Zinc Detoxification: A Functional Genomics and Transcriptomics Analysis in Drosophila melanogaster Cultured Cells

Cells require some metals, such as zinc and manganese, but excess levels of these metals can be toxic. As a result, cells have evolved complex mechanisms for maintaining metal homeostasis and surviving metal intoxication. Here, we present the results of a large-scale functional genomic screen in Drosophila cultured cells for modifiers of zinc chloride toxicity, together with transcriptomics data for wild-type or genetically zinc-sensitized cells challenged with mild zinc chloride supplementation. Altogether, we identified 47 genes for which knockdown conferred sensitivity or resistance to toxic zinc or manganese chloride treatment, and >1800 putative zinc-responsive genes. Analysis of the ‘omics data points to the relevance of ion transporters, glutathione (GSH)-related factors, and conserved disease-associated genes in zinc detoxification. Specific genes identified in the zinc screen include orthologs of human disease-associated genes CTNS, PTPRN (also known as IA-2), and ATP13A2 (also known as PARK9). We show that knockdown of red dog mine (rdog; CG11897), a candidate zinc detoxification gene encoding an ABCC-type transporter family protein related to yeast cadmium factor (YCF1), confers sensitivity to zinc intoxication in cultured cells, and that rdog is transcriptionally upregulated in response to zinc stress. As there are many links between the biology of zinc and other metals and human health, the ‘omics data sets presented here provide a resource that will allow researchers to explore metal biology in the context of diverse health-relevant processes.


