Hypoxia-Induced Changes in the Bioactivity of Cytotrophoblast-Derived Exosomes

Migration of extravillous trophoblasts (EVT) into decidua and myometrium is a critical process in the conversion of maternal spiral arterioles and establishing placenta perfusion. EVT migration is affected by cell-to-cell communication and oxygen tension. While the release of exosomes from placental cells has been identified as a significant pathway in materno-fetal communication, the role of placental-derived exosomes in placentation has yet to be established. The aim of this study was to establish the effect of oxygen tension on the release and bioactivity of cytotrophoblast (CT)-derived exosomes on EVT invasion and proliferation. CT were isolated from first trimester fetal tissue (n = 12) using a trypsin-deoxyribonuclease-dispase/Percoll method. CT were cultured under 8%, 3% or 1% O2 for 48 h. Exosomes from CT-conditioned media were isolated by differential and buoyant density centrifugation. The effect of oxygen tension on exosome release (µg exosomal protein/106cells/48 h) and bioactivity were established. HTR-8/SVneo (EVT) were used as target cells to establish the effect (bioactivity) of exosomes on invasion and proliferation as assessed by real-time, live-cell imaging (Incucyte™). The release and bioactivity of CT-derived exosomes were inversely correlated with oxygen tension (p<0.001). Under low oxygen tensions (i.e. 1% O2), CT-derived exosomes promoted EVT invasion and proliferation. Proteomic analysis of exosomes identified oxygen-dependent changes in protein content. We propose that in response to changes in oxygen tension, CTs modify the bioactivity of exosomes, thereby, regulating EVT phenotype. Exosomal induction of EVT migration may represent a normal process of placentation and/or an adaptive response to placental hypoxia.


Introduction
Placentation is an oxygen sensitive process. The events that occur from the time of implantation to maternal perfusion of the placenta are influenced and directed by site-specific oxygen tensions [1]. At the time of embryo implantation, the intrauterine oxygen tension is ,3% [2] while the decidua and myometrium oxygen tension is ,8-12% [3]. This standing oxygen gradient is thought to promote and direct the invasion of extravillous trophoblast cells (EVT) into the decidua and myometrium where they engage and remodel maternal spiral arterioles [4,5]. Intraluminal EVT occludes spiral arterioles to maintain a low oxygen tension environment that is requisite for normal early placental and fetal development. Towards the end of the first trimester, low resistance, high capacity flow is restored and the placental intravillous space is perfused with the maternal blood thus establishing effective materno-fetal exchange.
Factors that compromize trophobast cell (both cytotrophoblast (CT) and EVT) function during this critical period of development may dramatically effect subsequent fetal growth, outcome of pregnancy and the life-long disease risk profile of the newborn [6,7]. EVT migration is affected by oxygen tension [8,9]. Perturbation of intrauterine oxygen tensions, therefore, may compromise normal placental development. The molecular mechanisms by which oxygen tension regulates EVT function and cell-to-cell communication with maternal tissues remain to be fully elucidated.
Recently, the role of exosomes in cell-to-cell communication has been established [10][11][12][13]. Exosomes are nanoparticles (40-100 nm) membrane vesicles that are released following the exocytotic fusion of multi-vesicular bodies with the cell membrane [14]. They have been identified in plasma under both normal and pathological conditions [15] and their concentration increases with disease severity and/or progression [16], and in response to oxidative stress [17].
Recently, we demonstrated that exosomes are released from first trimester placental mesenchymal stem cells (pMSC) and increases endothelial cell migration and vascular tube formation [10]. In addition, the release of exosomes from pMSC was increased under low oxygen tension. These data are consistent with the hypothesis that in response to changes in the environmental milieu (such as oxygen tension) placental cells release exosomes that modify the phenotype of recipient cells. The role of cytotrophoblast cellderived exosome in cell-to-cell communication and, in particular, their effect on EVT has yet to be established. Similarly, the effect of oxygen tension on the release and bioactivity of cytotrophoblast exosomes is not known.
The aim of this study was to test the hypotheses that: (i) exosomes released by cytotrophoblast cells (CT) increase EVT proliferation and invasion; and (ii) the bioactivity and protein content of CT-derived exosomes is altered by oxygen tension. An in vitro treatment/control experimental design was used to test these hypotheses. Exosomes were isolated from primary first trimester placental villous CT. A first trimester EVT cell line (HTR-8/SVneo) was used to assess the effect of CT-derived exosomes on cell proliferation, invasion and bioactivity. The data obtained in this study are consistent with the hypothesis that exosomes from first trimester CT promote HTR-8/SVneo invasion and proliferation, and exosomal protein content is oxygen tension dependent.

