key: cord-0776852-xi9yj3uj authors: Khan, Azalea A.; T. M. de Rosales, Rafael title: Radiolabelling of Extracellular Vesicles for PET and SPECT imaging date: 2021-02-13 journal: Nanotheranostics DOI: 10.7150/ntno.51676 sha: 42354e26a8a4678548125213f0e3ad76f744f330 doc_id: 776852 cord_uid: xi9yj3uj Extracellular vesicles (EVs) such as exosomes and microvesicles have gained recent attention as potential biomarkers of disease as well as nanomedicinal tools, but their behaviour in vivo remains mostly unexplored. In order to gain knowledge of their in vivo biodistribution it is important to develop imaging tools that allow us to track EVs over time and at the whole-body level. Radionuclide-based imaging (PET and SPECT) have properties that allow us to do so efficiently, mostly due to their high sensitivity, imaging signal tissue penetration, and accurate quantification. Furthermore, they can be easily translated from animals to humans. In this review, we summarise and discuss the different studies that have used PET or SPECT to study the behaviour of EVs in vivo. With a focus on the different radiolabelling methods used, we also discuss the advantages and disadvantages of each one, and the challenges of imaging EVs due to their variable stability and heterogeneity. Extracellular vesicles (EVs) are cell-derived phospholipid bilayer enclosing vesicles. Once thought to be 'garbage bags' used by cells to excrete unwanted molecules, our current understanding of their function includes cell-to-cell communication and expanded their potential applications to the fields of medical diagnostics and therapy [1] [2] [3] [4] . One of the earliest reports on the potential of EVs as therapeutics was in 1996, when EVs were shown to trigger an adaptive immune response [5] , but it wasn't until 2007 that the first demonstration that EVs carry functional RNA, and can deliver it to other cells was reported [6] . Since then EVs have been shown to contain a variety of other cell-derived cytosolic molecules, such as proteins, lipids, nucleic acids, among others [7] . Consequently, later research revealed the important role EVs play in cell-cell communication [4, 8] and their association in various diseases, such as cancer and metastasis [9] [10] [11] , neurodegenerative diseases [12, 13] , diabetes [14] , and inflammatory conditions [15] [16] [17] . Recent advances in EV engineering have also demonstrated the possibility of chemically modifying these cell-derived vesicles to improve their therapeutic potential, including the introduction of targeting vectors, stability groups (e.g. PEGylation) as well as the possibility of drug loading, expanding their therapeutic potential even further [18] . For all these reasons, EVs have gained substantial recent attention as potential biomarkers of disease as well as nanotherapeutics. EVs have been classified in three subtypes, based on their cellular origin as well as their size (Fig. 1) : (i) exosomes (30-150 nm) , (ii) microvesicles (50-1000 nm), and (iii) apoptotic bodies (500-2000 nm). Exosomes are nanovesicles that are formed by inward budding of the endosomal membrane, and released into the extracellular space by exocytosis [19, 20] . Exosomal membrane is enriched with tetraspanins, such as CD63, CD81 and CD9; as well as endosomal proteins, such as Alix and TSG101 [7] . Unlike Ivyspring International Publisher exosomes, microvesicles are formed by outward budding of the cell membrane with abundant presence of phosphatidylserine in the outer layer of the bilayer [21, 22] . Apoptotic bodies are blebs formed when cells are undergoing apoptosis. However, a recent review outlined that apoptotic bodies are more than just membrane blebs, but are important modulators of immune response [23] . Theoretically, different EVs can be isolated from body fluids or cell culture supernatant by using different (ultra)centrifugation speeds [3] , size exclusion chromatography (SEC) or other techniques [24] . However, isolation of individual EV populations is challenging due to their overlap in size and complex physicochemical properties. For example, common exosome biomarkers CD63, CD81 and CD9 have been reported to be present in other EV sub-groups dependent on the cell type [25, 26] . Hence, adhering to the Minimal Information for Studies of Extracellular Vesicles (MISEV) 2018 guidelines, in this review EVs would be identified by their size [27] . Vesicles of sizes < 200 nm would be defined as small EVs (sEVs), vesicles in the 200-500 nm size range would be defined as medium EVs (mEVs), and vesicles > 500 nm would be defined as large EVs (lEVs). Of the different subtypes of EVs described above, sEVs or exosomes (< 200 nm) have been explored for their potential as therapeutic nanomedicines [28] [29] [30] . The goal of nanomedicine is to improve the therapeutic effects of drugs by improving targeting to disease sites and/or reducing their systemic toxic side effects [31] . Several properties make sEVs an attractive platform for nanomedicine. Compared to other platforms of synthetic origin (e.g. liposomes, polymers), natural EVs are antigen presenting, and depending on their cell of origin may have specific tissue-targeting properties [5, 32, 33] . sEVs can also be engineered to express ligands for better tissue/cell targeting and stability [34, 35] , can be used as drug delivery vehicles [36, 37] , and have been reported to accumulate in tumours due to the enhanced permeability and retention (EPR) effect [38, 39] . Another important property often cited is their innate ability of crossing the blood-brain barrier, which is a significant biological barrier for other nanomedicinal drug delivery platforms [40] [41] [42] . sEVs from immune cells may have therapeutic properties by themselves. For example, neutrophil-derived sEVs have been shown to induce inflammatory resolution [43] , while sEVs containing foreign antigens/molecules have the ability to trick the immune system, as is the case for many viral diseases [44] . Recent reports have also shown that sEVs can be more efficient as gene therapy and drug delivery vehicles than systems based on liposomes [45] [46] [47] . All these exciting results have resulted in many EV-based therapies progressing quickly towards clinical trials, with over 15 trials registered on NIH ClinicalTrials.gov, including 3 studies looking into treatment for COVID-19, and successful completion of six studies [48] [49] [50] [51] [52] [53] . However, there is still a lack of information regarding their fate in vivo after administration, particularly when these investigational therapeutic nanomedicines are used in humans. To tackle this lack of information and efficiently develop EV-based therapies, it is important to integrate imaging-based in vivo tracking techniques early in the developmental process, to help answer questions regarding their