In the present review, the neuromuscular unit is considered to be comprised of a skeletal muscle and the sensory neurons and motoneurons associated with that muscle. The primary issue being addressed in this review is the plasticity of the neural elements of this unit to chronic decreases or increases in neuromuscular activity levels, defined as alterations in the activation and/or loading conditions within the unit. The plasticity of skeletal muscles has been detailed in a large number of reviews. Therefore, it will be discussed only to highlight specific points.
MATCHING PROPERTIES OF MOTONEURONS AND INNERVATED MUSCLE FIBERS
Skeletal muscle fibers can be classified (typed) based on a number of schemes and the following will be used in the present review. Based on the adenosine triphosphatase (ATPase) activity after preincubation at different pHs, fibers can be classified as Type I (slow, fatigue resistant), IIA (fast, fatigue resistant), or IIB (fast, fatigable). A second classification system is based on the contractile and metabolic profile of the fibers: slow-twitch oxidative (SO), fast-twitch oxidative glycolic (FOG), and fast-twitch glycolytic (FG). A third system uses gel electrophoresis or immunohistochemical methods to identify molecular differences the myosin heavy chain (MHC) composition of the fibers. Four primary MHCs have been identified and are correlated to contractile speed from the slowest to the fastest as follows: MHC I, IIa, IIx (IId), and IIb. It should be noted that many fibers contain more than one MHC, even in muscles from control animals (13).
Motoneurons in the ventral horn of the spinal cord generally can be designated as α or γ motoneurons. α Motoneurons innervate extrafusal fibers that constitute the main mass of a skeletal muscle. Each α motoneuron has a single axon that branches to supply a variable number of fibers in the skeletal muscle forming a motor unit. The motoneuron group innervating a single muscle is collectively called a motor pool. Motor units are usually categorized into three types based on their mechanical properties: slow (S), fast, fatigue resistant (FR), and fast, fatigable (FF). Although there is overlap among the three types, S motor units generally have smaller motoneurons and axons, and innervate fewer fibers than FF motor units. Chronic EMG recordings from a single motor unit indicate that S motor units are recruited first and are active during most types of activity, whereas FF (and to some extent FR) units are recruited for higher force/shorter duration activities. For example, S units in the rat soleus may generate as many as ∼5500,000 impulses·d−1, whereas fast units in the tibialis anterior may produce ∼53000 or fewer impulses·d−1 (15). γ Motoneurons are relatively small and innervate intrafusal fibers that regulate the length and the tension properties of the muscle spindles. The firing rates of γ motoneurons are quite variable, i.e., ranging between 10–75 impulses·s−1.
The metabolic properties of both motoneurons and sensory neurons appear to be related to their targets (see (15) for review). There is an inverse relationship between cell body size and oxidative enzyme activity levels in a population of motoneurons, suggesting that the γ motoneurons and the smaller α motoneurons (presumably innervating the slow fibers of S motor units) have a higher oxidative potential than the larger motoneurons innervating the fast fibers of FF motor units. Consistent with this view, is the finding that α motoneurons in motor pools innervating predominantly slow fibers have a higher oxidative capacity than those innervating predominantly fast fibers, even within the same muscle. There also is a cell body size and metabolic enzyme activity level interrelationship among the sensory neurons of the dorsal root ganglion. In general, cell body sizes of sensory neurons belonging to groups I and II (muscle spindle, Golgi tendon organ, etc.) are relatively larger than those belonging to groups III and IV (mechanoreceptors, joint and skin receptors, etc.). In this case, the larger sensory neurons have higher oxidative enzyme activities than the smaller sensory neurons.
In summary, within the neuromuscular unit there appears to be a matching of sizes and metabolic properties of the motoneurons and the muscle fibers that they innervate. Within a given population of motor pools, the oxidative capacity appears to be higher in the sensory than in the motor neural elements.
ADAPTATIONS OF MOTONEURONS AND SENSORY NEURONS TO VARYING LEVELS OF CHRONIC NEUROMUSCULAR ACTIVITY
It is expected that modulation of neuronal activity would alter the metabolic capacity of a motoneuron if the number and frequency of action potentials have a significant impact on metabolic demands of motoneurons. Indirect evidence for this relationship is the higher level of oxidative enzymes in the sensory neurons compared with the motoneurons (15), with the former having higher firing rates than the latter. The energy demand on a motoneuron may be related to one or more measures of cell body size of the motoneuron because motoneuron functions including intraneuronal transport, resting membrane potential, cytoplasmic components, and release and reuptake of neurotransmitters are indirectly or directly affected by cell body size. Therefore, oxidative capacity, cell body size, and the relationship between these measures might be expected to adapt when energy demands are modulated chronically.
