ISOLATION AND ANALYSIS OF EXTRACELLULAR VESICLES FROM NANOFIBRILLAR CELLULOSE
3D CANCER CELL CULTURES
Heikki Kyykallio Master of Science thesis Master´s Degree Programme in Biomedicine
University of Eastern Finland Faculty of Health Sciences School of Medicine
19.5.2021
University of Eastern Finland, Faculty of Health Sciences, School of Medicine Master´s Degree Programme in Biomedicine
Heikki Kyykallio: Title of the Thesis
Master of Science thesis; 42 pages, supplementary material (3 pages)
Supervisors: Kirsi Rilla, Associate professor; Janne Capra, PhD, University of Eastern Finland
19.5.2021
Extracellular vesicles, 3D culture, nanofibrillar cellulose, hyaluronan, cancer
Abstract
The research of extracellular vesicles (EV) is heavily based on two-dimensional cell cultures although three-dimensional (3D) cell models are known to represent native tissues much more accurately. Recent research has shown that EVs from 3D cultured cells more closely represent in vivo EVs, but majority of current 3D models are not efficient in EV studies. In this work, a novel 3D culture model utilizing wood-derived nanofibrillar cellulose (NFC) was established and evaluated in study of EV secretion from MCF7 breast cancer cell spheroids. EV isolation from NFC cultures via enzymatic digestion was optimized and the impact of NFC derived material in EV analysis was assessed. Additionally, optically transparent anionic NFC (aNFC) was utilized in live cell imaging. Synthesis of abundant extracellular matrix component hyaluronan (HA) has been shown to induce shedding of EVs. NFC model was utilized to study the impact of secreted and external HA on EV secretion in 3D conditions by utilizing MCF7 cells with inducible GFP-Hyaluronan synthase 3 (HAS3) and NFC scaffolds with different HA concentrations. In addition, the impact of NFC matrix viscosity on EV secretion was analyzed.
The analysis of EV secretion, size, and marker expression was performed combining electron and confocal microscopy and nanoparticle tracking analysis. HA synthesis induced EV secretion was found to be decreased in 3D conditions. Additionally, increasing viscosity and HA content in the matrix was found to decrease EV secretion. It is shown that NFC supports efficient EV secretion and allows EV isolation and imaging of spheroid cultures.
1
Introduction
Extracellular vesicles (EV) are small lipid bilayer membrane particles which are secreted in large quantities into extracellular space by virtually all cell types and have been isolated from almost all biological fluids. The term ‘extracellular vesicle’ is generally used when referring to membrane particles secreted by a cell but EVs are in fact a highly heterogenous group of particles with varying size, origin, and molecular composition. In literature, EVs are usually divided into three main categories based on their size and biogenesis: exosomes, microvesicles, and apoptotic bodies. Exosomes (30-100 nm) originate from inside the cell as intraluminal vesicles in the lumen of multivesicular bodies and are released when the multivesicular body fuses with the cell membrane. More specifically, exosomes are formed by inward budding of the endosomal membrane of the multivesicular bodies. The biogenesis of exosomes is complex and not completely understood but it involves subunits from endosomal sorting complex required for transport (ESCRT) machinery, which acts in stepwise manner to first cluster ubiquitylated cargo to the site of biogenesis and then perform budding of the microdomain into an intraluminal vesicle (Van Niel et al, 2018). Exosome biogenesis can also be ESCRT- independent, but the mechanisms are poorly understood. Ceramides, which are formed by hydrolysis of sphingomyelins by sphingomyelinases, are proposed to form microdomains and promote membrane budding (Trajkovic et al, 2008). Tetraspanin-family of proteins are also known to directly participate to exosome biogenesis and cargo loading. Especially tetraspanins CD63, CD9 and CD81 are found to be enriched in exosomes and are commonly used as exosomal markers, although none of them are specific only for exosomes (Kowal et al, 2016).
Microvesicles (50-1000 nm) originate from the plasma membrane and are secreted via outward budding of the membrane. The biogenesis of microvesicles is also not well understood but it is thought to require changes in membrane lipid components and cytoskeletal regulation of actin and microtubules (Van Niel et al, 2018). Third main category of EVs are apoptotic bodies (1- 5 µm) which are released from the plasma membrane by blebbing during late stage of cell apoptosis and contain cytosolic molecules and cell organelles (Doyle & Wang, 2019).
Although the origins of EV groups are distinct, there are no known specific markers to distinguish them from each other. Therefore, it has been proposed that the classification of subpopulations of EVs should be divided by size into small (<100nm or <200 nm) and large particles (and >200 nm) or by their biochemical composition or cell conditions (Théry et al,
2 2018). In addition, cells secrete very heterogenous groups of EVs, and different EV isolation methods affect the subgroup of EVs collected (Tauro et al, 2012; Willms et al, 2016) making research of EVs challenging.
The research of EVs focuses on non-apoptotic vesicles as they are actively secreted by cells and carry bioactive molecules including proteins, lipids, different types of RNA, and DNA.
Other cells uptake EVs and alter their function based on the received cargo (Yáñez-Mó et al, 2015). Therefore, EVs are now considered to be an integral part of cellular signaling both in normal cell-to-cell communication and pathological states like tumor progression via their ability to modify tumor microenvironment and promote metastasis. EVs possess a great therapeutic potential as carriers of biomarkers, as their cargo resembles state of their originating cell and disease specific markers have been found from circulating patient EVs (Xu et al, 2018).
Due to their inherent ability to interact with and alter other cells, EVs are also considered ideal for drug delivery (Vader et al, 2016) and stem cell derived EVs also have potential to act as therapeutic entities (Rani et al, 2015).
Hyaluronan (HA) is a large glycosaminoglycan abundantly expressed in extracellular matrix (ECM). It regulates multitude of cellular processes including proliferation, migration, epithelial-to-mesenchymal transition, inflammation, wound healing, and cancer progression (Petrey & de la Motte, 2014; Tammi et al, 2019). In mammalians HA is synthesized from repeating units of UDP-D-glucuronic acid and UDP-N-acetyl-D-glucosamine monomers by three hyaluronan synthase isoenzymes (HAS1-3) localized on the plasma membrane. The synthases polymerize repeating monomers into a growing chain of HA which is translocated into the extracellular space through a pore structure in the HAS (Spicer & McDonald, 1998).
HA is natively a large molecule with molecular weight of 1000-8000 kDa in normal biological fluids and tissues but tumors may also secrete low molecular weight HA (Cowman et al, 2015).
Previous studies have shown that active synthesis of HA regulates membrane shape and increased synthesis by HAS2 and HAS3 induce growth of long membrane protrusions (Kultti et al, 2006) and enhances shedding of EVs both from the cell body and the tips of the formed protrusions (Rilla et al, 2013). Especially overexpression of HAS3 induces formation of long protrusions which share similar structure and protein composition with native filopodia and are dependent on ongoing HA synthesis (Koistinen et al, 2015). In addition, HAS3 overexpression via transfection has greatly increased total EV secretion in all utilized cell lines (Rilla et al, 2013; Arasu et al, 2020; Noble et al, 2020). The secreted EVs are positive for HAS3 and HA
3 receptor CD44 and are also characterized by a thick HA coat around them (Rilla et al, 2013;
Deen et al, 2016; Arasu et al, 2020). HAS3 overexpressing melanoma cell EVs were shown to activate Hedgehog signaling pathway and to induce increased proliferation and epithelial- mesenchymal transition in target cells (Arasu et al, 2020). HA coated EVs are also secreted by cells that naturally secrete high amount of HA, such as mesenchymal stem cells and cancer cells (Arasu et al, 2017; Paul et al, 2020). In addition, increased HA synthesis which occurs during epithelial-to-mesenchymal transition or wounding has also been found to increase secretion of EVs from rat mesothelial cells (Koistinen et al, 2017).