fate after administration into patients/subjects. This will facilitate not only their development but also the progress of clinical trials, allowing the early identification of the most suitable candidates to take forward. As a developmental tool, non-invasive imaging of EV therapies would be an ideal method to monitor and quantify EV biodistribution over time and enable elucidation of their pharmacokinetic profiles. Furthermore, taking into account patient and disease heterogeneity, integrating imaging with EV-based therapies could provide a tool to identify patients/lesions that are likely to respond to the treatment. Such personalised medicine approach would allow optimisation of treatment strategies to suit individual patients' needs [54] . Of the several imaging techniques that have been utilised to track EV biodistribution over time, optical imaging (OI) is the most popular for its simplicity, and for being both cost and time effective. For these reasons, it has been used in many studies to understand and characterise EV properties [55] . Nonetheless, OI suffers from inherent physical limitations, mainly poor tissue depth penetration hence its application is limited to the preclinical setting (for whole-body animal imaging), clinical intraoperative and in vitro/ex vivo imaging. Other imaging techniques don't suffer from this limitation, particularly computed tomography (CT), magnetic resonance imaging (MRI), and the nuclear medicine techniques -single photon emission computed tomography (SPECT) and positron emission tomography (PET). All of these allow whole-body imaging with unlimited imaging signal depth penetration [56] . Furthermore, they are available in both the preclinical and clinical setting. MRI and CT benefit from excellent spatiotemporal resolution but suffer from low sensitivity (i.e. large amount of contrast agent required to allow detection). This is a major limitation for EV imaging, as the concentration of contrast agent per EV required for efficient imaging signal can be potentially damaging to the vesicles. This is perhaps one of the reasons behind the low number of studies that used MRI for EV imaging [57] [58] [59] [60] [61] [62] . CT has only been used in combination with nuclear imaging techniques (SPECT or PET) for EV imaging. These two techniques offer higher sensitivity compared to CT/MRI (ca. 10 6 -fold), no background signal, and allow whole body imaging with accurate signal quantification. By radiolabelling EVs with appropriate radioactive isotopes, or radionuclides, it is possible to track/image EVs using either SPECT or PET. The main differences between SPECT and PET imaging lies in the type of radioisotope used, and how the signal is detected and converted into 3D images (Fig. 2) . SPECT detect radionuclides that decay by emission of γ photons. These γ photons have definite energies; for example 99m Tc, the most used radionuclide in nuclear medicine with a decay half-life (t 1/2 ) of 6 h, emits 140 keV γ photons. These radionuclides are detected by a rotating gamma camera to generate a 3D image. Collimators are used to only allow radiation at certain angles to reach the detector and thus determine the source ( Fig. 2A ) [63] . PET, on the other hand, uses radionuclides that decay by positron (β + ) emission. Positrons travel for a short distance (depending on their energy, sub-mm to several mm) before annihilating with an electron, an event that produces two γ rays of equal energy (511 keV) simultaneously in opposite directions. These γ photons are detected by a circular ring of detectors that maps out simultaneously arriving γ rays at 180°, a technique known as coincidence detection (Fig. 2B ) [64] . These differences in detection techniques and instrumentation lead to differences in image quality and sensitivity. In general terms, clinical PET is superior to SPECT in some aspects as outlined in Table 1 below. SPECT, however, has the major advantage that allows multi-radionuclide imaging with radionuclides that emit γ rays of different energies. In this review we focus on the different methods explored to date to radiolabel EVs that allow in vivo tracking with both PET and SPECT imaging. We critically review the main advantages and disadvantages of each method, and where possible discuss them into the context of their in vivo imaging properties. For reviews on imaging EVs using other imaging techniques, we point the reader to excellent publications available [41, 55, 66, 67] . A total of 16 published articles and pre-prints that included radiolabelling of EVs and EV-mimetics were found (July 2020, see Supporting Information for methodology) and analysed in this review, of which 11 were on SPECT imaging and 5 of them on PET imaging. These articles were further classified into the different types of radiolabelling methods discussed below (Fig. 3) . Due to the similarities in their physical structures, the different chemical concepts that allow radiolabelling of liposomes can also be applied to EVs [68] . Specifically, both EVs and liposomes are similar in size and consist of a phospholipid bilayer enclosing an aqueous core capable of carrying a chemical cargo. A recent review by Man et al. outlines the different methods that have been employed to radiolabel and image liposomal systems to date, including their pros and cons [69] . Similar to liposomes, there are also two primary methods for radiolabelling EVs -surface and intraluminal -that we will briefly describe below (Fig. 4) . Surface or membrane radiolabelling is the most common method to radiolabel EVs. This involves incorporation of the radionuclide directly into the lipid membrane or attached to the membrane proteins, either directly or via covalent chemical bond formation. Four main methods have been applied to achieve this: 1) genetic modification, 2) direct incorporation of radionuclide into the membrane, 3) radionuclide attachment via a chelator (i.e. radiometal-binding chemical group) on the surface, and 4) direct incorporation of radionuclide into membrane proteins (Fig. 4A) . Direct incorporation of radionuclides relies on non-specific affinities between the radionuclides and the EV membrane components. For the surface chelation methods the radiotracer is conjugated to the membrane, typically, via surface amine groups using standard bioconjugate chemistry techniques. More detailed descriptions behind these methods are described in the 'Radiolabelling of EVs using SPECT radionuclides' and 'Radiolabelling of EVs using SPECT radionuclides' sections below with the corresponding radiotracers (vide infra). A potential limitation of surface radiolabelling methods in general is that they could compromise the integrity of the EV's surface, particularly if they involve chemical modification of membrane proteins, or substantial modification of the membrane composition. The