WDHD1 is essential for the survival of PTEN-inactive triple-negative breast cancer

Triple-negative breast cancer (TNBC) is the most aggressive type of breast cancer that lacks the oestrogen receptor, progesterone receptor and human epidermal growth factor receptor 2, making it difficult to target therapeutically. Targeting synthetic lethality is an alternative approach for cancer treatment. TNBC shows frequent loss of phosphatase and tensin homologue (PTEN) expression, which is associated with poor prognosis and treatment response. To identify PTEN synthetic lethal interactions, TCGA analysis coupled with a whole-genome siRNA screen in isogenic PTEN-negative and -positive cells were performed. Among the candidate genes essential for the survival of PTEN-inactive TNBC cells, WDHD1 (WD repeat and high-mobility group box DNA-binding protein 1) expression was increased in the low vs. high PTEN TNBC samples. It was also the top hit in the siRNA screen and its knockdown significantly inhibited cell viability in PTEN-negative cells, which was further validated in 2D and 3D cultures. Mechanistically, WDHD1 is important to mediate a high demand of protein translation in PTEN-inactive TNBC. Finally, the importance of WDHD1 in TNBC was confirmed in patient samples obtained from the TCGA and tissue microarrays with clinic-pathological information. Taken together, as an essential gene for the survival of PTEN-inactive TNBC cells, WDHD1 could be a potential biomarker or a therapeutic target for TNBC.


Introduction
Breast cancer is the most common cancer type and the leading cause of cancer death in women worldwide 1 . Triple-negative breast cancer (TNBC) lacks the oestrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2), and accounts for between 10 and 20% of breast cancers [2][3][4][5] . TNBC is the most aggressive and high-grade breast cancer type with high risk of tumour recurrence and metastasis compared to the other breast cancer subtypes 6 . As TNBC lacks all three receptors, this causes more challenges for the treatment of the disease. Chemotherapy has been the only standard treatment option to improve the overall survival rate of TNBC patients for several years 7 . Therefore, it is important to study gene profiling by identifying different gene expression signatures in TNBC to discover a novel biomarker or targeted therapy for the disease. Atezolizumab (TECENTRIQ ® ), an antiprogrammed death-ligand 1 (PD-L1) monoclonal antibody (checkpoint inhibitor), was approved as the first breast cancer immunotherapy to be combined with chemotherapy (Abraxane; nab®-Paclitaxel) for PD-L1positive TNBC 8 . As a heterogeneous disease 9 , further gene profiling studies are required to identify novel biomarkers or therapeutic targets for TNBC.
TNBC shows frequent loss of phosphatase and tension homologue (PTEN) expression compared to the other molecular subtypes of breast cancer 10,11 . It has been shown that loss of PTEN expression was significantly associated with TNBC that shows poor prognosis and significant links with high-grade tumour, larger tumour size, lymph node metastasis and tumour recurrence 12 . PTEN was identified as a tumour suppressor gene (TSG), located on 10q23 chromosome band, which plays an essential role to control cell cycle, growth and survival 13 . Mechanistically, PTEN has a cytoplasmic lipid phosphatase role that can inhibit the phosphatidylinositol 3-kinase (PI3K)-AKT pathway 13,14 , and the nuclear phosphataseindependent role of PTEN which has been shown to maintain genomic stability 15,16 .
Targeting synthetic lethality is an alternative approach for cancer treatment 17 . To identify novel targeted therapies, synthetic lethality screens were performed, including RNA interference (RNAi) screens 18,19 . One of the well-known examples of synthetic lethality interaction is between BRCA1/2 and PARP. BRCA1/2 are TSGs that have a role in homologous-recombination-mediated DNA repair and PARP is involved in base excision repair. Tumours with BRCA1/2 deficiency depend on PARP1 for DNA repair. Thus, inhibition of PARP1 kills BRCA1/2deficient tumours 20,21 . Discovering PTEN synthetic lethal interactions in TNBC may provide potential biomarkers or targeted therapies for this breast cancer type that does not have successful treatment options.
In this study, candidate genes essential for the survival of PTEN-inactive TNBC cells were identified by the TCGA analysis and a whole-genome siRNA screen in isogenic PTEN-negative and -positive cells. Among them, WD repeat and high-mobility group box DNAbinding protein 1 (WDHD1) expression was increased in the low vs. high PTEN TNBC samples. It was also the top candidate gene whose knockdown significantly inhibited cell viability in PTEN-negative cells, which was further validated in 2D and 3D cultures. Mechanistically, WDHD1 was important to mediate a high demand of protein translation in PTEN-inactive TNBC. Finally, the importance of WDHD1 in TNBC was confirmed in patient samples obtained from the TCGA and tissue microarrays with clinic-pathological information.