Vast majority of current in vitro EV research is based on vesicles collected from two- dimensional (2D) cell cultures, such as culturing cells on a cell culture dish as a monolayer. In monolayer cultures the cells secrete EVs directly into the culture medium, which can be easily collected to harvest all the secreted vesicles. However, 2D cell cultures are hardly an ideal model to study complex biological processes like secretion of EVs as the cells lack the characteristics of in vivo healthy or tumor tissue and its microenvironment. In 2D cultures the morphology of cells is altered and cell-cell interactions are decreased. Monolayer culture also completely lacks the interactions with the ECM which occurs in native tissue. Intercellular and cell-matrix interactions are responsible for cell differentiation, proliferation, vitality, expression of genes and proteins, responsiveness to stimuli, drug metabolism and other cellular functions (Jensen & Teng, 2020). 2D culturing cells also leads to loss of diverse phenotype and polarity of cells which is a result of changed morphology and deprived cellular interactions.
3D cultures have numerous advantages compared to monolayer cultures. The cells preserve a more natural morphology and are organized as in native tissues. The lack of constant nutrients and proper cell-cell and cell-extracellular environment interactions lead to formation of environmental niches (Edmondson et al, 2014). 3D organization also allows formation of tissue like structures such as acinar structures of epithelial tissues (Vidi et al, 2013). In 3D cultures, cancer cells form cell aggregates or spheroids which contain multiple cell layers and mimic the key features of in vivo tumors, although they lack the cellular heterogeneity of a tumor tissue.
The outer cell mass containing the proliferative cells interacts with the ECM and harbors most of the oxygen, growth factors and signaling molecules while the inner cell mass receives less oxygen and nutrients leading to proliferative inactivity and eventually cell death and formation of a necrotic core mimicking growth of in vivo tumor tissue (Khaitan et al, 2006). Three- dimensional assembly of cancer cells also affects their sensitivity to drugs as 3D cultured
4 cancer cells show higher resistance to drugs like paclitaxel, doxorubicin, dacarbazine or cisplatin (Imamura et al, 2015; Fontoura et al, 2020).
The benefits of 3D cultures in EV research have already been indicated in studies with cancer cells. For example, 3D cultured colon carcinoma cells were found to secrete two distinct populations of exosomes expressing either apical or basolateral markers (Tauro et al, 2013).
However, the current knowledge of differences in EVs secreted by 2D and 3D cultures is still limited. This is mainly due to novelty of 3D cultures in research of EVs. In addition, the approaches to study the effects of the 3D culture environment vary considerably from different culturing methods to EV analysis methods (Abdollahi, 2021). Some studies have utilized established cell lines (Rocha et al, 2019; Thippabhotla et al, 2019; Yang et al, 2020; Villasante et al, 2016) while others have concentrated on patient derived tumor cells (Szvicsek et al, 2019;
Zeöld et al, 2020) or mesenchymal stem cells (Haraszti et al, 2018). Only some studies utilizing EVs from 3D cultures have assessed differences between 2D and 3D culture derived EVs. The common observations are increased secretion of EVs in 3D cultures (Bhattacharya et al, 2020;
Rocha et al, 2019; Thippabhotla et al, 2019; Haraszti et al, 2018; Yang et al, 2020), smaller size of 3D culture derived EVs (Villasante et al, 2016; Rocha et al, 2019; Thippabhotla et al, 2019) and changes in EV cargo (Rocha et al, 2019; Thippabhotla et al, 2019; Villasante et al, 2016; Zeöld et al, 2020). However, only a handful of publications have focused on comparison of EVs from 2D and 3D conditions. Rocha et al. utilized a novel agarose microwell array method to form two gastric cancer cell aggregates and compared the secreted small EVs to small EVs from 2D cultures. Differences in cargo of 2D and 3D culture derived EVs were found as the EVs exhibited distinct microRNA (miRNA) profiles depending on the culture conditions. Overall number of detected miRNAs in 3D culture EVs was found to be increased while proteins were mainly downregulated when compared to 2D culture derived EVs. A significant downregulation was found in ADP-Ribosylation Factor 6 (ARF6) signaling pathway proteins from 3D culture derived EVs (Rocha et al, 2019). ARF6 is a small GTPase, localized in cell membranes including endosomes, and regulates vesicular trafficking, cargo sorting, membrane transport and membrane remodeling (Donaldson & Jackson, 2011). ARF6 interacts with syntenin, a highly enriched protein in exosomes which stimulates exosome production via interaction with ALIX (Friand et al, 2015). ARF6-syntenin interaction is known to regulate biogenesis of syntenin enriched exosomes (Ghossoub et al, 2014). In addition, ARF6 pathways regulate shedding of microvesicles from tumor cells (Muralidharan-Chari et
5 al, 2009) and are recently shown to be involved in trafficking of miRNA cargo to sites of microvesicle biogenesis (Clancy et al, 2019).
EVs from 3D cultured cell lines are also found to be more similar to patient derived EVs compared to EVs from monolayer cultures. Thippabhotla et al. utilized peptide hydrogel as scaffold for a spheroid culture and compared small EVs from 2D and 3D cervical cancer HeLa cell cultures with cervical cancer patient plasma derived EVs. HeLa cell spheroid derived EV miRNA profiles were significantly different compared to their parent cells in both 2D and 3D cultures, while the miRNAs between 2D and 3D cultured cells remained similar. 3D culture derived EVs shared a significantly higher small RNA profile similarity with EVs from cervical cancer patient plasma samples compared to 2D culture EVs and similarity between 3D culture derived EV miRNAs and patient EV miRNAs was nearly 96% compared to 80% similarity between 2D culture EVs and patient EVs (Thippabhotla et al, 2019). Villasante et al. compared small EVs from 3D cultured Ewing’s sarcoma cell line grown in collagen I and hyaluronan scaffolds to small EVs from monolayer cultures and patient derived plasma EVs. The mean and mode sizes and size distribution of EVs from 3D cultures was similar to EVs from patient plasma while EVs from monolayer cultures were significantly larger compared to patient derived EVs. In addition, culturing the cells as polypropylene aggregates or as monolayer on top of collagen 1 or hyaluronan matrices was not sufficient to replicate the size of patient derived EVs. Both protein and mRNA-levels of EZH2, a methyltransferase known to mediate Ewing’s sarcoma tumorigenesis, were increased in 3D culture derived EVs and corresponded with EZH2 levels of Ewing’s sarcoma patient plasma (Villasante et al, 2016).
