Internationally harmonized guidelines are available and cover the quality, safety, efficacy and multidisciplinary issues required for drug manufacturing and development. The International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) attempts to connect regulatory authorities and the pharmaceutical industry in order to discuss scientific and technical aspects of drug registration at the global level. ICH’s mission is to achieve worldwide harmonization to ensure that safe, effective, and high-quality medicines can be developed and registered efficiently. In this article, we refer to these harmonized ICH guidelines, which are grouped, coded and available online in a logical, consistent and clear way (available online: http://www.ich.org/products/guidelines
3.1. Extracellular Vesicles (EVs) or Vesicular Secretome Fractions (VSFs)?
A critical element in the development of future therapeutics is that of regulatory compliance, obtaining a manufacturing license and, finally, a market authorization. For this, product specifications related to “purity, identity, quantity, potency and sterility” need to be defined in accordance with the regulations for pharmaceutical manufacturing. In order to be released for distribution and clinical use, each product batch has to meet predefined quality criteria. If release criteria are not reached because of “out of specification” results indicating deviations from defined and expected quality, the product has to be rejected, and distribution or clinical use of this special batch are prohibited. The issue of quality can be resolved in several ways; and as long as the procedures are validated and the specified ranges match the manufacturer’s requirements, there are few obstacles remaining.
Whatever potency assay is used, the outcome will refer back to product quantification and thus the two issues “quantity and potency” are interrelated. “Sterility” testing follows established procedures for the detection of microbiological contamination and endotoxin levels, and can be enumerated according to validated tests and acceptable threshold levels.
However, when it comes down to “purity and identity”, the issues are less straightforward. The more stringent the definitions for EV-based therapeutics are laid down, the more emphasis is put on purity and identity of the preparation. EVs include a broad variety of membrane-bounded vesicles, while exosomes are restricted, at least, by size and surface markers. EVs either from tissue or from in vitro expanded cell cultures are part of the secretome, which includes soluble molecules like proteins and lipids, extracellular RNA species, and membrane vesicles. Electron microscopy (EM) data show that even highly-purified EV preparations for analytical purposes contain co-purifying components [58
]. Considering that large-scale clinical manufacturing of EVs will be less stringent on segregating EVs from co-purifying components, we must expect an increasing fraction of secretome components in the final preparation. In this light the question arises if the definition of EV or exosome is still appropriate, and also how the various secreted components may influence therapeutic activity. If the product is termed “exosomes”, regulatory authorities will require a demonstration of purity, and the percentage of exosomes present in the final product may be in question. In the best case, the product will relate more to an “exosome-containing” preparation.
Such considerations can be discussed at great length, may increase the uncertainty about the vesicle-based therapeutics, and may eventually discourage scientists, manufacturers and potential investors from engaging in clinical trials. But how can this conundrum be resolved? One approach may be to accept the heterogeneity of secretome-based preparations and to find a terminology that embraces all biological components and therapeutic aspects without eliminating the central claim. Secretomes [59
] are crude and uncharacterized mixtures of soluble and vesicular components, and yet secretome-based therapeutics are already under pre-clinical investigation [60
]. Can secretomes be characterized better than EVs? No. But they are more broadly defined, and as such allow considerable flexibility with regard to purity and identity. From a cell biological standpoint, and not limiting the definition to a purely proteomic view, the secretome can be seen as the totality of organic molecules and inorganic elements secreted by biological cells into the extracellular space, either in a soluble or packaged form.
The manufacturing and enrichment process of EVs eliminates a large portion of soluble secreted proteome components, but the co-purifying fraction is difficult to determine. Nevertheless, the manufacturing strategy aims at enriching (not necessarily purifying) membrane-bounded vesicular structures. It may thus seem more appropriate to consider the resulting product to be a vesicular secretome fraction. The term “fraction” has already been used to identify subclasses of products (e.g., albumin fractions based on the original Cohn procedure) [61
]. This terminology can also accommodate fractions containing co-purifying serum components in those cases where the manufacturing process requires the use of (vesicle-depleted) serum.
