Trafficking and secretion of keratin 75 by ameloblasts in vivo

A highly specialized cytoskeletal protein, keratin 75 (K75), expressed primarily in hair follicles, nail beds, and lingual papillae, was recently discovered in dental enamel, the most highly mineralized hard tissue in the human body. Among many questions this discovery poses, the fundamental question regarding the trafficking and secretion of this protein, which lacks a signal peptide, is of an utmost importance. Here, we present evidence that K75 is expressed during the secretory stage of enamel formation and is present in the forming enamel matrix. We further show that K75 is secreted together with major enamel matrix proteins amelogenin and ameloblastin, and it was detected in Golgi and the endoplasmic reticulum (ER)–Golgi intermediate compartment (ERGIC) but not in rough ER (rER). Inhibition of ER–Golgi transport by brefeldin A did not affect the association of K75 with Golgi, whereas ameloblastin accumulated in rER, and its transport from rER into Golgi was disrupted. Together, these results indicate that K75, a cytosolic protein lacking a signal sequence, is secreted into the forming enamel matrix utilizing portions of the conventional ER–Golgi secretory pathway. To the best of our knowledge, this is the first study providing insights into mechanisms of keratin secretion.

Keratins comprise a large and diverse family of epithelial cytoskeletal proteins that evolved from two ancestral genes during the expansion of tetrapods into the land environments (1). There are 54 keratins in humans, divided into two major groups, type I (acidic) and type II (basic), which assemble into heterodimers (2). Based on their location and function, keratins are further divided into epithelial, hair follicle, and hair keratins (2). Keratins organize into heavily cross-linked networks of intermediate filaments, which provide mechanical rigidity to the cells and play important roles in cell-cell contacts as components of the desmosomal complexes (3,4). In addition to the major role in the cytoskeleton, recent studies suggest that keratins can have other important functions. Specifically, keratin 17 was detected in cell nuclei where it was shown to regulate gene expression (5). Hair follicle type II keratin 75 (K75 2 ; also known as K6hf) is expressed in ectodermal appendages of amniotes, such as hair and feather follicles, nail beds, and lingual papillae (6 -8). A single amino acid substitution, A161T, in this protein causes a common skin disorder, pseudofolliculitis barbae, colloquially called barber's itch (9). We recently demonstrated that K75 is present in dental enamel-forming cells, ameloblasts, and enamel (10). Moreover, it was found that dental enamel in individuals carrying the pseudofolliculitis barbaecausing polymorphism has altered structure and mechanical properties and is more susceptible to caries (10). Recently, a number of other epithelial keratins were discovered in enamel, and an association between mutations in these keratins and a higher susceptibility to caries has been established (11). Furthermore, in our recent study, K75 was identified in forming porcine enamel by Western blotting and mass spectrometry (MS) (12). Importantly, unlike other transient enamel matrix proteins (EMPs), K75 was not degraded by resident enamel proteases KLK4 and MMP20, strongly suggesting that it incorporates into mature extracellular enamel matrix (EEM).
Although hairs and teeth evolved independently, they share a number of regulatory pathways involved in their morphogenesis (13,14). For example, expression of K75 and several other keratins in ameloblasts, enamel-forming cells, is regulated by DLX3 transcription factor (10), which plays a major role in organogenesis of hairs and teeth (15)(16)(17). Mutations in DLX3 cause tricho-dento-osseus syndrome, manifested by severe hair and tooth defects. Specifically, enamel in tricho-dento-osseus syndrome patients lacks the rod pattern, which is the hallmark of mammalian enamel (18).
Mature enamel is the hardest tissue in the body; it comprises 95% carbonated hydroxyapatite mineral and ϳ1% organic matrix. The basic building blocks of enamel, enamel rods, are arrays of elongated crystals aligned with their crystallographic c axes (19 -21). The rods are woven into a very complex 3D structure, which is responsible for the unique mechanical resilience of this material (20,(22)(23)(24)(25). Although mature enamel is a highly mineralized material, it starts as a hydrogel containing arrays of tiny crystallites (26 -29). Enamel formation (amelogenesis) is carried out by specialized epithelial cells, ameloblasts, in a highly coordinated multistep process (30). Enamel matrix deposition begins at the surface of forming dentin by so-called presecretory ameloblasts. They differentiate into secretory ameloblasts, polarized cells with the nuclei located at the basal (proximal) pole, and the Tomes' processes, a specialized cellular secretory apparatus responsible for enamel rod formation, at the apical (distal) pole (31)(32)(33). The secretory EEM mainly consists of three proteins, amelogenin (AMELX) (34) comprising ϳ90% of the matrix, ameloblastin (AMBN) (35), and enamelin (ENAM) (36), and a metalloproteinase, MMP20 (37), which play important roles in the regulation of mineralization (38 -45). All EMPs possess a signal sequence and are secreted via the conventional ER-Golgi secretory pathway (46,47). Interestingly, several articles reported interactions between EMPs and cytosolic keratins, suggesting the presence of novel trafficking mechanisms in ameloblasts (48,49). Once the full thickness of enamel is deposited, the maturation stage begins, at which the majority of EMPs are removed and replaced by thickening enamel crystals (43). Only a small fraction of a heavily cross-linked fibrous material remains in the mature enamel (28). This material comprises rod sheaths, enveloping enamel rods, and enamel tufts, protein-rich structures in the inner enamel (50 -52) that contain K75 and other keratins (10,11).
Our discovery of K75 in enamel (10, 12) raises a number of questions regarding its roles in enamel formation and its functional properties. One question, with a potential impact beyond enamel research, is of particular interest, namely when and how this cytosolic protein lacking a lead sequence enters the extracellular space. There are only a few reports in the literature describing keratins outside cells (53,54), but these studies do not address the secretory pathway question. It is possible that K75 ends up in the enamel matrix as a part of cellular debris embedded in the matrix during enamel secretion (55). Alternatively, cytosolic proteins can be secreted via unconventional protein secretion mechanisms (56 -58), including Golgi bypass (56,59), secretory autophagy (60,61), secretory lysosomes (62), and endosomes (63). Conversely, studies show that a number of proteins lacking the signal sequence can enter the classical secretory pathway via posttranslational translocation into rough ER (rER) (64,65).
To gain insights into the mechanisms of K75 trafficking and secretion, we carried out a number of experiments to characterize the expression pattern and trafficking pathway of this protein during enamel formation. Importantly, unlike the majority of trafficking studies, we performed our experiments in vivo because ameloblasts, highly specialized polarized secretory cells, cannot maintain their differentiated phenotype in vitro.