Recent studies have also utilized patient derived tumor cell organoids as models for cancer derived EV research. The term organoid is used from three-dimensionally cultured heterogenous group of cells with organ or tissue-like assembly and function grown from pluripotent or tissue-derived adult stem cells (Kim et al, 2020) or from tumor tissue biopsy (Drost & Clevers, 2018). Tumor-derived organoids maintain the cellular heterogeneity and mutations of the tumor tissue they are derived from and therefore resemble the original tumor both phenotypically and genetically (Gao et al, 2014; Fujii et al, 2016). Studying EVs in vivo is very challenging and therefore organoids could provide an excellent model to study tumor derived EVs which closely resemble EVs from patients. So far EV release and cargo has been studied at least from colorectal adenoma and cancer (Szvicsek et al, 2019; Nagai et al, 2021) and ductal pancreatic adenocarcinoma (Zeöld et al, 2020) patient derived organoids. Organoid
6 cultures from three colorectal cancer patient tumors exhibited large variance in EV cargo between the organoid lines with only 45% of detected proteins being found in EVs from all organoid lines. It was also reported that miRNAs detected from secreted EVs differed between the organoid lines and both detected protein and miRNA content was influenced by the EV isolation method (Szvicsek et al, 2019). Mouse small intestinal organoids with Crispr-Cas9 induced Apc mutation, representing adenoma stage of intestinal cancers, displayed a massive increase in CD81-positive EVs after mutation. Similar effect was observed when Wnt pathway activation was induced by Wnt3a treatment or inhibition of GSK-3 with CHIR99021. As Apc mutation induces the Wnt pathway it was hypothesized that Apc mutation could be a factor increasing EV secretion in colorectal adenoma. In addition, no enrichment was found in Wnt activated genes and genes known to function in exosomal biogenesis with gene set enrichment analysis from publicly available data (Szvicsek et al, 2019). Five organoid lines from ductal pancreatic adenocarcinoma tumors secreted varying amounts of CD63-positive EVs and differed from their EV miRNA cargo. When compared to corresponding patient plasma derived EVs, small set of miRNAs (2.2-15.8%) was detected only in organoid EVs while 14-35% of detected miRNAs were shared between organoid and plasma EVs. However, majority of miRNAs detected in organoids were present in plasma of the patient (75.6 ± 9.7%). Large heterogeneity between individual patients resulted in a total of 8 miRNAs to be found to be present in all organoids and plasma samples. Two miRNAs (miR-21 and miR-195) were found to be increased in ductal pancreatic adenocarcinoma patients compared to control donors, although the corresponding miRNA levels were found to be similar in patients with chronic pancreatitis (Zeöld et al, 2020).
Current knowledge strongly supports the hypothesis that three-dimensional assembly is required for release of in vivo like EVs. However, there is a lack of standardized methods for 3D culture EV studies. Present studies have utilized a variety of isolation methods and types of 3D cultures. Some studies have utilized scaffold free 3D culture methods in form of low adhesion plates (Sadovska et al, 2018), agarose microwell arrays (Rocha et al, 2019), nanoculture plates (Eguchi et al, 2018; Taha et al, 2020) and for mesenchymal stem cells also microcarriers (Haraszti et al, 2018). Alternatively, various biological scaffold or hydrogel based models with Matrigel (Szvicsek et al, 2019; Zeöld et al, 2020), collagen I (Szvicsek et al, 2019; Zeöld et al, 2020; Yang et al, 2020), collagen I and hyaluronic acid (Villasante et al, 2016) and peptide hydrogel (Thippabhotla et al, 2019) have been used to model the ECM.
7 Biological scaffolds are biocompatible and porous materials which provide cells an in vivo-like ECM environment while hydrogels are composed of cross-linked polymeric material and exhibit a tissue-like stiffness which resembles natural ECM (Caliari & Burdick, 2016). While scaffold free cultures allow collection of all secreted EVs from the conditioned medium, many 3D culture materials have a gel-like structure which prevents isolation of EVs from inside the scaffold. In addition, it is not possible to dissolve all 3D scaffolds or hydrogels efficiently and without disturbing cellular material to retrieve the EVs or cells. Therefore, EVs can only be collected from the conditioned medium on top of the scaffold which leaves a fraction of EVs trapped inside the scaffold. Biologically derived scaffolds such as Matrigel are widely used in 3D cultures due to their high biocompatibility and in vivo ECM-like composition. However, the composition of such scaffolds is undefined and can vary between batches causing variability in experimental results (Hughes et al, 2010). Biological scaffold materials may also contain matrix bound vesicles which are functional EVs in connective tissue and shown to be present also in commercially available biological scaffolds (Huleihel et al, 2016) and could act as contaminating material in EV studies.
To overcome the issues of biological scaffolds, various synthetic or animal and human component-free materials have been introduced as alternatives. Some promising 3D culture scaffolds are nanocellulose hydrogels. Nanocellulose refers to cellulose material with one dimension being in nanometer range. It can be isolated from cellulose rich plant material such as wood, flax, cotton or hemp enzymatically, chemically, and mechanically with different production methods leading to formation of cellulose nanofibrils or nanocrystals. Alternatively, it can be produced by bacteria (bacterial nanocellulose) (Klemm et al, 2011). Typically, the diameter of nanocellulose units ranges between 5-100 nm and the length depends on the type of nanocellulose ranging from 20-100 nm for cellulose nanocrystals to several micrometers for fibrillated nanocellulose. Meanwhile bacterial nanocellulose exhibits a net-like structure of nanofibrils (Curvello et al, 2019). Bacterial and fibrillar nanocellulose form hydrogels in aqueous solutions. In addition, native nanocellulose is considered to have a low risk of cytotoxicity, making it ideal for in vitro and in vivo biomedical applications (Alexandrescu et al, 2013).
Nanofibrillar cellulose (NFC), also referred as nanofibrillated, microfibrillar, microfibrillated cellulose or cellulose nanofibrils, is typically produced from wood material with mechanical shearing in combination with pre-treatment by enzymatic hydrolysis (Pääkko et al, 2007),
8 TEMPO-mediated oxidation (Saito et al, 2006), carboxymethylation, or quaternization (Aulin et al, 2010) followed by homogenization. This results in long cellulose fibers with nanoscale diameter. NFC forms hydrogels already in low concentrations (0.1-0.2 wt.%) in water (Pääkko et al, 2007; Bhattacharya et al, 2012). NFC hydrogel was found to be optimal for 3D cell culturing as it exhibited similar viscoelastic properties with biological 3D scaffolds, ECM-like diffusion, and high biocompatibility with 3D HepaRG and HepG2 liver derived cell lines (Bhattacharya et al, 2012). Since, it has been utilized in multiple 3D culture applications including differentiation of human liver progenitor cells to organotypic cultures (Malinen et al, 2014), formation of spheroids of human pluripotent stem cells (Lou et al, 2015), maturation of human-induced pluripotent stem cell derived hepatocyte-like cells (Toivonen et al, 2016) and formation of breast cancer cell spheroids (Barnawi et al, 2019). In addition to providing optimal properties for ECM mimicking scaffold for 3D cell culture, NFC is compatible with imaging as it does not emit autofluorescence (Bhattacharya et al, 2012). More recently, anionic NFC (aNFC) with enhanced imaging properties was shown to support growth of human adipose tissue-derived stem cells (Sheard et al, 2019) and their osteogenic differentiation via electric stimulation (Bicer et al, 2020). Besides 3D cell culture, NFC has applications in wound dressing (Koivuniemi et al, 2020), 3D bioprinting (Stanco et al, 2020), and drug delivery (Paukkonen et al, 2017).
In this work, wood-derived NFC hydrogel was utilized as a 3D scaffold for breast cancer spheroid culture and for the first time in isolation of EVs secreted by the spheroids. Prior to this work, cellulose has been utilized in EV research only in semipermeable cellulose sulphate capsules which allowed small EVs from capsulated mesenchymal stem cells to be collected from culture medium (Zavala et al, 2020). In this work a 3D spheroid culture model in NFC was established and a workflow for isolation of EVs from NFC 3D cultures was optimized.
The concentration of the NFC hydrogel was altered to modify the viscosity of the matrix to mimic different in vivo extracellular environments. To isolate the cells and secreted EVs from the matrix, NFC was digested with commercially available optimized cellulase enzyme.
Cellulase enzymes degrade the nanofibrillar cellulose fibers into soluble glucose which allows recovery of cells and EVs from the culture efficiently and without disturbing the cells and cellular structures. The isolated particles were analyzed and characterized by scanning and transmission electron microscopy and by confocal imaging using fluorescent antibodies of classical EV markers. In addition, the established 3D model was used to study how HA
9 secretion, ECM viscosity and high molecular weight HA content in the matrix affect EV secretion and size of secreted EVs in 3D conditions using nanoparticle tracking analysis (NTA). Additionally, optically transparent aNFC (Sheard et al, 2019) was utilized in live cell imaging of the cultures. MCF7 breast cancer cell line with inducible GFP-HAS3 expression was used as a cell model. The cell line allows induction of HAS3 overexpression and increased HA secretion in cells which normally secrete very low amounts of HA (Deen et al, 2016).
