5.1 HYALURONAN IN PLASMA MEMBRANE PROTRUSIONS (I)
5.1.1 Myosin-‐‑X localizes in the tips of the hyaluronan-‐‑dependent plasma membrane protrusions
We wanted to study the role of Myo10 in HAS3-‐‑dependent plasma membrane protrusions.
In contrast to a previous study (Watanabe et al. 2010), Myo10 overexpression did not induce plasma membrane protrusions or hyaluronan production in MCF-‐‑7 cells. However, simultaneous overexpression of GFP-‐‑HAS3 and mCherry-‐‑Myo10, by double transfection, showed that Myo10 is specifically located on the tips of the HAS3-‐‑dependent plasma membrane protrusions (Study I, Figure 3B and C). Myo10 was not essential for the HAS3-‐‑
dependent plasma membrane protrusions, as using Myo10 constructs with the deleted pleckstrin homology domain, previously shown to prevent the protrusions (Plantard et al.
2010), did not inhibit the formation of HAS3-‐‑induced plasma membrane protrusions or hyaluronan synthesis. In addition, GFP-‐‑HAS3 transfection of MDA-‐‑MB231 cells with a stable Myo10 knockout did not show any change in the protrusions. These experiments demonstrate that although Myo10 has a specific location in HAS3-‐‑dependent plasma membrane protrusion, it is not essential for their formation.
5.1.2 Hyaluronan has a supportive role in plasma membrane protrusions
Because mCherry-‐‑Myo10 was found to specifically localize on the tips of HAS3-‐‑dependent protrusions, it was further utilized to visualize the tips of the protrusions to study the significance of the hyaluronan. Treatment of GFP-‐‑HAS3 and mCherry-‐‑Myo10 double transfected cells with Streptomyces hyaluronidase showed that while GFP-‐‑HAS3 disappeared from the plasma membrane of cell protrusions, the mCherry signal remained at the tips of the protrusions (Study I, figure 4A-‐‑B). 4-‐‑MU and glucose starvation (inhibitors of hyaluronan synthesis) were utilized to study the role of hyaluronan synthesis in the growth of the protrusions. Both treatments resulted in the disappearance of the protrusions and relocation of the mCherry-‐‑Myo10 and GFP-‐‑HAS3 signal to the cytosol (Study I, Figure 4E-‐‑F).
LP9 is a human mesothelial cell line that is known for its ability to produce spontaneously high amounts of hyaluronan and hyaluronan-‐‑dependent cell protrusions (Rilla et al. 2008). Also in LP9 cells, hyaluronidase treatment did not induce an immediate disappearance of the cell protrusions (study I, figure 8C and D). Instead, overnight treatment with 4-‐‑MU resulted in the disappearance of most of the cell protrusions (Study I, Figure 8E-‐‑F and Figure 9A and B). Removal of 4-‐‑MU resulted in quick regrowth of the protrusions (Study II, Figure 8G-‐‑H and Figure 9C) indicating that spontaneous cell protrusions are requirements for, and the consequence of, active hyaluronan synthesis.
Correlative light and electron microscopy is a powerful method for studying the living cellular structures first with confocal microscopy and to compare the same structures in scanning electron microscope in more detail. The effect of Streptomyces hyaluronidase was confirmed by using correlative light and electron microscopy as hyaluronidase treated and fixed GFP-‐‑HAS3 transfected cells showed a similar collapsed structure of cell protrusions like controls (Study II, figure 7) while 4-‐‑MU treatment induced the disappearance of the protrusions (Study II, figure 8 E-‐‑F).
These experiments showed that GFP-‐‑HAS3 escapes relatively quickly from the plasma membrane when not bound to the hyaluronan. Taking into account that the hyaluronan coat in the HAS3-‐‑positive microvilli is not dependent on CD44 (Kultti et al. 2006), these experiments indicate that it is the growing hyaluronan chain that retains the HAS3 enzyme on the plasma membrane and that the hyaluronan coat supports the protrusions.
