Polyploid mitosis and depolyploidization promote chromosomal instability and tumor progression in a Notch-induced tumor model

Abstract

Ploidy variation is a cancer hallmark and is frequently associated with poor prognosis in high-grade cancers. Using a Drosophila solid-tumor model where oncogenic Notch drives tumorigenesis in a transition-zone microenvironment in the salivary gland imaginal ring, we find that the tumor-initiating cells normally undergo endoreplication to become polyploid. Upregulation of Notch signaling, however, induces these polyploid transition-zone cells to re-enter mitosis and undergo tumorigenesis. Growth and progression of the transition-zone tumor are fueled by a combination of polyploid mitosis, endoreplication, and depolyploidization. Both polyploid mitosis and depolyploidization are error-prone, resulting in chromosomal copy number variation and polyaneuploidy. Comparative RNA-Seq and epistasis analysis reveal that the DNA damage response genes, also active during meiosis, are upregulated in these tumors and are required for the ploidy reduction division. Together, these findings suggest that polyploidy and associated cell-cycle variants are critical for increased tumor-cell heterogeneity and genome instability during cancer progression.

In brief

Wang et al. show that polyploidy is key to tumor initiation and progression in a Notch-induced Drosophila tumor model. Polyploid mitosis, endoreplication and depolyploidization promote genomic instability and polyaneuploidy. The ploidy-reduction division depends on genes involved in DNA damage response, which are also active in meiosis.

Introduction

Variation and heterogeneity in ploidy are general features of human cancers (Chen et al., 2019; Pienta et al., 2021). This variation can be an increase of the number of chromosome set (polyploidy), an increase or decrease of a small numbers of chromosomes (aneuploidy), or regional ploidy changes and copy number variations (CNVs). Increased aneuploidy and CNVs are associated with tumor progression, malignant transformation, and metastasis (Gordon et al., 2012; Martin et al., 2015; Sen, 2000; Tang and Amon, 2013; Walther et al., 2008). Polyploidy, detected frequently in human cancers, is also positively correlated with malignancy and poor prognosis (Bou-Nader et al., 2020; Coward and Harding, 2014). Genome-wide studies of large number of cancer specimen have revealed that whole genome doubling is among the most common molecular abnormalities in human cancers and is highly associated with other copy number alterations (Bielski et al., 2018; Priestley et al., 2019; Quinton et al., 2021; Zack et al., 2013).

The role of polyploidy in tumor initiation and progression appears to be complex (Mosieniak and Sikora, 2010). The identification of polyploid giant cancer cells (PGCCs) has prompted the hypothesis that polyploidization is a key event during the transformation of normal cells into cancer cells and is the source for different types of abnormal cell divisions (Amend et al., 2019; Chen et al., 2019). DNA ploidy is specifically amplified in hepatocellular carcinoma (HCCs) and highly polyploid HCC tumors are associated with poor prognosis (Bou-Nader et al., 2020). During the initial stages of liver tumorigenesis, the normally polyploid hepatocytes undergo ploidy reduction divisions, which has been shown to be an early step toward carcinogenesis (Matsumoto et al., 2021). However, polyploidy may also play a tumor-suppressive role by buffering genotoxic damages in the liver (Zhang et al., 2018).

The Drosophila tumor models, ranging from induced hyperplastic and/or neoplastic overgrowth in imaginal tissues to transplanted brain tumors that show metastatic behavior and immortality (Castellanos et al., 2008; Dekanty et al., 2012; Janic et al., 2010), have been instrumental in addressing questions fundamental to cancer biology (Chatterjee and Deng, 2019; Miles et al., 2011; Rudrapatna et al., 2012). We have recently reported a tumor model in the Drosophila larval salivary gland imaginal ring (ImR), where upregulated Notch signaling drives neoplastic tumor growth at a narrow epithelial transition zone (TZ) that resides at the posterior end of the ImR and borders the secretary polytene salivary gland cells (Figure 1A; (Yang et al., 2019)). These tumors show continued growth following transplantation in the host abdomen (Yang et al., 2019). Similar to imaginal disc cells, the ImR epithelial cells are precursors for adult structures. We found that Notch, an evolutionarily conserved pathway, controls the size of progenitor cell pool of the ImRs (Yang and Deng, 2018). At the TZ, there is a high level of endogenous Janus kinase (JAK)-signal transducer and activator of transcription (STAT) and Jun N-terminal kinase (JNK) signaling activity, which makes this region a tissue tumor hotspot for oncogenic Notch induced tumor formation (Tamori et al., 2016; Yang et al., 2019).

Figure 1. The salivary gland imaginal ring (ImR) transition-zone (TZ) cells are polyploid endoreplicating cells.

(A) A salivary gland stained with DAPI to show cell types and ploidy. The TZ cells were dotted by red line and show larger nuclei than their anterior neighbors. (B) Mmp1-Gal4 shows specific expression in the ImR TZ cells during the 3^rd larval instar (green, mCD8-GFP marks the TZ; red, phalloidin; blue, DAPI). (C) DAPI-assisted ploidy analysis of the haploid sperm (n=200), the anterior ImR cells (n=500), and the TZ cells (n=250). (D) The TZ cells (green and red dotted line) show sporadic BrdU incorporation but not pH3 staining. For BrdU analysis, we examined 1000 anterior (Ant.) cells from 45 ImRs, and 1000 posterior (Pos.) TZ cells from 254 ImRs. For mitotic index analysis, we examined 1053 anterior cells from 56 ImRs and 1548 TZ cells from 285 ImRs. (E) A salivary gland ImR with fly-FUCCI (#55124, CycB-FRP with nuclear localization signal) expression stained with Pha (white) and DAPI (blue). The TZ cells located at the posterior boundary of the ImR showed only GFP-E2f1 (green) expression, indicating these cells do not enter M phase. Scale bar, 20 μm.

Here we show that the tumor-initiating TZ cells undergo endoreplication to become polyploid during normal development. Upon continued Notch activation, these cells display a range of different cell cycle programs that include polyploid mitosis, endocycle and ploidy reducing depolyploidization, and show increased genome instability during tumor progression.

Results

The tumor-initiating imaginal ring transition-zone cells are polyploid cells

The salivary gland ImR transition-zone (TZ) cells show high levels of matrix metalloproteinase 1 (Mmp1) expression, in response to local JNK activation (Yang et al., 2019). A Gal4 insertion at the Mmp1 locus showed reporter expression in the TZ cells, similar to its antibody staining pattern (Yang et al., 2019) (Figures 1A, 1B). From 4′,6-diamidino-2-phenylindole (DAPI) staining, we found that the nuclei of the TZ cells appear to be slightly larger than those of their anterior neighboring ImR cells (Figure 1A). To determine whether the TZ cells are polyploid cells, we compared the relative intensity of DAPI signal of the ImR cells with that of the haploid Drosophila sperm and the wing imaginal disc cells. The sperm nuclei showed a single peak, indicating haploidy, whereas the anterior ImR cells are diploid (2N) with, two peaks at 2C and 4C of DNA content, similar to the diploid wing disc cells (Figures 1C and S1A–1C). In contrast, the TZ cells located at the posterior boundary of the ImR showed polyploidy, with three major peaks at 4C, 8C, and 16C of DNA content (Figures 1C and S1D). These findings suggest that the TZ cells are polyploid cells, different from their diploid anterior ImR neighbors and the polytene salivary gland cells (Yang and Deng, 2018).

