Assessment of Artificial MiRNA Architectures for Higher Knockdown Efficiencies without the Undesired Effects in Mice

RNAi-based strategies have been used for hypomorphic analyses. However, there are technical challenges to achieve robust, reproducible knockdown effect. Here we examined the artificial microRNA (amiRNA) architectures that could provide higher knockdown efficiencies. Using transient and stable transfection assays in cells, we found that simple amiRNA-expression cassettes, that did not contain a marker gene (−MG), displayed higher amiRNA expression and more efficient knockdown than those that contained a marker gene (+MG). Further, we tested this phenomenon in vivo, by analyzing amiRNA-expressing mice that were produced by the pronuclear injection-based targeted transgenesis (PITT) method. While we observed significant silencing of the target gene (eGFP) in +MG hemizygous mice, obtaining −MG amiRNA expression mice, even hemizygotes, was difficult and the animals died perinatally. We obtained only mosaic mice having both “−MG amiRNA” cells and “amiRNA low-expression” cells but they exhibited growth retardation and cataracts, and they could not transmit the –MG amiRNA allele to the next generation. Furthermore, +MG amiRNA homozygotes could not be obtained. These results suggested that excessive amiRNAs transcribed by −MG expression cassettes cause deleterious effects in mice, and the amiRNA expression level in hemizygous +MG amiRNA mice is near the upper limit, where mice can develop normally. In conclusion, the PITT-(+MG amiRNA) system demonstrated here can generate knockdown mouse models that reliably express highest and tolerable levels of amiRNAs.


Introduction
RNA interference (RNAi)-mediated gene silencing is a useful approach for rapidly obtaining hypomorphic phenotypes by repressing selected gene.Many different strategies are used for RNAi-mediated gene silencing that involve the expression of sequences such as short interference RNA (siRNA), short hairpin RNA (shRNA) or artificial microRNA (amiRNA).These sequences are designed to include the complementary sequence for a target mRNA with which they bind in RNA-induced silencing complex that will repress target gene expression (knockdown) both in vitro and in vivo [1,2].Because RNAi-mediated gene silencing is easier to use and gene function can be more rapidly analyzed compared to knockout techniques until recently when CRISPR/Cas9 system became popular, it has been widely used for analyzing a gene's function, for genome-wide genetic screening and also for gene therapy.
In order to assess the gene function reliably and satisfactorily, it is important to achieve sufficient knockdown of the target gene that can elicit a phenotype for further analysis.Several in vitro approaches have been employed by researchers to improve the efficiency of RNAi-mediated gene silencing.For example, McBride et al. screened shRNAs in vitro to select the most effective from among many designed shRNAs [3].Some groups optimized hairpin structures or miRNA backbones that could improve knockdown potency [4,5].Another group used multiple amiRNA in a single transcript [6].However, such strategies have not been systematically evaluated using in vivo models.In contrast, it is also reported that high shRNA expression levels can result in severe toxicity in some tissues (e.g., liver, central nervous system, and heart) and/or lethality in mice, rats, and dogs [3,7].These potential adverse effects of in vivo RNAi might be alleviated by using shRNA embedded within a natural miRNA backbone (also known as artificial microRNA/amiRNA) although in vivo toxicity of amiRNA architectures is not well studied.
The in vivo knockdown effects using RNAi have often been evaluated by administering siRNA or establishing transgenic (Tg) mice by using plasmids or virus vectors that contain shRNA-or amiRNA-expression cassettes [8].However, the knockdown levels in mice obtained with these methods vary and are often not reproducible due to transient inhibition by siRNA or unreliable expression of shRNA/amiRNA in those Tg mice generated by random integration-based transgenesis [8,9].Although these pitfalls can be circumvented by using targeted Tg mice generated via ES cell targeting [10], the ES cell-based methods are laborious, expensive and time-consuming.
We recently developed a novel system to generate Tg mice that we called Pronuclear Injection-based Targeted Transgenesis (PITT), for targeted insertion of a single-copy of a transgene into a predetermined genomic locus, such as Gt(ROSA)26Sor (Rosa26) [11].Because transgenes can be delivered to zygotes via pronuclear injection similar to conventional transgenesis methods, rapid, simple generation of Tg mice is achieved through PITT.Using this method, we have generated a series of Tg mouse lines that express a gene of interest driven by a ubiquitous CAG promoter.Using this technique, strong, reproducible and stable transgene expression can be achieved.We also demonstrated that this method can be used for generating knockdown mice [11], although, in our previous study, we had not characterized whether these mice exhibit reproducible, stable knockdown efficiencies.
In this study, we compared the efficiency of gene silencing of amiRNA expression cassettes with-or without-marker genes and also demonstrated that our PITT method is suitable for creating amiRNA-expressing mice with highly reproducible knockdown efficiency.We found that the simplest amiRNA-expression cassette that lacked a marker gene exhibited a higher degree of knockdown efficiency both in vitro and in vivo.The higher amiRNA expression levels (in constructs lacking marker gene) impaired the normal growth and viability of mice.We also demonstrate that marker gene containing amiRNA Tg mice generated using PITT showed higher and desirable levels of knockdown without the toxicity or viability issues.

