Intratumoral Cellular Heterogeneity Critically Determines Extracellular Vesicle Uptake in Colorectal Cancer

The majority of colorectal cancer (CRC) patients carry mutations in the APC gene, which leads to an unregulated activation of the Wnt pathway. Extracellular vesicles (EV) are considered as potential therapeutic tools. Although CRC is a genetically heterogenous disease, the signicance of the inter- and intra-tumor heterogeneity in EV uptake of CRC cells is not yet known. By using mouse and patient-derived organoids, the currently available best model of capturing cellular heterogeneity, we found that Apc mutation induces the expression of interferon-induced transmembrane protein 1 (Itm1), a membrane protein that plays a major role in cellular antiviral responses. Importantly, organoids derived from IFITM1 high CRC cells contained more proliferating cells. Whereas we found no difference in the uptake of liposomes between IFITM1 high and IFITM1 low/− cells, IFITM1 high cells had a markedly reduced uptake of broblast EVs as compared to IFITM1 low/− cells. In contrast, there was no difference in the intensity of EV release between CRC subpopulations with high and low IFITM1 level. Importantly, the difference in cell proliferation between these two subpopulations disappeared in the presence of broblast-derived EVs, proving the functional relevance of the enhanced EV uptake by IFITM1 low CRC cells. Collectively, we identied CRC cell subpopulations with functional difference in their EV, but not liposome uptake ability that must be taken into consideration when using EVs as therapeutic tools for targeting cancer cells.


Introduction
Colorectal cancer (CRC) is a leading cause of cancer-related death in developed countries. In the vast majority of CRC patients, APC mutation is a central genetic event leading to the ligand-independent activation of the Wnt pathway and to the uncontrolled proliferation of adenoma cells. CRC is both genetically and phenotypically diverse, and other driver mutations often occur such as KRAS, TP53, inactivation of the TGFβ signalling pathway etc. This accumulation of mutations results in the progression of adenomas to invasive carcinomas. Recently, a common clustering of CRC has been suggested with CMS1-CMS4 subtypes that have characteristic mutation pro les [1]. However, CRC not only shows a large diversity among patients, but also an intra-tumoral cellular heterogeneity that is critical in therapy. Of note, one of the most state-of-the-art methods to capture this intra-tumoral heterogeneity of human cancers is the organoid technology [2,3]. Organoids are derived from patient samples and they are cultured in 3D matrices, such as Matrigel under well-de ned conditions. Interestingly, the activity of the Wnt pathway is in uenced not only by mutations but also by other factors, such as hepatocyte growth factor (HGF) as well. As a consequence, Wnt signalling intensity is different among cells from the same tumor [4]. This leads to speci c Wnt target gene expression patterns and to tumor cells with functional diversity. For example, the Wnt target LGR5 is induced only at a high Wnt activity. Although nowadays the identity and the function of CRC stem cells are intensively discussed, LGR5 expression is agreed to mark a CRC cell population with stem cell properties [5,6].
The accumulation and activation of broblasts are hallmarks of some CRC subtypes. Importantly, there is a strong negative correlation between the amount of cancer-associated broblasts (CAF) and the time to disease relapse in CRC [7]. Furthermore, stromal gene expression signature predicts poor survival in patients [7]. CRC cells may undergo epithelial-mesenchymal transition and they may acquire a stromal expression pattern for a limited set of genes [8]. However, the predictive value of the stromal signature can at least partially be attributed to CAFs in the tumor tissues. Importantly, by secreting growth factors, such as HGF, CAFs contribute to the intra-tumoral cellular heterogeneity as well [4].
Extracellular vesicles (EV) are membrane-enclosed structures carrying biologically important molecules, such as proteins, lipids and RNAs between cells. Thus, they represent a special form of the intercellular communication. EVs can be classi ed according to their biogenesis, and they can be characterized by their biophysical or biochemical properties, too [9]. While exosomes are derived from the multivesicular bodies of the endosomal-lyosomal compartment, microvesicles are shed from the plasma membrane and they represent a larger EV population. In addition, large vesicles e.g. apoptotic bodies released by dying cells are regarded as EVs as well [10]. Since isolating EVs based on their cellular origin is di cult, they are often characterized by their size upon puri cation with differential centrifugation and ultracentrifugation [11]. Since EVs can be edited and loaded with speci c molecules which are then transported in a protected way, furthermore, EVs are e ciently taken up by target cells, they hold a great promise for cancer therapies as well [12]. However, whether intra-tumoral cellular heterogeneity has a role in EV uptake and whether speci c CRC subpopulations can be targeted by EVs more e ciently as compared to others, are not yet known.
Members of the interferon-induced transmembrane (IFITM) family regulate virus uptake in different compartments of the cells, thus, they critically contribute to cellular resistance against a wide range of membrane-surrounded viruses [13]. Among IFITM proteins, IFITM1 functions in the plasma membrane [14,15]. By using mouse intestinal and patient-derived organoids, here we provide evidence that differential IFITM1 expression marks an intra-tumoral cellular heterogeneity in CRC with a signi cantly different capacity for EV uptake. In addition, our results show that intra-tumoral cellular heterogeneity of CRC in IFITM1 expression has a major impact on broblast-derived EV function. Since the potential use of EVs in therapies is critically in uenced by their uptake in tumor cells, our results may have an important consequence when designing EV-based targeting tools.

