Systematic genetic mapping of necroptosis identifies SLC39A7 as modulator of death receptor trafficking

Regulation of cell and tissue homeostasis by programmed cell death is a fundamental process with wide physiological and pathological implications. The advent of scalable somatic cell genetic technologies creates the opportunity to functionally map such essential pathways, thereby identifying potential disease-relevant components. We investigated the genetic basis underlying necroptotic cell death by performing a complementary set of loss-of-function and gain-of-function genetic screens. To this end, we established FADD-deficient haploid human KBM7 cells, which specifically and efficiently undergo necroptosis after a single treatment with either TNFα or the SMAC mimetic compound birinapant. A series of unbiased gene-trap screens identified key signaling mediators, such as TNFR1, RIPK1, RIPK3, and MLKL. Among the novel components, we focused on the zinc transporter SLC39A7, whose knock-out led to necroptosis resistance by affecting TNF receptor surface levels. Orthogonal, solute carrier (SLC)-focused CRISPR/Cas9-based genetic screens revealed the exquisite specificity of SLC39A7, among ~400 SLC genes, for TNFR1-mediated and FAS-mediated but not TRAIL-R1-mediated responses. Mechanistically, we demonstrate that loss of SLC39A7 resulted in augmented ER stress and impaired receptor trafficking, thereby globally affecting downstream signaling. The newly established cellular model also allowed genome-wide gain-of-function screening for genes conferring resistance to necroptosis via the CRISPR/Cas9-based synergistic activation mediator approach. Among these, we found cIAP1 and cIAP2, and characterized the role of TNIP1, which prevented pathway activation in a ubiquitin-binding dependent manner. Altogether, the gain-of-function and loss-of-function screens described here provide a global genetic chart of the molecular factors involved in necroptosis and death receptor signaling, prompting further investigation of their individual contribution and potential role in pathological conditions.


Introduction
Regulated cell death programs are crucial for homeostasis in multicellular organisms by eliminating cells that have become obsolete, damaged or infected [1]. Cell death signaling can be induced extrinsically by death receptors of the TNF receptor superfamily, including tumor necrosis factor receptor superfamily member 1A (TNFR1), Fas cell surface Edited by S. Nagata These authors contributed equally: Astrid Fauster, Manuele Rebsamen * Manuele Rebsamen mrebsamen@cemm.oeaw.ac.at * Giulio Superti-Furga gsuperti@cemm.oeaw.ac.at death receptor (FAS), TNF-related apoptosis-inducing ligand receptor (TRAIL-R)1 and 2 [2]. TNFR1 stimulation can lead to NF-κB activation and survival, apoptosis, or necroptosis depending on the composition of the signal transduction complexes formed upon ligand binding [2]. Necroptosis can be triggered by different death and immune receptors in response to pathogen infection or in the context of sterile inflammation [2][3][4], and relies on activation of receptor-interacting serine/threonine-protein kinase (RIPK) 3 [5][6][7] and its substrate mixed lineage kinase domain-like protein (MLKL) [8,9], which mediates membrane rupture. Death receptor ligation triggers RIPK3 activation through RIPK1, in conditions where apoptosis is blocked [5,10,11]. Evidence for the involvement of necroptosis in different pathologies [4,[12][13][14][15], such as ischemia/reperfusionmediated injuries [16], has accumulated over the past years. While exacerbating these inflammatory conditions, necroptosis can be beneficial in other pathological contexts, particularly in restraining viral and bacterial infections [4,17]. Consequently, detailed insight into the molecular framework of necroptotic cell death entails the appealing prospect of therapeutic benefit [12,13]. Forward genetics with somatic cells constitutes a powerful, unbiased approach to unravel the genetic basis underlying fundamental biological processes. RNA interference approaches led to the identification of the core necroptosis pathway members RIPK3 and MLKL [5,6,9]. By allowing efficient generation of full knockouts through insertional mutagenesis, the near-haploid KBM7 cell line has further empowered screening approaches in human cells [18,19]. The scope of genetic screening has been broadened by the advent of CRISPR/Cas9 technology and its adaptation to gain-of-function screening modes, such as the development of synergistic activation mediator (SAM) libraries mediating transcriptional activation of endogenous genes [20][21][22]. In this study, we combine these technologies to investigate the genetic foundation of TNFα-induced necroptosis and provide a comprehensive mapping of the molecular factors controlling necroptosis signaling. We characterize the specific contributions of the zinc transporter SLC39A7 by demonstrating its requirement for death receptor trafficking, thereby affecting all aspects of TNFR1 signaling, and of the ubiquitin-engaging protein TNIP1 on necroptosis pathway activation.

Results
A KBM7 FADDcell line undergoes necroptosis upon treatment with TNFα or the SMAC mimetic birinapant We set out to map the genetic requirements for necroptosis signaling in human cells, employing the haploid myeloid leukemia KBM7 cell line [18,19]. In contrast to the related HAP1 cell line that lacks RIPK3 expression [23], KBM7 undergo necroptosis upon treatment with TNFα, the SMAC mimetic birinapant [24] and the pan-caspase inhibitor z-VAD-FMK (Fig. 1a, Supplementary Figure 1a). As apoptosis inhibition is required for death receptorinduced necroptosis [25], we genetically abrogated the extrinsic apoptosis pathway by deleting the signaling adapter Fas associated via death domain (FADD) by CRISPR/Cas9 gene editing ( Supplementary Figure 1b-c). After enrichment for resistance to FASL-induced and TRAIL-induced apoptosis, we selected a knockout clone carrying a >100 bp insertion in the sgRNA target site, abrogating FADD expression (Supplementary Figure 1c-e). As expected, absence of FADD did not affect TNFαinduced NF-κB activation (Supplementary Figure 1f). Necroptosis could be induced in KBM7 FADDcells using TNFα or a combination of TNFα and SMAC mimetic without requirement for caspase inhibition, and was blocked by the RIPK1 inhibitor Nec1-s [26] or by blocking MLKL using necrosulfonamide (NSA) [8], whereas z-VAD-FMK had no effect (Fig. 1a, b). Indeed, treatment with TNFα and SMAC mimetic triggered rapid phosphorylation of MLKL in KBM7 FADDcells, whereas it induced apoptosis in KBM7 wildtype cells, as evidenced by Caspase-3 cleavage (Supplementary Figure 1g). Interestingly, treatment with the SMAC mimetic birinapant alone sufficed to induce necroptosis in KBM7 FADDcells (Fig. 1a). In the following, these newly established FADD-deficient KMB7 cells were used to interrogate the genetic basis of necroptotic cell death by employing complementary forward genetics approaches (Supplementary Figure 1h).

