Peripheral Axons of the Adult Zebrafish Maxillary Barbel Extensively Remyelinate During Sensory Appendage Regeneration

Zebrafish husbandry

Wild-type “AB” zebrafish were bred and reared under standard conditions at 28°C at DePaul University and/or Children’s Memorial Research Center (Chicago, IL). Larval fish were fed live rotifers and/or a commercial powdered diet once daily. Juvenile and adult fish were fed fish flakes and/or brine shrimp once or twice per day. All animal protocols were approved by the Institutional Animal Care and Use Committees.

Surgical procedures and tissue collection

To collect developing barbel tissue, juvenile zebrafish (30-90 days postfertilization) were anesthetized in 0.015% buffered Tricaine (MS-222; Sigma, St. Louis, MO), then killed in ice water. Under a stereoscope, the head was removed with a small scalpel or razor blade; larger specimens were bisected sagittally before fixation. To collect regenerating barbel tissue, adult zebrafish 3-12 months postfertilization were lightly anesthetized as described above, and then the left maxillary barbel of each fish was removed by proximal amputation as previously described (LeClair and Topczewski, 2009). The contralateral barbel was left intact as a control. Fish were returned to the system after several hours and were kept for 10 days, 28 days, or 70 days after surgery. On the day of collection, the fish were anesthetized as described above, then killed in ice water. Under a stereoscope, regenerating and control barbel tissues were dissected away from the head and processed as needed for RNA isolation or tissue fixation.

RNA isolation from normal and regenerating maxillary barbels

Freshly dissected adult maxillary barbels were rinsed in saline and transferred with forceps into ~200 μl of RNAlater (Ambion, Austin, TX) in a glass depression slide. Within this solution, the barbels were trimmed under a dissecting scope. Normal barbels were trimmed to exclude the maxilla but otherwise contained all of the cell types in the normal adult appendage. Regenerated barbels were trimmed to exclude the proximal stump of non-regenerating tissue, leaving only cells that had grown out distal to the amputation plane. As previously described, this plane is easily seen under transmitted light (LeClair and Topczewski, 2010). After trimming, the desired fragments were collected with fine forceps, blotted on a Kimwipe to remove excess fluid, and transferred to 250 μl RNA isolation reagent (RNAzol; Molecular Research Center, Cincinnati, OH) at the bottom of a 1.5-ml RNase-free Eppendorf tube. The bottom of the tube was then snapfrozen in a bath of liquid nitrogen and stored at −80°C until all tissues had been collected. As a positive control for zebrafish myelin, we also collected whole larval zebrafish at 5 days postfertilization (5 dpf). Approximately 50 anesthetized embryos were pipetted into a 1.5-ml tube, and the supernatant was removed. Two hundred fifty microliters of fresh RNAzol was then added, and the tube was snap-frozen as described above. After all of the tissues had been collected and frozen, each 250-μl aliquot of RNAzol was individually thawed until it was just liquid, then ground on wet ice with a motorized RNase-free micropestle (Kontes Pellet Pestle). The final volume in each tube was adjusted upward to 1 ml either by adding fresh RNAzol or by combining up to four 250-μl aliquots of similar homogenates. Homogenates were stored at −20°C for up to several weeks, then processed for total RNA isolation according to the manufacturer’s protocol. RNA pellets were resuspended in nuclease-free water, checked for concentration and purity on a Nanodrop 2000c, and stored in aliquots at −80°C. On average, 1-2 μg total RNA could be obtained from the distal tips of ~30 regenerating barbels, compared with 3-4 μg from the intact contralateral controls from the same animals.

Reverse-transcription polymerase chain reaction

Gene-specific primers were designed to amplify zebrafish myelin basic protein (mbp), sox10, pou3f1 (=oct6), egr2a (=krox20a), and egr2b (=krox20b). Primers for mbp were taken from Buckley et al. (2010). Primers for other genes were designed with NCBI Primer BLAST and/or Primer 3. To avoid false-positive results from genomic contamination, we designed intron-spanning or exon-junction primers when possible. However, because one of our target genes (pou3f1) lacked introns, all RNA samples were additionally treated with DNase (DNA-Away kit; Ambion) prior to amplification. A list of all primer combinations, with expected fragment sizes, is given in Table 1.

