PAR-4/LKB1 prevents intestinal hyperplasia by restricting endoderm specification in Caenorhabditis elegans embryos

The master kinase PAR-4/LKB1 (STK11) is a notorious tumour suppressor, mutated in sporadic cancers as well as in Peutz–Jeghers syndrome, an inherited disorder that leads to the development of benign intestine hamartomatous polyps and a high frequency of malignant tumours (Alessi et al., 2006; Partanen et al., 2013). The ability of LKB1 to prevent polyp formation may be linked to its function in intestinal stromal cells. Indeed, tissue-specific deletions of LKB1 in mice have shown that the loss of LKB1 in the sole intestinal epithelial cells does not induce polyp formation. In contrast, LKB1 deletion in stromal cells gives rise to polyposis, possibly as a consequence of abnormal interleukin production (Katajisto et al., 2008; Ollila et al., 2018; Poffenberger et al., 2018). LKB1 has also been proposed to regulate intestinal polarity and brush border formation. In intestinal epithelial cancer cell lines, ectopic activation of PAR-4/LKB1 is sufficient to induce the polarized distribution of apical and basolateral proteins and the formation of apical microvilli-like structures (Baas et al., 2004). It should, however, be noted that deletion of the LKB1 kinase domain in mouse intestinal cells does not seem to affect epithelial polarity (Shorning et al., 2009). Thus, although LKB1 appears to be a major regulator of intestinal homeostasis, its exact in vivo functions remain to be elucidated.

To decipher the role of PAR-4/LKB1 during intestinal development, we took advantage of the Caenorhabditis elegans intestine. This very simple intestine constitutes an ideal model in which to characterize in vivo the different steps of intestinal development, including specification and differentiation, proliferation, polarization and brush border formation. The C. elegans intestine is composed of 20 cells, which form an antero-posterior tube surrounding the intestinal lumen. These 20 intestinal cells all derive from a single embryonic precursor cell, the E blastomere, which is born at the 8-cell embryonic stage and then undergoes a series of stereotyped cell divisions that give rise to the 20-cell stage intestine (Asan et al., 2016; Leung et al., 1999; Sulston et al., 1983) (Fig. 5A). The different intestinal development stages are named according to the number of E descendants present (E2, E4, E8, E16 and E20). Cell polarization occurs at the E16 stage (Achilleos et al., 2010; Leung et al., 1999; Totong et al., 2007). Shortly after, small gaps start to separate intestinal cells at the midline to eventually form the intestinal lumen (Asan et al., 2016; Bidaud-Meynard et al., 2021; Leung et al., 1999). Microvilli start to appear at the E20 stage, during the 1.5-fold embryonic developmental stage and form a regular brush border between the 2- and 3-fold stages (Asan et al., 2016; Bidaud-Meynard et al., 2021).

In 4-cell-stage embryos, asymmetric cell division of the EMS cell gives rise to the E and MS blastomeres (Fig. 5A). The restricted expression of two GATA transcription factors, END-1 and END-3, in the E blastomere triggers intestinal fate specification and expression of the differentiation GATA factors ELT-2 and ELT-7 (Fukushige et al., 1998; Maduro et al., 2005a; Sommermann et al., 2010; Wiesenfahrt et al., 2016; Zhu et al., 1997, 1998). end-1 and end-3 expression is induced by two other GATA factors, MED-1 and MED-2, which are themselves controlled by the bZIP-like factor SKN-1 (Maduro et al., 2005b, 2001). The SKN-1/MED-1/2 pathway plays a crucial role in intestinal specification (Bowerman et al., 1992; Maduro et al., 2005b, 2001) although additional parallel pathways, in particular the transcription factors POP-1/TCF and PAL-1, also trigger end-1 and end-3 expression (Maduro et al., 2005b; Shetty et al., 2005).

In C. elegans, PAR-4 plays a crucial role in the asymmetric localization of cell fate determinants in the early embryo (Kemphues et al., 1988; Tenlen et al., 2008) and is required for the formation of intestinal cells (Kemphues et al., 1988; Morton et al., 1992). In the 1-cell embryo, it also moderately regulates cortical polarity (Chartier et al., 2011; Hung and Kemphues, 1999) and controls cell cycle timing through the regulation of both CDC-25.1 and replication origins (Benkemoun et al., 2014; Rivers et al., 2008). Here, we examine the role of PAR-4 during intestinal development. We show that PAR-4 is not required for epithelial polarity establishment and has a minor role in brush border formation. By contrast, severe deformations of the intestinal lumen are observed in par-4 mutants. These deformations are associated with the presence of supernumerary intestinal cells. Importantly, we find that PAR-4 does not control intestinal cell number by regulating cell proliferation, but rather by restricting intestinal specification to the E blastomere.

In C. elegans, a faint and uniform cortical PAR-4 immunostaining has been observed in 1-cell embryos. As the embryos further divide, the cortical enrichment of PAR-4 becomes more robust, in particular at cell–cell boundaries (Watts et al., 2000). To characterize PAR-4 expression pattern further, we used CRISPR/Cas9 genome editing to endogenously label the two longest PAR-4 isoforms with the monomeric Neon Green (mNG) fluorophore. Consistent with a recent report (Roy et al., 2018), we observed that mNG::PAR-4 was ubiquitously expressed and cortically enriched in embryos, especially at cell–cell boundaries (Fig. 1A). In enterocytes, mNG::PAR-4 localized both at the apical cortex and at the basolateral cortex. Notably, during the early steps of intestine polarization, mNG::PAR-4 also localized on numerous subapical foci (Fig. 1A).

