Brain Research, 517 (1990) 7-18 Elsevier

7

BRES 15464

Preservation of integrative function in a perfused guinea pig brain George B. Richerson* and Peter A. Getting Department of Physiology and Biophysics, University of Iowa College of Medicine, Iowa City, 1A 52242 (U.S.A,) (Accepted 17 October 1989) Key words: Perfusion; Perfluorocarbon; Artificial blood; Respiration; Brainstem; Guinea pig

The mammalian brain has been one of the most difficult organs to maintain using artificial perfusion. Normal biochemistry, histology, and electrophysiology of the brain have been demonstrated for limited periods in vitro, but it has been more difficult to maintain complex, integrative neuronal activity such as the electroencephalogram (EEG) or programmed motor output. Normal motor output, other than reflex activity, has not previously been demonstrated in a perfused brain preparation. This paper reports the first preservation of normal function in a complete motor network, including intact afferent and efferent pathways, during perfusion of the mammalian brain. The brain, rostral spinal cord and peripheral nervous system of the guinea pig were perfused in situ using an artificial blood containing the oxygen carrier, perfluorotributylamine (FC-43). This preparation was maintained normothermic, whereas many other perfused brain preparations have been maintained hypothermic to prolong viability. Survival was enhanced by the addition of HEPES buffer to the perfusion medium, probably by increasing carbon dioxide transport. The duration of normal EEG was extended to 8 h. Spontaneous respiratory motor output with normal waveform and temporal pattern was recorded from the phrenic nerve for an average of 6 h. The respiratory motor output responded appropriately to blood pCO2, temperature, blood flow, drug concentrations, and electrical stimulation of vagal afferent fibers. This preparation represents a significant advance in the ability to preserve neural function during perfusion, and should offer advantages for studying cellular electrophysiology of intact, functioning neural networks, as well as neurochemistry and neuropharmacology. INTRODUCTION

In vitro preparations of intact non-mammalian nervous systems have proven instrumental in the analysis of neural networks, by permitting stable intracellular recording from neurons within functioning neural systems under controlled experimental conditions 58. The in vitro mammalian brain slice does not permit this s a m e t y p e of approach, because neural circuitry is disrupted during the slicing procedure. Recently, in vitro approaches using superfusion have been applied to isolated, semi-intact mammalian nervous systems for electrophysiology in the hemisected neonatal rat spinal cord 25'54 and the neonatal rat spinal cord and brainstem 64,67. Peffusion has been used as an alternative to this approach in the isolated adult guinea pig brain 47-49 and the kitten brain 62. These approaches involve the use of immature tissue, the use of hypothermia to preserve viability and/or the complete isolation of the CNS, thereby removing normal afferent and efferent pathways. We have developed an in situ, perfluorocarbon perfused guinea pig brain preparation, that differs from previous in vitro preparations by preserving normal afferent and efferent pathways at normothermia in the adult. The validity of data from a perfused brain preparation

first depends upon a rigorous demonstration that the nervous system continues to function normally under these conditions. Earlier preparations perfusing the mammalian brain with perfluorocarbon emulsions 9'19' 36,63 blood 1'22'27'42-44'69'72'73 or saline 47-49'62 did not demonstrate normal prolonged integrative function. A preliminary paper 57 demonstrated the presence of integrative CNS function in this perfused preparation. The goal of the present paper was to evaluate the normality of this integrative CNS function by studying the performance of the brainstem respiratory network, including its response to its major afferents. The mammalian respiratory rhythm is generated by a neural network located within the pons and medulla which in turn projects to phrenic and intercostal motoneurons by bulbospinal pathways 23. The respiratory rhythm is modulated by a variety of central and peripheral afferent systems including pulmonary stretch receptors, baroreceptors, chemoreceptors, and temperature systems. The production and modulation of a normal respiratory rhythm depends upon the integrity of sensory and motor integrative processes distributed at several levels of the neuraxis. A b n o r m a l rhythmic output can be produced after extensive neural damage, but the normal respiratory pattern is altered by even small physiologic

* Present address: Department of Neurology, Yale University School of Medicine, New Haven, CT 06510, U.S.A. Correspondence: G.B. Richerson, 100 Hemlock Road no. 8-1, Branford, CT 06405, U.S.A. 0006-8993/90/$03.50 (~) 1990 Elsevier Science Publishers B.V. (Biomedical Division)

c h a n g e s in v a r i a b l e s s u c h as p O 2, p C O 2 a n d p H . W e m e a s u r e d t h e v i a b i l i t y o f t h e p e r f u s e d p r e p a r a t i o n b y its ability

to

produce

a

normal

respiratory

pattern,

a

sensitive measure of CNS function.

MATERIALS AND METHODS

Surgical technique A total of 75 perfusion experiments were performed. Adult guinea pigs (400-600 g) were anesthetized with pentobarbital (25 mg/kg, i.p.) and Innovar-Vet (0.5 ml/kg, i.m.). A cannula was inserted into the trachea for mechanical ventilation. The skull was removed over the cerebellum, and the dorsal surface of the medulla was exposed by aspirating the medial cerebellum. The thorax was opened by cutting the ribs lateral to the sternum and removing the sternum. Positive pressure ventilation was used after the thoracotomy with a positive end expiratory pressure of 3-4 cm H20. The left phrenic nerve was tagged with suture (4-0 silk) just proximal to the diaphragm, and the nerve was cut distal to the suture. Suture (1-0 silk) was also placed loosely around the descending aorta rostral to the diaphragm. The pericardium was opened, and another suture (1-0 silk) was placed loosely around the root of the ascending aorta. The apex of the heart was cut, and a perfusion cannula inserted through the left ventricle into the ascending aorta. The cannula was secured using the ligature previously placed around the ascending aorta. Perfusion with artificial blood was initiated within 15 s of cutting the apex of the heart and was restricted to the rostral half of the body by tying the ligature around the descending aorta. Incisions were made in the superior vena cava and right atrium to allow venous drainage. Approximately 50-100 mi of perfusate were pumped through the animal before the venous outflow was collected, thereby discarding the animal's own blood. To avoid contamination of perfusate from non-perfused regions, the lungs and the body below the ribs were surgically removed. The heart was removed to reduce movement caused by residual myocardial contraction. A bipolar cuff electrode was placed on the left phrenic nerve in the thorax. The cervical portion of the vagus nerve was left intact, although the input from the heart, lungs, and abdominal viscera were eliminated since these organs were removed. The animal was placed in a prone position and secured with ear bars, a mouth clamp, and a spinal clamp.

Preparation of artificial blood Four hundred grams of the non-ionic detergent Pluronic F-68 (BASF Wyandotte, Wyandotte, MI) was dissolved in 6 liters of distilled water and filtered (0.2-0.8/~m membrane filter). Four kg of perfluorotributylamine9 ' 19 •29 •36 (FC-43; 3M, Minneapolis, MN) were added to the solution of Pluronic and homogenized in a blender, producing a crude emulsion. Distilled water was added to increase the total volume to 9 liters. The crude emulsion was sonicated (power approximately 200 W) using a horn sonicator with a flow-thru cell attachment (Heat Systems, Farmingdale, NY). During sonication, the emulsion was cooled by submersion in an ice bath. After sonic treatment for 15-20 h, the emulsion was filtered using a 5/~m membrane filter. Hydrogen fluoride is released during sonic treatment at a rate that is proportional to sonication power, temperature, and time, and indirectly related to pressure 5. During sonic treatment the pH dropped slowly, suggesting a rise in fluoride concentration over time. NaOH (10 N) was added every 1-2 h to maintain pH between 6 and 8. After sonic treatment, fluoride was removed by dialysis against distilled water with a hollow fiber dialysis unit (135 SCE, CD Medical, Miami Lakes, FL). In 2 cases the fluoride concentration was measured with a fluoride-sensitive electrode and was found to be less than 3.0 ppm (260 nM) after dialysis. Dialysis of the emulsion was used routinely to minimize the possibility of adverse effects of fluoride ions; however, the necessity of this step to ensure brain viability was not tested directly.

