If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Extracellular vesicles (EV) in milk, particularly exosomes, have attracted considerable attention as bioactive food compounds and for their use in drug delivery. The utility of small EV in milk (sMEV) as an animal feed additive and in drug delivery would be enhanced by cost-effective large-scale protocols for the enrichment of sMEV from byproducts in dairy plants. Here, we tested the hypothesis that sMEV may be enriched from byproducts of cheesemaking by tangential flow filtration (EV-FF) and that the sMEV have properties similar to sMEV prepared by ultracentrifugation (sMEV-UC). Three fractions of EV were purified from the whey fraction of cottage cheese making by using EV-FF that passed through a membrane with a 50-kDa cutoff (50 penetrate; 50P), and subfractions of 50P that were retained (100 retentate; 100R) or passed through (100 penetrate; 100P) a membrane with a 100-kDa cutoff; sMEV-UC controls were prepared by serial ultracentrifugation. The abundance of sMEV (<200 nm) was less than 0.3% in EV-FF compared with sMEV-UC (1012/mL of milk). Despite the low EV count, the protein content (mg/mL) of 100R (63 ± 0.02; ± standard deviation) was higher than that of 50P (0.75 ± 0.10), 100P (0.65 ± 0.40), and sMEV-UC (27 ± 0.02). There were 17, 14, 35, and 75 distinct proteins detected by nontargeted mass spectrometry analysis in 50P, 100R, 100P, and sMEV-UC, respectively. Exosome markers CD9, CD63, CD81, HSP-70, PDCD6IP, and TSG101 were detected in control sMEV-UC but not in EV-FF by using targeted mass spectrometry and immunoblot analyses. Negative exosome markers, APOB, β-integrin, and histone H3 were below the limit of detection in EV-FF and control sMEV-UC analyzed by immunoblotting. The abundance of the major milk fat globule protein butyrophilin showed the following pattern: 100R ≫ 100P = 50P > sMEV-UC. More than 100 mature microRNA were detected in sMEV-UC by using sequencing analysis, compared with 36 to 60 microRNA in EV-FF. Only 100R and sMEV-UC yielded mRNA in quantities and qualities sufficient for sequencing analysis; an average of 276,000 and 838,000 reads were mapped to approximately 14,600 and 18,500 genes in 100R and sMEV-UC, respectively. In principal component analysis, microRNA, mRNA, and protein in EV-FF preparations clustered separately from control sMEV-UC. We conclude that under the conditions used here, flow filtration yields a heterogeneous population of milk EV.
The International Society for Extracellular Vesicles defines extracellular vesicles (EV) as particles naturally released from the cell that are delimited by a lipid bilayer and cannot replicate (i.e., do not contain a functional nucleus;
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines.
). Communication is achieved by the transfer of proteins, lipids, and various classes of RNA such as microRNA and mRNA from exosome donor cells to adjacent or distant recipient cells, and also by the binding of exosomes to receptors on the surface of receptor cells (
MicroRNAs are absorbed in biologically meaningful amounts from nutritionally relevant doses of cow's milk and affect gene expression in peripheral blood mononuclear cells, HEK-293 kidney cell cultures, and mouse livers.
). Follow-up studies suggest that human and rodent gastrointestinal cells take up small milk EV (sMEV) and their microRNA cargos by endocytosis and secrete sMEV and cargos across the basolateral membrane (
The intestinal transport of bovine milk exosomes is mediated by endocytosis in human colon carcinoma caco-2 cells and rat small intestinal IEC-6 cells.
). The sMEV and microRNA accumulate primarily in the intestinal mucosa, liver, and brain; distinct species of microRNA have unique distribution patterns that may also differ from that of their sMEV shells (
). Encapsulation in sMEV protects microRNA cargos against degradation by harsh conditions as present in the gastrointestinal tract and dairy processing plants such as low pH and RNase (
The intestinal transport of bovine milk exosomes is mediated by endocytosis in human colon carcinoma caco-2 cells and rat small intestinal IEC-6 cells.
). The bioavailability and distribution of RNA other than microRNA remains to be determined.
The discovery that sMEV and their microRNA cargos are bioavailable has sparked an interest in 2 fields of study, nutrition and nanotherapy. In nutrition, sMEV and their microRNA cargos are studied for their potential roles as bioactive food compounds. These studies are based on observations that consumption of a sMEV and RNA-depleted diet elicited phenotypes such as increased levels of purine metabolites in body fluids, moderate loss of muscle grip strength, changes in gut microbial communities, and altered gene expression patterns in intestinal mucosa, liver, skeletal muscle, and placenta (
Concentrations of purine metabolites are elevated in fluids from adults and infants and in livers from mice fed diets depleted of bovine milk exosomes and their RNA cargos.
A diet defined by its content of bovine milk exosomes and their RNA cargos has moderate effects on gene expression, amino acid profiles and grip strength in skeletal muscle in C57BL/6 mice.
Dietary Depletion of milk exosomes and their microRNA cargos elicits a depletion of miR-200a-3p and elevated intestinal inflammation and CXCL9 expression in Mdr1a−/− mice.
