Decellularized extracellular matrix as scaffold for cancer organoid cultures of colorectal peritoneal metastases

ABSTRACT Peritoneal metastases (PM) from colorectal cancer (CRC) are associated with poor survival. The extracellular matrix (ECM) plays a fundamental role in modulating the homing of CRC metastases to the peritoneum. The mechanisms underlying the interactions between metastatic cells and the ECM, however, remain poorly understood, and the number of in vitro models available for the study of the peritoneal metastatic process is limited. Here, we show that decellularized ECM of the peritoneal cavity allows the growth of organoids obtained from PM, favoring the development of three-dimensional (3D) nodules that maintain the characteristics of in vivo PM. Organoids preferentially grow on scaffolds obtained from neoplastic peritoneum, which are characterized by greater stiffness than normal scaffolds. A gene expression analysis of organoids grown on different substrates reflected faithfully the clinical and biological characteristics of the organoids. An impact of the ECM on the response to standard chemotherapy treatment for PM was also observed. The ex vivo 3D model, obtained by combining patient-derived decellularized ECM with organoids to mimic the metastatic niche, could be an innovative tool to develop new therapeutic strategies in a biologically relevant context to personalize treatments.


Morphological evaluation of the decellularized matrices
3D-dECMs from normal peritoneum and PM lesions were washed twice with 1X PBS and placed in a 60 mm petri dish. Samples were illuminated with a widefield lamp laser to visualize the architecture of the collagen fibers. An image format of 1024x1024 pixels was used and all images were acquired with Leica Application Suite X, ver. 3 software. 3D-dECMs FFPE sections deriving from normal and PM peritoneum were used to perform polarized light microscopy (PLM). FFPEs were analyzed with an Olympus BX63 upright widefield microscope equipped with a motorized stage and a Hamamatsu OrcaAG camera, using Metamorph software. UplanSApo 4X/0.16 N.A objective was used to acquire the mosaics of the sections. Insets were acquired with UplanSApo 10X/0.4 N.A. and UplanSApo 20X/0.75 N.A. objectives. All experiments were perfomed at least in duplicate. Confocal reflection microscopy images were acquired with a Leica TCS SP8 laser confocal scanner mounted on a Leica DMi8 microscope through a HC PL FLUOTAR 20×/0.5 NA.

Nanoscale topographical analysis of 3D-dECMs
The topographical evaluation of the 3D-dECMs was performed by atomic force microscopy (AFM) analysis on samples deriving from normal peritoneum and PM of three different patients. Before the AFM analysis, the 3D-dECM slides were left for 30 minutes at RT to dissolve the optimal cutting temperature (OCT) compound. Then, the samples were carefully washed with ultrapure water and covered with 1X PBS buffer. AFM topographic measurements were carried out at RT using a NanoWizard3 AFM (JPK, Germany) coupled to an Olympus BX61 inverted microscope and equipped with tapping mode silicon ACTG AFM probes (APPNANO). The 50 µm thick tissue slices, instead, were mounted on polarized glass slides (ThermoFisher Scientific), left for 30 minutes at RT and carefully washed with ultrapure water. The topography of each tissue was characterized by collecting at least 10 areas (5x5 µm 2 ) of the sample surface with 512×512 points (scan speed 3,5 μm s −1 ).

ECM component quantification
Total collagen and sulphated glycosaminoglycan (sGAG) content in fresh and decellularized normal and PM peritoneum were quantified using the SIRCOL collagen assay (Biocolor, Carrickfergus, UK) and the Blyscan GAG assay kit (Biocolor), respectively. The experiments were performed in triplicate following the manufacturer's instruction. Data are the mean of three different neoplastic and normal-derived samples obtained from three different donors.

