Pancreatic Cancer 3D Cell Line Organoids (CLOs) Maintain the Phenotypic Characteristics of Organoids and Accurately Reﬂect the Cellular Architecture and Heterogeneity In Vivo

: Pancreatic cancer is a highly lethal disease. Therapeutic resistance to chemotherapy is a major cause of treatment failure and recurrence in pancreatic cancer. Organoids derived from cancer stem cells (CSC) are promising models for the advancement of personalised therapeutic responses to inform clinical decisions. However, scaling-up of 3D organoids for high-throughput screening is time-consuming and costly. Here, we successfully developed organoid-derived cell lines (2.5D) from 3D organoids; the cells were then expanded and recapitulated back into organoids known as cell line organoids (CLOs). The 2.5D lines were cultured long term into 2D established cell lines for downstream comparison analysis. Experimental characterisation of the models revealed that the proliferation of CLOs was slightly faster than that of parental organoids. The therapeutic response to chemotherapeutic agents in 3D CLOs and organoids showed a similar responsive proﬁle. Compared to 3D CLOs and organoids, 2D cell lines tended to be less responsive to all the drugs tested. Stem cell marker expression was higher in either 3D CLOs or organoids compared to 2D cell lines. An in vivo tumorigenicity study found CLOs form tumours at a similar rate to organoids and retain enhanced CSC marker expression, indicating the plasticity of CSCs within the in vivo microenvironment.

The poor progress in PDAC therapy is largely attributed to the asymptomatic nature of the disease in its early stages. By the time of diagnosis,~80% of the patients present at late stages with locally advanced or metastatic tumours, which are considered to be surgically non-operable. In patients with resectable tumours, the highest survival rates Nicotinamide (Sigma-Aldrich, N0636), 1× B-27 supplement (Life Technologies, 17504-044, Carlsbad, CA, USA), 10.5 µM Y-27632 (Sigma-Aldrich, Y0503) in DMEM-F12 (Merck, D642, Darmstadt, Germany) supplemented with 1 mM HEPES buffer (Merck, H3375) and 1× antibiotic-antimycotic (ThermoFisher, 15240062). The 1.5 × 10 5 cells were mixed in 100 µL of extracellular matrix (ECM) (Sigma, E1270, 8-12 mg/mL) diluted to 1 mg/mL and 50 µL was plated per well in a 24-well plate. Cells were fed 500 µL of the CHFM with the Rho kinase inhibitor (ROCKi). Cells were cultured until they began to adhere to the bottom of the plate, fed with CHFM. To passage cells, 1 mL of TrypLE (ThermoFisher, 12605010) was added to each well until the cells detached, and 1 mL of DMEM High Glucose GlutaMAX (ThermoFisher, 10566-016) with 10% FBS was added to stop trypsinisation. Cells were transferred to a 6-well plate, then upscaled to a T25 cm3 flask. After this, the media were changed from CHFM to 50:50 GlutaMax DMEM and L-WRN conditioned media with 10% FBS and 1% antibiotic-antimycotic. All experiments were performed within 10 passages.

Cell Line Organoid (CLO) Recapitulation to Organoid
Scaled-up early passage 2.5D organoid-derived cell lines were trypsinized, and 5 × 10 5 cells/well were resuspended in 20-50 µL ECM per well and plated onto a 24well polyHEMA (Sigma, P3932)-coated plate overlaid with CHFM with ROCKi, with media changed every 2 days without ROCKi. All experiments were performed within 10 passages.

Cell Viability Assays
Organoids and CLOs were seeded at 5 × 10 3 in 10 µL of ECM, and cell lines were seeded at a density of 1 × 10 4 -5 × 10 4 cells/mL in a 96-well plate. The plates were placed in the Incucyte Live Cell Imaging System (S3, Sartorius, Göttingen, Germany), and brightfield images were taken every 6 h over a 10-day period. To evaluate the cytotoxicity, 72 h (organoids and CLOs) and 24 h (cell lines) post seeding, the cell models were treated with increasing concentrations of 5-FU, cisplatin, and nab-paclitaxel. After five days' incubation with the drug, the viability was measured using CellTitre-Glo (Promega, G9682, Madison, WI, USA), as per the manufacturer's instructions. The luminescence was then read with an integration of 1.0 s and gain of 100 on a plate reader (Biotek) using Gen4 software. The percentage growth was calculated relative to an untreated control. All assays were performed in triplicate.