Decreased Neuromuscular Activity
Chronic unloading of the hindlimb muscles of rats (suspending the hindlimbs by lifting the tail) is one method for producing chronic unloading with minimal effects on the activation of the muscles (see (3) for a review). For example, the total daily EMG activity of two of the primary ankle extensors, i.e., the soleus and medial gastrocnemius, is reduced during the first few days of unloading, but returns to normal levels within 7 d and remains at that level for at least 28 d of unloading. The same period of unloading results in atrophy, a shift in phenotype from slow to fast, i.e., Type I to Type IIa and IIx, a decrease in force generation capability, and an increase in velocity of shortening speed at the whole-muscle and single-fiber levels. These adaptations are particularly evident for predominantly slow muscles, e.g., the soleus. In contrast, the distribution of cell body size, mean cell body size, and mean oxidative capacity in the soleus motor pool are unaffected in the same hindlimb unloaded rats (15). These results indicate that alterations in the phenotype and size of muscle fibers can occur without concomitant adaptations in the innervating motoneurons. These findings also indicate that chronic unloading of the musculature has little impact on the motoneurons innervating them, but do not address the role of neuromuscular activation on motoneuron properties.
Spinal cord transection and spinal cord isolation
A complete spinal cord transection reduces and spinal cord isolation, i.e., complete spinal cord transections at a mid-thoracic and a high sacral level plus bilateral deafferentation, eliminates the production of action potentials by the affected motor pools. For example, spinal cord transection reduces the daily integrated EMG activity by ∼575% and 50% in the soleus and lateral gastrocnemius muscles of adult cats, respectively. Spinal cord isolation virtually eliminates all EMG activity in the soleus, a normally highly active muscle, of rats and cats. It also is assumed that the loading of the hindlimb musculature is decreased in spinal-cord–transected and minimal in spinal-cord–isolated animals. Under these conditions, muscle fiber size and oxidative capacity in some muscles decrease. However, no changes in the oxidative capacity or cell body size of the associated motoneurons were observed (1). Similarly, there were no changes in number or size of motoneurons supplying the rat sciatic nerve after a complete spinal cord transection at T9 (11). Thus, there was a differential effect of these spinal cord manipulations on the neural and muscle elements of the neuromuscular unit suggesting that: 1) the metabolic work associated with motoneuron activity, i.e., the ionic pumping and neurotransmitter turnover associated with action potentials, represents a small proportion of the total energy demand within the motoneuron; 2) the oxidative capacity of a motoneuron is resistant to a chronic change in energy demand; and/or 3) the size of a motoneuron is not influenced by a change in the size of its target, i.e., total muscle fiber cross-sectional area within a motor unit. These conclusions are supported by the results from another model of inactivity: tetrodotoxin (a selective channel blocker) administration to the rat sciatic nerve for 2 wk results in atrophy and a decrease in the oxidative potential of fibers in the medial gastrocnemius, whereas a population of motoneurons in the medial gastrocnemius motor pool was unaffected (5). Interestingly, the lack of adaptation in the motoneurons in the spinal-cord–isolated and tetrodotoxin-treated rats suggests that the elimination of afferent information from the affected muscles has little importance in modulating the motoneuron properties (see Exposure to microgravity, below).
The energy demand that a cholinergic terminal places on its cell body may be related to the synthesis of the enzymes acetylcholinesterase and choline acetyltransferase and the maintenance of terminal structures such as the mitochondria, the macromolecules involved in vesicle release, and the plasma membrane. Even with changes in neuronal activity levels, however, these terminal energy demands on the cell body may be unaffected because the choline acetyltransferase level in a terminal seems to be unchanged after decreased activation of cholinergic neurons. The biosynthetic energy demand on the cell body, the presumed site of most protein synthesis within a motoneuron, would not be expected to change due to the reduced neural activity, because the frequency of neuronal activation does not seem to influence the half-lives of proteins. Another interpretation may be that because the connectivity between the motoneuron and muscle is maintained in these spinal models, activity-independent neurotrophic influence between these elements continues to have an impact on the motoneuron properties. This possibility is consistent with the results from the axotomy studies presented in the following section.
