scRNA-seq transcriptomics reveals NICD OE-derived NPPs

To investigate the transcriptomic changes underlying the differentiation-promoting effect of NICD-OE in follicle cells into NPPs at single-cell resolution, we analyzed scRNA-seq data from wild-type (w¹¹¹⁸) and NICD-OE ovaries (Fig 1A, C). Using Seurat [17], we clustered the cells (Fig 1B, D) and assigned identities based on established cell type-specific markers for Drosophila ovarian tissue [18,2,1,19] (S1A-C Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Single-cell RNA-seq analysis of adult Drosophila ovaries in wild-type and NICD-OE backgrounds.

(A, C) Confocal images of adult Drosophila ovaries expressing the Tj-GAL4^ts driver, which is phenotypically equivalent to the wild-type (w¹¹¹⁸) and serves as the control condition for panel (C). In (A), the yellow dashed line outlines the region occupied by fully differentiated eggs marking oogenesis completion in a wild-type ovary. In contrast, (C) shows a Tj-GAL4^ts>UAS-NICD ovary, where the yellow dashed line highlights the absence of late-stage egg chambers due to oogenesis arrest at stage 9, a phenotype induced by NICD OE [16]. DAPI staining displays cell nuclei. (B, D) UMAP projections showing the distribution of ovarian cell populations in (B) wild-type flies and (D) NICD-OE flies, highlighting shifts in population distribution between the two datasets. Abbreviations: GermCs, Germarium Cells; StCs, Stalk Cells; MBFCs, Main Body Follicle Cells. (E, F) Dot plots illustrating the expression of cell type-specific markers used to assign non-epithelial-cell related clusters in both w¹¹¹⁸ and NICD-OE datasets.

https://doi.org/10.1371/journal.pgen.1011953.g001

In the wild-type dataset, 20 clusters were identified, representing cells from the 14 developmental stages of egg chambers as well as adjacent tissues, including oviduct cells, muscle sheath, hemocytes, and adipocytes (Fig 1B). In the NICD-OE dataset, 14 clusters were identified, corresponding to cells from egg chambers up to early stage 10 (Fig 1D). Cells expressing the late-stage marker Femcoat were largely absent (S1B Fig), consistent with the oogenesis arrest at stage 9 caused by NICD-OE-driven GDAC [16].

Within the NICD-OE dataset, clusters 8 and 9 attracted particular attention due to their proximity to follicle cells clusters belonging to stage 8 and 9 egg chambers (clusters 5–7) but lacked stage-specific markers such as Vitelline membrane-like (Vml), palisade (psd), and defective chorion 1 (dec-1), suggesting a distinct cellular identity (S1B Fig).

To better characterize the unique effect promoted by NICD OE to promote the differentiation of epithelial follicle cells into NPPs, we decided to identify and remove non-epithelial cell types originating from adjacent tissues in both datasets by analyzing the expression of cell-type markers, such as Insulin-like peptide 6 (Ilp6), which is associated with adipocytes and highly expressed on wild-type dataset clusters 13, (2.990313449 avg_avg_log2FC, padj = 1.34E-218), vasa (vas), cluster 14 (4.829041 avg_log2FC, padj = 1.32E-276), 15 (4.17325 avg_log2FC, padj = 0) and 16 (2.295376 avg_log2FC, padj = 1.47E-37), oskar (osk), cluster 15 (2.067233 avg_log2FC, padj = 1.25E-96), 16 (6.976927 avg_log2FC, padj = 1.59E-124), Muscle protein 20 (Mp20), cluster 17 (5.128807 avg_log2FC, padj = 1.76E-151), abdominal A (abd-A), cluster 18 (9.525009 avg_log2FC, padj = 0), Hemolectin (Hml), cluster 19 (10.63374 avg_log2FC, padj = 0) (Fig 1E, S1 Table). On NICD-OE dataset, osk, cluster 10, (6.149274165 avg_log2FC, padj = 0), vas, cluster 10, (3.408142877 avg_log2FC, padj = 0), cluster 11, (4.026276583 avg_log2FC, padj = 0), Mp20, cluster 12, (4.985966844 avg_log2FC, padj = 0), Hml, cluster 13, (12.27544819 avg_log2FC, padj = 0) (Fig 1F, S2 Table).

Following the removal of non-epithelial cell types, we integrated the NICD-OE and wild-type datasets to highlight the transcriptomic changes caused by NICD-OE. This integrated dataset revealed 13 clusters (Fig 2A), which were identified using established cell type-specific markers [18,2,1,19] (Fig 2B). To delineate the transcriptional alterations driven by NICD-OE, we separated the integrated clusters into cells originating specifically from the w¹¹¹⁸ (control) and NICD-OE (stimulated) datasets (Fig 2C). We then analyzed the proportion of cells from each dataset within each cluster.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Integration of w¹¹¹⁸ and TjGal4^ts>UAS-NICD datasets reveals unique NPP clusters derived from NICD overexpressing cells.

(A) UMAP projection of integrated follicle cells from w¹¹¹⁸ and TjGal4^ts>UAS-NICD datasets, illustrating the overall cellular landscape. (B) Dot plot displaying the expression of cell type-specific markers across integrated data set cluster numbers. (C) UMAP embedding showing cluster distribution by sample ID, with w¹¹¹⁸ cells from the wild-type dataset and NICD-OE cells from the TjGal4^ts>UAS-NICD dataset (D) Bar graph representing the relative percentage of cells in each cluster, with bars segmented by dataset of origin: w¹¹¹⁸ (salmon) and NICD OE (navy blue). (E) UMAP projection illustrating the overlap between w¹¹¹⁸ and NICD-OE cells, with clusters highlighted in orange and red, predominantly derived from the NICD-OE dataset. (F-H) Confocal images showing staining of specific NPP markers, including Drpr, puc-lacZ and Fas3 in NICD-OE egg chambers undergoing GDAC compared to those not undergoing GDAC, named as GDAC egg chambers and Non-GDAC egg chambers respectively. DAPI stains for cells nuclei and GFP marks TjGal4^ts transcriptional activation. Rainbow2 LUT (Lookup Table) visualizes intensity variations in images by mapping different pixel values to a color scale, enhancing the comparison of staining intensity between images. (I-K) Bar plots with interquartile range showing the median for the normalized intensity mean values of Drpr (I), Fas3 (J), and puc-LacZ (K) in GDAC versus Non-GDAC egg chambers. Sample sizes for GDAC are N = 75 (I), N = 58 (J), and N = 73 (K), while Non-GDAC sample sizes are N = 52 (I), N = 51 (J), and N = 52 (K). P-values obtained from Mann Whitney test are indicated by **** (p ≤ 0.0001 for I and J) and *** (p ≤ 0.001 for K) above each plot.

https://doi.org/10.1371/journal.pgen.1011953.g002

Consistent with the known biological effects of NICD OE on stopping oogenesis at stage 9 egg chambers, clusters 6, 7, and 10, which correspond to cells from egg chambers at stages 9–14 and corpus luteum cells, were predominantly represented in the w¹¹¹⁸ dataset. Interestingly, clusters 8 and 9 were predominantly unique to the NICD-OE dataset (Fig 2D, E) and displayed distinct genetic signatures that differed from other known oogenesis cell types (Fig 2B, S3 Table).