Quantitative Reverse Transcription PCR (RT-qPCR)
Organoids and CLOs were grown as described for 10-14 days. TRI-reagent ® (Sigma, T9424) was used to lyse cells, and RNA isolation was performed using the Direct-zol RNA Miniprep Plus Kit (Zymo Research, K2072, Irvine, CA, USA). A High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, 4368814, Waltham, MA, USA) was used to synthesise cDNA from RNA. A G-Storm Thermal Cycler (Model GS1, Somerton Biotechnology Centre) was used to perform cDNA synthesis using the following program: annealing at 25 • C for 10 min, cDNA synthesis at 37 • C for 120 min, enzyme inactivation at 85 • C for 5 min.

Cell Lines
Cells were seeded in a glass-bottomed 8-well plate (Ibidi, 80827) and grown overnight. Media was removed from the cells and washed three times in a PBS 0.1% Tween 20 (Sigma, P1379) and 2% BSA (Sigma, A9418) (PTB) solution. Cells were fixed in ice-cold methanol for 5 min at 20 • C and washed in the PTB solution.

Organoids and CLOs
Organoids and CLOs were collected, fixed in formalin, and dehydrated using increasing concentrations of sucrose and then were placed in optimal cutting temperature (OCT) embedding matrix gel (Tissue-Tek, KMA-0100-00A) and placed at −80 • C overnight. Using a cryostat (Leica, CM 1900, Wetzlar, Germany), the OCT-embedded organoids were cut into 8 µm sections and placed onto Super Frost Plus slides (Thermo Fisher, 10149870, MA, USA) and stored at 4 • C until further use.
The immunofluorescence was observed using a Leica TCS SP8 STED super resolution microscope equipped with a CCD camera and 100× oil immersion objective. DAPI was excited with a 405 nm PicoQuant laser unit, and emission was captured between 387 and 474 nm. Alexa Fluor 488 was excited at 499 nm, with emission captured between 490 and 566 nm. Images were acquired in which combinations of excitation and emission wavelengths for specific dyes were applied sequentially. The fluorescence intensity was quantified using ImageJ software. The area of integrated intensity and mean gray value around the cells was measured, and the average background was subtracted from the images.

In Vivo Xenograft Study
Female SCID (CB17/lcr-Prkdcscid/lcrCrl) mice were obtained from Charles Rivers (UK), maintained in pathogen-free conditions, and fed a standard diet. Mice were kept at 25 • C with a 12 h light-dark schedule and free access to food and water. All experiments were approved by the DCU Research Ethics Committee (DCUREC/2012/202) and the Health Products Regulatory Authority (HPRA) under project authorisation AE19115/P023, and were carried out in accordance with the relevant guidelines and regulations and in compliance with the ARRIVE guidelines. PT291 organoids, CLOs, and cell lines [29] were seeded as described earlier and grown for one week prior to implantation. A cell suspension of 2 × 10 7 cells was prepared for the organoids, CLO, and 2D cell line. Cell pellets were resuspended in 1 mL of ECM. The ECM-cell suspension was plated in 400 µL ECM domes, allowed to solidify at 37 • C, and implanted subcutaneously on the flanks of the mice. Under anaesthesia (isoflurane, an O2 carrier gas), a small incision was made in the skin of the left flank of the animal. The ECM-cell dome was then drawn into a syringe and injected into the pocket under the skin, and the wound was sealed with a single staple. The animals were monitored post-surgery, and staple removal was completed within 10 days. Animals were monitored weekly for body weight and tumour development. Tumours were measured by height, width, and depth using electronic callipers. Tumour volume was calculated as outlined according to an equation: tumour volume = (height × width × depth)/1.9.
Mice were euthanized by cervical dislocation, and tumours were excised when the tumour reached a volume of >1600 mm 3 , a tumour axis >15 mm, or due to reaching humane endpoints. The tumour was then quickly divided into sections and preserved by liquid nitrogen flash freezing and formalin fixation. Flash-frozen samples were stored at −80 • C, and formalin-fixed samples were dehydrated in 50%, 70%, 90%, and 100% ethanol, followed by 100% xylene, and then embedded in paraffin. Formalin fixed paraffin-embedded (FFPE) blocks were stored at 4 • C until sectioning.
Sectioning of in vivo tumour samples was carried out using a Reichert-Jung 2030 microtome. Blocks were cut into 5 µM sections, which were floated in a water bath at 40 • C and mounted onto a SuperFrost Plus slide (ThermoFisher, 10149870) and allowed to dry at 60 • C for 2 h.

