JOURNALOF Vol. 66, No.

NEUROPHYSIOLOGY I, July 1991. Printed

in U.S.A.

Responsesof Olfactory Receptor Cells of Spiny Lobsters to Binary Mixtures. I. Intensity Mixture Interactions CHARLES

D. DERBY,

MARIE-NADIA

GIRARDOT,

AND

PETER

C. DANIEL

Department of Biology, Georgia State University, Atlanta, Georgia 30302-4010 SUMMARY

AND

CONCLUSIONS

I. Neural coding of chemical mixtures was studied with the use of the peripheral olfactory system of the spiny lobster. The occurrence of mixture interactions (i.e., where the observed response to a mixture deviates significantly from the expected response) in individual cells and the effect of such mixture interactions on the coding of odorant intensity by populations of cells were examined. 2. Extracellular recordings of spiking activity of 98 primary olfactory receptor cells in the antennules were measured in response to seven compounds [adenosine-S-monophosphate (AMP), betaine (Bet), L-cysteine (Cys), L-glutamate (Glu), ammonium chloride (NH,), DL-succinate (Sue), and taurine (Tau)] and their binary mixtures. To identify mixture interactions, observed responses to a range of concentrations of a binary mixture were compared with the predicted responses based on three mathematical models: a single receptor model, which assumes that the two compounds of a mixture bind to the same receptor site; a multiple receptor model, which assumes that the two compounds bind to two independent receptor sites; and a mixed composition receptor model, which incorporates our current state of knowledge of transduction processes in olfactory receptor cells of spiny lobsters. 3. Mixture interactions in individual cells were common: statistically significant mixture interactions were observed in 25% of the possible cases (Fig. 5). Suppression was much more common than enhancement. 4. Mixture interactions had significant effects on the absolute response magnitudes for a population of cells, which could be used as the neural code for stimulus intensity in this system. These effects are called intensity mixture interactions (Figs. 6- 11). Intensity mixture interactions occurred for -50% of the binary mixtures and were almost exclusively suppression (Figs. 12 and 13). The intensity mixture interactions were concentration independent. 5. The results suggest that mixture interactions in individual olfactory cells can result in intensity mixture interactions in the neuronal population such that there is a decrease in sensitivity to binary mixtures relative to what is expected based on the responses to individual components of the mixtures. INTRODUCTION

Olfaction and mixtures Olfaction is necessary for the expression of many important behaviors of aquatic and terrestrial invertebrates and vertebrates, including the production and maintenance of many components of feeding, reproductive, and social behaviors. Although single chemical compounds can be effective stimuli in eliciting behavior, especially species-specific social or reproductive behaviors evoked by pheromones, mixtures of chemicals are the most common chemical signals for animals in their natural environments (Carr 1988; 112

0022-3077/9

1 $1 SO Copyright

Laing et al. 1989). Therefore an understanding of mechanisms of function of olfactory systems can be best generated by using as stimuli chemical mixtures composed of behaviorally relevant components. Coding of odorant quality and quantity Information is transmitted by olfactory receptor cells to the brain in the form of action potentials or spikes. Odorant quality and odorant intensity are thought to be coded by separate components of the spiking activity produced by the population of olfactory receptor cells in the olfactory organ of the spiny lobster: odorant quality by the relative amount, or pattern, of activity produced across the neuronal population (across neuron patterns, or ANPs); and odorant intensity by the total amount of activity produced by the neuronal population (absolute response magnitudes, or ARMS). The evidence for this assertion comes from electrophysiological and behavioral studies using single compounds or 23- or 41.compound mixtures and can be summarized by three points. I) ANPs are greatly affected by changes in stimulus type but are only slightly affected by changes in stimulus concentration (Derby and Ache 1984a; Girardot and Derby 1988, 1990a). 2) ARMS are greatly affected by changes in stimulus concentration and are much less affected by changes in stimulus type (Derby and Ache 1984a; Girardot and Derby 1988, 1990a). 3) The ability of ANPs to resolve differences in four mixture types parallels the ability of lobsters to behaviorally discriminate these same mixture types (Daniel and Derby 1988; Derby et al. 1989, 1990; Fine-Levy et al. 1989). Chemical quality and intensity are thought to be coded in this way in other chemosensory systems, such as the rat taste system (Ganchrow and Erickson 1970), peripheral taste systems of insects (Dethier 1976), and probably also peripheral olfactory receptor cell systems of vertebrates (Duchamp and Sicard 1984; Kauer 1987) and insects (O’Connell 198 1; Tichy and Loftus 1983). Mixture interactions A common finding in neurophysiological, behavioral, and psychophysical studies of the chemical sensesis that the response to a mixture cannot be predicted, even when the responses to the individual components of the mixture are known. Such phenomena are called mixture interactions (Laing et al. 1989). Mixture suppression, where the response to a mixture is less than the predicted response, is by

0 199 1 The American

Physiological

Society

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

INTENSITY

MIXTURE

far more common than mixture enhancement, where the response to a mixture is greater than predicted. Mixture interactions have been identified in the olfactory and gustatory systems of vertebrates and invertebrates (e.g., Dethier 1987; Hyman and Frank 1980; Laing et al. 1989; Rifkin and Bartoshuk 1980), including lobsters (Ache 1989; Atema et al. 1989; Derby et al. 1989; Zimmer-Faust et al. 1984). Studies of mixture interactions in the Florida spiny lobster Panulirus argus using electrophysiological (Ache 1989; Ache et al. 1988; Derby and Ache 1984b; Derby et al. 1985, 1988; Gleeson and Ache 1985) and behavioral (Daniel and Derby 199 lb) techniques have revealed that mixture suppression is widespread and profound, whereas mixture enhancement, if present, is much less pervasive.