Cell culture and transfection
ES cells, eGFP-expressing ES cells [11] and E14tg2a [16] or E14.1 [17], were cultured in Dulbecco's modified Eagle's medium (Gibco, Grand Island, NY) supplemented with 15% fetal bovine serum, 0.5% penicillin/streptomycin, and 0.01% leukemia inhibitory factor (LIF; Chemicon, Temecula, CA) at 37°C in a humidified atmosphere with 5% CO 2 .Cells were seeded onto 12-well plates (1 × 10 5 cells/well) or 24-well plates (5.7 × 10 4 cells/well) one day before transfection.To compare the knockdown efficiencies among the knockdown constructs in ES cells, amiRNA expression vectors (at the same molar amounts of pCAG in control experiments: 1 μg/well in 24-well plates and 2 μg/well in 12-well plates) were transfected into cells using Lipofectamine 2000 (Invitrogen), according to the manufacturer's protocol, and with or without pG (0.63 μg/well in 24-well plates and 1.25 μg/well in 12-well plates) for eGFP knockdown.For some experiments, pL was co-transfected with these vectors to evaluate transfection efficiencies.After incubation for 48 hours, the cells were observed for fluorescent signals.
To test whether the knockdown efficiency was affected by the presence of a marker gene, eGFP-expressing ES cells were subjected to co-transfection using Lipofectamine 2000 (Invitrogen) with each of the PB-based amiRNA expression vectors (No. 14-16 in S2 Table ), and a PB transposase expression vector (pPBase), and/or a neo r gene expression PB vector (pPBN) [18,19].The eGFP-expressing ES cells had the eGFP-expression cassette (CAG promoter-eGFP-polyA) inserted into the Rosa26 locus by recombinase-mediated cassette exchange [11].At 24-48 h after transfection, cells were subjected to selection using G418 (300 μg/ml) for 1-2 weeks.Colonies that were considered to be derived from a single cell were picked and propagated further.These propagated cells were then subjected to Lipofectamine 2000-based transfection with a Dre expression vector (pDre).After about a week, colonies were picked and propagated.eGFP and amiRNA expression in each clone was analyzed by FACS and by real-time PCR, respectively.
To generate eGFP knockdown mice (double Tg mice, designated AWV Δex /eGFP, S3 Table ), AWV Δex mice were crossed with eGFP Tg mice [11].The resulting offspring were checked for double Tgs by PCR using the primer sets M411/M412 to identify an amiR-eGFP transgene and M245/M701 to identify an eGFP transgene derived from eGFP Tg mice.Wild-type Rosa26 alleles were also checked using the primer set M273/M274.
Genotyping for homozygous knockdown mice (produced by intercrossing hemizygous AWV Δex mice in S3 Table) was performed by PCR using the primer sets M124/M274 and M274/M273.
The AWK ex/Δex /FLPe mosaic mice tended to display signs of possible premature death.Therefore, we employed a "humane endpoint," which is defined as the time at which the animals were euthanized that was based on when the animals began to show clinical signs of illness (for example, reduced exploration, difficulty in obtaining food, and hunched posture).The animals were monitored daily, supplied with food on the bottom of the cage, and when they showed signs of illness, they were euthanized according to the Guidelines for the Care and Use of Animals for Scientific Purposes at Tokai University.

Removal of extra sequences via FLPe recombining
AWV ex and AWK ex mice were generated using second-generation donor vectors, pAWV and pAWK, which included a "CAG-FLPe-polyA" cassette to aid in the self-removal of extra sequences that would be expected to result in AWV Δex and AWK Δex mice (S3 Table) [14].Because self-removal of the extra sequences (such as the FLPe expression unit and the vector backbone) in the AWK ex founder mouse did not occur, this founder was then crossed with FLPe Tg mice.The offspring derived from AWV ex and AWK ex mice were analyzed for the removal of extra sequences by PCR using the primer sets M070/M124 and M212/M273.The presence of the FLPe transgene from FLPe Tg mice was also checked using the primer set M632/M244.