Immunohistochemistry
Immunohistochemistry (IHC) was performed on a DAKO Autostainer. Deparaffinisation and antigen retrieval were performed on slides using the DAKO PT Link and pH 6 (DAKO, S169984) or pH 9 (DAKO, S236784-2) antigen retrieval solutions. The slides were placed in the PT Link at 65 • C, heated to 95 • C, and maintained at this temperature for 20 min. The slides were then cooled to 95 • C and placed in the DAKO Autostainer. The reagents were applied as follows: Real  . Following the staining process, the slides were dehydrated in a fume hood using 70%, 90%, and 100% ethanol for two 3-min washes, followed by two 5-min xylene washes. Slides were mounted using DPX (Sigma, 44581), and the slides were allowed to air dry overnight in the fume hood.

Statistical Methods
Data were plotted and analysed using GraphPad Prism software (version 8.0, Graph-Pad, La Jolla, CA, USA), and the results are presented as the mean with standard deviation of three biological replicates. Statistical significance was determined by a two-tailed Student's t-test with Welch's correction and an ANOVA, where appropriate. * denotes p < 0.005, ** denotes p < 0.01, *** denotes p < 0.001.

Generation of 2.5D Organoid-Derived Cell Lines and 3D Cell Line Organoids (CLOs) from PDAC Organoids
We developed three new 2.5D organoid-derived cell lines and isogenic matched 3D CLOs from PDAC organoids. The 2.5D cells were developed by plating the organoids in lower concentrations of ECM, allowing the cells to migrate from the organoids, attach to the bottom of the plate, and proliferate over time ( Figure 1A). These early-passage 2.5D cells were then cultured using organoid conditions for two passages, which resulted in the generation of CLOs. The CLOs exhibited similar morphology patterns and rates to the original organoid ( Figure 1B). The newly developed cell lines were cultured long term and maintained their morphology after several passages similar to established PDAC cell lines ( Figure 1C). lower concentrations of ECM, allowing the cells to migrate from the organoids, attach to the bottom of the plate, and proliferate over time ( Figure 1A). These early-passage 2.5D cells were then cultured using organoid conditions for two passages, which resulted in the generation of CLOs. The CLOs exhibited similar morphology patterns and rates to the original organoid ( Figure 1B). The newly developed cell lines were cultured long term and maintained their morphology after several passages similar to established PDAC cell lines ( Figure 1C).

Relative Proliferation of Newly Established Primary 2D Cell Lines, Organoids, and CLOs
Proliferation assays were performed using the Incucyte Live Cell Imaging System to measure the growth rate of the newly generated 2D cell lines (>8 passages). The cell line PDM41 reached confluency within 4 days. PDM37 and PDM106 cell lines did not reach 100% confluency but achieved a maximal confluency of 47 ± 6% and 54 ± 6%, respectively, within 10 days, highlighting a slower rate of proliferation for these cell lines (Figure 2A). For organoids and CLOs, the total brightfield area of organoids and matched CLOs was calculated relative to day 0. Similar proliferation rates were observed between the isogenic matched organoids and CLOs. The PDM37 organoid displayed a relative size of 4.3 ± 1.6 compared to 4.5 ± 1.5 in PDM37 CLO. In addition, PDM106 organoid and CLO also displayed similar relative sizes of 6.0 ± 3.0 and 6.7 ± 1.5 after 9 days of growth, respectively ( Figure 2B,C). The PDM41-CLO proliferated faster with a relative size of 4.0 ± 1.0 compared to 2.3 ± 0.8 PDM41-organoid ( Figure 2D).
100% confluency but achieved a maximal confluency of 47 ± 6% and 54 ± 6%, respectiv within 10 days, highlighting a slower rate of proliferation for these cell lines ( Figure  For organoids and CLOs, the total brightfield area of organoids and matched CLOs calculated relative to day 0. Similar proliferation rates were observed between the isog matched organoids and CLOs. The PDM37 organoid displayed a relative size of 4.3 ± compared to 4.5 ± 1.5 in PDM37 CLO. In addition, PDM106 organoid and CLO also played similar relative sizes of 6.0 ± 3.0 and 6.7 ± 1.5 after 9 days of growth, respectiv ( Figure 2B and 2C). The PDM41-CLO proliferated faster with a relative size of 4.0 ± compared to 2.3 ± 0.8 PDM41-organoid ( Figure 2D).