Haploid genetic screens in KBM7 FADDcells identify the requirements for necroptosis
In order to identify genes required for necroptosis signaling by haploid genetic screening, KBM7 FADDcells were mutagenized with a retroviral gene-trap vector [18,19] and selected with a high dose of the SMAC mimetic birinapant, TNFα, or a combination thereof. Each of these screens resulted in significant (p-value < 10 -10 ) enrichment of disruptive insertions in 10-13 different genes compared to unselected mutagenized cells (Fig. 2a-c, Supplementary Table 1). The screens identified TNFRSF1A, RIPK1, RIPK3, and MLKL among the top hits with a high number of independent insertions, consistent with their wellestablished role in TNFα-induced necroptosis signaling and a recent loss-of-function screen in murine cells [27] (Fig. 2d, Supplementary Figure 2a,b). Interestingly, alongwith these known necroptosis effector proteins, the zinc transporter SLC39A7 scored among the most significant hits in all screens, while other genes significantly enriched in selected conditions, such as Tumor necrosis factor receptor superfamily member 1B (TNFRSF1B) and Sp1 (SP1) with the SMAC mimetic birinapant (Fig. 2d,  Supplementary Figure 2c). For follow-up analyses, we focused on the highly or selectively enriched genes, employing a CRISPR/Cas9-based multi-color competition assay (MCA) co-culture system for validation (Supplementary Figure 2d). Both SMAC mimetic as well as TNFα treatment strongly selected for GFP + sgSLC39A7 − harboring cells over control mCherry + cells harboring sgRen (targeting Renilla luciferase), indicating that loss of SLC39A7 conferred enhanced cell survival or outgrowth under necroptosis-inducing conditions (Fig. 2e). Among the other genes tested, we confirmed the selective advantage upon treatment with the SMAC mimetic birinapant of cells harboring sgRNAs targeting SP1, TNFRSF1B, and, to a lesser extent, PU.1 (SPI1) and Ragulator complex protein LAMTOR1 (LAMTOR1) (Fig. 2f, Supplementary Figure 2e).
Loss of SLC39A7 mediates resistance to TNFαinduced cell death by diminishing TNFR1 surface expression Next, we investigated how loss of SLC39A7 impacts on TNFα signaling, given that the proposed roles for this ER-resident zinc transporter did not readily explain its link to the necroptosis phenotype [28][29][30][31][32]. We isolated a KBM7 FADDclone carrying a 5 bp deletion in the SLC39A7 coding sequence, leading to a premature stop codon and loss of protein expression (Supplementary Figure 3a). SLC39A7 has been reported as an essential gene in a number of cell lines of different tissue origin [33,34], but does not form part of the core essentialome in KBM7 cells [35]. Yet, loss of SLC39A7 conferred a significant growth disadvantage as compared to sgRen control cells (Fig. 3a). In agreement with our screening results, SLC39A7 knockout cells were protected from TNFα and SMAC mimetic-induced necroptosis (Fig. 3b). Strikingly, SLC39A7cells failed to activate the canonical NF-κB pathway following TNFα stimulation, and this correlated with the loss of TNFR1 surface expression (Fig. 3c, d). We confirmed that SLC39A7 localizes to the ER [28] (Supplementary Figure 3b-c) and, to assess the cellular alterations resulting from its loss, we determined changes in protein levels by employing a proteomics approach focused on membrane compartments (Supplementary Figure 3d-e). Gene ontology (GO) term enrichment among the proteins significantly upregulated in SLC39A7cells and gene set enrichment analysis (GSEA) among all significantly changed targets respectively identified "response to endoplasmic reticulum stress" and the hallmark gene set "Unfolded_protein_response" (UPR) as top hits (Supplementary Figure 3f-h, Supplementary Table 2). These data strongly indicated that, similar to its Drosophila and murine orthologues [30,31,36], loss of human SLC39A7 impacts on ER homeostasis. Indeed, SLC39A7cells displayed ER stress at basal state, reflected in increased levels of BiP (ER chaperone 78 kDa glucose-regulated protein GRP), and exhibited higher sensitivity towards different chemical ER stressors as monitored by induction of CHOP (DNA damage-inducible transcript 3 protein DDIT3) (Fig. 3e). Given the crucial role of the ER in the synthesis, folding and shuttling of membrane proteins [37], we next explored how loss of SLC39A7 impinged on TNFR1 TNFα and 1 µM SMAC mimetic, and the indicated inhibitors. Cell viability was assessed using a luminescence-based readout for ATP (CellTiter Glo) and normalized to untreated control. Data represent mean value ± s.d. of two independent experiments performed in triplicates trafficking. Surprisingly, while TNFR1 was not detectable on the surface of SLC39A7cells (Fig. 3d), we found higher levels of intracellular TNFR1 protein in SLC39A7 knockouts (Fig. 3f). Using glycan maturation as readout for glycoprotein movement through the secretory pathway [38], we found that TNFR1 accumulating in SLC39A7cells exhibited sensitivity to Endo H, indicative of retained ER localization (Fig. 3g). Haploid genetic screens in KBM7 FADDcells identify genes required for necroptosis. a-c Circos plots of haploid genetic screens in KBM7 FADDcells with necroptosis induction by 10 µM SMAC mimetic birinapant (a) 100 ng/ml TNFα (b) and 1 µM SMAC mimetic and 100 ng/ml TNFα combined (c). Each dot represents a mutagenized gene identified in the resistant cell population, dot size corresponds to the number of independent insertions identified for each gene and distance from center indicates the significance of enrichment compared to an unselected control data set. Hits with an adjusted p-value < 10 -10 are labeled. d Bubble plot depicting the top hits over all three screens ranked according to adjusted p-value in the TNFα screen. Bubble size corresponds to the number of independent insertions identified and color gradient reflects the significance of enrichment. e, f Multi-color competition assay (MCA) of KBM7 FADD -SpCas9 cells transduced with a GFP marker (GFP + ) and sgRNAs targeting either SLC39A7 or RIPK1 (e), SP1 or TNFR2 (f), or Renilla luciferase (sgRen) as control, against cells transduced with sgRen and an mCherry marker (mCherry + ). The cell populations were mixed at 1:1 ratio, treated with SMAC mimetic (1 µM) or TNFα (10 ng/ml), and analyzed after 14 days by flow cytometry. Data represent mean value ± s.d. of two independent experiments performed in duplicates, n.d. (not determined) indicates wells with no outgrowth Orthogonal genetic screens and surface marker analysis define the specificity of SLC39A7 on receptor trafficking To study the specificity of the SLC39A7 loss on death receptor trafficking, we devised a set of targeted and genome-wide screens including the other prominent TNF receptor superfamily members, FAS and TRAIL-R1/2. The use of a focused CRISPR/Cas9 library targeting 388 members of the SLC family allowed us to efficiently screen various conditions and stimuli in multiple cell lines. KBM7 FADD -, KBM7 wildtype and HAP1 cells were lentivirally transduced with the library and subsequently selected for resistance towards cell death induction by TNFα, SMAC mimetic, TRAIL, or FASL (Supplementary Figure 4a). sgRNAs targeting SLC39A7 were top-enriched in screens with TNFα, SMAC mimetic, and, interestingly, FASL (Fig. 4a, Supplementary Table 4 and 5), highlighting the exquisite role of SLC39A7 among SLCs in affecting these signaling pathways. Intriguingly, loss of SLC39A7 did not confer a selective advantage when TRAIL was used to induce apoptotic cell death in KBM7 wildtype and HAP1 cells (  Table 1). In line with the screening results, treatment with FASL selected for GFP + sgSLC39A7-harboring cells, whereas we found no differential outgrowth in TRAIL-selected conditions in both Jurkat E6.1 and KBM7 cells (Fig. 4b, Supplementary Figure 4d). Similar to TNFR1, surface expression of FAS and TRAIL-R2 was reduced in SLC39A7 knockout cells, while, in contrast, TRAIL-R1 was still detectable and even slightly increased (Fig. 4c, Supplementary Figure 4e). Accordingly, total and, more importantly, Endo H-resistant levels of TRAIL-R1 showed a comparable increase in SLC39A7 knockout cells, differing therefore from the fully Endo H-sensitive accumulation of TNFR1 in these cells ( Supplementary Figure 4f). While explaining the retained sensitivity to TRAIL- Fig. 3 Loss of SLC39A7 mediates resistance to TNFα-induced cell death by diminishing TNFR1 surface expression. a Growth curve for KBM7 FADD -sgRen or SLC39A7cells. Equal cell numbers were seeded and cell growth monitored over 7 days. Data represent mean value ± s.d. of five independent experiments; statistical analysis by ttest, *P < 0.01. b Cell viability in KBM7 FADD -sgRen or SLC39A7cells treated overnight (16 h) with TNFα (10 ng/ml), SMAC mimetic (0.5 µM), and Nec-1s (50 µM) as indicated. Cell viability was assessed using a luminescence-based readout for ATP (CellTiter Glo). Data represent mean value ± s.d. of two independent experiments performed in triplicates. c KBM7 FADD -SpCas9 (empty, sgRen or SLC39A7 -) cells were stimulated for the time indicated with TNFα (10 ng/ml). Cells were then lysed and subjected to immunoblotting with the indicated antibodies. d Flow cytometry analysis for TNFR1 surface expression. KBM7 TNFRSF1Acells serve as negative control for background staining. Data shown are representative of two independent experiments. e KBM7 FADD -SpCas9 (empty, sgRen or SLC39A7 -) cells were treated for 7 h with Brefeldin A (0.5 µM), Tunicamycin (2 µM), Thapsigargin (0.5 µM), MG-132 (10 µM) or DMSO as control. Cells were then lysed and subjected to immunoblotting with the indicated antibodies. f KBM7 TNFRSF1A -, KBM7 FADDand KBM7 FADD -SLC39A7cells were lysed and subjected to immunoblotting with the indicated antibodies. g KBM7 FADD -SpCas9 sgRen or SLC39A7cell lysates were incubated for 1 h at 37°C in presence or absence of PNGaseF or EndoH, respectively, and analysed by immunoblot with the indicated antibodies. Immunoblots shown are representative of two independent experiments. Asterisk (*) indicates non-specific band induced cell death, the dichotomy between these two closely related receptors was surprising. Consequently, we monitored a wider panel of cell surface proteins present on KBM7 cells (Supplementary Figure 4g) [39]. Several markers, including CD31, CD11a, CD34, CD45, and CD55 exhibited altered surface expression in SLC39A7 knockout cells, while others, such as CD4, C3AR, or C5L2, remained unchanged. Taken together, these data indicate that specific membrane proteins are differentially affected by loss of SLC39A7.