In vitro fertilization
Sperm cryopreservation was performed for an AWK ex/Δex /FLPe double Tg male mouse.Unfertilized oocytes isolated from super-ovulated female mice (BDF1) were subjected to in vitro fertilization (IVF) using these spermatozoa based on a protocol by Nakagata (2011) [21].

FACS analysis
Cells that transiently or stably expressed amiRNA were washed twice with 1× phosphate buffered saline (PBS) and incubated with 0.05% trypsin-EDTA at 37°C in 5% CO 2 to remove cells from a cell culture surface.After 5 min, complete medium was added to the plates to stop cell dissociation.After centrifuging at 800 rpm for 5 min, cells were resuspended in complete medium and then passed through a 35-μm mesh filter (Falcon 352235, BD Biosciences, Franklin Lakes, NJ, USA) for FACS analysis.Spleen cells prepared from eGFP knockdown mice were cleared of red blood cells by osmotic lysis, resuspended in complete medium, and then filtered as described for cultured cells.
Prepared cells were subjected to flow cytometry using a FACSCalibur (BD Biosciences, San Diego, CA) or an LSRFortessa (BD Biosciences) to assess knockdown efficiency and/or a FAC-SAria (BD Biosciences, San Jose, CA) to sort cells of interest.Data were analyzed using FlowJo software (Tree Star, Inc., Ashland, OR, USA).
For the set of Tg lines, the expression levels of eGFP and tdTomato were measured as the mean fluorescence intensity (MFI) by FACS analysis.The MFIs for eGFP and tdTomato were normalized by dividing the MFIs by eGFP Tg and tdTomato Tg lines, respectively.The experiments were repeated three times.We examined the difference in the normalized MFIs for eGFP and tdTomato across three AWV Δex lines by analysis of variance (ANOVA) in which the effects of line and experiment were included as the independent factor variables.The two-way ANOVA was performed using R.
To evaluate the effect of removing a marker gene on knockdown efficiency, we measured the MFI of eGFP for a set of cell lines with (+MG) or without (−MG) the marker gene along with the wild-type and eGFP positive lines.The experiments were repeated six times for tdTomato marker gene, or seven times for neo r marker gene.The MFIs for eGFP were normalized by dividing the MFIs by the eGFP positive line within each experiment.The difference in the normalized MFI between +MG and −MG lines was evaluated by the paired t-test.

Real-time PCR
Total RNA was isolated from transfected cells using a mirVana miRNA Isolation Kit (Ambion, Austin, TX), according to the manufacturer's protocol.Purified RNA was then quantified using a NanoDrop 2000 (Thermo Scientific, Wilmington, DE). amiRNA cDNAs were synthesized from 10 ng of total RNA using a TaqMan microRNA RT Kit and miRNA-specific RT primer sets for target amiRNAs (eGFP123 and eGFP419) and endogenous control miRNA (sno202) (Applied Biosystems).The conditions used for cDNA synthesis were 16°C for 30 min, 42°C for 30 min, 85°C for 5 min, and 15°C pause.To quantify amiRNA, quantitative real-time PCR was performed using TaqMan small RNA Assays (Applied Biosystems), according to the manufacturer's instructions.In brief, the reaction mixture (10 μl) contained 4.5 μl of amiRNA cDNA (diluted 1:10, 1:20, 1:40, 1:80, 1:160, and 1:320 with nuclease-free water), 1× TaqMan Universal PCR Master Mix II, and a small RNA-specific TaqMan MGB probe (eGFP123, eGFP419 and sno202) (Applied Biosystems).Thermal cycling was conducted using StepOne Plus (Applied Biosystems) with an initial denaturation at 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min.The miRNA expression level was determined using the 2 −ΔΔCt method and normalized to sno202.Average threshold cycle (C) and SD values were determined from triplicate samples for each experiment, and were representative of at least three separate experiments.
Tissue biopsy samples (liver and muscle) that had been preserved in RNAlater (Qiagen, Hilden, Germany) were homogenized and total RNA was extracted using a mirVana miRNA Isolation Kit (Ambion), according to the manufacturer's protocol.Quantitative real-time PCR was conducted as described above.

Analyses of transfected cells and knockdown mice
Fluorescent signals in transfected cells were observed using a Keyence BZ-9000 fluorescence microscope with the filter sets OP-66836 for eGFP and XF175 for tdTomato.Fluorescent signals in eGFP knockdown mice were observed using a Leica M165 FC with filter sets for GFP and/or red fluorescence.Mouse eyes were inspected with an Olympus SZX12 stereomicroscope.

