PRAS40 Connects Microenvironmental Stress Signaling to Exosome-Mediated Secretion

ABSTRACT

Secreted exosomes carrying lipids, proteins, and nucleic acids conduct cell-cell communications within the microenvironment of both physiological and pathological conditions. Exosome secretion is triggered by extracellular or intracellular stress signals. Little is known, however, about the signal transduction between stress cues and exosome secretion. To identify the linker protein, we took advantage of a unique finding in human keratinocytes. In these cells, although transforming growth factor alpha (TGF-α) and epidermal growth factor (EGF) share the same EGF receptor and previously indistinguishable intracellular signaling networks, only TGF-α stimulation causes exosome-mediated secretion. However, deduction of EGF-activated pathways from TGFα-activated pathways in the same cells allowed us to identify the proline-rich Akt substrate of 40 kDa (PRAS40) as the unique downstream effector of TGF-α but not EGF signaling via threonine 308-phosphorylated Akt. PRAS40 knockdown (KD) or PRAS40 dominant-negative (DN) mutant overexpression blocks not only TGF-α- but also hypoxia- and H₂O₂-induced exosome secretion in a variety of normal and tumor cells. Site-directed mutagenesis and gene rescue studies show that Akt-mediated activation of PRAS40 via threonine 246 phosphorylation is both necessary and sufficient to cause exosome secretion without affecting the endoplasmic reticulum/Golgi pathway. Identification of PRAS40 as a linker protein paves the way for understanding how stress regulates exosome secretion under pathophysiological conditions.

INTRODUCTION

Secretion of extracellular vesicles (EVs) by cells under stress is an evolutionarily conserved phenomenon found in almost all types of cells and biological fluids. EVs have several synonyms, such as microvesicles, ectosomes, microparticles, and exosomes, which are widely used in independent studies (1, 2). The biologic origin and size differences account for the main distinctions made among these EVs. Exosomes belong to a subtype of EVs with loosely defined diameters between 30 and 150 nm and are derived from intraluminal vesicles (ILVs) within intracellular multivesicular bodies (MVB) (3, 4, 5). Nonetheless, due to technical limitations in purifying EV populations, the currently used term, exosomes, refers to a population of EVs of various sizes but with the majority being between 30 and 150 nm in diameter, instead of a single subtype of EV with a clearly defined population in size and origin of production (6).

Normal cells secrete exosomes under extracellular environmental stress, whereas tumor cells constitutively secrete exosomes driven by intracellular oncogenes (7). In response to environmental or oncogenic stress cues, the exosome-containing MVB or MVB-derived exosomes themselves directly fuse with the plasma membrane to release exosomes to the extracellular environment. The secreted exosomes can transfer their cargo molecules, including DNAs, mRNAs, microRNAs, lipids, and proteins, to other nearby cells of various cell types and affect the biological behavior of these target cells (1, 8, 9, 10). This new and increasingly recognized mechanism of intercellular communication has been demonstrated to play critical roles in host immune responses (11), tissue repair (1, 12), and tumor invasion and metastasis (7, 13, 14, 15).

Despite the importance of secreted exosomes, little is known about the regulation of exosome secretion by microenvironmental stress. The Rab27 small GTPases, including Rab27a and Rab27b, have been reported to regulate exosome biogenesis and secretion. Studies showed that Rab27a and Rab27b regulate distinct steps of multivesicular endosome (MVE) docking to the plasma membrane and exosome biogenesis, in which Rab27a regulates MVE breakdown and Rab27b regulates MVE distribution, formation, and secretion of exosomes in various types of cells (16, 17). On the other hand, Rab27 proteins do not appear to be strictly specific regulators of exosomes, since Rab27a also regulates secretion of MMP9 and growth factors through the conventional endoplasmic reticulum (ER)/Golgi pathway (13, 18). Sinha and colleagues showed that knockdown or overexpression of cortactin resulted in a respective decrease or increase in exosome secretion without altering exosome cargo content in cancer cells. They proposed that cortactin promotes exosome secretion via binding to Arp2/3 and stabilizing cortical actin-rich MVE docking sites (19). Nevertheless, it is not known if Rab27 or cortactin serves as a linker molecule that connect microenvironmental stress signals to exosome-mediated secretion. An interesting study by Wang and colleagues reported that hypoxia increases, via HIF-1α and HIF-2α, microvesicle formation by transcriptional activation of the Rab22A gene in several breast cancer cell lines (61), although it would be important to know whether HIF-1α/HIF2α directly or indirectly regulates Rab22A expression.

The proline-rich Akt substrate of 40 kDa (PRAS40) was initially identified as a direct substrate of Akt kinase and a binding partner for the 14-3-3 scaffolding molecule (20). Most studies focused on PRAS40's role in insulin, as well as nerve growth factor (NGF) and platelet-derived growth factor (PDGF), signaling to the mTOR (mammalian target of rapamycin) pathway (specifically mTORC1), which regulates cell metabolism, protein synthesis, and cell growth (21–28). In growth-arrested cells, PRAS40 was reported to bind, via the raptor subunit, to mTORC1 and inhibit mTOR kinase activity. Insulin stimulation activates Akt kinase mainly via threonine 308 (Thr-308) phosphorylation. The activated Akt kinase in turn phosphorylates PRAS40 on Thr-246. Thr-246-phosphorylated PRAS40 dissociates from mTORC1, resulting in activation of mTORC1, and (re)associates with 14-3-3 (23, 25, 26, 29). In addition to Akt, increased PIM1 kinase activity also correlated with increased PRAS40 phosphorylation. Activated mTORC1 phosphorylates PRAS40 at Ser-183, Ser-212, and Ser-221 (30, 31). Despite these reports, other studies suggest that PRAS40 is not a common regulator of mTOR activation in response to different extracellular signals (27, 32).

In this current study, we seized upon a unique property of human keratinocytes in culture and found a critical signaling molecule that connects microenvironmental cues to exosome secretion. While both transforming growth factor alpha (TGF-α) and epidermal growth factor (EGF) are known to utilize the same cell surface EGF receptor (EGFR) for transmembrane signaling and previously indistinguishable intracellular signaling networks, surprisingly we found that only TGF-α triggers secretion of heat shock protein 90 alpha (Hsp90α), a known exosome cargo protein. By comparing and deducting 43 intracellular signaling molecules/pathways in the same cells in response to TGF-α or EGF stimulation, we identified PRAS40 as a TGF-α-specific downstream target. We found that activated PRAS40 acts not only as a regulator of TGF-α-triggered exosome secretion but also as a common regulator of distinct microenvironmental and oncogenic signal-triggered exosome secretion in both normal and tumor cell types. PRAS40 is the first regulator identified for stress-induced exosome secretion.

