| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Report
1 Department of Physiology, University of Massachusetts Medical School, Worcester, Massachusetts 01655; 2 Department of Mechanical Engineering, University of Rome La Sapienza, 00184 Rome, Italy
Submitted 23 April 2003; accepted in final form 29 May 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
; Grigg 1996; Grigg and Hoffman 1996; Kumazawa and Perl 1977; Nordin 1994). However, stretching skin produces a number of mechanical states, such as tensile stress, tensile strain, strain energy density, and their rates of change, in the skin. In general, it has been unclear which of these states might act as the stimulus to the mechanoreceptor ending. Recently, however, several of us (Del Prete et al. 2003) described a method that allows for determining the association between spikes and multiple stimulus variables. This method, which is based on multiple logistic regression (Hosmer and Lemeshow 1989), also allows for determining whether there are memory effects in the association between variables and spikes. "Memory," in this sense, refers to the time between the production of a stimulus in the skin and the appearance of a response in the afferent.
The use of logistic regression in applications such as this has been described in detail with reference to mechanoreceptors from mouse skin (Del Prete et al. 2003). In addition, the method was used to re-analyze previously published data from rapidly adapting (RA) afferents in rat skin (Grigg and Del Prete 2002). In both preparations, spikes were strongly associated with the rate of change of tensile stress. However, the two experiments found differences in the memory times at which the effect of that stimulus was observed. Responses in mouse mechanoreceptors were most strongly associated with the rate of change of stress 8–10 ms before a spike. In contrast, the re-analyzed data from rat skin units showed that the rate of change of stress had its maximal effect on the order of 30 ms prior to a spike. In attempting to account for the differences between these two results, we note that there were methodological differences between the two experiments. The apparatus used in the rat study was different from that used in the study on mice, and had rather different operating characteristics. For example it had a lower bandwidth and had significant harmonic ringing in the input waveform. As a result, the stimuli used in the two studies were quite different. We hypothesized the differences between the two results were due to differences in the nature of the stimuli produced by the two apparati. In this experiment, we have re-investigated the mechanical sensitivity of rat RA cutaneous mechanoreceptors, using methods that overcame the limitations of the earlier study.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
) from this laboratory. Because the preparation of rat skin (Khalsa et al. 1997) and the apparatus (Del Prete et al. 2003) are described in detail elsewhere, they are only briefly described here.
All procedures involving animals were approved by the University of Massachusetts Medical School IACUC. Young adult Sprague-Dawley rats of either sex were anesthetized with pentobarbital (45 mg/kg ip). The skin on the inner side of one hindlimb was clipped and then dehaired using a chemical depilatory (Nair). A section of skin was excised from the leg along with a length of the sensory nerve that innervates it. The animal was then killed with an overdone of anesthetic. The skin-nerve specimen was removed to an apparatus where it was maintained in HEPES-buffered artificial interstitial fluid (Koltzenburg et al. 1997). The fluid was maintained at room temperature (20°C) and was not gassed: it was simply allowed to equilibrate with room air. Under these conditions, activity could be recorded in afferent fibers for very long periods of time.
The skin specimen was held in the apparatus by its edges. One end was attached to a 5-mm-wide clamp that was fixed to the apparatus. The other end was coupled, through a plastic tab that was glued to the skin, to the actuator (Fig. 1). The actuator was an Aurora Scientific model 300B servo-controlled motor. The motor had an arm on its shaft, and the plastic tab was coupled to the tip of the arm. The sides of the specimen were loosely coupled to the sides of the apparatus using tabs that were cut into the edges. A hole was punched in the end of each tab, and small hooks with lengths of thread attached were used to couple the tabs to the apparatus.
|
Operating the actuator stretched the sample along its long axis, which corresponded to the long axis of the leg. As in the corresponding experiment on mouse skin (Del Prete et al. 2003), there was no evidence that significant loads were generated along the axis orthogonal to the direction of stretch.
The nerve was drawn into a small chamber where it was kept under mineral oil, and where it was dissected into small filaments. Individual filaments were placed on a fine gold wire electrode, and neural signals were amplified with an EG&G 113 preamplifier. Dissection was continued until mechanical stimulation revealed a single, clearly identifiable action potential. If the evoked spike had a consistent shape, we assumed that it represented the activity of a single afferent. Responses of the afferent of interest were discriminated from other spikes and from any compound spikes using a template-matching algorithm (Signal Processing Systems, Prospect, S. Australia). Rejected spikes were excluded from analyses.
