tissue po and the effects of hypoxia on the … · 2 at various tissue depths in the isolated mouse...

TISSUE PO2 AND THE EFFECTS OF HYPOXIA ON THE GENERATIONOF LOCOMOTOR-LIKE ACTIVITY IN THE IN VITRO SPINAL CORD OFTHE NEONATAL MOUSE

R. J. A. WILSON,a T. CHERSAb AND P. J. WHELANb*aRespiratory Research Group, Department of Physiology and Bio-physics, University of Calgary, Calgary, Alberta T2N 4N1, CanadabNeuroscience Research Group, Department of Physiology and Bio-physics, University of Calgary, Calgary, Alberta T2N 4N1, Canada

Abstract—The neonatal mouse en bloc spinal cord-brainstempreparation used in combination with advances in mousegenomics provides a novel strategy for studying the spinalcontrol of locomotion. How well the mouse en bloc prepara-tion is oxygenated however, is unknown. This is an importantconsideration given that (a) other superfused mammalian enbloc preparations have anoxic cores and (b) hypoxia canhave profound effects on neuronal activity. Here we measurethe level of tissue oxygenation in the mouse preparation anddetermine how neuronal activity within the spinal cord isinfluenced by poor superfusion and/or low oxygen.

To measure tissue oxygenation, oxygen depth profileswere obtained (P0–1 and P2–3; Swiss Webster mice). AtP0–1, spinal cords were oxygenated throughout under rest-ing conditions. When fictive locomotor activity was evoked(5-HT 10 �M, dopamine 50 �M, NMA 5 �M), there was asubstantial reduction in tissue PO2 starting within 5 min ofdrug application. Following washout, the PO2 slowly returnedto control levels over a period of 30 min. The experimentsdescribed above were repeated using P2–3 preparations. Inthis older age group, the spinal cord preparations had ahypoxic/anoxic core that was exacerbated during metaboli-cally demanding tasks such as drug-evoked rhythmic activ-ity.

To examine how an anoxic core affects neuronal activitywithin the spinal cord we either altered the flow-rate or ma-nipulated superfusate PO2. When the flow-rate was reduced atransient disruption in the rhythmicity of drug-induced loco-motion occurred during the first 15 min (P0–1 preparations).However, the motor output adapted and stabilized. Duringprolonged superfusion with hypoxic artificial cerebrospinalfluid on the other hand, both the motor bursts in spinalnerves and the activity of most neurons near the center of thetissue were abolished.

Overall, this study suggests that while oxygenation ofP0–P1 preparations is adequate for studies of locomotorfunction, oxygenation of older preparations is more problem-atic. Our data also show that neonatal spinal neurons requireoxygen to maintain activity; and the spinal locomotor rhythmgenerator continues to function providing the peripheral tis-sue of the cord is oxygenated. Together, these results areconsistent with the results of a previous study which suggestthat the locomotor pattern generator is located close to the

surface of the spinal cord. © 2003 IBRO. Published byElsevier Science Ltd. All rights reserved.

Key words: pattern generation, CPG, anoxia, locomotion,respiration.

Mammalian superfused en bloc preparations have severaladvantages over other types of preparations for the study oflocomotor and respiratory function (Kerkut and Bagust, 1995;Kiehn and Kjaerulff, 1998). Some of the advantages includebetter access to the tissue, mechanical stability, and com-plete deafferentation. However, the adequacy of diffusion ofgases to and from the tissue surface in meeting metabolicdemand for oxygen of mammalian en bloc preparations hasbeen questioned. For example, superfused neonatal ratbrainstem and spinal cord preparations, while capable ofproducing rhythmic motor outputs, have anoxic cores (i.e.tissue PO2 of 0 Torr) which result in severe metabolic acido-sis (Okada et al., 1993; Brockhaus et al., 1993). The detri-mental effect of anoxia on local neuronal activity has beendocumented extensively (Luhmann et al., 1993; Czeh andSomjen, 1990; Ataka et al., 1996; Richter and Ballanyi, 1996;Krnjevic, 1999) and includes elevated extracellular K�, hy-perpolarization, disappearance of excitatory synaptic po-tentials, rise in intracellular calcium, fall in adenosinetriphosphate (ATP), and extracellular accumulation ofadenosine. The combined effect results in abnormal neu-ronal firing followed by a general cessation of activity.

Recently, the isolated spinal cord preparation of themouse has emerged as an attractive preparation to studythe neuronal control of locomotion (Whelan et al., 2000;Bonnot et al., 1998; Nishimaru et al., 2000; Jiang et al.,1999). There are two main reasons for using the mouse,firstly many genetically altered mouse models are avail-able (Erickson et al., 1996; Jacquin et al., 1999; Bou-Flores et al., 2000; Burnet et al., 2001; Cazalets et al.,2000) and secondly, calcium-imaging techniques havebeen successfully used in the en bloc preparation to re-solve neuronal activity (Bonnot et al., 2002). An additionalpossible advantage of the mouse preparation is that, dueto its small size, the isolated spinal cord may be less proneto hypoxia (PO2�20 Torr; Vovenko, 1999) and anoxia invitro. In the first part of this paper we test this assumptiondirectly, by measuring PO2 at various tissue depths in theisolated mouse spinal cord, before and after the onset ofdrug-induced locomotion. To the best of our knowledgethis is the first time that the effects of locomotor activity onPO2 have been examined in any in vitro preparation. In thesecond part of the paper we examine the functional con-

*Corresponding author. Tel: �1-403-220-4210; fax: �1-403-283-2700.E-mail address: [email protected] (P. J. Whelan).Abbreviations: ACSF, artificial cerebrospinal fluid; ANOVA, analysis ofvariance; ATP, adenosine triphosphate; CPG, central patterngenerator; NMA, N-methyl-DL-aspartate.

Neuroscience 117 (2003) 183–196

0306-4522/03$30.00�0.00 © 2003 IBRO. Published by Elsevier Science Ltd. All rights reserved.doi:10.1016/S0306-4522(02)00831-X

183

sequences of altering tissue PO2 on the performance ofthe mouse spinal locomotor network. We demonstrate thatin general, activity of neonatal neurons is oxygen depen-dent and that locomotor activity persists in the presence ofan anoxic core. Collectively, the findings suggest that im-portant elements of the neonatal locomotor pattern gener-ator are located in the outer shell of the spinal cord. Aportion of these results has been published in abstractform (Whelan and Wilson, 2001).