INTRODUCTION
Whereas metals such as mercury or cadmium are solely toxic to cells, other metals, such as zinc and manganese, are essential for cell viability and toxic only in excess.
Zinc is a structural component of many proteins and is also thought to act as a signaling molecule (FUKADA et al. 2011). Adding to the complexity of cellular zinc regulation, zinc is maintained at different levels in different organelles. Cells have evolved complex mechanisms for surviving zinc insufficiency, maintaining cellular and subcellular zinc homeostasis, and surviving exposure to toxic levels of zinc. The molecular mechanisms underlying regulation of zinc homeostasis and detoxification are in some cases zincspecific and in other cases, relevant to other metals. Methods used by cells to maintain zinc levels and/or survive metal toxicity include regulation of proteins required for metal influx (e.g. ZIP family importers of zinc), metal efflux (e.g. ZnT family exporters of zinc), or metal chelation (e.g. by metallothionines that chelate metals), as well as sequestration of zinc and/or biomolecules damaged by zinc in membrane-bound organelles such as the yeast vacuole or mammalian lysosome (KAMBE et al. 2015).
In addition to involving transporters and chelators, metal detoxification also involves more general detoxification strategies. Glutathione (GSH) has long been known to have the ability to form a complex with zinc or cadmium (PERRIN AND WATT 1971). Data from plants, yeasts, tunicates, fish, and other organisms suggest the relevance of GSH levels, conjugation, and/or transport to metal detoxification (PEREGO AND HOWELL 1997;PENNINCKX 2002;GHARIEB AND GADD 2004;FRANCHI et al. 2012;SETH et al. 2012).
Genetic evidence provides further support for a connection between GSH and metal detoxification. The ABCC-family transporter yeast cadmium factor (YCF1) is one example. YCF1 was originally identified based on a cadmium sensitivity phenotype and has been implicated in GSH-mediated detoxification of cadmium. The YCF1 protein is thought to be localized to the vacuole and to mediate transport of bis(glutathionato)cadmium into the vacuole (NAGY et al. 2006). Evidence suggests that the related protein YOR1, which is localized to the plasma membrane, can transport GSH-conjugated cadmium out of the cell. Consistent with this idea, ycf1, yor1 double mutant yeast strains are reportedly more sensitive to cadmium intoxication than either single mutant strain (NAGY et al. 2006).
The biology of zinc and other metals has many connections with human diseases. For example, genetic disruption of genes encoding metal transporters can lead to diseases of metal insufficiency or excess (CLAYTON 2017). In addition, accumulation of high levels of metals, which can occur following consumption of contaminated drinking water or through occupational exposure, can negatively impact human development and cause disease. Further, because metals are used by cells as 'weapons' in defense against pathogens (WEISS AND CARVER 2017), metal insufficiency can impact immune function.
Moreover, metals or metal-related genes have been implicated in, or levels correlated with, diseases such as diabetes, Parkinson's disease, and Alzheimer's disease (CHASAPIS et al. 2012). The zinc transporter ZNT8 is a common autoantigen in type 1 diabetes (ARVAN et al. 2012), for example, and ATP13A2 (PARK9), a Parkinson's disease gene, has been implicated in zinc homeostasis (KONG et al. 2014;PARK et al. 2014; TSUNEMI AND KRAINC 2014). Furthermore, adaptation of insect vectors of disease such as mosquitoes to metals might confer concomitant resistance to insecticides (POUPARDIN et al. 2008), such that understanding metal detoxification in insects might impact our understanding of disease vector control and the impact of polluted environments on the spread of insect-borne diseases (POUPARDIN et al. 2012).
Although yeast provides an excellent genetic platform for study of the cell biology of metal detoxification, using a single-celled organism has limited potential to model multicellular systems such as humans or insect vectors of disease. Drosophila presents many advantages as a genetic model system, including to study metal homeostasis and detoxification at the cellular and whole-organism levels; to study the effects of genetic or environmental perturbation of metal levels in the context models of human diseases (e.g. in Drosophila models of Parkinson's or Alzheimer's disease); and as a model of adaptation to metals and/or insecticides by insect vectors of disease. Work by several labs has established Drosophila as an in vivo model for study of zinc biology in a multicellular system (RICHARDS AND BURKE 2016; XIAO AND ZHOU 2016), as well as for evolutionary studies of metal-related genes (SADRAIE AND MISSIRLIS 2011;REMPOULAKIS et al. 2014). R. Burke and colleagues in particular performed a comprehensive genetic survey of in vivo ZIP and ZnT family zinc transporter functions using combined knockdown and over-expression approaches (LYE et al. 2012;LYE et al. 2013).
Moreover, R. Burke, B. Zhou, and others have established the fly gut as a system for the study of zinc and other metals (WANG et al. 2009;WANG AND ZHOU 2010;JONES et al. 2015). Studies in Drosophila have also identified a role for zinc in kidney stone disease (CHI et al. 2015), and Drosophila is an established model in which to study the effects of metal-containing nanomaterials (ALARABY et al. 2016).
Altogether, the existing literature suggests that Drosophila provides an excellent system in which to study the cellular and organismal biology of metal homeostasis and detoxification. Despite the growing body of work in Drosophila on zinc biology and other metal-related studies, however, there has remained a need for the application of highthroughput functional genomic methods to the study of zinc and other metals in Drosophila. Here, we describe the results obtained by applying two complementary 'omics approaches to the identification of genes relevant to metal homeostasis and detoxification. Specifically, we performed large-scale Drosophila cell-based RNAi screens to identify genes relevant to zinc or manganese detoxification, and performed a transcriptome-wide analysis of genes regulated in response to mild metal supplementation of wildtype or genetically zinc-sensitized cells. The results point to conserved genes and functions, and provide a resource for further study.

Cultured cell lines
The screen was performed using the DRSC isolate of the S2R+ Drosophila cell line.
Derivatives of this cell line newly generated in this work are available from the  Table).