5.2 HAS-INDUCED CELL PROTRUSIONS ARE DYNAMIC STRUCTURES (I) 5.2.1 The dynamics of the hyaluronan-‐‑dependent cell protrusions
Given that mCherry-‐‑Myo10 was a useful tool for the visualization of cell protrusions, mCherry-‐‑Myo10 and GFP-‐‑HAS3 double transfections were used for 3D time-‐‑lapse imaging in order to study the dynamics of the protrusions. Two diverse populations of the protrusions were seen according to their dynamics. Cell protrusions attached to the substratum were relative stable (Study I, asterisks in figure 5). However, the dorsally localized protrusions showed very rapid growth and retraction dynamics. In addition, intrafilopodial movement of Myo10 deposits was occasionally seen along the HAS3 positive cell protrusions, indicating that Myo10 may have a functional role in the intrafilopodial traffic.
5.2.2 Lateral mobility of GFP-‐‑HAS3 molecules on the cell protrusions is restricted
Fluorescence recovery after photobleaching (FRAP) technique was used to study the dynamics of the HAS3 movements in the cell protrusions, as compared to the plain plasma membrane. After bleaching the GFP tag, the recovery rate was significantly lower in cell protrusions than in the plain plasma membrane, with half-‐‑recovery times of 66,6 ± 16,5 s and 9,17 ± 2,7 s, respectively. The turnover of GFP-‐‑HAS3 may be lower in the cell protrusions because more hyaluronan chains are attached to it during ongoing synthesis.
5.3 STUDIES ON THE STRUCTURE OF THE HYALURONAN DEPENDENT CELL PROTRUSIONS
5.3.1 Hyaluronan exists on the luminal surface of the mesothelium (II)
The inner surface of the rat parietal peritoneum contained a one cell layer thick squamous mesothelial lining with flattened nuclei. The mesothelial cells were attached to the underlying connective tissue containing mainly collagen fibers. The mesothelium of the rat small intestine was structurally similar, but lay on two layers of smooth muscle cells.
Mesothelium is strongly positive for the mesothelial marker HBME-‐‑1 that recognizes an unknown microvillar epitope in mesothelial cells (Nga et al. 2008).
The mesothelium in the parietal peritoneum showed strong hyaluronan staining. Both the epithelial surface and the connective tissue were clearly positive for hyaluronan, which was also located between the striated muscle cell boundaries. In the small intestine, the mesothelial layer showed a clear positive surface pattern, with some DAB stain in the smooth muscle layers. However, the strongest hyaluronan stain was seen in the submucosa and between the glands of the small intestine. Hyaluronan stain was absent in the epithelium of the small intestine, a staining pattern similar to that described before (Alho and Underhill 1989).
Isolated primary mesothelial cells from the rat anterior parietal peritoneum showed typical cobblestone morphology in cell culture. The major proportion of the cell population expressed the mesothelial marker HBME-‐‑1, which was located on the plasma membrane
and its protrusions. The cells produced high amounts of hyaluronan, which was shown to be localized on plasma membranes and cell protrusions as previously reported (Rilla et al.
2008). The protrusions were especially numerous when EGF was present in the culture medium.
The mesothelial cells, rich in mitochondria and small vesicles near the plasma membrane of the luminal surface, were decorated with long and numerous plasma membrane protrusions. Some of the protrusions were perpendicular to the sectional plane. Scanning electron micrographs showed that the mesothelial plasma membrane protrusions often collapsed on the cell surface during sample preparation (Study II, Figure 4). The DAB deposits formed an electron dense, rather uniform layer over the luminal surface of the mesothelial cells (Study II, Figure 5).
5.3.2 Mesothelial and HAS3-‐‑induced cell protrusions are structurally similar (I)
Next, the structure of the protrusions on the mesothelium was compared in the small intestine brush border and in the HAS3 overexpressing MCF-‐‑7 cells (Kultti et al. 2006). The ultrastructure of the mesothelial and HAS3-‐‑dependent protrusions was similar (Study I, Figure 1). The average length and diameter of protrusions were also comparable in the mesothelium and the HAS3 overexpressing cell cultures, while the actin filament densities in both were lower than in the small intestine brush border (Study I, Table 1).
Hyaluronidase treatment did not affect the ultrastructure of the protrusions. (I)
The cytoskeleton-‐‑associated proteins between HAS3-‐‑dependent cell protrusions and those found in the LP9 mesothelial cell cultures were compared. It was found that both have a similar immunostaining pattern and contain proteins typically found in filopodia, such as ezrin and fascin, but not microvillus-‐‑specific proteins, like villin and espin (Study I, Table 2).
