The Intracellular Transport and Secretion of Calumenin-1/2 in Living Cells

Calumenin isoforms 1 and 2 (calu-1/2), encoded by the CALU gene, belong to the CREC protein family. Calu-1/2 proteins are secreted into the extracellular space, but the secretory process and regulatory mechanism are largely unknown. Here, using a time-lapse imaging system, we visualized the intracellular transport and secretory process of calu-1/2-EGFP after their translocation into the ER lumen. Interestingly, we observed that an abundance of calu-1/2-EGFP accumulated in cellular processes before being released into the extracellular space, while only part of calu-1/2-EGFP proteins were secreted directly after attaching to the cell periphery. Moreover, we found the secretion of calu-1/2-EGFP required microtubule integrity, and that calu-1/2-EGFP-containing vesicles were transported by the motor proteins Kif5b and cytoplasmic dynein. Finally, we determined the export signal of calu-1/2-EGFP (amino acid positions 20–46) and provided evidence that the asparagine at site 131 was indispensable for calu-1/2-EGFP stabilization. Taken together, we provide a detailed picture of the intracellular transport of calu-1/2-EGFP, which facilitates our understanding of the secretory mechanism of calu-1/2.

Extracellularly, calu-1 is reported to interact with the serum amyloid P component in the presence of Ca 2+ , indicating its possible role in amyloidosis [15]. Calu-1 may also be involved in autocrine and paracrine signaling since it decreases the expression level of septin 2 and actin fragments as well as regulating the cell cycle in fibroblasts [16]. Recently, thrombospondin-1, a secreted glycoprotein, is reported to form a complex with calu-1, suggesting that calu-1/2 play a potential role in hemostasis and thrombosis [2].
Despite the initial characterization of calu-1/2, many aspects are still elusive, including their intracellular transport and secretory process. Here, we used calu-1/2-EGFP, in which the EGFP would possibly mask the C-terminal HDEF retention signal, to study its secretion process. We reported the translocation of calu-1/2-EGFP into the lumen of the endoplasmic reticulum (ER), visualized their intracellular transport in the vesicles, and showed the secretion of calu-1/2-EGFP through either ''secretion after attachment'' or ''secretion after accumulation''. To investigate the underlying transport mechanism, we determined the roles of cytoskeleton network and motor proteins on the intracellular transport and secretion of calu-1/2-EGFP. Furthermore, we identified that Kif5b and cytoplasmic dynein were the motors that were responsible for their microtubule-dependent trafficking. Finally, mutational analysis revealed the export signal of calu-1/2-EGFP and amino acid point crucial for the calu-1/2-EGFP stability.

Calu-1/2 are translocated into the ER lumen and secreted
With our produced antibody, which recognize both calu-1 and calu-2 isoforms (data not shown), we found that a great deal of calu-1/2 were detected in the cultured medium ( Figure 1A). However, immunofluorescence assay showed that calu-1/2 localized throughout the cell in an ER-like manner rather than accumulated at the Golgi apparatus both in HeLa and HEK293T cells ( Figure S1). In order to specifically study the secretory process of calu-1/2, we used calu-1/2-EGFP, in which the C-terminal ERretaining signal HDEF was disrupted by the EGFP tag. A majority of calu-1/2-EGFP fusion proteins were detected in the cultured medium as expected ( Figure 1B). Meanwhile, calu-1/2-EGFP accumulated at the Golgi apparatus both in HeLa and HEK293T cells ( Figure 1C and 1D), suggesting that the ER-retaining pathway was blocked.
We then used digitonin, which selectively permeabilizes the plasma membrane but not the intracellular membranes [17], to treat the cells after fixation. Under this treatment, GM130, which peripherally attaches to the cytoplasmic surface of the Golgi apparatus, was detected by its antibody, whereas the ER luminal protein PDI could not be recognized by its own antibody ( Figure S2). Similarly, the cytosolic EGFP tag of APP-EGFP fusion protein was immunolabeled by anti-EGFP antibody, whereas b4GalT-1-EGFP, whose EGFP is on the luminal side of the ER and the Golgi apparatus, was not immunolabeled ( Figure 1E). Also, anti-EGFP antibody did not label the overexpressed calu-1/2-EGFP by immunofluorescence after digitonin treatment ( Figure 1E), suggesting that calu-1/2 localized in the lumen of the ER and the Golgi apparatus. Furthermore, to determine the signal responsible for the entry of calu-1/2 into the lumen, we generated several constructs of calu-2 and visualized their subcellular localization, and found that the 19 amino acids on the N-terminal were necessary and sufficient for the Golgi localization of calu-2-EGFP ( Figure 1F). These results, together with the previous report that the N-terminal 19 amino acids of calu-1 are cleaved [14], suggest that calu-1/2-EGFP are secreted to the extracellular space, and that the N-terminal 19 amino acids act as a signal peptide, which docks calu-1/2-EGFP to the ER and then translocates them into the lumen of the endomembrane system. Intracellular transport of calu-1/2-EGFP To monitor the intracellular transport and secretory process of calu-1/2-EGFP, we monitored the distribution and dynamic behavior of calu-1/2-EGFP in live HeLa cells ($30 cells) using an Andor spinning disk confocal microscope system. Both calu-1-EGFP and calu-2-EGFP localized to the secretory pathway, including the ER network and the Golgi apparatus ( Figure 2A). Calu-1/2-EGFP-containing vesicles were observed throughout the cytoplasm (Figure 2A). Some appeared to be spherical ( Figure 2B), while others were rod-shaped and tubular ( Figure 2C). These calu-1/2-EGFP-containing vesicles were intracellularly transported (Figure 2A-C; Movies S1 and S2).