importance of EV surface proteins and lipids in their behaviour has been demonstrated in various studies [32, [70] [71] [72] [73] [74] , and a recently published review described the different methods to modify the surface of EVs [75] . These alterations are usually applied to understand EV function or to improve tissue targeting; hence those aimed at enabling radiolabelling have significant potential to alter their natural physicochemical properties and biodistribution. [65] and James et al. [56] . Geometric efficiency (percentage of detected to emitted γ rays) (concentration of radiotracer needed) SPECT Collimator detection ~ 0.01% Minutes 8-10 mm 10 -10 to 10 -11 M PET Coincidence detection ~ 1% Seconds -minutes 5-7 mm 10 -11 to 10 -12 M An alternative approach to radiolabel EVs is to entrap the radiotracer inside the intra-vesicular space (Fig. 4B) . Thus, the lipid bilayer membrane is expected to protect the radionuclide from trans-chelation by extra-vesicular components, such as serum proteins. This is in contrast to surface radiolabelling, where the radionuclide is more exposed to extra-EV trans-chelation. To radiolabel EVs intraluminally, the radionuclide needs to cross the lipid bilayer and stay within the EV. To achieve this, two methods have been explored to date: 1) remote loading, and 2) ionophore-chelator binding. The first method takes advantage of endogenous intravesicular glutathione that is capable of transforming some complexes, such as [ 99m Tc]Tc-hexamethylpropyleneamine oxime ([ 99m Tc]Tc-HMPAO), from lipophilic to hydrophilic [76] . Once the lipophilic radiotracer complex passes through the lipid bilayer membrane, it is converted into its hydrophilic form and thus, gets trapped inside the aqueous core of the EV. The ionophore-chelator binding method exploits well-known ionophore ligands, such as tropolone and 8-hydroxyquinolin (oxine), that form a metastable and neutral complex with the radiometals, allowing them to be transported across the lipid membrane. This method is commonly used to radiolabel cells as well as liposomal nanomedicines, where the radiometal can bind to metal chelating moieties within the liposomal cargo [77] . In the case of EVs, the radiometal is expected to bind to intravesicular proteins and/or nucleic acids. The main disadvantage of intraluminal radiolabelling approaches, especially those based on ionophores, is that we lack the knowledge of exactly which component of the EVs' intraluminal space the radionuclide binds to. This may complicate interpretation of in vivo images, particularly at late timepoints when EV lipid bilayer fragmentation may be significant and due to the fact that some radionuclides -particularly radiometals such as 64 Cu -that accumulate in the same organs as EVs (e.g. liver, spleen) [69] . The two main radiolabelling categories described above, surface and intraluminal, have been employed with both SPECT and PET radionuclides with various degrees of radiolabelling capabilities. In the next sections we will describe the different EV radiolabelling studies reported to date, that we have classified in to SPECT radionuclides ( Table 2 ) and PET radionuclides ( Table 3 ). The first report reporting radiolabelling of sEVs was by Morishita et al. in 2014 (publication in 2015, but available online from Nov 2014) [78] . Using plasmid transfection, lactadherin -a glycoprotein found on cell membrane -was replaced with streptavidin on B16-BL6 cells and cultured for sEV isolation. Streptavidin containing B16-BL6 sEVs (70 ± 3 nm) were radiolabelled with biotin-conjugated 125 I (t 1/2 = 59.4 d), [ 125 I]I-IBB. Strong binding affinity (K d ~10 15 M -1 ) makes the streptavidin-biotin conjugate a popular pretargeting method in radioimmunotherapy [94] . The radiolabelling yield (RLY) of B16-BL6 sEVs with [ 125 I]I-IBB was ~80%, and in vitro serum stability was ~95% for up to 4 h. Despite blood clearance of 125 I-labelled B16-BL6 sEVs having the same profile as the free radiotracer ([ 125 I]I-IBB) in healthy mice (Fig. 5A) , their ex vivo biodistributions were significantly different with 125 I-labelled B16-BL6 sEVs showing accumulation in liver, lung and spleen (Fig. 5B) . Surprisingly, 125 I uptake in the thyroid was not mentioned in the paper. Thyroid is an organ of interest when using iodine as a sign of radiochemical instability with iodinated radiotracers, due to the presence of the human Na + /Isymporter (hNIS). Stomach uptake, another organ with high affinity for free iodide via hNIS was very low, suggesting in vivo radiochemical stability during the 4 h study. This study, however, is the only study to carry out mathematical pharmacokinetic modelling of radiolabelled sEVs. The same group later published another study based on the same radiolabelling method [79] , and reported higher retention of sEVs in xenograft tumours of up to 200 mm 3 volume compared to larger tumours of up to 500 mm 3 when injected intratumorally (Fig. 5C) . Although useful as a preclinical tool, genetic modification is a challenge in terms of clinical translation. Radiolabelling with iodine can easily be performed using iodination beads (iodo-bead method), which consists of a polystyrene bead coated with an oxidising agent facilitating the reduction of tyrosine residues and iodine substitution [95] , thus iodinating EV surface proteins. Using this method, Rashid et al. achieved > 80% RLY with 131 I (t 1/2 = 8 d) for sEVs derived from 4T1 cells [80] . Although up to 80% of the radiolabelled sEVs were stable in serum for 24 h; in vivo imaging at early time-points (ca. 3 h) showed high thyroid, stomach and bladder uptake for both tumour cell-and healthy cell-derived sEVs, which correlates with release of free 131 I (Fig. 5D ). It has been previously reported that radio-iodination using the iodo-bead method is prone to rapid deiodination in vivo as early as 2 h post injection [96] . Hence, it seems that sEVs radiolabelled using the iodo-bead method suffer from low in vivo radiochemical stability and cannot reliably be used to determine their biodistribution. Table 2 . Summary of reports of EV radiolabelling with SPECT radioisotopes. The hydrodynamic size of EVs are stated for unmodified EVs before radiolabelling, as appropriate. Radiolabelling condition column shows EV and radiotracer incubation time, temperature, and the amount of EVs used per reaction. Data shown as reported by the authors. RLY = radiolabelling yield, UC = ultracentrifugation, UF = ultrafiltration, SEC = size exclusion chromatography, RT = room temperature, iTLC = instant thin layer chromatography; * data taken from the figures. The long half-lives of radioiodines are ideal for long term in vivo tracking of EVs. Nevertheless, 99m Tc, with a shorter half-life of 6 h, is the most commonly used radionuclide for imaging of EVs probably due the availability/low