First, we tested the effect of overexpressing each core component of the Par-complex in S2 cells, which distribute these components evenly throughout the cytoplasm and divide symmetrically (Figure 1A). We found that all core components of the Par-complex cortically co-localized in an asymmetric manner when Par3 was overexpressed, but did not cortically localize, when Par6 or aPKC was overexpressed (Figure 1B,C, and data not shown). We overexpressed myc-Par3 (or Par3-mKate2) via the actin-promoter (act5c)-driven Gal4-UAS system (Act-Gal4 >UAS) by transfection (see Materials and methods), with or without actin-promoter-Par6-GFP (pAct-Par6-GFP) as a live marker, which was uniformly distributed in the cytoplasm in the absence of Par3 overexpression (Figure 1B). Among transfected cells that exhibited cortical Par-complex distribution, a fraction exhibited an asymmetrically localized Par-complex (Figure 1C), while the rest of the cells localized uniformly to the cortex (see below). Asymmetric distribution of the Par-complex induced by Par3 overexpression required endogenous aPKC and Par6 (Figure 1D). The expression level of exogenous Par3 was also important for S2 cell polarization. Cortical polarization was not observed (Figure 1B) when Par3 expression, directly driven by the actin promoter, was approximately 1/40 of that of the Act-Gal4 >UAS system (Figure 1E).

Figure 1

Download asset Open asset

We next examined the effect of endogenous Lgl that was largely localized uniformly along the cortex with a cytoplasmic distribution in S2 cells, prior to Par3 overexpression (Figure 2A). When Par3 was distributed asymmetrically along the cortex, Lgl and Par3 distributed in a complementary manner (Figure 2B). Knockdown of lgl via RNAi and the expression of Lgl3A, which aPKC is not able to phosphorylate, showed that Lgl and its phosphorylation by aPKC are required for asymmetric Par-complex localization in S2 cells (Figure 2C,D). We also confirmed that the other two components of Par-complex, Par6 and aPKC require lgl function to colocalize with Par3 along the cortex (Figure 2E,F).

Figure 2

Download asset Open asset

To evaluate the degree of polarization of transfected cells, we introduced the asymmetric index (ASI), a measure of the polarized Par-complex distribution, which, according to Derivery et al. (2015), indicates the degree of polarization of a fluorescent marker distributed along the equatorial cortex (Figure 3A). ASI distribution was compared with that of membrane-bound GFP (memGFP), which is essentially non-polarized (the control). The ASI value of memGFP ranged from 0 to 0.35 due to fluctuation. Par3 cortical distribution was categorized into two groups in comparison with memGFP. Cells with an ASI in the same range as that of memGFP were regarded as non-polarized. Those cells showing an ASI larger than that of the mem-GFP were interpreted as polarized. Among transfected cells showing cortical distribution of the Par complex, 39% were polarized, while 61% were non-polarized (Figure 3B).

Figure 3

Download asset Open asset

Figure 3—source data 1

Source data for the histogram in Figure 3.

https://doi.org/10.7554/eLife.45559.005

Download elife-45559-fig3-data1-v1.xlsx

Figure 3—source data 2

The individual histogram in each experiment in Figure 3.

https://doi.org/10.7554/eLife.45559.006

Download elife-45559-fig3-data2-v1.pdf

Furthermore, we examined the three-dimensional (3D) distribution of the Par complex in S2 cells by reconstructing serial images at steady state 2 days following transfection. Interestingly, the region where the Par-complex accumulated was not uniform but consisted of multiple large aggregates (Figure 3C). These large aggregates were termed ‘Par-islands’. These islands dynamically changed their arrangement on the surface of the S2 cells (Video 1).

Video 1

Download asset

We found that island structures of the Par-complex are also observed in the apical domain of Drosophila mitotic neuroblasts, where the Par-complex is asymmetrically localized (Figure 3D), suggesting that formation of Par-islands may be a common process during the polarized cortical distribution of the Par-complex irrespective of cell cycle phases.

To investigate temporal patterns of polarized distribution of the Par-complex, we induced Par3 expression via the Metallothionein promoter, which is activated in the presence of CuSO₄. Because the expression level of Par3 shortly after induction might be too low to visualize, we designed the experiments to monitor the cortical Par complex formation with Par6-GFP that had been expressed by the actin-promoter. To test rationality of this experimental design, we transfected S2 cells with pMT-myc-Par3-mKate2, pAct-Par6-GFP and pAct-aPKC. Because expression of Par6-GFP and aPKC reached steady levels 2 days following transfection (Figure 4—figure supplement 1), induction of Par3-mKate2 was initiated at this time. When the distribution of Par6-GFP and Par3-mKate2 along the S2 cell cortex were examined at 3 hr post-induction, where small Par-islands have been formed (Figure 4), we confirmed that both proteins show a nearly identical cortical pattern in islands and dots (Figure 4B,C,E) with a minor difference in the local fluorescent intensity gradient between them. We also checked the relationship of this overlapping cortical distribution of Par3 and Par6 with that of Cdc42, which is known to form a complex with aPKC-Par6 separately from Par3-Par6-aPKC complex. We confirmed cortical distribution of Cdc42 in a punctate or short string shape (Figure 4D) (Slaughter et al., 2013; Sartorel et al., 2018). Cdc42 distribution appeared to be independent of those of Par3-mKate2 and Par6-GFP, and only a small fraction of Cdc42 dots resided at the edge of Par-islands (Figure 4F,G). Thus, we concluded that monitoring Par complex formation with Par6-GFP works properly. Using this monitoring system, we observed that the Par-complex first emerged in dot form (designated Par-dot) in the cell cortex during the initial phase of myc-Par3-mKate2 elevation, 2–3.5 hr following induction (Figure 5A–C, and Video 2), as well known in C. elegans one-cell embryos. Interestingly, Par-dots emerged in a restricted region of the cell cortex (Figure 5C).