First trimester sample collection
Tissue collection was approved by the Human Research Ethics Committees of the Royal Brisbane and Women's Hospital, and the University of Queensland (HREC/09/QRBW/14). Written informed consent was obtained from women for the use of placental tissue for research purposes after clinically indicated termination of pregnancy in compliance with national research guidelines. All experimental procedures were conducted within an ISO17025 accredited (National Association of Testing Authorities, Australia) research facility. All data were recorded within a 21 CRF part 11 compliant electronic laboratory notebook (Irisnote, Redwood City, CA, USA).

Isolation of cytotrophoblast cells
First-trimester cytotrophoblast cells (CTs) were isolated from placentae (8-12 weeks) derived by the legal termination of pregnancy (n = 6 biological samples and 2 independent duplicate cultures per placenta) as previously described [18]. Briefly, 10-20 g of chorionic villi were washed in phosphate buffer saline (PBS) and were subjected to three sequential treatments with digestion buffer (0.25% trypsin (GibcoH Trypsin, Life Technologies TM , Carlsbad, CA) and 0.1 mg of DNase I (Sigma-Aldrich TM , Saint Louis, USA) per ml in HBSS (1x) + HEPES (25 mM). After each 15 min step, the supernatant fluid was layered over fetal calf serum (25 ml supernatant over 5 ml FCS). The supernatant fluid was centrifuged at 4006 g for 10 min and the cell pellet was suspended in Dulbecco's modified Eagle's medium (DMEM) (Life Technologies TM , Carlsbad, CA). A total of 6 ml of DMEM containing cytotrophoblast cells were layered onto a pre-formed colloidal silica gradient (Percoll, 10-70% (GE Healthcare Life Sciences, Piscataway NY, USA) diluted in HBSS 1X (Life Technologies TM , Carlsbad, CA) and centrifuged at 5006 g for 30 min. Cytotrophoblast cells displayed a buoyant density of 1.048 to 1.062 g/ml. Cells were collected and washed in 30 ml of DMEM plus 5 ml of FCS. CTs were further isolated by negative selection for monoclonal anti-HLA-A, B, C (clone W6/32, Accurate Chemical and Scientific Corp, Westbury, NY, USA) and negative selection for monoclonal anti-CD45 (Sigma-Aldrich) using magnetic beads (Dynabeads M450, Life Technologies) and characterized by specific markers: anti-cytokeratin-7 (clone 7H8C4, Sigma-Aldrich) and anti-vimentin (clone V9, Sigma-Aldrich).