RESULTS

TGF-α, but not EGF, selectively induces secretion of exosome cargo protein, Hsp90α, in human keratinocytes.

In wounded skin, the TGF-α levels rise from undetectable to ∼40 ng/ml and stimulate keratinocytes at the wound edge of epidermis to secrete Hsp90α, a known exosome cargo protein, presumably for promoting wound closure (33, 34, 35). Therefore, understanding how TGF-α regulates exosome-mediated Hsp90α secretion would allow for the identification of the linker molecule that connects extracellular environmental signals to the exosome trafficking pathway. Experimentally, we took advantage of a unique observation: while both TGF-α and EGF share the same cell surface receptor, EGFR, in primary human keratinocytes, only TGF-α stimulation induces Hsp90α secretion. As shown in Fig. 1A, secreted Hsp90α was detected by Western immunoblotting in the conditioned medium only from TGF-α-stimulated but not EGF-stimulated keratinocytes (row a, lane 3 versus lane 2). Accordingly, we detected a corresponding decrease (∼20% as measured by densitometry scanning) from the cytosolic pool of Hsp90α in the TGF-α-stimulated cells (row b, lane 3 versus lane 1). We confirmed that TGF-α-stimulated Hsp90α secretion is mediated by EGFR, since downregulation of EGFR (Fig. 1B, row d, lane 2 versus lane 1) completely blocked TGF-α-stimulated Hsp90α secretion (Fig. 1C, row f, lane 5 versus lane 2). The TGF-α signaling that leads to Hsp90α secretion is independent of de novo protein synthesis, since treatment of the cells with cycloheximide (CHX) did not affect the TGF-α-induced Hsp90α secretion (Fig. 1D, row g, lane 4 versus lane 3), whereas the CHX treatment completely blocked TGF-α-induced cyclin D1 expression (row h, lane 4 versus lane 3). These findings provided us with an opportunity to search for the direct downstream effector of TGF-α signaling that regulates Hsp90α secretion. As shown schematically in Fig. 1E, our approach was to deduct EGF-activated pathways from TGF-α-activated pathways in the same cells and identify the unique downstream target that is specific for TGF-α signaling. The question mark represents the putative unknown signaling molecule that mediates TGF-α signaling to Hsp90α secretion.

FIG 1.

TGF-α, but not EGF, stimulates Hsp90α secretion via EGFR in human keratinocytes. (A) Serum-starved primary human keratinocytes were either left untreated (−) or treated with TGF-α (20 ng/ml) or EGF (20 ng/ml) for 12 h. Conditioned media were collected for secreted molecules and Triton X-100 soluble extracts for total cytoplasmic proteins. Both fractions were subjected to immunoblot analyses with antibodies against the indicated target molecules. Densitometry scanning was performed on intracellular Hsp90α against GAPDH as a background control (rows b and c). (B) EGFR downregulation by lentiviral vector delivery of a control and anti-EGFR shRNA (row d, lane 2). (C) Downregulation of EGFR blocks TGF-α-stimulated Hsp90α secretion (row f, lane 5 versus lane 2). (D) Cycloheximide (CHX) blocks TGF-α-stimulated cyclin D1 induction (row h) but not Hsp90α secretion (row g). (E) A schematic interpretation of the results from panels A to D. The question mark represents the possible signaling molecule downstream of TGF-α but not EGF signaling. (F) TGF-α and EGF induce indistinguishable patterns of total protein tyrosine phosphorylation. (G) Both stocks of TGF-α and EGF are functional for activating ERK1/2 (row j). These results were reproducible in at least three repeated experiments (n > 3).

We first compared global protein tyrosine phosphorylation, since EGFR is a receptor tyrosine kinase. As shown in Fig. 1F, we obtained indistinguishable patterns between EGF- and TGF-α-induced tyrosine phosphorylation (lanes 2 and 3 versus 1), suggesting that the effector connecting TGF-α signaling to Hsp90α secretion is not among phosphotyrosine proteins. In this experiment, both TGF-α and EGF were fully functional, since both growth factors potently stimulated the activation of extracellular signal-related kinase 1/2 (ERK1/2) in the same cells (Fig. 1G, row j, lanes 1 to 8 versus lane 9). We concluded, therefore, that a broader search of the signaling networks was needed.

Identification of PRAS40 that mediates TGF-α signaling to Hsp90α secretion.

We subjected the lysates of TGF-α- or EGF-stimulated human keratinocytes to a global pathway screening with a human phosphokinase array (ARY003B; R&D Systems) that allows simultaneous detections of 43 independent signaling pathways, as schematically shown in Fig. 2A (see Materials and Methods for details). Our focus was on the early (minutes) events and de novo protein synthesis-independent (TGF-α and EGF) signaling events, as schematically shown in Fig. 2B, in which EGF and TGF-α stimulation was restricted to 2 min, when the highest level of EGFR activation in human keratinocytes is detected (36). The results of a representative experiment are shown in Fig. 2C. Within 2 min of stimulation, we detected five known molecules that exhibited increased phosphorylation by TGF-α stimulation, including PRAS40 (circle 1), EGFR (circle 2), ERK1/2 (circle 3), Akt (Thr-308) (circle 4), and ribosomal S6 kinase (RSK) (circle 6). One increased protein level, that for Hsp60 (circle 5), was ruled out from our search. Phosphorylations of EGFR, ERK1/2, and RSK were induced by both EGF and TGF-α and therefore were excluded from further consideration. We focused on the increased phosphorylation of PRAS40 at Thr-246 and phosphorylation of Akt at Thr-308, since they were detected only in TGF-α-stimulated (row b) but not in EGF-stimulated (row c) cells. EGF stimulation induced only Ser-473 phosphorylation of Akt in the same cells (see the next section).

FIG 2.

PRAS40 is the possible downstream effector of TGF-α but not EGF signaling by array screening. Serum-starved primary human keratinocytes were either left untreated (−) or treated with TGF-α (100 ng/ml) or EGF (100 ng/ml) for 2 min (5- to 10-fold-higher concentration of growth factor for shorter time stimulation to detect signaling molecules). The cell extracts were incubated with human phosphokinase array membranes from R&D Systems, and signals were detected according to the manufacturer's instructions (see Materials and Methods). (A) A schematic representation of the 43 pathways included in the array. (B) Experimental design and sequential steps. (C) ECL results of the membranes show major pathways activated by TGF-α, EGF, or both under the stimulation time and conditions.

