Investigating the in vivo biodistribution of extracellular vesicles isolated from various human cell sources using positron emission tomography

Abstract

Positron emission tomography (PET) is a powerful tool for investigating the in vivo behavior of drug delivery systems. We aimed to assess the biodistribution of extracellular vesicles (EVs), nano-sized vesicles secreted by cells isolated from various human cell sources using PET. EVs were isolated from mesenchymal stromal cells (MSC EVs), human macrophages (Mϕ EVs), and a melanoma cell line (A375 EVs) by centrifugation and were conjugated with deferoxamine for radiolabeling with Zr-89. PET using conjugated and radiolabeled EVs evaluated their in vivo biodistribution and tissue tropisms. Our study also investigated differences in mouse models, utilizing immunocompetent and immunocompromised mice and an A375 xenograft tumor model. Lastly, we investigated the impact of different labeling techniques on the observed EV biodistribution, including covalent surface modification and membrane incorporation. PET showed that all tested EVs exhibited extended in vivo circulation and generally low uptake in the liver, spleen, and lungs. However, Mϕ EVs showed high liver uptake, potentially attributable to the intrinsic tissue tropism of these EVs from surface protein composition. MSC EVs biodistribution differed between immunocompetent and immunodeficient mice, with increased spleen uptake observed in the latter. PET using A375 xenografts demonstrated efficient tumor uptake of EVs, but no preferential tissue-specific tropism of A375 EVs was found. Biodistribution differences between labeling techniques showed that surface-conjugated EVs had preferential blood circulation and low liver, spleen, and lung uptake compared to membrane integration. This study demonstrates the potential of EVs as effective carriers for various diseases, highlights the importance of selecting appropriate cell sources for EV-based drug delivery, and suggests EV tropism can be harnessed to optimize therapeutic efficacy. Our findings indicate that the cellular source of EVs, labeling technique, and animal model can influence observed biodistribution.

Graphical Abstract

Introduction

Extracellular vesicles (EVs) are secreted from nearly all cell types and are involved in cell-to-cell communication, thereby mediating physiological and pathological processes.^{1, 2} The various types of EVs are generally classified according to their size and biogenesis, including exosomes (~40 – 200 nm), ectosomes (alternatively, microvesicles; ~100 – 1000 nm), and apoptotic bodies (> 500 nm).³ The emergent applications of EVs in medicine have resulted from their ability to efficiently transfer diverse bioactive molecules, such as nucleic acids, proteins, and lipids, to influence the function of target cells.⁴ Additional advantages of EVs in drug delivery applications are that they are considered non-immunogenic, inherently biocompatible, biodegradable, with excellent scalability, and favorable characteristics compared to cell-based therapies.⁵ Consequently, EV-based therapies have become an area of intense interest for the treatment of numerous diseases, including cancer and degenerative diseases.^6–8

Understanding the in vivo biodistribution of EVs is paramount to translating them into successful clinical therapeutics. Tracking exogenous EVs in vivo has been complicated by their complex composition and small size. Moreover, the composition of EVs is influenced by the cell or tissue source, which also affects functionality.⁹ As examples, mesenchymal stromal cell (MSC)-derived and macrophage (Mϕ)-derived EVs have been investigated as therapeutics in regenerative medicine applications or cancer therapy.^10–17 Despite the vast applications of EVs, the in vivo biodistribution of EVs isolated from various tissue sources remains relatively unexplored, particularly using quantitative modalities such as positron emission tomography (PET) imaging.

Herein, we investigated the in vivo fate of EVs isolated from different cell types using PET, a highly sensitive, quantitative, noninvasive technique offering superior tissue penetration and scalability from small animals to humans. EVs isolated from two primary human cell types, MSC and Mϕ, and a metastatic melanoma cancer cell line (A375) were surface conjugated with a chelator (deferoxamine, Df) to enable efficient and stable radiolabeling with Zr-89 (t_1/2 = 74.8 h). PET studies compared the biodistribution of the different EVs in immunocompetent mice, immunodeficient mice, and in a tumor xenograft mouse model. By investigating the in vivo fate of EVs from diverse human cell sources, our study highlights the importance of organotropic EVs properties and their potential impact as therapeutics for various diseases.

Experimental Section

Isolation and Cultivation of cells

Bone marrow-derived mesenchymal stromal cells (MSCs) were first isolated as previously described, from the bone marrow (BM) of healthy human donors.^18–21 Briefly, leftover BM cells from transplant donors were washed with PBS, and mononuclear cells were isolated using Ficoll-Paque Plus density gradient separation. Ammonium-chloride-potassium (ACK) lysis buffer was used to lyse red blood cells (RBCs) with a 3-min incubation as needed. Mononuclear cells were suspended in α-minimum essential medium supplemented with 10% fetal bovine serum, 1% penicillin-streptomycin, 1x nonessential amino acids, and 4 mM L-glutamine. The identity of MSCs was confirmed using flow cytometry [CD90 (+), CD105 (+), CD34 (−), CD45 (−)].^{19, 22} Passage 4 to 8 MSCs were used for EVs isolation (MSC EVs).