To identify the cell types represented by clusters 8 and 9, we performed differential gene expression analysis, comparing their genetic profiles to the remaining 11 clusters. This analysis revealed significant upregulation of genes related to NPP phagocytosis, including the phagocytic receptors croquemort (crq) [20] (3.571624 avg_log2FC, padj = 0) and draper (drpr) [9,15,10,11] (1.417401 avg_log2FC, padj = 3.74E-180), Drpr’s ligand, pretaporter (prtp) [21], (0.98939 avg_log2FC, padj = 5.47E-58), the GTPase Rac2 [10,15] (1.729876 avg_log2FC, padj = 0), the JNK-regulated gene puckered (puc) (0.847344 avg_log2FC, padj = 1.29E-91) [9,15,10,11] and downstream transcription factors Jun-related antigen (Jra) (0.929437 avg_log2FC, padj = 2.14E-87) and kayak (kay) (1.266146 avg_log2FC, padj = 4.99E-163) coding genes [22]. Additionally, Fasciclin 3 (Fas3), a gene previously unreported in this context, was distinctly expressed and enriched in NPPs (1.7174 avg_log2FC, padj = 0) (Fig 2B, S3 Table).

To validate these findings, we performed immunofluorescence analysis to examine the expression and fluorescence intensity patterns of Drpr, Fas3, and Puc in NICD-OE-derived NPPs compared to control follicle cells (FCs) from egg chambers not undergoing GDAC. Quantification of normalized fluorescence intensity revealed significantly higher expression of Drpr (median FC = 0.35 in FCs vs. 0.58 in NPPs, n = 52 and 76 respectively, p < 0.0001) (Fig 2F, I), Fas3 (0.47 in FCs vs. 0.71 in NPPs, n = 51 and 58, p < 0.0001) (Fig 2G, J), and puc (0.53 in FCs vs. 0.67 in NPPs, n = 52 and 73, p = 0.0005) (Fig 2H, K). These results confirm the distinct and enriched expression of Drpr, Fas3, and Puc in NICD-OE-derived NPPs. Together, these findings support the conclusion that clusters 8 and 9 represent a population of NICD-OE-induced NPPs with a unique transcriptomic identity that sets them apart from other follicle cell types.

Distinct NICD-OE clusters exhibit transcriptomic changes associated with successive stages of NPP differentiation

During germline cell death, NPPs undergo a series of gradual morphological changes, including cytoplasmic expansion and cytoskeleton remodeling that enable the digestion of germline cell debris [9,15,10,11,23]. To determine whether our analysis could capture this chronological differentiation process in NICD-OE-derived NPPs clusters, we focused on clusters 8 and 9, due to the predominance of cells derived from the NICD-OE data set. We re-clustered these cells and identified three distinct sub-clusters labeled 0, 1, and 2 (Fig 3A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. NICD-OE-derived NPP clusters exhibit distinct gene expression profiles and differentiation trajectories.

(A) UMAP projection of re-embedded Tj-Gal4^ts>UAS-NICD derived NPP clusters, highlighting their unique cellular organization. (B) RNA velocity vectors overlaid on the UMAP plot, illustrating inferred cellular lineages and differentiation trajectories. (C) Pseudotime analysis showing the progressive differentiation of Tj-Gal4^ts>UAS-NICD-derived NPPs clusters. (D) Differentially expressed gene markers identified in the re-embedded NPP clusters, highlighting key genes associated with NPP differentiation. (E-G) Differential gene expression analysis for clusters 0, 1 and 2, reflecting distinct transcriptomic profiles at successive stages of NPP differentiation. Dot plots summarizing GO enrichment analysis for cluster 0, 1 and 2, respectively. The x-axis represents the gene ratio, while the y-axis lists enriched GO Biological Process terms. Dot size corresponds to the number of genes associated with each term, and the color gradient represents the adjusted p-value (p.adjust), with darker shades indicating higher statistical significance. (H) Confocal images depict distinct morphological changes that characterize the differentiation stages of NICD-OE NPPs during GDAC. These stages align with the biological processes assigned to the differentially expressed genes predominantly expressed in each NPP cluster. Yellow arrows indicate the nuclei of dying germline cells engulfed by NPPs. Yellow dashed line displays germline area during different stages of NPP differentiation. Nuclei are stained with DAPI (blue), and the cell membrane is marked with Armadillo (red).

https://doi.org/10.1371/journal.pgen.1011953.g003

To test whether these sub-clusters reflected successive stages of NPP differentiation, we performed RNA velocity and pseudotime analyses. The RNA velocity vectors demonstrated a clear directionality, starting from cluster 0, progressing to cluster 1, and culminating in cluster 2 (Fig 3B). Pseudotime analysis further confirmed this alignment, with cluster 2 displaying the highest pseudotime values (represented in yellow), indicative of the most advanced differentiation stage (Fig 3C).

Consistent with this, cluster 2 displayed the highest expression levels for NPP markers, including the phagocytic receptors crq (2.62 avg_avg_log2FC, padj = 2.12E-257) [20] and drpr (1.63 avg_avg_log2FC, padj = 3.41E-72) [9,21,11], Drpr’s ligand, prtp, (2.22 avg_avg_log2FC, padj = 1.70E-117) [21], and the GTPase Rac2 (1.22 avg_avg_log2FC, padj = 1.17E-57) [24], supporting cluster 2, as the NPPs on their most advanced differentiation stage (Fig 3D, S4 Table).

To further investigate whether the transcriptional profiles of these subclusters align with the known morphological and functional stages of NPP maturation, we performed differential gene expression analysis followed by Gene Ontology (GO) enrichment analysis [25]. Cluster 0 was enriched for genes associated with egg coat formation (GO:0035803; p.adj = 0.00049), vitelline membrane formation (GO:0030704; p.adj = 0.00049) and vitelline membrane formation involved in chorion-containing eggshell formation (GO:0007305; p.adj = 0.00049) (Fig 3E), suggesting that these cells retain characteristics of mid-stage follicle cells and represent an early transitional state toward NPP identity.

Moreover, cluster 0 cells displayed enriched expression of genes related to mitochondrial energy metabolism: aerobic respiration (GO:0009060; p.adj = 5.61E-33), mitochondrial ATP synthesis coupled electron transport (GO:0042775; p.adj = 8.98E-29), mitochondrial electron transport, NADH to ubiquinone (GO:0006120; p.adj = 1.52E-19), mitochondrial respiratory chain complex assembly (GO:0033108; p.adj = 1.20E-14) (Fig 3E, S5 Table). We validated the increase in mitochondrial activity and energy demand in both starvation and NICD-OE NPPs models, by immunostaining with the ATP5A antibody, which targets the alpha subunit of ATP synthase (complex V), a key component of the mitochondrial oxidative phosphorylation machinery [26]. Follicle cells undergoing the transition to NPPs exhibited significantly elevated ATP5A fluorescence intensity compared to control follicle cells in both NICD-OE (S2A, S2B Fig; mean FCs = 0.37 vs. 0.71 in NPPs, n = 91 and 73, p < 0.0001) and starvation model (S2C, S2D Fig; median FCs = 0.39 vs. 0.85 in NPPs, n = 75 and 80, p < 0.0001). Taken together, the transcriptional expression profile of cluster 0 cells resembles epithelial follicle cells undergoing normal development that have begun metabolic reprogramming through enhanced mitochondrial activity, likely to meet the increased energy demands required for their emerging phagocytic function as NPPs (Fig 3H’, S3A’ Fig).

