Loss of the Volume-regulated Anion Channel Components LRRC8A and LRRC8D Limits Platinum Drug Efficacy

In recent years, platinum (Pt) drugs have been found to be especially efficient to treat patients with cancers that lack a proper DNA damage response, for example, due to dysfunctional BRCA1. Despite this knowledge, we are still missing helpful markers to predict Pt response in the clinic. We have previously shown that volume-regulated anion channels, containing the subunits LRRC8A and LRRC8D, promote the uptake of cisplatin and carboplatin in BRCA1-proficient cell lines. Here, we show that the loss of LRRC8A or LRRC8D significantly reduces the uptake of cisplatin and carboplatin in BRCA1;p53-deficient mouse mammary tumor cells. This results in reduced DNA damage and in vivo drug resistance. In contrast to Lrrc8a, the deletion of the Lrrc8d gene does not affect the viability and fertility of mice. Interestingly, Lrrc8d−/− mice tolerate a 2-fold cisplatin MTD. This allowed us to establish a mouse model for intensified Pt-based chemotherapy, and we found that an increased cisplatin dose eradicates BRCA1;p53-deficient tumors, whereas eradication is not possible in wild-type mice. Moreover, we show that decreased expression of LRRC8A/D in patients with head and neck squamous cell carcinoma, who are treated with a Pt-based chemoradiotherapy, leads to decreased overall survival of the patients. In particular, high cumulative cisplatin dose treatments lost their efficacy in patients with a low LRRC8A/D expression in their cancers. Our data therefore suggest that LRRC8A and LRRC8D should be included in a prospective trial to predict the success of intensified cisplatin- or carboplatin-based chemotherapy. Significance: We demonstrate that lack of expression of Lrrc8a or Lrrc8d significantly reduces the uptake and efficacy of cisplatin and carboplatin in Pt-sensitive BRCA1;p53-deficient tumors. Moreover, our work provides support to confirm the LRRC8A and LRRC8D gene expression in individual tumors prior to initiation of intensive Pt-based chemotherapy.


Introduction
For over 40 years, platinum (Pt) compounds have been used as major components of chemotherapy regimens for several types of cancer (1). Even in the era of precision medicine and immunotherapy, Pt drugs remain among the most widely used anticancer drugs, due to their efficacy (2). Especially for cancer major cellular target (6). The three most frequently used Pt drugs in the cliniccisplatin, carboplatin, and oxaliplatin-all affect normal DNA functions by generating monoadducts as well as intrastrand and interstrand DNA cross-links (2), subsequently leading to cell death in case these adducts remain unresolved. Consistent with this finding, Pt drugs synergize with tumors that show defects in the DNA damage response (7). A useful example is breast cancer. Pt drugs are not a standard treatment for breast cancer, because their efficacy on most forms of breast cancer is modest. However, patients with breast carcinomas that are defective in DNA repair by homologous recombination (HR) due to the lack of function of BRCA1, BRCA2, or other repair proteins in the HR pathway, do benefit from Pt-based chemotherapy (8)(9)(10)(11). We also observed this in a genetically engineered Kcre;Brca F/F ;Trp F/F (KB1P) mouse model for BRCA-mutated breast cancer. The Brca −/− ;Trp −/− mammary tumors were highly sensitive to cisplatin or carboplatin (12,13).
Although HR deficiency scores provide useful information for patient stratification, the lack of additional reliable biomarkers for the prediction of Pt-based chemotherapy response still represents a major clinical limitation (2,14). Despite the long use of Pt drugs, precision medicine approaches to tackle this challenge are still in their infancy. Moreover, those patients with disseminated tumors who show a major initial response usually develop secondary drug resistance. The precise mechanisms of Pt drug resistance remain poorly defined (2,15,16). One mechanism that has been confirmed in patients with BRCAor BRCA-mutated cancers is the occurrence of secondary mutations in the BRCA or BRCA genes, leading to HR restoration (17,18). Nevertheless, this mechanism alone does not explain all cases of secondary resistance (19) and it is less suitable for predicting upfront therapy response of patients with HRdefective tumors who receive their initial anticancer therapy. In the past, a major focus of the drug resistance studies was put on active Pt drug influx or efflux using tumor cell lines selected in vitro with Pt drugs. However, no transporter has been unambiguously linked to clinical Pt drug resistance thus far (15,20).
In addition to transporters and diffusion, channels provide another route for Pt drugs to penetrate the cell membrane. Using genome-wide functional genetic screens for Pt drug resistance in haploid cells, we have identified volumeregulated anion channels (VRAC), composed of leucine-rich repeat containing (LRRC)8A and LRRC8D plasma membrane proteins, as the long sought-after plasma membrane entry points for cisplatin and carboplatin (21). VRACs, consisting of LRRC8 hexamers, contribute to the cellular volume regulation. The release of cellular solutes such as chloride and potassium ions reduces cell swelling under hypotonic conditions (22,23). For successful formation of the hexameric channel structure at the plasma membrane, the subunit LRRC8A is obligatory, as its knockout (KO) abolishes chloride currents even when the other paralogs (LRRC8B-E) are overexpressed (22,24). The composition of the other subunits (LRRC8B-E) is thought to determine the channel substrate specificity.
As the functional data on the role of LRRC8A and LRRC8D in Pt drug resistance are based on HR-proficient tumor cell lines thus far (21), we set out to study their role in the HR-deficient KB1P model. Indeed, loss of Lrrca or Lrrcd in KB1P tumors largely abrogated the high Pt drug sensitivity of these tumors in vitro and in vivo, even though the uptake of cisplatin or carboplatin was reduced only by about 50% in LRRC8A-or LRRC8D-deficient Brca −/− ;Trp −/− cells.
Using Lrrcd −/− mice, we show that intensified Pt therapy does eradicate KB1P tumors, whereas we are unable to do so in wild-type (WT) mice. Moreover, we corroborate the relevance of LRRCA and LRRCD gene expression for intensified Pt therapy in patients with HNSCC.

