Stereological Methods for Estimating the Myelin Sheaths of the Myelinated Fibers in White Matter

Authors

  • Chen Li,

    1. Department of Neurology, First Affiliated Hospital, Chongqing Medical University, Chongqing, People's Republic of China
    2. Laboratory of Stem Cell and Tissue Engineering, College of Basic Medical Sciences, Chongqing Medical University, Chongqing, People's Republic of China
    Search for more papers by this author
    • Chen Li and Shu Yang are the co-first authors.

  • Shu Yang,

    1. Department of Neurology, First Affiliated Hospital, Chongqing Medical University, Chongqing, People's Republic of China
    2. Laboratory of Stem Cell and Tissue Engineering, College of Basic Medical Sciences, Chongqing Medical University, Chongqing, People's Republic of China
    Search for more papers by this author
    • Chen Li and Shu Yang are the co-first authors.

  • Lin Chen,

    1. Laboratory of Stem Cell and Tissue Engineering, College of Basic Medical Sciences, Chongqing Medical University, Chongqing, People's Republic of China
    2. Department of Histology and Embryology, Chongqing Medical University, Chongqing, People's Republic of China
    Search for more papers by this author
  • Wei Lu,

    1. Laboratory of Stem Cell and Tissue Engineering, College of Basic Medical Sciences, Chongqing Medical University, Chongqing, People's Republic of China
    2. Department of Histology and Embryology, Chongqing Medical University, Chongqing, People's Republic of China
    Search for more papers by this author
  • Xuan Qiu,

    1. Laboratory of Stem Cell and Tissue Engineering, College of Basic Medical Sciences, Chongqing Medical University, Chongqing, People's Republic of China
    2. Department of Histology and Embryology, Chongqing Medical University, Chongqing, People's Republic of China
    Search for more papers by this author
  • Hans Jorgen G. Gundersen,

    1. Stereology and Electron Microscopy Research Laboratory, MIND Center, Aarhus University, Aarhus, Denmark
    Search for more papers by this author
  • Yong Tang

    Corresponding author
    1. Department of Neurology, First Affiliated Hospital, Chongqing Medical University, Chongqing, People's Republic of China
    2. Laboratory of Stem Cell and Tissue Engineering, College of Basic Medical Sciences, Chongqing Medical University, Chongqing, People's Republic of China
    3. Department of Histology and Embryology, Chongqing Medical University, Chongqing, People's Republic of China
    • Department of Neurology, First Affiliated Hospital, Chongqing Medical University, Chongqing 400016, People's Republic of China
    Search for more papers by this author

Abstract

In the present study, efficient and unbiased stereological techniques to investigate the myelin sheaths of the myelinated fibers in rat white matter were established. In the present design, four tissue blocks were obtained from the entire white matter of rat brain in a uniform, random fashion. Isotropic, uniform random (IUR) sections were ensured by the use of the isector technique. One section with the thickness of 60 nm was cut from the center of each epon block. Eight to 10 fields of vision were randomly photographed under a transmission electron microscope. The total length of the myelinated fibers and the total volume of the myelin sheaths in the white matter were the products of the length density, volume density, and the volume of the white matter obtained with the Cavalieri principle. The mean areas of the myelinated fibers profiles and myelin sheaths were estimated with the point counting technique. The inner and outer perimeters of the myelin sheaths were estimated by the use of a line grid, and the thickness of the myelin sheaths was estimated by direct orthogonal measurements in uniform, random locations. The described methods will provide very useful tools for future quantitative studies of changes in the myelin sheaths of white matter in various experimental conditions and in various neurodegenerative diseases. Anat Rec, 2009. © 2009 Wiley-Liss, Inc.