EXPERIMENTAL PROCEDURES

Preparations

Experiments were performed on Swiss Webster mice (CharlesRiver Laboratory), at two ages, P0–1 and P2–3. A total of 60 micewere used in this investigation. The protocol was approved by theUniversity of Calgary Animal Resource Centre. After induction ofanesthesia with methoxyflurane, the animals were rapidly decer-ebrated and eviscerated. All efforts were made to minimize thenumber of animals. The remaining tissue was placed in a dissect-ing chamber containing artificial cerebrospinal fluid (ACSF) (con-centrations in mM: 128 NaCl, 4 KCl, 1.5 CaCl2, 1 MgSO4, 0.5NaH2PO4, 21 NaHCO3, 30 D-Glucose), bubbled with 95% O2 and5% CO2 (pH�7.4) and maintained at room temperature. Theanimals were then pinned down onto a Sylgard base, and aventral laminectomy was performed. The dorsal and ventral rootswere cut and the spinal cord was isolated. An isolated spinal cordfrom T5-cauda equina was prepared. Following these procedures,the preparation was transferred to a recording chamber. Theseprocedures are described in greater detail elsewhere (Whelan etal., 2000).

Recording chamber

The recording chamber (volume: 2 ml) was custom built andbased on previously published specifications (Torgerson et al.,1997). The inflow was located at the bottom of the chamber andthe outflow at the top. The preparation was secured on a stainlesssteel or nylon mesh (using 7.0 silk suture thread looped thoroughthe mesh) between the inlet and outlet, thereby maximizing flowover the surface of the preparation. The ACSF in the chamber wasfed by gravity from a flask in which it was equilibrated with 5% CO2

and 95% O2 and re-circulated (total volume: 400 ml3). Flow ratewas maintained at 20–24 ml/min (see Results), unless otherwisestated. Once the preparation was in the chamber the ACSF wasgradually heated to 27 °C over the course of 30 min.

PO2 measurements

Measurement of tissue PO2 was performed with polarographicClarke style PO2 microelectrodes (tip size: 10 �m, Diamond Gen-eral Corporation, Ann Arbor, MI, USA) connected to a polaro-graphic amplifier (model 1900, A-M systems). Electrodes werecalibrated before and after each experiment, by immersing the tipsinto beakers containing ACSF (maintained at 27 °C) bubbled vig-orously for at least 30 min with either 100% nitrogen or 100%oxygen.

PO2 depth profiles

Depth profiles were performed to determine whether the tissuewas oxygenated throughout the preparation and to establish tis-sue PO2 at different depths (Wilson et al., 1999). For each profile,an electrode was placed 300 �m above the ventral surface of thespinal cord (L1–L2 segment) and advanced in 50 �m stepsthrough the tissue using a motor driven stepping apparatus (Nano-Stepper, model type B). In most cases, when the tip of the elec-

trode reaching the surface of the tissue, a slight dimpling wasobserved. In cases when the optics were inadequate to see dim-pling, the surface was detected by moving the electrode back andforth laterally (�50 �m) and observing a slight distortion of thesurrounding tissue. The electrode was slightly offset from thesagittal plane. After the tip of the electrodes had reached a tissuedepth where the PO2 began to increase (reversal point) it wasremoved in 50 �m steps until the surface of the spinal cord wasreached.

Effects of flow

In experiments to determine the effect of flow rate of the superfu-sate (either on tissue PO2 or on locomotor-like activity), flow wasadjusted by changing the height of the flask containing the ACSF.In experiments to determine the effects on oxygenation, tissuePO2 was measured 10 min after changes in flow rate. In experi-ments designed to investigate the effects of flow on locomotoractivity, the locomotor rhythm was initially established at 22 ml/minand then the flow was changed in steps to 6, 2, and back to 22ml/min. The rhythm was allowed to stabilize at the new flow ratefor 20–25 min. The flow rate was calibrated by measuring theinflow before and after each experiment.

Effect of spinal network activity

To determine how locomotor network activity affects tissue PO2 ofthe spinal cord we employed a protocol that began with a depthprofile under control conditions. The tip of the oxygen electrodewas advanced 400 �m into the L1–2 lumbar tissue and the effectsof spontaneous or dorsal root stimulus-induced activity on PO2

were monitored. The L1–2 region was selected since it has beensuggested that this region is important to locomotor rhythm gen-eration (Cazalets et al., 1995; Kjaerulff and Kiehn, 1996; Cowleyand Schmidt, 1997; Kremer and Lev-Tov, 1997). Locomotor-likeactivity was then evoked by bath application of a drug combinationcontaining: serotonin (5-HT), N-methyl-DL-aspartate (NMA), anddopamine. Bath concentrations were 10 �M, 5 �M and 50 �Mrespectively. Alternating motor activity was usually evoked within5–10 min of application. Once stable alternating activity had beenobtained (15–20 min), the tip of the oxygen electrode was re-moved from the tissue and an additional depth profile was per-formed. Following this procedure the tip of the electrode was againadvanced 400 �m into the tissue, the drugs were washed out (for20–30 min) and a final “wash-out” depth profile was completed.

Electrophysiological recordings and stimulation

Electrical activity from populations of motoneurons was recordedusing tight fitting plastic suction electrodes into which L1 or L2ventral roots were drawn. The resultant neurograms were ampli-fied (20,000 times), band-pass filtered (100–1 kHz, Dagan EX1Amplifiers), and digitized (sampling rate: 4 kHz, Digidata 1320,Axon Instruments, Foster City, CA, USA). In experiments involvingstimulation-evoked activity, alternating spinal ventral root dis-charge was evoked by electrical stimulation (300 �sec stimuluspulses, 2–10 V in amplitude, in 4 Hz trains lasting 10–20 sec) ofthe lumbar dorsal roots using tight fitting suction electrodes (sim-ilar to those used for recording).