Efects of m&w-e interactions on stimulus coding Mixture interactions change the information about the nature of chemical stimuli that is carried by chemosensory neurons. As a result, mixture interactions could have significant effects on ARMS and ANPs, which would in turn affect behavioral detection and discrimination of chemical quality and intensity. Mixture suppression would result in a decrease below expected in ARMS and a subsequent decrease below expected in the perceived intensity of or behavioral response to chemical stimuli. Mixture interactions could also affect ANPs. If mixture interactions occur and if the direction (i.e., suppression or enhancement) or magnitude of these mixture interactions varies across the members of a population of neurons, then the ANP for the mixture would deviate from expected (i.e., from the combination of the ANPs of the mixture’s components) and could potentially become very different from the ANPs for any of the mixture’s components.

Present study In the present study, responses of a population of 98 single olfactory receptor cells in the antennules of spiny lobsters to a set of seven compounds and their binary mixtures are examined to 1) identify mixture interactions for each neuron and for each binary mixture and 2) examine the effects of mixture interactions on the neural coding of chemical quality and chemical intensity. To identify mixture interactions we use two conventional mathematical models, and, additionally, we develop and use a new model that incorporates our current state of knowledge of receptor sites/transduction processes in olfactory receptor cells of spiny lobsters. We find that mixture interactions are common and that they affect coding of both stimulus intensity and quality. The effects of mixture interactions on ARMS (which we call intensity mixture interactions), and thus on coding of chemical intensity, are presented in this paper. The effects of mixture interactions on ANPs (which we call pattern mixture interactions), and thus on coding of chemical quality, are presented in the following paper (Derby et al. 1991). METHODS

Animals Male and female spiny lobsters (p. argus) were collected in the Florida Keys and were held in the laboratory at 20-25°C in

113

INTERACTIONS

aquaria containing recirculating Instant Ocean. Animals were fed a diet of shrimp and squid.

Chemical stimulants Stimuli were single compounds and binary mixtures of those compounds. The single compounds were adenosine-S-monophosphate (AMP), betaine (Bet), L-cysteine (Cys), L-glutamate (Glu), ammonium chloride (NH,), DL-succinate (Sue), and taurine (Tau). All compounds were obtained from Sigma Chemical or Boehringer-Mannheim (for AMP). Stock solutions wereprepared at 10 mM in artificial sea water (ASW) (Cavanaugh 1964), adjusted to pH 8.1, and frozen in aliquots (except for Cys, which was made fresh each day). Stimuli were serially diluted with ASW. ASW was used as the control stimulus. Binary mixtures of the stimuli were prepared each day from these stock solutions. These stimuli were selected because they are present in many natural prey of spiny lobsters (Carr and Derby 1986), represent different classes of compounds, and are physiologically or behaviorally active compounds for spiny lobsters (Daniel and Derby 199 1b; Derby and Ache 1984b; Fine-Levy and Derby 199 1).

Experimental methods ANTENNULAR RECORDINGS.

PREPARATION

AND

ELECTROPHYSIOLOGICAL

This preparation was described in detail elsewhere (Derby and Ache 1984a; Girardot and Derby 1988; Gleeson and Ache 1985). The distal end of an excised lateral filament of an antennule was inserted into a Teflon tube attached to an olfactometer in which ASW was continuously flowing at 4 ml/min. The proximal end of the antennule was inserted into a separate lucite recording chamber containing Pandims saline. Several annuli were removed from the cut proximal end of the filament, thus exposing both the antennular nerve and the antennular artery. The artery was then cannulated, and the preparation was perfused with oxygenated saline at a flow rate of 0.4-l .O ml/min. Fine-tipped suction electrodes (tip inner diameter -2-5 pm) were used to record responses from single cells in the antennular nerve. If necessary, single units were isolated from multiunit recordings by the use of a window discriminator. Activation of a valve in the olfactometer injected a 6-s pulse of the stimulus into the ASW flow. Responses were measured as the number of spikes generated by a stimulus during the 5-s period following initiation of the response, which usually occurred - 1 s after stimulus introduction. This delay between stimulus introduction and initiation of the 5-s counting period was held constant for all stimulations of a single cell. The number of spikes elicited by ASW was subtracted from the stimulus-induced responses to correct for responses to ASW, for spontaneous activity, and for any potential mechanical or contaminant-induced activity. Responses were standardized for each cell by expressing each response as a percentage of the highest response for that cell. EXPERIMENTAL PROTOCOL. Chemosensitive cells were identified by their altered firing rates in response to search stimuli. Search stimuli were the seven single compounds at 0.5 mM, presented in random order. Once a responsive cell was thus identified, its response spectrum was characterized by recording responses to the seven single compounds at 0.5 mM in random order. Stimuli were presented every 2.5-3 min. From this response spectrum, the “best compound” for that cell (i.e., the compound that is most excitatory) was determined. This designation is appropriate because olfactory receptor cells of spiny lobsters typically have high specificity when tested with an array of single compounds (e.g., Ache et al. 1988; Derby and Ache 1984a; Gleeson and Ache 1985). Next, the concentration-response (C-R) function for that best compound was determined from threshold to near saturation (up to 5 mM) by testing log-step concentrations of that