Cell RNAi screening
We screened in total four dsRNA reagent libraries for Drosophila cell-based RNAi screening from our Drosophila RNAi Screening Center (DRSC) collection (HU et al. 2017): the TM library targeting genes encoding transmembrane domain-containing protein library (17 unique 384-well assay plates), AUTGY library targeting genes encoding autophagy-related factors (3 plates), MBO1 library targeting genes encoding proteins associated with membrane-bound organelles (2 plates), and a customdesigned plate with candidate metal-related factors we refer to as the "Megadeath" plate (1 plate). In all cases, experimental dsRNAs are excluded from the outermost two wells of the final 384-well assay plate design to limit edge effects. Three replicates of each unique plate in the library (metal-supplemented conditions) or two replicates of each unique plate (control) were screened. To perform the screens, we added S2R+ cultured cells to dsRNAs-containing assay plates as described previously (ECHEVERRI AND PERRIMON 2006). We then incubated the plates in a 25˚C incubator with humidity control for four days. Next, freshly prepared ZnCl 2 or MnCl 2 (Sigma Aldrich) in solution or a control treatment (water) was added to the assay plates using a Formulatrix Mantis liquid handling robot to a final level of supplementation of 15 mM. Twenty-four h following metal supplementation or control treatment, cells were lysed and total ATP levels per well were determined using Promega Cell Titer Glo and a Molecular Devices Spectramax Paradigm luminometer. The step-by-step screen and assay protocols we used are available online at <https://fgr.hms.harvard.edu/fly-cell-rnai-384-well-format> and <https://fgr.hms.harvard.edu/fly-cell-total-atp-readout>. Relative luciferase values for each plate were normalized to the plate average, replicates were averaged, and average normalized relative luciferase values were then compared across plates by calculating Z-scores.

Generation of CRISPR knockout cell lines
The sgRNA sequence used to target ZnT63C was TGTGACCAATTCGATGGCTC; the sgRNA sequence used to target IA-2 was CGGCTGTTCCGCGTGCTCTCTGG (see also Supplemental Reagents Table). The sgRNAs were cloned and introduced into cells as described in (HOUSDEN et al. 2014;HOUSDEN et al. 2015). Briefly, following introduction of Cas9 and sgRNAs targeting ZnT63C or IA-2 and single-cell isolation, we used high-resolution melt analysis (HRMA) to identify gene-modified cells. The ZnT63C or IA-2 gene regions from colonies positive by HRMA were then amplified by PCR, PCR products were individually cloned by TOPO cloning, at least 10 isolates were subjected to Sanger sequencing, and sequence data was aligned and analyzed to confirm that all alleles contained frameshift mutations.

RNA preparation and RNAseq analysis
For RNAseq analysis of control or metal-treated cells, wildtype S2R+ or CRISPR modified mutant cell derivatives (IA2-KO and ZnT63C-KO) were first grown to confluency in 10 mL of media in T-75 flasks; then control samples were left untreated and experimental samples were supplemented to a final concentration of 1 mM ZnCl 2 or MnCl 2 ; and cells were incubated for 24 h, centrifugated, and resuspended in TRIzol reagent (Thermo Fisher). RNA was extracted as described previously using chloroform extraction and isopropanol precipitation (SONG et al. 2010). Each final total RNA solution was divided into two aliquots, one of which was freshly used for RNA Integrity Number (RIN) evaluation at the Harvard Medical School Biopolymers Facility. If the RIN was >6.7 and no RNA degradation was observed following analysis, the other aliquot (stored at -80˚C) was shipped on dry ice to the Columbia Genome Center (Columbia University, New York, NY) for standard sample processing and raw data analysis. The raw data files were processed by the Columbia Genome Center. In order to compare the results at the gene level, we took the average of the FPKM values for each gene for the two replicates done for each condition, then determined the log 2 ratio of FPKM levels for each genotype and treatment condition combination vs. FPKM levels in the same genotype. Prior to this analysis, we set to a value of "1" any average FPKM value between 0 and 1 to reduce the possibility that we get large ratio values for genes with negligible levels of detected transcript in both the experimental sample and the wildtype control (e.g. the ratio of FPKM 0.1 vs. 0.0001), as we assume those ratios are not likely to have biological relevance. A cutoff of two-fold change for all replicates was applied (log 2 >1 or <-1).

