The Caenorhabditis elegans homolog of the Evi1 proto-oncogene, egl-43, coordinates G1 cell cycle arrest with pro-invasive gene expression during anchor cell invasion

Cell invasion allows cells to migrate across compartment boundaries formed by basement membranes. Aberrant cell invasion is a first step during the formation of metastases by malignant cancer cells. Anchor cell (AC) invasion in C. elegans is an excellent in vivo model to study the regulation of cell invasion during development. Here, we have examined the function of egl-43, the homolog of the human Evi1 proto-oncogene (also called MECOM), in the invading AC. egl-43 plays a dual role in this process, firstly by imposing a G1 cell cycle arrest to prevent AC proliferation, and secondly, by activating pro-invasive gene expression. We have identified the AP-1 transcription factor fos-1 and the Notch homolog lin-12 as critical egl-43 targets. A positive feedback loop between fos-1 and egl-43 induces pro-invasive gene expression in the AC, while repression of lin-12 Notch expression by egl-43 ensures the G1 cell cycle arrest necessary for invasion. Reducing lin-12 levels in egl-43 depleted animals restored the G1 arrest, while hyperactivation of lin-12 signaling in the differentiated AC was sufficient to induce proliferation. Taken together, our data have identified egl-43 Evi1 as an important factor coordinating cell invasion with cell cycle arrest.

