One-week storage of refrigerated bovine milk does not affect the size, concentration, and molecular properties of extracellular vesicles

Milk extracellular vesicles (EVs) have gained extensive attention as promising diagnostic and therapeutic tools. Pre-analytical raw milk storage at low temperatures is an ordinary and usually necessary step after sample collection. It is known that direct freezing of unprocessed whole milk contaminates the native pool of milk EVs with other cell structures. However, less evidence is available regarding prolonged cooling at 4°C. The current study assessed whether pre-analytical storage of bovine raw milk for several days affected EV isolation and further analysis. To confirm the independence from the health status of the mammary gland, we analyzed milk samples stored at 4°C for 1, 2, 3, and 7 d past collection, respectively, from 2-quarters of the same cow with different somatic cell counts (SCC). Seven days of refrigeration did not change the milk EVs’ size, concentration, or morphology. We neither detected changes in the EV cargo regarding the amount of protein and RNA nor the specific EV markers TSG101, CD9, and CD81 in milk from quarters with high and low somatic cell counts. Overall, we observed fewer CD81 and CD9 markers in quarters with a high SCC. Moreover, there was no reduction in the mastitis-related miRNA bta-miR-223–3p, suggesting that refrigeration for several days up to one week is a possible storage option compatible with further EV analyses. The findings of this study enhance the confidence that milk EVs are highly stable in the raw milk matrix.


INTRODUCTION
Extracellular vesicles (EVs) are considered exploitable tools for diagnostic and therapeutic purposes in humans and livestock since their molecular composition reflects the physiological status of the producing cell (Ciferri et al., 2021).In particular, monitoring the inflammatory-related microRNA (miRNA) level could be critical for early disease diagnostic and prognostic observation (Jiao et al., 2021).It has been suggested that EVs in milk can provide helpful information about the mammary gland's cellular turnover and inflammatory status (Pollott et al., 2016).Previous publications demonstrated that the cargo of milk EVs changes upon Staphylococcus aureus stimulation in cattle.In particular, the EVs were significantly enriched in proteins (Reinhardt et al., 2013) and miRNA related to the immune response (de Montera et al., 2013;Sun et al., 2015;Cai et al., 2018;Ma et al., 2019;Stefanon et al., 2023).In this line, we have recently observed that chronic subclinical mastitis induces changes in milk EV miR-NA cargo, especially in the miRNA bta-miR-223-3p (Saenz- de-Juano et al., 2022).We, therefore, proposed bta-miR-223-3p as a potential indicator of subclinical mastitis progression and chronicity (Saenz-de-Juano et al., 2022).The miR-223 has exhibited multiple regulatory functions during the inflammation of various cell types, including granulocytes, macrophages, dendritic cells, T cells, endothelial cells, and epithelial cells (reviewed by (Jiao et al., 2021)).MiR-223 can regulate a pro-or anti-inflammatory response depending on the cell type and damage (Jiao et al., 2021).Specifically, in bovine mammary epithelial cells, bta-miR-223-3p likely mitigates the inflammatory progression by targeting CBLB and inhibiting the downstream PI3K/ AKT/NF-κB pathway (Han et al., 2020).
During an intra-mammary infection, immune cells are released from the bloodstream into the milk to fight the microorganisms and repair tissue damage (Wellnitz and Bruckmaier, 2012;Alhussien and Dang, 2018).The increase of immune cells within the mammary gland can be detected in the milk by performing a somatic cell count (SCC) analysis.In a previous experiment, we observed that subclinical mastitis did not induce changes in milk EVs' size or concentration.However, apart from more cells, subclinical mastitic milk contains higher ion concentrations and proteolytic enzymes caused by increased vascular permeability (Duarte et al., 2015), which could alter milk EV properties and cargo during prolonged storage.
In humans, it has been observed that freezing unprocessed breast milk samples led to the destruction of milk cells and contamination of the natural milk EVs population by the consequent formation of cell debris (Zonneveld et al., 2014).Therefore, for efficient recovery and reliable analysis of EVs post-sampling, it has been recommended to freeze milk supernatant devoid of cells and fat layer (Zonneveld et al., 2014).However, immediate sample processing is not always possible, and little is known regarding the effect of 4°C storage before skim milk freezing.In bovine colostrum, it has been shown that removing somatic cells, cream, and fat globules before long-term storage is necessary to obtain a higher yield of milk extracellular vesicles (Wijenayake et al., 2021).The problems associated with direct freezing might be related to the thawing process, where the milk cells can lyse and release proteases or RNases, which may affect the integrity and cargo of milk extracellular vesicles (Wijenayake et al., 2021).
Given the emerging role of milk EVs as potential diagnostic biomarkers in humans and animals, the need for storage of milk samples after collection, and the changes in milk properties due to inflammation, the current study aimed to evaluate the effect of preanalytical refrigeration on raw bovine milk EVs morphology, size populations, specific EV protein markers (TSG101, CD81, and CD9) and bta-miR-223-3p cargo in quarters with different SCC.

