Multi-resolution imaging using bioluminescence resonance energy transfer identifies distinct biodistribution profiles of extracellular vesicles and exomeres with redirected tropism

Extracellular particles (EP) including extracellular vesicles (EVs) and exomeres have been shown to play significant roles in diseases and therapeutic applications. However, their spatiotemporal dynamics in vivo have remained largely unresolved in detail due to the lack of a suitable method. We therefore created a bioluminescence resonance energy transfer (BRET)-based reporter, PalmGRET, to enable pan-EP labelling ranging from exomeres (< 50 nm) to small (< 200 nm) and medium and large (> 200 nm) EVs. PalmGRET emits robust, sustained signals and allows the visualization, tracking and quantification of the EPs from whole-animal to nanoscopic resolutions under different imaging modalities, including bioluminescence, BRET and fluorescence. Using PalmGRET, we show that EPs released by lung metastatic hepatocellular carcinoma (HCC) exhibit lung tropism with varying distributions to other major organs in immunocompetent mice. We further demonstrate that gene knockdown of lung-tropic membrane proteins, solute carrier organic anion transporter family member 2A1 (Slco2a1), alanine aminopeptidase (Cd13) and chloride intracellular channel (Clic1) decreases HCC-EP distribution to the lungs and yields distinct biodistribution profiles. We anticipate that EP-specific imaging, quantitative assays and detailed in vivo characterization to be a starting point for more accurate and comprehensive in vivo models of EP biology and therapeutic design.

The subtypes are lipid-bilayer encapsulated and released through two main pathways: i) fusion of the multivesicular bodies with the plasma membrane to release exosomes; or ii) outward budding and fission of the plasma membrane to release microvesicles. As EV subtype-specific isolation methods and terminology remain to be standardized [4] , the generic term "EV" is used here unless otherwise specified. Using asymmetric flow field-flow fractionation (AF4), cell-released EPs can be further resolved into Exo-S (60-80 nm), Exo-L (90-120 nm) small EVs (sEVs) and exomeres (< 50 nm), with the latter considered as non-EVs as they are non-membranous and exhibit a limited set of membrane-bound proteins [4,5] .
To facilitate intercellular communication, EVs and exomeres ferry bioactive cargoes between cells. These include DNA, RNA (e.g. mRNA, miRNA and lncRNA) and proteins (e.g. ligands and receptors), which have been associated with development, immune response, neurodegenerative diseases, developmental disorders and cancer progression [6][7][8][9][10][11][12] . Tumor exosomes have been shown to be organotropic in facilitating subsequent metastasis to the tropic organs [13] . Specific integrins have been found to redirect tumor exosomes to the lungs (ITGβ 4 ) and liver (ITGβ 5 ). Concurrently, EVs are being actively explored as a diagnostic medium for liquid biopsy, in addition to serving as an endogenous delivery vehicle for therapeutic applications. Specifically, EVs have been designed with cell-type-and cancer-targeting moieties to enable the targeted delivery of therapeutic EVs [14] .
However, the spatiotemporal dynamics of EVs in vivo remains elusive. For example, while organotropic and cell-targeting EVs may improve delivery to a particular site, their distribution to and relationship with other non-targeted tissues at organ systems level is largely unresolved.
Moreover, accurate quantification of EVs has been confounded by a lack of suitable dyes and technical limitations [15] . The biological understanding and therapeutic development of EVs would be greatly aided by the detailed characterization of EVs' spatiotemporal properties in vivo.
