Sds22, a PP1 phosphatase regulatory subunit, regulates epithelial cell polarity and shape [Sds22 in epithelial morphology]

Background How epithelial cells adopt their particular polarised forms is poorly understood. In a screen for genes regulating epithelial morphology in Drosophila, we identified sds22, a conserved gene previously characterised in yeast. Results In the columnar epithelia of imaginal discs or follicle cells, mutation of sds22 causes contraction of cells along their apical-basal axis, resulting in a more cuboidal morphology. In addition, the mutant cells can also display altered cell polarity, forming multiple layers in follicle cells and leaving the epithelium in imaginal discs. In yeast, sds22 encodes a PP1 phosphatase regulatory subunit. Consistent with this, we show that Drosophila Sds22 binds to all four Drosophila PP1s and shares an overlapping phenotype with PP1beta9c. We also show that two previously postulated PP1 targets, Spaghetti Squash and Moesin are hyper-phosphorylated in sds22 mutants. This function is shared by the human homologue of Sds22, PPP1R7. Conclusion Sds22 is a conserved PP1 phosphatase regulatory subunit that controls cell shape and polarity.


Background
Epithelial tissues are composed of polarised cells connected by adherens junctions to form continuous sheets with apical and basal surfaces. How epithelial cells maintain their polarity, adhesion and shape remains poorly understood. Polarity in epithelia is founded on the segregation of determinants into apical and baso-lateral membrane domains. Adherens junctions are located at the interface of these domains and connect the actin cytoskeleton to neighbouring cells. Actin filaments are visible around the entire plasma membrane, but a particularly prominent belt of actin filaments runs around the apical cortex, overlapping with the ring of adherens junctions. This apical contractile bundle of actin filaments is likely to be a critical element in organising the polarised form of epithelial cells (reviewed in [1]). Molecules regulating the spatial organisation of the actin cytoskeleton and generation of forces upon it are therefore of particular interest.
Erzin-Radaxin-Moesin (ERM) proteins link actin filaments with the plasma membrane and are necessary to organise the cortical actin cytoskeleton (reviewed in [2]).
Drosphila has a single ERM family member, Moesin, that is essential for maintenance of epithelial cell polarity and shape. Cells lacking Moesin are unable to maintain their polarised form, disassemble adherens junctions and leave the epithelium, ultimately undergoing apoptosis [3].
Myosin II can slide two actin filaments against each other to create tension (reviewed by [4]. This is the basis for muscle contraction in skeletal muscle, but also has important force-generating roles in non-muscle cells. Drosophila has a single non-muscle myosin II heavy chain encoded by zipper (zip) and a single non-muscle myosin II regulatory light chain (MRLC) encoded by Spagetti-Squash (Sqh). Analysis of mutant alleles of these genes has revealed that non-muscle myosin II is required for maintenance of epithelial cell shape, as well as other processes involving dynamic cell shape changes such as gastrulation movements and cytokinesis [5][6][7][8][9].
Both Moesin and Myosin II are activated by phosphorylation and concentrated at the apical membranes of Drosophila epithelial cells. Several kinases that phosphorylate Moesin (Slik kinase; [10]) and Sqh/MRLC (Rho kinase; see for example [9] have been identified). One of the four Drosophila PP1 phosphatases, PP1β9c, has been shown to antagonise Sqh/MRLC phosphorylation [11]. Here, we identify Sds22, a PP1 phosphatase regulatory subunit, that binds to all four Drosophila PP1 phosphatases and restricts the activity of both Sqh/MRLC and Moesin. We show that loss of Sds22 has a similar but stronger phenotype than loss of PP1β9c, disrupting both epithelial cell shape and polarity.