SLC39A7 transporter function is required for ER homeostasis, TNFR1 surface expression and cell death induction
To corroborate the link between SLC39A7 loss and the phenotypes observed, we reconstituted KBM7 FADD -SLC39A7 knockout cells with V5-tagged SLC39A7 or GFP (Fig. 5a). SLC39A7 reconstitution decreased BiP expression, restored TNFR1 surface levels and, consequently, sensitivity to cell death induction with TNFα and SMAC mimetic (Fig. 5b, c). To investigate the requirement of SLC39A7 transport activity, we performed analogous reconstitution experiments with SLC39A7 constructs bearing substitutions at conserved histidines (H329A or H358A) predicted to participate in an intramembranous zinc-binding site or at other conserved residues (H362A or G365R) within the same metalloprotease-like motif [40]. SLC39A7 H329A and H358A mutants failed to restore sensitivity to necroptosis and to relieve ER stress, whereas H362A and G365R behaved similar to the wildtype construct (Fig. 5d, e). These results are consistent with recent structural and transport data for SLC39A4, in which substitution of histidines corresponding to SLC39A7 H329 and H358, but not H362, was shown to affect transport activity [41]. Compared to the constructs able to functionally rescue SLC39A7 deficiency, H329A and H358A mutants showed reduced expression in KBM7 FADD -SLC39A7cells (Fig. 5e). This is likely a consequence of the unresolved ER stress, as all constructs were expressed at similar level in SLC39A7-proficient cells (Fig. 5e). While an intrinsic reduced stability of H329A and H358A mutants cannot be excluded, their expression was comparable to endogenous SLC39A7 levels, strongly suggesting that transport activity is required to restore ER homeostasis and necroptosis sensitivity. Genome-scale gain-of-function screens identify negative regulators of necroptotic cell death signaling Combination of the newly generated KBM7 FADDcell necroptosis model with recently developed CRISPR/Cas9based technologies mediating transcriptional activation of endogenous genes offered the opportunity to identify cellular inhibitors of necroptosis. We thus performed complementary gain-of-function screens employing the CRISPR/Cas9-based SAM approach [22]. KBM7 FADDcells expressing dCas9-VP64 and MS2-p65-HSF1 were transduced with the genome-scale SAM sgRNA lentiviral library and subsequently selected with TNFα or birinapant. NGS sequencing and downstream analysis allowed the identification of sgRNAs specifically enriched upon necroptosis induction, thereby revealing genes conferring resistance to necroptotic cell death when upregulated ( Fig. 6a, b, Supplementary Table 6 and 7). Genome-wide screens on birinapant-induced cell death revealed the SMAC mimetic target protein Cellular inhibitor of apoptosis (cIAP)2 (encoded by BIRC3) as top scoring gene and enriched for sgRNAs targeting cIAP1 (BIRC2), confirming the validity of our approach (Fig. 6b). TNFα treatment similarly revealed BIRC2 among other genes, and, notably, TNIP1 (TNFAIP3 interacting protein 1, ABIN-1) as the top scoring hit (Fig. 6a). Identification of TNIP1 was particularly interesting as this ubiquitin-binding protein affects multiple inflammatory pathways, including TNF receptor and Toll-like receptor signaling [42,43]. Indeed, TNIP1 was shown to negatively regulate NF-kB and MAPK activation [43,44] and to inhibit TNFα-induced apoptosis [45]. Moreover, single nucleotide polymorphisms (SNPs) in TNIP1 have been associated with several inflammatory diseases [42] and mice deficient for TNIP1 or knock-in for a ubiquitin-binding-defective mutant (TNIP1 [D485N]) develop autoimmunity [46][47][48][49]. We first generated KBM7 FADDcells expressing individual SAM-sgRNAs targeting TNIP1, BIRC3 or Renilla as control. Cells overexpressing TNIP1 or cIAP2 (BIRC3) showed increased resistance to TNFα-or birinapant-induced necroptosis, respectively (Fig. 6c, d, Supplementary FADD -sgRen and empty KBM7 FADD -SLC39A7serve as controls. Cells were treated overnight (16 h) with TNFα (10 ng/ml), SMAC mimetic (0.5 µM), or a combination thereof. Cell viability was assessed using a luminescence-based readout for ATP (CellTiter Glo). Data represent mean value ± s.d. of two independent experiments performed in triplicates. d Cell viability in KBM7 FADD -SLC39A7cells stably reconstituted with GFP or the indicated V5-tagged SLC39A7 constructs. Cells were treated as indicated for 24 h and cell viability was assessed as in c. Data represent mean value ± s.d. of two independent experiments performed in triplicates. e KBM7 FADD -SLC39A7or KBM7 FADDcells stably reconstituted with the specified constructs were lysed and subjected to immunoblotting with the indicated antibodies. Asterisk (*) indicates non-specific band  For each stimulus, screens were performed at low and high concentrations (TNFα: 10 ng/ml (green) or 100 ng/ml (purple); birinapant: 0.1 µM (blue) or 1 µM (orange)). Screen analysis was performed by identifying differentially enriched sgRNAs using DESeq2 and then aggregating sgRNAs to genes using Gene Set Enrichment Analysis. Identified hits are ranked according to the adjusted p-value of enrichment (-log10(p adj ). Bubble size corresponds to the average log2 fold-change (aLFC) of enrichment, color indicates the number of significantly enriched sgRNAs. Screens were performed in duplicate except the high concentration of birinapant in simplicate. c, d Cell viability in KBM7 FADD -SAM cells transduced with sgRNA targeting TNIP1, BIRC3 or Renilla luciferase (Ren). Cells were treated as indicated for 72 h and viability was assessed using a luminescence-based readout for ATP (CellTiter Glo). Data represent mean value ± s.d. of two independent experiments performed in triplicates. e KBM7 FADD -SAM cells transduced with the specified sgRNAs were lysed and subjected to immunoblotting with the indicated antibodies. Asterisk (*) indicates non-specific band upregulated cIAP2 was observed in the latter, in line with cIAP2 degradation being critically dependent on the presence of cIAP1 [50] (Supplementary Figure 5c). Confirming the on-target effects of TNIP1 sgRNAs, cells overexpressing wildtype TNIP1 via ectopic cDNA expression were resistant to TNFα-induced necroptosis and showed reduced phosphorylation of RIPK1 and MLKL, indicating that TNIP1 affects the necroptosis pathway upstream of these critical signaling events (Fig. 7a-c). Importantly, expression of a mutant (D472A) disrupting its ubiquitinbinding domain [47] was ineffective, demonstrating that protection requires TNIP1 ubiquitin-binding activity (Fig. 7a-c). In contrast, TNIP1 overexpression did not have any major effect on TNFα-induced NF-kB or MAPK pathway activation (Fig. 7d). Similar results were obtained upon sgRNA-mediated TNIP1 overexpression (Supplementary Figure 5c-d).
In summary, these data unveil a role for TNIP1 in controlling TNFα-induced cell death beyond apoptosis. Regulation of necroptosis could therefore contribute to the inflammatory phenotypes observed in TNIP1-deficient and TNIP1[D485N] knock-in mice, and be connected with its association with multiple inflammatory diseases.

