population, which starts to sequester cargo 30 min after exit from the ER. Vesicle capture showed that AP-1 vesicles contain transmembrane proteins found at the TGN and early/recycling endosomes, as well as lysosomal hydrolases, but very little of the anterograde adaptor GGA2. Together, our results support a model in which AP-1 retrieves proteins from post-Golgi compartments back to the TGN, analogous to COPI's role in the early secretory pathway. We propose that this is the function of AP-1 in all eukaryotes.

Figure 1.

Two models for AP-1 function. (A) In Model 1, AP-1 acts together with GGAs at or near the TGN, packaging a number of proteins into clathrin-coated vesicles for trafficking to peripheral endosomes. These proteins include hydrolases and their receptors, which are sorted by GGAs, with AP-1 playing an auxiliary role. AP-1 has a further role in retrieving proteins from peripheral endosomes, including empty hydrolase receptors and TGN-resident membrane proteins such as syntaxins 6, 10, and 16. Adapted from Hirst et al. (2012). (B) In Model 2, AP-1 and GGAs act independently. GGAs sort hydrolases and their receptors at or near the TGN for trafficking to endosomes, while AP-1 acts exclusively at endosomes to retrieve proteins, including empty receptors, back to the TGN. One or more of these proteins may be needed by GGAs for forward transport.

Figure 2.

System for investigating outgoing and incoming membrane traffic. (A) Schematic diagram of the system. A HeLa cell line was generated in which the endogenous AP-1 γ-adaptin subunit, encoded by AP1G1, was replaced by a mRuby2-tagged version. The cells were then stably transduced with a “Hook” construct, consisting of streptavidin with a KDEL sequence for ER retention, and one or more candidate cargo constructs, containing streptavidin-binding peptide (SBP) to keep the constructs in the ER until the addition of biotin. Most of these constructs were also tagged with GFP so that their appearance at the plasma membrane and subsequent endocytosis could be monitored with an anti-GFP nanobody. The green dotted lines with arrows show the pathways that might be taken by the SBP-GFP-cargo proteins after the addition of biotin, and the blue dotted lines with arrows show the pathways that might be taken by the nanobody after it has bound to SBP-GFP-cargo proteins on the cell surface. (B) Western blot of cells shown schematically in A, in which the SBP- and GFP-tagged cargo protein was the cation-dependent mannose 6-phosphate receptor (CDMPR). Addition of biotin causes a shift in the mobility of SBP-GFP-CDMPR (visualized with anti-GFP) as it moves through the Golgi apparatus. HaloTag-conjugated nanobody was added to all the cells for the final 30 min and it started to show a clear increase above background levels (which is most likely due to fluid-phase uptake) after 40 min in biotin. The AP-1 γ subunit is expressed at similar levels in the wild-type and transduced cells, but it has different mobilities (asterisks) because of the presence or absence of the mRuby2 tag. Clathrin heavy chain (CHC17) was used as a loading control. (C) Fluorescent images of the cells shown in B. After 15 min in biotin, most of the SBP-GFP-CDMPR has moved from the ER to the Golgi region, and it stays mostly Golgi-localized until after 40 min. Endocytosed nanobody starts to be detectable at 40 min. Scale bar: 10 μm. Source data are available for this figure: SourceData F2.

Figure S1.

Generation and characterization of cell lines and cell populations. (A) Initially, we attempted to tag the AP-1 γ subunit with mRuby2 using gene editing. Positive cells were sorted by flow cytometry and then individual clonal cell lines were isolated and analyzed by Western blotting. The blots show that although the tagging was successful, expression levels for the tagged protein (upper band) were always much lower than that of the wild-type protein in the parental HeLa cells. This is most likely because the gene we tagged, AP1G1, is present in multiple copies, and only very few were tagged in each cell line, while the others were frequently disrupted by indels. (B) After generating a clonal cell line in which the tagged AP-1 γ subunit was introduced by retroviral transduction and the endogenous gene was then deleted using gene editing, we added additional tagged membrane proteins using retroviral transduction. Cells were selected by flow cytometry and then routinely sorted by flow cytometry before scaling up for CCV isolation experiments. The dot plots show that there was some loss of mRuby2-tagged AP-1 γ with time and that although the three membrane proteins were under the control of the same LTR promoter and were sorted using the same gates, there were inherent differences in expression levels, with SBP-GFP-NAGPA being most strongly expressed, followed by SBP-GFP-CDMPR and then SBP-GFP-LAMP1. (C) Frames from Video 4, showing cells co-expressing SBP-HaloTag-NAGPA (red) and SBP-GFP-LAMP1 (green), treated with biotin for 28 min. Some of the LAMP1-containing tubules are decorated with AP-1, indicating that the presence of AP-1 on these tubules is not simply a consequence of overexpressing the AP-1 cargo protein NAGPA. Scale bar: 10 μm. Source data are available for this figure: SourceData FS1.