). In nanotherapy, exosomes at large are studied for their use in the delivery of drugs to tumor sites, including the delivery of small interfering RNA (
Both fields of study, nutrition and nanotherapy, would benefit from the availability of technologies for the purification of sMEV on a large scale, particularly if the technology lends itself to the purification of sMEV from byproducts (“waste”) in dairy plants. Purification of sMEV by tangential flow filtration (FF) is one possible strategy and has been successfully used to isolate exosomes from large volumes of fluids and mesenchymal stem cell cultures, respectively (
). The dairy industry generates large volumes of liquid waste during cheesemaking. Sweet whey originates from the rennet coagulation of milk proteins with a pH of at least 5.6 in the production of hard cheeses, whereas acidic whey is obtained when protein coagulation is performed at a pH of no higher than 5.1 in the production of cottage cheese (
The use of FF for enriching sMEV from dairy byproducts poses a challenging problem because sMEV have sizes (~100 nm) similar to milk compounds such as lipoproteins, small microvesicles (ectosomes), fat globules, and possibly casein micelles that escaped coagulation in cheesemaking (
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines.
). The study objective was to determine whether FF is suitable for enriching sMEV from byproducts of cheesemaking. To achieve our goal, we compared the size, morphology, and protein and RNA cargos in EV purified by tangential flow filtration (EV-FF) to reference sMEV purified by serial ultracentrifugation (sMEV-UC).
MATERIALS AND METHODS
Purification of EV by EV-FF and SMEV-UC and Iodixanol Density Gradient Purification of EV-FF Fractions and SMEV-UC
Three fractions of EV were prepared by FF by using the acid whey from processing cottage cheese (
). The whey contained 55% protein (by weight) and was further concentrated using a membrane with a 10-kDa cutoff to yield a whey protein concentrate (80% protein by weight). The concentrate was filtered through a micropore filter (800-kDa cutoff). The retentate was used to prepare 3 fractions by FF as follows. First, the retentate was subjected to FF using a membrane with a 50-kDa cutoff. The permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R; Supplemental Figure S1, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). The FF samples were provided by Land O'Lakes, Inc. Bovine milk (1% fat) from a local grocery store was used to isolate sMEV by serial ultracentrifugation as previously described with minor modifications and are denoted sMEV-UC (
The intestinal transport of bovine milk exosomes is mediated by endocytosis in human colon carcinoma caco-2 cells and rat small intestinal IEC-6 cells.
). Somatic cells, debris, and large fat globules were removed by centrifugation (12,000 × g at 4°C for 45 min). Fat-free supernatant was ultracentrifuged using a Fiberlite-F37L-8x100 rotor (Thermo-Scientific) at 80,000 × g at 4°C for 60 min to remove precipitated protein, residual fat globules, and EV larger than exosomes. Supernatant was collected and filtered sequentially using 0.45-μm and 0.22-μm membrane syringe filters (Milex) at 4°C. The sMEV were precipitated by ultracentrifugation of the filtered samples at 120,000 × g at 4°C for 90 min. The sMEV pellets were resuspended in sterile phosphate-buffered saline and filtered through 0.22-μm membrane filters. Sodium azide was added to produce a final concentration of 0.01%, and sMEV were stored at −80°C for up to 120 d. The sMEV-UC provide a reference for morphology, size, and cargos to be expected in milk exosomes. However, sMEV-UC should not be used to assess the efficiency of tangential flow filtration for preparing sMEV because sMEV-UC and EV-FF were obtained from distinct sources (see Discussion). In select experiments, EV-FF and sMEV-UC controls were isolated by ultracentrifugation and further purified by iodixanol density gradient centrifugation as previously described with minor modifications (
Cushioned-density gradient ultracentrifugation (C-DGUC): A refined and high performance method for the isolation, characterization, and use of exosomes.
). The protocol has been deposited in the EV-Track database (https://evtrack.org/) under accession ID EV210118 (EV-Track, 2017 #13589).
Analysis of EV Count, Size, and Charge by Nanoparticle Tracking Device, Dynamic Light Scattering, and Zeta Potential
The EV count and size distribution were assessed by using an NS-300 nanoparticle tracking analyzer (Malvern, Inc.) and a 200-nm cutoff. The EV-FF and sMEV-UC preparations were diluted 10-fold and 100,000-fold, respectively, with HPLC-grade water to achieve a density of 25 to 44 EV per frame. Instrument settings were 12 to 15 arbitrary units for camera level and 5 arbitrary units for detection threshold. No information was available regarding the starting volume of milk used to prepare the FF samples. Therefore, a comparison of absolute EV counts in EV-FF to counts in sMEV-UC was arbitrary. The EV counts were used to normalize for input when appropriate. The EV size, size distribution, polydispersity index, and surface zeta potential were assessed by dynamic light scattering using a Nano ZS Zetasizer (Malvern, Inc.).
Assessment of EV Morphology by Transmission Electron Microscopy
Phosphotungstic acid was used for negative staining of EV (
). Equal numbers of EV were used, which required diluting sMEV-UC 20,000-fold with HPLC-grade water. Proteins were crosslinked with 249.7 mmol/L (2.5%) glutaraldehyde (room temperature, 1 h); in some preparations, lipid membranes were stabilized with 5.9 mmol/L (1%) tannic acid (room temperature, 1 h) after glutaraldehyde fixation (
). Samples were stored overnight at 4°C. Thirty microliters of fixed EV-FF or sMEV-UC were applied to FCF-150-Cu formavar carbon film (150 mesh copper grids, Electron Microscopy Sciences, Inc.) and stained with 3.5 mmol/L (1%) phosphotungstic acid.
