Literature Search
As reported in Figure 1, the search strategy using the eight research strings identified 6,860 scientific manuscripts on PubMed® (MEDLINE®), 4,294 on EMBASE, 2,631 on BIOSIS™, and 5,120 on Scopus. The eight lists were compared within each search engine deleting all duplicates hence the final number of manuscripts was 875 for PubMed® (MEDLINE®), 1,441 for EMBASE, 844 for BIOSIS™ and 1,447 for Scopus. At this point the four lists were merged together, erasing again duplicates, with a definitive list of 2,073 records. Such list was individually screened by each author, relying on title and abstract, and excluding all non-English papers, congress abstracts and posters, letter to editors, clear reference to cancers other than BC or to other pathologies at all, and reviews. Reviews were however searched for other relevant references, but no other eligible papers were detected. In this phase 852 records were excluded, and 1,221 papers were assessed for full text eligibility, excluding all those in which there was no use of mouse models of BC, there was not in vivo imaging or ex vivo on whole organs, i.e., were excluded imaging techniques applied to histological samples. At this point, 170 papers were studied to prepare the quali-quantitative synthesis, further excluding all manuscript in which the animal models were used to study long non-coding, small interfering or other RNAs, as well as genes or other signaling molecules, without a clear link to a miRNA. The final data extraction was made on a total of 114 manuscripts.
Mouse models of breast cancer
Various factors play a role in the study of preclinical models, in particular the mouse strain, the cell line and the engraftment route. A summary of the murine strain used in the articles analyzed and the relative references are shown in Table 1, while Figure 2 shows the absolute number of experiments for each strain. The murine genetic background most used in miRNA studies have been resulted to be the Balb/C (46,7%), followed by different strains of athymic and/or nude mice (23,3%) and non-obese diabetic, severely immunocompromised strains (NOD/SCID) (14,2%) and SCID strains (9,2%). Only few experiments were performed on NOD/SCID gamma strains (NSG) (3,3%) and only one on NOD mice (0,8%). In two papers it was not possible to identify the murine strain used (1,7%) (22, 23). To be noticed, only one was a transgenic model, the vascular-endothelial growth factor receptor 2 (VEGFR2) -luc mouse. This model was generated, in the studied report, from an FVN/B strain and it harbors the luciferase gene downstream the VEGFR2 promoter region. In brief, anytime the VEGFR2 is transcriptionally activated, luciferase is transcribed as well, hence this model allows the direct, non-invasive and quantitative monitoring of VEGFR2 via bioluminescence imaging (BLI) (24).
Regarding cell lines, in most of the experiments human derived cell lines were used, and only few used syngeneic, i.e., mouse derived cell lines of BC. Figure 3 shows the absolute number of experiments for each cell line, and Table 2 shows the different specific modification to each cell line and the relative references. In details, MDA-MB-231 were used in most experiments (62,5%), followed by MCF- (16,7%). SKBR3 and SUM-derived cells were used in two experiments each (1,4% each), whereas R2N1d – labeled with green fluorescent protein (GFP) and transfected with miR–, BT549– transfected with miR– and T47D-TR (tamoxifen resistant) cell lines were used in one experiment each (0,7% each). Two experiments (1,4%) used breast cancer stem cells (BrCSCs): in one experiment, the BrCSCs were obtained after induction of differentiation of BC cells purified from fresh tissues from patients’ mastectomies and then transduced with GFP via lentivirus infection, prior to be used in an orthotopic model (25), in the second experiment BrCSCs were obtained from both patients tissues and from MDA-MB-231 and MCF-7 cell lines (26). One experiment (0,7%) used patients derived xenograft (PDX) labeled with luciferin and modulated for miR precursor expression (27) . The only syngeneic cell line used was 4T1 (13,8%).