K75 is expressed by secretory ameloblasts and is present in forming enamel
Taking advantage of the constantly growing murine incisors, containing all stages of enamel development (Fig. 1A), we were able to observe K75 expression pattern during amelogenesis and to compare it with AMELX and AMBN, two major EMPs with well-established expression profiles (46,66,67). Fig. 1B shows immunofluorescence (IF) images taken from the three boxed areas in Fig. 1A. The IF results indicate that only AMELX is expressed at the presecretory stage of amelogenesis, and all three proteins are expressed during the secretory stage ( Fig.  1B). At the transition between secretory and maturation stages, AMELX and K75 expression cease, whereas AMBN expression persists through the maturation stage (Fig. 1B). The expression profiles of AMELX and AMBN observed in this study are consistent with earlier findings (46,66,67). Importantly, our results clearly demonstrate that K75 protein is exclusively expressed during the secretory stage of amelogenesis. Higher magnification imaging of secretory-stage ameloblasts revealed that the K75 distribution in secretory ameloblasts was remarkably different from other typical cytokeratins, such as K14 (Fig. 1, C and  D). Although IF of K14 revealed a network of filaments throughout the ameloblast, K75 was found in large granular bodies located in the distal parts of ameloblasts and a distinct signal organized into a band in the central portion of the cytoplasm. Strong K75 signal was also present in Tomes' processes. In situ hybridization (ISH) studies revealed that Krt75 has a unique expression pattern in the craniofacial region. Its expression is confined to cells of hair follicles, lingual papillae, and enamel organ (Figs. 1E and S1, A and B). The presence of K75 in forming molar and incisal enamel from 14-day-old rats was also detected by Western blotting (WB) (Fig. 1F). A band around 70 kDa was observed in both molar and incisal enamel, and in the case of molar enamel, a higher-molecular-mass band between 110 and 160 kDa, likely corresponding to a K75 dimer with an acidic keratin, was detected (Fig. 1F). The fact that K75 is observed at a higher than nominal K75 molecular mass of 59 kDa suggests that it undergoes posttranslational modifications, potentially being glycosylated in forming enamel (68). The 70-kDa band was further analyzed using MS, which identified with 100% confidence five K75 fragments covering 8% of its sequence (Fig. S1C). Importantly no enamel matrix proteins were identified in this band. Overall, these results corroborate our earlier observation of K75 in enamel (10,12) and revealed that K75 expression is associated with secretory ameloblasts.

