Viral vectors could be administered both locally, and systemically and they include lentiviruses, adenoviruses, and adeno-associated viruses (Chen et al., 2015). Viral delivery of antisense construct expression vectors was used to inhibit miR-21 and miR-148a in animal models of pancreatic cancer (Bao et al., 2014; Bofill-De Ros et al., 2015), while expression vector for miR-1266/1207-5p was examined in replacement therapy in gastric carcinoma (Chen et al., 2014). Viruses are able to effectively deliver miRNA therapeutics (precursors, mimics, genes, or inhibitors) into the tumor cell, but their use could be associated with the risk of insertional mutagenesis, gain of replication competency of viral particles, or immune activation. Nucleic acid of adenoviruses (dsDNA viruses) and adeno-associated viruses (ssDNA viruses) usually do not integrate into the host cell genome, while lentiviral (ssRNA viruses) integrates (Soriano et al., 2013; Chen et al., 2015). Adeno-associated viruses are generally less immunogenic, but adenovirus-based delivery system could produce at least transient hepatotoxicity (Broderick and Zamore, 2011; Aslam et al., 2012) as was also observed by Bofill-De Ros et al. in animal model of pancreatic ductal adenocarcinoma (Bofill-De Ros et al., 2015).

In the selected studies, the most frequently used synthetic polymer was polyethylenimine (PEI). It was utilized to deliver mimics or expression vectors of miR-34a, miR-206, and miR-217 in animal model of pancreatic cancer, or miR-33a and miR-145 in the model of colorectal carcinoma (Tables 4, 5). PEI is cationic polymer able to produce nanoparticles. It has linear or branched structure and different molecular weight according to the reaction conditions during the synthesis. Due to the positive charge, PEI has high capacity for negatively charged oligonucleotides and nucleic acids which are moreover condensed after complexation with PEI, and thus protected from nucleases. The charge of PEI also facilitate cellular uptake by electrostatic interaction with negatively charged surface molecules (e.g., heparin sulfate proteoglycans), after which the particles enter the cell by endocytosis. PEI is able to disrupt the endosome and release the cargo into the cytoplasm, which grants this method high transfection efficacy. The disruption of endosome is achieved by protonization of PEI and buffering of acidic environment of the vesicle. These processes are followed by osmolarity changes and water intake which leads to the swelling and burst of the endosome (Höbel and Aigner, 2013; Zhang et al., 2013). Better capacity and efficacy is achieved by branched PEI but at the cost of higher non-specific cytotoxicity. In the presented studies, mostly linear PEI is used, like commercially available transfection reagents ExGen500™ (Euromedex, Mundolsheim, France) and in vivo-jetPEI™ (Polyplus Transfection, Illkirch, France) assigned for in vivo experiments. Other issues associated with PEI delivery are an aggregation of created nanoparticles, or opsonization in the plasma recognized by phagocytes. PEI with high density of positive charge could also trigger erythrocyte aggregation and thrombosis (Kanasty et al., 2012). PEI particles could be conjugated with various molecules [e.g., polyethylene glycol (PEG), or antibodies] to resolve such difficulties (Malek et al., 2009). PEI is not a biodegradable polymer, thus its toxicity is intensively discussed. It depends strongly on molecular weight and branching (Fischer et al., 1999) and also on the cargo, as it may neutralize the charge of PEI. There are studies describing immune activation in vivo, (Beyerle et al., 2011) hepatotoxicity and lethality in mice (Chollet et al., 2002) and increased apoptosis in vitro (Merkel et al., 2011), and also those that proved no immune response, or hepatotoxicity in mice (Bonnet et al., 2008).

Sun et al. used iron oxide nanoparticles to deliver miR-16 and overcome doxorubicin resistance in animal model of gastric adenocarcinoma (Sun Z. et al., 2014). Iron oxide nanoparticles (IONPs) are biocompatible and biodegradable particles with magnetic properties. They are composed of magnetite [Fe₃O₄, iron (II,III) oxide], or maghemite (Fe₂O₃, ferric oxide) and are usually coated with various other molecules (PEI, PEG, chitosan etc.) to improve their properties (Kievit and Zhang, 2011). IONPs could also serve as theranostics (i.e., substances with both diagnostic and therapeutic purpose). As well as other nanoparticles, IONPs protect nucleic acids from being cleaved by nucleases (Kievit et al., 2009) but could also be opsonized in the plasma and recognized by phagocytes, mainly by the reticuloendothelial system (RES). Non-coated IONPs are distributed in heart, liver, spleen, lungs, kidney, brain, stomach, small intestine, and bone marrow, while the highest concentration are reached in the liver and spleen due to the elimination by RES and macrophages (Wang et al., 2010). IONPs enter the cell by endocytosis and are degraded in the endosomes (Xie et al., 2009). Particles between 10 and 60 nm are the most effective, as they undergo limited kidney and liver/RES uptake, and are absorbed by tumor cells (Kievit and Zhang, 2011). Cytotoxicity of IONPs coated with PEI occurs in vitro in higher concentration than is needed for sufficient transfection (Lellouche et al., 2015). The administration in vivo could increase blood iron and intracellularly increase oxidative stress (Mahmoudi et al., 2011). Non-coated particles could produce hepatotoxicity, and lung or kidney damage (Hanini et al., 2011).