Animals
Five dairy cows were selected according to their SCC from the herd at Agrovet Strickhof (Eschikon-Lindau, Zurich).For their inclusion in the trial and to avoid cow biases, the cows had to present a quarter with Low SCC (50.000 cells/ml) and a quarter with High SCC (>150.000cells/ml) with a positive California Mastitis Test (CMT).Cows were in early or mid-lactation [DIM = 158.4 ± 45.73;mean ± SEM] and had between one and 6 parities.The cow's characteristics are shown in Supplementary Table S1.The study was approved by the Ethical Committee of Animal Experiments, Canton of Zurich (Switzerland) (ZH089/18).The study was carried out in compliance with the ARRIVE guidelines.

Milk collection, refrigeration storage, and processing before freezing
Foremilk milk samples (200 mL) were collected during the afternoon milking routine from High SCC and Low SCC quarters (Figure 1a), aliquoted, and stored at 4°C.Immediately after collection (d 0), the SCC was determined with a DeLaval cell counter (DCC, DeLaval).The first aliquot was processed 18 h after collection (d 1), the second after 42 h (d 2), the third after 66 h (d 3), and the last one 162 h after collection (d 7).Before processing each aliquot, the SCC was determined again.Then, samples were centrifuged at 300 g for 15 min at 4°C to remove milk cells.Finally, samples were centrifuged at 3000 g (3K) to remove the fat layer and the cell debris before freezing at −20°C.

Extracellular vesicles isolation
The EVs were isolated using differential centrifugation, acid precipitation, and ultracentrifugation as previously described with minor modifications (Saenz- de-Juano et al., 2022).Briefly, skim milk samples were centrifuged at 12'000 g (12K) for 1 h at 4°C.Then, to precipitate and separate the casein proteins from the whey, 250 µL of acetic acid (Sigma-Aldrich) was added to 25 mL of skim milk and centrifuged at 10'000 g for 15 min at 4°C.The whey was then filtered with a 0.45 µm filter and ultracentrifuged 210'000 g (210K) for 70 min at 4°C using a Beckman Coulter Optima XE-90 Ultracentrifuge with a 50.2Ti rotor.Then, the pellet was washed with phosphate buffer saline (PBS, Thermo Fisher) and ultracentrifuged 210'000 g again for 70 min at 4°C.The EV pellet was resuspended in 500 µL of PBS and centrifuged at 10'000 g for 5 min at 4°C.The supernatant was divided into aliquots and stored at −80°C.All the analyses were performed within one month after EV isolation.

Transmision Electronic Microscopy (TEM)
Extracellular vesicle visualization was performed by the Scientific Center for Optical and Electron Microscopy (ScopEM) service of ETH Zurich.Briefly, 3 microliters of the vortexed dispersion were placed on glow-discharged carbon-coated grids (Quantifoil, D) for 1 min, and negative contrast staining was done in 2% sodium phosphotungstate pH 7.2 for 1 s, followed by a second step for 15 s.After draining the excess moisture with filter paper, the air-dried grids were imaged in a TEM Morgagni 268 (Thermo Fisher) operated at 100 kV.

