All experiments were conducted in either C57BL/6J (stock #000664, B6) background or hemizygous male and homozygous Fmr1 knock-out strain (B6.129P2-Fmr1^{TM 1Cgr}/J stock #003025, Fmr1 KO) obtained from The Jackson Laboratory (Bar Harbor, ME, United States) (The Dutch-Belgian Fragile X Consorthium et al., 1994). Animals from both sexes were used in the experiments for both B6 and Fmr1 KO mice. There were no significant differences in any of the measures based on sex, so data for both sexes are combined for analysis (p-values shown are the main effect of sex; CARS p = 0.7454, OL count p = 0.9529, OPC count p = 0.2103, EM data was one animal of each sex so differences do not apply here). Exact number of animals used are listed in the figure legends for each experiment but ranged from 4–7 for each genotype and were between P72-P167 (postnatal day).

Mice were overdosed with pentobarbital (120 mg/kg body weight) and transcardially perfused with phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 1.76 mM KH₂PO₄, 10 mM Na₂HPO₄ Sigma-Aldrich, St. Louis, MO, United States) followed by 4% paraformaldehyde (PFA). After perfusion, the animals were decapitated, and the brains removed from the skull. Brains were kept overnight in 4% PFA before transferring to PBS. Brainstems were embedded in 4% agarose (in PBS) and sliced coronally using a Vibratome (Leica VT1000s, Nussloch, Germany) at either 200 μm thickness for myelination analysis with CARS or 70 μm thickness for labeling oligodendrocytes.

For EM imaging, mice were perfused with PBS followed by a solution containing 2.5% glutaraldehyde, 2% paraformaldehyde, and 0.1 M cacodylate buffer at pH 7.4. The brains were stored in the same solution for 24 h and sectioned at 500 μm using the same protocol as above. The sections were immersed in glutaraldehyde solution for a minimum of 24 h at 4°C. The tissue was then processed by rinsing in 100 mM cacodylate buffer followed by immersion in 1% osmium and 1.5% potassium ferrocyanide for 15 min. Next, the tissue was rinsed five times in the cacodylate buffer and immersed again in 1% osmium for 1.5 h. After rinsing five times for 2 min each in cacodylate buffer and two times briefly in water, en bloc staining with 2% uranyl acetate was done for at least 1 h at 4°C, followed by three rinses in water. The tissue was transferred to graded ethanols (50, 70, 90, and 100%) for 15 min each and then finally to propylene oxide at room temperature, after which it was embedded in LX112 and cured for 48 h at 60°C in an oven. Ultra-thin parasagittal sections (55 nm) were cut on a Reichert Ultracut S (Leica Microsystems, Wetzlar, Germany) from a small trapezoid positioned over the tissue and were picked up on Formvar-coated slot grids (EMS, Hatfield, PA, United States).

For staining oligodendrocytes, six to eight free-floating sections from each brain were submerged in L.A.B. solution (Liberate Antibody Binding Solution, Polysciences, Inc., Warrington, PA, United States, Cat No. 24310) for ten min to help expose epitopes related to labeling mature and precursor oligodendrocytes. Next, the slices were washed two times in PBS and blocked in a solution containing 0.3% Triton-X (blocking solution), 5% normal goat serum (NGS) and PBS for 1 h on a laboratory shaker. After blocking, slices were stained with primary antibodies rabbit anti-Aspartoacylase (GeneTex, Irvine, CA, United States; 1:1,000) and mouse Sox-10 (A-2) (Santa Cruz, Dallas, TX, United States; sc-365692, 1:500) diluted in blocking solution with 1% NGS and incubated overnight (Table 1). Slices were then washed three times (10 min each wash) in PBS and incubated in secondary antibodies (Table 2) diluted in blocking solution with 1% NGS for 2 h. After three additional washes in PBS (5 min each wash), Nissl (Neurotrace 425/435 Blue-Fluorescent Nissl Stain, Invitrogen, Carlsbad, CA, United States, 1:100) in antibody media [AB media: 0.1 M phosphate buffer (PB: 50 mM KH₂PO₄, 150 mM Na₂HPO₄), 150 mM NaCL, 3 mM Triton-X, 1% bovine serum albumin (BSA)] was applied for thirty min. Stained slices were then briefly washed in PBS and slide-mounted with Fluoromount-G (Southern Biotech, Cat.-No.: 0100-01, Birmingham, AL, United States). Slides were stored at 4°C. Slices used for myelination analysis with CARS (four to five slices per brain) were stained with Nissl as described above, then stored free-floating at 4°C until imaged. All antibody labeling was performed at room temperature.