Discussion
Despite its relatively recent discovery, evidence for the relevance of necroptosis in human physiology and pathology has accumulated over the past decade [4,13,14,51]. Thus, defining the functional genetic landscape of this fundamental process is expected to expand our understanding of this pathway and reveal potential therapeutic targets. The development of a cellular model efficiently recapitulating necroptosis, while conveniently enabling different types of somatic cell genetic screens, offered the opportunity of attempting a comprehensive survey of human genes involved in the process, both in terms of positive and negative regulation.
We found that deletion of FADD rendered KBM7 cells sensitive to SMAC mimetic as single agent. SMAC mimetics are considered promising anti-cancer agents and birinapant, having shown efficacy as a single agent against some tumors and leukemias [52,53], is currently being evaluated in clinical trials. SMAC mimetics are known to trigger TNFα production in certain cell lines [54][55][56][57] and we observed that phosphorylation of MLKL in KBM7 FADDcells upon SMAC treatment is delayed compared to TNFα stimulation, in line with the notion of autocrine TNF secretion (Supplementary Figure 1g). Intriguingly, we found that targeting of TNFRSF1B (encoding for TNFR2) in a pooled screen setting rendered cells more resistant to SMAC mimetic-mediated killing, pointing to a cell-intrinsic function. In contrast to TNFR1, TNFR2 can be fully activated only by recognition of membrane-bound, but not soluble TNFα [58,59]. Interestingly, activation of TNFR2 has been shown to sensitize macrophages for TNFR1mediated necroptosis [60]. In light of our findings, it will thus be interesting to evaluate whether sensitivity to SMAC mimetic treatment is correlated with TNFR2 expression.
Regarding the other resistance-conferring genes identified, it is tempting to speculate that some of the transcription factors may be directly or indirectly involved in transcriptional regulation of the different pathway components or TNFα itself. Indeed, the basal transcription factor SP1 has been shown to be recruited to the TNF promoter [61,62] and to mediate RIPK3 expression [63].
The prominent identification across multiple screens of SLC39A7, to date not linked to TNFα signaling, was of particular interest, considering the instrumental role of SLCs in influencing essential physiological processes by regulating metabolic fluxes between the environment and intracellular compartments [64][65][66]. SLC39A7 controls zinc efflux from the ER to the cytoplasm and its absence has been linked to disturbed zinc homeostasis and ER stress [29][30][31]36]. We here experimentally link the resistance to ligand-triggered cell death induction observed in SLC39A7cells to a trafficking defect of the respective receptors. This is reminiscent of the Notch trafficking abnormalities described for mutants in Catsup, the Drosophila ortholog of SLC39A7 [36]. Whereas we found that basic glycosylation of TNFR1 occurs in SLC39A7cells, the receptor does not transit to the Golgi apparatus, suggesting that proper folding is not attained. Several of the folding factors in the ER, including protein disulfide isomerases (PDIs) and the chaperones calnexin and calreticulin, rely on zinc as a cofactor and could therefore be affected by SLC39A7 deficiency. Indeed, PDI activity can be inhibited by high zinc concentration [31]. Importantly, reconstitution of SLC39A7cells with wildtype SLC39A7, but not with mutants predicted to impair its transport activity [40], restored all affected cellular functions indicating that the perturbations induced by SLC39A7 deficiency are reversible, and that substrate-transport is required for its function. Remarkably, the trafficking of specific membrane receptors is differentially affected by SLC39A7 deficiency. We did not identify obvious differences in their posttranslational processing with regard to glycosylation and disulfide bridge formation, and thus the definition of the molecular basis for this divergence remains an intriguing question to be addressed in future studies.
Our data show that SLC39A7 is the prominent SLC affecting TNFR1 and FAS receptor signaling and suggest a largely non-redundant role in ER homeostasis, central processes involved in the pathophysiology of the immune compartment and beyond. It is therefore tempting to speculate that loss-of-function or hypomorphic mutations in SLC39A7 may result in associated genetic diseases. Indeed, defective FASL-induced lymphocyte apoptosis underlies a group of primary immunodeficiencies (PIDs) denoted as autoimmune lymphoproliferative syndrome (ALPS) and our screen with FASL retrieved SLC39A7 amongst genes previously identified in ALPS-and ALPS-like PIDs such as FAS, FADD, and CASP8 (Supplementary Figures 4b) [67]. Mutations in TNFRSF1A, encoding TNFR1, can lead to the autoinflammatory disorder TNFR-associated periodic syndrome (TRAPS) [68], and some variants have been shown to result in abnormal oligomerization and ER retention [69]. Similarly, perturbation of the ER compartment can result in PIDs, as highlighted by the primary antibody deficiency due to plasma cell defects observed in patients with mutations in the Sec61 translocon subunit SEC61A1 [70].
Complementing the loss-of-function approaches, we used CRISPR/Cas9-based gene activation technology to perform genome-scale overexpression screens in analogous settings. Our data highlights a number of candidate genes conferring resistance to necroptotic cell death and revealed a novel role for TNIP1 in inhibiting necroptosis. TNIP1 has been shown to negatively regulate multiple inflammatory pathways through, in particular, its NF-kB inhibitory and anti-apoptotic activities [42,43,45]. We demonstrated that TNIP1 overexpression prevented TNFα-induced phosphorylation of RIPK1 and of the downstream effector MLKL via a ubiquitin-binding dependent mechanism, while NF-kB and MAPK signaling was largely unaffected. This expands the role of TNIP1 in controlling TNFα-induced cell death pathways beyond the proposed role in inhibiting apoptosis by interfering with the recruitment of caspase-8 by FADD [45]. Considering the association of TNIP1 with multiple inflammatory and autoimmune diseases [42] as well as the TNIP1-dependent phenotypes observed in animal models [46][47][48][49]71], it will be of importance to investigate the relative contribution of TNIP1-dependent necroptosis inhibition we described here. In line with our results, a recent study showed that TNIP1-deficiency sensitizes cells to TNFα-induced necroptosis by affecting RIPK1 activation [72]. Upon TNFα treatment, TNIP1 is recruited to the TNF receptor signaling complex by the action of the Met1-ubiquitylating complex LUBAC, resulting in the engagement of the deubiquitinating enzyme A20, which targets RIPK1 and limits its activation [72].
In summary, we presented here complementary genomescale loss-of-function and gain-of-function screens in a novel cellular model for necroptosis, and mechanistically validated the role of SLC39A7 and TNIP1 in death receptor trafficking and signaling. Highlighting multiple novel candidate regulators affecting TNFα responses and necroptotic cell death, our work provides the basis for future studies aiming at defining the individual contributions in these central immune processes and exploring the role in associated pathological conditions.