5.3.3 HAS3 overexpression relocates actin to the cell cortex and to the bases of the cell protrusions (I)
The localization and the impact of HAS3 overexpression on actin filaments in MCF7 cells were studied in more detail, by using double transfections in live cells. A cell overexpressing GFP-‐‑HAS3 has typical “Hedgehog-‐‑like” morphology. To confirm that HAS3 expression without GFP-‐‑tag induces protrusions, HAS3 GFP-‐‑actin double transfection was performed. To visualize actin filaments, Lifeact, a 17 amino acid actin binding compound that does not interfere with actin dynamics was used as a marker (Riedl et al. 2008). Transfection with Lifeact linked with control GFP showed that actin is localized rather uniformly in the cytoplasm. Instead, when transfecting the MCF-‐‑7 cells with GFP-‐‑
HAS3 and Lifeact, actin filaments were especially concentrated to the bases of the cell protrusions and the signal was weakening towards the tips. GFP-‐‑HAS3 signal was high in all parts of the protrusions and forms a knob-‐‑like structure on the tip (Study I, Figure 2).
5.4 HYALURONAN SYNTHASES AND CD44 IN THE INTACT MESOTHELIUM (II)
5.4.1 HAS2 is the main hyaluronan producing enzyme in mesothelium
All three HAS antibodies showed a positive staining pattern on mesothelium in addition to the smooth muscle layer, stromal cells, glands and epithelium. The intensity of the staining with each HAS antibodies varied between different tissue blocks. Comparison of the staining intensity of different isoenzymes was not relevant because the affinities of different HAS antibodies are not known, (Study II, Figure 2). Furthermore, Has mRNA levels have
been shown to correlate fairly well with hyaluronan synthesis (Jacobson et al. 2000;
Karvinen et al. 2003; Pienimäki et al. 2001; Yamada et al. 2004b). Therefore, in order to investigate in more detail which of the three HAS isoforms has the main role for producing hyaluronan in mesothelial cells, mRNA was isolated from the confluent primary mesothelial cell culture and anterior wall of the parietal peritoneum. Has2 was the most abundant isoform both in intact mesothelium and in cell culture comprising approximately 60-‐‑70 % of the total Has mRNA pool. There were no significant differences between Has1 and Has3 mRNA levels in the intact mesothelium. In contrast, in cultured primary cells, Has1 and Has3 proportions were 10 and 30 % respectively (Study II, Figure 7).
5.4.2 Mesothelium is negative for CD44
Hyaluronan receptor CD44 is often present in cells producing hyaluronan (Alho and Underhill 1989). Surprisingly, CD44 was absent in the intact rat mesothelium both in parietal peritoneum (Study II, Figure 1f) and in small intestine mesothelial cells (Study II, Figure 2h). Some of the fibrocytes in sub-‐‑mesothelium showed CD44 positivity. In the small intestine, cells with round nuclei between the two muscle layers, probably tissue macrophages, expressed CD44. Some cells in the bottom of small intestine glands showed a clear plasma membrane pattern of CD44, similar to a previous report (Alho and Underhill 1989; Hou et al. 2011), probably indicating small intestine epithelial stem cells (Hou et al.
2011).
5.5 EMT ACTIVATES HYALURONAN SYNTHESIS MACHINERY IN PRIMARY MESOTHELIAL CELLS (III)
5.5.1 EGF and wounding induces EMT in mesothelial cells
Epidermal growth factor (Cheng et al. 2012) and wounding (Arnoux et al. 2008) are capable of inducing epithelial to mesenchymal transition. The 100% confluent primary mesothelial cells showed typical cobblestone morphology with short microvilli on the apical surface (Study III, Figure 2). EGF (10 nm/ml) treatment for 24 h induced a spindle-‐‑like morphology (Study III, Figure 2C), as reported previously (Leavesley et al. 1999). The number and average length of cell protrusions was increased, especially in spindle-‐‑shaped cells (Study III, Figure 2D). In wounding experiments, the cells exhibited a migratory phenotype with prominent lamellipodia located on and near the wound edge (Study III, Figure 2F). In both treatments, long adhesive tethers were seen between the cells.
