A Cell Density-Dependent Reporter in the Drosophila S2 Cells

Cell density regulates many aspects of cell properties and behaviors including metabolism, growth, cytoskeletal structure and locomotion. Importantly, the responses by cultured cells to density signals also uncover key mechanisms that govern animal development and diseases in vivo. Here we characterized a density-responsive reporter system in transgenic Drosophila S2 cells. We show that the reporter genes are strongly induced in a cell density-dependent and reporter-independent fashion. The rapid and reversible induction occurs at the level of mRNA accumulation. We show that multiple DNA elements within the transgene sequences, including a metal response element from the metallothionein gene, contribute to the reporter induction. The reporter induction correlates with changes in multiple cell density and growth regulatory pathways including hypoxia, apoptosis, cell cycle and cytoskeletal organization. Potential applications of such a density-responsive reporter will be discussed.

Reporter activation occurs mainly through mRnA accumulation. Gene regulation can occur at many different levels including rate of transcription, mRNA degradation, as well as protein synthesis, modification, maturation and degradation. In order to distinguish whether the reporter activation occurs at mRNA or protein level, we performed reverse transcriptase-mediated PCR (RT-PCR) to assess the reporter mRNA level in low-and high-density cultures, respectively. As seen in Fig. 2A, RT-PCR followed by gel electrophoresis quantitation suggests a ~30x increase in the GFP mRNA level under high cell density over low density, as compared to the Actin 5C control. Quantitative reverse transcription PCR (RT-qPCR) was also used to measure the GFP mRNA and rp49 control mRNA levels. It showed a ~20x induction (Fig. 2B). These induction ratios are comparable to The total GFP fluorescence level is calculated as the percentage of the GFP positive cells multiplied by the mean GFP intensity of these cells. Left, GFP level in MG cells at low density (5 × 10 5 /mL) without CuSO 4 . This level is used as 1 to calculate fold of induction. Middle, fold of GFP induction in low-density MG cells after Cu ++ induction (see methods). Right, fold of GFP induction in high-density (1.6 × 10 7 /mL) MG cells in the absence of CuSO 4 . N indicates the number of biological replicates. The P-values for the difference between the GFP means of uninduced and induced conditions is marked above the induced data bar.
that of GFP protein induction as measured by FACS (Fig. 1H). Comparable protein and mRNA induction is also seen with the RFP reporter (see below). These results suggest that that reporter activation occurs mostly at the level of mRNA accumulation.

establishing stably transfected cell lines that exhibit density-responsive reporter activation.
A density responsive reporter is useful for biochemical studies and for identifying regulators of growth and proliferation. As such it is imperative to establish and characterize stably integrated transgenic cell lines that provide consistent cell sources and display well-calibrated reporter activation behaviors. P-element-based transposition has been shown to produce predominantly low copy number solitary insertions in stably transfected Drosophila cells 18 . We have also reported the use of the Drosophila P element transposon for improving efficiency of genomic integration and single copy transgene insertion in the S2 cells 9 . As the first step towards generating stable reporter lines that respond to cell density, we co-transfected S2 cells with the MG or MR transgenes and pTurbo, a plasmid encoding the P element transposase (Fig. S2) 9,19 . The GFP or RFP positive cells were enriched by consecutive FACS sorting over the course of eight months to reach a stable polyclonal cell population. Quantitative PCR indicates that the average transgene copy number in these polyclonal cells is ~10 per cell. Both populations show broad range of expression before and after induction, suggesting large variation in the transgene copy number. The GFP population is less pure and shows a similar crowding induction dynamics as the transiently transfected cells, both in changes in the frequency of positive cells, and in the mean GFP level ( Fig. S2A-D, compare to Fig. 1F-H). In contrast, the MR stable population is over 95% positive as judged by its response to Cu ++ . As a result, the MR population shows only a 3-fold increase in crowding induction, reflecting mainly the increase in mean fluorescence (Fig. S2E-H). In summary, both MG and MR transgenes exhibit strong reporter induction by high cell density as integrated copies in the genome and chromosomal environment. These stable transgenic populations are further characterized below to understand the density-mediated induction of reporters.