Cell line generation
HEK293T cells were transiently transfected with pGAG-POL, pVSV-G, pADVANTAGE and retroviral expression vectors, or psPAX2, pMD2.G and lentiviral expression vectors using Polyfect (301105, QIAGEN). After 24 h the medium was replaced with fresh medium. The viruscontaining supernatant was harvested 48 h later, filtered (0.45 μm), supplemented with 8 μg/ml protamine sulfate (P3369, Sigma-Aldrich) and added to 40-60% confluent target cells. Suspension cell lines were subjected to spinfection (2000 rpm, 45 min, RT). 24 h after infection, the medium was replaced with fresh medium supplemented with the respective antibiotics to select for infected cells. Where indicated, target gene expression was induced by adding 2 µg/ml doxycycline. For generation of KBM7 FADDcells pX330-U6-Chimeric_BB-CBh-hSpCas9 harboring a FADD-targeting sgRNA was electroporated into KBM7 cells using Nucleofector TM Technology (Lonza). In brief, 2 × 10 6 cells were spun down, resuspended in 100 µl Nucleofection solution (Amaxa Cell Line Nucleofector Kit V, Lonza), and transferred to an electrode cuvette. After addition of 1 µg of target vector DNA and electroporation (program Y-005), cells were immediately resuspended in 700 µl medium without antibiotics and transferred to a culture vessel with pre-warmed medium. Single cell knockout clones were derived by limiting dilution from KBM7 FADD-Cas9 cells transduced with LGPIG sgSLC39A7_618 or KBM7 FADD-transduced with pLentiCRISPRv2 sgSLC39A7_618.