Comparative analysis of different amiR-eGFP expression architectures for gene-silencing in ES cells
To establish robust knockdown system, we first investigated the optimal structure for amiRNA expression cassettes that could produce effective and desirable levels of gene knockdown.For this purpose, we generated a series of amiR-eGFP expression constructs and transfected them into eGFP-expressing ES cells to test their knockdown efficiency (Fig 1A and S2 Table).At 2 days after transfection, cells were harvested and analyzed by FACS to assess eGFP fluorescence.
Knockdown of eGFP fluorescence was detected in cells that were transfected with pKD-G3 and pKD-G4 (Fig 1B).The knockdown efficiency in pKD-G3 transfected cells was slightly greater than that of pKD-G4 transfected cells which can be due to differences in the efficiency of the amiRNA sequences (No.1 and 2 in Fig 1B).No differences in the knockdown levels were found between cells that had been transfected with either the vector that contained tandem amiR-eGFP cassettes (pKD-G7: amiR-eGFP123 and amiR-eGFP419) or the one with single amiR-eGFP cassette (pKD-G3: single amiR-eGFP123; No.1 and 3 in Fig 1B).In addition, knockdown efficiency did not seem to be influenced by either the position of amiR-eGFP in the expression construct (No. 4 to 6 in Fig 1C ) or the order of two tandemly arrayed amiR-eGFP sequences (eGFP123-eGFP419 or eGFP419-eGFP123: pKD-G5 and G6; G7 and G8; G9 and G10; and G11 and G12 in S1 Fig).
In contrast, the presence of a marker gene in an expression cassette affected its knockdown efficiency.An expression cassette without the tdTomato marker gene [without marker gene (−MG)] exhibited the highest knockdown efficiency (No.3 in Fig 1C).Even though the knockdown efficiencies among all the +MG vector architectures were comparably similar (Fig 1C ), the marker gene expression (tdTomato) was higher among the vectors in which the amiRNA was placed within the intron, but not in the 3 0 UTR region (compare No.

Effects of presence of a marker gene in an amiRNA expression cassette on knockdown efficiency in stable cell lines
It is possible that the differences in knockdown efficiency in ES cells observed in previous experiments may have been simply caused by differences in transfection efficiencies among the constructs of different sizes.To assess if knockdown efficiency was indeed affected by the presence of a marker gene, we generated stable cell lines that contained an amiR-eGFP expression cassette.For this purpose, a piggyBac transposon system was used in combination with a Dre-rox site-specific recombination system (Fig 2A).This strategy allowed a well-controlled comparison scenario between a construct with a tdTomato gene and one without this gene, because both constructs were inserted at the identical genomic locus as a single-copy configuration/locus.When the piggyBac transposon vector pPBKD-G1 that harbored a "tdTomato and amiR-eGFP123-419" expression cassette was transfected into eGFP-expressing cells together with a pPBase, two stable cell lines were obtained (+MG; 3 and 4), each of which exhibited different tdTomato fluorescence intensity.These lines were then administrated with a Dre recombinase expression plasmid, pDre, to obtain cells without a tdTomato gene (−MG).FACS analysis of all cell lines showed that eGFP fluorescence intensity in-MG cells was lower than that in +MG cells (Fig 2B ), indicating the higher knockdown of-MG architectures as observed in the previous experiment (Fig 1).The reduction was observed consistently in both the cell lines (Line 3, P = 1.8 × 10 −3 ; Line 4, P = 1.4 × 10 −5 ).In addition, real-time PCR analysis showed that the expression of amiR-eGFP123 in -MG cells was higher than that in +MG cells (Fig 2C ), which is in agreement with the increased knockdown in -MG architectures.Even though the real-time PCR analysis for amiR-eGFP419 expression (the second part of the cassette) showed a similar trend as amiR-eGFP123, the results were not statistically significant.We obtained similar results, of higher knockdown among -MG architectures, when the tdTomato reporter marker gene was replaced with a neomycin-resistant (neo r ) gene (S3 Fig) .Taken together, these results indicate that removing the MG, such as tdTomato or neo r , from an expression cassette results in increased target gene knockdown efficiency.

