Mechanisms of procollagen and HSP47 sorting during ER-to-Golgi trafficking

Efficient quality control and export of procollagen from the cell is crucial for extracellular matrix homeostasis, yet it is still incompletely understood. One of the debated questions is the role of a collagen-specific ER chaperone HSP47 in these processes. Most ER chaperones preferentially bind to unfolded polypeptide chains, enabling selective export of natively folded proteins from the ER after chaperone release. In contrast, HSP47 preferentially binds to the natively folded procollagen and is believed to be released only in the ER-Golgi intermediate compartment (ERGIC) or cis-Golgi. HSP47 colocalization with procollagen in punctate structures observed by immunofluorescence imaging of fixed cells has thus been interpreted as evidence for HSP47 export from the ER together with procollagen in transport vesicles destined for ERGIC or Golgi. To understand the mechanism of this co-trafficking and its physiological significance, we imaged the dynamics of fluorescently tagged type I procollagen and HSP47 punctate structures in live MC3T3 murine osteoblasts with up to 120 nm spatial and 500 ms time resolution. Contrary to the prevailing model, we discovered that most bona fide carriers delivering procollagen from ER exit sites (ERESs) to Golgi contained no HSP47, unless the RDEL signal for ER retention in HSP47 was deleted or mutated. These transport intermediates exhibited characteristic rapid, directional motion along microtubules, while puncta with colocalized HSP47 and procollagen similar to the ones described before had only limited, stochastic motion. Live cell imaging and fluorescence recovery after photobleaching revealed that the latter puncta (including the ones induced by ARF1 inhibition) were dilated regions of ER lumen, ERESs, or autophagic structures surrounded by lysosomal membranes. Procollagen was colocalized with HSP47 and ERGIC53 at ERESs. It was colocalized with ERGIC53 but not HSP47 in Golgi-bound transport intermediates. Our results suggest that procollagen and HSP47 sorting occurs at ERES before procollagen is exported from the ER in Golgi-bound transport intermediates, providing new insights into mechanisms of procollagen trafficking. ER HSP47. We created fluorescent constructs of type I procollagen and HSP47 and imaged them in a murine osteoblastic cell line MC3T3 together with protein markers of the ER lumen, COPII coat, ERGIC membranes, cis-Golgi, autophagic structures, and lysosomes. Osteoblasts, which are the cells responsible for making bone, are some of the most active secretory cells that produce and secrete massive amounts of type I collagen. we Golgi Upon Our results support a revised which latter in ERGIC53-postive transport

by preferentially binding to the triple helical rather than unfolded conformation [28][29][30][31]. Unlike other ER chaperones, HSP47 is accrued at multiple binding sites along the triple helix instead of being released upon completion of procollagen folding. It might also serve as an adapter for procollagen loading into ERESs by TANGO1 [32]. Thus, it is widely thought that HSP47 is co-transported with procollagen from the ER to ERGIC or cis-Golgi, followed by its release at lower pH [33,34], capture by KDEL/RDEL receptors, and retrograde traffic to the ER [25,26].
Previously, HSP47 was observed by immunofluorescence in vesicle-like procollagen puncta after brefeldin A [35] or GTPγS [36] disruption of ER-Golgi trafficking. It was also reported to be colocalized with TANGO1 and COPII in large procollagen carriers by immunofluorescence and by fractionation in a cell-free vesicle budding reaction [6]. However, these techniques cannot unambiguously identify bona fide transport vesicles. In fixed cells imaged by immunofluorescence, procollagen/HSP47 puncta might instead represent dilated regions of the ER, ERESs, autophagic structures, other cell compartments, or non-physiological structures caused by the disruption of cellular function with brefeldin A or GTPγS. Cell-free assays might produce vesicles that do not form inside the cell under physiological conditions. Unambiguous identification of any procollagen puncta as transport intermediates requires a proof of their disconnection from the ER, directional (nonstochastic) motion, and entry into Golgi. Having developed an assay for live cell imaging capable of monitoring these processes [37], we therefore decided to revisit procollagen trafficking regulation by HSP47.
