Derivation of pancreatic acinar cell carcinoma cell line HS‐1 as a patient‐derived tumor organoid

Abstract

Acinar cell carcinoma (ACC) of the pancreas is a malignant tumor of the exocrine cell lineage with a poor prognosis. Due to its rare incidence and technical difficulties, few authentic human cell lines are currently available, hampering detailed investigations of ACC. Therefore, we applied the organoid culture technique to various types of specimens, such as bile, biopsy, and resected tumor, obtained from a single ACC patient. Despite the initial propagation, none of these organoids achieved long‐term proliferation or tolerated cryopreservation, confirming the challenging nature of establishing ACC cell lines. Nevertheless, the biopsy‐derived early passage organoid developed subcutaneous tumors in immunodeficient mice. The xenograft tumor histologically resembled the original tumor and gave rise to infinitely propagating organoids with solid features and high levels of trypsin secretion. Moreover, the organoid stained positive for carboxylic ester hydrolase, a specific ACC marker, but negative for the duct cell marker CD133 and the endocrine lineage marker synaptophysin. Hence, we concluded the derivation of a novel ACC cell line of the pure exocrine lineage, designated HS‐1. Genomic analysis revealed extensive copy number alterations and mutations in EP400 in the original tumor, which were enriched in primary organoids. HS‐1 displayed homozygous deletion of CDKN2A, which might underlie xenograft formation from organoids. Although resistant to standard cytotoxic agents, the cell line was highly sensitive to the proteasome inhibitor bortezomib, as revealed by an in vitro drug screen and in vivo validation. In summary, we document a novel ACC cell line, which could be useful for ACC studies in the future.

By combining the organoid culture and xenograft formation, a novel cell line HS‐1 was eventually established from a biopsy of acinar cell carcinoma (ACC), a rare subtype of pancreatic cancer, for the first time as an organoid. HS‐1 is positive for carboxylic ester hydrolase (CEH), a highly specific acinar cell marker, and secretes trypsin abundantly, but is negative for the duct cell marker CD133, thereby retaining the features of acinar cells and the original tumor. HS‐1 harbors a missense mutation in EP400 and a 30‐Mb deletion encompassing CDKN2A. Drug screening identified the proteasome inhibitor bortezomib as a potential ACC therapeutic. This cell line will be useful for further ACC studies.

Abbreviations

ACC

acinar cell carcinoma

CEH

carboxylic ester hydrolase

CNV

copy number variation

CT

computed tomography

DGF

defined growth factors

ERCP

endoscopic retrograde cholangiopancreatography

EUS‐FNA

endoscopic ultrasound‐guided fine‐needle aspiration

FFPE

formalin fixed and paraffin embedded

GEM

genetically engineered mice

MBOC

Matrigel bilayer organoid culture

NGS

next‐generation sequencing

PDAC

pancreatic ductal adenocarcinoma

PDO

patient‐derived organoid

PDX

patient‐derived xenograft

SNV

single nucleotide variations

VAF

variant allele frequency

XDO

xenograft‐derived organoid

1. INTRODUCTION

Pancreatic carcinoma is a neoplasm arising from the epithelial cells of exocrine lineages. PDAC and ACC account for 99% and 1% of pancreatic carcinomas, respectively. ¹ , ² ACC shares common features with acinar cells in terms of histological properties and secretion of digestive enzymes. Elevated levels of serum lipase can cause subcutaneous panniculitis, a manifestation of hyperlipase syndrome, which is pathognomonic to ACC. ³ Approximately 30% of the cases also display endocrine property ⁴ , ⁵ suggesting a link between ACC and neuroendocrine tumors. In approximately 50% of the cases, metastatic diseases are already evident at the initial diagnosis, ³ , ⁵ , ⁶ , ⁷ and in approximately 75% of the cases, relapse after surgery is observed. ³ , ⁸ Although chemotherapy based on gemcitabine and fluorouracil is often selected, its efficacy is limited. In addition, no effective chemotherapeutic regimen or molecular‐targeted therapy has been developed. ⁷ , ⁹ Consequently, the prognosis of patients with ACC remains poor, with an overall survival rate of approximately 20 months. ³ , ⁴ , ⁵ , ⁸