Extracellular recording from single units were obtained usinga differential amplifier (Dagan EX1; low and high cutoffs were30–100 Hz and 0.5–1 kHz, respectively) and blunt glass micro-electrodes filled with 150 mM NaCl or 3 M KCl.

Data analysis

Data analyses were performed offline using PClamp8 software(Axon Instruments). Rectification and smoothing of dischargesrecorded from ventral roots of the spinal cord were performed

R. J. A. Wilson et al. / Neuroscience 117 (2003) 183–196184

digitally (10 Hz low pass Bessel filter). Extracellular spike fre-quency was calculated by using MatLab software (MathWorks).Data are presented as means with standard errors. The Student’st-test was used for single pairwise comparisons. Analysis of vari-ance (ANOVA) was used for analysis of multiple conditions. Fol-lowing the ANOVA, the Tukey method was used to test for post-hoc significant differences between conditions. For the series ofexperiments that examined the effects of oxygen and flow-rate onthe rhythm, cross and auto correlograms were performed using astatistics software package (STATISTICA).

Histology

Three P0 and three P3 preparations were fixed using 4% parafor-maldehyde for subsequent measurement of the diameter of thespinal cords. The spinal cords were paraffin embedded in wax, cutinto 8 �m sections, and stained with Cresyl Violet. The diameters ofthe spinal cords at the L2 segment were measured using ImageProsoftware and the means and standard deviations calculated.

RESULTS

We examined tissue oxygenation of the isolated in vitromouse spinal cord during activation of neural networks thatproduce locomotor-like motor output. Using ventral rootrecordings and single unit extracellular recordings, we thenexamined the functional consequences of manipulatingtissue PO2 on neuronal activity within the spinal cord.

Effects of flow rate on spinal cord PO2

We performed a set of experiments to determine the opti-mal flow rate to use for tissue oxygenation of the spinal

cord. Using Clark-style polarographic electrodes we mea-suring tissue PO2 at different flow rates. The electrode wasplaced 400 �m below the ventral surface at the level of theL1–2 lumbar segment (Fig. 1A), which was approximatelyin the center of the preparation (Fig. 1B). Despite the useof a dish that maximizes the flow over the surface of thepreparation, we found that flow rates less than 15 mL/minproduced a marked drop in tissue oxygenation. Oxygen-ation at 400 �m decreased from an average PO2 of around200 Torr at 15 ml/min to 80 Torr at 2 ml/min (Fig. 1). At 22ml/min tissue PO2 increased only slightly over that at 15ml/min (to 230 Torr) suggesting that a flow rate of 22ml/min was close to optimal. We decided against usingflows greater than 22 ml/min because mechanical stabilityof the preparation at very high flow rates would exclude itsuse for most electrophysiological applications. This flowrate was used to obtain the results described below unlessotherwise stated. In three P0–1 and three P2–3 prepara-tions we measured the dimensions of the spinal cord fol-lowing histological sectioning. These results are displayedin Fig. 1B (P0: 867�49 �m; P3: 1014�46 �m) and arecorrected for the approximate tissue shrinkage (20%) thatoccurs during the fixing and staining procedure.

Effects of spontaneous and induced locomotor-likeactivity on spinal cord PO2

Occasional spontaneous locomotor-like activity, consistingof alternating bursts of activity in left and right L2 roots,occurred spontaneously in most but not all preparations

Fig. 1. A. Graph plots the mean PO2 in P0–1 mice (n�4) as a function of superfusate flow rate. All PO2 measurements were made with themicroelectrode 400 �m below the surface of the spinal cord. The error bars represent S.E.M. B. Transverse 8 �m thick sections (fixed in 4%paraformaldehyde and stained with Cresyl Violet) of a P0 (left) and a P3 (right) L2 spinal cord segment. The mean length (n�three preparations)�S.D.from the dorsal to ventral surface is illustrated.

R. J. A. Wilson et al. / Neuroscience 117 (2003) 183–196 185

(Whelan et al., 2000; Bonnot et al., 1998). We examinedthe modulation of tissue PO2 at 400 �m below the ventralsurface of the spinal cord during spontaneous activity andduring two other rhythmogenic conditions: following elec-trical stimulation of the dorsal roots and following bathapplication of drugs (Whelan et al., 2000). Two age groupsof animals were studied (P0–1 and P2–3).

During periods of spontaneous activity, tissue PO2

fluctuated in most cases, decreasing modestly shortly afterthe onset of each bout of spiking activity (20–40 Torr; Fig.2A). Spontaneous activity was highly variable betweenanimals (Whelan et al., 2000; Bonnot et al., 1998), andthus, the effects on tissue PO2 were difficult to quantify.However, in no case did spontaneous activity cause tissueanoxia.

Electrical stimulation of the L5 dorsal root (trains: 4 Hz,10–40 s, Fig. 2B, C) resulted in robust tonic burstingactivity in animals of both age groups and was often fol-lowed by rhythmic activity (Whelan et al., 2000). Concom-itant with the increase in activity was a decrease in tissuePO2 (Fig. 2B). Although this decrease was larger than thatobserved during spontaneous bursting, in only one of thenine preparations studied (both age groups) did dorsal rootstimulation-induced bursting lead to transient anoxia. Onaverage, in P0–1 preparations, tissue PO2 fell followingelectrical stimulation from 328�99 to 287�111 Torr (n�4).In the P2–3 preparations tissue PO2 fell from 193�94 to135�80 Torr (n�4). The minimum PO2 in these prepara-tions (P2–3) occurred at 42�9 s following the onset ofdorsal root stimulation and, following the offset of the stim-

Fig. 2. Modulation of tissue PO2 with bursting activity in a P3 mouse. (A) Rectified and smoothed suction electrode recordings from the left and rightL2 ventral roots and tissue PO2 during spontaneous network activity. (B and C) Stimulation of the L5 dorsal root induced rhythmic activity incontralateral ventral roots (L2) and a fall in tissue PO2.


ulus, the PO2 returned to prestimulus conditions in160�35 s (n�4).