Here we describe dynamic imaging of puncta containing fluorescently tagged type I procollagen and HSP47 in live cells with ~120 nm spatial and up to 500 ms/frame time resolution, which reveals that procollagen is normally exported from osteoblast ER without HSP47. We created fluorescent constructs of type I procollagen and HSP47 and imaged them in a murine osteoblastic cell line MC3T3 together with protein markers of the ER lumen, COPII coat, ERGIC membranes, cis-Golgi, autophagic structures, and lysosomes. Osteoblasts, which are the cells responsible for making bone, are some of the most active secretory cells that produce and secrete massive amounts of type I collagen. Expression of the fluorescent procollagen and HSP47 constructs in these cells did not disrupt their function. The constructs faithfully reproduced the subcellular localization of the corresponding endogenous proteins observed by immunofluorescence as well as procollagen/HSP47 localization patterns previously reported in static snapshots of fixed cells. Upon live cell imaging, however, we found no evidence for HSP47 accompanying procollagen out of the ER in Golgi-destined transport intermediates. Rather, we observed HSP47 only in the ER and in stochastically moving vesicle-like structures, which were distinct from bona fide procollagen carriers that moved directionally toward the Golgi apparatus. Upon deleting or modifying the RDEL retention signal, HSP47 was co-trafficked with procollagen to Golgi in directionally moving vesicles. Our results support a revised trafficking mechanism in which HSP47 dissociates from procollagen at ERESs, before the latter is exported in ERGIC53-postive transport intermediates bound for the Golgi.

Fluorescently tagged HSP47, proα1(I), and proα2(I) constructs for live cell imaging of procollagen and HSP47 trafficking
We built fluorescent protein (FP)-tagged mouse HSP47 constructs, in which different color FPs were added on the N-terminus of the molecule, between the ER signal peptide and the start site of HSP47. GFP-HSP47 transiently transfected into the MC3T3-E1 mouse osteoblast line [38] co-immunoprecipitated with procollagen, suggesting that its binding to collagen was similar to endogenous HSP47 (K d ~ 1 μM, Fig. 1A). We observed complete colocalization of FP-tagged HSP47 molecules with each other and antibody-tagged HSP47, i.e. all spots positive for one of the tags were also positive for the other tags (Fig. 1B). Variation in the relative intensity of the tags at different spots appeared to be related primarily to antibody accessibility, since the largest difference was observed behind the cell nucleus. Localization of all FP-HSP47 constructs (TagBFP2, GFP, Venus, Apple, and Cherry) was similar to endogenous HSP47.
We performed procollagen/HSP47 sorting experiments separately with FP-proα1(I) and FP-proα2(I) constructs because of the difference in cell adaptation mechanisms responding to overexpression of proα1(I) and proα2(I) chains. Excessive synthesis of proα2(I) due to the exogenous Col1a2 produces unassociated proα2(I) chains that must be removed from the ER and degraded (only one proα2(I) can be incorporated into trimeric procollagen). In contrast, excess of proα1(I) relative to proα2(I) causes formation of proα1(I) homotrimers, which appear to be secreted normally by osteoblasts and cause no significant bone pathology even in individuals who do not produce proα2(I) at all [39,40].
The design and validation of the FP-proα1(I) and FP-proα2(I) constructs was described in more detail in our previous study [37]. Briefly, we found that both FP-proα1(I) and FP-proα2(I) chains assembled into procollagen heterotrimers, progressed through the normal secretory pathway, and after being secreted were incorporated into extracellular collagen fibers. In MC3T3 cells with low-to-moderate amounts of FP-procollagen (i.e., 12-24 h post transfection) we observed no overexpression artifacts and the results were the same for both constructs. In MC3T3 cells with excessive expression of FP-procollagen and after more than 24 h post transfection, we noticed the formation of large dilated ER regions filled by FP-procollagen chains (Supp. Fig. 1), which was probably due to irreversible aggregation of procollagen. The effects were more pronounced after FP-proα1(I) transfection, probably because of larger disruptions in procollagen folding and trafficking caused by incorporation of more than one FP-proα1(I) chain into the same molecule.
Since FP tags used for live cell imaging might perturb folding and trafficking of the corresponding proteins, we performed multiple control experiments. We utilized only the tags that have been well characterized before or additionally validated by colocalization with each other and endogenous protein antibodies (Fig. 1B, Supp. Fig. 2, and Experimental Procedures). To avoid ER disruption caused by slow accumulation of FP-tagged proteins, we did not use over-transfected cells with large regions of dilated ER and limited experiments to 18-24 h post transfection. We compared FP-proα1(I) and FP-proα2(I) transfections, exploiting the difference in the number of FP-tagged chains that can be incorporated into the same procollagen molecule. We performed key experiments with different combinations of FP tags, since different FPs might produce different artifacts. To emphasize the importance of such controls, most images in this manuscript are shown for both FP-proα1(I) and FP-proα2(I) transfections and different combinations of FP tags.