Results sds22 is required for epithelial morphology in imaginal disc epithelia
In a PiggyBac transposon-mutagenesis screen in the Drosophila eye imaginal disc, we recovered an insertion immediately upstream of the start codon of sds22 (PB1173) which disrupts epithelial morphology, causing lethality (Additional Fig 1). The lethality and mutant phenotypes of this allele were reverted when the PiggyBac transposon was excised. The PiggyBac transposon allows transcription (data not shown) but is expected to prevent translation of the sds22 mRNA. The sds22 PB1173 phenotype was also rescued by expression of a UAS.sds22-GFP transgene (Additional Fig 2). Finally, expression of a UAS.sds22-IR transgenic RNAi line (VDRC 11788) produced phenotypes highly similar to that of sds22 PB1173 (data not shown). Although we have not specifically established that sds22 expression was reduced or absent in the insertion mutant, we presume that this is the case as the mutant is rescued by an sds22 transgene and phenocopied by Sds22-RNAi. Together, these results show that the observed phenotypes are caused by disruption of the sds22 gene.
To examine the effect of this mutant more closely, we induced sds22 mutant clones in the developing wing imaginal disc, a commonly used model system for studying the growth and morphology of clones of cells. The wing disc is composed of a columnar epithelium, a pseudostratified monolayer, that is continuous with an overly-sds22 is required to maintain epithelial morphology in imagi-nal discs Figure 1 sds22 is required to maintain epithelial morphology in imaginal discs. Mitotic recombination clones were induced in developing wing imaginal discs with the hs.FLP system and marked by the absence of GFP. (A) Wildtype clones induced 84 hrs after egg laying (AEL) and allowed to grow for 2 days. (B) sds22 mutant clones induced 84 hrs AEL grow to a similar size as wildtype clones after 2 days. (C) sds22 mutant clones induced 60 hrs AEL are largely eliminated after 3 days. (D) Wildtype clones induced in a Minute background at 84 hrs AEL grow to large sizes after 2 days. DAPI marks nuclei. (E) sds22 mutant clones in a Minute background induced at the same developmental stage (84 hrs AEL) grow to smaller sizes and exhibit morphological defects after 2 days. DAPI marks nuclei. (F) sds22 mutant clones induced earlier in development (60 hrs AEL) and allowed to grow for 3 days are eliminated and often leave ectopic folds behind. DAPI marks nuclei. (G) sds22 mutant clones induced at 84 hrs AEL and allowed to grow for 2 days. The morphogenesis defect in mutant clones was visualised by staining for phalloidin and DAPI. The arrow points to an island of wild-type tissue surrounded by deeply folded sds22 mutant tissue. (H) Phalloidin channel of (G). (I) DAPI channel of (G).
ing peripodial epithelium, a squamous monolayer ( Fig  1A). We induced GFP negative clones of wild-type ( Fig  1A) and sds22 mutant (Fig 1B and 1C) cells and allowed the clones to grow for two (Fig 1A and 1B) or three days ( Fig 1C). The sds22 mutant clones survived and grew similarly to wild-type clones for the first two days (Fig 1A and 1B) but were eliminated by three days (Fig 1C). This phenotype suggests that the sds22 mutant cells are sensitive to cell competition, a phenomenon whereby weaker cells are killed off by their more strongly growing neighbours (see, for example, [12]).
We therefore gave sds22 mutant clones a growth advantage over their neighbours with the Minute technique. This technique slows the growth of cells neighbouring the clone by making these cells heterozygous mutant for a Minute gene, encoding a ribosomal subunit. With this method, wild-type clones grow to very large sizes ( Fig 1D) after 2 days. In contrast, sds22 mutant cells formed smaller clones over the same time period and exhibited defects in epithelial morphology (Fig 1E). After 3 days, the sds22 mutant cells were still largely eliminated from the tissue, being extruded basally from the epithelium and undergoing apoptosis (Fig 1F). Extrusion of the sds22 mutant cells left behind dramatic folds in the epithelium (Fig 1F), more easily appreciated when the filamentous actin cytoskeleton was visualised by staining with phalloidin ( Fig 1G, H To examine the cellular basis for these phenotypes, we examined confocal cross-sections of discs carrying sds22 mutant clones (Fig 2). By examining both smaller, younger clones (Fig 2A; marked by absence of GFP) and larger, older clones ( Fig 2B; marked by absence of GFP) we observed a progressively stronger phenotype. In smaller clones, the mutant cells appear abnormally short in their apical-basal axis, adopting a more cuboidal morphology than their pseudostratified columnar neighbours (Fig 2A). Although the mutant cells in these small clones are abnormally short, they retained their polarised epithelial character, with normal localisation of the adherens junction component, Armadillo (Arm; Fig 2A). In larger clones, abnormally short mutant cells were also visible, but, more dramatically, infolding and extrusion of the mutant cells created islands of wild-type epithelium, surrounded by mutant cells (Fig 2B; also visible in Fig 1G,  arrow). A large number of pyknotic nuclei, indicating apoptotic cells, were visible in the sds22 mutant clones (Fig 2B), some of which appeared to have left the epithelium entirely (Fig 2B, arrow). These results suggest that sds22 mutant cells first change shape, adopting a more cuboidal morphology, and later leave the epithelium and apoptose, leaving behind a deep infolding of the tissue.
We next tested whether extrusion of sds22 mutant cells from the epithelium was a cause or consequence of their apoptosis. We therefore prevented apoptosis of sds22 mutant cells by expression of the baculovirus caspase inhibitor, p35. We found that mutant cells were still extruded from the epithelium, collecting as a ball of round cells on the basal side of the disc (Fig 2C; positively marked by expression of RFP). These results show that sds22 is essential to maintain the epithelial integrity of sds22 mutant cells change shape and are extruded from imag-inal disc epithelia wing imaginal disc cells and suggest that extrusion of mutant cells is the cause of apoptosis.

