INPP5A phosphatase is a synthetic lethal target in GNAQ and GNA11-mutant melanomas

Elbatsh, Ahmed M. O.; Amin-Mansour, Ali; Haberkorn, Anne; Textor, Claudia; Ebel, Nicolas; Renard, Emilie; Koch, Lisa M.; Groenveld, Femke C.; Piquet, Michelle; Naumann, Ulrike; Ruddy, David A.; Romanet, Vincent; Martínez Gómez, Julia M.; Shirley, Matthew D.; Wipfli, Peter; Schnell, Christian; Wartmann, Markus; Rausch, Martin; Jager, Martine J.; Levesque, Mitchell P.; Maira, Sauveur-Michel; Manchado, Eusebio

doi:10.1038/s43018-023-00710-z

Download PDF

Article
Open access
Published: 17 January 2024

INPP5A phosphatase is a synthetic lethal target in GNAQ and GNA11-mutant melanomas

Ahmed M. O. Elbatsh¹,
Ali Amin-Mansour²^na1,
Anne Haberkorn¹^na1,
Claudia Textor³^na1,
Nicolas Ebel¹,
Emilie Renard¹,
Lisa M. Koch¹,
Femke C. Groenveld¹,
Michelle Piquet²,
Ulrike Naumann ORCID: orcid.org/0000-0001-7783-7675⁴,
David A. Ruddy²,
Vincent Romanet¹,
Julia M. Martínez Gómez⁵,
Matthew D. Shirley²,
Peter Wipfli³,
Christian Schnell¹,
Markus Wartmann ORCID: orcid.org/0009-0007-9822-3112¹,
Martin Rausch⁴,
Martine J. Jager⁶,
Mitchell P. Levesque ORCID: orcid.org/0000-0001-5902-9420⁵,
Sauveur-Michel Maira¹ &
…
Eusebio Manchado ORCID: orcid.org/0009-0009-3170-9788¹

Nature Cancer volume 5, pages 481–499 (2024)Cite this article

5089 Accesses
4 Altmetric
Metrics details

Subjects

Abstract

Activating mutations in GNAQ/GNA11 occur in over 90% of uveal melanomas (UMs), the most lethal melanoma subtype; however, targeting these oncogenes has proven challenging and inhibiting their downstream effectors show limited clinical efficacy. Here, we performed genome-scale CRISPR screens along with computational analyses of cancer dependency and gene expression datasets to identify the inositol-metabolizing phosphatase INPP5A as a selective dependency in GNAQ/11-mutant UM cells in vitro and in vivo. Mutant cells intrinsically produce high levels of the second messenger inositol 1,4,5 trisphosphate (IP3) that accumulate upon suppression of INPP5A, resulting in hyperactivation of IP3-receptor signaling, increased cytosolic calcium and p53-dependent apoptosis. Finally, we show that GNAQ/11-mutant UM cells and patients’ tumors exhibit elevated levels of IP4, a biomarker of enhanced IP3 production; these high levels are abolished by GNAQ/11 inhibition and correlate with sensitivity to INPP5A depletion. Our findings uncover INPP5A as a synthetic lethal vulnerability and a potential therapeutic target for GNAQ/11-mutant-driven cancers.

Targeting KRAS in cancer

Article 18 April 2024

Tumor-selective activity of RAS-GTP inhibition in pancreatic cancer

Article 08 April 2024

KRAS mutation: from undruggable to druggable in cancer

Article Open access 15 November 2021

Main

UM is the most recurring primary intraocular cancer in adults and the second most common melanoma subtype after cutaneous skin melanoma (CM)¹. Despite effective clinical management of the primary disease, 50% of patients with UM develop distant metastases and have an overall survival of less than 12 months². Localized primary tumors can be controlled by radiotherapy, phototherapy or enucleation, whereas metastatic malignancies are refractory to chemotherapy, MAPK-targeted inhibitors or immune-checkpoint blockades². Although both arise from melanocytes, uveal and cutaneous melanomas are highly distinct genetically and clinically³. UM tumors do not harbor mutations in BRAF or NRAS, the two most prevalent oncogenes mutations in CM⁴. Instead, mutually exclusive activating mutations in GNAQ and GNA11, encoding GNAQ and GNA11 G-proteins, occur in about 90% of UMs^5,6. The remaining 10% are driven by mutations identified in the GNAQ/11-linked G-protein-coupled receptor (GPCR), cysteinyl leukotriene receptor 2 (CYSLTR2) or phospholipase C β4 (PLCβ4), a direct downstream effector of GNAQ/11 (refs. ^7,8).

GNAQ/11 are the α subunits of the heterotrimeric Gq-protein family⁹. These highly homologous proteins function as the core GTPase machinery that relays signals from upstream GPCRs to downstream effectors to modulate numerous cellular processes¹⁰. Oncogenic mutations of GNAQ/11, occurring at residues Q209 (95%) and R183 (5%), give rise to GTPase-defective proteins that lack their intrinsic GTP-hydrolyzing activity¹. These oncoproteins are thus predominantly locked in an active configuration, resulting in constitutive activation of their downstream signaling¹¹. GNAQ/11 downstream effector, PLCβ, hydrolyzes phosphatidylinositol 4,5-bisphosphate (PIP2) at the plasma membrane to produce two cellularly active second messengers; IP3 and diacylglycerol (DAG)¹². IP3 mobilizes calcium to regulate various intracellular processes, whereas DAG is responsible for the plasma membrane tethering of protein kinase C (PKC), a serine/threonine kinase that is required for MAPK pathway activation¹³. In principle, all known UM oncogenic mutations lead to the constitutive production of both second messengers. While it is conceivable that DAG is essential for UM cells to activate the MAPK pathway to promote cell proliferation, the cellular and molecular implications of the sustained production of IP3 in the context of UM are largely unknown.

Unprecedented advances in high-throughput functional genomics and sequencing technologies have permitted the execution of large-scale profiling of almost all cancer types. Such projects generate extensive datasets that represent a rich repository to discover cancer-specific genetic vulnerabilities and unravel mechanisms of oncogenesis. In this study, we performed genome-scale CRISPR knockout screens on UM cells and analyzed publicly available cancer dependency and expression datasets to identify GNAQ/11-mutant-specific dependencies. We found inositol polyphosphate 5-phosphatase A (INPP5A) to be selectively required for the growth of GNAQ/11-mutant cells (GNAQ/11^Mut) in vitro and in vivo, yet dispensable for GNAQ/11 wild-type (GNAQ/11^WT) cells. In addition, we uncovered the underlying molecular mechanism of INPP5A dependency in GNAQ/11-mutant cells. Finally, we identified inositol tetrakisphosphate (IP4) as a predictive biomarker for INPP5A dependency and showed that it is elevated in GNAQ/11^Mut patient-derived xenografts and human tumors.

Results

Genetic screening for GNAQ/11-mutant UM dependencies

To probe selective genetic dependencies in UM cells with GNAQ/11 mutations, we performed genome-scale CRISPR-Cas9 screens in three independent GNAQ/11^Mut UM cell lines and concurrently analyzed datasets of projects AVANA and Score (performed using CRISPR-Cas9 technology) and projects Achilles and DRIVE (performed using RNA interference (shRNA) technology, respectively)^14,15,16,17. In addition, we interrogated the gene expression data generated by Cancer Cell Line Encyclopedia (CCLE) and The Cancer Genome Atlas (TCGA) (Fig. 1a)^18,19. Cas9-expressing 92.1, OMM1 and MP41 UM cell lines were transduced with a single-guide RNA (sgRNA) library and changes in library representation after 4, 14 and 21 d of culture were monitored using deep sequencing (Extended Data Fig. 1a). sgRNA read count distribution and depletion patterns of sgRNAs targeting known pan-lethal and nonlethal genes were comparable among the three screens, reflecting the overall high quality of our screening approach (Extended Data Fig. 1b). GNAQ and GNA11 sgRNAs were only depleted in cell lines harboring an oncogenic mutation in the corresponding gene, confirming that UM cells depend on continuous GNAQ or GNA11 activity (Fig. 1b).

**Fig. 1: Genome-scale CRISPR-Cas9 screens identify genetic dependencies in GNAQ/11-mutant UM cells.**

We then integrated the data of the three UM screens with publicly available cancer dependency maps using a multilayer bioinformatics analysis pipeline. The analysis entailed stepwise filtering criteria that ultimately led to the identification of essential genes specifically for GNAQ/11^Mut cells while excluding pan-cancer and melanocyte-lineage lethal genes (Fig. 1a and Extended Data Fig. 1c). In total, we identified 29 genes that passed all filtering steps at two time points (day 14 and day 21) and an additional 13 genes that scored only on day 21 (Fig. 1c and Extended Data Fig. 1d). The robustness of our approach was supported by the identification of RASGRP3, a validated GNAQ/11^Mut UM genetic dependency (Fig. 1b,c)²⁰. Functional annotation of the identified hits revealed their involvement in diverse biological processes, including MAPK and GPCR signaling, transcription regulation, immune response and inositol phosphate metabolism (Fig. 1c, Extended Data Fig. 1d and Supplementary Table 1).

To prioritize our gene list, we mapped the expression profile of GNAQ/11^Mut UM by leveraging the messenger RNA gene expression datasets of CCLE and TCGA. We defined a gene expression signature for GNAQ/11^Mut UM consisting of 213 genes (Supplementary Table 2). Of these, 193 and 20 genes displayed a significantly higher and lower mRNA expression in UM compared to all other solid tumor types, respectively (Fig. 1d,e and Extended Data Fig. 1e). Consistent with previous reports, Gene Ontology (GO) term enrichment analysis revealed processes that included melanogenesis, inflammatory response and oxidative phosphorylation, as well as enrichment of genes involved in inositol phosphate metabolism (Fig. 1f)^21,22. Notably, a few genes in the hit list of the genetic screens belong to this metabolic pathway, highlighting the relevance of inositol phosphate biology in GNAQ/11^Mut UM cells (Fig. 1c,f). Among these, inositol polyphosphate 5-phosphatase A (INPP5A) emerged as one of the top candidate genes (Fig. 1c). We thereby decided to follow up on INPP5A for several reasons. First, our comprehensive analysis revealed its selective essentiality in GNAQ/11^Mut UM, whereas none of the ten other inositol phosphatase genes encoded in the human genome was preferentially essential (Extended Data Fig. 1f). Second, the role of INPP5A in UM remains largely unknown. Last, INPP5A de facto is an enzyme, hence may represent an appealing druggable vulnerability in UM.

INPP5A is required for the growth of GNAQ/11-mutant UM

To validate INPP5A dependency, we evaluated the effect of INPP5A knockout in six GNAQ/11^Mut UM and seven GNAQ/11^WT UM and CM cell lines using two individual INPP5A sgRNAs (Extended Data Fig. 2a–c and Supplementary Table 3). Depleting INPP5A dramatically inhibited the clonogenic growth as well as the proliferation of all GNAQ/11^Mut UM cell lines to a similar effect as compared to the positive control, PLK1 (Fig. 2a and Extended Data Fig. 2d). In contrast, INPP5A knockout did not affect the viability of GNAQ/11^WT UM or CM cells, albeit comparable depletion efficiency among all cell lines (Extended Data Fig. 2e). We also transduced two UM cell lines with doxycycline-inducible INPP5A shRNAs (Extended Data Fig. 3a,b). Complementing the CRISPR-Cas9 data, knocking down INPP5A markedly diminished the viability of the mutant cells (Extended Data Fig. 3c,d). These results indicate that INPP5A is selectively essential for the viability of GNAQ/11^Mut UM cells.

**Fig. 2: INPP5A is synthetic lethal with GNAQ/11 oncogenic mutations.**

Our findings suggest that INPP5A dependence in GNAQ/11^Mut cells could likely be exploited through INPP5A inhibition. To assess whether the enzymatic activity of INPP5A is essential for GNAQ/11^Mut UM, we expressed shRNA-insensitive complementary DNA of either wild-type or phosphatase-deficient INPP5A in the doxycycline-inducible shINPP5A 92.1 cells (Fig. 2b). To generate this catalytic mutant, we introduced a missense mutation (Asp384Gly) in a highly conserved residue essential for the phosphatase activity of INPP5A (Extended Data Fig. 3e–g)^23,24. In cDNA complementation experiments, wild-type INPP5A rescued the lethality induced by depleting the endogenous protein, confirming that the lethal phenotype accompanying INPP5A loss is a consequence of its on-target inactivation (Fig. 2c,d). In stark contrast, the catalytic mutant failed to rescue the lethal phenotype, suggesting that the phosphatase activity of INPP5A is essential to maintaining the viability of GNAQ/11^Mut UM and can be exploited as a potential therapeutic target.