Monocytes were isolated from human peripheral blood mononuclear cells (PBMCs) obtained from the blood of healthy G-VCSF mobilized stem cell donors as described previously.^21–24 Briefly, PBMCs were separated using density gradient separation with Ficoll-Paque Plus, and RBCs were lysed using ACK lysis buffer. Platelet contamination in cell suspension was reduced by centrifuging the cell suspension at 300–700 rpm for 10 mins. CD14⁺ monocytes were collected in the resuspended cell suspension by incubating with anti-human CD14 microbeads (Miltenyi Biotec) for 15 mins at 4°C and separated using a MACS Pro Separator (Miltenyi Biotec). Monocytes were cultured using Iscove’s Modified Dulbecco’s media supplemented with 10% human serum blood type AB, 1x nonessential amino acids, 4 mM L-glutamine, 1 mM sodium pyruvate, and 4 μg/mL recombinant human insulin. For differentiation into macrophages, monocytes were cultured for 7 days at 37°C with 5% CO₂ without cytokines.

Human malignant A375 melanoma cells (ATCC; Manassas, VA) were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% FBS per vendor recommendations.

Isolation of EVs from cells

Cells (MSCs, macrophages [Mϕ], A375 cell line) were grown in their respective complete media in T-75 flasks. Prior to EV isolation, cells were washed twice with PBS and media replaced with MSC serum-free media (SFM; StemPro A103332-01, ThermoFisher Scientific). Cells were incubated for 18–24 h in SFM, and the collected conditioned media was centrifuged at a low-speed spin (2000 × g at 4°C for 20 minutes) to remove any detached cells and cell debris. The supernatant was then ultracentrifuged for 2 hours using Optima^™ L-80XP Ultracentrifuge (Beckman Coulter) at 100,000 × g at 4°C. Purified EVs (MSC EVs, Mϕ EVs, and A375 EVs) were then resuspended in PBS and stored at −80°C until further use.

EVs quantification and zeta potential measurements

EVs were quantified using the Nanosight NS300 available at the UW-Madison Wisconsin Center for NanoBioSystems. Data was acquired using a Green laser, camera level 16, syringe pump speed of 70. Three 60 s captures were used to determine the EV particle concentration (p/mL). Data was analyzed at a detection threshold of 4. Zeta potential measurements were acquired using a Zetasizer (Malvern Panalytical).

Surface Protein Analysis of EVs.

MACSPlex Exosome kit (Miltenyi Biotec; Bergisch Gladbach, Germany) was used for surface protein analysis MSC EVs and Mϕ EVs using flow cytometry to detect 37 protein markers and isotype controls. This assay was performed by Zen-Bio (Duham, NC) following manufactures guidelines. The median fluorescent intensities (MFI) of each surface marker were determined by subtracting the MFI of the respective isotype control.

EVs conjugation for PET

EVs were conjugated to p-SCN-Bn-Deferoxamine (Df; Macrocyclics) for PET imaging. In a typical reaction, EVs were conjugated to Df using isothiocyanate chemistry by dissolving Df in anhydrous DMSO and added to EVs at a ratio of 1 nmol to 5×10⁹ EVs in PBS that was pH adjusted to 8.5–9. The reaction was conducted at room temperature for 1–2 h. The conjugated EVs were purified using a PD-10 size exclusion chromatography column.

EVs fluorescent labeling

EVs were conjugated to Alexa fluor 647 NHS ester (ThermoFisher) by linkage to primary amines for IVIS imaging. For the reaction, Alexa fluor 647 NHS ester was dissolved in anhydrous DMSO and added EVs at a ratio of 2.5 × 10⁹ EVs per nmol in PBS pH adjusted to 8.5–9. EVs were labeled by membrane integration of the lipophilic dye 1,1’-Dioctadecyl-3,3,3’,3’-Tetramethylindodicarbocyanine, 4-Chlorobenzenesulfonate salt (DiD’; ThermoFisher) by dissolving in anhydrous DMSO adding to EVs at a ratio of 2.5 × 10⁹ EVs per nmol. Labeled EVs were purified using a PD-10 size exclusion chromatography column.

EVs radiolabeling

The Zr-89 was produced locally at UW-Madison. Conjugated EVs (Df-MSCs, Df-Mϕ EVs, or Df-A375 EVs) were radiolabeled with Zr-89 for PET studies (⁸⁹Zr-Df-MSCs, ⁸⁹Zr-Df-Mϕ EVs, or ⁸⁹Zr-Df-A375 EVs). For radiolabeling, 1–2 mCi (37–74 MBq) of Zr-89 was added to 1 M HEPES buffer (pH 7.5), and Df conjugated EVs were added at a ratio of 2×10¹⁰ EVs per mCi. The reaction mixture was incubated at 37°C for 1 h and labeled EVs purified using a PD-10 size exclusion chromatography column (GE Healthcare). Instant thin layer chromatography (iTLC) was used to determine the radiolabeling efficiency and radiostability in PBS and 50% serum. iTLC was developed using 100 mM EDTA as the mobile phase.