sds22 is required for epithelial morphology in ovarian follicle cells
To test whether the sds22 is specifically required in imaginal discs or more generally required in epithelia, we examined the follicular epithelium of Drosophila egg chambers ( Fig 3A). As in the wing, sds22 mutant clones showed a spectrum of defects in epithelial morphology, depending on clone size. In smaller clones, cells appear abnormally contracted along their apical-basal axis ( Fig 3A). In many larger clones, cells rounded up and tended to form double layers ( Fig 3B&C). When markers of apical-basal polarity were analysed in sds22 mutant clones (not shown) or in clones expressing RNAi against sds22 (UAS.sds22-IR; Fig  3D, E&F, clones marked by GFP expression), abnormal spreading of baso-lateral markers (anti-Dlg or anti-α-Spectrin 3A9 staining) around the cell into apical regions was detected. In some clones, the apical marker aPKC appeared relatively normal (Fig 3E), while in others aPKC staining appeared reduced and more diffuse (Fig 3D&F). These results indicate that follicle cell epithelia deficient for sds22 show both cell shape and polarity defects.

sds22 mutant cells have a defective actin cytoskeleton
We next examined the actin cytoskeleton in sds22 mutant cells in egg chambers. Staining with an antibody recognising both monomeric G-actin and filamentous F-actin, we found a striking punctate accumulation of actin in mutant cells in both the follicular epithelium and giant nurse cells (Fig 4A & B). A similar accumulation of F-actin was not observed with Phalloidin staining (Fig 3A, 4C) indicating that G-actin is preferentially accumulating in sds22 mutant cells. Notably, in the nurse cells, breakdown of the plasma membrane and underlying cortical actin cytoskeleton led to multinucleated nurse cells (stained with Phalloidin; Fig 4C). Strikingly, these sds22 mutant phenotypes, together with the apico-basal contraction phenotype described above, are similar to phenotypes reported for a mutant in the PP1β9c phosphatase [11]. sds22 encodes a highly conserved PP1 phosphatase regulatory subunit Experiments in yeast and mammalian cells suggest that Sds22 binds to and regulates PP1 phosphatases. We therefore tested whether this is also the case for Drosophila Sds22. Drosophila has four PP1 phosphatases, named after their isotype and cytological location: PP1 α96a, β87b, β13c and β9c. We expressed HA-tagged versions of each of these PP1s with GFP-tagged Sds22 in Drosophila cells and subjected cell lysates to immunoprecipitation with anti-HA antibodies (Fig 5A). GFP-tagged Sds22 was efficiently co-precipitated when any of the four HA-tagged PP1s was co-expressed. In the absence of HA-tagged PP1s, only a small background level of GFP-Sds22 associated with HA-sds22 is required to maintain epithelial cell shape and polarity in follicle cells beads. The results indicate that Sds22 binds to PP1s, confirming results in yeast and mammalian cells.

