Knockout of all ErbB-family genes delineates their roles in proliferation, survival and migration Free

The epidermal growth factor receptor (EGFR) gene was originally identified as the ErbB oncogene, and it became the first known member of the ErbB-family of receptors, which consists of ErbB1 (also known as ErbB, EGFR and HER1), ErbB2 (HER2), ErbB3 (HER3) and ErbB4 (HER4) (Linggi and Carpenter, 2006; Yarden and Sliwkowski, 2001) (Fig. S1A). ErbB1 binds to and is activated by a family of ligands that include epidermal growth factor (EGF), transforming growth factor α (TGFα), heparin-binding EGF-like growth factor (HBEGF), betacellulin (BTC), epiregulin (EREG), amphiregulin (AREG) and epigen (Harris et al., 2003). ErbB4 binds to HBEGF, EREG, BTC and neuregulins (NRGs) 1 to 4. ErbB3 binds to NRG1 and NRG2 but lacks tyrosine kinase activity. By contrast, ErbB2 has no known ligands and instead has a constitutively active conformation; it functions as a heterodimer with the other ErbB-family receptors, particularly ErbB3 (Moasser, 2007; Sliwkowski, 2003).

Previous extensive research has clarified the difference among the four ErbB-family receptors concerning binding affinity to the ligands, intracellular signal transduction pathways and the contribution to oncogenesis (Linggi and Carpenter, 2006; Singh et al., 2016; Sorkin and Goh, 2009; Wilson et al., 2009; Yarden and Sliwkowski, 2001). The physiological relevance of the ErbB-family receptors is exemplified by the fact that mice without any of the ErbB-family genes cannot survive until adulthood (Roskoski, 2014). On the other hand, their physiological roles at cellular resolution are not necessarily clear, partly because much of the current knowledge is based on exogenous bolus application of the ligands to tissue culture cells, and partly because both the regions upstream and downstream of the ErbB-family receptors are redundant.

In epithelial cells, the saturation cell density is achieved by growth arrest, inhibition of locomotion, and extrusion with or without cell death (McClatchey and Yap, 2012; Ohsawa et al., 2018; Yu et al., 2015). It is widely accepted that the Hippo pathway, comprising MST1 and MST2, LATS1 and LATS2, and Yap1 (the mammalian forms of Hippo, Wts and Yki proteins, respectively), plays a major role in contact inhibition of proliferation and regulation of apoptosis of a wide variety of animals (Yu et al., 2015). Meanwhile, inhibition of growth factor receptors is also known to contribute to the contact inhibition of cell proliferation (McClatchey and Yap, 2012). For example, E-cadherin at high cell density sequesters EGFR (ErbB1), thereby suppressing cell growth in Madin–Darby canine kidney (MDCK) cells (Qian et al., 2004). Extracellular signal-regulated kinase 1 and 2 (ERK1/2, also known as MAPK3 and MAPK1, respectively; hereafter collectively referred to as ERK) and AKT family protein activity waves act as survival signals that suppress apoptosis, which implies that the ErbB family contributes to maintain a dense epithelium (Gagliardi et al., 2021). Among signaling molecules downstream of EGFR (ErbB1), ERK exhibits an inverse correlation with the cell density (Aoki et al., 2013; Li et al., 2004).

ErbB-family receptors promote collective cell migration of MDCK cells by propagating ERK activation waves (Aoki et al., 2017; Hino et al., 2020). It was previously found that all four ErbB1 ligands expressed in MDCK cells contribute to the ERK activation waves in MDCK cells to some degree, indicating that there is substantial redundancy in their function (Lin et al., 2021). Here, to determine the roles played by each ErbB receptor in MDCK cells, we knocked out each of the four ErbB genes singly, multiple ErbB genes in various combinations or all four of the ErbB genes. We found that propagation of the ERK activation waves during collective cell migration was mediated primarily by ErbB1 and secondarily by the ErbB2–ErbB3 heterodimer. For the progression of G1/S, either ErbB1 or ErbB2–ErbB3 is sufficient. Finally, we noticed that ligand-independent ErbB2 activity is sufficient to generate an anti-apoptotic signal at saturation cell density. Our series of ErbB-knockout cell lines is expected to promote understanding of the physiological roles played by the endogenous ErbB-family receptors and their ligands.