There is significant age-related loss of the total length of the myelinated fibers in white matter (Tang et al., 1997; Yang et al., 2007). To detect the exact reason that leads to the decline of the myelinated fiber length in the white matter of aged brains, the myelin sheath ultrastructure of the myelinated fibers in white matter needs to be investigated with accurate quantitative methods. Beside the possible age-related changes of myelin sheaths, researchers have found that the changes of the myelin sheaths occurred as a result of experimental toxicity produced by Cuprizone, triethyltin, and isolecithin (Malamud and Hirano, 1973; Blakemore, 1978; Blaker et al., 1981; Chang, 1990; Ludwin, 1995; Irvine and Blakemore, 2006; Franco-Pons et al., 2007; Skripuletz et al., 2008), in patients with severe diabetes (Tamura and Parry, 1994; Myers, 1998), and in genetically engineered mice that had either an excess or a deficit of proteolipid protein (Monuki and Lemke, 1995; Anderson et al., 1998; Karim et al., 2007). It is well known that multiple sclerosis, a disease suffered by adults, is characterized by demyelination, inflammation, gliosis, and a variable loss of axons (Ludwin, 2000; Kuhlmann et al., 2008; Irvine and Blakemore, 2008). Therefore, precise methods to quantitatively investigate the myelin sheath structure in the nervous system are needed to study the changes of the myelin sheath structure in normal aging and various experimental and pathological conditions. To obtain accurate estimates of the myelin sheaths in white matter, unbiased stereological sampling and measuring principles must be applied. In the present study, unbiased stereological methods of estimating the myelin sheaths were investigated. The myelinated fibers of rat white matter were used as the example to illustrate the described methods.

MATERIALS AND METHODS

Animals

Five female Long-Evans rats (6–8-months old) obtained from Third Military Medical University, People's Republic of China were used for this study. The rats were housed three to four per cage with corncob bedding and were provided ad libitum access to food and water. They were kept under a constant 12-hour-light/12-hour-dark cycle at an ambient temperature of (22 ± 1)°C. The colony was certified specific pathogen-free of the following: mouse pneumonia virus, sendai virus, hepatitis virus, reovirus, lymphocytic choriomeningitis, Theiler murine encephalomyelitis virus, ectromelia, minute virus of rats, and mucoplasma pulmonis. Animal care and treatment followed the National Institute of Health Guide for the Care and Use of Laboratory Animals (NIH Publications No. 80-23, revised 1996).

Specimen Fixation and Sampling

The rats were anesthetized by an intraperitoneal injection of 2.5% chloral hydrate and were perfusion-fixed with a solution of 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffered saline through the left ventricle of the heart (pH 7.4). After perfusion, the cerebellum, brain stem, and cranial nerves under the pavimentum cerebri were cut, and the cerebral hemispheres were taken out. The two hemispheres were embedded in 6% agar and coronally cut into 2-mm-thick slabs, starting at the rostral pole. An average of 10 (CV = 0.04) slabs were obtained from each hemisphere (Fig. 1A,B). The right or left hemisphere was selected at random for analysis. A plastic sheet with equidistant points was placed randomly on all selected slabs of the sampled hemisphere. Tissue blocks about 1 mm3 were obtained from the white matter in which the points on the plastic sheet hit the white matter (Fig. 1C). Four tissue blocks were obtained per hemisphere. This sampling technique ensures a uniformly random distribution of the white matter samples and provides an equal sampling probability for all parts of the white matter. Thus, the final samples represented all parts throughout the entire white matter.

Figure 1.

A: An illustration of how to make systematic random slices of rat brain. The brain is embedded inside a small box with agar; a 2-mm scale paper is placed in the bottom of the box. After the agar hardens, the brain is systematically cut according to the scale paper using the same size small box without two bottom sides, the first one having been cut randomly. B: 2-mm systematic, random slices of rat brain are shown. C: The point grid is randomly put on the 2-mm brain slices. The tissue blocks are cut for the electron microscopy specimens where the points in the point grid hit white matter, as indicated by the arrow. D: (a) isector model: the numbers on the mould indicate that there are 5-mm in diameter spherical holes under the numbers. (b) A bent mould is illustrated. The spherical holes are filled with embedding media. After the embedding media hardens, the specimens become spheres. (c) An amplified image of a spherical specimen is shown. The white matter block was previously stained black by uranyl acetate.