Materials And Methods
Cell culture HT29 cells (ATCC HTB-38, American Type Culture Collection) and normal human colon broblasts (CCD-18Co, ATCC-1459) were cultured in DMEM containing 4500 g/L glucose (Gibco), 10% FBS (Biosera), glutamine (Sigma) and 1x penicillin/streptomycin (Gibco). Cells were washed with phosphate buffered saline (PBS) three times and cultured in serum-free medium for 2 days. Medium was then changed and EVs were collected for 2 days. Cell number was counted in a Burker chamber. We only used cells with a low (<p9) passage numbers after receiving them from ATCC. Cell cultures were regularly tested for Mycoplasma contamination with Hoechst staining and they were negative in our studies.

Human adenoma and CRC organoid cultures
The Medical Research Council of Hungary (ETT-TUKEB, No 51323-4/2015/EKU) approved the experiments with human samples and informed consent was obtained from the patients. CRC and adenoma patient-derived samples were processed according to previously published protocols [16]. Digested cell clusters were isolated by centrifugation at 650 g for 2 min and they were embedded into growth factor-reduced, phenol red-free Matrigel (Corning). In addition, we used the CRC organoid lines previously established in our research group [17]. Organoids were cultured in CRC medium containing basal medium (advanced DMEM/F12, 10mM HEPES (Sigma), penicillin/streptomycin, glutamine) and additionally B27 supplements (Gibco), 1mM N-Acetyl-Cysteine, 10 mM Nicotinamide, 50 ng/mL EGF (Peprotech), 500 nM A83-01 (Sigma), 10 uM SB202190-Monohydrochloride (Sigma). Furthermore, Y27632 (Rho kinase inhibitor, Sigma) was added for 3 days after passaging to avoid anoikis. Organoids were removed from Matrigel in every 5-6 days mechanically, they were centrifuged at 650g for 5 min and then digested with TrypLE (Thermo Fisher) for 5-10 min. CRC cell clusters were then washed with basal medium and embedded into Matrigel again in a 1:3 ratio. Clinical data of the patients and characterization of the organoids are in Table S1. Organoids #1-3 have already been published [17].

Normal human colon organoids
Normal colon tissue samples were collected from patients undergoing CRC surgical operation with a distance of more than 3 cm from the tumors [16]. Human colonic crypts were isolated according to [16] with some modi cations. Samples were cut into <0.5 cm pieces, washed with PBS ve times and incubated in PBS containing 2 mM EDTA (Sigma) for 30 min at 4 o C. The tubes had been pre-coated with 0.1% bovine serum albumin (BSA, Sigma) before the isolation procedure. Fractions were taken with PBS + 0.1% BSA into tubes containing 3 ml advanced DMEM/F12 medium (Gibco). Fractions with crypts were embedded into Matrigel and cultured in human organoid medium composed of CRC medium, supplemented with 100 ng/ml human noggin (Peprotech), 1000 ng/ml human R-Spondin1 (R&D Systems), 100 ng/ml murine Wnt3a (Peprotech) and 1 nM gastrin (Sigma). Medium was changed every 3 days.
To produce Apc-mutant organoids, we used a previously published sgRNA sequence (sgRNA4, [19]) and cloned it into the lentiCRISPR v2 plasmid (Addgene 52961). Apc-mutant organoids were produced according to [19] and [17] and they were selected by omitting R-Spondin1, EGF and noggin from the culture medium 3 days after transfection. Selected Apc-mutant organoids were used >6 days after removing the growth factors.