cost and favourable radiation properties of this radionuclide for imaging. The first use of 99m Tc- 3 ] + ) to label red blood cell-(RBC) derived sEVs (188 ± 11 nm) [81] . Unlike liposomes, that lack appropriate donor ligands and require surface modification [97] , EVs should be able to bind to [ 99m Tc(CO) 3 ] + via surface proteins, most likely involving histidine donors [98] . A RLY of 38.8 ± 6.2% was achieved after 30 min incubation at room temperature, a relatively low RLY which is not surprising as efficient binding to this inert complex requires high temperatures incompatible with biomolecules [99] . An in vivo imaging comparison between [ 99m Tc(CO) 3 ] + and [ 99m Tc(CO) 3 ] + -RBC-sEVs is consistent with efficient radiolabelling and high stability for the latter. However, the short imaging timeframe of the study (< 2 h) does not allow evaluation of the long-term in vivo stability of [ 99m Tc(CO) 3 ] + -RBC-sEVs (Fig. 6A-B) . A more recent report also exploited the use of [ 99m Tc(CO) 3 ] + to radiolabel human epidermal growth factor receptor 2 (HER2) targeted-HEK 293T sEVs [82] . Contrary to the previous study, a high RLY was achieved by incubation of [ 99m Tc(CO) 3 ] + with the sEVs at a relatively low temperature of 37 °C for 60 min. In vivo imaging showed expected high liver uptake, but no spleen, as well as significant kidney and intestine signal that could potentially indicate free [ 99m Tc(CO) 3 ] + . An alternative surface radiolabelling method with 99m Tc involves reduction of the unreactive [ 99m TcO 4 ]in the +7 oxidation state to 99m Tc 4+ using stannous chloride (SnCl 2 ), a commonly used RBC radiolabelling method [100] . Gonzalez et al. optimised this method for sEV radiolabelling and achieved 37 ± 9% RLY using 2 mM SnCl 2 and 75 µg of milk-derived sEVs (122 ± 1 nm) [83] . Using this method, 99m Tc was incorporated directly into the milk-derived sEV membrane, although the exact binding site of the 99m Tc 4+ ion is unclear and is likely to be non-specific binding to surface proteins. In vitro stability within 48 h in PBS was high, although it would have been interesting to test with more challenging conditions (e.g. in the presence of serum components). Looking into different administration routes the authors noted significant differences in the biodistribution depending on the injection route, with mainly liver and spleen uptake via intravenous (iv.) administration, unspecific abdominal distribution via intraperitoneal administration, and mainly digestive distribution after intranasal administration, with some minor brain uptake (Fig. 6C) . Despite being the most widely used radionuclide, the short half-life of 99m Tc only allows imaging for up to 24 h post administration, which is not suitable for long term in vivo tracking of EVs. This can be resolved by using 111 In (t 1/2 = 2.8 d), another clinically available gamma-emitting radionuclide that is the second most commonly used for EV imaging after 99m Tc (Fig. 3) . Smyth et al. used [ 111 In]In-oxinate (intraluminal labelling) to radiolabel sEVs derived from PC3 (140 ± 59 nm) and MCF7 (130 ± 57 nm) cells, with RLY of 81% and 67%, respectively [85] . These differences in RLY could be due to different affinities of the intraluminal composition of the two types of sEVs for the same radiometal. Blood clearance of the two sEVs was fast, and similar in tumour bearing mice as well as that of liposomes (Fig. 7A) . High uptake in liver, spleen and kidneys, but low tumour uptake was reported in ex vivo data at 24 h for both healthy and tumour bearing mice (Fig. 7B) . High kidney uptake could be a sign of "unchelated" 111 In release, but this was not evaluated. In a very interesting study, Rashid et al. also used intraluminal radiolabelling ([ 111 In]In-oxinate) to track CD206-positive M2 macrophage-specific sEVs (106 ± 14 nm) in vivo using SPECT imaging [86] . The authors used [ 111 In]In-oxinate to achieve very high RLY (98%) and in vitro stabilities, as assessed by thin layer chromatography. In vivo SPECT-CT images using M2-targeted sEVs showed the expected liver and spleen uptake seen with other sEVs, but also interesting uptake in lymph nodes (Fig. 7C ) -a finding that has also been observed with 89 Zr-labelled PANC1 sEVs (vide infra) [93] -and lungs. In addition, M2-targeted sEVs showed increased uptake in tumours, as well as significant kidney and bladder signal that the authors assign to excretion of radiolabelled sEVs. Although it would have been preferable to report image scales to allow image comparisons and quantify organ/tumour uptake using normalised standard units (ca. %ID/mass, %ID/volume, SUV), instead of %ID or counts/mass, these results show the potential of engineering sEVs for cell-specific targeting. Similar to [ 111 In] In-oxinate, [ 111 In]In-tropolone is another lipophilic radiotracer that allows intraluminal radiolabelling, but with higher inertness/stability compared to [ 111 In] In-oxinate that makes it more resistant to transchelation, and hence potentially less effective for intraluminal radiolabelling. In fact, [ 111 In]In-oxinate often results in higher cell RLYs when compared to [ 111 In]In-tropolone [101] . Faruqu et al. used [ 111 In]In-tropolone to achieve a RLY of 4.73 ± 0.39% with B16-F10 sEVs (132 ± 6 nm) [84] . In this study SEC using Sepharose ® CL-2B resin was used to separate unbound radiotracer, a technique that potentially results in losses of ca. 50% of the vesicles. The authors also performed membrane radiolabelling of the same sEVs using [ 111 In]In-DTPA (RLY = 19.2 ± 4.5%). Overall, the pharmacokinetics and tissue biodistribution differed slightly depending on whether they were radiolabelled on the membrane or intraluminally (Fig. 7D) . The relatively lower radiochemical stability of the [ 111 In]In-tropolone labelled sEVs was evident in the in vivo biodistribution with lower liver/spleen retention. This study also demonstrated that there are no significant differences in sEV biodistribution between immunocompetent (C57BL/6) and immunocompromised (NOD SCID gamma (NSG)) mice, except minor differences in tumour uptake, that could be explained by the smaller population of tumour associated macrophages in NSG mice. It is well established that intra-vesicular nucleic acids and proteins can be used as biomarkers for various diseases. Recently, it has been reported that the glycan profiles of EVs can also be used as cancer biomarkers [102] . Moreover, enrichment of specific glycoproteins, such as sialic acid [103] , allow iv. injected EVs to be captured by CD169 positive macrophages in spleen and lymph nodes [104] . Royo et al. evaluated the effect of glycosylation on sEV biodistribution by radiolabelling them with 124 I (t 1/2 = 4.2 d) [89] . Using iodination tubes (iodogen method), which allows oxidation of 124 I and thus radiolabels proteins on cell