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

114

DERBY,

GIRARDOT,

compound in ascending order. The usual number of concentrations tested was five or six. Subsequently, the C-R functions were determined for six binary mixtures; i.e., the best compound over the same concentration range as above plus 5 mM of one of the other six compounds. The order of presentation of these six sets of binary mixtures was randomized. Also tested for each binary mixture concentration series were the single components of the binary mixture alone at 5 mM and the ASW control. The C-R curve for the best compound alone was determined again after measuring the C-R curves for any two sets of binary mixtures. This was necessary to determine whether a cell’s chemosensitivity was maintained at a reasonably constant level during the course of the experiment. If the response to the best compound at any point in the experiment was ~50% of the original response, the data from that point in the experiment were not used. Therefore not every cell was tested with all six binary mixtures. [Of the 98 cells, 70% were tested with all six binary mixtures, and the mean (&SE) of the number of binary mixtures tested per cell was 5.0 t_ 0.17.1

Data analysis AMONG CELLS: CLUSTER ANALYSIS. Hierarchical cluster analysis (Bieber and Smith 1986; Ever&t 1980) was used to determine whether the response profiles of the olfactory receptor cells (or across chemical patterns) fell into distinct groups and, if so, whether such groups were correlated with the cells’ best-compound classification. The hierarchical cluster analysis (CSS Software, StatSoft) was derived from a proximity (distance) matrix of Pearson correlation coefficient values between any two cells, as determined by their standardized responses to the seven compounds at 0.5 mM. Use of Pearson correlation coefficient as the distance metric emphasized relative rather than absolute similarity in responses, thus placing less emphasis on the importance of differences in overall response magnitudes of the cells and focused on their across chemical patterns of activity (Bieber and Smith 1986). Ward’s method of linkage was used as the clustering algorithm in the cluster analysis. SIMILARITIES

CALCULATING

PREDICTED

RESPONSES

TO BINARY

MIXTURES.

Three mathematical models were used to predict expected responses to binary mixtures based on responses to the components of the mixtures: single receptor model, multiple receptor model, and mixed receptor model. I) Single receptor model. Modeling of observed C-R functions for single chemicals. For each cell all measurements of the C-R functions for that cell’s best compound were pooled and analyzed by curvilinear regression (CSS Software, StatSoft). Linear, quadratic, and cubic models of each C-R function were tested progressively for best empirical fit. The model of best fit was defined as the model with the largest number of regression coefficients all of which were significant (Sokal and Rohlf 198 1). According to this criterion the linear model was used in 67% of the cells and the quadratic model in 33%. Predictions of responses to binary mixtures. The assumption of the single receptor model is that the two components of the mixture competitively bind to the same receptor sites. To predict responses to mixtures under this assumption, we used the stimulussubstitution model (substitution model), which has been used previously to analyze electrophysiological, behavioral, and psychophysical data (e.g., Caprio et al. 1989; Carr and Derby 1986; Derby and Ache 1984b; Hyman and Frank 1980; Rifkin and Bartoshuk 1980). In this model it is assumed that a binary mixture will act as a higher concentration of either component. Thus predicted responses to each of the two components of a binary mixture at concentrations found in the mixture were determined for each cell from the regression model of the C-R data for the best compound for that cell. In cases for which a linear regression best described the C-R function for the best compound

AND

DANIEL

RAB= b,

X

log ([A] + [ 1O(RB-aA)/bA]) + aA

(0

where R,, is the predicted response to the mixture A + B; R, is the predicted response to compound B; [A] is the concentration of compound A, the best compound for that cell; and b, and aA are the slope and intercept, respectively, for the C-R response function for A. The above equation was used to calculate the predicted response to the mixture A + B in terms of the linear model for the response to chemical A. In caseswhere a quadratic regression best described the C-R function for the best compound, a quadratic version of Eq. I was used. 2) Multiple receptor model. The multiple receptor model assumes that each of the components acts independently of each other, each binding to a separate class of receptor sites. To predict responses to mixtures under this assumption, we used a polynomial model (Daniel and Derby 1987; Jackinovich 1982). This model assumes independence of responses to the components, such that binding events can temporally cooccur. This is unlike some other multiple receptor models, such as response summation models (Caprio et al. 1989; Carr and Derby 1986; Hyman and Frank 1980; Rifkin and Bartoshuk 1980), which assume mutual exclusivity of responses to components of a mixture. The polynomial equation is RAB = RA/lOO + R,/lOO

- (RA/lOO X RB/lOO)

(4

where R,, is the predicted response to mixture A + B; and R, and R, are the responses to compounds A and B, respectively. The polynomial model will not accept negative values, which our data set contains. (Negative values can occur because the raw response values to compounds are corrected for responses to ASW, which are occasionally greater than the responses to compounds.) Thus, if the response value to either component has a negative value, we used the simple sum of the responses to the two components as the predicted response to the binary mixture (response summation) (Carr and Derby 1986; Rifkin and Bartoshuk 1980). 3) Mixed receptor model. The single and multiple receptor models make extreme assumptions about the nature and distribution of receptor sites on cells. The single receptor model assumes that each cell contains only one type of receptor site, to which both components of a mixture can bind. The multiple receptor model assumes that each cell contains two types of receptor sites, each type specific for only one component of the binary mixture. It might be expected that most cells would possess many types of receptors with differing binding specificities, some with high specificity and others with low specificity. To model cells in such a system, we developed a mixed receptor composition model (mixed model). The logic of the mixed model is described in more detail in Daniel and Derby (199 la), although the mixed model used here for single cells is somewhat different from the mixed model used in analyzing behavioral data by Daniel and Derby ( 199 1a,b). The mixed model in this study is a weighted average of the single receptor model and the multiple receptor model. The weighting is based on our current state of knowledge, through cross-adaptation experiments using extracellular electrophysiological responses of single cells, of the nature and distribution of types of receptor sites on olfactory receptor cells of spiny lobsters (J. B. Fine-Levy, C. D. Derby, M.-N. Girardot, and P. C. Daniel, unpublished data). In the cross-adaptation studies, for any pair of stimuli, a cell was adapted to one chemical until that chemical elicited no or a low response. While the cell was in the adapted state, the response to the other chemical was recorded. Subsequent to disadaptation, the procedure was reversed, such that the other chemical became the adapting chemical, while the response to the previous adapting chemical was recorded. If there was no response to either test chemical under cross-adaptation conditions, there was complete and reciprocal cross adaptation. and we concluded