Therapeutic Response between 2D, CLOs, and Organoids
To assess the drug sensitivity between 2D and 3D CLOs and their original organo cell models were treated with different concentrations of 5-FU, cisplatin, and n paclitaxel. The viability of 3D CLOs and organoids decreased in a dose-dependent man and showed a similar responsive profile to each other (Figure 3), except PDM106 org oids were non-responsive to cisplatin drug treatment. The 2D cell lines tended to be responsive to all the drugs tested compared to the 3D CLOs and organoids. An IC50 v

Therapeutic Response between 2D, CLOs, and Organoids
To assess the drug sensitivity between 2D and 3D CLOs and their original organoids, cell models were treated with different concentrations of 5-FU, cisplatin, and nab-paclitaxel. The viability of 3D CLOs and organoids decreased in a dose-dependent manner and showed a similar responsive profile to each other (Figure 3), except PDM106 organoids were nonresponsive to cisplatin drug treatment. The 2D cell lines tended to be less responsive to all the drugs tested compared to the 3D CLOs and organoids. An IC50 value was not achieved in response to cisplatin and nab-paclitaxel in PDM41-CL and nab-paclitaxel in PDM106-CL (Table 1). This suggests our recapitulated 3D CLOs maintain the drug response profiles of their original 3D organoids and could be used as an easily expandable resource for drug screening.
was not achieved in response to cisplatin and nab-paclitaxel in PDM41-CL and nabpaclitaxel in PDM106-CL (Table 1). This suggests our recapitulated 3D CLOs maintain the drug response profiles of their original 3D organoids and could be used as an easily expandable resource for drug screening.

Stem Cell Marker Expression between Isogenic 2D Primary Cell Lines, Organoids, and CLOs
RT-qPCR was performed to examine the gene expression of stem cell-associated markers, NANOG, OCT4, and SOX2, in all models. These markers are key transcription regulators that maintain the self-renewal capabilities and pluripotency properties of stem

Stem Cell Marker Expression between Isogenic 2D Primary Cell Lines, Organoids, and CLOs
RT-qPCR was performed to examine the gene expression of stem cell-associated markers, NANOG, OCT4, and SOX2, in all models. These markers are key transcription regulators that maintain the self-renewal capabilities and pluripotency properties of stem cells [30,31]. A significant up-regulation of NANOG and OCT4 was observed in PDM37-CLOs and organoids relative to the 2D PDM37-CL (p < 0.05); however, there was a nonsignificant expression of SOX2 in both the PDM37 organoids and isogenic CLOs compared to the 2D PDM37-CL ( Figure 4A). NANOG expression was up-regulated in PDM41 CLO and organoids relative to the 2D PDM41-CL. OCT4 levels were non-significantly increased in PDM41 organoids and CLOs compared to the PDM41-CL. SOX2 was down-regulated in PDM41-CLOs (p < 0.01), but was increased in PDM41-organoids compared to the 2D PDM41-CL (p < 0.01) ( Figure 4B). In the PDM106 models, NANOG and SOX2 levels were all significantly higher in PDM106 organoids and CLOs compared to the 2D PDM106-CL, but OCT4 levels were not significantly increased in organoids and CLOs compared to the 2D PDM106-CL ( Figure 4C). In general, we observed an increased expression of two SC markers in both CLOs and organoids relative to 2D cell lines in all three models.
CLO and organoids relative to the 2D PDM41-CL. OCT4 levels were non-significantly increased in PDM41 organoids and CLOs compared to the PDM41-CL. SOX2 was downregulated in PDM41-CLOs (p < 0.01), but was increased in PDM41-organoids compared to the 2D PDM41-CL (p < 0.01) ( Figure 4B). In the PDM106 models, NANOG and SOX2 levels were all significantly higher in PDM106 organoids and CLOs compared to the 2D PDM106-CL, but OCT4 levels were not significantly increased in organoids and CLOs compared to the 2D PDM106-CL ( Figure 4C). In general, we observed an increased expression of two SC markers in both CLOs and organoids relative to 2D cell lines in all three models.