The emulsion was divided into 10 equal portions (900 ml each) and stored in sealed containers at 4 °C where it remained stable for months. For each experiment, salts, drugs, N-2-hydroxyethylpiperazine N'-2-ethanesulfonic acid (HEPES; Sigma Chemical Co., St. Louis, MO) and distilled H20 were added to 900 ml of the emulsion to obtain one liter of artificial blood with concentrations in the aqueous phase as listed in Table I. Since approximately 20% of the total volume consisted of the hydrophobic FC-43 (the density of FC-43 is aproximately twice that of H20), reagent weights were calculated using 80% of the total volume as an estimate for the aqueous phase. Prior to use, the blood was filtered again using a 5 /~m membrane filter.

Perfusion apparatus The perfusion apparatus is shown schematically in Fig. 1. The artificial blood (ART. BLD.), contained in a perfusate reservoir, was bubbled with 0 2 and CO 2 which were controlled independently to maintain pO 2 and pCO 2 at 600-700 mm Hg and 40-55 mm Hg, respectively, pO2 and pCO 2 were measured directly in some experiments using a blood gas analyzer, but in most experiments were calculated from the flow meters. To control foaming, anti-foam A (Dow Corning, Midland, MI) was baked onto stainless steel wool which was placed in a hole in the top of the perfusate reservoir where the gas escaped. The pH of the blood was monitored continuously and was initially 7.4 with p C O 2 = 40 mm Hg. The pH decreased to 7.33-7.36 when the pCO 2 was increased to 50-55 mm Hg. The blood was pumped by a non-pulsatile gear pump driven by a brushless DC motor (Micropump, Concord, CA; maximum flow rate = 200 ml/min). The blood was pumped through an in-line borosilicate glass filter (25/xm nominal pore diameter), through a flow meter and into a coil of stainless steel tubing submerged in a heated water bath. From the heating coil, the blood flowed into the top of a debubble chamber (a cylindrical tube 1.2 x 8 cm). Large air bubbles were trapped at the top of the chamber as the blood flowed out the bottom. Temperature of the blood was monitored by a thermistor (YSI, Yellow Springs, OH) in the debubble chamber and was maintained by feedback control of the heater to 37 + 0.1 °C, except during the temperature experiments. Perfusion

TABLE I

Composition of artificial blood Substance

FC-43 Pluronic NaC1 KCI MgSO4

CaCI2-2H20 NaHCO 3 NaH2PO4.H20 Glucose HEPES Dexamethasone Pentobarbital Galamine Gentamicin Ampicillin

Concentration (aqueous phase)

115.0 mM 4.5 mM 1.2 mM 1.75 mM 20.0 mM 1.2 mM 11.0 mM 20.0 mM 200.0 nM* 50.0/~M*

Mass~liter

Present in dialysate

400.0 g/l 40.0 g/l 5.38 g/l 0.27 g/1 0.12 g/1 0.21 g/l 1.34 g/l 0.13 g/l 1.60 g/1 3.82 g/l 0.20 mg/l 10.00 mg/l 6.00 mg/l 7.30 mg/l 11.00 mg/l

No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

* The concentration of these compounds was calculated assuming equal partition between the aqueous and micellar FC-43 phases of the emulsion. The remainder of the concentrations were calculated assuming strict partition between aqueous and hydrophobic phases with the volume of the aqueous phase equal to 80% of the total volume.

pressure, monitored by a transducer attached to the debubble chamber, was controlled to within +1 mm Hg by feedback regulation of the pump speed. The optimal perfusion pressure for maintaining viability with minimal edema was 40-45 mm Hg. This pressure is at the lower limit of the range of mean blood pressure measured in the anesthetized guinea pig in vivo52'56'65. Perfusion at higher pressures resulted in increased edema. Venous return was collected and pumped by a roller pump through another in-line borosilicate filter (25/~m pore diameter) prior to being dialysed with a hollow fiber dialysis unit (artificial kidney - - 135 SCE, CD Medical, Miami Lake, FL) against 9 liters of dialysate solution. The dialysate solution contained the same concentration of ions and drugs but was missing Pluronic and FC-43 (Table I). The dialysate reservoir was sealed to prevent transfer of water due to the difference in oncotic pressure of the dialysate and the blood caused by the Pluronic (mol. wt. = 8300 Da). A one-way valve from the perfusate reservoir (ART. BLD.) to the roller pump ensured that additional artificial blood could be drawn from the reservoir if the venous return flow was less than the roller pump rate.

Monitoring of anesthetic level The EEG and PN were monitored continuously in all experiments to evaluate the level of anesthesia. In some experiments, anesthesia was evaluated before paralysis with Flaxedil by observing the corneal reflex and reflex withdrawal to pinching of a front toe. Additional pentobarbital was added to the perfusate and dialysate to deepen the anesthetic level if: (1) the corneal reflex was brisk; (2) the animal responded to toe pinch by withdrawing the limb; (3) the rate or rhythm of the PN changed in response to toe pinch; or (4) the EEG developed activity at frequencies above 15 Hz in response to toe pinch12"37'4°'53'71.Pentobarbital at a concentration of 10 mg/l (approximately 50/~mol) was found to fulfill these criteria and was used routinely in all perfusion experiments. After the addition of Flaxedil, the frequency content of the EEG and the response of the respiratory rhythm to toe pinch were used alone to evaluate the anesthetic level.

RESULTS

Electrical recording and stimulation To record the EEG, a burr hole (2-3 mm diameter) was drilled over the left parietal cortex and an Ag/AgCI electrode was placed on the surface of the cortex. The electrode was fixed in place with dental acrylic (Duz All, Coralite Dental Products, Chicago, IL). The reference electrode was fastened onto the contralateral ear. The EEG was amplified (xl0,000; bandpass 1.6-50 Hz), continuously displayed on a chart recorder, and stored on magnetic tape. The phrenic neurogram (PN) was recorded using a bipolar silastic cuff electrode and was amplified (bandpass 0.5-10 kHz) rectified and low pass filtered (single pole, time constant = 100 ms) to produce the integrated phrenic neurogram (IPN). The effect of vagal afferents on the respiratory rhythm was examined by electrical stimulation of the vagus nerve. The cervical portion of either vagus nerve was placed on a pair of bipolar hook electrodes and stimulated with constant current pulses (50-100/~s duration). In two experiments, a second recording electrode was placed on the vagus nerve proximal to the stimulating electrodes. The stimulus current was increased to the threshold level for inhibition of inspiratory activity in the phrenic nerve. This current level produced a single, short latency volley in the vagus nerve typical of the activation of the most rapidly conducting fibers. Current pulses greater than these produced two additional peaks of activity in the vagus nerve with longer latencies typical of more slowly conducting fibers.

~HEATER CHAM~BER. DEBUBBLE

FLOW METER FILTER

CANNUL~A PUMP DIALYSATEI A --

VENOUS RETURN

CO2 -..~"].~ Fig. 1. Schematic diagram of perfusion apparatus. The perfusate reservoir (lower left) contained the artificial blood (ART. BLD.). The drawing in the upper right shows the arterial system and the point of insertion of the perfusion cannula in the ascending aorta. The descending aorta was ligated with suture at the level of the diaphragm.