This localization prompted us to investigate whether PAR-4 is necessary for in vivo polarization and brush border formation. Intestinal polarization starts with the apical accumulation of PAR-3, PAR-6 and PKC-3 at the E16 stage, preceding embryonic elongation (Achilleos et al., 2010; Totong et al., 2007). Consistent with this, in early lima bean embryos, prior to elongation, endogenously tagged PAR-6::GFP localized at the apical membrane (Fig. 1B). Basolateral proteins such as LET-413^Scribble initially localize on all membranes before being excluded from the apical membrane at the beginning of embryonic elongation (Legouis et al., 2000; Pickett et al., 2022). As a result, in elongated 2-fold embryos, endogenously tagged PAR-6::mKate2 and LET-413^Scribble::mNG localized at the apical and basolateral membranes of enterocytes, respectively (Fig. 1C). To inactivate PAR-4, we used the par-4(it47) temperature-sensitive allele (Kemphues et al., 1988; Watts et al., 2000) and shifted the embryos to a restrictive temperature (25°C) 2 h before imaging (Table S2). Inactivating PAR-4 2 h before the lima bean stage, i.e. prior to the beginning of enterocyte polarization, did not alter PAR-6 apical localization (Fig. 1B). Similarly, inactivating PAR-4 2 h before the 2-fold stage, i.e. during the first step of apical polarization, before the apical exclusion of LET-413^Scribble, did not prevent LET-413^Scribble from eventually being restricted to the basolateral membrane (Fig. 1C). Consistent with this, we also observed that the intestine-specific degradation of PAR-4 did not affect PAR-6 localization (Fig. S1A,B). Thus, PAR-4 does not seem to be required for the establishment of apico-basal polarity.

Once polarized, enterocytes assemble a regular array of microvilli, which forms the brush border (Asan et al., 2016; Bidaud-Meynard et al., 2021; Leung et al., 1999). Observation of 3/4-fold embryos expressing endogenously tagged ezrin (ERM-1::mNG) and intermediate filament (IFB-2::wScarlet) by super-resolution (SR) confocal microscopy allowed us to distinguish regularly organized apical microvilli containing ezrin and the subapical terminal web (Fig. 1D). Inactivating PAR-4 2 h before the 3/4-fold stage, i.e. during brush border formation, did not alter ERM-1 and IFB-2 localization and did not seem to prevent microvilli formation (Fig. 1D). Similarly, the intestine-specific degradation of PAR-4 did not seem to affect microvilli formation (Fig. S1C). To characterize further the effect of PAR-4 loss-of-function on brush border formation, we used transmission electron microscopy (TEM). At the 3/4-fold stage, control embryos displayed regular microvilli at the apical membrane of enterocytes (Fig. 1E). Consistent with our SR observations, we found that par-4(it47) embryos also developed an apical brush border (Fig. 1E). Nonetheless, measurements of microvilli features revealed mild defects of the brush border ultrastructure. Namely, microvilli density was slightly decreased and microvilli were slightly longer and thinner in par-4(it47ts) embryos compared with controls (Fig. 1F). It should, however, be mentioned that our analysis also revealed rather strong inter-individual variations (Fig. S1D-F). Altogether, our observations demonstrate that PAR-4 is not required for intestinal cell polarity and has a rather minor role in microvilli formation in C. elegans embryos.

Although par-4(it47) embryos do not have strong polarity or brush border defects, we found that they displayed striking lumen deformations. In contrast to control embryos, which had a regular elliptical intestinal lumen, par-4(it47) mutant embryos exhibited severe deformations of the apical membrane (Fig. 2A,B). At the 3/4-fold stage, such deformations were visible in 40% (n=22/55) and 78% (n=7/9) of par-4(it47ts) embryos observed either by confocal or TEM, respectively. Notably, these defects were not observed in 2-fold embryos (Fig. S2A). Importantly, we noticed that ERM-1 and IFB-2 correctly localized even in regions with strong lumen deformations (Fig. S2B), thus confirming the absence of apical polarity defects.

We next wondered whether these lumen deformations were associated with other defects in tissue organization and first examined apical junctions. C. elegans enterocytes have only one type of apical junction, the C. elegans apical junction (CeAJ), which is formed by two protein complexes, the cadherin/catenin complex (CCC) and the DLG-1–AJM-1 complex (DAC) (McMahon et al., 2001; Segbert et al., 2004). The CCC is composed of HMR-1 (E-cadherin; cadherin 1), HMP-1 (α-catenin) and HMP-2 (β-catenin) (Costa et al., 1998). The DAC localizes basally to the CCC and contains the Discs large homologue DLG-1 (McMahon et al., 2001) and the non-conserved protein AJM-1 (Köppen et al., 2001). To test whether PAR-4 loss of function had an effect on intestinal CeAJs, we used strains expressing the endogenously tagged CeAJ components DLG-1::mNG, HMR-1::GFP (or HMR-1::mKate2) and HMP-1::GFP (Heppert et al., 2018; Marston et al., 2016). DLG-1, HMR-1 and HMP-1 localized at CeAJs, both in control and par-4(it47) embryos (Fig. 2C, Fig. S2C). However, additional CeAJs were visible in about one-third of par-4(it47) embryos (Fig. 2C, Fig. S2C-D). In contrast to lumen deformations, these supernumerary CeAJs were present both in 2-fold and 3/4-fold embryos (Fig. S2D). The C. elegans gut arranges into rings composed of two enterocytes, except the anterior ring, which is made of four cells. Thus, on TEM transversal sections, control embryos display two, occasionally four, electron-dense structures corresponding to CeAJs, which maintain the cohesion between the two, or four, cells surrounding the lumen (Fig. 2B). By contrast, more than two CeAJs were observed on several sections of all par-4(it47) embryos and sections with more than four CeAJs were found in 56% of par-4(it47) embryo (n=5/9) (Fig. 2B). Notably, we also found that junction length was increased in all par-4(it47) embryos analysed by TEM (n=9) (Fig. S2E).

