Human vascular endothelial cells transport foreign exosomes from cow's milk by endocytosis

Abstract

Encapsulation of microRNAs in exosomes confers protection against degradation and a vehicle for shuttling of microRNAs between cells and tissues, and cellular uptake by endocytosis. Exosomes can be found in foods including milk. Humans absorb cow's milk exosomes and deliver the microRNA cargo to peripheral tissues, consistent with gene regulation by dietary nucleic acids across species boundaries. Here, we tested the hypothesis that human vascular endothelial cells transport milk exosomes by endocytosis, constituting a step crucial for the delivery of dietary exosomes and their cargo to peripheral tissues. We tested this hypothesis by using human umbilical vein endothelial cells and fluorophore-labeled exosomes isolated from cow's milk. Exosome uptake followed Michaelis-Menten kinetics (V_max = 0.057 ± 0.004 ng exosome protein × 40,000 cells/h; K_m = 17.97 ± 3.84 μg exosomal protein/200 μl media) and decreased by 80% when the incubation temperature was lowered from 37°C to 4°C. When exosome surface proteins were removed by treatment with proteinase K, or transport was measured in the presence of the carbohydrate competitor d-galactose or measured in the presence of excess unlabeled exosomes, transport rates decreased by 45% to 80% compared with controls. Treatment with an inhibitor of endocytosis, cytochalasin D, caused a 50% decrease in transport. When fluorophore-labeled exosomes were administered retro-orbitally, exosomes accumulated in liver, spleen, and lungs in mice. We conclude that human vascular endothelial cells transport bovine exosomes by endocytosis and propose that this is an important step in the delivery of dietary exosomes and their cargo to peripheral tissues.

micrornas (miRs, miRNAs) are small noncoding RNA, which are encoded by their own genes, or introns and exons of long nonprotein-coding transcripts (9, 16a, 42); miRNAs may silence genes via destabilizing of mRNA or preventing translation of mRNA (15, 26). The online database miRBase, release 21, lists 1,881 “high confidence” human miRNAs (33a), and ∼60% of the miRNA-binding sites in the human genome are evolutionarily conserved (21). Gene regulation by miRNAs has been implicated in numerous physiological (48) and pathological (13) conditions in humans. Mammalian miRNAs are encapsulated in extracellular vesicles such as exosomes, thereby conferring protection against degradation (23, 30, 39, 57) and a pathway for cellular uptake by endocytosis (17, 19).

Until recently, miRNAs have been considered endogenous regulators of genes, i.e., miRNAs synthesized in a given organism regulate the expression of genes in that host. In a recent publication we have refuted this paradigm and provided strong evidence that 1) humans absorb biologically meaningful amounts of miRNAs from nutritionally relevant doses of cow's milk, 2) milk miRNAs are delivered to peripheral human tissues, 3) physiological concentrations of milk miRNAs affect human gene expression in vivo and in cell cultures, and 4) endogenous synthesis of miRNAs does not compensate for dietary miRNAs deficiency in mice (7). That paper was the first report suggesting that miRNAs can be transferred between distinct animal species through dietary means. Our discoveries were corroborated in a recent report by investigators from the National Institutes of Health (NIH)-supported Genboree database, who detected numerous dietary miRNAs in 6.8 billion sequencing reads from 528 human samples (27). We also detected bovine-specific miRNAs in human plasma following milk consumption using next generation sequencing (J. Cui and J. Zempleni, unpublished observations). Studies by us and an independent laboratory suggest that human and rat intestinal cells transport cow's milk exosomes by endocytosis (51) and that milk exosomes may cross the intestinal mucosa without repackaging in mice and enter the peripheral circulation in intact form (35). In contrast to the above observations in exosomes and miRNAs of animal origin, previous claims by a single laboratory that miRNAs from plants affect human gene expression (56, 58) are highly controversial and have largely been dismissed by the scientific community (7, 8, 14, 44, 49).

On the basis of these previous observations it is reasonable to propose that a fraction of dietary exosomes from foods of animal origin enters the peripheral circulation, and that mechanisms exist for the transfer of these exosomes across vascular endothelia. Here we tested the hypothesis that human vascular endothelial cells transport cow's milk exosomes by a carrier-mediated process similar to the mechanism reported for the uptake of exosomes in intestinal cells, i.e., endocytosis (51).

MATERIALS AND METHODS

Exosome isolation, characterization, and labeling.