Phosphorylation of two potential PP1 targets, Sphaghetti Squash and Moesin, is increased in sds22 mutant cells
We next sought evidence that Sds22 is required for PP1 phosphatase activity in Drosophila. Previous work suggested that Sqh/MRLC is a target of PP1 β9c, because levels of phospho-Sqh are elevated in PP1 9c mutant cells [9,11]. In addition, evidence from mammalian cells suggested that the same phosphatase that targets Myosin II regulatory light chain may also target Moesin (Moe), the sole Drosophila ERM protein [13]. We therefore examined levels of phospho-Sqh and phospho-Moe in sds22 mutant cells by immunostaining with phospho-specific antibodies. Levels of both phospho-Sqh ( Fig 5B) and phospho-Moe ( Fig 5D) were elevated relative to total Sqh ( Fig 5C) and Moe (Fig 5E) in sds22 mutant cells. Note that the excess phospho-Sqh and phospho-Moe staining accumulates on both apical and baso-lateral membranes in mutant cells (in contrast to their apical concentration in wild-type cells). Thus, sds22 is required to restrict the phosphorylation of both Moe and Sqh/MRLC.

Conserved function of Sds22 in human cells
Sds22 is widely conserved in the animal kingdom, including a mammalian homologue PPP1R7. We therefore sds22 mutant cells accumulate G-actin and exhibit break-down of F-actin investigated if PPP1R7 performed similar functions in mammalian cells to Sds22 in Drosophila. A431 cells are keratinocytes that retain adherens junctions and grow in 'quasi-epithelial' clusters despite being transformed. First, we investigated the morphology of A431 cells lacking PPP1R7. Control A431 cells formed colonies with an average of 10 cells (Fig 6D&E) whereas cells depleted for PPP1R7 formed smaller clusters usually consisting of 4-6 cells (Fig 6D&E). Thus, PPP1R7 helps to maintain the cohesion of groups of human cells of epithelial origin.
We next tested whether the PPP1R7 knockdown phenotype reflected a conserved function in regulating the phosphorylation status of Ezrin, Radixin and Moesin (hereafter refered to as ERM proteins) and Myosin Light Chain (MLC). Depletion of PPP1R7 with siRNA oligos caused a dramatic increase in the phosphorylation of both ERM proteins and MLC (on Ser19; Fig 6A, B & E). This striking result indicates that the biochemical function of Sds22/ PPP1R7 is conserved between Drosophila and mammals. To verify the RNAi effect, we measured depletion of PPP1R7 mRNA levels in A431 cells by quantitative RT-PCR (Fig 6C). The mild variations observed between the extent of knockdown and phenotypic strength with different siRNA oligos are likely to result from experimental differences between the staining and PCR.