Analysis of cluster 1 revealed upregulation of genes involved in cell projection morphogenesis (GO:0048858; p.adj = 0.00079), actin cytoskeleton organization (GO:0030036; p.adj = 0.00494), cell migration (GO:0016477; p.adj = 0.00580) and growth (GO:0040007; p.adj = 0.0214) (Fig 3F, S6 Table). These biological processes are consistent with the behavior of early-stage NPPs, which exhibit cytoplasmic expansion and actively invade the germline area to engulf apoptotic debris following germline cell death [10] (Fig 3H’’, H’’’, S3A’’ Fig).

Finally, cluster 2 exhibited transcriptional enrichment in pathways related to protein degradation and nutrient stress responses, including protein catabolic process (GO:0030163, padj = 6.22E-13), autophagy (GO:0006914, padj = 5.25E-13), cellular response to starvation (GO:0009267, padj = 3.48E-06), regulation of autophagy (GO:0010506, padj = 1.00E-06), and apoptotic cell clearance (GO:0043277, padj = 1.85E-06) (Fig 3G, S7 Table). These processes reflect the metabolic shift required by differentiated NPPs to adapt to nutrient deprivation and perform the digestion of cellular debris, consistent with their fully differentiated state [27,28,29,30] (Fig 3H”, S3A”’ Fig).

Autophagy is required for NICD-OE-derived NPPs to perform germline cell debris clearance

The enrichment of autophagy-related genes in cluster 2 (Fig 3G), representing the most advanced stage of NPP differentiation, aligns with previous studies that report the activation of autophagy in mid-stage egg chambers under amino acid deprivation [27,28,29,30]. These studies state autophagy as essential for the proper clearance of germline cell debris by NPPs.

To determine whether autophagy plays a similar role in NICD-OE-derived NPPs, we first confirmed the enrichment of the autophagy reporter mCherry-Atg8a in these cells. NICD-OE-induced NPPs exhibited significantly higher mCherry-Atg8a signal compared to control follicle cells (Fig 4A-B; median FCs = 0.40 vs. 0.71 in NPPs, n = 63 and 114, respectively, p < 0.0001). This result was corroborated using Lysotracker staining, another autophagy-related marker [31], suggesting increased autophagic activity in NICD-OE-derived NPPs. (Fig 4C, D; median FCs = 0.90 vs. 1.16 in NPPs, n = 79 and 49, p < 0.0001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Autophagy is enriched in NPPs and required for effective germline debris clearance.

(A) Confocal images of Tj-Gal4^ts>UAS-GFP, UAS-NICD egg chambers undergoing GDAC and Non-GDAC conditions, showing expression levels of the autophagy reporters mCherry-Atg8a in follicle cells versus NPPs. (B) Bar plots with error bars showing normalized intensity median values for mCherry-Atg8a in GDAC versus Non-GDAC egg chambers under NICD-OE condition. Sample sizes are N = 114 and N = 63 for GDAC and Non-GDAC, respectively. P-values obtained from Mann Whitney test are indicated by **** (p < 0.0001) above the plot. (C) Confocal images of Tj-Gal4ts, UAS-GFP > UAS-NICD under Non-GDAC and GDAC conditions. LysoTracker staining highlights autophagy-related activity in follicle cells and NPPs. (D) Bar plot with error bars showing normalized Lysotracker intensity mean values in GDAC versus Non-GDAC egg chambers under NICD-OE condition. Sample sizes: NICD-OE GDAC (N = 49), Non-GDAC (N = 79). P-value from Mann-Whitney tests is indicated as **** (p < 0.0001). (E) Confocal images comparing Tj-Gal4^ts>UAS-GFP, UAS-NICD versus Tj-Gal4^ts>UAS-GFP, UAS-NICD + UAS-Atg3RNAi egg chambers. DAPI stains cell nuclei, GFP marks Tj-Gal4^ts transcriptional activation and Phalloidin (PHA), actin filaments. (F) Bar plots with error bars showing the percentage of egg chambers displaying defective phagocytosis for each genotype: NICD-OE (N = 52), NICD-OE + Atg3RNAi (N = 58). A p-value obtained from t-test is indicated by *** (p ≤ 0.001).

https://doi.org/10.1371/journal.pgen.1011953.g004

To test whether autophagy is functionally required for germline debris clearance in NICD-OE NPPs, we depleted Autophagy-related gene 3 (Atg3), which encodes a key protein required for autophagosome formation. Depletion of Atg3 led to defective phagocytosis in NICD-OE-derived NPPs. This defect was characterized by the accumulation of unprocessed, condensed germline DNA debris in the germline area. Furthermore, these cells failed to expand their cytoplasm and engulf the germline cell area, critical steps required for the complete phagocytosis of germline cell debris [15,10] (Fig 4E-F; mean NICD-OE = 0.229, Replicate 1: 0.201, N = 43, Replicate 2: 0.218, N = 55, Replicate 3: 0.262, N = 61; mean NICD OE + Atg3 RNAi = 0.951, Replicate 1: 1, N = 47, Replicate 2: 0.888, N = 72, Replicate 3: 0.966, N = 60, p ≤ 0.001).

Together, these findings reveal a transcriptional enrichment of autophagy-related genes in differentiated NPPs, supporting our proposed model (S3 Fig) and indicating that autophagy is essential for germline cell debris clearance in NICD-OE–derived NPPs. This is consistent with previous reports in starvation-induced NPPs [27,28,29,30]. The resemblance between these models further underscores their similarity and reinforces the utility of NICD OE for identifying shared mechanisms between the two systems.

SCENIC analysis identifies the JNK pathway transcription factor Jra as a regulator of NPP differentiation

After identifying NICD-OE-derived NPPs and their respective differentiation stages, we leveraged the advanced analytical capabilities of scRNA-seq, particularly Single-Cell rEgulatory Network Inference and Clustering (SCENIC), to uncover transcription factors potentially regulating NPP differentiation across different stages. Our analysis identified 11 regulons with distinct activity patterns across NPPs clusters 0, 1, and 2 (Fig 5A, S8 Table). Among these, we focused on Jra, the downstream transcription factor of the JNK signaling pathway, which has been established as essential for NPP differentiation [9,11]. However, the precise role of JNK in regulating the differentiation of follicle cells into NPPs remains poorly understood.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. SCENIC analysis reveals key transcription factors driving NPP differentiation.