Primers were ordered from Integrated DNA Technologies (Coralville, IA) and resuspended in Tris-EDTA (pH 7.5) to a stock concentration of 10 μM. RT-PCR was performed in a single-tube format (One-Step RT-PCR; Invitrogen, Carlsbad, CA) in an Applied Biosystems (Foster City, CA) 2720 Thermal Cycler. Each 10-μl reaction included 50-100 ng total RNA, 0.2 μl each of the forward and reverse primers, 5 μl of the RT-PCR master mix, and nuclease-free water. Amplification followed the kit’s recommended cycling parameters, with adjustments for the annealing temperature (T_m) of each primer pair. We used touchdown PCR for the first 20 cycles, decreasing annealing temperatures from T_m +8°C to T_m −2°C. The last 20 cycles were carried out at the final annealing temperature (T_m −2°C). After amplification, the entire reaction volume was run on a 1.5% agarose gel in lithium-boric acid (LB) buffer (Brody et al., 2004) with ethidium bromide. Images were obtained on an AlphaInnotech FluorChem HD2 gel station and saved as TIFF files.

RT-PCR specificity of the intron-spanning primers was confirmed by two methods. DNA-only reactions used the same reaction conditions but 50-100 ng of zebrafish genomic DNA instead of total RNA as a template. RT-minus reactions amplified the same amount of total RNA using PCR reagents alone (JumpStart RedTaq PCR Reagent; Invitrogen), without reverse transcription. For our intron-spanning primers, DNA-only reactions showed the larger bands expected from genomic DNA contamination; RT-minus reactions for these primers showed no products (not shown).

Glycol methacrylate sectioning and staining

Freshly dissected adult barbel tissues were fixed in 4% paraformaldehyde-phosphate-buffered saline (PF-PBS; pH 7.6) at 4°C overnight or at room temperature for 1-2 hours. After several rinses in PBS, tissues were dehydrated gradually over several hours to 70% ethanol and stored in the cold until use. For embedding, the specimens were dehydrated to 100% ethanol and infiltrated with cold-polymerizing glycol methacrylate (GMA; Technovit 7100; Electron Microscopy Sciences, Warrington, PA). Precise orientation of the specimens was obtained by pre-embedding in smaller blocks of resin (De Smet et al., 2004; Leroux et al., 2007). Infiltrated barbels were laid horizontally in a small amount of GMA embedding solution in shallow rectangular silicone molds, covered with Parafilm to exclude air, and polymerized overnight at room temperature. These blocks were then embedded with the long axis vertical in a 1.5-ml Eppendorf tube filled with additional GMA, followed by a second overnight polymerization.

For sectioning, the Eppendorf tube filled with solidified GMA was clamped in a vice, and the distal 4-5 mm of the plastic was trimmed off with a high-speed rotary tool (Dremel). Under magnification, the polymerized GMA was hand-trimmed with a razor blade so that the block face approached the specimen. The tube was then clamped in the chuck of an American Optical 820 rotary microtome, and 3-4-μm sections were cut on a tungsten-carbide blade. Sections were collected singly, floated on a room-temperature bath of distilled water, and mounted in groups of 10-12 on cleaned, coated glass slides (Color-frost Plus; Fisher Scientific, Fair Lawn, NJ). The freshly mounted sections were dried on a 60°C hot plate for 2 hours and stored desiccated at 4°C until use.