Figure 4 with 1 supplement see all

Download asset Open asset

Figure 5

Download asset Open asset

Video 2

Download asset

Par-dots continued to grow in size via self-expansion, repeated fusion and less frequent fission (Figure 5D). These then developed into ‘islands’ of various sizes and shapes 2.5–6 hr following induction (Figure 5A,B), as observed in the UAS-Gal4 system (Figure 3C). Par-islands dynamically developed by fusion with neighboring islands or dots as Par-dots did, and kept changing their shape and mutual position (Figure 5E). During this period, the distribution of Par-islands occurred via two separate processes, polar and non-polar clustering (Figure 5A,B), which corresponded with temporal changes in ASI values (Figure 6A). However, there was no significant difference between polarized cells and non-polar cells in either the time course of Par3-mKate2 expression levels or the Par6-GFP/Par3-mKate2 ratio (Figure 6B–D). While the average steady state amount of Par3-mKate2, driven by the Metallothionein promotor was approximately 1/16-fold of that driven by the Act-Gal4 >UAS system (Figure 6E, Figure 6—figure supplement 1), the appearance of the islands was similar between the two expression systems (Figures 3C and 5A). The typical process toward Par-complex clustering is summarized in Figure 6F.

Figure 6 with 1 supplement see all

Download asset Open asset

Figure 6—source data 1

Source data for the temporal change of S2 cells.

https://doi.org/10.7554/eLife.45559.014

Download elife-45559-fig6-data1-v1.xlsx

Once the Par3-mKate2 expression level steadied about 8 hr following induction, approximately 30% of cells with Par6-GFP localized cortically demonstrated polarized distribution of Par6-GFP, resulting in a crescent in the equatorial plane (ASI > 0.35), while the rest of cells showed a non-polarized cortical distribution (ASI ≦ 0.35); (Figure 7A). At steady state, Par-islands dynamically change their mutual positions (Figure 7B, and Video 3). However, their asymmetric clustering and non-polar distribution were largely maintained for at least several hours, once cells reached steady state (Figure 7C), suggesting that both polar and non-polar clustering of Par-islands was fairly stable. In both states of Par-island clustering, the distribution of Par-islands and Lgl in S2 cells were mutually exclusive (Figure 7D,E, and Videos 4 and 5), suggesting that Lgl plays a role in the stability of the two states. Interestingly, Par-islands were never unified into one large island regardless of their dynamic movements as well as fusion and fission (Figure 7B).

Figure 7

Download asset Open asset

Figure 7—source data 1

Source data for the histogram in the steady state.

https://doi.org/10.7554/eLife.45559.016

Download elife-45559-fig7-data1-v1.xlsx

Video 3

Download asset

Video 4

Download asset

Video 5

Download asset

Upon mitosis, cortical Par-islands disassembled and disappeared (Figure 8A). However, just prior to cell cleavage, small Par-dots reappeared preferentially in the region opposite the cleavage site (n = 11/14). Consistently, the position of the centrosome, which is normally located on the far side from the cleavage site, coincided weakly with the region of Par-dot reappearance (Figure 8B–D).

Figure 8

Download asset Open asset

Figure 8—source data 1

Source data for the position of centrosome.

https://doi.org/10.7554/eLife.45559.021

Download elife-45559-fig8-data1-v1.xlsx

A unique feature of the island form of the Par-complex was a slightly convex shape (Figures 5A,B and 7B, and Video 3). To quantify the curvature of Par-islands in comparison with the membrane curvature of S2 cells, we utilized memGFP driven by the actin-promoter to visualize the cell membrane, together with myc-Par3-mKate2 induced by the metallothionein-promoter (Figure 9A). The images for quantification were taken at 8 hr after induction. The curvature radius for individual islands was obtained from three different points along the plasma membrane on individual islands in the equatorial plane of the cell. On the other hand, the curvature radius of the plasma membrane was calculated from the three different points along the cell membrane of 5 μm long. For this measurement, we chose the ‘non-neighboring region’ defined as the regions along the cell membrane more than 5 μm away from the edge of neighboring Par-islands. This is to avoid the influence of the presence of Par-islands (see Materials and methods section for details). As shown in Figure 9B, the curvature radii of Par-islands were in the range 0.7 μm to 5 μm with the median of 1.70 μm. (n = 69 islands in 34 cells). In contrast, the majority of curvature radii for the plasma membrane span in the range between 2 μm and 20 μm with the median of 4.07 μm in the non-neighboring regions (n = 87 plasma membrane regions in 29 cells). We also measured the curvature radii of non-Par-islands cells lacking Par-island formation, which gave us the median of 4.56 μm with the range of 2 μm and 20 μm (n = 76 regions in 13 cells). These results indicate that the curvature radius of Par-islands is significantly smaller than that of cell membrane curvature in non-neighboring regions, which is virtually consistent with the radius of S2 cells that we used for experiments. Accordingly, Par-islands show a more convex shape than the ordinary plasma membrane curve of S2 cells.

Figure 9

Download asset Open asset

Figure 9—source data 1

Source data for the membrane curvature.

https://doi.org/10.7554/eLife.45559.023

Download elife-45559-fig9-data1-v1.xlsx

To better understand the organization of Par-islands, we investigated their structure at the super-resolution level. We performed super-resolution radial fluctuation (SRRF) analysis (Gustafsson et al., 2016), using confocal images of fixed samples, double-stained by Par3-mKate2 and Par6-GFP. This analysis revealed that both Par3-mKate2 and Par6-GFP exhibited polygonal shaped islands of various types (Figure 10A,B). In order to determine whether there was regularity in these structures, we measured contour lengths of the Par3-mKate2-stained meshwork, including separate rods and polygons. This measurement yielded distribution of lengths with multiple peaks in the density plot. We then searched the regularity of these multiple peaks. We found that it was well fitted with a combination of seven Gaussian curves, which exhibited a peak interval of 0.38 ± 0.06 μm (mean ± s.d., hereafter); (Figure 10C,D,F). We also performed spectral analysis of the density plot, and obtained a single major frequency of 2.4 μm^–1, which gives a peak interval of 0.42 μm (Figure 10E). These two analyses thus give consistent results with each other.