Regarding the murine model, the three models for cellular engrafting, i.e., subcutaneous and orthotopic xenografts as well as metastatic model obtained by intravenous injection (i.v.) of cancer cells, were all represented (Table 3). The most used model was the i.v. metastatic model (46,8%), which was obtained either via intravenous injection (68,9%) – in one paper it was indicated as intraarterial (28) – in the tail vein, or in the left ventricle (9,8%). Direct intratibial injection to study osseous metastasis was applied in few experiments (6,6%) as well as the direct intrapulmonary injection (1,6%). Finally, the development of spontaneous metastasis after orthotopic injection was obtained either after surgical resection of the primary nodule (4,9%) - with lymph node (29) or pulmonary metastasization (30, 31) - or with the primary orthotopic implant on site (8,2%). In two papers, it was not specified how the metastatic model was obtained (32, 33).
The orthotopic model, with injection in the second or forth mammary gland or fat pad, trans-cutaneously, after surgical exposure, or intra nipple, was the second most used model (29,2%). Subcutaneous xenograft, implanted in various sites, i.e., on the shoulder, the armpit, the flank and thigh, was used in 24% of the experiments.
When interpreting tables and results, it should be noted that various reports performed multiple experiments using different cell lines and/or multiple models and/or multiple mouse strains, for all of which imaging was applied (27, 30, 31, 34-55).
Mode and route of therapy administration
In this paragraph we systematically reported the different in vivo modality and route of miRNA therapy administration. In Figure 4 are showed the absolute number of experiments done for each delivery system, while in Table 4 are reported the miRNAs used, the specific formulation of vehicle system and the relative references.
The most of pre-clinical mice models (54,4%) were generated by injecting luciferase (Luc)- labelled BC cells transfected with DNA or lentiviral plasmids. In detail, a lentiviral vector was used to modulate the expression of miR: -206 (27), -1 (32), -124 (39), -211-5p (42), -494 (44), -1204 (46), -133b (47) (37), -101 (53), -630 (56), -150 (57), -133a-3p (58), -452 (59), -543(60), -96 (61), -29a (62), -455-3p (63), -30a (64), -100(25), -548j(45), -940(65), -429(43), -442a(66), -373(51), -509(67), -190(68), -125b(33, 69), -125a-5p(70, 71), -33a(49), -33b(30), -138(72), -27b(73), -454-3p(52), -23a(74), -218-5p(75), and of miR-30 family members (miR-30a-b-c-d-e)(28). A lentiviral vector was also generated to express a circular inhibitor miRNA (CimiRs) specific to silence the expression of miR-223 and miR-21 (76).
Moreover, BC cell line were transfected with DNA constructs encoding for the following miRNAs precursor and/or inhibitors: let-7a-5p(77), miR-196a(78), -205(79), -361-5p(50), -590-3p(80), -567(81), -106b-5p(82), -497(24), -135/-203(55), -29/-30(83), 14q32-encoded miRNAs(84), miR-191/425 cluster(34). The effect of miR-1 overexpression was studied both in mice injected with MDA-MB-231 -Luc cells stably transfected with miR-1 precursor and in tumor-bearing mice treated with the synthetic miR-1 mimic(85).
In 8 studies BC cells were transfected with different type of plasmid (lentiviral or DNA) encoding mimic and/or inhibitor specific for miR-200 family members (miR-200a, -200b, -200c, -141, -429)(86) (35) (87) (88) (89) (48) (40, 41).
In 1 experiment a doxycycline inducible vector was used to over-express miR-301a-3p(90).
Nanoparticles (NPs)-based delivery represent a promising strategy for BC treatment preventing miRNA degradation in bloodstream and improving the miRNA delivery in tissue-specific targeting. Indeed, we found that 29 experiments (25,4%) were conducted using different formulation of NPs including natural based lipid NPs (LNPs) and synthetic NPs composed by inorganic materials such as silica (SiO2), gold (Au) or polymer (e,i. polyamidoamine -PAMAM- dendrimers)(91) (Table 4).
Organic LNPs were generated to encapsulate: miR-203 mimic(92), AgomiR-143(93), AgomiR-186-3p(94) and the “edited” form of miR-379-5p(95). AntagomiR-214-3p was loaded into the osteoclast-targeting delivery system (D-Asp8-liposome)(96).
