Endothelial Exosome Plays a Functional Role during Rickettsial Infection

ABSTRACT Spotted fever group rickettsioses (SFRs) are devastating human infections. Vascular endothelial cells (ECs) are the primary targets of rickettsial infection. Edema resulting from EC barrier dysfunction occurs in the brain and lungs in most cases of lethal SFR, but the underlying mechanisms remain unclear. The aim of the study was to explore the potential role of Rickettsia-infected, EC-derived exosomes (Exos) during infection. Using size exclusion chromatography (SEC), we purified Exos from conditioned, filtered, bacterium-free media collected from Rickettsia parkeri-infected human umbilical vein ECs (HUVECs) (R-ECExos) and plasma of Rickettsia australis- or R. parkeri-infected mice (R-plsExos). We observed that rickettsial infection increased the release of heterogeneous plsExos, but endothelial exosomal size, morphology, and production were not significantly altered following infection. Compared to normal plsExos and ECExos, both R-plsExos and R-ECExos induced dysfunction of recipient normal brain microvascular ECs (BMECs). The effect of R-plsExos on mouse recipient BMEC barrier function is dose dependent. The effect of R-ECExos on human recipient BMEC barrier function is dependent on the exosomal RNA cargo. Next-generation sequencing analysis and stem-loop quantitative reverse transcription-PCR (RT-qPCR) validation revealed that rickettsial infection triggered the selective enrichment of endothelial exosomal mir-23a and mir-30b, which potentially target the endothelial barrier. To our knowledge, this is the first report on the functional role of extracellular vesicles following infection by obligately intracellular bacteria.

Recent studies provide emerging evidence that microRNAs are selectively sorted into EVs independently of their cellular levels (52,(55)(56)(57)(58)(59)(60)(61)(62). We reported that R. conorii infection induces significant upregulation of specific tRNA-derived RNA fragments in host cells, but no global changes of microRNAs in perfusion-rinsed mouse lung tissues were observed (80). Information regarding the potential role of extracellular RNAs during rickettsial infections is still lacking.
The aim of this study was to explore the potential role of rickettsia-infected, EC-derived Exos following infection. Using size exclusion chromatography (SEC), we purified Exos from conditioned, filtered, bacterium-free media collected from R. parkeri-infected human umbilical vein ECs (HUVECs) (R-ECExos) and plasma of R. australis-or R. parkeri-infected mice (R-plsExos). We observed that compared to noninfectious normal mouse plsExos and normal HUVECderived Exos, both R-plsExos and R-ECExos induced dysfunction of normal brain microvascular ECs (BMECs). The effect of R-plsExos on mouse recipient BMEC barrier function is dose dependent. The effect of R-ECExos on human recipient BMEC barrier function is dependent upon exosomal RNA cargos. Saponin-assisted active exosomal permeabilization pretreatment (81)(82)(83) of R-ECExos with RNase mitigated the effect of R-ECExos on human recipient BMEC barrier function. Next-generation sequencing analysis and stem-loop quantitative reverse transcription-PCR (RT-qPCR) validation revealed that R. parkeri infection triggered the selective enrichment of endothelial exosomal mir-23a and mir-30b, which potentially target the endothelial barrier.
Sizes and morphologies of isolated EVs from mouse plasma and EC culture media, respectively, were initially evaluated using transmission electron microscopy (TEM) (52,88) or atomic force microscopy (AFM) (89). The images captured using TEM and AFM show particles with typical exosomal morphology (arrowheads in Fig. 1a), as published previously (52,88,90). Using nanoparticle tracking analysis (NTA), the size distribution of isolated EVs was also confirmed to be in the range of 50 to 150 nm, which is the expected size of Exos ( Fig. 1b and c). We also verified the purity of isolated Exos using Western immunoblotting to detect traditional exosomal markers as shown in Fig. 1d (64,73,90).
