The cancer angiogenesis co-culture assay: In vitro quantification of the angiogenic potential of tumoroids

The treatment response to anti-angiogenic agents varies among cancer patients and predictive biomarkers are needed to identify patients with resistant cancer or guide the choice of anti-angiogenic treatment. We present “the Cancer Angiogenesis Co-Culture (CACC) assay”, an in vitro Functional Precision Medicine assay which enables the study of tumouroid induced angiogenesis. This assay can quantify the ability of a patient-derived tumouroid to induce vascularization by measuring the induction of tube formation in a co-culture of vascular cells and tumoroids established from the primary colorectal tumour or a metastasis. Furthermore, the assay can quantify the sensitivity of patient-derived tumoroids to anti-angiogenic therapies. We observed that tube formation increased in a dose-dependent manner upon treatment with the pro-angiogenic factor vascular endothelial growth factor A (VEGF-A). When investigating the angiogenic potential of tumoroids from 12 patients we found that 9 tumoroid cultures induced a significant increase in tube formation compared to controls without tumoroids. In these 9 angiogenic tumoroid cultures the tube formation could be abolished by treatment with one or more of the investigated anti-angiogenic agents. The 3 non-angiogenic tumoroid cultures secreted VEGF-A but we observed no correlation between the amount of tube formation and tumoroid-secreted VEGF-A. Our data suggests that the CACC assay recapitulates the complexity of tumour angiogenesis, and when clinically verified, could prove a valuable tool to quantify sensitivity towards different anti-angiogenic agents.


Introduction
Impaired control of angiogenesis is involved in numerous pathological conditions and is a hallmark of cancer progression and metastasis [1][2][3]. Solid tumours are constantly influenced by the distance to nearest capillary and growth beyond a diameter of 2 mm requires able to evaluate the combined contribution of both pro-and anti-angiogenic factors secreted by the tumour [34,35]. In this study we present our novel Cancer Angiogenesis Co-Culture (CACC) Assay (Fig 1). The assay is a modification of the organotypic co-culture assay of angiogenesis developed by Bishop et al. in which human umbilical vein endothelial cells (HUVECs) are co-cultured with juvenile foreskin fibroblasts for 14 days, during which time the endothelial cells (ECs) formes luminal tubules resembling capillaries in vivo [36]. The assay recapitulates the key aspects of angiogenesis such as extension of sprouts, branching of existing structures, migration by cord tip cells and tubulogenesis [36][37][38]. The assay modifications presented in this study have allowed for co-culture of fibroblasts, vascular cells, and tumoroids established from primary colorectal tumours or metastases and thereby enabled the study of tumour induced angiogenesis.

Preparation of a VEGF-expressing SW480 cancer cell line
A VEGF-expressing SW480 cell line was generated by lentiviral transduction. Human VEGF-A 165 (VEGF) was cloned into the lentiviral expression vector pLVX-Puro (Takara Bio, Gothenburg, Sweden). The expression vector was co-transfected with pMD2.G (AddGene, Watertown, MA, USA) and psPAX2 (AddGene) into Lenti-X™ 293T packaging cells (Takara Bio) using PEI (Sigma Aldrich) at a final concentration of 1uM. The produced virus particles were harvested 48 and 72hrs post transfection and filtered through a 0.4um filter. The SW480 cells were transduced with the produced viruses and a stable population of VEGF expressing cells was isolated by selection with puromycin (2ug/ml for 48h, Thermo Fisher Scientific).

