Zebrafish (Danio rerio) WT/AB line was used in this study. Zebrafish embryos were maintained in embryo medium on a 14/10 light/dark cycle at 28.5°C. In case of the formation of pigment, 0.2 mM N-phenylthiourea (PTU, sigma) was added to the embryo medium at 24 h post fertilization (hpf). All animal manipulations were preformed strictly following the guidelines and regulations presented by the University of Science and Technology of China (USTC) Animal Resources Center and University Animal Care and Use Committee. The protocol was approved by the Committee on the Ethics of Animal Experiments of the USTC (Permit Number: USTCACUC1103013).

To overexpress miRNAs, a construct containing pri-miR-133b/pri-miR-23a/pri-miR-21 was made by amplifying a genomic region containing the miR-133b/miR-23a/miR-21 precursor. The resulting PCR fragments were then inserted into the linearized pUAS-mCherry digested by NotI, locating at the 3′-UTR of mCherry.

To knock down miR-133b, we used the miRNA “sponge” assay, which presents an efficient and permanent miRNA loss-of-function by imperfectly binding to a miRNA of interest (Cohen, 2009). The plasmid pUAS-mcherry-8 × miR-133b sponge was designed by ourselves and then constructed by Sangon (Shanghai, China).

To generate overexpression of TPPP3 construct, full-length tppp3 was initially amplified from complementary DNA (cDNA) of the WT/AB zebrafish strain. The PCR fragment was inserted into a plasmid backbone containing UAS. Plasmid UAS-tppp3 was co-delivered with both pUAS-mCherry and pCMV-Gal4-VP16 while electroporation.

ShRNA design was performed using the siRNA design tool under the following website: http://www.genscript.com/design_center.html (Dong et al., 2013). We selected the top five shRNAs (shRNA1-shRNA5) for further experiment. ShRNA expression vector was constructed in the following way: The modified mir30e backbone (Dong et al., 2013) was firstly synthesized with PacI-NheI sites for cloning target shRNA oligos. This modified mir30e precursor was cloned into pmini-Tol2-UAS-tdTOM vector downstream of tdTOM ORF to generate pmT2-UAS-tdTOM-mir30e-ShRNA (SG1180-A). We then cloned the fragment containing miR-shRNA structures (guide sequence, loop sequence, target sequence, and the flanking sequences) into pUAS-mCherry plasmid, locating in mCherry 3′-UTR. Target shRNA structures were synthesized by Sangon (Shanghai, China) and then cloned into PacI-NheI site.

One-cell stage zebrafish embryos were injected with a solution consisting of 30 ng/μl CMV-Gal4-VP16 plasmid and 30 ng/μl pUAS-mCherry/pUAS-mCherry-mircoRNA/pUAS-mCherry-miR-shRNA. To detect miRNAs level, 3 days post fertilization (dpf) zebrafish larvae with relatively high mosaic red fluorescence were selected for total RNAs isolation by miRNA Isolation Kit (Tiangen), according to the manufacturer's protocols. Each sample was reverse-transcribed into cDNA by miRNA First-Strand cDNA Synthesis Kit (Tiangen) and was subjected to qRT-PCR analysis with qPCR Detection Kit (Tiangen). To detect mRNAs levels, 10 hpf zebrafish embryos expressing red fluorescence were selected to isolate total RNAs with the same kit mentioned above. Each sample was reverse-transcribed into cDNA with HiScriptII Q RT SuperMix (Vazyme) and was subjected to qRT-PCR analysis with AceQ qPCR SYBR Master Mix (Vazyme). Each experiment was carried out with three biological and experimental replicate. Results were shown as mean fold changes ±s.e.m. qRT–PCR primers were shown in Table S1.

Before electroporation, 4 dpf zebrafish larvae were embedded in 1% low-melting agarose gel on an electroporation chamber. Using a micropipette (WPI, USA) pulled by a micropipette puller (P-97, Sutter, USA) to electroporate plasmids into the M-cell soma by pushing the tip against it with a series of pulses at 14–16 V. CMV-Gal4-VP16 plasmid was co-delivered into the unilateral M-cell of zebrafish larva with pUAS-mCherry-microRNAs (plasmids used to overexpress specific miRNA)/pUAS-mCherry-microRNA sponge (plasmid used to inhibit specific miRNA)/pUAS-mCherry-miR-shRNA (plasmid used to inhibit tppp3). Each plasmid concentration is 120 ng/μl. Zebrafish electroporated with pCMV-Gal4-VP16 and pUAS-mCherry were treated as control. For the experiment to overexpress TPPP3, pUAS-tppp3 was delivered into cell soma with both pCMV-Gal4-VP16 and pUAS-mCherry. After electroporation, larvae were returned back to embryo medium containing PTU. Then we selected morphologically healthy zebrafish expressing red fluorescence in M-cells for later experiment.