Figure 3.

Comparison of SBP-GFP-tagged CDMPR, NAGPA, and LAMP1. (A) Like SBP-GFP-CDMPR, both SBP-GFP-NAGPA and SBP-GFP-LAMP1 are retained in the ER in the absence of biotin, and there is no detectable uptake of nanobody. After 90 min in biotin, both have reached their normal steady state localization and have endocytosed substantial amounts of nanobody, which was added for the final 30 min. Scale bar: 10 μm. (B) Cells expressing SBP-GFP-CDMPR, SBP-GFP-NAGPA, or SBP-GFP-LAMP1 were transfected with HaloTag-gadkin and then treated with biotin for 90 min. Both SBP-GFP-CDMPR and SBP-GFP-NAGPA show strong colocalization with gadkin at the cell periphery, while SBP-GFP-LAMP1 does not. This colocalization with gadkin is indicative of an association with AP-1. Scale bar: 10 μm.

Figure 4.

Live imaging of cells coexpressing mRuby2-tagged AP-1 and one or more tagged membrane proteins. (A) Frames from Video 1, showing cells co-expressing SBP-HaloTag-NAGPA (red) and SBP-GFP-LAMP1 (green), treated with biotin for 20 min. LAMP1 and NAGPA can be seen leaving the Golgi in the same tubules (arrowheads). The number of seconds between frames is indicated. Scale bar: 10 μm. (B) Frames from Video 2, showing cells coexpressing mRuby2-AP-1 (red) and SBP-GFP-NAGPA (green), treated with biotin for 33.5 min. NAGPA-positive tubules are decorated with AP-1 puncta. The number of seconds between frames is indicated. Scale bar: 10 μm. (C) Frames from Video 3 show cells coexpressing mRuby2-AP-1 (red), SBP-HaloTag-NAGPA (green), and SBP-GFP-CDMPR (blue), treated with biotin for 27 min. Although most of the CDMPR leaves the Golgi in vesicles rather than tubules, small amounts of CDMPR can sometimes be seen in NAGPA-positive tubules. These tubules are frequently decorated with AP-1 puncta (arrowheads). The number of seconds between frames is indicated. Scale bar: 10 μm. (D) A frame from Video 5, showing cells coexpressing mRuby2-AP-1 (red) and SBP-GFP-NAGPA (green), fed far-red transferrin (blue) for 1 h, and treated with biotin for 27 min. AP-1 puncta can be seen on both NAGPA-positive tubules (upper right arrowhead) and transferrin-positive tubules (other arrowheads). Transferrin-containing tubules at the cell periphery are frequently decorated with AP-1 puncta (see also Video 6). Scale bars: 10 μm for color images, 5 μm for black and white images.

Figure 5.