Protein Analysis by Immunoblotting and Targeted and Nontargeted Tandem Liquid Chromatography-Mass Spectrometry
Protein Quantification
Total protein was released from an equal number of EV by using radio-immunoprecipitation assaybuffer (HY-K1001, MedChemExpress); protease inhibitor cocktail (P8340, Sigma Aldrich) was added to the buffer before assay to produce a 10-fold dilution of the inhibitor cocktail. Total protein was quantified using the bicinchoninic acid assay (23225, ThermoFisher). If protein concentrations were below the detection limit of the bicinchoninic acid assay, the fluorometric Qubit protein assay (Q33211, ThermoFisher) was used.
Immunoblot Analysis
An equal number of EV (3 × 108) was mixed with 4X NuPAGE LDS buffer and heated at 72°C for 10 min before electrophoresis using a 4 to 12% gradient Bis-Tris gel (NPO322 Box, ThermoFisher) as previously described with minor modifications (
). Proteins were transferred to polyvinylidene fluoride membranes (Millipore). Membranes were blocked using commercial blocking buffer (LI-COR Biosciences) and probed using mouse anti-bovine CD63 (MCA2042GA; Bio-Rad), mouse anti-CD9 (ab61873; Abcam), goat anti-Alix (sc-49268; Santa Cruz Biotechnology), and rabbit anti-CD81 (ABIN2789419, Antibodies Online) as markers for exosomes (
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines.
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines.
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines.
). All primary antibodies were diluted 500-fold. Membranes were washed 3 times using cold phosphate buffer saline containing 1% Tween-20 for 5 min each. Bands were visualized using IRDye 800CW-labeled secondary antibodies (25,000-fold dilution; 926-32210, 926-32214, and 926-32211; LI-COR) in an Odyssey CLx infrared imaging system (LI-COR Biosciences).
Tandem Liquid Chromatography-Mass Spectrometry Sample Preparation and Analysis
An equal number of EV (3 × 108) from EV-FF preparations and control sMEV-UC was heated in NuPAGE LDS loading buffer (NP0007, ThermoFisher) at 95°C for 10 min; 15 µL were loaded per lane on a 4 to 12% gradient Bis-Tris gel (NPO322Box, ThermoFisher). Gels were stained using fresh Coomassie blue R-250 (190343, MP Biomedicals) and destained using 10% acetic acid, 20% methanol, and 70% water. Lanes were cut into 5 segments of equal size. Trypsin digestion, tandem liquid chromatography-mass spectrometry (LC-MS/MS) analysis, and protein identification were performed as described previously (
An equal number of EV (1.5 × 1012) from EV-FF and sMEV-UC controls was used to precipitate proteins. The EV samples were mixed with cold acetone (100%, 1:4 vol/vol) and incubated at −20°C for 1 h. Samples were centrifuged at 11,500 × g for 15 min to pellet proteins. Pellets were resuspended in 40 µL of a denaturation buffer containing 25 mM ammonium bicarbonate (pH 8.0), 10 mM Tris (2-carboxyethyl) phosphine hydrochloride, and 5% sodium deoxycholate and incubated at 60°C for 10 min. Five microliters of alkylation buffer (18 mg/mL iodoacetamide in water) were added and incubated in the dark at room temperature for an hour. Samples were diluted with 155 µL of 25 mM ammonium bicarbonate (pH 8.0) followed by trypsin digestion (1 µg/µL) at 37°C overnight. Ten microliters of 10% trifluoroacetic acid were added, incubated for 30 min at room temperature, and centrifuged at 15,000 × g for 5 min, and the supernatants were collected for multiple reaction monitoring analysis. The FASTA sequences for exosomes marker proteins CD9, HSP 70, and TSG101 were obtained from The UniProt database and imported into the Skyline software (
). The following parameters were used: trypsin was used as the enzyme in protein digestion; carbamidomethyl was used as modification; maximum peptide length was 25; ions were type p, b and y, and a product ion selection was set from ion 1 to last ion. Subsequently, the list of transitions was generated and used for multiple reaction monitoring analysis (Supplemental Table S1, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). Multiple reaction monitoring was performed on an Agilent 1260 LC system coupled to a triple quadrupole mass spectrometer (Agilent 6410 Series). The mobile phases were composed of 0.1% formic acid (solvent A) and 0.1% formic acid in acetonitrile (solvent B). Ten microliters of resuspended peptides were loaded onto a Halo 5 ES-C18 column (1.0 mm × 150 mm) at a flow rate of 50 µL/min. Separation of peptides was achieved using the following stepwise gradient: t = 0 min: 5% B; t = 6 min: 25% B; t = 29 min: 80% B; t = 29.10 min: 100% B; t = 31 min: 100% B, and t = 33 min: 5% B (column re-equilibration). Spectra were acquired in positive mode over a scan range of 300 to 1,200 m/z. The raw data files from multiple reaction monitoring analysis were analyzed using Skyline and peak areas were expressed as percent of sMEV-UC control.