Inorganic synthetic NPs was engineered to encapsulate miR-145 using PAMAM dendrimers modified with a thioaptamer (TA), a protein that binds CD44-receptors highly expressed on BC cells(97). Poly(ethylene glycol)–polyethylenimine (mPEG–PEI) was complexed with Molecular Beacon (MB) to detect miR-34a in BC(23).
Gold nanoparticles (AuNPs) were used to deliver miR-708(31) and miR-96/-182(98) mimics; other AuNPs were formulated with a photoacoustic (PA) nanoprobe that released a PA signal in the presence of the oncogenic miR-155(99). Magnetic (MN) NPs were engineered to the recognition of specific oncomiR in BC tissue(100, 101) or conjugated with locked nucleic acid (LNA) to inhibit the activity of miR-10b(29, 102). SuperparaMN iron oxide NPs (SPIONs) conjugated with Argonaute-2 protein (AGO2) were formulated to deliver miR-376B mimic in BC tissue(103).
In 5 independent studies, the activity of the tumor suppressor miR-34a was replenished using: i) hTERT promoter-driven VISA liposomal NPs(26); ii) polymeric hybrid nanomicelles simultaneously delivering Doxorubicin (Dox)(104); iii) Dextrin-PEI-CM nanoplex (DPC) delivering also cyclam monomer (a CXCR antagonist)(105); iv) silica dioxide NPs (SiO2NPs)(106), v) lipid core-shell nanocarrier coated with cationic albumin co-delivering docetaxel(107).
In 5 studies, miR-21 inhibition was obtained in vivo using: i) a core of phi29 pRNA- three-way junction motif (3WJ) harboring the RNA aptamer for EGFR (3WJ/EGFRapt/anti-miR21)(108), ii) a core of 3WJ harboring the aptamer binding to CD133 receptor (3WJ/CD133apt/anti-miR21)(109), iii) a polydopamine (PDA)-based NPs(110) , iv) tumor-extracellular vesicles complexed with gold-iron oxide NPs (TEV-GIONs)(111), v) RNA nanospheres into nanopompons(112).
In few studies, multiple miRNAs were simultaneously co-delivered using polymeric NPs triggered in BC tissue by the urokinase plasminogen activator peptide (uPA)(54), by ultrasound(113) or by RNA-triple-helix hydrogel scaffolds(114).
The combined delivery of miRNA and a chemotherapeutic drug into tumor sites was obtained using polymeric hybrid NPs (Dox + miR-34a)(104), polydopamine (PDA)-based NPs (Dox + antisense-miR-21)(110), magnetic NPs (Dox + miR-10b)(29), calcium/phosphate lipid NPs (Paclitaxel + miR-124)(115) and lipid nanocarrier coated by cationic albumin (Docetaxel + miRNA-34a)(107). Interestingly, specific NPs were developed to co-deliver photosensitizer indocyanine green (ICG) and the inhibitor of miR-21(116).
In 12 studied (10,5%) we found that to enhance the systemic delivery efficacy of mimic/inhibitor miRNA, in absence of a protective vehicle, synthetic small molecules or chemical modifications are added to miRNA increasing their stability in blood system. “CMM489” is a chemically modified mimic in which Uracil in the guide strand of miR-489 tumor suppressor was reply with 5-fluorouracil (5-FU)(117). Single-strand miRNA inhibitor (“AntagomiR”) and double-stranded mimic (“AgomiR”) are RNA harboring bases chemically modified to overcome the RNA instability. In this context, mice were treated with AgomiR-338-3p(118) or with AntagomiR-16-1-3p(119) or with AntagomiR-100(120). Additionally, FolamiR-34a is a modified mimic in which a folate group was attached to miR-34a sequence to directly bind the BC cells over-expressing the folate receptor(121). Another example of artificially synthesized nucleic acid is represented by the peptide nucleic acid (PNA) labelled with [99mTc] that recognize in vivo the presence of the oncomiR-155(122). Finally, the inhibition in the activity of miR-21(123-125), miR-210 (“Targapremir-210”)(126), miR-544(127) and miR-10b (“Linifanib”)(128) was obtained using small molecules compounds.