intron of the egl-43L locus, the insertion of the gfp cassette at the 5' end of the first exon labels exclusively the protein encoded by the long isoform (gfp::egl-43L), whereas the insertion of the gfp tag at the 3' end of the locus labels both, the short egl-43S and the long egl-43L isoforms (egl-43::gfp, Fig 1A). Furthermore, the gfp cassette inserted at the 5' end contained two flippase recognition target (FRT) sites in the gfp introns, permitting the tissue-specific inactivation of the egl-43L isoform (S1 Fig). For the expression analysis, we focused on the mid-L3 stage (the Pn.pxx stage, after the VPCs had undergone two rounds of cell divisions), the time when AC invasion normally occurs [9]. Both reporters, gfp::egl-43L and egl-43::gfp, were expressed in the invading AC and in the surrounding ventral and dorsal uterine (VU and DU) cells (Fig 1B, for a quantification of the AC expression levels of the different reporters, see S2 Fig). We found no obvious difference in the expression pattern of the two egl-43 reporters, suggesting that the long egl-43L isoform accounts for most of the expression observed. In order to test if the uterine expression is indeed caused by egl-43L, we designed two RNAi clones, one specifically targeting egl-43L (exons 4 and 5 of egl-43L), and the other targeting the 5' UTR of egl-43S, which is not included in the egl-43L transcript (Fig 1A). EGL-43::GFP expression was lost upon egl-43L RNAi, yet no difference in expression was observed after egl-43S RNAi (Fig 1C, S2C Fig), supporting the above conclusion that the uterine expression originates predominantly from the long egl-43L isoform.
As reported previously [11], egl-43L RNAi led to an invasion defect with a penetrance comparable to that of total egl-43 RNAi (92% for egl-43L (n = 53) and 94% for total egl-43 RNAi (n = 49), combined data from two independent RNAi experiments). Consistent with a role of egl-43 during AC/VU specification [10,11], we detected early L3 larvae with two ACs upon egl-43L or total egl-43 RNAi (Fig 1C). However, in 36 out of 49 cases, total egl-43 RNAi led to the formation of more than two ACs, a phenotype that cannot be explained by an AC/VU specification defect (Fig 1C). Similar to the invasion defects, the occurrence of multiple ACs was also caused by selective inhibition of the long egl-43L isoform, with 42 out of 53 worms containing more than 2 ACs (Fig 1C). Moreover, the number of ACs increased progressively with the age of the larvae, indicating an ongoing proliferation of the AC after the L2 stage (Fig 2A and 2B). This pointed to an additional role of egl-43L in preventing the proliferation of the AC after its specification, independently of its function during VU fate specification.
In order to specifically examine the role of the egl-43S isoform during AC invasion, we used CRISPR/Cas9 engineering to introduce an ATG!CTG mutation in the predicted egl-43S start codon (egl-43LΔS::gfp, Fig 1A and 1B). The egl-43LΔS::gfp strain was viable and showed a similar expression pattern as the egl-43::gfp and gfp::egl-43L strains (Fig 1B). Moreover, AC invasion occurred normally in all egl-43LΔS::gfp animals examined, and no AC proliferation was observed (n = 35), which confirms the predominant role of egl-43L in regulating AC invasion and proliferation.
Finally, we used the Flp/FRT system to generate an AC-specific knock-out of egl-43L [17]. Since two FRT sites had been inserted in introns of the gfp cassette at the 5' end of the egl-43L locus, the expression of the FLPase under control of the AC-specific lin-3 enhancer element deletion in the Fos-Responsive Element (FRE) and of the mutation in the initiation codon of the egl-43S isoform. Solid red lines refer to the regions targeted by RNAi. (B) Expression patterns of the different endogenous egl-43 reporters depicted in (A). The basement membrane (BM) was simultaneously labelled with the LAM-1::GFP marker to score AC invasion. At least 25 animals were examined for each reporter. All animals contained one GFP expressing AC and exhibited BM breaching. A quantification of the AC expression levels is shown in S2 Fig. (C) RNAi of the different egl-43 isoforms and FLP/FRT-mediated excision of egl-43L. Left panels show Nomarski (DIC) images and middle panels the fluorescence signals of endogenous EGL-43::GFP expression in the nuclei and the LAM-1::GFP BM marker. The right panels show the GFP signals merged with the ACs labelled by the lin-3 ACEL >mCherry reporter (rows 1-4) and cdh-3>mCherry::PH (row 5) in magenta. The black arrowheads point at the AC nuclei and the white arrows at the locations of the BM breaches. Control refers to animals exposed to the empty RNAi vector. The numbers to the right indicate the penetrance of the invasion defects. A quantification of the AC expression levels is shown in S2C Fig  Elevated AC expression of GFP::MCM-7 was detected after egl-43 RNAi and egl-43L RNAi (29/29 and 28/31 cases, respectively), but not after control or egl-43S RNAi (0/31 and 0/32 cases, respectively). The left panels depict Nomarski (DIC) images and the middle panels nuclear GFP::MCM-7 expression together with the LAM-1::GFP BM marker at the mid-L3 stage. The right panels show the GFP signals merged with the ACs labelled by the cdh-3>mCherry::PH reporter in magenta. (B) Quantification of the number of ACs formed after RNAi of the different egl-43 isoforms from the early L3 (Pn.p) until the mid-L3 (Pn.pxx) stage. The error bars indicate the standard deviation. (C) Expression of the S-phase marker RNR-1::GFP together with LAM-1::GFP in control and egl-43 RNAitreated mid-L3 larvae. No RNR-1::GFP expression was detected in 48 control RNAi animals, while 30/39 egl-43 RNAi-treated animals expressed the S-phase marker. (D) Elevated expression of the Cyclin D cyd-1>gfp reporter in the multiple ACs of egl-43 RNAi-treated mid-L3 larvae was observed in 52/58 cases, (lin-3 ACEL >flp) [18] specifically inactivated egl-43L in the AC (S1 Fig). Deletion of egl-43L in the AC led to invasion defects (9 out of 17 animals) and the formation of multiple ACs (4 out of 17 cases), similar to the phenotypes caused by egl-43 RNAi (Fig 1C). The relatively low penetrance observed after Flp/FRT-mediated excision compared to RNAi could be due to the perdurance of the EGL-43 protein in the AC, as faint GFP::EGL-43 expression in the AC of mid-L3 larvae could be observed in 2 out of 17 cases.
In summary, the long egl-43L isoform acts cell-autonomously to block AC proliferation and promote invasion.