To further investigate the functional relationship between INPP5A and GNAQ/11 oncoproteins in an UM lineage-independent context, we expressed GNAQ^WT or different mutants of GNAQ (Q209L, Q209P or R183Q) in HEK293A-Cas9 cells and assessed the effect of INPP5A knockout using competition-based proliferation assays (Fig. 2e and Extended Data Fig. 3h). INPP5A depletion inhibited the proliferation of all GNAQ^Mut-expressing cells but imposed negligible effects on the parental or GNAQ^WT HEK293A cells (Fig. 2f and Extended Data Fig. 3i). Suppressing GNAQ signaling activity using the GNAQ/11-specific inhibitor FR900359 (FR) rescued the lethality evoked by INPP5A depletion in the mutant cells (Fig. 2f)¹¹. Taken together, our findings reveal a lineage-independent synthetic lethal relationship between INPP5A loss and GNAQ/11 oncogenic mutants mediated through the aberrant signaling of these oncoproteins. Notably, INPP5A and mutant GNAQ/11 interaction compared favorably to other known synthetic lethal pairs such as NRAS mutants and SHOC2 or WRN helicase in microsatellite-instable cells (Extended Data Fig. 3j)^25,26.

IP3 signaling drives INPP5A dependency in GNAQ/11-mutant UM

We then sought to delineate the mechanistic basis of INPP5A dependence in GNAQ/11^Mut cells. INPP5A hydrolyzes IP3 and IP4 to inositol 1,4 bisphosphate (IP2) and inositol 1,3,4 trisphosphate, respectively (Fig. 3a)²⁷. IP3 is a second messenger generated by PLC through the activation of GNAQ/11 (Fig. 3a). We optimized an ion-pair chromatography/mass spectrometry method to dissect changes in intracellular levels of inositol phosphates upon INPP5A inactivation (Extended Data Fig. 4a and Supplementary Table 4)²⁸. INPP5A knockout did not result in noticeable alterations in inositol monophosphate (IP1) or IP2 levels in wild-type or mutant GNAQ/11 cells (Extended Data Fig. 4b–e). In contrast, depleting INPP5A led to a substantial accumulation of IP3 and IP4 in GNAQ/11^Mut UM cells, whereas it hardly affected their levels in GNAQ/11^WT UM or CM cells (Fig. 3b,c and Extended Data Fig. 4f,g). Consistently, we observed a marked elevation of IP3 and IP4 in GNAQ^Q209L-expressing but not in GNAQ^WT-expressing or parental HEK293A cells upon INPP5A loss (Fig. 3d,e). Importantly, IP3 and IP4 levels were restored by re-expressing wild-type but not phosphatase-deficient INPP5A, suggesting that INPP5A-mediated dephosphorylation is the principal route for their metabolism in GNAQ/11^Mut UM cells (Extended Data Fig. 5a,b).

We then probed whether the preferential accumulation of IP3 and IP4 in INPP5A-depleted mutant cells is the outcome of the constitutive signaling of the GNAQ/11 oncoproteins. To test this, we measured GNAQ/11 signaling output by quantifying the accumulation of IP1, a downstream metabolite of IP3 that is stabilized in the presence of LiCl, as a surrogate for IP3 production (Extended Data Fig. 5c)²⁹. GNAQ/11^Mut cells (UM or HEK293A-GNAQ^{Q209L/P or R183Q}) exhibited higher accumulation of IP1 compared to GNAQ/11^WT UM, CM or HEK293A cells, demonstrating that the mutant cells produce more IP3 than their wild-type counterparts (Extended Data Fig. 5d,e). Furthermore, inhibiting GNAQ/11 activity with FR significantly suppressed IP3 production and, most importantly, restored the elevated levels of IP3 and IP4 in INPP5A-depleted GNAQ/11^Mut cells (Fig. 3d,e and Extended Data Fig. 5d,e). This suggests that IP3 and IP4 preferential accumulation in INPP5A-depleted GNAQ/11^Mut cells is driven by the enhanced IP3 synthesis in these cells.

Because the key function of IP3 is to bind and activate IP3Rs to mobilize calcium, we hypothesized that IP3 accumulation and the consequent IP3 receptors (IP3Rs) activation might mediate the toxic effects of INPP5A depletion in GNAQ/11^Mut cells. To test this hypothesis, we first assessed the impact of CRISPR-mediated co-depletion of INPP5A and IP3R on the mutant cells. Mammalian cells express three distinct IP3R subtypes at variable levels³⁰. We performed our experiments on two independent UM cell lines, each predominantly expressing a distinct IP3R subtype (Extended Data Fig. 6a). Of note, depleting IP3R1 or IP3R3 markedly rescued the growth inhibition induced by INPP5A loss in OMM1^GNA11-Q209L or 92.1^GNAQ-Q209L UM cells, respectively, despite the sustained accumulation of IP3 and IP4 (Fig. 3f–i and Extended Data Fig. 6b,c). To further substantiate the notion that efficient termination of IP3 signaling is critical for GNAQ/11^Mut UM cells, we depleted ERLIN2, a member of the endoplasmic reticulum (ER)-associated degradation (ERAD) pathway that is responsible for degrading IP3Rs (Fig. 3j)³¹. ERLIN2 loss resulted in the upregulation of IP3R3 levels in all cells irrespective of their genotype and lineage (Fig. 3k); however, its depletion selectively impaired the viability of 92.1 and Mel202 GNAQ^Mut UM cell lines but did not affect Uacc62 GNAQ/11^WT CM cells (Fig. 3l and Extended Data Fig. 6d–f). In line with these results, analysis of TCGA expression data revealed that all three IP3R subtypes are expressed at lower levels in UM tumors compared to CM, supporting our hypothesis that tight regulation of IP3 signaling is crucial for the survival of GNAQ/11^Mut UM cells (Extended Data Fig. 6g–i). In summary, these results suggest that the enhanced synthesis of IP3 renders GNAQ/11^Mut UM cells sensitive to INPP5A inhibition by promoting the accumulation of IP3 and aberrant activation of IP3R signaling.

INPP5A loss increases cytosolic calcium in GNAQ/11-mutant UM

Previous work has shown that INPP5A suppression perturbs calcium homeostasis³². To evaluate the effect of INPP5A depletion on calcium signaling in GNAQ/11^Mut UM, we employed ratio-imaging microscopy using the high-affinity fluorescent calcium indicator fura-2 (Fig. 4a and Extended Data Fig. 7a,b). Using single-cell calcium measurements following fura-2 loading, we found that INPP5A depletion led to an increase in cytosolic calcium levels in GNAQ/11^Mut UM cells and in GNAQ/11^Q209L-expressing HEK293A cells (Fig. 4b,c and Extended Data Fig. 7c,d). Consistent with previous reports, we observed a minimal but significant increase in calcium levels of GNAQ/11^WT-expressing and parental HEK293A cells following INPP5A knockout (Fig. 4c)^32,33. In contrast, we did not detect any noticeable alterations in calcium levels in GNAQ/11^WT CM cells upon INPP5A depletion (Fig. 4b and Extended Data Fig. 7c,d).

**Fig. 4: INPP5A depletion preferentially results in elevated cytosolic calcium levels in GNAQ/11-mutant UM cells.**

It is well established that the disruption of intracellular calcium signaling triggers cell death³⁴. We thereby investigated whether the lethality induced by INPP5A loss in GNAQ/11^Mut cells is the consequence of perturbed calcium signaling. To this end, we engineered UM and CM cell lines with the genetically encoded calcium biosensor, dCys-GCaMP6 (ref. ³⁵). Upon calcium binding, dCys-GCaMP6-expressing cells produce GFP signal that allows us to track these cells to determine their fate (Fig. 4d). Using live-cell fluorescent imaging, we detected a progressive elevation of the GFP signal in GNAQ/11^Mut, but not GNAQ/11^WT cells, upon INPP5A depletion, validating our findings using the fura-2 indicator (Fig. 4e,f and Extended Data Fig. 7e). Accordingly, we observed a substantial increase in the percentage of dead cells in GNAQ/11^Mut, but not GNAQ/11^WT cells (Fig. 4g,h and Extended Data Fig. 7f). To associate this increase to changes in calcium levels, we performed single cell-tracking analysis (Fig. 4i). Whereas the majority of control cells underwent cell division, most INPP5A-depleted GNAQ/11^Mut cells died shortly after an increase in their GFP signal (Fig. 4i). Altogether, these results suggest that the aberrant activation of IP3/IP3R signal axis in INPP5A-depleted GNAQ/11^Mut cells leads to increased levels of cytosolic calcium and subsequent cell death.

INPP5A loss induces p53-mediated apoptosis in GNAQ-mutant UM

To gain further insight into the molecular mechanisms underlying the synthetic lethal interaction between INPP5A loss and GNAQ/11 mutations, we performed RNA sequencing upon INPP5A silencing (Extended Data Fig. 8a). We knocked down INPP5A in 92.1 UM cells and determined the transcriptional changes following the expression of empty vector, wild-type or phosphatase-deficient INPP5A. INPP5A loss induced substantial transcriptional changes in 84 genes (Fig. 5a and Supplementary Table 5). Expression of wild-type but not mutant INPP5A restored the normal transcription of the majority of the affected genes, suggesting that most, if not all, of INPP5A functions require its enzymatic activity (Fig. 5a). Examination of gene ontologies revealed that INPP5A depletion altered the expression of genes associated with GPCR signaling, inositol phosphate biology and calcium regulation, as well as genes involved in key cellular processes such as differentiation, transcription and cell cycle (Fig. 5b). We also performed gene set enrichment analysis (GSEA) of HALLMARK gene sets (Fig. 5c). Notably, p53 and apoptosis pathways were among the significantly upregulated signatures, whereas E2F targets and G2/M checkpoint gene sets were moderately repressed (Fig. 5c and Extended Data Fig. 8b,c).

**Fig. 5: INPP5A depletion induces p53-dependent apoptosis in GNAQ-mutant UM cells.**

In line with the transcriptomic data, we found that INPP5A depletion increased the levels of phosphorylated p53 at serine 15, a key post-translational modification required for p53 transcription activity (Extended Data Fig. 8d)³⁶. In addition, INPP5A loss in GNAQ^Mut UM cells led to an increase of the apoptotic marker cleaved PARP (c-PARP) and the percentage of apoptotic cells as assessed by annexin V staining (Fig. 5d and Extended Data Fig. 8d–f). To pursue the molecular basis underlying this apoptosis, we performed qPCR for the p53 transcription targets, NOXA, PUMA and CDKN1A³⁷. INPP5A depletion resulted in significant transcriptional upregulation of both proapoptotic genes, NOXA and PUMA (Fig. 5e). By contrast, changes in the expression of the cell-cycle regulator, CDKN1A, were limited, consistent with the modest increase in G1 cells observed following INPP5A depletion (Fig. 5e and Extended Data Fig. 8d,g–i). Knocking out TP53 in INPP5A-depleted GNAQ^Mut UM cells led to a remarkable rescue of all phenotypic traits, including suppression of the transcriptional upregulation of NOXA and PUMA, rescue of apoptosis and restoration of normal cell proliferation (Fig. 5f–i and Extended Data Fig. 8j–l). These results suggest that INPP5A depletion in GNAQ/11^Mut UM cells prompts a p53-dependent apoptotic response likely as a consequence of the elevation of cytosolic calcium levels.

INPP5A promotes UM tumor growth and metastasis development

We next investigated the impact of INPP5A inhibition on UM tumor growth in vivo. The 92.1 and Mel202 UM cells expressing doxycycline-inducible control and INPP5A shRNAs were subcutaneously transplanted into immunocompromised mice treated with doxycycline. Tumor xenografts expressing INPP5A shRNAs displayed reduced mRNA levels and significant tumor growth inhibition compared to those harboring a control shRNA (Fig. 6a,b and Extended Data Fig. 9a). The antitumor effect was consistent with a marked reduction in the proliferation marker Ki67 and accumulation of IP3 levels in 92.1 xenografts, suggesting that INPP5A inhibition evokes similar molecular responses in vivo and in vitro (Fig. 6c and Extended Data Fig. 9b–d). Importantly, regrowing tumors have restored their INPP5A mRNA levels, further substantiating that UM cells require INPP5A for tumor maintenance in vivo (Extended Data Fig. 9a).