The unique aspect of axotomy as a model of inactivity is that it involves the elimination of neuromuscular connectivity. For example, when the plantaris is functionally overloaded via the ablation of its major synergists, i.e., the soleus and gastrocnemius, the motoneurons innervating the ablated muscle lose their target tissue. Chalmers et al. (1) have shown that axotomized triceps surae motoneurons have a 28% lower oxidative capacity than intact triceps surae motoneurons in functionally overloaded cats. This reduction in oxidative capacity occurred in motoneurons of all sizes. In contrast, there was no change in the mean size of the motoneurons between the two groups. Because axotomized, ligated motor nerves have near normal activation patterns for up to 200 d postaxotomy, activity-dependent ion pumping per unit area of remaining membrane most likely was unaffected after axotomy. It is likely, however, that biosynthetic processes were reduced in the axotomized neurons, at least after an initial period of axonal elongation. It appears that once a motoneuron passes the initial postaxotomy phase, the biosynthesis demands on the cell body are greatly reduced because 60–80% of the cytoplasmic volume of a motoneuron innervating the hindlimb muscles is localized in the distal axonal tree and terminal branches (1). Accordingly, it appears that the principal modifications imposed by axotomy are an elimination of neurotransmission at the neuromuscular junction, a reduced biosynthetic demand, a decreased need for axoplasmic transport, and a loss of access to neurotrophic factors such as neurotrophin-3 and neurotrophin-4 in the muscle. These modifications appear to decrease the energy demand in the cell body of the motoneuron, as reflected by a reduction in oxidative capacity.
The mechanisms by which neuronal metabolism is reduced after axotomy are unknown. Loss of a trophic influence from the target muscles or other peripheral tissues (e.g., Schwann cells, muscle spindles) could be a stimulus to direct the cell body of the motoneuron to support regrowth metabolically as opposed to the alternative hypothesis of missing terminals. By whatever means the cell body of the motoneuron is signaled, however, the decrease in oxidative capacity after axotomy indicates a reduced energy demand.
Exposure to microgravity
It appears that exposure to microgravity induces adaptations that cannot be associated simply with chronic unloading of the hindlimb (see below). It is suggested that the exposure to microgravity may not induce altered supraspinal function such as is known to occur within the vestibular system which could, in turn, disrupt a specific group of motoneurons.
The most intriguing findings of neuronal adaptation to decreased levels of neuromuscular activity have been with spaceflight. Microgravity provides a unique environment, i.e., a 0 G environment. Thus, the constant influence of Earth’s gravitational forces on the organism is eliminated in microgravity. As with the other models of decreased activity or inactivity, the skeletal muscles atrophy and show shifts toward faster mechanical and phenotypic properties (10). Again, this is most evident in the slow, antigravity muscles, such as the soleus, because the normal function of these muscles is impacted the most in the microgravity environment. However, there also were some adaptations in the neural elements of the neuromuscular unit. For example, there was a selective decrease in the oxidative capacity of medium-sized motoneurons (small-sized α motoneurons) in the retrodorsolateral region of the ventral horn in the L5 and L6 spinal segments after exposure to 14 d of space flight (Fig. 1), and this effect persisted for at least 9 d after return to Earth (6). The location and cell body size of the affected motoneurons are consistent with these neurons representing the soleus motor pool and also would be consistent with the observed decrease in the mean size and oxidative capacity of the soleus fibers in the same rats after exposure to microgravity. This speculation is being addressed in a planned space flight experiment in which the soleus motor pool will be identified via retrograde neuronal labeling techniques.
Indirect evidence for the involvement of motor pools associated with antigravity muscles has been provided. The oxidative capacity and cell body size of motoneurons in the dorsomedial region of the ventral horn at the L5 and L6 spinal segments, i.e., motoneurons associated with the predominantly fast, low-oxidative bulbocavernosus and levator ani muscles that have no load-bearing function, were determined in the same rats described above (8). No changes in the oxidative enzyme activity of motoneurons in the dorsomedial region were observed after space flight or after recovery from space flight (Fig. 2). Combined, these data indicate that the capacity to support oxidative phosphorylation in a select population of motoneurons involved in posture and locomotion was reduced during a period of exposure to microgravity, whereas similar-sized motoneurons associated with muscles having no known function related to gravitational forces were unaffected.