The largest decrease in PO2 occurred when rhythmicalternating activity was evoked using bath-applied drugs(10 �M 5-HT; 5 �M NMA; 50 �M dopamine). Followingbath-application of the drugs there was an increase in tonicbursting activity that led to alternating rhythmic activity inthe contralateral L2 ventral roots (Fig. 3). This pattern wasobserved in all animals tested (both age groups). Followingaddition of the drugs tissue PO2 fell rapidly, reaching aplateau within 5–6 min (Fig. 3A). In P0–1 preparations,tissue PO2 at 400 �m below the ventral surface fell from303�76 to 89�70 Torr. In the P2–3 preparations, tissuePO2 fell from 192�127 to 61�62 Torr. Although alternat-ing rhythmic activity was recorded from contralateral ven-tral roots, no correlation between the PO2 and ventral rootbursts was observed (Fig. 3B).

Depth profiles of PO2 within the L1-2 lumbar spinalcord

The results presented thus far have compared the effectsof different metabolic challenges on PO2 at a constantdepth of 400 �m. In this section we quantified tissue PO2

at 50 �m depth intervals to ascertain the change withdepth and minimum levels.

The PO2 was measured starting at 300 �m above thetissue surface and the tip of the electrode was advancedthrough the tissue until a reversal in the PO2 gradient wasreached (reversal point). Separate depth profiles werecompleted during rest, drug-evoked rhythmicity and follow-ing washout in each animal. Figure 4 summarizes theresults from the two age groups. In all conditions and agegroups, the PO2 decreased as the surface of the tissuewas approached, suggesting an unstirred layer of super-fusate surrounding the tissue. These PO2 gradientsaround the tissue were small [14 Torr/100 �m (P0–1) and11 Torr/100 �m (P2–3)] demonstrating efficient superfu-sion at the high flow rate used.

The spinal cords of P0–1 preparations were oxygenatedthroughout (five of five preparations) under control conditionshaving a minimum tissue PO2 of 181�60 Torr (Fig. 4A). Infour of five preparations the reversal point occurred at 500–550 �m below the surface; in the remaining preparation thereversal point occurred at 350 �m. When depth profiles wererepeated 20–30 min after the application of rhythmogenicdrugs (sufficient time for the motor pattern to stabilize), theminimum tissue PO2 dropped to 65�46 Torr (P�0.05; Fig.4B) and the slope of the PO2 gradient 0–200 �m below theventral surface increased (PO2 gradient was 53 Torr/100 �mand 145 Torr/100 �m for control and rhythmic conditions

Fig. 3. Application of 5 �M NMA, 10 �M 5-HT and 50 �M dopamine reduces the tissue PO2. (A) Rectified and smoothed traces showing L2 ventralroot neurogram recordings along with tissue PO2 at a depth of 400 �m below the ventral surface of the spinal cord. The left and right arrows indicatethe times when the drugs were added to the bath and washed out, respectively. Note the increase in firing which was accompanied by a substantialreduction in tissue PO2. (B) Expansion of the boxed region in (A) illustrating the rhythmic alternation and lack of cycle by cycle modulation of tissuePO2.


respectively; P�0.05). Two out of five preparations devel-oped an anoxic core under these conditions extending towithin 425�75 mm of the ventral surface. Following washoutof the drugs (Fig. 4C) the minimum tissue PO2 (176�48 Torr;

P�0.1) and PO2 gradient (85 Torr/100 �m; P�0.1) werestatistically similar to control conditions. All of the prepara-tions were oxygenated throughout upon washout of thedrugs.

Fig. 4. Depth profiles of tissue PO2 in the L2 region of the ventral spinal cord of P0–1 (A–C) and P2–3 mice (D–F). Three separate profiles wereobtained in each preparation [before drugs (A and D), 20 min following addition of NMA, 5-HT and dopamine (B and E) and following washout (C andF). Each filled circle represents the mean from 4 to 5 preparations. The error bars represent the S.E.M.


In the P2–3 group of animals, under control conditions(Fig. 4D), the minimum tissue PO2 was 45�35 Torr. Oneof the four preparations had an anoxic core extending towithin 500 �m of the ventral surface. The PO2 gradientwas 50 Torr/100 �m (0–200 �m below the surface). Allfour preparations produced rhythmic alternating motor pat-terns in contralateral L2 ventral roots following applicationof the rhythmogenic drugs (Fig. 4E). One preparation re-mained oxygenated throughout in the presence of thedrugs, while the rest had anoxic cores (minimum tissuePO2 of the four preparations: 13�14 Torr) which extendedto within 350�57 �m of the ventral surface. The rhythmo-genic drugs increased the PO2 gradient to 194 Torr/100�m (0–200 �m below the surface; P�0.05). Followingwashout of the drugs (Fig. 4F) the PO2 gradients (53Torr/100 �m between 0 and 200 �m below the surface;P�0.1) and the minimum PO2 (50�50 Torr; P�0.1) weresimilar to control conditions. Two out of the four prepara-tions had anoxic cores following washout of the drugs.

Effects of superfusate flow-rate on locomotor-likeactivity

An important issue in studies of spinal cord pattern gener-ators is the repeatability of experiments from one labora-tory to the next. Flow-rate is rarely mentioned in the liter-ature, yet different flow-rates will affect tissue oxygenation(Fig. 1). In six preparations we examined the ability oflocomotor networks to adapt to different flow-rates andconsequently to changes in tissue oxygenation. In the firstset of experiments we evoked locomotor-like activity inP0–1 preparations by bath application of rhythmogenicdrugs (10 �M 5-HT, 5 �M NMA, 50 �M dopamine) at aflow-rate of 22 ml/min. We allowed the rhythm to stabilizefor 25 min following the initiation of alternating burst activ-ity. We then decreased the flow-rate to 6 ml/min then to 2ml/min and then increased it back to 22 ml/min. Flow ratewas maintained at each level for 25 min. Burst cycle peri-ods, averaged over the last 5 min of each step, did notchange significantly as a function of flow (P�0.01; n�sixpreparations). However, in 4/6 preparations, the rhythmicpattern became less regular at 2 ml/min as evidenced byan increase in the variance of the mean cycle periods[pooled mean and standard deviations cycle periods(n�four preparations): 22 ml/min, 2.97�0.16 s S.D., 2ml/min 4.41�1.14 s S.D., back to 22 ml/min: 2.92�0.25 sS.D.]. As shown in Fig. 5, at low flow rates alternatingbursts were, at times, interspersed with periods of tonicactivity and in-phase bursting. This instability in the rhythmled to a decrease in the cross-correlation coefficient com-pared with other flow rates (Fig. 5C). In the two remainingpreparations the rhythm remained quite regular as shownby the similarity in the standard deviation among condi-tions [Pooled Standard Deviations (n�two preparations):22 ml/min: 3.71�0.25; 2 ml/min: 3.24�0.23, back to 22ml/min: 3.03�0.234]. Although the rhythm recorded 25 minfollowing the change in flow from 22 ml/min to 6 ml/minwas regular, a transient disruption of the rhythm occurredduring the first 15 min (Fig. 6) with the rhythm becomingless regular (5/6 preparations). This can be appreciated in