Isolation and characterization of cytotrophoblast-derived exosomes
Isolation. Exosomes were isolated from CT-conditioned incubation medium as previously described [10,19]. In brief, CT-conditioned media was centrifuged at 3006 g for 15 min, 2,0006 g for 30 min, and 12,0006 g for 45 min to remove whole cells, and debris. The resultant supernatant fluid was passed through a 0.22 mm sterile filter (Steritop TM , Millipore, Billerica, MA, USA) and then centrifuged at 100,0006 g for 75 min (Thermo Fisher Scientific Ins., Asheville, NC, USA, Sorvall, SureSpinTM 630/36, Tube angle: 900). The pellet was resuspended in PBS, washed and re-centrifuged (100,0006 g for 75 minutes). The pellet was resuspended in PBS, layered on a continuous sucrose gradient (0.25-2 mM sucrose) (Sigma-Aldrich) prepared using a Hoefer SG30 gradient maker (GE Healthcare, NSW, Australia) and centrifuged at 110,000 g for 20 h. Fractions (10 in total) were collected with an 18-G needle and the density of each fraction was determined using the refraction index with OPTi digital refractometer (Bellingham+Stanley Inc., Lawrenceville, GA, USA). Fraction were diluted in PBS, and then centrifuged at 110,0006 g for 70 min. Recovered exosomes pellet was resuspended in 50 ml PBS and their protein concentration determined by the dye-binding assay [19].
Exosome protein quantification. Exosome-containing fractions (1.146 to 1.199 g/ml) from sucrose gradients were combined in a single tube and centrifuged at 110,0006 g for 70 min. Exosome pellets were resuspended in 50 ml PBS and their proteins determined by the DC TM Protein Assay (BIO-RAD, Hercules, CA, USA). Briefly, exosome samples (5 ml) were prepared by adding RIPA buffer (50 mM Tris, 1% Triton 6100, 0.1% SDS, 0.5% DOC, 1 mM EDTA, 150 mM NaCl, protease inhibitor) directly to exosomes suspended in PBS and sonicated at 37uC for 15 s three times to lyse exosome membrane and solubilize the proteins. Bovine serum albumin (BSA) diluted in RIPA buffer and PBS mixture (1:1) was prepared as protein standards (0, 200, 400, 600, 800, 1,000, 1,500 mg/mL). Standards and samples (exosomes) were transferred to 96-well plates. Alkaline copper tartrate solution (BIO-RAD, USA) and dilute Folin Reagent (BIO-RAD, USA) were added to the samples and incubated for 15 min. The absorbance was read at 750 nm with the Paradigm Detection Platform (Beckman Coulter, USA).
Transmission electron microscopy. Exosomes isolated by differential and continuous sucrose gradient centrifugation were analyzed by transmission electron microscopy. Exosome pellets (as described above) were fixed in 3% (w/v) glutaraldehyde and 2% paraformaldehyde in cacodylate buffer, pH 7.3. 5 ml of sample was then applied to a continuous carbon grid and negatively stained with 2% uranyl acetate. The samples were examined in an FEI Tecnai 12 transmission electron microscope (FEI TM , Hillsboro, Oregon, USA).

Effect of oxygen tension on exosome release
The effects of oxygen tension on the release of exosomes from CTs were assessed by incubating cells for 48 h in exosome-free culture medium (culture media was depleted of the contaminating exosomes using the same protocol for exosome isolation described previously and exosome-free culture media medium was confirmed by electron microscope) under an atmosphere of 5% CO2balanced N2 to obtain 1%, 3% or 8% O2 (pO2 ,6.75, ,20.25 or ,54 mmHg, respectively, n = 6 biological replicates in duplicate) in an automated PROOX 110-scaled hypoxia chamber (Bio-Spherics TM , Lacona, NY, USA). The three-compartment hypoxia chamber allowed the simultaneous analysis of CT-cells from individual placentae thus reducing inter-assay variation. Cell viability was determined by Trypan Blue exclusion and CountessH Automated cell counter (Life Technologies TM ). In all experiments, viability remained at .95% and was not significantly different between groups (p = 0.85). Incubation media pO2 and pH were independently confirmed using a blood gas analyzer (Radio-meterH, Brønshøj, Denmark) and NeoFox oxygen probe (Ocean Optics TM , Dunedin, FL, USA).

Time Course
To determine the effects of CT-derived exosomes on cell invasion, HTR-8/SVneo cells were incubated in the presence (treatment: 100 mg exosomal protein/ml) or absence (control) of exosomes for up to 24 h (n = 12). Cell invasion was monitored by scratch assays [10]. A scratch was made on confluent monolayers using a 96-pin WoundMaker TM (BioScience Inc, Ann Arbor, Michigan, USA) and layered with 3 mg/ml collagen type I (Life Technologies TM , Carlsbad, CA) and incubated at 37uC for 30 min to create a 3D matrix-gel. Wound images were automatically acquired and registered by the IncuCyteTM software system. CellPlayer TM 96-Well Invasion Assay software was use to fully automate data collection. Data were processed and analyzed using IncuCyte TM 96-Well Cell Invasion Software Application Module. Data are presented as the Relative Wound Density (Eizen, v1.0 algorithm). The rate of wound closure was compared using the half-maximal stimulatory time (ST50) and area under the time course curve (AUC) [10].

Dose Response
To assess the effect of exosome concentration on extra-villous trophoblast invasion and proliferation, HTR-8/SVneo cells were cultured in the presence of increasing exosomes concentration (5, 10, 20, 50 and 100 mg exosomal protein/ml) for 24 h. Data are presented as half-maximal effective concentration (EC50).