To examine the roles played by the ErbB-family receptors in MDCK cells, we employed CRISPR/Cas9-mediated knockout (Fig. 1A). The resulting cell lines with the knockout of a single ErbB-family gene were named as described and are denoted dErbB1, dErbB2, dErbB3, and dErbB4 hereafter. We next developed MDCK cells deficient in all ErbB-family receptors by sequential knockout. ErbB1 and ErbB4 were first knocked (denoted dErbB1/4), and then ErbB3 was additionally knocked out (denoted dErbB1/3/4). Finally, by knocking out ErbB2, an ErbB quadruple-knockout MDCK cell line was developed, and denoted Erbock (short for ErbB-knockout). We also developed Erbock cells by simultaneously introducing gRNAs for all four ErbB genes to avoid any cloning artifacts. These cell lines were designated using numbers, letters and/or capitals following the hash sign (#) to identify independent clones. Cell lines without a # in their names were not subjected to single-cell cloning. The sequences of the target genes and their karyotype, including the 4KO cells, were validated by genome sequencing (Figs S1B and S2). The 4KO MDCK cells are deficient in all ErbB1 ligands expressed in MDCK cells (i.e. EGF, TGFα, HBEGF and EREG) (Lin et al., 2021). In agreement with a previous report (Omeir et al., 2015), trisomy of chromosomes 14 and 17 was found in the wild-type (WT) MDCK cells. Two of the three Erbock clones – clones #5 and #D – exhibited normoploidy of chromosome 17. Because these two clones were developed in independent experiments (Fig. 1A), this alteration might have added an advantage for their survival during the single-cell cloning procedure. Next, we monitored the ERK activity with the FRET biosensor EKARrEV-NLS (Lin et al., 2021) (Fig. 1B). In comparison with the WT MDCK cells, the basal ERK activity was decreased in cells deficient in ErbB1. Consistent with the specificity of EGF to ErbB1, the response to exogenous EGF was lost in dErbB1#1 cells. The response to NRG1 was lost in dErbB2 and dErbB3 cells, indicating that the ErbB2–ErbB3 heterodimer is the sole receptor of NRG1 and that the ErbB1–ErbB3 heterodimer does not respond to NRG1 in MDCK cells. Because EGF, TGFα, HBEGF and EREG do not bind to ErbB3 (Harris et al., 2003), the ErbB1 and the ErbB2–ErbB3 pathways can be segregated at the level of ligands in MDCK cells. ErbB4 mRNA was barely detectable in MDCK cells (Fig. S2C; an analysis of RNA-Seq data from Watabe et al., 2023 preprint), hence dErbB4 cells were not subjected to the analysis. Decreased ERK activity in dErbB1, Erbock#5, and Erbock#12 cells was also confirmed by western blotting (Fig. 1C).

Note that dErbB2 is substantially regarded as a cell line preserving endogenous ErbB1 expression. First, the expression of ErbB4 is marginal in MDCK cells (Fig. S2C). Second, we could not detect the response of the ErbB1–ErbB3 heterodimer in MDCK cells (Fig. 1B). Third, any ligand-independent ErbB2 activity should be prevented by knocking out the ErbB2 gene.

For the analysis of collective cell migration, MDCK cells were seeded within a culture insert 1 day before observation. After the removal of the culture insert, cells were imaged in the presence of 10% serum (Fig. 1D). Clear ERK activation waves appeared beginning at 8 h after culture insert removal, as reported previously (Lin et al., 2021). ErbB1 deficiency in dErbB1#1 and dErbB1#2 suppressed ERK activation waves to a substantial degree, but not completely. Single knockout of the other ErbB genes did not affect the ERK activation waves to a detectable level. Almost complete elimination of ERK activation waves was achieved in dErbB134 and Erbock#D cells. Thus, we conclude that the ERK activation waves during collective cell migration are mediated primarily by ErbB1 and secondarily by the ErbB2–ErbB3 heterodimer. This is consistent with our observation of MDCK cells deficient in ErbB1 ligands. Note that the elimination of all four ErbB1 ligands reduced ERK activation propagation markedly, but not completely, suggesting that there is contribution by NRG1 (Lin et al., 2021).

Next, we examined the role of the ErbB-family receptors in cell proliferation (Fig. 2A). The doubling time of the WT MDCK cells was ∼14 h. Knockout of any single ErbB gene barely affected the doubling time. Meanwhile, the doubling time of dErbB134 and Erbock cells was ∼20 h. Because ErbB3 requires the other tyrosine kinase activity-competent ErbB-family receptors for its function, this observation indicates that ErbB1 or the ErbB2–ErbB3 heterodimer suffices for the promotion of cell proliferation.