Neither the N-terminal PR nor the ZF1 domains in EGL-43 are necessary for AC invasion
egl-43L encodes a transcription factor containing an N-terminal PRD1-BF1/RIZ-1 domain (PR) and two separate clusters of Zinc finger domain (ZF1 & ZF2) that could exhibit DNA binding activity [12]. PR domains are structurally similar to the SET (Su(var)3-9, Enhancer of zeste and Trithorax) domains, which contribute to histone lysine methyltransferase activity and have also been reported to mediate protein-protein interactions [19,20]. In order to dissect the requirement of the different EGL-43 domains during AC invasion, we generated in-frame deletions in the egl-43 coding region using the CRISPR/Cas9 system (Fig 1A). In the PR domain deletion allele (gfp:: egl-43ΔPR, deleted amino acids 2-62), we detected an approximately 75% decrease in GFP::EGL-43L expression levels in the AC (Fig 1B, S2A Fig). Despite this strong reduction, neither the proliferation nor the invasion of the AC were affected as all 25 animals scored showed normal AC invasion and no proliferation. Thus, the PR domain in EGL-43 is not necessary for AC invasion. Though, the reduced expression levels suggested that the PR domain is either required for egl-43 autoregulation (see below), protein stability, or that there exist additional regulatory elements in the deleted first intron that promote egl-43 expression in the AC (Fig 1A).
Furthermore, no AC invasion defects and no change in expression levels were observed in egl-43ΔZF1::gfp mutants, which carry an in-frame deletion removing amino acids 161-235, which encode the N-terminal Zinc Finger domains (ZF1) (Fig 1B, S2B Fig). Note that this allele also deletes the promoter of the egl-43S isoform. By contrast, the egl-43 null mutant tm1802, which contains a 659 bp deletion removing the C-terminal Zinc Finger domains (ZF2), displays severe developmental defects and early embryonic lethality [11]. Thus, the regulation of AC invasion and proliferation most likely depends on the C-terminal ZF2 domains.

egl-43 is required for the G1 arrest of the AC
Previous studies have shown that the AC must remain arrested in the G1 phase in order to invade [8]. The occurrence of multiple (i.e. more than two) ACs upon inactivation of egl-43L indicated that the AC had bypassed the G1 arrest and begun to proliferate. In order to test this hypothesis, we performed RNAi knock down of the different egl-43 isoforms in a strain carrying an endogenous GFP::MCM-7 reporter. mcm-7 encodes a subunit of the pre-replication complex (pre-RC) that is highly expressed in cycling cells but down-regulated in non-proliferating cells [21,22]. No GFP::MCM-7 expression was detectable in the AC after control or egl-while all 43 control larvae showed faint cyd-1>gfp expression in the single AC. (E) Expression of the CDK activity sensor in the AC of control and egl-43 RNAitreated mid-L3 larvae together with the LAM-1::GFP BM marker. The right panels show the CDK sensor signal merged with the ACs labelled by the lin-3 ACEL >mCherry reporter in magenta. (F) Quantification of CDK sensor activity in the AC. Scatter plots showing the cytoplasmic to nuclear intensity ratio as a measure of CDK activity. The error bars indicate the standard deviation and the vertical bars in (B) and horizontal bars in (F) the mean values. Statistical significance was determined with a Student's t-test and is indicated with �� for p<0.01, ��� for p<0.001, ���� for p<0.0001 and n.s for p> 0.05. In each graph, the numbers of animals scored are indicated in brackets. The black arrowheads point at the AC nuclei and the white arrows at the locations of the BM breaches. The scale bars are 5 μm.
https://doi.org/10.1371/journal.pgen.1008470.g002 43S RNAi. However, egl-43 and egl-43L RNAi resulted in elevated GFP::MCM-7 levels in the multiple ACs that formed, indicating that these ACs had re-entered the cell cycle (Fig 2A). Consistent with this conclusion, the average number of ACs increased during the development from the Pn.p (late L2/early L3) stage to the Pn.pxx (mid to late L3) stage (Fig 2B).
To further characterize the cell cycle state of the AC, we analyzed RNR-1::GFP expression, which serves as an S-phase marker [23]. While RNR-1::GFP was absent in the AC of control RNAi animals, it was expressed in the multiple ACs formed after egl-43 RNAi (Fig 2C). Moreover, we detected elevated levels of a transcriptional cyd-1>gfp Cyclin D reporter in the multiple ACs produced after egl-43i, while the single AC in control animals showed only faint Cyclin D expression (Fig 2D). Finally, we expressed a CDK biosensor in the AC to directly quantify CDK kinase activity [24,25]. This sensor monitors CDK activity via the phosphorylation-induced nuclear export of a GFP-tagged kinase substrate. Thus, a high cytoplasmic-tonuclear signal ratio indicates high CDK activity. The multiple ACs formed in egl-43 RNAitreated animals showed a significantly increased CDK sensor activity compared to the single AC in control RNAi animals (Fig 2E and 2F).
Thus, loss of EGL-43 function increases CDK activity and triggers cell cycle entry of the AC.