A command signal, a band-limited pseudo Gaussian noise (PGN) sequence, was generated in software using LabView. A series of random numbers was generated, and output as an analog voltage to the amplitude control of the actuator. Stimuli were random with respect to frequencies (see Fig. 2) but Gaussian with respect to amplitude. The normal stimulus had a bandwidth of 0–100 Hz. In some experiments, we downsampled the random number sequence and interpolated values so as to produce bandwidths as low as 0–10 Hz. The actuator controller was set to control applied force. Load and displacement signals were read from the actuator at 2-ms intervals and stored in files along with a "1" or a "0" indicating the presence or absence (respectively) of an action potential in that sampling period. In a previous study (Del Prete et al. 2003), we showed that there were no instances in which more that one spike was observed in a 2-ms sampling period. Data collection runs were 30–60 s in duration. Intertrial intervals were 3–4 min. With each afferent studied, stimulus intensity was varied between runs. Mean values of stimuli ranged from 3.4 to 21.8 kPa. With a number of afferents, we lowered the stimulus bandwidth so as to make it closer to the bandwidth of the apparatus used in the previous study (Del Prete and Grigg 1998).
|
The measured variables, load and displacement, were converted to stress and strain, respectively. Loads were converted to stresses by dividing the applied load (F) by an estimate of cross-sectional area (A): 
= F/A. Cross-sectional area was estimated using the measured width of the specimen and assuming a thickness of 0.3 mm (Hoffman and Grigg 2002). Lagrangian strain was calculated using the expression 
= 
l/l0 where l was the measured displacement, and l0 was the distance from the clamp to the plastic tab. Multiple logistic regression (MLR) was used to determine the strength of association between mechanical stimulus variables and spike responses. This method is described in detail elsewhere (Del Prete et al. 2003). Briefly, the measured variables, load and displacement, were differentiated to determine their rates of change, and the resulting four variables: stress (), strain (), and their rates of change, d/dt and d/dt were used as predictors in a logistic regression analysis. The model that we used for data analysis also included all six first-order interactions (listed in Fig. 3) between those four variables. Each variable: load, displacement, their rates of change, and the interaction terms, was normalized to a distribution with a mean = 0 and a SD = 1.0. The purpose of this was to make the odds ratios, which are the measure of the strength of association between spikes and the variables, comparable to each other (Del Prete et al. 2003). We tested for memory effects using "lag" analysis; i.e., by shifting spikes backward and performing separate analyses for each lag time. Spikes were shifted backward in increments of 2 ms for a total of 50 ms. The data were analyzed separately for each lag time. Thus for each data collection run there were 26 MLR analyses; one for each lag time from 0 to 50 ms. Logistic regression analyses were done with SPSS version 9.
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
To identify relationships between stimulus components and spikes, data from each run in each afferent were subjected to analysis with multiple logistic regression. We looked for memory effects at lag times 
50 ms. With each afferent, the odds ratios for predictors varied with memory time. Odds ratios from all runs, representing a variety of stimulus amplitudes and bandwidths, were averaged together to form the best estimate of the properties of each afferent. Results from all 25 afferents were then averaged together to form a population estimate of the relationship between spikes and stimulus variables (Fig. 3).
To see whether altering either stress bandwidth or stimulus intensity had an effect on which terms were significant, or on memory intervals, we systematically altered both variables in a number of runs in different afferents. Changing either variable, the bandwidth or the magnitude of applied stress, had the expected result of changing the number of spikes that were observed in a run (Fig. 4) but had no effect on either the memory interval over which variables were associated with stimulus components or which terms were significant.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
). Thus the earlier finding in rat skin (Grigg and Del Prete 2002), in which a significant effect of d/dt was reported at memory times of 30 ms, is incorrect in this regard. While both this study and the earlier study of rat afferents found associations with the same stimulus variables, the actual memory time for various stimulus components is in fact quite similar to that in mouse afferents.
The difference in results appears to have arisen from two factors in the previous experiment. One factor was control system instability, and the second factor was filtering of data that was done because of the instability. Examination of data from the earlier experiment revealed a significant harmonic component of the input waveform caused by mechanical ringing of the actuator. Mechanical ringing is a problem because logistic regression analysis is based on the assumption that the stimulus is random and contains no periodic components. We apply logistic regression as a "reverse correlation" method (Eggermont et al. 1983); i.e., we use it to seek relationships between spikes in the record and the average stimuli that precede them. Consider what happens when the actuator presents a particular stimulus and then rings. The stimulus causes a spike and also causes a second stimulus component (the ringing). The ringing also causes a spike. Both spikes are time-locked to the initial stimulus. The regression analysis finds an association between both spikes and the initial stimulus. Notably, the association with the second spike has a long memory time. The association is real but it does not reflect a causative relationship. The result is an apparent association at a long memory time.
The second factor that obscured the true results in the earlier study was that because of the ringing, the data were filtered heavily. Most data were analyzed using a running average filter of rank as high as 7 or 8. Heavy filtering tends to obscure the relationship between stimuli and responses by spreading it out in the time domain. The result of both of the preceding factors was apparent associations between spikes and d/dt, at long memory times. In contrast, in the mouse experiment (Del Prete et al. 2003) and in the data that we now describe, there was no detectable actuator instability and the data were analyzed unfiltered.
