Cell cycle expression of polarity genes features Rb targeting of Vang

Abstract

Specification of cellular polarity is vital to normal tissue development and function. Pioneering studies in Drosophila and C. elegans have elucidated the composition and dynamics of protein complexes critical for establishment of cell polarity, which is manifest in processes such as cell migration and asymmetric cell division. Conserved throughout metazoans, planar cell polarity (PCP) genes are implicated in disease, including neural tube closure defects associated with mutations in VANGL1/2. PCP protein regulation is well studied; however, relatively little is known about transcriptional regulation of these genes. Our earlier study revealed an unexpected role for the fly Rbf1 retinoblastoma corepressor protein, a regulator of cell cycle genes, in transcriptional regulation of polarity genes. Here we analyze the physiological relevance of the role of E2F/Rbf proteins in the transcription of the key core polarity gene Vang. Targeted mutations to the E2F site within the Vang promoter disrupts binding of E2F/Rbf proteins in vivo, leading to polarity defects in wing hairs. E2F regulation of Vang is supported by the requirement for this motif in a reporter gene. Interestingly, the promoter is repressed by overexpression of E2F1, a transcription factor generally identified as an activator. Consistent with the regulation of this polarity gene by E2F and Rbf factors, expression of Vang and other polarity genes is found to peak in G2/M phase in cells of the embryo and wing imaginal disc, suggesting that cell cycle signals may play a role in regulation of these genes. These findings suggest that the E2F/Rbf complex mechanistically links cell proliferation and polarity.

1. Introduction

Proper establishment of cell polarity is integral to development and homeostasis of all metazoans. Polarity is associated with asymmetric distribution of cellular components, and this distribution is important for defining and maintaining proper tissue architecture. Most epithelial cells display apical-basal polarity, which is critical for maintaining tissue integrity and function. In addition, epithelial cells can also display planar cell polarity (PCP), which coordinates the orientation of cellular structures orthogonal to the apical-basal axis. Planar polarity is manifest in the proximal- distal arrangement of trichomes in the fly wings, stereocilia bundles in the inner ear of mammals, as well as patterned organization of mammalian hair [1, 2].

Three distinct protein complexes play key roles in apical-basal polarity. Mutual antagonism among the Scribble/Discs large/Lethal giant larvae, the aPKC/Par6/Bazooka (Baz; Par3 in vertebrates), and the Crumbs/Stardust (Sdt; PALS1 in vertebrates)/Patj complexes ensures proper establishment and maintenance of polarity. At the heart of planar polarity are six evolutionarily conserved proteins. The transmembrane protein Frizzled (Fz; FZD in vertebrates) along with its cytoplasmic partners Dishevelled (Dsh; DVL in vertebrates) and Diego (Dgo; ANKRD6 in vertebrates) are localized distally, while Van Gogh (Vang, also known as Strabismus (Stbm); VANGL in vertebrates), also a transmembrane protein, along with its cytoplasmic partner Prickle (Pk) is localized in a complementary manner to the Fz complex on the proximal side. In addition, the transmembrane protein Flamingo (Fmi, also known as Starry night (Stan); CELSR in vertebrates) resides on both sides of the cell [3, 4]. Polarity protein complexes are regulated by inter-complex interactions in settings where apical-basal polarity intersects planar polarity. For example in the Drosophila embryo, Bazooka is planar polarized by the Fat-PCP pathway, while Vang is responsible for Baz polarization in the eye [5, 6]. In mammals, Scribble has been shown to physically interact with Vangl2, and loss of either one results in similar neural tube defects [7-9].

Similarly, coupling of polarity processes to the cell cycle also involves regulation of polarity proteins in the cytoplasm. Cycling mouse embryonic basal epidermis cells maintain polarity by endocytosing junctional PCP proteins during cell cycle to prevent cells from sending and receiving signals during cell division. The trans-endocytosed PCP proteins are subsequently degraded, and new PCP protein synthesis ensues to maintain proper levels during cytokinesis [10, 11]. These studies indicate that maintenance of proper levels of polarity proteins relies on their replenishment during cell cycle, which may involve a transcriptional component. Regulation of cell cycle and cell polarity has generally been studied separately; however, there are some indications that the polarization state of a cell affects its proliferative state. During Drosophila neuroblast development, the polarized mother cell stays in cell cycle and remains as progenitor stem cell while the non-polarized ganglion mother cell, with limited proliferative potential, divides only once to terminally differentiate into neurons or glial cells. Similarly, in the vertebrate nervous system, the non-polarized progenitors are closer to cell cycle exit than their polarized counterparts [12].