egl-43 and nhr-67 play distinct roles in inducing the G1 arrest of the AC
As reported previously [8], RNAi of nhr-67 led to the appearance of multiple ACs that could not breach the BM (Fig 3A). Since loss of nhr-67 led to a similar phenotype as inhibition of egl-43, we tested whether the expression of nhr-67 depends on egl-43, or vice versa. We first measured egl-43 and nhr-67 GFP reporter expression in early L3 larvae at the Pn.p stage, shortly after AC specification had occurred and the G1 arrest had been established. Quantification of endogenous GFP::EGL-43L reporter levels after nhr-67 RNAi revealed no significant change (S3A and S3B Fig). Also, the expression of an nhr-67::gfp reporter at the early L3 stage was not significantly changed by egl-43 RNAi (Fig 3C and 3D). Moreover, the expression of the histone deacetylase hda-1, which acts downstream of nhr-67 to promote AC invasion [8], was not changed by egl-43 RNAi (S3E and S3F Fig). However, in mid-L3 larvae (Pn.pxx stage), around the time of AC invasion, nhr-67 RNAi caused an approximately 30% reduction in GFP::EGL-43 expression levels, and a similar reduction in NHR-67::GFP expression levels was observed after egl-43 RNAi (Fig 3A-3D). Thus, at later stages egl-43 and nhr-67 may positively regulate each other's expression. Due to the high penetrance of the AC proliferation phenotype caused by single nhr-67 or egl-43 RNAi, we were unable to test if the simultaneous knock-down of both transcription factors had an additive effect. Consistent with a previous report [8], the AC-specific expression of the CDK inhibitor cki-1 together with an mNeonGreen (mNG) marker on a bi-cistronic mRNA under control of a cdh-3 enhancer/promoter fragment (cdh-3>cki-1::SL2::mNG) partially rescued the AC proliferation and invasion defects caused by nhr-67 RNAi (Fig 3E-3G). By contrast, CKI-1 overexpression did not rescue the invasion defects caused by egl-43 RNAi (Fig 3E and 3F), and it only slightly inhibited AC proliferation (Fig 3G). Notably, around 10% of egl-43 RNAi treated and CKI-1 overexpressing animals contained one AC, yet their BMs were not breached (Fig 3F).
These data suggested that that egl-43 acts in parallel with nhr-67 and hda-1 to maintain the G1 arrest and promote invasion of the AC.