Figure 10

Download asset Open asset

Figure 10—source data 1

Source data for the contour line length of SRRF images.

https://doi.org/10.7554/eLife.45559.025

Download elife-45559-fig10-data1-v1.xlsx

We also observed Par-islands of Par3-GFP and Par6-GFP separately via STED microscopy (Figure 11A) and found similar meshwork structures in the deconvoluted STED images (Figure 11B–E). The linear part of segments in the meshwork structures were measured (Figure 11C), and exhibited a length of 0.39 ± 0.09 μm (Figure 11F). Thus, these two methods essentially provided the same value for the segmental length. In addition, these segments had a fairly homogeneous diameter in STED microscopic images, where the mean half width of Par-segments was 0.22 ± 0.03 μm, (Figure 11G,H). These results raise the possibility that the Par-island meshwork contains a unit segment. Indeed, separate rod- or string-shape structures as well as open square structures were often observed in the earlier phases of the Par-complex aggregation time course (Figure 11I, and Video 2), supporting the notion that Par-islands are assembled from these elemental structures, generating regularity in the meshwork organization.

Figure 11

Download asset Open asset

Figure 11—source data 1

Source data for the length of segment of STED images.

https://doi.org/10.7554/eLife.45559.027

Download elife-45559-fig11-data1-v1.xlsx

Because the elevation of Par3 expression induced cortical polarization in S2 cells, we investigated the role of functional domains of Par3 by observing phenotypes with Par6-GFP following the overexpression of mutant Par3 forms via the Metallothionein promoter (Figure 12A). First, we tested the role of the CR1 domain responsible for self-polymerization in the polarized Par-complex assembly (Benton and St Johnston, 2003; Mizuno et al., 2003).

Figure 12 with 2 supplements see all

Download asset Open asset

Figure 12—source data 1

Source data for the ASI and island size of Par3ΔCR1 mutant.

https://doi.org/10.7554/eLife.45559.032

Download elife-45559-fig12-data1-v1.xlsx

Overexpression of Par3 lacking CR1(Par3ΔCR1) in the presence of the endogenous Par3 compromised the cortical Par-complex assembly significantly. The Par-complex was broadly distributed over the cell cortex in the initial stages, when dots were very faintly visible. Although brighter fluorescence spots (similar to Par-islands) formed later, they were relatively smaller than the Par-islands in size, and mostly faint with ambiguous contours, compared to those formed by wild-type Par3 expression (Figure 12B). Eventually, distribution of these Par-aggregates was little polarized (Figure 12C). These results suggested that the CR1 domain was important for all processes during the development of macro-scale structures of the Par-complex.

However, the endogenous Par3 expression may contribute to this phenotype under this condition. Therefore, we next examined the effect of Par3ΔCR1 overexpression after knocking down the endogenous Par3 by RNAi (Figure 12—figure supplement 1). RNAi treatment reduced the endogenous Par3 expression to 10% of the normal Par3 expression level. As the next step, we introduced silent mutations into the RNAi sequence part of the full-length Par3 and Par3ΔCR1 in the overexpression constructs to prevent the expression of these exogenous proteins from being knocked down by RNAi. We confirmed that the expression levels of the exogenous proteins bearing silent mutations, myc-Par3sm-GFP and flag-Par3ΔCR1sm-GFP, were not affected by simultaneous RNAi treatment of cells, and that the expression level of both exogenous proteins were nearly at the same level (Figure 12—figure supplement 2).

We first examined the distribution of myc-Par3sm-GFP and flag-Par3ΔCR1sm-GFP by STED super-resolution microscopy. We confirmed a various size of Par-islands including large ones with a regular meshwork structure when full-length myc-Par3sm-GFP was overexpressed (Figure 12D, compare Figure 11B). On the contrary, in the cells overexpressing flag-Par3ΔCR1sm-GFP, we often observed bright spots with an amorphous shape, and could rarely find clear island structures with a regular meshwork structure (Figure 12E). Thus, our STED observations strongly suggest that in the absence of Par3 CR1 domain, Par-islands consisting meshworks with unit-like segments were not formed.

We also measured the size of islands in S2 cells expressing Par3sm and that of bright regions of Par3ΔCR1sm at the steady state after RNAi suppression of the endogenous Par3 (see Materials and Materials and methods for details). The results show that the distribution of spot size after Par3ΔCR1 overexpression, which was independent of the knockdown of endogenous Par3 (data not shown), was quite different from that of Par-islands formed by the overexpression of Par3 (Figure 12F,G); approximately 90% of bright spots are smaller in area than 1 μm² (the mean value = 0.44 ± 0.37 μm² (s.d.)). The size distribution of bright spots of Par3ΔCR1 was indistinguishable from most bright spots of Par-dots 2.5 hr after myc-Par3-GFP induction, when only a small fraction becomes Par-islands (compare Figure 12G and H). These observations raise the possibility that Par3ΔCR1 still has an ability to form some cortical aggregates, either via (1) an un-identified domain in Par3 or (2) via Par6-aPKC, or (3) via the other components, while the development of Par3ΔCR1-containing aggregates is very limited.