qPCR analysis
Wildtype S2R+ cells were cultured under standard growth conditions in a 75 mL flask for 72 h. Next, the appropriate amount of a 1 M stock solution of ZnCl 2 or ZnSO 4 (or water) to reach a final supplementation concentration of 0, 1, 3, or 5 mM. Cells were then incubated for 24 h at 25˚C, centrifugated and resuspended in TRIzol (Thermo Fisher), treated with DNaseI, and cDNA was prepared using a QIAGEN RNeasy kit. The cDNA was analyzed by real-time quantitative PCR (qPCR) using SYBR Green and with tubulin as an internal reference gene control. All experiments were conducted in triplicate qPCR reactions. Data were analyzed with Bio-Rad CFX Manager software. Additional details regarding reagents and the sequence of oligonucleotide primer sequences used to detect experimental and control genes are indicated in the Supplemental Reagents Table.

Enrichment analysis of the RNAi screen and RNAseq datasets
The gene hits from RNAi screen and RNA-Seq profiling were analyzed for overrepresented gene sets using an in-house JAVA program based on hyper-geometric distribution. The gene sets we queried were assembled using gene ontology annotation, (<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE99332>).

Drosophila cell-based screen for modifiers of metal chloride toxicity
We reasoned that the application of high-throughput 'omics approaches in Drosophila cultured cells would provide a robust dataset that can inform future in vivo studies. To get a genome-scale view of genes relevant to zinc detoxification, we performed three large-scale RNA interference (RNAi) screens in parallel using Drosophila S2R+ cultured cells. We used total ATP levels as an indirect assay readout of cell viability and number, and screened under three conditions: control conditions, toxic zinc conditions (i.e. supplementation to a concentration of 15 mM ZnCl 2 ), and toxic manganese conditions (15 mM MnCl 2 ). Although our main goal was to identify genes relevant to zinc, we chose to screen in parallel under toxic manganese conditions for two reasons. First, we reasoned that screening with an equimolar concentration of a different metal chloride salt would help exclude the possibility that the genes we identify in the zinc intoxication screen are relevant to the chloride ion, changes in osmolarity, or other non-specific effects. Second, we reasoned that the data would provide a helpful high-throughput dataset relevant to the biology of manganese, a metal that, like zinc, is essential for cell viability and toxic in excess but underexplored in the literature.
In order to quickly identify high-confidence screen 'hits' (positive results) at the gene level, we screened RNAi reagent libraries from the Drosophila RNAi Screening Center (DRSC) collection that have multiple unique RNAi reagents per gene coverage and optimized layouts. Specifically, we screened four libraries of double-stranded RNAs (dsRNAs): a large library targeting known or predicted Drosophila transmembrane domain-containing protein-encoding genes and three smaller libraries targeting genes encoding putative autophagy-related factors (ZnCl 2 conditions only), conserved components of membrane-bound organelles, or candidate metal-related factors (total of 3885 dsRNAs targeting 1960 unique genes). Each screen condition was internally normalized (see Materials and Methods). How to access screen data is described in the data availability section of the Materials and Methods.
Several indicators suggest that the screens resulted in high-quality, metal-specific results. First, consistent with identification of condition-specific results, replicate assay plates within a treatment condition correlate well (Pearson correlation coefficients of 0.88 -0.93) but replicate assay plates from different treatment groups do not (Pearson correlation coefficients of 0.24 -0.50). Second, for many genes, two unique reagents targeting the same gene scored as hits under the same conditions and in the same direction. We define this set genes for which two unique reagents scored in the direction in a given condition as the set of 'high-confidence hits' (Fig. 1 and Supplemental Tables 1, 2, and 3). Third, the set of high-confidence hits in any given condition and direction include multiple members of a protein family (e.g. several tetraspanin family proteins are high-confidence hits in the same direction in the manganese screen), multiple members of a protein complex (e.g. several nuclear pore components are high-confidence hits in the same direction in the zinc screen), or multiple components of a given organelle (e.g. for both control and manganese conditions, mitochondrial proteins are among the highconfidence hits with lower ATP levels vs. the internal controls; Fig. 1, blue text).
Altogether, we identified 30 high-confidence hits in the control treatment group, 29 in the zinc toxicity treatment group, and 36 in the manganese toxicity treatment group ( Fig. 1 and Supplemental tables 1, 2, and 3). There is limited overlap among genes that scored as hits from each of the three screens. Three genes (CG5805,Ctr1A,RFeSP) are in common between screen hits for control conditions, decreased ATP ("down") direction and MnCl 2 conditions, down direction (Fig. 1A). Two genes (alphaCOP, hay) are in common between screen hits for control conditions, increased ATP ("up") direction and  1B). For the zinc toxicity screen, there was a clear bias in detection of highconfidence hits conferring higher ATP levels (28 high-confidence hits) as compared with lower ATP levels (1 high-confidence hit). We suspect that the zinc treatment conditions were so toxic it was difficult to detect a significant further reduction in total ATP levels vs. the internal control. In addition, some genes identified in the screens are 'frequent hitters' (Fig. 1, gray or green text), which we define here as hits scoring in >50% of Drosophila RNAi screens in the GenomeRNAi database (SCHMIDT et al. 2013).