Cortical activity during perfusion The E E G was defined as n o r m a l if the peak-to-peak amplitude was greater than 50/~V and there was no sharp activity. The frequency of activity was in the delta range, consistent with anesthesia. The average duration of normal cortical activity was 3.7 + 2.8 h ( m e a n ___ S.D.; n = 9). In the most successful experiment, the E E G r e m a i n e d n o r m a l for 8 h. I n 6 of 9 preparations the E E G was sustained for longer than 2 h. In one case, the E E G was a b n o r m a l from the onset of perfusion. The E E G was very sensitive to perfusion pressure and pCO2, which may have contributed to the large variability in m a i n t e n a n c e of a normal E E G . If the perfusion pressure fell below 40 m m Hg, or if p C O 2 was less than 40 m m Hg, the amplitude of the E E G decreased rapidly; therefore, the E E G rapidly became a b n o r m a l during experiments evaluating the effect of p C O 2 on respiratory motor output. Spectral analysis of the E E G 57 revealed two characteristics to be invariant in all animals that were sufficiently anesthetized to eliminate purposeful m o v e m e n t s in response to pinching of the toes. These were: (1) p r e d o m i n a n t frequencies in the 1-5 Hz range; and (2) the absence of activity above 15 Hz.

Spontaneous respiratory output during perfusion The pattern of respiratory activity recorded from the phrenic nerve in the cat 1° and guinea pig 56 is characterized by a sudden onset of activity at the beginning of inspiration followed by an increasing a m o u n t of activity throughout inspiration. The beginning of expiration is signalled by the rapid t e r m i n a t i o n of phrenic nerve activity with no activity during expiration except, on occasion, a short period of weak activity immediately following the end of inspiration. The perfused preparation also expressed spontaneous phrenic nerve activity with this pattern. The criteria used to judge ' n o r m a l ' function are critical

10 in evaluating the viability of a perfused preparation. This is particularly important for the respiratory system which responds very sensitively to small physiologic changes in p C O z and p O 2 , but continues to produce abnormal rhythmic activity in the face of severe ischemia. For this reason, the criteria we have used to judge respiratory activity as normal are presented in detail. Four quantitative measures were used to evaluate the normalcy of the respiratory pattern expressed by the perfused preparation. First, burst rate had to be within the normal range for breathing in the guinea pig (16-67 breaths/ min) 13. The average burst rate during perfusion was 28.4 + 6.7 bursts/min (mean + S.D.; n = 9, T = 37 °C, p C O 2 = 55 mm Hg) which compares favorably with the respiratory rate of intact, anesthetized guinea pigs (28.0 + 13 breaths/min) 56. Note that the standard deviation of respiratory rate in the perfused preparation was less than that seen in vivo. Second, the burst rhythm had to be regular. If the standard deviation of cycle period (Ttot) was greater than 20% of the mean Ttot, the rhythm was judged to be irregular. Third, the waveform of the IPN had to show a rapid onset, augmentation, and abrupt termination as described above. Fourth, there had to be less than one augmented burst (augmented breath) 3'7'31 per 10 cycles. These criteria were used to quantify the duration of normal respiratory motor output during perfusion. The phrenic nerve activity was normal for an average of 6 + 0.9 h (mean + S.D.; n = 9). The respiratory output remained stable during this period, but thereafter deteriorated rapidly. All activity ceased after an average of 6.9 + 0.8 h (mean + S.D.; n = 9). Four separate criteria were used to evaluate the respiratory output, because each could become abnormal in the absence of changes in the others. Fig. 2 shows examples of the types of patterns encountered which were judged to be abnormal. Fig. 2A shows an abnormally slow cycle rate (3.3/rain) despite a regular rhythm with normal IPN waveform and rare augmented bursts. In this case the only abnormality was a slow rate. Fig. 2B shows an example of an irregular rhythm. The average cycle period was 2.4 s (rate = 25 bursts/min), with a standard deviation of 1.5 s or 63% of the mean (range = 0.8-4.2 s). The standard deviation of the cycle period was greater than 20% of the mean; therefore, this pattern was judged irregular. There is clearly a repeating pattern present, but due to the large variation in expiratory duration from breath to breath (a respiratory 'bigeminy'), this pattern cannot be considered normal. Fig. 2C shows burst patterns with abnormal waveforms. The respiratory rate was also abnormal in this case (8 cycles/min). The most common abnormality was a large number of augmented bursts. Augmented breaths (deep breaths or sighs) normally occur at a low frequency 3'4'31, but can

increase in frequency under abnormal conditions including ischemia, hypoxia and hypercarbia, probably due to activation of lung irritant receptors and/or peripheral chemoreceptors e8. The phrenic output during augmented breaths (augmented bursts) is characterized by a biphasic trajectory during inspiration. The waveform of the IPN during early inspiration is the same as that during normal breaths. During late inspiration, there is an increase in rate of rise of the integrated phrenic nerve waveform and an increase in amplitude of the burst to near maximal levels. Fig. 2D shows an abnormal IPN pattern with augmented breaths every other cycle.

Modulatory influences The respiratory motor output is normally sensitive to a variety of afferent signals including those from mechano-, thermo- and chemoreceptors. Normal respiratory activity includes responses to modulatory influences. To determine if the perfused brain responded to these factors, pulmonary stretch receptor pathways were electrically stimulated, and p C O 2, temperature, bloodflow and drug concentrations were systematically varied while monitoring phrenic nerve activity. Vagal afferents. The lungs were removed during perfusion, so pulmonary stretch receptor input carried via the vagus nerves was absent. Electrical stimulation of the vagus nerve, in the absence of lung inflation, can be used to test the neural correlate of the Breuer-Hering inflation reflex if the stimulation is timed phasically to the phrenic nerve discharge 7°. Either the left or right vagus nerve was stimulated every third cycle, for 60 consecutive cycles, starting 75 ms after the onset of the inspiratory burst recorded from the phrenic nerve (Fig. 3A). Stimulus

A

B

~

C

Im

~|

I|

I|

wll l!

I!

I]

D

Fig. 2. Abnormal phrenic nerve activity patterns. The traces in each panel are the IPN (top) and corresponding PN (bottom). A: abnormally slow burst rate of 3.3 cycles/rain. B: irregular rhythm. C: abnormal waveforms with two peaks. Inspiratory bursts with 3 peaks were also seen. D: frequent augmented bursts (cycles 3, 5, 7, 10, 12 and 15 are augmented bursts). All scales = 6 s.

11 parameters were adjusted to activate the fastest conducting fibers, which include the slowly adapting pulmonary stretch receptors. During each cycle in which the vagus nerve was stimulated, the inspiratory burst in the phrenic nerve was shortened and the onset of the next burst was advanced (Breuer-Hering reflex). The average IPNs were calculated from the control and stimulated respiratory cycles (Fig. 3B). Vagus nerve stimulation resulted in a decrease in duration and maximum amplitude of the inspiratory burst with no change in the rate of rise of the IPN. The shortening of the inspiratory burst was highly reproducible from cycle to cycle, as indicated by the small standard deviations. A similar effect was observed in intact guinea pigs using the same stimulus paradigm 56, and has been reported in other species 7°. The effect of vagal stimulation on inspiratory burst duration was dependent on stimulus frequency. As the frequency of vagus nerve stimulation was increased from 50 to 200 Hz, inspiratory duration and maximum IPN amplitude were decreased in a graded manner. Maximal inhibition of

A

inspiratory activity was produced at approximately 200 Hz. Stimulation of the vagus nerve during expiration resulted in a prolongation of the expiratory phase and resetting of the inspiratory burst pattern (Fig. 4A; note stimulus artifact). With prolonged stimulation, the expiratory phase could be maintained for at least 30 s. This response is the neural correlate of the Breuer-Hering expiratory prolongation reflex. Stimulation of the vagus nerve during inspiration at higher current levels than those used above resulted in the production of an augmented burst (Fig. 4B). The initial portion of the IPN in the stimulated cycle was similar to the control bursts; however, approximately 150 ms after the onset of stimulation the rate of rise of the IPN increased and attained a greater maximum amplitude than the control bursts. This response represents the neural correlate of Head's paradoxical reflex, and is thought to result from activation of rapidly adapting pulmonary stretch receptor afferents 41 p C O 2. The effect of chemoreceptor input on respiratory motor output was studied by systematically varying perfusate p C O 2. Perfusate p H was allowed to change with p C O 2 (-0.044 p H units/1 m m H g pCO2). Changes in p H typically stabilized within 5 min following a step change in CO2/O 2 gas flow rate. Fig. 5A and B show