Secretion of angiogenic factors from cancer cell lines and tumoroids
The concentration of VEGF-A, bFGF and PDGF-BB was directly measured in culture supernatants using enzyme-linked immunosorbent assay (ELISA) kits following the manufacturer's recommendations. For cell line measurements R&D Systems (Minneapolis, MN, USA) and for tumoroids Thermo Fisher Scientific (VEGF and bFGF) and R&D Systems (PDGF-BB). Tumoroids were cultured in the CACC assay and supernatants collected on day 11. VEGFexpressing SW480 cells (SW480-V) and the parental line were cultured to confluence and allowed to condition the cell media for 48h. VEGF secretion was also examined qualitatively by immunofluorescence microscopy. Briefly, cells were fixed with 4% paraformaldehyde for 15 min and permeabilized with 0.5% Triton X-100 for 5 min. Cells were incubated with VEGF antibody (R&D systems) for 1h, followed by secondary Alexa-488 conjugated goat anti-mouse antibody (Invitrogen Carlsbad, CA, USA) for 45 min. Accordingly, the cells were incubated with DAPI for 5 min and mounted in Mowiol (Sigma-Aldrich) containing 2.5% 1,4-diazabicyclo[2.2.2]octane (DABCO) as an anti-bleaching agent. Images were obtained using a Leica SP5-AOBS confocal laser scanning microscope attached to a Leica DM I6000 inverted epifluorescence microscope.

Formation of cell line spheroids
SW480 and SW480-V cells (1.5 x 10 5 cells) were suspended in 3 ml of complete DMEM medium and mixed with 1.5 ml of a sterile suspension of methylcellulose (2.4% in DMEM). The suspended cells were dispensed in drops of 30 μl onto the lids of 10 cm culture dishes which were inverted in a humidified incubator for seven days. Subsequently, spheroids were harvested by suspension in PBS and gentle centrifugation at 160 x g for 90s. Next, the spheroids were resuspended in an ice-cold mixture of 30% Geltrex (Thermo Fisher Scientific) in complete EGM-2 medium (Lonza), for addition to the CACC assay. To visualize vessel formation the cells were fixed and stained with human anti-PECAM-1 antibody (R&D systems) in 1% BSA (Sigma-Aldrich) in PBS (1h, 37˚C), followed by Alkaline phosphatase-conjugated secondary antibody (Novus Biologicals, Littleton, CO, USA) diluted in 1% BSA in PBS (1h, 37˚C) and treated with the substrate SigmaFAST BCIP/NBT (5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium; Sigma-Aldrich) for 15-30 min at 37˚C. Tubes were imaged by widefield light microscopy.

Patient samples
Fresh tumour tissue was obtained from either surgically resected colonic tumours and liver metastases before routine histological processing of the specimen or by ultrasound-guided needle biopsies. The study protocols were approved by the Regional Committee on Health Research Ethics-Capital Region of Denmark (protocol no. H-1-2011-125 and H-16031106) and informed consent was obtained from all patients. Samples from primary tumours were collected at Bispebjerg Hospital, Denmark and samples from liver metastases were collected at Rigshospitalet, Denmark. Tumour tissue was placed in cold PBS supplemented with 2.5 μg/ml amphotericin B (Sigma-Aldrich) and antibiotics (500 U/ml and 500 μg/ml P/S and 100 μg/ml gentamicin (Sigma-Aldrich)) and subsequently transported to the laboratory on ice. For each patient age, gender, and TNM stage were recorded.

Establishment of patients-derived tumoroids
Patient-derived tumoroids were established as described by Jeppesen et al. [39] with minor modifications. In short, freshly resected tumour samples of 0.01-1 g were divided into 1 mm 3 pieces and digested for 20 min at 37˚C with 1 mg/ml collagenase type II (Thermo Fisher Scientific) diluted in STEM medium. Subsequently, the tissue suspension was sequentially filtered through a 70 μm cell strainer (VWR, Soeborg, Denmark) and a 30 μm pre-separation filter (MACS, Miltenyi Biotec, Bergisch Gladbach, Germany). Tissue retained by the 70 μm filter was collected and redigested for 10 min at 37˚C and passed through the filters again. This step was repeated until all tissue could pass through the 70 μm filter. Retained tumour fragments (30-70 μm) were seeded in STEM cell medium in petri dishes coated with 1.5% agarose (Sigma-Aldrich) and cultured at 37˚C in a 5% CO 2 humidified incubator. After 3-5 days the tissue fractions which had densified into rounded tumoroids with smooth surfaces could be collected for further culture. Tumoroids established from colonic tumours were denoted C after the patient number and tumoroids from liver metastases were denoted L.

