Journal of Applied Physiology
Vol. 83, No. 4, pp. 1174-1183, October 1997
CONTROL OF BREATHING, CIRCULATION, AND TEMPERATURE

Classic conditioning of the ventilatory responses in rats

Elise Nsegbe¹, Guy Vardon², Pierre Perruchet³, and Jorge Gallego¹

¹ Laboratoire de Neurologie et Physiologie du Développement, Hôpital Robert-Debré, Université de Paris-7, 75019 Paris; ² Unité de Recherches sur les Adaptations Physiologiques et Comportementales, Université de Picardie, 80036 Amiens; and ³ Laboratoire d'Etudes des Acquisitions et du Développement, Université de Bourgogne, 21000 Dijon, France

ABSTRACT

INTRODUCTION

METHODS

RESULTS

DISCUSSION

ACKNOWLEDGEMENTS

FOOTNOTES

REFERENCES

ABSTRACT

Nsegbe, Elise, Guy Vardon, Pierre Perruchet, and Jorge Gallego. Classic conditioning of the ventilatory responses in rats. J. Appl. Physiol. 83(4): 1174-1183, 1997.Recent authors have stressed the role of conditioning in the control of breathing, but experimental evidence of this role is still sparse and contradictory. To establish that classic conditioning of the ventilatory responses can occur in rats, we performed a controlled experiment in which a 1-min tone [conditioned stimulus (CS)] was paired with a hypercapnic stimulus [8.5% CO₂, unconditioned stimulus (US)]. The experimental group (n = 9) received five paired CS-US presentations, followed by one CS alone to test conditioning. This sequence was repeated six times. The control group (n = 7) received the same number of CS and US, but each US was delivered 3 min after the CS. We observed that after the CS alone, breath duration was significantly longer in the experimental than in the control group and mean ventilation was significantly lower, thus showing inhibitory conditioning. This conditioning may have resulted from the association between the CS and the inhibitory and aversive effects of CO₂. The present results confirmed the high sensitivity of the respiratory controller to conditioning processes.

control of breathing; carbon dioxide

INTRODUCTION

RECENT STUDIES provide new experimental evidence that ventilatory activity can be adapted to physiological requirements by learning processes (29, 30). Among these processes, a particular importance is attached to classic conditioning, i.e., the process by which a conditioned stimulus (CS) that has been paired with an unconditioned stimulus of breathing (US) is able to elicit a conditioned ventilatory response (CR) (13, 21, 24, 34). In particular, it has been reported that early conditioning during the postnatal period may have lasting influences on the breathing pattern (31, 32). In general terms, classic conditioning is interpreted as the acquired ability to anticipate forthcoming metabolic needs, which means that a CS signaling a respiratory US such as hypercapnia should normally elicit a conditioned increase in ventilation (33). This prediction was supported by early investigations of respiratory conditioning (16, 26). For example, Pogrebkova (26) reported that in dogs a sound previously paired with a hypercapnic stimulus of 9-10% triggered a conditioned increase in breathing frequency and amplitude.

However, the results of subsequent studies failed to confirm these early findings consistently. Weinstein and Fowle (37) observed no conditioning in pigeons, even after 400 paired presentations of a light or a sound with a 7.4% CO₂ stimulus. A still more contrasted outcome was reported by Biryukov et al. (5). These authors performed a series of experiments in monkeys, dogs, cats, rabbits, guinea pigs, pigeons, turtles, frogs, and rats, in which a sound or a light was paired with a 50% CO₂ stimulus. CRs were observed in most species (although not in cats) after different numbers of paired conditioned and unconditioned stimuli: 3-5 in dogs and monkeys, 6-10 in pigeons, and 20-25 in rabbits. The CR consisted of a decrease in breathing amplitude and frequency, sometimes reaching apnea. By itself, the fact that the CR and the unconditioned ventilatory response (UR) act in opposite directions is not an exception in classic conditioning, because such action has been reported in conditioning of body temperature, blood glucose levels, heart rate, etc.¹ However, in the specific framework of respiratory conditioning, this result was in total conflict, with both with previous findings and the current notion that conditioned ventilatory responses anticipate forthcoming metabolic requirements.