(A) Heatmap of the scaled and centered activity scores of regulons in clusters 0, 1, and 2. Both clusters (columns) and regulons (rows) are hierarchically clustered, revealing the relative similarity in regulon activity across these clusters. (B) Feature Plot illustrating the gene expression of Jra across cell clusters in the integrated dataset. Cluster numbers correspond to the identities shown in Fig 2A (C) Bar plots with error bars showing the percentage of egg chambers exhibiting defective phagocytosis in each replicate (Rep.) across different genotypes: mean Jra RNAi - Starvation = 0.9881 (Replicate 1: 0.9643, N = 28, Replicate 2: 1, N = 68, Replicate 3: 1, N = 72), mean Control - Starvation = 0.073 (Replicate 1: 0.042, N = 24, Replicate 2: 0.1034, N = 29, Replicate 3: 0.0741, N = 27), mean NICD-OE + JraRNAi = 0.7470 (Replicate 1: 0.7699, N = 452, Replicate 2: 0.76, N = 550, Replicate 3: 0.7111, N = 623), mean NICD-OE = 0.1295 (Replicate 1: 0.1414, N = 99, Replicate 2: 0.1436, N = 57, Replicate 3: 0.1068, N = 103). P-values obtained from t-tests are indicated by **** (p ≤ 0.0001). (D, E) Confocal images illustrating GDAC induction by starvation stimuli (D) or NICD-OE (E), comparing the effects of Jra depletion (RNAi) under both starvation conditions (D) and NICD-OE (E). DAPI stains for cells nuclei, GFP reports TjGal4^ts driver activity and PHA displays actin filaments.

https://doi.org/10.1371/journal.pgen.1011953.g005

To explore Jra’s role, we first validated the enriched expression of Jra observed in SCENIC analysis (Fig 5A) and NPPs clusters (Fig 5B) by examining Jra:GFP expression in NICD-OE (S4A, S4B Fig; early-stage egg chambers, median = 0.53, N = 71; mid-stage egg chambers, median = 0.74, N = 71; and GDAC egg chambers, median = 1.35, N = 71) and starvation-induced NPPs (S4C, S4D Fig; early-stage egg chambers, median = 0.72, N = 44; mid-stage egg chambers, median = 0.92, N = 45; and GDAC egg chambers, median = 1.5, N = 22). As expected, in both NICD-OE and starvation models, Jra:GFP intensity was significantly higher in NPPs compared to control follicle cells from early- and mid-stage egg chambers, confirming the expression pattern suggested by scRNA-seq analysis (Fig 5B).

Next, we tested the effect of depleting Jra on NPPs using RNAi, which effectively suppressed JNK pathway activity, as indicated by reduced TRE:GFP expression (S5A, S5B Fig; median Jra RNAi = 0.66 vs. 0.95 in Control, N = 35 and 34 respectively, **** (p ≤ 0.0001)). Jra-depleted cells failed to invade germline area and expand their cytoplasm to engulf apoptotic debris, leaving uncleaned debris in both starvation and NICD-OE model systems (Fig 5C, D, E; mean Control-Starvation = 0.0731, Replicate 1: 0.042, N = 24, Replicate 2: 0.1034, N = 29, Replicate 3: 0.0741, N = 27; mean Jra RNAi = 0.9881, Replicate 1: 0.9643, N = 28, Replicate 2: 1, N = 68, Replicate 3: 1, N = 72; mean NICD OE = 0.1295, Replicate 1: 0.1414, N = 99, Replicate 2: 0.1436, N = 57, Replicate 3: 0.1068, N = 103; mean NICD OE + Jra RNAi = 0.7470, Replicate 1: 0.7699, N = 452, Replicate 2: 0.76, N = 550, Replicate 3: 0.7111, N = 623, **** (p ≤ 0.0001)). In agreement with previous studies showing that JNK depletion disrupts NPP differentiation [9,11].

In contrast, overexpression of a constitutively active form of Jra (dJun^CA) resulted in cells exhibiting a stretched morphology, with enhanced cytoplasmic expansion and invasion into germline cells (Figs 6A-F; S5C; mean dJunCA = 0.9485, Replicate 1: 0.9434, N = 53; Replicate 2: 0.9535, N = 43; Replicate 3: 0.9487, N = 39) in a similar manner to the effects previously observed with hemipterous (hep^CA) activation [9]. However, no invasive phenotype was observed in control ovarioles (Fig 6A-F; S5C Fig; N = 62, 37, and 55, respectively, ****p ≤ 0.0001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. JNK activation induces follicle cell invasion and cytoplasmic expansion.

(A-C) Confocal images comparing dJUN^CA-expressing egg chambers to control egg chambers (hsflp; TRE:GFP; act>>RFP), with staining for Armadillo (A), Dlg (B), and PHA (C). TRE:GFP-positive cells indicate JNK signaling activation, DAPI marks cell nuclei, and RFP identifies cells with an active act>>Gal4 driver. Armadillo highlights cell junctions, Dlg indicates basolateral cell polarity, and PHA visualizes actin filaments. (A) Yellow arrows highlight differences in cell membrane organization as revealed by armadillo staining. (B) Yellow and pink arrows indicate basolateral (BL) polarity in control epithelial cells, while the red dashed line marks disrupted polarity in dJUN^CA-expressing cells. (C) Yellow dashed lines delineate the germline cell area in control versus dJUN^CA-expressing egg chambers, illustrating germline area invasion by dJUN^CA-expressing follicle cells. (D, E) Orthogonal projections of confocal images showing differences in follicle cell morphology and structural organization between control (D) and dJUN^CA-expressing egg chambers (E). (F) Schematic illustration depicting the stretched, migratory morphology of dJUN^CA-expressing cells compared to wild-type epithelial cells.

https://doi.org/10.1371/journal.pgen.1011953.g006

Interestingly, immunostaining analysis using antibodies against Armadillo (Arm, cell membranes) (Fig 6A), Discs large 2 (Dlg, basolateral polarity) (Fig 6B), and phalloidin (PHA, actin filaments) (Fig 6C-E), revealed that Jra-overexpressing cells displayed enhanced rearrangements of these features compared to control follicle cells. These changes resemble key characteristics of epithelial cells transitioning to a mesenchymal migratory state [32,18,33], further reinforcing the role of JNK in regulating cytoskeleton remodeling.

Jra regulates cytoskeleton rearrangement during NPP differentiation

To gain insights into how Jra orchestrates NPP differentiation, we analyzed its regulons, focusing on predicted downstream-regulated genes and their functions. Gene Ontology (GO) analysis revealed that Jra predicted targets, enriched in NPP cluster 1 (AUC: cluster 0 = 0.0158, cluster 1 = 0.0207, cluster 2 = 0.0159; Z-scored activity: cluster 0 = – 0.60, cluster 1 = 1.15, cluster 2 = –0.56) are involved in processes such as actin cytoskeleton organization (GO:0030036; p.adj = 1.22E-06), actin filament-based process (GO:0030029; p.adj = 1.66E-07), and actin filament organization (GO:0007015; p.adj = 0.00016) (S6A, S6B Fig, S9 Table). This aligns with our previous differential gene expression analysis pointing NPP cluster 1 as being associated with genes related to cell growth, migration, and actin cytoskeleton remodeling (Fig 3F).

