Establishment and Preclinical Application of Conditional Reprogramming Culture System for Laryngeal and Hypopharyngeal Carcinoma

Purpose Management of laryngeal and hypopharyngeal squamous cell carcinoma (LHSCC) remains highly challenging due to their anatomic location and highly variable therapeutic responses. We aim to establish a new in vitro model for LHSCC based on conditional reprogramming (CR), a novel cell-culture technique, and investigate its potential value on personalized cancer therapies. Methods Primary LHSCC cells were isolated from tumor specimens and cultured under CR conditions. The characteristics and malignant potential of cells were evaluated by histological staining, whole-exome sequencing and heterotransplantation. The responses of CR tumor cells to anticancer drugs and radiotherapy were tested using cell proliferation assay. CR cells could form xenografts and organoids, which were used for drug testing respectively. Clinical responses for certain patients were also compared with in vitro responses. Results A panel of 28 human LHSCC CR cells were established from 50 tumor tissues. They retain tumorigenic potential upon xenotransplantation and recapitulate molecular characteristics of LHSCC. Differential responses to anticancer drugs and radiotherapy were detected in vitro. CR cells can be transformed to xenograft and organoid, shared comparable drug responses. The clinical drug responses were consistent with in vitro drug responses. Conclusions The patient-derived CR cell model could promisingly be utilized in clinical decision-making and assist in the selection of personalized therapies for LHSCC.


Introduction
The larynx and hypopharynx are anatomically and functionally intimately associated adjacent structures and are lined by a strati ed mucosa that protects the underlying structures. Malignant tumors arising from the epithelium of the larynx and hypopharynx are predominantly squamous cell carcinomas, constituting the most frequent malignancies of head and neck squamous cell carcinomas (HNSCC). Head and neck cancer represents the seventh most frequently diagnosed malignancy with 890,000 newly diagnosed cases and 450,000 cancer-associated mortality worldwide in 2018 (Bray et al. 2018). Furthermore, approximately 40% of laryngeal and 60% of hypopharyngeal cancer patients is presented with advanced laryngeal and hypopharyngeal squamous cell carcinoma (LHSCC). Despite remarkable advances in the understanding of etiopathogenesis and enhancement in treatment strategies of LHSCC, long-term survival rates in patients with advanced LHSCC have not improved signi cantly over the past several decades (Shah et al. 1997). Typically, majority of patients with LHSCC are managed with surgical resection followed by ionizing radiation or chemoradiation, or chemoradiation alone.
However, the management of LHSCC remains very challenging due to the anatomical location of the tumor that complicates surgery and highly variable treatment response to radiotherapy and chemoradiation therapy. Moreover, resistance to radiotherapy or chemotherapy frequently occurs, one of the signi cant causes of tumor recurrence, and is associated with poor outcome. Although these curative therapeutic approaches can achieve high rates of locoregional tumor control, they affect laryngeal function, resulting in a permanent tracheostomy, which adversely affects the patient's quality of life. Thus, a precise and optimal personalized multimodality treatment strategy is highly desirable for patients with LHSCC.
Improved understanding of the molecular mechanism contributing to the initiation and progression of LHSCC and in vitro drug screening is highly desirable to identify molecular determinants of LHSCC and target molecules for targeted therapeutic strategies. The introduction of next-generation sequencing technologies and increasing tumor molecular pro ling in clinical settings has revolutionized personalized treatment strategy. However, doubts have been raised for negative evidence as less than 7% of patients had responses to targeted drugs paired with identi ed mutations (Prasad 2016), and only around a quarter of patients had treatments selected by individual genomic analysis (Tannock and Hickman 2016). These ndings indicate the signi cance of developing robust patient-derived primary cell systems to predict clinical e cacy during the in vitro drug screening studies (Godoy et al. 2013). However, success rates to establish HNSCC-derived cell lines were lower than 35% (Méry et al. 2017), fronting the obstacles of limited lifespan and contamination by microbes and broblasts. Moreover, the relationship betweenin vitro drug sensitivity and clinical responses remains elusive (Dohmen et al. 2015).
Conditional reprogramming (CR) system, a novel platform to establish a long-term culture of primary epithelium cells derived from normal and tumor tissues, has gained increasing attention in recent years (Liu et al. 2017; Liu et al. 2012). CR technique is a simple co-culture method with a Rho-associated kinase (ROCK) inhibitor (Y-27632), combined with broblast feeder cells, which rapidly expands both normal and malignant broblast feeder cells derived from various anatomic sites and to propagate human epithelial cells. Mechanistically, CR acts through several signaling pathways to increase telomerase activity and hTERT expression, and ROCK inhibitor, inactivating RB/p16 signaling pathway (Liu et al. 2017;Liu et al. 2012). The reported advantages of CR include high success rate, exponential growth, genotype stability, and ease of manipulation. These characteristics enabled CR as an exceptional in vitro model compared with the others, such as patient-derived xenotransplantation (PDX) and organoids. Owing to these advantages, CR has been documented by two American National Cancer Institute programs: PDMR (patient-derived cancer model repository) and HCMI (human cancer model initiatives) . However, the potential application of CR cell technology for long-term culture of laryngeal and hypopharyngeal tissues has not been reported. Although the CR technology is robust, it exhibits certain limitations, such as it preferentially proliferates non-malignant epithelial cells from nasopharyngeal carcinoma and non-small cell lung cancer specimens (Gao et al. 2017;Yu et al. 2017). Moreover, the clinical relationship of CR has not been completely validated. This study has attempted to explain and provide the solution to these limitations to a certain extent.
The present study aimed to establish a primary culture system that enabled the rapid ampli cation of genetically stable LHSCC cells with a high success rate. As a versatile in vitro model, CR could be transformed into organoids and be used to produce CR-derived xenografts. Both systems share comparable drug responses. This study also provided a preliminary investigation into the relationship between in vitro CR cell responses and clinical responses, which may contribute to its potential clinical implications. radiologists and physicians. Written informed consent was obtained from the patients and/or their authorized representatives. Inclusion criteria were as follows: primary or recurrent histologically con rmed LNSCC patients; aged above 18 years old; and fresh tissue available through either biopsy or surgical resection of the primary tumor site. Patients with cognitive impairment, mental health disorders, poor compliance, or allergic to chemotherapeutic agents were excluded. The CR cell viability and the clinical response were evaluated by technicians and clinicians double-blinded, respectively. All human tissue samples were obtained from diagnostic biopsies or therapeutic resections. Prior to surgery or biopsy, each patient signed written informed consent, allowing the excess tissue to be used for research studies.