We further verified the findings by Western immunoblot analysis. The time course of phosphorylation of Akt (Thr-308), Akt (Ser-473), PRAS40 (Thr-246), and RSK in human keratinocytes in response to TGF-α and EGF stimulation is shown in Fig. 3A. TGF-α induced a time-dependent phosphorylation of PRAS40 on Thr-246 (row a, lanes 2 to 5 versus lane 1), while EGF did not (lanes 6 to 9 versus lane 1). Both TGF-α and EGF stimulated phosphorylation of Akt on Ser-473 (row c), but interestingly, only TGF-α stimulated phosphorylation of Akt on Thr-308 (row b, lanes 2 to 5 versus lanes 6 to 9). These findings are important, since a previous report showed that only the Thr-308-phosphorylated Akt is the upstream kinase that phosphorylates Thr-246 in PRAS40 (39). Under the same conditions, TGF-α and EGF equally stimulated RSK phosphorylation (row d, lanes 3 to 5 versus lanes 7 to 9). Three corresponding protein-loading controls support the above-described conclusions (rows e to g). As expected, the TGF-α-induced phosphorylation of PRAS40 requires EGFR, since downregulation of EGFR completely blocked TGF-α-stimulated PRAS40 phosphorylation (Fig. 3B, row h, lane 5 versus lane 2).

FIG 3.

PRAS40 mediates connection of TGF-α signaling to Hsp90α secretion. (A) Serum-starved primary human keratinocytes were either left untreated (−) or treated with TGF-α (100 ng/ml) or EGF (100 ng/ml) for the indicated time. Total lysates were subjected to Western blotting with antibodies against the activated signaling molecules or controls. (B) TGF-α, but not EGF, stimulates phosphorylation of PRAS40 via EGFR. (C) A schematic representation of a hypothesis that PRAS40 mediates the TGF-α signaling leading to Hsp90α secretion. (D) PI3K inhibitor blocks TGF-α signaling regarding Hsp90α secretion. (E) Lentiviral infection (pHR-CMV-puro RNAi delivery system)-mediated downregulation of endogenous PRAS40. (F) Downregulation of PRAS40 blocks TGF-α-stimulated Hsp90α secretion (lane 5 versus lane 2). (G) Effect of downregulation of PRAS40 on cell survival. (H) Effect of downregulation of PRAS40 on cell motility. These experiments were repeated 3 times by different laboratory members on the author list.

The above-described findings pointed to PRAS40 as the potential linker molecule between TGF-α signaling, via pAkt-T308, and Hsp90α secretion, as schematically depicted in Fig. 3C. To test this possibility, as shown in Fig. 3D, we found that inhibition of an Akt upstream kinase, phosphatidylinositol 3-kinase (PI3K), by LY294002 completely blocked TGF-α-stimulated Hsp90α secretion (row j, lane 4 versus lane 2). Second, we used the lentiviral pHR-CMV-puro RNA interference (RNAi) delivery system (38) to deliver a short hairpin RNA (shRNA) against the 3′ untranslated region (UTR) of the human PRAS40 gene and achieved nearly complete downregulation of the endogenous PRAS40 protein in human keratinocytes following drug selection (Fig. 3E, row l, lane 2 versus lane 1). Under these conditions, TGF-α stimulation was no longer able to induce Hsp90α secretion (Fig. 3F, row n, lane 5 versus lane 2). Downregulation of PRAS40 affected neither cell morphology (Fig. 3G, row o versus row p) nor cell motility in response to serum (Fig. 3H, bars 3 and 4 versus bars 1 and 2). Intriguingly, the difference between TGF-α and EGF in promoting Hsp90α secretion was also detected in keratinocyte migration in response to these growth factors. TGF-α promotes an extra 10% more cell migration than EGF (bar 5 versus bar 7). PRAS40 downregulation selectively abolished this extra 10% enhancement of the cell motility by TGF-α (bar 6 versus bar 5) but had little effect on EGF-stimulated cell migration (bars 7 and 8). We found that this long-recognized extra 10% stimulation of cell motility by TGF-α was contributed by secreted Hsp90α, since addition of human recombinant Hsp90α recovered the extra 10% stimulation for TGF-α in PRAS40-downregulated cells (bar 10) and also synergized EGF stimulation to a level similar to that of TGF-α stimulation (bars 11 and 12).

More importantly, this new role for PRAS40 in exosome-mediated Hsp90α secretion was not restricted to growth factor-induced Hsp90α secretion. As schematically shown in Fig. 4A, we tested two additional common microenvironmental cues, hypoxia and oxidative stress. Under hypoxia (1% O₂), as shown in Fig. 4B, accumulation of the hypoxia-inducible factor 1alpha (HIF-1α) protein occurred in a time-dependent fashion, providing proof for hypoxia (row a). While the overall cellular level of PRAS40 remained unchanged (row b), hypoxia induced phosphorylation of PRAS40 on Thr-246 in a time-dependent manner, which declined after 6 h (row c). Oxidative stress, namely, H₂O₂ treatment, however, caused a sustained increase in phosphorylation on Thr-246 (Fig. 4C, row f), which was followed by increased PRAS40 protein levels at later times (row e). Nonetheless, downregulation of PRAS40 blocked both hypoxia-induced (Fig. 4D, row h, lane 4 versus lane 3) and H₂O₂-induced (Fig. 4E, row i, lane 4 versus lane 3) Hsp90α secretion. Finally, PRAS40 was also confirmed as a regulator of Hsp90α secretion via exosomes in a variety of cell types (see Fig. 7).

FIG 4.

PRAS40 connects different stress cues to Hsp90α secretion. (A) A schematic illustration of the model to be tested. (B) Serum-starved primary human keratinocytes were either left untreated (−) or treated with hypoxia (1% O₂) for the indicated time. Total cell lysates were immunoblotted with anti-HIF-1α (row a), anti-PRAS40 (row b), anti-phospho-PRAS40 (row c), and anti-GAPDH (row d) antibodies. (C) The same duplicate cells were either left untreated (−) or treated with H₂O₂ (30 μM). Total cell lysates were immunoblotted with anti-PRAS40 (row e), anti-phospho-PRAS40 (row f), and anti-GAPDH (row g) antibodies. (D) Conditioned media were collected from shLacZ- or shPRAS40-infected cells under either normoxia or hypoxia and subjected to immunoblotting with anti-Hsp90α antibody. (E) Conditioned media were collected from shLacZ- or shPRAS40-infected cells treated with H₂O₂ (in 3.3 μl) or the same volume of H₂O and subjected to immunoblotting with anti-Hsp90α antibody. The results were reproducible by two repetitive experiments.

FIG 7.

