morphology of the central projections of physiologically ... · of physiologically characterised...

J Comp Physiol A (1992) 170:101-120 Journal of Comparative ~ ' ~ ' Neural,

and

Physiology A ~,%,%" �9 Springer-Verlag 1992

Morphology of the central projections of physiologically characterised neurones from the locust metathoracic femoral chordotonal organ

Thomas Matheson*

Department of Zoology, University of Canterbury, Christchurch, New Zealand

Accepted September 25, 1991

Summary. The metathoracic femoral chordotonal organ of the locust (Locusta migratoria) is an internal pro- prioceptor composed of mechanosensory neurones which respond to tibial position, velocity, or acceleration, or to combinations of these parameters. Discrimi- nant function analyses confirmed the visual observation that neurones with different responses to tibial movements had different central branching patterns. Some aspects of the projections were consistent for all neurones (e.g., the path taken by the main neurite through the metathoracic ganglion), whereas other regions of branches were consistently reduced or missing in some response classes. Some position-and-acceleration receptors had no main branches off the main neurite, and must therefore make relatively restricted contact with motor neurones and interneurones. Phasic or tonic neurones which responded in ranges of tibial extension had branches which projected further medial in Dorsal Com- missures III and IV than similar neurones which responded in ranges of tibial flexion. I compare my results to previous studies of mapping in the insect CNS.

Key words: Insect - Locust - Chordotonal organ - Cen- tral projection - Mapping- Discriminant function analysis - Canonical function analysis

Introduction

The central branching patterns of physiologically similar afferent neurones often show variations in form. Map- ping occurs if these variations relate in a systematic way to aspects of the neurones' responses, to their peripheral location, or to the time of their development during the

Abbreviations: (ms) (mt)FCO (mesothoracic) (metathoracic) femoral chordotonal organ; A N O V A Analysis of Variance

* Present address: Department of Zoology, University of Cam- bridge, Downing Street, Cambridge CB2 3E J, England

animal's life. Several types of mapping have been described from the insect CNS: (1) tonotopic mapping is the orderly arrangement of afferent neurones according to their sensitivity to different frequencies of vibration. For example, the neurones which make up the ordered arrays of cell bodies in cricket tympanal organs (crista acoustica) project into the anterior intermediate sensory neuropile of the CNS in a pattern related to their physiological response (Oldfield 1983; R6mer et al. 1988). In contrast, the 4 groups of receptor cells from the locust ear project to 4 different areas of this same neuropile, but there is no discernible mapping within each area (R6mer et al. 1988); (2) somatotopic (topographic) mapping is the arrangement of afferent projections according to the position of their peripheral somata. For example, the terminal projections of identifiable cricket clavate hairs located on the cerci are unique and reproducible between individuals (Murphey et al. 1980). The branching pattern of a clavate hair neurone is correlated with the cell's 'birthday' (moult in which it first appears) and its position on the cercus. Newland (1991) has recently shown that tactile hairs on the legs of locusts have central branching patterns which correspond to the exact position of the hair on the leg. Thus, the surface of the leg is represented in a three dimensional map in the corresponding ganglion; (3) retinotopic mapping consists of an orderly projection of retinular fibres from the ommati- dia onto the lamina (Braitenberg 1972; Horridge and Meinertzhagen 1970), and further orderly representa- tions of the visual field in the lobula and lobulus (Braiten- berg 1972); and (4) developmental mapping arises when the time of differentiation during development correlates with the central projection pattern (e.g., Drosophila wing campaniform sensilla: Dickinson and Palka 1985; Palka et al. 1986). For these sensillla there is a three-way correlation between time of development, function, and central projection pattern.

Connective chordotonal organs are internal proprioceptors which monitor joint movements in arthropods (see reviews in Mill 1976). The metathoracic femoral chordotonal organ (mtFCO) spans the hind leg

102 T. Matheson: Chordotonal organ central projections

f e m u r - t i b i a j o i n t o f t he locus t , a n d r e s p o n d s to m o v e - m e n t s o f t he t i b i a r e l a t i ve to the f e m u r . I t c o n t a i n s ap- p r o x i m a t e l y 90 n e u r o n e s ( M a t h e s o n a n d F i e ld 1990) w h i c h r e s p o n d to t ib ia l p o s i t i o n , ve loc i ty , o r acce le ra - t i on , o r to c o m b i n a t i o n s o f these p a r a m e t e r s ( M a t h e s o n 1990a). By u s i n g w h o l e - n e r v e backf i l l i ng t e c h n i q u e s , F i e l d a n d Pf l i iger (1989) h a v e s h o w n t h a t t he f u n c t i o n s a n d cen t r a l p r o j e c t i o n s o f the t w o s c o l o p a r i a ( g r o u p s o f n e u r o n e s ) p r e s e n t in t he l o c u s t p r o - a n d m e s o t h o r a c i c F C O s a re qu i t e d i f f e r en t f r o m e a c h o the r . T h o s e a u t h o r s d id no t , h o w e v e r , a t t e m p t to s t a in o r c h a r a c t e r i s e in- d i v i d u a l n e u r o n e s . L i k e w i s e , B u r r o w s (1987) s h o w e d m t F C O p r o j e c t i o n s d e r i v e d o n l y f r o m w h o l e - n e r v e fills. T h i s m e t h o d d o e s n o t a l l o w a c c u r a t e i n v e s t i g a t i o n o f t he v a r i a t i o n b e t w e e n a f fe ren ts , a n d p r e c l u d e s a t t e m p t s to c o r r e l a t e s t r u c t u r e w i t h p h y s i o l o g i c a l r e sponse .

I n this p a p e r I p r e s e n t fo r t he first t i m e a de t a i l ed s t u d y o f t he c e n t r a l p r o j e c t i o n s o f i nd iv idua l , p h y s i o l o g - ica l ly c h a r a c t e r i s e d n e u r o n e s f r o m a j o i n t c h o r d o t o n a l o r g a n , t he l o c u s t m t F C O . D i s c r i m i n a n t f u n c t i o n a n a l y - ses a r e u s e d to h e l p r e v e a l r e l a t i o n s h i p s b e t w e e n phys io - log ica l r e s p o n s e p r o p e r t i e s a n d cen t r a l p r o j e c t i o n pa t - terns .

M a t e r i a l s a n d m e t h o d s

Adult locusts (Locusta migratoria) from our laboratory culture were used for all experiments. The results are based on 101 success- ful recordings from 119 animals.

Locusts were restrained ventral side up in plasticene. Move- ments of the mtFCO apodeme at controlled velocities mimicked tibial movements between 0 and 120 ~ (Matheson 1991). The flexor strand was held motionless at 60 ~ Matheson (1990a) discusses possible biases introduced by this method. More recent evidence (Field 1991) reveals that the flexor strand is morphologically simple, and has at most a minor and indirect effect on mtFCO neurones. This supports the argument that any biases will be small. More flexion sensitive neurones were recorded than extension sensitive neurones ('Fable 1), but it is not possible to determine if this apparent bias is due to the method of stimulation or whether it reflects the true proportions of these neurones. Neither interpretation alters the conclusions of the present paper, which is not concerned with relative numbers of each type of neurone.

A window cut in the ventral thoracic cuticle gave access to the metathoracie ganglion, which was supported on a wax-coated silver platform. A silver loop slightly larger in diameter than the ganglion was lowered over the ganglion to minimise movements, and to hold the large longitudinal tracheae away from the recording site in n5 (Matheson 1990a). Care was taken not to crush the connectives or other nerves entering the ganglion. The tracheal supply to the ganglion was left intact, but some air-sacs associated with the thoracic musculature were removed. A groove cut in a third silver platform, positioned under n5 close to the ganglion, provided great stability for intracellular recording. Drops of locust saline (pH 6.8) were added if required to prevent drying out of the preparation.