Preparation of Transmission Electron Microscopic Section

The tissue blocks were postfixed in 4% glutaraldehyde for at least 2 hr at 4°C and rinsed in 0.1 M phosphate buffered saline (pH 7.2) three times and then osmicated in 1% 0.1 M phosphate buffered osmium tetroxide (OsO4) at 4°C for 2 hr. The blocks were then gradually dehydrated through a series of 50%, 70%, and 90% ethanol, a 90% ethanol and 90% acetone mixture, and 100% acetone. The blocks were infiltrated with epoxy resin 618 (ChenGuang Chemical Engineering Institute, China). The infiltration steps were acetone: resin 1:1 (3 hr at room temperature) and absolute resin (2 hr at 37°C). The specimens were placed in a mould that had 5-mm-in-diameter spherical holes and was filled with embedding medium. After hardening, the medium formed a sphere containing the specimen. In this study, the tissue blocks were embedded in 5-mm spheres with epon (Fig. 1D). After hardening, the spheres were rotated randomly on the table before being re-embedded in an oven at 37°C (16 hr), 45°C (12 hr), and 60°C (14 hr). This method, known as the isector (Nyengaard and Gundersen, 1992), ensures that isotropic, uniform, and random (IUR) sections can be obtained and thus ensures that the fibers of each direction in three-dimensional space have an equal probability of being sampled.

One 60-nm-thick section was obtained from the center of each epon block stained with uranyl acetate and lead citrate and was viewed using a transmission electron microscope (TEM) (Hitachi-7500, Hitachi, Ltd., Japan). From each section, 8 to 12 fields of vision were photographed at a magnification of 20,000× in a simple, random sampling fashion.

Stereology Analysis

Estimation of the white matter volume.

The total volume of white matter was measured from the 2-mm-thick slabs using the Cavalieri principle (Gundersen et al., 1988; Tang and Nyengaard, 1997; Li et al., 2008). A transparent counting grid with an area of 0.39 mm2 associated with each grid point was placed at random on the caudal surface of each slab. The points hitting the white matter were counted under an anatomical microscope (Fig. 2A). The total volume of white matter, VWM, was calculated using Cavalieri's equation:

equation image(1)

where t equals the slab thickness, a(p) equals the area associated with each grid point, and ∑PWM is the total number of grid points hitting the white matter per rat.

Figure 2.

A: A rat brain slice under anatomic microscope is shown. An equidistant point grid superimposed on a rat brain slice is shown. B: The unbiased counting frame is put on the randomly captured electron microscopic image. The myelinated fibres are counted if they were completely inside the counting frame or partly inside the counting frame but only touching the counting lines (dotted lines), as indicated by arrows. The myelinated fibers are excluded if they touched the exclusion lines (solid lines), as indicated by stars. C: The point grid is randomly put on the randomly captured image. The points on the myelin sheaths and the points on the white matter are counted separately. D: The unbiased counting frame and the point grid are randomly put on the randomly captured electron microscope image. For those myelinated fibers sampled with the unbiased counting frame, the points hitting myelinated fiber profiles, the points hitting the axonal profiles, and the points hitting the myelin sheaths are counted separately. The areas of the myelinated fibers profiles, the axonal profiles, and the myelin sheaths are calculated from the total points hitting the related structures multiplied by the area associated with each point, a (p).

The estimation of the length density and total length of the myelinated fibers in the white matter.

The length density of the myelinated fibers in the white matter, LV (mf/wm), was estimated by randomly overlaying an unbiased counting frame (Gundersen, 1977), with an area of 6,000 mm2 on each photograph. The myelinated fiber profiles completely inside the counting frame or partly inside the counting frame but only touching the top and right lines (inclusion lines) were included in the counting and nerve fiber profiles touching the left line, bottom line, and the extensions of the right and left lines (exclusion lines) were not counted (Fig. 2B). LV (mf/wm) was calculated as (Gundersen et al. 1988; Tang & Nyengaard, 1997; Li et al. 2008):

equation image(2)

where 2 is a constant that pertains to isotropic, IUR sections, ∑Q (mf) is the total number of the myelinated fiber profiles counted per rat, a(frame) is the area of one frame, and the ∑frame is the total number of counting frames used per rat. The total length of the myelinated fibers in the white matter, L (mf, wm), was estimated from (Gundersen et al., 1988; Tang and Nyengaard, 1997; Li et al., 2008):

equation image(3)

The unit for [a(frame) × ∑frame] is mm2. As ∑Q (mf) is 0 dimension, the unit for Lv (mf/wm) is 0 per mm2. When the same dimension is multiplied, the unit for Lv (mf/wm) becomes mm/mm3. As the unit for VWM is mm3, the unit for L (mf, wm) is mm. As 1 kilometer (km) equals 106 mm, when 106 is divided, the unit for L (mf, wm) becomes km.