Animal studies

All animal studies were conducted on a protocol approved by the Institutional Animal Care and Use Committee at the University of Wisconsin-Madison. Female Hsd:ICR (CD-1^®) [ICR] mice were purchased from Envigo. Female Athymic nude mice were purchased from Jackson Laboratory. Male NOD scid γc^−/− (NSG) mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ) mice were purchased from Jackson Laboratory and bred at the University of Wisconsin-Madison.

PET/CT imaging studies

PET/CT images were acquired using an Inveon μPET/CT scanner. CT images were captured using the following parameters (80 kV, 900 uA, resolution of 105 μm). Mice were intravenously injected with approximately 5.55 MBq – 7.4 MBq (150 μCi – 200 μCi) of radiolabeled EVs (⁸⁹Zr-Df-MSC EVs, ⁸⁹Zr-Df-Mϕ EVs, ⁸⁹Zr-Df-A375 EVs) for PET studies. Static PET images were acquired at various timepoints post injection (2–3 h, 24 h, 48 h, 72 h, and/or 144 h) and reconstructed using an OSEM3D/MAP algorithm. The Inveon Research Workstation software was used to perform quantitative region of interest analysis of the PET images, with values reported in percent injected activity per gram of tissue (%IA/g). Time activity curves were constructed based on quantification of volumes of interest. Based on the data points and number of mice included in each group (n=3–4), in vivo half-lives were calculated by fitting to a single compartment model. After the final imaging time point, the mice were euthanized, and the major organs were collected for ex vivo biodistribution studies. A Wizard 2 (Perkin Elmer) gamma counter was used to quantify tracer uptake in organs. Biodistribution data was calculated in percent injected activity per gram of tissue (%IA/g) for all organs of interest.

A375 tumor model

To establish A375 tumor xenografts, female nude athymic mice were subcutaneously injected with 2×10⁶ A375 cells on the lower right flank of the mice. Mice were used for PET/CT imaging studies once tumors reached 150–300 mm³, measured by tumor $\mathrm{v}\mathrm{o}\mathrm{l}\mathrm{u}\mathrm{m}\mathrm{e}=(1/2)\times \mathrm{l}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{t}\mathrm{h}\times {\mathrm{w}\mathrm{i}\mathrm{d}\mathrm{t}\mathrm{h}}^2$.

IVIS Imaging

Fluorescent (FL) images were acquired using the In Vivo Imaging System (IVIS; Perkin Elmer). Equimolar amounts of labeled EVs were administered intravenously, based on the fluorophore concentration (400 pmol). Blood samples were collected 1 h, 6 h, and 24 h p.i. After the last blood collection, the mice were euthanized, and the major tissues were collected. Images were acquired at an excitation wavelength of 640 nm and emission wavelength of 680 nm.

Statistical Analysis

Statistical analysis was performed by two-tailed unpaired Student’s t tests. NS, non-significant (*P < 0.05, **P < 0.01, ***P < 0.001).

Results

EV isolation and radiolabeling characterization.

We investigated the in vivo trafficking of EVs derived from three distinctive cell sources: human MSC, Mϕ, and A375 melanoma cells (Figure 1a). As shown in Figure 1b, the average size of the isolated EVs was comparable (MSC EVs: 136 nm, Mϕ EVs: 117.6 nm, and A375 EVs: 123 nm) following sequential ultracentrifugation. Thus, differences in the in vivo biodistribution were likely attributed to the cell type used for EVs isolation rather than to EVs morphology.

Figure 1.

The in vivo biodistribution of EVs isolated from various human cell sources was investigated using positron emission tomography (PET). (a) Cells were isolated and cultured for EVs isolation using sequential ultracentrifugation. EVs were then conjugated with a deferoxamine chelator and radiolabeled with Zr-89 (t_1/2 = 78.4 h) for PET studies. Nanoparticle tracking analysis found that EVs isolated from (b) MSC EVs, (c) Mϕ EVs, and (d) A375 EVs were all similarly sized, with an average diameter of approximately 110–140 nm.