These processes are essential for NPPs as they expand their cytoplasm and invade the germline cell area to engulf and clear apoptotic debris. Consistently, Jra-depleted NPPs exhibited impaired cytoplasmic expansion, reduced invasion capacity, and an inability to fully clear germline cell debris (Fig 5C, D, E). Conversely, Jra-overexpressing follicle cells invaded the germline region, exhibited actin-cytoskeleton rearrangements and displayed a stretched morphology (Fig 6A-F; S5C Fig). These findings suggest that Jra’s downstream targets are involved in cytoskeletal rearrangement of follicle cell maturating into NPPs.

Functional analysis identifies the Arp2/3 complex as a key regulator of cytoskeleton remodeling in NPPs

To investigate the functional relevance of the Jra-predicted downstream targets identified by SCENIC analysis, we focused on the Arp2/3 complex given its evolutionary conserved role in cytoskeletal remodeling across species [34,35]. Notably, Arpc1, Arp2 and Arp3 are key components of this complex, and were all predicted Jra targets, each with a high normalized enrichment score (NES = 6.33; S6C Fig). To further corroborate Jra regulation of these genes, we examined publicly available Jra ChIP-seq data from wandering-stage Drosophila larvae [36]. This analysis revealed a strong Jra prominent peak at the promoter of Arp3 (S7C, S7D Fig), as well as intragenic binding sites within both Arp2 and Arp3 (S7B-D Fig). In contrast, Arpc1 did not exhibit any significant Jra binding within its promoter or gene body (S7A, S7D Fig), suggesting a more direct transcriptional regulation of Arp2 and Arp3 by Jra. Together, these findings highlight the Arp2/3 complex as a biologically relevant candidate for investigating how Jra may control cytoskeletal dynamics during epithelial-to-NPP differentiation.

We continued our analysis by examining the expression patterns of the eight genes listed in FlyBase (flybase.org) as encoding components of the Arp2/3 complex (Fig 7A, B). Most of these genes exhibited enrichment in polar cells (cluster 1) and NICD-OE-derived NPPs (clusters 8 and 9). Additionally, their expression was also elevated in stretched cells (cluster 11) and in centripetal and border cells (cluster 12), specialized epithelial cell types known to undergo extensive cytoskeletal remodeling during egg chamber morphogenesis [37,38,39,40] and, in the case of centripetal and border cells, active migration [41,42,40,43] (Supplementary table 10). We also observed that the majority of NICD-OE-derived NPPs co-expressed Arp3 and Jra in clusters 8 and 9 (Fig 7B), supporting the regulatory relationship predicted by SCENIC analysis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. JNK regulates cytoskeleton dynamics through the Arp2/3 complex.

(A) Feature Plot illustrating the gene expression of Arpc1, Arp2, Arpc2, Arp3, Arpc3A, Arpc3B, Arpc4 and Arpc5 across cell clusters in the integrated dataset. (B) Colocalization analysis showing cells co-expressing Jra (blue) and Arp3 (red), with co-expressed cells highlighted in purple, in the integrated dataset. (C) Arp3:GFP expression in egg chambers undergoing GDAC versus control egg chambers. DAPI stains cells nuclei, PHA stains actin filaments and RFP marks the expression of TjGal4^ts. Rainbow2 LUT (Lookup Table) was used to visualize staining intensity variations in images by mapping different pixel values to a color scale. (D) Representative confocal images of egg chambers with transgene expression in NPPs (green) expressing NICD-OE (Control), Arp2RNAi together with NICD-OE and Arp3RNAi combined with NICD-OE. Unengulfed germline regions are outlined with yellow dashed lines across genotypes to highlight areas not occupied by NPP cytoplasmic projections. Cells nuclei are displayed by DAPI (blue). F-actin is marked by Phalloidin (red) and TjGal4ts activity is shown by GFP. (E) Bar plot with error bars showing normalized Arp3:GFP Intensity Mean Value in NICD-OE egg chambers in different oogenesis stages, early stage egg chambers (stages 1-7), mid-stage egg chambers (stages 8-9), GDAC egg chambers. Sample sizes for early-stage egg chambers is N = 63, mid-stage egg chambers, N = 63, GDAC egg chambers, N = 62. P-values obtained from Mann Whitney test are indicated by **** (p ≤ 0.0001) and ns (non-significant). (F) Bar plots with error bars displaying percentage of egg chambers exhibiting defective phagocytosis in each replicate (Rep.) across different genotypes: NICD OE (Rep. 1, N = 54, Rep. 2, N = 43, Rep. 3, N = 57), NICD OE + Arp2 RNAi (Rep. 1, N = 89, Rep. 2, N = 102, Rep. 3, N = 173), NICD OE + Arp3 RNAi (Rep. 1, N = 141, Rep. 2, N = 73, Rep. 3, N = 60). A p-value obtained from t-test is indicated by * (p = 0.0157) and ** (p = 0.002) above bar plot. (G) Proposed working model generated in BioRender illustrating the transition of a follicle cell (FC) into a Non-Professional Phagocyte (NPP). The model highlights the regulatory role of Jra and the Arp2/3 complex, with Arp2 and Arp3 genes encoding components of the Arp2/3 complex, predicted by SCENIC and Chip-seq to be a downstream target of Jra.

https://doi.org/10.1371/journal.pgen.1011953.g007

To validate Arp2/3 expression in NPPs, we performed immunofluorescence using Arp3-GFP and confirmed its expression in both NICD-OE- and starvation-derived NPPs. Our results showed that Arp3-GFP expression gradually increases throughout oogenesis in follicle cells, reaching its peak in stage 9 egg chamber follicle cells and NPPs under both NICD-OE (Fig 7C, E; early-stage egg chambers, median = 0.82, N = 63; mid-stage egg chambers, median = 1.3, N = 63; and GDAC egg chambers, median = 1.4, N = 62) and starvation conditions (S8A, S8B Fig; early-stage egg chambers, median = 0.22, N = 66; mid-stage egg chambers, median = 0.44, N = 30; and GDAC egg chambers, median = 0.51, N = 24). This pattern, to some extent, resembles the gradual increase in Jra:GFP expression observed in egg chambers as they progress to more advanced developmental stages, ultimately culminating in NPP differentiation (S4A-D Fig).