Haploid genetic screens and deep sequencing analysis
Haploid genetic screening was conducted as described previously [19,74]. In brief, gene-trap virus was produced by transient transfection of low passage, subconfluent HEK293T cells with the gene-trap plasmid and packaging plasmids pGAG-POL, pVSV-G and pADVANTAGE using Lipofectamine 2000 (Thermo Fisher Scientific). Every 24 h for three consecutive times the virus-containing supernatant was collected and replaced with fresh medium. Virus was concentrated via ultracentrifugation and used to mutagenize 1 × 10 8 KBM7 cells by spinfection. Mutagenized cells were then expanded and directly used for genetic screens, while 1 × 10 8 gene-trapped cells were harvested as unselected control population. For screens, a total number of 1 × 10 8 gene-trapped cells/screen was selected with the indicated stimuli (TRAIL 100 ng/ml, FASL 250 ng/ml) in 96-well plates (1 × 10 5 cells/well in 100 µl). After 12 days, resistant clones were pooled, cleaned up using Lonza lymphocyte separation buffer (17-829E), expanded to a total number of 3-6 × 10 7 cells, and genomic DNA was isolated.
To map insertions by next generation sequencing, retroviral insertion sites in necroptosis screen samples and unselected reference pools were recovered via linear amplification-mediated (LAM)-PCR, and FASL and TRAIL screen samples were digested with NlaIII and MseI and circularized, followed by inverse PCR. Samples were loaded onto an Illumina HiSeq 2000 machine using custom sequencing primer [18] and sequenced 50 base pair singleend. The obtained sequences in the FASTQ data file were mapped to the human reference genome hg19 (UCSC hg19 build) using bowtie2 (version 2.2.4) with default parameters. Sequencing reads with multiple alignment as well as reads with low mapping quality (MAPQ < 20) were discarded. Duplicate reads were marked and discarded with Picard (version 1.118). Insertions 1 or 2 base pairs away from each other were removed to avoid inclusion of insertions due to mapping errors. Each uniquely mapped read corresponds to an insertion site and was annotated with gene build GRCh37.p13 (ENSEMBL 75-release February 2014) using bedtools (version 2.10.1) and custom scripts. Only insertions that are predicted to be disruptive to gene function were further tested for significant enrichment. The significance of enrichment of insertions in a given gene was calculated by comparing the number of insertions in the selected populations with the unselected control data set by applying a one-sided Fisher's exact test. P-values were adjusted for false discovery rate (FDR) using Benjamini-Hochberg procedure (5% threshold was considered significantly enriched). Screen results were visualized using Circos (version 0.66) software, summary bubble plots and gene-trap insertion plots were generated via custom scripts using R statistical environment.