Additionally, we generated an MC3T3 cell line, in which one endogenous Col1a2 allele was modified with CRISPR/CAS9 to produce the same proα2(I) chain as the GFP-proα2(I) construct used here. The endogenous GFP-procollagen was trafficked similar to the transfected one. It also gradually accumulated in the ER after stimulation of procollagen synthesis with ascorbic acid, eventually causing ER disruption and suppression of collagen synthesis or cell demise. This cell line provided no added benefits for the present study and its generation, characterization, and possible utilization will be described in detail separately.

Dynamic imaging of FP-HSP47 co-localization with procollagen
As we reported previously [37], FP-proα1(I) and FP-proα2(I) were observed in the ER lumen marked by ssFP-KDEL ( Fig. 2A,B) and in cis-Golgi marked by FP-GM130 (Fig. 2C,D). FP-HSP47 perfectly co-localized with FP-proα1(I) and FP-proα2(I) within the ER lumen. It was not observed in the cis-Golgi, unless the cells were subjected to excessive stress or over-transfected with FP-HSP47. The strict ER localization of FP-HSP47 under normal expression conditions suggested FP-HSP47 was either rapidly removed by retrograde transport back to the ER or never left the ER to enter the cis-Golgi. To distinguish between these two possibilities, we photobleached FP-HSP47 and FP-procollagen in the Golgi area to remove fluorescence in this area and then imaged transport vesicles moving into this region from distant ERESs.
Time lapse imaging of fluorescence recovery after photobleaching (FRAP) revealed rapidly moving FP-proα1(I)/α2(I) puncta entering into the cis-Golgi, but these puncta did not contain any FP-HSP47 ( Fig. 3; Movies 1 and 2). We also observed numerous vesicle-like puncta of FP-proα1(I)/α2(I) co-localized with FP-HSP47, but their restricted, stochastic-like movements were inconsistent with those of ER-to-Golgi transport intermediates.
To better visualize procollagen and HSP47 dynamics in the cell and trace rapidly moving procollagen puncta, we performed time-lapse imaging without photobleaching at 2 s/frame ( Fig. 4A; Movie 3), 1 s/frame ( Fig. 4B; Movie 4), and 500 ms/frame (Supp. Fig. 3, Movie 5). Regardless of the imaging speed, we observed no directional movement of FP-HSP47-labeled structures. Rather, their limited movements appeared to reflect stochastic fluctuations caused by rearrangements of the ER network. We traced numerous procollagen transport intermediates distinguished by rapid directional movement, likely along microtubules, some moving toward the cis-Golgi and some toward the plasma membrane (Movies 3-5). We detected HSP47 in just one or two out of the hundreds of vesicles that appeared to be Golgi-bound (for more accurate quantification, see below).

Effects of the RDEL sequence on FP-HSP47 localization and trafficking
One possible explanation for the absence of FP-HSP47 in transport vesicles delivering procollagen to Golgi is that HSP47 is retained in the ER because of its ER localization sequence, RDEL. To test this hypothesis, we made FP constructs of HSP47 in which this sequence was deleted (ΔRDEL) or mutated to RNGL. FP-HSP47ΔRDEL colocalized with FP-proα1(I)/α2(I) in the ER and cis-Golgi (Fig. 5A), but its expression caused cis-Golgi disruption and formation of large FP-GM130 spheres within the nucleus similar to droplets previously observed upon GM130 overexpression [41]. FP-HSP47RNGL also co-localized with FP-proα1(I)/α2(I) in the ER and cis-Golgi, causing less pronounced yet noticeable Golgi disruption and nuclear FP-GM130 spheres (Fig. 5B). Time lapse imaging revealed that FP-HSP47RNGL was present in many FP-proα1(I)/α2(I) transport vesicles ( Fig. 5C; Movie 6 and 7). Apparently, HSP47 release from the ER and trafficking to cis-Golgi was responsible for the Golgi disruption and mislocalization of FP-GM130 in the nucleus. This effect might be important for understanding how HSP47 mutations cause osteogenesis imperfecta [42,43], but its detailed analysis is beyond the scope of the present paper and will be reported elsewhere.