Cow's milk (1% fat) was obtained from a local grocery store. The milk was centrifuged at 12,000 g at 4°C for 30 min to remove somatic cells and debris. The fat-free supernatant was mixed with an equal volume of 0.25 M EDTA (pH 7.0) and incubated on ice for 15 min to precipitate casein and exosomes coated with casein (34). The suspension was ultracentrifuged at 80,000 g at 4°C for 60 min (Sorvall WX Ultra 80, F37L-8 × 100 rotor; Thermo Scientific) to remove precipitated protein, milk fat globules, and microvesicles larger than exosomes. Exosomes were collected by centrifugation at 120,000 g at 4°C for 90 min, resuspended in sterile phosphate-buffered saline, and filtered through a 0.22-μm membrane filter (Milex). Sodium azide was added to produce a final concentration of 0.01%, and exosomes were stored at 4°C if used the same day or at −20°C for up to 5 days.

Exosome purity and absence of aggregation was assessed as recommended by the International Society for Extracellular Vesicles (33). Briefly, whole protein extracts from exosomes were resolved by gel electrophoresis (10 μg protein/lane) as described previously (4), and membranes were probed using mouse anti-bovine CD63 (catalog no. MCA2042GA; AbD Serotec), mouse anti-CD9 (catalog no. ab61873; Abcam), goat anti-bovine Alix (catalog no. sc-49268; Santa Cruz Biotechnology) as markers for exosomes, rabbit antiserum to bovine α-s1-casein as a marker for the species of exosome origin (41), and goat anti-bovine histone H3 (catalog no. sc-8654; Santa Cruz Biotechnology) as a negative control (all at 1,000-fold dilutions). Bands were visualized using an Odyssey infrared imaging system (Licor) and IRDye 800CW-labeled secondary antibodies (50,000-fold dilution; catalog nos. 926-32210, 926-32214, and 926-32211; Licor). Anti-bovine α-s1-casein was raised in rabbits (Cocalico) using acetyl-AHSMKEGIHAQQKEPMIGVGC-amide as antigen (casein sequence underlined), coupled to keyhole limpet hemocyanin through a C-terminal cysteine. The anti-serum produced bands of the expected size with cow's milk exosomes, cow's milk, and the peptide antigen, but not with negative controls such as human breast milk, an unrelated synthetic peptide [platelet glycoprotein 1, amino acid sequence acetyl-APQLYPNTFMQGILNSFIKG(C)-amide], and exosomes from chicken egg yolk (Fig. 1). Absence of exosome aggregation in our preparations was confirmed using transmission electron microscopy (Hitachi H7500, Japan) in the Microscopy Core Facility in the University of Nebraska-Lincoln. ImageJ (http://imagej.nih.gov/ij/index.html) was used to analyze the exosome size distribution.

Fig. 1.

Rabbit anti-bovine α-s1 casein and gel electrophoresis were used to probe membrane blots of 1) cow's milk exosomes, 2) cow's milk, 3) human breast milk, 4) platelet glycoprotein 1 synthetic peptide, 5) α-s1 casein peptide, and 6) chicken egg yolk exosomes. Ten micrograms of milk and exosome protein were loaded per lane, whereas only 1 μg of synthetic peptides were loaded. M, molecular weight markers.

For studies in cell cultures, exosomes were labeled with FM-464 (Molecular Probes) as described previously (47). Unbound fluorophore was removed by pelleting the exosomes at 120,000 g and 4°C for 90 min, followed by three wash and ultracentrifugation cycles with sterile phosphate-buffered saline. For studies in mice, exosomes were labeled with a cyanine-based fluorophore, 1,1′-dioctadecyltetramethyl indotricarbocyanine iodide (DiR) as described previously (50). Exosome integrity and absence of aggregation was confirmed by transmission electron microscopy. The concentration of exosome protein was measured using a Nanodrop-1000 spectrophotometer (NanoDrop Technologies), and exosomes were diluted with F-12K media to produce the desired protein concentration.

Cell culture.