Reliable knockdown efficiency in mice generated by PITT
For reliable comparison of in vivo knockdown efficiency between amiRNA expression constructs, it is important to generate amiRNA Tg mice exhibiting knockdown of a target gene in a highly reproducible manner.We have established pronuclear injection-based targeted transgenesis (PITT) system that can produce Tg mice showing reproducible transgene expression [11].We previously constructed a donor vector, pAWV (S2 Table ), which harbored two amiR-eGFPs (eGFP123 and eGFP419) in the 3 0 UTR of a tdTomato expression cassette in addition to the mutant loxPs (JTZ17 and lox2272).This vector was used to generate Rosa26 targeted Tg mice using PITT approach as described before [14].We obtained three amiR-eGFP (+MG)  ) that were subsequently mated with eGFP Tg mice.These eGFP Tg mice used are proven to exhibit stable, reproducible, eGFP fluorescence [11].The resulting double Tg mice (i.e., eGFP knockdown mice) were designated "AWV Δex /eGFP mice" (S3 Table ) and were analyzed for eGFP knockdown.
As expected, eGFP fluorescent signals were un-detectable in any major organs, including the liver, kidney, intestine, heart, and muscle of the AWV Δex /eGFP mice derived from all three independent Tg lines (lines 5, 6 and 7 in Fig 3A).It is notable that eGFP fluorescence intensity in spleen cells derived from these founder lines, as analyzed by FACS, was reduced by approximately 50-fold as compared to that in spleen cells of a parental eGFP Tg mouse, and all the 3 lines exhibited similar decrease in fluorescence intensity in their spleen cells (5,6,and 7 in Fig 3B).The efficiencies for reducing the eGFP expression were not statistically different across these three lines (P = 0.45).Consistent with the results obtained by FACS analysis, the eGFP mRNA levels in the liver and kidney were reduced to 1-2% compared to the parental eGFP Tg mice (5,6,and 7 in Fig 3C).The tdTomato fluorescence intensities, that serve as a marker gene for amiRNA expression in these lines, were statistically similar to each other (P = 0.87).Notably, the tdTomato fluorescence intensities were lower than those in a tdTomato Tg mouse (approximately 20%).This may have been due to the placing of of amiRNAs in the 3 0 UTR (but not in the intron) of an expression cassette; similar observations are reported by other groups [23,24] (Fig 3D ).Taken together, these results indicate that PITT is a useful strategy for generating reliable knockdown mice.

Marker gene-free amiR-eGFP expressing mice are non-viable
In our original PITT system, the F0 animals will have extra sequences that get integrated at the Rosa26 locus (ex-allele).These sequences can be subsequently deleted by breeding the founders with FLPe-expressing mice (Δex-allele).In the second generation PITT vectors we overcome the necessity of extra breeding step by including a FLPe expression cassete that would automatically excise the extra sequence in F0 animals.We previously constructed pAWK second generation donor vector that carried a "CAG promoter -amiR-eGFP123 -amiR-eGFP419 -polyA" cassette.This vector was used to generate Rosa26 targeted Tg mice using PITT approach (Fig 4A) [14].Strangely, we noticed as described before that, in the pAWK vector-derived 'amiRNA Tg' model, the extra sequence was not deleted, even though it was derived from second generation PITT system [14].All the F1 offspring from this founder mouse also had this extra sequence.
Next, we tried to forcefully remove this extra sequence by crossing F0 or F1 mice with FLPe Tg mice because, we thought that the presence of extra sequences, including the vector sequence and the FLPe expression cassette, could affect the expected amiRNA expression.However, we could not obtain offspring that demonstrated the successful removal of this extra sequence.Only one double Tg male (AWK ex/Δex /FLPe), was found to have the amiRNA expression cassette together with the FLPe transgene.Interestingly, this mouse was mosaic and contained mixture of both cells that carried an Δex-allele or an ex-allele (Fig 4A).Notably, although this mosaic mouse was fertile, it failed to transmit the amiRNA expression cassettes (either the Δex-allele or the ex-allele) to the next generation by natural mating.
Further, to obtain AWK ex/Δex /FLPe offspring, we performed in vitro fertilization (IVF) using frozen sperm prepared from the AWK ex/Δex /FLPe mouse.We successfully obtained 20 newborns, and 5 pups (no.16-20 in  Quantitative realtime PCR analysis showed that the amiRNA expression levels were higher in mosaic mice compared to the mouse with the ex-allele, as recorded in the liver samples (Fig 4E).This was in agreement with the observation that no eGFP knockdown was detected in the liver of the AWK ex /eGFP mice (S6 Fig) .Based on all these results, we conclude that the difficulty in obtaining AWK Δex mice was due to the toxicity of overexpressed amiRNA itself from the amiR-eGFP (−MG) transgene of Δex-allele.