PRAS40 mediates exosome secretion in multiple cell types and in response to distinct stress cues. Downregulation of endogenous PRAS40 in primary mouse hepatocytes (A, row a), MLE (mouse lung epithelial) cells (B, row i), MDA-MB-231 breast cancer cells (C, row q), BJAB B cell lymphoma cells (D, row y), and SKNBE2 neuroblastoma cells (E, row d′). Hypoxia strongly induced secretion of Hsp90α (rows c and k), CD9 (rows d and l), and CD63 (rows e and m) in shLacZ-infected noncancer cells (lanes 3 versus lanes 1). In contrast, secretion of these markers was constitutive (lanes 3 versus lanes 1) in three cancer cell lines, MDA-MB-231 (rows s, t, and u), BJAB (rows a′, b′, and c′), and SKNBE2 (rows f′, g′, and h′). All of the secretion was blocked by downregulation of PRAS40 (lanes 2 and 4). Similarly, the H₂O₂ treatment greatly induced secretion of Hsp90α (rows f, n, and v), CD9 (rows g, o, and w), and CD63 (rows h, p, and x) in the control cells (lanes 3 versus lanes 1). Downregulation of PRAS40 blocked H₂O₂-induced secretion of all markers (lanes 2 versus lanes 4).

PRAS40 regulates exosome-mediated secretion in response to extracellular signals.

Secreted Hsp90α has been shown by proteomic analysis (60) and electron microscopy (40) to be present with secreted exosomes. We found that secreted Hsp90α was associated with the pellet fraction of exosomes and that little was left in the supernatant following centrifugation at 100,000 × g (Fig. 5A, lane 2 versus lane 1), although this finding did not exclude the possibility that some Hsp90α is associated with larger microvesicles. We therefore postulated that PRAS40 is, in fact, a regulator of stress-induced exosome secretion in which Hsp90α is just a cargo molecule. We utilized a sequential centrifugation technique to isolate the 100,000 × g pellet fractions from cell conditioned media, followed by nanoparticle tracking analysis (NTA), which measures the size and the amount of EVs, with the diameter range set for 10 to 1,000 nm in liquid suspension. As shown in Fig. 5B, the majority of the vesicles in the 100,000 × g pellets fell into the range of 30 to 150 nm in diameter, consistent with the size range for exosomes. In the control (shLacZ-infected) cells, TGF-α stimulated a dramatic increase in the quantity of secreted exosomes (peak 1 versus peak 2). However, the TGF-α-stimulated exosome secretion was completely blocked in the PRAS40-downregulated cells (peak 4). PRAS40 downregulation even eliminated the basal level of exosome secretion from the unstimulated parental cells (peak 3 versus peak 2). (Note that cell culture itself is a minor stress to the cells.) Quantitation of these data is shown in Fig. 5C, which clearly indicated a critical role for PRAS40 in TGF-α-induced exosome secretion (bars 3 and 4 versus bars 1 and 2).

FIG 5.

PRAS40 mediates induced exosome secretion. Serum-starved primary human keratinocytes with or without PRAS40 downregulation were either left untreated (−) or treated with TGF-α (20 ng/ml). Conditioned media were collected and subjected to sequential centrifugations (see Materials and Methods). The 100,000 × g pellet fraction was analyzed by the following methods. (A) Comparison of secreted Hsp90α in 100,000 × g exosomes and the exosome-depleted supernatant. (B) NTA analysis of 100,000 × g pellet fractions (from conditioned medium of 3 × 10⁶ cells) under the indicated conditions. (C) Quantitation of the NTA data. (D) Effects of PRAS40 downregulation on the intracellular and secreted EV/exosome markers as indicated. The results were reproducible in three independent experiments. (E) NTA analysis of 100,000 × g pellet fractions from conditioned medium of 3 × 10⁶ cells, as indicated. **, P < 0.05; ***, P < 0.005.

To verify the NTA data for the role of PRAS40 in exosome secretion, we carried out immunoblot (Western) analysis of (i) total cell lysates, (ii) 100,000 × g pellets, and (iii) the remaining supernatant after the centrifugation at 100,000 × g with antibodies against both exosome and nonexosome protein markers. As shown in Fig. 5D, as expected, TGF-α stimulated PRAS40 phosphorylation in control shLacZ-infected but not in PRAS40-downregulated cells (row a, lane 4 versus lane 2). Interestingly, PRAS40 was not detected in the exosome fraction (row c), supporting our hypothesis that PRAS40 acts as an early signaling molecule between stress and exosome secretion. Consistent with the NTA data, PRAS40 downregulation blocked TGF-α-induced secretion of exosome-associated protein markers and cargo molecules (lane 4 versus lane 2), including Hsp90α (row d), CD63 (row e), CD81 (row f), CD9 (row g), and flotillin-1 (row h). In contrast, the ER/Golgi pathway-secreted matrix metalloproteinase 9 (MMP9) was not detected in the exosome fractions (row i). In the exosome-depleted conditioned media, Hsp90α was no longer detectable (row j), whereas the presence of MMP9 was unaffected by PRAS40 downregulation (row k). Consistent with previous Western blotting data, NTA also confirmed that TGF-α, but not EGF, triggers exosome secretion in human keratinocytes (Fig. 5E). The quantity of secreted exosomes (i.e., E6 particles per milliliter in Fig. 5C and E) varies due to the total number of cells used for each experiment.

On the other hand, if Hsp90α is just one of the cargo molecules in exosomes, depletion of Hsp90α should not affect exosome secretion. We used CRISPR/cas9 to knock out both alleles of the Hsp90α gene in MDA-MB-231 breast cancer cells (41) that have been shown to constitutively secrete exosomes and Hsp90α (42, 43). As shown in Fig. 6A, Hsp90α knockout is evident (row l, lane 2). However, the presence of CD63 (row n), CD81 (row o), CD9 (row p), and flotillin-1 (row q) in the 100,000 × g exosome fraction of the conditioned medium was unaffected by Hsp90α knockout (lanes 2 versus lanes 1).

FIG 6.

Pharmacological studies with BFA and DMA. (A) Hsp90α knockout (row l) did not affect exosome secretion (rows n to q) in MDA-MB-231 cells that constitutively secrete exosomes. (B) TGF-α-stimulated phosphorylation of PRAS40 inside the cells could not be blocked by either an ER/Golgi pathway inhibitor, BFA (10 μg/ml), or an exosomal pathway inhibitor, DMA (25 μg/ml) (row a). PRAS40 was not found in isolated exosome fractions (row c). In contrast, TGF-α-stimulated secretion of Hsp90α, exosome-enriched tetraspanins, CD63, CD81, and CD9, and the more general EV-enriched marker flotillin-1 was blocked by DMA but unaffected by BFA (lanes 3) (rows d to h, lanes 2 and 4). The ER/Golgi pathway-secreted MMP9 was not detected in the exosome fractions (row i) but was present in EV-depleted supernatants and sensitive to BFA (row k).