These data demonstrate that purified EVs from rickettsia-infected mouse plasma or culture media used in these studies were free of bacteria or bacterial DNA, were intact and did not aggregate, and fell within the expected size range of Exos.
Exos are differentially induced and detected in mouse plasma and EC culture media in response to rickettsial infection. Exos from circulating blood have been identified as being heterogeneous and derived from multiple cell types, including ECs. Using Western immunoblotting, we detected Exo and EC markers (CD31 and VE-cadherin [91,92]) in mouse plsExos, as well as markers of other cells (CD45) (Fig. 2a), suggesting that the mouse plsExos used in these studies were derived from different types of cells, including ECs.
Exosomal particle counts were measured using NTA and showed that similar numbers of endothelial Exos were produced by mock and R. parkeri infection groups in vitro (Fig. 2b). However, the number of mouse R-plsExos was upregulated on day 4 p.i. (P = 0.02) in vivo (Fig. 2c) in the model of R. australis infection. Exos were also assessed using exosomal total protein content ( Fig. 2d and e) (88). The generation of R-plsExos was significantly upregulated on day 4 p.i. in both in vivo models of R. australis (P = 0.016) (Fig. 2e) and R. parkeri (P = 0.014) (Fig. S2a) infections. No difference in exosomal total protein content was observed between ECExos and R-ECExos in the R. parkeri infection model (Fig. 2d). Furthermore, we also compared the morphology of EC-derived Exos using TEM and AFM, which demonstrated no significant differences between normal ECExos and R-ECExos (Fig. 1a).
Endothelial Exosome during Rickettsial Infection ® Collectively, these data suggest that R. australis and R. parkeri infections increase heterogeneous plsExo release. However, endothelial Exo size, morphology, and production were not significantly altered after R. parkeri infection in vitro.
Recipient cells efficiently take up Exos. ECs are directly exposed to circulating substances and Exos, which are abundant in blood and are taken up by ECs (93,94). To confirm that ECs take up Exos in vivo, we intravenously delivered fluorescent PKH26prelabeled normal plsExos (1 Â 10 11 particles per mouse in 100 ml of phosphate-buffered saline [PBS]) to normal mice as described previously (90). As shown in Fig. 3a, colocalization between PKH26 (red) and CD31 (green, a marker of EC lineage) (arrowheads) were identified in multiple organs in mice, which were extensively perfused with PBS at 6 h postinjection, prior to fixation. These data suggest that plsExos directly interact with ECs in vivo.
Next, we examined HUVEC-derived ECExo uptake using normal recipient cells (i.e., human BMECs) in vitro. PKH26-prelabeled ECExos and nonlabeled controls were added to the cultured human BMECs. As early as 2 h after incubation, the uptake of PKH26prelabeled ECExos by BMECs was visualized using fluorescence microscopy (Fig. 3b).
These data suggest that vascular ECs efficiently take up Exos in our models. Effect of mouse R-plsExos on normal mouse recipient ECs. We next sought to evaluate the potential effect of R-plsExo on normal recipient ECs during rickettsial infection. Using SEC, R-plsExos from a mouse that was intravenously infected with a 2-LD 50 dose of R. australis (86,87) or R. parkeri (6,95) were isolated. Normal mouse recipient BMECs were treated with normal plsExos or R-plsExos at different doses (i.e., 400, 2,000, or 8,000 Exo particles/per cell) for 72 h before measurement of the transendothelial electrical resistance (TEER), an indicator for endothelial paracellular barrier function (96). We found that compared to normal mouse plsExos, mouse R-plsExos derived on day 4 p.i. with R. australis or R. parkeri reduced the TEER in normal mouse recipient ECs in different dose-dependent manners ( Fig. 4a and Fig. S2b).
This evidence suggests that mouse R. australis-or R. parkeri-plsExos induce dysfunction in normal recipient ECs in a dose-dependent manner.