**Fig. 6: INPP5A is required for UM tumor growth and metastasis development.**

We next determined whether suppression of INPP5A inhibits UM metastasis using a liver metastasis mouse model of UM³⁸. Luciferase-labeled UM cells transduced with doxycycline-inducible control or INPP5A shRNAs were transplanted into immunodeficient mice (Fig. 6d). Bioluminescent signal, ex vivo imaging and immunohistochemistry staining confirmed the ability of the injected UM cells to form metastatic tumors, predominantly in the liver (Extended Data Fig. 9e). To dissect the role of INPP5A in promoting UM metastasis formation and development, doxycycline treatment was initiated either directly following transplantation or after the formation of the macrometastasis (Fig. 6d, Extended Data Fig. 9f–i). Silencing INPP5A substantially inhibited both UM metastasis formation and growth of established metastases, even when treatment onset was delayed until the mice exhibited high metastatic tumor burden (Fig. 6e,f). Moreover, the anti-metastatic effect of INPP5A suppression extended the survival of treated mice as compared to control mice (Fig. 6g). Collectively, these findings demonstrate that INPP5A is required for the development of UM liver metastases.

IP4 is a predictive biomarker of INPP5A dependence

High levels of IP4 accumulate upon exogenous expression of activated GNAQ mutants³⁹. We hypothesized that the high steady-state levels of IP4 may be the consequence of enhanced IP3 synthesis and hence a predictive biomarker of sensitivity to INPP5A inhibition. We indeed found that GNAQ/11^Mut cells exhibit significantly higher steady-state levels of IP4 in comparison to GNAQ/11^WT cells (Fig. 7a and Extended Data Fig. 10a). In contrast, IP3 steady-state levels were comparable among all cell lines, indicating that IP3 metabolism can efficiently counteract the enhanced IP3 synthesis in GNAQ/11^Mut cells (Extended Data Fig. 10b,c). Inhibition of GNAQ/11 with FR led to reduced IP4 levels in GNAQ/11^Mut UM cells in vitro and in vivo, confirming that these elevated levels are the result of GNAQ/11-mediated enhanced IP3 synthesis (Fig. 7b,c). More notably, steady-state levels of IP4, but not IP3, positively correlated with sensitivity to INPP5A inhibition across all tested models, supporting the notion that IP4 can be used as a predictive biomarker for INPP5A dependence (Fig. 7d and Extended Data Fig. 10d).

To investigate the clinical significance of INPP5A as a therapeutic target, we examined the steady-state levels of IP4 in UM patient-derived xenografts (PDXs) and patients’ tumor samples (Fig. 7e–g and Extended Data Fig. 10e,f). All UM samples originated from metastatic tumors, some of which harbored inactivating mutations in BAP1, an event that frequently occurs in metastatic tumors and correlates with poor prognosis^3,21. Notably, steady-state levels of IP4 were significantly elevated in GNAQ/11^Mut UM PDXs and patients’ samples compared to GNAQ/11^WT CM samples indicating that patients’ GNAQ/11^Mut UM tumors intrinsically exhibit enhanced IP3 synthesis, the fundamental feature driving INPP5A dependence (Fig. 7e,g and Extended Data Fig. 10g,h). These results suggest that INPP5A depletion may act as a clinically relevant vulnerability for GNAQ/11^Mut UM.

Discussion

Identifying oncogene-induced vulnerabilities holds great promise for developing more efficacious and less toxic anticancer treatments⁴⁰. Here, we identified a synthetic lethal interaction between the inositol polyphosphate 5-phosphatase, INPP5A, and activating mutations in GNAQ/11. These mutations are the oncogenic driver events in over 90% of UM tumors, implying a broad applicability of INPP5A as a potential therapeutic target for UM tumors. Our findings provide key insights into the basis of the observed synthetic lethality. We propose a model in which the constitutive active state of oncogenic mutant GNAQ/11 results in enhanced synthesis of IP3, rendering the mutant cells highly dependent on INPP5A to metabolize this second messenger to terminate its signaling activity. Loss of INPP5A leads to IP3 accumulation, specifically in the mutant setting, resulting in the hyperactivation of IP3/Ca²⁺ signaling that promotes p53-dependent cell death by apoptosis (Fig. 7h). We also demonstrated that levels of IP4, a biomarker for enhanced IP3 synthesis, exquisitely correlate with sensitivity to INPP5A depletion and is elevated in metastatic GNAQ/11^Mut UM cell models, PDXs and patients’ tumors, underlying the potential relevance of INPP5A as a therapeutic target in the clinic.

Recent work has shown that other oncogenic alterations found in UM, such as mutations in CYSLTR2, a GNAQ/11-linked GPCR and in PLCβ4, a downstream effector of GNAQ/11, stimulate the overproduction of IP3 to a similar extent as GNAQ/11 mutations²⁹. Given that IP3 accumulation is a key driver of INPP5A dependence, we speculate that INPP5A may also act as a vulnerability for tumors harboring these other oncogenic mutations. In addition, molecular characterization of TCGA primary UM revealed that mutant GNAQ/11 signaling might be enhanced in tumors at high risk of metastasis²¹. As enhanced GNAQ/11 signaling translates into increased IP3 production, this analysis suggests that highly aggressive UM tumors that typically correlate with poor prognosis may be sensitive to INPP5A inhibition.

IP3 mobilizes calcium by binding to the IP3Rs at the endoplasmic reticulum⁴¹. It is subjected to two metabolic changes, dephosphorylation by INPP5A and phosphorylation by multiple IP3 kinases (IP3K), producing IP2 and IP4, respectively^27,42. Neither IP2 nor IP4 directly stimulates IP3R-mediated calcium release, indicating that both metabolic queues terminate IP3 signaling^43,44. Our findings, however, reveal that in GNAQ/11^Mut UM cells the two metabolic pathways are functionally non-redundant. IP3Ks do not compensate for INPP5A loss in mutant UM, pinpointing a dominant role for INPP5A in metabolizing IP3 in these cells. What determines the contribution of these metabolizing enzymes to IP3 flux is still elusive. One plausible factor could be the catalytic kinetics of IP3K and INPP5A. It has been shown that the kinase possesses higher affinity but lower processivity for IP3 than the phosphatase, suggesting that the kinases might act as the primary route for IP3 metabolism, whereas INPP5A is available as a rapid and durable way to get rid of any IP3 surplus⁴³. This division of tasks may explain the selective dependency of GNAQ/11^Mut UM cells, which produce excessive amounts of IP3, on INPP5A.

In addition to hydrolyzing IP3, INPP5A also metabolizes IP4 (ref. ⁴⁵). Indeed, IP4 accumulated in INPP5A-depleted cells; however, our results suggest that IP4 accumulation is unlikely to play a major role in driving the dependency of GNAQ/11^Mut cells on INPP5A as such. First, upon INPP5A loss, we observed IP4 accumulation in a few CM cell lines and in parental and GNAQ^WT-expressing HEK293A cells; however, none of these cells accumulated IP3 or suffered from loss of viability. Second, although the physiological role of IP4 is poorly understood, it is known that it does not trigger IP3R-mediate calcium release but instead imposes an inhibitory effect on the receptor’s activity^44,46,47. In our study, however, we showed that INPP5A depletion led to elevated cytosolic calcium preferentially in GNAQ/11^Mut UM cells, suggesting that the receptors are hyperactivated under these conditions. Finally, the co-depletion of INPP5A and IP3R rescued the lethal phenotype of INPP5A loss, confirming that the lethality is mediated explicitly through these receptors. Although we cannot formally rule out a contribution of IP4 to this genetic dependency, our data suggest that the lethality incurred by INPP5A depletion in GNAQ/11^Mut cells is mainly attributed to the accumulation of IP3.

Phosphatases are valuable therapeutic targets for various diseases, including cancer owing to their roles in regulating multiple vital cellular processes⁴⁸. INPP5A belongs to the inositol 5-phosphatase family, which consists of ten mammalian members, all containing a conserved catalytic domain capable of hydrolyzing (phospho)inositides⁴⁹. INPP5A was found to be lost in cutaneous squamous cell carcinoma, glioblastoma and leukemia, but its precise role in promoting tumorigenicity is yet to be demonstrated^42,50,51,52. Developing selective phosphatase inhibitors was deemed challenging owing to the highly charged and conserved nature of the phosphatases’ active sites⁵³; however, recent advances in drug discovery approaches allowed the development of several potent inhibitors, a few of which are currently under clinical assessement⁵⁴. For instance, instead of targeting conserved active sites, newly developed inhibitors bind at allosteric sites beyond the catalytic domains, as successfully demonstrated with the recent discovery of the allosteric inhibitors of the phosphatase SHP2 (ref. ⁵⁵). Moreover, selective irreversible covalent inhibitors have proven beneficial to selectively inhibit historically undruggable phosphatases such as the KRAS oncoprotein. Other approaches, such as protein–protein interaction disruptors or targeted protein degradation are also promising strategies, but they require further understanding of INPP5A biology.

The paucity of effective systemic treatments for primary and metastatic UM underscores the dire need to identify actionable drug targets for this lethal disease. Our study combined transformative functional genomics and comprehensive integrative bioinformatics coupled to profound molecular understanding to discover INPP5A as a synthetic lethal vulnerability in GNAQ/11^Mut UM cells. Through this powerful platform, we also identified other attractive gene candidates that warrant further investigation. They may provide major conceptual advances in UM biology and offer opportunities to uncover therapeutic targets against UM.

Methods

Animal ethics statement

All animal studies were approved by the local ethics committee (Ethikkommisslon Nordwest und Zentralschweiz) and were performed in accordance with the Federal Animal Protection Act and Order. The experiments were carried out according to the Association for Assessment and Accreditation of Laboratory Animal Care International guidelines. Mice were checked for clinical indications, tumor size and body weight as specified in the experimental licenses (BS-2712 and BS-1974). Mice were killed before reaching the approved humane end points of either tumor size limit of 1,500 mm³ or body weight loss of 15% or whenever they showed apparent clinical signs of pain. The maximum tumor size/burden was never exceeded in the studies. Source data are provided for all in vivo experiments.

Patient ethics statement

Patient biopsies from de-identified patients with melanoma were provided by the Department of Dermatology, University Hospital Zurich, Switzerland. Samples were collected and stored by the Biobank after obtaining a written informed consent (Biobank BASEC-NR_2017-00494). The analysis was conducted under the approval of the local ethics committee (Ethikkommisslon Nordwest und Zentralschweiz authorization no. 2021-02243) in accordance with good clinical practice guidelines and the Declaration of Helsinki.

Cell culture and cell engineering

Cell lines were obtained from ATCC: A375 (CRL-1619), MeWo (HTB-65), MP41 (CRL-3297) and MP46 (CRL-3298), Sigma: MEL202 (13012457), DSMZ: RHV421 (ACC127), Leiden University Medical Center – M.J. Jager: 92.1 (CVCL_8607), OMM1 (CVCL_6939), OMM2.5 (CVCL_C307), Mel285 (CVCL_C303) and Mel290 (CVCL_C304) or CCLE: Colo741 (CVCL_1133), K029ax (CVCL_8784) and UACC62 (CVCL_1780). All cell lines were cultured in RPMI 1640 (BioConcept). HEK293A and HEK293FT (Thermo Fisher) and A375 cells were cultured in DMEM high glucose (BioConcept). The culture medium was supplemented with 10–20% fetal bovine serum (Seradigm), 2 mM l-glutamine, 10 mM HEPES, 1 mM sodium pyruvate, 1% NEAA and 100 U g ml⁻¹ penicillin/streptomycin (BioConcept). Tetracycline-free FBS (F2442, Sigma) was used once cells were engineered with doxycycline-inducible constructs. Cas9-expressing cells were generated by the lentiviral delivery of Cas9 followed by selection with 10 µg ml⁻¹ blasticidin or neomycin (Invitrogen). Cas9 expression was determined by flow cytometry (Cas9 antibody, 7A9-3A3, 14697, CST). FACS analysis was performed using CytExpert v.2.4 software (Beckman Coulter). All Cas9-expressing cells used were polyclonal, except for the 92.1-Cas9 and Mel202-Cas9 cell lines, which were obtained from a single clone. A375 cells expressed a variant of Cas9 that is coupled to GFP. Cells expressing doxycycline-inducible shRNAs were obtained by lentiviral transduction of pLKO-TET-ON plasmid or a modified version with a U6 promoter and a GFP cassette.