membranes [95] , and a RLY of >15%, they have shown that modified glycosylation allows sEVs to accumulate and retain in the lungs after iv. injection; and to migrate through the lymph system after hock injection (injection in the joints), even though a large amount was retained in the injection site (Fig. 8A) . The in vivo PET imaging data showed that sEVs start accumulating in the liver as early as 30 s after iv. injection, and joint administration results in expected lymphatic drainage. Although it is difficult to compare the in vivo images without appropriate scale bars; there were only minor differences (lung uptake) between glycosylated or non-glycosylated sEVs, with both showing increasing signal in bladder and thyroid over time, the latter being a sign of in vivo deiodination (Fig. 8A-B) . Continuing with membrane labelling, Banerjee et al. used [ 64 Cu]Cu-DOTA-Maleimide (t 1/2 = 12.7 h) to radiolabel sEVs (~110 nm) [90] . DOTA was attached on the sEVs' surface using endogenous membrane thiol groups. This allowed ca. 20% RLY of umbilical cord blood cell sEVs. Interestingly, in vivo imaging showed brain uptake 20 -60 min post injection (Fig. 8C) . A considerable amount of sEV uptake was seen in the bladder, which increased over time as the liver signal decreased (Fig. 8D) . Interestingly adding a further SEC purification following ultracentrifugation increased sEV accumulation in liver and spleen but decreased in urine/bladder (discussed in the "Challenges" section below). Facilitated by the specific glycan profile, the rapid uptake of EVs by liver and spleen, results in very short blood circulation time when injected iv. [105, 106] . PEGylation is commonly used in nanomedicine to overcome such issues [107] . Likewise, PEGylated EVs have been shown to improve circulation, reduce liver sequestration and improve tumour uptake [108] . Shi et al. was the first group to study in vivo biodistribution of PEGylated EVs, using [ 64 Cu]Cu-NOTA [91] . Using the same (or less) amount of sEVs (~ 106 nm) as Banerjee and his group, they achieved considerably higher RLY (91.2 ± 0.2% vs. ~ 20%). This could be the result of the well-known high radiochemical stability of [ 64 Cu]Cu-NOTA complexes [109, 110] . PEGylation had the desired effect of increased sEV circulation, decreased liver uptake and increased tumour uptake, particularly after 24 h (Fig. 8E-F) . Uptake in the lymph nodes (ca. 2% ID/g) for both PEGylated and non-PEGylated was also identified from the ex vivo biodistribution data, but not discussed further (Fig. 8G) . In another report involving surface radiolabelling with PET radionuclide, Jung et al. isolated sEVs from 4T1 breast cancer cells (~ 100 nm) and radiolabelled them with 64 Cu and 68 Ga, via NOTA-isothiocyanate conjugation to sEV surface amine groups (from surface proteins) and also labelled with Cy7 to allow optical imaging [92] . RLY with 64 Cu was > 98%, which is comparable to what was reported by Shi et al. RLY and serum stability of > 95% by TLC were only reported for 64 Cu-labelled sEVs. Healthy mice were injected iv. or subcutaneously. As expected from this administration route, lymph node uptake was observed after subcutaneous injection, and the PET signal correlates well the fluorescent signal. High uptake in the lungs, liver and spleen was observed for iv. injected sEVs. Furthermore, presence of CD63 and Cy7 positive sEVs in liver, spleen and lymph nodes were confirmed by histological analysis. To date there is one report on intraluminal radiolabelling of sEVs using PET radionuclides [93] . Khan [77, 111, 112] , to radiolabel three different cancer cell-derived sEVs with RLYs as high as ca. 20% for 1 x 10 11 PANC1 sEVs (97 ± 4 nm) (Fig. 9A-B) , and ca. 76% radiochemical stability in PBS over 26 h. An alternative in vitro stability was demonstrated in a cell-uptake study in the presence of serum proteins, showing significant differences between cell lines over 4 h in the uptake of 89 Zr-PANC1 sEVs vs. control groups. In vivo PET imaging in healthy mice showed the expected biodistribution of sEVs with uptake in liver and spleen, and interestingly several lymph nodes and brain, particularly at early timepoints ( Fig. 9C-D) . Furthermore, significant differences were observed in the uptake of 89 Zr-PANC1 sEVs and heat-damaged 89 Zr-PANC1 sEVs. The latter group was aimed at denaturing the radiolabelled PANC1 sEVs to study the effect in the biodistribution. In particular, significant lower spleen uptake was observed at all time-points, leading to the suggestion that spleen/bone and liver/bone ratios may be used as in vivo markers of stability of sEVs radiolabelled using [ 89 Zr]Zr-oxinate (Fig. 9E) . Variability in EV production, isolation and radiolabelling properties can be overcome by using exosome mimetic vesicles (EMVs). EMVs are produced mainly by serial extrusion of cell membrane, and so far used for radiolabelling [113] , but can also be produced by other methods [68] . The main advantage of EMVs is that they can be mass produced in larger concentration than EVs. They are comparable to EVs in size, protein and lipid composition, biomarker expression, and retain tissue targeting properties. 10 MDA-MB-231.CD63-GFP sEVs and 1x10 11 PANC1 sEVs; data given as mean ± SD of n = 3. (C) In vivo PET-CT images of C57BL/6 mice injected iv. with 89 Zr-PANC1 sEVs, heat-damaged 89 Zr-PANC1 sEVs and 89 Zr-control, 1 h post injection; B = bladder, † = PET image scale for 89 Zr-control is 10 times that of the other images; adjusted for image clarity. (D) Image slices of a mouse injected with intact 89 Zr-PANC1 sEVs showing uptake in brain; image scale is the same as in C. (E) Ex vivo biodistribution of "intact" (n = 3) and "heat-damaged" (n=2) 89 Zr-PANC1 sEVs. Ratio of liver:bone uptake (n = 3) and spleen:bone uptake (n = 2) is shown on the right; data given as mean ± SD of the n values and analysed by Student's t-test. Figures taken from Khan et al. [93] . Using mouse macrophage derived EMVs (218 ± 8 nm), Hwang et al. managed to achieve RLY of >93% with [ 99m Tc]Tc-HMPAO (t 1/2 = 6 h) [87] . They utilised the endogenous intra-vesicular thiol groups to convert the lipophilic [ 99m Tc]Tc-HMPAO complex into a hydrophilic one, thus trapping the radionuclide in the intraluminal space, unlike the surface thiol groups used previously by Banerjee et al. In vivo biodistribution of two different types of 99m Tc-labelled EMVs were compared with mouse macrophage derived sEVs, and although comparisons are difficult from the images reported, it seems that significant release of the [ 99m Tc]-HMPAO complex and/or [ 99m TcO 4 ]is evident (Fig. 10A) . Later on, Gangadaran et al. also studied biodistribution of 99m Tc-labelled RBC derived EMVs (201 ± 16 nm). Their radiolabelling hypothesis is based on the RBC radiolabelling protocol, where 99m Tc 4+ reduced by SnCl 2 binds