Discussion
Our results show that Sds22, a protein previously identified as a PP1 phosphatase regulatory subunit in yeast, is essential for PP1 function in Drosophila tissues. We have also shown that the phosphorylation state of both Moesin and Sqh/MRLC depends upon the Sds22/PP1 phosphatase. Our experiments do not prove that Moesin and Sqh/MRLC are direct substrates of Sds22/PP1, but this is a distinct possibility, as previous work has shown that Sqh/ MRLC can be found in a complex with PP1. We have shown that this function of Sds22 is conserved in mammalian cells, consistent with previous biochemical evidence that PPP1R7 (the mammalian Sds22 homologue) binds to PP1 proteins.
Moesin and Sqh/MRLC are key regulators of the actin cytoskeleton. Our work favours a model in which restricted activation of Moesin and Sqh/MRLC maintains an apical contractile bundle of actin filaments that is essential for the shape and integrity of epithelial cells. Both Moesin and Sqh/MRLC are activated by phosphorylation at the apical membranes of epithelia. In the case of Moesin, this phosphorylation is essential for epithelial integrity and depends on the apically localised kinase Slik [10]. In the case of Sqh/MRLC, this phosphorylation is essential for cells to maintain their columnar shape and depends on Rho kinase [9]. Our work indicates that this restricted activation of these proteins at the apical mem-brane is complemented by dephosphorylation of Moesin in other parts of the cell by ubiquitous PP1 phosphatases containing the regulatory subunit Sds22. In sds22 mutants, these actin regulators are activated along the entire cell membrane at high levels, leading to an abnor- The function of Sds22 is conserved in human cells mal contraction of the cells along their apical-basal axis. A similar, but milder, phenotype is visible in PP1 9c mutant clones [11] and in clones overexpressing a phosphomimetic form of Sqh/MRLC [14,15].
A second phenotype observed upon depletion of Sds22 is a loss of apical-basal polarity, associated with cells rounding up and forming multi-layers. Baso-lateral polarity markers appeared to be more strongly affected than apical markers in follicle cell epithelia. In imaginal discs, loss of polarity in sds22 mutant clones would explain the extrusion and apoptosis of these cells. Importantly, both the loss of polarity in follicle cells and the extrusion of imaginal disc cells were most commonly visible in larger, older clones, indicating that this phenotype takes longer to manifest than the abnormal cell shape phenotype. This raises the possibility that Sds22 may not be a direct regulator of cell polarity, but rather is required for polarity as a consequence of its regulation of Sqh/MRLC and Moesin. Alternatively, there may be additional targets of Sds22 that regulate the localisation of polarity markers.

Evolution of Sds22 and PP1 functions
sds22 was first identified in the single-celled yeast, Schizosaccharomyces pombe where it encodes a nuclear protein that directly binds to and regulates a PP1 phosphatase [16,17]. In S. pombe, Sds22 is essential for this phosphatase to control events during mitosis [16,17]. In metazoans, the PP1 family has expanded and Sds22 has aquired additional functions in cell shape and polarity. Accordingly, in both Drosophila and mammals, Sds22 is not exclusively localised to the nucleus and is instead found throughout the cell (Additional Fig 3; [18,19]).

Discrepancies between sds22 and PP1 mutant phenotypes in Drosophila
Drosophila have four PP1 phosphatases, named after their isotype and cytological location:PP1 13c,PP1 87c, PP1 96a and PP1 9c. The potential for redundancy among these four genes complicates genetic analysis. Nevertheless, mutation of individual PP1 genes indicates that some may have unique functions. Mutant alleles of PP1α87c, which contributes 80% of the total PP1 phosphatase activity [20], show strong defects in mitosis [21]. Of the other three Drosophila PP1s, only one is essential, PP1β9c. Interestingly, the morphological phenotype of PP1 9c-mutant cells resembles that of sds22 mutant cells in both follicle cells and nurse cells [11]. In PP1 9c mutant clones of the follicular epithelium, cells are shortened in their apicalbasal axis and cytosolic levels of G-actin and phospho-Sqh are increased [11]. However, PP1 9c mutant cells were not reported to show defects in cell polarity.
Why do sds22 mutants not exhibit the mitotic defects found PP1 87c mutants? A likely explanation is that our sds22 mutant allele is not a null. In fact, depletion of sds22 by RNAi in Drosophila cells was recently reported to cause mitotic defects [22]; [D. Glover, personal communication]. Thus, our sds22 allele may be a hypomorph that reveals the function of Sds22 in regulating cell shape and polarity.
Why do PP1 9c mutants not exhibit the polarity defects found in sds22 mutants? A plausible explanation is that the presence of the three other PP1s masks this phenotype. PP1 87c mutants, which remove 80% of PP1 activity, might therefore be expected to show a polarity defect. However, the mitotic defects in these mutants prevent growth of clones, obscuring any potential polarity defect. Thus, sds22 mutations are a convenient way of modulating PP1 activity that can reveal otherwise hidden phenotypes.
Finally, other proteins have been identified which link PP1 phosphatases to target proteins in Drosophila. These include MYPT-75D, which appears to recruit Sqh to PP1 and its loss has a weaker phenotype than loss of Sds22 [11]. Another MYPT is MBS, whose loss of function phenotype even more strongly resembles that of Sds22 [23]. While these two related MYPTs may be partially redundant with one another, it is unlikely that they are also redundant with Sds22. Redundancy between Sds22 and MYPTs would likely result in Sds22 being non-essential. Further experiments will be necessary to establish whether the Sds22/PP1 complex and MYPT/PP1 complexes act independently or together in cells.