EV isolation for functional experiments
Serum-free medium from cultured cells was collected after 3 days. In some experiments, broblasts or HT29 cells were cultured with Vybrant TM Cell-Labeling Solution DiL dye for 10 min (according to the Cell in Suspension protocol of the manufacturer), the cells were extensively washed with DMEM-Glutamax medium (Gibco) twice to remove the free dye. After the medium was changed to serum-free after 2 days and labelled EVs were isolated after 3 days, the supernatants were serially centrifuged at 300 g for 5 min and 2000 g for 20 min at 16 °C to remove cells and cell fragments. Samples were then centrifuged at 12,500 g for 30 min. The pellet was washed with PBS, centrifuged again and used as the medium EV fraction (mEV, 12.5K fraction). Small EVs (sEV, 100K fraction) were ultracentrifuged (UC) at 100,000 g for 70 min at 4 o C, the EV-containing pellet was resuspended in PBS and UC again. The EV containing pellets were then resuspended in PBS and they were used for functional experiments.
2D experiments with organoid cells 2D experiments with organoid cells were carried out according to [20] with modi cations. Brie y, 24-well plates (Eppendorf) or 8-well culture slides (Falcon) were pre-coated with Matrigel and basal medium mixture in case of human CRC organoids 1:20 ratio, in case of mouse organoids 1:10 ratio for 1 hour at 37 °C. The liquid was then discarded and plates were dried at room temperature. 10 4 sorted CRC cells were plated into each well. The supernatant was changed to fresh medium after 1 day and the cells were treated with DiL labelled large EVs. The treated cells were xed with 4% paraformaldehyde (PFA) for 20 min the next day. Cells were then incubated with Phalloidin-iFluor 488 Reagent (Abcam) for 20 min. After covering samples with ProLong Diamond antifade mountant countaining DAPI (Thermo Fisher), images were taken with a Zeiss LSM800 or Leica TCS SP8 confocal microscopes. For mouse intestinal organoid 2D experiments, organoids were removed from Matrigel, they were centrifuged at 650g for 5 min and digested with TrypLE (Gibco) until we received a single cell suspension (5-15 min). Cells were then washed with PBS and they were cultured in SIM medium with the Rho kinase inhibitor Y27632 in 24-well plates (Eppendorf). Cells were used for EV uptake experiments at 24 h.

Nanoparticle Tracking Analysis (NTA)
Cells and organoids were cultured in FBS-free medium or in the chemically de ned CRC medium, respectively. The supernatants were harvested after 2 days, they were serially centrifuged at 300g for 5 min and at 2000g for 20 min. After centrifugation, 100 µL supernatant was diluted to 1 mL in PBS and particle concentrations and size distribution were measured on a ZetaView Z-NTA instrument (Particle Metrix). For each measurement, the cell positions were scanned at 25 o C. The following camera settings were used: auto expose, gain: 28.8, offset: 0, shutter: 100, sensitivity: 80. The videos were analyzed with a minimum area of 5, maximum area of 1000 and a minimum brightness of 20 by the ZetaView Analyze software 8.05.10. When the EV release from different experimental groups was compared, cells were cultured under the same conditions (medium volume, time, tissue culture dish format) and particle concentration data were then normalized to cell number.
Functional studies with EVs 10 8 small EVs in 10 µL were added to each well with CRC organoids (200 µL medium, 20 µL Matrigel, 48well tissue culture plates, Eppendorf). In some experiments, mEVs were incubated with the organoids in Eppendorf tubes in a total volume of 10µL for 1 hour before embedding them into the 3D matrix. DiL labelled mEVs or labelled sEVs were added to 2D cell cultures (10 6 cells in 500 µL medium) and the EV uptake was analyzed by confocal microscopy or ow cytometry (see below). The pixel intensity distribution on confocal images was calculated using a custom Python script (available from the authors upon request). Images were segmented using the Cellpose algorithm [21] to determine the cell boundaries for each cell. Pixel values were then extracted within the cell boundaries from the red channel to characterize EV uptake intensity.

Liposome production
The production and characterization of liposomes have been previously described [17]. The liposomes had a mean diameter of 105 nm. They were labelled with 2.5µL DiO uorescent dye (Thermo Fisher) in 500 uL PBS for 10 min at 37 o C. They were then ultracentrifuged at 100,000g for 70 min at 4 degree, washed with PBS and ultracentrifuged again. Cells were treated with 10 10 liposomes in 10µL PBS.

Flow cytometry and cell sorting
Organoids were removed from the 3D matrix, they were centrifuged at 650 g for 5 min, washed with PBS, mechanically disrupted by vigorous pipetting and they were then digested with TrypLE (Gibco, Thermo Fisher) until the organoids were dissociated into single cells (5-10 min). They were taken up in FACS buffer (PBS, 1mM EDTA, 25mM HEPES, 1% BSA). Cells were labelled with primary antibodies for 15 min and then with secondary antibodies for 15 min at room temperature. 10,000 events were measured with a FACSCalibur instrument (Beckton Dickinson) or cell subpopulations were sorted by a uorescent cell sorter (Sony SH800S) into tubes. Sorted cells were then centrifuged at 650 g for 10 min at 4°C and 10,000-20,000 cells were embedded into 20 µL Matrigel droplets. The same cell numbers were used among experimental groups within the same experiment.
Semi-quantitative analysis of EVs by anti-CD63 or anti-CD81-coated beads Conditioned media from broblasts or CRC organoids were harvested after 2 days and they were centrifuged at 300 g for 5 min and 2000 g for 20 min. EVs were then bound to beads coated with anti-CD63 (Thermo Fisher, 10606D) or anti-CD81 (Thermo Fisher, 10616D). Before incubating with the EV-containing samples, beads had been blocked with 0.1% BSA (Sigma) for 20 min. Six µL and 20 µL of the anti-CD81 or anti-CD63-coated beads were added to 200 µL supernatant, respectively. Beads were incubated overnight at 4 o C, they were then washed with PBS three times and the EVs were labelled with FITC-anti-CD81 or PE-anti-CD63. The percentage of positive beads was detected by ow cytometry (FACSCalibur) and they were normalized to cell number.