The mutation spectrum in human ACC is distinct from that in PDAC. Although the mutation rate in PDAC is approximately 90% for KRAS and approximately 50% for CDKN2A, SMAD4, and TP53, ¹⁰ , ¹¹ none of these mutations are detected in more than 20% of ACC cases. Mutations in APC and microsatellite instability are detected in approximately 25%. ¹² , ¹³ , ¹⁴ , ¹⁵ In the whole exome analysis of seven ACC cases, three and four mutations were found in BRCA2 and FAT, respectively. ¹⁶ Mutations in DNA repair‐related genes and BRAF‐ or RAF1‐fusion genes are recurrently detected in a mutually exclusive manner, at 45% and 23%, respectively. ¹⁷ With a higher grade of chromosomal aberration than PDAC, ACC exhibits distinct features in the genome‐wide copy number profile. ¹³ , ¹⁴

In vivo models of ACC have been developed in rodents. Upon chemical treatment, the rat is predisposed to ACC but not PDAC. ¹⁸ , ¹⁹ Among many rat cell lines, the AR42J cell has been most widely used, presumably because of its dual properties in both exocrine and endocrine lineages. ²⁰ It allows functional analysis as a substitute for normal islet cells and acinar cells, ²¹ which are still difficult to maintain in culture. Genetically engineered mice represent an alternative in vivo models, having pancreas‐specific overexpression of SV40 ²² and ablation of Tsc1, either alone ²³ or in combination with Trp53 deletion. ²⁴ In establishing cell lines, two‐dimensional (2D) culture has been usually adopted, but it was associated with a low success rate, potentially setting high pressure for clonal selection. Additionally, acinar cell traits were not strictly verified in previously described ACC cell lines, ²⁵ , ²⁶ , ²⁷ questioning their authenticity as ACC. Together with its low incidence, no reliable patient model has been available for ACC, hampering functional analysis and identification of effective therapeutics.

Organoid culture is an emerging technology that enables infinite propagation of tissue stem cells. ²⁸ , ²⁹ Due to its high success rate and physiological features, it has been applied to model many biological processes, including multistep carcinogenesis of the intestine, ³⁰ lung, ³¹ hepatobiliary system, ³² , ³³ pancreas, ³⁴ and female reproductive system. ³⁵ , ³⁶ By lentiviral gene transduction or chemical treatment ³⁷ followed by subcutaneous implantation in immunodeficient mice, tumors readily develop. More recently, PDOs have drawn much attention because they retain the morphological features, genetic abnormalities, and heterogeneity of the primary tumor. ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ Cancers from diverse organs can derive organoids, providing preclinical models for drug discovery and resources for precision medicine. However, no ACC PDOs have been documented to date.

In this study, we aimed to establish a novel ACC cell line. We obtained ACC organoids of the pure exocrine lineage that could propagate infinitely while retaining the features of the original tumor. Using this cell line, we conducted a drug screen, genome analysis, and functional analysis, thereby gaining novel insights into the biological features of ACC.

2. MATERIALS AND METHODS

Detailed information is described in Supplementary Materials and Methods.

2.1. Patient information

The patient was a 48‐year‐old woman with ACC who underwent surgery at the Chiba Cancer Center Hospital.

2.2. Organoid culture of clinical specimens

Bile and biopsy samples were collected endoscopically. The tumor was surgically resected. These specimens were subjected to MBOC as previously described. ³⁴ Briefly, enzymatically dissociated cells were embedded in a bilayer of Matrigel (BD Biosciences, Franklin Lakes, NJ) by two‐step procedures ⁴² and maintained in advanced DMEM/F12 (Thermo Fisher Scientific) supplemented with DGF including 50 ng/ml human epidermal growth factor (EGF) (Peprotech), 250 ng/mL R‐spondin1 (R&D), 100 ng/ml Noggin (Peprotech), 10 μM Y27632 (Wako, Osaka, Japan), and 1 μM Jagged‐1 (AnaSpec).

2.3. Tumorigenicity assay and in vivo drug treatment

Animal studies were performed with the approval of the Chiba Cancer Center for Ethics in Animal Experimentation. Organoids were injected into nude mice, as previously described. ³⁰ Briefly, organoids corresponding to 5 × 10⁵ cells were resuspended in 100 μl of 50% Matrigel and subcutaneously inoculated into nude mice. Bortezomib in saline was administered by i.p. injection, at a dose of 5 μg per mouse twice a week.

2.4. Pathological analysis

Organoids, xenografts, and resected tumors were FFPE tissues. Histopathological analysis was conducted as previously described. ³⁰

2.5. Western blotting

Organoids and culture supernatant were lysed using radioimmunoprecipitation assay (RIPA) buffer. In total, 5–10 μg of protein were analyzed by semi‐dry immunoblotting as previously described. ³⁰

2.6. RT‐PCR

Total RNA was extracted from organoids, followed by cDNA synthesis and PCR to detect fusion gene transcripts, as previously described. ³⁰

2.7. Next‐generation sequencing analysis

Genomic DNA was extracted from FFPE specimens and organoids. The Ion AmpliSeq Comprehensive Cancer Panel (Thermo Fisher Scientific) was used to analyze 409 cancer‐related genes as previously described, ⁴³ with slight modifications.