The primary antibody for mature oligodendrocytes (Aspartoacylase (ASPA), 1:1,000, GeneTex, Irvine, CA, United States; GTX113389; RRID:AB_10727411, Table 1) is a rabbit polyclonal antibody specific to mature oligodendrocytes (OLs). The ASPA antibody is specific to a recombinant protein encompassing a sequence within the center region of the human aspartoacylase. Aspartoacylase is responsible for the conversion of N-acetyl_L-aspartic acid (NAA) to aspartate and acetate. Hydrolysis by aspartoacylase is thought to help maintain white matter (Bitto et al., 2007). The ASPA antibody has been shown to label mature oligodendrocytes in mouse cerebral cortex and has been validated by protein overexpression and western blot analysis (Larson et al., 2018; Orthmann-Murphy et al., 2020). Additional western blot analyses with various whole cell extracts also showed that ASPA antibody detects aspartoacylase protein. The primary antibody Sox-10 [Sox-10 (A-2), 1:500, Santa Cruz Biotechnology, Heidelberg, Germany; sc-365692; RRID:AB_10844002] was used to label the entire oligodendrocyte lineage [oligodendrocyte precursor cells (OPCs) and OLs]. Sox-10 is a mouse monoclonal antibody specific to an epitope mapping between amino acids 2–29 at the N-terminus of the human Sox-10 gene. Sox-10 has been shown to specifically label the oligodendrocyte lineage in the CNS and in mouse brains (Zuo et al., 2018; Barak et al., 2019; Imamura et al., 2020). Both primary antibodies were visualized using two fluorescent-conjugated secondary antibodies which are listed in Table 2.

Brainstem slides for immunofluorescence were imaged using an Olympus FV1000 confocal microscope (Olympus, Tokyo, Japan) with lasers for 488, 543, and 635 nm imaging. Once MNTB was identified by the distinctive trapezoidal morphology and cell size, z-stacks were taken using a 20x objective (UPLSAPO20X, NA 0.75) so that the entire nucleus could be visualized and quantified. The nucleus was separated into medial, central, and lateral regions following the protocol in Weatherstone et al., 2016. Briefly, the MNTB was digitally extracted using FIJI software, and the tonotopic axis was estimated by drawing the longest possible dorsomedial-to-ventrolateral line through the nucleus (Weatherstone et al., 2016). This line was measured using FIJI and divided into thirds wherein perpendicular lines were drawn. These lines delineated the lateral, central and medial regions of MNTB. For CARS microscopy, brainstem sections were imaged using an Olympus FV-1000 (Olympus Corp., Tokyo, Japan) fitted with non-descanned detectors in both the forward and epi CARS directions. Sections were placed in a culture dish with coverslip (for inverted microscopy) and custom weight to keep tissue near coverslip. Z-stacks were taken with a 60X, 1.2NA infrared corrected water objective which served for collection of the CARS signal in the epi direction for medial and lateral MNTB. Signal in the forward direction was collected through an Olympus 0.55NA condenser. The APE picoEmerald laser set up contained an NKT fiber laser which provided the 80 MHz clock, and an OPO (Optical Parametric Oscillator) laser with a tunable range of 770–990 nm. Water absorption peak was at 1,388 cm^–1 which implies a 901.1 nm pump/probe beam with the fixed Stokes beam at 1,031 nm and a CARS signal at 801.6 nm. To excite the CH₂ stretch at 2,845 cm^–1 the APE picoEmerald lasers required a pump/probe beam set to 797.2 nm which resulted in a CARS signal at 649.8 nm, thus laser settings were always set to 797.2 nm prior to imaging. Pulse duration for the two lasers was approximately 2 ps and the polarization was linear and horizontal. The synchronization was primarily based on the OPO cavity length with the pulse pumped by the frequency doubled output from the NKT laser at 515.5 nm. The tunable pump/probe beam was used for the TPEF (two-photon excitation fluorescence), which was separated by a dichroic from the CARS signal. The NDD (non-descanned detector) epi-direction unit contained two PMTs of the same type. The CARS and fluorescence do not share PMTs. These settings were used for optimal lipid signal to selectively image myelination in this area.

As an additional control, one brain from a wild type mouse and one brain from a knockout were processed and imaged with electron microscopy. The main purpose of this control was to verify that CARS microscopy has sufficient resolution to discriminate between the small axon diameter differences that are studied here. Sections for EM were first imaged on a compound microscope, and MNTB was identified by the characteristic cell size and shape, as well as the location of the cells in the parasagittal slices. Once MNTB location was determined, slices were imaged on a FEI Tecnai G2 transmission electron microscope (Hillsboro, OR, United States) with an AMT digital camera (Woburn, MA, United States).