The presence of additional CeAJs prompted us to examine the number of intestinal cells in par-4(it47) mutants. To this end, we used strains expressing an intestinal-specific nuclear marker under the control of the elt-2 promoter, elt-2p::NLS::GFP::lacZ (Fukushige et al., 1998). Whereas most control 2-fold embryos had 20 intestinal cells, 28% of par-4(it47) embryos (n=17/60) had more than 22 enterocytes (Fig. 2D), indicating that PAR-4 prevents the appearance of supernumerary intestinal cells. These observations were initially made in embryos that had been shifted to a restrictive temperature (25°C) 2 h before the 2-fold stage. We next examined whether the timing of temperature shift could influence the number of intestinal cells. First, we counted intestinal cells in par-4(it47) embryos, which had been kept at permissive temperature (15°C) before observation and, surprisingly, found that 25% (n=17/69) had an excess of enterocytes (Fig. S3A). This result could be explained by previous observations showing that the par-4(it47) allele is already partially inactivated at 15°C (Morton et al., 1992). However, the similarity between the phenotypes observed at 15°C or after 2 h at 25°C also suggested that the presence of supernumerary cells was not due to the inactivation of PAR-4 in the 2 h preceding the 2-fold stage, but rather to an earlier effect of PAR-4. To test this hypothesis, we next tried to observe intestinal cells in par-4(it47) embryos after a longer shift at 25°C. Unfortunately, under those conditions, the par-4(it47) mutation induced pleiotropic embryonic defects, in particular an elongation arrest, precluding proper embryo staging and characterization. Interestingly, we also found that no extra intestinal cells were produced when PAR-4 was specifically degraded in the intestine (Fig. S3B), suggesting that the presence of supernumerary enterocytes in par-4(it47) mutants may not be due to the inactivation of PAR-4 in the E lineage itself, but rather reveals a previously unappreciated function of PAR-4 outside the E lineage. Notably, supernumerary enterocytes were also observed in 41% of par-4(it47) embryos expressing endogenously tagged ELT-2::mNG (n=23/56) (Fig. S3C). Finally, we found that intestinal nuclei are smaller in par-4(it47) embryos (Fig. S3D). Nevertheless, the presence of gut granules in all elt-2-expressing cells suggested that they were all properly differentiated enterocytes (Fig. S3E).

Our observations revealed the existence of different gut phenotypes in par-4(it47) embryos: mild brush border defects, increased junction length, intestinal lumen deformations, additional apical junctions and excess of enterocytes. We wondered whether these phenotypes were due to a unique function or to several roles of PAR-4. We first investigated whether the presence of supernumerary enterocytes could affect tissue organization and lead to the appearance of additional CeAJs and lumen deformations. To this end, we observed embryos co-expressing the intestinal nuclear marker elt-2p::NLS::GFP::lacZ and the CeAJ marker HMR-1::mKate2 and found that 23% (n=7/31) of par-4(it47) 2-fold embryos had extra intestinal nuclei. Importantly, all of them also displayed several additional CeAJs (Fig. 3A). Reciprocally, all embryos with additional CeAJs (n=7/31) also had supernumerary intestinal cells. Moreover, observation of embryos co-expressing HMR-1::mKate2 and ERM-1::mNG showed that 73% (n=32/44) of par-4(it47) 3/4-fold embryos displayed intestinal lumen deformations, and all of them (n=32/32) also had additional CeAJs (Fig. 3B). Reciprocally, 82% (n=36/44) of par-4(it47) embryos had additional CeAJs and 89% (n=32/36) of them displayed lumen deformations. Similarly, the presence of additional CeAJs was associated with the presence of lumen deformations on TEM images (Fig. 2B). We next used a plasma membrane marker specifically expressed in the intestine, vha-6p::GFP::PH, to assess both lumen and enterocyte organization. In control 3/4-fold embryos, vha-6p::GFP::PH allowed us to distinguish the regular and elliptical shape of the lumen and the presence of two enterocytes in the ring surrounding the lumen (Fig. 3C). By contrast, 51% (n=20/39) of par-4(it47) embryos displayed intestinal lumen deformations (Fig. 3C). Notably, all of them (n=20/20) also had additional enterocytes surrounding the lumen (Fig. 3C). Finally, our TEM analysis revealed the presence of more than four cells per intestinal ring on sections from 56% (n=5/9) of par-4(it47) embryos. In all those cases, the lumen appeared deformed and additional CeAJs were observed (Fig. 3D). Altogether, our observations reveal a strong correlation between the presence of extra intestinal cells, additional CeAJs and lumen deformation, suggesting that the presence of extra intestinal cells could explain the presence of additional CeAJs and lumen deformations. By contrast, our TEM analysis did not reveal any obvious link between microvilli or CeAJ length defects and the number of additional CeAJs (Figs S1D-F, S2E), suggesting that those latter defects may not be due to the presence of supernumerary enterocytes.

