Evaluation of endogenous miRNA reference genes across different zebrafish strains, developmental stages and kidney disease models

The majority of kidney diseases arise from the loss of podocytes and from morphological changes of their highly complex foot process architecture, which inevitably leads to a reduced kidney filtration and total loss of kidney function. It could have been shown that microRNAs (miRs) play a pivotal role in the pathogenesis of podocyte-associated kidney diseases. Due to their fully functioning pronephric kidney, larval zebrafish have become a popular vertebrate model, to study kidney diseases in vivo. Unfortunately, there is no consensus about a proper normalization strategy of RT-qPCR-based miRNA expression data in zebrafish. In this study we analyzed 9 preselected candidates dre-miR-92a-3p, dre-miR-206-3p, dre-miR-99-1, dre-miR-92b-3p, dre-miR-363-3p, dre-let-7e, dre-miR-454a, dre-miR-30c-5p, dre-miR-126a-5p for their capability as endogenous reference genes in zebrafish experiments. Expression levels of potential candidates were measured in 3 different zebrafish strains, different developmental stages, and in different kidney disease models by RT-qPCR. Expression values were analyzed with NormFinder, BestKeeper, GeNorm, and DeltaCt and were tested for inter-group differences. All candidates show an abundant expression throughout all samples and relatively high stability. The most stable candidate without significant inter-group differences was dre-miR-92b-3p making it a suitable endogenous reference gene for RT-qPCR-based miR expression zebrafish studies.

MicroRNAs (miRs) regulate protein expression by translational suppression via RNA interference. With their seed sequence they bind specifically to target mRNAs, thereby blocking translation and/or facilitating mRNA degradation 1 . In the recent past it could have been shown that microRNAs play a pivotal role in kidney development, physiology and pathology [2][3][4] .
Over the years, larval zebrafish have emerged a popular model organism, due to their small size, high reproductive potential and genetic accessibility 5 . The fact that zebrafish develop a fully filtrating glomerulus attached to a pair of tubules with high structural and molecular homology two days past fertilization (dpf) makes this model highly relevant for the investigation of human glomerular kidney diseases 6 . Most of these diseases arise from highly specialized glomerular cells, the podocytes. They are central players for glomerular permselectivity as they form the filtration slits together with neighboring podocytes. This is ensured by their complex interdigitating branching morphology. Interruptions in this morphology or the loss of podocytes subsequently leads to the development of nephrotic syndrome, a pathologic condition in which the size selectivity of the kidney filter gets impaired. Patients with nephrotic syndrome develop high-molecular weight proteinuria together with hypoproteinemic edema as typical characteristics 7 . A multitude of studies used larval zebrafish to model proteinuric glomerular diseases may it be with the help of morpholino-guided knockdown of podocyte genes 8,9 or treatment with podocyte-directed drugs [10][11][12] . Exemplary for morpholino-guided gene knockdowns, we targeted two disease-causing genes for which zebrafish models have been established. First, wt1a, a transcription factor essential for podocyte differentiation and maintenance 8 . Second, nphs1, the gene encoding nephrin, a cell-cell contact protein that is the important for glomerular filtration barrier function and which absence in patients leads to steroid-resistant nephrotic syndrome.
The relevance of miRs for glomerular pathophysiology has been proven several times. Studies have shown that the podocyte-specific knockout of the miR-processing ribonucleases Dicer and Drosha leads to proteinuria and glomerulopathy [13][14][15] . Beside their functional roles in the pathogenesis of all known types of glomerular diseases, miR deregulations can play an important role as biomarkers 16,17 .
Currently the gold standard in miR quantification is RT-qPCR 18 . For comprehensive analysis of RT-qPCR expression data, a reliable normalization strategy is required. The most widely applied method is the use of at least one endogenous reference gene 19,20 . The requirements for a suitable normalizer are sample type-independent abundant expression and a high stability throughout a certain sample set and it should not exhibit expressional differences between specific sub-and treatment groups.
Unfortunately, there is no consensus about a normalization strategy in zebrafish miR expression data. This is even more obvious when it comes to kidney research: Most studies make use of endogenous reference genes typically used in human and rodent studies. To the best of our knowledge there is no study specifically dealing with suitable endogenous normalizers focusing on zebrafish development and kidney disease models.
To address this, we analyzed the suitability of nine, from miR-sequencing preselected, miRs as endogenous reference genes for RT-qPCR based zebrafish miR expression data in different zebrafish strains, developmental stages and four different glomerular disease models.