Identification of vesicle-like structures containing FP-HSP47 and FP-proα1(I)/α2(I)
The observation of FP-HSP47ΔRDEL and FP-HSP47RNGL but not FP-HSP47 in procollagen transport vesicles was consistent with HSP47 retention in the ER. If true, this interpretation would also mean that vesicle-like structures containing both procollagen and HSP47 were either regions of the ER or autophagic structures derived from the ER, as suggested by their stochastic, ER-like dynamics.
To validate this deduction, we co-transfected the cells with ER and autophagy markers (Fig. 6, Movie 8). Some vesicle-like puncta containing FP-proα1(I)/α2(I) and FP-HSP47 were indeed colocalized with the ER lumen marker ssFP-KDEL (Fig. 6A, Movie 8) and surrounded by ER membrane (Supp. Fig. 1). Almost instantaneous recovery of ssFP-KDEL fluorescence after photobleaching in these puncta indicated that they were vesicle-like regions of ER lumen connected to the ER network rather than actual vesicles (Fig. 6A, Movie 8). Slower fluorescence recovery of FP-HSP47 and FP-procollagen was consistent with larger size and therefore slower diffusion of these molecules (Fig. 7C). Other procollagen/HSP47 puncta were autophagic structures colocalized with and/or surrounded by the lysosome and late endosome membrane marker FP-LAMP1 (Fig. 6B, Supp. Fig. 4), which could be formed by lysosomal engulfment of ER exit sites (ERES) [37], lysosomal engulfment of ER lumen [44], or ER-phagy [45]. Correlative light and electron microscopy supported the existence of ERESs engulfed by lysosomal membranes (Supp. Fig. 4 and Fig.  7 in Ref. [37]) as well as vesicle-like ER lumen regions containing procollagen (Supp. Fig.  4). These structures were not procollagen transport vesicles, yet static snapshots of such structures could be misinterpreted as transport vesicles.
We next revisited a frequently cited observation of accumulation of vesicle-like procollagen/ HSP47 structures after Golgi disruption by brefeldin A (BFA), which was interpreted as accumulation of transport vesicles that had nowhere to go [35]. We similarly observed accumulation of spherical, vesicle-like puncta containing FP-HSP47 and FP-proα1(I)/α2(I) after 1 h treatment with 5 μg/ml BFA (Fig. 7A, Supp. Fig. 5). However, these puncta were colocalized with ssFP-KDEL. Nearly instantaneous recovery of ssFP-KDEL fluorescence after their photobleaching revealed that they were vesicle-like ER regions rather than transport vesicles (Fig. 7C, Movie 9). Instead of accumulation, BFA caused disappearance of transport vesicles (Fig. 7A). Consistently, in cells expressing FP-HSP47RNGL, BFA caused accumulation of vesicle-like ER regions containing ssFP-KDEL and disappearance of Golgi and transport vesicles containing just FP-proα1(I)/α2(I) and FP-HSP47RNGL (Fig. 7B).
In addition to HSP47 retention in the ER, these effects suggested dependence of procollagen export from the ER on activation of ARF1 GTPase by GDP→GTP exchange, since BFA is an inhibitor of this exchange [46,47]. Given that GDP→GTP exchange activates membrane attachment of ARF1 and COPI coatomer recruitment, we examined ARF1 localization in cells co-transfected with FP-ARF1, FP-proα2(I), and FP-SEC23 (Supp. Fig. 6). Outside of the Golgi area, we found numerous punctate ARF1 structures, ~ 50±8% of which were colocalized with procollagen and most of which appeared to be associated with ERESs.

FP-HSP47 and FP-proα1(I)/α2(I) co-localization with ERGIC53
How could we reconcile the evidence pointing to HSP47 retention in the ER with the preferential HSP47 binding to natively folded procollagen in the ER? One hint is the putative role of HSP47 as an adapter for procollagen loading into the ERES by TANGO1 [32], which suggests that HSP47 might be loaded into ERESs as well. Indeed, colocalization of FP-HSP47, FP-proα1(I)/α2(I), and COPII coat marker FP-SEC23 in punctate structures supported the entry of HSP47 into ERESs together with procollagen ( Fig. 8A, Movie 10). Subsequent formation of transport intermediates at ERES could then be preceded by HSP47 release and retrieval back to the ER (enabled by distinct ERES environment).