2.8. Genomic PCR and Sanger sequencing

Target regions for the EP400 and CDKN2A genes were amplified using genomic PCR and subjected to direct sequencing.

2.9. Array‐based comparative genomic hybridization (aCGH) analysis

Hybridization was performed using a SurePrint G3 mouse comparative genomic hybridization (CGH) microarray 4 × 180K (G4826A, Agilent). Scanning and image analyses were performed using the Agilent Feature Extraction ver. 11.0 (Agilent) and a SureScan Microarray Scanner (G4900DA, Agilent).

2.10. Drug sensitivity assay and chemical screening

After drug treatment of the dissociated cells, cell viability was evaluated using CellTiter‐Glo3D (Promega). Alternatively, the images were captured with a scanner Cell3iMager Duos (Screen Holdings) to estimate the number of viable cells.

2.11. Lentiviral infection

Lentiviral particles were generated as previously described. ³⁰ The lentivirus expressing WT or mutant human CD133 ⁴⁴ was introduced into organoids as previously described. ⁴⁵

2.12. Statistical analysis

Data are represented as mean ± SD of three independent experiments. The statistical significance of the difference between mean values was tested using an unpaired two‐tailed Student's t‐test. Statistical significance was set at p < 0.05.

3. RESULTS

3.1. A solid tumor in the pancreatic head was diagnosed as ACC of the pure exocrine type

A female patient presenting with abdominal pain visited the hospital. Contrast‐enhanced CT detected a 30‐mm poorly enhanced lesion in the pancreatic head, and ERCP revealed bile duct stenosis caused by tumor progression (Figure 1A). A biopsy sample obtained by EUS‐FNA exhibited a solid structure with the occasional lumen and atypical cells with amphophilic cytoplasm (Figure 1B). Immunohistochemical analysis with the BCL10 mAb clone 331.3, which also detects CEH, a highly specific acinar cell marker, ⁴⁶ showed diffuse staining in the cytoplasm. In addition, the neuroendocrine marker synaptophysin was negative. Hence, the tumor was diagnosed as ACC of the pure exocrine lineage. The resected tumor was an encapsulated mass displaying an homogeneous white to pink color (Figure 1C). The tumor was histologically similar to the biopsy, although CEH expression displayed a more heterogeneous pattern (Figure 1D).

FIGURE 1.

Diagnosis of pancreatic acinar cell carcinoma (ACC). (A) Diagnostic images of an ACC case. (Left) Enhanced computed tomography (CT). The tumor is a low‐density mass in the pancreatic head (arrowheads). (Middle) Endoscopic retrograde cholangiopancreatography (ERCP). The bile duct stenosis (arrowheads) is caused by the tumor. (Right) Endoscopic ultrasound‐guided fine‐needle aspiration (EUS‐FNA). The tumor appears to be hypoechoic (arrowheads). A biopsy specimen is aspirated with the needle (arrow). (B) Histological analysis of the biopsy specimen. (Left) Hematoxylin and eosin (H&E) staining. (Middle) Immunostaining of carboxyl ester hydrolase (CEH) with anti‐BCL10 mAb. (Right) Immunostaining for synaptophysin. Scale bars, 50 μm. (C) The resected tumor. An encapsulated mass (arrowheads) is in the middle of the resected tissue. (D) Histological analysis of the resected tumor. (Left) H&E staining. (Right) Immunostaining of CEH with anti‐BCL10 mAb. Scale bars, 100 μm.

3.2. Multiple types of samples from the ACC patient initially gave rise to organoids

During diagnosis and surgery, we obtained specimens from three different sources; debris in the bile, a biopsy fragment, and a resected tumor. We conducted an MBOC ⁴² (Figure 2A) for each sample and obtained PDOs. Bile‐derived organoids (hereafter designated as bile‐PDO) showed a hollow structure, whereas biopsy‐ and surgery‐PDO exhibited more dense features (Figure 2B). To verify the presence of tumor‐initiating cells, bile‐ and biopsy‐PDOs from early passages were subcutaneously inoculated into nude mice. Only biopsy‐PDO formed a solid xenograft tumor (Figure 2C), which resembled the biopsy specimen in terms of histological features and abundant CEH expression (Figure 2D). Although primary culture and several subsequent passages were feasible, none of the PDOs from three different sources achieved long‐term propagation or tolerated cryopreservation, thereby declining the establishment of a cell line.