to intercellular haemoglobin [114] . In this case, 99m Tc bound to haemoglobin inside the EMVs (Fig. 10B) . With ~100% RLY after 20 min incubation, no further purification was used. Healthy C57BL/6 mice were injected iv. with 99m Tc-labelled EMVs and gamma camera imaging was performed, as opposed to SPECT. Nonetheless, no significant difference was observed between 1 h and 3 h post injection for the EMVs, unlike for free 99m Tc. Once again, significant signal from the liver/spleen and bladder was observed with the radiolabelled EMVs. (Fig. 10C-D) . Subsequently, due to the efficiency and stability of radiolabelling, the group used the 99m Tc-labelled EMVs to successfully radiolabel and track white blood cells in vivo [115] . [87] (B) Schematic representation of the protocol for radiolabelling red blood cell-derived EMVs with 99m Tc. (C) In vivo gamma camera images of iv. administered free 99m Tc and 99m Tc-EMVs in male C57BL/6 mice 1 h and 3 h post injection. (D) Quantification of free 99m Tc and 99m Tc-EMVs injected in organs of interest 1 h and 3 h post injection; data presented as mean ± SD of n = 4 and analysed by Student's t-test. Figures adapted with permission from Gangadaran et al. [88] . One of the biggest challenges in the in vivo imaging of EVs is their instability. A study by Clayton et al. demonstrated abundant presence of CD55 and CD59 on EVs, that could lend themselves longer survivability in vivo [72] . However, according to the EV radiolabelling studies discussed above, as well as studies using non-radiolabelled EVs [104, 116] , blood half-life of iv. administered EVs was found to be as short as < 2 min. However, presence of EVs could be detected in the reticuloendothelial system (RES) organs, particularly in liver and spleen, long after clearance from blood. After liver and spleen, the highest accumulation of imaging signals is observed in the bladder; only possible if the EVs are able to pass through the glomerular filtration in the kidneys. Using quantum dots, Choi et al. demonstrated that, nanoparticles >8 nm are not typically cleared by the kidneys [117] . This study suggested that the renal filtration size threshold maybe comparable to small proteins and was further demonstrated by antibody clearance data. Thus, renal clearance of intact antibodies is considered insignificant because the size of a typical antibody (e.g. ~150 kDa for IgG) is much larger than the glomerular filtration threshold (~55 kDa) (Fig. 11A ) [118] . Antibody fragments, on the other hand, are much smaller (for example, the size of the Fab fragment of IgG is ~50 kDa) and are cleared by kidneys [119] . Hence it seems to us that the kidney/bladder uptake observed in many SPECT/PET imaging studies with radiolabelled EVs is likely to be the result from fast disintegration of EVs in blood/serum. The radioactive hot-spots seen in bladder in the SPECT and PET images are either from the radiotracer itself (that taking into account that bladder uptake is often observed, regardless of radiolabelling method, makes this possibility improbable in many cases) or from EV components that are attached to the radiotracer. Interestingly, the supplementary data reported by Banerjee et al. [90] , showing that EV purification by combining ultracentrifugation with SEC leads to lower accumulation in urine, seems to support this hypothesis (Fig. 11B) . Previous studies have shown that sEVs isolated by ultracentrifugation can lead to co-precipitation of serum proteins (such as albumin) [120] , with lower purity as determined by the particle-to-protein ratio, compared to SEC [121] despite co-isolation of lipoproteins using this method [122] . Wei et al. have shown that by combining these two methods, it is possible to improve elimination of contaminating proteins, lipoproteins and to improve particle-to-protein ratio [123] . Nevertheless, further evidence is needed to determine if this is the case for kidney/bladder/urine uptake of EVs. Another of the main issues with many EV radiolabelling studies reviewed above is the use of instant thin layer chromatography (iTLC) to measure radiochemical stability in serum. When using appropriate stationary and mobile phases, for example Whatman N o 1 paper as the stationary phase and ethyl acetate or EDTA as the mobile phase for [ 89 Zr]Zr-oxinate, the lipophilic unbound radiotracer or leaked 89 Zr 4+ ions migrates to the solvent front (R f = 1) respectively, whereas radioactivity bound to the EVs stays at the origin (R f = 0) because EVs are not soluble and precipitate in this solvent system. This technique is appropriate when assessing radiochemical stability in aqueous buffers such as PBS. On the other hand, in serum, any trans-chelation of the radiotracer by the serum proteins would also be detected at the origin because proteins will also precipitate in the presence of organic solvents, even at low concentrations; making it difficult to distinguish whether the radioactivity is bound to EVs or serum proteins. The field of EV research and in particular the applications of EVs as nanomedicinal tools has sparked a recent remarkable interest from many researchers. From all the research carried out to this effect, it is clear that using imaging techniques, whether optical or radionuclide-based, is certainly beneficial in elucidating EV biology and behaviour, and should be integrated into this research early in the process to facilitate their development as potential therapies/diagnostic tools. In this review we focussed on EV research to date that includes radionuclide-based imaging, and in particular on the radiolabelling methods used and findings from SPECT/PET imaging. In these studies, two main categories of radiolabelling methods were used -intraluminal and membrane-based. Our review illustrates the importance of choosing the most appropriate radiolabelling method suitable for the downstream application, as each have their advantages and disadvantages. Of all these radiolabelling methods, two main ones stand out in terms of efficiency and stability: (i) direct conjugation of a bifunctional radionuclide-complex (e.g. NOTA-maleimide) onto surface components and (ii) the use of ionophores such as oxine for intraluminal labelling. Ideally, studies that compare these two methods such as that from Faruqu et al. should be performed prior to any PET/SPECT EV study to identify the best suited method for each EV and application [84] . It is important to emphasise that when radiolabelling EVs via the surface/membrane, we will be tracking these membrane components over time, whether they are still a part of the EV or not. Likewise, when we radiolabel intraluminally we will only be tracking the EVs as long as their contents have not been released. For this reason, regardless of the radiolabelling method, it is important to have a clear understanding of the