To test the latter hypothesis, we created, validated, and imaged the intracellular localization and dynamics of a fluorescently tagged membrane marker FP-ERGIC53 (Supp. Fig.  2A). This membrane protein is known to cycle between the ER and Golgi in transport intermediates believed to have reduced pH [48]. Live cell imaging showed FP-ERGIC53 was colocalized with procollagen in the area of cis-Golgi, at ERESs (including ones located far from the Golgi, Fig. 8B,C), and in procollagen transport intermediates destined for Golgi that exhibited rapid directional motion (Fig. 8D, Movie 11, Supp. Fig. 7A). FP-ERGIC53 labeling identified ER-Golgi transport intermediates that could not be traced to their origin or destination, distinguishing them from secretory vesicles. FP-HSP47 was detected in fewer than 1±3% of these intermediates and only in cells with higher FP-HSP47 expression (Supp. Fig. 7C). In contrast, FP-HSP47RNGL was detected in over 90±12% of procollagen transport intermediates regardless of FP-HSP47RNGL expression level (Fig. 8E, Movie 12, Supp. Fig. 7B). Therefore, HSP47 appeared to be only leaking rather than co-transported with procollagen to Golgi.

Discussion
Because HSP47 was shown to be a collagen-specific ER chaperone and its mutations were found to cause severe osteogenesis imperfecta with pronounced bone fragility and deformities, its interactions with procollagen have been extensively studied. Multiple observations have suggested co-trafficking of procollagen with bound HSP47, release of HSP47 at reduced pH in cis-Golgi (or ERGIC), and return to the ER by retrograde transport [25,26].
We tested this hypothesis by imaging trafficking of fluorescently tagged type I procollagen and HSP47 in live MC3T3 osteoblasts with up to ~120 nm spatial and 500 ms/frame time resolution. Unlike previous snapshots of fixed cells, time lapse imaging of live cells unequivocally distinguished procollagen transport intermediates from other vesicle-like punctate structures based on their directional vs. stochastic motion. Co-transfection of the cells with fluorescently tagged protein markers of different subcellular compartments, FRAP experiments, and tracking the origin and destination of transport intermediates revealed the nature and composition of different vesicle-like structures. Many post-Golgi procollagen carriers appeared to be larger and slower than ER-Golgi transport intermediates. Nonetheless, to avoid misidentification, we examined the composition of only those procollagen transport intermediates that could be traced to their origin, destination, or both. To quantify HSP47 content, we identified ER-Golgi procollagen carriers based on the presence of FP-ERGIC53 (observed in most ER-Golgi transport intermediates; Fig. 8, Supp. Fig. 7, Movies 11, 12, S2, S3). Multiple control experiments revealed no artifacts or off-target effects of low-to-moderate expression of the FP-tagged constructs in MC3T3 cells within 24 h of transfection.
Our observations suggest that HSP47 is released from procollagen before rather than after procollagen transport intermediate fission from ERESs ( Fig. 9) and that the RDEL sequence in HSP47 prevents the protein from being exported from the ER in such intermediates. The bona fide transport intermediates contained only procollagen and ERGIC53, emerged from ERESs, exhibited characteristic directional motion consistent with movement along microtubules, and merged into the cis-Golgi (Figs. 3 and 8A-C; Movies 1-5, 10, and 11). Their composition and dynamics were consistent with the previously described ERGIC-like carriers transporting proteins to the cis-Golgi [49,50]. With very few exceptions (< 1±3%), we observed no detectable HSP47 in these or any other (Fig. 4, Supp. Fig. 3, Movies 1-5) procollagen transport intermediates in the secretory pathway, unless the ER localization sequence RDEL of HSP47 was either deleted or mutated to RNGL (Figs. 5, 8E; Supp. Fig.  7B; Movies 6,7, and 12). We observed colocalization of procollagen, HSP47, and COPII proteins in a variety of vesicle-like structures, some of which had similar appearance to structures interpreted by other studies as procollagen transport vesicles in various human cell lines [6,51]. In MC3T3 cells, all these structures exhibited slow, limited, stochastic motion and none of them merged into the cis-Golgi ( Fig. 8A and Movie 10). These structures were either LAMP1-negative ER exit sites (Fig. 8A), larger dilated ER regions (some decorated with COPII-coated ERES), or LAMP1-positive autophagic structures (Fig. 6B), as illustrated in Fig. 9. We previously reported that all procollagen/COPII puncta that also contained ubiquitin, CUL3, or KLHL12 were positive for LC3, LAMP1, or both, indicating that they were degradative autophagic structures rather than secretory transport vesicles [37]. In this earlier study we found no evidence of directionally moving procollagen transport vesicles destined for Golgi that were marked by COPII, CUL3, or KLHL12.