Conventional EV-labelling methods mainly rely on fluorescent (FL) dyes. Organic dyes such as PKH dyes, 1,1′-dioctadecyl-3,3,3′3′-tetramethylindocarbocyanine perchlorate (DiI) and its derivatives or carboxyfluorescein succinimidyl ester (CFSE) have been widely used in the study of EVs due to their stable and high FL signals [16,17] . However, lipophilic FL dyes were later identified to spontaneously form nanometer sized micelles (i.e. false-positives) and carry an extended in vivo half-life (e.g. PKH2: 5-8 days, PKH26: > 100 days), resulting in inaccurate spatiotemporal detection of EVs [18,19] . Bioluminescent (BLI) reporters have been used to label EVs for the in vivo imaging and pharmacokinetic analysis of administered EVs [20,21] . However, the resolution is relatively poor when compared with that of FL dyes, and so conventional methods typically use BLI and FL reporters for in vivo and in vitro imaging, respectively. Bioluminescence resonance energy transfer (BRET) was developed to overcome some of these limitations [22] . Similar to the principle of fluorescence resonance energy transfer, BRET reporters consist of a pair of BLI and FL proteins conjugated in close proximity (30-70 Å) by a linker for intramolecular energy transfer. When treated with the substrate, BRET will emit both BLI and FL signals for detection. However, the use of BRET has been limited by BLI protein size, luminescence stability and intensity [23] . Nanoluc (Nluc) has recently overcome these obstacles to BRET; Nluc is a small (19 kDa), ATP-independent luciferase that provides one of the brightest glow-type bioluminescence for sustained excitation of the paired FL protein [23][24][25][26] .
Previous studies have used EV markers fused to FL proteins (e.g. CD63-GFP) to examine EV distributions, but have mostly been limited to in vitro observations [27] . Meanwhile, multiphoton microscopy enables high-resolution imaging of cells in vivo by exciting fluorophores with multiple photons at longer wavelengths for deeper tissue penetration (up to ~1 mm) [28] . However, it generally requires the imaging window to be fixed at one location to record cellular events longitudinally. We applied multiphoton microscopy to achieve in vivo visualization of EV release using pan-EV labelling FL EV reporters [29] . However, as EVs are circulated, they cannot be systemically tracked by multiphoton microscopy. Multiharmonic microscopy has also been applied successfully to imaging EVs in vivo, but it relies on the presence of specific metabolite signatures, and so is inapplicable to unspecified EVs [30,31] . On the other hand, radioactive tracers such as 99m Tc-HMPAO and 111 In-oxine have been applied to label EVs for in vivo imaging and EV biodistribution analysis using positron emission tomography-computed tomography and single-photon emission computed tomography, respectively [32,33] . However, the use of radioisotopes is limited to well-trained operators and certified facilities with available instruments, making these techniques costly and unsuitable for a general laboratory.
To overcome these limitations of EV visualization, we describe here a BRET-based EP reporter, PalmGRET, that enables multimodal, multi-resolution imaging and quantification of EPs from organismal to super-resolutions. PalmGRET was created by fusing a palmitoylation signal peptide sequence from growth-associated protein 43 (GAP43) [34] to the N-terminus of a GFP-Nluc BRET reporter (GpNluc) [23] . We demonstrate that PalmGRET can label multiple EP populations, including exomeres (< 50 nm) and small (< 200 nm; sEVs), medium and large (> 200 nm; mEVs and lEVs) EVs, making it a versatile reporter to visualize and monitor EPs released by cells. Moreover, PalmGRET specifically labels the EV inner membrane with a long-term and robust signal, thereby enabling live cell BLI and BRET microscopy, as well as super-resolution radial fluctuations (SRRF) nanoscopy [35] to observe EP dynamics in vitro. With its emitted BLI and BRET-FL signals, PalmGRET further allows in vivo imaging of administered EPs over an extended period of time (~20 min), as well as highly sensitive EP biodistribution analysis to detect minute EP changes in organs. Using PalmGRET, we identified lung-tropic proteins of EPs derived from lung metastatic hepatocellular carcinoma (HCC) in immunocompetent C3H mice, Slco2a1 (solute carrier organic anion transporter family member 2A1, Cd13 (Anpep, alanine aminopeptidase), and Clic1.
Knockdown of the lung-tropic proteins significantly reduced HCC EP lung tropism. Additionally, Slco2a1 and Cd13 protein knockdowns of HCAC EPs resulted in redirection to the kidney and heart.
Our results suggest that the in vivo function and therapeutic design of EPs should be considered at organ system level to account for their dynamic biodistribution.

Results and Discussion
PalmGRET labels EVs at the inner membrane with FL and BLI reporter activities. To develop a multimodal and multi-resolution EP reporter, PalmGRET was created by genetically fusing the palmitoylation signal peptide sequence [36] of GAP43 [34] to the N-terminus of a BRET pair, GpNluc [23] (Figure 1 a-c). Upon stable expression in human 293T cells, PalmGRET uniformly labelled the plasma membrane to reveal an uneven membrane surface with bud-like structures at the apical plane and protrusions of the cells, which may be precursors to EP release (Figure 1d). To examine whether PalmGRET can be applied to visualize EP release from multiple cellular populations, stable 293T-PalmGRET and 293T-PalmtdTomato cells [29] were co-cultured; both cell types were found to demonstrate EP release, thereby suggesting bidirectional EP-mediated intercellular communication (Figure 1e and Movie S1).