FIGURE 2.

Propagation of patient‐derived organoids (PDOs) and development of a xenograft. (A) Schematic view of a Matrigel bilayer organoid culture (MBOC). (B) Phase contrast images of PDOs from different sources. These PDOs are designated as bile‐, biopsy‐, and surgery‐PDO, respectively. Scale bars, 200 μm. (C) Xenograft development. The tumor originates from the biopsy‐PDO. Scale bar, 5 mm. (D) Histological analysis of the xenograft. (Left) H&E staining. (Right) Immunostaining of CEH with anti‐BCL10 mAb. Scale bar, 50 μm. (E) Phase contrast images of the xenograft‐derived organoid (XDO). (F) Histological analysis of the XDO. (Left) H&E staining. (Right) immunostaining of CEH with anti‐BCL10 mAb. Insets show a magnified image of representative organoids. Scale bars, 50 μm. (G) Western blotting of organoids. Note that anti‐BCL10 mAb (clone 331.3) detects both BCL10 (33 kDa) and CEH (105 kDa). α‐Tubulin serves as a loading control. NS, non‐specific bands. An image by long exposure is also shown (left). (H) Schematic summary of establishment of PDOs and the XDO (HS‐1). Color intensity of the arrows reflects proliferation rate. NT, not tested for inoculation in nude mice.

3.3. A novel ACC cell line was established from the xenograft originating from the biopsy‐PDO

Xenograft‐derived organoids robustly proliferated and exhibited solid features resembling biopsy‐ and surgery‐PDOs (Figure 2E). Histological examination revealed amphophilic cytoplasm, mostly solid structures with an occasional lumen formation, and prominent CEH expression (Figure 2F). As acinar and ductal cells were marked by CEH and CD133, ⁴⁶ , ⁴⁷ , ⁴⁸ respectively, we performed immunoblotting to confirm the cellular composition. Using the BCL10 mAb, which detects bands at 33 kDa and 105 kDa, corresponding to BCL10 and CEH, ⁴⁹ respectively, the CEH band was weakly positive in biopsy‐ and surgery‐PDOs, and prominent in the XDO, but negative in the bile‐PDO (Figure 2G). Conversely, CD133 expression was intense in bile‐PDO and modest in biopsy‐PDO, but undetectable in surgery‐PDO and XDO. Trypsin was detected in all samples except the bile‐PDO (Figure 2G). These observations suggested that CEH^high ACC cells in biopsy‐PDO might be purified through xenograft formation and organoid culture, and that surgery‐PDO and XDO exclusively comprised ACC cells, while bile‐PDO contained no ACC cells. The XDO tolerated cryopreservation and propagated for more than 9 months over many passages. Hence, we concluded the establishment of an ACC cell line, which we designated HS‐1 (Figure 2H).

3.4. HS‐1 robustly proliferated in 2D‐ or 3D‐culture condition with defined growth factors

To explore the utility of HS‐1, we investigated whether it could propagate in 2D conditions. Although the standard serum‐containing medium did not support cell proliferation, HS‐1 robustly propagated in the organoid culture medium supplemented with DGF (Figure 3A), with a doubling time of 1.7 and 2.4 days in 3D and 2D conditions, respectively (Figure 3B). Cells proliferated in 2D‐formed cell aggregates that were only weakly attached to the dish (Figure 3C). These aggregates often partially detached from the dish but remained largely viable (Figure 3D). The predisposition of HS‐1 to a semi‐suspension state (Figure 3E) prompted us to examine the expression level of trypsin, which is often secreted by ACC and widely used to detach cells from culture dishes. In HS‐1, high levels of trypsin were detected in the supernatant, whereas CEH expression was positive only in the whole lysate (Figure 3F). Moreover, trypsin in the supernatant increased over time (Figure 3G). These results suggested active trypsin secretion rather than cell rupture during organoid culture.

FIGURE 3.

Propagation of HS‐1 from the biopsy‐PDO. (A) Effect of culture conditions. Phase contrast images of HS‐1 cultured under various conditions. Organoid culture medium is supplemented with 10% FBS or defined growth factors (DGF) without serum. Scale bars, 250 μm. (B) Cell proliferation of HS‐1 in culture. The cell numbers were counted using the trypan blue dye exclusion method. (C) Phase contrast images of HS‐1 in the medium with DGF in the absence of Matrigel. Small aggregates are distributed on the surface of the well, while large aggregates expand upward (arrow). Scale bars, 50 μm. (D) Dual staining of cell aggregates. Semi‐floating and floating cells are stained. Dead cells (red) and viable cells (green) are differentially stained. (E) Schematic illustration of the HS‐1 propagation. Cells grew in a spherical shape in Matrigel, whereas the aggregate grew larger and eventually detached from the dish (arrow) in the absence of Matrigel. (F) Western blotting of the lysate and supernatant of HS‐1. Ponceau S staining of the membrane serves as a loading control. (G) Western blotting of the culture supernatant. Time‐lapse samples are analyzed for secreted trypsin.