Immunocytochemistry
Cells were xed in 4% PFA for 20 min, blocked and permeabilized in blocking buffer (PBS with 0.1% BSA, 5% FBS and 0.1% Triton X-100). Cells were then incubated with primary antibodies at 4 o C overnight and then in secondary antibodies in blocking buffer for 2 h at room temperature. After covering samples with ProLong Diamond antifade mountant containing DAPI (Thermo Fisher), images were taken with a Zeiss LSM800 or Leica TCS SP8 confocal microscope. In some experiments with the labelled medium-sized EV fraction, the blocking buffer contained 0.05% saponin (BD Biosciences) instead of Triton X-100. The antibodies used are listed in Table S2.
Whole-mount immunostaining CRC organoids were cultured in 8-well chamber slides (BD Biosciences), xed in 4% PFA for 20 min and washed with PBS. Blocking and permeabilization were carried out in whole-mount blocking buffer (WBB, containing 5% FBS, 0.2% BSA, 0.3% Triton X-100 in PBS) for 30 min. Samples were incubated with primary antibodies at 4 o C overnight in WBB. After washing in PBS, labelled secondary antibodies were added overnight at 4 o C. The organoids were then mounted with ProLong Diamond antifade mountant countaining DAPI (Thermo Fisher) and analyzed with Zeiss LSM800 or Leica TCS SP8 confocal microscopes. Images were evaluated by the ImageJ software. The antibodies used are listed in Table S2.

Protein concentration measurement and simple Western (WES) analysis
Cells were cultured in serum free medium, EVs were pelleted with serial centrifugation/ultracentrifugation, they were washed with PBS and then resuspended in 19 µl PBS and 1 µl cOmplete™ Protease Inhibitor Cocktail (Roche). For WES analysis from cells, they were taken up in 10 µl RIPA lysis buffer. The protein concentrations of the supernatants were measured with the Micro BCA Protein Assay Kit (Thermo Scienti c) and NanoDrop ND-1000 spectrophotometer (Thermo Scienti c). 3 µl of the lysates containing 0.5 µg protein were applied to Simple Western analysis WES (ProteinSimple) following the manufacturer's instructions. SM-W004 (for analysis between 12-230 kDa), DM-TP01 total protein detection kit, DM-001 anti-rabbit detection kit, DM-002 anti-mouse detection kit and PS-ST02EZ-8 EZ Standard Pack 2 were used (ProteinSimple). The primary antibodies are listed in Table S2. The results were evaluated with the Compass for SW4.0.1 software (ProteinSimple).

Transmission electron microscopy
The EV-containing pellet after UC was washed with PBS and then resuspended in 10 µL PBS. 5 µL droplet was dried on a 300 mesh grid (Electron Microscopy Sciences, USA). EVs were xed with 4% glutaraldehyde for 10 minutes and the grid was washed with water three times. Samples were treated with 2% phosphotungstic acid, they were dried at RT and imaged with a JEM-1011 transmission electron microscope (JEOL, Japan) equipped with a Morada digital camera (Olympus, Japan) using the iTEM software (Olympus, Japan).

RNA isolation and mRNA measurements from cells
Total RNA was isolated with the miRNEasy Micro Kit (Qiagen) according to the manufacturer's protocol in 15µL water. In some experiments, cells were directly uorescence sorted into Qiazol (Qiagen). RNA concentration was determined with a NanoDrop instrument. Half µg RNA (in 20 µL nal volume) was reverse transcribed with the SensiFAST cDNA Synthesis Kit (Bioline) and quantitative PCR reactions were carried out using the SYBRGreen method with the SensiFAST SYBR Hi-ROX Kit (Bioline) on an ABI 7900HT Fast real-time PCR instrument (384-well format, 5 µL/well volume). Results were evaluated with the following formula: relative expression level = 2 -ΔCt , where ΔCt = Ct(gene of interest) -Ct(housekeeping gene). The primers are in Table S3.