K75 is present in membrane-delineated vesicles in secretory ameloblasts
To better assess the intracellular localization of K75 in the secretory ameloblasts, immunogold transmission EM (IG-TEM) studies were carried out. Ameloblasts are polarized secretory cells with extensive rER and Golgi networks throughout the cell body (Figs. 2A and S2-S4). In addition to typical perinuclear rER and Golgi structures, a large cylindrical Golgi complex aligned with the cell axis exists in the midportion of an ameloblast (31, 33) (Figs. S2 and S3). The cis face of this central Golgi complex is mainly oriented toward the plasma membrane, and the trans face is mainly oriented toward the core of the cell, which is rich in vesicles. The space between the cis Golgi and plasma membrane is lined with the rER network. The very distal end of the ameloblast cell body, bordering the Tomes' process, is free of Golgi and rich in rER and secretory vesicles (33) (Fig. S4). Tomes' processes are free of rER and

Secretion of keratin 75 by ameloblasts
Golgi cisternae but are filled with small secretory granules and contain a tubulovesicular network (69,70) (Figs. 2D and S4). The IG-TEM analysis revealed K75 signal in the membranedelineated vesicles containing electron-dense material in the cell bodies and Tomes' processes (Figs. 2, B-D, S3, and S4). Two distinct groups of vesicles were identified based on their sizes. Large vesicles of various shapes (700 Ϯ 556 nm) were observed primarily in the apical region of the cell body (Figs. 2, A and B, and S3). Smaller circular vesicles were observed throughout the ameloblast body (89 Ϯ 20 nm) (Fig. 2, A and C, and S3) and in Tomes' process (102 Ϯ 35 nm) (Fig. 2, A and D). The differences between the two groups were not statistically significant, sug-gesting that they belong to the same population. Our observations indicate that the small vesicles are covered with one single membrane bilayer, and no membrane-delineated compartments were found within the vesicles, suggesting that these are not multivesicular bodies or autophagosomes. K75 signal was also detected in the EEM surrounding the Tomes' processes (Figs. 2D and 3, G and J). Our IG-TEM studies failed to identify association of K75 with tonofilaments, whereas other keratins, e.g. K14, were clearly associated with them (Fig. S5A). Furthermore, we were not able to identify this protein in multivesicular bodies. Collectively, these results suggest that K75 is translocated into vesicles in the cell bodies of ameloblasts and is exo-

Secretion of keratin 75 by ameloblasts
cytosed via Tomes' processes. To the best of our knowledge, this is the first observation of a keratin, a cytosolic protein lacking a signal sequence, in secretory vesicles.

K75 colocalized with two major EMPs, AMBN and AMELX, in secretory ameloblasts
The discovery of K75 in membrane-delineated secretory vesicles prompted us to carry out colocalization studies of K75 with AMBN and AMELX. IF revealed significant levels of colocalization of K75 with the EMPs in cell bodies and Tomes' processes (Fig. 3, A and B) as well as discrete structures, resembling strings of beads colocalizing with AMBN in the enamel rod sheaths (Fig. 3C). Importantly, all three proteins are present in structures resembling enamel tufts at the dentinoenamel junction in forming molars (Fig. 3B, white arrowheads).
A quantitative IF colocalization analysis in the bodies of secretory ameloblasts revealed that close to ϳ90% of K75 colocalizes with AMBN or AMELX, and ϳ50% of K75 overlaps with both EMPs (Fig. 3D). In contrast, only about 30% of either AMBN or AMELX signal overlapped with K75 ( Fig. S5, B, C, and F).
In addition to IF colocalization studies, a high-resolution IG-TEM analysis was also carried out. The results revealed that K75 colocalizes with AMBN and AMELX in small and large vesicles in the bodies of ameloblasts and in the Tomes' processes (Figs. 3, E-J, S3, and S4). The quantitative analysis of small secretory vesicles in the Tomes' processes revealed that

Secretion of keratin 75 by ameloblasts
the colocalization of K75 with AMBN as well as with AMELX is greater than 90%. Similarly, more than 90% of the vesicles containing either AMBN or AMELX contained K75 (Fig. S5, D, E,  and G). Overall, the results of the IF and IG-TEM colocalization studies strongly indicate that K75, despite the lack of a signal sequence, enters the main secretory pathway and is cotransported with the EMPs.

K75 colocalizes with ERGIC and Golgi markers
To investigate the trafficking route of K75 in secretory ameloblasts, IF colocalization of K75 and markers for rER (cal-reticulin peptide 1 (CP1)), ERGIC (ERGIC53; also known as LMAN1), Golgi (Golga 5 (G5)), and lysosomes (LAMP1) was studied. In parallel, colocalization experiments with the same markers were conducted with AMBN, which is secreted via the classical secretory mechanism (47). K75 and AMBN had low levels of colocalization with rER (Fig. 4, A and B). However, a significant degree of overlap for both AMBN and K75 was observed with ERGIC and Golgi (Figs. 4, E and F, and S6). In the ameloblasts cell bodies, ERGIC signal was present in large granules, which were often attached to the cis Golgi face (Figs. S3 and S6). Interestingly, a strong ERGIC signal overlapping with  K75 and AMBN was present in the apical portions of ameloblasts and in the Tomes' processes (Fig. 4, E and F).