3.5. Mutations in EP400 , CDKN2A , and ATRX were enriched in surgery‐PDO

To better characterize PDOs, we next conducted an NGS analysis of 409 cancer genes. Seven genes had mutations with VAF of more than 10% in at least one of the samples (Figure 4A). In the bile‐PDO, no somatic mutations were detected, consistent with its non‐malignant nature (Figure 2G). A missense mutation in EP400 (p.Pro882Ser) was detected with a VAF of 22% in the resected tumor, which increased to 99% in the surgery‐PDO and HS‐1. This finding strongly suggests that EP400‐mutant cells with LOH were positively selected during organoid culture, in line with the tumor‐suppressive roles of EP400. ⁵⁰ A nine‐base deletion in CDKN2A (p.Ala40_Asn42del) was detected with a VAF of 4% in the resected tumor, which increased to 30% in surgery‐PDO, but disappeared in HS‐1. A single‐base deletion in ATRX (p.Asn564fs) and a missense mutation in CSMD3 (p.D988H) were detected at similar VAF values between surgery‐PDO and HS‐1, suggesting their origins in the resected tumor and limited advantage in vitro and in vivo. Sanger sequencing validated the homozygous missense mutation in EP400 and the hemizygous nine‐base deletion in CDKN2A in surgery‐PDO (Figure 4B). We did not investigate the other mutations further because they were detected in only one sample.

FIGURE 4.

Genomic analysis of the ACC tumor and derived organoids. (A) Mutations in the resected ACC tumor and derived organoids. Genes with a variant allele frequency (VAF) >10% estimated using next‐generation sequencing (NGS) analysis are shown. Loss of a 9‐bp deletion in CDKN2A (asterisk) in XDO (HS‐1) was due to deletion of both alleles (see D and E). (B) Validation of somatic mutations in surgery‐PDO. Sanger sequencing verified a homozygous point mutation in EP400 (arrowhead) and a hemizygous 9‐bp deletion in CDKN2A (arrow). (C) Genome‐wide distribution of loss of heterozygosity (LOH). LOH is estimated based on the change in VAF and collectively shown as the line. Gray bars indicate areas with no corresponding data in NGS. (D) Genome‐wide copy number variation (CNV) by array CGH analysis. The regions in which the XDO (HS‐1) acquired copy number gain and loss compared with surgery‐PDO are indicated by (*) and (#), respectively. The 33‐Mb region in the chr. 9p is homozygously deleted in the XDO (HS‐1). Note that chr. 9p in surgery‐PDO seems to undergo slight copy number loss (arrowhead). (E) Genomic PCR of the chr. 9p region. CDKN2A and DNAJA1 are amplified, from inside and outside of the deleted region, respectively. Note that CDKN2A amplicon is faintly visible for XDO (HS‐1). (F) Schematic summary of the genome analysis of tumor and organoids.

3.6. The chromosome 9p region harboring CDKN2A was homozygously deleted in HS‐1

Changes in the VAF of single nucleotide polymorphisms (SNPs) can provide information on the LOH status. For example, if the VAF at any SNP was either 0% or 100% in cancer tissue, but approximately 50% in normal tissue, then LOH in cancer was inferred (Figure S1A). Using this method, LOH in the surgery‐PDO and HS‐1 was estimated to be distributed genome‐wide (Figure 4C). To evaluate genome instability with more accuracy, we conducted a CGH analysis. Both organoids showed extensive CNV in a similar pattern, with some additional alterations in HS‐1, such as copy number gain in chr. 3q, 8q, 12q, and 19, and loss in 5q, 9p and 20q (Figure 4D). The sharp decline of the copy number indicative of homozygous deletion in 9p emerged in HS‐1, harboring 150 genes including CDKN2A within the 33‐Mb region. Consistently, in HS‐1, the PCR amplicon for CDKN2A was extremely faint (Figure 4E), and NGS resulted in no read for CDKN2A (p.Ala40_Asn42del), but a few reads for the WT allele (Figure S2). These observations might well explain why the VAF of CDKN2A (p.Ala40_Asn42del) fell to 0% in HS‐1. Given that the copy number was slightly less than one in chr. 9p (Figure 4D), a fraction of tumor cells must be nullizygous in CDKN2A. Copy number gain was observed in chr. 2, 5p, 7, 8, 18–20, and X (Figure 4D) without LOH (Figure 4C) in both surgery‐PDO and HS‐1, suggestive of their origin in the resected tumor (Figure 4F). Transcripts for SND1‐BRAF, a recurrent fusion gene in ACC, ¹⁷ were not detected by RT‐PCR of HS‐1 (Figure S3).