To determine whether the TZ cells become polyploid through endoreplication, we analyzed the expression of the cell cycle markers. These cells exhibited sporadic incorporation of 5-bromo-2´-deoxyuridine (BrdU), a DNA synthesis (S) phase maker, but absence of mitotic (M) phase marker phospho-Histone H3 at Ser10 (pH3) staining (Figure 1D), suggesting that the TZ cells undergo endoreplication to become polyploid. Consistently, the Fly Fluorescent Ubiquitination-based Cell Cycle Indicator (Fly-FUCCI, (Zielke et al., 2014)) analysis showed that the TZ cells were positive of GFP-E2F1 but not CycB-RFP, indicating they are either in the gap (G) or S phase, but not in M phase (Figures 1E and S1E). Furthermore, we applied a lineage tracing technique, the Gal4 Technique for Real-time and Clonal Expression (G-TRACE) to determine whether the TZ cells divide during normal larval development (Evans et al., 2009). In the G-TRACE system, the Gal4 real-time expressing cells are marked with RFP and the lineage cells are labelled with GFP. Indeed, the Mmp1-Gal4 expressing TZ cells did not produce new cells during larval development (Figure S1F). Consequently, the number of TZ cells remained about the same during the third larval instar, as measured from 72 hours to 120 hours after egg laying (AEL) at 25°C (Figure S1G).

Taken together, these findings suggest that the TZ cells undertake two to three rounds of endoreplication to become polyploid, different from their anterior neighbors that undergo rapid mitosis during the third larval instar (Yang and Deng, 2018).

Upregulated Notch signaling drives polyploid transition-zone cells to re-enter mitosis

Endocycling cells bypass M phase and do not divide. However, the polyploid TZ cells apparently increased their cell number upon continued Notch activation (Figure 2A). When Mmp1-Gal4 was used to drive the expression of a constitutively active form of Notch (the Notch intracellular domain, NICD) at the TZ, neoplastic transformation was detected (Figure 2A), similar to the previously reported Act^ts>NICD tumors (Yang et al., 2019). Two days after NICD induction (a shift from 18°C to 29°C), the number of GFP-positive TZ cells (27.3 cells/ImR, n=15) was about three times as many as the control TZ cells (8.3 cells/ImR, n=22) (Figures 2A and S2A). Similar to that of Act^ts>NICD (Yang et al., 2019), NICD overexpression driven by Mmp1^ts-Gal4 resulted in a prolonged larval state and failed pupariation (43.4% of the larvae survived to 21 days at 29°C, n=129; Figure S2B), allowing examination of continued TZ cell proliferation. After five days of NICD overexpression, the average GFP-positive cell number in Mmp1^ts>GFP+NICD ImRs increased to about six times of the control (48.76 cells/ImR, n=29; vs 8.3 cells/ImR in controls; Figures 2A and S2A). The resultant cell mass displayed a tumorous phenotype with multilayering and disruption of epithelial polarity, as revealed by staining with DAPI, phalloidin to detect F-actin, and cell polarity markers Discs large (Dlg) and Dystroglycan (Dg) (Bergstralh et al., 2013; Deng et al., 2003) (Figures 2A, 2B). Dlg, a septate junction protein localized normally at the lateral membrane toward the apical side, was detected at the basal surface after NICD induction, whereas the basally localized ECM receptor Dg showed up at the apical and lateral surfaces in TZ cells upon NICD induction, indicating loss of apical-basal polarity (Figure 2B). These Mmp1^ts>GFP+NICD TZ tumors kept growing following transplantation into the host abdomen (Figure 2C), further suggesting the malignant transformation of TZ cells upon continued Notch activation.

Figure 2. Overexpression of NICD Notch in TZ cells promotes TZ cell proliferation.

(A) Mmp1^ts>NICD ImRs harvested at different time points after NICD induction were stained with Phalloidin (Pha) (red, upper panel; or white, lower panel), and DAPI (blue). Scale bar, 20 μm. (B) Epithelial polarity analysis of control (Mmp1^ts-GAL4) and tumor-bearing (Mmp1^ts>mCD8GFP+NICD) ImRs at different time points after NICD induction. The ImRs were stained with Dlg (red), Dg (green or white) and DAPI (blue). (C) Comparison of the Mmp1^ts>NICD tumor before transplantation (BTP, 7 days after NICD induction, left panel) and after transplantation (ATP, 16 days after transplantation, right panel). White, DAPI; green, GFP. Scale bar, 20 μm.

To determine whether the polyploid TZ cells enter mitosis to increase their number, we examined the expression of mitotic cyclins in the ImRs before and after NICD induction, and found that Cyclins A (CycA) or B (CycB) was detected in the tumor but not the wildtype TZ cells (Figure 3A). Using Fly-FUCCI, we identified cells at different mitotic cycle phases within the TZ tumors (Figures 3A and S2C). In addition, staining the tumors with an antibody against M-phase marker, pH3, showed that the pH3-positive cells displaying a range of different sizes of DAPI-stained nuclei. A small fraction of pH3-positive cells had a DNA content that was higher than 16C (Figure S2D). pH3 and DAPI co-staining also revealed the dividing polyploid cells at different phases of mitosis (prophase, metaphase, anaphase, and telophase) (Figure 3B). Furthermore, karyotyping analysis showed that the tumors contained polyploid mitotic cells with 4N, 8N and 16N metaphase chromosomes, respectively (Figure 3C). Together, these findings collectively suggest that the polyploid TZ cells re-enter mitosis upon Notch-induced neoplastic transformation to increase their cell number.

Figure 3. Polyploid transition-zone cells re-enter mitosis during Notch induced tumorigenesis.

(A) The expression of mitotic cyclins CycA or CycB upon NICD induction in the ImR. Upper panel, control TZ cells (arrows) fail to show CycA (red or white, anti-CycA staining, upper left) or CycB expression (red, CycB-RFP.nls, upper right). Lower panel: the TZ tumor cells show both CycA (red or white, anti-CycA staining, lower left) or CycB expression (red, CycB-RFP.nls, lower right). Fly-FUCCI (E2f1-GFP, CycB-RFP.nls) was used to show control and tumor cells at different cell-cycle stages (141 cells from 21 control ImRs; and 2276 cells from 18 ImR tumors). Scale bar, 20 μm. (B) Dividing polyploid tumor cells (yellow dotted lines) show different mitotic phases (prophase, metaphase, anaphase, and telophase). Red, anti-pH3; blue, DAPI. Scale bar, 10 μm. (C) Karyotype analysis of DAPI-stained TZ tumor cells with 2N, 4N, 8N and 16N ploidy, respectively. In total, we identified 65 nuclei with 2N, 32 nuclei with 4N, 11 nuclei with 8N, 6 nuclei with 16N, and 2 nuclei with 32N (total of 117 metaphase cells). The higher ploidies (8N, 16N and 32N) were estimated based on the number of the 4^th chromosome. (D) Analysis of stg in NICD induced ImR tumors. Left graph: stg transcription is upregulated in ImR tumors. Graph shows the relative log2 fold change of Notch (N) and stg mRNAs in the tumor sample compared with control. Right panel: stg overexpression induces mitosis entry of TZ cells. The salivary gland ImRs with Act^ts>GFP or Act^ts>stg stained with DAPI and pH3. Three days after Stg overexpression, the TZ cells (arrows) show positive Ph3 staining (red or white in right panels). Yellow outline marks the overproliferation of TZ cells after Stg overexpression. The mitotic index was analyzed from 689 TZ cells in Act^ts>stg ImRs (n=12). (E) Knockdown of stg reduced the size of NICD-induced TZ tumors. Left, Mmp1^ts>stg shows slightly enlarged TZ. Middle, Mmp1^ts>NICD shows a TZ tumor. Right, Mmp1^ts>NICD+stgRNAi shows a reduced tumor, but the TZ cells contain enlarged nuclei. Blue, DAPI; red Phalloidin (Pha); green, GFP. Scale bar, 20 μm.