The estimation of the volume density and total volume of the myelin sheaths in the white matter.

The volume density of myelin sheaths in the white matter, VV (ms/wm), was estimated by placing a transparent counting grid on the photographs (Fig. 2C). The points hitting the white matter and the points hitting the myelin sheaths were each counted. VV (ms/wm) was estimated as (Gundersen et al., 1988; Tang and Nyengaard, 1997; Li et al., 2008):

equation image(4)

where ∑P (ms) is the total number of the points hitting the myelin sheaths in the white matter per rat and ∑P (wm) equals the total number of points hitting white matter per rat. ∑P (ms) was included in ∑P (wm). The total volume of the myelin sheaths in the white matter, V (ms, wm), was calculated from (Gundersen et al., 1988; Tang and Nyengaard, 1997; Li et al., 2008):

equation image(5)

The estimation of the mean myelinated fiber profile area, mean axon profile area, and the mean area of myelin sheath cross section in the white matter.

The mean myelinated fiber profile area, the mean axon profile area, and the mean area of the myelin sheaths were estimated by placing a grid of systematically positioned test points randomly on the structure of interest (Fig. 2D). The myelinated fiber profiles were sampled by randomly overlaying an unbiased counting frame. The number of points that hit the structure, P, multiplied by the area associated with each point in the grid, a(p) (2.48 × 10−2 μm2), forms the area estimate, A (Gundersenn et al., 1988).

equation image(6)

The estimation of the mean perimeter of the myelin sheath cross section in the white matter.

On the randomly captured images from IUR sections, estimates of the outer perimeter, and inner perimeter of myelin sheaths were obtained with a randomized grid of parallel two-dimensional lines superimposed on the profiles sampled by an unbiased counting frame (Fig. 3). The number of intersections between the profile boundary and the test lines, I, was counted. The profile perimeter estimate, b(ms), was calculated as π/2 multiplied by the product of line separation, d (0.25 μm), and I:

equation image(7)
Figure 3.

The equidistant test lines with a line separation, d, are randomly put on the randomly captured image. For those myelinated fibers sampled with the unbiased counting frame as illustrated in this example, the number of intersections, ∑I, between the inner boundary as well as the outer boundary of the myelinated fiber and test lines is counted.

The outer perimeter of myelin sheaths was calculated when the intersections between the outer boundary of myelin sheaths and the test lines were counted. The inner perimeter of myelin sheaths was calculated when the intersections between the inner boundary of myelin sheaths and the test lines were counted.

The estimation of the mean diameter of the myelinated fibers in the white matter.

The external diameter of the myelinated fibers was estimated by measuring the longest profile diameter perpendicular to the longest axis of the myelinated fiber (Tang and Nyengaard, 1997; Fig. 4). The internal diameter of the myelinated fibers was estimated by measuring the longest profile diameter perpendicular to the longest axis of the axon (Fig. 4).

Figure 4.

Left: all myelinated fiber profiles completely inside the counting frame or partly inside it but not touching or intersecting the full-drawn lines are used for diameter measurement. Right upper: The inner diameter (d) of the myelinated fibers sampled with the unbiased counting frame is estimated by measuring the axonal profile diameter perpendicular to the longest axis of the axon (L). Right lower: The outer diameter (d) of the myelinated fibers sampled with the unbiased counting frame is estimated by measuring the myelinated fiber profile diameter perpendicular to the longest axis of the myelinated fiber (L).

The estimation of the thickness of the myelin sheath cross section in the white matter.

The points of intersection between the axon perimeter and the test lines used for perimeter estimation were used to obtain a systematic, IUR sample of locations for direct measurement of the myelin sheath thickness. The orthogonal myelin sheath thickness, t, for each axon was estimated as an average of four measurements. The points of intersections were numbered consecutively, the first position was chosen randomly in the first I/4 interval, the second position was I/4 + position 1, the third position was I/4 + position 2, and the fourth position was I/4 + position 3 (Fig. 5).

Figure 5.