Propagation and culture of patients-derived tumoroids
To increase the size and number of tumoroids per patient, tumoroids sized 30-70 μm were seeded in drops of 50 μl Matrigel matrix (VWR) diluted 1:2 with STEM cell medium in a 24-well dish. Following solidification (30 min, 37˚C), in a 5% CO 2 humidified incubator, 1 ml STEM cell medium was added to each well. After an average of 2 weeks, with intermediate medium change, the tumoroids exceeding 200 μm in size could be collected for experiments with a 200 μm pluristrainer (PluriSelect, Leipzig, Germany). Tumoroids used in this study were thawed from the 2cureX tumoroid biobank and cultured as described above before experiments.

Cancer angiogenesis co-culture assay
In the original co-culture assay of angiogenesis, established by Bishop et al. [36] and later modified by Mavria and colleagues [40], endothelial cells (HUVECs) are cultured in EGM-2 medium in co-culture with fibroblasts. Under these conditions the HUVECs form tube-like structures within the layers of fibroblasts and extra cellular matrix [36,37,40]. The Cancer Angiogenesis Co-Culture (CACC) Assay established for our studies follow the same procedure until day 7 after which the culture condition has been modified, since patient-derived tumoroids dissociate when cultured in EGM-2 medium. In short, NHDF were seeded in EGM-2 medium in 24-well plates at a density of 1.5-1.75x10 4 cells/well (batch and passage dependent density) and allowed to grow to confluence over five days with intermediate medium change. On day 5 the HUVEC cells were seeded onto the confluent layer of NHDF in EGM-2 medium at a density of 1.5x10 4 cells/well. After two days of HUVEC-NHDF co-culture (day 7) tumoroids were introduced to the assay (Fig 1). The tumoroids were isolated from the culture plates by gentle pipetting with cut-off P-1000 tips followed by filtration through a 400 μm and a 200 μm pluristrainer filter (PluriSelect). The 200-400 μm fraction of tumoroids were suspended in PBS and allowed to sediment without centrifugation. Subsequently, for each well 75 tumoroids were resuspended in 250 μl ice-cold matrix solution: Growth factor reduced (GFR) Matrigel (VWR) diluted 1:4 with reduced F12 medium (R-F12) consisting of DMEM/F12, 5% FBS, 0.45% BSA, 0.5% FAF BSA, 0.1 mM 2-mercaptoethanol, 100 U/ml and 100 μg/ml P/S, 50 μg/ml gentamicin, and 2.5 μg/ml amphotericin B. The tumoroid solution was carefully distributed on top of the HUVEC-NHDF co-culture and incubated at 37˚C for 30 min to ensure solidification of the matrix solution. Subsequently, 250 μl R-F12 medium ± treatment compounds were gently added on top of the matrix. Medium ± treatment was changed on day 9, and on day 11 the cells were fixed in 10% neutral buffered formalin containing 4% formaldehyde solution (Hounisen, Skanderborg, Denmark) for 2.5h at 4˚C with intermediate change of formalin after 30 min. To quantify vessel formation the cells were stained with human anti-PECAM-1 antibody (R&D systems) diluted in 1% BSA in PBS (1h, 37˚C) followed by Alexa Fluor 488 goat anti-mouse secondary antibody (Invitrogen, 1:500 in 1% BSA in PBS) for 1h at 37˚C and counterstained with Hoechst (Sigma, 5 μM in PBS) for 5 min at RT. Treatment compounds included VEGF-A 165 , bevacizumab, regorafenib (Selleck Chemicals, Houston, TX, USA), sorafenib (Selleck Chemicals), and sunitinib (Selleck Chemicals). Each experiment included positive controls obtained by treatment with 10 ng/ml VEGF-A 165. Microscope images (5x objective) of the wells were obtained with Cellomics Array Scanner V TI and a minimum of 8 representative images per sample were analysed with the proprietary BioSense Solutions and 2cureX tube algorithm.