TCGA analysis confirms PTEN expression is decreased in TNBC and correlates with clinical stages
It has been stated that PTEN inactivation occurs more frequently in TNBC than the other subtypes of breast cancer 11,12 Fig. S3b; P < 0.05).
By cross-referencing TCGA analysis with the wholegenome siRNA screen, 47 candidate genes essential for the survival of PTEN-inactive TNBC cells were identified (Fig. 1B, C and Supplementary Tables S1 and S2). Among them, WDHD1 expression was increased in the low vs. high PTEN TNBC samples (Supplementary Table  S1; P = 0.03). It was also the top candidate gene whose knockdown significantly inhibited cell viability in PTENnegative cells (Z score = −1.26) with mild effects on PTEN-positive cells (Supplementary Table S2; Z score = −0.32; P = 0.009). Given our findings that decreased PTEN levels are responsible for the high AKT activity in TNBC, we then determined if AKT is involved in the regulation of To reflect the functional consequence of PTEN status, we decided to check p-AKT_308 levels and the correlation with WDHD1 expression in TCGA. We demonstrated that there was a significant positive correlation between WDHD1, mRNA expression and pAKT_308 levels in the TCGA dataset ( Fig. 4E; r = 0.3321, P = 0.0296).
Taken together, our results demonstrate that WDHD1 expression is affected by PTEN status in TNBC cells and this is mainly achieved by AKT signalling.  whole-genome siRNA screen, depletion of WDHD1 selectively inhibited cell viability in PTEN null vs. WT TNBC cells with two individual siRNA oligos against WDHD1, although statistical significance for oligo 1# was not reached (P = 0.054) (Fig. 5A).
It is known that 3D cell cultures represent their in vivo counterparts better than 2D monolayer cell culture models 23,24 . To further validate the effects of WDHD1 knockdown in TNBC cells, 3D mammosphere assays with PTEN WT (BT20 and MDA-MB-231) and null type (HCC1395 and HCC1937) TNBC cell lines were performed. Images of spheres were analysed for sphere formation efficiency and sphere volume, and cell viability was determined using Cell-Titer Glo ® assays. WDHD1 depletion in PTEN WT TNBC cell lines (BT20 and MDA-MB-231) showed minimal effects on sphere formation efficiency, sphere volume and cell viability ( Supplementary Fig. S5). In contrast, a significant decrease in sphere formation efficiency, sphere volume and cell viability with two individual siRNA oligos against WDHD1 was observed in HCC1395 ( Fig. 5B; P < 0.05) and HCC1937 ( Fig. 5C; P < 0.05), both of which are PTEN null type TNBC cell lines.
These experiments showed that WDHD1 is preferentially required by PTEN-inactive TNBC cells for survival, but not for those harbouring WT PTEN.