These findings, summarized in Noatynska et al., (2013) [12] indicate that polarity and proliferation are coupled; however, a possible role in this coordinated process for transcriptional regulation of genes encoding the core polarity factors is unclear. Previously, we showed that a number of planar and apical-basal polarity genes were bound by the key cell cycle regulator protein complex, E2F/Rbf [13]. Retinoblastoma proteins (Rbf1 and Rbf2 in flies; RB, p130, p107 in vertebrates) are transcriptional corepressors that, in addition to cell proliferation, regulate cell differentiation, DNA repair, senescence, metabolism, and growth [14, and references therein]. Retinoblastoma proteins are recruited to target genes via heteromeric E2F/Dimerization partner (Dp) transcription factors, and it is these factors that permit cell cycle dependent activation of genes upon inactivation of RB proteins by cyclin/kinase complexes [15]. Multiple components of the Rbf-containing dREAM complex (Drosophila Rbf, E2F and Myb interacting proteins; mammalian homolog DREAM (Dimerization partner (DP), RB-like, E2F and MuvB)) are also found on the promoters of polarity genes such as Vang, aPKC, and baz [16]. The functional relevance of this occupancy was underscored by the significant upregulation of key apical-basal and planar polarity genes such as Vang, aPKC, and par6 upon depletion of Rbf1, which induced polarity defects in adult fly wings, ommatidia, and notum [13]. These results raised the possibility that regulation by E2F/Rbf provides a means to coordinate polarity and the cell cycle. However, E2F/Rbf can also control gene expression independent of the cell cycle, thus this coupling hypothesis remains to be tested. Additionally, although the physical presence of E2F/Rbf factors on polarity gene promoters favors a direct regulatory role, Rbf1 knockdown generates pleiotropic effects, which make it inherently difficult to differentiate direct from indirect effects. We thus further investigated the cell cycle aspects of polarity gene expression and explored direct E2F/Rbf interactions with a key polarity gene, Vang.

2. Results

2.1. Mitotically active wing discs show cell cycle phase dependent expression of Vang

To determine whether polarity genes are entrained into the cell cycle, we assessed polarity gene expression in distinct cell phases. We turned to recent single-cell RNA-seq datasets from Drosophila wing imaginal disc and the embryo, which contain proliferating cells in different parts of the cell cycle [17, 18]. Using the expression of 68 previously-characterized cell cycle marker genes as a guide, we inferred cycle stages [19-21]. Based on the relative expression of these predictive genes as a group, we sorted cells into G1, S, and G2/M classes. The relative expression levels of most of the individual marker genes correlated well with their assigned groups, with highest expression of S-phase markers in the S-phase group etc. although there were some outliers (Fig. 1 A). We then analyzed the expression of twenty polarity genes in each of the sorted phases of cells, including core planar polarity genes, apical basal polarity genes, and other known effectors of the polarity pathway. Strikingly, in the wing discs, eleven polarity genes exhibited a pronounced enrichment for G2/M phase expression. This pattern included Vang, fz, ft and their partners dsh, and ds (Fig.1B). Analyzing the distribution of cells within each of the three cell cycle categories with detectable Vang expression, we observed that while Vang transcripts were detected in all three groups, a modest but significantly higher proportion of cells with detectable levels of Vang were in G2/M (Fig. 1C). The median expression of Vang was also somewhat higher in cells classified as G2/M versus S phase (Fig. 1D, Table S1), which likely underlies the higher fraction of cells of this group that were scored as Vang positive. The bias for expression of Vang in G2/M phase cells was, however, not as dramatic as that exhibited by “classic” cell cycle genes such as Mcm5, an S-phase specific, or pav, a G2/M- specific gene. (Fig. 1E- H). Similar modest, but significant enrichment in fractions of cells scored, and median values was noted for other cell polarity genes; the more robust pattern for Mcm5 was also seen for other cell cycle genes (Suppl. Figs. S1- S4). Based on previous ChIP-seq and ChIP-chip data, we found that eight out of the eleven G2/M-enriched polarity genes are bound by E2F/Rbf proteins and other members of the dREAM complex, consistent with a direct transcriptional regulation by these cell cycle factors (Table S2) [16, 22- 24].

Figure 1.

Cell cycle dependent polarity gene expression in wing discs (A) Heat map showing the relative mean expression of canonical cell cycle phase marker genes in wing disc cells grouped by presumed phase. The marker row denotes the cell cycle phase for which the genes have been previously shown to be enriched. The color in the heatmap represents the z-score of the mean expression across the three phases, with white corresponding to average mean expression, red above average, and blue below average. The genes are ordered by hierarchical clustering; as expected, most genes previously characterized as enriched in G1, S, and G2/M show enriched expression in the corresponding group. (B) Heat map showing the relative mean expression of polarity genes in wing disc cells grouped by inferred cell cycle phase based on markers shown in (A). Many show a bias for higher expression in G2/M, and lower expression in S phase. The color in the heatmap represents the z-score of the mean expression across the three phases with white corresponding to average mean expression, red above average, and blue below average. The genes are ordered by hierarchical clustering. P-values were calculated using the Kruskal-Wallis one-way ANOVA. * denotes a Benjamini-Hochberg adjusted p-value < 0.05. (C) Stacked bar plot showing the number of wing disc cells with and without detectable Vang expression in each phase. P-values, denoting significance of differences in the fraction of cells with detectable expression, were calculated using a one-tailed Fisher’s exact test. (D) Dot plot showing the log₂ scaled normalized Vang expression in each wing disc cell, grouped by phase. The overlaying box plot (25%-75%) summarizes the data within the dot plot. The Benjamini-Hochberg adjusted p-value calculated by the Kruskal-Wallis one-way ANOVA is shown above the plot, and pairwise p-values from the Dunn’s Multiple Comparison test are shown in the plot area. (E) Numbers of cells with detectable expression of canonical cell cycle marker gene Mcm5. (F) Expression levels across the cell cycle for Mcm5. (G) Numbers of cells with detectable expression of G2/M specific gene pavarotti. (H) Expression levels across the cell cycle for pavarotti.