To understand how the polyploid TZ cells re-enter mitosis after NICD induction, we performed RNA-seq analyses to compare the transcriptomes of the 7-day-old TZ tumor (Act^ts>NICD) and the control ImR cells (Act^ts>GFP, L3 ImR). We found that the Drosophila Cdc25 homolog String (Stg), a M-phase inducer phosphatase (Edgar and O’Farrell, 1990), was significantly upregulated in the tumor sample (Figure 3D). Knockdown of stg in NICD induced TZ tumors decreased the tumor size significantly, suggesting Stg is a potential target of Notch in the TZ (Figures 3E, S2E and S2F). In Mmp1^ts>NICD tumors with Stg knockdown, the TZ cells appeared to have enlarged nuclei (Figure 3E), suggesting that stg knockdown blocks polyploid mitosis but not endoreplication in these TZ cells. To determine whether induced Stg upregulation is sufficient to drive mitosis re-entry of the polyploid TZ cells, we overexpressed Stg and found TZ cells started mitotic divisions after Stg overexpression, forming a small cell mass (Figures 3D, 3E, S2E and S2F). The larvae bearing Act^ts>stg ImRs, however, proceeded into metamorphosis during the normal developmental window, unlike the larvae with NICD overexpression which had extended larval stage to allow sufficient time for tumor formation. Together, these results suggest that stg is an important player in mitosis re-entry following Notch upregulation.

Continued endoreplication contributes to tumor growth

Interestingly, the Mmp1^ts>NICD tumor cells exhibit apparent heterogeneity in nuclear sizes, many of which are larger than those of the anterior ImR cells (Figure 2A). Reexamination of the Act^ts>NICD tumors also revealed heterogeneity in tumor-cell nuclear sizes (Figure S2D), suggesting that these tumor cells may have different ploidy levels. From the DAPI assisted ploidy measurement, we noticed that nearly 40% of tumor cells from 7-day-old tumors had ploidy greater than the maximum 16C detected in the original TZ cells (39.6%, n=2000), suggesting that these tumor cells have undergone additional rounds of endoreplication (Figures 4A and 4E). Indeed, the tumor sizes and the average tumor cell ploidy were reduced significantly when endoreplication regulators DNA replication-related element binding factor (Dref), Fizzy-related (Fzr), or Cyclin E (CycE) (Hirose et al., 1999; Sauer et al., 1995; Sigrist and Lehner, 1997) was knocked down individually in these NICD-driven tumors (Figures 4A–4E and S3A). Knockdown of Dref, Fzr, and CycE, decreased the sizes of 7-day-old tumors to about 15.3% (n=25), 39.2 % (n = 24), and 36.0 % (n=25) of the tumors with NICD-overexpression alone (n=22), respectively (Figures 4A–4D and 4F). As a control, knockdown of Dref, Fzr, or CycE individually with RNAi expression in wildtype ImRs, the TZ cells showed reduced ploidy as well (Figure S3B). In contrast, the cell number at the anterior ImR region did not show significant difference when Dref, Fzr, or CycE was knocked down in Act^ts>NICD larvae (Figure 4G). Taken together, these results suggest that endoreplication and the resultant ploidy increase are necessary for the volume increases of the NICD induced TZ tumors.

Figure 4. Endoreplication and ploidy analyses of the transition-zone tumors.

(A-D) Increased polyploidy in TZ tumor cells. Confocal images from (A) NICD, (B) NICD+Dref^RNAi, (C) NICD+Fzr^RNAi, and (D) NICD+CycE^RNAi were stained with DAPI (white). Orange dotted lines mark the tumor region in respective genotypes. (E), graphs show DAPI-assisted ploidy analysis of Act^ts>NICD (black line) and Act^ts>NICD+Dref^RNAi (red line) tumor cells. Scale bar, 20 μm. (F) Graph shows the tumor size in genotypes shown in A-D. Mean ± S.E.M.; One-way ANOVA Dunnett’s multiple comparisons test. Data from Act^ts>NICD as control. ***P < 0.001. (G) Graph shows the numbers of anterior cells from the ImRs with genotypes shown in A-D. Mean ± S.E.M.; One-way ANOVA Dunnett’s multiple comparisons test. Data from Act^ts>NICD as control.

Ploidy-reduction divisions in transition-zone tumor cells

Interestingly, measuring the DNA content in tumor cells at different intervals after NICD-induction revealed a change of ploidy level as tumors increased their volume (Figure 5A). As a control, the 2N anterior ImR cells had mean DAPI Integrated Density (IntDen) of 154.4 (n=40, Figure 5A). In contrast, the mean DAPI intensity of TZ cells 2 days after NICD induction was 488.8 (n=209), significantly higher than that of the control TZ cells (mean IntDen=368.8, n=39). However, the mean cellular DAPI intensity of 5-day (mean IntDen=342.6, n=154) and 12-day (mean IntDen=345.4, n=175) old tumor cells were lower than that of the wild-type TZ cells (Figure 5A), suggesting that polyploid TZ cells increase their ploidy level initially when NICD is upregulated, but a portion of the polyploid cells may have undergone a ploidy-reduction (i.e., depolyploidization) process as tumors develop.

Figure 5. Ploidy reduction of the ImR tumors.

(A) DAPI-assisted DNA intensity measurement of cells at the anterior (Ant.) or posterior (Pos.) regions of Mmp1^ts>GFP salivary gland ImRs, as well as TZ tumor cells from different intervals after NICD overexpression (Mmp1^ts>GFP+NICD). Two sample, unpaired two-tailed t-test. p = 0.0174. (B) FACS analysis of DNA content from G0 (primary), G4 and G11 (the 4th- and 11th-generation replanted) tumors (Act^ts>NICD).

To investigate the long-term trend of ploidy variation in these tumor cells, we examined the DNA content of transplanted tumors. The transplanted NICD TZ tumors can grow in the abdomen of host flies for up to three weeks ((Yang et al., 2019), Figure 2B) and can be further segmented into small pieces and replanted for more generations. Fluorescence-activated cell sorting (FACS) analysis of DNA intensity revealed that the replanted Act^ts-NICD tumors maintained a population of polyploid cells, but the average ploidy of tumor cells in replanted tumors at the 4th and 11th generation of transplantation (named as G4 and G11 tumors, respectively) appeared to be lower than that of the primary (referred as G0) tumors (G4=7.2C; G11=7.1C; G0=12.4C), though the overall ploidy distribution between G4 and G11 tumors was similar (Figure 5B). The percentages of large polyploid cells (>16C) in these replanted tumors (G4: 4.8%; G11: 5.9%) were significantly lower than those in the primary tumors (G0, 15.6%) (Figure 5B). In contrast, the proportion of tumor cells with lower ploidy (≤2C) was higher in replanted Act^ts-NICD tumors than in their primary counterparts (Figure 5B).

