Design-Based Stereology and Binary Image Histomorphometry in Nerve Assessment

Abstract

Background.

Stereology and histomorphometry are widely used by investigators to quantify nerve characteristics in normal and pathological states, including nerve injury and regeneration. While these methods of analysis are complementary, no study to date has systematically compared both approaches in peripheral nerve. This study investigated the reliability of design-based stereology versus semi-automated binary imaging histomorphometry for assessing healthy peripheral nerve characteristics.

New Method.

Stereological analysis was compared to histomorphometry with binary image analysis on uninjured sciatic nerves to determine nerve fiber number, nerve area, neural density, and fiber distribution.

Results.

Sciatic nerves were harvested bilaterally from 6 male Lewis rats, aged 8–12 weeks for comprehensive analysis of 12 nerve specimens. From each animal, sciatic nerve specimens were fixed, stained, and sectioned for analysis by light and electron microscopy. Both histomorphometry and stereological peripheral nerve analyses were performed on all specimens by two blinded and independent investigators who quantified nerve fiber count, fiber width, density, and related distribution parameters.

Comparison with existing methods.

Histomorphometry and stereological analysis provided similar outcomes in nerve fiber number and total nerve area. However, the light microscopy, but not electron microscopy, stereological analysis yielded higher nerve fiber area compared to histomorphometry or manual measurement.

Conclusion.

Both methods measure similar fiber number and overall nerve fiber area; however, stereology with light microscopy quantified less myelin area. Histomorphometry optimizes throughput and comprehensive analysis but requires user thresholding.

Graphical Abstract

1. Introduction

Reliable quantitative nerve assessment can be achieved through comprehensive, precise, and efficient analysis of axonal profiles. Advances in imaging and computing power have dramatically improved histological nerve assessment, but important questions remain regarding which approach is optimal for a specific scientific investigation. Currently, stereology and histomorphometry are the two most common methods in the field of peripheral nerve injury for characterizing nerve histology.^1–8 Unbiased design-based stereology⁹ and semi-automated binary image histomorphometry^6,10 have both been validated for nerve study; however, each method makes fundamentally different assumptions about nerve tissue anatomy. Since quantitative nerve characterization is critically important in peripheral nerve research, we performed this study to systematically compare stereology and histomorphometry for assessing axonal profiles.

Histomorphometry has long been used to comprehensively quantify peripheral nerves^11–14. Histomorphometry utilizes binary imaging to assess neuronal profile based on morphology or shape, allowing detailed study of neural tissue. It also enables complete description of cross-sectional neuronal characteristics. However, it requires the investigator to select a threshold. Thus, histomorphometry provides fine histological detail but requires some judgment in the area of tissue studied. Additionally, histomorphometry is not conducive to quantitative electron microscopy (EM) and is therefore only useful for assessing myelinated fibers by light microscopy.

Stereology, in contrast, utilizes mathematically derived models to estimate the structural features of tissues. It is used in a wide variety of biomedical applications and allows systematic, non-biased tissue sampling. Applying stereology to assess nerves is becoming more common.^9,15–17 Stereology can be performed on both light microscopy (LM) and electron microscopy (EM). Performing stereology on LM allows feasible quantification of number of nerve fibers, while stereology on EM allows assessment of nerve ultrastructure. Prior stereology efforts focused on optimization and calibration,¹⁸ probe validation¹⁸, and automated versus manual counting validation,¹⁹ but relatively little attention has been paid to its inherent strengths or limitations relative to histomorphometry. While both methods have been shown to be accurate and precise for evaluation of number of myelinated axons, little is known on the precision and accuracy of these two methods in quantifying sizes of myelinated axons. Indeed, characterizing nerve fiber size is an important aspect of studying nerve development.^20,21

To address this knowledge gap, the present study we evaluated histomorphometry and stereology in quantifying number and sizes of nerve fibers in rat sciatic nerve. We build upon our previous work of histomorphology method that tracks individual nerve fiber size, and compared it to that obtained from stereology. This work provides a robust comparison of two major methods of analysis for myelinated fibers in the peripheral nerve.