In phase sorted cells from stage 6 embryos, seven polarity genes showed significant cell cycle enrichment (Suppl. Figs. S5A, B). As in the wing disc, baz, Rok and ft were enriched in G2/M compared to S phase. A number of other genes whose expression was significantly enriched in G2/M in wing discs showed the same tendency in the embryo data, but did not reach statistical significance, including Vang, dsh, par6, fz, and ds. This relative expression indicates that there is also a cell cycle bias at this earlier developmental stage. However, unlike in the wing disc, for Vang, the fraction of cells scored as positive for expression was equivalent across all three cell cycle classes of the embryonic cells, and the median expression showed no statistically significant difference (Suppl. Figs. S5C, D; Table S1). A similar pattern was also seen for several other polarity genes. As in wing disc cells, Mcm5 and pav were enriched in S and G2/M phases respectively, with the fraction of Mcm5-positive or pav- positive cells also being higher in the respective phases (Suppl. Fig. S5E- G). However, the overall differences were much less striking than those observed in wing disc cells, and this trend was also seen for other cell cycle and polarity genes (Suppl. Figs.S5F; S6- S9). Overall, the lower fraction of polarity genes showing cell cycle enrichment, as well as the smaller magnitude of cell cycle signal for canonical markers, likely reflects the specific developmental stage sampled i.e. the onset of gastrulation, when cells have completed synchronous mitotic cycling and begin to show complex mitotic pattern in discrete domains. In the wing disc, where mitotic cycles are more prominent, the marked phase specific enrichment of Vang and other polarity genes point to a coordinate regulation of polarity determinants during the cell cycle.

2.2. Disruption of E2F binding motif in Vang promoter induces polarity defects in wing

We analyzed the Vang promoter for the presence of E2F motifs and identified a canonical E2F like motif (TTTGGCGG) at −70 bp (Fig. 2A). We disrupted the motif in vivo using CRISPR/Cas9 and obtained two alleles. Vang^TGA consists of a deletion of a thymidine residue and a substitution of G to A in the core motif (TT-GACGG), while Vang^TΔ displayed a single base deletion, with an otherwise intact core (TT-GGCGG). Flies homozygous for either of the two alleles were viable, and showed wing polarity defects (Fig. 2C, D). We did not observe ommatidial rotation defects such as those noted with Vang^stbm-6 or Vang^stbm-153 (data not shown). The subtle perturbation is evidently not sufficient to cause visible defects in this tissue, which may possibly also be less sensitive to modest changes in Vang expression. The misoriented hair pattern in the wing was observed predominantly in the posterior region of the wing (“region D”), while other regions showed wild type distal pointing hairs. Transheterozygous flies also displayed similar wing phenotypes. The polarity defect phenotype noted in the two CRISPR-induced promoter mutants was weaker than the phenotype noted in Vang⁶, a null allele, or the phenotype described for other alleles such as Vang^A3, Vang^TBS42 or Vang⁴⁰¹⁴. These alleles showed a more penetrant phenotype in comparison to the region specific phenotype noted in our promoter mutants [25]. Our phenotype was similar to the region specific defect noted in Humphries et al., (2020) when Vang was overexpressed [26].

Figure 2.

Vang promoter mutations in E2F motif induce polarity phenotype in Drosophila wings (A) Schematic representation of Vang 5’ region with E2F and CHR (cell cycle genes homology region) motifs boxed. Vang^TΔ and Vang ^TGA represent two alleles with specific changes noted. (B) Wild-type wing showing normal distally-pointing wing hairs. All wings are oriented distal to the right and anterior to the top. Region D, the region between the fourth and the fifth longitudinal vein, distal to the posterior cross vein is shown for all the images. Blue arrows indicate normal wing hair polarity. (C, D) Vang^TΔ and Vang ^TGA homozygous mutant wings exhibit abnormal hair polarity (red arrows) distal to the posterior cross vein. Normal hair polarity is observed along the vein margin and posterior edge. (E - G) Higher magnification (40x) of regions from wild type wing showing normal orientation of hairs (E), and of Vang^TΔ (F) and Vang ^TGA (G) wings showing abnormal orientation. (E’-G’) False colored images of wing regions shown in panels E to G to more clearly distinguish hair orientation. Counting was done blind; 75/92 wings from homozygous Vang^TΔ flies, and 95/130 wings from homozygous Vang flies showed abnormal hair polarity. From a transheterozygous cross (Vang^TΔ / Vang ^TGA), 66/79 wings showed polarity defects.