Together, the comparison of DNA content among the control TZ, the primary, and transplanted tumor cells further suggests that the polyploid tumor cells could depolyploidize to reduce their ploidy levels and that the replanted tumors maintain a pool of polyploid cells after many generations of transplantation.

Error-prone polyploid cell divisions in the transition zone tumor

Polyploid cell division has been reported to be error-prone and could cause chromosome instability (CIN) (Fox and Duronio, 2013; Fox et al., 2010). To determine whether aberrant cell divisions are associated with the dividing polyploid cells, we applied pH3 and DAPI staining and a fluorescence-tagged centrosome marker, Centrosomin (Cnn)-GFP (Zhang and Megraw, 2007), to examine the segregation of chromosomes in both the primary and transplanted tumors. We found that a majority of M-phase polyploid cells contained supernumerary (>2) Cnn-GFP foci (75.6%; n=114; Figure 6A). Using the anti-α-tubulin antibody to mark spindle we found that many cells showed multipolar spindles (65.6%; n=32) (Figure 6B). In contrast, the mitotically dividing diploid anterior ImR cells, which are non-tumorigenic and thereby serve as an internal control (Yang et al., 2019), only had two Cnn-GFP foci per cell at opposite poles of the spindle (Figure S4A). DAPI staining revealed frequent occurrences of chromosomal bridges (17 out of 62 dividing cells) and unequal separation of chromatin (21 out of 62 dividing cells) in the dividing polyploid cells (Figure 6C).

Figure 6. Error-prone polyploid cell divisions in transition-zone tumors.

(A) Supernumerary centrosomes (green, Cnn-GFP) are present in mitotic polyploid tumor cells (red, Phalloidin; blue or white, DAPI). Scale bar, 10 μm. (B) Abnormal spindle formation in dividing tumor cells. α-tubulin (red) and Cnn-GFP (green) indicate some mitotic spindles are multi-polar (right panels) (blue, DAPI). Scale bar, 10 μm. (C) Aberrant cell divisions. Left panel, DAPI staining shows a cell nucleus is cleaved unequally into two different-sized nuclei. Right panel, DAPI staining shows a chromosome bridge between two recently separated nuclei. Yellow arrowheads indicate the chromosome bridge. Scale bar, 10 μm. (D) Snapshots of a live imaging video of a dividing polyploid tumor cell (white dotted line) showing a lagging chromosome (green arrowheads) (red, Hoechst to label the DNA). Scale bar, 10 μm. (E) Snapshots of a live imaging video of Act^ts>NICD; Jupiter-GFP, Histone-mRFP tumors. At 210 minutes, a chromosome bridge (yellow arrowhead) was seen to link two separate nuclei. The panel shows the tumor cell underwent two consecutive divisions resulting in smaller nuclei (white arrows). Scale bar, 10 μm. (F) Primary (G0) and replanted (G13) tumors stained with γH2Av (red, or white) and DAPI (blue). Scale bar, 20 μm. (G) Single-cell DNA copy number variation (scCNV) analysis of the G0 (primary) or 9th generation replanted (G9) tumors. DNA copy numbers are color-coded to range from blue (lower copy number), through orange, to black (higher copy number). Groups i and ii were groups selected to show polyaneuploidy and high CNVs, respectively.

To detect polyploid tumor cell division directly, we performed live imaging analysis of 7-day-old primary tumors and found many aberrant division patterns, including asymmetric separation of the nuclear DNA, chromosomal bridges, and lagging chromosome (Figure 6D; Videos S1–S6). We noticed a polyploid tumor cell underwent two rounds of consecutive divisions within 9 hours (Figure 6E; Video S2) and formed smaller cells with apparently decreased ploidy. The division of the polyploid tumor cell appeared to be unequal that two daughter cells received different amount of chromosomal materials (Videos S3 and S4). In another, we found that the nucleus of a polyploid tumor cell appeared to have separated into three or more blocks at the same focal plan (Videos S3–S6). These observations suggest that polyploid tumor cells undergo aberrant cell divisions that may have led to depolyploidization, decreasing the cell ploidy and causing CINs.

To determine whether the aberrant polyploid cell divisions are associated with increased DNA damage, we applied an antibody against γ-H2Av (Lukas et al., 2011), a marker for DNA double-strand breaks (DSBs) to stain the primary and transplanted tumors. γ-H2Av signal was detected in both the primary and transplanted tumors. The γ-H2Av-positive cells appeared to be more numerous in tumors after multiple generations of transplantation. Among the primary tumors (n=13), we found about 12.1% tumor cells (n=1271) showed γ-H2Av staining, but in G13 replanted tumors (n=8), the γ-H2Av positive cells increased to 88.5% (n=1683), indicating increased DSBs in more advanced tumors (Figure 6F).

Increased chromosomal instabilities in replanted transition-zone tumors

To determine the extent of genome instability in replanted NICD tumors, we applied the single-cell DNA copy number variation (scCNV) profiling technique (Figure 6G), which allowed us to examine genomic CNV events at 2-Mb resolution at the single-cell level. We found that the both primary (G0) and replanted 9th-generation (G9) tumors had cells with a range of different ploidy levels (Figure 6G), consistent with the FACS analysis results (Figure 5B). In G9 tumors, we identified groups of polyploid cells with different numbers of chromosomes 2 and 3 (group i, n = 780, in total 4613 cells; Figure 6G), indicating polyaneuploidy (Pienta et al., 2021). In another group (group ii, n=410, Figure 6G), we found cells in which the left arm of chromosome 2 had disproportionately more copies than the right arm. Overall, CNVs are apparent across the genome but appeared to be more prevalent in chromosome 2. CNVs were also detected in the primary G0 tumors but to a less degree when compared with the G9 tumors. As the G0 tumors contained a significant number of the anterior ImR cells that were inseparable from the tumor cells during sample preparation, a large proportion of diploid cells with low CNVs were detected (Figure 6G). These data suggest that replanted NICD tumors display a high degree of genomic instabilities and that CINs and aneuploidy are widespread occurrences.

DNA damage response and repair genes promote depolyploidization and tumor progression

To determine the molecular mechanisms underlying increased genome-instability in NICD-induced TZ tumors, we compared the transcriptomes between tumor and control ImR cells, and found significant upregulation of genes involved in DNA damage response and repair (Figure 7A; Table S1). These include Replication Protein A 70 (RpA-70), a gene involved in DNA replication and the cellular response to DNA damage (Marton et al., 1994; Mitsis et al., 1993); meiotic recombination 11 (Mre11), which is involved in DNA damage checkpoint and telomere capping at the mitotic G2 phase (Ciapponi et al., 2004; Kondo and Perrimon, 2011); Structural maintenance of chromosomes 5 (SMC5), which is required for homologous DNA recombination (Chiolo et al., 2011; De Piccoli et al., 2006); meiotic 41 (mei-41), involved in DNA damage-dependent responses and chromosomal segregation during female meiosis (Banga et al., 1986; Brady et al., 2018); and many others (Table S1).

Figure 7. The involvement of DNA damage response and repair genes in transition-zone tumor progression.