Nanoindentation measurements by AFM
AFM mechanical analysis was carried out on 3D-dECMs deriving from normal peritoneum and PM of five patients. 3D-dECMs were embedded in OCT and frozen with nitrogen-cooled 2-propanol for 10 seconds. Slices of 100 µm thickness were cut with a microtome (Leica) and attached to positively charged poly-lysine coated glass coverslips (ThermoFisher Scientific), exploiting the electrostatic interaction. Nanomechanical tests were performed in liquid on samples covered by a PBS droplet confined by a circular ridge of hydrophobic two-component silicone paste (Leica). A Bioscope Catalyst AFM (Bruker) was used, which was resting on an active anti-vibration base (DVIA-T45, Daeil Systems) and put into an acoustic enclosure (Schaefer). The measurements were performed at RT. Custom monolithic borosilicate glass probes consisting of spherical glass beads (SPI Supplies), with radii R in the range of 7.5-12.5 µm, were attached to tipless cantilevers (Nanosensor, TL-FM) with nominal spring constant k = 3-6 N/m. Probes were fabricated and calibrated, in terms of tip radius, according to an established custom protocol (Indieri et al., 2011).
The spring constant was measured using the thermal noise calibration (Hutter and Bechoefer, 1993) and corrected for the contribution of the added mass of the sphere (Chighizola et al., 2021;Laurent et al., 2013;). The deflection sensitivity was calibrated in situ and non-invasively before every experiment by using the previously characterized spring constant as a reference, according to the SNAP procedure described in (Schillers et al., 2017).
The mechanical properties of the 3D-dECMs were obtained by fitting the Hertz model to sets of force versus indentation curves (simply force curves, FCs), as described elsewhere (Schillers et al., 2017;Nebuloni et al., 2016;Puricelli et al., 2015;Shimshoni et al., 2020), to extract the value of the YM of elasticity, which measures ECM rigidity. FCs were collected in Point and Shoot (P&S) mode, selecting the regions of interest from optical images, exploiting the accurate alignment of the optical and AFM images obtained using the Miro software module integrated in the AFM software.
Each set of FCs consisted of an array of typically 15x15=225 FCs spatially separated by 5-10 µm, each FC containing 8192 points, with ramp length L = 8-15 µm, maximum load Fmax = 150-1500 nN, and ramp frequency f = 1 Hz. The maximum load was chosen in order to achieve a maximum indentation in the range of 4-9 µm. Typical approaching speed of the probe during indentation was 16-30 µm/s. Five samples were characterized for each condition. In each sample, 3-10 P&S were acquired in macroscopically separated locations, for a total of 10-25 independent P&S per patient and condition (up to 2250-5500 FCs per patient and condition).

Stem cell maintenance, proliferation and apoptosis assays
Growing cells, stem cells and apoptotic cells were detected on FFPE sections. Growing cells, deriving from disaggregated TDO, were stained with anti-human Ki-67 monoclonal antibody (clone MIB-1) and DAPI, and their growth rate was expressed as the percentage of Ki-67-positive cells present in fields devoid of dead cells. Stem cells were stained with anti-human LGR5 monoclonal antibody (clone OTI2A2) and DAPI, and their density was expressed as the percentage of LGR5-positive cells present in fields devoid of dead cells. Apoptotic cells were stained with anti-human cCASPASE3 monoclonal antibody (clone 9661) and DAPI, and the apoptotic rate was calculated as the percentage of cCASPASE3-positive cells present in the field. The percentage of Ki-67-positive, LGR5-positive and cCASPASE3-positive cells was obtained by dividing the number of positive cell present in one field by the total number of cells in one field, multiplied by 100. Cells in three independent fields (40X magnification) were counted using ImageJ software. The experiments were performed in triplicate using three different neoplastic and normal-derived matrices obtained from three different donors.