Human umbilical vein endothelial cells (HUVEC, passages 38-45) were purchased from American Type Culture Collection (CRL-1730) and cultured in F-12K medium, supplemented with 0.04 mg/ml endothelial cell growth supplement, 0.1 mg/ml heparin, 100,000 U/l penicillin and 100 mg/l streptomycin (all from Sigma), and 10% exosome-free fetal bovine serum in a humidified atmosphere at 5% CO₂ and 37°C. Exosome-free fetal bovine serum was prepared by sonicating the serum in a water bath for 1 h to disrupt membranes, which granted milk RNases access to exosome RNAs and caused the degradation of 60%, 65%, and 86% of miR-15b, miR-21, and miR-200c, respectively. miRNAs were quantified in exosome extracts by using quantitative real-time PCR, U6 snRNA for normalization, and miSPIKE as internal standard (IDT Technologies), and miRNA-specific PCR primers (Table 1) as described previously (7). Media were replaced with fresh media every 48 h.

Table 1.

PCR primers used for the quantification of microRNAs

MicroRNA	Forward Primer*
miSpike	5′-CTC AGG ATG GCG GAG CGG TCT-3′
U6	5′-CGC AAG GAT GAC ACG CAA ATT-3′
miR-15b	5′-CAGCACATCATGGTTTACA-3′
miR-21	5′-GCTAGCTTATCAGACTGATGTTGA-3′
miR-200c	5′-TAA TAC TGC CGG GTA ATG ATG GA-3′

^*The reverse primer is a proprietary primer provided with the miScript II RT kit.

Binding studies.

In a typical experiment, 15 × 10³ HUVECs were seeded per well in a 96-well plate and allowed to adhere overnight. Fluorophore-labeled exosomes were added to the wells to produce the desired concentration of exosome protein. Cells were incubated for the lengths of time denoted in results. Media were removed and cells were washed three times with sterile PBS to remove extracellular exosomes. Controls were prepared by washing the cells immediately after addition of exosomes. Cell fluorescence (excitation 560 nm, emission 645 nm) was measured in a microplate fluorescence detector (BioTek). Cells were harvested using trypsin and counted using a hemocytometer. Units of fluorescence were converted into mass of exosome bound by labeling a known mass of exosomes (protein) with fluorophore, and quantifying the fluorescence after removing unbound fluorophore. In select experiments, we measured the effects of the following treatments on exosome binding: 1) cells were treated with 5 or 10 μg/ml of the endocytosis inhibitor cytochalasin D (GIBCO) for 30 min before adding exosomes (6); 2) cells were treated with 150 mM carbohydrate competitors d-glucose or d-galactose for 30 min before adding exosomes (40); and 3) milk exosomes were treated with 100 μg/ml of proteinase K at 37°C for 30 min to remove surface proteins (17). All assays were performed in three independent experiments, each in triplicate analyses. Binding kinetics were modeled using the Michaelis-Menten equation and nonlinear regression (GraphPad Prism 6.0; GraphPad Software, La Jolla, CA).

We formally excluded the remote possibility that adherence to cells, rather than uptake into cells, accounted for cell fluorescence using the following cell death-based and enhanced green fluorescent protein (eGFP) protocols. Milk exosomes were suspended in BTXpress electroporation buffer (BTX, Holliston, MA; final concentration 100 μg/μl protein), containing zero (negative control), 56 μM, or 223 μM puromycin in a 4-mm electroporation cuvette. Exosomes were loaded with puromycin or with a mammalian eGFP expression plasmid by electroporation in a Gene Pulser Xcell electroporator (Bio-Rad) using 250 V, 950 μF, and infinite resistance. Extra-exosomal puromycin was removed by ultracentrifugation and washing, and exosomes were resuspended in 300 μl of exosome-depleted cell culture media. In cell death assays, 100 μl of the suspension were added per well in a 96-well cell culture plate (containing ∼200 μl of media) and cells were cultured for 24 h, when viability was assessed using the MTT assay. Positive controls were created by adding puromycin directly to the media without encapsulation in exosomes, thereby producing a concentration of 1,837 μM puromycin. In eGFP assays, cells were cultured in media in which eGFP-loaded milk exosomes were substituted for exosomes in fetal bovine serum. Expression of eGFP was assessed by using confocal microscopy 3 days after initiation of cultures.

Distribution in mice.