In dErbB134 cells, which retain only ErbB2, the doubling time was extended to the level of the Erbock cells. Because ErbB4 is not expressed in MDCK cells, this observation indicates that either ErbB1 or the ErbB2–ErbB3 heterodimer is sufficient for the G1/S progression of MDCK cells. This data again agrees with the previous observation that NRG1 knockout markedly slows down replication of the 4KO MDCK cells that lack all four ErbB1 ligands (Lin et al., 2021). Notably, exogenous ErbB2 expression in Erbock cells restored the doubling time to the level of WT cells, indicating that overexpression of ErbB2 could override the lack of ligand-mediated activation of the ErbB2–ErbB3 complex. This observation is consistent with the observation that ErbB2 is overexpressed in many human cancer cells (Baselga and Swain, 2009).

For a more detailed analysis, we introduced the Fucci5 cell cycle sensor (Ando et al., 2023) into MDCK cells (Fig. 2B–D). The G1 period was prolonged from 7 h in WT to 12 h in Erbock cells. Meanwhile, the G2/M period was 7 h in both WT and Erbock cells. Thus, we concluded that the endogenous ErbB1 and the ErbB2–ErbB3 heterodimer redundantly contribute to the G1/S transition in MDCK cells.

During the above experiments, we noticed that the saturation density of Erbock cells was markedly lower than that of the WT cells. To confirm this finding, the cells were seeded at different densities and observed for 5 days (Fig. 3A). The WT cells reached a saturation cell density of 4×10³ cells mm⁻² irrespective of the initial cell density. Erbock cells reached a saturation cell density of 2×10³ cells mm⁻² at later time points. To gain an insight into the mechanism of the reduced saturation cell density, we plated Erbock cells at 4×10³ cells mm⁻². Because a large proportion of the seeded cells failed to attach to the dish in 24 h, we could not accurately count the live cells, but this observation suggested that the low saturation cell density of Erbock is at least partly caused by cell elimination. Re-expression of ErbB1 or ErbB2, but not ErbB3 or ErbB4, restored the saturation cell density of Erbock cells to that of the WT cells (Fig. 3B), suggesting that either ErbB1 or ErbB2 is sufficient for achieving the high cell density. This saturation cell density was affected by the serum concentration, but the dependence on ErbB genes was still significant (Fig. 3C). Because the re-expression experiment (Fig. 2B) suffers from the effect of overexpression, we examined whether endogenous ErbB1 and ErbB2 suffice for achieving high cell density. The dErbB134 cells, which express only ErbB2, retained the saturation cell density of the WT cells (Fig. 3D). We used dErbB2 cells to examine the contribution of ErbB1. This is because the expression of ErbB4 is negligible in MDCK cells, and ErbB3 would not function without ErbB2. We found that dErbB2 cells also retained the saturation cell density of the WT cells (Fig. 3D). Thus, either ErbB1 or ErbB2 is sufficient for the saturation cell density of the WT cells. To gain further insight into the mechanism of ErbB regulation of saturation cell density, we expressed ErbB1 mutants in Erbock#5 cells (Fig. 3E). The tyrosine kinase activity-deficient M721 mutant and the C-terminus truncation DEL1011 mutant failed to restore the saturation cell density, suggesting that tyrosine phosphorylation of the C-terminus of ErbB1 contributes to the maintenance of the saturation cell density of WT cells. F992-EGFR, which is known as an ERK activation mutant, restored the saturation cell density, implying the relevance of ERK activity.

Previous reports have shown that ErbB1 signaling is suppressed at high cell density and that the reduced sensitivity of ErbB1 to EGF is the cause of contact inhibition of proliferation (LeVea et al., 2004; Takahashi and Suzuki, 1996). In Drosophila pupal notum, suppression of ErbB1-ERK signaling causes the elimination of epithelial cells during development (Moreno et al., 2019). A similar compaction-induced apoptotic signal might operate in MDCK cells. Importantly, NRG1-independent basal ErbB2 activity suffices for the anti-apoptotic activity in MDCK cells.

At high cell density, YAP1 is phosphorylated and excluded from nuclei, shutting off YAP1-mediated transcription and cell proliferation (Yu et al., 2015). However, the YAP1 translocation to the cytoplasm was found to take place when cells were at a density of over 4×10³ cells mm⁻² (Fig. S3A,B), which is higher than the saturation cell density of Erbock cells. Therefore, we excluded YAP1 as a potential cause of the low saturation cell density of Erbock cells.