Analysis of zinc screen results
We found a number of zinc-related genes among the high-confidence hits in the zinc screen. For example, fear of intimacy (foi), which encodes a ZIP family zinc influx protein (MATHEWS et al. 2005), was identified as a high-confidence hit for increased total ATP levels ("up" hits) in the zinc screen but not in the other screens (Table 3). The zinc screen hits in the up direction also include CG32000, a putative ortholog of human ATP13A2 (PARK9); evidence from mammalian cells suggests a role for ATP13A2 in lysosomal zinc transport (KONG et al. 2014;PARK et al. 2014;TSUNEMI AND KRAINC 2014). The zinc screen hits in this direction further include IA-2 protein tyrosine phosphatase (IA-2), an ortholog of human PTPRN (better known as IA2), which, like the human zinc transporter ZNT8, is a common autoantigen associated with type 1 diabetes (ARVAN et al. 2012).
The single high-confidence hit in the lower ATP levels direction in the zinc screen (i.e. knockdown appears to confer zinc sensitivity; "down" hits) is a gene with the systematic name CG11897 that we rename red dog mine (rdog) after the large Alaskan zinc mine of that name. The rdog gene encodes a member of the ABCC sub-family of ABC transporters that also includes yeast YCF1. Five additional genes were initially categorized as low-confidence hits in the 'zinc treatment, lower ATP levels' (down) category: CG7627, CG3790, COX7AL, CR43469, and mthl3 (Supplemental file 1).
Notably, like rdog, CG7627 also encodes an ABCC family member, further supporting the idea that ABCC-type transporters are relevant to zinc chloride detoxification. Based on enrichment analysis (below), CG7627 was promoted to a 'moderate' confidence hit (Supplemental file 1).
We performed enrichment analysis for gene ontology (GO) 'cellular compartment' or 'biological process' terms; pathways as annotated in Reactome (CROFT et al. 2011); and protein complexes as annotated by COMPLEAT (VINAYAGAM et al. 2013). In all cases, we used the full set of hits in the analysis (i.e. both low-and high-confidence hits). We re-annotated low-confidence hits as 'moderate' confidence if they were members of a significant enrichment group (Supplemental file 1). Overall, the results of the enrichment analysis further suggest the quality and specificity of the screens. For the zinc toxicity screen, enrichment is driven by the presence of multiple components of the nuclear pore complex, suggesting the possible involvement of nuclear transport in zinc-induced cell death or another relevant process. We also note that there are related genes in the human genome for all of the high-confidence hits identified in the zinc screen (Supplemental Table 2).