Bioinformatical analysis
Data in the expressional data sets GSE37926 and GSE83513 were analyzed by the GEO2R online tool. Statistical signi cance for determining expression difference between two groups for individual genes was determined by the default moderated t-statistics and Benjamini and Hochberg false discovery rate with a p<0.05 threshold.

I tm1 expression is increased after Apc mutation
Previously we identi ed a set of genes that is activated by the loss of p53 and the activation of the Wnt pathway (the p53-suppressed invasiveness signature, PSIS) and which may account for the induction of invasiveness [22]. To nd genes that may be involved in the malignant behavior of CRC cells and in EV uptake by CRC cells, we focused on IFITM1, a member of the PSIS set and the interferon-induced gene family with cell surface localization. Oncomine data analysis (TCGA colorectal data sets, www.oncomine.org) showed that IFITM1 is highly overexpressed in CRC samples compared to normal colon and rectum (Fig. S1a), suggesting that IFITM1 is regulated by the Wnt pathway. To test this hypothesis, we established organoid cultures from wild type (WT) mouse small intestine and introduced Apc-mutation to constantly activate the Wnt pathway. As expected, we observed highly elevated RNA levels of the known Wnt targets Lgr5, Axin2 and Myc (Fig. S1b). In addition, Apc mutation resulted in a pronounced increase in the expression of Prox1, an intestine-speci c Wnt target gene [23] and we detected a more than 2,500-fold increase in the expression of I tm1 (Fig. S1c).
To further study the regulation of IFITM1, we isolated organoids from normal colon (NCO) and from the tumor (CRCO) of CRC patients. In addition, we obtained samples from patients diagnosed with colon adenoma (AO) ( Table S1). AOs and CRCOs were cultured without the Wnt-agonist R-Spondin1 and Wnt3a, thus, samples were selected for organoids carrying APC mutation and harbouring a constitutively active Wnt pathway. Interestingly, we detected higher RNA levels of not only AXIN2, LGR5 and MYC, but also of IFITM1 in both AOs and CRCOs as compared to NCOs (Fig. S1d).
LGR5 marks a stem cell population in CRC and in intestinal adenomas [5,6,24]. To decide whether I tm1 expression is speci c for the Lgr5+ cells, we analyzed microarray data from sorted Lgr5 high and Lgr5 low mouse intestinal adenoma cells [24]. Whereas the RNA levels of Cd44 and Cd133 (two other genes often considered as CRC stem cell markers) were higher in the Lgr5 high cell population, we found no difference in I tm1 (Fig. S1e). Furthermore, whereas the RNA level of CD133 was higher in LGR5 high compared to LGR5 low human CRC organoid cells in RNA expression datasets [6], we found no difference in IFITM1 expression between the two cell populations (Fig. S1f). Collectively, these data suggest that IFITM1 is under the regulation of the Wnt pathway, but it is not speci c for the LGR5 high CRC cells with stem cell features.

CRC patient-derived organoids have a heterogeneous expression of IFITM1
Although the membrane topology of IFITM1 has been unknown for a long time, recent publications suggested that the C-terminal part of the protein faces the extracellular space [14,15]. In line with these studies, we detected a ow cytometric signal with an antibody produced against the N-terminal part of IFITM1 only if cells had been permeabilized before labelling ( Fig. 1a-b). In contrast, the antibody against the C-terminal part labelled cells without permeabilization as well (Fig. 1a-b). To study both the intertumor and intra-tumor heterogeneity in IFITM1 expression, we used CRC patient-derived organoids.
Interestingly, we could not detect a large variation in the RNA level of this gene among samples (Fig. 1c). However, immunostaining proved the intra-organoid cellular heterogeneity in IFITM1 expression (Fig. 1d). Thus, IFITM1 expression is heterogeneous among CRC organoid cells.
The IFITM1 high CRC population contains more proliferating cells We next sorted IFITM1 high and IFITM1 low CRC organoid cells (Fig. 2a), checked the difference in the IFITM1 protein expression of the uorescence sorted cells by ow cytometry and capillary-based immunoblotting ( Fig. 2a-b) and subjected the samples to RNA analysis. Interestingly, whereas we found no difference in the RNAs of the Wnt target genes AXIN2, LGR5 or MYC, IFITM1 high cells expressed the mesenchymal marker genes VIM and ZEB1 at a lower level (Fig. 2c). Furthermore, we observed a decreased RNA expression of CD133, a gene that had been connected to stem-like cells in previous publications (Fig. 2c). Surprisingly, we detected no difference in the number of organoids initiated by the two sorted cell populations (Fig. 2d). However, the diameter of the organoids derived from IFITM1 high cells was signi cantly higher. They contained more KI67+ proliferating cells, but there was no difference in the percentage of active caspase3+ apoptotic cells (Fig. 2e-f and Fig. S2a). Parallel with these ndings, we observed lower RNA levels of the differentiation markers MUC2 (Goblet cells) and ALPI (enterocytes) in IFITM1 high cell-derived organoids (Fig. S2b), showing the shift between proliferating and other cell types. Thus, although classical stem-like cell markers are not highly expressed in IFITM1 high cells, they produce organoids with a higher proliferation potential compared to the IFITM1 low population.