Ideal labeling methods for imaging studies should not disrupt the physicochemical properties of the traced agent while providing high sensitivity and in vivo stability. Additionally, the labeling method will dictate the imaging modality and, thereby, the interpretation of EVs localization in vivo.²⁵ We chose to stably conjugate and chelate EVs with a long half-live positron-emitting radioisotope (Zr-89) for PET/CT studies to investigate the pharmacokinetics of EVs from different cell types. Using PET as our imaging modality incurs minimal EVs modification because of its high detection sensitivity, enabling longitudinal tissue uptake quantification. Deferoxamine (Df; Df-MSC EVs, Df-Mϕ EVs, or Df-A375 EVs) was conjugated to EVs for radiolabeling (⁸⁹Zr-Df-MSC EVs, ⁸⁹Zr-Df-Mϕ EVs, or ⁸⁹Zr-Df-A375 EVs; Figure 2a). The influence of chelator conjugation on EVs size was minimal, as both showed comparable diameter mean (116 nm vs. 118 nm) and mode (80 nm vs. 81 nm) for both MSC and Df-MSCs EVs, respectively (Figure 2b). Moreover, we measured the effects of conjugation on the EVs zeta potential, using MSC EVs and Df-MSC EVs as representative samples (Figure S3). Importantly, no significant zeta potential differences were found between MSC EVs (−12 ± 2 mV) and Df-MSC EVs (−11 ± 4 mV) (p = 0.68). EVs were successfully radiolabeled with Zr-89 with high efficiency. Radiochemical yield following one hour radiolabeling were 88.3 ± 1.4%, 89.6 ± 0.3%, and 96.5 ± 0.8% for ⁸⁹Zr-Df-MSC EVs, ⁸⁹Zr-Df-Mϕ EVs, and ⁸⁹Zr-Df-A375 EVs, respectively, corresponding to specific activity per particle of 1.63 × 10⁻⁹ MBq/EVs, 1.79 × 10⁻⁹ MBq/EVs for ⁸⁹Zr-Df-MSC EVs, ⁸⁹Zr-Df-Mϕ EVs, and ⁸⁹Zr-Df-A375 EVs, respectively (Figure 2c–d; Supporting information, Figure S1). Importantly, we determined the stability of the radiolabeled EVs in PBS and 50% mouse serum (Figure 2e). Using ⁸⁹Zr-Df-MSCs EVs as a representative sample, labeled EVs showed long-term stabilities of 91.6 ± 5.0% and 85.0 ± 2.9% after 72 h of incubation in PBS and 50% mouse serum, respectively. Thus, we demonstrated that EVs can be conjugated without altering physical properties and that they could be efficiently and stably radiolabeled with Zr-89, enabling in vivo PET imaging studies.

Figure 2.

Conjugation and radiolabeling of EVs for PET. (a) Schematic showing the conjugation of the chelator (Df) to the surface of EVs. (b) Nanoparticle tracking analysis showed no differences in the size distribution of MSC EVs and Df-MSC EVs after conjugation. (c) Representative instant thin layer chromatography used to determine the (d) radiolabeling efficiency, which exceeded 85% for all radiolabeled EVs (n=3). (e) Stability studies of ⁸⁹Zr-Df-MSC EVs showing that radiolabeling was stable in PBS and 50% mouse serum up to 72 h following incubation (n=3).

Biodistribution of radiolabeled EVs in immunocompetent mice.

The equivalent properties of native and conjugated EVs and their effective radiolabeling with Zr-89, enabled investigating the longitudinal biodistribution of EVs from different cell sources following intravenous injection by PET/CT imaging, first in healthy ICR mice (Figure 3). Volume of interest (VOI) analysis of serial PET/CT images quantified uptake in major organs of interest (Figure 4a–g; Supporting Information, Table S1–3). We observed prolonged circulation half-lives for all EVs, which largely cleared from the blood at 72 h post-injection (p.i.). Fitting the time-activity curves of the heart showed blood circulation half-lives of 12.4 ± 0.3 h, 14.4 ± 1.1 h, and 12.3 ± 0.6 h for ⁸⁹Zr-Df-MSC EVs, ⁸⁹Zr-Df-Mϕ EVs, and ⁸⁹Zr-Df-A375 EVs, respectively. Interestingly, these results contrast with previous studies reporting EVs blood circulation of less than 5 min.²⁶ The primary clearance route for all the EVs was hepatic, showing liver uptake of ⁸⁹Zr-Df-MSC EVs, ⁸⁹Zr-Df-Mϕ EVs, and ⁸⁹Zr-Df-A375 EVs reaching maximums at 2 h p.i. of 6.9 ± 0.9 %IA/g, 26.3 ± 4.4 %IA/g, of 7.7 ± 0.3 %IA/g, respectively. The liver uptake for all EVs slowly decreased throughout the imaging study. The most notable difference in distribution between the imaged EVs was seen in liver uptake of ⁸⁹Zr-Df-Mϕ EVs, which ranged from 3.3–4.5-fold higher across imaging timepoints compared to either ⁸⁹Zr-Df-MSC EVs or ⁸⁹Zr-Df-A375 EVs. Interestingly, no other notable differences between the EVs uptake in the spleen, lung, kidney, muscle, or bone were observed. All EVs had minimal uptake in the lungs or kidneys, suggesting no apparent aggregation or destabilization following EVs administration. Uptake in highly perfused organs, including the lung and kidney, decreased throughout the imaging study in conjunction with the blood clearance. Minimal radioactivity was detected in the bone, indicating low levels of bone-seeking “free” Zr-89 and confirming the in vivo stability of the radiolabeled EVs. Ex vivo biodistribution studies using harvested tissues following the final imaging time point at 72 h p.i. were performed to validate PET/CT results and to further quantify EVs uptake in other major organs of interest (Figure 4h; Supporting Information, Table S4). In line with the imaging data, the most prominent uptakes for each radiolabeled EV were found in the liver (⁸⁹Zr-Df-MSC EVs: 6.45 ± 0.55 %IA/g, ⁸⁹Zr-Df-Mϕ EVs: 29.04 ± 3.41 %IA/g, ⁸⁹Zr-Df-A375 EVs: 8.92 ± 0.52 %IA/g) and spleen (⁸⁹Zr-Df-MSC EVs: 5.93 ± 0.60 %IA/g, ⁸⁹Zr-Df-Mϕ EVs: 10.62 ± 4.26 %IA/g, ⁸⁹Zr-Df-A375 EVs: 6.76 ± 0.52 %IA/g). The uptake of the imaged EVs in the other major organs correlated to the VOI analysis and appeared non-specific, with all values less than 3.5 %IA/g. Based on these results, we conclude EVs exhibited excellent in vivo circulation and generally had similar biodistribution. Nonetheless, significant differences in the EVs biodistribution were observed for some specific organs, highlighting the importance of considering the cell source used for EVs production when developing therapeutic agents.