CRISPR/Cas9-based genetic screens
The SLC KO CRISPR/Cas9 library used is described in detail in [75] and Girardi et al. (in preparation). Briefly, a CRISPR/Cas9 library targeting 388 SLC genes with six sgRNAs per gene, together with a set of 120 sgRNAs targeting 20 genes essential in KBM7 and HAP1 cells [35], and a set of 120 non-targeting sgRNAs was cloned by Gibson cloning in the pLentiCRISPRv2 lentiviral vector. Viral particles were generated by transient transfection of low passage, subconfluent HEK293T cells with the SLCtargeting library and packaging plasmids pGAG-POL and pVSV-G using PolyFect (Qiagen). After 24 h the media was changed to fresh IMDM media supplemented with 10% FCS and antibiotics. The viral supernatant was collected after 48 h, filtered and stored at −80°C until further use. The supernatant dilution necessary to infect KBM7 FADDcells at a MOI (multiplicity of infection) of 0.2-0.3 was determined by puromycin survival after transduction as described in the ref. [76]. KBM7 FADDcells were infected in duplicates with the SLC KO library at high coverage (1000×) and after selection for 7 days with puromycin (0.5 μg/ml) an initial sample was collected to control for library composition. Cells were then treated with a set of stimuli (KBM7 FADD -: TNFα 50 ng/ml, 12 days; TNFα 10 ng/ml, 12 days; SMAC mimetic 5 μM, 12 days; SMAC mimetic 1 μM, 12 days; KBM7 wildtype: TRAIL 100 ng/ml, 14 days; FASL 100 ng/ml, 10 days; HAP1 wildtype: TRAIL 100 ng/ ml, 9 days; FASL 100 ng/ml, 7 days) and cell samples collected from both treated and control samples. Genomic DNA was extracted using the DNAeasy kit (QIAGEN) and the cassette containing the sgRNA sequence amplified with two rounds of PCRs following the procedure described in the ref. [76].
The amplified samples were sequenced on a HiSeq3000/ 4000 (Illumina), followed by processing with a custom analysis pipeline (see next section).
Genome-scale SAM screen was performed following an analogous procedure as previously described [22]. In brief, KBM7 FADDcells were transduced first with lenti MS2-P65-HSF1 (Addgene plasmid #61426; selection Hygromycin 800 μg/ml) and subsequently with lenti dCAS-VP64 (Addgene plasmid #61425, selection Blasticydin 20 μg/ml) to generate KBM7 FADD -SAM cells. The Human CRISPR/Cas9 Synergistic Activation Mediator (SAM) pooled sgRNA library (Addgene #1000000057) was amplified and packaged into lentivirus following the same procedure described above for the SLC KO CRISPR/Cas9 library. After viral titration, 2 × 10 8 KBM7 FADD -CRISPR-SAM cells were infected at 0.3 MOI and selected 48 h later with zeomycin (150 μg/ml) for 8 days. 5 × 10 6 cells were either harvested (time 0 control), treated in duplicate with two doses of TNFα (10 or 100 ng/ml) or SMAC mimetic (0.1 or 1 μM) or left untreated (untreated control). After 72 h cells were washed, expanded for 7 days before isolation of live cells by Lymphoprep (Stemcell technologies). Genomic DNA was extracted using the DNAeasy Blood & Tissue kit (QIAGEN) and the cassette containing the sgRNA sequence amplified using NEBnext High Fidelity 2X Master Mix (New England Biolabs) in 36 single-step PCR reactions of 23 cycles per sample with primers introducing sequencing adapters and barcodes (described in the ref. [22]). PCR products were pooled, purified and gel extracted. The amplified samples were then quantified, pooled (maximum of eight samples per sequencing lane) and sequenced on a HiSeq3000/4000 (Illumina), followed by processing with a custom analysis pipeline (see next section).

Analysis of CRISPR screens
Sequences of sgRNAs were extracted from RNA-Seq reads, matched against the original sgRNA library index and counted using an in-house Python script. To compensate for the noise and off target action of sgRNAs inherent to CRISPR screening approaches, we used a two-step differential abundance analysis. In a first step, differential abundance of sgRNAs was estimated with DESeq2 [77] using a two-factor design that accounts for both time and treatment variables. Subsequently, significantly enriched sgRNAs (adjusted p-value ≤ 0.05) were sorted by log2-fold change and aggregated to genes using Gene Set Enrichment Analysis [78], only considering genes with at least two significantly enriched sgRNAs.

T7 Endonuclease assay
To determine FADD targeting efficiency, 1-3 × 10 6 cells were spun down, washed in 1x PBS and genomic DNA isolated using the DNeasy Blood & Tissue Kit (QIAGEN) according to the manufacturer's instructions. The CRISPR/ Cas9 targeting site-containing locus was PCR-amplified (fw primer: GAGCTGACCGAGCTCAAGTTCCTAT; rv primer: CAAATCAAACCCGGCAAAGG; product size: 342 bp). The PCR products were denatured at 95°C for 2 min and re-annealed by ramping from 95-85°C at −2°C/s and from 85 to 25°C at −0.1°C/s. The annealed PCR products were digested with T7 Endonuclease I (NEB, M0302) for 20 min at 37°C, control digests without T7 Endonuclease I served as negative control. The fragmented PCR products were analysed by agarose gel electrophoresis.

Cell viability assays
Cells were seeded in 96-well plates at the appropriate cell density. For cell death induction, cells were incubated with the indicated compounds as stated or overnight (14-16 h

Flow cytometry-based multi-color competition assay (MCA)
To analyze the long-term cellular response upon treatment with different cell death inducers in CRISPR/Cas9-based knockout cell competition experiments, cell populations were marked with a fluorescent reporter coupled to individual sgRNAs, with sgRNA targeting Renilla luciferase (sgRen) serving as control. sgRen-mCherry + reporter cells (LGPIC lentiviral sgRNA vector) were mixed with sgRen-GFP + control or target gene sgRNA-GFP + cells (LGPIG lentiviral vector) in a 1:1 ratio. The mixed cell populations were treated with the cell death inducers indicated or left untreated as reference. After 14 days, the respective percentage of mCherry + and GFP + was determined by flow cytometry, gating on viable cells (FSC/SSC).
Triton X-114 phase separation for enrichment of membrane proteins Triton X-114 phase separation was performed as described previously [79]. Briefly, 60-75 × 10 6 cells were harvested and washed once with PBS. Cells were then resuspended in 500 μl PBS and 100 μl 6 % pre-condensed Triton X-114, mixed by inversion and incubated for 15 min on ice. After centrifugation for 1 min at 13,000 rpm, the supernatants were transferred to new tubes and incubated for 5 min at 37°C to induce phase separation. The upper aqueous phase was transferred to a new tube. To wash, 500 μl PBS were added to the detergent phase and 100 μl 6 % Triton X-114 to the aqueous phase, followed by incubation for 5 min on ice and 5 min at 37°C. Samples were centrifuged as before and the initial phases were kept for further processing. Proteins in the detergent phase were precipitated by adding 450 µl PBS, 500 μl methanol and 125 μl chloroform followed by vortexing. After centrifugation for 4 min at 13,000 rpm, 750 µl of the supernatant were removed, 400 µl methanol added and mixed by pipetting. Samples were centrifuged for 1 min at 13,000 rpm, the supernatant removed and the pellets dried at RT. Subsequently, the proteins were resuspended by sonication in 100 µl SDS-PAGE sample buffer without glycerol and bromophenol blue. Samples were then either subjected to immunoblot analysis or further prepared for mass spectrometric analysis via the FASP (filter aided sample preparation) method as described previously [80].