Detection of cell death in the CACC assay
To investigate the induction of cell death upon treatment with anti-angiogenic compounds the fixed culture was stained with human/mouse active caspase-3 antibody (R&D systems) together with human anti-PECAM-1 antibody (R&D systems) diluted with 1% BSA and 1% goat serum in PBS (1h, 37˚C), followed by the secondary antibodies Alexa Fluor 488 goat anti-mouse (1:500) and Alexa Flour 546 Goat anti-Rabbit (Invitrogen) for 1h at 37˚C. Subsequently, the samples were counterstained with Hoechst (5 min, RT, 5 μM in PBS). Positive controls were obtained by treatment with 1 μM staurosporine (Sigma-Aldrich) for 6h prior fixation. Microscope images (10x objective) were obtained with Cellomics Array Scanner V TI and a minimum of 10 representative images per sample were quantified with a proprietary 2cureX algorithm. The expression of active caspase-3 was investigated in areas covered with HUVEC cells (Active casp-3 masked Pecam-1) and in the entire well, including areas only covered by NHDFs (total casp-3 activation).

Statistical analysis
Data are presented as mean ± standard deviation (SD) or standard error of the mean (SEM) and analysed by two-tailed unequal variance t-test using Excel (Microsoft, Seattle, CA, USA). P-values less than 0.05 were considered statistically significant.

VEGF treatment increased tube formation in the CACC assay
To verify that the HUVECs retained the ability to form tubes and be stimulated by pro-angiogenic factors in the modified culture conditions, we treated with increasing concentrations of VEGF-A. In the untreated controls HUVECs grew in short tubes with only few branches and covered less than 10% of the investigated area (Fig 2A). VEGF-A treatment resulted in a dosedependent induction of tube formation. Accordingly, addition of 10 ng/ml VEGF-A induced a two-fold increase in tube area compared to the untreated control ( Fig 2B). Treatment with VEGF-A also increased branching.

VEGF-induced tube formation was inhibited by anti-angiogenic compounds
Treatment with known anti-angiogenic inhibitors resulted in a significant inhibition of tube formation at concentrations from 10 ng/ml bevacizumab (p<0.05), 25 nM regorafenib (p<0.05), 10 nM sorafenib (p<0.001), and 10 nM sunitinib (p<0.01) when compared with VEGF-A treated samples (3.75 ng/ml VEGF-A 165 without anti-angiogenic compounds) ( Fig  3A-3C). 3.75 ng/ml VEGF resembled the level observed in the tumoroid cultures with the highest amount of VEGF-secretion (patient 11-L and 12-L, Fig 5A). In addition, we found that concentrations from 50 nM sorafenib (p<0.005) and 25 nM sunitinib (p<0.05) resulted in a significant inhibition of tube formation when compared with vehicle treated controls (no VEGF-A).

Tube formation concentrated around VEGF secreting spheroids
To investigate if angiogenic factors secreted from cancer cells grown in co-culture with HUVEC cells affect tube formation we transduced the colon cancer cell line SW480 with Human VEGF-A 165 (Fig 4). Microscopic imaging of VEGF stained cell monolayers revealed no detectable VEGF in the non-transduced cells (SW480). In contrast, the VEGF transduced SW480 cells (SW480-V) contained VEGF-positive vesicles and a diffuse pattern of VEGF staining beneath the cells was observed (Fig 4A). The SW480-V cells secreted almost 10 times more VEGF than the parental SW480 as determined by ELISA (Fig 4B). Finally, in the co-cultures we found that tube formation was directed towards the VEGF secreting spheroids (Fig 4C).
Sphere size affects the level of accessible oxygen throughout the sphere and thereby also the level of the vegfa transcription factor hypoxia-inducible factor α (HIF α) [41][42][43]. We investigated the correlation between tumoroid diameter and VEGF secretion for three tumoroid cultures. The results showed a 35-fold increase in average VEGF secretion between small (70-100um) and large (200-400um) tumoroids (S1 Fig).