ER-Golgi trafficking inhibition using brefeldin A (BFA) does not affect colocalization of K75 with Golgi but significantly reduces its colocalization with ERGIC
To further assess the details of the K75 secretory route, we conducted trafficking inhibition experiments using BFA. BFA, a potent inhibitor of rER-Golgi transport, is widely used in studies of protein trafficking (46,71) . Mice were injected with BFA dissolved in ethanol. No obvious changes were found between the 1-h ethanol vehicle group, used as controls, when compared with the nontreated specimens (Fig. 5, A-D, versus Fig. 4, A-D). A quantitative colocalization analysis of secretory ameloblasts in the vehicle group showed that 14 Ϯ 2.5% of AMBN and 6 Ϯ 1% of K75 colocalized with rER marker CP1 (Fig. 5, A, B, and I). Although no significant overlap between K75 and CP1 is expected because K75 lacks the signal peptide, low colocalization levels of AMBN and CP1 can be due to a very short residence time of AMBN molecules in rER as previously reported (46). In contrast, substantial levels of colocalization of K75 (54 Ϯ 16%) and AMBN (45.3 Ϯ 8%) with Golgi marker G5 were found (Fig. 5, C, D, and J). In the BFA-treated group, 1 h after injection, the colocalization pattern of AMBN changed dramatically. Specifically, AMBN signal associated with the central Golgi complex significantly decreased (8.5 Ϯ 2%; p Ͻ 0.01), whereas its overlap with the rER significantly increased (30 Ϯ 5%; p Ͻ 0.01) (Fig. 5, E, G, I, and J). These results demonstrate that the disruption of rER-Golgi transport by BFA leads to the accumulation of AMBN in rER and its depletion from Golgi, which is anticipated for an extracellular protein secreted via the classical rER-Golgi-dependent pathway (57,72). The treatment with BFA, however, had a more limited effect on the localization of K75 in ER and Golgi. Colocalization of K75 and CP1 in the treatment group remained at low levels comparable with the control (10 Ϯ 1%) (Fig. 5, F and I), whereas the quantitative analysis revealed no significant changes in colocalization of K75 and G5 between the control and treatment groups (41 Ϯ 14 versus 54 Ϯ 16%; p ϭ 0.34) (Fig. 5, H and J). Colocalization studies using antibody against another Golgi marker, GM130, yielded similar results; namely the BFA treatment led to a dramatic decrease in overlap of AMBN and GM130, whereas the colocalization of K75 and GM130 remained essentially unchanged (Fig. S7).
The distribution of ERGIC53 and its colocalization with K75 and AMBN in the vehicle controls were similar to untreated samples (Figs. 4, E and F, and 6, A and B). After 1-h treatment with BFA, large granules positive for ERGIC53 disappeared, and instead a diffuse ERGIC53 signal was detected throughout the ameloblast body (Figs. 4, E and F, and 6, A and B, versus Figs.  6, C and D, and S8). The treatment also led to a significant decrease of ERGIC53 signal in the Tomes' processes. These observations prompted us to conduct quantitative analysis of The G5 signal appears as two well-organized long strips roughly parallel to each other, flanking the core of the ameloblast body. Both AMBN and K75 signals organized into bands adjacent to the two strips of Golgi complexes in the ameloblast core, partially overlapping with G5. Furthermore, large AMBN-and K75-positive granules were closely associated with the Golgi cisternae, potentially translocating their contents into Golgi for further processing. E and F, overlap with ERGIC marker ERGIC53. The ERGIC53 signal was associated with large granules in the ameloblast cell bodies where it colocalized with AMBN and K75. The ERGIC53 signal was also found in the distal portions of ameloblasts and Tomes' processes. Prominent overlap of K75 and AMBN with ERGIC53 was evident. G and H, overlap with lysosome marker LAMP1. LAMP1 signal was present throughout ameloblast cell bodies in the form of granules of different shapes and sizes. There was a very limited overlap of K75 with LAMP1, whereas AMBN was more prominently associated with LAMP1. Scale bar, 10 m. AB, ameloblast; SI, stratum intermedium; TP, Tomes' process.