To test the functional role of the Arp2/3 complex in NPPs, we knocked down Arp2 and Arp3 in NICD-OE-derived NPPs and starvation-induced NPPs. In both models, cytoplasmic expansion was impaired and reduced the ability of NPPs to invade the germline cell area. These defects limited the efficient removal of condensed DNA debris left by dying germline cells (Fig 7D, F; mean NICD-OE = 0.2343 (Replicate 1: 0.2593, N = 54, Replicate 2: 0.1628, N = 43, Replicate 3: 0.2807, N = 57), mean NICD-OE + Arp2 RNAi = 0.7708 (Replicate 1: 0.7865, N = 89, Replicate 2: 0.7450, N = 102, Replicate 3: 0.7808, N = 73), mean NICD-OE + Arp3RNAi = 0.6485, (Replicate 1: 0.5319, N = 141, Replicate 2: 0.6301, N = 73, Replicate 3: 0.7833, N = 60); S8C, S8D Fig; mean Control = 0.1547 (Replicate 1: 0.1346, N = 52, Replicate 2: 0.1556, N = 45, Replicate 3: 0.1739, N = 46), mean Arp2 RNAi = 0.641 (Replicate 1: 0.6585, N = 41, Replicate 2: 0.7435, N = 39, Replicate 3: 0.5208, N = 48), mean Arp3 RNAi = 0.8873 (Replicate 1: 0.8667, N = 60, Replicate 2: 0.8596, N = 57, Replicate 3: 0.9355, N = 31).

Single-cell transcriptomic atlas of NICD-OE-induced NPPs

The ability of epithelial cells to acquire phagocytic capabilities is fundamental for maintaining tissue homeostasis by promoting the clearance of apoptotic debris. While professional phagocytes such as macrophages specialize in this function, NPPs acquire phagocytic competence in response to developmental and environmental stimuli [15,44,45]. However, the transcriptional mechanisms regulating NPP differentiation need to be better understood.

Building on our previous study [16], which demonstrated that Notch-induced polyploidization transforms follicle cells into NPPs by permitting JNK pathway activation, we expanded this work by constructing a single-cell transcriptomic atlas. To accomplish this, we analyzed two previously generated scRNA-seq datasets from our laboratory [2,1] to investigate the transcriptional changes induced by NICD-OE during NPP differentiation, in comparison with wild-type ovarian follicle cells. Unlike traditional bulk RNA-seq, scRNA-seq enables cell-type-specific analysis, allowing us to identify subpopulations, differentiation trajectories, and stage-specific gene expression dynamics within NPPs on different stages of development.

While the starvation-induced model of NPP differentiation has been widely used and previously characterized by bulk RNA-seq [46], we observed that our approach utilizing NICD-OE [16] offers a more robust and uniform model of NPP induction. This strategy generates a larger and more homogeneous population of NPPs (S9 Fig), facilitating high-resolution single-cell transcriptomic analysis and improving the reproducibility of NPP differentiation profiling.

We recognize, however, that NICD OE represents an artificial induction system. To evaluate its physiological relevance, we compared NICD-OE-derived NPPs with starvation-induced NPPs and found overlap in key features. Both models showed enrichment of phagocytosis-associated genes such as crq, drpr, prtp, and Rac1 (Fig 2B, S3 Table), activation of autophagy (Fig 4A-F), and dependence on JNK signaling (Fig 5C-E). In addition, our analysis uncovered a novel JNK–Arp2/3 axis driving cytoskeletal remodeling (S6A-C, S7A-D Fig), which was functionally validated in both starvation- and NICD-OE-derived NPPs (Figs 7C-F, S8A-D). Thus, although artificial, NICD-OE recapitulates the essential transcriptional, signaling, and functional hallmarks of endogenous NPP differentiation.

Identification of the transcriptional programs coordinating epithelial-to-NPP transition

Our analysis identified a heterogeneous population of cells in the NICD-OE dataset that transcriptomically diverged from wild-type follicle cells, clustering near mid-stage egg chamber-derived follicle cells while exhibiting distinct transcriptional profiles. Differential gene expression analysis revealed an enrichment of NPP-associated markers, which we validated using fluorescent reporter lines and immunofluorescence, confirming the identity of NICD-OE-derived NPPs.

To determine whether these cells followed a differentiation trajectory, we sub-clustered NICD-OE-derived NPPs and identified three transcriptomically-distinct stages. RNA velocity and pseudotime analyses indicated that NPPs undergo a stepwise differentiation process. While these approaches support a linear model of progression, the inherent heterogeneity of follicle cells suggests that alternative trajectories, including branching or parallel routes, may also contribute to NPP maturation. To further characterize the transcriptional dynamics associated with each differentiation stage, we performed differential gene expression analysis followed by Gene Ontology enrichment analysis, which identified stage-specific transcriptomic profiles defining key competences acquired at each stage of NPP differentiation.

At the earliest stage, cluster 0, the majority of expressed genes was enriched for oogenesis and follicle cell differentiation, as well as genes associated with mitochondrial energy metabolism. This transcriptomical profile suggests that these cells retain characteristics of epithelial precursors while simultaneously undergoing a metabolic shift, likely to meet the energetic demands associated with the acquisition of phagocytic capabilities by NPPs. As differentiation progressed towards cluster 1, gene expression shifted toward cytoskeletal dynamics and cell motility, matching with the stage where NPPs undergo cytoplasmic expansion toward apoptotic germline debris. Finally, in the most advanced NPP population, cluster 2, we observed an upregulation of metabolic adaptation and autophagy-related pathways, suggesting that fully differentiated NPPs reprogram their metabolism to facilitate apoptotic debris degradation.

Through the integration of regulatory network analysis with ChIP-seq data, we propose a mechanistic link in which JNK, through its downstream transcription factor Jra, regulates the expression of Arp2 and Arp3, core components of the Arp2/3 complex, thereby orchestrating the cytoskeletal changes required for NPP phagocytic function. Together, these findings extend our understanding from the upstream prerequisites of NPP induction to the downstream transcriptional and cytoskeletal programs that drive full NPP differentiation.

Furthermore, resolving NPPs into distinct transcriptional states, our approach represents a major technical advancement, enabling the identification of key molecular signatures defining each phase of epithelial-to-NPP transitions. This refined model can now serve as a framework for studying NPPs in other biological contexts.

Single-Cell approaches as a powerful tool to study NPP differentiation across model systems

The strategy of leveraging scRNA-seq analytical tools to analyze the differentiation steps of epithelial cells into NPPs could enhance our understanding on several diseases development and identify potential therapeutical targets. For example, retinal pigment epithelium (RPE) cells are originally derived from epithelial cells and have acquired specialized phagocytic capabilities to perform the phagocytosis and turnover of photoreceptor outer segments. This process occurs in three stages: binding, endocytosis, and elimination. All essential to assure that RPE cells eliminate exfoliated POS properly and maintain normal renewal of visual cells [47,48,49,50].

While scRNA-seq has been largely deployed to investigate RPE-cells [51], we believe that sub-clustering RPE-phagocytic cells and tracking shifts across differentiation stages, can be a new approach to address the RPE-transcriptomic changes based on their main phagocytic competences (binding, endocytosis and elimination). Integrating techniques such as SCENIC regulatory network analysis could help identify the transcriptional programs undergoing each stage of RPE-mediated phagocytosis, providing insights into key-gene regulators governing RPE differentiation and function.