MCF7 cells formed spheroids and efficiently secreted EVs in NFC cultures. Digestion of NFC resulted in small particles which were detected with NTA but were much lower in number compared to isolated EVs. GFP-HAS3 mediated induction of HA synthesis increased EV secretion from the spheroids compared to non-induced spheroids but the effect in 3D conditions was not as prominent as in previous reports from 2D cultures. Increased viscosity by increasing the NFC concentration of the matrix slightly decreased EV secretion but did not affect the size of secreted EVs. Mixing high molecular weight HA in the NFC resulted in decreased cell number and EV secretion by cell as well as changes in size distribution of isolated EVs. The established 3D model system allows efficient secretion and isolation of EVs from 3D cultured cells. It also demonstrated how physical properties and composition of the extracellular environment may regulate secretion of EVs.
Materials and methods
Cell culture
Creation of MCF7 human breast cancer cell line with stable doxycycline-inducible GFP- Hyaluronan synthase 3 (HAS3) expression is previously described in (Siiskonen et al, 2013).
MCF7-GFP-HAS3 cell line was cultured in minimum essential medium alpha (MEMα, EuroClone, Pavia, Italy) supplemented with 5% fetal bovine serum (FBS) (Gibco, Thermo Fischer Scientific, Waltham, MA, USA), 2 mM glutamine (EuroClone), 50 µg/ml streptomycin sulfate and 50 U/ml penicillin (BioWhittaker, Lonza, Basel, Switzerland) and maintained with 50 µg/ml hygromycin (Invitrogen). Cells were passaged twice a week with a 1:5 split ratio using 0.05% trypsin (w/v) 0.02% EDTA (w/v) (Biochrom AG, Berlin, Germany). During experiments the cells were cultured in minimum essential medium alpha (EuroClone), supplemented with 5% EV-depleted FBS (Gibco), 2 mM glutamine (EuroClone), 50 µg/ml streptomycin sulfate and 50 U/ml penicillin (EuroClone).
10 EV-depleted FBS used for the experiments was prepared by ultracentrifugation at 110 000 x g for 16 h. The supernatant was then collected and sterile filtered with 0.22µm syringe filters (Guangzhou Jet Biofil, Guangzhou, China).
Poly-HEMA coating of culture wells
To prevent cell adhesion, wells of 48-well-plates were coated with Poly(2-hydroxyethyl methacrylate) (Poly-HEMA) followingly: 1.2% (w/v) Poly-HEMA (Sigma-Aldrich, St. Louis, MO, USA) was dissolved into 95% EtOH overnight at 65° C. The solution was then filtered with 0.22 µm filter (Guangzhou Jet Biofil). The culture wells were coated with the solution which was then allowed to evaporate inside laminar flow hood overnight. Coating was repeated on the next day. Before use, the wells were washed 3 times with PBS.
Nanofibrillar cellulose scaffolds
For the 3D cultures for EV isolation, 1.5% (w/v) GrowDex® nanofibrillar cellulose hydrogel (NFC) (UPM Biomedicals, Helsinki, Finland) was diluted with experiment medium into working concentrations between 0.5% and 0.9% (w/v). For confocal and scanning electron microscopy, 1% (w/v) GrowDex®-T anionic nanofibrillar cellulose hydrogel (aNFC) (UPM Biomedicals) was diluted with experiment medium into working concentrations between 0.33% and 0.5% (w/v).
NFC-Hyaluronan scaffolds were prepared by first diluting sodium hyaluronate of a size of 2MDa (Lifecore Biomedical, Chaska, MN, USA) with experiment medium into 1% and 2%
(w/v) concentrations. 1.5% (w/v) NFC was then added into the diluted hyaluronan solution to dilute the NFC into concentration of 0.5% (w/v).
3D cultures
After preparation of the NFC scaffolds, 250 000 cells/ml of prepared NFC scaffold were mixed and 300 µl of the NFC-cell mixture was added per well on a 48-well poly-HEMA coated plate.
300 µl of experiment medium was then added on top of the hydrogel. For live cell imaging 250 000 cells/ml of prepared aNFC were mixed and 100 µl of the mixture was added on a 16- well coverglass (Culturewell™ chamberSLIP 16, Grace Bio-Labs, Bend, OR, USA) per each well. Cultures were then grown in 37° C, 5% CO2 for 7 days to allow formation of spheroids.
The medium on top of the hydrogels was changed twice (on days 2 and 5) during the experiment. For induction of GFP-HAS3 overexpression, 0.5 µg/ml doxycycline (doxycycline
11 hydrochloride, Sigma-Aldrich) was added during the medium change. For live cell imaging, cell nuclei were stained with NucBlue live nuclear stain (Thermo Fischer)
EV isolation
NFC scaffold of the spheroid cultures was digested with 600 µg/mg (µg enzyme/mg cellulose) GrowDase® cellulase enzyme mix (UPM Biomedicals) by incubating in 37° C, 5% CO2 for 9h to 24h (specified in results). After digestion, the remaining culture medium was centrifuged at 600 x g for 10 min to pellet the spheroids, and the supernatant containing the EVs was collected. To calculate the number of cells from spheroids, the spheroid pellet was suspended into 0.5 ml of 0.05% trypsin and incubated in 37° C for ~15 min until the cells were detached.
The collected supernatant was centrifuged at 5000 x g for 10 min to remove cell debris and non-digested cellulose fibres and the pellet was discarded. The optional filtering step was performed with Minisart 5.0 µm syringe filters (Sartorius, Goettingen, Germany). EVs were isolated by ultracentrifugation at 189 000 x g for 90 min at 4° C. After centrifugation, the pellets were suspended into PBS which had been sterile filtered with 0.22µm syringe filters (Guangzhou Jet Biofil).
Nanoparticle tracking analysis
Isolated EV concentration and size distribution were measured using a nanoparticle tracking analyzer (NTA) (Malvern Instruments Ltd. Malvern, UK) with an NS300 view unit. EV concentration and size were analyzed from four replicated measurements with following settings for data acquisition: camera level 13, acquisition time 30 s and detection threshold 3.
NTA 3.1 Software (Nanosight, Amesbury, UK) was used for the analysis.
Immunofluorescent staining of EVs
Chambered Ibidi coverglass (Ibidi GmbH, Martinsried, Germany) was coated with 10 µg/ml Poly-D-lysine (Sigma-Aldrich) diluted in sterile H2O by adding 200 µl of Poly-D-lysine into each well under cell culture laminar hood. The coverglass was incubated at 37°C in 5% CO2, overnight. On the next day, Poly-D-lysine was removed and 200 µl of isolated EVs suspended in and diluted 1:2 with sterile filtered PBS was added per well and incubated at 37° C in 5%
CO2 for 3 h to allow the EVs to attach to the Poly-D-lysine coating. After the incubation, the PBS was carefully removed from the well. Alexa Fluor 594 anti-human CD63 antibody (#353033; BioLegend, San Diego, CA, USA) and FITC anti-human CD9 antibody (#312103;
12 BioLegend) were diluted 1:200 with 1% BSA in phosphate buffer and 200 µl of diluted antibodies was added per well and incubated for 2 h, RT in the dark before imaging.
Confocal imaging and image processing
Confocal imaging of spheroid cultures and isolated EVs was performed with Zeiss Axio Observer microscope (40 x NA 1.3 oil and 63 x NA 1.4 oil objectives) equipped with LSM800 confocal module (Carl Zeiss Microimaging GmbH, Jena, Germany). ZEN v2.5 Blue software (Carl Zeiss Microimaging GmbH) and ImageJ software (National Institute of Health, Bethesda, MD, USA) were utilized for image processing and 3D rendering.