Immunofluorescence Analysis of Cancer Stem Cell Markers
Immunofluorescence was carried out to determine whether the expression profiles of pancreatic cancer stem cell markers, ALDH1A1, CXCR4, HCAM, and EpCAM, were maintained between 2D cell lines, CLOs, and organoids in the PDM41 and PDM106 models. ALDH1A1 expression was higher in 2D cell lines compared to CLO in both models ( Figure 5A). Increased abundance of CXCR4 and HCAM was observed in CLOs and organoids compared to cell lines in both models ( Figure 5B,C). There was no significant increase in EpCAM expression between 2D and organoids in both PDM41 and PDM106. Contradictory findings of EpCAM expression between the 2D and CLO models, where PDM41-CLO expressed 3-fold higher levels of EpCAM and PDM106-CLO had a non-significant change compared to 2D, indicated that EpCAM expression may be cell specific ( Figure 5D).

Immunofluorescence Analysis of Cancer Stem Cell Markers
Immunofluorescence was carried out to determine whether the expression profiles of pancreatic cancer stem cell markers, ALDH1A1, CXCR4, HCAM, and EpCAM, were maintained between 2D cell lines, CLOs, and organoids in the PDM41 and PDM106 models. ALDH1A1 expression was higher in 2D cell lines compared to CLO in both models ( Figure 5A). Increased abundance of CXCR4 and HCAM was observed in CLOs and organoids compared to cell lines in both models ( Figure 5B,C). There was no significant increase in EpCAM expression between 2D and organoids in both PDM41 and PDM106. Contradictory findings of EpCAM expression between the 2D and CLO models, where PDM41-CLO expressed 3-fold higher levels of EpCAM and PDM106-CLO had a nonsignificant change compared to 2D, indicated that EpCAM expression may be cell specific ( Figure 5D).

In Vivo Tumorigenesis Comparison in Novel 2D Cell Line, CLO, and Organoid
In order to assess the ability of our models to form tumours in vivo, subcutaneous injection of the 2D cells, CLOs, and organoids from our previously established PT291 models [29] into SCID immunodeficient mice was performed. In this study, all mice in the organoid and CLO groups developed tumours; however, only one of the mice implanted with the 2D cell line developed a measurable tumour. The proliferation rates of the PT291 models are shown in Supplementary Figure S1. The original PT291 PDX in vivo tumour growth is included for comparison [32] (Figure 6A). There was no significant difference in weight change between models ( Figure 6B). A repeated measure ANOVA was used to determine statistical significance (* p < 0.05; ** p < 0.01; *** p < 0.005; **** p < 0.0001; ns non-significant).
HCAM were observed ( Figure 6C-F). Pancreatic cancer markers MASPIN and PDX1 were strongly expressed ( Figure 6G,H). Intense nuclear staining was observed for Ki67 in staining in CLO xenografts ( Figure 6I). These findings suggest that the CLOs have the capacity to form tumours in vivo and maintain the expression of pancreatic cancer and cancer stem cell markers.  H&E staining of the tumours showed the histopathology components of pancreatic cancer cells (Supplementary Figures S2 and S3). Immunohistochemistry was performed on CLOs for pancreatic cancer stem cell markers ALDH1A1, CXCR4, EpCAM, and HCAM; pancreatic cancer markers MASPIN and PDX1; and the proliferation marker Ki67. Strong expression of ALDH1A1, but low levels of CXCR4, EpCAM, and no expression of HCAM were observed ( Figure 6C-F). Pancreatic cancer markers MASPIN and PDX1 were strongly expressed ( Figure 6G,H). Intense nuclear staining was observed for Ki67 in staining in CLO xenografts ( Figure 6I). These findings suggest that the CLOs have the capacity to form tumours in vivo and maintain the expression of pancreatic cancer and cancer stem cell markers.