To characterize the sEV labelling properties of PalmGRET, sEVs isolated from stable 293T-PalmGRET and 293T-GRET cells (control) were examined by nanoparticle tracking analysis. The resulting sEVs, sEV-PalmGRET (147.7 ± 2.2 nm), sEV-GRET (143.9 ± 3.8 nm; control) and sEVwildtype (EV-WT; 150.0 ± 3.5 nm; control), were all similar in size (Figure 2a, b). Transmission electron microscopy with immunogold labelling further demonstrated that PalmGRET mostly localizes to sEV membranes (Figure 2c, d). To mitigate possible steric hindrance to ligand-receptor interaction on the outer EV membrane [13,37] , we designed PalmGRET to label the inner EV membrane using the palmitoylation signal. To confirm the inner-membrane specificity of PalmGRET, we subjected the labelled sEVs to dot blot analysis in the presence or absence of detergent (Figure 2e, f). Whereas antibodies can only recognize proteins on the outer sEV membrane in the absence of the detergent, the addition of TWEEN-20 (0.1% v/v) enables antibody entry and immunolabeling of proteins at the inner membrane by disrupting the EV membranes.
Therefore, as sEV-GRET showed a dim signal with the detergent and no signal in its absence, GRET appeared to be non-specifically packaged into the EV lumen (Figure 2e). By contrast, a strong sEV-PalmGRET signal was detected only in the presence of the detergent, thereby indicating inner membrane-specific labelling of sEVs (Figure 2f).
To verify the BLI function and EV-specificity of PalmGRET, isolated sEVs were subjected to sucrose density gradient fractionation followed by BLI assay. In both the sEV fractions (Figure 2g) and pelleted sEVs (Figure 2h), sEV-PalmGRET showed ~20 times the BLI activity of the sEV-GRET control. Western blot analysis of the pelleted sEVs further demonstrated that PalmGRET coincided with sEV markers (CD63 and Alix) [38] at the sEV-containing fractions (1.117-1.163 g·cm −3 ; Figure 2i and Figure S1). Overall, PalmGRET efficiently and specifically labelled the inner sEV membrane while functioning as a BLI reporter.
PalmGRET labels multiple bionanoparticle populations. An EV subtype-specific protein remains to be identified, but multiple tetraspanins (CD63, CD81 and CD9) have been recommended to identify subpopulations of sEVs [39] . While studies have commonly focused on a specific EV subpopulation to investigate its biological function, we envision multiple EP populations working simultaneously in mediating EP-based intercellular communication, especially in vivo. Therefore, we designed PalmGRET as a pan-EP labelling reporter to label and examine multiple EP populations. Using AF4, Zhang et al. subcategorized sEVs as Exo-L (90-120 nm), Exo-S (60-80 nm), and identified non-membranous exomeres (< 50 nm), each with different molecular profiles [40] .
To further examine the pan-labelling coverage of EP by PalmGRET, we performed AF4 on sEV-PalmBRET and characterized fractions corresponding to Exo-L (P4), Exo-S (P3) and exomeres (P2) (Figure 3a). To verify PalmGRET-labelling of the sEV and exomere fractions, the fractions (P2, P3 and P4) were subjected to BLI assay, which revealed ~10-fold stronger BLI and BRET-FL signals in PalmGRET-labelled Exo-L, Exo-S and exomeres when compared to that of GRET (Figure 3b, c).
While Exo-S and Exo-L of EV-PalmGRET exhibited no significant differences, exomeres showed significantly lower signals in both BLI and BRET-FL activities, which positively correlates to the particle abundance of the EPs as found by the quasi-elastic light scattering detector (Figure 3a).