Reporter induction is density dependable and reversible. We used the stable MG or MR cells to first examine the dynamics of the reporter induction during cell proliferation. We found that the GFP or RFP level began to rise even at low-intermediate cell concentration (1-2 × 10 6 /mL), when they were still actively proliferating (Fig. 3A,B,E,F,I). The reporter expression continued to increase exponentially with cell density, reaching a peak level at ~1.6 × 10 7 /mL (Fig. 3C,G,I). Similarly, the GFP mRNA exhibited a similar exponential increase with the cell density (Fig. 3J). Importantly, the GFP and RFP induction was reversible when cell density was reduced and maintain at 5 × 10 5 /mL through subculture (Fig. 3D,H,K). The GFP mRNA decreased rapidly, reaching the basal level within 48 hours (Fig. 3L). The GFP protein decayed more slowly, reducing to basal level in 3-4 days (Fig. 3K). This could be due to the perdurance of the fluorescent proteins. These observations are also consistent with the finding that the density-mediated induction occurs mainly at the transcript level.
Analysis of transgenes for DnA elements that mediate reporter activation. As the first step towards identifying the cis and trans components that mediate reporter activation at high cell density, we dissected the transgene DNA elements in reporter induction assays. Besides the reporter coding regions, the transgene plasmid contains three major functional elements: the MT enhancer, the eve basal promoter and the CaSpeR vector (Fig. 4A). The MT enhancer from the Drosophila metallothionine gene interacts with the metal responding factor MTF-1 to drive strong transcriptional activation of either homologous and heterologous genes both in vivo and in cultured cells 11,20 . The eve basal promoter contains a 42-bp upstream sequence and a canonical TATA box. It exhibits low basal activity but mediates robust activation when combined with a variety of enhancers in cultured cells and transgenic Drosophila 12,13,21 . The Casper vector contains Drosophila transposon P-elements. We have modified the original Casper vector to remove the adult fly selection marker gene miniwhite 9,22-24 . The GFP reporter was cloned into various combinations of enhancer, promoter and vector elements and introduced into Drosophila S2 cells via transient transfection. We first tested the role of the MT enhancer in either Cu ++ or density-mediated GFP induction ( Fig. 4B-D). Deletion of the MT enhancer abolished or greatly reduced both Cu ++ -and density-mediated GFP induction ( Fig. 4B-D). Replacing MT with 2PE, a mesoderm enhancer from the twist gene, which is known to be active in the S2 cells, largely restored the reporter induction by high cell density but not by Cu ++ (Fig. 4C) [25][26][27] . However, due to the higher basal expression with this enhancer, the fold of induction by high cell density is low (Fig. 4D). This result suggests that the cell density response is not dependent on the MT enhancer, which is known to mediate signals in multiple responses including heavy metal detoxification and hypoxia responses 20,28,29 . Rather, the MT and 2PE enhancers may serve to augment the density response.
We next replaced the eve basal promoter with the Alcohol dehydrogenase (Adh) or Heatshock protein 70 (Hsp70) promoter. The Adh and the Hsp70 basal promoters have been used widely to drive gene expression in both transgenic Drosophila and S2 cells 14,30 . Swapping the eve promoter with the Adh promoter caused a substantial loss in the magnitude of density-mediated induction but only a small reduction in the relative fold of induction due to a reduced basal transcription level. Replacing the eve basal promoter by the Hsp70 promoter also resulted in a higher background level in density-mediated induction. Taken together, although the Adh and hsp70 promoters both respond to cell density cues, the eve promoter provides the most robust response and more suited for diverse applications.
We further evaluated the contribution of the transgene vector by switching the CaSpeR vector with the pTOPO vector (Invitrogen). This dramatically reduced GFP responses to both density and metal inducing cues, although the fold of induction remained comparable to CaSpeR-based combinations ( Fig. 4B-D). This result suggests that sequences within the CaSpeR vector may mediate signals from high cell density. Taken together, the transgene dissection analysis indicates that multiple elements in the original MG transgene collaborate to maximize the reporter response to high cell density. Although none of the elements is essential for the density www.nature.com/scientificreports www.nature.com/scientificreports/ response, the CA-MT-eve combination appears to optimize the reporter induction by high cell density possibly in a synergistic fashion.