Essential roles of WDHD1 in cell cycle in PTEN null TNBC cell lines
In order to understand the functions of WDHD1, 92 TNBC samples from the TCGA were identified (Fig. 1A). The top 10% and bottom 10% of samples were separated into two groups: high and low WDHD1 expressing samples, respectively, and those genes with P values < 0.05 were considered as differentially expressed genes (DEGs). A heatmap of 3796 DEGs in the high vs. low WDHD1 groups (P < 0.05) was shown in Supplementary Fig. S6a. To investigate whether the significantly up-regulated 2069 genes in the high WDHD1 group were enriched in certain cellular functions, ToppGene, (https://toppgene.cchmc.org/), was used. We found that the regulation of cell cycle was enriched in the high WDHD1 TNBC samples ( Supplementary Fig. S6b).
To validate these findings, WDHD1 expression was depleted by two individual siRNA oligos in TNBC cell lines, followed by cell cycle analysis based on flow cytometry ( Supplementary Fig. S7). Interestingly, depletion of WDHD1 with two individual siRNA oligos significantly     Table S3). The top four functions are shown in Fig. 6B, with protein translation as the top one (Fig. 6B), which suggests a role of WDHD1 in protein translation in PTEN null TNBC cells.
To verify these findings, WDHD1 expression was depleted by two individual siRNA oligos in MDA-MB-468-TR-PTEN cells followed by puromycin incorporation assay to measure protein synthesis. Puromycin is commonly used to study translation 25,26 . Puromycin incorporation stops translation elongation and subsequently induces the release of puromycylated peptides from the ribosome 27 . Unlike radiolabelled amino acids and noncanonical amino acid analogues, puromycin incorporation is not significantly impacted by the endogenous methionine concentration nor the methionine content of proteins 26 . Puromycin thus incorporates relatively equally into all nascent polypeptides, making it a reliable tool for measuring global protein synthesis.
In this study, we utilised the puromycin incorporation assay, in which cells were treated with 2.5 µM puromycin for 5 min before sample collection. We were able to show a 25−30% reduction in global protein translation upon PTEN re-introduction or WDHD1 depletion (Fig. 6C, D; P < 0.05).  As a positive control, PTEN expression was induced in MDA-MB-468-TR-PTEN cells by addition of DOX, since it is known that PTEN inhibits protein translation through negative regulation of mammalian target of rapamycin (mTOR) (Fig. 6C) 28 .
As shown in Fig. 6C and D, depletion of WDHD1 with two individual siRNA oligos significantly inhibited global protein translation in MDA-MB-468 cells, reflected by the reductions in the puromycin labelling intensity (Fig. 6D; P < 0.05). The inhibitory effect of WDHD1 depletion on protein translation was similar to those achieved by reintroducing PTEN in MDA-MB-468 cells (Fig. 6C, D), indicating an important role of WDHD1 in protein translation in PTEN null TNBC cells. Interestingly, the phosphorylation level of mTOR was not affected by WDHD1 status (Fig. 6C), indicating that the impact of WDHD1 on protein translation is independent of mTOR. We further validated several interactions of WDHD1 with the potential binding partners (including RPS6 and eIF3β) identified via the IP-MS analysis (Fig. 6E), highlighting the interactions between WDHD1 and the components of translational machinery.

WDHD1 levels are increased in TNBC compared to normal breast tissues, and associate with tumour size and proliferation
The clinical importance of WDHD1 in TNBC was evaluated in samples from TNBC patients. From TCGA analysis, WDHD1 mRNA levels were significantly higher in TNBC than the normal breast samples ( Fig. 7A; P < 0.0001). In addition, the number of patients with T2 and above in the high WDHD1 group was significantly larger than the low WDHD1 group ( Fig. 7B; P = 0.027).
The association between WDHD1 and clinicpathological features in TNBC patients was further investigated by immunohistochemistry (IHC) staining of WDHD1 in a TNBC tissue microarray. We found that tumour grade (P = 0.03) and tumour size (P = 0.016) were significantly correlated with WDHD1 expression (Table 1). Representative images of high and low expression of WDHD1 in TNBC are shown in Fig. 7C. Moreover, a positive correlation between WDHD1 expression levels (reflected by its IHC scores) and Ki67 percentage (a proliferation marker) was observed in TNBC ( Fig. 7D; Pearson's correlation r = 0.3714; P = 0.0004), suggesting a role of WDHD1 in regulating cell viability, in consistence with the above in vitro findings.