Lead Contact and Material Availability
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Sven Rottenberg (sven.rottenberg@vetsuisse.unibe.ch).
Testing for Mycoplasma contamination was performed twice per year using PlasmoTest (InvivoGen, catalog no. rep-pt1).

Genomic DNA Isolation, PCR Amplification, and Tracking of Indels by Decomposition Analysis
To determine the modification rate of cell populations which had been transfected/transduced using individual sgRNA targeting the gene of interest, the gene region was sequenced and analyzed using the Tracking of Indels by

Clonogenic Assays
For clonogenic growth assays in a 6-well plate (TPP, catalog no. 92406) format, 2,000 KB1PM5 cells were seeded per well in DMEM-F12 complete medium. A total of 24 hours after seeding, the cells were treated with the indicated drug doses over the course of 24 hours. After 8 days, the wells were fixed with 4% paraformaldehyde (PFA)/PBS and stained with 0.1% crystal violet. Quantification of the wells was performed with ImageJ using the ColonyArea plugin

Growth Assays
For growth assays, KB1PM5 cells were seeded to 96-well plates (TPP, catalog no. 92696) at a density of 100 cells per well. Proliferation was measured on 7 consecutive days using CellTiter-Blue Cell Viability Assay (Promega, catalog no. G9241) following manufacturer's instructions.

Competition Assays
The polyclonal cell pools generated by transfection with individual sgRNAs targeting either Lrrca or Lrrcd were used for competition assays. In brief, cells were seeded to 12-well plates (TPP, catalog no. 92412) at 1,000 cells/well and drug selection using different Pt-based agents was applied at indicated concentrations for 24 hours. After 8 days of recovery, cells were either fixed with PFA or stained with crystal violet (Supplementary Fig. S3D) or harvested and the gDNA isolated and processed for subsequent TIDE analysis as described previously (28). The shift of mutated alleles was determined by comparing the modification rates of drug naïve and drug selected polyclonal cell populations.

Western Blotting
Cells were washed and scraped in cold PBS. After pelleting, cells were lysed in RIPA buffer (

In Vivo Validation of Resistance
All animal experiments were approved by the Animal Ethics Committee of The

Generation of Lrrc8d KO Mice
The Lrrcd conditional KO mice were generated by an established CRISPR/Cas9-mediated gene editing protocol (30). Briefly, LoxP sites were introduced to flank the first coding exon of the Lrrcd gene sequence (exon 3). Zygotes isolated from FVB mice were coinjected with a microinjection mix containing in vitro transcribed Cas9 mRNA, two sgRNAs targeting the different sites in the Lrrcd gene (5 -CACCGGCTTCAGGATTCGGTAAGT-3 and 5 -CACCGCAGGCACACCCACGTGCGG-3 ) and two homology-directed repair oligos, each containing a LoxP site. Zygotes that have further divided into two-cell embryos after overnight incubation at 37°C, 5% CO 2 , were surgically implanted into the oviduct of a pseudopregnant foster mother. Correct incorporation of the LoxP sites was confirmed by PCR. Subsequently, these mice were crossed with Cre recombinase expressing FVB mice, resulting in offspring with a 2,561 bp deletion in exon 3 of the Lrrcd gene.