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

INTENSITY

MIXTURE

that there is probably only one transduction process for these two chemicals and therefore presumably only one shared receptor. If the cell responded fully to the test chemicals under cross-adaptation conditions, there was no cross adaptation, and we concluded that there are separate transduction processes and presumably also separate receptors for the two chemicals. We realize that receptor sites are only one component of the transduction process in these cells, and therefore equating the two is overly simplistic. Nonetheless, analysis of spiking responses of single cells under cross-adaptation regimes is at present our only gauge of competition of compounds for receptor sites on these cells. If any error is made by the mixed model, it is in giving too much weight to the single receptor model and not enough to the multiple receptor model. Cross adaptation was quantified by the method of Caprio and Byrd (1984) and Rehnberg et al. (1989) by the use of the following equations CA A,

adaptB

=

{ 1 -

[cRA,

adaptB

-

RASW,

adaptB

)/tRA, -

INTERACTIONS ADAPTATION CHEMICALS AMP/BET AMP/CYS AMP/GLU AMP/NH4 AMP/SUC AM P/TAU BET/CYS BET/GLU

BET/NH4 BET/SUC E3ET/TAU CYS/GLU CYS/NHq CYS/SUC CYS/TAU GLU,‘NH4 GLU/SUC GLU/TAU NHa/SUC N H 4,‘TAU SUC,‘TAU

unadapt RASW,

unadapt

)I>

x

l O”

t3)

is the cross-adaptation value for stimulus A where cAA, adaptB under adaptation to B, RA,adaptBis the response to stimulus A under adaptation to B, R,, unadaptis the response to stimulus A in ASW background, RAsw a&@ is the response to control ASW during adaptation to B, and RAsw, unadaptis the response to control ASW in ASW background CA B, adaptA

=

{ 1 -

[cRB,

adaptA

-

RASW,

adaptA)/cRB,

ASW,unadapt)l)

x

loo

t4)

is the cross-adaptation value for stimulus B under cAB, adaptA adaptation to A, RB, a&ptAis the response to stimulus B under adaptation to A, Rg, “*adaptis the response to stimulus B in ASW backis the response to control ASW during adaptaground, RASW adaptA tion to A, and R ASW,unadapt is the response to control ASW in ASW background. Finally =

[fcAA

, adaptB)fNA

, adaptB)

+ (CA B, adaptA

MN

B, adaptA

)I/cNA,

adaptB

+

NB,

adaptA)

adaptA

l

Cross-adaptation data for 100 olfactory receptor cells of the spiny lobster yielded high cross-adaptation index values for pairs of compounds (Fine-Levy et al., unpublished data). (These 100 cells are all different from the 98 cells already described in this paper.) The mean cross-adaptation index values for pairs of compounds ranged from 0.7 to 1.05, with variation across cells. A cross-adaptation value of 0 signifies self adaptation but no cross adaptation; 1.0 signifies self adaptation equal in magnitude to cross adaptation; values slightly greater than 1.Oare possible when cross adaptation is slightly greater than self adaptation. The results are summarized in Fig. 1. The mixed receptor model uses these cross-adaptation index values to provide the weighting of the single receptor and multiple receptor models, as expressed in Eq. 6 below R AB,

mixed

=

(R AB , single

x

cAAB)

+

[RAB

9multiple

x

( 1 -

CA,)1

(6)

is the predicted response to A + B by mixed recepis the predicted response to A + B by single single is the predicted response to A + B by receptor mode1~ RAB, multiple multiple receptor model, and CA,, is the cross-adaptation index value for stimulus pair A and B. where tor

RAB,

mode1,

mixed RAB,

INDEX

1. Cross-adaptation index values used in the mixed receptor model. Values are given for each stimulus pair, and are means k SE, calculated according to Eq. 5. N is number of neurons used to calculate each cross-adaptation index value.

STATISTICAL ANALYSES OF INTENSITY TIONS. Differences between observed and

MIXTURE

INTERAC-

predicted response values were evaluated statistically by the use of paired t tests and a significance level of 0.05, both for single cells (Fig. 5) and for populations of cells, either concentration dependent (Figs. 6-l 1) or concentration independent (Figs. 12 and 13).

t5)

where c&B is the cross-adaptation index value for stimulus pair A and B, and is calculated as the weighted average of c&, a&@ and CA B, adaptA as the number of observations of with NA, adaptB CA A, adaptB as the number of observations of and NB, adaptA CA B,

1.0 ‘I.2

FIG.

where

cAAB

0.0 0.2 0.4 0.6 0.8

CROSS ADAPTATION

unadapt

-R

115

RESULTS

Neurons Responses from 98 olfactory receptor cells were used in these analyses. The rate of spontaneous spiking in these cells was 0.7 1 t 0.09 Hz (mean t SE, n = 98). The response spectrum for each of the 98 cells, based on responses to AMP, Bet, Cys, Glu, NH,, Sue, and Tau each at 0.5 mM, is shown in Fig. 2. Cells’ are arranged in this figure according to best-compound classification. There were 23 AMP-best cells, 3 Bet-best cells, 13 Cys-best cells, 8 Glu-best cells, 33 NH,-best cells, and 18 Tau-best cells; no Sue-best cells were found. Both excitatory and inhibitory responses were observed. Excitation was much more common than inhibition. Inhibitory responses occurred when a compound reduced the activity of a spontaneously active cell or when the response to a compound was less than the response to the ASW control stimulus. The percentage of responses that were excitatory (+ 1 to + 100% standardized response), inhibitory (- 1 to - 100% standardized response), and zero were 67, 23, and lo%, respectively (Fig. 2). The excitatory responses were of greater magnitude than inhibitory responses, which is evident in Figs. 2 and 4.