In order to confirm that the presence of extra intestinal cells is responsible for the presence of additional junctions and lumen deformations, we compared the different phenotypes of par-4(it47) embryos with those of cdc-25.1 gain-of-function mutant embryos. Previous studies had indeed revealed the presence of supernumerary enterocytes in cdc-25.1(rr31) and cdc-25.1(ij48) gain-of-function mutants (Clucas et al., 2002; Kostić and Roy, 2002). Immunostaining of the junction component AJM-1 also suggested the presence of additional CeAJs in the intestine of cdc-25.1(ij48) embryos (Clucas et al., 2002). We first observed cdc-25.1(rr31) gain-of-function mutants co-expressing the intestinal nuclear marker elt-2p::NLS::GFP::lacZ and the CeAJ marker HMR-1::mKate2 and confirmed that all cdc-25.1(rr31) embryos had supernumerary enterocytes and additional CeAJs (Fig. S4A). We next characterized the lumen morphology of cdc-25.1(rr31) embryos expressing vha-6p::GFP::PH and observed that all of them displayed intestinal lumen deformations (Fig. 4A). Lumen architecture defects were also present in late larval stages (Fig. S4B). Sections with more than four CeAJs and severe lumen deformations were observed in all cdc-25.1(rr31) embryos analysed by TEM (n=7) (Fig. 4B). Consistent with previous observations (Choi et al., 2017), we found that additional enterocytes could be organized into rings composed of four or more cells (Fig. 4). Our TEM analysis also revealed that cdc-25.1(rr31) mutants displayed a slightly increased microvilli density and thicker microvilli but microvilli length and CeAJs length were not affected (Fig. S4C,D). Thus, although par-4(it47) and cdc-25.1(rr31) embryos share the same phenotypes in terms of lumen and junction organization, they do not show the same microvilli and junction length defects. Altogether, these observations are consistent with our hypothesis that additional junctions and lumen deformations are linked to the presence of extra intestinal cells, but suggest that the defects of microvilli and CeAJs length observed in par-4(it47) mutant embryos are due to separate PAR-4 functions.

We next sought to decipher the mechanisms by which PAR-4 regulates intestinal cell number and first investigated whether the supernumerary enterocytes observed in par-4(it47) embryos were due to additional divisions in the E lineage. In control embryos, the 20 enterocytes all emanate from the E blastomere, which undergoes five series of stereotyped divisions (Fig. 5A) (Leung et al., 1999; Sulston et al., 1983). Acceleration of the cell cycle, as for instance reported in cdc-25.1 gain-of-function mutant embryos, can lead to the appearance of an extra round of divisions and supernumerary intestinal cells (Clucas et al., 2002; Kostić and Roy, 2002). We thus performed lineage experiments to determine whether PAR-4 also regulates the duration and number of intestinal cell divisions. To follow divisions, we used strains co-expressing two intestine-specific nuclear markers under the control of the end-1 or elt-2 promoters, end-1p::GFP::H2B and elt-2p::NLS::GFP::lacZ (Fukushige et al., 1998; Shetty et al., 2005). Signal from the end-1 reporter became visible in E2 cells whereas signal from the elt-2 reporter appeared during the E8 stage and persisted afterwards (Fig. 5B). In control embryos recorded at 25°C, it took around 3 h between the appearance of the end-1 reporter signal in E2 cells and the E16 stage (Fig. 5B). As previously reported (Kostić and Roy, 2002), we found that intestinal cells in cdc-25.1(rr31) gain-of-function embryos divided faster and underwent an additional division so that, after 3 h at 25°C, the intestine was composed of 32 cells (Fig. 5B,C). By contrast, no extra-division and no acceleration of the cell cycle occurred in the E lineage of par-4(it47) embryos observed during 3 h at 25°C (Fig. 5B,C). Thus, unlike what happens in cdc-25.1 gain-of-function mutants, the supernumerary intestinal cells observed in par-4(it47) mutants do not result from an overall shortening of the E lineage cell cycles and from the appearance of an extra division between the E2 and E16 stage.

Once embryos have reached the E16 stage, only four enterocytes of the E16 intestinal primordium, referred to as ‘star-cells’, divide to obtain the final E20 stage (Sallee et al., 2021). These star-cells usually divide between the late lima bean and the 1.8-fold stages (Sallee et al., 2021 and our observations). During our long-lasting lineage experiments, we however failed to record this late division, including in control embryos. This prevented us from determining whether the extra intestinal cells observed in par-4(it47) mutants could result from the late and abnormal division of non-star-cells. To circumvent this technical problem, we first imaged control and par-4(it47) embryos, which were at the lima bean stage and had 16 intestinal nuclei. We next observed the same embryos 1-2 h later: at this time point, most control embryos had 20 intestinal cells (n=15/18), and occasionally 21 (n=3/18) (Fig. 5D). Similarly, most par-4(it47) embryos exhibited 20 intestinal nuclei (n=32/37), occasionally 19 (n=1/37), 21 (3/37) or 22 (1/37) (Fig. 5D). They thus had normally completed their last intestinal cell division and did not undergo another round of division. Altogether, our results show that the presence of supernumerary intestinal cells in par-4(it47) mutants is not due to excessive cell proliferation in the E lineage.

Although our lineage experiments showed that the descendants of the E blastomere divide normally in par-4(it47) embryos, we occasionally observed expression of the end-1p::GFP::H2B reporter in two cells outside the E lineage (n=2/19). This ectopic expression of the end-1 reporter suggested that some cells that are not emanating from the E lineage adopt an intestinal-like fate and produce extra enterocytes. In these first lineage experiments, embryos were shifted to 25°C just at the start of imaging, i.e. at the E or E2 stage. We next tested whether an earlier shift would lead to the more frequent ectopic expression of the end-1 reporter and shifted the embryos to 25°C just prior to or during the first embryonic divisions (see Table S2). Under these conditions, we found that 53% (n=19/36) of par-4(it47) embryos ectopically expressed the end-1 reporter in one or two cells localized outside the E lineage (Fig. 6A). This ectopic expression appeared at the E2 or E4 stage. Notably, the presence of these cells ectopically expressing end-1 did not decrease cell cycle length in the E lineage (Fig. 6B). Importantly, endogenously tagged END-1::mNG was also expressed outside the E lineage in 32% (n=13/40) of par-4(it47) embryos (Fig. 6C).