Failure to obtain homozygous amiRNA (+MG)-expressing mice
To further confirm whether excessive amiRNA expression adversely affects mouse survival, we attempted to generate homozygous amiR-expressing mice (+MG) by inter-crossing the hemizygous AWV Δex mice described above.However, we failed to obtain homozygous AWV Δex mice so far (number of wild-type: hemizygote: homozygote = 5: 7: 0).We examined the deviation of the genotype frequencies from the expected Mendelian proportions (1: 2: 1) by using a multinomial test via EMT package in R. The suggested levels of the deviation were observed (P = 0.071).Then, we evaluated whether the number of mice for homozygous AWV-Δ ex were smaller than expected (25%) by using a binomial test.We found significantly greater decrease in the numbers of homozygous mice than expected (P = 0.046).This result suggests that it is difficult to generate homozygous AWV Δex mice, which is probably due to the deleterious effects on their survival by excessive amiRNA expression.

Discussion
Most of the previously reported knockdown mouse models are generated by random integration-based transgenesis (e.g., conventional microinjection methods and virus-mediated gene transfer) and they tend to show highly variable knockdown efficiencies [25].One of the goals of this study was to test if knockdown mice generated using PITT method show reliable expression of the knockdown cassettes.Indeed, the targeted AWV Δex /eGFP Tg mice, generated by PITT approach, exhibited highly reproducible knockdown efficiency.Because the expression of both "eGFP transgene" and "miRNA transgene" was very critical in this mouse model, our PITT method served as a perfect system to generate such a model since it can produce reliably expressing Tg lines.Furthermore, PITT method employs direct microinjection to target the transgene to a predetermined locus, it is less time-consuming and less expensive when compared to an ES cell-mediated method, which is typically used when one needs to generate a targeted knock-in of knockdown cassettes [10,26].Thus, PITT could be a powerful strategy for generating knockdown mice.We are currently developing a PITT-based strategy to generate inducible knockdown mice.
There are numerous examples of studies that attempted to obtain sufficiently high knockdown efficiency.Some studies focused on miRNA processing pathways [4,27,28], whereas others focused on the structures of shRNA/miRNA expression constructs [6,[29][30][31].In this study, we constructed a series of expression constructs and compared their knockdown potencies.We concluded that 1) simple expression constructs that lack a marker gene (−MG) resulted in higher knockdown efficiency than those with a marker gene (+MG); and 2) knockdown efficiency was not improved/affected by bicistronic miRNA, its order, or the location of miRNA within the expression cassette.Previous studies, that used cell culture models, show both contradictory [6,29,30] and supporting observations [31] to our findings in this study.Notably, there are no examples of previous in vivo studies because, the models that allow direct comparison, like those reported here, could not be generated with other techniques with relative ease as done using PITT, and therefore our study helped draw this important conclusion.
Hu et al. reported that the presence/absence of a reporter gene had no detectable effect on RNAi efficiency [29].They used the DsRed gene as a marker in only transiently transformed cells.In contrast, we used either tdTomato or a neo r gene as marker genes, and we examined knockdown efficiencies using both transiently transformed cells and stably transfected cells.Similar results were obtained with either of the marker genes in our experiments.In addition, we obtained consistent results both in vitro and in vivo.Considering the aforementioned factors, we are confident that the presence of a marker gene interferes with miRNA expression, at least in our experimental conditions.
Our results also suggested that the number and position of amiRNA cassettes in the transcription units had no obvious effect on knockdown efficiency, which was inconsistent in some previous studies [6,30].Regarding the number of amiRNAs used, our results were in agreement with those by Shan et al. [31] who showed that the knockdown efficiency obtained with three tandem amiRNAs expression cassette was at a level similar to that obtained by using the most effective amiRNA expressed by a single amiRNA expression cassette.One possibility for the contradictory results in Sun et al. [6], who used a miR-30 precursor backbone while we and Shan et al., [31] used a miR-155 precursor backbone.Future investigations will be necessary to better understand additional rules for optimal amiRNA construct design.