We determined whether intravenously administered, DiR-labeled milk exosomes cross vascular endothelia cells and accumulate in tissues. Briefly, 1 × 10¹¹ DiR-labeled exosomes/g body wt were injected retro-orbitally (intravenously) in female C57BL/6 mice. The mice were 11 wk of age (∼25 g body wt) and were fed Teklad Global 16% Protein Rodent Diet (catalog no. Teklad 2016, Envigo). Controls were injected with free DiR or unlabeled exosomes. Eighteen hours after injection, we assessed the distribution of exosomes using an iBox small animal imaging system in live mice and excised tissues. Dissected issues were flushed with cold saline to remove circulating exosomes prior to imaging. The experiments in mice were approved by the Institutional Animal Care Program at the University of Nebraska-Lincoln (protocol no. 963).

Statistical analysis.

Homogeneity of variances was confirmed using Bartlett's test (16, 54). Statistical significance of differences among treatment groups was assessed using one-way ANOVA and Tukey-Kramer's or Dunnett's post hoc test. Analyses were performed using GraphPad Prism. Differences were considered significant if P < 0.05. Means ± SD are reported.

RESULTS

Our exosome purification protocol yielded preparations of nonaggregated extracellular vesicles that were primarily composed of exosomes. When protein extracts were probed with anti-CD63, anti-Alix, anti-CD9, or anti-bovine α-s1-casein, strong bands were observed in western blots; in contrast, when protein extracts were probed with anti-histone H3 (negative control), no band was visible (Fig. 2A). The particle suspension was largely free of aggregates, and the shape and contour of exosomes suggested vesicle integrity (Fig. 2B). The average particle size was 69 ± 19.5 nm in diameter, as expected for exosomes (33). A few particles were detected that had a diameter less than that of exosomes (Fig. 1); these particles probably represent small fat globules (25).

Fig. 2.

Milk exosome preparations from cow's milk. A: exosome extracts were probed using anti-CD63, anti-CD9, anti-Alix, anti-α-s1 casein, and anti-histone H3. Protein extracts were run on the same gel, membranes were cut for probing with the three antibodies, and images were reassembled after probing. B: transmission electron microscope images of exosome preparations. The large field image was obtained with a 15,000-fold magnification; the insert depicts a single particle selected from the same image. M, molecular weight markers; E, exosome extract.

The uptake of milk exosomes into HUVECs is a carrier-mediated process. First, we established that exosome uptake was linear with time for up to 2 h if 20 μg exosome protein was added to 200 μl media (Fig. 3A), i.e., concentrations below transporter saturation (see below). Temporal patterns were similar when 70 μg exosome protein/200 μl media were used (data not shown). Subsequent transport studies were carried out using an incubation time of 1 h. Exosome uptake followed Michaelis-Menten kinetics (Fig. 3B): V_max = 0.057 ± 0.004 ng exosome protein × 40,000 cells/h and K_m= 18.0 ± 3.8 μg exosome protein/200 μl media. Exosome uptake depended on the incubation temperature (Fig. 4). When 100 μg of unlabeled exosomes was added to the cell cultures (equaling 5 times K_m), the uptake of fluorophore-conjugated exosomes decreased to 16.8 ± 7.2% of controls (P < 0.05, n = 3 biological replicates each measured in triplicate). When cells were treated with 5 μg/ml or 10 μg/ml cytochalasin D, exosome uptake decreased to 63.5 ± 21.3% and 40.8 ± 22.0%, respectively, of controls, consistent with endocytosis (P < 0.05, n = 3). Treatment of cells with puromycin-loaded exosomes provided compelling evidence that exosomes truly entered the intracellular space, as opposed to exosomes adhering to the cell surface. When cells were treated with puromycin-loaded exosomes, viability decreased 75% to 84% of controls (Fig. 5A). Not surprisingly, when 1,837 μM puromycin was added directly to media, viability decreased to 37% of puromycin-free controls (positive control). Cell death was caused by exosome-mediated delivery of puromycin as opposed to release of puromycin from exosomes, based on the following observation. When cells were cultured in media containing exosomes loaded with a plasmid coding for eGFP, they expressed eGFP protein (Fig. 5B).

Fig. 3.

Temporal (A) and saturation (B) kinetics of milk exosome uptake in human umbilical vein endothelial cells (HUVECs). The insert in A illustrates the temporal pattern for 120 min: y = 0.0003x + 0.009, R² = 0.98. In time course studies, 20 μg exosome protein/200 μl of media were used.

Fig. 4.

Temperature dependence of milk exosome uptake in HUVECs (^a,b,cP < 0.05 for bars not sharing a common letter).

Fig. 5.