A B 10080 zO.

m

60 40

B

2O

0.2 TIME

0.6 (sec)

1.0

Fig. 3. Effect of vagal nerve stimulation on phrenic nerve discharge pattern. A: IPN (top trace) and PN (middle trace) recordings during vagus nerve stimulation. The vagus nerve was stimulated electrically with 50/~s pulses at a frequency of 200 Hz for 1 s during the 2nd and 5th cycle, as indicated by the bars (bottom trace). Scale = 1 s. B: comparison of the IPN waveform with (STIMulated) and without (CONtrol) vagal stimulation. The solid lines show the mean IPN calculated from approximately 20 respiratory cycles. The dotted lines indicate the standard deviation around the mean IPN calculated at 10 ms intervals. The control IPN was calculated from every third respiratory cycle, using the burst immediately preceding vagal stimulation (for example, the 1st and 4th bursts of A). The arrow indicates the time of onset for vagal stimulation (75 ms after onset of the inspiratory burst). The scale on the left for IPN amplitude was normalized to the mean peak control IPN.

Fig. 4. A: prolongation of expiration by vagal stimulation. IPN (upper trace), PN (middle trace), and vagal stimulation (lower trace). The vagus nerve was stimulated (67 Hz, 50 gs pulses) for 12 s starting about 1 s into the normal expiratory phase. Scale = 4 s. B: augmented burst caused by stimulation (200 Hz, 50 gs pulses, for 600 ms) of the vagus nerve at high current intensity. The stimulation was timed to start 100 ms after onset of the inspiratory burst. Scale =ls.

12 phrenic nerve discharges at two different pCO 2 levels. An increase in pCO2 from 33 mm Hg (pH 7.43) to 64 mm Hg (pH 7.30) resulted in an increase in the rate of rise and maximum amplitude of the average IPN. The increase in IPN amplitude was graded with pCO 2 (Fig. 5C). Assuming a linear relationship, %IPN amplitude = 5.26 pCO 2 - 126%; r = 0.98 (IPN amplitude of 100% = IPN amplitude at a pCO2 of 43 mm Hg). Extrapolation of this line to the abscissa (dashed line) yielded an estimated apneic level of 24 mm Hg. The mean duration of inspiratory activity (Ti) , expiration (T~), and Ttot increased only slightly with increasing pCO 2. This experiment required 2 h to complete due to the time needed to reach equilibrium after changes in pCO 2. Large changes in pCO 2 resulted in several technical difficulties. High pCO 2 caused a large number of augmented breaths or abnormal IPN waveforms. Low pCO 2 resulted in permanent changes in the respiratory output and EEG, probably due to vasoconstriction and the resulting ischemia. One experiment was completed over the range of pCO 2 shown in Fig. 5. In 3 other experiments in which pCO 2

A

33

mm

Hg

changes were made over a more narrow range, the IPN amplitude also increased nearly linearly with increasing pCO2 with little effect on the temporal pattern. Temperature. The effect of temperature on the respiratory motor pattern was studied by systematically varying perfusate temperature. Decreasing temperature from 37 to 31 °C caused a decrease in burst rate. For the preparation shown in Fig. 6, the rate dropped from 33 to 18 bursts/min. In Fig. 6C, Ti is plotted as a function of Te at 4 different temperatures. The change in Ti and Te followed the relationship Ti = 0.49T~ - 0.235 s (r = 0.99) where Ti and Te are expressed in seconds. In 2 of 3 preparations, IPN amplitude decreased with temperature, as shown in Fig. 6; while in the third preparation, the amplitude remained constant. Increases in IPN amplitude with decreasing temperature were not observed. Respiratory activity ceased completely when the temperature was decreased below a critical level (from 27-31 °C in different preparations at p C O 2 = 40 mm Hg). Ischemia. The effect of ischemia was investigated in two experiments by turning off the perfusion pump. In one experiment, after turning the pump off, the phrenic nerve discharge initially developed primary apnea 45, then displayed a gasping pattern 5°'51'6°'61. The gasping activity

A B

64

rnm

T = 37Oc

Hg

H+H+H-HI ii

i/iiili

I llllllllll

B

T = 310C

A A ,d

C o. 200

J

-- 100

20

40 PCO 2 (ram Hg)

C %==

-,~

A

,A

&

,~

2.0 1.5

~1 °c 1.0

60

Fig. 5. Modulation of respiratory pattern by changes in p C O z at constant temperature (37 °C). A and B: IPN (top trace) and PN (bottom trace) at p C O 2 = 33 mm Hg (A) and p C O 2 = 64 mm Hg (B). Scale = 4 s. C: plot of peak IPN amplitude as a function of p C O 2 from 1 of 4 perfused preparations. Each point is the mean peak IPN amplitude from at least 20 contiguous bursts. The points (squares) were fit by a least-squares line given by % IPN amplitude = 5.26 p C O 2 - 126 (r = 0.98). The IPN amplitude did not return to the control level for the last data point (*) possibly due to ischemic damage caused by vasoconstriction at the low p C O 2 (33 mm Hg) of the previous data point.

~-~

35 °C.~

0.5 0

0.5

1.0 Te

1,5

210

2~5

(see)

Fig. 6, Modulation of respiratory pattern by temperature at constant pCO 2 (55 mm Hg). A and B: [PN (top trace) and PN (bottom trace) recorded at 37°C (A) and 3 ] ° C (B). Scale = 4 s. C: plot of inspiratory duration (Ti) versus expiratory duration (T=) at 4 temperatures. Each symbol is the mean ( + S.D.) T i and T¢ from at least 20 contiguous bursts. The least squares line is given by T i = 0.49T~- 0.235 (r = 0.99).

13

continued for 1.5 min before all activity ceased permanently (terminal apnea) 4s. A similar effect of ischemia has been reported in other species in vivo5°'51. In contrast to respiratory activity, the amplitude of the E E G declined rapidly after about 15 s of ischemia and remained isoelectric. In a second experiment, the perfusion pump was turned off transiently. Fifteen seconds after the onset of ischemia, the phrenic nerve discharge showed primary apnea and the E E G became isoelectric. The perfusion was restarted after 22 s of ischemia. The amplitude of the E E G returned to normal within 30-40 s; however, all PN activity was absent for 3 min. Five minutes after turning the perfusion pump back o n , rhythmic activity was present in the PN, but it was abnormal due to an irregular rhythm and a large number of inspiratory bursts with abnormal biphasic waveforms. The PN did not return to control until approximately 10 min after restarting perfusion. Pharmacologic manipulations. To demonstrate that this preparation responded to changes in perfusate composition, the concentration of pentobarbital was altered systematically and the respiratory output was recorded. One of the advantages of this type of preparation is tight control of drug concentrations due to the large volume of blood and dialysate compared to the animals body volume (dialysate + blood = 11 liters, animal's body volume = 0.2 liters). Fig. 8 shows the effect of changing the concentration of pentobarbital from 50 to 86 pM. The phrenic nerve activity decreased in amplitude, with little change in cycle frequency, similar to the effect produced by a decrease in pCO 2. This was accompanied by a change in the E E G from high amplitude continuous irregular delta activity to bursts of

50uMNembutal

86uMNembutal ~ 1 ~l~~~l~~t~

Artery

Artificial Blood

Without HEPES

Capillary

pCO2__4 0,.,.Hg~.~ pH=7.4 f

Vein

~ / ~ C O2= 94,.,.Hg ~ ~) p H = 7 . 0 3

Tissue COz Production = 2.4,~

With HEPES

pH=7.4

f

)