The gene expression levels of two typical molecular markers of EMT were studied. The results suggested that the levels of mesenchymal marker alpha-‐‑smooth muscle actin raised in both EGF-‐‑treated and wounded cultures; however, statistical significance was not reached (Study III, Figure 3A). The relative raise was smaller when cultured in 20% FBS. In immunostainings, α-‐‑SMA was clearly upregulated in both EGF and wounded cell cultures (study III, Figures 3D and 3E respectively). In EGF treated cells, the most intense α-‐‑SMA staining was seen in the spindle-‐‑like cells, however, in wounded cultures, α-‐‑SMA positivity was restricted to the cells near the wound edge. As expected, the E-‐‑Cadherin mRNA levels were decreased upon EGF-‐‑treatment and the decrease was statiscically significant (P<0.05 when cultured in 2% FBS). Interestingly, the E-‐‑cadherin levels seemed to be raised 16 h after wounding.
5.5.2 EGF and wounding induce a marked CD44 overexpression
Messanger RNA levels of CD44 in 100 % confluent control samples were low, however, 24 h EGF treatment increased the CD44 mRNA levels up to 8.3-‐‑ and 4.9-‐‑fold in 2% and 20%
FBS, respectively, and the changes were statistically significant. Likewise, in wounded cell
cultures, the changes seemed to be lower; 3.9-‐‑ and 1.7-‐‑fold in 2% and 20 % FBS, respectively, which were not statistically significant (Study III, Figure 4E).
The mRNA results were in line with those produced by immunohistochemistry. In control cultures, staining for CD44 was nearly negative (Study III, Figure 4B and F). EGF induced CD44 protein levels especially in spindle-‐‑like cells (Study III, Figure 4C and G). In wounded cultures, the changes were rather local. CD44 staining intensity was strong in 3-‐‑6 cell layers from the edge of the wound (Study III, Figure 4D and H).
5.5.3 Hyaluronan synthesis is increased by EGF and wounding treatments
Cell-‐‑associated hyaluronan and hyaluronan secretion to culture medium was quantitated with hyaluronan assay. Untreated confluent mesothelial cells showed low to moderate hyaluronan secretion rate. Because of the high variance between cell cultures originating from different individual rats in basic hyaluronan secretion, the effects of EGF or wound treatments were presented as relative changes compared with the control. EGF induced 3.8-‐‑
and 1.7-‐‑fold increase in the secreted hyaluronan levels in 2 % and 20 % FBS concentrations, respectively (Study III, Figure 5A). The changes were not statistically significant, probably because of the high variance. The cell-‐‑associated hyaluronan increased 1.4-‐‑ and 4-‐‑fold in 2
% and 20 % FBS, respectively (Study III, Figure 5B). Wounding did not induce hyaluronan secretion to cell culture medium, however in 20 % FBS a 2-‐‑fold increase was observed.
Confocal microscopy of the fluorescent hyaluronan binding protein confirmed that hyaluronan staining in control samples was rather low (Study III, Figure 5C), but higher compared to CD44 staining intensity. In the EGF samples, the cell associated hyaluronan and CD44 staining intensity were increased (Study III, Figure 5D and G). Hyaluronan and CD44 staining were clearly higher in the wound edge (Study III, Figure 5E and H).
5.5.4 Has2 is overexpressed during EGF or wounding treatments
mRNA levels of each HAS isoform were analyzed with qPCR to determine whether the increased hyaluronan secretion of EGF and wound-‐‑induced mesothelial cells was due to induced HAS levels. In EGF treated samples, Has1 and Has3 expression increased slightly whereas Has2 had a clear 4-‐‑fold raise. In wounded cultures, Has1 levels remained constant, Has2 levels raised about 2-‐‑fold and Has3 had a small raise.
Immunostainings confirmed the qPCR results. EGF or wounding had no evident effect on HAS1 or HAS3 protein staining intensity. However, in HAS2 immunostainings EGF and wounding clearly increased HAS2 staining intensity (Study III, Figure 6). These results are in line with previous studies on the effect of EGF (Chow et al. 2010; Erickson and Turley 1987; Heldin et al. 1989; Honda et al. 1991; Pasonen-‐‑Seppänen et al. 2003; Saavalainen et al.
2005; Tirone et al. 1997; Yamada et al. 2004a) or mesothelial wounding (Yung et al. 2000).