Exosomes are small extracellular vescicles (EVs) of 30–150 nm in diameter, which are released by cancer cells in tumor microenvironment to intercellular communication. In 6 studies (5,3%), researchers have exploited the possibility to use exosomes to encapsulate the following miRNAs: let-7(129), miR-210(130), -335(131), -159(132), -4443(38) and Anti-miR-21(111).
Recently, the anticancer activity of miRNAs derived from marine invertebrate marsupenaeus japonicus shrimp was analyzed in 2 experiments in which tumor bearing mice were fed with shrimp fed mja-miR-35- expressing bacteria(133) or treated with synthesized shrimp miR-34(134).
Therapy effect and efficacy
The potential role of miRNAs could be categorized based on their mode of action and of therapeutic efficacy established in pre-clinical BC mouse models. The number of experiments and references regarding the therapy effect and efficacy are summarized in Figure 5 and in Table 5.
Among the biological effects reported in mice, tumor growth alone (30,7%) or in combination with tumor metastasis (34,2%) are resulted to be the effects most studied.
Indeed, tumor growth inhibition occurred in tumor-bearing mice intravenous injected with several miRNAs (let-7, miR-145, -335, -34a, -203, -376B, -205/Anti miR-221, -379-5p, Anti miR-21) delivered using different approaches such as NPs(92, 95, 97, 101, 103, 106, 108-110, 113, 114), and extracellular vehicles(111, 129, 131, 132). The inhibition in tumor growth occurred in mice injected with BC cells transfected with miR-442a(66), -100(25), -27b(73), -567(81), -455-3p(63), -301a-3p(90), AntagomiR-138 (72), cirBulg21/223(76) compared to mice injected with BC cells transfected with a control plasmid. On the contrary, miR-196a over-expression in MDA-MB-231 -Luc cells promoted this capability (78).
Tumor growth was impaired in tumor bearing mice treated with: Linifanib(128), TargapremiR-210(126), small molecule “1” (specific for miR-544)(127), FolamiR-34a(121) , Trichostatin A (an inhibitor of histone deacetylase that up-regulates miR-125a-5p)(70), “CMM489” (a chemically modified miR-489)(117), or with Shrimp miR-34(134).
Interestingly, the injection of lipid vehicles loaded with AgomiR-186-3p(94) and with AgomiR-143(93) inhibited tumor growth and reduced the uptake of [ 18F]-fluoro-deoxyglucose ([18F]-FDG).
Regarding the effects of miRNA delivery on either tumor growth and lung metastasis, we found that luciferase expressing BC cells transfected with: miR-101(53), -141(89), -361-5p(50), -30a-5p(64), -125a-5p(71), -1(85), -211-5p(42), -190(68), -206(27), -33b(30), -33a(49), -96(61), -133b(47), -1(32), -494(44), -29b/-30d(83), Anti miR-1204(46), miR-191/-425 sponge(34) exerted antitumor and metastatic activity compared to BC cells transfected with an empty vector. Silencing of let-7a-5p(77) and of miR- 16-1-3p(119) in MDA-MB-231 and of miR-338-3p(118) in 4T1 cells influenced tumorigenesis and lung metastasis after implantation in nude mice.
The effects of miR-122 on glucose metabolism, tumor growth and metastasis were evaluated in different animal models using luciferase-labelled BC transfected cells or EV containing miR-122(36). Antitumor and antimetastatic effects was evaluated after injection of NPs loaded with specific miRNAs (-34a(105), -96/-182(98), -708(31), AntimiR-21/-10b(54), AntagomiR-10b(102)) or following the treatment with AC1MMYR2 (a specific small-molecule inhibitor of miR-21)(125), with AntagomiR-100(135) or with the antioxidant Pterostilbene(136). A novel approach was reported by Wu and colleagues in which the co-delivery of miR-21 inhibitor and indocyanine green (ICG) exerted anticancer activity photokilling MDA-MB-231 cells(116).
