Targeting TROP2 and Exploiting AUNIP as a Determinant of ADC Cytotoxicity and Chemosensitivity in HNSCC and ESCA Cancers

ABSTRACT

Head and neck squamous cell carcinoma (HNSCC) and esophageal carcinoma (ESCA) are anatomically contiguous, frequently co‐occurring malignancies that collectively account for substantial global cancer mortality. With half of patients progressing to recurrent or metastatic disease, alternative therapeutic strategies are urgently needed. Antibody‐drug conjugates (ADCs) offer promise, but their clinical impact requires identification of robust target antigens and elucidation of resistance mechanisms. Through single‐cell transcriptomic analysis followed by experimental validation, we identified TROP2 as a consistently and highly overexpressed cell‐surface antigen in HNSCC and ESCA. DS‐1062a, a TROP2‐directed ADC delivering the camptothecin (CPT)‐like topoisomerase I inhibitor deruxtecan (Dxd), demonstrated marked tumor regression in HNSCC and ESCA xenografts. This pronounced sensitivity may be attributable to prevalent homologous recombination deficiency (HRD) in these cancers, conferring vulnerability to DNA single‐strand break (SSB)‐inducing agents such as PARP inhibitors. However, therapeutic potential is constrained by an incomplete understanding of the regulators of drug sensitivity. To systematically dissect genetic modifiers of SSB‐induced cytotoxicity, we performed a genome‐wide CRISPR/Cas9 sgRNA knockout screen and identified AUNIP as a determinant of chemosensitivity. AUNIP loss profoundly enhanced DS‐1062a cytotoxicity and multiple SSB‐inducing chemotherapies across diverse HNSCC and ESCA models. Mechanistically, AUNIP deficiency led to impaired DNA repair, attenuated stress response signaling, and heightened apoptotic susceptibility. These findings establish TROP2 as a therapeutically actionable target and reveal AUNIP as a genetic rheostat governing ADC and chemotherapy response. Co‐targeting AUNIP represents a promising strategy to potentiate SSB‐inducing therapies and improve outcomes in patients with squamous carcinomas of the upper aerodigestive tract.

1. Introduction

Head and neck squamous cell carcinoma and esophageal carcinoma are anatomically contiguous malignancies that frequently co‐occur, reflecting shared risk factors. Collectively, they impose a substantial global burden: HNSCC ranks as the sixth most common cancer worldwide, with approximately 900,000 new cases and 450,000 deaths annually, while ESCA accounts for 500,000 new cases and 450,000 deaths annually [1]. Prognosis in HNSCC is highly stage‐dependent as early‐stage disease achieves 5‐year survival rates approaching 80%, but this figure plummets to approximately 20% for patients presenting with advanced stages [2]. Despite curative‐intent primary therapy, nearly half of all patients develop recurrent or metastatic disease [3]. Although the KEYNOTE‐048 trial established pembrolizumab plus chemotherapy as a standard‐of‐care regimen offering a survival benefit over cetuximab‐based therapy, such regimens are frequently limited by significant toxicity and diminishing options upon progression [4]. Thus, the development of alternative targeted therapeutic strategies remains an urgent unmet clinical need.

Poly (ADP‐ribose) polymerase (PARP) inhibitors have revolutionized the treatment of homologous recombination‐deficient (HRD) cancers. Notably, genomic studies revealed that many HNSCC tumors harbor alterations in HR‐related genes, including mutations in BRCA1, BRCA2, ATM, PALB2, and ATR, as well as promoter methylation of FANCF [5, 6, 7, 8, 9]. Preclinical evidence confirmed that HRD HNSCC models are sensitive to PARP inhibition, both as monotherapy and in combination with platinum‐based agents [10, 11, 12]. Clinically, the combination of olaparib, pembrolizumab, and carboplatin has shown promising anti‐tumor efficacy in HNSCC [13]. Nevertheless, the emergence of intrinsic and acquired resistance limits the long‐term utility of PARP inhibitors and other chemotherapies, underscoring the need for strategies to enhance their effectiveness.

Antibody‐drug conjugates (ADCs) have emerged as a transformative modality that leverages antibody selectivity to deliver potent cytotoxic payloads directly to tumor cells. The paradigm‐shifting success of ADCs such as trastuzumab deruxtecan (DS‐8201) has validated this approach and underscored advantageous features like the “bystander effect” [14, 15, 16], which is particularly relevant in the heterogeneous tumor microenvironment. Growing evidence also suggests that ADCs can synergize with immunotherapy by promoting immunogenic cell death and promoting antigen presentation [17, 18, 19, 20, 21]. Despite recent clinical advances, the successful deployment of ADCs in HNSCC and ESCA remains under clinical investigation [22, 23, 24, 25], and responses are frequently curtailed by resistance mechanisms across various cancer types [26, 27, 28, 29, 30, 31]. These limitations underscore the critical need for improved target identification or optimized ADC selection, and a deeper understanding of response determinants.

Through multi‐omic analysis, we identified TROP2 (encoded by TACSTD2) as a selectively overexpressed antigen in HNSCC cells. As a type I transmembrane glycoprotein overexpressed in diverse epithelial cancers and internalized upon antibody binding, TROP2 has emerged as a clinically validated target for ADC‐based therapies. Sacituzumab, a TROP2‐directed ADC, was evaluated in the TROPiCS‐03 basket trial, including 43 advanced HNSCC patients who had progressed on platinum‐based chemotherapy and anti‐PD‐(L)1 inhibitors. The objective response rate was 16% with seven confirmed partial responses, and median progression‐free survival was 4.1 months, demonstrating clinical potential in advanced HNSCC [24]. Consistently, we observed that DS‐1062a [32], another TROP2‐directed ADC that delivers CPT‐like payload Dxd, achieved marked tumor regression in HNSCC and ESCA xenografts.