According to their position within the embryo, the cells ectopically expressing the end-1p::GFP::H2B reporter or END-1::mNG in par-4(it47) embryos seemed to be P₂ or MS descendants (Fig. 6A,C; see legend for details). To identify the origin of ectopic cells more precisely, we followed embryonic cell divisions in embryos co-expressing the end-1 reporter and a ubiquitously expressed membrane marker (Fig. 6D). In this experiment, 47% (n=14/30) of par-4(it47) embryos showed ectopic expression of the end-1 reporter. This mostly occurred within the C lineage (n=9/14), especially in Cp or its descendants Cpa and Cpp (n=7/9) and occasionally in both Ca and Cp (n=2/9). In one embryo, we could not distinguish whether the end-1 reporter was expressed in the C or D lineage. In a few embryos (n= 4/14), the end-1 reporter was expressed only in MS descendants. Altogether, our data suggest that PAR-4 restricts intestinal specification to the E lineage by preventing end-1 expression in the C and MS blastomeres.

We next wondered whether the ectopic expression of end-1 in cells localized outside the E lineage is sufficient to induce elt-2 expression. In control embryos (Fig. 6E) and par-4(it47) embryos without end-1 ectopic expression (Fig. 6I), elt-2 expression was restricted to the E lineage. In contrast, 58% (n=11/19) of par-4(it47) embryos ectopically expressing end-1 in P₂ descendants also ectopically expressed the elt-2 reporter. The number of cells ectopically expressing elt-2 was variable, ranging from no cell to all cells ectopically expressing end-1 (Fig. 6F-I). Unlike P₂ descendants, MS descendants ectopically expressing end-1 did not express elt-2 (n=3/3) (Fig. 6G). Our data thus suggests that end-1 ectopic expression is sufficient to trigger some expression of elt-2 in P₂ descendants.

Strikingly, we noticed that par-4(it47) embryos also displayed a substantial variability in the number of E descendants expressing elt-2, with some embryos even lacking any elt-2 expression in the E lineage (Fig. 6G,I). Consistent with this, we had also observed that par-4(it47) 2-fold embryos expressing either elt-2p::NLS::GFP::lacZ or ELT-2::mNG occasionally lacked any elt-2 signal (Fig. S3 legend). Importantly, those embryos also lacked gut granules, indicating the absence of properly differentiated enterocytes (Fig. S5A). Those observations show that PAR-4 allows the robust expression of elt-2 in the E lineage and corroborate previous studies showing that it is required for proper intestinal differentiation (Kemphues et al., 1988; Morton et al., 1992).

We next investigated whether a reduction of end-1 expression could account for the variability of elt-2 expression inside the E lineage of par-4(it47) embryos. We found that the expression of endogenously tagged END-1::mNG was reduced in the E lineage of par-4(it47) embryos (Fig. S5B), with 32% (n=13/40) of them even showing no END-1::mNG expression inside the E lineage (Fig. S5B-D). Moreover, in contrast to control embryos, which all started to express END-1::mNG at the E2 stage (Fig. 6C), 30% (n=12/40) of par-4(it47) embryos started to express END-1::mNG only at the E4 stage (Fig. S5C,E). Finally, 10% (=4/40) of par-4(it47) embryos expressed END-1::mNG only in a subset of E descendants (Fig. S5F). Thus, PAR-4 is also required for the robust expression of end-1 in the E lineage.

Altogether, our results show that PAR-4 regulates the number of differentiating enterocytes in two ways: (1) by preventing C blastomeres from adopting an intestinal fate and (2) by ensuring the robust expression of end-1 and elt-2 in the E lineage.

We next set out to identify the mechanisms by which PAR-4 prevents the ectopic expression of end-1 and first tested the possible involvement of the kinase PAR-1, a major downstream effector of PAR-4 in the early C. elegans embryo (Benkemoun et al., 2014; Narbonne et al., 2010; Tenlen et al., 2008). We found that 35% (n=7/20) of par-1(zu310) temperature-sensitive mutant embryos expressed the end-1p::GFP::H2B reporter outside the E lineage, in P₂ descendants (n=7/7) (Fig. 7A) and occasionally also in MS descendants (n=1/7). Furthermore, 86% (n=6/7) of par-1(zu310) embryos ectopically expressing end-1 in P₂ descendants also ectopically expressed the elt-2p::NLS::GFP::lacZ reporter, in either some or all end-1 expressing cells (Fig. 7A,B). Those observations strongly suggest that PAR-4 prevents the ectopic expression of end-1 and elt-2 through PAR-1 regulation.

The GATA transcription factors MED-1 and MED-2 as well as their regulator SKN-1 play a major role in intestinal specification and end-1 expression (Bowerman et al., 1992; Maduro et al., 2005b; 2001). We thus asked whether the ectopic expression of end-1 in par-1(zu310) embryos was associated with the ectopic expression of med-1. In control embryos, med-1p::GFP::MED-1 was expressed in E and MS cells, as well as in their descendants (Fig. 7C). By contrast, 83% (n=10/12) of par-1(zu310) embryos, which ectopically expressed the end-1p::H1::mCherry reporter, also ectopically expressed med-1 (Fig. 7C; see legend for details). Thus, the ectopic expression of end-1 observed in par-1(zu310) embryos may result from the ectopic expression of med-1. We next examined skn-1 expression using a strain expressing endogenously tagged SKN-1::GFP. In control embryos, SKN-1::GFP was enriched in all nuclei in 4-cell stage embryos and was rapidly lost during the 8-cell stage. No abnormal accumulation of SKN-1 was observed in par-1(zu310) embryos (Fig. S6A; see legend for details), showing that the ectopic expression of med-1 and end-1 does not result from an increased or persistent accumulation of SKN-1 outside the E lineage. We nevertheless found that the ectopic expression of end-1 was dependent on SKN-1 activity. Indeed, whereas end-1p::GFP::H2B was ectopically expressed in 50% (n=11/22) of par-1(zu310) embryos treated with control RNAi, its expression was completely abolished following depletion of SKN-1 by RNAi, both inside and outside the E lineage (Fig. 7D; see legend for details). Collectively, our data strongly suggest that in par-1(zu310) embryos SKN-1 is active outside the E lineage, thereby triggering med-1 and end-1 ectopic expression.