To test for genetic interactions with other Vang alleles, we crossed the flies bearing the Vang promoter mutations to Vang⁶, a null, or Vang¹⁵³, a hypomorphic allele. We observed in all cases a suppression of the wing hair phenotype, consistent with the promoter mutations inducing expression of Vang in certain settings where Rb is otherwise active. It is possible that the CRISPR edits were in both cases accompanied by an off-target mutation, which was complemented in these crosses. However, the preponderance of evidence supports a direct effect of Vang; we previously showed that Vang expression increases in an Rbf1 knockdown, and as noted above the phenotype is consistent with Vang overexpression. We furthermore show below that Rb binding is lost in a Vang ^TGA mutant [13].

2.3. Loss of Rbf and E2F in vivo binding to mutant Vang promoter

To determine the effect of the mutation of the E2F site on in vivo E2F/Rb binding, we carried out chromatin immunoprecipitation (ChIP) analysis on the Vang promoter and a number of control loci. Developmental RNA-seq and developmental proteome life cycle data indicated that endogenous Vang transcripts and protein levels peak late in embryogenesis. Therefore, we performed ChIP on chromatin from 12 - 18 hr embryos, measuring Rbf and E2F interactions with known target genes DNAPol-a50, InR, Held out wings (How) and Arp53D, as well as Vang (Fig.3A-E). Binding of Rbf1, Rbf2, and E2F2 proteins were significantly reduced on the Vang promoter in Vang^TGA flies compared to wild type, while other E2F/Rb targets were not affected (Fig. 3A). E2F1 protein occupancy on the Vang promoter was very modest compared to that observed on DNAPol-a50, a classic E2F1 dependent S phase gene, and E2F1 remained unaltered in Vang^TGA flies compared to wild type control (Fig. 3A). Signals from E2F or Rb protein immunoprecipitations from an intergenic locus lacking E2F binding sites were low, and equivalent to those produced by nonspecific IgG immunoprecipitation, demonstrating the specificity of the assay (Fig. 3F). Thus, the E2F motif in Vang is directly involved in the recruitment of Rbf and E2F factors.

Figure 3.

Disruption of E2F motif in Vang promoter specifically reduces E2F and Rbf protein interaction without significant change in wing disc expression Interaction of Rbf1, Rbf2, E2F1, E2F2 proteins on various promoters assayed by chromatin immunoprecipitation. (A) Vang promoter region of Vang ^TGA mutants showed a significant loss of binding of Rbf1, Rbf2 and E2F2 proteins. No changes in occupancy of E2F or Rbf proteins were observed on promoters of four other E2F/Rbf target genes, (B) Polα50 (C) InR (insulin receptor) (D) How (held out wings), and (E) Arp53D (Actin-related protein 53D). (F) Negative control locus that contains no known or predicted E2F binding sites (Chr 2R: 9915750- 9915846, Zappia et al., 2019, [48]). Specific E2F and Rbf antibody signals were not higher than that of nonspecific IgG background in this region. Acetylated H3K27 levels remained unaltered between wild-type and Vang ^TGA flies for all of the promoter regions tested. Anti-IgG antibody served as negative control. n=3 biologically independent replicates per antibody; error bars represent SD, * p<0.05, ** p<0.01. Black dots represent values obtained for each biological replicate, for each antibody at the respective locus. Chromatin was prepared from 12-18hr embryos. (G) mRNA expression levels of Vang. n=4 biological replicates, error bars represent SD. Expression was measured using RNA extracted from 3^rd instar larval wing imaginal discs and 30-35 hrs. pupal wings.

We examined whether the reduced interactions of E2F/Rbf proteins in the mutant resulted in altered levels of Vang mRNA, reasoning that loss of Rbf might induce upregulation, or loss of E2F downregulation, depending on the state of the promoter. However, Vang expression in mutant larval wing imaginal discs was unaltered within the precision of our measurements (Fig. 3G). Vang levels measured in 12-18 hr embryos, larval salivary gland tissue, and adult ovaries were similarly equivalent in mutant and wild type backgrounds (data not shown). It is possible that the transcription levels are balanced out by loss of the repressor Rbf and activator E2F or that the effect is rather subtle, and too small for us to detect. Alternatively, it is possible that transcriptional defects leading to wing polarity phenotypes manifest themselves later in pupal development, after the time point we sampled. Indeed, pupal wing RNA-seq data show increasing levels of Vang between 42 and 72 hrs after puparium formation (apf), with peak levels observed at 72 hrs apf followed by a drop by 96 hrs apf. This pattern mirrors measured levels of E2F1 mRNA in the developing wing [27]. Interestingly, at an earlier time point (36 hr apf) when endogenous Vang levels are low, overexpression of E2F1 and DP significantly upregulated Vang [28]. These results support a model for direct regulation of the Vang promoter by the E2F/Rb proteins. Therefore, we analyzed Vang expression in 30-35 hr pupal wings (a stage when PCP signaling is initiating). We find that Vang transcript levels showed a subtle increase in three out of four biological replicates; however, this did not reach statistical significance (data not shown). It is likely that the mild phenotype is a reflection of a modest upregulation of Vang, and the precision of our measurements is limited.