All together, we conclude that the CR1 domain of Par3 is critical for Par-complex to form Par-islands with the regular meshwork and to asymmetrically localize in S2 cells. This conclusion is consistent with the previous finding that CR1 domain is an essential domain of the Par3 function to form the normal asymmetry in the cell cortex in the epithelium (Benton and St Johnston, 2003).

We next examined the effect of aPKC-dependent phosphorylation at Serine 980 in the CR3 domain, which is necessary for dissociation of Par3 from aPKC (Figure 13A) (Morais-de-Sá et al., 2010). Overexpressing the non-phosphorylatable form, Par3S980A, which tightly binds aPKC (Morais-de-Sá et al., 2010), rapidly promoted aPKC complex aggregation, and the polarized region initially assumed a bowl-like shape, in which clustering of the Par-islands was so tight under this condition that the island structure was not easy to be discriminated (see images at 3 hr 20 min in Figure 13B). This dense aggregation gradually separated into small and nested islands, and eventually resulted in an increase in the polarized cell population (4 hr 20 min in Figure 13B, and Figure 13C); 40% of cells with cortical Par3 showed an ASI > 0.35, and a degree of polarization with a mean ASI value of 0.52 ± 0.13 for polarized cells (Figure 13C). The initial dense packing of the Par-complex containing Par3S980A suggested that the turnover of Par3-aPKC association and dissociation might play a role in the normal clustering of Par-islands. This was similar to that of Drosophila epithelial cells, wherein Par3S980A colocalized with aPKC-Par6 in the apical domain with disorganized adhesion belts (Morais-de-Sá et al., 2010).

Figure 13

Download asset Open asset

Figure 13—source data 1

Source data for the ASI of Par3S980A mutant.

https://doi.org/10.7554/eLife.45559.034

Download elife-45559-fig13-data1-v1.xlsx

Next, we examined the effect of the membrane association region (MAR) of Par3 by overexpressing Par3ΔMAR (Krahn et al., 2010). The Par-complex no longer localized cortically, but formed several cytoplasmic aggregates, which coalesced into a single large sphere (Figure 13D) as previously observed in the ooctye (Benton and St Johnston, 2003). Thus, the functional domains of Par3 and the interactions between these domains, together, play a role in the properly polarized distribution of the Par-complex in the S2 cell system.

Lastly, we examined the effects of the actin cytoskeleton on islands. While ROCK inhibitor, Y27632, did not significantly affect the behavior of Par-islands (data not shown), an actin inhibitor, Latrunculin B, changed the islands into a spherical shape, which frequently formed membrane protrusions (Figure 13E, and Video 6), suggesting that the actin-membrane skeleton is necessary to balance the surface tension of Par-islands (see Discussion).

Video 6

Download asset

In our reconstruction system, cortical asymmetry began with the formation and growth of cortical dot-like structures, which were also reportedly associated with anterior localization of the Par-complex in C. elegans zygotes (Wang et al., 2017; Dickinson et al., 2017; Munro et al., 2004). Par-dots in the S2 cell system included all three Par-complex components. Thus, these dots appear to be the common initial process of Par-complex cortical aggregation. In parallel with the growth of Par-dots, string-like structures often emerged. The subsequent process of asymmetric localization proceeded in the form of Par-islands with amorphous and dynamic behavior. To our knowledge, this type of Par-complex form with such a dynamic behavior has not been explicitly described in cortical Par-complex assembly in C. elegans one-cell embryo or Drosophila neuroblasts. We indeed found the island-like structures in the Par complex distribution in Drosophila mitotic neuroblasts (Figure 3D). We also realized that a similar behavior of Par aggregates is detectable in the movies of C. elegans one cell embryo, whereas it is not explicitly described (Wang et al., 2017). Furthermore, Par3 is known to show punctate localization at the early stage of Drosophila epithelialization (Harris and Peifer, 2005; Harris and Peifer, 2007). Par3 is also localized as a small patch form on the niche side of interphase female germline stem cells in Drosophila, while its behavior is unknown (Inaba et al., 2015). Based on these, we suggest that Par-island structure is not specific to this artificial apolar S2 cells but a universal form of Par-complex aggregates.

Interestingly, initial dot formation appeared to be biased toward the region opposite the cleavage point, where the centrosome also appeared to be located, which was consistent with a recent study on Drosophila (Loyer and Januschke, 2018; Januschke and Gonzalez, 2010; Jiang et al., 2015). Thus, the cleavage point and/or the centrosome may be a general positional cue for the initiation of Par-complex-dependent cell polarity. In this context, polarization process of the S2 cell system is likely to be cell-autonomous and dependent on the induction of polarity proteins, wherein the orientation of polarity appeared to be dependent on internal cue(s).

Whereas our synthetic polarity shares a similar meshwork organization of Par-complex aggregates with Drosophila mitotic neuroblasts, there are two major differences between them: (1) cell cycle phase at which polarity is formed and (2) time constant of polarity formation (see the following section ‘Dynamic behaviors of Par aggregation and mechanisms for clustering’ for the time constant). In Drosophila neuroblasts, Par-polarity starts to form G2 phase and mainly operates in the mitotic phase, disappearing once they enter G1 phase. In contrast, in our artificial polarity system of S2 cells, Par-polarity is formed during interphase, and disappears upon mitotic entry. This difference in cell cycle phase of polarization affects the spatio-temporal relationship of Par-complex and other components; Lgl distributes cortically throughout interphase and prophase but relocalizes to the cytoplasm during prophase. Thus, Lgl cortically distributes when cell polarity is artificially formed in our S2 system, hence their complementary distribution lasts as long as Par-complex cortically distributes (Figures 2 and 7D,E). In contrast, in neuroblasts, their complementary distribution occurs only in a short period within the mitotic phase when Par complex localizes asymmetrically in neuroblasts, because Lgl rapidly relocalizes to the cytoplasm during prophase.