Importantly, similarly to par-4(it47) mutants, par-1(zu310) embryos displayed variable elt-2 expression in the E lineage (Fig. 7B). Interestingly, we found that par-1(zu310) embryos also had reduced expression levels of SKN-1 in EMS, the E precursor cell (Fig. S6B). Likewise, expression of MED-1 was reduced in the E lineage of par-1(zu310) embryos (Fig. S6C). Hence, PAR-1 appears to ensure the robust expression of SKN-1 and MED-1 inside the E lineage, which in turn are likely to account for the robust expression of end-1 and elt-2.

Altogether, our observations demonstrate that PAR-4 prevents intestinal hyperplasia without controlling cell proliferation in the E lineage, but rather by preventing the C blastomere from adopting a terminal intestinal fate. Restriction of end-1 and elt-2 expression to the E lineage by PAR-4 is likely to involve its effector PAR-1 as well as restriction of the activity of the SKN-1/MED-1 pathway. By contrast, PAR-4 and PAR-1 most likely also ensure robust intestinal differentiation by promoting SKN-1 and MED-1 expression inside the E lineage (Fig. 7E).

In this study, we have shown that PAR-4 is not required for the establishment of apico-basal polarity during C. elegans intestinal development and has a minor role during brush border formation. By contrast, PAR-4 is essential to prevent intestinal hyperplasia. PAR-4 inactivation indeed results in the presence of supernumerary intestinal cells and severe lumen deformations. This function of PAR-4 appears to be independent of cell proliferation regulation and involves the restriction of intestinal specification to the E lineage.

Our endogenously expressed mNG::PAR-4 construct confirms that PAR-4 is ubiquitously expressed in C. elegans embryos (Roy et al., 2018; Watts et al., 2000). In the intestinal epithelium, it localizes to both the apical and lateral cortices and is also present on subapical foci during polarization. Despite this intriguing localization, PAR-4 appears not to be required for the establishment of intestinal apico-basal polarity in C. elegans. Consistent with this, no obvious polarity defects were observed in mice intestinal cells lacking LKB1 activity (Shorning et al., 2009). PAR-4/LKB1 has been shown to modulate the asymmetric localization of PAR-3, PAR-6 and aPKC in several other in vivo systems, including in epithelial cells (e.g. Amin et al., 2009; Bonaccorsi et al., 2007; Chartier et al., 2011; Lee et al., 2007; Martin and St Johnston, 2003). Many cell types can nevertheless apparently polarize normally in its absence (e.g. Amin et al., 2009; Krawchuk et al., 2015), underlining that PAR-4/LKB1 is far from systematically having a crucial role in polarity establishment. Its requirement appears to be extremely sensitive to the cell context and may depend on the robustness of other polarization mechanisms at play in each cell type.

Our results show that PAR-4 is not strictly required for microvilli formation in C. elegans embryos but may slightly contribute to their robust shape. In intestinal cell lines, ectopic activation of LKB1 triggers the formation of microvilli-like structures through the activation of a signalling cascade involving the small GTPase Rap2 and ezrin phosphorylation (Baas et al., 2004; Gloerich et al., 2012; ten Klooster et al., 2009). In C. elegans, the phosphorylation of ERM-1 prevents its basolateral accumulation in embryonic intestinal cells (Ramalho et al., 2020). The normal apical localization of ERM-1 that we observed in par-4 embryos thus strongly suggests that ERM-1 phosphorylation is not affected in this context. Moreover, our TEM observations did not reveal microvilli defects in rap-2(gk11) mutant larvae (O.N., unpublished observations). It is thus unlikely that PAR-4 regulates microvilli growth through the RAP-2/ERM-1 pathway in C. elegans.

Besides its effect on microvilli morphology, we find that PAR-4 regulates CeAJ length. Interestingly, Drosophila Lkb1(−) photoreceptor cells also exhibit longer adherens junctions (Amin et al., 2009). Those are also frequently fragmented, a phenotype that we did not observe in C. elegans enterocytes. In C. elegans, LET-413^Scribble has been proposed to trigger CeAJs compaction and let-413 mutants display extended and fragmented intestinal CeAJs (Legouis et al., 2000; McMahon et al., 2001). However, loss of LET-413 also results in the strong mislocalization of the apical proteins PAR-3 and PAR-6 (McMahon et al., 2001). The absence of polarity defects in par-4(it47) embryos thus argues against a role of PAR-4 in regulating LET-413 activity.

Intestinal hyperplasia is the most striking phenotype that we observe in par-4 embryos. In 2-cell C. elegans embryos, par-4 mutants show an acceleration of the cell cycle in the P₁ blastomere (Morton et al., 1992). This effect can be explained by the ability of PAR-4 and its effector PAR-1 to inhibit CDC-25.1 nuclear accumulation as well as DNA replication in P₁ (Benkemoun et al., 2014; Rivers et al., 2008). Moreover, stabilization of CDC-25.1 in the intestine leads to cell cycle acceleration and extra enterocyte divisions (Clucas et al., 2002; Kostić and Roy, 2002). However, in the intestine, par-4 mutants exhibit neither cell cycle acceleration nor extra divisions, suggesting that PAR-4 does not regulate CDC-25.1 in that context and that intestinal hyperplasia is not linked to excessive cell proliferation.