- " ~ . . p H=7.32

Fig. 8. Calculations of venous pCO 2 and pH using equations in text. The arterial and venous pCO 2 and pH are shown for a perfiuorocarbon emulsion with and without a non-bicarbonate weak acid buffer (pK a = 7.5).

high amplitude delta activity separated by periods of relative attenuation of the background rhythms. This is consistent with deepening anesthetic level. These effects were graded with intermediate pentobarbital concentrations. Thyrotropin releasing factor (TRF) stimulates respiratory activity through a mechanism unrelated to its action in the hypothalamus 33'35. TRF is localized in nerve terminals in the nucleus tractus solitarius (NTS) 35, and produces bursting of NTS neurons in an in vitro brain slice 17, which may be related to its effects on respiration. To determine if TRF stimulated respiratory activity in the perfused preparation, TRF was added to the perfusate after suppression of respiratory motor output by pentobarbital. TRF at a concentration of 700 nM (pentobarbital concentration = 86/~M), resulted in an increase in amplitude of the phrenic nerve activity with a slight decrease in cycle frequency. The amplitude of the IPN returned to the baseline level observed prior to the addition of either agent. The effect of an intermediate concentration of TRF produced a graded response in the respiratory activity. There was little effect on the E E G at this concentration of T R E In two experiments, calcium was not added to the perfusate. Despite this, a respiratory rhythm continued for approximately 1.5 h after initiation of perfusion, suggesting that blood-brain barrier remained intact during perfusion.

Effect of HEPES

7OOnMT.R.F. Fig. 7. Effect of pentobarbital and TRF on respiratory activity and the E E G . The IPN is shown in the traces on the left. The E E G is

shown on the rightl. Top traces: [Nembutal] = 50 ~M: [TRF] = 0. Middle traces: [Nembutal] -- 86 ,uM; [TRF] -- 0. Bottom traces: [Nembutal] -- 86 ~M; [TRF] = 700 nM. Scale 5 s, 50 ~V.

One of the key factors which allowed us to demonstrate prolonged normality of integrative function during perfusion was the use of HEPES as a non-bicarbonate weak acid buffer. Perfluorochemical emulsions are efficient transporters of 0 2 at high pO2, but CO2 carriage is limited. In whole blood, CO 2 is carried mainly in the

14 form of HCO3-. The quantity of HCO 3- formed is increased by the presence of weak acid residues associated with hemoglobin and plasma proteins. Neither Pluronic F-68 nor FC-43 have significant weak acid buffering properties, so the CO 2 carrying capacity of artificial blood without a non-bicarbonate weak acid buffer is limited. To overcome this potential problem, the buffer HEPES was added to function as a non-bicarbonate, weak acid buffer, thereby increasing CO 2 transport. After the addition of HEPES, the duration of viability of perfused preparations significantly increased. In 9 experiments without HEPES, the duration of normal phrenic nerve activity was 0.8 + 0.6 h (mean + S.D.; maximum of 1.8 h). Augmented breaths occurred frequently (up to one every other cycle) in experiments in which HEPES was not added to the artificial blood, and the respiratory motor output did not respond to changes in p C O 2. The duration of normal E E G activity was 2.8 + 2.6 h (mean + S.D.; maximum of 6.5 h). With the addition of HEPES the duration of normal phrenic nerve activity increased to 6 + 0.9 h (mean + S.D., n = 9; range = 5-8 h) and augmented bursts were rare (usually absent). The duration of normal E E G increased to 3.7 + 2.8 h (mean + S.D.; maximum of 8 h). DISCUSSION All aspects of neural activity are not equally sensitive to ischemia/anoxia. Integrative functions of the CNS are the most sensitive. For example, consciousness in man can only be maintained for 6 s of ischemia 59, and the E E G in various mammals (including man) disappears after only 10-30 s of ischemia 11'34'46'6s. Synaptic transmission is much more resistant, continuing for more than 4 min of anoxia in the guinea pig hippocampal slice32 and for 15 min of anoxia in the superior cervical ganglion of the rat 2°. During anoxia, axonal conduction is extremely hardy, and can be maintained for 30 min in the sural nerve of the rabbit 24 and 25 h in the superior cervical ganglion of the rat 21. Biochemistry and histology are less sensitive than cellular electrophysiology. For example, arterial p O 2 levels of 25 mm Hg are required to deplete cellular ATP 26, and changes in cellular morphology at the light microscopic level do not become evident for 24-72 h after total occlusion of cerebral blood flow55. Thus, cellular electrophysiology, synaptic transmission, biochemistry and histology are less sensitive to ischemia/ anoxia than integrative functions of the nervous system. Validation of a claim of maintained function in a new preparation therefore requires a definition of which functional parameters are measured. It is often assumed that cortical function is more sensitive to ischemia/anoxia than respiratory output. It is

true that respiratory function is relatively preserved after recovery from global ischemia, such as occurs after cardiopulmonary arrest. The brainstem may autoregulate more effectively than the cortex, and brainstem neurons may have intrinsic properties allowing them to resist permanent damage from prolonged ischemia. Paradoxically, the respiratory motor output is extremely sensitive to minor hypoxia/ischemia, because the respiratory centers subserve the function of maintaining adequate blood oxygenation. The frequency of augmented bursts is known to be extremely sensitive to hypercarbia. In cats, the number of augmented bursts increases from less than one every 2 min to one every 15 s aspCO2 increases from 40 to 57 mm Hg 7. Hypoxia also stimulates augmented breaths and has a synergistic effect with hypercarbia. At a pCO 2 of 45 mm Hg and 10% 02, the number of augmented breaths increases to one every 10 s7. Hypercarbia and hypoxia of this degree only result in minor changes in E E G frequency. Thus, changes in frequency, amplitude and shape of the respiratory motor output are a sensitive measure of tissue perfusion. The difference between responses of the E E G and the respiratory motor output to transient and prolonged ischemia (see Results) support this view, with the respiratory output continuing (as a gasping pattern) for a longer period than the E E G during prolonged ischemia, but recovering more slowly than the E E G after transient ischemia. For probably related reasons, the PN was relatively more improved than the E E G by the use of HEPES. Previous techniques for perfusion of the mammalian brain have preserved reflex activity6'18"44, cellular electrophysiology including membrane potential and synaptic conduction 47-49, biochemistry42'73, and histology22'69'73. Perfusion has been less successful at maintaining integrative CNS function. E E G activity has been detected for up to 10 h of perfusion43; however, in this case as well as most other cases in which it has been recorded 1'19' 27.63,69.72 the E E G deteriorated and became abnormal within 1-2 h of perfusion. The longest that the E E G has been documented to be unchanged from baseline during perfusion at normothermia is 4 h 36. Normal motor output, other than reflex activity, has not been previously demonstrated in a perfused mammalian brain. For example, respiratory motor output has been observed during perfusion 18"69, but the characteristics of this activity were not analyzed, and gasping or other abnormal patterns of respiratory activity were not excluded. We have used two integrative CNS functions, the E E G and respiratory motor output, to evaluate CNS viability during perfusion. The spontaneous respiratory motor output was analyzed using four criteria (rate, rhythm, waveform and frequency of augmented bursts) to eval-

15 uate normality. Without applying these standards, the

At equilibrium, an aqueous solution containing strong

abnormal respiratory patterns shown in Fig. 2 may have been considered proof of successful perfusion.

ions, a weak acid and carbon dioxide must satisfy the following equations.