Reporter mRnA induction is contributed in part by hypoxia through the Mt enhancer. The role of the MT enhancer in density-driven reporter activation is intriguing. It has been reported that oxygen partial pressure in the pericellular space can be strongly affected by cell density and media height [31][32][33][34] . As a result, extreme hypoxia condition may exist in pericellular space under high cell density. Importantly, hypoxia has been shown to activate the metallothionein gene through the cooperative binding of MTF-1 and HF-1 factors on the MT enhancer 29,35 . To clarify the roles of hypoxia in reporter induction by high cell density, we compared the mRNA induction of GFP or RFP reporters with an in vivo hypoxia marker lactose dehydrogenase (LDH) under various hypoxic conditions. We first tested the MT-RFP transgene induction using the hypoxia-mimicking drug Deferoxamine Mesylate (DFO, Fig. S3). After a 24-hour treatment of 50, 100, or 200 µM DFO, concentrations known to induce hypoxia in cultured cells, we observed 3-4-fold increases in the LDH mRNA level (Fig. S3) [36][37][38] . The same conditions did not significantly alter the reporter mRNA level (Fig. S3). Next we tested the effect of low oxygen partial pressure using a hypoxia incubator chamber (see methods). After a 24-hour treatment under 1% O 2 partial pressure, the LDH mRNA level increases by over 300 fold (Fig. 5A,B). In comparison, 1% O 2 partial pressure only induced reporter mRNA ~4-fold in MT-GFP cells. In contrast, high cell density induced GFP by ~38 fold, but only induced LDH by ~4 fold. These results suggest that although hypoxia can weakly induce reporter mRNA, it cannot account for the dramatic increase of reporter transcription at high cell density. Deletion of the MT enhancer further reduced the reporter response to hypoxia and to crowding (ΔMT, Fig. 5C). Replacing MT with 2PE lead to an increase in the hypoxic response and a decrease in crowding response (2PE, Fig. 5D). This result again indicates that hypoxia is not the major contributor of reporter induction at high cell density.
Hypoxia has been known to suppress protein translation [39][40][41][42] . This suppression could potentially affect the reporter read out at high cell density where pericellular hypoxia occurs. To examine whether the mild induction by hypoxia at the mRNA level is reflected in reporter fluorescence, we conducted FACS analysis the MG cells under 1% oxygen partial pressure. We found only a 1-2 fold changes in the fluorescence level under hypoxia www.nature.com/scientificreports www.nature.com/scientificreports/ condition, compared to the strong GFP or RFP induction by high cell density (Fig. 5E). Taken together, our results indicate that hypoxia is not the major signal that activates the reporter expression at high cell density.
Reporter induction correlates with changes in several growth-related pathways. Besides hypoxia, many signaling pathways are involved in density sensing and growth control. To begin identifying the signals that activate our reporter expression, we surveyed the concurrent changes in several reporter-related cellular genes and transcription targets of growth-related pathways during the natural crowding process. We first compared the induction of GFP with those of metallothionein (MTA), and MTF-1, a transcription factor that binds to the MT enhancers and activates MTA 16,17 . We found that both GFP and MTA mRNAs increased significantly at high cell density (Fig. 6A). Induction of MTA by crowding is not surprising as similar density-dependent activation of MTA were reported in HeLa and lymphoma cells, although the changes are more moderate 28 . In addition, MTA is also activated by hypoxia via its metal response element 43 . Our results indicate that the transgene reporter is induced in parallel with the MTA gene by high cell density, likely through the MT enhancer and contributed by hypoxia. In contrast, the mRNA level of MTF-1 showed a strong decrease as cells reached medium density and remained low at high density, possibly due to negative feedback.