egl-43 inhibits AC proliferation by repressing lin-12 Notch expression
LIN-12 Notch signaling is not only critical during the AC/VU decision, but it also links differentiation to cell cycle progression in different tissues [4,26,27]. We therefore tested whether  [28]. In control animals at the mid-L3 stage, LIN-12::GFP expression had disappeared in the AC, while expression persisted in the adjacent VU cells (Fig 4A). By contrast, the multiple ACs that formed after egl-43 RNAi continued to express LIN-12::GFP (Fig 4A). The ACs in egl-43 RNAi-treated larvae still expressed a lin-3::mNG reporter, which serves as a marker to distinguish the AC from the VU fate [29], as well as a reporter for the LIN-12 ligand LAG-2 (S4A and S4B Fig) [30]. Thus, the inhibition of egl-43 did not cause a transformation of the AC into a VU fate, but rather resulted in the ectopic expression of LIN-12 in the proliferating ACs.
To test whether an over-activation of lin-12 Notch signaling in the AC is responsible for the AC proliferation phenotype, we performed double RNAi of egl-43 and lin-12 and scored the number of ACs, as well as their ability to invade. While 63% of egl-43 single RNAi-treated animals formed multiple (i.e. more than one) ACs, only 16% of egl-43; lin-12 double RNAi-treated animals contained multiple ACs. The average number of ACs per animals decreased from 2.6 in egl-43 single to 1.2 in egl-43; lin-12 double RNAi-treated animals (Fig 4B), and the penetrance of the AC invasion defect decreased from 72% to 14% (Fig 4C). Eight out of the 19 animals that had been treated with egl-43; lin-12 double RNAi and exhibited an invasion defect contained a single AC. This suggested that the egl-43 invasion phenotype is not exclusively caused by the over-proliferation of the AC.
To test if reducing lin-12 activity restored the cell cycle arrest of the AC, we examined the GFP::MCM-7 reporter, which is expressed exclusively in proliferating cells. Control or single lin-12 RNAi did not induce GFP::MCM-7 expression in the AC of mid-L3 larvae, while egl-43 RNAi resulted in the formation of multiple GFP::MCM-7 positive ACs in 91% of the cases (Figs 2A & 4D). By contrast, the single ACs formed in egl-43; lin-12 double RNAi-treated animals did not express GFP::MCM-7 in 68% of the cases (Fig 4D).
LIN-12::GFP expression was also up-regulated in the AC of nhr-67 RNAi treated animals (Fig 4A). However, inhibition of lin-12 only partially suppressed the AC over-proliferation caused by nhr-67 RNAi from 3.8 to 2.6 ACs per animal (Fig 4B) and only slightly reduced the nhr-67 invasion defects from 94% to 81% (Fig 4C).
Thus, reducing lin-12 activity suppressed the AC proliferation and invasion defects caused by egl-43 RNAi. By contrast, the nhr-67 phenotype was less sensitive to a reduction in lin-12 levels, suggesting that NHR-67 inhibits AC proliferation predominantly through another pathway.

Activation of LIN-12 Notch signaling in the differentiated AC triggers proliferation
These results led to the hypothesis that the ectopic activation of LIN-12 Notch signaling caused by loss of egl-43 function may trigger the re-entry of the AC into the cell cycle. In order to test the GFP signals merged with the ACs labelled by the lin-3 ACEL   if LIN-12 signaling is sufficient to induce AC proliferation, we expressed a fragment of the intracellular LIN-12 Notch domain, in which the C-terminal PEST degradation motif had been deleted (NICDΔCT) [26], under the control of the AC-specific cdh-3 enhancer/promoter fragment. Expression of NICDΔCT in the VPCs was shown to hyper-activate the Notch signaling pathway [26], and cdh-3-driven expression occurs only after the AC fate has been specified [31]. Three independent transgenic lines carrying the cdh-3>nicdΔct::SL2::mCherry transgene (co-expressing an mCherry marker on a bi-cistronic mRNA) exhibited an AC proliferation phenotype with an average of 3.4 ACs per animal, which is comparable to the phenotype observed after egl-43 RNAi (Fig 4E and 4F). In addition, the GFP::MCM-7 reporter was upregulated (Fig 4D), even though GFP::EGL-43L continued to be expressed in the multiple ACs of cdh-3>nicdΔct::SL2::mCherry animals (S4C Fig). In addition, NICDΔCT expression caused a penetrant AC invasion defect (Fig 4G).
Thus, the ectopic activation of the LIN-12 Notch pathway in the differentiated AC was sufficient to trigger cell cycle entry. Taken together, we propose that EGL-43 inhibits LIN-12 Notch expression to maintain the G1 arrest of the invading AC.