After physiological characterisation and dye-filling of a mtFCO afferent neurone (hexamminecobalt (III) chloride, 30 min, 1.5 Hz, 200 ms pulses, 2.5 V) the electrode and support platforms were removed, and the dye allowed to diffuse for a further 30 min. The metathoracic and mesothoracic ganglia were then dissected out together, and the cobalt visualised using standard procedures, in- eluding silver intensification (Bacon and Altman 1977). Neurones were drawn with the aid of a camera lucida. Preparations in which more than one cell stained have not been included in the analysis, with the following exception: if one cell in a multiple fill was much

Table 1. Summary of neurones successfully recorded and stained in the present study. Note that classes are numbered from 1-9, and subclasses from 0-9, and A T

Response type Class Sub- Abbreviation Number class recorded

Acceleration 1 0 ?A + 5 1 1 ?A + - 2 1 2 ? A - 3 1 3 ?AS 9

Position and 2 4 P~ 5 acceleration a 2 5 pmie?. A x 5

Position and 3 6 P~ 8 flexion velocity 3 7 P~ i

3 8 P~176 2 3 9 pmidV + 2

Position, 4 A P~ § ?A + 0 b flexion velocity, 4 B pmidV +?A + 2 and positive 4 c C p2~ 1 acceleration

Position and extension velocity

Position, extension velocity, and negative acceleration

Flexion velocity

Extension velocity

Extension velocity and negative acceleration

5 D P~ 0 b 5 E pmidV 3 5 F p12~ 5

6 G P~ - 0 b 6 H pmidV -?A - 9 6 I P12~ - 5

7 J V + 3 7 K V +wide 3

7 L V +~ 10 7 c M V+?A + 2

8 N V - 2 8 O V -mid 4 8 P V -wlde 2 8 Q v -12~ 0 b

9 R V-?A - 6 9 S V - mid?. A - 1 9 T V- wid~.A - 1

a The 4 neurones classified as pure position receptors in Matheson (1990a) probably also have an acceleration component, although in neurones with high tonic firing frequencies, any acceleration component (usually a single spike) is often obscured by the background activity b Because response classes and subclasses were assigned a priori (before dye-filled neurones were drawn), some subclasses contain no observations. In these cases, the neurones with appropriate responses did not stain satisfactorily c These neurones do not fit exactly into any of the response classes, so have been grouped here for simplicity

darker than the others it was assumed that this was the characterised cell, while the others had merely taken up traces of dye during short periods of recording from them earlier in the experi- ment. This possible ambiguity has been noted in the results where applicable. Poorly stained cells are included in the analysis, but are noted as such.

Selected ganglia were embedded in Paraplast wax for serial sectioning at 7 txm. Sections were counterstained in 0.1% Toluidine Blue and mounted in Eukitt. Cobalt filled branches were drawn in relation to recognisable tracts and neuropiles visible using inter- ference contrast optics. Illustrations are of individual sections, or in some cases, composites of up to 3 adjacent sections.

T. Matheson: Chordotonal organ central projections 103

Classification o f response types

Detailed descriptions and examples of the responses and classification of locust mtFCO neurones are presented in Matheson (1990a, b, and unpublished results). Abbreviat ions for the different response types follow those used by Hofmann et al. (1985). All neurones with a tonic discharge coded for position (i.e., the firing frequency differed at different set angles). These neurones were labelled 'P' . The angle causing the maximum tonic firing rate was added as a superscript (e.g., pTo, p0). If there was hysteresis in the response, then the two curves were averaged prior to determining this angle. Neurones which responded to velocity (V) or start of movement (?AS="accelera t ion at the start") were given a superscript ' + ', ' - ' or ' + - ' to indicate directional sensitivity (Fig. 1). A ' + ' (flexion) or ' - ' (extension) code indicates unidirectional sensitivity whereas a ~ + - ' code means that the neurone responded to movements in bo th directions. Putative acceleration receptors (?A) were given a superscript ' + ' or ' - ' to indicate the sign of the acceleration to which they responded (' + ' accelerations occur at the start of flexion ( + ) movements, and at the end of extension ( - ) movements, whi le ' - ' accelerations occur at the end of flexion ( + ) and start of extension ( - ) movements). When a neurone responded to more than 1 parameter of a movement the appropriate abbreviations were combined. For example, a neurone having a tonic discharge maximum at 70 ~ , and responding to flexion movements in a velocity dependent way was represented as pT0 V +.

Other superscripts have been appended to the response classifications of some velocity sensitive neurones. These are : wide (neurone responds across the full range of leg angles with high firing frequencies); mid (neurone responds only at mid-angles); 0 (neurone responds only close to 0~ 120 (neurone responds only close to 120~ For example, neurones classified as V - 0, V - mid, and V - x 20 are all extension sensitive, but respond in different ranges of leg angle. The difference between V - and V -Wide is slightly different, and more subjective. V - neurones respond to movements at most leg angles, but with relatively low frequencies of firing (50-250 Hz). Firing was stronger in some parts of the response range than in others. Neurones with the additional wide classification responded with higher firing frequencies (100-350 Hz), even to low velocity movements. In particular, the firing frequencies of 'wide' neurones were uniform across the full range of leg angles.

In the text a superscript x has been used in places to denote groups of response classes. For example, velocity sensitive neurones (extension or flexion) with any sort of acceleration component (including ?A § or ? A - ) could be referred to as Vx?A x.

Discriminant analyses

Theory. Discriminant analysis is a group of statistical procedures with the general function of revealing relationships between observations (in this case, neurones) placed in a priori groups (here based on physiological response classes) and a range of measured variables (aspects of morphology).

When many variables are measured from each of a large number of individuals comprising a number of distinct groups it is difficult to intuitively determine which variables are impor tant in characterising individuals in one group as different from individuals in another group. Generally there will be p variables (Xt, X2 . . . . , Xp) measured from each individual, and m possible groups. Canonical discriminant analysis determines functions of the variables X1, X2, .... Xp that in some sense separate the m groups as well as possible. The simplest method is to calculate a linear combinat ion of the variables:

Z = a l X l + a 2 X 2 + . . . + a p X p

The groups can be well separated using Z if the mean value changes markedly from group to group, but is relatively constant within each group. The coefficients al , a2, ..., ap are chosen so as

% A

I I I I I

cl

Anter ior

M e d i a l Latera l

sterior

n l

n5

~ 100/~m i

J

Fig. 1. Locust metathoracic ganglion containing the central projections of a single stained femoral chordotonal organ neurone. The ganglion is viewed from its ventral aspect, with anterior at the top. The midline is indicated by the vertical dashed line to the left. The main leg nerves are numbered (nl -n6) . The construction of the overlying grid is described in the text. All other illustrations of ganglia in this paper are in the same orientation as Fig. 1

to maximise the F ratio (Mb/Mw) (mean square between groups/ mean square within groups) in a one-way analysis of variance (ANOVA), i.e., a suitable function for separating the groups can be defined as the linear combinat ion for which the F ratio Mb/M w is as large as possible. It may be possible to determine several such linear functions for separating groups (generally the number of functions is the smaller o f p and m - l , say s). These are referred to as canonical discriminant functions. The first function

Z i = a l t X i + a i 2 X 2 + . . . + a i p X o

gives the maximum possible F ratio on a one-way A N O V A for the variation within and between groups. The second function

Z2 = a2tX1 + a22X2 +...a2pXp


gives the maximum possible F ratio on a one-way ANOVA subject to the condition that there is no correlation between Z1 and Z2 within groups. Further functions are defined in the same way, such that they are uncorrelated with all of the previous functions. The canonical discriminant functions Z1, Z2 . . . . , Z~ are linear combinations of the original variables chosen in such a way that ZI reflects group differences as much as possible; Z2 captures as much as possible of the group differences not displayed by Z1 ; Z3 captures as much as possible of the group differences not displayed by Z1 and Z2 ; etc. The first few functions are usually sufficient to account for virtually all the important group differences, thus reducing the complexity of the data to a form that can be represented in 1-3 dimensions by plotting individual observations on axes representing these first 1-3 functions. It is possible to test if each canonical function explains a significant proportion of the total variation. In addition, the relative importance of each variable in each canonical function is reflected by the size of its assigned coefficient.

Discriminant analyses in this study. In the present study, canonical discriminant analysis was used to discriminate between neurones placed a priori into physiological response classes (groups). The variables used to construct the discriminant functions were mea- sures of morphological features such as branch lengths and density of branching (see below).

The 9 physiological response classes and 30 subclasses of response which were used in the analyses are listed in Table 1. See Matheson (1990a, b) for further details of response types and terminology. A subsequent paper (Matheson, unpublished) examines in detail the range fractionation of mtFCO neurone responses.

The morphological variables used for the discriminant analyses were derived from cobalt-stained neurones drawn in wholemount from the ventral aspect. They were:

1. The presence of a branch in the anterior medial bundle; 2. The presence of a branch in the posterior medial bundle; 3. The presence of branches in both medial bundles; 4 The presence of primary dorso-later.~l brancb(es); 5. The distance from the midline to the termination of the

branch in the anterior bundle; 6. The distance from midline to the termination of the branch

in the posterior bundle; 7. The length of the longest dorso-lateral branch; 8. The area covered by branches; and 9-32. The density of branches in each of 24 grid squares (see below

for further details).

Distances were standardised as a percentage of the length of the ganglion (dashed line in Fig. 1), while area was expressed in arbi- trary units divided by the square of the length of the ganglion. This was considered necessary to counter variation in ganglion size related to different sized individuals, and caused by possible shrink- age during tissue processing. Where there was no branch in the anterior or posterior bundles, the distance given (at variable 5 or 6) was arbitrarily set to 25 (for anterior) or 35 (posterior) units - the approximate average standardised distance from the midline to the main neurite at the two levels.