Qpath analyses
Percentage estimation and cell counting were performed using Qupath software (https://qupath.github.io, version 0.2.3). The images used for Qpath analyses were acquired using Aperio Leica ScanScope XT (Leica Biosystems, Wetzlar, Germany). The slides were evaluated by an expert pathologist. The percentage of CK AE1/AE3, CK20, CK19, CDX2, Ki-67, and LGR5 positive cells was calculated by dividing the number of positive cells present in each field by the total number of cells in the same field. TDO-derived infiltrating cells were evaluated by calculating the total number of H&E stained cells. Three fields were counted per experiments.

RNA-seq analysis
Gene expression profiles were conducted on C1, C2 and C3 organoid cultures grown in Matrigel and on 3D-dECMs. Total RNA was extracted using TRIzol™ reagent (QIAGEN). Qubit fluorimeter (ThermoFisher Scientific) and Agilent Bioanalyzer 2100 (RIN > 8) were used to measure and assess RNA abundance and integrity, respectively. Indexed library preparation was performed starting with 500 ng total RNA with the TruSeq stranded mRNA (Illumina) according to the manufacturer's instructions. RNA-seq was performed in PE mode (2x75nt) on an Illumina NextSeq550 platform, generating an average of 55 million PE reads per sample. For every condition (Matrigel, normal 3D-dCM and neoplastic 3D-dECM), two replicates per organoid were sequenced, for a total of 18 data points. Raw reads were aligned to the human transcriptome (hg38) with STAR (Dobin et al., 2013) using the quantMode option to generate transcripts counts. Differentially expressed genes in the three growth conditions were identified with DESeq2 (Alshehri, 2018). All p-values were adjusted for false discovery rate with the Benjamini-Hochberg method.

Gene Set Enrichment Analysis
Gene Set Enrichment Analysis was performed with the enrichR R package (Chen et al., 2013) on deregulated genes (absolute fold change > 2 and adjusted p-value <0.05). In particular, the enrichment for the Matrisome database was assessed. This database provides live cross-referencing to gene and protein databases for every ECM and ECM-associated gene, also integrating experimental proteomic data on ECM and ECM-associated proteins and genes from the ECM Atlas (Naba et al., 2016). Gene sets with adjusted p-value <0.05 were considered significantly enriched.

Quantitative real-time polymerase chain reaction (qRT-PCR)
For gene expression analysis, cDNA was synthesized from 100 ng of total RNA using a High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific, Waltham, MA) and qPCR was carried out with gene-specific assays for MT1A (Hs00831826_s1), LOX (Hs00942480_m1), THY1 (Hs00174816_m1), FZD9 (Hs00268954_s1), SPP1 (Hs00959010_m1), and performed using the TaqMan FAST Universal PCR Master Mix, no AmpErase® UNG in a PRISM 7900HT Real-Time PCR system (Thermo Fisher Scientific). The expression values of the genes were normalized to GAPDH (Hs99999905_m1).

Dose-response curves for HIPEC treatment
To determine the IC 50 value of MMC and OXA, 5x10 3 C1, C2 and C3 TDO were suspended in 100 µl of culture medium and seeded on 96-well plates ( Table S4). The signals were detected using enhanced chemiluminescence, and protein levels were quantified using Imagelab software (Bio-Rad, Hercules, CA, USA). Each experiment was repeated at least three times.

Statistical analyses
Statistical analyses were performed using GraphPad Prism software (version 8.4.1 (676), GraphPad Software, San Diego, USA). Data are expressed as mean and SEM. A two-tailed Student's t test was used to compare paired groups. Differences among groups were evaluated using two-way ANOVA. In the case of AFM mechanical experiments, for each patient and each condition tested, the median values of the YM were extracted from each measured location (P&S) using the procedure described in Cramer et al (Shimshoni et al., 2020;Cramer, 1999). The distributions of the measured YM values were obtained by grouping all P&S measured in all locations, for each patient and each condition tested. The mean and median values and the corresponding standard deviations of the mean (as SEM) were calculated by averaging between P&S. The statistical significance of differences between normal and neoplastic conditions was estimated by applying the two-tailed t test. A p-value <0.05 was considered statistically significant.