We have also found the export of procollagen transport intermediates from the ER to be ARF1 GTPase dependent. In immunofluorescence studies, vesicle-like procollagen/HSP47 [35,36] or TANGO1/HSP47 [6] structures formed after SAR1 and ARF1 inhibition with GTPγS [36], ARF1 inhibition with brefeldin A (BFA) [35], COPI knockdown [6], or ARF1 Q71L transfection [6] were interpreted as transport intermediates containing HSP47. We reproduced the accumulation of such structures after BFA treatment (Fig. 7), which prevents GDP→GTP exchange at ARF1 and thereby ARF1 localization at membranes and COPI coatomer recruitment [52][53][54]. However, analysis of their composition, dynamics, and FRAP suggested that BFA blocked the secretory export of procollagen and HSP47 from the ER and that these structures were not transport intermediates but mostly dilated ER regions plus some ERESs and lysosomes (Figs. 6 and 7, Movies 8 and 9). At the same time, we previously found that BFA did not impede procollagen delivery from the ER to lysosomes for degradation (Fig. S3 in Ref. [37]). In the absence of BFA, we observed ARF1 colocalization with procollagen within elongated structures adjacent to COPII at ERES, which are consistent with transport intermediate precursors (Supp. Fig. 6). From quantitative analysis of N=9 cells in 3 such experiments with different FP combinations, we estimate that ~ 50±8 % of punctate ARF1 structures outside Golgi contain procollagen.
Taken together, these observations suggest that HSP47 release is accompanied by ARF1 and COPI dependent formation of procollagen transport intermediates at distal ERES regions (Fig. 9). This hypothesis is consistent with recent reports of COPII and TANGO1 rings at the ER-ERES interface [55,56], suggesting that only procollagen entry into the ERES but not its exit might be COPII and TANGO1 dependent [1]. It provides an alternative to ERGIC intermediates believed to form from COPII vesicles that bud from ERESs and lose their COPII coats [10,57], consistent with a recent observation of long tubular ERES membranes interacting with COPI proteins [58]. Another possible interpretation of ARF1dependent procollagen transport intermediates that contain ERGIC53 but not HSP47 is that these intermediates form upon fusion of ERGIC membranes with ERES [7]. Our findings are consistent with the latter study, yet we propose a somewhat different point of view.
We do see ARF1 structures that do not contain procollagen adjacent to SEC23 at ERES (Supp. Fig. 6). They might be arriving retrograde transport intermediates [59,60] or ERGIC membranes fusing with ERES [7]. We cannot distinguish the two interpretations and we are not sure that there is a substantive difference between them. None of the transport intermediates we tracked in this study stopped over at a distinct ERGIC compartment between ERES and cis-Golgi and none appeared to change their size or intensity, consistent with previously reported VSVG trafficking [61,62]. Therefore, we lean toward viewing transport intermediates as having all the attributes typically associated with ERGIC [61,62]. However, our data can be interpreted in either way.
Some of the questions raised by our study will have to be answered by future work that might require additional development of live cell imaging and other technologies. Indeed, it is presently unclear whether HSP47 dissociation from procollagen at ERES is mediated by reduced pH or other features of the ERES environment. We also do not know whether the procollagen transport intermediates we observed are rapidly moving vesicles like the megacarriers proposed by Santos et al [7], short tubular connections between the cis-Golgi and proximal ER like the short-loop connections proposed by McCaughey et al [9], long tubular regions of ERES membranes proposed by Weigel et al. [58], or all of the above. The existing technology might not be able to clearly answer these questions. On one hand, the time resolution required for visualizing highly dynamic transport intermediates necessitates imaging a single z-slice, in which tubular intermediates could be easily confused with transport vesicles (given that procollagen flow could occur in discrete bursts). On the other hand, the spatial resolution required for visualizing such structures in 3D necessitates imaging of fixed/frozen specimens, in which the dynamics can be only deduced but not shown. Nonetheless, we believe that a consistent picture of secretory procollagen trafficking is beginning to emerge.