The grid used for the last 24 variables was based on neurone morphology, rather than on landmarks visible on the surface of the ganglion (which are known to vary considerably between individuals (Burrows and Hoyle 1973)). Neuronal landmarks should be much more consistent because the paths taken during development by afferent axons are guided by genetically determined cues (stepping stones, and discrete labels on different pathways) (Bas- tiani et al. 1985; Bate 1976). It is apparent from the whole-organ locust FCO projections of Field and Pfl/iger (1989) (msFCO distal scoloparium) and Pfl/.iger et al. (1988) (mtFCO) that the most reliable features of filled FCO neurones are: (1) the location of the first main lateral branch (usually an anterior dorsal branch); and (2) the points where the 2 prominent medial branches leave the main neurite. The branch-point of the first main lateral branch was used as the starting location of the grid (* in Fig. 1). From this point a

straight line (A-B in Fig. 1) was projected along the axis of the main neurite (which was usually quite straight). A second (temporary) line (C-D in Fig. 1) was drawn along the anterior medial branch (where present) to intersect with the first line. The point of intersec- tion of these 2 lines (** in Fig. 1) was used as the second fixed point. This second point usually corresponded well with the location at which the anterior medial branch left the main neurite. The distance between the points was divided by 4 to give a convenient mesh size, and a grid of 24 squares was projected over the ganglion (Fig. 1). Three neurones did not possess 'main' lateral branches: in these cases the first visible lateral branch was used. Some neurones did not possess branches which clearly ran in the anterior medial bundle. In these cases it was possible to approximate the second fixed point by using the location where one or more fine branches left the main neurite to travel anteriorly and laterally for a short distance (arrow in Fig. 1). The main neurite generally ended at this point. Grid squares were numbered from 1 to 24 as shown in Fig. 1.

The density of branches in each grid square was scored on a 3 point scale: 0 = none; 1 =branches in less than 50% of square; 2 = branches throughout square.

Discriminant analyses were used to: (a) indicate neurones with branching patterns markedly different to others in the same response class; (b) test which morphological variables are the best predictors of response class, and (c) provide a graphical representation of the grouping of the response classes. The SAS procedures (SAS Institute Inc.) used were: (a) Proc discrim (discriminant function analysis based on the pooled covariance matrix); and (b) Proc Candisc (canonical discriminant analysis). These procedures were each run twice: once with the dataset divided into 9 main response classes; and again with the same dataset subdivided into a total of 30 subclasses.

Results

Background

Cobal t fills of m t F C O neurones were generally of good qual i ty (i.e., only a single axon stained, and there was good filling of fine branches), a l though a few appeared unusua l ly 'b lebby ' . To lay a g r o u n d w o r k for describing the project ions of indiv idual receptors I will first sum- marise the overall project ion pa t te rn described f rom whole-nerve cobal t backfills of the m t F C O by Burrows (1987) and Pfl/iger et al. (1988). The neurone i l lustrated in Fig. 1 (present paper) has branches in all regions occupied by neurones in whole-nerve fills, and therefore provides a useful reference for overall morphology. Note that the gangl ionic neuropi les and tracts referred to below are no t visible in Fig. 1, bu t are i l lustrated as landmarks in subsequent transverse sections (e.g., Fig. 3).

After enter ing the gangl ion via n5 the F C O axons immediate ly give off dorso-lateral branches which run anter ior ly and dorsally into the anter ior and poster ior lateral associat ion centres ( aLAC and pLAC) . The primary neuri tes con t inue to run anter ior -media l ly in a root o f n5 (probably 5ii or 5iii) giving off branches a long the way. Two parallel bundles of collaterals extending to wi thin 50 ~tm of the midl ine are p r o m i n e n t and consis tent (I term these bundles the anter ior and poster ior medial bundles) : the anter ior medial bund le passes between the ventral in termediate (VIT) and ventral lateral (VLT) tracts before jo in ing into dorsal commissure III (DCII I ) where it ends. The poster ior medial bund le passes between VIT and the dorsal in termediate and dorsal medial


(DIT, DMT) tracts before entering DCIV and terminat- ing approximately 50 ~m short of the midline. Most of the remaining branches lie at the level of VIT and the median ventral tract (MVT). No branches are found in the ventralmost or lateral ventral association centres (vVAC, IVAC).

The first discriminant analysis (SAS Proc Discrim) was used to identify misclassified observations (i.e., neurones with branching patterns significantly different from others in the same physiological response class). The SAS procedure identified misclassified observations by cal- culating (on the basis of morphological similarity) the probability that a given neurone would be classified into each possible class. Neurones which had been put by the investigator into a class other than that with the highest probability were termed misclassified. This preliminary result was used as a guide for reassessing individual neurone's classifications, and as a check of the accuracy of the dataset before running the canonical discriminant analyses described below. Neurones which the SAS procedure indicated were misclassified were only reclassified if a re-examination of their original response data revealed an error of interpretation (i.e., neurones were not reclassified solely on the basis of the discriminant analysis). Of 20 neurones indicated as 'misclassified', only 1 was found to be clearly in error, and was reclassified. The outcome of this initial discriminant analysis is not formally presented as Results.

Central projections of mtFCO neurones

Different neurones clearly had different patterns of central morphology. However the primary axon of each neurone ran as far anteriorly as the base of DCIII (the point where the anterior medial bundle turns medially and dorsally to enter the commissure). Neurones which had a branch in the posterior medial bundle always had a branch in the anterior medial bundle. Some neurones had anterior but not posterior medial branches.

The canonical discriminant analysis based on the 9 main response classes produced 8 canonical functions to describe the variation in the morphological data. The first 2 functions were able to explain significant amounts of this variation (P < 0.0005 for both), while the third was significant at the 1% level (P---0.095). The individual observations (neurones) were plotted on axes representing the first 2 canonical functions to yield the groupings shown in Fig. 2. It is clear that neurones with the same response classification were grouped together (groups encircled by solid lines). This showed that at least some of the morphological variables were correlated to response type. The analysis indicated the relative importance of each morphological variable in providing separation along each axis of Fig. 2. In this case, the 4 most important variables were the same for both plotted axes (length of posterior medial branch> presence of posterior medial branch > density of branches in grid 5 > length of anterior medial branch). On the first axis, the next (fifth) most important variable was the overall area covered by branches, while on the second axis, the

fifth-most important variable was the density of branches in grid 23.

The final groupings can be interpreted in terms of both morphology and response. The neurones in classes 5, 6, and 9 (right side of Fig. 2) have both anterior and posterior medial branches, and are extension sensitive (have a V- component in their response). These groups include a total of 29 neurones. Neurones in groups on the left side of Fig. 2 generally lack a posterior medial branch and are mostly flexion sensitive. Because the medial branches are the most prominent morphological features of many neurones (and are an important criterion for separation along the canonical axes) it is convenient to divide the results section on this basis also. I will therefore concentrate on the right-most groups of Fig. 2 (i.e., extension sensitive neurones) in the next section, before going on to examine the classes which fall further to the left.

Neurones grouped on the right side of Figure 2

The general form of the branching pattern of neurones in the right-most groups is illustrated in Fig. 3A (a V - ? A - neurone). Sections (Fig. 3Bi-v) were taken through this preparation at locations i-v. In all illustrations of sections the main neurite is indicated with an asterisk if it is clearly identifiable. The anteriormost branches were found at the level of, and lateral to VIT (Fig. 3Bi). The anterior medial branch in DCIII (Fig. 3Bii) terminated in a small cluster of very fine branches, some of which plunged ventrally, medial to VIT (Fig. 3Biii). The prominent soma and neurite of the FETi motor neurone can be seen in Fig. 3Bii, iii. A section cut at location iv of Fig. 3A (Fig. 3Biv) revealed the posterior medial branch running in DCIV to end near DMT. Dorso-lateral branches reached nearly to the level of DIT, while some ventral branches lay below VMT. A section (Fig. 3Bv) cut as the main neurite reached the edge of the neuropile contained further dorso-lateral branches, and one dorsal branch further medial (arrow).

Only 5 of the 38 neurones which had a V- component did not have branches in both medial bundles. Of these five, 3 were poorly filled, while 1 had an unusual response (it was very slow to stop firing following a stimulus). The single remaining exception is illustrated in Fig. 5B. The poorly filled neurone from class 6 was positioned much lower and further left in Fig. 2 than others in its class (dotted outline).