The above-described findings were further supported by a pharmacological approach using specific chemical inhibitors of the exosome and the ER/Golgi pathways, respectively. Dimethyl amiloride (DMA) and brefeldin A (BFA) are known specific chemical inhibitors of the exosome-mediated and ER/Golgi-mediated trafficking pathways, respectively. As shown in Fig. 6B, DMA showed the same effect as PRAS40 downregulation on exosome secretion (rows d to h, lane 4), whereas BFA inhibited the ER/Golgi-mediated MMP9 secretion (row k) but not exosome secretion (rows d to h, lane 3).

Finally, we found that the role for PRAS40 in stress-induced exosome secretion is conserved in various normal and tumor cell lines. As shown in Fig. 7, PRAS40 downregulation in primary mouse hepatocytes (Fig. 7A, row a) and MLE15 (mouse lung epithelial) cells (Fig. 7B, row i) blocked exosome secretion in response to hypoxia and oxidative stress (rows c to h and rows k to p, respectively). Many cancer cells show constitutively activated PRAS40 (44). We tested the role of PRAS40 in MDA-MB-231 breast cancer cells, BJAB Burkitt lymphoma cells, and SKNBE2 neuroblastoma cells, which all show constitutive PRAS40 phosphorylation at Thr-246 and constitutive expression of the HIF-1α oncogene. In these cells, as shown in Fig. 7C to E, PRAS40 downregulation dramatically decreased the constitutive exosome-mediated secretion even under normoxia (rows s to x for MDA-MB-231, rows y to c′ for BJAB, and rows d′ to h′ for SKNBE2).

Thr-246 phosphorylation of PRAS40 is necessary and sufficient to trigger exosome secretion.

The Thr-246 phosphorylation of PRAS40 by Thr-308-phosphorylated Akt1, as schematically shown in Fig. 8A, is a known extracellular signal-induced posttranslational modification (45). Since stress signals including TGF-α, hypoxia, and H₂O₂ all have been shown to activate Akt (46 to 48), we tested whether the Thr-246 phosphorylation of PRAS40 is the mechanism by which PRAS40 receives the extracellular stress signals for regulating exosome secretion. Specifically, we asked the important question of whether Thr-246 phosphorylation is necessary, sufficient, or both for PRAS40 to transmit microenvironmental signals to exosome-mediated secretion. We took a gene rescue approach by reintroducing wild-type PRAS40 (PRAS40-wt), dominant-negative PRAS40-T246A, and constitutively active PRAS40-T246E mutants into endogenous PRAS40-depleted cells. As shown in Fig. 8B, downregulation of the endogenous PRAS40 was nearly complete (row a, lane 2 versus lane 1). In these cells, each of the three PRAS40 constructs, as listed in Fig. 8A, was expressed at a level similar to that of the endogenous PRAS40 in the parental cells (lanes 3 to 5 versus lane 1). These cells were left untreated or treated with the indicated stress, their serum-free conditioned media collected, and the 100,000 × g pellet fractions of the conditioned media subjected to immunoblotting analysis for cargo and exosome markers. As shown in Fig. 8C, downregulation of PRAS40 blocked TGF-α-stimulated secretion of Hsp90α (row c, lane 4 versus lane 2), CD9 (row d, lane 4 versus lane 2), and CD63 (row e, lane 4 versus lane 2) compared to the control cells (lane 2). Reintroduced PRAS40-wt (lane 6), but not the dominant-negative PRAS40-T/A mutant (lane 8), rescued the TGF-α-induced secretion. Intriguingly, the constitutively active PRAS40-T/E mutant rescued Hsp90α and exosome marker secretion not only in the presence (lane 10) but also in the absence (lane 9) of TGF-α stimulation. Similar results were obtained in the same cells in response to H₂O₂ and hypoxia stress. Downregulation of PRAS40 blocked H₂O₂-triggered (Fig. 8D, lane 4 versus lane 2) and hypoxia-triggered (Fig. 8E, lane 4 versus lane 2) secretion of Hsp90α and exosome markers. Reexpression of PRAS40-wt (lane 6) but not PRAS40-T/A mutant (lane 8) rescued the secretion in the cells in response to both stress signals (lane 6). Again, the PRAS40-T/E mutant not only rescued stress-triggered secretion (lane 10) but also caused exosome secretion even in the absence of stress (lane 9). These findings suggest that activation of PRAS40 by Thr-246 phosphorylation is both necessary and sufficient to mediate stress signaling to exosome secretion.

FIG 8.

Threonine-246 phosphorylation of PRAS40 is necessary and sufficient for triggering exosome secretion. (A) List of the wt and two mutant constructs of PRAS40 cDNAs in a lentiviral vector, pRRLsinh-CMV, for overexpression. (B) Reintroduction of PRAS40-wt and the PRAS40-T/A and PRAS40-T/E mutants in endogenous PRAS40-downregulated human keratinocytes. The lysates of the cells were subjected to Western blotting with antibodies against the indicated targets. (C to E) Serum-free conditioned media of the cells treated with TGF-α (C), H₂O₂ (D), or hypoxia (E) were subjected to sequential centrifugations. The 100,000 × g pellet fractions were analyzed by Western blotting with antibodies against the indicated targets. This experiment was repeated three times with similar outcomes.

To directly demonstrate that PRAS40 regulates exosome secretion, we monitored CD63 translocation inside the parental and PRAS40-downregulated human keratinocytes by anti-CD63 immunostaining and confocal microscopy. As shown in Fig. 9A, in the parental cells, CD63 was found to cluster around the nuclei (rows b and c). TGF-α stimulation caused a clear spreading and translocation of CD63 to the periphery of the cells (rows e and f versus rows b and c). In the PRAS40-downregulated cells, as shown in Fig. 9B, TGF-α failed to cause CD63 translocation (rows k and l versus rows h and i). Quantitation of 80 to 120 randomly selected cells under each condition is presented underneath the images. The methods for quantitation are shown in Fig. 9C and D.

FIG 9.

PRAS40 regulates exosome translocation inside cells. (A and B) In 8-well chambered cell culture slides, serum-starved primary human keratinocytes without (A) or with (B) PRAS40 downregulation were either left untreated (−) or were treated with TGF-α (20 ng/ml) for 2 h. Cells were stained with DAPI for nucleus locations and CD63 for exosome locations and subjected to confocal microscopy analysis (see details in Materials and Methods). (C) An example of CD63's peripheral and total staining for computer-assisted quantitation. (D) The translocation percentage of CD63 was averaged from confocal readings of 60 cells per condition. Statistical analysis is described in Materials and Methods.