Quantification of OL’s and entire OL lineage [oligodendrocyte precursor cells (OPCs) and OLs] was performed using the optical fractionator approach and FIJI software (Schindelin et al., 2012). In a pilot study to determine the appropriate stereological parameters, five brains were sliced at 50 μm using a freezing microtome (SM2010R Leica, Nussloch, Germany) and z-stacks from five sections per brain (every other section) were taken with a 20x objective the same as above. Using FIJI each stack was brightened, the background subtracted, the image scaled and the MNTB was digitally extracted. Separate grid plugins were used to create the dissector and count frames for each image stack. To account for tissue shrinkage, the section thickness was measured at multiple points within MNTB using the Olympus FV1000 microscope and FV10-ASW microscopy software. Measurements were taken by recording the first z-level where tissue features came into focus, and then recording the last z-level where tissue features were in focus. The difference between the recorded levels for each measured area of MNTB was averaged and represented the section thickness for stereological calculations. Stacks were set to contain 20 slices regardless of the section thickness and the step size was kept constant within, but not between brains. Cells were counted by moving through the stack and using a cell counter plugin to mark the top of the cell when it first came into focus within the probe unless it contacted the exclusion lines of the count frame (Schmitz and Hof, 2005).

The pilot data indicated that selecting four 70 μm sections (every other section starting from a randomly selected slice) to give a section subfraction (SSF) of 0.280 was appropriate for calculating the total number of cells in a single MNTB. In addition, counts would use a probe area of (1,250 μm × 1,250 μm) and dissector (5,000 μm × 5,000 μm) giving an area subfraction (ASF) of 0.25. It should be noted that two brains had fewer than four slices (3.5 and three slices, respectively) which were suitable for quantification. OL analyses also included three brains from the pilot study that had adequate immunohistochemical staining for quantification. In these cases, the SSF was adjusted, or a fifth slice was counted to achieve an SSF that was closest to 0.280 (0.20–0.25). The height subfraction (HSF) was calculated separately for each brain using the mean dissector height (OL mean = 19.5–34.4 μm; OPC mean = 21.06–34.4 μm) and the mean section thickness (OL mean = 26.75–45.75 μm; OPC mean = 27.75–45.75 μm). Using these parameters, OL’s and OPC’s were counted and recorded in excel. OPC’s were identified as cells that were labeled with Sox 10, but not co-labeled with ASPA (Figure 1). Calculation of the total cell count for each complete MNTB nucleus was done by multiplying the counted cells (ΣQ^–) with the reciprocal sampling fractions such that:

Error in the stereological approach was calculated using the Gundersen-Jensen coefficient of error (CE) estimator and smoothness class of m = 1 (Gundersen et al., 1999). CE measures error as a function of the biological variance (noise) and the sampling variance from systemic uniformly random sampling. The coefficient of variance (CV) within and between brains was high (CV > 1) due to natural variations in the population of glial cells throughout MNTB. CE for OL and OPC estimates was greater than 10% (Mean CE = 0.51), but the sampling parameters chosen showed CE was negligible in terms of overall variation (CE²/CV² < 0.003) (Gundersen et al., 1999). Thus, the stereological parameters were determined to be sufficient for quantification estimates.

Analysis of myelination morphology was done using FIJI software. Myelin diameter was measured from CARS images by using the line tool to measure the fiber diameter (Figure 2A) for 20 axons per image. The resolution was not sufficient in CARS images to quantify thickness or g-ratio of axons. Tissue shrinkage was not corrected for in this analysis. In addition, diameters of 240 axons were also measured in the same way in EM images, however, myelin thickness was measured directly, and g-ratios were calculated from fiber and axon diameters (Figure 2B; Ford et al., 2015). G-ratios were calculated from EM images using inner and outer fiber perimeter ratio (g-ratio = inner fiber perimeter/outer fiber perimeter). EM data is representative of one animal from each genotype and was performed primarily to confirm CARS diameter changes.

Figures were generated in R (R Core Team, 2013) using ggplot2 (Wickham, 2016). Boxplots dislay the median and 25th–75th percentiles (or 1st and 3rd quartiles, respectively) the whiskers represent ± 1.5 times the interquartile range. Data that falls outside the range are plotted as individual points. Data for all experiments were analyzed using linear mixed-effects models [lme4; (Bates et al., 2014)]. To account for repeated measures and variability with animals, Animal was considered a random effect with fixed effects of genotype and location (medial, center, lateral, and total when appropriate). Additional analyses used fixed effects of age or sex to determine if these were contributing factors in the differences shown. It was expected that there may be differences between different regions of the MNTB (medial, center, lateral, total) therefore a priori, it was determined that estimated marginal means [emmeans; (Lenth, 2019)] would be used for pairwise comparisons between region and genotype. To control for multiple comparisons, emmeans implements a Tukey method for these contrasts. Asterisks are used to indicate statistical significance between the two genotypes, as follows: ^∗p < 0.05, ^∗∗p < 0.01, and ^∗∗∗p < 0.001. Figures were prepared for publication using Photoshop and Illustrator (Adobe, San Jose, CA, United States).