How do the Erbock cells stay at the low saturation cell density? The fraction of S/G2 cells was significantly higher in Erbock cells than the WT cells at saturation cell density, indicating that the G0/G1 arrest might not be the primary reason (Fig. 4A,B). Note that the size of the nuclei of Erbock cells appears to be almost twice that of WT cells (Fig. 4A); however, the height of the nuclei of Erbock cells was approximately half that of WT cells (Fig. S4C), suggesting that the volume of the nuclei of Erbock cells was comparable to that of WT cells. Moreover, we noticed that Erbcok cells were often excluded from the monolayer sheet at high cell density. Thus, we reasoned that the confluent Erbock cells might be eliminated by apoptosis. Using the FRET biosensor for caspase-3, SCAT3 (Takemoto et al., 2003), we found that these extruded cells were apoptotic (Fig. 4C,D). The increased CFP/FRET ratio of SCAT3-expressing cells was prevented by treatment with a pan-caspase inhibitor, confirming that the ratio change of SCAT3 was caused by apoptosis (Fig. S4A,B). Moreover, time-lapse imaging showed that caspase-3 activation preceded the extrusion of cells (Fig. 4E). To gain insight into the mechanism of apoptosis induction, we examined the effect of the EGFR inhibitor PD153035, the ErbB2 inhibitor Mubritinib, the MEK inhibitor Trametinib, and the phosphoinositide 3-kinase (PI3K) inhibitor LY294002 in confluent WT cells (Fig. 4F,G). Apoptotic cells were markedly induced by Trametinib as well as by the combination of PD153035 and Mubritinib treatment in confluent WT MDCK cells, but not by LY294002. We confirmed that phosphorylation of AKT family proteins, a substrate of PI3K, was markedly inhibited by LY294002 (Fig. S4D). Note that, at low cell density, the apoptosis induction by the MEK inhibitor was abrogated (Fig. 4G). Thus, we conclude that the ErbB-ERK pathway antagonizes apoptotic stress signals specifically at high cell density. Interestingly, dErbB134 cells grew as slowly as did Erbock cells (Fig. 2A) but reached a high cell density (a comparable level to WT cells) (Fig. 3D). Thus, ErbB2 might contribute more to the cell survival than the growth rate.

How can we integrate the wide range of information generated by ErbB-family receptors? (Fig. 4H) The reason that ErbB1 plays the principal role in ERK activation waves during collective cell migration probably involves the difference between the four ErbB1 ligands and the ErbB3 ligand NRG1, which is currently under investigation in our laboratory. The finding that ErbB2 requires ErbB3 for the G1/S progression, but not for anti-apoptotic activity at saturation cell density might be explained by a difference in the threshold of ERK activity. ErbB2 adopts an active conformation even in the absence of ligands (Baselga and Swain, 2009). This basal ErbB2 activity might be over the threshold that suffices for preventing apoptosis. Meanwhile, for the G1/S progression, the threshold of ERK activity could be higher, so strong signals from the ErbB2–ErbB3 heterodimer might be needed. We, therefore, examined the possibility that the difference in the downstream signaling proteins between ErbB1 and ErbB2 explains the difference in the ErbB3 requirement. Given that ErbB2 can potently activate PI3K (Baselga and Swain, 2009), we tested the anti-apoptotic activity of the PI3K pathway. However, we found that the MEK inhibitor, but not the PI3K inhibitor, induced apoptosis of MDCK cells at saturation cell density (Fig. 4G). Therefore, we conclude that the cell compaction-induced apoptotic signals in MDCK cells are primarily antagonized by the ErbB-ERK pathway, as reported in Drosophila (Moreno et al., 2019).

In conclusion, this study has delineated the contribution of each ErbB-family receptor to collective cell migration, cell proliferation and anti-apoptotic activity at saturation cell density. Considering the wide use of MDCK cells in cell biology, the series of mutants developed here will provide a useful platform to untangle the intercellular signal transduction mediated by ErbB-family receptors and their ligands.

Reagents, antibodies, plasmids and primers are described in Table S1.