PRAS40 regulates exosome secretion independent of mTORC1 and 14-3-3.

Prior to the current study, the only reported function for PRAS40 was its binding to the Raptor (regulatory-associated protein of mTOR) subunit and inhibiting mTORC1 activation (20, 23, 28, 29, 49). Therefore, we tested whether downregulation of PRAS40 alone leads to mTORC1 activation and, more importantly, whether Raptor, mTOR, and 14-3-3 have any roles in stress-triggered exosome secretion. First, we found, surprisingly, that downregulation of PRAS4, as shown in Fig. 10A (row a), neither resulted in activation of mTOR in the absence of stimulation (Fig. 10B, row c, lane 3 versus lane 1) nor affected stress-induced phosphorylation of mTOR (lane 4 versus lane 2) in human keratinocytes. Second, as shown in Fig. 10C, we individually silenced Raptor (row e, lane 3) and mTOR (row f, lane 2) expression in the cells. Under these conditions, as shown in Fig. 10D, stress-stimulated secretion of CD63, the most agreed-upon exosome marker (50), remained unaffected in the absence of either mTOR (row h, lane 4) or Raptor (lane 6) compared to the control cells (lanes 2). Similar results were obtained for secretion of Hsp90α (row i) and CD9 (row j). Finally, downregulation of another reporter, the PRAS40-binding partner 14-3-3 (14-3-3θ) (Fig. 10E, row k), did not affect (i) PRAS40 expression (row l), (ii) TGF-α-stimulated PRAS40 phosphorylation (Fig. 10F, row n, lane 4 versus lane 2), and (iii) TGF-α-induced secretion of CD63 (Fig. 10G, row p), Hsp90α (row q), and CD9 (row r). Taking these findings together, we concluded that PRAS40 plays a positive role in mediating stress-triggered exosome secretion via a yet-to-be-identified pathway. As expected, downregulation of Raptor, mTOR, or 14-3-3 showed little effect on TGFα-stimulated exosome secretion (Fig. 10H).

FIG 10.

PRAS40 regulates exosome secretion without mTOR and 14-3-3 participation. (A) Human keratinocytes with downregulation of PRAS40 (row a, lane 2). (B) The absence of PRAS40 alone did not cause augmented mTORC1 phosphorylation (row c, lane 3 versus lane 1). (C) Individual downregulation of Raptor (row e, lane 3) and mTORC1 (row f, lane 2). (D) Neither Raptor nor mTORC1 downregulation affects TGF-α-stimulated secretion of the 100,000 × g fraction containing CD63 (row h), Hsp90α (row i), and CD9 (row j). (E) Downregulation of 14-3-3θ (row k) and its effect on PRAS40 levels (row l). (F) TGF-α-induced phosphorylation of PRAS40 was unaffected in the presence (lanes 1 and 2) or absence (lanes 3 and 4) of endogenous 14-3-3θ. (G) 14-3-3θ is not required for TGF-α-induced secretion of the 100,000 × g fraction containing CD63 (row p), Hsp90α (row q), and CD9 (row r) (lanes 4 versus lanes 2). This experiment was repeated twice with similar outcomes. (H) NTA analysis of TGFα-stimulated exosome secretion in Raptor-, mTOR-, and 14-3-3-downregulated cells. **, P < 0.05.

DISCUSSION

The discovery of exosomes in cell-to-cell communication in the microenvironment of homeostasis and tumor progression is a recent breakthrough in biology. For instance, secreted exosomes can transfer molecules between dendritic cells and B cells to mediate adaptive immune responses to pathogens. Tumor cells constitutively secrete exosomes to direct and facilitate metastasis by creating a more favorable environment for the tumor cells. Under these conditions, secretion of exosomes is driven by either extracellular stimuli (tissue injury, hypoxia, nutrient deprivation, etc.) or intracellular stimuli (activated oncogenes, inactivated tumor suppressor genes, or stress). However, little was known about the signaling mechanism that links the stress cues to the exosome secretion pathway. In this study, we demonstrate that the ubiquitously expressed PRAS40 is a common signaling molecule that connects various kinds of microenvironmental stress cues to exosome-mediated secretion in normal and tumor cells. Extracellular stress signals activate Akt via Thr-308 phosphorylation or possibly another kinase(s), such as PIM1, which in turn phosphorylates PRAS40 on Thr-246. Thr-246-phosphorylated PRAS40 does not make direct physical contact with the exosomes. Rather, it triggers, via currently unknown intermediates, exosome secretion into the extracellular environment. We demonstrated that Thr-246 phosphorylation is both necessary and sufficient for connecting extracellular signals to exosome secretion. In addition, this new “positive” function of PRAS40 is independent from its previously reported inhibitory role of mTORC1 or its binding to the scaffolding protein, 14-3-3. Numerous studies have reported on the roles for PRAS40 in preventing cell death and promoting tumorigenesis, but little was known about the mechanism of action by PRAS40. We argue that this previously unrecognized function for PRAS40 in regulation of stress-induced exosome secretion is a major part of its biology. A schematic representation of these findings is depicted in Fig. 11. As illustrated, while this study has identified the first linker molecule between stress and exosome secretion, it remains to be studied how Thr-246-phosphorylated PRAS40 communicates with the exosome-secreting machinery (see below).

FIG 11.

Schematic representation of stress-triggered exosome secretion through PRAS40. Extracellular stress cues, including growth factors, hypoxia, and H₂O₂, activate an intracellular kinase, such as Akt, which in turn phosphorylates PRAS40 at Thr-246. The activated PRAS40 communicates with a currently unknown intermediate(s), leading to exosome secretion. Exosomes contain a variety of molecules for efficient cell-to-cell communication, unlike secretion of a single molecule, such as a hormone. PRAS40 is the first linker identified between stress and exosome secretion.

One of the most intriguing findings and the basis of this study is the unprecedented specificity with which TGF-α, but not EGF, triggers Hsp90α secretion only in primary human keratinocytes (the crucial epidermal cell type for skin wound closure). Neither TGF-α nor EGF triggers exosome or Hsp90 secretion in human dermal fibroblasts or human microvascular endothelial cells from the dermis (33). However, human dermal fibroblasts and human microvascular endothelial cells secrete Hsp90α under other microenvironmental stress signals, such as hypoxia (37, 38). These observations make biological sense. It is known that TGF-α levels are low or undetectable in intact skin and rise when skin is wounded. In contrast, the EGF levels remain unchanged before or after skin wounding (35). Mechanistically, while both TGF-α and EGF induce Ser-473 phosphorylation of Akt in keratinocytes, only TGF-α induces Thr-308 phosphorylation (the main activation of phosphorylation) of Akt. It has been shown that Thr-308-phosphorylated Akt is responsible for phosphorylating PRAS40 at Thr-246 (39), supporting our findings in this study. We have proposed that keratinocytes are the main source of secreted Hsp90a in wounded skin for promoting wound closure (35).