This pronounced sensitivity may be attributable to prevalent homologous recombination deficiency in these cancers, conferring vulnerability to SSB‐inducing agents such as PARP inhibitors. To systematically identify genes that modulate SSB‐inducing agents, we conducted a genome‐wide CRISPR/Cas9 sgRNA knockout (KO) screen upon PARP inhibition. This approach revealed AUNIP as a key determinant of sensitivity to multiple PARP inhibitors such as olapaprib, niraparib, and talazoparib, as well as other DNA damage response (DDR)‐targeting inhibitors such as topotecan and camptothecin. AUNIP has emerged as a key regulator of replication stress response and DNA double‐strand breaks (DSBs) repair, influencing DNA damage‐inducing drug response. It has been shown to bind branched DNA structures characteristic of stalled replication forks and interacts with CtIP, an endonuclease critical for HR initiation [33, 34]. However, the role of AUNIP in mediating the sensitivity to ADCs with CPT‐like payloads in HNSCC and ESCA remains largely unknown. Moreover, the global cellular changes accompanying AUNIP loss that may underlie chemosensitivity have not been systematically characterized. Here, we demonstrated that AUNIP loss profoundly enhances TROP2‐ADC cytotoxicity in HNSCC and ESCA models. Through mechanistic studies, we established that AUNIP loss synergizes with PARP inhibitors and other SSB‐inducing agents via multifaceted mechanisms, including impaired DNA repair, attenuated stress response signaling, and heightened apoptotic susceptibility. Furthermore, we linked high AUNIP expression to adverse clinical outcomes and activated DNA damage response pathways in patient tumors.

Collectively, these findings establish TROP2 as a therapeutically actionable target and reveal AUNIP as a genetic rheostat governing ADC and chemotherapy response. Co‐targeting AUNIP represents a promising strategy to potentiate SSB‐inducing therapies and improve outcomes in patients with squamous carcinomas of the upper aerodigestive tract.

2. Materials and Methods

2.1. Single‐Cell RNA Sequencing (ScRNA‐Seq) Data Processing and Analysis

ScRNA‐seq data from GSE164690 [35] were analyzed using the Seurat package (v5.3.0) in R (v4.4.3) [36]. Cells with 200–6000 detected genes and mitochondrial gene percentage (^MT‐) below 20% were retained. The top 2000 log‐normalized highly variable genes were identified using “vst” method. Harmony integration was applied to the first 30 principal components with “orig. ident” as the batch variable. Cells were clustered using FindNeighbors and the FindClusters with a resolution of 0.5. For visualization, UMAP was performed using 20 harmonized dimensions. Cell types were manually annotated based on canonical marker gene expression: endothelial cells (PECAM1, VWF, CDH5, PLVAP), cancer‐associated fibroblasts (COL1A1, FAP, ACTA2), T cells (CD3D, CD4, CD8A, FOX8A), NK cells (NKG7, GNLY), B cells (CD79A, MS4A1, CD19), myeloid cells (CD14, CD68, CD163), dendritic cells (CLEC4C), and tumor cells (EPCAM, KRT5, MKI67).

2.1.1. Cell Lines

All cell lines were authenticated by short tandem repeat (STR) profiling. The HEK 293 T/17 and CT26. WT cell lines were obtained from the American Type Culture Collection (ATCC). AC16, HMEC‐1, FaDu, KYSE‐30, KYSE‐150, and Hela cell lines were obtained from the Shanghai Cell Bank, Chinese Academy of Sciences. Cells were maintained in a humidified incubator at 37°C with 5% CO₂. HEK 293 T/17 cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS). All other cell lines were cultured in DMEM/F‐12 medium supplemented with 10% FBS. Cells were routinely tested for mycoplasma contamination.

2.1.2. Animal Studies

All animal experiments were conducted in strict accordance with the guidelines for the care and use of laboratory animals and were approved by the Institutional Animal Care and Use Committee (IACUC) of the National Institute of Biological Sciences. Six‐week‐old NCG mice with comparable body weights were selected for this study. A total of 5 × 10⁶ FaDu or 6 × 10⁶ KYSE‐150 cells, resuspended in a mixture of PBS and Matrigel (6:4 ratio), were subcutaneously inoculated into the right dorsal flank of each mouse. Tumor growth was monitored using a digital caliper. When tumors reached approximately 120 mm³ in volume, mice were randomly assigned to treatment groups and given unique identifiers (n =  6–7 per group). Tumor volume was calculated using the formula: Volume = (length × width²) / 2. Following the group assignment, body weight and tumor volume were measured at regular intervals. Mice received a single intravenous injection of either IgG1‐Dxd control (40 µg/mouse) or DS‐1062a (40 µg/mouse) for FaDu xenografts on day 1, or two injections on days 1 and 9 for KYSE‐150 xenografts. The humane endpoint was defined as tumor volume reaching approximately 1500 mm³ or the occurrence of ulceration, at which point mice were euthanized, and the experiment concluded. Tumor growth curves and body weight changes were plotted for each group. Data are presented as mean ± SEM. The p‐value was calculated by a non‐parametric unpaired two‐tailed Mann–Whitney t‐test.

2.1.3. Antibodies

The following antibodies were generated by Cell Signaling Technology: Rabbit anti‐Cleaved PARP (Asp214) (9544S); mouse anti‐β‐Actin (3700S); rabbit anti‐Vinculin (13901S); rabbit IgG (3900S). Rabbit anti‐AUNIP (PA5‐55964) was generated by Invitrogen. Human IgG1 isotype control was generated by Selleckchem (A2051), human IgG1 kappa isotype control was generated by TargetMol (T78242). The following antibodies were generated by MedChemExpress (MCE): human IgG1 kappa isotype control (HY‐P99001), datopotamab (HY‐P99843), enfortumab (HY‐P99016), and trastuzumab (HY‐P9907). Anti‐B7H3/CD276 (Ifinatamab Biosimilar) antibody was generated by AlpVHHS (400‐043‐001).

2.1.4. Flow Cytometry Staining

For flow cytometry analysis, cultured cells were harvested and washed with DPBS. Primary antibodies were diluted to a working concentration of 0.1 mg/mL in DPBS supplemented with 2% fetal bovine serum (FBS). Subsequently, 10,000 cells per sample were incubated with the diluted primary antibody at 4°C for 1 h in the dark. Following incubation, cells were washed three times with PBS containing 2% FBS. Cells were then incubated with the corresponding secondary antibody (diluted 1:200) under light‐protected conditions for 1 h, followed by another three washes with PBS containing 2% FBS. The stained cells were collected and immediately analyzed on a flow cytometer. Data analysis was performed using FlowJo software, and graphical representations were generated using GraphPad Prism (version 9.0, GraphPad). The following antibodies were used in this study: human IgG1 isotype control (Selleckchem, A2051), human IgG1 kappa isotype control (TargetMol, T78242), human IgG1 kappa isotype control (MedChemExpress, HY‐P99001), datopotamab (MedChemExpress, HY‐P99843), enfortumab (MedChemExpress, HY‐P99016), trastuzumab (MedChemExpress, HY‐P9907), and anti‐B7H3 antibody (AlpVHHS, 400‐043‐001).