Secretion of keratin 75 by ameloblasts
ERGIC53 colocalization with K75 and AMBN after BFA treatment. The quantitative colocalization analysis revealed significant reduction in K75 colocalization with ERGIC53 (29.5 Ϯ 5.8 versus 9.9 Ϯ 1.6%; p ϭ 0.003) (Fig. 6E). At the same time, the colocalization of AMBN and ERGIC53 remained unchanged (21.8 Ϯ 1.1% in vehicle control versus 26.5 Ϯ 6.3% for BFA treatment; p ϭ 0.5) (Fig. 6E). To rule out potential effects of the carrier (ethanol), we conducted similar experiments using PBS as the carrier, which lead to similar results but in a shorter time frame, possibly due to a slower degradation of BFA when using ethanol as a solvent (Fig. S9).
TEM analysis of the animals treated with BFA in ethanol revealed dilation of the rER, further confirming disruption of rER-Golgi trafficking (Fig. S10). IG-TEM of the control group detected AMBN in small vesicles, larger compartments of irregular shapes, potentially ERGIC, and Golgi apparatus,

Secretion of keratin 75 by ameloblasts
whereas no AMBN signal was detected in the rER (Fig. S10). After the treatment with BFA, we observed the accumulation of AMBN signal in rER; however, no signal was observed in the Golgi apparatus (Fig. S10), consistent with the results of IF studies. Overall, the results of BFA inhibition experiments indicate that although AMBN trafficking from rER to Golgi was blocked, as anticipated for a typical secreted protein, the BFA treatment did not affect association of K75 with Golgi.

Discussion
The presence of heavily cross-linked keratin matrix in enamel was first proposed in the late 19th century, and a num-ber of more recent papers suggested the presence of keratins in enamel (52). However, unequivocal evidence demonstrating the presence of keratin in enamel was lacking. Our recent studies, for the first time, demonstrate the presence of several keratins, including K75, in enamel matrix and that mutations in these genes cause structural and functional defects of enamel (10 -12, 73). The present results further corroborate our earlier observations of K75 in enamel (10,12) and revealed that K75 expression is associated with secretory-stage ameloblasts.
One of the arguments against the presence of keratins in enamel matrix was the fact that these are cytosolic proteins that lack a signal sequence and therefore cannot be secreted. There are several potential scenarios of translocation of cytosolic proteins into the extracellular space: they can be dislodged into the matrix as a part of cellular debris (55), or they can be secreted via the canonical ER-Golgi pathway (64,65) or by a number unconventional protein secretion mechanisms (56). Our colocalization studies demonstrate that secretion of K75, at least in part, utilizes the canonical secretory apparatus. Specifically, K75 is present in secretory vesicles together with two major EMPs, AMELX and AMBN, which possess a signal sequence and are secreted via the classical secretory pathway (46,47). K75 was also found in ERGIC and Golgi apparatus, cellular compartments associated with the classical secretory pathway. To the best of our knowledge, there are only a few other examples of cytosolic proteins lacking a signal peptide that utilize the conventional secretory pathway (64,65,74,75). Although there were a couple of reports showing keratins extracellularly (53,54), the mechanisms by which these keratin proteins enter the extracellular space were not previously studied. Our study is the first to demonstrate that K75 is cosecreted with AMBN and AMELX and that it utilizes, at least partially, the canonical ER-Golgi secretory pathway.
The fact that K75 colocalizes with AMBN and AMELX in secretory vesicles and the enamel matrix opens the possibility of interactions between K75 and EMPs. Ravindranath et al. (48,49) reported interactions between AMELX and K5 and K14. Further studies are needed to clarify this question. In another study using a yeast-two hybrid system, it was found that tuftelin, another protein expressed by ameloblasts, interacts with K6 and K5 (76).
ER-Golgi trafficking inhibition experiments using BFA provide further insights into the secretory pathway of K75 in ameloblasts. Although BFA treatment, which disrupts rER-Golgi transport, led to the accumulation of AMBN in dilated rER and its disappearance from Golgi as described previously (46), BFA treatment did not significantly affect the colocalization of K75 with Golgi. At the same time, even though ERGIC53 signal distribution in the cells changed, the degree of AMBN colocalization with ERGIC53 remained unchanged, whereas colocalization of K75 and ERGIC53 decreased significantly. Hence, although K75 utilizes portions of the canonical ER-Golgi secretory pathway, it behaves differently than typical secretory proteins containing a signal sequence. According to our model, K75 enters the classical secretory pathway not via rER, as do most secreted proteins, but is translocated via ERGIC. Once in the ERGIC it is cotransported with other secretory cargo through the Golgi apparatus and secreted from the