Scanning electron microscopy
For microscopy of isolated EVs, 13 mm cover glasses were coated with Poly-D-Lysin (Sigma- Aldrich) overnight and washed 3 times with PBS. EV preparations were added on the cover glasses and incubated in 4° C overnight. The EV preparations were fixed with 2%
Glutaraldehyde and 1% Osmium tetroxide and routinely dehydrated in ascending series of ethanol before drying with hexamethyldisilazane and coating with chromium.
For microscopy of spheroid cultures, cells were grown in 0.5% aNFC for 7 days. The aNFC scaffold containing the spheroids was collected and fixed with 2.5% glutaraldehyde in 4° C overnight. Pieces of fixed scaffold were routinely dehydrated with ascending series of ethanol and dried with hexamethyldisilazane. The samples were attached to a carbon tape and coated with gold. Imaging was done using a Carl Zeiss Sigma HD VP scanning electron microscope (Carl Zeiss NTS, Cambridge, UK) operated at 5 kV.
Transmission electron microscopy
The EV preparations were layered onto carbon-coated glow-discharged copper grids. Grids were then fixed in 2% paraformaldehyde for 10 min and contrasted using 2% neutral uranyl acetate for 15 min in the dark. Grids were embedded in 1.8% methyl cellulose (25 Ctp)/0.4%
uranyl acetate. Samples were imaged with a JEOL JEM 2100F transmission electron microscope (Jeol Ltd, Tokyo, Japan) operated at 200 kV.
Hyaluronan assay
Hyaluronan secretion of the 7-day spheroid cultures was measured from a 100 µl sample of EV isolate medium collected before ultracentrifugation. The hyaluronan-specific probe HABR
13 (Wang et al, 1996) and biotinylated hyaluronan bHABC (Hiltunen et al. 2002) were prepared as described previously. Maxisorp 96-well plates (Thermo Fischer) were coated with 1 µg/ml HABR in 50 mM sodium carbonate buffer (pH 9.5) for 2 h at 37° C. The plates were then washed with 0.5% Tween-PBS and blocked with 1% BSA-PBS for 1 h 37° C. Plates were washed with 0.5% Tween-PBS and stored in -20° C overnight. Standard hyaluronan (2.5-50 ng/ml) and medium samples were applied to the plate and incubated for 1h at 37° C. The plated were washed with 0.5% Tween-PBS and incubated with 1µg/ml bHABC in 1% BSA-PBS for 1 h at 37° C and washed. Horseradish peroxidase-streptavidin (#SA-5004, Vector Laboratories, Burlingame, CA, USA) was diluted 1:20 000 with PBS and added to the wells, incubated at 37°C for 1 h, and washed with 0.5% Tween-PBS. For colour reaction, substrate solution was prepared just before use. For the substrate solution, 0.5% 3,3’,5,5’-tetramethylbenzidine in DMSO was diluted into 0.02% solution with substrate buffer (0,1 sodium-acetate (trihydrate), 1,5 mM citric acid (monohydrate), 0,005 % H2O2, pH 6.0). The substrate solution was added to the plate and the plate was incubated RT in the dark until sufficient colour developed (~12 min). The reaction was stopped with 2 M H2SO4 and the plate was analysed with iEMS Reader MF plate reader (Labsystems, Helsinki, Finland) and absorbances were read at 450 nm. The concentration of hyaluronan was calculated using a linear standard curve and normalized to cell count.
Statistical analysis
Statistical analyses were performed using GraphPad Prism Software v.5.00 for Windows, (GraphPad, San Diego, CA, USA). Differences were considered significant when p < 0.05.
Results
Nanofibrillar cellulose 3D culture model for EV isolation and analysis
Nanofibrillar cellulose hydrogel (NFC) was utilized in creation of 3D cancer cell spheroid model (Fig. 1). In this model the spheroids were formed from single cells which were mixed into the NFC scaffold. The cells secreted EVs into the surrounding matrix and the secreted particles were trapped inside the scaffold. The spheroids and secreted EVs were isolated by digesting the NFC matrix enzymatically, which digests nanofibrillar cellulose fibers into soluble glucose without disturbing cell structures, allowing recovery of spheroids and EVs.
Digesting the matrix allows recovery of all secreted EVs, not only the EVs that have escaped
14 from the scaffold in the culture medium on top of the matrix. After the digestion the spheroids were easily recovered by slow centrifugation and the number of cells was calculated from the culture for normalization of secreted EVs by cell. Secreted EVs were isolated from the remaining supernatant. In this work, differential centrifugation was used for EV isolation, but other isolation methods are also compatible with the model. After isolation, EVs can be used for characterization and analysis. NFC is also compatible with imaging which offers possibilities to monitor spheroid growth and morphology and possibly even secretion and uptake dynamics of EVs. In this work the focus was on overall imaging properties of NFC hydrogels in the context of imaging EVs.
Figure 1. Schematic overview of the NFC 3D culture model. The figure represents the main steps of cell culture and EV analysis. The viscosity of NFC is first controlled by diluting the hydrogel into wanted concentration with culture medium. The cells are mixed in the hydrogel and the mixture is moved in culture wells. Wells are coated to prevent cell adhesion and ensure that all the cells are in 3D form. Culture medium is added on top of the hydrogel and the cells are grown allowing spheroid formation. To isolate the EVs secreted by the cells, the NFC matrix is digested into soluble glucose with cellulase enzymes. Spheroids are removed and used for cell counting and the EVs are isolated with differential centrifugation. Isolated EVs can then be analyzed with nanoparticle tracking analysis and imaging. NFC also allows imaging of both live and fixed cell cultures.
NFC supports the growth of GFP-HAS3 inducible MCF7 cell spheroids
Hyaluronan synthase 3 (HAS3) inducible MCF7 breast cancer cell line (MFC7-GFP-HAS3) was chosen as the cell model for the study. The cell line has a stable transfection of GFP-HAS3 which is only expressed when induced using doxycycline. The induction leads to overexpression of HAS3 resulting in greatly increased synthesis and secretion of hyaluronan (HA) which is known to enhance secretion of EVs (Rilla et al, 2013). For NFC 3D cultures the
15 cells were mixed in the NFC and grown for 7 days. During the culture time the cells formed spheroids with consistent size and morphology (Fig. 2A). The spheroid growth in NFC was quite slow, with 7-day cultures reaching a spheroid diameter of approximately 50 µm. Slower growth is most likely due to the lack of growth factors and ECM proteins that traditional ECM mimicking 3D scaffolds contain. The spheroid nuclei were stained to assess cell viability in spheroids. Cells showed high viability and cell division especially on the outer layer of the spheroids (Fig. 2D). No necrotic cores were formed during the 7-day culture.
To increase EV secretion of the spheroids the expression of transfected GFP-HAS3 was induced with doxycycline. The induction led to strong expression of GFP-HAS3 on the cell membrane of the MCF7 cells and formation of long filopodia-like protrusions. The high expression of GFP-HAS3 allowed visualization of the induced MCF7-spheroids in live cell cultures (Fig. 1C-E). In addition, a high number of GFP-HAS3 positive extracellular vesicles were detected around the spheroids.
The morphology of the GFP-HAS3 induced spheroids growing inside the cellulose matrix was further investigated with scanning electron microscopy (Fig 2B). For this purpose, aNFC spheroid culture was fixed and pieces of the scaffold were prepared for microscopy.
Morphology of the spheroids was well preserved, although some cell shrinking was detected when comparing cell size to live cells imaged with confocal microscopy. Long protrusions detected with confocal microscopy in live cultures were much less prominent, but areas of membrane blebs and ruffling were detected. In addition to membrane ruffles, multiple round vesicle-like structures of various size were found to be attached on the surface of the cells.