Three factors may account for these controversial data: the absence of appropriate control procedures, the detection of CO₂, and the inhibitory effects of CO₂. First, most early investigators used very few subjects, generally one or two, and evidence for conditioning was based on selected tracings of the ventilatory signal after the CS alone (2, 6, 11, 16, 26, 28, 36). Since these pioneering studies, new methodological concepts have led to substantial changes in conditioning designs (20, 27). According to present criteria, conditioning is not established unless it is shown that the response elicited by the CS is a specific consequence of the pairing of the CS with the US. This is particularly important in respiratory conditioning experiments because repeated exposure to hypercapnia or hypoxia may induce long-term physiological and behavioral changes, which may affect the response to any stimulus, including the CS, independently of any associative process. Without appropriate control procedures, it is impossible to decide whether ventilatory changes result from learning the association between the CS and the US (i.e., conditioning) or from nonassociative processes.

Second, the contradictory results of conditioning studies may be due to the ability to detect the US, especially CO₂, because of its sensory properties rather than its respiratory effects. This may strongly affect conditioning. Previous literature in fact showed that a conditioned activation of breathing occurred when CO₂ was administered intratracheally, i.e., when the probability of detecting CO₂ was low (16, 26), whereas conditioned inhibition of ventilation (5) or no conditioning at all (37) was reported when CO₂ was delivered through the upper airways. In fact, stimulation of the nasal mucosa by CO₂ may act as a CS, in addition to the auditory or visual stimuli experimentally designed to do so. What generally occurs when two CS are presented together with the US is that the "stronger" one, in terms of intensity, salience, and predictive value in relation to the US, may "overshadow" the weaker one (20). The US is preferentially associated with the stronger stimulus, and no CR occurs in response to the weaker stimulus. This may explain the negative outcome of some previous experiments (37).

Third, the contradictory results of conditioning experiments may result from the fact that the CO₂ stimulus, in addition to its stimulatory effects through chemosensitivity, also has inhibitory effects on breathing mediated by an upper airway sensory reflex (1, 4). Conditioning studies have shown that when a US has several different effects, they may be conditioned at different rates (2, 10). This raises the question of whether the inhibitory or stimulatory effect of CO₂ is predominantly conditioned. When CO₂ is delivered through upper airways instead of intratracheally, its inhibitory effects may be at first associated with a CS. This may account for the conditioned inhibition of breathing reported by some previous authors (5).

In view of the above considerations, we carried out a controlled experiment in which rats were submitted to paired presentations of CO₂ stimuli and tones. The control procedure consisted of submitting a control group to the same number of unpaired CS and US. Conditioning was examined by comparing the ventilatory responses to test trials with CS alone at identical times in the two groups. Accordingly, any difference in the response to test trials would unambiguously reveal conditioning. Second, we attempted to avoid the overshadowing of the CS by the CO₂ stimulus by the use of a continuous masking somatosensory stimulus. We postulated that such a stimulus would prevent the early detection of CO₂ and reduce the predictive value of its sensory effect, thus facilitating the CS-US association. This would ensure conditioning, even though CO₂ was delivered through the upper airways. However, no prediction was made about the direction of the CR. We postulated that conditioning would be either inhibitory or stimulatory, depending on which of the effects of CO₂ would be predominantly associated with the CS.

METHODS

Table  1.   Summary of the experimental design Experimental Group Control Group Day 1  Familiarization + 70 min recording Familiarization with AA + 70 min recording Same as experimental group Day 2  6 Phases comprising 7 trials each (trials 1-7) Same as experimental group Trial 1: baseline Same as experimental group Trials 2-6: paired CO2/tone (acquisition trials) Trials 2-6, unpaired CO2/tone Trial 7: tone alone (test trial) Same as experimental group Day 3  6 Phases comprising 7 trials each (trials 1-7): Trial 1: baseline Same as experimental group Trials 2-7: tone alone Familiarization lasted 140 min. Sessions in day 2 and day 3 comprised 6 phases each. Each phase lasted 70 min and comprised 7 trials. Each trial lasted 10 min. AA, acetic acid.

RESULTS

Table  2.   Baseline ventilatory data Group n Without Acetic Acid With Acetic Acid TT, ms VT, AU  I, AU TT, ms VT, AU  I, AU Experiment 9 764 ± 276  25 ± 3  40 ± 13  781 ± 241  24 ± 28  36 ± 10  Control 7 697 ± 253  23 ± 4  41 ± 14  766 ± 397  23 ± 4  37 ± 11 Values are group means ± SD; n, no. of rats. TT, total breath duration; VT, tidal volume; I, mean ventilation; AU, arbitrary units. Between-group differences were not significant. I was significantly higher with acetic acid than without (see text).