Tissue Processing
Upon receipt of fresh tissue, the tissue sample was into three parts for cryopreservation, xation, and digestion for primary cell derivation. For histology, a piece was removed and immediately xed in formalin. The xed tissue was processed and embedded in para n as described previously (Driehuis et al. 2019). For primary cell culture, tissue samples were minced and incubated at 37°C in 0.125% Trypsin (Sigma, catalog no. T1426) with high glucose DMEM (Life Technologies, catalog no. 12430-054) until digested. The tissue suspension was frequently agitated and monitored for up to 60 minutes. The suspension was strained through a 100 μm lter, centrifuged at 300 g and lysed with blood cell lysis buffer for 5 minutes. After washing twice with PBS, the resulting pellet was resuspended in Complete F medium and seeded in the CR culture system.

CR culture
Mouse embryonic broblast cell line 3T3-J2 (RRID: CVCL_W667; purchased from Otwo Biotech, Shenzhen, China) was cultured in complete DMEM with high glucose supplemented with 10% (v/v) FBS (Life Technologies) and 100 IU/ml penicillin, and 100 mg/ml streptomycin. In the CR system, 3T3-J2 were mitotically inactivated either by irradiation or by mitomycin C-treatment (2.5 h, 4mg/ml nal concentration, Sigma-Aldrich). Primary LHSCC cells were cultured in Complete F medium (Table 1) at 37℃ in a 5% CO2 humidi ed incubator. The medium was renewed every two days. The cell numbers of every passage were checked by a cell counter plate.