Figure 3.

PET imaging of EVs in ICR mice. Following conjugation and radiolabeling, ⁸⁹Zr-Df-MSC EVs, ⁸⁹Zr-Df-Mϕ EVs, and ⁸⁹Zr-Df-A375 EVs were intravenously injected in ICR mice and serial PET/CT images were acquired at various times post-injection. Maximum intensity projections of the fused PET/CT images are shown.

Figure 4.

Quantitative PET analysis of EVs distribution in healthy mice. Volume of interest quantification of PET images was performed on (a) blood, (b) liver, (c) spleen, (d) lung, (e) kidney, (f) muscle, and (g) bone at each imaging timepoint (n=3). (h) Ex vivo biodistribution studies quantified uptake in the major organs following the final imaging timepoint at 72 h post-injection (n=3).

Analysis of EVs surface protein composition.

To further investigate the basis for the observed differences in biodistribution between EVs isolated from various human cell types, we examined their surface marker composition. Using representative EVs showing the most divergent biodistribution (MSC EVs and Mϕ EVs), we analyzed the surface protein composition for differences in 37 established EV surface markers by MACSplex flow cytometry (Figure 5; Supporting Information, Table S5).²⁷ Interestingly, Mϕ EVs showed positive expression of 20 markers (CD3, CD4, CD9, CD11c, CD14, CD29, CD31, CD40, CD44, CD45, CD56, CD63, CD81, CD86, CD105, CD105, CD133/1, CD142, CD146, HLA-DR, HLA-ABC). In comparison, the MSCs EVs showed a common set of seven markers (CD9, CD29, CD44, CD63, CD81, CD105, CD146). Further comparison found differences in the tetraspanin, proteins enriched in EVs during their biogenesis,²⁸ with CD9 and CD63 enriched in Mϕ EVs (P = 0.02 and P=0.01, respectively). Mϕ EVs were positive for other markers specific to human macrophages or hematopoietic cells that were absent in MSC EVs, including CD11c (P = 0.05), CD14 (P = 0.3), CD40 (P=0.001), CD45 (P = 0.006), CD86 (P = 0.06), HLA-ABC (P = 0.35), and HLA-DR (P = 0.009), although not all differences were statistically significant. Other notable markers enriched in Mϕ EVs compared to MSC EVs were related to adhesion, including CD31 (P = 0.005) and CD44 (P = 0.009). Interestingly, only CD146 was enriched in the MSC EVs compared to the Mϕ EVs Such proteomic difference between EVs isolated from various cell types, have been previously demonstrated.²⁹ and indicate that surface protein composition, which seems to resemble that of it parental cells, may play an important role in dictating EVs biodistribution and should be further investigated in additional studies.

Figure 5.

Surface protein comparison of MSC EVs and Mϕ EVs. MACSPlex EV analysis found that the surface protein composition of MSC EVs and Mϕ EVs varied substantially.

EVs Biodistribution in immunodeficient mice.