16
Figure 2. MCF7-GFP-HAS3 spheroids in NFC and aNFC cultures. MCF7 cells grow into spheroids during 7 days of culture in 0.5% NFC (A). SEM image of a GFP-HAS3 induced spheroid growing in 0.5% aNFC (B) shows membrane blebbing and several vesicular structures being attached on the surface of the spheroid. 3D rendering of a live MCF7-GFP-HAS3 spheroid in 0.33% aNFC stained with NucBlue live nuclear stain (C, D, E).
The cells grow long protrusions and are highly positive for GFP-HAS3 as previously described in literature and secrete EVs into the surrounding NFC matrix. The cells show high viability and cell division on the outer spheroid layer (D).
Characterization of isolated particles
To assess whether cellulase enzyme digestion of NFC leaves scaffold derived particles in the isolates, 0.5% NFC scaffolds without cells were digested and put through the isolation protocol and particle concentrations were measured with NTA. As expected, NFC derived particles were detected in the isolates. The concentration of particles varied between repeated experiments.
Most of the concentrations were set between 1e8 – 5e8, but individual samples could also reach up to 2e9 particles per ml. This kind of variation made it challenging to assess the average
17 concentration of NFC derived particles after isolation and compare changes in digestion efficiency.
To enhance the digestion of the NFC, the recommended concentration of cellulase enzyme was doubled and the effect of digestion time from 9 hours to 24 hours was compared (Fig. 3A).
Longer digestion time resulted in slightly lower particle concentrations on average but even with the longer digestion time some individual samples remained with higher concentration of particles. To assess how the concentration of NFC affects particles detected in NTA, 0.7% and 0.9% NFC scaffolds were digested for 24 h and after isolation the particle concentration was measured (Fig. 3B). Increasing the NFC concentration from 0.5% to 0.7% did not increase the number of recovered particles. However, with the 0.9% concentration the number of particles increased considerably. In addition, the number of particles detected from 0.9% NFC scaffolds displayed higher variation and the samples were more likely to contain high numbers of particles.
To investigate whether the purity of EV samples could be increased by filtering the samples before ultracentrifugation, samples from 0.5% NFC cell culture and samples from empty 0.5%
NFC scaffolds were filtered as a part of isolation process. Pore size of 5.0 µm was used to minimize the loss of EVs and only filter out larger particles. Filtering step was done either 1) immediately after NFC digestion and removal of spheroids or 2) after 5000 x g centrifugation.
The particles were isolated with ultracentrifugation and the number of recovered particles was measured with NTA and compared with non-filtered samples (Fig. 3C). The point of filtration greatly affected especially the recovery of EVs with samples filtered after 5000 x g centrifugation containing over 2 times more particles compared to samples filtered after digestion. The loss of EVs was on average 65% for after digestion filtering and only 30% for after 5000 x g centrifugation filtering. Decreased particle numbers were also detected in samples with empty NFC. Filtering after digestion reduced particles by 55% and after 5000 x g centrifugation by 75%. Therefore, the ratio of NFC particles in EV samples was 11% for non- filtered, 17% after digestion filtering and 4% after 5000 x g centrifugation filtering, on average.
18
Figure 3. The effect of digestion time, NFC concentration and sample filtering to particles detected with NTA. (A) Longer NFC digestion time decreased the number of NFC derived particles but overall variation between samples was high with both digestion times. (B) High concentration of NFC resulted in increased number of NFC derived particles. (C) Comparison of isolated particle concentrations with different filtering steps from cell culture derived (EV) and empty NFC scaffold derived (NFC) samples. N=3-6, SEM.
To investigate if the NFC derived particles can be distinguished from EVs and whether filtering has an effect on the isolated particles, non-filtered and filtered isolates derived from NFC cell cultures and empty NFC scaffolds were compared with scanning electron microscopy (Fig. 4).
The cell culture derived samples were characterized with a large amount of heterogeneously sized particles, most of which had a round vesicle-like morphology. Interestingly, the filtered sample had an overall cleaner appearance with most of the particles having a smooth surface.
In contrast, some particles from the non-filtered sample seemed to have fiber-like particles attached to them. The same effect was visible when non-filtered and filtered samples from empty NFC scaffolds were compared. Non-filtered sample was characterized by larger structures with fiber like appearance. Filtered sample had much less particles which were also smaller in size. In addition, the fiber-like structures were again absent in the filtered sample, indicating that filtering had an effect on larger NFC derived particles. When comparing the overall number of particles between cell culture derived and empty NFC scaffold derived
19 samples, the cell culture samples were clearly much abundant in particles which also were much more heterogenous in size. Especially the non-filtered cell culture derived sample had a large amount of very small particles which were completely absent in empty NFC scaffold derived samples. On the other hand, the empty NFC scaffold derived samples contained some small round particles which could not be distinguished from EVs with scanning electron microscopy.
Figure 4. Comparison of non-filtered and filtered samples with scanning electron microscopy. Filtering samples reduced both the number of EVs and NFC derived particles. Samples were compared with scanning electron microscopy to characterize vesicles from NFC derived particles. EV samples contained a high number of small particles as well as larger round vesicle-like structures (arrows in EV and EV filtered). Filtering resulted in smoother appearance and reduced number of particles. Fiber-like structures found in non-filtered samples (NFC, arrow) were absent in filtered samples. Scale bar 1 µm.
To ensure that the isolated material is indeed EVs, the isolates were characterized with transmission electron microscopy (TEM). Isolates from non-induced and GFP-HAS3 induced cultures were compared. Both samples were characterized by a high number of particles with varying size and morphology. EVs with both round and collapsed cup-shaped morphology
20 (Fig. 5) as typically seen in EV samples were found from both non-induced and induced samples. No difference in EV number or general morphology was detected between non- induced and GFP-HAS3 induced EVs. Overall, larger EVs were rather rare which is, however, typical for TEM samples. Both isolates contained a high number of smaller particles most of which could not be recognized as EVs. However, this is mostly due to imaging quality. In addition, as seen from the scanning electron microscopy samples, the isolates contain many very small particles, and it is likely that the high particle concentration affects the imaging quality. Additionally, the smallest EVs only have a diameter of ~30 nm which makes imaging even more challenging. One more challenge are the NFC derived particles. As most of the smaller particles were not identified, it was not possible to estimate the purity of the samples by TEM.
Figure 5. NFC 3D culture derived EVs show typical morphological characteristics. Transmission electron microscopy images of EVs isolated from non-induced (a) and GFP-HAS3 induced (b) spheroid cultures. EVs of different size were found from both preparates and several vesicles having the typical collapsed cup-shaped morphology (arrows) were detected. Scale bar represents 200 nm.
EVs isolated from NFC are characterized by typical EV markers
For further characterization of the isolated EVs, ultracentrifuged isolates were stained against tetraspanin antibodies and imaged with confocal microscope. Antibody staining was performed for tetraspanins CD63 and CD9 which are commonly referred as classical EV markers and are widely used in characterization of EVs (Kowal et al, 2016). EVs from both non-induced and GFP-HAS3 induced culture isolates were allowed to attach on poly-D-Lysine coated coverglass and were stained before imaging (Fig 6). EVs from non-induced cultures were
21 double stained with CD63 and CD9 antibodies (Fig. 6A). The stained EVs abundantly expressed CD63 and CD9 and were heterogenous in size, although the diameter of an individual stained particle did not exceed 1 µm. The number CD9 positive EVs was slightly higher than the number of EVs positive for CD63, but the markers also seemed to colocalize in majority of the EVs. Some EVs only positive for either CD63 or CD9 were also detected.