Study of Hiraki et al. describes utilization of different inorganic nanomaterial, carbonate apatite nanoparticles. They used these particles as a delivery system for mature miR-4689 in animal model of colorectal adenocarcinoma (Hiraki et al., 2015). Carbonate apatite [Ca₁₀(PO₄)_6−X(CO₃)_X(OH)₂] is composed of calcium cations and phosphate and carbonate anions in defined ratios. It was firstly described as a transfection reagent and a delivery system for plasmid DNA by Chowdhury et al. (2006). Nanoparticles of carbonate apatite are stable in plasma (pH = 7.4), protect nucleic acids from nuclease cleavage, but in acidic environment of endosomes, they are quickly degraded. Their cargo is then released and could probably escape from endosomes, as high effectivity of this transfection method was proved for DNA (Wu et al., 2015) and RNA (Hossain et al., 2010). In mice, these nanoparticles are accumulated in tumor probably due to the EPR effect (discussed below), but slight accumulation was found also in the liver (Wu et al., 2015). As this method arose from calcium phosphate co-precipitation, which is known to produce certain level of cytotoxicity in vitro, adverse effects in animals were inquired. In mice, Wu et al. declared no mortality, weight loss, or histological damage in liver, kidney, and spleen after administration of common dose, and also after 2.5 and 5-fold higher doses. They also do not observed any urinary calculi in a mouse model of repeated administration. The team advanced to the evaluation of the delivery system on monkeys (macaques Macaca fascicularis, formerly M. cynomolgus). Monkeys received repeated i.v. infusions during a movement restraint. Equivocal results were obtained, as some animals had reversible increase in AST, ALT, LDH, and CPK enzymes, but from further analyses of isoenzymes, authors suggested that these increments might arise from the stress associated with body restriction rather from heart or liver damage (Wu et al., 2015).

For direct intratumoral treatment, atelocollagen was used in the study of Frampton et al. in pancreatic ductal adenocarcinoma to deliver miR-21, miR-23a, and miR-27a (Frampton et al., 2011). Atelocollagen is a biocompatible and biodegradable polymer. It was developed and tested in vivo for gene (plasmid) delivery with controlled release by Ochiya et al. (Ochiya et al., 1999; Hao et al., 2016). Atelocollagen is prepared from collagen extracted from bovine dermis. Natural collagen contains specific amino acid sequences on both C- and N-terminus (“telopeptides”), which are highly immunogenic. By digestion with pepsin, these telopeptides are cleaved. The polymer is liquid at low temperatures, but solidifies at temperatures above 30°C (Ochiya et al., 2001; Komatsu et al., 2016). After intramuscular injection of plasmid DNA in complex with glucose and atelocollagen in mice, the transfection was efficient and last more than 60 days. No apparent toxicity, or hematologic changes were observed in this study, (Ochiya et al., 1999) as well as Frampton et al. described neither changes in mice body weight, nor serious adverse effects (Frampton et al., 2011).