As for the disappearance of the cortical Par complex at mitosis, our artificial polarity system might resemble epithelial cells, in which apical components including aPKC, Par3 and Crumbs disappear from the apical side during the mitotic phase (Bergstralh et al., 2013). Neuroblasts that lose adherens junctions during delamination from epithelial layer must acquire a specific mechanism to maintain Par-polarity during mitosis. The Par-complex is activated by AuroraA kinase (Wirtz-Peitz et al., 2008). A possibility is that different levels of such mitotic kinases including AuroraA and Polo kinase may cause the disappearance of Par-complex clustering at mitosis in the S2 cell system.

Par-complex assembly at the cortex of S2 cells appears to stabilize the cell membrane because membrane filopodia extensively formed in areas where Par-islands were absent (Video 2). Also, cell membrane curvature was higher where Par-islands were attached, compared with that of the surrounding areas (Figures 5A,B and 9B). Membrane curvature may be determined by the balance between elasticity of the cortical cytoskeleton, the affinity of the Par-complex toward the cell membrane, and possibly the surface tension of the Par-island. Membrane affinity of the Par-complex is mediated by Par3 MAR, which interacts with phosphoinositides (PIPs) (Krahn et al., 2010) and/or by Par6-cdc42 interaction (Joberty et al., 2000). The convex shape of the Par-island and its higher membrane curvature reflects its relatively high surface tension. This is supported by the fact that disruption of the actin cytoskeleton by Latrunculin B treatment leads to a curled or spherical Par-island, inducing dynamic cell membrane protrusions. This phenomenon may be explained as follows; disruption of the cortical cytoskeleton leads to the loss of its elasticity, which had balanced the surface tension of the Par-island. The resulting imbalance in surface tension may cause the Par-island to shrink into a bowl or sphere shape, thereby bending the cell membrane outward and conferring protrusive activity to the cell membrane. In contrast, when membrane affinity is quite low, as in the case of Par3ΔMAR, Par-island shape is not affected by either cortical cytoskeleton elasticity or membrane affinity, and its shape would be determined only by the surface tension of Par3-islands. Under these conditions, we found that the Par-complex forms small cytoplasmic droplets, which subsequently coalesce into a spherical, densely packed structure, suggesting that phase separation takes place between the Par-complex aggregates and the cytoplasm (Hyman et al., 2014). A similar cytoplasmic droplet formation for a C-terminal truncated form of Par3 has been observed in Drosophila oocytes (Benton and St Johnston, 2003).

In this study, we revealed that a Par-island is a meshwork of various polygonal shapes, which appear to be built up of unit-like segments with an average length of approximately 0.4 μm. Isolated fragments such as single fragments and structures made up of a few connected fragments were observed during the development of Par-islands via live-imaging. These observations suggested that these isolated fragments assembled into a meshwork to form islands. Par-islands change shape rapidly during their movement along the cortex, and sometimes fuse with small pieces and also release pieces of different sizes, raising the possibility that Par-islands and small free fragments are mutually exchangeable. The factors that determine the size of these unit segments need further investigation.

Par3 is known to polymerize in vitro via the CR1 domain at its N-terminus to form a helical polymer of 8-fold symmetry (Zhang et al., 2013; Feng et al., 2007). Whether Par3 polymers are involved in the cortical cluster of the Par complex remain unclear. Our super-resolution microscopic observations and the ability of Par3 to form filaments lead to the simple hypothesis that the unit segment of a Par-island is formed by Par3 polymers as the core structure. This hypothesis is compatible with our observation that the overexpression of Par3ΔCR1 no longer forms a regular meshwork of Par-complex aggregates. While there are many possibilities via which Par3 filaments may form a unit segment, a single Par3 polymer may form a single segment. Another possibility is for Par3 polymers to be aligned along the long axis of the segment. Since Par6 and aPKC bind the PDZ and CR3 domains of Par3, respectively, Par6 and aPKC can act as cross-linkers between segments (Feng et al., 2007). Given the phenotype of ParS980A overexpression, the association of Par3 and aPKC by aPKC phosphorylation may confer flexibility and dynamism to the structure and/or assembly of the segmental elements. These hypotheses need to be tested in future studies. A surprising finding is that Par3ΔCR1 still has an ability to form Par-complex aggregates while those aggregates show a small size with no regular meshwork structure. We speculate that even in the absence of the CR1 domain, Par6 and aPKC cross-link multiple Par3ΔCR1 to form small aggregates via their multiple binding sites on Par3.

We have reproducibly obtained the data showing no significant effect of Y27632 on the behavior of Par-islands. Because cortical flow involving Myosin II is a major driving force for C. elegans polarization (Munro et al., 2004; Motegi and Sugimoto, 2006), the dynamic behavior of Par-islands is not likely to depend on the typical Myosin II cortical flow. Moreover, we observed no directional movement of Par-dots or Par-islands during the process of asymmetric clustering nor at the steady state. These two observations suggest no significant involvement of cortical flow in Par-island dynamics in the S2 cell reconstruction system, while we do not exclude the possibility that Y27632-insensitive Myosin might work for the dynamics of Par-islands even though it is unlikely to operate in a directional way.

How can Par islands cluster asymmetrically and dynamically behave? A possible mechanism underlying Par-island dynamics is diffusion and/or tension. We think that the situation where advection such as cortical flow surpasses diffusion for transportation is the following two cases. Firstly, the scale of the system is large enough. C. elegans one-cell embryo and Drosophila oocytes are good examples. In these cases, the random diffusion and capture mechanism without advection will take a much longer time (Goehring et al., 2011). While C. elegans one-cell embryo is approximately 50 µm long along the anterior-posterior axis, the size of S2 cells is approximately 10 μm in diameter (Figure 9B), which seems to be nearly marginal about which of advection and diffusion is efficient to polarize cells. The mutual capture of islands might work well to cluster islands in one pole.

