Page 90 - OR-1-1

allowed to solidify at 37°C for 30 min. The organoids were calculate the ratio of live to dead cells. Results were
then covered with cholangiocarcinoma organoid culture expressed as cell viability (number of live cells/total
medium (A211L11, Amoolo Biotech) and incubated at cell number) and subjected to statistical analysis to
37°C with 5% CO . To specifically support the proliferation evaluate the sensitivity of organoids to the drugs. The
2
of malignant cholangiocytes and sustain NOTCH pathway PDOs inhibition ratio is calculated by subtracting the
activation for maintaining a differentiation-supportive cell viability of the drug group from that of the control
tumor microenvironment in cholangiocarcinoma, we group using 1-cell viability.
supplemented the commercially available ICC organoid (ii) Cell viability assessment: During the exponential
culture medium (Amoolo Biotech) with recombinant growth phase of the organoids (typically on the 4 –
th
FGF10 (50 ng/mL, Gibco, US) and Jagged1 (100 ng/mL, 5 day of culture), the PDOs were exposed to drugs
th
Gibco). The medium was refreshed every 4 days. MycoAlert in a six-point 5-fold dilution series ranging from 50
Mycoplasma Detection Kit (LT07-710, Lonza, Switzerland) to 0.016 μmol/L for 72 h. After drug intervention,
was used to regularly check for mycoplasma contamination cell viability was also assessed using CellTiter-Glo
in PDOs, and short tandem repeat analysis was performed 3D (Promega), and the data were analyzed using
using a multiplex amplification kit (PowerPlex 21 system, GraphPad Prism 9.0 (GraphPad Software, USA), with
Promega, US) for identification.
the half-maximal inhibitory concentration (IC ) values
50
2.4. PDOs-based drug sensitivity testing calculated using non-linear regression (curve fitting).
PDOs-based drug sensitivity testing was conducted as 2.5. Treatment and follow-up
previously described. 11,12 In brief, the testing evaluated
with four chemotherapy regimens: GC (gemcitabine After hospitalization, ICC patients underwent
and cisplatin in a 1:1 ratio), GEMOX (gemcitabine and multidisciplinary team (MDT) consultations. If the
oxaliplatin in a 1:1 ratio), capecitabine, and S-1. Since MDT determined that the primary lesion was suitable
capecitabine is a prodrug of fluorouracil (5-FU), it is for R0 resection, the patient underwent surgical
metabolized into 5-FU in vitro to exert its antitumor effects. resection. All patients received adjuvant chemotherapy
Therefore, 5-FU was utilized as a substitute in the drug postoperatively, including GC, GEMOX, capecitabine, or
sensitivity testing in vitro. The active components of S-1 S-1. 5,6,13,14 Follow-up time was measured from the start of
that exert antitumor effects are the 5-FU prodrug tegafur chemotherapy to the last follow-up. The primary endpoint
and gimeracil. Similarly, the drug sensitivity testing in vitro of this study is RFS.
for S-1 used a combination of 5-FU and gimeracil (in a Patients were prospectively monitored for recurrence
5:2 ratio) as a substitute. All drug tests were performed in with serum tumor markers and imaging studies, including
triplicate, with each test including two technical replicates. ultrasonography, computed tomography, and/or magnetic
The evaluation approaches of PDOs-based drug sensitivity resonance imaging. Recurrence was defined as a biopsy-
included the following two methods: proven recurrent lesion or radiological evidence with cross-
(i) Live/dead fluorescence staining: Once the PDOs were sectional imaging and/or elevated levels of cancer antigen
successfully established, they were treated with the 19-9 and carcinoembryonic antigen. 15
proposed four drug regimens (the tested concentrations
of these drugs are shown in Table 1). Following 48 h of 2.6. Statistical analysis and machine learning
incubation, the organoids were washed 3 times with
phosphate-buffered saline to remove any unbound The Kaplan–Meier method was used to estimate RFS for
drugs. Subsequently, live/dead fluorescent staining patients in matched and unmatched group, and the log-rank
was performed using a Calcein-AM and Propidium test was used for comparison. Univariate and multivariate
Iodide Kit (P4864, Sigma-Aldrich, US). According Cox regression analyses were performed to identify
to the manufacturer’s instructions, Calcein-AM independent predictive factors for RFS. Statistical analyses
(2 μM) was added to the organoid culture medium were conducted using R Studio version 4.4.2 (R Project for
and incubated for 30 min to label live cells. Propidium Statistical Computing, Vienna, Austria, www.r-project.org).
iodide (4 μM) was then added and incubated for an Machine learning models were developed using Python
additional 10 min to label dead cells. After staining, the version 4.10. The eXtreme Gradient Boosting (XGBoost)
automated microscope (BioTek. Lionheart FX, BioTek, model was utilized to identify the most influential variables
US) was used to image and analyze the captured for predicting recurrence risk. The best hyperparameters
images of different drug treatment groups to determine are determined using the grid search strategy. Then, the
the spheroid viability, fluorescent intensities, and importance of each feature was ranked using the shap
morphology. ImageJ software (National Institutes package. A two-sided p<0.05 was considered statistically
of Health, USA) was utilized for image analysis to significant.

Volume 1 Issue 1 (2025) 3 doi: 10.36922/or.8571