For intratumoral administration, several studies used lipid-based transfection reagent Lipofectamine 2000 (Thermo Fisher Scientific, Waltham, USA) designed originally for in vitro experiments, e.g., Dong et al. in animal model of colorectal adenocarcinoma to deliver miR-133a (Dong et al., 2013). Pramanik et al. utilized DOTAP (N-[1-(2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methyl-sulfate) with co-lipids formula to deliver plasmid expression vector of miR-34a, and miR-143/145 cluster systemically in animal model of pancreatic ductal adenocarcinoma (Pramanik et al., 2011). Both delivery systems are composed of cationic lipids that form liposomes, vesicles with lipophilic bilayer and aqueous core able to encapsulate hydrophilic molecules (Mallick and Choi, 2014). By electrostatic interaction, cationic lipids have increased capacity for negatively charged nucleic acids (Xue et al., 2015). They enter the cells by endocytosis and are able to destabilize and breach endosomal membrane by interaction with its phospholipids (Zelphati and Szoka, 1996). Cationic lipids exert detergent effect on lipid membranes and interact also with enzymes, thus could irritate cells, decrease proliferation, alter gene expression, and even trigger cell lysis (Wu et al., 2001). They share the same advantages and disadvantages that account for positive charge as cationic polymers. They form aggregates with plasma proteins leading to RES elimination and accumulation in spleen and liver (Nchinda et al., 2002; Zhang et al., 2012). With decrease of positive charge, RNA encapsulation and transfection efficacy is decreasing. PEGylation increases blood circulation time of liposomes, (Pathak et al., 2011; Suk et al., 2016) but could lead to the formation of anti-PEG IgM antibodies (Ishida et al., 2006). Liposomes are generally less immunogenic than cationic polymers. But after processing of liposomes, some RNA molecules might remain on the surface of a particle. In the studies using siRNA, these residues lead to the significant immune activation (Xue et al., 2015). Inflammatory response in the liver followed by hepatotoxicity and with higher doses even lethality was described (Tan and Huang, 2002; Zhang et al., 2005).

Pharmacokinetics of therapeutically administered oligonucleotides is strongly driven by their physico-chemical properties. Generally, these properties are not sequence-specific in qualitative point of view, but can be quantitatively different from sequence to sequence, and could differ also between chemistries. Native oligonucleotides are small, negatively charged molecules, which means that the transfer through lipophilic membranes necessary for the absorption into systemic blood circulation and also later into the intracellular space is quite problematic.

The most common change in oligonucleotide chemistry is the replacement of phosphodiester bond with phosphorothioate bond in the backbone. This change goes usually hand in hand with chemical modification of the 2′ functional group on ribose in the nucleotide (2′-hydroxyl could be substituted e.g., to 2′-O-methyl, 2′-O-methoxyethyl, or 2′-fluoro group), and conjugation with cholesterol. These modifications either increase the stability of oligonucleotides modifying their susceptibility to RNAse cleavage (phosphorothioate bonds, 2′-O-modifications), or increase cellular uptake of the molecule (cholesterol conjugation). Cholesterol-conjugated 2′-O-methyl/methoxyethyl-modified oligonucleotides are sometimes termed “agomiRs” or “antagomiRs” depending on their mechanism of action, and they were utilized in some of the studies focused on gastrointestinal cancer presented in this article (Chang et al., 2013; Wang et al., 2013; Zhang et al., 2015; Zou et al., 2015). Modification on 2′ position could also change the affinity of oligonucleotides to plasma proteins which has a high impact on pharmacokinetics, most importantly on distribution and excretion (Crooke, 2007).

Another possibility to change oligonucleotide structure is chemical modification of the ribose forming a 2′,4′-bicyclic structure, which is termed locked nucleic acid (LNA) (Kumar et al., 1998). The most common type of LNA is oligonucleotide with one or more 2′-O-4′-methylene-β-D-ribosyl structure. This bicyclic bridge locks ribose in one of its conformation increasing binding affinity and decrease the susceptibility to nuclease cleavage (Braasch and Corey, 2001).

Various chemical modifications in the oligonucleotide structure are now available owing to the development of commercially available miRNA mimics. According to the information provided by manufacturers, miRNA mimics should possess higher affinity to miRISC and thus to the mRNA of interest. MiRNA-mimics should have no off-target biological activities due to the passenger strand, and should exert higher effect than native mature miRNAs. The chemistry modifications differ between passenger and guide strand, and the molecules could also be triple-stranded (e.g., Exiqon, Vedbaek, Denmark). Detailed information about the specific chemistry of the miRNA mimic are usually not released. Some evidences were published last year, that bring the commercially available miRNA mimics into focus because of non-specific dampening effect on overall gene expression, accumulation of non-endogenous high molecular weight RNA species and unintentional passenger strand loading into the RISC discovered after transient transfection of human cell lines. Søkilde et al. describe variations even between batches of a commercially available miRNA mimic obtained from one manufacturer (Søkilde et al., 2015). The authors emphasize the issue of a proper dosage of miRNA mimic and its optimization, and suggest to prefer viral and genetic approaches, as the created transcripts follow the physiological biosynthesis pathway and their mechanism of action could be considered as the very same as endogenous miRNAs (Jin et al., 2015).