2.1.5. Compounds and Drugs

The following compounds were used: olaparib (Selleckchem, S1060), topotecan (Selleckchem, S1231), niraparib (TargetMol, T3231), talazoparib (TargetMol, T6253), camptothecin (TargetMol, T1123), Dxd (TargetMol, T11249L), DS‐1062a (Selleckchem, D4050), IgG1‐Dxd (MCE, HY‐164152).

2.1.6. Bioinformatic Analysis of Drug Sensitivity From the GDSC Database

Drug sensitivity data, represented as natural log‐transformed half‐maximal inhibitory concentration (IC₅₀) values (ln(IC₅₀)), were obtained from the Genomics of Drug Sensitivity in Cancer (GDSC) database (release 8.5) [37, 38]. Cell lines were annotated using The Cancer Genome Atlas (TCGA) classifications; those designated as HNSC, ESCA, BRCA, or OV were selected for analysis. Additionally, cell lines with BRCA1 or BRCA2 mutations, as identified in the DepMap database, were included. Sensitivity profiles for a curated set of indicated drugs were extracted.

2.1.7. Cell Viability and Clonogenic Assays

Cell viability was assessed using the ATP‐based CellTiter‐Glo (CTG) Luminescent Assay (Promega, G7573) according to the manufacturer's instructions. Cells were seeded in 96‐well plates, allowed to adhere overnight, and then exposed to increasing concentrations of the indicated drugs for 96 h. The half‐maximal inhibitory concentration (IC50) was calculated with Graphpad prism. For clonogenic survival assays, cells were plated in 12‐well plates and incubated overnight. PARP inhibitors were added at the time of plating and maintained for 6 days. All other compounds were added 24 h post‐plating and incubated for an additional 5 days. At the experimental endpoint, cells were washed with PBS, fixed with 10% formalin, and stained with 0.1% crystal violet.

2.1.8. Whole Genome CRISPR/Cas9 SgRNA KO Screening

The whole‐genome CRISPR/Cas9 sgRNA library (Cellecta) was transduced into CAL‐33‐Cas9 cells, and selected with puromycin (3 µg/mL). The infected pool was maintained for about 2 weeks to allow for complete gene knockout. Then the pooled library‐transduced cells were treated with niraparib or left untreated, and genomic DNA of each group was subject to two‐round PCR‐based library construction according to the manufacturer's instructions. The resulting amplicon libraries were subject to sequencing, and the reads were demultiplexed and aligned to the reference sgRNA library. Bioinformatic analysis was done to determine the distribution of sgRNAs in each group, as well as to identify sgRNAs that were depleted or enriched in niraparib treated group compared to the control group.

2.1.9. Molecular Cloning

The sgRNA sequences targeting the human AUNIP and murine Aunip genes were designed using Benchling software (Tables S1 and S2). Controls included non‐targeting sgRNAs (sgNT_0, sgNT_1) and a targeting sgRNA against the non‐mammalian Copgfp gene. Forward and reverse primers (20 bp) were annealed and PCR‐extended to generate double‐stranded inserts, which were ligated into the pRSG16‐U6‐(sg)‐UbiC‐TagRFP‐Puro backbone (Cellecta) or engineered mYongHong or mBaoJin backbones using T4 DNA ligase. Ligated products were transformed into Stbl3 E. coli, and plasmid DNA was extracted using a miniprep kit (MN, 740490.250). Separately, a non‐targeting control sgRNA was cloned into the PRSG17‐U6‐(sg)‐UbiC‐TagGFP2‐Puro backbone (Cellecta) to generate the sgNT‐GFP control plasmid.

2.1.10. Lentivirus Transduction

Lentiviral particles were generated in HEK 293 T/17 cells. Cells were seeded in 6‐well plates 24 h prior to transfection. For each well, a transfection mixture was prepared in 200 µL Opti‐MEM (Gibco) containing 0.35 µg pRSV‐Rev, 0.18 µg pMD2.G, 0.35 µg pMDLg/pRRE, and 1.2 µg of the indicated construct with 4.2 µl TransIT‐X2 transfection reagent (Mirus) for 20 min before dropwise addition to the cells. The medium was replaced 4–6 h post‐transfection. Viral supernatants were collected at 48 and 72 h and stored at −80°C. For transduction, target cells were infected with lentiviral supernatant supplemented with 8 µg/mL polybrene (Millipore). Positive transduced cells were selected with puromycin or blasticidin.

2.1.11. Competition Assay

Competitive co‐culture assays were performed to assess the effect of gene knockout on drug sensitivity as previously described [39]. Cells expressing a gene‐targeting sgRNA (e.g., sgAUNIP) labeled with one fluorescent marker (RFP or GFP) were mixed 1:1 with control cells (non‐targeting sgRNA) labeled with the reciprocal marker. A parallel control mixture containing RFP‐ and GFP‐labeled control cells was also prepared. Co‐cultures were treated as indicated: PARP inhibitors, DS‐1062a, or IgG1‐Dxd were added at the time of plating and maintained for 6 days; all other compounds were added 24 h post‐plating and incubated for an additional 5 days. Cells were harvested and analyzed by flow cytometry. The percentages of RFP‐ and GFP‐positive cells were quantified using FlowJo, and the GFP‐negative/GFP‐positive or RFP‐positive/GFP‐positive ratio was calculated for each condition. Changes in this ratio upon drug treatment, relative to untreated controls, were used to infer modulation of drug sensitivity. Fluorescent markers were used interchangeably (e.g., mYongHong, mBaoJin), and the appropriate ratio (mYongHong/mBaoJin or mBaoJin/mYongHong) was calculated accordingly.

2.1.12. Assessment of Genomic Knockout Efficiency by TIDE

Knockout efficiency was assessed at the genomic level using Tracking of Indels by Decomposition (TIDE). Target sites for each sgRNA were identified via Benchling, and PCR primers were designed to amplify a 500–1000 bp region flanking the predicted Cas9 cleavage site. For sgRNAs targeting high‐GC regions (e.g., AUNIP sg9, sg10, sg11), PCR and Sanger sequencing were optimized using a high‐GC polymerase system to ensure robust amplification and sequencing quality. Purified PCR products were subjected to Sanger sequencing, and the resulting chromatograms were analyzed using the TIDE online tool to calculate indel frequencies. Primer sequences are listed in Tables S3 and S4.