Human R-ECExos induced dysfunction of normal human recipient ECs in an exosomal RNA-dependent manner. Endothelial markers were detected in plsExos (Fig. 2a). Given that ECs are the major target cells during rickettsial infection, HUVEC-derived ECExos were used to explore their effect on normal human recipient BMEC function.  It was first found that compared with normal (mock) ECExos (2,000 Exo particles), R-ECExos (2,000 Exo particles) reduced TEER in normal human recipient BMECs (Fig. 4b). Furthermore, R-ECExos (2,000 Exo particles/cell) weakened the tight junctional protein ZO-1 (arrowheads in Fig. 4c) of normal human recipient BMECs. (d and e) The concentration of exosomal total protein was determined using the micro-bicinchoninic acid (microBCA) protein assay (n = 5 per group). Statistical significance was determined using Student's t test. *, P , 0.05. Endothelial Exosome during Rickettsial Infection ® Exos contain many types of biomolecules, including proteins and nucleic acids, which contribute to disease pathogenesis (68). Active encapsulation techniques have been widely employed in the field of EV research, showing no significant impairment of exosomal constitution, integrity, or functionality (81)(82)(83). To identify the functional exosomal cargos during R. parkeri infection, we employed saponin-assisted active permeabilization (81)(82)(83) to pretreat exosomal cargos with 20 mg/ml of RNase in the presence of 0.1 mg/ml of saponin. Such pretreatment of R-ECExos mitigated the effect on TEER in normal human recipient BMECs compared to RNase in the absence of permeabilization or heat-treated RNase in the presence of saponin (Fig. 4b). Similar pretreatment of R-ECExos with RNase in the presence of saponin also mitigated the effect on the tight junctional protein ZO-1 in normal human recipient BMECs ( Fig. 4c and d).
We also isolated Exos from HUVEC culture media at 72 h after treatment with heatinactivated R. parkeri (MOI, 10) and observed no differences in exosomal total protein contents compared to the mock group (Fig. S5a). Such ECExos (2,000 particles/cell) induced no effect on the paracellular ZO-1 of normal human recipient BMECs ( Fig. S5b and c).
These data suggest that R-ECExos can induce normal human recipient EC barrier dysfunction in an exosomal RNA cargo-dependent manner.
R. parkeri infection upregulates exosomal mir-23a and mir-30b. EV RNA cargo mostly consists of sncRNAs, mainly microRNAs and tRNA-derived fragments (61,77,78). A growing number of reports have established that many effects of EVs are mediated by microRNAs (52,(55)(56)(57)(58)(59)(60)63). We characterized the exosomal microRNA cargo using next-generation sequencing (Fig. 5a). RNAs were isolated from Exos released from HUVECs infected with R. parkeri (at an MOI of 10) for 72 h or mock infected. R. parkeri is a biosafety level 2 (BSL2) pathogen, which lends itself to mechanistic studies. were administered to normal C57BL/6J mice intravenously (n = 3). After 4 h, organs were dissected for frozen sectioning after euthanasia and perfusion via the right ventricle. Representative immunofluorescent staining of ECs from liver, brain, and lung using an antibody against CD31 (an EC marker) is shown. Normal rabbit serum was used as an antibody during immunofluorescent staining (Fig. S3). The nuclei were stained with 49,6-diamidino-2-phenylindole (DAPI). Cells with red fluorescence indicate the uptake of PKH26 labeled Exos. Scale bars, 20 mm. (b) Purified ECExos were labeled with PKH26 (red) and added to the culture medium of human BMECs (2,000 particles per cell) as indicated. Pictures were taken using fluorescence microscopy after 2 h of ECExo incubation. Scale bars, 20mm.
There were no differences in total sncRNAs (,150 nucleotides) per Exo obtained from normal ECExos compared with R-ECExos. Seventy-two hours after R. parkeri infection, mir-23a and mir-30b exhibited the greatest induction of expression in R-ECExos, reaching 7.69-fold and 3.04-fold increases compared to controls, respectively (Fig. 5b).