DISCUSSION

This experiment showed that pairing a hypercapnic and an auditory stimulus elicits an inhibitory CR in rats after ~5-20 paired presentations of these stimuli. This response was characterized by higher TT and lower I values in the experimental compared with the control group. No significant effects were observed for VT. This response contrasted with the immediate response to the CS (a decrease in TT and an increase in I), which is typical of the physiological component of arousal. Therefore, the inhibitory CR neither potentiated the preexisting stimulating response to the tone nor was similar to the hypercapnic response to the US.

Our contention that the experimental group exhibited inhibitory conditioning in response to the auditory stimulus was based on two main arguments. First, because the only procedural difference between the two groups was the temporal contiguity between the CS and US, we postulated that the differences between the group responses to the CS were due to the learning of the association between the CS and US by the experimental group. Second, we attempted to ascertain whether the CR was specifically triggered by the CS or whether it was caused by contextual factors affecting the two groups differently. This issue is generally investigated by performing direct pre-post comparisons of the effects of the CS. However, in the present experiment, these comparisons were hampered because the breathing pattern of the experimental rats changed just before CS delivery as a result of learning. This is why the pre-CS ventilatory data did not provide a reference level suitable for assessing the effects of conditioning. By contrast, comparison of the breathing variables at CS delivery with the breathing variables for subsequent periods was relevant to the evaluation of the specificity of the CR in relation to the CS. This comparison showed that the group differences in the breathing patterns during the test trials (trial 7) were confined to a period of ~3 min after the CS and then vanished during the remainder of the test trial. It was, therefore, unlikely that the group differences after the CS were caused by a general contextual factor, because this would have affected breathing patterns throughout the entire test trial. Rather, we suggest that our data can be accounted for by specific effects of the CS.

Finally, because TT and I displayed baseline drifts, we addressed the possibility that the group differences observed in the test trials may have arisen because of long-duration changes brought about by the US. These changes may have had different effects in the two groups because, at the time of the test, the US of the preceding trial was given 3 min closer to the CS in the control than in the experimental group. We ruled out this possibility for the following reasons. First, the group differences in TT and I during the test trials were much higher than the baseline changes in these variables. Second, the fact that these group differences in the test trial emerged with practice in the late phases of day 2 could not be explained by aftereffects of the US, which would have been also observed in the early phases. In addition, had the group differences in the test been due to the fact that the US of the preceding trial was given 3 min closer to the CS in the control than in the experimental group, we would have observed parallelism between the ventilatory curves across the 10 min of the test trials with a lag of 3 min. There was no such trend in the data. Therefore, we ruled out the possibility that the group differences in the test were due to long-duration effects of hypercapnia.

Despite the above arguments in support of the specificity of the CR in relation to the CS, several aspects of our data may still go against this hypothesis. For example, during day 3, group differences in breathing patterns seemed unrelated to the CS. The protocols for day 2 and day 3 were markedly different. On day 2, each phase comprised five CS-US pairings (followed by 1 CS-alone trial), whereas on day 3, no US (i.e., no hypercapnia) was delivered. Therefore, there is no contradiction in the fact that days 2 and 3 data led to different results. In fact, day 3 data reflected the ventilatory behavior of the animals placed in the plethysmograph in which they had been subjected to repetitive CO₂ stimuli and receiving auditory stimuli. This context was far from neutral, and, as a matter of fact, in phase 1 the control group exhibited large transient decreases in TT and increases in I, which were not observed in the experimental group. However, these differences were not significant and, therefore, provided no further support to the conditioning effects observed on day 2. We do not totally rule out the possibility that contextual factors have a role in these nonspecific effects. The recent experiments by Mongeluzi et al. (23) provide a typical example of associations learned between an environment and the aversive effect of a 100% CO₂ stimulus. In these experiments, the conditioning and control environments differed as regards the size and lighting of the experimental room, the presence or the absence of an odor cue (vanilla), and the intensity of background noise. After exposure to a single test with 100% CO₂, freezing periods (i.e., the lack of any detectable body movement except for breathing, which was observed but not measured) were longer in the conditioning than in the control environment. In the present experiment, environmental cues (plethysmograph and visual or auditory cues from the laboratory environment) were strictly identical in the two groups and were, therefore, unlikely to explain between-group differences in breathing patterns. However, other contextual cues may have differently shaped the perceptual experience in the two groups. In particular, the different procedures applied to each group were associated with different levels of wakefulness and aversiveness. First, the sequential exposure to tones and hypercapnia in the control group yielded a higher rate of events, and possibly a greater arousal effect, with concomitant increases in breathing frequency and I. Second, the general aversiveness to the situation may have been greater in the control rats because, unlike the experimental rats, they were not warned of the oncoming aversive hypercapnic stimulus (20). The possibility that the contextual cues were associated with more highly aversive events in the experimental than in the control group may explain why the latter group displayed greater ventilatory activity. We may, therefore, postulate that our data can be explained, at least in part, by the association between the experimental context and stress.