CRC derived organoid culture
The primary cells were collected from the CR culture system when the cells reached 70-80% con uence; the feeder cells were removed following trypsinization for 1-minute. The cells at an indicated count were mixed with ice-cold Matrigel and then were loaded in the center of the well of the culture plate. After polymerization by incubating at 37°C for 30 min, a prewarmed organoid medium was added to the plate. The medium was changed every alternate day. The main organoid culture methods and composition of the organoid medium were as previously described Immuno uorescence staining For histological examination, excised patient tissues or heterotransplanted tumors from nude mice were xed overnight in 4% formaldehyde, dehydrated, and embedded in para n and followed with depara nization and standard hematoxylin & eosin (H&E) staining. Images were acquired on an inverted microscope (TH4-200, Olympus optical Co-Ltd, Tokyo, Japan). Cells slides were used for indirect immuno uorescence. Brie y, cells were seeded into a 24-well plate with round cover slides (a diameter of 1 cm) in the well. After reaching 60-80% con uence, cell slides were xed in paraformaldehyde for 15 minutes and acetone successively. After xation, heat-induced antigen retrieval was performed using either citric acid solution in a microwave. The slides were then permeabilized with 0.1% Triton X-100 (Sigma) for 10 minutes and blocked with 1% (w/v) bovine serum albumin (BSA) for 1 hour at room temperature. Following incubation overnight at 4˚C with a primary antibody (anti-pan-keratin, proteintech, 26411-1-AP; anti-CD44, proteintech, 15675-1-AP), the cells were washed with PBS and incubated with secondary antibodies (Invitrogen) at room temperature for 1 hour. The cells were then incubated with indicated additional stains (DAPI, life technologies D1306) for 5 minutes at room temperature. The samples were analyzed using a confocal microscope (LSM880; Carl Zeiss, Germany). For immuno uorescence staining of the organoids, the whole mount staining method was performed as described previously (Hu et al. 2018). Primary antibodies used for organoids included anti-KRT5 (Santa Cruz Biotechnology; sc-32721) and anti-p63 (Abcam; ab124762). Secondary antibodies included goat anti-rabbit IgG (Alexa Fluor R 594; Invitrogen; CA11012s) and goat anti-mouse IgG (Alexa

In vitro drug screening
The cells at a density of 2000 cells/well were seeded into 96-well culture plates. After 24 h, the cells were treated with different concentrations of the drugs. Control cultures received an equal amount of DMSO (0.01 to 0.1%). 72 h after treatment. The number of cell colonies was estimated by using the CCK8 assay with a slight modi cation. Speci cally, cells treated with different concentrations of the drugs were washed twice with PBS, and then CCK8 solution (100 μL, 10 mg/mL) (Dojindo) was added into each well at 37°C for 2 hours. Following incubation, the absorbance (optical density) was measured at 450 nm using a microplate reader (Thermo Scienti c Multiskan FC).
The values were normalized to the vehicle (100%) and baseline control (0%). For each test, if the calculated cell viability was higher than 70% or lower than 30%, an additional screen was performed for that particular drug with an adjusted dose of the drug for the cell line. Z factor score was used as the parameter for screen quality assessment using the following equation: Drug screens with a Z score of less than 0.3 were not used and repeated. Kill curves were generated using GraphPad® PRISM version 9.0 (Graph Pad Software, Inc., La Jolla, CA, USA), and the curves were tted using the log (Inhibitor) vs. response --Variable slope (four parameters)." The half-maximal inhibitory concentration (IC50), which is an essential indicator for drug sensitivity assay, was calculated by non-linear regression of the log of concentration versus the percentage of survival, implemented in GraphPad.

Radiation sensitivity
The cells at a density of 2000 cells/well were seeded in 96-well culture plates. After 24 h, cells were irradiated. The γray irradiation was performed from a cobalt-60 source at a dose rate of 0.59 Gy/min at room temperature. A separate plate was used for each radiation dose. Plates were sealed air-tight and irradiated with a single fraction of 2, 4, 6, 8, 10 Gy. After radiation, the medium was changed. Six days later, cell viability was measured using CCK8 assay. Kill curves were graphed following the method described above.