Secretion of keratin 75 by ameloblasts
Tomes' processes. At the same time, the fact that the BFA treatment, which leads to breakdown of ERGIC, does not significantly affect K75 colocalization with Golgi apparatus suggests that either BFA treatment prevents K75 from clearing Golgi or there is another ERGIC-independent mechanism of K75 translocation into the Golgi. To the best of our knowledge, this is the first demonstration of a trafficking route for a cytosolic protein that utilizes portions of the conventional rER-Golgi secretory pathway, and it is very different from other unconventional pathways utilized by cytosolic proteins (57,58).
Our observation of ERGIC53 in association with EMPs, in a distal rER-rich region of ameloblast and in the Tomes' processes, the ameloblast secretory apparatus, raises the possibility that some elements of the ERGIC system can be involved in a secretory pathway that bypasses Golgi. Earlier studies of ameloblasts revealed an extensive tubulovesicular system in the distal portions of ameloblasts and the Tomes' processes (69,70) that is potentially a part of ERGIC. One possible scenario is that some of these ERGIC structures are produced in the distal rERrich zone (Figs. S4 and S6) and transported into the Tomes' process directly. There are several examples of Golgi bypass (56,57,59); however, it is possible that, in ameloblasts, secreted proteins can simultaneously be transported via the conventional rER-Golgi pathway and through the Golgi bypass mechanism (59) involving ERGIC.
It is important to emphasize that ameloblasts are highly specialized supersecretory cells, and their secretory pathways are best studied in vivo. The fact that in tissue cultures ameloblastderived cells never attain their secretory phenotype presents a major challenge for studies of ameloblast cell biology. This constraint potentially applies to other cell types. The vast majority of cell trafficking and secretion studies are conducted in tissue culture systems, and often the data obtained in vitro are not in a good agreement with in vivo observations (57), potentially because cells in vitro are not able to differentiate properly. We anticipate that the mechanisms of secretion revealed in this study are not unique to ameloblasts but are much more universal, and other cell types with high secretory activity might utilize them as well.

Conclusions
Overall, these results further support our previous observations of K75 in ameloblasts and enamel and demonstrate that it is expressed during the secretory stage of amelogenesis. Furthermore, our data demonstrate that the majority of K75 is cotransported together with EMPs and is secreted from the Tomes' processes. Using cell compartment markers, we detected K75 in ERGIC and Golgi apparatus. We further conducted experiments with the ER-Golgi trafficking inhibitor BFA, which revealed that although the treatment disrupted colocalization of K75 with ERGIC, its association with Golgi remained unchanged. Collectively, these results demonstrate that K75, a cytosolic protein lacking a signal sequence, utilizes portions of the classical secretory pathway. Our results also suggest that some of the secreted proteins might utilize the Golgi bypass mechanism via ERGIC-associated vesicles. Together, these results provide novel insights into the K75 secretion processes and cellular trafficking mechanisms in ameloblasts and potentially other systems.

Experimental procedures
All methods were performed in accordance with the relevant guidelines and regulations. Four-week-old Sprague-Dawley rats (Charles River Laboratories, Wilmington, MA) and 4-week-old mice (C57BL/6J, The Jackson Laboratory, Bar Harbor, ME) were used in the study. All animal procedures described were approved by the University of Pittsburgh Institutional Animal Care and Use Committee.

Tissue preparation for ISH and IF and IG TEM studies
Four-week-old Sprague-Dawley rats (Charles River Laboratories) and 4-week old mice (C57BL/6J, The Jackson Laboratory) were euthanized with CO 2 . Mandibles were dissected and immediately submersed in 50-ml volumes of 4% paraformaldehyde in 10 mM PBS for IF or TEM or Karnovsky fixative (2% glutaraldehyde, 2% formaldehyde in 10 mM PBS) for TEM at 4°C in less than 3 min after euthanasia. After 24 -48 h of fixation at 4°C, samples were placed into demineralization solution containing 0.1 M EDTA (pH 7.2-7.4) for 1-2 weeks. The demineralization solution was changed every other day.
For IG-TEM experiments, animals were anesthetized with isoflurane and perfused through the left ventricle first with cold PBS for 30 s and then with cold 4% paraformaldehyde in 0.1 M phosphate buffer or cold 1% glutaraldehyde in 0.1 M phosphate buffer for 15-20 min. The mandibles were dissected and further fixed in the same fixative solution for another 8 -12 h at 4°C followed by demineralization as described above. For IF, whole demineralized jaws were dehydrated using a Leica ASP 300S automatic processer (Leica Biosystems, Buffalo Grove, IL) and paraffin-embedded using a standard protocol (77).
For TEM studies, after demineralization mandibular bone around the incisors was trimmed, and the molars were removed. The distal fragments of the jaws containing apical portions of the incisors were further cross-sectioned into 1-1.5-mm-thick pieces, processed, and embedded in LR White or Embed 812 (catalog numbers 14381 and 14120, Electron Microscopy Sciences, Hatfield, PA) according to published protocols (47). In brief, for Embed 812 processing, incisor pieces were postfixed in 1% ferrocyanide reduced osmium tetroxide for 1 h, washed in PBS, dehydrated in graded ethanol, and infiltrated with propylene oxide. The samples were embedded in Embed 812 and cured at 65°C for 2 days. For LR White processing, some of the samples were postfixed in osmium tetroxide, whereas others were not as osmication reduces the immunoreactivity of the samples. Incisor pieces were washed in PBS, dehydrated in graded ethanol, embedded in LR White, and cured at 60°C for 1-2 days.
For ISH, mice at postnatal day 1 were euthanized with CO 2 , and samples were fixed with 4% paraformaldehyde in 10 mM PBS for 24 h and embedded in paraffin as described above. For IF and ISH, the paraffin blocks were sectioned using a Leica RM 2225 microtome (Leica Biosystems, Buffalo, IL) into 10-mthick sections using a stainless steel microtome knife (Leica