Tunable Resistive Pulse Sensing (TRPS)
The concentration and size of isolated EVs were evaluated using qNano Gold (Izon Science) with an NP400 Nanopore (Izon Science).Calibration particles (CPC 400, Izon Science) and EV samples were diluted with filtered PBS in a ratio of 1:300 and 1:100, respectively.At least 1000 particles were measured in each sample.Particles were measured in 4 independent sessions using a 47.0 mm stretch of the nanopore with a voltage

Western Blot
Proteins were extracted using radioimmunoprecipitation assay buffer (RIPA, Thermo Fisher) with a 100X antiprotease cocktail (Thermo Fisher).Protein quantification was performed with the BCA protein assay kit (Thermo Fisher).
Western blots were carried out with 20 µg of protein.For TSG101, ANXA5, and Calnexin analysis, samples were reduced by adding 10% of β-mercaptoethanol (BioRad).After 10 min of incubation at 70°C, samples were run on an SDS-PAGE StainFree gel (BioRad) for 35 min at 200 V.The gel was activated by UV transillumination for 1 min using the ChemiDoc MP Imaging System (Biorad).Proteins were transferred from the gel to the nitrocellulose membrane using the TransTurbo system (BioRad) and the preassembled Trans-Blot Turbo Transfer Packs (Biorad).Blocking the membrane was performed with 5% skim milk (Sigma) in TBS-T buffer (TBS buffer containing 0.05% Tween 20, Biorad) for 1 h.Primary antibodies diluted in blocking buffer (Supplementary Table S2) were added to the membrane and incubated overnight at 4°C.After 3 washes of 5 min with TBS-T buffer, the membrane was incubated for 1 h with the secondary antibody.Following 3 more washes with TBS-T, an antibody signal was developed using the ClarityTM Western ECL kit (Biorad), and images were acquired using the ChemiDoc MP Imaging System.

RNA extraction and RNase protection assay
RNA was extracted from 200 µL of isolated milk EV using the miRNeasy Micro Kit (Qiagen).Afterward, RNA quantity was determined with Quantus Fluorometer and the QuantiFluor® RNA System kit (Promega).We performed an RNase protective assay to confirm that the majority of the extracted RNA was confined within the EVs (Almiñana et al., 2021).A pool of isolated EV aliquots was incubated with RNase A (PeqLab) to a final concentration of 0.1 mg/mL for 30 min at room temperature to degrade unprotected RNA.At the same time, another aliquot was treated with RNase (0.1 mg/mL) and 1% Triton X-100 (Sigma-Aldrich).One untreated aliquot was used as a control.After all the treatments (RNase/Triton +/+; +/−; −/−), EV samples were directly transferred to Qiazol (Qiagen), and the RNA isolation protocol was performed following the manufacturer's instructions.Afterward, the RNA profiles of selected samples were determined using the BioAnalyzer 2100 (Agilent Technologies) and the Picochip protocol (Agilent Technologies).

Statistical analysis
Statistical analyses were performed using GraphPad Prism 8.2 software.Logarithmic transformation was conducted if the gene expression values were not normally distributed, as verified by a Shapiro-Wilk test.A multiple linear regression test was performed to analyze EV concentration, size and protein amount, protein EV markers, RNA quantity, and bta-miR-223-3p ∆Cq.Differences were considered significant if P < 0.05.In the box-whisker plots, the line across the middle of the box shows the median, the upper and lower extremities of the box show the 25th and 75th percentile of the set of data, and the upper and lower black whiskers show the 5 and 95 percentiles.

Effect of cold storage on raw milk SCC
No significant differences were detected in the milk SCC between the time point of collection (d 0) or d 1, 2, 3, and 7 (Figure 1b-f).The coefficient of variation ranged between 4.78 and 6.13% for high SCC quarters and 13.81-74.33%for low SCC quarters.

Milk EVs characterization with TEM, TRPS, and WB
The morphology, integrity, and heterogeneity of EVs were confirmed by TEM on d 1, 2, 3, and 7 (Figure 1g), respectively.We observed populations of different-sized vesicles resembling EVs and shapeless aggregations with low electron density (Sedykh et al., 2019).The mean particle diameter was 161.3 ± 13.85 nm (mean ± SD) for High SCC and 160.1 ± 16.93 nm for Low SCC.Western blot analysis showed that EVs were positive for known markers such as TSG101, CD81, CD9, and ANXA5 (Figure 1h) (Théry et al., 2018).The absence of calnexin in EV samples, but not in the mammary gland cell lysates (Figure 1h), indicated the absence of cellular contamination and that most EVs belonged to the small subtype population (Théry et al., 2018).