MDCK (ECACC 84121903) cells were obtained from the European Collection of Authenticated Cell Cultures (ECACC) via the RIKEN BioResource Center (no. RCB0995). Lenti-X 293 T cells were purchased from Clontech (no. 632180). These cells were maintained in DMEM (no. 044-29765; FUJIFILM Wako Pure Chemical Corporation) supplemented with 10% FBS (no.172012-500ML; Sigma-Aldrich), 100 units ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin (no. 26253-84; Nacalai Tesque) in a 5% CO₂ humidified incubator at 37°C.

cDNAs of human ErbB-family genes were purchased from Addgene EGFR (no. 11011), ErbB2 (no. 40978), ErbB3 (no. 82114), and ErbB4 (no. 23875). cDNAs of EGFR mutants were obtained from Dr Noriko Gotoh (Kanazawa University, Japan; Gotoh et al., 1994). These cDNAs were inserted into pPB-derived vectors (Yusa et al., 2009).

For CRISPR/Cas9-mediated single or multiple knockouts of genes encoding ErbB-family receptors, single guide RNAs (sgRNA) targeting the exons were designed using CRISPRdirect (Naito et al., 2015). The gRNA sequences are listed in Table S1. Oligonucleotide DNAs for the sgRNA were cloned into the lentiCRISPRv2 (Addgene no. 52961; Sanjana et al., 2014), pX458 (Addgene no. 48138; Ran et al., 2013) or pX459 (Addgene no. 62988). The expression plasmids for sgRNA and Cas9 were introduced into MDCK cells by lentiviral infection or electroporation. For lentivirus production, the lentiCRISPRv2-derived expression plasmid, psPAX2 (Addgene no. 12260), and pCMV-VSV-G-RSV-Rev (Miyoshi et al., 1998) were co-transfected into Lenti-X 293T cells using polyethylenimine (no. 24765-1; Polyscience Inc.). The infected cells were selected with media containing the following antibiotics, depending on the drug resistance genes carried by the lentiCRISPRv2-derived plasmids: 100 µg ml⁻¹ zeocin (no. 11006-33-0; InvivoGen) or 2.0 µg ml⁻¹ puromycin (no. P-8833; Sigma-Aldrich). For electroporation, pX459-derived expression plasmids were transfected into MDCK cells by an Amaxa Nucleofector II (Lonza). The transfected cells were selected with 2.0 µg ml⁻¹ puromycin. For the production of MDCK cells deficient in all ErbB-family receptors, 10 ng ml⁻¹ HGF (Table S1) was added to the medium 12 h before electroporation, and also to the medium for single-cell cloning.

After single-cell cloning, genomic DNAs were isolated with SimplePrep reagent (no. 9180; TAKARA Bio) or DNeasy Blood & Tissue Kits (QIAGEN) according to the manufacturer's instruction. Genome libraries were prepared by using the TruSeq DNA PCR-free Library (Illumina). 150 bp paired-end sequencing was performed with NovaSeq6000 (Illumina). The average sequence depth was set at 1.7 or 30 times. Library preparation and whole genome sequencing were performed by Macrogen Japan. Whole genome-analysis data were trimmed using Trim Galore version 0.6.6 and Cutadapt version 2.8. Obtained reads were mapped to the ROS_Cfam_1.0 reference sequence using HISAT2 version 2.2.1, and sorted and indexed using SAMtools version 1.7. Pysamstats version 1.1.2 was used with the following options to calculate the depth of reads mapped to the reference sequence: --type coverage_binned, --window-size 1000000, --min-mapq 1, --no-dup. Pindel version 0.2.5b9 was used with the following option for the detection of insertions and deletions: insert-size 400. Among the insertions and deletions detected, those with a read count suggesting insertion or deletion in less than 10 samples out of all samples analyzed and those with a read count suggesting insertion or deletion in less than 1 sample were excluded from the analysis. Detection of insertions and deletions within exons was based on the ROS_Cfam_1.0 gene annotation file.

RNA-Seq analysis of MDCK cells was reported previously (Watabe et al., 2023 preprint). The sequence data are available in the DNA Data Bank of Japan Sequence Read Archive under accession numbers DRR014156 to DRR014161. The fastq files were trimmed with Trim Galore version 0.6.6 and Cutadapt version 2.8. FastQC version 0.11.9 was used for the quality check and filtering. Obtained reads were mapped to the reference sequence of canFam3.1 by using HISAT2 version 2.2.1. SAMtools version 1.7 was used for sorting and indexing. The relative amount of mRNA was calculated as FPKM with StringTie version 2.1.4. The amount of mRNA was normalized to the FPKM value.