Discussion
As TNBC is difficult to be targeted and is molecularly heterogeneous, further stratification is needed. TNBC has been subdivided into six distinct subtypes: basal-like 1 (BL1), basal-like 2 (BL2), immunomodulatory (IM), mesenchymal (M), mesenchymal stem-like (MSL) and luminal androgen receptor (LAR) 9 . Another study reclassified TNBC into five stable subtypes: BL1, IM, M, MSL and LAR 29 . PTEN inactivation was observed in the BL1 subtype 29 , which was further confirmed in a recent in silico analysis, showing exceedingly poor clinical outcome 30 .
Loss-of-function mutations in TSGs, such as PTEN, are major genetic alterations leading to more challenges to identify targeted drugs since it is difficult to restore their functions 31 . Therefore, studies have been focused to target downstream signalling pathways that are altered by inactivation of TSGs 18,31 . Targeting synthetic lethality provides an alternative approach 32 . As the second most mutated gene following p53 in various cancer types 33 , various studies have been performed to identify PTEN synthetic lethal interactions in a variety of cancer types. These include mitochondrial complex I inhibitors 34 and chromatin helicase DNA-binding factor CHD1 in PTENinactive prostate cancer cells 35 , polynucleotide kinase/ phosphatase (PNKP) in PTEN-deficient lung and colon cancer cells, and NUAK family kinase 1 (NUAK1) in PTEN-deficient breast cancer cells 36 . In this study, using TCGA analysis coupled with a whole-genome siRNA screen in isogenic PTEN-negative and -positive TNBC cells, we identified WDHD1 as a synthetic essential gene in PTEN-inactive TNBC cells.
WDHD1, an orthologue of Ctf4 in budding yeast 37 and Mcl1 in fission yeast 38 , is a DNA-binding protein 39 that is known to play important roles in DNA replication and cell cycle 37,[40][41][42][43][44][45][46] . We also observed an important role of WDHD1 in cell cycle, especially in PTEN-inactive TNBC cells. The selective killing of WDHD1 depletion in PTENinactive TNBC cells was further validated in both 2D and 3D cultures. In addition, using IP-MS analysis followed by bioinformatics, we identified a potential, yet unknown function of WDHD1 in protein translation in PTEN null TNBC cells, which was further validated with puromycin incorporation assay to measure global protein synthesis. Depletion of WDHD1 significantly inhibits global protein translation in PTEN null TNBC cells, which is independent of mTOR inhibition and potentially via directly interacting with the translational machinery. The impact of WDHD1 depletion on global protein translation is similar to the effect achieved by re-introducing PTEN. PTEN inactivation in TNBC leads to a high activity of mTOR 47 , which is linked to a high rate of protein synthesis, creating an "Achilles heel" of TNBC. Indeed, several clinical trials on Everolimus (a mTOR inhibitor) in TNBC are ongoing (clinicaltrials.gov), some of which showed positive results 48,49 . However, a common pattern seen in trial data is of a modest response to rapalog (rapamycin and its analogues) monotherapy, which does not lead to a significant improvement in patient outcomes. One of the likely reasons is that it is caused by reactivation of signalling pathways that drive the high rate of protein synthesis required by tumour growth. Inhibition of WDHD1 in a PTEN-inactive background reduces protein translation, suggesting that such a "synthetic sickness" approach may be applicable to PTEN-deficient tumours when rapalog resistance happens.
In addition, a potential role of WDHD1 in regulating the stemness of PTEN-inactive TNBC cells was investigated using a mammosphere formation assay, which is one of the assays used to determine cell stemness 50 . Given the impact of WDHD1 on cell cycle and protein translation, both of which play important roles in regulating cell stemness 51 , we presume that WDHD1 may control stemness in PTEN-inactive TNBC cells via its ability to regulate cell cycle and protein translation; however, this remains to be elucidated. We found WDHD1 expression is significantly higher in PTEN-inactive TNBC cells than in PTEN-active ones. A previous report from Sato et al. 44 and colleagues suggested that AKT kinase seems to phosphorylate and stabilise the WDHD1 protein in cancer cells. In addition to the reported effects of AKT on WDHD1 protein stability, we found the mRNA levels of WDHD1 are also regulated by the PTEN-AKT pathway. Together, these data suggest that WDHD1 expression is affected by PTEN-AKT signalling in TNBC cells at both mRNA and protein levels. The clinical importance of WDHD1 in TNBC was evaluated in samples obtained from TNBC patients, showing that its levels are increased in TNBC compared to normal breast tissues, and associates with tumour size, stage and proliferation, using Ki67 as a proliferation marker 52 . Moreover, recent reports demonstrated that overexpression of WDHD1 leads to cisplatin resistance in lung adenocarcinoma 53 and metastasis in cholangiocarcinoma 54 . Further studies are required to confirm these findings in TNBC. The data presented here suggest that inhibitors that can disrupt the interactions between WDHD1 and the protein synthesis machinery could target some of the most intractable tumour types, such as TNBC with PTEN-deficiency. The relatively mild effects of WDHD1 depletion in PTEN-positive cells suggests that on-target inhibition of this factor may also be relatively free from unwanted side effects. Short-interfering RNA (siRNA) oligos against WDHD1 (D-019780-02 and D-019780-03) was purchased from Dharmacon. Sequences are available from Dharmacon, or upon request. siGENOME RISC-Free siRNA (Dharmacon) was used as a negative control. Cells were transfected with the indicated siRNA oligos at a final concentration of 35 nM using Dharmafect 2 reagent (Dharmacon).