We next validated the enhanced expression of mir-23a and mir-30b in Exos using stem-loop RT-qPCR, which is a common method for detecting sncRNAs in EVs (52,61,97,98). In Fig. 5c, exosomal mir-23a was upregulated after rickettsial infection with a 3fold increase in expression compared to the mock group (P , 0.01). Similarly, mir-30b The TEER values of normal human recipient BMECs were measured after a 72-h treatment with normal ECExos (mock) or R-ECExos (2,000 Exo particles per cell), which were pretreated with 20 mg/ml of RNase in the presence or absence of 0.1% saponin. **, P , 0.01. (c) Immunofluorescence staining of tight junctional protein ZO-1 (red) in normal human recipient BMECs that were treated with different Exos for 72 h. The yellow arrowheads indicate the decreased signals of paracellular ZO-1. Nuclei of human recipient BMECs were counterstained with DAPI (blue). (d) Relative fluorescent intensities of paracellular ZO-1. Normal rabbit serum was used as negative reagent control during immunofluorescent staining (Fig. S4). Scale bars, 20 mm. Statistical significance was determined using one-way analysis of variance. *, P , 0.05.
Collectively, our data suggest that mir-30b and mir-23a are selectively sorted into R-ECExos following R. parkeri infection.
We have performed an additional experiment to examine mouse mir-23a and mouse mir-30b in mouse plsExos using stem-loop RT-qPCR. As shown in Fig. S6, levels of mouse plsExo mir-23a (P = 0.413) and mir-30b (P = 0.237) increased after R. parkeri infection, but the results were not statistically significant.

DISCUSSION
ECs are the primary mammalian host target cells of SFR infection (2,5,6). The most prominent pathophysiological effect during SFR infections is increased microvascular permeability, followed by vasogenic cerebral edema and noncardiogenic pulmonary edema with potentially fatal outcomes (2,5). Cellular and molecular mechanisms underlying endothelial barrier dysfunction in rickettsiosis remain largely unknown (7)(8)(9). The novel findings in the present study are that R. australis or R. parkeri infection increases the release of heterogeneous plsExos, but endothelial Exo size, morphology, and production are not significantly altered following infection. Mouse R-plsExos induced dysfunction of normal mouse recipient BMECs in a dose-dependent manner, and human R-ECExos induced dysfunction of normal human recipient BMECs in an exosomal RNA cargo-dependent manner. Next-generation sequencing and stem-loop RT-qPCR analyses suggested that mir-23a and mir-30b are selectively sorted into R-ECExos after R. parkeri infection. To our knowledge, this is the first report involving EVs in obligately intracellular bacterial infections.
Exos are in a size range similar to that of viruses (33,94) and contain many types of biomolecules, including proteins and nucleic acids, which contribute to diseases pathogenesis. Exos are being actively investigated in cancers, as biomarkers, and as potential therapeutics (33-38, 68, 70). Exos have been studied in the context of different infections (33-38, 70, 71). During infection, EVs released from the host can be derived from the pathogen or the host. It has been reported that pathogens can utilize different mechanisms to hijack host Exos to maintain their survival and increase their pathogenicity (33). Mycobacterium tuberculosis releases lipoarabinomannan into Exos to decrease the interferon response of the recipient macrophage (33). Exosomal gp63 from Leishmania has been shown to downregulate proinflammatory genes in dendritic cells and macrophages (105). Exos released from Leishmania donovani-infected macrophages block the formation of microRNA-122 in recipient hepatocytes, resulting in a higher parasite burden (106). However, most enveloped virions are the same size as Exos, and major exosomal surface markers CD63 and CD81 are enriched in enveloped viruses (54). Such similarities make the separation of virions and Exos in infected samples particularly challenging (54). Rickettsiae are strictly intracellular bacteria that are about 2.0mm in length (22,107,108). We succeeded in isolating and purifying bacterium-free plsExos and ECExos from rickettsia-infected mouse plasma and cell culture media, respectively, taking advantage of SEC technology, which now serves as the technical foundation for studying the potential role of Exos in the pathogenesis of rickettsiosis.