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

DERBY,

116

AMP Bet CYS Gltl NH4 sue Tau

AND

DANIEL

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

1 2 3 4 5 6 7 8 .oo@ooeooooe~oooe~ooeo. 0 0 l @ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

oo.oooooooooo~o oo.o.o.oooooo.@ 0 0.0 0 0 0 0 0 0 0 0 0.0

l .*.o...oooooooooo. .oooooo.ooo..o

0

0

0

0

0 0

0

0 0

0 0

.o..o.....oooooooo.....

24 25 26 AMP Bet CYS Glu NH4 sue Tau

GIRARDOT,

l

0

0.0 00.

0

l

0

0

40 41 42 43 44 45 46 47 00 .o. O.0 l . ..oooo

0.0 0

;ii;i;;;

0

0

l .0 ;;ooeo~~~;o~o

0 0 a 0

0

27 20 29 30 31 32 33 34 35 36 37 38 39 0 l .

0

O

l

0

.;:.

.

0

@

0

l

0

0

.

0

l

a

0

:

0

00

0

0

0

000~00

l

l

0

0

0

0.0

0

0

0

0

0

40 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73

AMP Bet CYS Glu NH4 sue Tau

AMP Bet CYS Glu NH4 sue Tau

0 0 00 0 0 0 0 0 0 0 0 0 0 l o..~ooo.oo o~eo0oooooooo.ooooooo~O.oO a .e .~~.o.o*oo*ooo~~oooo**o l

~o.o*@

l l

oooooooeoooooo.o.oooooooo O •~0o~ooo0ooo..ooo0ooOoOo

l oooooooi;i;;o~;;;;;~~oooo 74 75 76 77 78 79 00 0 0 0 0 0 l

0

0

l

0 0

0

0

0

0

&&&&i; 0

0

0

0

l

0

0

0

l

0

l

01 02 03 04 05 06 07 00 09 90 91 92 93 94 95 96 97 90 0

l

0

0

0

0

0

0

0

0

0

0

0

0

0

0

l ~@oo.~oeooooo.~.. •0~0~~~~~0~00

.:.:

0

0

:::::::I

0

0

0

0

0

0

0

0

0

00

.o ~&~~;mee;;i;;;i6~~

no. 2. Response spectra of 98 individual olfactory receptor cells. Shown are the responses of 98 cells to 7 compounds at 0.5 mM. Cells are arranged according to best-compound classification: cells I-23 are AMP-best cells, c&s 24-26 are Bet-best cells, cells 27-39 are Cys-best cells, cells 40-47 are Glu-best cells, cells 48-80 are NH&St cells, and cells 81-98 are Tau-best cells. Type and size of dot represent the responses as standardized to the best-compound. Closed dot represents excitatory response; open dot represents inhibitory response; no dot represents response of 0 Hz. Closed dots of 4 sizes are shown, and from smallest to largest they represent responses that are: l-25%, 26-50%, 5 l-75%, and 76- 100% of the maximum response ofthat cell. Open dots of 3 sizes are shown, and from smallest to largest they represent responses that are - 1 to -25%, -26 to -5O%, and -5 1 to -75% of the maximum response of that cell.

Ninety-one percent of all inhibitory response standardized values were -1 to -25%. Inhibition, although of low magnitude, was common for some compounds and for some cell types. For example, for AMP-best cells, 0.5 mM Glu inhibited 15 cells, excited 7 cells, and had no effect on 1 cell. The average effect of 0.5 or 5 mM Glu on AMP-best cells was -5 t 2% (mean t SE of standardized response; n = 19). The average effect of 5 mM Bet on NH,-best cells was - 13 t 6% (mean t SE of standardized response; n = 29). In general, in terms of response spectrum, cells were highly specific for a single compound (i.e., best compound). A reflection of the relatively high specificity is that most of the responses were of low magnitude: 7 1% of all of the standardized response values were between -25 and +25%. Many of the responses in this range, including those that we identified as excitatory or inhibitory, were so low as to represent statistical variation around zero (no response). For example, for a given cell, if ASW (the control stimulus) and

Tau elicited responses of 3 spikes/5 s and 2 spikes/5 s, respectively, we would identify the response to Tau as inhibitory, even though the difference between 3 and 2 spikes/5 s is within the variability for the ASW response. Hierarchical cluster analysis provided a quantitative measure of the tendency of the response profiles of the cells for the seven compounds to fall into clusters or groups. The results of cluster analysis are depicted as a dendrogram in Fig. 3. In dendrograms, cluster distances indicate the degree of similarity between the response profiles of cells or clusters of cells. The two most similar neuronal response profiles are separated by the least distance. Larger sets of neurons are interconnected at greater distances, indicating the greater degree of dissimilarity among their response profiles. Clusters are indicated by an abrupt increase in the distance at which clusters are linked. Scree diagrams are conventionally used to identify these discontinuities in clustering and therefore indicate the number of clusters (Bieber and Smith 1986). A scree diagram is shown for this

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

INTENSITY

MIXTURE

loo-@ W a\

Y 2

a0

c cl

60*

LL: K s

d

0

\ 40.

0 \

20.

l\ 0.