2.1.13. Protein Isolation and Western Blot

Protein isolation and western blot were done as previously described [39]. Briefly, protein lysates were prepared using RIPA buffer supplemented with protease and phosphatase inhibitors, and protein concentration was determined by BCA assay (Thermo Scientific) according to the manufacturer's protocol. Total protein was resolved by SDS‐PAGE on 4%–12% Bis‐Tris gels (Invitrogen), and transferred to nitrocellulose membranes (Cytiva). Membranes were blocked in 5% BSA in Tris‐buffered saline with 0.1% Tween 20 (TBST) and incubated overnight at 4°C with primary antibodies: anti‐cleaved PARP (1:1000, CST 9544S), anti‐AUNIP (1:250 or 1:500, Invitrogen PA5‐55964), anti‐β‐actin (1:6000, CST 3700S), or anti‐vinculin (1:4000, CST 13901S). After washing, membranes were incubated with HRP‐conjugated secondary antibodies and visualized using SuperSignal West Pico Plus or Femto substrates (Thermo Fisher). Band intensities were quantified using ImageJ. Background subtraction was performed using the Rolling Ball algorithm (radius 50 pixels), and the integrated density of each band was normalized to its corresponding loading control (vinculin or β‐actin) from the same lane.

2.2. In‐House RNA‐Seq and Data Analysis

FaDu was treated with niraparib for 72 h or topotecan for 48 h. The total RNA was isolated and 1 µg RNA was used for library construction with Fast RNA‐seq Lib Prep Kit V2 (Abclonal RK20306). Sequencing was performed on an Illumina NovaSeq X Plus, generating 30 million 2 × 150 bp paired‐end reads per sample. Adapters and low‐quality reads were trimmed by Trimmomatic v.0.39 using TruSeq. 3‐PE‐2 adapters, and then aligned to the UCSC hg38 genome by STAR v.2.7.11b (GENCODE v46) [40, 41]. Exon‐level counts were obtained using featureCounts in subread v2.0.6 [42]. Differential expression was analyzed with DESeq. 2 v1.44.0, followed by GSEA using clusterProfiler on log₂ fold change‐ranked protein‐coding genes [43, 44]. The raw sequence data have been deposited in the Genome Sequence Archive (GSA) in the National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Science [45, 46].

2.3. Assay for Transposase‐Accessible Chromatin With Sequencing (ATAC‐Seq)

ATAC‐seq was performed using the Chromatin Profile Kit for Illumina (Novoprotein Scientific; N248‐01A) following the manufacturer's protocol with minor adjustments. In brief, about 50,000 cells per sample were washed once in cold PBS and pelleted. Cells were incubated in lysis buffer to release nuclei, and nuclei were pelleted by gentle centrifugation. Nuclei were resuspended in the transposition mix supplied with the kit and incubated at 37°C for 30 min to allow Tn5‐mediated adapter insertion. After tagmentation, DNA was purified and then ligated with adapters, libraries were amplified by PCR, size‐selected and cleaned using magnetic beads, and quantified prior to sequencing on an Illumina platform.

2.4. ATAC‐Seq Data Analysis

Raw reads from the Illumina platform were trimmed and filtered, then clean reads were mapped to the human hg38 reference genome with Bowtie2 (v2.5.4) [47]. Duplicates, mitochondrial reads, and ENCODE blacklist regions were removed. To account for Tn5 insertion bias, reads were shifted (+4 bp for the positive strand and −5 bp for the negative strand) using alignmentSieve from deepTools (v3.5.6) [48]. Peak calling was performed using MACS2 (v2.2.9.1), yielding a set of high‐confidence accessible chromatin regions [49]. Read counts for each peak were quantified using featureCounts (Subread v2.0.3) [42]. Normalized peak count matrices were then subjected to PCA to evaluate overall sample relationships. Genome‐wide signal tracks (bigWig files) were generated using deepTools bamCoverage, and signal enrichment heatmaps and aggregate profile plots were produced using computeMatrix followed by plotHeatmap and plotProfile [48]. The raw sequence data have been deposited in the GSA in National Genomics Data Center, Chinese Academy of Science [45, 46].

2.5. The Cancer Genome Atlas (TCGA) Database Analysis

Clinical metadata and bulk tumor RNA‐seq data for head and neck squamous cell carcinoma and esophageal carcinoma cohorts were obtained from TCGA via the Genomic Data Commons (GDC). AUNIP expression levels were extracted and quantified as transcripts per million (TPM). For comparative analyses, TPM values were transformed to log₂ (TPM + 1). Differential expression of AUNIP between primary tumor tissue and matched adjacent normal tissue was assessed using a paired Wilcoxon signed‐rank test for the above two cohorts separately. Within the head and neck squamous cell carcinoma cohort, AUNIP expression was compared across pathological nodal (N) stages (N0 vs. N1/2/3). For pan‐cancer survival analysis, curated clinical annotations and harmonized AUNIP mRNA expression data (z‐scores relative to diploid samples) were downloaded from cBioPortal. Patients within each cancer type were stratified into AUNIP‐high and AUNIP‐low groups based on the optimal cutoff determined by the survminer R package. Kaplan‐Meier survival curves were generated, and differences in overall survival between expression groups were assessed using the log‐rank test. Hazard ratios (HR) and 95% confidence intervals (CI) were calculated using univariate Cox proportional hazards models. GSEA was conducted to identify biological pathways associated with AUNIP expression. Gene sets with a normalized enrichment score (NES) > 0 and adjusted p value < 0.05 were considered significantly enriched.

2.6. Pan‐Cancer Single Cell Analysis of AUNIP Expression

AUNIP expression was evaluated across cell lineages in a processed pan‐cancer scRNA‐seq dataset from Zenodo (DOI: 10.5281/zenodo.15554080). The dataset was normalized and log‐transformed using standard functions (sc. pp. normalize_total, sc. pp. log1p) in Scanpy (v1.10.4) [50]. Following this, dot plots were generated for each cancer type to visualize the variation in AUNIP expression across different cell types.