Overall, our observations provide the following mechanistic insights into ER-Golgi procollagen trafficking and the role of HSP47 in this process. Type I procollagen and HSP47 appear to be separated at ERESs followed by ARF1-dependent formation of Golgi-destined procollagen transport intermediates that contain ERGIC53 but no COPII coat and no HSP47. Perhaps the release of HSP47 enables entry of procollagen from ERES into transport intermediates, like the release of general ER chaperones enables entry of folded proteins from the ER lumen into ERES. This hypothesis is consistent with HSP47 retention at ERES by RDEL receptors and degradative targeting of procollagen together with HSP47 to lysosomes from ERESs. However, it has not been validated yet and its experimental testing is ongoing, e.g. the identity of the RDEL receptors for HSP47 and their relationship to cis-Golgi KDEL/RDEL receptors are presently unknown. In MC3T3 cells, HSP47 is exported from ERESs into the degradative but not secretory pathway, although some molecules likely escape this ERES sorting process and travel to the cis-Golgi, particularly under conditions of increased HSP47 expression or cell stress. Like general ER chaperones escaping from the ER lumen, the latter HSP47 molecules would then be captured by cis-Golgi KDEL/RDEL receptors and returned to the ER by retrograde transport. In the absence of cell stress or RDEL mutations, HSP47 leakage to cis-Golgi was detectable only in <1±3% of procollagen transport intermediates in a few cells with higher levels of FP-HSP47 expression. We do not know whether cells other than MC3T3 have a different mechanism for HSP47 release and procollagen export from ERES and whether other collagen types are trafficked similar to type I procollagen. Osteoblasts could be utilizing Golgi for HSP47 retrieval only as a backup (to capture molecules missed by the ERES sorting), because they produce more procollagen than other cells. However, from an evolutionary perspective, the ERES procollagen/HSP47 sorting and export mechanisms are unlikely to be collagen type or cell specific.
GFP or Venus-tagged FP-proα2(I) and FP-proα1(I) displayed a low transfection efficiency but were otherwise normally trafficked through the cell and secreted. Specifically, lowto-moderate expression of FP-procollagen appeared to have minimal or no impact on procollagen synthesis, localization inside the cell, trafficking, or cell function within 24 h of transfection. Over-transfected cells with high expression of FP-procollagen appeared to accumulate large procollagen pools within the ER lumen, causing severe ER dilation (Supp. Fig. 1). In cells with low-to-moderate expression of FP-procollagen, such procollagen pools usually did not accumulate during the first 24 h after transfection. Therefore, imaging experiments were performed within 12-24 h after transfection only with the latter cells. TagBFP2, Cerulean or Apple-procollagen constructs had similar features but lower transfection efficiencies. Cherry-procollagen was not used for this study, because it exhibited abnormal localization patterns inside the cell (likely caused by Cherry dimerization).
Transfection and treatments.-MC3T3 cells were transiently transfected with Fugene 6 (Promega) using the manufacturer's protocols. The DNA:Fugene ratio was optimized for MC3T3 cells, as was the total DNA:Fugene concentration in the cell culture media. Due to rapid MC3T3 proliferation, the cell density at the time of transfection was optimized as well, with 50-60% confluence being an ideal timepoint for 12-24 h transfection without cells becoming overconfluent. During transfection, cells were incubated with αMEM + Glutamax media supplemented with 100 μM ascorbic acid 2-phoshate (Sigma Aldrich) and 10% FBS from Valley Biomedical (Lot #2C0550 tested for supporting osteoblast differentiation). Cells were then imaged 12-24 h after the transfection. Brefeldin A (Cell Signaling) treatment was performed by adding an equal volume of 2x solution (in the same media with ascorbic acid 2-phoshate) to chamber slides at the timepoints indicated in the text, resulting in 5 μg/ml final concentration.
Imaging and image processing.-Live or fixed cell imaging and correlative light and electron microscopy (CLEM) were performed as previously described in [37]. Live cell imaging was performed using temperature-controlled incubation stage set to maintain 37°C and 5% CO 2 . Photobleaching experiments were performed with 10-30% laser power, with 405, 488, 514, or 561 nm lasers, as needed. Except for FRAP data, all time lapse sequences and videos were corrected for photobleaching using a Bleach Correction plugin with histogram matching within the FIJI image processing package [65]. Images were enhanced using linear brightness and contrast controls without any additional modification.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material. Fig. 1. Fluorescent protein-tagged HSP47 (FP-HSP47) interacts with procollagen and colocalizes with endogenous HSP47.