Patient-derived tumoroids induced tube formation
Ten different biobank cultures of large tumoroids (200-400 μm) were assessed for their ability to induce tube formation (Table 1). We found that 7 tumoroid cultures induced a significant (p<0.05) increase of tube formation ( Fig 5A) and these were accordingly classified as angiogenic. Interestingly, the two cultures originating from the same patient (pt8) had similar increases in the level of tube formation. The highest induction of tubes occurred in patient 10-L showing a 1.43-fold increase compared to the control without tumoroids (p = 0.01). Tumoroids originating from this patient also induced a high amount of tube branching ( Fig  5B) compared to the control. Tumoroids deriving from patient 5-C, 11-L and 12-L did not result in a statistically significant increased tube formation, and such cultures were therefore classified as non-angiogenic.

Increased tube formation did not correlate with secreted levels of antiangiogenic factors
All tumoroids secreted a measurable amount of VEGF-A (Fig 5A). The concentrations ranged from 0.99 ng/ml (patient 9-L) to 3.89 ng/ml (patient 12-L) with a median value of 1.76 ng/ml. The base level of VEGF-A secreted from NHDF, HUVEC cells, and medium was less than 0.1 ng/ml in all investigated control samples. Interestingly, there was no correlation between a high level of secreted VEGF-A and induction of tube formation.
The amount of tumoroid-secreted bFGF and PDGF-BB in the angiogenic tumoroid cultures were also investigated by ELISA (S2 Fig). The bFGF concentrations ranged between 59 and 90 pg/ml, with an average of 69 pg/ml. The PDGF-BB concentrations ranged between 189 pg/ml and 302 pg/ml, with an average of 224 pg/ml. These concentrations were similar to the control samples 80pg/ml and 203pg/ml for bFGF and PDGF-BB respectively (S2 Fig). are the average of triplicate samples ± SD. � p<0.05 for treated samples vs. control. # p<0.05 for anti-angiogenic treated vs. VEGF treated samples. https://doi.org/10.1371/journal.pone.0253258.g003

PLOS ONE
The angiogenic potential of tumoroids Tumoroids were introduced to the CACC assay on day 7. At the end of the experiment (day 11) the co-cultures were fixed and stained for the endothelial marker PECAM-1 to reveal the tubes. Tubes were quantified with the BioSense Solutions and 2cureX tube algorithm. (a) Left axis: Relative tube formation (RTF) for 10 patient-derived tumoroid cultures. Tube formation is expressed relative to control sample without tumoroids. Data are the average of triplicate samples ± SD, � p<0.05 for samples with tumoroids vs. control. Right axis: Tumoroid-secreted VEGF-A (VEGF) measured in culture medium collected from the CACC assay (day 11) and assayed for VEGF by ELISA. Data are the average of triplicate samples ± SD. (b) Representative images (5x objective) of tube formation on day 11 following addition of tumoroids from patient 1x0-L (high induction of tubes), patient 6-C (high to medium induction of tubes) and patient 12-L (no induction of tubes). Images were obtained with the Array Scanner V TI and are shown with green (tubes) and red (tumoroids) overlay from the analysis algorithm. Left

Tumoroid-induced tube formation could be inhibited by bevacizumab
After addition of 1 μg/ml bevacizumab, we found that the tumoroid-induced tube formation in 8 of the 10 cultures was significantly reduced compared with samples containing only tumoroids (p<0.05) (Fig 5B and 5C). In two of the cultures (4-C and 7-C) 1 μg/ml bevacizumab reduced the tube formation to a level significantly lower than the control (no tumoroids, no treatment, p<0.05), and addition of 3 μg/ml bevacizumab resulted in tube formation below the level of the control for 7 out of 10 patients (p<0.05). For patient 11-L and 12-L a significant reduction in tube formation was not observed after addition of bevacizumab.