The Cancer Genome Atlas (TCGA) data analysis
Expression of genes/proteins of interest, obtained from the cBioPortal for Cancer Genomics (https://www. cbioportal.org/) and UCSC Cancer Genome Browser (https://genome-cancer.ucsc.edu/), were analysed in each breast cancer molecular subtype along with normal breast samples (details provided in Supplementary Methods).
A whole-genome siRNA screen and data analysis The human siGENOME siRNA library-Genome (G-005005) was obtained from Dharmacon. siRNA transfection experiments were performed in 96-well format in antibiotic-free medium, using a reverse transfection employing 25 nM siRNA and 0.15 μl Dharmafect 2 (Dharmacon) per well together with a starting cell density  Triplicate data points from CherryFP channel (PTEN+) and GFP channel (PTEN−) screens underwent plate and position normalisation and Z score calculation using cellHTS software 56,57 . Differential Z scores (ΔZ score) between the two channels were subsequently used to create a gene hit list. Reproducibility of the replicates was analysed by performing Pearson correlation analysis in GraphPad Prism 8. P value < 0.05 was considered significant (details provided in Supplementary Methods).

Cell viability assay
siRNA transfected cells were plated into 96-well plate with a density of 8000 cells/well. CellTiter-Glo ® Luminescent cell viability assay (Promega) was performed 96 h post transfection according to the manufacturer's protocol using Glo-Max ® Discover Microplate Reader (Promega).
The mammospheres that were equal to or greater than 50 μm in diameter were counted to calculate the mammosphere formation efficiency (MFE%) with the following equation: (# of mammospheres per well)/(# of cells seeded per well) × 100. Additionally, the volumes of the mammospheres were also calculated using the formula of Volume = (4/3)πr 3 . ImageJ (version1.42q) was used to determine the MFE and volume of sphere.
CellTiter-Glo ® cell viability assay was performed with addition of 100 µl of CellTiter-Glo ® reagent into each well and incubated at room temperature for 1 h, followed by measuring using GloMax ® Discover Microplate Reader (Promega).

Western blot analysis
Western blot analysis was performed with lysates from cells lysed with urea buffer (8 M urea, 1 M thiourea, 0.5% CHAPS, 50 mM 1,4-Dithiothreitol (DTT) and 24 mM spermine). The bound proteins were separated on sodium dodecyl sulphate (SDS) polyacrylamide gels and subjected to immunoblotting with the indicated antibodies. For immunoprecipitations, the cells were lysed for 30 min at 4°C in pNAS buffer (50 mm Tris/HCl (pH 7.5), 120 mm NaCl, 1 mm ethylenediaminetetraacetic acid (EDTA) and 0.1% Nonidet P-40), with protease inhibitors. Anti-WDHD1 (Sigma-Aldrich) or control antibodies and Protein G magnetic beads (Thermo Fisher Scientific) were added to the lysate for 16 h at 4°C. Immunoprecipitates were washed four times with cold phosphate buffered saline (PBS) followed by the addition of SDS sample buffer. The bound proteins were separated on SDS polyacrylamide gels and subjected to immunoblotting with the indicated antibodies. Primary antibodies were from Cell Signalling Technology (