Differential centrifugation has been employed for Exo isolation for many years, but the technique suffers from aggregation and decreased integrity of Exos (38,53,(64)(65)(66). Recently, single-step SEC was employed successfully for Exo purification, with improved integrity, yield, and no aggregation (38,53,(64)(65)(66). In the present study, using SEC technology, we have successfully isolated Exos from plasma and culture media in BSL2/3 facilities and validated Exo quality using multiple EV-specific assays ( Fig. 1) demonstrating size, purity, and morphological integrity without aggregation. Exo size and morphology were not significantly changed after rickettsial infection (Fig. 1). Generation of plsExos was significantly upregulated after R. australis or R. parkeri infection, while no difference was detected in plasma protein concentrations. However, our in vitro endothelial R. parkeri infection model demonstrated no significant difference in exosomal generation between normal and rickettsia-infected ECs. Circulating Exos have been identified as heterogeneous and derived from multiple different types of cells, including ECs, epithelial cells, leukocytes, erythrocytes, and platelets (90). Our data suggest that ECExo size, morphology, and production were not significantly altered after infection. However, quantitative information regarding the production of SFG rickettsial infection-induced cell-type Exos in vivo remains elusive and requires further research.
Research on most species of lethal human SFG rickettsial infections, including R. rickettsii (3, 4), R. conorii (9), and R. australis (10), is restricted to BSL3 facilities. One key experimental method in our study was to employ novel SEC technology to isolate high-quality Exos from plasma and culture media. Vascular ECs are exceedingly thin (109). Preliminary experiments revealed that we obtained smaller amounts of Exo particles from ECs cultured in the same-size culture vessel at the same confluence for the same duration compared with other cell types. It is practical and feasible to isolate endothelial Exos from media using SEC on a relatively larger scale following infection with R. parkeri (5-7), which requires a lower biocontainment level than R. australis. R. parkeri, the cause of R. parkeri rickettsiosis (8), is proposed as an experimental pathogen for studying the pathogenesis of SFG rickettsiosis and can be experimentally handled at BSL2 (6,95). In a murine model of intravenous inoculation of R. parkeri (Atlantic Rainforest strain) using C3H/HeN mice, signs of illness begin on day 3 and fatalities occur on day 6 p.i. (95), during which time microvascular damage is detected in multiple organs (95); this is similar to the pathology observed in our murine model of R. australis infection (85,87). However, differences among various pathogen species must be considered in future mechanistic studies. Furthermore, current technological hurdles hamper us from isolating ECExos from plasma on a large scale for downstream experiments. To collect sufficient ECExos using SEC technology for mechanistic studies, culturing cells in media is the sole practical source for ECExo isolation. To investigate the functional role of ECExos in R. parkeri infection, we used HUVECs as the source of Exo-producing cells and BMECs as the recipient cells. Research to develop feasible celltype Exo isolation using the same pathogens in the same host models is warranted for future mechanistic studies.
Exos can induce functional responses using multiple mechanisms, including releasing bioactive components after internalization (67,68). In infectious disease biology, EVs from infected donor cells are associated with virulence of the pathogens in recipient cells (33)(34)(35)(36)(37)(38)(69)(70)(71). In the present study, both R-plsExos and R-ECExos weakened the barrier function of the normal ECs. Concomitantly, human R-ECExos induced disruption of the tight junctional protein ZO-1 in recipient human BMECs in an exosomal RNA-dependent manner. However, the underlying mechanism remains unclear.