Reversed-phase liquid chromatography mass spectrometry (LCMS)
Mass spectrometry was performed on a hybrid linear trap quadrupole (LTQ) Orbitrap Velos mass spectrometer (ThermoFisher Scientific, Waltham, MA) using the Xcalibur version 2.1.0 coupled to an Agilent 1200 HPLC nanoflow system (dual pump system with one pre-column and one analytical column) (Agilent Biotechnologies, Palo Alto, CA) via a nanoelectrospray ion source using liquid junction (Proxeon, Odense, Denmark). Solvents for LCMS separation of the digested samples were as follows: solvent A consisted of 0.4% formic acid in water and solvent B consisted of 0.4% formic acid in 70% methanol and 20% isopropanol. From a thermostatic microautosampler, 8 μl of the tryptic peptide mixture were automatically loaded onto a trap column (Zorbax 300SB-C18 5 μm, 5 × 0.3 mm, Agilent Biotechnologies) with a binary pump at a flow rate of 45 μl/ min. 0.1% TFA was used for loading and washing the precolumn. After washing, the peptides were eluted by back-flushing onto a 16 cm fused silica analytical column with an inner diameter of 50 μm packed with C18 reversed phase material (ReproSil-Pur 120 C18-AQ, 3 μm, Dr. Maisch, Ammerbuch-Entringen, Germany). The peptides were eluted from the analytical column with a 27 min gradient ranging from 3 to 30% solvent B, followed by a 25 min gradient from 30 to 70% solvent B and, finally, a 7 min gradient from 70 to 100% solvent B at a constant flow rate of 100 n;/min. The analyses were performed in a datadependent acquisition mode using a top 10 collisioninduced dissociation (CID) method. Dynamic exclusion for selected ions was 60 s. A single lock mass at m/z 445.120024 was employed. The maximal ion accumulation time for MS in the orbitrap and MS2 in the linear trap was 500 and 50 ms, respectively. Automatic gain control (AGC) was used to prevent overfilling of the ion traps. For MS and MS2, AGC was set to 10 6 and 5000 ions, respectively. Peptides were detected in MS mode at a resolution of 60,000 (at m/z 400). The threshold for switching from MS to MS2 was 2000 counts. All samples were analysed as technical, back-to-back replicates.

Mass spectrometry data processing and analysis
The acquired raw MS data files were processed with msconvert (ProteoWizard Library v2.1.2708) and converted into Mascot generic format (mgf) files. The resultant peak lists were searched against either the human or mouse SwissProt database v2014.07 (40,984 and 24,862 sequences, respectively, including isoforms obtained from varsplic. pl and appended with known contaminants) with the search engines Mascot (v2.3.02, MatrixScience, London, U.K.) and Phenyx (v2.5.14, GeneBio, Geneva, Switzerland). Submission to the search engines was via a Perl script that performs an initial search with relatively broad mass tolerances (Mascot only) on both the precursor and fragment ions (±10 ppm and ±0.6 Da, respectively). High-confidence peptide identifications were used to recalibrate all precursor and fragment ion masses prior to a second search with narrower mass tolerances (±4 ppm and ± 0.3 Da, respectively). One missed tryptic cleavage site was allowed. Carbamidomethyl cysteine and oxidised methionine were set as fixed and variable modifications, respectively. To validate the proteins, Mascot and Phenyx output files were processed by internally-developed parsers. Proteins with ≥2 unique peptides above a score T1 or with a single peptide above a score T2 were selected as unambiguous identifications. Additional peptides for these validated proteins with score >T3 were also accepted. For Mascot and Phenyx, T1, T2, and T3 peptide scores were equal to 14, 28, 10 and 4.2, 4.7, 3.5, respectively (p-value < 10 -3 ). The validated proteins retrieved by the two algorithms were merged, any spectral conflicts discarded and grouped according to shared peptides. A false discovery rate of <1% for protein identifications and <1% for peptides (including the ones exported with lower scores) was determined by applying the same procedure against a database of reversed protein sequences. Commonly-known contaminants including trypsin and keratin were removed. In each condition, four spectral counts were available for the median (two biological replicates and two technical for each). Average spectral counts of the replicates were used to determine log2 fold changes between KBM7 FADD-SpCas9 SLC39A7vs. sgRen control. p-values were calculated by a two-sided Welsch Two Sample t-test. Annotations for the GO term "GO:0034976 response to endoplasmic reticulum stress" were obtained from QuickGO (www.ebi.ac.uk/QuickGO). GO term enrichment was computed with R's topGO package for biological process ontology, using the default algorithm ("weight01") and the annotation file from geneontology.org as of October 2016. The study set consisted in all proteins identified with a log2 fold change ≥1 and a p-value < 0.05 (65 proteins). All reviewed human proteins in UniProtKB/Swiss-Prot were used as background population (20,348 proteins). p-values were corrected for multiple testing using Benjamini-Hochberg's procedure (False Discovery Rate). Gene Set Enrichment Analysis was performed using the GSEA software from the Broad Institute with default parameters. The gene list consisted in all proteins identified below a p-value of 0.05 (354 proteins), ranked by log2 fold change. The analysis was run against the Hallmarks gene set. Supplementary Table 2 gives results for LC-MSMS analysis, GO term enrichment, and GSEA.

Experimental design, data plotting and statistical rational
Proteomics experiments were performed as biological replicates (n = 2) and analyzed by LC-MSMS as technical duplicates. CRISPR/Cas9 library screens were performed in duplicate (n = 2). Flow cytometry-based MCA data are shown as mean value ± s.d. of at least two independent experiments (n ≥ 2) performed in duplicates. Data calculations were performed using Microsoft Excel (Microsoft, Redmond, WA, USA), data plotting and statistical analysis was done using GraphPad Prism 6 (GraphPad Software) if not otherwise stated. A normal distribution of data was assumed and appropriate tests were applied. Visualization of data was performed using R statistical environment.

Compliance with ethical standards
Conflict of interest The Superti-Furga laboratory at CeMM has an active research collaboration with Boehringer Ingelheim (Research Collaboration Agreement BICeMM 238114). The remaining authors declare that they have no conflict of interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.