Fig. 6 from the raw traces and the accompanying correlo-gram which shows a decrease in the correlation coeffi-cient.

Overall these data demonstrate that spinal patterngenerators appeared to be largely resistant to change inflow, providing the flow is maintained at or above 6 ml/min.However, we observed a transient disruption in the rhythmduring the first 15 min. This may have been due to neuronswithin the anoxic core that project to or belong to thecentral pattern generator (CPG) altering their firing prop-erties.

We tested the possibility that neonatal neurons in thespinal cord alter their firing properties under anoxic condi-tions by recording neuronal activity from spinal nerveswhile superfusing with hypoxic ACSF (equilibrated with5%O2, 5%CO2 bal N2). P0–1 preparations were used forthese experiments (Fig. 7A). Interestingly, rhythmogenesisin several preparations was transiently disrupted shortlyafter the change to the hypoxic superfusate, similar to thatwhich occurred when flow rate was reduced. After a shortperiod of hypoxic superfusion the rhythm recorded fromspinal nerves slowed (cycle period before hypoxia andduring the 2 min of exposure: 3.46s�0.21 and 5.01�0.56S.E., respectively; P�0.05; n�4). After prolonged expo-sure (25 min) the rhythm ceased in 2/4 preparations. In theremaining two preparations the rhythm was uncoordinatedbetween the two sides of the cord and substantially slowercompared with control conditions (Fig. 7B; 13.87s�3.16S.E.). Following return to hyperoxic superfusate, the pat-tern and the frequency of the rhythm was restored tocontrol conditions in 3/4 preparations (cycle period: 3.80s�0.67 S.E., n�three preparations; P�0.1). In the remain-ing preparation we were unable to reestablish a rhythm.

To determine the effects of hypoxia on the activity ofindividual neurons, we used extracellular glass microelec-trode to record from single units from P0–P3. Once a unitwas found and 5 min of baseline activity recorded, prepa-rations were superfused with hypoxic ACSF (5% O2, 5%CO2, 90% N; 14 units). Units that did not resume firing onreturning to control conditions could have been lost duringthe procedure, and therefore, were excluded from the dataset. Of the 12 units analyzed, 6/12 had activity correlatedto rhythm bursts in L2 roots during control conditions.During superfusion with hypoxic ACSF, 10/12 ceased tofire (e.g. Fig. 8A,B), and the firing frequency of the remain-der was reduced (From 1.47 Hz to 0.88 Hz and from4.64–3.94 Hz). All units resumed firing within 20 min fol-lowing the restoration of the gas mixture to control values(95% O2/5% CO2; e.g. Fig. 8C, D).

DISCUSSION

We have examined the oxygenation of the isolated invitro spinal cord and brainstem of the neonatal mouseunder several states of activity. Our main finding is thatat P0 –1 the spinal cord is oxygenated throughout, pro-viding high flow rates are used (Fig. 4). Spinal cordoxygenation deteriorates rapidly with age. Still, underquiescent conditions, a central hypoxic region (PO2�20


Fig. 5. Effect of changing flow rate on the locomotor rhythm (P0 preparation) recorded from L2 roots. Rhythm was evoked by bath application of 5�M NMA, 10 �M 5-HT, and 50 �M dopamine. A. The rhythm was initially evoked at a flow rate of 22 ml/min. The cross correlogram on the right showsthe high correlation cooefficient between left and right L2 ventral root bursts. The negative correlation cooefficient at 0 phase lag indicates that thebursts are out of phase (i.e. alternating). (B–D) Illustrate the rhythm 25 min following sequential changes in flow from 22 ml/min to (B) 6 ml/min (C),2 ml/min and finally (D) 22 ml/min. Note the decrease in the stability of the rhythm at 2 ml/min, illustrated by the lower cross correlation cooefficient.


Torr) is generally not present at P2–3. We also showthat spontaneous, afferent nerve-stimulated, and drug-induced locomotor activity are accompanied by substan-

tial reductions in tissue PO2. However, in the majority ofP0 –1 preparations, the increased oxygen consumptionduring fictive locomotor activity was insufficient to cause

Fig. 6. Transient disruption in the locomotor-like rhythm following a change in flow rate from 22 ml/min to 6 ml/min (P0 preparation). (Top set of traces)Control rhythm evoked by bath application of drugs (5 �M NMA, 10 �M 5-HT, 50 �M dopamine) at 22 ml/min. Graphs on the right indicates the crosscorrelograms between bursts in left and right segmental L2 ventral roots. (Middle and bottom sets of traces) Rhythm at 5 min, and 25 min followinga change in flow from 22 ml/min to 6 ml/min. Note the transient disruption in the rhythm.


tissue anoxia (0 Torr). In the second part of the paper,we demonstrate that most neonatal spinal neurons, in-cluding those involved in locomotor pattern generation,require oxygen to maintain electrical excitability. Thelocomotor pattern generator is remarkably resilient toconditions which are likely to result in a large anoxiccore, suggesting that sufficient components of the pat-

tern generator are located in the outer regions of thegray matter of the spinal cord (Figs. 5 and 6).