Drug treatment.
Embryos with a Tg(nphs2:GAL4), Tg(UAS:Eco.nfsB-mCherry) background obtained from double transgenic incrosses and selected for strong and homogenous mCherry fluorescence in podocytes at 72 h past fertilization (hpf), were podocyte-depleted by treating embryos with 5 mM metronidazole (MTZ) dissolved in 0.1% DMSO in E3 medium or 0.1% DMSO in E3 as a vehicle control from 96 to 120 hpf as described before 11 . To induce focal and segmental glomerulosclerosis (FSGS)-like disease in zebrafish embryos, partial podocyte depletion was performed as described before 12 . Briefly, double transgenic Tg(nphs2:GAL4), Tg(UAS:Eco.nfsB-mCherry) embryos were treated with 80 µM MTZ in 0.1% DMSO in E3 for 48 h starting at 96 hpf. For each treatment three independent biological replicates from individual clutches of embryos were set up.
Morpholino Injections. Antisense morpholino oligonucleotides were produced by Gene-Tools (Philomath, OR, USA). Following beforehand established and published morpholino sequences were used (MO IDs http:// www. zfin. org): MO3-wt1a (ZDB-MRPHLNO-071107-2), 5′-CAC GAA CAT CAG AAC CCA TTT TGA G-3′ 8 ; MO1-nphs1 (ZDB-MRPHLNO-051102-1), 5´-CGC TGT CCA TTA CCT TTA GGC TCC -3´9. Standard control morpholino targeting an intronic region of the human HBB gene: 5´-CCT CTT ACC TCA GTT ACA ATT TAT A-3´. Lyophilized morpholinos were reconstituted in ultrapure water to a stock concentration of 1 mM. Before injections, target and control morpholinos were diluted to 100 µM in injection solution containing 100 mM KCl, 10 mM HEPES pH 7.6 and 1% phenol red as a visual marker and incubated at 65 °C for 5 min to dissolve precipitates. Per embryo, 2 nl were injected in the yolk of 1-2 cell-stage embryos. Injected embryos were collected in E3 and transferred to 10 cm petri dishes with fresh E3. Embryos were checked daily for viability and medium was changed twice a day. For each target, three independent clutches of embryos were injected. Proteinuria assay. As described above, 1-cell embryos (Tg(fabp10a:gc-eGFP), mitfa w2/w2 , mpv17 a9/a9) ) expressing the 78 kDa gc-eGFP fusion protein in the blood plasma were injected with respective morpholino oligonucleotides. Under healthy conditions, this fusion protein is retained in the vasculature. In the case of glomerular leakage, intravascular protein is filtered and lost with the urine. At 96 hpf, randomly selected larvae were imaged with a P2-SHR Plan Apo 1 × objective attached to a SMZ18 fluorescence stereomicroscope equipped with a motorized Z-drive (Nikon GMBH, Düsseldorf, Germany) and an X-Cite Xylis LED (Excelitas, Göttingen, Germany) under exact the same illumination conditions for every group.

RT-PCR.
For RT-PCR, 1 µg of RNA was reverse-transcribed to cDNA using the Quantitect Reverse Transcription Kit (Qiagen) according to manufacturer's protocol. RT-PCR to assess MO knockdown efficiency was performed using the Phire Hot Start II DNA polymerase (Thermo Fisher Scientific) according to manufacturer's instructions with 1 µl undiluted cDNA-template plus 19 µl master mix containing primers targeting exon 24-26 of nphs1 and ef1a1l1 as a reference gene. Primer sequences were: nphs1_exon24_F: GTC TAT GTG GTG GTG ATC CTG, nphs1_exon26_R: CTG TGC CGA GGC GTT GAT AA, ef1a1l1_F: AAG GAG GGT AAT GCT AGC GG, ef1a1l1_R: GGG CGA AGG TCA CAA CCA TA. As control a -RT, no template control from reverse-transcription and no template control from PCR Endpoint RT-PCR was run together with target samples in a Mastercycler gradient Thermocycler (Eppendorf) under following conditions: initial denaturation 98 °C 3 min, 33 cycles: 98 °C 10 s, 60 °C 20 s, 72 °C 25 s. RT-PCR products were resolved on 3.5% low melting agarose (Biozym) in 1 × TBE-buffer containing 0.16 µg/ml ethidium bromide.