Discussion
The absence of representative in vitro tumour models is one of the obstacles to the study of disease progression and the development of effective treatments in PDAC. As it is simple to use, inexpensive, and highly replicable, 2D cell culture is a vital model in cancer research and drug screening; however, there are several limitations associated with it [33], as 2D culture can cause genetic mutations as well as rapid changes in the polarity and morphology of the cells. Furthermore, as cells adhere and grow into a single layer on the flat surface, traditional 2D cell culture does not offer an adequate in vivo environment as it fails to resemble the complex natural microenvironment experienced by cells in the tumour [34]. In addition, as cells in 2D monolayers receive the same amount of nutrients and growth factors in the medium, they tend to have homogenous growth and proliferation. This causes more cells to be in the same stage of the cell cycle, which does not mimic the heterogeneous nature of the tumours [33,35]. Novel 3D cell culture systems have overcome these limitations. They provide the spatial cell-cell interactions and cell-ECM interactions and represent more accurate in vivo models. These interactions play vital roles in different cellular functions, including cell differentiation, proliferation, vitality, expression of genes and proteins, and drug metabolism [36].
Although organoids are considered a more representative model of PDAC, their usefulness could be limited due to difficulties in obtaining patient samples as well as their time-consuming maintenance. In this present study, we illustrate the possibility of establishing a culture methodology for 2.5D cell lines for scale-up to CLOs and the long-term culture of 2D cell lines derived from established PDOs. We performed detailed comparative experiments assessing the morphology and phenotypic characteristics, such as proliferation rates, chemotherapeutic drug sensitivity, stem cell, and CSC markers, and in vivo growth and protein expression analysis of isogenic 2D cell lines, CLOs, and organoids.
Previous studies have shown that cells grown in 2D versus 3D have different proliferation rates and sensitivity levels to chemotherapeutic drugs, with 3D spheroid models showing increased levels of drug resistance [37,38]. In our study, the organoids and matched CLOs had similar proliferation rates and displayed increased sensitivity to treatment with 5-FU, cisplatin, and nab-paclitaxel compared to their isogenic 2D matched cell line. In accordance with our findings, a recent study by Christian et al. showed that 22 of 26 compounds in a drug screen were potent in the majority of PDAC organoid lines, but most compounds were not active in 2D cancer cell lines, confirming that the culture environment has a strong influence on drug responses [39].
Stem-cell-related pluripotency genes, such as NANOG, OCT4, and SOX2, are expressed in aggressive cancers that show resistance to clinical treatments, resulting in poor survival. These transcription factors control the expression of pluripotent factors and can reduce the expression of lineage-specific genes [31,32,40]. We observed an upregulation of key stem cell markers and transcription factors, NANOG, OCT4, and SOX2, in CLOs compared to 2D primary cultures, indicating a pluripotent stem cell state in these models. In addition, CSCs are considered one of the main drivers of tumour heterogeneity and malignancy. These cells are involved in cancer progression, recurrence, metastasis, and resistance to anti-cancer therapies. CSCs in various tumour types express specific surface markers, which are used to isolate and characterise these populations. Different surface markers such as CD24, HCAM, CXCR4, CD133, and EpCAM (epithelial specific antigen, ESA) are commonly used to identify CSCs in PDAC [14,41,42].
Molecular profiling of cancer stem cell markers CXCR4, EpCAM, HCAM, and ALDH1A1 was performed in our models. CXCR4 and CXCL12 of the chemokine signalling network promote PDAC tumour growth, invasion, chemoresistance, and metastasis by regulating the tumour microenvironment through complex crosstalk with Akt, ERK, c-myc, β-catenin, NFκβ, and p53 signalling pathways. CXCR4/CXCL12 expression is involved in the early stages of pancreatic carcinogenesis in PanINs and maintained into later phases during disease progression [43,44]. We observed consistent low expression of CXCR4 in the 2D cell line models compared to increased expression in the isogenic organoid and CLO models. Hypoxia and HIF1α have been shown to promote the expression of CXCR4 in PDAC and glioblastoma [45,46]. PDAC tumours tend to display significantly more hypoxia compared to other solid tumours; under oxygen limited conditions, hypoxia-inducible pathways are activated, initiating a cascade of signalling events that promote the aggressiveness of pancreatic cancer [47], and 3D organoids display a hypoxic core mimicking the tumour microenvironment gradients [48], which further validates our use of organoids and CLOs as in vitro models of pancreatic cancer.