Colony formation assay
The cells were seeded into 6-well culture plates immediately after exposure to 0 Gy and 4 Gy of γ-ray irradiation.
After 7 days of incubation, the colonies were xed in paraformaldehyde and stained with crystal violet solution.
Colonies containing more than 50 cells were counted; the relative colony-forming e ciency was calculated and plotted.

Whole-exome sequencing and bioinformatics analysis
Whole-exome sequencing data were mapped against human reference genome GRCh37, and variants were called under speci c pathogen-free conditions, at a temperature of 24 °C with a relative humidity of 50% -60%, under a 12-h-light/12-h-dark schedule. Animals were provided ad libitum access to standard rodent food and tap water. Mice were subcutaneously injected with 5 × 10 6 of primary cancer cells in the right ank (0.2 mL cell suspension per mouse). Six weeks after tumor cell inoculation, tumors were removed, and tumor tissues were xed in 4% formaldehyde, embedded in para n, and subjected to an H&E staining procedure.
For in vivo treatment assay, tumor-bearing mice were established following the above-mentioned method. When tumors reached approximately 200 mm 3 , the mice were randomized into three groups (n ≥3/group) according to tumor volumes and body weights. The treatments included vehicle control, 4 mg/kg cisplatin by intraperitoneal (i.p.) injection twice a week, 5Fu (100 mg/kg/week; i.p.), and paclitaxel (30 mg/kg/week; i.p.). Tumor volumes were measured using an electronic Vernier caliper and calculated with the formula . On the 28 th day after the rst treatments, mice were weighed and then euthanized with CO 2 asphyxia. Subsequently, the tumors were harvested, weighed, and photographed.

Statistical analysis
Statistical analysis was performed using GraphPad Prism 9.0 (GraphPad Software, Inc). All in vitro experiments were performed in triplicate and repeated three times. Data were expressed as mean ± standard deviation (SD). A two-tailed Student t-test was used to analyze differences between two groups. One-way analysis of variance (ANOVA) with Bonferroni correction was used to analyze multiple groups. The mean ± standard deviation was presented in all graphs, and raw data points were indicated. P values of <0.05 were considered statistically signi cant.