MicroRNA Sequencing Analysis
Small RNA were extracted using the miRNeasy Serum/Plasma Kit (217184, Qiagen) and shipped on dry ice to the Genomics Sequencing Core Facility in the University of Nebraska Medical Center (Omaha, NE). The quality and quantity of RNA were assessed by using a Fragment Bioanalyzer (Advanced Analytical Technologies, Inc.) and the Qubit microRNA Assay Kit (Q32880, ThermoFisher), respectively. Small RNA libraries were prepared by using 10 ng of microRNA per library following NEXTflex Small RNA Seq Kit v3 (BiooScientific) and sequenced using the Illumina NextSeq 500 platform in single-read mode. Raw sequencing data were deposited in the NCBI SAR database (https://www.ncbi.nlm.nih.gov/sra) under ID SUB7816703. The quality of sequence reads was assessed by using FastQC (
). Reads with lengths between 18 and 40 bp were mapped against the bovine reference genome (Bos taurus UMD3.1) with no mismatch allowed. MicroRNA expression analysis was conducted by using miRDeep2 (
). We considered real any mature microRNA that yielded at least 10 raw counts in at least 1 of the 3 biological repeats of a sample. When comparing preparations, we required that a microRNA was expressed in all biological replicates. A hierarchical clustering analysis was used to organize the identified microRNA based on similarities in their expression profiles in EV-FF and control sMEV-UC. Principal component analysis was conducted for an unsupervised assessment of the relationship between the microRNA content in EV from EV-FF and control sMEV-UC (
Total RNA were extracted using the miRNeasy Micro Kit (217084, Qiagen) and shipped on dry ice to the Genomics Sequencing Core Facility, University of Nebraska Medical Center. The quality of RNA was assessed by using a Fragment Bioanalyzer (Advanced Analytical Technologies, Inc.). The RNA libraries were prepared by using 1 ng of total RNA per sample following the SMART-Seq Stranded Kit (Takara Bio, Inc.), followed by deep sequencing using the Illumina NextSeq 500 platform. The mRNA-sequencing data analysis and quality control were performed using the nf-core/rnaseq pipeline (
). Gene quantification was performed with featureCounts, which counts the number of overlapping reads in a gene or a genomic feature of interest. The threshold to consider a gene real was when 10 raw counts were present in 1 of the 3 biological repeats of a sample. Raw sequencing data were deposited in the NCBI SAR database under accession SUB7816703. Clustering analysis was performed to organize the identified genes based on similarities in their abundance profiles in 100R and control sMEV-UC. Principal component analysis was conducted to assess the relationship between gene contents in 100R fraction of EV-FF and control sMEV-UC (
). The data variation was heterogeneous; therefore, data were log-transformed before statistical analysis. The statistical significance of differences among groups were analyzed using one-way ANOVA and Dunnett's posthoc test; control sMEV-UC preparations were designated the control in posthoc testing. Differences were considered significant if P < 0.05. Data were analyzed with the Graph Pad prism version 6.0. All experiments were performed in 3 biological replicates. Means ± standard deviations are reported.
RESULTS
The EV Count, Size, Charge, and Morphology
The EV count was 5 orders of magnitudes smaller in samples prepared by EV-FF compared with control sMEV-UC (Table 1). Although this information is important for purposes of assay normalization, it must not be used to assess the yield of EV by FF because the volume of the starting material in the EV-FF is unknown. The size of exosomes and possibly non-EV complexes in 50P, 100P, and 100R was similar to that of sMEV-UC (Table 1), although the size distribution was more heterogeneous in EV-FF than in sMEV-UC. We used a size of 200 nm as cutoff in NS-300 analysis because it is the size expected for exosomes, small microvesicles, CN micelles, and small fat globules (
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines.
) and because the quality of Nanosight calibration decreased with increasing size of authentic nanoparticle standards (Figure 1). By using 200 nm as cutoff in Nanosight analysis, we accepted a selection bias because large EV were eliminated from the calculation of mean sizes. We assessed the presence of EV larger than 200 nm by using dynamic light scattering and transmission electron microscopy. The EV-FF and sMEV-UC had the following sizes (mean ± SD): 50P = 157 ± 122 nm, 100P = 45 ± 17 nm, 100R = 1,108 ± 138 nm, and control sMEV-UC = 115 ± 31 nm. The polydispersity index of EV in FF preparations and sMEV-UC were (mean ± SD) as follows: 50P = 0.5 ± 0.4, 100P = 0.6 ± 0.1, 100R = 0.9 ± 0.02, and control sMEV-UC = 0.2 ± 0.02. The polydispersity index suggests that EV-FF contained particles of multiple sizes, whereas sMEV-UC controls were homogeneous (Supplemental Figure S2, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). The most prominent EV population in preparations 50P, 100P, and 100R had sizes of 67 ± 113 nm, 1,068 ± 1,288 nm, and 940 ± 220 nm respectively. The zeta potential in EV-FF was −21 ± −1.9, −14 ± −4.4, and −14 ± −0.6 in 50P, 100P, and 100R, respectively, whereas the zeta potential was −8.3 ± −0.6 for sMEV-UC controls. Purification of EV-FF 50P and 100P by using iodixanol density gradient yielded no visible layers, whereas a distinct layer was visible in gradient fraction 11 of preparation 100R (Supplemental Figure S3, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). Fraction 11 corresponded to a density of 1.255 g/mL. The size was 602 ± 308 nm for EV in fraction 11 of preparation 100R (Supplemental Table S2, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). When sMEV-UC were purified by using ultracentrifugation and subsequent gradient centrifugation, they accumulated in gradient fractions 4 to 8, corresponding to a density of 1.1 to 1.19 g/mL and consistent with densities previously reported for sMEV (
). For transmission electron microscopy, proteins in EV were crosslinked with glutaraldehyde without stabilizing the lipid bilayer with tannic acid; few EV were detected in EV-FF, but EV were abundant in sMEV-UC (Figure 2A). The few sMEV that were detected in EV-FF had a spherical shape with regular contours as expected for exosomes stained with phosphotungstic acid (
Milk-derived exosomes (MDEs) have a different biological effect on normal fetal colon epithelial cells compared to colon tumor cells in a miRNA-dependent manner.