In these same rats, there also was a selective decrease in the oxidative capacity of large-sized sensory neurons in the rat dorsal root ganglion after exposure to microgravity (Fig. 3), and this effect persisted for at least 9 d after return to Earth (7). These large sensory neurons are considered to belong to group Ia and II afferents, and therefore these changes are thought to reflect a decrease in the functional activity of primary afferent neurons, most likely emanating from the hindlimb musculature. These results are consistent with the finding that exposure to microgravity changes sensory function, and that these changes persist some days after return on Earth, as do some of the alterations in motor control (3).
During space flight, it is assumed that there is a general reduction or at least a severe alteration in afferent flow, particularly from the proprioceptors that respond to load bearing and kinematics of the limbs as well as from vestibular mechanisms. Furthermore, it appears that extensive reinterpretation of the afferent signals is required to accomplish motor tasks in the changed gravitational environment, e.g., to maintain a stable posture upon return to Earth. Edgerton and Roy (3) suggested that the integrated neuromotor response to exposure to microgravity be considered as a Gravitational Unloading Syndrome.
To begin to test the validity of this concept, soleus EMG and both afferent and efferent neurograms at the L5 segment of the spinal cord of adult rats were recorded during parabolic flights (9). Soleus EMG activity gradually increased when the gravity level was elevated from 1 to 1.5 G (+ 23%) and then to 2 G (+ 67%) during the ascending phase of the parabolic flight (Fig. 4 and 5). EMG activity in the soleus then immediately decreased to ∼528% of the 1-G level when the rats were exposed to microgravity during the descending phase and was maintained low during the entire 20-s epoch at microgravity. The EMG levels were restored immediately when the gravity level was increased to 1.5 G and then 1 G during the descending and recovery phases. The afferent neurogram was affected in the same direction as that of the soleus EMG, although the magnitude of the reduction during the microgravity phase was minor relative to that of soleus EMG (an ∼526% vs 72% decrease from 1-G levels for the afferent neurogram and EMG, respectively). The level of the efferent neurogram was decreased only slightly (∼59% from the 1-G level) during the 20 s of exposure to microgravity. Although these data suggest that the afferent input is closely associated with the gravity-dependent muscular activity, this simple relationship is not clearly evident in the efferent activity.
In summary, it appears that space-flight–induced adaptations in the neural elements of the neuromuscular unit cannot be associated simply with chronic unloading of the hindlimbs, because similar changes were not observed in animals chronically unloaded at 1 G (3,15). One possibility is that altered supraspinal function during space flight, such as in the vestibular system, could disrupt a specific group of motoneurons in the spinal cord (3,15). Regardless of the mechanism involved, these data clearly demonstrate that the space flight environment produces effects on the neural elements of the neuromuscular unit that are not observed in other models of decreased activity. Consequently, microgravity provides an excellent environment to study the effects of subnormal levels of neuromuscular activity on these central and peripheral neural elements.
Increased Neuromuscular Activity
Functional overload of a muscle by removal of its synergists increases both the activation and loading of the remaining musculature. The overloaded muscle hypertrophies, shows a shift toward a higher percentage of Type I fibers and assumes mechanical properties resembling a slow muscle (see (5) for a review). In contrast, the neural elements of the neuromuscular unit appear to be quite resistant to change. For example, no change in the cell body size or oxidative capacity of motoneurons in the cat plantaris motor pool was found after 12 wk of functional overload, although an increase in size and oxidative capacity of muscle fibers was observed (1). Similarly, no change in the oxidative capacity of rat plantaris motoneurons was observed after 10 wk of functional overload or functional overload plus endurance treadmill-running exercise although significant hypertrophy and a minimal effect on MHC adaptations of the muscle fibers were observed (14). In both of these studies, the plantaris was overloaded bilaterally and the data were compared with normal controls. When the plantaris is overloaded unilaterally and the results are compared with the contralateral control, a small decrease (∼59%) in cell body size of motoneurons has been reported (4). It is suggested that morphological and physiological parameters of motoneurons innervating a hypertrophied muscle were shifted toward those of normal slow motor units. On the other hand, a parallel decrease in the oxidative enzyme activity in the motoneurons and muscle fibers of the rat soleus motor pool was observed in the overloaded compared with the contralateral control muscle after 60 d of unilateral overload (12). The decreased oxidative capacity is considered to be a decreased ability to use oxidative metabolism for periods of short-term high-energy demands. The reason(s) for the differential response in uni- and bilaterally overloaded animals is unknown. However, these data are consistent with the view that the motoneurons are more resistant to adaptations in cell body size and oxidative capacity after functional overload.