Secretion of keratin 75 by ameloblasts
818, Leica, Germany). The sections were mounted on 3triethoxysilylpropylamine (440140, Millipore-Sigma)-coated glass slides. For TEM, the resin blocks were sectioned into 70-nm-thick sections using a Leica EM UC7 ultramicrotome (Leica Biosystems) equipped with a diamond knife (Electron Microscopy Sciences). The sections were mounted on carbon coated Ni grids (Electron Microscopy Sciences).

IF studies
An optimized IF procedure (77,78) developed in our laboratory, aimed at reducing false positives, was used. Briefly, after deparaffinization, sections underwent trypsin treatment for 5-30 min at 37°C or heat treatment in 10 mM sodium citrate buffer (pH 6.0) for 10 -20 min for antigen retrieval. The sections were blocked in 10% serum from the secondary Ab host animal in Tris-buffered saline (TBS) for 1 h at room temperature and incubated with primary Abs at 4°C overnight followed by incubation with secondary Abs for 45 min at room temperature. The sections were washed with TBS, incubated with 1.5% Sudan Black B (199664, Sigma) in 70% ethanol to minimize autofluorescence, and counterstained with 4Ј,6-diamidino-2phenylindole (DAPI).

In situ hybridization
ISH was performed following a published protocol (80,81). Briefly, the sections of P1 mice were deparaffinized, rehydrated, postfixed, and subjected to prehybridization on the 1st day.
Digoxigenin-labeled RNA probes to the sense and antisense strands of mouse Krt75 partial cDNA (421 bp) corresponding to exon 9, including the C-terminal tail domain and part of the 3Ј noncoding region (1494 -1996 bp of NCBI Reference Sequence NM_133357.3), were hybridized to the sections. The sections were blocked in 10% sheep serum for 2 h, incubated with 1:2000 anti-digoxigenin alkaline phosphatase Ab in 1% sheep serum mixture overnight at 4°C on the 2nd day, and washed and incubated in developing solution (1681451, Roche Applied Science) on the 3rd day. The reaction was stopped by washing in PBS. The slides were mounted in toluene (4112, Thermo Fisher) and observed under a light microscope in bright-field mode.

Western blotting
The enamel matrix was collected from apical portions of continually growing incisors and unerupted first molars from 14-day-old Wistar rats. For incisal enamel matrix preparation, mandibles of 14-day-old Wistar rats were dissected, and the basal bone was removed to expose unerupted portions of incisors under a dissecting microscope. The cellular components of enamel organ, including the ameloblast layer, were carefully removed, and the forming enamel tissue was scraped from the dentin surface. The matrix was transferred to TBS (Thermo Scientific) with proteinase inhibitor mixture (Thermo Scientific) at a 1:100 ratio. Molar enamel preparation was carried out according to published procedure (82). Maxillary and mandibular unerupted first molars were extracted, and pulp tissue was removed under a dissecting microscope. The crowns were cleaned of surrounding tissues, and enamel matrix proteins were extracted using a published method (82) with slight modifications. Briefly, the dissected molars were incubated with 0.17 M HCl, 0.98% formic acid for 2 h at 4°C, and the supernatant was concentrated using 3000-molecular-weight-cutoff centrifugal filtering unit (Millipore). Whiskers were processed according to a published procedure (83). Briefly, whiskers were plucked from the rat snout, and the hair follicles were collected in 200 l of protein extraction buffer with proteinase inhibitor mixture. The hair follicles were vortexed and incubated at 37°C overnight. For WB, the samples were centrifuged, and the supernatant was collected for analysis. The concentration of the proteins in the samples was detected using a BCA assay. Ten micrograms of proteins were loaded in a Bis-Tris 4 -12% gel (Life Technologies) and subjected to SDS-PAGE. The protein bands were transferred from the gel to a polyvinylidene difluoride membrane (Life Technologies), and bands containing proteins of interest were detected using ECL substrate (catalog number 32106, Thermo Scientific). Primary Abs used were anti-K75 guinea pig IgG antibodies at 1:1000 (20R-2647, Fitzgerald Industries International), rabbit anti-AMELX at 1:1000 (ABT260, Abcam), and rabbit anti-AMBN at 1:1000 (sc-50534, Santa Cruz Biotechnology). Secondary Abs used were donkey anti-guinea pig IgG conjugated with HRP at 1:2000 (706-035-148, Jackson ImmunoResearch Laboratories) and goat anti-rabbit IgG conjugated with HRP at 1:2000 (catalog number 32460, Thermo Scientific).