A positive regulatory feedback loop between egl-43 and fos-1 activates proinvasive gene expression
It has previously been reported that fos-1 positively regulates egl-43 expression in the AC and that egl-43 is required for the expression of zmp-1, cdh-3 and him-4 [10,11]. To further characterize the role of egl-43 in regulating pro-invasive gene expression, we investigated a possible mutual regulation of fos-1 and egl-43. The expression of the endogenous GFP::EGL-43L reporter in the AC was reduced approximately two-fold in homozygous fos-1(ar105) mutants compared to heterozygous fos-1(ar105)/+ control siblings at the mid-L3 stage (Fig 5A and  5B). Using CRISPR/Cas9 genome editing, we deleted the 11 bp (TTACTCATCTT) FOS-Responsive Element (FRE) [11] in the promoter region of the endogenous gfp::egl-43L reporter strain (ΔFRE>gfp::egl-43L, Fig 1A). In a heterozygous fos-1(ar105)/+ background, the expression of the mutant ΔFRE>gfp::egl-43L reporter was reduced to a similar extent as the wild-type gfp::egl-43L reporter was reduced in a homozygous fos-1(ar105) background. Since the levels of the FRE mutant reporter did not further decrease in homozygous fos-1(ar105) larvae, FOS-1 appears to control egl-43 expression mainly through this FRE (Fig 5A and 5B). While this experiment confirmed that FOS-1 up-regulates endogenous egl-43 expression, it also pointed at the existence of additional factors that activate egl-43 expression in the AC. In particular, the deletion of the FRE in egl-43 did not cause an obvious defect in AC invasion (all 23 animals scored showed normal AC invasion), suggesting that the reduction in egl-43 expression after the deletion of the FRE can be compensated by the AC.  Since egl-43 and fos-1 regulate some of the same target genes [11,15], we tested if egl-43 regulates fos-1a expression. The expression of a fos-1a::yfp reporter in the AC was reduced approximately three-fold by egl-43 RNAi (Fig 5C and 5D), indicating that egl-43 and fos-1 positively regulate each other's expression. Moreover, the expression of a transcriptional egl-43L reporter (egl-43L>gfp is a strain containing an insertion of the self-excising cassette [16] after the gfp coding sequences to terminate transcription 5' of the egl-43 coding sequences) was reduced by egl-43 RNAi (Fig 5E and 5F). Thus, egl-43 positively regulates its own expression in the AC.

Quantification of the AC numbers in mid-L3 larvae treated with the different RNAi combinations shown in (A). (C) Quantification of the AC invasion defects caused by the different RNAi treatments shown in (A). (D)
Unlike nhr-67 or egl-43, fos-1 was not required to maintain the G1 arrest of the AC, as neither the S-phase maker RNR-1::GFP nor the proliferation marker GFP::MCM-7 were up-regulated after fos-1 RNAi (S5A and S5B Fig). We thus speculated that the regulation of fos-1 by EGL-43 and the cell cycle inhibition via lin-12 repression represent two independent functions of EGL-43. Supporting this hypothesis, LIN-12::GFP expression was not up-regulated by fos-1 RNAi, while expression of NICDΔCT did not affect fos-1 expression levels in the AC (S5C-S5E Fig).
Taken together, the expression analysis indicated that egl-43 and fos-1 form a positive feedback loop that maintains high expression of both transcription factors in the AC to induce the expression of target genes required for invasion. The auto-regulation of egl-43 likely adds further robustness to this network.

EGL-43 binding is enriched at the fos-1, egl-43 and lin-12 loci
To test if the changes in fos-1, lin-12 and egl-43 reporter expression caused by egl-43 RNAi could be due to a direct regulation by EGL-43, we performed chromatin immuno-precipitation and sequencing (ChIP-seq) analysis of the endogenous EGL-43::GFP reporter at the L3 stage using anti-GFP antibodies, in two biological replicates (see extended methods). At this stage, most of the EGL-43::GFP expression was confined to cells in the somatic gonad and to approximately 30 head and 6 tail neurons [11] (and this study). The ChIP-seq analysis identified 6276 peaks of significant enrichment, 5257 of which we could associate with 3977 genes (S5 Table).
EGL-43 binding was found at the egl-43, fos-1 and lin-12 loci (Fig 6). Notably, EGL-43 was also enriched at the jun-1 locus, which encodes the homolog of the human c-JUN proto-oncogene that forms together with c-FOS the AP-1 transcription factor (S5 Table). Even though no function of JUN-1 in AC invasion has been reported, this observation might indicate that EGL-43 regulates AP-1 activity in other processes, such as ovulation or lifespan [32]. Moreover, EGL-43::GFP was enriched at other previously reported targets, including mig-10 [33], hlh-2 [34] and zmp-1 [11] (S5 Table). On the other hand, no specific binding to the nhr-67 [34] or cdh-3 [11] genes was observed, suggesting that these two genes may be indirectly regulated by EGL-43.