Activity-dependent ion pumping is most likely elevated after functional overload. The overloaded muscle must be recruited to perform the tasks normally accomplished by the ablated muscles and this is reflected in the elevated EMG activity and/or tendon forces recorded during locomotion and postural activities in rats and cats. Although neurotransmission also would increase in conjunction with the elevated number of action potentials, this increased demand may have little metabolic effect on the cell body or oxidative metabolism as described above for models of decreased activity. Functional overload does not necessarily modify either protein synthesis or axoplasmic transport (5). The lack of change in oxidative capacity of motoneurons and possibly the decrease in functionally overloaded animals may reflect only a modest increase in neuronal activity when considering the overall pattern of activity in a 24-h period. Interestingly, however, it would appear that functional overload would impose a significant metabolic stress on motoneurons that increase their activity levels. Thus, in a model in which the size and the metabolic properties of the target tissue are altered dramatically and the activity of the entire neuromuscular unit is elevated, these same properties are unaffected in the motoneurons.
Chronic electrical stimulation
Chronic electrical stimulation, usually at low frequency of stimulation, results in a shift in the phenotypic and mechanical properties toward those observed in a slow muscle. In addition, this paradigm usually results in atrophy and an increase in the oxidative capacity of the muscle fibers. Despite these muscle adaptations and the enormous number of imposed action potentials, the motoneuronal properties are unaffected. For example, no change in the oxidative capacity of motoneurons in cats after 56 d of chronic antidromic electrical stimulation was observed (2).
It is well documented that exercise results in a large number of adaptations in the morphological, metabolic, and functional properties of skeletal muscles that are specific to the type and amount of exercise training. In contrast, relatively few studies have examined the effects of exercise on the metabolic and size properties of motoneurons and the results have been equivocal (see Roy et al. (15) for a review). Swim training in rats for 52 d increased the oxidative capacity of an unspecified population of motoneurons, and endurance treadmill-running exercise for 8 wk slightly (∼54%) increased the oxidative potential of the motoneurons innervating the soleus muscle. On the other hand, 10 wk of endurance running showed adaptations (an increased oxidative enzyme activity in the soleus and an increased area of FOG fibers in the extensor digitorum longus) and caused an increased cell body size of soleus motoneurons, but not extensor digitorum longus motoneurons, although no difference in the oxidative enzyme activity of the soleus or extensor digitorum longus motoneurons was observed between the control and exercised rats. In contrast, 12 wk of daily voluntary plus endurance treadmill exercise increased the oxidative potential of Type I fibers and proportion of Type I and Type I + IIa fibers in the medial gastrocnemius, but had no effect on the oxidative potential of the innervating motoneurons. Combined with the data from the functional overload and chronic stimulation studies described above, it appears that the motoneurons are less responsive to chronic increases, as well as decreases (see above), in neuromuscular activity than the muscle fibers that they innervate.
Under normal conditions, there seems to be a close association between the size and metabolic properties of the neural and muscular elements within the neuromuscular unit. It appears, however, that the muscle fibers are, in most instances, much more adaptive than the motoneurons under a variety of conditions. In other words, the motoneurons appear to be more resistant to change than the muscle fibers that they innervate. Consequently, the cell body size and metabolic properties of motoneurons are not tightly coupled to the size or phenotype of the muscle fibers that they innervate when the activity level of the neuromuscular system is chronically perturbed. The mechanism(s) for this differential response is unknown.