Direct comparison of the hypercapnic response with the responses reported by previous authors was hampered by the fact that, in the present experiment, a tone was delivered during the hypercapnic stimulus. In fact, this response was roughly similar to previously reported responses (17); thus, the 8% CO₂ stimulus used in the present study elicited an ~100% increase in ventilation. However, repetition of the hypercapnic tests had different effects in the two groups. We observed that the experimental group exhibited higher I and lower TT at the end of the acquisition tests than did controls. The experimental rats may have anticipated the tone, as a result of having learned the association between the CO₂ stimuli and a temporal cue: the constant 10-min interval between two successive stimuli. Previous reports on the conditioning of rats to drug administration provide indirect support for this interpretation in terms of time conditioning. For example, conditioning of body temperature in rats, by using morphine as an US, yielded not only conditioned hyperthermia in response to a CS but also conditioned hypothermia 1 h before the expected morphine injection (10). However, in the present experiment, we feel that this possibility is unlikely because it would have occurred in the control group as well. In fact, we failed to observe any trace of a CR in the controls at the times corresponding to their hypercapnic tests. Alternatively, the changes occurring in the experimental group before the CS may be accounted for by effects of learning on the response to the hypercapnic stimuli. Previous authors have shown that, in some circumstances, the repetition of hypercapnic stimuli may induce changes in the VT and TT adopted to achieve a given level of I (for a review, see Ref. 14). These changes are poorly understood, but learning has been proposed as one of the underlying mechanisms (21). The respective merits of the two above interpretations are difficult to assess on the basis of the present data. However, this may easily be done in further experiments. A different design, in which the CS would be delivered at random time intervals, would prevent time conditioning from occurring but would not prevent the rats from learning to change their ventilatory response to hypercapnia.

The present finding of an inhibitory CR in the experimental group was consistent with the data of Biryukov et al. (5), although these authors also reported conditioned changes in VT, contrary to our present data. This difference may be explained by the stronger US used by these authors (50% CO₂), which presumably yielded a stronger conditioned inhibition than did the present 8% CO₂ stimulus. The significant results for inhibitory conditioning found here are at variance with other those of studies, in which the opposite stimulating effect was reported (16, 26), and also with the results of studies that failed to establish any conditioned change (37). The present experiment conditions differed from conditions in these studies in two major respects: the inclusion of a control group and the attempt to mask the onset of the US by acetic acid.

The control procedure consisted of unpairing the CS and US in a sequential control group. Our procedure minimized the total duration of the experiment by presenting the CS only 1 min after the ventilatory effects of the US had vanished. However, a general drawback to the sequential control procedure is that some conditioning may occur in the control subjects if they associate the CS with the US, despite the interval between them. In general, such an interval makes conditioning more difficult but does not necessarily prevent it (20). This possible conditioning of the control group may attenuate the differences between the two groups, thus leading to underestimation of the effects of conditioning. However, in the present study, within-group analysis of the controls did not reveal any conditioning in this group. Another drawback to our control procedure is that it may yield time conditioning, a possibility already considered above. An alternative control procedure would be to deliver the US and CS at random intervals to the control group, while excluding their simultaneous occurrence, to prevent any conditioning ("explicitly unpaired" control group). Under the present conditions, this control procedure would have lengthened the acquisition period, which might have been designed to cover successive days instead of a single day (day 2). However, despite the inherent limits of the sequential control used here, the present results show that the ventilatory effects of the tone were different, depending on whether they were previously paired with hypercapnia, which clearly established a conditioning effect.