0. 1

5

.a** 10

NUMBER

3LCy8 30-W* 81Tau

15

20

25

30

OF CLUSTERS

-

9STau 98Tau 86Tau 09Tau 97Tau 88Tau

84Tau 05Tau 91Tau 92Tau 94Tau 90Tau 93Tau

82Tau 83Tau 96Tau 87Tau 48JdH4 !5o,JIi4 SlJH4 54Nn4 49Jm4 55Jai4 66NH4 65N?I4 67NX4 73Jm4 77JaI4 s7m4 62NH4 74N?i4 76N?I4 61-NH4 58N?t4 64NH4 59NxI 68JIH4 6ONH4 79Jm4 7oHH4 S6NH4 8OJI?i4 69J4H4 78JW4 71Jd?r4

72NH4 63-NH4 7sNH4

FIG. 3. Cluster analysis of response profiles of 98 olfactory receptor cells. The analysis is based on standardized responses to AMP, Bet, Cys, Glu, NH,, Sue, and Tau at 0.5 mM as shown in Fig. 2. The cell numbers 1 to 98 correspond to the cells whose response profiles are shown in Fig. 2. The best compound for each cell is shown next to the cell number. The inset is a scree diagram for this cluster analysis. See text for explanation.

INTERACTIONS

117

cluster analysis (inset of Fig. 3). The scree procedure involves moving from left to right on the curve until one finds a point such that it and all remaining points to its right lie along a straight line (Bieber and Smith 1986). This point is often called the “elbow,” and it represents the most appropriate number of clusters for the data. Analysis of our scree diagram revealed a distinct elbow at six clusters. These six clusters as assigned by cluster analysis were correlated extremely well with the six groups of cells based on the bestcompound criterion. Only 3 of 98 cells were exceptions to this generalization: one Cys-best cell (#31) was included in the Bet-best cell cluster, and two NH,-best cells (52 and 53) were included in the Cys-best cell cluster. Thus this analysis supports the validity of our analyzing the data in the remainder of the paper according to best-compound cell classification. It should be stressed here that it is not our primary intention to identify cell types. The likelihood of identifying cell types depends on the type and number of stimuli tested. Although cell types may be identified based on their response spectra to a relatively small set of single compounds, as shown here, cell types are less easily identified when simple or complex mixtures are used as test stimuli (Derby et al. 1988, 1989; Girardot and Derby 1990b). Instead, it is our intention here to test empirically the validity of the best-compound cell groupings that were dictated by our experimental protocol and that were used in our data analysis. Mean population response spectra for the six best-compound cell groups are shown in Fig. 4. The narrow tuning of these cells, especially the AMP-best cells, NH,-best cells, and Tau-best cells, is apparent from this figure even though the responses are expressed as population means. The responses in this figure are expressed in terms of absolute values (mean number of spikes t SE), unlike the standardized values in Figs. 2, to show differences in response magnitudes for different cell groups: Bet-best cells and Cys-best cells tended on average to have lower response magnitudes to their best compound than did the other groups of cells. Individual cell analysis of binary mixture interactions The frequency of occurrence of mixture interactions across individual olfactory receptor cells is shown in Fig. 5. Results are separated by cell group (i.e., best-compound categorization) and by binary mixture and are based on the mixed model. For example, for AMP-best cells and for the mixture AMP + Bet, 24% of the cells showed mixture suppression, 10% showed mixture enhancement, and 66% showed no mixture interactions. For these AMP-best cells, of all the binary mixtures, AMP + Glu showed the most mixture interactions, all suppression, with 47% of the cells showing this effect. Overall, mixture interactions were fairly common. The enhancing and suppressing effect of a compound was typically reversible and of short duration. Mixture interactions were seen for 25% of all possible cases. Both suppression and enhancement occurred, but suppression was much more common. Suppression was most common in AMPbest cells, Bet-best cells, Cys-best cells, and Glu-best cells. Enhancement was especially rare in the above three cell

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

DERBY,

118

GIRARDOT,

AND 80

180~ 160--

DANIEL

-

AMP

cells

Bet Cells

(n=z3)

(n=3)

60

140

t

120 100 80 60 40

20 09 -20

AMP

f

Bet

Cys

Glu

NH4

Sue

Tau

1

-20

AMP

Bet

Cys

Glu

NH4

Sue

Tau

AMP

Bet

Cys

Glu

NH4

Sue

Tau

120

0 Q)

Cys Cells

(n=l3)

cn lOO-LD

\ z .-xQ u-l W 0

Z 0 nm W cr

-

t 160

80 60

80

40 20 03 AMP

-20

Bet

Cys

Glu

NH4

Sue

Tau

200

MOO

a

-40

NH4

160--

Cells

(n=33)

180

T

160

1

I

1

Tau Cells

(n=18) T

I

140 120 100 80 60 40

20 E

0I -20

t

AMP

Bet

Cys

Glu

NH4

Sue

Tau

-20

+

AMP

Bet

Cys

Glu

NH4

Sue

Tau

according to best-compound classification. Cell groups FIG. 4. Mean response spectra for the 6 cell populationsgrouped include AMP-best cells, Bet-best cells, Cys-best cells, Glu-best cells, NH,-best cells, and Tau-best cells, whose individual standardized response profiles are shown in Fig. 2. Response values shown here are mean (*SE) of absolute responses, i.e., number of spikes elicited in 1st 5 s of response. All compounds were tested at 0.5 mM. n is number of neurons in each best-compound class.