Interestingly, the PalmGRET-labelled exomere fraction showed a slightly greater hydrodynamic radius (~40 nm) than the GRET control (~35 nm), suggesting that PalmGRET slightly increases the size of exomeres while remaining under 50 nm in diameter. We further demonstrated that PalmGRET effectively labels EVs isolated from 0.22, 0.8 and 1.2 μ m-filtered media, as well as 10k and 100k pellets (Figure S2, S3). We therefore confirmed that PalmGRET labels varying sized EPs ranging from exomeres, and sEVs to lEVs.

PalmGRET-labelled EPs exhibit robust and long-term BLI and BRET-FL signals.
To characterize PalmGRET's BLI properties, sEV-PalmGRET was supplemented with furimazine (Fz) substrate and measured for BLI and BRET-FL activities over time. sEV-PalmGRET demonstrated a significantly higher signal in both the BLI and BRET-FL channels than either sEV-GRET or sEV-WT control (Figure 3d, e). Interestingly, 293T-GRET cells emitted higher BLI and BRET-FL signals than 293T-PalmGRET cells (Figure S4a, b). sEV-PalmGRET showed a similar BRET ratio to sEV-GRET, indicating that the labelling of GRET to sEVs using the palmitoylation moiety does not affect BRET efficiency (Figure 3f). The BRET ratio of 293T-PalmGRET and 293T-GRET cells were also similar, further supporting that PalmGRET does not affect reporter function ( Figure S4c).  (Figure S4d, e). The observed difference in PalmGRET activities between the labelled EVs and cells may be attributed to a variation in membrane-to-cytosol ratio: an EV particle has a higher membrane-to-cytosol ratio, which is higher for an EV particle than for a cell [41] . In this way, PalmGRET locates to the plasma membrane and hence is more enriched per EV, whereas GRET predominately distributes to the cytosol and therefore is more abundant per cell. Consequently, sEV-PalmGRET showed stronger BLI and BRET-FL signals (and lower half-lives) as more Nluc enzyme is available to oxidize Fz when compared with sEV-GRET, thereby yielding greater AUCs.
Similarly, 293T-GRET cells generate higher AUCs with more abundant Nluc available per cell. Visualization, tracking and quantification of EP-PalmGRET from whole animal to nanoscopic resolutions. While our previous GlucB reporter allows in vivo imaging of EVs, the signal half-life is short (< 5 min) due to the flash-type emission of Gaussia luciferase [21,42] . In addition to facilitating multi-resolution imaging, PalmGRET was designed to enable EP imaging for an extended period of time by employing Nluc, which has been reported to carry a BLI signal half-life greater than 2 h [43] . To examine in vivo imaging using PalmGRET, 100 μ g PBS-washed EP-293T- PalmGRET was IV-injected to C3H mice followed by Fz administration to image the EVs in vivo We further visualized EP distributions in the lungs, spleen, liver and kidneys using nanoscopy.
The EP-293T-PalmGRET-injected group showed a non-uniform and sporadic distribution of EPs in the lungs (Figure 5d). A similar distribution trend was also observed in the spleen, liver and kidneys, suggesting EP uptake by organs is tissue/cell-specific ( Figure S6). This observation corroborates the findings of Hoshino et al. that EVs derived from malignant breast cancer cells are taken up by specific cell types including SPC + epithelial cells and S100A4 + fibroblasts [17] .
Quantification of EP signals from seventy SRRF images per organ with ImageJ [44] showed that the lungs contained the highest EP signals followed by the spleen, liver and kidneys (Figure 5e and Figures S6 and S7), hence corroborating the EV distribution pattern observed by in vivo and ex vivo imaging (Figure 5b, c). The same imaging and analysis parameters were applied to all tissue samples. Taken together, PalmGRET accurately revealed EP biodistributions with quantification at multiple levels of organization from whole animals and organs to tissues and cells.

Redirected organotropisms of EPs yield distinct biodistribution profiles. Tumor EVs have been
reported to be tropic to pre-metastatic niches to promote subsequent tumor metastasis [13] . Other works have decorated EVs with cell type-and cancer-directing peptides such as rabies viral glycoprotein RVG and GE11, respectively, for targeted delivery of therapeutic EVs [45] . However, the spatiotemporal dynamics of EV biodistribution following redirection of EV tropism remains largely unknown. To explore this phenomenon, we applied the EP-PalmGRET system to elucidate the spatiotemporal properties of EP in the context of redirected EP tropism. We established the PalmGRET system in a mouse hepatocellular carcinoma cell line, HCA1, which exhibits spontaneous lung metastasis in immunocompetent C3H mice [46] , thereby generating stable HCA1-PalmGRET cells. The lung metastatic potential of HCA1-PalmGRET cells was confirmed by orthotopic implantation and detection of tumor formation in the lungs (Figure 6a, b).