biodistribution/pharmacokinetics of the radionuclide and/or the radio-complex used for EV radiolabelling. We think this is a crucial component of any EV PET/SPECT imaging experiment as many radionuclides/radio-complexes share biodistribution and excretion pathways with nanoparticulates (e.g. liver, lymph nodes, tumours) [69] . Ideally, the radiolabelling method of choice should not affect the natural properties of EVs (e.g. hydrodynamic diameter, membrane protein structure and composition, colloidal stability), and result in physicochemically and radiochemically stable products. In this regard, we believe there is a significant risk when using direct attachment of radionuclides or radionuclide-based bifunctional chelators onto the surface of EVs, as these are likely to affect the structure of important surface proteins/molecules that the EVs use for their function. One of the main findings of this review is the high number of studies that show accumulation of imaging signal/EVs in the bladder, making it unlikely that this is simply due to impurities from the radiolabelling process. As we discussed in the previous section, we believe that this may be linked to small EV fragments as a result of fast EV decomposition. Further studies are required to test this hypothesis, and we believe that radionuclide imaging will play an important role in evaluating this. Finally, it is commonly accepted in the EV field that even a carefully isolated pure microvesicle or exosome population can be highly heterogeneous, and contains subgroups of vesicles with distinct properties [26, 124, 125] . This highlights the benefits of integrating in vivo EV research with radionuclidebased imaging (that benefits from high sensitivity and quantification properties) to allow us improving our understanding of the basic EV biology of each of these subgroups and their interaction with other cells/tissues. In addition, being a clinically available imaging modality offers the possibility to develop/optimise the methodology for potential future clinical studies. Biogenesis, Secretion, and Intercellular Interactions of Exosomes and Other Extracellular Vesicles Biological properties of extracellular vesicles and their physiological functions Membrane vesicles as conveyors of immune responses Extracellular Vesicles: Unique Intercellular Delivery Vehicles B lymphocytes secrete antigen-presenting vesicles Exosome-mediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells Extracellular vesicles: exosomes, microvesicles, and friends Imaging exosome transfer from breast cancer cells to stroma at metastatic sites in orthotopic nude-mouse models Extracellular vesicles as modulators of the cancer microenvironment The key role of extracellular vesicles in the metastatic process Exosomes and their role in tumorigenesis and anticancer drug resistance The role of exosomes in the pathogenesis of Alzheimer' disease The emerging role of exosomes in mental disorders Extracellular vesicles in type 2 diabetes mellitus: key roles in pathogenesis, complications, and therapy Emerging role of extracellular vesicles in inflammatory diseases The Contrasting Role of Extracellular Vesicles in Vascular Inflammation and Tissue Repair Exosomes in inflammation and role as biomarkers How extracellular vesicles can enhance drug delivery Vesicle formation during reticulocyte maturation. Association of plasma membrane activities with released vesicles (exosomes) Exosomes: composition, biogenesis and function Activated platelets release two types of membrane vesicles: microvesicles by surface shedding and exosomes derived from exocytosis of multivesicular bodies and alpha-granules Microvesicles: mediators of extracellular communication during cancer progression Apoptotic Cell-Derived Extracellular Vesicles: More Than Just Debris Progress in Exosome Isolation Techniques Proteomic comparison defines novel markers to characterize heterogeneous populations of extracellular vesicle subtypes Reassessment of Exosome Composition Minimal information for studies of extracellular vesicles 2018 (MISEV2018): a position statement of the International Society for Extracellular Vesicles and update of the MISEV2014 guidelines Exosomes in Cancer Nanomedicine and Immunotherapy: Prospects and Challenges Perspective in Nuclear Theranostics Using Exosome for the Brain Stem Cell-Derived Exosomes as Nanotherapeutics for Autoimmune and Neurodegenerative Disorders Nanocarriers as an emerging platform for cancer therapy Glycosylated extracellular vesicles released by glioblastoma cells are decorated by CCL18 allowing for cellular uptake via chemokine receptor CCR8 Toward tailored exosomes: the exosomal tetraspanin web contributes to target cell selection Surface functionalized exosomes as targeted drug delivery vehicles for cerebral ischemia therapy Biological characteristics of exosomes and genetically engineered exosomes for the targeted delivery of therapeutic agents Extracellular vesicle-based drug delivery systems for cancer treatment Extracellular vesicles as drug delivery systems: Why and how? Adv Drug Deliv Rev, in press Dynamic biodistribution of extracellular vesicles in vivo using a multimodal imaging reporter Gold nanoparticle based double-labeling of melanoma extracellular vesicles to determine the specificity of uptake by cells and preferential accumulation in small metastatic lung tumors Delivery of siRNA to the mouse brain by systemic injection of targeted exosomes In Vivo Neuroimaging of Exosomes Using Gold Nanoparticles In Vivo Tracking of Multiple Tumor Exosomes Labeled by Phospholipid-Based Bioorthogonal Conjugation Neutrophil-derived microvesicles enter cartilage and protect the joint in inflammatory arthritis Exosomes in Viral Disease Exosomes facilitate therapeutic targeting of oncogenic KRAS in pancreatic cancer mTHPC-loaded extracellular vesicles outperform liposomal and free mTHPC formulations by an increased stability, drug delivery efficiency and cytotoxic effect in tridimensional model of tumors Exosomal delivery of doxorubicin enables rapid cell entry and enhanced in vitro potency A phase I study of dexosome immunotherapy in patients with advanced non-small cell lung cancer Vaccination of metastatic melanoma patients with autologous dendritic cell (DC) derived-exosomes: results of thefirst phase I clinical trial Phase I clinical trial of autologous ascites-derived exosomes combined with GM-CSF for colorectal cancer Dendritic cell-derived exosomes as maintenance immunotherapy after first line chemotherapy in NSCLC Umbilical cord mesenchymal stem cells derived extracellular vesicles can safely ameliorate the progression of chronic kidney diseases Effects of mesenchymal stem cells and their exosomes on the healing of large and refractory macular holes Image-guided drug delivery: preclinical applications and clinical translation Imaging