Conclusion
We have confirmed work in other organisms by showing that Sds22 acts as an essential subunit of PP1 phosphatases in both Drosophila and mammals and controls cell shape and polarity in epithelial cells. There is an interesting parallel between the sds22 mutant phenotype in Drosophila and the escape of metastasising tumour cells from epithelia to invade local tissues. Interestingly, PPP1R7 has been reported to be significantly downregulated in human squamous cell carcinomas (HNSCCs) and other cancers, such as melanoma and prostate cancer http://www.oncomine.org/. Thus, downregulation of Sds22/PPP1R7 may contribute to tumour progression in humans.

Generation of mutant sds22 alleles
The PiggyBac transposon based mutagenesis screen has been described [24,25]. PiggyBac insertion sites were mapped by inverse PCR. The PiggyBac is contains a genetrap module designed to prevent translation of downstream genes. Excision of the PiggyBac insertion in sds22 PB1173 was achieved by introduction of the PiggyBac transposase.

Molecular Biology
sds22 cDNA was amplified by PCR from the GM06266 cDNA clone (Berkley Drosophila Genome sequencing project), cloned into a pEGFP-N1 vector (Clontech Laboratories) to create a C-terminal GFP fusion, and then subcloned into a pUAST expression vector [26] for P-element transformation.

Fly genetics
The tubulin.Gal4, c204.Gal4 (BL3751) and engrailed.Gal4 fly lines are described in Flybase and available from the Bloomington stock centre.
Clones of genetically marked homozygous sds22 mutant cells were generated using the following genotypes:

Immuno-fluorescence in Drosophila
Imaginal discs and egg chambers were dissected on ice, fixed in 4% formaldehyde for 20 min and incubated with primary antibodies overnight at 4°C in BBT (PBS, 0.2% TritonX-100, 0.1% BSA; for the α-Sqh antibody: PBS, 0.3% Tween-20, 0.1% BSA). Secondary antibodies, Phalloidin and DAPI were added for 2 hrs at room temperature. The following antibodies and chemicals were used:

Immuno-precipitation
Drosophila S2 cells were transfected with combinations of Sds22-GFP and HA-tagged PP1 proteins, expressed from the pUAST vector by co-transfection of the Copper-inducible pMT-Gal4 plasmid. 2 days after induction with 700 μM CuSO 4 , cell lysates were harvested and incubated overnight at 4 degrees with agarose beads conjugated to mouse anti-HA antibodies. The beads were spun down and washed several times. Precipitated proteins were eluted from the beads by boiling in SDS-sample buffer and then analysed by western blotting with rat anti-HA and rabbit anti-GFP antibodies.
Immuno-fluorescent staining of cultured cells A431 cells were fixed using 4% PFA in PBS for 15 minutes at room temperature, washed with PBS and permeabilised with 0.2% Triton X100 in PBS for 15 minutes. They were blocked with 5%FBS in PBS prior to incubation with pERM or pS19MLC antibodies (Cell Signaling #3141 and #3671 respectively) or TRITC-phalloidin (Sigma #P1951).