As tumor EVs have been shown to target pre-metastatic niches [13] , we hypothesized that EPs released by HCA1-PalmGRET will be lung tropic to promote subsequent metastasis. To examine this possibility, we educated immunocompetent C3H mice with 30 μ g EP-HCA1-PalmGRET (or PBS as a control; Figure S8), and conducted BLI analysis to monitor EP biodistribution in the major organs. Remarkably, EP-HCA1-PalmGRET showed a greater distribution to the lungs and liver (50.7% ± 5.40% and 18.3% ± 3.24%, respectively) at 0.5 h post-injection (Figure 6c and signals, demonstrating the sEV-labelling specificity of PalmGRET ( Figure S10b). As C3H mice are immunocompetent, we also performed multiplex cytokine tests on mice serum, and confirmed that the EP injections did not induce an acute immune response, which may have led to a biased uptake in the lungs, livers or spleen (Table S3) [47] . In addition, since hepatocytes release lipoproteins including palmitoylated proteins [48,49] , PalmGRET may also label HCA1-PalmGRET-released lipoproteins and contribute to the observed biodistribution profiles. This warrants further investigation.
Given that EP-HCA1-PalmGRET exhibited lung tropism, we next redirected the EPs from the lungs, and investigated the subsequent distribution dynamics to non-targeted organs. We identified several lung tropic protein candidates by proteomic analysis of EP-HCA1-PalmGRET. Among the 521 detected proteins (Table S4), we screened for lung tropism-related membrane proteins using Uniprot [50] followed by association to HCC progression, lung metastasis and lung cancer progression, and selected four candidates: Slco2a1 [51] , Cd13 [52] , Clic1 [53] , and Nherf1 [54] . Following knockdown of the proteins in HCA1-PalmGRET cells (Figures S11, S12) and hence reduced expression of the derived EPs (Figure 6e), we administered the EPs at 10 μ g per 24 h for 72 h to investigate the tropism and biodistribution in C3H immunocompetent mice (Figure 6f). At 0.5 h post-injection, the shSlco2a1, shCd13 and shClic1 groups showed a significantly reduced percentages of EPs in the lungs compared with the Scramble control (Figure 6g, Figure S13).
Interestingly, shCd13 and shClic1 also exhibited increased EPs to the liver, whereas shClic1 and shSlco2a1 demonstrated elevated distribution to the spleen and kidneys at 0.5 h. At 72.5 h, all three groups maintained decreased EP distribution to the lungs. While shCd13 showed increased EP distribution to the spleen, kidneys and heart, shSlco2a1 displayed elevated distribution to the kidneys, heart and brain at 72.5 h. The shNherf1 group yielded neither a change in lung tropism nor an altered biodistribution profile when compared with the Scramble control. A high mRNA level and moderate-high expression of SLCO2A1 have previously been identified in primary and metastatic liver cancers [51] , and SLCO2A1 upregulation has been reported to promote lung cancer invasion using a PI3K/AKT/mTOR pathway [55] . Downregulation of CLIC1 has been shown to decrease the invasion and migration of HCC cells [56] , whereas CLIC1 overexpression in HCC correlates to larger tumor size, distal metastasis, aggressive pathologic tumor phenotype, nodes and poor survival rate [57] . Increased CD13 expression has been reported to relate to TGF-β-induced EMT-like phenomenon in liver cancer [58] . Therefore, the loss of lung tropism of EP-HCA1-PalmGRET by Slco2a1, Clic1 and Cd13 knockdowns suggests these proteins direct EP-HCA1-PalmGRET to the lungs for the subsequent lung metastasis of the HCA1-PalmGRET tumor, and warrants further investigation. Additional study will also be required to elucidate the specific tissues and cells facilitating EV uptake by each organ. In summary, we demonstrate that while redirected EP tropism reduces delivery to the lungs, its distribution to other major organs is also dynamically altered, possibly yielding a different function at organ systems level.