2.7. Statistical Analysis

Statistical analyses were performed using GraphPad Prism (v.9). Comparisons between groups were conducted using two‐tailed unpaired Student's t‐tests, as indicated in the figure legends. For omics datasets, the specific statistical methods are described in the corresponding legends. For mouse studies, statistical comparisons between groups were performed using the non‐parametric unpaired two‐tailed Mann–Whitney t‐test for final tumor volume. Details on replicates and data presentation are provided in the figure legends. Significance is denoted as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

3. Results

3.1. TROP2 Is Highly Expressed in HNSCC and ESCA Cells

To identify cell surface proteins enriched in HNSCC, we analyzed public scRNA‐seq data from 18 human HNSCC specimens [35]. Cells were classified into lineages using canonical markers (Figure 1A). Differential expression analysis of membrane protein‐encoding transcripts revealed upregulation of TACSTD2 in the tumor cell compartment compared to normal cell types (Figure 1B). Pan‐cancer analysis of bulk RNA‐seq data from the Human Protein Atlas (HPA) [51], confirmed that TACSTD2 expression was markedly elevated in head and neck and gallbladder cancers (Figure S1A).

Figure 1.

TROP2 is highly expressed on head and neck and esophageal cancer cell lines. (A) UMAP projection of single‐cell transcriptomes from the GSE164690 dataset, showing major cellular subsets identified by unsupervised clustering and annotation. (B) Feature plot showing log‐normalized TACSTD2 expression overlaid on the UMAP embedding. (C–K) Flow cytometry histograms showing TROP2 membrane protein levels in a panel of head and neck cancer, esophageal cancer, and normal immortalized cell lines. Data are representative of three independent experiments.

We next assessed surface protein abundance of TROP2 and several other established ADC target controls, including Nectin‐4, HER2, and B7‐H3 in various solid tumors [52, 53, 54] in a panel of HNSCC (CAL‐33, FaDu, SNU‐1076), ESCA (KYSE‐30, KYSE‐70, KYSE‐150), and immortalized normal cell lines (AC16, HEK293T/17, HMEC‐1). While Nectin‐4, HER2, and B7‐H3 showed no selective expression, TROP2 was highly expressed in most HNSCC and ESCA lines relative to normal cells, supporting its suitability as an ADC target (Figure 1C–K, S1B).

3.2. TROP2‐directed ADC Demonstrated Therapeutic Efficacy in HNSCC and ESCA Xenografts

To evaluate the therapeutic potential, we tested DS‐1062a in established subcutaneous FaDu and KYSE‐150 xenografts. The drug‐to‐antibody ratio (DAR = 4) was designed to enable targeted payload delivery with minimal systemic release. DS‐1062a treatment induced marked and durable tumor regression in both models compared with IgG1‐Dxd (DAR = 4) control. Serial body weight measurements revealed no differences between treatment groups, and no overt toxicity was observed, indicating a favorable tolerability profile (Figure 2A–G).

Figure 2.

DS‐1062a demonstrates therapeutic efficacy in FaDu and KYSE‐150 xenograft models. (A) Schematic of the in vivo experimental strategy. (B) Tumor growth curves of FaDu xenografts treated with IgG‐Dxd (DAR = 4) or DS‐1062a. Data are presented as mean ± SEM. Statistical significance was determined by the non‐parametric unpaired two‐tailed Mann–Whitney t‐test. ***p < 0.001 compared to the IgG‐Dxd control group. (C) Mouse body weights throughout the study. Data are shown as mean ± SEM. (D) Images of tumors excised from mice at the study endpoint. (E) Tumor growth curves of KYSE‐150 xenografts treated with IgG‐Dxd or DS‐1062a. Data are presented as mean ± SEM. Statistical significance was analyzed by the non‐parametric unpaired two‐tailed Mann–Whitney t‐test. **p < 0.01 compared to the IgG‐Dxd control group. (F) Mouse body weights in the indicated mice. Data are shown as mean ± SEM. (G) Images of tumors excised from mice at the experiment endpoint.

3.3. Diverse Response to Single‐Strand DNA Damages in HNSCC and ESCA Cells

The pronounced sensitivity to DS‐1062a may be attributable to the prevalence of HRD in these cancers. Given this, we hypothesized that HNSCC and ESCA tumors may harbor broader vulnerability to agents that induce DNA single‐strand breaks. To test this, we systematically analyzed DepMap data across more than 30 HNSCC and more than 30 ESCA cell lines. This revealed diverse responses to DNA damage‐inducing drugs, including SSB‐inducing agents such as PARP inhibitors and topoisomerase I inhibitors, similar to patterns observed in BRCA1/2‐mutant lines (Figure S2A). This heterogeneous response extended to breast and ovarian cancer lines regardless of BRCA status (Figure S2B). We validated these findings experimentally using clonogenic survival assays, which demonstrated diverse drug responses to olaparib, niraparib, and topotecan across multiple HNSCC and ESCA lines (Figure S3A–E). These results were corroborated by CellTiter‐Glo (CTG) assays showing dose‐dependent decreases in cell viability (Figure S4A–O).

3.4. AUNIP Determines Chemosensitivity and ADC Cytotoxicity

To discover determinants of SSB‐induced cytotoxicity, we performed a genome‐wide CRISPR knockout (KO) screen in CAL‐33 cells treated with niraparib (Figure 3A). The screen achieved high coverage, successfully recovering over 99.95% of the library's sgRNAs, confirming its robustness and quality (Figure 3B). A plot of log₂ read counts (control) against log₂ fold‐change (niraparib/control) for each sgRNA revealed a pronounced depletion of sgRNAs targeting the AUNIP gene in the treated group, suggesting that loss of AUNIP may potentiate the sensitivity to niraparib (Figure 3C). To validate this finding, we designed 18 sgRNAs targeting exons 1–3 of AUNIP (Figure 3D, Table S1) and transduced them, along with three control sgRNAs (sgNT_0, sgCOPGFP, sgNT_1), individually into FaDu cells. TIDE analysis confirmed efficient editing, and immunoblotting verified substantial AUNIP reduction (Figure 3E, S5A–R). Competitive co‐culture assays demonstrated that AUNIP depletion did not impair basal proliferation but markedly enhanced sensitivity to niraparib and topotecan (Figure 3F–H). This phenomenon could be applied to other HNSCC and ESCA models, including CAL‐33, SNU1076, and KYSE‐70, as AUNIP loss markedly enhanced sensitivity to those SSB‐inducing agents such as olaparib, niraparib, talazoparib, camptothecin, or topotecan (Figure S6–8). Importantly, since DS‐1062a delivers camptothecin‐like payload Dxd, we also observed that AUNIP loss potentiates the sensitivity to DS‐1062a but not IgG1‐Dxd in FaDu or KYSE‐150 cells (Figure 4A–N).