Secondly, the time scale of clustering in S2 cell system is an order of hours, while the polarization of C. elegans one-cell embryo or of Drosophila neuroblasts occurs in an order of minutes (within 15 min, and 5 min, respectively). Such a big difference in the time scale of Par-island polarization, together with its non-directional movement, is likely explained by the notion that polarized clustering of Par-islands occurs through diffusion and mutual capture that might be driven by phase separation with surrounding cortex.

In conclusion, while we do not exclude the involvement of myosin motor in Par island dynamics, it is more likely to be driven by diffusion mechanism on the two-dimensional cortex.

An interesting property of Par-islands is that they are not unified into one large island under the cell membrane, even when polarized. Overexpression of Par3ΔMAR or Par3S980A is an exception. In the latter case, rapid and enhanced formation of the cortical Par-complex does not seem to initially permit separate island formation, and a large, transient dome is formed instead. In the former case, the Par-complex aggregates to form one large sphere. This cytoplasmic phenomenon is likely to be due to a phase separation between the Par-complex and the cytoplasm. Considering this property of the Par-complex, the unique feature of Par-islands associated with the cell membrane may reflect phase separation in two dimensions.

Steady-state Par-island distribution in a cell may be classified into two different states, polarized and non-polarized. While we failed to identify a single parameter correlating these two states (Figure 5A,B), our analysis shows that the two states of island distribution are nearly fixed during the formation of islands (Figure 6A). Because the position of island formation appears to be stochastic, variation in the position of Par-island formation across the cell may explain the two localization patterns of Par-islands. Since Lgl distribution is largely complementary to dots and islands, this molecule may contribute to stabilize the two states of island distribution at the cellular scale (Betschinger et al., 2003; Guo and Kemphues, 1995). Thus, these two different states of Par-island distribution may be the outcome of two stable solutions of the reaction diffusion system (Chau et al., 2012; Goehring et al., 2011), where a negative regulator Lgl is involved (Betschinger et al., 2003). The initial condition, which is possibly determined by a stochastic distribution of islands, may select one of the two stable patterns in a cell. We propose that such cell-scale patterning is coupled with local phase separation of Par-islands as previously described for the membrane lipid domain (John and Bär, 2005).

In summary, we have developed a potential Par complex-polarization system upon induction of Par3 in non-polar S2 cells, which provides a useful model for cell-autonomous cell polarization, allowing us to easily manipulate gene expression and image at the super-resolution level. One intriguing challenge will be the coupling of mitosis with cell polarization in this system to induce asymmetric division.

S2 cell (Invitrogen:R69007) culture and transfection were performed as previously described (Ogawa et al., 2009). Expression vectors were transfected at 2 days prior to microscopic or western blot analysis. For induction of the Metallothionein promoter, 100 mM CuSO₄ solution was added to a medium at a final concentration of 1 mM.

Cells were mounted on a 35-mm glass-bottom dish coated with 15 µg/ml poly-L-ornithine and incubated at 25°C for 30 min, followed by microscopic analysis. Images were taken at a 1 μm z-interval with a spinning disk confocal microscopy CSU-W1 (Yokogawa, Tokyo, Japan) equipped with a sCMOS camera Neo (Andor, Belfast, Northern Ireland) and MetaMorph software (Molecular Devices, San Jose, CA).

For immunostaining of S2 cells, transfected cells were mounted on a poly-L-ornithine-coated cover slip and fixed with 4% paraformaldehyde in PBS for 15 min at room temperature. Cells were washed with PBS, followed by treatment with 0.1% Triton-X100 in PBS for 15 min. After washed with PBS, cells were treated with a blocking buffer containing 2% BSA in PBS for 30 min and incubated with primary antibodies in the blocking buffer for 30 min, followed by incubation with secondary antibodies for 30 min. Immunostained cells were embedded in mounting medium PermaFluor (Thermo Fisher Scientific) and analyzed with a confocal microscopes LSM510 (Zeiss, Oberkochen, Germany). For super-resolution microscopy, samples were embedded in ProLong Glass Andifade Mountant (Thermo Fisher Scientific) and analyzed with a comfocal microscopes LSM880 (Zeiss, Oberkochen, Germany).

For immunostaining of Drosophila neuroblasts, brains isolated from third instar larvae were fixed with 4% paraformaldehyde in PBS for 20 min at room temperature. Samples were treated with the blocking buffer for 1 hr, followed by incubation with primary antibodies and secondary antibodies for overnight and 2 hr, respectively. Samples were then embedded in Vectashield H-1000 (Vector Laboratories) and analyzed with a confocal microscope FV1000 (Olympus, Tokyo, Japan).

Primary antibodies used were anti-aPKC (rabbit polyclonal, used at 1:1000, Santa Cruz), anti-Par-3 (rabbit polyclonal, used at 1:1000, or mouse monoclonal, used at 1:100) (Ohshiro et al., 2000), anti-Par-6 (rabbit polyclonal, used at 1:1000) (Izumi et al., 2004), anti-Myc (chicken polyclonal, used at 1:1000, Bethyl), anti-Flag (mouse monoclonal, used at 1:1000), anti-Miranda (mouse monoclonal, used at 1:100) (Ohshiro et al., 2000), anti-Lgl (rabbit polyclonal, used at 1:1000) (Ohshiro et al., 2000), anti-GFP (chicken polyclonal, Betyl, mouse monoclonal, Chemicon, and rabbit polyclonal, MBL, used at 1:1000), anti-Cdc42 (guinea pig polyclonal, used at 1:100, a kind gift from U. Tepass, University of Toronto, Canada). Secondary antibodies used were anti-Rabbit Alexa Fluor488 (donkey polyclonal, Jackson Immuno Research), anti-mouse Alexa Fluor488 (donkey polyclonal, Jackson Immuno Research), anti-chicken Alexa Fluor488 (donkey polyclonal, Jackson Immuno Research), anti-rabbit Cyanin3 (Donkey polyclonal, Jackson Immuno Research), anti-mouse Cyanin3 (donkey polyclonal, Jackson Immuno Research), anti-chicken Cyanin3 (donkey polyclonal, Jackson Immuno Research), anti-rabbit Alexa Fluor647 (donkey polyclonal, Jackson Immuno Research), anti-guinea pig Cyanin5 (donkey polyclonal, Jackson Immuno Research). All secondary antibodies were used at 1:4000 or 1:200 for staining of S2 cells or Drosophila 3^rd instar larvae, respectively.