Rather, we find that supernumerary cells arise from specification defects and are due to abnormal expression of the intestinal specification factor END-1 in the C lineage. Most C cells ectopically expressing end-1 also in turn activate the expression of the differentiation factor ELT-2. The kinase PAR-1, a major effector of PAR-4, appears to be involved in preventing the ectopic expression of end-1 as well as of its upstream regulator med-1. Furthermore, the ectopic expression of end-1 in the C lineage depends on SKN-1 activity (Fig. 7E). In wild-type 4-cell-stage embryos, SKN-1 is present both in EMS and P₂ blastomeres but its transcriptional activity is repressed by PIE-1 in P₂, preventing SKN-1 from inducing intestinal differentiation in P₂ descendants (Mello et al., 1992; Tenenhaus et al., 2001). Notably, PAR-1 has been shown to regulate PIE-1 asymmetry in the 1-cell embryo (Cuenca et al., 2003), raising the possibility that a reduction of PIE-1 activity in the P lineage of par-4 or par-1 embryos is responsible for the ectopic activity of SKN-1.

Notably, our findings show that PAR-4 and PAR-1 not only prevent the abnormal expression of end-1 and elt-2 in the C lineage but also ensure their robust expression inside the E lineage, most likely by allowing the proper expression of SKN-1 and MED-1 (Fig. 7E). Altogether, those results are likely to explain why we and others observed that PAR-4 and PAR-1 are required for proper intestinal cell differentiation (Kemphues et al., 1988; Liro et al., 2018; Morton et al., 1992).

At the time when C descendants start to express end-1 and then elt-2, they are located in the vicinity of E descendants, but they have the astonishing ability to rearrange and become incorporated with E descendants to form a continuous intestine. As previously described in cdc-25.1(gof) mutants (Choi et al., 2017), supernumerary cells eventually accommodate to form rings with more than two cells surrounding a deformed lumen. The ability of cdc-25.1(gof) mutants to survive until adulthood despite the persistence of abnormal lumen architecture suggests that those defects do not severely impair feeding. Intriguingly, lumen deformations become apparent only at the 3/4-fold stages. One possible explanation is that this is due to the progressive and late incorporation of supernumerary cells to form intestinal rings containing more than two cells. Alternatively, supernumerary cells may already be incorporated into intestinal rings at the 2-fold stage and induce weak lumen deformations. Those small and early lumen deformations may then be amplified and become detectable as the lumen grows. The presence of additional CeAJs in both 2- and 3/4-fold embryos support this latter hypothesis.

In conclusion, our work reveals a dual function of PAR-4 and PAR-1 in regulating intestinal specification. Consistent with previous studies showing that PAR-4 and PAR-1 are required for the formation of differentiated intestinal cells, our results demonstrate that PAR-4 is required for the robust expression of intestinal specification and differentiation factors in the E lineage. Additionally, they also show that PAR-4 and PAR-1 prevent C blastomeres from adopting an intestinal fate. These findings may shed new light on the role of LKB1 in intestinal homeostasis, in particular on its non-epithelial function in preventing polyposis.

Strains were grown on agar plates containing NGM growth media and seeded with Escherichia coli (OP50 strain). Worms were maintained at 20°C, except strains carrying the par-4(it47) and par-1(zu310) temperature-sensitive mutations, which were kept at 15°C and shifted to the restrictive temperature (25°C) during the course of our experiments. Note that par-1(zu310) was shown to be a slow-inactivating temperature-sensitive allele (Liro et al., 2018), explaining why experiments performed with this allele required longer temperature shifts. The strains used in this study are listed in Table S1 and the temperature conditions used for each experiment are detailed in Table S2.

CRISPR/CAS9-genome edited mNG::PAR-4, PAR-4::mKate2::AID, LET-413c::mNG and ELT-2mNG were generated at the SEGiCel facility (Université Lyon 1, UMS3421, Lyon, France). mNG::PAR-4 was obtained by tagging the two longest isoforms of PAR-4 (PAR-4a and PAR-4c) with mNeonGreen at their common N terminus. PAR-4::mKate2::AID was obtained by inserting a GASGASGAS linker, mKate2 and an auxin-inducible degron (AID; MPKDPAKPPAKAQVVGWPPVRSYRKNVMVSCQKSSGGPEAAAFVK) (Zhang et al., 2015) at the N terminus of the short PAR-4 isoform (PAR-4b). This insertion also tags the two PAR-4 long isoforms (insertion after Met142 in PAR-4a and PAR-4c). LET-413c::mNG was obtained by inserting a GASGASGAS linker and mNeonGreen to the C-terminus of the LET-413c isoform. ELT-2::mNG was obtained by inserting a GASGASGAS linker and mNeonGreen to the C-terminus of ELT-2.

Intestine-specific degradation of PAR-4 was induced by incubating at 20°C embryos expressing PAR-4::mKate2::AID and the auxin receptor TIR-1 under the control of the elt-2 promoter in 5 mM acetoxymethyl indole-3-acetic acid (IAA-AM, a cell permeable form of auxin that is able to trigger degradation in embryos; Negishi et al., 2019). Control experiments showed that PAR-4 was absent from intestinal cells after 1 h 15 min auxin treatment. PAR-6 localization, microvilli formation and the number of intestinal cells in embryos depleted of PAR-4 were assessed after 4 h, 6 h 30 min, and 6 h auxin treatment, respectively.

RNAi was performed by feeding worms with RNAi clones from the Ahringer-Source Bioscience library (Kamath et al., 2003) for 48 h on 1 mM IPTG plates. L4440 was used as a control.