Single vesicle analysis of cells coexpressing mRuby2-tagged AP-1 and GFP-tagged membrane proteins. (A) Overview of the technique. A CCV-enriched fraction is prepared from stably transduced cells and aliquoted onto slides that were precoated with 100-nm beads. The beads fluoresce in four wavelengths and are used as fiducial markers for channel alignment. Widefield images are collected in four channels and analyzed using a newly written script. The analysis includes channel alignment, quantification of fluorescence in each spot, and elimination of the beads, which are the only particles that fluoresce in Channel 4. Data are plotted as the percentage of AP-1-containing spots (Channel 1) that also contain another protein or proteins. (B) A CCV-enriched fraction was prepared from cells co-expressing mRuby2-tagged AP-1 and SBP-GFP-tagged CDMPR, treated with biotin for 2 h, and incubated with HaloTag-conjugated nanobody for the final 30 min. Aliquots were spotted onto slides that had been precoated with beads. The figure shows cropped images in each wavelength, with the beads circled. The full image can be seen in Fig. S2. AP-1 fluoresces in red, CDMPR in green, and nanobody in far-red, shown as blue in the merged image. Scale bar: 5 μm. (C) Western blot from an experiment similar to the one in B, showing the whole cell homogenate and CCV-enriched fractions for four different time points in biotin. CHC17 was used as a loading control. (D) Single vesicle analysis of the experiment is shown in C. The data represent the means from three independent pooled experiments, with at least 10 images analyzed in each experiment for each condition, and 1,000–10,000 discrete spots per image. The error bars indicate the standard deviation. The presence of CDMPR in AP-1 vesicles continues to increase throughout the time course, so that by 2 h in biotin, about three-quarters of the AP-1 vesicles contain detectable CDMPR, and nearly half of these also contain detectable endocytosed nanobody. There are no vesicles containing nanobodies that do not also contain CDMPR, as expected because the nanobody enters the cell by piggybacking on the CDMPR. Complete datasets for all the single vesicle analysis experiments are shown in Fig. S3. (E) Western blot of cell homogenates and CCV-enriched fractions from cells expressing SBP-GFP-NAGPA, treated with biotin for varying lengths of time and incubated with nanobody for the final 30 min. CHC17 was used as a loading control. (F) Single-vesicle analysis of the experiment is shown in E. The data represent the means from three independent pooled experiments, with at least 10 images analyzed in each experiment for each condition, and 1,000–10,000 discrete spots per image. The error bars indicate the standard deviation. The presence of NAGPA in AP-1 vesicles plateaus at 40 min, with about one-third of the AP-1 vesicles containing detectable NAGPA. Nearly half of these also contain detectable endocytosed nanobody, which accumulates more quickly in NAGPA-expressing cells than in CDMPR-expressing cells, consistent with NAGPA leaving the Golgi in LAMP1-positive tubules (see Fig. 4 A). (G) Western blot of cell homogenates and CCV-enriched fractions from cells expressing SBP-GFP-LAMP1, treated with biotin for varying lengths of time and incubated with nanobody for the final 30 min. CHC17 was used as a loading control. (H) Single-vesicle analysis of the experiment shown in G. The data represent the means from three independent pooled experiments, with at least 10 images analyzed in each experiment for each condition, and 1,000–10,000 discrete spots per image. The error bars indicate the standard deviation. Relatively little LAMP1 accumulates in AP-1 vesicles. Source data are available for this figure: SourceData F5.

Figure S2.

Example of a widefield image used for single vesicle analysis. The region shown in Fig. 5 B is indicated. Scale bar: 10 μm.

Figure S3.

Results from all 18 of the experiments used for single vesicle analysis. Three biological repeats were carried out for each condition. All of the cells used in these experiments expressed mRuby2-tagged AP-1 (red) but their GFP-tagged and far-red proteins were different (shown in green and blue, respectively). The key in the upper left shows the different combinations of labels, based on how they appear in the images (e.g., R+G for red and green only, shown as yellow; R+FR for red and far-red only, shown as magenta, etc.). On the actual graphs, the right-hand “AP-1 + other label(s)” portions (R+G/yellow, R+FR/magenta, and R+G+FR/white) were used to generate Figs. 5 and 6. The middle portion of each graph (red) shows the percentage of vesicles that were positive for AP-1 only. The right-hand portions (G, FR, G+FR/cyan) show vesicles that were negative for AP-1 but positive for other fluorescent proteins. These were normalized to total AP-1-containing vesicles, and thus there are some instances where the percentage is >100. Although the AP-1-containing vesicles were very consistent between experiments, there was much more variability in non-AP-1 puncta. The error bars indicate the standard deviation.