Apparently, microgravity presents a unique environment for the neuromuscular unit. Exposure to microgravity may induce altered supraspinal function such as is known to occur in the vestibular system which could, in turn, disrupt a specific group of motoneurons, e.g., slow-type motoneurons. It is clear, however, that microgravity-induced adaptations in the neural elements of the neuromuscular unit cannot be associated simply with chronic unloading of the hindlimbs. Further studies are needed to elucidate the mechanism for neuronal adaptations after exposure to space flight because it is suggested that there are unique factors that affect a selected group of spinal motoneurons presumably associated with posture and locomotion and disrupt the metabolic homeostasis of motoneurons in the microgravity environment. In general, these studies lead to the conclusion that motoneurons and dorsal root ganglion neurons either are relatively stable in response to variations in activity levels or may be quite adaptive in parameters other than cell size and metabolic properties reflecting their oxidative phosphorylation potential.
A large portion of this work was supported by NIH Grant NS16333 (to V.R.E. and R.R.R.), NASA Grant NAG-2–717 and −438 (to V.R.E. and R.R.R.), and a grant from the Japan Space Forum and the National Space Development Agency of Japan (to Y.O. and A.I.).
1. Chalmers, G.R., Roy, R.R.and Edgerton. V.R. Adaptability of the oxidative capacity of motoneurons
. Brain Res. 570: 1–10, 1992.
2. Donselaar, Y., Kernell, D.and Eerbeek. O. Soma size and oxidative enzyme activity in normal and chronically stimulated motoneurones of the cat’s spinal cord. Brain Res. 385: 22–29, 1986.
3. Edgerton, V.R., and Roy. R.R. Neuromuscular adaptations to actual and simulated spaceflight. In: Handbook of Physiology, Section 4. Environmental Physiology, III. The Gravitational Environment, edited by Fregly M.J. and Blatteis. C.M. New York: Oxford University Press, 1996, pp. 721–763.
4. Finkelstein, D.I., Lang, J.G.and Luff. A.R. Functional and structural changes of rat plantaris motoneurons
following compensatory hypertrophy of the muscle. Anat. Rec. 229: 129–137, 1991.
5. Gardiner, P.F. Effects of exercise
training on components of the motor unit. Can. J. Sport Sci. 16: 271–288, 1991.
6. Ishihara, A., Ohira, Y. Roy, R.R. Nagaoka, S. Sekiguchi, C. Hinds, W.E.and Edgerton. V.R. Influence of spaceflight on succinate dehydrogenase activity and soma size of rat ventral horn neurons. Acta Anat. 157: 303–308, 1996.
7. Ishihara, A., Ohira, Y. Roy, R.R. Nagaoka, S. Sekiguchi, C. Hinds, W.E.and Edgerton. V.R. Effects of 14 days of spaceflight and nine days of recovery on cell body size and succinate dehydrogenase activity of rat dorsal root ganglion neurons. Neuroscience. 81: 275–279, 1997.
8. Ishihara, A., Ohira, Y. Roy, R.R. Nagaoka, S. Sekiguchi, C. Hinds, W.E.and Edgerton. V.R. Comparison of the response of motoneurons
innervating perineal and hind limb muscles to spaceflight and recovery. Muscle Nerve. 23: 753–762, 2000.
9. Kawano, F., T. Nomura, A. Ishihara, I. Nonaka, and Y. Ohira. Afferent input-associated reduction of muscle activity in microgravity environment. Neuroscience
. (in press, 2002).
10. Ohira, Y. Neuromuscular adaptation to microgravity. Jpn. J. Physiol. 50: 303–314, 2000.
11. McBride, R.L., and Feringa. E.R. Ventral horn motoneurons
10, 20 and 52 weeks after T-9 spinal cord transection
. Brain Res. Bull. 28: 57–60, 1991.
12. Pearson, J.K., and Sickles. D.W. Enzyme activity changes in rat soleus motoneurons
and muscle after synergist abaltion. J. Appl. Physiol. 63: 2301–2308, 1987.
13. Pette, D., and Staron. R.S. Mammalian skeletal muscle fiber type transitions. Int. Rev. Cytol. 170: 143–223, 1997.
14. Roy, R.R., Ishihara, A. Kim, J.A. Lee, M. Fox, K.and Edgerton. V.R. Metabolic and morphological stability of motoneurons
in response to chronically elevated neuromuscular activity. Neuroscience. 92: 361–366, 1999.
15. Roy, R.R., Edgerton, V.R.and Ishihara. A. Chapter 10. Influence of endurance training and detraining on motoneurone and sensory neurone morphology and metabolism. In: Endurance in Sport, edited by Shephard R.J. and Astrand. P.-O. London: Blackwell Science, 2000, pp. 136–157.