Lentivirus production and knockout cell line generation

HEK293FT cells were seeded at a density of 5 × 10⁶ in 75-cm² cell culture flasks (Corning). Cells were transfected with 17.5 µg of ready-to-use lentiviral packaging plasmid mix (Cellecta) and 7–10 µg of the indicated constructs using TransIT-293 transfection reagent (Mirus). Viral supernatant was collected 48 and 72 h post-transfection, filtered through 0.45-µm cellulose acetate filters (Millipore) and stored at −80 °C for subsequent use.

TP53 lenti-CRISPR-Cas9 knockout cell lines were generated using a pair of sgRNAs each targeting a different exon. TP53-KO#1 was generated using sgTP53#1 and sgTP53#2, whereas TP53-KO#2 was generated using sgTP53#3 and sgTP53#4. To generate INPP5A, ITPR1, ITPR3 or ERLIN2 CRISPR knockout cell lines, a single sgRNA targeting an exon region was used. The GNAQ–Q209P construct was purchased from Twist Bioscience and the GNAQ–Q209L and GNAQ–R183Q constructs were generated using the Quickchange Lightning Site-Directed Mutagenesis kit following the manufacturer’s protocol (Agilent Gernomics). pDONR-HA-GNAQ (Origene) was used as a template. The DpnI-treated PCR amplicon was transformed into XL1-blue chemocompetent bacteria and mutations were confirmed by Sanger sequencing. The sequences of oligonucleotides are provided in Supplementary Table 3.

Genome-wide CRISPR genetic screen

Genome-wide CRISPR genetic screens were performed as described⁵⁶. In brief, Cas9-expressing UM cells were plated in CellSTACK culture chambers (Corning) and infected with the pooled sgRNA lentivirus at a representation of 1,000 cells per sgRNA with a multiplicity of infection of 0.4. Cells were selected with 2 µg ml⁻¹ puromycin (GIBCO) for 4 d and the percentage of sgRNA-infected mCherry⁺) cells was assessed by FACS. Selected cells were propagated and collected at the indicated time points for genomic DNA isolation. DNA sequences containing sgRNA templates were amplified by PCR and fragments were purified with the Agencourt AMpute XP beads (Beckman Coulter Life Sciences) for deep sequencing using the Illumina HiSeq 4000 platform.

Bioinformatic analysis of genetic screens

For data processing of UM CRISPR screens, the total number of sgRNA reads of each sample was counted and scaled according to the library size. Read counts were normalized using the TMM method in the edgeR Bioconductor package v.3.6.2. For publicly available datasets, CRISPR and RNAi screen data were obtained from DepMap v.2020Q4 (refs. ^17,57). Essentiality and cell line dependency data were downloaded from project Score¹⁴. Gene-summary log FC for the three UM screens were calculated. Data analysis proceeded along the following steps:

1.
Genes with a median logFC of >−0.5 from each of the three UM CRISPR screens were excluded.
2.
Genes with a dependency score from project AVANA (CRISPR), project DRIVE (RNAi) or Achilles (RNAi) of <−0.4 were removed.
3.
Genes with a priority score or cell line dependency score of >0.4 from project Score were excluded.
4.
To filter out melanocyte-lineage essential genes, genes with a median dependency score <−0.5 in CM cell lines were removed and then any gene that has a dependency score of <−0.5 in Mel285 or Mel290 (GNAQ/11^WT UM cell lines) were eliminated.
5.
P values were determined by performing RSA analysis⁵⁸ and were adjusted against multiple testing using an independent hypothesis weighting (IHW) method as described⁵⁹. An adjusted P value cutoff of <0.01 was applied.
6.
Overlapping genes among the three UM screens were identified as final hits.

TCGA and CCLE data analysis

mRNA expression data of TCGA was obtained from the GDC portal and CCLE data from the DepMap portal (v.2022Q1)^19,60. After excluding liquid tumor data, the 5th and 95th percentiles of mRNA expression (TPM values) were calculated for every single gene for all non-UM tumors and median gene expression was computed from UM samples. If the median gene expression in UM was below the fifth percentile gene expression value, we considered it a lowly expressed gene, whereas if it was higher than the 95th percentile, we considered it a highly expressed gene. To exclude melanocyte-specific lineage genes, the median gene expression per tumor lineage was calculated. z-scores were computed and genes with a difference in z score between UM and CM of less than 0.5 were excluded. Synthetic lethality comparisons presented in Extended Data Fig. 3j were analyzed from DepMap v.2022Q1.

RNA sequencing

Cells were plated in six-well plates and treated with 100 ng ml⁻¹ doxycycline to induce INPP5A knockdown. At the indicated time point, cells were collected using 350 µl RLT buffer and lysates were homogenized using QIAshredder spin columns (QIAGEN, 79656). Total RNA was isolated using RNeasy kit according to the manufacturer’s protocol (QIAGEN, 74104). RNA quality and concentration were determined using the Agilent RNA 6000 Nano kit on the Agilent 2100 BioAnalyzer (Agilent). Then, 200 ng RNA was used as input for the library preparation. RNA-seq libraries were generated using the TruSeq RNA Sample Prep kit v.2 and NGS was performed on a HiSeq2500 (Illumina). Gene-level expression estimates were performed using the Pisces algorithm v.0.1.3.1 as described⁶¹, using Ensembl release 97 gene models. Differential gene expression changes between no doxycycline and doxycycline samples were calculated using DESeq2 v.1.26.0 (ref. ⁶²). Common genes between control and INPP5A shRNAs were excluded. Overlapping genes of shINPP5A#1 and shINPP5A#2 were identified and shared genes between the empty vector or wild-type or D384G INPP5A were eliminated. To find significantly altered pathways, GSEA (http://software.broadinstitute.org/gsea/index.jsp) and Fisher’s exact tests were performed with the MSigDB gene set annotations v.7.5.1 using the FGSEA package v.1.22.0 (ref. ⁶³).

Cell growth assays

For clonogenic assays, Cas9-expressing cells were transduced with sgRNAs by lentiviral transduction. At 24 h later, cells were selected with 1 µg ml⁻¹ puromycin (GIBCO) and 72 h after transduction, cells were seeded at a density of 1,000–10,000 cells per well in six-well plates depending on the growth rate of each cell line. Cell lines expressing inducible shRNA constructs were treated with DMSO or 100 ng ml⁻¹ of doxycycline (Sigma-Aldrich) 16 h after seeding. After 10–21 d, colonies were fixed with 25% glutaraldehyde (Fluka, G0068), stained with 0.5% crystal violet (Sigma, CVC3886) and dissolved with 10% acetic acid for quantification.

For knockdown-rescue experiments, shRNA-resistant cDNAs of wild-type or phosphatase-deficient (D384G) HA-tagged INPP5A were cloned into the pLNCX2 retroviral constructs. Cells expressing the doxycycline-inducible control or INPP5A shRNAs were transduced with the retrovirus of the cDNA or empty expression vectors and expression levels were examined by western blotting.

For Incucyte live-cell proliferation assays, indicated cell lines were infected with lentiviral particles of the sgRNAs. At 24 h post-infection, cells were selected with 1–4 µg ml⁻¹ puromycin (Gibco) for 48 h and then seeded at a density of 500–4,000 cells per well in 96-well plates (Corning 3904) depending on the growth rate and the experimental design. Cells were imaged every 6 h using the Incucyte S3 platform (Essen Bioscience). Phase-contrast images were analyzed to assess cell proliferation based on confluence using the Incucyte built-in software.

Proliferation assay was used to determine the effect of the GNAQ/11 inhibitor FR900359 (FR) on the proliferation of examined cell lines. Then, 3,000 cells were seeded into 96-well plates and treated with FR at a concentration range of 0.01 to 100 nM using the HP D300e drug dispenser. After incubation for 96 h at 37 °C, cell proliferation was assessed using the resazurin sodium salt dye reduction assay (alamarBlue assay) using a Synergy HT plate reader (BioTek).

For competition-based proliferation assays, Cas9-expressing HEK293A cells were infected with lentiviral particles of the indicated sgRNAs. Three days after infection, cells were collected and 10,000 cells were plated into two 96-well plates (6–8 replicates per condition) and treated with DMSO or FR (10 nM) if applicable. The remaining cells were stored at −20 °C to measure their inositol phosphate levels by mass spectrometry. Six days after infection, baseline levels of sgRNA-expressing cells were determined by measuring the fraction of mCherry⁺ or GFP⁺ cells by flow cytometry (FACS Calibur, BD Biosciences). At the indicated time point, the percentage of mCherry⁺ or GFP⁺ cells relative to baseline was determined.

Ion-pair chromatography and mass spectrometry

To measure levels of inositol phosphates from cells, 100,000 cells were plated in six-well plates. At the indicated time points, cells were centrifuged at 142g at 4 °C for 5 min, pellets were resuspended in 100 µl extraction buffer (1 M perchloric acid with 3 mM EDTA) and stored at −20 °C. For ion-pair high-performance liquid chromatography-tandem mass spectrometry (IP-HPLC–MS/MS), Precellys ceramic beads (1.4 mm, Bertin Instruments, P000927-LYSK0-A) were added before sonication in an ultrasonic bath for 5 min. Then, 60 µl neutralization buffer (1 M Na₂CO₃ with 3 mM EDTA) was added and supernatants were collected and transferred to a 0.2-µm Captiva ND 96 filter well plate (Agilent, A5969002) mounted on a vacuum manifold and prewashed with 200 µl ACN:water (1:1, v/v). Eluates were collected in 96-well plates and 6 µl 2.5 µg ml⁻¹ internal standard (IS) was added for normalization.

For tumor samples, isolated tumors were homogenized with twofold (w/v) ACN:water (2:8, v/v) and 3 mM EDTA for four cycles of 20 s at 2,460g at 4 °C in Precellys tubes (Bertin Instruments, P000916-LYSK0A) using a Precellys 24 Tissue Homogenizer (Bertin Instruments, 000669-PR240-A). Then, 100 µl extraction buffer was added to 20 µl tumor homogenate in protein LoBind tubes and samples were sonicated for 5 min with Precellys ceramic beads (2.8 mm, Bertin Instruments, P000926-LYSK0A). The extraction was repeated twice and supernatants were combined into 1.2 ml 96-deep-well plates to which 120 µl neutralization buffer with 8 µl IS was added. Plates were stored overnight in a precooled auto sampler before the start of the LC–MS analysis. Prepared samples were analyzed using the Nexera X2 HPLC system platform (Shimadzu) coupled to a triple quadrupole mass spectrometer QTRAP 6500+ (Sciex). Data acquisition was performed with Analyst v.1.7 software.

Quantitative real-time PCR

The 92.1-doxycycline-inducible shRNA cells were plated in six-well plates at a density of 250,000 cells per well. At 72 h after doxycycline treatment (100 ng ml⁻¹), mRNA was extracted using QIAshredder spin columns (QIAGEN, 79656) and RNeasy mini kit (QIAGEN, 74106) according to the manufacturer’s protocol. For quantitative PCR reactions, 32 ng mRNA, GAPDH probe (Hs.PT.39a.22214836), INPP5A probe (Hs00897218_m1) (Integrated DNA Technologies) and iTaq Universal Probes One-Step kit (BioRad, 172-5140) were pipetted in 384-well plates. Quantitative real-time PCR (qRT–PCR) was performed using the probe-based assay with a 7900HT thermal cycler (Applied Biosystems). GAPDH mRNA levels were used for normalization and relative expression levels of genes were calculated with the 2^−ΔΔCt method. To quantify mRNA levels of NOXA, PUMA and CDKN1A, 20,000 92.1-doxycycline-inducible parental or TP53 knockout (KO#1 or KO#2) cells were plated in six-well plates. Then, 7 and 11 d after doxycycline treatment, the total RNA was extracted and cDNA synthesis was performed using MultiScribe reverse transcriptase (Thermo Fisher, 4311235). qRT–PCR was performed using FastStart Universal SYBR green master (Roche Rox, 4913850001). Analysis was performed using the Sequence Detection System (SD) v.2.4.1.