The in vitro mouse preparation is promising for theinvestigation of mammalian locomotor network function(Whelan et al., 2000; Jiang et al., 1999; Nishimaru et al.,2000) since it allows the use of (a) genetic models (Caza-lets et al., 2000) and (b) real-time calcium imaging tech-

Fig. 7. Effect of hypoxia on the locomotor-like network activity. (A) Changing the gas mixture to 5% O2 from 95% O2 decreased the populationmotoneuronal bursting activity recorded from segmental L2 ventral roots (activity evoked by application of 5 �M NMA, 10 �M 5-HT, 50 �M dopamine).Reapplication of 95% O2 reestablished the level of bursting activity observed in control conditions. (B) Control alternating rhythm recorded from L2ventral roots with 95% O2. Locomotor rhythm established by bath application of 5 �M NMA, 10 �M 5-HT and 50 �M dopamine. (C and D) Rhythmrecorded 6 min (C) and 30 min (D) following initiation of hypoxic condition with 5% O2. E. Restoration of alternating rhythm following a return to 95%O2 conditions.


niques for monitoring neuronal activity in large populationsof neurons (Bonnot et al., 2002). Our results suggest thatan additional utility of the neonatal mouse preparation isthat, at P0–1 it can be oxygenated throughout, making itwell suited for studying the perinatal spinal cord locomotornetwork.

In superfused preparations, oxygenation is depen-dent on the diffusion of oxygen from the tissue surface.When the dimensions of the tissue are small and/or themetabolic rate is low, superfusion is sufficient to ensurethorough oxygenation (Torgerson et al., 1997). How-ever, in the en bloc neonatal rat preparation diffusion ofoxygen from the tissue surface is insufficient to preventa large anoxia region extending from the center of thepreparation to within �600 �m of the surface. Tissuewithin this anoxic core is exposed to mild hyperkalemiaand severe metabolic acidosis (�pH 6.4) (Richter et al.,1978; Melton et al., 1991; Ballanyi et al., 1996). Incomparison, our data demonstrate that the spinal cord ofthe P0 –1 mouse preparation is oxygenated throughout.The difference between the rat and mouse data arelikely to be due to the small dimensions of the neonatal

mouse spinal cord, and our use of high flow rates (seebelow), and a recording chamber that reduces the un-stirred layer around the tissue (Torgerson et al., 1997).We consider it less likely that the metabolic rate of spinalcords differ between the neonatal rat and mouse, al-though this was not measured.

We performed a series of control experiments to ex-amine the effects of flow rate on PO2 when the electrodewas inserted 400 �m below the surface (Fig. 1). Given theaverage size of the P0 spinal cords is 865 �m (Fig. 1B), thePO2 electrode was likely located in the intermediate gray.These experiments established that flow rates below 10ml/min led to a rapid falloff in tissue PO2, and that optimalflow rates for oxygenation were around 22–23 ml/min. Wenote that this flow rate is greater than that used in most enbloc brainstem-spinal cord studies (Ballanyi et al., 1996;Smith and Feldman, 1987).

Effect of neuronal activity on tissue PO2

Another variable that affects tissue oxygenation is neuro-nal activity. In studies using in vitro preparations, several

Fig. 8. Extracellular recording of two separate units following the imposition of hypoxic conditions. Control rhythm evoked by bath application of drugs(5 �M NMA, 10 �M 5-HT, 50 �M dopamine) at 22 ml/min. (A and B) Following a change in the gas mixture from 95% O2 to 5% O2 (left arrows) theunits ceased to fire. Restoring the gas mixture to 95% O2 (right arrows) restored the firing of the units. (C and D) Same unit shown in (B), with anexpanded scale, before and 10 min after termination of hypoxia.


methods have been used to elicit a locomotor rhythm,including afferent nerve stimulation and bath application ofdrugs. However, the effects of employing these methodson tissue oxygenation are unknown. In preparations fromP0–1 animals (Figs. 3 and 4), in which the central core ofthe spinal cord was hyperoxic (i.e. PO2�90; Vovenko,1999), the addition of rhythmogenic drugs decreased theminimum tissue PO2, and increased the PO2 gradient. Inonly two out of five preparations did the drugs lead to ananoxic core. In P2–3 preparations, 3/4 of the animals hadan anoxic core following the addition of rhythmogenicdrugs (Fig. 4E). Bath application of rhythmogenic drugslikely recruits populations of neurons and possibly glia notdirectly involved in locomotor pattern generation. Thewidespread excitation of neurons could explain the largerdecrease in PO2 following application of rhythmogenicdrugs (Figs. 3 and 4) as compared with that produced byspontaneous activity or electrical stimulation (Fig. 2).These results are consistent with previous reports on theeffects of spontaneous respiratory-like activity on oxygen-ation of the in vitro medulla of the neonatal rat (Brockhauset al., 1993).

Spinal locomotor activity in preparations with anoxiccores

Our data reveal two interesting properties of the spinallocomotor circuits. Firstly, locomotor output is severelydisrupted if preparations are superfused with hypoxic su-perfusate (Fig. 7). However, coordinated activity can berecovered if control hyperoxic superfusion is returned. Thisremarkable resilience to hypoxia has been described pre-viously in other regions of the neonatal CNS.

Secondly, while neurons in the central core of the graymatter may contribute to rhythmogenesis they do not ap-pear to be essential. Three lines of evidence support thisidea. Firstly, our data show that in three of four of the P2–3spinal cord preparations, which had an anoxic core in thepresence of rhythmogenic drugs, coordinated left-rightrhythmicity was still observed. Secondly, the locomotorrhythm was only transiently affected by decreases in theflow rate (i.e. to 6 ml/min) likely to produce a moderateanoxic core in P0 preparations. Thirdly, the spike activity of10/12 units within the central core, 6/12 of which hadactivity correlated with locomotor output, was eliminatedwhen the preparation was superfused with hypoxic ACSF(Fig. 8) (Khazipov et al., 1993).