extracellular vesicles: current and emerging methods A Molecular Imaging Primer: Modalities, Imaging Agents, and Applications Magnetic Resonance Imaging of Melanoma Exosomes in Lymph Nodes Magnetic resonance imaging of ultrasmall superparamagnetic iron oxide-labeled exosomes from stem cells: a new method to obtain labeled exosomes Labeling and Magnetic Resonance Imaging of Exosomes Isolated from Adipose Stem Cells Biodistribution of gadolinium-and near infrared-labeled human umbilical cord mesenchymal stromal cell-derived exosomes in tumor bearing mice Imaging of extracellular vesicles derived from human bone marrow mesenchymal stem cells using fluorescent and magnetic labels Visualization of exosomes from mesenchymal stem cells in vivo by magnetic resonance imaging The Gamma Camera: Basic Principles Positron emission tomography: An overview PET versus SPECT: strengths, limitations and challenges Tracking Exosomes in Vitro and in Vivo To Elucidate Their Physiological Functions: Implications for Diagnostic and Therapeutic Nanocarriers Phenotypic analysis of extracellular vesicles: a review on the applications of fluorescence Exosomes and Exosome-Inspired Vesicles for Targeted Drug Delivery Nuclear imaging of liposomal drug delivery systems: A critical review of radiolabelling methods and applications in nanomedicine Clinical applications of exosome membrane proteins An emerging focus on lipids in extracellular vesicles Antigen-presenting cell exosomes are protected from complement-mediated lysis by expression of CD55 and CD59 Adhesion and signaling by B cell-derived exosomes: the role of integrins Tumour exosome integrins determine organotropic metastasis Surface functionalization strategies of extracellular vesicles A simple method for producing a technetium-99m-labeled liposome which is stable In Vivo Exploiting the Metal-Chelating Properties of the Drug Cargo for In Vivo Positron Emission Tomography Imaging of Liposomal Nanomedicines Quantitative Analysis of Tissue Distribution of the B16BL6-Derived Exosomes Using a Streptavidin-Lactadherin Fusion Protein and Iodine-125-Labeled Biotin Derivative After Intravenous Injection in Mice Accelerated growth of B16BL6 tumor in mice through efficient uptake of their own exosomes by B16BL6 cells Differential in vivo biodistribution of 131I-labeled exosomes from diverse cellular origins and its implication for theranostic application Radiolabeling of Extracellular Vesicles with (99m)Tc for Quantitative In Vivo Imaging Studies 99mTc-radiolabeled HER2 targeted exosome for tumor imaging Radioactive Labeling of Milk-Derived Exosomes with 99mTc and In Vivo Tracking by SPECT Imaging Membrane Radiolabelling of Exosomes for Comparative Biodistribution Analysis in Immunocompetent and Immunodeficient Mice -A Novel and Universal Approach Biodistribution and Delivery Efficiency of Unmodified Tumor-Derived Exosomes Generation of Novel Diagnostic and Therapeutic Exosomes to Detect and Deplete Protumorigenic M2 Macrophages Noninvasive imaging of radiolabeled exosome-mimetic nanovesicle using In vivo Non-invasive Imaging of Radio-Labeled Exosome-Mimetics Derived From Red Blood Cells in Mice Modification of the glycosylation of extracellular vesicles alters their biodistribution in mice A positron-emission tomography (PET)/magnetic resonance imaging (MRI) platform to track in vivo small extracellular vesicles Copper-64 Labeled PEGylated Exosomes for In Vivo Positron Emission Tomography and Enhanced Tumor Retention Highly Sensitive Identification of Lymphatic and Hematogenous Metastasis Routes of Novel Radiolabeled Exosomes Using Non-invasive PET Imaging Allows Direct Radiolabelling and PET Imaging of Small Extracellular Vesicles Pretargeted radioimmunotherapy of cancer: progress step by step Radiopharmaceuticals and General Methods of Radiolabelling. Fundamentals of Nuclear Pharmacy Iodination of Annexin V for Imaging Apoptosis Radiolabeling and Quantitative In Vivo SPECT/CT Imaging Study of Liposomes Using the Novel Iminothiolane-99mTc-Tricarbonyl Complex Development of Annexin V Mutants Suitable for Labeling with Tc(I)-Carbonyl Complex Tc(CO)3]+ chemistry: a promising new concept for SPET? Radionuclide-labeled red blood cells: Current status and future prospects Neutrophil labeling with indium-111: tropolone vs. oxine Glycoconjugates from extracellular vesicles: Structures, functions and emerging potential as cancer biomarkers Glycosylation of extracellular vesicles: current knowledge, tools and clinical perspectives CD169 mediates the capture of exosomes in spleen and lymph node Pharmacokinetics of Exosomes-An Important Factor for Elucidating the Biological Roles of Exosomes and for the Development of Exosome-Based Therapeutics Extracellular vesicle in vivo biodistribution is determined by cell source, route of administration and targeting Nanoparticle PEGylation for imaging and therapy Cancer cell-type tropism is one of crucial determinants for the efficient systemic delivery of cancer cell-derived exosomes to tumor tissues Comparison of (64)Cu-complexing bifunctional chelators for radioimmunoconjugation: labeling efficiency, specific activity, and in vitro/in vivo stability Promising bifunctional chelators for copper 64-PET imaging: practical (64)Cu radiolabeling and high in vitro and in vivo complex stability In Vivo PET Tracking of 89Zr-Labeled Vγ9Vδ2 T Cells to Mouse Xenograft Breast Tumors Activated with Liposomal Alendronate PET Imaging of Liposomal Glucocorticoids using 89Zr-oxine: Theranostic Applications in Inflammatory Arthritis Bioinspired Exosome-Mimetic Nanovesicles for Targeted Delivery of Chemotherapeutics to Malignant Tumors Characteristics of Specific Radiopharmaceuticals. Fundamentals of Nuclear Pharmacy White blood cell labeling with Technetium-99m (99mTc) using red blood cell extracellular vesicles-mimetics Visualization and in vivo tracking of the exosomes of murine melanoma B16-BL6 cells in mice after intravenous injection Renal clearance of quantum dots Antibody Pharmacokinetics and Pharmacodynamics Therapeutic antibody fragments with prolonged in vivo half-lives Isolation of Exosomes from Blood Plasma: Qualitative and Quantitative Comparison of Ultracentrifugation and Size Exclusion Chromatography Methods How pure are your vesicles? Comparison of small extracellular vesicles isolated from plasma by ultracentrifugation or size-exclusion chromatography: yield, purity and functional potential Combination of Size-Exclusion Chromatography and Ultracentrifugation Improves the Proteomic Profiling of Plasma-Derived Small Extracellular Vesicles Cells release subpopulations of exosomes with distinct molecular and biological properties Exosomes purified from a single cell type have diverse morphology Supplementary materials. http://www.ntno.org/v05p0256s1.pdf The authors have declared that no competing interest exists.