MDCK cells expressing EKARrEV-NLS were reported previously (Lin et al., 2021). SCAT3NES, tFucci(CA)5, and ErbB-family receptors were stably introduced into MDCK cells by transposon-mediated gene transfer (Yusa et al., 2009). H2B-iRFP was introduced by lentivirus-mediated gene transfer. The expression plasmids and established cell lines are summarized in Table S1.

Fluorescence images were acquired essentially as described previously (Aoki and Matsuda, 2009). Briefly, cells cultured on glass-base dishes were observed under an IX81 inverted microscope (Olympus) equipped with a UPlanAPO 10×/0.40 objective lens (Olympus), an IX2-ZDC laser-based autofocusing system (Olympus), an electric XY stage (Sigma Koki), a QI-695 CCD camera (Molecular Devices), CoolLED pE (Molecular Devices), and a stage top incubator (Tokai Hit). The filters and dichromatic mirrors used for time-lapse imaging were as follows: for FRET imaging, a 438/24 excitation filter, an FF458-Di02-25×36 dichromatic mirror (Semrock), and FF01-483/32-25 and FF01-542/27-25 emission filters (Semrock) for CFP and FRET, respectively.

We also used an ECLIPSE Ti2 inverted microscope (Nikon Instruments Inc.) equipped with PlanFluor Ph1 10×/0.30 (Nikon), S Plan Fluor LWD ADM Ph2 20×/0.70 (Nikon), and S Plan Fluor ELWD ADM Ph2 40×/0.6 (Nikon) lenses, a Perfect Focus System (Nikon), a TI2-S-SE-E XY stage (Nikon), an ORCA Fusion CMOS camera (Hamamatsu Photonics K.K., Hamamatsu, Japan), an X-Cite TURBO LED (Excelitas Technologies), and a stage top incubator (Tokai Hit). The filters and dichromatic mirrors used for time-lapse imaging were as follows: for FRET imaging, a 434/32 excitation filter, a 455 dichromatic mirror, and 480/40 and 535/30 emission filters (Nikon) for CFP and FRET, respectively.

In some experiments, an incubator-type FV10i confocal microscope (Olympus) was also used. The LD laser, filters, and wavelength ranges used for time-lapse imaging were as follows: for FRET imaging, a 405 nm laser diode and 460–510 nm emission range and 515–615 nm emission range for CFP and FRET, respectively. For h2-3, a 473 nm laser diode and 490–590 nm emission range. For Azalea B5, a 559 nm laser diode and 570–600 emission range. For iRFP, a 635 nm laser diode and 660–760 nm emission range.

Image processing for FRET/CFP ratio images was performed with Fiji essentially as described previously (Lin et al., 2021). The background intensity was subtracted by using the subtract-background function. The processed images were subjected to image calculation and the ratio values were binned into eight steps to obtain 8-color FRET/CFP ratio images. To convey the brightness of the original images to the FRET/CFP ratio images, the 8-color FRET/CFP ratio images were multiplied by the corresponding intensity-normalized grayscale image. For the 8-bit pseudocolor FRET/CFP ratio images, CFP images were subjected to thresholding with the Otsu method by Fiji to obtain binary images of the nucleus. Then, ratio images without binning of the ratio values were multiplied by the binary images of the nucleus to remove signals in the cell-free regions. To measure the time course of the FRET/CFP ratio in each cell, the FIJI TrackMate plug-in (Tinevez et al., 2017) was applied to the CFP fluorescence images to acquire the track of each cell. To obtain the kymographs, values were averaged along the y-axis in a defined region of the images, providing an intensity line along the x-axis.

MDCK cells were plated on six-well plates (Thermo Fisher Scientific) at a saturation cell density and cultured for 24 h without FBS. Cells were lysed in lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 5 mM EDTA, and 10% glycerol). Insoluble material was removed by centrifugation at 20,000 g for 10 min at 4°C. The supernatant was mixed with SDS sample buffer and subjected to SDS-PAGE. Proteins were transferred to a PVDF membrane (Merck Millipore), probed with the primary and secondary antibodies (Table S1) diluted at 1:1000 or 1:10,000, respectively, in Odyssey blocking buffer (LI-COR Biosciences), and detected with an Odyssey Infrared Imaging System (LI-COR Biosciences). Whole images of blots form this paper are shown in Fig. S5.