Grimm et al. reported that strong shRNA expression resulted in in vivo toxicity [7,32] which suggest that other options, such as use of amiRNA, can be considered to avoid the pitfalls of using shRNA cassettes.However, even when amiRNA was used to produce knockdown mice, we observed, for the first time to our knowledge, in vivo toxicity.First, we failed to generate Tg mice with an amiRNA (−MG) transgene even though we could generate amiRNA (+MG) Tg mice.Second, homozygous knockdown mice with a marker gene (+MG) were not obtained.Because an amiRNA (−MG) expression cassette tended to transcribe large amounts of amiRNA, compared to an amiRNA (+MG) expression cassette, we concluded that our inability to produce amiRNA (−MG) Tg mice could be due to reasons such as (i) saturation effect for an endogenous microRNA pathway [7,27,32,33] or (ii) excessive off-target effects of amiRNA-eGFP used in this study.If the former reason is true, then in vivo toxicity could be circumvented and homozygous knockdown mice could be obtained by supplying rate-limiting factors of the endogenous microRNA pathway (e.g., Exportin-5 or Argonate-2) [27,28,34].
The lethality of AWK Δex /FLPe mice was partially rescued in AWK ex/Δex /FLPe mosaic mice (Fig 4 and S4 Fig) .In our previous reports, we showed that gene expression in PITT method generated Tg mice with an ex-allele was weak and often sporadic, whereas those with an Δex-allele showed strong and stable expression [11].Therefore, cells with the ex-allele were considered to be amiRNA low-expression cells even when they had an amiRNA (−MG) transgene, and this transgene's expression increased after extra sequence was removed for generating Δex-allele.The mosaic mice with both amiRNA (−MG) high-expression cells (Δex-allele) and low-expression cells (ex-allele) exhibited growth retardation and were susceptible to cataract formation.These results further strengthened our conclusion that high amiRNA expression resulted in in vivo toxicity.
Finally, our results showed that the presence of a marker gene in an amiRNA expression construct affected the knockdown efficiency, which could be enhanced after removing the marker gene from the expression construct.This suggests that knockdown levels can be altered by simply including or excluding a marker gene.In addition, we found that the in vivo toxicity of amiRNA was due to its higher level of expression.Although this problem remains to be addressed further, it is possible to express amiRNA at near highest tolerable levels in a reproducible and predictable manner by generating hemizygous amiRNA (+MG) Tg mice by our PITT method.
In most cases, RNAi-based knockdown does not completely abolish target gene expression.The previous study showed that only a 20% reduction in tumor suppressor PTEN gene expression was sufficient to show a cancer development phenotype [35].In addition, approximately 30% of knockout mouse lines are presumed to be embryonic or perinatal lethal [36] and a large proportion of these mutants are expected to express phenotypes in heterozygotes.Considering these observations, it is highly likely that analyzing hypomorphic mice generated by our knockdown approach would be appropriate for many experimental settings.
In conclusion, this work demonstrates that (i) knockdown efficiency was increased when a marker gene from an amiRNA expression cassette was excluded, (ii) PITT-based generation of Tg knockdown mice using amiRNA cassettes without a marker gene may be difficult because of the non-tolerable levels of amiRNA expression and its toxicity in mice, (iii) the PITT can be a suitable method to generate reliable knockdown mouse models showing highest tolerable levels of amiRNA expression without any toxicity and viability issues, when a marker gene-linked amiRNA cassette is used as a heterozygous mouse.
5 and No.6 with No.4 in Fig 1D).We observed similar effects when different experimental design (using a wild-type ES cell co-transfected with different amiRNA cassettes and an eGFP-expressing plasmid) was used (S2 Fig).