groups; but it was more common for NH,-best cells and Tau-best cells (in fact, in these cells, enhancement was as common as suppression). Apparent in this figure is that a given mixture may produce mixture suppression in some cells and mixture enhancement in other cells, even in the same best-compound group. In addition (not shown in this

figure), a single cell can show suppression to some mixtures and enhancement to other mixtures. In summary, there was variability across individual olfactory receptor cells in the occurrence and type of mixture interactions. Not all cells of a given best-compound type showed the same mixture interactions, and not all binary

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

INTENSITY AMP

Cells

loo-'

(n=23)

MIXTURE

INTERACTIONS

119

Bet

Cells

(n=3)

655l SUPPRESSION I NO INTERACTION IZZIENHANCEMENT

80VI Ii

60

60 -

40

40-

8 o\o

20-

OA+B

A+C

A+G

Cvs

Cells

A+N

A+S

A+T

B+A

(n=13)

B+C

B+G

Glu Cells

B+N

B+S

B+T

(n=B)

80-

60-

40-

T \ \ C+A

C+B

C+G

C+N

C+S

G-tA

C+T

G+B

G+C

Tau

Cells

T+B

T+C

G+N

G+S

G+T

T+N

T+S

(n=18)

FIG. 5. Intensity mixture interactions for individual cells. Shown is the percentage of cells of each best-compound cell group whose response to their best compound (AMP for AMP-best cells, Bet for Bet-best cells, etc.) is suppressed (-), not affected (o), or enhanced (H) by addition of each compound shown on the abscissa. Statistically significant mixture interactions for each cell were determined for each binary mixture by comparing the observed response and the predicted response from the mixed receptor model for the same binary mixture, using comparisons at different concentrations as observations in a paired t test (P < 0.05). y2is number of neurons in each best-compound class; however, not all binary mixtures were tested on all cells of a class, so sample size for each binary mixture is in some cases less than n. A = AMP, B = Bet, C = Cys, G = Glu, N = NH,, S = Sue, and T = Tau.

80

60

N+A

N+B

N+C

N+G

N+S

N+T

T+A

T+G

STIMULI mixtures produced the same m ixture interactions across cells). For clarity, we show in these figures only the predicted responses from the mixed receptor model. The recells. As a rule, though, mixtu re suppression domin ated. sults of the single and multiple receptor models are generally statistically similar to those of the mixed receptor Cell population analysis: concentration-response functions model. The results of all three models are summarized in for observed and predicted responses to binary mixtures Fig. 13. The previous section examined mixture interactions in Observed and predicted responses of the population of individual cells. This section examines how mixture inter- AMP-best cells to binary mixtures of AMP plus each of the actions affect populations of cells, that is, their average ef- other compounds are shown in Fig. 6. Asterisks indicate fects. comparisons between observed responses and responses INTENSITY MIXTURE INTERACTIONS. Concentration-dependent predicted from the mixed receptor model, at any given conanalysis. Figures 6-l 1 show the comparisons of the ob- centration, that are significantly different. For example, the served and predicted responses to binary mixtures as a observed response to AMP + Bet was significantly different function of concentration. The results are presented for from the predicted value at only one concentration, AMP each best-cell group (i.e., AMP-best cells, Bet-best cells, 0.5 mM + Bet 5 mM. Because the observed value was less Cys-best cells, Glu-best cells, NH,-best cells, and Tau-best than the predicted value, this type of intensity mixture inDownloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

120

DERBY,

GIRARDOT,

AND

DANIEL

AMP CELLS: AMP+BET

AMP CELLS: AMP+CYS 100,

100 H---

l

*

P

=-’

-

.AMP+BET

MIXED

AMP OBSERVED CYS 5 mM OBSERVED -oAMP+CYS 5 mM OBSERVED -0AMP+CYS MIXED MODEL

l l

---.

oAMP+BET 5 mM OBSERVED

o-

80

.. AMP OBSERVED BET 5 mM OBSERVED

80

MODEL

60

60

k

i

13

i

Lo / / 0

40.

-

>

20-

O\

No\

/-;-@A: ; --------------

2

-10

W v, Z

0 11 ul

-9

-8

-7

-6

-5

-4

-3

AMP CELLS: AMP+

-2

-1

0

1

GLU 5 mM OBSERVED -oAMP+GLU 5 mM OBSERVED

W e

60

n w

4o

7

/

/i

N n 2o nc 4 0 n Z~ -20 -10

-8

/ 0

-?L-1-------------------. 12

-6

-5

-4

-3

-2

-1

0

1

1

P

.=--¤ AMP OBSERVED .--. NH4 5 mM OBSERVED -o--AMP+NH4 5 mM OBSERVED 80 .-.AMP+NH4 MIXED MODEL

.j .A. * /

60

o/i

0

-7

17 m

f

AMP CELLS: AMP+NH4 100

8o

-9

GLU

100

9

-10

14

0 / l

15

.m

/

/

3q7?*p8

'7 0-o

;Y\

0-1 15 4 _-------------------------------------,

.-----

.---_---------------------------------------.

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

-10

1

I -9

1 -8

I -7

2 1 -6

I -5

11 1 -4

I -3

I -2

I -1

I 0

1

tCA

AMP CELLS: AMP+SUC

AMP CELLS: AMP+TAU

100

100

m---m AMP OBSERVED ---. SUC 5 mM OBSERVED -o---AMP+SUC 5 mM OBSERVED 80 l -•AMP+SUC MIXED MODEL

AMP OBSERVED ---. TAU 5 mM OBSERVED -o--AMP+TAU 5 mM OBSERVED 80 l -.AMP+TAU MIXED MODEL

60

60

8

.=-.

P

I

40

20

0 -10

_--------_----------------------------------~ I I I I

-9

-8

-7

-6

I

I

I

I

I

I

-5

-4

-3

-2

-1

0

1

/

\

I nE?!Tf%M-------.