Understanding the transcriptional mechanisms that enable RPE cells to acquire phagocytic function could provide critical insights into the molecular changes underlying defective RPE phagocytosis, which contribute to diseases such as age-related macular degeneration (AMD) [52,53], retinitis pigmentosa (RP) [52], and Stargardt disease [54].

JNK signaling as a key regulator of NPP differentiation

JNK signaling is a conserved sub-branch of the MAPK signaling kinase present across metazoans [55,56,57,58,59]. The role of MAPK signaling kinase has been largely reported in mammalian professional and non-professional phagocytes [60,61,62,63]. Our study proposed to answer how JNK signaling orchestrate mid-stage ovarian NPP differentiation. Through SCENIC analysis, we identified, the JNK downstream transcription factor, Jra as one of the most active network associated with cell migration and cytoskeleton remodeling, required functionalities by early maturating NPPs. We confirmed that depletion of Jra resulted in impaired NPP invasion and cytoplasmic expansion, preventing cells from invade germline cell area and clearing dead-germline cells debris. On the other hand, overexpressing Jra induced follicle cells to acquire invasive and stretched morphology, supporting the role of JNK activation on driving cytoskeleton remodeling in NPPs.

Further supporting the role of JNK signaling in cytoskeletal remodeling, SCENIC, in association with ChIP-seq analysis, identified Arp2 and Arp3, which encode components of the Arp2/3 complex, as part of the Jra regulon. Given that Arp2/3 is a well-known regulator of actin polymerization [64], we tested the functionality of the Arp2/3 complex on NPP differentiation by depleting two of its main components, Arp2 and Arp3, which partially impaired NPPs of performing proper cytoplasm expansion and engulfment of apoptotic debris. Consistent with these results, both MAPK signaling and the Arp2/3 complex have been implicated in coordinate cell motility and cytoskeleton rearrangements in conserved-different cell types including professional phagocytes such as neutrophils and NPPs like humans fibroblasts [65,66,67,68,69,70].

Thus our approach not only identified conserved regulators of cytoskeleton dynamics but also potentially uncovered a previously unreported transcriptional link between JNK signaling and the Arp2/3 complex that underlies actin remodeling and phagocytic competence in NPPs. This approach offers valuable insight into the mechanisms of non-professional phagocytosis and can help uncover conserved regulatory connections across diverse biological systems.

Fly Stocks and Husbandry

Flies were maintained under standard laboratory conditions at 25 °C and fed with yeast for three days prior to dissection. For starvation experiments, flies were fed with yeast for two days and starved for one day prior to dissection. To induce mosaic clones, we used the FLP-Out Gal4 system, with flies heat-shocked at 37 °C for 15 minutes to achieve a partial mosaic induction.

To confirm JNK pathway activation during dJUN^CA overexpression in follicle cells, we used TRE-GFP as a reporter. For these experiments, hsflp;;act>>RFP was preferred over TjGal4^ts, as it carries an non-occupied second chromosome, facilitating the insertion of TRE-GFP. In this context, flies were heat-shocked at 37 °C for 40 minutes to achieve a more dominant mosaic expression of dJUN^CA. However, for quantitative evaluation of dJUN^CA’s effect on follicle cell invasion of germline cells, TjGal4^ts was preferred due to its more predominant expression in follicle cells compared to the hsflp;;act>>RFP induced mosaic system.

For temperature sensitive experiments utilizing tub-Gal80^ts associated with tj-Gal4 flies crosses were expanded at 18°C, upon F1 flies’ selection, flies were fed with yeast, transferred to a 29°C incubator, and kept for 3 days for transgene induction. Following lines used in this study, UASp-mCherry-Atg8a (#37750), UAS-Atg3-RNAi (#34359), UAS-Jra-RNAi (#31595), UAS-kayDN (#7214), Jra:GFP (#50755), kay:GFP (#38657), UAS-Arp3:GFP (#39722), UAS-Arp2-RNAi (#27705), UAS-Arp3-RNAi (#53972), UAS-GFP (#4775), UAS-RFP (#30556), tubP-GAL80[ts](#7019), were ordered from Bloomington Drosophila Stock Center at Indiana University. TjGal4 driver (Kyoto Stock Center, #104055).

We thank Dr. Singh for providing UAS-dJUN^CA [71], Dr. Bohmann for TRE-GFP [72] and puc-lacZ[E69] [73], and Dr. Schüpbach for generously sharing UAS-NICD [74], which was designed on the Drosophila second chromosome. Additionally, UAS-NICD [75], designed on the Drosophila third chromosome, was generated in our lab and hsflp;;act>>RFP is a tool line routinely used in our lab.

Immunofluorescence staining and imaging

Flies were dissected at room temperature in 1x PBS, followed by fixation utilizing 4% paraformaldehyde (PFA) for 15 minutes. Upon completed fixation, ovaries were washed for 30 minutes in 1x PBT (PBS with 0.2% Triton X-100). Followed by overnight incubation with primary antibody diluted in PBT. Primary-antibody stained ovaries were then washed with 1x PBT for 30 minutes, subsequently secondary antibody, diluted on PBT, was added for either overnight incubation at 4°C or 2 hours at room temperature. Secondary-antibody-stained ovaries were washed for 30 minutes with 1x PBT, followed by 10–15 minutes of DAPI staining (Invitrogen, 1 µg/mL) to stain cells nuclei. Finally, 80% glycerol mounting solution were added and samples were mounted on microscopy slides compatible with confocal miscroscopy.

For LysoTracker staining, samples were dissected in PBS. Following dissection, LysoTracker was added at a 1:1000 dilution. Samples were incubated and covered from the light for 10 minutes. After incubation, PBS was removed, and a PBS wash was performed to reduce background staining. Fix solution was then added, and staining proceeded as normally. The primary antibodies used in this study were obtained from Abcam and the Developmental Studies Hybridoma Bank (DSHB). From Abcam, we used mouse anti-ATP5A (ab14748, 1:500 dilution), Antibodies from DSHB included: mouse anti-arm (N27A1, 1:40), mouse anti-Draper (5D14,1:50), mouse anti-Fas3 (7G10,1:10), Mouse anti-β-gal (40-1a-s,1:5 dilution) and mouse anti-Dlg (4F3,1:50). The secondary antibodies were: Alexa Fluor 488, 546, and 633 (1:400, Invitrogen). Following dies were obtained from Invitrogen: Phalloidin Flour 633 (A22284), LysoTracker Red DND-99 (L7528).

Statistics and reproducibility

Representative images, shown in the Figs, of egg chambers and their respective cells across different genotypes were acquired using Zeiss LSM 800 confocal microscopy. These images are representative of at least three independent experiments, with a minimum of 20 ovaries dissected per genetic cross. The genotypes used for each experiment are indicated in the corresponding Fig legends.

Images utilized for quantitative fluorescence analysis, intensity mean values for antibody staining, dye staining, and fluorescent proteins (e.g., GFP) were obtained using Zeiss LSM 980 confocal microscopy. To ensure reproducibility and comparability, all samples were prepared under standardized experimental conditions, and fixed imaging settings were applied across all samples. Image saturation was avoided during image taking, to ensure consistent and accurate signal acquisition.