Importance of a weak-acid buffer After the addition of the weak-acid buffer H E P E S to the perfusion emulsion, the length of time that normal respiratory output was produced ,increased from a mean of less than 1 h to a mean o ( 6 h. The number of augmented bursts decreased, and the respiratory output responded appropriately to changes in p C O 2. This improvement was consistent with an increase in CO 2 transport and concomitant increase in tissue pH with the use of HEPES. Several lines of evidence suggest that perfluorocarbon emulsions do not transport CO2 adequately. Clark et al. 9 observed an increased arterio-venous difference in p C O 2 during whole body perfusion of dogs with solutions of perfluorocarbon emulsions, but not with blood. An uncompensated respiratory acidosis was suspected, but the addition of a weak acid buffer was not attempted. Instead arterial p C O 2 was maintained at very low levels to decrease tissue p C O 2 and to prevent acidosis. Deutschmann et ai.18 measured arterial and venous p C O 2 in a peffused spinal cord and brain preparation of the adult rat. When perfused with a perfluorocarbon emulsion, the arterio-venous difference o f p C O 2 was 31.5 mm Hg (compared to the normal value of 5 mm Hg in vivo). The following theoretical considerations explain why the presence of a non-bicarbonate weak acid buffer results in an increase in CO 2 transport and a decrease in the arterio-venous difference in p C O 2. As CO 2 diffuses into blood in the capillaries, it combines with H 2 0 to form H C O 3- and H ÷ as shown in the following equation:

Water dissociation equilibrium:

CO 2 + H20 ~

H C O 3- + H +

Most carbon dioxide carriage in whole blood in vivo is in the form of HCO3-. As CO 2 diffuses into blood in the capillaries, little H C O 3- forms unless H ÷ is buffered by a non-bicarbonate weak acid, driving the above reaction to the right. The change in p C O 2 per change in total CO 2 content (dpCO2/dTco2) for whole blood at p C O 2 = 40 mm Hg is I mm Hg/voi.%. In contrast, the dpCOE/dTco 2 for a 20% w/v emulsion of FC-43 without a nonbicarbonate weak acid is 10 m m H g / v o l . % 29. T h u s , the increase in blood p C O 2 in response to the production of a given amount of CO 2 by the tissue, would be 10 times greater for a 20% w/v emulsion of FC-43 without a non-bicarbonate weak acid buffer compared to whole blood. This would result in a corresponding greater decrease in pH.

[H +] x [OH-] = K'w Weak acid dissociation equilibrium: [H+] x [A-I = KA x [HAl Weak acid conservation: [HA] + [A-] = [Atot] H C O 3- formation: [H+] x [HCO3- ] = K c x p C O 2 CO3 2- formation:

[H +] x [CO32-] = K 3 × [HCO3- ] Electrical neutrality: SID + [H +] - [HCO3- ] - [A-] - 2[CO32-] - [OH-] = 0 where K" w = H 2 0 dissociation constant = 4.4 x 10 -14 (Eq./l)2; K A = weak acid dissociation constant; H A = undissociated weak acid; Ato t = total weak acid; Kc = equilibrium constant for H C O 3- formation = 2.58 x 10-11 (Eq./l)E/mm Hg; K 3 = equilibrium constant for CO32- formation = 6.0 x 10 -11 Eq./l, and; SID -- strong ion difference = [strong cations] - [strong anions]. In this set of equations SID, Ato t and p C O 2 a r e the only independent variables. An iterative computer program derived from these equations was used to calculate the theoretical venous p C O 2 in the cerebral vessels during perfusion with a 20% w/v emulsion of FC-43 with and without H E P E S buffer. Assumptions were made for arterial flow rate (55 ml/min/100 g), CO 2 production by the brain (2.4 mmol CO 2 per liter of blood), arterial p C O 2 (40 mm Hg), and arterial pH (7.4) (derived from ref. 39). Using the equations above, cerebral venous p C O 2 w a s calculated to be 94 mm Hg (pH 7.03) in the absence of H E P E S buffer. If a non-bicarbonate weak acid buffer with p K a = 7.5 (e.g. HEPES) was present at a concentration of 40 mM, v e n o u s p C O 2 would only be 52 mm Hg (pH 7.32; see Fig. 8). Thus the addition of H E P E S would decrease the change in p C O 2 and the change in pH from arterial to venous blood as a result of tissue CO 2 production, and thus help prevent tissue

16 acidosis. This role of non-bicarbonate weak acid buffers in facilitating CO 2 transport has not been appreciated in previous attempts to use perfluorochemical emulsions for perfusion of organs in animals or transfusions in humans.

Comparison with other in vitro and perfused preparations The technique described here differs from other in vitro techniques using neonate 25'54'64'67 or adult systems 47-49. In our approach, the brain is left in situ, reducing trauma to the CNS, peripheral nerves, and vascular supply during surgery. Cannulation of the aorta permitted rapid initiation of perfusion, minimizing ischemic time to less than 15 s. Complete isolation of the nervous system requires substantial dissection and a concomitant increase in ischemic time. Using in situ perfusion, connections between the CNS, peripheral nerves, sensory receptors, and motor apparatus are preserved, making it possible to study the processing of afferent information. The use of artificial blood with HEPES increased viability and allowed the preparation to be maintained at a normal physiological temperature (37 °C). Many mammalian superfused 25'54'64'67 and perfused 1'19'36'48'49'62'63'72 CNS preparations have required lowered temperature (15-32 °C) to achieve prolonged survival; however, lowered temperature can alter CNS function, as illustrated by the temperature dependence of the respiratory rhythm (Fig. 6). The preparation described in this paper uses adult tissue. Neonatal or fetal tissue preparations survive more easily than adult tissue in vitro due to a greater resistance to anoxia. Neonatal preparations are important for studying neural development, but a complementary armamentarium of preparations using adult tissue is also necessary to place developmental data in the context of the mature system. Effect of changes in pCO 2 and temperature An increase in pCO 2 resulted in a nearly linear increase in IPN amplitude, with no significant change in cycle period or timing (Fig. 5). This finding is consistent with results obtained in the cat after vagotomy8'3°, and supports the hypothesis that timing of the respiratory rhythm generator in the absence of vagal feedback is independent of amplitude. During most experiments, the pCO 2 was maintained at 55 mm Hg because the higher level of pCO 2 resulted in recruitment of medullary respiratory neurons. Although this level of pCO 2 is higher than that seen in awake, unanesthetized animals, the pattern of respiratory activity was not qualitatively different from that seen at pCO 2 = 40 mm Hg. In addition, the higher level of pCO 2 was found to enhance

viability, possibly by cerebral vasodilation and thus increased tissue perfusion. Decreasing temperature (Fig. 6) resulted in a dramatic increase in cycle period (decrease in respiratory rate). When pCO: was maintained constant at 40 mm Hg, respiratory activity ceased entirely at temperatures below 27-31 °C. Altering pCO 2 or temperature affected different aspects of respiratory pattern generation, with the predominant effect of changes in pCO 2 on IPN amplitude, and of temperature on cycle timing. Any theory of central respiratory rhythm generation must explain this independence of cycle rate and IPN amplitude after vagotomy, as well as the exquisite sensitivity of respiration to temperature.

Control of blood composition An important property of a perfused preparation is the ability to control the blood composition. This is difficult in whole animals due to tissue absorption, metabolism and protein binding. A small increase in pentobarbital concentration (which is freely permeable across the blood-brain barrier) rapidly resulted in a marked suppression of the IPN and the EEG. The neuropeptide TRF (which may cross the blood-brain barrier via membrane carrier systems or by slow diffusionTM)antagonized the effect on the IPN of pentobarbital, but did not affect the EEG. A similar effect of low concentrations of TRF has been noted in intact animals, with a dissociation between the effects on the respiratory output and the EEG, while higher concentrations of TRF result in cortical activation 2. Higher concentrations of TRF were not used in the perfused preparation in an attempt to activate the E E G , due to the concern that this would result in reversal of the anesthetic state. CONCLUSIONS We have previously demonstrated that intracellular recording is possible from medullary respiratory neurons in this perfused preparation 56'57. The in situ perfused brain preparation combines the advantages of mechanical stability, and maintenance of intact structure and integrative function. We envision this preparation as serving as a bridge between studies of cellular properties in reduced preparations (e.g. the brain slice) and investigation of network function of integrative systems in whole animals. The cellular properties of neurons within the dorsal respiratory group 14-17 and ventral respiratory group 38 of guinea pigs have been studied using an in vitro brainstem slice preparation. Stable intracellular recording in a perfused preparation during the expression of the respiratory rhythm should allow a direct assessment of how these cellular properties interact with synaptic drive

17 to shape the respiratory rhythm. A similar approach should be applicable to electrophysiology of other integrated neural systems as well as to pharmacology and biochemistry.