LRRC8D Protein Levels in Mouse Kidneys
To determine the LRRC8D protein levels in WT, heterozygous, and homozygous KO mice, protein was isolated from whole kidneys by metal bead homogenization in PBS and subsequent lysis in T-PER Tissue Protein Extraction Reagent (Thermo Fisher Scientific, catalog no. 78510) and 1x Halt protease inhibitor cocktail (Thermo Fisher Scientific, catalog no. 87786). A total of 80 μg protein was subjected to electrophoresis and Western blotting.
Antibody incubations using the rabbit anti-LRRC8D polyclonal antibody provided by T. Jentsch and subsequent imaging were performed as described previously.

Cisplatin Adduct and yH2AX IHC Staining of Mouse Kidneys After Treatment
Littermate WT, heterozygous, and homozygous Lrrcd KO mice were intravenously treated with 6 mg cisplatin per kg (Teva, catalog no. 4333164). After 6 hours, the animals were anesthetized with isoflurane, sacrificed with CO 2 followed by organ harvest. Kidneys of vehicle-or cisplatin-treated mice were fixed in 4% PFA and further embedded in paraffin. Each paraffin block was sectioned at 2.5 μm and IHC was performed on an automated immunostainer (Leica Bond RX, Leica Biosystems, catalog no. 95735-848538). The following antibodies were used for IHC staining of tumors: the homemade NKI-A59 antibody for the detection of cisplatin DNA adducts as described previously (31)  mouse ± SD (two-way ANOVA, followed by Tukey multiple comparisons test). A pathologist who was blinded concerning the sample identity carried out the quantification of both stainings.

Imaging Mass Cytometry Measurement of Pt Content in the Kidneys
The same kidney samples as for the Pt-adduct antibody IHC stainings were used for the imaging mass cytometry (IMC) analysis.

Drug Toxicity in Lrrc8d KO Mice
To determine the tolerability of higher cisplatin doses, littermate WT and homozygous Lrrcd KO mice were subjected to 6, 9,  . Briefly, Z-stacks of the individual channels were projected using the "sum slices" projection. All nuclei were detected by the "analyze particles" command using the DAPI channel.
This region of interest (ROI) selection was then used to determine the raw integrated density of the NKI-A59 antibody staining of each nucleus. ROI touching the edges of the images were excluded. Data were plotted in GraphPad Prism software 9 and significance was calculated using two-way ANOVA followed by Tukey multiple comparisons test.
For the yH2AX foci detection, the cells were seeded to coverslips and treated with 2 μmol/L cisplatin for the indicated timepoints. After treatment, cells were washed with PBS and fixed with 4% PFA/PBS for 20 minutes at 4°C. Briefly, Z-stacks of the individual channels were projected using the "max intensity" projection setting. All nuclei were detected by the "analyze particles" command using the DAPI channel projection and this ROI selection was then used to determine the number of yH2AX foci of each nucleus by the "finding maxima" command on the FITC channel. ROI touching the edges of the images were excluded. Data were plotted in Graph-Pad Prism 9 software and significance was calculated using ordinary one-way ANOVA followed by Tukey multiple comparisons test.

Statistical Analysis HNSCC
The collection of the HNSCC biopsies was approved by the Institutional Review Board of the Netherlands Cancer Institute. All patients signed an informed consent for the collection and analysis of the HNSCC samples and the study was conducted in accordance with International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS). CNV data of LRRCA or LRRCD were obtained by shallow DNA sequencing of HNSCCs dataset (35). Gene expression data were obtained by polyA RNA sequencing (RNA-seq; ref. 36). Patients were classified into low or high LRRCA expression by the cutoff at 20 rpkm.