). When particles from EV-FF were stabilized with tannic acid, the number of particles were greater compared with glutaraldehyde alone (Figure 2B), suggesting that a substantial percentage of the EV detected in the nanotracking analysis were lipid globules (
Ultrastructural discrimination of lipid droplets and vesicles in atherosclerosis: Value of osmium-thiocarbohydrazide-osmium and tannic acid-paraphenylenediamine techniques.
J. Histochem. Cytochem.1988; 36 (2458408): 1319-1328
). As expected, the apparent count of EV in sMEV-UC was similar in glutaraldehyde-treated sMEV-UC compared with sMEV-UC treated with glutaraldehyde and tannic acid. Note that tannic acid–treated FF preparations stained positive when probed with 1% phosphotungstic acid, suggesting that EV were damaged and permeabilized during flow filtration (
). In contrast, sMEV-UC stained negative if probed with 1% phosphotungstic acid (i.e., shapes and contours were distinct in sMEV-UC compared with EV-FF samples;
Milk-derived exosomes (MDEs) have a different biological effect on normal fetal colon epithelial cells compared to colon tumor cells in a miRNA-dependent manner.
Table 1Extracellular vesicle (EV) counts, sizes, and proteins in tangential flow filtration (FF) preparations and small milk EV purified by serial ultracentrifugation (sMEV-UC); values are means ± SD (n = 3)
Figure 1Nanoparticle tracking analysis of extracellular vesicles (EV) purified by tangential flow filtration (EV-FF) and small milk EV purified by serial ultracentrifugation (sMEV-UC). Size distribution and concentration of EV are shown for tangential flow filtration (FF) fractions 50P (A), 100P (B), 100R (C) and control sMEV-UC (D). Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R). Data shown are representative of 3 independent experiments each analyzed in triplicate. Values = mean ± SD (n = 3).
Figure 2Transmission electron microscopy images of extracellular vesicles (EV) purified by tangential flow filtration (EV-FF) and small milk EV purified by serial ultracentrifugation (sMEV-UC). (A) Membrane proteins were crosslinked with glutaraldehyde before negative staining of the following samples with phosphotungstic acid of 50P (I), EV in 100P (II), EV in 100R (III), and control sMEV-UC (IV). Approximately 1.0 × 109 particles were fixed before staining. (B) Membrane proteins were crosslinked with glutaraldehyde and lipid membranes were stabilized with tannic acid before negative staining of following samples with phosphotungstic acid of 50P (I), EV in 100P (II), EV in 100R (III), and control sMEV-UC (IV). Approximately 1.0 × 109 and 5.0 × 106 particles from EV-FF and control sMEV-UC, respectively, were used. The EV-FF showed positively stained particles. White arrows indicate EV < 100 nm and red arrows indicate positively stained particles. Large frames are images that were magnified 20,000-fold, and inserts represent sections of the large frames that were magnified 50,000-fold. Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R).
The content of total protein, purified from an equal number of EV, was greater in 100R compared with sMEV-UC, whereas negligible protein was noted in 50P and 100P samples (Table 1, Supplemental Figure S4, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). When an equal number of particles was analyzed by untargeted LC-MS/MS, we detected 17, 14, 35, and 75 distinct proteins in 50P, 100P, 100R, and sMEV-UC, respectively (Supplemental Table S3 and Figure S5, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). The 10 most abundant proteins in 50P, 100R, 100P, and control sMEV-UC are shown in Table 2. In multiple reaction monitoring analysis, the exosome marker proteins CD9, HSP70, and TSG101 were detectable in EV-FF, albeit their abundance was 62 to 99% lower compared with control sMEV-UC (Table 3). Immunoblot analysis produced similar patterns. Classical exosome markers such as CD9, CD63, and CD81 were identified in control sMEV-UC, whereas negative exosome markers such as histone H3, β-integrin, and APOB were largely absent (Figure 3). In contrast, none of the classical exosome markers were detected in EV-FF.
Table 2The 10 most abundant proteins in extracellular vesicles (EV) purified by tangential flow filtration (FF) and small milk EV prepared by ultracentrifugation (sMEV-UC) using untargeted proteomics (n = 3)
Peptide count: peptide count were semiquantitatively assessed by MASCOT (https://www.matrixscience.com/). Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R).
Maximum confidence score of individual protein detected in control sMEV-UC.
50P
100P
100R
Control sMEV-UC
Q2UVX4
Complement C3
34
P80457
Xanthine dehydrogenase/oxidase
18
25
2,547
P24627
Lactotransferrin
3
23
1,229
Q95114
Lactadherin
1
14
2,325
P02662
Alpha-S1-casein
5
4
5
9
19,237
P02754
Beta-lactoglobulin
4
2
8
9
23,432
P02666
Beta-casein
3
2
2
6
12,562
P02668
Kappa-casein
3
2
6
8,582
P11151
Lipoprotein lipase
18
6
894
P81265
Polymeric immunoglobulin receptor
3
6
905
1 Peptide count: peptide count were semiquantitatively assessed by MASCOT (https://www.matrixscience.com/). Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R).