Establishment of patient-derived matched primary normal and LHSCC CR cells
We obtained normal and tumor samples from patients with a histopathologically con rmed diagnosis of laryngeal or hypopharyngeal squamous cell carcinoma. The media compositions were optimized to propagate primary cells based on the published protocols for CR (Liu et al. 2017). Conditions that were successful in growing breast and prostate epithelium were re ned on head and neck tissues. Brie y, the epithelial layer from the surgical specimen was microdissected to eliminate fat and muscle, digested in trypsin, and ltered. The resulting cell suspension was then seeded onto an NIH-3T3 broblast feeder layer in Complete F Medium (Fig. 1a). The detailed composition of Complete F Medium was listed in Table 1. Within the rst few days after inoculation, the adherent LHSCC cells exhibited small colonies and progressed into cell islands. As illustrated in Fig. 1b Table 2). The established criteria for CRC were long-term (>10 passages) proliferation and successful cryopreservation and recovery. These tumor CRCs were designated as T1, T2, T3, and so on, and the corresponding normal CRCs were indicated by N1, N2, N3, and so on.
The histological features were compared between CRCs and original tumors using H&E sections. CRCs exhibited similar histological patterns to their original tumors (Fig. 1b). In consideration of the heterogeneity of tumor tissues, the purity of epithelial-derived cells in CRCs was evaluated using immuno uorescence. Pan-keratin was used as an epithelial marker. Positive staining of pan-keratin was found in all cells (Fig. 1B). The expression of CD44, a wellknown stem cell marker for head and neck cancer (Leinung et al. 2015), was higher in tumor CRCs than in its matched normal CRCs (Fig. 1b).
CR method conditionally induces long-term and stable proliferation. On average, CRC could form small colonies within 2-5 days and be passaged within 5-7 days. After the rst passage, CRC typically proliferated at a stable rate, being passaged every 2-5 days with a split ratio of 1:2. For instance, as shown in Fig. 1c, primary cells were growing at different rates, which were passaged every 2-7 days after reaching con uence in CR co-culture and continued to proliferate at a steady rate for at least 40 days with 10 population doublings till the end of the experiment. STR analysis con rmed that the primary cells were genetically stable up to a span of around 10 passages (Table 3).
The tumorigenic potential of the cultured LHSCC cells was evaluated by subcutaneous transplantation of the tumor CSCs into nude mice. Transplantation in all four tumor lines yielded macroscopically visible tumors after 6 weeks in at least 1 of 4 mice (n =4 for each CRC line; Fig. 1d), while injection with the normal CSCs did not result in outgrowth. H&E staining of the tumors revealed strati cation and keratinization characteristics of LHSCC (Fig. 1e).
The tumor cells exhibited atypia, as observed in cancerous cells. Nuclear pleomorphism was also observed. Besides, muscle invasion was detected in one case. Together, LHSCC CRC lines retain tumorigenic potential and form xenografts with similar characteristics to the parental tumor.
CRC as a platform for chemotherapy and radiotherapy sensitive assay We exposed multiple CRC lines to cisplatin, 5Fu, paclitaxel, and cetuximab, drugs currently used to treat patients with LHSCC. The in vitro concentration was based on the peak exposures observed at the highest clinically recommended doses delivered as a single administration (Liston and Davis 2017). We observed differential sensitivity of the CRC lines to these compounds (Fig. 2a-d). Based on the measured IC50 values, we ranked the CRC lines tested for cisplatin, 5Fu, and paclitaxel (Fig. 2a-c). For the sensitivity of cetuximab, the area under the curve (AUC) was applied as an alternative to IC50 values because the curvature of the kill curve was not suitable to calculate IC50 (Fig. 2d). To assure the quality of the drug screening experiments, a Z-score, a parameter of assay quality (Zhang et al. 1999), was implemented and calculated for each drug screen (Fig. 2e). The average Z-score Radiotherapy is another major treatment modality for LHSCC. Thus, we investigated the sensitivity of CRCs to ionizing radiation. Kill curves of radiotherapy were drawn, and AUC values were calculated and ranked among the ve tested CRC lines (Fig. 2g). Differential radiotherapy responses were observed among the CRC lines. The ndings suggested that CRCs had the potential to re ect patients' clinical responses to radiotherapy.

Tumor CRCs Recapitulate Genetic Alterations Identi ed in LHSCC
To determine whether the CRCs recapitulated genetic alterations found in LHSCC, whole-exome sequencing was conducted on matched tumor and normal CRCs from 3 patients with LHSCC. In general, the mean sequencing depth was 96.34×, and a mean of 90.99% of the target sequence was covered to a depth of at least 20× ( Table 4). The somatic mutation load per subject varied signi cantly in LHSCC (mean 208, range 187-248; Fig. 3a). The spectrum of mutations was also illustrated in Fig. 3b.
A mutation lists were ltered for those genes that are most commonly affected in LHSCC (Stransky et al. 2011). We scrutinized all single nucleotide variants (SNV) and small insertions or deletions (Indel) throughout the genome in the tumor and normal CRC lines. Using this criterion, we detected pathogenic mutations in 2, 5, and 7 LHSCC cancerassociated genes, in T1, T2, and T7 lines, respectively (Fig. 3c). The tumor CRC lines also revealed SNVs and Indels that were absent from the normal CRC lines. The most commonly mutated gene in LHSCC, TP53, was genetically altered in 2 of the 3 tumor lines. The other tumor line, though without TP53 mutation, exhibited the mutation of MDM2, an oncogene encoding protein MDM2 binding and inhibiting P53. Thus, all the 3 tumor lines suffered from the functional alteration of P53. Notch1 was altered in 2 of the 3 CRC lines. Genes affected in one case include DICER1, FAT1, IRF6, MDM2, NOTCH2, PKHD1L1, RIMS2, SI, SMARCB1, and SYNE2. In the T6 line, we sequenced different mutation types in FAT1, NOTCH1, and TP53, such as missense, frameshift, deletion, or splice variant. Normal CRC lines lacked these genetic alterations, con rming they consisted of nontumor cells. The copy number variants of cancer-associated genes were also represented in Fig. 3d. TP53 gene loci of the above CRC lines were sequenced. In detail, T1 exhibited no TP53 mutation. T2 harbored TP53 exon deletion mutation; T6 harbored TP53 exon deletion mutation and a distinct SNV, Pro72Arg, which was a de ned bad mutational status indicating cisplatin-resistance (Bergamaschi et al. 2003). According to previous studies, resistance to cisplatin-based chemotherapy was positively correlated with the TP53 mutation burden (Bergamaschi et al. 2003). T1, T2, and T6 CRC lines were used to test the cisplatin treatment sensitivity. We found that CRC line T1 had the lowest IC50 than the others. Notably, T6 with two TP53 mutations propagated better than T2 with a single TP53 mutation under the same concentration of cisplatin, indicating that T6 was resistant to cisplatin treatment, and T2 was relatively sensitive to cisplatin (Fig. 2a). These ndings suggested that wholeexome sequencing and CRC lines derived from patient samples could predict response or resistance to individual therapy and could be used to evaluate target therapy based on gene mutation.