3.7. An in vitro drug screen identified bortezomib as a therapeutic candidate for ACC

We next performed drug sensitivity assays with HS‐1. For convenience, we adopted a monolayer organoid culture on Matrigel (Figure 5A). Doxorubicin, cisplatin, and 5‐fluorouracil did not show high cytotoxicity, whereas paclitaxel and gemcitabine reduced cell viability by approximately 80% and 60%, respectively (Figure 5B). Dead and viable cells could be automatically distinguished using imaging software, if the parameters for a filter of the scanner were adjusted (Figure 5C). Moreover, the estimation of the area occupied by the viable cells correlated well with the luminescence intensity obtained by the ATP‐based assay for the aforementioned therapeutics (Figure 5D). To explore effective drugs for ACC, we conducted a drug screen with 364 known chemicals related to the signaling pathways relevant to cancer. At an initial dose of 1 μM, more than half of the reagents were highly effective (Figure S4A, Table S1), suggesting an excessively high drug concentration. Nonetheless, the area of viable cells and the fluorescence intensity using the ATP assay were well correlated (Figure S4B), prompting us to conduct an imaging‐based assay to streamline the procedure. Even at a dose of 10 nM, several agents showed high efficacy (Figure 5E). Among the top‐scoring reagents (Table S2), we selected six candidates for one‐by‐one validation. Although rapamycin and lapatinib had moderate effects, bafilomycin A1 and bortezomib had high cytotoxicity (Figure 5F). As bortezomib was the most potent, we examined its efficacy in vivo. Upon i.p. administration to nude mice with inoculated HS‐1, the growth of xenografts was significantly suppressed (Figure 5G)

FIGURE 5.

Drug sensitivity assay with HS‐1 in vitro and in vivo. (A) Schematic presentation of the drug screening. Organoids grown on Matrigel were treated. (B) Sensitivity of HS‐1 to conventional chemotherapeutic drugs. Error bars depict SD (n = 3). (C) Imaging‐based evaluation of cell viability. Viable cells (green) and dead cells (red) after pneumothorax (PTX) treatment are computationally detected. Scale bar, 500 μm. (D) Correlation between the area of viable organoids and luminescence intensity. Data obtained in the ATP‐based viability assay in (B) were compared with the area of viable cells estimated by computational detection. (E) Imaging‐based drug screening. In total, 364 reagents were administered at concentration of 10 nM. The bars corresponding to bafilomycin and bortezomib are indicated by arrows. (F) Imaging‐based validation of hits in the drug screening. Error bars depict SD (n = 3). (G) Antitumorigenic effect of bortezomib in vivo. After inoculation of HS‐1 to nude mice, 5 μg of bortezomib in saline was administered twice a week (arrows). Error bars depict SD (n = 3 each). *p < 0.05.

3.8. CD133 differentially affected drug sensitivity of the ACC cell line

CD133, a marker of normal ductal cells, ⁴⁷ , ⁴⁸ is also known as a cancer stem cell marker that mediates stem cell‐like features through AKT phosphorylation. ⁴⁴ , ⁵¹ Given that the HS‐1 lacks CD133 expression (Figure 2G), we reasoned that it would serve as an ideal platform for evaluating the effects of CD133, particularly on the ACC stem cell‐like features such as tumorigenicity and drug resistance, and trans‐differentiation potential to PDAC. We first verified the high efficiency of lentiviral gene transduction (Figure 6A), and transduced HS‐1 with CD133^WT, CD133^EE, and CD133^FF, in which two tyrosine residues were substituted to a normal, constitutively active, and inactive form in phosphorylating AKT, respectively. ⁴⁴ Despite the robust expression of these constructs (Figure 6B), no obvious morphological changes were observed (Figure 6C). Upon inoculation in nude mice, HS‐1 with CD133^EE tended to develop larger tumors (Figure 6D), but not significantly in terms of tumor size (Figure 6E) or histological features (Figure 6F). We then investigated whether CD133^EE could confer drug resistance. Whereas no significant change was observed in the sensitivity to bortezomib, transduced HS‐1 became resistant to gemcitabine and was sensitized to rapamycin (Figure 6G), in line with CD133‐mediated augmentation of cancer stem cell‐like features and AKT activation, respectively.