2. Material and Methods

2.1. Surgical procedures

Lewis rats (n = 6) weighing between 200–300 grams (The Jackson Laboratory, Bar Harbor, ME) were anesthetized by subcutaneous ketamine (75 mg/kg; Fort Dodge Animal Health, Fort Dodge, IA) and medetomidine (100 mg/kg; Pfizer, NY) injection. A skin _incision was made parallel and 2 mm posterior to the femur. The sciatic nerve was exposed with a muscle splitting incision followed by microsurgical dissection exposing the sciatic notch proximally and the neural trifurcation. The sciatic nerve was then resected from its origin at the sciatic notch to 5 mm proximal to the trifurcation. Nerves were then placed in fixative as outlined below, and animals were humanely euthanized. All experiments were performed according to protocols approved by the Division of Comparative Medicine at the Washington University School of Medicine and conducted in strict compliance with the National Institutes of Health Guide for the Care and Use of Laboratory animals.

2.2. Tissue preparation

Harvested nerve samples (n ₌ 6 per group) were fixed in 3% EM grade glutaraldehyde (Polysciences, Warrington, PA) at 4°C, postfixed with 1% osmium tetroxide, and serially dehydrated in ethanol. Specimens were embedded in Araldite 502 (Polysciences), and cut into 1 μm cross-sections with an ultramicrotome (LKB III Produkter A.B., Bromma, Sweden). For imaging by light microscopy, sections were stained with 1% toluidine blue dye, a reproducible method of quantitative peripheral nerve assessment, and mounted on slides.²² For analysis of LM with manual measurement, histomorphometry or stereology, the same nerve sections (1 per animal) was used. For EM, sections were stained with uranyl acetate and lead citrate and sectioned to a 90 nm thickness.

2.3. Semi-automated histomorphometry

Histomorphometry analysis was only performed on LM images. Quantitative analysis was performed with a semi-automated digital image analysis system linked to Clemex PE software adapted for nerve morphometry (Clemex Technologies Inc., Longueuil, Quebec).⁶ Clemex PE allows generation of analysis method using sequential selection of standard workflows and thus negate the need for programming. The workflow used is adapted from our previous work⁶, and is available from us in Clemex PE upon request. Briefly, sectioned nerve images were divided into 100 × 100 μm² non-overlapping frames and evaluated using eight-bit plane digital pseudo-coloring and thresholding-based algorithms to identify myelinated axons, calculate their density, area, myelin thickness, and diameters.⁶

Data were analyzed by two blinded and independent investigators to evaluate the reliability of sampling methods. Binary image analysis histomorphometry used an algorithm that counted myelinated fibers that bordered on two sides of the square counting frame and excluded them on the other two sides to adjust for nerves that straddle a border.

2.4. Stereology analyses

Quantitative analyses were performed with MicroBright Field Stereo Investigator software (MBF Bioscience StereoInvestigator version 7.0, Williston, Vermont). For electron micrographs (EM), 15 ultramicrographs were captured with a Zeiss 902 electron microscope (Zeiss Instruments, Chicago, IL) at 4360x magnification, scanned at 400 dots per inch resolution, and evaluated by stereology the sampling frame grid was 150 × 150 μm², and axon areas and diameters were calculated via the Fractionator and Nucleator protocols.²³ For light micrograph (LM) the size of the counting frame used was 1600μm². The size of the grid between counting frame was 100 × 100 μm², The methodologies enable unbiased nerve fiber quantification (Figure 1A). The user determines the number of counting frames and the results CE (Coefficient of the Error) reflects the accuracy and validation of the design setup. The CE was set to < 0.08 per prevailing design-based stereology practices⁵. The average of CE for all samples was 0.0468. For the nucleator method, four isotropic line ray from the profile center were generated and intercepts placed on myelin and axon boundaries.

Figure 1: Schematic of stereological method for quantification of myelinated axons.

A) Stereology allows for unbiased quantification of nerve fibers. In this example, any nerve on the green line would be counted, and anything on the red line would not be counted, analogous to in binary image analysis. Any fibers inside the box and not touching a line would also be counted. The number of counting frames is determined by the user, with the CE (Coefficient of the Error) of the results reflecting the accuracy and validation of the design setup. B) Stereological analysis with 2D nucleator showing rays approximation of cross-sectional area. The fractionator\ nucleator protocol was used to quantify cross sectional areas and total fiber numbers from light and electron microscopy images, marking myelin and axon boundaries. The sum of the length of the 2 rays (diameters) passing through center of the axon is divided by 4 to yield and estimate a radius, r. With assumption of approximate circularity, ╥r² is then used to estimate profile area. C) Schematic of stereology analysis – Stereology allows for quantification of 3-dimensional structures, with exclusion of counts on 3 of the six faces of each cubic region evaluated. Note that rays project beyond confines of the counting frame, with implications for exclusion of irregularly shaped axonal figures. While stereology was performed on 2-dimensional images in the present study, this depiction illustrates the rules and procedures that are applied to characterizing 3-dimensional structures. Scale bar on light microscopy image represent 50 μm. Scale bar on EM image represent 2 μm.