Figure 3.

Genome‐wide CRISPR/Cas9 screening reveals and validates that AUNIP deficiency sensitizes HNSCC cells to PARP inhibitors. (A) Schematic of the genome‐wide CRISPR/Cas9 sgRNA screening strategy. (B) Distribution and coverage of sgRNA reads in untreated and niraparib‐treated groups. (C) Scatter plot comparing sgRNA abundance between treatment groups. Each dot represents a unique sgRNA from the library. The x‐axis shows the log₂‐transformed read counts in the untreated group. The y‐axis shows the log₂ fold change (niraparib‐treated vs. untreated). (D) Distribution of eighteen sgRNAs targeting AUNIP across its exons. (E) Western blot analysis confirming AUNIP loss in FaDu cells. *, non‐specific band. (F) Relative cell proliferation of indicated groups measured by competition co‐culture assay in FaDu cells (end point vs. start point). Data are presented as mean ± SD. n = 3. Statistical significance was analyzed using two‐tailed unpaired Student's t‐test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 versus sgNT_1 group. (G, H) Drug sensitivity of indicated FaDu cell groups to niraparib (G) and topotecan (H) monitored by in vitro competition assay. Data represent changes in GFP−/GFP+ ratio (drug‐treated vs. untreated control). Data are presented as mean ± SD. n = 3. Two‐tailed unpaired t‐test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 versus sgNT_1 group.

Figure 4.

AUNIP loss enhances the responsiveness of FaDu and KYSE‐150 cells to Dxd and DS‐1062a. (A) Relative proliferation of indicated FaDu cell groups measured by competition assay. (B, C) Drug sensitivity of indicated groups to IgG‐Dxd (B), DS‐1062a (C), monitored by competition assay in FaDu cells. (D–F) Trace decomposition showing the spectrum and frequency of indels in FaDu cells. (G) Relative proliferation of indicated FaDu cell groups measured by in vitro competition assay. (H, I) Drug sensitivity of indicated groups to Dxd (H), DS‐1062a (I), monitored by competition assay in FaDu cells. (J, K) Trace decomposition showing the spectrum and frequency of indels in KYSE‐150 cells. (L) Relative cell proliferation of indicated KYSE‐150 cell groups measured by in vitro competition assay. (M, N) Drug sensitivity of indicated groups to IgG‐Dxd (M), DS‐1062a (N), monitored by competition assay in KYSE‐150 cells. In (A, G, L), data represent changes in GFP−/GFP+ ratio (end point vs. start point). In (B, C, H, I, M, N), data represent changes in GFP−/GFP+ ratio (drug‐treated vs. untreated control at the end point). In (A–C, G–I, L–N), data are presented as mean ± SD. n = 3. Statistical analysis was done with two‐tailed unpaired t‐test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 versus sgNT_1 group.

To exclude potential artifacts related to specific fluorescent proteins, we generated modified pRSG16 backbones in which the fluorescent marker was replaced with mYongHong (red) or mBaoJin (green) (Figure S9A). AUNIP depletion in FaDu cells using pooled sgRNAs delivered via these alternative vectors consistently sensitized cells to topotecan (Figure S9B–F). Likewise, AUNIP knockout in KYSE‐70 cells using the same approach increased the sensitivity to niraparib (Figure S9G–K). These results confirm that the observed synthetic lethality is independent of the fluorescent tag used for cell tracking.

3.5. AUNIP Loss Potentiated PARP Inhibitor‐Induced Apoptosis

To investigate the mechanism by which AUNIP loss sensitizes cells to DNA‐damaging agents, we assessed apoptosis induction in AUNIP‐deficient FaDu cells following PARP inhibition. Immunoblotting revealed a marked increase in cleaved PARP, a canonical apoptosis marker, specifically in AUNIP knockout cells treated with niraparib (Figure 5A). In contrast, minimal cleaved PARP was detected in untreated AUNIP‐deficient cells or in niraparib‐treated controls, indicating that apoptosis is substantially triggered by the combination of AUNIP loss and PARP inhibition. Densitometric quantification confirmed significantly elevated cleaved PARP levels in the AUNIP knockout, niraparib‐treated group compared to all other conditions (Figure 5B). These results establish that synthetic lethality between AUNIP inactivation and PARP inhibition drives apoptotic cell death.

Figure 5.

AUNIP loss promotes apoptosis upon niraparib treatment in FaDu cells, and sensitizes mouse cancer cells to PARP inhibition. (A) Representative western blot of cleaved PARP in control and AUNIP‐depleted FaDu cells. (B) Quantification of cleaved PARP levels normalized to β‐Actin. The error bars indicate the mean ± SD; n = 3. p values were calculated by tow‐tailed unpaired t‐test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 versus the sgNT_0 group. (C–G) Trace decomposition showing the spectrum and frequency of indels in CT26 cells. (H) Relative cell proliferation of the indicated mouse CT26 cell groups measured by in vitro competition assay. The error bars indicate the mean ± SD. P values were calculated by two‐tailed unpaired t‐test. n = 3. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 versus the sgNT_0 group. (I) The sensitivity of indicated groups to niraparib was evaluated by competition assay in CT26 cells. Mean and SD are shown. n = 3. p values were analyzed using unpaired t‐test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 versus the sgNT_0 group.

3.6. Conserved Function of AUNIP in Murine Cells

To determine whether the role of AUNIP in modulating PARP inhibitor sensitivity is evolutionarily conserved, we generated a pooled Aunip knockout CT26. WT mouse colon carcinoma cells using five independent sgRNAs, with confirmed gene disruption by TIDE analysis (Figure 5C–G). Consistent with human cell line observations, Aunip ablation did not impair proliferative capacity (Figure 5H) but confer increased sensitivity to niraparib (Figure 5I). These findings demonstrate that the synthetic lethal interaction between AUNIP loss and PARP inhibition is functionally conserved across species.