Two types of calu-1/2-EGFP secretion
To further visualize the secretory process of calu-1/2-EGFP, we observed in more detail the in vivo behavior of calu-1/2-EGFP at the periphery of the cells, and found two main types of secretion. Some the calu-1/2-EGFP -containing vesicles moved toward the cell periphery ( Figure 4A; Movies S1 and S2). One or two seconds later, the calu-1/2-EGFP signal faded rapidly and finally vanished, suggesting that the calu-1/2-EGFP proteins were secreted into the extracellular space. Interestingly, we noted another secretory manner, in which vesicles were transported towards the periphery of the cell and accumulated at cellular processes ( Figure 4B; Movies S1 and S2). We also noted that these cell-tip-heading vesicles seemed to fuse with each other before being secreted together ( Figure 4C), indicating that calu-1/2-EGFP proteins were secreted collectively after their accumulation in the tip of cells.
Calu-1/2-EGFP-containing vesicles are transported by Kif5b and cytoplasmic dynein along microtubules Having visualized the intracellular transport and secretion of calu-1/2-EGFP, we then further investigated the molecular mechanism of calu-1/2 transport. To determine whether the secretion of calu-1/2-EGFP depends on the cytoskeleton, we treated the calu-1/2-EGFP-overexpressing cells with different drugs to interfere with the polymerization of microtubules or actin filaments. Nocodazole binds to b-tubulin and blocks tubulin selfassembly, thus interfering with the polymerization of microtubules [18]. Cytochalasin blocks both assembly and disassembly of individual actin monomers by binding to the plus ends of microfilaments, thus inhibiting the normal polymerization and elongation of actin filaments [19]. After nocodazole treatment, the protein level of calu-2-EGFP markedly deceased in the cultured medium, while increased a little in the cell lysate ( Figure 5A). On the contrary, cytochalasin treatment did not alter the secretory level ( Figure 5A). Besides, using immunofluorescence assay, we found calu-1/2-EGFP green dots localized along microtubules to cellular processes ( Figure 5B and S3). Therefore, the secretion of calu-1/2-EGFP is microtubule dependent but is not affected by the integrity of actin filament.
In eukaryotes, there exist two kinds of microtubule motor proteins, kinesin and dynein, which widely participate in vesicle transport [20,21]. The velocity of kinesin heavy chain (Kif5b) motor and cytoplasmic dynein both varied a lot in previous reports, but many researches showed that both Kif5b and cytoplasmic dynein move at ,1 mm/s in vivo [22][23][24][25][26][27]. Considering that Kif5b and cytoplasmic dynein moved at different directions, we measured the inward and outward velocity of calu-1/2-EGFPcontaining vesicles separately. We found that calu-1/2-EGFPcontaining vesicles heading the Golgi apparatus moved at ,0.9 mm/s, while those moving towards the cell periphery moved at ,1.2 mm/s ( Figure 6C). These two velocities fit well with the two peaks and are similar to the previously reported velocities of Kif5b and cytoplasmic dynein, suggesting that calu-1/2-EGFP- To further confirm our hypothesis, we performed dominant negative experiments. Overexpressed p50, one subunit of the dynactin complex, acts as a dominant-negative inhibitor of dyneindynactin, and disrupts cytoplasmic dynein-dependent organelle movement [28]. Kif5b interacts with its cargoes through kinesin light chain via the kinesin-light-chain binding domain (KLCBD, amino acids 760-830), or through cargo binding domain (CBD, amino acids 831-964). Overexpression of the KLCBD and CBD of Kif5b cause a dominant negative phenotype of Kif5b, through their competitive cargo binding effect [29,30]. In the cells transfected with