Statistical Analysis Ovarian Cancer
The GSE32063 dataset consists of 40 advanced-stage high-grade serous ovarian cancer samples. This study was performed after approval by the Institutional Review Board. All patients provided a written informed consent for the collection and analysis of these samples and the study was conducted in accordance with International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS). All patients were treated with a combined platinumtaxane standard chemotherapy. Gene expression levels were obtained by whole human genome microarray sequencing (Agilent-014850 4X44K G4112F; ref. 37). Patients were classified into low or high LRRCA or LRRCD expression by a cutoff of 33% (lower and upper tertile).

Data Availability Statement
The HNSCC dataset was previously published by Essers and colleagues (35).
The Dutch multicenter cohort CNA and RNA-seq data that support the findings of this study are available in the European Genome-Phenome Archive (EGA) at https://ega-archive.com under the EGA study numbers EGAS00001004090 (RNA-seq data), EGAS00001004091 [low-coverage whole-genome sequencing (WGS)] and dataset numbers EGAD00001005716, EGAD00001005715 and EGAD00001005719, EGA D00001005718 for the RNA-seq and low-coverage WGS, respectively.
The expression data of the ovarian cancer dataset were previously published by Yoshihara and colleagues and is available from Gene Expression Omnibus data repository by the accession number GSE32063 (37).

Loss of Lrrc8a or Lrrc8d Induces Cisplatin and Carboplatin Resistance in BRCA1;p53-deficient Mouse Mammary Tumor Cells
To investigate the effects of LRRC8A or LRRC8D defects on Pt drug sensitivity in HR-deficient tumors, we generated CRISPR/Cas9 KOs in cell lines derived from a genetically engineered mouse model for hereditary BRCAmutated breast cancer (25). Because of the irreversible Brca deletion these cells are highly sensitive to Pt drugs and thereby provide a useful tool to study mechanisms of Pt drug resistance that are independent of a restoration of BRCA1 function (38). Using a paired gRNA approach to generate AACRJournals.org Cancer Res Commun; 2(10) October 2022 big deletions in the Lrrca or Lrrcd genes (Supplementary Fig. S1A and S1B), we obtained monoclonal cell lines that lost expression of LRRC8A Fig. 1A). KO of Lrrca or Lrrcd did not affect cell growth overall (Fig. 1B). When we treated these Lrrca −/− and Lrrcd −/− cells with cisplatin ( Fig. 1C and D) and carboplatin ( Fig. 1E and F), we observed an increased survival using clonogenic assays, the Lrrca −/− cells being more resistant than the Lrrcd −/− cells. In contrast, only a minor effect was observed in response to oxaliplatin in the LRRC8D-deficient cells (Fig. 1G  and H). Furthermore, we observed resistance of the KO cell lines to the protein synthesis inhibitor blasticidin S ( Supplementary Fig. S2A and S2B). The uptake of this drug is known to be LRRC8D dependent (39). The reintroduction of the We further corroborated these data using polyclonal cells that we generated with single Lrrca-or Lrrcd-targeting sgRNAs ( Supplementary Fig. S3). With the help of the TIDE analysis (28), we quantified the presence of WT and Fig. S3A and S3B). These polyclonal cell populations, that contain about 50%

Lrrca/d-modified alleles in the polyclonal cell populations (Supplementary
WT alleles, were then treated with the Pt-based agents cisplatin, carboplatin, and oxaliplatin ( Supplementary Fig. S3C). As expected by the presence of WT alleles, the resistance to cisplatin and carboplatin was milder compared with the KO clones presented in Fig. 1 (Supplementary Fig. S3D). More importantly, when we measured the frequency of frameshift modifications following drug treatment, we observed a clear selection in favor of the Lrrca-and Lrrc8dmutated alleles following cisplatin and carboplatin. Regarding oxaliplatin, only for the Lrrcd-mutated alleles, we observed a modest positive selection when treating with the 0.5 μmol/L drug concentration ( Supplementary Fig. S3E and S3F).
In our previous study using human HAP1 cells, we found that less carboplatin enters the cells if they contained VRACs that were LRRC8A or LRRC8D deficient (21). We also measured the Pt uptake in our Lrrca −/− and Lrrcd −/− KB1PM5 cells using CyTOF, which allows the use of more physiologic drug concentrations than were tested previously in the HAP1 cells. After incubation with 0.5 μmol/L of cisplatin or 4 μmol/L carboplatin for 24 hours, the Pt content was reduced by more than 65% in the Lrrca −/− cells for both cisplatin and carboplatin and about 25% for cisplatin and 35% for carboplatin in the Lrrcd −/− cells ( Fig. 2A and B). The decrease could be reversed by the reintroduction of Lrrca or Lrrcd (Fig. 2C). The decrease in intracellular Pt accumulation can be expected to result in less Pt-DNA adducts. To test this, we used the NKI-A59 antibody (31), which detects cisplatin-induced DNA adducts. Consistent with the uptake data, less Pt-DNA adducts were formed in the LRRC8Aand LRRC8D-deficient cells (Fig. 2D and E). Again, the effect was stronger in the Lrrca −/− cells than the Lrrcd −/− cells. The lower amount of Pt-DNA adducts resulted in less DNA damage, as measured by γH2AX foci formation ( Fig. 2F and G). Hence, the high cisplatin sensitivity of the BRCA1-deficient cells is alleviated in the absence of LRRC8A or LRRC8D, due to reduced drug uptake.