The choice of acetic acid to mask CO₂ was based on previous studies showing that a somatosensory stimulus such as CO₂ could be masked by an irritant acting on the trigeminal fibers (15, 19). Among the variety of irritants exerting such action, we chose acetic acid, one of the less noxious (3). However, our contention that CO₂ was actually masked by acetic acid under the specific conditions of the present experiment was not based on independent validation studies. Given the rats' sensitivity to somatosensory stimulation (without any masking agent, a rat is able to perceive 0.52% CO₂) (38), we do not rule out that the rats detected the delivery of CO₂, even though this detection was delayed by acetic acid. In addition, CO₂ does not belong to the rats' habitual sensory repertoire, which probably made this stimulus more salient than the tone. Despite this, the present data show, first, that the sensory effects of CO₂ did not overshadow the tone enough to prevent conditioning from occurring. Second, they suggest that the negative findings of the previous experiments using a CO₂ stimulus delivered through the upper airways may be due to the overshadowing of the experimentally controlled CS by the sensory effects of CO₂ (37).

The present result of an inhibitory CR may be due to the inhibitory and aversive effects of CO₂. In addition to its strong activation of breathing through chemosensitivity, the CO₂ stimulus, at least above 8%, has inhibitory effects on breathing, presumably through the upper airway sensory reflex (1). In addition, the increase in VT, which is one component of the hypercapnic response, elicits an inhibitory vagal reflex by activating stretch receptors in the lung and chest wall. This raises the possibility that these inhibitory effects of hypercapnia were predominantly associated with the CS during the conditioning process. Ventilatory conditioning using aversive stimuli such as inhibitors of breathing has previously been reported. In the experiments of Orem and Trotter (25), the cats were trained to stop breathing in response to a small puff of ammonium hydroxide vapor at the onset of inspiration, and additional ammonium hydroxide was given if the cat failed to stop inspiration within 500 ms. The major outcomes were that conditioned apneas were associated with the inactivation of cells in the ventral and dorsal ventilatory groups and that some cells within that system became active during these apneas. We believe that these findings may also account for the present inhibitory effect.

We do not deny that the CR might have been mediated by instrumental contingencies between breathing and the aversive CO₂. This possibility is supported by two arguments: first, for rats, CO₂ concentrations are aversive above 8% (35), the level reached in the present experiment. Second, rats are capable of "voluntary" control of breathing to obtain a reward or avoid punishment, as shown by experiments in which changes in the breathing pattern were followed by either rewarding electric brain stimulations (13) or aversive electric shocks (22). The possibility that the present CR was an avoidance response to the aversive CO₂ constitutes an alternative interpretation of our data.

An objection may be raised that the inhibitory conditioning that stems from the aversive sensation caused by CO₂ is poorly suited to model the automatic processes that govern breathing patterns in normal humans. However, it is not impossible that spontaneous breathing patterns may be governed, at least partly, by responses shaped by the optimization of respiratory comfort. This is in line with the notion developed by Chonan et al. (7) that the minimization of respiratory sensations may play a substantial role in the adjustment of breathing patterns. In fact, these authors observed that voluntary changes in breathing frequency or minute ventilation at given levels of PCO₂ systematically intensified the sensation of dyspnea, suggesting that spontaneous patterns normally minimize dyspnea. Under natural conditions, respiratory sensations are not necessarily minimized through conscious and voluntary processes but rather through unconscious processes, possibly as a result of some kind of automatization. Within this framework, the conditioned inhibition of breathing as described in the present study may be relevant to the investigation of these processes.

In conclusion, this experiment showed that pairing an auditory stimulus with an 8% CO₂ stimulus led to an inhibitory CR. This conditioning probably resulted from the learning of an association between the auditory stimulus and the aversive or inhibitory effects of CO₂. The present results confirmed the high sensitivity of the respiratory controller to conditioning processes. However, the ability of conditioning by CO₂ inhalation to account for the conditioning processes that may occur under natural conditions requires confirmation by further experiments.

ACKNOWLEDGEMENTS

The authors are grateful to Stéphane Delavaud for technical contribution.

FOOTNOTES

Address for reprint requests: J. Gallego, Laboratoire de Neurologie et Physiologie du Développement, Hôpital Robert-Debré, 48 Bd Sérurier, 75019 Paris, France.

Received 2 April 1996; accepted in final form 5 June 1997.

REFERENCES