Immunohistochemistry on whole mounts and sections

Primary antibody characterization

The mouse monoclonal anti-tubulin antibody was obtained from Sigma (T7451) and has been extensively characterized (JCN Antibody Database). It was used at 1:1,000-1:2,000 for this study. The other primary antibody was a polyclonal rabbit anti-zebrafish myelin basic protein (anti-MBP) kindly donated by Drs. Clare Buckley and William Talbot. This reagent has the ZFIN antibody identifier ZDB-ATB-111104-1 and was used at 1:50-100 for this study. The antibody was generated commercially (Sigma Genosys), the immunogen being a zebrafish-specific MBP peptide (CSRSRSPPKRWSTIF) conjugated to keyhole limpet hemocyanin (KLH) via m-maleimidobenzoyl-N-hydroxysuccinimide (MBS; Lyons et al., 2005). To validate the tissue visualization of myelin in this study, we tested this antibody by whole-mount staining of 4-5-days-postfertilization zebrafish embryos; these yielded staining patterns consistent with known myelinated tracts in the embryonic zebrafish CNS (not shown). We additionally tested this antibody on GMA sections of adult zebrafish optic tectum fixed and stained as described above. These sections showed expected staining of known myelinated regions (see Fig. 3C,D). All antimyelin staining patterns detected for barbel tissue were closely associated with the tracts of adjacent axons, as detected by anti-tubulin staining. The only MBP staining not closely associated with axons was seen in some cells of the epithelial taste buds, possibly the “type I” cells (glial-type cells) of this organ (see Fig. 5G).

Transmission electron microcopy

Regenerating maxillary barbels and their contralateral normal controls were fixed by immersion in cold 2.5% paraformaldehyde/2.5% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4 (Electron Microscopy Sciences) for 3-4 days in a 4°C refrigerator. After three rinses in cacodylate buffer, they were transferred to 2% osmium tetroxide in 0.1 M sodium cacodylate for 1.5 hours, rinsed three times with distilled water, then stained en bloc with 3% uranyl acetate for 1 hour. Samples were then dehydrated in a graded ethanol series (50-100%), embedded in epoxy resin, and cut on a Leica UC6 ultramicrotome. For light microscopy, 1-μm-thick sections were stained with toluidine blue O. For transmission electron microscopy (TEM), 70-nm sections were mounted on 200-mesh copper grids and poststained with uranyl acetate and Reynolds lead citrate. Sections were examined in a JEOL JEM 1220 and/or a FEI Tecnai Spirit G2 transmission electron microscope.

Image processing and analysis

Images were adjusted in Adobe Photoshop for brightness and contrast. To obtain complete high-resolution cross-sections of the maxillary barbel, multiple transmission electron micrographs were assembled into composites. Each frame carried its own embedded micrometer scale, ensuring that all images were kept at the same magnification. To compare normal and regenerated axons quantitatively in these images, we used Image J (Rasband, 1997-2012) to measure all visible ventral axons in 3 normal and 3 regenerated maxillary barbels (28 days postsurgery), respectively. Axon areas were measured by using the freehand tool to outline the light gray portion of the cytoplasm within the axoplasm membrane. Nonmyelinated and myelinated axons were distinguished by inspection. For myelinated axons, myelin thickness was measured by drawing a line segment through a region of compact, well-preserved myelin for each axon, typically at the 12 o’clock position. Axon diameter was back-calculated from axon area, assuming that each axon was a perfect circle:

Fiber diameter was calculated as (axon diameter + [2 myelin thickness]). g-ratio, a measure of myelin thickness with respect to axon size, was calculated by dividing axon diameter by fiber diameter. Two observers (A.K.H., E.E.L.) performed the digitization, obtaining similar results. Graphing and statistical analyses were performed in GraphPad Prism. Data distributions were assessed for normality by using the D’Agostino and Pearson omnibus normality test. Percentages of myelinated axons were compared by using the two-sample, two-tailed z-test of proportions. Differences in central tendency were tested by using the nonparametric equivalent of the two-tailed, two-sample t-test (Mann-Whitney U). Equality of continuous distributions was tested by using the two-sample, two-tailed Komolgorov-Smirnov test (http://www.physics.csbsju.edu/stats/KS-test.n.plot_form.html). To correct for multiple, nonindependent tests while maintaining an overall type I error rate of approximately 5%, we adjusted our alpha value downward by using the Bonferroni correction, namely, (0.05/8 tests performed) = P < 0.006 to declare statistical significance for each test (Sokal and Rohlf, 2011).