Possible explanations for rhythm generation in thepresence of an anoxic core

We propose that the most parsimonious explanation ofpersistent rhythmogenesis in the presence of an anoxiccore is that the neurons contributing to locomotor-like ac-tivity in the mouse are located close to the tissue surface.A similar argument has also been used to explain respira-tory-like rhythmogenesis of the en bloc neonatal rat prep-aration, which is thought to reside in the PreBotzingercomplex located close to the ventral surface of the brain-stem (Brockhaus et al., 1993; Okada et al., 1993). Thisexplanation is consistent with results obtained from spinal

cord lesioning experiments, which suggests that the ven-tral area of the neonatal rat spinal cord contains sufficientneurons to support coordinated left-right rhythmicity(Kjaerulff and Kiehn, 1996).

Other possibilities should also be considered. For ex-ample, we can not entirely rule out the possibility thatrhythmogenesis occurs in a population of anoxia resistantneurons located in the core region. This possibility seemsunlikely, however, given that only 2 of 12 of the units werecorded show sustained spiking under anoxic conditions.

Another possibility is that the locomotor network canreconfigure in response to loss of input from neurons lo-cated in the anoxic core. Consistent with this idea, a de-crease in the flow-rate was accompanied by a transientdisruption of the rhythm. This disruption was most evident5–15 min following the change in flow and therefore, wasunlikely to be caused by mechanical perturbations. Thefact that the rhythm stabilizes, suggests that the networkcan compensate for changes in neuronal activity that occurduring the formation of a limited anoxic core. Thesechanges could include loss of rhythm generating elementsand/or tonic inputs from neurons located within the core.There is evidence for this kind of homeostatic networkplasticity in many areas of the brain including the spinalcord (Galante et al., 2000; Feller, 1999; O’Brien et al.,1998). For example, blockade of action potentials by tetro-dotoxin increased the amplitude of �-amino-3-hydroxy-5-methyl-4-propionate receptor currents in spinal cord neu-rons in organotypic cultures (Turrigiano et al., 1998;O’Brien et al., 1998). Similarly, in the embryonic chick,blockade of excitatory or inhibitory neurotransmissionleads only to a transient blockade of spontaneous activity(Chub and O’Donovan, 1998).

Implications

In the neonatal rat literature, there has been generalagreement that the L1/2 region of the spinal cord has agreater rhythmogenic capacity than other segments of thespinal cord (Kremer and Lev-Tov, 1997; Kjaerulff andKiehn, 1996; Cowley and Schmidt, 1997; Cazalets et al.,1995). It also has been suggested that there is a rostro-caudal gradient of rhythmogenic potential as one movescaudally toward the L5 segment. However, these ideashave originated in large part from data obtained from therat in vitro preparation. A possible confounding factor inthese experiments is that the diameter of the spinal cordincreases as one moves caudally. Since an anoxic corelikely exists in the rat in vitro preparation during fictivelocomotion, the relative size of the anoxic core would beexpected to increase in a rostrocaudal fashion. Our datasuggest that this will decrease the population of neuronsthat continue bursting. As shown in Fig. 5, conditions thatexacerbate an anoxic core interfere with network function.We still favor the existence of an in vivo rhythmogenicrostrocaudal gradient in neonatal animals. However, ourstudies strongly suggest that a functional assessment ofthe influence of PO2 in the in vitro neonatal rat preparationis required to discount the possibility that an anoxic corecontributes to the reported phenomena.


CONCLUSIONS

Our data suggest that under quiescent conditions the neo-natal mouse spinal cord is well oxygenated in vitro be-tween the ages of P0–1. This represents the first demon-stration of an en bloc mammalian in vitro preparationwithout a central anoxic core region. However, duringdrug-induced locomotor-like activity, when metabolic de-mands for oxygen increase substantially, anoxic conditionscan occur in P2–3 but not generally in P0–1 animals. Aninteresting finding is that patterned motor output persistsfollowing reduction in the neuronal activity within the coreof the spinal cord. This observation is consistent with aprevious in vitro study which suggests the locomotor cen-tral pattern generator (CPG) is located ventrally (Kjaerulffand Kiehn, 1996). However, our data also indicate thatlocomotor networks adjust to changes in tissue oxygen-ation. At this time we cannot rule out the possibility that thepopulation of neurons comprising the spinal locomotorrhythm generator is dynamic and depends on the state ofthe preparation. This is a further factor that should beconsidered in the hunt for the CPG for locomotion.

Acknowledgements—This project was supported by grants fromthe University of Calgary, the Alberta Heart and Stroke Founda-tion, the CIHR, the Robertson Fund, and the Parker B FrancisFoundation.

REFERENCES

Ataka H, Murakami M, Goto S, Moriya H, Hayashi F, Fukuda Y (1996)Effects of hypoxia on the ventral root motor-evoked potential in thein vitro spinal cord preparation. Spine 21:2095–2100.

Ballanyi K, Volker A, Richter DW (1996) Functional relevance ofanaerobic metabolism in the isolated respiratory network of new-born rats. Pflugers Arch 432:741–748.

Bonnot A, Morin D, Viala D (1998) Genesis of spontaneous rhythmicmotor patterns in the lumbosacral spinal cord of neonate mouse.Brain Res Dev Brain Res 108:89–99.

Bonnot A, Whelan PJ, Mentis GZ, O’Donovan MJ (2002) Spatiotem-poral pattern of motoneuron activation in the rostral lumbar and thesacral segments during locomotor-like activity in the neonatalmouse spinal cord. J Neurosci 22:RC203.

Bou-Flores C, Lajard AM, Monteau R, De Maeyer E, Seif I, Lanoir J,Hilaire G (2000) Abnormal phrenic motoneuron activity and mor-phology in neonatal monoamine oxidase A-deficient transgenicmice: Possible role of a serotonin excess. J Neurosci 20:4646–4656.

Brockhaus J, Ballanyi K, Smith JC, Richter DW (1993) Microenviron-ment of respiratory neurons in the in vitro brainstem- spinal cord ofneonatal rats. J Physiol 462:421–445.

Burnet H, Bevengut M, Chakri F, Bou-Flores C, Coulon P, Gaytan S,Pasaro R, Hilaire G (2001) Altered respiratory activity and respira-tory regulations in adult monoamine oxidase A-deficient mice.J Neurosci 21:5212–5221.