Relationship of drug responses between in vitro CRCs and xenografts in nude mice
Following in vitro ndings of the routine chemotherapy agents against LHSCC, we then validated their e cacy in suppressing the growth of xenograft tumors in vivo. As con rmed in vitro, T2 was relatively sensitive to cisplatin and 5Fu, while T6 was relatively resistant to the two drugs ( Fig. 2a-b). Hence, the same doses of cisplatin or 5Fu were administered to nude mice bearing xenografts of T2 or T6 and compared with their corresponding vehicles ( Fig. 4a, 4d). The tumor weight and mouse body weight at the endpoint were also measured (Fig 4b-c, 4e-f). Under the same treatment, the growth of T2 xenografts was inhibited evidently, while the growth of T6 xenografts was not hampered signi cantly compared with normal controls. Moreover, T6 was relatively sensitive to paclitaxel compared to the others according to the previous in vitro drug response test (Fig. 2c). The growth of xenografts was markedly suppressed with paclitaxel administration, rather than cisplatin or 5Fu (Fig. 4d). Hence, the CRCs and xenografts exhibited consistent drug responses.
CRC derived organoids and the application for drug testing Traditionally, 3D cultures have represented a widespread system to recapitulate the structural organization of primary tissues. CRCs could be transformed into organoids using the embedded method. In brief, the feeder cells were removed by rst trypsinization from the CR culture system, and then the primary cells were collected by second trypsinization. Next, cells were mixed with Matrigel, and the Matrigel-cell mixture seeded into the well of the culture plate. After solidifying the gel, a dome-like structure was formed to provide the cells with a 3D growing environment.
The viable cells could proliferate into spheres under this condition: CRC derived organoids (Fig. 5a). Consistent with previous studies performed on head and neck cancers (Driehuis et al. 2019), CRCs successfully formed spheres that resembled the morphology. The non-malignant cells expanded and re-associated to spheres of approximately 100 µm with a mass-like morphology and polarized growth (Fig. 5b), while smaller spheres distinguished their malignant counterpart and a relatively slow growth rate. Immuno uorescence (IF) analysis revealed that the organoids were composed of basal cells expressing KRT5 and p63 in the outer cell layer. In contrast, keratinized and differentiated cells with enlarged nuclei were inside the organoids (Fig. 5b). Furthermore, cisplatin was used to treat tumor organoid T1. Bright-eld images of the organoid under different drug concentrations were shown in Fig. 5c. Organoids treated with 1 µM cisplatin were smaller and sparsely distributed than those treated with 0.1 µM cisplatin.
Kill curve was plotted together with the IC50 value of 0.5 µM for cisplatin in T1 (Fig. 5d).