For the super-resolution radial fluctuations (SRRF) method, S2-cells were transfected with pMT-myc-Par3-mKate2, pAct-Par6-GFP and pAct-aPKC. For SRRF method, confocal imaging was performed using LSM880 (Zeiss) with an objective lens Plan-Apochromat 63x/1.4 Oil DIC M27 (Zeiss). A series of 200 frames was obtained for each cell with a pixel size of 53 nm and 160 ms exposure time. Drift-correction and reconstruction of SRRF images were performed with an ImageJ plug-in NanoJ-SRRF (Gustafsson et al., 2016).

Using SRRF-processed images, Par3 contour lengths along the meshwork were manually traced with Fiji. Each image was overlaid by an edge-enhanced image generated with the Sobel filter, to highlight Par3 contour shapes. Lengths between their terminal ends and/or branching points were measured. A histogram and a density plot were generated from all contour lengths, and the shape of the density plot was fitted with a linear combination of 7 Gaussian curves by a fitting function implemented in R with the non-linear least square method. Power spectral density of the second derivative of the density plot was calculated using fast Fourier transform method with R.

For Stimulated emission depletion (STED) imaging, S2-cells were transfected with pMT-myc-Par3-GFP, pAct-Par6 and pAct-aPKC, or pMT-myc-Par3, pAct-Par6-GFP and pAct-aPKC, followed by CuSO₄ addition for induction 2 days following transfection. Cells were fixed for immune-staining for GFP at 8 hr post-induction. STED imaging was performed using TCS SP8 STED 3X microscope (Leica, Wetzlar, Germany) with an objective lens HC PL APO 93X/1.30 GLYC (Leica). Deconvolution was performed with a deconvolution software package Huygens Professional (version 17.10, Scientific Volume Imaging, Hilversum, Netherlands).

Deconvoluted STED images were used for the analyses of Par3 segment lengths and widths. The segment length was defined as a shortest length between terminal ends, corners and/or branching points of Par3 contours, and manually traced with Fiji. The segment width was given by the full width at half maximum (FWHM) of a Gaussian-fitted signal distribution orthogonal to each Par3 segment.

The equatorial z-plane of each cell was analyzed for the estimation of asymmetric index (ASI) (see also Figure 3A). The cell perimeter was traced by a 0.5 µm-width line and the signal intensity along the line was measured with Fiji. The signal intensities were summed up along the half (L) of the total perimeter length (2L). The difference between this value and that of the other half was calculated and normalized by the total signal intensity along the perimeter. This measurement was done starting from every pixel along the perimeter (one pixel = 0.108 µm), The maximum value of them was defined as ASI. Cell with an ASI larger than 0.35 was defined as polarized cell. Each experiment to measure the distribution of ASI was independently duplicated, and the results of the two experiments were combined to make a single histogram for ASI distribution. The statistical significance of polarized cell population was analyzed by Fisher's exact test with post-hoc Bonferroni correction for multiple comparisons (Figures 3A, and 10E,F). Statistical analyses were performed with R software.

The equatorial z-plane of each cell was analyzed for the estimation of membrane curvature radius. The cell membrane was defined by membrane-GFP signal. A Par-island region was defined as a continuous compartment of the cell membrane that co-localized with continuous myc-Par3-mKate2. A non-neighboring region was defined in cells with Par-islands as a continuous membrane compartment of less than 5 µm at a distance of more than 5 µm from Par-island regions. A membrane region in non-island cells was defined as a continuous membrane compartment of less than 5 µm in cells without Par-islands. For non-island cells, 5–6 regions were analyzed per cell. Coordinates of two ends and a mid-point of each region were taken. The circle that passes through the three points was solved from the coordinates, and its radius was shown as a curvature radius in Figure 9. Measurements was performed with Fiji. Calculations and statistics were performed with R.

Bottom surface images of cells were analyzed to measure the size of Par aggregates. Region of interests (ROI) were made along the perimeters of Par aggregates by tracing manually with Fiji. Aggregates that contacted with a limit of an optical section were excluded from the analysis. The area of each ROI was measured, and shown as the size of Par aggregates in Figure 12.

Whole cell extracts of the untransfected S2 cells and the transfected S2 cells were subjected to SDS-polyacrylamide gel electrophoresis. Primary antibodies used were anti-Par3 antibody (rabbit polyclonal, used at 1:1000), anti-alpha-tubulin (rat monoclonal, Santa Cruz). Secondary antibodies used were horseradish peroxidase (HRP)-conjugated anti-mouse antibody (sheep polyclonal, used at 1:3000, GE Healthcare), HRP-conjugated anti-rabbit antibody (sheep polyclonal, used at 1:3000, GE Healthcare) and HRP-conjugated anti-chicken antibody (donkey polyclonal, used at 1:250, SA1-300, Thermo Fisher Scientific). Protein level was analyzed by chemiluminescence with Chemi-Lumi One L (Nacalai tesque, Kyoto, Japan) and quantified with an image analyzer LAS-3000 system (Fujifilm, Tokyo, Japan). To compare the expression level of the overexpressed fluorescent protein per cell between two different transfectants or with the endogenous Par3 proteins, transfection efficiency for each sample was calculated by counting fluorescence-positive cells and negative cells. The ratio of the expression level per cell was calculated by dividing the measured staining intensity on the western blot by the transfection efficiency.