Maxillary barbel axons express mbp and are wrapped in thick myelin sheaths

To determine whether maxillary barbel nerves were myelinated, we used RT-PCR to detect transcripts of mbp. mbp is strongly expressed in the zebrafish CNS and PNS from embryonic day 5 onward, making it a common molecular marker of myelinating activity in the zebrafish nervous system (Buckley et al., 2010; Pogoda et al., 2006). Two previously identified alternatively spliced mbp transcripts were detected in both the normal adult barbel shaft and the distal tips of regenerating adult barbels at 10 and 14 days postsurgery (Fig. 2).

To examine the distribution of myelin in normal maxillary barbel neurons, we prepared thin (1-4 μm) plastic sections of adult tissue for light microscopy and immunohistochemistry. Slides stained with toluidine blue O revealed many small (1-2 μm), dark rings in the dorsal and ventral neural regions (Fig. 3A, dn and vn), suggesting abundant myelination. To confirm that these areas contained myelin, we probed tissue sections with a zebrafish-specific anti-MBP primary antibody and a fluorescent secondary. The result was a high-contrast pattern of minute fluorescent punctae clustered within the axon-rich ventral region (Fig. 3B). Smaller clusters of punctae were observed in the dorsal nerve region and at the bases of epithelial taste buds. The best resolution of myelin in situ was achieved by preparing ultrathin sections for TEM. These samples confirmed that most axons of the maxillary barbel are indeed thickly myelinated, with numerous tightly packed lamellae (Fig. 4).

Early barbel buds are innervated but lack myelination; myelination proceeds from proximal to distal and from ventral to dorsal to establish the adult pattern

Having detected abundant myelination in the adult maxillary barbel, we sought to discover when this myelination becomes established during development. To examine the early stages of barbel development, we used whole-mount immunohistochemical techniques to stain axons and/or myelin in a series of juvenile zebrafish.

As previously reported (LeClair and Topczewski, 2010), the maxillary barbel bud is innervated from its very first extension with a central tract of fibers descending from the facial nerve (cranial nerve VII). Antiacetylated tubulin staining confirmed this pattern (Fig. 5A-F); however, at early juvenile stages (<0.3 mm barbel length), myelin within the barbel bud was undetectable (not shown). Only barbels that had grown past the edge of the maxilla, protruding several hundred micrometers (>0.5 mm barbel length), showed proximal myelination in the large (ventral) nerve tract (Fig. 5G,H). Myelin staining was weak or absent distally. The ventral nerves, which invade first, were myelinated prior to the dorsal nerves, which invade later (not shown).

Although developmental myelination of the maxillary barbel nerves lagged behind the extension of axons, adult maxillary barbels stained with the same antibodies showed abundant myelination in most areas where axons occur (Fig. 6). In whole mounts, almost all fibers in the ventral and dorsal nerve tracts appeared myelinated along most of their length. However, individual nerve fibers within these tracts, small fibers crossing between the dorsal and the ventral tracts, or sections of otherwise myelinated nerves could be unmyelinated.

Early-regenerating axons are nonmyelinated or have disorganized myelin, and collagenous matrix and fibroblasts predominate

We next wanted to know whether the adult pattern of extensive myelination was restored during maxillary barbel regeneration and, if so, in what period of time. As previously shown, regenerating maxillary barbels have a normal epithelium but an abnormally thick, matrix-rich core that resembles scar tissue (LeClair and Topczewski, 2010). We thus speculated that this matrix might be a barrier to regenerating axons and/or migrating Schwann cells. To study axon regeneration in this context, we collected regenerating maxillary barbels at 10 days postsurgery and prepared TEM sections just distal to the amputation plane (Fig. 7). The following observations are based on five or six sections per barbel from eight barbels collected at this time point.