Transcriptomics analysis of wildtype and zinc-sensitized cells
The use of genetically zinc-sensitized strains of Drosophila has helped uncover mechanisms of zinc homeostasis in vivo (LYE et al. 2012;LYE et al. 2013). We reasoned that production of mutant cell lines lacking activity of the zinc exporter ZnT63C would similarly result in cultured cells genetically sensitized to zinc supplementation and allow for detection of zinc-related genes. Such an approach would allow us to capitalize on the advantages of performing transcriptomics studies in a relatively homogenous cultured cell line, allowing for robust detection of down-or up-regulated genes, as well as minimize the need to treat cells with high ionic strength solutions. We used a CRISPR-Cas9 strategy to target ZnT63C, which encodes a zinc efflux protein, and also targeted IA-2, which encodes the Drosophila ortholog of human PTPRN/IA2, which, like the human zinc transporter ZNT8, is a common autoantigen in type 1 diabetes (ARVAN  Table 1. We observed a larger overall transcriptional response in ZnT63C-KO cells treated with ZnCl 2 than in the other genotypes and conditions (i.e. a larger total number of genes with significant log 2 values as compared to the control and a large fold-change among the top hits), consistent with the idea that ZnT63C-KO are more sensitive to zinc supplementation than wildtype S2R+ cells. Supplemental file 2 includes a list of the genes the data suggest are down-or up-regulated in each genotype and condition. How to access the RNAseq data is outlined in the data availability section of the Materials and Methods.
As summarized in Table 1, we identified >1800 putative zinc-responsive genes. Several indicators suggest that the strategy of genetically sensitizing cells to zinc supplementation was successful in identifying high-confidence zinc-responsive genes, including the observation that several genes down-or up-regulated in two or more zincsupplemented S2R+, IA2-KO, and/or ZnT63C-KO genotypes (Fig. 2). Not surprisingly, the set of zinc-responsive genes include metallothionines, which act as metal chelators, as well as a number of heat shock proteins (Fig. 2B). Interestingly, at least for some Table 4, Supplemental file 2). As compared with zinc, supplementation with low levels of MnCl 2 of any genotype resulted in a small number of genes showing significant down-or upregulation as compared with wildtype untreated cells (see Table 1, Fig. 3 and Supplemental file 2), consistent with the idea that knockout of ZnT63C sensitizes cells to zinc but not manganese. Despite the smaller numbers, however, some genes are in common between the zinc and manganese datasets ( Fig. 2 and 3, green text, and Supplemental file 2), suggesting that these genes might be chloride-responsive or general factors.

Enrichment analysis of transcriptomics data from zinc-treated cells
As for the RNAi screen data, we performed enrichment analysis of the transcriptomics data to detect GO terms, molecular pathways, or protein complexes significantly enriched in these datasets (Supplemental file 2). Enrichment among genes downregulated in response to ZnCl 2 supplementation includes processes or protein complexes related to ribosomes (e.g. Reactome "translation," p-value 6.88 x 10 -15 in zinc-treated wildtype cells), and pathways or cellular components of respiratory electron transport. Enrichment among genes up-regulated in response to ZnCl 2 supplementation includes processes or complexes involving heat-shock proteins and starvin (stv), an ortholog of the human "BCL2 associated athanogene 3" or BAG3 gene. Significant enrichment was also seen in zinc-treated wildtype cells for the GO molecular function heme-copper terminal oxidase activity; the 3 of 21 genes in this group that result in enrichment are COX4L and COX7AL, which encode subunits of cytochrome c oxidase, and CG42376, which encodes an ortholog of human "cytochrome c oxidase assembly factor 6" or COA6. For zinc-treated ZnT63C-KO cells, up-regulated genes are also enriched for glutathione-related activities (e.g. KEGG "glutathione metabolism," p-value 3.62 x 10 -9 in ZnT63-KO cells). Indeed, 9 of 11 GstD sub-family genes, other glutathione S-transferase genes, and Glutamate-cysteine ligase catalytic subunit (Gclc), which is a rate-limiting enzyme in the glutathione synthesis pathway, are up-regulated in zinctreated ZnT63C-KO cells.
Altogether, the transcriptomics data show that zinc stress results in up-regulation of metal chelators and heat-shock proteins, and suggests that zinc stress has specific impacts on mitochondrial function that elicit compensatory transcriptional responses. 1 Moreover, the data obtained using a zinc-sensitized genotype suggest that under high zinc stress conditions, there is a significant need for conjugation of substrates to GSH. This is consistent with a recent report that glutathione S-transferase activity is relevant to methyl mercury toxicity in Drosophila (VOROJEIKINA et al. 2017) and with results obtained for other species that associate metal detoxification with GSH conjugation and/or flux (PEREGO AND HOWELL 1997;PENNINCKX 2002;GHARIEB AND GADD 2004;NAGY et al. 2006;FRANCHI et al. 2012;SETH et al. 2012).