IFITM1 high and IFITM1 low CRC cells do not differ in their intensity of EV release
The IFITM1 cell surface protein is involved in inhibiting the uptake of some membrane-surrounded virus particles [25,26], raising the possibility that this molecule regulates EV tra c as well. To study this question, we rst determined the stability of IFITM1 expression in sorted cell populations. Organoidderived CRC cells were viable on Matrigel-coated plates (Fig. S2c) and we found no difference in cell surface IFITM1 expression between cells cultured either 2D or 3D (Fig. S2d). Importantly, sorted cells maintained their IFITM1 high or IFITM1 low expression patterns at day 7, determined by RT-qPCR (Fig. 3a).
Furthermore, IFITM1 high CRC cells showed a higher IFITM1 expression by immunocytochemistry even after 3 or 7 days in 2D and 3D cultures, respectively (Fig. 3b). Thus, the pattern of IFITM1 expression is maintained both in short-term 2D and 3D cultures.
Previously, we proved that small EVs (sEV, with a diameter of 50-100 nm) are preferentially released from 3D cultures into the medium [17]. sEVs can be captured from the organoid supernatants by anti-CD63 or anti-CD81-coated beads that are two widely used sEV markers and the percentage of positive beads can be detected by ow cytometry [17,27]. We proved the presence of sEVs in CRC organoid-derived supernatants by antibody-coated beads, Nanoparticle Tracking Analysis (NTA), transmission electron microscopy and capillary-based immunoblot after ultracentrifugation (Fig. S3a-d). Importantly, calnexin that is not associated with sEVs [9], was detected only in cell lysates, but not in sEV preparations. In contrast, the sEV marker TSG101 was present in the ultracentrifuged pellet (Fig. S3c). We next compared the sEV release from IFITM1 high and IFITM1 low cell-derived organoids. Unexpectedly, we found no difference in the sEV concentration between the two sorted cell population-derived organoids either with the bead-based semi-quantitative method (Fig. 3c-d) or with NTA ( Fig. 3e-f). We obtained similar data when sorted cells were cultured in another matrix, collagen I or in a mixture of collagen I and Matrigel ( Fig. S4a-d). These results suggest that IFITM1 high and IFITM1 low CRC cells do not have an altered release of sEVs.