Results
Zebrafish kidney disease models: morpholino-mediated gene knockdown. As established and widely used kidney disease models, we injected 1-cell-stage embryos with antisense morpholino oligonucleotides targeting wt1a translation initiation (translation blocking MOs: TBM) and nphs1 splicing of the intron between nphs1 exons 24-25 (splice blocking MOs: SBM). While the transcription factor wt1a plays an important role during glomerular development, nphs1 is a main component of the mature glomerular filtration barrier. Generally, as shown in Fig. 1A, injection of both MOs in 1-cell stage of zebrafish embryos resulted in typical pericardial edematous alterations 8,9 . Edema was graded in four categories from 0 to normal phenotype to 3 to severe whole-body edema with bent body axis. As shown in Fig. 1B,C, a statistically significant proportion of embryos injected with 2 nl of 100 µM anti-wt1a or anti-nphs1 morpholinos developed statistically significant pericardial or periocular edema. While the proportion of embryos showing edema of any severity was similar in both groups, the phenotype of the nphs1 knockdown embryos was generally more severe compared to the wt1a knockdown (Fig. 1B). To investigate whether wt1a and nphs1 knockdown induced high-molecular weight proteinuria, we injected 1-cell stage embryos expressing a 78 kDa gc-eGFP fusion protein under control of the hepatocyte-specific fabp10a promotor. When produced, gc-eGFP is secreted in the blood plasma where it circu- www.nature.com/scientificreports/ lates and is retained under healthy conditions. Upon glomerular injury and impairment of the glomerular filtration barrier, this fusion protein will be rapidly cleared from the vasculature resulting in a decrease of vascular eGFP fluorescence 34 . As shown in Fig. 1D, at 96 hpf, control MO-injected embryos showed eGFP-fluorescence in the vasculature while wt1a or nphs1 MO-injected larvae showed significantly lower eGFP fluorescence indicative for leakage of protein (Fig. 1D, images with correlative bright field micrographs are shown in Suppl. Figure 1). As shown in the exon-intron scheme in Fig. 1E, the SBM used for the knockdown of nphs1 blocks the splice donor site leading to a truncated protein due to integration of the intron between exons 24 and 25. RT-PCR (oligo positions depicted in Fig. 1E) amplifying exons 24-26 showed integration of the intron only after injection of the nphs1 SBM (889 bp) with down-regulation of the wild type allele (210 bp) not containing the intron. As a reference gene ef1a1l1 was used which was stably expressed across samples (Fig. 1F). In contrast to that, a lower concentration of MTZ depleted only a subset of podocytes and leads to a prolonged disease course. This mimics human FSGS with its diverse features (progressive proteinuria, parietal epithelial cell activation, extracellular matrix deposition) as we have shown before 12 . As shown in Fig. 2A, the phenotype progressively developed during the disease course over 4 days after the washout of MTZ with a lower concentration with increasing lethality until 9 dpf (Fig. 2B).

Stability analysis. The IRC corrected miRNA expression data was analyzed by the combined online tool
RefFinder. For the analysis we used all unsorted values together as well as subclassifications such as only strains, only MTZ treatments and only morpholino-mediated knockdowns as input. The data was analyzed by Best-Keeper, Genorm, Normfinder and DeltaCt.
Inter-group differences. Since a certain endogenous control should not only be stable throughout a specific sample set but should not show significant expression differences between sample subgroups and treatment groups, we tested the expression data of the single miRs for the aforementioned differences. We could not find statistically significant expression differences between the three different strains for any tested miR. Additionally, there were no significant differences between the MTZ-treated groups themselves as well as between the treatment-and the control-groups. This was also true for the morpholino groups, where we could not detect any significant differences between the nphs1-and wt1a MOs groups as well as between them and the control group (Fig. 6).
Developmental differences. Beside specific developmental zebrafish research, many studies are dealing with zebrafish larvae at different developmental stages. In case of miR-based RT-qPCR data this fact sets the requirement for an endogenous control that shows either a high stability or no significant differences between developmental stages. To address this, we tested our data set for differences between the single developmental stages. For dre-miR-92a-3p, dre-miR-92b-3p and miR-126a-5p we could not detect any significant differences. The other six candidates showed differences between different developmental stages. Dre-miR-206-3p showed significant differences between 72 and 120 h and between 96 and 120 h. The next candidate, dre-miR-99-1, showed significant differences between 24 h and the other time points. Dre-miR-363-3p had significant expression differences between 120 and 72 h and between 120 and 48 h.  (Fig. 7).