Next we examined the changes in Cyclin E (CycE), the cyclin that promotes G1-S transition and is overexpressed in many cancer cells 44,45 . We found that the CycE mRNA level reduces as cell density increases and www.nature.com/scientificreports www.nature.com/scientificreports/ possibly enter quiescence at high density. These changes are consistent with the function of CycE in growth regulation 46 . We also examined the level of Drosophila proapoptotic gene reaper (rpr), a FasC death factor receptor homolog [47][48][49] . It decreased during the initial phase of growth, and then remained unchanged during the latter phase of growth to high density. In Drosophila embryos, rpr can be activated by induced hyperplastic growth to restore normal cell density and tissue stoichiometry 50 . However, this was not observed at the high cell density we tested in S2 cells. We next examined Expanded (Ex) and Kibra, two proteins involved in the contact inhibition of cell proliferation. Ex and Kibra can signal to the Hippo pathway, leading to the inactivation of Yorkie (Yki) and the down regulation of growth and proliferation [51][52][53][54] . Both Ex and Kibra genes are also targets of Yorkie activation and therefore are expected to down-regulate at high cell density. This is indeed what we observed -both Ex and Kibra mRNA showed significant reduction at high cell density (Fig. 6C). Another important pathway related to cell density and proliferation is the cytoskeletal organization. Cancer cells have different cytoskeleton organization 55,56 . Mechanical forces in tissue and cellular environment have been known to act through integrin and actin network to regulate growth [57][58][59] . The alpha-smooth-muscle Actin (alpha-SMA) and the stress fibers it forms can regulate the Hippo pathway by promoting Yki/YAP nuclear translocation in mammals 52,60-62 . Cytoplasmic beta-and gamma-actin are also known to be down-regulated at high cell density in mouse fibroblasts, often concomitant with F-actin depolymerization 63 . We observed a continued reduction in the mRNA level of Actin 88F, the Drosophila alpha-SMA, at high cell density (Fig. 6D). We further examined the mRNA level of gelsolin, an actin-binding protein that controls the actin filament assembly and disassembly [64][65][66][67] . We found it to be also reduced as cell density increases (Fig. 6D). Down-regulation of gelsolin contributes to the disorganized cytoskeleton present in many cancer cells. Our results suggest that both the amount and the organization of the actin network are decreased when S2 cells reach post-confluence density. Taken together, we have examined changes in several cellular pathways related to cell density and growth regulation, hoping to identify transcriptional targets that correlate with the reporter induction both in timing and amplitude. Among the genes we tested, the MTA mRNA level responds to cell density in a parallel fashion as the GFP reporter. In comparison, markers of cell cycle regulation, apoptosis, contact inhibition and cytoskeletal organization showed significant but less dramatic changes.

Discussion
We have characterized a novel reporter system that responds to cell density in transient and stably transformed Drosophila S2 cells. We showed that the transgenic reporters are induced in a culture density-dependent and reporter-independent fashion. The rapid and reversible induction of the GFP or RFP transgene is reflected both in the mRNA level and in the mean fluorescence of the cells. These results suggest a quantitative signal(s) that correlates with cell density, and argue against a threshold or all-or-none response. Our results indicate that multiple components within the transgene DNA respond to the density-dependent signals to possibly synergistically induce the reporter expression. Among the elements we tested, the MT enhancer appears to mediate metal www.nature.com/scientificreports www.nature.com/scientificreports/ response, hypoxia response and high cell density response. Replacing MT with the 2PE enhancer eliminates the responses from the metal and hypoxia but not density-mediated reporter activation, suggesting that distinct signals contribute to reporter induction via these DNA sequences. We further show that the eve promoter and the CaSpeR vector sequences in the transgene also help to optimize the density-induced reporter activation. We found that the reporter induction correlates with changes in several growth-related pathways during density increase. These include pathways components involved in cell cycle regulation, apoptosis, contact inhibition and cytoskeleton organization. However, further work is needed to identify the specific signal(s) that directly trigger the reporter induction. Taken together, our analysis suggests that multiple cis-and trans-factors intercept with cell density and growth regulation signals, supporting the potential of developing versatile and specific reporters that respond to different regulators of cell proliferation. the role of the Mt enhancer in cell crowding and hypoxia mediated gene regulation. The metallothionein (MT) genes encode small cysteine-rich proteins that chelate divalent metal ions like Zn ++ , Cu ++ and Cd ++ 68-70 . They are found widely in eukaryotic species, often in multiple copies and expressed constitutively in most tissues and organs. They are involved in metal metabolism and detoxification, reactive