Comparison of functional screen and transcriptomics data
We reasoned that genes encoding proteins normally involved in zinc influx would be expected to score in the 'up' direction in the screen (higher ATP values as compared with the internal control, consistent with resistance to zinc treatment) and 'down' in response to zinc supplementation in the transcriptomics analysis. Consistent with this, we found that the high-confidence screen hit foi was down-regulated in zinc-treated S2R+ and zinc-treated IA2-KO as compared with untreated cells of the same genotype, and down-regulated in zinc-treated ZnT63C-KO cells as compared with wildtype untreated cells. In addition, CG32000 was down-regulated in zinc-treated ZnT63C-KO cells as compared with wildtype untreated cells. We also compared the data for components of the nuclear pore. The low-confidence screen hit Nup107 was downregulated in zinc-treated wildtype and zinc-treated ZnT63C cells as compared with genotype controls, and in zinc-treated IA2-KO cells as compared with untreated cells of the same genotype. In addition, the high-confidence hit Nup93-1 was down-regulated in zinc-treated ZnT63C-KO cells as compared with the wildtype untreated control, and the additional nuclear pore component-encoding genes Nup43,Nup44A,Nup50,Nup54, and Nup160 were down-regulated in ZnT63C-KO cells as compared with untreated cells of the same genotype.
We next explored the converse prediction: that genes encoding proteins protective against zinc intoxication would be expected to score in the 'down' direction in the screen and to be up-regulated in response to zinc supplementation. Despite the relatively small number of 'down' direction hits in the zinc screen ( Fig. 1 and Supplemental File 1), we did find overlap between the RNAi 'down' direction hits and zinc-responsive gene lists.
The low-confidence hit COX7AL was significantly up-regulated in zinc-treated S2R+ and zinc-treated ZnT63C-KO cells as compared with untreated genotype controls, and in all three as compared with untreated wildtype cells. In addition, the one high-confidence 'down' direction RNAi screen hit, rdog, scored as significantly up-regulated in zinctreated ZnT63C-KO cells; the log 2 values for rdog were 1.21 for zinc-treated IA2-KO cells and 3.53 for zinc-treated ZnT63C-KO cells as compared with genotype controls.

rdog is up-regulated in response to zinc in Drosophila S2R+ cells
We further confirmed that rdog is up-regulated in response to zinc supplementation using a graded series of ZnCl 2 to supplement the culture media of wildtype S2R+ cells followed by quantitative real-time PCR (qPCR), as shown in Fig. 3. As expected, under Based on identification of rdog in the ZnCl 2 but not MnCl 2 -supplemented screen and transcriptomics datasets, we suspected that the effect is zinc-specific, rather than being attributable to the chloride ion. To further test this experimentally, we performed qPCR analysis on wildtype S2R+ cells supplemented with ZnSO 4 . The trends for control and rdog transcript levels are similar (Fig. 3B), demonstrating that rdog expression is upregulated by zinc in Drosophila S2R+ cells.