Apoptosis assays

Cells expressing doxycycline-inducible shRNAs were treated with either DMSO or 100 ng ml⁻¹ doxycycline for 3 d before seeding in 96-well plates. The medium was refreshed every 3 d. At the indicated time points, cells were collected and apoptosis was determined using the annexin V/Dead Cell Apoptosis kit (Invitrogen, V13245 and A23204) according to the manufacturer’s protocol. In brief, samples were incubated with 100 µl apoptosis reagent (97.5 µl 1× annexin-binding buffer, 0.5 µl 100 µg ml⁻¹ propidium iodide and 2 µl Alexa Fluor 647 annexin V) for 15 min on ice. Stained cells were examined using flow cytometry using the CytoFLEX-S analyzer (Beckman Coulter). Untreated or staurosporine-treated cells were used as negative and positive control for apoptosis, respectively.

For live-cell analysis of apoptosis, 92.1-Cas9 parental or TP53 knockout (KO#1 and KO#2) cells were infected with lentiviral particles of the indicated sgRNAs. Two days after puromycin selection, cells were plated in 96-well plates (Corning, 3904) and treated with Incucyte annexin V dye (1:200 dilution, Sartorius, 4642). Cells were imaged every 4 h using the Incucyte SX3 or SX5 platforms (Essen Bioscience). Red and phase-contrast images were analyzed to quantify red objects (apoptotic events) and percentage of confluence using the Incucyte built-in software. The number of red objects was normalized to confluence percentage and the cumulative count of normalized red objects (obtained by summing up the number of red objects every 8 h) was plotted.

Cell cycle analysis

For cell cycle analysis, 92.1 cells were seeded in six-well plates. At 16 h post-seeding, DMSO or 100 ng ml⁻¹ doxycycline was added to the cells. At the indicated time points, cells were pulsed with 10 µM EdU (Sigma-Aldrich, BCK-FC647) for 1 h before harvesting. Samples were processed according to the manufacturer’s protocol and then incubated with 3 µM 4′,6-diamidino-2-phenylindole for 5 min before imaging by flow cytometry.

HTRF-based IP1 accumulation assay

To assess IP3 synthesis, accumulation of IP1 was measured as a surrogate of IP3 production using HTRF technology⁶⁴. IP3 is converted to IP1, which is degraded to D-myo-inositol. IP1 degradation is inhibited by LiCl, resulting in its accumulation. Endogenous IP1 accumulation leads to a decrease in FRET signal as it competes with exogenously added acceptor-labeled IP1 for the binding to a donor-labeled anti-IP1 antibody (CisBio, 62IPAPEB). Experimentally, 2,500 cells were plated in 96-well Nunclon plate (Corning) and treated overnight with FR900539 (10 nM). Cells were washed, 70 µl stimulation buffer containing LiCl was added and the plate was incubated at 37 °C for 1 h. Then, 15 µl exogenous IP1 and 15 µl antibody-cryptate were added and the plate was incubated for 1 h at room temperature. The readout was performed with a Tecan M1000pro reader.

Phosphatase activity assay

To assess the phosphatase activity of INPP5A-D384G mutant, 92.1-doxycycline-inducible shRNA cells expressing empty expression vector, wild-type or mutant INPP5A (D384G) cDNAs were plated and treated with 100 ng ml⁻¹ doxycycline. At 72 h post-induction, protein samples were extracted in 25 mM TRIS-HCl, 250 mM sucrose, 500 mM KCl, 1 mM dithiothreitol (DTT), 10 mM MgCl₂ and 2 mM EGTA. Supernatants were collected after centrifuged for 45 min at 16,627g at 4 °C. Then, 10 µg protein was incubated with 1 µg inositol 1, 4, 5 trisphosphate (Cayman Chemical) for 2 min and a malachite green phosphate assay kit was used to measure the levels of inorganic free phosphate according to the manufacturer’s protocol (Cayman Chemical, 10009325).

Microscopy

For cytosolic calcium imaging with fura-2, sgRNA-infected cells were plating in µ-Slide eight-well ibiTreat precoated plates (Ibidi). At the indicated time points, cells were incubated with 5 µM fura-2-AM dye (Invitrogen, F1221) in phenol-red free RPMI medium (BioConcept) for 45 min at 37 °C. Images were acquired on an Olympus inverted IX83 frame equipped with an incubator maintained at 37 °C and 5% CO₂ (CellVivo, PECON). Images were recorded with an Olympus silicon-immersed ×30 objective. For UV excitation, a X-Cite Exact 200W Hg light source (Excelitas Technologies) was used. Images were acquired using a Hamamatsu Orca Flash 4.0 CCD camera. Then, 2 µM ionomycin (Sigma-Aldrich, I9657-1MG) was used as a positive control. Image analysis and quantification was done using the Olympus cellSens v.3.1.1software.

For live-cell calcium imaging and cell-fate analysis, a dCys-GCaMP6 calcium biosensor was used, which is composed of calmodulin calcium-sensing and M13 peptide-binding elements and a GFP moiety that is activated by binding to calcium. dCys-GCaMP6-expressing 92.1-Cas9, MeWo-Cas9 or Uacc62-Cas9 cells were seeded into 96-well CellCarrierUltra plates (PerkinElmer). At 16 h later, cells were transduced with sgRNAs by lentiviral transduction. At 24 h post-transduction, cells were treated with DMSO or thapsigargin 1.5 nM (Thermo Fisher, T7458) before imaging. Confocal time-lapse images were captured every 6 h for 3 d on an Opera Phenix High Content Imager (PerkinElmer) using ×20/1.0 water objective with sequential exposures for GFP and mCherry. A total of 53 fields were acquired per well and analyzed using the Harmony image analysis software package v.4.9 (PerkinElmer). For analysis, primary cell objects were identified based on mCherry positivity. Mean GFP intensity for each cell was calculated. Single cell values were used to calculate the well median of mean GFP intensity per cell (Spotfire, Tibco). Round cells were classified based on area and roundness thresholds to calculate percentage of round (dead) cells; at least 660 objects per well were counted. For single-cell-tracking analysis, fields were selected randomly and the mean GFP signal intensity was recorded per time frame until the cell either died or divided. Cells that did not change their fate until the end of recording were excluded.

Western blot

Total cell lysates were incubated with lysis buffer consisting of 50 mM Tris-HCl pH 7.8, 0.01% NP40 (Fluka, 74385), 120 mM NaCl (Fluka, 71380), 25 mM NaF (Merck, 106450), 40 mM β-glycerol phosphate (Fluka, 50020), 100 µM sodium metavanadate (Sigma, 590088), 1 mM DTT (Sigma, 43815), 100 µM phenylmethyl sulfonyl fluoride (Sigma, P7626), 1 mM benzamidine (Sigma, B6506) and 1 µM microcystin (Alexis Biochem, 350012). Protein levels were determined using a Bradford assay (BioRad, 500-0006). Equal protein amounts were loaded into NuPage precast polyacrylamide gels (Invitrogen), transferred to methanol-activated PVDF membranes (Millipore) and 5% milk-blocked membranes were incubated with the following antibodies at 4 °C overnight; GNAQ (1:1,000 dilution, D5V1B, 14373, CST); α-tubulin (1:1,000 dilution, T9026, Sigma-Aldrich); α-actinin (1:1,000 dilution, E7U1O, 69758, CST); ERLIN2 (1:500 dilution, NB100-1884, Novus Biologicals); IP3R3 (1:1,000 dilution, 610312, BD Biosciences); IP3R1 (1:500 dilution, NB120-5908, Novus Biologicals); c-PARP (1:1,000 dilution, 9541, CST); p53 (1:500 dilution, 2527, CST); p21 (1:1,000 dilution, DCS60, 2946, CST); p53-phosphoSer15 (1:1,000 dilution, 9284, CST); HA (1:1,000 dilution, ab9110, Abcam); actin (1:20,000 dilution, C4, MAB1501, Millipore); BAP1 (1:1,000 dilution, D7W70, 13271, CST); pERK (1:1,000 dilution, 9101, CST) and pRASGRP3 (1:1,000 dilution, 124823, Abcam). Membranes were washed with PBS with 0.1% Tween-20 solution and incubated with secondary antibodies; anti-rabbit IgG (1:5,000 dilution, NA931-VS, Amersham); anti-mouse IgG (1:2,500 dilution, 7076, CST) or anti-goat IgG (1:20,000 dilution, A5420, Merck) for 1 h at room temperature. Membranes were imaged using the Fusion FX7 imager (Vilber) after incubation with ECL solution (Amersham, NA931-VS).

In vivo mouse experiments

Animals were kept under optimal hygiene conditions in Allentown mice cages (maximum of five animals per cage) with a 12-h dark–light conditions and they were fed sterilized food and water ad libitum. For mouse xenograft experiments, naive female athymic nude mice, 6–8 weeks, 20–25 g (Charles River Laboratories) were injected subcutaneously with 5 × 10⁶ of 92.1 or Mel202 cells expressing doxycycline-inducible shRNAs suspended in 200 µl 50% Matrigel (Corning, 354234) and 50% HBSS (Sigma, H6648). Experiments were initiated once the mean size of tumors reached around 200 mm³ (13–19 d post-implant) by administering doxycycline (25 mg kg⁻¹) daily using oral gavage. Tumors were measured using caliper and tumor volumes were calculated using a modified ellipsoid formula: tumor volume = L × W² × π/6, where L is the longest axis of the tumor and W is perpendicular to L.

For liver metastasis experiments, 6–8-week-old NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) female mice (Taconic) were used. Liver metastasis tumors were established by injecting 2 × 10⁶ 92.1 expressing firefly luciferase and doxycycline-inducible shRNAs i.v. via the lateral tail vein. Doxycycline (25 mg kg⁻¹) was given daily starting at the indicated time points using oral gavage or in drinking water. For in vivo imaging, mice received d-luciferin (150 mg kg⁻¹, 10 ml kg⁻¹) through tail vein injections before being anesthetized using isoflurane (2%). Firefly bioluminescence was measured by an In Vivo Imaging System (IVIS) Lumina XR (Xenogen) 6–10 min after luciferin injection using a highly sensitive, cooled CCD camera. Imaging and quantification of signals were conducted by image software Living Image v.4.7.4 (64-bit) (Xenogen). Bioluminescence was measured at days 10, 21, 31, 41, 51 and 61 post-tumor cell injections.

Histopathology and immunohistochemistry

For patient tumors, the corresponding FFPE blocks of the slow-frozen biopsies were cut (4-µm thickness) and processed as described⁶⁵. Tumor xenograft FFPE blocks were sliced into 3-µm sections and H&E staining and immunohistochemistry were prepared as described⁶⁶. For patient samples, slides were stained with H&E and Melan A (1:200 dilution, Novus, NBP1-30151). Tumor content was evaluated by two independent pathologists. Staining was performed on a Leica BondRX platform (Leica Microsystems). Antibodies were diluted with diluent antibody and alkaline phosphatase chromogen (Leica Bond kit) was used for detection (Bond Polymer Refine Red Detection kit). Ki67 Immunohistochemical staining was performed using Epitope Retrieval 2 (ER2) pretreatment conditions for 20 min at 100 °C. Refine Red kit (Leica Biosystems) was used for detection and the primary antibody was anti-human Ki67 (1:2,000 dilution, SP6, Neomarker, RM9106). Slides were then hematoxylin counterstained, dehydrated, cover-slipped and digitalized with Aperio AT2. Ki67 quantification was conducted using the HALO image analysis platform (v.3.1.1076.437) (Indica Labs) CytoNuclear IHC algorithm.

Patient-derived xenografts

UM PDXs were purchased as snap-frozen samples from Xenostart and Charles River laboratories. CM PDXs and one UM PDX were grown in-house and are the property of the Novartis Institute for Biomedical Research.

Statistics and reproducibility

Data are presented as indicated in the respective figure legends. In vitro experiments were performed in biological triplicate or duplicate. No statistical method was employed to predetermine sample size. Data collection and analysis were not performed blind to the conditions of the experiments. Data distribution was assumed to be normal, but this was not formally tested. For in vivo experiments, mice were randomized into control and treatment groups within their genotype, while maintaining comparable tumor volume averages per group. No data were excluded from the analysis, except for Fig. 6a, where one mouse was excluded as it did not develop a tumor, and Fig. 6d, when one mouse died under anesthesia. A few patient biopsies were excluded as they failed to meet the quality control criteria (Reporting Summary provides details).