While a body of reports showed that PRAS40 is an inhibitor of the mTORC pathway and inhibitors of mTORC are often known as tumor suppressors, an equal number of studies showed that PRAS40 has two positive roles in both normal and tumor cells, including (i) preventing stress-triggered normal and tumor cell apoptosis and (ii) supporting tumor progression in vitro and in vivo. An earlier study showed that PRAS40 inhibits cell apoptosis by preventing caspase 3 cleavage (21). Madhunapantula and colleagues showed that Akt3 phosphorylates PRAS40 and upregulates the PRAS40 levels in melanoma cells to prevent the cancer cells from undergoing apoptosis. Downregulation of PRAS40 or inhibition of its upstream Akt3 decreases the anchorage-independent growth of cells in culture and tumor development in mice (51). Similar findings were reported in breast and lung cancer cells (52). Kazi et al. showed that silencing PRAS40 reduced proliferation of C2C12 cells due to a cell cycle arrest in the G₁ phase (53). Huang et al. showed that PRAS40 is a target gene for Ewing sarcoma protein, a transcription factor, and promotes development of Ewing sarcoma (52). Havel and colleagues reported a protumorigenic effect of PRAS40 by suppressing p53-mediated cellular senescence (54). In normal cells, Yu et al. reported that elevated PRAS40 levels protect motor neurons from spinal cord injury-induced cell death (55). Similarly, Shin et al. showed that overexpression of PRAS40 prevents brain ischemic insult and oxidative stress-induced brain cell death (56). Obviously, these findings cannot be explained by the reported role for PRAS40 as an inhibitor of the mTOR pathway. The critical question, then, is how PRAS40 exerts these two positive functions. Our current finding that Thr-246-phosphorylated PRAS40 regulates exosome secretion provides a possible mechanism for how PRAS40 protects cells from apoptosis and supports tumor progression via secreted exosomes. A recent study from our laboratory showed that the exosome-mediated secretion of Hsp90α prevents tumor cells from hypoxia-induced cell death (59). Moreover, Zou and colleagues showed that exosome-mediated secretion of Hsp90α plays an essential role not only in de novo tumor formation but also in the further expansion of already-formed tumors in mice (41). Besides the cargo of Hsp90α, one could anticipate that other types of exosome cargo (proteins, DNA, miRNA, mRNA, and lipids) modulate other specific biological events in various tissues and cells.

Results of our study also suggest that distinct stress signals activate PRAS40 via Thr-246 phosphorylation by different kinases or distinct mechanisms. For instance, TGF-α-stimulated Thr-246 phosphorylation of PRAS40 is inhibited by LY29200 (Fig. 3D), but H₂O₂-induced Thr-246 phosphorylation of PRAS40 is not (data not shown). Instead, H₂O₂ treatment dramatically increases the cellular PRAS40 protein levels, in addition to Thr-246 phosphorylation (Fig. 4C). These observations suggest that under oxidative stress a different upstream kinase or kinases phosphorylates PRA40 at Thr-246. This phosphorylation or a completely independent mechanism causes the increased cellular PRA40 protein levels, triggering increased exosome secretion. Consistent with this notion, increased PRAS40 levels were reported to correlate with later stages of melanoma progression due to activated Akt3 (51).

The critical question of how Thr-246-phosphorylated PRAS40 communicates with the exosomal trafficking pathway remains to be investigated. Since the activation of PRAS40 is an early signaling event in response to extracellular signal stimulation, we believe that there must be additional intermediate signaling events. One possibility is that PRAS40 uses its proline-rich domains/motifs to connect with Src homology 3 (SH3) domain-containing signaling molecules. For instance, it has recently been reported that the SH3-containing protein cortactin promotes exosome secretion by stabilizing cortical actin-rich MVE docking sites (19). We did not, however, detect PRAS40 directly binding to cortactin (data not shown). Amzallag et al. reported a role for TSAP6, a p53-inducilbe transmembrane protein, in exosome-mediated secretion of TCTP (translationally controlled tumor protein) from its preexisting pool in the cells (57). Yu and colleagues extended this finding to show that gamma radiation-induced DNA damage activates p53, resulting in increased expression of TSAP6 and exosome-mediated secretion of proteins that are not processed by the ER/Golgi classical protein trafficking pathway (55). Moreover, Lespagnol et al. reported that TSAP6 knockout mice, otherwise developmentally normal, showed defects in DNA damage-induced and p53-dependent exosome secretion (58). As mentioned previously, the Rab27 small GTPase family is widely reported to play a role in exosome biogenesis and secretion. Currently, limited information is available for what their exact roles are in those two sequential and distinct processes. Based on exosome-mediated secretion of Hsp90α, normal cells do not secrete exosomes under physiological conditions. Rather, they secrete exosomes only under environmental stress cues. In contrast to normal cells, many tumor cells constitutively secrete exosomes driven by their intrinsic oncogenic signals (7, 35). It would be of a great interest to test whether PRAS40 has any relationship with TSAP6 or Rab27.

In conclusion, tissue microenvironmental stress cues, such as hypoxia, nutrient paucity, injury-released cytokines, and oxidative stress, all can trigger cells to vesiculate and secrete EVs. Under physiological conditions, for instance, secreted exosomes can transfer molecules between dendritic cells and B cells to mediate adaptive immune responses to pathogens. Under pathological conditions, such as in a tumor microenvironment, malignant cells, vascular cells, stromal cells, and immune cells surrounded by extracellular matrices (ECMs) and soluble factors use the cargo within secreted exosomes to engage in cell-to-cell communication to support tumor progression and metastasis (7). The identification of PRAS40 as a pivotal regulator of exosome secretion will hopefully lead to new insights about how microenvironmental and intracellular stress induces cells to secrete exosomes. In the domain of cancer biology, PRAS40 may also serve as a target for therapy.

MATERIALS AND METHODS

Cell lines.