(A) Heat map of putatively selected genes that show upregulation in TZ tumors from three independent replicates. Gene expression levels are color-coded to range from blue (lower), through yellow (medium), to red (higher expression). All samples are from salivary gland ImRs. Control, 3^rd instar Act^ts>GFP. Tumor, Act^ts>GFP+NICD (7-day old). (B) Genetic epistasis analysis of DNA damage response and repair genes in ImR TZ tumors. The genotype of each sample is indicated in the panel. Red, Phalloidin (Pha), Green, or white, DAPI. Arrowheads indicate large polyploid cells. Scale bar, 20 μm. Graph shows the tumor size from the indicated genotypes. Mean ± S.E.M.; One-way ANOVA Dunnett’s multiple comparisons test. Data from Act^ts>NICD as control. ***P < 0.001. (C) Transplanted tumors with mre11(middle panel) or SMC5 (right panel) knockdown, harvested 14 days after transplantation, showed giant polyploid nuclei (arrows). White, DAPI. Scale bar, 20 μm. Graph shows the relative tumor size at 14 days after transplantation. Mean ± S.E.M.; One-way ANOVA Dunnett’s multiple comparisons test. Data from Act^ts>NICD as control. ***P < 0.001. (D) A diagram showing a combination of endoreplication and error-prone polyploid mitosis and ploidy-reduction (depolyploidization) divisions in NICD induced TZ tumors.

To determine whether the upregulation of these genes is functionally involved in tumor growth, we performed genetic epistasis experiments to knock down these genes in NICD-expressing TZ tumors. Individual depletion of these genes alone in salivary gland ImRs revealed no obvious phenotype (Figure S5) but knockdown of RpA-70, Mre11, mei-41, or SMC5 in NICD-tumors all resulted in smaller tumor sizes than with the NICD-overexpression alone (Figure 7B). Surprisingly, we found that these tumors, albeit smaller, frequently contained cells with significantly increased nuclear size based on DAPI staining (i.e., RpA-70, 28.6%, n=21; mre11, 41.4%, n=29; mei-41, 36.7%, n=30; and SMC5, 30.8%, n=26) (Figure 7B). These findings suggest the DNA damage response genes are involved in the depolyploidization of tumor cells. To test this possibility further, we transplanted the TZ tumors with mre11 or SMC5 knockdown into the adult host abdomen, respectively, to detect the ploidy level after extended periods of tumor growth. Consistently, the transplanted tumors with mre11 or SMC5 knockdown showed slower growth and smaller overall tumor sizes than did the control, and they contained polyploid giant cells that normally are not observed in transplanted tumors (Figure 7C), confirming the involvement of DNA damage response and repair genes in the ploidy reduction division and tumor progression.

Discussion

In this study, we show that tumorigenesis in salivary gland ImRs is induced in the naturally occurring polyploid cells, which are normally considered a developmental dead end. The re-entry of mitosis and continued division of these polyploid TZ cells in response to elevated Notch signaling are key to tumor initiation and progression. The tumor cells display a high degree of variations in ploidy levels and show marked increase in CINs and aneuploidy as the tumors continue to grow following transplantation. The dynamics and heterogeneity of ploidy levels are caused by a mixture of endoreplication, polyploid mitosis, and activation of a ploidy reduction program in polyploid tumor cells. Both polyploid cell mitosis and depolyploidization are error-prone and can increase genome instability during tumor progression (Figure 7D). These tumors, after repeated transplantation, also possess a large proportion of polyaneuploid cells, aneuploid cells with more than two copies of genomic content (Figure 6G). Polyaneuploid cancer cells (PACCs), commonly observed in cancers, have been suggested as a critical player in response to therapeutic stress and are related to cancer lethality (Pienta et al., 2020; Pienta et al., 2021). The aberrant divisions by PACCs could be a major source of increased CINs and DNA damages in replanted tumors (Figure 6).

Polyploidy and tumorigenesis

A few examples suggest that polyploidy play important roles during the initial stages in tumorigenesis. In Barrett’s esophagus, a premalignant condition, there is a significant amount of tetraploid cells associated with an early loss of P53 (Galipeau et al., 1996). In mouse liver, loss of Hippo signaling promotes hepatocyte polyploidy through the Akt-Skp2 axis. Combined loss of Hippo signaling and p53 leads to greatly increased polyploidy and higher incidence and early onset of liver tumors. The division of these polyploid cells causes oncogenesis and genomic instability (Zhang et al., 2017). A recent study reported that the polyploid liver cells undergo ploidy reduction divisions that are critical during early stages of liver tumorigenesis (Matsumoto et al., 2021). In epithelial ovarian tumors, the observation of mononucleated or multinucleated giant cancer cells following chemotherapy in high grade ovarian serous carcinoma and their subsequent division to produce more cancer cells have prompted the hypothesis of polyploid giant cancer cells (PGCCs) in cancer initiation (Bharadwaj and Yu, 2004; Zhang et al., 2014a; Zhang et al., 2014b).

Several Drosophila tumor models indicate that polyploidization is an important step in tumor formation in genetically induced tumors. In the wing imaginal disc, co-expression of EGFR and a conserved microRNA miR-8 causes a failure in cytokinesis to produce a polyploid intermediate and then neoplastic transformation. These polyploid cells gain competitive advantages to outcompete neighboring cells during tumorigenesis (Eichenlaub et al., 2016). Also, in imaginal discs, mutation of endocytic pathway genes rab5, vps25, erupted, or avalanche results in the activation of JNK and Yorkie to downregulate CycB to induce endoreplication and generation of polyploid giant cells, which is followed by tumorigenesis. In the same study, they found imaginal disc tumors induced by a combination of Ras expression and cell polarity mutations also consist of polyploid cells. Blocking endoreplication in these tumors suppresses tumor growth and metastasis (Cong et al., 2018).

A crucial step for endocycling polyploid cells to become tumorigenic is the re-entry into the mitotic cycle. NICD induced mitosis re-entry in the salivary gland TZ is through upregulation of the Cdc25 homolog, Stg. Knockdown of Stg expression in the TZ tumor prevented tumor from growing (Figure 3D). Although Stg overexpression also induced mitosis re-entry in TZ cells (Figures 3D and 3E), these cells nonetheless failed to form tumors, as their transplantation into host abdomen did not show continued growth, suggesting re-entry into mitosis by polyploid cells does not necessarily lead to tumor formation. In Drosophila rectum, the polyploid cells re-enter mitosis during normal development, although those cells frequently exhibit extended anaphases, chromosome bridges, and lagging chromosomes, tumors do not form following the polyploid mitosis (Fox et al., 2010). To induce the polyploid cells to undergo tumorigenesis, Notch activation may have triggered other changes in the TZ cells, eg., the disruption of apical-basal polarity (Figure 2B).

Polyploidy and genome instabilities

Polyploid cells can be a source of CINs and aneuploidy during tumorigenesis (Coward and Harding, 2014; Davoli and de Lange, 2011; Fox and Duronio, 2013; Storchova and Pellman, 2004). Following anticancer drug treatment, it has been reported that the polyploid giant cells among the MCF7 (p53 wild-type) breast carcinoma cells increased significantly, these cells show increased metastasis and cancer recurrence (Mirzayans et al., 2018). The danger of Polyploidy lies in the inaccuracy when polyploid cells divide, even in developmentally programmed polyploid division. It has been reported that mitosis re-entry in polyploid cells is error prone due to lacking precise checkpoints during cell division (Chen et al., 2016; Duncan et al., 2010; Fox et al., 2010; Hassel et al., 2014). In the TZ tumor model, polyploid cell division is frequently associated with lagging chromosomes, chromosome bridges, and uneven dividing, whereas increased DSBs and CNVs were found in replanted tumors. With the increased DNA repair genes expressed in the tumor cells, it will be interesting to determine how these DSBs are fixed.