Unlike a tightly organized normal barbel (see, e.g., Fig. 3A), early-regenerating barbel tissue appeared patchy, with many voids or holes in both the ectodermal and the mesodermal layers (Fig. 7A). Under TEM, these electrolucent voids were filled with loose collagen fibers, microtubules, and/or fragmentary cell membranes (Fig. 7B,C). There were no large, intact axons. A few small intact axons were observed, most of which were unmyelinated. The most myelin observed was eight to 10 wraps of loose myelin around a disorganized core (Fig. 7D). Thick, dark sheaths of myelin were not present. Schwann cells were not evident, nor were there any obvious inflammatory cells. The most abundant cells in the mesodermal compartment at this stage were large, irregular cells full of rough endoplasmic reticulum and Golgi (Fig. 7E). This is typical of fibroblasts, a major source of tissue collagen. Consistent with this interpretation, these fibroblastic cells surrounded and in some cases were cytoplasmically continuous with extracellular collagenous fibers, suggesting cellular secretion and/or absorption. Numerous collagen bundles were observed in cross, oblique, and longitudinal sections (Fig. 7F), indicating a mixed or random orientation of fibers in the regenerating tissue.

Late-regenerating axons are altered in position, but axon morphology and myelination are restored

Because few intact axons were observed in distal barbel tissue at 10 days postsurgery, we repeated our surgical experiments, this time allowing the tissues to regenerate for 28 or 70 days, respectively (Fig. 8A). Results from both time points were similar. For brevity, only the 28-day results are shown.

Regenerating barbels at 28 days appeared similar to the original appendages, having grown out several millimeters (see, e.g., Fig. 1B). In contrast to the patchy, fluid-filled tissue seen at 10 days postsurgery, these appendages had a largely normal histology. Most features of the normal adult barbel were restored, including a thick, glandular epithelium with numerous taste buds and goblet cells; a simple vascular system of two or three endothelially lined fluid-filled tubes; and a large, centrally located nerve bundle (Fig. 8B,C). Consistent with our previous descriptions based on paraffin histology, 100% of the regenerated barbel cores had a disorganized mesodermal matrix, which was expanded into a ring around the central soft tissues. Unlike the loose collagen bundles surrounded by fibroblasts, this matrix was dense and largely acellular; occasional nucleated cells were embedded within. In regenerated barbels, the major nerve trunks were also repositioned. Unlike the case in normal barbels, we consistently observed one large bundle of axons ventrally and few or no axons dorsally.

Higher magnification of the ventral bundle showed six to eight smaller nerve fascicles, each with 10-20 intact axons (Fig. 8D). Cross-sections through these axons showed well-organized electrolucent axoplasms packed with abundant microtubules. Most of the regenerated axons were thickly myelinated, in close association with Schwann cells (Fig. 8E).

Regenerated axons are similar to normal axons in fiber type and caliber but have slightly reduced myelination

To compare normal and regenerated axons further, we digitized all visible ventral axons from complete cross-sections of three normal and three regenerated (28 days postsurgery) adult maxillary barbels. Nerves outside of the dermal region, such as at the bases of the tastebuds, or within the epidermis, were not examined. Over 900 axons were measured for axoplasm area (in μm²), and, if the axons were myelinated, we measured the thickness of the corresponding myelin layer (in μm). The absolute difference in axon number was not considered important; adult fish were not strictly size matched, and larger individuals, such as females, would be expected to have more axons (Bailey et al., 2007). Axon number would also be expected to vary proximodistally along the appendage; small variations in section placement, therefore, might sample different numbers of axons. Rather, we sought to compare percentages (for categorical variables) or distributions (for continuous variables) between all normal and all regenerated axons, respectively.

Normal adult barbels had between 122 and 223 ventral axons, an almost twofold difference between individual samples (Table 2). Despite this range of axon number, the proportion of myelinated fibers in each normal appendage was much less variable, between 59% and 65%. A test of proportions on the pooled, within-group data showed that the percentage of myelinated fibers did not differ between normal and regenerated tissue (316/499 = 63.4%_normal vs. 306/483 = 63.3%_regenerated; two-sample, two-tailed z-test = 0.76, P = 0.93).