Research applications often require using immunodeficient mice to establish human cancer models or investigate the efficacy of different therapies.²¹ Therefore, we used serial PET/CT imaging to examine the impact of a dysfunctional adaptive immune system on the biodistribution of ⁸⁹Zr-Df-MSC EVs using healthy immunocompromised NSG mice lacking lymphoid T-cells and B-cells, and NK cells (Figure 6a).³⁰ As we previously observed, ⁸⁹Zr-Df-MSC EVs displayed prolonged circulation and minimal uptake in the lungs or liver in NSG mice. Using VOI analysis (Figure 6b; Supporting Information, Table S6), we found that ⁸⁹Zr-Df-MSC EVs were predominately cleared from the blood by 72 h p.i., with a blood circulation half-life of 11.6 ± 0.9 h. High spleen uptake of ⁸⁹Zr-Df-MSC EVs in the NSG mice was evident from the initial imaging timepoint The ⁸⁹Zr-Df-MSC EVs exhibited hepatic excretion with liver uptake plateauing at 24 h at 6.20 ± 1.01 %IA/g. As in immunocompetent mice, minimal uptake of free Zr-89, peaking at 2.0 ± 0.5 %IA/g 24 h p.i., was observed in the bones at all time points. Ex vivo biodistribution studies after the final imaging time point at 144 h p.i. determined exosome uptake in the major tissues of interest (Figure 6c; Supporting Information, Table S7). The spleen had the most prominent uptake of ⁸⁹Zr-Df-MSC EVs (55.6 ± 11.2 %IA/g), which could be attributed to the EVs glycosylation status.³¹ The significant difference between the spleen uptake quantified from VOI analysis and ex vivo biodistribution studies can be explained by partial volume effects associated with the small spleen size in NSG mice and limited spatial resolution of PET.³² The uptake of ⁸⁹Zr-Df-MSC EVs in the other major organs, however, correlated to the VOI analysis and appeared non-specific, with only the liver having uptake greater than 5 %IA/g. This imaging study confirmed the excellent in vivo circulation of MSC-derived EVs, but the observed prominent spleen uptake in NSG mice reaffirmed the necessity for performing distribution studies in different mouse models. These results are critical to inform therapeutic studies involving MSC-derived EVs that require utilization of NSG mice, as divergent biodistribution patterns in this mouse strain might confound interpretation of experimental results.

Figure 6.

PET imaging of EVs in immunocompromised NSG mice. (a) Serial maximum intensity projection PET/CT images of healthy NSG mice intravenously injected ⁸⁹Zr-Df-MSC EVs. (b) Volume of interest quantification of ⁸⁹Zr-Df-MSC EVs at various timepoints post-injection (n=3). (d) Ex vivo biodistribution studies of ⁸⁹Zr-Df-MSC EVs following the final imaging timepoint at 144 h post-injection (n=3). Marked liver and spleen uptake were observed in these mice. Incongruences in spleen uptake between imaging and ex vivo results are likely attributed to partial volume effects.

PET imaging of EVs in cancer xenograft model.

Due to the potential of EVs as cancer therapies,³³ we explored the tumor tropism of MSC EVs and A375 EVs using an A375 melanoma xenograft tumor model. PET/CT images were acquired after intravenously injecting ⁸⁹Zr-Df-MSC EVs or ⁸⁹Zr-Df-A375 EVs in mice bearing A375 tumors (Figure 7a). Tissue uptake of the radiolabeled EVs was quantified by VOI analysis (Supporting Information, Table S8–9). The blood circulation of ⁸⁹Zr-Df-MSC EVs and ⁸⁹Zr-Df-A375 EVs was observed up to 48 h p.i., with half-lives of 15.0 ± 1.5 h and 13.1 ± 0.9 h. Notably, tumor uptake of ⁸⁹Zr-Df-MSC EVs and ⁸⁹Zr-Df-A375 EVs was observed at all time points and reached maximums at 24 h p.i. of 5.0 ± 0.8 %IA/cc and 4.3 ± 0.4 %IA/cc, respectively, and gradually declined until the final time point at 72 h p.i. (Figure 7b). No statistical differences in tumor uptake were found at any of the imaging timepoints (P > 0.15 for all timepoints). Relatively low uptake of the radiolabeled EVs was found in the liver and spleen, which reached maximums at 3 h p.i. when blood-pool activity was highest and was then slowly eliminated up to 72 h p.i.. Ex vivo biodistribution studies of major tissues after the terminal imaging time point at 72 h p.i. quantified the uptake of ⁸⁹Zr-Df-MSC EVs and ⁸⁹Zr-Df-A375 EVs (Figure 7c; Supporting Information, Table S10). The highest tissue uptake was observed in the liver (⁸⁹Zr-Df-MSC EVs = 6.4 ± 1.8 %IA/g, ⁸⁹Zr-Df-A375 EVs = 7.5 ± 1.9 %IA/g; p = 0.3) and spleen (⁸⁹Zr-Df-MSC EVs = 6.4 ± 2.0 %IA/g, ⁸⁹Zr-Df-A375 EVs = 5.0 ± 1.3 %IA/g; p = 0.23). The uptake in the A375 tumors was comparable, which was measured as 4.4 ± 2.4 %IA/g for ⁸⁹Zr-Df-MSC EVs and 3.0 ± 1.0 %IA/g for ⁸⁹Zr-Df-A375 EVs (p=0.3).

Figure 7. PET imaging of ⁸⁹Zr-Df-MSC EVs and ⁸⁹Zr-Df-A375 EVs in a murine A375 tumor model.