Relationship between in vitro CRCs responses and clinical responses: special cases
To demonstrate the relationship between in vitro and clinical responses, 4 special cases were selected. Their corresponding CRC lines were derived from their biopsy tissues. After diagnosed with LHSCC, they received chemo/radiotherapy prior to surgery. Before and after the chemo/radiotherapy, imaging examinations were conducted twice to evaluate their clinical responses to the treatment according to the RECIST criterion (Eisenhauer et al. 2009). The timeline of diagnosis and treatment procedure was depicted in Fig. 6a.
The patient from whom T11 was derived had hypopharyngeal carcinoma (stage T3N0M0). The patient was treated with chemoradiotherapy (2 sessions of cisplatin and radiotherapy dose of 40 Gy) prior to surgery because of the strong personal willingness of the laryngeal preservation approach. The tumor partially responded to the treatment as assessed from imaging examination (Fig. 6b), and a laryngeal preservation surgery was conducted later.
Fortunately, a pathologic complete response (pCR) was achieved after postoperative pathologic examination of the resected tissue (Fig. 6b). This indicated that T11 was sensitive to the treatment of cisplatin and radiotherapy in vivo. Indeed, CRC line T11 was relatively sensitive to cisplatin from our in vitro drug screening assay. Similarly, patient T13, who exhibited the highest sensitivity to cisplatin, was diagnosed with hypopharyngeal carcinoma (stage T4N1M0) and was also treated with 2 sessions of cisplatin and a radiotherapy dose of 40 Gy prior to surgery. After the removal of the residual primary tumor, the pathologic examination demonstrated a pCR (Fig. 6c).
The patient of T2 presented with laryngeal squamous cell carcinoma (stage T4N2M0) was treated with preoperative radiotherapy. However, the patient showed progressive disease status shortly after radiotherapy of 40 Gy, and a total laryngectomy was performed. Postoperative radiotherapy of a dose of 26 Gy was administered because of lymph node metastasis. Unfortunately, the patient succumbed to locoregional recurrence 4 months later. Patient T7 was diagnosed with hypopharyngeal carcinoma (stage T4N2M0). Progressive disease status was observed after preoperative radiotherapy of 40 Gy, hence partial hypopharyngectomy and total laryngectomy were performed on the patient. Postoperative radiotherapy of a dose of 26 Gy was given because of positive margins. The patient was last followed up 1 year after the end of treatment. There were no signs of recurrence to this point. Further, follow-up would be conducted to observe any remission. The in vitro sensitivity to radiation, colony formation assay, and cell proliferation assay were performed on the CRCs. After exposing to 4 Gy radiation, T1 exhibited a signi cant inhibition in colony formation e ciency, while T2 did not (Fig. 6d). Besides, CRCs T1, T2, T7, and T11 were treated with 4 Gy radiation. After 3 days' culture, their relative cell proliferation was detected. The result indicated that relative cell proliferation of T2 and T7 was signi cantly higher than that of T1 and T11 after radiation (Fig. 6e).
Based on these ndings, T2 and T7 presented resistance to radiotherapy. Consistently, the clinical history showed concordance with the in vitro ndings.