All data are presented as mean ± s.e.m. No assumptions on data distribution were made. Statistical analysis was performed with GraphPad Prism 9 software using a two-tailed, unpaired Student’s t-test, one-way ANOVA, two-way ANOVA or multiple Student’s t-test. Corrections for multiple comparisons were performed according to the recommended method by the software. Correlation analysis was performed using Pearson’s correlation test and R² values are displayed. P values of <0.05 were considered statistically significant for all experiments. Box plots show the 25th to 75th percentiles, the center line depicts the median and whiskers depict minimum and maximum values.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

RNA-seq data presented in this study were deposited into the Sequence Read Archive under accession no. PRJNA860930. RNA expression data derived from TCGA data were downloaded from the GDC data portal (GDC (cancer.gov)). Cancer cell line dependency and RNA expression data were downloaded from DepMap (DepMap: The Cancer Dependency Map Project at Broad Institute) and internal RNA-seq datasets. UM CRISPR screen data are shown in Supplementary Table 1. All other data that support the findings of this study can be made available from the corresponding author upon request. Source data are provided with this paper.

Code availability

No custom algorithms were used in this study. Open-source software was used to analyze data. Details of software versions are specified in the Methods.

References

Jager, M. J. et al. Uveal melanoma. Nat. Rev. Dis. Primers 20, 18–20 (2020).
Google Scholar
Carvajal, R. D. et al. Advances in the clinical management of uveal melanoma. Nat. Rev. Clin. Oncol. 20, 99–115 (2023).
Article PubMed Google Scholar
Shain, A. H. et al. The genetic evolution of metastatic uveal melanoma. Nat. Genet. 51, 1123–1130 (2019).
Article CAS PubMed PubMed Central Google Scholar
Zuidervaart, W. et al. Activation of the MAPK pathway is a common event in uveal melanomas although it rarely occurs through mutation of BRAF or RAS. Br. J. Cancer 92, 2032–2038 (2005).
Article CAS PubMed PubMed Central Google Scholar
Van Raamsdonk, C. D. et al. Frequent somatic mutations of GNAQ in uveal melanoma and blue naevi. Nature 457, 599–602 (2009).
Article PubMed Google Scholar
Van Raamsdonk, C. D. et al. Mutations in GNA11 in Uveal Melanoma. New Engl. J. Med. 363, 2191–2199 (2010).
Article PubMed Google Scholar
Johansson, P. et al. Deep sequencing of uveal melanoma identifies a recurrent mutation in PLCB4. Oncotarget 7, 4624–4631 (2016).
Article PubMed Google Scholar
Moore, A. R. et al. Recurrent activating mutations of G-protein-coupled receptor CYSLTR2 in uveal melanoma. Nat. Genet. 48, 675–680 (2016).
Article CAS PubMed PubMed Central Google Scholar
Campbell, A. P. & Smrcka, A. V. Targeting G protein-coupled receptor signalling by blocking G proteins. Nat. Rev. Drug. Discov. 17, 789–803 (2018).
Article CAS PubMed PubMed Central Google Scholar
Kostenis, E., Pfeil, E. M. & Annala, S. Heterotrimeric Gq proteins as therapeutic targets? J. Biol. Chem. 295, 5206–5215 (2020).
Article CAS PubMed PubMed Central Google Scholar
Onken, M. D. et al. Targeting nucleotide exchange to inhibit constitutively active G protein a subunits in cancer cells. Sci. Signal. 11, eaao6852 (2018).
Article PubMed PubMed Central Google Scholar
Berstein, G. et al. Phospholipase C-β1 is a GTPase-activating protein for Gq/11, its physiologic regulator. Cell 70, 411–418 (1992).
Article CAS PubMed Google Scholar
Hubbard, K. B. & Hepler, J. R. Cell signalling diversity of the Gqα family of heterotrimeric G proteins. Cell Signal. 18, 135–150 (2006).
Article CAS PubMed Google Scholar
Behan, F. M. et al. Prioritization of cancer therapeutic targets using CRISPR–Cas9 screens. Nature 568, 511–516 (2019).
Article CAS PubMed Google Scholar
McDonald, E. R. et al. Project DRIVE: A compendium of cancer dependencies and synthetic lethal relationships uncovered by large-scale, deep RNAi screening. Cell 170, 577–592 (2017).
Article CAS PubMed Google Scholar
Tsherniak, A. et al. Defining a cancer dependency map. Cell 170, 564–576 (2017).
Article CAS PubMed PubMed Central Google Scholar
Meyers, R. M. et al. Computational correction of copy number effect improves specificity of CRISPR-Cas9 essentiality screens in cancer cells. Nat. Genet. 49, 1779–1784 (2017).
Article CAS PubMed PubMed Central Google Scholar
Barretina, J. et al. The cancer cell line encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 483, 603–607 (2012).
Article CAS PubMed PubMed Central Google Scholar
Ghandi, M. et al. Next-generation characterization of the cancer cell line encyclopedia. Nature 569, 503–508 (2019).
Article CAS PubMed PubMed Central Google Scholar
Chen, X. et al. RasGRP3 mediates MAPK pathway activation in GNAQ mutant uveal melanoma. Cancer Cell 31, 685–6966 (2017).
Article CAS PubMed PubMed Central Google Scholar
Robertson, A. G. et al. Integrative analysis identifies four molecular and clinical subsets in uveal melanoma. Cancer Cell 32, 204–220 (2017).
Article CAS PubMed PubMed Central Google Scholar
Karlsson, J. et al. Molecular profiling of driver events in metastatic uveal melanoma. Nat. Commun. 11, 1–13 (2020).
Article Google Scholar
Tresaugues, L. et al. Structural basis for phosphoinositide substrate recognition, catalysis, and membrane interactions in human inositol polyphosphate 5-phosphatases. Structure 22, 744–755 (2014).
Article CAS PubMed Google Scholar
Whisstock, J. C. et al. The inositol polyphosphate 5-phosphatases and the apurinic/apyrimidinic base excision repair endonucleases share a common mechanism for catalysis. J. Biol. Chem. 275, 37055–37061 (2000).
Article CAS PubMed Google Scholar
Chan, E. M. et al. WRN helicase is a synthetic lethal target in microsatellite unstable cancers. Nature 568, 551–556 (2019).
Article CAS PubMed PubMed Central Google Scholar
Sulahian, R. et al. Synthetic lethal interaction of SHOC2 depletion with MEK inhibition in RAS-driven cancers. Cell Rep. 29, 118–134 (2019).
Article CAS PubMed PubMed Central Google Scholar
Laxminarayan, K. M., Matzaris, M., Speed, C. J. & Mitchell, C. A. Purification and characterization of a 43-kDa membrane-associated inositol polyphosphate 5-phosphatase from human placenta. J. Biol. Chem. 268, 4968–4974 (1993).
Article CAS PubMed Google Scholar
Zhang, S., Yang, W., Zhao, Q., Zhou, X. & Fan, Y. Rapid method for simultaneous determination of inositol phosphates by IPC-ESI–MS/MS and its application in nutrition and genetic research. Chromatographia 80, 275–286 (2017).
Article CAS Google Scholar
Ma, J., Weng, L., Bastian, B. C., Chen, X. & Bastian, B. C. Functional characterization of uveal melanoma oncogenes. Oncogene 40, 806–820 (2021).
Article CAS PubMed Google Scholar
Prole, D. L. & Taylor, C. W. Structure and function of IP3 receptors. Cold Spring Harb. Perspect. Biol. 11, 1–18 (2020).
Google Scholar
Pearce, M. M. P., Wang, Y., Kelley, G. G. & Wojcikiewicz, R. J. H. SPFH2 mediates the endoplasmic reticulum-associated degradation of inositol 1,4,5-trisphosphate receptors and other substrates in mammalian cells. J. Biol. Chem. 282, 20104–20115 (2007).
Article CAS PubMed Google Scholar
Speed, C. J., Neylon, C. B., Little, P. J. & Mitchell, C. A. Underexpression of the 43 kDa inositol polyphosphate 5-phosphatase is associated with spontaneous calcium oscillations and enhanced calcium responses following endothelin-1 stimulation. J. Cell Sci. 112, 669–679 (1999).
Article CAS PubMed Google Scholar
Malek, M. et al. Inositol triphosphate-triggered calcium release blocks lipid exchange at endoplasmic reticulum-Golgi contact sites. Nat. Commun. 12, 2673 (2021).
Article CAS PubMed PubMed Central Google Scholar
Orrenius, S., Zhivotovsky, B. & Nicotera, P. Regulation of cell death: the calcium-apoptosis link. Nat. Rev. Mol. Cell Biol. 4, 552–565 (2003).
Article CAS PubMed Google Scholar
Cai, B. et al. A cell-based functional assay using a green fluorescent protein-based calcium indicator dCys-GCaMP. Assay Drug Dev. Technol. 12, 342–351 (2014).
Article CAS PubMed PubMed Central Google Scholar
Loughery, J., Cox, M., Smith, L. M. & Meek, D. W. Critical role for p53-serine 15 phosphorylation in stimulating transactivation at p53-responsive promoters. Nucleic Acids Res. 42, 7666–7680 (2014).
Article CAS PubMed PubMed Central Google Scholar
Villunger, A., Michalak, E. M., Coultas, L., Adams, J. M. & Strasser, A. p53- and drug-induced apoptotic responses mediated by BH3-only proteins puma and noxa. Science 302, 1036–1038 (2003).
Article CAS PubMed Google Scholar
Sugase, T. et al. Development and optimization of orthotopic liver metastasis xenograft mouse models in uveal melanoma. J. Transl. Med. 18, 1–14 (2020).
Article Google Scholar
Otto, J. C., Kelly, P., Chiou, S. T. & York, J. D. Alterations in an inositol phosphate code through synergistic activation of a G protein and inositol phosphate kinases. Proc. Natl Acad. Sci. USA 104, 15653–15658 (2007).
Article CAS PubMed PubMed Central Google Scholar
Huang, A., Garraway, L. A., Ashworth, A. & Weber, B. Synthetic lethality as an engine for cancer drug target discovery. Nat. Rev. Drug Discov. 19, 23–38 (2020).
Article CAS PubMed Google Scholar
Irvine, R. F. & Schell, M. J. Back in the water: the return of the inositol phosphates. Nat. Rev. Mol. Cell Biol. 2, 327–338 (2001).
Article CAS PubMed Google Scholar
Speed, C. J., Little, P. J., Hayman, J. A. & Mitchell, C. A. Underexpression of the 43 kDa inositol polyphosphate 5-phosphatase is associated with cellular transformation. EMBO J. 15, 4852–4861 (1996).
Article CAS PubMed PubMed Central Google Scholar
Irvine, R. F., Lloyd-Burton, S. M., Yu, J. C. H., Letcher, A. J. & Schell, M. J. The regulation and function of inositol 1,4,5-trisphosphate 3-kinases. Adv. Enzyme Regul. 46, 314–323 (2006).
Article CAS PubMed PubMed Central Google Scholar
Irvine, R. F., Letcher, A. J., Lander, D. J. & Berridge, M. J. Specificity of inositol phosphate-stimulated Ca2⁺ mobilization from Swiss-mouse 3T3 cells. Biochem. J. 240, 301–304 (1986).
Article CAS PubMed PubMed Central Google Scholar
Irvine, R. F., Letcher, A. J., Heslop, J. P. & Berridge, M. J. The inositol tris/tetrakisphosphate pathway-demonstration of Ins(1,4,5)P3 3-kinase activity in animal tissues. Nature 320, 631–634 (1986).
Article CAS PubMed Google Scholar
Taylor, C. W. & Tovey, S. C. IP3 Receptors: toward understanding their activation. Cold Spring Harb. Perspect. Biol. 2, 1–23 (2010).
Article Google Scholar
Bird, G. S. J. & Putney, J. W. Effect of inositol 1,3,4,5-tetrakisphosphate on inositol trisphosphate-activated Ca2⁺ signaling in mouse lacrimal acinar cells. J. Biol. Chem. 271, 6766–6770 (1996).
Article CAS PubMed Google Scholar
Stanford, S. M. & Bottini, N. Targeting protein phosphatases in cancer immunotherapy and autoimmune disorders. Nat. Rev. Drug Discov. 22, 273–294 (2023).
Article CAS PubMed PubMed Central Google Scholar
Ooms, L. M. et al. The role of the inositol polyphosphate 5-phosphatases in cellular function and human disease. Biochem. J. 419, 29–49 (2009).
Article CAS PubMed Google Scholar
Sekulic, A. et al. Loss of inositol polyphosphate 5-phosphatase is an early event in development of cutaneous squamous cell carcinoma. Cancer Prevent. Res. 3, 1277–1283 (2010).
Article Google Scholar
Milinkovic, V. et al. Identification of novel genetic alterations in samples of malignant glioma patients. PLoS ONE 8, e82108 (2013).
Article PubMed PubMed Central Google Scholar
Mengubas, K. et al. Inactivation of calcium ion-regulating inositol polyphosphate second messengers is impaired in subpopulations of human leukemia cells. Leukemia 8, 1718–1725 (1994).
CAS PubMed Google Scholar
Stanford, S. M. & Bottini, N. Targeting tyrosine phosphatases: time to end the stigma. Trends Pharmacol. Sci. 38, 524–540 (2017).
Article CAS PubMed PubMed Central Google Scholar
Mullard, A. Phosphatases start shedding their stigma of undruggability. Nat. Rev. Drug Discov. 17, 847–849 (2018).
Article CAS PubMed Google Scholar
Chen, Y. N. P. et al. Allosteric inhibition of SHP2 phosphatase inhibits cancers driven by receptor tyrosine kinases. Nature 535, 148–152 (2016).
Article MathSciNet CAS PubMed Google Scholar
Liu, H. et al. Tumor-derived IFN triggers chronic pathway agonism and sensitivity to ADAR loss. Nat. Med. 25, 95–102 (2019).
Article CAS PubMed Google Scholar
Dempster, J. M. et al. Extracting biological insights from the project Achilles genome-scale CRISPR screens in cancer cell lines. Preprint at bioRxiv https://doi.org/10.1101/720243 (2019).
Article Google Scholar
König, R. et al. A probability-based approach for the analysis of large-scale RNAi screens. Nat. Methods 4, 847–849 (2007).
Article PubMed Google Scholar
Ignatiadis, N., Klaus, B., Zaugg, J. B. & Huber, W. Data-driven hypothesis weighting increases detection power in genome-scale multiple testing. Nat. Methods 13, 577–580 (2016).
Article CAS PubMed PubMed Central Google Scholar
Weinstein, J. N. et al. The Cancer Genome Atlas pan-cancer analysis project. Nat. Genet. 45, 1113–1120 (2013).
Article PubMed PubMed Central Google Scholar
Shirley, M. D., Radhakrishna, V. K., Golji, J. & Korn, J. M. PISCES: a package for rapid quantitation and quality control of large-scale mRNA-seq datasets. Preprint at bioRxiv https://doi.org/10.1101/2020.12.01.390575 (2020).
Article PubMed PubMed Central Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 1–21 (2014).
Article Google Scholar
Korotkevich, G. et al. Fast gene set enrichment analysis. Preprint at bioRxiv https://doi.org/10.1101/060012 (2021).
Nørskov-Lauritsen, L., Thomsen, A. R. B. & Bräuner-Osborne, H. G protein-coupled receptor signaling analysis using homogenous time-resolved Förster resonance energy transfer (HTRF) technology. Int. J. Mol. Sci. 15, 2554–2572 (2014).
Article PubMed PubMed Central Google Scholar
Sakr, R. A. et al. Protocol for PTEN expression by immunohistochemistry in formalin-fixed paraffin-embedded human breast carcinoma. Appl. Immunohistochem. Mol. Morphol. 18, 371–374 (2010).
Article CAS PubMed PubMed Central Google Scholar
Chapeau, E. A. et al. Resistance mechanisms to TP53-MDM2 inhibition identified by in vivo piggyBac transposon mutagenesis screen in an Arf^−/− mouse model. Proc. Natl Acad. Sci. USA 114, 3151–3156 (2017).
Article CAS PubMed PubMed Central Google Scholar
Wang, T., Birsoy, K., Hughes, N. W., Krupczak, K. M., Post, Y., Wei, J. J., Lander, E. S. & Sabatini, D. M. Identification and characterization of essential genes in the human genome. Science 350, 1096–1101 (2015).
Harbour, J. W., Onken, M. D., Roberson, E. D. O., Duan, S., Cao, L., Worley, L. A., Council, M. L., Matatall, K. A., Helms, C. & Bowcock, A. M. Frequent mutation ofBAP1in metastasizing uveal melanomas. Science 330, 1410–1413 (2010).
Hichri, H., Rendu, J., Monnier, N., Coutton, C., Dorseuil, O., Poussou, R. V., Baujat, G., Blanchard, A., Nobili, F., Ranchin, B., Remesy, M., Salomon, R., Satre, V. & Lunardi, J. From Lowe syndrome to Dent disease: correlations between mutations of the OCRL1 gene and clinical and biochemical phenotypes. Hum. Mutat. 32, 379–388 (2011).