Fig 1 .
Fig 1.Comparison of knockdown efficiencies of different amiR-eGFP architectures in eGFP-expressing ES cells.(A) Structures of amiR-eGFP expression vectors.Two amiRNAs, each of which was designed to target different positions of the eGFP gene (hairpin structures in pink, for eGFP123, and blue, for eGFP419), were used (See S1 Table for details).(B-D) Fluorescence intensity determined after transfecting eGFP-expressing ES cells with each vector (No. 1 to 6 in A).At 2 days after transfection, the fluorescence intensity of eGFP (B and C) and tdTomato (D) was assessed by FACS.Orange lines (arrowheads in B and C) are eGFP fluorescence intensities of non-transformed eGFP-expressing ES cells (control: C).The mean fluorescence intensities (MFIs) are indicated in the upper right-hand corner of each graph in (B-D).Numbers 1 to 6 in (B-D) correspond to those shown in (A).doi:10.1371/journal.pone.0135919.g001

Fig 2 .
Fig 2. Removing the tdTomato marker gene from an amiRNA expression cassette enhances knockdown efficiency.(A) Schematic representation of the experimental strategy.An amiR-eGFP expression vector (pPBKD-G1) that contained a tdTomato marker gene flanked by rox sites was co-transfected into eGFP-expressing ES cells along with a pPBase and a pPBN.Stably transformed clones that exhibited tdTomato fluorescence (+MG clone) were isolated first.Then, these clones were administrated Dre recombinase to remove the tdTomato gene, which resulted in generating a marker-less clone (−MG clone).(B) eGFP fluorescence intensity in ES cells.The experiments were repeated six times and the representative histogram is shown (left).Cell lines 3 (solid line) and 4 (dot line) were used for analysis.Red and blue lines represent eGFP fluorescence in a "+MG clone" and a "−MG clone," respectively.Green and black lines indicate eGFP fluorescence intensity in eGFP-expressing and wild-type (eGFP-negative) ES cells, respectively.The box plots show the normalized MFI obtained in six experiments (right); *P = 1.8 × 10 −3 , **P = 1.4 × 10 −5 (paired t-test).(C) Quantitative real-time PCR of amiR-eGFP123 (left) and amiR-eGFP419 (right) in cell lines 3 (+MG and −MG) and 4 (+MG and −MG); The experiments were repeated three times; *P = 2.6 × 10 −3 , **P = 6.9 × 10 −3 (two tailed Student's t-test).The differences are not significant in amiR-eGFP419 expression (P > 0.05).doi:10.1371/journal.pone.0135919.g002

Fig 3 .
Fig 3. Knockdown efficiency in AWV Δex mice.(A) eGFP and tdTomato fluorescent signals in mouse organs.Mouse genotypes (No. 1 to 7) are indicated on the right.Mice 5 to 7 are AWV Δex /eGFP mice, in which the transgene included in each mouse had been derived from a different founder (F0) mouse (lines 1 to 3).An amiRNA expression cassette in AWV Δex Tg mice is shown (upper right).(B) eGFP fluorescent signals in spleen cells analyzed by FACS.(C) Quantitative real-time PCR of eGFP mRNA expression in mouse liver and kidney tissues.(D) tdTomato fluorescent signals in spleen cells analyzed by FACS.Numbers 1 to 7 in (B-D) correspond to those shown in (A).doi:10.1371/journal.pone.0135919.g003 Fig 4B) died soon after birth.Genotyping analysis of the stillborn pups showed that all were double Tg mice (AWK Δex /FLPe) that carried both Δexallele and the FLPe transgene but not ex-allele (Fig 4B and S3 Table).In contrast, 3 out of the 15 living newborns (no.3, 5, and 7 in Fig 4B) were amiR-eGFP Tg mice that contained cells with an ex-allele only or cells with both an ex-allele and an Δex-allele in a mosaic pattern (Fig 4B).Additional attempts to generate mice with the Δex-allele alone failed, while mosaic mice with cell populations with the ex-allele or the Δex-allele were obtained occasionally among the offspring of AWK ex and FLPe Tg mice matings.The mosaic mice often showed growth retardation (Fig 4C) and premature death (S4 Fig).All surviving mice (n = 9) had cataracts, which became severe as they aged (Fig 4D and S5 Fig).They also failed to transmit the transgene to the next generation (S4 Fig).

Fig 4 .
Fig 4. Generation of marker gene-less (−MG) amiR-eGFP-expressing mice.(A) Schematic representation of the strategy used to eliminate extra sequences in AWK mice.Extra sequences including vector sequence (shown in moss green line) and an FLPe expression cassette in the ex-allele, were removed by administering FLPe, which resulted in generating an Δex-allele.An FLPe expression cassette in FLPe Tg mice is shown below.Red arrows indicate primer-binding sites used for genotyping analysis.(B) Genotyping results of pups obtained by in vitro fertilization using frozen sperm derived from an AWK ex/Δex /FLPe mouse (lanes 1 to 20).Lane 21: negative control.PCR conducted using the primer set shown in (A) was used to assess the presence of transgenes.Fifteen individuals (1 to 15) were alive, whereas 5 pups (16 to 20) died just after birth [Note: for the sample No. 20 with primer set (M632/M244), the PCR product was loaded onto a separate lane in the same gel and rearranged in this Figure for clarity].(C) Body weights of littermates obtained by crossing an AWK ex mouse with an FLPe Tg mouse.The AWK ex/Δex /FLPe mouse contained both type of cells (with the AWK Δex allele or AWK ex allele) classified as a mosaic mouse.(D) Cataracts developed in the AWK ex/Δex /FLPe mouse.All mice, similar to those observed in (C), were examined at 66 weeks of age.Cataract phenotypes in additional AWK ex/Δex /FLPe mice are shown in S5 Fig. (E) Scatter plot of quantitative real-time PCR of amiR-eGFP123 expressions (log2 of normalized data) in nine mosaic mice and one AWK ex mouse including littermates as in (D).Each circle and triangle represents an individual mouse.The values are expressed as mean ± SD represented by the green line for mosaic mice.doi:10.1371/journal.pone.0135919.g004