Before axotomy, 6 dpf zebrafish larvae expressing red fluorescence in unilateral M-cells were anesthetized in MS222 (Sigma, USA) and fixed in 1% low-melting agarose. A Zeiss microscope (LSM710, Germany) was used to ablate the M-cell axons over cloacal pores. We normally set the 800 nm two-photon laser at an intensity of 12–15% to damage axon over ~1.5 s (Xu et al., 2017).

Before imaging, embryos were anesthetized by MS222 and then embedded in 1% low melting point agarose in embryo medium containing MS222. All images and time-lapse movies were taken from lateral views of the spinal cord, anterior to the left, and dorsal toward the top.

To observe M-cells regrowth after ablation at 6 dpf, anesthetized zebrafish were imaged at 1–2 days post-axotomy (dpa) using Olympus FV1000 confocal microscope (Olympus, Tokyo, Japan) equipped with a 40x, 0.8 N.A. water-immersion objective at 2-μm intervals. All images well spliced using with Photoshop CS4 (Adobe, USA). We defined the starting point of regrowth as the ablated site of axons just above cloacal pores, and the axonal terminal of regeneration was stipulated as the end point of regrowth axons. In this article, regeneration length refers to the maximum regenerated axon length of one branch, while total regeneration length refers to all the regenerated axon branches length combined. All regenerative length was calibrated to convert pixels into distance using FV10-ASW 4.2 viewer software.

For investigating mitochondrial transport in single M-cell in vivo, zebrafish larva electroplated with pUAS-mito-eGFP (plasmid used to label mitochondria) were imaged at 6 dpf using a confocal microscope with a 60x, 0.9 N.A. water-immersion objective. 2.5-min movies of the axonal area, locating within 200 nm proximal to the site above the cloacal pores, were taken with an imaging frequency about 1.5 s, and the imaging length of axons was ~43 mm at the site of the axon. All images were processed with Fiji/ImageJ (National Institutes of Health, USA). The quantification of mitochondrial dynamics were measured as previously described (Misgeld et al., 2007; Plucinska et al., 2012; Takihara et al., 2015; Xu et al., 2017). Mitochondrial motility was defined as the percentage of moving mitochondria, which were identified to move more than 2 μm, during the 2.5-min time-lase movies. The velocity of a moving mitochondrion referred to the total moving distance of a mitochondrion divided by its observed moving time.

In vitro transcription of EGFP-tppp3 3′-UTR, EGFP-tppp3 mut-3′-UTR and mCherry mRNAs were performed with mMESSAGE mMACHINE T7 Ultra Kit (Invitrogen) and these synthesized mRNAs were purified with MEGAclearTM Kit (Invitrogen). Zebrafish embryos at one-cell stage were injected with a combing solution of sensor mRNA and mCherry mRNA. When applicable, 10 μM miR-133b duplex was added as an experimental group, while 10 μM non-sense duplex was added as a control. EGFP fluorescence was quantified at 24–28 h post-fertilization (hpf) using software Fiji-imageJ.

The distribution of data points was expressed as mean ± standard error of the mean (S.E.M.), or as relative proportion of 100% as mentioned in the appropriate legends. Depending on the number of the groups and independent factors, student's t-tests, one-way analyses of variance (ANOVA) and non-parametric tests were used as indicated in the figures. Results were classed as significant as follows: ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001.

We have identified in our previous study that M-cells have strong regenerative capacity (Xu et al., 2017). More than 90% of two-photo ablated M-cells could regenerate a certain length in our experiments. To explore the role of miR-133b in M-cell axon regeneration, we performed cell type-specific overexpression. A vector-based miRNA expression was used to achieve enduring expression of the miRNA during our experimental time window. We constructed a vector containing dre-pri-miR-133b sequence (miRBase Accession: MI0001994) in the 3′-UTR of mCherry, which conveniently marked the cells that expressed the miR-133b (Figure 1A). The plasmid UAS-mCherry-miR-133b was co-injected with pCMV-Gal4-VP16 into one-cell zebrafish embryos. As a control, embryos were injected with pUAS-mCherry and pCMV-Gal4-VP16. We then selected zebrafish larvae with relatively high mosaic red fluorescence at 3 dpf to isolate the total RNA (Figure 1B). Our qRT-PCR data showed that miR-133b in experimental group (EG) was more than three times of that in control, indicating that our constructed plasmid UAS-mCherry-miR-133b could successfully drive overexpression of miR-133b (Figure 1C).