A second canonical discriminant analysis was carried out on 30 subclasses of response (Fig. 4A). The first axis of this analysis was significant at P < 0.0005 level, while the second axis was only significant at P < 0.1 level. The order of importance of the first 5 morphological variables for placing neurones along the two plotted canonical axes are respectively: (axis 1: length of posterior medial branch > presence of posterior medial branch > density of branching in grid 5 > area > density of branches in grid 12; axis 2: presence of posterior medial branch > length of posterior medial branch > length of anterior medial branch>presence of anterior medial


0

.o o

t~ ~

0

~J

4

3

2

0

--I

--2

--3

--4

--5

--5

1 2 2

2

2

11

1

8,

7 7

8

. 3 7

- - '4 -- '3 -- '2 --' 1

3 3

\

6 i g g Canonical function one

Fig. 2. Neurones plotted on the first two canonical functions derived from the analysis of response CLASSES. The position of each neurone is shown by its class number. For example, each position- and-acceleration sensitive (class 2) neurone is represented by the symbol 2 (all are plotted towards the upper left). Solid lines have

been drawn around the boundaries of each response class. Five neurones which fell well away from others in their class are indicated by dashed circles, and are mentioned in the next. All the neurones with an extension-velocity sensitive component are contained within classes 5, 6, 8, and 9 (upper right)

branch > area). Figures 4B-E are simply subsets of the groupings shown in Fig. 4A, and are referred to in the following descriptions for the sake of clarity. Individual observations have been removed f rom these figures, and the subclass names and responses added. Figure 4A, on the other hand provides an overview of all the groups.

Classes 5, 6, and 9 (Fig. 2) include subclasses E, F, H, I, R, S, and T (Fig. 4B). For the purposes of this more detailed description it is useful to include subclasses of class 8 (so that all neurones with a V - component are considered together): these are subclasses N, O and P.

No neurones in subclasses D, G, or Q were successfully stained (although several have been physiologically characterised in these and other experiments), so these subclasses are not represented in Fig. 4B.

Pure extension-velocity sensitive units

Pure extension-velocity units fell into 3 groups related to their subclasses (O, N, P). As suggested by their location on the right of Fig. 4B, neurones in subclass O (V -told)

A

-- ii~ l

IV

J V

DCIV dD~MT

/

Bi- S"

Biii ~ ~ ~__~_/5~e ~ (S~

Bv

0 0 ~ ~) a - ~ ,

100,um

Fig. 3. A Central projections of a V - ? A - (class 9) neurone. The exit point of the medial trachea from the ganglion is shown as a dashed circle overlapping with the branches in the anterior medial bundle.

Sections were taken at locations i-v. B Transverse sections through the ganglion illustrated in A. Each section is mentioned in detail in the text


A 6

5

4

3

o 2

.o I

0 ~

0 --1

" R

--2

--4"

--5

�9

n

i �9 �9 w ,

--5 --4 --,3 --2 --I

Canonica l Fig. 4A-E. Neurones plotted on the first 2 canonical functions derived from the analysis of response SUBCLASSES. A The location of individual observations within subclasses. Each neurone is represented by its subclass number. For example, P12~ (subclass F) neurones are represented by the symbol F (all are plotted near the lower right). A break-down of this figure is presented in B--E, in which individual observations are omitted, but subclass designa- tions and response types are added. B Outlines of extension-veloc-

0 1 2 .3 4 5

funct ion one ity, position-and-extension-velocity, and extension-velocity-and- acceleration sensitive subclasses. C Outlines of flexion-velocity, and flexion-velocity-and-acceleration subclasses. D Outlines of position- and-flexion-velocity, and position, -flexion-velocity, -and-acceleration sensitive subclasses. E Outlines of acceleration, and position- and-acceleration subclasses. The classification of subclasses bounded by dashed lines is discussed in the text

had branches in both medial bundles (Fig. 5A). The 2 neurones in subclass N (V-) had relatively sparse branching, and no posterior medial branch (Fig. 5B). One of these had a branch in the anterior medial bundle, but the other (which was poorly filled) did not. Two V -wid~ (subclass P) neurones were stained (1 poorly). Both had a short branch in each medial bundle, and sparse branching elsewhere (Fig. 5C).

Posit ion-and-extension-velocity sensitive neurones

Neurones sensitive to position and extension velocity (Pmi~V- (subclass E), and P12~ (subclass F)) fall into 2 groups to the right of Fig. 4B (no P~ (subclass D) neurones were successfully stained). PmidV- neurones (E) lie higher than P12~ (F) neurones. Neurones with maximal tonic sensitivity near 120 ~ generally had longer


o

.o

.R

o

6

5

4

,3

2

1

0

- 1

- 2

- 3

- 4

- 5

6

5

4

.5

2

1

0

- 1

- 2

- 3

- 4

- 5

B

T: V-'~?A -

s: v-o,0,:A- o \ I /"- t :=,b N: V ' : t', E: P " i " V ' ~ /

H "'~ R: V-?A- ' = = ~ . J / H: Pm'dV-?A-

1: P l 2 ~

D C: W~ -

B: Pm"JV+ ?A +

C ~ 8 : P~176

C

J J: V*

K:V +wide

\ I %,. ~.j L: V +~ " ~

E

5: Pmid?Ax ..-'~ ~ . ~ . "]3: ?AS 2: ?A-

4: P~

- ' 5 - ' 4 - 3 - ' 2 - ' 1 () i 2 ; A ;

Fig. 4B-E Canonical function one

branches in the medial bundles, and longer dorso-lateral branches (i.e., reaching further anterior) than did mid- position sensitive neurones (compare Fig. 6A and B). Because of the small sample sizes (n = 3,5) a t test could not be used to determine the statistical significance of these specific differences. However, the clear division into two groups in Fig. 4B shows that these subclasses do have real morphological differences.

Position, -extension~veloeity, -and-acceleration sensitive neurones

Neurones sensitive to position, extension-velocity, and acceleration (pmidV-.9A- (subclass H), and P12~

(subclass I) also form 2 groups in Fig. 4B (no P~ - (subclass G) neurones were stained). The distribution pattern is similar to that for position-and-velocity receptors, with mid-range units (subclass H) plotted higher on the Y axis than corresponding flexion-range (I) units. The loci of pmidV- and pmidV-?A - neurones overlap, as do the loci of Plz~ and P12~ neurones. The morphology of pxV-?A- units is subjectively very similar to that of pxv- units (Fig. 6A, B), and is therefore not illustrated separately. Neurones in this group with tonic maxima at 120 ~ had longer anterior medial branches than those with maximal tonic firing at mid- angles (t test: P < 0.03, n = 13). The lengths of posterior medial branches and dorso-lateral branches did not differ

A B

"-'2~


C

J Fig. 5A-C. Central branching patterns of extension-velocity sensitive neurones (Class 8). A V-mid (subclass O) neurone. B V- (subclass N) neurone. C V wid~ (subclass P) neurone

A

Fig. 6A, B. Position-and-extension-velocity sensitive neurones (Class 5). A pmidV- (subclass E) neurone. Compare the length of the medial branches with those of the neurone in B. B px2~ (subclass F) neurone

significantly between the types (t test: P- -0 .16 and 0.23 respectively).

Extensh)n-velocity-and-acceleration sensitive neurones

The groupings of extension-velocity-and-acceleration sensitive units ( V - ? A - (subclass R), V-~id?A - (S), and V-wido ?A- (T) did not overlap on the canonical discriminant plot (Fig. 4B), but their branching patterns appeared very similar. A representative V - ? A - neurone is illustrated in Fig. 3A. This morphology is similar to that

of the position-and-velocity neurones described above (e.g., Fig. 6A, B. Also note the large extent of overlap between subclasses R (V -?A- ) , H (pmidV-?A-), and E (pmidV-) in Fig. 4B).

Relationship between phasic and tonic position sensitivity and branching pattern

The results of the canonical discriminant analysis suggested that a further investigation of the relationships between the length of various branches and positional sensitivity would be valuable. All neurones with branches in both the anterior and posterior medial bundles which had clear position-dependent maxima in either their tonic or phasic components were pooled. The standardised length of the longest dorso-lateral branch, and the standardised distance from the midline of the ganglion to the termination of each medial branch were regressed against the leg angle giving maximal tonic or phasic response (Fig. 7A-C). The slopes of the regressions for both medial branches (Fig. 7A, B) are significantly different from zero (anterior: P < 0.05, posterior: P < 0.0001), in- dicating that these branches tend to terminate closer to the midline in extension-velocity sensitive neurones with phasic or tonic firing maxima near 120 ~ . There is considerable variation about both lines (r z values are 0.14 and 0.56 respectively). The slope of the regression for length of dorso-lateral branches (Fig. 7C) was not significantly different from zero ( P = 0.25).