FIGURE 6.

Effects of exogenous CD133 expression on HS‐1. (A) Efficient gene transduction in HS‐1 organoids. Images obtained 2 days after infection of the lentivirus encoding GFP are shown. (Left) a bright field. (Right) a fluorescent field. Scale bar, 500 μm. (B) Western blotting. Transduced HS‐1 before and after inoculation are analyzed. Empty backbone vector (pCDH) and the wild type (CD133^WT), constitutively active (CD133^EE) and inactive (CD133^FF) forms of human CD133 are introduced. β‐Actin serves as a loading control. (C) Phase contrast images of transduced HS‐1. Scale bars, 300 μm. (D) Subcutaneous tumor development. Upper panel, nude mice inoculated with transduced HS‐1. Lower panels, isolated xenograft tumors. Results from three independent experiments are shown. Two tiny nodules labeled using asterisks are the results of a massive leak upon injection, which was excluded from the analyses. Scale bars, 1 cm. (E) Tumor volume of each xenograft. Error bars depict SD (n = 3). (F) Histological analysis of the xenografts. H&E staining is shown. Scale bars, 100 μm. (G) Effect of CD133^EE in XDOs on drug sensitivity. Error bars depict SD (n = 3). *p < 0.01, **p < 0.05.

4. DISCUSSION

The establishment of ACC cell lines has been challenging due to the lack of an optimized protocol, let alone its rare incidence. This was also the case in this study: initial attempts to propagate PDOs from the three different sources all resulted in a gradual decline within several passages. Nevertheless, we were eventually able to establish a novel cell line, HS‐1, by xenograft tumor formation from biopsy‐PDO followed by the organoid culture. HS‐1 still failed to survive under standard 2D culture conditions with FBS, in agreement with the few successful cases in establishing cell lines. ²⁵ , ²⁶ , ²⁷ Regardless of 2D or 3D conditions, it robustly propagated in the organoid culture medium, highlighting the relevance of the DGF, at least for HS‐1. Further optimization of the protocol will be required for stable culturing of primary ACC.

By integrated genome analysis, we identified the inactivation of CDKN2A toward the establishment of HS‐1. The target sequencing identified CDKN2A (p.Ala40_Asn42del) in the resected tumor, which was more enriched in surgery‐PDO suggestive of its advantage, but unexpectedly lost in HS‐1. Array CGH analysis of HS‐1 uncovered a sharp decline in copy number in chr. 9p indicative of homozygous deletion, which could definitely account for the loss of CDKN2A (p.Ala40_Asn42del). In addition, a slight decrease in copy number of fewer than one copy and LOH in chr. 9p strongly suggested the presence of CDKN2A‐null cells in surgery‐PDO. This observation pointed toward the notion that the resected tumor contained a fraction of CDKN2A‐null cells, rather than homozygous deletion occurring independently in surgery‐PDO and biopsy‐PDO. Because we lost viable biopsy‐PDO, it was no longer feasible to retrospectively investigate when and how homozygous deletion occurred toward the establishment of HS‐1. However, we infer that this might be the likely scenario. Nonetheless, the fact that CDKN2A was targeted by two independent mechanisms in the PDOs underscores its critical role in establishing cell lines as a negative regulator of the senescence program. ⁵² It is also possible that bi‐allelic inactivation of EP400, another tumor suppressor gene implicated in senescence, played critical roles by cooperating with extensive CNVs, which requires further investigation.

Consistent with the ACC origin, HS‐1 secretes high levels of trypsin. Considering its common use in scraping cells from the culture dish, the low adhesiveness of HS‐1 might be an inherent property of ACC. If trypsin is produced in excess, it should overwhelm FBS as an inhibitor of trypsin, thereby promoting the detachment of cells from the dishes. Thus, trypsin autosecretion might underlie the low success rate in establishing ACC cell lines. HS‐1 in this study was derived from the major type of ACC with a pure exocrine lineage, in sharp contrast with AR42J cells, the mixed lineage rat ACC cell line. Given that the culture of normal acinar cells remains a huge challenge, ⁵³ the human ACC cell line might also be valuable as a surrogate for normal acinar cells, particularly for in vitro studies of pancreatitis, in which digestive enzymes secreted by acinar cells play a major role. ⁵⁴