Download references

Acknowledgements

We thank the following members of the Novartis Institute for Biomedical Research, Basel, Switzerland; members of the Manchado group for helpful discussions, E. Dammassa, S. Tobler, K. Daniel and J. Loretan for technical assistance, Y. Mesrouze for help with the structural model, A. Bill and R. Lopes for sharing plasmids, E. Billy, P. Megel and O. Weber for providing reagents and advice on CRISPR screens. We are grateful to G. Galli and C. Fritsch for discussions and T. Schmelzle for discussion and critical reading of the manuscript. We thank D. Bonenfant for her help with earlier experiments. The authors received no specific funding for this work.

Author information

These authors contributed equally: Ali Amin-Mansour, Anne Haberkorn, Claudia Textor.

Authors and Affiliations

Oncology, Novartis Institute for Biomedical Research, Basel, Switzerland
Ahmed M. O. Elbatsh, Anne Haberkorn, Nicolas Ebel, Emilie Renard, Lisa M. Koch, Femke C. Groenveld, Vincent Romanet, Christian Schnell, Markus Wartmann, Sauveur-Michel Maira & Eusebio Manchado
Oncology, Novartis Institute for Biomedical Research, Cambridge, MA, USA
Ali Amin-Mansour, Michelle Piquet, David A. Ruddy & Matthew D. Shirley
PK Sciences, Novartis Institute for Biomedical Research, Basel, Switzerland
Claudia Textor & Peter Wipfli
Chemical Biology and Therapeutics, Novartis Institute for Biomedical Research, Basel, Switzerland
Ulrike Naumann & Martin Rausch
Dermatology Department, University Hospital Zurich, University of Zurich, Zurich, Switzerland
Julia M. Martínez Gómez & Mitchell P. Levesque
Department of Ophthalmology, Leiden University Medical Center, Leiden, The Netherlands
Martine J. Jager

Authors

Ahmed M. O. Elbatsh
View author publications
You can also search for this author in PubMed Google Scholar
Ali Amin-Mansour
View author publications
You can also search for this author in PubMed Google Scholar
Anne Haberkorn
View author publications
You can also search for this author in PubMed Google Scholar
Claudia Textor
View author publications
You can also search for this author in PubMed Google Scholar
Nicolas Ebel
View author publications
You can also search for this author in PubMed Google Scholar
Emilie Renard
View author publications
You can also search for this author in PubMed Google Scholar
Lisa M. Koch
View author publications
You can also search for this author in PubMed Google Scholar
Femke C. Groenveld
View author publications
You can also search for this author in PubMed Google Scholar
Michelle Piquet
View author publications
You can also search for this author in PubMed Google Scholar
Ulrike Naumann
View author publications
You can also search for this author in PubMed Google Scholar
David A. Ruddy
View author publications
You can also search for this author in PubMed Google Scholar
Vincent Romanet
View author publications
You can also search for this author in PubMed Google Scholar
Julia M. Martínez Gómez
View author publications
You can also search for this author in PubMed Google Scholar
Matthew D. Shirley
View author publications
You can also search for this author in PubMed Google Scholar
Peter Wipfli
View author publications
You can also search for this author in PubMed Google Scholar
Christian Schnell
View author publications
You can also search for this author in PubMed Google Scholar
Markus Wartmann
View author publications
You can also search for this author in PubMed Google Scholar
Martin Rausch
View author publications
You can also search for this author in PubMed Google Scholar
Martine J. Jager
View author publications
You can also search for this author in PubMed Google Scholar
Mitchell P. Levesque
View author publications
You can also search for this author in PubMed Google Scholar
Sauveur-Michel Maira
View author publications
You can also search for this author in PubMed Google Scholar
Eusebio Manchado
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

A.M.O.E., A.H., E.R., L.M.K. and F.C.G. performed biological experiments. A.A.M. conducted the computational analyses under the supervision of M.D.S. N.E., V.R. and C.S. supervised in vivo experiments. U.N. and D.A.R. supported NGS. C.T. performed mass spectrometry experiments under the supervision of P.W. M.W. and M.R. supervised microscopy experiments. M.P.L. and J.M.M.G. provided human biopsies. M.J.J. provided cell lines. S.M. and E.M. supervised the whole project. A.M.O.E. and E.M. wrote the manuscript with input from all authors. Illustrations in Figs. 1a, 3a,j, 6d and 7h and Extended Data Fig. 5c were created with BioRender.com.

Corresponding author

Correspondence to Eusebio Manchado.

Ethics declarations

Competing interests

All authors except M.J.J., J.M.M.G. and M.P.L. are/were employees of Novartis Pharma during their time of contribution to this study. The authors declare no competing interests.

Peer review

Peer review information

Extended Data Fig. 6 Tight regulation of IP3 receptors is crucial for GNAQ/11-mutant UM cells survival.

a, Table summarizing CCLE mRNA expression (TPM values) data of ITPR1, ITPR2 and ITPR3 (encoding IP3R1, IP3R2 and IP3R3 subtypes, respectively) in OMM1 and 92.1 cells. IP3-receptor subtype with the highest expression in each cell line is shown in red. b,c, Levels of inositol phosphates (IP1, IP2, IP3 and IP4) in 92.1-Cas9 (b) and OMM1-Cas9 (c) cells expressing the indicated sgRNAs. Cells transduced with sgITPR3#1 or sgITPR3#2 (92.1-Cas9) and sgITPR1#1 or sgITPR1#2 (OMM1-Cas9) were co-transduced with sgINPP5A#1 or non-targeting sgRNA and harvested 72 h later. Results are shown relative to NT control sgRNA and dashed line depicts equivalent levels to control, n = 3 independent experiments. NT, non-targeting. d, e, f, Growth of Mel202-Cas9 (d), 92.1-Cas9 (e) and Uacc62-Cas9 (f) cells determined by Incucyte live-cell analysis following the transduction with the indicated sgRNAs. ERLIN2 is depleted with two independent sgRNAs (sgERLIN2#1 and #2), sgNT and sgPLK1 are negative and positive control sgRNAs, respectively, n= 2 independent experiments. NT, non-targeting. g, h, i, Box plots showing TCGA mRNA expression data (TPM values) of ITPR1 (encoding IP3R subtype 1) (g), ITPR2 (encoding IP3R subtype 2) (h) and ITPR3 (encoding IP3R subtype 3) (i) in human UM and CM tumors. UM, uveal melanoma; CM, cutaneous melanoma. All data are presented as mean ± s.e.m.; P values were determined by two-way analysis of variance (ANOVA) with multiple comparisons (b, c), one-way ANOVA with multiple comparisons (d, e, f), or two-tailed, unpaired Student’s t-test (g, h, i). Box plots show 25th to 75th percentiles, center line depicts median, and whiskers depict minimum and maximum values (g, h, i).