3.7. Transcriptomic Profiling Revealed Pathway Alterations Induced by AUNIP Loss Under Genotoxic Stress

To define transcriptional consequences of AUNIP loss under genotoxic stress, we performed RNA‐seq on FaDu cells expressing two independent AUNIP‐targeting sgRNAs or two non‐targeting controls, following treatment with niraparib, topotecan, or left untreated. Principal component analysis revealed enhanced segregation by treatment condition, with secondary separation by AUNIP genotype and tightly clustered biological replicates, confirming experimental reproducibility (Figure 6A–C). Differential expression analysis showed that AUNIP knockout alone induced minimal transcriptional changes under basal conditions (Figure 6D), indicating that AUNIP is not a major regulator of the basal transcriptome.

Figure 6.

Transcriptomic changes and pathway enrichment analysis upon AUNIP knockout in FaDu cells under SSB‐inducing drug treatment. (A–C) Principal component analysis (PCA) of RNA‐seq comparing AUNIP KO versus WT cells without treatment (A), treated with niraparib (B) or treated with topotecan (C) in FaDu cells. (D–F) Volcano plots showing the change of gene expression upon AUNIP KO under different treatment conditions. Significantly differential expression genes (DEGs) were defined as adjusted P value less than 0.05 and absolute log2 fold change greater than 0.5. (G) Heatmap displaying the row‐scaled score of each DEGs in FaDu upon niraparib or topotecan treatment. (H) Ridgeline plot showing significantly negative enrichment of the DNA repair‐related signatures in FaDu cells with AUNIP KO upon niraparib or topotecan treatment. Normalized enrichment scores (NESs) and adjusted P values were determined by GSEA. (I) Dot plots showing other representative signatures in FaDu cells treated with niraparib or topotecan. NESs and adjusted P values were determined by GSEA.

Given this established connection between chromatin dynamics and DDR [55, 56, 57], and because our transcriptomic analyses revealed only minimal transcriptional changes under basal conditions, we sought to investigate whether AUNIP loss might influence chromatin accessibility as a potential mechanistic explanation. To this end, we performed ATAC‐seq in AUNIP‐deficient and control FaDu and HeLa cells. Libraries exhibited expected fragment size distributions (Figure S10A), with high reproducibility between replicates (Figure S10B,C). Consistent with RNA‐seq findings, PCA showed no separation by AUNIP genotype (Figure S10D), and only minimal differentially accessible regions were identified (Figure S10E). These data suggest that AUNIP ablation does not globally rewire basal gene expression or chromatin accessibility.

In contrast, combined AUNIP loss and drug treatment provoked substantial transcriptomic remodeling. Hundreds of genes were differentially expressed in AUNIP‐deficient cells treated with niraparib or topotecan compared to untreated controls (Figure 6E–G). Gene set enrichment analysis revealed significant suppression of core DDR pathways, most prominently base excision repair (Figure 6H), alongside downregulation of cell cycle and proliferation‐associated programs including MYC targets, E2F targets, and pathways governing translation initiation, chromosome maintenance, and DNA replication. Consistent with increased apoptosis, GSEA confirmed significant enrichment of apoptosis‐related gene sets in drug‐treated AUNIP knockout cells. Collectively, these data indicate that AUNIP deficiency profoundly alters the transcriptional response to genotoxic agents, characterized by coordinated suppression of DNA repair and proliferative programs coupled with activation of apoptotic pathways.

3.8. AUNIP Expression Correlates With Adverse Clinical Outcomes and DNA Repair Pathway Activation

To evaluate clinical relevance, we analyzed AUNIP expression and patient outcomes using public transcriptomic databases. TCGA analysis revealed significantly elevated AUNIP mRNA levels in tumor versus matched normal tissues in both HNSCC and ESCA cohorts (Figure 7A). Single‐cell RNA‐seq data across multiple cancer types indicated that AUNIP expression is largely confined to the epithelial compartment (Figure 7B). Higher AUNIP expression in HNSCC was significantly associated with advanced lymph node metastasis (Figure 7C), and pan‐cancer survival analysis demonstrated that elevated AUNIP transcript levels correlate with worse overall patient survival (Figure 7D).

Figure 7.

AUNIP may serve as a potential prognostic marker in human cancers. (A) AUNIP is markedly upregulated in malignant versus adjacent normal tissues in HNSC and ESCA (TCGA). (B) Pan‐cancer single‐cell RNA‐seq profiling reveals consistently elevated AUNIP expression across diverse cancer types. (C) AUNIP expression increases with advancing nodal stage in HNSC, indicating an association with metastatic progression (TCGA). (D) Kaplan‐Meier analysis shows that elevated AUNIP mRNA levels are associated with significantly poorer overall survival (TCGA). (E, F) GSEA in TCGA‐HNSC (E) and TCGA‐ESCA (F) cohorts. Median mRNA expression level of AUNIP was used as the threshold to define AUNIP ^high and AUNIP ^low patients. Differential analysis was performed with DESeq. 2. NESs and adjusted P values were determined by GSEA.

GSEA of TCGA data revealed strong enrichment of DNA repair signatures in tumors with high AUNIP expression. Hallmark DNA repair and Reactome homology‐directed repair pathways were notably activated in both HNSCC and ESCA (Figure 7E,F). Additionally, high AUNIP expression was associated with upregulation of proliferative programs, including E2F targets, G2M checkpoint, and chromosome maintenance pathways (Figure S11A,B). These clinical and bioinformatic analyses establish that elevated AUNIP expression may be a feature of aggressive disease, correlating with metastatic progression and poor survival. The strong association with active DNA repair and cell cycle pathways aligns with our mechanistic data, supporting a model in which AUNIP may promote genomic stability and tumor proliferation. Collectively, these findings position AUNIP as a key mediator of DNA damage sensitivity and a promising therapeutic target.

4. Discussion

The emergence of antibody‐drug conjugates has transformed the therapeutic landscape for multiple malignancies, yet their application in squamous cell carcinomas of the upper aerodigestive tract has been hindered by a paucity of validated targets and an incomplete understanding of resistance mechanisms. Here, we establish TROP2 as a therapeutically actionable target in HNSCC and ESCA based on its expression profile and antitumor efficacy with the TROP2‐directed ADC DS‐1062a delivering Dxd. The pronounced sensitivity to TROP2‐directed ADC may be attributable to the prevalence of homologous recombination deficiency in these cancers, as genomic studies have revealed that a substantial subset of HNSCC tumors harbor alterations in HR‐related genes [5, 6, 7, 8, 9]. Clinically, the combination of olaparib, pembrolizumab, and carboplatin has shown promising antitumor efficacy in HNSCC, further supporting the relevance of DNA damage response pathways in these malignancies [13]. Following the identification of AUNIP as a mediator of PARP inhibitor sensitivity in HNSCC cells through an unbiased genome‐wide CRISPR/Cas9 knockout screen, our current study significantly extends the understanding of AUNIP by establishing its previously unrecognized role in determining ADC sensitivity. This finding is particularly timely given the expanding clinical use of ADCs employing camptothecin‐derived topoisomerase I inhibitors as payloads, exemplified by Dxd.