A possible problem for polyploid cell division is the formation of polytene chromosomes. Polytene chromosomes have been observed in human diseases such as spontaneous abortions, muscular dystrophy, and many types of cancer (Stormo and Fox, 2017). Our karyotyping analysis of the ImR TZ tumors suggests that some polyploid tumor cells contain polytene or partial-polytene chromosomes, particularly in those with higher ploidy (eg., 16 N, Figure 3C). The polytene cells can be a source of CINs and aneuploidy during tumorigenesis since chromosome segregation is a challenge for cells with multiple chromosome sets that are clumped together (Figures 3C and S3G).

The error-prone feature of polyploid divisions might be related to the increase of centrosome numbers in polyploid cells (Storchova and Pellman, 2004). Centrosome amplification has been commonly observed in tumors, and it contributes to tumor initiation and progression (LoMastro and Holland, 2019). Studies in Drosophila and cell lines also showed that extra centrosomes can cause CINs by promoting chromosome missegregation, which gives rise to cells with aneuploidy (Ganem et al., 2009; Hassel et al., 2014; Schoenfelder et al., 2014). Also, cytokinesis failure can cause these cells to acquire extra centrosomes, those cell with extra centrosomes can further lead to tumorigenesis together with Yki and JNK (Gerlach et al., 2018). In abdominal epidermal cells in Drosophila pupae, Myb has been shown to be involved in suppressing the formation of supernumerary centrosomes to maintain genome instability (Fung et al., 2002). In the study reported here, the TZ tumor cells frequently show supernumerary centrosomes and exhibit mitotic aberrations that include asymmetric division, multipolar division, which has been considered as a source of aneuploidy and CINs. Also, live imaging analysis showed that polyploid cells can be separated into three or more blocks in one division (Videos S2 and S4). This is probably due to the formation of multipolar spindles that pull chromosomes in different directions, causing unequal and inaccurate separation of sister chromatids and homologous chromosomes. Indeed, our results showed that ImR tumor cells show high ploidy variations in replanted tumors (Figure 6). Similarly, centrosome dysfunction is associated with increased polyploidy and aneuploidy in transplanted Drosophila larval brain tumors caused by mutations of polo, aurA and dsas-4 in neural stem cells (Castellanos et al., 2008). In Drosophila larval brain, cells with extra centrosomes could generate metastatic tumors (Basto et al., 2008).

A noticeable phenotype is the vast amount of DNA DSBs in the tumor cells, which show continued increase in transplanted tumors (Figure 6F). It has been reported that polyploid cells are resistant to cell death (Hassel et al., 2014; Mehrotra et al., 2008), we observed very little cell death in these tumors despite of massive amount of DSBs (Figure S4B). These findings are consistent with the report that polyploid cells in the adult brain are more resistant to DNA damage induced cell death when compared with their diploid counterparts (Nandakumar et al., 2020). In papillar cells, polyploid cells depends on Fanconi anemia proteins to segregate damaged chromosomes (Bretscher and Fox, 2016). The TZ tumor cells may use a similar mechanism for segregating acentric fragments, avoiding cell death.

Is a meiotic-like program activated during depolyploidization?

The requirement of DNA damage response and repair genes for tumor cells to undergo depolyploidization suggest that this process shares similar mechanisms with meiosis, because the same genes, mre11, SMC5, mei-41, etc., are also involved in meiotic division that results in reduction of the ploidy level during gamete production. In a Drosophila brain tumor model with l(3)mbt mutation, meiotic W68, the Drosophila orthologue of SPO11 is upregulated and has a pro-tumor role by causing DNA damages in the tumor forming tissue (Janic et al., 2010; Rossi et al., 2017). Meiosis requires the generation of DSBs, which are frequently detected in the NICD-driven primary and transplanted tumors (Figure 6F). The genes involved in the recombination of homologous chromosomes and other key steps during meiosis are also upregulated in the tumor and play critical roles in ploidy regulation in the tumor cells. Therefore, triggering a meiotic-like program to regulate the ploidy level may play critical roles in the progression and evolution in our polyploid tumor model. Given the prevalent occurrences of ploidy changes in human solid tumors (Priestley et al., 2019), a similar meiotic-like program activation is plausible. The depolyploidization process may increase the tumor-cell heterogeneity, just as the recombination of homologous chromosomes during meiosis is essential for the increase in genetic heterogeneity among the gametes. Furthermore, meiosis is involved in the production of multipotent cells and resets the epigenome to support the development of the next generation (Villeneuve and Hillers, 2001). The involvement of a meiosis-like program during depolyploidization is probably also key to the maintenance of the immortality of these tumor cells.

Irrespective of the origin of polyploid tumor cells, the combination of different cell-cycle programs (endoreplication, mitosis, meiosis-like depolyploidization, etc.) undertaken by these cells promotes CNV, chromosomal instability, and aneuploidy, which are prominent features of mammalian cancer cells (Figure 7D). These cell-cycle variations could contribute to intratumor heterogeneity, complexity, and cancer evolution. Since polyploid cells have been suggested to be more resistant to drugs and cell death (Amend et al., 2019; Niu et al., 2017; Zhang et al., 2014b), and their entry into error-prone mitosis or meiosis-like depolyploidization could produce cells with increased genetic diversity, the maintenance of a mixture of cells with different states of ploidy and different cell cycle programs is likely beneficial for continued survival and evolution of cancer cells.

Limitations of Study

Although we show that ploidy-reduction divisions take place in the TZ tumor model and that genes active during DNA damage response and meiosis are involved in depolyploidization, the mode of ploidy-reduction divisions is still unclear because live imaging analysis captured a limited number of dividing polyploid cells. More samples for live imaging studies will help elucidate the precise pattern(s) of the ploidy-reduction divisions.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Wu-Min Deng (wdeng7@tulane.edu).