LRRC8A and LRRC8D Defects Abrogate the In Vivo Efficacy of Carboplatin
Many clinicians prefer carboplatin to cisplatin, as the toxicity profiles of the two drugs differ, including a lower nephrotoxicity of carboplatin than cisplatin (40). To validate the role of LRRC8A and LRRC8D in the Pt drug response in vivo, we tested the carboplatin response of the tumors in our breast cancer KB1P model, using the 3D organoid technology (26). For this purpose, KB1P4N organoids, derived from a Brca −/− ;Trp −/− mammary tumor (41), were transduced with lentiviruses carrying Lrrca or Lrrcd targeting pLENTiCRISPRv2 vectors. Control organoids were generated by transduction with pLentiCRISPRv2, encoding a nontargeting sgRNA. LRRC8A/D-deficient organoids were then transplanted orthotopically into the mammary fat pad of mice. The outgrowing tumors were analyzed by TIDE and indeed showed a high frameshift modification percentage ( Supplementary Fig. S4A and S4B). When the organoid-derived control or Lrrc8a/d-targeted tumors reached a volume of 75 mm 3 , we treated them with the MTD of 50 mg/kg carboplatin i.v. on days 0 and 14 (Fig. 3). The tumor volume was monitored throughout the whole experiment ( Supplementary Fig. S4C and S4D). As shown in Fig. 3, depletion of Lrrc8a or Lrrc8d significantly lowered the tumor response to carboplatin and resulted in a decreased overall survival (OS; P = 0.002 for Lrrc8a, P = 0.0131 for Lrrc8d; Fig. 3). Consistent with the in vitro data, the effect was stronger in the LRRC8A-deficient tumors than in the LRRC8D-deficient ones. These data demonstrate that loss of Lrrc8a or Lrrc8d renders BRCA1;p53-deficient tumors resistant to carboplatin in vivo.