For the ratio imaging, 10⁴ MDCK cells were seeded into a 96-well glass-bottom dish (Matsunami Glass Ind.) coated with 0.3 mg ml⁻¹ type I collagen (Nitta Gelatin). After 22.5 h of incubation, the medium was replaced with Medium 199 (11043023; Life Technologies) supplemented with 100 units ml⁻¹ penicillin and 100 µg ml⁻¹ streptomycin. At 2.5 h after the medium replacement, cells were observed under the ECLIPSE Ti2 microscope. During observation EGF or NRG1 was added to 100 ng ml⁻¹.

2.8×10⁴ MDCK cells were seeded into a two-well culture insert placed on a collagen-coated glass-bottom dish and incubated for 20 h. After removing the culture insert, the medium was replaced with Medium 199 supplemented with 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin and 10% FBS. At 30 min after the medium replacement, cells were observed under the ECLIPSE Ti2 inverted microscope.

WT and Erbock#5 cells expressing tFucci(CA)5 and H2B-iRFP were plated into a two-well culture insert placed on a collagen-coated glass-bottom dish at a density of 2.5×10² cells mm⁻², and incubated for 23 h. The medium was replaced with Medium 199 supplemented with 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin and 10% FBS. Beginning at 1 h after the medium replacement, cells were observed under the FV10i every 10 min for 24 h. Image analysis was performed with FIJI. By applying the TrackMate plug-in to H2B-iRFP images, a series of nuclear intensity data of AzaleaB5 (M/G1) and h2-3 (S/G2/M) was acquired. The border of G1 to S/G2/M phases was set to the time points when h2-3 intensity exceeded AzaleaB5 intensity and vice versa. When neither G1 to S transition nor M to G1 transition was observed during the observation periods and when the tracking periods were longer than 8 h, either the observation start or end time was used as the surrogate of the transition times.

For time-lapse imaging, MDCK cells expressing H2B-iRFP were seeded into a two-well culture insert placed on a collagen-coated glass-bottom dish. After 23 h of incubation, the medium was replaced with Medium 199 supplemented with 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin, 100 µg ml⁻¹ zeocin and 10% FBS. Beginning at 1 h after the medium replacement, cells were observed under an FV10i confocal microscope every 60 min for 5 days. An equal volume of the same medium was added 2 days after the initiation of imaging.

For the measurement of saturation cell density, MDCK cells were seeded into a collagen-coated 96-well glass-bottom dish at a density of 3.3×10² cells mm⁻². After 23 h of incubation, the medium was replaced with Medium 199 supplemented with 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin, and 10% FBS. After 2 days, the medium was refreshed. After an additional 3-day incubation, cells were fixed with 4% paraformaldehyde for 30 min at room temperature. Nuclei were stained with 10 µg ml⁻¹ Hoechst 33342 in PBS. Cells were observed under the ECLIPSE Ti2 inverted microscope equipped with an S Plan Fluor ELWD ADM Ph2 40×/0.6 lens. Cells were counted essentially as described in the cell doubling time measurement.

MDCK WT or Erbock#5 cells expressing H2B-iRFP were plated at 1.6, 3.3, 6.7, or 10×10² cells mm⁻² on 96-well glass-bottom plates (Matsunami Glass Ind.) precoated with 0.3 mg ml⁻¹ type I-C collagen (Nitta Gelatin) and incubated for 2 days. Immunofluorescence analysis was performed as described previously (Hino et al., 2020). Briefly, cells were fixed with 4% paraformaldehyde for 30 min at room temperature and permeabilized with 0.2% Triton X-100 for 5 min. After blocking with 1% BSA in PBS for 1 h, cells were incubated with anti-YAP1 rabbit antibody (cat. no. 4912, Cell Signaling Technology) diluted with 1% BSA in PBS (1:100) overnight at 4°C. After washing, cells were stained with Alexa Fluor 488-labeled goat anti-rabbit-IgG antibody (Molecular Probes) diluted with 1% BSA in PBS (1:1000) for 1 h at room temperature. Cells were observed under the ECLIPSE Ti2 inverted microscope with the S Plan Fluor ELWD ADM Ph2 40×/0.6 lens.