(a) Serial maximum intensity projection PET images in A375 tumor bearing mice intravenously injected ⁸⁹Zr-Df-MSC EVs and ⁸⁹Zr-Df-A375 EVs show marked tumor uptake that peaked at 24 h post-injection (p.i.). Yellow arrow indicates tumor location. (b) Volume of interest quantification of ⁸⁹Zr-Df-MSC EVs and ⁸⁹Zr-Df-A375 EVs uptake at various timepoints p.i. (n=4). Interestingly, no increased tumor uptake was observed for EVs derived from the A375 cells over MSC cells. (c) Ex vivo biodistribution studies of ⁸⁹Zr-Df-MSC EVs and ⁸⁹Zr-Df-A375 EVs following the final imaging timepoint at 144 h p.i. (n=4) found predominate hepatic clearance and similar tumor uptake of the radiolabeled EVs, as seen in PET.

Comparison of EV labeling strategies using fluorescence imaging.

As a prerequisite for in vivo imaging, EVs must be modified with a labeling agent. Our previous PET imaging studies employed a covalent surface modification strategy which may influence our results. IVIS imaging was utilized to compare imaging of EVs labeled covalently with a fluorophore (Alexa fluor 647 [AF647]) or using by membrane integration with a lipophilic dialkylcarbocyanine dye (DiD’), the most common EV labeling strategy.^{26, 34} Differences in the biodistribution of EVs labeled by covalent modification (AF647-MSC EVs or AF647-Mϕ EVs) or by using a membrane integrating dye (DiD’@MSC EVs) were evaluated in blood samples collected up to 24 h p.i. (Figure 8a–b; Supporting Information, Table S11) and in the major organs of collected ex vivo 24 h p.i. (Figure 8c–d; Supporting Information, Table S12). Quantification of blood samples found AF647-MSCs ranged from 3.7-fold to 6.5-fold higher than DiD’@MSC EVs across the timepoints (p=0.00011, 0.00014, 0.04 at 1 h, 6 h, 24 h, respectively). These results suggest significantly prolonged blood circulation of EVs when labeled by covalent modification compared to membrane integration. Importantly, initial signal in the blood of AF647-MSC EVs was significantly higher than AF647-Mϕ EVs (p=0.02), which correlates with our results from PET. Further differences in biodistribution of EVs labeled using the different methods was found by semi-quantitatively analyzing tissues ex vivo. Notably, signal in the liver and spleen of DiD’@MSC EVs was 3.6-fold and 11.2-fold higher than AF647-MSC EVs, respectively. Thus, we attributed decreased DiD’@MSC EVs blood circulation to increased liver and spleen uptake. However, signal in the liver was 1.5-fold higher for AF647-Mϕ EVs compared to AF647-MSC EVs, which correlated to the increased liver uptake of Mϕ EVs observed using PET/CT imaging.

Figure 8. Effects of EV labeling strategies on the observed biodistribution.

IVIS images of (a, b) blood and (c, d) ex vivo tissues from mice administered MSC EVs or Mϕ EVs labeled by surface conjugation with AF647 (AF647-MSC EVS or AF647-Mϕ EVs) or by membrane integration of a lipophilic dye (DiD’@MSC EVs). Quantification of IVIS image signal from (e) blood and (f) tissues found significant differences in the observed biodistribution between labeling strategies (n=3).

Discussion

Understanding the in vivo biodistribution and pharmacokinetics, which may be influenced by the parental cell type, is fundamental to successfully developing EV-based therapeutic agents. In addition, properly understanding these properties is key to establishing EVs dosing route and regimen, understanding their mechanism of action, and exposing potential off-target effects leading to unintended toxicity. Utilizing PET approaches is ideal for understanding the biodistribution of EVs over time because of its quantitative ability, high sensitivity, full tissue penetration, and information on biological and biochemical processes. Investigating the pharmacokinetic profile through noninvasive imaging informs platform and model selection, tumor uptake, and off-target targeting toxicity concerns. Through these studies, safer and more effective therapeutics can be designed, which ultimately may accelerate clinical translation.

Our PET study suggested an interdependence between EVs biodistribution profiles and its originating cell type. ⁸⁹Zr-Df-MSC EVs and ⁸⁹Zr-Df-A375 EVs showed similar biodistribution with minimal uptake in the liver, kidneys, spleen, and lungs. However, ⁸⁹Zr-Df-Mϕ EVs showed fast and prominent accumulation in the liver, with minimal decline in uptake over the 72 h observation period. The contrasting results in the liver and spleen could be explained by an inherent tropism of the Mϕ-derived EVs to these macrophage-rich tissues. As such, EVs organotropism may be influenced by the composition of their membranes, favoring organs that are hosts to the parental cells from which they are derived. Studies have described similar cell tropisms in biological processes or drug delivery applications using cell carriers, a property that may be likely passed to the cellular-derived EVs.^35–37 Moreover, we found our the surface protein composition varied amongst the MSC EVs and Mϕ EVs, which may dictate the observed differences in biodistribution. Thus, this intrinsic tropism of cellular sources of EVs is an essential factor to consider when optimizing the design of drug delivery platforms for a specific disease.