p50, Kif5b-CBD or Kif5b-KLCBD&CBD, the protein level of calu-2-EGFP in the cultured medium reduced significantly ( Figure 6D and 6E). Besides, overexpression of Kif5b-KLCBD did not change the secretion level ( Figure 6D), suggesting that the CBD, but not the KLCBD of Kif5b, is involved in the recognition of calu-1/2-EGFP-containing vesicles. However, we could not exclude the possibility that the overexpressed Kif5b-CBD directly inhibited the activity of endogenous Kif5b through the binding to its motor domain [31]. Therefore, we knocked down Kif5b, detected the reduced secretion level of calu-2-EGFP ( Figure 6F), further suggesting that Kif5b participated in transporting calu-1/2-EGFP. Similarly, the knockdown of cytoplasmic dynein also reduced the secretion level of clau-2-EGFP ( Figure 6F). Collectively, these data suggested that cytoplasmic dynein and Kif5b were involved in the transport of calu-1/2-EGFP-containing vesicles.
To further demonstrate the motor-cargo association, we performed pull-down and immunoprecipitation analysis. Calu-2-EGFP was pulled down by GST-Kif5b-CBD ( Figure 6G). Immunoprecipitation assay also confirmed the association between Export signal of calu-1/2-EGFP Soluble cargo proteins in the intracellular lumen are exported by either bulk flow or receptor-mediated transport [32][33][34][35][36][37]. To investigate whether calu-1/2-EGFP is passively or actively exported into the extracellular space, we expressed the fusion proteins SP19-EGFP, which are the EGFP-tagged N-terminal 19 amino acids of calu-1/2 ( Figure 8A). If the secretion of calu-1/2-EGFP is through bulk flow, SP19-EGFP should have the same percentage secretion as full-length calu-1/2-EGFP. However, we found that the percentage secretion of SP19-EGFP was dramatically decreased compared to the calu-2-EGFP ( Figure 8A). To exclude the possibility that N-terminal signal peptide cleavage of SP19-EGFP was disrupted, thus leading to its retention, we further studied SP22-EGFP. SP22-EGFP showed similar percentage secretion to SP19-EGFP ( Figure 8A), suggesting that calu-1/2 may not be secreted through bulk flow but via receptor-mediated secretory pathway.
Next, to identify the export signal for the calu-1/2 secretion, we constructed transcripts of calu-2(D1-D6)-EGFP ( Figure 8A). The percentage secretions of calu-2(D1-D3)-EGFP were nearly the same as that of calu-2-EGFP; on the contrary, calu-2(D4-D6)-EGFP exhibited a significant reduction of secretory level ( Figure 8A). These data suggested that a sorting signal, which was crucial for the receptor-mediated export of calu-1/2-EGFP, was located between the amino acid positions 20 and 46. Furthermore, within the sorting signal, we noted that the proline at site 21 was reported to be an ER export signal [14]. To determine whether this site alone plays a vital role in calu-1/2-EGFP secretion, we mutated it to alanine, and measured the percentage secretion. The percentage secretion of calu-2-P21A-EGFP was only slightly decreased compared with the control (Figure 8B and 8C), suggesting that the proline at site 21 is not the only key element for calu-1/2-EGFP secretion but may contribute to the integrity of the sorting signal between amino acids 20-46. Asparagine at site 131 is important for calu-1/2-EGFP stability The asparagine at site 131 is predicted to be an N-glycosylation site [8,11]. To test this hypothesis, we investigated whether this site affects calu-1/2-EGFP secretion. Surprisingly, we found that when the asparagine at position 131 was mutated to alanine, calu-2(N131A)-EGFP was degraded extensively both in the intracellular and extracellular space ( Figure 8B), suggesting that the Nglycosylation at this site plays an important role in calu-1/2-EGFP stability, which needs further investigation.