For characterization of EVs from GFP-HAS3 induced cultures the EV samples were stained with CD63 antibody. The total amount of detected CD63 positive EVs was clearly higher compared to EV samples from non-induced cultures. The number of GFP-HAS3 positive particles was distinctly lower compared to particles positive for CD63 indicating that only a fraction of EVs secreted by GFP-HAS3 induced cells were carrying GFP-HAS3. Majority of particles that were positive for GFP-HAS3 were also positive for CD63.
Figure 6. EVs isolated from NFC 3D culture are positive for classical EV markers. EVs from non-induced (A) and GFP-HAS3 induced cultures (B) stained with CD63 and CD9 fluorescent antibody stains. Area marjed with a smaller square is shown magnified on the top right of each image. EVs from non-induced cultures (A) were widely positive for both CD63 and CD9 which were also colocalized in majority of the particles (A, arrows). EVs from GFP-HAS3 induced cultures (B) showed increased number of CD63-positive EVs when compared to non- induced EVs. EVs positive for GFP-HAS3 were detected but were lower in number compared to EVs positive for CD63. GFP-HAS3 expressing particles were also typically positive for CD63 (B, arrows). Scale bars: 5 µm, 2 µm (magnified areas).
22 GFP-HAS3 induction increases EV secretion in 3D culture
Increased synthesis of HA has been shown to enhance EV shedding and especially overexpression of HAS3 leads to increased EV secretion in monolayer cultures (Kultti et al, 2006; Arasu et al, 2020). However, due to the limitations in 3D culture models the HAS induced secretion of EVs has not been previously studied in 3D conditions. To find out whether HAS3 overexpression increases total EV secretion in 3D culture, EVs from non-induced and GFP-HAS3 induced 0.5% NFC cultures were compared with nanoparticle tracking analysis.
When total particle concentrations were compared, the GFP-HAS3 induced spheroids secreted on average 40% more EVs compared to non-induced spheroids (Fig. 7B). However, when the particle concentrations were normalized to corresponding cell counts the difference was only about 23% (Fig. 7C). The secreted particle numbers per cell were very high, 27 000 for non- induced and 33 000 for GFP-HAS3 cells, indicating that 3D conditions support efficient secretion of EVs. However, the particle number per cell varied highly between experiments.
This variation seemed to be more of a result of variation in cell number than in particle concentration which remained relatively similar between experiments. There were no major differences between size distribution of non-induced and GFP-HAS3 induced EVs, although the mean diameter of non-induced EVs was slightly larger compared to the GFP-HAS3 induced EVs (Fig. 7D -E).
To compare activity of HA synthesis between non-induced and GFP-HAS3 induced cultures, total amount of HA was measured from medium of the spheroid cultures after digestion of the NFC matrix. The amount of secreted HA in non-induced MCF7 cells was very low and massively increased with GFP-HAS3 induction (Fig. 7A) which was consistent with previous reports (Deen et al, 2014).
23
Figure 7. Induction of GFP-HAS3 increases EV secretion. GFP-HAS3 induction greatly increased HA secretion of MCF7 cells (A) and enhanced secretion of EVs. Concentration of EVs was significantly higher in GFP-HAS3 induced samples (B) but normalizing to cell number diminished the difference and caused higher deviation between samples (C). The size distribution of detected particles was slightly different in non-induced (D) and GFP-HAS3 induced (E) samples, SEM shown as error bars. HA secretion: n=4, SEM, student’s T-test.
EV concentration: n=5, SEM, One-way ANOVA with Tukey’s post hoc. * p < 0.05, *** p < 0.001.
Increasing viscosity of NFC decreases EV secretion and hyaluronan synthesis
To investigate whether the viscosity of NFC scaffold affects EV secretion and activity of HA synthesis, three different concentrations (0.5%, 0.7% and 0.9%) of NFC were prepared mimicking low, medium, and highly viscose matrices, respectively. MFC7 cells induced to express GFP-HAS3 were grown for 7 days and secreted EVs were collected and isolated. The EV concentrations were measured with nanoparticle tracking analysis (Fig. 8). No difference in total particle concentration was found between the different NFC concentrations. However, when the particle concentration was normalized to cell count from the spheroids, the number of particles per cell was slightly lower in samples from higher NFC concentration scaffolds.
Interestingly, the cell count average was highest in 0.7% NFC indicating that the MCF7 cells preferred the medium viscosity matrix for cell growth, while cell counts for 0.5% and 0.9%
NFC scaffolds were lower (Fig. S3B, supplementary material). As detected earlier, the higher concentration of NFC resulted in increased number of NFC derived particles. To take these non-cellular particles into account, the average particle concentration derived from empty NFC
24 scaffolds for each NFC concentration was subtracted from the total particle count of EV samples. When the particles were then normalized with the cell count there was a trend towards a decrease in particle number per cell with increased NFC viscosity. NTA was also utilized in measurement of size distributions of isolated EVs. The EVs displayed highly similar size distributions between repeated samples and did not differ between EVs isolated from different NFC concentrations (Fig. S1, supplementary material). The mean size of particles also remained unchanged between the different NFC matrixes (Table 1, supplementary material).
Figure 8. The concentration of NFC affects the number of secreted EVs. Isolated EVs from three NFC concentrations (0.5%, 0.7% and 0.9%) mimicking low, medium, and high viscosity matrix respectively were analyzed with NTA. No difference in total particle concentration was found (A) but normalizing particles by cell revealed a decrease in particles secreted by cell in medium and high viscosity matrices (B). To take the NFC derived particles into account, average counts of NFC derived particles from the three concentrations were subtracted from the EV counts (C), which further increased the difference between low and high viscosity matrices. Cells in low viscosity matrices secreted significantly more hyaluronan compared to high viscosity matrices (D). For EV experiments: n=5, SEM, One-way ANOVA with Tukey’s post hoc test. HA secretion: n=4, SEM, T-test, * p < 0.05.
As the GFP-HAS3 induction greatly increases the synthesis and secretion of HA, the amount of secreted HA between low and high viscosity scaffolds was compared (Fig. 8D). HA secretion was measured from culture medium after NFC digestion and normalized with cell
25 count. Interestingly, in low viscosity scaffold the cells secreted significantly higher levels of HA compared to high viscosity scaffold.
Increasing hyaluronan content in NFC scaffold alters EV secretion
HA is an abundant component of breast tissue and is increased in the ECM of invasive breast cancers (Schwertfeger et al, 2015). As HA synthesis is known to increase vesicle secretion, we proceeded to investigate whether HA concentration in the ECM would have a similar effect.
To mimic matrices with different HA composition high molecular weight (2 MDa) HA was mixed with 0.5% NFC in 1% and 2% (w/v) concentrations. The aim was to create a combination gel where NFC acts as a supporting element holding the scaffold together. As a water binding molecule, HA forms a gel when diluted with 1% HA causing matrix to become of medium viscosity and 2% HA gel being highly viscose. Prior to experiments it was not known whether mixing HA with NFC would affect digestion of cellulose by cellulases.
However, no difference in digestion effectiveness or amount of NFC derived particles was detected.
After the digestion of NFC, the secreted EVs were collected and isolated. Spheroid cell numbers were calculated and concentration of EVs was measured using NTA (Fig. 9). The increasing HA concentration resulted in decreased cell number in the HA containing scaffolds (Fig. S3C, supplementary material). Similarly, total EV concentration decreased as concentration of HA increased and the decrease was significant with 2% HA concentration. As there was a decrease in cell number, normalization of EV number to cell count compensated the difference between the scaffolds but a clear trend of decreased EV number remained between NFC and NFC + HA scaffolds (Fig. 9B). The decreasing EV count was also reflected to the size distribution of isolated EVs and was most notable in 2% HA culture EVs (Fig. S2, supplementary material). Interestingly, in 2% HA there seemed to be a decrease especially in smallest EVs. Changed size distribution also affected EV mean size (Table 2, supplementary material).