Undesirable physico-chemical properties of oligonucleotides could be attenuated by delivery systems mentioned before, which subsequently influence pharmacokinetic processes of miRNA-based therapeutics.

After being absorbed or injected into systemic circulation, charged molecules of oligonucleotides bind to various plasma proteins, above all on albumin and α₂-macroglobulin (Cheng et al., 2013). The binding and the distribution is non-linear, saturable, changes slightly with length and sequence of oligonucleotides and is different in rodents and in human. Distribution to the tissues is very quick and prevails over metabolic degradation (Levin, 1999). Naked oligonucleotides accumulate in the liver, kidneys, spleen, bone marrow and lymphatic nodes, while they do not cross the blood-brain barrier, placental barrier and they are not present in testes.

In the treatment of cancer, accumulation of a drug in the tumor tissue or in the metastasis site is a desirable state. MiRNA-based therapeutics could achieve this due to enhanced permeability and retention (EPR) effect of a tumor. Enhanced permeability of new vessels and relative lack of lymphatic vessels in the tumor site was firstly described by Matsumura and Maeda (1986). Charge-neutral small particles complexed or loaded with miRNA-based therapeutics have enhanced extravasation and could accumulate in the tumor. Metastatic sites are generally less accessible, as their EPR effect is not so significant (Maeda, 2015).

According to the technology of the delivery system used, miRNA-based therapeutic could accumulate also extratumorally in various tissues. All cells capable of phagocytosis accumulate naked oligonucleotides, liposomes, or nanoparticles, e.g., RES cells present in the liver (Kupfer cells) and in the circulation, tissue monocytes and macrophages, and proximal tubular cells (Chen et al., 2015). In this case, the delivery system alone as a protection could be insufficient, because in plasma, these particles get coated by proteins recognized by the RES. The most common defense against RES is PEGylation, binding of polyethylene glycol substituents on the surface of a nanoparticle or liposome, which prevent binding of opsonization proteins and became very common. Contrarily, the excess of PEG on the surface of a delivery system particle could diminish cellular uptake, therefore the process of PEGylation should be optimized (Seto, 2010).

Oligonucleotides, liposomes and polymer-based nanocarriers enter the cell by active mechanism, endocytosis. Escape from endosomes is desired to reach the interaction of miRNA with mRNA, however, this is another obstacle in miRNA-based therapy. Some of the carriers could enhance endosomal escape by steric or osmotic effects. pH sensitive molecules could change structure in relatively acidic environment due to electrostatic interactions, which is leading to the mechanical disruption of the vesicle and release of miRNA into cytoplasm (Ju et al., 2014). Other molecules are accepting H⁺ (proton sponges) and by alteration of ion homeostasis cause swelling and burst of the endosome (Akinc et al., 2005; Chen et al., 2015).

Ubiquitous nucleases begin to degrade oligonucleotides shortly after administration. According to the chemistry changes, free oligonucleotides are metabolized by 3′- and 5′-exonucleases or by endonucleases, and the rate of metabolism depends on the chemical modifications. Endonuclease cleavage is slower and takes place only when 3′ and 5′ end of oligonucleotide is protected by methoxyethyl-modified nucleotides. As was mentioned before, modifications on 2′-hydroxyl on ribose or structural changes in the backbone such as LNA structure can decrease the affinity of nucleases to cleave miRNA-based therapeutics. Also the complexes of oligonucleotides with nanoparticles or liposomes have modified susceptibility to nuclease cleavage.

The metabolites of nuclease cleavage are weakly bound to the plasma proteins and therefore are rapidly excreted in urine. Oligonucleotides do not undergo liver oxidation by cytochrome P450, or conjugation processes (Levin, 1999; Crooke, 2007).

Oligonucleotides not bound to proteins are excreted in the urine, while binding to plasma proteins, or other delivery systems like liposomes and nanoparticles of specific parameters (e.g., hydrodynamic diameter up to 5–6 nm) results in protection from being urinary excreted (Crooke, 2007; Cheng et al., 2013). As oligonucleotides accumulate also in the liver, they could be excreted by both these organs. About 10% of the administered dose of naked oligonucleotides, and 80% of the metabolites are urinary excreted. The remaining are excreted by faeces, or endure bound to the tissue, or inside the cells. The elimination half-life of oligonucleotides is 1–30 days depending on the type of tissue. This attribute allows designing a therapeutic regimen comfortable for potential patients with one dose administration for a week, 2 weeks, or a month (Crooke, 2007).