Twenty animal models (17,5%) were done studying the effects of miRNA delivery on lung metastasis, and only few experiments were performed analyzing bone (7,9%), brain (1,7%) and liver metastasis (0,87 %).
Lung metastasis were suppressed when BC-Luc cells were transfected with the following miRNAs: miR-630(56), -452(59), -590-3p(80), -150(57), -543(60), -133a-3p(58), -133b(37), 14q32 microRNA cluster (84), or transfected with the inhibitors for miR-106b-5p(82), -23a(74), -454-3p(52) or when mice were injected with Shrimp miR-35(133), with a small molecule that bind the precursor of miR-21 activating its destruction(124). On the contrary, an increased incidence of metastasis was established in mice injected with BC cells over-expressing miR-29a(62), -373(51). miR-548j overexpression increased the metastatic potential of BC cells without affecting tumor growth (45). Five studies reported that miR-200 family members (miR-200a, miR-200b, miR-200c, miR-429, miR-141) play an important role in the primary tumor formation and in the metastatic phenotype of BC (35, 40, 41, 86, 88).
The co-delivery of miRNA and small-molecule chemotherapy drugs in tumor site represents a promising strategy to fight the cancer progression in mice. In this context, the co-delivery of Dox with miR-34a(104) or with miR-159(132) in cancer site suppressed tumor growth. A regression of lung metastasis disease was established by the cotreatment of miR-10b and Dox(29). The combination treatment of Taxol and AC1MMYR2 (a small molecule that reduce miR-21 expression)(123), or of miRNA-34a and docetaxel (107) impaired tumor growth and metastasis. Paclitaxel and miR-124 coloaded in lipid nanosystem impaired lung metastasis formation in orthotopic mice(115). Co-delivery of miR-96/-182 with cisplatin, using NPs, reduced primary tumour and prevented lung metastasis formation(98).
Two experiments reported that brain metastasis formation was affected in vivo by the modulation of miR-509(67) and of miR-141(48). In only 1 study liver metastasis was impaired by the administration of EV carrying miR-4443 inhibitor(38). Bone metastasis was impaired by the over-expression in BC cells of miR-124(39), -429(43), -205(79), -940(65), -125b(33), -30 family members(28) or by the inhibition of miR-218-5p (75) or by intratumorally injection of synthetic miR-135/-203 mimics(55) or of the osteoclast-targeting AntagomiR-214-3p using (D-Asp)8-liposome(96).
Currently, the detection of miRNAs in cancer tissues could help to monitor the progression of cancer. From our research, biodistribution studies were found in 6 articles (5,3%). miR-155 expression was monitored in 2 different studies by intravenously injection of PA nanoprobe(99) and by the synthesized peptide nucleic acid (PNA) mimic loaded with [99mTc](122). Molecular beacon (MB) circuit was developed to monitor the expression of miR-34a in BC tissue with high sensitivity(23). A nanosensor conjugated with a MN-NPs allowed to discriminate BC cells from non-tumoral cells based on miR-10b expression(101). Monitoring the expression of miR-200c (87) and of miR-14/-21/-9(100) in tumor bearing mice was useful to determine the therapeutic approach. Finally, tumor angiogenesis was evaluated in 3 studies reporting that miR-497 exhibited anti-angiogenesis and anti-tumor effects targeting VEGFR2(24), miR-210 promoted angiogenesis(130), while miR-125a-5p affected tumorigenesis, metastasis, and angiogenesis in vivo(71).
Molecular Imaging
Most of the known preclinical imaging techniques have been applied in studying miRNAs delivery and/or efficacy. Figure 6 shows the absolute number of experiments for each imaging modality, and Table 6 shows the number of experiments for each specific modality and aim, and the relative references.