2.4. Expression of Vang relies strongly on the presence of an intact E2F binding element in S2 cells

To further investigate the functional significance of the E2F motif for Vang promoter activity, we created luciferase reporters extending from −305 to +336 from the Vang TSS, and generated mutant forms in which we targeted the E2F motif (Fig.4A). The luciferase activity obtained from the wild type reporter gene was robust, on a similar level to that of the previously studied PCNA promoter [29]. The E2FΔ reporter bearing an 8bp deletion of the E2F sequence and the TGA reporter with two E2F site mutations were only about 50% as active as the wild type promoter, demonstrating the function this motif plays in driving expression in S2 cells (Fig.4B). The overall activity of the TΔ reporter was only modestly reduced consistent with the less extensive change in sequence, although these differences were not statistically significant using single pairwise T tests, when considered as a group the lower activity was significant (ANOVA, p<0.001). Taken together the weaker activity of the mutants supports the functional importance of this motif in regulating Vang expression in cultured cells. We overexpressed Rbf or E2F to assess their abilities to regulate the wild type and mutant promoters. Interestingly, except on the TGA mutant, Rbf1 or Rbf2 expression did not significantly impact the activity of the reporters. This result is in agreement with our previous study, where we observed that overexpression of Rbf1 and Rbf2 in the embryo did not result in a significant change in the steady state levels of Vang [29, and unpublished data], possibly because there Vang promoter is fully bound by endogenous Rbf proteins.

Figure 4.

E2F motif required for Vang activity in cultured S2 cells (A) Schematic representation of luciferase reporter plasmids. Vang wt possesses an intact E2F motif; this motif is deleted to create the Vang E2FΔ reporter plasmid. Vang TGA and Vang TΔ reporters show changes identical to CRISPR alleles. (B) Normalized reporter activity. Vang wt, Vang E2FΔ, Vang TGA or Vang TΔ plasmids were co-transfected with pAX, Rbf or E2F plasmids, and luciferase reporter activity was measured. Compared to wildtype, Vang E2FΔ and Vang TGA reporter activities were significantly reduced while the activity of Vang TΔ was similar to that of the control. Except for Vang TGA that showed a modest response to Rbf2, all of the reporters were insensitive to overexpression of Rbf1 or Rbf2 proteins. E2F1 expression repressed both Vang wt and mutant reporters; E2F2 strongly induced E2FΔ reporter while the activity of others remained unaffected. Luciferase readings were normalized to PCNA reporter transfections performed in parallel on the same day. n=6 biological replicates, error bars represent SD, * p<0.05, ** p<0.01. Black dots show values obtained for each replicate under individual transfection conditions.

In contrast, E2F1 and E2F2 expression had significant and opposite effects on Vang activity. On both wild type and mutant promoters, expression of E2F1 negatively regulated activity by 50%-60%. For this effect, E2F1 may rely on other promoter elements such as an additional E2F motif located at −190 bp. On PCNA and other G1/S promoters, E2F1 is an activator; however, on certain G2/M-active promoters such as CycB, repression by E2F1 has been reported [30]. Interestingly, the activity of E2F2 was more context dependent; on the wild type promoter, E2F2 had no effect, while E2F2 significantly upregulated the deletion mutant reporter, and not others. This exception to E2F2’s function as a repressor is also found for regulation of CycB, where it exhibits potent activation potential, possibly by antagonism of E2F1-Rbf complexes.

3. Discussion

Coupling of cell division and polarity is frequently observed in diverse developmental settings, including in mouse basal epidermis, where, following cell division, Vangl2, Fzd6, and Celsr1 are endocytosed and degraded, followed by reestablishment of polarity utilizing newly synthesized proteins [11]. Cell cycle-related mitotic kinases and cyclins may play a role in this coupling, based on their roles in regulation of asymmetric division machineries in C. elegans and Drosophila [12, and references therein]. These examples do not indicate that such functional interconnections might rely on common transcriptional signals, however. Our previous identification of polarity-related genes as physical targets of Rb proteins suggested that such a transcriptional relationship may exist, linked by Rb-binding E2F transcription factors [22, 24]. Indeed, mutations in Rb as well as Rb depletion have been found to induce polarity defects in Drosophila and mouse [13, 31]. E2F/Rbf proteins are indeed essential regulators of the cell cycle, but the landscape of E2F targets is extensive, and encompasses a wide variety of processes outside of the cell cycle, including growth, metabolism, differentiation and death [32, and references therein]. Thus, E2F/Rb-related transcriptional processes affecting polarity genes might be uncoupled from cell cycle control; this aspect of regulation has been an important point to clarify.

Previous studies that tied Rb function to polarity phenotypes, and upregulation of genes such as Vang and par6, relied on global disruption of Rb expression, necessarily inducing pleiotropic effects that made it difficult to determine if the regulation of polarity gene expression was direct. Here, we show evidence that strongly points to a cell cycle related expression for a number of polarity genes, and, by targeting a key E2F site on Vang, that at least some of this regulation involves direct action on the promoter.