Next, we used this vector system to overexpress miR-133b in individual M-cells at 4 dpf via single-cell electroporation. We selected the zebrafish with red fluorescence in unilateral M-cell at 6 dpf for two-photon laser axotomy and visualized axon regeneration at 2 dpa (Figure 1D). Our imaging data showed that most M-cells in control could regenerate a certain length, while M-cell overexpressing miR-133b could hardly regenerate [control: 243.7 ± 32.9 μm, n = 33 fish vs. miR-133b overexpression (OE): −14.8 ± 20.7 μm, n = 16 fish] (Figures 1E,F). To further verify the specific role of miR-133b in regulating axon regeneration, we overexpressed another two miRNAs, miR-23a and miR-21, with the same assay as mentioned above. Together with qRT-PCR results confirming that miR-23a and miR-21 were indeed overexpressed in zebrafish via vector-based miRNA expression assay (Figure 1C), we found out that miR-23a, a miRNA that has not been reported to be associated with axon regeneration, had no obvious effect on M-cell axon regeneration; while miR-21, which has been shown to promote regeneration in different organs (Strickland et al., 2011; Han et al., 2014; Hoppe et al., 2015), remarkably promoted M-cell axon regeneration (control: 243.7 ± 32.9 μm, n = 33 fish vs. miR-23a OE: 229.0 ± 62.4 μm, n = 10 fish vs. miR-21 OE: 778.4 ± 60.8 μm, n = 12 fish; Figures 1E,F).

Since researches on dre-miRNAs often explore their roles in different processes using miRNA duplex, to further verify miR-133b's role on axon regeneration, we also expressed the miR-133b duplex in M-cell by single-cell electroporation. M-cells expressing only rhodamine-dextran (3,000 molecular weight, Invitrogen) (named None) seemed to have similar outgrowths to those expressing non-sense duplex (named Negative Control), while both explicated a slight increase, even though without significant discrepancy, compared to M-cells in experimental group (expressing miR-133b duplex), not matter at 1 dpa or 2 dpa (1 dpa: None: 142.3 ± 19.0 μm, n = 26 fish; Negative control: 140.0 ± 15.8 μm, n = 24 fish; miR-133b duplex: 102.7 ± 17.6 μm, n = 23 fish; 2 dpa: None: 464.8 ± 40.5 μm, n = 20 fish; Negative control: 396.7 ± 32.5 μm, n = 22 fish; miR-133b duplex: 373.4 ± 33.9 μm, n = 23 fish; Figure S1). This result was consistent with the results obtained by vector-based system, indicating that miR-133b has negatively effects on M-cell axon regeneration. Together, these results indicate that the reduction of M-cell regenerative capability by miR-133b is specific and cell intrinsic.

To determine whether loss of miR-133b in single M-cells could also regulate its axon regeneration, we needed an assay that could achieve long-term miRNA loss-of-function. MiRNA sponges have been shown to efficiently bind to endogenous miRNAs and block their silencing activity with bulged miRNA binding sites (Ebert et al., 2007; Cohen, 2009; Otaegi et al., 2011). Moreover, the bulged sites can protect against cleavage and degradation of sponge RNA by the Ago2 component of the RISC (Ebert et al., 2007; Ebert and Sharp, 2010), which can satisfy our experimental requirement.

We constructed a plasmid containing 8 bulged target sites complementary to miR-133b in 3′-UTR of mCherry reporter gene driven by the UAS promoter (Figure 2A). To testify the ability of this plasmid in blocking miR-133b activity in zebrafish, we examined the expression of a known miR-133b target gene, mps1 (Yin et al., 2008), in 10 hpf zebrafish embryos injected with a combination of pUAS-mCherry-8 × miR-133b sponge and pCMV-GAL4-VP16 at one-cell stage. The mps1 mRNA level increased in zebrafish larvae expressing the miR-133b sponge, suggesting that it could reduce miR-133b activity in zebrafish (Figure 2B).