From the perspective of the manufacturer, the advantages of EV/VSF-based therapeutics over cellular counterparts relate to the possibility of filter sterilization of the final product immediately prior to aseptic filling; a considerable flexibility in the choice of storage buffers; and the reduced demands on the freezing and storage conditions, which substantially reduces the overall costs of the manufacturing process. The disadvantage, in comparison to stem cell therapy, of an extended manufacturing and processing period seems to be well compensated for by the above benefits.
3.2. On the Importance of Working with Highly Purified EV Populations
Heterogeneity among EVs not only exists with regard to vesicle size (separating apoptotic bodies, microvesicles and exosomes), but also among exosomal subgroups [62
]. It can be anticipated by extrapolation of the natural heterogeneity of the producing cells, and supported by detailed analyses that EV preparations comprise different vesicle populations [63
]. Functionality and therapeutic value of EVs or VSFs, however, can only be correlated with their composition once the entire range of secreted EV subpopulations is isolated for thorough analysis and comprehensively described. Thus, it remains to be established how heterogeneity affects the therapeutic effect of a putative biological EV- or VSF-based product [66
For various investigations, including the definition of a MoA and product release criteria, but also for toxicity, biodistribution or pharmacokinetic studies, highly purified and homogenous EV fractions seem necessary. The technical challenges associated with purifying EVs to homogeneity are evident. For purely analytical purposes, affinity chromatography solutions may be suitable (shiga or cholera toxin binding [63
], or heparin affinity binding [67
]) and provide satisfactory results. However, such approaches fall short of complying with current GMP requirements and the demand for large-scale enrichment and purification procedures of clinical doses of EVs or VSFs. It appears necessary to define an acceptable purity and identity of EV or VSF preparations with a view to available and manageable purification schemes, and based on serial filtration, tangential flow filtration (TFF) or polymer precipitation. Useful information on the current purification regimens for EVs has been provided, recently [66
]. In any case, the overarching perspective must be guided by functionality and clinical efficacy of the therapeutic product, which sets the issue of purity to a second-level of relevance.
3.4. How to Purify
At present, there is no consensus on the most suitable method for EV enrichment and purification [70
]. Major factors influencing the choice of methodology are the starting volume and the intended use of the final product. With the focus of this article on therapeutic application, the logics underlying the choice of enrichment or purification strategy differ from that of analytical and diagnostic use. EV and VSF enrichment from cell culture supernatants (plastic adherent cultures, static or dynamic bioreactors or hollow fiber perfusion reactors) requires an initial volume reduction from the scale of several liters to a few milliliters. For pharmaceutical manufacturing, raw materials or consumables must comply with GMP regulations. Ideally the process is scalable to accommodate future large-scale manufacturing to eventually reduce the costs of goods and to circumvent the risk of unexpected hurdles brought about by procedural changes in late stages of product development. Bearing these parameters in mind, the initial purification steps that follow the clearance of cell debris by low-speed centrifugation and 0.22-µm filtration should include high-throughput filtration steps such as TFF. Standardized GMP-compliant TFF systems are available on the market, offering the possibility of validated process control and documentation.
The TFF process can be used to reduce an initial volume of 10 L to about 200 mL in less than 2 h, depending on the harvest medium used and the molecular weight cutoff chosen. A 750 kDa pore (equaling a 13–15 nm pore size) will eliminate a large portion of soluble, not aggregated, proteins. A buffer change by diafiltration using the same column is advised at this stage. Further volume reduction can be achieved by either additional filtration on smaller sized columns with reduced dead volume, or by low-pressure track etch membrane filtration [71
A final step employing size exclusion chromatography (SEC) would at this point be suitable for removing protein aggregates and lipoprotein particles. This step, however, commonly causes a significant drop in the total particle number (by 30–70%), owing to the limited recovery, and comes at the expense of a volume increase by a factor of 1.5 [72
]. Although upscaling is possible for SEC, this method is currently not suitable for the initial volume reduction required for cell culture-derived EVs and VSFs.