Which promoter characteristics might endow polarity genes with cell cycle responsiveness, and in particular, enriched expression in G2/M? Most polarity genes show E2F/Rb binding. Clearly, this binding is a necessary but not necessarily a sufficient condition to predict G2/M activity. The promoter regions of seven out of the eleven genes with G2/M expression (including Vang, baz, par6, and msn) also showed occupancy by dREAM components (Table S2). However, it appears to be the case that, similar to classical cell cycle genes such as cyclin E, RnrL, RnrS, PCNA and DNAPol-a50, most polarity genes bound by Rbf proteins are also targeted by dREAM complex proteins, suggesting similarities in the regulatory complexes on these genes [16]. Thus, interaction with dREAM protein per se, which may be critical for their regulation, does not appear to be a unique feature that endows these promoters to be active during the G2/M phase. However, the presence of other E2F interacting elements such as the cell cycle homology region (CHR) and cell cycle dependent elements (CDE) may distinguish polarity genes from S phase specific genes. As discussed in Muller and Engeland, 2010 [33], promoters of cell cycle genes such as CyclinB1 and CyclinB2 that show peak G2/M expression feature CHR and CDE elements. Indeed the upstream promoter region of Vang and other polarity genes features a canonical CHR motif. The CHR element is typically correlated with binding of the dREAM complex necessary for repression and activation of late cell cycle genes [34, 35], and it is possible that this element might direct the assembly of promoter specific dREAM complexes. The unusual repression activity of E2F1 on the Vang reporter mimics the response of another canonical G2/M promoter, CycB, which similarly is repressed by E2F1 overexpression, a transcription factor that robustly induces expression of classical S phase genes such as PCNA [29]. These promoters may thus invoke similar G2/M responses through common regulatory mechanisms.

Our results indicate that polarity genes are entrained into cell cycle signals. One outstanding question, however, is how much of the total transcriptional output of polarity genes may be driven by cell cycle signals. At one extreme, many classic S phase promoters are essentially silent prior to cell cycle induction; however, such on/off behavior appears to be less pronounced for a number of G2/M active genes identified in our analysis of the scRNA-seq data (Suppl. Fig. S3) including jumu, Diap2, aurA, Klp10A, which have expression levels similar to that of Vang [36, 37]. Thus, it is possible that the set of G2/M-enriched polarity genes represent the typical dynamics of later cell cycle genes.

Cell cycle signals do not appear to be the only source of regulation for polarity genes, however. Polarity genes are transcribed in post-mitotic cells, including the increasing expression of Vang, fmi, and fz in terminally differentiated fly pupal wings, and of their mammalian counterparts Vangl2, Celsr1, and Fzd3 in post-mitotic mouse cochlear cells [27, 38, 39]. Thus, it is likely that some proportion of polarity gene transcription represents the action of regulatory elements independent of cell cycle signals. The regulation of fj by Fat and Hippo signaling represents one example of the types of additional signals controlling polarity gene expression [40].

We have suggested that the Rb-mediated regulation of widely expressed genes such as those encoding ribosomal proteins genes is necessarily modest, providing a “soft repression” that regulates activity of these highly active promoters within tight physiologically-relevant boundaries [41]. Consistent with this picture, for polarity genes, a G2/M regulatory feature representing “soft” repression inputs of Rb proteins would provide a fillip of expression relevant to a temporary need for extra polarity components as cells divide. The combination of widespread expression coupled with context-specific important, but modest regulatory fluctuations are the product of complex transcriptional controls of the insulin receptor gene [42, 43]. Whether polarity genes similarly possess such elaborate sets of transcriptional enhancers remains to be determined.

3.1. Limitations of the study

In this study we focused on the significance of the E2F/Rb interaction on Vang, but did not test the effect of disrupting E2F interactions on other polarity genes. However, the global cell cycle related regulation observed for a number of polarity genes in the single cell data, together with the physical association of E2F/Rb and dREAM machinery on many of these promoters points to a role for the E2F/Rbf complex as master coordinators of two fundamental processes- cell polarity and cell cycle. Significantly, the loss of polarity control is an important step in cancer progression. Disruption to E2F/Rb pathways may thus directly transcriptionally impact both processes, thus it will be important to examine how E2F and Rb mutations affect the function of these complexes in regulating polarity and proliferation in normal and pathological conditions.

A shortcoming of CRISPR- Cas9 technology is its associated off target effects. Our promoter mutations when tested over Vang loss of function alleles led to a suppression of the phenotype. While this leads us to believe that we have generated a gain of function allele as a result of Rb binding loss, we do not rule out the possibility that the complementation was due to genetic interaction at another locus that was disrupted as a result of nonspecific Cas9 editing.

4. Materials and Methods

4.1. Downloading of single cell RNA-seq expression data and preprocessing

The single-cell RNA-seq expression data generated and described in Deng et. al., 2019 [18] was downloaded from GEO (GSM3902311_frt82b_normalization_data.csv.gz). The deposited data was already normalized by counts per million before the natural log was taken (ln(CPM+1)). In order to arcsine normalize the data we took e to the power of the original normalized value subtracted 1 and then took the arcsine using the asinh() function in R (asinh(e^( ln(CPM+1)) - 1)).

The single-cell RNA-seq expression data generated and described in Karaiskos et. al., [17] was downloaded from GEO (GSE95025_high_quality_cells_digital_expression.txt.gz). Genes with zero expression across all cells were filtered out. To normalize for differences in total count numbers across cells, the counts were multiplied by 1e6 and divided by the total number counts per cell. The resulting values were arcsine normalized with the asinh() function in R.