Analysis of parallel studies using manganese chloride
As mentioned, we performed the RNAi screens and transcriptomics studies with MnCl 2 in parallel to help distinguish zinc-specific factors from general factors, and to provide an additional metal intoxication-related data resource. For the MnCl 2 toxicity screen, enrichment among genes conferring higher ATP levels upon knockdown is driven by the presence of multiple components of the vacuolar H+ ATP transport machinery ( Fig. 1B and Supplemental Table 2). This suggests the possible relevance of proton transport to manganese-induced cell death or another related process. Several tetraspanin family proteins are also hits in the manganese screen. This is intriguing, as three tetraspanin family proteins were detected as co-regulated by trans-eQTLs following feeding of flies with lead (Pb) (RUDEN et al. 2009), suggesting the possibility of a general role for tetraspanin family proteins in detection or responses to metals or metal-induced stress.
Enrichment analysis of genes conferring lower ATP levels in the MnCl 2 screen points to the relevance of mitochondria. With regards to transcriptomics analysis, we found that the results for MnCl 2 -treated samples were more useful as a comparison set for ZnCl 2 treatment than for detection of Mn-specific factors. Overall, the fold-change values were modest and a relatively small number of genes surpassed the cutoff values; for example, only 23 genes were down-regulated and 33 up-regulated in MnCl 2 -treated wildtype cells (Table 2). Nevertheless, overlap between these and the zinc treatment group was observed ( Fig. 2 and 3, Supplemental file 2), suggesting that the manganese treatment group provides a useful filter for further refinement of a list of candidate zincrelated factors.

Concluding remarks
The functional genomics and transcriptomics data sets described here provide a genome-scale resource for the study of zinc biology in Drosophila. Despite the fact that we performed the RNAi screens under high metal supplementation conditions, we were able to identify factors known to be relevant to zinc homeostasis at physiological levels (i.e. foi and CG32000). This is consistent with known overlap between metal homeostasis and detoxification genes, and suggests the validity of the approach. In addition, despite assay bias, we were able to detect one high-confidence gene, rdog, for which knockdown of results in lower ATP values as compared with the internal control.
We further found that rdog, an ortholog of yeast YCF1, is up-regulated in genetically zinc-sensitized cells following mild zinc supplementation. Identification of rdog in the cell-based screen, as well as identification in the transcriptomics data of rdog, Gclc, and genes encoding glutathione S-transferase family proteins (SAISAWANG et al. 2012), supports the idea that GSH is relevant to zinc detoxification in Drosophila. Moreover, the observation that another gene encoding ABCC family member, CG7627, was also identified as a zinc sensitivity screen hit in this work, together with the fact that a third Drosophila ABCC family member, dMRP, has previously been implicated in methylmercury toxicity (PRINCE et al. 2014), suggest a general role for ABCC-type transporters in metal detoxification in Drosophila. Altogether, we expect that the 'omics data presented here will guide further research into the mechanisms underlying metal homeostasis and detoxification in Drosophila and other systems. For example, the data provide a focused set of candidates for in vivo analyses of wildtype and genetically metal-sensitized flies under normal or metal-supplemented conditions, as well as for in vivo analyses in fly models of human diseases such as diabetes or neurodegeneration.      hits as shown here, each of the designs resulted in a Z-score >1.5 or <-1.5 and in the same direction as what was found for other designs targeting the same gene. 2 Down, decreased total ATP levels following plate-based normalization within a treatment group; up, increased total ATP levels following plate-based normalization within a treatment group.

ACKNOWLEDGMENTS
2 Number of unique dsRNAs in the library that target the gene. For all high-confidence hits as shown here, all of the designs resulted in a 'hit' as defined by a Z-score >1.5 or <-1.5 and in the same direction as what was found for other designs targeting the same gene. 2 Down, decreased total ATP levels following plate-based normalization within a treatment group; up, increased total ATP levels following plate-based normalization within a treatment group. 1 2 Number of unique dsRNAs in the library that target the gene. For all high-confidence hits as shown here, all of the designs resulted in a 'hit' as defined by a Z-score >1.5 or <-1.5 and in the same direction as what was found for other designs targeting the same gene. --, results not significant in this genotype.