2 Maximum confidence score of individual protein detected in control sMEV-UC.
Peak areas of multiple reaction monitoring response peptide per protein are expressed as percent of small milk extracellular vesicles prepared by ultracentrifugation (sMEV-UC) to compare extracellular vesicles (EV) purified by tangential flow filtration (EV-FF) and control sMEV-UC; values are means ± SD (n = 3).
Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R).
(% of sMEV-UC)
50P
100P
100R
sMEV-UC
50P
100P
100R
sMEV-UC
CD9
EEVIK
309.2
236.1
3.7 ± 3.1
5.7 ± 2.3
92.5 ± 14
156.0 ± 29.5
2.4
3.6
59.3
100
HSP70
DNNLLGR
267.8
175.1
84.3 ± 60
83.3 ± 11.0
177.5 ± 26
1,353.7 ± 132.3
6.2
6.2
13.1
100
TSG101
TAGLSDLY
420.2
205.6
148.0 ± 21
320.0 ± 4.0
95.5 ± 11
402.7 ± 11.2
36.8
79.5
23.7
100
1 Peak areas of multiple reaction monitoring response peptide per protein are expressed as percent of small milk extracellular vesicles prepared by ultracentrifugation (sMEV-UC) to compare extracellular vesicles (EV) purified by tangential flow filtration (EV-FF) and control sMEV-UC; values are means ± SD (n = 3).
2 Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R).
3 Uniprot identifiers: P30932 for CD9; Q27975 for HSP70, and A3KN51 for TSG101.
Figure 3Immunoblots of conventional extracellular vesicles (EV) markers in tangential flow filtration (FF) preparations and small milk EV purified by serial ultracentrifugation (sMEV-UC). Total protein from an equal number of EV was loaded per lane. Gels are representative of 3 independent experiments. Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R).
An average of ~9.1 million reads was obtained for the libraries, ranging from 7,109,610 to 11,046,276 raw reads per sample. After performing quality control and read-filtering, an average of ~4.6 million remained for mapping to the bovine genome. On average, 508,660 reads were successfully mapped, resulting in the identification of 139 mature microRNA for all samples combined (Table 4; Supplemental Table S4, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). MicroRNA were less abundant in FF samples compared with control sMEV-UC (Figure 4A). Members of the let-7 family were abundant in sMEV-UC, which is consistent with previous reports suggesting that these microRNA are abundant in both bovine and human milk exosomes (Figure 4B;
). In principal component analysis, the microRNA detected in EV-FF clustered separately from the microRNA in control sMEV-UC (Figure 5). The number of distinct mature microRNA detected in all 3 biological replicates was 4.5 to 9 times lower in EV-FF compared with control sMEV-UC (Table 5; Supplemental Figure S6A, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). Only 15 microRNA overlapped between control sMEV-UC and EV-FF (Supplemental Figure S6A). If accepting any microRNA detected in at least 1 of the 3 replicates as real, up to 60 microRNA were identified in EV-FF preparations, whereas 111 microRNA were detected in control sMEV-UC (Table 5). More than half of microRNA identified in control sMEV-UC were absent in EV-FF (Supplemental Figure S6B). The 10 most abundant microRNA in control sMEV-UC accounted for 77% of the total microRNA, whereas the same 10 microRNA accounted for only 30%, 19%, and 46% of the total microRNA in 50P, 100P, and 100R, respectively (Table 6). We identified 20, 16, 19, and 17 novel microRNA in all biological replicates of 50P, 100P, 100R, and control sMEV-UC, respectively (Supplemental Table S5, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
Table 4Sequencing data: number of raw, filtered and mapped reads in biological replicates of extracellular vesicles (EV) purified by tangential flow filtration (FF) preparations and small milk EV purified by serial ultracentrifugation (sMEV-UC) preparations and control sMEV-UC; values are raw counts
Figure 4MicroRNA profiles in extracellular vesicles (EV) purified by tangential flow filtration (EV-FF) and small milk EV purified by serial ultracentrifugation (sMEV-UC). Cluster analysis (A) and heatmap (B) of the 30 most abundant microRNA in control sMEV-UC compared with EV-FF. Data shown are representative of n = 3 independent samples. Columns and rows denote samples and microRNA, respectively. Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R).
Figure 5Principal component (PC) analysis of mature microRNA in extracellular vesicles purified by tangential flow filtration (EV-FF) and small milk EV purified by serial ultracentrifugation (sMEV-UC).
Table 5Number of microRNA identified in extracellular vesicles (EV) purified by tangential flow filtration (FF) preparations and control small milk EV purified by serial ultracentrifugation (sMEV-UC) samples (n = 3)
Table 6Cumulative abundance of the 10 most abundant microRNA in control small milk extracellular vesicles prepared by ultracentrifugation (sMEV-UC) compared with extracellular vesicles purified by tangential flow filtration (FF) preparations (n = 3); sMEV-UC were used as reference when identifying abundant microRNA
After miRDeep2 analysis, individual microRNA were ranked according to their abundance in control sMEV-UC in comparison to FF samples. Cumulative percentages were calculated based on raw counts.
Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R).
50P
100P
100R
sMEV-UC
bta-let-7a-5p
4.9
4.5
11.4
37.7
bta-miR-320a
7.9
6.2
27.0
51.3
bta-miR-200c
8.1
6.8
31.2
57.1
bta-let-7c
20.4
9.2
37.4
61.0
bta-let-7f
25.0
12.4
38.1
65.1
bta-miR-11975
25.0
12.4
39.1
67.0
bta-miR-92a
27.7
14.6
41.1
70.2
bta-miR-423–5p
27.8
15.1
43.8
73.2
bta-miR-26a
29.5
17.8
45.2
76.0
bta-let-7g
30.8
19.0
46.4
77.4
1 After miRDeep2 analysis, individual microRNA were ranked according to their abundance in control sMEV-UC in comparison to FF samples. Cumulative percentages were calculated based on raw counts.
2 Permeate (50P) and the retentate (50R) were subjected to FF using a membrane with a 100-kDa cutoff, yielding a permeate (100P) and retentate (100R).
Despite the use of 20 times more EV when extracting total RNA from EV-FF compared with control sMEV-UC (2 × 1010 vs. 1 × 109), the yield of RNA was 2.5 to 16 times lower for FF preparations (ng/µL): 50P = 0.63 ± 0.06, 100P = 0.50 ± 0.33, 100R = 3.30 ± 1.50, and control sMEV-UC = 8.00 ± 1.02. The RNA from preparations 50P and 100P did not pass the quantity and quality check and were excluded from sequencing analysis. The 100R and control sMEV-UC yielded 30,709,769 ± 14,834,86 and 28,421,723 ± 14,141,62 clean reads, respectively. We found 21 and 59% of reads in 100R and control sMEV-UC, respectively, mapped uniquely to the bovine genome. The read counts that mapped to genes was lower in 100R (14,628 genes) than in control sMEV-UC (18,520 genes; Figure 6A; Supplemental Table S6, https://data.mendeley.com/datasets/768cwtj8h4/1,
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
). The 2 preparations had in common 14,190 genes. Top 30 most abundant genes in control sMEV-UC compared with EV-FF is shown in Figure 6B. The genes identified in 100R clustered separately from the genes in sMEV-UC (Figure 7).
Figure 6Gene profiles in 100R and small milk extracellular vesicles purified by serial ultracentrifugation (sMEV-UC). Cluster analysis (A) and heatmap (B) of the 30 most abundant genes in control sMEV-UC compared with 100R. Data shown are representative of n = 3 independent samples. Columns and rows denote samples and genes, respectively. Permeate and the retentate were subjected to tangential flow filtration using a membrane with a 100-kDa cutoff, yielding a permeate and retentate (100R).
Figure 7Principal component (PC) analysis of genes in 100R and small milk extracellular vesicles purified by serial ultracentrifugation. Permeate and the retentate were subjected to tangential flow filtration using a membrane with a 100-kDa cutoff, yielding a permeate and retentate (100R).
This is the first report assessing the merit of purifying sMEV-UC from byproducts of cheesemaking by using FF. Our findings suggest that the preparation of EV by ultracentrifugation, a widely used strategy for enriching exosomes (
), and FF yielded distinct populations of EV. Although EV isolated by ultracentrifugation of milk tested positive for exosome marker proteins, these proteins were below the limit of detection in EV isolated from whey by using FF. Principal component analysis of microRNA and mRNA signatures are consistent with the conclusion that EV-FF are distinct from sMEV-UC controls. The RNA in EV-FF preparations clustered discreetly from those in sMEV-UC controls. The EV other than exosomes contain RNA (
). We speculate that a large proportion of mRNA signatures observed in preparation of the 100R fraction was associated with milk fat globules.
The low abundance of sMEV in EV-FF is somewhat surprising because FF has been successfully used in the isolation of exosomes from cell culture supernatants (
). We offer the following, not mutually exclusive, interpretations. First, the low abundance of sMEV in EV-FF preparations might be due to a low sMEV count in the starting material, acidic whey. In that scenario, FF might have led to an enrichment of sMEV but without reaching a point at which sMEV marker proteins became detectable. Previous studies have reported the presence of sMEV in whey, prepared by depleting cell debris and fat by centrifugation at 21,500 × g and filtration through a 0.22-µm membrane (
). These studies need to be interpreted with caution as sMEV were not authenticated following the protocols that are now the accepted standard in the EV community (
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines.
Second, it is important to recognize that the starting material used to obtain FF preparations was distinct from that used to obtain sMEV-UC. For example, the whey fraction from making cottage cheese was fermented in a dairy plant, whereas sMEV-UC samples were obtained from undiluted milk that was not fermented. With that said, sMEV-UC provide a reference for properties and cargos when assessing whether FF preparations contained exosomes in large numbers, but are not a reference when assessing the yield of exosomes in FF preparations.
Third, we speculate that number of non-sMEV, sizes less than 200 nm, greatly exceeded the number of sMEV in FF preparations. Candidates include lipoproteins, small microvesicles, fat globules, and perhaps CN micelles that escaped precipitation in cheesemaking (
). Butyrophilin was abundant in EV-FF, suggesting that a large fraction of EV in these preparations were fat globules. In addition, homogenization causes changes in the size and composition of particles in milk (
). In that scenario, the great number of non-sMEV outcompetes sMEV in downstream analyses. The FF protocol used here might require an additional delipidation step to increase the percentage of sMEV in EV-FF fractions.