4.2. Cell-cycle phase prediction

To predict the phase of the cell cycle each cell was in when it was assayed, we used the CellCycleScoring() function in Seurat v3 [44] with custom lists of G2M and S-phase genes. See supplemental table S1. Gene names from Tirosh et. al., [19] were converted from human Ensembl gene symbols to corresponding D. melanogaster Ensembl gene symbols using the bioMart package in R [45]

4.3. Comparison of gene expression by phase

To determine if polarity gene and cell-cycle gene expression differed by cell-cycle phase we used the Kruskal-Wallis ANOVA with a post-hoc Dunn’s test for pairwise comparisons. See supplemental table S1. The results were visualized using box plots overlaid on dot plots or were summarized using heatmaps.

The dot/box plots were made using ggplot2 in R. The middle bar of each box plot shows the box plots median, the lower and upper hinges correspond to the first and third quartiles (the 25th and 75th percentiles). The upper whisker extends from the hinge to the largest value no further than 1.5 * IQR from the hinge (where IQR is the interquartile range, or distance between the first and third quartiles). The lower whisker extends from the hinge to the smallest value at most 1.5 * IQR of the hinge.

The heatmaps show the mean expression value for each gene scaled across the phases. The genes are ordered via hierarchical clustering. Heatmaps were made using the pheatmap package in R, with the options scale = “row” and cluster_rows = TRUE.

To determine if there are more cells with detectable expression of a gene in a particular phase than expected by chance, we labeled cells with non-zero expression as “above detection level” and cells with zero expression as “below detection level”. We then performed Fisher’s exact test comparing the number of cells in each group in a particular phase with the total number of cells in each group. The results were visualized with stacked bar plots made using ggplot2 in R.

4.4. Design of guide RNA sequence for CRISPR, and generation of transgenic flies

Guide RNA sequences were selected using the FlyCRISPR tool (http://flycrispr.molbio.wisc.edu/); each contain 20 nucleotides excluding the PAM, and each was predicted to have zero off-target hits. gRNA sequences that showed overlap with the predicted E2F motif in Vang, were chosen to target the E2F motif. Annealed oligonucleotides containing the guide RNA sequences targeting ebony and Vang were then ligated into the Bbs1 digested pU6-Bbs1-chiRNA vector (Addgene #45946) to generate the transgenic constructs [46]. The sequences for ebony guide RNA are 5′-CTTCGCCACAATTGTCGATCGTCA-3′ and 5′-AAACTGACGATCGACAATTGTGGC. The sequences for Vang guide RNA are 5′-CTTCGAATTTAAACTATCAGTTTGG-3′ and 5′-AAACCCAAACTGATAGTTTAAATTC. The gRNA containing constructs were then injected into y[1] M{w[+mC]=nos-Cas9.P}ZH-2Aw[*] embryos. Injections were performed by BestGene (https://www.thebestgene.com). Fly crosses and screening were carried out using a co-CRISPR strategy adapted from Kane et al., 2017, [47]. Briefly, injected adult flies (F0), were crossed to the double balancer stock w[1118]/Dp(1; Y)y[+]; CyO/Bl[1]; TM2, e/TM6B, e, Tb[1] (Bloomington Drosophila Stock Center No. 3704). Flies in the F1 generation were scored for ebony body color and curly wings. These progeny were crossed to a second chromosome balancer stock (w[*]; P{w[+m*]=GAL4-ey.H}4-8/CyO), and were scored in the F2 generation for wing phenotype. These were then crossed inter se to generate homozygous and balanced fly lines. DNA extraction and PCR amplification of the target region was performed on homozygote and heterozygote flies, and Vang mutations were confirmed using Sanger sequencing. Vang^TGA represents deletion of a thymidine residue and G to A substitution in the core motif (TTTGGCGG to TT-GACGG), and Vang^TΔ represents a single nucleobase deletion in an otherwise intact core (TT-GGCGG). Vang^stbm-6 (stock number 6918) or Vang^stbm-153 (stock number 6919) stocks were obtained from the Bloomington stock center.

4.5. ChIP-qPCR

12 – 18 hour embryos were used for making chromatin extract following the protocol described in Acharya et al., 2012 [22]. Immunoprecipitation was performed using Rbf1 and Rbf2 antibodies [48]. E2F1 and E2F2 antibodies were gifts from. Dr. Maxim Frolov (University of Illinois, Chicago). Input genomic DNA and immunoprecipitated DNA was quantified using qPCR. qPCR was performed using Perfecta SYBR green fast mix (Quanta Bio) on a QuantStudio 3 (Applied Biosystems). Enrichment was calculated as percentage of immune precipitated DNA relative to the input DNA before precipitation for each antibody. Negative control locus represents region on Chr 2R: 9915750- 9915846, and contains no known or predicted E2F binding sites [49]. Three independent biological samples were used for each antibody.