Image analysis and processing were conducted using ZEN Blue (Zeiss). The regions of interest (ROIs) corresponding to egg chambers undergoing GDAC were manually outlined using the spline contour tool and compared to control (non-GDAC) egg chambers. Mean fluorescence intensity values for each selected ROI were recorded and analyzed using GraphPad Prism version 9.5.1 for Windows (GraphPad Software, Boston, Massachusetts, USA, www.graphpad.com). The colocalization of RFP-expressing cells with TPA response element-GFP (TRE:GFP) expression was analyzed using the colocalization toll provided by ZEN blue and the Colocalization Coefficient was selected as the method to quantify colocalization.

To ensure statistical robustness, colocalization and intensity mean values identified as outliers were excluded prior to analysis. Normality tests were then performed to determine the appropriate statistical test. If data followed a normal distribution, unpaired t-tests with Welch’s correction were applied. For data sets that did not meet the assumptions of normality, the non-parametric Mann-Whitney U test was used. In both cases, a p-value < 0.05 was considered statistically significant.

Fluorescence intensity values were normalized to DAPI mean intensity after confirming that DAPI levels did not significantly vary across samples. Bar plots representing normalized fluorescence intensity or colocalization values across genotypes were generated using GraphPad Prism. Error bars represent interquartile range (IQR) for Mann-Whitney U tests and standard error for t-tests, with corresponding p-values indicated.

Single-cell RNA-sequencing data processing

The scRNA-seq dataset used for the analysis of TjGal4^ts>NICD-OE cells was originally generated and published by our lab [1] and is publicly available in the SRA database (SRX11836918). After running Cell Ranger processing pipeline, we obtained an estimated number of 33,703 cells with an average of 10,683 reads per cell and a median of 1,014 genes per cell, with a total of 360,058,745 reads.

Similarly, the scRNA-seq dataset used for the analysis of w¹¹¹⁸ ovarian cells in this study was also originally published by our lab [2] and is publicly available in the SRA database (SRX7814226). Following Cell Ranger processing, we obtained estimated 37,928 cells with an average of 11,333 reads per cell and a median of 907 genes per cell, with a total of 459,855,892 reads.

Cell ranger processing

Raw FASTQ files obtained from NCBI SRA (experiment accession numbers SRX11836918 and SRX7814226) were processed using Cell Ranger v7.1.0. The reference genome used for alignment was the Berkeley Drosophila Genome Project 6 (BDGP6.46.dna.toplevel). The Cell Ranger count pipeline was used for read alignment, filtering, barcode assignment, and unique molecular identifier (UMI) counting, generating a multidimensional feature-barcode matrix for both w¹¹¹⁸ and NICD-OE ovarian cells.

Following Cell Ranger processing, BAM files from the NICD-OE ovarian dataset were further analyzed using Velocyto for annotation of unspliced and spliced reads, following the 10x Genomics Trajectory Analysis pipeline (https://www.10xgenomics.com/analysis-guides/trajectory-analysis-using-10x-Genomics-single-cell-gene-expression-data) as described by La Manno et al. [76].

Seurat analysis

Single-cell RNA sequencing data were processed and analyzed using Seurat v5.1.0 [17]. Quality control filtering was applied to remove low-quality cells and potential doublets/multiplets, with dataset-specific thresholds established based on gene count, UMI count, and mitochondrial gene expression.

For the w¹¹¹⁸ dataset, cells with more than 3,200 detected genes and total RNA counts exceeding 30,000 were removed to prevent the inclusion of potential doublets or multiplets. Additionally, cells exhibiting mitochondrial gene expression above 5% were excluded to filter out low-quality or dying cells.

For the NICD-OE dataset, an upper threshold of 3,400 genes per cell and 31,000 RNA counts per cell was applied to eliminate potential doublets or multiplets. Similar to the w¹¹¹⁸ dataset, cells with mitochondrial gene expression exceeding 5% were removed. Mitochondrial genes were identified based on the ‘mt:’ prefix in the annotation.

After preprocessing, 8,871 cells were retained for the w¹¹¹⁸ dataset, while 25,122 cells were obtained for the Tjts>NICD-OE dataset. The gene expression matrices for these cells were normalized, scaled, and subjected to linear dimensionality reduction following the Seurat pipeline. Cell-cycle scoring was performed using the list of cell-cycle genes proposed by [2], and cell cycle scores were regressed out during data scaling to prevent clustering and downstream analyses from being biased by cell cycle effects.

To further ensure the retention of high-quality cells, the ScDblFinder package was used to identify and remove predicted doublets. Cells were then embedded in low-dimensional space using UMAP with parameters in according to the Seurat pipeline. Cells that were displayed outside their original clusters were considered outliers and were removed to maintain cluster integrity and improve cluster visualization. As a result of these quality control steps, a final dataset of 8,037 cells was obtained for w¹¹¹⁸, and 21,704 cells for Tjts>NICD OE.

To focus on follicle cell differentiation into NPPs, non-follicle cell types were identified and removed based on the expression of marker genes defined by previous studies [18,2,1,19]. After plotting follicle cell only related clusters with UMAP, additional outliers were identified and removed. Following this final refinement, the dataset consisted of 4,264 high-quality follicle cells for w¹¹¹⁸ and 17,304 high-quality follicle cells for Tjts>NICD OE.

To integrate Tjts>NICD OE and w¹¹¹⁸ follicle cells, we used the FindIntegrationAnchors() and IntegrateData() functions [77], specifying 2000 genes for anchor finding and 50 dimensions for Canonical Correlation Analysis (CCA). We selected CCA for dataset integration as it better captured shared variance while minimizing batch effects in our dataset. While Reciprocal PCA (RPCA) is computationally more efficient, CCA provided superior integration, preserving known biological clusters. Additionally, CCA was chosen over RPCA because both the NICD-OE and w¹¹¹⁸ datasets display sufficient overlapping features, making it an appropriate method for aligning shared cell populations while maintaining genotype-specific differences.

After integration, UMI counts in the integrated assay were normalized, log-transformed, and scaled. Cells were clustered using 40 principal components (PCs) and a resolution parameter of 0.325 was chosen. To visualize the integrated dataset, cells were projected onto a lower-dimensional UMAP space, using 50 nearest neighbors and a minimum distance of 0.7 to approximate the local manifold structure.

To perform differential gene expression (DGE) analysis, we used the “RNA” assay from each Seurat object. For integrated datasets in which the “RNA” assay contained multiple expression layers (corresponding to control and NICD overexpression conditions), we applied the JoinLayers() function (Seurat v5) to merge the layers into a unified expression matrix, enabling DGE analysis using FindMarkers().

To recover missing gene count values in the RNA assay, we applied ALRA imputation [78]. The imputed matrix was stored in the “alra” assay and was used for generating gene enrichment plots. To improve visualization of cluster-specific marker enrichment, the plots were scaled between the range of 5^th and 95^th quartiles.

Postprocessing of single-cell RNA-sequencing data