Discussion
An uncontrolled activation of cell invasion is one of the hallmarks of malignant cancer cells that form metastases [1]. Genetic studies in model organisms have indicated that invasive left panels show Nomarski (DIC) images, the middle panels the GFP signals of the indicated reporters in green (in (A) and (C) together with the LAM-1::GFP BM marker) and the right panels the GFP reporter signals merged with the ACs labelled with the cdh-3>mCherry::PH reporter in magenta. The black arrowheads point at the AC nuclei and the white arrows at the locations of the BM breaches. The error bars indicate the standard deviation and the horizontal bars the mean values. Statistical significance was determined with a Student's t-test and is indicated with n.s. for p>0.05, � for p<0.05 and ���� for p<0.0001. The numbers in brackets refer to the numbers of animals analyzed. The scale bars are 5 μm. https://doi.org/10.1371/journal.pgen.1008470.g005 tumor cells re-activate the same molecular pathways that control cell invasion during normal animal development. AC invasion in C. elegans has served as an excellent model to dissect the genetic pathways regulating the invasive phenotype of a single cell [3].
Here, we report that the egl-43 gene, the C. elegans ortholog of the human Evi1 proto-oncogene, functions in a regulatory network together with the transcription factors NHR-67 and FOS-1 to control AC invasion (Fig 7). Our results are consistent with a recent report by Medwig-Kinney et al. [34] demonstrating similar interactions between egl-43, nhr-67 and fos-1 during AC invasion. The inclusion of the LIN-12 NOTCH pathway in our model further expands Firstly, EGL-43 maintains the AC arrested in the G1 phase of the cell cycle by repressing the expression of the LIN-12 Notch receptor. The ectopic activation of Notch signaling in the differentiated AC was sufficient to induce proliferation in the presence of EGL-43, while reducing lin-12 expression efficiently suppressed the AC proliferation and invasion defects caused by inhibition of egl-43. Thus, LIN-12 is an essential downstream target of EGL-43 that can reactivate the cell cycle in the AC. A recent study in C. elegans has highlighted the importance of the LIN-12 Notch pathway in keeping an equilibrium between the proliferation and differentiation of somatic cells [27]. Furthermore, the regulation of different cell cycle genes by the Notch pathway has been reported in several cases. For example, Notch signaling induces cyclin D1 expression in mammalian kidney and breast epithelial cells [35,36] and activates dE2F1 and cyclin A expression in the photoreceptor precursors of the Drosophila eye to promote S-phase entry [37].
It was previously shown that NHR-67 maintains the G1 arrest of the AC by inducing the expression of the CDK inhibitor CKI-1 [8]. While overexpression of CKI-1 efficiently rescued the AC proliferation and invasion phenotype caused by nhr-67 RNAi, CKI-1 only slightly suppressed the AC proliferation phenotype and had no effect on the invasion defects caused by egl-43 RNAi. However, it should be noted that the transgene we used to overexpress CKI-1 did not fully suppress the nhr-67 AC proliferation and invasion phenotypes, while a different cki-1 overexpression transgene used by Medwig-Kinney et al. [34] caused a complete rescue of nhr-67 RNAi, probably due to higher levels of CKI-1 expression. It is therefore possible that a further increase in CKI-1 levels beyond the concentration we reached may also suppress the egl-43 phenotype. Taken together, we suggest that the AC proliferation phenotype caused by inhibition of egl-43 is less sensitive to an increase in the CKI-1 dosage than the nhr-67 phenotype.
On the other hand, the nhr-67 AC proliferation and invasion phenotypes were less sensitive to lin-12 inhibition when compared to the egl-43 phenotype, even though NHR-67 RNAi also increased lin-12 expression in the AC. We thus propose that EGL-43 inhibits AC proliferation predominately by repressing LIN-12 NOTCH signaling, while NHR-67 acts primarily by enhancing CKI-1 expression (Fig 7). One possible explanation for the different sensitivities could be that the hyper-activation of LIN-12 signaling caused by loss of egl-43 results in elevated CDK/Cyclin activity, which overcomes a threshold set by CKI-1-mediated cell cycle inhibition. The inhibition of nhr-67, on the other hand, may reduce the threshold by decreasing CKI-1 expression. Since nhr-67 and egl-43 positively regulate each other's expression in the proliferating ACs, EGL-43 may promote cki-1 expression indirectly via nhr-67.
The egl-43 AC invasion and proliferation phenotypes did not completely correlate, as a fraction of egl-43 depleted animals -especially in combination with lin-12 RNAi-contained a single AC that failed to invade. Hence, EGL-43 appears to have an additional function besides merely preventing AC proliferation. EGL-43 likely exerts the proliferation-independent function through its interaction with fos-1, which is not required for the G1 arrest of the AC but necessary to induce the expression of pro-invasive genes. Deleting the FOS-1 responsive element (FRE) in the endogenous egl-43 locus confirmed our earlier findings based on transgenic reporter analysis that egl-43L expression is positively regulated by FOS-1 [11]. FOS-1 is not absolutely required for the expression of EGL-43, because additional factors such as HLH-2 activate egl-43 expression in the AC [10]. Moreover, EGL-43 positively regulates fos-1 as well as its own expression in the AC. Thanks to the positive feedback loop between egl-43 and fos-1, low levels of either of the two transcription factors may be sufficient to induce a stable expression of both transcription factors and thereby irreversibly determine the invasive fate of the AC. Interestingly, a similar self-activating function has been described for mammalian Evi1 in hematopoietic stem cells [12].
In summary, EGL-43 coordinates the expression of pro-invasive genes with the cell cycle arrest of the AC by inducing fos-1 and inhibiting lin-12 Notch expression. Thus, EGL-43 is a central component in a regulatory network, which decides whether to divide or invade.