Effect of refrigerated/cold storage on the particle size distribution and concentration
The particle size distribution analysis revealed that most of the particles in both High SCC and Low SCC quarters had a diameter below 200 nm (Figure 2a,b).Over the 7 d of storage, no significant differences in particle size diameter were observed (Figure 2c).No effects were found regarding the cow, the TRPS measurement session, or the SCC.Similarly, no significant differences were observed in the particle concentration of isolated EVs (Figure 2d), and no effect of cow, TRPS measurement session, or SCC was detected.Finally, as observed by TEM, despite the cold storage before the EV isolation, we did not observe differences regarding the preservation of EV morphology, aggregates, or impurities in the samples (Supplementary Figure S1).

Effect of refrigerated/cold storage on milk EV protein abundance and EV markers TSG101, CD81, and CD9
An estimated EV quantification was performed based on protein concentration (Théry et al., 2018;Almiñana et al., 2021).We observed no significant differences in protein concentration due to the storage time in 3K, 12K, or 210K pellets (Figure 3a-c).When we analyzed the specific amount of TSG101, CD81, and CD9, we observed no significant differences regarding the day of storage (Figure 3d-f).However, in the case of TSG101, we found a statistically significant cow effect, where Cow 4 and Cow 5 had a lower amount of TSG101 than the reference cow (Cow 1, Figure 3g).For CD9 and CD81, there was also a significant effect on the cow and the SCC.In both cases, all cows differed from the reference cow, and the High SCC quarters had fewer EV markers than the Low SCC quarters (Figure 3h, i).

Effect of prolonged cold storage at 4°C on RNA abundance
The RNase protective assay confirmed that most of the extracted RNA was confined within the EVs.After the RNase treatment, most of the RNA remained intact, while the 18S and 28S ribosomal RNA peaks disappeared (Figure 4a,b).Then, when we added Triton X-100 with the RNase, the EV membranes were disrupted, and the RNA released was degraded by the RNase (Figure 4c).The Bioanalyzer profiles likewise indicated that most of the RNA belonged to the small RNAs population, while larger RNAs' contribution was minimal.This distribution of RNA did not significantly change throughout the storage (Figure 4d, Supplementary Figure S3a).
As measured by the Quantus TM fluorometer, the concentration of extracted RNA ranged between 4.9 and 42.6 ng/µL with no effect on the storage time.The Bioanalyzer measurements confirmed this result on d 1 and d 7 samples (Figure 4e).There was also no effect of the SCC on the RNA concentration, but there was a cow effect where the RNA isolated from Cow 2 was significantly lower than the rest (Figure 4f).

Effect of refrigerated/cold storage on specific miRNAs in High SCC quarters
The relative expression of bta-miR-223-3p in milk EV from quarters with High SCC was evaluated with RT-qPCR.We observed that the day of storage did not affect the raw Cq of bta-let7a-5p, bta-miR-200c-3p, and bta-miR-223-3p (Supplementary Figure S3b), or the relative amount of bta-miR-223-3p when using the geometric mean of bta-let7a-5p and bta-miR-200c-3p as reference (Figure 4g).