Profile area was calculated by π × (modeled radius)² (modeled radius is the length of a line extending to the central point of the profile), assuming a circular profile. Neural density was determined by computing the mean values of area occupied by nerve fibers per counting frame area. For EM, the sampling frame grid was set to 150 × 150 μm² and axon area and diameter calculated. Stereology assessed area, density, and myelin thickness based on rays (Figure 1B). With the nucleator method, only fibers inside a box and not touching a stair-step ray were counted/included, analogous to the approach used in morphometry (Figure 1C).

2.5. Manual evaluation

Fixed and stained nerve sections were imaged using Nanozoom at 40x objective magnification to capture the entirety of nerve histology. Images were then exported to ImageJ (NIH). The contour of nerve fibers was traced and measured using the “measurement” tool. All nerve fibers were counted.

2.6. Statistical analysis

Identical statistical procedures were used for light microscopy analysis by axonal profile stereology and histomorphometry. Data were first assessed for normality using a Kolmogorov-Smirnov test. After normality was established, analysis of variance (ANOVA) was performed to determine differences between multiple groups with post hoc comparisons by a Newman-Keuls test. Student’s t test was performed to analyze differences between 2 groups. For distribution analysis, ANOVA with multiple comparison was performed with Bonferroni to control for false discovery. Significance was established at p < 0.05. All results are reported as mean ± standard deviation (SD).

3. Results

The relevant characteristics of myelinated axons in uninjured rat nerves were characterized by three methods: histomorphometry, stereology with light microscopy, and stereology with EM (Table 1). Quantitative results are summarized for all parameters in Figures 3A-B. There were no statistically significant differences between the two methods for counting myelinated axons, or the percent of neural tissue within the sciatic nerves (Figure 2A). To further characterize myelinated axons, we measured the average size of axon, myelin, and myelinated fibers (axon + myelin). We found stereology on light micrographs determined higher values for fiber size, myelin area, and axon areas compared to histomorphometry or stereology on EM images (Figure 2B). To determine the accuracy of stereology on EM images, we measured average size of axon, myelin and fibers using stereology and compared them to manual exhaustive counting using the same EM images, and found no difference (Figure 4). To then further assess the differences in fiber size as measured by stereology on LM and histomorphometry, we compared the distribution of nerve fibers counted by either histomorphometry or stereology. We performed this comparison in two ways. First, the average and distribution of fiber area for each rat was measured using stereology, histomorphometry, and manual exhaustive measurement (Figure 3A-B). Both histomorphometry and manual measurement yielded similar results and were both significantly lower than that of stereology on LM. We also analyzed the distribution of fiber size and found an increase in proportion of smaller sized nerve fibers, and reduction in proportion of larger nerve fibers measured by histomorphometry, but no significant overall differences (Figure 4C).

Table 1:

Ranking nerve morphometry metrics assessed using Stereology vs. Histomorphometry

	Histomorphometry	Stereology (LM)	Stereology (EM)
Unmyelinated fiber #	n/a	n/a	+
Myelinated fiber #	+	+	n/a
Percent Neural tissue	+	+	n/a
Myelin area	+	+	+
Axon area	+	+	+

Figure 3. Characterizing uninjured rat sciatic nerve using stereology or manual measurement of electron micrograph.

Representative EM image of rat sciatic nerve is shown (scale bar represent 2 μm), as well as fiber area, myelin area and axon area as measured using stereology or manually. No statically significant difference was detected. N=6 for all groups. Scale bar represent 2 μm.

Figure 2. Characterizing uninjured rat sciatic nerve using histomorphometry or stereology.

A) Schematic of the techniques used to quantify the myelinated axons within sciatic nerve of rats. B) Total number of myelinated axons and % of nerve tissue within rat sciatic nerves. P values are shown, 6 animals per group. Only light microscopy stereology and histomorphometry were used. B) The size of axons, myelin, and myelinated fibers within rat sciatic nerves as measured by LM stereology, histomorphometry, and EM stereology. * represent p<0.05 compared to LM measurements. Error bar represent standard deviation. N=6 for all groups. Scale bar on light microscopy images represent 20 μm, scale bar on electron microscopy represent 2 μm.

Figure 4: Difference in measured fiber size using stereology.