Additional mouse models, including immunocompromised mice and tumor models were utilized to further investigate the EVs biodistribution. PET studies in immunocompromised NSG mice found significantly increased spleen uptake. Previous studies have found similarly high spleen uptake of human monoclonal antibodies in NSG mice, attributed to Fc region – Fc receptor binding on myeloid cells in these mice with low to no endogenous immunoglobulins.³⁸ Moreover, Fc region – Fc receptor interaction and overall spleen uptake were correlated to the degree of glycosylation on the Fc region of the antibody.³⁹ Glycans are a fundamental component on the surface of exosomes, and the importance of glycosylation status on the observed spleen uptake of NSG mice should be investigated in future studies.^{31, 40} Separate PET studies demonstrated efficient delivery of systemically administered EVs to tumors, corroborating the potential of EVs as carriers of drugs or engineered bioactive molecules for cancer therapy. Interestingly, no significant tropism resulting in increased A375 tumor uptake for A375 EVs compared to MSC EVs was observed. Thus, it is likely that the tumor uptake in this model was governed by the enhanced permeability and retention effect, which may be increased by adding a targeting component.^{41, 42} Engineering EVs for cancer therapies are among the most promising therapeutics undergoing clinical trials (ClinicalTrials.gov identifiers: NCT05156229, NCT04592484).^43–45 The tumor accretion could be further improved by engineering EVs to express molecular binders targeting tumor antigens, which may further increase their specificity and potency as drug delivery platforms. This approach will require an in-depth understanding of the in vivo distribution profile of the engineered EVs using noninvasive PET methodologies like the ones described herein.

In our PET study, EVs were labeled by covalent attachment to the primary amine groups present on surface proteins, which affix the imaging label to the EVs surface to limit background signal.^46–49 Importantly, we found negligible effects of the surface conjugation on the size or surface charge of the conjugated EVs. Other strategies have been investigated for labeling EVs, with the most common strategy being membrane integration. Membrane integration labeling most frequently employs lipophilic molecules, such as dialkylcarbocyanine dyes (DiD, DiR, etc.), PKH dyes (PKH26, PKH67, etc.), or radioactive labels that integrate into the EVs lipid bilayer non-specifically.^50–56 Using lipophilic membrane incorporating labels can result in the exchange between EVs and cells in the subject that confounds image interpretation. Although simple, this labeling method is limited by background from dissociated labels, induction of EVs clumping, and inability to distinguish between lipid proteins and micelles. Using this labeling strategy, a previous study investigated the influence of cell source, route of administration, and targeting on EVs biodistribution.⁵⁷ However, the FL imaging employed was neither quantitative nor able to reliably track EVs biodistribution in vivo with accurate spatiotemporal resolution because of the limited penetration of light in tissues, poor label stability, and low sensitivity requiring the loading of a large number of dyes onto the EVs. Thus, FL imaging only showed minor differences in EVs biodistribution isolated from different cell types. Moreover, differences in the observed biodistribution were found when using membrane-incorporated dyes, which could be attributed to label dissociation from the EVs lipid bilayer. Released or transferred label may concentrate in liver and spleen cells or membrane. In comparison, covalently modified EVs affix the label to the EVs surface. Thus, the advantages and limitations of various imaging labeling strategies must be carefully considered in any application with conditions optimized to attain meaningful results describing the in vivo biodistribution of EVs.

Our study focuses on the EV properties that can influence molecular imaging studies, specifically EV’s cellular source (i.e., Mϕ EVs, MSC EVs, or A375 EVs, mouse model (i.e., ICR, athymic nude, or NSG mice), labeling strategy (i.e, covalent surface modification or lipid integration), and imaging modality (PET vs fluorescent imaging). Other reviews have also highlighted that many variables in experimental design can influence EV imaging studies,³⁴ such as labeling methodology, cell source, imaging modality, administration route, among others. These studies have found major differences in the biodistribution of EVs in tissues such as liver, lung, bladder, blood, tumor, etc. Our work further highlights the results of imaging studies should be interpreted with attention to the specific variable of that study.

Conclusion

Herein, we studied the in vivo biodistribution of EVs derived from various human cell types using PET. To this end, we efficiently and stably radiolabeled EVs without altering their physical properties. Our PET studies in immunocompetent mice show labeled EVs had excellent in vivo circulation, with the liver being the most prominent target organ. Additionally, we detected some organotropic biodistribution for EVs that was attributed to surface protein composition differences. Additionally, PET studies using immunodeficient mice revealed that the biodistribution of EVs was significantly elevated spleen uptake compared to that of immunocompetent mice, which may be due to the myeloid cell composition in the spleen of NSG mice and glycosylation status of EVs. Proper utilization of PET imaging in the development of EVs should significantly influence and accelerate their clinical translation. Our results demonstrate the significant potential of EVs in diverse therapeutic applications and the importance of selecting the right cell source and animal model for a given application. Noninvasive PET imaging is a powerful tool for answering drug development questions, which will hopefully spur additional interest in EV-based technologies. More importantly, the methods described herein can be employed to noninvasively interrogate the in vivo distribution of any other exosomes.