Bioluminescence resulted the most used tool (64%); this technique was used as a surrogate of tumor growth for efficacy treatment or for the evaluation of tumorigenicity in miRNA transfected cells (29,9%); for tracking, evaluation of engraftment and response to therapy in metastatic models (50,6%); for both the afore mentioned aims in the same experiment, evaluating metastasis either in vivo or ex vivo on whole organs (16,1%). As already reported, in one experiment (1,1%), a transgenic VEGFR2-luc mouse was used to evaluate the expression of VEGFR by non-invasive bioluminescence, and to evaluates the effect of miRNA-mimic treatment as anti-angiogenetic therapy(24). Bioluminescence was also used for vector uptake and intercellular target repression (2,3%), although most of these experiments were performed by fluorescence imaging.
Fluorescence imaging was the second most used technique (21,3%) and was used primarily to trace vector biodistribution (73,2%) by using different strategies, e.g., by directly conjugating the miR to the fluorophore, or simply uploading the fluorophore within the vector. In one interesting report, the vector was neither a NP nor an extracellular vesicle nor a liposome, but the vector was a folate, directly linked to the miR as well as to a near-infrared (NIR) fluorophore for fluorescent detection(121). Fluorescence was rarely used for tumor growth evaluation (11,5%), analysis of tumor persistence, after direct intratumoral injection, of the miR labeled with fluorophore(116) or within fluorescent SiO2 NPs(106) (7,7%), and for cell tracking (3,8%). One interesting experiment (3,8%) showed the ability of a molecular beacon to detect and image endogenous miRNAs with a high level of specificity in vivo(23). Besides these two mostly used imaging techniques, other tools were used to study biodistribution or different aspects of miRNAs treatment efficacy. Micro Computed Tomography (µCT) was used to analyze in vivo or ex vivo osteolytic lesions in metastatic bone models or to identify pulmonary metastases (5,2%). The former evaluation was performed with standard radiography (1,5%) in two other experiments(28, 39). Magnetic Resonance Imaging (MRI) was used in 2,9% of the experiments, mainly for detection of magnetic NPs biodistribution, and only in one experiment for the evaluation of invasiveness of adjacent tissue(38). Positron Emission Tomography (PET)/CT was applied with [18F]- FDG administration to evaluate tumor growth, in term of tumor glucose metabolism, or for detection of pulmonary metastases (2,2%). High Frequency Ultrasonography (HFUS) was performed to evaluate tumor growth or microbubbles-mediated nanoparticles delivery as therapeutic intervention (1,5%). Photoacoustic (PA) (0,7%) imaging was used to determinate the ability of self-assembling nanoprobes to identify specific miRNA. In brief, in presence of the specific miRNA, aurum aggregation from the nanoprobes, via a hybridization chain reaction, allowed identification of the PA signal(99). Finally, single photon emission computed tomography (SPECT) (0,7%) was used to label and track molecular probe, and to evaluate both the specificity in detecting the selected miR and both for biodistribution purposes(122).
In addition to what has been already stated a multimodal imaging approach, i.e., the use of multiple imaging technologies to evaluate different aspects or models within the same manuscript, was used and evaluated in 19 papers (16,7%) (28, 29, 31, 33, 38, 39, 43, 75, 84, 93, 96, 98, 101, 102, 104, 108, 111, 130, 137). Finally, it is important to highlights that in in some manuscripts in which multiple animal model are developed, the primary tumor growth was evaluated exclusively by tumor caliper measurement or tumor weighting ex vivo, whereas imaging (in vivo or ex vivo) was applied only for metastasis evaluation(28, 32, 34, 35, 42, 44, 46, 47, 50, 61, 64, 69, 77, 83, 85, 89, 118, 119, 134). In other studies, the therapeutic effects of miRNA delivery were evaluated independently from their biodistribution visualization obtained with preclinical imaging (92, 103-107, 109, 110, 112, 115, 121, 129).