We first examined the diameter of axons in the medial and lateral MNTB using CARS microscopy (Figure 3). CARS microscopy for myelination imaging was used over electron microscopy (EM) due to its speed and compatibility with immunofluorescence (Wang et al., 2005), in this case we used a Nissl stain for simultaneous cell body staining. We measured axon fiber diameters in both medial and lateral MNTB because previous work has shown the potential for differences in myelin based on tonotopic location (medial or lateral) (Ford et al., 2015; Stange-Marten et al., 2017). Representative images for both location (medial or lateral columns) for B6 and Fmr1 KO (rows) are shown in Figures 3A–D. We found that there was a significant decrease in fiber diameter in Fmr1 KO compared to B6 for both medial and lateral MNTB (Figure 3E). Not surprisingly when data were summed between medial and lateral locations, there was still a significant decrease in fiber diameter in Fmr1 KO mice.

To confirm that the CARS technology was sensitive enough to determine differences between fiber diameters in Fmr1 KO and B6 mice, we also performed EM controls on one animal of each genotype (across multiple sections with multiple measurements, Figure 4). The advantage of EM, while costly and time consuming, is that it allows for higher resolution and thus more detailed description of changes to myelination than CARS microscopy techniques. Since alterations in fiber diameter seen in CARS data was not dependent on medial or lateral localization, and it is difficult to determine location with parasagittal EM sections, these data are not separated by location. Representative EM images for B6 and Fmr1 KO are shown in Figures 4A,B, respectively. Consistent with CARS analysis, EM data showed smaller fiber diameters in Fmr1 KO animals compared to B6 (Figure 4C). Additionally, axon diameter and myelin thickness are also decreased in Fmr1 KO compared to B6 (Figures 4D,E). Lastly, g-ratio, an often-used index of optimal functional and structural myelination, was increased in Fmr1 KO animals compared to B6 (Figure 4F; Chomiak and Hu, 2009). These results are consistent with the relatively large axon diameters in the Fmr1 KO mice compared to fiber diameters (a measure for g-ratio) despite thinner myelin and smaller axons than B6 mice.

Oligodendrocytes (OLs) are the glial cells responsible for myelinating axons in the central nervous system. FMRP has been shown to be expressed in OLs, potentially providing a mechanism through which loss of FMRP can lead to changes in myelination (Wang et al., 2004; Giampetruzzi et al., 2013). For OL counts, the MNTB was divided into three sections, medial, central, and lateral. Figures 5A,B show representative images of the entire MNTB stained with ASPA (a marker for mature OLs) and sox-10 (a marker for the entire OL lineage) for B6 and Fmr1 KO mice, respectively. As shown by arrowheads, OLs are represented by staining with ASPA (which also coincides with sox-10 staining). Cells that are only stained for sox-10 (magenta) and not ASPA (yellow) are quantified as OPCs (see below and Figures 1, 5). Like CARS data, there was no difference in OL number based on specific tonotopy of the MNTB (medial, center, or lateral) however, when data was combined for all regions, there was a significant increase in OL number in Fmr1 KO mice compared to B6 (Figure 5C top panel). Using the sampling Q count technique, to estimate number of cells within the entire MNTB, similar results were found to raw counts. Specifically, there was no difference in medial, center, or lateral estimates, but total combined regions showed a highly significant (p < 0.0001) increase in OLs in Fmr1 KO animals compared to B6. In addition to changes in number of mature OLs, another possible mechanism underlying alterations to myelination could be incomplete or impaired maturation of OLs.

Other brain areas (cerebellum, neocortex, and others) show changes in myelination in FXS related to impaired maturation or function of oligodendrocyte precursor cells (Wang et al., 2005; Pacey et al., 2013; Jeon et al., 2017). Like OL count measures, the MNTB was divided into medial, central, and lateral subsections (Figure 5D), with representative images of the entire MNTB shown in Figure 5 for B6 (Figure 5B) and Fmr1 KO (Figure 5A). As described above, OPCs were quantified as cells that were positive for the sox-10 (magenta) marker without ASPA (yellow) staining (Figures 5A,B arrowheads, OPC label). Like OL counts, there was no significant difference between Fmr1 KO and B6 mice based on tonotopic location (medial, center, or lateral, Figure 5D), however, when data was summed for a total MNTB count, there was a significant increase in OPC number in Fmr1 KO mice compared to B6 (Figure 5D, top panel). Consistent with data from OLs, whole nucleus Q counts were consistent with raw counts such that there was no difference between genotypes based on tonotopy, but total estimates were significantly higher in Fmr1 KO animals compared to B6 (p < 0.0014). Together, these OPC and OL data show that changes in myelination observed through EM and CARS are not due to inherent decreases in numbers of mature or immature oligodendrocytes, however, there are changes in overall number of these cells compared to wildtype animals.