IFITM1 high CRC cells take up less EVs
Next we studied whether cell populations with different I tm1 expression have different EV uptake abilities. To address this question, we cultured cells from WT and Apc mutant mouse intestinal organoids in 2D conditions for short term and we added EVs isolated by centrifugation at 12,500g (medium EV fraction, mEV) from broblasts that had been pre-treated with a membrane labelling dye (Fig. S5a).
Importantly, we found no increase in the percentage of active caspase+ cells after labelling, showing that this treatment does not induce apoptosis in broblasts (Fig. S5b). Since EVs were not directly labelled and cells were washed after applying the dye, this precluded the possibility that we measured dye aggregates. Importantly, we detected EVs in broblast culture supernatant by NTA measurements when isolating mEVs or sEVs (Fig. S5c). Furthermore, transmission electron microscopy proved the identity of EVs in samples after ultracentrifugation (Fig. S5d). In addition, capillary based-immunoblot showed the presence of the sEV marker TSG101 only in sEV preparations, whereas we could detect calnexin in the mEV fraction (Fig. S5e) [28].
Interestingly, Apc mutant organoids that express a higher level of I tm1 accumulated less broblastderived mEVs compared to wild type intestinal organoids (Fig. S5f). To study whether EV uptake intensity differs between CRC cell subpopulations as well, we next sorted IFITM1 high and IFITM1 low cells and we cultured them with labelled mEVs or sEVs. Of note, less IFITM1 high cells took up EVs both in case of HT29 CRC cell-line-derived and human colon broblast-derived mEVs as compared to IFITM1 low CRC cells, detected by confocal microscopy (Fig. 4a-b). In addition, IFITM1 high cells that had taken up broblastderived EVs showed a lower signal intensity for mEVs than IFITM low cells (Fig. 4c). Thus, these results suggest that the difference in EV uptake is not restricted to one speci c cell type-derived EVs. Importantly, we proved the presence of mEVs and sEVs in the conditioned medium of HT29 cells with NTA and transmission electron microscopy ( Fig. S5g-h). In addition, when samples were treated with Triton X-100, the EV signal disappeared from the cells (Fig. 4d), proving that we visualized membrane-enclosed EVs in our experiments.
As the next experiment, we treated CRC organoid cells with broblast-derived labelled EVs in 2D cultures and then sorted cells with the highest and lowest uorescent signal, representing cell populations with high and low EV uptake ability, respectively (Fig. 4e). Whereas we found no difference in the expression of AXIN2 and MYC, we measured a signi cantly higher RNA level of IFITM1 in cells with low EV uptake (Fig.  4f). Since we found no increase in the expression of IFITM1 in CRC cells after treatment with EVs (Fig.   4g), this con rms again that IFITM1 high cells take up less EVs. In addition, the majority of cells with high IFITM1 level were negative for the uorescent signal after treatment with labelled broblast-derived mEVs (Fig. 4h). Furthermore, we observed the accumulation of IFITM1 low cells within the mEV+ cell population, measured by ow cytometry (Fig. 4i). When focusing on cells that had taken up mEVs, we detected a higher signal intensity, characterizing a higher mEV uptake, within the IFITM1 low cell population (Fig. 4j). Importantly, we obtained similar results when CRC organoid cells were treated with the sEV fraction collected from labelled broblasts or HT29 cells (Fig. 4i-j). Collectively, all these data prove i) the presence of cell subclones in CRC with different mEV and sEV uptake ability and ii) that IFITM1 high marks cells with a lower EV uptake.
To study whether this difference in uptake is speci c for EVs, we produced uorescently labelled synthetic liposomes and we added them to sorted IFITM1 high and IFITM1 low CRC cells after ultracentrifugation. In contrast to the labelled or unlabelled EV preparates, we detected no signal when only the uorescent dye was ultracentrifuged (Fig. S5i), thus, ruling out the possibility that we used dye aggregates in our experiments. Surprisingly, we found no difference in the signal intensity between the two cell populations after adding uorescently labelled synthetic liposomes (Fig. 4k-l). Thus, the difference in EV uptake ability between the two CRC cell subpopulations is speci c for the EVs.
Fibroblast-derived EVs result in a marked increase in the proliferating cell number of IFITM1 low CRC organoids To test the functional relevance of differential EV uptake between CRC cell subpopulations, we rst added broblast-derived mEVs or sEVs to CRC cells and we then detected IFITM1 expression and KI67 in the organoids after 7 days. Importantly, neither sEVs nor mEVs resulted in a change in the percentage of IFITM1+ CRC cells (Fig. 5a). However, we detected a markedly higher increase in the proportion of proliferating cells within the IFITM1-cell population than in IFITM1+ cells in the presence of EVs (Fig. 5b).
Furthermore, adding broblast-derived mEVs or sEVs to sorted IFITM1 low cells resulted in organoids with a markedly higher increase in the number of KI67+ cells as compared to IFITM1 high cell-derived organoids (Fig. 5c). In addition, the initial difference in the percentage of proliferating cells between the two CRC cell subpopulations completely disappeared after treatment with mEVs or sEVs (Fig. 5c). On the other hand, we detected no changes in the percentage of active caspase-3+ apoptotic cells (Fig. 5d). Thus, these results indicate that differential mEV or sEV uptake by CRC cell subpopulations critically modi es the percentage of proliferating cells in an organoid model.