C. elegans culture and maintenance
C. elegans strains were maintained at 20˚C on standard nematode growth plates as described [38]. The wild-type strain was C. elegans Bristol, variety N2. We refer to translational protein fusions with a :: symbol between the gene name and the tag, while transcriptional fusions are indicated with a > symbol between the promoter/enhancer and the tag. The genotypes of the strains used in this study are listed in S1 Table. The construction of the plasmids, oligonucleotides and sgRNAs used to generate the different reporters are described in the Extended Methods in S1 Text and in S2-S4 Tables.

Scoring the AC invasion phenotype
AC invasion was scored in mid-L3 larvae after the VPC had divided twice (Pn.pxx stage) as described [11]. We monitored the continuity of the BM by DIC or fluorescence microscopy using the qyIs10[lam-1>lam-1::gfp] transgene as a marker.

RNA interference
RNAi interference was done by feeding dsRNA-producing E. coli [39]. Larvae were synchronized at the L1 stage by hypochlorite treatment of gravid adults and plated on NGM plates containing 3mM IPTG. P0 animals were analyzed after 30-36 hours of treatment. For double RNAi experiments, bacteria of the indicated clones were mixed at a 1:1 ration. RNAi clones targeting genes of interest were obtained from the C. elegans genome-wide RNAi library or the C. elegans open reading frame (ORFeome) RNAi library (both from Source BioScience). RNAi vectors targeting egl-43L and egl-43S were subcloned into the L4440 vector by Gibson assembly (S3 Table). The empty L4440 vector (labelled "control" in the figures) was used as negative control in all experiments.

Microscopy and image analysis
Fluorescent and Nomarski images were acquired with a LEICA DM6000B microscope equipped with a Leica DFC360 FX camera and a 63x (N.A. 1.32) oil-immersion lens, or with an Olympus BX61 wide-field microscope equipped with a X-light spinning disc confocal system, a 100x Plan Apo (N.A. 1.4) lens, a lumencor light engine as light source and an iXon ultra888 EMCCD camera. Worms were imaged with 100 x magnification and z-stacks with a spacing of 0.1 to 0.8 μm were recorded. The Fiji software [40] was used for image analysis and fluorescent intensity quantifications using the built-in measurement tools as follows. To quantify expression of the different reporters, deconvolved optical z-sections across the AC were used to generate summed z-projections. The region of the AC nucleus was manually selected, and the integrated signal intensity was measured in the AC nucleus, from which a background value measured in an identically sized region outside of the animal was subtracted. To quantify the CDK sensor activity, images were processed with a Gaussian blur filter (sigma = 50), and a single mid-sagittal z-slice through the AC nucleus was selected for the measurements. The average of the integrated intensities in three equally sized and randomly selected areas, each in the cytoplasm (I c ) and the nucleus (I n ) of the AC were measured to calculate the cytoplasmic to nuclear (I c /I n ) intensity ratios, which are plotted in Fig 2F.