It is worthwhile to point out that control sMEV-UC showed characteristics expected for milk exosomes. For example, classical markers of exosomes were detected in control sMEV-UC including CD9, CD63, CD81, HSP70, TSG101, and annexin (
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines.
). Control sMEV-UC also contained other proteins previously reported for milk exosomes (e.g., NADH-cytochrome b5 reductase, β-1,4-galactosyltransferase, CD36, lactoperoxidase, RAB18, MUC1, sodium-dependent phosphate transport protein, polymeric immunoglobulin receptor, and lactotransferrin and xanthine dehydrogenase/oxidase;
A limitation of this study is that we did not identify the identity of EV in preparations 50P and 100P, despite extensive characterization efforts. In contrast, our analyses suggested that milk fat globules are the quantitatively most important particle in preparation 100R. Given that preparation 100R was obtained from preparation 50R in FF, we speculate that 50R contains milk fat in large quantities. We can say with certainty that the expression of conventional exosome markers is low yet detectable in all FF preparations; therefore, sMEV-UC are of low abundance in these preparations.
We conclude that the content of sMEV is low when using the whey fraction from making cottage cheese as starting material and FF as purification protocol. That said, FF is a proven technology for isolating exosomes from cell cultures (
), and it is possible that the content of sMEV is low in milk whey compared with milk. One might ask what the potential use might be for FF fractions of whey from cheese making. Although anhydrous milk fat is used in bakery products and confectionary, we remain to be convinced that the cost of purifying milk fat by FF is economically rewarding.
ACKNOWLEDGMENTS
J. Z. conceived the idea. B. T. developed the FF protocol and provided samples. J. Z. and J. A. designed the research. S. S. wrote draft versions of the paper. S. S. and C. P. B. conducted research. S. S., C. P. B., T. T. A., J. A., J. C., and J. Z. analyzed data. J. Z. wrote the final version of the paper and had primary responsibility for final content. All authors read and approved the final manuscript. The granting agencies had no influence on the study design; the collection, analysis, and interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication. The authors acknowledge the use of the Biomedical and Obesity Research Core (NIH 1P20GM104320; NIH, Bethesda, MD) and the assistance of You Zhou and Julia Russ in the Nebraska Center for Integrated Biomolecular Communication (P20 GM113126), both at the University of Nebraska-Lincoln, and the services provided by the DNA Sequencing Core at the University of Nebraska Medical Center (NIH P20GM103427, 1P30GM110768, and P30CA036727). The authors acknowledge the Holland Computing Center at the University of Nebraska-Lincoln for providing computational support. This project was supported by Land O'Lakes, Inc. (Gray Summit, MO; no grant number available), The National Institutes of Health under grant 1P20GM104320, NIFA (National Institute of Food and Agriculture, Kansas City, KS) under grant 2016-67001-25301, USDA (Kansas City, KS) Hatch under grant 1011996, and USDA multistate group under grant W-4002. The authors declare no conflict of interest. J. Z. serves as consultant for PureTech Health Inc. (Boston, MA).
REFERENCES
Abels E.R.
Breakefield X.O.
Introduction to extracellular vesicles: Biogenesis, RNA cargo selection, content, release, and uptake.
Concentrations of purine metabolites are elevated in fluids from adults and infants and in livers from mice fed diets depleted of bovine milk exosomes and their RNA cargos.
MicroRNAs are absorbed in biologically meaningful amounts from nutritionally relevant doses of cow's milk and affect gene expression in peripheral blood mononuclear cells, HEK-293 kidney cell cultures, and mouse livers.
Ultrastructural discrimination of lipid droplets and vesicles in atherosclerosis: Value of osmium-thiocarbohydrazide-osmium and tannic acid-paraphenylenediamine techniques.
J. Histochem. Cytochem.1988; 36 (2458408): 1319-1328
A diet defined by its content of bovine milk exosomes and their RNA cargos has moderate effects on gene expression, amino acid profiles and grip strength in skeletal muscle in C57BL/6 mice.
Cushioned-density gradient ultracentrifugation (C-DGUC): A refined and high performance method for the isolation, characterization, and use of exosomes.
Milk-derived exosomes (MDEs) have a different biological effect on normal fetal colon epithelial cells compared to colon tumor cells in a miRNA-dependent manner.
Supplemental Data_JDS_“Isolation of extracellular vesicles from byproducts of cheese making by tangential flow filtration yields heterogeneous fractions of nanoparticles,” Mendeley Data, V1.
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines.
The intestinal transport of bovine milk exosomes is mediated by endocytosis in human colon carcinoma caco-2 cells and rat small intestinal IEC-6 cells.
Dietary Depletion of milk exosomes and their microRNA cargos elicits a depletion of miR-200a-3p and elevated intestinal inflammation and CXCL9 expression in Mdr1a−/− mice.
00710-4&id=gr1.jpg){kind=link}
00710-4&id=gr2.jpg){kind=link}
00710-4&id=gr3.jpg){kind=link}
00710-4&id=gr4.jpg){kind=link}
00710-4&id=gr5.jpg){kind=link}
00710-4&id=gr6.jpg){kind=link}
00710-4&id=gr7.jpg){kind=link}