A scalable anion exchange chromatography method yields functionally active EVs [68
]. A drawback of this technique, however, is the fact that most serum proteins also bind to the resin and co-elute at 500 mM NaCl with the vesicular fraction. The resulting solution thus equals a vesicle-enriched secretome fraction. Heparin affinity columns [67
] seem to more selectively associate with membrane components of EVs, but elution at 2 M NaCl and retrieval of functional EVs appears questionable, and the entire process is time consuming in its present form.
Dynamic culture conditions based on large-scale computer-controlled stirred suspension bioreactors have been applied in an attempt to develop future strategies to manage neurodegeneration involving the use of human MSC secretomes. Secretomes from dynamically cultured BM-MSCs induced a higher number of human neural progenitors to differentiate into neurons compared to MSC secretomes collected under static conditions, and increased the secretion of several neuroregulatory molecules and miRNAs. BM-MSC dynamic secretome further induced neurogenesis, as well as a robust increase in neuronal cell differentiation. These outcomes were associated either with the exclusive presence, or increased expression, of neuroregulatory molecules and miRNAs within the dynamic secretome [56
3.5. How is Identity and Purity Defined in EV Preparations?
Irrespective of the technology that is applied for the measurement of EVs, there are pertinent problems to this mode of characterization. Nanoparticle tracking analysis (NTA) and other optical flow-based approaches may quantify the particulate fraction in a solution to a satisfactory level, but are unable to discriminate between particulate and vesicular (membrane-bounded) events. Electron microscopy can solve that problem, but this method is not suitable for quantitative and high-throughput analysis. NTA and flow cytometry techniques combine fluorescent labeling primarily of membrane lipids or transmembrane protein components with enumerable detection of total events. Preliminary developments are promising, but at present do not satisfactorily address the need for a reliable, stable, reproducible and GMP-compliant technology.
A commonly accepted mode of concentration determination is that of total protein content. This approach can only be applied with confidence once the purification process and producer cell lines are standardized and validated. Determining the identity of MSCs is essentially based on the minimal criteria of the International Society for Cellular Therapy (ISCT) set forward in 2006 [73
]. The ISCT criteria are still the basic reference for the majority of clinical trials with MSCs and even these rather common surface marker profiles are sometimes diluted out.
If MSCs are measured by flow cytometry and their characteristics are evaluated according to the above ISCT criteria, the majority of MSC preparations reveal an almost 100% MSC identity, as do some fibroblasts (see Table 2 in [74
]). However, by adding only 1 or 2 other surface markers, this seemingly homogeneous population reveals a striking heterogeneity [75
As a consequence, the majority of MSC populations should be regarded as non-clonal and heterogeneous with unpredictable properties [77
] and an average composition of at least three different cell populations. If such cell pools secrete only two functionally different EV populations (e.g., exosomes and microvesicles), which may further comprise two or three different exosome fractions, the identity of one particular EV fraction in a seemingly homogeneous EV pool drops below 10%. These considerations once more emphasize the need to reconsider the applied terminology and to perhaps use the term VSF instead of EV. As long as identity and purity of a therapeutically active drug cannot be better defined, a common principle for early development of biologicals predicts that “the process is the product”
. We suggest that this consideration is also valid for EV- or VSF-based therapeutics.
3.6. Release Criteria
If indeed the process is the product, then the release criteria for MSC-EVs must encompass both the producing cells and the enrichment scheme. Minimal acceptance and release criteria may thus include the following:
MSCs display an ISCT-compliant surface marker profile at the time of secretome harvest [73
EVs within the secretome fraction must comply with the minimal criteria of ISEV, at least for a number of membrane markers [78
The size range should be in the range of exosomes (50–150 nm).
Sterility and endotoxin levels must comply with regulatory requirements.