In both normal and regenerating appendages, myelinated axon cross-sectional areas were strongly skewed rightward, indicating many small but few large axons of this subtype (Fig. 9A,B; D’Agostino-Pearson omnibus normality test, P < 0.0001). Most myelinated axons were less than 2 μm². Unmyelinated axons were smaller in caliber, less than 1 μm², and were also nonnormally distributed (P < 0.0001). There were no large, unmyelinated axons. The calibers of myelinated and unmyelinated fibers were overlapping but not identical. Although they were smaller, all unmyelinated axons were within the size range of myelinated axons.

The median cross-sectional areas of myelinated and unmyelinated axons were similar between normal and regenerated tissue, indicating that, after 28 days, the typical caliber of each type of axon was restored (Mann-Whitney U; myelinated_{normal vs. regenerated}, 0.72 vs. 0.75 μm², U = 44788, P = 0.11; unmyelinated_{normal vs. regenerated}, 0.22 vs. 0.27 μm², U = 14719, P = 0.13). The cross-sectional areas of each fiber type were also not significantly different when compared by using a highly sensitive test for equality of continuous distributions (Komolgorov-Smirnov; myelinated axon areas_{normal vs. regenerated}, D = 0.09, P = 0.14; unmyelinated axon areas_{normal vs. regenerated}, D = 0.11, P = 0.19). A cumulative frequency plot comparing the crosssectional areas of normal and regenerated myelinated axons confirmed almost identical curves along the same x-axis scale (Fig. 9E).

Median myelin thickness was slightly less in regenerated tissue relative to controls, a highly significant result (Fig. 9C vs. D; 0.11 vs. 0.13 μm, Mann-Whitney U = 34,567; P < 0.0001). A significant difference was also detected when the two groups were compared for equality of continuous distributions (Komolgorov-Smirnov; D = 0.25, P = < 0.0001). This difference was further confirmed by a cumulative frequency plot of the two distributions, showing that the myelinated axons in regenerated maxillary barbels have thinner myelin in almost all size classes (Fig. 9F). Proportionally, these regenerated axons have myelin sheaths about 15% thinner, on average, than normal axons, the absolute between-group median difference being approximately 0.02 μm (20 nm). Stated another way, regenerated maxillary barbel axons restore, after 28 days, approximately 85% of normal myelin levels.

Another common measure of myelination is the g-ratio, or the inner axonal diameter divided by the outer axonal (“fiber”) diameter. Lower g-ratios indicate thicker myelin. The g-ratios for normal and regenerated axons were both negatively skewed (D’Agostino and Pearson omnibus normality test, P < 0.0001), indicating a small number of “hypermyelinated” axons in each population. Unsurprisingly, given the slightly thinner myelin detected in regenerated axons, the g-ratios of these regenerated fibers were slightly higher relative to controls (Fig. 10A vs. B; Mann Whitney U, g-ratio_normal = 0.79 vs. g-ratio_regenerated = 0.83; U = 32,530, P < 0.0001). There was, however, considerable variation both within and among groups. Notably, the g-ratio range for normal axons was quite large (0.55-0.91), and almost identical to the g-ratio range for regenerated axons (0.57-0.96). Scatterplots of g-ratio vs. axon diameter also revealed considerable overlap between the axons in the two groups (Fig. 10C vs. D). There was a similar, slightly curvilinear relationship in each plot, suggesting that smaller axons tend to be more thickly myelinated than larger ones. This relationship was quite noisy, however, particularly in normal fibers, for which a larger scatter was observed.

Regenerating maxillary barbels express multiple promyelinating transcription factors

To confirm that the myelin sheaths that we observed in regenerating barbels represented de novo myelin synthesis, we isolated total RNA from distal barbel tissue 28 days postsurgery (Fig. 11A) and used RT-PCR to screen for multiple promyelinating transcription factors. This analysis strongly amplified all four targeted transcription factors, sox10, pou3f1, and egr2a/b, from distal regenerating barbel tissue (Fig. 11B). Positive (RT⁺) and negative (no RNA) controls for a typical housekeeping gene (β-actin) gave expected results of one band and no band, respectively. The entire experiment was repeated on a second, independently isolated pool of regenerating barbel tissue, producing similar results (not shown).