Effects of oxygen tension on CT-derived exosome bioactivity
To determine the effect of oxygen tension on exosome bioactivity, exosomes were obtained from CTs incubated under 1%, 3% and 8% O2 (as detailed above) for 24 h. The effects of exosomes on HTR-8/SVneo cell invasion and proliferation were assessed as detailed above.

Proliferation
Cell proliferation was measured by the change in confluence in the presence of exosomes (100 mg exosomal protein/ml) isolated from 1%, 3% and 8% O2 using a standard confluence algorithm (IncuCyte TM ) as we have previously published [10,23].

Proteomic analysis of cytotrophoblast derived-exosomes by mass spectrometry (MS)
Isolated exosomes were solubilized in 8 M urea in 50 mM ammonium bicarbonate, pH 8.5, and reduced with DTT for 1 h. Proteins were then alkylated in 10 mM iodoacetic acid (IAA) for 1 h in the dark. The sample was diluted to 1:10 with 50 mM ammonium bicarbonate and digested with trypsin (20 mg) at 37uC for 18 h. The samples were desalted by solid phase extraction using a STAGE tip protocol (Stop and go extraction tips for matrix-assisted laser desorption/ionization, nano-electrospray, and LC/MS sample pre-treatment in proteomics). The eluted peptides were dried by centrifugal evaporation to remove acetonitrile and redissolved in Solvent A. The resulting peptide mixture was analysed by Liquid Chromatography (LC)/Mass Spectrometry (MS) LC-MS/MS on a 5600 Triple TOF mass spectrometer (AB Sciex, Framingham, U.S.A.) equipped with an Eksigent Nanoflow binary gradient HPLC system and a nanospray III ion source. Solvent A was 0.1% formic acid in water and solvent B was 0.1% formic acid in acetonitrile. MS/MS spectra were collected using Information Dependent Acquisition (IDA) using a survey scan (m/z 350-1500) followed by 25 datadependent product ion scans of the 25 most intense precursor ions. All mass spectra were analysed using the Mascot and Protein Pilot search engines against the Swissprot-swissprot database with the species set as human (scores greater than 30). False discovery rate (FDR) was estimated using a reversed sequence database. Finally, proteins identified were submitted to bioinformatic pathway analysis (Ingenuity Pathway Analysis [IPA]; Ingenuity Systems, Mountain View, CA; www.ingenuity.com).

Statistical Analysis
Data are represented as mean 6 SEM, with n = 6-12 different cell cultures of CTs isolated from first trimester pregnancies. Comparisons between two group means were performed by unpaired Student's t-test. Multiple groups were compared using the analysis of variance (ANOVA) post hoc analyses were used for pair-wize comparisons (Bonferroni correction test). Statistical significance was defined as p,0.05.

Characterization of cytotrophoblast-derived exosomes
Characterization included analysis of the physical properties of exosomes on a continuous sucrose gradient and protein composition by Western blot. CD63, CD9, CD81 and PLAP positive nano-paticles displayed a buoyant density of 1.146 -1.199 g/ml ( Figure 1A). Each exosome-containing fraction (5 to 8) was separated by SDS-PAGE and stained with SimplyBlue TM Safe-Stain (Invitrogen). Fractions 5 to 8 displayed a similar protein profile that was distinct from that observed for cytotrophoblast total cellular protein ( Figure 1B). The exosomal fractions isolated from CTs were pooled and examined under transmission electron microscopy. Exosomes where identified as vesicles sized between 40-100 nm ( Figure 1C).

CTs-derived exosomes induce EVT invasion
Time course. Representative photomicrographs of HTR-8/ Svneo wound closure for treatment and control experiments are presented in Figure 3A. The effect of CT-derived exosomes on HTR-8/SVneo cell invasion is presented as relative wound density (percent) over time (Figure 3

CTs-derived exosomes increase EVT proliferation
A real-time imaging system (IncuCyte TM ) was used to measure cell proliferation using non-label cell monolayer confluence approach. The proliferation ratio (+exosomes/-exosomes for each Figure 3. Cytotrophoblast cell-derived exosomes increased EVT invasion. EVT cells were grown to confluence in complete media. A wound was made using 96 well WoundMaker and then overlaid to form a 3D matrix-gel (see Methods). EVT invasion was measured in absence (white circles) or presence (black circles) of 100 ug/ml of exosomes from cytotrophoblast cells exposed to 1% O2 (exo-CTs-1%) for   hour) was significantly higher (p,0.05) compared to the control in the absence of exosomes ( Figure 5). At 24 hours, exo-CTs-1% increased EVT proliferation by ,1.5 fold compared to those in absence of exosomes (control).