All myelinated fiber profiles completely inside the counting frame or partly inside it but not touching or intersecting the full-drawn lines are sampled. The number of intersections, ∑I, between the inner boundary of the myelinated fiber profile and the equidistant test lines are counted (in this illustrated figure, there are 16 intersections between inner boundary and test lines). In this illustrated figure, the shortest distances from the inner boundary to the outer boundary at four positions are measured, which are indicated by t1, t2, t3, and t4. The mean thickness of myelin sheath equals to the mean value of t1, t2, t3, and t4. The first position is randomly chosen in the first I/4-interval. (In the given example, I is 16 so that the first position should be chosen at random from numbers 1 to 4.) In the illustrated figure, intersection 3 is randomly chosen as the first position. The second position is I/4 + position 1 (Intersection 7). The third position is I/4 + position 2 (Intersection 11) and the fourth position is I/4 + position 3 (Intersection 15).

Tissue Shrinkage

To estimate the tissue shrinkage induced by electron microscopy processing, the tissue blocks used to measure the shrinkage were taken from the white matter of the same hemisphere where electron microscopy tissue blocks were randomly sampled. One tissue block was randomly taken from each brain's white matter. The area of the coronal surface of each block was estimated with point counting:

equation image(8)

where ∑P is the number of points hitting tissue and a(p) is the area associated with each point of the counting grid.

The tissue was then processed the same way as in the experiment. After being processed, the area of the coronal surface of each block, A(after), was measured again by using point counting.

The measurements were then compared with see if any shrinkage had occurred. The amount of shrinkage was estimated as:

equation image(9)

RESULTS

The processing-induced shrinkage in the white matter was 7.5% and was not significantly different from zero. The areal shrinkage was not statistically significant, so it was not converted into volume shrinkage. Therefore, the total length and total volume of the myelinated fibers in the white matter were not corrected for the processing-induced tissue shrinkage.

The results from five rat brain white matter are presented in Table 1. The mean volume of the white matter [mean ± standard error of mean (SEM)] was 115 ± 5.43 mm3. Using 30 unbiased counting frames, on average, 212 myelinated fiber profiles were counted per rat white matter and 509 points hitting the myelin sheaths were counted per rat white matter. The mean length density of the myelinated fibers in rat white matter was 0.99 ± 0.07 km/mm3 and the mean volume density of the myelin sheaths was 0.29 ± 0.01. The mean total length of the myelinated fibers in rat white matter was 115 ± 12.5 km. The mean total volume of the myelin sheaths was 33.8 ± 2.74 mm3. The mean area of the myelin sheath cross section was 0.32 ± 0.02 μm2. The mean external perimeter and the mean internal perimeter of the myelin sheaths in the white matter were 3.39 ± 0.32 μm and 2.60 ± 0.30 μm, respectively. The mean arithemic diameter of the myelinated fibers and the axons were 0.67 ± 0.03 μm and 0.43 ± 0.03 μm, respectively. The mean thickness of the myelin sheaths in the white matter was 0.12 ± 0.00 μm. The absolute distribution of the myelinated fiber length in the white matter is shown in Fig. 6, which indicates that most of the myelinated fibers had myelin sheath thickness from 0.08 to 0.15 μm.

Figure 6.

The absolute fiber distribution of the total length of the myelinated nerve fibers in a young male rat's white matter.

Table 1. The stereological results of the white matter and the myelinated fibers and the myelin sheaths in the white matter
Animal12345MeanSEMCE (%)
  1. V(wm), the mean volume of white matter; L(mf), total length of myelinated fibers; V(ms), total volume of myelin sheaths; A(ms), mean area of myelin sheaths; A(a), mean area of axonal profiles; Ob(ms), outer perimeter of myelin sheaths; Ib(ms), inner perimeter of myelin sheaths; D(mf), arithemic mean diameter of nerve fibers; D(a), arithemic mean diameter of axons; t(ms), thickness of myelin sheaths; CE, coefficient of error for the estimation of the quantity before it.