oxygen species scavenging, stress response and neuronal growth regulation 71,72 . Besides responding to metals, certain isoforms of mammalian MT's (MT-1 and 2) are induced by glucocorticoids while others, such as MT-3, inhibit neurite outgrowth and neuronal survival, and are deficient in patients with Alzheimer's disease 72 . Therefore, the MT genes are likely to respond to diverse signaling inputs and perform distinct function in different tissues. The MT enhancer we used in this study is isolated from the regulatory region of the Drosophila MTA gene 9,11 . MTA does not directly correspond to a specific mammalian ortholog but it shares the basic structure and function characteristics of the MT proteins. Previous studies indicate that MTA could be induced by several signaling pathways through its enhancer/promoter sequences. Our finding that the MT enhancer responds to Cu ++ , hypoxia, and density-induced signals are also consistent with the previous observations. However, since significant level of reporter activation remains in the absence of the MT enhancer, additional cis-and trans-signals may also be involved in reporter induction. Further fragmentation or internal deletions may reveal sequences that specifically respond to signal from high cell density.
Hypoxia response is an important mechanism that can drastically alter cell physiology and gene expression profiles. It also affects cell fate decisions including growth and differentiation [73][74][75][76] . Our MG reporter responds to low oxygen partial pressure less dramatically than LDH, an in vivo hypoxia marker (Fig. 5B,C). But it responds to high cell density more dramatically than LDH, suggesting that although pericellular hypoxia exists under high cell density, it plays a minor role in reporter induction that other density-related signal [31][32][33][34] . This is consistent with the observation that the ΔMT transgene, which no longer respond to hypoxia, still respond to high cell density. Importantly, hypoxia is known to suppress protein translation [39][40][41][42] . Under severe hypoxia (1% P O2 ) bulk protein synthesis could reduce by over 75% in just two hours 39 . This down regulation is known to be independent of PI3/Akt and HIF-1alpha, but mediated through the repression of mTOR, which acts as an oxygen sensor 77,78 . Consistent with these findings, we observed only a 2-fold increase in the florescent protein level after hypoxia despite the strong increases in the mRNA level. The translation suppression can be alleviated by RNAi knockdown of elF2B-alpha and Tsc-2, two proteins responsible for inhibiting translation under hypoxia, is know to rescue GFP protein expression under hypoxia 39 .
Developing a cell-based screening platform for regulators of cell density signaling. Cell culture system provides powerful advantages that complement in vivo studies. The homogeneity of the cell populations allows consistent and uniform response, thus permitting sensitive and quantitative assessment both in biochemical and cell biological behaviors. The cell-based assays are amenable to chemical and molecular genetic treatment such as drug and small chemical screens, genome-wide RNAi and CRISPR-based screens for identifying genes and pathways regulating a variety of cellular function [79][80][81][82][83] . Transgene reporters can integrate into the genome both via the classic non-homologous insertion of long tandem arrays or through transposase-mediated single copy insertion in different cell clones 84,85 . As a result, these clones may response with different intensity and timing to various signaling events. In particular, the density reporter we reported here provide a foundation for screens for signaling components and regulators of cell cycle progression, apoptosis, contact inhibition of proliferation and cell migration, as well as cytoskeletal organization. The system could be customization at the level of transgene DNA composition, which allows selective response to cell-density and/or hypoxia differentially. Combination of reporter lines containing different transgenes allows multiplex screening for different density-responsive targets. RNAi knock down of elF2B-alpha and Tsc-2 may allow use in hypoxia -related screen. Additional reporters, such as luciferase, could be used for more quantitative readout. The system could also be tested in other cells. In addition to the S2 cells, we have also observed similar reporter response in the Drosophila Kc cells (data not shown). Both of these two cells lines are presumed to be mesoderm-derived macrophage-like cells. Although regulation of growth and proliferation occurs in all cell types, different signaling components may be employed to accomplish such regulation. Therefore the assay platform could be tested in additional tissue types to optimize searches for diverse growth-related cell behaviors such as contact inhibition. www.nature.com/scientificreports www.nature.com/scientificreports/ transfection, pTurbo plasmid containing the P-element transposase was mixed with the assay construction at a ratio of 1 to 10. The transfection cocktail was replaced with fresh medium after 5 hour of incubation. Cells were normally induced with 1 mM CuSO 4 24 hours after transfection. For polyclonal stably transfected cells, transgenic cells were passaged and frequently FACS enriched for eight months until population reach >95% transgenic.