Proteomic analysis of CTs-derived exosomes
Mass spectrometry analysis identified over 160 exosomal proteins (Table 1). We identified unique proteins for each condition ( Figure 6). The biological relevance of differentially expression proteins was analyzed using Ingenuity Pathway Analysis (IPA) software. Exosomal proteins isolated from CT exposed to different oxygen tensions were associated with cellular movement and morphology, immune cell trafficking and cellular assembly and organization in accordance with IPA analysis. The canonical pathways associated with our exosomal proteins defined by IPA Core comparison analysis showed that the score (-log [pvalue]) for proteins associated with HIF-a signalling ( Figure 6B) and IL-8 signalling ( Figure 6C) were inversely correlated to oxygen tension. Finally, we investigated the molecular network that can be activated by the unique proteins identified in exosomes isolated from cytotrophoblast cells exposed to 1% O 2 (31 proteins) (Figure 7). We found molecules involved in cellular movement such as MMP9, TGF-b, P38 MAPK, VEGF and others.

Discussion
Extravillous trophoblast invasion into the maternal tissue is a critical process in placentation. Hypoxia is a risk factor for complications of pregnancy and may adversely affect placentation and development of the materno-fetal vascular exchange. In particular, during early pregnancy low oxygen tension may impact on EVT migration and interactions with the maternal spiral arterioles [24][25][26]. Exosomes from cytotrophoblast cells may interact with EVT and modify their invasiveness, however, the effect and role of exosomes from placental cells has yet to be defined. The aim of this study was to establish the effect of oxygen tension on the release and bioactivity of CT-derived exosomes on EVT invasion and proliferation in vitro. The data obtained are consistent with the hypothesis that exosomes released from cytotrophoblast cells incubated under low oxygen tension promote HTR-8/SVneo invasion and proliferation. While the role of CT exosomes in vivo remains to be established, their release under hypoxic conditions within the placenta may be an adaptive response to promote proliferation and invasion of extravillous trophoblast.
In this study, placental cell exosomes were isolated from cellconditioned media using differential and buoyant density centrifugation according to our previously published method [10]. CTderived exosomes displayed: a buoyant density between 1.146 and 1.199 g/ml; positive expression for exosome markers, such as CD63, CD9 and CD81; and a diameter of than less 100 nm analysed under electron microscope. These characteristics are consistent with previously published data [15,27]. The release of exosomes from CTs was measured as total CD63 positive, particulate protein with a buoyant density 1.146 to 1.199 g/ml.
The release of exosomes from CTs was inversely correlated with oxygen tension. When CTs were incubated under 1% oxygen, exosome release into the incubation medium was 3-fold greater than that observed when cells were incubated under 3% and 8% oxygen. This increase in exosome release was not associated with a loss of cell viability. Our observations that exosome release is affected by oxygen tension are consistent with previously published data obtained from other cell types [28][29][30], including first trimester placental mesenchymal stem cells [10]. While the mechanism(s) by which low oxygen tension increases the release of exosome release remains to be established, oxygen-sensing transcriptional factors, such as HIF-a may be involved [28]. Jauniaux et al., (2000) measured in situ oxygen tension during early pregnancy (at 60 days) within the chorionic placenta, intervillous space (IVS), and maternal endometrium underlying the placenta [31]. Oxygen tension was ,3%, 1% and 8%, respectively. Low oxygen tension within the placenta and IVS at this stage of pregnancy may promote the release of exosomes from cytotrophoblast cells and enhance cell-to-cell communication. In particular, exosomes released from CTs may induce functional changes in EVTs that promote cell invasion and proliferation. Consistent with this hypothesis, in this study we demonstrated that CT-derived exosomes increase invasion and proliferation of the EVT cell line HTR-8/SVneo.