Materials availability

Drosophila lines described in this study are available from the Bloomington Drosophila Stock Center (BDSC), the Vienna Drosophila Resource Center (VDRC), or from the Lead Contact. Antibodies are available from the sources listed in the Key Resources Table.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-phospho-Histone H3 (Ser10)	Millipore	#06–570; RRID: AB_310177
Mouse anti-BrdU	BD Biosciences	#555627; RRID: AB_395993
Mouse anti-alpha-tubulin	Sigma-Aldrich	#T9026; RRID: AB_477593
Mouse anti-gamma-H2Av	DSHB	#UNC93–5.2.1; RRID: AB_2618077
Mouse anti-Hts	DSHB	#1B1
Mouse anti-Dlg	DSHB	4F3
Rabbit anti-Dg	Deng et al. 2003	N/A
Donkey secondary antibodies conjugated with Alexa 488	Invitrogen	#A32790; RRID: AB_2762833
Donkey secondary antibodies conjugated with Alexa 546	Invitrogen	#A10040; RRID: AB_2534016
Goat secondary antibodies conjugated with Alexa 633	Invitrogen	#A-21052; RRID: AB_2535719
Phalloidin conjugated with Alexa 488	Invitrogen	#A12379
Phalloidin conjugated with Alexa 546	Invitrogen	#A22283; RRID: AB_2632953
Phalloidin conjugated with Alexa 633	Invitrogen	#A22284
Chemicals, Peptides, and Recombinant Proteins
Hoechst 33342	Thermo Scientific	#62249
Fetal Bovine Serum (FBS)	Fisher Scientific	#SH3007102
Penicillin/Streptomycin	Thermo Fisher	#11548876
Critical Commercial Assays
Reagents for single-cell copy number variation (scCNV)	10X Genomics	CG000153 Rev B
Experimental Models: Organisms/Strains
D. melanogaster: w1118 strain w[1118]	Bloomington Drosophila Stock Center	BDSC#5905
D. melanogaster: UAS-NICD yw; UAS-NICD	Yang et al., 2009	N/A
D. melanogaster: temperature-sensitive ubiquitous Gal4 driver yw; Act-Gal4/CyO; Gal80ts/TM6B	Yang et al., 2009	N/A
D. melanogaster: Mmp1 Gal4 driver y[1] w[*]; Mi{Trojan-GAL4.1}Mmp1[MI03010-TG4.1]	Bloomington Drosophila Stock Center	BDSC#76161
D. melanogaster: Centrosome marker w; UAS-GFP-Cnn	Gift from Timothy Megraw (Florida State University) Zhang and Megraw, 2007	N/A
D. melanogaster: Jupiter-GFP, mRFP-histone	Gift from Timothy Megraw (Florida State University)	N/A
D. melanogaster: Fly-FUCCI w[1118]; Kr[If-1]/CyO, P{ry[+t7.2]=en1}wg[en11]; P{w[+mC]=Ubi-GFP.E2f1.1–230}5 P{w[+mC]=Ubi-mRFP1.NLS.CycB.1–266}12/TM6B, Tb[1]	Bloomington Drosophila Stock Center	BDSC#55124
D. melanogaster: Fly-FUCCI w[1118]; Kr[If-1]/CyO, P{ry[+t7.2]=en1}wg[en11]; P{w[+mC]=Ubi-GFP.E2f1.1–230}5 P{w[+mC]=Ubi-mRFP1.CycB.1–266}4/TM6B, Tb[1]	Bloomington Drosophila Stock Center	BDSC#55099
D. melanogaster: GFP overexpression y[1] w[*]; P{w[+mC]=UAS-mCD8::GFP.L}LL5, P{UAS-mCD8::GFP.L}2	Bloomington Drosophila Stock Center	BDSC#5137
D. melanogaster: Stg overexpression w[*]; P{w[+mC]=UAS-stg.HA}2	Bloomington Drosophila Stock Center	BDSC#56562
D. melanogaster: GFP overexpression w[1118]; P{w[+mC]=UAS-EGFP}5a.2	Bloomington Drosophila Stock Center	BDSC#5431
D. melanogaster: Dref-RNAi y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.JF02232}attP2	Bloomington Drosophila Stock Center	BDSC#31941
D. melanogaster: CycE-RNAi y[1] sc[*] v[1] sev[21]; P{y[+t7.7] v[+t1.8]=TRiP.HMS00060}attP2	Bloomington Drosophila Stock Center	BDSC#33654
D. melanogaster: mre-RNAi y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.HMC02995}attP2	Bloomington Drosophila Stock Center	BDSC#50628
D. melanogaster: mei-41-RNAi y[1] sc[*] v[1] sev[21]; P{y[+t7.7] v[+t1.8]=TRiP.GL00284}attP2	Bloomington Drosophila Stock Center	BDSC#35371
D. melanogaster: SMC5-RNAi y[1] sc[*] v[1] sev[21]; P{y[+t7.7] v[+t1.8]=TRiP.HMC04343}attP2	Bloomington Drosophila Stock Center	BDSC#56035
D. melanogaster: RpA-70-RNAi y[1] sc[*] v[1] sev[21]; P{y[+t7.7] v[+t1.8]=TRiP.GL00350}attP2	Bloomington Drosophila Stock Center	BDSC#35426
D. melanogaster: Fzr-RNAi	Vienna Drosophila Resource Center	VDRC#25550
Software and Algorithms
ImageJ	National Institute of Health	https://fiji.sc/
R (v.4.04)	R Core Team	https://www.r-project.org/
DESeq2	Love et al., 2014	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
RStudio (v.1.4.1106)	RStudio Team	https://www.rstudio.com/

Data and code availability

No software or custom code was generated for this study. The accession number for RNA-seq data reported in this paper is GEO: GSE175415.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Fly strains and genetics

Drosophila lines were maintained and crossed at 21–22°C on the BDSC cornmeal food (https://bdsc.indiana.edu/information/recipes/bloomfood.html), unless otherwise indicated. Control and tumor salivary-gland imaginal-ring samples harvested from both larval sexes were not differentiated in this study. For tumor transplantation studies, NICD-TZ tumors were xenografted in w⁻ female adults to avoid sex-based host differences on tumor growth. UAS-NICD, Act-GAL4, and Gal80^ts (Yang et al., 2019) and Mmp1-GAL4 ((Lee et al., 2018), BDSC#76161) were used for tumor induction in the transition zone. UAS-GFP-Cnn was used to mark centrosomes (Zhang and Megraw, 2007). The Jupiter-GFP, mRFP-histone stock was a gift from Dr. T. Megraw. Fly-FUCCI ((Zielke et al., 2014), BDSC#55124, #55099), UAS-mCD8-GFP (BDSC#5137), UAS-stg (BDSC#56562), UAS-GFP (BDSC#5431), UAS-Dref^RNAi (BDSC#31941), UAS-CycE^RNAi (BDSC#33654), UAS-mre11^RNAi (BDSC#50628), UAS-mei-41^RNAi (BDSC#35371), UAS-SMC5^RNAi (BDSC#56035), and UAS-RpA-70^RNAi (BDSC#35426) were obtained from Bloomington Drosophila Stock Center, and RNAi lines were generated from the Transgenic RNAi Project (Perkins et al., 2015). UAS-Fzr^RNAi (VDRC#25550) was obtained from the Vienna Drosophila Resource Center.

Tumor induction and temperature shift information

Gal80^ts was used to control the time for gene expression as described previously (Yang et al., 2019). Briefly, flies laid eggs at 18°C for 1 day, and eggs were cultured at 18°C for 7 days, then shifted to 29°C for additional days as indicated. Detailed genotypes were shown in the supplementary table (Table S2).

The two Gal4 lines, Act^ts-Gal4 and Mmp1^ts-Gal4 were both used to drive UAS-NICD express to induce primary tumors specifically at the TZ. No significant difference was detected in their ability in tumor induction. However, after tumor transplantation, Act^ts>NICD tumors grow faster and steadier than Mmp1^ts>NICD tumors, probably because Mmp1 expression is dependent on JNK activity (Yang et al., 2019), which may vary during tumor progression.

Transplantation assays

The transplantation procedure was performed as previously described (Gong et al., 2021; Yang et al., 2019). Salivary glands from larvae were dissected in Grace’s Medium (Sigma). After carefully removing the secretary polytene salivary-gland cells and the anterior duct cells, the remaining ImR tissue was transplanted into the abdomen of 1- or 2-day-old w¹¹¹⁸ female adult hosts by means of Nanoject II (Drummond Scientific Company). The host flies were incubated at room temperature for 1 day and then transferred to 29°C for continued incubation. The tumors were harvested from host flies after incubation for a specified period of time for further analysis. For tumor replantation, tumors grown in the host abdomen for 14 days were dissected into small pieces and injected into new w¹¹¹⁸ adult hosts.