Figure 6.

Heterogeneity of AP-1 vesicles. (A) Fluorescent images of cells coexpressing SBP-GFP-CDMPR and SBP-HaloTag-NAGPA after 90 min in biotin. Although both localize to juxtanuclear membranes, the fine details are different (inset), and only CDMPR localizes to peripheral membranes, where AP-1 is also found (arrowheads). Scale bars: 10 μm. (B) Western blot of cells co-expressing SBP-GFP-CDMPR and SBP-HaloTag-NAGPA, showing the whole cell homogenate and CCV-enriched fractions for four different time points in biotin. CHC17 was used as a loading control. (C) Single-vesicle analysis of the experiment shown in B. The data represent the means from three independent pooled experiments, with at least 10 images analyzed in each experiment for each condition, and 1,000–10,000 discrete spots per image. The error bars indicate the standard deviation. Most of the NAGPA-containing AP-1 vesicles (∼80%) also contain CDMPR, but not vice versa. The percentages are deduced from the white and magenta portions of the bars: in every case, the white portion is ∼4× taller than the magenta portion. (D) Single vesicle analysis of cells co-expressing SBP-GFP-CDMPR and SBP-HaloTag-CDMPR, showing the means from three independent pooled experiments, with at least 10 images analyzed in each experiment for each condition, and 1,000–10,000 discrete spots per image. The error bars indicate the standard deviation. The differences are most likely due to differences in expression levels because the localization patterns of SBP-GFP-CDMPR and SBP-HaloTag-CDMPR are virtually identical (see Fig. S4 B). (E) Single vesicle analysis of cells expressing SBP-GFP-CDMPR, treated with biotin for various lengths of time, and in every case fed far-red transferrin for 1 h and then processed immediately. The data represent the means from three independent pooled experiments, with at least 10 images analyzed in each experiment for each condition, and 1,000–10,000 discrete spots per image. The error bars indicate the standard deviation. The percentage of AP-1 vesicles that contain both CDMPR and transferrin increases with time in biotin. (F) Fluorescent image of the CCV-enriched fraction from cells expressing mRuby2-tagged AP-1 and labeled after fixation with an antibody against CHC17. Scale bar: 5 μm. (G) Single-vesicle analysis of experiments similar to the one shown in F, with at least 10 images analyzed in each of three experiments, and 1,000–10,000 discrete spots per image. The error bars indicate the standard deviation. Nearly all of the AP-1 vesicles are also positive for CHC17. Source data are available for this figure: SourceData F6.

Figure S4.

Further characterization of AP-1 cargo constructs. (A) Western blots comparing expression levels of GFP- and HaloTag-constructs. The cells were all incubated with biotin for 40 min. In every case, the constructs labeled with GFP are expressed at ∼threefold higher levels than the same constructs labeled with HaloTag. (B) Fluorescent images of cells coexpressing GFP-CDMPR and HaloTag-CDMPR. Although there is variability in relative expression levels when different cells are compared, the fine details are virtually identical (insets). Scale bars: 10 μm. Source data are available for this figure: SourceData FS4.

Figure 7.

Stills from Video 7 showing cells co-expressing mRuby2-tagged AP-1 and GFP-tagged GGA2. Arrowheads indicate some of the puncta that are positive for GGA2 only. The circled area indicates a moving structure that is positive for both AP-1 and GGA2, but the two localize to different regions. Scale bars: 2 μm.

Figure 8.