Uptake of milk exosomes loaded with puromycin (Puro) or enhanced green fluorescent protein (eGFP) plasmid. A: effects of puromycin-loaded milk exosomes on HUVEC survival. Cells were treated with puromycin-free exosomes (left), puromycin-loaded exosomes (middle), and free puromycin (right) (^a,bP < 0.05 for bars not sharing a common letter). B: expression of eGFP after 3 days of HUVEC culture with eGFP plasmid-loaded exosomes.

Imaging studies in mice also suggest that milk exosomes can cross vascular endothelial cells for delivery to tissues. Eighteen hours after intra-orbital injection of DiR-labeled milk exosomes, the majority of exosomes was cleared from circulation and accumulated in a region near the liver (Fig. 6A); no signal was detected in mice injected with free DiR or unlabeled exosomes. In excised tissues, strong signals were detectable in liver and spleen when the exposure time was 20 s (Fig. 6B), whereas traces were detectable in intestine, stomach, and lungs only when the exposure time was increased from 20 s to 30 s under the experimental conditions (Fig. 6C). Note that the signal in liver and spleen was fully saturated at 20 s, i.e., the extension of the exposure time caused an artificial bias towards a stronger signal in intestine, stomach, and lungs.

Fig. 6.

Distribution of milk exosomes in mice. C57BL/6 mice received retro-orbital injections of 1,1′-dioctadecyltetramethyl indotricarbocyanine iodide (DiR)-labeled exosomes, free DiR, or unlabeled exosomes. A: whole mice, 18 h after injection. B: excised tissues, 20 s exposure time. C: same sample as shown in B but exposure time increased to 30 s. Scale bars depict fluorescence intensity in units of percent saturation.

Surface proteins played an important role in facilitating exosome uptake into HUVECs. When exosomal surface proteins were removed by treatment with proteinase K, exosome uptake decreased to ∼50% of controls (Fig. 7A). Likewise, the carbohydrate competitor galactose, but not glucose, caused a significant decrease in exosome uptake (Fig. 7B).

Fig. 7.

Effects of surface proteins on milk exosome uptake in HUVECs. A: treatment of exosomes with proteinase K (100 μg/ml). B: effects of carbohydrate competitors. ^a,bP < 0.05 for bars not sharing a common letter.

DISCUSSION

Evidence is accumulating in support of the theory that dietary miRNAs may cross the intestinal mucosa and have biological activity in humans (7, 27, 35, 55). However, the mechanisms of intestinal transport and subsequent delivery to tissues are unknown. To the best of our knowledge, this is the first paper to propose that the transport of cow's milk exosomes across vascular endothelial cells is mediated by endocytosis and that proteins on the surface of milk exosomes are compatible with proteins on the surface of human vascular endothelial cells. This study further corroborates the notion that dietary miRNAs have biological activity in humans.

This study has far-reaching implications for human nutrition and health. The National Cancer Institute defines bioactive compounds as “a type of chemical found in small amounts in plants and certain foods [ . . . ]. Bioactive compounds have actions in the body that may promote good health. They are being studied in the prevention of [ . . . ] diseases” (36). Milk miRNAs meet that definition, based on our previous studies which suggest that cow's milk microRNAs regulate genes in circulating cells and peripheral human tissues (7). Future studies will need to reveal the extent to which dietary miRNAs contribute to the total miRNA body pool. Our studies suggest that dietary exosomes are cleared primarily by uptake into liver and spleen.

miRNAs have been implicated in virtually all aspects of human health including bone health (32, 43), female and male reproduction (1, 38), arthritis and inflammatory bowel disease (5, 24, 37), metabolic syndrome (20, 31, 45, 53), and cancer (3). Note that miRNAs also may have effects detrimental to human health. For example, the plasma concentrations of miR-210 are significantly higher in patients with pancreatic cancer compared with healthy controls (29); plasma miR-141 and miR-25 are elevated in prostate cancer and esophageal squamous cell carcinoma, respectively (28); plasma miR-21 is elevated in various types of cancer (52); and the urinary excretion of miR-126 and miR-182 is greater in bladder cancer patients compared with healthy controls (22). Additional areas of health relevance include the possible use of milk exosomes as vehicle for the oral delivery of unstable drugs, and the potential role of dietary miRNAs as confounders in biomarker studies that rely on miRNAs in body fluids. Evidence suggests that dietary preferences affect miRNA signatures in human plasma (46). While the distribution of exogenously administered extracellular vesicles (derived from mammalian cells) and their miRNA cargo depends on the cell source of the vesicles and the route of administration, the majority of vesicles appears to accumulate in liver, spleen (macrophages), and the gastrointestinal tract (50). We speculate that vesicles of dietary origin follow that distribution pattern.