Tumoroid-induced tube formation could be inhibited by regorafenib, sorafenib, and sunitinib
Three of the patient-derived tumoroid cultures were chosen to investigate if other anti-angiogenic compounds than bevacizumab could reduce tumoroid-induced tube formation (Fig 6). All of the investigated TKIs induced a significant reduction in tube formation at concentrations of 50 nM and above (p<0.05) when compared with the untreated tumoroid control samples. The most potent inhibitor was sorafenib, which induced a significant reduction in tube formation even at 10 nM (p<0.05). Addition of sorafenib also resulted in tube formation below the level of untreated controls (without tumoroids) at a concentration of 25 nM for culture 6-C (p<0.01) and 10-L (p<0.005). This was observed in all cultures for sunitinib and regorafenib at 50 nM and 100 nM, respectively (p<0.05).
In addition to the ten tumoroid cultures from the 2cureX tumoroid biobank, we also established two new tumoroid cultures from liver metastases biopsies (13-L and 14-L) and investigated the effect of adding these cultures to the CACC assay. On average, 13-L induced a 1.55-fold increase in tube formation compared with the control without tumoroids (p<0.005), column: Tumoroids in control medium. Right column: Tumoroids treated with 1μg/ml bevacizumab (Bev) from day 7 of the CACC assay. Bar 200 μm. (c) Relative tube formation upon addition of patient-derived tumoroids and treatment with 1 or 3 μg/ml bevacizumab from day 7 to 11 in the CACC assay. Tube formation is expressed relative to samples with tumoroids (no treatment).  while 14-L induced a 1.79-fold increase in tube formation (p<0.05) (Fig 7). The amount of tube formation was investigated after addition of increasing concentrations of bevacizumab (0.01, 0.1, and 1 μg/ml), regorafenib (10, 50, and 100 nM) and sorafenib (10, 25 and 50 nM). Interestingly, 13-L tumoroid-induced tube formation was inhibited by less than 20% when treated with 1 μg/ml bevacizumab while tube formation was reduced by 54% in patient 14-L at the same drug concentration.

Inhibition of tube formation was not driven by endothelial cell death
None of the angiogenic inhibitors induced a significant increase in HUVEC or NHDF cell death (active casp-3) at the investigated concentrations (Fig 8). Concurrently, the positive control which was treated with the known cell death inducer staurosporine induced at more than two-fold induction of cell death in both NHDFs and HUVEC covered areas.