DISCUSSION
Immediate centrifugation after milk collection on the farm is rather difficult or impossible from a practical point of view.Direct freezing of unprocessed milk at −80°C reduces the native pool of EVs in human breast milk (Zonneveld et al., 2014;Leiferman et al., 2019) and bovine colostrum (Wijenayake et al., 2021).Thus, refrigeration at 4°C is the preferred storage option.However, keeping the raw milk for several days raises the concern that apoptotic cells might disintegrate, forming membrane-enclosed vesicles and contaminating the pool of naturally present milk EVs (Zonneveld et al., 2014).In addition, milk contains enzymes such as proteases and lipases (Fox and Kelly, 2006) that could degrade the native bulk EVs and difficult the detection of possible milk EV biomarkers.Thus, evaluating the effect of a prolonged bovine raw milk storage period at 4°C for several days before EV analyses is mandatory.We chose 18 h as a starting point to simulate an actual waiting of the collected samples before reaching the laboratory and being analyzed.
We did not observe a significant variation in the SCC after 7 d at 4°C.The DCC Delaval equipment individually assesses the somatic cells by identifying nuclei stained with propidium iodide (PI) (Gonzalo et al., 2006).Therefore, we could assume no significant upregulation of nuclei fragmentation or cell proliferation during cooled storage.These results agree with previous observations confirmed by flow cytometry that in bovine milk, somatic cell viability remains stable through 72h of refrigeration at 4°C (Li et al., 2015).The authors suggested milk is a better conservation medium than PBS because milk resembles blood regarding osmotic de-Juano et al.: Effect of raw milk storage on EVs pressure, pH, and nutrient supply (Li et al., 2015).We tried to confirm the cell viability stability in our samples using the trypan blue exclusion assay (data not shown).However, the abundant presence of cellular debris, milk fat globules, and other impurities produced inaccurate counting results, as previously reported (Gleeson et al., 2022).
We suspected that the longer the cooling time and the higher the SCC, the larger the EVs detected due to the appearance of apoptotic bodies (ApoBD).We tried to determine the particle's size and concentration in 3K and 12K pellets, but the abundant presence of cellular debris and fat clogged the nanopore.However, the protein concentration of these pellets did not show a significant difference due to storage time.After the acid precipitation, we filtered the milk with 0.4 µm filters to allow a bigger range of particles to pass and reach the 210K pellet.Most recovered EVs were still lower than 200 nm.This result is in line with previous observations in human breast milk and commercial bovine milk, in which exosome-sized (<200 nm) counts did not significantly change after 2 weeks of 4°C storage (Leiferman et al., 2019).It has been observed that even the dead cells maintain their overall shape, likely keeping most of the enzymes within the intracellular compartment (Li et al., 2015).Indeed, even direct ultracold storage of unpasteurized milk did not influence isolated EVs' particle size, purity, or morphology after ultracentrifugation (Wijenayake et al., 2021).However, as other apoptotic vesicles comprise exosome and microvesicle sizes, ApoExo and ApoMV, respectively (Kakarla et al., 2020), the latter share many EV markers with the non-apoptotic vesicles.Since we did not test for any specific apoptotic marker, we cannot discard the pres- The storage time did not impact the amount of EV markers TSG101, CD9, and CD81.However, we found that the amount varied depending on the individual cow.This result agrees with our previous work, where we discovered a cow-individual variability reflected in the milk EV miRNA (Saenz-de-Juano et al., 2022, 2023).Because cows were selected depending on the SCC value, our statistical model did not include the individual cow's characteristics such as breed, age, milk yield, days of lactation, or parities.Indeed, further ex- periments should consider these to explain the resulting milk EV variance.
Furthermore, we observed a significant SCC effect on CD81 and CD9 protein amounts.These tetraspanins belong to the 100K milk EV subpopulation, mainly involved in galactosidase activity, glycosyl transferase, and peptidase activity (Benmoussa et al., 2019a).Therefore, it will be interesting to perform additional analyses that evaluate the CD81 and CD9-specific milk EV subpopulations and their enriched proteins to see whether their function can be perturbed during mastitis.