Next, we examined the consequence of knocking down miR-133b activity in axon regeneration. Remarkably, most axons regenerated with supernumerary branches (Figure 2C). The longest regeneration length of a single axon had no significant difference between control and experimental group (control: 253.7 ± 34.9 μm, n = 30 fish vs. miR-133b sponges: 296.7 ± 40.5 μm, n = 20 fish; Figure 2D). However, the total regeneration length, all branches combined, was significantly different (control: 486.7 ± 78.8 μm, n = 30 fish vs. miR-133b sponges: 835.2 ± 131.4 μm, n = 20 fish; Figure 2E). The experimental group had significantly more axonal branches than the control (control: 2.13 ± 0.28, n = 30 fish vs. miR-133b sponges: 4.50 ± 0.83 μm, n = 20 fish; Figure 2F). Collectively, these results demonstrate that blocking the function of miR-133b promotes M-cell axon outgrowth, which is a phenotype that is complementary to overexpressing miR-133b in M-cells.

Typically, one miRNA can suppress the expression of many genes by interacting with the 3′-UTR or the coding regions of the targets mRNAs (Lewis et al., 2005; Duursma et al., 2008; Forman et al., 2008). We searched several databases, including TargetScan Fish, miRBase and microcosm Targets, and identified potential targets containing complementary regions to miR-133b seed sequences in their 3′-UTR. We focused on one gene, tppp3, which has a single binding site for miR-133b at its 3′-UTR. In addition, tppp3 corresponds perfectly to nucleotides 2–7 of the mature miR-133b in zebrafish (Figure 3A). TPPP3 is a member of tubulin polymerization promoting protein family. Previous studies identified TPPP3 as a potent inducer of tubulin polymerization (Vincze et al., 2006) and human TPPP3 binds and stabilizes microtubules (MTs; Oláh et al., 2017). Since regulation of axonal microtubule (MT) dynamics influence axon regeneration (Sengottuvel and Fischer, 2011; Bradke et al., 2012; Hur et al., 2012), and pharmacological stabilization of MTs by paclitaxel or related molecules promotes axon regeneration in vitro and in vivo (Hellal et al., 2011; Sengottuvel et al., 2011; Ruschel et al., 2015), we hypothesized that miR-133b might regulate axon regeneration through directly modulating tppp3 mRNA in vivo.

We firstly detected the mRNA level of tppp3 in miR-133b overexpressed or miR-133b sponge expression zebrafish embryos. Our qRT-PCR results showed that the mRNA level of tppp3 in 10 hpf zebrafish embryos overexpressing miR-133b was lower than that in control (Figure 3B), while tppp3 mRNA level was increased in embryos expressing miR-133b sponge compared with that in control (Figure 3C). We then used zebrafish embryo sensor assays (Giraldez et al., 2005). Two mRNAs were synthesized, one encoding enhanced green fluorescent protein (EGFP) with 3′- UTR of tppp3 and the other composed of mCherry fluorescent protein with a poly(A) alone. These mRNAs were co-injected into one-cell zebrafish embryos, in the presence of miR-133b RNA duplex or non-sense duplex (GenePharma). Injections of these two mRNAs along with a non-sense RNA duplex (negative control) resulted in both high EGFP expression and mCherry expression. However, when a synthesized duplex of miR-133b was co-injected, EGFP signals were dampened by almost 50% with no detective changes in mCherry signals (negative control: 100.0 ± 7.0%, n = 10 fish vs. miR-133b duplex: 43.5 ± 3.7%, n = 10 fish; Figures 3D,E). When the seed sequence in the 3′-UTR of tppp3 was mutated, we found no difference in EFGP signals between non-sense RNA duplex and miR-133b RNA duplex (negative control: 100.0 ± 11.9%, n = 10 fish vs. miR-133b duplex: 117.1 ± 12.0%, n = 10 fish; Figures 3F,G).

In conclusion, our results indicate that tppp3 is a downstream gene of miR-133b in vivo.

Given the effects of miR-133b on tppp3 expression and the role of miR-133b in neurite outgrowth, we next planned to investigate the effects of gain or loss-of-function of tppp3 on regenerative axon growth. We firstly overexpressed tppp3 by electroplating into M-cell at 4 dpf a plasmid containing the zebrafish tppp3 cDNA. As a control, pUAS-mcherry and pCMV-GAL4 was delivered. Consistent with the effects of miR-133b sponge on axonal regeneration (Figure 4A), overexpression of tppp3 in M-cell significantly increased the total regeneration length (Regenerative length: control: 255.6 ± 37.2 μm, n = 27 fish vs. TPPP3 OE: 382.9 ± 66.6 μm, n = 15 fish; total regeneration length: control: 476.2 ± 83.2 μm, n = 27 fish vs. TPPP3 OE: 855.2 ± 177.4 μm, n = 15 fish; Figures 4B,C), although there is no significant difference in branching number (control: 2.11 ± 0.29, n = 27 fish vs. TPPP3 OE: 3.60 ± 0.73, n = 15 fish; Figure 4D).