Long double-stranded RNAs (dsRNAs) were used for RNA interference (RNAi) in S2 cells as previously described (Bettencourt-Dias and Goshima, 2009). For knocking-down Par6 or aPKC, dsRNA was synthesized with MEGAscript T7 Transcription Kit (Ambion, Thermo Fisher Scientific) according to the manufacturer’s instructions, by using pBS-T7/Par-6/T7 or pBS-T7/aPKC/T7 plasmid directly as a template, which contains the full-length Par6 or aPKC ORF flanked by two T7 promoters, respectively. For knocking-down Lgl or Par3, dsRNA was synthesized by using a PCR amplicon consisting of the full-length Lgl ORF or a part of the Par3 ORF (906–1226 nucleotides) flanked by T7 promoters as a template. PCR was performed by using pUAS-Flag-Lgl or pUAS-myc-Par3 as a template and primers shown in below:

Lgl, 5’-TAATACGACTCACTATAGGGATGGCAATAGGGACGCAAACAGGGGCTTTAAAAGTT-3’ and 5’-TAATACGACTCACTATAGGGTTAAAATTGGCTTTCTTCAGGCGCTGTTTTTGGCGTTCCAA-3’; Par3, 5’- TAATACGACTCACTATAGGGTGAATCCATCAGGGAGAAGG-3’ and 5’- TAATACGACTCACTATAGGGCTCGGCCACCTTAGAGTCAC-3’.

The dsRNA sequence for knocking-down Par3 was selected according to an RNAi database DRSC (DRSC25558) (Mohr et al., 2015).

dsRNAs were added to the culture media at a final concentration of 4.5 µg/ml at 2–3 hr following transfection of expression plasmids.

To construct expression vectors under control of an actin (act5c) promoter, Drosophila Par6, Par3 or aPKC ORF, or Par6, aPKC or Fyn myristoylation tag (5’- ATGGGCTGTGTGCAATGTAAGGATAAAGAAGCAACAAAACTGACG-3’) conjugated with GFP at the C-terminus (Par-6-GFP, aPKC-GFP, membrane-GFP) was inserted into pAc5.1/V5-His B plasmid (Invitrogen, Thermo Fisher Scientific).

To construct an expression vector for Par3 under control of the Gal4-UAS system, Par3 conjugated with Myc or Flag epitope and mKate2 at the N-terminus and C-terminus, respectively, or Lgl^3A conjugated with Flag epitope was inserted into pUAST plasmid (Brand and Perrimon, 1993), and Drosophila gal4 ORF, which was subcloned from pAC-GAL4 plasmid, was inserted into pDAMCS plasmid. To construct pDAMCS expression plasmid, BglII-XhoI fragment of pAct5C0 plasmid (Thummel et al., 1988) containing actin 5C promoter and poly(A) addition signals (and a small region of hsp70 promoter) was cloned between BamHI and SalI site of pUC19 plasmid. Then, a synthetic double-stranded oligonucleotide containing a multiple cloning site was inserted into the BamHI site between the actin 5C promoter and the poly(A) addition signals.

To construct expression vectors for Par3 under control of the induction system, Par3 conjugated with Myc epitope and mKate2 or GFP at the N-terminus and C-terminus, respectively, was inserted into pMT plasmid (Invitrogen, Thermo Fisher Scientific).

To construct expression vectors for RNAi-resistant form of Par3 (Par3sm), silent mutations that do not alter amino acids were introduced into the dsRNA-targeted region in the Par3 ORF as below, in which mutated nucleotides were underlined, 5’-TGAAAGCATTAGAGAAAAAG ACGGCGAGATGTTGTTGATTATAAATGAATATGGATCGCCCTTGGGCTTGACCGCCCTGCCCGATAAAGAACACGGAGGCGGTTTGCTGGTCCAGCACGTCGAACCCGGAAGTCGCGCAGAAAGGGGCCGCTTGCGCCGCGACGACCGCATTCTCGAAATTAACGGCATTAAATTGATCGGCCTGACGGAGAGCCAAGTCCAGGAACAGTTGCGCCGCGCCCTGGAGTCCAGCGAACTGAGGGTGCGGGTCCTCCGCGGCGACCGCAACCGCCGCCAGCAACGCGATTCCAAAGTCGCTGA-3’. A mutated DNA fragment (GeneArt, Thermo Fisher Scientific) was amplified by PCR and replaced with the corresponding region of Par3WT-GFP or Par3∆CR1-GFP in pMT plasmid by using In-Fusion.

pAC-GAL4 was a gift from Liqun Luo (Addgene plasmid # 24344; http://n2t.net/addgene:24344; RRID:Addgene_24344; Potter et al., 2010). pUbq-Spd2-GFP was a gift from Jordan Raff (University of Oxford, UK). Plasmids used for transfection were purified with Wizard Plus SV Minipreps DNA Purification System (Promega, Madison, WI) or NucleoBond Xtra Midi (Macherey-Nagel, Düren, Germany).

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank Y Tsunekawa for technical advice, Y Otsuka for technical assistance, U Tepass for ant-Cdc42 antibody, J Raff for plasmids, T Nishimura for fly strains, S Hayashi, T Nishimura, and K Kawaguchi, and S Yonemura for critical discussion. Some of the imaging experiments were performed at the RIKEN Kobe Light Microscopy Facility. IF is a recipient of RIKEN Special Postdoctoral Researcher Program.

© 2019, Kono et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.