Lrrc8d −/− Mice are Viable and Provide a Useful Model to Study High-dose Cisplatin Therapy
It was previously shown that Lrrca KO mice are severely compromised and show an increased mortality in utero and postnatally (42). To test whether the Lrrcd KO is tolerable in mice, we introduced a large 2561 bp deletion of the protein coding sequence in exon 3 using CRISPR/Cas9-mediated gene editing of FVB/N zygotes (Fig. 4A). This completely abrogated Lrrcd expression at the RNA and protein level in homozygous mice, and also significantly reduced RNA and protein levels in heterozygous animals ( Fig. 4B; Supplementary  Fig. S5A and S5B). In contrast to the Lrrca KO mice, we did not see any alterations regarding fertility or histomorphology of the Lrrcd −/− mice that we followed for at least 6 months. We therefore conclude that the Lrrcd −/− mice are fully viable. In FVB/N mice, we previously reported an MTD of 6 mg/kg cisplatin i.v. on days 0 and 14 (38). We could not escalate the dose above MTD levels by bone marrow reconstitution, and 3 days after treatment using 9 or 12 mg/kg cisplatin i.v., the body weight dropped below 90% and animals had to be sacrificed. In contrast, in the Lrrcd −/− mice we could double the dose to 12 mg cisplatin per kg and the average weight of the mice stayed above 90% (Supplementary Fig. S5C). A main cytotoxic side effect of cisplatin is the increased death of tubular epithelial cells in the renal cortex. We therefore directly measured the Pt content in the kidneys of the Lrrcd-proficient and homozygous KO mice 6 hours after treatment with 6 mg/kg cisplatin i.v. For the analysis, the amount of Pt was measured in three kidneys per group using IMC. The mean 194 Pt values per section were normalized to the 134 Xe + background signal intensity. Indeed, the amount of Pt that we found in the kidneys of Lrrcd −/− mice was lowered by about 50% compared with the WT mice ( Fig. 4C and D). With less cisplatin entering the kidneys, we expected a reduction in the cisplatin-DNA adducts to be formed in the nuclei of tubular epithelial cells, which cause the nephrotoxicity. As presented in Supplementary Fig. S5D and S5E, these cells (Continued) The data represent the mean ± SD of three independent replicates where approximately 50,000 cells per condition and cell lines were acquired (two-way ANOVA followed by Tukey multiple comparisons test, ****, P < 0.0001). C, CyTOF-based measurement of Pt uptake after 24 hours treatment with 0.5 μmol/L cisplatin, 4 μmol/L carboplatin, or 0.5 μmol/L oxaliplatin of selected clonal Lrrc8a-, or Lrrc8d-high expression rescue lines compared with the empty vector transduced KO cell line. The data represent the mean ± SD of three independent replicates where approximately 50,000 cells per condition and cell lines were acquired (two-way ANOVA followed by Tukey multiple comparisons test, ****, P < 0.0001).
D, Representative images of the average nuclear staining in ntg or Lrrc8aor Lrrc8d-KO cell lines using the NKI-A59 antibody against cisplatin-adducts in the presence or absence of 10 μmol/L cisplatin treatment for 6 hours; scale bar, 10 μm. E, Quantification of the raw integrated density per nucleus of the NKI-A59 cisplatin adduct staining, mean with ± 95% confidence interval of three independent replicates are shown. Per replicate approximately 100 nuclei were quantified. The significance was determined using two-way ANOVA followed by Tukey multiple comparison test. ****, P < 0.0001. can be detected with the NKI-A59 antibody against Pt-DNA adducts 6 hours after treatment of WT mice with 6 mg/kg cisplatin i.v. The amount of Pt-DNA adducts was also substantially decreased in the kidneys of Lrrcd −/− mice and to a lesser extent also in the Lrrcd +/− animals, in comparison with the WT mice ( Fig. 4E and F). This decrease of Pt-DNA adducts correlated with a decrease in DNA damage, measured by γH2AX foci formation (Fig. 4E-G). These data show that also in vivo less cisplatin enters LRRC8D-deficient cells.
Despite the high cisplatin sensitivity of the KB1P tumors, the tumors are not easily eradicated, not even with repeated treatments using the MTD (38). We previously showed that residual G 0 -like tumor cells that transiently avoid entering the cell cycle can escape the cisplatin-induced DNA damage, even without functional BRCA1 (38). The fact that Lrrcd −/− mice tolerate the double MTD of the WT mice allowed us to address the basic question whether a high dose of cisplatin eradicates the KB1P tumors containing functional VRACs. As shown in Fig. 4H   the Lrrc8d KO mice were used. Statistical analysis was performed with the log-rank test (Mantel-Cox). **, P < 0.01; ***, P < 0.001.