Neurones grouped on the left side of Figure 2

This section examines the groups which lie to the left of Fig. 2. These neurones are either flexion sensitive, or are position-and-acceleration sensitive. Neurones in class 8 (V-), some of which lie to the left of Fig. 2, have been presented along with other extension sensitive neurones in the previous section.


A 18-

16-

o 141

.~ 12-

lOq

= 4-

2 i

0 0

B 18- 16

~ t4

! 8 .

z'0 i0 8'0 8'0 160 it0

�9 �9 �9

lO-

6q

2-

o o z.0 io 6'0 g0 160

C 30 @

15 I

O/ 0 z'0 4'o 8'o go 16o

Leg angle giving m a x i m u m fir ing

Fig. 7A-C. Linear regressions of medial and dorsal branch lengths against an aspect of physiological response: the leg angle giving maximal phasic or tonic firing. All neurones with a branch in both medial bundles were pooled. A Length of anterior medial branch (Y = --0.023 X+9.1 . r 2 = 0.14). B Length of posterior medial branch (Y = - 0 . 0 6 3 X + 1 2 . 1 . r 2 = 0.56). C Length of dorso- lateral branch (Y = 0.022 X + 17.1. r 2 = 0.22)

4C. Units which fire only as the leg reaches full flexion (40-0~ V+~ subclass L) are grouped separately from those which fire over much of the range (V + : subclass J). Neurones which fire equally strongly during flexions at all angles (V+wlde: subclass K) are clustered in a central location, overlapping the other two distributions.

Seven of the 10 V +~ (subclass L) neurones stained well. All 7 had a short branch in the anterior medial bundle (e.g., Fig. 8A). These neurones had a small but dense area of branching in the dorso-lateral neuropile, with few other branches anterior to the main axon. Two neurones, however, had 2 branches extending anterior to the main axon. One of these neurones had an unusual response: its phasic discharge continued for a short time after cessation of the movement stimulus (other V + neurones only fired during movements). Extension sensitive neurones generally had a region of branching anterior and lateral to the base of the branch in the anterior medial bundle (see arrow in Fig. 1), but this was poorly developed or absent in flexion sensitive neurones (Fig. 8A).

Two V § (subclass K) neurones had a morphology similar to that just described for V §176 neurones (see Fig. 8A), and were grouped accordingly in Fig. 4A. The third neurone classified as V +wide had a different morphology, and was placed higher in Fig. 4A by the analysis. It had a well developed dorso-lateral region of branching but no major branches further medial (Fig. 8B). The main axon was well filled with dye, giving no cause to suspect that this pattern was an artifact. This neurone closely resembles some WV + and px?A neurones described later, and had erratic 'tonic' firing somewhat similar to that of pxV+ or Px?A units. It was not possible to unambiguous- ly determine if any of this ongoing activity was real tonic firing (because of superimposed bursts of spikes caused by leg tracheal movements), and so the classification as V +w~d* was tentative. Its morphology suggests that it would be better classified as a phasic-tonic unit.

Three neurones were classified as V § (subclass J in Fig. 4C). All these neurones were placed higher in Fig. 4C than any V § o neurones. However, one of these cells was poorly stained, and another (uppermost J neurone in Fig. 4A) had a questionable identity because of prob- lems during dye-injection. Disregarding this last cell, there is then little distance separating V + (subclass J) neurones and V + o (subclass L) neurones in Fig. 4C. Their morphologies are correspondingly similar (Fig. 8C), although the V + neurone had a prominent branch anterior to the main axon (Fig. 8C, arrow).

Flexion-velocity sensitive neurones

Neurones with a flexion-velocity sensitive (V +) component in their response were included in classes 3 (PxV+), 4 (PxV+?A+), and 7 (V+). Canonical discriminant analysis based on response classes placed them in 3 largely overlapping groups to the lower left in Fig. 2. The canonical analysis based on subclasses (Fig. 4C, D) provided further separation for some groups.

Pure velocity-sensitive units. Pure V + units (class 7, subclasses J -L) are spread out along the vertical axis of Fig.

Flexion~velocity-and-aceeleration sensitive neurones. The two V+?A + neurones stained (class 7, subclass M) have very different morphologies from one another (Fig. 9A, B). This is reflected by their wide separation along the first canonical axis in Fig. 4C. Both cells were well filled, and clearly characterised. The neurone shown in Fig. 9A responded to flexions in the range 0-120 ~ (but was far less sensitive to velocity than was the V + wide neurone illustrated in Fig. 8B (100 Hz cf. 200 Hz), while the other V + ?A + neurone (Fig. 9B) only responded to movements close to 0 ~ It was also relatively insensitive to velocity (firing frequency of < 100 Hz during a 500~ - 1 movement). The


100urn

Fig. 8A-C. Central projections of flexion-velocity sensitive neurones (Class 7). A V §176 (subclass L) neurone. B V +wi~e (subclass K) neu-

rone with branching largely restricted to the dorso-lateral neuropile. C V + (subclass J) neurone

Fig. 9A, B. Comparison of the two V + ?A + (subclass M) neurones stained. The neurone in (A) responded to movements between 0-120 ~ while the neurone in (B) only fired during movements close to 0 ~

branching of this neurone resembles that of the V +~ neurone illustrated in Fig. 8A, but it has an additional branch in the posterior medial bundle. This is unusual for a flexion sensitive neurone.

Posi t ion-and-f lexion-veloci ty sensitive neurones, pxv+ neurones (class 3 in Fig. 2) comprise subclasses 6, 7, and 8 in Fig. 4A, D.

P~ neurones (subclass 6) have a branch in the anterior medial bundle (often longer than the examples in Fig. 10A, B), and well developed dorso-lateral rami- fications (Fig. 10A). Most also have a branch near the

posterior medial bundle, but in many cases it appears not to project far enough anterior to join properly into the bundle (i.e., DCIV). In a section cut at location A of Fig. 10B, branches lie lateral to VIT (Fig. 10C). This neurone has no medial branches in DCIV (Fig. 10D). One lateral branch (arrow) lies ventral to a group of neurites emerg- ing from ventral lateral somata. The anterior-most dorso-lateral branch appears in aLAC. Further posterior (Fig. 10E) the 'posterior medial' branch can be seen lateral to, and just below the level of DIT. Some dorso- lateral branches are close to roots of n3. The ventralmost branch continues to run ventral (but lateral) to MVT. Further posterior (Fig. 10F), dorso-lateral and 2 medial branches are apparent in pLAC.

Two P~ neurones did not have a branch in the anterior medial bundle. In one case this was the result of a poor fill. In the other case, the neurone was well filled, but had only very sparse branching medially, and 2 dorso-lateral branches (Fig. 11A). This morphology is therefore very similar to that of the V +wlde neurone illustrated in Fig. 8B. The P~ neurone illustrated in Fig. 11A was also sensitive to movements throughout the leg's arc, responding clearly to velocities as slow as 20Os - 1 with bursts of spikes. Its response and morphology were quite similar to those of the neurone classified as P~176 (Fig. 11B) and those classified as pmidW+ (Fig. 11 C, D) (all of which also responded phasically throughout the leg's arc of movement, and had tonic firing which was only weakly influenced by leg angle). All these neurones should probably be classified together as being relatively insensitive to position, but very sensitive to flexion velocity.

Two neurones which were classified as P~ o (subclass 8 in Fig. 4D) were placed by the analysis slightly lower on the canonical discriminant plot than P~ (subclass 6) neurones, but there was no obvious subjec-

A j ' B ~ A

B

C

lO0,um

E ~ F

\ D~MT

IVAC

Fig. 10A-F. Position-and-flexion-velocity sensitive neurones (P~ § class 3, subclass 6). A, B Two examples of the central branching pattern. Sections illustrated in C-F were cut at locations A - D of B.

L~ 1 fo ~ x ; ~

i

p~c \ 1

) N2

100urn

C-F Sections cut through the P~ (subclass 6) neurone of B. Each section is discussed in the text. The arrow in D points to a ventral branch of the neurone

114

A

100tim

,- \ ' . ' t

C

B

100/~m

"d

T. Matheson: Chordotonal organ central projections

2',

B C - D

~-, E ( ~" F ;

/

D

lOOvm

Fig. IIA-D. Position-and-flexion-velocity sensitive neurones (Class 3). A Central projections of a P~ (subclass 6) neurone with an unusual branching pattern compared to other subclass 6 neurones. B P~ § wide (subclass 7) neurone. C, D Two examples of the central projections of l~idV § (subclass 9) neurones. All these (A-D) neurones had strong velocity responses, but only weak position sensitivity

tive difference in their branching morphology f rom the pat tern illustrated for P~ + neurones (Fig. 10A, B). They are therefore not illustrated separately.