The discovery of extracellular sncRNAs in the blood, despite the abundant presence of RNases, led to the proposal of a scenario in which sncRNAs are encapsulated in EVs (55,(72)(73)(74) or in the form of circulating ribonucleoproteins (75,76). EV-enclosed mRNAs are mostly fragmented, and extracellular RNAs are enriched in sncRNAs (77). Despite a previous report that the average copy number of miRNAs in each EV is low (110), accumulating evidence suggests a critical function of EV-containing miRNAs. EV RNA cargo mostly consists of sncRNAs (77). A growing number of reports have established that many, if not all, of the effects of EVs are mediated by microRNAs (52,(55)(56)(57)(58)(59)(60)63), which remain functional to regulate cellular behaviors of the recipient cell (79). Exosomal microRNAs are of particular interest due to their participation in posttranslational regulation of gene expression. A single microRNA can regulate many target genes to affect biological function (111). Recent studies provide evidence that microRNAs are selectively sorted into EVs, independent of their cellular levels (52,(55)(56)(57)(58)(59)(60)(61)(62). We observed no significant differences in total sncRNAs (,150 nucleotides) per Exo between normal ECExo and R-ECExo. Seventy-two hours after R. parkeri infection, expression levels of mir-23a and mir-30b were remarkably upregulated in R-ECExos, but no change was observed in the mock-infected cells. These data suggest that mir-23a and mir-30b are selectively sorted into R-ECExos during R. parkeri infection. The underlying mechanism is yet to be elucidated.
Analysis of the interactions among enriched exosomal microRNAs and potential mRNA targets will provide putative mRNA candidates for future studies. Given that the molecular and functional effects of mir-23a (100-103) and mir-30b (103,104) have been documented to target endothelial barrier functions, further research into the selective sorting mechanism(s) and functional roles of exosomal mir-23a and mir-30b may provide new insights into the pathogeneses of SFR. Furthermore, additional research may validate specific exosomal microRNAs as impactful druggable targets for the prevention and treatment of fatal human diseases caused by Rickettsia and other pathogens.

MATERIALS AND METHODS
Mouse models of R. australis and R. parkeri infections. All animal experiments were performed according to protocols approved by the Institutional Animal Care and Use Committee of the University of Texas Medical Branch (UTMB). C57BL/6J mice were obtained from the Jackson Laboratory (Bar Harbor, ME). Mice used for R. australis infections were 8-to 12-week-old males. C57BL/6J mice are highly susceptible to R. australis, becoming ill on day 3 p.i. and succumbing to infection on day 6. Therefore, this organism was chosen as the SFG rickettsial agent for the in vivo studies (10). Male C57BL/6 mice infected with R. australis are an established animal model of human SFG rickettsiosis because the pathology involves disseminated endothelial infection and pathological lesions, including vasculitis in multiple organs, similar to what is observed in human SFG rickettsiosis (10,87). However, use of R. australis requires BSL3 containment. An ordinarily lethal dose of 2 LD 50 of R. australis (the LD 50 is 1 Â 10 6 PFU) was injected through the tail vein (87), and whole-blood samples were collected on day 4 p.i. for plasma isolation.
Liu et al.

®
R. parkeri is proposed as an experimental pathogen for studying the pathogenesis of SFG rickettsiosis and is similar to R. australis (10,84,86,87), but it can be experimentally handled using BSL2 containment (6,95). For isolation of endothelial Exos from infected media on a large scale, it is more practical and feasible to use R. parkeri. In a murine model of intravenous inoculation of R. parkeri (Atlantic Rainforest strain) using C3H/HeN mice, signs of illness began on day 3 and fatalities occurred on day 6 p.i. (95), during which time microvascular damage was detected in multiple organs (95), which is similar to the pathology observed in the murine model of R. australis (85,87). C3H/HeJ mice (8-to 12-week-old males) were obtained from the Jackson Laboratory. An ordinarily lethal dose of 2 LD 50 of R. parkeri (Atlantic Rainforest strain) was injected through the tail vein (87), and blood samples were collected on day 4 p.i. for plasma isolation.