V(wm) (mm3)113122112129981155.434.7
L(mf) (km)121139841428711512.510.9
V(ms) (mm3)35.638.131.339.624.433.82.748.1
A(ms) (μm2)0.270.300.380.320.340.320.025.9
A(a) (μm2)0.250.300.500.350.290.340.0413.0
Ob(ms) (μm)2.432.894.223.633.773.390.329.5
Ib(ms) (μm)1.712.153.342.793.012.600.3011.4
D(mf) (μm)0.700.690.760.650.550.670.035.1
D(a) (μm)0.440.420.510.420.340.430.036.3
t(ms) (μm)0.120.120.120.120.120.120.001.0

DISCUSSION

Till now, there has been no quantitative study of the myelin sheaths in brain white matter. Therefore, the present study examines efficient and unbiased stereological techniques for quantitatively investigating the myelin sheaths in white matter. In this study, systematic, uniform, and random sampling was done in entire white matter so that all parts of white matter and the myelinated fibers inside the white matter were equally as likely to be sampled. As the myelinated fibers in white matter were not arranged randomly, the myelinated fibers in white matter must have an equal probability of being sampled in three-dimensional space when the length and diameter of myelinated fibers and the perimeter and thickness of myelin sheaths are estimated. To make sure that all myelinated fibers in three-dimensional space have an equal probability of being sampled, isotropic, uniform, and random sections are needed. In the current study, isotropic, uniform, and random sections were ensured with the isector technique (Nyenggaard and Gundersen, 1992). In the present study, unbiased counting frame was used to sample and count myelinated fibers so that all sizes of myelinated fibers had an equal probability of being sampled. In this study, the problem associated with density estimates has been solved by the use of the design-based stereological technique, the Cavalieri principle, to estimate the reference volume. The total length of the myelinated fibers and the total volume of the myelin sheaths were obtained by multiplying the density estimates with the volume of white matter. The transmission electron microscopy technique was used to avoid the phenomenon called overprojection (Gundersen, 1979). In the present study, one 60-nm-thick section was cut from each epon block using an ultramicrotome. The ultrathin sections were then viewed under TEM. The very thin sections make the overprojection minimal. The design-based estimators were used to obtain estimates of the sizes of myelinated fibers, axons, and myelin sheaths. The methods described here enable observation of absolute size distributions. The size estimators in the present study provide information about the mean size of the myelinated fibers, axons, and myelin sheaths in the white matter. As the total length of the myelinated fibers in white matter was estimated, it permits one to observe the absolute size distribution (Fig. 6). In the absolute size distribution, the ordinate shows the actual length of the myelinated fibers for each class in contrast to the traditional relative size distribution where the ordinate is depicted with relative frequencies. A preferential change of a subpopulation of myelinated fibers can be detected in the absolute size distribution whereas the relative size distribution is hard to interpret.

The tissue-processing-induced shrinkage might affect the results obtained in the current study. To obtain the estimations of the total length and total volume of the myelinated fibers in the white matter, the length density and volume density have to be multiplied by the white matter volume. To verify the assumption that the reference volume is the same for both estimates, any changes in tissue volume, between the tissue used for the Cavalieri estimation of white matter volume and the tissue used for the estimations of the volume density and length density, should be estimated. We embedded the tissue in epoxy resin 618. The mean areal shrinkage induced by the tissue processing, measured only in x, y-directions, was 7.5%, which was not significantly different from zero. This shrinkage is probably negligible. One problem, however, for the estimation of area, perimeter, and thickness of myelin sheaths in the white matter, is that the results may be influenced to an unknown degree by the heterogeneous shrinkage of myelin sheaths, that is, the shrinkage may be not the same for the entire white matter and the myelin sheaths inside the white matter. Currently, this is a major limitation when interpreting results obtained from the processed tissue.

Another factor that might affect the current results is electron microscope (EM) sampling. In this study, the uniform, random sampling strategy was done in three dimensions. However, it needs to be emphasized that random sampling at the EM level must also be ensured. In the present study, 8 to 12 fields of vision were photographed in a simple, random sampling fashion. In practice, the field of vision was first sampled under lower magnification so that one could not observe the myelinated fiber density in the sampled area. Then, the myelinated fibers were photographed at very high magnification in the sampled field of vision.

CONCLUDING REMARKS

In conclusion, this study presents, for all practical purposes, unbiased stereological sampling techniques for estimating the total length of myelinated fibers, the total volume of myelin sheaths, mean sizes, and absolute sizes of the myelinated fibers and the myelin sheaths in rat white matter. The presented methods will prove to be very useful tools for future quantitative studies on the changes in the myelinated fibers and myelin sheaths of brain white matter in various experimental conditions and in various neurodegenerative diseases.

Acknowledgements

The authors thank all the staff in the Department of the Electron Microscope, Chongqing Medical University, People's Republic of China, for their assistance in the TEM technique.

Ancillary