Studies with exosomes transfected with puromycin and GFP plasmid provide strong evidence that exosomes entered HUVECs, as opposed to adsorbing to the cell surface. We recognize that it would have been desirable to demonstrate that exosomes and their miRNA cargo are transferred across the vascular endothelium in this study. We sought to assess endothelial transfer using transwell plates (12). However, HUVECs failed to form a tight monolayer, judged by Lucifer Yellow rejection values that were consistently below 90% (not shown). Our observations are consistent with those by others who concluded that the permeability of endothelial cell monolayer cultures is 10–100 greater than in intact endothelia and there is no evidence of restricted diffusion into the subendothelial space (2). On the basis of these observations, it was not feasible to assess the transfer of exosomes across the endothelial layer in cell cultures. Despite this limitation we are confident that endothelial transfer of dietary miRNAs occurs, based on the following observations. 1) Human colon carcinoma Caco-2 cells form tight monolayers. Our studies of intestinal transport of milk exosomes using Caco-2 cells and transwell plates (Lucifer Yellow rejection rate >99.8%) suggest that miRNAs are transferred across intestinal epithelia (51). 2) Previous studies used miRNA reporter genes and quantified the abundance of miRNA in human kidney cell cultures, circulating primary human cells, and mouse livers to demonstrate that dietary miRNAs are delivered to cells (7). 3) When vesicles are administered subcutaneously, intraperitoneally, or intravenously they accumulate in liver and peripheral tissues (50). 4) When fluorophore-labeled milk exosomes are administered orally (gavage) or intravenously, the majority of exosomes accumulates in liver and spleen [(35), this study]. 5) Studies using puromycin-loaded and eGFP plasmid-loaded exosomes suggest that milk exosomes enter cells, as opposed to merely adsorbing to the cell surface.

Some uncertainties remain and will need to be addressed in future studies. First, the identity of the glycoproteins that facilitate the endocytosis of milk exosomes in endothelial cells is unknown and is an area of investigation to be pursued. Second, it is possible that an excess of endogenous exosomes might compete with the endocytosis of dietary exosomes. We cannot formally assess this possibility until the plasma concentration of dietary exosomes has been established; distinct glycoproteome profiles on the surfaces of cells from different tissues might be a confounder in such studies. Third, it is unknown whether dietary exosomes from species remotely related to humans will be recognized by surface proteins in human cells. Fourth, HUVECs from passage number 38-45 were used in this study. While these cells showed no signs of senescence, e.g., decreased rate of proliferation, we did not formally test whether the relatively high number of passages used here altered transport kinetics. The number of passages was dictated by the cells provided by the supplier and the minimal number of passages required to obtain cells for transport studies. Recently, we have launched a database, which is devoted to dietary miRNAs and their analysis by a variety of bioinformatics tools (10). This database will address some of the above uncertainties on an ongoing basis.

We conclude that this study provides an important mechanistic framework for future studies of dietary extracellular vesicles and the roles of dietary miRNAs in human health and disease. In particular, due to the controversy surrounding the bioavailability of plant-borne miRNAs in humans, resources need to be devoted to this promising field of research. A representative example would be the identification and validation of molecular signatures for assessing the dietary intake of vesicles and their cargo.

GRANTS

This material is based on work that is supported by the National Institute of Food and Agriculture, US Department of Agriculture under award number 2015-67017-23181, National Institute of General Medical Sciences Grants 1P20GM104320 and P20GM103427, the Gerber Foundation, the Egg Nutrition Center, the University of Nebraska Agricultural Research Division (Hatch Act), USDA multistate group W3002, and USAID PRESTASI.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

R.J.K., S.M., T.F., S.S., and C.N. performed experiments; R.J.K., S.M., T.F., S.S., C.N., and J.Z. analyzed data; R.J.K., S.M., and J.Z. interpreted results of experiments; R.J.K. drafted manuscript; R.J.K., S.M., T.F., S.S., C.N., and J.Z. approved final version of manuscript; J.Z. conception and design of research; J.Z. prepared figures; J.Z. edited and revised manuscript.