5.6 STUDIES ON HYALURONAN-COATED MICROVESICLES
5.6.1 Active hyaluronan synthesis induces hyaluronan-‐‑coated microvesicles (IV)
Several cell types including immortalized human mesothelial cells (LP9), rat primary mesothelial cells and human chondrosarcoma cells form numerous hyaluronan coated cell protrusions. When these structures were studied with confocal microscopy using the fluorescent hyaluronan binding probe, fHABR, small hyaluronan-‐‑positive particles were detected around the cells, (Study IV, Figure 1). Using melanoma and MCF-‐‑7 cells transfected with GFP-‐‑HAS3, the deposits also showed GFP fluorescence, indicating that these structures were covered with plasma membrane, and further suggesting that they may represent microvesicles (Study IV, Figure 1f). By using 3D time-‐‑lapse confocal microscopy, GFP-‐‑HAS3-‐‑positive structures were found to detach from the tips of the
plasma membrane protrusions (Study IV, Figure 2). GFP-‐‑HAS3-‐‑transfected MDCK cells, which synthesize high levels of hyaluronan and form long hyaluronan-‐‑positive protrusions, were used to produce microvesicles for further studies. Culture medium from the transfected cells was collected and ultracentrifuged. Confocal microscopy of the pellet showed numerous GFP-‐‑HAS3-‐‑ and hyaluronan-‐‑positive microvesicles. Treating these vesicles with hyaluronan hexasaccharides, which blocks hyaluronan-‐‑CD44 interaction, did not diminish the hyaluronan coat. This suggests that interaction with CD44 is not required for the hyaluronan coat of these microvesicles (Study IV, Figure 3D).
5.6.2 Microvesicles bud off from the apical surface of the cell cultures (IV)
In order to study the correlation between hyaluronan production and microvesicle secretion, glucose starvation was used to inhibit hyaluronan synthesis. When cultured in 5 mM glucose, GFP-‐‑HAS3-‐‑transfected cells produced 30 ng of hyaluronan per 10,000 cells during 24 h, but in the absence of glucose, hyaluronan production was diminished by 88%.
Microvesicle shedding was also studied in a 3D matrix prepared from a basement membrane extract, where MDCK cells differentiate into a hollow cysts. Glucose starvation resulted in approximately 80% decrease in the shedding of the GFP-‐‑HAS3-‐‑positive microvesicles, indicating that hyaluronan production correlates well with the shedding of microvesicles (Study IV, Figures 4 and 5).
5.6.3 Hyaluronan-‐‑dependent microvesicles are formed by two distinct ways (IV)
Transmission electron microscopy revealed that the microvesicles were formed in two distinct ways; some of the vesicles seemed to bud directly from the plasma membrane, whereas others originated from the tips of the plasma membrane protrusions, as seen already in the time-‐‑lapse confocal microscopy. Numerous globular particles on the plasma membranes of LP9 and GFP-‐‑HAS3-‐‑transfected MDCK cells were seen in scanning electron microscopy, confirming the results from transmission electron microscopy (Study III, Figures 6 and 7).
5.6.4 EGF and wounding enhance the production of extracellular vesicles (III)
Next, the effect of EMT on microvesicle production was investigated. Scanning electron micrographs showed some big vesicles in control cultures (Study III, Figure 7A). EGF induced variable sized vesicles (Study III, Figure 7B). Wounding, however, did not have an obvious effect on vesicle formation, but some of the migrating cells had a high number of budding vesicles (Study III, Figure 7C and H). The SEM results were confirmed by using nano tracking analysis (NTA) from culture media of different treatments. EGF increased extracellular vesicles in the culture media by 8.6-‐‑fold, and wounding by 4.6-‐‑fold. The increases were, however, not statistically significant (Study III, Figure 7D). The mean diameter of the secreted vesicle population was equal in all treatments, approximately 150 nm (Study IV, Figure 7E). High resolution light microscopy imaging was used to study if the cell-‐‑associated vesicles were positive for hyaluronan, HASs and CD44. Live cell imaging using fluorescent HABC confirmed that the levels of the cell-‐‑associated hyaluronan were relatively low in control samples (Study III, Figure 8C). An increase of the cell-‐‑associated hyaluronan was seen in the EGF-‐‑treated and wounded samples (Study III, Figure 8D-‐‑E). Interestingly, the immunostainings showed that HAS3 protein accumulated on the plasma membranes of extracellular vesicles in the EGF-‐‑treated cells, while HAS1 or HAS2 were not detected (data not shown). CD44 accumulation was seen on the membranes of the vesicles from the EGF treated cells (Study III, Figure 8B).
Superresolution images (Airyscan) showed budding of CD44-‐‑positive vesicles (Study III, Figure 8F and I). After fixation most of the extracellular vesicle-‐‑associated hyaluronan coat was shed away, or clustered (Study III, Figure 8G-‐‑H).