Position, - f lexion-velocity, -and-acceleration sensitive neurones. The two neurones in subclass B (PmidV+?A +) and the single neurone in subclass C (p2~ (Fig. 4A, D) all had a small region of branching in the dorso- lateral neuropile, and no major branches further medial (i.e., similar morphology to that illustrated in Fig. 11 for PxV+ units). The wide separation between the Pz~ - neurone and the Pmi~v'+?A+ neurones in Fig. 4A, D

Fig. 12A, B. Two examples of the central projections of pure acceleration (class 1, subclass 3) neurones. Sections illustrated in Fig. 13 were taken from locations A-F

appears to be the result of a slightly different path fol- lowed by the main neurite of the p2~ neurone.

Acceleration sensitive neurones

Pure acceleration units. Neurones with pure acceleration responses (class 1 in Fig. 2) formed a large group overlapping several other response types. Pure acceleration units had in common well developed dorso-lateral branches and a branch in the anterior medial bundle. Different subclasses of acceleration receptors, however, had consistent differences f rom one-another (presented in later paragraphs). Sections cut through a ?AS (subclass 3) neurone at the locations indicated in Fig. 12A revealed the anterior medial branch in D C I I I (Fig. 13A), the dorso-lateral branch anterior to the soma of FETi (Fig. 13B), and ventral branches below the level o f VIT (Fig. 13C). A section cut f rom a similar position in the ganglion of Fig. 12B (subclass 3 neurone) illustrated these ventral branches reaching as far medially (but ventral to) the centre of VIT (Fig. 13D). Further posterior (Fig. 13E) dorsal branches were evident in pLAC, while some medial branches projected towards MVT. A section cut as the main neurite reached the margin of the neuropile (Fig. 13F) revealed a large medial branch projecting as far dorsal as VLT. A lateral dorsal branch (arrow) reached even further dorsal. All pure acceleration receptors had similar degrees of dorso-lateral branching, and the density of other branches was also fairly consistent. There were however, differences in the lengths of medial branches which related to response classes. These differences must therefore be important in providing the clear separation between subclasses 0, 1, 2, and 3 Fig. 4E, and are presented in the following paragraph.

All 3 neurones which responded only to negative accelerations (A- (subclass 2) in Fig. 4A, E) had well

A ~ B ~ ~ \ /

O @ @

" , p/

C

~ T

Ob

E

oO ,k-

J ?

F

'1

Fig. 13A-F. Sections through the pure acceleration (subclass 3) neurones illustrated in Fig, 12A, B. Each section is described in the text. Asterisks indicate the main neurite where it can be clearly

distinguished from other branches. Arrow in F points to a dorsal lateral branch

116 T. M a t h e s o n : C h o r d o t o n a l o rgan central project ions

A

100 l~m

B F - 'r

lOOum

Fig. 14. A Cent ra l pro jec t ions o f an ?A - (subclass 2) neurone . B Cent ra l pro jec t ions o f an ?A + (subclass 0) neu rone

developed branches in both medial bundles (Fig. 14A). Four out of the 5 neurones which responded to positive accelerations (?A + (subclass 1) had a well developed branch only in the anterior bundle (Fig. 14B). The other ?A + neurone did not have either medial branch (but was similar in all other respects). Neurones with a ?AS response (subclass 3) exhibited variations between these extremes: one had both medial branches well developed, while others had (to varying degrees) shorter posterior medial branches (n = 8) (these intermediate forms are not illustrated separately because the only apparent difference from the morphology shown in Fig. 14A, B is the length of the posterior medial branch).

Position-and-acceleration receptors. Neurones in class 2 (Fig. 2) responded to accelerations and leg position. Subclasses 4 and 5 were based on the angle giving the highest tonic firing (P~ 4; P m'~.A x= subclass 5). The canonical discriminant analysis based on subclasses revealed considerable overlap between these two groups (Fig. 4A, E). There are, however, two distinct morphological types in each subclass, and in retrospect it is possible to attribute these to a different aspect of response: the strength of the relationship between leg angle and firing frequency (i.e., the accuracy with which the firing frequency codes for different static leg posi- tions). Some px?Ax units have a low frequency response (< 5 Hz) which is only weakly dependent on tibial position, while other px?Ax units fire at relatively high frequencies (20~40 Hz), and are strongly influenced by tibial angle. Neurones from either subclass which had low frequency tonic firing (n=4) had the same branching pattern as pure acceleration receptors (cf. Fig. 14B: i.e., well developed dorso-lateral branches, a branch in the anterior medial bundle, and a short (or no) branch in the posterior medial bundle). This pattern is not re-illustrated. None of these units had an ?A- component, and

none therefore had well developed branches in both medial bundles. One other unit had the same response, but a different branching pattern. However, this preparation was ambiguous because 2 cells had been stained with cobalt. In contrast, position-and-acceleration sensitive neurones which had high frequencies of tonic firing (n = 6) had a completely different morphology to that just described: they had no main branches leaving the axon anywhere along its length, and no branches in the medial bundles (Fig. 15A). Sections cut at the locations indicated in Fig. 15A indicated that the main neurite runs medially at about the same level as other FCO axons. In Fig. 15B this neurite can be seen just below the lateral margin of VIT. The most prominent medial branch clearly did not enter DCIV (Fig. 15C). A section further posterior (Fig. 15D) showed that the only significant lateral branch (arrow) runs slightly ventral to the main neurite, and not dorsal to it as might be expected.

One Pmid?A+ neurone had a different branching pa t - tern and an intermediate type response, but has been ignored because the quality of the recording declined suddenly just prior to staining (spike size decreased from 30 to less than 5 mV). It is likely that the electrode moved into (and stained) a cell other than the one that was recorded from.

Summary of gross morphological features

1. All neurones have a main neurite which extends as far anterior as DCIII, and all except some position-and- acceleration receptors have branches in the dorsal lateral neuropile (aLAC, pLAC).

2. Extension sensitive units, and acceleration sensitive units which fire at the start of extensions have branches in both medial bundles, and many large branches off the main neurite (Fig. 3A).

3. Tonic or phasic extension sensitive neurones with maximal responses near full extension have longer medial branches than those with maximal responses near flexed angles (Fig. 7A, B).

4. Neurones which respond to flexion velocity only, and neurones which respond to flexion velocity and position have a branch in the anterior medial bundle, and many large branches off the main neurite (Fig. 10A, B). Most have a short branch in or near the posterior medial bundle.

5. Neurones which respond to flexion velocity across the full range of leg angles, but which have a weak tonic (position) response (i.e., tonic firing frequency changes little when the leg is set at different angles) have branches in the dorsal lateral neuropile, but no other large branches off the main neurite (Fig. 11A-D).

6. Neurones with a strong position response and a response to acceleration have no large branches off the main neurite (Fig. 15A-D).

Discussion

To understand the functioning and role of sensory systems it is important to understand not only the responses


A )

A

B

C

100ttm

B

Y0

vLT

~ llX)

Fig. 15A-D. Central projections of a position-and-acceleration (class 2, subclass 5) neurone with restricted central branching (and high frequency tonic firing). Sections illustrated in B-D were cut at

locations A-C. Each section is mentioned in the text. Asterisks indicate the main neurite. Arrow in D points to the largest lateral branch, which is ventral to the main neurite

of the individual transducers, but also the functional connections they make with interneurones and motor neurones. When investigating a sense organ which contains only a single afferent neurone (e.g., some tactile hairs), it is possible to record from identified afferents in many animals while searching for and characterising the postsynaptic neurones (e.g., Hamon et al. 1990). The connections made by afferent neurones from multineu- ronal sense organs such as chordotonal organs are considerably more difficult to investigate because it is often not possible to record from individually identified sensory neurones in successive experiments, so each afferent must be morphologically characterised in every experi- ment. In addition, it is generally not possible to stimulate individual afferents separately (i.e., a physiological stimulus will activate many afferent neurones within the population, providing a barrage of inputs onto central neurones).

Because of these complications, even the most thor- ough studies of the central connections of joint chordotonal organs to date (e.g., Burrows 1987; Burrows et al. 1988; Biischges 1989, 1990; Laurent and Burrows 1988) have generally relied on extracellular recordings of the entire organ's response to imposed leg movements, rather than intracellular recordings from single sensory units. This approach allows many more interneurones to be investigated, but precludes interpretation of the specific connections made by individual FCO afferents and identified interneurones. In particular, although the FCO is known to contain neurones which respond to position, velocity, or acceleration, or to combinations of these parameters (Hofmann and Koch 1985; Hofmann et al. 1985; Matheson 1990a; Zill 1985), no studies which have demonstrated central connections of FCO neurones have illustrated the morphology of the afferent neurones. It has not yet been determined if all FCO neurones that


provide information about different aspects of a movement make the same connections with any given inter- neurone (Burrows et al. 1988). However, it is known that all extension sensitive neurones do not synapse onto one particular flexor motor neurone (Burrows 1987).