26
Figure 9. External HA content in the matrix decreases EV secretion. EV secretion from cultures with different concentrations of HA mixed in the NFC scaffold was compared with NTA. Increasing HA content reduced the amount of secreted EVs especially in 2% HA hydrogel (A). Significant difference was found in total particle content between non-HA and 2% HA containing gels. Increasing HA content also slightly decreased cell growth (Fig S3C, supplementary material). In addition, presence of HA in the scaffold decreased particles secreted per cell indicating that the reduction was not completely dependent on the decreased cell number (B). n=5, SEM, One-way ANOVA with Tukey’s post hoc test. * p < 0.05.
Imaging properties of the NFC hydrogels
Imaging properties of NFC and optically transparent aNFC were compared by growing GFP- HAS3 induced MCF7 spheroids in low and high concentrations of NFC and aNFC and imaging with confocal microscope. For NFC 0.5% and 0.9% concentrations were used and compared with corresponding dilutions for aNFC, 0.33% and 0.6%, respectively (Fig. 10). Imaging was possible with both hydrogels but aNFC was superior in imaging quality, displaying higher intensity of fluorescence and resulting in overall sharper image. In addition, filopodia and singular EVs were clearly visible with aNFC. While there was a clear difference between the two hydrogels, the concentration had little effect to image quality in both hydrogels. No autofluorescence was detected with either hydrogel.
27
Figure 10. Comparison of image quality in different concentrations of NFC and aNFC. Image quality in low and high concentration (0.5% and 0.9% for NFC, 0.33% and 0.6% for aNFC) hydrogels were compared by imaging GFP-HAS3 induced MFC7 spheroids. Optically transparent aNFC provided superior image quality in both concentrations making it possible to detect single EVs without autofluorescence. 3D projection from 0.33%
aNFC displays how secreted GFP-HAS3 positive EVs can be detected around the spheroids.
Discussion
The use of 3D cell culture models has been constantly increased and gained interest in different areas of research. The ability to provide more in vivo like cell culture conditions and mimic tissue structures and extracellular environment is essential for understanding both normal tissue behavior and many disease mechanisms. Additionally, development of more complex 3D cultures can provide a useful middle ground between in vitro and in vivo experiments and advancements on the established models can also help decrease the need of using animals in experiments. Being able to create cell culture models which accurately represent the state and environment of cells in in vivo tissues is a key element in development of efficient drugs and therapies but also provides useful tools for basic research in better understanding of cellular behavior and development of diseases. The current research of EVs still heavily based on 2D cultures which are cost effective and allow simple recovery of secreted EVs via collection of the culture medium. However, recently it has demonstrated that three-dimensional assembly of cells is necessary for in vivo like secretion of EVs (Thippabhotla et al, 2019). Studying EVs in vivo is extremely challenging, which leaves a need for cell models which can accurately mimic in vivo conditions. However, many 3D culture systems have limitations regarding matrix
28 composition and EV isolation. In this work a novel 3D culture model for simple and efficient isolation and imaging of EVs was introduced. The properties of NFC are excellent for EV studies, as the hydrogel is completely plant based and therefore does not contain any animal derived contaminants that could possibly interfere in imaging or measurements of EV cargo.
In addition, biologically derived hydrogels vary from their batch-to-batch protein composition and can contain unknown amounts of residual growth factors which can affect cell growth and behavior and therefore the reproducibility of acquired results (Hughes et al, 2010). As NFC is free of such issues, in this work its potential as a scaffold for 3D culture derived EV studies was evaluated.
A 3D culture of MCF7 breast cancer cells grown in NFC hydrogels demonstrated that NFC supports growth of spheroids and efficient secretion of EVs in 3D conditions. MCF7 cells formed round spheroids during 7 days of culture and showed high viability without formation of necrotic cores when imaged live in cultures. These observations are in agreement with previous literature stating that the physical properties of NFC hydrogel resemble natural ECM and support growth of cells (Bhattacharya et al, 2012). Here, for the first time the secretion of EVs in NFC cultures was assessed. EV secretion of MCF7 spheroids was found to be very efficient with the cultures secreting up to 40 000 EVs per counted cells during the 7 days of culture. The high number of particles detected with NTA were supported by electron microscopy which showed a high number of small particles in EV samples that were absent in NFC-only isolates and were shown to have a typical morphology of EVs with TEM. In addition, similar observation of high EV numbers were done with confocal microscopy of fluorescently labelled EVs. Although MCF7 EV secretion from 2D cultures was not studied in this work, the high EV numbers indicate that 3D culturing is associated with increased EV secretion. In addition, the EV secretion of MCF7 cells is quite low compared to many other cancer cell types, including other breast cancer cells (Hurwitz et al, 2016), further supporting the evidence that 3D assembly is beneficial for efficient secretion of EVs.
Synthesis of HA and glycoproteins such as mucin have been shown to regulate cell membrane shape and induce formation of protrusions, blebs, and EVs (Rilla et al, 2013; Shurer et al, 2019). While the exact mechanisms of the glycosaminoglycan and glycoprotein -driven membrane shaping are not understood, one convincing hypothesis is that the biopolymers attached on the cell surface cause physical forces that cause membrane curvature and formation of membrane shapes. This was demonstrated using native and synthetic mucin polymers which
29 were able to form different numbers and morphologies of membrane extensions in a cell- surface biopolymer concentration dependent manner (Shurer et al, 2019). In this work, the effect of HA increased synthesis on total EV secretion was studied for the first time in 3D conditions with an ECM mimicking scaffold. The GFP-HAS3 induction led to its high expression on the cell membrane and formation of up to 20 µm long protrusions also positive for GFP-HAS3, consistent with previous reports from multiple different cell lines (Rilla et al, 2013; Kultti et al, 2006; Arasu et al, 2020). Scanning electron microscopy of GFP-HAS3 induced spheroids revealed intense membrane ruffling and high number of pearl or vesicle like structures on the spheroid cell membrane. The membrane shaping caused by GFP-HAS3 induction synthesis is similar as has been demonstrated previously with MCF7 cells in monolayer (Rilla & Koistinen, 2015) and also resembles membrane remodelling caused by mucin polymers in monolayer cultured cells (Shurer et al, 2019). Prior to this work, the effect of GFP-HAS3 induced increase of EV secretion has not been comprehensively studied in 3D conditions. The HAS3 induction led to massive induction in hyaluronan synthesis of MCF7 spheroids and increase in secreted EVs was observed but the increase in EV secretion was not significant when EV numbers were normalized to cell counts. In 2D cultured cells HAS3 induction has been reported to nearly double the number of secreted EVs regardless of cell type (Arasu et al, 2020; Noble et al, 2020). While the difference can be partly explained with cell line, it is more likely that the 3D conditions affect the HA synthesis -driven membrane shaping and EV secretion.
Upon induced expression of HAS3, it has been shown that HAS3 is shed into extracellular vesicles (Deen et al, 2016). In this work, live cell imaging of GFP-HAS3 induced spheroid cultures revealed a high number of secreted GFP-HAS3 positive EVs around the spheroids and the GFP-HAS3 EVs were also detected after EV isolation with confocal microscopy. HAS3 resides on the cell membrane and therefore GFP-HAS3 positive EVs have been considered to be mostly microvesicles as some of those EVs shed from the tips of protrusions (Rilla et al, 2013). Here, using fluorescent tetraspanin antibodies, individual EVs could be detected and characterized by their EV marker expression. Interestingly, it was found that majority of the GFP-HAS3 carrying EVs seemed to be positive to CD63 which is classically considered mainly as an exosomal marker but depending on cell type and conditions has also been found from microvesicles (Crescitelli et al, 2013). In this study, however, it should also be noted that induction of GFP-HAS3 causes an HAS3 overexpression. This also increases the amount of