REFERENCES 1 Andjus, R., Suhara, K. and Sloviter, H.A., An isolated, perfused rat brain preparation, its spontaneous and stimulated activity, J. Appl. Physiol., 22 (1967) 1033-1039. 2 Andry, D.K. and Horita, A., Thyrotropin-releasing hormone: physiological concomitants of behavioral excitation, Pharmacol. Biochem. Behav., 6 (1977) 55-59. 3 Bartlet, D., Jr., Origin and regulation of spontaneous deep breaths, Respir. PhysioL, 12 (1971) 230-238. 4 Bendixen, H.H., Smith, G.M. and Mead, J., Pattern of ventilationin young adults, J. Appl. Physiol., 19 (1964) 195-198. 5 Berliner, S., III., Application of ultrasonic processors, American Biotechnology Laboratory, March, 1984. 6 Brown-S6quard, E., J. Physiol. Homme Anim. (Paris), 1 (1858) 117. [Ref. in Pike, EH., Guthrie, C.C., and Stewart, G.N., J. Exp. Med., 10 (1908) 371-418. 7 Cherniack, N.S., von Euler, C., Glogowska, M. and Homma, I., Characteristics and rate of occurrence of spontaneous and provoked augmented breaths, Acta Physiol. Scand., 111 (1981) 349-360. 8 Clark, EJ. and von Euler, C., On the regulation of depth and rate of breathing, J. Physiol. (Lond.), 222 (1972) 267-295. 9 Clark, L.C. Jr., Kaplan, S., Becattini, E and Benzing, G., Perfusion of whole animals with perfluorinated liquid emulsions using the Clark bubble-defoam heart-lung machine, Fed. Proc., 29 (1970) 1764-1770. 10 Cohen, M.I., Phrenic and recurrent laryngeal discharge patterns and the Hering-Breuer reflex, Am. J. Physiol., 228 (1975) 1489-1496. 11 Cohn, R., Convulsive activity in gerbils subjected to cerebral ischemia, Exp. Neurol., 65 (1979) 391-397. 12 Courtin, R.F., Beckford, R.G. and Faulconer, A. Jr., The classification and significance of electro-encephalographic patterns produced by nitrous oxide-ether anesthesia during surgical operations, Proc. Staff Meet. Mayo Clinic, 25 (1950) 197. 13 Crosfill, M.L. and Widdicombe, J.G., Physical characteristics of the chest and lungs and the work of breathing in different mammalian species, J. Physiol. (Lond.), 158 (1961) 1-14. 14 Dekin, M.S. and Getting, P.A., Firing pattern of neurons in the nucleus tractus solitarius: modulation by membrane hyperpolarization, Brain Research, 324 (1984) 180-184. 15 Dekin, M.S. and Getting, P.A., In vitro characterization of neurons in the ventral part of the nucleus tractus solitarius. II. Ionic basis for repetitive firing patterns, J. Neurophysiol., 58 (1987) 215-229. 16 Dekin, M.S., Getting, P.A. and Johnson, S.M., In vitro characterization of neurons in the ventral part of the nucleus tractus solitarius. I. Identification of neuronal types and repetitive firing properties, J. Neurophysiol., 58 (1987) 195-214. 17 Dekin, M.S., Richerson, G.B. and Getting, P.A., Thyrotropinreleasing hormone induces rhythmic bursting in neurons of the nucleus tractus solitarius, Science, 229 (1985) 67-69. 18 Deutschmann, W., Wellh6ner, H.H. and Erdmann, G., Perfusion of the cervical spinal cord in situ of adult rats using a perfluorocarbon emulsion, Brain Research, 280 (1983) 239-249. 19 Dirks, B., Krieglstein, J., Lind, H.H., Rieger, H. and Sch0tz, H., Fluorocarbon perfusion medium applied to the isolated rat brain, J. Pharmacol. Methods, 4 (1980) 95-108. 20 Dolivo, M. and Rouiller, C.H., Changes in ultrastructure and synaptic transmission in the sympathetic ganglion during various metabolic conditions, Prog. Brain Res., 31 (1969) 111-123. 21 Dolivo, M., Metabolism of mammalian sympathetic ganglia,

Acknowledgements. We thank Drs. Michael O'Donovan, Michael Dekin, Corey Cleland, Michael Jennings and Jack Feldman for their helpful advice and criticisms. This work was supported by NIH Grants NS15350 and HL32336, and constituted a portion of a Ph.D. thesis (G.B.R.).

Fed. Proc., 33 (1974) 1043-1048. 22 Drewes, L.R., An improved apparatus for blood perfusion of the canine cerebral vasculature, Neurochem. Res., 5 (1980) 551-560. 23 Feldman, J.L., Neurophysiology of breathing in mammals. In F.E. Bloom (Ed.), Handbook of Physiology, Section: The Nervous System IE, Volume: Intrinsic Regulatory Systems of the Brain, APS, 1986, pp. 463-524. 24 Frankenhauser, B., Ischaemic paralysis of a uniform nerve, Acta Physiol. Scand., 18 (1949) 75-98. 25 Fulton, B. and Walton, K., Electrophysiological properties of neonatal rat motoneurones studied in vitro, J. Physiol. (Lond.), 370 (1986) 651-678. 26 Gibson, G.E., Hypoxia. In D.W. McCandless (Ed.), Cerebral Energy Metabolism and Metabolic Encephalopathy, Plenum, New York, 1985, pp. 43-78. 27 Gilboe, D.D., Cotanch, W.W. and Glover, M.B., Isolation and mechanical maintenance of the dog brain, Nature (Lond.), 206 (1965) 94-96. 28 Glogowska, M., Richardson, P.S., Widdicombe, J.G. and Winning, A.J., The role of the vagus nerves, peripheral chemoreceptors and other afferent pathways in the genesis of augmented breaths in cats and rabbits, Respir. Physiol., 16 (1972) 179-196. 29 Green Cross Corp., FC-43 Emulsion, Technical Information Ser. No. 3, Osaka, Japan, September 4, 1976. 30 Grunstein, M., Younes, M. and Milic-Emili, J., Control of tidal volume and respiratory frequency in anesthetized cats, J. Appl. Physiol., 35 (1973) 463-476. 31 Haldane, J.S., Meakins, J.R. and Priestley, J.G., The effects of shallow breathing, J. Physiol. (Lond.), 52 (1919) 433-453. 32 Hansen, A.J., Hounsgaard, J. and Jahnsen, H., Anoxia increases potassium conductance in hippocampal nerve cells, Acta Physiol. Scand., 115 (1982) 301-310. 33 Hedner, J., Hedner, R., Jonason, J. and Lundberg, D., Central respiratory stimulant effect by thyrotropin releasing hormone in the rat, Neurosci. Lett., 25 (1981) 317-320. 34 Hirsch, H., Euler, K.H. and Schneider, M., Ober die Erholung und Wiederbelebung des Gehirns nach Ischamie bei Normothermie, Pflagers Arch., 265 (1957) 281-313. 35 Hokfelt, T., Fuxe, K., Johansson, O., Jeffcoate, S. and White, N., Thyrotropin releasing hormone (TRH)-containing nerve terminals in certain brain stem nuclei and in the spinal cord, Neurosci. Lett., 1 (1975) 133-139. 36 H611er, M., Breuer, H. and Fleischhauer, K., An improved method for perfusion of the isolated brain of the rat - - influence of perfusion conditions and application of analeptic and anticonvulsant drugs, J. Pharmacol. Methods, 9 (1983) 19-32. 37 Hug, C.C., Jr., Monitoring. In R.D. Miller (Ed.), Anesthesia, Volume I, 2nd edn., Churchill Livingstone, New York, 1986, pp. 411-464. 38 Johnson, S.M., Electrophysiological and Anatomical Character-