0 -10

-9

AMP CONCENTRATION

-8

-7

-6

-5

-4

-3

-2

-1

0

1

(log x 5 mM)

FIG. 6. Intensity mixture interactions in population of AMP-best cells. C-R functions representing mean values for all AMP-best cells are given for the following: observed values for AMP alone (Wn); observed values for AMP + 5 mM of a 2nd compound [either Bet, Cys, Glu, NH,,, Sue, or Tau (o o)]; predicted values for AMP + 5 mM of the 2nd compound from the mixed receptor model (o l ). Also given is the observed value for 5 mM of the 2nd compound [either Bet, Cys, Glu, NH,, Sue, or Tau (- - -)]. The numbers below the points represent numbers of neurons from which observations were made at each concentration. Asterisks indicate concentrations where the observed response and the response predicted by the mixed receptor model are significantly different (paired t test, P < 0.05).

teraction would be suppression. For AMP-best cells, the mixture of AMP + Glu showed the greatest prevalence and extent of intensity mixture interactions, all of the suppressive nature; these occurred from 5 X 10e5to 5 X lo-’ mM. AMP + Sue also showed mixture suppression, especially at

concentrations of AMP above 5 X 10m4mM. The same trend occurred at 5 X low6 and 5 mM, but the sample size was too small to show significance. The population of Bet-best cells (Fig. 7) showed no statistically significant mixture interaction. For two cases, there

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

INTENSITY BET

CELLS:

MIXTURE

INTERACTIONS BET CELLS:

BET+AMP

BET+CYS

100 BET OBSERVED AMP 5 mM OBSERVED BtT+AMP 5 mM OBSERVED BET+AMP MIXED MODEL

W-

80-,: l

-

. =-= BET OBSERVED ---* CYS 5 mM OBSERVED -o-oBET+CYS 5 mM OBSERVED 80 .--•BET+CYS MIXED MODEL

0

60-

60 40

20

0 -7

-6

-5

-4

W v-l

clv, I I I

-2

0

-1

1

-7

-6

100

-4

-3

-2

-1

0

1

100

.' ..--

-5

BET CELLS: BET+NH4

BET CELLS: BET+GLU

7

L0

-3

BET OBSERVED

80-oz: ..-.

:k:+%t”50,BM;E:::;RVED BET+GLU MIXED

80

,.,-. MODEL

i 60

60-

o--BET+NH4

5 mM OBSERVED

1

-7

-6

-5

-4

-3

-2

-1

-7

0

-6

-5

BET CELLS: BET+SUC

-3

-2

-I

0

BET CELLS: BET+TAU

T

.-=

BET OBSERVED - SUC 5 mM OBSERVED -oBET+SUC 5 mM OBSERVED BET+SUC MIXED MODEL

80

-4

0

-. TAU 5 mM OBSERVED

60

40

20

0

I

1

-8

-7

1

-6

1

I

I

I

-5

-4

-3

-2

-20

I --I

0

-5

BET CONCENTRATION

1

1

1

I -4

I -3

I -2

1 -1

(log x 5 mM)

FIG. 7. Intensity mixture interactions in population of Bet-best cells. C-R functions representing mean values for all Bet-best cells are given for the following: observed values for Bet alone (= 1); observed values for Bet + 5 mM of a 2nd compound [either AMP, Cys, Glu, NH,,, Sue, or Tau (o o)]; predicted values for Bet + 5 mM of the 2nd compound from the mixed receptor model (o +). Also given is the observed value for 5 mM of the 2nd compound [either AMP, Cys, Glu, NH4, Sue, or Tau (- - -)]. The numbers below the points represent numbers of neurons from which observations were made at each concentration. Asterisks indicate concentrations where the observed response and the response predicted by the mixed receptor model are significantly different (paired t test, P < 0.05).

were trends toward mixture interactions-suppression for Bet + Glu and Bet + Tau and enhancement for Bet + AMP -but the low number of Bet-best cells (13) precluded finding any significant differences at any one concentration. Mixture suppression was common in the population of

Cys-best cells (Fig. 8), occurring for all but one mixture (Cys + AMP) at least at one concentration. The population of Glu-best cells (Fig. 9) showed mixture interactions, especially for Glu + Bet, Glu + Cys, and Glu + NH4, each at several concentrations.

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (137.154.019.149) on January 12, 2019.

122

DERBY,

GIRARDOT,

AND

DANIEL

CYS CELLS: CYStAMP

CYS CELLS: CYS+BET 120

AMP 5 mM

&m CYS OBSERVED BET 5 mM OBSERVED loo-o-oCYS+BET 5 mM OBSERVED l -.CYS+BET MIXED MODEL

OBSERVED

.I

.-m.

80

80-

60

60-

40 20 0 -4

-3

-2

-1

I -3

04 -4

1

0

I -2

I -1

I 0

I 1

W

CYS CELLS: CYS+NH4

CYS CELLS: CYS+GLU CYS OBSERVED GLU 3 mM OBSERVED CYS+GLU 5 mM OBSERVED CYS

W w

80

n w

CYS OBSERVED NH4 5 mM OBSERVED CYS+NH4 5 mM OBSERVED CYS+NH4 MIXED MODEL

*

6o

60-

N n

40

40-

K < n

20

20

6

10

-'3

-'2

*

6

/

ol

-4

Responses of olfactory receptor cells of spiny lobsters to binary mixtures. I. Intensity mixture interactions.

1. Neural coding of chemical mixtures was studied with the use of the peripheral olfactory system of the spiny lobster. The occurrence of mixture inte...
4MB Sizes 0 Downloads 0 Views