Discussion
EVs can be edited and loaded with speci c molecules. Although EVs are considered as a promising tool for cancer therapy, it is still unknown which CRC cell populations take up EVs preferentially. By using patient-derived organoids, here we identify CRC subpopulations with different EV uptake capability.
Importantly, whereas CRC cells expressing the cell surface protein IFITM1 at a high level (IFITM1 high ) and at a low level (IFITM1 low ) do not differ in their EV release intensity, IFITM1 high cells take up signi cantly less EVs. Fibroblast-derived EVs result in a marked increase in the proliferating cell number of only IFITM1 low CRC organoids, showing the functional signi cance of the difference in EV uptake. We also show that I tm1 expression is increased after Apc mutation, similarly to human adenomas when compared to normal colonic epithelium. CRC patient-derived organoids have a heterogenous intraorganoid expression of IFITM1 and the IFITM1 high CRC population contains more proliferating cells.
IFITM family members have been identi ed as molecular markers in human colorectal tumors [29]. Other published studies have also connected IFITM1 to the progression of colorectal cancer, showing that it modi es the proliferating, invasive and metastatic capability of CRC cell lines [30]. However, the majority of these previous data were derived from models using cell lines cultured 2D. Importantly, our study is the rst where patient-derived organoids were used to characterize IFITM1 + CRC cells. We identi ed the existence of CRC intra-tumor cellular heterogeneity in IFITM1 expression and we observed that IFITM1 high sorted cells form organoids with more proliferating cells, suggesting that IFITM1 high and IFITM1 low subpopulations differentially contribute to CRC tumorigenesis.
IFITM genes are generally regulated by type I and type II interferons. Interestingly however, IFITM3 is constitutively expressed in tissue-resident T cells in some tissues, such as the airways and the lungs [31]. In T cells, IFITM2 and IFITM3 expressions are regulated by T-cell receptor (TCR) signal transduction [32], and hedgehog (Hh) signalling modi es the expression of IFITM genes in some systems [33]. This shows diverse regulatory mechanisms for this gene cluster. Here we prove that human organoids derived from adenomas or CRC patients express IFITM1 at a higher level compared to normal colon. By using genetically modi ed mouse intestinal organoids, we provide evidence that Apc mutation is a critical factor of the regulation of I tm1 expression. Our ndings concur with those of Lickert H et al [34] who found a regulatory effect of the Wnt pathway on I tm genes during gastrulation. In addition, we con rmed the results by Andreu P et al [35], showing the regulation of many I tm members in mouse intestinal adenoma models.
The IFITM genes encode for a group of small homologous proteins that are localized in the membranes of the endosomal-lysosomal compartment or the plasma membrane [36]. While there are seven I tm genes in mice, only ve of them have been identi ed in the human genom. However, IFITM1 is among the genes shared by both species. IFITM proteins provide a cellular resistance against a wide variety of both enveloped and non-enveloped viruses not only in vitro, but in vivo as well [37,38]. Similarly to certain viruses, EVs contain a membrane envelope. Although the mechanism of action of the IFITM proteins is largely unknown, some studies suggested that they exert their anti-viral effects via changing the uidity or other physical properties of cell membranes [39,40] which may apply for EV uptake as well. Of note, here we show that IFITM1 high cells take up less EVs compared to cells with low IFITM1 expression, however, this difference was not characteristic for synthetic liposomes. Thus, our results clearly indicate the existance of functional heterogeneity in EV, but not liposome uptake within CRC. In addition, these data provide further evidence for the relationship between certain viruses and EVs not only in the mechanism of their production, but in their uptake as well [41]. The biological importance of this differential EV uptake is illustrated by our data that broblast-derived EVs have a more pronounced stimulatory effect on cell proliferation in the IFITM1 low CRC subpopulation that accumulate more EVs.
Unlike liposomes and other synthetic drug nanoparticle carriers, EVs contain proteins in their membranes that enhance their uptake, thus, providing a more powerful delivery system of drugs or RNAs in RNA interference (RNAi)-based approaches. The usability of this approach has recently been proved by two elegant studies where RNAi was delivered to speci cally target oncogenic KRas in pancreatic tumors in mouse models [42,43]. Importantly, these results open the possibility for engineered EV-based therapies targeting oncogenes in other cancers, such as CRC. However, the proper and e cient use of EVs delivering the engineered cargo critically depends on their uptake by tumor cell subpopulations. As an important step on this way, we provide evidence for the intra-tumor cellular heterogeneity of EV uptake intensity in CRC. Interestingly, the IFITM1 high cell population with a higher proliferation potential accumulates less EVs, a fact that must be considered when designing targeting EVs for clinical applications in CRC.  The IFITM1high CRC cell-derived organoids contain more proliferating cells than IFITM1low organoids. A) The uorescence sorting strategy for IFITM1high and IFITM1low cells and their analysis after sorting with ow cytometry. B) IFITM1 protein levels of cells sorted from two CRC organoid lines (capillary-based immunoblotting). Note that actin was used as housekeeping control. C) Relative RNA levels of the indicated genes in sorted cells (RT-qPCR, n=4). D) The relative number of organoids initiated by 20,000 sorted cells. Each dot represents an individual sorting experiment. E) Representative images (left panel) and the diameter of CRC organoids (right panel) derived from IFITM1low or IFITM1high sorted cells. F)

Declarations
The percentage of KI67+ proliferating cells. Representative images and their quanti cation from confocal images (whole mount immunostaining). For E) and F), quanti cations were carried out from ten images taken from four sorting experiments. Scale bars: 50 µm (E, F). Paired t-test (C, D) and Mann-Whitney Utest (E, F) were used. ***<0.005, n.s.: p>0.05. The IFITM1high and IFITM1low CRC cell-derived organoids do not differ in their EV release. A) Relative IFITM1 RNA level of sorted cells (d0) and organoids derived from the sorted cells (d7). The RNA levels (normalized to housekeeping) of IFITM1low samples were taken as 1 in each experiment (RT-qPCR, n=4). B) Immunostaining for IFITM1 directly after sorting or after culturing in 2D or 3D conditions for the indicated time periods (immunostaining and whole-mount immunostaining, confocal microscope