4.6. RT- qPCR

Total RNA was extracted using TRIzol from the third instar larval wing imaginal discs. At least 35 wing discs were used to extract total RNA, and cDNA was prepared using high capacity cDNA reverse transcription kit (Applied Biosystems). qPCR analysis was performed using Perfecta SYBR Green Fastmix (Quanta Bio). The reference genes RpL32, CG8636, NucB1 were validated as control genes. Normalization used the geometric mean of these reference genes. Five independent biological samples were used, and each sample was measured twice. In all experiments normalized expression is relative to wild type control.

Sequences of primers used to measure chromatin occupancy (ChIP-qPCR) and mRNA expression

Gene	Forward primer	Reverse primer	Use
Vang	GGCGGTTGAATTGCCAAACTGA	ACACAATCAGCAGCACGGTC	ChIP-qPCR
Polα50	GCGCGAAAATATGGGAACAG	ATCTAGTGGCTTAAATACGGTTAGAG	ChIP-qPCR
InR	TCCTTGGATTTTGATCATTTTACCC	GCTACGTTTTACTGCTTTCCG	ChIP-qPCR
How	GCGAAACAAAATGCCGCTTG	ATTAAGACACTGCGCTGTGG	ChIP-qPCR
Arp53D	TAGCTGCTTACGTATCGACTGC	TGTTGTGTGCTGTGTTCCAG	ChIP-qPCR
Neg.cont	TGTGTATGCCTTGCTTGCAC	TCTATGCACACGCTCTACTGAG	ChIP-qPCR
Vang	CAGCACACACTCCTCGAAGTC	CTGCCGTCTCCTGTCATATTAAC	RT-qPCR
pk	GCACATGAACCACTTTGCCT	CCTCCCGCATGATATACCGC	RT-qPCR
fz	TTGGCCATACCAAGCAGGAG	GTCATCACTGCAGCCGATCT	RT-qPCR
fmi	CCGCGTACTTGAGTGACA	GGTAAGGCGACTCCGACTTG	RT-qPCR
ft	GACAGGGAGACACAGAACCG	TGCCTCGTCAGTGTCGTTAC	RT-qPCR
aPKC	CCAGCTGGATAGTCCACAGTC	TGAATGGAACTGACCACCGC	RT-qPCR
baz	CGGCGGCTTCAAATGTAAGT	AGGCTGTCCTGTGGTGTTTAAGA	RT-qPCR
Rbf1	ACCAGATGGTCTACAGTTTTTGC	CCGCGTACCTTTTCGTTC	RT-qPCR
RpL32	ATCGGTTACGGATCGAACAAGC	GTAAACGCGGTTCTGCATGAGC	RT-qPCR
CG8636	GATCCGCTGCTAGATCCCAC	CCCTTGTACGGGCAGTTGA	RT-qPCR
NucB1	CTTGAGGCGGACCCTGAGTT	TAGTCCAGCTCCTGTGCGAT	RT-qPCR

4.7. Design of Vang reporter constructs

The Vang promoter region (−305 to +336) was cloned into the AscI and SalI sites in the pAC2T-luciferase vector [50]. The Vang E2FΔ bearing an 8bp deletion of the E2F motif was created using site-directed mutagenesis. The E2F motif described previously [22] was used to identify the Vang motif using MAST (MEME-suite v.5.0.5) using P < 0.001 and P < 0.005 cutoffs. The PCNA-luciferase reporter (a gift from the Nick Dyson laboratory) was previously described [50, 51]. The pIE-E2F1 and pIE-E2F2 vectors were a kind gift from Maxim Frolov [52]. The pAX-Rbf1 and pAX-Rbf2 were previously described [50].

4.8. Luciferase reporter assays

Drosophila S2 cells were cultured in Schneider's medium (Gibco) supplemented with 10% HIFBS and penicillin-streptomycin (100 units/mL penicillin and 100 μg/mL streptomycin, Gibco). 1.5 million cells were transfected using Effectene transfection reagent (Qiagen) with 400 ng of each Vang reporter vector. An equal amount of PCNA-luciferase reporter vector was transfected in separate wells on the same day, and the activities of Vang reporters were normalized to PCNA reporter vector as noted in our earlier study [29]. Transfection of Vang reporters along with pAX-Rbf1, pAX-Rbf2, pIE-E2F1, pIE-E2F2 or pAX control vector was done similarly. 250ng of pAX-Rbf1, pAX-Rbf2, pIE-E2F1, pIE-E2F2 or pAX were used for cotransfection with reporter plasmids. To control for day-to-day variations in transfection efficiency, luciferase values were first normalized to the PCNA reporter, and then re-normalized to the pAX control vector to take into account effects of introduction of the Actin5C promoter on this plasmid. Renilla luciferase was not used as a control, as the activity of this gene is affected by co-transfected effectors. Cells were harvested 72 h post-transfection, and luciferase activity was measured using the Dual-Glo luciferase assay system (Promega).

Highlights.

• Polarity genes show G2/M phase specific expression.

• E2F motif in the key polarity gene Vang directly recruits E2F/Rbf proteins.

• Intact E2F motif is required for Vang activity in S2 cells

• E2F1 activator functions as a negative regulator of Vang

Data availability

This article includes all analyzed data.