For a stable and compliant process, protein concentration as a means to determine the concentration of given EV preparation may not be a suitable release criterion. Co-purifying proteins from the secretome or serum components prevent a precise analysis of the EV-associated protein moiety. Moreover, if an anticipated therapeutic activity and EV quantification are based on, for example, MHC class II molecules [3
], the question may arise as to why the preparation is not defined as an MHC II product containing lipids and RNA, instead of a vesicle-containing product. While the protein profiles of MSCs and their EVs or VSFs remain astonishingly constant despite largely divergent culture conditions and enrichment procedures, the miRNA content changes within a few h when culture conditions are altered. Thus, a basic profile for miRNA should be established for each method of cell cultivation to accommodate all preferences (serum type, medium type, oxygen levels, etc.). The reproducibility of miRNA profiles may be best suited to report on the persistence of a stable manufacturing procedure (irrespective of a complete lack of prediction for a MoA by the miRNAs).
3.7. Naive vs Loaded EVs
An alternative approach to using unmanipulated MSC-EVs is that of loading vesicles with a known and characterized active component (proteins or nucleic acid-based compounds). However, similar regulatory requirements with regard to robustness, reproducibility and scalability of the loading and manufacturing process also apply for loaded EVs [79
] and the loading methods. A series of siRNA loading technologies have been tested including transfection, electroporation (can lead to RNA aggregation), co-incubation of cholesterol-conjugated siRNA [79
], and liposome fusion with hybrid biocompatible carriers (Anjarium’s proprietary Hybridosome™ technology) that comprise structural and bioactive elements and a tunable fusogenic moiety. DNA loading via electroporation seems to be the current standard, resulting in the presence of several hundred linear DNA molecules (less than 1000 bp) per EV [15
]. Vesicle loading with chemotherapeutics by electroporation, incubation and mild sonication has gained increasing popularity, and encouraging results for Paclitaxel loading demonstrate that exoPTX performs better than Taxol, as judged by the antineoplastic efficacy in a Lewis Lung Carcinoma mouse model [80
Irrespective of these seemingly direct therapeutic approaches, ex vivo loaded EVs also have to be prepared from source cells and cannot (at present) be synthesized from artificial components. The most popular cell lines currently employed are RAW 264.7 macrophages, various dendritic cells, glioblastoma and lung carcinoma lines, and HEK 293 cells. It appears necessary to point out that all co-purifying components in enriched vesicle fractions from these cell sources may contribute to the overall therapeutic effect in addition to that of the loaded substance.
It has to be considered that tumor cells are able to transform not only neighboring cells by transferring proteins and nucleic acids, but also the extracellular matrix by releasing several metalloproteases, potentially via EVs. Therefore, the use of tumor cells in the manufacturing of EVs or VSFs for therapy may convey pro-tumorigenic effects that have to be investigated with sufficient caution prior to clinical evaluation. Furthermore, any side effects and immunological reactions may be also caused by the vesicular drug delivery systems and effects of such combined therapeutic products must be well characterized prior to application for market authorization.
3.8. Aseptic Filling, Storage and Stability of the Final Product
Aseptic filling in non-automated manufacturing sites requires an A-in-B cleanroom environment (class 100 in class 1000 by U.S. standards). Potential manufacturers should bear this in mind during the planning phase. The biophysical properties of EVs and VSFs require low adsorption materials to be used as storage containers. Depending on the formulation of the final product, such packaging material must be validated, and the suitability confirmed. Currently, some low-protein binding synthetic materials are available from several suppliers, but recommendations supported by conclusive data from manufacturers of EV therapeutics are still missing.
If EVs are stored as highly-concentrated, ready-to-use liquid formulations, the storage temperature has to be determined in order to preserve the monodisperse suspension and to avoid aggregation or degradation over time. While in most cases a temperature between −40 and −80 °C seems appropriate, the lower temperature range raises concerns regarding the ease of shipment and distribution of the product, as well as available storage options at the points of care. Thus, stability testing programs need to be executed that address not only the long-term storage conditions but also the potential influence of changing temperatures during storage and distribution until the administration of the product to patients. For all the above issues we recommend a thorough and well-documented risk-based analysis to comply with practical issues of clinical application and regulatory requirements.