Discussion
In this study, we showed that calu-1/2-EGFP are translocated into the lumen with the help of their signal peptide, transported intracellularly in vesicles, and finally secreted into the extracellular space (Figure 9), during which the sorting signal on calu-1/2 is indispensable. We examined the velocities of calu-1/2-EGFPcontaining vesicles, and identified the motor proteins that participate in powering calu-1/2-EGFP-containing vesicles. Most interestingly, besides the traditional mode of secretion, we also observed another type of secretion, in which calu-1/2-EGFPcontaining vesicles accumulated in cellular processes and were secreted collectively.
During the secretory process, protein with an N-terminal signal peptide is synthesized in the rough ER-bound ribosomes and then translocated into the ER lumen. After proper folding and modification, the protein is delivered to the Golgi apparatus for further modification and finally released into the extracellular space through vesicular transport [38]. Among the various motor proteins responsible for the intracellular transport of diverse cargoes, the microtubule motor protein cytoplasmic dynein moves towards the minus end of the microtubules, participating in anterograde transport from the ER to the Golgi apparatus [21,39], and Kif5b, a microtubule plus end-directed motor, is responsible for post-Golgi traffic [20,40]. Previous reports about the velocities of Kif5b and cytoplasmic dynein in vivo varied a lot, from 0.2 mm/s to larger than 3 mm/s [22][23][24][25][26][27][41][42][43][44]. In our experiments, calu-1/ 2-EGFP-containing vesicles also moved at different speeds, with some moved at 0.3-0.5 mm/s while some fast vesicles reached 3 mm/s. This is consistent with previous reports. Besides, we found that calu-1/2-EGFP-containing vesicles moved at ,0.9 mm/s inward and ,1.2 mm/s outward. We hypothesized that cytoplasmic dynein and Kif5b were responsible for the intracellular transport of calu-1/2-EGFP at different speeds. Using dominant negative and RNAi methods, we found that the blockage of Kif5b or cytoplasmic dynein significantly inhibited the secretion of calu-1/2-EGFP ( Figure 6D-F). Moreover, GST pull-down and immunoprecipitation analysis further confirmed the association between Kif5b/cytoplasmic dynein and calu-1/2-EGFP-containing vesicles ( Figure 6G-H). Therefore, it is possible that cytoplasmic dynein mediates the ER-to-Golgi transport of calu-1/2-EGFP at a relatively slower speed of ,0.9 mm/s, while Kif5b executes the transport of calu-1/2-EGFP-containing vesicles at ,1.2 mm/s during post-Golgi transport (Figure 9). Also, we do not exclude the possibility that calu-1/2-EGFP-containing vesicles are bidirectionally transported through spatial-temporal regulation of Kif5b and dynein cooperation [45,46], and some other motor proteins may also be involved in the transport of calu-2-EGFP.
In our experiments, we found two types of secretion of calu-1/2-EGFP. In the model of ''secretion after attachment'' ( Figure 9B), which is traditionally reported [42], the release of the secretory product is thought to be regulated by the network of actin filaments. It was suggested that the secretory vesicles traveling to the periphery of the cell are inhibited to fuse with each other and the plasma membrane by a dense web of microfilaments, and the stimulation of the rapid depolymerization of these actin filaments leads to the following vesicle fusion and secretion [47,48].
However, our results show that the disruption of actin filament network does not enhance the secretory level of calu-2-EGFP ( Figure 5A), suggesting most calu-1/2-EGFP may not use this secretory mode, but are secreted through the model of ''secretion after accumulation'' ( Figure 9C), which is also observed in mRNA trafficking [49]. Moreover, the projection of the green fluorescent dots ( Figure 4B) not only stands for the trafficking paths of calu-1/ 2-EGFP-containing vesicles, but may also represent the microtubule tracks. This is further supported by the immunostaining data, in which microtubules pointed to cellular processes with calu-2-EGFP dots aligning along them ( Figure 5B and S3). Therefore, we propose that it is the microtubules directed towards cellular processes that facilitate the accumulation of calu-1/2-EGFPcontaining vesicles. Previous reports show that exocytotic vesicles are targeted to the ELKS-containing patches where MT plus ends are attached [42,50]. So it is possible that the cellular processes are abundant with ELKS and attract calu-1/2-EGFP-containing vesicles.
Soluble proteins with ER exit signals are recognized by transmembrane receptors, which recruit the coat proteins, and mediate the transport from the ER to the Golgi apparatus, and finally to the cell periphery [32,33,35]. Besides, some proteins without ER exit signals also enter secretory vesicles and are secreted [34,37]. Our data indicate that calu-1/2-EGFP are secreted through the receptor-mediated pathway, with their exit signal located at amino acid positions 20-46. In addition, we identified the asparagine at position 131, which was predicted to be a glycosylation site [8,11], as essential for calu-1/2-EGFP stability. The asparagine to alanine mutation lead to  extensive degradation of calu-2-EGFP both intracellularly and extracellularly. This can be explained by two possibilities: either the glycosylation at this site stabilize calu-2-EGFP, protecting it from being cleaved by protease, or the mutation itself leads to misfolding of calu-2(N131A)-EGFP, its stuck in the ER and subsequent degradation via the ERAD pathway. Further experiments are needed to uncover the detailed mechanism.