Source data

Extended Data Fig. 7 INPP5A depletion preferentially results in elevated cytosolic calcium levels in GNAQ/11-mutant UM cells.

a, Cumulative time course illustrating the response of fura-2 signal to changes in intracellular calcium levels. The curves depict fura-2 fluorescence signal measured at excitation wavelength of 340 nm (left y axis) and 380 nm (right y axis). Ionomycin (2 µM) was added to cell culture medium at ~40 seconds after the start of imaging to induce changes in cytosolic calcium levels. b, Cumulative time course illustrating intensities of fura-2 fluorescence ratio (F₃₄₀/F₃₈₀) produced by excitation at two wavelengths (340 nm and 380 nm). The increase in cytosolic calcium levels results in an increased ratio of ion-bound indicator (measured at 340 nm) over ion-free indicator (measured at 380 nm). c, d, Quantification of fura-2 fluorescence signal ratio (F₃₄₀/F₃₈₀) 24 h (c) and 48 h (d) after transduction with the indicated sgRNAs in the depicted cell lines. Each point represents a cell; n indicates the cell number used for quantification in each group. A representative of n = 3 for 92.1 and OMM1 and n = 2 for Mel202 and Colo741 independent experiments is shown. e, Quantification of normalized GFP signal intensity of Uacc62-Cas9 cells transduced with the indicated sgRNAs. The calcium ATPase inhibitor Thapsigargin (1.5 nM) was added before the start of imaging. Measured signal represents median per well, n = 2 independent experiments. f, Quantification of the percentage of dead cells of Uacc62-Cas9 cells transduced with the indicated sgRNAs and treated as shown in (e), n = 2 independent experiments. All data are presented as mean ± s.e.m.; P values were determined by one-way analysis of variance (ANOVA) with multiple comparisons (c, d).

Source data

Extended Data Fig. 8 Cellular phenotypes induced by depleting INPP5A in UM cells.

a, Scatter plots of Log₁₀ fold change gene expression as determined by RNA-seq upon depletion of INPP5A in 92.1 cell line using two independent shINPP5A (#1 and #2). Cells expressing doxycycline-inducible shRNA and overexpressing empty expression vector (left), phosphatase-deficient INPP5A (INPP5A-D384G) cDNA (middle) or wild-type INPP5A (INPP5A-WT) cDNA (right) were harvested 7 days following doxycycline treatment. Correlations between Log₁₀ FC of shINPP5A #1 and shINPP5A #2 are shown for all detected genes (depicted as gray dots and correlation shown as black lines) or only for significant differentially expressed genes (DEGs; n = 84, depicted as blue dots and correlation shown as red lines). Profiles of cells transduced with shINPP5A#1 (x axis) and shINPP5A#2 (y axis) showed good correlation for the DEGs. b, c, Gene set enrichment analyses (GSEA) signature enrichment plots for E2F targets (b) and G2/M checkpoint (c) HALLMARK gene sets for the DEGs depicted in (a). NES, normalized enrichment score. d, Immunoblot showing levels of cleaved PARP (c-PARP), p53-phosphorylated serine 15 (p53-pS15) and p21 in doxycycline-inducible shRNA-expressing 92.1 cells 11 days after doxycycline treatment. α-tubulin is used as loading control. e, Stacked bar graph showing percentage of early and late apoptotic cells based on annexin V and propidium iodide staining in doxycycline-inducible shRNA-expressing 92.1 cells at 4, 8 and 11 days after doxycycline treatment. A representative of n = 3 independent experiments is shown. Dox, doxycycline. f, Bar graph showing percentage of apoptotic cells based on annexin V and propidium iodide staining in 92.1-Cas9 cells at day 11 after transduction with the indicated sgRNAs, n = 3 independent experiments. g, Quantitative real-time PCR (qRT–PCR) analysis of CDKN1A in parental cells and two independent TP53 knockout (TP53-KO1 and TP53-KO2) 92.1-Cas9 cell lines after transduction with the indicated sgRNAs, n = 3 biologically independent samples. h, FACS plots depicting the gating strategy for determining different cell cycle phases. Debris and dead cells were excluded according to the forward scatter-area (FSC-A) and side scatter-area (SSC-A) gating. Singlets were identified based on FSC-A and forward scatter-height (FSC- H) profiles. Then, they were analyzed for propidium iodide (DNA content) and 5-ethynyl-2 deoxyuridine (EdU) coupled to Alexa Fluor 647 (EdU–647) staining intensities. As EdU is incorporated during active DNA synthesis, EdU⁺ cells were classified as ‘S phase’. EdU⁻ cells were classified as ‘G1 phase’ or ‘G2/M phase’ based on their DNA content. i, Cell cycle analysis 7 days following shRNA induction with doxycycline in doxycycline-inducible shRNA-expressing 92.1 cells. Percentage of cells at a specific cell cycle phase from total number of live singlet cells is shown. A representative experiment of n = 2 independent experiments is shown. j, k, l, Quantitative real-time PCR (qRT–PCR) analysis of INPP5A (j), NOXA (k) and PUMA (l) in parental cells and two independent TP53 knockout (TP53-KO1 and TP53-KO2) 92.1-Cas9 cell lines after transduction with the indicated sgRNAs, n = 3 biologically independent samples. All data are presented as mean ± s.e.m.; P values were determined by two-way analysis of variance (ANOVA) with multiple comparisons (e) or one-way ANOVA with multiple comparisons (f, g, j, k, l). Western blot was repeated twice with similar results.

Source data

Extended Data Fig. 9 INPP5A is required for UM tumor growth and metastasis development.

a, Quantitative real-time PCR (qRT–PCR) analysis of INPP5A mRNA levels of 92.1 xenografts tumors on day 3 and 45 post doxycycline treatment. Results are shown as percentage of no doxycycline control. shCtl, n = 3 mice, shINPP5A#1 and shINPP5A#2 day 3 and 45, n = 6 mice, except shINPP5A#2 day 45, n = 5. Ctl, control, INP, INPP5A. b, Quantification of Ki67 immunohistochemistry staining in 92.1 xenograft tumors at day 10 post-induction of the indicated shRNAs; n = 3 mice for all conditions, except shCtl + dox, n = 2 mice. Ctl, control, INP, INPP5A. c, d, Quantitative analysis of IP3 (c) and IP4 (d) levels of 92.1 xenograft tumors 72 h after treating the mice with doxycycline (25 mg/Kg q.d.) to induce the expression of the indicated shRNAs. Results are shown relative to no doxycycline control. Dashed line depicts equal levels to control, n = 7 mice for all conditions. Ctl, control, INP, INPP5A. e, Bioluminescence images of NSG mice transplanted with 92.1-luciferase-labeled cells represented at day 21 post-transplantation (top). Cells were transduced with the indicated shRNAs before injection and mice were treated with doxycycline as indicated. Ex vivo bioluminescence images of histological sections of the liver of the mice (middle). The liver signal confirms that injected cells could colonize the liver and form tumor metastases. Immunohistochemistry staining of haematoxylin and eosin (H&E) and the melanocytic marker gp100 confirm that the metastases are originating from the injected UM cells (bottom). Scale bar, 500 µm. f, Quantitative real-time PCR (qRT–PCR) analysis of INPP5A mRNA levels of liver metastases formed upon transplantation of 92.1-lucifearse-labeled cells transduced with the indicated shRNAs. Results are shown as percentage of no doxycycline, each black dot represents an individual animal, shCtl +Dox and shINPP5A -Dox, n = 7, and shINPP5A +Dox, n = 4. Data confirm that doxycycline treatment induced the knockdown of INPP5A in the liver metastases. g, Quantification of bioluminescence signal of the liver of the mice after transplantation of 92.1-luciferase-labeled cells transduced with shCtl or shINPP5A. Images were taken on days 10, 21, 31, 41, 51 and 61 post-transplantation. Data show that doxycycline treatment does not impact the bioluminescence signal and cells transduced with control or INPP5A shRNAs have similar growth rates in the liver, n = 8 mice per condition. h, i, Quantification of bioluminescence signal after transplantation of 92.1-luciferase-labeled cells transduced with shCtl or shINPP5A at day 21 (h) and day 31 (i) post-transplantation. Mice were treated with doxycycline as indicated. Data represent average flux as photons per second for the liver area. The results show that the signal was comparable between control groups (shCtl + dox at day 0 and shINPP5A -Dox) and the treatment groups (shINPP5A + Dox at day 21 or day 31) at the start of the treatment. For treatment group of day 21, shCtl + Dox, n = 8, shINPP5A - Dox, n = 12, and shINPP5A + Dox, n =12, for treatment group of day 31, shCtl + Dox, n = 8, shINPP5A - Dox, n = 11, and shINPP5A + Dox, n =12. All data are presented as mean ± s.e.m.; P values were determined by two-tailed, unpaired Student’s t-test (a, b, f), two-way analysis of variance (ANOVA) with multiple comparisons (c, d) or one-way ANOVA with multiple comparisons (h, i).

Source data

Extended Data Fig. 10 Steady-state levels of inositol phosphates in UM cell lines, patient-derived xenografts, and patient tumor biopsies.

a, b, Graphs showing steady-state levels of IP4 (a) and IP3 (b) displayed as µg per million cells in HEK293A-Cas9 parental cells (gray) or cells expressing wild-type (green) or Q209L mutant (blue) HA-GNAQ cDNAs, n = 3 independent experiments. c, Box plots showing steady-state levels of IP3 displayed as µg per million cells in GNAQ/11^WT cell lines (n = 8) and GNAQ/11^Mut cell lines (n = 7). Each dot represents an individual cell line. d, Scatter-plot illustrating the correlation of between cell viability upon INPP5A depletion (y axis) and steady-state levels of IP3 as µg per million cells (x axis) for GNAQ/11^WT cell lines (green, n = 7) and GNAQ/11^Mut cell lines (blue, n = 6). e, Table summarizing relevant information for the patient-derived xenograft (PDX) samples analyzed. f, Representative images of haematoxylin and eosin (H&E) and Melan A immunohistochemistry staining of CM patients’ tumor samples (n = 8 biopsies). Percentage of tumor content and oncogenic mutation are depicted for each biopsy. Scale bars, 50 μm. g, Box plots of steady-state levels of IP3 shown as µg per gram of tumor tissue of GNAQ/11^WT CM patient-derived xenografts (PDX) (n = 4) and GNAQ/11^Mut or CYSLTR2^Mut UM patient-derived xenografts (n = 5). h, Box plots of steady-state levels of IP3 shown as µg per gram of tumor tissue of GNAQ/11^WT CM patients’ tumor samples (n = 8 biopsies) and GNAQ/11^Mut UM patients’ tumor samples (n =7 biopsies). All biopsies are represented by two distinct histological sections (indicated as matching-color dots), except two biopsies per tumor type each is represented by one section. All data are presented as mean ± s.e.m.; P values were determined by one-way analysis of variance (ANOVA) with multiple comparisons (a, b), two-tailed, unpaired Student’s t-test (c, d, g, h). Two-tailed Pearson’s correlation with 95% confidence interval (d). Box plots show 25^th to 75^th percentiles, center line depicts median, and whiskers depict minimum and maximum values (c, g, h).

Source data

Supplementary information

Reporting Summary

Supplementary Table

Supplementary Tables 1–5

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/.

Reprints and permissions

About this article

Cite this article

Elbatsh, A.M.O., Amin-Mansour, A., Haberkorn, A. et al. INPP5A phosphatase is a synthetic lethal target in GNAQ and GNA11-mutant melanomas. Nat Cancer 5, 481–499 (2024). https://doi.org/10.1038/s43018-023-00710-z

Download citation

Received: 25 July 2022
Accepted: 14 December 2023
Published: 17 January 2024
Issue Date: March 2024
DOI: https://doi.org/10.1038/s43018-023-00710-z

Subjects

Abstract

Similar content being viewed by others

Main

Results

Genetic screening for GNAQ/11-mutant UM dependencies

INPP5A is required for the growth of GNAQ/11-mutant UM

IP3 signaling drives INPP5A dependency in GNAQ/11-mutant UM

INPP5A loss increases cytosolic calcium in GNAQ/11-mutant UM

INPP5A loss induces p53-mediated apoptosis in GNAQ-mutant UM

INPP5A promotes UM tumor growth and metastasis development

IP4 is a predictive biomarker of INPP5A dependence

Discussion

Methods

Animal ethics statement

Patient ethics statement

Cell culture and cell engineering

Lentivirus production and knockout cell line generation

Genome-wide CRISPR genetic screen

Bioinformatic analysis of genetic screens

TCGA and CCLE data analysis

RNA sequencing

Cell growth assays

Ion-pair chromatography and mass spectrometry

Quantitative real-time PCR

Apoptosis assays

Cell cycle analysis

HTRF-based IP1 accumulation assay

Phosphatase activity assay

Microscopy

Western blot

In vivo mouse experiments

Histopathology and immunohistochemistry

Patient-derived xenografts

Statistics and reproducibility

Reporting summary

Data availability

Code availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Extended data

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

Search

Quick links