During the first trimester of pregnancy, EVT cells invade the decidua and myometrium and regulate the flow of maternal blood into the IVS. EVTs co-localize with maternal spiral arterioles and are present within the lumen of these vessels (where they are thought to prevent flow into the IVS). Subsequent perfusion of the IVS is associated with the transformation of these arterioles from high resistance, low capacity to low resistance, high capacity vessels. EVTs play a role in remodelling these vessels [32,33]. Abnormal EVT function may result in failure to transform these vessels resulting in compromised placental perfusion and hypoxia [34]. What regulates EVT invasion and/or function and their interactions with maternal vessels remains to be clearly established. The data obtained in this study, however, establish that exosomes released form CTs increase EVT cell invasion in a concentrationdependent manner, and that the activity of exosomes is increased under low oxygen tension.
In this study, HTR-8/SVneo were cultured under low oxygen tension for at least 48 h before experimental manipulation and incubated in the presence of CT-derived exosomes for 24 h. An automated real-time imaging system was used to maintain cells in optimal conditions for quantifying cell invasion and proliferation. Exosome-treatment reduced EVT ST50 compared to control incubations (i.e. absence of exosomes). Furthermore, exosomes isolated from CTs incubated under low oxygen tension (i.e. 1%) displayed greater activity (per unit exosomal protein) than exosomes obtained from cells incubated at higher oxygen tensions (i.e. 3 and 8%). Similar, effects (but of lower magnitude) were observed for EVT cell proliferation.
The effect of low oxygen tension on EVT invasion remains controversial and disparate data have been reported [8,33,35,36]. Studies performed on EVT cells isolated from placental tissue (5 to 10 weeks of gestation) established that low oxygen tension reduces invasion of EVT cells through decreased MMP-2 [36] and appears to be mediated by urokinase plasminogen activator (PLAU) system [8]. In contrast, low oxygen tension (,1% O 2 ) increased HTR-8/ SVneo cell invasion when compared to cell incubated under 20% O 2 [37].
The data obtained in the current study establish a role for exosomes in the intercellular communication between placental cells and in regulating EVT cell invasion in an oxygen-dependent manner. Low oxygen tension increased exosome release and  Table 1 List of proteins identified in exosomes from CT exposed to different oxygen level. All mass spectra were analysed using the Mascot and Protein Pilot search engines against the Swissprot-swissprot database with the species set as human (score over 30). Exosomal proteins identified by mass-spectrometry were analyzed with the Ingenuity Pathway Analysis (Ingenuity Systems, www.ingenuity.com). False discovery rate (FDR) was estimated using a reversed sequence database. List of total exosomal protein from cytotrophoblast cells exposed to different oxygen level are presented as Protein ID, Symbol, Entrez Gene Name, Location and type. doi:10.1371/journal.pone.0079636.t001 modified their protein content. When CTs were incubated under different oxygen tensions, exosomal protein content was altered significantly. Ingenuity Pathway Analysis (IPA) of exosomal proteins identified oxygen-dependent changes in HIFa and IL-8 signalling pathways. In addition, when CTs were incubated under low oxygen tension, exosomal proteins identified were predominantly associated with pathways involved in the activation of MMP-9, TGF-b, MAPK, VEGF, p38MAPK, TIMP1 and ERK1/2. It remains to be established whether or not specific changes in exosomal protein content are causally related to changes in EVT invasion and proliferation. EVT cells have a crucial role in placentation, characterized by their invasion of spiral uterine arteries to establish a low-resistance, high-capacity perfusion system. The mechanism involved in EVT invasion is not completely understood, however, there is a consensus that invading EVT cells up-regulate proteins such as MMPs, integrins (a5b1 and a1b1) and VE-cadherin which support uterine wall invasion. Our data suggest that the exosome released from cytotrophoblast cells also promote EVT invasion including the activation of MMPs, MAPK and invasiveness pathways. We are only beginning to develop an understanding of the role of placental-derived exosomes (e.g. cytotrophoblast cells) in early pregnancy events and, in particular, how they might affect the function of key cell-types (e.g. EVT) involved in the development of the placenta and its vascular communication with both mother and fetus. In summary, the release of exosomes from primary culture of cytotrophoblast cells is oxygen tension-dependent. CT-derived exosomes increase EVT cell invasion and proliferation in a concentration and oxygen-dependent manner. Exosomal protein content is altered in response to oxygen tension, with the enhancement of signals involved in cellular invasion and migration. The release of CT-derive exosomes under hypoxic conditions within the placenta may be an adaptive response to promote proliferation and invasion of extravillous trophoblast cells.