Discussion
In the present paper we determined the angiogenic potential and anti-angiogenic treatment sensitivity of 12 tumoroid cultures established from primary colorectal tumours or liver metastasis, thus opening the door for personalized therapy. They were all evaluated for their sensitivity towards bevacizumab and five of the cultures were also evaluated for their sensitivity towards three different TKIs. Patient-derived tumoroids and organoids are important 3D model systems since they present individual drug response profiles and can predict the response to chemotherapy with high sensitivity and specificity [39,[44][45][46][47]. Such microtumour cultures can be propagated in vitro and frozen in biobanks [44,45,48]. The main target of most anti-angiogenic agents is the tumour vasculature and accordingly the ECs. These cells have not been incorporated into previous tumoroid and organoid chemosensitivity assays, and prediction of sensitivity to antiangiogenic agents was therefore not feasible. Many 3D in vitro methods for studying tumour angiogenesis use embedding of cancer cells, cancer cell line spheroids or freshly resected cancer tissue into a matrix gel with or without addition of ECs or stromal cells such as fibroblasts and pericytes [34,35,49,50]. Such assays are adequate models of in vivo angiogenesis, but their ability to deliver rapid and quantifiable readouts are inferior since the resulting tube formation and infiltration are scattered throughout the matrix. Our CACC assay is a novel 3D in vitro model of tumour angiogenesis employing patient-derived tumoroids co-cultured with tube forming HUVECs and fibroblasts. The confluent layer of fibroblasts is only 3-5 cells in thickness and thereby enables microscopy in a single plane while concurrently allowing the ECs and tumoroids to be cultured in a microenvironment that includes stromal cells and extra cellular matrix (ECM) [38].
The Z' factor is a measure of the quality of a screening assay and ranges from -1 to 1. An ideal assay would have a Z'-factor of 1, and a screening assay is generally classified as excellent if 1 > Z' � 0.5 [51,52]. Z' incorporates the dynamic range (the difference between the means of the negative and positive control) and the separation band (the difference between the SD of the negative and positive control). Based on the data from Fig 2, we calculated a Z' factor of 0.54 for the CACC assay meaning that the assay qualifies as an excellent screening assay. An important aspect when developing a screening assay to be utilized in a clinical setting is reproducibility, speed, and cost-effectiveness. The CACC assay was developed using HUVECs as these ECs are widely available and reported to represent a robust source of primary human ECs in the organotypic co-culture assay of angiogenesis [53,54]. This is supported by the fact that different batches of HUVEC cells demonstrated the same angiogenic potential upon treatment with pro-angiogenic factors. The assay also proved highly reproducible when the same tumoroid cultures were investigated in repeated independent experiments (S3 Fig).
Another important aspect during assay development was to verify that the observed inhibition of tube formation was not mediated by inducing cell death in the ECs. We therefore investigated whether treatment with bevacizumab, regorafenib, sorafenib or sunitinib led to increased apoptosis in both fibroblasts and more importantly the ECs of the CACC assay. The results (Fig 8) show that none of the anti-angiogenic treatments resulted in significant increase in apoptosis in neither the fibroblast nor the ECs at the investigated concentrations. This supports that the treatments reduced tube formation through their anti-angiogenic properties and not by inducing apoptosis in the ECs.
The majority of our results were obtained using biobanked tumoroid cultures. To investigate whether the CACC assay results were affected by the biobanking procedure, we performed the CACC assay on two fresh tumoroid cultures (pt13-L and pt14-L) (Fig 7). The results show that the tube formation with fresh tumoroids was morphologically indistinguishable from the biobanked tumoroid cultures. The two fresh cultures had the highest observed RTF but whether this was due to the age of the tumoroid cultures is unknown. These results emphasize the value of using biobanked tumoroid cultures for assay development. Nine tumoroid cultures were classified as angiogenic and three as non-angiogenic. Tumoroid cultures presented with individual angiogenic profiles since the induced relative tube formation values ranged from 0.93 to 1.79 (Figs 5A and 7). Interestingly we observed that tumoroids deriving from the primary and metastatic lesions from the same patient (pt8) exhibited a similar angiogenic potential. It is rare to obtain samples from both primary and metastatic lesions; however, it would be interesting to investigate the angiogenic potential in more of these dual lesion patients.