(A) Western blots of cell lysates before and after IP of GFP-HSP47 with Anti-GFP beads. Green arrow shows transfected GFP-HSP47; red arrow shows endogenous HSP47. Significant enrichment of procollagen relative to vinculin on the beads revealed that it co-immunoprecipitated with GFP-HSP47. Since the estimated procollagen concentration in cell lysates was lower than 1 μM and HSP47 binds procollagen with K d ~ 1-10 μM [30], the co-immunoprecipitation indicated minimal or no disruption of HSP47-procollagen interaction by the fluorescent tag in GFP-HSP47. Because multiple endogenous HSP47 and GFP-HSP47 molecules would be bound to the same procollagen triple helix, a significant amount of endogenous HSP47 was found on the Anti-GFP beads. (B) Confocal imaging of colocalization between two FP-HSP47 constructs co-transfected into MC3T3 cells and HSP47 antibodies that label both transfected and endogenous HSP47. All spots labeled with anti-HSP47 were also positive for GFP-HSP47 and Cherry-HSP47, indicating complete colocalization of transfected and endogenous molecules in all subcellular compartments. Variations in relative fluorescence intensity of different tags are affected by antibody accessibility, uneven photobleaching during sample preparation and imaging, and other uncontrolled factors. Here and throughout the paper, different channels in multichannel images are merged using RGB pseudo color scheme and the corresponding color labels for individual channels. The scale bar is 10 μm; N=6 cells were examined.   Cherry-HSP47. Some of procollagen transport vesicles identified by rapid directional movement are outlined by white circles in the time-lapse videos and also marked by arrows in the still frames. These transport intermediates contain procollagen but not HSP47. Stochastically moving vesicle-like structures containing procollagen and HSP47 are outlined by magenta circles in the videos and also marked by arrowheads in the still frames. Scale bars = 10 μm (whole cell) and = 1 μm (zoom); N>70 (≥14 experiments with different FP combinations) for each panel.  Zoomed still frames show the same area as the inset in Movie 8. In these zoomed frames, 3 vesicle-like structures marked by Venus-proα1(I), Cherry-HSP47, and ssCFP-KDEL are traced by thin white lines. One of these structures is also highlighted by a white circle in the movie inset. The bright green structure in 45 s and 1 min frames is likely a procollagen transport vesicle containing no Cherry-HSP47 and no ssCFP-KDEL, which entered the field of view between 30 and 45 s post-bleaching (the 15s/frame imaging rate in this experiment was not fast enough for more definitive identification). Quantitative analysis of FRAP kinetics in shown in  [37] or ER lumen [44], since they are surrounded by LAMP1-positive lysosomal membrane and contain internalized procollagen, HSP47, and LAMP1. Lysosomes 6 and 7 contain little or no internalized HSP47 and thereby might have a different origin, e.g. endocytosis of secreted procollagen molecules. Scale bars = 10 μm (whole cell) and = 1 μm (zoom).   Grey structure at the bottom represents interconnected network of rough ER cisternae. Procollagen is loaded into ERES together with HSP47, as indicated by colocalization of procollagen, HSP47, and SEC23 (Fig. 8A, Movie 10). Maturation of procollagen transport intermediate precursors at distal ERES is accompanied by HSP47 release and accumulation of ERGIC53, since both ERGIC53 and HSP47 colocalize with procollagen at ERES and only ERGIC53 colocalizes with procollagen in Golgi-bound transport intermediates (Fig.  8, Movies 10-12). Unless its RDEL sequence is deleted or mutated, HSP47 is found in < 1±3% ERGIC53/procollagen transport intermediates, indicating that HSP47 is primarily returned to the ER from ERES (Supp. Fig. 7C). Only few escaping HSP47 molecules are likely to be captured by cis-Golgi KDEL/RDEL receptors and returned to the ER by retrograde trafficking. As reported before, ERESs containing misfolded procollagen and HSP47 might be engulfed by lysosomes and degraded (ERES microautophagy pathway) [37]. Under a microscope, procollagen ERESs (left), dilated ER regions (middle), and ERES microautophagy intermediates (right) appear as either small puncta or larger vesicle-like structures containing procollagen and HSP47 (Figs. 6 and 7, see also Figs. 1-7 in Ref. [37]). They accumulate upon inhibition of procollagen export from the ER by brefeldin A and are easy to confuse with transport vesicles, unless their motion and/or composition is imaged.