Immunoblotting analysis and percentage secretion
Proteins were separated by SDS-PAGE and then transferred onto PVDF membranes (Millipore) in a semidry transfer cell (Bio-Rad). The membranes were blocked and subsequently probed with primary and secondary antibodies as previously reported [57].
For detecting the protein level in the cultured medium of HeLa cells, we substituted the medium with the serum-free DMEM 24 hours after transfection, and collected the total cell lysate (TCL) and the corresponding serum-free cultured medium (CM) from each cell culture dish 48 hours after transfection. As for experiments in HEK293T cells, we changed the medium with serum-free medium 12 hours post transfection and collected for 24 hours. Both TCL and CM were compressed to 50 mL, and a fifth of the total sample (10 mL) was loaded and subjected to Western blotting. The intensity of each target band was quantified using Image J analysis software (NIH). The percentage secretion is the ratio of the band intensity in the CM to the total intensity of the target band in both TCL and CM. Besides, in some experiments, cells were treated with DMSO, nocodazole (2 mg/ml), and cytochalasin B (1 mg/ml) in serum-free medium, and the TCL and CM were collected 24 hours later for Western blotting analysis.

Fluorescence microscopy and time-lapse imaging
Transfected HeLa cells were fixed with freshly prepared 4% paraformaldehyde in PBS, and then subjected to direct observation under a fluorescence microscope (Olympus) or a confocal microscope (Leica). For immunofluorescence, the fixed cells were incubated with antibodies after permeabilization (0.15% Triton X-100 in PBS) and blocking. For digitonin treatment, the cells were permeabilized with 25 mg/ml digitonin in KHM buffer on ice as previously reported [17], and then fixed with 4% paraformaldehyde. For live cell imaging, the transfected cells were subjected to spinning disk laser microscopy (Andor) 24 hours after transfection. Images were collected using Andor software.

RNA interference
For knockdown experiment in HeLa cells, siRNAs targeting Kif5b and cytoplasmic dynein according to previously report [42] were synthesized, and transfected into the HeLa cells using Lipofectamine 2000 (Invitrogen) as the manufacture's instruction. Control scramble sequence were designed by GenePharma Co.,Ltd.

Pull-down and immunoprecipitation
For pull-down analysis, the fusion protein GST-Kif5b-CBD was prokaryotically expressed and attached to glutathione-sepharose 4B beads (Amersham). The total cell lysate of calu-2-EGFP transfected HEK293T cell was then collected and incubated with the affinity column in lysis buffer (20 mM HEPES, 120 mM NaCl, 5 mM EGTA, 2 mM MgCl2 and 1 mM DTT, pH 7.4) at 4uC overnight. The combined protein was harvested and tested by Western blotting.
For immunoprecipitation assay, the calu-2-EGFP and myc-p150 transfected HEK293T cell was lysed in IP buffer (20 mM HEPES, 250 mM sucrose, 10 mM KCl, 1 mM EGTA, 1 mM EDTA, 1.5 mM MgCl2 and 1 mM DTT, pH 7.4) containing protease inhibitor cocktail (Roche). Then, nucleus were removed by centrifugation at 3,000 rpm for 5 minutes. The supernatant was further centrifugated at 8,000 rpm for 15 minutes to remove mitochondria. The remained cytosolic and membrane fraction was then incubated with anti-myc antibody at 4uC overnight. The protein samples that combined with the Protein A Sepharose (Amersham) were collected and subjected to Western blotting.

Statistical analysis
Images analyzed using Andor software and Image J (NIH). Vesicle motility (with a velocity exceeding 0.3 mm/s), relative fluorescence intensity, and vesicle path projection were acquired as previously reported [24,42]. The velocities of moving vesicles were measured using the Image J software (NIH). The distance that one dot moved every two frames was calculated, with dots tracked manually, and the corresponding velocity was measured according to the interval time (0.376 seconds). Approximately 2000 dots ($10 cells) were measured for each group. As for the calculation of the number of moving vesicles, we manually counted the number of moving dots in each cells for one minute ($10 cells in each group).
Results are presented as mean 6 standard deviation. Comparisons were analyzed by unpaired, two-sided, independent student's t test with or without equal variance assumption, according to the result of an F test. * means p,0.05, ** means p,0.01, *** means p,0.001.