Discussion
Currently, there is no consensus about endogenous reference controls for RT-qPCR derived miR expression data in zebrafish larvae. To best of our knowledge, there are no studies explicitly approaching different drug treatments, knockdowns and/or developmental stages like it is the case for mRNAs 30,31 . In the present study we investigated the nine preselected candidates dre-miR-92a-3p, dre-miR-206-3p, dre-miR-99-1, dre-miR-92b-3p, dre-miR-363-3p, dre-let-7e, dre-miR-454a, dre-miR-30c-5p and dre-miR-126a-5p for their suitability as endogenous reference genes. As it has been shown before, zebrafish embryos exhibit highly dynamic miR expression patterns during development among different developmental stages in three different genetic zebrafish strains 2,32 . We started at 24 h past fertilization (hpf), because lower time points have only a little relevance for kidney research 6  www.nature.com/scientificreports/ abundance of maternal miRs in the yolk 32 . Additionally, to different background strains and the developmental series, we applied four different models, successfully established to model glomerular diseases in zebrafish: The first two models are based on the morpholino-guided knockdown of genes important for glomerular development and/or maintenance of the glomerular filtration barrier. The first one hindered translation initiation of wt1a, a transcription factor that regulates proper glomerular development and is still active in adult podocytes. The second one targeted proper splicing of nphs1, a gene encoding the protein nephrin which is a crucial part of the slit diaphragm and therefore maintaining glomerular filtration barrier function. In line with our results, knockdown of both genes has been shown to result in disrupted glomerular development as well as in early onset of high molecular weight proteinuria in larval zebrafish 8,9 .
The other models represent a pharmacogenetic form of highly specific podocyte depletion. They are grounded on the NTR/MTZ model of targeted tissue ablation which has been translated to podocyte research. In this model expression of the NTR, a bacterial enzyme, is transgenically driven by tissue-specific promotor fragments, such as in our case the nphs2 promotor. When zebrafish embryos are treated with the antibiotic MTZ, the NTR converts this prodrug into a cytotoxin which leads to rapid apoptosis in the targeted tissue. Initially this model has been established to deplete pancreatic beta-cells and has later been modified to a dose-dependent depletion of podocytes 10,33 . We have shown that upon treatment with high dose MTZ (5 mM) zebrafish rapidly develop proteinuria and classic morphologic changes known from human nephrotic syndrome such as severe podocyte foot process effacement and later podocyte detachment 11,34 . Lately, we could show that when dose-dependently only a smaller subset of podocytes is depleted, a more chronic course of disease is initiated. In that model, zebrafish www.nature.com/scientificreports/ embryos resemble the phenotype known from human FSGS including activation of parietal epithelial cells that migrate to the glomerular tuft and deposit extracellular matrix 12 .
After RT-qPCR based measurements, the expression levels of candidate miRs showed a high homology in general in the described models. There were no statistically significant differences detectable between the single subgroups, except for developmental stages. The SSS revealed dre-miR-206-3p as the most stable endogenous reference gene. Dre-miR-206-3p has been shown to play a crucial role in angiogenesis of zebrafish larvae 35 . It was followed by dre-miR-30c-5p which is known as a key player in pronephric development in Xenopus laevis and Danio rerio, where it shows a pronephros specific expression within the first days of development 36,37 . It has also been used as an endogenous reference gene before 38 . Unfortunately, there were significant differences in the expression of dre-miR-206-3p and dre-miR-30c-5p between different developmental stages, so they had to be discarded as suitable normalizers for this type of experimental set-ups. The only miRs that showed no differences in their expression levels between developmental stages were dre-miR-92a-3p, dre-miR-92b-3p and dre-miR-126a-5p. Dre-miR-92b-3p was the third most stable candidate closely after dre-miR-30c-5p. Additionally, it was exclusively ranked first in all subgroups by Bestkeeper. Dre-miR-92b-3p is known to be abundantly and ubiquitously expressed in zebrafish up from 24 hpf 39 and is essential, together with dre-miR-92a, for the earliest steps in zebrafish embryogenesis 40 .
The present study shows that all tested candidate miRs have a high homology within the tested experimental setups and that the most stable normalizer was dre-miR-206-3p. However, if it comes to developmental research studies or comparisons of zebrafish larvae at different ages it is rather unsuitable. Our analysis shows that dre-miR-92b-3p is the best candidate to be used as an endogenous reference gene for RT-qPCR-based miR expression in zebrafish larval studies.