WT, Erbock#5, and Erbock#12 cells expressing tFucci(CA)5 and H2B-iRFP were plated into a two-well culture insert placed on a collagen-coated glass-bottom dish at a density of 3.2×10² cells mm⁻², and incubated for 23 h. The medium was replaced with Medium 199 supplemented with 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin, 100 µg ml⁻¹ zeocin, 10 µg ml⁻¹ blasticidin and 10% FBS. Cells were set under an FV10i confocal microscope and observed every 60 min for 5 days. An equal volume of the medium was added on day 3. For the quantification, the tFucci(CA)5-expressing MDCK cells were seeded into a 96-well glass-bottom dish at a density of 3.3×10² cells mm⁻², and incubated for 23 h. The medium was replaced with Medium 199 supplemented with 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin, and 10% FBS. At 2 days later, the medium was replaced with the same fresh medium. After 3 more days, cells were fixed in 4% paraformaldehyde for 30 min at room temperature and observed with an ECLIPSE Ti2 inverted microscope equipped with an S Plan Fluor LWD ADM Ph2 20×/0.7 lens. The percentages of S, G2 and M phase cells were quantified as described for the cell cycle analysis at low cell density.

MDCK cells expressing SCAT3NES were plated into a two-well culture insert placed on a collagen-coated glass-bottom dish at a density of 1.3×10³ cells mm⁻² and incubated for 23 h. The medium was replaced with Medium 199 supplemented with 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin and 10% FBS. After 2 days, the medium was replaced with the same fresh medium. Beginning at 24 h after the medium replacement, cells were observed under an FV10i confocal microscope every 10 min for 24 h. For the quantification of apoptosis, images were segmented by using the Cellpose segmentation algorithm (Stringer et al., 2021). Cells with a CFP/FRET ratio >0.6 were scored as cells undergoing apoptosis.

For the detection of cell extrusion, MDCK cells expressing SCAT3NES were plated into a two-well culture insert placed on a collagen-coated glass-bottom dish at a density of 3.2×10² cells mm⁻², and incubated for 23 h. The medium was refreshed with Medium 199 supplemented with 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin, and 20% FBS at 23 h, 3 days, and 5 days. Beginning at 1 h after the final medium change, cells were observed under an SP8 confocal microscope with 0.68 µm Z-sectioning every 10 min for 15 h.

To examine the effect of inhibitors on apoptosis, MDCK cells expressing SCAT3NES were plated into a two-well culture insert placed on a collagen-coated glass-bottom dish at a density of 3.2×10² or 2.6×10³ cells mm⁻² and incubated for 47 h. The medium was refreshed with Medium 199 supplemented with 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin and 10% FBS. Beginning at 1 h after the medium replacement, cells were observed with the FV10i confocal microscope every 10 min for 20 h. At 3 h after the initiation of observation, DMSO, 2 µM PD153035, 0.1 µM Mubritinib, 200 nM Trametinib or 20 µM LY2940002 was added. To confirm the effect of Q-VD-OPH, SCAT3NES-expressing Erbock cells were seeded on collagen-coated 96-well glass-bottom plates and incubated for 28 h. The medium was then changed to Medium 199 containing 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin, 10% FBS, 10 µg ml⁻¹ cycloheximide, 50 ng ml⁻¹ TNFα, and either DMSO or 50 µM Q-VD-OPH. Beginning at 30 min after the medium exchange, the cells were observed under an IX81 inverted microscope every 30 min for 16 h. To inhibit apoptosis of Erbock cells at a saturated cell density, a two-well culture-insert was placed on a collagen-coated glass-bottom dish coated with 0.3 mg ml⁻¹ type I-C collagen in 1 mM HCl. SCAT3NES-expressing Erbock cells were seeded at 1.3×10³ cells mm⁻² in the two-well culture-insert and incubated for 23 h. The medium was then changed to Medium 199 containing 100 units ml⁻¹ penicillin, 100 µg ml⁻¹ streptomycin, 10% FBS, and either DMSO or 50 µM Q-VD-OPH. The medium was refreshed daily for 4 days, followed by observation with an FV10i confocal microscope every 10 min for 20 h.

Probability (P) values were determined by using a two-sample unpaired t-test implemented in MATLAB. The sample number for this calculation (n) is indicated in each figure legend.

We thank M. Miura for SCAT3NES, the members of the Matsuda Laboratory for their helpful input and encouragement, K. Hirano and K. Takakura for their technical assistance, T. Kondo and Y. Sando for supporting the RNA-seq analysis, and the Medical Research Support Center of Kyoto University for in vivo imaging. We also thank Prof. Ian G. Macara for his technical advice.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261199.reviewer-comments.pdf