Immunohistochemical and H/E staining and scoring
Tissue microarray of TNBC patients with information of clinic-pathological parameters was purchased from Outdo Biotech (HBreD090Bc01; Shanghai, China). Tissue samples were pre-stained with Ki67. All procedures were approved by the Ethical Committee of Tongji Hospital, China. Informed consent was obtained from all subjects. For immunohistochemical staining, antigen retrieval, blocking of non-specific binding and incubation of primary antibodies at 4°C overnight were conducted sequentially. The primary antibody of anti-WDHD1 (HPA001122, Sigma-Aldrich, 1:500) was used. After incubation with secondary goat anti-rabbit immunoglobulin conjugated to peroxidaselabelled dextran polymer (SV0002; Boster) at 37°C for 1 h, visualisation, counterstaining with haematoxylin and mounting were performed. Semi-quantitative evaluations of protein expression were scored on the basis of the intensity and the percentage of WDHD1-positive tumour cells as previously described [59][60][61][62] .

Flow cytometry
For cell cycle analysis, 48 h post transfection, cells were fixed with 70% ethanol and kept at 4°C for up to 2 weeks. Cells were treated with 0.25% Triton-X-100, 200 µg/ml RNAse A and 50 µg/ml propidium iodide (PI), and analysed by FACS, Guava.

Immunoprecipitation-mass spectrometry (IP-MS) analysis
For immunoprecipitations of endogenous WDHD1, the cells were lysed for 30 min at 4°C in pNAS buffer (50 mm Tris/HCl (pH 7.5), 120 mm NaCl, 1 mm EDTA and 0.1% Nonidet P-40), with protease inhibitors. Anti-WDHD1 (Sigma-Aldrich) or control antibodies and Protein G Sepharose (GE Healthcare) were added to the lysate for 16 h at 4°C. Immunoprecipitates were washed four times with cold PBS followed by mass spectrometry analysis (details provided in Supplementary Methods).
Two repeats of WDHD1 and two repeats of IgG control samples were combined in RStudio (version 3.4.4), and the proteins with NA values in more than two samples were removed. The average of peptide numbers for WDHD1 and IgG control samples was calculated and ratio of peptide numbers for each sample group was calculated. The proteins which had two times higher peptide number in WDHD1 compared to the control samples were chosen as threshold and used to perform pathway analysis in ToppGene website as described below.

Bioinformatics
For pathway analysis, ToppGene Suite (https:// toppgene.cchmc.org/) was used to detect functional enrichment of the mRNAs or proteins. The pathways were sorted from lowest P value and top 15 pathways were chosen for TCGA data. We then produced a histogram plot with the top 15 pathways in GraphPad Prism 8. The pathways for IP-MS data were sorted from lowest P value and the histogram was plotted with top four pathways in GraphPad Prism 8.

Statistical analysis
Two tailed, unpaired Student's t test for the TCGA data and two paired, paired Student's t test for the wholegenome siRNA screening data were performed in RStudio (version 3.4.4). Codes are available upon request. Unless stated otherwise, comparison of two groups was statistically calculated by two paired, unpaired Student's t test in GraphPad Prism 8 software. Ordinary one-way ANOVA was conducted to statistically compare more than two groups in GraphPad Prism 8 software. Correlation analysis was conducted by Pearson's correlation in GraphPad Prism 8 software. χ 2 test was used to analyse the association of PTEN and WDHD1 with clinical features of TNBC samples in the TCGA breast invasive carcinoma data in GraphPad Prism 8 software. χ 2 test or Fisher's exact test was used to evaluate the relationship of WDHD1 and clinic-pathological parameters of TNBC patient samples in IHC using SPSS (version 19.0). Data were shown as box and whisker plot with minimum and maximum individual values, mean ± SD or mean ± SEM, indicated in figure legend.