BrdU labeling

BrdU incorporation analysis was modified from a previous study (Sun and Deng, 2005), except that tissues were incubated with BrdU for 1 hour after dissection. Briefly, samples were dissected in the Schneider Drosophila Medium (Genesee Scientific, #25–512), followed by incorporation of 0.1 mg/ml BrdU in the Schneider Drosophila Medium for 1 hour. Samples then fixed with 4% formaldehyde for 20 minutes at room temperature. After washing, samples were treated with 5 units of DNase I at 37°C for 1 hour. After DNase I treatment, samples were blocked and stained with antibody.

Live imaging

Salivary gland ImRs were dissected from larvae in the Schneider Drosophila Medium (Genesee Scientific, #25–512), then mounted with low-melting agarose (Tsao et al., 2017). During imaging, control ImR and TZ tumor samples were cultured in freshly prepared imaging medium (Schneider Drosophila Medium with 15% fly extract, 0.5% penicillin-streptomycin, and 20μg/ml insulin). Hoechst (1:1000, Thermo Fisher Scientific #33342) or mRFP-histone was used to mark the nucleus. Live images were taken with a Zeiss LSM 980 Confocal Microscope with Airyscan. Z-stack images were acquired at 2-μm intervals every 10 minutes. The imaging raw data were processed with the Airyscan processing function of the ZEN 3.0 software.

Immunohistochemistry and Confocal Imaging

Tissue samples were dissected in phosphate buffered saline (PBS), then fixed in 4% formaldehyde in PBS for 20 minutes. After washing with PBS with 0.2% Triton X-100 (PBT), the samples were incubated in PBT with primary antibodies at 4°C overnight with shaking and then washed in PBT three times for 15 minutes each. Primary antibodies were used at the following dilutions: rabbit anti-pH3 (1:200, Millipore); mouse anti-BrdU (1:50, BD Biosciences^™); mouse anti-α-tubulin (1:400, Sigma); mouse anti-γH2Av (1:200, DSHB); mouse anti-Hts (1:20, DSHB). The secondary antibodies conjugated with Alexa 546 or 633 (Invitrogen) were diluted 1:400 and incubated at room temperature for 2 hours. Nuclei were labeled with DAPI (Invitrogen, 1:1000), and F-actin with Phalloidin (1:50, Invitrogen). After washing, samples were mounted and imaged with Zeiss LSM 800 or Zeiss LSM 980 Confocal Microscope.

DAPI intensity analyses

Images were captured with Zeiss LSM 800 confocal microscopes. ImageJ software or Zen microscope software (Zeiss) was used for image analyses and processing. DNA-content analyses were modified from a previous study (Fox et al., 2010). DAPI (1:20000) was used for nuclear DNA staining. The DAPI intensity was then measured quantitatively by the function “integrated density” in ImageJ.

Fluorescence-Activated Cell-Sorting (FACS) Analysis

Analysis of cellular DNA content by flow cytometry has been described previously (Tamori and Deng, 2013). Tissue and tumor samples were dissected in Grace’s Insect Medium (Sigma). For transplanted tumors, the GFP signal was used to distinguish between the transplanted tissue and the host tissue. Only transplanted tissues were used for further steps. The cell dissociation procedure was modified from a previous study (Jevitt et al., 2020). Samples were treated with 50 U/ml papain with intermittent pipetting up and down at room temperature until tissues were fully dissociated. Cell suspensions were passed through a 40-μm filter and pelleted at 3,500 rpm for 12 minutes in an Eppendorf MiniSpin, then samples were fixed with 4% formaldehyde in PBS with Vybrant DyeCycle Violet Stain (1:500, Life Technologies) for 30 minutes and washed twice with EBSS. Cell ploidy was determined by a flow cytometer (FACSAria, Becton Dickson) based on the excitations of Vybrant DyeCycle stain at 407 nm and GFP at 488 nm. CS&T beads (BD Biosciences) and SPHERO Rainbow Fluorescent Particles (Spherotech) were used as the calibration standards.

RNA-seq and data analysis

Total RNAs were extracted from 7-day-old tumors (Act^ts>NICD) and the control ImRs (Act^ts>GFP, L3) using the Zymo RNA preparation kit. Three replicates were collected for each genotype. NEBNext Poly(A) mRNA Magnetic Isolation Module and NEBNext Ultra II RNA Library Prep Kit for Illumina were used for library preparation. The libraries were sequenced using an Illumina HiSeq 2500 system, obtaining 40 million reads for each sample. Raw reads were aligned to the Drosophila melanogaster reference genome (dos Santos et al., 2015). FeatureCounts and DESeq2 were then used to assign gene identity to genomic features, and to analyze fold change (log2(FC)) from count data (Liao et al., 2014; Love et al., 2014).

Single cell DNA copy number variation (CNV)

Cell suspensions were obtained in the same way as for the FACS sample preparation, except that they were resuspended in PBS with 0.04% BSA. After cell number estimation by hemocytometer (Hausser Scientific), the resuspended cells were used immediately for library preparation. The library was prepared by means of the Chromium Single Cell DNA Library & Gel Bead Kit according to the protocol from 10X Genomics. After sequencing, the data were analyzed with Cellranger DNA (version 1.0.0), according to the instructions from 10X Genomics. The Drosophila reference genome was prepared from the published sequencing from NCBI (Release 6 plus ISO1 MT). Results were visualized in the Loupe scDNA Browser (version 1.0.0).

QUANTIFICATION AND STATISTICAL ANALYSIS

Data analysis was conducted in the software GraphPad Prism. Unpaired t-test was used for two-sample comparisons, one-way ANOVA Dunnett’s multiple comparison test and One-way ANOVA with Tukey’s multiple comparison test were used for multiple-sample comparisons. Specific statistical approaches for each figure are indicated in the figure legend.

Supplementary Material

2

Table S1. Upregulation of DNA damage response and repair genes in NICD-TZ tumors. Related to Figure 7.

3

Table S2. Detailed genotypes for each experiment, related to STAR Methods.

4

Video S1. Live imaging from Act^ts>NICD+Jupiter-GFP ImR showed a cell with lagging chromosome during cell division, Related to Figure 6D. Yellow arrowhead indicates the lagging chromosome. Hoechst was used to label the DNA.

5

Video S2. Live imaging from Act^ts>NICD+Jupiter-GFP ImR showed ploidy reduction, Related to Figure 6E. One cell continually divides twice. Hoechst was used to label the DNA.

6

Video S3. Live imaging from Act^ts>NICD+Jupiter-GFP ImR showed uneven division and ploidy reduction, Related to Figure 6. Yellow arrows shown cells got different amount of DNA after division. White arrow indicated one nuclear separated into three blocks. Hoechst was used to label the DNA.

7

Video S4. Live imaging from Act^ts>NICD+Jupiter-GFP ImR showed uneven division (white arrow) and ploidy reduction (yellow arrow), Related to Figure 6. Hoechst was used to label the DNA.

8

Video S5. Live imaging from Act^ts>NICD+Jupiter-GFP ImR showed ploidy reduction, Related to Figure 6. Yellow arrows showed one nuclear separated into multiple blocks.

9

Video S6. Live imaging from Act^ts>NICD+Jupiter-GFP showed a cell with lagging chromosome and ploidy reduction (yellow arrow), Related to Figure 6.

Highlights.

Active Notch induces polyploid imaginal ring cells to re-enter mitosis

Polyploid mitosis, endoreplication and depolyploidization together promote tumor growth

Ploidy-reduction division depends on genes involved in DNA damage response and repair

Tumor progression correlates with increased CNVs, ploidy heterogeneity and polyaneuploidy