EpCAM/ESA is specifically expressed in epithelial tissue and overexpressed in numerous epithelial-derived tumours such as those of the pancreas, stomach, lung, kidney, breast, and liver [49]. EpCAM is thought to participate in cellular proliferation, cell signalling pathways, migration, and differentiation, as well as in metastasis and cancer stem cells possibly through regulation of the PTEN/AKT/mTOR pathways [40,50]. In pancreatic cancer, overexpression of EpCAM is associated with a shorter overall survival rate [51].
Li et al. found that the subpopulation of CD44 + CD24 + ESA + pancreatic cancer cells had stem cell-like characteristics of self-renewal, exhibited increased tumorigenic potential, and had elevated expression of the developmental signalling molecule sonic hedgehog [41]. The expression of the transmembrane glycoprotein EpCAM was retained in our 2D and 3D models but showed cell-dependent expression. Although EpCAM is known to be highly expressed in PDAC and established cell lines [52], studies showed that breast cancer cells with a mesenchymal phenotype are independent of EpCAM in invasion processes and tumour progression [53]. Moreover, overexpression of EpCAM in vivo inhibited the invasive growth of fibroblastic L-cells and dedifferentiated mammary carcinoma L153S cells [54].
HCAM, also known as CD44, is an adhesion molecule involved in cell-cell and cellmatrix adhesion, as well as cellular growth and differentiation. HCAM is a known CSC marker, plays an important role in inducing EMT, and correlates to a more mesenchymal phenotype in cancer [55]. Expression of HCAM was highly expressed in all three organoid and CLO models compared to isogenic 2D cell lines. We observed a similar low HCAM expression in the PT291 2D cell line and matched CLO (Supplementary Figure S3), which was also confirmed in the in vivo PT291-CLO xenograft. Studies in prostate cancer showed that higher expression of HCAM may be cell type dependent rather than related to culture conditions, suggesting that PT291 may have a more epithelial than mesenchymal phenotype [56].
Functional markers are also used to identify CSCs within the tumour. A clinical study of patients with pancreatic cancer showed that ALDH + pancreatic tumour cells are associated with metastatic disease progression and shortened overall survival, suggesting ALDH expression as a prognostic marker of PDAC [57]. ALDH1A1 is involved in retinoic acid metabolism and has a role in proliferation and differentiation [58]. High expression levels of ALDH1A1 were observed in the PDM41 organoids but reduced expression in the isogenic 2D cell lines and CLOs. High expression levels of ALDH1A1 were also observed in the PT291 PDO xenograft.
Finally, to assess if the organoids, CLOs, and 2D cell lines retained their original tumour characteristics after being propagated long-term as organoids, the PT291 2D cell line, CLO, and organoid models were re-implanted as subcutaneous xenografts. All cell models resulted in tumour growth, with measurable tumours for the organoid and CLO tumours forming. Interestingly, organoid tumours grew faster than the matched CLO tumours. H&E staining showed clear morphological differences between the tumours developed from the PT291 organoid, cell line, and CLO. The organoids and cell lines developed large cell clusters with no tumour-like structures forming; however, the CLO tumours formed adenocarcinomas with atypical cells, glands, and/or a desmoplastic stroma, and are more representative of an in vivo tumour. Sylvia et al. showed that orthotopic implantation of PDOs resulted in the organoid progressing through all stages of tumour development, from PanIN to a PDAC tumour [59].

Conclusions
In conclusion, this study is the first to generate PDAC 2D and 3D CLOs that are isogenic and derived from PDAC PDOs. We show that the CLOs represent the histological architecture of PDAC and retain the phenotypic and molecular features of PDOs. The CLOs showed a similar sensitivity response to chemotherapeutic agents with their derived 3D PDOs compared to 2D cell lines. Furthermore, the CLOs demonstrated tumorigenesis in vivo and maintained the expression pattern of PDAC markers. These results suggest that our established CLO culture method can be used as an expandable, easy scale-up, affordable, and less time-consuming research model mirroring 3D organoid studies to investigate the mechanisms of PDAC.