A) Myelinated fiber area measured per animal using LM, histomophometry and manual measurement. Represented as average with standard deviation. B) Average myelinated fiber area as measured by LM, histomophometry and manual measurement, error bar represent standard deviation, p value shown. C) Distribution of fiber size as measured by LM stereology and histomophometry. * represent p<0.05. No significant differences between the two distributions. N=6 for all groups

4. Discussion

Interpreting neuronal profiles correctly requires a comprehensive and efficient analysis of considerable amounts of information from neuronal profiles, such as, but not limited to, axon morphology and the surrounding neural architecture.⁶ Despite technological imaging advances and increased computational power, successful analysis is still heavily influenced by the investigative tissue quality and morphology. Both analytical technologies were created and adapted in the context of developer-specific needs, which vary with different pathologies and fields in peripheral nerve research. They can differ in methodology, examined profile attributes, and limitations. While histomorphometry more rapidly measures nerve fiber profiles, stereology does not require subjective threshold settings, which may increase inter-operator variations.⁷

In this paper, we compared two methods of semi-automated nerve histology assessment: binary histomorphometry and unbiased stereology. While both methods have been used to quantify nerve histology, there has been limited data on how they compare to each other. Here, the rat sciatic nerve was evaluated using both methods. We first compared the two methods on myelinated fiber counts. We obtained similar values for the total number of myelinated axons using either method. We also compared the two methods for measuring nerve fiber size, another important characteristic of nerves.²⁴ We observed that histomorphometry yielded lower values for axon size, myelin size, and fiber size (axon + myelin). This is likely due to the assumption of unbiased stereology that nerve fibers are circular when they may be more convoluted and irregular (Figure 5). By assuming nerve fibers are circular, stereology thus likely over-estimates the size of nerve fibers.^5,25 Improvement of the reliability of stereology to that of histomorphometry would require an increased number of isotropic 4 ray lines (area), to account for the fact that peripheral nerve profiles are not circular. This would also subsequently require a substantial increase in analysis time. Stereology does, however, afford greater freedom from selection subjectivity.

Figure 5:

Normal sciatic nerve, 1000x oil immersion light microscopy of nerve section showing typical range of fibers reflected in Figure 4. Arrow points to myelinated fibers that are not spherical A) Uninjured rat nerves, B) injured rat nerves repaired with ANAs. Note the increased number of non-spherical myelinated fibers in the injured nerves. Scale bar represent 20 μm.

The requirement for manual structure identification in stereology is a potential source of bias and errors. In one study, misidentification was shown to cause variation comparable to those between biological samples, reaching as high as 11%, with most false positives being the misidentification of Schwann cell nuclei as axons¹⁸. Moreover, the amount and area of counting frames needed to obtain meaningful estimates depends on intrinsic properties of the field, which can vary between samples. The precision (CE<0.08) can only be assessed once all counts are completed. Consistent adequate framing requires experience from the researcher. Binary histomorphometry also allowed for simultaneous and more comprehensive analysis of nerve parameters without the need for different probes and multiple assessments, unlike stereology. In one assessment, the operator can directly measure multiple parameters on true histological profile (fiber count, myelin/axon/fiber area, axon/fiber diameter, fiber/axon perimeter) and provide calculated information (fiber density, myelin thickness and g-ratio), including a detailed distribution for the same nerve components.

Stereology and morphometric approaches will likely increasingly be combined in peripheral nerve studies,²⁶ partly because stereology is conducive to analyzing profiles and structures that cannot be differentiated based on thresholding alone.^27,28 For example, quantifying glial or inflammatory cell within injured nerves has become increasingly common. Histomorphometry with its linear thresholding, cannot reliable identify such cell populations. Rigorous characterization of these non-neuronal cells using stereological methods on either LM or EM can provide information not obtainable via histomorphometry. Overall, while this study current assessed healthy nerve, this analysis comparison on injured nerve and various types of pathologic nerve would be invaluable in future studies to understand the advantages of these various methodologies.

5. Conclusion

Both histomorphometry and stereology can accurately quantify myelinated axon counts. Stereology using light microscopy measured lower axon and myelin size compared to both histomorphometry and manual counting methods. Therefore, histomorphometry has potential advantages to the assessment of healthy nerve over stereology when stereology is limited to light microscopy alone.

HIGHLIGHTS.

• Stereology on LM and histomorphometry are comparable in measuring number of nerve fibers

• Stereology on EM accurately measures fiber area

• Stereology on LM over-estimates the true nerve fiber area