The work presented here is the first step towards understanding the specific connections of mtFCO sensory neurones. In conjunction with a previous paper (Matheson 1990a) it provides the first comprehensive survey of the responses and central projections of these neurones. It is clear from this work that different mtFCO neurones have different branching patterns within the metathoracic ganglion, and that the specific pattern of branching is related to the response of the neurone.

The canonical discriminant analyses provided a useful way of objectively detecting any relationship between branching pattern and response. The analyses allowed many variables to be considered simultaneously, and 'suggested' which variables would be worth further detailed analysis (i.e., those which were shown to be most important in each significant canonical function).

Previous attempts to classify neurones on morphological grounds have generally relied on numbering of major recognisable branches (e.g., Burrows and Wat- kins 1986; Nagayama 1989), or on subjective criteria such as overall shape (Nagayama 1989). Nagayama (1989) used a coarse grid of 4 rectangles to help classify the branching of local spiking interneurones in the locust. The most common approach has been to illustrate a few representative neurones in whole mount and in sections (e.g., Hamon et al. 1990; Murphey et al. 1980; R6mer et al. 1988). This approach certainly presents a simple pic- ture, but it does not allow any objective test of a suggested relationship between morphology and other parameters. Because of journals' space constraints, only a few of the many neurones recorded in a study may be drawn in this way. In contrast, by quantifying morphological parameters and presenting the results as a plot of canonical functions, all the neurones recorded may be presented (albeit in a symbolic form). Such plots quickly convey an idea of the relative separation between the proposed groupings. Drawings of representative neurones then illustrate the features typical of each group. The canonical discriminant analysis proved to be able to discriminate between some groups even when the branching patterns appeared subjectively similar. For example, although there was no subjective difference between the branching of P~176 and P~ neurones, they were clearly separated by the analysis (compare sublasses 8 and 6 in Fig. 4A). The analysis thus supports the a priori distinction between these response types, and provides enough evidence to justify further investigation of the underlying morphological differences.

Burrows (1987) first suggested that the projections of mtFCO neurones varied, but his 6 published stains were derived from a single whole-nerve cobalt backfill, and were therefore not physiologically characterised neurones. In addition, 4 of the patterns that Burrows illustrated have not been recorded in the present survey: they appear to be partial stains because in all cases they have fewer rather than more branches than similar neurones recorded here.

All mtFCO neurones stained in the present study had a main neurite extending as far anterior as DCIII (but not necessarily entering the tract itself), and (except for position-and-acceleration receptors with a high frequency tonic discharge), dorsal branches in the lateral neuropil (aLAC, pLAC). Pflfiger et al. (1988) note that mechanoreceptors which have collaterals in these lateral association centres usually lack branches in any of the other well-defined neuropiles (e.g., the ventral association centres (aVAC, pVAC, vVAC). This is apparently the case for femoral chordotonal organ neurones.

Neurones with a branch in the posterior medial bundle (DCIV) always had a branch in the anterior bundle (DCIII), but the converse did not hold true. The length of branches in the medial bundles was related to the response of the neurone. Extension sensitive neurones which responded most strongly (either phasically or ton- ically) near full extension had significantly longer anterior medial branches, and slightly longer posterior medial branches (not statistically significant) than similar neurones which responded maximally at more flexed angles. For pure acceleration receptors the length of the posterior branch (in DCIV) was most strongly related to response: posterior medial branches were long in A- units, but short or missing in A + units. Pure acceleration receptors which fired in response to accelerations of both signs (?AS) formed a continuum between these two extremes.

The present study was designed to investigate differences in central branching morphology between response classes, and therefore required a rather broad approach. A further survey concentrating on staining more neurones from a selected number of response classes is now needed to describe in detail the variation within classes and subclasses. In particular, the 3-dimensional nature of the branching pattern has not been quantified in the present study. For the dorso-lateral branches at least (and probably also for some medial branches), variation along the dorsal-ventral axis is likely to be at least as important as anterior-posterior or medial-lateral variation in determining the synaptic connections made by FCO afferents. Such a detailed study would uncover more subtle differences in morphology, and may help to distinguish between (for example) PW-, PxV-?A x, and V- ?A x neurones which subjectively appear very similar.

Neurones in two response classes had very restricted areas of branching. Position-and-flexion-velocity (PW +) neurones which had low uniform tonic firing over all angles, and which fired rapidly (>200 Hz) during flexions at all angles, had only a few branches in the dorso-lateral region, and no major branches further medial. These neurones must make many fewer synaptic contacts than the majority of FCO afferents. Some phasic-tonic neurones which responded to acceleration, and strongly to position (Px?A), had similar but even more restricted branching: they lacked the dorso-lateral branches. It is surprising that these groups are so morphologically similar, yet have virtually no response characteristics in common (i.e., weak position + strong flexion-velocity sensitive versus strong position+acceleration sensitive).

Neurones which are known to receive connections


from FCO afferents include motor neurones, spiking local interneurones (Burrows 1985, 1987), nonspiking local interneurones (Burrows et al. 1988), and intersegmental interneurones (Laurent and Burrows 1988; New- land 1990). All of these central neurones have at least some branches in the dorso-lateral neuropiles, and all except the non-spiking interneurones have medially di- rected branches. It is possible to predict that FCO neurones with restricted dorso-lateral branching are unlikely to make connections with non-spiking interneurones. Tibial flexor muscle motor neurones which are excited by tibial extensions have prominent and dense regions of branching near the 'posterior medial' bundle of fibres from the mtFCO (i.e., entering or very close to DCIV) (Burrows 1987). FCO afferents which project to this region are all extension sensitive. Some extensor motor neurones also have branches in this region, but they appear not to extend as far medial (compare Fig. 13B, D in Burrows and Pflfiger 1988). An intersegmental inter- neurone excited by tibial flexion has branches restricted mainly to lateral neuropiles, while another which is excited by both extension and flexion has, in addition, many more medial branches (Laurent and Burrows 1988) which could receive inputs from the extension sensitive FCO afferents which also project to this medial region.

In vertebrates and more recently in invertebrates it has been shown that at least some classes of afferent neurones project into the CNS in orderly arrays. In insects somatotopic (topographic), tonotopic, retinotopic, and developmental mapping have been described.

Murphey et al. (1980) showed that clavate hairs on the cercus of a cricket have central branching patterns that are determined by the 'birthday' and peripheral location of each neurone. Older hairs (those appearing at earlier moults) project further into the CNS and have more extensive branches than younger cells. It would be most interesting therefore, to determine if the FCO neurones with very restricted branching reach the CNS later in development than other FCO neurones. To date there has been no attempt to study the timing of development of neurones within the FCO.

Neurones in the insect visual system (Bate 1978), and auditory receptors (see below) are known to make orderly connections in the CNS. Interneurones postsynaptic to crayfish antennal chordotonal organ neurones have been shown to be arranged in an ordered way in the antennal neuropile (Taylor 1975). Unfortunately these interneurones were simply characterised by extracellular recordings, and no attempt was made to visualise their morphology. The central projections of the presynaptic antennal chordotonal organ neurones are not known. The most recent example of topographic mapping is that presented by Newland (1991): locust mechanosensory hairs on the proximal femur project further anterior in the metathoracic ganglion than do distal hairs on the femur, or hairs on the tibia and tarsus. In addition, leg hairs at any given distance from the body project central- ly in a way that reflects their position on the circum- ference of the leg: hairs on the medial surface of the leg project further medial than lateral hairs. The CNS therefore contains a three-dimensional representation of the surface of the leg.

Metathoracic FCO neurones with similar responses to one-another have somata in similar locations in the FCO (Matheson 1990a). The correlation of both soma location (Matheson 1990a) and central branching pattern (this paper) with physiological response does not necessarily lead to a meaningful correlation between soma location and central branching. For example, although acceleration (?A +, ? A , ?AS) and position-and-flexion- velocity (P~ sensitive neurones have somata in a very similar region of the chordotonal organ (near the dorsal attachment point), their central projections differ markedly (cf. Figs. 11B and 14A, B).