3.9. Biodistribution, Bioavailability, Cytotoxicity and Pharmacokinetics
Administration of EVs in patients has already been tested in a small number of phase I clinical trials, revealing an overall low toxicity and considerable stability of EVs in the circulation [4
]. Currently, the original research information on the biodistribution of EVs is extraordinarily scarce [83
]. The use of clinically approved radioisotopes for labeling EVs seems a fruitful approach. Supramagnetic particle loading is an alternative, but both techniques bear considerable limitations. A significant contribution to this matter was shown recently [86
]. Based on this work and that of others [83
], it appears that EVs are rapidly sequestered (within minutes) by circulating macrophages. Thus, systemic (intravenous) application of EVs may not be the best route for achieving significant therapeutic effects. More solid data from both pre-clinical and early clinical phase I studies are urgently required to inform the community about this important aspect, and to improve future therapeutic applications.
3.10. Predictive Potency Assays, Mode of Action (MoA) and Proof-of-Concept (PoC)
A basic biological readout to evaluate the in vitro and in vivo therapeutic potency of EVs and VSFs seems mandatory. However, there is currently no consensus about the biological relevance of EVs. If we focus only on one possible active EV-component, such as RNAs, and on how one may effectively and reliably characterize them in an EV preparation, the current knowledge prevents the implementation of universally acceptable standards for EV-RNA analysis as part of the product release criteria [88
], and this seems to be true for the entire EV field. Combinatorial potency assays should be developed to account for the multifactorial regenerative instructive potential that can be exerted by MSC-EVs. As for the few model target diseases that we discuss in this article, it is understood that predictive potency assays for bone regeneration should not only depict the investigational drug’s capacity for inducing osteoblast proliferation, but should also take into account a potentially required osteoclast activity and/or angiogenesis. Potency assays for skin repair in EB have to go beyond a 2D keratinocyte migration assay, and in the case of acute SCI models potency testing for neuronal outgrowth may have to be supplemented with assays depicting the cellular re-myelination propensity, as well as an overall immunomodulatory and neuroprotective capacity.
One more essential question that needs to be addressed is the proposed MoA of EVs or VSFs in any of the intended applications. At least a rudimentary knowledge of a potential MoA is obligatory for designing a suitable test environment for the examination of a novel drug substance in relevant animal models, and to provide meaningful PoC studies [68
However, even if no definitive MoA can be described and the active component of a given novel biological product that is responsible for an observed therapeutic activity remains ill-described, it may be possible to engage in early clinical trials for indications of a definitive unmet medical need. Poorly characterized biological therapeutics in early developmental stages can be clinically tested after thorough safety evaluation and the reproducible and convincing demonstration of a therapeutic effect in vitro and in relevant animal models.
A prerequisite for this strategy, however, is that the investigative product is of reproducible quality and biological effect, irrespective of the precise content and function of the individual factors. Once again, we may have to look at the manufacturing process as the therapeutic product.
Prockop and colleagues recently provided a striking correlation between MSC treatment and the (dose-dependent) presence of the anti-inflammatory protein TSG-6 [76
]. The group confirmed a proposed anti-inflammatory MoA via TSG-6 that is responsible for the positive therapeutic effect of MSCs after myocardial infarction and chemical cornea injury in mouse models, but this criterion is not yet broadly applied when therapeutically active MSCs are characterized.
Investigating the potency of MSC-derived EVs in post ischemic stroke, a PoC for the potential neuroprotective and neuroregenerative capacity of EVs was provided [53
]. EVs successfully recapitulated the effect of MSCs, and no adverse reactions were reported. In a compassionate clinical use experiment to treat GvHD, a similar level of safety and efficacy for MSC-derived EVs was demonstrated. This first-in-man approach suggests a promising future for phase I and II clinical studies to test the systemic application of MSC-EVs [90