Mass spectrometry
For MS, the bands of interest were dissected from the gels and submitted to the University of Pittsburgh Biomedical Mass Spectrometry Center for analysis.

IG-TEM studies
The IG-TEM studies were performed on 70-nm-thick sections mounted on carbon-coated Ni grids. Immunolabeling was performed according to published procedures (46,47). Briefly, grids were blocked in the same blocking solution as in the IF protocol described above for 1 h, incubated with primary antibody for 1 h, washed with TBS, then incubated with secondary antibody for 1 h, washed, and counterstained with 2% uranyl acetate for 10 min. Primary Ab concentrations 10 times higher than in the IF assays described above were used. Six-nanometer colloidal gold-conjugated donkey anti-guinea pig IgG at 1:30 (706-195-148, Jackson ImmunoResearch Laboratories) and 12-nm colloidal gold-conjugated donkey anti-rabbit IgG at 1:30 (711-205152, Jackson ImmunoResearch Laboratories) were used as secondary Abs. Guinea pig serum (88R-1015, Fitzgerald Industries International) and rabbit serum (011-000-001, Jackson ImmunoResearch Laboratories) were used as isotype controls. All grids were observed with JEM 1400Plus TEM (JEOL USA Inc.) at 80 kV, located in the Center for Biologic Imaging at University of Pittsburgh. To assess the distribution of K75 in vesicles of different sizes in the ameloblast cell bodies and the Tomes' processes, sections from three rats were labeled with anti-K75 Abs and analyzed. The vesicles containing at least one gold particle were counted and measured, and at least 50 vesicles per animal were analyzed. For IG-TEM colocalization of K75 with EMP in Tomes' processes, LR White sections from three rats were double-labeled for K75 and either AMBN or AMELX. Vesicles with at least one gold particle were counted, and more than 400 vesicles were analyzed.

Brefeldin A experiments
BFA powder (B-8500, LC Laboratories, Woburn, MA) dissolved in DMSO at 25 mg/ml concentration was used as a stock solution. BFA stock solution was further dissolved in 30% ethanol or PBS to a final concentration of 0.5 mg/ml prior to the experiments. Four-week-old mice were used in this study; all animals were anesthetized by isoflurane. Three mice received intraperitoneal injection of 0.5 mg of BFA in ethanol, and three mice injected with ethanol only were used as vehicle control. The mice were euthanized 1 h after the injection. In another set of experiments, three mice were injected with 0.5 mg of BFA in PBS. These animals were euthanized 15, 30, and 60 min after the injection. One mouse was injected with the vehicle (control) and euthanized 60 min after the injection. The animals were euthanized in CO 2 1 h after the injection, and samples were fixed in 4% paraformaldehyde according to the IF or TEM procedures as described above. An independent two-sample t test was used to analyze the colocalization differences between the vehicle control and BFA/ethanol-treated mice (n ϭ 3).

Quantitative IF colocalization analysis
Quantitative colocalization analysis was performed using a well-established methodology (84,85). The colocalization anal-ysis was conducted using proprietary NIS-Elements software. The quantitative colocalization analysis was conducted on sections from three animals. Colocalization was calculated based on the number of pixels in confocal slices from a z-stack containing two channels. The colocalization was assessed according to established procedures (84 -86) using Pearson correlation analysis and Manders' coefficients, which represent a ratio of overlapping pixels to the total number of pixels in each channel. For background adjustment, adjacent sections were treated with naïve sera (isotype control), and signal intensity from these sections was used to determine the background fluorescence. The colocalization analysis was conducted using proprietary NIS-Elements software. The quantitative colocalization analysis was conducted on sections from three animals. The data were analyzed using a t test assuming unequal variance in the OriginPro 2015 software package (OriginLab Corp.).