In summary, it is very likely that the relationship that we now show was obscured in the earlier study by both the periodic components of the stimulus and the filtering of the data. We were led to this conclusion by some observations made in the present study in which we used position-controlled stimuli. When position was the controlled variable, there was a small resonant (i.e., periodic) component in the force response. As described in the preceding text, the resonant component occasionally resulted in spikes. This resulted in an apparent long memory time for stress variables. Because of this, we did not include analyses of position-controlled runs in this paper.
Rat RA afferents have a strong association with the rate of change of tensile stress with a peak at a memory time 
10 ms. In mouse cutaneous RAs, a similar association with d/dt was observed with a peak 8 ms. We attribute the difference in memory time to differences in conduction time along the axon because conduction time is subsumed in memory time in this experiment. The length of the nerve in the rat preparations was on the order of 30 mm as compared with 10 mm in the mouse. It was very difficult to measure conduction time in these experiments due to the fact that stimuli were applied to the skin in the saline bath, which both shorted out the stimulus current and created large artifacts. The short conduction distance meant that fast-conducting action potentials were usually obliterated by the stimulus artifact. If conduction time is estimated at 30 m/s, then the shift in the memory peak can be largely explained by conduction time.
RA afferents such as those we describe undoubtedly arise from hair follicles. However, it is not clear what kinds of hair afferents they are. If hairs (even clipped hairs) are left intact, they become wetted by the bathing solution and form a dense mat in which it has not been possible to identify the single hairs that drive the afferent. Hence we simply refer to them as RA afferents.
There are some significant differences between the properties of mouse and rat RA afferents. In rat afferents the strongest relationship was with static stress. In addition, relationships with interactions differed between rat and mouse. The only interaction term that was significant in this study was d/dt x . This interaction term was also significant in mouse afferents, but in mouse the odds ratio was smaller. At this time, we have no explanation for the differences between mouse and rat afferents other than to suggest that they are caused by species differences.
|
DISCLOSURES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
|
FOOTNOTES |
|---|
Address for reprint requests: P. Grigg, Dept. of Physiology, University of Massachusetts Medical School, 55 Lake Ave., Worcester, MA 01655 (E-mail: Peter.Grigg{at}umassmed.edu).
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
Del Prete Z and Grigg P. Responses of rapidly adapting afferent neurons to dynamic stretch of rat hairy skin. J Neurophysiol 80: 745–754, 1998.
Eggermont JJ, Johannesma PIM, and Aertsen AMHJ. Reverse-correlation methods in auditory research. Q Rev Biophysics 16: 341–414, 1983.[ISI][Medline]
Grigg P. Stretch sensitivity of mechanoreceptor neurons in rat hairy skin. J Neurophysiol 76: 2886–2895, 1996.
Grigg P and Del Prete Z. Stretch sensitivity of cutaneous afferent neurons. Behav Brain Res 135: 35–41, 2002.[Medline]
Grigg P and Hoffman AH. Stretch-sensitive afferent neurons in cat knee joint capsule: sensitivity to axial and compression stresses and strains. J Neurophysiol 75: 1871–1877, 1996.
Hoffman AH and Grigg P. Using uniaxial pseudorandom stress stimuli to develop soft tissue constitutive equations. Ann Biomed Eng 30: 44–53, 2002.[ISI][Medline]
Hosmer DW and Lemeshow S. Applied Logistic Regression. New York: Wiley, 1989.
Khalsa PS, LaMotte RH, and Grigg P. Tensile and compressive responses of nociceptors in rat hairy skin. J Neurophysiol 78: 492–505, 1997.
Koltzenburg M, Stucky CL, and Lewin GR. Receptive properties of mouse sensory neurons innervating hairy skin. J Neurophysiol 78: 1841–1850, 1997.
Kumazawa T and Perl ER. Primate cutaneous sensory units with unmyelinated (C) afferent fibers. J Neurophysiol 40: 1325–1338, 1977.
Nordin M. Intrafascicular recordings of afferent multi-unit activity from the human supraorbital nerve. Acta Physiol Scand 151: 507–514, 1994.[Medline]
This article has been cited by other articles:
|
|
 |
|
  B. B. Edin Quantitative Analyses of Dynamic Strain Sensitivity in Human Skin Mechanoreceptors J Neurophysiol, December 1, 2004; 92(6): 3233 - 3243. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Grigg and D. R. Robichaud II Rat Cutaneous RA Afferents Activated by Two-Dimensional Skin Stretch J Neurophysiol, January 1, 2004; 92(1): 484 - 491. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