Construction of DNA plasmids used in S2 cell reporter expression. The pCA-MT-eb-GFP (MG)
and pCA-MT-eb-RFP (MR) plasmids were described previously 9,10 . To generate the pCA-Delta-GFP plasmid, the MT element was removed through an EcoRI digestion and the resulting vector religated. To generate the pCA-2PE-GFP plasmid, the MG plasmid was digested with Bam H1 to remove the MT-eve promoter, and ligated to a Bam HI fragment containing 2PE-eve promoter fragment. For Adh and Hsp70 promoter swap plasmids, the 1.6-kb Hsp70 promoter and the 1.4-kb Adh promoter was PCR cloned and excited out from the pTOPO vector as an EcoR I-BamH I fragment and ligated MT and GFP region (see Supplemental Table 1 for PCR primers). For pTOPO (Invitrogen Cat# 451641) based constructs, the enhancer-promoter-reporter region was removed from CaSpeR-GFP constructs using Hind III and Pst I sites and ligated into the pTOPO vector (Invitrogen Cat# 451641).
Reporter induction and quantitation. For Cu ++ induction, CuSO4 is added to the cell media to a final concentration of 1 mM 24 hours prior to FACS analysis or mRNA preparation for qRT-PCR analysis. For crowding induction, cells were cultured for 5-8 days to reach 1.6 × 10^7/mL prior to FACS analysis or mRNA preparation for qRT-PCR analysis. Epifluorescence microscopy and flow cytometry analysis were done 24 hours after induction. Cell images were taken with a digital camera attached to a Zeiss Axioplan 2 fluorescence microscope. FACS was performed using a FACSCalibur flow cytometer (Becton Dickinson Immunocytometry Systems) on 2.5-5 × 10 4 cells each sample. For each sample, three or more biological replicates were performed. Data analysis was done using the FlowJo software. The fold of induction is calculated as (frequency × mean fluorescence) induced / (frequency × mean fluorescence) uninduced . For all bargraphs, the number of biological replicates (N) is indicted and the error bar represents standard error of the mean (SEM). P-values for the mean are calculated using the Student's t-test or ordinary one-way ANOVA multiple comparison in GraphPad Prism or JMP applications. The significance of the differences between means at 95% confidence interval is indicated as follow: four asterisks for P value smaller than 0.0001, three asterisks for P value greater than 0.0001 but smaller than 0.001, two asterisks for P value greater than 0.001 but smaller than 0.01, one asterisks for P value greater than 0.01 but smaller than 0.05, and ns (not significant) for P value greater than 0.05.
RnA preparation, cDnA preparation, Rt-pcR and qRt-pcR. 15569018) following the protocol therein. Multiplex RT-PCR reactions using gene-specific primers (see Supplemental Table 1 see primer sequences) were performed using isolated RNAs as templates. Gel electrophoresis was performed on a 2% agarose gel and the specific bands of the product was quantitated using a BioRad gel imager. For the rest of the manuscript, mRNA quantitation was performed using qRT-PCR. Briefly, cDNA library was generated using oligo-d(T) primer and the Superscript III reverse transcriptase Kit (Invitrogen). Quantitative PCR (qPCR) analysis was done to quantify gene expression using SYBR Green Supermix (Bio-Rad). For each sample, at least three biological replicates and two qPCR replicates for each biological replicates were performed. Sequence information for primers used in DNA cloning and RT-PCR is in Supplemental Table 1.