Insect auditory receptors have been studied in great detail. The tuning of cricket crista acoustica neurones correlates with their projections into the prothoracic ganglion (Oldfield 1983). These neurones possess scolopi- dia, and are therefore chordotonal neurones by defini- tion. The crista acoustica, however, does not span a joint in the way that the FCO does, instead it rests on a tracheal airsac apposed to the tibial tympanum. Crista acoustica neurones which are tuned to frequencies lower than 16 kHz project to the anterior of the ganglion, while those tuned to higher frequencies project further posterior within the anterior ring tract (aRT=auditory neuropil=mVAC of Pflfiger et al. 1988). Cells tuned to near the calling-song frequency (16-20 kHz) also have larger central arbors (Oldfield 1983; Rrmer 1985). Other chordotonal (vibration) receptors from the subgenual and intermediate organs (near the crista acoustica) project to a different region of the CNS, but also form an orderly array dependent on their frequency response (Rrmer 1985). In the locust the auditory receptor is the tympanal organ located in the abdomen (the crista acoustica is part of the tibial tympanal organ in Grylli- dae, Gryllacrididae, Tettigoniidea, and Stenopelmati- dae). The locust's abdominal tympanal receptors are divided into 4 groups on the basis of their peripheral morphology. Three groups respond to low frequencies, while the other group responds best at high frequencies. All 4 groups project to the aRT (= mVAC), where they arborise in distinct regions along an anterior-posterior axis (R6mer 1985; R6mer et al. 1988). Neurones which respond best to low frequencies have branches further anterior than do neurones which respond best to high frequencies. In addition, neurones with lower thresholds tend to be further posterior than neurones with higher thresholds at the same frequency (Rrmer 1985). It is difficult to draw comparisons between these results and those presented in my work because the stimulus proto- cols differ greatly (vibration versus linear stretch at different velocities). It may be useful to include some vibration stimuli in any future characterisation of FCO neurones.

In summary, I have shown that the central projections of the FCO are shaped on the basis of (a) the mechanical stimulus modality (position, velocity, and acceleration) monitored by each unit, and (b) (for some) the angle of the femur-tibia joint which produces maximal tonic or phasic firing. Taylor's (1975) work on crayfish antennal interneurones suggested that modalities of a stimulus monitored by a chordotonal organ may be represented in discrete locations within the CNS, but the present


work is the first to directly address this ques t ion for the afferent neurones themselves.

Acknowledgements. I thank Dr Larry Field for his enthusiastic support and encouragement throughout the duration of this work. Dr Russell Death's expert help with the statistics made it possible for me to contemplate the analyses presented in this paper. Dr Phil Newland and 2 anonymous referees made useful comments on a draft of the manuscript.

References

Bacon JP, Altman JS (1977) A silver intensification method for cobalt-filled neurones in wholemount preparations. Brain Res 138 : 359-363

Bastiani MJ, Doe CQ, Helfand SL, Goodman CS (1985) Neuronal specificity and growth cone guidance in grasshopper. Trends Neurosci 8 : 257-266

Bate CM (1976) Pioneer neurons in an insect embryo. Nature 260: 54-56

Bate CM (1978) Development of sensory systems in arthropods. In: Jacobson M (ed) Development of sensory systems. (Handbook of sensory physiology vol. IX). Springer, Berlin Heidelberg New York, pp 2-53

Braitenberg, V (1972) Periodic structures and structural gradients in the visual ganglia of the fly. In: Wehner R (ed) Information processing in the visual systems of arthropods. Springer, Berlin, pp 3-15

Burrows M (1985) The processing of mechanosensory information by spiking local interneurones in the locust. J Neurophysiol 54: 463-478

Burrows M (1987) Parallel processing of proprioceptive signals by spiking local interneurons and motor neurons in the locust. J Neurosci 7:1064-1080

Burrows M, Hoyle G (1973) Neural mechanisms underlying behav- iour in the locust Schistocerca gregaria. III Topography of limb motoneurons in the metathoracic ganglion. J Neurobiol 4:167-186

Burrows M, Pfliiger HJ (1988) Positive feedback loops from proprioceptors involved in leg movements of the locust. J Comp Physiol A 163:425-440

Burrows M, Watkins BL (1986) Spiking local interneurones in the mesothoracic ganglion of the locust: homologies with metathoracic interneurones. J Comp Neurol 245 : 29-40

Burrows M, Laurent GJ, Field LH (1988) Proprioceptive inputs to nonspiking local interneurons contribute to local reflexes of a locust hindleg. J Neurosci 8:3085-3093

Btischges A (1989) Processing of sensory input from the femoral chordotonal organ by spiking interneurones of stick insects. J Exp Biol 144:81-111

Biischges A (1990) Processing of sensory input from the femoral chordotonal organ by nonspiking pathways in a joint-control loop of the stick insect Carausius morosus. J Exp Biol 151 : 133-160

Dickinson M H, Palka J (1985) Physiological characteristics of sensory neurons of Drosophila wing correlate with central projections and developmental timing. Soc Neurosci Abstr 11:164

Field LH (1991) Mechanism for range fractionation in chordotonal organs of Locusta mioratoria (L.) and Valanga sp. (Orthoptera: Acrididae). Int J Insect Morphol Embryol 20:25-39

Field LH, Pfliiger HJ (1989) The femoral chordotonal organ: A bifunctional Orthopteran (Locusta migratoria) sense organ? Comp Biochem Physiol 93A:729-743

Hamon A, Guillet JC, Callec JJ (1990) A gradient of synaptic efficacy and its presynaptic basis in the cercal system of the cockroach. J Comp Physiol A 167:363-376

Hofmann T, Koch UT (1985) Acceleration receptors in the femoral chordotonal organ in the stick insect Cuniculina impiora. J Exp Biol 114:225-237

Hofmann T, Koch UT, B/issler U (1985) Physiology of the femoral chordotonal organ in the stick insect Cuniculina impiora. J Exp Biol 114:207-223

Horridge GA, Meinertzhagen IA (1970) The accuracy of the patterns of connexions of the first- and second-order neurons of the visual system of Calliphora. Proc R Soc Lond B 175:69-82

Laurent G, Burrows M (1988) A population of ascending intersegmental interneurones in the locust with mechanosensory inputs from a hind leg. J Comp Neurol 275:1-12

Matheson T (1990a) Responses and locations of neurones in the locust metathoracic femoral chordotonal organ. J Comp Physiol A 166:915-927

Matheson T (1990b) An experimental and descriptive analysis of the responses and peripheral and central organisation of the locust metathoracic femoral chordotonal organ - a leg joint position and movement detector. Ph.D. Thesis, University of Canterbury, New Zealand, 172 p

Matheson T, Ditz F (1991) A simple computer-controlled analogue ramp generator for producing multiple ramp and hold stimuli. J Neurosci Methods 39:45-52

Matheson T, Field LH (1990) Innervation of the metathoracic femoral chordotonal organ of Locusta migratoria. Cell Tissue Res 259 : 551-560

Mill PJ (Ed) (1976) Structure and function of proprioceptors in the invertebrates. Chapman & Hall, London, 686 pp

Murphey RK, Jacklet A, Schuster L (1980) A topographic map of sensory cell terminal arborizations in the cricket CNS : correlation with birthday and position in a sensory array. J Comp Neurol 191 : 53-64

Nagayama T (1989) Morphology of a new population of spiking local interneurones in the locust metathoracic ganglion. J Comp Neurol 283:18%211

Newland P (1990) Morphology of a population of mechanosensory ascending interneurones in the metathoracic ganglion of the locust. J Comp Neurol 299:242-260

Newland P (1991 Morphology and somatotopic organisation of the central projections of afferents from tactile hairs on the hind leg of the locust. J Comp Neurol 312:493-508

Oldfield BP (1983) Central projections of primary auditory fibres in Tettigoniidae (Orthoptera: Ensifera). J Comp Physiol 151 : 389-395

Palka J, Malone MA, Ellison RL, Wigston DJ (1986) Central projections of identified Drosophila sensory neurons in relation to their time of development. J Neurosci 6:1822-1830

Pfliiger HJ, Br/iunig P, Hustert R (1988) The organization of mechanosensory neuropiles in locust thoracic ganglia. Phil Trans R Soc Lond B 321 : 1-26

R6mer H (1985) Anatomical representation of frequency and inten- sity in the auditory system of Orthoptera. In: Kalmring K, Elsner N (eds) Acoustical and vibrational communication in insects. Paul Parey, Hamburg, pp 25-32

RSmer H, Marquart V, Hardt M (1988) Organization of a sensory neuropile in the auditory pathway of two groups of Orthoptera. J Comp Neurol 275:201-215

SAS User's Guide: Statistics. Version 5 Edition (1985) SAS In- stitute Inc. Cary, NC, USA

Taylor RC (1975) Integration in the crayfish antennal neuropile: Topographic representation and multiple channel coding of the mechanoreceptor submodalities. J Neurobiol 6:475-499

Zill SN (1985) Plasticity and proprioception in insects. I. Responses and cellular properties of individual receptors of the locust metathoracic femoral chordotonal organ. J Exp Biol 116:435-461

morphology of the central projections of physiologically ... · of physiologically characterised...

Documents