Cazalets JR, Borde M, Clarac F (1995) Localization and organizationof the central pattern generator for hindlimb locomotion in newbornrat. J Neurosci 15:4943–4951.

Cazalets JR, Gardette M, Hilaire G (2000) Locomotor network matu-ration is transiently delayed in the MAOA-deficient mouse. J Neu-rophysiol 83:2468–2470.

Chub N, O’Donovan MJ (1998) Blockade and recovery of spontaneousrhythmic activity after application of neurotransmitter antagonists tospinal networks of the chick embryo. J Neurosci 18:294–306.

Cowley KC, Schmidt BJ (1997) Regional distribution of the locomotor

pattern-generating network in the neonatal rat spinal cord. J Neu-rophysiol 77:247–259.

Czeh G, Somjen GG (1990) Hypoxic failure of synaptic transmission inthe isolated spinal cord, and the effects of divalent cations. BrainRes 527:224–233.

Erickson JT, Conover JC, Borday V, Champagnat J, Barbacid M,Yancopoulos G, Katz DM (1996) Mice lacking brain-derived neuro-trophic factor exhibit visceral sensory neuron losses distinct frommice lacking NT4 and display a severe developmental deficit incontrol of breathing. J Neurosci 16:5361–5371.

Feller MB (1999) Spontaneous correlated activity in developing neuralcircuits. Neuron 22:653–656.

Galante M, Nistri A, Ballerini L (2000) Opposite changes in synapticactivity of organotypic rat spinal cord cultures after chronic block ofAMPA/kainate or glycine and GABAA receptors. J Physiol 523 Pt3:639–651.

Jacquin TD, Sadoc G, Borday V, Champagnat J (1999) Pontine andmedullary control of the respiratory activity in the trigeminal andfacial nerves of the newborn mouse: An in vitro study. Eur J Neu-rosci 11:213–222.

Jiang Z, Carlin KP, Brownstone RM (1999) An in vitro functionallymature mouse spinal cord preparation for the study of spinal motornetworks. Brain Res 816:493–499.

Kerkut GA, Bagust J (1995) The isolated mammalian spinal cord. ProgNeurobiol 46:1–48.

Khazipov R, Bregestovski P, Ben Ari Y (1993) Hippocampal inhibitoryinterneurons are functionally disconnected from excitatory inputs byanoxia. J Neurophysiol 70:2251–2259.

Kiehn O, Kjaerulff O (1998) Distribution of central pattern generatorsfor rhythmic motor outputs in the spinal cord of limbed vertebrates.Ann NY Acad Sci 860:110–129.

Kjaerulff O, Kiehn O (1996) Distribution of networks generating andcoordinating locomotor activity in the neonatal rat spinal cord invitro: A lesion study. J Neurosci 16:5777–5794.

Kremer E, Lev-Tov A (1997) Localization of the spinal network asso-ciated with generation of hindlimb locomotion in the neonatal ratand organization of its transverse coupling system. J Neurophysiol77:1155–1170.

Krnjevic K (1999) Early effects of hypoxia on brain cell function. CroatMed J 40:375–380.

Luhmann HJ, Kral T, Heinemann U (1993) Influence of hypoxia onexcitation and GABAergic inhibition in mature and developing ratneocortex. Exp Brain Res 97:209–224.

Melton JE, Chae LO, Neubauer JA, Edelman NH (1991) Extracellularpotassium homeostasis in the cat medulla during progressive brainhypoxia. J Appl Physiol 70:1477–1482.

Nishimaru H, Takizawa H, Kudo N (2000) 5-Hydroxytryptamine-in-duced locomotor rhythm in the neonatal mouse spinal cord in vitro.Neurosci Lett 280:187–190.

O’Brien RJ, Kamboj S, Ehlers MD, Rosen KR, Fischbach GD, Huga-nirRL (1998) Activity-dependent modulation of synaptic AMPA re-ceptor accumulation. Neuron 21:1067–1078.

Okada Y, Muckenhoff K, Holtermann G, Acker H, Scheid P (1993)Depth profiles of pH and PO2 in the isolated brain stem-spinal cordof the neonatal rat. Respir Physiol 93:315–326.

Richter DW, Ballanyi K (1996) Response of the medullary respiratorynetwork to hypoxia. In: Tissue Oxygen Deprivation (Haddad GG,Lister RE, eds), pp. 751–777. New York: Marcel Dekker Inc.

Richter DW, Camerer H, Sonnhof U (1978) Changes in extracellularpotassium during the spontaneous activity of medullary respiratoryneurones. Pflugers Arch 376:139–149.

Smith JC, Feldman JL (1987) In vitro brainstem-spinal cord prepara-tions for study of motor systems for mammalian respiration andlocomotion. J Neurosci Methods 21:321–333.

Torgerson CS, Gdovin MJ, Kogo N, Remmers JE (1997) Depth pro-files of pH and PO2 in the in vitro brainstem preparation of thetadpole Rana catesbeiana. Respir Physiol 108:205–213.


Turrigiano GG, Leslie KR, Desai NS, Rutherford LC, Nelson SB (1998)Activity-dependent scaling of quantal amplitude in neocortical neu-rons. Nature 391:892–896.

Vovenko E (1999) Distribution of oxygen tension on the surface ofarterioles, capillaries and venules of brain cortex and in tissue innormoxia: An experimental study on rats. Pflugers Arch 437:617–623.

Whelan PJ, Bonnot A, O’Donovan MJ (2000) Properties of rhythmic

activity generated by the isolated spinal cord of the neonatalmouse. J Neurophysiol 84:2821–2833.

Whelan PJ, Wilson RJ (2001) Depth profiles of PO2 within the isolatedin vitro mouse spinal cord during locomotor-like activity. Soc Neu-rosci Abstr 297.3.

Wilson RJ, Straus C, Remmers JE (1999) Efficacy of a low volumerecirculating superfusion chamber for long term administration ofexpensive drugs and dyes. J Neurosci Methods 87:175–184.

(Accepted 16 October 2002)


tissue po and the effects of hypoxia on the … · 2 at various tissue depths in the isolated mouse...

Documents