Conclusions
Accurate visualization and tracking of EV distributions are critical to understanding the role of EV-mediated cell-to-cell communication in diseases and therapy. Here we established a multimodal and multi-resolution PalmBRET method to enable pan-EP labelling and imaging and therefore quantification in live cells, whole animals, and preserved tissues. Addgene plasmid #70185) [23] . For comparison between sEVs and m/lEVs, CM from 293T-GRET or 293T-PalmGRET were sequentially centrifuged at 300 × g and 2,000 × g at 4 °C for 10 minutes to remove cells and cell debris, respectively. The supernatants were centrifuged at 10,000 × g, 4°C for 30 min to collect 10k pellets (e.g. microvesicles; m/lEVs), and the remaining supernatants were further centrifuged at 100,000 × g, 4°C for 90 min to harvest 100k pellets (e.g. exosomes; sEVs) [39,60] .
Nanoparticle tracking analysis. NanoSight NS300 (Malvern Panalytical, United Kingdom) was used to determine the size and particle concentration of isolated EVs. NTA Version: NTA 3.  Imaging System (Topbio). Antibody dilutions and hosts are provided in Table S2.

HCA1-PalmGRET orthotopic implantation in C3H mice. C3H mice (Biolasco) were
anesthetized, and 1 x 10 6 cells of HCA1-PalmGRET were mixed with 10 μ L Matrigel (Corning, New York, USA) then orthotopically injected to the liver of 5-6 weeks old C3H mice as previous described [46] . streptomycin (100 μ g/mL, Hyclone) for 48 h at 37 °C with 5% CO 2 in a humidified incubator followed by EP isolation. Timelines for EP administration (10 µg of EP injected at each specified time point) and organ harvest are described in Figure 6. Proportions of detected EP signal was calculated as follows: (RLU/specific organ) / (RLU/all organs) x 100%.
Immunohistochemistry. The kidneys, livers, lungs, and spleen were embedded in Tissue-Tek  Table S2. All sections were imaged using Axio Observer 7 epifluorescent microscope (Zeiss) equipped with iXon Ultra 888 EMCCD (Andor Technology) with or without SRRF with the same imaging parameters. Images (n = 70 per organ) were subjected to Fiji (ImageJ, NIH) for EP quantitative analysis with the same threshold. 5 x 5 tiled images were selected from 6 x 6 tiled images to remove uneven edges following tiling by Fiji (NIH).
Bioluminescence assays. GloMax® Discover Microplate Reader (Promega) with auto-injectors was set for Nluc (450/10 nm) and BRET-FL (510/10 nm) readings, and BRET ratio was calculated by dividing the acceptor signal (510/10 nm) with the donor signal (450/10 nm). To assess EV BL, 5 µL of crude EVs (triplicates), 5 µL of each fraction (triplicates), or 3 µL of pelleted fraction 2-9 (duplicates) were added into a 96-well white plate (Greiner Bio-One, Germany). Furimazine Statistical analyses. Statistical analyses were performed using GraphPad Prism Version 7.05. Twotailed Student's t-test was used for comparison between two groups. Two-way analysis of variance (ANOVA) followed by Tukey post hoc test, or ordinary ANOVA followed by Tukey or Dunnett post hoc test was used for comparison of three or more groups. Values were normally distributed, and the variance was similar between compared groups. Error bars for all the graphs represent mean ± SEM. P-value < 0.05 was considered statistically significant.

Supporting Information
Supporting Information is available from the Wiley Online Library.
Project (AS-CFII108-113)] for assistance on the cell sorting service. We thank Inflammation Core Facility (Institute of Biomedical Sciences, Academia Sinica, Taiwan) for assistance on multiplexed murine cytokine assay. We thank the National RNAi Core Facility (Academia Sinica, Taiwan

Conflict of interest
The authors declare no competing interests.     Meanwhile, shSlco2a1 increased EP distribution to the kidney, heart and brain, whereas shCd13 elevated EP distribution to the kidney, heart and spleen at 72.5h. Proportions for each organ at different time points were acquired by . *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001 with one-way ANOVA followed by Dunnett's post hoc test compared with the Scramble control. N = 3 mice per group with technical triplicates.

Table of contents
PalmGRET, a bioluminescence resonance energy transfer (BRET)-based reporter for extracellular