Capture of AP-1 vesicles using magnetic beads. (A) A CCV-enriched fraction was prepared from HeLa cells expressing mRuby2-tagged AP-1 γ, and then the AP-1-positive vesicles were captured just before the final centrifugation step using a rabbit antiserum against mRuby2 followed by protein A-coated magnetic beads. CCVs (arrowheads) can be seen to be associated with the beads; the two top ones are shown magnified in the inset. Scale bars: 100 nm. (B) Coomassie blue-stained gel of the whole cell homogenate, proteins captured on magnetic beads, and the CCV-enriched fraction from cells expressing mRuby2-tagged AP-1 γ. As a control for specificity, the preimmune serum from the same rabbit was also incubated with protein A beads. The samples containing beads and the CCV-enriched fraction were made up to the same volume so the efficiency of capture could be assessed, and the sample containing cell homogenate was made up to the same protein concentration as the sample containing the CCV fraction. H and L refer to immunoglobulin heavy and light chains; the bands labeled 1–4 were excised and found to correspond to CIMPR, CHC17, the AP-1 γ subunit, and the AP-1 β1 subunit, respectively. (C) Western blot of the samples shown in B probed with various antibodies. The dotted lines indicate the position on the gel for each antigen. In addition to the expected proteins, a small amount of AP-2 was also captured by the beads. GGA2 has a similar profile to AP-2, with relatively little captured by the beads. Source data are available for this figure: SourceData F8.

Figure S5.

Protein rank plots. The graphs show the mean rank (0 is lowest intensity) versus the mean log₁₀ intensity for each of the 632 proteins analyzed in Fig. 9. The L (light) channel is the input, i.e., the total CCV-enriched fraction, while the H (heavy) channel is the bead-captured proteins. CHC17 (CLTC) is the most abundant protein in both channels and ribosomal proteins (black) are also abundant in both channels. However, AP-1, as well as AP-1-associated proteins identified in our previous knocksideways study (red) (Hirst et al., 2012), shift up the rank distribution in the heavy channel. The L (light) channel (A) is the input, i.e., the total CCV-enriched fraction, while the H (heavy) channel (B) is the bead-captured proteins.

Figure 9.

Proteomic analysis of the bead-captured samples. CCV-enriched fractions were prepared from HeLa cells expressing mRuby2-tagged AP-1 γ, grown in either SILAC heavy (H) or SILAC light (L) medium. AP-1-positive vesicles were captured from the H fraction before the final centrifugation step, while the L fraction was centrifuged to provide an input reference. Heavy-to-light ratios were calculated as a measure of enrichment in the bead-captured samples over the input and statistical analysis was performed for 632 proteins that passed quality control. (A) Data are presented as a volcano plot. For each protein, a one-sample t test was performed against a hypothetical ratio defined as the mean ratio for CLTC, the most abundant protein in both the H and L samples (see Fig. S5). The proteins on the right side of the plot are potential components of AP-1 vesicles, with the most enriched hits being AP-1 subunits. The three hydrolase receptors and various hydrolases also score highly. Proteins on the left side are unlikely to be components of AP-1 vesicles, and they include ribosomal proteins. AP-2 subunits are generally to the left of clathrin, with the exception of AP2B1 (arrowhead), which is known to associate promiscuously with AP-1 (Page and Robinson, 1995). Other AP-1 cargo and accessory proteins are not indicated in this plot. (B) Protein identities in the boxed region in A, which contains the most enriched hits. The proteins shown in color were also hits in our previous knocksideways analyses (Hirst et al., 2012, 2015). The arrowheads indicate the syntaxins STX6, STX10, and STX16. The complete dataset is available in Table S1. Table S2 contains proteomic data from unlabeled cells, in which the CCV-enriched fraction was incubated with beads coated with either anti-mRuby or preimmune serum, as further confirmation of the specificity of the AP-1 vesicle capture.

Figure 10.

A third model for AP-1 function. This model is similar to Model 2, with AP-1 and GGAs acting independently. However, AP-1 vesicles bud from compartments (post-Golgi and early/recycling endosomes) where the pH is not low enough for hydrolases and their receptors to dissociate, and therefore the entire hydrolase–receptor complex is packaged into vesicles, together with TGN resident proteins. These vesicles return to the juxtanuclear TGN. GGAs sort hydrolases and their receptors at the juxtanuclear TGN for trafficking to a later endosome, with an intralumenal pH below 6, and empty receptors are recycled from this compartment by retromer and sorting nexins. In non-opisthokonts like plants, which do not have GGAs, we propose that the TGN-to-endosome pathway is mediated by AP-4. The juxtanuclear TGN, equivalent to the Golgi-associated TGN in plant cells, matures into the post-TGN compartment, which in plant cells would be called the Golgi-independent TGN.