There were no differences in RNA quantification, and, as the Bioanalyzer profiles reflect, most of the RNAs belonged to the small RNAs population.The contribution of larger RNAs was minimal.Moreover, this RNA fragment distribution did not change with the storage duration.Therefore, we provide evidence that most isolated RNA was confined within the EVs by processing a pool of our samples with RNase and  Triton X-100 before RNA isolation and quantification (Almiñana et al., 2021).With the RNase protection assay, we saw how most RNA remained intact after the RNase treatment while the 18S and 28S ribosomal RNA peaks disappeared.When we added Triton X-100, the EV membranes were disrupted, and the RNA released was degraded by the RNase A. Our Bioanalyzer profiles were similar to those obtained for equine uterine EVs (Almiñana et al., 2021).
There is no established endogenous RNA control for ncRNA expression data obtained by qPCR in milk EVs.Thus, synthetic/exogenetic RNA is usually added as a control to normalize the ncRNA expression (Zeng et al., 2021).Milk EVs have been identified and characterized in many species, including cow, human buffalo, pig, wallaby, horse, camel, rat, and panda (van Herwijnen et al., 2018), showing remarkable conservation of specific miRNAs in milk EVs from different species (van Herwijnen et al., 2018;Benmoussa and Provost, 2019).Bta-let7a-5p is among the top 5 most abundant miR-NAs in humans, cows, and pigs' milk EVs (van Herwijnen et al., 2018;Zeng et al., 2021) and represents more than 15% of all miRNAs in bovine milk (Benmoussa et al., 2019b).is between the top 20, depending on the species (van Herwijnen et al., 2018).Our previous experiments observed that bta-let7a-5p and bta-miR-200c-3p were secreted by primary bovine mammary epithelial cells (Silvestrelli et al., 2021), and their amount did not change during subclinical mastitis (Saenz- de-Juano et al., 2022).Thus, we used them as a reference control for the qPCR.We only included High SCC milk EVs in the analysis since we demonstrated that bta-miR-223-3p is a subclinical mastitis biomarker not present in healthy milk EVs (Saenz- de-Juano et al., 2022).We did not observe any differences in the bta-miR-223 amount after 7 d of storage at 4°C.Moreover, we detected no miRNA as low abundant as > 35 Cq, confirming no additional miRNA degradation.Although it was not significant, the cow with the highest SCC had the highest bta-miR-223-3p expression, which agrees with considering bta-miR-223-3p as an EV marker for subclinical mastitis (Saenz- de-Juano et al., 2022).
In conclusion, the current study showed that refrigerated milk storage at 4°C did not affect milk SCC measurement, EV protein amounts, and the EV markers TSG101, CD9, and CD81.Importantly, across all sampling time points, we detected the mastitis-related miRNA bta-miR-223-3p.Therefore, rather than direct freezing, storage at 4°C for up to one week could be considered the preferred option for milk EV analysis when immediate centrifugation of the milk sample is not possible.
Figure 1.a) Experimental design.SCC: somatic cell count.b-f) Milk SCC of each cow's aliquots immediately after milking (d 0) or after 1, 2, 3, and 7 d of storage at 4°C.Data is shown on a logarithmic scale.g) Transmission Electron Microscope (TEM) images of milk extracellular vesicles derived from the same milk sample (exemplary Cow 3, Low SCC) at 1, 2, 3, and 7 d of storage at 4°C.Additional TEM images with higher magnification are shown in Suppl.Fig. S1.Black arrows indicate microvesicles and exosomes.H) Western blot images showing the presence of EV markers TSG101, tetraspanins CD9, CD81, and annexin 5 (ANXA5), and the absence of calnexin (CANX).L: Ladder; 1-10: Milk EVs pellets.MG: Mammary gland tissue; MC: Milk cells.Original complete Western blots are shown in Suppl.Fig. S2.
de-Juano et al.: Effect of raw milk storage on EVs Figure 2. a,b) Particle size distribution in milk from quarters with High and Low SCC, respectively.c) Mean particle diameter size in milk EVs after 1, 2, 3, and 7 d of storage at 4°C. d) Concentration of particles from milk stored in the fridge for 1, 2, 3, and 7 d of storage at 4°C.
Figure 3. a-c) Milk protein amount in 3K, 12K, and 210K pellets after 1, 2, 3, and 7 d of storage at 4°C. d-f) Milk EV protein mount of TSG101, CD81, and CD9 from milk stored in the fridge for 1, 2, 3, and 7 d of storage at 4°C. g-i) Milk EV protein mount of TSG101, CD81, and CD9 in each cow.
de-Juano et al.: Effect of raw milk storage on EVs

Figure 4 .
Figure 4. a) -c) Exemplary RNA Bioanalyzer profile of a) a non-treated EV sample; b) a sample treated with RNase but not Triton X-100, and c) a sample treated with Rnase and Triton X-100.d) RNA Bioanalyzer profiles of the same exemplary sample after storage at 4°C for 1, 3, and 7 d.e) RNA concentration in milk EVs of d 1 and d 7 samples obtained using the Quantus Fluorometer TM (Promega) and the Agilent 2100 BioAnalyzer (Agilent), respectively.f) RNA concentration in milk EVs from each cow using the Quantus Fluorometer TM .g) Relative gene expression of bta-miR-223-3p in milk EVs stored at 4°C for 1, 2, 3, and 7 d, respectively.The geometric average of bta-let7a-5p and bta-miR-200c was used as reference gene expression.