Previous studies have claimed the establishment of cell lines from ACC patients’ samples, although their validity as ACC retrospectively appears elusive, because BCL10 mAb was not available when these studies were conducted. First, the HPC‐Y0 cell, ²⁶ established along with many other PDCA cell lines, harbored a mutation in KRAS, ⁵⁵ which is unusual in ACC but highly prevalent in PDCA, pointing toward a concern of cross‐contamination. Although proliferation in a semi‐suspension state and expression of various digestive enzymes were observed, this is not pathognomonic of ACC. Second, Panc‐4 cells, which secrete lipase in the supernatant, were not characterized in detail in vitro. ²⁵ Although its acinar properties seemed verified in histological analysis when grown as xenografts, ⁵⁶ sufficient in vitro data were not provided. Third, Faraz‐ICR cells express mesenchymal markers such as vimentin and desmin, but not epithelial markers, which are unlikely to represent ACC. ²⁷ Collectively, it is unclear whether it is justified to use these cells as authentic ACC cell lines.

Using an in vitro chemical screen of HS‐1 followed by an in vivo validation, we identified bortezomib, a proteasome inhibitor, as a potential candidate for the treatment of ACC. Owing to its antitumor effects, the inhibitory actions of bortezomib on three targets are known: the NF‐κB pathway, angiogenesis, and unfolded protein response (UPR). ⁵⁷ In this study, NF‐κB pathway inhibitors did not show cytotoxicity, as demonstrated by its specific inhibitor N‐acetyl‐L‐cysteine, IKK inhibitor BMS‐345541, and IKK‐2 inhibitor VI (Table S2). Additionally, the effects on angiogenesis are not unlikely to be recapitulated in organoid culture. Consequently, UPR inhibition could be a plausible mechanism of action, which requires further investigation. As bortezomib has already been used in the treatment of multiple myeloma, its use in patients with ACC might be worth considering, although its adverse effects have prevented its approval for PDCA. ⁵⁸ , ⁵⁹ Bafilomycin A1, an inhibitor of lysosomal acidification and autophagy, was highly effective, and the mTOR pathway inhibitor rapamycin had a moderate effect, as previously documented in ACC developed in KO mice for Tsc1. ²³ These observations raise the possibility that interfering with the fine‐tuned balance between protein degradation and synthesis might be an effective therapeutic strategy for ACC.

By taking advantage of the ACC cell line that lacks CD133 expression, we investigated the functional relevance of CD133. Although acinar‐to‐ductal trans‐differentiation and augmentation of tumorigenicity were not observed, drug sensitivity was affected in two directions; resistance to cytotoxic drug gemcitabine and sensitization to mTOR inhibitor rapamycin, providing novel insights into the roles of CD133 in ACC. Conversely, the major limitation of this study is that only a single ACC cell line was established. In addition, CDKN2A loss in HS‐1 could confer drug resistance. Consequently, it remains to be seen to what extent it represents ACC. To address this issue, it is necessary to establish more ACC cell lines from primary sources through optimization of the culture conditions. Alternatively, as shown in this study, the use of a PDX as a starting material for organoid culture might help overcome the low incidence of new cases.

In conclusion, we established a novel ACC cell line from a single patient. Future efforts to increase the number of available ACC cell lines will allow us to obtain more fundamental knowledge and useful data on personalized medicine and drug discovery for this deadly disease.

AUTHOR CONTRIBUTIONS

DH, EK, YM, and HK conducted experiments. DH, RY, and MI reviewed pathological samples. EK, KN, KS, and AT provided clinical samples. OS provided the reagent. YN and YT conducted the NGS analysis. DH, EK, and YH wrote the manuscript. YM, TU, MF, TY, and YH supervised the study. YH edited the manuscript.

FUNDING INFORMATION

This work was supported in part by research grants from Chiba Prefecture and Grant‐in‐Aid for Scientific Research KAKENHI (16K09402, 17K08755, and 19K07518) of the Japan Society for the Promotion of Science (JSPS).

CONFLICT OF INTEREST

Yoshitaka Hippo is an associate editor of Cancer Science.

ETHICS STATEMENT

Approval of the research protocol by an Institutional Review Board: The research protocol was approved by an Institutional Reviewer Board (IRB number 26‐63).

Informed consent: Written informed consent was obtained from the patient.

Registry and the Registration No. of the study/trial: N/A.

Animal Studies: Animal Studies were approved by Institutional Animal Research Committee (Approved number 15‐1).

Supporting information

Appendix S1

Figure S1‐S4

Table S1

Table S2

Hoshi D, Kita E, Maru Y, et al. Derivation of pancreatic acinar cell carcinoma cell line HS‐1 as a patient‐derived tumor organoid. Cancer Sci. 2023;114:1165‐1179. doi: 10.1111/cas.15656