istics of Neurons within the Nucleus Ambiguus of Adult Guinea Pigs, Ph.D. Thesis, University of Iowa, 1989. 39 Keele, C.A., Neii, E. and Joels, N., Samson Wright's Applied Physiology, 13th edn., Oxford Medical Publications, New York, 1982. 40 Kiersey, D.K., Bickford, R.G. and Faulconer, A. Jr., Electroencephalo-graphic patterns produced by thiopental sodium during surgical operations: description and classification, Br. J. Anesth., 23 (1951) 141-152. 41 Knowlton, G.C. and Larrabee, M.G., A unitary analysis of pulmonary volume receptors, Am. J. Physiol., 147 (1946)

18 100-114. 42 Krieglstein, G., Krieglstein, J. and Stock, R., Suitability of the isolated perfused rat brain for studying effects on cerebral metabolism, Naunyn-Schmiedebergs Arch. Pharmacol., 275 (1972) 124-134. 43 Krieglstein, G., Krieglstein, J. and Urban, W., Long survival time of an isolated perfused rat brain, J. Neurochem., 19 (1972) 885-886. 44 Krieglstein, J. and Stock, R., The isolated perfused rat brain as a model for studying drugs acting on the CNS, Psychopharmacologia, 35 (1974) 169-177. 45 Lawson, E.E. and Thach, B.T., Respiratory patterns during progressive asphyxia in newborn rabbits, J. Appl. Physiol., 43(3) (1977) 468-474. 46 Ljunggren, B., Ratcheson, R.A. and Siesjo, B.K., Cerebral metabolic state following complete compression ischemia, Brain Research, 73 (1974) 291-307. 47 Llintis, R., Yarom, Y. and Sugimori, M., Isolated mammalian brain in vitro: new technique for analysis of electrical activity of neuronal circuit function, Fed. Proc., 40(8) (1981) 2240-2245. 48 Llinhs, R. and Muhlethaler, M., An electrophysiological study of the in vitro, perfused brain stem-cerebellum of adult guineapig, J. Physiol. (Lond.), 404 (1988) 215-240. 49 Llin~s, R. and Muhlethaler, M., Eiectrophysiology of guinea-pig cerebellar nuclear cells in the in vitro brain stem-cerebellar preparation, J. Physiol. (Lond.), 404 (1988) 241-258. 50 Lumsden, T., Observations on the respiratory centres in the cat, J. Physiol. (Lond.), 57 (1923) 153-160. 51 Lumsden, T., Observations on the respiratory centres, J. Physiol. (Lond.), 57 (1923) 354-367. 52 Marshall, L.H. and Hanna, C.H., Direct measurement of arterial blood pressure in the guinea pig', Proc. Soc. Exp. Biol. Med., 92 (1956) 31-32. 53 Morgan, R.J. and Turbes, C.C., Experimental studies of ongoing power spectral analysis of normal and abnormal EEGs under anesthesia, Biomed. Sci. Instrum., 11 (1975) 85-88. 54 Otsuka, M. and Konishi, S., Electrophysiology of mammalian spinal cord in vitro, Nature (Lond.), 252 (1974) 733-734. 55 PulsineUi, W.A., Brierley, J.B. and Plum, F., Temporal profile of neuronal damage in a model of transient forebrain ischemia, Ann. Neurol., 11 (1982) 491-498. 56 Richerson, G.B., The Respiratory Centers of the Guinea Pig: In Vivo and Perfused Brain Studies, Ph.D. thesis, University of Iowa, 1987. 57 Richerson, G.B. and Getting, P.A., Maintenance of complex neural function during perfusion of the mammalian brain, Brain Research, 409 (1987) 128-132. 58 Roberts, A. and Roberts, B.L., Neural Origin of Rhythmic Movements, Cambridge Univ. Press, Cambridge, 1983. 59 Rossen, R., Kabat, H. and Anderson, J.P., Acute arrest of cerebral circulation in man, Arch. Neurol. Psychiatr., 50 (1943) 510-528.

60 St. John, W.M., Bledsoe, T.A. and Sokol, H.W., Identification of medullary loci critical for neurogenesis of gasping, J. Appl. Physiol., 56 (1984) 1008-1019. 61 St. John, W.M. and Knuth, K.V., A characterization of the respiratory pattern of gasping, J. Appl. Physiol., 50 (1981) 984-998. 62 Shapovalov, A.I., Shiriaev, B.I. and Tamarova, Z.A., A study of neuronal activity of mammalian superfused or intra-arterially perfused CNS preparations. In G.A. Kerkut and H.V. Wheal (Eds.), Electrophysiology of Isolated Mammalian CNS Preparations, Academic Press, New York, 1981, pp. 367-394. 63 Sloviter, H.A. and Kamimoto, T., Erythrocyte substitute for perfusion of brain, Nature (Lond.), 216 (1967) 458-460. 64 Smith, J.C. and Feldman, J.L., In vitro brainstem-spinal cord preparation for study of motor systems for mammalian respiration and locomotion, J. Neurosci. Methods, 21 (1987) 321-333. 65 Spitzer, A. and Edelmann, C.M., Jr., Maturational changes in pressure gradients for glomerular filtration, Am. J. Physiol., 221(5) (1971) 1431-1435. 66 Stewart, P.A., How to Understand Acid-Base. A Quantitative Acid-Base Primer for Biology and Medicine, Elsevier, New York, 1981. 67 Suzue, T., Respiratory rhythm generation in the in vitro brain stem-spinal cord preparation of the neonatal rat, J. Physiol. (Lond.), 354 (1984) 173-183. 68 Swaab, D.E and Boer, K., The presence of biologically labile compounds during ischemia and their relationship to the EEG in rat cerebral cortex and hypothalamus, J. Neurochem., 19 (1972) 2843-2853. 69 Thompson, A.M., Robertson, R.C. and Bauer, T.A., A rat head-perfusion technique developed for the study of brain uptake of materials, J. Appl. Physiol., 24 (1968) 407-411. 70 Trenchard, D., Role of pulmonary stretch receptors during breathing in rabbits, cats and dogs, Respir. Physiol., 29 (1977) 231-246. 71 Turbes, C.C., Morgan, R.J., Schneider, G.J. and Simard, J.M., Some physiological aspects of analysis of frequency spectra of the EEG, Biomed. Sci. lnstrum., 12 (1976) 57-62. 72 White, R.J., Albin, M.S. and Verdura, J., Preservation of viability in the isolated monkey brain utilizing a mechanical extracorporeal circulation, Nature (Lond.), 202 (1964) 10821083. 73 Woods, H.F., Graham, C.W., Green, A.R., Youdim, M.B.G., Grahame-Smith, D.G. and Hughes, J.T., Some histological and metabolic properties of an isolated perfused rat brain preparation with special reference to monoamine metabolism, Neuroscience, 1 (1976) 313-323. 74 Zlokovic, B.V., Lipovac, M.N., Begley, D.J., Davson, H. and Rakic, L., Slow penetration of thyrotropin-releasing hormone across the blood-brain barrier of an in situ perfused guinea pig brain, J. Neurochem., 51 (1988) 252-257.