NTA. To determine the size and concentration of EVs, nanoparticle tracking analysis (NTA) was performed at the Nanomedicines Characterization Core Facility (University of North Carolina, Chapel Hill, NC). Briefly, isolated Exo samples were diluted to a concentration of 5 Â 10 9 to 1 Â 10 11 particles/ml in filtered PBS. The samples were then processed on a NanoSight NS500 (NanoSight, Malvern Instruments, Westborough, MA) to capture particles moving by way of Brownian motion (camera type, scientific complementary metal oxide semiconductor [sCMOS]; camera level, 16; detection threshold, 5). The hydrodynamic diameters were calculated using the Stokes-Einstein equation. The 100-nm standard particles and the diluent PBS alone were used for reference.
MicroRNA quantification in HUVEC Exos. As previously reported (52), RNAs were extracted from purified HUVEC Exos. Small RNAs (6 to 150 nucleotides) and microRNA fractions (10 to 40 nucleotides) were quantified using high-resolution small RNA analysis (2100 Bioanalyzer system; Agilent, Santa Clara, CA) at the Biopolymer Facility (Harvard Medical School, Cambridge, MA). To determine the concentration of small RNAs and microRNAs per Exo, the quantified sncRNA/microRNA value was normalized to the Exo count, which was evaluated using NTA.
Bioinformatic analysis of sequencing data. Sequencing was done using an Illumina NextSeq as single-end 75-bp reads generating between 4.8 and 40.9 million reads per sample. Quality control of the samples was performed using Qiagen CLC Genomics Workbench 20.0. Raw sequencing reads were trimmed to remove Qiagen 39-AACTGTAGGCACCATCAAT and 59-GTTCAGAGTTCTACAGTCCGACGATC adapters, as well as filtered based on initial quality assessment. Reads dominated by low-quality base calls, and longer than 55 nucleotides, were excluded from the downstream analyses. Atomic force microscopy (AFM). The purified and concentrated EV samples were diluted at 1:10, 1:100, and 1:1,000 with molecular-grade water. Glass coverslips were cleaned three times with ethanol and acetone and then three times with molecular-grade water. Each coverslip was correctly labeled, placed in the hood to dry under laminar flow for 1 h, and subjected to coating with the diluted EV samples on the designated area for 30 min. EV samples were washed away gently with molecular-grade water, and the coverslip was dried for 1 h.
The coverslip coated with EV samples was examined using an AFM (CoreAFM; Nanosurf AG, Liestal, Switzerland) using contact mode in the air. A PPP-FMR-50 probe (0.5 to 9.5 N/m, 225 mm in length, and 28 mm in width; Nanosensors, Neuchatel, Switzerland) was used. The parameters of the cantilever were calibrated using the default script from the CoreAFM program using the method of Sader et al. (115). The cantilever was approached to the sample under the set point of 20 nN (115), and topography scanning was done using the following parameters: 256 points per line and 1.5 s per line in a 5-mm by 5-mm image.
Rickettsiae, cell culture, and rickettsial infections. R. australis (Cutlack strain) (87) and R. parkeri (Atlantic Rainforest strain) (95) were prepared as described previously. Uninfected Vero cells were processed as mock-infected control material using the same procedure. All biosafety level (BSL)2/3 and animal BSL2/3 experiments were performed in CDC-certified facilities in the Galveston National Laboratory at UTMB, Galveston, TX, using established procedures.
A standard protocol to isolate brain microvascular endothelial cells (BMECs) from C57BL/6 and C3H/ HeN mice (116) was used. Human umbilical vein endothelial cells (HUVECs; Cell Applications, Inc.) or BMECs were cultivated in 5% CO 2 at 37°C on type I rat tail collagen-coated round glass coverslips (12mm diameter; Ted Pella, Redding, CA) until 90% confluence was observed. HUVECs were infected with R. parkeri at an MOI of 10. Uninfected ECs were used as mock controls and were subjected to the same procedure. All experiments were performed in triplicate. Normal mouse or rabbit IgGs were used as negative controls.