Discussion
Management of LHSCC is highly complicated and mandates multidisciplinary care (Ringash 2015). Precision oncology aims to identify and target tumor-speci c aberrations with effective therapeutic strategies for individual cancer patients. Currently, in precision oncology, the recommendation of molecularly targeted drugs is primarily based on the genomic pro le of a drug-target gene as a therapeutic indicator (Bailey et al. 2018). The drive toward precision oncology has signi cantly increased attention in adapting in vitro tumor models for patient-speci c therapies, drug response assessment, and clinical management. Various in vitro models for LHSCC were developed to guide for patient-speci c therapies in the past decades, including patient-derived primary tumor cells, 3D culture spheres, patient-derived xenotransplantation, and tumor organoids (Dohmen et  ampli cation in the CR culture system, we were able to obtain 10 6 cells from each individual endoscopic biopsy or tumor sample within 2 weeks. The cells could be cultured for more than 40 days and 10 passages, with a stable proliferation rate during the extended period. Moreover, the proliferation of cells was not affected by repeated cycles of cryopreservation and resuscitation. Collectively, the CR system contented the criteria for LHSCC continuous cell culture from tumor samples. Besides, su cient cell numbers could be acquired for further assays such as drug screening, sequencing, or xenografts. As a milestone for the CR technique, Liu et al. successfully generated continuous cell cultures from tumor samples in a patient with recurrent respiratory papillomatosis and identi ed Vorinostat as a therapeutic agent using in vitro chemosensitivity testing (Yuan et al. 2012). CR technique played a crucial role in guiding clinical administration, though only 1 case was involved. In the present study, the drug and radiation sensitivity of some established CRC lines was tested. Various killing effects were observed. These ndings motivated to explore the clinical responses using CR techniques. Most of the patients included in this study received surgical resection of the tumor as primary treatment, making it impossible to assess their chemoradiotherapy response outcomes by RECIST criterion (Eisenhauer et al. 2009). Only 4 cases treated with preoperative chemo/radiotherapy were selected to validate the relationship between in vitro and clinical responses. And closely matched responses were observed. However, more cases are required to validate this relationship robustly. These results implied that CR cultured LHSCC could predict patients' clinical responses to chemotherapy or radiotherapy and might serve as an excellent preclinical model for precision oncology.
Based on the whole-exome sequencing of the 3 pairs of CRC lines, resistance to cisplatin-based chemotherapy was also positively correlated with TP53 mutation burden, consistent with a previous study (Bergamaschi et al. 2003).
Epidermal growth factor receptor (EGFR) was overexpressed in 50%~90% of the tumors (Bossi et al. 2016), and about 15% of patients carry gene ampli cation of EGFR (Network 2015). None of the sequenced cells carried EGFR mutation in this study, and only one of them (T2) had EGFR gene ampli cation. Cetuximab, a monoclonal antibody targeting EGFR, was used in drug screening. However, the cetuximab sensitivity of the CRC lines could not be re ected by their EGFR expression levels. Although con icting, the result was in agreement with previous studies (Bossi et al. 2016).
After being heterotransplanted subcutaneously in nude mice, the primary tumor CRC lines could form tumors successfully, while the normal paired ones could not. This demonstrated the malignant potential of the tumor cells.
These CR cell-derived xenografts retained the primary tumors' histopathologic characteristics and could be used for personalized treatment just as patient-derived xenografts (PDX) did (Karamboulas et al. 2018). Moreover, CRC lines could be expanded, cryopreserved, and resuscitated in vitro and are repeatedly used for xenografts, which made it much more exible and convenient than the PDX model. Patient-derived organoids (PDOs), recapitulating the primary tissues' genetic and molecular characteristics, have been applied to conduct high-throughput drug screening and predict the treatment responses of HNSCC (Driehuis et al. 2019). We could not ignore its disadvantage as aiming at potential clinical application. Firstly, the success rate of LHSCC organoids culture was relatively low, according to the previously published literature (Driehuis et al. 2019). We also tried to culture tumor organoids following previously described methods, but a dissatisfying success rate of 12.5% (2/16) was achieved. Secondly, a clinically signi cant time window was not available as required for personalized treatment decision-making, which is generally less than 2 to 3 weeks for preoperative and postoperative chemotherapy. It is not su cient time for tumor organoids to proliferate to optimal cell number for drug testing. Thirdly, the organoid culture system relied on multiple expensive growth factors in the culture medium and extracellular matrix substitutes. The high expense hampered its extensive application for a less supported institute. Finally, organoid associated processes, unlike classical monolayer cell culture techniques, were speci c and complicated to get started. Furthermore, to demonstrate the stemness of CRC lines, we dissociated the CR primary cells embedded in Matrigel and cultured as organoids as previously described (Driehuis et al. 2019).
Sphere-shaped organoids could be cultured from the primary cells. A polarized growth was observed as the basal cells strati ed the outer layer of the organoids while the differentiated and keratinized cells located inside the organoids. Furthermore, these CR derived organoids could also be used to test drug sensitivity.
In summary, we established LHSCC primary cell lines using the CR technique. The CR cell lines had retained histological and molecular characteristics and heterogeneity of the parental LHSCC. CR cell lines could be transformed into xenografts and organoids, serving as versatile in vitro models. Collectively, patient-derived cell model system using CR technology could be promisingly utilized in clinical decision-making and help identify personalized therapies for LHSCC. Con icts of interest The authors declare that they have no con ict of interest.

Declarations
Availability of data and material The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Code availability Not available. Consent to participate Written informed consent to participate in the study was obtained from the patients and/or their authorized representatives.
Consent for publication The authors a rm that human research participants provided informed consent for publication of the images in Figure 6b and