Primary human keratinocytes (HKCs) were cultured in EpiLife medium with added growth factor supplements (Thermo Scientific, MA). The third or fourth passages of cells were used throughout this study. Primary mouse hepatocytes, the mouse lung epithelial cell line MLE15, and the human triple-negative breast cancer cell line MDA-MB-231 were obtained from the laboratories of Bangyan Stiles, Zea Borok, Pinghui Feng, and Michael Press (University of Southern California, Los Angeles, CA). All four types of cells were cultured in Dulbecco's modified Eagle's medium (DMEM) with high glucose supplemented with 10% fetal bovine serum (FBS) (Thermo Scientific, MA).

Antibodies and reagents.

TGF-α and EGF were purchased from Fitzgerald Industries International (Acton, MA). Human PRAS40 cDNA (wt) was purchased from Addgene (plasmid 14950). Anti-PRAS40 (MAB6408), anti-phospho-PRAS40 (T246) (MAB6890), and anti-phospho-RSK S380 (MAB79671) antibodies were from R&D Systems (Minneapolis, MN). Anti-CD63 (EXOAB CD63-A1), anti-CD9 (13403), anti-flottilin-1 (3253), anti-CD81 (EXOAB CD81A-1), anti-phospho-Akt S473 (4060), anti-phospho-Akt T308 (4087), and anti-EGFR (4267) antibodies were from System Biosciences (Mountain View, CA) and Cell Signaling Technology (Danvers, MA). Mouse monoclonal antibodies against human Hsp90α (CA1023) and human Hsp90β (SMC107) were from Calbiochem (Billerica, MA) and Stressmarq Biosciences (Victoria, BC, Canada), respectively. Anti-cyclin-D1 (GTX61845) and anti-glyceraldehyde-3-phosphate dehydrogenase (anti-GAPDH) (GTX28245) antibodies were from Genetex (Irvine, CA). LY294002 was from Cell Signaling (catalog no. 9901; Danvers, MA). BFA and DMA were purchased from Sigma-Aldrich (St. Louis, MO).

Human phosphokinase antibody array.

Human keratinocytes were grown to 80% confluence in 15-cm tissue culture dishes and incubated in serum-free medium overnight. Cells were stimulated with growth factors for the indicated time. The stimulation was stopped by addition of ice-cold phosphate-buffered saline (PBS) buffer, and the cells were lysed on ice. The postnuclear extracts were subjected to a proteome profiler human phosphokinase array (ARY003B; R&D Systems, Minneapolis, MN) according to the manufacturer's instructions. In this protocol, the most critical thing is to synchronize all of the steps for differentially treated samples from the time of incubation with the kinase array membranes all the way to ECL development of the results in order to compare the relative intensities of the dots.

Stress treatments.

An OxyCycler C42, from BioSpherix (Parish, NY), was used as the oxygen content controller. The medium used for hypoxia experiments was preincubated in a hypoxia (1% O₂) chamber for 16 h prior to its use to replace the normoxia culture medium (37). O₂ (1%) was used throughout the study. Hydrogen peroxide (H₂O₂) was purchased from VWR Analytical (Radnor, PA), and 10 μM was chosen for the treatment of the cells.

Exosome purification, characterization, and analyses.

Conditioned media were collected (33) and spun at 300 × g at 4°C for 10 min to remove floating cells. Dead cells and microvesicles were removed by centrifugation at 2,000 × g for 10 min, followed by centrifugation at 10,000 × g for 30 min. Finally, the cleared supernatant was centrifuged at 100,000 × g for 70 min to collect the exosomes in the pellets. The exosome fractions were washed in 10 ml of PBS and centrifuged again at 100,000 × g for 70 min to get rid of any contaminating particles left. The size distribution and concentration of the exosome fractions were analyzed using Nanosight (Malvern Instruments) aided by NTA software. Molecular markers for exosomes and Hsp90α were verified using Western immunoblot analyses.

Site-directed mutagenesis of PRAS40.

The QuikChange II XL site-directed mutagenesis kit (200521-5) from Agilent Technologies was used to mutate the T246 site of PRAS40. The primer sequence GGAAGTCGCTGGCGTTAAGCCGCGGC (sense) was used to generate the T246A mutation, and the primer sequence GCTTCTGGAAGTCGCTTTCGTTAAGCCGCGGCCGTGG (sense) was used to generate the T246E mutation.

Lentiviral systems for up- or downregulation of target genes.

The protocols for using lentiviral systems for gene downregulation and gene upregulation, including virus packaging, isolation, infection, and analyses, were as previously described (33, 37, 42). The pRRLsinh-CMV system was used to overexpress exogenous genes such as the wt and mutant PRAS40 cDNAs. The pHR-CMV-puro RNAi delivery system was used to deliver shRNA. The shRNA sequence of PRAS40 was GCTGAGTTCTAAGCTCTAA (sense), for EGFR was AGAATGTGGAATACCTAAGG (sense), and for 14-3-3θ was GTGCAGTACTGCTGTAGA (sense).

Confocal microscopic analysis of immunostaining.

For immunostaining with anti-CD63 antibody, approximately 15,000 cells of passage 2 to 3 primary human keratinocytes were seeded in 8-well chambered cell culture slides (Corning, NY) and grown to ∼80% confluence. The cells were serum starved overnight and either left untreated or treated with human recombinant TGF-α (20 ng/ml) for 2 h, the earliest time point at which induced exosome secretion was detectable. Cells were washed twice with ice-cold PBS buffer, fixed in acetone for 5 min, washed with PBS, and incubated in blocking reagent (10% normal goat serum, 0.05% Tween 20, 0.05% Triton X-100, and 1% BSA in PBS) for 60 min. The slides were washed twice with PBS and incubated with anti-CD63 antibody (19281; Thermo Fisher Scientific, CA) for 2 h, washed three times with PBS, and incubated with fluorescein isothiocyanate (FITC)-conjugated secondary antibody. Lastly, the slides were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) solution (2 μg/ml DAPI in 40% glycerol), mounted on coverslips, and sealed with nail polish. The images were taken under a confocal microscope (Eclipse C1; Nikon, Japan) with sequential applications of the following fluorochromes: green (FITC) and blue (DAPI). The images were taken by Photoshop software (San Jose, CA) in JPEG format. Sixty cells from 10 to 20 randomly chosen fields (60×) per experimental condition were evaluated for image analysis by Image J software (NIH). Green fluorescence around the peripheral membrane of the cells was localized by drawing the region of interest. The intensities of the green fluorescence within the membrane periphery and of the total cell were measured by ImageJ software. The translocation percentage was calculated by dividing the fluorescence intensity in the membrane area by the total cell fluorescence.

Statistics.

Data are based on three or more independent experiments and presented as means ± standard deviations. Statistical significance for comparisons was evaluated by the Student two-tailed t test for comparisons of two groups or analysis of variance for comparisons of more than two groups. A P value equal to or less than 0.05 was considered statistically significant.