To test whether knockdown of tppp3 might cause regenerative defects similar to miR-133b overexpression, we used designed shRNAs to silence tppp3 based on the miR-ShRNAs system. MiR-shRNAs have now been widely used in mammals and zebrafish (De Rienzo et al., 2012; Dong et al., 2013; Shinya et al., 2013), in vitro and in vivo (Giraldez et al., 2005; Zuber et al., 2011), due to its higher efficiency than simple hairpin designs. We designed shRNAs employing the primary miR-30 backbone. Based on the Web-based shRNA design tool (https://www.genscript.com), five shRNAs (shRNA1-shRNA5) targeting the tppp3 gene were selected (Figure S2). mCherry was used as a fluorescent reporter to mark the zebrafish embryos that expressed the miR-shRNA (Figure 5A). To valid the function of these shRNAs, we injected the miR-shRNA expressing plasmids combing with pCMV-GAL4 into one-cell stage embryos and isolated mRNA of these embryos exhibiting red fluorescence at 10 hpf to examine the tppp3 mRNA level. We found that, among these five shRNAs, shRNA-5 exhibited the significant reduction of tppp3 mRNA level (Figure 5B). We then investigated the effects of shRNA-5 on axonal regeneration by delivering it into M-cells via single-cell electroporation at 4 dpf. To avoid the effects of other miR-shRNA structures (guide sequence, loop sequence, and the flanking sequences) on the capability of regeneration, cells expressing shRNA-1, which had little effects on reducing tppp3 mRNA (Figure 5B), were used as an additional control. Both imaging and quantitative results indicated that shRNA-5 diminished the regenerative length of damaged axons, while shRNA-1did not (control: 278.2 ± 33.1 μm, n = 26 fish vs. miR-shRNA-1: 314.2 ± 42.7 μm, n = 8 fish vs. miR-shRNA-5: 97.6 ± 47.7 μm, n = 20 fish; Figures 5C,D).

Taken together, these results indicate that tppp3 is critical to promote axon outgrowth.

Mitochondria plays a critical role in axon regeneration, a highly energy-demanding process. Our previous study has indicated that mitochondrial trafficking is associated with axon regenerative capacity, suggesting that axons having more motile mitochondria regenerate better than those having less ones (Xu et al., 2017). Moreover, another research group finds out that mature injured axons in mice can regenerate by enhancing mitochondrial motility via genetic manipulation, which helps remove damage mitochondria and recruit new ones to meet the energy demands at injury sites during regenerative process (Zhou et al., 2016).

To examine whether miR-133b overexpression had any effects on mitochondrial dynamics, we co-transfected pUAS-mito-EGFP and pUAS-mcherry-miR-133b driven by the expression of pCMV-GAL4 via single-cell electroporation at 4 dpf and visualized the movement of mitochondria at 6 dpf via in vivo time-lapse confocal imaging, through which stable vs. mobile mitochondria could be discerned (Figure 6A, Video S1). By counting and analyzing mitochondria in M-cells, we identified that the percentage of motile mitochondria was much lower in miR-133b overexpressing conditions than in control (Figure 6B, Video S2), and this reduction was more significant in retrograde than in anterograde directions (Total: control: 20.31 ± 2.34%, n = 11 fishes vs. miR-133b OE: 10.70 ± 2.14%, n = 13 fishes; antero: control: 13.21 ± 1.89%, n = 11 fishes vs. miR-133b OE: 7.91 ± 1.92%, n = 13 fishes; retro: control: 7.10 ± 1.11%, n = 11 fishes vs. miR-133b OE: 2.79 ± 0.77%, n = 13 fishes; Figure 6C). Moreover, mitochondrial velocity in the miR-133b overexpression group was slower in both transport directions compared with that in control, though in retrogradely moving mitochondria it did not reach significance (Total: control: 0.501 ± 0.018 μm/s, n = 54 mitos from 11 fishes vs. miR-133b OE: 0.404 ± 0.015 μm/s, n = 55 mitos from 13 fishes; antero: control: 0.488 ± 0.015 μm/s, n = 39 mitos from 11 fishes vs. miR-133b OE: 0.396 ± 0.017 μm/s, n = 41 mitos from 13 fishes; retro: control: 0.535 ± 0.052 μm/s, n = 15 mitos from 11 fishes vs. miR-133b OE: 0.4294± 0.034 μm/s, n = 14 mitos from 13 fishes; Figure 6D). Together, our results suggest that miR-133b is an important cell intrinsic regulator of mitochondrial dynamics during M-cell axon regeneration.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.