Low Expression of LRRC8A or LRRC8D Correlates with Decreased OS and Reduced Recurrence-free Survival in Patients with HNSCC Treated with Cisplatin-based Chemoradiotherapy
In the clinic, the use of Pt drugs for the treatment of patients with BRCAdeficient breast cancer has only recently been expanded, and currently the sample availability for this specific subgroup is scarce. Another tumor type in which BRCA mutations are frequently found and which is highly sensitive to Pt-based chemotherapy is ovarian cancer. Using The Cancer Genome Atlas and Patch and colleagues datasets, we have previously reported a lower survival of patients with ovarian cancer who have a low LRRCD gene expression (21). We could confirm this in another independent data set of patients with ovarian cancer (GSE32063; ref. 37). As shown in Supplementary Fig. S6, patients expressing low LRRCD levels had a significantly decreased OS (P = 0.009) and nonsignificantly shortened progression-free survival (P = 0.096).
As the relevance of LRRC8A or LRRC8D function for Pt drug uptake is independent of BRCA1/2 status, we also studied LRRC8A/D in patients with HNSCC, for which cisplatin is a standard therapy in combination with radiotherapy (Fig. 5). In contrast to the ovarian cancer cohort, we had both copy-number variation (CNV) and gene expression data available from the tumors of these patients with HNSCC. CNV analysis identified 20 of 166 tumors with a loss of the LRRCA gene. In these patients, the loss of LRRCA is associated with a lower OS (P = 0.014) and increased tumor progression (P = 0.028) compared. Moreover, the loss of LRRCA in these patients is correlated with an increased risk for failure of locoregional control (P = 0.0046) and distant metastasis (P = 0.051) after treatment (Fig. 5A). To test for alterations in the level of LRRCA transcripts, gene expression data of these patients were analyzed. Samples were classified to low (<20 rpkm, N = 58) and high (>20 rpkm, N = 129) subgroups (Fig. 5B). Low LRRCA gene expression correlated significantly with poor OS in these patients (P = 0.0057; Fig. 5C).
It has previously been shown that patients benefit from cumulative cisplatinbased radiotherapy regimens above 200 mg/m 2 (43,44). We therefore stratified patients according to their cumulative dose, and their OS was analyzed according to their LRRCA expression levels (for low LRRCA expression N = 43 in ≥200 mg/m 2 and N = 24 in <200 mg/m 2 ). The loss of LRRCA expression resulted in a decrease of the OS curves to the levels of lower and less effective cisplatin doses (Fig. 5D). The data suggest that HNSCC patients with high LRRCA expression are likely to benefit from higher cisplatin doses.
Regarding LRRCD, we also identified a group of 61 patients with low gene expression, and these patients are different from those expressing low levels of LRRCA (Fig. 5E). The stratification of the tumors into low (N = 61), medium (N = 60), and high (N = 60) LRRCD gene expression subgroups revealed decreased OS (P = 0.051) and poor tumor progression outcome (P = 0.054) in the low expression group. No effects on locoregional control of the tumor were found (P = 0.84; Fig. 5F). Moreover, low LRRCD expression correlated with an increased distant metastasis occurrence (P = 0.030; Fig. 5G). To assess the impact of LRRCD expression on varying cumulative cisplatin doses, patients were split into ≥200 mg/m 2 and <200 mg/m 2 receiving subgroups. Low expression was defined as the lowest tertile expressing group in both treatment classes (N = 37 in ≥200 mg/m 2 and N = 23 in <200 mg/m 2 ). Low LRRCD gene expression was associated with increased distant metastasis, especially in the ≥200 mg/m 2 subgroup of patients (P = 0.0054; Fig. 5H). This association was strongest in the distant metastasis formation (Fig. 5I).
Hence, in cisplatin-treated patients with HNSCC, low LRRCA and LRRCD expression is associated with poor outcome. On the basis of our experimental work, this is explained by poor drug uptake of these tumors.