ECExo and plsExo isolation, concentration, and permeabilization. (i) ECExo isolation and concentration. Donor HUVECs in T75 flasks were infected using R. parkeri at an MOI of 10 or were mock infected for 72 h, and 11 ml of medium was collected. The media were passed through 0.2-mm syringe filters twice. Following the manufacturer's instructions, 10 ml of filtered medium was subjected to the qEV10 column (Izon, New Zealand) for SEC isolation. Four fractions (the number 7 to 10 fractions) were collected as the Exo-enriched fractions, which were concentrated using 100,000-molecular-weight-cutoff Endothelial Exosome during Rickettsial Infection ® (MWCO) polyethersulfone (PES) Vivaspin centrifugal filters (Thermo Fisher Scientific). Exo samples (in 200 ml of PBS) were stored at 280°C prior to use in downstream assays.
(ii) plsExo isolation and concentration. For plasma isolation, blood samples were collected in anticoagulation tubes 4 days following infection with R. australis, R. parkeri, or saline (mock infected). The plasma sample (200 ml) was passed through 0.2-mm syringe filters twice. Following the manufacturer's instructions, filtered plasma was placed onto the qEVoriginal column (Izon, New Zealand) for SEC isolation. The number 7 to 9 fractions were collected as the Exo-enriched fractions, which were concentrated using 100,000-MWCO PES Vivaspin centrifugal filters (Thermo Fisher Scientific). Exo samples (in 200 ml of PBS) were stored at 280°C prior to use in downstream assays. Following the manufacturer's instructions, three fractions (the number 13 to 15 fractions) from the qEVoriginal column isolation were collected as the Exo-free, high-protein plasma fractions, which were concentrated using centrifugal filters and used in Western immunoblotting as a control.
Distribution of Exos in vivo and in vitro. Using a published approach, recipient cell uptake of Exos was assessed in vivo and in vitro (90). Briefly, following incubation using materials from the PKH26 red fluorescent cell linker kit (Millipore Sigma, St. Louis, MO), the PKH26 red-prelabeled Exos were washed three times with PBS prior to ultracentrifugation at 100,000 Â g for 20 min at 4°C using a Beckman L7-80 and SW41 rotor (Beckman Coulter, Indianapolis, IN) to remove unbound stain. PBS without Exos was processed using the same steps as the mock PKH26-labeled tracer. A single injection of PKH26-labeled exosomes (about 1 Â 10 11 particles in 100 ml of PBS) via the tail vein of a normal mouse was done to observe the distribution of Exos in the lungs, liver, and brain 6 h after injection. Immunofluorescence staining was done using frozen sections with rabbit antibodies to CD31. For the in vitro assessment, living BMECs were exposed to PKH26-prelabeled ECExos (2,000 particles/cell) in the culture media of normal human BMECs. After 2 h, cells were fixed and immediately subjected to fluorescence microscopy. All solutions of PKH26-labeled ECExos were filtered with a 0.2-mm filter. Fluorescent images were analyzed using Olympus BX51 epifluorescence and a Nikon A1R MP ECLIPSE Ti confocal microscope with NIS-Elements imaging software version 4.50.00 (Nikon, Tokyo, Japan).
Statistics. Statistical significance was determined using Student's t test or one-way analysis of variance. Results were regarded as significant if two-tailed P values were ,0.05. All data are expressed as means 6 standard errors of the means.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.03 MB.

ACKNOWLEDGMENTS
We gratefully acknowledge Pragnesh Patel for his contributions establishing the capacity of the exosomal size distribution analysis. We gratefully acknowledge Kimberly Schuenke for her critical review and editing of the manuscript. We thank Hugo Samano for input during the planning phases of the experiments. The sponsors had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
We declare that we have no conflicts of interest.