For standard and SR confocal microscopy, embryos were mounted on 2% agarose pads in a drop of M9 medium enriched in OP50 bacteria to immobilize embryos. Larvae were mounted on 10% agarose pads in a solution of 100 nm polystyrene microbeads (Polysciences Inc.) to stop worm movement.

Standard confocal images were acquired with a SP8 confocal microscope (Leica) equipped with an HC Plan-Apo 63×, 1.4 NA objective and the LAS AF software. z-stacks were acquired with 400 or 500 nm steps. Images of autofluorescent gut granules were obtained by illuminating embryos with a 405 nm diode and acquired with 800 nm z-steps.

SR images were acquired with a LSM 880 microscope (Zeiss), equipped with a Plan-Apo 63×, 1.4 NA objective and ZEN Black software. The Airy Scan module was used to obtain SR images. z-stacks were acquired with 180 nm steps (optimal step size).

For lineage experiments, adult hermaphrodites were dissected in M9 medium. Embryos were then transferred to a 4 µl drop of M9 medium on a 35 mm glass-bottom dish (poly-D-lysine coated, 14 mm microwell, No 1.5 coverglass, MatTek). Excess of M9 was then removed to ensure adhesion of embryos to poly-D-lysine and 2 ml of mineral oil [Light Oil (neat), Sigma-Aldrich] was used to cover the drop of M9. Recordings were performed on a Leica DMi8 spinning disc microscope equipped with a HCX Plan-Apo 63×, 1.4 NA objective and a photometrics Evolve EMCCD camera. The setup was controlled by the Inscoper Imaging Suite. Embryos were maintained at 25°C during recordings. Images were acquired at 3 or 5 min intervals (Figs 5B,C, 6C,D, 7C,D, Figs S5B-F, S6) or every 5 min during the first 2 h of recordings and then every 10 min (Figs 6A,B,E-I, 7A,B). z-stacks were acquired with 400 nm steps.

Temperature conditions used for each experiment are detailed in Table S2.

Control and mutant C. elegans embryos were shifted to 25°C 2 h before fixation. They were fixed by high-pressure freezing with the EMPACT-2 system (Leica Microsystems). Cryosubstitution (freeze substitution, FS) was carried out in anhydrous acetone containing 1% OsO4, 0.5% glutaraldehyde and 0.25% uranyl acetate for 60 h in an FS system (AFS-2; Leica Microsystems). The embryos were embedded in an Epon-Araldite mix (EMS hard formula). Adhesive frames were used (11560294 GENE-FRAME 65 L, Thermo Fisher Scientific) for flat embedding, as previously described (Bidaud-Meynard et al., 2019). Ultrathin sections (70 nm thickness) were cut on an ultramicrotome (UC7; Leica Microsystems) and collected on Formvar-coated slot grids (FCF2010-CU; EMS). Each sample was sectioned at seven to ten different places with ≥5 μm between each grid, to ensure that different gut cells were observed. TEM grids were then observed using a JEM-1400 transmission electron microscope (JEOL) operated at 120 kV, equipped with a Gatan Orius SC1000 camera and piloted by the Digital Micrograph 3.5 program (Gatan).

Images were assembled for illustration using Fiji and Inkscape 1.1 software. All embryos are oriented with the anterior pole to the left. Images obtained with the Airy Scan module were processed to enhance resolution using ZEN Black software. Maximum intensity z-projections, orthogonal views and 3D reconstructions were obtained with Fiji. Note that in Figs 2A, 3B,C and 4A, SR images are shown for illustration, but quantifications were obtained with a standard confocal microscope. Measurements of END-1, SKN-1 and MED-1 nuclear levels were performed with Fiji. END-1 levels were measured in the four E descendants at the E4 stage, SKN-1 levels in the EMS blastomere at the 4-cell embryonic stage and MED-1 levels in the E blastomere at the 8-cell embryonic stage as well as in its descendants Ea and Ep at the E2 stage. In all cases, the intensity of the cytoplasmic background signal was subtracted from measured nuclear intensities.

TEM micrographs were analysed using Fiji. Microvilli length was measured from the tip to the point where their base intersected with the apical pole. Microvilli width was measured at mid-height. Microvilli density was defined as the number of microvilli over 1 µm of lumen perimeter. Junction length corresponds to the length of the electron-dense structure.

Statistical analysis and graphical representations were performed using GraphPad Prism 8.0.1 software. Violin plots show all individual values (dots) and the median (bold black line). Details of the statistical tests used are indicated in each figure legend. Two-tailed, unpaired Student's t-tests were performed when data passed the Shapiro and Kolmogorov–Smirnov normality tests and when sample variance was equal; two-tailed Welch's t-tests were performed when data passed the Shapiro and Kolmogorov–Smirnov normality tests and when sample variance was unequal; two-tailed, non-parametric Mann–Whitney tests were performed when data did not pass the Shapiro or Kolmogorov–Smirnov normality test.

We thank M. Boxem, B. Grant, J. Rothman, M. Soto and the Caenorhabditis Genetics Center (funded by the National Institute of Health Office of Research Infrastructure Programs, P40 OD010440) for providing worm strains. The SEGiCel Facility (SFR Santé Lyon Est, CNRS UAR 3453, Lyon, France) generated CRISPR strains and M. Kanemaki provided us with modified auxin. We thank Anaëlle Raoul who helped with the characterization of intestinal cell number in par-4 mutants as well as the Michaux and Pecreaux labs for helpful discussions. Imaging was performed at the Microscopy Rennes imaging Center (MRiC Photonics and TEM, Biosit, Rennes, France), a member of the national infrastructure France-BioImaging supported by the French National Research Agency (ANR-10-INBS-04).

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202205.reviewer-comments.pdf