Discussion
Using a mouse model for BRCA-mutated breast cancer, we show here the relevance of LRRC8A-and LRRC8D-mediated cisplatin and carboplatin uptake to kill Pt drug-sensitive tumors. The potential relevance to the treatment of human cancer is suggested by our finding that patients with ovarian cancer as well as HNSCC with low gene expression levels of LRRCA or LRRCD in their tumors have a reduced benefit of Pt-based chemotherapy. Hence, the absence of LRRC8A or LRRC8D in tumor cells may be a helpful marker to avoid the use of cisplatin or carboplatin.
At present, there is no biomarker routinely applied in the clinic to predict the outcome of Pt-based chemotherapy. In part, this may be due to the fact that cisplatin influx into tumor cells was long thought to be completely due to passive diffusion (15). Although active influx via SLC31A1-the mammalian homolog of the budding yeast copper transporter (CTR1)-or via OCT2 (SLC22A2)an organic cation transporter-was suggested, these transporters could not be unambiguously proven to cause reduced uptake and drug resistance when absent (45,46). OCT2 has been shown to mediate cisplatin-induced nephrotoxicity (47,48), but its expression is largely limited to the kidney, explaining why it may not be involved in tumor uptake. Decreased accumulation of Pt drugs may also be caused by increased drug export. In this context, two Cu AACRJournals.org Cancer Res Commun; 2(10) October 2022 transporters-ATP7A and 7B-were put forward (49). Despite some correlations that have been reported between poor cisplatin response and high ATP7B levels in patients, the role of these efflux transporters in Pt drug resistance remains to be clarified (2,15,20). Moreover, low-level cisplatin resistance was reported in cells overexpressing MRP, but in human tumors, there is no consistent correlation between MRP expression and cisplatin resistance (15,46). In summary, no transporter has been explicitly linked to clinical Pt drug resistance, be it importer or exporter (2,20).
In addition to diffusion or transporters, channels provide another route for Pt drugs to enter cells. Already in 1993, Gately and Howell (50) concluded that a fraction of cisplatin enters via a channel, because there are several inhibitors that decrease cisplatin entry. The authors suggest that about half of the cisplatin uptake is by passive diffusion through the membrane and the other half through a dedicated channel. More than 50% inhibition by any inhibitor has not been seen, resulting in this 50% channel estimate. Using genome-wide functional genetic screens for carboplatin drug resistance in human HAP1 cells, we have recently identified VRACs composed of LRRC8A and LRRC8D proteins as these long sought-after plasma membrane entry points for cisplatin and carboplatin (21). Loss of LRRC8D was also a major hit in a genome-scale CRISPR-Cas9 KO screen for cisplatin resistance using BRCA-mutated ovarian cancer cells (51).
Here, we confirm the relevance of these proteins for sensitivity to Pt drugs, using Pt drug-sensitive BRCA1;p53-deficient mouse mammary tumors and cell lines derived from this model. Despite their strong drug sensitivity due to an irreversibly deleted Brca gene, the effect of cisplatin and carboplatin was largely abrogated in the absence of LRRC8A or LRRC8D. We found that about 50% of carboplatin and cisplatin uptake depended on LRRC8A and LRRC8D (Fig. 2), which is consistent with the assumption of Gately and Howell (50) and our observations in HAP1 cells (21). According to recent Cryo-EM studies of the LRRC8 subunits, the substrate specificity for larger osmolytes is likely dependent on the presence of LRRC8D in the channel composition. LRRC8D is the largest isoform, presenting the longest extracellular loop between transmembrane domain 1 and 2 (52,53). Of note, homo-hexameric LRRC8D structures contain a wider pore diameter than structures consisting of solely LRRC8A (53,54). Together with LRRC8A, LRRC8D seems to be responsible for the cellular uptake of Pt-based drugs, in contrast to the other subunits (LRRC8C, LRRC8E; ref. 21). However, the effect of LRRC8A depletion on Pt drug resistance is not seen in all cell lines though, most likely because LRRC8A is essential for some of them (24,55). Lrrca KO mice are severely compromised and show an increased mortality in utero and postnatally, as well as infertility (42,56). In contrast, we found that Lrrcd KO mice are viable and breed normally. Although further experiments are required to investigate Pt drug pharmacokinetics, renal excretion and nephrotoxicity, we clearly observed less cisplatin and Pt-DNA adducts in the kidneys of Lrrcd −/− mice, resulting in reduced DNA damage. Because of reduced cisplatin uptake, their MTD is doubled, which allowed us to build a model for high-dose Pt drug-based chemotherapy. Both cisplatin-sensitive and -resistant LRRC8A-and LRRC8D-proficient KB1P tumors were completely eradicated when the cisplatin MTD was augmented 2-fold (Fig. 4). This is consistent with our previous findings using nimustine (38), that drug-tolerant cells can be eliminated, if sufficient damage is inflicted. Also in our cohort of patients with HNSCC, the correlation between low LRRCA or LRRCD gene expression and poor cisplatin-based therapy response was more pronounced in the high-dose groups. While further analyses need to be made to investigate whether this result can be reproduced using carboplatin, these data strongly indicate that it would be useful to assess the LRRCA and LRRCD expression of tumor cells before applying high-dose cisplatin-or carboplatin-based therapies. Clinicians might be able to avoid the serious side effects, if insufficient drug amounts can be expected to reach the tumor cell DNA, the main cellular target of Pt drugs. An alternative approach for cancers with low LRRC8A levels may be oxaliplatin. In our KB1P model, we only observed a modest reduction in oxaliplatin uptake and consequent therapy resistance in Lrrcd-deficient cells, suggesting oxaliplatin as useful alternative when patients are stratified on the basis of LRRCA expression.
Given the frequent use of Pt drugs in daily clinical practice, our data highlight the importance of further validation of LRRC8A and LRRC8D status as a predictive biomarker in prospective clinical trials. In addition to classical gene or protein expression correlations, our results also suggest the use of CyTOF-based measurements to quantitatively evaluate Pt uptake in patient-derived primary 2D or 3D cell cultures.