The CACC assay only depicts angiogenesis but some tumours use e.g. vessel co-option to supply vessels. Bevacizumab was designed to inhibit sprouting angiogenesis and not tumour co-option of mature peritumoral vessels [31]. In mCRC up to 30% of liver metastases predominantly present with the histopathological growth pattern (HGP) termed replacement-HGP. This pattern is characterized by co-opting of the peritumoral vessels and may explain why treatment with bevacizumab does not prolong the OS of some patients [31,55,56]. A retrospective explorative study investigating the predictive value of HGP showed that the OS was significantly shorter for the cohort of bevacizumab treated patients with replacement-HGP [31]. However, HGP patterns can only be determined on resected liver tumours [56], and is therefore not a suitable biomarker for the choice of anti-angiogenic agents in patients with non-resectable tumours. Possibly, the non-angiogenic tumoroid cultures (patient 5-C, 11-L and 12-L) derive from co-opting tumours and these patients may not benefit from bevacizumab treatment.
The tube formation detected in the CACC assay represents the cumulative effect of both pro-and anti-angiogenic factors in the sample. We observed no correlation between the amount of tumoroid-secreted VEGF-A and induction of tubes (Fig 5A) confirming that VEGF-A is not the only angiogenic factor regulating the process of angiogenesis. These findings are consistent with several clinical studies showing that VEGF-A levels did not correlate with the functional effect of bevacizumab treatment [27,[57][58][59]. The ELISA assay measured the combined amount of the VEGF-A isoforms and differences between isoform levels within each tumoroid culture are therefore unknown. It would be interesting to investigate how isoform patterns differ between patients and whether this has an impact on the tube induction.
Since the non-angiogenic tumoroids (patient 5-C, 11-L, and 12-L) secreted VEGF-A, the tube formation in these samples are probably inhibited by endogenous anti-angiogenic factors.
The tumoroids from patient 13-L induced the second highest tube formation among the 12 investigated tumoroid cultures. However, the 13-L culture was largely non-sensitive to bevacizumab (Fig 7) suggesting that VEGF-A was probably not a major angiogenic driver for this tumour. Likewise, it is unlikely that the lack of response to bevacizumab was caused by tumour co-option but rather the interplay of several pro-angiogenic factors since the TKIs regorafenib and sorafenib could obliterate tube formation in this culture.
To investigate the potential effect of other angiogenic factors the amount of tumoroidsecreted bFGF and PDGF-BB were also evaluated by ELISA (S2 Fig). The measured concentrations of both bFGF and PDGF-BB for angiogenic cultures were all similar to that of the controls. These results indicate that neither bFGF nor PDGF-BB are likely to have significantly contributed to the higher tube formation observed in the angiogenic cultures. For bFGF this correlates well with published results in a similar tube formation assays where co-cultures of HUVEC and NHDF were treated with increasing concentrations of bFGF. The EC 50 value of bFGF was found to be 1400 pg/ml [60], which is more than 14 times higher than the highest bFGF concentration found in our angiogenic cultures. We only measured the soluble forms of VEGF-A, bFGF and PDGF-BB and it is unknown whether ECM bound forms are sequestered within the matrix and how they would influence tube formation.
We tested three different TKIs (regorafenib, sorafenib, sunitinib) for their ability to inhibit tube formation. For all investigated tumoroid cultures we saw a dose dependent inhibition of tube formation for each of the TKIs (Figs 6 and 7). The three TKIs tested are multi-kinase inhibitors targeting several of the kinases associated with angiogenesis [22][23][24]. As discussed above tumoroids from patient 13-L were largely insensitive to bevacizumab but showed sensitivities towards regorafenib and sorafenib which were equal to that of the other investigated tumoroid cultures (Fig 7). This once again emphasises that angiogenesis is regulated by multiple pro-angiogenic factors and suggest that treatment with more broad-spectrum kinase inhibitors in some cases could improve the overall treatment response compared to treatment with more targeted drugs such as the monoclonal antibody bevacizumab.
The present study shows that angiogenesis, as measured in our CACC assay, is a complex interplay between multiple inducers and inhibitors of tube formation and that the assay can stratify tumoroid cultures as sensitive or resistant to anti-angiogenic therapy thus potentially profile the individual patients´expected drug responses in vitro. The secretion data emphasises that measurements of single drug targets for either monoclonal antibodies or TKIs (e.g. VEGF, bFGF or PDGF-BB) cannot account for the cumulative effect of all present pro-and antiangiogenic factors.
The patients included in this study were all curatively resected, consequently, it was not possible to correlate our results with patient outcome as the patients were not offered antiangiogenic treatments. Validation of the ability of the test to predict resistance or sensitivity towards a specific anti-angiogenic treatment in the individual patient requires prospective clinical trials comparing the assay results with patient outcome and such trials will hopefully be conducted in the near future. Tumoroids were introduced to the CACC assay on day 7. At the end of the experiment (day 11) the co-cultures were fixed and stained for the endothelial marker PECAM-1 to reveal the tubes. Tubes were quantified with the BioSense Solutions and 2cureX tube algorithm. Tube formation is expressed relative to control sample without tumoroids ± SD. Data are the average of 4 independent experiments for Pt 6-C, Pt 7-C and Pt 10-L, and 3 independent experiments for Pt 13-L and 14-L, with triplicate samples in each independent experiment. (TIF)