Based on our integrated transcriptomic analyses, AUNIP deficiency synergizes with genotoxic stress through multifaceted mechanisms. These include downregulation of critical DNA repair pathways, most notably base excision repair, attenuated stress response signaling, and suppression of proliferative programs such as E2F and MYC signaling. Concurrently, apoptotic pathways are robustly activated. This shift from repair and proliferation toward cell death provides a systems‐level explanation for the observed synthetic lethality beyond DNA repair capacity alone. Elucidating the precise molecular function of AUNIP in the DNA damage response remains an important direction for future investigation. Subsequent studies should investigate whether AUNIP is directly recruited to sites of SSB‐specific damage, as well as its role in modulating response to other DDR‐related agents such as TOP II inhibitors, DNA synthesis inhibitors, or replication stress inducers.

The clinical relevance of these findings is supported by bioinformatic analyses linking high AUNIP expression to adverse clinical outcomes, including lymph node metastasis and poor overall survival. Tumors with elevated AUNIP expression also exhibited activated DNA repair and cell cycle signatures, suggesting that cancers most dependent on AUNIP‐mediated genomic stability may be precisely those with the worst prognosis, highlighting a potential therapeutic window for AUNIP‐targeted strategies. While previous reports have positioned AUNIP as a pan‐cancer oncogenic factor associated with advanced disease, including in oral squamous cell carcinoma [58, 59], its role within the molecularly heterogeneous landscape of HNSCC requires further delineation. The efficacy of AUNIP‐targeting may vary significantly across HNSCC subtypes defined by HPV status, ancestral background, and distinct mutational profiles. A key priority will be to systematically evaluate AUNIP dependency across a comprehensive, molecularly annotated panel of HNSCC models and to identify robust predictive biomarkers for patient selection.

The translational promise of AUNIP inhibition lies in its potential to overcome or prevent clinical resistance to ADCs and SSB‐inducing agents. Our experimental observations support a model in which AUNIP is dispensable for basal proliferation but becomes an optimal synthetic lethal partner under genotoxic stress. In principle, as long as these agents exhibit selectivity for cancer cells over normal cells, the toxicity of such combination strategies may be manageable. Supporting this tumor‐specific context, our TCGA analysis revealed elevated AUNIP mRNA levels in tumor versus matched normal tissues in HNSCC and ESCA cohorts, and single‐cell RNA‐seq data indicated that AUNIP expression is largely confined to the epithelial compartment. Nevertheless, we acknowledge the potential role of AUNIP in normal cells and the possible toxicity of combining AUNIP inhibition with these agents. Future work should systematically assess these concerns in preclinical models. Given our emphasis on minimizing normal tissue toxicity, co‐targeting AUNIP with ADCs may represent a compelling strategy that maximizes antitumor efficacy while limiting systemic toxicity.

In conclusion, our study establishes TROP2 as a therapeutically actionable target in HNSCC and ESCA and identifies AUNIP as a pivotal genetic determinant governing ADC and chemotherapy response. By demonstrating that AUNIP loss disrupts DNA repair and licenses apoptosis specifically under genotoxic stress, we provide a mechanistic foundation for a dual‐target therapeutic strategy. These findings warrant the development of specific AUNIP inhibitors and their rigorous preclinical evaluation in combination with ADCs or chemotherapy, with the ultimate goal of translating this synergistic approach into improved outcomes for patients with squamous carcinomas of the upper aerodigestive tract.

Author Contributions

Guotai Xu, Liangfa Liu, and Jiaqi Bai conceived and supervised the study. Jiaqi Bai, Feifei ji, Xiyu Zhao, Zishu Yang, Jiahui Wei, Siming Chen, Wenxin Zhang, Haoran Wang, Xiaofeng Huang, Peijie Zhang, and Xiaoyu He performed the experiments and data analysis. Jiaqi Bai, Guotai Xu, and Liangfa Liu wrote the manuscript.

Ethics Statement

All animal experiments were conducted in strict accordance with the guidelines for the care and use of laboratory animals and were approved by the Institutional Animal Care and Use Committee (IACUC) of the National Institute of Biological Sciences.

Consent

All authors have approved the publication of this submission.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: TACSTD2 expression profiling in human cancer cell lines. Figure S2: The sensitivity of cancer cell lines across multiple tissues to DNA damage‐inducing agents or other chemotherapies through the GDSC database. Figure S3: The sensitivity of head and neck and esophageal cancer cell lines to DNA SSB‐inducing drugs by clonogenic survival assay. Figure S4: The sensitivity of head and neck and esophageal cancer cell lines to DNA SSB‐inducing drugs by cell viability assay. Figure S5: TIDE analysis of across eighteen AUNIP sgRNA constructs in FaDu cells. Figure S6: AUNIP loss potentiates the responsiveness of human CAL‐33 cells to SSB‐inducing drugs. Figure S7: AUNIP loss potentiates the responsiveness of SNU‐1076 cells to SSB‐inducing agents. Figure S8: AUNIP loss potentiates the responsiveness of KYSE‐70 esophageal cancer cells to SSB‐inducing chemotherapies. Figure S9: AUNIP loss enhances the sensitivity of SSB‐inducing drugs in FaDu or KYSE‐70 cells across backbones expressing different fluorescent proteins. Figure S10: AUNIP ablation does not globally rewire chromatin accessibility. Figure S11: AUNIP expression is linked to cell cycle or chromosome dysregulation in the HNSC and ESCA cohorts. Table S1: hAUNIP sgRNA sequences. Table S2: m Aunip sgRNA sequences. Table S3: Primers for hAUNIP TlDE analysis. Table S4: Primers for m Aunip TlDE analysis.

Data Availability Statement

The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive of the BIG Data Center at the Beijing Institute of Genomics, Chinese Academy of Science, under accession number HRA015208 (accessible at http://bigd.big.ac.cn/gsa‐human). Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.